Original Article

Retinoic Acid-Induced Inner Ear Teratogenesis Causedby Defective Fgf3/Fgf10-Dependent Dlx5 Signaling

Wei Liu,1 Giovanni Levi,2 Alan Shanske,3,4,5 and Dorothy A. Frenz1,6�

1Department of Otorhinolaryngology Head and Neck Surgery, Albert Einstein College of Medicine, Bronx, New York2Evolution des Regulations Endocriniennes, CNRS, UMR5166, Museum National d’Histoire Naturelle, Paris, France

3Department of Pediatrics, Albert Einstein College of Medicine, Bronx, New York4Department of Pathology, Albert Einstein College of Medicine, Bronx, New York

5Department of Obstetrics and Gynecology and Women’s Health, Albert Einstein College of Medicine, Bronx, New York6Department of Anatomy and Structural Biology, Albert Einstein College of Medicine, Bronx, New York

Background: Retinoic acid (RA) is essential for inner ear development. However, exposure to excess RA at a criticalperiod leads to inner ear defects. These defects are associated with disruption in epithelial–mesenchymal interactions.METHODS: This study investigates the role of Dlx5 in the epithelial–mesenchymal interactions that guide otic capsulechondrogenesis, as well as the effect of excess in utero RA exposure on Dlx5 expression in the developing mouse innerear. Control of Dlx5 by Fgf3 and Fgf10 under excess RA conditions is investigated by examining the developmentalwindow during which Fgf3 and Fgf10 are altered by in utero RA exposure and by testing the ability of Fgf3 and Fgf10 tomitigate the reduction in chondrogenesis and Dlx5 expression mediated by RA in high-density cultures of perioticmesenchyme containing otic epithelium, a model of epithelial–mesenchymal interactions in which chondrogenicdifferentiation of periotic mesenchyme ensues in response to induction by otic epithelium. RESULTS: Dlx5 deletionalters expression of TGFb1, important for otic capsule chondrogenesis, in the developing inner ear and compromises theability of cultured periotic mesenchyme containing otic epithelium, harvested from Dlx5 null embryos, to differentiateinto cartilage when compared with control cultures. Downregulation in Dlx5 ensues as a consequence of in utero RAexposure in association with inner ear dysmorphogenesis. This change in Dlx5 is noted at embryonic day 10.5 (E10.5), butnot at E9.5, suggesting that Dlx5 is not a direct RA target. Before Dlx5 downregulation, Fgf3 and Fgf10 expression ismodified in the inner ear by excess RA, with the ability of exogenous Fgf3 and Fgf10 to rescue chondrogenesis and Dlx5expression in RA-treated cultures of periotic mesenchyme containing otic epithelium supporting these fibroblast growthfactors (FGFs) as intermediary genes by which RA mediates its effects. CONCLUSIONS: Disruption in an Fgf3, -10/Dlx5signaling cascade is operant in molecular mechanisms of inner ear teratogenesis by excess RA. Birth Defects Res (Part B)83:134–144, 2008. r2008 Wiley-Liss, Inc.

Key words: retinoic acid; inner ear; Dlx5; Fgf; teratogenesis; epithelial–mesenchymal interactions

INTRODUCTION

Retinoic acid (RA) is a biologically active metabolite ofvitamin A that is crucial for normal morphogenesis andorgan development. The normal function of RA isachieved only at optimal homeostatic concentrations,such that when RA concentrations deviate from normal,teratogenesis is produced. Infants born to mothersexposed to oral retinoids at a critical period of develop-ment have an increased risk of craniofacial malforma-tions (William et al., 2004) and inner ear embryopathiesthat resemble Michel’s aplasia and the Mondini-Alex-ander defect (Morris, 1972; Lammer et al., 1985; Moerikeet al., 2002). Similar inner ear embryopathies, in whichthe developing cochlea, semicircular canals, utricle,saccule, endolymphatic duct, cochleovestibular ganglia,and otic capsule are abnormal, are produced in themouse by in utero exposure to teratogenic doses of RA at

embryonic age 9 days (E9) (Frenz et al., 1996). In extremecases, the inner ear fails to develop beyond a cystic-likestructure (Frenz et al., 1996).

The gamut of in utero RA-induced inner ear mal-formations is critically dependent on both the dose of RAadministered and the developmental stage during whichthe otic primordia is exposed to excess RA (Jarvis et al.,1989; Frenz et al., 1996). Middle ear anomalies, includingmalformations of the malleus, incus, and tympanic ring,

Published online in Wiley InterScience (www.interscience.wiley.com)DOI: 10.1002/bdrb.20154

*Correspondence to: Dorothy A. Frenz, Department of Otorhinolaryngol-ogy Head and Neck Surgery, Albert Einstein College of Medicine, KennedyCenter 301, 1410 Pelham Parkway South, Bronx, NY 10461.E-mail: frenz@aecom.yu.edu

Contract Grant sponsor: NIH/NIDCD; Grant number: R01DC04706; Grantsponsor: March of Dimes Birth Defects Foundation.

Birth Defects Research (Part B) 83:134–144 (2008)& 2008 Wiley-Liss, Inc.

ensue when RA is administered at earlier developmentaltimes within a narrow temporal framework (E8.5–E8.75)(Zhu et al., 1997; Mallo, 1997) due to effects on thesignaling epithelium of the first pharyngeal arch (Vieux-Rochas et al., 2007), which gives rise to these structures(Mallo and Gridley, 1996). Defects of the external ear, alsomalformed by early exposure to excess RA (Granstrom,1990), are associated with anomalies of the first phar-yngeal arch (Verloes et al., 1991).

The inner ear originates from a simple otic vesicle,which is dependent on inducing signals from mesodermand neuroectoderm (Torres and Giraldez, 1998), andgives rise to complex cochlear and vestibular apparati forhearing and balance, respectively. Surrounding the oticvesicle is periotic mesenchyme, from which the oticcapsule, a cartilaginous structure which surrounds andprotects the epithelial-derived otic vesicle, is derived.Exposure to teratogenic doses of RA at E9 can causeinner ear dymorphogenesis in the mouse and impairhearing function by disrupting inductive epithelial–mesenchymal interactions that normally guide inner eardevelopment (Butts et al., 2005). However, the molecularmechanisms underlying this RA action are not wellunderstood. Dlx5, a homeodomain transcription factorrelated to Drosophila Distal-less, is among the earliestexpressed genes in the otic placode, i.e., the ectodermalthickening from which the otic vesicle develops, todefine the subsequent vestibular morphogenetic pro-gram (Merlo et al., 2002; Simeone et al., 1994) and toparticipate in inner ear epithelial–mesenchymal control(Depew et al., 1999). Targeted inactivation of Dlx5produces a complex embryonic phenotype characterizedby craniofacial abnormalities involving branchial archderivatives (Acampora et al., 1999). Dlx5 null embryoscan be recognized by abnormal development of themaxilla, short snouts, and in some cases exencephaly.These anomalies are often observed after exposure toaberrant levels of RA (Frenz et al., 1996), suggesting thatDlx5 mutation may affect tissues that are also affected byinappropriate RA exposure. Similar to conditions of RAexcess, in which the formation of the semicircular canals,cochlea, endolymphatic duct, utricle, and otic capsule iscompromised (Frenz et al., 1996), the Dlx5 null inner eardemonstrates semicircular canals which fail to properlyform, with lack of distinct anterior and posterior canals,and reduction of the horizontal canal. A shortenedendolymphatic duct and an enlarged malformed utriclehave been observed (Merlo et al., 2002). Incompletecoiling of the cochlea and a hypoplastic malformed oticcapsule, regions devoid of Dlx5, characterize the Dlx5null inner ear, suggesting that morphogenetic defectsextend to regions of the inner ear in which Dlx5 is notexpressed.

