The development of the gnathostome (jawed vertebrate) skull is characterized by several universal traits. First, regardless of the particulars of adult (or reproductive) phenotype, a basic seminal structural bauplan (unitary plan) is initially ontogenetically followed (see Reichert, 1837; Parker, 1866, 1869, 1871, 1873, 1876– 79, 1881– 83, 1885a, b; Huxley, 1869; Gegenbaur, 1888; Wiedersheim and Parker, 1897; Gaupp, 1898, 1899; Howes and Swinnerton, 1901; Gregory, 1904, 1913, 1933; Reynolds, 1913; Kindred, 1921; Wilder, 1923; Kingsley, 1925; Kingsbury, 1926; de Beer, 1937; Paterson, 1939; Nelsen, 1953; Romer, 1956, 1966; Jollie, 1957, 1962, 1977; Goodrich, 1958; Romanoff, 1960; Young, 1962; Schmalhausen, 1968; Allin, 1975; Presley and Steel, 1976; Crompton and Parker, 1978; Barghusen and Hopson, 1979; Bellairs and Kamal, 1981; Moore, 1981; Kuhn and Zeller, 1987; Zeller, 1987; Radinsky, 1987; Carroll, 1988; Langille and Hall, 1989; Vorster, 1989; Allin and Hopson, 1992; Couly et al., 1993; Novacek, 1993; Schultze, 1993; Trueb, 1993; Zusi, 1993; Kimmel et al., 1995; Cubbage and Mabee, 1996; Kuratani et al., 1997; Kuratani, 2003a, b, 2004, 2005; Depew et al., 2002a, b, 2005; Shigetani et al., 2005). One recognizes, for example, the Meckel's cartilage (MC) or dentary of a human, a mouse, or a macaw as such, as members of the MC and dentary homeotypes (i.e., as a members of these classes—MCs and dentaries—of elements) and does not confuse them with frontals, palatines, or any other elements. Second, to meet the particular demands (trophic, reproductive, cognitive, locomotory, and so on) of a taxon, varied ontogentic elaborations of this bauplan have manifested the great diversity in size, shape, and articulations of skeletodontal elements and overall forms observed. Again, looking at the lower jaws of a man, a mouse, and a macaw, it is clear that each is specialized for the particular demands of its possessor.
The assertion of these traits is the result of the application of the concept of homology onto a comparison of cephalic structural units as recognized by the embryologists and anatomists who have studied vertebrate skulls. Originally defined by Richard Owen as “the same organ in different animals under every variation of form and function”, homology as a concept underlies all of comparative biology (Hall, 1994). Yet, for a notion that is so central to biology, a healthy literature re-defining and re-nuancing this notion continues to accumulate. Although this is not the place for a protracted history of the notion of “homology” or of the related concept “homoplasy” (or the occurrence of similarity in structure not due to common descent), it is expedient for a discussion of comparative craniofacial developmental studies to here briefly touch on a few pertinent ideas (for expended discussions of homology, homoplasy, and so on, see de Beer, 1938, 1971; Hall, 1994, 2003; Wake, 1996; Gould, 1977, 2002; and references therein).
First, the concept of homology has evolved from a typological idea tied to structure into a concept encompassing other levels of organization within and between organisms, including the developmental processes and mechanisms used to achieve structural organization. Hall (2003) defines homology, for example, as the “continuous occurrence of the same feature (gene, structure or behavior) in two organisms whose common ancestor also possessed the feature.” Second, a “feature” or “character state” is any trait of the phenotype; thus, one's recognition of a “feature” dictates in large part the invocation of homology. Together, these mean that character states and, thus homology, can be approached hierarchically. For some investigators, this means that homology at one level of organization need not necessitate homology at another. Because ontogeny is a temporal process, developmental processes at one stage necessary for the elaboration of a structural foundation or scaffold may be conserved and homologous, whereas those sequent are not. For instance, to use an example from Hall (2003), whereas the autopodal digits shared by two organisms and their common ancestor may be homologous as structures, and the process of generating the mesoderm in the limb bud may be similar, the developmental mechanisms used to individuate digits (e.g., interdigital apoptosis versus differential growth of the digit primordia) may differ. Such mechanisms are genetically regulated processes; hence, genotypes and phenotypes may dissociate. Third, developmental mechanisms may be conserved (deep homology), despite the eventual absence of associated structure. Developmental mechanisms may be canalized (or modular) and be co-opted or re-induced to act and be seen as resulting in “homoplastic” structures due to a temporal break in the appearance of the feature. (As argued by Hall [ 2003], this would in effect make such homoplasy a subset of homology). Lastly, phenotypic alteration due to change in developmental processes have traditionally been attributed to heterochronic (change in the timing of an event) or heterotopic (change in the topology of an event) shifts, although change due to heterofacience (change in the active capacetence, or the elaboration of capacity, of a developmental event) is also recognized. Two terms introduced by Arthur (2000), heterotypy (change in the type of an event) and heterometry (change in amount), are forms of heterofacience. A change in the affinity of a morphogen receptor or the level of transcription and translation of a morphogen inhibitor leading to differential induction and response are examples of heterofacience, as would be the isochronic, isotopic utilization of related but not directly orthologous genes during a developmental process. Changes in cis-regulatory elements can conceivably lead to any of the above (although it cannot explain all possible changes, such as with a modification in affinity).
Just as homology is an old concept amended for modern purposes, so are many of the questions regarding the development of the skull, and it is with due deference to Geoffroy-St. Hilaire, Cuvier, Owen, Lankester et al., that we still may ask: How are the above two traits—fidelity of initial bauplan and elaboration of final design—maintained and coordinated to generate the amazing diversity in cranial morphologies (both extant and extinct)? What establishes and maintains pattern in the skull? Are there universal developmental mechanisms underlying gnathostome autapomorphic (derived) structural traits (e.g., the gnathostome cranial bauplan)? Can we detect and identify the etiologies of heterotopic, heterochronic, and heterofacient changes in craniofacial development within and between taxa? What are the developmental mechanisms and pathways underlying homoplastic traits?
Seeking an understanding of the development and fundamental nature and organization of the skull has been a key endeavor for developmental and evolutionary biologists for well over 150 years. Although the conceptual frameworks for approaching such understanding have themselves evolved over time, one point has remained clear: jaws are fundamental, functional cranial units that unite the organization of the splanchnocranium (the branchial arch [BA] -derived cranial skeleton) with that of the neurocranium (the cranial skeleton supporting and protecting the central nervous system [CNS] and primary sensory organs). The relative diversity of jawed vertebrates attests to the expansive potentiation the acquisition of jaws effected during vertebrate evolution. Significantly, the capacity, engendered by the presence of jaws, to take advantage of new, varied trophic enterprises had to be coordinated with the capacity to modify any system in place to detect such sources, move to the sources, ingest, masticate, and swallow the trophic factor and to integrate each step of the process (Gans and Northcutt, 1983). In discerning the comparative developmental paradigms of jaws, then, we may hope to elucidate an example of the coordination of bauplan fidelity and design elaboration. Using a model of jaw development as a platform, we explore some of these questions and other issues in modern neontological (comparative embryological and anatomical) studies of the vertebrate skull. Although homologous structures need not of necessity be generated by homologous developmental mechanisms, the essence of modern neontology is the discovery of the presence or absence of correlations between such mechanisms and morphologic outcome. Thus, we begin the present discussion of the development of jaws with a look at the morphologic outcome of the gnathostome bauplan.
JAWS: BAUPLAN AND STRUCTURE
The realization of a bauplan is, in large part, the realization of structure, which is understood through the recognition of homology. Structurally, gnathostomes can be viewed as ontogenetically possessing several skulls: typically, an embryonic/larval, a perinatal/juvenile, and an adult/reproductive (e.g., Gregory, 1933; de Beer, 1937; Goodrich, 1958; Romer, 1966; Schmalhausen, 1968; Moore, 1981; Zeller, 1987; Presley, 1993). The gnathostome bauplan is initially revealed with the appearance of the chondrocranium (the cartilaginous, or endocranial elements; Figs. 1, 2). Two chondrocranial components are distinguished: the neurocranium and the splanchnocranium (also viscerocranium or branchiocranium). The neurocranium can be subdivided into the posterior parachordal (paired struts underlying the caudal CNS forming from mesoderm lateral to, and induced to develop by, the notochord) and occipital (somitic) elements, the anterior trabecular (the paired, cranial neural crest (CNC) -derived struts developing anterior to the parachordals and underlying the CNS rostral to the hypophysis), and sensory capsular divisions (the cartilaginous capsules enveloping the olfactory, optic, and otic sensory organs). The splanchnocranium is formed of those elements derived from the BAs. Embryonic chondrocranial elements have numerous possible ontogenetic fates: endochondral ossification, direct investment by dermal bone, degeneration (resorption), quiescence, or transdifferentiation (de Beer, 1937; Goodrich, 1958; Barghusen and Hopson, 1979; Depew et al., 2002b). The manifestation of the chondrocranium clearly is a conserved gnathostome feature.
A second phase of skull ontogeny generally arises with the advent of the nascent dentition and the dermatocranium, which is composed of those intramembranous (dermal) elements that surround and develop in association with the chondrocranium and dorsal CNS. These bones are typically classified topologically (Fig. 1G,H). Osteichthyeans are characterized by large numbers of dermal bones, whereas tetrapods are characterized by large-scale reductions in number (known as “Williston's Law”; Gregory, 1933) and chondrychtheans do not possess any as such. The skull of the adult develops with the refined modeling and remodeling of the cranial elements to fit endpoint demands.
Primary jaws develop within the context of this bauplan: they are articulated, prehensile oral apparatuses principally derived from the splanchnocranial, dermatocranial, and associated dental elements of the anterior most BA (mandibular, or BA1) usually with a small yet significant contribution from the olfactory placode-associated frontonasal prominences (FNPs; Fig. 3). Secondary, pharyngeal jaw systems, usually using skeletodontal structures derived from caudal BA in apposition to elements underlying the neurocranial base and palate, also have evolved sporadically. Such systems, however, are found in addition to the primary jaws as defined above. Although variously modified, expanded, or reduced, by definition all gnathostomes possess jaws.
With respect to the functional architecture of jaws, several universal traits are apparent (see Reichert, 1837; Parker, 1866, 1869, 1871, 1873, 1876; Parker, 1878–79, 1881–83, 1885a, b; Huxley, 1869; Gegenbaur, 1888; Gaupp, 1898, 1899; Wiedersheim and Parker, 1897; Howes and Swinnerton, 1901; Gregory, 1904, 1913, 1933; Reynolds, 1913; Fawcett, 1917, 1922; Kindred, 1921; Wilder, 1923; Kingsley, 1925; Kingsbury, 1926; de Beer, 1937; Paterson, 1939; Nelsen, 1953; Romer, 1956, 1966; Jollie, 1957, 1962, 1977; Goodrich, 1958; Romanoff, 1960; Young, 1962; Schmalhausen, 1968; Allin, 1975; Presley and Steel, 1976; Crompton and Parker, 1978; Barghusen and Hopson, 1979; Bellairs and Kamal, 1981; Moore, 1981; Kuhn and Zeller, 1987; Zeller, 1987; Radinsky, 1987; Carroll, 1988; Langille and Hall, 1989; Vorster, 1989; Allin and Hopson, 1992; Couly et al., 1993; Novacek, 1993; Schultze, 1993; Trueb, 1993; Zusi, 1993; Kimmel et al., 1995; Cubbage and Mabee, 1996; Kuratani et al., 1997; Depew et al., 2002a, b, 2005; Shigatani et al., 2005). First, there are, apodictically, two appositional units that are articulated (Fig. 4A). That is, a point of articulation (a joint or hinge) between the jaws exists that is placed such that it allows relative motion of the two jaw units. Polarity is, therefore, an inherent character state of jaws: there is “hinge,” and there is “other.” Polarity additionally exists within each “other:” minimally there is “other closest-to-hinge” and “other furthest-from-hinge.” A consequence of this polarity is the potential for the modularity of elements within a particular jaw unit, one manifestation of which is the appearance of differential fates within initially singular elements. This finding is clear when looking at either splanchnocranial element (MC or palatoquadrate, PC) of the jaws. Taking a mammalian example, the murine MC appears as a single element early in ontology yet has four distinct, genetically regulated fates along its proximodistal axis (Depew et al., 2002b; Fig. 4B). Morphologic modularity, moreover, exists across all gnathostome phylogenies (Wiedersheim and Parker, 1897; Reynolds, 1913; Wilder, 1923; Gregory, 1933; de Beer, 1937; Romer, 1956, 1966; Jollie, 1957, 1962, 1977; Goodrich, 1958; Romanoff, 1960; Young, 1962; Schmalhausen, 1968; Barghusen and Hopson, 1979; Bellairs and Kamal, 1981; Moore, 1981). Indeed, various taxa, such as Ceratopsian dinosaurs, are identified and classified by the manifestations of modules as discrete neomorphic elements (e.g., “pre-dentaries” distal to dentary and “rostrals” distal to premaxillae) at the endpoints (“other-furthest-from-hinge”) of the their jaws (Dodson and Currie, 1990; Fig. 4D). Similar manifestations are known in a wide range of taxa, particularly teleostean and amphibian lineages.
A salient question regards the nature of the mechanisms that individuate modules along the jaw axes. Several mechanisms thought to achieve this aim have been suggested, including some seeking to find meaningful boundaries in the patterns of nested BA genes such as the Dlx genes (see below; Depew et al., 2005). This issue is also at the heart of a recent study by Svensson and Hass (2005) that addresses the appearance of “infrarostral” and “suprarostral” elements at the distal midlines (again, “other-furthest-from-hinge” of above) of the developing anuran tadpole jaws. In seeking a plausible etiology, they have reasonably proposed that the development of a joint between MC and the infrarostral may be correlated to the heterotopic elaboration of a developmental mechanism for specifying joint position involving the vertebrate orthologues of the Drosophila bagpipe gene that have been correlated with the development of joints in mice and chicks (e.g., Wilson and Tucker, 2004). Such is also the nature of ongoing work in our own laboratory, in collaboration with that of Abigail Tucker, that seeks to use snakes as models for the manifestation of modules in the form of modified cranial articulations. As stated above, most gnathostome jaws functionally incorporate derivatives of the frontonasal prominences, such as the premaxillae and associated dentition. These elements, functionally and topographically forming “other-furthest-from-hinge” within the upper jaw, must be kept in functional registration with the remainder of the upper jaw as well as the lower jaw. Modularity is inherently clear in the dermatocranial and dental elements of the jaws; thus with respect to the hinge, jaws have polarity and are modular in their construction.
