Canonical Wnt signaling regulates soft palate development by mediating ciliary homeostasis

Organ morphogenesis is a complex process involving homeostatic cellular activities as well as reciprocal tissue-tissue interactions with signals relayed by secreted factors (Strand et al., 2010). The soft palate is an anatomical structure with unique functions, and its embryonic development involves interplay among cranial neural crest (CNC)-derived mesenchymal cells, mesoderm-derived myogenic cells and ectoderm-derived epithelial cells, making it an excellent model for unraveling the intricate morphogenetic tissue-tissue communications during craniofacial development (Grimaldi et al., 2015; Li et al., 2019; Sugii et al., 2017). Of particular interest is the role of CNC-derived mesenchymal cells and factors derived from them, which mediate communication with myogenic precursors during early craniofacial muscle development. The majority of these factors, mechanistic links and molecular crosstalk are yet to be discovered (Noden and Francis-West, 2006; Sambasivan et al., 2011; Sugii et al., 2017; Ziermann et al., 2018).

Cleft lip with or without cleft palate, which occurs in 1/700 live births, is one of the most common birth defects and significantly affects quality of life (Dixon et al., 2011). The mechanism underlying cleft soft palate is less understood than that of cleft hard palate, but soft palate clefting imposes a significant burden on individuals and healthcare providers, and may require treatment from infancy through adulthood (Wehby and Cassell, 2010). The soft palate is part of the oropharyngeal complex with important functions in swallowing, breathing, hearing and speech, which are disrupted in individuals with clefts. Soft palatal functions are not sufficiently re-established in 30% of cleft patients after surgery (Marrinan et al., 1998; Monroy et al., 2015; Von den Hoff et al., 2018). Owing to the delicate structure of the soft palate, its functional restoration requires long-term multidisciplinary treatment (Monroy et al., 2012).

Studies have identified Dlx5, Fgf10, Mn1, Tbx22, Runx2 and Twist1 as important regulators of soft palate development (Han et al., 2021; Liu et al., 2008; Sugii et al., 2017). Wnt signaling is of particular interest for further investigation as it is a known mediator of TGF-β signaling during soft palate development (Iwata et al., 2014) but its mechanisms of action in this context have not yet been elucidated. Furthermore, comprehensive screening of expression patterns related to signaling pathways potentially involved in soft palate development indicate that Wnt signaling is predominantly expressed in CNC-derived cells surrounding the myogenic soft palatal cells, making it a strong candidate for mediating communication between them (Janeckova et al., 2019). In general, Wnt signaling is crucial for craniofacial development in humans and mice, and its disruption causes severe craniofacial defects, including cleft palate (Brault et al., 2001; Chen et al., 2009; He et al., 2011; Huelsken et al., 2000; Reynolds et al., 2019). Independently, Wnt signaling has also been shown to play a role during various phases of myogenesis (Suzuki et al., 2018; Zhong et al., 2015).

In the present study, using the soft palate as a model, we highlight how canonical Wnt signaling acts upstream of ciliogenesis. Primary cilia serve as essential mediators of the complex tissue-tissue interactions accompanying soft palate development. We show that persistence of the primary cilia, caused by their dysfunctional disassembly in the CNC-derived mesenchymal cells due to a lack of canonical Wnt signaling, prevents cell cycle progression. This reduces the number of proliferating CNC-derived mesenchymal cells, influences their differentiation status and ultimately results in underdevelopment of the soft palatal muscles. Our study emphasizes the connection between Wnt signaling and ciliogenesis in regulating tissue-tissue interactions to control organogenesis.

To identify how cell signaling regulates soft palate development, we screened several signaling pathways, including Wnt, Fgf and Hh (Janeckova et al., 2019). Wnt signaling emerged as a strong candidate as it was primarily activated in the CNC-derived mesenchymal cells surrounding the mesoderm-derived myogenic cells during the early development of the soft palate. This corroborated a previously published study, in which it was shown that mesenchymal Wnt signaling mediates TGF-β epithelial signaling to guide myogenesis in the soft palate (Iwata et al., 2014).

To test whether canonical Wnt signaling is functionally required for regulating tissue-tissue interactions in soft palate development, we deleted β-catenin, the key mediator of Wnt signaling, in the CNC-derived mesenchymal cells, without affecting Wnt signaling in epithelial or myogenic cells in Osr2-Cre;β-catenin^fl/fl mice (Chen et al., 2009) (Fig. 1A,B,E,F). We confirmed efficient reduction of β-catenin expression in the CNC-derived mesenchymal cells (Fig. 1C,D,G,H). In contrast, β-catenin expression was not affected in the epithelial cells, which are not targeted by Osr2-Cre (Fig. 1C,D,G,H). Loss of β-catenin in palatal CNC-derived mesenchymal cells resulted in defects in the palatine process of the maxilla and the palatine bone in Osr2-Cre;β-catenin^fl/fl mice (Fig. 1I,J). Furthermore, the histology revealed complete cleft palate (Fig. 1K,L). The hard palatal shelves failed to grow horizontally towards each other and to fuse together. These phenotypes resulted in postnatal lethality and signified that canonical Wnt signaling in the CNC-derived mesenchymal cells is indispensable for palatogenesis.

