Development 112, 541-550 (1991)Printed in Great Britain © The Company of Biologists Limited 1991

541

Hyaluronan as a propellant for epithelial movement: the development of

semicircular canals in the inner ear of Xenopus

CATHERINE M. HADDON AND JULIAN H. LEWIS

Imperial Cancer Research Fund, Developmental Biology Unit, Department of Zoology, University of Oxford, South Parks Rd, Oxford,0X1 3PS, UK

Summary

The membranous labyrinth of the inner ear, with itsthree semicircular canals, originates from a simplespheroidal otic vesicle. The process is easily observed inXenopus. The vesicle develops three dorsal outpocket-ings; from the two opposite faces of each outpocketingpillars of tissue are protruded into the lumen; and thesepaired 'axial protrusions' eventually meet and fuse, toform a column of tissue spanning the lumen of theoutpocketing like the hub of a wheel, with a tube ofepithelium forming the semicircular canal around theperiphery. Each axial protrusion consists of epitheliumencasing a core of largely cell-free extracellular matrixthat stains strongly with alcian blue. In sections, at least60 % of the stainable material is removed by treatmentwith Streptomyces hyaluronidase. When Streptomyceshyaluronidase is microinjected into the core of aprotrusion in vivo, the protrusion collapses and the

corresponding semicircular canal fails to form. Hyaluro-nan (hyaluronic acid) in the core of the protrusiontherefore seems to be essential in driving the extension ofthe protrusion. Autoradiography with tritiated glucosa-mine indicates that the hyaluronan-rich matrix issynthesised by the epithelium covering the tip of theprotrusion; the basal lamina here appears to bediscontinuous. These findings indicate that the epi-thelium of the axial protrusion propels itself into thelumen of the otocyst by localised synthesis of hyaluro-nan. Hyaluronan may be used in a similar way in thedevelopment of other organs, such as the heart and thesecondary palate.

Key words: hyaluronan, inner ear, semicircular canals,Xenopus.

Introduction

All vertebrates are equipped in a similar way with anelaborately shaped inner ear. This has its beginnings ina patch of thickened ectoderm, the otic placode, on theside of the head of the embryo adjacent to thehindbrain. The placode invaginates and pinches offfrom the surface to become the otic vesicle, which thendevelops into the membranous labyrinth of the innerear through an extraordinary process of epithelialgrowth and remodelling, as shown in Fig. 1. The simplespheroidal vesicle is transformed into a shape withprecisely positioned bulges and projections and, moststrikingly, with three toroidal 'handles', the semicircu-lar canals. In this paper, we examine how thesemicircular canals are created. They provide a veryclear illustration of how a localised biochemical activitybrings about a transformation of shape.

The geometry of semicircular canal formation hasoften been described before (von Noorden, 1883;Krause, 1903; Streeter, 1907; Paterson, 1948; Knowl-ton, 1967; Waterman and Bell, 1984; Hertwig, 1987).The otic vesicle develops a set of three flattenedoutpocketings, one for each future duct. The centres of

the two opposite faces of each outpocketing becomeindented, forming axial protrusions into the lumen ofthe vesicle from opposite sides. These protrusionseventually contact one another and, at the site ofcontact, the two apposed epithelial sheets fuse andreorganise, creating a squat pillar of tissue spanning thelumen of the outpocketing (see Fig. 2). Around thispillar, like the tube of a car tyre around the wheel hub,lies the epithelium of the newly formed semicircularcanal, derived from the peripheral part of the epi-thelium of the outpocketing, and communicating ateach end with the lumen of the body of the otic vesicle.The topological transformation is the same in allvertebrates; there are, however, variations in thequantitative details of the geometry. In higher ver-tebrates, the initial outpocketings are pronounced andstrongly flattened, and the axial protrusions are merelylow, broad mounds, whereas in fish (Waterman andBell, 1984) and amphibians, including Xenopus (onwhich we focus here), the outpocketings are much lessapparent, and the axial protrusions are finger-shapedand much more clearly defined (Fig. 3; see alsoPaterson (1948) and Hertwig (1987)).

The protrusions are the crucial structures in the

542 C. M. Haddon and J. H. Lewis

Fig. 1. Clay models of a right inner ear of Xenopus atthree successive stages in the formation of the semicircularcanals. Lateral views above, medial views below. (A) Stage44: slight indentations represent the axial protrusionsbeginning to extend into the ear lumen. (B) Stage 47: threestubby semicircular canals have formed through the fusionof the axial protrusions. The dorsal structure is theendolymphatic sac at the end of the endolymphatic duct.(C) Stage 48+: The semicircular canals have extended andbecome thinner. (Endolymphatic duct and sac not shown).

Fig. 2. Schematic cut-away views of three stages in theformation of a semicircular canal. (A) Initial outpocketingof otocyst epithelium. (B) Axial protrusions extend intolumen. (C) Axial protrusions meet and fuse, creating asemicircular canal.

standard methods (Godsave et al. 1988) and allowed todevelop in dechlorinated tap water. They were stagedaccording to the developmental tables of Nieuwkoop andFaber (1967).

