Hand1 phosphoregulation within the distal arch neural crest is essential for craniofacial morphogenesis Free

Neural crest cells (NCCs) are a multipotent cell population that specifies within the dorsal lip of the neural tube and subsequently delaminates, migrates and populates the pharyngeal arches (PAs), before ultimately differentiating into a wide spectrum of structures/tissues along the anterior-posterior (AP) axis of vertebrate embryos (Clouthier et al., 2010; Minoux and Rijli, 2010; Ruest and Clouthier, 2009; Trainor, 2005; Gitton et al., 2010). The NCCs that migrate into the first and second PAs are primarily responsible for orchestrating craniofacial development. Well-integrated signaling programs function through transcription factors to define tissue patterning and NCC differentiation. The complex signals that govern craniofacial morphogenesis involve a number of input pathways, including Fgf, Shh, Wnt, Bmp, Pdgf, retinoic acid (RA) and endothelin signaling (Abe et al., 2008; Abzhanov and Tabin, 2004; Clouthier et al., 2003; Jiang et al., 2006; Kurihara et al., 1995; Macatee et al., 2003). Dysregulation of NCC migration, proliferation and patterning can result in craniofacial abnormalities observed in numerous human syndromes (Chai and Maxson, 2006; Jiang et al., 2006; Noden and Trainor, 2005; Clouthier et al., 2013). Indeed, cleft lips and cleft palates are among the most common congenital defects observed in newborns. Understanding the signaling pathways and downstream transcription factors that govern craniofacial morphogenesis is crucial in defining the etiology of these common congenital defects.

Members of the basic helix-loop-helix (bHLH) transcription factor superfamily play dominant roles in cell specification, differentiation and tissue patterning throughout embryonic development (Massari and Murre, 2000). bHLH transcription factors affect transcription by forming either a bHLH homo- or heterodimer and by binding DNA via a cis-element termed E-box. Loss-of-function studies in the Twist sub-class of bHLH factors, which include Twist1, Twist2, Hand1 and Hand2 (Barnes and Firulli, 2009), reveal that these factors play key roles in craniofacial morphogenesis. Targeted disruption of Twist1 in NCCs leads to defects in upper and lower jaw development (Bildsoe et al., 2009). NCC-specific deletion of Hand2 results in disruption of lower jaw and tongue development (Barron et al., 2011). These factors appear highly conserved during vertebrate evolution (Barnes and Firulli, 2009) and probably play similar roles in human facial development. Indeed, human mutations in TWIST1 cause Saethre–Chotzen syndrome (SCS). In addition to craniosynostosis, palatal anomalies are commonly observed in SCS patients (Stoler et al., 2009).

Compelling evidence shows that Twist family bHLH factors mediate biological function by dimer choice (Castanon et al., 2001; Firulli et al., 2003, 2005, 2007). Dimer choice is regulated, in part, by a threonine and serine pair that is evolutionarily conserved among all Twist family members. Mimicking Hand1 hypophosphorylation through mutations in residues T107 and S109 enhances homodimer formation, whereas mimicking Hand1 phosphorylation at T107 and S109 enhances formation of E-protein heterodimers (Firulli et al., 2003). Indeed, changes in bHLH dimer choices affect craniofacial development (Connerney et al., 2006). Dysregulation of Twist1 phosphorylation at these threonine and serine residues causes SCS (Cai and Jabs, 2005; Firulli et al., 2005). Dimerization of Twist-family bHLH factors is governed by their co-expression within a cell, their colocalization within the nucleus, their level of relative expression and the phosphorylation of the conserved threonine and serine residues within Helix I of the bHLH domain (Firulli and Conway, 2008; Firulli et al., 2003, 2005, 2007). Overexpression studies alter the stoichiometry of this bHLH pool. Thus, to validate the consequences of dysregulating this post-translational control mechanism within the bHLH dimer pool in vivo, it is essential to manipulate phosphorylation but not expression levels.

By contrast to the dominant roles exhibited by Twist1 and Hand2 in craniofacial development, the related factor Hand1, although expressed within the distal most NCC-derived PA mesenchyme (Clouthier et al., 2000), shows no observable phenotypes in NCC loss-of-function analysis. However, there is a clear gene dosage effect when Hand1 NCC conditional-null mice are placed on a Hand2 heterozygous background (Barbosa et al., 2007). This suggests a crucial Hand gene dosage that, when disrupted, perturbs the bHLH dimer pool within cranial NCCs, resulting in morphogenic defects. To test whether Hand1 phosphorylation-dependent dimer regulation alters the composition of the bHLH dimer pool, which influences craniofacial morphogenesis, we engineered two Hand1 conditional-activation knock-in alleles designed to either mimic Hand1 hypophosphorylation by replacing both threonine 107 and serine 109 with alanines (PO₄−), or to mimic Hand1 hyperphosphorylation by replacing these residues with aspartic acids (PO₄+). Activation of these phospho-mutant alleles within NCCs results in severe mid-facial clefting. Mutant embryos display abnormal cell death within the developing PAs, resulting in reduced outgrowth of the maxillary processes and aberrant craniofacial development. The majority of structural defects occur within tissues where Hand1 is not expressed and where Hand1-lineage cells are not detected. Gene expression analyses show that both fibroblast growth factor 8 (Fgf8) and sonic hedgehog (Shh) signaling pathways are affected. Interestingly, removing the wild-type Hand1 allele on both the Hand1^PO4− and Hand1^PO4+ backgrounds results in reduced levels of cell death and improved craniofacial development. Together, these results show that, although Hand1 is dispensable for development of craniofacial structures, a misregulation of Hand1 dimer choice within the NCC-derived distal PA ectomesenchyme disrupts the function of other factors required for craniofacial morphogenesis. These findings suggest that Hand1 and its putative bHLH dimer partners transcriptionally control one or more of the theorized cap signals central to the ‘Hinge and Caps’ model of PA patterning (Depew and Compagnucci, 2008; Fish et al., 2011).

