the differential sensitivities of inner ear structures to retinoic acid during development

The Differential Sensitivities of Inner Ear Structuresto Retinoic Acid during Development

Daniel Choo, Jean-Luc Sanne, and Doris K. Wu1

Laboratory of Molecular Biology, National Institute on Deafness and Other CommunicationDisorders, National Institutes of Health, Rockville, Maryland 20850

In order to examine the mechanisms that underlie development of the inner ear, the normal processes were perturbed usingall-trans-retinoic acid (RA). By implanting a resin exchange bead saturated with RA into stage 16 (Hamburger and Hamilton,1951, J. Morphol. 88, 49–92) embryonic day 2.5 chick ears, it was possible to analyze its in vivo effects on inner eardevelopment. There is a temporal window during which the developing chick inner ear is particularly susceptible to theeffects of RA (stages 16–19). This RA period of sensitivity precedes evidence of gross morphologic or histologicdifferentiation by at least 24 h, suggesting that mechanisms controlling formation of key inner ear structures are already inprogress. There is a dose dependence on RA, with increasing doses of RA generating increasingly severe phenotypicabnormalities. Data indicate that these effects are due to differential sensitivities of the various inner ear structures to RAduring their formation. In general, the vestibular structures were more susceptible to RA effects than the cochlear duct.Furthermore, nonsensory structures such as semicircular canals seemed to display a greater susceptibility to RA thantheir associated sensory structures (i.e., cristae). Among the three semicircular canals, the superior canal was the mostsusceptible to RA treatment, whereas the common crus was particularly resistant, suggesting that the molecularmechanisms for each structure’s formation are different. The defect in semicircular canal formation is due to problems inthe initial outgrowth of the canal plate which in turn is related to a down-regulation of early otocyst cell proliferation. Thisperturbation model provides valuable insight into the processes involved in producing the intricate patterning of the innerear.

Key Words: chick embryo; inner ear; retinoic acid; perturbation; dose-dependent; sensory organ; BMP4; SOHo-1; Msx-1.

INTRODUCTION

Normal development of the inner ear involves a highlycomplicated series of events that have proven difficult toanalyze. In the chick, the development of eight sensoryorgans (three cristae, the utricular and saccular maculae,the macula neglecta, the basilar papilla, and the lagena) andtheir associated nonsensory structures, as well as the en-dolymphatic apparatus, must occur correctly for all in orderto form a normally functioning inner ear. Sorting out thecausal relationships between different morphogeneticevents during development is critical to the understandingof inner ear formation.

This study utilized all-trans-retinoic acid (RA), a primarymetabolite of vitamin A, as a perturbing agent. The terato-

genic effects of either an excess or a deficiency of maternalvitamin A in humans have been recognized since themid-1900s. Developmental defects such as anophthalmia,endocardial cushion defects, and skeletal defects, as well asthymic, thyroid, and craniofacial abnormalities, have allbeen described (Warkany and Schraffenberger, 1946; Wilsonet al., 1953). Developmental effects and biologic activityhave also been observed with other endogenously generatedretinoids (such as RA, 13-cis-RA, or 9-cis-RA) or syntheticretinoid metabolites (Eckhardt and Schmitt, 1994; Charp-entier et al., 1995). Perturbation of the normal developmen-tal process using these various retinoid compounds hasproven to be a useful technique in dissecting complexdevelopmental systems such as limb formation (for reviewsee Paulson, 1994; Tabin, 1995) and retina development(Marsh-Armstrong et al., 1994).

Several lines of evidence support a role for RA in thedeveloping inner ear. RA has been shown by an in vitroreporter assay to be produced endogenously in the inner ear

1 To whom correspondence should be addressed at NIDCD, 5Research Court, Room 2B34, Rockville, MD 20850. Fax: (301)402-3470. E-mail: [email protected].

DEVELOPMENTAL BIOLOGY 204, 136–150 (1998)ARTICLE NO. DB989095

136 0012-1606/98 $25.00

(Kelley et al., 1993). Transcripts of cellular retinoic acidbinding protein-1 have been shown in the developing oticepithelium (Sanne and Wu, 1994) while nuclear receptorsfor RA have also been demonstrated in the region of thedeveloping inner ear region (Smith and Eichele, 1991; Lee etal., 1995). Compound retinoic acid receptor null mutantstudies have subsequently shown that these nuclear RAreceptors are critical for normal inner ear development(Lohnes et al., 1994). In other experiments, in vitro cultureof the organ of Corti have shown that RA can induce thegeneration of supernumerary hair cells (Kelley et al., 1993).In vitro assays using cultured chick otocysts indicate thatRA can inhibit mitogen-induced cellular proliferation andalter normal otocyst morphogenesis as well as inducepremature histologic differentiation of the cultured earepithelium (Represa et al., 1990; Leon et al., 1995). How-ever, there are limitations on the extent to which one canextrapolate from such in vitro experiments to normal invivo events. This is particularly problematic because cul-tured inner ear tissues undergo limited morphogenesis.Alternatively, when RA is administered systemically topregnant mice, a variety of inner ear malformations areinduced in the offspring (Frenz et al., 1996). However, thesystemic administration of RA in vivo makes it difficult todifferentiate indirect interactions that produce secondarychanges in the inner ear from direct effects of the perturbingagent.

This study of RA effects was performed using an in vivochick model of inner ear development. By employing aRA-saturated bead delivery system (Eichele et al., 1984), itwas possible to deliver RA directly to the ear, thus mini-mizing the known systemic effects of RA that mightindirectly influence inner ear development. Our resultsindicate that even at low doses, exogenous RA can perturbnormal chick inner ear development when delivered at lowlevels by this system. A dose-dependent effect was seen,with lower doses affecting vestibular structures and higherdoses affecting both vestibular and auditory structures.These effects occurred during a restricted period of devel-opment when the inner ear is particularly susceptible toexogenous RA and occurred at least 24 h prior to any signsof gross or histologic differentiation. Expression patterns oftwo genes normally expressed in the inner ear at the time ofRA treatment were found to be altered as a result of RAbead implantation, suggesting possible molecular pathwaysby which RA exerts its effects.

MATERIALS AND METHODS

RA Bead Preparation

The following procedure was empirically determined usingtritium-labeled RA ([3H]RA; sp act 52.1 Ci/mmol; DuPont, Inc.).One milligram of positively charged resin exchange beads (AGIX-8,Bio-Rad, Inc.) was suspended in 300 ml of dimethyl sulfoxide(DMSO). A 100-ml aliquot of the bead suspension, which corre-sponded to approximately 100 beads, was centrifuged briefly. The

pellet was resuspended in solutions of RA (in DMSO) ranging from0.0625 to 1.0 mg/ml. After a 20-min incubation, the beads werewashed once with DMSO and twice with phosphate-buffered saline(PBS). Beads ranging in diameter from 85 to 100 mm were used forimplantation. Studies using labeled RA showed that under theconditions outlined above, the amount of RA bound to the beadswas saturated after 20 min of incubation and that the amount of RAbound to the beads was linear within the range of RA solutionsused. The amount of RA loaded per bead when using a 0.125, 0.25,or 0.5 mg/ml RA solution was equivalent to 4, 8, or 12 ng of RA,respectively. Scintillation counts taken after washing the beadsovernight in PBS showed no significant change in the amount of[3H]RA bound to the beads.