Dlx5 is controlled during inner ear development byfibroblast growth factor (FGF) signaling. Targeted in-activation of Fgf3 and/or Fgf10 alters Dlx5 expression(Wright and Mansour, 2003), with levels of Dlx5 beingmoderately affected by Fgf3 deletion and dramaticallyreduced in Fgf10 null mutant inner ears (Hatch et al.,2007; Pauley et al., 2003). Genetic ablation of Fgf10, whichis expressed in the developing otic cup (Pirvola et al.,2000), causes severe dysmorphogenesis of the inner ear,noted by overall reduction in size, absence of recogniz-able anterior and posterior semicircular canals, a reducedhorizontal canal, and smaller saccule and cochlea sensory

epithelia (Pauley et al., 2003). Defects affecting cochlearcoiling and posterior semicircular canal and endolym-phatic duct formation frequently accompany targetedFgf3 deletion (Mansour, 1994). Taken together, thesedefects recapitulate the vestibular and cochlear anoma-lies that typify the Dlx5 null inner ear, supporting Dlx5 asa downstream target by which Fgf3 and Fgf10 canfacilitate inner ear developmental control. Embryoslacking both Fgf3 and Fgf10 fail to form otic vesiclesand demonstrate markedly aberrant patterns of oticmarker expression, including Dlx5, indicating that theseFgf signals are redundantly required for otic inductionand establishment of normal patterns of inner ear geneexpression (Zelarayan et al., 2007; Wright and Mansour,2003). Inactivation of Fgf8 during induction of the oticplacode in a homozygous Fgf3 null background leads to areduced or absent otic vesicle and associated loss of Dlx5expression (Zelarayan et al., 2007).

In the present study we explore the possibility that theinner ear anomalies evoked by conditions of excess RAcould derive from altered Fgf3, -10/Dlx5 signaling. Ourfindings suggest disruption in an Fgf3, -10/Dlx5 signal-ing cascade as a likely explanation for the inner eardysmorphogenesis produced when RA concentrationsdeviate from normal.

MATERIALS AND METHODS

Experimental Animals

C57/BL6 female mice were intercrossed with CBAmales (Charles River Laboratories, Wilmington, MA).Gestational age of embryos was estimated by the vaginalplug method, with the day of plug occurrence designatedas day 1(E1). After death of the gravid females byisoflurane inhalation followed by cervical dislocation,embryos were harvested and immediately placed intoDulbecco’s phosphate-buffered saline (PBS) (Gibco,Grand Island, NY). Embryonic age was determined bya combination of external features and somite count(Theiler, 1972). Dlx5 mutant mice were generated asdescribed (Acampora et al., 1999). Heterozygous micewere cross-mated and offspring genotyped by a combi-nation of external features and PCR analysis using tail tipDNA and the described primers (Merlo et al., 2002).

Immunohistochemistry

Inner ear specimens were fixed in methacarn fixative,dehydrated in methanol, cleared in Histoclear (NationalDiagnostics, Rochester, NY), and paraffin-embedded.Deparaffinized sections (5–6 mm) were processed forimmunolocalization using the avidin-biotin complex(ABC) method (Vector, Burlingame, CA), as described(Frenz et al., 1992). The deparaffinized sections werepretreated for 30 min in 100% methanol 10.3% peroxide,after the blocking of nonspecific binding sites with serumfrom the ABC kit (Vector). Sections were covered with anoptimal dilution of TGFb 1 antibody (1:300; Ab9566; R&DSystems, Minneapolis, MN) and incubated overnight in ahumidified chamber at 41C. Controls were prepared byomission of primary antibody or by replacing it withbovine serum albumin (BSA). Sections were incubatedwith secondary biotinylated antisera and ABC reagent,then subjected to stable DAB (Research Genetics,Carlsbad, CA). Sections were stained for 1–3 min, then

135Fgf/Dlx5 SIGNALING IN RA TERATOGENESIS

Birth Defects Research (Part B) 83:134–144, 2008

counterstained with Mayer’s hematoxylin (Sigma) andmounted in Crystalmount (Biomedia, Foster City, CA),with final mounting in Permount (Fisher Scientific,Hanover Park, IL). The immunostain distribution patternwas qualitatively evaluated by 2 investigators. Evalua-tions were performed on 5–7 sections from each of theinner ears of 3 Dlx5 null mutant or wild-type embryos.

Maternal Retinoic Acid Treatment

A solution of all-trans RA (Sigma, St. Louis, MO) wasmade as 5 mg crystalline RA in 0.8 ml of absolute alcoholin 9.2 ml sesame oil and stored in the dark at 41C. A totalof 6 gravid female mice (mean weight 25 g) received RAtreatment as described (Frenz et al., 1996). Briefly, gravidmice were administered 2 consecutive doses of 25 mg/kgRA by feeding needle at 10:30 AM and 2:30 PM on E9.Control gravid mice (n 5 6) received equivalent doses ofthe vehicle (i.e., alcohol in sesame oil) at correspondingtimes or were untreated.

Probe Preparation

Dlx5-encoding plasmid (provided by Dr. MichaelShen) was linearized with HindIII and transcribed withT3 RNA polymerase. For the sense probe, the plasmidwas linearized with BamH1 and transcribed with T7RNA polymerase. Probes are digoxigenin (DIG)-labeled.

Whole-Mount In Situ Hybridization

Embryos were dissected in cold PBS and fixed over-night in 4% paraformaldehyde (PFA) in calcium/magnesium-free DEPC-PBS at 41C. The tissues werewashed in PBS with 0.1% Tween-20, dehydrated (metha-nol series), and stored until use at �201C. Embryos wererehydrated, treated with 10 mg/ml proteinase K (5 min),refixed in fresh 4% PFA-0.2% glutaraldehyde (15 min),prehybridized (3 h, 651C), and incubated with hybridiza-tion mix, including antisense or sense DIG-labeled RNAprobe (final probe concentration, 200 ng/ml) at 701Covernight. A no-probe control was performed. Anti-DIG-AP conjugate (Boehringer-Mannheim, Palo Alto, CA)was used to detect the RNA probe, and color wasdeveloped with nitro blue tetrazolium (NBT) and 5-bromo-4-chloro-3-indolyl phosphate (BCIP). Embryoswere washed in PBS with 1% Triton X-100 (3� ) to stopthe staining reaction. Expression patterns of Dlx5 wereobserved in 4 embryos from 3 different litters for eachembryonic age and treatment (RA, control).

Reverse-Transcription PCR

Total RNA was isolated (RNassay Kit; Qiagen, Valen-cia, CA) from E9.5 and E10.5 mouse otocysts. Total RNAwas reverse-transcribed to first-strand cDNA using theSuperScript Preamplification System (Invitrogen, Carls-bad, CA). The second DNA strand was synthesized usingTaq PCR Core Kit (Qiagen) and oligomers specific forDlx5. The PCR conditions were 951C (5 min); 35 cycles of941C (30 s), 601C (1 min), 721C (45 s); and 721C (10 min).Aliquots (25ml) of each amplified product were electro-phoresed on an ethidium bromide-stained gel. Theoligomers were as follows: Dlx5 antisense, 50-ATCTAATAAAGCGTCCCGGA-30; Dlx5 sense, 50-CTATGACAGCAGTGTTTGAC-30.

Genotyping

Genotypings of Dlx5 mutant mice were made by PCRanalysis. The Dlx5-specific primers were as follows:

* For detection of wild-type embryos:J primer 1: 50-GACAGGAGTGTTTGACAGAAGAG

TCCC-30

J primer 2: 50- GTAGTCGGCATAAGCCTTGGC-30,generating a band of 280 bp;

* For detection of null mutant embryos:J primer 3: 50-GCCCATCTAATAAAGCGTCCCGG-30

J primer 4: 50-TGCTGTGTTCCAGAAGTGTT-30,generating a band of 1100 bp

High-Density Culture

Otocysts were harvested with their associated perioticmesenchyme from either E10.5 CBA C57/BL6 embryosor E13 Dlx5 mutant embryos, and dissociated cells werecultured according to standard procedures (Frenz andVan De Water, 1991). Briefly, mesenchymal and epithelialtissues were dissociated with 0.05% trypsin-EDTA(Gibco, Grand Island, NY) and mesenchymal cells wereresuspended in Ham’s F-12 culture medium (Gibco,Grand Island, NY) supplemented with 10% fetal bovineserum (FBS) at a density of 2.5� 107 cells/ml. Equivalentamounts of otic epithelial tissue per culture were mixedinto the cell suspension, and 10-ml droplets of cellsuspension were plated in the centers of wells of a4-well tissue culture plate (Nunc, Naperville, IL). Aftera 1-h incubation period at 371C, 1 ml Ham’s F-12 mediumplus 10% FBS was added to each well. Some cultureswere treated with 10�10 M all-trans RA (days 1–7), a doseknown to produce teratogenic effects in capsule cartilagein vitro that mimic RA-mediated teratogenesis of capsulecartilage in vivo (Frenz and Liu, 1997, 2000). Controlcultures received an equivalent amount of vehicle orremained untreated. Rescue cultures received RA treat-ment on days 1–7, then supplementation on day 2 withFgf3 (30 ng/ml; 48 h) (Cambridge Research Biochemicals,Billingham, UK), Fgf10 (30 ng/ml; 48 h) (R&D Systems,Minneapolis, MN) or a combination of Fgf3 and Fgf10(20 ng/ml total; 48 h). All cultures were maintained in ahumidified 5% CO2 atmosphere at 371C for a 7-dayperiod. Nutrient solution was exchanged every other day.