We then analyzed the soft palatal defects in Osr2-Cre;β-catenin^fl/fl mice. The levator veli palatini (LVP) was our focus as the main soft palatal muscle (Monroy et al., 2015). Intraoral whole-mount views and soft tissue CT scans showed complete cleft palate in Osr2-Cre;β-catenin^fl/fl mice, in comparison with intact well-developed control palates (Fig. 2A-D). Histological analysis further confirmed that the soft palatal shelves were not developed at E18.5 in the Osr2-Cre;β-catenin^fl/fl mice (Fig. 2E-H). The LVP was significantly affected and no myogenic fibers were detected (Fig. 2I-L). To determine when the muscle phenotype first appears in these mice, we analyzed embryos at different stages and found that the palatal shelf development and the number of MyoD⁺ positive cells at the level of LVP were comparable between controls and mutants at E13.5 (Fig. 2M-P,U), suggesting there were no migration defects affecting the arrival of these myogenic cells at the palatal shelves. The difference in the number of MyoD⁺ positive cells became apparent in Osr2-Cre;β-catenin^fl/fl mice relative to controls at E14.0 (Fig. 2Q,S,V) and persisted at E14.5 (Fig. 2R,T,W). Therefore, we considered E13.5-E14.0 as the time of phenotype onset, and we used samples at these stages for further analyses.

Balanced cellular activities, such as proliferation and apoptosis, are crucial for mesenchymal cells during soft palate development. In order to determine why soft palatal shelf growth was stunted in Osr2-Cre;β-catenin^fl/fl mice, we analyzed cellular activities of CNC-derived mesenchymal cells at E13.5 and E14.0, when the palatal shelves were still comparable between controls and mutants at the level of the LVP. At this stage, the palatal shelves consisted predominantly of CNC-derived mesenchymal cells with only a few myogenic cells. Our analysis showed decreased proliferation of soft palatal mesenchymal cells in Osr2-Cre;β-catenin^fl/fl mice at E13.5 (Fig. 3A,B,D,E,I). However, apoptosis was not affected; very few cells were apoptotic in controls or mutants at E13.5 (Fig. 3C,F). To uncover any changes in cell cycle progression between control and Osr2-Cre;β-catenin^fl/fl mice, we colocalized the S-phase marker BrdU with the M-phase marker pH3 to identify cells that progress from S- to M-phase during the short BrdU labeling. Our analysis revealed fewer BrdU⁺/pH3⁺ cells in the palatal shelves of Osr2-Cre;β-catenin^fl/fl mice at E14.0 (Fig. 3G,H,J,K,L). Therefore, fewer S-phase cells progressed into M-phase in Osr2-Cre;β-catenin^fl/fl mice. These results indicated that Wnt signaling has a specific role in maintaining appropriate proliferation and cell cycle progression of CNC-derived mesenchymal cells in the soft palate.

To uncover how canonical Wnt signaling regulates cell proliferation in the soft palatal mesenchyme, we performed RNA sequencing (RNAseq) analysis of Osr2-Cre;β-catenin^fl/fl and control mice at E14.0 (Fig. 4A-C) to identify significant changes in the transcriptome and downstream targets in the absence of Wnt signaling in soft palatal mesenchymal cells. We identified significantly affected genes associated with myogenesis and ciliogenesis in the soft palate of Osr2-Cre;β-catenin^fl/fl samples compared with controls (Fig. 4D). The decrease in soft palatal myogenesis in Osr2-Cre;β-catenin^fl/fl mice indicated that Wnt signaling in CNC-derived mesenchymal cells is required for myogenic cells through tissue-tissue interactions, consistent with the defects observed in histological analysis (Fig. 2E-L). A striking change in Osr2-Cre;β-catenin^fl/fl mice was increased ciliogenesis (Fig. 4D). Primary cilia are complex, unique organelles that play an important role in regulating organogenesis and function as cellular antennas (Wheway et al., 2018). The potential involvement of primary cilia in soft palate development has not been explored. To comprehensively validate the ciliary phenotype, we analyzed ciliary structure, intraflagellar transport and post-translational modifications of cilia.

To analyze the ciliary structure in Osr2-Cre;β-catenin^fl/fl mice, we used γ-tubulin (a marker of the ciliary basal body) and Arl13b (a marker of the ciliary axoneme). We identified increased expression of γ-tubulin and Arl13b in the palatal mesenchyme of Osr2-Cre;β-catenin^fl/fl mice in comparison with controls (Fig. 4F-N). As cilia are dynamically regulated during cell cycle progression (Kasahara and Inagaki, 2021) and our analysis showed that palatal mesenchymal cells are not progressing to M-phase in Osr2-Cre;β-catenin^fl/fl mice, we analyzed the RNAseq data with a focus on genes that are essential for ciliary stability and homeostasis (the balance between assembly and disassembly). One of the genes we identified is Ttll3, which encodes a glycylase that serves as a post-translational modifier of the ciliary axoneme (Gadadhar et al., 2017). Increased expression of Ttll3 was confirmed through RNAScope in situ staining of soft palatal mesenchyme (Fig. 4O-R). Another post-translational modifier, Ttll6, also showed increased expression in the palatal mesenchyme of Osr2-Cre;β-catenin^fl/fl mice (Fig. S1A-D). Furthermore, we discovered that the expression of Ift88, which is involved in ciliary intraflagellar transport, was also increased in the palatal mesenchyme of Osr2-Cre;β-catenin^fl/fl mice (Fig. 4S-V). In summary, in the absence of canonical Wnt signaling in mesenchymal cells, myogenesis and ciliogenesis are adversely affected in Osr2-Cre;β-catenin^fl/fl mice.

