Genetic interaction of Pax3 mutation and canonical Wnt signaling modulates neural tube defects and neural crest abnormalities

Mouse models provide opportunities to investigate genetic interactions that cause or modify the frequency of neural tube defects (NTDs). Mutation of the PAX3 transcription factor prevents neural tube closure, leading to cranial and spinal NTDs whose frequency is responsive to folate status. Canonical Wnt signalling is implicated both in regulation of Pax3 expression and as a target of PAX3. This study investigated potential interactions of Pax3 mutation and canonical Wnt signalling using conditional gain‐ and loss‐of‐function models of β‐catenin. We found an additive effect of β‐catenin gain of function and Pax3 loss of function on NTDs and neural crest defects. β‐catenin gain of function in the Pax3 expression domain led to significantly increased frequency of cranial but not spinal NTDs in embryos that are heterozygous for Pax3 mutation, while both cranial and spinal neural tube closure were exacerbated in Pax3 homozygotes. Similarly, deficits of migrating neural crest cells were exacerbated by β‐catenin gain of function, with almost complete ablation of spinal neural crest cells and derivatives in Pax3 homozygous mutants. Pax3 expression was not affected by β‐catenin gain of function, while we confirmed that loss of function led to reduced Pax3 transcription. In contrast to gain of function, β‐catenin knockout in the Pax3 expression domain lowered the frequency of cranial NTDs in Pax3 null embryos. However, loss of function of β‐catenin and Pax3 resulted in spinal NTDs, suggesting differential regulation of cranial and spinal neural tube closure. In summary, β‐catenin function modulates the frequency of PAX3‐related NTDs in the mouse.

genetic heterogeneity, and the potential influence of environmental factors. Hence, for common birth defects such as neural tube defects (NTDs), the genetic cause is not known in most individuals.
NTDs, including anencephaly and spina bifida, result from incomplete closure of the neural tube during embryonic development Nikolopoulou, Galea, Rolo, Greene, & Copp, 2017). Failure of closure in the cranial region leads to exencephaly, the developmental precursor of anencephaly, while failed closure of the spinal neural tube leads to spina bifida. Mouse genetic models provide insight into the molecular requirements for neural tube closure, with NTDs arising in more than 200 different mutant and knockout strains (Harris & Juriloff, 2007Nikolopoulou et al., 2017). Mouse models also provide opportunities to investigate the potential for multigenic causes of NTDs. Examples include strains in which there is a major risk allele with contributions from modifier genes, as in the curly tail strain in which NTDs result from a hypomorphic allele of Grhl3, with penetrance of NTDs influenced by contributions from other loci including Lmnb1 (De Castro et al., 2012;Gustavsson et al., 2007;Sudiwala et al., 2016;Ting et al., 2003). Alternatively, a number of studies have identified additive interactions of heterozygous genetic mutations which individually do not result in NTDs. For example, the Vangl2 Lp mutation causes a tail flexion defect in Vangl2 Lp/+ mice but interacts with heterozygous mutations in genes such as Celsr1 (encoding a component of the planar cell polarity signaling pathway) (Murdoch et al., 2014), as well as other genes including loss-and gain-of-function alleles of Grhl3 (Caddy et al., 2010;De Castro et al., 2018).
In the current study, we investigated the potential for modulation of NTDs caused by mutation of Pax3 by genetic alteration of canonical Wnt signaling. Pax3 encodes a member of the paired-and homeodomain-containing family of PAX transcription factors that play roles in a variety of developmental contexts (Blake & Ziman, 2014).
Key functions of PAX3 during embryogenesis are revealed by analysis of splotch mice, which carry mutations in Pax3 and are named for the characteristic belly spot present in heterozygotes (Auerbach, 1954;Greene, Massa, & Copp, 2009). At least nine spontaneous and radiation-induced alleles of splotch have been identified, corresponding to a range of Pax3 mutations and deletions and including the functionally null Sp and Sp 2H alleles (Epstein, Vekemans, & Gros, 1991;Epstein, Vogan, Trasler, & Gros, 1993). Several further knock-in alleles have been generated by gene targeting and recapitulate splotch phenotypes Mansouri, Pla, Larue, & Gruss, 2001;Zhou, Wang, Rogers, & Conway, 2008).
