Temporospatial sonic hedgehog signalling is essential for neural crest-dependent patterning of the intrinsic tongue musculature

The mammalian tongue is a highly specialised muscular organ situated in the oral cavity and oropharynx that contributes to multiple essential functions, including airway maintenance, phonetic articulation, oral sensation, mastication and swallowing. Normal function of the tongue requires the coordinated activity of both extrinsic and intrinsic muscles and their associated tendons, an extensive vasculature and complex sensory and motor innervation. The morphological and functional variation seen in the tongue of mammals is reflective of the significant environmental adaptation observed amongst these organisms (Iwasaki, 2002).

In the mouse embryo, tongue formation is heralded by the appearance of paired buds on the oral side of the first (mandibular) pharyngeal arch around embryonic day (E)10.0. These buds grow and ultimately fuse with a medial lingual swelling to produce an early primordium situated in the midline of the oral cavity floor by E11.0. The primordium is initially populated by cranial neural crest cells (CNCCs) (Han et al., 2012); however, myogenic progenitor cells subsequently enter this region from E11.5 following their migration from the caudal occipital somites as the hypoglossal cord (Czajkowski et al., 2014; reviewed by Noden and Francis-West, 2006). Further development of the tongue requires interaction between the overlying oropharyngeal epithelium, CNCCs and myogenic progenitors to regulate cell proliferation, differentiation and survival. By E13.5, a prototype tongue structure is established, with a symmetrically arranged intrinsic musculature and a specific group of extrinsic muscles suspending the tongue between the skull, palate, mandible and hyoid – all supported by the midline lingual septum and peripheral aponeurosis linguae tendons within the tongue dorsum (Standring, 2016). Although the gross embryological processes involved in forming the tongue have been understood for many years (Sadler, 2012) it is only relatively recently that some insight has been gained into the complex molecular signalling events that coordinate development (Parada et al., 2012).

Sonic hedgehog (SHH) is a secreted signalling protein that plays a key role in many diverse biological events extending from early development through to postnatal tissue homeostasis (reviewed by Briscoe and Thérond, 2013; Ingham and McMahon, 2001; Ingham et al., 2011; McMahon et al., 2003). SHH signals from three key regions of ectoderm in the early head, which includes the ventral forebrain, frontonasal process and oropharynx (reviewed by Helms et al., 2008; Marcucio et al., 2011; Petryk et al., 2015; Tapadia et al., 2005; Xavier et al., 2016). Specifically, Shh is expressed from E9.5 in the pharyngeal endoderm and oropharyngeal epithelium (Billmyre and Klingensmith, 2015; Firulli et al., 2014; Xu et al., 2019) with this expression maintained in the primordial tongue epithelium before localising to fungiform papillae of the anterior tongue from E12.5 (Jung et al., 1999; Sagai et al., 2009) (Fig. S1), where SHH plays a key role in regulating lingual epithelial cell fate (Castillo et al., 2014; Miura et al., 2014). Blocking SHH in embryonic rat pharyngeal explant culture arrests development of the early tongue primordium (Liu et al., 2004) and Nkx2.5-Cre;Shh^flox/flox mice lacking SHH from the early pharyngeal endoderm and oropharyngeal epithelium have aglossia and micrognathia (Billmyre and Klingensmith, 2015). Moreover, Gas1^−/−; Boc^−/− mice with reduced SHH signalling secondary to loss of co-receptor function have a diminutive and cleft tongue (Seppala et al., 2014). SHH signal transduction to CNCCs is essential for tongue development. Wnt1-Cre; Smo^n/c and Wnt1-Cre; Kif3^flox/flox mice lack SHH responsiveness in CNCCs from their point of migration through abrogated function of the essential smoothened (SMO) transducer or the primary cilium, respectively, and both have aglossia (Jeong et al., 2004; Millington et al., 2017). In addition, Hand2-Cre; Smo^c/c embryos lacking SMO function in the mandibular arch from around E9.5 display a tongue that is only rudimentary and associated with significantly reduced numbers of muscle precursor cells (Xu et al., 2019). Collectively, these studies have demonstrated a role for SHH transduction within CNCCs during establishment of the early tongue primordium, although the local contribution of SHH during subsequent development of this organ is less clear.

The long-range cis-regulatory enhancers mammal fish conserved sequence 4 (MFCS4) and mammal reptile conserved sequence 1 (MRCS1) located 600-900 Kb upstream of the mouse Shh locus drive regional Shh expression in the epithelial lining of the oral cavity and pharynx (Sagai et al., 2017, 2009). MFCS4 is conserved between mammals and teleost fishes (Sagai et al., 2009) and is an orthologue of the human-fugu conserved sequence SHH_1 (Goode et al., 2005; Woolfe et al., 2005), whereas MRCS1 is conserved amongst mammals, chicken and lizard but not Xenopus or teleost fishes (Sagai et al., 2009). MFCS4 directs Shh expression in epithelium of the primordial tongue, pharynx and larynx from E11.0, with strong expression in the epiglottis and arytenoid swelling by E13.5, whereas MRCS1 regulates expression in the incisor and molar teeth at E12.5, and the palatal rugae and anterior fungiform papillae at E13.5. Targeted deletion of MFCS4 in mice (MFCS4^−/−) results in severely downregulated Shh expression in epithelia of the pharyngeal and laryngeal regions, leading to truncation of the soft palate, slight deformation of the tongue, loss or reduced size of the epiglottis and hypotrophy of the arytenoid accompanied by hypoplasia of multiple cartilaginous laryngeal elements (Sagai et al., 2009). However, some hedgehog signalling activity does remain in anterior regions of the tongue in these mice, consistent with the predominantly posterior pharyngeal and laryngeal phenotype, and suggestive of a requirement for SHH signalling in tongue development after establishment of the early primordium. These mice therefore represent a useful experimental tool for investigating the role of SHH during tongue formation.

