Recent advances in craniofacial morphogenesis


  • Yang Chai,

    Corresponding author
    1. Center for Craniofacial Molecular Biology School of Dentistry University of Southern California, Los Angeles, California
    • Center for Craniofacial Molecular Biology, University of Southern California, 2250 Alcazar Street, CSA 103, Los Angeles, CA 90033
    Search for more papers by this author
  • Robert E. Maxson Jr

    1. Department of Biochemistry and Molecular Biology, USC/Norris Comprehensive Cancer Center and Hospital, Keck School of Medicine, University of Southern California, Los Angeles, California
    Search for more papers by this author


Craniofacial malformations are involved in three fourths of all congenital birth defects in humans, affecting the development of head, face, or neck. Tremendous progress in the study of craniofacial development has been made that places this field at the forefront of biomedical research. A concerted effort among evolutionary and developmental biologists, human geneticists, and tissue engineers has revealed important information on the molecular mechanisms that are crucial for the patterning and formation of craniofacial structures. Here, we highlight recent advances in our understanding of evo–devo as it relates to craniofacial morphogenesis, fate determination of cranial neural crest cells, and specific signaling pathways in regulating tissue–tissue interactions during patterning of craniofacial apparatus and the morphogenesis of tooth, mandible, and palate. Together, these findings will be beneficial for the understanding, treatment, and prevention of human congenital malformations and establish the foundation for craniofacial tissue regeneration. Developmental Dynamics 235:2353–2375, 2006. © 2006 Wiley-Liss, Inc.


Development of the craniofacial region is a complex process with many features that reflect strong evolutionary forces controlling morphology. The vertebrate craniofacial region houses and protects the brain and provides the scaffold on which the sensory and feeding organs are located. The ability to sense and devour prey is fundamental to animal survival. Variations in craniofacial anatomy and function provide the major driving force in evolutionary adaptation.

One of the key features of craniofacial development is the formation of cranial neural crest (CNC) cells. The specification, emigration and migration, proliferation, survival, and ultimate fate determination of the CNC play an important role in regulating craniofacial development. Unlike the trunk neural crest, CNC cells give rise to an array of cell types during embryonic development. For example, CNC cells form most of the hard tissues of the head such as bone, cartilage, and teeth, whereas hard tissues in the rest of the body are formed from mesoderm cells. Genetic disorders, environmental insults, or the combination of both can alter the fate determination of CNC cells and result in craniofacial malformations. Significant progress has been made in recent years toward the understanding of how this important population of pluripotent cells is initially established in the early embryo and of the molecular mechanisms that mediate neural crest cell lineage segregation, differentiation, and final contribution to a particular tissue type (Shah et al.,1996; LaBonne and Bronner-Fraser,1999; Chai et al.,2003; Le Douarin et al.,2004).

The tissues of the head are composed of cells from all three germ layer origins: ectodermal, endodermal, and mesenchymal. As seen in the development of many organs, craniofacial morphogenesis depends upon continuous and reciprocal tissue–tissue interactions, with tooth, palate, and mandible development as classic examples. Research on the development of the head requires a thorough understanding of normal morphology, cell movement, cell signaling, gene/gene interaction, and transcriptional regulation in time and space.

Investigation of craniofacial development uses different animal species as models as with other areas of research in developmental biology. Studies in mice combine the power of genetics and genome manipulation together with in vitro organ culture techniques, leading to great progress in recent years. Avian embryos (chicken and quail) are easily accessible and are used for grafting/transplantation experiments. Zebrafish craniofacial developmental studies have emerged more recently, and mutant screens have led to the identification of new cell signaling interactions (Trainor and Krumlauf,2000; Yelick and Schilling,2002). Central to these models and techniques lies morphology. The head is a very complex structure, and a detailed analysis and appreciation of morphology are essential to understanding the mechanism of craniofacial malformations. Ultimately, “molecular morphology,” the combination of classic morphology and molecular biology, provides the basis for understanding the evolutionary changes in head structure and formation that in turn helps to clarify key developmental principals as well as the mechanism of craniofacial malformations.

Mouse models are extremely valuable in our effort to gain a better understanding of human craniofacial birth defects. The remarkable progress being made in the human genome, in parallel with exquisite functional genomic investigations in various mouse models, has resulted in the discovery of morphoregulatory genes that determine craniofacial morphogenesis. For example, we have uncovered genes that are critical for determining cranial–caudal axis, dorsal–ventral patterning, left–right symmetry and segmentation in early forming neurulation as well as branchial arches. More recently, many animal models with specific craniofacial malformations have facilitated human genetic linkage analysis, in which a genetic defect has been identified as the cause of congenital malformation(s) (Thyagarajan et al.,2003; Murray and Schutte,2004). Overall, genetically mutated mouse models highlight the enormous challenges of uncovering complex genetic mechanisms underlying craniofacial development and malformations.

Progress and understanding of the key control processes of head development have advanced to an extent where developmental biologists are interacting with tissue engineers to devise cell-based approaches to treat clinical malformations in humans. The prospect of harnessing developmental processes to repair or replace damaged or diseased craniofacial tissues and organs is an exciting new area that links basic animal research with human genetics and provides great promise in our effort to reduce the pain and suffering associated with craniofacial malformations.

In this review, we highlight some recent advances in our understanding of evo–devo (evolutionary–development) as it relates to craniofacial development, the fate determination of cranial neural crest cells, craniofacial patterning and organogenesis. We also review new discoveries on the development of craniofacial bones, signaling interactions, and specificity in craniofacial morphogenesis. Finally, we will discuss how research advancements will be beneficial for the understanding, treatment, and prevention of human congenital malformations and prospectus of tissue engineering.


Perhaps no other anatomical feature more closely epitomizes vertebrates than the head. Comprising paired sensory elements, a muscularized masticatory apparatus, and a cartilaginous or bony braincase, the head develops from complex and massive movements of mesenchymal cells derived from mesoderm and neural crest. These cells interact with each other and with a variety of craniofacial epithelia to produce the intricate structures of the face, skull, teeth, and jaw. How this trajectory of inductive interactions occurs in molecular terms is a key question in developmental biology; how it has been modified during evolution to produce the stunning variety of craniofacial structures in vertebrates is of major interest in evolutionary biology.

We begin our review with a brief consideration of some current issues concerning the evolution of the head (for comprehensive treatments, see Santagati and Rijli,2003; Kuratani,2004, 2005; Northcutt,2005; Morriss-Kay and Wilkie,2005; Depew et al.,2005). We focus on recent findings that pertain to two broad questions: First, how did the head evolve in the chordate ancestor of vertebrates? Second, how, over shorter evolutionary periods, did modifications of the head developmental program produce the evolutionary novelties that made possible the huge array of craniofacial morphologies evident in vertebrates?

Although the details of how the head first evolved are as elusive as fossils of the early chordates that once lived in pre-Cambrian seas, there is broad agreement with a scenario put forward by Northcutt and Gans (1983) and modified recently by Northcutt (2005). Their New Head hypothesis postulates that the ancestral vertebrate was an animal similar to the modern day cephalochordate Amphioxus. Like Amphioxus, this animal was a filter feeder that lacked pharyngeal arch muscles to move water through the gills. Also lacking was a braincase and the structures characteristic of the rostral head of vertebrates, including olfactory bulbs and telencephalic vesicles.

According to the New Head hypothesis, this stem organism underwent an evolutionary transition from filter feeding to active predation (Fig. 1). Underlying this transition were several key innovations, including the development of muscularized jaws and gill arches, a nerve plexus that enabled the animal to detect and capture prey, and cartilaginous and skeletal elements that provided a fixed spatial organization for this plexus. In principle, these new structures could be produced in two different ways: the anterior portion of the trunk could be restructured, or a section could be added onto the trunk de novo. The latter possibility—that the head is a neomorph—is a central proposal of the New Head hypothesis. Also key to this hypothesis is the idea that the cell type most critical for the evolution of the new head was the neural crest.

Figure 1.

The New Head and the evolution of the jaw. Schematics of the oral region of Amphioxus, a lamprey (jawless vertebrate), and a gnathostome (jawed vertebrate) are shown with their phylogenetic relationships (redrawn from Shigetani et al.,2002, and Holland et al., 2004). As proposed by Shigetani et al. (2002), the neural crest-derived mesenchyme of lampreys and gnathostomes is subdivided into postoptic (po) and mandibular arch (ma) subdomains. Growth factors are secreted by the epidermis (red line). In the lamprey, these induce homeobox genes (brown and blue) over the entire length of the neural crest-derived mesenchyme, whereas in the gnathostome, such signals are effective only in the mandibular component. Thus, it is proposed that a shift in epithelial–mesenchymal interactions results in the evolutionary emergence of the jaw. Ulp, upper lip; llp, lower lip; prc, prechordal cranium.

The neural crest is induced by an interaction between cells of the neural plate and adjacent surface ectoderm (reviewed by Meulemans and Bronner-Fraser,2004). Nascent neural crest cells delaminate from the developing neural tube, take on a mesenchymal character and migrate to distant sites in the developing embryo. In the craniofacial region, they give rise to a wide variety of cell types and tissues, including intramembranous bone, cartilage, muscle, and nerves (Creuzet et al.,2005; Noden and Trainor,2005). Northcutt and Gans (1983) proposed that the neural crest emerged in an early vertebrate ancestor and served as a key source of evolutionary novelty that made the New Head possible. They based this view on the apparent uniqueness of neural crest to vertebrates and on the ability of neural crest both to contribute to structures that form the new head as well as to serve as a source of patterning information (Noden,1978; Couly et al.,1993; Jiang et al.,2002; Schneider and Helms,2003).

That the fate of the neural crest includes elements of the rostral most portion of the new head and that the neural crest has a key role in patterning these tissues are clear (Couly et al.,1993; Jiang et al.,2002; Gross and Hanken,2005; Evans and Noden,2006). What is less clear is whether the neural crest is unique to vertebrates. Efforts to identify neural crest-like cell populations in extant nonvertebrate chordates have produced mixed results. Yu et al. (2002) found that amphioxus has cells that express an amphioxus homologue of the crest marker FoxD3. AmphiFoxD is expressed in the anterior neural plate but not in cells bordering the neural plate, as is the case for FoxD3 in vertebrates. AmphiFoxD is also expressed in axial and paraxial mesoderm. Yu and coworkers speculate that an amphiFoxD congener in a common ancestor of vertebrates and amphioxus had a role in the development of the mesoderm, and subsequently, during early vertebrate evolution took on a function in the early neural crest.

