Proc. Natl. Acad. Sci. USAVol. 92, pp. 10590-10594, November 1995Biochemistry

The versican C-type lectin domain recognizes the adhesionprotein tenascin-R

(J1-160/180/janusin)

ANDERS ASPBERG, CHRISTOPH BINKERT*, AND ERKKI RUOSLAHTItCancer Research Center, La Jolla Cancer Research Foundation, La Jolla, CA 92037

Communicated by Elizabeth D. Hay, Harvard Medical School, Boston, MA, July 25, 1995 (received for review March 31, 1995)

ABSTRACT The core proteins of large chondroitin sul-fate proteoglycans contain a C-type lectin domain. The lectindomain of one of these proteoglycans, versican, was expressedas a recombinant 15-kDa protein and shown to bind toinsolubilized fucose and GIcNAc. The lectin domain showedstrong binding in a gel blotting assay to a glycoprotein doubletin rat brain extracts. The binding was calcium dependent andabolished by chemical deglycosylation treatment of the ligandglycoprotein. The versican-binding glycoprotein was identi-fied as the cell adhesion protein tenascin-R, and versican andtenascin-R were both found to be localized in the granularlayer of rat cerebellum. These results show that the versicanlectin domain is a binding domain with a highly targetedspecificity. It may allow versican to assemble complexescontaining proteoglycan, an adhesion protein, and hyaluro-nan.

Versican, aggrecan, neurocan, and brevican constitute a familyof chondroitin sulfate proteoglycans that form aggregates withhyaluronan (1-4). The C termini of these proteoglycans con-tain a C-type lectin domain, as well as epidermal growthfactor-like and complement regulatory-like sequence motifs.This domain arrangement is similar to that of the selectins,which are adhesion receptors involved in leukocyte homingand extravasation at inflammatory sites (5-8).The C-type animal lectins are defined by the presence of a

calcium-dependent carbohydrate-recognition domain in whicha pattern of 32 amino acid residues, spaced over a stretch of- 120 amino acids, is highly conserved (9). The lectin domainsof human and chicken versican show 96% amino acid identity(10, 11), and the human and rat versican lectin domains differin only one amino acid (Thr-2257 in human versican replacedby Ala; Joan Lemire and Thomas N. Wight, personal commu-nication), suggesting that the versican family lectin domainsmay have an important conserved function. However, verylittle is known about the function of these lectin domains. Theaggrecan lectin domain has been shown to bind to fucose andgalactose in a calcium-dependent manner (12, 13). Recently,the recombinant C-terminal part (epidermal growth factor-like, lectin, and complement regulatory-like domains) ofchicken versican was shown to bind to simple carbohydrates, aswell as to heparin and heparan sulfate, in affinity chromatog-raphy (14). A C-terminal fragment of neurocan has beenreported to bind tenascin and the cell-cell adhesion proteinsN-CAM and Ng-CAM (15-17), but it is unclear if the lectindomain is responsible for these interactions.We have studied the potential of the lectin domain of

versican to function as a binding domain. Here we report thatrecombinant versican lectin domain, produced in mammaliancells has carbohydrate-binding activity and that it also bindsthe extracellular matrix glycoprotein tenascin-R (also knownas J1-160/180 and janusin).

EXPERIMENTAL PROCEDURESMaterials. Antiserum against the versican lectin domain was

raised against a synthetic peptide (KMFEHDFRWTDG-STLQYEN), corresponding to amino acid residues 2244-2262(10), and affinity purified by using the peptide. The anti-versican antiserum used in immunofluorescence staining wasraised against recombinant intact versican (1) and affinitypurified by using a bacterially expressed fragment of versican,corresponding to amino acid residues 51-354 (10). Mousemonoclonal anti-tenascin-R antibody #596 (18) was a kind giftfrom Melitta Schachner (Eidgenossiche Technische Hochs-chule, Zurich), and anti-tenascin-C antiserum was from TeliosPharmaceuticals (La Jolla, CA). Agarose beads conjugated tovarious monosaccharides or to streptavidin were from Sigma.