To understand the role of Hand1 phosphoregulation within NCCs, we engineered two targeted conditionally active Hand1 alleles, wherein residues T107 and S109 were replaced with either alanines (PO₄−) or aspartic acids (PO₄+; Fig. 1A). By incorporating a Stop-flox cassette containing a neomycin resistance gene downstream of the Hand1 transcriptional start site, but upstream of the Hand1-initiating methionine, we were able to create null, conditional phospho-mutant Hand1 alleles. In tissues where the Cre recombinase and endogenous Hand1 expression directly overlap, the Stop-flox cassette is removed, allowing expression of the Hand1^PO4− or Hand1^PO4+ mutant alleles. Hand1 expression within NCC-derived mesenchyme of the distal PAs is first detectable at E9.5. Expression cannot be detected at E9.0 using the sensitive Hand1^lacZ knock-in allele (Firulli et al., 1998; Fig. 1B-E). Hand1 expression initiates in the most posterior/caudal portions of the distal PA before expanding anteriorly/rostrally, but not laterally. At E13.5, Hand1-expressing cells mark the medial tongue (t) and fusion point of the two mandibular halves (md; Fig. 1F,G). Although Hand1-lineage analysis reveals minor lateral spreading of these distal cells, Hand1 expression is lost as the cells move laterally from the midline. Craniofacial structures of the calvarium do not express Hand1 or contain Hand1 lineage cells (Barnes et al., 2010).

We tested the efficacy of our conditional Hand1 phospho-mutant alleles by activating mutant expression in NCCs using the Wnt1-Cre allele (Danielian et al., 1998). Whole-mount in situ hybridization (ISH) of Hand1 detects distal expression in the forming PAs of control embryos (Fig. 1H-J). To demonstrate that the conditional alleles show the expected Hand1 spatio-temporal pattern, we crossed both the Hand1^PO4− and Hand1^PO4+ mutant alleles onto the Hand1 conditional (Hand1^fx) allele (Barbosa et al., 2007; McFadden et al., 2005), and then recombined both conditional alleles in NCCs using the Wnt1-Cre allele. Using Hand1^PO4−/fx;Wnt1-Cre(+) and Hand1^PO4+/fx;Wnt1-Cre(+) embryos, we demonstrate that mRNA expression from either of the Hand1 phospho-mutant alleles is indistinguishable from wild-type Hand1 or the Hand1^lacZ expression by Hand1 ISH (Fig. 1I,J). However, Wnt1-Cre-mediated expression of both Hand1 phospho-mutant alleles results in neonatal death accompanied by 100% penetrant mid-face clefts (Fig. 1L-N).

To determine when we could first detect a facial phenotype, we collected mutant embryos from timed pregnancies between E10.5 and E19.5. Hand1^+/PO4−;Wnt1-Cre(+), Hand1^+/PO4+;Wnt1-Cre(+), Hand1^PO4−/fx;Wnt1-Cre(+) and Hand1^PO4+/fx;Wnt1-Cre(+) phenotypes are first morphologically identifiable at E10.5, wherein distances between the lateral nasal prominences (lnp) and the olfactory pits (op, black line) are extended and maxillary processes (mp) are symmetrically reduced in size (supplementary material Fig. S1). As development proceeds, the extent of this phenotype becomes more evident. Surprisingly, removal of the wild-type Hand1 allele reduces the severity of both hypophosphorylated and phosphorylation-mimicking phenotypes. Histological analysis at E14.5 shows that structures such as tooth primordia (tp), Meckel's cartilage (mc) and nasal cavities (nc) form normally in the Hand1 phospho-mutant embryos (supplementary material Fig. S2). The nasal capsule is also present but is deviated down the midline with the rest of the facial structures. By contrast, Hand1 phospho-mutant heterozygotes [Hand1^+/PO4−;Wnt1-Cre(+) and Hand1^+/PO4+;Wnt1-Cre(+)] lack a patent nasal septum (does not fuse at the midline; ns; black asterisk; supplementary material Fig. S2G,I). Furthermore, the palatal shelves (ps) are not fused, resulting in aberrant communication between the nasopharynx and oral cavity (supplementary material Fig. S2B,D). Single-copy point mutants [Hand1^PO4−/fx;Wnt1-Cre(+) and Hand1^PO4+/fx;Wnt1-Cre(+)] show improvement in nasal septum development (supplementary material Fig. S2H,J), and in the case of Hand1^PO4+/fx;Wnt1-Cre(+) embryos, palatal shelf fusion is observed in mixed penetrance, although clefting is still fully penetrant (supplementary material Fig. S2E,J).

