Cell and Developmental Biology

13
Development 107, 309-319 (1989) Printed in Great Britain © The Company of Biologists Limited 1989 309 Jl/tenascin-related molecules are not responsible for the segmented pattern of neural crest cells or motor axons in the chick embryo CLAUDIO D. STERN 1 , WENDIE E. NORRIS 1 , MARIANNE BRONNER-FRASER 2 , GEOFFREY J. CARLSON 1 , ANDREAS FAISSNER 3 , ROGER J. KEYNES 4 and MELITTA SCHACHNER 3 department of Human Anatomy, South Parks Road, Oxford 0X1 3QX, UK 2 Developmental Biology Center, University of California, Irvine, Ca. 92717, USA ^Department of Neurobiology, University of Heidelberg, 69, Heidelberg, West Germany 4 Department of Anatomy, Downing Street, Cambridge CB2 3 DY, UK Summary It has been suggested that substrate adhesion molecules of the tenascin family may be responsible for the segmented outgrowth of motor axons and neural crest cells during formation of the peripheral nervous system. We have used two monoclonal antibodies (M1B4 and 578) and an antiserum [KAF9(1)] to study the expression of Jl/tenascin-related molecules within the somites of the chick embryo. Neural crest cells were identified with monoclonal antibodies HNK-1 and 20B4. Young somites are surrounded by Jl/tenascin immunoreactive ma- terial, while old sclerotomes are immunoreactive pre- dominantly in their rostral halves, as described by other authors (Tan et al. 1987 - Proc. natn. Acad. Sci. U.S.A. 84, 7977; Mackie et al. 1988 - Development 102, 237). At intermediate stages of development, however, immuno- reactivity is found mainly in the caudal half of each sclerotome. After ablation of the neural crest, the pattern of immunoreactivity is no longer localised to the rostral halves of the older, neural-crest-free sclerotomes. SDS-polyacrylamide gel electrophoresis of affinity-puri- fied somite tissue, extracted using M1B4 antibody, shows a characteristic set of bands, including one of about 230xlO 3 , as described for cytotactin, Jl-200/220 and the monomeric form of tenascin. Affinity-purified somite material obtained from neural-crest-ablated somites reveals some of the bands seen in older control embryos, but the high molecular weight components (120-230X10 3 ) are missing. Young epithelial somites also lack the higher molecular mass components. The neural crest may therefore participate in the expression of Jl/tenascin-related molecules in the chick embryo. These results suggest that these molecules are not directly responsible for the segmented outgrowth of precursors of the peripheral nervous system. Key words: tenascin, cytotactin, Jl, segmentation, somites, neural crest, motor axons, peripheral nervous system, substrate-adhesion molecules (SAMs). Introduction The peripheral nervous system of higher vertebrate embryos becomes segmented because neural crest cells and motor axons emerging from the developing spinal cord are restricted in their migration to the rostral half of each adjacent sclerotome (Keynes and Stern, 1984; reviewed by Keynes and Stern, 1988). The study of the interactions between precursors of the peripheral ner- vous system and somite tissue therefore constitutes an apparently simple model system for the investigation of factors controlling patterning in the developing nervous system. Recently, our laboratories and others have set out to find macromolecules that are differentially ex- pressed in the two halves of the sclerotome, with the aim of identifying those that have a function in control- ling the segmental pattern (Stern et al. 1986; Tosney et al. 1987; Tan et al. 1987; Mackie et al. 1988; Layer et al. 1988; Norris et al. 1989; Halfter et al. 1989). One recent paper (Tan et al. 1987) reported that the glycoprotein cytotactin (main M r = 190-220xlO 3 ; Gru- met et al. 1985; Hoffman et al. 1988) is restricted to the rostral halves of the sclerotomes of the chick embryo, while a cytotactin-binding proteoglycan (CTB-proteo- glycan; Hoffman and Edelman, 1987) is restricted to the caudal halves. Tan etal. (1987) suggested that cytotactin and its proteoglycan ligand 'may control the pattern of neural crest migration'. However, several of their results

Transcript of Cell and Developmental Biology

Page 1: Cell and Developmental Biology

Development 107, 309-319 (1989)Printed in Great Britain © The Company of Biologists Limited 1989

309

Jl/tenascin-related molecules are not responsible for the segmented

pattern of neural crest cells or motor axons in the chick embryo

CLAUDIO D. STERN1, WENDIE E. NORRIS1, MARIANNE BRONNER-FRASER2,

GEOFFREY J. CARLSON1, ANDREAS FAISSNER3, ROGER J. KEYNES4

and MELITTA SCHACHNER3

department of Human Anatomy, South Parks Road, Oxford 0X1 3QX, UK2Developmental Biology Center, University of California, Irvine, Ca. 92717, USA^Department of Neurobiology, University of Heidelberg, 69, Heidelberg, West Germany4 Department of Anatomy, Downing Street, Cambridge CB2 3 DY, UK

Summary

It has been suggested that substrate adhesion moleculesof the tenascin family may be responsible for thesegmented outgrowth of motor axons and neural crestcells during formation of the peripheral nervous system.We have used two monoclonal antibodies (M1B4 and578) and an antiserum [KAF9(1)] to study the expressionof Jl/tenascin-related molecules within the somites ofthe chick embryo. Neural crest cells were identified withmonoclonal antibodies HNK-1 and 20B4. Young somitesare surrounded by Jl/tenascin immunoreactive ma-terial, while old sclerotomes are immunoreactive pre-dominantly in their rostral halves, as described by otherauthors (Tan et al. 1987 - Proc. natn. Acad. Sci. U.S.A.84, 7977; Mackie et al. 1988 - Development 102, 237). Atintermediate stages of development, however, immuno-reactivity is found mainly in the caudal half of eachsclerotome. After ablation of the neural crest, thepattern of immunoreactivity is no longer localised to therostral halves of the older, neural-crest-free sclerotomes.

SDS-polyacrylamide gel electrophoresis of affinity-puri-fied somite tissue, extracted using M1B4 antibody,shows a characteristic set of bands, including one ofabout 230xlO3, as described for cytotactin, Jl-200/220and the monomeric form of tenascin. Affinity-purifiedsomite material obtained from neural-crest-ablatedsomites reveals some of the bands seen in older controlembryos, but the high molecular weight components(120-230X103) are missing. Young epithelial somitesalso lack the higher molecular mass components. Theneural crest may therefore participate in the expressionof Jl/tenascin-related molecules in the chick embryo.These results suggest that these molecules are notdirectly responsible for the segmented outgrowth ofprecursors of the peripheral nervous system.

Key words: tenascin, cytotactin, Jl, segmentation, somites,neural crest, motor axons, peripheral nervous system,substrate-adhesion molecules (SAMs).

Introduction

The peripheral nervous system of higher vertebrateembryos becomes segmented because neural crest cellsand motor axons emerging from the developing spinalcord are restricted in their migration to the rostral halfof each adjacent sclerotome (Keynes and Stern, 1984;reviewed by Keynes and Stern, 1988). The study of theinteractions between precursors of the peripheral ner-vous system and somite tissue therefore constitutes anapparently simple model system for the investigation offactors controlling patterning in the developing nervoussystem. Recently, our laboratories and others have setout to find macromolecules that are differentially ex-

pressed in the two halves of the sclerotome, with theaim of identifying those that have a function in control-ling the segmental pattern (Stern et al. 1986; Tosney etal. 1987; Tan et al. 1987; Mackie et al. 1988; Layer et al.1988; Norris et al. 1989; Halfter et al. 1989).

One recent paper (Tan et al. 1987) reported that theglycoprotein cytotactin (main Mr = 190-220xlO3; Gru-met et al. 1985; Hoffman et al. 1988) is restricted to therostral halves of the sclerotomes of the chick embryo,while a cytotactin-binding proteoglycan (CTB-proteo-glycan; Hoffman and Edelman, 1987) is restricted to thecaudal halves. Tan etal. (1987) suggested that cytotactinand its proteoglycan ligand 'may control the pattern ofneural crest migration'. However, several of their results

Page 2: Cell and Developmental Biology

310 C. D. Stern and others

suggest that the relationship between neural crestmigration and the expression of these macromoleculesis not straightforward. First, cytotactin is not localisedto the rostral halves of the sclerotomes present beforestage 16 (Hamburger and Hamilton, 1951) or so, atwhich time the neural crest has already entered therostral halves of these sclerotomes (Thiery et al. 1982;Rickmann etal. 1985; Bronner-Fraser, 1986; Loring andErickson, 1987; Teillet et al. 1987. cf. Crossin et al.1986); second, paradoxically, cytotactin substrates in-hibit neural crest migration and spreading in vitro;third, it seems curious that while cytotactin is restrictedto the rostral half of the sclerotome, its proteoglycanligand becomes localised to the opposite half after theappearance of cytotactin.