In this study, we have utilised a series of genetic approaches to investigate the temporospatial contribution of SHH signalling during tongue formation, focusing on timing and the local effect on tissues responding to this pathway within the early tongue. We find that production of SHH ligand in tongue epithelium from E10.5 to E12.5 is crucial for lingual septum tendon formation and a prerequisite for normal patterning of the intrinsic musculature, occurring primarily through local signalling to CNCCs within the tongue primordium. Moreover, we provide further insight into the tongue phenotypes seen in humans with disrupted hedgehog signalling, including the ciliopathies, for which the relative contribution of local signalling interactions and global timing of developmental events during tongue development is currently unclear.

To explore the function of SHH signalling within the tongue primordium, we first utilised Shh^MFCS4/− mutant mice, in which one allele lacks MFCS4, a key cis-enhancer element for Shh expression in tongue epithelium activated at E11.0 (Sagai et al., 2009) and Shh is conventionally deleted in the other allele. It was predicted that an almost complete loss of Shh would occur in tongue and pharyngeal epithelia of the mouse, whereas the remaining tissues would be heterozygous for Shh expression. We confirmed that, in comparison to wild-type (WT) littermates, SHH signalling was significantly decreased in the developing tongue of Shh^MFCS4/− mice by E11.5, as demonstrated by reduced Shh and patched 1 (Ptch1) expression (Fig. 1A-D) and consistent with the known period of MFCS4 activity (Sagai et al., 2009). In addition, we also observed slightly decreased Gli1 expression in the mutant (Fig. 1E,F) (Millington et al., 2017). At this stage of development, Myf5-positive occipital somite-derived myoblasts had reached the developing tongue primordia in both WT and Shh^MFCS4/− mice (Fig. 1G,H), confirming that myoblast migration into the tongue primordium was not affected in these mice and in contrast to mice with disrupted SHH signalling in CNCCs through loss of primary cilia (Millington et al., 2017). The presence of myoblast differentiation was also confirmed in WT tongue through the detection of Myod1 and desmin in anterior and posterior regions at E11.5 (Fig. 1I-L). Interestingly, myoblasts present anteriorly expressed Myod1 but not desmin at this stage (Fig. 1I,K, anterior panels), whereas those located more posteriorly expressed both markers (Fig. 1I,K, posterior panels). Desmin is a marker of more advanced myoblast differentiation than Myod1, thus myoblast differentiation progresses from posterior to anterior during normal tongue development. The expression pattern of Myod1 and desmin was not altered in Shh^MFCS4/− mutants (Fig. 1J,L). At E12.5, analysis confirmed that SHH signalling was decreased but myogenic differentiation was not affected (data not shown). These observations were confirmed by quantitative RT-PCR (Fig. 1M).

At E13.5, the establishment of myotube organisation in both intrinsic and extrinsic muscles was indicated by the expression pattern of Myod1 and desmin in WT (Fig. 1N,P). In Shh^MFCS4/− mice, this arrangement was disrupted and distinct in the superior and inferior longitudinal, vertical and transverse intrinsic muscles, whereas in lateral regions where the bilateral extrinsic styloglossus muscles run, there was no significant disruption (Fig. 1O,Q). All Shh^MFCS4/− mutants died between postnatal day (P)0 and P0.5 with cleft palate, and at this stage the mutant also exhibited disorganised intrinsic muscle fibres based upon anti-smooth muscle actin (SMA) immunohistochemistry (Fig. 1R-U). Interestingly, the disorganised intrinsic muscle arrangement was rarely found in the tongues of MFCS4 homozygous-deleted mice (Shh^MFCS4/MFCS4) (Fig. S2).

The cleft palate observed in Shh^MFCS4/− animals was caused by a failure of palatal shelf elevation, which requires a combination of intrinsic elevation forces within the shelves and extrinsic factors, such as clearing of the tongue from the roof of the mouth due to involuntary and swallowing movements. Maxillary organ culture showed that, although the soft palate was shorter in an anterior-posterior direction, the palatal shelves of Shh^MFCS4/− foetuses successfully elevated and were able to fuse in the midline (Fig. S3). Based on these data as well as previous studies on tongue obstruction leading to cleft palate (Iseki et al., 2007; Tsunekawa et al., 2005), we hypothesised that decreased SHH signalling in the tongue results in disorganised intrinsic muscle arrangement, which leads to cleft palate. This idea was supported by the observation that the penetrance of cleft palate was limited in Shh^MFCS4/MFCS4 mice [13.5% in a previously published study (Sagai et al., 2009) versus 100% in Shh^MFCS4/−].