We next employed micro-computed tomography (micro-CT) scans of heads from P0 Hand1^+/PO4−;Wnt1-Cre(+), Hand1^+/PO4+;Wnt1-Cre(+), Hand1^PO4−/fx;Wnt1-Cre(+) and Hand1^PO4+/fx;Wnt1-Cre(+) embryos. In Hand1^PO4−/+ mutant embryos, structural defects were observed throughout the skull. On lateral view, the premaxilla (pm, dark purple) is shortened, and both the overlying nasal bone (n, green) and a portion of the maxilla (mx, blue) closest to the frontal bone (f, red) are missing (Fig. 1L). The jugal bone (j, dark red), the middle bone of the zygomatic arch (with the others being the zygomatic processes of the maxilla and squamosal bones), is hypoplastic, and the squamosal bone (sq, light purple) is absent (Fig. 1K). The proximal mandible (md, bronze) is also hypoplastic. In dorsal view (Fig. 1L), the interparietal (i, turquoise) and frontal bones are hypoplastic, the latter leading to a large gap between the frontal bones. The upper incisors (magenta, white arrow) are readily visible due to this hypoplasia and the absence of the nasal bone. Both sagittal sutures are also aberrantly fused (asterisk). The hypoplasia of the mandible is more obvious from the dorsal view, with the two halves failing to meet at the midline. A ventral view (Fig. 1M) reveals more significant changes to the skull base. Whereas the midline defects around the premaxilla are obvious, most bones appear hypoplastic. The basisphenoid (bs, uncolored) and pterygoid bones (pt, uncolored) are severely underdeveloped, with the missing squamosal bone more obvious. Also missing is the lamina obturans (the bony portion of the future alisphenoid that abuts the squamosal bone). In addition, the midline cleft defect resulted in the failure of the palatine bones and the palatal processes of the maxilla to fuse (pl, yellow). To examine cartilage derivatives, we stained E17.0 control and Hand1^+/PO4+;Wnt1-Cre(+) embryos with alizarin red and alcian blue to visualize bone and cartilage, respectively. Compared with control embryos (Fig. 1N), the malleus and incus (mandibular arch) as well as the stapes (second PA) appear normal in Hand1^+/PO4+;Wnt1-Cre(+) embryos (Fig. 1O). However, Meckel's cartilage (mandibular arch) is truncated at its origin with the malleus, and the cartilage anlage of the hyoid bone (second arch) is poorly ossified and deformed at the midline.

Defects in Hand1^PO4−/fx;Wnt1-Cre(+) embryos, in which only the mutant Hand1 allele is present, are less severe. In lateral view (Fig. 1K), the squamosal, jugal and tympanic ringbones appear normal, as does the premaxilla and the mandible. The nasal bone is present but hypoplastic. In dorsal view (Fig. 1L), the midline defects in the front bones are apparent, as hypoplasia of the nasal bone allows visualization of the upper incisors. The sagittal sutures appear normal. In ventral view (Fig. 1M), there is modest improvement in the pterygoid bones, although the basisphenoid is still hypoplastic. The size of the lamina obturans and squamosal bones are also improved compared with Hand1^+/PO4− embryos. The trabecular basal plate is also better developed, although there is still a significant cleft between the palatal processes of the palatine and maxilla bones and between the two halves of the premaxilla.

In Hand1^+/PO4+;Wnt1-Cre(+) embryos, a lateral view shows that the premaxilla, nasal, and the squamosal bones are better formed, but are still slightly hypoplastic. The proximal mandible also appears slightly smaller. In dorsal view, the distal mandible appears fused at the midline and the sagittal sutures appear normal. By contrast, the midline cleft in the maxilla appears more severe than observed in Hand1^+/PO4− mutants, with the defect extending into the parietal bones. The gap between the upper incisors is noticeably enlarged. Ventral changes in Hand1^+/PO4+ mutants are nearly identical to those observed in Hand1^+/PO4− mutants. One additional difference is the absence of at least a portion of the ala temporalis portion of the alisphenoid.

In lateral view, Hand1^PO4+/fx;Wnt1-Cre(+) mutants appear nearly identical to Hand1^+/PO+;Wnt1-Cre(+) mutants, although the proximal mandible is slightly hypoplastic, as is the squamosal bone. On dorsal view, phenotypes are similar to Hand1^+/PO4+ mutants. On ventral view, the middle-ear ossicles are either hypoplastic or absent, although portions of the alisphenoid appear better developed. Interestingly, the palatal processes of the palatine and maxilla appear to fuse normally along the midline, although both structures are smaller than those in control embryos. The ectopic bone observed in Hand1^PO4+/fx;Wnt1-Cre(+) mutants is not present in Hand1^+/PO+;Wnt1-Cre(+) embryos. A detailed measure of bone sizes is shown in supplementary material Table S1.

