Two endothelin 1 effectors, hand2 and bapx1,pattern ventral pharyngeal cartilage and the jaw joint

The skeleton of derived vertebrates consists of variously shaped bones and cartilages separated by joints. Experiments with the chick elbow joint showed that normal joints formed following excision of the embryonic forelimb distal to the joint; thus, joint specification appeared to have some local autonomy(Holder, 1977). Consistent with this idea, genetic analysis in the mouse has revealed that skeletal elements and joints are specified by partially separable mechanisms; whereas some genes are essential for joint formation, other genes are essential for particular bones (Brunet et al.,1998; Davis et al.,1995). However, these processes are intricately linked and share certain regulators. For instance, defects in joints and cartilage formation are both seen in Gdf5 mutant mice(Gruneberg and Lee, 1973;Storm et al., 1994;Storm and Kingsley, 1996;Storm and Kingsley, 1999). Noggin (Nog) mutant mice lack joints, but also exhibit hyperplastic cartilage and defective cartilage maturation(Brunet et al., 1998). However,humans heterozygous for certain NOG mutations can have more specific joint defects, arguing these processes are at some level genetically separable(Gong et al., 1999).

GDF5 and another signalling molecule of the TGFβ superfamily, BMP5,are required for partially overlapping subsets of joints in the mouse axial skeleton, suggesting the skeleton is assembled piecemeal by partially redundant TGFβ signalling molecules(Storm et al., 1994;Storm and Kingsley, 1996;Storm and Kingsley, 1999).Gdf5 negatively regulates its own expression, and thus refines where the joint is positioned (Storm and Kingsley, 1999). Despite being required for certain joints, GDF5 is not sufficient to induce ectopic joints(Storm and Kingsley, 1999;Merino et al., 1999;Francis-West et al., 1999a). In chick embryos, another secreted molecule, Wnt14, is sufficient to induce ectopic joints, and in addition inhibits nearby joints(Hartmann and Tabin, 2001). Little is known about upstream factors that control the expression of these joint-promoting signaling molecules.

The jaw joint forms in the first, or mandibular, pharyngeal arch,articulating the upper and lower jaw. Work in mice, chicks and zebrafish has begun to unravel molecular mechanisms responsible for jaw development. In both mice and zebrafish, the secreted peptide endothelin 1 (encoded by theedn1 or sucker gene in zebrafish) is required for development of the jaw, as well as skeletal elements of the second, or hyoid,pharyngeal arch (Kurihara et al.,1994; Miller et al.,2000; Miller and Kimmel,2001). In mice, targeted inactivation of the Edn1 receptor, EdnrA,produces a similar phenotype as Edn1 inactivation, namely loss of the mandible and severe malformations of other pharyngeal skeletal elements(Clouthier et al., 1998). Pharmacological inactivation of EdnrA in chick embryos results in a similar disruption of lower jaw formation (Kempf et al., 1998). Thus, Edn1-EdnrA signaling is required for lower jaw formation in chicks as well as mice and fish. Within the early pharyngeal arch primordia, secretion of Edn1 from paraxial mesodermal cores and surrounding epithelia, both surface ectoderm and pharyngeal endoderm, is received by the EdnrA receptor, which is broadly expressed in the postmigratory cranial neural crest (CNC) cylinder. Edn1 signaling sets up a dorsoventral prepattern and promotes the specification of ventral fates within this cylinder of postmigratory CNC (reviewed byKimmel et al., 2001a). In zebrafish, graded reduction of Edn1 function with edn1 antisense morpholino oligonucleotides (Edn1-MOs) results in graded reduction in ventral pharyngeal cartilage formation. The pharyngeal joints also require Edn1 and are more sensitive to Edn1 reduction, because ventral cartilage, but not joints, form in animals injected with lower doses of Edn1-MOs(Miller and Kimmel, 2001). Thus, in zebrafish Edn1-signaling is required for both joint and ventral pharyngeal fates, with joint fates being more sensitive to Edn1 reduction.

The requirement for Edn1 in activating expression of hand2 (also known as dHAND) in the ventral arch primordia is conserved between mice and zebrafish (Thomas et al.,1998; Miller et al.,2000). hand2 mutant mice die before skeletal differentiation occurs with severe heart and circulatory defects(Srivastava et al., 1997;Thomas et al., 1998;Yamagishi et al., 2000). In the hand2 mutant mouse pharyngeal arch primordia, CNC fails to adopt ventral fates and undergoes apoptosis(Thomas et al., 1998). Thus,hand2 is an excellent candidate effector of Edn1-mediated pharyngeal arch patterning.

