Mapping of a Defined Neurocan Binding Site to Distinct Domains of Tenascin-C*

Neurocan is a member of the aggrecan family of proteoglycans which are characterized by NH2-terminal domains binding hyaluronan, and COOH-terminal domains containing C-type lectin-like modules. To detect and enhance the affinity for complementary ligands of neurocan, the COOH-terminal neurocan domain was fused with the NH2-terminal region of tenascin-C, which contains the hexamerization domain of this extracellular matrix glycoprotein. The fusion protein was designed to contain the last downstream glycosaminoglycan attachment site and was expressed as a proteoglycan. In ligand overlay blots carried out with brain extracts, it recognized tenascin-C. The interaction was abolished by the addition of EDTA, or TNfn4,5, a bacterially expressed tenascin-C fragment comprising the fourth and fifth fibronectin type III module. The fusion protein directly reacted with this fragment in ligand blot and enzyme-linked immunosorbent assay procedures. Both tenascin-C and TNfn4,5 were retained on Sepharose 4B-linked carboxyl-terminal neurocan domains, which in BIAcore binding studies yielded aK D value of 17 nm for purified tenascin-C. We conclude that a divalent cation-dependent interaction between the COOH-terminal domain of neurocan and those fibronectin type III repeats is substantially involved in the binding of neurocan to tenascin-C.

Increasing evidence suggests that the formation of central nervous system structures is, at least in part, regulated by specific interactions of neural cells with their neighbors and the pericellular microenvironment. Both adhesive and inhibitory interactions seem to be of importance, and various gene families have been implicated in these events (1). In this context, an increasing number of extracellular matrix glycoproteins and proteoglycans has been described which seem involved in the control of neuron migration and axon growth and guidance (2)(3)(4)(5). Among these, neurocan has been identified as a member of the aggrecan family of chondroitin sulfate proteoglycans (6,7). The constituents of this family are characterized by an NH 2 -terminal globular hyaluronan-binding domain and a COOH-terminal globular domain with a C-type lectin-like motif. They differ, however, considerably in their central regions (8). Aggrecan, a large proteoglycan primarily expressed in cartilage forms large aggregates with hyaluronan through its NH 2 -terminal region. This interaction is stabilized by link protein, which is homologous to the NH 2 -terminal aggrecan domain and induces a dense association of the proteoglycan along the hyaluronan strands (9). The copurification of link protein with neurocan (6,10) and other biochemical and electron microscopical studies (11,12) suggest that lamp-brush-like aggregates, comparable to those which have been observed in cartilage, might also be formed in the central nervous system. Neurocan was found to interact with cell adhesion molecules of the Ig superfamily such as L1 and N-CAM and to modify their binding properties (13,14). Furthermore, neurocan might be involved in the organization of the extracellular environment by interaction with the extracellular matrix glycoprotein tenascin-C, which is up-regulated during neural development and influences both neuronal and glial cell behavior in various ways (15)(16)(17).
Hyaluronan/proteoglycan aggregates would appear as macromolecular structures with multiple COOH-terminal domains exposed at their surface. Structures with multimerized C-type lectin domains can actually be found quite frequently, probably to potentiate the weak adhesive forces of singular domains (for review, see Ref. 18). Therefore, attempts to identify potential ligands by biochemical approaches should rely on the combination of more than one lectin-type domain in detection probes. For example, IgG-Fc fusion proteins containing dimers of the NH 2 -terminal extracellular domains of selectins have proven useful for the isolation of complementary ligands (19,20). These NH 2 -terminal domains are composed of the same motifs as those contained in the COOH-terminal sequence typical for the aggrecan family, although in a different arrangement. In a recent study, an IgG-Fc fusion protein encompassing the complete COOH-terminal homology domain of PG-M/versican was used to detect proteins co-migrating with the extracellular matrix protein tenascin-R in overlay blots (21). The COOHterminal domains of PG-M/versican and neurocan display analogous structures composed of two EGF 1 -like, one lectin-like, and one complement regulatory-like (or sushi-) motif (7,22). Despite this overall resemblance, two potentially significant differences between these molecules can be observed. Neurocan exhibits a functional glycosaminoglycan attachment site * This work was supported in part by German Research Council Grants DFG, Ra 544/3-1 (to U. R.) and SFB 317/A2 (to A. F.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  closely preceding the first EGF-like domain by a few amino acids (7,23), whereas all known versican isoforms lack such an attachment motif at this place. Furthermore, neurocan contains two additional cysteines, one in the second EGF motif and one as the COOH-terminal amino acid, which might facilitate a COOH-terminal loop structure. Because it is not possible to conserve the latter structural feature, which is unique for neurocan among the proteoglycans of the aggrecan family, in conventional Ig-Fc fusion proteins, a different strategy of multimerization was developed. Taking advantage of the hexameric structure of tenascin-C under native conditions (24), multimers of COOH-terminal neurocan parts were generated by fusing the corresponding sequence of the neurocan with the NH 2terminal sequence of tenascin-C. The resulting fusion proteoglycan was named TENCAN and used as a tool to uncover specific ligands. We show here that TENCAN prominently interacts with the extracellular matrix glycoprotein tenascin-C and map mutual binding regions to the carboxyl terminus of neurocan and distinct domains of the glycoprotein, namely to the fibronectin type III repeats 4 and 5 (TNfn4,5) of tenascin-C.