On the other hand, recent work from Jeffery's group has shown that a urochordate (ascidian) possesses cells that exhibit neural crest-like behavior (Jeffery et al.,2004). These cells migrate from the neural tube and express the neural crest markers hnk and Zic1. Ultimately they give rise to pigment cells, leading Jeffery and coworkers to propose that the neural crest first arose as pigment cell precursors in a common ancestor of vertebrates and ascidians, and later acquired additional fates.

Whether such neural crest-like cells exist in other nonvertebrate chordate groups remains to be determined. Results to date show that, although cells with neural crest-like properties are present in some nonvertebrates, cells possessing the full spectrum of neural crest behavior are unique to vertebrates. Thus, the idea that the neural crest is a major source of New Head structures remains intact (Northcutt,2005).

Evolutionary Novelty in the Craniofacial Complex

The first vertebrates exhibiting a “New Head” were probably similar to modern day jawless fishes (agnathans; Fig. 1). These organisms, which include lampreys and hagfish, are similar to higher vertebrates in their embryology, but lack jaws, possessing instead a filter-feeding apparatus (Kuratani et al.,2001). A major question is what were the sources of evolutionary novelty that made possible changes that led to more complex craniofacial morphologies of later vertebrate groups? One approach to this problem has been to carry out comparative embryological studies on the lamprey.

Kuratani and colleagues have examined marker gene expression in the Japanese lamprey, comparing expression patterns with those seen in gnathostomes (Shigetani et al.,2002). A generally accepted view of jaw evolution is that the mandibular arch, the most rostral of the pharyngeal arches, was modified to produce the mandible and maxilla. In ancient vertebrates, the arches were morphologically similar. The lamprey lacks a jaw but has upper and lower lips with distinct morphologies. These authors examined the expression of Dlx and Msx homeobox genes, which, in amniotes, are expressed in neural crest-derived mesenchymal cells along the future proximal–distal axis of the mandibular arch. Additional markers with region-specific expression included Fgf8 and Bmp4.

The expression patterns of these genes in lamprey embryos appeared at first to support the view that lamprey larval lips and gnathostome jaws are homologous. However, DiI labeling of neural crest populations suggested a more complex picture. Neural crest cells contributing to the upper and lower lips are derived from forebrain and midbrain crest populations, unlike the situation in gnathostomes in which mandibular arch crest derives from more posterior populations. The lamprey Dlx1 homologue is expressed in different regions of the neural crest compared with gnathostomes. This shift in expression is correlated with a shift in expression of Fgf8, an upstream regulator of Dlx1 in both lamprey and gnathostomes, in the epidermis. This shift is seen as the cause of a corresponding shift in the fate of subpopulations of neural crest.

An evolutionary change such as this one in a jawless gnathostome ancestor could have led to the emergence of the jaw. If this scenario is correct, then the gnathostome jaw is an evolutionary innovation resulting from a change in the topographic location of an epithelial–mesenchymal interaction. This change resulted in a shift in fate of neural crest cells and, thus, to the innovation that became the vertebrate jaw.

Recent work from the Tabin and Chuong labs has identified potential sources of evolutionary novelty in the epithelial–mesenchymal interactions that pattern the avian beak (Abzhanov et al.,2004; Wu et al.,2004). This work suggests that evolutionary changes in beak morphology in birds may be caused by similar shifts in epithelial–mesenchymal signaling. Darwin's finches, in the course of their radiation into various ecological niches in the Galapagos, evolved diverse beak morphologies. By examining the expression of various growth factors among Galapagos finches, Tabin's group found that the expression of Bmp4 in the mesenchyme of the upper beak was closely associated with a particular morphology, a broad shape (Abzhanov et al.,2004). By misexpressing bone morphogenetic protein-4 (Bmp4) in chick embryos, a similar broad shape could be produced.

The Chuong group (Wu et al.,2004) addressed a related question by examining the basis of morphological differences between the shapes of chicken and duck beaks. Ducks have two growth zones in the frontonasal mass, whereas chickens have only one. This difference accounts for the difference in beak morphology between these two birds. Bmp4 is expressed in the growth zones. By manipulating the location of Bmp4 expression, Chuong's group was able to modulate the shape of the beak. Together, these studies on beak development demonstrate that subtle changes in the topography of Bmp signaling within the developing beak can account for dramatic evolutionary shifts in beak morphology.

Schneider and Helms (2003) addressed the role of the neural crest in beak patterning and the evolution of beak shape. These authors exchanged neural crest cells fated to contributed to the beak between quail and chicks, which exhibit distinct beak morphologies. They sought to test the tripartite hypothesis that neural crest cells contain patterning information for beak morphology, that such cells possess an autonomous program by which they express this information, and that such cells influence the fates of non-neural crest cells.

Reciprocal transplantation experiments confirmed each of these points. Thus, for example, quail neural crest cells, when transplanted into duck embryos gave rise to beaks like those found in quail. That donor neural crest cells pattern host structures was shown by an examination of egg tooth morphology. Duck and quail egg teeth, which are derived from epidermis, not neural crest, exhibit distinct morphologies. The results showed that the morphology was characteristic of the transplanted tissue. Transplanted neural crest cells express molecular markers characteristic of the donor species, confirming the idea that such cells express an autonomous molecular program. It is not hard to envisage that evolutionary shifts in neural crest populations could produce profound morphological changes.

Tissue Boundaries in Cranial Evolution

The finding that the neural crest can provide patterning information suggests that evolutionary shifts in neural crest–non-neural crest boundaries could be an additional source of novelty (Fig. 2). Recent studies have addressed the significance of boundaries in the development and evolution of the skull vault (Jiang et al.,2002; Merrill et al.,2006; Evans and Noden,2006).

Figure 2.

Neural crest boundary relationships in skull vaults of mouse, chicken, and frog. Skulls are shown in the dorsal aspect. In the mouse, the coronal suture (CS) marks a neural crest–mesoderm boundary (blue line; neural crest anterior) (Jiang et al.,2002). Note also that the frontal (F) and interparietal (IP) bones are neural crest-derived (blue). In the chicken, the neural crest–mesoderm boundary lies within the frontal bone (F) and does not coincide with a suture (Noden,1975; Le Lievre,1978; Noden and Trainor,2005; Evans and Noden,2006). The solid blue frontoparietal (FP) bone of the frog, Xenopus, indicates that neural crest contributes over its entire length (Gross and Hanken,2005). The extent to which mesoderm contributes to this bone is not known. See Figure 3 for a detailed depiction of neural crest- and mesoderm-derived bones of the skull vault. P, parietal bone.

Figure 3.

Contribution of ectoderm, mesoderm, and endoderm during craniofacial development.A: Neural crest cells are formed at the junction of neural and surface ectoderm. These cells undergo epithelial–mesenchymal transformation, become ectomesenchyme, and travel into multiple destinations. B: Side view of an E9.5 mouse embryo shows unsegmented paraxial mesoderm in the head and mesoderm-derived somites in the trunk. OP, optic vesicle; OV, otic vesicle. C: Transverse section of the developing first branchial arch that is covered by surface ectoderm. The core of the first arch contains cranial neural crest (CNC) -derived (blue) and paraxial mesoderm-derived (pink) cells. The pharyngeal endoderm (yellow) lines the inner aspect of the branchial arch. D: Schematic drawing of an adult mouse skull shows both the CNC- and paraxial mesoderm-derived elements (modified from Noden and Trainor,2005). Mesoderm-derived cells are in pink, and CNC-derived cells are in blue.

As discussed above, the skull vault develops from populations of mesenchymal cells with distinct embryological origins, neural crest, and head mesoderm. The extent of the contribution of neural crest to calvarial bones of birds has been controversial. Several studies using quail–chick chimeras or retroviral infection to produce fate maps (Noden,1975; Le Lievre,1978; Evans and Noden,2006) suggest that the neural crest–non-neural crest boundary lies within the frontal bone (reviewed in Noden and Trainor,2005). The frontal bones develop after the fusion of two intramembranous centers (Jollie,1981; discussed in Noden and Trainor,2005). The more rostral of these is of neural crest origin, the caudal of mesoderm origin. Thus, the boundary lies at the interface of these two ossification centers. Later studies, however, have suggested that the entire chick cranial vault, including the frontal and parietal bones, is of neural crest origin (Couly et al.,1993).

Similarly, in Xenopus, dye-marking experiments have shown that crest contributes to the full length of the frontoparietal bone, the major bone of the skull vault in anuran amphibians (Gross and Hanken,2005). In mouse, in contrast, the frontal bone appears to be derived entirely from neural crest, whereas the parietal bone is derived from mesoderm (Jiang et al.,2002; Ishii et al.,2003). Thus, in the mouse, the interface between the frontal and parietal bones—the coronal suture—is a boundary between neural crest and mesoderm (Merrill et al.,2006).

If we make the assumption that the boundary in the chick is within the frontal bone (Noden,1975; Le Lievre,1978; Evans and Noden,2006), then, as pointed out by Noden and Trainor (2005), the situation may not be substantially different from the mouse. On the other hand, if, in the chicken, crest indeed contributes to the entire skull vault, as it apparently does in Xenopus, then there may have been an evolutionary shift in the neural crest–mesoderm boundary in the ancestor of birds and mammals (Gross and Hanken,2005).

It is also possible that, as pointed out by Gross and Hanken (2005), amphibians, birds, and mammals may each exhibit distinct contributions of neural crest to the skull vault. We note as well that there remains some uncertainty about the homology relationships between the frontal, parietal, and frontoparietal bones of these three taxa—an uncertainty that clearly complicates the interpretation of neural crest labeling studies.

In the mouse, the coronal suture is located at the neural crest–mesoderm interface and is a major growth center of the skull vault. Boundaries between dissimilar cell populations often serve as important signaling centers that function in growth control (reviewed in Dahmann and Basler,1999). The location of the tissue boundary at the coronal suture in mammals might represent an evolutionary innovation that makes possible, for example, new modes of mastication, locomotion, or cranial growth. Further analysis of contributions of neural crest to the skull vault in extant vertebrates will be required to illuminate this issue.


The vertebrate neural crest is a pluripotent cell population derived from the lateral ridges of the neural plate during early stages of embryogenesis (Fig. 3). The functions of neural crest cells include coordination of various visceral activities, such as in the peripheral nervous system; protection of the body from external conditions, such as by melanocytes; and participation in craniofacial skeletal development (Le Douarin et al.,2004).

