HPE is a common malformation of the face in humans. Its etiology is highly heterogeneous, likely due to the involvement of multiple genetic factors that are required in a variety of craniofacial tissues and at various times during facial growth and patterning [17, 73]. Disruption of the Hh-signaling pathway is the most frequently reported deficit underlying HPE, or its closely related clinical microforms of the disorder, in humans . For this reason, multiple studies have sought to experimentally demonstrate the requirements for the Hh-signaling pathway in craniofacial tissues and facial precursors. In the zebrafish, these previous studies have either focused solely on the development of the skull and upper jaw [18, 23, 33], or utilized severe loss-of-function alleles representing extreme cases of cranial skeleton loss . Here we provide a detailed characterization for the role of Hh-signaling in the PA skeleton by utilizing a hypomorphic allele of Hh-signaling (con/disp1). Indeed, it has been suggested by others that these less-severe alleles may provide very useful insights into the etiology of Hh-associated HPE as they display characteristic craniofacial abnormalities in the absence of more debilitating phenotypes, such as cyclopia and nervous system deficits [2, 74].
Our study revealed multiple requirements for Hh-signaling during CNCC development that influences the final facial skeletal pattern. First, Hh-signaling controls jaw outgrowth by influencing both patterning and chondrogenic differentiation of postmigratory-CNCC that reside in the anterior arches. Using the Hh-inhibitor Cya we revealed that Hh-signaling is needed during gastrulation for proper jaw formation. Hh-inhibition during gastrulation in zebrafish and chick embryos is also associated with cyclopia and mid-facial dysplasia (data not shown, ), and has also been shown to influence patterning and chondrogenesis in the zebrafish dorsal skeleton [18, 23]. These phenotypes mimic severe loss-of-function mutations within human SHH in HPE, which is characterized by facial anomalies including cyclopia, cleft lip, and underdeveloped jaw . Despite impairing jaw outgrowth, multiple cranial structures remained in the lower jaw of con/disp1 larvae. This not only revealed that some cranial structures were more sensitive to Hh-attenuation (i.e. the dorsal jaw and jaw support elements), but also allowed for detection of deficits in the jaw joint.
Our work further revealed a requirement for Hh-signaling in cartilage development within the posterior gill-forming PA. con/disp1 mutant posterior arch-residing CNCC fail to chondrify, despite becoming properly patterned within well-formed endodermal pouches. Cya treatments revealed the single and crucial time for Hh-signaling in this event is during the late-pharyngula stage, well beyond the time when CNCC have migrated into the PA primordia. It is not surprising that Hh-signaling is required by postmigratory-CNCC in the posterior arches as Hh-ligands become widely expressed throughout the facial epithelial tissues shortly after the time in which CNCC migrate into the face. By examining CNCC differentiation in these arches, we found that certain genes are specifically downregulated in the absence of Hh-signaling, and histological preparation of the gill arches suggest that CNCC are transfated to fibrous-connective tissue as a result. It is clear from these findings that multiple signaling interactions involving the Hh-pathway must occur for proper craniofacial development, and this likely reflects the variability in human phenotypes associated with disruptions to the pathway 
Hh-signaling is required for mandibular arch patterning and chondrogenesis
Frontonasal development and the outgrowth of anterior craniofacial structures are strongly influenced by the facial oral ectoderm [21, 23, 75]. Fate mapping has revealed that dorsal CNCC precursors (neurocranium and pq) condense upon the oral ectoderm roof upon their arrival to the face, while ventral CNCC precursors (Mc and a subset of pq) migrate to, and condense upon, the oral ectoderm floor [18, 41]. The ability of CNCC to condense upon the oral stomodeum is negatively affected by the loss of Hh-signaling , likely resulting in the aberrant localization of first arch CNCC. In our studies of fli1GFP:con/disp1 mutants, CNCC that have migrated into the first arch were conspicuously absent just anterior to the stomodeum, and rather these cells preferentially localized posterior to the stomodeum. Interestingly, the anterior-most CNCC in the fli1GFP:wild type siblings never localized posterior to the stomodeum, suggesting that Hh-signaling may normally act to restrict CNCC from this region, either through promoting CNCC-stomodeum adhesion and/or by inducing midline tissue expansion posterior to the stomodeum that would serve as a physical barrier.
