TAG-1/axonin-1 is a high-affinity ligand of neurocan, phosphacan/protein-tyrosine phosphatase-ζ/β, and N-CAM

Proteoglycans appear to play an important role in modulating cell-cell and cell-matrix interactions during nervous tissue histogenesis. The nervous tissue-specific chondroitin sulfate proteoglycans neurocan and phosphacan/protein-tyrosine phosphatase-η/β were found to be high-affinity ligands of the neural cell adhesion molecule TAG-1/axonin-1, with dissociation constants of 0.3 nM and 0.04 nM, respectively. Phosphacan binding was decreased by ∼70% following chondroitinase treatment, whereas binding of neurocan was not affected. The contribution of chondroitin sulfate chains to the binding of neurocan and phosphacan to TAG-1/axonin-1 is therefore the opposite of that previously observed for their binding to two other Ig-superfamily neural cell adhesion molecules, Ng-CAM/L1 and N-CAM. Moreover, whereas phosphacan interactions with certain proteins are mediated at least in part by N-linked oligosaccharides on the proteoglycan, N-deglycosylation of phosphacan had no effect on its binding to TAG-1/axonin-1. In addition to the chondroitin sulfate proteoglycans described above, we have demonstrated that N-CAM is a high-affinity ligand of TAG-1/axonin-1 (Kd ∼1 nM), and specific binding of TAG-1/axonin-1 to tenascin-C was also observed (Kd ∼9 nM). Immunocytochemical studies of embryonic and early postnatal nervous tissue showed an overlapping localization of TAG-1/axonin-1 with all four of these ligands, further supporting the biological significance of their ability to interact in vitro.

We have previously described the specific, high-affinity interactions of neurocan and phosphacan (1) with neural cell adhesion molecules (Ng-CAM/L1 and N-CAM) and the extracellular matrix protein tenascin-C (2)(3)(4). These studies also demonstrated their binding to neurons and that they are potent inhibitors of neuronal and glial adhesion and of neurite growth. Neurocan is synthesized by neurons (5) and is a member of the family of multidomain hyaluronan-binding chondroitin sulfate proteoglycans that also includes aggrecan, versican, and brevican (6), whereas the astroglial proteoglycan phosphacan (7) is an mRNA splice variant that represents the entire extracellular domain of a receptor-type protein-tyrosine phosphatase (RPTP/␤ 1 ; 8 -10) which occurs as a chondroitin sulfate proteoglycan in brain (11,12). The N-terminal carbonic anhydrase-like domain of phosphacan/RPTP/␤ has also been reported to bind to contactin (13). Both neurocan and phosphacan undergo extensive developmental changes in concentration, molecular size, glycosylation, sulfation, and immunocytochemical localization (14 -17). It would appear that through their association with cells and with neural cell adhesion and extracellular matrix proteins, these proteoglycans play an important role in modulating cell-cell and cell-matrix interactions during nervous tissue histogenesis. Because this is a complex process that undoubtedly depends on the relative levels and locations of a variety of cell surface and extracellular matrix proteins at different developmental stages, we have searched for other ligands of neurocan and phosphacan and have identified the neural cell adhesion molecule TAG-1/axonin-1, which we found to also be a high-affinity ligand for N-CAM.
The gene product termed TAG-1 in the rat and axonin-1 in the chicken belongs to a group of nervous tissue glycoproteins of the immunoglobulin (Ig) superfamily. It is present predominantly on the axons of specific nerve fiber tracts that have a restricted distribution pattern during neural development, and TAG-1/axonin-1 can serve as a substrate for neurite outgrowth (for references, see Ref. 18). The Ig superfamily of cell adhesion molecules can be subdivided into several subgroups (18,19), one of which comprises proteins having six Ig-like domains and four fibronectin type III repeats, but that lack a transmembrane region and may be attached to the membrane by a glycosylphosphatidylinositol anchor. In terms of sequence homology, this group can be further divided into several subclasses (20). The TAG-1/axonin-1 subclass, whose members have been shown to occur in both soluble and membrane-bound forms, includes chicken axonin-1 (21), rat TAG-1 (22), and human TAX-1 (23,24). Axonin-1 and TAG-1 have 75% amino acid sequence identity, and TAX-1 has 91% identity to TAG-1 and 75% to axonin-1.