The formation of the neural crest is a classic example of embryonic induction, in which tissue–tissue interactions and the concerted action of signaling pathways are critical for the induction of neural crest precursor cells. After induction, neural crest cells disperse from the dorsal surface of the neural tube, undergo epithelial–mesenchymal transformation, and migrate extensively through the embryo, giving rise to a wide variety of differentiated cell types. The migration, proliferation, and differentiation of neural crest cells along multiple distinctive pathways have been studied extensively in various animal models (Bronner-Fraser,1993; LaBonne and Bronner-Fraser,1999; Le Douarin et al.,2004; Basch et al.,2004). Recent studies suggest that fate determination of the neural crest is strongly influenced by environmental cues. CNC cells interact with and are consequently instructed by pharyngeal endoderm, ectoderm, and mesoderm before giving rise to various types of tissues in the craniofacial region.

During craniofacial development, neural crest cells migrate ventrolaterally as they populate the craniofacial region. The proliferative activity of these crest cells produces the frontonasal process and the discrete swellings that demarcate each branchial arch. As these ectodermally derived cells migrate, they contribute extensively to the formation of mesenchymal structures in the head and neck. Cell labeling studies have demonstrated that neural crest cells arising from rhombomeres 1–3 (r1–3) of anterior hindbrain migrate into the first branchial arch and, thereafter, reside within the maxillary and mandibular prominences (Osumi-Yamashita et al.,1990; Serbedzija et al.,1992; Bronner-Fraser,1993; Selleck et al.,1993; Lumsden and Krumlauf,1996). The migration of these rhombencephalic crest cells is regulated by growth factor signaling pathways and their downstream transcription factors before the CNC cells become committed to several different tissue types such as bone, cartilage, tooth, and cranial nerve ganglia (Noden,1983, 1991; Lumsden,1988; Graham and Lumsden,1993; Le Douarin et al.,1993; Echelard et al.,1994; Imai et al.,1996; Trainor and Krumlauf,2000). Recent studies provide strong supportive evidence that the CNC cells are developmentally “plastic,” i.e., their fate is not predetermined before they reach their final destination; rather, these progenitor cells must be instructed by signals from other tissues to generate skeletal elements of appropriate shape and size in the craniofacial region. Tissues that provide the instructive signaling for CNC fate specification include, but are not limited to, the pharyngeal endoderm, the branchial arch ectoderm, and the isthmic organizer at the midbrain–hindbrain boundary (Baker and Bronner-Fraser,2001; Trainor et al.,2002; Couly et al.,2002; Le Douarin et al.,2004).

The Pharyngeal Endoderm

Most of our knowledge about the critical function of pharyngeal endoderm in regulating the fate of CNC cells and the patterning of middle and lower face derives from studies using chick, quail, or zebrafish embryos as models. These embryos allow relatively easy manipulation compared with the mouse model. Surgical removal of pharyngeal endoderm during early stages of chick embryogenesis resulted in defects in facial bone and cartilage development (Couly et al.,2002). During the development of the first branchial arch, pharyngeal endoderm is thought to prepattern the orofacial epithelium, which in turn will provide instructive signals to pattern the CNC-derived mesenchyme (Haworth et al.,2004). In zebrafish studies, fibroblast growth factor (FGF) signaling has been shown to be critical for the development of pharyngeal endoderm itself and for the mediation of the endoderm to regulate facial skeletal morphogenesis (Ruhin et al.,2003; Crump et al.,2004; Helms et al.,2005). On the other hand, a recent study has shown that pharyngeal endoderm is not critical for the normal development of the upper and lower face. Instead, it is the ectoderm that is critical for providing the instructive information for facial morphogenesis (Aoki et al.,2002). Furthermore, CNC cells also contain intrinsic information that can affect facial development, although this contribution may depend upon the collective number of neural crest cells present at a given time and position (Schneider and Helms,2003; Tucker and Lumsden,2004). Overall, there is overwhelming evidence to support the notion that the pharyngeal endoderm has a critical role in regulating the fate of CNC. Discovery of the critical signaling pathway involved in this regulatory process awaits the development of mouse pharyngeal endoderm-specific gene inactivation models.

Orofacial Ectoderm

Early patterning of the oral ectoderm is independent of the neural crest (Veitch et al.,1999). In the mouse model, the branchial arch epithelium is correctly patterned, despite the CNC migration defect (Gavalas et al.,2001). After the interaction with pharyngeal endoderm, the orofacial ectoderm is roughly divided into proximal and distal domains. The establishment of the ectodermal domain greatly influences the fate determination of migrating CNC cells as they populate the branchial arch. At later stages of embryonic development, CNC cells interact with and provide instructive signals to the overlying epithelium and have an effect on epithelial patterning (Miletich and Sharpe,2004). Tooth development is a clear example of the consistent shift of instructive signal between orofacial ectoderm and the CNC-derived mesenchyme (see below for detailed discussion).

Concerted Action of Growth and Transcription Factors in Determining the Fate, Expansion, and Survival of CNC Cells

Recent studies have addressed the influence of growth factors on the fate of multipotent progenitor cells, such as the neural crest, during embryogenesis. It turns out that signaling centers, such as the isthmic organizer, rely on growth factor signaling in regulating the fate of CNC cells. Specifically, FGF8 signaling from the isthmic organizer can alter Hoxa2 expression and consequently control branchial arch patterning, demonstrating that neural crest cells are patterned by environmental signals (Trainor et al.,2002). Several growth factor signaling pathways have been shown to be critical for CNC fate determination, of which transforming growth factor-beta/BMP (TGF-β/BMP) signaling is a classic example. Members of the TGF-β superfamily of growth factors are expressed at sites where neural crest cells commit to form particular cell types. TGF-β superfamily members promote alternative fates for trunk neural crest cells; BMP signaling promotes neurogenesis by inducing MASH1 expression, whereas TGF-β signaling favors smooth muscle differentiation (Shah et al.,1996). In contrast, CNC cells react to TGF-β signaling differently. TGF-β controls the differentiation of CNC cells to glial cells, and TGF-β, BMP, and Wnt together control chondrocyte differentiation. Apparently, one of the major differences between CNC and TNC is the expression of Hox genes, which may account for the differential responsiveness between CNC and TNC to the same environmental cues (Abzhanov et al.,2003). Furthermore, the function of TGF-β in regulating neural crest cell differentiation is sensitive to the TGF-β expression level, such that TGF-β may promote alternative cell fates or induce apoptosis (Hagedorn et al.,2000). During early mouse craniofacial development, TGF-β subtypes are present in the CNC-derived mesenchyme during critical epithelial–mesenchymal interactions related to the formation of various organs (Lumsden,1984; Hall et al.,1992; Chai et al.,1994; Lumsden and Krumlauf,1996; Chai et al.,2003). Targeted null mutation of Tgfb2 or haploinsufficiency of Smad2 results in a wide range of developmental defects, including craniofacial malformations such as small mandible, dysmorphic calvaria, and cleft palate (Sanford et al.,1997; Nomura and Li,1998). Significantly, many affected tissues have neural crest-derived components and simulate neural crest deficiencies; thus, TGF-β signaling may provide significant instructive information to specify the fate of CNC during early craniofacial development.

Members of the TGF-β superfamily regulate the expression of transcription factors to influence cell fate decisions instructively during embryogenesis (Shah et al.,1996; Dorsky et al.,2000). For example, the expression patterns of Tgfb2 and transcription factor Msx1 have significant overlaps during early tooth and palate development when CNC-derived cells become specified to form dental and palatal mesenchyme, respectively, suggesting an epistatic relationship between these two genes (Ferguson,1994; Ito et al.,2003). In the palatal mesenchyme, overexpression of TGF-β suppresses transcriptional activity of the Msx1 gene (Nugent and Greene,1998). We have shown recently that compromised TGF-β signaling affects the expression of Msx1, which in turn controls the progression of the CNC cell cycle by regulating cyclin D1 expression (Ito et al.,2003). In parallel, in vitro studies also suggest that Msx1 gene expression maintains cyclin D1 expression and holds cells in an undifferentiated state by promoting proliferation (Hu et al.,2001). In general, TGF-β signaling specificity in regulating downstream target gene expression is determined by the interaction with other factors (such as BMP, Wnt, FGF, and so on) in a temporal- and spatial-specific manner (Massague,2000). The close intertwining of TGF-β signaling with other pathways appears to be a critical component of specific cell fate determination mediated by TGF-β family members.

In addition to regulating the fate of CNC cells, the concerted action of growth and transcription factors also controls cell proliferation and death. For example, BMP signaling controls Msx gene expression by directly regulating Msx2 promoter activity (Brugger et al.,2004). There is also an apparent feedback for BMP signaling by the Msx genes, because loss of Msx1 and Msx2 results in altered Bmp4 expression in the CNC cells (Ishii et al.,2005). Msx1 and Msx2 function redundantly in regulating the survival and proliferation of CNC cells during craniofacial development. This regulation is most likely carried out through the control of cell cycle progression (Hu et al.,2001; Han et al.,2003, Ishii et al., 2005). Recently, studies in Xenopus have shown that a Myc-mediated Id3 expression level is critical for determining whether the neural crest cells will proliferate or die, thus determining the size of the neural crest population (Kee and Bronner-Fraser,2005; Light et al.,2005). It is currently unknown whether Id3 plays a similar role during mouse development.

Finally, it is important to address the issue of prepatterning of premigratory neural crest cells. Noden's experiment suggested that chick CNC cells are predetermined according to their rostrocaudal origin in the neural tube (Noden,1983). However, recent work has demonstrated that adjacent tissues, such as the isthmus, provide instructive signals to regulate downstream target genes to determine the fate of CNC cells (Trainor et al.,2002). Furthermore, by providing critical feedback to the surface ectoderm, CNC can provide species-specific patterning information during craniofacial development, highlighting the importance of tissue–tissue interaction in regulating organogenesis (Schneider and Helms,2003; Helms et al.,2005). In the mouse model, it has been demonstrated elegantly that CNC cells retain a remarkable degree of plasticity, even after their migration into the branchial arch (Zhao et al.,2006). The second arch CNC cells have a cell-autonomous requirement for the Hoxa2 gene for their intrinsic patterning program (Santagati et al.,2005). In vitro studies have also begun to address the plasticity of postmigratory CNC cells (Zhao et al.,2006). Taken together, it is clear that the fate of mouse postmigratory CNC cells is determined through the reciprocal signaling between neural crest mesenchyme and the surrounding environment, where timing is an essential component.


The principal goal of developmental biology is to understand how tissues are induced and patterned to generate different organs at the correct location and time. Evidence suggests that signals from the anterior visceral endoderm, anterior neural ridge, and the mesoderm are required for the development of head structures (Spemann,1938; Shimamura and Rubenstein,1997; Beddington and Robertson,1999; Depew et al.,2002a). Furthermore, the formation of vertebrate “New Head” depends upon the presence of CNC cells and additional sensory placodes. Multiple molecules have been identified as critical regulators at these signaling centers. However, the challenge remains to determine how various signaling centers coordinate with each other and build complicated structures that make the head.

