Rostral and caudal pharyngeal arches share a common neural crest ground pattern

A large part of the vertebrate craniofacial skeleton and the entire pharyngeal skeleton derive from cranial neural crest cells (NCCs) that arise along the dorsal edge of the forming neural tube(Creuzet et al., 2005). Cranial NCCs originating from rostral brain regions migrate into the frontonasal process, whereas at hindbrain level, NCCs migrate out of the segmented rhombomeres (R) and colonize the pharyngeal arches (PAs), another series of metameric structures that give rise to maxillary, mandibular, middle ear,hyoid and neck bones (Santagati and Rijli,2003). Therefore, a segmental pattern underlies the development and morphogenesis of the pharyngeal region of the vertebrate head(Graham, 2007; Kuratani et al., 1997). Nonetheless, fundamental differences exist in the deployment and contribution of NCCs to the PAs between pre-otic (R1-R4) and postotic (R6-R8) regions.

At pre-otic level, NCCs migrate into separate streams: NCCs from the posterior mesencephalon, R1 and R2 fill the first (mandibular) arch, whereas the second (hyoid) arch is colonized by NCCs originating mainly from R4(Lumsden et al., 1991; Sechrist et al., 1993). As both R3 and R5 provide little if any NCC contribution to PA derivatives(Graham et al., 1993; Sechrist et al., 1993), this leads to the segregation of first, second and post-otic arch NCCs into distinct streams. By contrast, within post-otic NCCs, there is no segmental registration between their rhombomeric origin and PA contribution(Shigetani et al., 1995), such that NCC subpopulations entering arches 3, 4 and 6 are generated from the entire rostrocaudal extent of the post-otic hindbrain and are partially overlapping, although arch 3 mainly contains NCCs of R6 origin(Shigetani et al., 1995). Such a distribution is also reflected in the overlapping deployment of post-otic NCCs to form skeletal elements of multi-rhombomeric origin, whereas pre-otic NCC subpopulations maintain strict segregation on skeletal derivatives of the first two arches according to their rhombomeric origin(Kontges and Lumsden, 1996). It is therefore unclear whether first and second arches are patterned by molecular mechanisms distinct from those acting in more posterior arches, or whether all PAs belong to a homology series sharing a common ground patterning program.

At the molecular level, NCC subpopulations contributing to distinct PAs express various combinations of Hox genes, providing each arch with distinct regional identities along the anteroposterior (AP) axis(Santagati and Rijli, 2003; Trainor and Krumlauf, 2001). Hoxa2 is the only Hox gene expressed in the skeletogenic NCCs of the second PA, whereas first arch NCCs are Hox-negative(Couly et al., 1998). In mouse,the inactivation of Hoxa2 in NCCs resulted in homeotic transformation of second arch derivatives into morphologies characteristic of Hox-negative first arch-derived structures(Gendron-Maguire et al., 1993; Rijli et al., 1993; Santagati et al., 2005). Therefore, Hoxa2 promotes second arch skeletal development by modifying an underlying Hox-negative, first arch-like ground (default)patterning program, shared by the skeletogenic NCCs of the first and second arches (Rijli et al., 1993). Such a proposal has been further supported by functional studies in other vertebrate systems, including zebrafish, Xenopus and chick(Baltzinger et al., 2005; Crump et al., 2006; Grammatopoulos et al., 2000; Hunter and Prince, 2002; Miller et al., 2004; Pasqualetti et al., 2000).

By contrast, third and fourth arch NCC-derived skeletal structures were not affected in Hoxa2^-/- or Hoxb2^-/-single or compound mutants (Barrow and Capecchi, 1996; Davenne et al., 1999; Gendron-Maguire et al., 1993; Rijli et al.,1993), suggesting that paralog group 2 (PG2) genes are dispensable or functionally compensated by other Hox genes in arches posterior to the second. Indeed, third arch NCCs express both PG2 Hox (Hoxa2 and Hoxb2) and PG3 (Hoxa3, Hoxb3 and Hoxd3) genes (see Fig. S1 in the supplementary material). Functional inactivation of Hoxa3, in contrast to both Hoxb3 or Hoxd3, resulted in malformations of the greater horn of the hyoid bone and fusions to the thyroid cartilage (Chisaka and Capecchi,1991). Such defects were exacerbated in Hoxa3^-/-/Hoxd3^-/- or Hoxa3^-/-/Hoxb3^-/- mutants, whereas Hoxb3^-/-/Hoxd3^-/- compound mutants displayed milder defects when compared with single Hoxa3^-/- mutants(Condie and Capecchi, 1994; Manley and Capecchi, 1997). Consequently, although dose-dependent genetic interactions occur between Hoxa3, Hoxb3 and Hoxd3, there is a prevalent role for Hoxa3 in patterning third arch NCC derivatives. However, no homeotic transformations were observed in any of the described Hox PG3 mutants. Similarly, single or compound PG4 Hox (Hoxa4, Hoxb4, Hoxc4 and Hoxd4) mutants did not result in any homeosis of NCC derivatives(Boulet and Capecchi, 1996; Horan et al., 1995; Ramirez-Solis et al., 1993). The question thus remained as to whether or not pre-otic NCCs share an underlying patterning program distinct from that of post-otic NCCs. Alternatively, all pharyngeal arch NCC skeletal derivatives may be built on the top of the same Hox ground patterning program, yet posterior to the second arch the observation of this ground-state would necessitate the deletion in cis of Hox genes belonging to multiple groups of paralogy.