Construction of Expression Vectors, Transfections, Cell Culture, and Protein Purification-A description of the constructs for neurocan fragments 359H, 773M, L639, D925, and T950 and their purification is given elsewhere (23). For the construction of the tenascin-C/neurocan fusion protein (TENCAN), the NH 2 -terminal part of the tenascin-C sequence was fused at aspartic acid residue 733 (32) with the COOHterminal part of the rat neurocan sequence at aspartic acid residue 925 (7) using the endogenous tenascin-C cDNA BamHI site, and a BglII site, artificially introduced into the neurocan sequence by polymerase chain reaction. The mouse tenascin-C constructs TN200 and TN250 represent the short splice variant containing 8 FNIII repeats and the long splice variant containing 13 FNIII repeats (32), respectively. Positive clones of human embryonic kidney cells (293, American Type Culture Collection) transfected with these construct in the pRC/CMV vector (Invitrogen) were identified by SDS-PAGE and maintained as described (23). The construction of the vectors, the expression and the purification of the tenascin-C fragments expressed in Escherichia coli has been reported (33).
Analytical Methods-SDS-PAGE was performed on 10% slab gels (34) and stained with Coomassie Blue (Serva, Heidelberg, Germany) according to standard protocols. Protein concentrations were determined with the Micro BCA reagent (Pierce) according to the manufacturer's protocol. Digestion with protease-free chondroitinase ABC (Seikagaku, Tokyo) was carried out for 1 h at 37°C in 100 mM Tris/HCl, pH 8.0, 30 mM sodium acetate using 0.5 milliunit of enzyme/g of proteoglycan. Overlay blots were performed by transfer of proteins separated by SDS-PAGE to supported nitrocellulose (Bio-Rad, Hercules, CA) in Tris glycine buffer containing 10% methanol for 1 h with 100 V with the Bio-Rad mini-gel system. Membranes were blocked for 30 min at room temperature with TBSTCM buffer (10 mM Tris/HCl, pH 8, 150 mM NaCl, 0.05% (w/v) Tween 20, 2 mM CaCl 2 , and 2 mM MgCl 2 ) containing 1% (w/v) BSA. For the incubation with TENCAN at 4°C overnight, mixtures of blocking buffer and TENCAN-containing 293 cell supernatant replenished with at least 50% (w/v) blocking buffer were used. All washes, generally three after each incubation, were performed with TBSTCM buffer. Incubations with the monoclonal anti-tenascin-C J1/tn3 and the alkaline phosphatase-conjugated anti-rat antibodies were carried out in blocking buffer. Blots were finally developed in 100 mM Tris/HCl, pH 9.5, 100 mM NaCl, and 50 mM MgCl 2 containing 4.5 l of nitro blue tetrazolium and 3.5 l of 5-bromo-4-chloro-3-indoyl phosphate/ml. Blotting of proteins for NH 2 -terminal amino acid sequencing was carried out in CAPS buffer as described (7). Sequencing was carried out with an ABI sequencer model 492 according to the manufacturer's instructions.
Affinity Chromatograpy-The COOH-terminal neurocan fragment T950 or BSA were separately coupled to about 2 ml of CNBr-activated Sepharose each according to the manufacturer's protocol. The supernatants of the 293 cells transfected with the TN200 and TN250 constructs, whether analyzed separately or combined, were mixed with an equal volume of TBSCM buffer before incubation with the affinity matrix. Aliquots of the supernatants were supplemented with 0.1% Triton X-100 and precipitated with trichloroacetic acid (10% w/v). After washing the matrices with TBSCM, the bound proteins were eluted with TBS containing 20 mM EDTA and 1 M NaCl. Aliquots of the first two 3-ml fractions were precipitated with trichloroacetic acid as described above. Samples of each fraction were separated on a 7.5% SDS-PAGE gel, blotted on nitrocellulose, incubated with J1/tn3, and developed with alkaline phosphatase-conjugated secondary antibody, as described above. The same affinity columns were used for binding studies with the bacterially expressed tenascin-C peptides. In general, 75 g of these tenascin-C fragments were incubated overnight with 2 ml of affinity matrix in a total volume of 4 ml of TBSTCM. The slurry was transferred to a column and the resin was washed with 5 volumes of TBSTCM (fractions 2-6), 3 volumes of TBSTCM with 25 mM EDTA (fractions 7-9), and 3 volumes of TBSTCM containing 25 mM EDTA and 1 M NaCl (fractions 10 -12). 500-l aliquots were precipitated with trichloroacetic acid, separated on a 12% SDS-PAGE, and stained in the gel with Coomassie Blue. In one experiment (Fig. 10G), the fragments were incubated overnight with 2 ml of T950-Sepharose in TBSTCM containing 25 mM EDTA, washed with 8 volumes of the same buffer and 3 volumes of TBSTCM containing 25 mM EDTA and 1 M NaCl.
