GATA3 is essential for separating patterning domains during facial morphogenesis

Gnathostomes (hinged-jaw vertebrates) account for >99% of all vertebrates. Separation of the upper and lower jaws requires establishment and maintenance of gene expression boundaries in the first pharyngeal arch during early embryo development, ensuring that the mandible and maxilla articulate with the skull in a manner that allows the formation of a hinged jaw (Clouthier et al., 2010; Medeiros and Crump, 2012; Neben and Merrill, 2015). Disruption of this process leads to fusion of the upper and lower jaws, a condition called syngnathia (OMIM: 119550).

One current theory that explains how differential patterning of the first pharyngeal arch is achieved is referred to as the ‘hinge and caps’ model (Depew et al., 2002, 2005; Depew and Compagnucci, 2008). In this model, ectodermal signaling centers (caps) in the ventral (distal) mandibular arch and the junction of the maxillary prominence and the olfactory placode (the lambdoidal junction) provide instructive signals dorsally (proximally) towards the maxillomandibular junction (the hinge), instructing development of the first arch (Fig. 1). Signals from the maxillomandibular junction also participate in patterning the two prominences while ensuring that mandibular and maxillary cap signals remain separated. This is important, as NCCs within the maxillary and mandibular portions of the first arch are initially competent to respond to signals normally found in the other half of the arch (Ferguson et al., 2000; Sato et al., 2008; Tavares and Clouthier, 2015).

Multiple cap signals establish patterning domains within the first arch of mice and zebrafish, with these domains effectively defining the dorsoventral axis of the arch. These include endothelin 1 (EDN1), bone morphogenetic proteins (BMPs), SIX1 and PITX1 (Liu et al., 2003, 2005; Clouthier et al., 2010; Alexander et al., 2011; Compagnucci et al., 2011). However, from a functional standpoint, separation of cap signals by the hinge region is not as well understood. One of the key regulators thus far identified in this process in the mouse is FOXC1, which genetically interacts with Fgf8 in the hinge region to ensure normal separation of the upper and lower jaw and establishment of the temporomandibular joint (TMJ) (Inman et al., 2013). Although loss of Foxc1 leads to fusion of the mandible and maxilla (syngnathia), expression of ventral mandibular arch markers, including Hand2, are unaffected, indicating that the FOXC1 gene regulatory network acts specifically in the maxillomandibular junction. It is less clear how cap signals interact with hinge signals and how the two domains are integrated.

GATA3 belongs to the evolutionarily conserved GATA family of zinc-finger transcription factors. Mutations in GATA3 are associated with autosomal dominant hypoparathyroidism, sensorineural deafness and renal anomaly (HDR) syndrome (reviewed by Van Esch and Devriendt, 2001). Targeted deletion of Gata3 in mice have supported these findings, with defects observed in cardiac, renal, ocular, craniofacial, sympathetic neuron and immune system development (Pandolfi et al., 1995; Lim et al., 2000; Ho and Pai, 2007; Grote et al., 2008; Hendershot et al., 2008; Maeda et al., 2009; Raid et al., 2009). Although the basis of renal, ocular and cardiovascular defects has been characterized, the cause of the craniofacial defects remains unclear. Here, we demonstrate that loss of Gata3 disrupts patterning of post-migratory NCCs within pharyngeal arch one owing to early expansion of a BMP GRN and subsequent intrusion of cap signals into the maxillomandibular junction, resulting in syngnathia. Interestingly, the severity of skeletal defects is asymmetric, suggesting that Gata3 mutant embryos may be a useful model in which to investigate how facial symmetry is disrupted in human conditions in which facial asymmetry is lost, including hemifacial (craniofacial) microsomia (OMIM: 164210).

To assess the role of GATA3 in craniofacial development, we examined mice in which a lacZ reporter cassette was inserted into the Gata3 gene, resulting in a null allele (Lakshmanan et al., 1999). Because loss of Gata3 leads to embryonic lethality around embryonic day (E) 13.5 owing to noradrenaline deficiency resulting from defects in sympathetic neuron development (Lim et al., 2000), we employed a pharmacological rescue strategy in which pregnant dams were given water containing the α- and β-adrenergic receptor agonists isoproterenol and phenylephrine (Hendershot et al., 2008). Using this approach, Gata3^z/z embryos survived to E18.5, allowing a more complete analysis of bone and cartilage defects. Compared with E18.5 Gata3^+/+ embryos (Fig. 2A,C), E18.5 Gata3^z/z embryos had a narrow jaw region and presented with micrognathia and microstomia (Fig. 2B,D) (n=5 for each group). Skeletal staining of these embryos demonstrated that, compared with Gata3^+/+ embryos (Fig. 2E, Fig. S1A,C), most of the proximal mandible and maxilla was absent in Gata3^z/z embryos (Fig. 2F, Fig. S1B,D), with the distal mandible fused with the maxilla (syngnathia) (Fig. 2F, arrow). In Gata3^z/z embryos, the amount of proximal craniofacial bone present was asymmetric, with one side always having more bone than the other, although there was not a specific sidedness to this finding (n=2 with more bone on the right; n=3 with more bone on the left; Fig. S1B,D). This asymmetry resembles craniofacial microsomia. Cleft palate was also present (see below).

