Cell fate specification in the lingual epithelium is controlled by antagonistic activities of Sonic hedgehog and retinoic acid

The interaction between signaling pathways is a central question in the study of organogenesis. Using the developing murine tongue as a model, we uncovered unknown relationships between Sonic hedgehog (SHH) and retinoic acid (RA) signaling. Genetic loss of SHH signaling leads to enhanced RA activity subsequent to loss of SHH-dependent expression of Cyp26a1 and Cyp26c1. This causes a cell identity switch, prompting the epithelium of the tongue to form heterotopic minor salivary glands and to overproduce oversized taste buds. At developmental stages during which Wnt10b expression normally ceases and Shh becomes confined to taste bud cells, loss of SHH inputs causes the lingual epithelium to undergo an ectopic and anachronic expression of Shh and Wnt10b in the basal layer, specifying de novo taste placode induction. Surprisingly, in the absence of SHH signaling, lingual epithelial cells adopted a Merkel cell fate, but this was not caused by enhanced RA signaling. We show that RA promotes, whereas SHH, acting strictly within the lingual epithelium, inhibits taste placode and lingual gland formation by thwarting RA activity. These findings reveal key functions for SHH and RA in cell fate specification in the lingual epithelium and aid in deciphering the molecular mechanisms that assign cell identity.


Introduction
The tongue is a muscular organ that plays critical roles in mastication, speech and taste, and the regulatory mechanisms that specify the diverse cell types and structures of the tongue are of great interest.The mature tongue (Fig 1A and 1B) is covered by a mucosa made of lingual epithelium (LE) and lingual mesenchyme (LM).The dorsal LE of the anterior 2/3 of the tongue (oral tongue) is a stratified, squamous epithelium primarily comprised of mechanosensory filiform papillae.The gustatory units of the LE, the taste buds (TBs), develop in three different types of papillae: fungiform papillae, foliate papillae, and circumvallate papillae.Fungiform papillae house a single TB and are distributed between filiform papillae over the dorsal surface of the oral tongue, whereas foliate papillae house several TBs and develop posteriorly at the lateral edges of the oral tongue.In rodents, a single circumvallate papilla harbouring numerous TBs forms at the junction between the oral tongue and the posterior 1/3 of the tongue, known as pharyngeal tongue [1].The LE also produces minor salivary glands of a mixed sero-mucous type in the pharyngeal tongue, as well as a purely serous type, von Ebner's glands, which arise from the epithelium of the circumvallate papilla [2,3].
The distribution of mature taste papillae depends on coordinated signaling events during early stages of tongue development [4].From embryonic day (E) 10.5-E11.5,Shh is expressed throughout the LE and signals both within the LE and to the LM [5][6][7][8].At E12.5, fungiform papilla development is heralded by fungiform placodes (FPs), localized thickenings of the LE made of post-mitotic TB precursors expressing Shh, Wnt10b, and Bmps.At E12.5 to E14, the interplacodal epithelium is devoid of Shh expression, which is restricted to taste placodes.Canonical signaling through these pathways is critical for regulating the size, number, and spacing of fungiform placodes [4].As the tongue grows, new FPs develop until E14.5, at which time fungiform papilla morphogenesis begins and differentiating TBs become morphologically and molecularly visible at the tip of fungiform papillae.After E14.5 and during early postnatal life, only developing TB cells within gustatory papillae produce SHH [4].
Several components of the Hedgehog signaling pathway play key roles to ensure properly calibrated spatio-temporal Hedgehog inputs [8][9][10][11].Upon ligand binding to the Hedgehog receptor Patched (PTCH1), Smoothened (SMO), an obligatory Hedgehog transducer, translocates to the primary cilium, unleashing a signaling cascade culminating in transcriptional regulation of Hedgehog target genes by GLI proteins [10].Gli1 and Ptch1 are themselves direct targets of Hedgehog signaling, and hence their expression enables identification of cells responding to Hedgehog signals.Loss of SMO function abrogates all Hedgehog signaling even in the presence of copious amounts of ligands [9].
Studies of tongue organ cultures have shown that pharmacological inhibition or activation of SHH signaling causes increased or decreased size and number of FPs, respectively,  (C') and (D') are enlarged images of the boxed areas in (C) and (D), respectively.The oral tongue of mutants exhibits epithelial hyperplasia, lacks filiform papillae (FiP), and displays serous (sHG) and sero-mucous (smHG; mucous cells stain blue) heterotopic glands.(E,F) Anti-Keratin 8-stained (K8; purple) parasagittal sections of E18.5 control (E) and mutant (F) tongues.The oral tongue of the mutant shows K8-positive heterotopic glands (HG).Adhesion between the tongue and floor of the oral cavity in the mutant (arrowheads in F). (G,H) Alcian blue van Gieson-stained sections across the circumvallate papilla (CvP) of E18.5 control (G) and mutant (H) tongues.The mutant shows overgrown von Ebner's glands (vEG) exhibiting mucous metaplasia (arrows in H).The inset in (H) is an enlarged image of the boxed area in (H).(I-N) Frontal (I,J) and parasagittal (K-N) tongue sections from controls (I,K,M) and mutants (J,L,N) at P0 (I-L) and E18.5 (M,N) immunostained for Keratin 15 (K15), Keratin 6 (K6) and K8 (purple).The mutant tongues exhibit epithelial foci with weak or no K15, K6, and K8 staining (arrows in J,L and N).CvP, circumvallate papilla; FiP, filiform papillae; FuP, fungiform papillae; IME, intermolar eminence; LE, lingual epithelium; LM, lingual mesenchyme; TB, taste bud; pLG, posterior lingual glands.Scale bars: 1 mm (C-F), 100 μm (G,H), and 50 μm (C',D', I-N). https://doi.org/10.1371/journal.pgen.1006914.g001 indicating that SHH inhibits FP formation [12][13][14][15].However, whether and how SHH performs such a function in vivo is unknown [6,16].Moreover, it remains unclear which lingual tissue SHH acts upon, as in vitro manipulations may disrupt the LE and LM which are both SHH-responsive [5,6,8].Furthermore, how different lingual structures such as TBs and glands are specified within a seemingly homogeneous epithelial sheet remains a critical open question.That SHH signaling is active in both the oral and pharyngeal tongue during the earliest steps of tongue formation [6,8] suggests this molecule plays a yet unknown role in cell fate specification within the LE.
To address these issues, we utilized mouse genetics to modulate SHH signaling in the LE, LM, or both.We found that SHH plays a dual role during tongue development by acting in the LM to control growth and morphogenesis and in the LE to control patterning and cell fate determination.We revealed an unknown, highly sensitive balance between SHH and retinoic acid (RA) signaling that regulates the specification of taste placodes and glands.Upon loss of SHH signaling, the augmented RA inputs direct the LE to (1) form minor salivary glands in the oral tongue, (2) overproduce oversized FPs and TBs, and (3) sustainably generate taste placodes.We found that both FP specification and antero-posterior patterning of the LE are regulated by antagonizing activities of SHH and RA.We show that SHH antagonizes RA signaling by maintaining expression of Cyp26a1 and Cyp26c1.Unexpectedly, we found that in the absence of SHH signaling, the LE aberrantly produced Merkel cells.Finally, we identified two novel TB markers, Homer1 and RALDH1.