Second, positionally relative to the rest of the organism, the two jaws are distinct: the upper jaw is placed in more intimate association with the neurocranium than the lower (Fig. 4C). Importantly, several strategies for suspending and bracing the jaws against the neurocranium have evolved, including autostylic, holostylic, amphistylic, and dermatostylic. However, regardless of the nature of the suspensorium, or the extent to which they connect to the neurocranium, the PQ cartilage derivatives (and associated dermal ossifications) of the upper jaw are always collectively closer to, and substantially more intimate with, the neurocranium than the lower jaw MC and dermal bones. This further generates polarity within the jaws.
Third, again regardless of the nature of the upper jaw's connection to the neurocranium or the jaw's active (primary articular) players, the articulation of the jaws is found in the same relative developmental position. As defined by the elements directly involved, gnathostomes possess one or both of two types of jaw articulation. Pleisiomorphic for gnathostomes is the primary jaw articulation that occurs between the quadrate portion of the palatoquadrate splanchnocranial element (homologous to the mammalian incus) and the articular portion of the splanchnocranial Meckel's cartilage (homologous to the mammalian malleus); this is a universal articulation (Figs. 2, 4C). In all gnathostomes but mammals and some mammal-like reptiles, this is also the functional joint of the jaws acting as a hinge between the upper and the lower. In mammals (and some of their lineage antecedents), a secondary (although functionally primary) jaw articulation occurs in addition; this is between the dermal dentary and the dermal squamosal. Most important, however, is the location of this mammalian neomorphic jaw articulation: the dentary develops adjacent to Meckel's cartilage, whereas the squamosal develops in close association with the palatoquadrate derivatives. The actual mammalian hinge, then, is developmentally topologically juxtaposed to the primary malleal–incal (articulo–quadrate) articulation.
JAWS: BAUPLAN AND DEVELOPMENT
Whereas the above structural traits are autapomorphic (and homologous as traits) for gnathostomes, can we say the same for their generative developmental mechanisms? Are all primary jaws built in the same way? Clearly, the generation and placement of the joint between the upper and lower jaws is critical to their function, the above traits, and the survival of an organism. As the jaw joint develops from structures that derive from within the primordia of the first branchial arch, regulated development of BA1 is key to understanding its patterning.
BAs are transient, iterative structures in the embryonic vertebrate head arising from the ventrolateral surfaces (Fig. 3). Notably, they are delimited by points of apposition between the ectoderm and endoderm and are filled in all gnathostomes studied, in stereotypic manner, with CNCs surrounding a core of cephalic mesoderm (Hall, 1999; Kimmel et al., 2001). In gnathostomes, BA1 (commonly, the “mandibular” arch) has two principal subdivisions, the maxillary (mxBA1, proximal) and mandibular (mdBA1, distal) branches, that contribute to the upper and lower jaws, respectively. The BA1 splanchnocranium has two major components, the derivatives of the PQ (mxBA1 derivative) and MC (mdBA1derivative; Fig. 4C). These elements, furthermore, are associated with an ordered series of dermatocranial bones (Fig. 1E–H). Together, the chondrocranial splanchnocranium and its associated dermal skeleton form the BA1 portion of gnathostome jaws.
Recent interest has arisen as to the disposition of the maxillary branch of BA1, prompting us to attempt a brief clarification of what has traditionally constituted “maxillary.” The maxillary branch garnered its appellation from the recognition of several understood facts: (1) that jaws articulate, the lower jaw being referred to, collectively, as the mandibular jaw; (2) that all gnathostomes posses within their developing lower jaws an MC, the proximal end of which forms an articulation with the quadrate (including its incal homologue) portion of the PQ cartilage; (3) that it is pleisiomorphic in the Gnathostomata for the quadrate portion of the PQ to directly participate in the functional primary jaw articulation, and, although its exaptation into the middle ear as an incus is autapomorphic for mammals, its fundamental topologic relations have not changed (e.g., it still articulates with the end of MC); (4) that the PQ derivatives are more intimately associated with the neurocranium than the Meckelian cartilages, contacting (frequently synchondrotically) the neurocranium in numerous locales, including the otic capsule, the trabeculae cranii, the optic capsule, and the nasal capsule; (5) that the PQ and its derivatives may form but one component of the upper jaws and that there are a variable number of nonmandibular, nonfrontonasal-derived dermatocranial contributions to the upper jaws, including many intimately associated with the PQ derivatives and the floor of the functioning palate (many of which contain functioning dental elements); (6) that the olfactory-associated frontonasal primordia contribute the premaxillae to the upper jaw arcade in all but the chondrychthyans; (7) that the position of the external nasal openings and, in tetrapods, the nasolacrimal groove and primary choanae delineate the boundary between the premaxillary component from the remainder of the upper jaw apparatus; (8) that all evidence indicates the BA1 origins of the actual jaw articulations; and (9) that the upper jaw of humans is dominated by bones known as “maxillae.” These, together, lead to the classic understanding and choice of the term “maxillary” as referring to those developing primordia that give rise to those jaw-associated elements delimited by the articular of the mandibular jaw on one side and the nasal-associated structures on the other.
The context and applicability of the use of the term “maxillary” has been called into question in two recent papers (Cerny et al., 2004; Lee et al., 2004). The data presented, however, are fully consistent with this classic understanding. Both of these studies are fate mapping experiments in which the authors have labeled cells within a region that lies between the distal tip of mdBA1 and the developing frontonasal structures. Cerny et al. used in situ hybridization to detect transcripts of the snail gene (thought to label CNC cells) in the axolotl, Amblytoma mexicanum. At stage 34, they found snail-positive cells throughout the presumptive BA1 and surrounding the posterior eye primordia, whereas at stage 35/36 snail expression was found in two discrete positions: one subjacent to the eye, which they labeled “maxillary,” and one in the ventral half of BA1, which they labeled “mandibular,” with an unexplained gap in between. They conducted two further types of experiments: fluorescein isothiocyanate (FITC) -labeled or green fluorescent protein (GFP) -labeled neural folds were transplanted to unlabelled hosts or 1,1′, di-octadecyl-3,3,3′,3′,-tetramethylindo-carbocyanine perchlorate (DiI) was applied to the presumptive mandibular arch and the fate of the cells followed. From these experiments, they found that both the PQ and MC were labeled by cells originating within mdBA1, with cells from the ventral-most (distal) mdBA1 contributing to MC and that cells proximal to these in the medial portion of BA1 contributed to the PQ. This finding is in line with classical understanding as set out above. Cerny et al. concluded, however, that this was unexpected, as it appears to contradict the previously held assumption that the upper jaw (PQ) cartilage arises from the “maxillary” condensation of the first arch in all vertebrates. This, however, is not the “classical assumption”; rather, the classical understanding is that the maxillary branch of the first arch gives rise to the PQ cartilage. What has defined the maxillary arch is its position between the premaxillary (nasal) primordia and the mandibular primordia and not whether cells that contribute to the palatoquadrate are found throughout this topography or where and when cell concentrations are densest. Indeed, although the one constant in the development of the gnathostome PQ has been the presence of the quadrate portion that articulates with the end of MC, the presence and continuity of the PQ processes variably extended toward the neurocranium are idiosyncratic to specific taxa. The relative size of the PQ does not, in itself, define the maxillary region. (Strictly speaking, the classical understanding is, furthermore, independent of the presence of the naso-optically orientated proximal BA1 and bent shape of BA1 typical to gnathostome.) What would have been a surprise would have been finding that labeled PQ cells were ventrosdistal to those of MC or that the quadrate portion of the axolotl PQ did not arise adjacent to the developing MC (although even this would not of necessity lead to the repudiation of the use of the term “maxillary” in the traditional sense).
Cerny et al. further wanted to understand the fate of their snail-positive “maxillary condensation” and so labeled cells subjacent to the eye. Although having found no labeled cells in the PQ, they indicate that they found labeled cells next to the eye in what they have taken as the trabeculae cranii. Whereas they do not present data on the labeling of those cells in between their “maxillary” and “mandibular” positions, finding labeled cells in cartilage next to the eye is not terribly surprising: the trabeculae of ambystomatid salamanders are widely separated from the midline (leaving an extensive basicranial fontenelle eventually overlain by the parasphenoid, the origin of which is of interest in its own right), laterally displaced toward the eye, and continuous (synchondrotic) with the orbital (optic) cartilages (de Beer, 1937; Trueb, 1993). Moreover, at some point during its ontogeny the ambystomatid PQ extends an ascending process to fuse with the conjoined trabeculae–orbital cartilages; and, as with salamanders in general, ambystomatids have an open temporal region, lack a check region, and posses a reduced upper jaw arcade (although they posses both a maxillae and a premaxillae; Bonebrake and Brandon, 1971; de Beer, 1937; Trueb, 1993). Whereas distinguishing the relative contributions of synchondrotic structures is problematic at best, labeling cells around the developing eye of an ambystomatid and subsequently finding labeled cells associated with the trabeculae is not. While Cerny and colleagues do not present data on the positional origins of the dermatocranial jaw elements, they do label stage 13–14 chick embryos for comparison. Cerny et al. concluded that the chick develops in an analogous manner to Ambystoma: e.g., the PQ develops from cells found proximal to those contributing to MC in the ventral BA1. In all, although some evidence of cell mixing is presented, no evidence is presented that would call for a countermand to the traditional usage of “maxillary.”
Lee et al. (2004) labeled the developing BA1 and FNP of stage 13, 15, and 17 chick embryos with DiI and/or DiO and examined the position of the labeled cells after 48 hours. They demonstrated that both “cranial” (juxtaposed to the developing nasal apparatus) and “caudal” (adjacent to the mandibular branch) maxillary prominence cells were found between the nasolacrimal groove and the position of those cells contributing to MC (although a restricted swath of cells subjacent to the eye were found toward the midline of the oral cavity). Moreover, they show that cells from the presumptive FNP are not later found within the maxillary prominence. Again, this finding is coincident with a traditional understanding of what is maxillary. Lee et al. make the astute observation that the maxillary prominence makes a major contribution to the upper jaws of amniotes, and then pose the question of whether more “primitive” gnathostomes also posses this prominence. Despite the direct statements of the authors (e.g., “The mandibular arch has distinct maxillary and mandibular processes, and the future mouth is visible as a “v”-shaped notch.”, p 63) and the explicit evidence in the micrographs, Lee et al. cite a scanning electron study of the actinopterygian fish, Polyodon spathula, by Bemis and Grande (1992) as suggesting that fish do not reveal a separate maxillary bud projecting from the side of the oral cavity. Whereas a correlation between a robust maxillary prominence and a robust upper jaw may turn out to be valid for some taxa, it is not because it is an amniote autapomorphy.
We believe that they also err in their assessments of the fundamental lessons about jaw development that the Hoxa2−/− and Dlx5/6−/− mouse mutants teach. Lee et al. indicate that only mandibular elements are duplicated in the Hoxa2−/− mutants, which could be argued at some level if one believed that the incus (PQ) and the squamosal—both clearly duplicated in part or in whole in the mutants—were also mandibular structures. The squamosal in mice, however, bears the responsibility of acting as the upper jaw articulation and, hence, cannot by definition also be lower, or mandibular, jaw. It is true that mandibular structures are duplicated, but clearly only a restricted subset, in particular those that develop in close apposition to the first pharyngeal plate (pouch + cleft). Moreover, they appear as mirror-image duplications, strongly suggestive of a point of positional information being interpreted by the CNC around the plate. Lee et al. also suggest that the homeotic transformation seen in the Dlx5/6−/− mutants, rather than having been a duplication of the maxillary part of the first arch at the expense of the mandibular first arch (as interpreted by the authors of the knockouts; Beverdam et al., 2002, and Depew et al., 2002a), shows there has been a loss of first pharyngeal arch identity and its replacement by a nonpharyngeal arch maxillary prominence. Regardless of the name one gives to those tissues between the mandibular BA1 and the FNP, or if you believe that the maxillary prominence is not pharyngeal/branchial in nature, this fails to recognize the bias of the transformations as being centered around structures that they apparently do not even consider to be “maxillary,” such as the PQ derivatives (e.g., the incus and ala temporalis) and (ostensibly) the associated squamosal. And whereas we (Depew et al., 2002a) have never claimed, as asserted by Lee et al., that the Dlx5/6−/− mutant phenotype is atavistic, we believe that the essence of the observation made by Beverdam et al. (2002) that the mirror-image duplication seen in the Dlx5/6−/− mutant is reminiscent of the mirror-image seen in various fossils is appropriate: it suggests a conserved aspect of where at least some patterning information for the developing jaws is positioned. Perhaps lost sight of by semantics is that the Dlx5/6−/− mutants demonstrate that the complement of Dlx expression in the cranial neural crest dictates the morphogenic response to the patterning information Dlx-positive cells encounter, and (based on the mirror-image duplication of the structures generated) that one such source of patterning information is likely to be found at the junction of the upper and lower jaw articulations.
Whereas we do not believe that the data presented in either of these papers necessitates a paradigm shift in what has traditionally been understood as, and called, “maxillary” or “mandibular,” we do believe that they serve to highlight what will be a productive approach to understanding the comparative development of the vertebrate skull—the active, focused analysis of the comparative position (heterotopy), timing (heterochrony) and nature (heterofacience) of patterning information and its interpretation between and within taxa. With the exception of enamel, which is derived from oropharyngeal epithelium, lineage tracing, and fate mapping studies in fish, amphibians, birds, and mammals indicate that the skeletal elements of the jaws develop from CNC cells (reviewed by Hall, 1999, and Le Douarin and Kalcheim, 1999). Clearly, CNC subpopulations need not be specified by the same mechanisms, and the relative importance to jaw development of differences in CNC cell induction and behavior needs to be better understood. For instance, do the heterochronic differences in the timing of delamination and migration of cells, relative to neural fold apposition, seen in various gnathostomes have specific importance for the eventual morphologic outcome? As colonizers of the splanchnocranium, rostral neurocranium and dermatocranium, understanding the degree to which cell fate and regional pattern is generated before, during, and/or after the migration of CNC cells, and the degree to which this is conserved, has historically been paramount. In this regard, a gradually building consensus concerning the cumulative experimental evidence in vertebrates suggests that the development of BA1 hard tissues involves a combination of both prepattern and epigenetic mechanisms and the elaboration of developmental programs is consensual at heart. Clearly, CNC cells are bringing with them both factors crucial for their interpretation of their environment and the malleability (plasticity) to interpret the distinct, specific focal inductive influences in their environment; apparently, so too can the epithelia be characterized (e.g., Couly et al., 2002; Depew et al, 2002a, b; 2005; Trainor et al., 2002; Santagati and Rijli, 2003; Schneider and Helms, 2003; Graham et al., 2004; Tucker and Lumsden, 2004).
