An endothelin-1 switch specifies maxillomandibular identity

Abstract

Articulated jaws are highly conserved structures characteristic of gnathostome evolution. Epithelial-mesenchymal interactions within the first pharyngeal arch (PA1) instruct cephalic neural crest cells (CNCCs) to form the different skeletal elements of the jaws. The endothelin-1 (Edn1)/endothelin receptor type-A (Ednra)→Dlx5/6→Hand2 signaling pathway is necessary for lower jaw formation. Here, we show that the Edn1 signaling is sufficient for the conversion of the maxillary arch to mandibular identity. Constitutive activation of Ednra induced the transformation of upper jaw, maxillary, structures into lower jaw, mandibular, structures with duplicated Meckel's cartilage and dermatocranial jaws constituted by 4 dentary bones. Misexpression of Hand2 in the Ednra domain caused a similar transformation. Skeletal transformations are accompanied by neuromuscular remodeling. Ednra is expressed by most CNCCs, but its constitutive activation affects predominantly PA1. We conclude that after migration CNCCs are not all equivalent, suggesting that their specification occurs in sequential steps. Also, we show that, within PA1, CNCCs are competent to form both mandibular and maxillary structures and that an Edn1 switch is responsible for the choice of either morphogenetic program.

The gnathostome skull is characterized by the presence of articulated jaws consisting of the palatoquadrate dorsally and the Meckelian cartilage (MC) ventrally. These prehensile cranial units derive from the maxillary and mandibular processes of the first pharyngeal arch (PA1). Most cartilaginous and dermatocranial derivatives of PA1 are formed by Hox-negative cephalic neural crest cells (CNCCs) emigrating from the mesencephalic and anterior rhombencephalic neural folds (1–6). Before migration, CNCCs lack the topographic information needed to unfold the morphogenetic process of the jaws (7), and receive these instructions from epithelial structures of PA1. In particular, surgical deletion and grafting of different parts of the foregut endoderm has shown that this epithelium harbors instructive signals directing CNCC to form polarized jaw structures such as MC (4, 7, 8). The molecular nature of these instructive signals remains to be determined, but experimental evidence indicates that FGFs, bone morphogenetic proteins (BMPs), endothelin-1 (Edn1), and Sonic hedgehog are involved (9). In particular, loss of Edn1-endothelin receptor type-A (Ednra) signaling (10, 11) results in the homeotic transformation of lower into upper jaw structures, indicating the role of this signaling pathway in the specification of mandibular structures (12, 13). This transformation is accompanied by the down-regulation of Dlx5 and Dlx6 (10, 12–14), and is strongly reminiscent of that observed after targeted inactivation of these 2 genes (15, 16). Dlx homeobox genes, vertebrate orthologues of Drosophila Distal-less, have a fundamental role in the specification of the dorsoventral patterning of PA1 derivatives (17, 18). Edn1 is expressed in the epithelium and mesodermal core of the mandibular part of PA1, whereas Ednra is broadly expressed in CNCC-derived ectomesenchyme of the head [supporting information (SI) Fig. S1] (10, 19, 20), suggesting that all CNCCs might be competent to respond to Edn1 signaling. To test this hypothesis, we generated mice in which Edn1 was knocked-in into the Ednra locus by recombinase-mediated cassette exchange (RMCE) in embryonic stem cells (21), resulting in Ednra constitutive activation throughout the head and pharyngeal mesenchyme. Here, we demonstrate that constitutive activation of Ednra induces the transformation of maxillary structures into mandibular structures with duplicated MCs and dermatocranial jaws constituted by 4 dentary bones. A similar transformation is obtained by forcing the expression of Hand2, a downstream target of the Edn1 pathway, in the Ednra domain. Skeletal transformations are accompanied by neuromuscular remodeling. Thus, within PA1, CNCCs are competent to form both mandibular and maxillary structures and that an Edn1 switch is responsible for the choice of either morphogenetic program.

Results

Constitutive Activation of Ednra Induces the Transformation of Upper Jaw Structures to Lower Jaw Structures.

