talpid³ is a naturally occurring chicken mutant. The talpid³ mutation is recessive and leads to leaky blood vessels among other defects. This is causing very high embryonic mortality as reported in both chicken and mouse Talpid³ models (Davey et al., 2006; Ben et al., 2011). Although limbs and organ defects have been reported in talpid³ chicken embryos, there has been a lack of detailed analysis of the GI defects in these mutants. As previously described, Buxton et al. (2004) E10.5 talpid³ chick embryos were smaller than controls, showed generalized edema and displayed a wide range of congenital abnormalities including short ribs, polydactylous paddle shaped limbs, organ defects (lung hypoplasia, liver fibrosis, and cholestasis) and craniofacial abnormalities (hypotelorism, reduction and anterior displacement of the frontonasal process) (Figure 1A). At E10.5, the GI tract of talpid³ chick embryos was significantly reduced in length compared to controls (Figure 1B). As previously described, no left right asymmetry or rotation defect was observed in the stomach of talpid³ chick embryos (Stephen et al., 2014). Transverse sections at the level of the neck showed abnormal connection (fistula) between the esophagus and the trachea or a total absence or narrowing (atresia) of either of the structures (Figures 1G,O). Additionally, talpid³ embryos had an open hindgut (Figure 1B – inset-, J,R). Lastly, gut epithelium thickness varied between controls and mutants, as shown in Figures 1M,Q insets.

To gain further insight into the GI defects of talpid³ embryos we performed immunohistochemistry using molecular markers to highlight the neuromuscular organization of the GI tract in controls and talpid³ mutants. In E8.5 controls, smooth muscle actin (SMA) immunostaining revealed a compact ring of immunopositive cells, corresponding to the presumptive circular muscle layers, encircling the gut epithelium (Figures 1C–F). Notably, this muscular ring was located in the outer mesenchyme, with a distinct separation between the muscle and the gut epithelium (Figures 1C–F, double arrows). In talpid³ gut, despite the presence of SMA staining surrounding blood vessels, SMA positive cells were discontinuous or absent around the esophagus (Figure 1G). In the stomach and the intestine, the compact muscular ring seen in controls was replaced by diffuse staining that extended across the entire gut mesenchyme, with SMA positive cells abutting the gut epithelium (Figures 1H,I, arrowheads). In the colorectum, the circular muscular pattern was disrupted by the open hindgut phenotype (Figure 1J). In E8.5 controls, TuJ-1 immunostaining revealed ENCC-derived neurons arranged in characteristic plexuses (Figures 1K–N). In the esophagus and the stomach, the presumptive ENS was organized in a single plexus located in the outermost mesenchyme (Figures 1K,L, double arrows). In the intestine and colorectum, ENCC were organized in two plexuses: the myenteric plexus (MYP) and the submucosal plexus (SMP) (Figures 1M,N). The nerve of Remak (Rem), an avian specific nerve derived from sacral ENCC, was also positively labeled (Figure 1N). In E8.5 TALPID3 embryos, ENCC were present throughout the intestine (Figure 1Q) and colorectum, although well-defined plexuses were not apparent in the distal gut due to the open hindgut phenotype (Figure 1R). Although ENCC were observed alongside the esophagus at E6.5 (data not shown), they were not found around the esophagus at E8.5 and did not form a plexus (Figure 1O). At this stage, ENCC were present in the stomach and intestine regions, but they failed to organize in plexuses and were scattered throughout the gut wall (Figures 1P,Q). In the colorectum, the nerve of Remak was also smaller in diameter and/or comprised several bundles (Figures 1Q,R). The presence of ENCC in distal parts of the GI tract demonstrated that TALPID3 is not required for ENCC migration per se. knowing the role of TALPID3 in ciliogenesis, we looked at the presence of primary cilia on migrating ENCC (Supplementary Figure 1). In chicken neural tube cultured in vitro, primary cilia were readily evident on the neural tube cells. However, we failed to observe them on migrating vagal ENCC, either anatomically by SEM or molecularly using immunofluorescence (Supplementary Figures 1A–D). In vivo, despite primary cilia being readily detected in E6.5 chicken gut sections, only 11% of HuC⁺ ENCC showed a detectable primary cilium (Supplementary Figures 1E,F). Overall, these results showed regions such as the esophagus being devoid of both ENS and smooth muscle in talpid³ mutants, whereas regions such as the ventral stomach and the intestine showed scattered ENS and extended regions of smooth muscle differentiation across the mesenchyme.

Due to the well-established functional connection between TALPID3 and Hh signaling, we investigated changes in this pathway in control and talpid³ mutants using in situ hybridization. In situ analysis of shh expression in the GI tract of E6.5 control and talpid³ embryos revealed transcripts in the gut epithelium of both type of tissues, demonstrating expression of the shh gene in the talpid³ mutants (Figures 2A–H). Expression of SHH was confirmed in the endoderm of the intestine using an anti-SHH antibody (Supplementary Figure 2). To visualize the readout of Hh signaling in the gut wall, mRNA expression of the Hh receptor PTCH1 was used. In E6.5 control embryos, the pattern of expression of PTCH1 was composed of one or two concentric gradients surrounding the epithelium (Figures 2I–L). In the esophagus, the ventral part of the stomach and the intestine, two concentric gradients were present; the first was located adjacent to the epithelium in the sub-epithelial mesenchyme, and the second more distally located in the outer mesenchyme (Figures 2I–K). In the dorsal part of the stomach and the colorectum a single sub-epithelial gradient was present (Figures 2J,L). Strikingly, the discrete characteristic gradients of PTCH1 expression were absent in the GI tract of talpid³ mutants. Instead a diffuse homogenous expression was observed throughout the gut wall and the epithelium. The same diffuse PTCH1 pattern was observed at all levels of the GI tract (Figures 2M–P). Together these results showed that, despite epithelial expression of SHH in talpid³ mutants, the precise integration of the signal in the surrounding mesenchyme was lost, as shown by the aberrant PTCH1 expression.