HistologyEmbryos to be used for histology were either fixed at 4°C inhalf-strength Karnovsky fixative (2% paraformaldehyde,1.25% glutaraldehyde, 0.25% calcium chloride in 0.1Mcacodylate buffer; Karnovsky, 1965) and subsequently em-bedded in Araldite or Fibrowax (BDH), or were fixed inBodian's fixative (75 % absolute alcohol, 18 % distilled water,2 % formaldehyde, 5 % glacial acetic acid) or Sainte Mariefixative (1 % glacial acetic acid in 95 % alcohol) and thenembedded in Fibrowax (Tuckett and Morriss-Kay, 1988).Araldite sections were cut at 3 fim and stained with toluidineblue. Fibrowax sections of embryos fixed in half-strengthKarnovsky fixative were cut at 6/an and stained forglycosaminoglycans (GAGs) with alcian blue at pH2.5 (1%alcian blue 8GX, 3 % glacial acetic acid, in distilled water) for30min (Humason, 1972). Some sections were further counter-stained with neutral red. To assess how much of the alcian-blue stainable material was hyaluronan, sections wereincubated, before alcian-blue staining, with Streptomyceshyaluronidase (Sigma, type IX) at concentrations rangingfrom 0.015 to l.Omgml"1 in sodium acetate buffer (pH5.2,0.1M), for 4h at 37 °C. These sections were compared withadjacent control sections incubated in buffer alone. The stainintensity was quantified by spot-metering of the lighttransmitted through the regions of interest, using the accuratephotomultiplier-based digital light-meter incorporated in aLeitz Vario-Orthomat camera system on a Leitz Diaplanmicroscope, with a xlOO objective. The diameter of themetered spot was 11 /im. The intensity of light, Ls, transmittedthrough the stained core of a protrusion was compared withthe intensity, Lo, transmitted through the empty lumen of theotic vesicle, and the absorption coefficient, A=(LO—Ls)/Lo,was calculated. For our specimens, values of A were relativelysmall - in the range 0.1 to 0.3 depending on the enzymetreatment used - and should therefore be very nearly linearlyproportional to the amount of stained material present. Thefraction of stained material remaining after enzyme treatmentis therefore given by the ratio Aenzyme/Acontroi, i.e. the ratioof the A value for the enzyme-treated specimen to the A valuefor an adjacent control section treated with buffer only. Thisratio is the 'relative stain intensity' referred to in the Resultssection below.

development of semicircular canals, and our concernhere is with the mechanism that drives their formation.We give evidence that they are created by localisedproduction of an extracellular matrix, of which hyaluro-nan (hyaluronic acid) is an essential component. Thisspace-filling molecule is apparently produced by theotic epithelium, with enhanced levels of synthesis at thesites destined to form the axial protrusions. Synthesis ofhyaluronan-rich matrix lifts the cells of these specificregions away from the mesenchyme and propels theminto the lumen of the otic vesicle.

Materials and methods

Xenopus laevis embryos were obtained and dejellied by

ImmunohistochemistryFor the localisation of chondroitin sulphate, wax sections ofSte. Marie-fixed embryos were stained with the CS56monoclonal antibody (Sigma, C8035) followed by a goat anti-mouse IgM, /x-chain specific FITC conjugate (Sigma, F9259),according to the method of Morriss-Kay and Tuckett (1989).Sections were counter-stained with DAPI (S^gml"1 in PBS)and mounted in gelvatol (Canning and Stern, 1988). Tomonitor the laminin distribution, we used a rabbit anti-laminin antiserum (ICN, 68-125) on Fibrowax sections oflarvae fixed in Bodian's fixative, detected with an affinity-purified, goat anti-rabbit IgG-specific, FITC-conjugated sec-ondary antibody (ICN, 61-640-1). The sections werecounterstained with DAPI and mounted in gelvatol as above.For precise localization, some of the immunofluorescentsections were viewed using a BioRad MRC 500 confocalscanning laser microscope.

Hyaluronan in Xenopus ear morphogenesis 543

Electron microscopyFor scanning E.M., embryos were dissected to expose thelumen of the otic vesicle and the tip of an epithelial protrusionand fixed at 4°C in half-strength Karnovsky fixative; they werethen critical-point dried and sputter-coated in the usual way.Specimens for T.E.M. were similarly fixed in half-strengthKarnovsky fixative, post-fixed in osmium tetroxide andstained with uranyl acetate and lead citrate.

In vivo injectionsLiving tadpoles were anaesthetised in MS222 (Sandoz) andimmobilized in gelatin wells at stage 46, when axialprotrusions were well defined but not yet fused. The enzymeschondroitinase ABC (Sigma, C2905) and Streptomyceshyaluronidase (Sigma, type IX, H1136) were dissolved in full-strength NAM saline (Slack et al. 1973) at varying concen-trations, with the addition of a small amount (approximately0.01%) of trypan blue to make the injected fluid visible.According to the manufacturer's data, the Streptomyceshyaluronidase had a specific activity of 1050 hyaluronidaseunits mg"1, while the chondroitinase ABC had a specificactivity of 0.5 chondroitinase units mg"1. To avoid confusionbetween the two types of activity unit, based on quitedifferent assays, we specify enzyme concentrations in mg ml"1

in this paper. Enzyme solution was microinjected by mouth-applied pressure into the acellular core of a protrusion using aglass microelectrode, bevelled to a tip diameter of roughly3 ;«n. Approximately 0.01 nl fluid was injected although thevolume was hard to regulate exactly. Control injections weremade using NAM medium with trypan blue but without anyenzymes. The effect of the injections on the morphology ofthe axial protrusions and the subsequent semicircular canaldevelopment was monitored by light microscopy of livingembryos and by analysis of transverse sections.

AutoradiographyTo localise sites of GAG synthesis by incorporation of aradioactive GAG precursor, we used stage 46 larvae in whichsemicircular canal protrusions are just beginning to elongate.To facilitate the uptake of the radioactive precursor by theinner ear, some of these were anaethetised with MS222, killedsurgically and used to prepare explants taken in the form ofthick transverse sections that included both ears, togetherwith the neural tube and ventral tissues; other larvae werekept alive and intact. Specimens were placed in either half-strength NAM solution or in a solution containing 60% L15medium (Sigma), 10% fetal calf serum and 30% distilledwater. To both solutions we added tritiated glucosamine(glucosamine hydrochloride,D-[l,6-3H(N)], specific activity52.7Cimmon1, from NEN Research Products) diluted togive a final activity of O.WmCiml"1. The larvae and tissueexplants were incubated at 22°C in the labelled media for lhand either fixed straight away or incubated for a further 20 h inunlabelled media (whole tadpoles in NAM/2, explants inL15/fetal calf serum medium) before fixing. We used half-strength Karnovsky fixative at 4°C and embedded in Araldite;sections were cut at 3/mi, dipped in Ilford K2 emulsion andleft to expose for 2 to 3 weeks at 4°C before developing inKodak D19 developer (Wilkinson and Green, 1990), counter-staining with toluidine blue and mounting in DePeX (BDH).