In the facial region, each branchial arch contains a central blood vessel, the aortic arch, which is surrounded by cells of paraxial mesoderm. The mesoderm core is enveloped by the more peripherally located CNC cells. The inner surface of the branchial arch contains cells derived from the pharyngeal endoderm, while the outer surface is covered by the ectoderm (Fig. 3). The spatial relationship of the boundary of cells with different embryonic origin is crucial for normal organogenesis and may reflect the regulatory interactions that control the development of complex structures in the craniofacial region.


Craniofacial ectoderm plays a critical role in regulating the fate of CNC cells during craniofacial morphogenesis, whereas the establishment of ectoderm identity is independent of CNC cells. Later, the continuous and reciprocal interaction between the ectoderm and the CNC-derived ectomesenchyme controls the position, size, and shape of craniofacial organs during embryogenesis. It is imperative to appreciate the inductive capability of ectoderm and CNC cells in the context of time. This concept is best illustrated by the seemingly contradictory conclusions regarding the ability of the oral ectoderm or the CNC-derived ectomesenchyme to induce tooth morphogenesis.

Tissue recombination experiments show that the tooth inductive signal first resides in the oral ectoderm and then shifts into the underlying CNC-derived ectomesenchyme at a later developmental stage. In mouse, dental epithelium before embryonic day (E) 12 is capable of inducing tooth formation when combined with nondental mesenchyme, whereas dental mesenchyme after E12 can induce tooth formation when combined with nondental epithelium (such as the second branchial arch epithelium; Mina and Kollar,1987).

Growth and transcription factors are responsible for establishing the patterning of craniofacial development. In the first branchial arch oral ectoderm, FGF8 and BMP4 are critical for setting up the proximal–distal axis during development. In the upper and middle face, where structures derive from the frontonasal process, Sonic hedgehog (Shh) and FGF signaling appear to be critical for setting up a boundary in the neural and surface ectoderm (Helms et al.,2005). In craniofacial patterning, definitive evidence only supports the critical function of ectodermal FGF signaling, as conditional inactivation of the Fgf8 signaling resulted in the disappearance of the proximal portion of the mandible (Trumpp et al.,1999). More recently, studies have shown that signaling through FGFR1 is critical for the neural crest independent patterning of the pharyngeal ectoderm. Ectoderm FGF signaling patterns the pharyngeal region to create a permissive environment for the entry of CNC cells (Trokovic et al.,2003,2005).

Craniofacial ectoderm regulates the expression of transcription factors to specify the fate of CNC cells. For example, FGF8 signaling controls the expression of two Lim-homeobox domain genes, Lhx6 and Lhx7, in the CNC-derived ectomesenchyme. Lhx6 and Lhx7 are mainly expressed in the oral side of ectomesenchyme, while Gsc is expressed in the aboral region (Tucker et al.,1999). Endothelin-1 in the mandibular epithelium controls the expression of Gsc. Mutations in either Endothelin-1 or Gsc result in mandibular development defects, demonstrating the endothelin-mediated Gsc expression is critical for controlling the patterning of mandible (Clouthier et al.,1998). In addition, BMP signaling has been shown to be a critical regulator for Msx1 function, which is exclusively expressed in the CNC-derived ectomesenchyme (Chen et al.,1996; Tucker et al.,1998a). Conditional inactivation of Bmpr1a in the oral ectoderm results in tooth agenesis (Andl et al.,2004). Inhibition of BMP signaling by noggin causes ectopic Barx-1 expression in the distal, presumptive incisor ectomesenchyme and a transformation of tooth identity from incisor to molar, thus demonstrating the significance of BMP signaling in patterning of craniofacial structures (Tucker et al.,1998b). Taken together, it is clear that ectoderm-mediated signaling plays an important role in determining the patterning of craniofacial structures. At this point, one cannot help but ask the question of what is responsible for setting up the signaling domain in the ectoderm. It turns out that pharyngeal endoderm is critical for this process.

Pharyngeal Endoderm

The pharyngeal endoderm makes a limited contribution to craniofacial development, but it serves as an indispensable inducer during tissue–tissue interactions, controlling craniofacial development (Fig. 3). The traditional view that the neural crest plays the key role in patterning the branchial arches must be reconsidered, because studies have found that the early ablation of CNC cells did not affect the development and function of endoderm cells. These crestless branchial arches were patterned normally and had a sense of individual identity, strongly suggesting that the patterning of endoderm is not dependent on CNC cells (Veitch et al.,1999). Hox genes are known to be critical for the identity of neural crest cells. Hoxa1 and Hoxb1 double mutant mice had patterning defects in the hindbrain; specifically, rhombomere 4 lost its ability to generate neural crest cells. Consequently, the second branchial arch crest population was not generated. Despite the neural crest defect, the formation of the second branchial arch endoderm and epithelial regionalization was normal (Gavalas et al.,2001).

Limited information is available regarding the patterning of pharyngeal endoderm. Retinoic acid signaling clearly plays a crucial role in regulating pharyngeal endoderm development as inactivation of retinaldehyde-specific dehydrogenase type 2 (Raldh2), the retinoic acid synthetic enzyme, resulted in the absence of all branchial arches caudal to the first (Niederreither et al.,1999). Inhibition of retinoid action by pan-retinoic acid receptor (pan-RAR) specifically perturbed the development of the third and fourth branchial arches, although the first and second branchial arches formed normally (Wendling et al.,2000). Recent studies have shown that Raldh2 is expressed in the lateral mesoderm flanking the endoderm during early embryonic development and may pattern the pharyngeal endoderm development (Niederreither et al.,2003; Graham et al.,2005). Tbx-1 is another important gene for pharyngeal endoderm development. Tbx-1 mutant mice show failure to form caudal pouches, whereas the first pharyngeal pouch develops normally (Jerome and Papaioannou,2001; Lindsay et al.,2001). Of interest, Tbx-1 is also expressed in the lateral mesoderm flanking the endoderm before the patterning of the pharyngeal endoderm (Brown et al.,2004).

The pharyngeal endoderm exerts its regulatory function through tissue–tissue interactions. For example, the pharyngeal pouches generate several specialized epithelial structures, such as the thyroid, parathyroid, and thymus (Graham et al.,2005). The development of these organs involves the interaction between pharyngeal endoderm and its flanking cranial neural crest cells. Compromised retinoic acid signaling from the pharyngeal mesoderm affects the development of pharyngeal endoderm, which in turn causes defects in CNC migration and development of pharyngeal pouch-derived organs, such as the thymus and parathyroid glands (Niederreither et al.,2003). Clearly, the pharyngeal endoderm plays an important role in regulating the morphogenesis of pharyngeal arch derivatives.


The cranial paraxial mesoderm has an organization different than the trunk paraxial mesoderm. The cranial paraxial mesoderm can be roughly divided into (1) preotic head mesoderm, which lacks any overt sign of segmentation and never forms somites; and (2) occipital somites, which are caudal to the otic vesicle and give rise to epaxial and hypaxial muscles of the neck, the pharyngeal and laryngeal muscles that develop in the caudal branchial arches, and the tongue muscle (Fig. 3B; Noden,1983; Couly et al.,1992; Huang et al.,1999; Mootoosamy and Dietrich,2002).

There are profound morphological and molecular differences between the cranial and trunk paraxial mesoderm. In comparison to the clearly segmented somite development within the trunk mesoderm, studies have suggested that head mesoderm has become arrested evolutionarily as somitomeres (Jacobson,1963; Packard and Meier,1983). Despite several gene expression analyses suggesting some degree of regionalization in the cranial paraxial mesoderm (such as Paraxis, Tbx1, and Hoxb-1), there is no evidence supporting the existence of metameric patterns. The importance of somitomeric head mesoderm is yet to be determined (Noden and Trainor,2005). At the molecular level, genes that drive mesoderm segmentation in the trunk are missing from the pre-otic mesoderm in the head. When cranial paraxial mesoderm was grafted to the trunk region, activation of the myogenic program in this ectopic position was inhibited, suggesting that there are distinct regulatory cascades acting in the development of trunk and head muscles, possibly reflecting their distinct function and evolution (Mootoosamy and Dietrich,2002). Of interest, however, heterotopic grafting of paraxial mesoderm cells to different regions of the cranial mesoderm in the mouse showed no restriction in cell potency in the craniocaudal axis, revealing considerable plasticity in the fate of the cranial mesoderm (Trainor et al.,1994). Signals from surface ectoderm and endoderm as well as CNC may influence the fate of mesoderm cells (Trokovic et al.,2003).

In terms of patterning capability, cranial paraxial mesoderm provides a permissive substratum for the migrating CNC cells to populate the branchial arch. Studies using chick embryos suggest that cranial paraxial mesoderm is able to direct CNC cell movement independently of their epithelial or mesenchymal organization (Noden,1986; Ferguson and Graham,2004). Of interest, recent cell fate mapping analysis has suggested that myoblast precursors contain positional identity inherited from their somatic mesenchymal stem cell precursors and can help to determine skeletal homologies that are based on muscle attachments (Matsuoka et al.,2005). Conversely, CNC cells, which provide most of the connective tissues and tendons in the head, may pattern and shape the individual cranial muscle (Noden,1986; Kontges and Lumsden,1996). The intimate relationship between cranial paraxial mesoderm and the CNC is clearly established from the moment that CNC cells enter the branchial arch and is maintained throughout craniofacial development. In parallel, studies have also demonstrated that cranial mesoderm is capable of providing crucial signals for endoderm development.

To date, there is not a clear demonstration of the patterning potential of cranial paraxial mesoderm to regulate craniofacial development in mammals. This finding is largely because of (1) lack of information on molecular signals involved in the interaction between cranial paraxial mesoderm and adjacent tissue and (2) our inability to generate tissue-specific gene inactivations. Recently, using the Cre/loxp recombination approach, we have learned that Myf5-Cre- or Mesp1-Cre-mediated recombination can be used to generate a targeted gene inactivation in the cranial paraxial mesoderm-derived first branchial arch myogenic core (our unpublished data). In Myf5-Cre;R26R mouse embryos at E9.5, the first branchial arch myoblasts are lacZ-positive, accurately reflecting their origin from the paraxial mesoderm (Fig. 4). Targeted gene inactivation in the cranial paraxial mesoderm-derived mesenchyme will reveal important information on its patterning potential in regulating craniofacial development.

