Cranial Neural Crest and Development of the Head Skeleton

Introduction

The skull and jaws were key innovations in vertebrate evolution, vital for a predatory lifestyle.¹ Much of the skull and all of the pharyngeal skeleton, including jaws, hyoid and gill structures, also have a unique embryonic origin from CNC, unlike the more posterior axial and appendicular skeletons which are derived from mesoderm. These CNC-derived cartilages and bones are modified in different vertebrate lineages, such that in mammals the branchial elements form laryngeal bones in the throat, in contrast to the primitive function in fish where they support the gill arches.² Similarly, the mammalian homologue of the hyoid bone that primitively supports the jaw in fish, instead forms the stapes of the mammalian middle ear. These remarkable changes in function and shape of vertebrate cranial skeletal elements reveal how subtle differences in patterning of CNC in the embryo can result in significant morphological differences between species.

Such changes are best understood if the skeletal elements are considered as units within larger head segments that are established in the embryo.³ The pharyngeal skeleton forms within a reiterated series of arches that surround the anterior foregut. Several factors influence the fate of CNC within the arches. These include the site of origin of the CNC in the neural tube, position within the arch following migration and proximity to local signaling centres including the surface ectoderm and the pharyngeal endoderm. Interpretation of these signals provides the spatial information required for the development of the final skeletal pattern.

Specification and Migration of the Skeletogenic CNC

A CNC origin for the skull was first suggested near the end of the 19th century by ablation experiments in amphibians.^4-6 These results were treated with skepticism at the time, since they contradicted the germ layer theory in which skeletal tissues were thought to be exclusively of mesodermal origin. Lately, new debates on the origin and significance of this “ectomesenchymal” population have emerged, suggesting that cranial skeletal elements arise from a migratory cell population distinct from the CNC.⁷ Fate mapping studies in chick first showed that CNC contributes directly to the “neurocranium”, which surrounds the brain, as well as to all of the pharyngeal arches of the “viscerocranium”,^8-10 and these results have been confirmed in other species.^11-13 CNC cells delaminate from the ectoderm overlying the dorsal neural tube and migrate as separate streams into the pharyngeal arches (fig. 1). Neurocranial precursors emerge from midbrain levels and migrate between the eyes to form the palatal shelves, while viscerocranial precursors emigrate in three distinct streams from the hindbrain into the mandibular (stream 1), hyoid (stream 2), and five branchial (stream 3) arches. Recent data also reveal a contribution of the most anterior (mandibular) stream to the ventral neurocranium^14,15 suggesting that the mandibular stream has a more significant contribution to the skull than has been appreciated. As discussed below, this has implications relating to signals received early in development during CNC migration.

In all vertebrates, segmental differences between streams of CNC are conferred through the nested expression of homeobox (Hox and Otx) genes (fig. 1). This is thought to coordinate A-P patterning of the hindbrain with the arches that it innervates (reviewed in ref. 16). For example, CNC cells migrating in stream 2 (hyoid) are the most anterior neural crest cells to express Hox genes. They express only Hox group 2 genes, and arise from the Hox-2 expressing region of the hindbrain. In contrast, CNC cells in stream 1 (mandibular) that will give rise to the jaws do not express Hox genes and arise from Hox-negative regions of the anterior hindbrain and midbrain. Loss of Hox group 2 gene function results in a homeotic transformation of hyoid skeletal elements derived from CNC of stream 2, to a Hox-negative, mandibular fate.^17,18 Grafts of CNC along the neural tube from Hox-negative regions to more posterior Hox-positive levels in the chick results in ectopic mandibular structures at inappropriate A-P positions.¹⁹ In contrast, overexpression of Hox genes in CNC of stream 1 results in ectopic hyoid skeletal elements in place of the jaw.^20,21 Taken together, these results suggest that the Hox expression status of CNC confers a subsequent positional identity, which is dictated by the site of origin of the CNC in the hindbrain.

