Sonic hedgehog signalling from foregut endoderm patterns the avian nasal capsule

The nasal capsule is dorsoventrally divided into two parts: the upper part,the ectethmoid, serves olfaction and is composed of the lamina cribosa, the crista galli apophysis and the conchae. The lower part, the mesethmoid, is a thick cartilage bar extending from the corpus sphenoidalis to the rostral extremity of the nose (Fig. 1A-B′). In the avian embryo, the mesethmoid constitutes the cartilage primordium of the upper beak.

Lineage experiments have shown that Hox-negative cephalic neural crest cells (CNCCs) emigrating from the prosencephalic and the anterior mesencephalic neural folds give rise to the nasal capsule and to first pharyngeal arch (PA) structures (Couly et al., 1993; Creuzet et al.,2002; Kontges and Lumsden,1996; Noden, 1992; Trainor and Tam, 1995). By contrast, Hox-positive NCCs, which are only found posteriorly to rhombomere 2,do not contribute to the facial skeleton, but generate more-posterior structures of the embryo (Couly et al.,1996; Kontges and Lumsden,1996; Santagati and Rijli,2003). Premigratory, Hox-negative CNCCs behave as an equivalence group and lack the topographic information needed to give rise to the different structures of the craniofacial skeleton(Couly et al., 2002). We have shown by surgical deletion and grafting of different parts of the foregut endoderm that this epithelium harbours the instructive signals(Couly et al., 2002; Noden, 1992; Ruhin et al., 2003), although their molecular nature remains to be determined. Among candidate cues, sonic hedgehog (Shh) is expressed in the most anterior part of the endoderm, endoderm zone I (EZ-I), of the early chick embryo(Brito et al., 2006). Later,incoming CNCCs express the Shh receptor patched 1, indicating a potentially active Shh signalling (Jeong et al.,2004). The importance of Shh signalling in the control of different aspects of craniofacial development has been demonstrated in several models including zebrafish (Wada et al.,2005), mouse (Chiang et al.,1996; Jeong et al.,2004) and chicken (Brito et al., 2006; Cordero et al.,2004; Helms et al.,1997; Hu et al.,2003). Furthermore, human syndromes with nasal malformations, such as foetal alcohol syndrome, Smith-Lemli-Opitz syndrome and holoprosencephaly,have been associated with defective SHH signalling(Herman, 2003; Traiffort et al., 2004; Yamada et al., 2005).

Fertilised eggs were obtained from Morizeau Farms, France (chicken, Gallus gallus) or Cailles de Chanteloup Farms, France (quail, Coturnix coturnix japonica) and incubated at 38°C in a humidified atmosphere for approximately 32 hours to reach the 5-somite stage HH8+(Hamburger and Hamilton, 1992; Teillet et al., 1998). Embryos were collected and dissected at room temperature in phosphate-buffered saline(PBS).

Immunoperoxidase detection was performed as previously described(Couly et al., 2002). Quail nuclei were detected with the quail-specific monoclonal antibody QCPN at 1/500(obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by The University of Iowa, Department of Biological Sciences, Iowa City, IA 52242). Chick and quail Gli1 proteins were detected using rabbit polyclonal antibody #2553S (Cell Signaling Technologies)at 1/200. Shh transcripts were detected by in situ hybridisation as previously described (Couly et al.,2002). Whole-mount skeletons were visualised according to standard staining protocols using Alcian Blue for cartilage and Alizarin Red for bone(Couly et al., 2002).