To assess these changes better, the heads of stained skeletons were then scanned using micro-computed tomography (µCT). These scans showed hypoplasia and fusion of the mandible and maxilla in Gata3^z/z embryos (Fig. 2H) compared with Gata3^+/+ embryos (Fig. 2G). In addition, although lower incisors were present in Gata3^z/z embryos (Fig. 2H,J), they appeared smaller compared with the incisors of control embryos (Fig. 2G,I).

To define the basis of the syngnathia, we examined earlier skeletal changes in E16.5 Gata3^+/+ and Gata3^z/z embryos. Compared with Gata3^+/+ embryos (Fig. 3A), the maxilla in Gata3^z/z embryos were present but hypoplastic (Fig. 3B) (n=4 for each group). However, the jugal bone, which articulates with the zygomatic process of the maxilla and the zygomatic process of the squamosal bone in Gata3^+/+ embryos (Fig. 3A), failed to extend posteriorly in Gata3^z/z embryos, instead extending caudally and fusing with the mandible (Fig. 3B, black arrow). Mandibular hypoplasia was already evident, including absence of bone in the proximal mandible. The mandible and maxilla of all stained E16.5 Gata3^z/z embryos (n=4) also skewed to the left side (Fig. 3A,B, insets).

We next performed histological analysis using frontal sections through the head of E16.5 Gata3^+/+ and Gata3^z/z embryos stained with Hematoxylin and Eosin (H&E) (n=5 for both groups). Compared with Gata3^+/+ embryos (Fig. 3C), the distal oral cavity was almost completely closed with oral ectoderm (Fig. 3D, arrow). More proximally in the oral cavity, the mandible of Gata3^+/+ embryos appeared symmetric (Fig. 3E), whereas the mandible of Gata3^z/z embryos was skewed to the left and was fused to the maxilla, resulting in microstomia (Fig. 3F) (n=5). The mandible and Meckel's cartilage were dysmorphic (Fig. 3F) and the tongue was hypoplastic, with disorganized connective tissue and muscle fibers (Fig. 3H) compared with the organization observed in the tongues of Gata3^+/+ embryos (Fig. 3G). Although the upper and lower incisors showed some size variation in Gata3^z/z embryos, overall development appeared normal (Fig. 3F, Fig. S2B) compared with Gata3^+/+ embryos (Fig. 3E, Fig. S2A). In contrast, both upper and lower molar development appeared to be arrested at the early bud stage (Fig. 3H, arrows). Cleft palate was also present in all Gata3^z/z embryos; in sectioned embryos, the right palatal shelf was not elevated and the left shelf was absent (Fig. 2J; n=4/5, with the fifth embryo ambiguous in shelf development). Proximal jaw structures were absent, including the mandibular condyles and the temporomandibular joint (Fig. S2D) compared with Gata3^+/+ embryos (Fig. S2C). The tympanic ring and gonial bone (Fig. S2F), malleus (Fig. S2H) and incus/stapes (Fig. S2J) were all present but dysmorphic compared with Gata3^+/+ embryos (Fig. S2E,G,I). It is not possible to determine whether dysmorphology of middle ear structures was a primary event or due secondarily to changes in other jaw elements.

Disruption in early pharyngeal arch patterning signals often also affect cranial ganglia development. When Gata3^+/+ embryos were stained with an antibody against NF160, a neuronal marker, the ophthalmic, maxillary and mandibular branches of the trigeminal ganglia (V) and the facial (VII) nerve were normal in appearance and projection (Fig. 4A) (n=5). In contrast, most of the maxillary branch of the trigeminal ganglia was absent in stained Gata3^z/z embryo, although a small piece of nerve tissue remained below the optic placode (Fig. 4B, white arrow) (n=5). In addition, the facial nerve (VII) showed aberrant branching (Fig. 4B, black arrows).

By E13.5, in contrast to Meckel's cartilage in Gata3^+/+ embryos (Fig. 4C), Meckel's cartilage in E13.5 Gata3^z/z embryos contained a gap in the left Meckel's cartilage of varying length (Fig. 4D, area between black arrows; n=4/4), with the symphysis curved downward. A smaller gap was present on the right Meckel's cartilage (n=3/4). In H&E-stained frontal sections through the head of E13.5 embryos, the distal microstomia observed in E16.5 Gata3^z/z embryos (Fig. 3D,F) was already apparent at this earlier age (Fig. 4F) compared with Gata3^+/+ embryos (Fig. 4E). At the level of the molars, upper and lower molars were present in Gata3^+/+ embryos (Fig. 4G), with downward-oriented palatal shelves and Meckel's cartilage also present. In Gata3^z/z embryos, molar tooth buds were not obvious (Fig. 4H). In addition, palatal shelves were not readily apparent (Fig. 4H; n=4), suggesting slowed palatal shelf development.