After identifying Ttll3 as the key factor responsible for dysfunctional ciliary disassembly in Osr2-Cre;β-catenin^fl/fl mice, we further analyzed this gene. To uncover how Ttll3 expression is controlled by canonical Wnt signaling, we searched its promoter region for potential binding sites of canonical Wnt-related transcription factors Tcf7, Tcf7l1, Tcf7l2 and Lef1, all of which contain a β-catenin-binding domain (Cadigan and Waterman, 2012). Binding of β-catenin to these transcription factors in the nucleus enables activation or repression of Wnt-responsive genes (Ip et al., 2012; Kim et al., 2017; Ohsugi et al., 2017). Bioinformatic analysis using transcription-binding profiles from the UCSC/JASPAR database showed that Tcf7 and Tcf7l2 can potentially bind to the Ttll3 promoter region (Fig. 5A), indicating that Ttll3 may act downstream of canonical Wnt signaling to control ciliary disassembly on mesenchymal cells and ultimately influence soft palate development.

To determine whether Tcf7 or Tcf7l2 presented a better candidate for further analysis, we analyzed our scRNAseq data from E13.5 soft palatal shelves (Fig. 5B) (Han et al., 2021) using the R package SCENIC (Aibar et al., 2017). This allowed us to identify the main genetic regulatory networks (regulons) in early soft palate development to guide identification of the transcription factors that play biological roles in specific clusters identified by scRNAseq (Aibar et al., 2017). Our data revealed that the Tcf7l2 regulon, not that of Tcf7, is present in perimysial cells (Fig. 5C), a population that closely interacts with myogenic progenitor cells during early development of the soft palate and is essential for further myogenesis (Han et al., 2021). These data underscore the importance of canonical Wnt signaling in early soft palate development and point to Tcf7l2 being an important transcription factor during this process. We identified a motif (chr6113388069-113388082) that enables binding of Tcf7l2 to the Ttll3 promoter region (Fig. 5D). We used a CUT&RUN (cleavage under targets and release using nuclease) assay (Skene and Honikoff, 2017) and confirmed direct binding of Tcf7l2 to the predicted region at the Ttll3 promoter site (Fig. 5E). These data suggest that Ttll3 is a direct target of canonical Wnt signaling in the soft palate, mediated by Tcf7l2.

Next, we examined the expression of Tcf7l2 in vivo. Tcf7l2 was expressed primarily in the palatal mesenchymal cells in controls, and this expression was decreased in Osr2-Cre;β-catenin^fl/fl mice (Fig. 5F-I). To confirm whether Tcf7l2 is responsible for changes in Ttll3 expression, we performed in vitro analysis on primary soft palatal mesenchymal cells isolated from E13.5 control soft palates. Following siRNA-mediated knockdown of Tcf7l2 in these cells (Fig. 5J), the expression of Ttll3 was significantly increased (Fig. 5K), suggesting that Tcf7l2 represses Ttll3 expression in soft palatal mesenchymal cells. In summary, Tcf7l2 can functionally regulate Ttll3 expression by binding to its promoter.

To support our findings, we used siRNA targeting β-catenin and confirmed its knockdown efficiency (Fig. 5L). Similar to Tcf7l2 siRNA, decreasing β-catenin caused increased Ttll3 expression in soft palatal mesenchymal cells (Fig. 5M). Combined knockdown of β-catenin and Tcf7l2 (β-catenin siRNA:Tcf7l2 siRNA=1:1) increased Ttll3 expression (Fig. 5N), similar to β-catenin and Tcf7l2 alone (Fig. 5K and M), suggesting the collaboration of β-catenin and Tcf7l2 in regulating Ttll3 expression. Thus, in Osr2-Cre;β-catenin^fl/fl mice, decreased levels of both β-catenin and Tcf7l2 in CNC-derived mesenchymal cells cause stabilization of cilia and prevent these cells from progressing through the cell cycle, whereas, in normal conditions, β-catenin and Tcf7l2 are able to repress Ttll3 for ciliary homeostasis. These results reveal the mechanism by which Wnt signaling regulates the precise balance of ciliary assembly and disassembly that is indispensable for soft palate development.