Here we present functional analysis of two edn1-dependent genes,bapx1 and hand2, during zebrafish pharyngeal arch development. edn1 expression is complementary to and surrounded byhand2-expressing ventral arch CNC, whereas bapx1 expression defines an intermediate presumptive joint domain, ventral to somedlx2 expression and dorsal to hand2 expression. The loss of the jaw joint in edn1 mutants can in part be explained by a failure to upregulate expression of bapx1, whose function is required for multiple aspects of skeletal development of the jaw joint region including the joint itself, the retroarticular process of Meckel's cartilage, and the retroarticular bone. In the developing jaw joint, bapx1 is required for expression of chordin and gdf5. The loss of ventral pharyngeal cartilage in edn1 mutants can in part be explained by a second edn1 target gene, hand2. Similar to edn1,hand2 is required for formation of almost all ventral pharyngeal cartilage. Despite this phenotypic similarity to edn1 mutants, the early ventrally restricted edn1-dependent expression of dlx3,EphA3, gsc, msxe and msxb in cartilage precursors are all present in hand2 mutants. Further, msxe and msxbare upregulated in the ventral arches of hand2 mutants. However,similar to edn1 mutants, hand2 mutants lack late first arch expression of gsc. bapx1 expression in hand2 mutants is ectopically expanded ventrally, suggesting that hand2 helps position the jaw joint by repressing expression of bapx1. Finally, we show that both hand2 and edn1 restrict eng2 expression to dorsal mesoderm. Thus the specification of ventral pharyngeal arch fates involves the repression of other ventral, joint and dorsal arch fates. Collectively, our results identify bapx1 and hand2 as critical effectors of Edn1 in patterning the jaw joint and ventral pharyngeal cartilage.

Fish were raised and staged as described(Westerfield, 1995;Kimmel et al., 1995).edn1(sucker)^tf216b(Piotrowski et al., 1996;Miller et al., 2000) were maintained on *AB background, whereas hand2^s6(han^s6) was inbred on the original background described by Yelon et al. (Yelon et al.,2000). hand2^s6 heterozygotes were incrossed,and mutants sorted from clutches by lack of beating heart tissue at 28 hours postfertilization (hpf) (Yelon et al.,2000).

Degenerate PCR primers designed against the Nk3/Bap homeobox region 5′-TGGARMGNMGYTTYAAYCAYCA-3′ and 5′-TTRTAN-CKNCKRTTYTGRAACCA-3′ were used with zebrafish genomic DNA as a template and 48 cycles of: 94°C for 30 seconds, 48°C for 30 seconds and 72°C for 30 seconds, which generated a 117-bp band. This PCR product was cloned using a TOPO TA kit (Invitrogen) and sequencing revealed a Bapx1-like homeodomain. These primers and conditions were then used to screen DNA pools from an arrayed genomic DNA PAC library(Amemiya and Zon, 1999), which identified a single positive PAC, 91M18. This PAC was isolated and subcloned,and sequencing subclones yielded the sequence of the second exon and the 3′ end of the intron. Using gene-specific primers 5′-GATCTTGACCTGCGTCTCG-3′ and 5′-GCGTTATCTCTCCGG-ACCG-3′ from this PAC sequence to amplify a 72 bp band, phage dilution pools of a 15-19 hpf zebrafish cDNA library(Appel and Eisen, 1998) were screened by PCR. A single phage was isolated, which contained a 1357 bp insert, containing the first exon and predicted full-length ORF of abapx1 gene (Accession Number, AY225416).

Alcian staining and in situ hybridizations were performed as described(Miller et al., 2000). After Alcian staining, some wholemounts were treated with a solution of 3% hydrogen peroxide and 1% potassium hydroxide for 10 minutes to remove pigmentation. Bone labeling using calcein was performed as described(Yan et al., 2002;Kimmel et al., 2003). For sectioning, embryos were embedded in Epon and sectioned at 5 microns.

A 1.2 kB PCR fragment of bapx1 genomic DNA, containing the entire second exon and part of the 3′ UTR, was used for all in situs.chd and gdf5 probes are described in Schulte-Merker et al.(Schulte-Merker et al., 1997)and Bruneau et al. (Bruneau et al.,1997), respectively. All other riboprobes are described or referenced in Miller et al. (Miller et al., 2000).

edn1 morpholino oligo (edn1-MO) 5′-GTAGTATGCAAGTCCCGTATTCCAG-3′(31 to 7 nucleotides 5′ to predicted translation start site), (seeMiller and Kimmel, 2001),bapx1 morpholino oligo (bapx1-MO1)5′-GCGCACAGCCATGTCGAGCAGCACT-3′ (ATG start complementary sequence underlined) and bapx1-MO2(5′-GCGGAGCATTAGGGTTAAGATTACG-3′, complementary to 52 to 28 nucleotides 5′ of the predicted ATG start codon) were purchased from Gene Tools, Inc., and diluted to 25 mg/ml in 1 × Danieau buffer. Subsequent dilutions were made in 0.2 M KCl and 0.2% Phenol Red. These dilutions were injected into the yolk of 1-8 cell zebrafish embryos,approximately 5 nL per embryo. bapx1-MO1, which seemed less toxic and gave cleaner phenotypes (see below), was used for all phenotypic analyses. The inbred *AB line was used for all MO-injections into wild types.