Comparable results were obtained by another groupof workers (Mackie et al. 1988), who investigated thedistribution of tenascin (Chiquet and Fambrough, 1984;Mackie et al. 1987), a hexamer of glycoproteins ofsimilar molecular mass to cytotactin (Vaughan et al.1987), in the sclerotome of the quail embryo. They, too,found that tenascin is restricted to the rostral halves ofthe sclerotomes after stage 16 (Mackie etal. 1988), andthat tenascin substrates inhibit neural crest migrationand spreading in vitro when cultured on basal laminae(Mackie et al. 1988; Halfter et al. 1989).

It seems likely that cytotactin and tenascin, as well astwo other glycoproteins, Jl 200/220 (Kruse et al. 1985;Faissner et al. 1988) and GMEM (glioma extracellularmatrix molecule; Bourdon etal. 1985) are related, if notidentical (see Aufderheide and Ekblom, 1988; Faissneret al. 1988; Friedlander et al. 1988; Hoffman et al. 1988;Mackie et al. 1988; Halfter et al. 1989). Their relativemolecular masses appear to be the same; all of themcomprise several polypeptides of various relative mol-ecular masses between 110 and 260 X103, with a pre-dominant band at Mr = 220xl03, and both cytotactinand tenascin are assembled as a hexamer, called a'hexabrachiori (Friedlander et al. 1988). Tenascin andJl have been reported to be related immunologically,and cytotactin and tenascin show apparently identicalpatterns of immunohistological localisation in varioustissues (see Mackie et al. 1988), as well as identicaleffects on neural crest cells in vitro. At least whenisolated from brain, cytotactin and its proteoglycanligand can carry the HNK-1/L2 carbohydrate epitope(Hoffman and Edelman, 1987; Kunemund et al. 1988;Friedlander et al. 1988; Hoffman et al. 1988), as does Jl(Kruse et al. 1985; Faissner et al. 1988).

In this study, we have used three different antibodies,two monoclonals (M1B4 - Chiquet and Fambrough,1984 and MAb578 - Faissner et al. 1988 and in prep-aration) and one antiserum (KAF9(1); Faissner et al.1988 and in preparation), to study the distribution ofJl/tenascin-related molecules in the sclerotome of thechick embryo in more detail. We performed neuralcrest ablations to examine the relationship between theposition of neural crest cells and the expression of thesemolecules, and constructed an M1B4 affinity column topurify the antigens for subsequent identification bySDS-polyacrylamide gel electrophoresis.

Materials and methods

Embryos and operationsFertile hens' eggs (Rhode Island Red hybrids) were incubatedat 38 °C for 2-3 days until they had reached Hamburger andHamilton (1951) stages 10-20. Embryos destined for ablationof the neural crest were treated as follows: at about stage10-12, a l-5xl-5cm window was cut in the egg shell with ascalpel, the embryo floated up to the level of the shell withcalcium- and magnesium-free Tyrode's saline (CMF), andabout 50^1 of a 1:10 solution of Indian ink (Pelikan FountIndia) in CMF injected into the subblastodermic cavity, toimprove contrast between the embryo and the underlyingyolk. Removal of the neural crest was then performed, undera drop of CMF, using a microsurgical knife (Week, 15° angle),Dewecker's scissors and a steel needle made from a N°lentomological pin. The dorsal half of the neural tube spanningthe region between the caudal end of the segmental plate andthe 4th most recently formed somite (corresponding to some14 segments in length) was excised. The ventral half of thetube and the underlying notochord were left in place. Afterthe operation, the CMF was removed from above the embryo,and about 2 ml thin albumen withdrawn with a syringe from ahole in the blunt end of the egg; this lowered the embryo backinto the shell. The shell was then sealed with PVC tape andthe egg incubated at 38°C for a further 36-48 h, by which timethe embryos had reached stages 17-22. Three additionalembryos were incubated for a longer period, until they hadreached stages 25-27.

Control embryos were treated in an identical manner: thevitelline membrane was cut above the embryo and, in mostcases, deep cuts were made on either side of the neural tube.After these manipulations, the embryos were lowered into theshell, the eggs sealed and incubated exactly as for operatedembryos. All of the 107 operated embryos and 116 controlssurvived the subsequent period of incubation.

ImmunohistochemistryControl and operated embryos destined for immunohisto-chemistry were fixed in 100% ethanol for lh or in bufferedformol saline (4 % formaldehyde in phosphate-buffered saline[PBS]) for 30 min, and washed well with PBS. They were thenplaced in a 5 % solution of sucrose in PBS for 1-3 h, then in20% sucrose/PBS for 8-24h, and finally infiltrated for 2-6hat 38°C with a solution containing 7-5% gelatin (Sigma, 300Bloom) and 15 % sucrose in PBS. They were then brought toroom temperature, to allow the blocks to set. These blockswere stored at 4°C for no longer than a week prior to serialsectioning (5-10pun) at -20°C in a Bright cryostat. Thesections were collected on gelatinised glass slides, air driedand stored at 4°C until required.

Prior to immunohistochemical staining, the sections wererehydrated and the gelatin removed by placing them in PBS at38°C for a few minutes, then washed extensively with PBS.They were then blocked with 3% bovine serum albumin(BSA, Sigma) for 20 min, placed in the appropriate antibody(see below) for lh at room temperature, washed extensivelywith PBS, placed in the appropriate labelled secondaryantibody for 1 h at room temperature, washed again, and thenprocessed for immunofluorescence or immunoperoxidase.

For immunofluorescence, the slides were mounted in anonquenching medium (14% Gelvatol 20/30 [Fisons] con-taining 8'5ragmr' diazobicyclo-octane [DABCO, Aldrich,as anti-quenching agent], 30% glycerol and 350/.tgml"1

sodium azide as preservative in PBS, pH6-8), and viewed andphotographed under epifluorescence optics in an Olympus

Page 3: Cell and Developmental Biology

JlItenascin in neural crest and somites 311

Vanox-T microscope, using supplementary exciter and barrierfilters to minimise spill-over between fluorescein and rhoda-mine wavelengths. The supplementary filter combinationsused were: excitation EY455/emission B460 for fluorescein(blue excitation, green pass) and excitation EO530/emissionO590 for rhodamine (green excitation, red pass). Lack ofspill-over may be confirmed by the reader by viewing thecolour figures with red/green glasses.

For immunoperoxidase, the slides were washed briefly inO-lM-Tris-HCl and then incubated in a 500jUgml~' solutionof diaminobenzidine (Aldrich), to which H2C>2 was added to afinal concentration of 0-3 %. The slides were incubated in thisfor about lOmin at room temperature, then washed inrunning tap water for 30min, rinsed in distilled water andmounted in Aquamount medium (BDH).

AntibodiesNeural crest cells present within the sclerotome of chick andquail embryos display the HNK-1/L2 epitope, and thereforemonoclonal antibodies specific for this epitope (such as NC-1or HNK-1; Tucker et al. 1984) are often used to detect thepresence of neural crest cells in histological sections. Bronner-Fraser (1986) and Teillet et al. (1987) have investigated theexpression of the HNK-1 epitope in neural crest cells in thechick embryo in some detail, conclude that almost all the crestcells that traverse the sclerotome express HNK-1 immuno-reactivity (unlike some crest cells in the head and some thattake the dorsolateral pathway between the ectoderm and thesomite); HNK-1 is therefore a good marker for neural crestcells present in the sclerotome at this stage of development.Since Jl/tenascin-related molecules can carry the HNK-1/L2epitope, we have used antibody 20B4 (which recognises adifferent, as yet uncharacterised, epitope expressed by neuralcrest cells; Dr J. Denburg, unpublished data) in addition toHNK-1 to detect neural crest cells and to control for thepossibility of cross-reaction between the HNK-1 antibody andJl/tenascin-related molecules.