We postulated that the midline defects in facial structures could be the result of a defect in cell migration, the inability of the structures to fuse, decreased cell proliferation or increased cell death within the cranial NCC. We looked at NCC migration by utilizing ROSA26^lacZ reporter lineage-tracing and observed no obvious migration defects for the NCCs entering the PAs or cardiac outflow tract (supplementary material Fig. S3). Defects in tissue fusion seem unlikely, as fusion is observed in the Hand1^PO+/fx;Wnt1-Cre(+) mutants (Fig. 1; supplementary material Fig. S2), and the assessment of the fusion markers Jagged2 and Mmp13 reveal no significant variations between E14.5 control and Hand1^+/PO4−;Wnt1-Cre(+) or Hand1^+/PO4+;Wnt1-Cre(+) embryos (supplementary material Fig. S4). 5-Ethynyl-2′-deoxyuridine (EdU) incorporation analysis reveals no significant difference in NCC proliferation between control and all Hand1 phospho-mutant embryos at E9.5 and E10.5 (data not shown). By contrast, significantly elevated levels of cell death are observed within the PAs of Hand1 phospho-mutant embryos compared with control littermates (Fig. 2). At E9.5, whole-mount LysoTracker staining of control embryos reveals two dorsally localized domains of normal developmental cell death (black arrows) with low levels of cell death detected within the PA mesenchyme (Fig. 2A, white arrow). This is confirmed by section terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) analysis (Fig. 2C, yellow arrows indicating developmental cell death). E9.5 Hand1^+/PO4−;Wnt1-Cre(+) mutant embryos exhibit a marked reduction in the developmental dorso-lateral cell death domains while displaying a marked increase in PA cell death as assayed by both LysoTracker and TUNEL staining (Fig. 2E,G). Correlating with the less severe clefting phenotype, Hand1^PO4−/fx;Wnt1-Cre(+) mutants display levels of normal dorso-lateral developmental cell death similar to controls, while at the same time exhibiting decreased cell death within the PA mesenchyme (Fig. 2I,K). Similar findings are observed in Hand1^PO+/fx;Wnt1-Cre(+) embryos (Fig. 2M,O). At E10.5, the extent of cell death within the PA of the phospho-mutants is reduced when compared with control embryos (Fig. 2B,D,F,H,J,L,N,P,R,T). Given that the Hand1 expression domain (Fig. 1) does not directly overlap with the observed domain of PA cell death, we reasoned that a cell signaling pathway is disrupted in Hand1 phospho-mutant embryos and that the observed craniofacial phenotypes are generated non-cell autonomously.

During craniofacial formation, numerous signaling pathways intersect to govern normal morphological patterning and include Fgf, Shh, Wnt, Bmp and RA signaling (Clouthier et al., 2010; Trainor, 2005). We first looked at altered RA signaling as Rdh10 mutant mice exhibit similar mid-facial clefting (Sandell et al., 2007). By intercrossing Hand1 phospho-mutant mice with the RARE-lacZ reporter line (Rossant et al., 1991), we found that RARE-lacZ patterning was largely unaffected in both Hand1^+/PO4−;RARE-lacZ(+);Wnt1-Cre(+) and Hand1^+/PO4+;RARE-lacZ(+);Wnt1-Cre(+) embryos at E10.5 and E11.5 (supplementary material Fig. S5). Likewise, expression of both Bmp2 and Bmp4 shows no appreciable difference between mutant and control embryos (data not shown).

We next examined Fgf8 expression in control and Hand1 mutant embryos between E9.5 and E12.5 (Fig. 3). Mouse models that exhibit variation in Fgf8 gene dosage and Fgf8 gain-of-function studies in chicks both present with mid-face clefts that are accompanied by early cell death (Abzhanov and Tabin, 2004; Chen et al., 2012; Griffin et al., 2013; Moon and Capecchi, 2000). Fgf8 expression between E9.5 and 12.5 shows a marked expansion in Hand1 phospho-mutants when compared with control embryos. Fgf8 expression within the rostral portion of the first arch, the maxillary processes and surrounding the nasal pits is expanded and becomes more pronounced over time (Fig. 3: compare D with H, L, P and T). Levels of Fgf8 expression appear more persistent in E12.5 mutant embryos where olfactory pit expression is more robust and expression at the edge of the maxilla is still visible (Fig. 3D,H,L,P,T, white arrows).

