sucker encodes a zebrafish Endothelin-1 required for ventral pharyngeal arch development

In all vertebrate embryos, segmental streams of cranial neural crest cells migrate to form the pharyngeal arches and differentiate into cartilages and bones of the head skeleton. In gnathostomes, the head skeleton has a dorsoventral (DV) polarity from its first appearance: dorsal and ventral cartilages of different shapes (such as the upper jaw and lower jaw in the first arch) are separated by joints. Fate-mapping studies in the avian embryo have shown that these dorsal and ventral skeletal elements arise from different anteroposterior levels of midbrain and hindbrain neural crest (Köntges and Lumsden, 1996), suggesting that before migration, the cranial neural crest is prepatterned with respect to its DV fate. Since the cranial neural crest has been shown to pattern morphogenesis of the paraxial mesodermally derived pharyngeal musculature (Noden, 1983a,b; Schilling et al., 1996), patterning information in the head periphery could be primarily contained within the premigratory neural crest. However, other experiments have established that neural crest cells are patterned by their environment after migration. For example, cartilage histogenesis of the neural crest requires contact with pharyngeal endodermal epithelium (reviewed in Hall, 1978) and chondrification of the lower jaw is inhibited by surface ectoderm (Kollar and Mina, 1991; Mina et al., 1994). Thus, pharyngeal arch development requires bidirectional signaling between the cranial neural crest and its environment.

Targeted mutations in mice have revealed a signal present in the environment of the cranial neural crest which is required for ventral neural crest-derived skeletal fates. Inactivation of genes encoding either the 21-amino-acid secreted ligand Endothelin-1 (Et-1), the Endothelin type A receptor (EdnrA), or the Endothelin converting enzyme (Ece-1), produces an identical craniofacial phenotype in which cartilages in the first and second arches are deleted or mispatterned (Kurihara et al., 1994; Clouthier et al., 1998; Yanagisawa et al., 1998). This common phenotype of the Et-1 and EdnrA mutant mice is remarkable given the complementary expression profiles of these genes in the arch primordia. While Et-1 is expressed in a central mesenchymal core of arch paraxial mesoderm, EdnrA is expressed in a surrounding shell of mesenchymal postmigratory neural crest cells. Et-1 is also expressed in epithelia (both surface ectoderm and pharyngeal endoderm) surrounding this entire concentric arrangement of mesenchyme (Maemura et al.,1996; Clouthier et al., 1998). The germ layer derivations of the mesenchymal populations expressing these genes are inferred from fate-mapping studies, which have established that paraxial mesoderm occupies the core of an arch and is surrounded by neural crest (Trainor et al., 1994; Trainor and Tam, 1995; Hacker and Guthrie, 1998). In both Et-1 and EdnrA mutant mice, skeletal defects are preceded by much earlier defects in gene expression; for instance, dHAND, which encodes a bHLH transcription factor, is not expressed as it normally would be in ventral arch presumptive postmigratory crest (Thomas et al., 1998). Hence, the mutant phenotypes and expression patterns suggest a model in which Et-1 in the environment of the cranial neural crest specifies ventral crest fates.

This model is further supported by work in avian embryos, which has shown that pharmacological inactivation of EDNRA results in severe ventral arch one and two deletions, creating a chicken with no lower beak (Kempf et al., 1998). As in mice, chick Et-1 and EdnrA are expressed in arch cores and epithelia, and postmigratory neural crest, respectively (Nataf et al., 1998; Kempf et al., 1998). Thus, both this anatomical arrangement of EdnrA-expressing cylinders of neural crest cells surrounding a central core of Et-1-expressing paraxial mesoderm and a required function for EDNRA in lower jaw development have been conserved between birds and mammals and, thus, likely date back at least 350 million years to the Carboniferous, the time of the last common ancestor of both mammals and the diapsids from which birds derive. The appearance of jaws earlier in vertebrate evolution, around 450 million years ago in the Ordovician, sparked the gnathostome radiation, which generated most of the vertebrates alive today (reviewed in Mallatt, 1997). The conservation of Et-1 signaling for lower jaw development in chicks and mice raises the possibility that Et-1’s requirement for lower jaw development is much more ancient and shared within all gnathostomes. If so, Et-1 should be required for lower jaw development in the most divergent gnathostomes, sharks and bony fish.

Genetic screens in the zebrafish have revealed a large number of loci required for pharyngeal arch development (reviewed in Schilling, 1997). Four recessive loci isolated in the Tübingen screen were placed into a single phenotypic class based on their similar mutant phenotypes (Piotrowski et al., 1996; and see Discussion). sucker (suc) mutants have the most severe phenotype of the four: the lower jaw and other ventral cartilages are drastically reduced, misshapen and fused to the relatively unaffected dorsal cartilages of the same arch (Piotrowski et al., 1996; Kimmel et al., 1998).

Here we present molecular and phenotypic characterization of suc. We show that a missense mutation in the secreted domain of an et-1 ortholog cosegregates with the suc mutant phenotype, and wild-type, but not mutant, et-1 rescues suc mutants. Together these results indicate that the suc mutant phenotype is due to this mutation in et-1. Injection of human ET-1 into an arch at stages after neural crest has migrated also rescues the suc mutant phenotype; thus, suc/et-1 is not required for migration of most, if not all, neural crest. Rather, suc/et-1 is required in the postmigratory environment of the neural crest where, like chick and mouse Et-1, suc/et-1 is expressed in arch cores and epithelia. Zebrafish, like mice, require et-1 signaling for the ventral arch expression of dHAND, msxE, gsc and dlx3. We find that ventral expression of dlx2 and EphA3 also requires suc/et-1, while dorsal domains of dlx2 and gsc are suc/et-1-independent. Patterning of cranial mesodermal and endodermal tissues also requires suc/et-1 since ventral premyogenic condensations, ventral pharyngeal muscles and the hyomandibular pouch are all mispatterned in suc/et-1 mutants. With mosaic analyses, we demonstrate that suc/et-1 is nonautonomously required in cranial neural crest cells for both early dHAND and EphA3 expression, and for later formation of ventral pharyngeal cartilages. suc/et-1 is also nonautonomously required in mesoderm to make correctly patterned ventral muscles. Our findings show that Et-1 signaling plays an essential role in lower jaw formation, which has been widely conserved within gnathostomes. Further, the results support a model in which the signaling occurs during a conserved transient anatomical arrangement: hollow neural crest cylinders surround central cores of paraxial mesoderm, and ventral neural crest requires Et-1 from its surrounding environment to make a lower jaw and other ventral arch fates.