Results and discussion
Glandular metaplasia, epithelial hyperplasia, and failure of filiform papilla formation upon loss of SMO in the lingual epithelium To decipher the function of SHH in patterning of and cell type specification in the LE, we first irreversibly ablated SHH signaling in epithelial cells that express Shh and their progeny by ShhGFPCRE-mediated removal of Smo in mutants carrying Smo floxed (f) alleles (ShhGFPCRE/ Smo f/f ).The ShhGFPCRE/Smo f/f mutants die shortly after birth due to unknown causes.Littermates not carrying the ShhGFPCRE knockin and/or the Smo floxed alleles and heterozygotes, which are viable and phenotypically normal, were used as controls.Tongues from embryos carrying the R26R transgene (ShhGFPCRE/R26R) showed robust CRE activity in the LE (S1A-S1D Fig) .Probes targeting Ptch1 and Gli1, which are readouts of Hedgehog activity [6,9,[17][18][19], confirmed the absence of Ptch1 and Gli1 transcripts in the ShhGFPCRE/Smo f/f LE, indicating efficient Smo ablation in this tissue.By contrast, and as expected, the LM retained endogenous hedgehog signaling (S1E-S1H' Fig) .Consistent with these findings, RT-qPCR revealed downregulation of Ptch1 in ShhGFPCRE/Smo f/f tongues (S1I Fig).
To assess the impact of epithelial SMO ablation on tongue development, we compared control and ShhGFPCRE/Smo f/f tongues at E17.5, E18.5, and shortly after birth (postnatal day P0).Alcian blue van Gieson staining and immunohistochemistry for Keratin 8 (K8) revealed striking defects within the LE and its derivatives in the ShhGFPCRE/Smo f/f cohort with 100% penetrance.Specifically, the oral tongue, a territory normally bereft of glands, exhibited heterotopic serous and sero-mucous glands (Fig 1C -1F).Furthermore, the LE was hyperplastic, lacked filiform papillae (Fig 1C The ShhGFPCRE/Smo f/f tongue undergoes Merkel cell metaplasia Closer examination revealed abnormal formation of K8-positive (+) (Fig 2A and 2B) and SOX2+ (Fig 2C and 2D) solitary cells in the basal layer of the LE in all ShhGFPCRE/Smo f/f mutants examined at E17.5 to P0.We postulated that these cells could be stray TB cells or ectopic epithelial tactile cells called Merkel cells, as both cell types express high levels of Sox2 and Krt8 [20][21][22].Merkel cells do not express Shh (Fig 2P) [21], whereas all TB cells express SHH at least up to the perinatal period (Fig 2E and 2F) [4].K8/SHH immunostaining revealed absence of SHH+ solitary cells in the basal layer of the ShhGFPCRE/Smo f/f LE and showed that the K8+ solitary cells were SHH-negative (-) (Fig 2F ; S1 Table ), suggesting that these were not TB cells.To further assess this notion, we immunostained sections for Homer1, a scaffolding protein found in skeletal muscle and brain [23] which we serendipitously identified as a reliable TB marker in the course of this study; Homer1 expression in the LE of control and ShhGFPCRE/Smo f/f tongues is not only restricted to the TBs, but also readily detectable in all TB cell types (Fig 2G -2J; S1 Table ), differing from other marker proteins that are only expressed in subsets of TB cells [1,4].Neither orthotopic K8+ Merkel cells in whisker pads (Fig 2Q) [21] nor the K8+ solitary cells in the basal layer of the ShhGFPCRE/Smo f/f LE (Fig 2K ; S1 Table ) were Homer1+.Thus, two TB markers, SHH and Homer1, excluded a TB cell type and supported a Merkel cell identity of the solitary cells in the mutant LE.To confirm that these cells were Merkel cells, we examined expression of Rab3c, a protein known to be enriched in Merkel cell cytoplasmic vesicles (Fig 2R) [21].We found that while the basal layer of the LE in controls as well as K8+ TBs in controls and ShhGFPCRE/Smo f/f mutants were Rab3c(-), the ectopic K8+ solitary cells within the ShhGFPCRE/Smo f/f LE were Rab3c+ (Fig 2L and 2M; S1 Table ).These data show that the K8+, SOX2+, and Rab3c+ cells were indeed ectopic Merkel cells, indicating that the LE, a tissue normally devoid of Merkel cells in humans and mice, underwent Merkel cell metaplasia upon loss of SMO.
In mice, Merkel cells develop in whisker pads and touch domes of hairy and glabrous skin [24] and become innervated postnatally [24,25].Remarkably, we found that the majority of the K8+ ectopic Merkel cells in the ShhGFPCRE/Smo f/f LE were innervated by NF-200+ neurites (Fig 2N ; S1 Table ), suggesting that they might be functional in this foreign environment.TBs in fungiform papillae are innervated by P2X2+ gustatory axons [26], however, none of the K8 + basally located solitary cells in the ShhGFPCRE/Smo f/f LE were innervated by P2X2+ axons (Fig 2O ; S1 Table ), further supporting our conclusion that these cells are ectopic Merkel cells and not TB cells.
Unlike in the glabrous skin where Merkel cells develop in the absence of SHH inputs, Merkel cell specification in the hairy skin is dependent on SHH signaling [27].Here we showed that in the absence of SMO, a subset of cells in the LE are diverted to a Merkel cell fate.Taken together, these observations suggest a context-dependent requirement for SHH signaling for Merkel cell specification.Polycomb genes have been shown to repress Sox2 which encodes a crucial factor for Merkel cell specification in the epidermis [28].However, the LE expresses Sox2 mRNA and protein from early stages of tongue development onwards [22,29], and the ectopic Merkel cells in the ShhGFPCRE/Smo f/f LE express SOX2 protein.These observations likely indicate that Polycomb genes are not effectors of the repression of Merkel cell formation in the LE.TBs and Merkel cells both degenerate after denervation [30,31] and share common markers such as Krt8, K20, Sox2, Advillin, Satb2, Cpe, Claudins 6/7, Mash1, Hes6, and Snap25  E,F) Sections of control (E) and mutant (F) tongues after double staining for K8 (green) [1,21,22,[32][33][34].This raises the possibility that a subset of TB progenitors converted into Merkel cells in the SMO-deficient LE.

