The vertebrate face contains bones that differentiate from mesenchymal cells of neural crest origin, which colonize the median nasofrontal bud and the first branchial arches. The patterning of individual facial bones and their relative positions occurs through mechanisms that remained elusive. During the early stages of head morphogenesis, an endodermal cul-de-sac, destined to become Sessel’s pouch, underlies the nasofrontal bud. Reiterative outpocketings of the foregut then form the branchial pouches. We have tested the capacity of endoderm of the avian neurula to specify the facial skeleton by performing ablations or grafts of defined endodermal regions. Neural crest cells that do not express Hox genes respond to patterning cues produced regionally in the anterior endoderm to yield distinct skeletal components of the upper face and jaws. However, Hox-expressing neural crest cells do not respond to these cues. Bone orientation is likewise dependent on the position of the endoderm relative to the embryonic axes. Our findings thus indicate that the endoderm instructs neural crest cells as to the size, shape and position of all the facial skeletal elements, whether they are cartilage or membrane bones.
INTRODUCTION
Most of the head skeleton is formed by membrane and cartilage bones of neural crest origin. The paraxial mesoderm that contributes to the posteriormost region of the skull (including the otic capsule, the occipital, part of the sphenoid and post-orbital bones) covers only the midbrain and hindbrain. Cell lineage studies carried out in the avian embryo have revealed that, from an early developmental stage, the prosencephalon becomes covered by neural crest cells emigrating from the posterior diencephalon, mesencephalon and the first rhombencephalic segments. These cells are at the origin of the frontal, parietal, squamosal and basi-presphenoid bones, which belong to the neurocranium. The nasal bud, the first branchial arch (BA1), from which the nasal capsule, maxillary bone, lower jaw and tongue (entoglossum) skeleton develop, are populated by cells arising from the mesencephalic neural fold and from the two first rhombomeres (r1 and r2). The rest of the viscerocranium (i.e. the hyoid cartilage) originates from the crest of more caudal rhombomeres that colonizes branchial arches 2, 3, 4 (BA2, BA3, BA4) (Fig. 1) (Couly et al., 1996; Köntges and Lumsden, 1996).
One of the crucial issues related to head morphogenesis concerns the developmental mechanisms that underlie the patterning of various skeletal structures of the face, namely the bone shapes and sizes that exhibit a considerable variability in the vertebrate series.
Segmentation of the branchial arches close to the hindbrain into rhombomeres constitutes the first visible step in the patterning process of the facial and hypobranchial structures. The rhombencephalon becomes metamerized into eight rhombomeres (Vaage, 1969; Lumsden and Keynes, 1989) and an iterated series of endodermal pouches form in the pharynx. Specific Hox genes, as well as others, such as Krox20 (Wilkinson et al., 1989) and valentino (Moens et al., 1996), are known to be involved in hindbrain segmentation. The Hox genes expressed in the hindbrain belong to the four first paralogue groups, and the anteriormost limit of Hox gene expression corresponds to the boundary between r1 and r2 (Prince and Lumsden, 1994) (Fig. 1A). As a general rule, the rhombomeres and the neural crest cells that they produce express the same set of Hox genes, forming a ‘Hox code’; this rule holds for the crest cells that colonize BA2, BA3 and BA4 (Hunt et al., 1991). Thus, two domains can be distinguished in the neural crest cells that are at the origin of the facial and visceral skeleton: a rostral Hox gene non expressing domain (Hox negative), which gives rise to the membrane bones of the neurocranium, the nasal capsule, the maxillary bone and the lower jaw; and a Hox-positive domain, which yields the hyoid cartilages (except the entoglossum). It has been shown (Couly et al., 1998) that Hox-expressing neural crest cells transposed anteriorly to the Hox gene non-expressing domain fail to differentiate into cartilage and bone. By contrast, the neural crest cells of the Hox-negative domain transplanted posteriorly respond normally to the local cues and participate in the formation of normal hyoid cartilage while maintaining their Hox-negative status.