The finding of some residual Shh expression in the Shh^MFCS4/− tongue after the stage at which MFCS4 was activated (E11.0) suggested the presence of other factors regulating Shh in this region. Thus, use of the Shh^MFCS4/− mutant in this study allowed us to examine the consequences of severely decreased epithelial Shh expression combined with a decrease of SHH signalling within the tongue primordium.

CNCCs within the developing tongue have been suggested to act as a scaffold during the organisation of lingual myoblasts and myotubes (Parada et al., 2012). Patterning of the tongue musculature occurs simultaneously with tendon development and includes formation of the midline lingual septum tendon within the tongue dorsum as a flat broad vertical sheath of midline fibrous tissue. We investigated expression of the tenocyte marker scleraxis (Scx) (Schweitzer et al., 2001) using section in situ hybridisation and found low-level expression from E11.5 in the posterior tongue, but not the anterior (Fig. 2A and data not shown), consistent with the pattern of muscle localisation in the WT. At E12.5, an M-shaped expression domain was seen (Fig. 2C), coinciding with the expression pattern seen in whole mounts (Fig. 2E). In contrast, there was little evidence of Scx transcripts in Shh^MFCS4/− mice at E11.5 (Fig. 2B) and only faint expression at E12.5 (Fig. 2D), which was confirmed in whole mounts (Fig. 2F). In addition to expression in the midline at the developing lingual septum tendon (Fig. 2G, double-headed arrow), a thin layer of bilateral widespread low-level expression was also found in the tongue dorsum at E13.5 in the WT, indicating early formation of the tendinous aponeurosis linguae (Fig. 2G, dotted line). Beneath the tongue, there was strong bilateral expression of Scx at sites of tendon formation associated with the paired extrinsic genioglossus muscles (Fig. 2G, arrowheads). In contrast, Shh^MFCS4/− mice failed to develop Scx-positive tendons for the intrinsic muscles at E13.5 (Fig. 2H), despite clear expression in the extrinsic genioglossus muscle tendons (Fig. 2H, arrowheads). Expression of Scx in the tongue dorsum was significantly decreased; however, low-level Scx transcripts were occasionally observed in lateral regions of the mutant tongue dorsum (Fig. 2F,H, arrows). As Shh expression and MFCS4 activation were found in the dorsal lingual epithelium, these observations suggested involvement of other factors in tenocyte differentiation on the lateral sides of the tongue. Detection of CD31 (Pecam1) and synaptophysin, markers for vascularisation and innervation, respectively, did not demonstrate any particular differences between WT and Shh^MFCS4/− mice at E13.5 (Fig. S4).

We further analysed CNCC differentiation in Shh^MFCS4/− mice through the expression of Sox9, a marker for CNCCs as well as a common representative transcription factor for chondrocyte, ligament cell and tenocyte differentiation (Mori-Akiyama et al., 2003; Spokony et al., 2002). Sox9 expression was transiently evident during the early stages of tongue development in the WT, with strong expression established in the future lingual septum-forming region by E12.5 (Fig. 2I, arrow); however, by E13.5 Sox9 expression had dissipated in the WT and been replaced by Scx marking the lingual tendon (data not shown). In contrast, Sox9 was only weakly expressed at E12.5 in the mutant (Fig. 2J). Collectively, these data suggest that in the process of tenocyte differentiation from CNCC-derived tongue primordium mesenchyme there is a transition stage, during which WT mesenchyme expresses both Sox9 and Scx, whereas the mutant fails to properly express either gene.

In accordance with the Scx expression pattern, type I collagen (Col1a1) and fibronectin (Fn1), which mark the extracellular matrices of tendon tissues, were clearly transcribed in WT tendon, including the aponeurosis and septum (Fig. 2K,M), whereas their expression in the Shh^MFCS4/− tongue was weak and not well patterned (Fig. 2L,N). We next studied cell proliferation in the tongue mesenchyme at E11.5 and E12.5, during which tenocyte specification occurs in CNCCs. However, there were no significant differences in proliferative activity of CNCC-derived mesenchyme or myoblasts during this period (Fig. 2O) and no differences in patterns of cell death (assayed through active caspase 3 detection; data not shown) in WT or mutant embryos.

