Jagged-Notch signaling ensures dorsal skeletal identity in the vertebrate face

The facial skeleton arises from cranial neural crest cells (CNCCs) that populate a series of pharyngeal arches in all vertebrate embryos. CNCCs of the mandibular and hyoid arches are further divided into dorsal and ventral domains that generate distinctly shaped cartilages and bones. In the larval zebrafish, ventral mandibular CNCCs generate the lower jaw Meckel's (M) cartilage, and dorsal mandibular CNCCs give rise to part of the palatoquadrate (Pq) cartilage (Crump et al., 2006). However, the pterygoid process (Ptp) of Pq, which functions as the larval upper jaw, arises from maxillary CNCCs (Eberhart et al., 2006). In the hyoid arch, ventral CNCCs give rise to the ceratohyal (Ch) and symplectic (Sy) cartilages and the branchiostegal ray (Br) bone, and dorsal CNCCs generate the hyomandibular (Hm) cartilage and opercle (Op) bone that support the gill covering (Fig. 1A,D). In general, dorsal mandibular and hyoid cartilages have plate-like morphologies, whereas their ventral cognates have rod-shaped morphologies. Moreover, the dorsal Op bone has a fan-shaped morphology that is distinct from the finger-shaped ventral Br bone.

Within the mandibular and hyoid arches, the secreted ligand Edn1 plays a central role in specifying ventral identity. In zebrafish and mice lacking Edn1 or the Endothelin type-A receptors (Ednras), the ventral facial skeleton either fails to develop or is transformed to a dorsal morphology (Kurihara et al., 1994; Clouthier et al., 1998; Miller et al., 2000; Ozeki et al., 2004; Nair et al., 2007). By contrast, Edn1 misexpression transforms the dorsal facial skeleton to a ventral morphology (Kimmel et al., 2007; Sato et al., 2008). Edn1 is thought to promote ventral skeletal development in part by activating the earlier expression of a network of ventral-specific genes in the mandibular and hyoid arches. Edn1 targets include Dlx3/dlx3b, Dlx5/dlx5a, Dlx6/dlx6a, Msx1/msxe and epha4b (rtk2) in ventral CNCCs of each arch, bapx1 (nkx3.2) in dorsoventral (DV)-intermediate CNCCs of the mandibular arch, and Hand2 (dHand)/hand2 in the most ventral CNCCs of each arch (Thomas et al., 1998; Clouthier et al., 2000; Miller et al., 2000; Miller et al., 2003; Walker et al., 2006). Moreover, several of these Edn1 targets have been shown to be required for development of the ventral face. Compound Dlx5^–/–; Dlx6^–/– mutant mice display transformations of the ventral mandibular skeleton (Beverdam et al., 2002; Depew et al., 2002), zebrafish lacking bapx1 fail to form the jaw joint (Miller et al., 2003), and mutations in Hand2/hand2 result in ventral skeletal loss in mice and zebrafish (Miller et al., 2003; Yanagisawa et al., 2003). However, whether patterning of the dorsal facial skeleton occurs simply by default (i.e. in the absence of ventral signaling) has remained unclear. Here, we show that dorsal skeletal identity requires active repression of ventral fates by Jagged-Notch signaling.

The Notch pathway is widely used during animal development to determine cell fates. Notch signaling occurs when transmembrane ligands of the Delta and Jagged/Serrate families engage Notch receptors on adjacent cells. Ligand binding then triggers cleavage and release of a Notch intracellular domain that translocates to the nucleus and activates the transcription of genes such as those of the Hey/Her/Hes class. In a process termed lateral inhibition, differential Notch signaling causes neighboring cells to adopt distinct fates. In other contexts, such as the fly wing, Notch signaling patterns fields of cells in organ primordia (Diaz-Benjumea and Cohen, 1995). In vertebrates, Jagged-Notch signaling has been implicated in the development of diverse organs, including the ear (Brooker et al., 2006; Kiernan et al., 2006), liver (Geisler et al., 2008; Lozier et al., 2008), pancreas (Golson et al., 2009) and cardiovascular system (High et al., 2008).

The role of Jagged-Notch signaling in craniofacial development is less clear. Several components of Jagged-Notch signaling are expressed in facial skeletal precursors, including zebrafish jag1b (Zecchin et al., 2005), mouse and human Jag1/JAG1 (Mitsiadis et al., 1997; Kamath et al., 2002b), zebrafish and mouse jag2/Jag2 (Jiang et al., 1998; Zecchin et al., 2005), and mouse Notch2 (Higuchi et al., 1995; Mitsiadis et al., 1997). Heterozygous mutations in human JAG1 or NOTCH2 are linked to Alagille syndrome, which is characterized by defects in multiple visceral organs, an abnormal facial appearance and occasional craniosynostosis and deafness (Li et al., 1997; Oda et al., 1997; Kamath et al., 2002b; Kamath et al., 2002a; Le Caignec et al., 2002; McDaniell et al., 2006). Whereas Jag1^–/– mice are embryonic lethal (Xue et al., 1999), Jag2^–/– mice die at birth from cleft palate (Jiang et al., 1998). In zebrafish, combined reduction of jag1b and jag2 function with morpholino oligonucleotides (MOs) has been reported to result in general reductions of facial cartilage (Lorent et al., 2004). However, a potential function of Jagged-Notch signaling in regional patterning of the facial skeleton has not been previously investigated.