Results

Sequence of anatomical eventsThe inner ear develops rapidly in Xenopus and the head

of the tapdole soon becomes transparent so that thelater stages of ear morphogenesis can be observedeasily in vivo, as seen in Fig. 3. The sequence andtiming of the key events, based on the standarddescriptions of Nieuwkoop and Faber (1967) and ofPaterson (1948) as well as on our own observations, is asfollows for animals reared at 22-24°C. The otic vesiclecloses and pinches off from the surface ectoderm atstages 27-28 (31-32.5h). The first signs of sensorydifferentiation are seen at stage 33/34 (44 h), whensensory hair cells begin to be visible by scanning E.M.At stage 41 (76 h), a slight outpocketing of theepithelium on the lateral wall of the dorsal part of thevesicle marks the first step in the formation of thehorizontal (lateral) semicircular canal. A little later, atstage 42 (80 h), a similar but more extended outpocket-ing develops along the topmost edge of the vesicle; thissoon becomes divided by a slight depression into ananterior and a posterior outpocketing, corresponding tothe future anterior (superior) and posterior semicircularcanals.

The paired axial protrusions belonging to the lateraloutpocketing appear first, at stage 43 (87 h), growtowards one another, and finally fuse at their tips atabout stage 46 (106 h) to form the hub of the lateral ductsystem. The other two semicircular canals develop inthe same way, the anterior usually a little before theposterior; both their pairs of protrusions originate atapproximately stage 45 (98 h) and have fused by aboutstage 46.5 (120 h) although there is some randomvariability in the timing. For each of the three ducts, thepaired protrusions, having fused at their tips, thenbecome moulded over the course of a few hours intosmooth pillars linking opposite faces of the originaloutpocketings. Transverse sections illustrating the stepsin this process are shown in Fig. 4.

Each pillar has a largely acellular core filled withfibrous matrixOne can see by viewing the living tadpole withinterference contrast optics that each axial protrusion iscovered by a thin shell of epithelium and that its core isusually empty of cells, or contains at most one or two.This is confirmed by serial sections, which show theprotrusions to have a core of extracellular matrix,visible also by scanning E.M., as shown in Fig. 5.Alcian-blue staining at pH2.5 (Fig. 6A) indicates thatthe matrix is rich in glycosaminoglycans (GAGs)(Pearse, 1968). In at least one other system - thedeveloping heart - where an epithelium becomes locallyseparated from mesenchyme by an accumulation ofacellular matrix, the matrix has been shown to consistlargely of one particular GAG, hyaluronan (Markwaldet al. 1978). To test whether hyaluronan was a majoringredient of the matrix in the cores of our protrusions,we treated sections with the highly specific hyaluroni-dase from Streptomyces (Pintar, 1978; Derby andPintar, 1978) before staining with alcian blue (Fig. 6B),and compared the resultant intensity of staining withthat seen in adjacent control sections that had not beentreated with the enzyme. The amount of stain bound in

544 C. M. Haddon and J. H. Lewis

the core of the protrusion in treated and controlsections was measured photometrically for a series ofdifferent concentrations of hyaluronidase, each appliedfor 4h, as described in Materials and methods. Theresults are shown in Fig. 7. This indicates that forconcentrations of Streptomyces hyaluronidase above

Fig. 4. Araldite sections showing the formation of asemicircular canal. (A) A pair of axial protrusions (arrows)filled with extracellular matrix grow into the ear lumen.Note the thickened appearance of the epithelium coveringthe rounded tips of the protrusions. (B) The protrusionsextend and meet, fusing at their tips. Cell death occurs atthe fusion plate (arrow). (C) Epithelial remodelling occurscreating a smooth cylinder of epithelium covering a columnwhose core, initially acellular, later becomes seeded withmesenchymal cells. The completed canal lies perpendicularto the plane of the page. Scale bar=80,um.

0-5

Enzyme Concentration (mg/ml)10

Fig. 5. S.E.M. of an axial protrusion with its apicalepithelium removed, viewed end-on from the luminal side.Note the finely fibrous matrix in the core. Scalebar=10/«n.

Fig. 7. Graph showing the amount of alcian-blue stainablematerial remaining in the core of an axial protrusionfollowing a 4-h digestion with Streptomyces hyaluronidaseat various concentrations. The intensity of stain wasmeasured photometrically as described in 'Materials andmethods'; the 'relative stain intensity' plotted along theabscissa is the stain intensity for the given enzymeconcentration divided by the intensity for an adjacentcontrol section exposed to zero enzyme concentration.(Thus the relative stain intensity for zero enzymeconcentration is exactly 1.0 by definition). The solid circlesrepresent a series of measurements made on multiplesections of a single tadpole. The hollow square atl.Omgml"1 represents a set of measurements on sectionsof 10 different tadpoles. The error bars represent thestandard error of the mean. The curve is drawn by eye.

0.5mgml~l, at least 60% of the alcian-blue-stainablematerial is lost. We can conclude that this proportion ofthe GAGs in the core of the protrusion either consists ofhyaluronan or depends on binding to hyaluronan for itsretention in situ.