In search of a mechanistic explanation for the lack of ENS plexus formation in talpid³ mutants, we analyzed expression of components of the gut ECM known to influence neuronal behavior (Tennyson et al., 1990; Siebert et al., 2014; Nagy et al., 2016). First, we used the CS-56 antibody, which stains the glycosaminoglycan portion of native chondroitin sulfate proteoglycans (CSPG). CSPG are ECM components regulated by Hh signaling (Nagy et al., 2016). In E6.5 control tissues, CS-56 staining comprised two concentric gradients: the first immediately adjacent to the gut epithelium (sub-epithelial mesenchyme) and the second located in the outer gut wall (outer mesenchyme) (Figure 3A). This expression pattern was very similar to PTCH1 as described above. In E6.5 talpid³ gut tissues, CS-56 expression was absent from the outer mesenchyme, with only cells adjacent to the epithelium staining positive (Figure 3F). Additionally, we examined the expression of Collagen 9 (Coll9), a ECM component expressed in the developing gut, which has been specifically shown to elicit avoidance behavior by neural crest cells in vitro (Ring et al., 1996; Nagy et al., 2016). Consistent with the CS-56 staining, Coll9 pattern of expression was also composed of two concentric areas, one within the sub-epithelial mesenchyme and one in the outer mesenchyme (Figures 3B,D). Interestingly, the Coll9 expression pattern appeared to define exclusion zones for the migrating ENCC, which were only present outside Coll9 positive areas, as shown by N-cadherin (NCadh) staining (Figures 3B,C; double arrows). As seen with CS-56, there was no Coll9 expression in the outer mesenchyme of E6.5 talpid³ gut samples and only the sub-epithelial mesenchyme was stained (Figures 3G,H). Interestingly, and in correlation with the lack of outer mesenchyme Coll9 expression, NCadh-positive ENCC were scattered throughout the gut mesenchyme, with some cells located adjacent to the epithelium (Figures 3G,H; arrowheads in inset). We also investigated the expression of Coll9 respective to smooth muscle differentiation. We found that SMA and Coll9 have distinct, yet partially overlapping, patterns of expression in control tissues (Figures 3D,E). In E6.5 control esophagus, the distal gradient of Coll9 corresponded with the inner boundary of the smooth muscle ring (Figure 3D). In the stomach, Coll9 and SMA were mostly mutually exclusive apart from a subdomain in the ventral region where both were co-expressed (Figure 3E). In E6.5 talpid³ esophagus, both the Coll9 distal gradient and the SMA ring were absent (Figure 3I). In the talpid³ stomach, most of the Coll9 expression domain was absent compared to control, while the SMA-positive domain was extended (Figure 3J). These results show that expression of CSPG in the gut mesenchyme is regulated by talpid³ (likely indirectly via the dysregulation of the Hedgehog pathway), as its absence led to significant loss in expression of these ECM components. Moreover, the changes in ECM components expression were concurrent with mislocalization of migrating ENCC.

To investigate the role of TALPID3 in ENS plexus formation, we attempted to rescue normal ENS patterning in talpid³ mutants by grafting wild type neural tubes, including neural crest, into talpid³ host embryos (Figure 4). GFP chicken tissues were used as donors for the transplants, so that vagal ENCC had a functional TALPID3 protein and could also be traced in the chimeric embryos using GFP expression. When compared to GFP > wild type transplanted controls (Figures 4A–D) or stage matched talpid³ embryos (Figures 4I–L), transplanted GFP + ENCC behaved similarly to ENCC of talpid³ embryos and were unable to rescue ENS patterning (Figures 4E–H). At the level of the esophagus, instead of surrounding the esophagus in a presumptive circular plexus, as seen in controls (Figure 4A), transplanted GFP + ENCC clustered in the dorsal region (Figure 4E), in a pattern similar to that seen in talpid³ embryos (Figure 4I). In the stomach and intestine, transplanted GFP + ENCC migrated similar distances along the GI tract as observed in talpid³ embryos but were scattered throughout the mesenchyme (Figures 4F,G,J + inset G) and not arranged in plexuses, as seen in controls (Figures 4B–D + inset C). Epithelium and smooth muscle thickness (stained with DAPI and Phalloidin, respectively) in intestine and stomach sections were measured in stage matched GFP > wild type and GFP > talpid³ chimera, as well as talpid³ mutants. Measurements from GFP > talpid³ chimeric tissues and talpid³ mutant were both statistically (^***p < 0.001) different from the controls (Supplementary Figure 3). The non-parametric Spearman’s Rho test was used to measure the correlation between GFP > wild type and GFP > talpid³ chimera, and talpid³ mutants measurements. The correlation between GFP > wild type and GFP > talpid³ transplants was 0.670, whereas the correlation between the GFP > wild type transplant and the talpid³ mutant was 0.715. Importantly, the correlation between the GFP > talpid³ and the talpid³ mutant was 0.942, an extremely high score. The fact that measurements from chimera and mutant clustered together and were statistically different from the control supported histological findings of a lack of rescue. Overall, wild type transplanted vagal ENCC did not rescue ENS plexus formation or the smooth muscle phenotype in a talpid³ environment.