Here, we employ mutant and transgenic analyses in zebrafish to demonstrate a novel role for Jagged-Notch signaling in patterning the dorsal face. In particular, we find that Jagged-Notch signaling limits the dorsal extent of ventral gene expression and helps determine dorsal skeletal morphology in the mandibular and hyoid arches. We further show that Jagged-Notch positive feedback and Edn1 inhibition are integrated through jag1b expression to restrict Notch activity to dorsal skeletal precursors. Moreover, compound mutant analysis reveals that a major function of Edn1 in ventral skeletal development is the repression of Jagged-Notch signaling. Together, our work defines a crucial role for Jagged-Notch signaling in DV facial patterning that might help to explain some of the craniofacial anomalies seen in Alagille syndrome.

Zebrafish (Danio rerio) embryos were raised at 28.5°C and staged as described (Kimmel et al., 1995). The jag1b^b1105 allele was identified in an ENU mutagenesis screen in which parthenogenic diploid progeny were analyzed for skeletal defects, and the sucker/edn1^tf216b mutant is as described (Miller et al., 2000). Tg(hsp70I:Gal4)^kca4/+ (Scheer and Campos-Ortega, 1999) and fli1a:GFP (Lawson and Weinstein, 2002) zebrafish are as described. To create UAS:JAG1^el108 transgenic zebrafish, we used the Gateway (Invitrogen) Tol2kit (Kwan et al., 2007). Full-length human JAG1 cDNA (Open Biosystems, clone 30528888) was inserted into pDONR221 by PCR using primers hJAG1-L2 (5′-GGGGACAAGTTTGTACAAAAAAGCAGGCTGAATTCCGCGGCGCAGCGATGCGTT-3′) and hJAG1-R2 (5′-GGGGACCACTTTGTACAAGAAAGCTGGGTACTAGTCCCGCGGTCTGCTATACG-3′). The resultant pME-JAG1 vector was combined with p5E-UAS, p3E-polyA and pDestTol2CG2 to create UAS:JAG1, which also contains a cmlc2:EGFP transgene that expresses GFP in the heart (cmlc2 is also known as myl7 – Zebrafish Information Network). UAS:JAG1 was injected with Tol2 transposase RNA into one-cell stage embryos and stable line el108 was isolated. For heat-shock induction, embryos were placed in a 40°C incubator at 20 hours post-fertilization (hpf) and then transferred to 28.5°C at 28 hpf.

One-cell stage embryos were injected with 3 nl of jag1b-MO (400 μM), notch2-MO (800 μM) or edn1-MO (27 μM) (Gene Tools, Philomath, OR, USA). jag1b-MO (previously known as jag3-MO) and notch2-MO have been demonstrated to block translation and mRNA splicing, respectively, and we confirmed inhibition of notch2 splicing as described (Lorent et al., 2004). The concentration of edn1-MO used causes partial loss of Edn1 function (Miller and Kimmel, 2001).

Skeletal staining with Alcian Blue and Alizarin Red (Walker and Kimmel, 2007), live bone staining with Calcein Green (Kimmel et al., 2003), and colorimetric in situ hybridization experiments (Crump et al., 2004) were performed as described. For fluorescence in situ experiments, two modifications were made to the published protocol (Welten et al., 2006): hybridizations were conducted at 68°C and antibody concentrations were 1:500 anti-DIG-POD and 1:200 anti-DNP-peroxidase.

jag1a, jag1b, jag2, notch2, hey1 and ednra2 probes were synthesized with T7 RNA polymerase from PCR products using the following primers (shown 5′ to 3′): Jag1a-L, CCGCGTATGTTTGAAGGAGT; Jag1a-RT7, GCTAATACGACTCACTATAGGGCAGTTCTGTCCGGAGTAGC; Jag1b-3L, CACGTGACGAGTTCTTTGGA; Jag1b-4RT7, GCTAATACGACTCACTATAGGGACACCGGTATCCATTCACC; Jag2-L, TGGGATGGGATAACTCCAC; Jag2-RT7, GCTAATACGACTCACTATAGGTCAAAGCCATTTTCCAGGTC; Notch2-L, ACCCTGTCATCATGGCAAAT; Notch2-RT7, GCTAATACGACTCACTATAGGACAGGTTCCCTGATTCATGC; Hey1-L, TCATTTAAAGATGCTTCATGCTG; Hey1-RT7, GCTAATACGACTCACTATAGGGTCTGTTTCTGTGCATCTGTTCA; Ednra2-L, CAATCATTTCCTGCATCGTG; and Ednra2-RT7, GCTAATACGACTCACTATAGGCAAGAGTTCACAGTCGCCAA. Published probes include dlx2a and dlx3b (Akimenko et al., 1994), bapx1 (Miller et al., 2003), epha4b [referred to as EphA3 by Xu et al. (Xu et al., 1995)], msxe (Akimenko et al., 1995), dlx5a and dlx6a (Walker et al., 2006), hand2 (Angelo et al., 2000) and edn1 (Miller et al., 2000).