Evidence that Hox gene expression is involved in patterning the visceral skeleton has come from experiments carried out in the mouse, where some genes of the anterior paralogue groups were either mutated or misexpressed. Thus, in Hoxa2-null mutant mice, the neural crest cells of BA2 behave like their Hox-negative counterpart that populate BA1 and form pieces of the lower jaw skeleton (Rijli et al., 1993; Gendron-Maguire et al., 1993). Further studies have indicated that in normal development, Hoxa2 acts by inhibiting chondrogenesis in the rostral part of BA2 where no other Hox gene is expressed (Kanzler et al., 1998). This is in line with the deficit in chondrogenesis observed when Hox-positive neural crest cells are transplanted to a Hox-negative domain (Couly et al., 1998). These and other results (Schilling and Kimmel, 1994; Sechrist et al., 1994) support the notion already put forward on the basis of earlier transplantation experiments (Noden, 1983) that the neural crest cells themselves possess the intrinsic information required for determining the identity of the various bones and cartilages that form the facial skeleton.
It has to be emphasized, however, that other tissue components, the endoderm, mesoderm and superficial ectoderm, participate in facial and visceral arch morphogenesis and may influence neural crest development. It has been shown that neural crest cells can differentiate into cartilage only in the presence of pharyngeal endoderm, whereas ectoderm is crucial for membrane bone differentiation from ectomesenchymal cells (Le Douarin, 1982; Takahashi et al., 1991). The paraxial mesoderm, in which a discrete segmentation into somitomeres has been described, exerts an effect on neural crest cell migration and differentiation (Trainor and Krumlauf, 2000). It has recently been shown that the segmental characteristics of the endodermal pharyngeal pouches develop independently of the presence of neural crest cells (Veicht et al., 1999). Moreover, endodermal segmentation depends upon the activity of specific genes. For example, in vgo null mutant zebrafish, larval hindbrain segmentation proceeds normally but the endodermal gill slits do not form and the distinct streams of neural crest cells exiting from the rhombencephalon ventrally fuse and do not form the individual skeletal elements of the viscerocranium (Piotrowski and Nusslein-Volhard, 2000). This suggests that at least part of the information for patterning the face could originate from signals of endodermal origin.
We have further explored the origin of the developmental signals responsible for determining the shape of the individual facial bones, as well as their relative positions. We first established that the neural crest cells that form the bones and the cartilages do not themselves possess all the information necessary for patterning the facial skeleton. We then tested the capacity of the foregut endoderm to specify facial bone identity in the avian neurula by performing ablations or grafts of defined endodermal regions.
MATERIALS AND METHODS
Quail (Coturnix coturnix japonica) and chick (Gallus gallus) eggs from commercial sources were used throughout this study. Microsurgery was performed on embryos at 5 to 6 somite stages [ss; HH9 (Hamburger and Hamilton, 1951)], corresponding to approximatively 32 hours of incubation in a humidified atmosphere at 38°C.
Excision and grafts of the neural fold in the Hox-non expressing domain of the neural crest
Chick embryos were prepared for in ovo surgery as described (Teillet et al., 1998). Concurrently, stage-matched quail embryos were harvested and isolated in sterile PBS. The chick embryos were subjected to the surgical ablation of the bilateral neural folds from the mid-diencephalon down to the presumptive r2/r3 limit (Fig. 2A). In certain experimental series, different fragments of the quail neural fold were transplanted into the chick at the posterior diencephalic level (Fig. 2B-D). The fate map of the cephalic neural primordium established by Grapin-Botton et al. (Grapin-Botton et al., 1995) (Fig. 1A) served as a reference to determine the presumptive level of the various encephalic structures: anterior and posterior diencephalon and mesencephalon and rhombomeres (r) 1 to 8 (Fig. 1). Chimeric embryos were allowed to grow from embryonic day (E) 4 to E8, depending on the experimental series. In another set of experiments, we ablated unilaterally the diencephalic neural fold, leaving the contralateral side unperturbed. The space resulting from the excision was filled with a fragment of equivalent length taken from the quail neural fold at the diencephalic, posterior mesencephalic or rhombencephalic (r4-r6) level (Fig. 2E-G).
Preparation of the endodermal stripes
The experiments were carried out on chick embryos at 5-6 ss in ovo. Superficial ectoderm was incised on the right side of the neural tube from the rostral tip to the prospective level of r3 included and reflected laterally to expose the endoderm and paraxial mesoderm (Fig. 2H-J). Four transverse stripes of endoderm were defined corresponding to the levels of the diencephalon (stripe I), anterior (stripe II) and posterior mesencephalon (stripe III) and metencephalon (stripe IV), according to previous fate maps of the neural primordium (Couly and Le Douarin, 1987; Grapin-Botton et al., 1995). Thus, the pharyngeal endodermal pocket was divided into approximately equal (150 μm width) regions from its rostral end (zone I) to the level underlying the caudal end of r3 (zone IV) (Fig. 2I,J).