In the present study we have identified TAG-1/axonin-1 as an additional ligand for neurocan and phosphacan and found that its interactions involve different types of proteoglycan glycosylation from those previously shown to affect the binding of these proteoglycans to Ng-CAM/L1, N-CAM, and tenascin-C. Our studies have also demonstrated that N-CAM is a highaffinity ligand for TAG-1/axonin-1, and that all of the proteins that we have shown to interact in vitro have overlapping localizations in nervous tissue.

EXPERIMENTAL PROCEDURES
Proteins and Antibodies-Neurocan, phosphacan, and phosphacan-KS (previously designated the 1D1, 3F8, and 3H1 proteoglycans, respectively) were isolated from rat brain by ion exchange chromatography, gel filtration, and immunoaffinity chromatography (14). Axonin-1 was isolated from E14 chicken brain membranes or ocular vitreous fluid by immunoaffinity chromatography using the SC2 monoclonal antibody (25) and was generously provided by Drs. Martin Grumet and Takeshi Sakurai (New York University Medical Center), as were chick Ng-CAM and N-CAM isolated as described previously (4), and Nr-CAM that was isolated by immunoaffinity chromatography using the 23A7 monoclonal antibody (26). Human tenascin-C was a kind gift from Dr. Mario Bourdon (La Jolla Institute for Experimental Medicine). Other proteins were obtained from the sources described in Milev et al. (4).
The 3F8 monoclonal antibody (14) and a polyclonal rabbit antiserum to phosphacan (4) have been described previously, as has the 1F6 monoclonal antibody to an N-terminal epitope of neurocan (15). A rabbit antiserum to tenascin-C was obtained from Telios, and an antiserum to N-CAM was provided by Dr. Martin Grumet. The 3.1C12 monoclonal antibody to TAG-1 (27) was obtained from the Developmental Studies Hybridoma Bank (maintained by the Dept. of Pharmacology and Molecular Science, Johns Hopkins University School of Medicine, Baltimore, MD, and the Dept. of Biology, University of Iowa, Iowa City, IA, under contract NO1-HD-6 -2915 from the NICHD, National Institutes of Health).
A polyclonal antiserum specific for neurocan was prepared by immunizing rabbits with a glutathione S-transferase fusion protein expressed in E. coli transformed with the pGEX plasmid containing an insert corresponding to nucleotides 1991-2625 of rat neurocan (6). Because these code for amino acids in the C-terminal portion of the central nonhomologous domain of neurocan, this antiserum would not be expected to recognize other members of the family of hyaluronanbinding chondroitin sulfate proteoglycans (i.e. aggrecan, versican, and brevican), and immunoblots of rat brain chondroitin sulfate proteoglycans demonstrated reactivity only with full-length neurocan and its previously described proteolytic processing products (14,15).
Binding Assays-Binding assays were performed as described previously (4). Briefly, proteins were coated in removable Immulon-4 wells using a concentration of 1-2 g/ml for binding of proteoglycans to axonin-1, Ng-CAM and N-CAM, and 10 g/ml when testing the binding of axonin-1 to various proteins. Binding of labeled protein was measured in 20 mM Tris, pH 7.4, 2 mM CaCl 2 , 2 mM MgCl 2 , 0.02% NaN 3 containing 1 mg/ml heat-treated BSA and either 50 or 150 mM NaCl, following incubation with gentle shaking (45 rpm) for 3-4 h at room temperature. Scatchard plots were generated and dissociation constants were determined using the Macintosh version of the Ligand program (28). Proteins were labeled to a specific activity of 10 18 -10 19 cpm/mol with 125 I by the lactoperoxidase/glucose oxidase method using Enzymobeads (Bio-Rad). Typically, 10 -25 g of protein were labeled per reaction, and free iodine was removed using a PD-10 column (Pharmacia Biotech Inc.). Labeled proteins were eluted in 30 mM Tris-HCl buffer, pH 7.4, containing 150 mM NaCl, 10 mM NaI, and 8 mg/ml heat-treated BSA.
