Complex-type asparagine-linked oligosaccharides on phosphacan and protein-tyrosine phosphatase-zeta/beta mediate their binding to neural cell adhesion molecules and tenascin.

Phosphacan, a soluble nervous tissue-specific chondroitin sulfate proteoglycan, is an alternative splicing product representing the entire extracellular domain of a transmembrane receptor-type protein-tyrosine phosphatase (RPTP zeta/beta) that also occurs as a chondroitin sulfate proteoglycan in brain. We have previously demonstrated that phosphacan binds with high affinity to neural cell adhesion molecules (Ng-CAM/L1 and N-CAM) and to the extracellular matrix protein tenascin and that it is a potent inhibitor of cell adhesion and neurite outgrowth. Tryptic digests of 125I-labeled phosphacan contain two glycopeptides that bind to Ng-CAM/L1, N-CAM, and tenascin. The larger of these (17 kDa) begins at Gln-209 near the end of the carbonic anhydrase-like domain of phosphacan/RPTP zeta/beta, whereas a 13-kDa glycopeptide begins at His-361 located in the middle of the fibronectin type III-like domain. Treatment of phosphacan with peptide N-glycosidase under nondenaturing conditions reduced its binding the neural cell adhesion molecules and tenascin by 65-75%, whereas endo-beta-N-acetylglucosaminidase H had no effect, and peptide N-glycosidase treatment both decreased the molecular sizes of the tryptic peptides to congruent to 11 kDa and abolished their binding. Based on the amino acid sequence of phosphacan, it can be concluded that each of the tryptic peptides contains one potential N-glycosylation site (at Asn-232 and Asn-381), and analyses of the isolated glycopeptides demonstrated the presence of sialylated complex-type oligosaccharides. Our results therefore indicate that the interactions of phosphacan/RPTP zeta/beta with neural cell adhesion molecules and tenascin are mediated by asparagine-linked oligosaccharides present in their carbonic anhydrase- and fibronectin type III-like domains.

We have previously described the isolation and biochemical properties of phosphacan, a nervous tissue-specific chondroitin sulfate proteoglycan that is synthesized by astrocytes (Rauch et al., 1991). Cloning of phosphacan (Maurel et al., 1994) demonstrated that it is an alternative splicing product representing the entire extracellular domain of a receptor-type protein-tyrosine phosphatase (RPTP) 1 named RPTP/␤ (Krueger and Saito, 1992;Levy et al., 1993;Maurel et al., 1995). A keratan sulfate-containing glycoform (phosphacan-KS) also occurs in postnatal brain (Rauch et al., 1991;Maurel et al., 1994;Meyer-Puttlitz et al., 1995). Phosphacan binds reversibly and with high affinity to the neural cell adhesion molecules Ng-CAM/L1 and N-CAM (K d ϳ0.1 nM) and to the extracellular matrix protein tenascin (K d ϭ 3 nM), but not to over a dozen other cell membrane and extracellular matrix proteins tested Grumet et al., 1994). These studies also demonstrated that phosphacan is a potent inhibitor of neuronal and glial adhesion and of neurite outgrowth. Because of its potential importance both as a developmentally regulated extracellular matrix proteoglycan of nervous tissue and as the ligand-binding domain of a transmembrane protein-tyrosine phosphatase (which also occurs in the form of a chondroitin sulfate proteoglycan in brain; Shitara et al., 1994), we have attempted to identify the region of phosphacan/RPTP/␤ that participates in its interactions with nervous tissue proteins, using the glycosylated proteoglycan that is synthesized by rat brain in vivo. Phosphacan contains an N-terminal carbonic anhydrase-like domain followed by a single fibronectin type III sequence and a glycosaminoglycan linkage region to which are attached three to four chondroitin sulfate chains and (in the case of phosphacan-KS) a similar or greater number of keratan sulfate chains which vary in length and fine structure (Rauch et al., 1991;Maurel et al., 1994). We have found that two tryptic glycopeptides derived from the carbonic anhydrase-and fibronectin type III-like domains of phosphacan/RPTP/␤ bind to the neural cell adhesion molecules and tenascin and that their interactions are mediated by asparagine-linked oligosaccharides.