When hyaluronidase is injected into the core of anaxial protrusion in vivo, it collapses, aborting thedevelopment of the semicircular canalA further experiment served to test whether thehyaluronan is functionally important. We immobilisedtadpoles in anaesthetic at stage 46, when the anteriorand posterior pairs of axial protrusions are usually well-formed (about twice as long as they are wide) but havenot yet both fused. Using a fine glass micropipette, wethen injected roughly 0.01 nl of enzyme solution into thecore of the lateral protrusion of either the posterior orthe anterior pair. Two different GAG-degradingenzymes were tested: chondroitinase ABC and Strepto-myces hyaluronidase, both at a variety of concen-trations. Chondroitinase ABC acts on several sub-strates, degrading chondroitin sulphates A and C,dermatan sulphate and hyaluronan (Derby and Pintar,

Fig. 3. Dorsal view of an intact tadpole at stage 46, showing the location of the ears adjacent to the hindbrain. A bluepigment (monastral blue) has been injected into the lumen of the ear to make its outline clear. Scale bar=100/nn.

"urn.

6A

Fig. 6. Fibrowax sections stained with alcian blue at pH 2.5 and counterstained with neutral red. (A) Control section.(B) Adjacent section treated with Streptomyces hyaluronidase (lmgml"1) for 4h at 37°C before alcian blue staining. Theenzyme has removed most of the alcian-blue-stainable material from the core of the axial protrusion (arrow). Scaleb 7 5

Fig. 9. Collapse of an axial protrusion after Streptomyces hyaluronidase (0.01 nl, lmgml ') is injected into its core, as seenin dorsal views of the ear of a live, anaesthetised stage 46.5 tadpole. (A) Just before injection of enzyme into the lateralposterior axial protrusion (arrow). (B) The same specimen Imin after injection: the protrusion is already beginning todeflate. (C) After 10min the protrusion has collapsed to less than half its former size. Scale bar=70,um.

Hyaluronan in Xenopus ear morphogenesis 545

1978; Morriss-Kay and Tuckett, 1989) while Strepto-myces hyaluronidase degrades hyaluronan only (Ohyaand Kaneko, 1970). As a control, buffer devoid ofenzymes was injected into the same site in siblingtadpoles at the same developmental stage.

Injection of buffer in all 16 control specimens had noapparent effect: the puncture wound made by themicropipette healed rapidly and the protrusion con-tinued to elongate normally, resulting in the formationof a morphologically normal canal. By contrast, in 13out of 17 specimens injected with Streptomyces hyalu-ronidase at a concentration of either 0.5 or lmgml"1,the injected protrusions had collapsed to less than halftheir original length within 4 h and usually much sooner.No recovery was seen over the following 48 h and thesemicircular canal failed to form, as shown in Fig. 8.The other protrusions in the same ear usually alsoretracted, although not so rapidly, presumably throughthe effects of enzyme that had diffused from the site ofinjection. A typical example of the collapse is shown inFig. 9, for a specimen injected with the hyaluronidase ata concentration of lmgml"1. The injected protrusionhas begun to retract within lmin, and by lOmin hascollapsed to a low pimple. When Streptomyces hyalu-ronidase was injected at a concentration of onlyO.lmgml"1, effects were still apparent but muchmilder; 3 out of 5 injected protrusions showed a partialcollapse or slight delay of outgrowth. 12 out of 16specimens injected with chondroitinase ABC at concen-trations of 20 or 40mgml~1 caused protrusions tocollapse as rapidly as Streptomyces hyaluronidase at 0.5or l.Omgml"1; similarly, injections of chondroitinaseABC at 2 mg mP1 caused a very mild effect in 3 out of 5protrusions, analogous with O.lmgmP1 Streptomyceshyaluronidase injections. The chondroitinase ABC thushad a collapsing activity about 20 times lower (per mg ofprotein) than the Streptomyces hyaluronidase. (Notethat according to Morriss-Kay and Tuckett (1989), thedegradative activity of chondroitinase ABC on hyaluro-

nan is approximately l/50th of its activity on chondroi-tin sulphates A or C.)

Chondroitin sulphate is absent from the cores of theprotrusions but present on their epitheliumIt is still possible that other extracellular matrixmolecules, and in particular other charged GAGs suchas chondroitin sulphate, may also contribute to theinflation of the protrusions. We examined the distri-bution of chondroitin sulphate by immunofluorescencewith the CS56 monoclonal antibody, which bindsspecifically to the GAG moiety of chondroitin sulphateproteoglycans (Morriss-Kay and Tuckett, 1989). Atypical result is shown in Fig. 10. The cartilage of the

Fig. 8. Transverse section of a tadpole 2 days after aninjection of Streptomyces hyaluronidase (—0.01 nl,lmgml"1) into the lateral axial protrusion in the right ear.On the control (left) side, the axial protrusions have metand fused to form the hub (arrow) of a semicircular canal(asterisk). By looking at serial sections, we confirmed thatnothing remained of the injected protrusion and its partnerexcept the two low pimples visible in this section. On theinjected side, the protrusions have regressed and no canalhas been formed. Scale bar=160/«n.

Fig. 10. (A) The chondroitin-sulphate distribution in the inner ear at stage 46.5 as revealed by immunofluorescence withthe CS56 antibody. (B) The pattern of cell nuclei revealed by DAPI staining. Note that there is CS56 immunoreactivity inthe epithelium of an axial protrusion (arrow) and as expected in the cartilage around the ear but not in the core of theprotrusion. Scale bar=50,um.

546 C. M. Haddon and J. H. Lewis

BFig. 11. T.E.M. sections showing (A) a portion of the squamous epithelium near the base of an axial protrusion, and (B) aportion of the rather thicker epithelium covering the tip of the same protrusion. In (A) one can see a well-defined basallamina; in (B) only an amorphous basal fuzz is present. Scale bar=2^m.

developing otic capsule is strongly labelled as expected,since chondroitin sulphate is known to be the predomi-nant GAG in cartilage matrix. By contrast, noimmunostaining is seen in the core of the adjacentprotrusion. The epithelium covering the protrusion,however, is strongly labelled; the rest of the oticepithelium is also labelled but not quite so strongly.Evidently chondroitin sulphate is absent, or occurs onlyat very low levels in the cores of the protrusions but ispresent in their epithelial covering.

