Molecular profiling of the vestibular lamina highlights a key role for Hedgehog signalling

Odontogenesis proceeds through a series of characteristic stages, which are controlled by multiple molecular interactions between the epithelium and underlying mesenchyme. In mammals, the first sign of tooth development is the appearance of a band of thickened epithelium marked by the expression of Pitx2 and Shh (Yu et al., 2020) This band goes on to form the dental lamina (DL), from which the tooth germs will emerge. Lateral to the DL is a second epithelial lamina, known as the vestibular lamina (VL) in mammals. In humans, the VL and DL have a complex relationship during early development, with the two laminas merging and branching as they run along the jaw margin (Hovorakova et al., 2005; 2007). In the anterior parts of the lower jaw, the DL and VL share a common origin, emerging from the same Shh-positive epithelial thickening in both mouse and human (Hovorakova et al., 2007, 2016; Qiu et al., 2020). As these structures diverge in development, Shh signalling is associated with the forming tooth germ, but turns off in the VL (Hovorakova et al., 2016). While the DL produces the teeth, the VL splits down the middle to form a space between the teeth and lips and teeth and cheeks, thereby creating the vestibule of the oral cavity. Given this role, the VL is also known as the labio-gingival ridge, lip-furrow band or cheek furrow ridge (Bolk, 1921; Schour, 1929; Peterková, 1985). During development, the lingual (tongue side) and buccal/labial (cheek/lip) sides of the VL are distinct in both mouse and human, with differences in rates of proliferation and expression of epithelial markers (Qiu et al., 2020; Qiu and Tucker, 2022). These differences may influence the opening of the lamina and they are later reflected by differences in molecular identity of the different parts of the oral mucosa.

The developing mammalian VL has a varied structure across mammals. In human and sheep, the VL is in general thick and wide, whereas in the vole and mouse, the VL is relatively thin and narrow (Pavlikova et al., 1999; Witter et al., 2005; Hovorakova et al., 2005, 2016; Qiu et al., 2020; Qiu and Tucker, 2022). The human VL is prominent throughout the upper and lower jaws and regionalized in parallel with the DL (Hovorakova et al., 2005, 2007). In contrast, the mouse only has a prominent VL in the anterior mandible, and the maxillary VL is almost absent (Qiu and Tucker, 2022). The relationship of the DL and VL also varies, with the DL branching off the VL in humans, while in mice both the DL and VL have a separate connection to the oral cavity. These differences are likely to correspond to the different morphologies of the vestibule and reflect differences in diet and chewing patterns.

A similar laterally situated lamina is also observed in reptiles. This lamina also appears to share a common origin with the adjacent DL. In the chameleon, this more lateral lamina develops into the dental glands that lubricate the teeth, while in poisonous snakes it forms the venom gland (Vonk et al., 2008; Tucker, 2010). It has been suggested that the mammalian VL and the reptilian dental gland lamina have evolved from laterally positioned DLs, similar to those that create the inner and outer dental arches in extant axolotls (Soukup et al., 2021) As the dentition became restricted to a single row of teeth in many tetrapods, the redundant DLs would have been lost or, in the case of the dental glands and VL, repurposed to create other ectodermal structures (Hovorakova et al., 2020).

Despite its shared origin with the DL, the VL does not normally form teeth. However, overexpression of Wnt signalling in the VL does lead to ectopic tooth formation in this structure in mice (Wang et al., 2009; Popa et al., 2019) and may explain the occasional occurrence of odontomas situated in the vestibule of human subjects (Hovorakova et al., 2020). Vestibular deficiencies have been reported in a number of human syndromes; in particular, in the ciliopathy Ellis-van Creveld (EvC) (MIM #225500), where development of the VL is defective, and the gums adhere to the upper lip and cheek with multiple associated frenula (Sasalawad et al., 2013). Recently, VL anomalies along with tooth abnormalities have also been observed in a patient with cryptophthalmos (MIM #123570) due to mutation in the FREM2 gene (Kantaputra et al., 2022). In addition, a shallow (<4 mm) vestibule and frenula anomalies in the lower incisor region have been noted in a study of gingival phenotypes in healthy children (Kus-Bartoszek et al., 2022). Such deficiencies in frenulum attachment were suggested to be associated with certain periodontal diseases (Placek et al., 1974).