HINGE AND CAPS MODEL
To address whether jaws are all made in a like manner (and if not then how not), one needs a starting point for the sake of comparison. Arguments for prepattern or epigenesis in CNC elaboration of jaw form both necessitate proximate sources of information: this information might come, for example, in concert with, or secondary to, specification of the CNC, neurectoderm, surface cephalic ectoderm, endoderm, and/or mesoderm (Depew et al., 2002b). To (1) make sense of these possibilities and (2) form a foundation for which a comparison between and within taxa can eventually be framed, a hypothesis for jaw development is useful. To this end, we present here a “hinge and caps” model that places the articulation, and subsequently the polarity and modularity, of the upper and lower murine jaws in the context of CNC competence to respond to positionally located epithelial signals (Depew et al., 1999, 2002a, b, 2005; Trumpp et al., 1999; Shigetani et al., 2000, 2002).
This model expands on an evolving model of polarity within the mandibular arch (e.g., Sharpe, 1995; Neubuser et al., 1997; Thomas et al., 1998, 2002; Tucker et al., 1998, 1999; Trumpp et al., 1999) and seeks to explain a developmental patterning system that apparently keeps gnathostome jaws in functional registration yet tractable to potential changes in functional demands over evolutionary time. It relies upon a system for the establishment of positional information where pattern and placement of the “hinge” is driven by factors common to the junction of mxBA1 and mdBA1 and of the “caps” (i.e., signals from the primordia of “other furthest-from-hinge” of above) by the signals emanating from the distal-most BA1 midline and the lambdoidal junction (where the mxBA1 meets the olfactory placode-associated FNP). In this particular model, then, the functional registration of jaws is achieved by the integration of “hinge” and “caps” signaling, with the “caps” sharing at some critical level a developmental history that potentiates their own coordination. Whereas the relative level of importance and contributions toward the manifestation of specific structures by these centers of patterning information is by nature idiosyncratic to a species and its ontogenetic functional demands, that they are integrated is patent. Moreover, patterning information contained in the medial epithelia from Rathke's pouch through to the commissural plate and the derivatives of the anterior neural ridge is essential for jaw development; and although the changing relationships of these tissues may have been significant in the transition of agnathans to gnathostomes (see Kuratani, 2003a, b, 2004, 2005), for convenience the model initially places this patterning information in the context of upper cap lambdoidal junctional patterning events. For the purposes of future comparative assessments, the model places an emphasis in jaw development on the relative timing, position, nature, and capacity of the epithelial (both ectodermal and endodermal) patterning centers and the CNC capacity to respond. In short, the emphasis is on the identification and comparison of tissues acting in like inductive and responsive roles apparently necessary to generate pattern and polarity in developing jaws; hence, it does not emphasize particular players, but rather the effects categories of players have on the cellular behaviors necessary to generate functioning jaws.
Evidence of a Hinge-Associated Source of Patterning Information
What is the evidential foundation for such a model? Indirect evidence for a hinge signal comes from gene expression profiles at critical embryonic stages, whereas more direct evidence comes from targeted mutations and other experimental manipulations that result in morphologic transformations around the hinge within BA1. At some level, the hinge can be conceived of as a fulcrum of symmetry: thus, molecular evidence of its existence ought to reflect this symmetry. Indeed, a general symmetrical pattern of gene expression, centered around the mxBA1 and mdBA1 junction that gives rise to the jaw articulation, exists, as exemplified by the expression patterns of a wide range of genes (e.g., as with Dlx1, Dlx2, Barx1, Lhx6, Lhx7, Tbx2, Tbx3, Emx2, Spry, Gbx2, and Twist expression in the mesenchyme and Fgf8 and Pitx expression in the overlying epithelium; Fig. 5B; Depew et al., 2002b). Rather than seeking to exhaustively review all avenues and aspects of the direct evidence, however, we here limit ourselves to factors for which comparative data are most readily available or attainable and so hope to direct an appreciation of the framework of the model for which comparisons between taxa eventually can meaningfully be made.
Transformations around the hinge can be exemplified by loss of function of Dlx gene family members in mice. Dlx genes are expressed in appendages, or outgrowths from the body axis of vertebrates, including the BAs. Six murine Dlx genes have been described: Dlx1, Dlx2, Dlx3, Dlx4 (previously Dlx7), Dlx5, and Dlx6 (Fig. 6A; Stock et al., 1996; Panganiban and Rubenstein, 2002). Based on studies of genomic organization, Dlx genes are arranged as tightly linked, convergently transcribed (tail-to-tail), bi-gene second-order paralogue clusters. DNA sequence similarity (mainly outside of the homeodomain) and chromosomal location indicates that Dlx genes can be placed into two paralogous groups: Dlx1, 6, & 7 and Dlx2, 5, & 3. First-order paralogous Dlx genes share nested expression patterns within the mesenchyme of the BAs: Dlx1/2 are expressed throughout most of the PD axis of the BA, whereas Dlx5/6 and Dlx3/7 share progressively restricted domains distally (Fig. 6B). Importantly, the proximal boundary of the nested Dlx5/6 pair in mice coincides in large part with the junction of mxBA1 (i.e., the upper jaw) and mdBA1 (i.e., the lower jaw). Based on the correlation of this nested pattern of expression and the segmental nature of the skeletal units within a BA, it has been hypothesized that a combinatorial code of these genes controls the proximodistal growth and patterning of the BA units (reviewed in Depew et al., 2005).
This hypothesis was initially genetically examined by the generation of mice bearing nonfunctional alleles of Dlx1 and/or Dlx2 (Qiu et al., 1995, 1997). Two main points came from these initial studies of the Dlx1−/−, Dlx2−/−, and Dlx1/2−/− mice. First, despite being expressed both in mxBA1 and mdBA1 mesenchyme, phenotypic alterations did not appear to extend to the lower jaw structures. Both the chondrocranial and dermatocranial elements representing the upper jaw hinge were significantly altered. The incus lacked a normal articulation with the malleus (loss of the primary jaw joint) and was instead fused to an ectopic cartilaginous strut that ran rostrad while the ala temporalis portion of the PQ was generally lost (Fig. 6D). In the topographic region of the normal squamosal and lamina obturans were four dermal bones, the ventral-most two bearing anterior running zygomatic processes, but neither possessed a glenoid fossa where the condyle of the dentary makes the secondary (primary functional) jaw joint in mammals. Meckel's cartilage, the ectotympanic, gonial, and dentary appeared normal. Although subsequent assessment of the mutant phenotype has revealed weak phenotypic changes in the scale of some proximal mandibular structures (e.g., ectotympanic and gonial), clear distinctions in modification between the articulation components of the upper and lower jaws due to the loss of Dlx1 and Dlx2 are seen. Second, whereas alterations included a transformation of the PQ-derivatives and associated dermal bones (e.g., palatal bones) and teeth, they did not include the entirety of the upper jaw apparatus as those structures developing in close association with the FNPs (including portions of the maxillae that develop around the nasal capsules) were normally developed. Thus, a selective transformation around the upper jaw articulation contribution to the hinge is exhibited in the these mutant mice.
Further transformations about the hinge are seen with the targeted disruption of both the endothelin signaling pathway and its potential targets, the nested Dlx5 and Dlx6 genes (Kurihara et al., 1994; Clouthier et al., 1998; Yanagisawa et al., 1998; Acampora et al., 1999; Depew et al., 1999; Charite et al., 2001; Ruest et al., 2004). Dlx5 is expressed both in early head mesenchyme (mdBA1 CNC) and epithelia (e.g., the early anterior cephalic ectoderm, ANR, and olfactory and otic placodes; Simeone et al., 1994; Qiu et al., 1997; Yang et al., 1998; Depew et al., 1999). Dlx5−/− mutants die shortly after birth, with regional defects in their nasal and otic capsules and proximal mdBA1 structures (Acampora et al., 1999; Depew et al., 1999). MC is shortened and its path back toward the middle ear is disrupted. Near the proximocaudal end of the dentary, the body of MC sharply deviates laterad only to abruptly reorient caudomedially again for a short distance whereafter it splits. A medial branch forms a strut toward the pterygoid, basisphenoid, and ala temporalis, while a lateral branch runs (at the level of the processus folii) to the malleus. By postnatal day (P) 0, this deviated cartilage is invested by ectopic intramembranous bone (Fig. 6E). This ectopic bone also forms a synovial joint with the misshapen gonial, and sutures with the anterior crus of the tympanic. The malleus has a smaller than normal head and is caudally extended and thickened at the level of the manubrium. The tympanic is slightly smaller and thicker. A short and dysmorphic proximal end dentary develops around the abnormal MC (Fig. 6E). The proximal lamina of the coronoid is absent, and the condylar and angular processes are shortened, misshapen, and juxtaposed. Clearly, neither Dlx1 nor Dlx2 are capable of full compensation for a loss of Dlx5 in proximal mdBA1 development. Such drastic changes in morphology are not seen in the non-nasal–associated upper jaw articulation elements. Disruption of the distally expressed Dlx5, thus, is consistent with a Dlx code, and these alterations to jaw development occur on the opposite side of the hinge than those seen with the Dlx1−/−, Dlx2−/−, and Dlx1/2−/− mutants.
A relevant prediction based on the Dlx combinatorial code model is that mdBA1 structures ought to be proximalized with the simultaneous loss of a nested linked-pair such as Dlx5/6 (Depew et al., 2002a). This postulation was also genetically tested in mice (Beverdam et al., 2002; Depew et al., 2002a). Being foreshadowed by the alterations evident in the Dlx5−/− mice, the functional loss of both Dlx5 and Dlx6 leads to a homeotic transformation at and around the jaw joint (Fig. 6F). Within MC, the body and, frequently, the rostral process fail to form. The body is replaced instead by a second ala temporalis (a mxBA1 PQ-derivative). A malleus as such does not form; rather, a cartilage more similar to two fused incudi is in evidence. This trend is continued with the dermal skeleton of mdBA1: for example, the dentary is replaced by a dermal bone with characteristics of a mxBA1-derived maxillae, and a second squamosal develops. Although the upper jaw hinge-associated elements are slightly affected by the modifications of their lower jaw partners and the loss of FNP-associated elements (due to the loss of Dlx5/6 expression in the epithelium of the olfactory region), they are maintained. Transformations are also seen within the soft tissues of the lower jaws, including the presence of ectopic vibrissae and ruggae, two ectodermal structures associated with mammalian upper jaws. Thus, phenotypic transformations above and about the hinge region can be experimentally induced. These morphologic transformations are accompanied by the loss of mdBA1-specific molecular identities (e.g., the loss of dHAND and mesenchymal Pitx1 expression) and the acquisition of maxillary-like identities (e.g., up-regulated Meis2 and Wnt5a expression; Beverdam et al., 2002; Depew et al., 2002a).
The orientation and placement of this homeotic transformation is significant: it renders a mirror-image duplication of the upper and lower jaws at the point of articulation (Fig. 6F). There are numerous means to generate such a mirror image. For example, a signal could emanate from around the point of image convergence (i.e., in the region of the hinge); patterning information from two separate signals, either similar or distinctly dissimilar in nature, could simultaneously emanate from points further from the point of image convergence; or a combination of both may be involved. The most parsimonious model for rendering a mirror image, however, suggests the presence of a signaling source of positional information in the BA1 epithelium associated with the hinge region that is subsequently interpreted in a Dlx-dependent manner by the adjacent CNC. Thus in mice Dlx genes are crucial for the elaboration of the articular, hinge region and supports the notion of a hinge-associated source of patterning information that is interpreted by the CNC. (It should be clear however that the hinge and caps model does not depend on the nesting of the Dlx genes nor do the boundaries generated by such nesting need to correspond to any obvious anatomical landmark.)
That the nonlinked Dlx genes work in concert across the hinge has been demonstrated by the generation of compound Dlx2−/−;5−/− mice (Depew et al., 2005). Dlx2−/−; Dlx5−/− mutants are striking at birth, having small domed heads and, usually, cleft mandibles. The Dlx2−/−; Dlx5−/− mandible is severely truncated, appearing as a small projection beneath the eye. The external auditory pinnae, which usually develop out of, and around, the first pharyngeal cleft between BA1 and BA2, are mostly absent and are represented by the barest of a hillock. Although compound Dlx2−/−; Dlx5−/− mutants have several phenotypes distinct from the single mutants, alterations idiosyncratic to either the Dlx2−/− or the Dlx5−/− single mutants are generally in evidence; this includes the loss of most of the alisphenoid and the presence of altered side-wall dermal ossifications, an ectopic PQ related structure, and lateral projections from the basitrabecular plate rostral to the basisphenoid (as seen in the Dlx2−/− mutants), and the (occasional) exencephaly and otic and (asymmetric) nasal capsular defects (as seen with the Dlx5−/− mutants).
Notably, in regions derived from where both Dlx2 and Dlx5 are extensively co-expressed—that is, the distal BAs—the greatest alterations and transformations have taken place (Fig. 6G). MC, except for the midline rostral process, is essentially absent. Clearly identifiable incudes and mallei are not present. The dentary is represented by only the barest remnant of a molar alveolus, a diminished incisive alveolus, and one or two aberrant incisors. Neither a gonial nor a definitive ectotympanic are ever observed. The greatly truncated dentary develops a closely apposed, abnormal articulation with the truncated molar alveolus of the maxilla; thus, the dentary–squamosal secondary jaw articulation is likewise lost. Hence, both the primary and secondary jaw articulations have disappeared.