Fish were raised under standard conditions (Westerfield, 1995) and staged as described (Kimmel et al., 1995). suc^tf216b mutants (Piotrowski et al., 1996) were outcrossed to the Oregon ABC strain for phenotypic analyses. For some of the analyses at larval stages, mutants were outcrossed to golden (gol) (Streisinger et al., 1981) to obtain double mutant embryos with reduced pigmentation. No additional phenotypic differences were observed between suc mutants and suc; gol double mutants.

Cartilage staining, dissection and flat-mounting were done as described (Kimmel et al., 1998).

In situ hybridization was performed essentially as described (Thisse et al., 1993) with the following modifications: probes were not hydrolyzed, the glycine stop step was omitted, hybridization temperature was 66.5°C, two 30 minute 0.1× SSC washes were done after the 0.2× SSC washes, and older embryos were permeabilized by treating with 10 μg/ml Proteinase K treatment for 1 to 30 minutes depending on age. Probes used were: dlx2 and dlx3 (Akimenko et al., 1994), dHAND (Angelo et al., 2000), msxE (Ekker et al., 1997), gsc (Schulte-Merker et al., 1994), EphA3 (Xu et al., 1995), MyoD (Weinberg et al., 1996), and shh (Krauss et al., 1993). For suc/et-1 in situs, probe was made from EST clone fb14d01 (see below). Deyolking was done manually with tungsten insect pins. Embryos were cleared in 70% glycerol, mounted on bridged coverslips and photographed on a Zeiss Axiophot microscope. Occasionally, embryos were raised in 0.003% PTU (1-phenyl 2-thiourea) to inhibit melanogenesis (Westerfield, 1995). Two-color in situs were done as described (Jowett and Lettice, 1994) with the modifications of Hauptmann and Gerster (1994) and the same modifications as above. For sectioning, embryos were embedded in Epon and sectioned at 5 μm.

Larval muscles were stained with the 1025 anti-myosin antibody (generous gift of Drs S. Hughes and H. Blau) as described by Schilling and Kimmel (1997).

We mapped suc to LG19 using a single large early pressure (EP)-derived gynogenetic clutch (Streisinger et al., 1981) obtained from a suc heterozygous female, which was generated by crossing suc heterozygotes to the WIK mapping strain (Knapik et al., 1996). Of 95 total diploid animals, 18 were by morphology suc mutants. DNA was prepared from each of these animals essentially as described (Johnson et al., 1994) and used for centromere linkage analysis (Johnson et al., 1996) with mapped z-markers and gene RFLPs (Knapik et al., 1998; Postlethwait et al., 1998) using PCR conditions as described (Johnson et al., 1994).

EST clone fb14d01 (M. Clark and S. Johnson, WUZGR; http://zfish.wustl.edu) was purchased from Research Genetics. This clone contains a 563 bp ORF of a zebrafish Endothelin-1 gene (GenBank accession number AF281858). Using gene-specific primers 5′-GTGACCACAGAAATGGCGATTA-3′ and 5′-TCTGCA-ATCAGGGACTCTAG-3′ from the sequence of this EST, we screened a PAC library (Amemiya and Zon, 1999) by PCR (3 minute denaturation at 94°C followed by 37 cycles of 20 second 94°C denaturation, 20 second 53°C annealing, 30 second extension at 72°C). All PCR reactions were done on a MJ Research PTC-100 or PTC-200 using conditions essentially as described (Johnson et al., 1994). A single PAC, 16A1, was isolated which contained all five exons spanned by this EST. cDNAs were obtained from clutches of suc mutant embryos by RT-PCR. Total RNA was isolated using standard methods (Chomczynski and Sacchi, 1987). Primer 5′-CCTGAAATGCATGACGTGTG-3′ was used for first-strand cDNA synthesis using Superscript II reverse transcriptase (Gibco). RT-PCR was done with this reverse primer and forward primer 5′-AATACGGGACTTGCATACTACA-3′ (3 minute 94°C denaturation, followed by 36 cycles of 15 second denaturation at 94°C, 15 second annealing at 56°C, 1 minute extension at 72°C). These 659 bp RT-PCR products were cloned with a TOPO TA kit (Invitrogen) into pCR2.1, and this wild-type cDNA clone is referred to as pET1. All sequencing was done on an ABI automated sequencer.

For DNA injections, DNA was diluted in 0.2 M KCl and 0.25% Phenol Red and pressure-injected asymmetrically into the yolk at the 2-to 4-cell stage. A 677 bp EcoRI fragment of pET1 (wild-type Et-1 cDNA), containing 69 bp upstream of the predicted ATG start codon and 65 bp downstream of the predicted stop codon, was subcloned into the EcoRI site of pCS2, to make pCS2-ET1. Injecting this construct allows unambiguous genotyping of injected animals with intronic primers flanking the et-1 MseI RFLP, since the intronic primers do not amplify the injected wild-type cDNA. Point mutagenesis was done on pCS2-ET1 using Promega’s Gene Editor Site-Directed Mutagenesis kit and mutagenic primer 5′-GCAAGTTTTCTGGTTAAAGAGTGCGTCTAC-3′ (mismatch in bold) to create pCS2-ET1D8V. Injected animals were screened at 4 days for head morphology, and all potentially rescued mutants were anesthetized, sacrificed and bisected, their heads fixed for Alcian staining of cartilages and their tails used for making DNA for PCR-genotyping with primers 5′-GCGACAAATTCAATCATCTCTAG-3′ and 5′-ACAGTTCATAACTGATGGTATTTG-3′. A PCR program of an initial 3 minute 94°C denaturation, followed by 35 cycles of a 20 second 94°C denaturation, a 15 second 55°C annealing, and a 25 second 72°C extension generates a 300 bp band, which, when digested with MseI, allows genotyping for the Asp-to-Val polymorphism (see Fig. 2E,F). Since injected DNA is inherited mosaically (see Westerfield et al., 1992; Kroll and Amaya, 1996), only a fraction of animals injected with this construct have expression driven in the head (data not shown).