EXPERIMENTAL PROCEDURES
Isolation of Phosphacan and Enzyme Digestions-Phosphacan was isolated from PBS extracts of 7-day and adult rat brain by DEAEcellulose ion exchange chromatography and Sepharose CL-6B gel filtration, followed by immunoaffinity chromatography using the 3F8 monoclonal antibody (Rauch et al., 1991). The proteoglycan was labeled to a specific activity of ϳ10 19 cpm/mol with 125 I by the lactoperoxidase/ glucose oxidase method using Enzymobeads (Bio-Rad). Typically, 25 g of protein were labeled per reaction, and free iodine was removed by gel filtration on a PD-10 column (Pharmacia Biotech Inc.).
Trypsin and chondroitinase ABC digestions were performed at 37°C in 100 mM Tris-HCl buffer, pH 8.0, containing 30 mM sodium acetate. A range of trypsin to phosphacan ratios was tested, from which it was found that the optimal conditions for generation of peptides capable of binding to Ng-CAM were 2-h digestion at an enzyme-substrate ratio of 1 to 20 (w/w) or 18 h at a ratio of 1 to 200 -400. L-1-tosylamido-2phenylethyl chloromethyl ketone-trypsin was obtained from Worthington, and following digestion the enzyme was inactivated using 0.5 mM N ␣ -p-tosyl-L-lysine chloromethyl ketone (Sigma).
For removal of asparagine-linked oligosaccharides, phosphacan was treated for 6 h at 37°C with either recombinant peptide N-glycosidase * This work was supported by Grants NS-09348, NS-13876, and MH-00129 from the National Institutes of Health. 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 U.S.C. Section 1734 solely to indicate this fact.
Binding Assays-Chicken Ng-CAM, rat L1/NILE, and chicken and rat N-CAM were purified from 14-day embryonic chicken brains and 7-day postnatal rat brains by immunoaffinity chromatography as described previously  and generously provided by Dr. Martin Grumet. Human tenascin was obtained from Dr. Mario Bourdon. Binding assays were performed in removable Immulon-4 wells in 16 mM Tris, pH 7.2, 50 mM NaCl, 2 mM CaCl 2 , 2 mM MgCl 2 , 0.02% NaN 3 containing 1 mg/ml heat-treated BSA as described previously , and the same buffer was used for blocking. For binding assays with labeled tryptic peptides, wells were coated with a higher concentration of Ng-CAM and N-CAM (10 g/ml) than used for solid-phase binding of native phosphacan (2 g/ml) and incubated with gentle shaking (45 rpm) for 4 h at room temperature. For SDS-PAGE, the wells were washed and bound material was extracted with hot SDS sample buffer.
Peptide Sequencing-For N-terminal sequencing of the 13-and 17-kDa peptides that bound to the neural cell adhesion molecules, 52 g of unlabeled phosphacan was mixed with 780,000 cpm of 125 I-labeled phosphacan (without added BSA) and treated for 2 h with 80 milliunits of chondroitinase ABC (Seikagaku America Inc.) and then for 18 h with trypsin at an enzyme-substrate ratio of 1 to 400. SDS-PAGE (on a 1.5 mm, 12.5% minigel) and transfer to a ProBlott membrane (Applied Biosystems) were performed as described previously (Rauch et al., 1991). Due to the poor Coomassie Blue staining of phosphacan, the positions of the 13-and 17-kDa bands were determined by autoradiography before excision from the membrane.
Monosaccharide Analysis of Tryptic Glycopeptides-Tryptic glycopeptides were prepared from 50 -100 g of phosphacan, transferred to a ProBlott membrane after electrophoresis on a 15% minigel, and the 13-and 17-kDa bands were excised as described above. Bands were hydrolyzed for 4 h at 100°C in 2 M trifluoroacetic acid for analysis of neutral sugars and hexosamines or for 30 min in 0.1 N HCl at 80°C for sialic acid, and monosaccharides were quantitated by high-performance anion exchange chromatography with pulsed amperometric detection (Weitzhandler et al., 1993;Margolis and Margolis, 1994). A single hydrolysis was used for both neutral and amino sugars to allow calculation of molar ratios, and hexosamine values were corrected for release under these conditions. Due to the presence in the membrane of apparently polymeric material that after hydrolysis yielded a large interfering peak at the elution position of glucose, the ProBlott membrane was prehydrolyzed for 4 h in 2 M trifluoroacetic acid and washed with water before use for transfer of the tryptic glycopeptides.