Experiments were carried out in ovo on windowed chick embryos at the 5-somite stage HH8+ (Couly et al.,2002). Two bilateral excisions were first performed on the superficial ectoderm on each side of the neural tube at the level of the prosencephalon and down to the mesencephalon (see Fig. S1 in the supplementary material). A very fine curved tungsten microknife was then passed under the neural tube and the notochord, resulting in their separation from the dorsal foregut endoderm. A transversal incision was then performed on the neural plate at the level of the posterior mesencephalon. The neural epithelium was reclined rostrally in order to gain access to the endoderm. The ventral endoderm was separated from the ventral ectoderm by passing a very fine curved tungsten microknife between the two tissues. The rostral-most attachment between the ventral endoderm and the anterior neural fold was then cut. At this point, the ventrolateral endoderm was free of any attachment and we could then easily remove the EZ-I by performing a transversal incision at a distance of ∼150 μm from its rostral end, and ∼100 μm in front of the anterior intestinal portal vein. At this stage, the caudal domain is EZ-II(Couly et al., 2002). The neural tube and the ectoderm were then delicately placed back in their initial position, where they rapidly reconnected with the rest of the embryonic tissues, resuming their development with no obvious defects. In a series of experiments, before closing the embryo, we placed a heparin acrylic bead(Sigma) in the vacant region where the EZ-I had been ablated. Embryos were reincubated until stage HH35 for morphological and chondrocranial analysis.

Grafts of EZ-I were obtained from 5-somite quail embryos following the same dissection procedure described for EZ-I ablation in the chick. EZ-I grafts were transplanted into chick embryos at the 5-somite stage in ovo after performing a unilateral ectodermal incision at the appropriate axial level. The graft was placed either in the anteroventral mesenchyme (below the level of the prosencephalon) or into the presumptive mesenchyme of the first branchial arch. Some embryos were fixed and sectioned 24 hours after operation(HH14) to analyse Shh expression, to detect quail nuclei using the QCPN antibody, and to examine Gli1 expression. In some cases, a simultaneous graft of both EZ-I and anterior neural crest was used to reveal, by QCPN immunostaining, the relative position of migrating CNCCs and of the grafted endoderm.

Experiments were carried out in ovo on windowed chick embryos at the 5-somite stage (Couly et al.,2002). For rescue experiments using Shh, heparin beads (120 μm diameter; Sigma, St Louis, MO) were soaked in 100 μg/ml mouse recombinant Shh (R&D systems, Minneapolis, MN) in PBS for 1 hour at 37°C, then rinsed three times in PBS immediately prior to use. Beads soaked in 0.1% BSA were used as control. EZ-I was ablated as previously described and then one bead was placed in the vacated territory. For experiments with cyclopamine,heparin beads were soaked in crystalline 11-deoxyjervine (Toronto Research Chemicals, Toronto, Canada) at 4 mg/ml in 95% ethanol, and then rinsed three times in PBS prior to use (Watkins et al.,2003). Control beads were soaked in PBS.

Microdissected tissue fragments from transverse domains of the ventrolateral endoderm or bilateral neural plate apical ridges were collected as illustrated in Fig. 4A(Couly et al., 2002; Ruhin et al., 2003). Endodermal stripes and neural crest fragments were washed in PBS, then either deposited onto 0.4 μm porosity Millipore nylon inserts(Gitton et al., 1999) or onto glass Lab-Tek multichambered slides (Nunc) for up to 48 hours and cultured in Dulbecco's Modified Eagle Medium (Gibco, France). This defined medium was supplemented with a 1:1 mix of serum replacement medium (Sigma) and B27 solution (Gibco), a 1:1 mix of penicillin and streptomycin antibiotic mix (25μg/ml and 25 U/ml) and l-glutamine (2 mM, Gibco) as previously described (Dahmane et al.,2001). Pharmacological treatments included 20 μM cyclopamine(R&D Systems) or 10 μM Shh [recombinant mouse N-terminal fragment(Shh-N), R&D Systems] diluted in the culture medium. Significant cell emigration was observed around neural crest explants as early as 12 hours after incubation.

At the end of the culture period, the explants and surrounding cells were gently washed in culture medium, scraped and detached from the support and collected for total RNA extraction using the RNeasy Microkit (Qiagen). To obtain equivalent amounts of extracted material, equivalent numbers of tissue fragments were used, i.e. two EZ-1 or two bilateral CNCC fragments were equated with each co-culture of EZ-1 and CNCCs. For random hexamer-primed reverse transcription into total cDNA, we used the RT-PCR First-Strand Synthesis System (Invitrogen) including DNaseI treatment. Negative controls that omitted the Superscript II reverse transcriptase demonstrated that all samples were free of contaminant nucleic material (data not shown). Total cDNAs were analysed by PCR in the linear amplification range (30 cycles). Expression of chicken Gapdh, Shh and Gli1 was monitored by PCR; primers and conditions are available upon request.