To further assess the role of TALPID3 during ENS development and to test its cell autonomous requirement in ENCC, we performed the converse experiment to that described above, by grafting talpid³ vagal ENCC into GFP-expressing chick embryos. In the resulting chimeric embryos, the transplanted ENCC were devoid of functional TALPID3 protein and were identified by lack of GFP expression (GFP^–ve; Figure 5 inset). Due to the extremely high mortality of this type of chimera, only one specimen survived to E7.5 for the gut tissues to be analyzed. Surprisingly, transplanted vagal talpid³ ENCC migrated throughout the GI tract and patterned normally into the characteristic MYP and SMP of the ENS, as shown by HNK1 and GFP staining at E7.5 (Figures 5A–H). Importantly, talpid³ENCC colonized the entire length of the GI tract and were found in the colorectal region (Figure 5H + inset). Here, vagal GFP^–ve ENCC (red) were present alongside GFP⁺ (yellow) ENCC (Figure 5H left inset) that were likely sacral ENCC entering the gut prematurely (Burns and Le Douarin, 2001). In contrast to the ENS, which was grossly normal in the chimeric embryo, there were obvious defects in smooth muscle differentiation as revealed by SMA staining: instead of the characteristic SMA⁺ ring of differentiating smooth muscle cells seen at E7.5 in control chimeras (Figures 5I–P + insets J,N), there was thickened and diffuse SMA staining in regions of the esophagus and the stomach, suggesting impaired differentiation (Figures 5I,J,M,N double arrows + insets J,N; Figure 6H). The characteristic SMA⁺ ring of cells was also missing in the intestine of the chimeric embryo (Figure 5O). Normal SMA staining was observed in some areas around the esophagus and within the colorectum, which coincided with the presence of normal vagal ENCC from the region adjacent to the graft and wild type sacral ENCC, respectively (Figure 5H inset). Additionally, the architecture of the mucosa was disrupted in the chimeric embryo, with the esophageal epithelium consisting of a thin folded monolayer, unlike controls where a thickened circular epithelium was present (Figures 5A,E,I,M doted lines). Using DAPI, SMA, and TuJ-1 staining, epithelial cells, smooth muscle cells and enteric neurons were counted on stomach sections (n = 3 sections minimum, as shown in Figures 5B,F,J,N). Numbers were combined with equivalent murine data for chicken/mouse cross-species statistical analysis of chimeric animals where Talpid3 was knocked down in ENCC, as described below (Figure 6H).

Because only one talpid³ > GFP chicken chimera could be analyzed (due to extremely high mortality) we wished to confirm the chicken results using an alternative approach. For this, we engineered a conditional knock out of the Talpid3 gene in a mouse model by crossing a Wnt1:Cre line (Danielian et al., 1998), with a floxed Talpid3 knock-out line. Resultant Wnt1:Cre;Talpid3 ^flox/flox embryos, which do not express TALPID3 in NCCs, were often embryonic lethal or died at P0, due to presumptive respiratory problems, as their lungs failed to fill with air (Supplementary Figure 4). P0 pups showed a craniofacial phenotype characteristic of hypoplastic neural crest derivatives (Supplementary Figures 3D,F). In mutants, the GI tract morphology was grossly normal compared to control littermates (Supplementary Figures 4B,E). As seen in the chicken model, Talpid3 ENCC colonized the entire length of the gut and formed an apparently normal ENS all along the GI tract (Figures 6A–F,A′′–F′′). Although the ENS was grossly normal, quantitative analysis of the ENS in the stomach region showed a statistically significant (p = 0.045) increase of 29% in the number of ENCC in the mutants (n = 8 sections) (Figure 6G). Of note, and again in accordance with the defects observed in the chicken model, the number of cells in both the epithelium and muscle layers was affected, in a non-cell autonomous manner, by the lack of TALPID3 in the ENCC. Indeed, quantitative analysis of cell numbers in both tissues showed statistically significant 39% increase in the epithelium (p = 0.009; n = 7 sections) and 67% increase in the muscle layers (p = 0.002; n = 7 sections), with an 80% increase in the longitudinal muscle (p = 0.002) and 55% increase in the circular muscle (p = 0.018). Interestingly, the smooth muscle myoblasts appeared misshaped in the conditional Talpid3 mutant (Figures 6A′–F′′,A′′–F′′ and Supplementary Videos 1, 2). The transgenic mouse measurements were combined with equivalent chicken data from the talpid³ > GFP transplant for mouse/chicken cross-species analysis. Cell counts for epithelium showed a statistically significant (^**p = 0.009) increase of 39% in the chimeric animals compared to controls. Smooth muscle cell count showed a highly statistically significant (^**p = 0.002) increase of 67% in chimeric animals compared to control. The relative increase observed in ENS cell numbers was not statistically significant in this cross-species analysis (Figure 6H). Overall, apart from altered cell numbers, knocking out talpid³ in ENCC had little effect on early ENS development and patterning in both chicken and mouse models, but it altered growth and differentiation of smooth muscle and mucosa in a non-cell-autonomous manner.