To discriminate between these possibilities, we have deleted all Hoxa genes in NCCs and report here evidence for the occurrence of homeotic transformations of third and fourth arch derivatives towards first arch-like Meckel's cartilage morphology, with additional appearance of ectopic first arch-specific membranous bone elements. These results demonstrate that a common Hox-free ground patterning program is shared caudally to the first and second arches, and that Hoxa2 and Hoxa3 act synergistically to pattern third and fourth arch NCC derivatives. Moreover, additional deletion of the entire Hoxd cluster in the context of the NCC-specific Hoxa inactivation did not significantly enhance such a homeotic phenotype, but resulted only in an increase of the frequency of ectopic squamosal bones in posterior pharyngeal arches, suggesting a preponderant role of Hoxa genes in patterning skeletogenic NCCs in mammals. We discuss these results both in terms of Hox-mediated pharyngeal arch patterning and from an evolutionary standpoint.

Generation of conventional Hoxa2 and conditional Hoxa2^flox mutant alleles has been described previously(Rijli et al., 1993; Ren et al., 2002). Mice mutant for the Hoxd cluster were as reported previously(Zakany et al., 2001). The floxed Hoxa cluster allele was that described by Kmita et al.(Kmita et al., 2005) and the Wnt1::Cre mice were those described by Danielian et al.(Danielian et al., 1998). For the generation of mutant fetuses, Hoxa^flox/+/Hoxd^del/+ specimens were intercrossed, with one parent carrying in addition the Wnt1::Cretransgene. The R4::Cre and Z/AP reporter transgenic lines have been described previously (Oury et al.,2006; Lobe et al.,1999).

Newborns were skinned and eviscerated. Skeletons were fixed overnight in 95% ethanol and then stained in Alcian Blue (750 μg/ml in 80 ml of 95%ethanol and 20 ml of glacial acetic acid) for at least 24 hours. Skeletons were cleared in 1% KOH overnight and then stained in Alizarin Red (100μg/ml in 1% KOH) overnight. Further clearance was performed in 20%glycerol/1% KOH. Skeletons were stocked in 50% glycerol/50% water.

In situ hybridization on sections and on whole-mount embryos were performed as previously described (Santagati et al.,2005). The following RNA probes were used: Hoxa2(Ren et al., 2002), Hoxb2 (Vesque et al.,1996), Hoxa3, Hoxb3, Hoxd3(Manley and Capecchi, 1997), Hoxa4, Hoxb4, Hoxd4 (Gaunt et al., 1989), CRABP1(Maden et al., 1992), Pax1 (Wallin et al.,1996), Alx4 (Qu et al., 1997), Pitx1(Lanctot et al., 1999) and Barx1 (Tissier-Seta et al.,1995).

E15.5 embryos were fixed overnight in 4% paraformaldehyde (PFA), rinsed in phosphate-buffered saline (PBS), equilibrated in 20% sucrose and included in Cryomatrix (Thermo Electron Corporation). Cryostat sections (20 μm) were cut in the frontal plane of the embryo to visualize the great and lesser horns of the hyoid bone. The sections were then fixed for 1 hour in 4% PFA, rinsed in PBS at room temperature and then incubated in PBS at 65°C during 1 hour to inactivate the endogenous AP. After the inactivation, sections were rinsed in a solution of NTMT (NaCl 0.1 M, Tris HCl 0.1 M (pH 9.5), MgCl₂0.05 M and Tween 20 0.1%). NBT (3.5 μl; Roche 1383213) and 3.5 μl BCIP(Roche 1383221) per ml of NTMT were used for the staining. The sections were rinsed first in water, next in 100% ethanol and mounted onto slides.