For protein rescue, human recombinant mature ET-1 (Sigma) was diluted in distilled water (0.5 mg/ml) and pressure-injected into pharyngeal arch primordia through a glass micropipette using a picospritzer (Applied Instruments). Injected embryos were raised to 4 days for Alcian staining of head cartilages and PCR-genotyping of tails as described above.

suc mutant heterozygous carriers were intercrossed and donor embryos labeled at 1-to 2-cell stages with 3% biotinylated dextran. Cells were transplanted into unlabeled sibling hosts either at late blastula (mesoderm) or 3 to 5 somites (neural crest), by mounting them in 3% methyl cellulose and transferring cells using a suction micropipette as described previously (Hatta et al., 1990 for mesoderm transplants; Schilling et al., 1996 for crest transplants). Donor cells were detected with an ABC kit (Vectostain) using either a DAB substrate, or a tyramide substrate, which was detected by fluorescence (Moens and Fritz, 1999). Images were captured on a Zeiss Axioplan 2 fluorescence microscope using a CCD camera and Macintosh G3 equipped with Improvision imaging software.

sucker (suc) is one of a class of four mutations that disrupts patterning in the two anteriormost pharyngeal arches, the mandibular and hyoid. In homozygous suc mutant larvae, ventral cartilages (Meckel’s cartilage in the mandibular arch and the ceratohyal in the hyoid) are severely reduced and fused to the dorsal cartilages of the same arches (Fig. 1, Piotrowski et al., 1996; Kimmel et al., 1998). Quantification of this phenotype shows that the expressivity of severe ventral reductions in the first two arches is 100% in suc mutants (Table 1). The ventral cartilages in arches 3 and 4 are also dramatically reduced with high expressivity in suc mutants, while cartilages in the three most posterior arches are spared (Table 1).

The ventral pharyngeal cartilage reduction seen in suc mutants is reminiscent of the phenotype seen in mice with targeted mutations in Endothelin-1 (Et-1), where homologous ventral (distal) skeletal elements (such as Meckel’s cartilage) are dramatically reduced (Kurihara et al., 1994). As an initial step towards the molecular identification of the sucker locus, we mapped suc to linkage group 19 (LG19) (Fig. 2A), distal to three previously mapped zebrafish genes that are homologs of human genes on chromosome 6 (Fig. 2B). In humans, ET-1 maps to chromosome 6, distal to the syntenic region described above (Fig. 2B). Based on the phenotypic similarities between the Et-1 mutant mice and suc mutants, as well as this potential synteny, we tested a zebrafish et-1 as a candidate for suc. An EST encoding a predicted protein highly similar to mammalian ET-1 (Fig. 2C) was generated by Washington University Zebrafish Genome Resources (http://zfish.wustl.edu) and mapped close to suc on LG19. Eighteen of the 21 amino acids in the predicted secreted domain are conserved between human and zebrafish (Fig. 2C,D). Sequencing of the et-1 cDNA and genomic coding regions in suc mutant embryos revealed a missense mutation which replaces aspartate with valine at the highly conserved 8th residue of the 21 amino acid secreted domain (Fig. 2D,E). This transversion creates a MseI site, which cosegregated with the suc phenotype in over 500 diploid animals (Fig. 2F and data not shown). The cosegregation places this MseI RFLP and suc within 0.2 cM of one another, suggesting that the suc mutant phenotype is due to this Asp8-to-Val8 missense mutation.

Supporting the hypothesis that suc is et-1, injection of wild-type et-1 DNA rescued pharyngeal defects in suc mutants. Injection of a genomic fragment containing the entire wild-type et-1 locus (PAC 16A1) asymmetrically into 2-to 4-cell embryos rescued ventral cartilage defects unilaterally in suc mutants (Table 2). Similar injections of a construct driving expression of wild-type et-1 RNA (pCS2-ET1) also unilaterally rescued the cartilage phenotype in suc mutants (Fig. 3A-F and Table 2). Despite the inherent mosaicism of DNA injections (see Westerfield et al., 1992; Kroll and Amaya, 1996), injections of pCS2-ET1 rescued ventral cartilage formation in a high percentage (nearly one-quarter, Table 2; Fig. 3G) of suc mutants. An otherwise identical construct with the Asp8-to-Val8 missense mutation in et-1 (pCS2-ET1D8V) did not rescue ventral cartilage formation in suc mutants under similar injection conditions (Table 2). These data, together with our sequencing and cosegregation results indicate that the suc mutant phenotype is due to this missense mutation in et-1, and for clarity in this paper we refer to this gene as suc/et-1.

The Et-1 signal might be required early for migrating precartilaginous neural crest or later within the arch primordium. To determine if Et-1 is sufficient to rescue ventral cartilage formation at a stage after neural crest migration, we injected human recombinant ET-1 protein into suc/et-1 mutant arch primordia at 20 or 28 hours, two stages after most cranial neural crest has migrated (Schilling and Kimmel, 1994). Injections at both time points rescued formation of first and second arch ventral cartilages (Meckel’s and ceratohyal) (Table 2, Fig. 3H). Thus, even though this experiment does not reveal the critical period when suc/et-1 normally functions, it suggests that Et-1 is not required for ventral neural crest migration, but rather for correct specification of ventral postmigratory neural crest.