Along the dorsoventral (DV) axis of the zebrafish early larval pharyngeal skeleton, separate cartilages form dorsally and ventrally (such as the upper and lower jaw in the first arch), and are separated by joints at an intermediate position (such as the jaw joint in the first arch). In zebrafish,edn1 patterns the DV pharyngeal arch axis: graded reduction ofedn1 in zebrafish results in joint loss with mild reduction and joint and ventral cartilage loss with more severe reduction. At high levels ofedn1 reduction, dorsal fates are also affected, although to a relatively lesser degree (Miller et al.,2000; Miller and Kimmel,2001). Here we investigate the genetic circuitry downstream ofedn1 controlling DV pharyngeal arch patterning. We previously described a DV prepattern set up in the postmigratory CNC. Expression ofhand2, which encodes a bHLH transcription factor, is confined to a ventral subset of dlx2-expressing CNC, defining a ventral domain of the prepattern. We also described edn1 expression in ventral arch mesenchyme and epithelia (Miller et al.,2000). Here we extend those analyses and ask how, on a detailed cellular level, edn1 expression in the ventral arches relates to this DV prepattern. For instance, does edn1 expression extend past the DV interface, or is edn1 expression contained within the ventral domain?We chose to focus on a postmigratory stage within the probable time of function of the Edn1 signal (Miller et al., 2000).

We analyzed the pharyngeal arch expression domains of hand2, edn1,and the more broadly expressed postmigratory CNC marker, dlx2, in serial sections of 32 hpf embryos stained for the expression of each of these genes (Fig. 1). These serial section studies strongly support our previous findings in wholemounts(Miller et al., 2000), and add cellular resolution to the expression patterns. hand2-expressing cells are a ventral subset of dlx2-expressing CNC cells, and are closely apposed to cells of three different tissues expressing edn1:ventral surface ectoderm, ventral mesodermal cores and pharyngeal endodermal epithelia. In the first two arches, hand2-expressing cells cover dorsally the edn1-expressing arch cores, which are in extreme proximity to the yolk. edn1 expression is not appreciably detected in the first pharyngeal pouch at this stage. The ventral surface ectodermal domain of edn1 expression seems to extend to, but not beyond, the dorsal extent of the adjacent hand2 expression domain(Fig. 1E,F,H,I). Thus, in the first two arches at this stage, edn1 expression is restricted to the ventralmost tissues of the arches, appearing slightly more ventrally localized than hand2 expression.

In this DV postmigratory CNC prepattern, approximately the ventral third of the dlx2-expressing CNC cylinder expresses hand2, and probably includes precursors of the ventral cartilages. We hypothesized that joint-forming cells, although slightly farther away from the Edn1 source,respond to Edn1 signaling, given the requirement of edn1 for pharyngeal joint primordia (Miller and Kimmel, 2001). We thus began a search for markers of pharyngeal joint primordia. In Xenopus embryos, thebapx1/bagpipe-related NK3 superfamily homeobox gene Xbap is expressed in a large region in the intermediate first arch, encompassing the jaw joint region (Newman et al.,1997). Using degenerate PCR with a zebrafish genomic DNA library,followed by gene-specific PCR with a 15-19 hpf zebrafish cDNA library, we cloned a zebrafish bapx1 gene. Phylogenetic analyses of this zebrafish and other bagpipe-related genes reveals that this zebrafish gene is orthologous to Xbap and other vertebrate bapx1(nkx3.2) genes (data not shown).

Similar to expression of the Xenopus Xbap gene(Newman et al., 1997),expression of zebrafish bapx1 is present in mesenchyme of the mandibular arch primordia. Mandibular expression begins at approximately 30 hpf (see Fig. 2H), and at 32 hpf a large patch of expression is present in the posterior intermediate first arch postmigratory CNC cylinder (Fig. 2A,B). This domain persists through 54 hpf. By this stage, these intermediate bapx1 domains clearly mark the jaw joints because the upper and lower jaw cartilages have begun to chondrify, and bapx1expression is present in cells within and surrounding the jaw joint(Fig. 2C,D). By 54 hpf,additional bapx1 expression domains are present in the midline of the first two arches (Fig. 2C). Expression is also detected in pharyngeal endodermal epithelia at 32 through 54 hpf (Fig. 2A,B,F). Otherbapx1 expression domains include putative sclerotomal derivatives at 48 hpf, the pectoral fin at 54 hpf, and cells closely apposed to the eye at 36 and 42 hpf (data not shown).

We next asked how bapx1 expression in the putative joint region primordium relates to the dlx2/hand2 DV prepattern in the zebrafish pharyngeal arch primordia. Double-labeling experiments show thatbapx1 expression is ventral to a large dorsal domain ofdlx2-expressing CNC cells and dorsal to hand2-expressing CNC cells (Fig. 2G,H). Thus at this early stage, from 30-32 hpf, in the first arch primordia, bapx1 marks an intermediate or presumptive joint region.

Because the jaw joint region is particularly affected in 4-day old animals with reduced edn1 function(Miller and Kimmel, 2001), we next asked whether bapx1 downregulation prefigures the edn1mutant phenotype. At 36 hpf, first arch mesenchymal bapx1 expression is undetectable in edn1 mutants(Fig. 3A,B). These defects are not simply because of developmental delay, because by 54 hpf, the midline domains of bapx1 expression in both arch one and two are present inedn1 mutants, whereas the jaw joint domains of bapx1expression remain undetectable (Fig. 3C,D).