Immunocytochemistry-Immunocytochemistry of Vibratome sections of perfusion-fixed postnatal rat brain using peroxidase-conjugated second antibody was performed as described previously (14). Localization of proteins in embryos was performed by double immunofluorescence of cryostat sections (16) using the 3.1C12 monoclonal antibody to TAG-1 and polyclonal antibodies to neurocan, phosphacan, N-CAM, and tenascin-C, in conjunction with fluorescein-conjugated anti-rabbit IgG and rhodamine-conjugated anti-mouse IgG (Jackson Immunoresearch, West Grove, PA).

TAG-1/Axonin-1 Interactions with
Proteoglycans-There was significant binding of neurocan and phosphacan to axonin-1 isolated from either brain membranes or vitreous fluid, and phosphacan binding to TAG-1/axonin-1 was decreased by ϳ70% following chondroitinase treatment, whereas the bind-ing of neurocan was not affected (Fig. 1). The contribution of chondroitin sulfate chains to the binding of neurocan and phosphacan to TAG-1/axonin-1 is therefore the opposite of that observed for their binding to Ng-CAM/L1 and N-CAM, in which neurocan binding was decreased significantly following chondroitinase treatment whereas this had only a slight effect on the binding or biological effects of phosphacan (2,4). These results suggest that for neurocan and phosphacan, which share at least three neural cell adhesion molecule ligands, glycosylation provides an important means for regulating their specific interactions. It is also of interest that whereas ionic strength has a significant effect on the binding of neurocan and phosphacan to Ng-CAM/L1 and N-CAM (2, 4), binding of both neurocan and phosphacan to vitreous TAG-1/axonin-1 and of phosphacan to membrane TAG-1/axonin-1 was only slightly greater in 50 mM NaCl (data not shown) as compared to binding in 150 mM NaCl (Fig. 1).
The specificity and affinity of the proteoglycan-TAG-1/axonin-1 interactions were also studied. Binding of neurocan and phosphacan was found to be saturable, reversible, and of high affinity, with dissociation constants of 0.3 nM and 0.04 nM, respectively (Figs. 2 and 3). Phosphacan binding to TAG-1/ axonin-1 was almost completely inhibited by neurocan, phosphacan, and the phosphacan core glycoprotein, to the extent of ϳ80% by chondroitinase-treated neurocan, and little or not at all by chondroitin sulfate and fibronectin (Fig. 4). These results suggest that the TAG-1/axonin-1 binding sites for neurocan and phosphacan may be relatively close to one another and are similar to those from earlier studies in which we have found reciprocal inhibition of neurocan and phosphacan binding to Ng-CAM/L1 and N-CAM by the proteoglycans or the phosphacan core glycoprotein. 2 We previously reported that the interactions of phosphacan (but not neurocan) with Ng-CAM/L1, N-CAM, and tenascin-C are mediated by N-linked oligosaccharides on phosphacan (29). However, binding of phosphacan to TAG-1/axonin-1 was not affected by peptide N-glycosidase 2 P. Milev and R. U. Margolis, unpublished results.
FIG. 1. Binding of 125 I-labeled neurocan and phosphacan to axonin-1 prepared from chick brain membranes and ocular vitreous fluid. Axonin-1 was coated in removable wells at a concentration of 1 g/ml, and labeled proteoglycans were used at ϳ140,000 cpm/well. Chondroitinase digestions were performed as described under "Experimental Procedures," and "control" refers to incubation in enzyme buffer alone. The percent bound represents specific binding (total cpm bound minus cpm bound to wells coated with BSA) using binding buffer containing 150 mM NaCl, and all values are means of duplicate determinations Ϯ S.E. treatment to remove N-linked oligosaccharides (data not shown), which is consistent with evidence described above that the chondroitin sulfate chains play an important role in this process.