Epithelial cells at the site of fusion are removed bycell deathSections of tadpoles fixed within a couple of hours afterthe moment of fusion show pycnotic cells and debris atthe site of fusion (see Fig. 4B), clearly indicating thatthe majority of epithelial cells here are eliminated bycell death, although it is possible that a few may beretracted back into the epithelium as occurs in themouse (P. Martin, personal communication). Throughremoval of cells at the site of fusion, the two opposingaxial protrusions become joined into a single squatpillar, with a continuous core of extracellular matrixand a cladding of epithelium. The matrix, initiallymostly acellular, soon becomes invaded by fibroblasts.

The basal lamina appears discontinuous beneath theepithelium covering the tips of the axial protrusionsTransmission E.M. (Fig. 11) reveals a continuous basallamina beneath most of the otic epithelium and inparticular beneath the epithelium at the bases of theaxial protrusions. But towards the tips of the pro-trusions the basal lamina seems to be absent ordiscontinuous. The findings were similar (Fig. 12) whenwe stained sections by immunofluorescence with ananti-laminin antiserum. The basal lamina appears to beinitially intact, but as the protrusion extends the laminaseems to break down under the apex of the protrusionwhile remaining intact elsewhere.

A utoradiographsWhich cells make the matrix in the cores of the axialprotrusions - the cells of the epithelium or those of the

Fig. 12. Immunofluorescence with an anti-lamininantiserum on fibrowax sections, viewed with a confocalscanning microscope. (A) An axial protrusion justbeginning to form: there appears to be a continuous layerof laminin beneath the epithelium. (B) A well-developedaxial protrusion: the layer of laminin appears discontinuousbeneath the epithelium covering the distal part of theprotrusion. Scale bar=45f«n.

mesenchyme? To find out, we examined, by autoradi-ography, the pattern of incorporation of tritiatedglucosamine, a precursor that is incorporated intoGAGs including hyaluronan and into glycoproteins.The labelled compound was supplied for 1 h at stage 46,when the anterior and posterior pairs of axial pro-trusions are about half-way in their outgrowth. Uptakeby intact tadpoles was very weak irrespective of themedium used, but dissected tissue explants (comprising

Hyaluronan in Xenopus ear morphogenesis 547

ears, hindbrain and associated ventral tisues) showedstrong incorporation. In autoradiographs of specimensfixed immediately after pulse-labelling, silver grainswere found concentrated over the epithelium of theprotrusions, and present at a much lower density overthe epithelium forming the rest of the otic vesicle(Fig. 13A). Immediately beneath the heavily labelledepithelium of the protrusion there seemed to be a thinzone of labelling of the extracellular matrix. Elsewherein the cores of the protrusions and in the mesenchymebeneath them the labelling was scarcely above back-ground levels. In autoradiographs of specimens inwhich the radioactively labelled medium was removedafter one hour and replaced with nonradioactivemedium for a further 20 h before fixation, fewer silvergrains were seen over the epithelium of the protrusionsbut now there were silver grains over the whole of theextracellular-matrix cores of the protrusions (Fig. 13B).These observations strongly suggest that GAGs (pre-sumably including hyaluronan) in the cores of the axialprotrusions are made by the epithelium, and that theepithelium of the protrusion is distinguished from therest of the otic epithelium by a high rate of incorpor-ation of glucosamine into GAGs. This fits in with thehistological picture: if the mesenchyme cells weremaking the matrix, one would expect them to beembedded in it, whereas in fact the material in the coresof the protrusions is largely acellular.

Discussion

We have demonstrated that formation of the semicircu-lar canals depends on movements of the otic epitheliumthat are driven by localised accumulation of extracellu-lar matrix at specific sites in each ear. The matrix atthese locations occupies a cell-free space between theepithelium and its underlying mesenchyme, causing thetwo to separate and thereby forcing the epithelium toprotrude into the lumen of the vesicle. Hyaluronan is anessential component of this matrix: when the hyaluro-nan is destroyed enzymatically, the cell-free spacedisappears, the protrusion collapses, and the corre-sponding semicircular canal fails to form. From thepattern of incorporation of glucosamine, it appears thatthe matrix in the protrusions is made by the epithelium:localised GAG synthesis by the epithelium seems toprovide the driving force for the localised epithelialmovement.

Comparison with previous observations of eardevelopmentWhile many authors have described the morphology ofsemicircular canal formation, only a few (von Noorden,1883, Krause, 1903; Hertwig 1987 and Waterman andBell, 1984) have commented on the accumulation ofextracellular matrix beneath the developing axialprotrusions. Waterman and Bell (1984) have reportedin the zebrafish a pattern of events very similar to that inXenopus; but while they note the presence of anacellular matrix in the cores of the protrusions they do

Fig. 13. Autoradiographs showing the pattern of[3H]glucosamine incorporation in extending axialprotrusions and adjacent tissues. Bright-field views on theleft, dark-field views on the right. (A) Explant pulsed with[3H]glucosamine at stage 46 and fixed after lh. Theepithelium of two rudimentary axial protrusions is heavilylabelled; neighbouring regions of otic epithelium are not.Scale bar=50,um. (B) Explant pulsed for lh with[3H]glucosamine at stage 46 and then cultured for 20 h innonradioactive medium before fixation. Most of the labelhas cleared from the epithelium and is seen instead in theextracellular matrix, both in the acellular core of theprotrusion and in other tissues such as cartilage, which isalso rich in GAGS. In this specimen, the starred region(less heavily labelled) is not a growing axial protrusion butan already formed axis of a semicircular canal, withflattened, inactive epithelium and mesenchymal cellsalready populating its core. Scale bar=50^m.

not investigate its composition. They do suggest,however, that the matrix is made by the epithelium atthe tips of the protrusions, which they find to be rich inrough endoplasmic reticulum, and they note theabsence of basal lamina here. McPhee etal. (1987) haveshown that hyaluronan is synthesized by the inner ear of

548 C. M. Haddon and J. H. Lewis

the mouse, with a peak at E12.5, approximately thetime when semicircular canals are forming, althoughthey have not reported the distribution of the hyaluro-nan or related it to semicircular canal formation. Weourselves see, in analogy with Xenopus, cell-free,matrix-filled spaces at sites of semicircular canalformation in the chick embryo (unpublished data); suchspaces are also visible in Knowlton's published photo-graphs (1967), though she makes no comment on them.