Figure 4.

Two component genetic system for indelibly marking the fate of cranial neural crest- (CNC), or mesoderm-, or ectoderm-derived cells. A: Contribution of CNC cells during early craniofacial development. CNC-derived cells (blue) populate the frontonasal process (FN) and the first branchial arch (arrow). B: Cross-section of an embryonic day (E) 9.5 mouse embryo (Wnt1-Cre/R26R) shows CNC-derived cells (blue) in the first branchial arch. The arrow indicates the paraxial mesoderm-derived myogenic core in the first arch. The arrowhead indicates ectoderm. C: Contribution of paraxial mesoderm-derived cells during early craniofacial development. The arrow indicates the paraxial mesoderm-derived myogenic core (blue) in the first branchial arch of the Myf5-Cre/R26R sample. D: Cross-section of the first and second branchial arch contains paraxial mesoderm-derived myogenic cells (blue; arrow). Notice these cells are pink in B. E: Contribution of ectodermal cells during craniofacial development. Ectoderm-derived cells (blue) cover the surface of the K14-Cre/R26R embryo. The arrow indicates the first branchial arch surface ectoderm. F: Cross-section of the first branchial arch shows ectoderm-derived cells (arrow) on the surface.

Collectively, it is clear that craniofacial development requires continued interaction and contribution by cell populations derived from the ectoderm, the neural crest, the paraxial mesoderm, and the endoderm. These interactions occur en route and at the terminal site of tissue and organ genesis. There is no example of any craniofacial developmental event that can occur in a totally cell autonomous or completely dependent manner. Such a dichotomous view can only impede our progress toward a better understanding of the regulatory mechanism of craniofacial development. Instead, we need to focus our efforts to unveil the molecular signals involved in tissue–tissue interactions at each critical time point and to gain a better understanding of tissue boundary establishment, because a disturbed tissue boundary during early embryonic development can lead to craniofacial malformations.


Recent studies have uncovered specific signaling cascades that play crucial roles in regulating the patterning and morphogenesis of craniofacial organs. One of the best-studied models of craniofacial organogenesis is tooth development, which involves continuous tissue–tissue interactions between the ectoderm-derived enamel organ epithelium and the cranial neural crest-derived ectomesenchyme (Slavkin et al.,1968; Kollar,1972; Lumsden,1988; Jernvall and Thesleff,2000; Tucker and Sharpe,2004, and references therein). Multiple growth and transcription factors belonging to several signaling families have been identified as critical regulators at the initiation and throughout all stages of tooth development ( Furthermore, most of the signaling networks are used reiteratively throughout tooth development and are common to the regulatory systems critical for governing the development of other organs (such as the development of feather, hair, mammary gland, salivary gland, and pancreas; see reviews by Nuckolls et al.,1999; Shum et al.,2000). The growing scientific evidence suggests a highly conserved biological mechanism to regulate the patterning and morphogenesis of craniofacial organs, but studies are now beginning to reveal the unique features of the signaling network in regulating tooth morphogenesis. This topic will be the focus of our discussion here.

Patterning of the Mammalian Dentition

The patterning of dentition depends on the proper development of the oral cavity, where maxillary and mandibular teeth are housed and involves the determination of location, shape, number, and size of tooth development. Multiple developmental decisions are made in the patterning and development of mammalian dentition. The development of the oral cavity begins with the establishment of the frontonasal prominence and the first branchial arch. The migrating CNC cells are the major driving forces for branchial arch development. As the first branchial arch extends ventral–medially, it gives rise to both mandibular and maxillary prominences (although recent study suggests that the maxillary prominence has a separate origin from the mandibular prominence in chicken; Lee et al.,2004). The primitive oral cavity forms as a consequence of the fusion among intermaxillary segments of the frontonasal prominence and the paired maxillary and mandibular prominences. The oral epithelium, in contrast to the aboral (suboral) epithelium, becomes thickened to form the dental lamina, marking the time and location for tooth development to begin.

Before the initiation of tooth development, the mandibular ectoderm can be roughly divided into the proximal domain, which expresses FGF8 and gives rise to molars, and the distal domain, which expresses BMP4 and gives rise to incisors (Fig. 5). Significantly, FGF and BMP act antagonistically to restrict Barx1 and Dlx2 expression to the proximal domain of the first arch ectomesenchyme and Msx1 and Msx2 expression to the distal domain, respectively. The biological significance of such a regional molecular specification has been demonstrated elegantly with inhibition of BMP signaling resulting in ectopic Barx-1 expression in the distal, presumptive incisor mesenchyme and producing transformation of tooth form from incisor-form to molar-form (Tucker et al.,1998b). Recently, study has shown that Barx1 played a key role in the development of the dentition and digestive system, which was vital for the evolution of mammals (Miletich et al.,2005). Overall, FGF and BMP growth factor gradients are critical determinants for specifying the initiation sites of tooth formations as well as for specifying the CNC-derived dental mesenchyme to become incisor-form vs. molar-form tooth organs.

Figure 5.

Patterning of the first branchial arch.A: The frontal view of a scanning electronic microscopic image of an embryonic day (E) 9.5 mouse embryo (25–29 somites) shows the frontonasal process (FN), the maxillary (max), and the mandibular (mand) process. B: Schematic drawing of A. C: The first branchial arch can be divided into proximal/distal, oral/aboral, and buccal/lingual (see D) domains. In the distal domain, bone morphogenetic protein (BMP) signaling in the ectoderm regulates the expression of Msx1 and Msx2. It also provides inhibition for Barx1 expression in the cranial neural crest (CNC) -derived mesenchyme in the proximal domain. In the proximal domain, fibroblast growth factor (FGF) signaling regulates the expression of multiple transcription factors (modified from Tucker and Sharpe,2004). D: Molar teeth (m1, m2, m3) show different cusp patterns along buccal and lingual aspects of the jaw.

Besides the proximal and distal domains within the oral ectoderm, the mandibular arch mesenchyme can also be divided into an oral and aboral (rostral–caudal) axis (Fig. 5). Signals from the oral ectoderm appear to be responsible for coordinating this division. Of interest, the postmigratory CNC cells have an intimate association with the oral ectoderm (Chai et al.,2000). This CNC distribution pattern significantly overlaps with the expression of two specific Lim-homeobox domain genes, Lhx6 and Lhx7, within the CNC-derived ectomesenchyme, suggesting that they may have critical functions in directing CNC cells to reach their destination (Grigoriou et al.,1998). In the aboral region, the homeobox gene Goosecoid (Gsc) is expressed where Lhx6 and Lhx7 are excluded (Tucker et al.,1999). Apparently, FGF8 is responsible for regulating the expression of both Lhx and Gsc genes. The mandibular arch mesenchyme can be roughly divided into the Lhx-positive rostral (oral) domain and the Gsc-positive caudal (aboral) domain (Fig. 5).

Clearly, patterning of the mammalian dentition is a three-dimensional process. One aspect that is beginning to be explored is the determination of lingual–buccal (medial–lateral) dental cusp patterning (Fig. 5). Lingual and buccal cusps of the molar are different (in number and shape), and left and right first molars, for example, have a mirror image in patterning. It will be fascinating to learn how signals are set up to achieve this harmony during tooth development.

The development of each dental cusp is under the control of a transient signaling center known as the enamel knot, which is a dense population of epithelial cells without any proliferative activity (Fig. 6). The enamel knot contains multiple signaling molecules (such as Shh, BMP, and FGF) and controls the size and shape of cusp formation. Thereafter, it is removed by programmed cell death (Jernvall and Thesleff,2000; Tucker and Sharpe,2004). Unlike the incisor tooth organ, molar development relies on additional enamel knots (secondary and tertiary) to form multicusp patterning.

Figure 6.

Schematic drawing of reciprocal and reiterative signaling in regulating epithelial–mesenchymal interactions throughout sequential stages of tooth morphogenesis. A precise spatial and temporal orchestration of multiple growth and transcription factors (as listed within each box) is critical for the initiation as well as patterning of tooth germ development. These signaling molecules tightly regulate both the enamel organ epithelia and the cranial neural crest-derived dental mesenchyme. Specifically, when tooth development is initiated with formation of dental lamina, its underlying mesenchyme is almost entirely populated with cranial neural crest (CNC) -derived cells (dark blue). As tooth germ develops from the bud to the cap stage, CNC-derived cells are concentrated (dark blue) at the interface with the enamel organ epithelium, whereas the peripheral portion of the dental sac is populated with both CNC- and non–CNC-derived cells (light blue). Molecular signaling residing within the enamel organ epithelium and the CNC-derived dental mesenchyme are engaged in a constant and reciprocal dialogue to mediate tooth morphogenesis. The dark red dot represents enamel knot-a signaling center within the enamel organ epithelia, whereas light pink represents enamel organ epithelium (modified from Jernvall and Thesleff,2000; Chai et al.,2000).

BMP signaling is involved in regulating the distance between adjacent secondary enamel knots to control the positioning of cusps. Two recently published studies have provided important information regarding the contribution of BMP signaling inhibition to the regulation of cusp patterning. Ectodin is a secreted BMP inhibitor with an inverse expression pattern to p21, a hallmark cell cycle regulator expressed in the enamel knot during tooth development (Kassai et al.,2005). Loss of ectodin results in enlargement of the enamel knot, highly altered cusp patterns, fusion of molars, and extra teeth. Interestingly, the most dramatic changes in cusp patterning occurs along the buccal side of the cusps, suggesting that ectodin functions in the evolution of lateral bias in teeth. Noggin is another widely distributed BMP inhibitor (Chen et al.,2004). Overexpression of Noggin in the dental epithelium blocked the development of all mandibular and maxillary third molars at the bud stage and severely altered the patterning of maxillary molars (Plikus et al.,2005). The differential defects between mandibular and maxillary molars suggest a requirement for varying thresholds of BMP signaling.

Control of Signaling Level and Tooth Development

Recent studies show that local feedback provides a tightly controlled FGF and BMP gradient to ensure proper tooth development. For example, Islet1 positively regulates the expression of Bmp4, whereas a range of Pitx2 levels can differentially regulate the expression of Fgf8 and Bmp4 during initial tooth development (Lu et al.,1999; Mitsiadis et al.,2003; Liu et al.,2003). Pitx2 is restricted to the dental epithelium throughout tooth morphogenesis (Fig. 6). Mutation of the human PITX2 gene results in Rieger's syndrome, an autosomal dominant disorder, that leads to the absence of certain teeth and defects of the eye (Rieger,1935; Semina et al.,1996). Loss of Pitx2 results in retarded tooth development at the initiation/early bud stage, clearly demonstrating the important function of Pitx2 gene in regulating tooth development (Lu et al.,1999). Significantly, loss of one copy of the Pitx2 gene can also affect tooth development in mice (Gage et al.,1999). Therefore, Pitx2 acts in a dosage-specific manner in humans and mice. The expression of Pitx2 gene is sensitive to the level of BMP4 or FGF8 signaling, indicating a feedback loop among these signaling molecules (St. Amand et al.,2000).