RA Bead Implantation

One RA-soaked bead was implanted directly into the rightotocyst of each stage 16, E2.5 (Hamburger and Hamilton, 1951)chick embryo. At this stage, one bead filled the majority of theotocyst cavity (Fig. 2). After implantation, eggs were resealed withcellophane tape and incubated at 38°C until the desired develop-mental time point. Embryos were harvested at various stagesbetween E3 and E7. For analysis of otoliths, some embryos wereallowed to incubate until E12 or E14.

For delayed implantation experiments, precision in staging ofotic development was critical. Accordingly, embryos (and in par-ticular, the otocysts) were first exposed and staged at the normalimplantation time. This allowed a selection of embryos that wereat the same otic developmental point, thereby reducing the vari-ability noted even when eggs were incubated for identical periods.The eggs that showed correct developmental times (28–31 somites)were returned to the incubator for an additional 8 h. At that point,they were reopened and one bead was implanted into each rightotocyst. Embryos were harvested at E7 for paint injection.

For late-implantation experiments, stage 19 embryos (38–40somites) were implanted according to the same protocol.

For bead-removal experiments, the eggshell windows were re-opened after the planned time interval, and a tungsten needle wasused to atraumatically remove the bead. Eggs were then resealedand reincubated until E7.

Embryo Harvesting/Fixation and Processing/PaintInjection

Embryos were harvested at various stages between E3 and E7 andimmediately fixed in Bodian’s fixative, then dehydrated in ethanoland cleared in methyl salicylate. The embryonic ears were thenvisualized by injecting a 0.5% white latex paint in methyl salicy-late to fill the inner ear labyrinth (Martin and Swanson, 1993;Bissonnette and Fekete, 1996; Morsli et al., 1998).

In Vivo RA Release

[3H]RA was used to determine the kinetics of RA binding andrelease from the resin exchange beads. Beads were prepared usingthe same technique as described above except for the addition of 2mCi of [3H]RA. These beads were implanted into stage 16 chickotocysts and then removed for scintillation counting after differentimplantation time intervals. In this manner, an estimate of theactual in vivo release of RA was obtained.

137Retinoic Acid and Inner Ear Phenotypes

Copyright © 1998 by Academic Press. All rights of reproduction in any form reserved.

Whole-Mount and Serial Section in SituHybridization

In situ hybridization experiments were focused on identifyingearly changes in gene expression as a result of RA treatment.Accordingly, embryos were implanted with a single 0.250 mg/mlRA bead in the right ear and then harvested 24 h later andimmediately fixed in 4% paraformaldehyde (pH 7.4) for in situhybridization. The 24-h interval was chosen in hopes of identifyingmore immediate changes in gene expression resulting from RAbead implantation. The left ear of each embryo served as thecontrol for each specimen.

A full-length BMP4 cDNA (approximately 1.5 kb) (Roberts et al.,1995) and a 550-bp fragment of Msx-1 cDNA downstream of thehomeobox domain (Suzuki et al., 1991) were used to generatedigoxigenin-labeled sense and antisense RNA probes. Similarly,digoxigenin-labeled RNA probes for SOHo-1 were generated from a1.2-kb PvuII-digested fragment which included approximately 200bp of coding sequence and 1 kb of a 39 untranslated sequence butdid not include the homeobox domain (Kiernan et al., 1997). In situhybridization was carried out as described (Wu and Oh, 1996).

Evaluation of the in situ hybridization data was performedindependently by two investigators in the lab. A third investigatorblinded to the treatment of the embryos also evaluated some of thespecimens.

Whole-Mount Hair-Cell-Specific Antigen Staining

To assess sensory organ formation in chick ears treated with RA,a whole-mount immunohistochemical staining using a monoclo-nal antibody against a hair-cell-specific antigen (anti-HCA; Good-year et al., 1995) was performed. Embryonic ears were implantedwith RA beads as above, harvested at E7, and processed forwhole-mount staining as described (Wu et al., 1997). A total of 31RA-treated embryos were analyzed by whole-mount anti-HCAimmunostaining.

Otolith Examination

In order to examine the effects of high-dose RA on otolithformation associated with macular organ development, beadstreated with 0.75 and 1.00 mg/ml RA were implanted into stage 16ears and the embryos harvested at E12–14. These specimens werefixed overnight in 4% paraformaldehyde and then grossly dissectedout of the temporal bones. The presence or absence of the otolithscould be scored by direct observation of the inner ear in thismanner. To confirm the nature of the otolithic sensory structures,these same specimens underwent anti-HCA labeling as describedabove. Overall, a total of 22 embryos were implanted with high-dose RA beads and then analyzed for otolith formation.

Assays of Cell Proliferation and Programmed CellDeath

Cell proliferation was examined by 5-bromo-29-deoxyuridine(BrdU) incorporation. Each embryo was treated with a bead soakedin a 0.25 mg/ml RA solution as per the routine and then pulsedwith BrdU (3 mg/ml) 3.0 h prior to selected harvest times whichranged from E3.5 to E5. After the inner ears were paint-filled, four(RA-treated) embryos that displayed malformed inner ears, as wellas six RA-treated E3.5 embryos, were embedded in paraffin, and7-mm alternating serial sections were taken at the level of the ear.

For E3.5 embryos, every other section was scored for BrdU label.For E5 embryos, every fourth section from the dorsal half of theinner ear, excluding the endolymphatic duct and sac, was scored forBrdU label. The E5 scored region was further divided into ananterior and posterior domain using the endolymphatic duct as themedian line (see straight line in Figs. 6C and 6D).

Programmed cell death was assayed using a modified terminaldeoxynucleotidyl transferase-mediated dUTP nick end-labeling(TUNEL) staining technique (Apotag Plus, Oncor, Inc). Embryoswere similarly implanted with 0.250 mg/ml RA beads and thenharvested at selected time points between E3.5 and E5. Tissueprocessing was performed essentially as recommended by themanufacturer. Eight embryos were sectioned and analyzed forapoptotic profiles.

RESULTS

Release of RA in Vivo

In order to interpret the effects of adding exogenous RA tothe developing ear, it is necessary to know the rate at whichRA leaches from the delivery vehicle into the target tissue.With a bead soaked in 1.00 mg/ml of RA, approximately11% of the original amount of RA remained in the beadafter 8 h implantation in vivo (Fig. 1). Similar RA releasecurves were observed with beads soaked in lower doses ofRA (data not shown).