To validate this observation, we next looked at the expression of sprouty homolog 1 (Spry1) and ets variant 5 (Etv5), both downstream mediators of Fgf8/FgfR1 signaling (Firnberg and Neubüser, 2002; Lunn et al., 2007; Minowada et al., 1999; Raible and Brand, 2001), in Hand1^+/PO4−;Wnt1-Cre(+) and Hand1^+/PO4+;Wnt1-Cre(+) embryos at E10.5 (Fig. 4). Similar to Fgf8 expression, Spry1 expression is expanded around the first PA and surrounding the olfactory pits in both Hand1^+/PO4−;Wnt1-Cre(+) and Hand1^+/PO4+;Wnt1-Cre(+) mutants when compared with control littermates (Fig. 4B,C, asterisks). Expression of Etv5 is also expanded. In control embryos, Etv5 mRNA is observed to be more robust within the rostral portion of the first PA (Fig. 4D, black arrow). In comparison, Etv5 expression within the first PA of both the Hand1^+/PO4− and Hand1^+/PO4+ phospho-mutant embryos shows a clear caudal expansion (Fig. 4E,F, white arrows). The expression of the FgfR1 shows no significant differences in expression patterns (Fig. 4G-I). We next performed qRT-PCR on cDNA reverse transcribed from RNA isolated from E10.5 control and Hand1 point mutant heads (Fig. 4J). Results confirm qualitative observations, showing that Fgf8, Spry1 and Etv5 are significantly upregulated (*P≤0.05; **P≤0.01 n≥4) in both Hand1^+/PO4− and Hand1^+/PO4+ phospho-mutant embryos when compared with controls. Interestingly, both Etv5 and FgfR1 expression are elevated in Hand1^+/PO4− mutants but not in Hand1^+/PO4+ mutant embryos, and we observe no significant difference in Etv4 expression (Fig. 4J).

In light of the expanded Fgf8 expression, we next examined Shh signaling in Hand1 phospho-mutant embryos. Shh is known to act synergistically with Fgf8 to drive cartilage outgrowth and cranial formation (Abzhanov and Tabin, 2004). Moreover, expanded Shh signaling is associated with a widening of the head (Abzhanov et al., 2007; Brugmann et al., 2006; Cobourne et al., 2009; Hu et al., 2003; Huang et al., 2007; Jeong et al., 2004; Jiang et al., 2006; Lan and Jiang, 2009). Hand1 phospho-mutants exhibit a similar head expansion between the nasal pits (supplementary material Fig. S1). At E9.5, expanded expression of Shh is observed within Hand1 phospho-mutant embryos when compared with controls (Fig. 5A-E, white asterisks). E11.5 control embryos (Fig. 5F) show undetectable Shh expression within the medial head ectoderm of the sinus cavity region, although robust expression is observed in the tooth primordia. By contrast, Shh expression is persistent in the sinus cavity of Hand1 phospho-mutant embryos (Fig. 5G-K, white arrows) in addition to the tooth primordia (tp). Persistent Shh medial head ectoderm expression is observed up to E12.5 (Fig. 5K-O), with removal of the Hand1 wild-type allele reducing the observed persistent expression. Additionally, strong Shh E12.5 expression is observed within the forming tooth primordia, rugae (r) and mystacial vibrissae (mv).

We next looked at the expression of patched 1 (Ptch1) in E9.5 embryos (Fig. 6A-C). When compared with control embryos, Ptch1 expression appears unchanged in the forming head mesenchyme and first arch (black asterisk) of both Hand1^+/PO4−;Wnt1-Cre(+) and Hand1^+/PO4+;Wnt1-Cre(+) embryos. Enhanced PA expression (black asterisk) of the Shh-regulated transcription factor Gli1 is observed in Hand1 heterozygous phospho-mutants; however, no significant changes are observed in the expression of Gli3 (Fig. 6D-I). To confirm these observations, we performed qRT-PCR on Hand1^+/PO4−;Wnt1-Cre(+) and Hand1^+/PO4+;Wnt1-Cre(+) phospho-mutant head cDNA (Fig. 6J). Results confirm that both Shh and Gli1 expression are significantly enhanced in both Hand1 phospho-mutants; however, Ptch1 and Gli3 expression levels are not significantly affected. Collectively, these data demonstrate non-cell-autonomous changes in both Fgf8 and Shh signaling pathways in Hand1 phospho-mutant embryos.

As Twist1 is a known integrator of Fgf and Shh signaling (Hornik et al., 2004), we next investigated the expression of Twist1 in control and Hand1 phospho-mutant embryos (Fig. 7). Twist1 loss-of-function mutations are associated with craniofacial defects during embryogenesis (Chen and Behringer, 1995) and the disease SCS (Barnes and Firulli, 2009; Firulli et al., 2005; Jabs, 2004). At E10.5, Twist1 is robustly expressed within the NCC of the medial and lateral nasal processes (lnp), the maxillary process (mp), as well as the first (I) and second (II) PAs (Fig. 7A). In E10.5 Hand1 phospho-mutant embryos, the Twist1-expressing tissues are visibly reduced in size, including the maxillary process (Fig. 7B-E, white arrows), the medial and lateral nasal processes (Fig. 7B-E) and the visibly smaller first and second PAs (Fig. 7B-E, asterisks). The frontal view of E11.5 control embryos shows the correct juxtaposition of the medial and lateral nasal processes, whereas the Hand1 phospho-mutant embryos show an obvious reduction of Twist1-expressing mesenchyme (black arrows) within both the lateral and medial nasal processes of Hand1 phospho-mutant embryos (Fig. 7F-J) and in the sinus cavity (Fig. 7P-T, asterisks), suggesting a loss of tissue. As expression levels of Twist1 mRNA do not appear uniformly reduced, we employed qRT-PCR analysis to quantitate the Twist1 levels more accurately (Fig. 7U). Results show that indeed the levels of Twist1 are significantly increased in both E10.5 Hand1^+/PO4− and Hand1^+/PO4+. Thus, it appears that a loss of tissue is responsible for the altered Twist1 expression pattern and that Twist1 expression in both the remaining NCC- and non-NCC-derived mesenchyme is increased in Hand1 phospho-mutants.