A Yates' Chi-square test was used to compare the frequency of appearance of each ectopic membranous bone (i.e. supernumerary tympanic, squamosal and dentary bones) between the group formed by Wnt1::Cre/Hoxa^flox/flox and Wnt1::Cre/Hoxa^flox/del, and the group formed by Wnt1::Cre/Hoxa^flox/flox/Hoxd^del/+ and Wnt1::Cre/Hoxa^flox/flox/Hoxd^del/del mutant newborns. The analysis were carried out using Statistica 7.1 software package(Statsoft)

To analyze Hox gene expression in PAs, we performed whole-mount in situ hybridization on E10.0 mouse embryos using antisense probes for PG2(Hoxa2 and Hoxb2), PG3 (Hoxa3, Hoxb3 and Hoxd3) and PG4 (Hoxa4, Hoxb4 and Hoxd4) genes. In addition to arch 2, PG2 Hox genes were strongly expressed in arches 3 and 4 NCCs (see Fig. S1A,B in the supplementary material). PG3 Hox genes were expressed in the dorsolateral circumpharyngeal (CP) ridge (see Fig. S1C-E in the supplementary material) from which post-otic NCCs stream into each posterior arch (Shigetani et al.,1995). Among PG3 Hox genes, Hoxa3 was highly expressed in arches 3 and 4 (see Fig. S1C in the supplementary material), whereas Hoxb3 and Hoxd3 transcripts were detected at low level and/or in subsets of arch 3 cells and at relatively high levels in arch 4 (see Fig. S1D-E,I in the supplementary material). Interestingly, in keeping with a previous study on human embryos(Vieille-Grosjean et al.,1997), we found that Hoxa4 and Hoxb4 were not expressed at significant levels in arches 3 and 4, but only in more posterior regions (see Fig. S1F-G,J in the supplementary material). By contrast,however, we found Hoxd4 relatively highly expressed in arch 4 NCCs(see Fig. S1H,K in the supplementary material). To summarize, at E10.0, arch 3 mostly expressed Hoxa2, Hoxb2 and Hoxa3, whereas Hoxa2,Hoxb2, Hoxa3, Hoxb3, Hoxd3 and Hoxd4 transcripts were scored in arch 4.

Given the above Hox expression patterns, we decided to look at the morphogenesis of post-otic NCC skeletal derivatives after deleting the entire Hoxa and Hoxd gene clusters, either separately or in combination. Cre-loxP mediated targeted deletion of the entire Hoxd cluster(Zakany et al., 2001) resulted in Hoxd^del/del newborns with no significant defects at the level of NCC-derived skeletal structures (not shown). By contrast, deletion of the entire Hoxa cluster resulted in premature lethality at early to mid-embryonic stages (Kmita et al.,2005), thus preventing the analysis of NCC derived skeletal elements.

To overcome such an early lethality, we conditionally deleted the entire Hoxa cluster in NCCs, based on a previous observation showing that, at least in the case of Hoxa2, selective inactivation in NCCs was sufficient to recapitulate the skeletal phenotype of the full inactivation(Santagati et al., 2005). We used a Wnt1::Cre transgenic mouse line in which the Wnt1promoter drives Cre recombinase expression in NCC progenitors(Danielian et al., 1998) to cross with a conditional Hoxa^flox allele(Kmita et al., 2005). Wnt1::Cre/Hoxa^flox/flox or Wnt1::Cre/Hoxa^flox/del fetuses survived throughout fetal development, up to birth, similar to Wnt1::Cre/Hoxa^flox/flox/Hoxd^del/+ or Wnt1::Cre/Hoxa^flox/flox/Hoxd^del/del mutant animals. Newborn mutants, however, eventually died within a few hours, and suffered from several skeletal abnormalities (see below). To assess the efficiency of the Wnt1::Cre mediated recombination, we carried out whole-mount in situ hybridization on E10.0 mutant embryos. We have previously shown that the Wnt1::Cre driver mated to a Hoxa2 floxed allele efficiently induces lack of Hoxa2 expression in NCCs(Ren et al., 2002; Santagati et al., 2005). Similarly, we tested the expression of Hoxa3 in Wnt1::Cre/Hoxa^flox/del mutant embryos and observed, as expected, a selective loss of Hoxa3 transcripts from the NCCs contributing to the third and more posterior PAs, whereas the other sites of Hoxa3 expression were unaffected (see Fig. S2A-D in the supplementary material), thus validating the conditional deletion approach.