To assess whether the role of TALPID3 is conserved throughout evolution and is relevant to human GI tract development and patterning, we examined human fetal GI tissues obtained from a 26 weeks human fetus with a homozygous for a 1815G > A mutation in KIAA0586, the human ortholog of talpid³ (chicken) and Talpid3 (mouse), as previously described (Alby et al., 2015). Previous anatomical description of the fetus listed shortened ribs, micromelia, lingual hamartomas, postaxial and preaxial polydactyly, temporal polymicrogyria and an occipital keyhole defect (Alby et al., 2015; Cocciadiferro et al., 2020). Gross examination of the GI tract revealed an elongated and tubular stomach, but otherwise apparently normal GI tract. Histological analysis showed a portion of the intestine (“segment 1”) with grossly normal neuromuscular pattern (Figures 7A,D,G) consistent with that observed in a 26 weeks control fetus (Figure 7C). However, another portion (designated as “segment 2”) showed massive overgrowth of the smooth muscle layers (∼140 mm in the mispatterned “segment 2,” compared to ∼20 mm in “segment 1,” and 26 weeks control) and the mucosa (∼25 mm in the mispatterned “segment 2,” compared to ∼6 mm in “segment 1,” and 26 weeks control). The smooth muscle overgrowth in “segment 2” corresponded to an increase in thickness of 6.5 times (Figures 7B,E,H). Additionally, immunostaining for TuJ1 showed enteric neurons scattered throughout the gut wall (Figure 7E), in a pattern similar to that observed in the talpid³ chicken model (Figures 1O–R). The striking similarities found in histological observations were confirmed by combining equivalent measurements from human and chicken intestinal tissues for human/chicken cross-species analysis. This analysis showed a statistically significant increase in (i) ENS cell numbers (40.8%,^**p = 0.004) and (ii) epithelium and smooth muscle tissue thickness (77%,^***p = 0.00006 and 413%, ^****p < 0.0001, respectively) in mutant tissues compared to stage matched controls (Figure 7I). We also investigated expression of human chondroitin sulfate proteoglycans using CS56 (Figures 7F–H). 19 and 26 weeks controls showed expression of CS56 in the serosa the submucosa and the mucosa (Figure 7F and Supplementary Figure 5B). In “segment 1,” CS56 expression was also observed in the serosa and the mucosa whereas, strikingly, in “segment 2” no CS56 expression was detected (Figures 7G,H). To investigate further the intestinal phenotype of this fetus, we examined the expression of both Patched and Cytokeratin (Figures 7J–O). Even though cytokeratin was expressed in the fetus bearing the KIAA0586 mutation, the staining was not as strong and uniform as in the control. Likewise, Patched staining was reduced in “segment 1” of the fetus bearing the KIAA0586 mutation compared to the control, whereas “segment 2” showed no staining. The altered patched expression was confirmed by in situ hybridization (Supplementary Figures 4C–E). Both epithelial stainings pointed to delayed/altered gut epithelial differentiation. Overall, the phenotype observed in “segment 2” of the fetus was strikingly similar to that of the talpid³ chicken model with scattered enteric neurons, smooth muscle and mucosa overgrowth as well as impaired differentiation. Other defects observed in both human and the chicken model included: (i) altered Hh pathway (as shown by a lack of Patched expression and loss of mRNA expression pattern) (ii) impaired epithelium differentiation (as shown by Cytokeratin) and (iii) lack of CSPG components of the ECM.

We specifically examined the role of TALPID3 in neuromuscular patterning of the developing GI tract, as it was clear from the histological defects in both chicken and human GI tissues that lack of TALPID3 led to severe disruption of this developmental process. Our analysis showed that the lack of TALPID3 consistently affected the neuromuscular patterning of the gut. Interestingly though, the resulting phenotypes varied, depending on the location along the A–P axis of the gut. In the chicken model, lack of TALPID3 in the esophagus led to complete absence of smooth muscle and ENS, whereas in the stomach and the intestine lack of TALPID3 led to ectopic smooth muscle differentiation and misplaced enteric neurons. Likewise, one portion of the human intestine from a fetus bearing a KIAA0586 mutation showed grossly normal neuromuscular patterning, whereas another had dramatic muscle overgrowth and scattered enteric neurons. These observation, when combined with quantifications of enteric neuronal, smooth muscle and epithelial cell numbers, suggest that, albeit with species differences, TALPID3 is part of a conserved mechanism controlling the appropriate spatial differentiation of smooth muscle and the correct positioning of enteric neurons. Moreover, TALPID3 functions as part of regional-specific mechanisms regulating the correct neuromuscular patterning at different levels of the GI tract. Perhaps this is not surprising since the spatiotemporal development of GI smooth muscle has been shown to have regional-specific differences (Bourret et al., 2017; Graham et al., 2017). Our results are in accordance with recent findings in the related chicken mutant talpid2 (Brooks et al., 2021).

