Amniotes have in common an egg with a particular arrangement of extraembryonic membranes (amnion, chorion and allantois), and primitive members of this class were the first life forms to live exclusively on land. The most ancient amniotes arose during the Early Carboniferous period and had the appearance of small lizards. There are two main groups of Amniota: Synapsida (mammals and their extinct relatives) and Sauropsida (snakes, lizards, crocodilians, turtles, birds, and their fossil relatives including dinosaurs; Fig. 1). These classifications are based primarily on the fenestration of the temporal region of the skull (Carroll, 1988; Pough et al., 2005). Within the group of Sauropsida, two main skull morphologies can be recognized: Diapsida (two temporal openings both posterior to the orbit: includes birds, snakes, lizards, and crocodiles) and anapsida (no temporal openings: turtles). The Synapsida had a single temporal opening dorsal to the eye, which in modern mammals is merged with the orbit. Here it is important to mention that turtles are in a debated position in amniote evolution depending on whether molecular (Hedges and Poling, 1999) or morphological data are sampled (Hill, 2005). On purely molecular phylogenetic studies, turtles are closely related to crocodilians and birds (Hedges and Poling, 1999; Cao et al., 2000; Matsuda et al., 2005). Such close relationships between turtles and Archosauria (birds, crocodiles, and extinct dinosaur relatives) would mean that turtles lost their temporal fenestration during the course of evolution (Rieppel, 1999). Resolution of these debates requires an integration of physical characteristics from extant and extinct amniotes with molecular data from as many types of amniotes as possible (Hill, 2005). As more genes from turtles and reptiles are sampled, we may be able to finally resolve the position of the turtle in the amniote tree.
Analysis of the temporal sequence of the appearance of the major amniote subgroups, Lepidosaura (lizard, snakes, and tuatara), Archosaura (birds, crocodiles, and possibly turtles), and Synapsida, shows that the separation of the Synapsid branch occurred very early after the initial appearance of basal amniotes (Romer and Parsons, 1977; Reisz, 1997). Stated another way, nonmammalian amniotes are not more primitive or less derived than are the Synapsida. Of all amniotes, Aves are considered to be the most recent radiation, with the first feathered birds appearing during the Jurassic period (Zhou, 2004).
The most commonly used models for developmental studies of the face are rodents and birds, representing only two of the main amniote groups. There are many elegant developmental studies on the molecular and tissue interactions that pattern the mandible (Tyler and Hall, 1977; Hall, 1980; Atchley and Hall, 1991; Tucker et al., 1999; Ferguson et al., 2000; Tucker et al., 2004; Wilson and Tucker, 2004). In addition, lower jaws rather than upper jaws have been used for classifying extant and extinct amniote groups (De Beer, 1937). Considerably less attention has been paid to the ontogeny and phylogeny of the upper jaw. The amniote upper jaw has undergone a wide array of specializations in response to different ecological challenges (e.g., Fig. 1). Here, we review the anatomical relationships, embryonic origins, and the ontogeny of the upper jaw bones in representatives of the major amniote groups. Finally, we propose models by which morphological variation has arisen in the upper jaw of different amniote taxa.
MAJOR DIFFERENCES IN ANATOMY OF THE UPPER JAW BONES BETWEEN AMNIOTES
In this review, we use the term “upper jaw” to include the premaxilla, maxilla, vomer, jugal, ectopterygoid, quadratojugal, and pterygoid bones. All of these bones directly ossify from mesenchyme (intramembranous ossification) rather than by replacement of a cartilage template (endochondral ossification). During the course of evolution, facial bones in different taxa were lost or gained in response to selective pressures. To appreciate some of these major differences, we next compare the position and morphology each of the upper jaw bones in representative, modern amniotes.
“Extra” Bones in Reptiles
It is generally true that non-avian reptiles have a greater number of skull and jaw bones than do birds and mammals, and in some cases, the extra elements are remnants of an ancestral phenotype. One such example is the septomaxilla, which is associated with the nasolacrimal duct, and is represented in many primitive amniotes. It is interesting that the timing of appearance of the septomaxilla in the fossil record coincided with the move of tetrapods onto land. However, the relationship between having a septomaxilla and living on land is not a strict one. This bone has been lost in birds, crocodilian, and the majority of mammals with the curious exception of monotremes (platypus) and edentates (armadillo; Fuchs, 1911; De Beer, 1937). In contrast, snakes and some lizards retain the septomaxilla where it contacts the premaxilla distally and the nasal bone medially (Figs. 1, 2A,B). As a result, the septomaxilla forms the floor of the nasal cavity and encloses the vomeronasal organ (Romer, 1956). The snake may require this arrangement so that the prefrontal and septomaxilla can support the nasolacrimal duct while leaving other palatal components (pterygoid, palatine, and maxilla) to move independently. Clearly other bones such as the nasal, vomer and nasolacrimal have taken over the function of the septomaxilla in modern mammals.