In all experiments, genotyping of embryos confirmed the observed phenotypes. For jag1b^b1105 genotyping, we amplified product using primers Jag1b-IDL (5′-GTACCAAATCCGGGTGACCT-3′) and Jag1b-IDR (5′-GTGGCTTTTTGGGTCATTATCA-3′) and digested with BtsCI to generate a 206 bp fragment in mutants and 134/72 bp fragments in wild types. edn1^tf216b genotyping was performed using primers Edn1-IDL (5′-AGCGCGACAAATTCAATCAT-3′) and Edn1-IDR (5′-CAAAAGTAGACGCACTCGTTA-3′), followed by digestion with HpaI to produce 178/20 bp fragments in mutants and a 198 bp fragment in wild types. The presence of hsp70I:Gal4 was detected by PCR using primers Gal4-IDL (5′-CTCCCAAAACCAAAAGGTCTCC-3′) and Gal4-IDR (5′-TGAAGCCAATCTATCTGTGACGG-3′). UAS:JAG1 embryos were selected by heart GFP, and hsp70I:Gal4-negative UAS:JAG1 siblings were used as controls.

Unilateral tissue transplantations were performed as previously described, with the non-recipient side acting as an internal control (Crump et al., 2004). Briefly, donor tissue from fli1a:GFP embryos injected with Alexa 568-dextran (Molecular Probes) was transplanted into different fate-map regions of jag1b^b1105; fli1a:GFP or notch2-MO; fli1a:GFP hosts at 6 hpf. Targeting of Alexa 568-positive donor tissue was assessed at 36 hpf by localization relative to fli1a:GFP. For endoderm transplants, donor embryos were also injected with Tar^* RNA to promote endoderm targeting.

Skeletons and in situ hybridization embryos were photographed on a Zeiss Axioimager.Z1 microscope using Axiovision software. Fluorescence images were captured on a Zeiss LSM5 confocal microscope and, except where indicated otherwise, z-stacks of ∼40 μm were flattened into single projections. Levels were adjusted in Adobe Photoshop CS2, with care taken to apply identical adjustments to images from the same data set and to avoid removing information from the image. Dissected skeletons were cropped to remove surrounding soft tissue.

JMP 7.0 software (SAS) was used for one-way analysis of variance. A Tukey-Kramer honestly significant difference (HSD) test (α=0.05) showed significance for all dlx3b and dlx5a expression differences and for the following comparisons in the skeletal analysis of Edn1 and Jagged-Notch interactions: M and Ch (edn1 versus all others), Hm (edn1; jag1b versus all others) and Op (edn1 versus jag1b and edn1; notch2-MO).

As part of an ENU mutagenesis screen conducted at the University of Oregon, we isolated a mutation, b1105, that displays defects in the facial skeleton (Fig. 1B). Linkage analysis, phenocopy with a jag1b-MO and gene sequencing revealed that the b1105 lesion is a G-to-A transition in the zebrafish jag1b gene (see Fig. S1 in the supplementary material). The jag1b^b1105 mutation converts tryptophan 223 to a premature stop codon, truncating the Jag1b protein within the extracellular DSL domain required for Notch binding (Fig. 1C) (Cordle et al., 2008). At 5 days post-fertilization (dpf), jag1b^b1105 mutants have variable facial defects that ranged from mild reductions to more striking shape changes of dorsal hyoid and mandibular skeletal elements (Fig. 1I). In the hyoid arch of the most severely affected jag1b^b1105 larvae, the dorsal Hm cartilage became more rod-shaped, partially resembling the ventral Ch, and the normally fan-shaped Op bone adopted the finger-shaped morphology of the ventral Br bone to which it occasionally fused (Fig. 1E,H). In the jag1b^b1105 mandibular arch, the dorsal portion of Pq was truncated rather than transformed. Contrary to previous jag1b-MO studies (Lorent et al., 2004), we observed no defects in the ventral Ch, Sy and M cartilages or in the Br bone in jag1b^b1105 larvae. The maxillary-derived Ptp and the neurocranium, which is derived from maxillary and frontonasal CNCCs (Wada et al., 2005), were likewise unaffected. jag1b^b1105 larvae died by 7 dpf, precluding an examination of the DV morphology of later-forming facial bones. Nonetheless, our loss-of-function data suggest that, particularly in the hyoid arch, Jag1b is required for dorsal skeletal morphology.