The stages of development of the VL have recently been characterized in mouse and human, with an analysis of proliferation, cell death and epithelial differentiation (Qiu and Tucker, 2022); however, no other data is available regarding the genes involved in shaping and directing the development of this structure. In addition, although the VL and the dentition share a common origin, whether these tissues interact during later development to coordinate oral development is unknown. To address this, a transcriptome of the murine VL at embryonic day (E) 14 was created to identify the molecular signature of this structure and compare it with the neighbouring cap stage incisor tooth germ. This allowed a comparison of odontogenetic and non-odontogenic tissue and identification of genes unique to the VL. At E14, the VL has extended into the lower jaw as a sheet of epithelium with higher proliferation on the labial and buccal sides away from the developing teeth (Qiu and Tucker, 2022). This suggested potential interactions between the tooth and VL. We, therefore, used the transcriptomic dataset to focus on signalling molecules expressed in the tooth and readouts in the VL to further understand how the two tissues might interact. Mouse mutants and explant culture were then used to test the function of identified pathways.

From this analysis, we identified previously unreported markers associated with the VL and highlighted Hedgehog (Hh) signalling from the developing tooth as playing a potential role in patterning the VL during development. Functional studies confirmed the role of Hh signalling in patterning the VL and demonstrated that signals from the tooth have a key role in orchestrating development of neighbouring structures, coordinating development of the teeth and the oral cavity.

The VL has odontogenic potential but does not normally form teeth (Popa et al., 2019). To identify the molecular signatures that distinguish the VL from the early tooth bud (TB) during development, we dissected the VL and TB from the incisor region of the murine mandible at E14 for analysis through bulk RNA-sequencing (RNA-seq) (Fig. 1A). At this stage, the anterior VL and incisor TB have developed their distinct morphologies and are clearly observed in sagittal slices, facilitating precise dissection. However, despite their morphological differences, the VL can still be induced to form tooth-like structures after stabilization of β-catenin at this late stage (Popa et al., 2019). Fate is, therefore, not yet determined in this structure at E14. The epithelium and mesenchyme were dissected together to provide a comprehensive characterization of the whole structure. The RNA-seq dataset included VL (n=5, VL1-5) and TB (n=5, TB1-5) samples, with the boxplot, violin plot and density plot showing consistent distributions of normalized read counts (counts per million; CPM) (Fig. S1A-C). A correlation plot confirmed that there was a higher correlation between the transcriptomic profile of VL-VL and TB-TB compared with TB-VL (Fig. S1D). Levels of vimentin (a general mesenchymal marker) and Cdh1 (a general epithelial marker) were used to estimate the relative amounts of mesenchymal and epithelial tissue included in the analysis. No significant difference was observed in expression of these two markers between the TB and VL groups, suggesting that the dissected regions contained approximately equivalent numbers of mesenchymal and epithelial cells (Fig. S1E,F). Principal component analysis (PCA) presented distinct gene expression distributions between the VL and TB datasets at E14 (Fig. 1B). In total, 1788 differentially expressed genes (DEGs) were detected, including 788 upregulated (TB) and 1000 downregulated (VL) DEGs. The top DEGs were visualized by a volcano plot, where absolute value of log2 fold change (abs log2FC)>2 and P-value<0.01 (Fig. 1C). The DEGs were highlighted in a heatmap where abs log2FC>2 and P-value<0.001 (Fig. 1D). Overall, these results confirmed that the VL and TB were significantly different at the molecular level by E14. Our large-scale analysis of gene signatures confirmed that genes known to play a role in early tooth development were upregulated in the TB relative to the VL, including Shh, Fgfs, Pax9, Dlx homeobox gene-family members, Runx2, Rspo1 and Scube1 (Zhao et al., 2000; Seppala et al., 2007; Xavier et al., 2009; Kawasaki et al., 2014) (Fig. 1C,D), with gene expression patterns confirmed by RNAscope (Fig. S2). In addition, the screen highlighted a number of genes upregulated in the VL relative to the TB (Fig. 1C,D), with expression in the VL again confirmed by RNAscope (Fig. S3), or by checking expression on Genepaint (https://gp3.mpg.de/), a digital atlas highlighting gene expression in the embryo at E14.5 (Fig. S4) (Visel et al., 2004). These included Meis1/2, Otx1 and Nr4a2 transcription factors, and signalling pathway members Cd44 and Wnt7b, which have not previously been associated with the VL. Interestingly, in keeping with the observed lingual-buccal/labial differences in the VL epithelium at this stage (Qiu and Tucker, 2022), some highlighted genes had clear lingual-buccal/labial differences in expression, such as Otx1, which was expressed on the labial side of the VL away from the tooth (Fig. S3L′). In addition, differential expression was evident in the oral and aboral parts of the VL, suggesting additional levels of compartmentalization (Fig. S4). Using this data, we were, therefore, able to create the first molecular signature for the VL.