Endodermal stripes were cleared of paraxial cephalic mesoderm as completely as possible. However, a few adherent mesodermal cells may be transplanted together with the endodermal layer. In previous experiments (Couly et al., 1992) (G. Couly, unpublished) transplantations or excisions of definite regions of cephalic mesoderm in 5-6 ss chick and quail embryos never perturbed morphogenesis of the facial skeleton. We, thus, assumed that the effects observed after foregut manipulations were attributable to the endoderm (or to the endoderm together with the few mesodermal cells that may still adhere to it). The endodermal stripes that are either excised or grafted in these experiments must include the endodermal floor of the foregut. Experiments involving only the dorsal aspect of the foregut had no effect on head morphogenesis.
Manipulations of the foregut endoderm
The endodermal stripes already defined were first subjected to unilateral ablation on the right side of the embryo (Fig. 5A,D, Fig. 6A). Then, the ectoderm was replaced and the embryos were further incubated until E6 to E9 and their skeleton was observed. In some cases, embryos were treated at E5 for in situ hybidization.
In a second series, similar stripes of endoderm were removed from 5-6 ss quail embryos and grafted underneath the cephalic ectoderm on one side or under the floor of the endogenous foregut of the stage-matched chick embryos in the migration pathway of the cephalic neural crest cells. Their dorsoventral and anteroposterior (AP) axes were either maintained or inverted (Fig. 7A, Fig. 8A,F,G, Fig. 9A, Fig. 10A,D,G,J). The AP position of the graft with respect to the foregut transverse level of the host varied according to the experimental series being considered (Fig. 11A,C,E). The endogenous chick foregut was untouched in all these experiments.
Analysis of chimeras
The embryos were sacrificed from 3 hours after grafting up to E6 for histology, and at E8, E9 and E12 for whole staining of the skeleton. Embryos were processed for immunocytochemistry with the quail-specific monoclonal antibody (Mab) QCPN (Developmental Studies Hybridoma Bank), according to the method described by Catala et al. (Catala et al., 1996), for whole skeletal stains (Simons and van Horn, 1971), or for whole-mount in situ hybridization with the Pitx1 probe at E5 as described (Henrique et al., 1995) and on sections with the chicken Hoxa2 (Etchevers et al., 2001). Chick Hoxa2 was linearized with HindIII and transcribed using T3 (Prince and Lumsden, 1994). Chick PitxI was linearized with EcoR1 and transcribed with T7 (Lanctôt et al., 1997).
RESULTS
The Hox-negative neural crest possesses the exclusive capacity to generate the facial skeleton but does not carry the information to pattern it
Previous experiments based on the construction of quail/chick chimeras have shown that the facial skeleton is derived from neural crest cells originating from the posterior diencephalic, mesencephalic and anterior rhombencephalic (r1, r2) neural folds (Couly et al., 1996; Köntges and Lumsden, 1996) (Fig. 1). Moreover, ablation of fragments of the neural fold in the avian neurula was shown to be followed by a regulation originating from the anterior and posterior neural crest cells left in situ (Couly et al., 1996). We have explored the capacity of the rhombencephalic neural folds from the Hox gene expressing domain of the neural axis to replace cells of the Hox-negative domain and to generate facial bones. For this purpose, the bilateral extirpation of the neural folds from the presumptive level of the mid-diencephalon down to the r2/r3 limit was performed in 5-6 ss chick embryos: this operation was followed by a high level of mortality. The twelve (out of 104 operated) embryos that survived until E8, were smaller than the unoperated controls and lacked the facial structures normally visible at that stage, i.e. the nasofrontal and maxillo-mandibular buds (Fig. 3A-C).
In the next series of experiments, excision of the same neural fold segments was followed by the orthotopic bilateral graft of the posterior diencephalic neural fold from a stage-matched quail embryo. The chimeric embryos examined at E8 (n=4) exhibited normal development of their skeletal structures and facial morphology (Fig. 3D-F). Histological examination of the tissues after immunostaining with QCPN Mab revealed that the neural crest derived structures were exclusively made up of quail cells that had migrated from the implanted diencephalic neural folds (not shown). Thus, the diencephalic neural folds alone can generate enough neural crest cells to form the complete facial skeleton and head connective tissue normally derived from the diencephalic, mesencephalic and anterior rhombencephalic (r1, r2) crest. The same result was obtained when the posterior mesencephalic neural folds were implanted at the diencephalic level after complete removal of the crest down to r3 (Fig. 3G-I). In previous experiments (Couly et al., 1998), the mesencephalic neural folds grafted to r4-r6 level were also found to be able to replace the excised Hox-negative domain of the neural crest, thus showing that the level of the neural axis at which the competent neural fold fragments (of about 150 μm length) are implanted is not crucial [Fig. 5F,G by Couly et al. (Couly et al., 1998)]. By contrast, in the same experiments r4-r6 neural folds failed to generate a facial skeleton [Fig. 4G by Couly et al. (Couly et al., 1998)]. Therefore, the capacity of a 150 μm fragment of the neural fold from the non Hox-expressing domain to generate a complete facial skeleton depends neither upon its level of origin along the AP axis nor its position after grafting.