To determine the spatiotemporal profile of embryonic suc/et-1 expression, we examined suc/et-1 mRNA distribution by whole-mount in situ hybridization. Whereas cranial neural crest first begins to migrate at about 12 hours in zebrafish (Schilling and Kimmel, 1994), suc/et-1 expression in the head periphery was not detected until 16-18 hours. At this stage, small bilateral groups of mesenchymal cells adjacent to the midbrain and hindbrain express suc/et-1 (Fig. 4A,B). Expression persists in mesenchymal cells in this location at 24 hours and, by this time, appears segmental, with one suc/et-1-expressing cluster of cells in each of the first two arches (Fig. 4C). At this stage, expression is also detectable in ventral pharyngeal endodermal epithelia of the second and third pharyngeal pouches (Fig. 4C) and in ventral surface ectodermal epithelium (data not shown).

In each of the first two arches at 30 hours, the suc/et-1- expressing mesenchyme has coalesced into a discrete cluster located ventrally (Fig. 4D). Ventral views and horizontal sections show that these mesenchymal clusters lie in a central ‘core’ position (Fig. 4E,F; see Introduction). By 30-32 hours, a mesenchymal cluster of suc/et-1-expressing cells is also detected in the core of arch 3 (Fig. 4D,F). These arch mesenchymal cores of suc/et-1-expressing cells are surrounded by non-expressing mesenchymal cells (Fig. 4E,F), which by position appear to be neural crest (see below). Epithelial expression persists and, at 30 hours, expression is present in pharyngeal pouches 2 and 3 and is now detected in the fourth pharyngeal pouch (Fig. 4D). At these stages, suc/et-1 expression in ventral surface ectoderm appears continuous with expression in the pharyngeal pouches (Fig. 4F).

At 36 hours, the ventral mesenchymal cores of the arches continue to express suc/et-1 and appear to have elongated in the mediolateral direction (Fig. 4H). Epithelial expression in both ventral surface ectoderm and pharyngeal pouches 2-4 is still detectable at this stage (Fig. 4G). Within the pharyngeal pouches, expression is restricted to posterior, ventral epithelia (Fig. 4G).

At 48 hours, we no longer detect suc/et-1 expression in central arch mesenchymal domains. However, expression persists in pharyngeal pouches 2-4 and is now detected weakly in the first pharyngeal pouch (data not shown).

suc/et-1 is also expressed in other tissues, including many associated with the developing vasculature. From 24 to 36 hours, suc/et-1 is expressed in cells lining the paired dorsal aortas and transverse sections show many of these have an endothelial morphology (Fig. 4I and data not shown). At 48 hours, suc/et-1 expression was detected in cells lining blood vessels around the eye and in the midbrain (data not shown).

suc/et-1 expression is also detectable in cells in the otic vesicle from 24 to 48 hours (Fig. 4C,D, data not shown), in anterior cells of the pectoral fin rudiment at 48 hours (Fig. 4J), and in discrete bilateral clusters of cells adjacent to the boundary of the yolk ball and yolk extension at 24 hours (Fig. 4K).

In PCR-genotyped suc/et-1 mutants, the onset of suc/et-1 mRNA expression at 18 hours appears normal and expression is detectable through 48 hours, although the expressing tissues are disorganized at later stages. At 30 hours in suc/et-1 mutant embryos, the mesenchymal core domains express suc/et-1, as does both ventral surface ectoderm and pharyngeal pouches 2-4 (Fig. 4L,M). Expression in other tissues such as the otic vesicle is also not noticeably reduced in suc/et-1 mutant embryos (Fig. 4L and data not shown).

Since the pharyngeal cartilages affected in suc/et-1 mutants are derived from cranial neural crest (Schilling and Kimmel, 1994), we analyzed expression of neural crest markers to determine at what stage this population is mispatterned in suc/et-1 mutant embryos. Early stages appear normal: three markers of presumptive premigratory cranial neural crest, fkd6 (Odenthal and Nüsslein-Volhard, 1998), sna2 (Thisse et al., 1995) and dlx2 (Akimenko et al., 1994) showed no defect in suc/et-1 mutants from 10-14 hours (data not shown).

Two genes encoding transcription factors that are required for pharyngeal arch development in mice and expressed in postmigratory cranial neural crest are dlx2 and dHAND (Qiu et al., 1995; Thomas et al., 1998). In wild-type zebrafish, orthologs of these genes are also expressed in postmigratory arch crest (Akimenko et al., 1994; Angelo et al., 2000). dHAND expression is confined to a ventral subset of dlx2-expressing cells (Fig. 5A). Horizontal sections of dlx2 expression at 28 hours reveal that, within each arch, dlx2-expressing cells surround a central core of non-expressing cells (Fig. 5B). Similarly, viewing whole-mount embryos from a ventral aspect reveals that, in each arch, dHAND-expressing cells are also arranged cylindrically and surround a non-expressing core of cells (Fig. 5C). Ventrally, dHAND expression is strikingly complementary to suc/et-1 expression: core mesenchyme expresses suc/et-1 but not dHAND, and peripheral mesenchyme expresses dHAND but not suc/et-1 (Fig. 5C,D). The surrounding epithelia (both ventral surface ectoderm and pharyngeal endoderm) also express suc/et-1, but not dHAND (Fig. 5D).

No dlx2 expression defect was detected in suc/et-1 mutants at early premigratory stages (see above). However, by 28 hours, we could unambiguously sort mutants (verifying genotypes by PCR, see above) by a reduction in the dorsoventral length of the dlx2 expression domains in the mandibular and hyoid arches (Fig. 5E,F). Dorsal cells in suc/et-1 mutant arches express dlx2 at normal levels (Fig. 5E,F). Hence the reduction is of the ventral domain, due to either an absence of cells ventrally, or a failure of ventral cells to express dlx2. The latter appears to be the case, since in sections of dlx2 expression in suc/et-1 mutants, ventral first and second arch crest cells appear to be present but fail to express dlx2 (Fig. 5G,H). In accord with this interpretation, in suc/et-1 mutants, unlabeled mesenchymal cells surround the suc/et-1-expressing cores (Fig. 4M). Furthermore, preliminary TUNEL labeling (data not shown) revealed no significant cell death in the arches of suc/et-1 mutants. Therefore three lines of evidence suggest that postmigratory neural crest cells are present in the ventral arches of suc/et-1 mutants.

dHAND arch expression is severely reduced in the Et-1 mutant mice (Thomas et al., 1998). Similarly, in zebrafish, the expression of dHAND in ventral arch mesenchyme is dramatically reduced in suc/et-1 mutant embryos at 28 hours (Fig. 5I-L). Expression of dHAND is often maintained at a low level in the posterior half of the second arch in suc/et-1 mutants (Fig. 5K-L), suggesting additional genes are required for arch dHAND expression. In wild types, while dHAND is also expressed in the developing heart and fin, neither of these domains are noticeably affected in suc/et-1 mutants (Fig. 5K-L and data not shown).

