Recovery of taste organs and sensory function aftersevere loss from Hedgehog/Smoothened inhibitionwith cancer drug sonidegibArchana Kumaria, Alexandre N. Ermilovb, Marina Grachtchoukb, Andrzej A. Dlugoszb,c,1, Benjamin L. Allenc,1,Robert M. Bradleya,1, and Charlotte M. Mistrettaa,1,2

aDepartment of Biologic and Materials Sciences, University of Michigan School of Dentistry, Ann Arbor, MI 48109; bDepartment of Dermatology, Universityof Michigan Medical School, Ann Arbor, MI 48109; and cDepartment of Cell and Developmental Biology, University of Michigan Medical School, Ann Arbor,MI 48109

Edited by Linda M. Bartoshuk, University of Florida, Gainesville, FL, and approved October 11, 2017 (received for review July 19, 2017)

Striking taste disturbances are reported in cancer patients treatedwith Hedgehog (HH)-pathway inhibitor drugs, including sonidegib(LDE225), which block the HH pathway effector Smoothened (SMO).We tested the potential for molecular, cellular, and functionalrecovery in mice from the severe disruption of taste-organ biologyand taste sensation that follows HH/SMO signaling inhibition.Sonidegib treatment led to rapid loss of taste buds (TB) in bothfungiform and circumvallate papillae, including disruption of TBprogenitor-cell proliferation and differentiation. Effects were selec-tive, sparing nontaste papillae. To confirm that taste-organ effects ofsonidegib treatment result from HH/SMO signaling inhibition, westudied mice with conditional global or epithelium-specific Smo dele-tions and observed similar effects. During sonidegib treatment,chorda tympani nerve responses to lingual chemical stimulation weremaintained at 10 d but were eliminated after 16 d, associated withnearly complete TB loss. Notably, responses to tactile or cold stimulusmodalities were retained. Further, innervation, whichwasmaintainedin the papilla core throughout treatment, was not sufficient to sustainTB during HH/SMO inhibition. Importantly, treatment cessation led torapid and complete restoration of taste responses within 14 d asso-ciated with morphologic recovery in about 55% of TB. However,although taste nerve responses were sustained, TB were not restoredin all fungiform papillae even with prolonged recovery for severalmonths. This study establishes a physiologic, selective requirementfor HH/SMO signaling in taste homeostasis that includes potentialfor sensory restoration and can explain the temporal recoveryafter taste dysgeusia in patients treated with HH/SMO inhibitors.

taste receptor cell | fungiform papilla | circumvallate papilla |taste regeneration | chorda tympani nerve

Cancer patients treated with Hedgehog (HH) pathway in-hibition (HPI) drugs experience severe taste disturbances (1–

5). The Food and Drug Administration-approved HPI drugsonidegib (LDE225) blocks HH signaling at the Smoothened(SMO) receptor (Fig. 1A) (6). We had reported that oral gavagewith sonidegib in mice for 16 or 28 d results in progressive fungi-form papilla (FP) taste-organ disruption, taste bud (TB) loss, andelimination of taste nerve responses to chemical stimuli (7). Nowwe present data for the initial time course of HPI effects on tasteorgans and cell types and on sensation; study of Smo deletion; andthe presence of the HH ligand in the nerve fibers of taste organs.Importantly, the potential for and nature of recovery from HPIeffects in taste organs and taste neurophysiology are demonstrated.Lingual taste organs are composed of (i) the TB, about 80–

100 modified epithelial/neuroepithelial cells of several types;(ii) the sensory innervation to the TB and surrounding epithe-lium; and (iii) the papilla residence, a stratified epithelium over acore of stromal cells and connective tissue elements (8). Precisecontrols within each compartment are essential to integrate thecell processes that support taste organs and sensation. Here weinvestigated homeostatic lingual taste organ regulation by HH/SMO

signaling in FP and circumvallate (CV) papillae and tested thehypothesis that morphologic and functional recovery would occurfollowing cessation of HH/SMO inhibition by sonidegib treatment.Taste cells turn over continuously in adults (9, 10), but TB are

eliminated when essential signaling pathways, including HH, aredisrupted (11–13). However, the taste cells are resilient in specificcontexts, as illustrated by degeneration after nerve cut in adultmammals and regeneration as nerves grow back into the tastepapilla epithelium (14). The taste organ therefore is ideally suitedto explore fundamental questions about the biology of tissue ho-meostasis and renewal, the mechanisms that sustain sensation inconstantly renewing organs comprised of heterogeneous cell types,and the potential restoration of organs and function after TBelimination resulting from deregulation of essential controls.Although the HH pathway is a major regulator in several organ

systems (15), there is sparse fundamental knowledge about thefunction and timing of HH signaling in taste organs, especially themechanisms of cell regulation, the regulation across anterior andposterior tongue tissues, and the effects on taste sensation per se.Whereas transcriptional-level regulation of HH/GLI signaling hasbeen studied by genetic manipulation of Gli2 (11), the conse-quences of HH signal disruption at the cell surface remain largelyunexplored, although most pharmacologic HH inhibitors act atthis level (16).

Significance

Hedgehog pathway-inhibitor drugs effectively treat basal cellcarcinoma, a common skin cancer. However, many patientstaking such drugs report severe taste disturbances that impairtheir quality of life. To understand the biology behind theseadverse effects, we studied the consequences of Hedgehogpathway inhibition on taste organs and neural sensation in mice.Taste bud progenitor-cell proliferation and differentiation werealtered, resulting in taste bud loss. Nerve responses to lingualtaste stimuli were also eliminated, while responses to touch andcold stimuli remained. After stopping Hedgehog pathway in-hibition, taste buds and sensory responses recovered. This studyadvances our understanding of Hedgehog signaling in tastehomeostasis and the reported taste recovery after clinicaltreatments with Hedgehog pathway-inhibiting drugs.

Author contributions: A.K., A.N.E., A.A.D., B.L.A., R.M.B., and C.M.M. designed research;A.K., A.N.E., R.M.B., and C.M.M. performed research; M.G. contributed new reagents/analytic tools; A.K., A.N.E., A.A.D., B.L.A., R.M.B., and C.M.M. analyzed data; and A.K.,A.A.D., B.L.A., R.M.B., and C.M.M. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Published under the PNAS license.1A.A.D., B.L.A., R.M.B., and C.M.M. contributed equally to this work.2To whom correspondence should be addressed. Email: chmist@umich.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1712881114/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1712881114 PNAS | Published online November 13, 2017 | E10369–E10378

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 2

3, 2

021

SMO is the core signal transduction component of HH signaling(Fig. 1A) (17, 18). In HH signaling, pathway activity is repressed inthe absence of ligand via the transmembrane receptor PTCH,which represses the pivotal HH signaling protein SMO (19–21).HH binding to PTCH1 blocks its inhibition of SMO; downstreamsignaling is then initiated, resulting in the modulation of GLItranscription factors (GLI1, GLI2, GLI3) and leading to thetranscription of target genes, including Gli1 and Ptch1 (Fig. 1A).In the tongue, HH pathway components are in distinct compart-ments of adult taste papillae and TB (Fig. 1B), including activelyproliferating TB progenitor cells (8, 13). The secreted SHH ligandin epithelium is produced principally within TB cells (13, 22),whereas the HH targets, Ptch1 and Gli1, are expressed withinperigemmal cells around the TB, in basal cells of the stratifiedepithelium that lines the papilla core, and in stromal or connectivetissue cells of the papilla core (Fig. 1B) (11, 13). Paracrine sig-naling from SHH in TB cells to surrounding epithelial and con-nective tissue cells has been proposed (13, 23).Sonidegib and other HPI drugs block the HH pathway at SMO

and are effective in treating patients with advanced BCC (3, 16,24–26). However, these patients experience dysgeusia and ageusiathat compromise treatment adherence and quality of life (4, 5, 27,28). There are reports that patients have some recovery 2–3 moafter stopping treatment (5), but the nature and extent of thisrecovery are unclear. Regenerative potential after HPI drugtreatment has not been explored in animal models, so the basicbiology of patient recovery is not known.Here we used the cancer drug sonidegib in mice, in parallel with

conditional Smo deletion targeting the whole body or epithelium,to test the main site of inhibitory effects and discern the mecha-nisms for HH/SMO inhibition in FP and CV taste organs and insensory responses from the chorda tympani nerve that innervatesTB in the FP. Further, we assessed taste organs and nerve re-sponses for periods of several months after cessation of HPI drugtreatment to determine whether recovery is possible. We demon-strate coordinated cell proliferation and differentiation regulatedby HH/SMO signaling in taste papillae and TB, selective regula-

tion of oral sensory modalities of taste, touch, and temperature,and the recovery of taste organs and sensation. Our data provideinsight into the regenerative biology and clinical consequences inpatients treated with sonidegib who experience dysgeusia.