To determine the function of EZ-I in the control of craniofacial development, we first ablated this endodermal territory in ovo from 5-somite chick embryos, well before CNCC migration(Fig. 1C; see Fig. S1 in the supplementary material) and analysed their skeletal morphology at Hamburger Hamilton stage 35 (HH35). In all surviving embryos (8/19), the mesethmoid cartilage was absent or severely reduced, whereas the ectethmoid and the infraorbital septum were always present and of relatively normal size and shape (Fig. 1D,D′). The adjacent ectoderm, neural tube and first arch CNCC derivatives developed normally.

Next we transplanted a supplementary EZ-I from 5-somite quail embryos into the presumptive nasal capsule territory of stage-matched chick embryos and analysed skeletal morphology at HH35. All survivors (12/21) displayed one ectopic, supernumerary mesthemoid-like cartilage(Fig. 2A,B,B′) next to the normal and complete nasal capsule. Heterotopic quail EZ-I grafts,implanted within the 5-somite chick presumptive first PA territory, induced a supernumerary mesthemoid-like element in 10/12 survivors(Fig. 2C,D). Dermatocranial elements [putatively premaxillary bones(Couly et al., 1993)] formed in close proximity to the supernumerary cartilage. All other first and second PA derivatives developed normally. Both homotopic and heterotopic grafts suggest that EZ-I is necessary and sufficient to induce the differentiation of any Hox-negative CNCC contingent into a mesethmoid cartilage. Ectopic cartilage elements consistently failed to develop when we grafted EZ-I into the Hox-positive neural crest domain (data not shown).

We then examined whether Shh expression by EZ-I(Fig. 3A,B)(Brito et al., 2006) could be responsible for its capacity to induce mesethmoid formation. First, we doubly transplanted EZ-I and CNCCs from 5-somite quail into stage-matched chicken embryos. In all cases, the grafted EZ-I was strongly Shh-positive and was in close contact with grafted post-migratory CNCCs (Fig. 3A-C). We then transplanted EZ-I into the presumptive first PA close to the ocular region and observed a strong ectopic induction of Gli1, a Shh target(Dahmane et al., 2001) not normally expressed in this area (see Fig. S2 in the supplementary material). To determine whether Shh expression from EZ-I directly contributed to mesethmoid induction, we ablated EZ-I from 5-somite embryos and implanted a Shh-loaded bead in the vacant territory(Fig. 3D). In all surviving embryos (18/36), the nasal capsule presented a mesethmoid-like cartilage correctly located under a normal ectethmoid(Fig. 3E,E′). Control BSA-loaded beads did not rescue the lack of mesethmoid resulting from EZ-I ablation (n=6).

To assess whether Shh-Gli1 signalling was necessary to generate a mesthemoid cartilage, we then implanted a cyclopamine-loaded bead in contact with EZ-I at the 5-somite stage. In all survivors (14/36), the mesethmoid was absent, whereas the ectethmoid could be recognised (see Fig. S3 in the supplementary material).

To address whether direct CNCC-EZ-I interaction could trigger Shh-Gli1 signalling, we examined organotypic cultures of endoderm and neural crest fragments from 5-somite embryos(Fig. 4A). We cultured the explants for 24 hours either alone or in combination. RT-PCR analysis of EZ-I explants alone indicated a weak constitutive Shh expression accompanied by a barely visible Gli1 signal. Conversely, neither Shh nor Gli1 transcription could be seen in CNCC explants. Co-cultures of EZ-I and CNCC fragments displayed substantially higher levels of Gli1 transcripts, as compared with either component alone. Shh levels were also increased. Cyclopamine inhibition of Shh-Gli1 transduction in the co-cultures prevented Gli1expression and reduced Shh levels(Fig. 4A). Furthermore,co-cultures of CNCC explants and endoderm zone II (EZ-II) failed to display Gli1 expression, even after 48 hours in vitro (data not shown).