Overlapping Localization of TAG-1/Axonin-1 with Neurocan and Phosphacan-To determine whether TAG-1/axonin-1 and the proteoglycans are present together at the same sites and developmental periods, which is a necessary requirement for their molecular interactions to be biologically meaningful, we performed immunocytochemical studies on their localization in the embryonic and early postnatal rat central nervous system. Double immunofluorescence studies showed a colocalization of TAG-1 and phosphacan in the dorsal root entry zone of the spinal cord, the trigeminal ganglion, and the cortex at embryonic day 13 (E13), and in the cortex, lateral olfactory tract, optic nerve, and retina at E16 (Figs. 5 and 6). There was also colocalization of TAG-1 and neurocan in the dorsal root ganglion and entry zone, optic nerve, and retina at E13-16 ( Fig. 7), and an overlapping staining pattern for TAG-1, neurocan, and phosphacan was seen in postnatal day 7 rat cerebellum (Fig. 8). These results demonstrate that there are many sites at which TAG-1/axonin-1 and the proteoglycans have the opportunity to interact during central nervous system development, although the distribution of TAG-1/axonin-1 is considerably more restricted than that of other neural cell adhesion molecule ligands of neurocan and phosphacan such as Ng-CAM/L1 and N-CAM.
Phosphacan-KS-A glycosylation variant of phosphacan (designated phosphacan-KS) contains keratan sulfate in addition to chondroitin sulfate chains. Four different forms of phosphacan-KS have been isolated based on their relative affinities for the 3H1 monoclonal antibody to the keratan sulfate chains and partly reflecting developmental changes in glycosylation (7,14). We therefore compared the binding of these four populations of phosphacan-KS to TAG-1/axonin-1, Ng-CAM/L1, N-CAM, and tenascin-C. There was binding of all forms of phosphacan-KS, but only the 0.5 M NaCl eluate from 7-day brain showed a significantly greater degree of binding to these ligands (data not shown). Endo-␤-galactosidase treatment had no effect on the binding of phosphacan-KS, indicating that the keratan sulfate chains do not play a significant role in this process, and neuraminidase treatment of either the proteoglycans, neural cell adhesion molecules, or tenascin-C also had little or no effect on their ability to interact. However, as expected from the results previously obtained with phosphacan itself (29), peptide N-glycosidase treatment of phosphacan-KS significantly decreased its binding to Ng-CAM, N-CAM, and tenascin-C (data not shown). We can therefore conclude from TAG-1/Axonin-1 Is a Ligand of Neurocan, Phosphacan, and N-CAM these data that N-linked oligosaccharides play a major role in the binding of phosphacan and phosphacan-KS to Ng-CAM/L1, N-CAM, and tenascin-C, whereas the presence of chondroitin sulfate is important for interactions of these proteoglycans with TAG-1/axonin-1, and that sialic acid and keratan sulfate do not appear to have a significant role.
Because relatively little was known concerning the differences between the molecular forms of phosphacan-KS, we also examined the electrophoretic pattern of their 125 I-labeled core glycoproteins. After treatment with chondroitinase and endo-␤-galactosidase to remove both the chondroitin sulfate and keratan sulfate glycosaminoglycan chains, one obtains varying proportions of two core glycoproteins with apparent molecular sizes on SDS-polyacrylamide gel electrophoresis of 315 and 380 kDa. When the four different preparations of phosphacan-KS were tested for binding to previously identified ligands of phosphacan, it was found in each case that only the 380-kDa component bound to axonin-1 and tenascin-C, whereas both the 315-and 380-kDa components bound to Ng-CAM/L1 and N-CAM, demonstrating an additional level at which the binding of phosphacan may be modulated by glycosylation.
Homophilic and Other Heterophilic Interactions of TAG-1/ Axonin-1-We also investigated the binding of TAG-1/axonin-1 to itself and to other potential ligands. Although some (but not all) previous studies have demonstrated homophilic binding of TAG-1/axonin-1 and heterophilic interactions with Ng-CAM/L1 and Nr-CAM (24, 25, 30 -33), because of the different types of assays used it is not possible to make quantitative comparisons of binding to different ligands. Using a radioligand assay, we detected only a relatively low degree (3-4%) of homophilic binding of TAG-1/axonin-1 or heterophilic binding to Ng-CAM/L1 and Nr-CAM (Fig. 9). These values are nevertheless significant insofar as they exceeded background binding to BSA by 4-to 10-fold. Unexpectedly, we found much greater binding to several previously unreported ligands for TAG-1/ axonin-1 including N-CAM (11%), tenascin-C (10%), laminin (12%), and collagens I and II (8 -15%), whereas there was no significant binding to a number of other cell surface or extracellular matrix proteins such as fibronectin, vitronectin, mer-  It should be noted that the polyclonal antiserum to neurocan used for this double immunofluorescence study stains structures in embryonic nervous tissue that are not recognized by the 1F6 monoclonal antibody (15,16) prepared to proteoglycans from postnatal brain. Bars, 200 m.