Shh, another important epithelial signaling molecule, controls the proliferation of enamel organ epithelial cells during initiation of tooth development (Fig. 6; Bitgood and McMahon,1995; Dassule and McMahon,1998; Hardcastle et al.,1998; Cobourne et al.,2001). The Wnt signaling pathway interacts with Shh signaling to establish boundaries during tooth development. Specifically, Wnt-7b acts to repress the expression of Shh in nondental oral ectoderm, whereas Shh expression is restricted to the dental ectoderm and can instruct, permit, or induce tooth bud formations (Sarkar et al.,2000).

Determination of Tooth Number

Humans have a dental formula of in the permanent dentition [two incisors, one canine, two premolars (bicuspids) and three molars, with “/” marking the front midline]. Some rodents, such as mice, have a dental formula of The initiation of each tooth germ is marked by the formation of dental lamina. Ectodysplasin (EDA) signaling cascade molecules, which belong to the tumor necrosis factor family of ligands, are critical regulators for the determination of tooth number during embryonic development (Fig. 6). Mice with a compromised EDA signaling cascade, such as Tabby (EDA), Downless (EDAR, EDA receptor), and Crinkled (EDARADD, EDA intracellular adaptor protein) mutations, show abnormal tooth numbers and defects in the development of other ectodermal organs (hair follicles and exocrine glands; Sofaer,1969,1977; Headon et al.,2001; Pispa and Thesleff,2003; Tucker et al.,2004). On the other hand, overexpression of EDA signaling molecules results in an expansion of the molar tooth field and the formation of a supernumerary tooth distal to the first molar (Mustonen et al.,2003; Tucker et al.,2004). Of interest, the supernumerary tooth has the cusp patterning of a premolar, suggesting that dental cusp patterning is highly dependent on the exact position in the oral cavity where tooth development occurs.

BMP signaling is another morphoregulator for the number, size, and shape of tooth development. Loss of BMP receptor 1a (Bmpr1a) caused retarded tooth development (Andl et al.,2004). Overexpression of BMP inhibitor Noggin in the oral ectoderm resulted in miniaturization of maxillary first and second molars, and retarded maxillary third and mandibular molar development at the lamina stage (Plikus et al.,2005). The selective loss of molars in Noggin-overexpressing mice, along with the abnormal buccal cusp patterning in ectodin null mutant mice, strongly suggest that various tooth germs and dental cusps have differential requirements for the level of BMP signaling. Modulation of BMP signaling may account for one of the mechanisms responsible for changes in tooth number, size, and shape through evolution.

Continued Epithelial–Mesenchymal Interactions and the Success of Tooth Morphogenesis

After the initiation of tooth development, there is continued interaction between the dental epithelium and the CNC-derived ectomesenchyme throughout tooth morphogenesis. The initial tooth generating potential resides within the epithelium and is capable of inducing nontooth forming ectomesenchyme to develop teeth (Mina and Kollar,1987; Jernvall and Thesleff,2000). Soon after development, this tooth-forming potential shifts to the dental mesenchyme. This shift of tooth-forming potential coincides with a shift in BMP signaling from the epithelium to the mesenchyme. Morphologically, at this time, tooth development advances into the bud stage (Fig. 6). Multiple signaling molecules within the enamel organ epithelium have been implicated in the regulation of transcription factors in the surrounding ectomesenchyme during the bud to the cap stage transition (Fig. 6; Jernvall and Thesleff,2000; Tucker and Sharpe,2004). Interestingly, individual null mutations of Msx1, Lef1, or Pax9 all result in retarded tooth development at the bud stage, indicating the critical roles of these transcription factors during the transition from the bud to cap stage tooth development (Satokata and Maas,1994; Kratochwil et al.,1996; Peters et al.,1998; Sasaki et al.,2005). Ectopic overexpression and tissue recombination experiments have shown that these transcription factors are regulated by epithelial signals, can control the expression of growth factors and other transcription factors in the ectomesenchyme and provide instructive feedback to the enamel organ epithelium. Thus, they are essential determining factors during the cap stage tooth development.

To date, no null mutation experiments have resulted in failure of initiation of tooth development (thickening of the dental lamina). Some of the compound null mutations, however, have resulted in retardation of tooth development at an early bud stage, such as Msx1−/−/Msx2−/−, Dlx1−/−/Dlx2−/−, and Gli2−/−/Gli3−/−. Clearly, there is functional redundancy among different members of the same transcription factor family in regulating the advancement of tooth morphogenesis. Of interest, some of the null mutations reveal regional specificity of certain signaling molecules in regulating tooth morphogenesis. For example, only maxillary molar tooth organs were affected in Dlx1−/−/Dlx2−/− double mutants. Furthermore, loss of Dlx1 and Dlx2 genes resulted in a change of cell fate in the dental mesenchyme region from odontogenic to chondrogenic (Thomas et al.,1997; Tucker and Sharpe,1999). This finding also raises the possibility that a subpopulation of CNC cells carrying Dlx1/Dlx2 genes migrates into the maxillary molar region and, upon receiving the proper instruction from the dental epithelium, can contribute to the formation of tooth organ. By using the two-component genetic system (Wnt1-Cre/R26R) for indelibly marking the progenies of CNC cells (Fig. 4), it may be feasible to test the hypothesis that a subpopulation of CNC cells designated for the maxillary molar region is affected in Dlx1/Dlx2 double-mutant mice (Chai et al.,2000; Han et al.,2003). The outcome of such experiments may address the speculation that a predetermined subpopulation of CNC cells possesses instructive functions in the induction of tooth development.

The activin βA null mutation reveals another example of region-specific signaling in regulating tooth development. All teeth, except the maxillary molars, are arrested at the bud stage in activin βA−/− mice, a reverse phenotype compared with the Dlx1/Dlx2 double mutant (Ferguson et al.,1998). Because activin βA is expressed in the CNC-derived ectomesenchyme, it may help to specify the fate of CNC cells during tooth development.

Overall, continuous and reciprocal epithelial–mesenchymal interaction is the key for successful tooth organ morphogenesis. From initiation to odontoblast/ameloblast differentiation to matrix formation, this interaction provides precise communication between two adjacent tissue types and governs the number, size, and shape of tooth formation. Our understanding of these biological processes may serve as a foundation for the future design and fabrication of tooth regeneration.


The vertebral skull represents an excellent model for the investigation of development and evolution. In mammals, both CNC- and mesoderm-derived cells contribute to the development of the skull (Fig. 3D). Different elements of the skull are established by the formation of tissue boundaries, which reflect an evolutionary contribution and are critical for the pre- and postnatal dynamic development of the head and face. Recent studies are now beginning to address the molecular regulation of patterning and size determination of different elements of the skull. Multiple excellent reviews have addressed the regulation of skull vault development (Wilkie and Morriss-Kay,2001; Santagati and Rijli,2003; Morriss-Kay and Wilkie,2005), but there are very few reviews that address the molecular regulatory mechanism of facial bone development. Here, using mandible and palate development as examples, we summarize recent advancements toward the understanding of the regulatory mechanism of craniofacial bone development.

The skull consists of the neurocranium and viscerocranium. The neurocranium (skull vault and base) surrounds and protects the brain. In humans, eight bones compose the neurocranium: the paired temporal and parietal bones and the singular frontal, sphenoid, ethmoid, and occipital bones. Fourteen bones compose the viscerocranium (the jaws and other pharyngeal arch derivatives): the paired nasal bones, maxillae, palatine bones, lacrimal bones, zygoma and inferior nasal conchae, along with the singular vomer and mandible. The viscerocranium derives from the neural crest, forms the face, and supports the functions of feeding and breathing.

Until recently, the tissue origin of the skull vault has been controversial because of conflicting reports using quail–chick grafting. Noden (1978,1988) reported that the CNC only contributes to the rostral portion of the frontal bones and that the remainder of the skull vault is of mesoderm origin. In contrast, Couly and coworkers (1993) concluded that the skull vault is entirely neural crest derived. In the mouse model, it has been demonstrated elegantly that frontal bones are neural crest-derived and parietal bones are of mesoderm origin (Jiang et al.,2002). The posterior part of the skull vault (supraoccipital and exoccipital bones) derives from the occipital somites (Fig. 3D). The dura mater that underlies the frontal and parietal bones is also neural crest derived (Jiang et al.,2002; Ito et al.,2003; Sasaki et al.,2006). The coronal and sagittal sutures are of mesoderm origin (Morriss-Kay and Wilkie,2005). Clearly, the skull vault elements are developed at the boundary between CNC- and mesoderm-derived tissue. This boundary is of paramount importance in mediating tissue–tissue interaction that control skull vault development. When there is mixing of the two cell populations as the result of a gene mutation (such as in the ephrin-B1 or Twist mutant), the boundary between CNC and the mesoderm is lost, resulting in the premature fusion of cranial sutures known as craniosynostosis (Twigg et al.,2004; Merrill et al.,2006).


Patterning of the branchial arches requires the establishment of both interbranchial arch and intrabranchial arch identities (Depew et al.,2002a). It is well established that Hox, Pbx, and Otx homeobox genes are critical for the normal patterning of interbranchial arch identities (Gendron-Maguire et al.,1993; Rijli et al.,1993; Matsuo et al.,1995; Selleri et al.,2001). For example, there clearly is a Hox gene code that controls branchial arch development. To give rise to the derivatives of the first branchial arch, the structure needs to be Hox gene negative. Until recently, however, less information has been available about the genetic control of the establishment of the intrabranchial arch identities. This is an important area in craniofacial development, because human mandibular dysmorphogenesis appears to be a common malformation and appears in multiple congenital birth defect syndromes, ranging from agnathia (agenesis of the jaw) to micrognathia to patterning malformations.

Based on the mouse model, we know now that the first branchial arch (mandibular arch) becomes apparent at E8.0–E8.5 (6–8 somites) as small swellings on the side of the developing head. As CNC cells migrate into and proliferate within the mandibular arch, it develops rapidly toward the ventral midline. The CNC-derived cells are localized immediately subjacent to the covering epithelium (Chai et al.,2000). Mandibular development depends upon the interaction between the oral ectoderm and the CNC-derived mesenchyme within the first branchial arch. Within the oral ectoderm, signaling molecules, such as BMP, TGF-β, and FGF, are expressed in a region-specific manner (Chai et al.,1994,1997; Trumpp et al.,1999; Ito et al.,2002; Liu et al.,2005). They may regulate homeobox-containing genes (such as Dlx, Lhx, and Gsc) within the CNC-derived mesenchyme to generate early polarity and are responsible for patterning of the first branchial arch.