Two other transcription factors, Msx1 and Gsc, are required for mammalian craniofacial development (Satokata and Maas, 1994; Yamada et al., 1995; Rivera-Perez et al., 1995). gsc, like dHAND, requires Et-1 signaling for expression in mouse arch primordia (Clouthier et al., 1998). In contrast, msx1 arch expression is unaffected in the Et-1 mutant mice, although msx1 has been indirectly shown to be downstream of Et-1 signaling (Thomas et al., 1998; and see Discussion). To test if this genetic hierarchy has been conserved between fish and mammals, we examined expression of msxE and gsc in suc/et-1 mutant embryos. In wild-type zebrafish embryos at 24 hours, msxE is expressed similarly to dHAND, in ventral but not dorsal postmigratory arch neural crest (Ekker et al., 1997; Fig. 6A). Expression in suc/et-1 mutants is severely reduced, although other non-arch domains are unaffected (Fig. 6B).

goosecoid (gsc) expression in zebrafish marks a complex pattern of dorsal and ventral arch mesenchyme from 26-42 hours (Schulte-Merker et al., 1994; Fig. 6C and data not shown). In the hyoid arch of 30 hour wild types, these dorsal and ventral gsc-expressing domains of cells are separated by a non-expressing domain, which seems to prefigure the joint between the dorsal and ventral cartilages (Fig. 6C). While gsc is expressed dorsally in suc/et-1 mutant embryos, ventral expression is largely abolished (Fig. 6D).

The homeobox transcription factor dlx3 (Akimenko et al., 1994) and the ephrin receptor tyrosine kinase EphA3 (Xu et al., 1994) are also expressed in zebrafish ventral arch postmigratory neural crest and we find that expression of both requires suc/et-1 function (Fig. 6E-H). In wild types, arch expression of dHAND, msxE, dlx3, and EphA3 are all initiated from 18-24 hours, well after dlx2 (Ekker et al., 1997; Akimenko et al., 1994; data not shown). Defects in msxE, dlx3 and EphA3 were detected at 24 hours, while the gsc defect was apparent at 26 hours, when its ventral arch expression is first initiated in wild types (data not shown). dHAND, msxE, dlx3 and EphA3 are all also expressed in ventral postmigratory neural crest in more posterior arches (Figs 5, 6). In suc/et-1 mutants at 24 hours, gene expression defects are usually localized to the first two arches but, by 36 hours, severe defects are also seen in the third and fourth arches (data not shown).

To assay potential requirements for suc/et-1 in cranial development outside of the neural crest, we examined arch muscles and endodermal derivatives in suc/et-1 mutants. In wild types, both dorsal and ventral pharyngeal muscle precursors are detectable at 54 hours by expression of myoD (Schilling and Kimmel, 1997). However, in suc/et-1 mutants at the same stage, the first and second arch ventral myoD-expressing muscle precursors were severely reduced (Fig. 7A-D). A few scattered myoD-expressing cells are present in the ventral arches of suc/et-1 mutants, but the ventral premyogenic condensations fail to form (Fig. 7B,D). In contrast, forming dorsal muscles of the mandibular and hyoid arches express myoD and appear unaffected in suc/et-1 mutants (data not shown).

Differentiated ventral arch muscles are also severely reduced in suc/et-1 mutant larvae, as determined by immunostaining of 5-day-old larvae with the 1025 antibody to muscle myosins (Fig. 7E-H). The ventral first arch muscles (intermandibularis anterior and posterior) are reduced to a few fibers that span the midline (Fig. 7F), and the ventral second arch muscles (interhyoideus and hyohyoideus) fail to extend to the midline as they do in wild types and are greatly reduced in size. In contrast to these ventral disruptions, dorsal muscles are relatively unaffected and, as normal, attach to the dorsal cartilages (Fig. 7G,H).

Defects are also present in the pharyngeal endoderm of suc/et-1mutants. sonic hedgehog (shh) is expressed in wild-type pharyngeal endodermal cells beginning around 33 hours (Krauss et al., 1993), at which time no defect in suc/et-1 mutant embryos was detected (data not shown). However, by 54 hours, a shh expression defect is detectable: expression in the first pharyngeal pouch does not extend as far laterally in mutants as it does in wild types (Fig. 7I,J). Furthermore, a diverticulum forming at the ventral midline of the arch one/two pharyngeal endoderm boundary, perhaps the thyroid rudiment, is visible in lateral views of 54 hour embryos but is malformed in suc/et-1 mutant embryos (Fig. 7K,L). Thus, in addition to suc/et-1’s early requirement for ventral neural crest patterning, suc/et-1 is also required to pattern later arch mesodermal and endodermal derivatives.

Since ventral arch neural crest is severely mispatterned in suc/et-1 mutant embryos, yet suc/et-1 expression was not detected in the neural crest, we predicted neural crest cells would require suc/et-1 function nonautonomously. To test this prediction, we performed mosaic analyses, transplanting labeled suc/et-1 mutant premigratory cranial neural crest cells into unlabeled wild-type hosts (Fig. 8A). Neural crest cells from suc/et-1 mutants grafted into wild-type crest at early somite stages migrated into the peripheral, dlx2-expressing region of the arches and, when located ventrally, expressed ventral markers such as dHAND and EphA3 (Fig. 8C,E; Table 3). In similar suc/et-1 mutant-to-wild type neural crest transplants, suc/et-1 mutant cells also contributed to normal ventral cartilages in wild-type larvae (Fig. 8G, Table 3). Thus suc/et-1 functions nonautonomously in the neural crest, and the skeletal defects apparently arise indirectly because of a missing signal in the crest environment.