Dose-Dependent Effects of RA

Once it was established that the beads provided a gooddelivery vehicle for RA to the inner ear, beads that had beensaturated in RA solutions ranging from 0.0625 to 1.00mg/ml were implanted into stage 16 inner ears. To assessthe effects of RA administration, white latex paint wasinjected into the inner ear cavities of embryos that had beenimplanted with a RA bead to the right ear, allowed toincubate until E7, and then harvested and fixed as above. Aspectrum of phenotypes was observed, which for the pur-poses of analysis were grouped into five types (Fig. 3), withtype 1 being the mildest abnormality and type 5 being themost severe malformation. Beads similarly prepared, butwithout RA, served as control beads. No inner ear malfor-mations were observed in the specimens implanted withcontrol beads (n 5 9; Fig. 3A). In the type 1 phenotype (Fig.3B), only one semicircular canal was affected, most com-monly the superior semicircular canal (SSC). In the type 2phenotype (Fig. 3C), two canals were affected, most com-monly the superior and lateral semicircular canals (LSC)although rarely the superior and posterior semicircularcanals (PSC) were affected. In the type 3 phenotype (Fig.3D), all semicircular canals were absent while the cochlearduct appeared normal. The type 4 phenotype (Fig. 3E)demonstrated an absence of all semicircular canals and ashortened or poorly differentiated cochlear duct. And last,the type 5 phenotype (Fig. 3F) showed only a rudimentaryinner ear mass, typically with some type of endolymphaticapparatus projection and a common crus.

138 Choo, Sanne, and Wu


When the concentration of RA was varied, a dose-dependent relationship was apparent (Table 1). The lowestdose of RA (0.0625 mg/ml) produced only one mild pheno-type in one of nine embryos. A low dose (0.125 mg/ml)generated normal or mild phenotypes (i.e., only one canalaffected) approximately 87% of the time. High doses (1.00mg/ml), in contrast, generated a large proportion of type 4and 5 phenotypes (86%) and only rarely a lesser phenotype.Intermediate doses of RA showed a transitional pattern.The 0.250 mg/ml dose, for example, generated approxi-mately 53% type 1 phenotypes, but also started to producemore severe phenotypes. The 0.500 mg/ml dosage generatedinner ear malformations in a somewhat bimodal distribu-tion. While this dose of RA did produce more type 4 and 5phenotypes (43% type 4 and 5), there were also a significantnumber of type 1 phenotypes produced (23%). Interestingly,there were few type 3 phenotypes generated with any of the

doses used in these experiments. Of 155 embryos examinedin these experiments, only five type 3 phenotypes wereobtained (3%).

Subtle abnormalities of the endolymphatic apparatuswere noted but could not be adequately analyzed andcompared to controls based solely upon paint-injectionexamination. However, increased size of the sac, increasedlength of the whole apparatus, and rare duplication of theapparatus were observed in these specimens. No obviousdose-dependent phenomenon was noted for these endolym-phatic apparatus abnormalities.

As seen in Fig. 2, the common crus appeared to be fairlyresistant to RA perturbation. Even in the most severephenotypes (Fig. 3F), a thin common crus projection can benoted.

These data indicate a differential sensitivity to RA amongthe various inner ear structures. For instance, among the

FIG. 1. RA remaining in beads over time of implantation. Beads were soaked in 1.0 mg/ml RA solutions containing 0.2 mCi [3H]RA andprocessed as described under Materials and Methods. After the beads were implanted for the specified time, they were removed, crushed,and processed for scintillation counting. Regardless of dose used, the majority of the RA is released within the first 4 h. Each point is themean value from several beads (n ^ 6) with ranges representing the mean 6 standard error of the mean.

TABLE 1Distribution of Chick Inner Ear Phenotypes According to Concentration of RA Solution Used to Saturate Implanted Beads

Conc. RA(mg/ml) Type 0 Type 1 Type 2 Type 3 Type 4 Type 5 Total

0.000 9 0 0 0 0 0 90.0625 8 1 0 0 0 0 90.125 15 18 4 0 1 0 380.250 8 24 4 2 3 4 450.500 5 7 2 3 2 11 300.750 1 0 1 0 0 7 91.00 1 1 0 0 3 10 15

Total 155



nonsensory structures of the chick inner ear, the semicir-cular canals seem particularly susceptible to RA. The SSCcan further be identified as the most susceptible of thecanals. Interestingly, other nonsensory structures such asthe endolymphatic apparatus and common crus show aparticular resistance to RA effects as evidenced by theirgrossly normal formation even in the type 5 phenotypes.Specific sensory organs such as the macula utriculi andmacula sacculi similarly showed a resiliency to the effectsof RA as evidenced by their formation after high-doseimplants in type 5 inner ears.

Sensory Organ Generation and Otolith Formation

In the inner ear, transduction of sensory stimuli takesplace at the hair-cell level. Groups of these specialized haircells are organized into discrete patches (cristae and macu-lae in the vestibular portion and a basilar papilla in theauditory portion) housed within associated specializedstructures. Cristae, for example, are normally found withinthe ampullae of the semicircular canals. Paint-injectiondata typically showed normal or partial development ofampullae in mild phenotypes while no discrete ampullaewere observed in more severe phenotypes. To determine

whether sensory organs still developed despite the absenceof associated ampullae and/or semicircular canals, immu-nohistochemical staining of whole-mount inner ears wasperformed using a monoclonal antibody against hair-cell-specific antigen (anti-HCA) (Fig. 4).

In the type 1 phenotype (Fig. 4A), all eight patches ofsensory hair cells were observed (two macular organs, onemacula neglecta, three cristae, the basilar papilla, and thelagena). The superior crista in some cases, however, wasnoted to be significantly smaller and misshapen in compari-son to control ears (control ear not shown). Anti-HCA-labeled inner ears with intermediate phenotypes (types 2and 3) demonstrated the capacity to form normal-appearingcristae despite the absence of associated semicircular canalstructures (data not shown). Even in the most rudimentaryinner ears (type 5, Figs. 4B and 4C), multiple discrete haircell patches were readily identified by anti-HCA immuno-staining, including two particularly prominent hair cellpatches which resembled the two maculae. To verify theidentity of these sensory organs, embryos treated withhigh-dose RA beads were harvested at E12 to allow time forotolith formation. Such embryos (Fig. 4C) confirmed theidentity of the two prominent hair cell patches as rudimen-tary macula utriculi and macula sacculi by the generationof macular-specific otolithic structures. In addition, veryrudimentary and malformed cristae-like patches were fre-quently noted in type 5 phenotypes (Fig. 4C—small arrows).

Taken together, these data show that sensory organs(such as the maculae and cristae) are more resistant to RAtreatment than (associated) nonsensory inner ear structures(such as the canals). The maculae, in particular, seemespecially resistant to RA, differentiating and forming oto-lithic structures even after high-dose implantation, whileother sensory organs fail to develop properly.