In addition to patterning defects, our studies revealed cell differentiation defects in the anterior-most CNCC. Local morphogenetic signals from the oral ectoderm, such as shh, are likely involved in promoting mesenchymal differentiation and skeletal outgrowth in the jaw and neurocranium. In chick embryos, Shh from the oral ectoderm influences expression of critical CNCC markers to promote maxillary and frontonasal outgrowth [21, 76, 77]. By closely examining the cell differentiation defects in the con/disp1 mutants, we found significant variability in the expression of CNCC markers between mesenchymal condensations in different DV domains within the lower jaw (PA1-2). Notably, neural crest-specific gene expression often appeared to be reduced in con/disp1 dorsal condensations within anterior PA, while expression in ventral condensations were less affected. This is similar to the anterior craniofacial development of Shh-/- mouse embryos wherein severe hypoplasia is seen in the proximal (dorsal) mandibular arch and maxillary mesenchyme, while the more distal (ventral) mandibular arch regions are less affected . Furthermore, a shortened, hypoplastic ventral, first arch Mc element is the only remaining structure when Hh-responsiveness is specifically removed in all cranial precursors (CNCC) in the mouse . Collectively, this suggests that the dorsal jaw, dorsal jaw support and dorsal neurocranium are more sensitive to Hh-attenuation than the ventral jaw, a hypothesis that is supported by the suggestion that dorsal and ventral mandibular identities are controlled by specific sets of transcription factors and signaling molecules . This has been validated by the existence of a number of zebrafish jaw mutants where the dorsal and ventral structures within the first two arches are affected to differing degrees (chameleon, sucker, schmerle) .
Finally, our findings that Hh-signaling is required during gastrulation (4-10 hpf) for proper ventral jaw development is similar to previous findings that this time period is required for proper neurocranium development . Thus, patterning of all anterior craniofacial structures (both dorsal and ventral) is likely controlled by the same Hh-signaling event that originates in the ventral brain primordium during gastrulation, as hypothesized by Eberhart et al., 2006. However, unlike dorsal neurocranium precursors that continue to require Hh-signaling for subsequent chondrogenesis up to 2 dpf , we show that jaw precursors require a single Hh-signaling event during gastrulation for both patterning and chondrogenesis (see Figures 8 and 9). Importantly, this early Hh-signal also influences later gene expression in the oral ectoderm, such as pitx2  and shh itself (data not shown), which are good candidate genes for signals required by jaw precursors for CNCC chondrogenesis. However, more work will be required in order to determine the specific Hh-dependent signals for the proper development of both the neurocranium and jaw in the zebrafish.
Hh-signaling is required for patterning the jaw joint region in the mandibular and hyoid arch
Our studies here reveal multiple requirements for disp1 and Hh-signaling in the development of the jaw joint region. Larval cartilage preparations show absence of the jaw joints, both bilaterally in the anterior arches and within the midline of the mandibular arch. Additional morphological defects included the loss of the RAP extension of the Mc element in the mandibular arch. These phenotypes were reminiscent of bapx1-morphant larvae . bapx1 is expressed in arch mesenchyme and prospective joint domains within the first and second arch and its absence results in a failure of the joints to become properly specified, potentially through influencing gdf5 expression which is required in mouse appendicular and axial joint formation [63, 66, 67]. The joint specifying marker bapx1 was either reduced (bilaterally) or completely absent (midline) in con/disp1 mesenchyme. These bapx1 gene disruptions were sufficient to reduce gdf5 expression to barely detectable amounts in the con/disp1 embryos. Collectively, this data provides a genetic mechanism for Hh-signaling in both the development of jaw joint regions as well as the ventral midline bh element, which is also lost in con/disp1 larvae and requires gdf5 function .
Furthermore, careful analysis of the bilateral joints in the hyoid arch revealed two consistent, varying phenotypes: either ectopic cartilage cells forming an aberrant joint between the hs and ch elements or the complete absence of cartilage cells at the would-be joint site, leading to the failure of hs and ch elements to properly articulate. The presence of these two phenotypes in the mutant hyoid arch cartilages likely reflects two separate requirements for Hh-signaling in CNCC and neighboring cells that reside within future joint sites; first, in CNCC chondrogenesis to promote cartilage formation (as described in Figures 1 and 3) and secondly in joint organization within cells that will eventually chondrify (as described above). The different mutant phenotypes may reflect very slight differences in the amount of Hh-signaling present in the joint regions, as we occasionally visualized both of the described PA2 mutant phenotypes occurring on opposite sides of the same con/disp1 larvae (left and right sides of single larvae shown in Figure 6J, K).
Hh-signaling controls chondrogenic differentiation of CNCC in the posterior arches
Unlike con/disp1 CNCC that have migrated into the mandibular arch, CNCC migrating into the hyoid and posterior arches condense normally. These crest cells become segmentally patterned within well-formed endodermal pouches, indicating that Hh-signaling is not involved in endodermal pouch morphogenesis nor is it required for invading CNCC to adhere to the PA2-7 epithelia. However, our studies reveal that those cells residing in the posterior arches require Hh-signaling to undergo chondrogenic differentiation. This is not the case in the hyoid arch, however, wherein normal CNCC differentiation and subsequent chondrification occurs. Variations in the potential of PA2-7 CNCC to differentiate may be due to intrinsic differences in the ability of these cell groups to respond to Hh-signals, or Hh-dependent co-factor(s), in the facial primordia. CNCC-responsiveness to local facial patterning cues depends in part on their Hox gene expression, which is determined by the original location of these cells in the hindbrain. Moreover, Hox status differs between CNCC in the second arch (hoxa2) and the posterior arches (hoxa2/hoxb3) . In chick embryos, grafts of facial ectoderm or foregut endoderm, into the PA2 region has no influence on the fate of this arch [4, 21]. Rather, PA2 mesenchyme is thought to be influenced by the pharyngeal ectodermal margin, and shh from this tissue has been implicated (directly or indirectly) in hyoid outgrowth [82, 83].