Plasmid Construction and Expression of the Versican Lec-tin Domain. The human versican lectin domain (amino acidresidues 2177-2305; ref. 10) was amplified from cDNA clone2B (19) by PCR with upstream primer 5'-GTCAACTC-GAGACAAGATACCGAGACATGTGAC-3' and down-stream primer 5'-TGCTCTAGATCAGGTCGACCCTT-TCTTGCACGTATAGGTG-3'. The lectin-coding sequencewas then fused to an immunoglobulin signal peptide (20) andinserted into pcDNA I/Neo (Invitrogen). The sequence of theconstruct was determined by using Sequenase II (UnitedStates Biochemical). CHO/DG44 cells (21) were cotrans-fected with the versican lectin domain construct and pSV2-dhfr (22) by using calcium phosphate transfection (CellPhect;Pharmacia). Transfected cells were selected in nucleoside-freea-modified minimal essential medium (GIBCO/BRL) con-taining 9% (vol/vol) dialyzed fetal calf serum, and lectinexpression in selected colonies was amplified with methotrex-ate according to standard procedures. Cell supernatants werescreened for versican lectin secretion by immunoblotting withanti-lectin-domain antiserum.

Purification and Biotinylation of Versican Lectin Domain.Cells expressing the versican lectin were adapted to growth inserum-free medium (Nutri Base + Vita Cyte serum supple-ment; J. Brooks Laboratories, San Diego). Proteins fromconditioned serum-free medium were precipitated in thepresence of 5 mM EDTA with ammonium sulfate (50%saturation) and dissolved in Hepes-buffered saline (HBS; 50mM Hepes, pH 7.4/0.88% NaCl) containing 5 mM CaCl2.Thereafter, the lectin was enriched through affinity chroma-tography on a fucose-agarose column and eluted with HBScontaining 20 mM EDTA. After adding CaCl2 to 50 mM andNaOH to pH 8.5, the lectin was biotinylated with NHS-LC-biotin (Pierce) and purified by C4 reverse-phase HPLC. Thesamples were injected onto a 214TP54 column (Vydac Sepa-

Abbreviations: ELC, epidermal growth factor-like, lectin, and com-plement regulatory-like domains; TFMS, trifluoromethanesulfonicacid; PNGase F, glycopeptide N-glycosidase; Endo Hf, protein fusionof endo-,B-N-acetylglucosaminidase and maltose-binding protein.*Present address: Samuel Lunenfeld Research Institute, Mount SinaiHospital, 600 University Avenue, Toronto ON, Canada M5G 1X5.tTo whom reprint requests should be addressed.

10590

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement" inaccordance with 18 U.S.C. §1734 solely to indicate this fact.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

Proc. Natl. Acad. Sci. USA 92 (1995) 10591

rations, Hesperia, CA) equilibrated with 0.06% trifluoroaceticacid (TFA). A linear acetonitrile gradient (0-80% acetonitrilein 0.06% TFA) was applied, and the protein was eluted at 44%acetonitrile. Lectin-containing fractions were lyophilized andresuspended in HBS containing 5 mM CaCl2 and 0.1% bovineserum albumin.

Construction and Expression of Versican C-Terminal Do-main-IgG Fusion Protein. The 3' part of versican cDNA,corresponding to the epidermal growth factor-like, lectin, andcomplement regulatory-like domains (ELC; nucleotides 6530-7493; ref. 10), was amplified by PCR from clone 2B and fusedin frame with the cDNA coding for the signal peptide and thefirst eight amino acids of cr5 integrin (residues 1-49; ref. 23).This fragment was fused to human genomic DNA encoding theFc region of IgGl (24), inserted into plasmid pcDNA I/Neo,and sequenced. The plasmid was introduced, together withpSV2-dhfr, into CHO/DG44 cells by using Lipofectamine(GIBCO/BRL). The ELC-IgG fusion protein was purifiedfrom cell culture medium by affinity chromatography onprotein A-Sepharose FF (Pharmacia).