To understand how the tooth bud might impact on development of the VL, we focused on signalling pathways highlighted by the transcriptomics. As expected given its conserved role in tooth development, Shh was highlighted as differentially expressed in the tooth germ, along with patched 1 (Ptch1) and Gli1 (Fig. 2A,B; Fig. S5A-F) (Buchtová et al., 2008; Seppala et al., 2017). Ptch1 is a transmembrane domain protein that acts as the principal receptor for Shh, but also provides a readout of Shh signalling activity, whereas Gli1 acts as a universal downstream transcriptional target. The Shh pathway is activated when Shh binds to Ptch1, which is facilitated by a number of co-receptors, including growth arrest-specific 1 (Gas1) and Ig/fibronectin subfamily transmembrane proteins Boc and Cdon (Cdo) (Tenzen et al., 2006; Seppala et al., 2022). Gas1 is a vertebrate-specific glycosylphosphatidylinositol-anchored membrane protein proposed to be a co-receptor for Shh and positive regulator of the signal pathway in vertebrates (Martinelli and Fan, 2007; Seppala et al., 2022). In contrast to Shh and Ptch1, Gas1, Cdon and Boc were all significantly more highly expressed in the VL (Fig. 2A,B). To further investigate the expression of these Shh pathway-related genes, we compared the expression of Ptch1, Cdon, Shh, Gas1 and Boc in these two laminas at E12.5 (before obvious morphological differences in these structures) and at E15.5 (when the two structures are clearly morphologically distinct) (Fig. 2E-P). The VL runs as a continuous band anteriorly around the incisor tooth germs and, therefore, can be viewed both in sagittal plane for the labial lamina (Fig. 2C) and frontal plane for the buccal lamina (Fig. 2D). At E12.5, the VL (lined by green dashed line) and DL (lined by white dashed line) were evident as two side-by-side epithelial thickenings (Fig. 2E) with Shh expression restricted to the DL (Fig. 2H). Ptch1 was expressed in the epithelium of the VL, DL and strongly in the surrounding mesenchyme (Fig. 2F). In contrast, expression of Gas1 was mainly observed in the epithelium of the VL and the mesenchyme on the buccal side (Fig. 2I). Cdon and Boc showed similar expression patterns to Gas1 (Fig. 2G,J). At E15.5, the VL curves as it extends into the mesenchyme so that it grows under the developing incisors and Meckel's cartilage, while the neighbouring incisor tooth germ has developed from a thickening to reach the late cap stage (Fig. 2K). At this stage, strong transcription of Shh was detected in the inner enamel epithelium of the incisors (Seppala et al., 2017); however, the adjacent VL was negative for Shh expression (Fig. 2N). Ptch1 was highly expressed in the developing tooth, in the epithelium and surrounding mesenchyme, but relatively lower Ptch1 expression was observed in the VL, mainly localized to the lower part of the VL (Fig. 2L). In contrast to the tooth germ-dominant expression of Shh and Ptch1, Cdon, Gas1 and Boc were reduced in the developing tooth germs but were strongly expressed on the buccal side of the VL in the buccal epithelium and mesenchyme (Fig. 2M,O,P). This buccal/labial bias of Gas1 expression in the VL epithelium was confirmed at E14 by RNAscope (Fig. S5D-F).

Given the distinct expression pattern of Gas1 in the VL, we focused on this component of the Shh pathway through the analysis of Gas1 mutant mice. Gas1 has been shown to facilitate Shh signalling, particularly with low signal levels at long distance from the source (Martinelli and Fan, 2007). A lack of Gas1 is associated with a series of midline craniofacial phenotypes including maxillary incisor fusion, midfacial hypoplasia and cleft palate due to reduced Shh signalling in the early craniofacial region (Seppala et al., 2007). At E13.5, the VL (contoured by green dashed lines) has started to extend into the mesenchyme under the forming incisors in wild-type (WT) mice, with full extension under the tooth germ, towards the midline, by E15.5 (Fig. 3A,F). In contrast, in the Gas1 null mutants (Gas1^−/−), the VL was shorter and failed to pass under the forming incisors (Fig. 3B,G) (truncated VL observed in N=15/15 Gas1^−/− embryos from E13.5 to E15.5). Interestingly, a similar truncated VL was observed in some Gas1^+/− embryos at E13.5, although the severity of the defect varied between heterozygous mutants (Fig. 4J-L) (truncation observed in N=5/7 Gas1^+/− embryos at E13.5). The reduction in the VL was significant in both homozygous and heterozygous mutants when measured at E13.5 (Fig. 4G).