Studies in non-mammalian vertebrates assessed that the hyoid bone complex has a composite origin from both pre-otic and post-otic NCCs (e.g. Kontges and Lumsden, 1996; Couly et al., 1998). In mammals, the precise contribution of pre- and post-otic NCCs to the mammalian hyoid complex has not been mapped by long-term tracing experiments. It is generally assumed that the lesser horns of the hyoid bone derive from the second, and the greater horns from the third, arch NCCs. Moreover, the prevalent opinion in literature is that second arch pre-otic NCCs contribute to the upper part of the body of the hyoid bone, whereas third arch post-otic NCCs colonize the lower part (Larsen,1993).

Given the relevance for our present analysis, we set up to define the pre-otic versus post-otic NCC composition of the hyoid bone apparatus in the mouse. To this aim, we mated a previously described R4::Cre mouse line (Oury et al., 2006) to the Z/AP transgenic reporter (Lobe et al.,1999). Upon Cre-mediated recombination, the expression of the alkaline phosphatase (AP) gene was permanently activated in the pre-otic NCC progenitors specifically emerging from R4 and colonizing the second pharyngeal arch (Fig. 1G; see Fig. S3A,B in the supplementary material). AP staining was present throughout the arch mesenchyme, except in the central mesodermal core(see Fig. S3B in the supplementary material). Conditional inactivation of Hoxa2 with the R4::Cre driver was sufficient to induce the full extent of the Hoxa2 knockout mutant phenotype (see Fig. S3C-F in the supplementary material). Namely, all second arch-derived skeletal elements, including the lesser horn of the hyoid bone, were absent and replaced by a duplicated set of first arch-like elements in reverse polarity. Altogether, these data strongly indicated that AP⁺ cells represent the vast majority of NCC contribution to the second PA, even though minor contributions from R3- and/or R5-derived NCCs could not be ruled out (e.g. Kontges and Lumsden, 1996). By contrast, post-otic NCCs migrating into third and more posterior arches were devoid of reporter gene expression (Fig. 1G, arrow; see Fig. S3A,B in the supplementary material). Thus,mapping the fate of AP⁺ versus AP^- mesenchyme on the hyoid complex essentially allows the assessment of the spatial deployment and contribution of pre-otic (i.e. second arch-derived) versus post-otic (i.e. third arch-derived) NCCs.

AP staining of tissue sections at E15.5 revealed that the lesser horn was entirely made of AP⁺ NCCs, whereas the greater horn was entirely AP^-, i.e. of post-otic NCC origin(Fig. 1A-C,H). Interestingly,as for the body of the hyoid bone, we found that pre-otic AP⁺ and post-otic AP^- NCCs were deployed in a striped pattern(Fig. 1A-C,H). Pre-otic AP⁺ NCCs contributed to the central part of the body, as well as to two symmetrical regions articulating with the lesser horns, which abutted the post-otic AP^- greater horn of the hyoid bone. In summary, our data confirm that the lesser horn is a second arch derivative, whereas the greater horn is most likely of third arch origin. In addition, the body of the hyoid bone is a composite of pre-otic and post-otic NCCs that remain segregated in an alternate striped pattern.

Newborns lacking the Hoxa cluster in NCCs displayed an absence of the external ear and cleft palate (not shown), much like the original Hoxa2^-/- mutants(Gendron-Maguire et al., 1993; Rijli et al., 1993). Moreover,in the second arch, the phenotypes of skeletal derivatives in Wnt1::Cre/Hoxa^flox/flox(Fig. 2E,F,I,J; Fig. 4B), Wnt1::Cre/Hoxa^flox/del(Fig. 4C), Wnt1::Cre/Hoxa^flox/flox/Hoxd^del/+(Fig. 2G,H; Fig. 4D) or Wnt1::Cre/Hoxa^flox/flox/Hoxd^del/del (not shown)mutant fetuses were all similar to those observed upon Hoxa2inactivation (Fig. 2C,D), as expected from Hoxa2 being the sole gene expressed in this arch. In particular, the stapes, styloid process and lesser horns of the hyoid bone were absent and were replaced by mirror-image duplications of first arch-like structures, including partially duplicated Meckel's cartilage (MC2), incus(I2), malleus (M2), tympanic (T2) and pterygoid (P2) bones, as well as transformed gonial (G^*) and ectopic squamosal (SQ2) bones. Such an outcome at a skeletal level further confirmed the efficiency of our conditional Hoxa cluster deletion.