Further evidence for a hinge signal emanating from around the above-mentioned point of image convergence (i.e., from the jaw joint) comes from studies in mice of Fgf8, a secreted signaling factor expressed in the epithelium at the mxBA1–mdBA1 junction and the first pharyngeal plate (Fig. 5B). Fgf8 has a dynamic spatiotemporal pattern of expression in several known signaling centers that regulate pattern and morphogenesis, including the primitive streak, the apical ectodermal ridge of the limb bud, the midbrain–hindbrain isthmus and anterior neural ridge (ANR), the intraembryonic coelom (to form the heart), olfactory placodes, pharyngeal plates, and oral ectoderm of BA1 (Heikinheimo et al., 1994; Crossley and Martin, 1995; Trumpp et al., 1999; Crossley et al., 2001). Notably, Fgf8 is first detected in the presumptive mxBA1–mdBA1 ectoderm and the pharyngeal plates before the influx of CNC into the arch (Crossley and Martin, 1995). Ectopic Fgf8 protein, moreover, has been shown to induce and maintain BA ectomesenchymal expression patterns, including those of the Dlx genes (Neubuser et al., 1997; Bei and Maas, 1998; Thomas et al., 1997, 2000; Tucker et al., 1999a, b; Mina et al., 2002).
Loss-of-function analysis of Fgf8 provides a clear demonstration of its involvement in craniofacial and hinge development. Two genetic strategies have been used to circumvent the embryonic death at gastrulation seen in Fgf8−/− mouse mutants (Sun et al., 1999): tissue-specific loss and decreased dosage due to the generation of hypomorphic alleles (Trump et al., 1999; Abu-Issa et al., 2002; Frank et al., 2002; Macatee et al., 2003). We have previously used a Cre/loxP strategy to inactivate Fgf8 within the peri-oral ectoderm of mxBA1 and mdBA1 by E9.0 (Trumpp et al., 1999). Significantly, gene inactivation in the caudal, peri-cleftal mdBA1 ectoderm, however, occurred later than in the rest of BA1 (Trumpp et al., 1999). This strategy resulted in an agnathic mouse (Fgf8Nes-Cre−/−) with a dramatic loss of most of the BA1-derived skeletal elements, in particular elements associated with the jaw articulation and hinge region (Fig. 7A). With regard to the hinge elements, the entire maxillary (incus and ala temporalis) and most of the mandibular splanchnocranium failed to form. Of MC, only the rostral process (with associated dermal alveolar bone and diminutive incisors) and much of the malleus (including a well-formed manubrium and processus brevis) developed. The elongate body of MC did not. Most of the hinge-associated BA1 dermatocranial elements were agenic (e.g., dentary, palatine, pterygoid, lamina obturans, jugal, palatal and molar portions of the maxillae, glenoid cavity of the squamosal) or severely hypoplastic (e.g., nasal capsule-associated parts of the maxilla, ectotympanic, gonial, and nonarticulating portions of the squamosal). In summary, the hinge is lacking in these mutants.
BA1 is clearly hypoplastic by E9.5 in these mutants, but this is not due to a lack of CNC cell migration into BA1 (based on Cad6, CRABPI, and AP2.2 expression patterns in the CNC). Although cell proliferation assays suggested mutant levels were similar to wild-type embryos, extensive BA1 cell death was detected in Fgf8Nes-Cre−/− mice. Loss of Fgf8 lead to disruption of molecular patterning in both the ectoderm and mesenchyme in Fgf8Nes-Cre−/− mice: for example, BA1 ectoderm maintained oral expression of Pitx1 but lost ET-1 (Trumpp et al., 1999). Moreover, within the mandibular “cap” associated distal midline (which will eventually yield the rostral process, incisors, and associated dermal bone) expression of both Pax9 and Bmp4 was maintained, whereas the ectoderm of the molar field associated with the body of MC and much of the dentary lost Pax9 expression. Expression patterns of Msx1 were maintained in the underlying distal midline mesenchyme as were those of Msx2, eHAND, and dHAND. Notably, loss of Msx1 leads to the absence of those mandibular midline structures that persist in the Fgf8Nes-Cre−/− mice—namely, the rostral process and incisors (Satokata and Maas, 1994; Houzelstein et al., 1997). Although at E9.5 the Fgf8-inducible genes Dlx2 and Dlx5 were expressed in BA1, it is not clear that their expression was indeed Fgf8-dependent: Dlx2 is expressed in migrating CNC cells before their exposure to BA1 localized Fgf8 and Dlx2 and Dlx5 are also Bmp—which is expressed at the distal midline—inducible (Neubuser et al., 1997; Bei and Maas, 1998; Thomas et al., 1997, 2000; Tucker et al., 1999a, b; Mina et al., 2002). Moreover, whether Fgf8 is necessary for later (e.g., by E10.5) maintenance of the Dlx genes in the BA was left unclarified in this study.
Those BA1 elements that do form in Fgf8Nes-Cre−/− mice are important. The portions of the maxillae and premaxillae that do arise in these mutants develop in close association with the nasal capsules and frontonasal processes, hence, the upper jaw lambdoidal junction cap region. The malleus and rudimentary ossification in the region of the juxtaposition of the ectotympanic, gonial, and malleal processus folii developed out of a region of BA1 where the gene inactivation in the caudal, peri-cleftal mdBA1 occurred later than the rest of BA1. As focal expression of ET-1, Barx1, and Gsc in this spot transiently remained, the cells contributing to these structures may have been exposed to an elaborated Fgf8 signal at the critical point of their ontogeny. The expression of Bmp and Msx in the mandibular midline suggests the presence of an Fgf8-independent, Bmp/Msx-dependent field at the midline (Trumpp et al., 1999; Depew et al., 2002b).
These studies appear to be corroborated by the phenotype of mutants from a second tissue-specific loss of Fgf8 (Fgf8AP2α-IRESCreI), where expression is lost in nearly all of the early embryonic surface cephalic ectoderm (Macatee et al., 2003). Although not fully characterized, Fgf8AP2α-IRESCreI mutants seemingly lack much of their craniofacial structures (see Fig. 4C, Macatee et al., 2003), including those associated with the Fgf8-positive upper cap, lambdoidal junction region. Notably, whereas a report of a Tbx1-Cre–driven tissue-specific loss of Fgf8 in the endoderm of the pharyngeal pouches suggested that an endodermal source of Fgf8 was essential for the proper development of the smooth muscle of the proximal aorta, it failed to disrupt the aortic arches or grossly evince craniofacial defects (Brown et al., 2004).
Members of the Sprouty (Spry) gene family encode RTK inhibitors, including Fgfr RTKs (Hacohen et al., 1998; Casci et al., 1999; Kramer et al., 1999), and Spry2 has been identified as a regulatory target of FGF8 signaling in BA1 (Minowada et al., 1999; Trumpp et al., 1999). Moreover, analysis of the loss-of-function of multiple members of this family shows that the hinge region is susceptible to gene dosage (Depew, Basson, Rubenstein, and Martin, unpublished observations).
Cumulative data from the Fgf8 signalling system support the idea that, as outgrowths, BA1 receives patterning signals from both the rostral, peri-oral and the caudal, pharyngeal plate-associated epithelia. The pharyngeal plate region is known to express patterning signals (Begbie et al., 1999; Depew et al., 2002b) and is a region where mirror image patterns of gene expression (e.g., Dlx3, Bapx1, Teashirt, etc.) and duplications of some hinge-associated elements are known to occur. Such structural duplications are exemplified by the phenotype of the Hoxa2−/− mutants (Rijli et al., 1993; Gendron-Maquire et al., 1993). Although Hoxa2 is expressed in R2 and more caudal rhombomeres, BA1 CNC is Hoxa2-negative. BA2, in contrast expresses Hoxa2 in both the CNC and ectoderm. Hoxa2−/− mutants exhibit mirror-image, homeotic-like transformations of the BA2-derived structures into proximal, hinge-associated BA1-like structures. The stapes and styloid process fail to form; instead, duplicated proximal MC and associated structures (i.e., tympanic, gonial, malleus, and processes folii) appear. The transformed structures are fused to the normal, BA1-derived structures that they replicate. This is likewise the case for the incus and squamosal, while an ectopic cartilaginous strut runs toward the basisphenoid. Notably, the dentary portions of the hinge were not duplicated. That there is a mirror-image duplication suggests the likelihood that the CNC receive positional information around a point source in the region of the pharyngeal plate.
Thus, a pertinent question arises: How do these two regions—the rostral, peri-oral ectoderm and the caudal, pharyngeal plate—interact to coordinate jaw hinge development? We hypothesize that they act in concert to stabilize (much like a staple would) the positioning of the elements so crucial to the jaw articulation and, therefore, to function. Furthermore, we are testing the postulation that modulation of the strength of the Fgf signaling in these epithlia de-stabilizes the positions of the jaw articulations.
What is required for the initial expression of Fgf8 in the cephalic, peri-oral ectoderm? This is, of course, an essential issue to address. Although not necessary for the initial expression of Fgf8, Pitx2 is required for its maintenance in mice and subsequently for the development of the hinge region (Lin et al., 1999; Lu et al., 1999; Liu et al., 2003). Vertebrate Pitx2 is expressed in the surface ectoderm that will cover the stomodeum, including associated hypophyseal and dental placodes and the peri-oral ectoderm (Mucchielli et al., 1997; St. Amand et al., 1998, 2000; Campione et al., 1999; Lin et al., 1999; Lu et al., 1999; Schweickert et al., 2001a, b; Boorman and Shimeld, 2002; Liu et al., 2003; Jackman et al., 2004). Whereas the skeletal defects have not been extensively reported, murine mutants lacking Pitx2, much like the Fgf8Nes-Cre−/− mice, maintain the distal rostral process of MC but lack most of the body of MC (Liu et al., 2003). By both physical appearance and patterns of gene expression, E10.5 Pitx2−/− embryos appear have the distal midline cap fused to the lambdoidal junction, having lost the peri-oral hinge region (but maintaining some of the pharyngeal plate hinge region). Indeed, Fgf8 expression in the peri-oral ectoderm, present at E9.5, is not maintained in these mutants and is lost by E10.5. Thus, the phenotype may be partially explained by the loss of Fgf8. Roles for Pitx genes in stomodeal and surface ectoderm and head development of other vertebrates have also been demonstrated, as for example in Xenopus where Pitx1 and Pitx2 appear to be necessary for the development of endogenous or ectopic cement glands (Schweickert et al., 2001a, b). Clearly, however, this is not the only influences on the expression of Fgf8 (see below).
In mice, then, modulation of either Dlx or Fgf expression demonstrates their requirement for the manifestation of proper, functioning jaws. The complement of Dlx genes acts to direct the individuation of the upper jaw from the lower jaw at the hinge, whereas Fgf acts to maintain the hinge crest population.
Evidence of Caps Signals
As with the hinge, indirect evidence for “caps” signalling comes from pharyngeal stage gene expression patterns. Indeed, numerous genes are simultaneously expressed in both the distal midline of mdBA1 and the juxtaposition of the mxBA1 and FNPs at the lambdoidal junction with (at least initially) a hinge region gap, including Dlx2, Dlx3, Islet-1, Bmp4, and Shh in the ectoderm and Alx3, Alx4, Cart1, Msx1, Msx2, Foxd1, Foxf1, Foxf2, Prx1, Prx2, Ptch1, and Smo in the mesenchyme (Fig. 5C,D; Depew et al., 2002b; Jeong et al., 2004; Lu et al., 1999).
Loss-of-function studies of most of these genes and many others, either singly or in combination with related genes, have revealed their regulatory roles in the development of regions distal to the hinge. Msx1, for example, has a high level of expression in cap regions and (with the exception of a small, weak patch in the mdBA1 subjacent to the pharyngeal plate; Fig. 5C) is lacking in the hinge region. Msx1−/− mutant mice lack teeth; they also have a differential loss of the midline of the dentary and rostral process of MC and defects in FNP-associated nasal capsular development (personal observations; Satokata and Mass, 1994; Houzelstein et al., 1997). Hinge regional development is for the most part intact (except a small part of the malleus). Attesting to the sensitivity and intricacy in the regulation of the cap system that integrates the FNPs and mxBA1, cleft lip with or without cleft palate (CL/P) forms the most prevalent congenital malformation of the human face, and variant MSX1 alleles in humans have been implicated in both syndromal and isolated cleft lip and palate cases (Cox, 2004). Msx2 is similarly expressed in the cap regions, and compound Msx1−/−; Msx2−/− mutants demonstrate a genetic interaction with regard to midline cap development (Satokata et al., 2000).
Genetic interactions are also seen with the paired-like, Aristaless-related genes Alx3, Alx4, and Cart1 (Qu et al., 1999; Beverdam et al., 2001). Alx4−/−;Cart1−/− mouse mutants are micrognathic as mandibular structures associated with the distal midline (cap) are truncated or lacking (Qu et al., 1999). Although defects in the integration of the upper jaw were not explicitly characterized, mutants have distinctly cleft faces as the two nasal capsules have failed to integrate across the midline; significantly, the cleft is rostral to the presphenoid. Alx3−/−; Alx4−/− murine mutants likewise suffer from a cleft upper face and a distal truncation of their mandibles while sparing the hinge region (Beverdam et al., 2001).