The primary antibodies used (dilutions in 0-3 % BSA/PBS)were:M1B4 supernatant (Mouse igG, originally raised by DrDouglas Fambrough and obtained from the DevelopmentalStudies Hybridoma Bank). This was used at concentrationsbetween 1 and lO^gm!" (between 1:10 dilution and undi-luted). The antibody recognises the original 'myotendinousantigen' (= tenascin = hexabrachion) described by Chiquetand Fambrough (1984; see also Friedlander et al. 1988) and anepitope near the carboxy-terminal end of chain I of thecytotactin molecule (Friedlander et al. 1988).578 purified ascites (Rat IgG monoclonal against Jl 200/220;Faissner et al. 1988 and in preparation). Used at 1:100

1(gKAF9(J) (IgG fraction of rabbit antiserum directed againstmouse Jl 200/220-tenascin; Faissner et al. 1988 and inpreparation). Used at 1:200 (25/*gml"').HNK-1 supernatant (Mouse IgM, originally raised by Abo andBalch, 1981). Used 1:2 or undiluted (about 50/igml"1). Theneural crest specificity of this antibody is discussed by variousauthors (e.g. Tucker et al. 1984; Rickmann et al. 1985;Bronner-Fraser, 1986; Teillet et al. 1987; Canning and Stern,1988).20B4 supernatant (Mouse IgG, originally raised by Dr JeffDenburg, to chick dorsal root ganglion cells and obtainedfrom the Developmental Studies Hybridoma Bank). Usedundiluted (20^gmr').

In double-immunofluorescence studies, it was important touse secondary antibodies that did not cross-react across

species or immunoglobulin types. After extensive tests toascertain this, the following antibodies were selected:For mouse IgG: (1) affinity-purified tetramethylrhodamineisothiocyanate (TRITC)-labelled goat anti-mouse IgG (Fcfragment specific; Sigma), used at 1:50 or (2) IgG fraction ofaffinity-purified fluorescein isothiocyanate (FITC)-labelledgoat anti-mouse IgG (Fc fragment, y-chain specific; Cappel).Used at 1:50 or 1:100.For mouse IgM: (1) affinity-purified FITC-labelled goat anti-mouse IgM (ju-chain specific; Sigma), used at 1:100, or (2)TRITC-labelled sheep anti-mouse IgM (Fc fragment specific;Serotec), used at 1:40.For rat IgG: affinity-purified TRITC- or FITC-labelled goatanti-rat IgG (Fc fragment specific, tested for the absence ofcross-reactivity to mouse IgG by the manufacturers; Cappel1213-1721). We confirmed the absence of cross-reactivitybetween this antibody and mouse IgG ourselves in sectionsthat had been incubated with mouse monoclonals M1B4 or20B4. Used at a dilution of 1:100 (30^gml~').For rabbit IgG: affinity-purified TRITC- or FITC-labelledgoat anti-rabbit IgG (Fey chain specific; Cappel), used at 1:50(20/igmr1).

To absorb out any remaining nonspecific binding, allsecondary antibodies were made up in 0-3% BSA in PBScontaining 1-5% normal goat serum, and the final workingsolution absorbed against either fresh or formalin-fixed 2-daychick embryo tissues for 30min under gentle agitation at roomtemperature; after this, the working solution was centrifugedfor 3min at lOOg in a Micro-Centaur centrifuge. Theseprocedures were necessary to remove components of thesecondary antisera that were found to bind to various chicktissues. In double-immunofluorescence experiments, the twoprimary antibodies were added together in 0-3 % BSA/PBS;the two secondary antibodies were also used in a singlesolution.

In each experiment, the entire control or neural-crest-ablated embryo (i.e. all the serial sections) was stained withone combination of antibodies. At least two control and twooperated embryos were studied in this way for each antibodycombination. Table 1 shows the combinations of primaryantibodies used in single- and double-immunofluorescenceexperiments.

Control sections were included in each experiment; inthese, incubation in 3 % BSA/PBS replaced the incubation inprimary antibody.

Electrophoretic characterisation of antigensWe were not able to detect antigens recognised by antibodiesM1B4, 578 or KAF9(1) in immunoblots made from polyacryl-amide gels of unfractionated somite tissue under eitherreducing or nonreducing conditions. For this reason, we hadto resort to affinity-extraction of the antigens from thesetissues. 45 /ig of M1B4 IgG were purified by affinity fromculture supernatant using goat anti-mouse IgG agarose(Sigma), following the manufacturer's protocol. The boundIgG was eluted using 0-1 M-glycine and 045 M-NaCl (pH2-4),

Table 1. Combinations of antibodies used forimmunofluorescence studies

M1B4578KAF9(1)HNK-120B4Total = 59

M1B4

423

18-

578

24252

KAF9(1)

32452

HNK-1

1855-4

20B4

_2244

Page 4: Cell and Developmental Biology

312 C. D. Stern and others

and the fractions collected were neutralised with a smallvolume of lM-Trizma base (Sigma). This purified IgG wascoupled to 0-5 ml CNBr-activated Sepharose-4B (Sigma)following the manufacturer's instructions.

Strips of somite tissue (with some adherent ectoderm andendoderm) were dissected from unoperated embryos or fromembryos from which the neural crest had been ablated36-48 h earlier (as described above). Equivalent regions weretaken from both operated and unoperated embryos, ignoringthe first and last three pairs of somites opposite the operatedregion; both control and operated embryos were at stages18-22. Two types of sample were dissected from youngerembryos. In one, the last four pairs of somites plus someadherent tissues were dissected from embryos at stages 12-14and treated identically to the other samples. In the other, thesame region was dissected but the neural tube and notochordwere included in the sample. In all cases, tissues werecollected in PBS containing a cocktail of protease inhibi-tors: lmM-phenymethylsulphonylfluoride (PMSF; Sigma),lmM-N-ethyl maleimide (NEM; Sigma), lO^gml"1 soybeantrypsin/chymotrypsin inhibitor (STI; Sigma), lmM-EDTAand 1 mM-EGTA, and quickly frozen on solid CO2.

The frozen tissue samples were thawed and homogenised in200 ,ul solubilization buffer, made up of 0-25 % sodium deoxy-cholate, lOmM-Tris, 150mM-NaCl (pH8-3) containing thecocktail of protease inhibitors described above. They werethen centrifuged at 11600g for 5min and the supernatantscollected for loading onto the affinity column. Samples wereincubated in the column for 45 min at room temperature, afterwhich the column was washed with 15 column volumes ofsolubilisation buffer containing 300mM-NaCl. Elution wasthen performed using a buffer containing 0-1 M-diethylamine,150mM-NaCl and 0-25% sodium deoxycholate (pHll-5).100 iA fractions were collected and a protein assay performedas a dot blot on Immobilon membrane (Millipore) usingAuroDye (Janssen). Protein-containing fractions were pooledand concentrated by pressure dialysis against 0-1% deoxy-cholate, 20mM-Tris, 150mM-NaCl, 1 mM-EGTA and lmM-EDTA(pH8-3).

Approximately equal amounts of protein were loaded pertrack of 4-15% SDS-polyacrylamide gradient minigels(9 cm x 9 cm x 0-5 mm) using the Laemmli (1970) buffer sys-tem in an apparatus designed as described by Matsudaira andBurgess (1978). Electrophoresis was performed at 150 V for90 min, or until the tracking dye was 1-5 cm above the bottomof the gel. The resulting gels were silver-stained as describedpreviously (Canning and Stern, 1988; Norris et al. 1989).

14C-labelled, colour-coded molecular weight standards(Amersham, Rainbow Markers) were run in one or two tracksin each gel. A track containing a sample of purified mouseJl/tenascin was also included in each gel for comparison. Thispreparation of Jl/tenascin was prepared by affinity purifi-cation as described previously (Faissner et al. 1988), usingpostnatal mouse brain tissue as starting material. The elutedfractions were dialysed against water and lyophilised.