In addition, however, newborns lacking the Hoxa cluster in NCCs displayed ectopic skeletal elements (compare Fig. 2E,F,I,J and Fig. 4B-D with Fig. 2C,D), observed neither in single Hoxa2^-/- or Hoxa3^-/- specimen, nor in compound Hoxa3^-/-/Hoxd3^-/- mutant animals(Chisaka and Capecchi, 1991; Condie and Capecchi, 1994; Manley and Capecchi, 1997). In all analyzed mutants (see Table S1 in the supplementary material), the greater horn of the hyoid bone (GH) significantly extended dorsally and resembled a partially triplicated Meckel's cartilage (MC3), whereas MC2 extended further ventrally than in single Hoxa2^-/- specimen (compare Fig. 2E,F with 2C,D). In addition, in some mutants (see Table S1 in the supplementary material), the dorsal end of such ectopic cartilage displayed a morphology resembling an ectopic first arch-like malleus (M3) (Fig. 2E,F).

A significant fraction of Wnt1::Cre/Hoxa^flox/floxmutants (nine out of 20) also displayed a notable transformation of the thyroid cartilage, which we interpret as a partial quadruplication of Meckel's jaw cartilage (MC4) (Fig. 2I,J; Fig. 3B). In these mutants, the lateral process (LP) of the thyroid cartilage was reduced and transformed into an ectopic cartilage that projected dorsally and fused with the transformed greater horn of the hyoid bone (MC3). Furthermore, in the mutant presented in Fig. 2I,J, the ectopic MC4 and MC3 directly fused with an MC2 that unusually extended ventrally(posteriorly), further supporting their identity as supernumerary jaw cartilages.

The frequency of MC3 or MC4 was not enhanced upon additional deletion of one or two alleles of the Hoxd cluster (see Table S1 in the supplementary material; and not shown). The Wnt1::Cre/Hoxa^flox/flox or Wnt1::Cre/Hoxa^flox/flox/Hoxd^del/+ mutant newborns that did not bear the ectopic MC4 displayed fusions between hyoid bone and thyroid cartilage (Fig. 2G,H, arrowhead), reminiscent of the Hoxa3 inactivation phenotype (Chisaka and Capecchi,1991), whereas Wnt1::Cre/Hoxa^flox/flox/Hoxd^del/del mutants displayed a reduction of the thyroid cartilage (data not shown), similar to the Hoxa3^-/-/Hoxd3^-/- double knockout phenotype(Condie and Capecchi, 1994). Conversely, deletion of only one Hoxa^flox allele in NCCs,in the context of a full Hoxd deletion(Wnt1::Cre/Hoxa^flox/+/Hoxd^del/del) did not result in morphological transformation, neither of the hyoid bone, nor of the thyroid cartilage, but only in mild malformations (not shown), indicating that the remaining copy of Hoxa^flox allele was sufficient to compensate for the loss of both Hoxd alleles.

Finally, supernumerary ectopic membranous bones also appeared in a fraction of Hoxa and Hoxa/Hoxd deleted mutants (see Table S1 in the supplementary material; Fig. 2G,H; Fig. 4B-D) that were never observed in wild type, Hoxa2^-/- or Hoxa3^-/- mutants, nor in double Hoxa3^-/-/Hoxd3^-/- newborns. Caudal to SQ2, such ectopic elements may reflect a partial triplication of squamosal bones (SQ3)(Fig. 4C,D). Additional ectopic membranous elements were also present in the proximity of MC3 and M3,respectively, and may thus represent supernumerary dentary (DB2) and tympanic(T3) bones (Fig. 2G,H). Such supernumerary membranous bones were scored with variable occurrence (see Table S1 in the supplementary material) and often only on one side of the mutants,indicating stochastic compensation for the lack of Hoxa genes in NCCs. The deletion of either one or both Hoxd cluster alleles in a Hoxa mutant background did not significantly enhance this phenotype, with the exception of increasing the frequency of the ectopic SQ3 membranous bones (see Table S1 in the supplementary material) (P=0.01 with the Yates' Chi-square test),thus indicating redundancy with Hoxa genes in repressing the formation of this structure in posterior arches.

Altogether, these results strongly suggested that, in the absence of Hoxa genes in NCCs, the third and fourth arch post-otic NCC derived structures partially underwent homeotic transformations into first arch-like morphologies, similar to arch 2 NCCs, thus pointing to a potential common Hox-free ground patterning program shared by NCCs originating from both pre-and post-otic domains.