We next investigated which Notch receptor mediates Jag1b signaling in the face. In humans, heterozygosity of either NOTCH2 or JAG1 can result in Alagille syndrome, and Notch2 genetically interacts with Jag1 in mouse (McCright et al., 2002). Moreover, of the four zebrafish Notch genes (notch1a, notch1b, notch2 and notch3), we found that only notch2 is expressed in the pharyngeal arches at DV patterning stages, 28-36 hpf (data not shown). We thus tested the requirement for Notch2 in facial skeleton patterning. Using a notch2-MO to block notch2 mRNA splicing (Lorent et al., 2004), we found that Notch2 reduction results in partial transformation of dorsal Hm to a ventral rod-like morphology, the transformation of dorsal Op to a ventral finger-like morphology, and the truncation of dorsal Pq (Fig. 1F,H). notch2-MO larvae also infrequently developed ectopic cartilage near the DV boundaries within the mandibular and hyoid arches, suggesting an additional role for Notch2 in suppressing skeletal development at the DV interface (see Fig. S1 in the supplementary material). Overall, the skeletal phenotypes of notch2-MO larvae are comparable to those of jag1b-MO larvae, but weaker than those of jag1b^b1105 mutants, which might be due to incomplete reduction of Notch2 function by the MO or partial compensation by other, weakly expressed Notch receptors (Fig. 1I). Nonetheless, as Notch2 reduction results in dorsal-specific skeletal defects that are similar to, although less severe than, those seen in jag1b^b1105 mutants, we conclude that Notch2 at least partially mediates Jag1b signaling during DV facial patterning.

We next tested whether Jagged-Notch signaling is also sufficient to promote a dorsal skeletal morphology. Using a heat shock-inducible Gal4/UAS system (Scheer and Campos-Ortega, 1999) to induce human JAG1 expression throughout zebrafish embryos at early facial patterning stages (20-28 hpf), we found that JAG1 misexpression results in ventral-specific defects of the mandibular and hyoid skeletons (Fig. 1G). In the most severely affected larvae, ventral M and Ch cartilages adopted plate-like morphologies similar to those of dorsal Pq and Hm cartilages. In addition, the ventral Sy cartilage and mandibular jaw joint were lost, and the ventral Br bone fused to the dorsal Op bone. By contrast, dorsal Hm and Pq cartilages and the maxillary-derived Ptp were unaffected. In less severe examples, the ventral Br bone was strikingly transformed to a mirror-image duplicate of the fan-shaped dorsal Op, the jaw joint was lost, and ectopic cartilage formed near Pq and the midline (Fig. 1H,I). As loss and gain of Jag1b function result in reciprocal DV skeletal transformations, we conclude that Jag1b is both necessary and sufficient for dorsal skeletal morphology in the hyoid and, to a lesser extent, the mandibular arches.

We next examined whether Jag1b-Notch2 signaling might control dorsal skeletal character by regulating earlier patterns of DV gene expression in CNCC-derived skeletal precursors. Whereas dlx2a was expressed throughout mandibular and hyoid CNCCs, double-fluorescence in situ hybridizations showed that dlx3b and dlx5a expression is restricted to more ventral CNCCs of 36 hpf wild-type embryos (Fig. 2A,E). By contrast, we observed a moderate dorsal expansion of dlx3b and dlx5a expression in jag1b^b1105 and notch2-MO embryos (Fig. 2B,C,F,G). In particular, the dorsal expansion of dlx3b and dlx5a was more prominent in the hyoid arch, correlating with the stronger transformations seen in the dorsal hyoid skeleton. In order to rule out the possibility that dlx3b and dlx5a expansion is simply due to changes in arch size, we measured the areas of dlx2a, dlx3b and dlx5a expression in the hyoid arches of wild-type, jag1b^b1105 and notch2-MO embryos. Normalization of hyoid arch size by measuring the ratio of dlx3b to dlx2a (wild type, 43%; jag1b^b1105, 61%; notch2-MO, 55%) and of dlx5a to dlx2a (wild type, 41%; jag1b^b1105, 55%; notch2-MO, 54%) confirmed that the percentage of the hyoid arch expressing dlx3b and dlx5a increases in both jag1b^b1105 and notch2-MO embryos. By contrast, dlx3b and dlx5a expression was severely reduced in JAG1-misexpression embryos (Fig. 2D,H). Thus, Jagged-Notch signaling is also sufficient to inhibit dlx3b and dlx5a expression in ventral CNCCs.