Brain Extracts. Adult, female Sprague-Dawley rats wereeuthanized with carbon dioxide, and the brains were imme-diately removed and minced in ice-cold 0.9% NaCl. The tissuewas homogenized with a Polytron in nine volumes of 4 mMHepes, pH 7.5/0.3 M sucrose, containing 0.25 mg of N-ethylmaleimide per ml, 1 mM EDTA, and 0.2 mM phenyl-methylsulfonyl fluoride as protease inhibitors. The homoge-nate was centrifuged at 12,000 x g for 30 min. The supernatantwas transferred to new tubes and centrifuged at 378,000 x g.The resulting supernatant was aliquotted, quick frozen inliquid nitrogen, and stored at -70°C until used.

Protein Blotting. Brain extract was boiled 5 min in samplebuffer, electrophoresed through precast SDS/4-12% poly-acrylamide gels (NOVEX, San Diego) with prestained molec-ular mass markers (GIBCO/BRL), and electroblotted ontonitrocellulose membranes (Schleicher & Schuell). The mem-branes were treated with 3% bovine serum albumin in TTBS(25 mM Tris-HCl, pH 7.4/0.88% NaCl/0.1% Tween 20) toinhibit nonspecific binding, incubated with biotinylated versi-can lectin or antibodies in 0.3% bovine serum albumin inTTBS, washed with TTBS, incubated with avidin-conjugatedhorseradish peroxidase (Vectastain ABC Elite; Vector Lab-oratories) or peroxidase-conjugated secondary antibody,washed with TTBS, and visualized by chemiluminescence(ECL; Amersham). For the lectin blots, all solutions containedeither 5 mM CaCl2 or 5 mM EDTA.

Deglycosylation of Brain Extracts. Aliquots of brain extractwere precipitated in acetone, dissolved in appropriate buffersfor enzyme digestions, and incubated with or without recom-binant glycopeptide N-glycosidase (PNGase F; EC 3.5.1.52;New England Biolabs), recombinant protein fusion of Endo-,B-N-acetylglucosaminidase (EC 3.2.1.96) and maltose-bindingprotein (Endo Hf; New England Biolabs), or Vibrio choleraeneuraminidase (EC 3.2.1.18; Boehringer Mannheim). Thebuffers for these incubations were provided by the manufac-turer or prepared according to the manufacturer's instructions.After a 12-h incubation, the samples were boiled in SDS/PAGE sample buffer. Deglycosylation with trifluoromethane-sulfonic acid (TFMS) was performed as described (25).

Precipitation of Tenascin-R with Versican Lectin. Brainextract was incubated for 1 h on ice with biotinylated versicanlectin in the presence of 5 mM CaCl2 or 20 mM EDTA.Streptavidin-conjugated agarose beads were then added, andthe incubation was continued for 1 h at 4°C with gentlerotation. The beads were washed four times in Tris-bufferedsaline (TBS; 25 mM Tris-HCl, pH 7.4/0.88% NaCl/5 mMCaCl2 or 20 mM EDTA) and boiled in SDS/PAGE samplebuffer containing 2.5% 2-mercaptoethanol. The samples wereelectrophoresed and electroblotted, and tenascin-R was de-tected as described above.

Immunofluorescence Staining. The cerebellum from anadult, male Sprague-Dawley rat was dissected, removed, em-bedded in O.C.T. compound (Miles), frozen, and cut into 5-,umcryosections. The sections were air dried for 20 min andrehydrated in TBS. Nonspecific binding was blocked by incu-bating sections in 5% (vol/vol) goat serum/TBS. Primaryantibodies were then applied in 2% (vol/vol) goat serum/TBS.The sections were washed, incubated with fluorescein-conjugated secondary antibodies raised in goat, washed again,and mounted in Vectashield mounting medium (Vector Lab-oratories). The slides were inspected in a Zeiss AxioVertmicroscope equipped for fluorescence microscopy, and pic-tures were taken with Kodak T-Max 400 film.