Quantitation of Mesenchymal CellCondensations

Cultures were monitored daily by phase contrast andHoffman modulation contrast microscopy for identifica-tion of condensed mesenchymal cells. On culture day 3,cell condensations were counted and used as an earlyindex of chondrogenesis (Frenz et al., 1989).

Quantitative Alcian Blue Staining of Cultures

Cultures were fixed on day 7 with a solution of 10%formalin containing 1% cetyl pyridinium chloride, thenstained with Alcian blue 8GX, pH 1.0, a stain which atpH 1.0 binds specifically to sulfated glycosaminoglycans(S-GAG) (Lev and Spicer, 1964) in the matrix ofchondrifying cells. After washing cultures with a 3% aceticacid solution to remove unbound stain, matrix-bound stainwas extracted overnight with a solution of 8 M guanidine

136 LIU ET AL.

Birth Defects Research (Part B) 83:134–144, 2008

hydrochloride and measured by spectrophotometric quan-titation (Hassell and Horigan, 1982) on a Biotek microplatereader equipped with a 600-nm wavelength filter, asdescribed (Frenz et al., 1992). Values for optical density ofbound Alcian blue stain in Table 1 were normalized tocontrol for differences in the amount of otic epithelium atcell seeding between experimental setups.

Cell Transfection

Cultures were prepared as indicated above and grownovernight in 1 ml Ham’s F-12 culture medium supple-mented with 10% FBS. Transfection was carried outusing GeneJammers Transfection Reagent (Stratagene,La Jolla, CA), which is effective in serum-containingmedium, offering high efficiencies with minimal cyto-toxicity. Transfection mixture containing Dlx5 antisenseexpression construct (pCMV5-Dlx5-AS), constructed andprovided by Dr. H. Ryoo (Lee et al., 2003), was preparedaccording to the manufacturer’s instructions. Eachtransfection assay was performed with 0.5 mg of theDlx5 expression construct. The transfection mixture wasadded dropwise to the culture dish, which was gently

rocked to distribute the mixture evenly. At 6 h aftertransfection, medium was changed and cells cultured foran additional 24 h, after which the transfection wasrepeated. This timepoint corresponds to when endogen-ous Dlx5, devoid in precondensing mesenchyme, be-comes expressed in chondrifying cells. The cultures werethen maintained for an additional 2 days in Ham’s F-12.Transfection controls consisted of (1) cell control with noGeneJammer reagent and no plasmid DNA (nontrans-fected); (2) reagent control with GeneJammer reagentand an empty vector (mock-transfected; no plasmidDNA); and (3) DNA control with no GeneJammerreagent, but containing DNA. Cell morphology wasmonitored to ensure no signs of toxicity. At the end ofculture (day 7), the extent of chondrogenesis wasassayed using binding of Alcian blue stain as an indexof chondrogenesis. Transfection efficiency was deter-mined by fluorescence microscopy to detect expressionof plasmid-encoded GFP. Expression of Dlx5 RNA wasdetermined in transfected and nontransfected controlcultures by RT-PCR.

Indirect Immunofluorescent Staining of Cultures

Cultures were grown on Lab-Tek culture chambers(Nunc, Rochester, NY) and fixed in 4% paraformalde-hyde in 0.1 M phosphate buffer (pH 7.4) (PBS). Afterwashing the fixed cultures in phosphate buffer, antibodydirected against Dlx5 (Chemicon, Billerica, MA) wasapplied (1:50 dilution) at room temperature for 1–2 h.Controls were prepared by omission of primary anti-body. Cultures were washed in phosphate buffer, and anFITC-conjugated secondary antibody was applied(30 min, room temperature). Specimens were examinedon a Zeiss Axiophot, using a FITC epifluorescence bluewavelength (450–490 nm) excitation filter set.

RESULTS

Molecular Characterization of Dlx5Loss-Of-Function Otic Capsule Defects

The periotic mesenchyme from which the otic capsuleforms does not express Dlx5, indicating that anomalouscapsule development cannot be directly accounted for byDlx5 deletion. Known regulators of otic capsule forma-tion were therefore evaluated to understand the mole-cular basis underlying the Dlx5 null capsule phenotype,which was previously characterized as hypoplastic anddysmorphic (Merlo et al., 2002; Acampora et al., 1999;Depew et al., 1999). Of particular interest was TGFb1,a signaling molecule that mediates otic capsule chon-drogenesis (Frenz et al., 1992, 1994). At E14 days, TGFb1

was evident in the chondrifying otic capsule of wild-typeembryos. In contrast, a dramatic reduction in TGFb1

immunostain was observed in the malformed inner earsof Dlx5 null mutant littermates (Fig. 1).

Defective Capsule Chondrogenesis Due to Dlx5Diminution

The reduction in TGFb1 due to Dlx5 deletion, takentogether with the dysmorphic development of the Dlx5null otic capsule (Merlo et al., 2002), suggested that theDlx5 null capsule phenotype may reflect compromisedchondrogenesis. We addressed this by ascertainingwhether there are consequences of diminished Dlx5

Table 1Effect of Dlx5 Mutation on Otic Capsule Chondrogenesis

in Culture

A. Dlx5 mutant otic epithelium 1 mutant periotic mesenchyme

Epithelium 1

mesenchymeCondensation

No.OD Alcian

blue

CBA C57/BL6 9774 0.93870.02 (n 5 4)Dlx51/�; 1/1 9976 0.94370.03 (n 5 4)Dlx5�/� 3573� 0.33070.01� (n 5 3)

B. Dlx5 mutant otic epithelium 1 control periotic mesenchyme

Experiment Epithelium Mesenchyme Condensation No.

1 CBA C57/BL6 CBA C57/BL6 92Dlx51/�; 1/1 Dlx51/�; 1/1 88

Dlx5�/� Dlx51/�; 1/1 17

2 CBA C57/BL6 CBA C57/BL6 87Dlx51/�; 1/1 Dlx51/�; 1/1 91

Dlx5�/� Dlx51/�; 1/1 23

3 CBA C57/BL6 CBA C57/BL6 90Dlx51/�; 1/1 Dlx51/�; 1/1 86

Dlx5�/� Dlx51/�; 1/1 14

A. Periotic mesenchyme and otic epithelium were harvestedfrom E13 Dlx5 null (�/�) mutant embryos, E13 Dlx5 hetero-zygous (1/�) and wild-type (1/1) embryos or control E13CBAC57/BL6 embryos, and cultured at high density. Valuesrepresent the mean number of condensations (day 3) and themean optical density of matrix-bound Alcian blue stain (day 7)7 SEM.�Statistical significance (Po0.05, Student’s t-test).B. Otic epithelium from E13 Dlx5 null mutant embryos wasmixed in culture with E13 control (1/1; 1/�) perioticmesenchyme. Parallel-run cultures were comprised of E13control (1/1; 1/�) otic epithelium and control (1/1; 1/�)periotic mesenchyme or E13 CBAC57/BL6 otic epithelium andCBAC57/BL6 periotic mesenchyme. Values represent the num-ber of condensations per culture per experiment.