Consistent with the observation that endoderm-derived Shh activates Gli1 in CNCCs, recombinant Shh-N protein induced Gli1 in cultures of CNCCs alone, thereby reproducing the endoderm inductive effect. By contrast, exposure to either Shh-N or cyclopamine repressed Shhexpression in cultures of endoderm explants.

Collectively, our in ovo and ex ovo data suggest a reciprocal,cyclopamine-sensitive interaction between EZ-I and uncommitted CNCCs. Shh expression increases in the rostral-most endoderm following contact with post-migratory CNCCs (Fig. 4B). Shh, in turn, strongly induces Gli1 transcription in CNCCs. This inductive process leads to chondrogenesis, resulting in the formation of the mesethmoid cartilage (Fig. 4C), and later to the formation of the corresponding dermatocranial elements.

Our previous results suggested a very precise topographical engraving of endodermal information as the orientation of the endodermal grafts was precisely reflected in the shape and orientation of CNCC-derived ectopic cartilages (Couly et al.,2002).

Although the nature of these molecular clues is not yet elucidated,candidates include Bmp4, Fgf8, Shh, endothelin 1, noggin and retinoic acid(Foppiano et al., 2007; Hu et al., 2003; Lee et al., 2001; Song et al., 2004; Vieux-Rochas et al., 2007). The importance of Shh signalling for craniofacial morphogenesis(Brito et al., 2006; Haworth et al., 2007) and CNCC survival (e.g. Ahlgren and Bronner-Fraser,1999; Brito et al.,2006; Charrier et al.,2001; Cordero et al.,2004) has been reported previously, but the topological and chronological significance of this pathway remains to be determined. In particular, it has been shown that the maintenance of a boundary between the territories of expression of Fgf8 and Shh in the frontonasal process depends on reciprocal inhibitory interactions(Abzhanov et al., 2007; Abzhanov and Tabin, 2004; Haworth et al., 2007).

Our observations imply that endoderm-derived Shh signalling, which takes place already at the 5-somite stage, predates the ectodermal Shh contribution to craniofacial patterning (Hu and Helms,2001; Hu et al.,2003). Globally, these results indicate that whereas early mesethmoid differentiation requires EZ-I Shh expression, ectodermal expression of Fgf8, Shh and Bmp4 controls the size and shape of the beak at later stages of development (Cordero et al., 2004; Hu et al.,2003; Wu et al.,2004). Shh is both a long-range diffusible morphogen and a short-range contact-dependent factor(Chuang and McMahon, 1999; Johnson and Tabin, 1995). Here, ablation of EZ-I, a topographically restricted source of Shh in the embryo, leads to a morphological lesion limited to the mesethmoid. This implies that, at least in this context, the action of Shh takes place in a localised paracrine, possibly contact-mediated, fashion. The human mesethmoid cartilage gives rise to the nasal septum and the vomer. It thus contributes to determining the proximodistal size of the nose and the position of the premaxillary bone. Our results could help to clarify the origin of a number of malformations of the nasofrontal bud(Couly, 1981) that are characterised by abnormal mesethmoid development, with normal ectethmoid. These pathologies include, for example, cases of hypo- and hyper-septoethmoidism such as a flat nose, short nose or very pronounced nose and a median nasal cleft without cerebral anomalies. Thus, although as yet we cannot answer Cyrano de Bergerac's question “De quoi sert cette oblongue capsule?” (What use this oblong capsule?), we may begin to understand where this oblong capsule comes from.

We gratefully acknowledge the helpful comments of Prof. Nicole Le Douarin,Prof. Barbara Demeneix and Dr Anne Grappin-Botton, and the logistical help of Dr Marie-Aimée Teillet. We thank Sophie Gourmet and Michel Fromaget for illustrations. This research was partially supported by the EU Consortium CRESCENDO (LSHM-CT-2005-018652) and the ANR project `GENDACTYL'.