RESULTSBinding of the Versican Lectin Domain to Monosacchar-

ides. The versican lectin domain was expressed in CHO cells,and a band corresponding to the expected molecular mass ofthe lectin (14.7 kDa) in culture medium was detected byimmunoblotting with anti-versican lectin antibodies. To deter-mine if the lectin-domain protein could bind carbohydrates,the culture medium from the CHO cells was incubated withmonosaccharides coupled to agarose beads. The lectin domainbound to beads coupled with fucose or GlcNAc but not tobeads coupled with GalNAc, mannose, or glucose (Fig. 1).Chelation of calcium with EDTA or EGTA eliminated binding(data not shown). Thus, the versican lectin domain appears tobe a functional carbohydrate-binding domain.

Versican Lectin Domain Binds to a Brain Protein. Thebiotinylated versican lectin domain bound to two proteins ofapproximate molecular masses 170 and 190 kDa in rat-brainextracts (Fig. 2A). Rat brain was chosen as a primary targetsince versican is prominently expressed in the brain (26, 27), asare the other members of this proteoglycan family (2, 4, 28).The 75- and 130-kDa bands on the blots appear to be non-specific because they are also detected by the avidin-horseradish peroxidase conjugate alone (Fig. 2A, control).The binding of the lectin to the 170- and 190-kDa bands wasinhibited in the presence of EDTA (Fig. 2A).The 170- and 190-kDa bands were also detected when blots

were probed with ELC-IgG recombinant fusion protein, andthis binding was also calcium dependent (Fig. 2B).The 170- and 190-kDa Glycoproteins React with Anti-

Tenascin-R Antibodies. The sizes of the 170- and 190-kDaglycoproteins we detected with the biotinylated lectin sug-gested to us that they might be tenascin-R (29). Furthermore,its presence in the brain and the reduction in apparentmolecular mass of "20 kDa by PNGase F treatment (see Fig.

o c)

61 cZ ZCA OoD CD~

S B S B

CD0En 0

c 0 (0C

2 a L c

0

S B S B SB 0

674330

20.114.4

FIG. 1. Versican lectin domain binds specifically to fucose andGlcNAc. Culture supernatants from versican-lectin-domain-expressing CHO cells were incubated with monosaccharide-conjugated beads, the beads were pelleted by centrifugation, and beads(B) and supernatants (S) were analyzed for versican lectin by SDS/PAGE and immunoblotting with anti-versican-lectin-domain-specificantibodies. The positions of molecular mass markers (in kDa) are onthe left. The control lane contains untreated culture supernatant.

Biochemistry: Aspberg et aL

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

10592 Biochemistry: Aspberg et al.

± <+\ + <

A:A >u >u °u B { cm

203215-

105-71 0651 -

4329

28 18-14

FIG. 2. Calcium-dependent binding of versican lectin domain to170- and 190-kDa proteins in rat brain extracts. (A) Rat-brain extractswere electrophoresed, transferred to nitrocellulose, and probed withbiotinylated versican lectin domain (VcL) in the presence of 5 mMCaCl2 or EDTA. Bound lectin was detected with avidin-horseradishperoxidase, chemiluminiscent substrate, and autoradiography. In thecontrol lane, the lectin was omitted to show nonspecific binding byavidin-horseradish peroxidase. (B) As in A, but probed with versicanELC-IgG fusion protein and horseradish peroxidase-conjugated anti-human IgG antibodies. The control lane (IgG) was probed with humanIgG instead of ELC-IgG.

5) are in agreement with the reported expression pattern oftenascin-R and the reduction of its molecular mass afterremoval of N-linked carbohydrates (18).