To assess the status of NCC migratory streams in Hoxa-deleted mutant embryos, we carried out whole-mount in situ hybridization. As a marker for migrating cranial NCCs, we used an antisense probe for cellular retinoic acid-binding protein 1 (Crabp1)(Maden et al., 1992) and found no difference in Crabp1 expression patterns in E9.5 Wnt1::Cre/Hoxa^flox/flox or Wnt1::Cre/Hoxa^flox/flox/Hoxd^del/del mutant embryos, when compared with a wild-type specimen(Fig. 5A,B; and not shown). We then assessed the pattern of PA segmentation by analyzing the expression pattern of Pax1, which specifically labels the endodermal pharyngeal pouches, from their initial formation until late stages(Wallin et al., 1996). In E10.0 Wnt1::Cre/Hoxa^flox/flox and Wnt1::Cre/Hoxa^flox/flox/Hoxd^del/del mutant embryos, Pax1 was normally expressed in the endoderm of each pharyngeal pouch and no fusions occurred between adjacent arches, when compared with wild-type embryos (Fig. 5C,D; and not shown). From these results, we concluded that the combined deletion of Hoxd genes in all cells and Hoxa genes in NCCs, had altered neither pharyngeal segmentation nor the migration patterns of cranial NCCs.

We next investigated the molecular identity of pharyngeal arch NCCs in Wnt1::Cre/Hoxa^flox/flox mutant embryos. In E10.5 wild-type embryos, the expression domains of Pitx1, a marker of Meckel's cartilage (Lanctot et al.,1999), and Alx4 are restricted to the first arch(Fig. 5E,H). In Wnt1::Cre/Hoxa^flox/flox embryos, both Pitx1 and Alx4 were ectopically expressed in the second arch, similar to Hoxa2^-/- mutants (compare Fig. 5F-G,I-J with 5E,H)(Santagati et al., 2005). In addition, however, Alx4 and Pitx1 were ectopically expressed either in the third, or in both the third and fourth PAs, respectively, of Wnt1::Cre/Hoxa^flox/flox mutant embryos(Fig. 5G,J). Notably, the Pitx1 ectopic expression domains in PA2-4 were confined posteriorly,thus opposite to the wild-type pattern in anterior PA1 (compare Fig. 5E with 5G). This observation might well correlate with the potentially reversed polarity of the ectopic MC3 and MC4 in these mutants, in addition to the mirror-imaged MC2 expected from the Hoxa2 deletion in PA2 (Fig. 2E-J). Barx1, a marker of chondrocyte differentiation and condensation (Sperber and Dawid,2008), is expressed in all pharyngeal arches of wild-type embryos,although with arch-specific differences in its spatial distribution(Fig. 5K). Indeed, at E10.5 Barx1 is highly expressed in the mandibular and maxillary portions of PA1 and in ventral PA2, as well as at lower levels in ventral PA3 and PA4(Fig. 5K). Barx1transcripts are also present in a small domain at the dorsoanterior margin of PA2 (Fig. 5K, arrowhead). In Hoxa2^-/- mutant embryos, such domain was extended ventrally and posteriorly (similar to PA1)(Fig. 5L). Similarly, this ectopic Barx1 domain was also observed in Wnt1::Cre/Hoxa^flox/flox mutant PA2(Fig. 5M, arrow). In addition,an ectopic expression domain of Barx1 appeared in the dorsal part of PA3 (Fig. 5M, arrow). Altogether, these results further support a view whereby the deletion of the Hoxa cluster in NCCs results, in addition to PA2 molecular changes, in transformation of PA3 and PA4 gene expression patterns into a first arch,jaw-like pattern, thus indicating that post-otic NCCs acquired partial first arch-like identities.

