Most of the head skeleton is derived from the neural crest. Such is the case for the facial and hyoid cartilages and bones (Couly et al., 1993; Le Douarin and Kalcheim, 1999). Investigations using quail–chick chimeras have shown that the skeletogenic domain of the neural crest, extending from the level of the mid-diencephalon down to that of the fourth somite pair (i.e., rhombomere 8, r8), is divided in an anterior region from which arise the cells forming the nasal capsule and superior beak as well as the lower jaw skeleton (including the entoglossum and the rostral part of the basihyal) and a posterior region that yields the remaining of the hyoid bone (Couly et al., 1996; Köntges and Lumsden, 1996). Both regions produce mesectodermal cells able to differentiate into connective tissues, cartilages, and bones, but differ as far as expression of Hox genes is concerned: Hox gene expression does not take place in the anterior domain (Hox-negative; H-), whereas the caudal one is the site of expression of Hox genes of the four first paralogous groups (Hox-positive; H+). Transposition of neural fold fragments along the neural axis showed that if Hox-expressing neural crest cells of the posterior domain are rostrally transplanted to replace the Hox- neural fold, no facial skeleton develops (Couly et al., 2002; Creuzet et al., 2002). Moreover, forced expression of Hox genes (Hoxa2, Hoxa3, Hoxb4) in the anterior domain of the cephalic neural crest prevents more or less completely the development of the facial skeleton (Pasqualetti et al., 2000; Grammatopoulos et al., 2000; Creuzet et al., 2002). In contrast, neural fold from the Hox-negative anterior domain translocated posteriorly yields crest cells able to participate in the formation of normal hyoid cartilages (Couly et al., 1998).
Another trait of the anterior Hox-negative domain of the neural crest is that it behaves as an “equivalence group” because, after ablation of the complete facial skeletogenic domain of the crest (extending from the mid-diencephalon down to r3), a small part of this domain (representing approximately one fourth of the entire Hox-negative skeletogenic crest) is sufficient to regenerate a complete facial skeleton as well as the connective tissue of neural crest origin taking part in the construction of the face (Couly et al., 2002). This means that the neural crest cells arising from this domain do not possess the information required to specify the various pieces of the facial skeleton. We have shown that this information is provided to the neural crest cells by signals arising from the foregut endoderm (Couly et al., 2002).
The anteriormost region of the foregut is involved in the patterning of the nasal septum. Posteriorly, the transverse region of the pharyngeal endoderm corresponding to the level of the mesencephalon and to r1 and r2 (at the five- to six-somite stage; 5–6 ss) specifies the lower jaw cartilages (articular, quadrate, Meckel's) as well as the entoglossum and the anterior part of the basihyal. This finding was demonstrated in two ways: (1) by extirpation experiments in which definite endodermal areas were removed; (2) by grafting supernumerary endodermal regions of quail foregut in the migration pathway of the cephalic neural crest cells of chick embryos (Couly et al., 2002).
In the present work, we have extended our investigations to more posterior regions of the foregut endoderm in embryos at 6–9 ss corresponding to the level of the mid-rhombencephalon (r4–r5) and the posterior rhombencephalon (extending from r6 down to the anterior part of r8) from which the neural crest cells populate the second, the third, and fourth branchial arches (BA; Fig. 1A). We found that the lateroventral endoderm plays a role in the specification of this crest-derived ectomesenchyme of the second to fourth branchial arches to yield the formation of the hyoid bone.
Patterning of the Hyoid Bone Is Controlled by the Foregut Endoderm
The hyoid bone is formed by the juxtaposition of several pieces of cartilage distributed along the anteroposterior axis in the following order: (1) the entoglossum formed by a pair of cartilages medially located; (2) the basihyal and the basibranchial (which as well as the entoglossum form the tongue skeleton) situated in the sagittal plan of the body; (3) two lateral branches from the articulation between basihyal and basibranchial form the ceratobranchials and epibranchials (Fig. 1G). This complex bone is formed by cephalic neural crest cells originating from the two Hox-negative and Hox-positive domains mentioned above (Fig. 1H).
The demonstration that the endoderm is involved in the patterning of these bones was brought about by ablation and grafting experiments. Ablation of zone II and III as reported in Couly et al. (2002) resulted in the truncation of the axis of the hyoid cartilage within the basihyal. The anterior part of the basihyal and the entoglossum were missing (see Fig. 5D–F of Couly et al., 2002).