Temporal Factors in RA-Induced Phenotypes

To determine whether the timing of RA exposure playeda significant role in the generation of inner ear phenotypes,we varied the stage of RA bead implantation. RA beads wereimplanted between stages 16 and 21. Consistent with ourfirst experiments described above, when 0.250 mg/ml beadswere implanted at stage 16, only 17% of the resultingembryos had normal inner ears. In contrast, by stage 19(38–40 somites) the inner ear is markedly less sensitive toexogenous RA. When same-dose beads were implanted intostage 19 ears, 88% of inner ears were normal (Table 2).Implantation of RA beads was also attempted at earlier timepoints (stage 15). However, the presence of an open otic cup,which persists until stage 16, made implantation erratic.Results at stage 15 were therefore inconsistent due todislodging of the bead from the original implantation site.Accordingly, while we could not define the onset of sensi-tivity to RA, our results indicate a temporal windowextending at least between stages 16 and 19 during whichthe chick inner ear is particularly sensitive to the effects ofRA.

FIG. 2. Embryonic day 2.5 chick in ovo showing the typicalappearance of the right otocyst before bead implantation. A typicalRA bead is shown adjacent to the otocyst. The inset shows theappearance after a bead has been implanted into the otocyst.



The most straightforward interpretation of our study ofRA dose dependence was that there is a differential sensi-tivity of the various inner ear structures. Accordingly, therequirement for higher doses of RA in order to generate

cochlear malformations would indicate that the cochlearduct is less sensitive to RA than other structures such asthe semicircular canals. However, another plausible inter-pretation of our results was that the period of sensitivity of

FIG. 3. RA-induced phenotypes in the developing chick inner ear. (A–F) Paint-filled right inner ears from E7 chick embryos treated withvarying doses of RA and shown in an anterior–lateral view. (A) Normal right inner ear. (B) Type 1 phenotype with absence of the SSC. (C)Type 2 phenotype with absence of two semicircular canals (SSC and LSC). (D) Type 3 phenotype with absence of all semicircular canalstructures. Note that an identifiable cochlear projection and endolymphatic apparatus are still present. A common crus (cc) is alsoidentifiable. (E) Type 4 phenotype with malformation of the cochlea in addition to the semicircular canals. (F) Type 5 phenotypedemonstrating a rudimentary inner ear mass. Only the common crus and the endolymphatic apparatus are apparent in this specimen.Abbreviations: ssc, superior semicircular canal; psc, posterior semicircular canal; lsc, lateral semicircular canal; s.amp, superior ampulla;p.amp, posterior ampulla; l.amp, lateral ampulla; cc, common crus; els, endolymphatic sac. The orientation axes shown in (A) are similarfor all specimens. A, anterior; D, dorsal; M, medial. Scale bars, 100 mm.



the different inner ear structures to RA is staggered suchthat the optimal sensitivity of the cochlear duct lies slightlyafter the stage when RA was applied. An earlier sensitivityfor the semicircular canals than for the cochlear duct wouldfit well with morphologic and histologic data which showthat in general, the vestibular components of the inner eardifferentiate noticeably earlier than the auditory compo-nent. Our observed kinetics of RA release from the beadcould also be consistent with this hypothesis. Our in vivorelease experiments showed that approximately 11% of theoriginal RA remains in the bead after 8 h, which would givea lower effective dose to a late-differentiating structuresuch as the cochlea. To test this possibility, experimentswere conducted removing the bead after 8 h and, alterna-tively, delaying implantation of the bead by 8 h.

To eliminate effects induced by residual RA, high-dosebeads (0.500 mg/ml) were implanted at stage 16 and thenremoved after 8 h. Removing the beads after 8 h generatedmore normal and mild phenotypes than parallel specimensin which the beads were left in the ears (Table 3). Inparticular, the generation of cochlear phenotypes was mark-edly reduced by bead removal. In routine 0.500 mg/mlimplants, 43% type 4 and 5 (cochlear) phenotypes wereobserved. In contrast, when the beads were removed, only3% (1 of 29) showed a type 4 or 5 phenotype.

In a second series of experiments, bead implantation wasdelayed 8 h. In a model in which cochlear patterning occurslater than vestibular patterning, such late implants wouldbe expected to affect the cochlear duct to a greater extentwhile perhaps sparing the already patterned semicircularcanals. In these experiments, isolated cochlear phenotypeswere never observed. Data show a trend toward milderphenotypes (data not shown) and significantly, no canal andcochlear phenotypes were obtained when using either 0.125or 0.250 mg/ml beads.

Taken together, these results support a model in whichthe different inner ear structures have a differential sensi-tivity to RA during a window of susceptibility betweenstages 16 and 19. The dose-dependent phenotypes we ob-served are the result of each structure’s unique sensitivityto RA. Eight-hour bead removal as well as 8-h-delayedexperiments suggest that perturbation of the cochlea re-quires not only a high-dose implant, but also a sustainedexposure (longer than 8 h).

Malformation of Semicircular Canals

In normal inner ear development, the semicircular canalsform from two outpouches or “canal plates,” one for the

FIG. 4. Sensory organs and otolith formation in RA-induced phenotypes. (A) Whole-mount immunostaining of a type 1 inner ear using amonoclonal anti-HCA antibody. The labeled patches of hair cells identify the sensory organs visible on this frontomedial view of the innerear. Note that the superior crista (s.cr) in this inner ear lacks an associated SSC. The crista is identifiable but appears small and lacks theusual w shape. All other patches of sensory hair cells appear normal. The posterior crista and macula neglecta are not visible from thisperspective but were noted to be normal. (B) A type 5 inner ear after whole-mount anti-HCA immunostaining shown in a medial view. TheRA bead is visible in the ear, as are two anomalous and unusually small cristae (arrows). Two large heterogeneous patches ofanti-HCA-labeling are marked (arrowheads). The possibility of these patches representing a rudimentary utricle and saccule was examinedby looking for the otoliths normally associated with these two sensory organs. (C) A type 5 inner ear from another specimen. Two separateotolithic structures are identified. Arrowheads mark a plate-like otolith when viewed medially. Small arrows mark a second andabnormally globular-appearing otolith. psc, posterior semicircular canal; lsc, lateral semicircular canal; els, endolymphatic sac; s.cr,superior crista; l.cr, lateral crista; utr, utricle; sac, saccule; bp, basilar papilla; lg, lagena. Orientation axes and scale bar are shown in (C) andare the same for all images. D, dorsal; A, anterior. Scale bar, 100 mm.



LSC and one common pouch for the SSC and PSC. Thecanal plate for the SSC and PSC starts forming around E4.By early E6, the opposing epithelia of the canal plate cometogether in the central region, fuse, then are “resorbed”leaving the mature canal structure. Proposed mechanismsresponsible for this canal morphogenesis have includedretraction of epithelial cells and epithelial-to-mesenchymaltransformation (Martin and Swanson, 1993). Recently, pro-grammed cell death (PCD) has been demonstrated to play arole in this patterning process as well (Fekete et al., 1997).