ResultsTreatment with HPI Drug Sonidegib Alters FP Taste-Organ MorphologyWithin 10 D. Before testing recovery from HPI drug treatment,it was important first to determine the temporal aspects ofHH/SMO signaling inhibition in mice gavaged with sonidegib for5–36 d. We quantified effects by characterizing FP and TB mor-phology as category I (typical FP/TB), II (atypical FP/TB), or III(atypical FP/no TB) (Fig. 1C). Compared with vehicle-treatedmice, there were no differences in FP/TB categories after 5 d ofsonidegib gavage (Fig. 1D). However, after just 10 d of sonidegibtreatment the proportion of category I FP (typical FP/TB) FP wasreduced by half, and category II FP (atypical FP/TB) had in-creased several fold to 40% of all FP, compared with ∼2% invehicle treatment. After 16 d of sonidegib gavage only ∼1% of allFP were category I (typical FP/TB) FP; category II (atypical FP/TB) FP comprised almost 70%, and category III (atypical FP/noTB) FP accounted for ∼30% of all FP (Fig. 1D). With 28 d ofsonidegib gavage, category I (typical FP/TB) were essentiallyeliminated, replicating our prior results (7); notably, after 36 d,virtually all FP were category III, atypical FP/no TB (Fig. 1D).Therefore, in a detailed time course, HPI drug duration of 10–16 deliminated typical FP and TB, whereas 5 d was not effective, andin progressive effects by 36 d only atypical papillae that lack TB(category III) remained on the tongue. (Statistics for data in Fig.1D are given in Fig. S1A.)Importantly, we quantified all FP numbers in a standard 600-μm

region of vehicle- and sonidegib-treated tongues and found noreduction in papilla density across treatments (Fig. S1B). That is,FP were not eradicated, and we were not categorizing fewer FP atlong treatment durations. Rather, these duration-specific resultssuggest that HPI drug effects are related to the disruption ofphysiologic cell renewal in taste organs.

Fig. 1. Sonidegib alters FP and TB morphology andreduces all TB cell types. (A) The HH pathway in OFFand ON signaling and INHIBITED at SMO by sonidegib.(B) Schematic: FP and filiform papillae with HH sig-naling components. (C) H&E staining for FP/TB cate-gories. Dotted lines demarcate TB or regions of TBremnants. Blue dashed surface lines indicate epithelialkeratinization in FP; a red dashed line indicates epi-thelial keratinization in filiform papilla in type IIIatypical FP/no TB. (Scale bar: 50 μm.) (D) Percentage ofFP/TB categories: typical FP/TB (category I), atypical FP/TB (category II), and atypical FP/no TB (category III),after vehicle or sonidegib treatment. Bars are mean ±SEM. Numbers of tongues per group are in paren-theses. Brackets indicate significant differences fortreatment durations (two-way ANOVA and Tukey’sHSD post hoc tests); ###P < 0.001 for vehicle vs. soni-degib treatments. Complete F and P values are givenin Fig. S1A. (E) Antibody detection of TB cell types:NTPdase2 (type 1), PLCβ2 (type 2), SNAP25 (type 3)after vehicle or sonidegib treatment. Cell nuclei areidentified with DAPI (blue). (Insets) TB cell types 2 and3 were eliminated after 28 d of sonidegib treatment.Type 1 cells (NTPdase2) were eliminated after 36 d.(Scale bar: 50 μm.) (F) Number of TB cell types after16 or 28 d of sonidegib treatment compared withvehicle treatment. Data are mean ± SEM. Numbers oftongues are in parentheses in bars. Brackets indicatesignificant differences (one-way ANOVA with Tukey’sHSD post hoc tests). Complete F and P values are givenin Fig. S1C.

E10370 | www.pnas.org/cgi/doi/10.1073/pnas.1712881114 Kumari et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 2

3, 2

021

To further understand the progression of TB loss in FP, weused immunostaining to test whether all TB cell types weresusceptible to sonidegib treatment or whether specific taste celltypes (type 1: NTPDase2+; type 2: PLCβ2; type 3: SNAP25+)(29) were lost in a particular sequence or time course or wereretained. Whereas cell types 2 and 3 were essentially eliminatedin TB after 28 d, type 1 cells were not completely lost until 36 dof sonidegib gavage (Fig. 1 E and F). The type 1 cells are mostnumerous in the TB and have a long turnover time, up to 24 d(29–31), which could account for their longer retention after HPItreatment. However, turnover for no TB cell type was resistant tothe HPI drug, suggesting effects on progenitor(s) for all TB cells.(Complete statistics for data in Fig. 1F are in Fig. S1C.)Cell proliferation and cell death. To test whether HPI altered turn-over in lingual epithelium, we studied cell proliferation within FPand the nontaste filiform papillae. In FP we designated three re-gions, basal, apical, and perigemmal zones of FP basal cells (Fig.2A, vehicle), and quantified Ki67+ cells for proliferation. Cells inall these regions include TB progenitors (13, 32). In basal cells ofthe apical FP wall there were substantial reductions in Ki67+ cellsas early as 3–5 d of HPI treatment (Fig. 2 A and B), that is evenbefore the FP/TB organs had noticeable morphological disruption(Fig. 1D). This early decrease in the number of Ki67+ TB pro-genitors in the FP was sustained after 16 and 28 d of drug treat-ment (Fig. 2 A and B). Alterations in proliferating cell numberswere not observed in basal or perigemmal regions of FP epithelialcells (Fig. 2 A and B). Note that perigemmal counts were madeonly within category I and category II FP, which have TB or TBremnants, because there were no TB in type III FP (Fig. 2A, 28 dand Fig. 2B). Notably, there was no reduction in Ki67+ cells in thebasal cells of filiform papillae (Fig. 2 A and B).We did not observe differences in the extent or distribution of

cell death markers [TUNEL assay, Cleaved caspase 3 (CC3)antibody] in FP or filiform papillae of vehicle- or sonidegib-

treated tongues (Fig. S2 A and B). That is, accelerated or alteredcell death did not contribute to TB cell loss in HPI.Cell differentiation. When TB were eliminated from the FP aftersonidegib gavage, as seen in category III FP (atypical FP/no TB),the papilla apex acquired a particular conical shape (Figs. 1C and2A, 28 d, Inset). To learn whether TB progenitor cells were withinthis conical FP region, we used immunostaining for the basal cellmarker keratin 5 (K5). K5+/K14+ cells are among the HH-responding progenitors for TB cells in FP (13, 32), and K5+/K14+

cells in CV differentiate into mature K8+ TB cells (33). In vehicle-treated mice a few cells within the TB were dual-labeled for the TBcell marker K8 (34) and K5 (Fig. 2C). After sonidegib treatment for5, 10, and 16 d there was a reduction in K8+ TB cells and an in-crease in K5+ cells in the perigemmal region; a few cells within theTB were dual-labeled for K8 and K5 (Fig. 2C, Insets). However, at28 d in category III FP (atypical FP/no TB), where there were noTB or TB remnants, the FP had K5+ cells in suprabasal layersthroughout the conical apex (Fig. 2C, 28 d sonidegib). Thus, K5+

cells occupy tissue territory where TB had been located.We also assessed the expression of K14+ and p63+ cells in the

FP apex and observed both K14+ and p63+ cells in suprabasalcells of the apex after 28 d of sonidegib treatment (Fig. S2C), aswith K5 cells. Ki67+ cells were sometimes found in the FP apexbut generally were not suprabasal (Fig. S2D). As the K5+/K14+/p63+ cells in the tongue differentiate to TB, their expressions arelost (32). After sonidegib treatment, at the apex of category IIIFP (atypical FP/no TB), the suprabasal expression of K5+/K14+/p63+ points toward a differentiation defect from a normal pro-gression of K5+/K14+/p63 expression to K8+ cells.