Hyaluronan as apropellant for epithelialmorphogenesisHyaluronan is a linear glycosaminoglycan polymer thatdiffers fundamentally from other GAGs in respect of itsgiant size, the disproportionately large volume that itoccupies, and its mode of synthesis. A single moleculeat the upper end of the hyaluronan size range has arelative molecular mass (Mr) of about 4x10 , corre-sponding to 104 repetitions of the disaccharideTV'-acetyl-D-glucosamine+D-glucuronate, and forms arandom-coil configuration with a diameter of about0.5/an (Comper and Laurent, 1978). Thus hyaluronanacts as a space-filler, and by synthesizing relatively smallquantities of it a cell can quite efficiently push itselfaway from other structures. In particular, an epitheliumsitting on a stationary substratum can propel itself awayfrom that substratum by emitting hyaluronan basally.

For this function, it is necessary that the hyaluronanshould not diffuse freely away from the site of itsproduction. Hyaluronan is exceptional among theextracellular polymers found in animal tissues in that itis not accumulated inside the cell and then secreted, butrather is spun out from the cell surface by a synthetaselocated in the plasma membrane (Prehm, 1984).Throughout the synthetic process, the molecule istherefore tethered to the exterior of the cell. Once it hasbeen released from the synthetase, the hyaluronan maybe retained by binding to another cell-surface hyaluro-nan-binding protein such as the hyaluronan receptor,CD44 (Aruffo et al. 1990), or through sparse protein-mediated linkages to other components of the extra-cellular matrix (Underhill, 1989; Toole, 1990).

Hyaluronan as an agent of morphogenesis in othersystemsHyaluronan has been implicated in several othermorphogenetic movements in embryos (Toole, 1981).In the developing chick cornea, for example, itaccumulates between the external corneal epitheliumand the corneal endothelium, creating a space betweenthese layers that is then invaded by neural crest cells(Toole and Trelstad, 1971; Bard and Abbott, 1979).Similarly, it accumulates beside the neural tube andbeneath the ectoderm of the very early rodent embryo,and creates a space through which neural crest cells willmigrate (Pintar, 1978; Derby and Pintar, 1978; Morriss-Kay and Tuckett, 1986). It has also been suggested thathyaluronan plays some part in bringing about theclosure of the neural tube (Morriss and Solursh, 1978;Schoenwolf and Smith, 1990; Copp and Bernfield, 1988;but see also Morriss-Kay and Tuckett, 1986) and in the

moulding of the nasal folds (Burk, 1985). Its accumu-lation in the palatal shelves is thought to be necessary todrive their elevation to form the secondary palate (Prattet al. 1973; Brinkley and Morris-Wiman, 1987a,b;Ferguson, 1988). In addition, it has a multitude ofeffects on cell differentiation and proliferation (Toole,1990).

There is an especially close parallel between oursystem and the developing heart. The heart originatesas a simple bilayered tube, which becomes subdividedinto separate chambers through the formation of septaand valves. These are derived from the endocardialcushions (Markwald et al. 1981; Van Mierop et al.1963), protrusions of the endothelial inner lining of theheart tube that extend into the lumen, analogous to theaxial protrusions in the inner ear. The formation of thecushions is driven by localised accumulation of anacellular, avascular extracellular matrix, the cardiacjelly, which consists largely of hyaluronan (Markwald etal. 1978). When hyaluronidase is injected into the jelly(Nakamura and Manasek, 1981), the endocardialcushion collapses, just like the protrusions in the innerear.

Hyaluronan-driven morphogenesis is coupled withbasal lamina breakdown and changes of epithelialcohesionThere is a further parallel between our system andseveral of the others where epithelial morphogenesisinvolves hyaluronan accumulation: the movement ofthe epithelium as a coherent sheet is followed by arearrangement of the contacts between the epithelialcells. Thus in the inner ear, when the tips of a pair ofprotrusions meet, they adhere and the epithelium at thesite of contact breaks down and disappears to allowcontinuity between the core of one protrusion and thatof the other. This means that the epithelial cells mustform new junctions at their luminal surfaces with theiropposite numbers on the other protrusion and must losetheir old lateral attachments to their original neigh-bours on the same protrusion. A similar rearrangementof epithelial cell contacts occurs in the heart when twoendocardial cushions meet and fuse to form a septum(Hay, 1978; Kinsella and Fitzharris, 1980), and duringneurulation, when the neural folds come together andfuse (Schoenwolf and Alvarez, 1989), and in thedevelopment of the palate, when the two palatal shelvesmeet and fuse (Fitchett and Hay, 1989). We have shownhere for the axial protrusions in the ear, and others haveshown for the other systems listed, that the epithelialrearrangements are accompanied by a local breakdownof the basal lamina. Moreover, in each of the latterthree systems, some cells are released from theepithelium, become mesenchymal and migrate into thehyaluronan-rich matrix (Markwald et al. 1978; Ber-nanke and Orkin, 1984; Krug et al. 1985; Fitchett andHay, 1989; Pintar, 1978). Thus there seems to be somegeneral correlation between hyaluronan production,dissolution of the basal lamina, and changes inepithelial cohesion.