Craniofacial bone hypoplasia/aplasia and the absence of the maxillary branch of the trigeminal nerve in Gata3^z/z embryos are suggestive of defects in NCC specification, migration or differentiation. To assess these possibilities, we examined expression of Sox10 (which marks migrating NCCs at E8.5 before becoming restricted to cranial ganglia at E9.5; Marmigere and Ernfors, 2007) and Tfap2 (a transcription factor involved in multiple facets of cranial NCC development; Brewer et al., 2004). In E8.5 embryos, Sox10 expression was observed in both Gata3^+/+ (Fig. 5A,B) and Gata3^z/z (Fig. 5C,D) embryos along the neuroepithelium of the posterior midbrain and hindbrain, extending down to the preotic sulcus (pos), although expression was weaker in Gata3^z/z embryos (Fig. 5C,D). Tfap2a expression was also observed along the neuroepithelium of the posterior midbrain and hindbrain in E8.5 Gata3^+/+ (Fig. 5E,F) and Gata3^z/z (Fig. 5G,H) embryos, although this was reduced in Gata3^z/z embryos, with a more dramatic reduction caudal to the pos (Fig. 5G,H). By E9.5, Sox10 expression was restricted to the developing cranial ganglia in Gata3^+/+ (Fig. 5I) and Gata3^z/z (Fig. 5J) embryos, but was weaker in Gata3^z/z embryos. Similarly, Tfap2a expression in E9.5 Gata3^+/+ (Fig. 5K) and Gata3^z/z (Fig. 5L) embryos was broadly observed in the mandibular portion of arch one, extending towards the maxillary prominence (Fig. 5K,L, black arrows), although, like Sox10 expression, overall staining was reduced in Gata3^z/z embryos. These findings indicate that although there is not a major disruption in NCC migration in Gata3^z/z embryos, either a subset of NCCs does not reach the arches or normal gene expression within the NCCs is disrupted.

We next examined whether there was a difference in cell death or proliferation in post-migratory NCCs within the mandibular arch at E9.5. Whole-mount terminal deoxynucleotidyl transferase biotin-dUTP nick end labeling (TUNEL) analysis in E9.5 embryos revealed that the relative level of cell death between Gata3^+/+ (Fig. 6A) and Gata3^z/z (Fig. 6B) embryos was similar in the area near and within the pharyngeal arches. To quantify cell death better, TUNEL analysis was performed on sections through the mandibular arch of E9.5 embryos that were collected 1 h after the pregnant female was injected with 5-ethynyl-2′-deoxyuridine (EdU). This analysis revealed that, as a percentage of total cells, there was no statistically significant difference in cell death in the NCC-derived mandibular arch mesenchyme between Gata3^+/+ (Fig. 6C,G) and Gata3^z/z (Fig. 6D,G) embryos, although the overall number of TUNEL-positive cells was low for both groups (a range of 0-19 positive cells per section in Gata3^+/+ embryos and 0-7 positive cells per section in Gata3^z/z embryos). EdU incorporation, which identifies cells in S phase, was also quantified on the same sections used for TUNEL analysis. Similarly, there was no statistically significant difference in proliferating cells, as a percentage of total cells, in the mandibular arch mesenchyme between Gata3^+/+ (Fig. 6E,H) and Gata3^z/z (Fig. 6F,H) embryos. For both proliferation and cell death, asymmetric or localized differences in labeled cells were not detected. However, the total number of mandibular arch mesenchyme cells, as detected by DAPI staining of the sections used for TUNEL and EdU analysis, was significantly reduced in Gata3^z/z mutants compared with that in Gata3^+/+ embryos (Fig. 6I). This supports the idea that fewer NCCs populated the mandibular arch of Gata3^z/z embryos.

We have previously shown that Gata3 is expressed in the ventral mandibular arch and the maxillary prominence/nasal placode by E9.5, with this pattern being more prominent by E10.5 (Ruest et al., 2004). Although the ventral mandibular arch mesenchyme domain of Gata3 corresponds to the Hand2 domain (Clouthier et al., 2000), Hand2 expression is not regulated by GATA3 (Ruest et al., 2004); in fact, in E10.5 Gata3^z/z embryos, ventral mandibular arch Hand2 expression expands in a dorsorostral direction, approaching the maxillomandibular junction (Ruest et al., 2004). These findings suggest that GATA3 may regulate a gene regulatory network (GRN) that directly or indirectly prevents encroachment of cap gene expression into the junction, thus allowing for development of distinct upper and lower jaws.

To investigate this possibility further, we examined the expression of genes that pattern the dorsal mandibular arch/hinge region. For all in situ hybridization (ISH) expression assays described below (Figs 7 and 8), asymmetry in gene expression was not observed in either Gata3^+/+ or Gata3^z/z embryos. We first examined Pitx2 expression, as PITX2 acts in a dosage-dependent manner to regulate the expression of several genes around the maxillomandibular junction (Liu et al., 2003). In E9.5 and E10.5 Gata3^+/+ embryos, Pitx2 expression extended along the ectoderm from the maxillary prominence to the rostral half the mandibular arch (Fig. 7A,C; extent denoted by arrows) (Liu et al., 2003). A similar expression pattern was observed in E9.5 and E10.5 Gata3^z/z embryos (Fig. 7B,D; extent denoted by arrows), although both the maxillary and mandibular prominences appeared smaller compared with Gata3^+/+ embryos.

