Identification of Hyaluronan-binding Domains of Aggrecan*

Aggrecan, a large cartilage proteoglycan, interacts with hyaluronan (HA), to form aggregates which function to resist compression in joints. The N-terminal region of aggrecan contains two structurally related globular domains, G1 and G2 separated by IGD domain. The G1 domain consists of three subdomains, A, B, and B′, structural features characteristic to many other HA-binding proteoglycans. Here, we studied the interaction of aggrecan domains with HA using recombinant proteins expressed in 293 cells, an embryonal kidney cell line. Deglycosylation of the recombinant aggrecan fragment reduced the HA binding activity. We found that both the B and B′ subdomains were required for HA binding and that a single module of A, B, or B′ was unable to bind HA. The A subdomain increased the HA binding activity of the B-B′ region. The G2 domain had no HA binding activity confirming previous reports. Studies of HA-binding properties using a BIAcoreTM biosensor system revealed that the K D of recombinant aggrecan fragment (AgW) consisting of G1, IGD, and G2 was 0.226 μm, whereas the K D of another HA-binding protein, native bovine link protein, is 0.089 μm. In contrast, AgMut11 which lacked subdomain A showed little HA binding activity. AgMut12 consisting of only B-B′ had a 3.4-fold lower affinity and AgMut13 containing A-B-B′ was 1.5-fold lower than AgW. These results suggest that carbohydrates are essential for high level aggrecan binding to HA and that the A subdomain of aggrecan functions in a cooperative manner with subdomains B and B′.

Hyaluronan (HA) 1 is an ubiquitous repeating disaccharide chain. HA has various functions, including tissue morphogenesis, wound repair, cell migration, tumor invasion, and immune recognition (for review, see Refs. 1 and 2). These functions of HA are mediated through specific interactions with HA-binding molecules. A number of these extracellular matrix and cell-surface molecules have been identified and possible mech-anisms of interaction with HA have been proposed (for review, see Refs. [2][3][4]. Several studies have suggested that a proteoglycan tandem repeat (PTR), acts as a functional site of interaction with HA (5,6). Most of the HA-binding molecules, including link protein and aggrecan, contain PTRs. For example, link protein contains three looped domains: A, B, and BЈ, in which B and BЈ contain PTR modules. Aggrecan (7) is a member of a family of large extracellular matrix proteoglycans, which includes versican/proteoglycan-mesenchyme (8,9), neurocan (10), and brevican (11). These proteoglycans have Nterminal globular domains, G 1 , whose structure is homologous to link protein. Like link protein, G 1 consists of an A subdomain and PTR-containing B and BЈ subdomains. Aggrecan is the only member of this family containing an additional globular domain, G 2 , which has a structure similar to G 1 and is separated from it by the interglobular domain, IGD, at the N terminus. The G 2 domain also contains B and BЈ subdomains but without the A subdomain and lacks HA binding activity. The HAreceptor CD44 (12) and the arthritis-associated protein tumor necrosis factor-stimulated gene-6 (TSG-6) (13) have been reported as HA-binding molecules. These molecules contain one PTR motif, suggesting that a single PTR can interact with HA.
Although various HA-binding molecules with PTR have been reported, the HA-binding mechanism of PTR has not yet been well elucidated. For example, it is not clear which one of the PTRs of link protein or of the proteoglycans interacts with HA. The role of the A subdomain for HA binding is also unknown. Various expression systems have been employed to produce recombinant PTR molecules. However, studies of the HA-binding function using these systems have been hampered by insolubility of the PTRs, ternary structure formed by disulfide bonds, and possible effects of glycosylation on HA-binding function. Link protein (14) and TSG-6 (15) have been expressed in Escherichia coli. Since bacteria have no protein disulfide isomerases, disulfide bonds occur at random and the recombinant proteins do not form a correct structure. Therefore, refolding of the molecule with glutathiones (14) or laborious purification steps to obtain molecules with correct structure were performed (15). Link protein has also been expressed in a baculovirus system (16). Since the glycosylation machinery of insect cells is different from that of mammalian cells, the HA binding of the recombinant protein may not be properly assessed. Recent nuclear magnetic resonance (NMR) studies of the PTR structure of TSG-6 show that the ternary structure is similar to the C-type lectin domain (17). Since the TSG-6 was expressed in bacteria, these studies have not considered the role of glycosylation on structure and activity.
Various assays have previously been established such as the transblot assay using labeled HA (18 -20), HA-Sepharose column chromatography (21), co-precipitation using HA-Sepharose or cetylpyridinium chloride (CPC) (13), and enzyme-linked immunosorbent assay (22). Since HA has a highly negative charge, it may interact with any positively charged molecules or basic amino acid residues in a protein. Also by virtue of its viscous nature, HA may trap molecules nonspecifically. In addition, previous studies that used synthetic peptides to identify an active sequence of link protein for HA-binding (23), were inconclusive (24).