Extirpation of the lateral aspect of stripe V (corresponding to the level of r4–r5; number of experiments n1 = 64, number of surviving and examined embryos at embryonic day [E] 8 n2 = 4; number of embryos showing the phenotype described n3 = 4) led to the truncation or complete absence of the ceratobranchial (Fig. 2A,B), whereas that of stripe VI suppressed the development of the epibranchial (n1 = 64; n2 = 4; n3 = 3; Fig. 2C,D). When both the lateral zones V and VI were removed together (n1 = 91; n2 = 6; n3 = 4), the ceratobranchial and epibranchial were absent (Fig. 2E,F). One of the embryos did not show any phenotype. The mortality of the embryos after these experiments is higher than when endoderm excision concerns more rostral zones. This finding may be related to the fact that the anterior intestinal portal corresponds to the area where the heart is forming at that stage.
Homotopic grafts of the corresponding endodermal stripes induce the formation of supernumerary hyoid cartilages.
Quail endodermal ventromedial zones II and III were ventrally engrafted to the host foregut at the level of the chick anterior mesencephalon. These grafts led to the formation of the entoglossum associated with a small cartilage rod corresponding to the anterior part of the basihyal that developed in the chick's lower jaw (n1 = 85; n2 = 12; n3 = 9; Fig. 3A–C). In three embryos, no phenotype was observed, probably due to loss of the grafted tissue.
Homotopic graft of quail endodermal ventromedial stripes III and IV underneath the host ventral foregut corresponding to the level of the presumptive mesencephalon of the chicken host generated the formation of a cartilaginous piece that we interpret as being a rostral portion of basihyal fused with the endogenous basihyal (n1 = 46; n2 = 5; n3 = 3; Fig. 3D–F).
The ventromedial area of the more posterior endodermal stripes V and VI taken from the quail donor foregut, were homotopically grafted ventrally to their chick endogenous counterparts; they induced rods of basihyal (caudal portion) and basibranchial cartilages, which in some cases fused with the host hyoid cartilages (Fig. 4A–F; n1 = 121, n2 = 10, n3 = 5). Note that these cartilaginous rods are never associated with membrane bones. The graft of lateral regions of the ventral foregut endoderm (see right or left of zones V and VI on Fig. 1D) induced supernumerary cartilages corresponding the ceratobranchial and epibranchial as shown in Figure 4G–L (n1 = 156; n2 = 19; n3 = 12).
CONCLUSIONS AND DISCUSSION
Role of Foregut Endoderm in Specifying the Hyoid Cartilage
These experiments show that the posterior foregut endoderm (as it stands in the 6 to 9 ss avian embryo) exerts a patterning activity upon the neural crest cells that form a large part of the hyoid cartilages. In contrast to the facial skeleton that can only be constructed by Hox-negative neural crest cells, the hyoid cartilages originate from both the Hox-negative and Hox-positive domains of the cephalic neural fold (Fig. 1H).
The results reported in Couly et al. (2002) and in the present article are summarized on Figure 5A where the endodermal regions specifying the various cartilages forming the primary facial and visceral chondrogenic skeleton are indicated. It has to be underlined that the limits of each zone so defined are not precisely determined, hence, the overlaps represented in the figure.
Three anteroposterior levels can be distinguished in the primary facial skeleton: (1) the nasal septum (derived from the nasal process); (2) the lower jaw and the anterior hyoid cartilages (entoglossum and anterior part of basihyal, derived from BA1); (3) the posterior hyoid cartilages (basibranchial, ceratobranchials, epibranchials, derived from BA2–BA3–BA4).
These skeletal pieces are respectively specified by the foregut of the following transverse regions: zone I of foregut endoderm specifies the nasal septum; zones II, III, and IV of the ventral endoderm specify the first branchial arch skeleton: Meckel's, quadrate, articular, and anterior hyoid cartilages (entoglossum and anterior basihyal); zone IV of endoderm has a patterning activity that concerns part of the basihyal and the quadrate; zone V of endoderm specifies the posterior part of the basihyal and ceratobranchials; zone VI of endoderm specifies the basibranchial in its midline and epibranchials laterally.
The role of the endoderm was demonstrated by experiments involving extirpation of definite regions of the lateroventral foregut endoderm. If the same experiments are restricted to the dorsal side of the pharynx, no incidence on the development of the facial and visceral skeleton are noticed (data not shown).