Immunoblots of purified mouse brain tenascin were alsomade. Each blot included radioactive relative molecular massmarkers and samples of purified Jl/tenascin, under reducedand non-reduced conditions. Transfer to Immobilon mem-branes was performed as described previously (Primmett et al.1989). Each membrane was probed with one of the antibodiesused in this study, namely M1B4, 578 or KAF9(1), anddetected with either rabbit anti-rat followed by gold-labelledgoat anti-rabbit Ig for MAb578, or gold-labelled goat anti-rabbit Ig for KAF9(1), or gold-labelled goat anti-mouse IgGfor M1B4. All gold-labelled antibodies were purchased fromJanssen (AuroProbe system).

Results

ImmunohistochemistryIdentification of neural crest cells with antibodies20B4 and HNK-1

Monoclonal antibody 20B4 is reported to be specific fordorsal root ganglion cells and migrating neural crestcells; it also detects an antigenic determinant in theectoderm of chick embryos (Dr J. Denburg; data fromsheet obtained with supernatant supplied from theDevelopmental Studies Hybridoma Bank). In ourhands, 20B4 was not as useful as HNK-1 in revealing thedistribution of neural crest cells in the sclerotomebecause 20B4 staining was always of lower intensity andit only recognised a subset of cells stained by HNK-1.Nevertheless, the overall pattern of staining seen withthis antibody in embryo sections was similar to that seenwith HNK-1.

Jl I tenascin-immunoreactivityMost of the immunocytochemical procedures involveddouble staining, using one antibody directed to J l /tenascin-related molecules and one directed to neuralcrest cells. The combinations of antibodies used in theseexperiments are summarised in Table 1. Identical re-sults were obtained using all three antibodies directedagainst Jl/tenascin [M1B4, 578 and KAF9(1)]. Thiswas confirmed in double-staining experiments combin-ing the three antibodies, one pair at a time: in everycase, the staining patterns in the rhodamine and fluor-escein wavelengths were identical.

In all somites of the youngest embryos (stage-13 orearlier) and in the most caudal 2-3 pairs of (epithelial)somites of older embryos (stage 16-22), staining wasrestricted to the extracellular matrix surrounding eachsomite and to the matrix associated with mesenchymalcells within the lumen of the epithelial somites, asdescribed by Crossin et al. (1986), Tan et al. (1987) andMackie et al. (1988). Also in accordance with Tan et al.(1987) and Mackie et al. (1988), staining within thesclerotomes of the most cranial (older) somites ofembryos older than stage 16 was restricted to the rostralhalf of each sclerotome; most of the cells expressingJl/tenascin immunoreactivity were not HNK-1- or20B4-positive (Fig. 1 E, F). In addition, Jl/tenascin-immunoreactive filamentous material was seen close tothe border between adjacent sclerotomes; this materialappeared to be associated with the intersegmentalvessels (Fig. 1 B). Immunoreactive extracellular ma-terial was also seen between the dermomyotome andthe ectoderm at all segmental levels. Strong stainingwas also seen between adjacent dermomyotomes, es-pecially at the caudal margin of each (Fig. 1 A).

The pattern of staining seen in younger sclerotomes(the 4-6 most caudal sclerotomes, between 4 and 10segments rostral to the segmental plate) of embryosbeyond stage 13 was different from that described byTan et al. (1987) and by Mackie et al. (1988). In theseyoung somites, which have already begun to be col-onised by migrating neural crest cells (Rickmann et al.

Page 5: Cell and Developmental Biology

Fig. 1. Changes in the expression of Jl/tenascin-related molecules during development of the sclerotome. (A) Coronalsection through the two most caudal sclerotomes of a stage-14 embryo, stained with M1B4 (red) and HNK-1 (green). Notethe intensely HNK-1-positive perinotochordal matrix, and the HNK-1-positive neural crest cells in the rostral (left) half ofthe sclerotomes. MlB4-positive material, on the other hand, is more intense in the caudal than the rostral halves.(B) Sagittal section through two caudal sclerotomes of a stage-15 embryo, stained with HNK-1 (red) and KAF9(1) (green).More intense KAF9(1) staining is seen in the caudal (right) halves of the sclerotomes, while many HNK-1-positive neuralcrest cells have already entered into the rostral halves. KAF9(1) staining is also seen surrounding each somite and near theintersegmental vessels. (C) Oblique transverse section through a stage-15 embryo, stained with KAF9(1) (green) and HNK-1(red), about 9 segments cranial to the segmental plate. The section passes through the rostral half of the sclerotome on theleft and through the adjacent caudal half on the right. Note the abundance of KAF9(l)-positive material between thedermomyotome and the ectoderm, between the neural tube and the sclerotome, and within both halves of the sclerotome.KAF9(l)-positive material is more abundant in the caudal half at this level. HNK-1-positive cells have entered deep into therostral half of the sclerotome. (D) Coronal section through two more rostral sclerotomes (about 12 segments rostral to thesegmental plate) of a stage-15 embryo, stained with HNK-1 (red) and MAb578 (green). MAb578-immunoreactivity begins topredominate in the rostral half. (E) Coronal section through three more rostral sclerotomes (about 15 segments rostral tothe plate) of a stage-17 embryo, stained with M1B4 (red) and HNK-1 (green). Within the sclerotome, M1B4immunoreactivity is now confined to the rostral (left) half, except in the most ventromedial portion of the sclerotome,adjacent to the notochord. (F) Higher power view of one of the sclerotomes shown in (E). HNK-1 (green) and M1B4 (red)immunoreactivity do not coincide except in very few areas. In each figure, the head of the embryo lies towards the left ofthe photograph. Scale bars: 50^m (A-E), 25^m (F). d, dermomyotome; m, myotome; n, notochord; s, sclerotome; t,neural tube; v, intersegmental vessel.

Page 6: Cell and Developmental Biology

Fig. 2. Neural crest ablation and expression of Jl/tenascin-related molecules. (A) Low-power view of an oblique coronalsection through a stage-15 embryo after ablation of six segments worth of dorsal neural tube at stage 11, stained with HNK-1and TRITC-labelled secondary antibody. Note the presence of neural crest cells in most of the sclerotomes in the operatedregion. (B) Phase-contrast micrograph of a sagittal section through two neural-crest-free sclerotomes of a stage-16 neural-crest-ablated embryo. Note that the rostral (left) halves display a reduction in cell number as compared with thecorresponding caudal halves. (C) Same section as in B above, stained with M1B4 (red) and HNK-1 (green). Note theabsence of HNK-1-positive cells, and the predominance of M1B4 immunoreactivity in the caudal half-sclerotome.(D) Sagittal section through neural crest-free older sclerotomes (about 12 segments rostral to the segmental plate) of a stage-17 embryo, stained with HNK-1 (red) and KAF9(1). Note the predominance of KAF9(1) immunoreactivity in the caudalhalves of the sclerotomes. [Compare the pattern of staining with that shown in Fig. 1E,F, which passes through the samelevel of a similarly-staged but unoperated embryo]. (E) Sagittal section through neural-crest-free sclerotomes of a stage-17embryo stained with HNK-1 (red) and KAF9(1) (green). In this embryo, both halves of the neural-crest-free sclerotomesdisplay KAF9(1) immunoreactivity. Scale bars: 100/im in A 50 ,um in B-E. The head of the embryo lies towards the left ofeach photograph, a, aorta; d, dermomyotome; n, notochord; t, neural tube.

Page 7: Cell and Developmental Biology

Jl/tenascin in neural crest and somites 313

1985; Bronner-Fraser, 1986; Teillet et al. 1987), thecaudal half of each sclerotome is more intensely stainedby the antibodies than the rostral half (Fig. 1 A-C). Inabout half the cases, moreover, staining was restrictedto the caudal half. The relationship between the distri-bution of migrating neural crest cells (as revealed byantibodies HNK-1 and 20B4) and that of Jl/tenascinimmunoreactivity was revealed in double-immunoflu-orescence experiments (Fig. 1): at this stage in thedevelopment of the somite, when neural crest cells arestarting to invade the sclerotome, Jl/tenascin immuno-reactivity predominates in the opposite half of thesclerotome to that occupied by the neural crest cells(Fig. 1 A-C). Jl/tenascin immunoreactive material inthe sclerotome was most abundant dorsally near thedermomyotome, where it had a filamentous appearance(Fig. 1 A, B), and ventrally in the vicinity of thenotochord, where it was associated with the peri-notochordal sheath. It was also associated with thebasal lamina surrounding the neural tube (Fig. 1 C).