Gas1 has also been reported to interact with the Hh co-receptor Boc in different developmental contexts (Allen et al., 2011). We therefore further explored the function of Boc during VL development using mutant mice. Interestingly, analysis of a Boc mutant showed a normal VL indistinguishable from WT (Fig. 3C,H) [normal VL observed in N=4/4 Boc^−/− embryos from E13.5 to postnatal day (P) 0]. Loss of a single Boc allele in Gas1^−/− mice did not significantly exacerbate the phenotype compared with Gas1^−/− mutants (Fig. 3D,I; Fig. 4G) (truncated VL observed in N=9/9 Boc^+/−Gas1^−/− embryos from E13.5 to P0). In contrast, Boc/Gas1 double knockouts had exacerbated VL and TB phenotypes, with a severely truncated VL and a developmentally arrested incisor tooth germ (Fig. 3E,J) (severely truncated VL observed in N=3/3 Boc^−/−Gas1^−/− embryos at E13.5 and E15.5). Gas1 and Boc, therefore, appear to be able to compensate for each other to some extent during VL development.

The identification of a shorter VL in Gas1^−/− mice suggested that the VL is dependent on Shh signalling for normal development. To confirm this, we assessed expression of Ptch1 and Gli1 in these mutants at E13.5, the stage when a clear VL defect is first identifiable in these mice. Ptch1 was expressed on the labial/buccal side of the VL in WT embryos but, in Gas1^−/− mice, this expression was lost in the truncated lamina but retained in other regions, including the aboral mesenchyme, underlying the developing whisker follicles (Fig. 4A,D). In the WT, Gli1, as viewed by RNAscope, was similarly expressed at significantly higher levels on the buccal side of the VL compared with the lingual side of the VL (Fig. 4B,C,H). Quantification of the RNAscope in the Gas1^−/− mice revealed that the levels of expression in the buccal epithelium were significantly reduced (Fig. 4E,F,I; Fig. S6) (N=3). Given that Gas1^+/− mice showed a variable truncated VL defect, we assessed changes in Gli1 in heterozygotes with and without a phenotype at E13.5. In a Gas1^+/− mouse with no phenotype at this stage, Gli1 expression appeared as for the WT (Fig. 4M,N). In contrast, Gas1^+/− mice with a truncated VL still showed Gli1 expression but had lost the normal buccal-lingual polarity (Fig. 4O). The buccal-lingual difference in Shh response was, therefore, lost in mice with a truncated VL.

To analyse the cause of the truncated phenotype, we investigated the level of proliferation during early development of the VL before the defect in epithelial extension using bromodeoxyuridine (BrdU) to detect cells in the S phase of the cell cycle. At E12.5, the Gas1^−/− VL and DL were evident as two distinct thickenings, although they appeared to be closer to each other when compared with WT littermates, suggesting part of the phenotype is determined very early in VL development (Fig. 5A-C,G). E-cadherin was used to outline the epithelium and the number of BrdU-positive cells were counted within a defined region, taken from the deepest projection of the DL to the VL (N=3) (Fig. 5D-F,H-J, area outlined in D,E,H,I highlights region counted). Proliferation in the epithelium was significantly reduced in the Gas1^−/− embryos (Fig. 5J,K), with clear reduction of proliferation in the mesenchyme around the forming VL at this stage (Fig. 5D,H).

Given the strong expression of Gas1 in both the buccal epithelium and mesenchyme of the VL during development (Fig. 2I,O; Fig. S5E,F), and reduction of proliferation in both tissues, we wanted to address whether truncation of the VL was caused by loss of Gas1 in the mesenchyme alone. For this, Wnt1-Cre;Gas1^fl/fl conditional mutant mice were used. At E16.5, similar to E15.5, the WT VL (labelled by green dashed lines) had extended under the incisors with the ends almost touching each other in the midline in the anterior, and slightly less extended more posteriorly (Fig. 6A,C) (Qiu and Tucker, 2022). Interestingly, Wnt1-Cre;Gas1^fl/fl mice mimicked the VL phenotype observed in Gas1^−/− mice, with a significantly truncated VL that failed to reach past the developing incisors (Fig. 3G compared with Fig. 6B,E) (truncation in N=3/3). Wnt1-Cre mice have been shown to have some brain defects but no craniofacial defects (Heuzé et al., 2014), and in keeping with this the VL was normal in all Wnt1-Cre controls (N=5/5) (Fig. 6C-E). These findings highlight that loss of Gas1 in the neural crest-derived mesenchyme alone is sufficient to cause a truncated VL phenotype.