Overproduction of oversized fungiform placodes upon epithelial loss of SMO
In vitro modulation of SHH signaling in the developing rodent tongue impinges upon FP patterning [12][13][14][15], however, it is currently unclear whether this is due to altered SHH activities in the LE, LM, or both.Furthermore, whether SHH has a role in FP patterning in vivo is still unknown.To address these points, we analyzed the effects of loss of epithelial SMO on FP development by assessing the expression patterns of Shh mRNA and protein, an established taste placode marker.At E12.5, the LE and LM showed robust and weak SHH immunostaining, respectively, in both control and ShhGFPCRE/Smo f/f tongues.The staining of the LM and interplacodal epithelium was the result of protein diffusion from taste placodes which produce SHH (Fig 3A and 3B).At E14, instead of being restricted to FPs as in control tongues, SHH staining was expanded in the LE of the ShhGFPCRE/Smo f/f tongues (Fig 3C and 3D).At E14.5 FP induction normally ceases, and Shh transcripts become confined to developing TBs [4].While at E14.5 and E15.5 the control tongues showed this expected Shh expression pattern, the ShhGFPCRE/Smo f/f tongues displayed a remarkable expansion of Shh expression (Fig 3E -3H) and overproduced enlarged Shh-expressing spots (FPs and/or TBs), some of which developed ectopically (Fig 3G -3L).Quantification in E15.5 mutant and control tongues revealed increased numbers of Shh+ spots in the mutants (147 ± 8 in mutants vs 101 ± 7 in controls; mean ± SD; P = 0.000).Thus, loss of SMO in the LE not only caused Merkel cell and glandular metaplasiae, but also engendered mispatterning of FPs as a result of their overproduction.Despite displaying FP patterning defects, the ShhGFPCRE/Smo f/f tongues developed K8+ TBs that were innervated by P2X2 neurites (Fig 3M and 3N; S1 Table ).
Temporal requirement for SHH inputs in the developing tongue for growth, morphogenesis, patterning and cell type specification Having uncovered a stringent requirement for epithelial SHH signaling for patterning and differentiation of the LE, we sought to decipher the role of SHH inputs in the LE and LM by assessing the impact of loss of SHH production.We generated ShhCreER T2 /Shh f mutant mice carrying both a Shh floxed allele and the tamoxifen (TAM) -inducible ShhCreER T2 knockin allele.Without exposure to TAM, the ShhCreER T2 /Shh f mice are viable and phenotypically normal.Ablation of Shh in the mutants was induced at E10.5.Immunostaining and in situ hybridization confirmed that TAM exposure led to a dramatic decrease of SHH protein in the LE and severe downregulation of Gli1 expression in the LE and LM (Fig 4A -4D) of the ShhCreER T2 /Shh f tongues, indicating loss of SHH signaling.This was further confirmed by RT-qPCR for Ptch1 (Fig 4O).Upon dissection, we found that relative to their control and Sonic Hedgehog (SHH; red).TBs are K8+/SHH+.The K8+ ectopic Merkel cells in the mutant are SHH-negative (-).(G,H) Anti-Homer1-stained (purple) sections of control (G) and mutant (H) tongues showing Homer1(+) TBs.(I-K) Sections of control (I) and mutant (J,K) tongues after double staining for K8 (red) and Homer1 (green), showing K8+/Homer1+ TBs (I,J) and K8+/Homer1(-) ectopic Merkel cells (K).(L,M) Sections of control (L) and mutant (M) tongues after double staining for K8 (red) and Rab3c (green).The ectopic Merkel cells (arrowheads in M) are K8+/Rab3c+, whereas TBs are K8+/Rab3c(-).littermates, the E10.5 TAM-induced ShhCreER T2 /Shh f mutants developed abnormally small tongues (microglossia) with a bifid tip as well as epithelial foci of squamous hyperplasia (Fig 4A -4F).Of note, bifid tip of the tongue has been described in patients with Ellis-van Creveld syndrome, which is caused by mutations of the EVC and EVC2 genes that lead to decreased Hedgehog signaling [35][36][37][38].In situ hybridization with a Shh riboprobe targeting both deleted (exon2) and non-deleted (exon1) Shh-coding sequences [39] revealed enlarged Shh+ spots, likely TBs or FPs, some of which were ectopically located along the midline of the E10.5 TAM-induced ShhCreER T2 /Shh f  ).However, unlike the ShhGFPCRE/Smo f/f tongues, the E10.5 TAM-induced ShhCreER T2 /Shh f tongues were exempt from glandular metaplasia, i.e. no heterotopic glands developed in the oral tongue (Fig 4L ), and displayed only a few scattered ectopic Merkel cells (Fig 4I and 4J).Thus, SHH is required for growth and morphogenesis of both the LE and LM, but it is dispensable for FP induction.
Remarkably, in ShhCreER T2 /Shh f mutants first exposed to TAM at E11.  ).Thus, unlike in the central nervous system where SHH can act as an axonal chemoattractant [41], SHH is not required for local guidance of gustatory neurites towards fungiform papillae.
To further dissect the role of SHH in lingual cell fate specification, we analysed Wnt1-CRE/ Smo f/f mutants lacking SHH signaling specifically in the LM.As shown previously [6], the Wnt1-CRE/Smo f/f tongues were abnormally small and cleft, and we found that they developed TBs and were free from glandular metaplasia (S2A-S2D Fig) .The overlapping tongue anomalies in the E10.5 TAM-induced ShhCreER T2 /Shh f mutants and the Wnt1-CRE/Smo f/f mutants demonstrate that SHH activity in the LM is necessary for tongue growth and morphogenesis but not for TB induction.The lack of glandular metaplasia in the oral tongue of the Wnt1-CRE/ Smo f/f mutants as well as in the E10.5 and E11.5 TAM-induced ShhCreER T2 /Shh f mutants, as compared to the ShhGFPCRE/Smo f/f mutants, suggests the requirement for SHH-dependent mesenchymal inputs for ectopic gland formation.This situation is akin to that of the skin where induction of glandular metaplasia was found to require SHH signaling in the mesenchyme [19].
We also found that proper tongue development necessitates SHH inputs during a restricted period since the ShhCreER T2 /Shh f mutants and the ShhCreER T2 /Smo f/f mutants developed normal tongues following an initial TAM exposure at E12.  [42], and taste placodes as well as TBs do not express K14 [29].Together, these findings point to a restricted period before E12.5 during which loss of SHH signaling induces tongue defects and support previous findings in rat tongue organ cultures [15].