We next examined expression of the PITX2 target gene Fgf8 (Liu et al., 2003). At E9.5 and E10.5, Fgf8 expression in Gata3^+/+ embryos was observed along the ectoderm of the maxillary prominence and rostral half of the mandibular arch, spanning the maxillomandibular junction (Fig. 7E,G; extent denoted by arrows) (Trumpp et al., 1999; Liu et al., 2003). This expression pattern was similar in E9.5 Gata3^z/z embryos (Fig. 7F), but was less pronounced and appeared broader on the maxillary prominence compared with Gata3^+/+ embryos. In contrast, Fgf8 expression in E10.5 Gata3^z/z embryos expression was undetectable in the first arch ectoderm (Fig. 7H). This absence suggests that changes in Fgf8 expression are unrelated to PITX2 activity.

We next examined Barx1, a prominent FGF8 target (Trumpp et al., 1999), expression of which can be repressed by BMP4 (Mitsiadis et al., 2003). In E9.5 Gata3^+/+ embryos, strong Barx1 expression was observed in the dorsal and intermediate domains of the mandibular arch mesenchyme (Fig. 7I) (Clouthier et al., 2000) but was almost absent in E9.5 Gata3^z/z embryos (Fig. 7J). At E10.5, Barx1 expression in Gata3^+/+ embryos was observed in the mesenchyme of the mandibular and maxillary portions of arch one (spanning the maxillomandibular junction) and in the mesenchyme of arch two (Fig. 7K) (Clouthier et al., 2000). In E10.5 Gata3^z/z embryos, expression in arch two was similar to that of Gata3^+/+ embryos, although arch one expression was restricted to an area immediately surrounding the maxillomandibular junction (Fig. 7L).

Pitx1 expression is sensitive to Pitx2 gene dosage (Liu et al., 2003), although its expression relies also on Fgf8 expression (Trumpp et al., 1999). At E9.5, Pitx1 expression appeared along the ectoderm of the maxillary prominence, across the maxillomandibular junction and onto the mandibular prominence in Gata3^+/+ (Fig. 7M) (Lanctot et al., 1997) and Gata3^z/z (Fig. 7N) embryos. By E10.5, Pitx1 expression in Gata3^+/+ embryos remained along the ectoderm of the maxillary and mandibular arches and also extended into the intermediate mandibular arch mesenchyme (Fig. 7O) (Lanctot et al., 1997; Liu et al., 2003). In contrast, although ectoderm expression of Pitx1 was unchanged in E10.5 Gata3^z/z embryos, Pitx1 expression in the mandibular arch mesenchyme was reduced and shifted dorsally, localizing around the maxillomandibular junction (Fig. 7P). Thus, loss of Gata3 disrupts a FGF8 GRN that includes Barx1 and Pitx1, with expression of these genes either greatly reduced or shifted more dorsally.

Fgf8 expression in the maxillomandibular junction in part establishes the dorsal identity in the first arch, although this expression relies on the exclusion of BMP signaling (Liu et al., 2003), as does the mesenchymal expression of Barx1 (Mitsiadis et al., 2003). We thus examined the expression of Bmp4 in E9.5 and E10.5 Gata3^+/+ embryos, finding prominent Bmp4 expression on the ectoderm of the dorsal maxillary and mandibular portions of arch one, although expression was excluded from the maxillomandibular junction (boundaries denoted by black arrows in Fig. 8A,C) (Liu et al., 2005). In contrast, Bmp4 expression in E9.5 and E10.5 Gata3^z/z embryos was observed on the ectoderm of the maxillary and mandibular portions of arch one (Fig. 8B,D, arrows), including the ectoderm of the maxillomandibular junction (Fig. 8B,D). These findings suggest that aberrant Bmp4 expression in the maxillomandibular junction may contribute to gene expression changes in this domain.

We next examined expression of two BMP4-induced genes, Lhx8 and Msx1 (Liu et al., 2003; Bonilla-Claudio et al., 2012). In E9.5 Gata3^+/+ embryos, Lhx8 expression was observed in the ventral mesenchyme of the mandibular and maxillary prominences but not in the mesenchyme near the maxillomandibular junction (Fig. 8E; boundary of expression denoted by arrows) (Tucker et al., 1999). In E9.5 Gata3^z/z embryos, Lhx8 expression shifted to the area around the maxillomandibular junction (Fig. 8F) in a pattern similar to aberrant Bmp4 expression in Gata3^z/z embryos (Fig. 8B). A similar shift in gene expression was observed for Msx1 in Gata3^z/z embryos. Whereas Msx1 is normally expressed in the ventral mandibular arch mesenchyme of Gata3^+/+ embryos at E9.5 (Fig. 8G) (Tucker et al., 1999) and E10.5 (Fig. 8I) (Ruest et al., 2004), Msx1 expression in Gata3^z/z embryos was observed in the mesenchyme adjacent the ectoderm along the maxillary and mandibular prominences, including around the maxillomandibular junction at both time points (Fig. 8H,J). Hand2 expression in E9.5 Gata3^+/+ embryos was confined to the ventral area of the arch (Fig. 8K) (Clouthier et al., 2000) whereas expression in E9.5 Gata3^z/z embryos extended dorsally along the rostral half of the arch (Fig. 8L). A similar shift in Hand2 expression was previously observed in E10.5 Gata3^z/z embryos (Ruest et al., 2004).