Injecting morpholino antisense oligos (MOs) into embryos has been shown to be highly successful at reducing gene function in vivo (reviewed byHeasman, 2002). This technique efficiently works to downregulate genes involved in zebrafish head skeletal development, as shown by highly penetrant phenocopy of the edn1mutant phenotype upon injection of edn1-morpholinos(Miller and Kimmel, 2001). To assess bapx1 function in pharyngeal skeletal patterning, morpholino oligos complementary to the region around the predicted translation start site of bapx1 were injected. Animals injected with bapx1-MOs display dose-dependent loss of the jaw joint (Fig. 4, Table 1). Injection of 1 mg/ml (total of 5 ng) of bapx1-MO1 resulted in 51% of injected animals lacking jaw joints, whereas injection of 3 mg/ml (total of 15 ng)raised this frequency to 80% (Table 1). The loss of features of the jaw joint region in bapx1-MO1-injected animals was graded in severity. Frequently the cartilaginous retroarticular process (RAP), which projects ventrally from the posterior end of Meckel's cartilage, was present, but just dorsally, the jaw joint itself was lost with the dorsal and ventral cartilages locally fused(Table 1,Fig. 4A-H). In more severely affected animals RAP was missing, and the joint was completely missing and filled in with ectopic cartilage (Fig. 4C,D). Unlike animals with mild edn1 reduction, and consistent with the absence of second arch joint bapx1 expression,the second arch joint is unaffected. However, the second arch midline skeletal element, the basihyal, which is prefigured by bapx1-expressing cells(see Fig. 3C), was characteristically reduced in bapx1-MO1-injected animals(Table 1).

To confirm the specificity of this morpholino, we injected a second non-overlapping bapx1-MO (bapx1-MO2). Although injection of MO2 caused other possibly non-specific phenotypes including loss of the branchial cartilages(data not shown, see Discussion), this MO also caused highly penetrant,dose-dependent loss of the jaw joint (Table 1). To further confirm the specificity of these jaw joint region MO phenotypes, we injected these two bapx1-MOs together, at relatively lower concentrations. These combinatorial injections caused more frequent jaw joint loss at concentrations lower than that of the singles alone. Although injection of 5 ng of MO1 and MO2 alone resulted in 51% and 2% loss of the jaw joint, respectively, injecting half as much of each MO combinatorially enhanced the penetrance of this phenotype to 77%. Coinjection of MO1 and MO2 also enhanced the basihyal phenotype, resulting in frequent deletion of this element (Table 1). These phenotypic enhancements further suggest that these non-overlapping MOs are specifically reducing function of bapx1.

Slightly later in development, at around 6 dpf, the retroarticular bone(RAB) begins ossifying perichondrally on RAP of Meckel's cartilage(Fig. 4I)(Cubbage and Mabee, 1996) (C. B. K., unpublished). Because RAP is often missing in bapx1-MO-injected fish(see above), we asked if skeletal defects in bapx1-MO-injected animals also included defects in RAB. To examine potential defects in pharyngeal bones of bapx1-MO-injected animals, we stained bones in uninjected and injected larvae with the fluorescent dye Calcein (Kimmel et al., 2003; Yan et al.,2002). In bapx1-MO-injected animals, we observed a highly penetrant RAB loss (Fig. 4J;111/117, or 95% of injected animals lacking RAB vs. 14/51 or 28% of uninjected siblings lacking RAB). Thus, bapx1 is required for at least three aspects of skeletal development in and around the jaw joint: (1) local inhibition of chondrogenesis at the jaw joint, (2) specification of the RAP of Meckel's cartilage, and (3) ossification of the RAB.

To provide a potential genetic mechanism for bapx1's role in jaw joint development, we asked whether markers of tetrapod limb joints also mark the zebrafish jaw joint, and whether such domains require bapx1function. In tetrapods, Gdf5, which encodes a TGFβ-related signaling molecule, is expressed in early cartilage condensations and later in developing joints (Chang et al.,1994,Storm et al.,1994; Storm and Kingsey, 1996;Merino et al., 1999;Francis-West et al., 1999a). A subset of mouse appendicular and axial joints requires Gdf5 function(Storm et al., 1994;Storm and Kingsley, 1996;Merino et al., 1999). Another marker of developing tetrapod limb joints is the BMP-antagonistchordin (chd)(Francis-West et al., 1999b;Scott et al., 2000). In 56 hpf zebrafish, chd expression, which is localized to the jaw joint in wild types (Fig. 5A), is absent in bapx1-MO-injected animals (Fig. 5B). At 78 hpf, gdf5 expression is detected in cells within the jaw joint, and these expression domains are severely reduced in bapx1-MO-injected animals. gdf5 expression was also detected in a triangular cluster of cells at the second arch ventral midline, seemingly prefiguring the unpaired ventral midline basihyal cartilage, and similar to the second arch midline domain of bapx1 expression at 54 hours(Fig. 5C,Fig. 2C). This second arch midline domain of gdf5 expression was also downregulated in bapx1-MO-injected animals (Fig. 5D), correlating with the reduced basihyal cartilage phenotype(Table 1). These expression defects in bapx1-MO-injected animals were specific to the joint region,because other expression domains of both genes, including cells around the second arch joint for both genes, the heart and ears for chd, and ceratohyal and palatoquadrate perichondrial cells for gdf5, were not affected by bapx1-MO injection. Thus, bapx1 is required for development of the jaw joint region: RAP, RAB and the jaw joint itself all require bapx1 function, as does the early expression of two genes,chd and gdf5, within the developing jaw joint.