Recent evidence, however, suggests that this identity is not completely fixed prior to migration into the periphery. The maintenance of Hox expression in migrating CNC depends on the Hox expression of immediately neighbouring cells and other signaling centers within the arches.^22,23 This implies that the positional origin of CNC in the neural tube is important for subsequent CNC fates in the pharyngeal arches, but that this is not sufficient to confer identity along the A-P axis. Indeed, the molecular regulation of Hox group 2 gene expression in migrating CNC is independent to that in the hindbrain, indicating that CNC do not simply inherit their A-P identities and patterns of Hox expression based on their hindbrain origins.²⁶

Recent genetic studies in zebrafish have provided several insights into the mechanisms of Hox gene regulation in CNC. For example, mutations in the histone acetyltransferase moz (monocytic leukemia zinc finger) disrupt both hoxa2 and hoxb2 expression in the hyoid CNC, but not in the hindbrain.²⁴ This results in partial transformations of hyoid arch cartilages to a mandibular fate, resembling a loss of hox2 function (fig. 2A-C).²¹ The first signs of these tranformations are changes in early gene expression in CNC condensations within each arch, prior to overt skeletal differentiation (fig. 3). Two examples of this are goosecoid (gsc), expressed in medial first arch and lateral second arch CNC condensations of cartilage precursors and bapx1, an orthologue of Drosophila bagpipe, expressed only in first arch joint precursors (asterisks in fig. 3). moz mutants develop mirror-image patterns of gsc expression in first and second arch medial cartilage precursors, and bapx1 expression in both the first and second arch joints. Such mirror-image transformations are also seen in the skeleton in Hoxa2^-/- mice and hox2-deficient zebrafish, pointing to the existence of a polarizing influence from cells at the junction between first and second arches (fig. 2C).

The zebrafish lockjaw (low) mutant disrupts the transcription factor AP-2 alpha (tfap2a), which also regulates Hox gene expression. Like moz, homozygous low mutants exhibit segment-specific changes in CNC fate in the arches, but not in the hindbrain.²⁵ low mutants develop partial hyoid-to-mandibular transformations, particularly in the dorsal region of the arch, which closely correlate with reductions in hoxa2 expression (fig. 2D).²⁷ hoxa2 gene expression is lost in CNC of the hyoid arch in mutants, but expression in the hindbrain remains unaffected. This CNC-specific requirement for tfap2a in hoxa2 regulation is conserved with mammals; the mouse Hoxa2 promoter contains an AP-2 binding site essential for expression in the CNC but not in the hindbrain.²⁶ Disruption of specification during CNC development correlates with a lack of neural crest-derived pigment cells, enteric neurons, and craniofacial cartilage in low (tfap2a) mutants. These correlate with early defects in the specification of the premigratory neural crest (expression of foxd3 and sox9a is reduced in tfap2a mutants) and cell survival in low mutants. These studies in low and moz support the paradigm of independent regulation of hox2 genes in CNC and the hindbrain and reveal that the final fate of the CNC is not dictated simply by its hindbrain level of origin, but requires signals from adjacent tissues.

Patterning Influences of Skeletogenic CNC on Surrounding Tissues

The most anterior CNC cells that form the upper and lower jaws as well as the anterior neurocranium do not express Hox genes (fig. 1). However this CNC, perhaps more than any other, possesses an identity independent of adjacent tissues once it has migrated into the head periphery. This was dramatically demonstrated by Noden¹⁹ in a classical series of transplantation experiments in avian embryos, in which midbrain CNC (Hox negative) was grafted to more posterior hindbrain (Hox positive) regions. Surprisingly, grafted cells retained the identity appropriate for their original positions in the donor embryo and formed an ectopic mandibular skeleton. Notably, these grafts reorganized the surrounding mesoderm to form an ectopic set of mandibular muscles, indicating that CNC is instructive for muscle development.^19,28 Similar results were obtained more recently with interspecific transplants of CNC contributing to the frontonasal process, between duck and quails, which showed that at least some aspects of beak shape are dictated by the CNC from the donor.^29,30 Likewise, in this case the transplanted CNC reorganized surrounding soft tissues including feather placodes derived from ectoderm and muscles derived from the mesoderm, indicating an instructional influence of the CNC on these tissues.