Smo Deletion Mimics HPI Drug Effects on FP Taste Organs. To es-tablish that the effects observed in sonidegib-treated mice reflectedthe blockade of SMO, the HH signaling effector targeted by thedrug, we generated mice to conditionally (doxycycline-regulated)delete Smo globally (R26M2rtTA/+;tetO-cre;Smofl/fl, referred to as

Fig. 2. Sonidegib treatment reduces apical epithelialcell proliferation in FP and leads to the accumulationof suprabasal K5+ cells. (A) Immunofluorescent anti-body detection of E-cadherin (Ecad, red) and Ki67(green) in basal epithelial cells of FP and filiform pa-pillae after vehicle or 5 d, 16 d, or 28 d sonidegibtreatments. The image for FP, vehicle, shows threeregions for quantifying Ki67+ cells (apical, basal, andperigemmal). (Inset) Twenty-eight days sonidegib il-lustrates the absence of Ki67+ perigemmal cells insome category III FP (atypical FP/no TB). The vehicleimage for filiform papillae illustrates Ki67+ cells, clus-tered at the base of rete ridges (arrows). White dottedlines indicate the surface of the epithelium. (Scale bar:50 μm, also applies to C.) (Magnification: Inset, 0.9×.)(B) Number of Ki67+ cells in FP regions in vehicle- andsonidegib-treated mice. Counts in the perigemmalregion were done only when TB or TB remnantsremained in the papilla, i.e., in category I (typical FP/TB) or category II (atypical FP/TB) FP. In filiform pa-pillae Ki67+ proliferating cells were quantified in a300-μm length of tongue. Numbers of tongues are inparentheses. Statistical analysis was performed withone-way ANOVA with Tukey’s HSD post hoc tests(*P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001). F and P valuesare shown in the table at the right of the graphs.(C) Antibody detection in FP of K5 (green, arrowspoint to K5+ cells in TB or TB remnants) in TB cellsafter vehicle or 5, 10, 16, or 28 d sonidegib treatment.(Insets) K5 coexpression with a few K8+ TB cells (red)indicated by arrows. (Magnification: Insets, 3.5×.)

Kumari et al. PNAS | Published online November 13, 2017 | E10371

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 2

3, 2

021

“cSmoKO” mice) or in the epithelium (K5rtTA;tetO-cre;Smofl/fl,referred to as “cSmoKO-epi” mice), diagrammed in Fig. 3A. IncSmoKO mice, the category I FP (typical FP/TB) were reduced toless than 10% of all FP after 16 d of Smo deletion (Fig. 3B). Withepithelial deletion of Smo, in cSmoKO-epi mice, there were noeffects at 5 d after gene deletion, but after 16 d only 15% of FPwere category I (typical FP/TB) (Fig. 3B). After 24 d, more than40% of all FP were category III (atypical FP/no TB), which iscomparable to the 16-d effects in cSmoKO mice. Therefore themajor target cell population on which sonidegib acts to alter FPand TB is likely to be epithelial. Statistical analyses for the data inFig. 3B are in Fig. S3A. The FP density did not alter with time ineither Smo deletion model (Fig. S3B).Overall the data from two cSmoKO models are similar in time

course, extent, and effects after sonidegib treatment. Further,when comparing sonidegib and Smo-deletion experiments a sim-ilar phenotype was seen in category III FP (atypical FP/no TB)(Fig. S3C, H&E). That is, a heavily cornified, conical papilla apexwas observed with a discernible cell collection where the K8+ tastecells had been located. As with sonidegib treatment, in the cate-gory III papilla after Smo deletion, K5+ cells occupied the conicalpapilla apex (Fig. S3C, K5/K8). With HPI, therefore, K5+ cellsreplace differentiated cells of the apical FP that typically includeTB cells and accumulate as suprabasal, stratified epithelial cells.

After Smo Deletion, SHH+ and K8+ Cells Are Reduced in FP, and HH-Responding Cells Are Eliminated from the Epithelium but Remain inthe Connective Tissue Core. Associated with the loss of TB cells(using K8 as a taste-cell marker), the SHH+ TB cells also werereduced in cSmoKO and cSmoKO-epi mice (Fig. 4A, SHH/K8).Furthermore, Gli1lacZ-positive HH-responding cells were notretained in the FP epithelium, in either the perigemmal cells orbasal cells of the FP epithelial wall (Fig. 4A, Gli1lacZ). In thepapilla core, however, HH-responding cells remained but werereduced in number after global Smo deletion compared withcontrols. The remaining HH-responding cells in the FP core ofcSmoKO mice might relate to noncanonical signaling; alterna-

tively, the efficiency of Smo deletion in stromal cells might beless than in epithelial cells.As HH signaling activity was eliminated in the epithelium,

there was an apparent decrease in Ki67+ proliferating cells in thebasal cell layer of the apical papilla wall but not in the basalregion (Fig. 4B, Ecad/Ki67). This demonstrates a disruption inthe supply of a subset of TB cell progenitors.

After Smo Deletion, Nerves Remain Within the FP Connective Tissue Core.Whereas TB cells were absent after Smo deletion in both cSmoKOand cSmoKO-epi mice, innervation remained within the FP corefrom the lingual nerve that typically innervates the FP walls andperigemmal regions (neurofilament, NF+, nerve fibers) and fromthe chorda tympani nerve that typically innervates TB cells (P2X3+taste nerve fibers) (Fig. 4C, K8/NF and K8/P2X3). In addition,X-Gal staining for Gli1lacZ-positive cells plus NF immunofluores-cence demonstrates a close physical association of nerves and HH-responding cells in the FP connective tissue core in control andcSmoKO mice (Fig. 4D), suggesting potential interactions betweenthe papilla innervation and stromal cells. Furthermore, the resultsdemonstrate that nerve fibers sustained within the FP are not ableto maintain TB in the context of HPI effects in lingual epithelium.

HPI Leads to Decreased TB in CV on Posterior Tongue.We found thatsonidegib gavage and Smo gene deletion each substantially re-duced TB and altered morphology in ectoderm-derived FP of theanterior tongue. To test whether HH/SMO inhibition could alsoaffect TB and morphology of the CV (an organ with endodermalderivation), we extended studies to include a major posterior-tongue taste organ. We used H&E staining to quantify CV sizeand TB numbers. TB were identified as cell collections with a tastepore (Fig. 5A, Inset, arrows) or as collections of cells with orwithout a taste pore, termed “TB profiles” (Fig. 5A, Inset, dottedlines encompass TB profiles). CV depth (all sections in which thepapilla appeared, multiplied by section thickness) and the lengthof papilla walls (Fig. 5A, bars denote wall length measure) wereassessed for size determinations. Across sonidegib-treated andSmo-deletion mice, there was not an apparent gross alteration inCV form (Fig. 5B). Nor were depth or wall length different, exceptin cSmoKO-epi mice, in which CV depth was reduced over timeand there was a trend for wall length to be shorter (Fig. 5C).(Statistical analyses for data in Fig. 5C are presented in Fig. S4A.)Importantly, in pharmacologic and genetic HPI models,

numbers of TB profiles were reduced, and numbers of TB withpores decreased by 25–75% after 16 d (Fig. 5D). Notably, theeffects were more pronounced in Smo-deletion mice than aftersonidegib gavage. In the CV of cSmoKO-epi mice we examinedearly-time-point effects and found that the number of TB withpores was not reduced after 5 d treatment, but there was a sig-nificant decrease after 16 d (Fig. 5D). This is similar to effects inFP/TB. (Detailed statistics for Fig. 5D are given in Fig. S4B.)Overall, the TB loss is substantial and demonstrates that HPI

affects the CV as well as FP taste organs and, therefore, can altertwo major sets of TB on the anterior and posterior tongue. Althoughnot all of the more than 100 TB in the CV were totally eliminated,as they were in the FP, which have a single TB each, the effects aresubstantial. In assessing the efficacy of HPI in cSmoKO-epimice, wefurther observed that Gli1lacZ-positive HH-responding cells wereeliminated in the CV epithelium but not from the stroma (Fig. 5Eand Fig. S4C). Similar to effects in the FP, HH-responding cellswere reduced in the stroma in cSmoKO mice (Fig. S4C). Further-more, even with loss of TB cells, innervation to the CV was retainedin physical association with HH-responding cells (Fig. S4C).