De novo taste placode formation upon loss of SHH signaling
The active phase of FP patterning during which FPs are induced normally occurs between E12.5 and E14, and the FPs subsequently form TBs. Strikingly, from E15.5 onwards, unlike in We next used a probe targeting Wnt10b which, like Shh, marks taste placodes [4].In contrast to control tongues, which were devoid of Wnt10b transcripts after E14.5, the ShhGFPCRE/Smo f/f as well as the E10.5 and E11.5 TAM-induced ShhCreER T2 /Shh f tongues abnormally exhibited Wnt10b expression separated by Wnt10b(-) regions within the basal layer of the LE (Fig 6K -6P).That the basally located placodes expressing Shh mRNA and protein (Shh/SHH) and Wnt10b were overlain by Shh/SHH(-) and Wnt10b(-) suprabasal epithelial cells (Fig 6A -6P) supports de novo induction of FPs.These data show sustained formation of taste placodes upon loss of SHH signaling, at least up to birth, and are reminiscent of the finding that K14-CRE/Smo f/f mutants undergo de novo hair follicle formation [19].Since de novo taste placode induction occurred in the ShhGFPCRE/Smo f/f and TAMinduced ShhCreER T2 /Shh f mutant tongues, it could be concluded that loss of SHH signaling in the LE is the main cause of this phenomenon.
Our work here utilized a genetic approach to reveal that upon loss of SHH signaling, the mutant tongues transcended the time when FP induction normally ceases (E14.5) and formed de novo taste placodes.This phenomenon was not observed in previous studies using rodent tongue organ cultures, since in those studies the assessment of the impact of loss of SHH signaling on taste placode formation was limited to the active phase of FP patterning [12][13][14][15].Furthermore, our findings that in vivo SHH inhibits FP induction during the active period of FP patterning not only concur with prior work in vitro [12][13][14][15], but they also demonstrate that SHH signaling is required to prevent sustained taste placode induction and that SHH operates strictly within the LE to fulfill these functions, without inputs from the LM.Similarly, genetic studies have demonstrated that Wnt/β-catenin signaling within the LE is crucial for FP induction [1].Loss of Wnt/β-catenin activity downregulates Shh in FPs and Ptch1 in FPs and their associated mesenchyme, whereas activation of Wnt/β-catenin signaling engenders expansion of Shh expression [14,43].Moreover, findings from tongue organ cultures suggest that SHH inhibits Wnt/β-catenin activation, whereas Wnt/β-catenin signaling induces its own inhibitor, i.e.SHH [14].However, how SHH impinges upon Wnt/β-catenin activity is unknown.Our findings of ectopic and anachronic induction of Wnt10b expression in mutants lacking SHH inputs suggest that SHH antagonizes Wnt/β-catenin signaling at least in part through inhibition of Wnt10b expression.Further supporting this notion is the finding that SHH inhibits Wnt10b expression in cultured teeth and whisker pads [44].