RESULTS
Tryptic (glyco)peptides prepared from 125 I-labeled phosphacan gave a complex pattern on an SDS-PAGE gradient gel (Fig.  1A). When this mixture was added to plastic wells coated with chick Ng-CAM, its rat homolog L1/NILE, or rat or chick N-CAM, two bands with apparent molecular sizes of 13 and 17 kDa were seen on autoradiographs of SDS gels of the bound radioactivity ( Fig. 1D and data not shown), and proteins of these sizes were depleted from the unbound material (Fig. 1C). The bound 13-and 17-kDa peptides were also seen without chondroitinase treatment and following elution and electrophoresis in the absence of mercaptoethanol. These results are consistent with both peptides originating from regions of phosphacan before the first potential chondroitin sulfate attachment site at Ser-595 (Maurel et al., 1994) and demonstrate that each peptide binds independently, rather than as a result of disulfide linkage to a single peptide that contains the binding site. There was much less binding of tryptic peptides to immobilized tenascin, which probably reflects the 30-fold lower af-finity of the phosphacan-tenascin as compared with the phosphacan-Ng-CAM interactions Milev et al., 1994). However, the 13-and 17-kDa peptides could also be identified by SDS-PAGE and autoradiography after elution from tenascin-coated wells.
To determine the origin of these peptides within the phosphacan sequence, unlabeled proteoglycan was mixed with iodinated material, digested with trypsin, fractionated by SDS-PAGE, and transferred to a ProBlott membrane. The positions of the 13-and 17-kDa peptides were identified by autoradiography, and the excised bands were used for N-terminal microsequencing. The 13-kDa peptide gave a sequence (HEFLT DGYQDLGAILNNLIP-MSYV) beginning at His-361 of phosphacan, whereas the 17-kDa peptide sequence (QA----FILQNLLPISTDKYY) began at Gln-209 ( Fig. 2A).
In view of the fact that two peptides were observed to bind to the neural cell adhesion molecules and tenascin and that both peptides contained potential N-glycosylation sites (at Asn-232 and Asn-381), we considered the possibility that oligosaccharides might be involved in these interactions. Moreover, there was a blank at Asn-381 in our amino acid sequence of the 13-kDa tryptic peptide (see above), indicating that this glycosylation site was utilized. Following peptide N-glycosidase treatment of phosphacan, the electrophoretic mobility of the tryptic peptides increased and appeared as a probably overlapping broad band of the two peptides at approximately 11 kDa (Fig. 1, E-H). Based on this result and the residue weights, it can be estimated that the C termini of the two peptides are probably at Lys-303 and Lys-456. Although the decrease in molecular size following deglycosylation of the larger glycopeptide is greater than can be accounted for by the size of a complex oligosaccharide, this result is nevertheless consistent with the anomalously slow electrophoretic migration of certain glycopeptides and glycoproteins, which varies with oligosaccharide and protein structure, due to decreased binding of SDS (Segrest and Jackson, 1972;Leach et al., 1980).
Peptide N-glycosidase treatment of phosphacan reduced its binding to the neural cell adhesion molecules and tenascin by 65-75% but had little or no effect on the binding of neurocan  A and E-H) or on a 12.5% minigel (lanes B-D). Lanes A and B, total tryptic peptides; lane C, peptides that did not bind to Ng-CAM; D, peptides bound to and eluted from Ng-CAM. Bound peptides eluted from L1/NILE (the rat homolog of Ng-CAM) and from rat or chick N-CAM and tenascin gave a pattern identical to that seen in lane D (not shown). Peptide N-glycosidase treatment of 125 I-labeled phosphacan isolated from 7-day postnatal (lanes E and F) or adult (lanes G and H) rat brain results in the disappearance of the 17-and 13-kDa tryptic glycopeptides and the appearance of a new band at ϳ11 kDa (lanes F and H).