Recovery of FP and TB Cells and Molecular Phenotypes When SonidegibTreatment Is Discontinued. To test potential FP taste-organ resto-ration after HPI, we gavaged mice with sonidegib for 16 d andthen discontinued treatment and maintained mice for recoveryperiods of 7, 14, or 21 d and for 3, 5, or 9 mo (Fig. 6A). A recoveryperiod of 7 d was not sufficient to restore papillae to category Imorphology (typical FP/TB); rather most papillae were atypical, in

Fig. 3. Smo deletion alters FP morphology and reduces TB. (A) Diagramsshowing the conditional deletion of Smo from all tissues (cSmoKO) or fromepithelial cells and progeny (cSmoKO-epi). The gray/shaded region on thecSmoKO-epi diagram indicates normal Smo expression. (B) Percentage ofcategory I (typical FP/TB), category II (atypical FP/TB), and category III (atypicalFP/no TB) FP in cSmoKO or cSmoKO-epi mice. Bars are mean ± SEM. Numbersof tongues are in parentheses. Brackets indicate significant differences (two-way ANOVA with Tukey’s HSD post hoc tests); ###P ≤ 0.001 for control vs.cSmoKO or cSmoKO-epi. Complete F and P values are given in Fig. S3A.

E10372 | www.pnas.org/cgi/doi/10.1073/pnas.1712881114 Kumari et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 2

3, 2

021

category II or III (Fig. 6B). Remarkably, however, 14 d weresufficient to restore about 50% of all papillae to category I mor-phology, and only 5% or fewer of all FP were category II FP(atypical FP/TB). About 40% of all papillae were categoryIII (atypical FP/no TB).We used prolonged recovery times to determine whether an in-

creased proportion of FP was restored to category I after severalmonths. Notably, the proportions of typical FP/TB ( category I) andatypical FP/no TB (category III) did not change from 14 d through9 mo of recovery (Fig. 6B). This suggests that a proportion of FP isnot resilient and cannot recover after HPI drug treatment, whereasanother FP subset recovers and maintains TB after sonidegibtreatment is discontinued. Details of FP/TB morphology duringtreatment and the recovery transition period from 7 to 14 d afterstopping sonidegib treatment, and after 9 mo of recovery are seenwith H&E staining in Fig. S5A. The category III (atypical FP/noTB) phenotype persists throughout treatment and recovery. Thiscategory, which loses all TB cells, apparently cannot recover fromHPI effects, whereas category II FP (atypical FP/TB), which retainTB cell remnants, presumably can recover.Statistical analysis for FP/TB categories across recovery in Fig.

6B is presented in Fig. S6A, and additional data are included todemonstrate that FP numbers per tissue region do not differacross sonidegib treatment and recovery periods (Fig. S6B).To test whether all taste cell types in the FP/TB recover after

discontinuing sonidegib treatment or whether some types wereresistant to recovery, we used immunostaining and quantified celltypes. As noted, all cell types were eliminated after the HPI drug(Fig. 1D), but when treatment was discontinued the three cell typeswere restored, with similar time courses, after 14 d in category I(typical FP/TB) and category II (atypical FP/TB) FP to the valuesseen in vehicle-treated FP (Fig. S6C). Category III FP (atypical FP/no TB), were not quantified because no TB were present.During recovery from HPI treatment, which occurs between

7 and 14 d after stopping sonidegib gavage, there was an increasein SHH+ and K8+ cells in the reconstituted taste buds (Fig. 6C,SHH/K8), associated with recovery of Gli1lacZ-positive cells inthe perigemmal region and FP epithelial walls (Fig. 6C, Gli1lacZ).A gradual restoration of Gli1lacZ-positive cells was apparent inthe varied expression patterns seen at 7 d recovery, when HH-responding cells were observed in perigemmal regions only orpartially patterned in perigemmal cells and papilla walls (Fig. 6C,7 d recovery and Inset). Overall, the HH ligand and respondingcells returned to FP/TB in concert at 14 d after drug cessation.Recovery of TB in the CV. TB in the posterior tongue CV papillaewere decreased after 16 d of sonidegib treatment (Fig. 5). Whensonidegib gavage was discontinued, there was no restoration ofTB with pores after 7 d (Fig. S5B), but there was a trend towardincreased numbers at 14 d. By 3–9 mo of recovery the numbers of

TB with pores were not different from the numbers in vehicle-treated mice. Therefore, there was recovery from HPI effects inthe CV as well as in the FP.Innervation in the FP. Innervation, shown with NF+ and TB-specificP2X3+ fibers, was retained within the FP core during sonidegibtreatment and also was obvious and robust during recovery fromHPI (Fig. 7 A and B). The nerve fibers were observed even incategory III papillae (atypical FP/no TB) that have no taste buds(Fig. S7A) and in FP after prolonged HPI for 36 d (Fig. S7B).There were no detectable differences in the extent of innervationin papillae treated with vehicle or the HPI drug (Fig. S7C).However, nerves alone could not maintain or restore TB in FPwith epithelial effects of HH/SMO inhibition.We found that HH-responding cells and nerve fibers remain

within the FP connective core after HPI and continue in recoveryperiods (Figs. 6C and 7 A and B). Nerve fibers and neurons in thedorsal root ganglion are a source of SHH in signaling to Gli1-expressing cells in the hair follicle (35). To determine whetherfibers in the FP and their soma express SHH, we studied genic-ulate ganglion (GG) and trigeminal ganglion (TG) neurons in Shhreporter mice. SHH expression has been reported in TG neurons(36), and we confirmed that finding (Fig. 7C). Importantly, wefurther observed Shh expression within GG neurons and Shh+

fibers within the tongue and FP and projecting into TB (Fig. 7C).To test whether SHH+ neurons were altered after HH pathwayinhibition, we used immunoreactions for SHH in ganglia aftervehicle or sonidegib treatment and quantified SHH+ neurons withNeuN coexpression in the GG. No differences were detected (Fig.7D). Although not quantified, similar results were seen in the TG(Fig. 7E). This indicates that a GG ganglion source of SHH ligandremains during pathway inhibition, although the lingual epithelialSHH associated with TB cells is eliminated. The GG- and TG-derived SHH sources could potentially maintain HH signaling inpapilla stroma via secretion from nerves.

Responses to Chemical Stimuli from the Chorda Tympani NerveRecover After Sonidegib Treatment Is Stopped, and Tactile andCold Responses Are Maintained Throughout. To determine earlysensory effects after HPI and to test whether taste sensation isrestored in concert with taste-organ recovery, we recorded fromthe chorda tympani nerve that innervates TB in FP and stimu-lated the tongue with chemicals including salts, acids, sucrose,and quinine to elicit taste responses. Compared with vehicletreatment, after 10 d of sonidegib treatment the taste responseswere fully maintained (Fig. 8). Therefore, although 10 d of HPIdrug treatment effectively reduced the proportion of category I(typical FP/TB) FP to about 60% of the proportion in vehicletreatment (Fig. 1D), neurophysiological responses were not al-tered. However, after 16 d of sonidegib, responses to chemicals

Fig. 4. Smo deletion leads to reduced TB, HH ligand,HH-responding cells and reduced proliferation in FP,while innervation is retained. (A) Immunofluorescentantibody detection of SHH (red) and K8 (green) in TBcells; X-Gal staining (blue) of Gli1lacZ-positive cells fromcontrol, cSmoKO, and cSmoKO-epi mice. (Insets)Merged images for SHH and K8. No TB cells remainin the cSmoKO image (white box, Inset). (B) Anti-body detection of E-cadherin (Ecad, red) and Ki67(green) in control and Smo deletions. Ki67+ cellswere reduced in apical FP walls shown by white bars.(C) Antibody detection of K8+ (red) TB cells andnerves (NF, green or P2X3, green) in cSmoKO orcSmoKO-epi mice. Arrows indicate P2X3 fibers in theFP core. (Scale bar: 50 μm, all images.) (D) Antibodydetection of nerves. NF (green) with X-Gal staining(blue) of Gli1lacZ in control and 16-d cSmoKOmice. (Inset)Enlarged image of the boxed region shows the associa-tion of Gli1lacZ-positive, HH-responding cells with nerves(NF+) in FP stroma (arrows). Dotted lines in in A, C, and Dindicate the basal lamina. (Magnification: D, Inset, 5×.)