Although bapx1 is a required edn1 effector of joint patterning, ventral pharyngeal cartilage formation outside of the joint region is largely unaffected in bapx1-MO-injected animals. Thus, we next sought anedn1 effector of ventral cartilage patterning. We focused onhand2, an excellent candidate for three reasons. First, in the mouse embryo, hand2 expression requires Edn1 function andhand2 is required for ventral pharyngeal arch patterning(Thomas et al., 1998). Second,hand2 expression requires edn1 function in zebrafish,correlating with the edn1 mutant ventral cartilage loss(Miller et al., 2000). Third,as shown above, hand2 expression exquisitely complementsedn1 expression in the ventral pharyngeal arch primordia.

hand2 mutant zebrafish were found in a screen for mutations affecting heart development, including a null allele completely deleting thehand2 locus (han^s6)(Yelon et al., 2000). Despite lacking a heart and circulating blood, han^s6 mutants (for clarity hereafter referred to as hand2 mutants) live for at least four days and make differentiated pharyngeal cartilage in the mandibular and hyoid arches (Fig. 6A,B). Unlike mutants of the anterior arch class (e.g. edn1 mutants)cartilages of the more posterior pharyngeal arches never develop. Dorsal anterior arch cartilages of reduced size, but relatively normal shape, form inhand2 mutants, whereas only a tiny amount of ventral cartilage forms(Fig. 6C-I). Mutant dorsal cartilages have readily recognizable components, including the pterygoid process and parallel stacks of chondrocytes forming the lateral plate of the palatoquadrate cartilage in arch one, and the symplectic (SY) and hyomandibular regions of the hyosymplectic cartilage in arch two. Mutant ventral cartilages, in contrast, are almost absent in the anterior arches(Fig. 6C-I). Although similar to edn1 mutants, hand2 mutants lack jaw joints, less cartilage is present ventrally in hand2 mutants, and the two upper jaws typically are connected by a small disorganized cartilage bridge spanning the midline (Fig. 6C-E;Table 2). In contrast and also unlike edn1 mutants, the hand2 mutant second arch has well-formed morphological joints between the dorsal and severely reduced ventral second arch cartilages (Fig. 6C-I). Hence the differential requirement of hand2 for joint formation reveals a key patterning difference between arch one and two.

We previously showed that edn1 function is required for the ventral arch expression of hand2 and four other genes: dlx3,EphA3, gsc and msxe (Miller et al., 2000; Miller and Kimmel; 2001). Homologs of dlx3, gsc and msxeare all required for proper mammalian craniofacial development(Kula et al., 1996;Price et al., 1998;Rivera-Perez et al., 1995;Yamada et al., 1995;Jabs et al., 1993;Satokata and Maas, 1994;Satokata et al., 2000).EphA3, the Ephrin transmembrane receptor tyrosine kinase, is expressed in Meckel's cartilage of rats(Kilpatrick et al., 1996). Thus, all four of these genes have potentially conserved pharyngeal arch domains transducing the Edn1 signal.

Because absence of ventral cartilage is shared between hand2 andedn1 mutants (Piotrowski et al.,1996; Kimmel et al., 1996;Miller et al., 2000) and because in the mouse pharyngeal arch msx1 expression requireshand2, we expected that ventral arch expression of some or all of these edn1 target genes would also require hand2. Instead,early ventral arch expression of dlx3, EphA3, gsc and msxeare robustly present in hand2 mutants(Fig. 7). We see three classes of effects on these genes in hand2 mutants: no effect or mild upregulation, arch-specific requirement and clear upregulation. In the first class, the ventrally restricted pharyngeal arch expression domains ofdlx3 and EphA3 expression are present in hand2mutants, and appear slightly upregulated(Fig. 7A-D).

In the second class, gsc expression requires hand2function in ventral arch one, but not ventral arch two. At 32 hpf, dorsal and ventral domains of gsc expression are present in the second arch, and both domains are present in hand2 mutants at 32 hpf(Fig. 7E,F). Later at 38 hpf, a ventral arch one domain of gsc expression is present, and this domain of gsc expression is absent in hand2 mutants(Fig. 7G,H). Both the early second arch ventral domain and the late first arch ventral domain ofgsc expression are missing in edn1 mutants(Miller et al., 2000) (data not shown).

In the third class, two msx genes are upregulated inhand2 mutants. Expression of msxe in ventral first and second arch CNC at 30 hpf is present in hand2 mutants and appears upregulated, possibly in other cells, but seemingly in the same cells that express msxe at lower levels in wild types(Fig. 7I,J). A second zebrafishmsx gene, msxb, is more sparsely and weakly expressed in wild-type ventral arch CNC at 30 hpf (Fig. 7K). This expression, similar to that of msxe, requiresedn1 but not hand2 function(Fig. 7L,M). Similar tomsxe, but more dramatically so, msxb expression appears upregulated in hand2 mutants. This msxb upregulation inhand2 mutants appears, similar to the msxe upregulation, to involve upregulation of transcription in cells that normally expressmsxb at lower levels, and in addition seems to involve the ectopic expression of msxb in ventral CNC cells in which msxbexpression is normally not detectable by in situ hybridization (i.e. in cells in the anterior ventral first and second arch, compareFig. 7K,L). Thus, in stark contrast to edn1 mutants, and despite ventral cartilage being almost absent later, early expression of dlx3, EphA3, gsc, msxe andmsxb are all present in the ventral arches of hand2 mutants. However, similar to edn1 mutants, hand2 mutants lack later first arch ventral gsc expression. These findings reveal complexity in the genetic network controlling ventral pharyngeal cartilage development. Based on the genes we have examined, we suggest that mostedn1-dependent signaling in the early pharyngeal arch primordia occurs independently of hand2 function.