Oligosaccharides Mediate Binding of Phosphacan/RPTP/␤ 24651 ( Fig. 3), which is a member of the aggrecan-versican family of hyaluronic acid-binding proteoglycans and has no known structural similarity to phosphacan (Rauch et al., 1992). These results are in good agreement with our previous demonstration that, in distinction to phosphacan , the interactions of neurocan with neural cell adhesion molecules are to a large extent mediated by its chondroitin sulfate chains . Similar effects of N-deglycosylation were observed when binding was measured in buffer containing 150 mM rather than 50 mM NaCl, although as demonstrated previously the percent bound is somewhat lower in 150 mM NaCl . The residual phosphacan binding after peptide N-glycosidase treatment probably reflects incomplete deglycosylation of the unreduced protein under nondenaturing conditions in which the oligosaccharides may not be completely accessible to the enzyme (Tarentino et al., 1985;Hirani et al., 1987;Nuck et al., 1990), and this is apparent in the small amounts of the 17-and 13-kDa glycopeptides that are obtained from glycosidase-treated phosphacan (Fig. 1, lanes F  and H). As expected, there were only traces of radioactivity bound to Ng-CAM and N-CAM using tryptic peptides prepared from peptide N-glycosidase-treated phosphacan (data not shown). Because there are developmental changes in the glycosylation of phosphacan (Rauch et al., 1991), we also tested binding and the effects of N-deglycosylation using proteoglycan isolated from both 7-day postnatal and adult brain, but no age-related differences were found (Figs. 1 and 3). Electrophoresis of glycosidase-treated phosphacan showed no evidence of protease activity in the enzyme preparation, and control experiments showed that residual glycosidase present in the diluted proteoglycan used for the binding assay did not affect binding by acting on the CAM or tenascin during the binding incubation period. In fact, treatment of Ng-CAM with peptide N-glycosidase increased the binding of phosphacan by nearly 100%, but peptide N-glycosidase treatment of N-CAM and tenascin had no effect on phosphacan binding (data not shown). Moreover, glycosidase treatment of Ng-CAM had no detectable effect on the binding of phosphacan tryptic glycopeptides, suggesting that N-linked oligosaccharides on Ng-CAM may modulate its interactions with phosphacan by affecting Ng-CAM conformation or through steric factors.
Monosaccharide analyses of the two tryptic glycopeptides showed in both cases the presence of glucosamine, mannose, galactose, and sialic acid in molar ratios characteristic of tri-antennary oligosaccharides (Table I). A large portion of the oligosaccharides on the 17-kDa glycopeptide also contain fucose. These results are consistent with the finding that treatment of phosphacan with Endo H, which releases high-mannose and hybrid oligosaccharides, had no effect on its binding to Ng-CAM, N-CAM, or tenascin (data not shown). We have previously found that phosphacan contains oligosaccharides with 3-sulfated residues that are recognized by the HNK-1 monoclonal antibody (Rauch et al., 1991), as well as Lewis X oligosaccharides (Allendoerfer et al., 1995) that are not present on other soluble chondroitin sulfate proteoglycans of rat brain. 2 However, staining of immunoblots demonstrated that neither of the tryptic glycopeptides that bind to neural cell adhesion molecules and tenascin show HNK-1 or Lewis X reactivity, indicating that these structures are not involved in the interactions we have described.