TAG-1/Axonin-1 Is a Ligand of Neurocan, Phosphacan, and N-CAM
osin, collagens III and V, and epidermal growth factor or fibroblast growth factor-2 receptors (Fig. 9). These values all refer to the binding of vitreous fluid TAG-1/axonin-1 in 150 mM NaCl. Brain membrane TAG-1/axonin-1 gave a similar binding profile but a slightly lower percent binding (data not shown).
The binding to N-CAM was of high affinity (K d ϳ1 nM), saturable, and reversible (Figs. 3 and 10). There was also saturable and reversible binding of 125 I-axonin-1 to immobilized tenascin-C with a calculated K d of ϳ0.3 nM (data not shown), but because binding, although reversible, was surprisingly not inhibited by even relatively high concentrations of unlabeled axonin-1, this value cannot be considered as reliable. However, binding of 125 I-tenascin-C to axonin-1 was saturable, reversible, and inhibitable, but indicated a lower affinity with a K d of ϳ9 nM (Figs. 3 and 10). Immunocytochemical studies demonstrated an overlapping localization of TAG-1 with N-CAM in the spinal cord, dorsal root ganglion, optic chiasm (Fig.  11), and retina, and there are areas in the developing brain (e.g. the lateral olfactory tract of E16 cortex) in which we observed a colocalization of TAG-1 and tenascin-C (not shown).
Binding of axonin-1 to laminin and certain collagens was also observed (Fig. 9). In further studies of the interactions with laminin, it was found that, as described above for tenascin-C, binding of 125 I-axonin-1 was not inhibited by unlabeled ligand in 50 mM NaCl (and was not tested in 150 mM NaCl, in which binding decreases by 65-70%). Because laminin and collagen are very minor and transient components of central nervous tissue parenchyma and because the colocalization of laminin with TAG-1 could not be demonstrated in preliminary immunocytochemical studies, we did not investigate the specificity and affinity of these interactions in greater detail. However, it is possible that they are functionally significant in the peripheral nervous system.
The interactions between proteoglycans, neural cell adhesion molecules, and tenascin-C, and the differences in the binding affinities of these complexes, suggest that an important role of neurocan and phosphacan may be to competitively inhibit cell interactions with neural cell adhesion and extracellular matrix proteins. We have previously demonstrated that neurocan and phosphacan inhibit homophilic interactions of Ng-CAM and N-CAM (34). These inhibitory effects may be mediated by com- FIG. 8. Immunoperoxidase staining of TAG-1 (A), phosphacan (B), and neurocan (C) in adjacent sections of postnatal day 7 rat cerebellum, using the 3.1C12, 3F8, and 1F6 monoclonal antibodies, respectively. All three proteins are found in the fiber tracts of the prospective white matter (WM) and in the molecular layer (arrow), but are absent from the external granule cell layer (asterisk). Bar, 200 m.

FIG. 9.
Binding of axonin-1 to neural cell adhesion molecules, extracellular matrix proteins, and growth factor receptors. Proteins were coated in removable plastic wells (neural cell adhesion molecules using a concentration of 1-2 g/ml; other proteins at 10 g/ml), and 125 I-axonin-1 was used at ϳ110,000 cpm/well. The percent bound represents specific binding (total cpm bound minus cpm bound to wells coated with BSA) using binding buffer containing 150 mM NaCl, and all values are means of duplicate determinations Ϯ S.E.