Ligand Overlay Blot and Solid Phase ELISAs with TENCAN and Recombinant TN-C Domains-2.5 g of recombinant TN-C domains were separated on a reducing 17.5% SDS gel and transferred to a nitrocellulose membrane (BAB5, Schleicher and Schü ll, Dassel, Germany) by semidry blotting. TBSTCM buffer was used as basic buffer for all solutions. Unspecific binding sites on the membrane were blocked for 90 min at room temperature with 1% (w/v) blocking solution (Boehringer Mannheim, Federal Republic of Germany) in TBSTCM, and the filters were incubated overnight at 4°C with TENCAN (diluted 1:2 in blocking buffer). Unbound TENCAN was removed with three TBSTCM washing steps of 10 min each. Binding of TENCAN to recombinant TN-C domains was detected with the mAb J1/tn3, peroxidase-derivatized goat anti-rat antibodies and ECL reagent. To assess the role of divalent cations, incubation was carried out in TBST (TBSTCM without divalent cations), including 20 mM EDTA, which served as basic buffer through all steps of the procedure. For the solid phase ELISA, polyvinylchloride plates were coated with recombinant TN-C domains (1 g/ml in 0.1 M NaHCO 3 , pH 8.0, 100 l/well) overnight at 4°C, washed 3 times with NaHCO 3 , and blocked with 3% (w/v) BSA in TBSTCM for 1.5 h at 37°C. 100 l of TENCAN (1:2 diluted in blocking buffer) were added to each well and incubated for 2 h at 37°C. After three washing steps with TBSTCM, bound TENCAN was detected with J1/tn3 and horseradish peroxidase-derivatized secondary antibodies. Color reaction was developed with ABTS.
Quantitation of Affinities of Neurocan Interactions with Tenascin-C with the BIAcore System-The BIAcore system (BIAcore, Uppsala, Sweden) has been described in detail (35). Experiments were performed and affinities were calculated in essentially the same way as described earlier (36), in part, with the identical covalently immobilized ligands.

Construction of Tenascin-Neurocan Fusion Proteins-Mouse
tenascin-C monomers consist of a NH 2 -terminal heptad/crosslinking domain, followed by 14 1 ⁄2 EGF-like repeats, 8 -13 FNIII repeats, and a COOH-terminal fibrinogen-like globule (32). Specific functions in cell recognition and adhesion to other molecules have been mainly attributed to FNIII repeats and the COOH-terminal knob (16,33). In the fusion protein all but the first of these repeats and the COOH-terminal knob were removed and substituted by the COOH-terminal part of neurocan. The remaining first FNIII repeat and the EGF-like repeats were expected to maintain a sufficient degree of flexibility for the fused neurocan domains to interact with multiple ligands (Fig. 1). The segment of the neurocan sequence incorporated into the fusion protein (fragment D925) had previously been expressed in eucaryotic cells where it was modified with chondroitin sulfate chains (23). The analysis of the proteins precipitated from the supernatants of cells transfected with the fusion construct also revealed that the major fraction of the resulting proteins was derivatized with chondroitin sulfate (not shown). To emphasize this proteoglycan characteristic, the fusion protein was named TENCAN.