These changes in the BMP GRN could reflect a gain in BMP signaling in the maxillomandibular junction area and/or inappropriate patterning of ventral mandibular arch cells due to closer apposition of ventral cells with the maxillomandibular junction. The latter would occur if there was an absence or reduction in dorsal domain cells. To examine this question, we determined whether the expression of Hand1, one of the ventral-most markers in the mandibular arch (Clouthier et al., 2000), was altered. In E10.5 Gata3^+/+ and Gata3^z/z embryos, Hand1 expression was confined to the ventral arch mesenchyme (Fig. 8M,N). This suggests that the expansion of ventral and intermediate gene expression into the maxillomandibular junction is not simply due to proximity of the ventral cap to the maxillomandibular junction.

FOXC1 is a transcription factor that functions through a genetic interaction with FGF8 to position the hinge region, with loss of FOXC1 activity leading to syngnathia (Inman et al., 2013). We thus examined whether Foxc1 expression was affected by loss of GATA3. At E9.25, Foxc1 expression was similarly expressed in the first arch mesenchyme of Gata3^+/+ (Fig. 8O) and Gata3^z/z (Fig. 8P) embryos. By E10.5, Foxc1 expression in both Gata3^+/+ (Fig. 8Q) and Gata3^z/z (Fig. 8R) embryos was present in the caudal mesenchyme of the mandibular arch and the rostral mesenchyme of arch 2. As in E9.25 embryos, discernible differences were not apparent. These findings suggest that, although GATA3 and FOXC1 may impact a similar downstream GRN, Foxc1 is not transcriptionally downstream of Gata3.

The expansion of ventral gene expression towards the maxillomandibular junction illustrates that the absence of GATA3 results in a dorsal expansion of a ventral GRN. We therefore examined whether targeted inactivation of Gata3 also affected the transcriptional control of Gata3 itself. To accomplish this, we took advantage of the lacZ gene knocked-in to the Gata3 locus. At E8.5, β-galactosidase (β-gal) activity was observed in pharyngeal arches of both Gata3^+/z (Fig. 9A) and Gata3^z/z (Fig. 9B) embryos. Additional staining in Gata3^z/z embryos was observed extending from the neural tube to the arches/circumpharyngeal region (Fig. 9A,B, white arrows). At E9.5, β-gal staining was observed in the ventral domain of the pharyngeal arches. In pharyngeal arch one, staining was observed in the ventral half of the mandibular arch in both Gata3^+/z (Fig. 9C,D) and Gata3^z/z (Fig. 9E,F) embryos. Staining was also apparent in the otic placode and outflow tract in both genotypes (Fig. 9C-F).

At E10.5, β-gal staining in Gata3^+/z embryos was still confined to the ventral mandibular arch (Fig. 9G-K; purple dashed line in Fig. 9H,K), with staining present in the ectoderm and underlying mesenchyme of the arch (Fig. 9K). β-Gal staining was also present in the lambdoidal junction (λ) (Fig. 9H). In E10.5 Gata3^z/z embryos, however, β-gal activity in the ventral domain extended along the rostral half of the mandibular arch (Fig. 9I-L; purple dashed line in Fig. 9J,L) into the maxillomandibular junction (Fig. 9J,L), where it met the staining extending from the lambdoidal junction. Together, these findings indicate that loss of GATA3 leads to an upregulation of a ventral GRN (that includes Gata3 itself) in the maxillomandibular junction.

We demonstrate here that GATA3 is crucial for craniofacial development, with loss of GATA3 both reducing the total number of post-migratory NCCs in the early mandibular arch and disrupting the balance between BMP and FGF8 GRNs along the mandibular and maxillary portions of pharyngeal arch one. Together, our results illustrate that GATA3 is essential for separating the morphogenetic programs that define the upper and lower jaws and that cap signals can function in both positive and negative manners to ensure maxillomandibular (hinge) region development.

One of the first changes observed in Gata3^z/z embryos is hypoplasia of the mandibular arch at E9.5. Although cell proliferation and cell death are unchanged in E9.5 post-migratory NCCs, the total number of cells within the mandibular arch are significantly reduced. Msx1^−/−;Msx2^−/− embryos also exhibit craniofacial bone hypoplasia that is preceded by arch hypoplasia at E9.5 (Ishii et al., 2005). However, although the expression of Tfap2 is delayed in Msx1^−/−;Msx2^−/− embryos, expression of Tfap2 and Sox10 in Gata3^+/+ and Gata3^z/z embryos between E8.5 and E9.5 appears temporally similar, although staining always appears weaker in Gata3^z/z mutant embryos at this stage. GATA2/3 are key modulators of a transcriptional circuit that positions the neural plate border (NPB) and act directly downstream of early BMP signaling (Linker et al., 2009; reviewed by Simões-Costa and Bronner, 2015). It is thus possible that loss of GATA3 shifts the boundary of the NPB, thus allowing an expansion of the neural plate or placodal domain at the expense of NCCs, resulting in fewer NCCs populating the pharyngeal arches. Further experiments are required to test this possibility directly.