To further manipulate Shh signal levels during early development of the VL, we moved to an ex vivo explant culture system, which has previously been used to successfully culture the VL (Qiu and Tucker, 2022). Here, the mandible harvested at E13.5 was chopped longitudinally into 200 μm thick live slices, and the midline slices containing the VL and incisor tooth germ were selected. Cyclopamine was added at a concentration of 20μM and 50μM to block Shh signalling by interfering with the obligate smoothened receptor (Li et al., 2016). At 20μM whisker follicle development was clearly affected in the slices (Fig. S7), but no obvious VL phenotype was evident (Fig. 7A-D). We therefore increased the concentration to 50μM. Cyclopamine-treated cultures at 50μM continued to grow over the 2 day culture period, with development of a cap stage tooth, but the treated cultures had significantly shorter VLs when compared with slices treated with the carrier solution alone (Fig. 7E-H; Fig. S8) (N=6 control, 6 treated). Confirming loss of Shh signalling, addition of cyclopamine caused an almost complete loss of Gli1 and Ptch1 expression (Fig. S10A,C,D,F). To enhance Shh signalling, cultures were treated with SAG (smoothened agonist) at 5μM, which acts to stabilize the key Hh transducer smoothened in the primary cilium. The SAG-treated cultures displayed severe retardation of tooth development and a significant thickening of the forming VL (Fig. 7I-L; Fig. S9) (N=7 control, 7 treated) and was accompanied by an upregulation of Gli1 in the mesenchyme around the VL and tooth germ and Ptch1 in the thickened VL (Fig. S10B,E). Targeted loss of Shh signalling at E13.5, therefore, impacted VL development.

An important aim of this work was to investigate the molecular signature of the VL. The bulk RNA-seq data at E14 provided an interesting set of genes that are likely to play a role in VL development and can be used as markers of the VL in future experiments. Interestingly Wnt7b was expressed in the VL, and this gene has been proposed to inhibit Shh expression in non-dental regions (Sarkar et al., 2000). Meis1/2 were also highlighted in the VL. Meis1 has been linked to stem cell maintenance and overlaps with Sox2-expressing regions, with both Sox2 and Meis1 expressed at high levels in the VL (Sanz-Navarro et al., 2019). The VL may, therefore, have some stem cell properties.

One perhaps unexpected result was that many genes exhibited differential expression in the lingual and labial/buccal sides of the VL, suggesting that the epithelium is spatially patterned at this stage, more than 1 week before the VL opens to form the vestibule (Qiu and Tucker, 2022). This differential expression, however, agrees with the observed differences in proliferation rates of the labial/buccal and lingual sides of the VL, which have been proposed to drive the shape of the elongating VL (Qiu and Tucker, 2022). Differences in oral and aboral parts of the VL were also evident, highlighting further compartmentalization of the VL during development. Overall, this is the first molecular analysis of the VL and provides a number of interesting pathways for further exploration.

Here, we determined that, although the VL does not express Shh after the earliest placodal stages, its development is dependent on Shh signals from the developing tooth. The two laminas are therefore patterning by the same signal. The Shh co-receptors Gas1, Boc and Cdon were shown to have enhanced expression in the forming VL, and loss of Gas1 in mouse mutants resulted in truncation of the forming VL. Failure in extension of the VL was driven by a reduction/misexpression of readouts of Hh signalling and a reduction in proliferation. Interestingly, Gas1 heterozygous mice showed a variable defect in the VL, correlating to changes in Gli1 expression. Gas1 heterozygous mice are viable and feed as normal, suggesting that earlier VL defects may be able to correct themselves during later development. This may be linked with developmental stalling, whereby mutant mice display initial defects and a slowing down of development but are able to recover with later accelerated development (Miletich et al., 2011). In both homozygotes and heterozygous Gas1 mutants the truncated VL was linked to a loss of the differential expression of Gli1 in the epithelium, suggesting that the lingual-buccal/labial differences are key drivers of vestibular lamina outgrowth.

Gas1 mutants have previously been shown to have tooth defects, which include fusion of the first and second molars, formation of supernumerary teeth in the diastema and cusp defects (Seppala et al., 2022). Loss of one copy of Boc in Gas1 mutants did not exacerbate the phenotype but loss of two copies in the compound mutants resulted in very severe VL defects. Boc and Gas1 have been shown to form distinct complexes with Ptch1 (Izzi et al., 2011). Boc^−/− are viable, with subtle changes to the cerebellum and a background-dependent mild widening of the facial region (Izzi et al., 2011; Echevarrıá-Andino and Allen, 2020). Interestingly, although in many tissues loss of both Boc and Gas1 leads to a worsening of the single mutant phenotype (Seppala et al., 2014), with regards to the width of the face, Boc and Gas1 double mutants had a less severe craniofacial phenotype, suggesting context-dependent effects of these two co-receptors (Echevarrıá-Andino and Allen, 2020). Similarly, Gas1 can restrain, rather than enhance, Hh signalling in the mouse diastema and in presomitic mesoderm explants (Lee et al., 2001a; Cobourne et al., 2004). In the VL, however, Boc and Gas1 appear to act by facilitating Shh signalling in tissue at a distance from the Shh source, with a significant downregulation of Gli1 in the VL of Gas1 mutants, particularly on the buccal/labial side. A proposed model of Hh signalling in the VL is shown in Fig. 8, highlighting the known role of the primary cilium. This agrees with findings in the heart, neural tube and limb (Martinelli and Fan, 2007). In these tissues, Shh has been shown to downregulate Gas1 to constrain its own activity, which may explain the heightened expression of Gas1 in the VL compared with the tooth in WT embryos.