Our findings add to the body of evidence linking TALPID3 to the Hh signaling pathway (Davey et al., 2007; Bangs et al., 2011; Ben et al., 2011; Alby et al., 2015; Li et al., 2017). The link between TALPID3 and Hh signaling in the gut can be observed in (i) the gross anatomy of the GI tract (ii) the smooth muscle and (iii) ENS defects of the mutants. (i) First, the gross anatomical abnormalities in the TALPID3 chicken GI tract are consistent with a loss of Hh signaling. Severe reduction in GI tract size with normal gross anatomy has been reported using a conditional approach to remove both Shh and Ihh functions from early mouse gut endoderm (Mao et al., 2010). Likewise, mouse knockouts of Shh, Gli2 and Gli3 all display tracheo-esophageal atresia/fistula, a phenotype also described here in the TALPID3 chicken mutant (Motoyama et al., 1998; Ramalho-Santos et al., 2000). Additionally, regulation of gut epithelium homeostasis has been linked to the Shh signaling pathway and both talpid³ chicken and KIAA0585 human gut tissue showed altered epithelium growth/differentiation (Mao et al., 2010; Ben-Shahar et al., 2019). Lastly, Hedgehog signaling is critical for normal anorectal development and anorectal malformations have been reported in shh knockout mice and in humans with polymorphisms in Hedgehog genes (Ramalho-Santos et al., 2000; Gao et al., 2016). The open hindgut phenotype of the talpid³ chicken model offers an additional model for researching such malformations. (ii) It has been shown that Shh signaling regulates the concentric architecture of the intestine. In particular, SHH affects the mesenchyme immediately adjacent to the epithelium, where it restricts smooth muscle differentiation to allow the connective tissue of the submucosa to develop (van den Brink, 2007; Huycke et al., 2019). The increased domain of smooth muscle differentiation in the stomach and intestine of the talpid³ chicken suggests that this Hh-dependant regulation is lost. Moreover, the results from ptch in situ hybridization in talpid³ mutants directly demonstrate that SHH expressed by the endoderm is not correctly integrated throughout the radially surrounding mesenchyme. The lack of PATCHED protein immunostaining in “segment 2” suggests that this part of the gut is unable to respond to SHH signaling. The loss of the proximal SHH readout in the talpid³ mutants can be directly correlated with the subsequent differentiation of smooth muscle adjacent to the epithelium in the stomach and intestine. This result demonstrates that mesenchymal cells surrounding the epithelium are competent for induction into smooth muscle, but are normally prevented from doing so by the proximal SHH gradient via a TALPID3-dependent mechanism. In the esophagus, the lack of TALPID3 does not lead to an extension of the smooth muscle differentiation domain, but rather to a lack of smooth muscle differentiation altogether, pointing to a different region-specific readout of the Hh signal. This failure of smooth muscle differentiation in the esophagus may explain why migrating ENCC do not halt their migration and differentiate into enteric neurons, as they lack target tissue to innervate. Indeed, ENCC were observed at the level of the esophagus at stage E6.5 in the talpid³ chicken mutant, but no cells were present in this region at E8.5. TALPID3 is therefore necessary for the correct integration of SHH throughout the gut mesenchyme and to define the concentric architecture of the smooth muscle differentiation domain. Our results fit in with a recent study showing that Hedgehog acts through Bmp signaling to inhibit subepithelial smooth muscle and that levels of Hedgehog signaling regulate differentiation of the inner smooth muscle layer (Huycke et al., 2019). This suggests that TALPID3 is part of this regulatory mechanism. (iii) The influence of SHH on NCC migration and proliferation is well established, but many paradoxical results point to a complex role for Hh during ENS development with species and developmental differences, as well as direct and indirect regulatory mechanisms (Ramalho-Santos et al., 2000; Fu et al., 2004; Reichenbach et al., 2008; Biau et al., 2013; Jin et al., 2015). Recent results clearly show, both in chicken and mouse, that the SHH receptor Patched is not expressed by ENCC and point to indirect regulatory mechanisms (Nagy et al., 2016). Our findings are in agreement with an indirect regulation of ENCC by SHH through its role in modifying the gut ECM and thereby the environment through which ENCC migrate, as we discuss below.

In search of a mechanistic explanation for the lack of ENS plexus formation in talpid³ mutants, we analyzed expression of CGPG molecules, which are components of the gut ECM. Expression of ECM components is known to be regulated by both primary cilia and Hh signaling (Seeger-Nukpezah and Golemis, 2012; Nagy et al., 2016), making them good candidates for further investigation. Alterations in ECM expression have also been described in human ciliopathies (Ramalho-Santos et al., 2000; Seeger-Nukpezah and Golemis, 2012). Importantly, CSPG molecules provide guidance cues for neuronal behaviors such as migration, axon outgrowth and axon termination (Ring et al., 1996; Carulli et al., 2005; Siebert et al., 2014). We specifically investigated Coll9, a CSPG expressed in the developing gut, which has been shown to elicit avoidance behavior by NCC in vitro (Ring et al., 1996; Nagy et al., 2016). In correlation with the loss of Hh signaling in the GI tract of both the talpid³ chicken and the fetus bearing a KIAA0586 mutation, expression of CSPG molecules was lost. Importantly, double staining of CSPG and NCC showed that the disappearance of the CSPG expression correlated tightly with ectopic localization of ENCC, suggesting that the lack of repellent molecules such as Coll9 is a direct mechanism underlying the lack of ENS plexus formation in talpid³ GI tract. This finding is in agreement with previous studies showing that the regulation of CSPG by SHH in both chicken and mouse models can modify the behavior of ENCC and subsequently ENS patterning (Nagy et al., 2016). To investigate a possible connection between smooth muscle differentiation and the expression of CSPG, we performed double immunofluorescence with SMA and coll9 antibodies. We found that SMA and Coll9 have distinct, yet partially overlapping, patterns of expression. This demonstrates that some myoblasts express Coll9. However, Coll9 is mainly expressed in mesenchymal cells and it is this wider CSPG expression domain, and not the SMA domain, that correlates best with the regional localization of ENCC. Additionally, our chicken transplantation experiments demonstrated the inability of wild type ENCC to rescue ENS formation in a TALPID3-null gut environment. This finding highlights the role of environmental cues in directing ENS formation and shows that ENS plexus development does not rely on self-organizing properties of ENCC. This is in agreement with recent work showing that disruption of the ENCC environment can disrupt ENS patterning (Graham et al., 2017). Our study also demonstrates that TALPID3 expression in the gut is essential for proper expression of guidance cues that direct ENS plexus patterning. Importantly, we also show that this role is conserved during human fetal gut development.