In this study, we developed a novel mammalian expression system and expressed a soluble recombinant human aggrecan fragments consisting of various portions of the G1ϳG2 domains linked to the C-terminal portion of the laminin ␥1 chain. This system allows us, for the first time, to study the HA-binding function of recombinant aggrecan domains expressed in mammalian cells. We analyzed these recombinant proteins for HAbinding by several different methods including the HA-Transblot assay and lipid-conjugated HA-binding assay (HA-PE assay). We also used a BIAcore TM biosensor instrument to evaluate the kinetics of binding of recombinant aggrecan fragments and bovine link protein to HA. By measuring the changes in surface plasmon resonance signal of aggrecan domain-and subdomain-specific mutants associating to and dissociating from the immobilized HA on the sensor chip, the binding characteristics of these molecules were monitored in real-time. This technique allows accurate and highly reproducible calculation of their kinetic rate constants. Our results suggest that both B and BЈ are required for HA binding and that the A subdomain and carbohydrates are important for high level interaction between HA and aggrecan.

EXPERIMENTAL PROCEDURES
Construction of the Expression Vectors-A basic expression vector pBFX was constructed from the pcDNA3 vector (Invitrogen, San Diego, CA) with several modifications. The pBFX vector contains the following sequences: cytomegalovirus promoter, a rabbit ␤-globin splicing site (25), a human interleukin-2 receptor signal peptide sequence (26), a FLAG epitope tag (Kodak Scientific Imaging Systems, Rochester, NY), a factor Xa cleavage site, and a segment of mouse laminin ␥1 chain (27). The splicing signal was placed in the 5Ј-untranslated region since it is necessary for high-level expression of the recombinant protein in stably transfected cells. 2 The fusion protein can be purified with anti-FLAG antibody affinity column, and the recombinant protein can be released from the ␥1 chain by factor Xa cleavage. The following is a detailed description for these modifications. First, polymerase chain reaction (PCR) was performed to generate a 207-amino acid cDNA segment of the mouse laminin ␥1 chain using pfuDNA polymerase (Stratagene, La Jolla, CA). The forward primer was 5Ј-GGATCCATCGAAGGTCGTGC-CATCAACCGGACCATAGCT-3Ј (residues 4417 to 4437) with a BamHI linker site and factor Xa cleavage site at 5Ј-end and the reverse primer was 5Ј-GGGCCCTTAGCCGGTTGGTAGGGTCTTCTT-3Ј (residues 6037 to 6007) with an ApaI linker site and a translation termination codon at 5Ј-ends. The reaction program was 25 cycles of 94°C for 12 s, 58°C for 30 s, and 72°C for 1 min (Gene Amp PCR System 9600, Perkin-Elmer, Norwalk, CT). The PCR-amplified DNA fragment that encodes factor Xa cleavage site (Ile-Glu-Gly-Arg) followed by the mouse laminin segment (27) was inserted at BamHI and ApaI sites of pcDNA3. The interleukin-2 receptor and FLAG sequences were prepared by PCR with the plasmid cytomegalovirus-interleukin-2 receptor (26) as a template using the following set of primers: 5Ј-GGTACCCCAAGGGTCAG-GAAGATGGAT-3Ј with a KpnI linker site; 5Ј-GAATTCCCGGAATTC-CTTGTCATCGTCGTCCTTGTA-3Ј with an EcoRI linker site. The ␤globin splicing signal was prepared by PCR with the plasmid pKCR (25) and was cloned into the HindIII site. The resulting plasmid designated as pBFX was used as the basic vector for further aggrecan gene constructions.
DNA segments for various regions of aggrecan were generated by PCR with pSA005 (7) as a template using a set of primers carrying a XhoI linker site for the forward primer and a BamHI linker site for the reverse primer as shown in Table I. These primers are as follows: F1 and R1 for AgW; F2 and R1 for AgMut11; F2 and R2 for AgMut12; F1 and R2 for AgMut13; F1 and R3 for AgMut14; F1 and R4 for AgMut16; F2 and R3 for AgMut17; F3 and R2 for AgMut18; F4 and R1 for AgMut19. For construction of AgMut20, PCR was performed with primers F5 and R1, both of which carry a BamHI linker site. The amplified DNA was inserted into the BamHI site of AgMut16. The mutant proteins, AgMut21 and AgMut22, were constructed using a Quick-Change TM site-directed mutagenesis kit (Stratagene). The cDNA insert from AgMut13 was subcloned into pBluescript II SK(Ϫ) and the plasmid was used as a template for mutagenesis. The primers used for AgMut21 were F6 and R5. In these primers, a TGG codon for tryptophan is replaced with a GCG codon for alanine. The primers for Ag-Mut22 were F7 and R6. The sequence GCCGCC in the forward primer and GGCGGC in the reverse primer encoding two threonines are replaced by ACCACC encoding two alanines. The PCR fragments with the mutations were subcloned into the XhoI and BamHI sites of pBFX. The sequence of all expression constructs was confirmed by DNA sequencing using an automated DNA sequencer (Applied Biosystems, Foster City, CA).