Our results suggest that the disturbed cell cycle progression of the palatal mesenchymal cells in Osr2-Cre;β-catenin^fl/fl mice is caused by dysfunctional ciliary disassembly, disrupting ciliary homeostasis. Therefore, we hypothesized that restoring the ciliary homeostatic balance could rescue cell proliferation in the soft palate of Osr2-Cre;β-catenin^fl/fl mice. To test this hypothesis, we treated soft palatal mesenchymal cells isolated from Osr2-Cre;β-catenin^fl/fl mice with Ttll3 siRNA and observed a decrease in ciliated cells, as shown by reduced staining of ciliary axonemes by Arl13b (Fig. 6A-E). This treatment also increased proliferation of CNC-derived soft palatal cells (Fig. 6A-D,F), suggesting that ciliary assembly can restrain their cell cycle progression. Moreover, Ttll3 siRNA treatment restored proliferation of CNC-derived palatal mesenchymal cells from Osr2-Cre;β-catenin^fl/fl mice to a level comparable with that of controls (Fig. 6G-K).

Furthermore, to investigate whether reducing Ttll3 expression could rescue the myogenic defect in Osr2-Cre;β-catenin^fl/fl mice, we optimized an in vitro system to culture 300 µm slices of embryonic heads (Feng et al., 2022) and microinjected Ttll3 siRNA into Osr2-Cre;β-catenin^fl/fl soft palatal shelf mesenchyme to restore the Ttll3 expression level. After 3 days in culture, we confirmed that Ttll3 expression decreased following Ttll3 siRNA microinjection in comparison with control siRNA (Fig. 6L-O). We observed an increase of MyoD⁺ cells in Osr2-Cre;β-catenin^fl/fl samples injected with Ttll3 siRNA in comparison with Osr2-Cre;β-catenin^fl/fl samples that were injected with control siRNA (Fig. 6L-P). Collectively, these results indicate that strict control of ciliary homeostasis is essential for proper proliferation, cell cycle progression of CNC-derived mesenchymal cells and development of the soft palate muscles.

To discover the mechanism of communication among CNC-derived mesenchymal and myogenic cells, we analyzed our E13.5 soft palatal scRNAseq data (Han et al., 2021) because the altered cell proliferation phenotype in Osr2-Cre;β-catenin^fl/fl mice appears immediately after this stage. We used CellChat to identify predicted ligand-receptor interactions between the CNC-derived mesenchymal and myogenic cells (Jin et al., 2021). Among the highly ranked predicted outgoing signaling patterns from the perimysial cells (clusters 2, 4 and 9) was Notch signaling (Fig. 7A), especially the Dlk1/Notch3 ligand and receptor combination (Fig. 7B), which represents non-canonical Notch signaling that could be received by the myogenic cells. Supporting this notion, we found decreased expression of Dlk1 in the palatal mesenchyme of Osr2-Cre;β-catenin^fl/fl mice (Fig. 7C-F). As primary cilia function as a signaling hub, we hypothesized that, apart from influencing the cell cycle and its progression, primary cilia could also be responsible for the differentiation status of the perimysial cells that are essential for the development of soft palatal muscles. Furthermore, in our scRNAseq analysis, Dlk1 was identified as one of the top three molecules enriched in the perimysial cells, and it can function both as a transmembrane and secreted protein, making it a good candidate for mediating the communication among CNC-derived mesenchymal and myogenic cells (Feng et al., 2022). After sufficiently decreasing the expression of Ttll3 using siRNA in the primary soft palatal mesenchymal cells in vitro (Fig. 7G), we observed increased expression of Dlk1 (Fig. 7H). These findings are consistent with our in vivo finding that overexpression of Ttll3 caused a decrease in Dlk1 expression.

To identify the functional requirement of Dlk1 for the myogenesis of the LVP, we also used the in vitro slice culture system (Feng et al., 2022) to culture slices from E13.5 control embryonic heads containing the LVP, followed by microinjection of control or Dlk1 siRNA. Although the MyoD⁺ precursors for LVP continued to develop in the soft palatal shelf injected with control siRNA (Fig. 7I-J), the number of MyoD⁺ precursors did not increase in the palatal shelves following Dlk1 siRNA treatment (Fig. 7K-M). Furthermore, the Dlk1 expression was restored in Osr2-Cre;β-catenin^fl/fl slice cultures injected with Ttll3 siRNA in comparison with those injected with control siRNA (Fig. 7N-Q), corresponding to a rescue of the number of MyoD⁺ cells (Fig. 6P). These experiments demonstrate that Ttll3 expression regulates the expression of Dlk1 in the soft palatal mesenchymal cells and ultimately myogenesis through cell-cell interaction. Therefore, in the soft palate, the homeostasis of primary cilia influenced by canonical Wnt signaling coordinates both the proliferation and cell cycle progression of the CNC-derived mesenchymal cells as well as the differentiation of perimysial cells, which ultimately drives soft palatal myogenesis (Fig. 8).

The soft palate is an essential part of the oropharyngeal complex, fulfilling vital functions, including breathing, hearing, swallowing and speech, all of which are affected in individuals with cleft palate (Monroy et al., 2015). Cleft soft palate is less commonly studied than cleft hard palate, yet it accompanies the majority of cleft cases, and functional restoration of the soft palatal tissue has a low surgical success rate (Monroy et al., 2015; Von den Hoff et al., 2018). Moreover, it imposes a significant burden on patients, families and healthcare providers, as multidisciplinary treatment may be required into adulthood (Von den Hoff et al., 2018). Clinically, individuals with soft palatal clefts exhibit a reduced number of myogenic cells, together with misorientation, lack of fusion, misalignment, atrophy and fibrosis of the soft palatal muscles (Li et al., 2019; Monroy et al., 2015). The detailed molecular mechanisms involved in regulating soft palate development still need to be elucidated, including the reciprocal interactions among the CNC-derived mesenchymal cells, mesoderm-derived myogenic cells, ectoderm-derived epithelial cells and mediators of these communication channels, such as the secreted factors conveying the message or special organelles that may be involved (Li et al., 2019). A better understanding of these signaling mechanisms may contribute to the development of improved, biologically informed soft palatal muscle restoration strategies.