The analysis of Shh^MFCS4/− mutant mice demonstrated an essential role for SHH signalling in the tongue primordium; however, Shh expression is present in the oropharyngeal epithelium before early tongue primordium formation and continues until a prototype intrinsic lingual muscle arrangement is established (see Fig. S1) (Sagai et al., 2009). Furthermore, some SHH activity is still present in Shh^MFCS4/− embryos, mediated through Shh transcription independently of MFCS4 activity (Fig. 1B,D,F). Therefore, we investigated the temporal requirements of SHH during tongue formation using stage-specific abrogation of gene function. Specifically, we crossed pCag-CreER^TM mice (Hayashi and McMahon, 2002) with a line harbouring a conditional (floxed) Shh^c allele (Dassule et al., 2000). Single maternal tamoxifen administration was carried out at E10.5, E11.5 and E12.5, respectively, with the tongue phenotype analysed at E14.5. In comparison with WT littermates (Fig. 3A,E), significant disorganisation of the intrinsic musculature was present within the tongue of mutant mice, which was grossly dependent upon the timing of signal loss (Fig. 3B-D,F-H). In mice treated with tamoxifen at E10.5, in comparison with WT the normal striated architecture of the intrinsic muscles was lost and tongue size reduced (Fig. 3A,B,E,F), whereas deletion at later stages resulted in a progressively less severe disruption of myogenic pattern (Fig. 3C,D,G,H; Fig. S5). In situ hybridisation for Myod1 suggested that myoblast differentiation had occurred in both the intrinsic and the extrinsic musculature, even in the most severely affected mutants (E10.5 treatment, Fig. 3J), but that organisation of the myotubes was defective (Fig. 3I-L). In comparison with WT, Scx expression was lost in the midline septum tendon and significantly downregulated in aponeurosis linguae-forming regions of the dorsum, whereas expression remained in the developing genioglossus and geniohyoid tendons of the mutant (Fig. 3M,N). Scx expression in the midline dorsum was progressively increased in foetuses exposed to later injections at E11.5 and E12.5 (Fig. 3O,P). Collectively, these data demonstrate an important timing-dependent role for SHH signalling after E10.5 in mediating normal tendon architecture and organisation of the intrinsic musculature of the tongue, with low levels of SHH signalling present in Shh^MFCS4/− mice sufficient to support gross growth of the tongue. Interestingly, examination of K14-Shh transgenic embryos, which overexpress Shh in the developing tongue epithelium from around E11.5 (Cobourne et al., 2009) revealed a gross distortion of tongue architecture at E15.5 but a seemingly normal arrangement of the intrinsic muscles within the body of the tongue (data not shown). Therefore, myoblast organisation and differentiation is able to occur in the presence of excess SHH signal from the tongue epithelium during development.

Given that myoblast migration and differentiation was not affected by decreased SHH signalling in the tongue after E11.0, but that tendon formation and myotube arrangement specifically were, we hypothesised that CNCCs were the direct recipients of SHH signals. We therefore investigated the spatial relationship between SHH-responding cells, CNCCs and myoblasts in the developing tongue (Fig. 4A-E). Specifically, we examined the tongue of Wnt1-Cre; R26R embryos, which constitutively express β-galactosidase in CNCCs (and their progeny) from their point of exit at the neural tube (Chai et al., 2000). At E12.5, X-gal staining for β-galactosidase revealed the distribution of CNCCs in the developing tongue (Fig. 4A-C), and the expression of Ptch1 on adjacent sections indicated that CNCC-derived mesenchymal cells were the primary target of SHH signal transduction during tendon formation and myotube arrangement (Fig. 4D). In contrast, Myf5 was expressed in a broadly complementary pattern, suggesting that mesodermal cells are not the direct target of SHH in the tongue (Fig. 4E). Interestingly, Gli1 showed a more uniform expression within the tongue, whereas the SHH co-receptor-encoding genes Gas1, Cdon and Boc were detected in the mesenchymal core, including myoblasts and more peripheral regions in domains consistent with a negative correlation between co-receptor expression and SHH signal levels (Martinelli and Fan, 2007) (Fig. S6).

The primary cilium is recognised as a key cellular organelle necessary for normal hedgehog reception and signal transduction within receiving cells (Huangfu and Anderson, 2005). Consistent with this, humans with ciliopathic loss-of-function orofaciodigital 1 (OFD1) mutations have multiple craniofacial anomalies, including cleft palate and tongue defects that involve clefting, cystic formation and hamartoma (Fenton and Watt-Smith, 1985). Animal models of Ofd1 also show a ciliogenesis defect and loss of SHH signal reception (Adel Al-Lami et al., 2016; Ferrante et al., 2006) and mice lacking function of the ciliary protein Kif3a in CNCCs have aglossia (Millington et al., 2017). We generated CNCC-specific deletion in Ofd1 using Wnt1-Cre; Ofd^fl/Y mice and found decreased levels of Ptch1 expression in CNCC-derived mesenchyme at E11.5 and 12.5 (Fig. 4F,G,I,J). In contrast, Mesp1-Cre; Ofd^fl/Y embryos, which have Ofd1 deletion in mesodermal cells from the onset of gastrulation (Saga et al., 1999) had relatively normal levels of Ptch1 expression in the tongue (Fig. 4H,K). Significantly, there was a complete loss of normal myotube arrangement and hypoglossia in Wnt1-Cre; Ofd^fl/Y embryos (Fig. 4L,M), whereas in Mesp1-Cre; Ofd^fl/Y mutants myotube arrangement was largely unaffected (Fig. 4N). These observations further suggested that muscle patterning in the tongue requires continued SHH signalling from epithelium to CNCCs after tongue primordium establishment.