In Vitro Expression, and Purification of Recombinant Proteins-293 cells (CRL 1573, ATCC, Rockville, MD) were cultured in Dulbecco's modified Eagle's medium containing 10% (v/v) fetal bovine serum, penicillin-streptomycin, and L-glutamine (Dulbecco's modified Eagle's medium-10). The cells at ϳ70% confluency in 6-well plates were transfected with 2 g of expression vector DNA and 6 l of LipofectAMINE TM (Life Technologies, Gaithersburg, MD) per well according to the manufacturer's instructions. Forty-eight h after transfection, the medium was replaced with Dulbecco's modified Eagle's medium-10 containing 650 g/ml G418 (Life Technologies). The cells were cultured for 10 days in the presence of G418 and a pool of stable transfectants was further grown to confluency. For preparation of recombinant proteins, 10 dishes (15 cm in diameter) of the confluent cells were used. The cells were rinsed three times with phosphate-buffered saline (PBS) and collected in 10 ml of 20 mM Tris-HCl, pH 7.4, containing 0.15 M NaCl, 0.05% Brij 35, 10 mM EDTA, 2 mM phenylmethanesulfonyl fluoride, and 10 mM N-ethylmaleimide. The protein fractions were extracted twice from the cell lysate by homogenization and centrifugation for 30 min at 10,000 ϫ 2 Y. Yamada, unpublished data. g. The sample was sonicated on ice (W-385, Heat Systems-Ultrasocic, Inc.), and then was ultracentrifuged at 100,000 ϫ g for 1 h. The supernatant was applied to an anti-FLAG M2 column (1 ml) pre-equilibrated with the buffer mentioned above. The column was washed twice with 5 ml of 20 mM Tris-HCl, pH 7.4, containing 1 M NaCl and 10 mM EDTA (TBSE), and twice with 5 ml of 4 mM Tris-HCl, pH 7.4, containing 30 mM NaCl, and 2 mM EDTA (one-fifth ϫ TBSE) and eluted with 5 ml of 150 g/ml FLAG peptide (Kodak Scientific Imaging Systems) in one-fifth ϫ TBSE. The eluate was concentrated five times using a SpeedVac. The eluate from the anti-FLAG M2 column was further purified on a fast protein liquid chromatography system equipped with a Superdex 200 HR 10/30 column (Pharmacia Biotech Inc., Piscataway, NJ). Fractions with the recombinant protein, monitored by immunoblot analysis, were combined and stored at Ϫ80°C. The purity of the recombinant proteins was assessed by SDS-PAGE and silver staining. The concentration of the purified proteins was measured using a BCA protein assay kit (Pierce, Rockford, IL). Immunoblot Analysis and HA-Transblot Assay-The sample was separated by SDS-PAGE under both reducing and non-reducing conditions and was electrotransferred onto a polyvinylidene difluoride membrane. The membrane was blocked for 1 h at room temperature in 20 mM Tris-HCl, pH 7.4, containing 0.15 M NaCl, 0.05% Tween 20 (TBST), and 5% instant non-fat dry milk (Super G, Inc. Landover, MD). For immunoblot analysis, the membrane was incubated with 2.5 g/ml mouse monoclonal anti-FLAG M2 antibody (Kodak Scientific Imaging Systems) in TBST containing 5% dry milk for 1 h at room temperature, then reacted with 2 g/ml horseradish peroxidase-conjugated goat antimouse IgG antibody (Pierce). ECL detection reagents (Amersham, Cleveland, OH) were used to visualize protein bands. For the HA-Transblot assay, the membrane was incubated with 10 g/ml biotinylated HA prepared as previously reported (28) for 2 h at room temperature after blocking (Biotinylated HA for the initial experiment was a gift from Dr. M. Yoneda, Aichi Medical University). The membrane was then reacted with 2 g/ml horseradish peroxidase-conjugated streptavidin (Pierce, Rockford, IL) for 1 h, and treated with ECL to detect proteins that bound to biotinylated HA. The band density was quantitated using a densitometer (PDSI-PC, Molecular Dynamics, Sunnyvale, CA) for HA binding activity.
Lipid-conjugated HA-binding (HA-PE) Assay-A 96-well microtiter plate was coated with 50 l (100 g/ml) of HA conjugated to phosphatidylethanolamine dipalmitoyl (HA-PE, a gift from Seikagaku Corp., Tokyo, Japan) (29) in PBS at 4°C overnight. The plate was then blocked with 1% bovine serum albumin in PBS for 1 h at room temperature. After washing the wells with PBS containing 0.05% Tween 20 (PBST) three times, 50 l of the sample solution was added to each well and was incubated for 1 h at room temperature. The anti-FLAG M2 antibody (1:1,000) was used as the first antibody and horseradish peroxidaseconjugated goat anti-mouse IgG antibody (1:2,000) as second antibody. After washing the well with PBST, 50 ml of the substrate solution (100 g/ml o-phenylenediamine, 1% (v/v) methanol, 0.01% (v/v) H 2 O 2 ) was added and read at A 490 nm using a microtiter plate reader. As a control, bovine link protein or a biotinylated hyaluronan binding region (HABR) prepared from bovine nasal cartilage (30) was reacted with HA. For bovine link protein, mouse monoclonal antibody, 8A-4 (a gift from Dr. B. Caterson, University of Wales), was used instead of anti-FLAG M2 antibody. For HABR, horseradish peroxidase-streptavidin was used instead of the antibodies. For studies of specificity, the same amount of chondroitin sulfate-PE or heparan sulfate-PE (a gift of Seikagaku Corp.) (29) as HA-PE was used.