Kumari et al. PNAS | Published online November 13, 2017 | E10373

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 2

3, 2

021

were essentially eliminated, replicating our prior data for 16-dtreatment (7), and category I (typical FP/TB) papillae were re-duced to less than 3% of all FP (Fig. 1D).We further tested whether neural responses could be restored

after HPI that had effectively eliminated intact FP taste organs.When sonidegib treatment was stopped after 16 d, a recovery pe-riod of 7 d was not sufficient to restore chorda tympani nerve re-sponses to lingual taste stimuli (Fig. 8). Nor is this recovery periodsufficient for restoring category I (typical FP/TB) papillae to thetongue in large proportion (Fig. 6B). In contrast, after 14 d recoveryand continuing through 9 mo, chorda tympani nerve responses totaste stimuli were fully restored. However, during this period thecategory I (typical FP/TB) papillae were ∼55% and category III(atypical FP/no TB) taste organs were ∼40% of all papillae (Fig.6B). Therefore, normal chorda tympani nerves responses can beobtained with only about 55% of intact FP taste organs recovered.(Quantified analysis for data in Fig. 8 is presented in Fig. S8A.)Modality-specific effects. Importantly, we also stimulated the ante-rior tongue with tactile (stroking with a wooden rod) and cold(water at 4 °C) stimuli. Although HPI for 16 d can eliminate chorda

tympani nerve taste responses, responses to tactile and cold mo-dalities were retained (Fig. 8), replicating our prior data (7). Thisparticularly noteworthy result demonstrates that HH differentiallyregulates oral sensory modalities and suggests that the receptororgans for touch and cold are not within the TB cells per se.The tactile and cold responses were sustained throughout all

the recovery periods, from 7 d through 9 mo (Fig. 8). We haveobserved Shh+ nerve fibers in ShhCre reporter mice that extendin apical FP outside the bounds of TB cells (Fig. 7C) and pro-pose that these might represent fibers that respond to touch and/or cold FP stimulation or fibers that connect to epithelial re-ceptors for tactile and temperature.Response/concentration series and chloride salt stimuli. We challengedthe taste organs and neurophysiological properties of the chordatympani nerve with ability to generate graded responses to a seriesof NaCl (from 0.05–1.0 M) during HPI and with recovery. Aftersonidegib treatment for 16 d, nerve responses were essentially notdiscernible, except as very small responses to 0.5 and 1.0 M NaCl(Fig. S9A). A 7-d recovery period after sonidegib treatment wasnot sufficient for restoration of responses. However, after 14 d

Fig. 5. Pharmacologic and genetic HH pathway blockade decreases TB in CV papillae. (A) H&E staining for CV/TBmorphology in vehicle-treated mice. Bars delimitthe CV walls, for measuring wall length. (Inset) An enlarged image of the boxed region illustrating TB profiles (dotted lines) and TB pores (arrows). (Scale bars:50 μm.) (B) H&E staining for CV/TB morphology after 16 d of HPI by sonidegib or whole-body (cSmoKO) or epithelial Smo (cSmoKO-epi) deletion. (Scale bar: 50 μmfor all images in A and B.) (C and D) Graphs for depth and wall length (C) and TB profiles and TB pores (D). Brackets in D denote significant differences (one-wayANOVA with Tukey’s HSD post hoc tests). Bars are mean ± SEM. F and P values are given in Fig. S4 A and B. (E) X-Gal staining (blue) for Gli1lacZ illustrates HH-responding cells in epithelium and stroma in control and 24-d cSmoKO-epi mice. Dotted lines outline the epithelium. (Scale bar: 50 μm for both images.)

E10374 | www.pnas.org/cgi/doi/10.1073/pnas.1712881114 Kumari et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 2

3, 2

021

graded responses to lingual stimulation with increasing concen-trations of NaCl were apparent (Fig. S9A). The timing of recovery,between 7 and 14 d, matches that for responses to a broad rangeof taste stimuli (Fig. 8). Further, response/concentration functionsdemonstrated typical curves after 14 d of recovery and through9 mo (Fig. S9B).We also used a series of 0.5 M chloride salts, which have

distinctive “tastes” and receptor mechanisms, after HPI and withrecovery. After sonidegib treatment the chorda tympani nervedid not respond to stimulation with Mg2+, Ca2+, K+, NH4

+, orNa+ chloride salts (Fig. S8B). However, recovery in responses tovalues comparable to vehicle treatment was apparent aftersonidegib had been discontinued for 14 d.Therefore, chorda tympani nerve responses to a range of chemical

taste qualities, a concentration series of NaCl, and high-concentrationdisparate chloride salts were eliminated by HPI but recovered by2 wk after treatment was discontinued. On the other hand, re-sponses to lingual stimulation with touch or temperature modal-ities were not affected by treatment with the HPI drug sonidegib.

DiscussionWe tested the potential for restoration of taste homeostasis afterHH signaling deregulation and show that taste organs and sen-sory responses can recover after severe loss from HH/SMO in-hibition with the cancer drug sonidegib. Indeed, after eliminationof TB and of chorda tympani nerve responses to taste stimuli,there is a remarkable restoration of TB in taste organs and full

recovery of taste nerve responses. TB elimination based on thederegulation of essential HH signaling controls for cell pro-liferation and differentiation and loss of all TB cell types is rapid,within 2 wk. Whereas nerves remain during HH/SMO inhibitionand recovery, they are not able to sustain or restore TB in thecontext of selective taste papilla epithelial effects. However, whensonidegib treatment is stopped, the taste organs are restored, andneural taste sensation returns to typical response patterns within2 wk. Although restoration of a full complement of TB numbers isnot achieved, the recovery of chorda tympani nerve responses canbe supported by about 55% of all FP taste buds. Redundancy inthe taste periphery has long been noted (37, 38) as being crucialfor life-essential sensations but has not previously recognized asimportant in recovery from HPI drug effects. We suggest that thereversible taste disruption in HPI-treated patients is, therefore, aspecific, directed effect reflecting the physiologic requirement forHH/SMO signaling in taste-organ homeostasis.Whereas HH signaling is required for TB maintenance and

restoration, it is not clear why some papillae can recover butothers cannot after HPI is stopped. We suggest that as long assome TB cells remain, and some HH-responding cells are in theepithelium (in category II, atypical FP/TB papillae), there ispotential for TB reconstitution. However, if TB are eradicatedalong with putative TB stem cells, then there is no TB restorationeven up to 9 mo after stopping treatment. This places TB stemcells within the TB. In fact, basal cells of the TB have been in-dicated as stem/progenitor cells (13, 22, 30). In addition, after

Fig. 6. Cessation of sonidegib treatment results inthe recovery of FP/TB morphology, SHH ligand, andHH-responding cells within 14 d and continuing upto 9 mo. (A) Time line for studying recovery from theeffects of 16 d sonidegib treatment. (B) Percentageof category I (typical FP/TB), category II (atypical FP/TB), and category III (atypical FP/no TB) FP after 16 dof sonidegib treatment followed by recovery periodsof 7, 14, or 21 d and 3, 5, or 9 mo. Examples of FP/TBmorphology are shown in Fig. S5A. Bars are mean ±SEM. Numbers in parentheses are number of miceanalyzed. Brackets denote significant differences fortreatment durations (two-way ANOVA with Tukey’sHSD post hoc tests); ###P ≤ 0.001 for vehicle vs.sonidegib treatments. Complete F and P values aregiven in Fig. S6A. (C) Antibody detection of SHH(red) and K8 (green, Inset) for TB cells and X-Galstaining (blue) for Gli1lacZ after 16 d of sonidegibtreatment and recovery for 7, 14, or 21 d. The Gli1lacZInset at 7 d recovery is included to illustrate thevariability in the recovery phenotype at 7 d. Dottedlines demarcate the basal lamina. (Scale bar: 50 μm.)(Magnification: C, Lower, Inset, 0.6×.)