We next examined whether Jag1b might regulate the expression of a broader cohort of ventral-specific genes. dlx6a, epha4b and msxe were expressed in a ventral-specific pattern, similar to that of dlx3b and dlx5a at 36 hpf. Compared with wild types, we found that the expression of dlx6a, epha4b and msxe also extended more dorsally in the hyoid and mandibular arches of jag1b^b1105 mutants and was severely reduced in JAG1-misexpression embryos (Fig. 3A-I). The expression of the mandibular joint marker bapx1 was also expanded dorsally in jag1b^b1105 mutants and lost in JAG1-misexpression embryos (Fig. 3J-L). By contrast, the expression of hand2, one of the most ventrally restricted genes in the arches, was unaffected in jag1b^b1105 embryos, although it was reduced in JAG1-misexpression embryos (Fig. 3M-O). Furthermore, the arch expression of edn1 and its receptor, ednra2, were unaffected in jag1b^b1105, notch2-MO and JAG1-misexpression embryos (Fig. 3P-U and data not shown). We therefore conclude that the role of Jag1b and Notch2 in dorsal skeletal patterning correlates with an earlier requirement in limiting the dorsal extent of most, but not all, ventral gene expression in facial skeletal precursors.

In order to understand where Jagged-Notch signaling functions to repress ventral gene expression, we analyzed the expression of jag1b, notch2 and the Notch target gene hey1 during arch patterning stages. At 28 hpf, double-fluorescence in situ hybridizations of jag1b with notch2, and of hey1 with the CNCC-specific dlx2a probe, showed that jag1b and hey1 are expressed in the dorsal-most CNCCs of the mandibular and hyoid arches and in pouch endoderm (Fig. 4A,H). From 32 to 36 hpf, jag1b and hey1 expression continued to be dorsally restricted, yet extended more ventrally to abut dlx3b and dlx5a expression; concomitantly, expression became more prominent in posterior CNCCs in each arch and endoderm expression disappeared (Fig. 4B,C,I,J and see Fig. S2 in the supplementary material). By 36 hpf, hey1 also began to be expressed in ventral arch mesoderm. Double-fluorescence in situ hybridization revealed that hey1 expression extended just ventral to that of jag1b, as predicted if Jag1b is activating Notch2 in adjacent cells, and extensive colocalization of jag1b with dlx2a confirmed that within the dorsal arches, jag1b is expressed primarily in CNCCs (see Fig. S2 in the supplementary material). In addition, jag2, but not jag1a, was co-expressed at low levels with jag1b in dorsal CNCCs at 36 hpf (see Fig. S3 in the supplementary material).

We also confirmed that hey1 is a bona fide target of Jag1b-Notch2 signaling in CNCCs, as hey1 CNCC expression was greatly reduced in jag1b^b1105 and notch2-MO embryos and upregulated in JAG1-misexpression embryos (Fig. 4K-M). Of note, the expression of hey1 in ventral arch mesoderm was lost in notch2-MO, but not in jag1b^b1105, embryos, suggesting a specific function of Jag1b in regulating hey1 expression within CNCCs. Furthermore, in contrast to the dorsal-specific expression of jag1b, we found that notch2 is more widely expressed throughout the pharyngeal arches from 28 to 36 hpf. Whereas higher expression of notch2 was observed in ventral CNCCs that also express dlx3b and dlx5a, notch2 was co-expressed at weaker levels with jag1b, hey1 and dlx2a in dorsal CNCCs (Fig. 4A-C and see Fig. S2 in the supplementary material). This widespread expression of notch2 was confirmed with two independent probes (data not shown). Thus, despite the stronger ventral expression of notch2, hey1 expression indicates that Jag1b activates Notch2 specifically in dorsal CNCCs, consistent with the observed function of Notch2 in repressing ventral gene expression in dorsal skeletal precursors.

As jag1b and notch2 are expressed in multiple arch tissues, we used mosaic rescue experiments to test in which tissues Jag1b and Notch2 are sufficient for facial skeletal patterning. In order to create tissue mosaics, we transplanted wild-type fli1a:GFP precursors at early gastrulation stages (6 hpf) into different fate-map domains of jag1b^b1105; fli1a:GFP or notch2-MO; fli1a:GFP hosts. fli1a:GFP specifically labels the CNCC component of the pharyngeal arches (Lawson and Weinstein, 2002), allowing us to assess the correct targeting of donor tissue to CNCCs, endoderm or ectoderm of the arches. Whereas transplantation of wild-type CNCC precursors rescued facial skeletal patterning in 30/39 jag1b^b1105 embryos, transplantations of wild-type endodermal (0/5) or ectodermal (1/8) precursors did not reliably rescue (Fig. 5B-D). Similarly, wild-type CNCC precursor transplants rescued skeletal defects in 12/15 notch2-MO embryos (Fig. 5E). We therefore conclude that Jag1b and Notch2 function predominantly in CNCCs, and not in the surrounding endoderm or ectoderm, to pattern the dorsal facial skeleton.