Retinoic acid (RA) balance in the developing embryo is critical because either a deficiency or an excess leads to developmental defects. The effects of vitamin A deficiency syndrome include microophthalmia and cleft lip, palate, and/or face defects that can be prevented with RA administration (Webster et al., 1986; Wedden et al., 1988; Morriss-Kay, 1993; Mark et al., 1995; Morriss-Kay and Skolova, 1996; Mallo, 1997). The retinoic acid RAR and RXR receptors both exist in three forms—α, β, and γ—with isoforms of each. For RAR, this includes α1 and α2, β1 to β4, and γ1 and γ2. RARα transcripts are apparently nearly ubiquitous developmentally, RARβ more restricted, and RARγ in specific mesenchymal populations, including the frontonasal and branchial arch ectomesenchyme (Dolle et al., 1989, 1990; Ruberte et al., 1990, 1991; Leroy et al., 1991; Mendelsohn et al., 1991, 1994). RARα−/−;γ−/− double mutant neonates have diminished eyes, shortened snouts, median facial clefts, occasional exencephaly, and agenic auditory pinnae (Lohnes et al., 1994; Mark et al., 1995). Extensive cell death is seen in the frontonasal CNC at E10.5, and although an olfactory pit forms the frontonasal processes are fused to the ipsilateral maxillary process and never at the midline. The trabecular basal plate, moreover, is widely split and the nasal capsules and ethmoid are represented by rods of cartilage without any midline structures. The associated dermatocranial elements (e.g., premaxillae, nasals, vomers, lacrimals, and frontals) are partially or completely agenic. The orbitotemporal region and those BA-derived elements that develop in intimate association with the nasal capsules (including the maxillae) are deficient and malformed. The incus is fused to the ala temporalis of the alisphenoid. Experimental manipulation of FNP mesenchyme in mice further supports a crucial role in RA signaling in nasal morphogeneis (LaMantia et al., 2000). Thus, RA signaling is required for proper coalescence and development of the upper jaw primordia in mice.
Shh is expressed in many important ventral, midline tissues, including the node, stomodeal ectoderm, notochord, and floor plate, and ventral forebrain and foregut endoderm. In the developing craniofacial primordia, Shh is generally restricted to epithelial layers—often in sites of epithelial–epithelial contacts (Echelard et al., 1993; Jeong et al., 2004). Both gain- and loss-of-function studies suggest that Shh has organizing potential, in particular in the axial mesendoderm (i.e., the notochord and prechordal plate; Echelard et al., 1993; Riddle et al., 1993; Johnson et al., 1994a–c; Marti et al., 1995; Chiang and Flanagan, 1996). Targeted disruption of Shh has devastating effects on cranial development: severe cyclopic holoprosencephaly (i.e., loss of midline structures) and, despite early (∼E9.5) BA development, a near complete loss of the BA-derived structures (Chiang and Flanagan, 1996). Thus, the entire chondrocranium (parachordal, trabecular, and splanchnocranial) is compromised. In all, these defects appear to result from a loss of Shh in midline and unelaborated dorsoventral/mediolateral cephalic development. A proboscis-like “nasal” protrusion forms, however, without discernible skeletal elements. The early loss of midline Shh obscures whatever role that ectodermal SHH signals may play in the refinement of later craniofacial development, a problem that is being overcome in several manners. For instance, Jeong et al. (2004) addressed Shh signaling by conditionally knocking out its receptor, Smo, in presumptive CNC through using the Wnt-1–Cre mouse. Although much of the head skeleton in Wnt-1-Cre; Smon/c mice was altered, it appears that the greatest malformations of structure and losses of gene expression are concentrated along the epithelium of the mandibular and neurocranial midlines (i.e., the caps regions).
Based on their patterns of expression and their capacity to activate gene expression in circa-stomodeal tissues, Bmp genes (Bmp2, Bmp4, and Bmp7 in particular) have featured prominently in models of craniofacial development, pattern, and polarity (reviewed in Depew et al., 2002b; Francis-West et al., 2003). Bmp2 and Bmp4 are expressed in discrete spatiotemporal patterns within the ectomesenchyme and overlying ectoderm (in particular along the oral midline and pharyngeal plate) of the early craniofacial primordia, as well as dorsal midline of the neural tube (Francis-West et al., 1994; Bennett et al., 1995; Liu et al., 2005a, b). Ectopic application of Bmps in mandibular explants has been shown to induce the expression of Msx and Dlx genes (Bei and Maas, 1998; Thomas et al., 1997, 2000; Tucker et al., 1998). Bmp7 is expressed in many cranial tissues, including in the stomodeal and surface ectoderm covering BA1, over and between the frontonasal processes, in the otic vesicle, the commissural plate between the telencephalic vesicles, and the neural tube at the cephalic flexure (Lyons et al., 1995). The effect of genetic inactivation of either Bmp2 or Bmp4, however, is embryonic death (Winnier et al., 1995; Zhang and Bradley, 1996).
To overcome the problems of early lethality and the reciprocal patterns of expression between epithelial and mesenchymal expression of Bmps, several tissue-specific strategies have been used in mice. Dudas et al. (2004) attenuated Bmp signaling in the CNC through loss of the Alk2 receptor by generating a Wnt-1-Cre; Alk2flox mouse. Although some alterations of morphology are seen in structures derived from peri-cleftal tissues and the FNP, the most striking phenotype is the loss of structures associated with the mandibular midline (mandibular cap), such as the rostral process of MC. Conditional loss of Bmp4 in the mandibular ectoderm and associated pharyngeal endoderm was achieved through the use of the Nkx2.5cre mouse (Liu et al., 2005a). A graded series of defects centered along the mandibular midline are apparent in this mouse as well. The Bmpr1a receptor has been floxed using the same Nestin-Cre strategy used to flox Fgf8 (Trumpp et al., 1999; Liu et al., 2005b), which resulted in compromised lambdoidal junctional signaling and a cleft lip.
Transcription factors expressed in the ectoderm are also clearly important to the elaboration of caps-associated patterning (Depew et al., 2002b). For instance, Dlx5 and Dlx6 are both expressed in the anterior cephalic ectoderm, including ectoderm that will eventually cover the BA1 and will form the olfactory placode; whereas expression in the ectoderm that will eventually cover the BA1 is lost as CNC cells are migrating into the arch, expression is maintained in the olfactory placode and nasal pits (Yang et al., 1998; Depew et al., 1999). Loss of Dlx5/6 results in a disruption of the olfactory ectoderm and subsequently in the loss of the nasal capsules and much of the trabecular plate (Fig. 6F). Significantly, although truncated incisor rudiments develop, premaxillary bones also fail to form in Dlx5/6−/− mice. Moreover, the maxillae, which usually form in close association with the nasal capsules and premaxillae, are altered by the loss of the capsules: although they possess clear maxillary characteristic (e.g., frontal process with an infraorbital foramen, molar alveolae, and so on), their morphology has as much in common with the ectopic maxillae developing out of mdBA1 as they do a wild-type maxillae (Fig. 6F; Depew et al., 2002a). This is not so surprising when one sees the nature of the clear mirror-image duplication between the developing upper and lower jaws of the Dlx5/6−/− mutant mice (Fig. 6F).
Comparing the Components of “Hinge and Caps” Model Between Taxa
With our consciously simple model, we return to the question: If homology is the basis of comparative biology, are there homologous developmental mechanisms used to generate homologous gnathostome structures? How robustly can the “hinge-and-caps” model be applied to the development of other taxa? Obviously, one must attempt a meaningful comparison of the relative timing, position, and nature of the epithelial patterning centers (in short, their own etiologies and capacities) as well as those of the CNC. Thus, both the level of conservation between species of the patterns of gene expression in (1) the CNC (as it migrates and as it lines the BA) and (2) the craniofacial epithelia and the function of these genes must be addressed. Although, as stated above, it is not our purpose here to do so by cataloguing the entirety of the comparative expression profiles and functional data of all gnathostomes (indeed, such an endeavor necessitates a review of tome-like length), appreciation of this (or any other) model does require some form of comparison.
Examination of the expression patterns of genes involved in the signaling cascades of Bmp, Fgf, RA, and Shh, as well as the loss- and gain-of-function experimental manipulations of these molecules in chicks, generally support their homologous roles (at least at a gross level), suggesting some level of conservation between birds and mammals in their elaboration of jaw development (e.g., Bally-Cuif et al., 1995; Wall and Hogan, 1995; Barlow and Francis-West, 1997; Helms et al., 1997; Hu and Helms, 1999; Sun et al., 2000; Lee et al., 2001; Schneider et al., 2001; Wilson et al., 2001; Ashique et al., 2002; Mina et al., 2002; Hu et al., 2003; Schneider and Helms, 2003; Abzhanov and Tabin, 2004; Abzhanov et al., 2004; Song et al., 2004; Tucker and Lumsden, 2004; Wu et al., 2004; Marcucio et al., 2005; Havens et al., 2005). The information gleaned from the many studies of the application or repression of signaling cascades in the facial primordial of chicks highlight an important point regarding the hinge and caps model and jaw development: such studies collectively emphasize the synergy and juxtaposition of combinations of these signaling molecules—in particular in the caps regions and pharyngeal plate—in how such molecules regulate jaw development. The model does not depend, moreover, upon the presence or absence of specific players (such as a specific gene transcript); rather, it emphasizes the net effect on cellular behaviors (such as cell death, migration, or proliferation) at specific locations and times of multiple factors. Thus, when comparing the centers of information, the cumulative nature of the centers must be a concern.
These investigations also highlight our growing capacity to address heterochronic, heterotopic, and heterofacient changes in the hinge and caps patterning centers between different levels of taxa outside of studies of cis-regulatory elements. This can be exemplified in two types of recent studies on avian embryos. In the first type, isochonic and isotopic tissue grafts between different, but related, species have demonstrated that the CNC is largely responsible for the species-specific characteristics of jaw-associated elements, in particular the upper beak (Schneider and Helms, 2003; Tucker and Lumsden, 2004; Wu et al., 2004). Such results were neither unexpected nor novel (after all, decades-old xenograft experiments between amphibians and between birds had clearly demonstrated this principle; see expanded discussions in Huxley and de Beer, 1934; Hall, 1988); nor was the demonstration that the CNC could effect a wholesale change in the nature of the ectodermal derivatives. What really made these studies highly significant was the productive use of these graft assays to address potential manifestations of heterotopic and heterochronic differences in the generation of jaws between the species involved, which might then explain subsequent species-specific differences. In another set of experiments, correlations between morphology and heterotopic and heterofacient expression of genes and molecules were drawn. For example, in examining the expression profiles of upper jaw-associated signaling in Darwin's finches, Abzhanov et al. (2004) noted a correlation between the timing (heterochrony) and level (heterofacience) of Bmp4 expression and the eventual size of the upper beak that formed; they then used this correlation to attempt to experimentally recapitulate the plausible effects a change in Bmp4 expression would have on beak morphology. These studies further highlight the predictive value that stems from basic descriptive embryology (neontology).
By no means is our emphasis thus far on mice and chicks indicative of a lack of studies on other taxa. For example, studies on apparently homoplastic structures, such as the pharyngeal teeth in teleosts, have been emphasizing the neo-deployment of developmental cascades and the potential heterochronic, heterotopic, and heterofacient changes that may be accompanying (e.g., Fraser et al., 2004). Moreover, were examined, the importance of Fgf, endothelin, RA, Bmp, and Shh for cephalogenesis has also been established in amphibians and zebrafish (e.g., Ellies et al., 1997; Reifers et al., 1998; Degitz et al., 2000; Shanmugalingham et al., 2000; Miller et al., 2000; Koide et al., 2001; Sasagawa et al., 2002; Crump et al., 2004; Furthauer et al., 2004, Wilson and Tucker, 2004; Groppelli et al., 2005; Albertso and Yelick, 2005; Reversade et al., 2005).
What of the comparative expression profiles of the CNC partners in the signal-response two-step of jaw development? Although basic comparisons have been made, these have to be more cohesively assessed in relation to timing, position, and nature of the signaling centers. The nested expression of Dlx genes in mice, for instance, has been correlated with the segmentation of BA-derived elements and the evolution of jaws (reviewed in Depew et al., 2005); moreover, lampreys, which do not have segmented BA structures, do not have nested patterns of Dlx gene expression. However, one must ask whether such nesting is an obligate apomorphy of gnathostomes. Along with the laboratories of David Stock, Clare Baker, Pip Francis-West, Abigail Tucker, Bethan Thomas, and others we are systematically profiling the expression of Dlx genes in class level taxa, including in sharks, snakes, chicks, and mice, while concentrating on possible evidence for heterochronic, heterotopic, and heterofacient differences in expression (manuscript in preparation). For a family of transcription factors known to genetically interact, such an understanding is crucial to further understanding of their biology, but by no means are they the only CNC factors of importance or investigation.
The First Steps to Getting a Head: Proximate Events in the Induction of Hinge and Caps Signaling
We are not limited, however, in furthering our understanding of the model and the comparative development of jaws until a comprehensive comparative examination of gene expression and function has been completed. We have already suggested that arguments for either “pre-pattern” or “epigenetic” programming in CNC elaboration of jaw form each necessitate proximate sources of information and that a gradually building consensus is that the development of BA1 (and FNP) hard tissues involves a combination of both prepattern and epigenetic mechanisms. However, this still leaves the need to explain the nature and organization of proximate sources of developmental information. We must still ask: Are there specific, significant factors or mechanisms involved in determining the specific constellation of genes BA1 and FNP CNC cells express before emigration (or while they migrate) or which they turn on (or off) when they reach the BA? If so, what are these factors and are they conserved? What are the proximate sources of induction and/or refinement of the circa-stomodeal epithelial signaling centers (i.e., hinge and the caps) and their particular patterns of gene expression? What positions the instructive signals that we know take part in jaw pattern and development?
In our model, we have emphasized the notion that the functional registration of gnathostome jaws is achieved by the integration of “hinge” and “caps” signaling, with the “caps” sharing at some critical level a developmental history that potentiates their own coordination. Several factors point toward the pharyngeal plate participation in the establishment of the jaw articulation. First, the most parsimonious explanation for the capacity of the hinge region primordia to generate mirror-image duplications of structure centered around the jaw articulation would indicate the presence a focal source of impellent information (or two closely juxtaposed sources) directing the development of the jaw articulation (hinge). Such a strategic placement of the source of information would serve to facilitate, localize, and coordinate the development of both the upper and lower components of the hinge so crucial to jaw function. Second, the work of Drew Noden, Fillipo Rijli, and others clearly has demonstrated that mirror-image duplications of primary jaw articulation elements (e.g., articular/malleus and quadrate/incus) can be centered around the plate and include both BA1 and BA2. Third, the attachment of the jaws to the neurocranium has historically involved the second arch splanchnocranium. Again, the placement of informative signals at the plate would serve to facilitate, localize, and coordinate the connections of the first and second BA. Hence, one is compelled to address what dictates the development, competence, and activity of the endodermal–ectodermal contacts of the first pharyngeal plate.