The ectopterygoid bone is found in non-avian reptiles (Figs. 1, 2A,B) but not in birds or mammals (Figs. 1, 2C–H; De Beer, 1951; Romer and Parsons, 1977). One theory is that the ectopterygoid has been incorporated, perhaps by direct fusion, into the mammalian medial pterygoid process and contributes to the hamular cartilage (Presley and Steel, 1978). We question how the exclusively intramembranous reptilian ectopterygoid bone could be homologous to a cartilaginous hamular process. Based on close proximity in the jaw and the observation that the ectopterygoid and pterygoid form one plate of bone in primitive amniotes (Carroll, 1988), we hypothesize that the reptilian ectopterygoid derives from a similar region in the embryonic face to the pterygoid.
Determining Homology Among Amniote Facial Bones Despite Differences in Function, Position, and Morphology
One of the distinctive features of the mammalian and avian upper jaw is the arrangement of the cheek bones. The avian jugal bone (malar or zygomatic bone in mammals) forms part of an arch of bone, curving outward from the brain case (Figs. 1, 2C,D). The design of the zygomatic arch in mammals increases the efficiency and force of jaw closing. The masseter inserts into the zygomatic arch, and the temporalis fits underneath the arch. By projecting the jugal bone lateral to the brain case, there is increased space available for these muscles, leading to expansion of muscle size and consequently increased muscle force (Pough et al., 2005). Because birds have a different phylogenetic lineage than mammals, the avian jugal bar arose independently from the zygomatic arch of mammals. In aves, jaw closing is mainly achieved by contracting the pterygoideus muscle, which originates on the pterygoid and palatine bones and inserts into the mandible. The jugal bar may serve to lighten the skull, because flightless birds lack this feature. In contrast, most reptiles either have a jugal bone that is a part of the braincase, lying posterior to the eye or they lack this bone entirely (as in snakes, Figs. 1, 2A,B). Thus, the reptilian muscles of mastication must insert directly on the sides of lower jaw and lateral skull wall (Romer, 1956). This primitive jaw closing apparatus may mean that there is less force generated on closing or that other muscles, such as the pterygoideus, may provide additional power.
Similar to the jugal, where functional homology between mammals and birds is not present despite similar morphology, there are other bones of the upper jaw where it is difficult to find homologous characteristics. There is a long-standing controversy about whether it is appropriate to homologize the mammalian pterygoid plates and pterygoid bones of Diapsida. In all Sauropsida (extant and extinct) as well as in primitive Synapsida, the pterygoid is a separate bone in the posterior secondary palate (Figs. 1, 2A–D). In contrast, in modern mammals, there are no distinct pterygoid bones but rather there are two small, vertical plates attached to the sphenoid bone (Fig. 1). The two alternative hypotheses are as follows: (1) that the mammalian pterygoid plates are homologous to reptilian pterygoids but have been reduced and assimilated into the sphenoid bone (Presley and Steel, 1978), or (2) the pterygoid bone has been lost in mammals (De Beer, 1951). Comparative studies of the developing pterygoid in mammals, birds, and reptiles are required to resolve these debates.
Other bones for which homologies are more obvious are the premaxillary, maxillary, and palatine. The premaxillary bone is always the most distal in the upper jaw, the maxilla is an intermediate position and the palatine bones are the most proximal. Primitive amniotes such as the Captorhinids (Reisz, 1997) and Pelycosaurs (Carroll, 1988) show this characteristic arrangement (Fig. 1). In most amniotes, the premaxilla, maxillary, and palatine bones are joined by sutures that fuse, immobilizing the upper jaw. However, in snakes, the premaxilla is not in contact with the maxillary bone (Frazzetta, 1966). The uniqueness of the jointed upper jaw of snakes is an example of a relatively recent, derived trait that helps to increase the gape of these reptiles.
The relative contributions of premaxillary, maxillary, and palatine bones to the upper jaw in major amniote groups is quite different. For example, in mammals and alligators (Ferguson, 1981), the maxillary bone comprises most of the upper jaw with the palatine bones and premaxillary bones making minor contributions (Figs. 1, 2E–H). In aves, the largest bone is the palatine, whereas the maxillary bone is very small (Figs. 1, 2C,D). The other major difference between birds and other amniotes is the size and prominence of the premaxilla, the major bone defining upper beak shape. In Lepidosauria (lizards and snakes), turtles, and mammals, the premaxilla is small and relatively recessed, so much so, that the nasal cavity is flush or projects outward beyond the premaxilla (Fig. 2A,B,E–H). In addition, there are some major differences in the articulations between these three palatal complex bones among amniotes (Table 1). In mammals, the palatine bone is posterior to the maxillary bone and makes no contact with the premaxilla (Figs. 1, 2E,F), whereas in Lepidosauria and birds, palatine bones extend from the pterygoid to the premaxilla (Fig. 1; 2A–D; Table 1). In some cases, there are additional bones that comprise the amniote palatal complex. In Diaspida (except Crocodilia), the vomer is exposed and provides major support for the roof of the oral cavity (Fig. 1). In contrast, the mammalian vomer is within the nasal cavity and does not directly contribute to the palate.
Table 1. Anatomical Relationships of Upper Jaw Bones in Selected Amniotesa
The articulations of each of the upper jaw bones is listed to complement Figure 1 and to provide more three-dimensional information. Where the exact bone name is not present in a certain species, the homologous structures are listed (e.g., Pterygoid and Ectopterygoid; Presley and Steel, 1978). This table also highlights bones for which no homologous structure is present (indicated by not present, NP).