Immunoblotting of brain extracts fractionated by SDS/PAGE with either an anti-tenascin-R monoclonal antibody orthe biotinylated lectin domain resulted in virtually identicalstaining patterns with reduced samples (Fig. 3A). When un-reduced samples were probed, the lectin and the anti-tenascin-R antibody stained the same bands, but an additional band wasdetected with the antibody (Fig. 3B). This band may representan alternative glycoform or splicing product (29) that is notreactive with the versican lectin domain and that is not seen inthe reducing gel because of comigration with one of the other

A u cr: +

z Z

JJ1 4J1 -iC C uCI X >

B n, + <:- U LU

4-X

m > >

04JC0u

bands. Immunoblotting with anti-tenascin-C antiserum gaveonly weakly staining bands in the adult rat-brain extracts (datanot shown); a strongly staining 230-kDa band and a weaklystaining triplet between 210 and 220 kDa were seen in extractprepared from brains of 11-day-old rats (Fig. 3A). The versicanlectin did not bind tenascin-C in brain extracts prepared fromeither adult or 11-day-old rats (data not shown).To confirm the identity of the gel bands detected with the

versican lectin domain as tenascin-R, the ligand was precipi-tated from brain extracts by employing biotinylated versicanlectin domain and streptavidin-conjugated agarose beads, andanalyzed by immunoblotting. As shown in Fig. 4, the lectindomain-bound material reacted with anti-tenascin-R. Nobands were detected in the lectin domain-bound fraction whenEDTA had been added to the brain extract prior to theincubation with the lectin (Fig. 4, lane 3) or when the extractwas incubated with beads without the lectin domain (Fig. 4,lane 4).Chemical Deglycosylation of Tenascin-R Eliminates Versi-

can Lectin Domain Binding. When all carbohydrates wereremoved from brain-extract proteins through chemical degly-cosylation with TFMS, versican lectin domain binding totenascin-R was abolished (Fig. 5). Incubation with PNGase F,which removes N-linked carbohydrates, reduced the apparentmolecular mass of tenascin-R but did not decrease lectindomain binding (Fig. 5). Incubation with Endo Hf had noeffect on versican lectin binding, although concanavalin Astaining of a parallel blot was decreased after the Endo Hftreatment (data not shown). V cholerae neuraminidase causeda slight reduction in the apparent molecular mass withoutaffecting lectin domain binding (Fig. 5). The effect of TFMSdid not seem to have been caused by degradation of the proteinbackbones in the extracts, as judged from Ponceau S stainingof the blots (data not shown) and the continued presence of the75- and 130-kDa avidin-binding bands in the TFMS-treatedlane (Fig. 5). The size of the tenascin-R core protein in theTFMS treatment was evaluated by stopping the reaction afterdifferent incubation times. As shown in Fig. 6B, the versican

:

M LULo U

EVPm

1 23 4

203 - tV

105 -

71 - _

44 -

28 -

Reducing

w ww

Non red ucing

FIG. 3. Versican lectin domain ligand comigrates with tenascin-R.Rat-brain extract was boiled in SDS/PAGE sample buffer with (A;Reducing) or without (B; Nonreducing) 2-mercaptoethanol, electro-phoresed, transferred, and probed with biotinylated versican lectindomain (VcL), as in Fig. 2, with mouse monoclonal anti-tenascin-Rantibody #596 (anti-TN-R), or with polyclonal anti-tenascin C (anti-TN-C). Tenascin-C staining in adult rat-brain extract was very weak;thus, the anti-tenascin-C lane inA shows staining in postnatal day 11rat-brain extract, while the sample in all other lanes is adult rat-brainextract. The control lane shows the nonspecific staining by avidin-horseradish peroxidase in the absence of versican lectin domain.

FIG. 4. Precipitation of tenascin-R by biotinylated versican lectindomain and streptavidin beads. Brain extract was incubated withbiotinylated versican lectin in the presence of 5 mM CaCl2 or 20 mMEDTA or without versican lectin (beads only). The lectin was thencollected by addition of streptavidin-conjugated agarose beads, whichwere then washed and boiled in SDS/PAGE sample buffer containingreducing agent, and the solubilized sample was analyzed by immuno-blotting with the anti-tenascin-R monoclonal antibody. Lane 1, brainextract only; lane 2, brain extract, lectin, CaCl2, and beads; lane 3, brainextract, lectin, EDTA, and beads; and lane 4, brain extract, CaCl2, andbeads.