We investigated the cell autonomous requirement for TALPID3 during ENCC migration and gut patterning by knocking out Talpid³ specifically in ENCC, either by neural tube transplantation in chicken or using a Wnt1-Cre driven Talpid³ conditional knockout in mouse. In both models, TALPID3 was found not to be required for ENCC migration or normal ENS plexus formation as the gross morphology of the ENS was unaffected. The only alteration to the ENS in these conditions was the increase in cell number observed in the chicken model and in the mouse model, which we quantified in stomach sections. Alteration of NCC numbers could be the result of subtle migration or differentiation defects changing the relative number of ENCC in specific locations. Alternatively, it could be linked to a TALPID3-dependent alteration of the SHH pathway and, specifically, the influence of SHH on NCC as a mitogen (Fu et al., 2004; Reichenbach et al., 2008; Roper et al., 2009; Nagy et al., 2016). Remarkably, knocking out Talpid³ in ENCC led to wider non-cell autonomous defects in other tissues such as smooth muscle and epithelium. Both the mouse and chicken phenotypes revealed a neural crest TALPID3-dependent mechanism controlling growth and differentiation of mesenchyme and epithelium. The most striking effect was alteration of smooth muscle differentiation and increased numbers of myoblasts which we quantified in stomach sections in chicken and mouse. The alteration of smooth muscle differentiation was particularly evident in the stomach region of mice. Variations in epithelium shape and cell numbers were also evident both in mouse, chicken. Mouse/chicken cross-species analysis confirmed that these differences in smooth muscle and epithelial cell numbers were statistically significant when comparing chimeric animals to controls. Interestingly, in addition to its role in smooth muscle development, the Hh pathway also plays an important role in mucosal growth (Mao et al., 2010; Ben-Shahar et al., 2019). Indeed shh and gli3 mutant mice have mucosal hyperplasia in the stomach (Ramalho-Santos et al., 2000; Kim et al., 2005). It has been previously postulated that vagal ENCC could act as a mediator in the mesenchymal-epithelial interactions that control stomach development (Faure et al., 2015). Additionally it has been shown that ENS influences the differentiation and growth of other cell types within the gut, as demonstrated by smooth muscle overgrowth in the aganglionic portion of EdnrB mice, as well as alteration of goblet cell differentiation (Spencer et al., 2007; Thiagarajah et al., 2014). Recent work has also shown that implantation of neural crest cells within tissue-engineered small intestine altered the transcriptome of a wide variety of gastrointestinal cell types, demonstrating the necessity of the neuronal lineage to be able to recapitulate gut organogenesis in vitro (Schlieve et al., 2017). Our findings emphasize the importance of the neural component for correct smooth muscle and mucosa development both during embryonic gut development and for regenerative medicine research. Moreover our study shows the central role played by TALPID3 in neuronal-mesenchymal-epithelial interactions necessary for normal GI tract development.

The neural crest specific talpid³ knockout results described here are in agreement with non-cell autonomous effects seen in other tissues after NCC-specific gene knockout. For example, knocking out the intraflagellar protein Kif3a in NCC led to non-cell autonomous striated muscle defects during tongue development (Millington et al., 2017). Additionally, loss of NCC disrupts the distribution of second heart field cells in the pharyngeal and outflow regions (Bradshaw et al., 2009). Likewise, conditional neural crest Rac1 knockout, using a Rac1/Wnt1-Cre line, shows excessive proliferation of SMA+ cell wall around the aortic sac and ventral aorta (Waldo et al., 2005; Thomas et al., 2010). Investigating the gut phenotype of Rac1/Wnt1-Cre mutants for smooth muscle defects would be informative. Interestingly, Rac1 is downstream of the non-canonical SHH pathway (Mulligan, 2014). TALPID3 and Rac1 have a comparable function in vesicular trafficking (Stenmark, 2009), centrosome regulation (May et al., 2014) and in cell-matrix interactions (Thomas et al., 2010). Rac1 is also downstream of RET, a tyrosine kinase receptor essential for ENS development (Natarajan et al., 2002; Fu et al., 2010; Mulligan, 2014).

Having established that TALPID3 was not required cell autonomously for ENCC migration and gross ENS patterning, and considering the role of TALPID3 in ciliogenesis, it follows that primary cilia might not be required for ENCC migration. ENCC are a highly proliferative cell population and it is their mitogenic activity, which is driving their migration and invasiveness (Simpson et al., 2007). It is known that there is a reciprocal regulation of cilia and the cell cycle, so that the primary cilium is dismantled in replicating cells, which makes it unlikely for highly proliferative cells to extend a primary cilium (Ishikawa and Marshall, 2011; Ford et al., 2018). Interestingly, migrating interneurons within the developing murine brain do not show an extended primary cilium (Higginbotham et al., 2012). In our study, we failed to observe primary cilia in migrating chicken vagal ENCC using in vitro neural tube culture. Performing Immunohistochemistry on tissue sections at E6.5, migrating ENCC bearing a primary cilium were seldom observed (11% of HuC⁺ ENCC). It has been shown that differentiated ENS neurons bear a primary cilium (Junquera Escribano et al., 2011; Luesma et al., 2013). Other studies describe primary cilia on YFP+ cranial NCC using the Wnt1:Cre; Rosa:YFP transgenic line (Brugmann et al., 2010; Millington et al., 2017). In this transgenic line, YFP is expressed in both the neural tube (which is ciliated) and the neural crest derivatives. This lack of discrimination is problematic to decipher this issue (Cassiman et al., 2006; Murdoch and Copp, 2010). Our study suggests that extending primary cilia mainly happens as ENCC start to differentiate into enteric neurons and might not be required for migration per se.