The 4-5 sclerotomes rostral to the 4-6 most recentlyformed (situated between the 10th and 15th segmentrostral to the segmental plate) displayed an intermedi-ate pattern of staining: both halves of these sclerotomesshowed approximately equal levels of immunoreactivity(Fig. 1 D). The region of the sclerotome that liesadjacent to the notochord is immunoreactive for J l /tenascin along the entire rostrocaudal axis (i.e. in bothhalves of the sclerotome) rostral to the 10th segmentbeyond the segmental plate (e.g. Fig. 1 E).

Neural crest removalIn a preliminary series of experiments, a strip of dorsalneural tube, 3-15 segments in length, was removedfrom a region opposite the caudal paraxial mesoderm.The operated embryos were then incubated for 36-48 h,fixed, sectioned and stained with monoclonal antibodyHNK-1 to visualise the distribution of the neural crestcells and to assess the effectiveness of the operation. Inagreement with earlier investigations (Yntema andHammond, 1954; Rickmann et al. 1985; Teillet et al.1987), we found that neural crest cells are capable ofmigrating along the rostrocaudal axis for a distance ofthree segments in each direction (e.g. Fig. 2 A). There-fore, the sclerotomes in the operated region of embryosin which the neural crest removal spanned six segmentsor less always became colonised by HNK-1-immuno-reactive cells.

In another series of preliminary experiments, alength equivalent to 3-15 segments of the entire dorso-ventral extent of the neural tube was removed, and thedistribution of HNK-1-positive cells examined in sec-tions after 36-48h incubation. In these embryos, theHNK-1-positive cells did not migrate rostrocaudallyfrom the edges of the operation site. However, theintegrity of the somites opposite the operated regionwas affected: the somites were always fused in themidline and the segmental arrangement of somiticmesoderm was often disrupted.

For these reasons, all our neural crest ablationexperiments were performed on embryos from which

the dorsal half of the neural tube was excised over10-15 segments. In these embryos, after 36-48 h post-operative incubation, the somites appeared normal andat least five sclerotomes opposite the operated regionwere always found to be devoid of HNK-1-immuno-reactive cells.

The distribution of M1B4, MAb578 or KAF9(1)staining in sclerotomes devoid of neural crest cells atstages 18-22 was rather different from that seen inunoperated sclerotomes, and different from the patternreported by Tan et al. (1987). Young sclerotomes in theoperated region displayed immunoreactivity predomi-nantly in the caudal half of the sclerotome (Fig. 2 C) (asin unoperated embryos). However, in most embryos(39/56; 70%), this pattern was unchanged in oldersclerotomes opposite the operated region (Fig. 2 D). Insome embryos (13/56; 23 %), immunoreactive materialwas seen in both halves of the sclerotomes (Fig. 2 E), inthree embryos (5%) the sclerotomes appeared to bedevoid of immunoreactivity in the operated region andin one embryo staining was restricted to the rostralhalves of sclerotomes devoid of HNK-1-immuno-reactive cells.

To investigate whether the difference between oper-ated and control embryos in the pattern of Jl/tenascinimmunoreactivity is due to a contribution of the neuralcrest cells or to a developmental delay of the operatedembryos, three embryos from which the neural cresthad been removed as described above were incubatedfor a total of 3-5 days after the operation, until they hadreached stages 25-27. The pattern of Jl/tenascin immu-noreactivity in these embryos was similar to that seen inoperated embryos incubated to stages 18-22: in everycase, Jl/tenascin immunoreactive material was foundin both halves of the neural-crest-free sclerotomes.

Biochemical characterisation of Jl/'tenascin-relatedantigensPurification of Jl/tenascin-related antigens was per-formed using a column of immobilised M1B4 IgG,followed by SDS-polyacrylamide gel electrophoresis(SDS-PAGE) to identify the eluted antigens undernonreducing conditions.

In preliminary experiments, the specificity of thethree antibodies (M1B4, KAF9(1) and MAb578) forJl/tenascin-related molecules was assessed in immuno-blots of a sample of purified mouse brain tenascinseparated by SDS-PAGE under nonreducing con-ditions. Silver-stained gels obtained from this sampleshowed a characteristic pattern of bands, summarised inTable 2. In immunoblots, both KAF9(1) and MAb578recognised many of the bands seen in silver-stainedgels, of which a band of Mr = 230xl03 was mostprominent (Fig. 3). The avian-specific antibody M1B4,by contrast, did not recognise any components in thismouse tenascin preparation.

In samples of somite tissue obtained from unoperatedembryos at stage 18-22, several bands were seen insilver-stained gels of eluted material. Their Mr rangedfrom 39 to 230xl03 (Fig. 4). Table 2 summarises therelative molecular masses of the polypeptides seen in

Page 8: Cell and Developmental Biology

314 C. D. Stern and others

Table 2. Apparent relative molecular masses of Jl/tenascin-related molecules (Mr x JO 3)

Mouse braintenascin

>300230

185

140

705g

•47

39

stage-18unoperated

230195

150140130

120115

10094/96

7870585249

4539

stage-18— crest

105/110

70585249

39

st.13 epith.somites+ NT & notochord

190

140130125120

105/110

94/967870585249

39

st. 13 epith.somites- NT & notochord

160

947870585249

39

these gels. The highest relative molecular mass com-ponent seen appeared to be similar to the majorcomponent (c. 230xl03) found in silver-stained gels andimmunoblots of samples of purified mouse tenascin(Figs 3,4). The sample of purified tenascin contained ahigher relative molecular mass component which mayrepresent a multimer (hexamer, Mr approximately

MrX 10

200-

116 -

66 -

4 5 - —,

12 3 4 5 6Fig. 3. Western blots of mouse brain tenascin showingspecificity of antibodies used for Jl/tenascin.Immunoreactivity was detected using the AuroProbesystem. Odd-numbered lanes: non-reduced mouse braintenascin. Even-numbered lanes: reduced mouse braintenascin. Lanes 1 and 2 were probed with KAF9(1). Lanes3 and 4 were probed with MAb578. Lanes 5 and 6 wereprobed with M1B4 (which is specific for avian tenascin).

106 = 'hexabrachion'; Chiquet and Fambrough, 1985;Mackie et al. 1988), but this was not seen in samplesobtained from chick embryo somites.

Samples obtained from neural-crest-ablated embryosdisplayed a different pattern of silver-stained bandsafter elution from the M1B4 column (Table 2). The sixbands of highest relative molecular mass (120, 130,140,150, 195 and 230X103) seen in unoperated embryoswere missing. Some lower relative molecular masscomponents were also absent. Some bands, however,were present in samples from both unoperated andneural-crest-ablated embryos (A/r 39, 49, 52, 58 and70X103).

Antigens were also detected in samples of epithelialsomites from stage-13 to -14 embryos subjected toaffinity purification through the M1B4 column; somedifferences were found between those samples thatincluded the neural tube and notochord and those thatdid not. Gels are shown in Fig. 4 and the molecularmasses of the bands summarised in Table 2. The patternof bands seen in this material was different from thatseen in similarly treated sclerotome from unoperatedembryos. The most conspicuous difference was theabsence of the 230X103 component, as seen in neural-crest-ablated embryos. However, other componentsabsent from neural-crest-ablated sclerotomes were seenin the epithelial somite preparations (Table 2).