Gas1 was expressed in both the buccal/labial epithelium of the VL and the surrounding mesenchyme. Perhaps unexpectedly, loss of Gas1 in the neural crest-derived mesenchyme alone led to truncation of the VL. Expression in the epithelium is therefore not sufficient for VL extension. Tooth number has similarly been found to be dependent on Gas1 expression in the mesenchyme, with supernumerary teeth developing in the diastema of Wnt1-Cre;Gas1^fl/fl mice (Seppala et al., 2022).

Vestibule defects are associated with EvC syndrome, where the gums adhere to the upper lip and cheek with multiple associated frenula (Sasalawad et al., 2013). Interestingly, Evc also has a role in the Shh pathway, interacting with smoothened in the primary cilium (Fig. 8). Evc regulates Hh signalling by promoting Sufu/Gli3 dissociation and Gli3 ciliary traffic (Caparrós-Martín et al., 2013). A lacZ Evc reporter has previously shown expression of Evc around the teeth and VL at E15.5 (Ruiz-Perez et al., 2007). The VL has not been studied in Evc mutant mice, but a similar truncated VL might be predicted based on a reduction in the Hh pathway and the phenotype from the Gas1 mutants. VL defects would also be predicted to occur in other ciliopathies and, in keeping with this, frenula defects have been noted in oral-facial-digital (OFD) syndromes (Bruel et al., 2018) and in Joubert syndrome (Penon-Portmann et al., 2022). OFD syndromes can be caused by defects in the ciliary basal body, ciliogenesis or post-ciliary microtubule functions, with at least 20 genes identified in patients (Bruel et al., 2018). It would, therefore, be interesting to study the VL in mouse models where primary cilia function is disrupted. In addition, changes in the vestibule, with defects and accessory frenula, may represent an important additional clinical character for defining ciliopathies in patients.

Our culture experiments allowed the development of the VL and tooth germ to be followed and the Hh pathway to be manipulated at specific time points. Addition of cyclopamine or SAG to target smoothened led to striking defects in the VL. Agreeing with the results from the Gas1 mutants, addition of cyclopamine caused a shorter VL that failed to extend into the surrounding mesenchyme, strengthening the results that Gas1 promotes Hh signalling in the VL. Addition of SAG led to a broadening of the VL, which we hypothesize was caused by an extension of the tissue that was able to respond to Shh produced by the neighbouring tooth germ (Fig. 8). The width of the VL varies considerably across mammals, therefore, the wider VL of humans may be driven by enhanced Hh signalling in this tissue compared with mammals such as the mouse, where the VL is very thin.

The results highlight that the tooth germ, through Hh signalling, plays a key role in development of the VL. These two neighbouring laminas, therefore, not only share a common origin in some parts of the jaw but are later patterned together. Defects in the tooth are, therefore, likely to result in defects in the VL. The VL is a fairly unstudied structure, therefore, many published mouse mutants with dental defects may have unreported defects in the VL. From an evolutionary perspective, changes to the dentition might not change in isolation but involve secondary effects on the vestibule, linking the evolutionary history of these two structures.

Gas1 mutant mice were generated as previously described (Lee et al., 2001b; Seppala et al., 2022). Gas^+/− (Lee et al., 2001b) and Boc^+/− (Okada et al., 2006), Wnt1-Cre^+/− (Danielian et al., 1998) and Gas1^fl/+ (Jin et al., 2015) were crossed to generate compound and conditional mutant mice, respectively. The mice analysed in this research are listed below: WT (E12.5 n>9; E13.5 n>9; E14.5 n>3; E15.5 n>9; E16.5 n>3), Gas1^−/−N=18 (E12.5 n=3; E13.5 n=8, E14.5 n=4, E15.5 n= 3), Gas1^+/−N=13 (E12.5 n=3; E13.5 n=7; E14.5 n=3), Boc^−/−N=4 (E13.5 n=1; E14.5 n=1; E15.5 n=1; P0 n=1); Boc^+/−Gas1^−/−N=9 (E13.5 n=4; E15.5 n=3; E18.5 n=1; P0 n=1); Boc^−/−Gas1^−/−N=3 (E13.5 n=1; E15.5 n=2); Wnt1-Cre;Gas1^fl/fl N=3 (E16.5), Wnt1-Cre N=5 (E16.5). Mutant mice and littermate controls were on a mixed 129sv/C57Bl6 background (Seppala et al., 2022). CD1 mice for transcriptomics, explant culture and in situ hybridization were obtained at the Institute of Animal Physiology and Genetics, Czech Academy of Sciences, and King's College London. E0.5 was considered as the day when the plug was detected. All animal procedures were carried out under the guidelines of the Institute of Animal Physiology and Genetics, and King's College London, with Home Office approved Schedule 1 culling methods and conform to ARRIVE guidelines. For proliferation assays, BrdU (30 mg/kg) was injected into the pregnant mouse 2 h before culling.