Ectopterygoid (homologous to hamular cartilage of pterygoid process in mammals)
Jugal (also zygomatic/malar)
maxilla; temporal; frontal; sphenoid
maxilla; pterygoid; prefrontal
premaxilla; maxilla; pterygoid
maxilla; pterygoid; vomer; frontal
maxilla; vomer; sphenoid
Pterygoid (homologous to medial and lateral pterygoid plates of sphenoid in mammals)
The above discussion highlights some of the difficulties in discerning homology between skeletal structures based on postnatal structure and function. One of the recurring themes is that the most convincing evidence of homology may come from ontogenetic similarities rather than postnatal form (Donoghue, 2002). For example, despite the differences in palatal design discussed above, the initial budding of the palatal shelves from the maxillary prominences is conserved throughout amniote phylogeny. During embryogenesis, the palatal shelves initially grow outward into the oral cavity but in Lepidosauria and aves (Fig. 3D,F), they do not contact and fuse (Shah and Crawford, 1980; Koch and Smiley, 1981). In mammals, the palatal shelves grow to be quite large and then reorient to extend cross the oral cavity (Fig. 3G,H). After contact, a bilayered epithelial seam forms (Fig. 3H) and then degenerates leaving a solid mesenchymal bridge (Fig. 3J). A similar type of fusion in the secondary palate is present in Crocodilia (Ferguson, 1981). Ossification centers for the palatine processes of the maxillary bone and the horizontal plate of the palatine bones form on either side and eventually meet in the center to form the midline suture. Thus, the presence or lack of fusion at early stages sets the stage for the morphology of the bones supporting the secondary palate. Because large variations in postnatal morphology can occur with very subtle changes in early stages of development, we next review the ontogeny of the upper jaw.
ORIGINS AND FATES OF THE FACIAL PROMINENCES
The vertebrate upper jaw bones originate from Hox-negative cranial neural crest cells from rhombomeres 1 and 2, the mesencephalon, and the diencephalon (reviewed in Le Douarin, 2004). The unique feature of cranial neural crest cells is their skeletogenic capacity, which is related in part to the lack of Hox gene expression (Abzhanov et al., 2003). Ultimately, neural crest-derived facial mesenchyme forms virtually all of the intramembranous bone of the face.
Cranial neural crest cells move away as a sheet from the neural tube and are split into two portions once the eye is encountered (Creuzet et al., 2005). The neural crest cells that are cranial to the eye move ventrally and contribute to the mesenchyme in the center of the face (preoptic neural crest cells). The neural crest cells that pass inferior to the eye move into the maxillary and mandibular regions (postoptic and mandibular; Johnston, 1966; Noden, 1975; Lumsden et al., 1991). The main influences on migratory, preoptic neural crest cells include signals from the forebrain and the surface epithelium such as retinoids, fibroblast growth factors (FGFs), Sonic Hedgehog (SHH), and WNTs (vertebrate genes related to drosophila Wingless; Schneider et al., 2001; Cordero et al., 2004; Junghans et al., 2005; Marcucio et al., 2005). The migratory, postoptic neural crest cells encounter signals provided by the endoderm and floor plate of the brain (SHH; Ahlgren and Bronner-Fraser, 1999; Schneider and Helms, 2003) as well as from the ectoderm (FGF8, bone morphogenetic proteins [BMPs]; Shigetani et al., 2000; Schneider et al., 2001; Creuzet et al., 2004; Marcucio et al., 2005). After neural crest cells reach their destinations in the face, the mesenchymal cells interact with tissues in their local environment, and these reciprocal interactions guide their eventual fate.
Close to the time when cranial neural crest cells have reached the ventral side of the neural tube, the pharyngeal arches form. The first and most superior arch is commonly thought to give rise to maxillary and mandibular processes and ultimately to most of the upper and lower jaws (Kuratani, 2005; Shigetani et al., 2005). However, surprisingly little experimental work had been done to test this assumption. Studies on intermediate stages of development, between when neural crest cells begin to migrate and skeletal condensations form, were required. The first such experiments were done in the lab of Kuratani (Shigetani et al., 2000) and involved focal dye labeling of the avian stage 13 (Hamburger and Hamilton, 1951) first arch and postoptic region (Fig. 4A), a stage at which most neural crest cell migration has ceased (Baker et al., 1997). Embryos were followed until the maxillary prominence had just begun to emerge (stage 18). Dye put into the maxillomandibular cleft at stage 13 labels both the proximal maxillary and mandibular prominences (Shigetani et al., 2000), whereas dye injected midway between the eye and maxillomandibular cleft remains closely associated with the periocular tissues. The origins of at least part of the maxillary and mandibular mesenchyme are similar, although mandibular mesenchyme also has a major contribution from the distal first arch. A more extensive study from our lab (Lee et al., 2004) produced slightly different results. After label of the postoptic region at stage 13, we were able to resolve an additional small group of cells that had entered the maxillary prominence (Fig. 4A). Because our endpoint was later (stage 24) than that of Shigetani et al. (stage 18), the maxillary bud was more obvious, thus making it easier to see whether mesenchyme was labeled (Fig. 4A). Furthermore, in another series of experiments conducted at stage 15 (6–8 hr older than stage 13), we show that cells labeled between the eye and first arch hardly change position and become incorporated into the maxillary prominence (Figs. 4B, 5A). The lack of dye spread is consistent with the formation of early mesenchymal condensations. These data pave the way for future studies in which condensation formation is specifically perturbed.