Pax3 is expressed in the dorsal neuroepithelium of the neural folds and closed neural tube, in populations of neural crest cells and in the developing somites (Goulding, Chalepakis, Deutsch, Erselius, & Gruss, 1991). Corresponding with its expression domain, homozygous mutation of Pax3 in mouse embryos leads to multiple defects including cranial and spinal NTDs, muscular defects, and abnormalities in neural crest derivatives in the heart, gut innervation, and melanocytes (Auerbach, 1954;Conway, Henderson, Anderson, Kirby, & Copp, 1997;Lang et al., 2000;Lang et al., 2005).
Pax3-related abnormalities are assumed to result from dysregulated transcription owing to the role of PAX3 as a transcriptional activator, as a homodimer or heterodimer with other transcription factors, or as a transcriptional inhibitor with co-repressors (Boudjadi, Chatterjee, Sun, Vemu, & Barr, 2018). A large number of targets have been identified in different cell types, but the molecular mechanisms underlying Pax3-related NTDs have yet to be determined. Among PAX3 targets, Wnt1 and Wnt3a appear to be regulated by Pax3 during neurulation and in neural crest development (Conway et al., 2000;Fenby, Fotaki, & Mason, 2008;Monsoro-Burq, Wang, & Harland, 2005). Conversely, the presence of binding sites for Wnt signaling mediators, TCF/LEF, in intron 4 of Pax3 as well as regulatory elements that confer indirect response to β-catenin in the proximal promoter suggests that Pax3 may itself be regulated by canonical Wnt signaling (Degenhardt et al., 2010;Moore et al., 2013). This is supported by findings in neural crest and neural tube development (Taneyhill & Bronner-Fraser, 2005;Zhao et al., 2014).
Canonical Wnt signaling involves the repression of activity of the Axin-GSK3-APC-containing destruction complex, such that phosphorylation and ubiquitination of β-catenin is diminished, and β-catenin is free to translocate to the nucleus (Nusse & Clevers, 2017). Owing to the potential role of canonical Wnt signaling both upstream and downstream of Pax3, we tested whether the frequency of NTDs caused by PAX3 mutation is modified by gain-or loss-of-function of β-catenin. In the current study, we found that β-catenin gain of function exacerbates both cranial and spinal neural tube closure, leading to more frequent NTDs than with Pax3 mutation alone. In contrast, cranial NTDs resulting from Pax3 mutation are partially rescued by β-catenin ablation.
2 | RESULTS 2.1 | β-Catenin gain of function exacerbates the effect of Pax3 mutation on neural tube closure We explored potential effects of β-catenin gain of function on Pax3related phenotypes using the Ctnnb1 floxE3 allele, in which cremediated deletion of exon 3 (Ctnnb1 ΔE3 ) leads to production of a stabilized β-catenin protein (Harada et al., 1999). Hence, we crossed mice of genotype Pax3 Sp2H/+ ; Ctnnb1 floxE3/+ with Pax3 cre/+ mice to generate litters containing embryos with combinations of Pax3 and Ctnnb1 alleles, for comparison of rates of NTDs. Litters include embryos of Pax3 Cre/+ ; Ctnnb1 ΔE3/+ and Pax3 Sp2H/Cre ; Ctnnb1 ΔE3/+ genotype, which have β-catenin gain of function in the Pax3 lineage in the dorsal epithelium and NCC derivatives. Within these litters, embryos of Pax3 Sp2H/Cre genotype (either Ctnnb1 wild type or gain of function) are functionally null for Pax3 as the cre knock-in ablates Pax3 exon 1.