137Fgf/Dlx5 SIGNALING IN RA TERATOGENESIS

Birth Defects Research (Part B) 83:134–144, 2008

expression on otic capsule chondrogenesis using high-density cultures of periotic mesenchyme and otic epithe-lium (referred to as periotic mesenchyme 1 otic epithe-lium) to model the chondrogenic process. Culturedperiotic mesenchyme 1 otic epithelium were harvestedfrom mouse embryos of embryonic age E10.5 days andtransfected with Dlx5 antisense expression constructpCMV5-Dlx5-AS. As a control, parallel-run E10.5 cultureswere mock transfected with an empty vector (no plasmidDNA) or nontransfected. E10.5 was selected for studybecause it is a developmental stage that corresponds toongoing epithelial–mesenchymal interactions. Represen-tative transfected cultures were assayed in 2 independentstudies by RT-PCR on culture day 4 to confirm diminutionin Dlx5 mRNA expression (Fig. 2). A total of 10 culturesper treatment (Dlx5 transfected cells, control cells)comprised each RT-PCR sample, with each RT-PCR runyielding consistent and reproducible results. On day 7, theextent of chondrogenesis was assessed in transfected andcontrol cultures using binding of Alcian blue stain, pH 1.0,as an index of chondrogenesis. Mean values for opticaldensity (OD) of matrix-bound Alcian blue stain weresignificantly decreased in Dlx5 antisense transfectedcultures compared with untreated control and transfectioncontrol cultures (Fig. 3), suggesting that critical levels ofDlx5 are required for otic capsule chondrogenesis.

Defective Capsule Chondrogenesis in Dlx5 NullMutant Cultures

It was reasoned that if a reduction in Dlx5 expressionby transfection with antisense construct can modify

chondrogenesis, absence of Dlx5 due to targeted deletionis also likely to produce a compromised chondrogeniceffect. To address this, we tested the ability of perioticmesenchyme 1 otic epithelium harvested from Dlx5 nullmutant embryos to differentiate in vitro. To ensuresufficient tissue, mesenchyme and epithelium wereisolated at E13, a developmental stage which has beencharacterized for chondrogenic potential in CBA C57/BL6 embryos (Frenz and Van De Water, 1991). Tissuefrom only 4 Dlx5 null otocysts was used to prepare 3Dlx5 null mutant cultures. In all cases, culturescomposed of a combination of periotic mesenchymeand otic epithelium from Dlx5 wild-type and hetero-zygous littermates were cultured in parallel as controlsto the null mutant cultures. On culture day 3, mesench-ymal condensations, i.e., aggregates of mesenchymewhich presage the formation of chondrogenic foci, werecounted and used as an early index of chondrogenesis,then on day 7, cultures were stained with Alcian blue(pH 1.0). Comparable mean values for condensationnumber and for OD of Alcian blue stain were noted incultures consisting of a combination of Dlx5 hetero-zygous and wild-type inner ear tissue (mesenchyme,epithelium) and inner ear tissue from CBA C57/BL6embryos of the same developmental age (E13) (Table 1),confirming absence of an effect of Dlx5 heterozygousand wild-type genetic background on chondrogenesis.Mean values for condensation number and for opticaldensity (OD) of Alcian blue stain were dramaticallyreduced in Dlx5 null mutant cultures in comparisonwith parallel-run control cultures composed of acombination of Dlx5 heterozygous and wild-type tissue,

Fig. 1. E14 Dlx5 mutant inner ears, immunostained for TGFb1 A: Wild-type specimen, showing immunostaining for TGFb1 in thechondrifying otic capsule (c) (dorsal aspect shown). The developing utricle (u) and a semicircular duct (sd) are indicated. However, inthe malformed Dlx5 null inner ear (B), which appears as an ovoid vesicle (O), the periotic mesenchyme (m) which will form the oticcapsule is not yet condensed, and immunostaining for TGFb1 is not evident. Scale bar 5 200mm.

Fig. 2. RT-PCR analysis of Dlx5 expression in culture, day 4. A: Lane 1, Dlx5 antisense transfected cultures; lane 2, nontransfected controlcultures. Comparison of lane 1 with lane 2 demonstrates a reduction in band intensity (874 bp) corresponding to Dlx5 after transfectionwith Dlx5 antisense construct. B: b-Actin control, showing comparable bands (349 bp) in Dlx5 antisense transfected (lane 1) andnontransfected (lane 2) cultures.

138 LIU ET AL.

Birth Defects Research (Part B) 83:134–144, 2008

indicating a reduction in chondrogenesis. This chondro-genic reduction could be rescued when a Dlx5 nullmutant culture was supplemented with TGFb1 (1 ng/ml) 24 h after seeding, i.e., a timepoint equivalent to E14when TGFb1 was shown to be diminished by Dlx5deletion in vivo (OD Alcian blue, Dlx5 null cul-ture 5 0.196; OD Alcian blue, Dlx5 null culture withadded TGFb1 5 0.265; OD Alcian blue, Dlx5 controlculture 5 .317), suggesting that TGFb1 may be a down-stream target by which Dlx5 may control otic capsulechondrogenic differentiation.

Altered Epithelial–Mesenchymal Interactionsby Dlx5 Deletion

The compromised chondrogenic potential of Dlx5 nullmutant cultures may reflect an inability of Dlx5 null oticepithelium to effectively initiate otic capsule chondro-genesis. To address this, otic epithelium was isolatedfrom E13 Dlx5 null mutant embryos and mixed inculture with periotic mesenchyme from heterozygousand wild-type littermates. Parallel-run control culturesconsisted of periotic mesenchyme and otic epithelium

from Dlx5 heterozygous and wild-type embryos. Sincethe mesenchymal condensation process is induced inresponse to epithelial signaling, condensation formationwas used as a measure of compromised chondrogenicinduction by Dlx5 deletion. In each culture testedcontaining Dlx5 null otic epithelium, a marked reductionin condensation number occurred by day 3 whencompared with cultures containing otic epithelium fromcontrol (heterozygous, wild-type) littermates (Table 1).Correspondingly, when a culture containing Dlx5 null oticepithelium (culture from Experiment 3; Table 1) wasmaintained for 7 days in vitro, binding of Alcian bluestain (pH 1.0) was dramatically reduced (OD Alcianblue 5 0.093) compared with its respective control culture(OD Alcian blue 5 0.526).

High-Dose RA and Its Effect on Dlx5 Expression

The inability of Dlx5 null otic epithelium to effectivelyinitiate epithelial–mesenchymal interactions in vitro isreminiscent of that of in utero RA-exposed otic epithe-lium (Frenz and Liu, 1997). We therefore analyzed theexpression of Dlx5 in the otic epithelium of in utero

Fig. 3. Cultured E10.5 periotic mesenchyme 1 otic epithelium, day 5 in vitro. A: Nontransfected control culture. B: Mock-transfectedcontrol culture. C: Dlx5 antisense-transfected culture. Comparison of A with B shows a comparable extent of chondrogenesis, withregions of chondrification appearing whitish against the gray background. In contrast, the transfected culture (C) shows a reduction inchondrogenesis when compared with the control cultures in A and B. Arrows demonstrate location of otic epithelium within thecultures. Values below each figure represent the mean optical density of matrix-bound Alcian blue stain7SEM for 3 cultures perrepresented experimental group. �Statistical significance (Po0.05; Student’s t-test). Scale bar 5 35mm.

Fig. 4. Whole-mount in situ hybridization for Dlx5 in control and RA-exposed mouse embryos. A: E9.5, showing comparable levels ofDlx5 expression in control and all-trans (at) RA exposed otocysts (arrows). B: E10.5, showing a reduction in Dlx5 in the atRA-exposedotocyst compared with the control otocyst (arrows). Magnified view of each otocyst is provided below its respective embryo.

139Fgf/Dlx5 SIGNALING IN RA TERATOGENESIS

Birth Defects Research (Part B) 83:134–144, 2008

RA-exposed mouse embryos. Before Dlx5 expressionanalysis, RA-exposed inner ears were evaluated stereo-microscopically to confirm RA teratogenic effects,apparent at E9.5 and E10.5 by the reduced size of theotic vesicle (Frenz et al., 1996). Whole-mount in situhybridization revealed comparable patterns of Dlx5expression throughout the otic vesicle of RA-exposedand control embryos at E9.5 (Fig. 4A). In contrast,expression of Dlx5 was markedly diminished in theRA-exposed otocyst at E10.5 days compared with age-matched control otocysts (Fig. 4B). To ensure that theobserved diminution in Dlx5 expression at E10.5 daysdid not reflect size differences of RA-exposed otocysts,RT-PCR was performed and confirmed the in situhybridization findings (Fig. 5A,B), noting that otocyststhat were not malformed by excess RA exposure did notdemonstrate a reduction in Dlx5, suggesting specificityto teratogenicity (Fig. 5B).