203 -

105 -

71 -

44 -

28 -

Proc. Natl. Acad. Sci. USA 92 (1995)

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

Proc. Natl. Acad. Sci. USA 92 (1995) 10593

0cn

C-o

Co

U) co Cr13OI E

F- w Z

+_ +- +-

A anti-versican B anti-tenascin R

ML

PCL

GL

203-

105-

71 - ^s '- i

44-

28 -

FIG. 5. Effect of deglycosylation on versican lectin domain bindingto brain extract proteins. Rat brain extract was chemically deglycosy-lated with TFMS or enzymatically deglycosylated with differentglycosidases, electrophoresed under reducing conditions, and analyzedby lectin blotting as in Fig. 2. Aliquots of brain extract were incubatedin the same buffer with (+) or without (-) enzyme or TFMS. Thearrow indicates the position of the unaltered tenascin R. Endo H, EndoHf.

lectin domain binding to tenascin-R was lost after a 10-secincubation in the acid, whereas anti-tenascin-R antibodiesdetected a band with the size of the intact core polypeptideeven after 120 min (Fig. 6A).

Versican and Tenascin-R Immunolocalization in Rat Cer-ebellum. Immunofluorescence staining of rat cerebellumshowed versican immunoreactivity predominantly in the gran-ular layer and to some extent also in the white matter tracts(Fig. 7A), in agreement with previous reports (27, 30). Tena-scin-R staining was mainly found in the granular layer, butstaining could also be found in the molecular layer and to alesser extent in the white matter (Fig. 7B), also in agreementwith previous findings (18).

DISCUSSIONWe show that the lectin domain of versican interacts withfucose and GlcNAc and that it binds specifically and in a

2

A°<,= -c S B ° U'-cS~~~~~~~~~~~~cA u Eln E E

OQOQQc~~~~~~~~~~~i OO Oru~~~~~~~~~4aco o o o 0 0o 0o

fK~~~~~~~~~~~~~~~~~........................

_. ..,i ,,' .s! :.,:.......*.-......~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~. S... ... .. .. ...98 -......

64 -

36 -

anti-TN-R

- 250

- 98

6450

- 36

VcL-overlay (Ca2+)

FIG. 6. Effect of chemical deglycosylation on tenascin-R versicanbinding and core protein size. Aliquots of rat brain extract weredeglycosylated with TFMS for the indicated times, electrophoresedunder reducing conditions, and analyzed by immunoblotting and lectinoverlay blot assay as in Fig. 2. (A) Anti-tenascin-R (anti-TN-R)immunoblot. (B) Versican lectin domain (VcL) binding to an identicalfilter to the one in A.

WM

ML

PCL

GL

WM

C Rabbit lgG D Mouse IgG

FIG. 7. Both versican and tenascin R are localized in the granularlayer of rat cerebellum. Adjacent cryosections of rat cerebellum wereincubated with affinity-purified anti-versican antiserum (A) or anti-tenascin-R monoclonal antibody (B) followed by fluorescein-conjugated secondary antibodies. Control sections, where the primaryantibodies were replaced with rabbit (C) or mouse (D) IgG at the sameconcentrations, were completely devoid of staining. ML, molecularlayer; PCL, Purkinje cell layer; GL, granular layer; and WM, whitematter.

calcium-dependent manner to tenascin-R. The carbohydratebinding of versican lectin domain was demonstrated in twoways. First, the lectin domain (as well as the ELC-IgGchimera; data not shown) bound to some, but not all, simplecarbohydrates. This binding was dependent on divalent cat-ions. Similar binding to simple sugars has been observed forthe lectin domain of aggrecan, another member of the sameproteoglycan family (12, 13), and recently also for the C-terminal part of the chicken homolog of versican (14). In thelatter study, only a fragment containing the ELC was found tobe active in carbohydrate binding, whereas a smaller fragmentcomposed of the epidermal growth factor-like and lectindomains was inactive. The reason for the inactivity of thesmaller fragment may be that, as a bacterially producedprotein, it may not have folded correctly. We have attemptedbacterial expression of the versican lectin domain in twodifferent bacterial expression systems, but in both cases theproduct has-been inactive (unpublished results). In contrast,the eukaryotic expression system used here yielded versicanlectin domain that has carbohydrate-binding activity.The second line of evidence supporting the function of the