Loss of Shh signaling during tongue development enhances retinoic acid signaling
The ShhGFPCRE/Smo f/f tongues not only underwent glandular metaplasia, but they also formed von Ebner's glands displaying mucous metaplasia.We postulated that the dramatic tongue anomalies caused by disrupted SHH signaling were a result of enhanced retinoic acid (RA) signaling.Components of the RA pathway are expressed during tongue development [45][46][47][48][49][50], and it is well-established that excess RA or vitamin A (a RA precursor) induces glandular/mucous metaplasia of various epithelia [51][52][53][54][55][56].Strikingly, rodent embryos overexposed to RA/vitamin A exhibit tongue anomalies phenocopying those occurring in the Shh n/n [57] and the various mutants described in this work, including aglossia, microglossia, bifid tongue, and adhesion of the LE to epithelia of the oral mucosa [58][59][60][61].
To test our hypothesis, we first assessed the expression of RARb and RARg, well-established transcriptional targets of RA signaling that encode the nuclear retinoic acid receptors RARβ and RARγ, respectively [62].We found that relative to controls, the LE of the ShhGFPCRE/ Smo f/f mutants (Figs 7A-7D' and S3A-S3L) as well as that of the E10.These data show that the LE and its derivatives produce RARγ protein and that RA signaling is enhanced in the LE upon loss of SHH inputs.
Enhanced RA signaling upon loss of SHH signaling during tongue development is due to loss of SHH-dependent Cyp26a1/c1 expression Transcripts encoding the RA synthesizing enzymes RALDH1-3 have been shown to be expressed in the LE at E14.5 and E16.5 [49].To assess the distribution patterns of RALDH1-3 during tongue development and to determine whether they are altered in the ShhGFPCRE/ Smo f/f tongues, we immunostained sections from controls and mutants from E11.5 onwards.We found that RALDH1-3 were expressed in the LE from E11.5 onwards as well as in the orthotopic and heterotopic lingual glands (S5 and S6 Figs).We also found that FPs were RALDH3+ (S5Q and S5R RA signaling is tightly regulated by activities of RALDH1-3 and the RA catabolic enzymes belonging to the CYP26 family [63,64].As the ShhGFPCRE/Smo f/f mutant tongues had no major alterations in RALDH distribution, we reasoned that the increased RA signaling could be caused by decreased production of CYP26A1 and CYP26C1 which are expressed in the LE [65][66][67].We found that Cyp26a1/c1 transcripts were dramatically reduced in developing Cyp26b1, another CYP26 family member, is expressed exclusively in the LM [65].We found that Cyp26b1 levels (Fig 9U ) in the mutant tongues were either essentially unaltered (ShhGFPCRE/Smo f/f and E10.5 TAM-induced ShhCreER T2 /Shh f ) or upregulated (E11.5 TAM-induced ShhCreER T2 / Shh f ), confirming enhanced RARb and RARg levels in the LE of the mutants.Unlike in both the ShhGFPCRE/Smo f/f and in the E10.5 and E11.5 TAM-induced ShhCreER T2 /Shh f mutant tongues, Cyp26a1/c1 levels were unaltered in E12.5 TAM-induced ShhCreER T2 /Shh f mutant tongues (Fig 9O -9R).Thus, a critical temporal window of SHH activity is necessary for expression of Cyp26a1/ c1 in the LE.These data, together with our findings indicating a temporal window during which loss of SHH signaling causes FP mispatterning, suggest that early loss (before E12.5) of SHH signals is necessary for downregulation of Cyp26a1/c1 expression in the LE during FP patterning (E12.5-E14) and for the entailing patterning defects.
While our data indicate that Cyp26a1/c1 expression in the LE requires SHH inputs, it is unlikely that the transcripts are direct targets of SHH signaling.At E11.5, despite the presence of significant SHH activity in the tongue primordium [6,8], only weak Cyp26a1/c1 signals were detectable (Fig 9A) [67].Furthermore, treatment of tongue/mandible explants with SAG, a SMO agonist, enhanced Cyp26a1 signals in the LE but failed to induce ectopic Cyp26a1 expression in epithelia of the pharyngeal tongue and mandible (Fig 9S and 9T).These observations suggest that SHH signaling is required for maintenance and/or reinforcement of Cyp26a1/c1 expression in the LE but not for their initial induction.
The virtual absence of Cyp26a1/c1 transcripts in taste placodes and TBs amid high expression levels in the interplacodal LE is a conundrum in view of the data indicating the requirement of SHH signaling for Cyp26a1/c1 expression.However, taste placodes and TBs, similar to other SHH-producing signaling centers including the notochord, neural floor plate and tooth enamel knots [18,68,69], did not exhibit higher levels of Ptch1 and Gli1 than the nearby Shhnegative-SHH-responsive tissues.Furthermore, Gli1-lacZ mice revealed that Shh-expressing cells in taste placodes and papillae are less SHH-responsive than the neighbouring epithelial and mesenchymal cells [70].Moreover, taste placodes, the notochord, and tooth enamel knots are induced after loss of Shh or Smo [our present data; 9,18,40,71].In light of these observations, we propose that SHH signaling is attenuated in taste placodes and TBs and that this, combined with inputs from yet to be identified factor(s), leads to absence of Cyp26a1/c1 in these structures.controls, to a large extent, appear as punctae, whereas in the LE of the mutants they appear as globular aggregates, indicating higher levels of RARg expression in the mutants relative to the controls.Epithelial foci in the mutants with weaker RARg signals than the rest of the LE (arrow in H').(I-K) RT-qPCR analysis for RARb and RARg relative to Actb (β-actin).(I) Upregulation of RARb (P = 0.0000) and RARg (P = 0.0000) levels in tongues from E13.5 ShhCreER T2 /Shh f mutants (n = 6) as compared to tongues from controls (n = 6) first exposed to TAM at E11.5.(J) Upregulated RARb (P = 0.0000) and RARg (P = 0.0003) levels in tongues from E13.5 ShhCreER T2 /Shh f mutants (n = 6) as compared to tongues from controls (n = 6) first exposed to TAM at E10.5.(K) Upregulation of RARb (P = 0.0015) and RARg (P = 0.0000) levels in tongues from E12.5 ShhGFPCRE/Smo f/f mutants (n = 6) as compared to tongues from controls (n = 6).The data are mean values ± SD.CvP, circumvallate papilla; T, tongue.Scale bars: 500 μm (A-H) and 50 μm (A'-H').
https://doi.org/10.1371/journal.pgen.1006914.g007RA signaling is required for orthotopic lingual gland formation, and enhanced RA activity causes glandular but not Merkel cell metaplasiae in the SMO-deficient LE That RA signaling is activated upon loss of SHH signaling strongly suggests that the glandular metaplasia, i.e. cell fate switch prompting the LE in the ShhGFPCRE/Smo f/f oral tongue to form heterotopic glands, is a direct consequence of hyperactivation of RA signaling.Indeed, it is well-established that excess RA and Vitamin A induces glandular/mucous metaplasia of various epithelia [51][52][53][54][55][56].To functionally test this scenario, we cultured E11.5 tongues and the surrounding mandibular structures in the presence of retinoids and inhibitors of RA signaling.We found that treating explants from control embryos with retinoids, including all-trans RA, CD1530 (RARγ selective agonist) and CD2314 (RARβ selective agonist) induced formation of K8+ heterotopic glands in the oral tongue, mimicking the ShhGFPCRE/Smo f/f in vivo phenotype (S7A-S7E Fig; S3 Table ).ShhGFPCRE/Smo f/f tongue explants treated with vehicle (DMSO) recapitulated the in vivo defects, including glandular and Merkel cell metaplasiae as well as formation of K8(-) epithelial foci (S7F and S7I Fig; S3 Table).Importantly, exposure to BMS493 (pan-RAR antagonist) and DEAB (RA synthesis inhibitor) inhibited heterotopic gland formation in the ShhGFPCRE/Smo f/f explants (S7F-S7K Fig; S3 Table).Neither BMS493 nor DEAB inhibited lingual Merkel cell metaplasia in the mutant explants (S7H and S7J Fig) .Thus, lingual Merkel cell metaplasia is not caused by enhanced RA signaling.Furthermore, BMS493 and DEAB failed to inhibit development of the K8(-) foci in the mutants (S7H and S7J Fig), consistent with our findings suggesting that these are subjected to low RA signaling owing to persistent Cyp26a1/c1 expression.To determine whether RA signaling is required for development of von Ebner's glands and posterior lingual glands, which normally develop in Cyp26a1/c1-free domains, we treated E12.5 control tongues/mandibles with DMSO or BMS493 (S3 Table ).While the DMSO-treated explants formed K8+ von Ebners' glands and K8+ posterior lingual glands, the BMS493-treated explants failed to develop von Ebners' glands and displayed developmentally arrested posterior lingual glands (S7L-S7M' Fig; S3 Table ).These data demonstrate that abnormal RA activation in the ShhGFPCRE/Smo f/f tongues causes glandular metaplasia and that development of von Ebners' glands and posterior lingual glands requires RA inputs.They also show that the LE responds to RA.Despite this evidence and well-established expression of components of the RA pathway during tongue development, various transgenic mice reporting RA activity failed to show transgene activity in the LE and its derivatives [72][73][74][75].Likewise, we were unable to detect RA activity in the LE except in subsets of desquamating cells of controls and ShhGFPCRE/Smo f/f mutants carrying the RARE-hsp68-LacZ transgene, another reporter of RA activity [76]  inactive in the LE and its derivatives.Accordingly, several studies pointed to lack of accuracy of such transgenic mice for visualization of RA activity in various tissues [77][78][79].