F I G U R E 1 β-catenin gain of function exacerbates cranial NTDs and delays spinal neural tube closure in Pax3 mutant embryos. (a) Embryos carrying combinations of functionally null alleles of Pax3 (Pax cre , Pax3 Sp2H ) and the Ctnnb1 floxE3 allele (recombined by Pax3 cre to Ctnnb1 ΔE3 ) were analyzed at E9.5 or E10.5 for the presence of cranial NTDs. Numbers of embryos per group are shown in each bar (the functionally equivalent Ctnnb1 floxE3 and Ctnnb1 + alleles are combined). ** indicates significant difference between embryos of the same Pax3 genotype with/without Ctnnb1 ΔE3 (p < .001 Chi-Square). ## indicates significant effect of Pax3 genotype among embryos that are wild type for Ctnnb1 (p < .001 chisquare). (b-g) Compared with wild type in which the cranial NT is closed (b, e), and Pax3 cre/Sp2H in which the cranial region is either closed (d) or has a moderately sized NTD (G), Pax3 cre/Sp2H ; Ctnnb1 ΔE3/+ embryos show severe cranial and spinal NTDs at E9.5 (c) and E10.5 (f). The region of open neural folds in the cranial region is indicated by white arrowheads and the rostral extent of the enlarged PNP is indicated by black arrowheads. (h) Embryos were scored for spinal NTDs (spina bifida) on the basis of failed PNP closure at E10.5 (30 or more somites). (i) Analysis at earlier stages (E9.5) revealed significant increase in anterior-posterior length of the PNP in Pax3 cre/Sp2H embryos, suggesting exacerbation of spinal closure. * indicates significant difference between embryos of the same Pax3 genotype with/without Ctnnb1 ΔE3 (p < .05, ANOVA) In embryos with β-catenin gain of function, the open cranial neural folds typically encompassed a more extensive region, including the entire hindbrain (Figure 1b-g), midbrain, and sometimes the forebrain.
Analysis of litters at E10.5 showed that spinal NTDs occurred with high frequency in Pax3 null embryos, irrespective of Ctnnb1 genotype, and also occasionally in heterozygotes ( Figure 1h). We observed a non-significant trend toward more frequent spinal NTDs with β-catenin gain of function, but the high frequency of spinal NTDs in Pax3 cre/Sp2H embryos diminished sensitivity to detect exacerbation of these defects at E10.5, by which stage closure has either succeeded or failed ( Figure 1h). We therefore evaluated spinal neurulation during the period of closure by measuring the length of open posterior neuropore (PNP) at E9.5, shortly after delay of closure (indicated by increased anterior-posterior length of the PNP) first becomes apparent in Pax3 mutant compared with wild-type embryos (Sudiwala et al., 2019). Among stage-matched Pax3 null embryos, we observed significant enlargement of the PNP when the Ctnnb1 ΔE3 allele was also present, indicating further delay of closure imposed by β-catenin gain of function ( Figure 1i).
The high rate of cranial NTDs in Pax3 cre ; Ctnnb1 ΔE3/+ embryos suggests that there is an additive genetic interaction between PAX3 loss of function and β-catenin gain of function, in the dorsal neuroepithelium in which Pax3 is expressed. To further investigate this hypothesis, we asked whether β-catenin activation is sufficient to induce NTDs in the absence of a co-occurrent Pax3 null allele by using Sox1 cre to recombine Ctnnb1 floxE3 throughout the cranial neuroepithelium. Among litters from a cross of Pax3 Sp2H/+ ; Ctnnb1 floxE3 mice with Sox1 cre/+ , cranial NTDs arose among 25% (one out of four) of Sox1 cre/+ ; Ctnnb1 ΔE3 , and 20% (3 out of 15) Sox1 cre/+ ; Ctnnb1 +/+ embryos. The presence of a low frequency of exencephaly in Sox1 cre/+ embryos was surprising and may relate to the presence of the Sox1 cre knock-in allele. Nevertheless, these findings suggest that β-catenin gain of function alone is not responsible for the high rate of NTDs in the Pax3 cre ; Ctnnb1 ΔE3 embryos. In contrast, the two embryos from this cross in which a Pax3 mutant allele was present (Pax3 Sp2H/+ ; Sox1 cre/+ ; Ctnnb1 ΔE3 ) both exhibited cranial NTDs (2/2; 100%), consistent with an additive effect of Pax3 mutation with β-catenin gain of function.