How do CNC cells that form the skeleton interact with the muscles to which they attach? In the head, this coordination appears to be regulated by specialized attachment cells (putative tendon or ligament precursors) that are also derived from CNC (fig. 3D). These resemble attachment cells in the limb, which derive from lateral plate mesoderm and are able to assemble a correct spatial pattern in the complete absence of myocytes.³¹ Within each arch primordium the CNC cells surround a central core of mesoderm that will form the muscles (fig. 3B). Here the initial “pioneer” muscle cells form attachments at precise locations along the skeleton through CNC-derived tendons. Fate mapping studies of CNC in chick-quail chimeras revealed a striking correlation in the embryonic origins of these two tissues. Tendon precursors attaching a muscle to a particular bone originated from the same A-P location in the CNC as the precursors of that bone.³² This suggests a mechanism whereby muscle attachment and bone shape are coordinated through a common process that determines the A-P identities of CNC cells prior to their migration, possibly through their Hox expression (or lack thereof ). Through this process, CNC-derived skeletal and tendon precursors may coordinate the formation of the entire functional skeletomuscular system.

Mutant studies in zebrafish also support the model that CNC cells play instructive roles in patterning cranial muscles. Histological analysis of the chinless (chn) mutant revealed that the pharyngeal skeleton and cranial muscles fail to form entirely.³³ Surgical replacement of the mandibular CNC with wild-type cells rescued both skeletal and muscle development. Rescued myocytes were confined to regions immediately adjacent to CNC derived from the donor, demonstrating the local nature of the interaction. Likewise, in moz and low (tfap2a-/-) mutants, reductions/transformations of the second arch (hyoid) skeleton correlate with disruptions of cranial muscles.^24,25

Patterning within Pharyngeal Arch Primordia

After CNC migration, each pharyngeal arch is organized cylindrically, with a core of mesoderm surrounded by CNC (fig. 3B). These in turn are surrounded by endodermal and ectodermal epithelia. Separate dorsal and ventral condensations (proximal and distal in mice) form in the CNC and these are serially reiterated in every arch (fig. 3A; reviewed in ref. 35). For example, the ventral/distal element (V1) of the first arch (mandibular) forms Meckel's cartilage (mc), the template for the lower jaw, while the dorsal/proximal element (D1) forms the palatoquadrate (pq), or upper jaw. Until recently, less was known about the trabeculae (tr) of the so-called “premandibular” skeleton (fig. 3C). Fate maps both in amphibian and avian embryos now suggest that adjacent condensations in the mandibular arch contribute not only to the jaws, but also to the neurocranium (arrow in fig. 3A). Dorsal condensations of the mandibular CNC give rise to cartilage of the neurocranium and fronto-nasal process, while both upper and lower jaws arise from more ventral condensations.^14,15 At face value, this challenges established notions of first arch development in which the only derivatives are thought to be mandible and maxilla. It also appears to differ from previous fate maps, including maps of mandibular CNC in zebrafish.³⁶ The precise locations of condensations, however, with respect to arch boundaries are difficult to compare between such widely divergent species and need to be defined relative to patterns of gene expression (as discussed below).