We generated Edn1 knocked-in mice by RMCE in embryonic stem cells (Fig. S2) (21). All heterozygous (Ednra^Edn1/+) mice died perinatally with severe craniofacial defects. At E18.5, Ednra^Edn1/+ mice were externally characterized by loss of vibrissae, open eyelids, and an anterior shift of the ear position (Fig. 1 A and B). Skeletal preparations revealed that the maxillary components of Ednra^Edn1/+ mice were replaced by a second set of mandibular elements (Fig. 1 C–J). Notably, a dentary-like bone carrying an incisor was formed in the maxillary region (Fig. 1 K–N), resulting in jaws constituted by 4 mandibles (Fig. 1 C–H; Fig. S3 and Fig. S4). The tip of this ectopic dentary was fused to the premaxilla (Fig. 1 L and N). The 2 dorsal mandibular-like structures did not form a midline symphysis, possibly due to hindrance by the frontonasal process (Fig. 1 E–H; Fig. S3 and Fig. S4). MC and associated dermal bones (gonial and ectotympanic) were also duplicated in Ednra^Edn1/+ mice (Fig. 1 I and J). The incus was fused to the malleus and morphologically changed to a malleus-like structure in mutants (Fig. 1 I and J). Mandibular components were only mildly affected with shortening of coronoid and condylar processes and hyperplasia of the anterior part of the tympanic ring (Fig. S4). The palatine, pterygoid, lamina obturans, and ala temporalis were missing or severely deformed, the squamosal was reduced and divided in 3 to 4 parts as in Dlx1/2 double mutant mice (22), whereas midline structures including the nasal capsule, vomer, and presphenoid were normal (Fig. 1 G and H; Fig. S5 A–C). The premaxilla, the frontal process of maxilla, and the lacrimal bone were maintained in Ednra^Edn1/+ mice. The stapes, a derivative of the second PA (PA2), was deformed and fused to the styloid process (Fig. S4).

Fig. 1.

Transformation of maxillary components in Ednra^Edn1/+ mice. (A and B) Facial appearance of E18.5 wild-type (A) and Ednra^Edn1/+ (B) mice. Vibrissae are absent in mutant mice that have open eyelids (arrows) and an anterior shift of the ear position (arrowheads). (C–L) Bone and cartilage elements of E18.5 wild-type (C, E, G, I, and K) and Ednra^Edn1/+ (D, F, H, J, and L) skulls. (C and D) Lateral views of wild-type and Ednra^Edn1/+ craniofacial skeletons. (E and F) The Ednra^Edn1/+ mutant shows mirror-image duplication of the mandibular elements (dentary, MC, and malleus) at the expense of the maxillary elements (maxilla, jugal, squamosal, and incus). (G and H) Caudal views of wild-type and Ednra^Edn1/+ craniofacial skeletons. The dentaries are removed to show the normal and transformed maxillae. Large parts of the palatine, pterygoid, lamina obturans, and ala temporalis are missing or severely deformed in the Ednra^Edn1/+ mutant. (I and J) Ectotympanic and gonial bones are also duplicated in association with the ectopic MC in Ednra^Edn1/+ mutants. (K and L) Caudal views of the premaxilla-maxilla junction. The tip of the transformed maxilla-dentary in the Ednra^Edn1/+ mutant contains an ectopic incisor and fuses to the premaxilla. (M and N) Parasagittal sections of E18.5 wild-type (M) and Ednra^Edn1/+ (N) mice stained by Mallory trichromic. The Ednra^Edn1/+ mutant shows an ectopic incisor in addition to the orphotopic 1. Fusion between the premaxilla and transformed maxilla-dentary is observed in the mutant. at, ala temporalis; bs, basisphenoid; dnt, dentary; etm, ectotympanic; fmx, frontal process of maxilla; gn, gonial; hy, hyoid; in, incus; jg, jugal; lo, lamina obturans; ma, malleus; mx, maxilla; pl, palatine; pmx, premaxilla; ps, presphenoid; ptg, pterygoid; sq, squamosal; tg, tongue; UI, upper incisor; vbf, vibrissae follicle; vm, vomer; *, ectopic structure.

Maxillomandibular Identity Correlates with Regional Edn1-Ednra Signaling Activity.