The only amino acid sequence similarity between the two tryptic glycopeptides resides in the octapeptides ILQNLLPN (beginning at residue 215) and ILNNLIPN (beginning at residue 374), which have 75% identity ( Fig. 2A). A synthetic peptide containing the ILNNLIPN sequence and corresponding to a lysine followed by the first 21 amino acids of the 13-kDa glycopeptide produced up to 60% inhibition of phosphacan binding to Ng-CAM. However, there was even greater inhibition by a scrambled control peptide having the same amino acid composition, and binding inhibition studies using the first or last ten amino acids of this peptide sequence as well as a number of unrelated peptides also failed to support the involvement of these octapeptide sequences in phosphacan binding (Fig. 4). The highly variable inhibitory effects of the tested peptides probably reflects their ability to interact nonspecifically with the oligosaccharides involved in phosphacan binding. DISCUSSION Although there has been considerable interest in the function of glycans in molecular recognition, protein sorting, and developmental processes, the dramatic decrease in binding of phosphacan/RPTP/␤ to Ng-CAM/L1, N-CAM, and tenascin fol- FIG. 2. A, amino acids 201-500 of phosphacan (Maurel et al., 1994) showing the trypsin cleavage sites used for generation of the 17 (filled arrowhead)-and 13 (open arrowhead)-kDa glycopeptides, potential Nglycosylation sites (asterisks), the estimated C termini of the two peptides (vertical bars), and the two similar octapeptide sequences (boxed). B, diagram showing the location of the tryptic glycopeptides in relation to the carbonic anhydrase-and fibronectin type III-like domains of phosphacan/RPTP/␤. The numbers indicate the amino acid residues corresponding to the N termini and the estimated C termini of the peptides and to the domain boundaries. Oligosaccharides Mediate Binding of Phosphacan/RPTP/␤ 24652 lowing removal of N-linked oligosaccharides represents one of the still relatively few instances in which they have been demonstrated to play a role in protein interactions in vertebrate tissues. Because glycosylation is tissue-and cell type-specific (Parekh et al., 1987(Parekh et al., , 1989, we chose to utilize for our studies phosphacan purified from brain, since recombinant forms may not address the function of the native proteoglycan, and this is presumably the reason why carbonic anhydrase and fibronectin type III domains of phosphacan/RPTP/␤ expressed as Fc fusion proteins in COS7 or 293 cells were not capable of binding to Ng-CAM, N-CAM, or tenascin (Peles et al., 1995). The lack of binding of a recombinant protein containing the phosphacan/ RPTP/␤ amino acid sequence up to Leu-415 provides additional evidence for the role of specific oligosaccharides in the binding process.
Our results suggest that the neural cell adhesion molecules and tenascin, all of which are involved in cell interactions, contain a lectin sequence that binds the phosphacan/RPTP/␤ oligosaccharides, and like certain other lectins, the binding of phosphacan by tenascin requires the presence of divalent cations. 2 The phosphacan binding site may be located in the Iglike domains of Ng-CAM and N-CAM, since it has previously been reported that the fourth Ig-like domain of N-CAM contains a recognition site for high-mannose oligosaccharides on L1 (Horstkorte et al., 1993) and that a soluble immunoglobulin fusion construct of CD22, a B cell-specific receptor of the Ig-superfamily, binds to the trisaccharide Sia(␣2-6)Gal(␤1-4)GlcNAc that is present in N-and O-linked oligosaccharides (Powell et al., 1995).
Other oligosaccharides may also participate in the binding process in addition to those on the two identified tryptic glycopeptides. The affinity of phosphacan/RPTP/␤ for neural cell adhesion molecules and tenascin is several orders of magnitude greater than that for most lectin interactions with monovalent oligosaccharide ligands, which are usually in the micromolar range (Lee and Lee, 1994), and is therefore consistent with multiple interactions. A more detailed structural characterization of the phosphacan/RPTP/␤ glycans is clearly of interest, both because this information may be useful for the design of agents that affect interactions of a receptor-type protein-tyrosine phosphatase with its ligands, and insofar as the asparagine-linked oligosaccharides involved in the binding of phosphacan to neural cell adhesion molecules and tenascin are also likely to mediate its potent inhibitory effects on cell adhesion and neurite outgrowth that we reported previously . Present Present a Because the samples for hydrolysis and high performance liquid chromatographic analysis of neutral and amino sugars were from different gel lanes and bands (see "Experimental Procedures") than those used for determination of sialic acid, it is not possible to reliably calculate the molar ratio of N-acetylneuraminic acid in relation to that of the other monosaccharides, but significant amounts were detected in both glycopeptides.