RA signaling promotes fungiform placode formation
We found that loss of SHH signaling elicits enhanced development of FPs amidst increased RA activity.This strongly suggests that RA signaling promotes FP formation.To test for direct evidence of the phenomenon, we treated tongue explants from control embryos with retinoids and found that they induced development of oversized Shh+ placodes (

Conclusions
In this study, we adopted in vivo genetic approaches and experimental manipulation in tongue organ cultures in vitro which enabled us to reveal a temporal requirement for SHH signaling for proper tongue development.Before E12.5, SHH signaling in the LE controls patterning and cell fate specification, whereas in the LM, SHH inputs are required to drive growth and morphogenesis (S9 Fig) .We uncovered unanticipated relationships between SHH and RA signaling during patterning of the LE (S9 Fig) .We discovered that genetic loss of SHH activity in the LE leads to enhanced RA signaling and that in the absence of SHH inputs, the LE overproduces oversized FPs and TBs and can even undergo de novo taste placode formation.We showed that Cyp26a1/c1 expression in the interplacodal epithelium requires SHH inputs and that retinoids or RA signaling inhibitors induce or inhibit FP formation, respectively.These findings provide compelling evidence that SHH inhibits FP induction by abating RA activity in the interplacodal epithelium.
It is well-established that RA signaling plays a central role in controlling antero-posterior patterning of the limb, hindbrain, gut, and trunk [80][81][82][83].Our genetic studies showed that loss of SHH signaling in the LE not only leads to formation of heterotopic glands in the oral tongue, but also causes overgrowth and mucous metaplasia of von Ebner's glands.Furthermore, in tongue explants retinoids and RA inhibitors induced and inhibited lingual gland formation, respectively.These findings strongly suggest that antero-posterior patterning of the LE is controlled by antagonistic SHH and RA activities, where SHH inhibits and RA promotes lingual gland formation in the oral and pharyngeal tongues (S9 Fig) .To our knowledge, neither de novo taste placode induction nor cell fate switch causing glandular and Merkel cell metaplasiae in the oral tongue, as in the ShhGFPCRE/Smo f/f mutants, have been reported in mouse models or in experimental studies in vitro, suggesting that SHH and RA pathways play key roles in cell fate specification within the LE (S9 Fig) .Our findings may provide insight into the etiopathogenesis of Merkel cell carcinoma of the tongue which remains an enigma [84].We found that genetic loss of SHH signaling in the LE causes ectopic expression of Shh amidst abnormal activation of RA signaling and that retinoids induce expansion of Shh expression in tongue organ cultures.As normal levels of Shh expression in FPs require Wnt/β-catenin inputs [14,43], these observations raise the question as to the relationships between RA and Wnt/β-catenin signaling during FP patterning.The Hedgehog and RA pathways are crucial for development and homeostasis of a vast array of tissues and organs, and deregulated Hedgehog signaling generates neoplasia [10,11,[81][82][83].Our findings thus constitute a basis for future research aimed at deciphering the interactions between these pathways during development and disease, a largely unexplored topic.

Ethics statement
The experiments using mice were reviewed and approved by the Animal Research Ethics Committee in Go ¨teborg, Sweden (Dnr.230-2010, 174-2013 and 40-2016).Mouse studies were also carried out under approved protocols in strict accordance with the policies and procedures established by the University of California, San Francisco (UCSF) Institutional Animal Care and Use Committees (UCSF protocol AN084146).Mice were euthanised by cervical dislocation.