To analyze the effect of Edn1/Ednra signaling on the dorsoventral specification of PA1, we then compared chondrocrania of wild-type, Ednra^lacZ/lacZ and Ednra^Edn1/+ embryos at E15.5. As previously described, Ednra^lacZ/lacZ embryos showed loss of MC and duplication of the ala temporalis (Fig. 2 A and B; Fig. S5 A and B) (21), indicating a ventral-to-dorsal transformation of PA1. By contrast, Ednra^Edn1/+ embryos showed a duplicated MC in place of the ala temporalis (Fig. 2C; Fig. S5 C–I), indicating that an opposite dorsal-to-ventral transformation had taken place. The duplicated MC extends beyond the ectopic dentary bone forming laterally to it, as in case with their orthotopic counterparts (Fig. S5H). The ectopic MC was discontinuous to the transformed incus-malleus in some mutants (Fig. S5H), but was continuous to it in others (Fig. S5I). Thus, Edn1/Ednra signaling is necessary to specify mandibular identity within PA1 and CNCCs colonizing both the maxillary and the mandibular arch are competent to respond to Edn1 signaling.

Fig. 2.

Comparison of craniofacial skeletons among E15.5 Ednra-null (Ednra^lacZ/lacZ) (A), wild-type (B) and Edn1-misexpressing (Ednra^Edn1/+) (C) mice. Activated Edn1/Ednra signaling (colored in pink) correlates to the formation of MC and associated mandibular structures. at, ala temporalis; dnt, dentary; hy, hyoid; in, incus; ma, malleus; mx, maxilla; pl, palatine; pmx, premaxilla; *, ectopic structure.

Edn1-Induced Skeletal Transformations Are Accompanied by Neuromuscular Remodeling.

Histological analysis of heads of Ednra^Edn1/+ embryos at E18.5 (Fig. 3 A and B) not only confirmed the duplication of the dentary bone and MC, but also revealed the presence of an ectopic profound muscle mass juxtaposed to the normal masseter muscle. This supernumerary muscle, which connected the 2 dentary bones, could be interpreted as a duplicated masseter muscle. To reinforce this notion, neurofilament immunostaining of cranial nerves of wild-type and Ednra^Edn1/+ embryos at E10.5 (Fig. 3 C and D) showed that the motor root of the trigeminal nerve, which normally enters the mandibular arch and innervated the masseter, divided in 2 symmetric branches which innervated both the mandibular and the transformed maxillary arch of mutant embryos.

Fig. 3.

Histological and whole-mount neuromuscular analysis of Ednra^Edn1/+ mice. (A and B) Frontal sections of E18.5 wild-type (A) and Ednra^Edn1/+ (B) mice stained by Mallory trichromic. Constitutive activation of Edn1/Ednra signaling results not only in the transformation of maxillary skeletal elements, but also in the appearance of new muscular components associated to the ectopic dentary, which can be interpreted as a duplicated masseter muscle. (C and D) Cranial nerves of E10.5 wild-type (C) and Ednra^Edn1/+ (D) embryos stained by antineurofilament antibody. Ectopic branches of the mandibular (arrow) and facial (arrowheads) nerves are indicated. dnt, dentary; mm, masseter muscle; pl, palatine; tg, tongue; gV, trigeminal ganglion; V₁, ophthalmic nerve; V₂, maxillary nerve; V₁, mandibular nerve; VII, facial nerve; IX, glossopharyngeal nerve; *, ectopic structure.

Mandibular Arch-Specific Genetic Program Is Activated in the Maxillary Arch by Edn1 Signaling.

The dorsal-to-ventral transformation of PA1 induced by ectopic Edn1 signaling in the maxillary region suggested that a shift in the genetic program specifying jaw identity might have taken place. To test this hypothesis, we examined the expression of genes involved in PA specification. Dlx2 was expressed in the mesenchyme of PA1 and PA2 in both E10.5 wild-type and Ednra^Edn1/+ embryos (Fig. 4 A and B). Fgf8 was similarly expressed in wild-type and Ednra^Edn1/+ embryos in the ectoderm around the maxillomandibular junction (Fig. 4 C and D). Ednra^Edn1/+ embryos exhibited an extension of the territory of Dlx5/Dlx6 expression to the maxillary (dorsal) region of PA1 (Fig. 4 E–H). Consistently, E10.0 Ednra^Edn1/+ embryos carrying the mI56i-lacZ transgene, in which the mI56i enhancer from the intergenic region of the Dlx5/6 cluster drives lacZ expression (23), revealed transgene expression extending to the dorsal region of PA1 and PA2 (Fig. S6), supporting the notion that the mI56i enhancer may mediate the Dlx5/Dlx6 induction by the Edn1/Ednra signaling even in dorsal ectopic locations. The expression of Dlx3 (Fig. 4 I and J), Pitx1 (Fig. 4 K and L), and Goosecoid (Fig. 4 M–P), mandibular markers down-regulated in Dlx5/Dlx6-null and in Edn1/Ednra-null embryos (10, 12, 15, 16), was activated in the maxillary region of Ednra^Edn1/+ mutants. Hand2, a downstream target of Dlx6 (24), was also up-regulated in the mutant maxillary arch, although the expression level appeared lower than in the mandibular arch (Fig. 4 Q–T). Ectopic Hand2 expression in the maxillary arch was observed along the nasolacrimal groove with the highest intensity in the medial region abutting the frontonasal process (Fig. S7).