KIAA0586 is the human ortholog of talpid³. Some homozygous mutations of KIAA0586 have been shown to be embryonic lethal and to lead to severe developmental abnormalities such as Hydrolethalus syndrome and short-rib polydactyly (Alby et al., 2015). Here we show that the phenotypic spectrum of KIAA0586 mutations extends to defects in the GI tract. Apart from a tubular stomach the gross anatomy of the GI tract of the fetus was normal. However, our histological analysis showed severe alteration of the gut patterning with smooth muscle and mucosa hyperplasia, as well as scattered enteric neurons in a sub-section of the intestine we designated “segment 2.” In accordance with the animal models, we show that this phenotype is the consequence of disruption of the Hh pathway and loss of normal ECM expression. Our findings shed light on the central role of KIAA0586 in patterning of the gut during human fetal development. Recently KIAA0586 mutations have also been identified in Joubert syndrome patients. JBTS is defined by three primary findings: (i) underdevelopment of the brain cerebellar vermis, (ii) brain stem defects giving the appearance of the molar tooth sign (MTS) and (iii) Hypotonia (Abdelhamed et al., 2013; Sanders et al., 2015; Stephen et al., 2015). Importantly also, patients typically have a perturbed respiratory pattern in the neonatal period and severe psychomotor delay. Although very rare, association with Hirschsprung disease and problems with bladder and bowel control (incontinence) have been reported in Joubert patients with milder forms of the disease (Shian et al., 1993; Ozyurek et al., 2008; Akhondian et al., 2013; Purkait et al., 2015). Considering the role of KIAA0586 in tissue patterning our study unfolds, it is also tempting to speculate that other symptoms of JBTS (like psychomotor delay, hypotonia and respiratory difficulties), could be linked to neuronal-mesenchymal-epithelial patterning defects in other parts of the body. Interestingly, in this study, Wnt1:Cre; Talpid3 ^fl/fl P0 mutants died at birth from presumed respiratory failure, as their lungs did not appear to inflate. This points to a possible failure in the developmental interactions of neural crest, smooth muscle and epithelium, as lung neural crest cells share developmental origins with ENCC (Burns and Delalande, 2005; Freem et al., 2012). Finally, the ENS is often referred to as the “second brain” and it is possible that brain ECM defects, similar to the ones we describe in this study, could underlie abnormal brain patterning defects, such as cortical heterotopias, commonly seen in Joubert syndrome, but also in Bardet–Biedl syndrome or Meckel–Gruber syndrome (Willaredt et al., 2008; Abdelhamed et al., 2013; Higginbotham et al., 2013). In mice, a large variety of CSPGs represent major components of the ECM in the brain (Horii-Hayashi et al., 2015). In human, defects in ECM, leading to impaired neuronal guidance, could underlie other types of brain function abnormalities. A recent CNS conditional Talpid³ knockout mouse model for JBTS showed cerebellar defects due to granule cell proliferation, migration and differentiation, three aspects of cellular behavior also affected in our models (Bashford and Subramanian, 2019). This study, however, did not investigate ECM components.

Fertile chicken eggs were obtained from commercial sources within the UK. talpid³ and transgenic (http://topics.sciencedirect.com/topics/page/Green_fluorescent_protein) GFP chicken eggs were obtained from The Roslin Institute, The University of Edinburgh (McGrew et al., 2004; Davey et al., 2006). Eggs were incubated at 37°C and staged according to the embryonic day of development (E), and by using the developmental tables of Hamburger and Hamilton (Hamburger and Hamilton, 1951). The talpid³ mutation leads to leaky blood vessels causing very high mortality (Davey et al., 2006). This high mortality meant that specimens were collected over more than 10 years. Our study presents oldest time point at which we could reliably analyse gut tissues (E6.5 for most of the study). Beneficial outcrossing of the Talpid flock improved survival over time. It became possible to analyse later time points, when the gut is more developed, as shown in Figure 1 with E8.5 and E10.5 specimens.

For each combination of grafting experiment (chick^GFP- talpid³; talpid³-chick^GFP; chick^GFP- wild type), the neural tube adjacent to somites 2–6 inclusive (and its associated neural crest) was microsurgically removed from the host embryos at embryonic day E1.5 and replaced with equivalent stage-matched tissue, as previously described (Burns and Le Douarin, 2001; Delalande et al., 2015). Following grafting, eggs were returned to the incubator, and embryos allowed to develop to the appropriate stage. To ascertain the genotype of the Talpid³ neural tube transplant onto a GFP host embryo, the remaining tissues of the donor were used for DNA extraction, PCR amplification and sequencing of the region flanking the a366 mutation on exon7 using the following primers: Forward: CATTAGCTCTGCCGTCAACA Reverse: GGTAGGCAGACCACTGGAAG (Figure 5) (Davey et al., 2006). A total of 30 GFP neural tube grafts into TALPID3 hosts were made, of which 2 homozygote talpid³ mutant hosts were fixed for analysis at E6.5. After series of “blind” grafting experiments, a total of 5 talpid³ neural tube grafts onto GFP hosts were identified after genotyping, of which only 1 chimera survived and was fixed for analysis at E7.5.