In the following series of experiments looking at theeffects of RA on PCD, cell proliferation, and gene expres-sion, we utilized an intermediate RA dose (0.25 mg/ml)based upon the frequency of phenotypes generated with thisdose bead. As shown in Table 1, a 0.25 mg/ml bead could beexpected to yield approximately 53% type 1 inner ears andmuch fewer normal or severely malformed inner ears. Thisapproach was used in hopes of focusing our attention oncanal defects and allowing us to probe possible mechanismsinvolved in generating such canal defects.

Paint injection from E3.5 to E4.5 showed no gross ana-tomical difference between RA-treated and control ears(data not shown). A defect of the primordial SSC was first

evident at early E5 (Fig. 5) (21 of 37 paint-injected embryos)which is an age before epithelial fusion and PCD normallyoccur. To determine whether a premature induction of PCDwas involved in generating the observed phenotypes, amodified TUNEL assay was employed. Figures 6E and 6Fshow representative sections from the canal plate of E5control and RA-treated ears. Only scattered apoptotic pro-files are noted for both control and experimental ears.Embryos were also TUNEL stained at E3.5, 24 h after RAbead implantation (Fig. 7). Again, the number of labeledcells in both control and RA-treated ears was scarce. Takentogether, these data suggest that PCD does not play asignificant role in establishing the primordial canal defect.

Alternatively, we examined whether a down-regulationof cell proliferation was involved in producing the primor-dial canal defect noted at E5. A BrdU-incorporation assaywas used to detect cell proliferation. Embryos implantedwith a 0.250 mg/ml bead were harvested at E5 after a 3-hpulse with BrdU. These specimens were first paint-injectedto confirm the phenotype, then sectioned and immuno-stained to detect BrdU incorporation (Figs. 6A–6D). Countsof BrdU-labeled cells from four specimens showed a 38%reduction of proliferating cells in the anterior region of thecanal plate (a % 0.025). However, since there was a grossreduction in the size of the anterior canal plate in RA-treated inner ears (compare Figs. 6A, 6C, and 6E with 6B,

TABLE 2The Effects of RA on the Developing Chick Inner Ear Dependupon the Stage of Implantation

Stage at implantationNormal

phenotypeAbnormal

phenotypea

Stage 16 (28–31 somites)b 8 37Stage 19 (38–40 somites)b,c 14 2

Note. x2 5 21.93, a , 0.005.a Abnormal phenotype means anything other than a normally

formed inner ear (i.e., phenotypes 1–5).b The same intermediate dose (0.250 mg/ml) of RA was used for

both experimental groups.c Results were pooled from two independent experiments.

TABLE 3Removal of Implanted RA Beads after 8 h Dramatically Reducesthe Generation of Cochlear Phenotypes

Implantationmethod

Noncochlearphenotype

(normal and types1, 2, and 3)

Cochlearphenotype

(types 4 and 5)

Normal implanta 17 138-h bead removala,b 19 1

a The same 0.500 mg/ml RA dose was used for both experimentalgroups.

b Results generated from two independent experiments.

FIG. 5. Paint-injected right inner ears from control and RA-treated specimens harvested at embryonic day 5. The control ear(A) shows the normal outgrowth anterosuperiorly (white arrows)that forms the superior canal primordium. The white arrowheadsidentify the primordium for the posterior canal. The RA-treated ear(0.250 mg/ml bead) in (B) displays a reduced outgrowth of tissuedorsally as well as anterosuperiorly (compare white arrows in A andB). The posterior and lateral canal regions, the early endolymphaticapparatus, and the cochlea, in this specimen, show no appreciabledifference from the control. At E5, the lateral canal is still rudi-mentary. Black arrows (in A and B) mark the very early outgrowthof tissue for the developing lateral canal. ela, endolymphaticapparatus. Scale bars, 100 mm.



6D, and 6E), we then digitized the anterior and posteriorepithelial areas and calculated the relative density of pro-liferating cells (number of BrdU-labeled cells/mm2) for con-trol and RA-treated ears. This analysis showed no statisti-cally significant difference in the density of BrdU-labeledbetween RA-treated and control ears (a ^ 0.10). Of note, thenumber and density of BrdU-labeled cells in the posteriorregion was not significantly reduced (a ^ 0.10).

To address the possibility that RA treatment might haveaffected cell proliferation earlier in development, we ana-lyzed BrdU incorporation 24 h after RA bead implantation(E3.5) (Fig. 8). At this stage, there were no obvious differ-ences in size and/or shape of RA-treated and control oto-cysts. However, counts of the total number of BrdU-labeledcells showed a 23% reduction among the RA-treated ears(compared to controls) (a &lE; 0.005) (Table 4). Based on ouranalysis, the spatial pattern of BrdU-labeled cells at differ-ent levels of the inner ear remained comparable betweentreated and control otocysts (Fig. 8).

Gene Expression Affected by RA Treatment

In an attempt to explore genes that may be affected by RAtreatment, we investigated the expression patterns of threegenes: SOHo-1, BMP4, and Msx-1, 24 h after RA treatment(Fig. 9). These genes were chosen because each is a markerfor some of the inner ear structures affected by RA treat-ment. SOHo-1, for example, is an early marker for the threesemicircular canals and their cristae (Kiernan et al., 1997).BMP4 is a marker for all presumptive sensory organs in thechick inner ear, and Msx-1 a marker for the endolymphaticduct (Wu and Oh, 1996). Our interest in examining Msx-1expression was further piqued by previous studies, whichdemonstrated the common localization of this gene inregions demonstrating susceptibility to exogenous RA(Lyons et al., 1992). In addition, all three of these genes arenormally activated during the time of RA treatment, mak-ing them possible modifying targets of RA.