VL and neighbouring incisor tooth germs were isolated from the same slices at E14. The dissected VL and tooth germs from a single litter (9-15 embryos) were pooled to obtain enough RNA for sequencing. In total, tissue from 11 litters was isolated at this stage. RNA extraction was performed using the RNeasy Plus Mini Kit (Qiagen, 74136) and pooled samples from five litters with the highest quality scores were used for analysis. These are referred to here as TB1-5 and VL1-5. Sequencing libraries were prepared from total RNA using the Smarter Stranded Total RNA-seq Kit v2 Pico Input Mammalian (Takara), followed by size distribution analysis in the Agilent 2100 Bioanalyzer using a High Sensitivity DNA Kit (Agilent, 5067-4626). Libraries were sequenced using the Illumina NextSeq 500 instrument. Samples were sequenced by the Genomics service at the Institute of Molecular Genetics. Tuxedo protocol was performed for Genome-guided assembly.

The RNA-seq datasets were visualized using R (v4.0.2). Duplicated genes and genes with zero read counts were removed. The counts per gene were normalized to CPM transformed with log2 using an offset of 1. Expression plots such as box plot, violin plot and density plot were generated using package ggplot2 (v3.3.5). Two-tailed, unpaired t-test was performed for statistical significance of comparisons in gene expression visualized in box plots. PCA was performed using package pca3d (v0.10.2). Correlation matrix was visualized using package corrplot (v0.90). For differential expression analysis, package limma (v3.46.0) was used to identify the DEGs in TB and VL groups (N=5 in each group). The cut-off criteria for DEGs were set up with the abs log2FC>1.5 and P-value<0.05. Heatmaps and volcano plots were depicted to visualize the results using the packages pheatmap (v1.0.12), ggplot2 (v3.3.5), dplyr (v1.0.7) and ggrepel (v0.9.1). To prioritize genes of interest, two groups of criteria were established in volcano plots with the abs log2FC_1>1, P-value_1<0.05, and abs log2FC_2>2, P-value_2<0.01. The genes were filtered with settings of the abs log2FC>2 and P-value<0.001 in Heatmap showing ranked genes with highest differential expression.

Embryonic heads were fixed in 4% paraformaldehyde (PFA) overnight at 4°C followed by gradient dehydration in ethanol, xylene clearance, wax immersion and paraffin embedding. Samples were then sectioned at 5-8 μm using a microtome (Leica RM2245) and mounted on the charged slides in series sections. For histological analysis, the slides were stained with Haematoxylin and Eosin (H&E) staining, or trichrome staining (Sirrus Red, Haematoxylin and Alcian Blue) using standard protocols. The Nikon Eclipse 80i light microscope was used to photograph the stained slides.

Immunostaining was performed using standard protocols as previously described (Qiu et al., 2020; Qiu and Tucker, 2022). The primary antibodies used on sections were: mouse anti-E-cadherin (Abcam, ab76055, 1/400) and rat anti-BrdU (Abcam, ab6326, 1/500). Secondary antibodies were then used to incubate the sections for 1 h at room temperature (RT) at a dilution of 1/500 in the dark: Alexa Fluor donkey anti-mouse 488 (Invitrogen, A21202) and Alexa Fluor donkey anti-rat 647 (Invitrogen, A21247). Slides were mounted with Fluoroshield containing DAPI (Sigma-Aldrich, SLBV4269) and visualized using a Leica TCS SP5 confocal microscope. For BrdU immunofluorescence staining, the mice were injected with BrdU labelling reagent (30 mg/kg, Life Technologies, 000103) 2 h before collection. For E-cadherin immunofluorescence on slices, rabbit anti-E-cadherin (Abcam, ab15148, 1:100) with secondary Fluor goat anti-rabbit 488 (Invitrogen, A-11008, 1:500). Control groups were set up to confirm immunofluorescence staining. Each antibody analysis was carried out at least three times independently.