One of the key signals establishing the maxillomandibular junction is FGF8 (Fig. 10A). In chick embryos, Fgf8 in the early ectoderm defines the maxillomandibular junction, with Bmp4 expression in the ventral ectoderm preventing a more ventral expansion of Fgf8 expression (Shigetani et al., 2000). Targeted inactivation of Fgf8 using Nestin-Cre in mouse embryos leads to NCC-derived mesenchyme cell death and a dramatic reduction in the size of the mandibular arch, suggesting that the patterning program induced by ectodermal FGF8 signaling in the pharyngeal arch is either directly or indirectly required for mesenchyme cell survival (Trumpp et al., 1999). However, in Fgf8;Nes-Cre mutant embryos, early Bmp4 expression does not spread dorsally, indicating that in Gata3^z/z embryos (Trumpp et al., 1999) downregulation of Fgf8 expression in the maxillomandibular junction is not responsible for the upregulated Bmp4 expression. In fact, that Pitx2 expression is not dramatically changed in Gata3^z/z embryos suggests the opposite: changes in the Fgf8 GRN occur due to an expansion of BMP signaling. BMP4 bead implantation in mandibular arch explants represses Fgf8 expression, indicating that BMP4 may repress Fgf8 transcription more directly (Stottmann et al., 2001) as it does Barx1 expression (Mitsiadis et al., 2003). Similarly, targeted expression of Bmp4 in mouse NCCs also leads to downregulation of Fgf8 and an upregulation in the expression of BMP-responsive genes, including Hand2 and Msx1, resulting in bony syngnathia (He et al., 2014). Further, BMP4 overexpression induces a BIG (BMP-induced gene) profile that includes Hand2, Gata3, Msx1 and Hand1 (Bonilla-Claudio et al., 2012). In contrast, loss of Bmp4 within the arch ectoderm led to a reduction in the expression of these BIG genes (except for Hand1; discussed below). Thus, changes in Bmp4 expression appear to be a major driver of the gene expression changes in Gata3^z/z embryos (Fig. 10B). The next big challenge is identifying the signals induced by GATA3 activity in the caps that normally act in a non-cell-autonomous manner to allow patterning of the hinge region.

If loss of GATA3 results in the inappropriate upregulation of a BMP GRN within the maxillomandibular junction, what then is responsible for syngnathia and how does this fit into the hinge and caps model? Several proteins are involved in jaw separation, with their disruption causing syngnathia. One of these is the gene encoding FOXC1, a forkhead box transcription factor that is expressed in the early ectoderm and mesenchyme of the first arch (Inman et al., 2013). Disruption of Foxc1 leads to syngnathia, with variations in phenotype involving a genetic interaction between Fgf8 and Foxc1 (Inman et al., 2013). However, Gata3 expression is independent of FOXC1 (Inman et al., 2013). We have shown here that Foxc1 expression is independent of GATA3, suggesting that Gata3 and Foxc1 function independently to influence a common downstream GRN. The most likely candidate is FGF8, as Fgf8 expression in the maxillomandibular junction is reduced in both Gata3^z/z (here) and Foxc1^−/− (Inman et al., 2013) embryos. Further, both Gata3^z/z (here) and Foxc1^−/− (Inman et al., 2013) embryos have a leftward skew to the mandible, a loss of facial laterality that might be expected if Fgf8 expression was disrupted (Albertson and Yelick, 2005). Together, these results suggest that, in Gata3^z/z embryos, a BMP-driven downregulation of a Fgf8 GRN in the maxillomandibular junction derails boundary separation between the mandibular and maxillary prominences and, hence, syngnathia (Fig. 8).

Of the genes for which mandibular arch expression was examined, only Hand1 expression remained confined to the ventral domain. Although confinement of ventral cap gene expression is likely a requisite event in the hinge and caps model, multiple mechanisms may underlie this restriction. In the case of Hand1, ventral cap expression requires the direct overlap of BMP signaling and HAND2 transcriptional activity (Vincentz et al., 2016). However, whereas Hand2 expression in Gata3^z/z embryos spreads in a dorsorostral direction towards the maxillomandibular junction (Ruest et al., 2004), Hand1 expression does not. This illustrates that although BMP and HAND2 are both required for Hand1 expression (Barron et al., 2011; Vincentz et al., 2016), repressive mechanisms also exist that can override these inductive mechanisms. The Hand1 pharyngeal arch enhancer (Hand1^PA/OFT enhancer; Vincentz et al., 2016) contains several validated DLX-binding elements, and introducing DLX5 in reporter assays disrupts BMP/HAND2 synergy in vitro, thus preventing expression of Hand1^PA/OFT enhancer-driven transgenes (Vincentz et al., 2016). Given that Dlx5 expression in Gata3 mutant embryos expands in a pattern similar to Hand2 (Ruest et al., 2004), the repressive activity of DLX likely prevents dorsal Hand1 expansion. A targeted deletion approach utilizing CRISPR to mutagenize DLX DNA-binding consensus residues in the Hand1^PA/OFT enhancer is required to address this question.