Comparative studies of vertebrate embryos suggest that many of the key vertebrate innovations evolved through changes in patterning along the D-V axis. It is interesting to speculate that such apparent differences in CNC condensations within the arches underlie: (1) the origins of jaws, (2) divergent patterns of jaw articulation between teleosts and amniotes, and (3) the eventual evolution of ear ossicles from dorsal skeletal components of the arches.^2,3,37 Teleosts exhibit a “hyostylic” jaw suspension, in which the mandibular arch is not attached directly to the neurocranium, but instead its dorsal element (D1; pq) forms a joint with the dorsal hyoid arch (D2; hs), and this in turn articulates with the skull (fig. 3C). In contrast, in amniotes the suspension is “autostylic”, with the lower jaw (V1) hinging directly upon the skull. Currently, it is thought that modification of a hypothetical primitive autostylic suspension in an ancestral vertebrate, led to a “amphistylic” condition seen in some primitive sharks, in which there is a direct articulation of the jaw (V1) with the braincase plus support from the dorsal hyoid arch (D2; hs). Subsequent modifications of this amphistylic articulation led toward hyostyly in teleost fish and autostyly in tetrapods with the consequence of the eventual freeing of mammalian homologues of the palatoquadrate (incus) and hyosymplectic (stapes) to function exclusively as sound transducers. Thus, modifications in the spatial and temporal sites of these D-V skeletal condensations have had major consequences for modes of feeding and hearing during vertebrate evolution.

One signaling molecule that plays a critical role in D-V skeletal patterning in the arches is Endothelin 1 (Edn1), and recent studies in fish and in mice have revealed conserved functions in formation of the lower jaw (fig. 4; reviewed in ref. 35). Edn1 acts through a type-A, G-protein coupled endothelin receptor (EdnrA), which is expressed in CNC and required for lower jaw development.^38-41 Kimmel et al⁴² have argued based on phenotypic changes in bony elements of the zebrafish edn1 mutant sucker, that Edn1 acts as a “morphogen” to specify fates along the D-V axis of an arch. This is based on the observation that while some sucker mutants simply lack the lower jaw, others show what appear to be partial duplications of dorsal skeletal elements in the ventral arch, possibly the result of a ventral to dorsal transformation caused by a loss of Edn1. Likewise, EdnrA receptor mutant mice show proximal-distal (D-V) transformations that result in small ectopic dorsal (maxillary) structures in place of the mandible and ectopic whisker barrels ventrally on the chin.⁴³ This suggests that Edn1 acts to pattern D-V identities of CNC in the arches, as well as surrounding facial ectoderm, and that cells exposed to the highest concentrations of Edn1 form ventral fates such as Meckel's cartilage.

Consistent with this model, expression of a number of transcription factors in ventral arch CNC is lost in Edn1 and EdnrA mutants. Of particular note are Dlx5 and Dlx6 (fig. 4). In mice, these genes are coexpressed broadly throughout the arches, though not to such a dorsal (proximal) extent as Dlx1/2. Mice lacking Dlx1/2 function lack proximal (maxillary) bones within the arches. Conversely, the combined loss of Dlx5/6 function results in a loss of distal (mandibular) structures and duplicated maxillae. Similar to EdnrA^-/- mutants, not only bones but also dorsal ectodermal tissues (marked by the whisker barrels) expand into the ventral arch.^44-46 The similarity between the transformations of cartilage and bone in Dlx5/6 and EdnrA mutant mice, points to a role for Dlx genes in potentiating the Edn1 signal. This is achieved through a nested pattern of Dlx gene expression along the D-V axis of the arch (fig. 4), which subsequently confers positional identity in combination with other ventrally expressed genes, such as HAND2, another target of Edn1 signalling.⁴⁷ D-V fates of CNC within the arch in this model are induced by a graded Edn1 signal from the distal part of the arch that activates expression of different combinations of Dlx genes. This distal Edn1 signal arises from either the pharyngeal ectoderm or endoderm, suggesting that the CNC does not possess an intrinsic D-V identity (ref. 41, unpublished results). These results amid others, have led to a reappraisal of the importance of other tissues in subsequent patterning of the CNC. Recent studies in chick and zebrafish now suggest that much of the positional information within an arch is initiated independent of the CNC.