Detection of Proteins in Overlay Assays-Protein fractions obtained from postnatal rat brain were separated by SDS-PAGE, blotted on nitrocellulose, and incubated with the conditioned supernatant of 293 cells secreting TENCAN. For the detection of TENCAN molecules bound to blotted proteins, the monoclonal rat anti-mouse tenascin-C antibody J1/tn3, which recognizes an epitope in the EGF-like repeat region of mouse tenascin-C, was used. This mAb, which did not recognize rat tenascin-C, was expected to interfere only marginally with the ability of the neurocan-derived part of TENCAN to interact with potential ligands. The supernatant of untransfected 293 cells did not yield specific signals in these blotting experiments (Fig. 2). In contrast, the supernatant of 293 cells transfected with the short splice variant of tenascin-C (TN200) revealed a different pattern of rat brain proteins than the supernatant of 293 cells transfected with TENCAN (Fig. 3). Thus, TN200 containing supernatant did not only serve as a control supernatant, but also as a specific detection aid for potential ligands of tenascin-C. The 250 and 750 mM NaCl fractions of rat central nervous system proteins which were used in this study were prepared by extracting the supernatant of detergent-free brain homogenates with DEAE-Sephacel in TBS (containing 150 mM NaCl), and eluting the ion exchange resin with salt concentrations of 250 and 750 mM NaCl, respectively. Subsequently, the 750 mM NaCl fraction was depleted of intrinsic neurocan by 1D1-affinity chromatography. After SDS-PAGE of these protein fractions under reducing conditions and blotting onto nitrocellulose, TENCAN was recognized in both fractions proteins with an apparent molecular mass of about 250 kDa (Fig.  3, C and D). In the 250 mM NaCl fraction, a protein band at about 160 kDa was revealed in addition to the 250-kDa component (Fig. 3, D). In the 750 mM fraction, a diffuse band at about 190 kDa could be detected only after chondroitinase treatment, whereas both samples displayed minor components of 80 and 100 kDa (Fig. 3, C). The 100-kDa protein was also weakly detectable in the 250 mM fraction. Amido Black staining revealed that this protein represented one of the major constituents of the 750 mM NaCl fraction which contained few distinct proteins. Thus, by transfer of this fraction to a polyvinyldifluoride membrane it was possible to determine the NH 2 -terminal amino acid sequence of the 100-kDa protein without further purification. The obtained sequence VKLAKAGKTHGE(S)KK-MAP identified the 100-kDa band as nucleolin (37).
The most likely candidate molecule for the material detected at about 250 kDa, the upper part of which appeared to be sensitive to chondroitinase ABC treatment, was tenascin-C. This hexameric glycoprotein has been shown to be a ligand of neurocan and a component of soluble postnatal rat brain proteoglycan fractions (15). Furthermore, it has previously been observed in chicken (38) and also in rat brain (15), that subfractions of tenascin-C might be modified by the attachment of chondroitin sulfate chains. Thus, an analogous set of protein samples (Fig. 3, E and F) was developed with polyclonal tenascin-C antibodies, which do not show any reactivity in Western blots with tissue samples of tenascin-C knock-out mice (29). This serum reacted strongest with bands that comigrated with those recognized by TENCAN at about 250 kDa, indicating that these indeed represent rat tenascin-C.
The supernatant of 293 cells transfected with TN200 revealed a distinct set of bands. Material at 350 -400 kDa was detected in the 750 mM NaCl eluate upon treatment of this fraction with chondroitinase ABC and absent under control  (23). TN200 and TN250 were constructed by joining tenascin-C cDNA fragments without and with five alternatively spliced FNIII repeats (32) to full-length tenascin-C molecules. Structural elements used for the fusion protein TENCAN were the central nodule, all EGF-like repeats, the first FNIII repeat, and presumably two of the seven ␤-sheats of the second FNIII fold of mouse tenascin-C. These were linked to the last 25 amino acids of the central region and the COOH-terminal globular domain which consists of two EGF repeats, a C-type lectin-like module, a sushi motif, and the final COOHterminal sequence of rat neurocan. conditions, indicating that binding occurred to a chondroitin sulfate proteoglycan core protein (Fig. 3, A). Tenascin-C binds via its COOH-terminal fibrinogen-like domain to oligosaccharide structures N-linked to phosphacan (39,40). Since phosphacan is a major component of the rat brain proteoglycan fraction (6), the core protein of this proteoglycan appears to be the most likely candidate molecule for the 400-kDa band. The major bands recognized by TN200 in the 250 mM NaCl fraction were visible at 220, 135, and 115 kDa (Fig. 3, B).