The fate of neural crest cells from either the diencephalic or mesencephalic levels was strikingly different when the quail neural folds were implanted at the diencephalic level into a chick embryo that had not been subjected to the complete cephalic neural fold excision as before. After orthotopic substitution of the posterior diencephalic neural fold by its quail counterpart, quail cells could be seen around the optic vesicles, and participating in the ciliary muscles, the corneal endothelium, the sclerotic and the cartilages of the nasal septum (Fig. 4A-D). They were also found as pericytes associated with blood vessels in the prosencephalic meninges and in the choroid membrane. This wide distribution of quail cells was regularly observed in all the embryos (n=6) examined at E6 and E8. It reflects the normal fate of the diencephalic mesectodermal neural crest cells.
When the same experiment was performed, using quail mesencephalic neural folds transplanted to the diencephalic level, the grafted neural crest cells contributed to the same structures as the diencephalic derived crest cells, as observed at E4 (n=3), E5 (n=2), E7 (n=2) and E8 (n=3).
The posterior rhombencephalic (r4-r6) neural fold from 5-6 ss quail implanted at the same diencephalic level into stage-matched chicks yielded neural crest cells that colonized the anterior cephalic area where they contributed to the prosencephalic meninges and pericytes. However, they did not participate in the sclerotic that was exclusively composed of chick host cells originating from the mesencephalon (Fig. 4E-H). These embryos were observed at E4 (n=3) and E5 (n=2). At later stages (E6, n=5; E8, n=8) quail cells were found lining peripheral nerves and forming the connective component of ocular muscles; these quail cells contained Hoxa2 transcripts, showing that they had maintained their original molecular identity while migrating within an environment where Hox genes are not activated (Fig. 4H,K). In the nasal bud, some quail cells formed abnormally dense aggregates but did not participate in the cartilage of the nasal septum that developed, by compensation, from endogenous neural crest of more posterior origin (Fig. 4I-K).
In conclusion, the cartilages and bones of the face originate from the anterior domain of the neural crest in which the genes of the Hox-clusters are not activated. After complete ablation of this domain, the facial skeleton fails to develop, owing to the incapability of the neural crest cells of more posterior origin to replace them. The neural folds of the anterior Hox-negative domain possess, by contrast, a high capacity of regulation: a 150 μm long fragment of the neural fold (corresponding to about a quarter of the total length of the neural folds that normally generate the facial skeleton) is sufficient to ensure the normal development of the head. Under the experimental conditions just described, the territory extending from the diencephalon down to r3 can be considered as an ‘equivalence group’ in which each fragment of the neural fold is endowed with a virtually similar development potential.
These results indicate that the information for the specific morphogenesis of the facial bones is not carried by the neural crest cells themselves, as previously suggested (Noden, 1983). We then investigated the possible role of the foregut endoderm in this process. In a first step, transverse stripes of foregut endoderm were removed and the consequences of this operation on the facial skeleton were recorded.
Ablation of endodermal stripes of foregut endoderm
Stripe I
Ablation of stripe I was performed on the right side of 14 embryos, of which eight could be recovered at E8. Four were anatomically normal. Four had an abnormal nasal capsule in which the nasal septum was deflected to the left side (not shown). In the next series, stripe I was removed on both sides. This resulted in the strong reduction (or absence) of the nasal bud and later on of the nasal capsule, nasal septum and upper beak (n=5) (Fig. 5A-C).
Stripe II
Ablation of stripe II was performed on 70 embryos (Fig. 5D); 12 survived until E8-E9, in all of which Meckel’s cartilage was missing in the first branchial arch on the operated side. In one out of 12, part of the articular cartilage was also missing (Fig. 5E,F).