The absence of cartilage elements in the posterior arches of 96 hpf con/disp1 mutant larvae is not due to changes in cell survival or proliferation as we were unable to find any variation in cell death or cell proliferation in the craniofacial region between con/disp1 mutants and their wild type siblings. This is consistent with zebrafish studies utilizing both smu/smo mutants and Shha;Shhb double morphant embryos, which lack dorsal cartilages yet display no changes in cell death or proliferation [18, 23]. Rather, we found that Hh-signaling is involved in maintaining the chondrogenic differentiation potential of posterior arch mesenchyme. Specifically, critical transcription factors, sox9a and dlx2a, become down-regulated in these cells during the second day of development. Notably, other critical transcription factors such as hand2, barx1, msxb, msxe were not downregulated in these cells indicating that these cells persist in the posterior arches in a differentiated state that still resembles their neural crest origin.
Genetic analysis on the sox9a/jellyfish (jef) mutant has revealed a critical function for this gene in cartilage development within the posterior arches, potentially by influencing normal col2a1 expression in a similar fashion to the con/disp1 mutant . The critical importance of sox9a in directing PA3-7 prechondrogenic condensations to develop as chondrocytes is underscored by the findings that other chondrogenic factors, like runx2b, are expressed at normal levels in the sox9a/jef and con/disp1 mutants alike [ and our studies].
So what then happens to the mesenchyme within the posterior arches of con/disp1 mutants? Our histological sections on 5 dpf con/disp1 mutants suggest that CNCC in PA3-7 give rise to fibrous connective tissue, a common derivative of pluripotent mesenchymal neural crest cells. We hypothesize that these ectopic fibrous-connective tissue cells are of CNC origin and have committed to a mesenchymal lineage, instead of a neural one, independently of Hh-signaling. These cells then require the Hh-pathway to adopt, or maintain, a chondrogenic fate. Consistent with this, exogenous Shh can induce multipotent CNCC and pluripotent mesenchymal stem cells to adopt a chondrogenic fate in vitro [84–87]. Examining later skeletal development in the con/disp1 mutant pharyngeal arches is frustrated by early embryonic death by 6-7 dpf in these animals.
Our stage-specific Hh-signaling inhibition studies revealed that the critical time for Hh-signaling to maintain chondrogenic differentiation in posterior arch-residing CNCC is the late pharyngula stage (32-48 hpf). At this time, shh and disp1 are coexpressed in facial ectoderm and endoderm, as well as the ventral brain, making it difficult to interpret the crucial Hh-source for posterior arch cartilage formation. An intriguing tissue source of shh in this process is the pharyngeal endoderm. In casanova mutants, which fail to make endoderm, all PA-derived cartilages are lost . Further, surgical removal of foregut endoderm in chick embryo prior to CNCC migration results in a loss of ventral arch structures, yet arch identity is restored when Shh-soaked heparin beads are placed in the region of the excised tissue . More studies will be required to determine the cell autonomy of Hh-ligands in posterior arch cartilage formation, as well as whether the signal must be interpreted by mesenchymal-neural crest and/or the surrounding epithelia.
Conserved and non-conserved roles for Hh-signaling in PA development
Zebrafish PA skeletal defects in con/disp1 mutants resemble loss-of-function mutations to Shh in mouse [89, 90] and treatments with Shh-blocking-antibodies in chick embryos , indicating a conserved role for Hh-signaling in PA development. However, unlike in fish, the primary defect attributing to the loss of mouse and chick PA skeleton is a high degree of cell death in CNCC as they migrate into the pharyngeal pouches. Additionally in Shh-/- mouse, further cell death is seen in the pharyngeal pouch endoderm , indicating a pro-survival role for Hh-signaling in epithelial and mesenchymal facial tissue.
Hh-signaling may mediate its pro-chondrogenic (zebrafish) or pro-survival (mouse, chick) effects within CNCC through its influence on Sox9 gene activity. We noticed a strong correlation between the loss of sox9a expression in con/disp1 CNCC and the failure of those same cells to chondrify. In Shh-/- mouse there is a striking pattern of Sox9 gene expression loss in streaming CNCC just prior to cell death within the same regions . Inverse experiments, where ectopic Shh was used to stimulate mouse or human chondrocytes revealed that increased levels of Shh and downstream target genes in chondrocytes correlated with higher levels of Sox9 expression [91, 92]. In zebrafish, due to gene duplication there are two orthologs of Sox9, sox9a and sox9b, which are co-expressed in postmigratory-CNCC and they control chondrocyte morphogenesis and survival, respectively [13, 44, 93–95]. These roles are likely controlled by the single tetrapod Sox9 in higher vertebrates and can account for the cell death seen in these animals upon gene loss, whereas we show here that, for the most part, sox9b (survival) gene expression is independent of Hh-signaling in CNCC within the PA.