In addition, we show that the information transferred from the ventral endoderm to the neural crest cells are regionalized not only along the anteroposterior axis but also along the mediolateral axis: the ventromedial endoderm specifies (from anterior to posterior) the entoglossum and most of Meckel's cartilage, the basihyal, basibranchial, and parts of the ceratobranchials (Fig. 5A). The lateral endoderm bears the information to construct the articular, quadrate cartilages and the cerato- and epibranchials as schematically reported on Figure 5A.
Our conclusions are also based on the finding that the definite regions of endoderm that have been thus defined are able to induce the formation of specific cartilages when grafted in the pathway of migration of the cephalic neural crest cells. Such grafts of additional stripes of endoderm generated the formation of specific supernumerary pieces of cartilage as a result of a chondrogenic differentiation of the neural crest cells that are in contact with the transplant. We could thus localize the endodermal epithelium corresponding to the level of r4 and r5 (zone V) and demonstrate that, if transplanted ectopically, it induces the formation of the medial segments of the hyoid bone (i.e., the posterior part of the basihyal) and the ceratobranchials. In normal development, these cartilages are actually built up with neural crest cells from r4–r5 (Couly et al., 1996; Köntges and Lumsden, 1996). The endoderm located at 8–9 ss at the transverse level of r6 to r8 (zone VI) specifies the basibranchial in its medial region and the epibranchials laterally when ectopically transplanted as it does in normal development (Fig. 4).
Hox Genes and Development of the Vertebrate Head
In our previous work (Couly et al., 2002), we have demonstrated that the rostral foregut endoderm, located at the level of the diencephalon down to the metencephalon (r1–r2), is able to instruct the corresponding Hox-negative cephalic neural crest cells to generate the facial skeleton. Furthermore, Hox-positive neural crest cells grafted anteriorly are unable to respond to the anterior foregut endodermal cues, and no facial skeleton develops. In contrast, Hox-negative neural crest cells transplanted posteriorly can generate a normal hyoid cartilage (Couly et al., 1998). The Hox-positive neural crest cells that colonize the posterior branchial arches (BA2–BA4)—although they are not able to respond to the cues of anterior foregut endoderm—are patterned by signals emanating from the posterior foregut endoderm. Thus, Hox-negative neural crest cells respond equally well to signals arising from anterior and posterior foregut endoderm. Whereas, Hox-positive neural crest cells can express their skeletogenic potencies only if instructed by the posterior foregut endoderm. These and other experiments point to the fact that the requirements for cartilage and bone formation are different for Hox-positive and Hox-negative neural crest cells. The Hox-negative domain that corresponds to the prosencephalon, mesencephalon, and metencephalon has expanded considerably during evolution of the vertebrate phylum together with the development of forebrain to which it provides both skeletal protection and vascularization (Couly et al., 1993; Le Douarin and Kalcheim, 1999; Etchevers et al., 1999, 2001).
Early Segmentation of the Foregut Endoderm
Our experiments together with those of Couly et al. (2002), demonstrate that the patterning activity of the endoderm is distributed along the rostrocaudal axis in a segmental manner that corresponds to the migration flows of neural crest cells that colonize the facial processes and branchial arches (Fig. 5B,C).
Early in development, the foregut endoderm exhibits a metamerization that becomes evident when the facial processes and branchial arches develop. Such a metamerization also exists in the ectoderm, which forms the “ectomeres” covering the branchial arches as proposed by Couly and Le Douarin (1990). The ectomeres, as well as the stripes of endoderm defined in this and in our precedent article (Couly et al., 2002), correspond to the same levels of the encephalic vesicles as they stand at the early stages of neurogenesis.
Thus, a cryptic segmentation of the vertebrate head is already present in the neural plate and in the ectoderm and endoderm of the head, early in development, before it is morphologically expressed in the branchial arches and in the developing brain (Rubinstein et al., 1994; Lumsden and Krumlauf, 1996).
It is interesting to notice that, in zebrafish, null-mutation for van gogh without perturbing the segmentation of the hindbrain, resulted in a deficit in the segmentation of the endoderm, which fails to form discrete reiterated outpockets leading to the formation of branchial arches (Piotrowski and Nüsslein-Volhard, 2000). In these mutants, the pharyngeal cartilages fuse with each other anteriorly and are absent posteriorly, meaning that segmentation of the branchial skeleton primarily depends on signal of endodermal origin.