BMP4 has been shown to be an early marker for sensoryorgans in the developing chick inner ear (Wu and Oh, 1996). Its

FIG. 6. Cell proliferation and programmed cell death in E5RA-treated and control ears. Transverse sections from the controlear (A, C, E) and RA-treated ear (B, D, F) of an E5 specimen.(Schematic insets in C and D show the levels from which thesesections were taken.) Note the normal clustering of BrdU-labeledcells in the posterior region and the sparser signal in the anterior

region of the control ears (A, C). BrdU incorporation was noted ina similar pattern in the RA-treated ears (B, D). Also note thereduced anterior–posterior dimension of the RA-treated ear at theselevels. The solid black lines in C and D mark the approximateanterior–posterior median using the endolymphatic duct as areference point. Scattered BrdU-labeled cells are apparent in thedeveloping endolymphatic duct (ED). Representative TUNEL-stained sections from an E5 specimen are shown in E and F. Arrowsmark the positively labeled cells. On similar sections through thecanal plate of both RA-treated and control ears, only scatteredapoptotic profiles could be demonstrated. ED, endolymphatic duct;arrows, TUNEL-labeled cells. Scale bars, 100 mm. Orientation axesin A are similar for C and E. Orientation axes in B apply for Dand F.



expression is already restricted during the time when RAbeads were implanted. Twenty-four hours after RA treatment,a down-regulation of BMP4 was observed by in situ hybridiza-

tion in an anterior domain in 6 of 11 specimens (Figs. 9A and9B), corresponding to the presumptive anterior crista (Wu andOh, 1996). This is consistent with the proposed role of BMP4

FIG. 7. Programmed cell death in the E3.5 inner ear. (A) Schematic diagram of the E3.5 ear. “B”, “D”, “E”, and “F” indicate the approximatelevels of sections B, D, E, and F. Sections from the RA-treated ear are shown in B, D, and F. Note the relatively rare and scattered TUNEL-labeledcells. The pattern and frequency of these labeled cells were similar to those seen in control ears (representative section shown in E). Rat mammarygland sections (C) were used as positive controls. Orientation axes in B apply for D and F. Scale bar, 100 mm.



FIG. 8. BrdU incorporation in the E3.5 inner ear. (A, C, E) Transverse sections from the control ear of a RA-treated specimen. (B, D, F)Transverse sections from the RA-treated ear taken at levels corresponding to those in A, C, and E. Note that the patterns and regions ofthe otocyst that have labeled cells are similar for both RA-treated and control ears. In particular, the medial aspect of the ears, where theotic epithelium appeared thicker, consistently showed a larger number of proliferating cells than the thinner lateral aspect. Strong BrdUsignal could also be seen in all areas of the mesenchyme surrounding the developing ear. Similarly, obvious cell proliferation could bedemonstrated in the hindbrain on all sections. Orientation axes in E apply for A, C, and E, while the orientation axes in F also apply forB, D, and F. Scale bar, 100 mm.



in specifying the presumptive crista and the small, abnormallyshaped anterior cristae seen in some specimens. The posteriorsignal was unchanged after RA treatment.

SOHo-1, a sensory organ homeobox gene, is normallyexpressed in a differential manner in the developing chickinner ear (Kiernan et al., 1997). This pattern includes thelateral otic epithelium that is believed to contribute tosemicircular canal formation (Figs. 9C and 9D). Surpris-ingly, despite the susceptibility of the canals to RA, nochanges in SOHo-1 expression were observed 24 h after RAtreatment (n 5 8). Based upon the dose–phenotype datashown in Table 1, we anticipated that a 0.250 mg/ml RAbead should have generated more than 50% type 1 pheno-types. Accordingly, if SOHo-1 gene expression were affectedby RA treatment, we would have observed changes inapproximately half the specimens.

Msx-1 also demonstrates specific expression domains inthe developing chick inner ear during the developmentaltimes being studied here (Wu and Oh, 1996). Our interest inMsx-1 expression after RA treatment was further stimu-lated by prior studies demonstrating a RA-responsive ele-ment within the promoter region of the human Msx-1 gene(Shen et al., 1994). Such a regulatory element implies afairly rapid inducibility of this gene in response to elevatedRA levels. Normal Msx-1 expression includes robust hy-bridization signal in the developing endolymphatic duct, inthe edges of the neural tube, and later on in the developingcristae as well as the adjacent mesenchyme and portions ofthe semicircular canals (Suzuki et al., 1991; Wu and Oh,1996). Increased Msx-1 signal was seen in the posteriorperiotic mesenchyme as a result of RA treatment in 9 of 15evaluated specimens (arrowheads in Figs. 9E and 9F). Ex-pression of Msx-1 in the endolymphatic duct and adjacenthindbrain was unchanged by RA exposure (thin arrow andmultiple arrows in Figs. 9E and 9F).

DISCUSSION

RA-Induced PhenotypesUsing RA as a perturbing agent provided insight into the

processes governing normal development of the inner ear.

There is a narrow window during development duringwhich the inner ear is sensitive to RA and this periodprecedes most overt histologic differentiation by at least24 h. Since the various vestibular structures and the co-chlear duct demonstrate a similar RA sensitivity window,early molecular events responsible for the formation ofthese structures must occur at about the same time indevelopment, despite the earlier gross differentiation ofvestibular structures in comparison to the cochlea. Sincecochlear phenotypes were only generated with higher dosesof RA and under conditions that always generated vestibu-lar phenotypes as well, we cannot rule out from theseresults that formation of vestibular and cochlear structuresmay be linked. However, the existence of multiple mutantsin mice and humans that have isolated vestibular or co-chlear duct defects indicate that, at some point, theirdevelopmental pathways diverge.

Our study repeatedly demonstrated specific effects of RAon inner ear development. For example, the vestibularcomponents of the inner ear were more sensitive to RAthan the auditory component. Also, the nonsensory struc-tures such as the semicircular canals seemed more vulner-able to RA than their corresponding sensory organs, thecristae. Among all the nonsensory components of the innerear, structures such as the endolymphatic apparatus andcommon crus were much more resilient to RA treatmentthan the semicircular canals. The eight sensory organs alsoshowed a differential sensitivity to RA treatment with themaculae showing a striking resistance to RA perturbation,even at high doses. Furthermore, despite the fact thatnormal morphogenesis is largely impaired with higherdoses of RA, some sensory patches still differentiatedenough for hair-cell formation as evident by the anti-HCAstaining. This is consistent with mutant zebrafish data(Kalicki et al., 1996; Whitfield et al., 1996) and other reports(Swanson et al., 1990) which show that hair-cell differen-tiation still occurs despite the lack of normal morphogen-esis.

Although RA beads were implanted directly into theotocyst, we cannot rule out the possibility that the effectsof RA could be mediated by indirect effects on adjacenttissues (e.g., hindbrain or mesenchyme), which then influ-ence the otic epithelium. However, in limited experimentsin which RA beads were implanted into the hindbrain(region of rhombomeres 5 and 6), only occasional mild innerear abnormalities could be demonstrated by paint-injectionassay.

Possible Mechanism of Canal Malformation

Our results indicate that the defect in semicircular canalformation is due to problems in the initial outgrowth of thecanal plate, a conclusion supported by the demonstration ofa down-regulation of otic epithelial cell proliferation 24 hafter RA treatment. We could not identify any specificregions of the E3.5 otocyst that displayed a focal down-regulation of BrdU incorporation. It is possible that our

TABLE 4RA Reduces the Total Number of BrdU-Labeled Cells in E3.5Inner Ears

Specimen Control ear RA-treated eara

1 3811 31662 2599 18233 2811 20424 2405 20565 2455 16346 2398 2026

Note. t 5 7.22 (t0.005 5 4.03).a Using 0.25 mg/ml RA bead.