Given the upregulation of Fgf8, Shh and Twist1, we reasoned that we could improve the mid-face clefting phenotype by reducing the gene dosage of Twist1. To test this, we crossed the Hand1^+/PO4+;Wnt1-Cre(+) alleles onto the Twist1^fx allele and looked for improvement of mid-face clefting at E11.5 (Fig. 8A-F). Results show no improvement to mid-face clefting in Hand1^+/PO4+; Twist1^fx/+;Wnt1-Cre(+) embryos, and Hand1^+/PO4+;Twist1^fx/−;Wnt1-Cre(+) superimposes Twist1-mediated exencephaly (Chen and Behringer, 1995; Soo et al., 2002) on the Hand1 phospho-mutant clefting phenotype (Fig. 8E,F). We also reasoned that Hand2 expression levels could influence the Hand1 phospho-mutant phenotypes. We therefore crossed Hand1^fx/PO4−;Wnt1-Cre(+) mice onto the Hand2^+/neo background; however, lowering Hand2 gene dosage had no visible effects on Hand1 phospho-mutant craniofacial phenotypes (Fig. 8G-I).

We next reasoned that inhibition of Shh signaling could potentially improve Hand1 phospho-mutant mid-face clefting, so we injected pregnant dams with cyclopamine (Kasberg et al., 2013; Lipinski et al., 2008) to antagonize Shh pathway activation and harvested embryos at E13.5 and E14.5 (Fig. 8J-U). Partial phenotype rescue is observed in the most severe Hand1^+/PO4−;Wnt1-Cre(+) mutant phenotype (three out of 15 mutant embryos isolated), as compared with the 100% penetrance of pronounced mid-face clefting observed in Hand1^+/PO4−;Wnt1-Cre(+) mice (Fig. 8I,L,M,R,S). The least severe Hand1^PO4+/fx;Wnt1-Cre(+) mutants show more efficient rescue (seven out of nine mutants; Fig. 8N,O,T,U). It is established that abnormally low and high levels of Shh can also result in facial abnormalities (Lipinski et al., 2010), and the improved phenotype observed further supports the notion that the increased Shh level observed in Hand1 phospho-mutants contributes directly to the mid-face cleft phenotype.

The bHLH transcription factor Hand1 is clearly dispensable for craniofacial formation (Barbosa et al., 2007). However, altering the phosphoregulation of Hand1, which influences its bHLH dimerization, within the distal-most pharyngeal arch mesoderm results in a severe non-cell-autonomous increase in PA NCC death that ultimately manifests in a mid-face cleft phenotype. In addition to increased cell death, essential signaling pathways required for proper craniofacial morphogenesis (Fgf8 and Shh) show expanded expression in Hand1 phospho-mutant embryos. Treatment in utero with the Shh antagonist cyclopamine can improve the severity of the resultant mid-face cleft (Fig. 8), thus supporting the hypothesis that the increases observed in these signaling pathways drive the resultant phenotype.