And what of those portions of the jaw apparatus that are derived “furthest from the hinge”? Are there parsimonious arguments for their coordinated etiology? We suggest that there are and hypothesize that the frontonasal and distal mandibular cap regions, in fact, do share a developmental history crucial to their coordination, one greatly influenced by the anterior mesendoderm and the processes of head induction. Ideally, the structures of the CNS (i.e., the contents) and the neurocranium (i.e., the container) are both developmentally and functionally integrated. Such integration may be reflected by the number of genes expressed early in embryogenesis in both the nascent neural plate and the surface ectoderm that will eventually cover the craniofacial complex and from which patterning information for the skull is known to eventually emanate. Among these genes are a subset known to play roles in the development of both the forebrain and the neurocranium, including Dlx5, Pax6, and Otx2 (Simeone et al., 1992; Yang et al., 1998; Depew et al., 2002b). We postulate that genes which are expressed in both the nascent anterior cephalic ectoderm and the anterior neural plate function to coordinate and integrate the development of the CNS and the skull that provides it physical protection. This coordination and integration may happen through multiple steps and may act to (1) actively demark the early ectoderm as an “anterior” field; (2) protect both the presumptive anterior cephalic ectoderm and the anterior neurectoderm from posteriorizing or extraembryonic influences; and (3) act as a mechanism for the coordinated recognition of an interacting partner across an organizing center such as the anterior neural ridge. Because jaws are, of course, integral parts of the head, the organization of their germinal centers of patterning begins with the general regionalization of the embryo and the seminal organization of the head. Thus, to address the question of the proximate sources of hinge and caps information, we must examine the germination of the gnathostome head and then ask whether there are aspects conserved between gnathostomes that may explain the level of coordination seen in the midlines of the upper and lower jaws typical in gnathostomes.
Ahead of Jaw Development
Through manipulation of embryos, it has been variously understood for close to a century that a second body axis can be experimentally induced, an activity eventually attributed to an embryonic “Organizer” (Spemann and Mangold, 1924; Spemann, 1931; Huxley and de Beer, 1934; Nieto, 1999; Gerhart, 2001). This organizer was initially coherently described after it was found that cells of the upper (dorsal) blastopore lip of gastrulae-staged urodele amphibian embryos could generate, by changing the fate of surrounding cells, a second body axis when transplanted into an indifferent embryonic region (Spemann and Mangold, 1924; Spemann, 1931). This axis was found to be organized both rostrocaudally and dorsoventrally; thus the organizer not only instructed neighboring cell and tissue fate it further directed the global morphogenetic cues that regulate four-dimensional embryonic development (Spemann and Mangold, 1924; Hamburger, 1988; Harland and Gerhardt, 1997; Nieto, 1999; Camus and Tam, 1999; Gerhart, 2001; Stern, 2005). Equivalent embryonic regions subsequently have been identified in teleost (the shield), avian (Henson's node), and mammalian (node) embryos (Waddington, 1932/ 3; Oppenheimer, 1936a, b; Beddington, 1994; Kessler and Melton, 1994; Shih and Fraser, 1996; Camus and Tam, 1999; Foley and Stern, 2001).
Significantly, in these seminal “organizer” experiments, head structures rostral to the level of the otic capsule were not induced (Spemann and Mangold, 1924; reviewed by Gerhart, 2001). These and subsequent experimental embryological and teratogenic experiments, mainly in amphibians, lead to the concept of a distinct “head organizer” for the organizer activity responsible for anterior, cephalic development, whereby a second head could be induced (Spemann, 1931; Mangold, 1933; Niehrs, 1999; Nieto 1999; Stern, 2005). For example, grafts of the dorsal lip of “early” amphibian gastrulae or “early” Henson's node of chicks are able to induce secondary axes possessing a full range of identities, whereas grafts of late gastrulae induced only posterior structures (Spemann, 1931, 1938; Mangold, 1933; Holfreter, 1938; Dias and Schoenwolf, 1990; Storey et al., 1992; Foley and Stern, 2001; Gerhart, 2001). Fate maps in gnathostomes have typically demonstrated that the axial mesoderm (or mesendoderm) migrating out of the dorsal lip and node comes to underlie the future neural plate and that this mesendoderm can induce neural tissue (see below).
It generally has been emphasized that organizer-induced secondary heads posses patterned neural plate derivatives as well as primary sensory apparatuses (e.g., optic) whereas the step-wise developmental and morphologic details of these induced heads have frequently been less characterized. In reflection of what Spemann (1938) termed the problem of “wholeness” (and secondary inductions), a head connotes a skull and jaws, and often lost in the noting of a patterned nervous system is the notion that, if secondary heads have been induced, such heads would apparently also have to posses embryonic cephalic skeletal elements, including inchoate jaws (Fig. 8A). The organizer is integral to the process of gastrulation wherein the gnathostome embryo elaborates axes, and definitive endoderm and mesoderm, from epiblastal ectoderm. With regard to the vertebrate head, from this ectoderm must come appropriate, developmentally competent, specified, committed, and partitioned neural plate neurectoderm, anterior neural ridge and neural fold (both neural crest generative and barren populations), cephalic surface ectoderm with primary sensory (olfactory, optic, and otic), cranial nerve, and stomodeal (e.g., adenohypophyseal and dental) placodes, buccopharyngeal membrane, pharyngeal clefts and pouches, various peri-oral specializations (such as cement and hatching glands, balancers, egg teeth, vibrissae, and ruggae), as well as the embryonic endoderm and mesoderm. Whether actively or passively regulated, the epiblastal ectoderm also contains the embryonic–extraembryonic boundary, and, although not a cell population particularly well fate-mapped across species, it must also develop the ectoderm that covers the heart and midline body wall positioned initially between the neural plate and the presumptive future umbilicus or point of gut closure. This places a population of midline oral ectoderm between the neural plate and the ectoderm over the developing heart (Fig. 8B,C). In short, such inductive organizer manipulations would have to result in a re-patterned ectoderm (with induced neural plate, placodal, stomodeal, and neural crest populations) as well as mesoderm and endoderm (including a second prechordal plate and pharyngeal endoderm).
Although extensive effort has been made to understand ectodermal subdivisions within the neural plate, there has been a much less of an effort to understand early regional pattern within the cephalic surface ectoderm outside of those investigations examining the origination of the neural crest, the cranial placodes, and some perhaps generally underappreciated work involving the cement gland in amphibians (Gammill and Sive, 1997, 2000, 2001; Fang and Elinson, 1999; Baker and Bronner-Fraser, 2001; Schweickert et al., 2001; Sauka-Spengler et al., 2002; Wardle and Sive, 2003; Streit, 2004; Nokhbatolfoghahai and Downie, 2005). Analyses of the development of the early vertebrate epiblast have typically been concerned either with (1) when and how the distinction between neural and non-neural ectoderm is established or (2) how is positional information, i.e., anterior (head) and posterior (trunk), within the neural plate established (Wilson and Edlund, 2001; Stern, 2002, 2005; Wilson and Houart, 2004), and have lead to several questions, including: Is the epiblast ubiquitously competent to become either neurectodermal or surface ectodermal, and if so, is the specification and eventual commitment to neural and surface ectodermal achieved simultaneously, or does one precede the other? Is there is a default state, either neural or surface ectodermal (epidermal in general) in character, in the early epiblast, and if so, does the nondefault state come about by means of an active induction, a repression, or a de-repression? With regard to anterior positional identity and neural induction, does one precede the other or does the neural plate come about with a default anterior character? Although these issues have been detailed and reviewed elsewhere, particularly in relation to neural induction (Niehrs, 1999; Foley and Stern, 2001; Wilson and Edlund, 2001; Stern, 2002, 2005; Wilson and Houart, 2004), with regard to the formation of jaws it is expedient to make here note of several points that follow from them.
First Perambulations: Polarity and Pattern in the Pregastrulation Epiblast
First, polarity and, therefore, pattern appears to already exist in the epiblastal ectoderm before gastrulation as evinced by asymmetric patterns of gene expression typically encountered within epiblast (e.g., Bmp4, Otx2, Fgf8, T [Brachury], Cripto, Wnt3, Nodal, Hesx1, Hex, Cerberus-like [Cerl], VE-1, Lim1 [Lhx1], GATA, Dlx, and Dkk1; Lawson et al., 1991; Lawson and Pedersen, 1992; Dono et al., 1993; Simeone et al., 1993; Ang and Rossant, 1994; Crossley and Martin, 1995; Belo et al., 1997; Varlet et al., 1997; Eyal-Giladi, 1997; Beddington and Robertson, 1998; Biben et al., 1998; Ding et al., 1998; Shawlot et al., 1998; Thomas et al., 1998; Liu et al., 1999; Pera et al., 1999; Crossley et al., 2001; Tian et al., 2002; Chapman et al., 2002). This asymmetry is mirrored by the underlying hypoblast (visceral endoderm; e.g., Otx2, Lim1, Hesx1, Hex, HNF3β, Lefty1, Nodal (Acampora et al., 1995; Thomas and Beddington, 1996; Belo et al., 1997; Filosa et al., 1997; Varlet et al., 1997; Thomas et al., 1998; Meno et al., 1999; Shimono and Berhinger, 1999; Kimura et al., 2001).
Acquisition of an “Anterior” Character State in Gnathostome Presumptive Head Epiblast
Second, evidence is accumulating which suggests that the gnathostome presumptive head epiblast first acquires an “anterior” character state before the demarcation of the anterior neural plate–cephalic surface ectodermal boundary (Viebahn et al., 1995; Beddington and Robertson, 1998; Knoetgen et al., 1999; Pera et al., 1999; Gerhart, 2001; Withington et al., 2001; Stern, 2002; Chapman et al., 2003; Stern, 2005). Several studies into the nature of the murine node have suggested that maintenance if not induction of anterior structure in mice requires information not contained in the node (but plausibly contained in the AVE; Tam and Steiner, 1999; Camus et al., 2000). For example, Beddington (1994) found that, whereas the murine node could induce a second axis, at no time could it induce an axis including a head, and that extirpation of the anterior visceral endoderm greatly affected the development of the head (Thomas and Beddington, 1996).
Furthermore, the striking total loss of head structures anterior to the otic capsules—not simply neural plate, but neural plate, cephalic surface ectodermal derivatives, and associated mesenchyme—has been achieved through targeted deletions in mice of several genes, including among others Lim1 (Lhx1; Shawlot and Berhinger, 1995), Otx2 (Acampora et al., 1995, 1998; Matsuo et al., 1995; Ang et al., 1996; Tian et al., 2002), HNF3β (Rossant et al., 1994; Filosa et al., 1997; Dufort et al., 1998; Kinder et al., 2001c), and Dkk1 (Mukhopadhyay et al., 2001). Importantly, such mutant mice can make it to term as their cardiovascular systems develop. The heart field develops in anterior-most spreading mesoderm, and cardiogenesis is thought to be induced by anterior definitive endoderm (Schnieder and Mercola, 1999; Marvin et al., 2001; Harvey, 2002); significantly, it is overlain by epiblastal ectoderm initially anterior to the midline facial cephalic ectoderm (including the buccopharyngeal ectoderm and midline mandibular arch) and prosencephalic (forebrain) neural plate. Thus, that these mutant mice have ectoderm that covers across the rostral end of the truncated head demonstrates they have not simply lost all “anterior” ectoderm.
Most of these genes, however, have dynamic expression patterns in the early epiblast, the node, and its derivatives, and the anterior visceral endoderm (AVE, hypoblast equivalent in mice), a fact that complicates analysis of the correlation of gene expression and gene function. Fortunately, chimeric embryos can be generated wherein the epiblast is essentially either wild-type or genetically modified, whereas the visceral endoderm is of the opposite genotype (Robertson and Beddington, 1989). Chimeras with defective Otx2, HNF3β, Lim1, or Nodal genes have demonstrated their necessity in the epiblast and AVE for normal anteroposterior development (Rossant et al., 1994; Filosa et al., 1997; Varlet et al., 1997; Dufort et al., 1998; Rhinn et al., 1998; Shawlot et al., 1999; Kinder et al., 2001; Perea-Gomez et al., 2001a, b, 2002), where they variously act (and interact) to help establish an “anterior” region initially protected from the “posteriorizing” influence of the forming node and streak by means of the relative movements of the AVE and epiblast and induction of gene expression within the anterior (head) epiblast. Dkk1 is apparently dispensable in the AVE (Mukhopadhyay et al., 2001).
From Organizer to Mesendoderm to Pharyngeal Endoderm: Successive Derivative Influences on Anterior Cephalic Ectoderm and Jaws
Third, the nodal organizer (NO) and its anterior axial mesendoderm (AME, prechordal plate) and pharyngeal endodermal derivatives further instruct the development of the anterior epiblastal ectoderm, including the cephalic surface ectoderm. The mesendoderm is so named as it is composed of a coherent group of cells that will give rise to both endoderm and mesoderm: this includes the entirety of the axial midline and, therefore, consists of the notochord and the AME that lies rostral to the notochord beneath the developing forebrain near the junction of the hypophysis and the buccopharyngeal membrane (Adelmann, 1922; Aasar, 1931; Meier, 1981; Seifert et al., 1993; Sulik et al., 1994; Foley et al., 1997; Pera and Kessel, 1997; Niehrs, 1999; Kiecker and Niehrs, 2001; Roessler and Muenke, 2001). Lineage tracing studies suggest that the first cells through the gastrulating nodal organizer are the anterior definitive endoderm and AME and that, where the AME is initially a singular, coherent cellular mass, it eventually segregates into separate mesodermal and endodermal populations: the mesoderm subsequently gives rise to perioccular musculature and the endodermal subpopulation integrates with the nascent anterior definitive endoderm contributing to the anterior pharyngeal foregut (Lawson et al., 1991; Foley et al., 1997; Tam et al., 1997; Warga and Nusslein-Volhard, 1999; Tam and Berhringer, 1997; Kiecker and Niehrs, 2001; Gerhart, 2001; Kinder et al., 2001; Lawson and Schoenwolf, 2003).