FIG. 10. Saturation curves and Scatchard plots for the binding of 125 I-axonin-1 to N-CAM and of 125 I-tenascin-C to axonin-1.
Binding values represent specific binding as defined in the legend to Fig. 1. Axonin-1 was tested at 0.3-14 ng/well (4 ϫ 10 18 cpm/mol) and tenascin-C at 9 -152 ng/well (10 18 cpm/mol). Points in the saturation curves are averages of duplicate determinations Ϯ S.E., and Scatchard plots were drawn using the Ligand program (28). petition for the same receptors on the basis of affinity or, because neurocan and phosphacan are large bulky molecules, they may also function by binding to receptors that are not in close proximity to the actual sites that mediate homophilic or heterophilic binding. In view of the newly identified axonin-1/ N-CAM interaction described above, we studied the effects of neurocan, phosphacan, and other soluble molecules on the binding of labeled axonin-1 to immobilized N-CAM (Fig. 12) and found that their relative inhibitory potencies were in good agreement with their binding affinities for axonin-1 and N-CAM (i.e. phosphacan Ͼ neurocan Ͼ axonin-1 Ͼ tenascin-C Ͼ fibronectin and vitronectin). There was also a concentrationdependent inhibitory effect of neurocan and phosphacan on the binding of labeled axonin-1 to tenascin-C (IC 50 for phosphacan ϭ 42 nM, and 41% inhibition by neurocan was observed at 57 nM, which was the highest concentration tested). The higher concentrations of neurocan and phosphacan required for inhibition of axonin-1 binding to tenascin-C are consistent with the lower affinity of both proteoglycans for tenascin-C. DISCUSSION TAG-1/Axonin-1 Interactions with Proteoglycans-We have previously described the interactions of neurocan and phosphacan/RPTP/␤, which are major chondroitin sulfate proteoglycans that are specific to nervous tissue (1), with two neural cell adhesion molecules (Ng-CAM/L1 and N-CAM) and the extracellular matrix protein tenascin-C, whereas there was no significant binding to a large number of other potential ligands that were tested (2)(3)(4). Here we have presented evidence for the overlapping localization and specific, high-affinity binding of neurocan and phosphacan/RPTP/␤ to TAG-1/axonin-1, another Ig-superfamily cell adhesion molecule, and for the interaction of TAG-1/axonin-1 with N-CAM and tenascin-C.
Although TAG-1/axonin-1 is the third neural cell adhesion molecule that has been identified as a ligand for neurocan and phosphacan, these interactions are apparently mediated by different molecular mechanisms, insofar as phosphacan binding to TAG-1/axonin-1 (but not to Ng-CAM/L1 or N-CAM), and neurocan binding to both Ng-CAM/L1 and N-CAM (but not to TAG-1/axonin-1) is greatly decreased by the enzymatic removal of their chondroitin sulfate chains ( Fig. 1 and Refs. 2 and 4). Moreover, interactions of phosphacan (but not neurocan) with Ng-CAM/L1 and N-CAM have been shown to be mediated by N-linked oligosaccharides (29), whereas N-deglycosylation has no effect on its interactions with TAG-1/axonin-1. These differential effects of glycans on the interactions of neurocan and phosphacan with neural cell adhesion and extracellular matrix proteins are summarized in Fig. 13.
A recombinant Fc-fusion protein representing the N-terminal carbonic anhydrase-like domain of phosphacan/RPTP/␤ has also been reported to bind to contactin but not to Ng-CAM/ L1, N-CAM, or tenascin-C (13). Since a tryptic glycopeptide derived from the carbonic anhydrase domain of phosphacan binds to these latter ligands by a mechanism dependent on its N-glycosylation (29), the interaction of phosphacan with contactin is also likely to involve different structural features. Moreover, previous studies demonstrated that neither proteoglycan bound to two other Ig-superfamily neural cell adhesion molecules, the myelin-associated glycoprotein (2, 4) and neurotrimin (35). 3 It may be significant in this connection that neither the myelin-associated glycoprotein nor neurotrimin contain fibronectin type III repeats, which are present in all of the identified ligands of neurocan and phosphacan.