Transforming growth factor β (TGFβ) signalling has previously been shown to mediate interactions between CNCCs and myoblasts during cell proliferation, differentiation and tongue muscle organisation (Han et al., 2014; Hosokawa et al., 2010) and TGFβ receptors are transcribed in developing tongue mesenchyme (Hosokawa et al., 2010; Millington et al., 2017), which suggests potential cross-talk between TGFβ and SHH signalling. We therefore examined transcription of the ligands Tgfb1, -2 and -3 at E12.5 in WT and Shh^MFCS4/− mice using section in situ hybridisation (Fig. 5A-F). Tgfb1 did not show any defined expression within tongue mesenchyme of WT mice, but Tgfb2 was mainly detected in the myoblasts and Tgfb3 in the lingual septum tendon (Fig. 5A,C,E). Although Tgfb1 did not show any expression in Shh^MFCS4/− mutants either, Tgfb2 and -3 were both downregulated in the intrinsic musculature (Fig. 5B,D,F). However, these changes were subtle and, although qPCR analysis confirmed this trend, the changes were not significant (Fig. S7). Therefore, we further investigated TGFβ signalling in pCag-CreER^TM; Shh^flox/flox mice using temporally controlled deletion of Shh. As SMAD2 and -3 are essential components of the TGFβ intracellular signal transduction pathway, their phosphorylation (pSMAD2/3) was examined by western blotting of tongues obtained from pCag-CreER^TM; Shh^flox/flox mice treated with tamoxifen at E10.5 and analysed at E13.5. Interestingly, pSMAD2/3 was significantly reduced in all mutants compared with controls (Fig. 5G). It has also been demonstrated that fibroblast growth factor (FGF) signalling contributes to tenocyte differentiation in other parts of the body (Brent et al., 2005; Sukegawa et al., 2000; Du et al., 2016). We found that Fgfr1 was expressed in the lingual septum tendon and aponeurosis linguae in WT mice; however, Fgfr1 transcripts were significantly reduced in Shh^MFCS4/− mice as revealed by both qRT-PCR comparison and in situ hybridisation (Fig. 5H) confirming the potential involvement of FGF signalling in tendon formation. Finally, we considered the possibility that WNT signalling could also be affected in our mutant models as previously reported (Zhu et al., 2017). However, neither RT-PCR analysis of the WNT target gene Axin2 in Shh^MFCS4/− mice at E12.5, nor mRNA in situ hybridisation for Axin2 at E14.5 in pCag-CreER^TM; Shh^flox/flox mice treated with tamoxifen at E10.5 showed any significant differences (Fig. S7).

Taken together, these data demonstrate that a loss of SHH or decreased SHH in the developing tongue epithelium from the time of tongue primordium establishment results in disrupted patterning of the lingual tendon and intrinsic tongue musculature, which is potentially mediated through TGFβ and FGF signalling.

Tongue anomalies are a common finding amongst the multiple defects that can affect the developing craniofacial region (Cobourne et al., 2019). In this study, we have investigated the temporospatial contribution of SHH signalling in the developing tongue through the analysis of multiple mouse mutants. During normal development, SHH signals from the early tongue primordium, pharyngeal and laryngeal epithelium to the underlying CNCCs and is required for normal formation of the lingual tendons and patterning of the intrinsic musculature. MFCS4^−/− mice have downregulated SHH in the pharyngeal and laryngeal regions with soft palate truncation, deformation of the posterior tongue and loss of the epiglottis, and Shh^MFCS4/− mice have an almost complete loss of SHH function in these regions and a corresponding lack of the lingual tendons and disorganised intrinsic muscle. In pCag-Cre; Shh^fl/fl mice engineered with loss of SHH from E10.5, there is absence of the lingual tendons, disorganised intrinsic muscles and severe hypoglossia. Moreover, Wnt1-Cre; Ofd^fl/Y mice with reduced SHH signal response in CNCCs have severe hypoglossia, clefting of the tongue and poorly organised muscular clumps within the dorsum. SHH from the tongue epithelium is therefore required for the specification and differentiation of CNCC-derived tenocytes, with formation of the lingual tendons providing a scaffold for myoblast cell populations and facilitating normal patterning of the intrinsic musculature. These findings are summarised in Fig. 6.

Shh is expressed in multiple regions of the early pharyngeal epithelium; global Shh signal disruption prior to tongue anlage formation leads to aglossia (Billmyre and Klingensmith, 2015) and early transduction to migrating CNCCs is essential for normal development (Jeong et al., 2004; Millington et al., 2017; Xu et al., 2019). However, local Shh expression is maintained in epithelium of the early tongue primordium and the significance of this signalling has remained unclear.

Here, we have shown that loss of SHH from E10.5 onwards causes significant disruption to the intrinsic musculature with hypoglossia, but not aglossia, demonstrating an essential temporal role for local signalling. We also demonstrate that SHH directly induces formation of the CNCC-derived lingual septum tendon, which is required for normal myotube arrangement within the intrinsic musculature. The crucial time point appears to be from E11.0, as expression of Shh in the epithelium via the MFCS4 enhancer is important for this process, with the removal of enhancer activity from one allele and Shh expression from the other resulting in severely decreased SHH signalling in CNCC-derived mesenchyme and defective tendon formation. Collectively, these results suggest that SHH from the oropharyngeal epithelium is required to induce tongue primordium formation before E10.5, with subsequent signalling from the oropharyngeal and tongue primordium epithelium required for differentiation of the lingual tendon during intrinsic tongue muscle patterning.