Radioactive section in situ hybridization was performed as previously described (Wilkinson, 1992; Seppala et al., 2014). RNA probes were synthesized using ³⁵S-UTP, and signals were recognized using silver emulsion, which presents positive signals as white grains when viewed under dark-field. Bright-field and dark-field images were photographed using a Nikon Eclipse 80i light microscope. The images were then merged in Photoshop (Adobe 2020), with the colour in dark-field artificially changed to red. The following restriction and polymerase enzymes were used for plasmid DNA linearization and mRNA probe synthesis; EcorI and T7 for Shh, BamHI and T3 for Ptch1, EcorI and T7 for Gas1, XhoI and T7 for Cdon, SalI and T7 for Boc. We would like to thank Andrew McMahon (Harvard University, USA) for the Shh plasmid, Matthew Scott (Stanford University, USA) for the Ptch1 plasmid, Chen-Ming Fan (Carnegie Institution of Washington, USA) for the Gas1 plasmid and Robert Krauss (Icahn School of Medicine at Mount Sinai, USA) for the Cdon and Boc plasmids.

Paraffin-embedded embryonic head sections (5 μm) were processed as described above for multiplex fluorescent in situ hybridization. Commercially available Multiplex Fluorescent Reagent Kit v2 (323100, Advanced Cell Diagnostics) and RNAscope probes (1:50; Mm-Gli1, 311001; Mm-Meis1, 436361; Mm-Meis2, 436371; Mm-Cd44, 476201; Mm-Nr4a2, 423351; Mm-Wnt7b, 401131; Mm-Otx1, 536041; Mm-Dlx5, 478151; Mm-Runx2, 414021; Mm-Rspo1-O1, 479591; Mm-Scube1, 488131; Mm-Shh, 314361; Mm-Gas1, 547201; Mm-Ptch1, 402811; Advanced Cell Diagnostics) were used for transcript detection according to the manufacturer's protocol. The hybridized probes were visualized using the fluorescein (NEL741001KT, Perkin Elmer) and TSA-Plus Cyanine 3 (NEL744001KT, Perkin-Elmer) system, according to the manufacturer's protocol. All incubation steps were completed using the ACD HybEZ II Hybridization System (321721) at 40°C. In brief, the slides were deparaffinized in xylene, applied in RNAscope Hydrogen Peroxide (322335) at RT for 10 min and submerged in RNAscope Target Retrieval (322001) at 99°C for 30 min. A hydrophobic barrier was then created to circle the interested areas using the ImmEdge Hydrophobic Barrier Pen (310018). Slides were then incubated with RNAscope Protease III (322340) for 15 min followed by the hybridization, amplification and detection steps according to the protocol. Slides were mounted with Fluoroshield with DAPI (Sigma-Aldrich, SLBV4269) and stored at 4°C in the dark.

Quantification of Gli1 expression levels from RNAscope is outlined in Fig. S6. Morphometrics of structures were performed using Fiji/ImageJ. The length of the VL was assessed by drawing a line parallel to the curved VL from the top to the end using the line tool in Fiji. The width of the VL was measured by drawing a line parallel to the oral epithelium from the junction of VL and DL to the edge of the buccal VL. Results were plotted in GraphPad Prism software (V.8.0.2). Statistical significance was analyzed by IBM SPSS Statistics software (V.25.0) using two-tailed, unpaired t-tests. The P-values were considered statistically significant if the P-value<0.05, with *P<0.05, **P<0.01 and ***P<0.001. Error bars represent s.e.m.

Slice culture for the tooth germ was carried out as previously described (Alfaqeeh and Tucker, 2013). In brief, embryonic lower jaws were isolated at E13.5, when a very clear VL/DL bud can be observed. The mandibles were then chopped sagittally at 200 μm using a McIlwain Tissue Chopper, and slices in the incisor region were selected and cultured on permeable filters (pore size 0.4 μm, BD Falcon cell culture inserts) held by a steel mesh in Advanced Dulbecco's Modified Eagle Medium F12 (DMEM F12) (Invitrogen) with 1% penicillin-streptomycin (Sigma-Aldrich, Merck) and 1% Glutamax (Invitrogen). To investigate the role of Shh during VL development, slices were cultured with Shh signalling inhibitor cyclopamine (Sigma-Aldrich) at 20 μM or 50 μM (10 mg/ml stock in ethanol), or SAG (smoothened agonist) (Sigma-Aldrich) at 5 μM (stock dissolved in H₂0). The explants were cultured at 5% CO₂ and 37°C for 2 days before harvest. Littermate control groups were set up to compare with the experimental groups and cultured under the same conditions but with carrier alone. Litter age can vary depending on mating time, so cultures were only compared within litters. Slices were photographed at day 0 and day 2 using a Leica dissecting microscope to record the morphology. All experiments were repeated more than six times for SAG and 50 μM cyclopamine, and four times for 20 μM cyclopamine.

We thank Chen-Ming Fan (Carnegie Institution of Washington) for generously providing us with the Gas1 mutant line and compound mutant embryos.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.201464.reviewer-comments.pdf.