Interactions of TAG-1/Axonin-1 with N-CAM and Tenascin-C-Our preliminary survey of the interactions of TAG-1/axonin-1 with itself and with other potential ligands using a solidphase binding assay yielded several interesting results, which are summarized in Fig. 14. Although in previous studies only weak homophilic binding was detected using axonin-1 bound to Covaspheres (25), binding could be demonstrated following heterologous expression of chicken axonin-1 or its human homologue TAX-1 on the surface of transfected myeloma cells (24,31). Microbeads coated with TAG-1 also bound homophilically to TAG-1 immobilized on a nitrocellulose membrane (32). TAG-1/Axonin-1 Is a Ligand of Neurocan, Phosphacan, and N-CAM regard to heterophilic interactions, Kuhn et al. (30) showed that axonin-1 binds to Ng-CAM using a microbead coaggregation assay, whereas Felsenfeld et al. (32) did not detect any heterophilic binding of TAG-1 to either L1/Ng-CAM or N-CAM using an assay involving the aggregation of transfected Drosophila S2 cells. However, they did obtain some indirect evidence for interactions between TAG-1 and L1/Ng-CAM, insofar as antibodies to L1 (but not to N-CAM) inhibited neurite outgrowth on a TAG-1 (but not on a laminin) substrate, and they noted that their negative results do not preclude interactions of TAG-1 with L1 or N-CAM if the binding affinities are below the detection threshold of the S2 cell aggregation assay. Suter et al. (33) have also recently reported the binding of axonin-1 to Nr-CAM based on a Covasphere aggregation assay. It is difficult to directly compare these previous results with those from our radioligand binding assays, which may be both more sensitive and also allow a better quantitation of the extent and specificity of binding. However, our studies suggest that the interactions of TAG-1/axonin-1 with itself and with Ng-CAM or Nr-CAM may be relatively minor in quantitative (but not necessarily in functional) terms, whereas N-CAM may represent an equally or more important ligand.
Binding of tenascin-C to TAG-1/axonin-1 was quantitatively significant (Fig. 9), saturable, and reversible, with a K d of ϳ9 nM (Fig. 10), and there is an overlapping localization of TAG-1/axonin-1 and tenascin-C in the developing central nervous system (data not shown). Although the affinity is approximately one-third of that previously found for the binding of tenascin-C to neurocan and phosphacan (3) and several orders of magnitude lower than that for the proteoglycan interactions with neural cell adhesion molecules (2,4), the basic conditions are present for TAG-1/axonin-1 interactions with tenascin-C to play a functional role during nervous tissue histogenesis.
The saturation experiments in which we studied the binding of labeled axonin-1 to N-CAM involve two molecules that selfaggregate. Self-aggregation of ligand, leading to dimer or multimer formation, may alter the ligand conformation in such a way as to reduce affinity for the receptor, as has been postulated for insulin dimers and higher aggregates (36). It is also possible that aggregate formation may occlude binding sites on the ligand and/or receptor, thus resulting in lower apparent affinity. On the other hand, our measurements may overestimate the affinity of a monomeric axonin-1/N-CAM interaction because a multimeric aggregate may bind with a high overall avidity. For example, in the interaction of the immunoglobulin superfamily cell adhesion molecules CD2 and CD48, a K d of 60 -90 M was measured with monomeric CD48 ligand compared to a K d of 1 M obtained with a preparation of CD48 containing multimeric aggregates (37), and similar results were obtained from studies on monomeric or multimeric LFA-3 binding to cell surface CD2 (38). Factors such as these may explain why labeled axonin-1 bound to tenascin in a saturable but noninhibitable manner with a K d of 0.3 nM (probably based on avidity), whereas labeled tenascin (which does not selfaggregate) bound to axonin-1 with a K d of 9 nM (affinity).