Developmental biologists have long held that the AME (historically, the anterior archenteron roof/prechordal plate) is in some way integral to the normal formation of the head (e.g., Adelmann, 1922, 1936; Adams, 1924; Dart, 1924; Aasar, 1931; Waddington, 1932, 1933, 1936, 1937, 1962; Mangold, 1933; Huxley and de Beer, 1934; Wright and Wagner, 1934; Spemann, 1938; Nieuwkoop, 1952; Juriloff et al., 1985; Sulik and Johnston, 1982; Elinson and Kao, 1993; Niehrs, 1999; Roessler and Muenke, 2001). Significantly, there is a long history of evidence for an inductive capacity in the AME: heterotopic grafting of the AME in amphibians or chicks has been shown to lead to ectopic head structures (Spemann, 1931, 1938; Mangold, 1933; Pera and Kessel, 1997; Zoltewicz and Gerhart, 1997; Kieker and Niehrs, 2001). In mice, grafts of the axial mesendoderm in conjunction with a grafted AVE have been shown to induce anterior ectodermal markers such as Otx2 (Tam and Steiner, 1999). Regardless of its inductive capacity, moreover, the AME serves a role beyond induction.
Not withstanding that extirpated early nodal organizer-related tissue heals and reforms (in line with the notion of the Organizer being a position and not a dedicated set of cells), physical ablation of early anterior endodermal and AME populations in amphibians, zebrafish, chicks, or mice leads to alteration of anterior cephalic ectodermal and forebrain markers (including Hesx1 [Ganf], BF1, and Fgf8) and subsequent deficits in anterior development, mostly midline in character, including defective forebrain vesicles, cyclopia, midline stomodeal loss, and cardiac malformations (Mangold, 1931; Adelmann, 1936; Shih and Fraser, 1996; Thomas and Beddington, 1996; Niehrs, 1999; Li et al., 1997; Pera and Kessel, 1997; Foley et al., 1997; Shimamura and Rubenstein, 1997; Schneider and Mercola, 1999; Pera et al., 1999; Camus et al., 2000; Saude et al., 2000; Gerhart, 2001; Kinder et al., 2001a–c; Withington et al., 2001).
Although many analyses of ablation do not emphasize the development of skeletal elements, the midline character of the defects suggests that signaling from the caps, in particular, is likely to be effected/defective. From the seminal attempts of Etienne and Isidore Geoffrey St. Hillaire in the early 19th century to generate and classify developmental anomalies, it has been recognized that abnormalities of the head could be systematically categorized (reviewed in Wright and Wagner, 1934; Adelmann, 1936). It has also been widely noted that, with such nosologically and etiologically diverse developmental cranial anomalies (both teratogenic and naturally occurring) involving the midline of the head, including otocephaly, cyclopia, and holoprosencephaly, there is the tendency toward the manifestation of midline oral malformations, including micrognathia (or agnathia) and the collapse, truncation, or loss of the frontonasal and maxillary region (Wright and Wagner, 1934; Adelmann, 1936; Demyer et al., 1964; Sulik and Johnston, 1982; Juriloff et al., 1985; Elinson and Kao, 1993; Cohen, 2001, 2002). Notably, with such malformations, the hinge region is relatively spared, whereas the cap (and associated midline) regions are more affected. This finding is exemplified with the murine oto (otocephaly) mutant (Fig. 7B; Juriloff et al., 1985; Zoltewicz et al., 1999), suggesting a relative commonality in the origins of the caps and distinction of hinge development.
This classic understanding has generally been supported by experiments wherein signals from the AME are effected. Signaling deficiencies centered in the AME or the anterior ectoderm that responds to them have typically been blamed for such anomalies, but the nature of these deficiencies has not always been clear. Work over the past 15 years, fortunately, has made the conflux of pre- and postgenerative molecular and cellular influences of the AME on head development less caliginous if no less byzantine (see de Souza and Niehrs, 2000; Kiecher and Niehrs, 2001; Roessler and Muenke, 2001; Schier, 2003; Tam et al., 2003). As noted above, the pre-gastrulation, pre-organizer–influenced epiblast is a polarized tissue, in large part as a consequence of developmental events initiated by the germinal establishment of asymmetry and resulting in the focalization of “determinants” that subsequently induce the formation of the vertebrate Organizer itself (Nieuwkoop, 1985, 1997; Eyal-Giladi, 1997; Joubin and Stern, 1999; Beddington and Robertson, 1999; Brennan et al., 2001; Gerhart, 2001; Kiecher and Niehrs, 2001). These “determinants” eventually result, directly or indirectly, in the graded activity (principally in the epiblast) of the transforming growth factor (TGF-β) -related secreted signaling molecule Nodal (Brennan et al., 2001; Gerhart, 2001; Bertocchini and Stern, 2002; Schier, 2003; Tam et al., 2003).
Briefly, Nodal ligands are secreted molecules that bind the Activin receptors ActRIB (ALK4), ActRIIA, ActRIIB, and possibly ALK7 (Kumar, 2001; reviewed in Schier and Shen, 1999; Whitman, 2001; Schier, 2003). This binding appears to necessitate EGF–CFC coreceptors: Cripto and Cryptic in mammals, oep (one-eyed pinhead) in zebrafish, FLR-1 in frogs, and CFC in chicks. Receptor activation is thought to result in Smad phosphorylation, in particular of Smad2 (and likely Smad3), whereafter specific transcription factors then interact with the phosphorylated Smad2 to regulate downstream targets of the Nodal signal. Perhaps the best characterized such transcription factor is FoxH1 (Fast1), for which evidence suggests that FoxH1–Smad2 complexes mediate Nodal induction of several genes, including goosecoid, Lefty2, Pitx2, Lim1, Mix2, an Nodal itself (Yamamoto et al., 2001). Nodal signaling is attenuated by extracellular inhibitors of the Lefty (including antivin, Lefty1 and Lefty2) and Dan (including Cerberus and Cerberus-like [Cerl]) gene families, the latter of which further acts to antagonize both the bone morphogenetic protein (BMP) and Wnt signaling pathways (Picolo et al., 1999; Belo et al., 2000). Lefty expression is induced by Nodal signaling and generally follows Nodal expression patterns; hence, Nodal autoregulation and induction of its Lefty antagonists sets up a system of feedback inhibition. Cerberus and Cer1, however, are expressed in the hypoblast/visceral endoderm and AME and act upon the adjacent ectoderm. There is further evidence that, like other TGF-β–related molecules, active Nodal signaling is regulated both by convertases acting on pro-proteins and latent binding proteins in the extracellular matrix (Schier, 2003). Nodal signaling activity, thus, is potentially regulated at multiple levels.
Evidence from studies on Xenopus, chick, zebrafish, and mice all demonstrate that vertebrate Nodal (and TGF-β superfamily) -directed signaling is essential to the initial formation within the epiblast of the NO (as its name would suggest), to the subsequent determination of NO-derivative cell fate, and, not surprisingly, to head and jaw development (Iannaccone et al., 1992; Conlon et al., 1994; Matzuk et al., 1995; Jones et al., 1995; Nomura and Li, 1998; Feldman et al., 1998; Sampath et al., 1998; Osada and Wright, 1999; Schier and Shen, 2000; Song et al., 1999; Kiecher et al., 2000; David and Rosa, 2001; Lowe et al., 2001; Roessler and Muenke, 2001; Schier and Talbot, 2001; Schier, 2003; Vincent et al., 2003; Lu and Robertson, 2004).
Some of the clearest examples for the role and impact for this system specifically on jaw development come from studies in mice of the genetic attenuation of the signaling system through differential allelic loss of the Nodal ligand, Activin receptors and Smad transcriptional mediators. For example, mice harboring one hypomorphic and one null allele of Nodal exhibit a loss of the prechordal plate, a truncated anterior foregut, and, except for a small fleshy proboscis, loss of the facial midline (Lowe et al., 2001). Cre-mediated mosaicism of Nodal in the epiblast, resulting in reduced Nodal activity, leads to the reduction or abrogation of the anterior expression of Shh, Hex, Hesx1, and Otx2, and results in defective AME development and a truncated embryonic head (Lu and Robertson, 2004). Whereas ActRIIA−/− mutants are micrognathic, with the distal midlines of the dentaries hypoplastic and fused (Matzuk et al., 1995), ActRIIA−/−; Nodal+/− mice are either cyclopic or anophthalmic (and also apparently lacking nasal capsular-associated structures) with severe micrognathia (Song et al., 1999); although differential skeletal staining was not reported, such mice develop outer ears, suggesting at least some development around the first pharyngeal cleft remains.
Nodal also interacts genetically in mice with Smad2, whose protein is its putative signaling mediator (Nomura and Li, 1998). Ten percent of Smad2+/− heterozygous mice have variable anophthalmia and micrognathia of the dentaries, whereas their nasal structures are also likely affected (Nomura and Li, 1998). Targeted loss of Smad2 in the epiblast results in anterior defects, including loss of ANR expression of Fgf8 and Shh and Foxa2 in the AME (Vincent et al., 2003). More than 50% of the compound Smad2+/−; Nodal+/− mutants, however, harbor severe craniofacial defects, including cyclopia (morphologically distinct, however, from the cyclopia associated with ActRIIA−/−; Nodal+/−) or severe truncation of the head but are in possession of outer ears. Compound Smad2+/−; Smad3+/− mice also exhibit exacerbated craniofacial defects (Dunn et al., 2004; Liu et al., 2004), with a range of phenotypes the most severe of which is cyclopia with a dorsal proboscis (Liu et al., 2004). These mice appear to have defective endodermal development, with an enlarged pericardial cavity, and Shh is absent from the prechordal plate; they also lack Fgf8 expression in the commissural plate but maintain it within the pharyngeal ectoderm.
Studies in zebrafish and Xenopus in large part corroborate these murine studies. Two zebrafish nodal-related genes, cyclops and squint, are essential for normal organizer and anterior development (reviewed in de Souza and Niehrs, 2000; Schier and Talbot, 2001). Mutations in the individual genes leads to prechordal plate deficiencies and cyclopia (with loss of attendant gene expression such as gsc), whereas compound mutants fail to form an NO (shield) and, hence, systemically lack mesendodermal derivatives (Hatta et al., 1991; Thisse et al., 1994; Brand et al., 1996; Heisenberg and Nusslein-Volhard, 1997; Talbot et al., 1998; Feldman et al., 1998; Rebagliata et al., 1998a, b; Sampath et al., 1998; Schier and Talbot, 2001). Moreover, overexpression of Cyclops and squint induces endoderm and mesoderm (Gritsman et al., 2000; Kikuchi et al., 2000). In Xenopus, at least five mesendodermal-inducing, nodal-related factors, the Xnrs, have been isolated (reviewed in Whitman, 2001; Schier, 2003). Ectopic expression of Nodal inhibitors (Antivin or Cerberus-short) or dominant-negative Nodal proteins (i.e., cmXnr2) leads to anterior defects and limited mesendodermal development (Whitman, 2001).
Differential Nodal activity, moreover, appears to lead to differential cell fates within the NO derivatives: initially, high levels appear to induce the formation of the AME and endoderm, whereas lower levels induce greater mesoderm development (reviewed in Osada and Wright, 1999; Joseph and Melton, 1998; Kiecker and Niehrs, 2001; Whitman, 2001; Green, 2002; Schier, 2003). These studies further serve to highlight the idea that the appropriate balance of NO-derivative cell fates (endoderm, mesoderm, and mesendoderm) must be achieved. Balanced Nodal signaling must be achieved in such a manner that the “anterior” ectoderm already in place—both presumptive neurectoderm and cephalic surface ectoderm—is protected (by Nodal antagonists such as Cerberus and Lefty) from a signaling cascade that converts ectoderm into endoderm and mesoderm (Piccolo et al., 1999; Niehrs, 1999; Thisse et al., 2000; Perea-Gomez et al., 2002; Yamamoto et al., 2004; Kuroda et al, 2004). Protection of the anterior ectoderm appears to be one function of the NO-derived AME (Glinka et al., 1997, 1998; Shimamiura and Rubenstein, 1997; Bachillier et al., 2000; de Souza and Niehrs, 2000; Kazanskaya et al, 2000; Kiecker and Niehrs, 2001; Mukhopadhyay et al., 2001; Anderson et al, 2002; del Barco Barrantes et al., 2003).
The AME is the source of a range of BMP (e.g., Noggin, Chordin, and Cerberus homologues), Nodal (e.g., Cerberus homologues), and Wnt (e.g., Dkk1, Tlc, Cerberus, Frzbs, and Cresent homologues) signaling antagonists critical to normal head development (Chapman et al., 2004; Wilson and Houart, 2004). Dkk1 is expressed in the AVE, NO, and AME equivalents of amphibians, zebrafish, chick, and mouse (Glinka et al., 1998; Hashimoto et al., 2000; Shinya et al., 2000; Mukhopadhyay et al., 2001; Niehrs et al., 2001; Chapman et al., 2004) and acts to inhibit β-catenin–mediated Wnt signaling through competitively binding to the LRP5/6 Wnt-co-receptors (Bafico et al., 2001; Mao et al., 2001; Semenov et al., 2001). Overexpression of dkk1 in blastula-stage embryos of Xenopus or zebrafish results in enlarged heads bearing enlarged forebrains, eyes, and circum-oral structures (e.g., cement glands; Glinka et al., 1998; Hashimoto et al., 2000; Shinya et al., 2000). Conversely, variable loss of dkk1 expression results in anterior malformations, including acephaly or microcephaly with cyclopia and agenesis of circa-oral structures (e.g., the cement glands in Xenopus; Glinka et al., 1998; Kazanskaya et al., 2000; Mukhopadhyay et al., 2001; Niehrs et al., 2001). Dkk1−/− mice exhibit a striking loss of head structures anterior to the otic capsules (Mukhopadhyay et al., 2001). Although expressed in the AVE, chimera analysis suggests that Dkk1 is not necessary in the AVE but rather in the AME. Mutant mice develop hearts but, significantly, fail to form perforated buccopharyngeal membranes, olfactory-associated prominences, or clear BA1s. Notably, however, they appear to form appropriately placed ectodermal invaginations at the first pharyngeal cleft and external ears develop (Fig. 1 in Mukhopadhyay et al., 2001). The presence of ears suggests the possibility that some hinge development may occur. The effects of the loss of Wnt inhibition can be ameliorated by the additional loss of the LRP6 Wnt-co-receptor (MacDonald et al., 2004); clearly then, without balanced Wnt signaling jaws will not form although anterior ectoderm can.