| Canonical Wnt signaling appears unaffected by Pax3 mutation
We confirmed that canonical Wnt signaling was increased among embryos carrying the Ctnnb1 ΔE3 allele by qRT-PCR ( Figure 2a) and whole mount in situ hybridization for the target gene Axin2 (Figure 2b-e). In contrast, in the absence of this allele we did not observe any effect of Pax3 genotype on Axin2 expression suggesting that there was not a pre-existing over-activation of canonical Wnt signaling in Pax3 mutants that is exacerbated by β-catenin gain of function. Lack of an alteration in Wnt signaling in Pax3 mutants was also consistent with findings obtained using the BAT-Gal reporter, in which LacZ is expressed under the control of LEF/TCF-regulatory elements (Maretto et al., 2003). After breeding the reporter into the Pax3 Sp2H line, relative LacZ expression did not differ between Pax3 +/+ and Pax3 Sp2H/Sp2H embryos at E9.5 ( Figure S1a). Staining for β-galactosidase activity in BAT-Gal positive embryos at E9.5 confirmed that the domain of canonical Wnt signaling activity ( Figure S1b 2.2 | β-Catenin loss of function lowers the frequency of cranial NTDs in Pax3 null embryos but causes spinal NTDs Having observed an effect of β-catenin gain of function on neural tube closure, we next tested whether the frequency of Pax3-related NTDs was altered by β-catenin loss of function, using a conditional allele in which cre-mediated recombination deletes exons 2-6 of Ctnnb1, creating a null allele, Ctnnb1 ΔEx2-6 (Brault et al., 2001). The frequency of NTDs was compared within litters from a cross of Pax3 Sp2H/+ /Ctnnb1 floxEx2-6 with Pax3 cre/+ /Ctnnb1 floxEx2-6 . Cranial NTDs occurred at low frequency when Ctnnb1 was deleted in the Pax3 domain, generating Pax3 cre/+ ; Ctnnb1 ΔEx2-6/ΔEx2-6 (β-catenin loss of function in the Pax3 domain), but this did not differ significantly from embryos carrying the Pax3 cre/+ allele with Ctnnb1 +/ΔEx2-6 or Ctnnb1 +/+ genotype (Figure 3a). However, while Pax3 null (Pax3 cre/Sp2H ) embryos that were wild type or heterozygous for Ctnnb1 ΔEx2-6 showed cranial NTDs with an expected frequency of around 45%, this was significantly lower among Pax3 cre/Sp2H ; Ctnnb1 ΔEx2-6/ΔEx2-6 embryos that were homozygous for the floxed loss of function β-catenin allele (Figure 3a; p < .05; z-test). Homozygous Pax3 mutants displayed a high rate of spinal NTDs as expected, irrespective of Ctnnb1 genotype. Similarly, β-catenin loss of function in the dorsal neuroepithelium of Pax3 cre/+ ; Ctnnb1 ΔEx2-6/ΔEx2-6 embryos also caused spinal NTDs with high frequency (Figure 3b), consistent with previous findings (Zhao et al., 2014).
Pax3 has been reported to be a target of Wnt signaling (Zhao et al., 2014). We therefore evaluated transcription from the Pax3 locus by qRT-PCR using primers which amplify the wild-type and Pax3 cre alleles. We found significant reduction of expression from the Pax3 locus in Ctnnb1 conditionally deleted embryos ( Figure 4a) and this was replicated using allele-specific primers, which amplify the wild-type and Pax3 Sp2H alleles ( Figure S2). We tested whether β-catenin conditional gain of function may have a reciprocal effect on Pax3 expression. However, we did not find that presence of the Ctnnb1 ΔE3 allele led to altered Pax3 expression (Figure 4b).

| β-Catenin gain of function exacerbates the effect of Pax3 mutation on neural crest cell migration
In addition to NTDs, PAX3 loss of function leads to defects in NCCderived tissues including peripheral nervous system, cardiac outflow tract, melanocytes, and limb musculature (Goulding, Lumsden, & Paquette, 1994;Greene et al., 2009). We therefore asked whether NCC specification and/or migration were exacerbated by β-catenin gain of function in Pax3 mutant embryos, by analyzing expression of ErbB3, a marker of migrating NCC at E9.5-10.5 ( Figure 5). ErbB3 expression was diminished in Pax3 Sp2H/Cre compared with Pax3 +/+ embryos when wild type for Ctnnb1 (Figure 5a, f,k compared with d, i, n). The precursors of the dorsal root ganglia (DRG) did form in Pax3 mutant embryos, but were abnormally small, and segmentation was visible (Figure 5d, i, n).