We next investigated how Notch activity is established throughout dorsal skeletal precursors. As Notch positively regulates the expression of Jagged/Serrate in other contexts (de Celis and Bray, 1997; Daudet et al., 2007), we examined whether Notch signaling also regulates jag1b expression in CNCCs. Indeed, we found that jag1b expression is severely reduced in dorsal CNCCs of jag1b^b1105 and notch2-MO embryos and is expanded into ventral CNCCs of JAG1-misexpression embryos at 36 hpf (Fig. 4D-F). Conversely, the stronger ventral expression domain of notch2 was expanded in jag1b^b1105 and notch2-MO embryos and reduced in JAG1-misexpression embryos. We therefore conclude that Jagged-Notch signaling activates jag1b expression and represses strong notch2 expression in dorsal CNCCs of the mandibular and hyoid arches.

As Edn1 is known to promote ventral gene expression, we examined whether Edn1 also inhibits the ventral expression of Jagged ligands. Similar to what we observe in JAG1-misexpression embryos, we found that jag1b expression expands into ventral arch CNCCs in edn1 mutants (Fig. 4G). jag2 expression also expanded ventrally in edn1 mutants (see Fig. S3 in the supplementary material), whereas ventral notch2 expression was reduced. Concomitantly, hey1 expression was upregulated in ventral CNCCs (Fig. 4N), suggesting that ectopic jag1b and/or jag2 expression results in increased Notch activity in ventral CNCCs of edn1 mutants. Thus, Edn1 acts oppositely to Jagged-Notch signaling to inhibit Jagged ligand expression and hence Notch activity in ventral skeletal precursors.

As we found that Edn1 represses Jag1b-Notch2 signaling, we reasoned that inappropriate Notch activity in the ventral domain might contribute to the ventral skeletal defects of edn1 mutants. edn1 mutants exhibit nearly complete loss of ventral M, Ch and Sy cartilages, ventral Br bone and joints, and the dorsal Op bone is variably reduced or expanded (Fig. 6C) (Miller et al., 2000; Kimmel et al., 2003). Remarkably, reduction of Jag1b or Notch2 function in edn1; jag1b^b1105 and edn1; notch2-MO larvae substantially rescued ventral cartilage (Fig. 6D-G). Whereas the `rescued' ventral mandibular M cartilage was morphologically abnormal, the ventral hyoid Ch and Sy cartilages were restored to a nearly normal morphology in some edn1; jag1b^b1105 and edn1; notch2-MO larvae. Interestingly, heterozygosity of jag1b also partially rescued the ventral cartilage defects of edn1 mutants at a low frequency, underscoring the critical balance of Jagged-Notch and Edn1 signaling required for DV skeletal patterning (Fig. 6E,G). Consistent with the rescue of ventral skeletal defects, we found that the earlier ventral expression of dlx3b, dlx5a and dlx6a is partially restored in edn1; jag1b^b1105 mutants (Fig. 6H). Of note, hey1 expression was reduced, but not completely absent, in ventral CNCCs of edn1; jag1b^b1105 embryos, potentially reflecting the presence of residual low-level Notch signaling mediated by Jag2 (see Fig. S4 in the supplementary material).

In contrast to the rescue of edn1^–/– phenotypes by loss of jag1b, partial reduction of Edn1 function with a low dose of edn1-MO did not rescue the dorsal Hm, Op and Pq cartilage defects of jag1b^b1105 mutants (see Fig. S5 in the supplementary material). However, upon nearly complete loss of Edn1 function in edn1; jag1b^b1105 mutants, there was a slight rescue of Hm shape and an increase in the frequency of expanded Op compared with that seen in jag1b^b1105 single mutants. Thus, although the dorsal skeletal defects of jag1b^b1105 mutants are not simply the consequence of increased Edn1, the presence of Edn1 appears to influence the penetrance of dorsal transformations in jag1b^b1105 mutants. Nonetheless, our genetic analysis shows that for ventral skeletal patterning, Edn1 functions primarily as an upstream inhibitor of Jagged-Notch signaling.

Here, we demonstrate a novel function of Jagged-Notch signaling in ensuring dorsal identity in the mandibular and hyoid arches (Fig. 7). Several lines of evidence indicate that Jag1b-Notch2 signaling acts oppositely to Edn1 and ventral Dlx genes in DV facial patterning. In particular, reduction of Jag1b-Notch2 signaling in jag1b^b1105 and notch2-MO larvae results in dorsal-specific defects that are similar to those seen upon Edn1 overexpression, including dorsal-to-ventral transformations of hyoid cartilage (Hm to Ch-like) and dorsal reductions of mandibular cartilage (Pq truncation) (Kimmel et al., 2007). Conversely, JAG1 misexpression results in ventral-specific skeletal defects similar to those seen upon reduction of Edn1 or combined Dlx3b/Dlx5a function, including loss of ventral Sy cartilage and joints and striking homeotic transformation of the ventral Br bone into a dorsal Op-like morphology (Miller and Kimmel, 2001; Walker et al., 2006). Moreover, JAG1 misexpression results in partial transformations of ventral M and Ch cartilages to dorsal plate-like morphologies, phenotypes that are not observed in zebrafish edn1 mutants but which are reminiscent of the ventral mandibular transformations of Edn1^–/–, Ednra^–/– and Dlx5^–/–; Dlx6^–/– mice (Beverdam et al., 2002; Depew et al., 2002; Ozeki et al., 2004; Ruest et al., 2004). Thus, Jag1b-Notch2 signaling is specifically required for dorsal skeletal morphology and is also sufficient to alter the morphology of the ventral hyoid and mandibular skeleton.