Characterization of TENCAN and TN200 Ligands with Defined Antisera-To evaluate whether proteins in the 250 mM NaCl fraction identified by TENCAN or TN200 might represent fibronectin, tenascin-R, or fibulin-2, analogous protein samples were electrophoresed, blotted in parallel, and probed with the respective antisera (Fig. 4A). The result of the Western blot renders it likely that the 220-kDa band recognized by TN200 represents fibronectin. The polyclonal anti-tenascin-R antibodies resulted in very slowly developing signals and, for this reason, the 250 mM NaCl fraction was probed with a mixture of two mAbs to tenascin-R in a separate experiment. These antibodies reacted with protein bands corresponding to the material migrating at around 180 kDa and faintly stained with the antiserum, but not with the more prominent 160-kDa component also recognized by TENCAN (results not shown). When purified chicken tenascin-R was transferred to nitrocellulose and probed with TENCAN, only a very faint staining was apparent (Fig. 4B). In addition to its consistently observed interaction with the 250-kDa band, in a particular rat brain protein sample TENCAN clearly recognized also a 200-kDa component, presumably a tenascin-C isoform which lacks the alternatively spliced FNIII domains (Fig. 4B). Anti-brevican antibodies and a mixture of two mAbs specific for the cell adhesion molecule F11 failed to develop any protein band recognized by TENCAN or TN200 (results not shown). Thus, the identities of the potential tenascin-C-binding proteins of 115 and 135 kDa, which might be a rat homologue of CALEB (41), remain presently undefined.
TENCAN Binding Depends on Divalent Cations-To evaluate whether the recognition of the protein bands contained in the 250 mM NaCl fraction was dependent on divalent cations or glycosaminoglycan chains, TENCAN binding studies were performed in the presence of potentially inhibitory substances (Fig. 5). The interaction with the 250-kDa band was inhibited completely in the presence of 12.5 mM EDTA. No significant inhibition could be observed in the presence of chondroitin sulfate. The presence of chondroitinase ABC during the incubation with the fusion protein did not negatively affect the detection of the 250-kDa band and, surprisingly, enhanced considerably the overall staining of the lane. These results demonstrate that the recognition of the 250-kDa band by TEN-CAN depends on the presence of divalent cations, which could indicate a critical contribution of the C-type lectin-like domain present in the COOH-terminal neurocan fragment. In contrast, the recognition of the 160-kDa band appears to depend strongly on chondroitin sulfate, since the presence of these glycosaminoglycans or of chondroitinase ABC exerted inhibitory effects.
Interaction of TENCAN with Tenascin-C Domains in Overlay Ligand-Binding Assays-To further examine which particular domains of tenascin-C are participating in the interaction with neurocan, several recombinant tenascin-C fragments spanning the FNIII repeat region were scanned for interactions using solid phase ELISAs (Fig. 6) and the ligand overlay assay (Fig.  7) with TENCAN as soluble binding partner. These FNIII domains embody several functional sites for neural and other cell types, as documented previously (17,33), and do not contain disulfide bridges which might cause folding problems in bacterial expression systems. In the solid phase ELISA a prominent binding to TNfn4,5, a recombinant protein representing the 4th and 5th FNIII repeat of tenascin-C, could be observed (Fig. 6). This interaction could not be observed in the absence of divalent cations (results not shown). An EDTA-sensitive binding of TENCAN to the blotted TNfn4,5 fragment could also be observed in the ligand overlay assay (Fig. 7). Similar to the binding of TENCAN to the 250-kDa band (Fig. 5), this interaction could not be abolished by the presence of chondroitin sulfate or chondroitinase ABC (results not shown). When four of the recombinant tenascin-C fragments, TNfn1-3, TNfn4,5, TNfnA1,2,4, and TNfn7,8, were compared for their ability to compete with immobilized ligands in a TENCAN overlay assay, only TNfn4,5 could efficiently inhibit the binding of TENCAN to the 250-kDa band of the 250 mM NaCl fraction (Fig. 8).
Affinity Chromatograpy of Recombinant Tenascin-C Isoforms and Tenascin-C Fragments on the Individually Expressed COOH-terminal Neurocan Domain-The structure of the TEN-CAN fusion protein would suggest that the interaction with tenascin-C occurs via the COOH-terminal globule of the proteoglycan. To evaluate this possibility directly, the recombinantly expressed neurocan fragment T950 was immobilized to CNBr-activated Sepharose and used for affinity chromatography studies. It has been shown previously that the recombinant fragment T950 displays the expected globular shape of the COOH terminus of neurocan and is not modified by addition of chondroitin sulfate side chains (23). The T950 affinity matrix was incubated with the supernatant of 293 cells transfected with the large and the small isoform of mouse tenascin-C, TN200 and TN250. Both variants of tenascin-C were applied separately, or as a mixture containing comparable amounts of both proteins. In all cases, these molecules were depleted from the supernatant by the T950 affinity matrix (Fig. 9, and results not shown). No binding of tenascin-C isoforms could be observed when the 293 cell supernatants were incubated with CNBr-activated Sepharose derivatized with BSA (Fig. 9). Binding of tenascin-C to the T950 affinity matrix was confirmed by eluting the matrix with buffer containing 1 M NaCl and 20 mM EDTA. Having established that the T950 column interacts with intact tenascin-C, recombinant tenascin-C fragments were incubated with the affinity matrix and analyzed for their elution behavior. The four analyzed fragments which contained FNIII modules, TNfn1,2,3, TNfnA1,2,4, TNfn4,5, and TNfn7,8, showed a differential retention to the T950 matrix. Only minute amounts of fragments TNfn1-3, TNfnA1,2,4, and TNfn7,8 were eluted from this matrix by addition of EDTA and additional NaCl to the buffer (Fig. 10A, C, and D). By comparison, tenascin-C fragment TNfn4,5 was considerably retarded, but efficiently eluted upon exposure to chelating agents (Fig. 10E). No retardation of this fragment could be observed when a BSA affinity matrix was used or when EDTA was included from the beginning (Fig. 10, F and G). Thus, in this approach TNfn4,5 exhibited the most stringent interaction, consistent with the results of the overlay and ELISA assays.