Kumari et al. PNAS | Published online November 13, 2017 | E10375

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 2

3, 2

021

pathway inhibition we observed the loss of HH-responding cellsin FP walls, known regions of TB progenitors (13, 32); thus,another proposed progenitor/stem cell population is eliminated.We believe that there is more than one source of TB stem/pro-genitor cells and that these are regulated by HH signaling (8, 11).

HH Pathway Regulation of Proliferation and Differentiation. Withcomplementary pharmacologic and genetic approaches to inhibitSMO-dependent HH signaling, we found consistent phenotypesacross models. There were rapid effects on TB loss in bothectoderm-derived anterior tongue FP taste organs and endoderm-derived posterior tongue CV taste organs, but there were no dis-cernible effects on the nontaste filiform papillae. Thus, HH/SMOsignaling is an essential and selective regulator of gustatory epi-thelia and TB that turn over constantly in dynamic papilla organs.Cell cycles of the three TB cell types range from 3 to 30+ days (9,10, 31), with an average TB cell life span of 10 d (9). TB elimi-nation began after 10 d and derived from the loss of all three taste-cell types in FP and a reduction in proliferation among TB cellprogenitors in the apical wall of FP. HH/SMO signaling regulationof all TB cell progenitors is indicated.Progenitor differentiation to taste cells also is affected in HPI. As

TB cells are eliminated, the apical integrity of FP is compromisedby the acquisition of a heavily cornified, conical cap over epithelialcell clusters that are devoid of orientation and stain intensely withhematoxylin. We identified these cells as K5+ and K14+, charac-teristic of lingual and FP basal epithelial cells that include pro-genitors for TB (13, 32). Typically, in stratified epithelia the K5+

cells include transit-amplifying cells that leave the basal cell layerand differentiate to suprabasal cells that are not proliferative andexpress different keratins (39). In lingual papillae these differenti-ated cells include those of the TB, which are K8+ cells. Withprolonged HPI and attendant TB loss, the epithelial SHH ligandproduction is correspondingly eliminated. HH-responding cells are

reduced or eliminated. We propose that in the absence of HHsignaling, the K5+ and K14+ cells do not differentiate to K8+ cellsbut instead occupy the apical FP epithelial layers where the TB hadbeen located. Alternatively, K8+ cells could revert to K5+ expres-sion in a transdifferentiation process seen in other tissues (40).Overall, K5+ cell differentiation is disrupted when HPI deregulatestaste organs. Notably this notion is consistent with a genetic over-expression of SHH, where a role for HH signaling in regulating TBdifferentiation has also been proposed (41).The acquisition of a conical apex in the FP that accompanies

the loss of TB cells is characteristic of reduced HH pathwayactivity, as previously reported (7, 8) and as observed here aftersonidegib administration or Smo deletion. This reproduciblephenotype with HH signaling suppression or inhibition is remi-niscent of an acquired filiform papilla-like form observed in FPafter prolonged periods of denervation (42). A key feature in allthese experimental models is the loss of TB accompanied by theloss of HH-responding epithelial cells. This indicates that HHsignaling is a principal regulator not only of TB maintenance butalso of FP as distinct from filiform papilla morphology.

Functional Effects of HPI on Lingual Sensation. The rapid effects ofHPI on taste organs were associated with loss of neurophysio-logical taste responses. Here we found that the altered neuraltaste responses to stimuli representing all taste qualities relatedirectly to the loss of all taste-cell types. Further, effects wereobserved even though the TB were challenged with stimulationacross high concentrations of varied chloride salts that mighthave elicited nonspecific effects.Interestingly, however, typical taste nerve responses were still

obtained after 10 d of drug HPI, although only about 50% of FPwere category I (typical FP/TB) at this time point. Whereaschorda tympani nerve taste responses were absent after 16 d ofsonidegib gavage, it is notable that nerve responses to lingual

Fig. 7. Retained innervation during and after soni-degib treatment and identification of Shh expressionin GG and TG and in anterior tongue and FP nervefibers. (A and B). Immunofluorescent antibody de-tection of TB cells (K8, red) and nerves (NF, green, inA; or P2X3, green, in B). Even in category III (atypicalFP/no TB) FP, with no K8+ cells retained during soni-degib exposure or recovery periods, the innervationwas intact, as illustrated in Fig. S7A. Dotted lines in-dicate the basal lamina. Arrows in B point toP2X3 fibers. (Scale bars: 50 μm.) (C) RFP (red) expres-sion in constitutive ShhCre;R26RFP mice and aftertamoxifen (TAM) administration in ShhCreER;R26RFPmice. In both GG and TG, all cell bodies expressedShh. Anterior tongue TB include Shh+ cells and theirprogeny. Arrows denote Shh+ nerve fibers in theanterior tongue, FP core, and surrounding the TBreaching into the apical epithelium. Dotted lines de-marcate the basal lamina. (Scale bars: 50 μm.)(D) Immunofluorescent antibody detection of SHH(red) demonstrates SHH+ cells in GG in vehicle- orsonidegib-treated mice. Insets show merged andmagnified images for SHH and NeuN (green). (Scalebar: 50 μm.) The graph indicates that number ofSHH+ cells in four noncontiguous 10-μm sections ofGG did not change after 16 d sonidegib treatmentcompared with vehicle (t test; t = 1.32, P = 0.26). Barsrepresent mean ± SEM. Numbers in parenthesesare the number of mice per group. (Magnification:Insets, 2×.) (E) Immunofluorescent antibody de-tection of SHH (red) demonstrates SHH+ cells in TG.Insets show merged and magnified images for SHHand NeuN (green). (Scale bar: 50 μm.) (Magnification:Insets, 3×.)

E10376 | www.pnas.org/cgi/doi/10.1073/pnas.1712881114 Kumari et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 2

3, 2

021

tactile or cold stimuli were not altered. Previously we suggestedthat the receptors for responses to tactile and cold stimuli mustnot be TB cells per se, because these have been eliminated, and weindicated that mechanoreceptor or thermal receptor endingsmight be adjacent to TB (7). Now we have shown nerve endingsnext to TB that are Shh+. Studies of receptive fields of GG neu-rons (soma of chorda tympani nerve fibers) have demonstrated FPthat respond only to tactile or temperature stimuli and not taste(43). This suggests that there are GG/nerve fiber subsets, possiblywith end organs that are specifically somatosensory. HH signalingpresumably has different roles in maintaining epithelial sense or-gan specializations vs. receptor nerve endings.

FP Innervation and HH Signaling.Whereas TB are dependent on anintact innervation (14), the retention of nerves within FP in theface of TB elimination indicates that nerves alone are not suf-ficient to maintain TB in an epithelium that is devoid of SHHligand and HH-responding cells. Similarly, nerves in the CV la-beled with NF immunostaining are retained in the papilla corebut apparently are not able to sustain the full complement of TB.Notably these results substantiate our data after the suppressionof HH/GLI signaling (11).Here we have shown physical association between papilla in-

nervation and Gli1lacZ-positive HH-responding cells during ge-netic HPI in both FP and CV stroma. This affords an opportunityfor nerve interactions with HH-responding stromal cells in thepapilla, not previously proposed. SHH secreted from dorsal rootganglion nerve fibers activates signaling via Gli1+ HH-respondingcells in touch domes (44) and maintains homeostasis in the hairfollicle (35). Because we have observed SHH protein expression inneurons of the GG and TG that include the soma for nerves in-nervating FP and have observed Shh+ expression in fibers withinthe FP, we propose interactions between SHH+ nerve fibers andGli1lacZ-positive stromal cells in maintaining the FP structure.Furthermore, Shh+ fibers contact basal lamina components of theFP and remaining epithelial cells; these could participate in sig-naling interactions within a taste-organ niche (8).SHH likely has varied roles in taste-organ development and

maintenance. SHH is a morphogen in developing FP placodes (45)but might also serve as an axon-guidance molecule or chemo-attractant via noncanonical signaling (46, 47) to direct growingnerves specifically into the developing papillae (48). In the adulttaste organ, HH signaling is required for epithelial cell proliferationand differentiation in FP and TB homeostasis. We suggest thatSHH, retrogradely secreted from nerve fibers in the FP core, signalsto HH-responding cells and tissue elements in the connective tissue.