These considerations do not, however, minimize the significance of the saturable binding of axonin-1 to N-CAM. Even if the calculated K d of 1 nM is an overestimate based on avidity binding, such a value may be highly relevant to the in vivo situation in which aggregates of these proteins may be formed on membranes, thus allowing tight binding on the basis of avidity. In studies of reconstituted vesicles composed of N-CAM and brain membrane lipids (39), the markedly nonlinear aggregation rate observed with respect to N-CAM concentration in the lipid bilayer may be due to a cis-interaction of N-CAM (i.e. on the plane of one membrane), and a cis-binding of axonin-1 on membranes has been suggested on the basis of its homophilic interaction (31). Glycosylphosphatidylinositol-anchored proteins such as TAG-1/axonin-1 have an approximately 10-fold higher lateral membrane mobility compared to transmembrane proteins, and this higher mobility has been shown to enhance cell adhesion both by accumulation of CAM at the cell contact area and by increasing the rate of receptorligand bond formation (40).
The affinities for interactions between immunoglobulin superfamily cell adhesion molecules measured by radioligand or other assays are generally quite low (in the micromolar and even submillimolar range; Refs. 37 and 41). These lower affinities are thought to provide a mechanism for de-adhesion when transient interactions are required and for stable adhesion (based on high multimer avidity) such as that involved in fascicle formation. It is likely that neurocan and phosphacan, which bind to Nr-CAM, 4 Ng-CAM/L1, N-CAM, and TAG-1/ axonin-1, as well as to tenascin-C which interacts with TAG-1/axonin-1, may all affect transient and stable adhesions during development and plasticity based on a complex interplay of binding affinities, avidity modulation, and formation of com-4 L. Karthikeyan, P. Milev, and R. U. Margolis, unpublished results. plexes. To our knowledge, the only reported dissociation constant for homophilic or heterophilic binding between immunoglobulin superfamily neural cell adhesion molecules is for the homophilic interaction of rat N-CAM, for which Moran and Bock (42) determined a K d of 1.23 ϫ 10 Ϫ6 M for polysialylated newborn rat brain N-CAM and 6.9 ϫ 10 Ϫ8 M for adult N-CAM using a solid-phase radioligand assay. Although their data may not be directly comparable with ours, the K d for the interaction of TAG-1/axonin-1 with embryonic (polysialic acid-rich) N-CAM is 3 orders of magnitude lower than that for N-CAM homophilic interactions. If these in vitro data reflect the situation in vivo, the N-CAM/axonin-1 interaction may be preferred over homophilic N-CAM interactions. This may also be important for axonal path finding, by analogy with the results of a recent study that demonstrated that antibodies to axonin-1 or Nr-CAM injected in ovo prevented commissural axons from crossing the midline (43).
Conclusions-Numerous in vitro studies and antibody perturbation experiments have demonstrated that TAG-1/axonin-1 is an excellent substrate for neurite growth and is involved in nerve fasciculation (30,32,33,44,45). These studies point to the importance of both homophilic and heterophilic interactions of TAG-1/axonin-1 in neural morphogenetic events and to the complexity of these processes, which may involve associations between several different neural cell adhesion molecules. Because neurocan and phosphacan/RPTP/␤ have been shown to be high-affinity ligands for many of these proteins, they may prevent, modulate, or possibly even enhance adhesive processes (e.g. soluble TAG-1/axonin-1 bound to proteoglycans may no longer be capable of inhibiting fasciculation). Proteoglycans may also affect the formation of protein complexes on the plane of the same membrane or serve as guidance cues on the basis of local gradients or complexes with other molecules. In the latter case, the relative abundance of the proteoglycans and their ligands could facilitate or inhibit the formation of higher order complexes which may direct or modulate neurite growth and fascicle formation.
Our previous studies and the present investigation have therefore demonstrated that the nervous tissue-specific chondroitin sulfate proteoglycans neurocan and phosphacan/ RPTP/␤ interact with several cell adhesion and extracellular matrix proteins. However, the distinctive properties of each proteoglycan (e.g. their relative affinity for different ligands, the consequences of developmentally regulated proteolytic processing, and the modulatory effects of chondroitin sulfate and N-linked oligosaccharides) all suggest that they are components of a multidimensional mechanism for the regulation of cell-cell and cell-matrix interactions at different sites and periods during nervous tissue histogenesis. Some of these interactions may be confined to very restricted areas and/or relatively brief developmental stages, and the multiplicity of ligands with differing properties could provide a means for the fine tuning of various regulatory processes.