BIAcore Analysis of the Tenascin-C Neurocan Interaction-The ligand overlay and affinity chromatography approaches allowed detection of an interaction between the COOH-terminal neurocan globule and the pair of FNIII domains TNfn4,5 of tenascin-C. To establish the affinities of interactions between neurocan, tenascin-C, and derived fragments, a direct analysis was performed with the BIAcore system. Attempts to immobilize mouse brain tenascin-C on the carboxymethylated dextran matrix of the BIAcore chip were unsuccessful, as well as attempts to immobilize proteoglycans without prior treatment with chondroitinase ABC. Tenascin-C used as a soluble ligand showed significant binding to the core protein of rat brainderived neurocan in the presence of divalent cations (Fig. 11A, Table I). The K D value of 3 nM derived from the observed association and dissociation kinetics of tenascin-C was similar to the K D value obtained previously in a solid phase radioligand binding assay (15). Additional tenascin-C binding studies were performed with covalently immobilized neurocan and derived fragments which had recombinantly been expressed in the eucaryotic cell line 293 (23,36). Characteristic association profiles were observed with recombinant neurocan and the fragments T950 and L639, which encompass the COOH-terminal globule and the COOH-terminal half of the proteoglycan, respectively (Fig. 11A, Table I). Different from these, the two fragments lacking the COOH-terminal globular domain displayed rather uncharacteristic plasmon resonance profiles, indicating a dissociation of ligand already during the course of its application (Fig. 11B). When recombinant tenascin-C fragments were analyzed for their interaction with the core protein of rat brain-derived neurocan, no significant plasmon resonance signals could be obtained. When fragment T950 was immobilized to the carboxymethylated dextran matrix of the BIAcore chip a significant retention of not only fragment TNfn4,5 could be observed, but notable binding could also be documented for TNfn1,2,3 and TNfn7,8, while fragment TN-fnA1,2,4 remained inert. A calculation of K D values established almost equal affinities with K D values of about 700 nM for fragments TNfn1,2,3 and TNfn4,5, and 7 M, 1 order of magnitude weaker, for TNfn7,8 (Table I). DISCUSSION Design and Properties of the TENCAN Probe-To identify presumed neural extracellular matrix ligands of the carboxylterminal domain of neurocan, a probe was designed which should display optimal affinity for potential ligands by containing multiple copies of this neurocan domain in a single molecule. To achieve this, the carboxyl-terminal end of neurocan was fused to the amino-terminal region of tenascin-C. The resulting chimeric molecule was named TENCAN. We have used TENCAN as probe in various solid phase ligand binding assays performed with detergent-free neural tissue extracts or purified proteins to identify interaction partners of the carboxyl terminus of neurocan. Neurocan bears chondroitin sulfate chains in neural tissues which might be involved in ligand binding. Hence this carbohydrate modification was retainedin the TENCAN fusion protein by conserving the last glycosaminoglycan attachment site of the central region of the proteoglycan. Because of the strong polyanionic character of the glycosaminoglycan complement, the resulting TENCAN proteoglycan could be expected to develop charge-dependent interactions with proteins which exhibit a strong basic character, but do not necessarily constitute biological ligands of neurocan. Considering this possibility, surprisingly few proteins were detected with the TENCAN probe in the 250 and 750 mM NaCl DEAE fractions of the soluble rat brain proteins. One of these molecules could be identified by NH 2 -terminal amino acid sequencing as nucleolin, an abundant intracellular molecule with clusters of acidic and basic amino acids (37). These features might also be responsible for the emergence of nucleolin in previous ligand overlay approaches carried out with other probes (42,43). Other proteins recognized by TENCAN were observed at molecular sizes of 160 and 250 kDa. Earlier reports have already pointed out that neurocan interacts with tenascin-C, and that both molecules are expressed in the central nervous system with partly overlapping local and developmental expression patterns (15). In agreement with these data, the 250-kDa protein was confirmed as a high molecular mass splice variant of tenascin-C (see below). A 160-kDa protein, which has first been suspected to represent tenascin-R, another member of the tenascin gene family expressed in the central nervous system (5), apparently interacts with chondroitin sulfate chains. Protein bands of the same apparent molecular mass could also be revealed with TENCAN in analogous 250 mM NaCl DEAE fractions of thymus, lung, and kidney tissue of 7-day-old rats. 2 Interest in chondroitin sulfate-binding proteins might emerge from the recent observation that chondroitin sulfate chains exhibit neuronal survival promoting activities (44).