One control embryo was sacrificed at E5 for whole-mount in situ hybridization with the Pitx1 probe. Pitx1 is strongly expressed by the stomodeal ectoderm and the mesenchymal cells of the proximal part of the BA1 at E5 (Fig. 5G). In the embryo subjected to removal of stripe II endoderm on the right side, Pitx1 was normally expressed in the left branchial arch. By contrast, Pitx1 expression was strongly reduced on the operated side while the corresponding BA1 was underdeveloped (Fig. 5H). The remaining small distal zone of Pitx1 expression may correspond to the presence of a distal short piece of Meckel’s cartilage observed in most of the Alcian Blue stained embryos that were subjected to this operation (see Fig. 5E,F).
Stripe III
Ablation of stripe III was performed on 20 embryos. Five embryos were observed at E9. All were devoid of the articular, the quadrate and the proximal part of Meckel’s cartilages (Fig. 6A,B). Ablation of stripes II and III together (n=3) resulted in the total absence of the entire skeleton of the first branchial arch (not shown).
Stripe IV
Ablation of stripe IV was performed on 12 embryos; five were observed at E8 and showed a strong reduction of the articular and quadrate cartilages (not shown).
In conclusion, removal of transverse stripes of endoderm from the extreme tip of the foregut (corresponding to the level of the prosencephalon) did not perturb the development of the jaw but affected that of the nasal capsule and upper beak. Ablation of the endoderm of zone II, III and IV, corresponding (respectively) to the levels of the mesencephalon (II, III) and of r1-2 (IV) prevented the morphogenesis of Meckel’s (zone II and III) and of articular and quadrate cartilages (zone III-IV). Moreover, when cartilaginous bones failed to develop, nearby membrane bones, which normally ossify later directly from the neural crest mesenchyme of the nasofrontal bud or of BA1, did not form either. Thus, in the absence of the nasal septum, the other bones of the nasal capsule are missing. Without the quadrate, articular and Meckel’s cartilages, the squamosal, pterygoid, quadratojugal, angular, supra-angular, opercular and dentary bones of the jaw did not appear.
The absence of skeletal structures after ablation of defined regions of the endoderm could have been due either to mechanical prevention of neural crest cell migration or to the absence of a specifically localized patterning effect of the endoderm on the ectomesenchyme. To clarify this, we implanted defined regions of 5 ss quail foregut endoderm into the head of stage-matched chicken embryos. In such chimeras, donor cells can be identified immunohistologically with the quail-specific QCPN Mab. In these chimeras, migrating cephalic neural crest cells, exiting from the Hox-negative domain of the encephalic vesicles, are subjected to the influence of ectopic pieces of endoderm, even though the host endoderm is left intact.
Ectopic grafts of defined foregut endodermal areas in the migration pathway of cephalic neural crest cells induce the formation of specific cartilages and bones
Grafts of stripe II quail endoderm were made laterally to the mesencephalon of chicken hosts, thus overlying their endogenous counterpart (Fig. 7A). At E3, the grafts formed vesicles embedded within the host neural crest-derived mesenchyme of BA1 (Fig. 7B). At E5, in situ hybridization showed that Pitx1, usually expressed only within the stomodeal ectoderm and mesenchyme of BA1 (Lanctôt et al., 1997), was reproducibly induced by the stripe II endodermal grafts in an ectopic, subocular position (Fig. 7C, n=3). Consistent with the assumption that BA1 identity is specified in the mesenchyme surrounding the endodermal graft, a supernumerary Meckel’s cartilage formed near the eye, as seen at E8 (n=2), E9 (n=3) and E12 (n=1). These ectopic cartilages, examined at E6 in two chimeras, developed from host neural crest cells (not shown). The corresponding supernumerary membrane bones of the lower jaw also formed (Fig. 7D).
A bilateral stripe II of quail endoderm was implanted underneath and around endogenous stripes II (Fig. 8A-C) (n=5). An ectopic lower beak was induced in the host (Fig. 8D). It contained not only the two Meckel’s cartilages but extra lower jaw membrane bones that differentiated in contact to the supernumerary cartilages (Fig. 8E). If the stripe II endodermal fragments were implanted lateral to the endogenous foregut and neural tube [as represented in Fig. 8F,G (n=1)], an extra lower beak with the corresponding skeleton and the two part of entoglossum developed hanging lateral to the lower beak of the host (Fig. 8H,I).