Casanova mutants are defective for the endodermal derivatives and for the ventral viscerocranium. Grafting experiments of wild-type endoderm to these mutant embryos rescued the development of the head skeleton. The involvement of foregut endoderm as a source of morphogenetic molecules required for crest-derived skeleton has been particularly well established for the development of the posterior branchial arches. The role of FGF3 as a signalling molecule in this process has been demonstrated (David et al., 2002).
In conclusion, these experiments confirm that the neural crest cells, as they start migrating from the neural primordium, do not carry the information to specify the morphology and pattern of the various bones and cartilages composing the facial skeleton. The ventral foregut endoderm plays an early role in this process. The signalling molecules that mediate the interactions taking place between endoderm and crest cells are still to be discovered. FGFs are certainly critical for the development of facial and hypobranchial structures. Where do they act in the cascade of interactions required to build up these structures is not yet known. Moreover, later in development, other tissue interactions take place to complete the skeletogenic process. Among them, an interplay between ectoderm and ectomesenchyme has been demonstrated (Schneider and Helms, 2003), showing that the complex process of facial and visceral skeletogenesis involves several successive steps that require further investigations.
The experiments were performed on chick and quail embryos developed from eggs from commercial sources. The embryos were at 6–9 ss (stage 8 to 9 according to Hamburger-Hamilton (1951; HH8 to HH9). The principle of these experiments was the same as in Couly et al. (2002) but concerned the lateroventral side of more posterior levels of the foregut endoderm at 7 to 9 ss and its ventromedial aspect at 6 to 9 ss.
In our previous work, we had defined four (I to IV) transverse stripes of endoderm (in 5–6 ss chick and quail embryos) corresponding, respectively, to the level of the posterior diencephalon (stripe I), anterior mesencephalon (stripe II), posterior mesencephalon (stripe III), and anterior rhombencephalon (rhombomeres 1 and 2, r1–r2, stripe IV). In the presently described experiments, similar endodermal stripes corresponding to the mid-rhombencephalon (r4–r5, stripe V) and the posterior rhombencephalon (r6–r8, stripe VI) were defined in embryos at 6–7 ss (see also Couly and Le Douarin, 1987; Grapin-Botton et al., 1995, for the definition of the brain levels at these early stages; Fig. 1A,D,F).
Extirpation of Endodermal Stripes in Chick Embryos In Ovo
The endodermal stripes were exposed on the dorsal side of the chick embryos in ovo: the superficial ectoderm, cut laterally to the neural folds limiting the neural plate was externally reflected, thus exposing the dorsal foregut endoderm. The lateroventral foregut endoderm is the tissue that carries the patterning information for bone morphogenesis as demonstrated in Couly et al. (2002). Excision of endodermal stripes was carried out from dorsal to ventral (Fig. 1A). In a first experimental series, the lateral aspects of stripes V, VI, and stripes V+VI, were unilaterally removed in the chick embryo in ovo (Fig. 2A,C,E). Then the embryos subjected to surgery were incubated until they reached 8–12 days of embryonic life (E8–E12). The mesial limits to extirpate the lateral zone correspond to the level of the lateral aspect of the neural tube.
Supplementation of Quail Endodermal Stripes Into Chick Embryos
To explore the specifying capacities of the endoderm located in the ventral midline of the foregut, definite regions of the ventral foregut were isolated from quail embryos lying on their dorsal side in a dish (Fig. 1B–D). The ventral ectoderm and mesoderm were carefully discarded (Fig. 1B,C) to expose the foregut endoderm in which the territories designated in Figure 1 were cut off and transversely transplanted to the ventral foregut of normal stage-matched chicks (Figs. 3A,D, 4A,E,G,I,K).
The supernumerary pieces of cartilage that resulted from graft experiments were identified according to morphologic criteria based on the shape and articular surfaces of the endogenous structures. The basihyal consists in a short rod of cartilage with two rounded tips; the basibranchial is recognizable as a short rod with a proximal rounded tip and a distal spindle-shaped one; the ceratobranchial appears as an elongated and curved cartilage with a nodular proximal tip; in that respect, it differs from the epibranchial, which is an elongated spindle-shaped cartilage with a flat rostral tip.
Analysis of the embryos was performed from E8 to E12 by using Alizarin red–Alcian blue staining to evidence the skeleton. Figure 1G shows the normal facial skeleton in a E9 chick embryo.
We thank Marie-Françoise Meunier for preparing the manuscript, Sophie Gournet for illustrations, and Francis Beaujean and Michel Fromaget for photographs. S.C. received a fellowship from the Fondation pour la Recherche Médicale.