Loss-of-function studies of Noggin and Chordin demonstrate that inhibition of BMP signaling is likewise essential for craniofacial development (Bachiller et al., 2000; Stottmann et al., 2001; Anderson et al., 2002; del Barco Barrantes et al., 2003; Kuroda et al., 2004). Both Noggin and Chordin are expressed in the NO-derived AME and the ANR (Bachiller et al., 2000; Anderson et al., 2002; Kuroda et al., 2004), and mice homozygous null for either results in increased BMP signaling and a range of moderate craniofacial malformations; Chd−/−; Nog−/− and Chd−/−; Nog+/− mice, however, have severe head and craniofacial defects, including cyclopia with a single nasal opening and agnathia, but possess external ears (Bachiller et al., 2000; Stottmann et al., 2001; Anderson et al., 2002). Chd−/−; Nog−/− mice lack early anterior axial Shh expression (as do some Chd−/−; Nog+/−) and have a reduced anterior pharynx and no trachea. Gene dosage in BMP inhibition also appears to be important: Chd−/−; Nog+/− mice exhibit a wide range of phenotypes from close to normal to nearly headless (Stottmann et al., 2001; Anderson et al., 2002), perhaps suggesting an effect on a small number of cells that usually exert a potent graded influence over their neighbors. Anderson et al. (2002) describe three general categories of Chd−/−; Nog+/− mutants, including one in which the midline derivatives of the frontonasal prominences and BA1 are lacking, and note the correlation of such defects with concomitant loss (leading to holoprosencephaly) of the anterior midline neural tissues. Such loss of midline structures is in line with the loss of gene expression in both the distal midline of mdBA1 (Ptch1), ANR, and commissural plate (e.g., Fgf8 and BF1 [Foxg1]). Mice null for Twisted gastrulation (Twsg1), which is thought to act as a BMP inhibitor, also shows a similar range of midline truncations (Petryk et al., 2004). In similar manner, increased BMP signaling by means of application of ectopic BMP proteins to the anterior, ventral forebrain yields cyclopic chicks and reduces regional Shh and Fgf8 expression (Golden et al., 1999; Ohkubo et al., 2002).
It has been demonstrated clearly in both Xenopus and mouse that the systems of Wnt and BMP inhibition act in concert (Glinka et al., 1997, 1998; Niehrs, 1999; Niehrs et al., 2001; del Barco Barrantes et al., 2003). Although some Dkk1+/−; Nog+/− mice survive to adulthood, many are perinatal lethal evincing a wide range (from upper jaw midline malformations and anophthalmia to acrania rostral to the parietals) of craniofacial defects. These malformations are accompanied by the early anterior loss of Fgf8, Hesx1, Pax6, and Six3—each necessary for some aspect of anterior craniofacial development. Similarly, in Xenopus embryos co-injected with anti-Dkk1 antibodies and morpholinos to both noggin and chordin, the loss of anterior structures (e.g., cement glands) is synergized.
Legacy for the Anterior Pahryngeal Endoderm
As stated above, the endodermal component of the AME integrates with the anterior endoderm to contribute to the anterior pharyngeal endoderm (APE). This integration raises several questions: What is (are) the eventual fate(s) of the endodermal AME cells? Are the AME-derived endodermal cells maintained as a cohesive unit within the APE or do they disperse within the APE? Does the AME-derived endoderm continue to actively participate in head development by exerting, perhaps in a planar manner, an influence on the APE?
Although it is unclear whether the AME-derived endoderm actively patterns the APE, it is clear that the APE itself possesses patterning information. The APE is the source of several familiar signaling molecules, including Shh and various Fgfs and Bmps, suggestive of an active patterning role (Crossley and Martin, 1995; Begbie et al., 1999; Graham, 2001; Smith, 2003). Bmp7 signaling emanating from the pharyngeal pouches, for example, appears to induce the epibranchial placodes in the surface ectoderm (Begbie et al., 1999). The APE also has the capacity to regulate the development of cranial hard tissues. Whereas the relative contributions of endoderm and ectoderm to individual vertebrate dentitions have been debated for some time, endodermally derived pharyngeal teeth have a long history, one possibly separate from the evolution of jaws (Smith, 2003). Moreover, the cytodifferentiation of BA cartilage in certain amphibians has been thought to be endoderm-dependent (Hall, 1999), and revisiting these studies with an eye to the model will be fruitful.
Couly and colleagues (2002) have demonstrated recently that ablation of the anterior most APE (i.e., rostral half of APE anterior to the caudal-most bend of the head-fold bay or their zone I stripe) in five- to six-somite chicks (Hamburger and Hamilton stage 8) resulted in the reduction or loss of the nasal bud, septum, capsules, and upper beak (the fate of the entirety of trabecular plate or of the distal midline rostral process was unclarified). Ablation of the next-most rostral section, zone II stripe (still rostral to the caudal-most bend of the head-fold bay), resulted in the loss of the body of MC: significantly, both the anterior-most rostral process of MC remained as did the articular portion of MC and the associated dermal bone, the supra-angular, remained. Ablation of the next strip, zone III, the first post head-fold, resulted in the loss of the splanchnocranial elements of the hinge, the articular and quadrate.
Couly and colleagues (2002) also used chick–quail xenografting experiments at the same five- to six-somite stage to demonstrate that the APE has the capacity to induce patterned ectopic bone and cartilage in the head (Couly et al., 2002). These experiments are at once both probative and problematic: patterned hard tissues were clearly and importantly induced, but the nature of the molecular, cellular, and morphogenic sequences and participants and events is not so clear. For example, although possibly infeasible, isotopic and heterotopic APE replacement grafts were not conducted, nor was any evidence of a complete proximal to distal (e.g., proximal articular to distal mdBA1 midline rostral process) induction of elements demonstrated. In APE rotation experiments, hinge elements such as the articular or quadrate were never found ectopically forming at the distal midline; in fact, the result of all such grafts examining polarity appears to be that hinge region elements emanate from the hinge region regardless of their direction. Furthermore, a case can be made that the elements induced were either hinge or cap in nature but not both. Ultimate understanding of such experiments will have to entail a fuller molecular and cellular characterization of the induced CNC and surface ectoderm, of the effects on and by endogenous endoderm, and of the APE region involved. Indeed, in seeking to add to the fate maps the cephalic ectoderm generated by Couly and Le Douarin (1990) and, furthermore, to identify proximate sources for the oral ectodermal Fgf8 expression, Haworth et al. (2004) extirpated sections of the anterior foregut endoderm and found that this led to the loss of ectodermal Fgf8 expression.
While regionalization of the APE is patent, the mechanisms underlying the establishment of this regionalization are not (Graham, 2001). One of the first signs of such regionalization is the establishment of endodermal–ectodermal contacts, beginning with the buccopharyngeal membrane and running through the series of pharyngeal plates. What establishes the number and nature of these endodermal–ectodermal contacts has been a long-standing, salient question (e.g., see Spemann, 1938; Goodrich, 1958; Waddington, 1962; Graham, 2001). Certainly, unbalanced levels of RA have been shown to affect these contacts by effecting the pharyngeal endoderm. Among the enzymes responsible for retinol conversion to RA are the members of the retinaldehyde dehydrogenases (Raldh1, Raldh2, and Raldh3); CYP enzymes subsequently degrade RA. RA binds to an array of nuclear receptors (RARs, RXRs) that then mediate transcriptional activities, cellular behaviors, and development. Temporospatially balanced levels of RA are critical to normal development: either too much (leading to such defects as microcephaly, anteroposterior neuraxial defects, spina bifida, encephaloceles, exencephaly, hydrocephaly, micrognathia, maxillomandibular hypoplasia, anterior BA fusions, cleft lips and palates, deformed or absent outer ears, microphthalmia or exophthalmia, and cardiovascular and aortic defects) or too little (leading to loss caudal pharyngeal structures, facial clefting) result in severe developmental defects (reviewed in Morriss-Kay, 1993; Soprano and Soprano, 1995; Brickell and Thorogood, 1997; Collins and Mao, 1999).
Although much of the early work seeking mechanisms for RA mediated defects in the head sought explanations in the neural plate and CNC, more recent work, driven in part by more detailed understanding of the expression of RA converting enzymes, has concentrated on the AME and APE. With regard to the head development, loss of RA-mediated signaling has been experimentally studied in several systems, including in vitamin A-deficient quail, and murine gene knockouts of RA receptors (RARs and RXRs), and converting enzymes (Raldh1/2/3). Notably, these studies demonstrate a pronounced, selective sensitivity to the loss of RA by the caudal pharyngeal endoderm, in particular the pouches and those structures that require their development (Niederreither et al., 1999; 2000, 2003; Wendling et al., 2000; Begemann and Meyer, 2001; Quinlan et al., 2002; Vermot et al., 2003). In such experiments, however, the buccopharyngeal membrane and the first pouch, cleft, and arch form.
Conversely, ectopic RA (or RAR agonists) has significant effects on the anterior ectodermal development, including on anterior endodermal–ectodermal contacts (e.g., Lee et al., 1995; Kuratani et al., 1998; Escriva et al., 2002; Matt et al., 2003). Matt et al. (2003) showed that by using a RARβ specific agonist the BA defects associated with RA tetratogenesis were largely replicated; they also applied the agonist to mice carrying a RA-responsive element (RARE) that normally gives expression only up to the second pharyngeal pouch and found reporter expression throughout the APE. Thus, the lack of proper anterior endodermal–ectodermal contacts and development were correlated with abnormal, anterior RA signaling. Notably, Raldh2 has been shown to be expressed in the AME and, thus, many of the embryopathies associated must be considered in this context (Halilagic et al., 2003). In Xenopus, RA administration is known to lead to the deletion of the cement glands (Durston et al., 1989; Sive et al., 1990; Blitz and Cho, 1995; Conlon, 1995; Pannese et al., 1995; Gammill and Sive, 1997). Ectopic RA is further correlated with a down-regulation in the expression of Otx2 homologues in the anterior surface ectoderm and AME and pharyngeal endoderm of numerous species (Bally-Cuif et al., 1995; Pannese et al., 1995; Gammill and Sive, 1997; Kuratani et al., 1998; Hinman and Degnan, 2000). In Xenopus, ectopic Xotx2 will induce additional cement glands (Blitz and Cho, 1995; Pannese et al., 1995), even in the presence of ectopic RA (Gammill and Sive, 1997). In zebrafish, RA application also transforms the head skeleton (Ellies et al., 1997) Thus, RA appears to affect the development of both the endoderm and ectoderm of the early head, a state apparently that it shares with many other signaling molecules.
We have presented here what will likely be axiomatic to all, namely that the proper elaboration of signals and responses within the developing embryo is essential to the development of its skull. This process is not static and begins with the germination of the anterior ectoderm that includes both the neural plate and surface cephalic ectoderm, and involves the activity of the nodal organizer and its mesendodermal and pharyngeal endodermal descendants. We have further put the signaling systems known to generate pattern and polarity in the jaws in the context of these early influences on the developing skull.
Neontologists have long sought to understand whether there are universal developmental mechanisms underlying gnathostome autapomorphic structural traits and how bauplan fidelity and elaboration of design can be maintained and coordinated to generate the amazing diversity seen in cranial designs and forms. To address whether jaws are in fact all made in a like manner, one needs a starting point for the sake homologizing. To this end, we have laid out here a “hinge and caps” model that places the articulation, and subsequently the polarity and modularity, of the upper and lower jaws in the context of cranial neural crest competence to respond to positionally located epithelial signals. This model relies upon a system for the establishment of positional information where pattern and placement of the “hinge” is driven by factors common to the junction of the maxillary and mandibular branches of the first arch and of the “caps” by the signals emanating from the distal-most first arch midline and the lambdoidal junction (where the maxillary branch meets the frontonasal processes). The model dictates that the functional registration of jaws is achieved by the integration of “hinge” and “caps” signaling, with the “caps” sharing at some critical level a developmental history that potentiates their own coordination. An important aspect of the model is that the emphasis is on the identification and comparison of tissues acting in like inductive and responsive roles. These roles are postulated to be necessary to generate pattern and polarity in developing jaws; hence, the model does not emphasize particular players per se, but rather the effects categories of players have on the cellular behaviors necessary to generate functioning jaws. Moreover, in creating a foundation for comparison, it is designed to facilitate the detection and identification of heterotopic, heterochronic, and heterofacient changes during craniofacial development within and between taxa.
We have examined some of the evidential foundation for this model in mice as well as the robustness with which it can be applied to the examination of other taxa. As a developmentally based model, it carries with it the inherent capacity to generate testable predictions (even some that are ludicrously grand) about the developmental basis underlying the great diversity of skull forms. Thus, the developmental mechanisms yielding both the elaborate coupling and registration of the upper and lower jaws, as seen with the Bow Mouth Shark Ray (Rhina anklystoma; Fig. 8D) and the Ganges River Dolphin (Fig. 8G), or its distinct uncoupling, as is seen with the bulldog jaw (Fig. 8E) and the Babyrusa tusk (Fig. 8F), can be understood through the hinge and caps model. For instance, the generation of the exquisite matching of the peaks and valleys of the upper and lower jaws of Rhina anklystoma can be viewed in terms of the hinge and caps model. It is possible that the matching may be generated by an out-of-phase wave propagation, or interpretation, of a signal emanating from the hinge region to the upper and lower jaw primordia. Achievement of this phase shift could come though change in length, or the character, of one jaw primordia relative to the other or by a change in one of the potency of the countering midline signals; in either case, only a single change in the competence to interpret from a default system were both upper and lower jaw primordia received the same strength signal would be required. It is possible that the properties of the developmental mechanisms that constrain gnathostome development, thus ensuring bauplan fidelity, are also the very mechanisms that inherently permit the greatly varied elaboration of the bauplan. This, and other issues, are being addressed as modern neontologists seek to detect and identify heterotopic, heterochronic, and heterofacient changes during craniofacial development within and between taxa.
We thank Claudia Companucci for her assistance. We thank S. Zoltewicz and A. Peterson for the Oto skull, and our neontological collaborators, especially Abigail Tucker, Bethan Thomas, Clare Baker, Gail Martin, John Rubenstein, Kim Haworth, Licia Selleri, and Phil Crossley, for discussions on various points. M.J.D. was funded by ARCS and the Royal Society, UK.