After screening for expression patterns of several signaling pathways, we uncovered that Wnt signaling is activated primarily in the CNC-derived mesenchymal cells surrounding the myogenic cells during soft palatal muscle development. Although mutations in Wnt signaling genes are well described in orofacial clefting in humans, as well as mice, and this pathway is independently known to play a role in regulating myogenesis, here, we have focused on its function in the soft palate (Reynolds et al., 2019; Suzuki et al., 2018). Studying Wnt signaling in murine palatogenesis is otherwise very challenging due to embryonic lethality and failure of craniofacial development after conditional deletion of β-catenin using the more broadly expressed Wnt1-Cre driver (Brault et al., 2001). In the current study, loss of canonical Wnt signaling only in the CNC-derived palatal mesenchymal cells using the Osr2-Cre mouse model created an opportunity to study the role of Wnt signaling in regulating tissue-tissue interactions during soft palate development.

We discovered that canonical Wnt signaling is essential for proper soft palatal muscle development, and in particular, its regulation of ciliogenesis is indispensable for the proliferation, cell cycle progression and differentiation of CNC-derived mesenchymal cells. These findings agree with previously published data establishing that cilia must disassemble before cells enter the M phase of the cell cycle; otherwise, they function as a cell cycle brake (Liu et al., 2021). Indeed, in Osr2-Cre;β-catenin^fl/fl mice, the CNC-derived mesenchymal cells do not progress into M phase and exhibit increased expression of Ttll3, which encodes a glycylase required for ciliary stability (Gadadhar et al., 2017). Restoring ciliary homeostasis by knocking down expression of Ttll3 rescues these cellular defects in CNC-derived mesenchyme and ultimately rescues myogenesis in Osr2-Cre;β-catenin^fl/fl mice. Canonical Wnt signaling therefore influences ciliary post-translational modifications and disassembly, cell cycle progression of the CNC-derived mesenchymal cells, and consequently soft palatal muscle development through cell-cell interactions.

As canonical Wnt signaling is a complex pathway, it is important to determine which of its mediators is the furthest downstream, such that it can bind with β-catenin to directly regulate the expression of ciliary genes. Transcription factors Tcf7, Tcf7l1, Tcf7l2 and Lef1 each contain a β-catenin-binding domain and this binding determines the specificity of canonical Wnt signaling in activating or repressing canonical Wnt-responsive genes in various tissues and organs (Cadigan and Waterman, 2012; Ip et al., 2012; Kim et al., 2017; Ohsugi et al., 2017). We discovered that both β-catenin and Tcf7l2 can regulate genes related to ciliogenesis, which play important roles in regulating the integrity of the soft palatal shelves and muscle development. Moreover, Tcf7l2 is an important regulon in perimysial cells, further supporting the importance of canonical Wnt signaling in early soft palate development. These data are consistent with the previously reported role of Tcf7l2-mediated canonical Wnt signaling in early embryonic limb muscle development (Kardon et al., 2003). In addition, TGF-β-mediated reduction of Tcf7l2 occurs in the mesenchymal stromal cells but not in myoblasts, also supporting the role of Tcf7l2 in cells surrounding the muscle cells during myogenesis (Contreras et al., 2020).

Intriguingly, the relationship between Wnt signaling and ciliogenesis is unclear. To date, limited studies have reported Wnt signaling being upstream of ciliogenesis. It has been shown, on the basis of Kupffer's vesicle motile cilia in zebrafish (Caron et al., 2012) and in vitro experiments on human telomerase reverse transcriptase-immortalized retinal pigment epithelial cells (Kyun et al., 2020), that Wnt stimulation promotes primary ciliogenesis. The effects of cilia on Wnt signaling in the reverse direction are uncertain. After manipulating cilium-related genes, Wnt signaling may be restricted (Corbit et al., 2008) or unchanged (Ocbina et al., 2009), and it has also been proposed that cilia serve as a switch between canonical and non-canonical Wnt signaling (Simons et al., 2005). Our study sheds light on the intricate relationship between Wnt signaling and ciliogenesis. We demonstrate the direct involvement of the Wnt signaling pathway in the regulation of ciliary genes during soft palate development.