Since we previously fate-mapped head mesoderm on the zebrafish gastrula (Kimmel et al., 1990), we could test a mesodermal role for suc/et-1 by performing mesodermal transplants between wild-type and suc/et-1 mutant embryos at the gastrula stage (Fig. 8B). Both wild-type and suc/et-1 transplanted mesendodermal cells spread widely throughout the mesendoderm of a wild-type host and cells in the arches tend to occupy the central region of the arch primordium at 24 hours (Fig. 8D, data not shown), consistent with the hypothesis that paraxial mesoderm occupies the core of each arch. Transplanted cells that contributed to posterior branchial arches, which are developmentally younger, suggest that initially within the arch primordium, the neural crest is located lateral to the paraxial mesoderm (Fig. 8F).

Although these mesendodermal transplants might have contained a few pharyngeal endoderm cells, we did not observe induction of ventral neural crest markers dHAND or EphA3 in suc/et-1 mutant hosts in response to the presence of wild-type mesendoderm (Table 3). Furthermore, in the converse transplant, transplanted suc/et-1 mutant mesodermal cells were able to form normal ventral muscles in a wild-type host (Fig. 8H, Table 3) and did not result in abnormal cartilage development. Thus, suc/et-1 appears to also function nonautonomously in the paraxial mesoderm.

We have cloned and characterized zebrafish sucker (suc), mutation of which results in severe reduction of the lower jaw and other ventral pharyngeal arch cartilages. Our sequencing, cosegregation and rescue experiments indicate that the suc mutant phenotype is due to a missense mutation at a conserved residue within the secreted domain of Endothelin-1. Like mouse and chick Et-1, suc/et-1 is expressed in pharyngeal arch mesenchymal cores and epithelia. These mesenchymal cores are surrounded by dlx2 and dHAND-expressing putative neural crest cells, suggesting that the cores consist of paraxial mesoderm. Like mouse Et-1, suc/et-1 is required for ventral arch fates, including expression of dHAND, msxE, gsc and dlx3. Additionally, in zebrafish ventral expression of dlx2 and EphA3 also requires suc/et-1. Dorsal expression of dlx2 and gsc is suc/et-1 independent, and later patterning of arch mesodermal and endodermal derivatives also requires suc/et-1 function. Lastly, our mosaic analyses demonstrate that suc/et-1 functions nonautonomously in both neural crest cells and paraxial mesoderm. Collectively our results support a model (Fig. 9) in which ventrally-restricted Et-1 expression in both paraxial mesodermal arch cores and surrounding epithelia specifies ventral arch crest fates, including the ventral expression of dlx2, dHAND, msxE, gsc, dlx3 and EphA3 and the later formation of ventral cartilages.

Biochemical evidence indicates that the Asp8-to-Val8 missense mutation in suc^tf216b is likely to be a loss-of-function mutation. Alanine scanning of the 21-amino-acid human mature ET-1 showed that Asp8 is one of only five residues which, when replaced with alanine, resulted in a nonfunctional protein (Tam et al., 1994). Hence, replacing this highly conserved charged residue of zebrafish Et-1 with a noncharged residue such as valine would be expected to create a nonfunctional protein. Our rescue experiments also suggest tf216b is a loss-of-function mutation, since injection of pCS2-ET1 (wild-type Et-1), but not pCS2-ET1D8V (Et-1 with Asp-to-Val missense mutation at residue 8 of secreted domain) rescued ventral cartilage formation in suc mutants. Further testing of the strength of the tf216b allele will be possible once additional alleles of suc are found, or a Df(suc) becomes available. Since suc/et-1 is still expressed in suc/et-1 mutants, this missense mutation does not appear to result in fewer or more unstable transcripts, and suggests that suc/et-1 does not indirectly autoregulate its own transcription.

Mutations of three other ‘anterior arch’ loci (schmerle, sturgeon and hoover) cause similar craniofacial defects as those seen in suc/et-1 mutants (Piotrowski et al., 1996). We subsequently described a common phenotypic spectrum of flat-mounted larval hyoid cartilage shapes in sucker, schmerle and sturgeon mutants and noted the resemblance of these mutant phenotypes to phenotypes seen in mice with mutations in Et-1 signaling components (Kimmel et al., 1998). In mice, mutation of Et-1, EdnrA or Ece-1 produces similar craniofacial defects, suggesting that, in zebrafish, some of the other anterior arch loci represent ednrA or ece-1 genes, a possibility that we are currently exploring. However, the Ece-1 mutant mice also have pigment and enteric neuron defects (Yanagisawa et al., 1998) that have not been reported for any of the zebrafish anterior arch mutants. Furthermore, the EdnrA and Ece-1 mutant mice have defects in cardiac neural crest-derived tissues such as the outflow tract (Clouthier et al., 1998; Yanagisawa et al., 1998), while no circulatory system defect has been reported for any of the anterior arch mutants. Two of the other anterior arch mutations (schmerle and hoover) result in shorter pectoral fins (Piotrowski et al, 1996), which is interesting in light of suc/et-1 expression in the developing pectoral fin. Although suc/et-1 mutants display no obvious defects in the pectoral fin or other non-arch suc/et-1-expressing tissues, these tissues have not been examined comprehensively in suc/et-1 mutants. suc/et-1 mutants live up to 9 days, long enough to detect defects in derivatives of these suc/et-1-expressing tissues. These tissues in zebrafish might not require Et-1 signaling. Alternatively, the additional genomic duplication in the teleost lineage (Postlethwait et al., 1998; Amores et al., 1998) might have generated genes redundant for these developmental requirements. Redundant functions can be revealed by double mutant analyses and further motivates the isolation of more anterior arch mutants. Forward genetic screens in zebrafish could also reveal genes that have not yet been implicated in the Et-1 signaling pathway in mice. Our ongoing head cartilage screen has uncovered a fifth anterior arch locus (C. T. M. and C. B. K., unpublished), yet no more alleles of the existing four, suggesting that the screens have not approached saturation.