Fig. 4.

Gene expression analysis of the pharyngeal arches in Ednra^Edn1/+ embryos. (A–T) Whole-mount in situ hybridization for Dlx2 (A and B), Fgf8 (C and D), Dlx5 (E and F), Dlx6 (G and H), Dlx3 (I and J), Pitx1 (K and L), Goosecoid (M–P), and Hand2 (Q–T) in E10.5 wild-type (A, C, E, G, I, K, M, O, Q, and S) and Ednra^Edn1/+ (B, D, F, H, J, L, N, P, R, and T) embryos. Arrows indicate the junction of maxillary and mandibular arches. Yellow arrowheads indicate ectopic gene expression in the maxillary arch. hy, hyoid arch; md, mandibular arch; mx, maxillary arch.

These gene expression switches between the maxillary and the mandibular arch are opposite to those observed in Edn1/Ednra-null embryos. Collectively, they indicate that CNCCs colonizing PA1 can respond to Edn1 signaling by activating the genetic program specifying mandibular identity both in dorsal (maxillary) and ventral (mandibular) PA1.

Forced Expression of Hand2 Partially Mimicked Edn1-Induced Skeletal Transformation.

To explore the role of the molecular pathway of jaw specification downstream to the Edn1/Ednra-Dlx5/6 signaling, we used RMCE to replace Ednra with Hand2 (Fig. S2) and examined the phenotype of PA1-derived structures. Because no chimeric Ednra^Hand2/+ mice survived after birth, we directly analyzed the phenotype of chimeras. Among 93 chimeras, 53 (57%) exhibited mirror-image polydactyly, as previously observed by forced Hand2 expression in the limb bud (25, 26). Chimeras with polydactyly also had craniofacial abnormalities (Fig. 5A; Fig. S8 and Fig. S9). At birth, Ednra^Hand2/+ chimeras resembled Ednra^Edn1/+ mice in lacking vibrissae follicles and having open eyelids (Fig. 5A). Maxillary structures were often transformed into dentary-like bones fused to the premaxilla and carrying a small incisor at the distal end (Fig. 5 B–D; Fig. S8). A MC-like rod-shaped cartilage was present in the maxillary region of 66% (29/44) E15.5 (Fig. 5 F–I) and E16.5 (Fig. S9) Ednra^Hand2/+ chimeras. However, in Ednra^Hand2/+ chimeras, the ectopic MC-like cartilage often coexisted with a smaller and deformed ala temporalis and proximal structures of the mandibular and maxillary arches (gonial, ectotympanic, and squamosal) were often malformed or absent rather than duplicated (Fig. 5E; Fig. S9). The premaxilla, the frontal process of maxilla, and the lacrimal bone were maintained in Ednra^Hand2/+ chimeras (Fig. 5C; Fig. S8). The skeletal abnormalities present in Ednra^Edn1/+ and in Ednra^Hand2/+ mutants were never observed in control Ednra^Neo/+ chimeric mice (n = 8) (Fig. 5B; Fig. S8).

Fig. 5.