FIG. 9. Patterns of BMP4, SOHo-1, and Msx-1 expression in stage 19 normal and RA-treated chick inner ears. (A–F) Whole-mount in situhybridization of chick embryos that underwent 0.250 mg/ml RA-bead implantation in the right ear at stage 16 and were then harvested 24 hlater. (B,D,F) RA-treated right ears. (A,C,E) Control left ears from the same embryos. RA treatment results in a change of expression patternof BMP4 at the anterior crista (ac) in (B) while hybridization signals from other sensory patches such as the macula sacculi (sac), lateral crista(lc), and posterior crista (pc) remain unchanged. Instead of a large, well-defined patch of signal on the control side (ac) in (A), a much smallerspot is appreciable on the RA side (ac) in (B). Hybridization signals dorsal to the inner ear are contributed by BMP4 expression in thesurrounding mesenchyme in (A) and (B); see also Figs. 6A and 6C in Oh et al. (1996). SOHo-1 expression in the developing ear showed noobvious changes as a result of RA treatment. Arrowheads in (C) and (D) mark the area of normally weak or absent SOHo-1 expression atthe anterior ventral aspect of the stage 19 otocyst that is unchanged after RA exposure. Note the strong signal encompassing the lateralotocyst and extending ventrolaterally. Arrowheads in (E) and (F) mark corresponding regions of Msx-1 signal at the posterior oticmesenchyme (see also Fig. 4B Wu and Oh, 1996). Stronger Msx-1 expression is appreciable on the RA-treated side in comparison to thecontrol. Signals in the endolymphatic duct (arrows in E and F) are unchanged. Normal expression at the edges of the neural tube (smallarrows) are unchanged as well. ac, anterior crista; lc, lateral crista; pc, posterior crista; sac, saccule. Scale bars, 100 mm.



analysis of the BrdU-labeled cells may not be sensitiveenough to detect subtle differences within discrete regionsof the otocyst. However, analysis of over 15 specimens atvarious stages between E3.5 and 5.0 showed the pattern oflabeled cells to be remarkably consistent between controland RA-treated inner ears, even though the total number ofproliferating cells was different.

Furthermore, even though the three semicircular canalsare similar in morphology and develop at approximately thesame time, their differential sensitivity to RA suggests thatthe molecular mechanisms underlying each canal’s forma-tion may be different. This hypothesis is supported by aclinical finding that a patient with Goldenhar’s syndromehas an intact superior and posterior semicircular canal withthe isolated absence of the common crus (Manfre et al.,1996).

Molecular Patterning in the Developing ChickInner Ear

The expression patterns of three genes (BMP4, SOHo-1,and Msx-1) were examined 24 h after RA treatment. Thereduction in BMP4 expression in the presumptive superiorcrista correlates with the smaller sized cristae observed atE7 and is consistent with the postulated role of BMP4 insensory organ specification (Wu and Oh, 1996). It is notclear if the changes in BMP4 expression are direct effects ofRA since BMP4 expression in the other sensory organs wasnot equally affected. However, the altered Msx-1 expressionobserved here may very well be a direct consequence of RAtreatment since RA-responsive elements are present in thepromoter region of the human MSX-1 gene (Shen et al.,1994), and Msx-1 gene expression is often colocalized withthe gene expression of RA receptors in other organ systems(Lyons et al., 1992). It is not clear how the change in Msx-1gene expression in the periotic mesenchyme contributedto the observed phenotypes. However, epithelial–mesenchymal interactions are believed to play a significantrole in the shaping of the inner ear.

Other members of the SOHo gene family, such as Nkx5-1(Hmx3), play an important role in semicircular canal andampulla formation (Hadrys et al., 1998; Wang et al., 1998).SOHo-1 most likely serves a function similar to that ofNkx5-1 based upon their similar, but not identical, expres-sion patterns (Kiernan et al., 1997; Herbrand et al., 1998) Itis interesting that there was no appreciable difference inSOHo-1 gene expression 24 h after RA treatment, eventhough the semicircular canals were the structures mostfrequently affected by RA.

Finally, even though there is no direct evidence that RAacts to regulate normal inner ear development, the specificand dose-dependent phenotypes elicited by RA stronglysuggest that this important embryonic signal plays a role inpatterning these structures. The particular sensitivity of thevestibular structures and cochlear duct indicate that preciselevels of RA may be essential for their normal patterning.Consistent with this view, both RA receptors and RA-

binding proteins are specifically expressed in the developingear and therefore, the endogenous role of this importantmolecule warrants further study.

ACKNOWLEDGMENTS

The authors acknowledge Drs. Cliff Tabin, James Battey, andSusan Sullivan for critically reviewing the manuscript and Ms.Mirene Boerner for her editorial assistance. The generous gift of theanti-HCA antibody (G. Richardson) is also gratefully acknowl-edged.

REFERENCES

Bissonnette, J. P., and Fekete, D. M. (1996). Standard atlas of thegross anatomy of the developing inner ear of the chicken.J. Comp. Neurol. 368, 620–630.

Charpentier, B., Bernardon, J. M., Eustache, J., Millois, C., Martin,B., Michel, S., and Shroot, B. (1995). Synthesis, structure–affinityrelationships, and biological activities of ligands binding toretinoic acid receptor subtypes. J. Med. Chem. 38, 4993–5006.

Eckhardt, K., and Schmitt, G. (1994). A retinoic acid receptor alphaantagonist counteracts retinoid teratogenicity in vitro and re-duced the incidence and/or severity of malformations in vivo.Toxicol. Lett. 70, 299–308.

Eichele, G., Tickle, C., and Alberts, B. M. (1984). Microcontrolledrelease of biologically active compounds in chick embryos: Beadsof 200 mm diameter for the local release of retinoids. Anal.Biochem. 142, 542–555.

Fekete, D. M., Homburger, S. A., Waring, M., Riedl, A., and Garcia,L. (1997). Involvement of programmed cell death in morphogen-esis of the vertebrate inner ear. Development 124, 2451–2461.

Frenz, D. A., Liu, W., Galinovic-Schwartz, V., and van de Water,T. R. (1996). Retinoic acid-induced embryopathy of the mouseinner ear. Teratology 53, 292–303.

Goodyear, R., Holley, M., and Richardson, G. (1995). Hair andsupporting cell differentiation during the development of theavian inner ear. J. Comp. Neurol. 351(1), 81–93.

Hadrys, T., Braun, T., Rinkwitz-Brandt, S., Arnold, H. H., andBober, E. (1998). Nkx5-1 controls semicircular canal formation inthe mouse inner ear. Development 125(1), 33–39.

Hamburger, V., and Hamilton, H. (1951). A series of normal stagesin the development of the chick embryo. J. Morphol. 88(1),49–92.

Herbrand, H., et al. (1998). Two regulatory genes, cNkx5-1 andcPax2, show different responses to local signals during oticplacode and vesicle formation in the chick embryo. Development125(4), 645–654.