Trainor and Tam (1995) showed that, in mammalian embryos, neural crest surrounds the paraxial mesodermal core of the arch. In chick embryos, Noden (1983b) demonstrated that head paraxial mesoderm makes most of the muscles of the pharynx and Hacker and Guthrie (1998) described the migration of cranial paraxial mesoderm into the center of each pharyngeal arch. Our mosaic analyses support a neural crest origin for peripheral mesenchyme as well as a paraxial mesodermal origin for the central arch mesenchymal core.

Gene expression patterns in amniote embryos also reveal separate central and peripheral arch mesenchymal populations. In mice and chicks alike, Et-1 is expressed in a central arch core surrounded by non-Et-1-expressing mesenchyme (Maemura et al., 1996; Clouthier et al., 1998; Nataf et al., 1998). In the arches of mice, dlx2 is expressed in mesenchyme surrounding a central non-expressing core (Bulfone et al., 1993). We show that, in zebrafish, suc/et-1 is expressed in central arch mesenchyme, while dlx2 is expressed in peripheral mesenchyme. Thus, the general spatial arrangement of Et-1 and dlx2 expression in arch mesenchyme has been conserved between fish and amniotes. A higher resolution analysis of dlx2 and Et-1 expression, with serial sections and double-labeling experiments, would further test this model of distinct arch mesenchymal populations revealed by expression of these two genes. We propose that the suc/et-1-expressing ventral arch cores in the first two arches correspond to the premyogenic condensations intermandibularis and constrictor hyoideus ventralis, respectively, described by Edgeworth (1935). According to this proposal, suc/et-1 expression marking ventral muscle precursors parallels engrailed expression marking dorsal muscle precursors (constrictor dorsalis) in the first arch (Miyake et al., 1992; Hatta et al., 1990; Edgeworth, 1935).

Together, our expression and mosaic analyses support a model (Fig. 9) in which arch postmigratory neural crest cells form a hollow cylinder surrounding central cores of paraxial mesoderm. Within each arch, ventral paraxial mesoderm expresses suc/et-1 and is surrounded by ventral postmigratory neural crest. dlx2 and gsc are expressed in both dorsal and ventral postmigratory arch crest, although only the ventral domains are suc/et-1-dependent. dHAND, msxE, dlx3 and EphA3 expression is ventrally-restricted and also requires suc/et-1. Both these gene expression patterns and these tissue arrangements argue for a conserved anatomical arrangement of arch mesenchyme that occurs transiently during arch development in all gnathostomes.

The Et-1 mutant mice have reduced dHAND arch expression while the dHAND mutant mice have no arch msx1 expression, leading Thomas et al. (1998) to propose a pathway involving these three genes. Surprisingly, the Et-1 mutant mice have normal msx1 expression, possibly due to low levels of dHAND expression, which are sufficient to activate msx1 at normal levels (Thomas et al., 1998). Our results demonstrate that, in zebrafish, both dHAND and msxE expression require suc/et-1. The orthologies of the five zebrafish msx genes and the three mammalian msx genes are unclear (Ekker et al., 1997).

However, since msx2 expression in both the Et-1 and dHAND mutant mice is normal (Thomas et al., 1998), we propose that, in the arches, msxE is the functional equivalent of mammalian msx1 and that, in zebrafish, expression of this msx gene more directly requires suc/et-1.

In zebrafish, gsc expression requires suc/et-1 ventrally, but not dorsally. Similarly, gsc expression is undetectable in E10.5 EdnrA mutant mice (Clouthier et al., 1998). Although no ventral specificity to this defect has been reported in mice, later in development, gsc is expressed in dorsal arch tissues (Gaunt et al., 1993) and gsc mutant mice have both dorsal and ventral craniofacial defects (Yamada et al., 1995; Rivera-Perez et al., 1995). Perhaps a later dorsal gsc expression domain in mice is homologous to the unaffected dorsal domain that we see in suc/et-1 mutants. Alternatively, a second zebrafish et-1 might activate dorsal gsc expression. A third possibility is that the suc/et-1- independent dorsal gsc domain is not homologous to any mammalian gsc expression domain. Regardless, the abolished ventral domains of gsc in suc/et-1 mutant embryos show that, in zebrafish and mice alike, some expression of gsc requires et-1.

Interestingly, these dorsal and ventral domains of zebrafish gsc expression, separated by a non-expressing intermediate region, combined with ventrally restricted gene expression patterns (e.g. dHAND), suggest that at this early precondensation stage, at least three arch mesenchymal fates have been specified: dorsal, intermediate or presumptive-joint, and ventral. This apparent specification of joints at a stage before they form is reminiscent of Gdf5 expression in the mouse, which prefigures joints in the axial and appendicular skeleton (Storm and Kingsley, 1996). Since the joints between dorsal and ventral arch cartilages in suc/et-1 mutants are also eliminated (Piotrowski et al., 1996; Kimmel et al., 1998; Fig. 1), we predict that markers discovered to be expressed in these presumptive-joint regions will be downregulated in suc/et-1 mutants.

dlx2 expression in cranial neural crest cells in suc/et-1 mutant embryos is unaffected until a postmigratory stage, when ventral neural crest cells fail to express dlx2. Likewise, the EdnrA mutant mice have no early dlx2 expression defects. However, at later stages, the EdnrA mutant mice fail to maintain dlx2 expression specifically in the second arch (Clouthier et al., 2000). Thus, while these results suggest that neither fish nor mice require et-1 signaling for early dlx2 expression, at later stages, fish and mice require et-1 signaling for dlx2 expression in different subpopulations of postmigratory arch neural crest.