The additional presence of the Ctnnb1 ΔE3 allele in Pax3 Sp2H/Cre ; Ctnnb1 ΔE3/+ embryos resulted in lack of ErbB3 staining in the trigeminal F I G U R E 2 Upregulation of Axin2 in embryos carrying the β-catenin gain of function allele. (a) Expression of Axin2 was assessed by qRT-PCR using RNA extracted from whole embryos at E10 (22-24 somites). Expression did not differ between wild-type and Pax3 mutant embryos in the absence of the Ctnnb1 ΔE3 allele, whereas embryos in which Ctnnb1 ΔE3 was expressed in the Pax3 domain showed significant upregulation of Axin2 (* significantly different from embryos with Ctnnb1 +/+ genotype, p < .05 ANOVA; n = 3-4 per genotype). (b) Whole mount in situ hybridization at E9.5 showed more intense Axin2 staining in embryos carrying the Ctnnb1 ΔE3 allele and this was particularly evident in the pharyngeal arches (arrowhead in d). Scale bar represents 250 μm F I G U R E 3 β-catenin loss of function has opposing effects on Pax3-related cranial and spinal NTDs. Experimental litters were generated by intercross of Pax3 Sp2H/+ ; Ctnnb1 floxE2-6/+ with Pax3 cre ; Ctnnb1 floxE2-6/+ and analyzed at E9.5-10.5 for (a) cranial and (b) spinal NTDs. In Ctnnb1 +/+ embryos, NTDs occurred more frequently among Pax3 cre/Sp2H than wild-type as expected. (a) Within Pax3 cre/+ genotype, the frequency of cranial NTDs did not differ significantly with Ctnnb1 genotype, whereas (b) deletion of Ctnnb1 in the Pax3 domain led to a significant increase in spina bifida frequency (**, p < .01 z-test). (a) Pax3 cre/Sp2H ; Ctnnb1 ΔEx2-6/ΔEx2-6 embryos exhibited a lower rate of cranial NTDs than Pax3 cre/Sp2H embryos carrying the Ctnnb1 +/ΔEx2-6 or +/+ alleles (*p < .05; z-test). Development of the NCC-derived peripheral nervous system was further evaluated by immunostaining for β-tubulin III (TuJ1) (Figure 6).
Staining in the DRG of Pax3 Sp2H/Cre embryos was less defined than in Pax3 +/+ , corresponding with the previously reported diminished size of DRGs (Auerbach, 1954;Conway, Henderson, Anderson, et al., 1997;. However, in Pax3 +/Cre ; Ctnnb1 ΔE3/ + and Pax3 Sp2H/Cre ; Ctnnb1 ΔE3/+ embryos the spinal DRGs showed disrupted patterning, segmentation was poorly defined and the vagus nerve appeared disrupted (Figure 6f compared with e and h compared with g). Hence, gain of function of β-catenin resulted in abnormalities of NCC derivatives, although the presence of ErbB3 expression until E10 (24-25 somite stage) suggested that some migration did occur.

| DISCUSSION
Our findings reveal the potential for a multigenic cause of NTDs involving Pax3 mutation and dysregulated canonical Wnt signaling. We find a significant deleterious effect of activated β-catenin function on neural tube closure when present in combination with heterozygous or homozygous Pax3 mutation. This is sufficient to cause a high frequency of cranial NTDs, even though enhanced β-catenin function is present only in the dorsal neuroepithelial component of the neural folds corresponding to the domain recombined in embryos carrying the Pax3 cre allele. Notably, conditional knock-out of β-catenin (using the same Pax3 cre allele) led to partial rescue of cranial NTDs in Pax3 null embryos.
Compared with cranial closure, there was a lesser effect of β-catenin gain of function on spinal closure, but further exacerbation of PNP closure defects did occur in homozygous Pax3 mutants. As previously reported (Zhao et al., 2014), conditional knock-out of β-catenin in the Pax3 domain is also sufficient to cause spinal NTDs in Pax3 cre ; Ctnnb1 ΔEx2-6/ΔEx2-6 embryos. Hence, unlike in the cranial region, a genetic interaction of Pax3 and either loss or gain of Ctnnb1 can contribute to spinal NTDs, suggesting differential regulation of closure at cranial and spinal levels.