versican lectin domain as a lectin is the loss of binding totenascin-R in chemically deglycosylated brain extract. Thus,the versican lectin domain would appear to be a lectin, thebinding of which to tenascin-R is calcium dependent andcarbohydrate directed, as is characteristic of C-type lectins (9).However, these data regarding the possible carbohydratenature of the versican lectin binding site in tenascin R have tobe interpreted cautiously. At least one C-type lectin, the IgE

Biochemistry: Aspberg et aL

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

10594 Biochemistry: Aspberg et al.

receptor FceR2, binds its ligand in a carbohydrate-independentmanner (31). Moreover, the chemically deglycosylated tenas-cin-R could have lost its ability to bind versican lectin domain asa result of a modification in the protein backbone. Thus, whetherthe tenascin-R binding is purely carbohydrate mediated or mightalso depend on protein recognition remains to be determined.The highly selective binding of the versican lectin to tena-

scin-R in brain extracts suggests that this interaction is phys-iologically significant. The similarity of the tissue localizationsof versican and tenascin-R in cerebellum provides an addi-tional indication of such significance. We found strong stainingfor both proteins in the granular layer of cerebellum and alsodetected both of them in the white matter. In addition,tenascin-R, but not versican, staining was observed in themolecular layer with the anti-tenascin-R used in this study.This antibody reacts with all of the isoforms of tenascin-R (18).As one of the isoforms of tenascin-R detected in nonreducingSDS/PAGE did not react with versican lectin domain, it ispossible that one or more of the tenascin-R isoforms are notpresent in the molecular layer and that versican-binding tena-scin-R codistributes with versican more completely than isapparent from our study.The various proteoglycan lectin domains may have different

ligand specificities and functions; the versican lectin domainbound to both fucose and GlcNAc, whereas aggrecan lectindomain binds to fucose and galactose but only weakly toGlcNAc (12, 13). Thus, the saccharide-binding specificities ofthe proteoglycan lectins may be related but distinct. As neu-rocan may bind tenascin-C (15, 32, 33), versican and neurocanmay each bind to a different member of the tenascin proteinfamily. Both tenascins and proteoglycans have been classifiedas antiadhesive and have been found to be expressed in barriersof neuronal outgrowth in vivo (for review see ref. 34). Suchspecialized matrixes consisting of a proteoglycan, tenascin-family glycoprotein, and hyaluronan may be formed as a resultof the lectin interactions described in this paper.

We thank Dr. Melitta Schachner for providing the anti-tenascin-Rantibody, Dr. Brian Seed for the IgG-fusion plasmid, and Drs. HudsonFreeze and Minoru Fukuda for helpful discussions. We are alsograteful to our colleagues at the La Jolla Cancer Research Foundationfor comments on the manuscript. This work was supported by GrantsCA42507, CA28896, and Cancer Center Support Grant CA30199(E.R.) from the National Cancer Institute. A.A. was supported byfellowships from the Swedish Cancer Research Fund and the SwedishInstitute. C.B. was supported by fellowships from the EuropeanMolecular Biology Organization (ALTF 129-1990) and the SwissNational Foundation (823A-033322.92).

1. LeBaron, R. G., Zimmermann, D. R. & Ruoslahti, E. (1992) J.Biol. Chem. 267, 10003-10010.

2. Rauch, U., Karthikeyan, L., Mauel, P., Margolis, R. U. & Mar-golis, R. K. (1993) J. Biol. Chem. 267, 19536-19547.

3. Doege, K. J., Sasaki, M., Kimura, T. & Yamada, Y. (1991)J. Biol.Chem. 266, 894-902.

4. Yamada, H., Watanabe, K., Shimonaka, M. & Yamaguchi, Y.(1994) J. Bio. Chem. 269, 10119-10126.

5. Bevilacqua, M. P. & Nelson, R. M. (1993) J. Clin. Invest. 91,379-387.

6. Springer, T. A. (1994) Cell 76, 301-314.7. Varki, A. (1994) Proc. Natl. Acad. Sci. USA 91, 7390-7397.8. Lasky, L. A. (1992) Science 258, 964-969.9. Drickamer, K. (1988) J. Biol. Chem. 263, 9557-9560.