Morphological transformation in Ednra^Hand2/+ ES cell-derived chimeric mice. (A) Representative surface appearance of E18.5 Ednra^Hand2/+ chimeric mice. Mutant mice demonstrate loss of vibrissae follicles, open eyelids (arrows), and polydactyly (arrowheads). (B–E) Craniofacial bone and cartilage elements of E18.5 Ednra^Neo/+ ES cell-derived (B) and Ednra^Hand2/+ ES cell-derived (C–E) chimeric mice. Whereas Ednra^Neo/+-chimeric mice show normal skeletal morphology (B), Ednra^Hand2/+-chimeric mice demonstrate transformation of the maxilla into dentary-like bone (C), which fuses to the premaxilla and contains an ectopic incisor at the tip (D). The condylar and angular processes of the dentary, gonial, ectotympanic, and squamosal bones were often malformed or absent (E). (F–I) Craniofacial skeletons of E15.5 wild-type (F and H) and Ednra^Hand2/+-chimeric (G and I) mice. Ectopic formation of MC-like, rod-shaped cartilage is present often with small and deformed ala temporalis in Ednra^Hand2/+-chimeric mice. The malleus, incus, stapes, and styloid process are deformed. agp, angular process; at, ala temporalis; cdp, condylar process; crp, coronoid process; dnt, dentary; etm, ectotympanic; fmx, frontal process of maxilla; gn, gonial; hy, hyoid; in, incus; jg, jugal; ma, malleus; mx, maxilla; pmx, premaxilla; sp, styloid process; sq, squamosal; st, stapes; UI, upper incisor; *, ectopic structure.

Discussion

In the gnathostome craniofacial development, environmental signals have central roles in the fate determination of CNCCs and regional patterning. CNCCs constituting PAs and adjacent facial prominences are highly plastic at early stages and their identities can be separately specified by signaling cues (27–29). Here, we show that a single Edn1 signal can convert the maxillary arch into a mandibular fate, indicating a dorsal-to-ventral transformation within PA1. Previously, we and other groups have reported a ventral-to-dorsal transformation by Edn1/Ednra-null mutations in mice (12, 13) and zebrafish (30, 31). These findings indicate that an Edn1 switch is necessary and sufficient for the determination of mandibular identity.

The significance of endoderm- and ectoderm-derived epithelia as sources of instructive signals for facial skeletal patterning has also been demonstrated (7, 32, 33). Edn1 is produced in the ventral aspect of the pharyngeal endoderm (20), a region which can induce the formation of MC and associated membranous bones (7), suggesting that Edn1 could be the endodermal signal that specifies ventral PA identity. Notably, distal-to-proximal polarity within PA1 was maintained in Ednra^Edn1/+ mutants. Determination of intra-PA polarity is essential for modular development of upper and lower jaws (9, 34). The nested expression of Dlx genes seems to have a central role in determining the dorsoventral polarity of PA1, whereas the proximodistal polarity of the maxillary and mandibular regions appears to depend on the integration of signals from the proximal “hinge” and distal “caps” territories of PA1 (9). Thus, dorsoventral identity and intraregional polarity may be separately regulated within PA1.

The fact that constitutive activation of the Edn1/Ednra signaling pathway confers mandibular identity to maxillary CNCCs indicates that CNCCs within PA1 are competent to respond to Edn1. By contrast, the effects of Ednra constitutive activation, such as Dlx5/Dlx6 induction, were mostly restricted to PA1, although Edn1expression took place throughout the head mesenchyme and in other regions of the embryo including the limb bud (data not shown). This fact suggests that the competence to respond to Edn1/Ednra signaling may be limited to a subpopulation of Hox-negative CNCCs, although the possibility that Edn1 peptide ligands may not be efficiently produced outside PA1 cannot be completely excluded. The limited, but not negligible, response of PA2 CNCCs to the constitutive activation of Endra suggests that the absence of Hox expression is not necessary for Edn1 competence. It is worth remembering that targeted disruption of Hoxa2 results in the transformation of PA2 into a mandibular PA1 (35), suggesting that the expression of Hox genes might have an active role in determining the morphogenetic consequence of Edn1 signaling to CNCCs.

The transformation of the maxillary skeleton was accompanied by changes in the associated musculature and corresponding innervation. Previous reports have suggested a tight communication between CNCCs and cranial mesodermal cells involved in patterning the craniofacial neuromuscular system (3, 5, 36). Thus, in Ednra^Edn1/+ mutants, changes in CNCCs may cause the muscle duplication and concurrent changes in regional axon guidance signals. Indeed, the changes in innervation are likely to be due to changes in such guidance signals in the transformed region rather than changes intrinsic to motor neurons, because Ednra is not expressed in neurons in the E9.5–10.5 neural tube (21); therefore, Edn1 does not directly act on them. These neuromuscular changes may affect muscle attachment to the bone and subsequently the formation of coronoid processes. Analysis of differential gene expression in Ednra^Edn1/+ mutants and RMCE-induced knock-in experiments used in this study may give a clue to the mechanism underlying the region-specific muscle patterning and axon guidance.