Ectodermal FGF signaling is critical for CNC cell survival in the mandibular arch as conditional inactivation of Fgf8 in the ectoderm caused increased apoptotic activity and a dramatic loss of all first branchial arch skeletal structures (except the most distal portion of the mandible; Trumpp et al.,1999). During early embryonic development, however, FGF signaling is not required for cell proliferation and survival, but instead is required for cell migration. Therefore, FGF signaling has differential functional specificity in different developmental contexts (Sun et al.,1999).

In the proximal domain of the first branchial arch, FGF signaling directly or indirectly regulates the expression of homeobox genes in the mesenchyme to control the development of mandible (see previous discussion). In the distal domain of the first branchial arch, BMP signaling appears to be a crucial regulator for mandibular morphogenesis. BMP4 is expressed within the distal ectoderm that covers the first branchial arch mesenchyme at E9.5. The postmigratory CNC cells are exposed to the instructive signaling (such as BMP) from the ectoderm and become committed to give rise to structures associated with the distal portion of the first arch. As embryonic development progresses, there is a shift of the instructive capability for patterning the mandible from the ectoderm to the CNC-derived mesenchyme. The patterning of the proximal–distal domain of the first branchial arch is achieved through the antagonistic interaction between BMP and FGF, which controls the mesenchymal expression of signaling molecules, such as Msx1 and Barx1, respectively (Tucker et al.,1998b; Tucker and Sharpe,2004). BMP signaling clearly has an important role in regulating mandibular morphogenesis. Specifically, loss of BMP signaling in the oral ectoderm and the pharyngeal endoderm results in extreme phenotypes, ranging from an almost completely missing mandible to severe defects in the distal region (Liu et al.,2005). BMP target genes Msx1 and Msx2 have differential dose requirements for BMP4 signaling, and they act in a functionally redundant manner. Both Msx1 and Msx2 are expressed in the midline region, and loss of BMP4 signaling results in compromised Msx gene expression. Mutation of Msx1 and Msx2 genes results in midline cleft of the first arch and severe defects in mandibular morphogenesis (Satokata et al.,2000; Ishii et al.,2005; our unpublished data).

Mandibular development requires proper modulation of BMP and FGF signaling in the ectoderm. This modulation is achieved through the antagonistic interaction between BMP and FGF signaling, the inhibition of BMP signaling by Noggin or Chordin and by the feedback from the underlying CNC-derived mesenchyme (Stottmann et al.,2001; Wilson and Tucker,2004; Liu et al.,2005). For example, BMP signaling is required for maintaining FGF signaling in the proximal domain of the mandibular arch, but it represses FGF signaling in the distal domain (Liu et al.,2005). Furthermore, BMP signaling is subject to multiple points of regulation, at the ligand, receptor, Smad, and transcription complex level (Massague et al.,2005). Clearly, tightly controlled BMP and FGF signaling is of paramount importance for mandibular morphogenesis.

Similar to the proximal–distal domains within the ectoderm of the first branchial arch, there are discrete populations within the underlying CNC-derived mesenchyme. Dlx genes are known to play important roles in regulating the patterning of the developing jaw. Six Dlx genes (Dlx1, Dlx2, Dlx3, Dlx5, Dlx6, and Dlx7) have been described in mice (Dolle et al.,1992; Robinson and Mahon,1994; Simeone et al.,1994; Qiu et al.,1995,1997; Stock et al.,1996). During the development of the branchial arches, Dlx genes show nested expression patterns and play important roles in establishing the identity along the proximodistal axis within each branchial arch (Depew et al.,2002a,2005). Specifically, Dlx1/2 are expressed in the CNC-derived ectomesenchyme both in the proximal and distal region of the branchial arch, whereas Dlx5/6 and Dlx3/7 show progressively restricted expression domains toward the distal region of the branchial arch.

Recent studies have shown that a combinatorial Dlx code regulates the establishment of the distinct skeletal elements within a given branchial arch unit (Qiu et al.,1995,1997; Depew et al.,2002a,b; Cobourne and Sharpe,2003). Loss of Dlx5 and Dlx6 results in a homeotic transformation of the lower jaw into an upper jaw, supporting a model of patterning within the branchial arch that relies on a nested pattern of Dlx gene expression (Depew et al.,2002b). In addition, jaw development is sensitive to the dosage of Dlx genes, as haploinsufficiency of single or multiple Dlx genes has a gradient effect on mandibular development (Depew et al.,2005). This observation clearly suggests that the expression of Dlx genes must be tightly regulated. To date, little information is available about the direct regulation on Dlx gene expression. Although FGF8-soaked beads placed in the first arch epithelium are able to induce Dlx2 and Dlx5 expression in the mandibular mesenchyme, loss of Fgf8 in the ectoderm shows unaltered Dlx2 and Dlx5 expression in the first branchial arch (Trumpp et al.,1999). Similarly, BMP-soaked beads are able to induce Dlx gene expression, but the endogenous BMP and Dlx expression patterns do not suggest a direct regulatory relationship. Overall, Dlx genes are clearly important for intrabranchial arch patterning, and further studies in this area will advance our understanding of their role and the regulatory mechanism in this process.


The mammalian palate develops from two primordia: the primary palate and the secondary palate. The primary palate represents only a small part of the adult hard palate. The secondary palate is the primordium of the hard and soft parts of the palate. Palate development is a multistep process that involves palatal shelf growth, elevation, midline fusion of palatal shelves, and the disappearance of midline epithelial seam (Fig. 7). The palatal structures are composed of the CNC-derived ectomesenchyme and pharyngeal ectoderm (Ferguson,1988; Shuler,1995). The epithelia that cover the palatal shelves are regionally divided into oral, nasal, and medial edge epithelium (Fig. 7D). The nasal and oral epithelia differentiate into pseudostratified and squamous epithelia, respectively, whereas the medial edge epithelium (MEE) is removed from the fusion line by means of programmed cell death and cell migration (Martinez-Alvarez et al.,2000; Vaziri Sani et al.,2005).

Figure 7.

Anatomy of palatogenesis and five categories of palatal shelf defects that result in cleft palate. A: Mouse palatogenesis starts at embryonic day (E) 11.5. By E13.5, palatal shelves (P) are on both sides of the tongue (T). Each palatal shelf can be divided into the medial (m) and lateral (l) aspects. t, tooth germ. B,C: Between E13.5 and E14.0, palatal shelves turn horizontally above the tongue and face each other along the midline. D: At E14.5, palatal fusion begins to take place. The arrow indicates the midline epithelial seam. n, nasal epithelium; o, oral epithelium. E,F: From E15.5 to E16.5, palatal fusion is completed throughout the entire palate. Five different categories of palatal shelf defects that result in cleft palate (also see Zhang et al.,2002). P, palatal shelf.

Mouse palatogenesis initiates at E11.5 as marked by the formation of palatal shelves extending from the internal aspects of maxillae. After initiation, the palatal shelves project downward on each side of the tongue between E12.5 and E13.5 (Fig. 7A,B). As the jaws develop, the tongue descends, thus providing space to accommodate the horizontal apposition of palatal shelves above the tongue (Fig. 7C). The fusion of palatal shelves occurs at E14.5, resulting in the formation of continuous palate (Fig. 7D). Both the elevation and fusion of the palatal shelves occur in an anterior to posterior sequence. The complete fusion of palatal shelves results in the separation of the oral cavity from the nasal cavity (Fig. 7E,F).

Fate of Medial Edge Epithelial Cells and the Contribution of Cranial Neural Crest Cells During Palatogenesis

There has been a tremendous interest in the fate of medial edge epithelial cells during palatal fusion. Apoptosis is clearly one of the important mechanisms for eliminating MEE cells during palatal fusion (Farbman,1968; Saunders,1966; Dudas and Kaartinen,2005). At the molecular level, TGF-β and RA are critical inducers for apoptosis of the MEE. The dosage of these signaling molecules may differentially control the progression of the cell cycle in the MEE (Cuervo et al.,2002; Martinez-Alvarez et al.,2000; Cuervo and Covarrubias,2004). An alternative fate for the MEE is migration along the midline toward the nasal and oral epithelia, resulting in the loss of MEE (Carette and Ferguson,1992, Hilliard et al., 2005). Studies from multiple laboratories have suggested that epithelial–mesenchymal transformation (EMT) is an important cellular mechanism for the disappearance of MEE (Fitchett and Hay,1989; Griffith and Hay,1992; Shuler et al.,1992). These studies were largely based on cell lineage tracing using membrane-intercalating dye and epithelial and mesenchymal cellular markers. Recently, however, studies using genetic cellular markers have challenged the theory of EMT during palatal fusion. Both in vivo and in vitro studies show that EMT does not occur during palatal fusion (Cuervo et al.,2002; Vaziri Sani et al.,2005; Xu et al.,2006). Apoptosis and cell migration may be sufficient to account for the cellular mechanism for the disappearance of MEE cells.

CNC cells are critical for palatogenesis. Until recently, however, little has been known about the fate of the CNC-derived palatal mesenchyme or the molecular mechanisms that regulate the epithelial–mesenchymal interactions during palate development. The lack of information is largely due to the difficulty in CNC cell labeling and fate analysis in the palate. Significantly, using the Wnt1-Cre/R26R model, we have systematically followed the migration, proliferation, and differentiation of CNC cells throughout embryogenesis. We show that the CNC-derived ectomesenchyme contribute significantly to the palatal mesenchyme and that there is a dynamic distribution of these CNC cells during palatogenesis (Chai et al.,2000; Ito et al.,2003). This two-component genetic system provides the opportunity to integrate analysis of the fate and function of the mammalian neural crest with mouse molecular genetics in both normal and abnormal embryonic development.

Mouse Models for Investigating the Molecular Regulatory Mechanism of Palatogenesis

Multiple genetically mutated mouse models have made significant contributions to our understanding of the gene pathways involved in palate development and the nature of signaling molecules that act in a tissue-specific manner at critical stages of embryonic development. Interestingly, however, most of genetic mutations cause multiple structural and functional defects in the developing embryo. Consequently, assessment of the role of a particular gene in regulating palatogenesis has been a challenge. To further complicate the issue, cleft lip and/or palate is a complex trait caused by multiple genetic and environmental factors (Murray,2002). Nevertheless, these animal models have advanced our understanding of some important gene functions in human palate development.