CV Papillae and HPI. In our current experiments, we did not recordfrom the glossopharyngeal nerve that innervates taste buds in theCV. Because we noted TB loss in the CV after 16 d of HPI drugtreatment and the recovery of about 80% of TB after 14 d, wepredict that any loss of posterior tongue sensation would recoveralso. In mice treated with the HPI drug vismodegib for 15 wk(49), statistically significant but biologically small effects wereseen on TB cell numbers in the CV, whereas we found a sub-stantial loss of TB (up to about half of control values) after 4 wkof sonidegib gavage. The only other reports of altered HH sig-naling on TB in the CV are from HH/GLI suppression in mice,where TB were substantially reduced/lost (11). Here we observedeffects in TB of both the FP and CV, that is, in two major lingualtaste regions. Although there were no reported behavioral ef-fects in two bottle preference/avoidance tests for sucrose ordenatonium benzoate in vismodegib-treated mice (49), we pre-dict behavioral effects after HH/SMO signaling inhibition.

Implications for Adverse Taste-Disturbance Effects in Patients Treatedwith HPI Drugs. This work is at the confluence of HH/SMO sig-naling as a principal regulator of taste organs and taste sensation,deregulated HH signaling in basal cell carcinoma (BCC), and BCCtreatment with drugs that inhibit HH signaling with associatedsevere taste disruption (4, 5, 26, 28, 50) that can alter patientquality of life (27). The demonstrated duration-dependent effectson TB in the FP and CV and the profound loss of chorda tympaninerve responses to all taste stimuli are the biological bases forsevere taste disturbances in patients. We found substantial effectsin anterior and posterior tongue TB in both the FP and CV, whichhave different innervation and chemical response profiles (51).Effects in patients taking HPI drugs therefore are likely to be wideranging across taste chemical stimuli and oral regions, and thereported ageusia in >20% of patients (5) is not unexpected. Fur-ther, based on our neurophysiological data, it is not likely that evenvery strong chemicals would elicit responses in patients duringtreatment with HPI drugs. However, our modality-specific resultssupport consideration of overall lingual sensation, and patient di-ets focused on the texture and temperature characteristics of nu-trients could temper the loss of taste sensation in overall flavorof meals.Notably, a majority of taste organs and TB regain cell in-

tegrity, and taste nerve responses rapidly return to normal pat-terns in the post-HPI drug recovery period in mice. Therefore,the reported recovery from taste disturbances after treatment isdiscontinued in patients (5) is related directly to the recovery oftaste organs and sensory responses. Importantly, management of

Fig. 8. Chorda tympani nerve responses to lingualstimulation with chemical, tactile, and cold stimuliafter HPI with sonidegib and recovery. Chemical re-sponse: Integrated electrophysiological whole-nerverecordings in response to chemical stimuli applied tothe tongue for 20 s followed by a distilled water rinsefor 30 s. The height at the steady-state integratedrecording for an individual stimulus above baselinewas measured as response. Tactile response: Whole-nerve, not integrated, responses to tactile stimuli,which consisted of five strokes with a wooden rod, onthe anterior tongue. Cold response: Integrated re-cordings to water at 4 °C applied for 20 s.

Kumari et al. PNAS | Published online November 13, 2017 | E10377

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 2

3, 2

021

reported decreased food and fluid intake in patients should ad-dress flavor perception. We are investigating responses to gus-tatory and olfactory stimuli in patients treated with HPI drugs tolearn about possible contrasting effects in taste and smell.

Materials and MethodsAnimals. Animal use and care procedures were performed according to theguidelines of the National Institutes of Health and approved protocols of theUniversity of Michigan Institutional Animal Care and Use Committee. C57BL/6 mice were treated for 3–36 d by daily oral gavage with sonidegib [NVP-LDE225 diphosphate salt; Chemietek catalog no. CT-LDE225] dissolved invehicle, PEG 400/5% dextrose in water (75:25 vol/vol), at a dose of 20 mg/kg,or with vehicle alone. Dose determination was after Kumari et al. (7). Forrecovery experiments, mice were treated with sonidegib or vehicle for 16 d;then the drug was discontinued, and animals were maintained for 7, 14, or21 d or for 3, 5, or 9 mo in standard housing. Mouse genotypes and sourcesare given in SI Materials and Methods.

Tissue Analyses. Tongues were collected and prepared for tissue analysis, or re-cordings were made from the chorda tympani nerve and then tongues were dis-sected. Tissue processing, immunohistochemistry and neurophysiological protocols,and data and quantification analyses are described in SI Materials and Methods.

Statistics. Animal numbers and statistical tests are included in the graphs andfigure legends. Details are supplied in SI Materials and Methods.

ACKNOWLEDGMENTS. We thank Ariell M. Joiner in the B.L.A. laboratoryand Libo Li in the C.M.M. laboratory for technical assistance with tissue collec-tion, analyses, and immunostaining, and the staff in the University of MichiganDentistry Histology Core for technical support. This work was supported by NIHfunding from the National Institute on Deafness and Other CommunicationDisorders Multi Principal Investigator Grant R01 DC014428 (to B.L.A., R.M.B.,A.A.D., and C.M.M.); National Institute of Arthritis and Musculoskeletal andSkin Diseases Grant R01 AR045973 (to A.A.D.); University of Michigan Compre-hensive Cancer Center Core Grant P30 CA046592 (to A.A.D.); and a Universityof Michigan Center for Organogenesis Postdoctoral Fellowship (to A.K.).

1. Migden MR, et al. (2015) Treatment with two different doses of sonidegib in patientswith locally advanced or metastatic basal cell carcinoma (BOLT): A multicentre,randomised, double-blind phase 2 trial. Lancet Oncol 16:716–728.

2. Rodon J, et al. (2014) A phase I, multicenter, open-label, first-in-human, dose-escalation study of the oral smoothened inhibitor Sonidegib (LDE225) in patientswith advanced solid tumors. Clin Cancer Res 20:1900–1909.

3. Sekulic A, et al. (2012) Efficacy and safety of vismodegib in advanced basal-cell car-cinoma. N Engl J Med 366:2171–2179.

4. Tang JY, et al. (2012) Inhibiting the hedgehog pathway in patients with the basal-cellnevus syndrome. N Engl J Med 366:2180–2188.

5. Fife K, et al. (2017) Managing adverse events associated with vismodegib in thetreatment of basal cell carcinoma. Future Oncol 13:175–184.

6. Pan S, et al. (2010) Discovery of NVP-LDE225, a potent and selective smoothenedantagonist. ACS Med Chem Lett 1:130–134.

7. Kumari A, et al. (2015) Hedgehog pathway blockade with the cancer drugLDE225 disrupts taste organs and taste sensation. J Neurophysiol 113:1034–1040.

8. Mistretta CM, Kumari A (2017) Tongue and taste organ biology and function: Ho-meostasis maintained by hedgehog signaling. Annu Rev Physiol 79:335–356.

9. Beidler LM, Smallman RL (1965) Renewal of cells within taste buds. J Cell Biol 27:263–272.

10. Hamamichi R, Asano-Miyoshi M, Emori Y (2006) Taste bud contains both short-livedand long-lived cell populations. Neuroscience 141:2129–2138.

11. Ermilov AN, et al. (2016) Maintenance of taste organs is strictly dependent on epi-thelial hedgehog/GLI signaling. PLoS Genet 12:e1006442.

12. Gaillard D, Xu M, Liu F, Millar SE, Barlow LA (2015) β-catenin signaling biases multi-potent lingual epithelial progenitors to differentiate and acquire specific taste cellfates. PLoS Genet 11:e1005208.

13. Liu HX, et al. (2013) Multiple Shh signaling centers participate in fungiform papillaand taste bud formation and maintenance. Dev Biol 382:82–97.