Mouse lines and tamoxifen induction of alleles
The different transgenic or knockin mice used in this work include the K14-CRE [40], ShhGF PCRE and ShhCreER T2 [85], Wnt1-CRE [6], and R26R [86] mice.We crossed K14-CRE, ShhGFPCRE or Wnt1-CRE males with females carrying the Shh or Smo floxed alleles (Shh f/f and Smo f/f ) as described previously [17,18,40,87,88].Thereafter K14-CRE/Smo +/f , ShhGFPCRE/ Smo +/f , ShhCreER T2 , ShhCreER T2 /Smo +/f , or Wnt1-CRE/Smo +/f males were crossed with Shh f/f or Smo f/f females to generate mutants.K14-CRE, or ShhGFPCRE males were crossed with R26R females to generate reporter embryos carrying the CRE and R26R alleles (K14-CRE/R26R and ShhGFPCRE/R26R).The RARE-hsp68-LacZ transgenic mice were as described [76].We generated ShhGFPCRE/Smo +/f /RARE-hsp68-LacZ males which were thereafter crossed with Smo f/f females to generate ShhGFPCRE/Smo f/f /RARE-hsp68-LacZ mutants and controls (without the CRE and/or the Smo floxed alleles).ShhGFPCRE/Smo +/f (or without CRE) without the RARE-hsp68-LacZ transgene were used as negative controls.For tamoxifen (TAM) induction of the ShhCreER T2 /Shh f and ShhCreER T2 /Smo f/f mutants, pregnant females received intraperitoneal injections of TAM (dissolved in corn oil) every other day (excluding the day of embryo harvest) with the first injection consisting of 0.2 mg/g body weight (bw) TAM and the subsequent injections consisting of 0.1 mg/g bw.All mutants and their control littermates were identified by genotyping using DNA from the tip of the tail or a piece of embryonic tissue and/or, when possible, by their external phenotypes.

Histology, immunostaining, in situ hybridization, Lac-Z staining and GFP imaging
The number of tongues from control and mutant embryos processed for in vivo studies is described in S1 Text.Heads or mandibles/tongues from the various mutants and their control littermates were prepared for sectioning or used as whole-mounts.For histological staining (alcian blue van Gieson), immunohistochemistry, and immunofluorescence, specimens were fixed in ethanol-acetic acid as previously described [89,90] or in 4% paraformaldehyde in 1X PBS (PFA/PBS).For in situ hybridization, specimens were fixed in 4% PFA/PBS at 4˚C.The specimens were thereafter processed for whole-mount in situ hybridization or embedded in paraffin.Six-μm paraffin sections were used for immunostaining and in situ hybridization.For histology, immunostaining, and in situ hybridization the specimens were processed as previously described [18,19,[89][90][91].Sections were counterstained with Richardson's azure II-methylene blue and methyl green after in situ hybridization and immunostaining, respectively.For green fluorescent protein imaging, tongues from ShhGFPCRE/+ (controls) and ShhGFPCRE/ Smo f/f mutants were fixed overnight in 1% PFA/PBS at 4˚C, rinsed in PBS, and imaged as whole-mounts.For visualization of LacZ activity, specimens were fixed overnight in 2% PFA/ PBS at 4˚C.Whole-mounts and cryostat tissue sections were subsequently processed for βgalactosidase histochemistry as previously described [19,90].

Quantification of taste buds and ectopic Merkel cells
Quantification of Shh-expressing taste buds/taste placodes.The number of Shh-expressing taste buds/taste placodes was determined by counting Shh+ spots in micrographs of tongues processed for whole-mount in situ hybridization.Photoshop's pencil tool was used to mark the spots counted.Quantification was performed in E15.5 control (n = 2 from the same litters as the mutants and n = 3 from other litters) and ShhGFPCRE/Smo f/f mutant (n = 3) tongues, as well as in E15.5 control (n = 3) and ShhCreER T2 /Shh f mutant (n = 4) tongues from embryos first exposed to tamoxifen at E11.5.Student's t-test was used for statistical analyses.
Quantification of Keratin 8-expressing (K8) taste buds and ectopic Merkel cells after double immunofluorescence for various markers.Ribbons consisting of 3-4 parasagittal sections of paraffin-embedded tongues were collected onto slides.Each slide held 3-4 ribbons of sections, and each ribbon on a given slide was taken every 234-240 μm.Taste buds (TBs) and ectopic Merkel cells were counted in one section per ribbon.All the other TBs and Merkel cells (different from those already counted) that were visible in the remaining sections were also included in the quantification.The number of K8+ solitary cells in the basal layer of the lingual epithelium (ectopic Merkel cells) and the number of K8+ TBs expressing Sonic Hedgehog (SHH), Homer1 and Rab3c were determined after K8/SHH, K8/Homer1 and K8/Rab3c immunostaining.The number of K8+ TBs and K8+ ectopic Merkel cells that were innervated by gustatory neurites was determined after K8/P2X2 immunostaining.The number of ectopic Merkel cells that were innervated by NF-200+ neurites was determined after K8/NF-200 immunostaining.The number of tongues from control and mutant embryos processed for immunofluorescence is described in S1 Quantification of K8-expressing cell clusters in fungiform papillae.Paraffin-embedded tongues from E18.5 control (n = 3) and ShhCreER T2 /Shh f mutant (n = 3) embryos first exposed to tamoxifen at E11.5 were sectioned in their entirety in sagittal/parasagittal planes.After immunohistochemistry for K8, K8+ cell clusters in fungiform papillae were counted in all serial sections and averaged per section.Student's t-test was used for statistical analyses.
'-1D'), and displayed overgrown von Ebner's glands transformed into a sero-mucous type (Fig 1G and 1H; see also S6 Fig), as well as epithelial foci virtually devoid of K8, Keratin 6, and Keratin 15 immunostaining (Fig 1I-1N).7.6% of mutants also showed adhesions between the LE and epithelia of the floor of the oral cavity (Fig 1F).Taken together, these findings indicate that an altered antero-posterior patterning and differentiation program in the SMO-deficient LE is translated into development of glands in the oral tongue, abnormal development of von Ebner's glands, and formation of epithelial regions exhibiting a molecular signature different from that of the normal LE.