Effect of Excess RA on Fgf3, -10 Expression

Diminution in Dlx5 by in utero RA exposure is not animmediate response, suggesting that Dlx5 may not be adirect RA target. Since Dlx5 is controlled by Fgf3 andFgf10 during otic morphogenesis (Wright and Mansour,2003), we investigated the possibility that expression ofthese Fgf genes may be altered by RA exposure beforeRA-mediated downregulation of Dlx5. RT-PCR analysisreveals that in comparison with control specimens, Fgf3and Fgf10 are markedly reduced in the RA-exposedotocyst at E10.5 and E9.5 days, i.e., a timepoint (E9.5)preceding diminished Dlx5 expression by RA (Fig. 6A,B).To rule out the possibility that Wnt signaling, which canalso regulate Dlx5 expression (Riccomagno et al., 2005),may participate in control of Dlx5 by RA, expression ofb-catenin, a mediator of canonical Wnt signaling, wasevaluated. Comparable expression of b-catenin wasnoted in RA-exposed and control otocysts at E9.5 andE10.5 days (Fig. 6C).

Rescue of RA-Mediated ChondrogenicSuppression by Fgf3 and Fgf10

It was reasoned that if the effects of excess RA on innerear development are facilitated through downregulationof Fgf3 and Fgf10, then overexpression of these Fgfsshould provide a rescue of RA-mediated defects. Oticcapsule chondrogenesis, which is suppressed both invivo and in culture by exposure to excess RA (Frenzet al., 1996; Frenz and Liu, 1997), was used as a model inwhich to test the rescue capability of Fgf3 and Fgf10.Cultured periotic mesenchyme 1 otic epithelium wastreated with RA (1�10�10 M; days 1–7), then supple-mented on day 2 with Fgf3 (30 ng/ml; 48 h), Fgf10(30 ng/ml; 48 h) or Fgf3 110 (20 ng/ml total; 48 h).Cultures treated with RA demonstrated a markedreduction in chondrogenesis, evidenced by a mean valuefor OD of Alcian blue stain that was decreased by 63% incomparison with untreated cultures (Table 2). In RA-treated cultures supplemented with either Fgf3 or Fgf10,mean values for OD of bound Alcian blue stain wereincreased by 23% and 40%, respectively, in comparisonwith cultures treated with RA alone, suggesting a partialrescue of chondrogenesis. Chondrogenesis was almostcompletely restored in RA-treated cultures when Fgf3 1

10 were each added at a concentration of only 10 ng/ml.

This was demonstrated by a 57% increase in the meanvalue for OD of Alcian blue in RA-treated culturessupplemented with Fgf3 110 in comparison withRA-treated cultures not supplemented with Fgf3 110.

Enhancement of chondrogenesis was also producedwhen Fgf3 110 was added to control cultures. However,this increase in chondrogenesis (14%) was markedly lessthan that which occurred between RA-treated culturesand RA-treated cultures supplemented with Fgf3 110(57%) (Table 2). This finding suggests that rescue ofchondrogenesis by Fgf3 110 in RA-treated mesenchyme1 epithelium is not simply due to a differentiation effectof added growth factor. Ability of Fgf3 110 to rescue RA-mediated Dlx5 repression, as described below, supportsthis hypothesis.

Fig. 5. RT-PCR analysis of Dlx5. A: Lane 1, E9.5 control otocysts;lane 2, E9.5 RA-exposed otocysts; lane 3, E10.5 control otocysts;lane 4, E10.5 RA-exposed otocysts. A band corresponding to thepredicted size of the amplified Dlx5 product (874 bp) wasproduced from E9.5 control and RA-exposed otocysts. However,intensity of the band was markedly diminished in E10.5 RA-exposed otocysts compared with their age-matched controlspecimens, indicating diminished expression of Dlx5. B: RT-PCRanalysis of E10.5 otocysts. Lane 1, Control; lane 2, RA-exposedand malformed; lane 3, RA-exposed, nonmalformed; lane 4; RA-exposed, nonmalformed. Reduction in band (874 bp) intensityfor Dlx5 was only produced when otocysts were malformedby in utero RA exposure. C: Analysis of b-actin in E9.5 control(lane 1), E9.5 RA-treated (lane 2), E10.5 control (lane 3) and E10.5RA-treated (lane 4) otocysts served as a control.

140 LIU ET AL.

Birth Defects Research (Part B) 83:134–144, 2008

Rescue of RA-Mediated Repression of Dlx5 byFgf3 and Fgf10

To address the hypothesis that Fgf3 and Fgf10 actupstream of Dlx5 to mediate RA teratogenesis, we testedthe ability of Fgf3 and Fgf10 to rescue downregulation ofDlx5 by RA. E10.5 periotic mesenchyme 1 otic epithe-lium was treated with RA and/or Fgf3 110 as describedabove. On day 4, cultures were fixed and immunofluor-escently labeled for endogenous Dlx5. Day 4 was selectedas a timepoint representative of active cartilage differ-entiation in vitro, since Dlx5 is not expressed inprecondensing periotic mesenchyme (Acampora et al.,1999). Our findings show that Dlx5 was expressed inregions of culture where cartilage was already formed(Fig. 7A,B). However, treatment with RA suppressedchondrogenesis and reduced Dlx5 expression to a faintpattern of immunofluorescence even at sites wherecondensations had formed (Fig. 7C,D). Supplementationof RA-treated cultures with Fgf3 110 restored expressionof Dlx5 (Fig. 7E,F). Furthermore, addition of Fgf3 110 tountreated (no RA) control cultures, which was shown toaugment chondrogenesis, resulted in an increased

intensity for endogenous Dlx5. RT-PCR analysis con-firmed our immunofluorescence findings, showingreduction in band intensity corresponding to Dlx5(874 bp) in cultures treated with RA, and restoration byexogenous Fgf3 110 (Fig. 8).

DISCUSSION

It is well established that exposure to excess RA at acritical period of development has profound effects onmorphogenesis and organogenesis. The inner ear isparticularly sensitive to perturbations in RA, with severeinner ear anomalies resulting from exposure to excess RAat a defined window of time. However, the mechanismsunderlying RA inner ear embryopathies are not wellunderstood. In this study, we provide an explanation forthe inner ear malformations, in particular otic capsuleanomalies, induced by exposure to high-dose RA.

Dlx5 is essential to inner ear morphogenesis, directlyaffecting development of the vestibular apparatus(semicircular canals) and indirectly affecting formationof the cochlea and otic capsule (Acampora et al., 1999;Depew et al., 1999). We show that inactivation of Dlx5leads to diminution in TGFb1, a signaling molecule thatmediates otic epithelial–periotic mesenchymal interac-tions (Frenz et al., 1992) (Fig. 1) and whose down-regulation may account for the compromised andhypoplastic development of the Dlx5 null otic capsule.BMP4, a TGFb-related molecule also diminished by Dlx5deletion (Acampora et al., 1999), participates in oticepithelial-periotic mesenchymal interactions, impactingboth otic capsule formation (Liu et al., 2003; Chang et al.,1999) and semicircular canal development (Gerlachet al., 2000; Cole et al., 2000; Morsli et al., 1998; Ohet al., 1996). To address the role of Dlx5 in epithelial–mesenchymal signaling, mesenchyme and otic epithe-lium was harvested from the Dlx5 null inner ear, thencultured at high-density. A reduction in chondrogenesisensued in mutant cultures devoid of Dlx5 in comparisonwith control cultures expressing Dlx5 (Table 1). Thisreduction in chondrogenesis could be restored bysupplementation with exogenous TGFb1, supportingTGFb1 as a downstream target by which Dlx5 mayfacilitate chondrogenic control. TGFb1 is initially derived

Fig. 6. RT-PCR analysis for (A) Fgf10, (B) Fgf3, and (C) b-catenin.A: Lane 1, E10.5 control otocysts; lane 2, E10.5 RA-exposedotocysts; lane 3, E9.5 control otocysts; lane 4, E9.5 RA-exposedotocysts; lane 5, E9.5 RA-exposed embryos (no otocysts). A bandcorresponding to the predicted size of the amplified Fgf10product (569 bp) was produced from both control and RA-exposed specimens, but markedly reduced by RA-exposure. B:Lane 1, E9.5 control otocysts; lane 2, RA-exposed otocysts.Intensity of the band corresponding to the amplified Fgf3product (145 bp) was markedly reduced in RA-exposed otocystscompared with control otocysts. C: Lane 1, E10.5 controlotocysts; lane 2, E10.5 RA-exposed otocysts; lane 3, E9.5 controlotocysts; lane 4, E9.5 RA-exposed otocysts. Comparable bandscorresponding to b-catenin (427 bp) were produced from RA-exposed and control otocysts at each age examined.