Animals used for this study were maintained and the experiments were performed in accordance with local approvals and the United Kingdom Animals (Scientific Procedures) Act, 1986 under license from the Home Office (PPL70/7500). Talpid3^f/f mice were acquired from Prof. Malcolm Logan (King’s College London). Wnt1-cre;R26R-YFP/YFP mice (Srinivas et al., 2001; Druckenbrod and Epstein, 2005), in which NCC express yellow fluorescent protein (YFP), were crossed to Talpid3^f/f to generate a neural crest conditional knockout of Talpid³. Subsequent Wnt1:Cre; Talpid3^flox/flox mouse tissues were examined at postnatal day 0 (P0).

Human material was sourced via the Joint MRC/Wellcome Trust Human Developmental Biology Resource (HDBR) under informed ethical consent with Research Tissue Bank ethical approval (08/H0712/34 + 5 and 08/H0906/21 + 5) (Gerrelli et al., 2015). Staging of embryos was carried out according to the Carnegie system. GI tissues from a fetus with Short-rib polydactyly and bearing a homozygous null mutation in KIAA0586 were obtained from case II:5 family 4, as previously described in Alby et al. (2015). Informed consent was obtained for all participating families, and the study was approved by the ethical committee of Paris Ile de France II.

Transverse sections were cut from whole embryos and dissected gut segments. Sections were obtained at a thickness of 6 μm from wax blocks and 10–15 μm from frozen blocks. All sections were placed on Superfrost Plus microscope slides (BDH Laboratories).

For labeling of chicken and human dissected GI samples, the tissues were cryoprotected in 15% sucrose in PBS and frozen in liquid Nitrogen. Frozen sections were cut at 12 μm using a Leica CM1900 cryostat at −22°C. Briefly, for immunofluorescence, antibody blocking solution (10% sheep serum, 1% Triton-X-100 in PBS) was applied for 1 h at room temperature then samples were rinsed extensively in PBS and incubated primary antibodies (listed in Table 1) diluted in antibody blocking solution overnight at 4°C. Samples were then washed three times in PBS for 20 min and incubated with fluorescently tagged secondary antibodies (listed in Table 2) for 4 h at room temperature. Samples were washed for 1 h, and stained for 10 min with DAPI, before being mounted under a coverslip using Vectashield mounting medium (Vector Laboratories) as previously described (Wallace and Burns, 2005; Delalande et al., 2014) in situ hybridization was performed as previously described (Burns and Delalande, 2005).

Tissues were imaged using confocal microscopy (Zeiss LSM 710 confocal microscope). Images of gut sections double immunostained with relevant tissue antibodies and DAPI. Equivalent fields of relevant sections were examined in controls, mutants and/or chimeras (minimum n = 3). Cells were counted and tissue thickness measured using the Image-J Fiji cell counter plugin and measuring tool, respectively. Data was plotted to a histogram or interval plot. The results for all samples were normalized to baseline. The outcome was set as the percentage of the difference between the baseline and the experimental conditions. Multiple linear regression was used to study the effects of tissue, species and gene mutation simultaneously (Schneider et al., 2010; Lunt, 2013). This allowed us to assess the relative contribution of each predictor to the total variance that our model explained. Interaction terms were also considered with stricter p-value cut offs. The models were validated using the adjusted R squared and by performing residual analysis. To assess possible tissue rescue in the GFP > taplid³ transplantation experiment, the non-parametric Spearman’s Rho test was used to measure the correlation between control, chimera and mutant. All raw data and statistical analysis calculations are available here: https://datadryad.org/stash/share/xNqbJ7teCQMoatb4CDKgsdbtIPGZ5QH1eGBvz_g-fVY.

Migrating vagal NCC for scanning electron microscopy imaging were obtained using in vitro neural tube cultures as previously described (Delalande et al., 2008). Samples were then fixed overnight in 2% glutaraldehyde, 2% paraformaldehyde in 0.1 M phosphate buffer, pH7.4, at 4°C, post-fixed in 1% OsO₄/1.5% K₄Fe(CN)₆ in 0.1 M phosphate buffer at 3°C for 1.5 h. After rinsing with 0.1 M phosphate buffer and distilled water, specimens were progressively dehydrated to 100% ethanol, then washed once in acetone. The samples were then critical point dried using CO₂ and mounted on aluminum stubs using sticky carbon taps. Samples were then coated with a thin layer of Au/Pd (2 nm thick) using a Gatan ion beam coater and imaged with a Jeol JSM-6480LV high-performance, Variable Pressure Analytical Scanning Electron Microscope.

Wnt1:Cre; Talpid3^fl/fl P0 mice and control littermates were fixed in 90% ethanol, then skinned and eviscerated. Staining with Alcian Blue (0.05%) was performed in 70% ethanol with 20% acetic acid, followed by staining with Alizarin Red (0.15%) in 1% KOH. Soft tissue was cleared with 1% KOH with 20% glycerol and skeletons were stored in 80% glycerol.