Next we asked whether hand2 mutants, similar to edn1mutants that also lack jaw joints, have early bapx1 expression defects. Despite lacking a jaw joint, hand2 mutants have expanded intermediate first arch bapx1 expression(Fig. 8B). However and remarkably, bapx1 expression ectopically expands into the ventral first arch of hand2 mutants, where hand2 is normally expressed (Fig. 8B). To determine whether this ectopic bapx1 expression in hand2mutants requires Edn1 function, we injected Edn1 morpholinos (Edn1-MOs)(Miller and Kimmel, 2001) into clutches of hand2 mutants to obtain animals lacking bothhand2 and edn1 function. The expansion of bapx1expression in hand2 mutants is edn1-dependent, becausehand2 mutants injected with Edn1-MOs, similar to edn1mutants, completely lack first arch mesenchymal expression of bapx1(Fig. 8C).

The expanded domain of bapx1 expression in hand2 mutants suggests hand2 functions to repress the specification of joint fates. Because the joint forms at an intermediate DV location (see above), we finally asked whether even more dorsal fates are also repressed by hand2. Although we currently know of no marker restricted to dorsal arch postmigratory CNC in zebrafish, engrailed2 (eng2) expression is restricted to dorsal first arch myoblasts (seeHatta et al., 1990;Ekker et al., 1992;Kimmel et al., 2001b). In bothhand2 and edn1 mutants, eng2 expression expands ventrally, apparently revealing an unsubdivided arch mesodermal core(Fig. 8D-F). Thus,edn1 and hand2 specify ventral fates at least in part by repressing dorsal arch fates.

We show that two edn1 target genes, hand2 andbapx1, pattern ventral pharyngeal cartilage and the jaw joint during zebrafish development. Although shown to be required for patterning throughout the DV extent of the pharyngeal arch, edn1 expression is restricted to ventral tissues complementary to hand2 expression. First archbapx1 expression defines an intermediate or presumptive joint domain that requires edn1 function. We uncover multiple roles ofbapx1 in patterning the jaw joint region, including a requirement for morphological jaw joint formation and early expression of chd andgdf5 in the developing jaw joint. We show that a second edn1target gene, hand2, is required for ventral cartilage formation. The early, ventrally restricted, edn1-dependent pharyngeal arch expression of dlx3, EphA3, gsc, msxe and msxb does not require hand2. Instead, both msx genes appear upregulated inhand2 mutants. Moreover, bapx1 and eng2 expression expand ventrally in hand2 mutants. Thus, the specification of ventral pharyngeal fates is achieved at least in part through the repression of joint and dorsal fates.

The serial section analyses we present here place edn1-expressing cells as ventrally confined at a postmigratory stage. The ventral surface ectoderm, paraxial mesodermal arch cores, and pharyngeal endodermal epithelia expression domains of edn1 are all complementarily contained within the hand2-expressing ventral region of the arch primordia. Because the Edn1 signal is thought to be secreted and Edn1 function is required for patterning throughout the DV extent of the pharyngeal arch, perhaps Edn1 acts as a morphogen in patterning DV fates in the pharynx.

Our finding that a day before chondrogenesis, bapx1 is expressed in an intermediate region of the first arch, ventral to some dlx2expression and dorsal to hand2 expression, raises the possibility that at least three domains of the resultant skeleton (dorsal, joint and ventral) fate map to stereotypical domains of the pharyngeal arch primordia. Preliminary experiments with uncaged fluorescein support this idea, because uncaged dorsal and ventral spots in the arch primordia gave rise to labeled cells in dorsal and ventral cartilages, respectively (C. T. M., C. B. K., S. Cheesman and S. Hutchinson, unpublished). Powerful new tools including green fluorescent protein (GFP) lines and 4D confocal microscopy promise to allow high resolution fate mapping of the postmigratory CNC cylinders (J. G. Crump and C. B. K., unpublished), and will test the proposals that the DV prepattern prefigures dorsal and ventral cartilages, and that the early intermediatebapx1 expression domain at 30 hpf prefigures the jaw joint.

Our bapx1-morpholino experiments reveal multiple requirements forbapx1 in formation of the jaw joint region. Morphologically, the jaw joint itself, the nearby RAP of Meckel's cartilage, and the RAB all requirebapx1 function. These phenotypes, all associated with the jaw joint region, suggest bapx1 functions to specify multiple fates within the intermediate first arch primordia. Because all three of these skeletal phenotypes are seen in edn1 mutants(Miller et al., 2000;Kimmel et al., 2003), andbapx1 first arch expression requires edn1 signaling,bapx1 is a critical effector of edn1 in patterning the intermediate first arch.

bapx1 expression was detected at the ventral midline of the second arch, i.e. in the position of the later basihyal cartilage, consistent with the report of Xenopus Xbap expression(Newman et al., 1997). Midline expression of gdf5 in the second arch was downregulated in bapx1-MO-injected animals, providing a potential earlier molecular correlate of the basihyal reduction phenotype. Thus, in zebrafish, bapx1 is required for jaw joints and a specific ventral midline cartilage, the basihyal.