Table 2Rescue of RA-Mediated Chondrogenic Suppression

by Fgf3,-10

Treatment OD Alcian blue stain

Control 0.43770.05 (n 5 5)RA 0.16170.03� (n 5 5)RA 1 Fgf3 0.20770.09 (n 5 2)RA 1 Fgf10 0.26770.01 (n 5 2)RA 1 Fgf3,-10 0.37670.06 (n 5 4)Fgf3,-10 0.50670.12 (n 5 4)

Cultured E10.5 periotic mesenchyme 1 otic epithelium wastreated with RA (10�10 M; days 1–7), Fgf3 1 Fgf10 (FGF3,-10;10 ng/ml each; days 2–3) or a combination of RA and Fgf3(30 ng/ml), RA and Fgf10 (30 ng/ml) or RA and Fgf3 110(10 ng/ml each). Control cultures were untreated (no RA or Fgf).Values represent the mean OD of Alcian blue stain 7 SEM.�Statistically significant difference when compared withuntreated control or RA 1 Fgf3,-10-treated cultures (Po0.05).

141Fgf/Dlx5 SIGNALING IN RA TERATOGENESIS

Birth Defects Research (Part B) 83:134–144, 2008

from the otic epithelium (Frenz et al., 1992). Thisobservation, combined with the fact that otic epitheliumexpresses Dlx5, but periotic mesenchyme is devoid ofthis signaling molecule (Acampora et al., 1999), suggeststhat deficits in otic capsule chondrogenesis due to Dlx5deletion may be due to deviant interactions of Dlx5 nullotic epithelium with periotic mesenchyme (Depew et al.,1999). When interacted in culture with control perioticmesenchyme, Dlx5 null otic epithelium could noteffectively initiate the chondrogenic process (Table 1).This finding suggests that in the absence of Dlx5 andconsequently its downstream target genes (e.g., TGFb),otic epithelium is aberrant in its ability to inducechondrogenesis in surrounding periotic mesenchyme,leading to perturbation in the epithelial–mesenchymalinteractions that guide inner ear development. Even justdiminished levels of Dlx5, produced by transfection ofcultures with Dlx5 antisense construct (Fig. 2), aresufficient to affect chondrogenesis (Fig. 3).

Several lines of evidence suggest that the morpholo-gical effects of RA treatment could derive from areduction in Dlx5 signaling. The phenotype of the

developing inner ear of in utero RA-exposed embryos(Frenz et al., 1996) resembles the Dlx5 null inner earphenotype (Acampora et al., 1999; Depew et al., 1999),both being characterized by absent or reduced semicir-cular canals, incomplete coiling of the cochlea, mal-formed utricle, shortened endolymphatic duct anlagen,and hypoplastic otic capsule. Epithelial–mesenchymalinteractions are disrupted by exposure to RA (Frenz andLiu, 1997) and by targeted Dlx5 deletion (Table 1), withdownregulation in TGFb1 (Fig. 1) and BMP4 (Acamporaet al., 1999), i.e., signaling molecules that control theseinner ear tissue interactions (Frenz et al., 1992, 1994; Liuet al., 2002), being common to both (Butts et al., 2005;Thompson et al., 2003). Consistent with our hypothesis,in utero exposure of the developing mouse embryo to RAat a dose that perturbs inner ear development leads todownregulation in Dlx5 expression at E10.5 days, but notat E9.5 (Figs. 4 and 5). In accord with these findings,treatment of zebrafish embryos with RA results in astage-dependent loss of Dlx expression in cartilageelements, including that affiliated with the inner ear(Ellies et al., 1997a), while excess RA-mediated

Fig. 7. Cultured E10.5 periotic mesenchyme 1 otic epithelium, day 4 in vitro. A,C,E,G: Differential interference contrast micrographsshowing sites of mesenchymal condensation (arrows) and epithelial tissue (E). B,D,F,H: Immunofluorescent photomicrographs of thesame fields as A,C,E,G, respectively, showing regions of culture where Dlx5 is expressed. A,B: Control culture. C,D: RA-exposed culture.E,F: RA-exposed culture supplemented with Fgf3 110. G,H: Control culture supplemented with Fgf3 110. Comparison of B with Ddemonstrates a marked decrease in immunofluorescent staining for Dlx5 in response to RA treatment, even at sites where mesenchymalcondensations have formed (arrows). This decrease could be rescued by exogenous Fgf3 110 (cf. D and F). Supplementation of controlculture with Fgf3 110 leads to an increase in immunofluorescent stain (cf. B and H). Scale bar 5 35 mm.

Fig. 8. RT-PCR analysis of Dlx5 in Fgf rescue cultures. A: Lane 1, Control cultures; lane 2, RA-treated cultures supplemented with Fgf3 1

10; lane 3, RA-treated cultures. Expression of Dlx5 is downregulated by treatment of cultured E10.5 periotic mesenchyme 1 oticepithelium with RA (10�10 M), but could be rescued by supplementation with Fgf3 1 Fgf10. B: A b-actin control is provided, showingcomparable bands (349 bp) in control cultures (lane 1), RA-treated cultures supplemented with Fgf3 110 (lane 2) and RA-treated cultures(lane 3).

142 LIU ET AL.

Birth Defects Research (Part B) 83:134–144, 2008

craniofacial defects affecting most of the first pharyngealarch are a consequence of reduction in the level ofexpression of Dlx genes (Vieux-Rochas et al., 2007).

Given that gravid mice received RA administration atE9, downregulation of Dlx5 at E10.5 (Figs. 4 and 5) doesnot reflect a timepoint indicative of an immediateresponse to RA, suggesting that Dlx5 is not likely to bea direct RA target. Consistent with this hypothesis,sequence analysis of Dlx3, -4 reveals elements thatresemble RA response elements (RARE) but none thatmatch exactly (Ellies et al., 1997b), while the 50 flankingregion of Xenopus Dlx2 does not contain sequences thatmatch consensus RAREs (Studer et al., 1994). Sincedirect regulation of Dlx by RA has yet to be demon-strated, we looked into control factors upstream of Dlx5that may be affected by exposure to this teratogen. Fgf isresponsive to RA, with excess RA leading to modulationof Fgf ligand expression and Fgf-mediated cell activities,including proliferation and differentiation (Shao et al.,2005; Kosaka et al., 2001; Kusaka et al., 1998; Laeng et al.,1994). Inactivation of Fgf3 and/or Fgf10 leads todysmorphogenesis of the developing semicircular ca-nals (Wright and Mansour, 2003; Alvarez et al., 2003)and an associated downregulation of Dlx5 in the dorsalotic ectoderm (Hatch et al., 2007; Wright and Mansour,2003; Pauley et al., 2003). It was therefore reasoned thatFgf3 and Fgf10 may be likely candidates for retinoidcontrol of Dlx5 expression, a hypothesis supported byRA-mediated downregulation of Fgf3 and Fgf10 beforeDlx5 (Fig. 6) and by sequence analysis of the Fgf3inner ear enhancer, which indicates the presence of 9different putative sites where retinoid signaling coulddirectly control Fgf3 expression (Powles et al., 2004).Potential retinoid binding sites were also found in a 4.5-kb fragment of 50 upstream region of Fgf10 (Ohuchiet al., 2005). Wnt1 and Wnt3a are required for oticexpression of Dlx5 (Riccomagno et al., 2005), however,expression of b-catenin, which facilitates canonical Wntsignaling, is not altered by RA exposure at either E9.5 orE10.5 (Fig. 6), indicating that Wnt is not likely toparticipate in perturbation of Dlx5 expression byaberrant RA.