This study provides novel insights into the molecular mechanisms underlying craniofacial patterning. We report that the conditional deletion of the Hoxa cluster in NCCs induced a more extensive phenotype than the mere combination of Hoxa2 and Hoxa3 loss of functions. In such conditional mutant mice, multiple sets of first arch-like skeletal elements appeared,including partial quadruplication of the jaw cartilage and supernumerary mallei, tympanic, dentary and squamosal bones. We present evidence indicating that such supernumerary elements arose through partial transformation of third and fourth arch post-otic NCC segmental identities. We first show that Hoxa-deleted embryos displayed normal PA segmentation and NCC migration. Subsequently, we document that ectopic jaw-like elements result from a morphological alteration of the greater horn of the hyoid bone and the lateral process of the thyroid cartilage, third and fourth arch derived structures,respectively. Finally, molecular markers normally expressed in the PA1 of wild-type embryos were ectopically induced in the third and fourth PAs of Wnt1::Cre/Hoxa^flox/flox mutant specimen. We conclude that in the absence of Hoxa genes in NCCs, the third and fourth arches adopt a partial first arch-like identity, in addition to the homeotic transformation induced in the hyoid arch by the absence of Hoxa2. Notably, partial anterior homeotic transformations of PA3 and PA4 were reported in zebrafish that were mutant for the histone acetyltransferase moz (monocytic leukemia zinc finger). This mutations is also associated with reduced Hox expression levels (Miller et al.,2004). Altogether, our findings provide strong evidence that post-otic and pre-otic NCCs share the same Hox-free ground patterning program(model in Fig. 6).

Our results revealed an unprecedented role for Hoxa2, as well as synergistic interactions between Hoxa2 and Hoxa3, in the patterning of third and fourth PA skeletal derivatives. We have previously shown that Hoxa2 performs its main patterning role in post-migratory skeletogenic NCCs (Santagati et al.,2005). Given the absence of Hoxa4 expression in arch 3 and 4 NCCs (see Fig. S1F in the supplementary material)(Behringer et al., 1993), our data imply that the phenotypes obtained in Wnt1::Cre/Hoxa^flox/flox mutant embryos can entirely be accounted for by lack of Hoxa2 and Hoxa3 in NCCs. However,the role of Hoxa2 only becomes apparent in the absence of Hoxa3, indicating non-equivalent functions between Hoxa2 and Hoxa3. In the absence of Hoxa2, third and fourth arch-derived structures were indeed normally patterned(Rijli et al., 1993),suggesting that Hoxa3 is able to compensate for the lack of Hoxa2. By contrast, in the absence of Hoxa3, morphological alterations of hyoid and thyroid cartilages were detected(Chisaka and Capecchi, 1991)showing a prevalent role for Hoxa3 over Hoxa2, in agreement with the posterior prevalence concept(Duboule and Morata, 1994). Homeotic transformations were nevertheless not observed under these conditions. Here, we show that Hoxa2 function also significantly contributes to patterning the third and fourth arches, as the concomitant removal of Hoxa2 from a Hoxa3-deficient background generated dramatic homeosis.

Because Hoxa2 and Hoxa3 have overlapping spatial expression domains in third and fourth arch NCCs (see Fig. S1A,C in the supplementary material), it is nonetheless unclear how third or fourth arch specific patterns can be achieved. Hoxa2- or Hoxa3-specific functions could result from a difference in their relative expression levels(quantitative difference) and/or target specificity (qualitative difference). In addition, qualitative and/or quantitative differences of expression of other Hox genes in NCCs and/or in PA epithelia may contribute to establish arch specific pattern. Indeed, Hoxd4 expression overlaps with Hoxd3 in arch 4 (see Fig. S1E,H,K in the supplementary material). Hoxd3 is in addition expressed at low levels in subsets of third arch NCCs (see Fig. S1E in the supplementary material)(Manley and Capecchi, 1997)and genetically interacts with Hoxa3(Condie and Capecchi, 1994). Such genetic interaction was supported by the deletion of the Hoxd cluster in the context of the conditional inactivation of Hoxa genes in NCCs, resulting in mild thyroid cartilage reduction and occasional fusions to hyoid cartilage(Fig. 2; not shown), and additional ectopic squamosal bones (Fig. 4; supplementary material Table S1).

However, the removal of the Hoxd cluster in a NCC-specific Hoxa-deleted background did not increase the extent of the homeotic phenotype by generating, for example, a supernumerary incus(Fig. 2). This might be due to the persistent expressions of Hoxb2 and Hoxb3 (see Fig. S4 in the supplementary material; and not shown), although their single or compound deletions were not sufficient to alter third or fourth arch skeletal pattern (Medina-Martinez et al.,2000). Similarly, the Hoxa2 knockout phenotype was not extended posteriorly by the additional deletion of Hoxb2(Davenne et al., 1999),despite its strong expression in third and fourth arches (see Fig. S1B in the supplementary material). Finally, Hoxa2 and Hoxa3 expression in the ectoderm and endodermal pouches of third and fourth arches could also have an influence on the patterning of skeletal derivatives. These observations emphasize the specificity of given Hox clusters for strong vertebrate adaptive traits, such as for the limbs or the teguments, as if ancestral genome duplications, which occurred at the base of the vertebrate radiation, had made it possible for Hox clusters to acquire specific roles(Duboule, 2007). In this context, the Hoxa cluster arguably has a primary role in the biology of skeletogenic neural crest cells. More specifically, our data strongly support a pivotal role for Hoxa3 and Hoxa2 in patterning third and fourth arch skeletogenic NCCs, whereas Hoxb and Hoxd genes may provide a`quantitative backup', which may become functionally relevant only in the absence of Hoxa genes (see also Rijli and Chambon, 1997).