Furthermore, we discovered a previously unappreciated role of primary cilia in regulating soft palate development. Several studies have reported craniofacial phenotypes, including cleft palate, that are associated with ciliopathies, especially Joubert syndrome, Meckel-Gruber syndrome, Oro-facial-digital syndrome and Ellis-van Creveld syndrome (Brugmann et al., 2010b; Schock and Brugmann, 2017). Analysis of Arl13, which marks the ciliary axoneme, indicated an increased number of cilia in CNC-derived soft palatal mesenchymal cells in Osr2-Cre;β-catenin^fl/fl mice and its mutation is linked to Joubert syndrome that manifests with cleft palate (Cantagrel et al., 2008; Caspary et al., 2007; Poretti et al., 2012). Primary cilia have been shown to play a role in regulating hard palate development (Ift88, Kif3a and Tmem107) and skeletal muscle development (Brugmann et al., 2010a; Fu et al., 2014; Tian et al., 2017), but whether they are involved in soft palate development has not been reported. Our study highlights primary cilia as key mediators of soft palatal muscle development.

Cilia have also been suggested as key regulators for cell-cell interactions (Wang and Barr, 2018). Our analysis has shown that there is strict interdependence of individual signals and collective corresponding cellular responses, with primary cilia serving as the integration hub. Previously, cilia have been identified as key for maintaining the balance between proliferation and differentiation, e.g. in kidney epithelial cells; they fulfil this function through paracrine signaling (Boletta et al., 2000; Irigoin and Badano, 2011). In our model, primary cilia influence both the proliferation and differentiation of CNC-derived mesenchymal cells to establish and maintain interactions with myogenic cells during morphogenesis. In this study, Notch signaling has been identified as one of the top enriched signaling pathways involved in interactions between perimysial-derived signaling molecules communicating with the myogenic cells in soft palate development. Moreover, Dlk1 is one of the top three molecules enriched in palatal perimysial cells (Feng et al., 2022). We found decreased expression of Dlk1 in the soft palate of Osr2-Cre;β-catenin^fl/fl mice, suggesting changes in differentiation of perimysial cells. These changes ultimately lead to defective palatal muscle development. Our data agree with previously published work supporting the role of Dlk1 in skeletal muscle development and regeneration (Shin et al., 2014; Waddell et al., 2010).

Overall, our study points to the complexity and combination of signaling pathways, cellular processes and their interaction with organelles, such as cilia, coming together to orchestrate embryonic development of the soft palatal region. Canonical Wnt signaling influences the proliferation and ciliary homeostasis of soft palatal mesenchymal cells, enabling their differentiation and ultimately promoting proper muscle development. Our study describes primary cilia as the missing mechanistic link that relays diverse signals from CNC-derived mesenchymal cells to myogenic cells to enable the congruency of the whole system, including the interactions that initiate sustained differentiation of the CNC-derived mesenchymal cells and their functional diversity. These results represent important findings that ultimately will lead to more effective functional muscle restoration for individuals with orofacial clefts.

All animal handling was performed according to federal regulations with approval from the Institutional Animal Care and Use Committee (IACUC) at the University of Southern California (protocol 9320). C57BL/6J mice (Jackson Laboratory, 000664) were used for this study. Osr2-CreKI mice (Chen et al., 2009) (a gift from Rulang Jiang, Cincinnati Children's Hospital, OH, USA) were crossed with β-catenin^fl/fl (Brault et al., 2001) (Jackson Laboratory, 004152) and tdTomato conditional reporter mice (Jackson Laboratory, 007905) to generate Osr2-Cre;β-catenin^fl/+ males or Osr2-Cre;tdTomato^fl/+ embryos. The Osr2-Cre;β-catenin^fl/+ males were further crossed with β-catenin^fl/fl females to generate Osr2-Cre;β-catenin^fl/fl embryos. As control samples, either β-catenin^fl/+ or β-catenin^fl/fl were used. Pregnant females were euthanized by carbon dioxide overdose followed by cervical dislocation. Embryos (regardless of their sex) were collected at various developmental stages [embryonic day (E) 13.5, E14.0, E14.5 and E18.5, and postnatal day (P)0] and fixed overnight in 10% neutral buffered formalin solution (Sigma-Aldrich, HT501128). E18.5 and P0 samples were decalcified in ethylenediaminetetraacetic acid [EDTA (pH 7.1-7.3), Alfa Aesar, A15161) for 1 week and processed for cryosectioning or paraffin wax embedding and sectioning (8 μm).

MicroCT scans of P0 β-catenin^fl/fl and Osr2-Cre;β-catenin^fl/fl mice were obtained from a SCANCO mCT50 scanner at the University of Southern California Molecular Imaging Center with the X-ray source at 70 kVp and 114 mA, and all the data were collected at a resolution of 10 mM.