Although not an amniote, the axolotl has very clear mesenchymal condensations shortly after the cessation of neural crest cell migration (Cerny et al., 2004). In recent dye-labeling studies (Cerny et al., 2004), the maxillary condensation gives rise to trabecular cartilage (parachordal cartilage), whereas mandibular condensation forms Meckel's cartilage and the palatoquadrate (reduced to the quadrate in birds and reptiles). Thus, similar to our results in the chicken, the maxillary condensation is distinct from the mandibular condensation. Furthermore, the condensations fated to form the facial skeleton in axolotls are present by the time neural crest cells have ceased migration.
It is not known whether mammals have shared or distinct maxillary and mandibular condensations. In mice at embryonic day (E) 8.5, the early first arch is relatively large compared with the rest of the head, extending from the primitive eye to the heart. In addition, the first arch swelling is present during active neural crest cell migration (Serbedzija et al., 1992; Osumi-Yamashita et al., 1994; Trainor and Tam, 1995). Thus, it is likely in mammals that very early maxillary and mandibular condensations are both contained in the first arch. In contrast, in chicken, the mesencephalic and diencephalic neural crest cells have all but completed migration before formation of the first arch (stage 13; Lumsden et al., 1991; Baker et al., 1997). Thus, one must be careful about extrapolating results from aves to mammals, especially with such important differences in developmental histories of the first pharyngeal arch.
It is also important to remember that the early neural crest-derived mesenchymal condensations are distinct from the mesodermal core of the pharyngeal arches, whereas neural crest cells are intermixed with parasomitic mesoderm in the presumptive maxillary regions (Serbedzija et al., 1992; Osumi-Yamashita et al., 1994; Trainor and Tam, 1995). The mesoderm is fated to give rise to muscles rather than skeletal tissue in the face (Couly et al., 1992; Noden and Trainor, 2005) and patterning of muscle insertions is coordinated with neural-crest derived skeletal elements (Le Lievre, 1978; Noden, 1978; Couly et al., 1993; Kontges and Lumsden, 1996). Despite the interactions with mesoderm and other tissues from beginning to end of migration, neural crest-derived mesenchymal condensations are not fully specified until they become incorporated into facial prominences.
Fate of Facial Prominences
The majority of avian craniofacial fate maps were originally constructed at the neurula stage of development and carried forward to when full skeletal pattern was evident (Le Lievre, 1978; Noden, 1978; Couly et al., 1993; Kontges and Lumsden, 1996). However, the relationships of the facial prominences to the facial bones are not as well studied. Through a combination of grafting and other types of labeling experiments, we know that in the chicken, frontonasal mesenchyme gives rise to the premaxilla, prenasal cartilage and frontal bone (Wedden, 1987; Richman and Tickle, 1989; Lee et al., 2004; Fig. 4C). The lateral nasal prominences give rise to the nasal conchae and nasal bones (MacDonald et al., 2004). The maxillary prominence in birds and probably also in other amniotes gives rise to the maxillary, palatine, jugal, and quadratojugal bones (Fig. 4C; Lee et al., 2004). The areas proximal to but not actually contained within the maxillary bud are likely to form more proximal upper jaw elements such as the pterygoid and perhaps part of the quadratojugal, however, this finding has not been determined directly (Fig. 4C). There is no other information on the fate of facial prominences in other diaspids. Because lepidosaura eggs are oviposited well after facial prominences have fused, other models (crocodilian and turtles) would be more suitable for fate mapping studies. Information on the fate of mammalian facial prominences is starting to become available. Promoter elements that restrict expression of Cre recombinase to specific facial prominences or regions of facial prominences have been generated as part of the process of creating conditional knockouts (Ruest et al., 2003a, b; Nelson and Williams, 2004; Liu et al., 2005a, b). Careful analysis of such transgenic models crossed with reporter mice will determine whether there are differences in contributions of facial prominences to jaw bones in mammals as compared with birds.
The identity of the facial prominences (upper or lower jaw) is fixed shortly after they begin to bud outward. Exchanging ectoderm and mesenchyme between prominences in avian embryos shows that the mesenchyme dictates what type of skeletal element will form (Richman and Tickle, 1989). Although not directly tested, the temporal aspects of specifying facial mesenchyme fate are likely to be similar in other amniotes.