Fig 5 .
Fig 5. Tamoxifen induction at E11.5 causes overproduction of oversized, innervated taste buds in ShhCreER T2 /Shh f mutant tongues.(A-N) Analyses of tongues from control and ShhCreER T2 /Shh f mutant embryos first exposed to tamoxifen at E11.5.(A,B) In situ hybridization with a
Fig), whereas TBs were RALDH1+ (S5S-S5T and S5Y-S5Z Fig).Thus, the LE and its derivatives in controls and mutants are exposed to RALDH-derived RA.However, besides abnormal enhancement of RALDH3 and RALDH2 staining in the LE of the mutants at E11.5 (S5F Fig) and E16.5 (S5V Fig), respectively, there were no major alterations in RALDH1-3 distribution patterns in the mutants.

Fig 7 .
Fig 7. Upregulation of RARb and RARg expression in ShhGFPCRE/Smo f/f and ShhCreER T2 /Shh f mutant tongues, indicating enhanced retinoic acid signaling.(A-H') RARg in situ hybridization with oligonucleotide probes (brown) in parasagittal tongue sections.The dotted lines highlight the junction between the lingual epithelium and lingual mesenchyme.(A-D') Sections from E12.5 (A,B) and E13.5 (C,D) control (A,C) and ShhGFPCRE/Smo f/f mutant (B,D) embryos.(A'-D') are enlarged images of the boxed areas in (A-D).(E-F') Sections from E13.5 control (E) and ShhCreER T2 /Shh f mutant (F) embryos first exposed to tamoxifen (TAM) at E10.5.(E') and (F') are enlarged images of the boxed areas in (E) and (F), respectively.(G-H').Sections from E14 control (G) and ShhCreER T2 /Shh f mutant (H) embryos first exposed to TAM at E11.5.(G') and (H') are enlarged images of the boxed (red) areas in (G) and (H), respectively.The inset in (H') is an enlarged image of the boxed (green) area in (H).The RARg hybridization signals in the lingual epithelium (LE) of the

Fig 8 .
Fig 8. Dramatic downregulation of Cyp26a1 and Cyp26c1 expression in the lingual epithelium of ShhGFPCRE/Smo f/f and ShhCreER T2 /Shh f mutant tongues.(A-N') Parasagittal tongue (T) sections from controls and mutants after in situ hybridization with a 35 S-UTP-labelled riboprobe for Cyp26c1 (A-F') and oligonucleotide probes for Cyp26a1 (G-N'; black).Genotypes are indicated in the panels.(A-D) Dark-field images showing Cyp26c1 expression (shiny dots) in E12.5 (A,B) and E14 (C,D) control and ShhGFPCRE/Smo f/f mutant tongues.(A'-D') Enlarged bright-field images (hybridization signals appear as black dots) of the boxed areas in (A-D).(E,F) Dark-field images showing Cyp26c1 expression in tongue sections from E14 control and ShhCreER T2 / Shh f mutant embryos first exposed to tamoxifen (TAM) at E10.5.(E') and (F') are enlarged, bright-field images of the boxed areas

Fig 9 .
Fig 9. SHH signaling is required for maintenance/reinforcement but not for induction of Cyp26a1 and Cyp26c1 expression.(A-R) In situ hybridization for Cyp26c1 and Cyp26a1 in the developing tongue (dark blue/brown).(A,B) E11.5 control (A) and ShhGFPCRE/Smo f/f mutant (B) tongues showing Cyp26c1 expression in the control tongue (arrow in A) and absence of Cyp26c1 expression in the mutant tongue (arrow in B). (C,D) Cyp26c1 expression in E12.5-E13 control (C) and ShhGFPCRE/Smo f/f mutant (D) tongues.(E,F) Cyp26c1 expression in tongues from E14-E14.5 control (E) and ShhCreER T2 /Shh f mutant (F) embryos first exposed to tamoxifen (TAM) at E10.5.The inset in (E) is an enlarged image of the specimen in (E).(G,H) Cyp26a1 expression in E14 control (G) and ShhGFPCRE/Smo f/f mutant (H) tongues.(I,J) Cyp26a1 expression in tongues from E14.5 control (I) and ShhCreER T2 /Shh f mutant (J) embryos first exposed to TAM at E10.5.(K-N) Cyp26a1 (K,L) and Cyp26c1 (M,N) expression in tongues from E15 control (K,M) and ShhCreER T2 /Shh f mutant (L,N) embryos first exposed to TAM at E11.5.Severe downregulation of Cyp26a1/Cyp26c1 expression in all mutants, except in epithelial foci (ef).Fungiform placodes or taste buds, which appear as unstained spots (C,G, inset in E), and the circumvallate papilla (CvP) are Cyp26a1/ Cyp26c1-negative.Artefact due to loss of the thin epithelium of the intermolar eminence during tissue processing (arrowheads in G, I and K).(O-R) Unaltered Cyp26c1 (O,P) and Cyp26a1 (Q,R) expression in tongues from E14.5 control (O,Q) and ShhCreER T2 /Shh f mutant (P,R) embryos first exposed to TAM at E12.5.(S,T) Cyp26a1 expression in tongue explants from E12 control embryos after in vitro culture for 2 days with DMSO (S) and 200 nM SAG (T).SAG Fig 10C-10M; S4 Table).Furthermore, explants exposed to CD2314 and CD1530 exhibited ectopic Shh expression along the tongue midline and enhanced Shh signals, including in enlarged spots within two domains flanking the prospective intermolar eminence (Fig 10J, 10L and 10M).By contrast, explants from controls (Fig 10A and 10B; S4Table) and ShhGFPCRE/Smo f/f mutants (Fig 10N-10Q; S4 Table) exposed to BMS493 or DEAB exhibited large areas with weak or no Shh signals.Taken together with our findings of occurrence of RA signaling in taste placodes (expression of RALDHs amid absence of Cyp26a1/c1), these data show that RA promotes taste placode formation.