The opposite effect of Jag1b-Notch2 and Edn1 signaling on skeletal morphology is also reflected at the level of DV gene expression. Edn1 has been shown to positively regulate a broad cohort of ventral genes (dlx3b, dlx5a, dlx6a, epha4b, msxe and bapx1) (Miller et al., 2000; Miller et al., 2003; Walker et al., 2006), and here we show that Edn1 also negatively regulates two newly characterized dorsal-specific genes, jag1b and hey1. By contrast, our gain- and loss-of-function studies demonstrate that Jag1b-Notch2 signaling negatively regulates these same ventral genes and positively regulates dorsal jag1b and hey1. Although we find that Jag1b and Notch2 function tissue autonomously within CNCCs for DV patterning, we do not know whether Jag1b-Notch2 signaling inhibits ventral gene expression directly through transcriptional repressors such as hey1, or indirectly via other downstream targets. Moreover, whereas changes in DV skeletal morphology correlate with earlier changes in DV gene expression in jag1b^b1105 mutants and JAG1-misexpression animals, we cannot rule out the possibility that Jag1b and/or Notch2 have additional roles in the proliferation, survival and/or differentiation of skeletal precursors (Crowe et al., 1999; Nakanishi et al., 2007). Indeed, the stronger expression of notch2 in ventral CNCCs suggests additional roles of Notch2 in arch development that are independent from its function in mediating Jag1b-dependent repression of ventral gene expression in dorsal skeletal precursors. Nonetheless, the mirror-image homeotic changes of hyoid bone seen in jag1b^b1105 and JAG1-misexpression larvae, combined with the opposite effects on dorsal versus ventral gene expression and cartilage morphology, indicate a clear role for Jag1b-Notch2 signaling in promoting dorsal identity in the mandibular and hyoid arches.

Whereas Jag1b-Notch2 signaling plays a crucial role in ensuring dorsal identity in the face, except for the previously discussed hyoid bones, loss and gain of Jag1b-Notch2 signaling results in only partial transformations of DV skeletal character. There are several reasons why skeletal elements might adopt morphologies similar, but not identical, to their DV cognates upon Jag1b-Notch2 manipulation. First, the lack of full transformations in jag1b^b1105 larvae could be due to residual Jagged-Notch signaling mediated by Jag2. Similarly, the partial transformations of notch2-MO larvae could be attributed to incomplete efficacy of the MO. However, our analysis of jag1b^b1105; jag2 double mutants has not revealed any striking enhancement of facial defects over jag1b^b1105 single mutants (data not shown), suggesting that the partial nature of the transformations is not due to residual Jagged-Notch activity. Second, cell-intrinsic changes in identity often do not lead to homeotic transformations of an identical nature, as a field of cells that adopts the identity of another may encounter different types of extrinsic signals or spatial constraints. For example, the loss of Hox paralog 2 genes, which function as homeotic selectors in anterior-posterior patterning, results in only partial duplications of the mandibular facial skeleton in the more posterior hyoid arch (Gendron-Maguire et al., 1993; Rijli et al., 1993; Miller et al., 2004). Analogously, Edn1 overexpression results in only partial transformations of dorsal skeletal elements, similar to what we observe in jag1b^b1105 mutants (Kimmel et al., 2007; Sato et al., 2008). Third, the partial transformations of the dorsal skeleton in jag1b^b1105 and notch2-MO embryos correlate with the only moderate expansion of dlx3b, dlx5a, dlx6a, epha4b, msxe and bapx1 expression into the dorsal arches, with the expression of the most ventrally restricted gene, hand2, being unaffected. Hence, rather than inhibiting ventral identity throughout the dorsal domain, the function of Jagged-Notch signaling might be to refine the dorsal limit of a subset of ventral genes that are expressed up to the DV border. Therefore, as discussed below, it might be that other signaling pathways act in parallel with Jagged-Notch to repress ventral gene expression in dorsal skeletal precursors.

Our analysis also reveals that Jagged-Notch signaling has a more restricted role in patterning a DV axis of the mandibular and hyoid arches that is distinct from the maxillary-mandibular axis (Fig. 7B). Whereas jag1b and hey1 are expressed in `dorsal mandibular' CNCCs anterior to the first pouch and in dorsal hyoid CNCCs between the first and second pouches, they are not expressed in maxillary CNCCs anterior to the oral ectoderm. Concomitantly, ventral gene expression expands into the dorsal mandibular and hyoid domains, but not the maxillary domain, of jag1b^b1105 and notch2-MO embryos. Our previous fate maps of wild-type arches (Crump et al., 2006; Eberhart et al., 2006) also help to explain why Hm and Op are strikingly transformed in shape, yet only a portion of Pq is lost, in jag1b^b1105 mutants. Whereas Hm and Op derive entirely from dorsal hyoid CNCCs, only a portion of Pq derives from dorsal mandibular CNCCs, with the Ptp process of Pq deriving from maxillary CNCCs that are unaffected in jag1b^b1105 mutants.