Endoderm Patterns CNC and Is Required for Skeletal Development

All vertebrates form a series of pharyngeal pouches or pharyngeal slits that subdivide the arches in the embryo, and these are derived from endoderm (fig. 5). The pouches form as slender, bilateral outgrowths from the lining of the foregut during early segmentation of the embryo. Classic extirpation studies in amphibians and avians showed a requirement for this endoderm in pharyngeal cartilage formation.^48,49 Genetic studies in zebrafish have recently confirmed a role for pharyngeal endoderm in cartilage development and patterning of the CNC-derived head skeleton. Mutants that lack endoderm, such as the sox32 mutant, casanova (cas), fail to form pharyngeal cartilage entirely, while more dorsal neurocranial cartilages remain unaffected. Furthermore, these skeletal defects can be restored by transplantation of wild-type endoderm into cas mutants.⁵⁰ This suggests that CNC cells contributing to pharyngeal cartilages depend on the presence of pharyngeal endoderm whereas those forming the neurocranium do not. Cartilage patterning is also disrupted in tbx1 (vgo) mutants; pharyngeal pouches are variably lost and cartilages are reduced or fused.⁵¹ Similar to cas, the cartilage defects in vgo can be rescued by transplantation of wild-type endoderm, revealing the importance of the endoderm in both skeletal differentiation and patterning.⁵² Segmental development and patterns of gene expression within the endodermal pharyngeal pouches, on the other hand, occur independently of the presence of CNC.⁵³

To identify signals from the pharynx required for cartilage development, David et al⁵⁰ chose to examine fibroblast growth factor 3 (fgf3), which is expressed specifically in the endoderm of the pouches during cartilage condensation. By creating mosaic animals in which Fgf3 function was removed specifically from the endoderm, they showed that fgf3 is required for the formation of posterior, branchial cartilages (fig. 5D). Similar defects are seen in the fgf3 mutant lia.⁵⁴ Subsequent work suggests that Fgf3 acts together with Fgf8 in pharyngeal patterning.^55,56 Loss of both Fgf3 and Fgf8 function, following gene knockdown using antisense morpholino oligonucleotides, results in a loss of all head cartilages, including most of the neurocranium. In mice, loss-of-function mutations in Fgf8 targeted to the facial ectoderm of the first arch also disrupt skeletal patterning (ref. 57, see below). Thus, although these and other Fgfs are expressed in multiple cranial tissues throughout development, Fgf3 and Fgf8 in particular play critical roles in the epithelial-mesenchymal interactions that pattern the CNC-derived head skeleton.

When pharyngeal pouches are missing or malformed, streams of migrating CNC often do not separate from one another, pointing to a role for the pouches in guiding CNC migration. This does not affect the segmental identity of the CNC streams initially, as assayed by Hox expression during migration, but does result in subsequent loss of regional gene expression in the arches. The shape of the pouch also seems crucial in molding the shapes of certain skeletal elements, and for at least one pouch in the zebrafish this is an integrin-dependent process (fig. 5B).⁵⁸ In this respect, endoderm may stabilize adjacent skeletal precursors in the CNC. In the integrin-α5 mutant polypterus, the first pharyngeal pouch is reduced as are the adjacent dorsal cartilages of the hyoid arch. Time-lapsed movies of CNC development, using a transgenic line in which GFP is expressed in the postmigratory neural crest (fli1-GFP), revealed that cartilage precursors align along the edge of the pouch. Twists and bends in the pouches in the absence of integrin-α5 function lead to misalignment of GFP-positive CNC. These data suggest that the endoderm acts not only to pattern the CNC regionally and induce chondrogenesis through Fgf signaling, but may also impose the subsequent shape of each cartilage element.