GENETIC (AND EPIGENETIC) REGULATION OF THE NUMBER, POSITION, AND TIMING OF APPEARANCE OF UPPER JAW BONES
Recognition of Early Mesenchymal Condensations That Differentiate Into Skeletal Elements
Neural crest-derived preosseous condensations are contained within the facial prominences and give rise to membranous bones. There are several important differences between prechondrogenic and preosteogenic condensations. First, preosteogenic condensations are defined as being subadjacent to epithelia (Fig. 5B) and second, as having no overt histiogenic signs of bone formation such as Von Kossa or Alizarin Red staining (Dunlop and Hall, 1995). Markers for these preosteogenic condensations are scant (as discussed later). In contrast, prechondrogenic condensations are centrally located (removed from the epithelium) and express the transcription factor Sox9 (Eames et al., 2004) as well as a host of extracellular matrix markers (Hall and Miyake, 2000). Other properties of preosteogenic condensations are the requirement for epithelium to be present for a certain period of time in order for bone to differentiate, whereas in some amniotes cartilage can form in the absence of this interaction (Hall, 1980). Condensations fated to make bone are defined by the relatively higher cell proliferation immediately subadjacent to facial epithelium (Minkoff and Kuntz, 1978; Bailey et al., 1988; MacDonald et al., 2004; Wu et al., 2004). Spatially restricted proliferation may help to recruit cells into the condensation (Fig. 5B). In addition, programmed cell death at the edges of the facial prominences may help to sculpt the preosseous condensations (Barlow and Francis-West, 1997; McGonnell et al., 1998; Hirata and Hall, 2000; Song et al., 2004). Finally, it is possible that some type of cell sorting takes place to preferentially group preosteoblasts together, but we are only just beginning to explore such mechanisms (Kawakami et al., 2006).
The problem of distinguishing pre- or early osseous condensations from cartilaginous condensations has been a barrier to progress in understanding how jaw patterning is modified during evolution. The studies from the Hall group that were published a decade ago were done without some of the newer molecular markers (Dunlop and Hall, 1995). Here the authors used peanut agglutinin (PNA) and several other criteria (i.e., position relative to epithelium, timing of appearance) to highlight areas that will form intramembranous bone in stage 26 avian embryos. To put this stage into a developmental perspective, it is 2.5 days after the end of neural crest cell migration and approximately 3.5 days before overt osseous differentiation. More recently, a detailed molecular analysis of the same stage avian developing mandible (Eames and Helms, 2004) showed that a bone transcription factor, Runx2 (Runt-related transcription factor; also known as Cbfa1), is expressed in the same areas of mesenchyme close to the epithelium as recognized by Dunlop and Hall (1995). Runx2 is required for all bone formation, including endochondral bone (Ducy et al., 1997; Komori et al., 1997; Otto et al., 1997). When the data are compared (forgiving differences in plane of section) Runx2 transcripts are more localized compared with those areas stained with PNA, suggesting that PNA is not a very specific marker for intramembranous bone condensations. Mouse Runx2 expression domains are subadjacent to the epithelium and in the distal mesenchyme of the mandibular and maxillary prominences at E10.5 (Stricker et al., 2002). The mouse data are consistent with this transcription factor being expressed well before osteoblast differentiation occurs. It is true that, later in development, Runx2 is also expressed in differentiating cartilage (Karsenty and Wagner, 2002), but this expression is easily distinguished from the more peripheral expression in the preosseous condensations. It is important to find a good early marker for the condensations that give rise to bone so we can study stages when evolutionary changes in morphology are hypothesized to have occurred (Atchley and Hall, 1991; Hall and Miyake, 2000) and when it is likely that some of the craniofacial skeletal phenotypes in transgenic mice are initiated.
Epithelial Signals That Determine the Fate of Preosseous Condensations
The interactions of mesenchyme with epithelium are an important early signal in the induction of osteogenic condensations. Once this signal is received, the mesenchyme can go on to form bone even in the absence of epithelium (Tyler and Hall, 1977; Tyler, 1978; Hall, 1980; Van Exan and Hall, 1984; Dunlop and Hall, 1995). In mice, the requirement for epithelium to induce bone is present until E10.5 (Hall, 1980), and in chickens, epithelium must be present up until stage 22 (Tyler and Hall, 1977). In addition to the general requirement for bone induction, frontonasal mass epithelium at certain stages has been shown to be capable of reorganizing the pattern of the underlying skeleton (Hu et al., 2003). In this experiment, duplicated upper or lower beak structures were produced from epithelial grafts into the frontonasal mass or mandible, suggesting that supernumerary condensations were formed. The growth factors expressed in the grafted epithelium include Fgf8, Shh, and Bmp4, and these same signals are also expressed in other facial epithelia. The transcripts for Bmp4 are initially restricted to the edges of the frontonasal mass epithelium, close to areas that will later form the premaxillary bone (Francis-West et al., 1994; Ashique et al., 2002) and the superior half of the maxillary prominence (Barlow et al., 1999; Ashique et al., 2002; Song et al., 2004), an area that gives rise to the maxillary and jugal bones (Lee et al., 2004). In older embryos (stage 24 and up), Bmp4 expression is located in the mesenchyme; thus, the timing is similar to the epithelial requirement for inducing bone (Tyler and Hall, 1977; Tyler, 1978) Two studies have shown that relatively early (Abzhanov et al., 2004) and prolonged (Wu et al., 2004) BMP expression in the presumptive premaxillary preosseous condensation (expression directly under the frontonasal mass epithelium) is correlated with either thickened finch beaks or broad duck bills, respectively. The temporal and spatial regulation of Bmp4 expression in different bird beaks suggests that this gene is one of the functional determinants of osseous condensations.
Several experiments reinforce the role for BMPs in patterning preosseous condensations. When exogenous BMPs are applied to the stage 24 developing maxillary prominence mesenchyme at stage 20, 1 day before formation of condensations, a duplication of the palatine bone occurs (Barlow and Francis-West, 1997). The proteins released from beads are usually exhausted within 24 hours, before condensations were previously thought to begin in this part of the face (Dunlop and Hall, 1995). It is reasonable to assume that, to make a second bone, would require either a second condensation or expansion of the original one. A similar experiment, where increased levels of BMPs in the frontonasal mass had deleterious effects on bone formation (Ashique et al., 2002), suggested that the response of mesenchyme to BMPs is biphasic and that there are position-specific effects of BMPs.