Likewise, a graft of ectopic stripe III endoderm in addition to its host counterpart (Fig. 9A) led to the formation of vesicles of graft origin lateral to the mesencephalon by E3 and the condensation of chick host-origin quadrate cartilage around the graft at E6 (Fig. 9B,C). A supernumerary membranous squamosal bone was associated with the ectopic quadrate at E9 (Fig. 9D,E). Thus, precise regions of the pharyngeal endoderm instruct neural crest cells in BA1 to become the cartilaginous bones of the jaws, and the patterning of adjacent membranous bones. These cartilages are necessary for the subsequent patterning of adjacent membranous bones. The instruction given by ectopic endodermal grafts, placed in the migratory path of BA1 neural crest cells, supersedes that of other local signals for differentiation and patterning.
Facial bone orientation relative to the embryonic axes also depends on an endodermally derived information
When a piece of quail endoderm, including both stripes II and III was implanted above the chicken stripes II and III in its normal AP and mediolateral (ML) orientation (Fig. 10A), a supernumerary lower jaw developed above that of the host, as expected (Fig. 10B,C, Fig. 8). Rostrocaudal inversion of the graft (Fig. 10D) gave rise to a duplicate lower jaw developing toward the back of the head (Fig. 10E,F). Meckel’s cartilage resulting from a 90º rotation of the endodermal graft to the left (anterior becoming medial, posterior becoming lateral, Fig. 10G), pointed toward the top of the head, at the expense of the ipsilateral eye (Fig. 10H,I). After a 90º rotation of the graft to the right (Fig. 10J), the induced jaw was oriented caudally (Fig. 10K,L). The rostrocaudal position of the early foregut endoderm underlying BA1 therefore translated into the orientation of the distoproximal axis of the jaw.
The foregut endoderm patterning cues influence only the non-Hox-expressing cephalic neural crest cells
As mentioned before, two distinct populations of neural crest cells differentiate into the components of the facial and visceral skull. A rostral, Hox-negative population gives rise to all of the membrane bones of the brain case, face and jaws, as well as the cartilaginous bones of the face; a caudal, Hox-positive population yields the hyoid cartilage (except the entoglossum) (Couly et al., 1996). To test whether the endoderm involved in patterning the Hox-negative neural crest can influence the Hox-expressing, chondrogenic rhombencephalic crest cells, foregut stripes II (Fig. 11A) or III from quail embryo were implanted adjacent to r7-r8 of 5 ss chick recipients. Small nodules of cartilage of chick origin formed around the grafts (n=5) in the host cervical region by E9 (Fig. 11B). However, when the quail anterior mesencephalic neural fold accompanied the endodermal graft (Fig. 11C,E), recognizable cartilages and bones rudiments could be observed at E9. Although reduced in size if compared with their normal endogenous counterpart, Meckel’s, articular and quadrate cartilages, and some associated membrane bones could be identified at the level of the upper neck (Fig. 11D,F,G; n=3). Note that the shape and size of the skeletal pieces that form in these ectopic locations are as close to their normal counterpart as when the grafts are placed in the facial region (c.f. the experiments described above).
Interestingly, supernumerary bones were strictly of quail origin (not shown) and therefore constructed by the grafted non-Hox-expressing neural crest cells. Supernumerary stripe III endoderm, grafted with a posterior mesencephalic neural fold from the same quail donor, induced quadrate (and squamosal) formation in the neck from donor neural crest cells (Fig. 11E-G). Localized information for constructing specific facial skeletal components hence resides in the anterior pharyngeal endoderm. However, only the ‘equivalence group’ of cephalic neural crest cells that do not express Hox genes can interpret this information.
DISCUSSION
The foregut endoderm exerts a patterning activity on skeletogenic neural crest cells of the anterior Hox-negative domain
The experiments reported above demonstrate that the foregut endoderm of the avian neurula displays a regional activity essential in specifying the identity and orientation of the neural crest-derived bones forming the vertebrate facial skeleton.
The anteriormost region of the endoderm, although fated to form Sessel’s pouch and thus to degenerate, is crucial for specification of the nasal septum and later the nasal capsule, the ethmoid bone, and the upper beak. Stripes of endoderm corresponding to the transverse level of the mesencephalon (stripe II and III) and metencephalon (stripe IV) are necessary for the morphogenesis of the bones derived from the maxillary and mandibular buds forming the first branchial arch (Table 1). The foregut endoderm patterning capacities lie in the lateroventral aspect of the foregut endodermal layer. Thus, the size, shape and orientation of the cartilages forming in the facial anlage of the vertebrate embryo seem to be prefeatured in the endoderm of the early head primordium. Although much progress has been made in the understanding of the mechanisms that underlie cell differentiation, the questions raised by morphogenesis have remained largely unaddressed and the fact, disclosed here, that a definite region of foregut endoderm precisely specifies the shape and orientation (with respect to body coordinates) of every given bone of the face is striking.