Tissue-grafting studies in chick have also shown that endoderm plays an instructive role in maintaining the identities of CNC along the A-P axis and establishing the skeletal pattern.⁵⁹ Simply reversing the A-P orientation of the endoderm at the level of the midbrain can lead to ectopic mandibular structures. The anteriormost region of the endoderm is vital for formation and orientation of the nasal capsule, while more posterior endoderm promotes formation of the first pharyngeal arch. Importantly, it is only Hox-negative CNC cells that respond to the endoderm in this manner, suggesting that the Hox-positive CNC cells derived from hindbrain levels are patterned by different signals (see fig. 1). Those CNC cells derived from the Hox-negative parts of the neural tube exhibit a larger range of plasticity than those expressing Hox genes, which may in part reflect their origins near the mid-hindbrain boundary.²⁸ This highlights the fundamental differences in how different populations of CNC respond to environmental signals, which depends in part on Hox gene function.

Pharyngeal Ectoderm Patterns CNC

A second set of epithelial-mesenchymal interactions in cranial skeletal development occur between CNC and the surface ectoderm, which forms epidermis. Ectoderm was first shown to promote chondrogenesis in CNC by culturing facial ectoderm together with chick mandibular and maxillary mesenchyme.⁶⁰ Removal of the surface ectoderm correspondingly results in a failure of skeletal development in adjacent CNC mesenchyme. Only CNC cells that migrate superficially within the arches contact overlying surface ectoderm, but ectodermal influences (like those from endoderm) extend throughout the skeletogenic population (fig. 3B, fig. 5C). A key ectodermal signal in the first arch is Fgf8. Mice with a targeted loss of Fgf8 function in the mandibular ectoderm lack all but the most distal (ventral) skeletal elements (Meckel's cartilage), implying that an Fgf signal is necessary for proximal (dorsal) mandibular development.⁵⁷ This requirement for Fgf8 by proximal skeletal elements in the mandibular arch correlates with the proximity of CNC to a domain of Fgf8 expression in the oral ectoderm. More distal mandibular elements (Meckel's cartilage) do not develop adjacent to Fgf8 expressing ectoderm in mice and may respond instead to other Fgfs in the ectoderm, as well as other sources of Fgfs.

In zebrafish, CNC cells migrating into the mandibular arch do not express the Fgf inducible gene erm until after migration, and expression is first detected adjacent to the first pharyngeal pouch.⁵¹ How is this pattern regulated by Fgfs originating from the surface ectoderm versus Fgfs from the endoderm? Tissue-specific ablation of Fgf8 function in either ectoderm or endoderm during early mouse development indicates that the primary requirement for Fgf8 is in the ectoderm.⁶¹ Removal of Fgf8 function in mandibular ectoderm causes reductions and fusions of arch elements. CNC cells die through programmed cell death in these conditional mutants, similar to hypomorphic alleles of Fgf8.^57,62 In the zebrafish, fgf8 is also expressed in the mandibular ectoderm and the patterning of pharyngeal cartilages is disrupted in acerebellar (ace) mutants, which lack a functional Fgf8.⁶³

Recent studies on the function of tfap2a and its close relative, tfap2b, also support a role for the ectoderm in craniofacial patterning in zebrafish. In both mouse and zebrafish Tfap2a is expressed in nonneural ectoderm as well as CNC. Tfap2a loss-of-function mutants exhibit craniofacial defects, including a lack of second arch (hyoid) skeletal elements and these, at least in part, are cell autonomous requirements in CNC (fig. 2D).^25,64-68 However, transplants of wildtype CNC into zebrafish tfap2a mutants rescues hyoid arch outgrowth, but never completely rescues cartilage development, suggesting that there may be other functions for tfap2a in arch development, perhaps in the ectoderm. In zebrafish tfap2b is coexpressed with tfap2a in pharyngeal ectoderm but is not expressed in CNC.⁶⁸ Surprisingly, the combined disruption of tfap2a and tfap2b function using morpholino oligonucleotides results in the loss of both pharyngeal and neurocranial cartilages, a phenotype much more severe than loss-of-function mutations in tfap2a alone. This suggests that these two closely related genes play redundant roles in chondrogenesis, and that both are required in the ectoderm. Consistent with this model, the facial ectoderm shows elevated levels of apoptosis in embryos lacking tfap2a and tfap2b function (Tfap2 deficient), and transplantation of wild-type ectoderm into these embryos rescues cartilage development. This demonstrates a nonautonomous role for Tfap2 genes in CNC through their roles in promoting survival and patterning of the pharyngeal ectoderm.