More investigations of the role of BMP signaling in facial morphogenesis have been carried out in mouse models. Genetic deletion of BMPs results in early lethality, but more recently, several face-specific conditional knockouts have been described. For instance, Nestin-Cre deletion of Bmp4 and Bmp receptor1a highlights the importance of BMP signaling in upper facial formation, since cleft lip and cleft palate results (Liu et al., 2005a, b). The bones were not analyzed in these conditional knockouts, but it is likely that there were defects in maxillary and premaxillary bones, secondary to the clefts.
Fgfs are also expressed in facial epithelia and could be important for bone condensations of the face. Fgf8 is expressed in complementary patterns to Bmp4 in the face (Barlow et al., 1999; Shigetani et al., 2000; Liu et al., 2005a). There is a well-described antagonism between FGF and BMP signaling in embryo development including in teeth and limbs (Niswander and Martin, 1993; Neubuser et al., 1997) and a similar interaction is possible in the face. The opposing effects may help to control the positions of osseous condensations. Genetic deletion of Fgf8 with Nestin-Cre recombinase after neural crest cells have arrived in the facial prominences led to abnormally small or absent maxillary prominences (Trumpp et al., 1999). The mandibular prominences were also very small. However, the medial and lateral nasal prominences were normal. When embryos were allowed to develop further, significant defects in the bones were observed. Deletion of Fgf8 in the face led to loss of almost all the upper jaw bones derived from the maxillary prominence, including maxilla, palatine, and jugal (Trumpp et al., 1999). The premaxillary bone continued to form however, even though Fgf8 is also highly expressed in the medial nasal prominences (Crossley and Martin, 1995). Further analysis of where the Nestin-Cre recombinase is acting in the face (Liu et al., 2005b) revealed that there is a partial excision in the medial nasal prominence epithelium and mosaic excision in the maxillary prominence mesenchyme plus epithelium. Surprisingly, especially in view of the severe Fgf8 mandibular phenotype, Cre recombination is less complete in the mandibular prominence. In the Fgf8 mutants, there was an early increase in apoptosis, which may have decreased the critical mass of cells necessary to make the condensations. It is equally possible that some of the morphogenetic effects on the bones are indirect and happen later than the condensation stages. Because the osteogenic condensations were not evaluated, we are only able to evaluate the final result on fully formed bones.
Several members of the Wnt gene family are expressed in the epithelium and mesenchyme of the developing facial region (Dealy et al., 1993; Parr et al., 1993). Moreover, there are high levels of endogenous Wnt–β-catenin signaling in the E10.5 mouse face as determined by a reporter mouse (TOPGAL; Smith et al., 2005). Knockouts of Wnt5a (Yamaguchi et al., 1999), Wnt9b (Carroll et al., 2005), or forebrain-specific deletion of β-catenin signaling (Junghans et al., 2005) lead to facial bone defects. It is interesting that both noncanonical (Topol et al., 2003) and canonical (by means of β-catenin; Brault et al., 2001; Junghans et al., 2005) Wnt signaling are involved in jaw morphogenesis. The roles that Wnt family members have in epithelial–mesenchymal interactions and in endochondral bone formation (Gaur et al., 2005; Glass et al., 2005), together with the short distance signaling characteristics (Logan and Nusse, 2004), make it highly likely that they are involved in some of the early and later steps of intramembranous ossification (Fig. 5A–D).
In addition to epithelium as a source of signals for inducing bone condensations, the foregut endoderm is also an important signaling center. Excision of the most anterior forgut endoderm leads to ablation of the upper beak (Couly et al., 2002). Furthermore, the endoderm has an important role signaling to the presumptive facial epithelium and inducing Fgf8 expression in the first pharyngeal arch (Haworth et al., 2004). Deciphering the expression profile of the endoderm will be critical to understanding the tissue interactions that instruct osteogenesis in the intramembranous bones of the face.
CONTROL OF BONE IDENTITY AND MORPHOLOGY
Endothelin–Dlx Pathway and the Control of Maxillomandibular Identity
There are several genes that have specifics roles in controlling the identity of the jaws. These genes include several transcription factors that are expressed in the neural crest-derived facial mesenchyme before differentiation of bone and result in craniofacial phenotypes in gene targeting experiments (Francis-West et al., 2003). Of these, there is a subset of genes that have direct roles in patterning the upper jaw. The Distaless (Dlx) transcription factors are a family of six genes that are expressed in nested domains in the facial prominences at early mesenchymal condensation stages (Depew et al., 2005). Unlike Runx2, the Dlx genes are not restricted to the subepithelial mesenchyme and, therefore, are not considered to be markers of the later forming, preosseous condensations. The Dlx genes do appear, however, to confer identity on early mesenchymal condensations. Most of the Dlx genes have been knocked out either singly or in combination and all of these give rise to craniofacial skeletal changes, many of them in the upper jaw (Depew et al., 2005). Of particular note is the result of Dlx5/6 deletion (normally only expressed in the mandibular prominence) where the lower jaw is transformed into a recognizable maxillary bone and associated elements (Beverdam et al., 2002; Depew et al., 2002). Several other combinations of Dlx mutant mice with similar duplications of the maxillary derivatives are discussed fully in Depew et al. (2005). Partial identity changes occur in single knockouts and various compound knockouts (Depew et al., 2005). For example, targeted deletion of Dlx5 leads to ectopic palatine bones that are located more caudal to the normal palatine shelves, supernumerary struts of membranous bone projecting laterally from the basisphenoid, and abnormalities of the pterygoid plates (Depew et al., 1999). Deletion of Dlx 1 and 2 in single or compound mutants also affects the pterygoid by reducing the size, causing separation from the basisphenoid and delaying the onset of ossification (Qiu et al., 1997). In addition, there were large lateral projecting intramembranous bones, connected to the basisphenoid whose identity could not be confirmed. Whether these extra bones represent a partial change in identity or recapitulation of a more primitive amniote is not clear.