14. Guagliardo NA, Hill DL (2007) Fungiform taste bud degeneration in C57BL/6J micefollowing chorda-lingual nerve transection. J Comp Neurol 504:206–216.

15. Lee RTH, Zhao Z, Ingham PW (2016) Hedgehog signalling. Development 143:367–372.16. Rimkus TK, Carpenter RL, Qasem S, ChanM, Lo HW (2016) Targeting the sonic hedgehog

signaling pathway: Review of smoothened and GLI inhibitors. Cancers (Basel) 8:E22.17. Stone DM, et al. (1996) The tumour-suppressor gene patched encodes a candidate

receptor for sonic hedgehog. Nature 384:129–134.18. van den Heuvel M, Ingham PW (1996) smoothened encodes a receptor-like serpentine

protein required for hedgehog signalling. Nature 382:547–551.19. Briscoe J, Thérond PP (2013) The mechanisms of hedgehog signalling and its roles in

development and disease. Nat Rev Mol Cell Biol 14:416–429.20. Hui CC, Angers S (2011) Gli proteins in development and disease. Annu Rev Cell Dev

Biol 27:513–537.21. Ingham PW, McMahon AP (2001) Hedgehog signaling in animal development: Par-

adigms and principles. Genes Dev 15:3059–3087.22. Miura H, Scott JK, Harada S, Barlow LA (2014) Sonic hedgehog-expressing basal cells are

general post-mitotic precursors of functional taste receptor cells. Dev Dyn 243:1286–1297.23. Miura H, Kusakabe Y, Harada S (2006) Cell lineage and differentiation in taste buds.

Arch Histol Cytol 69:209–225.24. Jain S, Song R, Xie J (2017) Sonidegib: Mechanism of action, pharmacology, and

clinical utility for advanced basal cell carcinomas. Onco Targets Ther 10:1645–1653.25. Silapunt S, Chen L, Migden MR (2016) Hedgehog pathway inhibition in advanced basal

cell carcinoma: Latest evidence and clinical usefulness. Ther Adv Med Oncol 8:375–382.26. Jacobsen AA, Aldahan AS, Hughes OB, Shah VV, Strasswimmer J (2016) Hedgehog path-

way inhibitor therapy for locally advanced and metastatic basal cell carcinoma: A sys-tematic review and pooled analysis of interventional studies. JAMADermatol 152:816–824.

27. Lacouture ME, et al. (2016) Characterization and management of hedgehog pathwayinhibitor-related adverse events in patients with advanced basal cell carcinoma.Oncologist 21:1218–1229.

28. Tang JY, et al. (2016) Inhibition of the hedgehog pathway in patients with basal-cellnevus syndrome: Final results from the multicentre, randomised, double-blind,placebo-controlled, phase 2 trial. Lancet Oncol 17:1720–1731.

29. Chaudhari N, Roper SD (2010) The cell biology of taste. J Cell Biol 190:285–296.

30. Sullivan JM, Borecki AA, Oleskevich S (2010) Stem and progenitor cell compartmentswithin adult mouse taste buds. Eur J Neurosci 31:1549–1560.

31. Perea-Martinez I, Nagai T, Chaudhari N (2013) Functional cell types in taste buds havedistinct longevities. PLoS One 8:e53399.

32. Okubo T, Clark C, Hogan BL (2009) Cell lineage mapping of taste bud cells and ker-atinocytes in the mouse tongue and soft palate. Stem Cells 27:442–450.

33. Asano-Miyoshi M, Hamamichi R, Emori Y (2008) Cytokeratin 14 is expressed in im-mature cells in rat taste buds. J Mol Histol 39:193–199.

34. Knapp L, Lawton A, Oakley B, Wong L, Zhang C (1995) Keratins as markers of dif-ferentiated taste cells of the rat. Differentiation 58:341–349.

35. Brownell I, Guevara E, Bai CB, Loomis CA, Joyner AL (2011) Nerve-derived sonichedgehog defines a niche for hair follicle stem cells capable of becoming epidermalstem cells. Cell Stem Cell 8:552–565.

36. Zhao H, et al. (2014) Secretion of shh by a neurovascular bundle niche supportsmesenchymal stem cell homeostasis in the adult mouse incisor. Cell Stem Cell 14:160–173.

37. Bartoshuk LM (1993) Genetic and pathological taste variation: What can we learnfrom animal models and human disease? Ciba Found Symp 179:251–262, discussion262–267.

38. Spector AC, Smith JC (2004) Neurobiology of Food and Fluid Intake, Handbook ofbehavioral neurobiology, eds Stricker E, Woods SC (Kluwer Academic/Plenum Pub-lishers, New York), Vol 14, pp 63–87.

39. Fuchs E (2008) Skin stem cells: Rising to the surface. J Cell Biol 180:273–284.40. Lafkas D, et al. (2015) Therapeutic antibodies reveal Notch control of trans-

differentiation in the adult lung. Nature 528:127–131.41. Castillo D, et al. (2014) Induction of ectopic taste buds by SHH reveals the competency

and plasticity of adult lingual epithelium. Development 141:2993–3002.42. Oakley B, Wu LH, Lawton A, deSibour C (1990) Neural control of ectopic filiform

spines in adult tongue. Neuroscience 36:831–838.43. Yokota Y, Kumari A, Mistretta CM, Bradley RM (2016) Thermal and tactile responses of

the rat chorda tympani nerve. Neuroscience Meeting Planner. Available at http://www.abstractsonline.com/pp8/index.html#!/4071/presentation/8790. Accessed October 27, 2017.

44. Peterson SC, et al. (2015) Basal cell carcinoma preferentially arises from stem cellswithin hair follicle and mechanosensory niches. Cell Stem Cell 16:400–412.

45. Mistretta CM, Liu HX, Gaffield W, MacCallum DK (2003) Cyclopamine and jervine inembryonic rat tongue cultures demonstrate a role for Shh signaling in taste papilladevelopment and patterning: Fungiform papillae double in number and form innovel locations in dorsal lingual epithelium. Dev Biol 254:1–18.

46. Yam PT, Charron F (2013) Signaling mechanisms of non-conventional axon guidancecues: The Shh, BMP and Wnt morphogens. Curr Opin Neurobiol 23:965–973.

47. Yam PT, Langlois SD, Morin S, Charron F (2009) Sonic hedgehog guides axons througha noncanonical, Src-family-kinase-dependent signaling pathway. Neuron 62:349–362.

48. Mbiene JP, Mistretta CM (1997) Initial innervation of embryonic rat tongue and de-veloping taste papillae: Nerves follow distinctive and spatially restricted pathways.Acta Anat (Basel) 160:139–158.

49. Yang H, Cong WN, Yoon JS, Egan JM (2015) Vismodegib, an antagonist of hedgehogsignaling, directly alters taste molecular signaling in taste buds. Cancer Med 4:245–252.

50. Dummer R, et al. (2016) The 12-month analysis from basal cell carcinoma outcomes withLDE225 treatment (BOLT): A phase II, randomized, double-blind study of sonidegib inpatients with advanced basal cell carcinoma. J Am Acad Dermatol 75:113–125.e5.

51. Colbert CL, Garcea M, Spector AC (2004) Effects of selective lingual gustatory deaf-ferentation on suprathreshold taste intensity discrimination of NaCl in rats. BehavNeurosci 118:1409–1417.

52. Bai CB, Auerbach W, Lee JS, Stephen D, Joyner AL (2002) Gli2, but not Gli1, is requiredfor initial Shh signaling and ectopic activation of the Shh pathway. Development 129:4753–4761.

53. Diamond I, Owolabi T, Marco M, Lam C, Glick A (2000) Conditional gene expression inthe epidermis of transgenic mice using the tetracycline-regulated transactivators tTAand rTA linked to the keratin 5 promoter. J Invest Dermatol 115:788–794.

54. Perl AK, Wert SE, Nagy A, Lobe CG, Whitsett JA (2002) Early restriction of peripheraland proximal cell lineages during formation of the lung. Proc Natl Acad Sci USA 99:10482–10487.

E10378 | www.pnas.org/cgi/doi/10.1073/pnas.1712881114 Kumari et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 2

3, 2

021