In zebrafish, dlx3 arch expression is ventrally-restricted and requires suc/et-1. Similarly, in mice, dlx3 arch expression is ventrally-restricted and requires EdnrA (Robinson and Mahon, 1994; Qiu et al., 1997; Clouthier et al., 2000). Qiu et al. (1997) suggest that the phenotypes of the Dlx2 and Dlx1;Dlx2 mutant mice are specific to the dorsal arches because Dlx3 compensates for Dlx1 and Dlx2 ventrally. Although Dlx3 mutant mice die too early to assess a craniofacial requirement (Morasso et al., 1999), humans with a missense mutation in DLX3 have tricho-dento-osseous (TDO) syndrome (Price et al., 1998), which includes a craniofacial component (Kula et al., 1996).

Expression of the zebrafish ephrin receptor tyrosine kinase EphA3 is also ventrally restricted and requires suc/et-1. Similarly, a mouse EphA3 gene is expressed in Meckel’s cartilage (Kilpatrick et al., 1996), although no function for EphA3 has been demonstrated in ventral arch development. We described a ventral arch morphogenesis defect in suc mutants (Kimmel et al., 1998). Since all other genes thus identified as downstream of suc/et-1 are transcription factors, EphA3 promises to help connect this hierarchy of transcription factors with the more immediate ‘effectors of morphogenesis’ (Holder and Klein, 1999).

Since dHAND, msx1 and dlx3 are all also ventrally-restricted within mouse arches, it seems that much of the genetic network that specifies ventral arch postmigratory neural crest (Fig. 9) has been largely conserved between mice and fish. Once mutations are found in each of the six zebrafish genes whose ventral expression requires suc/et-1, these genes can be ordered into genetic pathways by examining the time course of expression of the other ventral arch markers in each mutant. Furthermore, by examining the mutant phenotypes, the specific functional requirements of each of these genes for craniofacial development can be determined.

The ability of suc/et-1 mutant neural crest cells to adopt ventral arch fates (e.g. express dHAND or EphA3 or contribute to normal ventral cartilages) in a wild-type host shows that the defects in suc/et-1 mutant neural crest cells are nonautonomous. Our results show that, beginning at approximately 12 hours, transplanted suc/et-1 mutant neural crest cells migrate into the arches of a wild-type host, mixing extensively with host cells. Thus, suc/et-1 is not required in neural crest cells for their ventral migration. Furthermore, markers of premigratory and migratory neural crest are unaffected in suc/et-1 mutant embryos, and injections of human ET-1 protein into arch primordia rescue defects in suc/et-1 mutant embryos. Therefore, suc/et-1 appears to function after neural crest migration, to specify ventral fates in the postmigratory arch primordia. Similarly, since the EdnrA mutant mice also lack defects in markers of migratory neural crest, Clouthier et al. (2000) conclude that EdnrA is required not for neural crest cell migration, but for postmigratory fates. This later requirement for Et-1/EdnrA function at a postmigratory stage contrasts to the earlier role of EdnrB in neural crest migration demonstrated in mouse embryos by Shin et al. (1999).

Transplants of wild-type mesendodermal cells fail to rescue the suc/et-1 mutant phenotype, indicating that either small numbers of transplanted wild-type mesendodermal cells are not sufficient to induce ventral markers or rescue cartilage or alternatively, the surface ectoderm domain of suc/et-1 expression is required for ventral arch fates. Tissue-specific promoters that drive suc/et-1 expression in one or a subset of its expression domains could be used to assay which suc/et-1- expressing tissues control craniofacial development.

Conversely, the ability of suc/et-1 mutant mesodermal cells to contribute to normal ventral muscles (which are mispatterned in suc/et-1 mutants) of a wild-type host suggests that paraxial mesoderm does not autonomously require suc/et-1 for the abililty to form normal ventral muscles. To explain this, we propose bidirectional signaling occurs during arch development. First, Et-1 in the postmigratory neural crest environment specifies ventral arch neural crest fates. Later, neural crest-derived cells signal back to the paraxial mesoderm to allow correct formation of ventral muscles. This signal would be independent of Et-1, since neural crest does not express et-1 (Maemura et al., 1996; Clouthier et al., 1998; Nataf et al., 1998; this paper). Consistent with this, EdnrA is not expressed in arch paraxial mesoderm in mice and chicks (Clouthier et al., 1998; Kempf et al., 1998; Nataf et al., 1998). Likewise, the endodermal defect in suc/et-1 mutants suggests a similar suc/et-1-dependent feedback to the hyomandibular pouch, a tissue that in mice and chicks does not express EdnrA (Clouthier et al., 1998; Kempf et al., 1998; Nataf et al., 1998). In this model, both arch mesodermal and endodermal defects in suc/et-1 mutants are secondary to earlier defects in neural crest specification. However, the specificity of both the mesodermal and endodermal defects in suc/et-1 mutants suggests that these defects are revealing specific intertissue signaling events, ultimately dependent on suc/et-1 and required for proper pharyngeal arch development.

Our findings, combined with the pharmacological inhibition of the avian EDNRA receptor (Kempf et al., 1998) and the Et-1/EdnrA/Ece-1 mutant mice (Kurihara et al., 1994; Clouthier et al., 1998; Yanagisawa et al., 1998) suggest that Et-1’s requirement for jaw development is evolutionarily ancient and dates back to the common ancestor of teleosts and amniotes. Perhaps recruitment or modification of et-1 signaling in the anterior arches of agnathans played a role in the appearance of jaws during evolution. Examining expression of et-1 signaling components, and other genes like dlx, msx and dHAND in lamprey anterior arches, could shed light on the homologies of agnathan skeletal elements and hence on jaw evolution itself.

We thank Angel Amores and Bruce Draper for generously providing PAC DNA pools; Yan-Ling Wang for DNA sequencing; John Postlethwait, Don Kane, and Rachel Warga for mapping advice; Hervé Kempf, Dave Raible, and Debbie Yelon for helpful discussions, and Bruce Draper, Bill Jackman, and Lisa Maves for critical reading of the manuscript. This research was supported by NIH grants HS17963 and HD22486. C. T. M. was supported by an NSF predoctoral fellowship and NIH 5-T32-HD07348, T. F. S. was supported by the HFSP and the Wellcome Trust (RCDF 055120), K. H.-L. was supported by NIH grant K08 HL03371, and J. P. was supported by a HHMI summer research fellowship.