Discussion

We have used two monoclonal antibodies (M1B4 and578) and one antiserum [KAF9(1)] to study the ex-pression of Jl/tenascin-related molecules in somitetissue of the chick embryo. Young somites are sur-rounded by immunoreactive material, while old sclero-tomes display expression of these molecules predomi-

Page 9: Cell and Developmental Biology

Jl/tenascin in neural crest and somites 315

Mrx1CT3

200-*—'

I'• • .-4

116 -

92-5 -

66-H

45* *

31 -

IFig. 4. Silver-stained gels of somitematerial purified by affinity to M1B4antibody. Lane 1, relative molecularmass markers. Lane 2, mouse braintenascin (nonreduced; not affinitypurified through M1B4 column).Lane 3, sclerotomes from unoperatedstage-18 embryos. Lane 4,sclerotomes from neural-crest-ablatedstage-18 embryos. Lane 5, stage-13epithelial somites with neural tubeand notochord. Lane 6, stage-13epithelial somites without other axialtissues. The molecular masses of thebands seen in these samples are givenin Table 2.

nantly in their rostral halves, as previously described byother authors (Tan et al. 1987; Mackie et al. 1988).Sclerotomes at intermediate stages of development,however, often display immunoreactivity in the op-posite (caudal) half of the sclerotome to that occupiedby neural crest cells (Fig. 5). Our results thereforediffer from those of Tan et al. (1987), who stated that 'atthe time of sclerotome formation, cytotactin began to beexpressed in the rostral half of the sclerotome, and neuralcrest cell invasion occurred in this area\ and from thoseof Mackie et al. (1988). Also in contrast to Tan et al.'s(1987) observations, we find that after neural crestablation, immunoreactivity is not restricted to therostral half of the sclerotome. SDS-PAGE of affinity-purified somite tissue reveals a characteristic pattern ofbands of various relative molecular masses, includingtwo bands at about 195 and 230xl03, as described forcytotactin, Jl-200/220 and the monomeric form oftenascin (Chiquet and Fambrough, 1984; Grumet et al.1985; Kruse et al. 1985; Vaughan et al. 1987; Faissner etal. 1988; Hoffman et al. 1988). After ablation of theneural crest, somite material analysed in the same waystill shows some of the bands seen in unoperatedembryos, but several components are missing, includingthe six of highest relative molecular mass(120-230 xlO3).

Ablation of the neural crestIn agreement with other workers (e.g. Yntema andHammond, 1954; Rickmann et al. 1985; Teillet et al.1987), we have found that ablation of a short length (sixsegments or less) of the dorsal half of the neural tubeallows neural crest cells from the cut ends of the tube tocolonise the rostral halves of the sclerotomes in theoperated region. Under these conditions, neural crestcells appear to be able to migrate for about threesegments along the rostrocaudal axis. If the wholethickness of the neural tube is ablated, moreover, theintegrity of the adjacent somites is disrupted, asreported by Strudel (1955) and Teillet and Le Douarin(1983). In order to obtain neural-crest-free and mor-phologically normal sclerotomes, therefore, it wasnecessary to extirpate a length of dorsal neural tube atleast seven segments in length.

In Tan et al.'s (1987) experiments, the neural crestwas ablated by removing a six-segment length of anunspecified depth of neural tube. In our hands, if onlythe dorsal portion of the tube was removed, such anoperation would have allowed neural crest cells tocolonise all the sclerotomes in the operated region. If,on the other hand, the whole depth of neural tube wasremoved, the operation would have caused the somitesto display abnormal segmentation. Moreover, Tan et al.

Page 10: Cell and Developmental Biology

316 C. D. Stern and others

©HNK-1 staining

M1-B4 staining

unoperated

neural crest removed

Ohsomiteformation

10 hsclerotomeformation; crestenters somite

6 somites(epithelial)

12-15 segments

36 -48 hmotor axonsenter somite

Fig. 5. Diagram summarising the changes in the distribution of Jl/tenascin immunoreactivity in relation to the onset ofneural crest migration and motor axon outgrowth through the sclerotome in the presence and absence of neural crest cells.Each of the two embryos is depicted at about stage 16, with its caudal end towards the left of the picture, so that youngersomites appear towards this end. The upper diagram shows the distribution of neural crest cells and of Jl/tenascinimmunoreactivity in an unoperated embryo, while the lower cartoon illustrates the pattern of Jl/tenascin after neural crestremoval. In this lower diagram, the results are presented as if the neural crest were absent over the entire extent of theembryonic axis. The line below the diagrams gives an approximate indication of the time scale of the changes that occur andpoints to the level at which the motor axons are starting to enter the sclerotome.

(1987) ablated the neural tube opposite the last six pairsof segments of 15- to 18-somite embryos; at this stage,the most rostral one or two of these pairs alreadycontain neural crest cells (Rickmann et al. 1985).

Tan et al. (1987) argue that, after ablation of theneural crest, cytotactin expression in the sclerotomes issimilar to that seen in unoperated embryos, that is,restricted to the rostral half of each sclerotome. Ourresults do not agree with this interpretation. While thepattern of expression of Jl/tenascin-related moleculesis still segmental after ablation of the neural crest,immunoreactivity fails to become restricted to therostral halves of the sclerotomes in the absence ofneural crest cells.

Molecular nature of Jl/tenascin-related antigensRemoval of the neural crest leads to a change in thedistribution of Jl/tenascin immunoreactivity and in theproportions of the various molecules extracted byaffinity to M1B4 antibody. In the absence of neuralcrest cells, the higher relative molecular mass bands(120-230X103) characteristic of cytotactin, Jl-160/180,Jl-200/220 and monomeric tenascin are absent, andlower relative molecular mass components predomi-nate. Epithelial somites from younger embryos do notdisplay the higher molecular mass (230 xlO3) band.Many of the lower relative molecular mass components

are present in a purified sample of tenascin preparedfrom mouse brain.

Several possibilities must be considered in attemptingto identify the various components seen in gels. First, itis possible that at least some of them are degradationproducts of the larger macromolecule. Although this isprobably responsible for some of the lower molecularmass components, it seems unlikely to be the onlyfactor accounting for our results, because expression ofthe higher relative molecular mass components can bedissociated from expression of the lower relative mol-ecular mass bands (in the absence of neural crest and inyounger somites). While it might be argued that theoperation of neural crest removal itself releases proteo-lytic enzymes, control embryos destined for biochemi-cal investigations were subjected to cuts in and aroundthe neural tube (without removing the neural crest).Moreover, sections of operated embryos did not showany differences in the pattern of immunoreactivityoutside the operated region, suggesting that diffusibleproteases are not released from the operation site.

A second possibility to consider is that the smallercomponents are not related to Jl/tenascin directly, butare ligands for molecules containing the epitope recog-nised by M1B4, which copurify with M1B4 antigensduring affinity extraction. However, staining is seenwith all three antibodies in the absence of neural crest

Page 11: Cell and Developmental Biology

Jl/tenascin in neural crest and somites 317

cells and in young somites, despite the finding that thehigher molecular mass components are absent fromthese tissues.

A final possibility is that the molecules recognised bythe antibodies used in this study are not single mol-ecules but rather complexes, or associations of severalmolecules of various molecular masses. The possibilityof a single promiscuous epitope present on manydifferent molecules can probably be excluded becausethe two monoclonal antibodies (M1B4 and 578) recog-nise different epitopes. It is possible, however, that themultiple bands seen in silver-stained gels from affinity-purified Jl/tenascin-related molecules represent wholeportions of the larger molecule(s), which may be usedby the embryo in different contexts, for example asprecursors for the assembly of the higher molecularmass components.

The results of this study therefore suggest that agroup of molecules, immunologically related to J l /tenascin, is expressed in the sclerotome of the chickembryo, and that the expression of the characteristic120-230xl03 components may be regulated indepen-dently of the expression of the lower relative molecularmass forms. Further speculations on their function willhave to await characterisation by antibodies that candistinguish between them, as well as knowledge of theiramino-acid sequence.

Are Jl/tenascin-related molecules produced by neuralcrest or by sclerotome cells?It was found both by Tan et al. (1987) and by Mackie etal. (1988) that, in vitro, neural crest cells cultured alonedo not produce cytotactin or tenascin, while somitecultures do. However, it is not clear whether Tan et al. 's(1987) somite cultures contained neural crest cells,which begin to colonise the somitic mesenchyme assoon as the sclerotome can be distinguished (see Rick-mann et al. 1985; Bronner-Fraser, 1986; Teillet et al.1987). Jl/tenascin immunoreactivity in older sclero-tomes is localised both to some of the cells expressingthe HNK-1 and/or 20B4 epitopes (the two neural crestmarkers) and to some not expressing them (Fig. 1 E,F). Since recently formed epithelial somites, which donot yet contain any neural crest cells, already containcytotactin and tenascin immunoreactivity (Tan et al.1987; Mackie et al. 1988) within the somite lumen,Mackie et al. (1988) suggested that the somite cellsthemselves secrete these molecules into the lumen. Ourresults suggest that the Jl/tenascin-immunoreactivematerial detected in and around epithelial somites isdue to molecular components different from thosedetected in older sclerotome/neural crest tissue. Takentogether, therefore, these findings suggest that thepresence of neural crest cells is required for expressionof the higher molecular mass Jl/tenascin-related com-ponents. One possibility, suggested by our results onthe relative molecular masses of the components iso-lated from epithelial somite tissue samples with andwithout neural tube/notochord, is that the somite cellsproduce some of the precursor chains of the tenascinmolecule while the neural crest or other notochord or

neural tube-derived tissues supply other chains. Thisaccounts for our conclusion that both the neural crestand the somite are required for expression of the highermolecular weight components of Jl/tenascin.