Deglycosylation-Chemical deglycosylation of the protein was performed using trifluoromethane sulfonic acid (TFMSA) as described previously (31). Briefly, 100 ng of the dry sample was resuspended in 10 l of anisol (Fluka, Ronkonkoma, NY). Ninety l of TFMSA was added to the reaction and it was incubated at 4°C for 2 h. The sample was precipitated and washed five times with ice-cold diethyl ether. The precipitate was vacuum dried. For immunoblot analysis or HA-binding transblot assay, the sample was dissolved in SDS sample buffer. The Glycoshift De-N-Glycosylation kit (Oxford GlycoSystems, Bedford, MA) was used for enzymatic deglycosylation. Briefly, 200 ng of the recombinant proteins was incubated with 0.4 units of peptide N-glycosidase F for 1 h at 37°C in TBS.
BIAcore TM Biosensor--For immobilization of HA to a SA sensor chip (Biacore, Inc., Piscataway, NJ), a solution of biotinylated HA at 40 g/ml in 20 mM Tris-HCl, pH 7.4, containing 0.3 M NaCl, 0.005% Tween 20 was injected into the flow cell at a flow rate of 5 l/min. The amount of biotinylated HA immobilized in each flow cell was controlled by varying the injection volume of the protein solution. In the kinetic studies, the binding assays were performed at 25°C with a constant flow rate of 50 -100 l/min in both association and dissociation phases. In brief, a series of protein concentrations ranging from 0.5 to 3.0 mM in running buffer was injected into the flow cell, and the change in response unit was recorded. After each run, regeneration of the sensor chip surface was accomplished by two successive injections of 15 l of 5 mM HCl. The values for rate constants were determined by nonlinear regression analyses using BIAevaluation 2.1 software provided by the manufacturer and as described in detail by Karlsson et al. (32,33). Association rate constants (k a ) were calculated from the linear portion of sensorgrams during the early association phase. Dissociation rate constants (k d ) were calculated from the early portion of the dissociation phase about 20 s after the completion of the sample injection during the wash-out period. The apparent equilibrium dissociation constant (K D ) was calculated as the ratio of k d /k a . The kinetic constants were determined by six to 10 independent experiments.

Expression of Recombinant Aggrecan Fragments with Hya-
luronan Binding Activity-The pBFX vector was constructed for expression of recombinant aggrecan fragments (Fig. 1A). A recombinant aggrecan fragment (AgW) containing the N terminus of human aggrecan core protein spanning from G 1 to G 2 domains was expressed in 293 cells, a human embryonal kidney cell line. AgW is a fusion protein consisting of a segment of the mouse laminin ␥1 chain at the C terminus and the FLAG tag sequence at the N terminus. The cleavage sequence for factor Xa was inserted between the aggrecan G 2 domain and the segment of the laminin ␥-1 chain. The FLAG tag was used for identification and purification of AgW. Expression of the recombinant protein as a fusion protein was necessary because the N-terminal portion of aggrecan without the ␥1 chain was not expressed in 293 cells probably due to rapid degradation of the protein (data not shown). Hence, all recombinant aggrecan molecules in this study were expressed as a fusion protein. The AgW monomer was purified to an apparent homogeneity by two-step column purification with an anti-FLAG M2 antibody and a Superdex-200 FPLC column (data not shown). Approximately 200 g of the recombinant protein was obtained from 10 confluent tissue culture dishes (15 cm in diameter). The expression of the recombinant protein was examined by immunoblot using an anti-FLAG M2 antibody (Fig. 1B). The HA binding activity of AgW was examined by several methods. First, the HA-Transblot assay confirmed the HA binding activity of AgW (Fig. 1B). Reduced AgW did not bind to HA (data not shown), suggesting the importance of disulfide bonds in the G 1 domain for the activity as previously reported for link protein (23). To examine the effect of the laminin ␥1 chain on HA-binding, the fusion protein (AgW) was cleaved by activated factor Xa and examined for its HA binding activity by the HA-Transblot analysis (Fig. 1C). Western blotting with anti-FLAG M2 antibody showed that the fusion protein (M r 138,000) was cleaved to a protein consisting of FLAG sequence and the G 1 and G 2 domains (M r 86,000) with factor Xa. HA-Transblot showed that there is no difference in HA binding ability between the fusion protein and the factor Xa-cleaved 86-kDa product (Fig. 1C). These results indicate that the laminin ␥1 chain does not affect HA binding to AgW.