10. Zimmermann, D. R. & Ruoslahti, E. (1989) EMBO J. 8, 2975-2981.

11. Shinomura, T., Nishida, Y., Ito, K. & Kimata, K. (1993) J. Bio.Chem. 268, 14461-14469.

12. Saleque, S., Ruiz, N. & Drickamer, K. (1993) Glycobiology 3,185-190.

13. Halberg, D. F., Proulx, G., Doege, K., Yamada, Y. & Drickamer,K. (1988) J. Biol. Chem. 263, 9486-9490.

14. Ujita, M., Shinomura, T., Ito, K., Kitagawa, Y. & Kimata, K.(1994) J. Biol. Chem. 269, 27603-27609.

15. Grumet, M., Milev, P., Sakurai, T., Karthikeyan, L., Bourdon, M.,Margolis, R. K. & Margolis, R. U. (1994) J. Biol. Chem. 269,12142-12146.

16. Grumet, M., Flaccus, A. & Margolis, R. U. (1993) J. Cell Biol.120, 815-824.

17. Friedlander, D. R., Milev, P., Margolis, R. K., Margolis, R. U. &Grumet, M. (1994) J. Cell Bio. 125, 669-680.

18. Pesheva, P., Spiess, E. & Schachner, M. (1989) J. Cell Bio. 109,1765-1778.

19. Krusius, T., Gehlsen, K. R. & Ruoslahti, E. (1987)J. Bio. Chem.262, 13120-13125.

20. Rogelj, S., Weinberg, R. A., Fanning, P. & Klagsbrun, M. (1988)Nature (London) 331, 173-175.

21. Urlaub, G. & Chasin, L. A. (1980) Proc. Natl. Acad. Sci. USA 77,4216-4220.

22. Subramani, S., Mulligan, R. & Berg, P. (1981) Mol. Cell. Biol. 1,854-864.

23. Argraves, W. S., Suzuki, S., Arai, H., Thompson, K., Piersch-bacher, M. D. & Ruoslahti, E. (1987) J. Cell Biol. 105, 1183-1190.

24. Aruffo, A., Stamenkovic, I., Melnick, M., Underhill, C. B. &Seed, B. (1990) Cell 61, 1303-1313.

25. Horvath, E., Edwards, A. M., Bell, J. C. & Braun, P. E. (1989) J.Neurosci. Res. 24, 398-401.

26. Perides, G., Rahemtulla, F., Lane, W. S., Asher, R. A. & Big-nami, A. (1992) J. Biol. Chem. 267, 23883-23887.

27. Bignami, A., Perides, G. & Rahemtulla, F. (1993)J. Neurosci. Res.34, 97-106.

28. Yamagata, M., Shinomura, T. & Kimata, K. (1993) Anat. Em-bryol. 187, 433-444.

29. Fuss, B., Wintergeist, E.-S., Bartsch, U. & Schachner, M. (1993)J. Cell Bio. 120, 1237-1249.

30. Bignami, A. & Dahl, D. (1986) Proc. Natl. Acad. Sci. USA 83,3518-3522.

31. Vercelli, D., Helm, B., Marsh, P., Padlan, E., Geha, R. S. &Gould, H. (1989) Nature (London) 338, 649-651.

32. Hoffman, S. & Edelman, G. M. (1987) Proc. Natl. Acad. Sci. USA84, 2523-2527.

33. Hoffman, S., Crossin, K. L. & Edelman, G. M. (1988)J. Cell Biol.106, 519-532.

34. Faissner, A. & Steindler, D. (1995) Glia 13, 233-254.

Proc. Natl. Acad. Sci. USA 92 (1995)

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0