In our system, one can see a plausible mechanism

Hyaluronan in Xenopus ear morphogenesis 549

that might explain this coupling: if the epithelial cellsare secreting large volumes of hyaluronan at their basalsurface, they can hardly fail to push away or at leastdisrupt the basal lamina on which they sit. Since contactwith the basal lamina is thought to be important inmaintaining the polarity of epithelial cells and coordi-nating their organisation, one would expect a break-down of epithelial structure to be a common result. Ofcourse, other factors such as TGF /3 (Potts and Runyan,1989; Fitzpatrick et al. 1990) and chondroitin sulphate(Morriss-Kay and Tuckett, 1989) may also play a part insuch transformations.

The ear, the heart, the neural tube and crest, and thesecondary palate are at first sight radically differentstructures. Yet in all of them we find a similar pattern ofevents, in which localised hyaluronan accumulationgoes hand in hand with, and in the former two systemsappears to cause, a localised movement of epithelium asa coherent sheet, and this movement is followed by achange of cohesiveness and a rearrangement of theepithelial cell-cell contacts. This hints at the existenceof a common developmental subroutine that involveshyaluronan production and is evoked at many differentsites of epithelial morphogenesis in the embryo.

The example of semicircular canal developmentshows especially clearly how, for such systems, theproblem of explaining the pattern of morphogeneticmovement reduces to a problem of chemical patternformation. To understand how the labyrinth acquires itsstrange shape, we have now to discover what causescells at a restricted set of sites in the epithelium of theotic vesicle to manufacture large quantities of hyaluro-nan.

We are indebted to Paul Martin, Julie Adam, Ian McKayand Gillian Morriss-Kay for discussions, comments, andpractical help. We also thank Jonathan Slack and themembers of his laboratory for Xenopus embryos and BarbaraLuke for help with electron microscopy.

References

ARUFFO, A., STAMENKOVIC, I., MELNICK, M., UNDERHILL, C. B.AND SEED, B. (1990). CD44 is the principal cell surface receptorfor hyaluronate. Cell 61, 1303-1313.

BARD, J. B. L. AND ABBOTT, A. S. (1979). Matrices containingglycosaminoglycans in the developing anterior chambers of chickand Xenopus embryonic eyes. Devi Biol. 68, 472-486.

BERNANKE, D. H. AND ORKIN, R. W. (1984). Hyaluronidaseactivity in embryonic chick heart muscle and cushion tissues andcells. Devi Biol. 106, 351-359.

BRINKLEY, L. L. AND MORRIS-WIMAN, J. (1987a). Computer-assisted analysis of hyaluronate distribution duringmorphogenesis of the mouse secondary palate. Development100, 629-635.

BRINKLEY, L. L. AND MORRIS-WIMAN, J. (1987ft). Effects ofchlorcyclizine-induced glycosaminoglycan alterations on patternsof hyaluronate distribution during morphogenesis of the mousesecondary palate. Development 100, 637-640.

BURK, D. T. (1985). Morphological effects of Streptomyceshyaluronidase treatment on the outgrowth of the nasal processesin mouse embryos. /. craniofac. Genet, dev. Biol. 5, 385-398.

CANNING, D. R. AND STERN, C. D. (1988). Changes in theexpression of the carbohydrate epitope HNK-1 associated withmesoderm induction in the chick embryo. Development 104,643-655.

COMPER, W. D. AND LAURENT, T. C. (1978). Physiological functionof connective tissue polysacchandes. Physiol. Rev. 58, 255-315.

COPP, A. J. AND BERNFIELD, M. (1988). Accumulation of basementmembrane-associated hyaluronate is reduced in the posteriorneuropore region of mutant (curly tail) mouse embryosdeveloping neural tube defects. Devi Biol. 130, 583-590.

DERBY, M. A. AND PINTAR, J. E. (1978). The histochemicalspecificity of Streptomyces hyaluronidase and chondroitinaseABC. Histochem. J. 10, 529-547.

FERGUSON, M. (1988). Palate development, Development 103Supplement, 41-60.

FrrcHETT, J. E. AND HAY, E. D. (1989). Medial edge epitheliumtransforms to mesenchyme after embryonic palatal shelves fuse.Devi Biol. 131, 455-474.

FITZPATRICK, D. R., DENHEZ, F., KONDAIAH, P. AND AKHURST, R.J. (1990). Differential expression of TGF beta isoforms inmurine palatogenesis. Development 109, 585-595.

GODSAVE, S. F., ISAACS, H. V. AND SLACK, J. M. W. (1988).Mesoderm-inducing factors: a small class of molecules.Development 102, 555-566.

HAY, D. A. (1978). Development and fusion of the endocaxdialcushions. Birth Defects: Original Article Series 16, 69-90.

HERTWIG, I. (1987). Morphogenesis of the inner ear of Ranatemporana (Amphibia, Anura). Zoomorph. 107, 103-114.

HUMASON, G. L. (1972). Animal Tissue Techniques, SanFransisco: W. H. Freeman.

KARNOVSKY, M. J. (1965). A formaldehyde-glutaraldehyde fix ofhigh osmolarity for use in electron microscopy. J. Cell Biol. 27,137a-138a.

KINSELLA, M. G. AND FITZHARRIS, T. P. (1980). Origin of cushiontissue in developing chick heart: cinematographic recordings ofin situ formation. Science 207, 1359-1360.

KNOWLTON, V. Y. (1967). Correlation of the development ofmembranous and bony labyrinths, acoustic ganglia, nerves, andbrain centers of the chick embryo. J. Morph. 121, 179-208.

KRAUSE, R. (1903). Entwickelungsgeschichte des Gehororgans. InHandbuch der Entwickelungslehre (ed. O Hertwig), pp. 83-138.Jena.