The lipid-conjugated HA-binding (HA-PE) assay was also used to measure the activity of AgW and to compare its activity with bovine link protein and with the HABR that represents the G 1 domain prepared by partial proteolytic digestion of native bovine aggrecan (30). In this method, PE dipalmitoyl, a lipid, was covalently coupled to the reduced terminus of the HA chain (HA-PE) and coated on a 96-well dish (29). PE has a high affinity for plastic while HA does not, which leads to the preferential attachment of PE to the dish, and leaves the HA portion of the HA-PE conjugate free in solution and available for binding. AgW had approximately 50 and 25% of HA binding activity when compared with the bovine aggrecan fragment and bovine link protein ( Fig. 2A). A control recombinant protein, FLAG-laminin ␥1 chain, showed no HA binding activity. AgW did not bind to chondroitin sulfate or heparan sulfate (Fig.  2B), indicating specific interaction of AgW with HA. Increasing salt concentrations did not inhibit interaction of AgW with HA, suggesting that HA-binding is independent of ionic interactions (Fig. 2C).
Expression of Domain-specific Recombinant Proteins and Their HA-binding Ability-A series of domain-specific constructs were prepared and expressed in 293 cells (Fig. 3A). The recombinant proteins were partially purified by anti-FLAG-M2 columns. The proteins were separated by 4 -20% SDS-PAGE and analyzed by Western blot using anti-FLAG antibody (Fig.   3B) and by HA-Transblot (Fig. 3C). All recombinant proteins showed a single monomeric band except AgMut14 which had an additional slow migrating band that appeared to be a dimer (Fig. 3B). HA-Trasblot analysis demonstrated that AgMut11 and AgMut12, which contain both B and BЈ subdomains of G 1 but lack A subdomain had HA binding activity although Ag- FIG. 1. Expression of recombinant aggrecan domains G 1 -G 2 . A, design of a mammalian expression construct is shown. The prototype expression construct is derived from the basic vector (pBFX) with a backbone of pcDNA3 and contains the cytomegalovirus promoter/enhancer, rabbit globin splicing signals, interleukin-2 receptor signal sequence, FLAG tag sequence, human aggrecan G 1ϳ -G 2 domains, and the C-terminal mouse laminin ␥1 chain. A factor Xa cleavage sequence is inserted between the aggrecan segment and the truncated laminin ␥1 chain. B, the recombinant aggrecan fragment, AgW, was extracted from cell lysates and purified as indicated under "Experimental Procedures." The purified protein was separated by 4 -12% SDS-PAGE and analyzed with an immunoblot using anti-FLAG M2 antibody or with a HA-Transblot assay for HA binding activity with biotinylated HA. C, AgW was digested by factor Xa for different times and analyzed by immunoblot and HA-Transblot assays. The band density was compared and no significant effect of the laminin ␥1 chain was found.
FIG. 2. Enzyme-linked immunosorbent (HA-PE) assay for the recombinant aggrecan fragment. A, using HA-PE assay, the HA binding activity of recombinant aggrecan fragment AgW was compared with that of bovine link protein and of HABR prepared from a proteolytic digest of bovine aggrecan. HA-PE was coated onto a 96-well plate and incubated with the sample. After washing, bound AgW was reacted with anti-FLAG M2 antibody. Link protein or HABR bound to HA was treated with anti-link protein, 8A-4, or horseradish peroxidase-streptavidin as described under "Experimental Procedures." Note that AgW Mut11 showed a significantly lower level of HA-binding than AgMut12. AgMut13 consisting of A, B, and BЈ subdomains showed substantial HA binding activity. The A (AgMut16), B (AgMut17), or BЈ (AgMut18) subdomain alone did not have HA binding activity. A recombinant protein consisting of A and B (AgMut14) also showed no activity even after prolonged exposure to a x-ray film (Fig. 3C, bottom panel). Faint bands seen in the prolonged exposure in lanes 12 and 14 in Fig. 3C represent nonspecific reactions since the band did not correspond to the molecular weight of recombinant AgMut12 and AgMut14. The B-BЈ segment from G 2 (AgM19) did not show HA binding activity. The presence of the A subdomain (AgMut20) also failed to bind to HA. Similar results were obtained from HA-PE analysis (data not shown). These results indicate that the minimal segment for HA-binding is the B-BЈ loops of the G 1 domain.