Some of the most direct upstream signals that control Dlx gene expression are Endothelins. Blocking endothelin receptors with specific antagonists (Fukuhara et al., 2004) or targeting the gene coding for the ligand or the receptor (endothelin-1, endothelin receptor-A; Ozeki et al., 2004; Ruest et al., 2004) lead to loss of Dlx5 and Dlx6 expression. Most strikingly, the deletion of endothelin-1 or ETRA leads to homeotic transformation of the mandible into an upper jaw (Ozeki et al., 2004; Ruest et al., 2004) similar to that produced in the Dlx5/6 double knockout. Other evidence that endothelin acts upstream of Dlx5/6 is that conditional inactivation of the Endothelin receptor in neural crest cells causes an ectopic strut of bone to form lateral to the basisphenoid (D. Clouthier, personal communication) similar to what is reported in the Dlx1/2 knockouts (Depew et al., 2005). This ectopic bone may recapitulate the large lateral flange of the pterygoid bone in reptiles and primitive amniotes, a bone that is a major component of the palate (Figs. 1, 2A). With the striking consistency of pterygoid defects in the various Dlx and Endothelin mutants, it is possible that these genes have a hitherto unexplored role in pterygoid patterning, something that should be examined in amniotes that have a discrete pterygoid bone.
In addition to endothelins, other signals that act during the early mesenchymal condensation stage are retinoids and BMPs. We have shown that a different type of homeotic transformation can be induced in the avian face by increasing retinoid concentration and blocking BMP signaling (Lee et al., 2001). In this case, the maxillary condensation was respecified to form a frontonasal mass prominence resulting in an ectopic set of frontonasal mass elements replacing the maxillary derived skeleton. The other interesting point is that cartilage and bone were produced in the ectopic beaks. These results indicate that early condensations are not restricted to one skeletal fate nor is identity firmly established.
Later Determinants of Bone Morphology
Once an osteogenic condensation reaches a required stage of maturity, the cells begin to differentiate into osteoblasts, then to deposit bone-specific organic matrix, followed by mineralization. These areas are called ossification centers and are distinguished from the preosseous condensations described earlier (Fig. 5D). It is possible that an additional level of morphological control is exerted at the ossification center stage.
One practical benefit to having multiple ossification centers may be to expand the opportunities to vary morphology. In reptiles, there are many instances of intrabony joints that do not exist in mammals or birds and these joints correlate in some instances with the number of ossification centers. The maxilla of the colubrid snake Elaphe obsoleta forms from two separate ossification centers (Haluska and Alberch, 1983), and ultimately, a hinged joint forms between the two sections. In contrast, the maxillary bones of other venomous snakes such as vipers, rattlesnakes, and the monocled cobra (Naja kaouthia; Jackson, 2002) have but a single ossification center and a one-piece maxillary bone. Unfortunately, the data on how many ossification centers exist for the maxillary bone of other amniotes is not as clear. This uncertainty is due to difficulties obtaining material at appropriate stages and in obtaining a three-dimensional picture of such an irregular structure as an ossification centre. It is worthwhile revisiting this question with some of the more accessible experimental models to determine accurately whether early-mineralized regions are truly separate or are connected.
Finally, we need to consider that preosseous condensations for a single bone may derive from different facial prominences in different amniotes. For example, it is possible that, in Aves, the palatine bone originates partly from the frontonasal mass and partly from the maxillary prominence, whereas in mammals, the origins may be entirely from the maxillary prominence. It is necessary to carry out more fate mapping studies of these subepithelial osseous condensations to more clearly determine their contributions to specific bones.
Epigenetic and Intrinsic Influences on Bone Patterning
At various times during osteogenesis, bone shape is determined by multiple and interacting factors that include the outgrowth of other supporting tissues, and, in the case of the jaws, the forces exerted by masticatory muscles (Herring, 1993). This is best demonstrated experimentally, when pieces of the facial prominences are grown in isolation from the rest of the head, freeing them from influences of sensory capsule growth, brain growth, and muscle function. The parts of the face that contain the anlagen for a cartilage element such as the frontonasal mass and mandibular prominences develop good morphology in the ectopic graft sites. Mandibular bones surrounding Meckel's cartilage or premaxillary bones surrounding prenasal cartilage are well formed (Richman and Tickle, 1989, 1992; Mina et al., 2002; MacDonald et al., 2004). The maxillary prominence is the only example where membranous bone formation occurs without the influence of a cartilage model. In these grafts, partially recognizable palatine, maxillary, or jugal bones form (Lee et al., 2004). Some processes are clearly able to form even in the absence of normal articulations (e.g., the maxillary process of the palatine bone). Although a degree of intrinsic maxillary bone patterning must already be present in the preosseous condensations, there remains a significant component that relies on local influences of the adjacent tissues.