The effects of excess RA on otic capsule chondrogen-esis and Dlx5 expression could be mitigated in culture bytreatment with Fgf3 and/or Fgf10, with a combination ofFgf3 110 resulting in a near-complete rescue of chon-drogenesis. This rescue supports a hierarchy in whichFgf3, -10 act as intermediary genes through which RAtargets Dlx5 and suppresses chondrogenesis. In accordwith our findings, addition of Fgf10 to RA-treatedcultures of mouse embryonic pancreas could rescueexocrine differentiation and branching morphogenesis(Shen et al., 2007). No significant changes in Fgfexpression were noted in Dlx5/Dlx6 double null otocysts(Robledo and Lufkin, 2006), indicating that the reciprocalsituation, Dlx control of Fgf, is not operant in thedeveloping inner ear. Besides the developing inner ear,Fgf and Dlx signaling is required in a number ofembryonic tissues that are sensitive to aberrant concen-trations of RA, including the palate, teeth, and limb (Leviet al., 2006; Kraus and Lufkin, 2006; Yokohama-Tamakiet al., 2006; Rice et al., 2004; Jackman et al., 2004; Brittoet al., 2002; Zhao et al., 2000). Disruption of this signalingcascade may therefore have broader implications as amodel for assessment of teratogenic agents on other

embryonic tissues which are also dependent on Fgf andDlx signaling for developmental control.

SUMMARY

In summary, our Fgf rescue studies, together with theRA-mediated changes in Fgf3, -10 expression, and thedevelopmental timepoints at which these changes wereobserved with respect to Dlx5, implicate an Fgf/Dlx5signaling cascade in underlying mechanisms of RAteratogenesis. We propose a model whereby inner eardysmorphogenesis is produced when exposure to excessRA disrupts Fgf3 and Fgf10 signaling, leading tomodification in Dlx5 expression and its downstreamtarget genes.

ACKNOWLEDGMENTS

This work was supported by NIH/NIDCD grantR01DC04706 (to D.A.F.), and by the March of DimesBirth Defects Foundation (to D.A.F.). The authors thankDr. Michael Shen for the Dlx5 encoding plasmid and Dr.H. Ryoo for the Dlx5 antisense construct. We alsoacknowledge the generous financial support of theFleisig family. This work was partially supported bythe EU Consortium CRESCENDO LSHM-CT-2005-01852(to G.L.). We are grateful to Rose Imperati for assistancein the preparation of this manuscript and to the Institutefor Communicative Disorders, Albert Einstein College ofMedicine.

REFERENCES

Acampora D, Merlo GR, Paleari L, Zerega B, Postiglione MP. 1999.Craniofacial vestibular and bone defects in mice lacking the Distal-less-related gene Dlx5. Development 126:3795–3809.

Alvarez Y, Alonso MT, Vendrell V, Zelarayan LC, Chamero P, Theil T, BosiMR, Kato S, Maconochie M, Riethmacher D, Schimmarg T. 2003.Requirements for FGF3 and FGF10 during inner ear formation.Development 130:6329–6338.

Britto JA, Evans RD, Hayward RD, Jones BM. 2002. Toward pathogenesisof Apert cleft palate: FGF, FGFR, and TGF beta genes aredifferentially expressed in sequential stages of human palatal shelffusion. Cleft Palate Craniofac J 39:331–340.

Butts SC, Liu W, Li G, Frenz DA. 2005. Transforming growth factor-b1

signaling participate in the physiological and pathological regulationof mouse inner ear development by all-trans retinoic acid. BirthDefects Res A Clin Mol Teratol 73:218–228.

Chang W, Nunes FD, De Jesus-Escobar JM, Harland R, Wu DK. 1999.Ectopic noggin blocks sensory and nonsensory organ morphogenesisin the chicken inner ear. Dev Biol 216:369–381.

Cole LK, Le Roux I, Nunes F, Laufer E, Lewis J, Wu DK. 2000. Sensoryorgan generation in the chicken inner ear: contributions of bonemorphogenetic protein 4, serrate1, and lunatic fringe. J Comp Neurol424:509–520.

Depew MJ, Liu JK, Long, JE, Presley R, Meneses JJ, Pederson RA,Rubenstein JL. 1999. Dlx5 regulates regional development of thebranchial arches and sensory capsules. Development 126:3831–3846.

Ellies DL, Langille RM, Martin CC, Akimenko M-A, Ekker M. 1997a.Specific craniofacial cartilage dysmorphogenesis coincides with aloss of dlx gene expression in retinoic acid-treated zebrafishembryos. Mech Dev 61:23–36.

Ellies DL, Sock DW, Hatch G, Giroux G, Weiss KM, Ekker M. 1997b.Relationship between the genomic organization and the overlappingembryonic expression patterns of the zebrafish dlx genes. Genomics45:580–590.

Frenz DA, Liu W. 1997. Effect of retinoic acid on otic capsulechondrogenesis in high-density culture suggests disruption ofepithelial–mesenchymal interactions. Teratology 56:233–240.

Frenz DA, Liu W. 2000. Treatment with all-trans-retinoic acid decreaseslevels of endogenous TGF-b1 in the mesenchyme of the developingmouse inner ear. Teratology 61:297–304.

Frenz DA, Jaikaria NS, Newman SA. 1989. The mechanism of precartilagemesenchymal condensation: a major role for interaction of the cell

143Fgf/Dlx5 SIGNALING IN RA TERATOGENESIS

Birth Defects Research (Part B) 83:134–144, 2008

surface with the amino-terminal heparin binding domain offibronectin. Dev Biol 13:97–103.

Frenz DA, Van De Water TR. 1991. Epithelial control of perioticmesenchyme chondrogenesis. Dev Biol 144:38–46.

Frenz DA, Galinovic-Schwartz V, Flanders KC, Van De Water TR. 1992.TGFb1 is an epithelial-derived signal peptide that influences oticcapsule formation. Dev Biol 153:324–336.

Frenz DA, Liu W, Williams JD, Hatcher V, Galinovic-Schwartz V, FlandersKC, Van De Water TR. 1994. Induction of chondrogenesis: require-ment for synergistic interaction of basic growth factor andtransforming growth factor-beta. Development 120:415–424.

Frenz DA, Liu W, Van De Water TR. 1996. Retinoic acid-inducedembryopathy of the mouse inner ear. Teratology 53:292–303.

Gerlach LM, Hutson MR, Germiller JA, Nguven-Luu D, Victor JC, BaraldKF. 2000. Addition of BMP4 antagonist, noggin, disrupts avian innerear development. Development 127:45–54.

Granstrom G. 1990. Retinoid-induced ear malformations. OtolaryngolHead Neck Surg 103:702–709.

Hassel JR, Horigan FA. 1982. Chondrogenesis: a model developmentalsystem for measuring teratogenic potential of compounds. TeratogCarcinog Mutagen 2:325–331.

Hatch EP, Noyes CA, Wang X, Wright TJ, Mansour SL. 2007. Fgf3 isrequired for dorsal pattering and morphogenesis of the inner earepithelium. Development 134:3615–3625.

Jackman WR, Draper BW, Stock DW. 2004. Fgf signaling is required forzebrafish tooth development. Dev Biol 274:139–157.

Jarvis BL, Johnston MC, Sulik KK. 1989. Congenital malformations of theexternal, middle, and inner ear produced by isoretinoin exposure inmouse embryos. Otolaryngol Head Neck Surg 102:391–401.

Kosaka C, Sasaguri T, Komiyama Y, Takahashi H. 2001. All-trans retinoicacid inhibits vascular smooth muscle cell proliferation targetingmultiple genes for cyclins and cyclin-dependent kinases. HypertensRes 24:579–588.

Kraus P, Lufkin T. 2006. Dlx homeobox gene control of mammalian limband craniofacial development. Am J Med Genet A 140:1366–1374.

Kusaka K, Kothary PC, Del Monte MA. 1998. Modulation of basicfibroblast growth factor effect by retinoic acid in cultured retinalpigment epithelium. Curr Eye Res 17:524–530.