Paraffin wax-embedded sections were used for Hematoxylin and Eosin staining. The histological protocol is described in detail in our previous publication (Janeckova et al., 2019). For immunohistofluorescence, cryosections were air-dried at room temperature and then at 37°C for 15 min each. Distilled water was used for removing OCT (5 min wash) and slides were treated with 3% hydrogen peroxide solution (10 min), citrate-based antigen unmasking solution (Vector, H-3300-250) for 15 min at 99°C, blocking reagent (Thermo Fisher Scientific, B40922) for 1 h at room temperature and primary antibody overnight, followed by washes in PBST. Incubation with secondary antibody was performed for 2 h at room temperature with three subsequent washes in PBST and final counterstaining with DAPI (Thermo Fisher Scientific, D1306) for 5 min. Primary antibodies used in this study were as follows: rabbit monoclonal active β-catenin (Cell Signaling Technology, 19807S, 1:100 with TSA), mouse monoclonal, myosin heavy chain, MHC (DSHB, MF20, 1:10), rabbit polyclonal anti-phospho-histone H3, pH3 (Sigma-Aldrich, 06-570, 1:100), rabbit monoclonal cleaved caspase 3 (Cell Signaling Technology, 9664S, 1:100), rabbit monoclonal MyoD (Abcam, ab133627, 1:200 with TSA), rabbit polyclonal γ-tubulin (Sigma-Aldrich, T5192, 1:100), rabbit polyclonal Arl13b (Proteintech, 17711-1-AP, 1:100), rabbit polyclonal red fluorescein protein, RFP (Rockland, 600-401-379, 1:500 with TSA) and rat monoclonal BrdU (Abcam, ab6326, 1:100). The following secondary antibodies were used in this study: Alexa Fluor 488 anti-mouse (Thermo Fisher Scientific, A11001, 1:200), Alexa Fluor 568 anti-rabbit, (Thermo Fisher Scientific, A11011, 1:200) and anti-rabbit HRP (Vector Laboratories, PI-1000, 1:200) (Table S1). For active β-catenin, MyoD and RFP immunostaining, TSA Plus FITC or Cy3 (Akoya Bioscience, NEL7744001KT or NEL771B001KT) was applied for 3 min before DAPI counterstaining.

Cryosections were air dried at room temperature and at 37°C for 15 min each. Distilled water was used for removing the OCT (5-min wash) and slides were treated with a pre-heated target retrieval reagent (Advanced Cell Diagnostics, ACD, 322000) for 15 min at 99°C. Standard RNAScope reagents were used according to the manufacturer's instructions using RNAScope Multiplex Fluorescent Detection Kit v2 (Advanced Cell Diagnostics, 323110). The following probes from Advanced Cell Diagnostics were used in this study: Mm-Myod1 (316081), Mm-Ttll3 (586791), Mm-Ttll6 (570931), Mm-Ift88 (420211), Mm-Tcf7l2 (466901) and Mm-Dlk1 (405971).

A short 2 h pulse injection of BrdU (500 µg per 10 g body weight) was administered intraperitoneally at E14.0 to β-catenin^fl/fl and Osr2-Cre;β-catenin^fl/fl mice. EdU (10 mg/1 ml PBS) was used on cell cultures for 2 h before the fixation of the cells overnight in 10% neutral buffered formalin solution (Sigma-Aldrich, HT501128). EdU was visualized using a Click-iT Plus EdU Cell Proliferation Kit (Thermo Fisher Scientific, C10637) followed by DAPI staining (Thermo Fisher Scientific, D1306, 5 min).

The posterior part of the secondary palate (last third of the secondary palate, i.e. the region of the soft palate) was dissected into RNALater (Thermo Fisher Scientific, AM7020). Samples were stored at −80°C, and when sufficient control and mutant samples were obtained (minimum of n=3 per genotype), RNeasy Plus Micro Kit (Qiagen, 74034) was used to isolate RNA. UCLA Technology Center for Genomics and Bioinformatics performed determination of the RNA quality and subsequent cDNA library preparation and sequencing (single-end reads with 75 cycles, Illumina Nextseq 500 platform). Differential expression was calculated by selecting transcripts with fold change ≤−1.5 or ≥1.5 and a significance level of P <0.05 using PartekFlow. The RNA sequencing data generated in this study have been deposited in GEO (Edgar et al., 2002) under accession number GSE208619. The soft palatal tissue was also used for primary cell culture followed by qPCR analysis; after RNA isolation, cDNA was transcribed using an iScript cDNA Synthesis Kit (Bio-Rad, 1708891). The qPCR reaction was run using SsoFast EvaGreen Supermix (Bio-Rad, 172-5202) on a Bio-Rad CFX96 Real-Time System. Values were normalized to Gapdh using the 2^−ΔΔCt method. The primer sequences used in this study were obtained from PrimerBank (Wang et al., 2012) and produced by IDT (Integrated DNA Technologies) as follows: Gapdh (forward primer 5′-AGGTCGGTGTGAACGGATTTG-3′, reverse primer 5′-TGTAGACCATGTAGTTGAGGTCA-3′), Tcf7l2 (forward primer 5′-GCACACATCGTTCAGAGCC-3′, reverse primer 5′-GGGTGTAGAAGTGCGGACA-3′), Ttll3v2 (forward primer 5′-ATGGGCCGACTCAGAAACG-3′, reverse primer 5′-CGCAAGAGACACCGGATCA-3′), Dlk1 (forward primer 5′-CCCAGGTGAGCTTCGAGTG-3′, reverse primer 5′-GGAGAGGGGTACTCTTGTTGAG-3′) and β-catenin (forward primer 5′-ATGGAGCCGGACAGAAAAGC-3′, reverse primer 5′-CTTGCCACTCAGGGAAGGA-3′).