Our results differed from previous reports on recombinant link protein expressed in insect SF9 cells (16) and in bacteria (14) in that a single loop of B or BЈ was sufficient to bind HA using the HA-Sepharose or CPC precipitation method. We, therefore, examined HA binding activity of the recombinant aggrecan fragments using these same methods. In the HA-Sepharose method, the recombinant proteins were incubated with HA-Sepharose beads and bound proteins were analyzed. We found that all recombinant proteins shown in Fig. 3A were bound to HA-Sepharose even after extensive washes with different conditions (data not shown). We also found that a control protein containing FLAG and the C terminus laminin ␥1 chain (FLAG-laminin ␥1) derived from the basic vector pBFX, bound to the HA-Sepharose beads although another control protein, bovine serum albumin, did not bind to HA-Sepharose (data not shown). These results suggest that the HA-Sepharose method was not appropriate to evaluate specific binding of recombinant aggrecan fragments to HA. In the CPC assay, the recombinant proteins were incubated with HA and bound proteins were precipitated with 1% CPC. All recombinant proteins except the negative controls, FLAG-␥1 and bovine serum albumin, were co-precipitated with HA. However, even under reducing conditions these recombinant proteins were also co-precipitated with HA (data not shown). Thus, the CPC method is incapable of detecting conformation-dependent HA binding.
Effect of the A Subdomain on HA Binding-Although AgW, AgMut11, AgMut12, and AgMut13 were active for HA binding, their binding activity levels differed. HA binding activity of these recombinants was compared quantitatively using immunoblot and HA-Transblot. Four serially diluted samples from each recombinant aggrecan fragment were separated on a set of two SDS-PAGE gels under nonreducing conditions. One of the gels was analyzed by immunoblot to estimate the amount of the proteins and the other by HA-Transblot for HA binding activity. Fig. 4A (AgW and AgMut11) and Fig. 4B (AgW, Ag-Mut13, and AgMut12) show representative patterns of the immunoblots and HA-Transblots. The HA-Transblot showed much less relative HA binding activity of AgMut11 than AgW, compared with the same ratio between AgW and AgMut11 in immunoblots. Similarly, AgMut12 showed less binding reactivity to HA as compared with AgMut13. The affinity to HA was quantitated by measuring the density of the bands and the ratio of the reactivity was statistically analyzed from two independent experiments (Fig. 4C). The HA binding ability of AgMut11, AgMut12, and AgMut13 were calculated as 3.1 Ϯ 1.9%, 32.1 Ϯ 16.3%, and 84.0 Ϯ 22.9% of AgW, respectively. These data suggest that the A subdomain significantly enhances HA binding activity of the B-BЈ segment.
The A subdomain structure is characteristic of an immunoglobulin (Ig) type-fold with three and four ␤-sheets in parallel orientation (6). As the structure of the hypervariable region is dependent on interaction with antigens, the segment of the A loop may be important for the enhancement of HA binding activity. The tryptophan (Trp-75) in the center of the Ig-like fold functions as a hinge of these ␤ sheets and is critical for its ternary structure (Fig. 5A). To examine whether the conformation of the A subdomain is essential for its enhancing effect on the HA-binding function of aggrecan, a mutant protein, Ag-Mut21, was generated with the tryptophan replaced with alanine. Using immunoblot and HA-Transblot assays (Fig. 5, B and C), HA binding activity of AgMut21 was determined to be 45 Ϯ 9% that of AgMut13. A similar result was obtained with the HA-PE assay (Fig. 5D). Double substitutions (AgMut22) at two threonines (Thr-61 and Thr-62) with alanines reduced HA binding activity to 31 Ϯ 15% that of AgMut13 (Fig. 5, B and C). By HA-PE assay, a decrease in the HA binding affinity was similarly observed (Fig. 5D). These results suggest that the structure of the A subdomain plays an important role in high level HA binding.
Effect of Deglycosylation on HA Binding Activity-The G 1 domain has been shown to have O-and N-linked carbohydrate chains (34,35). We examined the role of these carbohydrate chains in HA binding activity by removing them either chem-ically or enzymatically. TFMSA treatment which removed both O-and N-linked carbohydrates, reduced the molecular weight of AgMut13 significantly, indicating that AgMut13 is glycosylated. Since a pilot experiment revealed a significant reduction in HA binding of deglycosylated AgMut13, 5-fold more was applied on a SDS-PAGE gel and compared with unglycosylated AgMut13 in immunoblot and HA binding assays in Fig. 6A. Densitometric calculation of the immunoblot with anti-FLAG M2 antibody showed that 97% of AgMut13 was deglycosylated and 3% remained glycosylated (Fig. 6A). HA-Transblot showed that the ratio of the intensity of the bands for HA binding is 36.5 to 63.5 for deglycosylated to glycosylated proteins. Hence, HA binding activity of deglycosylated AgMut13 was about 2% that of glycosylated AgMut13. Peptide N-glycosidase F treat- ment removed N-linked carbohydrates. Equal amounts of the sample treated or non-treated with the enzyme was analyzed for immunoblot and HA binding in Fig. 6B. HA-Transblot showed that the intensity of the deglycosylated protein was reduced to 30% of the non-glycosylated (Fig. 6B). The enzymatic deglycosylation did not affect the level of immunoreactivity of AgMut13 with anti-FLAG M2 antibody. These results indicated that both O-and N-linked carbohydrates are involved in the high level of HA binding of aggrecan.