Do Jl/tenascin-related molecules play a controllingrole in neural crest migration and/'or axon ougtrowththrough the sclerotome?Our findings appear to resolve the paradox presentedby the results of Tan et al. (1987), Mackie et al. (1988)and Halfter et al. (1989). These authors suggested thatcytotactin and tenascin, respectively, are restricted tothe rostral half of each sclerotome, but that thesemolecules are inhibitory to neural crest migration andspreading in vitro. Our results show that, at the time ofonset of neural crest migration through the rostralhalves of sclerotomes in embryos between stages 13 and16, Jl/tenascin-related molecules are present in thecaudal halves of the sclerotomes, a finding that is moreconsistent with an inhibitory role of these molecules forneural crest migration in vivo. It may be interesting tonote in this context that the notochord, which inhibitsthe migration of neural crest cells (Newgreen etal. 1986;Guillory and Bronner-Fraser, in preparation; Stern andBronner-Fraser, in preparation), is immunoreactivewith antibodies that recognise Jl/tenascin-related mol-ecules (Fig. 1 B-C, E-F); it is therefore possible thatthese molecules contribute to the notochordal inhi-bition of neural crest migration.

The distribution of the cytotactin-binding proteogly-can ('CTB proteoglycan'), reported by Tan etal. (1987)to become localised to the caudal halves of the sclero-tomes, should also be considered. Because CTB pro-teoglycan binds peanut lectin (PNA), and because thislectin has been shown to stain the caudal halves of thesclerotomes of the chick embryo (Stern et al. 1986),CTB proteoglycan could be responsible for the patternof staining seen with PNA. However, peanut lectinbinding is never seen in both halves of the sclerotome,as described for CTB proteoglycan at early stages ofsomite development. Furthermore, preliminary exper-iments have shown that at the stages of developmentwhen PNA binding to caudal sclerotome cells is seen,several PNA-binding molecules of diverse molecularmasses from 17x10 upwards are expressed differen-tially in the two halves of the sclerotome (Davies,Cook, Norris, Keynes and Stern, in preparation).Neither of these observations supports the hypothesisthat CTB proteoglycan is responsible for the pattern ofPNA staining in the sclerotome.

If cytotactin/Jl/tenascin played a controlling role indetermining the migration pathways of trunk neuralcrest cells, one would expect their temporal and spatialpatterns to correlate. However, this is not the case.When the first neural crest cells enter the rostral halvesof the first sclerotomes to form (at about the 8 somitestage, stage-9; Thiery etal. 1982; Rickmann etal. 1985;Bronner-Fraser, 1986; Lim et al. 1987; Loring andErickson, 1987; Teillet et al. 1987; Tosney, 1988), thesesclerotomes display very little, if any, cytotactin/Jl/tenascin immunoreactivity (Crossin et al. 1987; Tan et

Page 12: Cell and Developmental Biology

318 C. D. Stern and others

al. 1987; Mackie et al. 1988 and the present obser-vations). Neural crest cells entering newly formedsclerotomes between stage-13 and stage-16 do so whenthese sclerotomes display immunoreactivity mainly intheir caudal halves (present observations). Finally,neural crest cells entering the sclerotomes that formafter stage-16 do so when Jl/tenascin immunoreactivitybegins to predominate in the rostral halves of thesesclerotomes (Tan et al. 1987; Mackie et al. 1988 and thepresent observations). Thus, some neural crest cellsenter sclerotomes devoid of Jl/tenascin or the rostralhalves of sclerotomes expressing Jl/tenascin in theircaudal halves, while yet others enter the Jl/tenascin-containing halves. These results argue strongly thatcytotactin/Jl/tenascin-related molecules cannot besolely responsible for the segmented pattern of mi-gration of neural crest cells through the sclerotome.

Motor axon outgrowth through the rostral half-sclerotome, however, begins at about stage 17 (Keynesand Stern, 1984), shortly after the onset of cytotactinand of Jl/tenascin expression in the rostral half-sclero-tome. Is it possible that these molecules determine thesegmented pattern of motor axon outgrowth? Ourresults argue strongly that this is not the case. Ablationof the neural crest leads to a change in the spatialpattern of expression of Jl/tenascin-related moleculesin the sclerotome, but the same operation does notdisturb the pattern of axon outgrowth through therostral half-sclerotome (Rickmann et al. 1985). There-fore, it can be concluded that Jl/tenascin-related mol-ecules are not responsible for the segmented pattern ofeither neural crest migration or motor axon outgrowth.

There remains the possibility, nevertheless, thatJl/tenascin-related molecules play a role in the sub-sequent morphogenesis of neural crest derivatives, forexample during dorsal root ganglion formation. Thispossibility is suggested by our finding that the neuralcrest appears to participate in the production of mol-ecules capable of inhibiting its own migration. The timeinterval between the entrance of neural crest cells intothe rostral half of a sclerotome and the change in thelocalisation of Jl/tenascin-immunoreactivity appears tobe equivalent to about 8-10 somites, in turn equivalentto about 12-15 h. It is possible that such a mechanismacts to limit the number of neural crest cells entering thesclerotome. It may also prevent the further migration ofneural crest cells in preparation for dorsal root ganglio-genesis, which occurs at about stage 21 (Lallier andBronner-Fraser, 1988), and may therefore determine abalance between the number of neural crest cellsdestined to contribute to the autonomic nervous systemand the number destined to form sensory neurones.

ConclusionsOur results provide strong evidence that Jl/tenascin-related molecules are not responsible for the segmentedpattern of neural crest cell migration or motor axonoutgrowth. Rather, they suggest that the neural crestmodulates the expression of these molecules in thesclerotome, although it is not clear whether the neuralcrest itself is responsible for their production. J l /

tenascin-related molecules seem to have an inhibitoryeffect on neural crest migration both in vivo and in vitroand could play a role in gangliogenesis and the latermorphogenesis of neural crest derivatives. Our resultssuggest that the names 'tenascin', 'Substrate AdhesionMolecules (SAMs)' and 'cytotactin' given to these mol-ecules (e.g. Grumet et al. 1985) do not suit theirapparent functions as judged by their pattern of ex-pression in the developing chick embryo. While theseterms suggest strong adhesion, cellular affinity andattraction of migrating cells, our results point towards apossible inhibitory role of these molecules in patterningthe peripheral nervous system after the initial stages ofits development.

This work was supported by a grant from Action Researchfor the Crippled Child to CDS. MBF, who is a SloanFoundation Research Fellow, is supported by grants fromUSPHS (HD-15527), March of Dimes (1-896), and theMuscular Dystrophy Association. AF and MS are supportedby the German Research Society. MS is a member of theSwiss Federal Institute of Technology, Zurich.

References

ABO, T. & BALCH, C. M. (1981). A differentiation antigen ofhuman NK and K cells identified by a monoclonal antibody(HNK-1). J. Immunol. 127, 1024-1029.

AUFDERHEIDE, E. & EKBLOM, P. (1988). Tenascin during gutdevelopment: appearance in the mesenchyme, shift in molecularforms, and dependence on epithelial-mesenchymal interactions.J. Cell Biol. 107, 2341-2349.

BOURDON, M. A., MATTHEWS, T. J., Pizzo, S. V. & BICNER, D. D.(1985). Immunochemical and biochemical characterization of aglioma-associated extracellular matrix glycoprotein. J. cell.Biochem. 28, 183-195.

BRONNER-FRASER, M. (1986). Analysis of the early stages of trunkneural crest cell migration in avian embryos using monoclonalantibody HNK-1. Devi Biol. 115, 44-55.