Another key point resulting from the present work is that, whatever the nature of the endoderm-derived signals, they trigger a response only from the anterior domain of the neural crest in which no genes of the Hox clusters are expressed. More posterior Hox-expressing neural crest cells are unable to generate facial bones when they are transplanted into a Hox-negative environment. As shown before (Couly et al., 1998), fragments of neural fold heterotopically transplanted along the AP axis maintain their Hox code, while yielding crest cells that remain closely associated in large groups. Thus, when Hox-negative neural crest cells are transferred to the level of BA3-BA4, they participate in the formation of the hyoid cartilages without being induced to express genes of the Hox clusters. Inversely, posterior Hox-positive neural crest cells transplanted anteriorly follow the migration routes taken by the endogenous neural crest arising from the same level and continue to express their specific Hox code, although they have reached ectopic locations. However, in this position, they are unable to generate the skeletal structures specific to their novel environment. It is interesting to note that experiments of the same type performed in the mouse have revealed that maintenance of Hox code in transposed neural crest cells seem to depend upon a community effect. If posterior cells are transferred anteriorly in small number, they may lose their Hox gene expression when mixed up with the endogenous Hox-negative cells. The eventual fate of such cells (i.e. whether they die or differentiate according to their novel position) is not known, as the embryos could not be cultured long enough to be examined at the appropriate stage (Trainor and Krumlauf, 2000).
We have also investigated the possible patterning capabilities of the other – non neural crest and non endoderm-derived constituents of the branchial arches. By using a similar microsurgical approach, defined fragments of the ectoderm covering the cephalic region of the five-somite stage avian neurula, which correspond to previously defined ‘segments’ called ectomeres (Couly and Le Douarin, 1990) were either removed or ectopically transplanted. In no case did these experiments perturb the development of the facial skeleton (G. Couly, unpublished). Similar investigations carried out with the cephalic paraxial mesoderm showed that, at these early stages, the various AP regions of the cephalic mesoderm (Couly et al., 1992) are similarly interchangeable. Thus, in strong contrast to the foregut endoderm, neither the ectoderm nor the mesoderm of the early neurula display patterning properties on the skeletogenic process that leads to the formation of the facial bones (G. Couly, unpublished).
The cells of the Hox-negative domain of the cephalic neural crest behave as an equivalence group
In the Hox-negative domain, the cephalic neural crest cells exhibit a large range of plasticity:
(1) their proliferative capacity can be considerably increased in certain of the experimental situations described above. For example, if a large part (three quarters) of the neural fold that normally participates in head morphogenesis is removed, the remaining segment that is left in situ can regulate the deficiencies caused by the operation and construct the complete skeletal and connective tissue components of a normal head
(2) individual areas can be identified in the diencephalic, mesencephalic and metencephalic neural crest that are normally devoted to participate in a definite set of skeletal structures. Nevertheless this regionalization does not correspond to a strict determination of the cells to a given fate. Limited territories of the anterior Hox-negative neural fold have the capacity to regenerate the whole cephalic neural crest and to yield the entire facial skeleton. Thus, although the anterior domain of the neural crest exhibits a certain degree of specificity in being the only one able to form the facial skeleton, it does not contain the information required to specify each of its constitutive bones and cartilages.
The fact that Hoxa2–/– mice crest cells originating from r4 and colonizing BA2 form first arch structures shows that, even in this posterior region, the cephalic neural crest can display the same developmental capabilities as their more anterior counterpart (provided they do not activate Hoxa2). This may also imply that the pharyngeal endoderm of the second branchial pouch is as able, like that of the first one, to specify bones of the lower jaw if it is in contact with Hoxa2-negative neural crest cells. The fate map of the foregut endoderm and the determination of the spatial areas participating in each branchial pouch in the pharynx has not yet been established. The exact contribution of zones II to IV to pharyngeal pouches at later stages, when facial morphogenesis is under way, is not yet known, but work is being carried out to clarify this.