It is noteworthy that while the duplicated Meckel's (MC2) was always truncated in single Hoxa2^-/- mutants, it was posteriorly extended and occasionally fully duplicated and fused directly to the MC3 in Wnt1::Cre/Hoxa^flox/flox newborns(Fig. 2E,F,I,J). This indicates that the absence of Hoxa3 function in the third PA enhanced the second PA phenotype observed in the Hoxa2 knockout. In support of this, the inactivation of Hoxa3 function alone already resulted in a reduction (or an absence) of the lesser horns of the hyoid bone(Chisaka and Capecchi, 1991), a structure of second arch origin. As the expression of Hoxa3 is not detected in the second PA, Hoxa3 function can thus influence the patterning of the second arch structures in a non cell autonomous manner,supporting the observation of patterning interactions between neighboring PAs(e.g. Gavalas et al.,1998).

Finally, an important conclusion of this work is that pre-otic and post-otic skeletogenic NCCs may share a common Hox-free ground pattern (model, Fig. 6). In all vertebrates analyzed, including jawless lampreys(Takio et al., 2004), the first mandibular arch is devoid of Hox gene expression, whereas the second arch expresses only PG2 Hox genes. Functional inactivation of PG2 Hox genes in mouse, as well as knockdown approaches in Xenopus and zebrafish(Baltzinger et al., 2005; Hunter and Prince, 2002) have revealed that mandibular and hyoid arches share a ground pattern, manifested in the absence of Hox gene function. Accordingly, Hoxa2 acts as a selector gene by modifying the underlying ground pattern to yield a hyoid specific pattern. Such a proposal was further supported by ectopic Hox PG2 gain-of-function in the first arch of amphibian and bird embryos(Grammatopoulos et al., 2000; Pasqualetti et al., 2000),which was sufficient to elicit the opposite transformation, namely to change mandibular into hyoid identity. These experiments highlighted the wide conservation, throughout vertebrate evolution, of the presumed Hox-free ground pattern between the first two arches.

However, and even though all PAs clearly belong to a metameric series,loss-of-function experiments involving PG3 Hox genes did not allow the assessment of whether or not posterior arch segments share the same ground pattern with the two rostralmost elements of the series. Here, we find that third arch patterning requires functional contributions from both PG3 and PG2 Hox genes, and that in the absence of Hoxa2 and Hoxa3function, third and fourth arch derivatives are partially replaced by jaw-like and other first arch-like structures. This observation thus extends the ground pattern corresponding to the rostralmost element of the metameric series, the mandibular arch, to more posterior segments. In the transition from jawless agnathan to jawed gnathostome vertebrates, the first arch underwent dramatic morphological changes. Our findings indicate that molecular changes that occurred in first arch skeletogenic NCCs and which resulted in a modification of the morphological ground pattern, e.g. the appearance of the jaw, were concomitantly `built-in' within the second and more posterior arches. Arch-specific Hox codes in turn induced particular morphological changes, thus linking morphological evolution of the mandibular arch to concomitant morphological changes of hyoid and posterior arches.

In the vertebrate hindbrain, the rhombomere ground pattern also corresponds to the rostralmost element of the series, i.e. rhombomere 1, which is devoid of Hox gene activity (Waskiewicz et al.,2002). The finding that hindbrain and PA ground patterns can be modulated by Hox gene functions to select segment-specific morphologies further underscores the metameric assembly of the pharyngeal region of the vertebrate head. Interestingly, in the short germ band beetle Tribolium, the complete deletion of Hox genes resulted in all head,trunk and abdominal embryonic segments acquiring the morphology of the rostralmost segment carrying antennae(Brown et al., 2002). Morphological co-evolution of a metameric series sharing the same ground pattern corresponding to that of the most anterior segment might thus be a rather widespread strategy in animal evolution.