We did not find evidence that activation of β-catenin is sufficient to positively regulate Pax3, whereas reduced Pax3 expression was found when canonical Wnt signaling was diminished, consistent with previous studies (Zhao et al., 2014). We also confirmed that conditional knock-out of Ctnnb1 using Pax3 cre causes spinal NTDs. These defects do not appear to be solely due to the loss of Wnt signaling, independent of the Pax3 cre knock-in allele, as Pax3 over-expression was found to rescue NTDs in an equivalent genotype (Zhao et al., 2014). Therefore, the reduction in Pax3 expression that accompanies loss of Ctnnb1 may contribute to spinal NTDs in double mutants, by further suppression of residual Pax3 expression. In addition, there may be an additive effect of diminished Wnt signaling and heterozygosity for Pax3 acting via distinct mechanisms. For example, although mice that are heterozygous for Pax3 mutations do not develop NTDs, an additive effect with other NTD-predisposing mutations has also been found with genetic cross of the Pax3 Sp2H allele into the curly tail (ct/ct) strain, which led to increased frequency of spina bifida compared with ct/ct alone (Estibeiro, Brook, & Copp, 1993).
PAX3-related NTDs have been found to be associated with diminished cell cycle progression, and premature neuronal differentiation, in the dorsal neuroepithelium corresponding to the Pax3 expression domain (Sudiwala et al., 2019). Cranial NTDs in Pax3 mutant (Sp 2H ) embryos can be prevented by folic acid supplementation and exacerbated by maternal folate deficiency (Burren et al., 2008;Copp, Fleming, & Greene, 1998). Notably, rescue of cranial NTDs by supplemental F I G U R E 4 Expression from the Pax3 locus is diminished by β-catenin loss of function but not stimulated by β-catenin gain of function. Expression of wild-type and Pax3 cre alleles was compared between genotypes by qRT-PCR among stage-matched embryos (23-24 somite stage) at E10 (n = 3-4 embryos per group). (a) Conditional β-catenin loss of function (Ctnnb1 ΔEx2-6 ) led to a significant decrease in Pax3 expression (#p < .05; ANOVA), whereas (b) conditional gain of function (Ctnnb1 ΔEx3 ) did not affect Pax3 expression. Presence of the Pax3 Sp2H allele (truncated transcript not detected by this primer pair) led to a significant reduction in Pax3 transcript as expected (*indicates significant difference from Pax3 +/+ ; Ctnnb1 +/+ , p < .05; one-way ANOVA) folic acid was associated with correction of the proliferation defect, with treated embryos showing enhanced progression through S-phase (Sudiwala et al., 2019). These findings suggest that the causative mechanism for NTDs in Pax3 mutant embryos involves diminished proliferation in the dorsal neuroepithelium. It will be of interest to determine whether the modulation of NTD frequency by canonical Wnt signaling is also mediated through effects on cellular proliferation.

| Mice
The Pax3 Sp2H allele carries a 32 bp deletion in exon 5 (encoding the paired-type homeodomain), which generates a premature stop codon encoding a truncated and functionally null protein (Beechey & Searle, 1986). Pax3 cre is a functionally null knock-in allele in which Cre replaces exon1 . β-catenin gain of function was achieved using the Ctnnb1 floxE3 allele in which exon 3 is flanked by loxP sites. Recombination deletes exon 3, which encodes the GSKβ phosphorylation domain such that Ctnnb1 ΔE3 is not targeted for ubiquitination and is stabilized (Harada et al., 1999). Conditional β-catenin loss of function was achieved using Ctnnb1 tm2Kem (Ctnnb1 floxEx2-6 ), in which recombination of loxP sites flanking exons 2-6 leads to deletion of these exons and creation of a null allele (Brault et al., 2001). The Sox1 cre allele carries an in-frame knock-in cre into exon 1 of Sox1 (Takashima et al., 2007). The BAT-Gal reporter was previously described (Maretto et al., 2003). Pax3 cre , Ctnnb1 ΔE3 , Ctnnb1 flox2-6 , and Sox1 cre were all on a C57BL/6 background. The splotch (Pax3 Sp2H ) mice have been maintained as a closed colony for more than 50 generations on a mixed background derived from CBA/Ca, 101 and C3H/He. Compound heterozygotes (Pax3 Sp2H/+ ; Ctnnb1 flox/+ ) used for experimental matings were the F1 generation (50% C57BL/6).