We have shown here that loss of Gata3 leads to a disruption in lateral symmetry. Although both sides of the facial skeleton are dysmorphic, the developmental defects in bone and cartilage structures were always more severe on one side. These morphological changes resemble those in individuals with craniofacial microsomia [also known as hemifacial microsomia (HFM), Goldenhar syndrome or oculoauriculovertebral spectrum], which often includes unilateral defects in first and second pharyngeal arch-arch derived structures, including the mandible, temporomandibular joint, middle ear bone, ear pinna, maxilla, zygoma and muscles of mastication (Tiner and Quaroni, 1996; Barisic et al., 2014). The vast majority of these individuals do not have a genetic diagnosis. Although BAPX1 (NKX3-2) (Fischer et al., 2006), MYT1 (Lopez et al., 2016; Berenguer et al., 2017), TCOF1 and SALL1 (Huang et al., 2010) have been implicated in HFM individuals, none has been validated (Thiel et al., 2005; Fischer et al., 2006). Further, mouse mutants for Bapx1 (Tribioli and Lufkin, 1999; Akazawa et al., 2000; Tucker et al., 2004), Myt1 (Wang et al., 2007) and Tcof1 (Dixon et al., 2006) do not have unilateral craniofacial defects. One of the few mouse models reported as a model for HFM resulted from a transgenic insertional event on mouse chromosome 10 (Naora et al., 1994; Cousley et al., 2002). Although this locus was named the hemifacial microsomia-associated (Hfm) locus, the underlying genetic lesion has not been described. Goosecoid (Gsc) is located within the insertional region, although mutations in GSC were not found in two groups of HFM individuals (Kelberman et al., 2001). In addition, Gsc^−/− mouse embryos do not develop a Hfm phenotype (Yamada et al., 1995; Rivera-Perez et al., 1995, 1999). Our results here provide a new gene to explore in craniofacial microsomia. Indeed, genome-wide association studies have identified the locus containing GATA3 as a susceptibility locus for craniofacial microsomia (P=6.58×10⁻⁹) (Zhang et al., 2016).

Generation and genotyping of the Gata3^+/z line (which contain a lacZ gene) have been previously described (Lakshmanan et al., 1999). Mice were maintained on a FVB background (Taconic). Genotyping of embryos was conducted using yolk sac DNA and PCR using the primers 5′-TCCTGCGAGCCTGGCTGTCGGA-3′ and 5′-GTTGCCTTGACCATCGATGTT-3′ to detect the wild-type allele and 5′-GACACCAGACCAACTGGTA-3′ and 5′-GCATCGAGCTGGGTAATAAC-3′ to detect the lacZ allele. Generation of Gata3^z/z embryos for timed matings was conducted by breeding Gata3^+/z animals, with the day of vaginal plug counted as E0.5. Targeted deletion of Gata3 is embryonic lethal around E12.5 owing to disruption of sympathetic neuron development, leading to reduction of epinephrine/norepinephrine (Lim et al., 2000). To rescue Gata3^z/z embryo development, pregnant females were given water ad libitum containing the adrenergic receptor agonists L-phenylephrine (100 μg/ml; P6126, Sigma-Aldrich) and isoproterenol (100 μg/ml; I5627, Sigma-Aldrich), along with 2 mg/ml ascorbic acid (A0278, Sigma-Aldrich) beginning at E8.5. This approach has previously been used to rescue Gata3^z/z embryos (Lim et al., 2000) and Hand2^flneo/flneo and Hand2^fl/fl;Wnt1-Cre embryos (Hendershot et al., 2008) from pre-term death due to reduced or absent norepinephrine. Based on these studies, agonist treatment does not appear to affect the rate or extent of osteogenesis. Embryos were collected as described below; all embryos examined (Gata3^+/+ and Gata3^z/z) came from treated females. All studies were approved by the Institutional Animal Care and Use Committee at the University of Colorado Anschutz Medical Campus.

Analysis of both bone and cartilage development in E18.5 embryos using Alizarin Red and Alcian Blue (Sigma-Aldrich) (Ruest et al., 2004) and cartilage analysis in E14.5 embryos using Alcian Blue (Clouthier et al., 1998) was performed as previously described.

For histological analysis using H&E (Sigma-Aldrich), embryos were collected and fixed in 4% paraformaldehyde (Fisher Scientific) overnight before being processed through graded ethanol solutions and xylene and then embedded in paraffin as previously described (Tavares et al., 2017). Seven-micron-thick sections were cut on a Leica microtome, with staining and subsequent analysis performed as previously described (Tavares et al., 2017).