Hand1 expression is observed only within the most distal arch mesenchyme and becomes detectable between E9.0 and E9.5 (Fig. 1; Firulli et al., 1998; Ruest et al., 2004). As the Hand1^PO4− and Hand1^PO4+ mutants are knock-in alleles, these Hand1 dimer mutants are expressed at endogenous levels in a spatio-temporal pattern consistent with wild-type Hand1. At E9.5, Hand1 phospho-mutants display robust pathological cell death in the PAs, indicating that the deleterious effects from the expression of the mutant allele are nearly immediate within the arch mesenchyme. Indeed, phenotypic changes are present by E10.5 (supplementary material Fig. S1). Histological analysis indicates that right and left structures are essentially normal, although, due to the loss of tissue from the early cell death, the outgrowth of the right and left sides of the forming face fail to fuse (supplementary material Fig. S2). This conclusion is supported by the normal expression of the fusion markers jagged 2 (Jag2) and matrix metallopeptidase 13 (Mmp13) (supplementary material Fig. S4) and the direct initiation of fusion observed in Hand1^PO4+/fx;Wnt1-Cre(+) embryos (Fig. 1N; supplementary material Fig. S2), which exhibit the least amount of early cell death and a less severe phenotype. Additionally, by utilizing the P0-Cre mouse strain to activate Hand1 mutant allele expression within the NCC during their migration [a full day later than Wnt1-Cre allelic activation (Yamauchi et al., 1999)], the severity and penetrance of the mid-face cleft is greatly reduced (supplementary material Fig. S6). This reinforces the idea that the deleterious effects of Hand1 phospho-mutant proteins are immediate to their expression and that even a subtle delay in expression largely limits the observed phenotype. Combined, these data demonstrate that Hand1 phospho-mutant facial structures are capable of fusion and, considering the lack of significant changes in cell proliferation, suggest that a non-cell-autonomous increase in arch mesenchyme cell death causes these phenotypes. Given that reduced cell death is observed in the distal most arch mesenchyme where the Hand1 dimer mutants are expressed (Fig. 2), it is probable that Hand1 phospho-mutants interfere with other bHLH dimer choices within this limited Hand1 expression domain. Consequently, this causes the non-cell-autonomous cell death, leading directly to the changes in Fgf8 and Shh signaling pathway expression and ultimately resulting in the defects in the upper bones of the skull. Outside the Hand1 expression domain, we observe a loss of Twist1-expressing tissue but an increase in Twist1 expression (Fig. 7), which could lead to increased formation of Twist1 homodimers. Indeed, increased Twist1 homodimer expression is directly associated with premature ossification and suture fusion accompanied by enhanced Fgf signaling (Connerney et al., 2006, 2008). Conversely, later-stage deletion of Twist1 using Tyr-Cre also yields a mid-faced cleft (Bildsoe et al., 2009), reinforcing the notion that altered gene dosage in conjunction with dimer choice can dramatically effect development. In genetic experiments, lowering the gene dosage of Twist1 on the Hand1^+/PO4− background does not improve the mid-face phenotype, and complete NCC conditional knockout of Twist1 on the Hand1^+/PO4− via Wnt1-Cre results in a more severe phenotype than either mutant separately (Fig. 8).

Another possible interpretation of these findings is that a bHLH-dependent distal arch-signaling center could integrate Fgf8 and Shh signaling pathways. The incongruity of the tightly restricted domain of Hand1 within the distal midline of the first mandibular PA and the alterations in gene expression observed in Hand1 phosphoregulation mutants is consistent with the ‘Hinge and Caps’ model proposed by Depew (Compagnucci et al., 2013; Depew and Compagnucci, 2008; Depew and Simpson, 2006; Fish et al., 2011). In this model, reciprocal signaling between the putative ‘Hinge’, or the oral ectoderm of the proximal first PA, and two ‘Caps’ [corresponding to (1) the medial frontonasal processes and (2) the distal mandibular midline] pattern the nascent jaw along the proximo-distal, medio-lateral and rostro-caudal axes. This reciprocal signaling could explain how altered bHLH-dependent transcriptional mechanisms within the distal mandibular midline can alter gene expression profiles non-cell autonomously within the medial frontonasal processes and cause ectopic cell death within the proximal first arch. We can rule out the alternative possibility of cell migration from regions of Hand1 expression in the mandibular arch to the maxillary prominence, as Hand1^Cre lineage analysis shows that Hand1 daughter cells do not enter the maxillary prominence (Barnes et al., 2010).

Initially, we were surprised by the observation that both the hypophosphorylated and phosphorylation-mimicking Hand1 mutants exhibit a less severe phenotype when the wild-type Hand1 allele is deleted. This observation highlights the mechanics of bHLH factors and their prerequisite homodimer or heterodimer formation conveying transcriptional activity on target genes. It is clear that complete deletion of Hand1 does not significantly affect the bHLH dimer pool within the distal mesenchyme, as such Hand1 loss-of-function mutants are phenotypically normal, viable and fertile (Barbosa et al., 2007). The absence of Hand1-mediated defects might be the result of functional compensation by the related bHLH factor Hand2 within the cranial NCC. This is an established mechanism that functions in both sympathetic neurons and the developing heart (Barbosa et al., 2007; Hendershot et al., 2008; Howard, 2005; McFadden et al., 2005; Vincentz et al., 2012). By contrast, altering Hand1 phosphoregulation alters the ability of Hand1 to interact with potential bHLH dimer partners (Firulli et al., 2003). By interfering with the function of other crucial bHLH factors via direct dimerization and/or by titration of available E-protein levels, the result is a dramatic increase in PA cell death. The finding that wild-type Hand1, in the presence of the Hand1 phospho-mutant alleles, contributes to these deleterious transcriptional effects when the bHLH dimer pool is altered further confirms the role of Hand gene dosage phenotypes that are well established in a number of developmental systems, including the heart and limbs (Barbosa et al., 2007; Firulli et al., 2010, 2005; McFadden et al., 2005). The complexity of regulating the bHLH pool lies in determining all combinations of Hand bHLH dimer complexes within the extremely narrow temporal window when this dysfunction is occurring. Additionally, we have to consider that altering the Twist-family dimer regulation also affects DNA-binding affinities (Firulli et al., 2007). Therefore, it is also possible that Hand1 mutant proteins could target promoters inappropriately and/or facilitate other bHLH factors to do likewise, thus further effecting gene expression.