F I G U R E 5 β-catenin gain of function disrupts neural crest development as visualized by expression of ErbB3. Whole mouse in situ hybridization for ErbB3 at (a-e) E9.5 (18-20 somites), (f-j) E10 (24-26 somites), and (k-o) E10.5 (around 30 somites). In wild-type embryos, ErbB3 is expressed in the DRG primordia (black arrowheads), trigeminal ganglia (black arrow), and otic vesicle (red arrow). The pattern is similar in Pax3 heterozygotes (b, g, l) but staining is less intense in the DRGs of Pax3 Sp2H/cre embryos (black arrowheads in d, i, n). Although ErbB3 expression in the DRGs of Pax3 cre/+ ; Ctnnb1 ΔE3/+ appeared comparable to wild-type at E9.5 (c), staining was less intense by E10 (h) and absent by E10.5 (m), by which stage expression in the vagus and trigeminal also appeared abnormal. Pax3 cre/Sp2H ; Ctnnb1 ΔE3/+ embryos (e, j, o) had even weaker ErbB3 expression in the developing DRGs and this was absent by E10.5 (o), as was expression in the vagus and trigeminal ganglia. Scale bars represent 500 μm and 2-3 embryos per genotype were analyzed at each stage Animal studies were carried out under regulations of the Animals  Embryos were rinsed in PBS and either frozen at À80, prior to RNA extraction for RT-PCR or fixed in 4% paraformaldehyde (PFA), dehydrated in a methanol series and stored at À20 C prior to in situ hybridization. Embryos were genotyped by PCR of yolk sac genomic DNA.

| Whole mount in situ hybridization and immunostaining
Whole-mount in situ hybridization (De Castro et al., 2018) was performed using sense and anti-sense digoxygenin-labeled riboprobes for Axin2 (Andoniadou et al., 2007) and Erbb3 (Henderson et al., 2001) were generated using a digoxygenin RNA-labeling kit (Roche) and purified on Chroma spin columns (Clontech).

| Real-time quantitative RT-PCR (qRT-PCR)
RNA was extracted using Trizol reagent (Invitrogen) and used for first strand cDNA synthesis using VILO Superscript cDNA synthesis kit (Invitrogen). qRT-PCR was performed using Mesa Blue qPCR Master Mix Plus for SYBR assay on an ABI 7500 Real-Time PCR machine (Applied Biosystems) with β-actin used as a housekeeping gene to normalize expression (De Castro et al., 2012;De Castro et al., 2018).
Primers for amplification of Axin2 were 5 0 AAGCCTGGCTCCA GAAGATCACAA and 5 0 TTTGAGCCTTCAGCATCCTCCTGT. Primers F I G U R E 6 Neuronal differentiation is disrupted by β-catenin gain of function. Whole mount immunostaining for β-tubulin III (TuJ1) identifies differentiating neurons at (a-d) E10 (23-24 somites) and (e-h) E10.5 (approximately 30 somites). In wild-type embryos, neuronal differentiation is extensive in the cranial region at E10 and is progressing in a rostral to caudal direction in the spinal region (a). At E10.5 (e), the regular pattern of DRGs is evident in the spinal region (yellow arrowhead) and the vagus nerve is well defined (white arrowhead). In Pax3 cre/Sp2H ; Ctnnb1 +/+ embryos (c, g), the vagus nerve is well defined and DRGs form but are smaller except in the caudal region where the neural folds remain open. In contrast, in the presence of Ctnnb1 ΔE3/+ , the pattern of cranial nerves appears disrupted and there is a failure of DRG segmentation, an effect that is more evident at E10.5 (f, h) that E10 (b, d) for amplification of Pax3 were designed to: (a) exons 5-6, 5 0 GGCTTTCGAGAGAACCCACT and 5 0 AGGTCTCCGACAGCTGG TAT (to evaluate expression of the wild-type and Pax3 cre alleles) and (b) exons 1-2, 5 0 GTGCTCGCTTTTTCGTCTCG and 5 0 CAGAGGC CTGCCGTTGATAA (to evaluate expression of the wild-type and Pax3 Sp2H alleles).
Statistical analysis was performed using Sigmastat version 3.5 (Systat Software). Multiple groups in qRT-PCR experiments were compared by One Way ANOVA with Holm-Sidak post-hoc test.

ACKNOWLEDGMENTS
This work was funded by the Wellcome Trust (087525) and by a Child Health Research PhD Studentship. NDEG and AJC were supported by Great Ormond Street Children's Charity.