PERSPECTIVES AND CONCLUSIONS
An Earlier Start to Bone Morphogenesis—Being at the Right Place and at the Right Time
The model proposed by Hall and colleagues (Atchley and Hall, 1991; Hall and Miyake, 2000) in which osseous condensations in the mandible are the morphological units that respond to selective pressures is a useful way to think about the evolution of the upper jaw. Our addition to this model is the presence of an earlier type of mesenchymal condensation that forms just at the end of neural crest cell migration (Fig. 5A). These condensations are not yet fated to become skeletal tissue but are destined for specific facial prominences. Regional signaling begins to act on these groups of cells providing global patterning information. We and others, have shown that perturbing the signalling or the downstream transcription factors that act at these early stages leads to transformation of facial prominence identity (Lee et al., 2001; Beverdam et al., 2002; Depew et al., 2002; Ozeki et al., 2004; Ruest et al., 2004). Therefore, the mesenchymal condensations appear to be relatively plastic at this time of development.
Early mesenchymal condensations become incorporated into facial prominences as they form, by which time mesenchymal patterning is fixed (Wedden, 1987; Richman and Tickle, 1989, 1992). It is at these stages that fine-tuning of bone size and shape is taking place. The relative size and position of epithelial signals in facial prominences differs among amniotes (Fig. 5B), and these expression differences may correlate with condensation size. The size of the initial condensation and the relative timing of osteogenic center formation interact to produce bone morphology. For example, a condensation that makes an earlier start on the ossification process may result in a smaller bone by completing the developmental program sooner (Fig. 5 Ci). The corollary is that a condensation that starts at a similar time but remains in an undifferentiated, proliferative state longer would form a bigger mass of cells, which in turn would then give rise to a bigger bone (or several bones; Fig. 5 Cii). It is possible that additional ossification centers could give rise to additional processes and through this mechanism alter morphology. These hypotheses can be tested once the signals that initiate osseous condensations are worked out. In this way, it should be possible to see whether ectopic condensations merge to form alternate bone morphologies.
Because many skeletal defects occur despite normal neural crest cell migration and normal facial prominence morphology, investigations of condensation stages merit high priority. It is important to consider the possibility that the experimental manipulation have direct rather than secondary effects. Because many of our experiments (Lee et al., 2001; Ashique et al., 2002) and those of others (Barlow and Francis-West, 1997) begin around the time of osseous condensation formation but well before the onset of mineralization, they demonstrate the feasibility of targeting early intramembranous bone condensations in the avian embryo. In addition to work with oviparous amniote models, so much more can be learned from gene-targeted mice. Thus far, the tendency of the developmental biology community is to only characterize the early gene expression changes and subsequent skeletal defects in the various knockouts. The result is that valuable experimental material is not being collected or analyzed. We would like to encourage our colleagues to study embryos at intermediate stages and at the same time, to develop more markers for the early mesenchymal and preosseous condensations. It should be possible to develop models in which molecular determinants are altered in a controlled manner (through local or genetic manipulations) so that morphogenetic changes in the skeleton are induced in a predictable manner. Such studies could ultimately induce reptilian or mammalian features in avian embryos and vice versa. In this manner, we can integrate molecular information with skeletal patterning and determine which aspects of skeletal ontology have been conserved during evolution.
Finally, to fully understand evolution of the craniofacial complex, reptilian models should be studied. Descriptions of development can now go beyond the morphology studies of the past to include gene expression analyses. The mere isolation of developmentally relevant coding sequences from reptiles will provide a much-needed missing link. Oviparous species such as alligator and turtle offer opportunities for in ovo fate mapping, setting up organ cultures (inter or homospecific; Ferguson et al., 1984; Nagashima et al., 2005) or for performing experimental manipulations in ovo. Interspecific grafts would allow one to test whether conservation of tissue interactions and gene function exists between the major amniote groups. By carrying out more studies on non-avian reptiles, we can better understand amniote evolution, especially because reptiles are the most closely related group to basal amniotes. Moreover, by studying the development of the reptilian embryo at the molecular level, we will have a great opportunity to unlock the puzzle of how differences in gene regulation resulted in the jaw being redesigned so many different ways during the course of evolution.
We thank K. Fu and V.M. Diewert for providing technical support for the python embryos. Tortoise, chameleon, and python wet specimens were provided by Z. Knotek (UFVS, Czech Republic); python embryos were provided by The Reptile Refuge, Surrey, BC; mouse and human whole-mount stained skulls were provided by T.M. Underhill and V.M. Diewert, UBC, Canada. The authors thank R. Cerny, D. Clouthier, T. Williams, V.M. Diewert, S. Kuratani, and R. Reisz for stimulating discussions. J.M.R. is funded by grants from the Canadian Institutes of Health and is a Michael Smith Distinguished Scholar.