Laeng P, Decimo D, Pettmann B, Janet T, Labourdette G. 1994. Retinoicacid regulates the development of oligodendrocyte precursor cells invitro. J Neurosci Res 39:613–633.

Lammer EJ, Chen DT, Hoar RM, Agnish N, Benke PJ. 1985. Retinoic acidembryopathy. N Engl J Med 313:837–841.

Lee M-H, Kwon TG, Park HS, Wozney JM, Ryoo H-M. 2003. BMP-2induced osterix expression is mediated by Dlx5 but is independent ofRunx2. Biochem Biophys Res Commun 309:689–694.

Lev R, Spicer SS. 1964. Specific staining of sulfate groups with Alcian blueat low pH. J Histochem Cytochem 12:309–319.

Levi G, Mantero S, Barbieri O, Cantatore D, Paleari L, Beverdam A,Genova F, Robert B, Merlo GR. 2006. Msx1 and Dlx5 actindependently in development of craniofacial skeleton, but convergeon the regulation of Bmp signaling in palate formation. Mech Dev123:3–16.

Liu W, Li G, Chien JS, Raft S, Zhang H, Chiang C, Frenz DA. 2002. Sonichedgehog regulates otic capsule chondrogenesis and inner eardevelopment in the mouse embryo. Dev Biol 248:240–250.

Liu W, Oh SH, Kang YK, Li G, Doan TM, Little M, Li L, Ahn K, CrenshawEB, Frenz DA. 2003. Bone morphogenetic protein 4 (BMP4): aregulator of capsule chondrogenesis in the developing mouse innerear. Dev Dynam 226:427–438.

Mallo M. 1997. Retinoic acid disturbs mouse middle ear development in astage-dependent fashion. Dev Biol 184:175–186.

Mallo M, Gridley T. 1996. Development of the mammalian ear: coordinateregulation of formation of the tympanic ring and the externalacoustic meatus. Development 122:173–179.

Mansour SL. 1994. Targeted disruption of int-2 (fgf-3) causes develop-mental defects in the tail and inner ear. Mol Reprod Dev 39:62–68.

Merlo GR, Paleari L, Mantero S, Zerega B, Adamska M, Rinkwitz S, BoberE, Levi G. 2002. The Dlx5 homeobox gene is essential for vestibularmorphogenesis in the mouse embryo through a BMP4-mediatedpathway. Dev Biol 248:157–169.

Moerike S, Pantzar JT, De Sa D. 2002. Temporal bone pathology in fetusesexposed to isotretinoin. Pediatr Dev Pathol 5:405–409.

Morris GM. 1972. Morphogenesis of the malformations induced in ratembryos by maternal hypervitaminosis. Am J Anat 113:241–250.

Morsli H, Choo D, Ryan A, Johnson R, Wu DK. 1998. Development ofthe mouse inner ear and origin of its sensory organs. J Neurosci18:3327–3335.

Oh SH, Johnson R, Wu DK. 1996. Differential expression of bonemorphogenetic proteins in the developing vestibular and auditorysensory organs. J Neurosci 16:6463–6475.

Ohuchi H, Yasue A, Ono K, Sasaoka S, Tomonari S, Takagi A, Itakura M,Moriyama K, Noji S, Nohno T. 2005. Identification of cis-elementregulating expression of the mouse Fgf10 gene during inner eardevelopment. Dev Dynam 233:177–187.

Pauley S, Wright TJ, Pirvola U, Ornitz D, Beisel K, Fritzsch B. 2003.Expression and function of FGF10 in mammalian inner eardevelopment. Dev Dynam 227:203–215.

Pirvola U, Spencer-Dene B, Xing-Qun L, Kettunen P, Thesleff I, Fritzsch B,Dickson C, Ylikoski J. 2000. FGF/FGFR-2(lllb) signaling is essentialfor inner ear morphogenesis. J Neurosci 20:6125–6134.

Powles N, Marshall H, Economou A, Chiang C, Murakami A, Dickson C,Krumlauf R, Maconachie M. 2004. Regulatory analysis of themouse Fgf3 gene: control of embryonic expression patterns anddependence upon sonic hedgehog (Shh) signaling. Dev Dynam230:44–56.

Rice R, Spencer-Dene B, Connor EC, Gritli-Linde A, McMahon AP,Dickson C, Thesleff I, Rice DP. 2004. Disruption of FGF10/Fgfr2b-coordinated epithelial–mesenchymal interactions causes cleft palate.J Clin Invest 113:1692–1700.

Riccomagno MM, Takada S, Epstein DJ. 2005. Wnt-dependent regulationof inner ear morphogenesis is balanced by the opposing andsupporting roles of Shh. Genes Dev 19:1612–1623.

Robledo RF, Lufkin T. 2006. Dlx5 and Dlx6 homeobox genes are requiredfor specification of the mammalian vestibular apparatus. Genesis44:425–437.

Shao RX, Otsuka M, Kato N, Taniguchi H, Hoshida Y, Moriyama M,Kawabe T, Omata M. 2005. Acyclic retinoid inhibits humanhepatoma cell growth by suppressing fibroblast growth factor-mediated signaling pathways. Gastroenterology 128:86–95.

Shen CN, Marguerie A, Chien CY, Dickson C, Slack JM, Tosh D. 2007.All-trans retinoic acid suppresses exocrine differentiation andbranching morphogenesis in the embryonic pancreas. Differentiation75:62–74.

Simeone A, Acampora D, Pannese M, D’Esposito M, Stornaiuolo A,Gulisano M, Mallamaci A, Kastury K, Druck T, Huebner K,Boncinelli E. 1994. Cloning and characterization of two membersof the vertebrate Dlx gene family. Proc Natl Acad Sci USA 91:2250–2254.

Studer M, Papperl H, Marshall H, Kuroiwa A, Krumlauf R. 1994. Role ofa conserved retinoic acid response element in rhombomere restric-tion of Hoxb-1. Science 265:1728.1732.

Theiler K. 1972. The house mouse: development and normal stages fromfertilization to 4 weeks of age. Berlin: Springer-Verlag. p 168.

Thompson DL, Gerlach-Bank LM, Barald KF, Koenig RJ. 2003. Retinoicacid repression of bone morphogenetic protein 4 in inner eardevelopment. Mol Cell Biol 23:2277–2286.

Torres M, Giraldez F. 1998. The development of the vertebrate inner ear.Mech Dev 71:5–21.

Verloes A, Seret N, Bernier V, Gonzales M, Herens C, Koulischer L. 1991.Branchial arch anomalies in trisomy 18. Ann Genet 34:22–24.

Vieux-Rochas M, Coen L, Sato T, Kurihara Y, Gitton Y, Barbieri O,LeBlay K, Merlo G, Ekker IV, Kurihara H, Janvier P, Levi G. 2007.Molecular dynamics of retinoic acid-induced craniofacial malforma-tions: implications for the origin of gnathostome jaws. PLoS ONE2:e510.

William SS, Mear JP, Liang H-C, Potte SS. 2004. Large-scale reprogram-ming of cranial neural crest gene expression by retinoic acidexposure. Physiol Genom 19:184–197.

Wright TJ, Mansour SL. 2003. Fgf3 and Fgf10 are required for mouse oticplacode induction. Development 130:3379–3390.

Yokohama-Tamaki T, Ohshima H, Fujiwara N, Takada Y, Ichimori Y. 2006.Cessation of Fgf10 signaling, resulting in a defective dental epithelialstem cell compartment, leads to the transition from crown to rootformation. Development 133:1359–1366.

Zelarayan LC, Vendrell V, Alvarez Y, Dominquez-Frutos E, Theil T,Alonso MT, Maconochie M, Schimmang T. 2007. Differentialrequirements for FGF3, FGF8 and FGF10 during inner ear develop-ment. Dev Biol 308:379–391.

Zhao Z, Stock D, Buchanan A, Weiss K. 2000. Expression of Dlx genesduring the development of the murine dentition. Dev Genes Evol210:270–275.

Zhu CC, Yamada G, Blum M. 1997. Correlation between loss of middleear bones and altered goosecoid gene expression in the branchialregion following retinoic acid treatment of mouse embryos in vivo.Biochem Biophys Res Commun 235:748–753.

144 LIU ET AL.

Birth Defects Research (Part B) 83:134–144, 2008