The skeletogenic capacities of neural crest cells depend upon Hox gene expression not only in neural crest cells but also in the other cellular constituents of the branchial arches
The experiments described in this and other articles (Gendron-Maguire et al., 1993; Rijli et al., 1993; Couly et al., 1998; Kanzler et al., 1998; Pasqualetti et al., 2000) indicate that the effect of Hox-selector genes on the morphogenesis of neural crest derivatives involves cellular interactions between the neural crest cells and the environment within which they differentiate. This is particularly clear for the duplication of the lower jaw, as it can be obtained not only by the inactivation of Hoxa2 in the mouse but also by the anteroposterior transfer of the mesencephalic neural crest to the BA2 level (Noden, 1983; Couly et al., 1998). By contrast, if Hox-negative neural crest cells are transferred to the levels of BA3 or BA4, formation of an ectopic lower jaw does not ensue. However, formation of BA1 specific cartilages can be obtained even in this posterior location, provided that Hox-negative neural crest cells are transferred together with this appropriate region of foregut endoderm (see Fig. 11). It is characteristic that BA2 endoderm does not express Hoxa2 or any other Hox gene, whereas that of BA3 and BA4 do. It thus appears that the lower jaw skeleton can develop only if Hoxa2 remains unactivated in both the neural crest cells and their immediate environment. By contrast, forced expression of Hoxa2 in Xenopus embryo induces the homeotic transformation of BA1 into BA2 skeleton (Pasqualetti et al., 2000). In such a case, not only BA1 neural crest cells but also the surrounding tissues, including the endoderm, do express Hoxa2 from a stage preceding crest cell emigration up to complete facial morphogenesis. This experiment supports the notion that Hoxa2 gene expression is not compatible with facial skeleton development. Similarly, in the chick embryo, homeotic transformation of BA1 into BA2 structures (although at a more discrete level than in Xenopus) could be obtained when the constitutive tissues of BA1 were all subjected to Hoxa2 overexpression. Thus, a bifurcation of tongue skeleton could be seen and was assumed to arise form modified BA1 skeletogenic neural crest cells. Sometimes this extra bone was fused to the remains of Meckel’s cartilage, thus further supporting the view that it arose as a result of homeosis. By contrast, this did not occur when only the neural crest was targeted before migration. In the latter case, first arch skeletal structures were strongly reduced or absent (Grammatopoulos et al., 2000). In the experiments that involved the anterior transposition of a Hox-positive (r4-r6) fragment of the neural fold, the neural crest cells were surrounded by an environment where no Hox gene was expressed. Although yielding normal neural derivatives, these neural crest cells were unable to form recognizable cartilage structures. Apoptotic figures observed in certain zones where the grafted neural crest cells were aggregated (S. Creuzet, unpublished) support the contention that, in the absence of an appropriate Hox-positive environment, these cells undergo apoptosis. The homeotic transformations of BA1 into BA2 cartilages, observed in the experiments carried out in Xenopus by Pasqualetti et al. (Pasqualetti et al., 2000) and on chick by Grammatopoulos et al. (Grammatopoulos et al., 2000), seem therefore to indicate that overexpression of Hoxa2 exerts, in these circumstances, an antiapoptotic role on the crest-derived chondrogenic precursors.
About the nature and mode of action of the endodermal signal(s) on neural crest cells
The nature of the signal(s) arising from the endoderm is so far unknown but it is interesting to see that the presumptive first arch endoderm (zone II) triggers in the neural crest the expression of Pitx1, a gene normally activated in BA1 ectoderm and neural crest cells during the skeletogenic process leading to lower jaw formation. Moreover, the fact that not only the shape but also the proximodistal and anteroposterior polarity of the skeleton is dictated by the endoderm in our ectopic graft experiments, is in line with the observation of Veitch et al. (Veitch et al., 1999) that the polarity of the branchial pouch endoderm is established in the embryo in the absence of neural crest cells. This was deduced from the observation that the expression of genetic markers such as Bmp7, Fgf8 and Pax1 occurs in the same regions of the pouch endoderm whether neural crest cells immigrate or not. Our experiments show that these polarities are already determined within the foregut endoderm at the early neurula stage, i.e. well before the branchial pouches are formed.
Acknowledgements
The authors warmly thank Drs Andrew Lumsden and Jacques Drouin for providing them with Hoxa2 and Pitx1 probes, respectively. They are grateful to Marie-Françoise Meunier, Evelyne Bourson, Francis Beaujean, Michel Fromaget and Sophie Gournet for their participation in preparing the manuscript. This work was supported by the CNRS and a grant from the ARC. Sophie Creuzet received fellowships from la Fondation des Treilles and from les Amis des Sciences.