Analysis of HA-binding Function Using BIAcore TM System-A BIAcore TM biosensor was used to analyze real-time interactions of various recombinant aggrecan fragments and HA. Prior to kinetic studies, control experiments with different concentrations of the proteins were injected over the streptavidin surface to determine nonspecific binding and baseline responses of these proteins (reference surface). The results from these reference surface runs were subsequently subtracted from the results obtained from experiments to eliminate the contributions of nonspecific binding and sample refractive index changes. For kinetic measurements, a low concentration of biotinylated HA, ranging from 62 to 102 response units, was captured on the SA sensor chip surface to facilitate kinetic studies and to minimize mass transport and rebinding effects. Fig. 7 shows a representative overlay of sensorgrams used for kinetic measurements of bovine link protein.
Each injection of different concentrations of link protein gave a sharp rise when injected at A due to a bulk refractive index change at the sensor chip surface. The gradual increase in response from A to B was due to binding of bovine link protein to immobilized HA, in which A indicated the beginning phase of the association phase and B the beginning of the dissociation phase. Our results indicated that the association of AgW and bovine link protein with HA was faster than AgMut12 and AgMut13, as shown in Table II, with association rate constants of 1.76 ϫ 10 4 M Ϫ1 s Ϫ1 and 1.48 ϫ 10 4 M Ϫ1 s Ϫ1 , respectively. Therefore, AgMut12 and AgMut13 bound HA 1.7 to 1.2-fold more slowly than AgW or link protein. Interestingly, AgMut11 did not bind to hyaluronan at all, even at a very high protein concentration. Dissociation rate constants of AgW and other proteins were moderately slow ranging from 1.31 ϫ 10 Ϫ3 s Ϫ1 (bovine link protein) to 6.56 ϫ 10 Ϫ3 s Ϫ1 (AgMut12). Other recombinant proteins including AgMut14 -20 did not bind to HA, which is consistent with the results obtained by HA-Transbot and HA-PE assays. DISCUSSION Using a mammalian expression system, we expressed recombinant proteins containing various domains of human aggrecan and tested their activity for HA binding. A part of the mouse laminin ␥1 chain was fused to the C terminus of the N-terminal segment of aggrecan. The ␥1 chain fusion is required for stable expression of the recombinant proteins and the highly soluble property chain contributes to the increase in solubility of the recombinant proteins. The ␥1 chain and the FLAG tag had no effect on HA binding to aggrecan globular domains.
In this study, we used several different assays for HA binding, including HA-Transblot, HA-PE, and a BIAcore TM biosensor system. Essentially similar results were obtained from these assays. We found that both the B and BЈ subdomains in the G 1 domain are required for HA binding. Individual subdomains, A, B, or BЈ alone are inactive for HA binding. We also found that the A subdomain plays a critical role in enhancing the HA binding activity of the G 1 domain of aggrecan. There is a more profound effect of the A subdomain on HA binding for a longer recombinant molecule consisting G 1 , IGD, and G 2 (Ag-Mut11) than for a smaller molecule containing B and BЈ (Ag-Mut12). In the BIAcore TM assay, this is more evident in that AgMut11 did not show any HA binding activity, whereas inclusion of the A subdomain in AgMut11 (AgW) resulted in HA binding activity to nearly the same level as native link protein.
In the absence of the A subdomain, the IGD and G 2 domains may block the B-BЈ loops of the G 1 domain and interfere with B, the beginning of dissociation phase. The sharp rise at A in response to each link protein injection is due to the bulk refractive index change at the sensor chip surface; the gradual increase from A to B is due to binding of bovine link protein to immobilized HA.

TABLE II
Summary of rate and equilibrium dissociation constants of AgW and AgW mutants interactions with HA Summary of kinetic parameters for the interaction between AgW mutants and bovine-LP and HA determined by BIAcore ™ biosensor. Apparent association (k a ) and dissociation (k d ) rate constants and dissociation equilibrium constants (K D ϭ k d /k a ) are means of 6 to 10 independent experiments. Association and dissociation rate constants were calculated using BIA evaluation 2.1 software provided by the manufacturer.  (14,16,17). This discrepancy is probably due to different assays and expression systems. To evaluate these differences, we analyzed HA binding activity of the recombinant aggrecan molecules with the same assays used in previous studies with recombinant proteins expressed in bacteria and in baculovirus systems. These assays included the HA-Sepharose binding and CPC precipitation methods. In the HA-Sepharose binding assay, the recombinant aggrecan proteins, AgW, AgMut11-14, 16 -18, and the control FLAG-laminin ␥1 bound to HA-Sepharose. In the CPC-precipitation assay, recombinant proteins bound to HA even under reducing conditions (data not shown). These results indicate that, under these assay conditions, recombinant aggrecan molecules interact with HA in a nonspecific manner and may not reflect an in vivo mechanism that involves the ternary structure of PTR. Since HA is highly negatively charged, basic amino acid-rich regions of the proteins have a tendency to interact nonspecifically with HA.