Inactivation of Edn1-dependent Hand2 expression in the ventrolateral PA regions results in malformations of MC and associated bones (37). Our results confirm the importance of the Edn1/Ednra→Dlx→Hand2 regulatory cascade for mandibular specification as forced expression of Hand2 in the Ednra domain induces the partial transformation of upper jaw structures into lower jaw elements. Apparently, incomplete penetrance of the phenotype is probably due to different contribution of ES cells to each chimeric mouse. The coexistence of the MC-like cartilage and the ala temporalis in Hand2-knock-in chimeras may reflect such variation, and suggest that Hand2-positive and negative cells might behave differently within the same maxillary region. It is also noteworthy that the gonial and ectotympanic bones were malformed in Ednra^Hand2/+ chimeras. This fact indicates that Hand2 may antagonize the genetic program for the formation of the proximal mandibular structures.

Together with the grafting experiments of Couly et al. (7), in which small aliquots of premigratory crest could rescue the deletion of all CNCCs, our data would suggest a multistep specification of CNCCs. Before migration, CNCCs constitute an equivalence group. The first step in their specification would be the definition of large craniofacial territories, such as the Edn1-responsive PA1 and the Edn1-unresponsive frontonasal prominence. Successive signaling from restricted epithelial regions would then specify specific subregions such as the maxillary and mandibular territories of PA1.

Materials and Methods

Mice.

Mice carrying the Ednra^lacZ allele (21) and mI56i-lacZ transgenic mice (23) have been previously described. To obtain mice carrying the Ednra^Edn1 (Edn1-knock-in) or Ednra^Hand2 (Hand2-knock-in) allele, we performed RMCE on the Ednra^neo/+ ES cells, in which an exchangeable floxed site was introduced into the Ednra locus as previously described (21). Briefly, PCR-amplified fragments encoding the ORF of mouse Edn1 and Hand2 cDNA were introduced into the knock-in vector p66–2272 containing multiple cloning sites between lox66 and lox2272 (38). The resultant plasmids were transfected into Ednra^neo/+ ES cells with Cre-expressing adenovirus (39). Targeted ES clones were injected into ICR blastocysts to generate chimeras. To obtain mice carrying the Ednra^Edn1 allele, 2 independent male germ-line chimeras were crossed with ICR females. Ednra^Edn1/lacZ mice, derived by crossing the germ-line chimeras and Ednra^lacZ/+ females heterozygous for Ednra-null mutation, exhibited the typical Ednra-null phenotype, confirming that Ednra was responsible for the effect of ectopic Edn1 signaling. All of the chimeras from Ednra^Hand2 ES cells were died before or at the time of birth; therefore, chimeric embryos from Ednra^Hand2 ES cells and wild-type ICR blastocysts were prepared at different stages for phenotype analysis. Genotypes were determined by PCR by using primers specific for RMCE-mediated recombination. All of the animal experiments were performed in accordance with the guidelines of the University of Tokyo Animal Care and Use Committee.

Skeletal Staining.

Alizarin red/alcian blue staining was performed, as previously described (40).

Histological Analysis.

Embryos were fixed in Bouin's solution and embedded in paraffin. Sections were then subjected to Mallory trichromic staining.

Whole-Mount Immunostaining.

Whole-mount neurofilament staining was performed and visualized with 3–3′-diaminobenzidine tetrahydrochloride/NiCl₂, as previously described (41).

In Situ Hybridization.

Whole-mount in situ hybridization was performed as described previously (42). Probes for Hand2 (43) and Goosecoid (44) were generously provided by D. Srivastava (University of California, San Francisco, CA) and G. Yamada (Kumamoto University, Kumamoto, Japan), respectively. Other probes were prepared by RT-PCR.

β-Galactosidase Staining.

LacZ expression was detected by staining with X-gal for β-galactosidase activity. Whole-mount staining was performed as previously described (41) with minor modifications.