The Carboxyl Terminus of Neurocan Interacts with Tenascin-C-The 250-kDa band which was developed with the TENCAN construct in ligand overlay assays performed with the 250 and 750 mM fractions of detergent-free brain extracts were attributed to tenascin-C because: (i) specific polyclonal antibodies to 2 U. Rauch, unpublished observation. FIG. 9. T950-affinity chromatography of TN200 and TN250. A mixture of TN200 and TN250 supernatants diluted with an equal volume of TBSCM was incubated with BSA-Sepharose (A) and the unbound material was subsequently incubated with T950-Sepharose (B). The supernatants before (C) and after (D) incubation with the respective resins were analyzed for the presence of recombinant tenascin-C by immunoblotting using antibody J1/tn3 (clone 630) and APderivatized secondary antibodies. After washing of the affinity resins with TBSCM, bound proteins were eluted with TBS containing 20 mM EDTA and 1 M NaCl. Aliquots of the first two fractions (E1 and E2) were analyzed for the presence of the recombinant tenascin-C molecules as described above. tenascin-C recognize the same band patterns in these preparations, (ii) TENCAN reacts with defined bacterially expressed tenascin-C domains, and (iii) an interaction of the carboxylterminal domain of neurocan with intact tenascin-C and derived recombinant proteins could also be observed in other assay systems. The interaction with tenascin-C proved resistant to the addition of chondroitin sulfate, in contrast to the binding observed with the other proteins, e.g. nucleolin or the elusive 160-kDa band. The overall organization of the TEN-CAN fusion protein would suggest that the carboxyl terminus of neurocan is exposed and harbors the interaction site with tenascin-C. This assertion is supported by the finding that neurocan fragment T950 which encompasses the carboxyl-terminal domain of the proteoglycan by itself reacts with intact tenascin-C purified from mouse brain, with the TN-200 and TN-250 variants of mouse tenascin-C expressed in 293 cells and with tenascin-C domains expressed in bacteria, as shown by affinity chromatography and BIAcore analysis with immobilized T950 protein. However, the higher affinity obtained with the full-length neurocan core proteins indicate that the COOH-terminal domain might not be the only part of the neurocan molecule which is able to interact with tenascin-C. The interaction of neurocan with tenascin-C, as well as the binding of TENCAN to tenascin-C, is divalent cation-dependent and can be abolished by addition of the chelator EDTA. The divalent cation dependence of this interaction suggests that the C-type lectin-like module in the COOH-terminal neurocan domain could be critically involved. However, since the seques-tration of divalent cations with EDTA might be able to induce a structural alteration of the entire COOH-terminal neurocan domain, it remains to be determined whether this module might be involved directly or indirectly. C-type lectin-like modules are best known for their interactions with carbohydrate structures (18), but specific interactions with protein structures have also been reported (45). The results obtained in the interaction studies with the tenascin-C fragments indicate that this might also be the case for the C-type lectin-like domain of neurocan. The ability of the amino terminus of neurocan to bind to hyaluronic acid has already been shown (23). Assuming that in brain hyaluronan/neurocan aggregates form in a way reminiscent of the lamp-brush-like structures known from cartilage, the COOH-terminal domains of the proteoglycan would be the most conveniently accessible components for interactions with other large molecules. As determined by rotary shadowing, the length of neurocan is comparable to the size of a tenascin-C monomer (23,46,47). Thus, the diameter of hypothesized hyaluronic acid-neurocan complexes would be similar to the size of tenascin-C hexamers. It is tempting to imagine that this arrangement would permit a simultaneous binding of several arms of tenascin-C to multiple neurocan COOH-terminal domains accessible on the surface of lampbrush-like aggregates, thereby stabilizing or even cross-linking resulting superstructures.