Hand1^{stopfloxHand1T107;S109A} (Hand1^PO4−) and Hand1^{stopfloxHand1T107;S109D} (Hand1^PO4+) mice were generated from embryonic stem cells targeted with the constructs described in Fig. 1. Genotyping was performed by Southern blot as described previously (Firulli et al., 1998), or with the allele-specific primers H1 (5′-CTGCCATTGGCTCCGGCTAGAGGT-3′) and PGK (5′-GGCTGCTAAAGCGCATGCTCCAGACTG-3′), using PCR conditions of 94°C for 1 min, 60°C for 1 min and 72°C for 1 min for 35 cycles. B6.129S4-Gt(ROSA)26Sortm1Sor/J (ROSA26R-β-gal homozygous; R26R^lacZ) mice were genotyped using a probe located 5′ of the Stop-Flox (provided by Dr Phillippe Soriano, Mount Sinai Hospital). Both Hand1^+/PO4− and Hand1^+/PO4+ alleles were bred onto a R26R^lacZ homozygous background, and females of this genotype were crossed to Wnt1-Cre [Tg(Wnt1-cre)2Sor] (Danielian et al., 1998) or Wnt1-Cre(+);Hand1^fx/fx males to generate either Hand1^+/PO4 or Hand1^fx/PO4 embryos. RARE-lacZ mice [Tg(RARE-Hspa1b/lacZ)12Jrt/J] were obtained from the Jackson Laboratories. P0-Cre mice (Yamauchi et al., 1999) were obtained from Simon J. Conway (Indiana University, USA). Cyclopamine (Sigma) was resuspended at 1 mg/ml in 45% w/v solution in water 2-hydroxpropyl-beta-cyclodextrin solution and administered at a dosage of 20 mg/kg bodyweight twice daily to pregnant dams at E8.5-E10.5 via intraperitoneal injection.

Embryos (E9.5-E18.5) were fixed in 4% paraformaldehyde, dehydrated, embedded, sectioned and Hematoxylin and Eosin (H&E) stained as described (Firulli et al., 2010). A minimum of four viable embryos per genotype was used for all analyses. All data were collected on a Leica DM5000 B compound fluorescence microscope.

Digoxygenin (DiG)-labeled section and whole-mount ISHs were carried out as described (Barnes et al., 2011; Firulli et al., 2010; Vincentz et al., 2008). qRT-PCR was performed on a LightCycler 480II (Roche) using TaqMan primers (Life Technologies) recognizing the following transcripts: Fgf8, Spry1, Fgfr1, Etv5, Etv4, Shh, Ptch1, Gli1, Gli3 and Twist1. Heads cut between the first and second PAs from viable embryos were flash-frozen for RNA isolation and genotyped from the yolk sac DNA. Total RNA was isolated using a High Pure RNA Tissue Kit (Roche) and cDNA was prepared using a High-Capacity cDNA Reverse Transcription Kit (Life Technologies) following the manufacturer's protocol. Error bars denote s.e.m. Statistical significance was determined using Student's t-test. P-values ≤0.05 were regarded as significant and marked in all graphs by a single asterisk, and P-values ≤0.01 are denoted by a double asterisk. n≥4 in all experiments.

TUNEL analysis on sectioned embryos was performed as described (Firulli et al., 2010). LysoTracker (Life Technologies) was incubated with embryos as per the manufacturer's instructions. Embryos were imaged in a well slide. Cell proliferation was assayed using the Click-IT EdU Imaging Kit (Life Technologies). Pregnant dams were injected with EdU (200 mg/kg body weight) 1 h prior to embryo harvest.

Craniofacial morphology of P0 mice was assessed using high-resolution desktop X-ray microtomography (micro-CT) SkyScan 1172 imaging system (SkyScan). Skulls were dissected, fixed in 10% neutral buffered formalin and stored in 70% ethanol. Skulls were scanned with an isotropic voxel size of 8 µm, with an energy level of 50 kV and an aluminum 0.5 mm filter. A lower energy source was used to capture regions of undermineralized bone. Two-dimensional (2D) cross-sectional grayscale slices (∼600-800 slices per skull) from each skull were reconstructed using NRecon reconstruction software (SkyScan). Reconstructed slices were saved as individual TIFF images and converted to a DICOM (Digital Imaging and Communications in Medicine) format. DICOM files were used to create 3D models using OsiriX version 5.6, imaging processing software for DICOM images (Medical Imaging Software). All 3D images were created using identical grayscale thresholds, with scaling of each image conserved. Overlaying skeletal structures were removed using a bone removal tool to identify structures of the palate (Fig. 1). n≥4 per genotype scanned and reconstructed. Measurements of bones were performed on all scanned heads and significance (P≤0.05) determined by Student's t-test. Skeletal preparations were performed as described (Firulli et al., 2005).

We thank Danny Carney for technical assistance. Infrastructural support at the Herman B Wells Center for Pediatric Research is in part supported by the generosity of the Riley Children's Foundation, and the Carrolton Buehl McCulloch Chair of Pediatrics.