Species-specific differences might also account for the relative importance of Jagged-Notch signaling in facial patterning. Whereas the majority of the craniofacial skeleton of the larval zebrafish derives from mandibular and hyoid CNCCs, in mammals most of the facial skeleton derives from frontonasal, maxillary and ventral mandibular CNCCs, with dorsal mandibular and hyoid CNCCs contributing prominently to the ossicles of the middle ear. Recently, Pofut1^flox/–; Wnt1-Cre mice have been generated that lack Notch signaling throughout CNCCs owing to the tissue-specific deletion of O-fucosyltransferase 1, an essential component of Notch signaling (Okamura and Saga, 2008). The lack of severe craniofacial defects in Pofut1^flox/–; Wnt1-Cre mice is consistent with our findings in zebrafish that Jag1b and Notch2 are not required for the patterning of maxillary and frontonasal CNCCs. As we find that Jag1b and Notch2 inhibit ventral skeletal identity only in the dorsal mandibular and hyoid domains, it will be interesting to examine whether a role of JAG1 in development of the ossicles of the middle ear contributes to the conductive hearing loss seen in some Alagille syndrome patients (Le Caignec et al., 2002).

Our analysis of jag1b transcriptional regulation also indicates a potential mechanism by which Notch patterns a broad field of dorsal skeletal precursors. In contrast to secreted morphogens, Jagged and Notch are transmembrane proteins and do not generally act at a distance. In the Drosophila wing, Notch signaling induces the expression of morphogens, in particular Wingless, which act at a distance to pattern the dorsal domain (Diaz-Benjumea and Cohen, 1995). However, our observation that the Notch2 target hey1 is expressed throughout dorsal skeletal precursors suggests a more direct role for Notch in dorsal facial patterning. Similar to what is observed during early inner ear development (Daudet et al., 2007), we find that Jagged-Notch transcriptional feedback serves to extend jag1b expression more ventrally over time. Such a mechanism would propagate Notch activity throughout the dorsal facial domain, ensuring that each cell within the dorsal field experiences Notch activity at some point during development.

Whereas Jagged-Notch feedback extends Notch activity, we find that Edn1 signaling prevents Notch activity from spreading into the ventral domain by repressing jag1b expression. Previous studies have demonstrated a nearly complete loss of the ventral facial skeleton in zebrafish edn1 mutants, suggesting that Edn1 is essential for development of the ventral facial skeleton (Miller et al., 2000). By contrast, we find that in the absence of Jag1b-Notch2 signaling, Edn1 is partially dispensable for development of the ventral face. Thus, the ventral gene expression and skeletal defects of edn1 mutants can at least partially be attributed to aberrant jag1b expression and hence Notch activity in the ventral domain. However, the partial nature of the rescue in edn1; jag1b^b1105 mutants suggests that Edn1 might also have Notch-independent roles in promoting ventral gene expression and skeletal development. Moreover, we find that partial reduction of Edn1 signaling fails to rescue jag1b^b1105 defects, and the expression of edn1 and its receptor ednra2 are not regulated by Jag1b-Notch2 signaling. Thus, the dorsal-to-ventral transformations of jag1b^b1105 mutants are not due to increased Edn1 signaling. What, then, might account for the striking ability to form a fairly normally patterned ventral hyoid skeleton even in the complete absence of Edn1? Studies in mouse suggest that unknown redundant signals function in parallel with Edn1 to promote ventral skeletal fates (Ruest et al., 2004). In the absence of Edn1, these redundant signals might be unable to sustain ventral skeletal development owing to ectopic repressive Notch activity in the ventral domain. However, in the absence of both Edn1 and Notch signaling, redundant ventralizing signals might now be able to partially support ventral skeletal development. Further investigation of ventral patterning in the absence of both Edn1 and Jagged-Notch signaling should help to reveal the extent of functional redundancy in specifying the DV axis of the facial skeleton.

We thank Gabriel Finch, Megan Matsutani, Pablo Castillo and Corey Gingerich for fish care; Rob Maxson, Henry Sucov, Yang Chai, B. Frank Eames and Charles Kimmel for comments; and Jared Talbot for the fluorescence in situ protocol. Will Talbot performed the initial mapping of jag1b^b1105. hsp70I:Gal4 fish were obtained from the Zebrafish International Resource Center at the University of Oregon. Research was supported by an NIH F31 grant to E.Z. and NIH R01 DE018405 and March of Dimes grants to J.G.C. NIH DE13834 and HD22486 grants funded the University of Oregon skeletal screen. Deposited in PMC for release after 12 months.