Mapping of Neurocan Binding to TNfn4,5-To determine the domains of tenascin-C involved in the interaction with the carboxyl terminus of neurocan, a library of bacterially expressed tenascin-C FNIII repeats (33) was scanned with the TENCAN probe by ligand overlay and ELISA. In these experiments, the pair of FNIII domains 4 and 5 (TNfn4,5) which is part of the constitutive basic variant of tenascin-C was identified as the prominent binding partner. The interaction of TNfn4,5 could also be demonstrated for the carboxyl-terminal neurocan domain T950 in affinity chromatography. The importance of TNfn4,5 is underlined by the finding that the soluble recombinant protein is sufficient to compete the binding of the TENCAN fusion to intact tenascin-C recovered in brain extracts. Furthermore, the binding to TNfn4,5 is divalent-cation dependent, as predicted from the studies with the integral proteins. A critical contribution of fragment TNfn4,5 for the interaction with neurocan would be in general accordance with earlier attempts to map the binding site of chicken tenascin-C to a "cytotactin-binding proteoglycan," which is likely to represent the chicken homologue of neurocan. In these experiments, FabЈ fragments of antibodies raised against a 35-kDa CNBr peptide were able to block efficiently the interaction of tenascin-C not only with the proteoglycan, but also with fibronectin (48). The 35-kDa CNBr peptide turned out to be derived from the alternative FNIII repeats B and D (Vb and Vc) and the constitutive FNIII repeat 6 (49). Therefore, the TNfn4,5 fragment would be in even closer vicinity to the antibody protected fragment than the determined major binding site of fibronectin, the third FNIII repeat (50). In a more recent investigation, deletion mutants of tenascin-C expressed in mammalian cells were coated on microwell plates and tested for their ability to FIG. 11. BIAcore analysis of tenascin-C interactions with neurocan and neurocan fragments. Neurocan and neurocan fragments (for a sketch, see Fig. 1) containing (A) or lacking (B) the COOHterminal globular domain were immobilized on the carboxymethylated dextran matrix. The observed plasmon resonance signal is proportional to the binding of mouse brain tenascin-C to matrices modified with tissue-derived neurocan (NC), recombinant neurocan (rNC), the COOHterminal globular domain (T950), the COOH-terminal half of the molecule (L639), the NH 2 -terminal globular domain alone (359H) or in combination with the first two thirds of the central region (773M). retain radioiodinated rat brain-derived neurocan and phosphacan (40). Whereas all reported experimental results indicated an essential interaction of phosphacan with the COOH-terminal globular domain of tenascin-C, the results obtained with neurocan were rather indicative of a more complex interaction. Interestingly, although it was shown that neurocan binds to coated recombinant tenascin-C molecules which missed the FNIII domains, it was stated that an inhibition of neurocan binding to integral tenascin-C was only achieved with mutant tenascin-C molecules containing the FNIII domains. In those investigations, it was observed that neurocan binding to fulllength native and recombinant tenascin-C was calcium dependent, whereas the absence of divalent cations had only minor and variable effects on neurocan interactions with tenascin-C deletion variants. It was also noted that binding of neurocan to full-length tenascin-C was not linked to a proper arrangement of the disulfide bonds in tenascin-C (40). These results imply a divalent cation-dependent and dithiothreitol treatment-insensitive interaction of neurocan with tenascin-C domains. In the absence of the tenascin-C FNIII domains a divalent cationindependent interaction emerged, which also proved insensitive to the reduction of the tenascin-C disulfide bonds. This latter observation is notable, because the COOH-terminal fibrinogen-like globe of tenascin-C contains four cysteine residues. In agreement with these findings, we also observed moderate binding of TENCAN to the COOH-terminal globular domain of tenascin-C in the ELISA assay. 3 Yet, because it could not be taken for granted that the disulfide bonds are correctly arranged in our bacterial expression product, we concentrated on the FNIII domains.
Conclusions-In conclusion, the mapping studies suggest that the carboxyl terminus of neurocan interacts with the pair of tenascin-C FNIII domains TNfn4,5 in a divalent cation-dependent manner. Additional interactions between other parts of these molecules, especially the fibrinogen-like globe of tenascin-C, are likely to exist. In our or other (40) experiments with full-length tenascin-C molecules, however, no indication for an inhibitory effect of the alternatively spliced region could be revealed. Such an effect has been inferred for the interactions of tenascin-C with fibronectin and contactin/F11 (51,52). The differential expression of the alternatively spliced domain might modulate the preferences of tenascin-C for particular ligands in environments where more than one partner would be available. The functional implications of neurocan interactions with tenascin-C are presently not well understood. TNfn4,5 is repulsive for embryonic day 18 hippocampal neurons and some glial cell types, and comprises a stop signal for dorsal root ganglion axons (17,33,53). Thus, formation of neurocan-tenascin-C complexes might modify the functional properties of tenascin-C. Furthermore, the interaction could contribute to the immobilization of tenascin-C in discrete distribution patterns, e.g. in the subplate of the developing cortex (54). The biological consequences of the formation of extracellular matrix superstructures constitute a challenging subject for future studies.