Although mammalian Bapx1 is expressed in mandibular arch mesenchyme and joints in the appendicular skeleton, no defects have been reported in these tissues in Bapx1 mutant mice. Bapx1 mutant mice have defective vertebrae and basal skulls, and are asplenic (Tribioli et al., 1999; Lettice et al.,1999; Akazawa et al.,2000). Calcein labeling revealed no detectable differences in early stages of vertebral formation in bapx1-MO injected-animals (data not shown, but see below) and we did not examine spleen development in bapx1-MO-injected zebrafish.

Although our morphological and molecular analysis of bapx1-MO-injected animals demonstrate that bapx1 is required for the jaw joint, we cannot be certain that MO injections completely eliminate bapx1function, especially at later larval stages when the MO is probably significantly diluted. Thus, the lack of detectable differences in early vertebral development in bapx1-MO-injected animals could reflect the late stage this event occurs and the dilution of the injected MO. Once true genetic nulls of zebrafish bapx1 are discovered, the role of bapx1in patterning the axial skeleton and spleen in zebrafish can be examined.

Morphologically, synovial joints are associated with a complex array of cell types, including articular cartilage, tendons and ligaments (seeKingston, 2000). Once markers are discovered for these tissue types in zebrafish, their formation can be assayed in anterior arch mutants and bapx1-MO-injected animals. Although GDF5 is insufficient to induce ectopic joints, GDF5 is sufficient to induce tendons and ligaments (Wolfman et al.,1997). Perhaps gdf5 and/or bapx1 in the developing jaw joint region also function to pattern connective tissues.

In chicks, RAP is derived from CNC emanating from the r4 level(Kontges and Lumsden, 1996). Thus, it will be particularly interesting to determine the axial level of origin of RAP in zebrafish. Genetic null alleles of bapx1 will allow mosaic analyses to determine which phenotypes (e.g. RAP loss) are cell-autonomous.

In zebrafish, the hand2 mutant cartilage phenotype resembles theedn1 mutant phenotype (i.e. ventral cartilage in the first two arches is severely reduced, displaced ventrally and posteriorly, and the jaw joint is missing). However, one notable difference is that the dorsal cartilages inhand2 mutants are less affected than in edn1 mutants. Organized stacks of palatoquadrate and SY chondrocytes are seen inhand2 mutants, but not in edn1 mutants, which typically lack the SY cartilage altogether (Kimmel et al., 1998; Miller et al.,2000). Thus, edn1 affects a broader pharyngeal arch domain than hand2 (see below).

Early edn1-dependent expression of dlx3, EphA3, gsc, msxeand msxb occurs in ventral postmigratory CNC of hand2zebrafish mutants, showing that none of these genes are sufficient for ventral cartilage formation. Late ventral arch one gsc expression fails to be initiated in hand2 mutants (Fig. 2H,I). This defect, shared with edn1 mutants, may contribute to the shared loss of ventral cartilage and/or the jaw joint inhand2 and edn1 mutants.

Given the phenotypic similarity of ventral cartilage reduction inedn1 and hand2 mutants and that in the mouse pharyngeal arch expression of msx1 requires hand2 function(Thomas et al., 1998), we expected that a subset of the edn1-dependent ventral genes would also require hand2. Although late ventral arch one gsc expression failing to be initiated in hand2 mutants meets this prediction, it is perhaps surprising that not only are the early edn1-dependent genes expressed, but that msxe and msxb are upregulated. Precedents exist for Hand2 functioning as a repressor, because in the mouse limb bud, Hand2 represses expression of Gli3 and Alx4(te Welscher et al., 2002). We suggest that in the zebrafish pharyngeal arches, the repression ofmsxe by hand2 is largely in the same ventral arch cells that in wild types express both genes. msxb, however, appears to be upregulated in hand2 mutants both in cells that in wild types expressmsxb, and in cells that in wild types do not contain detectablemsxb transcript by in situ hybridization (e.g. cells in the anterior ventral first two arches, see Fig. 7). Once antibodies specific to zebrafish Hand2, MsxB and MsxE are available, double-labeling experiments in sections could reveal the exact cellular relationship of these expression domains in pharyngeal arches of wild types and hand2 mutants. The difficulty in establishing orthology of the zebrafish msx genes with mammalian msx genes, two of which are required for craniofacial development, and one of which is downstream to Edn1 signaling, complicates predicting the head skeletal consequences of altered msxb and msxe expression in zebrafish (Ekker et al., 1997;Satokata and Maas, 1994;Satokata et al., 2000;Thomas et al., 1998). Functional analyses of the zebrafish msx genes in head skeletal patterning could reveal specific roles, if any, for msxb andmsxe. Because msx1 and msx2 in mice have redundant roles in chondrogenesis and osteogenesis(Satokata et al., 2000),perhaps the zebrafish msx genes do also.