Kelley, M. W., Xu, X. M., Wagner, M. A., Warchol, M. E., andCorwin, J. T. (1993). The developing organ of corti containsretinoic acid and forms supernumerary hair cells in response toexogenous retinoic acid in culture. Development 119, 1041–1053.

Kiernan, A. E., Nunes, F., Wu, D. K., and Fekete, D. M. (1997).Unique expression domain of two members of a novelhomeobox-containing gene family suggests a role in semicircularcanal formation during inner ear development. Dev. Biol. 191,215–229.

Lee, Y. M., Osumi-Yamashita, N., Ninomiya, Y., Moon, C. K.,Eriksson, U., and Eto, K. (1995). Retinoic acid stage-dependently



alters the migration pattern and identity of hindbrain neuralcrest cells. Development 121, 825–837.

Leon, Y., Sanchez, J. A., Miner, C., Ariza-McNaughton, L., Represa,J. J., and Giraldez, F. (1995). Developmental regulation of Fos-protein during proliferative growth of the otic vesicle and itsrelation to differentiation induced by retinoic acid. Dev. Biol.167, 75–86.

Lohnes, D., Mark, M., Mendelsohn, C., Dolle, P., Diereich, A.,Gorry, P., Gansmuller, A., and Chambon, P. (1994). Function ofthe retinoic acid receptors (RARs) during development. I. Cranio-facial and skeletal abnormalities in RAR double mutants. Devel-opment 120, 2723–2748.

Lyons, G. E., Houzelstein, D., Sassoon, D., Robert, B., and Buck-ingham, M. E. (1992). Multiple sites of Hox-7 expression duringmouse embryogenesis: Comparison with retinoic acid receptormRNA localization. Mol. Reprod. Dev. 32, 303–314.

Malicki, J., Schier, A. F., Solnica-Krezel, L., Stemple, D. L., Neuhauss,S. C. F., Stanier, D. Y. R., Abdililah, S., Rangini, Z., Zwartkruis, F.,and Driever, W. (1996). Mutations affecting development of thezebrafish inner ear. Development 123, 275–283.

Manfre, L., Genuardi, P., Tortorici, M., and Lagalla, R. (1996).Absence of the common crus in goldenhar’s syndrome. Am. J.Neuroradiol. 18, 773–775.

Marsh-Armstrong, N., McCaffery, P., Gilbert, W., Dowling, J. E.,and Drager, U. C. (1994). Retinoic acid is necessary for thedevelopment of the ventral retina in zebrafish. Proc. Natl. Acad.Sci. USA 91, 7286–7290.

Marshall, H., Nonchev, S., Sham, M. H., Muchamore, I., Lumsden,A., and Krumlauf, R. (1992). Retinoic acid alters hindbrain hoxcode and induces transformation of rhombomeres 2/3 into a 4/5identity. Nature 360, 737–740.

Martin, P., and Swanson, G. J. (1993). Descriptive and experimentalanalysis of the epithelial remodelling that control semicircularcanal formation in the developing mouse inner ear. Dev. Biol.159, 549–558.

Morsli, H., Choo, D., Ryan, A., Johnson, R., and Wu, D.K. (1998).Development of the mouse inner ear and its sensory organs.J. Neurosci. 18, 3327–3335.

Oh, S. H., Johnson, R., and Wu, D. K. (1996). Differential expressionof bone morphogenetic proteins in the developing vestibular andauditory sensory organs. J. Neurosci. 16, 6463–6475.

Paulson, D. F. (1994). Retinoic acid in limb-bud outgrowth: Reviewand hypothesis. Anat. Embryol. 190, 399–415.

Represa, J., Sanchez, A., Miner, C., Lewis, J., and Giraldez, F. (1990).Retinoic acid modulation of the early development of the innerear is associated with the control of c-fos expression. Develop-ment 110, 1081–1090.

Roberts, D. J., Johnson, R. L., Burke, A. C., Nelson, C. E., Morgan,B. A., and Tabin, C. (1995). Sonic hedgehog is an endodermal signal

inducing BMP4 and Hox genes during induction and regionalizationof the chick hindgut. Development 121, 3163–3174.

Sanne, J-L., and Wu, D. K. (1994). Expression of retinoic acidbinding protein in the developing chick otocyst. 34th AnnualMeeting of the American Society for Cell Biology, p. 230a.[Abstract 1336]

Shen, R., Chen, Y., Huang, L., Vitale, E., and Solursh, M. (1994).Characterization of the human MSX-1 promotor and an enhancerresponsible for retinoic acid induction. Cell. Mol. Biol. Res. 40(4),297–312.

Smith, S. M., and Eichele, G. (1991). Temporal and regionaldifferences in the expression pattern of distinct retinoic acidreceptor-b transcripts in the chick embryo. Development 111,245–252.

Suzuki, H. R., Padanilam, B. J., Vitale, E., Ramirez, F., and Solursh,M. (1991). Repeating developmental expression of g-hox 7, anovel homeobox-containing gene in the chicken. Dev. Biol. 148,375–388.

Swanson, G. J., Howard, M., and Lewis, J. (1990). Epithelial au-tonomy in the development of the inner ear of a bird embryo.Dev. Biol. 137, 243–257.

Tabin, C. (1995). The initiation of the limb bud: growth factors,Hox genes, and retinoids. Cell 80, 671–674.

Wang, W., Van De Water, T., and Lufkin, T. (1998). Inner ear andreproductive defects in mice lacking the Hmx3 homeobox gene.Development 125, 621–634.

Warkany, J., and Schraffenberger, E. (1946). Congenital malforma-tions induced in rats by maternal vitamin A deficiency. I. Defectsof the eye. Arch. Ophthalmol. 35, 150–169.

Whitfield, T. T., Granato, M., van Eeden, F. J. M., Schach, U.,Brand, M., Furutani-Seiki, M., Haffter, P., Hammerschmidt, M.,Heisenberg, C-P., Jiang, Y-J., Kane, D. A., Kelsh, R. N., Mullins,M. C., Odenthal, J., and Nusslein-Volhard, C. (1996). Mutationsaffecting the development of the zebrafish inner ear and lateralline. Development 123, 241–254.

Wilson, J. G., Roth, C. B., and Warkany, J. (1953). An analysis of thesyndrome of malformations induced by maternal vitamin Adeficiency. Effects of restoration of vitamin A at various timesduring gestation. Am. J. Anat. 92, 189–217.

Wu, D. K., Nunes, F., and Choo, D. (1997). Axial specification inthe developing chick inner ear. Development 125, 11–20.

Wu, D. K., and Oh, S. H. (1996). Sensory organ generation in thechick inner ear. J. Neurosci. 16, 6454–6462.

Received for publication August 20, 1998Revised September 23, 1998

Accepted September 23, 1998



the differential sensitivities of inner ear structures to retinoic acid during development

Documents