An obvious target of Tfap2 genes in the ectoderm is the Fgf signaling pathway, given its role in chondrogenesis. Despite this, neither fgf3 nor fgf8 expression is perturbed in Tfap2-deficient animals. It is possible that Tfap2 genes regulate expression of other fgf genes in the ectoderm. Of the more than 25 known Fgf genes in mammals, several are expressed in the facial ectoderm, and may have overlapping roles with Fgf3 and Fgf8.⁶⁹ In support of a role for Tfap2 proteins in regulating Fgf signaling, the expression of a known Fgf target gene, sef, is reduced in CNC of the arches of Tfap2-deficient zebrafish.⁶⁸ This expression can be rescued by wildtype ectoderm, further confirming the nonautonomous requirements for AP-2 proteins in CNC development.

Several other signaling molecules expressed in the facial ectoderm are known to influence the craniofacial skeletal pattern. Like Fgfs, these include factors also expressed in endodermal and mesodermal tissues, such as Edn1 (see above), Bone morphogenetic proteins (Bmps; reviewed in ref. 70) and Sonic hedgehog (Shh; reviewed in ref. 71) (fig. 5C). A specialized region of facial ectoderm in the hyoid arch, known as the posterior ectodermal margin (PEM), expresses Shh. This pattern is conserved in the chick and mouse, and correlates with accelerated growth of the hyoid relative to the other arches. shh expression is lost in low mutants, which correlates with a failure of hyoid arch outgrowth.²⁷ Transplantation of wild-type CNC into low restores shh expression in the adjacent PEM, confirming the interaction and suggesting that it is direct. Ectodermal cells of the PEM may serve a role similar to the apical ectodermal ridge (AER) of the limb bud, maintaining growth and patterning of the underlying mesenchyme.³⁴ In turn, the reciprocal maintenance of the PEM by underlying CNC also resembles the AER. By comparison with the limb, we suspect that such epithelial-mesenchymal interactions coordinate development of various craniofacial tissues.

Shh and Fgf8 synergize in promoting cranial cartilage outgrowth in the chick.⁷² Domains of Bmp4 expression in the maxillary and mandibular ectoderm lie both anterior and posterior to Fgf8 expression and this appears to restrict the response of the underlying mesenchyme to Fgf8.^70,71 Interestingly, recent evidence suggests that domains of early Fgf8 and Bmp4 expression in ectoderm are specified surprisingly early in the chick, prior to CNC migration, and are regulated by interactions with endoderm.⁷³ Subsequent Bmp and Fgf signaling from this ectoderm induces the expression of two homeodomain transcription factors, Msx1 and Barx1 respectively, in the underlying CNC, both in mice and chick.^74-77 These prefigure the sites of tooth buds in mice, such that Barx1 specifies molars and Msx1 specifies incisors.^78,79 However, this regional specification of fate in the underlying mesenchyme is not just restricted to tooth type, but as mentioned earlier, also plays a role in defining the identity of the underlying cartilages and bones. This interaction of Fgf and Bmp signaling in specifying identity in the mandibular mesenchyme appears to be a conserved feature in all vertebrates including the sister group of the jawed vertebrates, the jawless lampreys.