It is likely that some residual SHH signalling activity remains in Shh^MFCS4/− mice because Shh and Ptch1 expression was not completely lost (see Fig. 1A-D) and loss of Scx did not always occur. One possibility is that additional Shh enhancers or regulatory factors may compensate for loss of MFCS4 function; however, the other known oropharyngeal epithelium enhancer MRCS1 is unlikely to be responsible because it only activates Shh after E12.5 (Sagai et al., 2009) when expression of Scx has already been initiated in the tendon-forming area. Therefore, the precise regulatory mechanisms underlying Shh expression in the oropharyngeal epithelium remain to be fully elucidated (Anderson et al., 2014; Irimia et al., 2012). A slightly reduced tongue size was observed in pCag-CreER^TM; Shh^flox/flox mice with loss of SHH from E11.5, suggesting that the remaining signalling in Shh^MFCS4/− mice functions in tongue growth, which is supported by the observed cell proliferation assays (see Fig. 2O).

The contribution of SHH signalling to tendon formation has been reported in a variety of anatomical systems, but not the tongue (Schweitzer et al., 2001; Subramanian and Schilling, 2015). Tendons are specialised connective tissues that assemble musculoskeletal tissues and anchor force-generating muscles to the skeleton, which facilitates locomotion and mobility in vertebrates. Consistent with our findings, a common feature of early tendon formation is the relationship between SHH and FGF signalling. During chick axis tendon formation, Shh expressed from ventral midline structures, such as the floor plate and notochord, inhibits induction of the syndetome, the dorsolateral compartment of the somite from which tendon cells are generated. However, SHH indirectly induces Scx expression through activation of FGF in the dermomyotome, which promotes Scx transcription in the somite with FGF signalling required for differentiation of tenocyte precursors in mice (Brent et al., 2005). In the chick digestive system, expression of Scx in two tendon domains that develop in close relation to visceral smooth muscles also depends upon FGF signalling (Le Guen et al., 2009) and Shh expressed from the endoderm is involved in regulation of smooth muscle cell and tenocyte differentiation (Sukegawa et al., 2000). A recent study has reported that some FGFs are increased during the period of lingual tendon formation (Du et al., 2016), suggesting involvement of FGF signalling in lingual tenocyte differentiation. In support of this, we detected diminished expression of Fgfr1 in the lamina propria, future aponeurosis and lingual tendon of the Shh^MFCS4/− tongue. Thus, we suggest that the differentiation of tenocytes derived from CNCCs is positively regulated by SHH via FGF signalling.

It is established that normal CNCC function is required for tongue initiation through SHH signalling (Jeong et al., 2004; Millington et al., 2017). These investigations have shown that disrupting hedgehog responsiveness in CNCCs from their point of migration results in large-scale apoptosis within the first arch, with an accompanying failure of mesoderm-derived muscle precursor cell migration into the tongue anlage and aglossia. However, the role of SHH during regulation of subsequent developmental events, such as myoblast differentiation, is less clear. Pax3-positive muscle progenitors begin migration at E9.75 and reach the pharyngeal region subsequent to CNCCs by around E11.5 (Relaix et al., 2004). Our data point to a key step in tongue development whereby CNCCs receive epithelial SHH cues in the tongue primordium, which are essential for the normal organisation of intrinsic myogenic precursors, but not for myoblast differentiation. However, the expression of Shh and related pathway components (Fig. S6) suggests involvement in muscle formation and requires further investigation.

We considered a number of molecular pathways known to regulate early tongue formation that may act downstream of SHH. Disruption of canonical TGFβ signalling in CNCCs through loss of Tgfbr2 results in microglossia secondary to abnormal myogenic precursor cell proliferation and organisation via loss of Fgf10 (Hosokawa et al., 2010). The differentiation and proliferation of CNCCs is also disrupted (Hosokawa et al., 2010; Iwata et al., 2013), partially mediated by non-canonical TGFβ signalling acting through TGFβ-activated kinase (TAK1; Map3k7) (Song et al., 2013) and FGF/bone morphogenetic protein (BMP) signalling through ABL1 (Iwata et al., 2013). The disruption of Tgfβr1 (ALK5) in CNCCs also leads to severely disrupted tongue muscle formation via a lack of BMP4-mediated myogenic proliferation and Fgf4/6-mediated myogenic differentiation (Han et al., 2014). We found no significant change in Tgfb2 and -3 expression although non-significant changes in Tgfβr1 and -r2 have previously been reported in Wnt1-Cre; Kif3a^flox/flox mice (Millington et al., 2017). A change in pSMAD2/3 was detected through western blot in the tongues of pCag-CreER^TM; Shh^flox/flox embryos, which might conceivably be due to potential changes in BMP signalling rather than TGFβ. Indeed, hedgehog-SMO-GLI1 signalling is required for transcription of the Forkhead box genes Foxf1, -f2, -d1 and -d2 in CNCCs (Everson et al., 2017; Jeong et al., 2004; Millington et al., 2017) and negatively regulates BMP signalling in CNCCs through Foxf1 and Foxf2 during early patterning and survival of this cell population in the mandibular arch (Xu et al., 2019).