It is of interest to note that CD44 and TSG-6, which each contain a single PTR, bind to HA by enzyme-linked immunosorbent assays (22) and CPC coprecipitation methods (13). It has also been shown that the HA binding activity of CD44 is regulated by glycosylation levels and by its clustering on the cell surface (for review, see Ref. 2). Different glycosylation levels of CD44 are achieved by alternative splicing of mRNA. The correlation of the clustering and HA binding activity of CD44 suggests that dimerization of the molecule may be involved in its HA binding activity. The sequence of CD44 shows little homology to link protein or to the G 1 domain of aggrecan except for a PTR motif. It is also possible that the flanking sequence of the PTR of CD44 may contribute to HA binding of a single PTR. It would be of interest to know whether dimer formation of CD44 is a prerequisite for HA binding. The function of TSG-6, a secreted protein found in arthritic joints has not yet been identified. Since its domain structure is more similar to CD44 than to link protein, TSG-6 may bind to HA through a mechanism similar to that of CD44.
The BIAcore TM instrument allows real-time interaction analysis without any labeling such that association and dissociation rate constants can be measured directly. We used a fast flow rate and a low concentration of HA on the sensor chip surface for kinetic measurements. These measures minimize the effect of size and diffusion rate differences among these proteins that would affect the rate constant calculations. Our data indicated that binding of AgW and native link protein to HA have a similar relatively fast on-rate despite the fact that AgW is three times larger than link protein. This suggests that diffusion rates do not play a role in the kinetic rates. AgW has a moderately fast off-rate while link protein has the slowest off-rate of all the recombinant proteins examined in this study. Thus, the affinity of AgW to HA (K D , 0.226 M), which is consistent with previous report (36), is about three times lower than link protein (K D , 0.089 M). The half-life (t1 ⁄2 ) for dissociation of AgW from the complex is given by the equation t1 ⁄2 ϭ ln2/k off . The half-life of the AgW-HA complex and link protein are calculated as 175 s and 529 s, respectively. We tested the HA binding property of all the other recombinant proteins. Only AgMut12 and AgMut13 bound to HA, although their affinity to HA is significantly lower than that of either AgW or link protein.
Proteins that bind to HA without a PTR have been identified such as the receptor for HA-mediated motility (37), the intercellular adhesion molecule-1 (35), and the cumulus extracellular matrix stabilizing factor (39). Receptor for HA-mediated motility binds to HA with high affinity (K D ϭ 10 Ϫ8 M) via clusters of basic amino acids in the molecule. Denatured and reduced receptor for HA-mediated motility can bind to HA in the transblot assay, suggesting that ionic interactions between the two molecules appear to be more important (28). Application of the BIAcore TM biosensor system to HA binding analysis for these molecules may lead to a better understanding of the binding properties.
We found that the A subdomain significantly enhanced the HA binding activity of the aggrecan G 1 domain. The A subdomain forms an Ig-fold which consists of 7 ␤-strands. Tryptophan in the strand C and a disulfide bond are important for its conformation (40). Replacement of the tryptophan with alanine indicates that the ternary structure of the A loop is necessary for its enhancing effect on the HA binding function of the G 1 domain. The Ig-fold contains loops that correspond to hypervariable regions of immunoglobulin. A Thr-Thr-Ala-Pro sequence in the A loop located in the loop corresponding to L1 of immunoglobulin has an O-linked carbohydrate side chain(s) (35,41) (6). Substitution of these threonine residues with alanines to remove possible O-linked carbohydrates decreased aggrecan HA binding activity, suggesting that carbohydrate side chain(s) in the A loop may fortify the HA binding affinity of the G 1 domain. Our results demonstrate, for the first time, the significant contribution of the N-linked and O-linked carbohydrate side chains of the G 1 domain for its HA binding function. It has been reported that the carbohydrate side chains become elongated in preference to keratan sulfate chains with age (34,41). Carbohydrate side chains, by regulating charge interactions in the microenvironment, may enhance interactions with HA and thereby strengthen the cartilage matrix structure. Studies with mutations in glycosylation sites of CD44 have also suggested that glycosylation regulates its lectin activity (42).
It is intriguing that the aggrecan G 2 domain did not bind to HA. Our results confirmed the previous finding that the G 2 domain of pig aggrecan failed to bind to HA (43). Attachment of the A subdomain at the N terminus of the G 2 domain (Ag-Mut20) did not enhance HA binding function. The G 2 domain shows a high homology of amino acid sequence to B-BЈ of the G 1 domain (44) and contains the same disulfide bond patterns as those of the G 1 domain (45). It has been suggested that an extra N-linked oligosaccharide may exist in the BЈ subdomain of G 2 , 4 amino acid residues after the first cysteine residue, and may disrupt the folding of the domain (46). Creation of chimeric proteins and substitution mutations would give us clues for the identification of critical differences between G 1 and G 2 for HA binding.