CANNING, D. R. & STERN, C. D. (1988). Changes in the expressionof the carbohydrate epitope HNK-1 associated with mesoderminduction in the chick embryo. Development 104, 643-655.

CHIQUET, M. & FAMBROUGH, D. M. (1984). Chick myotendinousantigen. II. A novel extracellular glycoprotein complex consistingof large disulfide-linked subunits. J. Cell Biol. 98, 1937-1946.

CROSSIN, K. L., HOFFMAN, S., GRUMET, M., THIERY, J.-P. &EDELMAN, G. M. (1986). Site-restricted expression of cytotactinduring development of the chicken embryo. J. Cell Biol. 102,1917-1930.

FAISSNER, A., KRUSE, J., CHIQUET-EHRISMANN, R. & MACKIE, E.(1988). The high-molecular-weight J l glycoproteins areimmunologically related to tenascin. Differentiation 37, 104-114.

FRIEDLANDER, D. R., HOFFMAN, S. & EDELMAN, G. M. (1988).Functional mapping of cytotactin: proteolytic fragments active incell-substrate adhesion. J. Cell Biol. 107, 2329-2340.

GRUMET, M., HOFFMAN, S., CROSSIN, K. L. & EDELMAN, G. M.(1985). Cytotactin, an extracellular matrix protein of neural andnon-neural tissues that mediates neuron-glia interactions. Proc.natn. Acad. Sci. U.S.A. 82, 8075-8079.

HALFTER, W., CHIQUET-EHRISMANN, R. & TUCKER, R. P. (1989).The effect of tenascin and embryonic basal lamina on thebehavior and morphology of neural crest cells in vitro. Devi Biol.132, 14-25.

HAMBURGER, V. & HAMILTON, H. L. (1951). A series of normalstages in the development of the chick. J. Morph. 88, 49-92.

HOFFMAN, S., CROSSIN, K. L. & EDELMAN, G. M. (1988).Molecular forms, binding functions, and developmentalexpression of cytotactin and cytotactin-binding proteoglycan, aninteractive pair of extracellular matrix proteins. J. Cell Biol. 106,519-532.

Page 13: Cell and Developmental Biology

Jl Itenascin in neural crest and somites 319

HOFFMAN, S. & EDELMAN, G. (1987). A proteoglycan with HNK-1antigenic determinants is a neuron associated ligand forcytotactin. Proc. natn. Acad. Sci. U.S.A. 84, 2523-2527.

KEYNES, R. J. & STERN, C. D. (1984). Segmentation in thevertebrate nervous system. Nature, Lond. 310, 786-789.

KEYNES, R. J. & STERN, C. D. (1988). Mechanisms of vertebratesegmentation. Development 103, 413-429.

KRUSE, J., KEILHAUER, G., FAISSNER, A., TIMPL, R. & SCHACHNER,

M. (1985). The J l glycoprotein-a novel nervous system celladhesion molecule of the L2/HNK-1 family. Nature, Lond. 316,146-148.

KONEMUND, V. , JUNGALWALA, F. B . , FlSCHER, G., CHOU, D . K.H., KEILHAUER, G. & SCHACHNER, M. (1988). The L2/HNK-1carbohydrate of neural cell adhesion molecules is involved in cellinteractions. J. Cell Biol. 106, 213-223.

LAEMMLI, U. K. (1970). Cleavage of structural proteins during theassembly of the head of bacteriophage T4. Nature, Lond. 227,680-685.

LALLIER, T. E. & BRONNER-FRASER, M. (1988). A spatial andtemporal analysis of dorsal root and sympathetic ganglionformation in the avian embryo. Devi Biol. 127, 99-112.

LAYER, P. G., ALBER, A. & RATHJEN, F. G. (1988). Sequentialactivation of butyrylcholinesterase in rostral half somites andacetylcholinesterase in motoneurones and myotomes precedinggrowth of motor axons. Development 102, 387-396.

LIM, T. M., LUNN, E. R., KEYNES, R. J. & STERN, C. D. (1987).The differing effects of occipital and trunk somites on neuraldevelopment in the chick embryo. Development 100, 525-533.

LORING, J. F. & ERICKSON, C. A. (1987). Neural crest cellmigratory pathways in the trunk of the chick embryo. Devi Biol.121,220-236.

MACKIE, E. J., TUCKER, R. P., HALFTER, W., CHIQUET-EHRISMANN,R. & EPPERLEIN, H. H. (1988). The distribution of tenascincoincides with pathways of neural crest cell migration.Development 102, 237-250.

MATSUDAIRA, P. & BURGESS, D. R. (1978). SDS microslab lineargradient polyacrylamide gel electrophoresis. Analyt. Biochem.87, 386-396.

NEWGREEN, D. F., SCHEEL, M. & KASTNER, V. (1986).Morphogenesis of sclerotome and neural crest in avian embryos.In vivo and in vitro studies on the role of notochordalextracellular material. Cell Tissue Res. 244, 299-313.

NORRIS, W. E., STERN, C. D. & KEYNES, R. J. (1989). Moleculardifferences between the rostral and caudal halves of thesclerotome in the chick embryo. Development 105, 541-548.

PRIMMETT, D. R. N., NORRIS, W. E., CARLSON, G. J., KEYNES, R.J. & STERN, C. D. (1989). Periodic segmental anomalies inducedby heat-shock in the chick embryo are associated with the cellcycle. Development 105, 119-130.

RICKMANN, M., FAWCETT, J. W. & KEYNES, R. J. (1985). The

migration of neural crest cells and the growth of motor axonsthrough the rostral half of the chick somite. J. Embryol. exp.Morph. 90, 437-455.

STERN, C. D. , SISODIYA, S. M. & KEYNES, R. J. (1986). Interactions

between neurites and somite cells: inhibition and stimulation ofnerve growth in the chick embryo. J. Embryol. exp. Morph. 91,209-226.

STRUDEL, G. (1955). L'action morphogene du tube nerveux et de lacorde sur la differentiation des vertebres et des musclesvertebraux chez l'embryon de poulet. Arch. Anat. Microsc.Morph. Exp. 44, 209-235.

TAN, S.-S., CROSSIN, K. L., HOFFMAN, S. & EDELMAN, G. M.

(1987). Asymmetric expression in somites of cytotactin and itsproteoglycan ligand is correlated with neural crest celldistribution. Proc. natn. Acad. Sci. U.S.A. 84, 7977-7981.

TEILLET, M.-A., KALCHEIM, C. & LE DOUARIN, N. M. (1987).

Formation of the dorsal root ganglion in the avian embryo:segmental origin and migratory behavior of neural crestprogenitor cells. Devi Biol. 120, 329-347.

TEILLET, M.-A. & L E DOUARIN, N. M. (1983). Consequences ofneural tube and notochord excision on the development of theperipheral nervous system in the chick embryo. Devi Biol. 98,192-211.

THIERY, J. P., DUBAND, J. L. & DELOUvfiE, A. (1982). Pathwaysand mechanisms of avian trunk neural crest cell migration andlocalization. Devi Biol. 93, 324-343.

TOSNEY, K. W. (1988). Somites and axon guidance. ScanningMicrosc. 2, 427-442.

TOSNEY, K. W., WATANABE, M., LANDMESSER, L. & RUTISHAUSER,U. (1987). The distribution of NCAM in the chick hindlimbduring axon outgrowth and synaptogenesis. Devi Biol. 114,437-452.

TUCKER, G. C , AOYAMA, H., LIPINSKI, M., TURSZ, T. & THIERY, J.P. (1984). Identical reactivity of monoclonal antibodies HNK-1and NC-1: Conservation in vertebrates on cells derived from theneural primordium and on some leukocytes. Cell Differ. 14,223-230.

VAUGHAN, L., HUBER, S., CHIQUET, M. & WINTERHALTER, K. H.

(1987). A major, six-armed glycoprotein from embryoniccartilage. EMBO J. 6, 349-354.

YNTEMA, C. L. & HAMMOND, W. S. (1954). The origin of intrinsicganglia of trunk viscera from vagal neural crest in the chickembryo. J. comp. Neurol. 101, 515-541.

{Accepted 26 June 1989)