WNT signal transduction from the epithelium is also essential for tongue development (Lin et al., 2011). A loss of WNT secretion results in disrupted development of the epidermis, tenocytes and internal musculature, likely due to a requirement for signalling to underlying CNCC-derived connective tissue and muscle progenitor cells (Zhu et al., 2017). Interestingly, we did not observe altered Axin2 expression in Shh^MFCS4/− or pCag-CreER^TM; Shh^flox/flox mice suggesting that WNT signalling was not altered in the mutants used in this study. However, further investigation of any association between SHH and WNT signalling in tongue development is required.

As our hypothesis was that CNCCs are key recipients of SHH signalling within the developing tongue, we considered models of human OFD1 mutations that are predicted to result in neurocristopathies. OFD1 mutations lead to a spectrum of craniofacial phenotypes, including gingival frenulae, lingual hamartomas, cleft palate and, significantly, cleft and/or lobulated tongue (Franco and Thauvin-Robinet, 2016). Moreover, ablation of the Kif3a ciliary protein in CNCCs is associated with aglossia (Millington et al., 2017). Wnt1-Cre; Ofd^fl/Y embryos lacking Ofd1 function in CNCCs had a severely hypoplastic tongue and complete disruption of myotube arrangement. The Ofd1 gene is located on the X chromosome and encodes a component of the centrosome and basal body of primary cilia, a key mediator of SHH signalling (Satir et al., 2010). Impaired function of the cilium results in a variable effect on signal transduction, affecting not only SHH but also other pathways, such as WNT, depending upon the molecular context (Bangs et al., 2015). Again, although further investigation is required to clarify the cross-network interactions between SHH, FGF, TGFβ, BMP and WNT signalling during tongue formation, we note that the majority of phenotypes can be attributed to Shh expression from the epithelium. It is of further note that, in contrast to Wnt1-Cre; Kif3a^flox/flox embryos (Millington et al., 2017), Wnt1-Cre; Ofd^fl/Y embryos did not show aglossia but hypoglossia, which is suggestive that SHH signals can be transduced to some extent in these mice and may provide a useful tool for further analysis of SHH signalling in orofacial development.

Taken together, investigation of these temporospatial tissue-tissue interactions provides new insight into formation of the tongue, specifically the lingual septum tendon and intrinsic musculature, and reveal how localised signalling can influence gross structure and function of this highly adapted organ. This has relevance for further understanding of the molecular basis of multiple craniofacial disorders associated with tongue anomalies.

All animal experiments were approved by the Institutional Animal Care and Use Committees of Tokyo Medical and Dental University (0170238A), King's College London (PPL7007441 and PPL P8D5E2773, KJL), and National Institute of Genetics (28-7). MFCS4^+/− (Sagai et al., 2009), Shh^−/+ (Amano et al., 2009), pCag-CreER^TM (Hayashi and McMahon, 2002), Shh^flox/+ (Dassule et al., 2000), Wnt1-Cre (Danielian et al., 1997), Mesp1-Cre (Saga et al., 1999) and ROSA26 reporter (R26R) (Soriano, 1999) mice were maintained in a C57BL/6N background, whereas Ofd^fl/Y (Ferrante et al., 2006) were maintained in a CD-1 background. In compound heterozygote Shh^−/+; MFCS4^+/− mice (here referred to as Shh^MFCS4/−) the (MFCS4) long-distance Shh enhancer is deleted on one allele, whereas the (Shh) gene remains intact on the same allele. pCag-CreER^TM; Shh^flox/+ mice were mated with Shh^flox/+ mice, with pregnant mice receiving tamoxifen by intraperitoneal injection (75 mg/kg, equivalent to 3 mg per 40 g body weight) through the maternal body at the appropriate developmental stage. All comparisons were performed between the mutants above versus WT littermates.

Specimens were fixed with Bouin's solution for Haematoxylin and Eosin (H&E) staining or fixed in 4% paraformaldehyde (PFA) in PBS for embedding in paraffin or OCT compound (Sakura Finetek) for other histological analyses. Sections were taken at 5 μm (paraffin) or 12 μm (frozen) thickness. All paired images for comparison were derived from littermates and were representative of at least three independent experiments.

For immunohistochemistry, anti-desmin antibody (clone D33, 413651, Nichirei Biosciences) at 1:1 dilution, anti-smooth muscle actin (SMA) antibody (RRID: AB_476701, clone 1A4, A2547-100UL, Sigma-Aldrich) at 1:1000 dilution, anti-bromodeoxyuridine (BrdU) antibody (clone BMC9318, 11 170 376 001, Roche Diagnostics) at 1:100 dilution, anti-MYF5 antibody (RRID: AB_10744494, polyclonal, SAB4501943, Sigma-Aldrich) at 1:100 dilution, anti-CD31 antibody (RRID: AB_726362, polyclonal, ab28364, Abcam) at 1:100 dilution, and anti-synaptophysin antibody (RRID: AB_2198854, clone SY38, ab8049, Abcam) at 1:100 dilution were used. For visualisation, corresponding secondary antibodies from the Vectastain ABC Kit (RRID: AB_2336827, AK-5000, Vectastain) and diaminobenzidine (DAB) were applied, or corresponding fluorescence-conjugated secondary antibody was applied [RRID: AB_2630356, goat anti-rabbit IgG H&L (Alexa Fluor 488), ab150077 and RRID: AB_2576208, goat anti-mouse IgG H&L (Alexa Fluor 488), ab150113, both Abcam]. Haematoxylin was used to counterstain DAB-labelled sections and Hoechst 33342 was used to stain DNA in fluorescent sections.