CNC Development and Evolution

Evolution of a biting jaw was a crucial step in the evolution of jawed vertebrates (gnathostomes) from their jawless ancestors (agnathans) over 400 million years ago. The classical model of jaw evolution arises from the segmental theory of Goodrich,³ who proposed that a single cartilaginous element in the vertebrate ancestor became modified into a dorsal and a ventral element, the precursors of the upper and lower jaws in the mandibular arch. In accordance with this theory, the two extant agnathans, lampreys and hagfish (which form the sister group to all gnathostomes), have a segmented pharynx and CNC, but their arches are not clearly subdivided into dorsal and ventral components (reviewed in refs. 80, 81). One theory suggests that the jaw could have arisen by modification of the ventral cartilage element through changes in D-V patterning. Consistent with this idea, some genes that show ventrally restricted expression in the arches of gnathostomes, such as Dlx5/6 and Dlx3/7 (fig. 4), are expressed throughout the D-V extent of the arches during lamprey embryogenesis.^82,83 This lack of nested Dlx gene expression in CNC of the arches is reminiscent of Edn1 signalling mutants in which ventral Dlx gene expression is lost, leading to a loss of a clear ventral skeletal identity. This suggests that progressive restrictions in gene expression evolved into discrete D-V domains within the arches, possibly through the acquisition of new ventral signals.

D-V identity of CNC, however, is imposed by Dlx genes in all of the arches and does not confer a specific mandibular identity. What processes led to segment-specific changes in the mandibular arch that led to an articulated jaw during vertebrate evolution? Hox genes are obvious candidates, due to their highly conserved roles in imposing A-P identity along the body axis in all bilaterian animals. There are no Hox genes expressed in the mandibular arch (see fig. 1), and ectopic expression of Hoxa2 in this arch results in a loss of the mandible.^84,85 This suggests that the jaw evolved, in part, as a consequence of a loss of Hox expression in the anterior CNC that contributes to the mandibular arch. Consistent with this model, the brook lamprey, Lampetra fluviatilis, appears to express one Hox gene (HoxL6) in the mandibular arch.⁸⁶ However, more recent studies of a larger number of lamprey Hox genes do not support the model. In the Japanese lamprey, Lampetra japonicum, Hox expression is restricted to the hyoid and branchial CNC streams and arches, similar to gnathostomes.⁸⁷

This reveals that agnathans and gnathostomes show a similar conserved pattern of early Hox expression and CNC patterning. This implies that there are dramatic differences in later events that determine the eventual fates of CNC-derived skeletal structures between jawed and jawless vertebrates. These may include differences in signals from adjacent tissues, such as the endoderm or the ectoderm, as well as in CNC responses to these signals. Shigetani and colleagues⁸⁸ tested if Fgf and Bmp signaling from the pharyngeal ectoderm to the underlying CNC is conserved between gnathostomes and lampreys. They showed that the same target transcription factors, Dlx and Msx genes (see above), are similarly induced in the underlying CNC in both groups, but importantly, that this activation differs spatially in relation to CNC condensations. Thus, despite the conserved manner of CNC migration into the mandibular arch in agnathans, certain parts of the lamprey pharyngeal apparatus may not be directly comparable to the jaws of gnathostomes. Both share the same molecular cues for patterning CNC in the arches, but they differ in their spatial interpretation. Central to these features of jaw evolution may have been the interpretation of signals from adjacent epithelia to condense at specific locations and assume identities based on their spatial position within the arch. Future investigations into the nature of these signals will lead to a better understanding of both the development and evolution of the skeleton in different vertebrate lineages.

Summary and Conclusions

CNC cells enter the pharyngeal arches to become the cartilage and bones of the jaw and middle ear in mammals. They also form the bones of the frontonasal process, the dentine of the teeth and the peripheral neurons and glia of the cranial nerves. Specification and migration of the skeletogenic CNC depends on interactions with surrounding tissues. Paracrine factors from these regions induce the expression of transcription factors that control CNC migration and differentiation. This is particularly well known for patterning along the D-V axis within the pharyngeal arches, whereby Edn1 signaling and the nested expression of Dlx genes define distinct domains within each pharyngeal arch segment. Other signals received by CNC cells in the arches include Fgfs, Bmps and Shh which are secreted by the facial ectoderm and pharyngeal endoderm in a dynamic fashion. Changes in the sources of these signals and the responses by the CNC have been central to the evolution of craniofacial development.