Hu et al. (Hu et al., 2001)show that msx genes maintain cells in a proliferative state by blocking exit from the cell cycle, thus inhibiting differentiation. These findings could provide an explanation for why the same phenotype (ventral cartilage loss) is seen in edn1 and hand2 mutants, despite opposite effects on msx expression. Perhaps in edn1 mutants,the early failure to specify the entire ventral arch domain, including expression of msxe and msxb, results in the almost complete absence of tissues derived from this domain because of lack of proliferation. Conversely, in hand2 mutants, perhaps excess and ectopic msxexpression prevents ventral arch CNC from differentiating into chondrocytes.

The ventrally expanded expression domain of bapx1 inhand2 mutants indicate that the ventral first arch in hand2mutants has partially adopted a joint fate. However, because the jaw joint later fails to form in hand2 mutants, bapx1 appears to be insufficient for formation of a differentiated jaw joint in this context. Because the expanded ventral domain of bapx1 in hand2mutants also requires edn1 function, the positioning ofbapx1 to the intermediate first arch is accomplished at least in part through the positive regulation of edn1 and the repression ventrally by hand2.

In wild-type zebrafish embryos, eng2 expression in the head periphery is restricted to a dorsal first arch paraxial mesodermal core,constrictor dorsalis (reviewed by Kimmel et al., 2001b). Expression of eng2 expands ventrally in both hand2 and edn1 mutants, showing that mandibular mesoderm is dorsalized in edn1 and hand2 mutants, and indicating that in the early pharyngeal arch primordia, these ventral specifiers repress dorsal arch fates. Testing the initial proposal of Piotrowski et al. (Piotrowski et al.,1996) that dorsal skeletal identity is expanded ventrally in anterior arch mutants, awaits the identification of zebrafish dorsal-specific postmigratory CNC markers. Excitingly, a dorsal second arch dermal bone, the opercle, expands in animals with reduced edn1 function, providing another line of evidence that Edn1 signaling represses dorsal pharyngeal arch fates (Kimmel et al.,2003).

The downregulation of bapx1 expression in the jaw joint primordium and of dlx3, EphA3, gsc, msxe and msxb expression ventrally(this work) (Miller et al.,2000) in edn1 mutants reveals that Edn1 patterns both ventral and joint fates. edn1 expression appears contained within the ventral arch and at least some bapx1 expression does not overlap withhand2 expression. This raises the hypothesis that Edn1 functions as a morphogen in patterning the arch primordia, with the ventrally localized secreted Edn1 signal specifying ventral and joint fates at high and intermediate thresholds, respectively. An Edn1 gradient model, combined with the repression of bapx1 by hand2, provides an attractive mechanism for the positioning of the jaw joint. Analysis of dermal bone phenotypes in edn1 mutants is consistent with a gradient model(Kimmel et al., 2003). Embryological studies involving focal misexpression of Edn1 would directly test the gradient model.

Based on our results, we propose the following genetic model for DV pharyngeal arch patterning (Fig. 9). Specification of ventral is in part performed by theedn1-dependent activation of hand2. Specification of the jaw joint is performed by the positive regulation of bapx1 byedn1, acting at a distance from the ventral edn1, source. In the first arch, hand2 restricts the jaw joint by repressingbapx1 in the ventral cartilage-forming domain, hence delimiting the position of the jaw joint. bapx1 positively regulates jaw joint expression of chd and gdf5, which might play roles in pharyngeal joint development.

The localized expression of bapx1, chd and gdf5 to the zebrafish jaw joint (this work) and tetrapod appendicular joints(Tribioli et al., 1997;Francis-West et al., 1999a;Francis-West et al., 1999b;Scott et al., 2000;Storm and Kingsley, 1996;Merino et al., 1999) combined with the requirements of bapx1 (this work) and gdf5 for formation of particular joints (Storm and Kingsley, 1996) suggests that a conserved genetic network controls joint formation in both the pharyngeal and appendicular skeleton. Whether pharyngeal or appendicular joints arose first during evolution is currently not clear, but some early vertebrates such as placoderms and acanthodians had a clearly functional jaw joint (Janvier,1996).

Because bapx1 in vertebrates and invertebrates is expressed in gut-associated mesenchyme (Tribioli et al., 1997; Tribioli and Lufkin, 1999; Azpiazu and Frasch, 1993), this is probably an ancient role forbapx1, one present before the jaw or skeletal joints evolved. The localized expression of bapx1 in the jaw joint is particularly fascinating given the transformation this region underwent during vertebrate evolution. The expression and function of zebrafish bapx1 suggests a specific genetic network exists for jaw joint formation and immediately raises the question of whether agnathan lampreys have a first arch mesenchymalbapx1 expression domain. If so, perhaps modification ofbapx1 downstream targets, or modification of Bapx1 function,facilitated evolution of the jaw. If not, perhaps co-opting bapx1expression in the first arch of agnathans played a role in the appearance of jaws and joints during gnathostome evolution.

We thank Angel Amores and Allan Force for generously providing PAC DNA and phage pools, respectively; Jewel Parker for expert sectioning; and Gage Crump and Lisa Maves for comments on the manuscript. C. T. M. was supported by a NSF graduate research fellowship. D. Y. was an Amgen fellow of the Life Sciences Research Foundation and the recipient of a Burroughs Wellcome Fund Career Award. Research was supported by NIH RO1DE13834 and PO1HD22486 (C. B. K.); and NIH HL54737, AHA and the David and Lucille Packard Foundation (D. Y. R. S).