For in situ hybridisation, specimens were hybridised in whole mounts or section using digoxygenin-labelled RNA probes specifically designed to be complementary to the partial mRNA of mouse Shh, Ptch1, Gli1, Myf5, Myod1, Sox9, Scx, ColIa1, Fn1, Fgfr1, Ptch2, Cdon, Boc, Gas1, Hhip, Tgfb1, -2, and -3, and Axin2, followed by incubation with anti-digoxygenin-AP conjugate. Nitro blue tetrazolium chloride (NBT)/5-bromo-4-chloro-3-indolyl phosphate, toluidine salt (BCIP) were used for visualisation. Section ³⁵S (Perkin Elmer) radioactive in situ hybridisation was carried out for Shh and Ptch1 as previously described (Cobourne et al., 2004). Light- and dark-field images of sections were photographed using a Zeiss Axioskop microscope and merged in Adobe Photoshop CS6. All commercial reagents for in situ hybridisation were purchased from Roche Diagnostics. Probe sequences are listed in Table S1.

For the detection of β-galactosidase activity using X-gal staining, sections were incubated with 5-bromo-4-chloro-3-indolyl-β-D-galactoside (X-gal) in phosphate buffer (pH 7.3) supplemented with 2 mM MgCl₂, 5 mM potassium ferrocyanide [K₄Fe(CN)₆-3H₂O] and 5 mM potassium ferricyanide [K₃Fe(CN)₆] at 30°C after fixation in 4% PFA. Nuclear Fast Red was used for counter staining.

The dissected maxilla (with vertically oriented palatal shelves) of E13.5 Shh^MFCS4/− or WT littermates were cultured in Dulbecco's Modified Eagle's Medium Nutrient Mixture F-12 (Sigma-Aldrich) and BGJb Medium (Life Technologies) for 48 h with continuous supply of 95% O₂+5% CO₂ at 37°C using a rotary culture system.

Sixty minutes before dissection, BrdU at 100 mg/ml was injected intraperitoneally into pregnant females at 10 mg/kg on the designated day. Every sixth of seven sections through the tongue primordium of a specimen were used for the cellular proliferation analysis. BrdU incorporation and MYF5 localisation were detected by immunohistochemistry. External cells scored as epithelial cells were confirmed by basement membrane staining. Myoblast cell lineage was determined as MYF5-positive cells, and MYF5-negative cells were considered to be CNCC derived. A proliferation index was calculated from the number of BrdU-positive cells divided by the total number of cells of each population and statistical significance was examined by two-tailed Student's t-test for three individual experiments for each genotype.

RNA was extracted from the tongues of Shh^MFCS4/− and WT littermate embryos at E12.5 using the Direct-zol RNA MiniPrep Kit (R2050S, Zymo Research) following the product protocol. mRNA (250 ng) was transcribed to cDNA by ReverTra Ace (TRT-101, Toyobo Life Science). Real-time PCR was performed with LightCycler 480 High Resolution Melting Master (04909631001, Roche Diagnostics). The expression was normalised to the β-actin gene and relative expression to the littermate WT was shown. Statistical significance was examined by two-tailed t-test for three individual experiments for each genotype. The primer sequences are listed in Table S2.

Whole tongue lysates were obtained from pCag-CreER^TM; Shh^flox/flox and WT littermate embryos in radioimmunoprecipitation assay (RIPA) buffer (Thermo Fisher Scientific) (n=3 for each). Equivalent amounts of protein lysate were run on a 4-12% gradient Norex gel (Thermo Fisher Scientific) and separated proteins electro-transferred onto a nitrocellulose membrane. Phosphorylated SMAD2/3 (pSMAD2/3) and β-actin (β-actin) were detected using the corresponding antibodies [phospho-SMAD2 (Ser 465/467)/SMAD3 (Ser 423/425) (D27F4) rabbit monoclonal antibody; β-actin (13E5) rabbit monoclonal antibody; RRID: AB_2631089 and AB_10694076, respectively, Cell Signalling Technology] both at 1:1000 dilution. Bands were analysed using the Chemidoc MP imaging system (Bio-Rad) and intensity was measured using Image lab software (RRID:SCR_014210, version 5.2.1). Band intensities were normalised against β-actin and data plotted on a histogram. The integrated volume of pSMAD2 and -3 bands were divided by that of β-actin band and statistically examined for comparison.

We thank Zoe Webster and Jane Sealby (Embryonic Stem Cell Facility, MRC Clinical Sciences Centre, Imperial College London) and Paul Sharpe (King's College London) for generation of K14-Shh embryos through microinjection. We also thank William Barrell for expertise with schematics.