KRUG, E. L., RUNYAN, R. B. AND MARKWALD, R. R. (1985).Protein extracts from early embryonic hearts initiate cardiacendothelial differentiation. Devi Biol. 112, 414-426.

MARKWALD, R. R., FITZHARRIS, T. P., BANK, H. AND BERNANKE,D. H. (1978). Structural analyses on the matrical organization ofglycosaminoglycans in developing endocardial cushions. DeviBiol. 62, 292-316.

MARKWALD, R. R., KROOK, J. M., KITTEN, G. T. AND RUNYAN, R.B. (1981). Endocardial cushion tissue development: structuralanalyses on the attachment of extracellular matrix to migratingmesenchymal cell surfaces. Scan. Elect. Microsc. 2, 261-274.

MCPHEE, J. R., VAN DE WATER, T. R. AND SU, H. X. (1987).Hyaluronate production by the inner ear during otic capsule andperilymphatic space formation. Am. J. Otolaryngol. 8, 265-272.

MORRISS, G. M. AND SOLURSH, M. (1978). Regional differences inmesenchymal cell morphology and glycosaminoglycans in earlyneural-fold stage rat embryos. /. Embryol. exp. Morph. 46,37-52.

MORRISS-KAY, G. M. AND TUCKETT, F. (1986). The effects ofStreptomyces hyaluronidase on tissue organization and cell cycletime in rat embryos. J. Embryol. exp. Morph. 98, 59-70.

MORRISS-KAY, G. M. AND TUCKETT, F. (1989).Immunohistochemical localisation of chondroitin sulphateproteoglycans and the effects of chondroitinase ABC in 9- to 11-day rat embryos. Development 106, 787-798.

NAKAMURA, A. AND MANASEK, F. J. (1981). An experimental studyinto the relation of cardiac jelly to the shape of the earlyembryonic chick heart. J. Embryol. exp. Morph. 65, 235-256.

NIEUWKOOP, P. J. AND FABER, J. (1967). Normal Table of Xenopuslaevis (Daudin). Amsterdam: North Holland.

NOORDEN, C. VON (1883). Die Entwickelung des Labyrinthes beiKnockenfischen. Arch. Anat und Phys., 235-264.

OHYA, T. AND KANEKO, Y. (1970). Novel hyaluronidase fromStreptomyces. Biochim. biophys. Acta 198, 607-609.

PATERSON, N. F. (1948). The development of the inner ear ofXenopus laevis. Proc. Zool. Soc. Lond. 119, 269-291.

550 C. M. Haddon and J. H. Lewis

PEARSE, A. G. E. (1968). Histochemistry, Theoretical and Applied,vol. 1. (3rd ed.). London; J. A. Churchill Ltd.

PINTAR, J. E. (1978). Distribution and synthesis ofglycosaminoglycans during quail neural crest morphogenesis.Devi Biol. 67, 444-464.

POTTS, J. D. AND RUNYAN, R. B. (1989). Epithelial-mesenchymalcell transformation in the embryonic heart can be mediated, inpart, by transforming growth factor fi. Devi Biol. 134, 392^01.

PRATT, R. M., GOGGINS, J. F., WILK, A. L. AND KING, C. T. G.(1973). Acid mucopolysaccharide synthesis in the secondarypalate of the developing rat at the time of rotation and fusion.Devi Biol. 32, 230-237.

PREHM, P. (1984). Hyaluronate is synthesised at plasmamembranes. Bwchem. J. 220, 597-600.

SCHOENWOLF, G. C. AND ALVAREZ, I. S. (1989). Roles ofneuroepithelial cell rearrangement and division in shaping of theavian neural plate. Development 106, 427-439.

SCHOENWOLF, G. C. AND SMITH, J. L. (1990). Mechanisms ofneurulation: traditional viewpoint and recent advances.Development 109, 243-270.

SLACK, C., WARNER, A. E. AND WARREN, R. L. (1973). Thedistribution of sodium and potassium in amphibian embryosduring early development. J. Physiol., Lond. 232, 297-312.

STREETER, G. L. (1907). On the development of the membranouslabyrinth and the acoustic and facial nerves in the humanembryo. Am. J. Anat. 6, 139-165.

TOOLE, B. P. (1981). Glycosaminoglycans in morphogenesis. In

Cell Biology of Extracellular Matrix (ed. E. D. Hay), pp.259-294. New York: Plenum Press,

TOOLE, B. P. (1990). Hyaluronan and its binding proteins, thehyaladherins. Curr. Opin. Cell. Biol. 2, 839-844.

TOOLE, B. P. AND TRELSTAD, R. L. (1971). Hyaluronateproduction and removal during comeal development in thechick. Devi Biol. 26, 28-35.

TUCKETT, F. AND MORRISS-KAY, G. M. (1988). Alcian blue stainingof glycosaminoglycans in embryonic material: effects of differentfixatives. Histochem. J. 20, 174-182.

UNDERHILL, C. B. (1989). The interaction of hyaluronate with thecell surface: the hyaluronate receptor and the core protein. InThe Biology of Hyaluronan (ed. P. Evered and J. Whelan),Ciba Foundation Symp. 143, 87-106.

VAN MIEROP, L. H. S., ALLEY, R. D., KAUSEL, H. S. ANDSTRANAHAN, A. (1963). The anatomy and embryology ofendocardial cushion defects. J. Thorac. Cardiovasc. Surg. 43,71-83.

WATERMAN, R. E. AND BELL, D. H. (1984). Epithelial fusionduring early semicircular canal formation in the embryoniczebrafish, Brachydanw rerio. Anat. Rec. 210, 101-114.

WILKINSON, D. AND GREEN, J. (1990). In situ hybridisation and thethree-dimensional reconstruction of serial sections. In Post-Implantation Mammalian Embryos: A Practical Approach (ed.A. J. Copp and D. L. Cockroft), pp. 164-169. Oxford: IRLPress at OUP.

{Accepted 21 March 1991)