Embryos previously stained with Alizarin Red and Alcian Blue (see above) were also imaged using µCT. These embryos were placed individually in polypropylene cryovial tubes filled with PBS and scanned with a Skyscan 1275 Micro-CT (Bruker BioSpin Corporation) using the following parameters: 10 µm resolution, 40 kV, 200 µA, 45 ms exposure, 0.3° rotation step, 180° imaging, 4 frame averaging. Raw images from all scans were reconstructed using NRecon software (Bruker BioSpin Corporation). Reconstructed scan data were imported into Drishti volume exploration software (version 2.63) (Limaye, 2012) for 3D rendering. Rendering settings were optimized for visualization and phenotypic assessment of mineralized tissues. Selected images from the renderings were saved and optimized for contrast, color and background using Photoshop (Adobe).

Analysis of nerve development in E10.5 embryos was performed using a monoclonal anti-neurofilament 160 (NF160) antibody (N5264, Sigma-Aldrich) as previously described (Clouthier et al., 1998). The antibody was used at a dilution of 1:100 and diaminobenzidine (D7304-1SET, Sigma-Aldrich) was used as the substrate. Images were captured on a SZX12 (Olympus) stereomicroscope fitted with a DP21 camera (Olympus) and then converted to gray scale in Photoshop (Adobe).

Pregnant mice were injected intraperitoneally with 200 mg/kg body weight of 5-ethynyl-2′-deoxyuridine [contained in the Click iT EdU Imaging Kit (C10338, Thermo Fisher)] containing Alexa Fluor 594 as the dye 1 h before embryo collection at E9.5. Embryos were collected and fixed in 4% paraformaldehyde (Thermo Fisher Scientific) on ice for 1 h before rinsing in PBS, dehydrating, and embedding in paraffin. Somite counts were performed before embedding to ensure accurate embryo staging, with three control and three mutant embryos used for the analysis. Embryos were sectioned at 7 m; three or four sections separated by 35-40 m through the mandibular arch of each Gata3^+/+ or Gata3^z/z embryo were used for the assay. After rehydrating slides through graded ethanol solutions and rinsing in PBS, slides were subjected to antigen retrieval by placing slides in 200 ml of citrate buffer (pH 6.0) and microwaving for 5 min at 50% power. After rapid cooling with 80 ml of ddH₂O followed by rinsing in PBS, TUNEL analysis was performed using the Roche In Situ Cell Death Detection Kit, Fluorescein (11684795910, Sigma-Aldrich) following the manufacturer's recommendations. Slides were then rinsed in PBS and incorporated EdU was detected using the Click iT EdU Imaging Kit according to the manufacturer's recommendation. After staining, sections were counterstained using DAPI (Sigma-Aldrich) to mark all cell nuclei. After coverslipping, sections were examined and recorded using an Olympus BX50 compound microscope fitted with appropriate fluorescence cubes and a DP72 camera. Labeled EdU, TUNEL and DAPI cells were counted, with the TUNEL-positive cells/total cells and EdU-positive cell/total cells calculated. Statistical analyses and graph generation were performed in Prism 9 (GraphPad), with the final graph processed through Adobe Illustrator. The red EdU signal generated with Alexa Fluor 594 was replaced with magenta using Photoshop (Adobe).

Whole-mount analysis of cell death using TUNEL was performed as previously described (Abe et al., 2007).

Whole-mount gene expression analysis was performed as previously described (Clouthier et al., 1998) using digoxigenin-labeled RNA riboprobes against Tfap2a (Feng et al., 2008), Barx1 (Tissier-Seta et al., 1995), Bmp4 (Furuta and Hogan, 1998), Dlx2 and Dlx3 (Robinson and Mahon, 1994), Dlx6 (Charité et al., 2001), Fgf8 (Trumpp et al., 1999), Hand1 and Hand2 (Srivastava et al., 1997), Lhx8 (Tucker et al., 1999), Msx1 (Thomas et al., 1998), Pitx1 and Pitx2 (Liu et al., 2003). Four embryos of each genotype were examined for each probe. Bound probes were detected with either BM Purple (Sigma Aldrich) or 4-nitro blue tetrazolium chloride (NBT)/5-bromo-4-chloro-3-indolyl phosphate (BCIP) (Roche). Matching embryo pairs were processed together and developed for the same length of time.

To examine β-galactosidase (β-gal) staining in whole embryos, E8.5, E9.5 and E10.5 Gata3^z/z and Gata3^+/z embryos were collected and fixed for 1 h in 4% paraformaldehyde. Embryo staining and photography was performed as previously described (Ruest et al., 2003). E10.5 embryos were then processed through graded alcohols and embedded transversely in paraffin. Seven-micron-thick sections were cut, mounted on Superfrost Plus slides (Fisher Scientific), stained with Nuclear Fast Red, coverslipped and photographed (Ruest et al., 2003).

We would like to thank Katherine Kuhn and Drs Andre Tavares, Bruno Ruest, Francie Hyndman and Gwinn Vonnahme for technical assistance and scientific input, and Dr Katherine Fantauzzo for critical reading, suggestions and encouragement.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.199534