The soft palate from E13.5 embryos was dissected and digested using Multi Tissue Dissociate Kit 3 (Miltenyi Biotec, 130-110-204) at 37°C in a Thermomixer (600 rpm for 15 min, Thermo Fisher Scientific, 2231000269). After adding the medium to stop the digestion process, filtering through a 70 μm Pre-Separation Filter (Miltenyi Biotec, 130-095-823) and centrifuging at 4°C (300 g, 5 min), cells were seeded on a 24-well plate (0.5×10⁵ cell/well) in growth medium (DMEM medium, Thermo Fisher Scientific, 10569010; fetal bovine serum, VWR, 97068-085; 1% penicillin and streptomycin, Thermo Fisher Scientific, 15140122). The 24-well plates were collagen coated (Millipore Sigma, 125-50) before plating the cells. For siRNA treatment, Tcf7l2 Silencer Select Pre-designed siRNA (Thermo Fisher Scientific, 4390771; s74838), Ttll3 Silencer Select Pre-designed siRNA (Thermo Fisher Scientific, 4390771; s97693), β-catenin Silencer Select Pre-designed siRNA (Thermo Fisher Scientific, 4390771; s63417) and Silencer Select Negative Control #1 siRNA (Thermo Fisher Scientific, 4390843) were delivered by using Lipofectamine RNAiMAX Transfection Reagent (Thermo Fisher Scientific, 13778-075) according to the manufacturer's instructions. For β-catenin and Tcf7l2 double siRNA treatment, a half dose of each siRNA was used. The cells were treated with siRNA for 72 h before further analysis.

Embryonic mouse heads were dissected and embedded on ice in a mixture of preheated gelatin and Accell siRNA delivery culture medium (Horizon, B-005000-500) and sectioned into 300 µm slices using MicroSlicer Zero 1N (Ted Pella, 10111N). These slices were placed on 0.4 µm Millicell cell culture inserts (30 mm, hydrophilic PTFE, Sigma-Aldrich, PICM03050) which were inserted into 6-well culture plates (VWR, 10062-892). Micro-injection of control siRNA (Accell Control siRNA – Green; Horizon; K-005000-G1-02 or Accell Control siRNA – Red; Horizon; K-005000-R1-02), Accell Mouse Ttll3 siRNA, SMARTPool (20 nmol; Horizon; E-063939-00-0020) or Accell Mouse Dlk1 siRNA, SMARTPool (20 nmol; Horizon; E-043253-00-0020) was performed at a concentration of 100 μM. Cultures were maintained at 37°C in a 5% CO₂ incubator for 3 days and the culture medium was changed every day.

The UCSC/JASPAR bioinformatics database was used to search the promoter regions of individual genes in order to identify the binding profiles of transcription factors related to canonical Wnt signaling. The predicted Tcf7l2-binding sites to the promoter region of Ttll3 gene were identified at chr6:113388069-113388082. The following primers were used for the CUT&RUN assay: forward, TTGAACATCATGTGGCCCAGT; reverse, CACTGTAAACCAGGGACTGCT. The soft palatal tissue from E13.5 embryos was collected and fixed according to the manufacturer's instructions based on CUT&RUN assay kit (Cell Signaling Technology, 86652). 10 μl of IgG isotype control (Cell Signaling Technology, 3900) or Tcf7l2 (Thermo Fisher Scientific, MA5-14935) were used for each reaction. Embryos from two litters were used for each replicate.

Previously generated E13.5 soft palatal tissue single cell analysis data (Han et al., 2021) were processed using Seurat R package. RunPCA and RunUMAP visualization were used for cluster visualization. The data are available in GEO under accession number GSE155928. This dataset was processed with R package SCENIC to identify gene regulatory networks (Aibar et al., 2017). GENIE3 was used to identify the individual transcription factors that were later scored by AUCell. R package CellChat (Jin et al., 2021) was used to recognize the predicted ligand-receptor interactions between individual cell populations in the soft palate. The CellChat workflow was used with standard parameters (pre-processing functions: identifyOverExpressedGenes, identifyOverExpressedInteractions and projectData, core functions: computeCommunProb, computeCommunProbPathway and aggregateNet). netVisual_bubble and netAnalysis_signalingRole_heatmap were used to visualize the predicted results.

Immunofluorescence images were captured using a Leica DMI3000 B research microscope. Bright-field images were captured using a Keyence BZ-X710 system at the Center for Craniofacial Molecular Biology, University of Southern California.

GraphPad Prism Software (Prism 9) was used for statistical analysis. We used n=3 for all experiments unless otherwise stated. Each data point in the bar graphs represents one biological replicate. For each qPCR experiment, at least three technical replicates were analyzed. For histological staining, immunofluorescence and RNAScope, we used samples from at least three individual mice, or embryos, and representative images are shown. All bar graphs display mean±s.e.m. Statistical comparisons were carried out using an unpaired t-test with two-tailed calculations. One-way ANOVA was used for analyzing data for comparison of four independent groups. P<0.05 was treated as statistically significant for all analyses.

We are grateful to Bridget Samuels for critical reading of this manuscript, and to Kimi Nakaki and Giselle Mejia for the schematic illustrations. We acknowledge USC Libraries Bioinformatics Service for supporting with analysis of the sequencing data. The bioinformatics software and computing resources were funded by the USC Office of Research and the Norris Medical Library.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.201189.reviewer-comments.pdf