One of the important discoveries has been the existence of genetic heterogeneity along the anterior–posterior and medial–lateral axes of the developing palate (for review, see Hilliard et al.,2005). This heterogeneity may provide differential regulatory mechanisms for the fusion of the anterior vs. posterior region of the palate. For example, MEE cells begin to undergo apoptosis at different times during palatal fusion, depending on their location. It has been shown that apoptosis of MEE cells is triggered by palatal shelf contact in the anterior region, whereas it is initiated before any contact between the opposing shelves in the posterior region (Cuervo et al.,2002). This may be the result of differential molecular signals in the palatal mesenchyme along the anteroposterior (A-P) axis that instruct different fates to the palatal epithelium (Ferguson and Honig,1984). More recent studies have demonstrated that constant and reciprocal interactions between the palatal epithelium and the CNC-derived mesenchyme are responsible for setting up this genetic heterogeneity along the A-P axis and are crucial for normal palatal development and fusion (Zhang et al.,2002; Murray and Schutte,2004; Rice et al.,2004).

Multiple genes have been found to be critical for the development of the anterior region of the palate. For example, Msx1, Bmp4, Bmp2, Fgf10, and Shox2 show restricted expression patterns in the anterior region of the palate (Rice et al.,2004; Hilliard et al.,2005). Loss of Msx1 results in a cell proliferation defect in the CNC-derived palatal mesenchyme in the anterior region of secondary palate. BMP4 functions downstream of Msx1 and controls the expression of Shh in the palatal epithelium. Shh in turn regulates the expression of Bmp2 in the mesenchyme to promote cell proliferation (Zhang et al.,2002). Meanwhile, FGF10 is expressed in the anterior palatal mesenchyme and functions in a paracrine manner through its receptor FGFR2 in the palatal epithelium to mediate Shh expression, which in turn regulates Bmp2 expression in the mesenchyme to promote cell proliferation. So the BMP and FGF signaling pathways converge on Shh signaling in the epithelium to control the growth of the anterior region of the palatal shelf. The Shox2 gene is exclusively expressed in the anterior palatal mesenchyme. Loss of Shox2 results in an incomplete cleft of the anterior hard palate, whereas fusion of the posterior palate is normal (Yu et al.,2005). Significantly, this study clearly demonstrates that there are different regulatory mechanisms that control the development and fusion of the anterior and posterior parts of the palate and that a successful fusion of the posterior part of the palate can occur, despite that there is a failure of anterior palate fusion.

We know less about the specific gene expression patterns in the posterior region of the palate. Fgfr2 is expressed in the epithelium and the CNC-derived mesenchyme in the middle and posterior palate. FGF8 signaling selectively induces the expression of Pax9 in the posterior region of palatal mesenchyme. Loss of Pax9 results in a palatal shelf development defect and cleft palate (Peters et al.,1998; Hilliard et al.,2005). To date, there is little known about the regulation of Pax9 expression in the developing palate. Therefore, the biological significance of FGF8-mediated Pax9 expression during palatogenesis remains to be determined. In addition to the differential gene expression patterns along the A-P axis of the developing palate, there is also mesenchymal heterogeneity between the medial and lateral regions of the palatal shelf (Fig. 7A). For example, the odd skipped-related genesOsr1 and Osr2 are expressed in a medial–lateral gradient in the palatal shelf. Significantly, mutation of the Osr2 gene results in the compromised development of the medial aspect of palatal shelf development and retards palatal shelf elevation (Lan et al.,2004). The expression of Fgfr2 is focused on the medial aspect of the developing palatal shelf, suggesting a possible functional significance in regulating the development and elevation of palatal shelf.

In analyzing different mutant animal models with cleft palate, we propose to divide the palatal shelf development defects into the following five categories. (1) Failure of palatal shelf formation (Fig. 7). Although this is a severe type of palatal shelf development defect, it has a rare occurrence. Mutation of activin-βA causes a severe facial primordia development defect, which is likely responsible for the retardation of palatal shelf development and complete cleft palate (Matzuk et al.,1995). The Fgfr2 mutation also affects the initial development of the palatal shelf and results in complete cleft palate (Rice et al.,2004). (2) Fusion of the palatal shelf with the tongue or mandible (Fig. 7). For example, loss of function mutation of Fgf10 results in anterior palatal shelf fusion with the tongue, whereas the middle and posterior part of the palatal shelf adheres to the mandible, thus preventing the elevation of the palatal shelf (Alappat et al.,2005). In humans, mutations in TBX22 have been reported in families with X-linked cleft palate and ankyloglossia (Braybrook et al.,2001). Similarly, Tbx22 is expressed in the developing palate and tongue in mice, suggesting an important role of Tbx22 in regulating palate and tongue development (Bush et al.,2002). (3) Failure of palatal shelf elevation (Fig. 7). Studies have shown that mutations of Pax9, Pitx1, or Osr2 can lead to failed palatal shelf elevation and cleft palate defect (Peters et al.,1998; Szeto et al.,1999; Lan et al.,2004). The cellular defect is mainly associated with the CNC-derived palatal mesenchyme, suggesting important functions of these transcription factors in regulating the fate of CNC cells during palatogenesis. (4) Failure of palatal shelves to meet after elevation (Fig. 7). By far, this is the most common type of cleft palate defect documented in animal studies. For example, mutations in Msx1 and Lhx8 and conditional inactivation of Tgfbr2 in CNC cells or Shh in the epithelium all result in retarded palatal shelf development (Satokata and Maas,1994; Zhao et al.,1999; Ito et al.,2003; Rice et al.,2004). (5) Persistence of medial edge epithelium (Fig. 7). In Tgfb3 or Egfr mutant mice, there is an alteration of the fate of MEE cells (Kaartinen et al.,1995; Proetzel et al.,1995; Miettinen et al.,1999). In Tgfb3 null mutant mice, MEE cells fail to undergo apoptosis and persist along the midline to prevent normal fusion. In addition to its function in regulating the fate of MEE, TGF-β3 is also critical for proper proliferation of the CNC-derived palatal mesenchyme (our unpublished data). Furthermore, mutations in TGF-β3 have been associated with cleft palate in humans, underscoring the crucial function of TGF-β signaling in regulating palatogenesis (Lidral et al.,1998).


Each year, approximately 250,000 infants born in the United States have some mental or physical defects. Three fourths of all malformations seen at birth involve craniofacial dysmorphogenesis, affecting the development of head, face, or neck. These malformations are particularly devastating, as our faces are our identity—they are how we see ourselves and how others see us. In recent years, research has progressed so that we know the precise genetic error that leads to many craniofacial birth defects.

At the conclusion of a recent Gordon Research Conference on Craniofacial Morphogenesis and Tissue Regeneration (Ventura, CA), it was clear that there has been tremendous interest and development in recent years toward a better understanding of the molecular regulatory mechanism of craniofacial development. Developmental and evolutionary biologists as well as tissue engineers are working together to investigate and compare the tissue origin, patterning, and growth of various craniofacial organs in an effort to reproduce and/or repair defective tissue in the craniofacial region.

To date, only approximately one third of mouse mutations associated with craniofacial malformations have been linked to mutations of the orthologous human genes with similar dysmorphogenesis (Wilkie and Morriss-Kay,2001). It is not known what defects might arise from mutations of the other two thirds as some of these mutations in mice result in early embryonic lethality. This lethality demonstrates the important function of these genes in early embryonic development but makes it impossible to investigate the functional significance of that particular gene in regulating craniofacial development. It is conceivable that the equivalent mutations in human will also result in a lethal phenotype. Nevertheless, an important conclusion from studies using various animal models suggests that there are no “craniofacial genes.” The same morphogen (BMP, TGF-β, FGF, Shh, and so on) that controls limb development, for example, also plays a crucial role in regulating craniofacial morphogenesis. Clearly, the functional specificity and the developmental outcome of a signaling molecule is determined by the cell upon which the morphogen acts, as different cell types are exposed to a different combination of growth and transcription factors and may respond differently to the same growth factor signaling (“Ask not what TGF-β can do for the cell, ask what the cell can do with TGF-β,” Joan Massague, personal communication). Another important lesson from animal studies is the available dosage of an important regulatory gene, as haploinsufficiency or gain of function of a critical gene is often associated with congenital malformations in humans. Overall, studies using mutant mice have significantly improved our understanding of human embryogenesis in general and craniofacial development in particular. It is of paramount importance that there are constant interactions among researchers involved in both human and animal studies as this will help to ensure the rapid progress toward a better understanding, treatment, and prevention of human congenital malformations.

Tissue regeneration is another area with many exciting new developments based on research advancements in developmental and evolutionary biology (Chai and Slavkin,2003). The recent convergence of the human genome project and scientific advances toward understanding the molecular regulation of craniofacial morphogenesis, stem cell biology, and biotechnology offer unprecedented opportunities to realize craniofacial tissue regeneration. One of the next critical steps will be to apply our knowledge of molecular regulation of tooth morphogenesis, for example, to manipulate adult stem cells toward an odontogenic phenotype.

Significant progress has been made in stem cell biological research, which has advanced our understanding in the area of hematopoiesis, tissue engineering, and biomaterials (Lovell-Badge,2001). Recently, studies have shown that adult stem cells have a much higher degree of developmental potential than previously thought, and this has prompted considerations to explore the potential of stem cell mediated muscle, bone, cartilage, and dentin regeneration (Gronthos et al.,2000; Bianco et al.,2001; Bianco and Robey,2001). Stem cells are truly remarkable. They have the potential to grow into an array of specialized cells and hold great promise for treating medical and dental conditions. Systemically injected mouse bone marrow derived cells have given rise to muscle, cartilage, bone, liver, heart, brain, lung, alveolar epithelium, intestine, and, of course, hematocytes (Ferrari et al.,1998; Lagasse et al.,2000; Woodbury et al.,2000; Kotton et al.,2001). Although most of these animal studies serve now as precursors for future human clinical trials in treating certain medical and dental conditions, the basic scientific principles learned from these current analyses certainly have advanced our understanding of the biological regulation of tissue engineering. The 21st century provides us with tremendous opportunities to enable biological solutions to biological problems.


We thank Pablo Bringas, Jr., Jun Han, Ryoichi Hosokawa, Julie Mayo, Kyoko Oka, and Xun Xu for their assistance with this manuscript. We apologize to those colleagues whose publications were not cited due to space limitations. Both Dr. Chai's lab and Dr. Maxson's lab were supported by the NIH.