Evidence for the covalent binding of SHAP, heavy chains of inter-alpha-trypsin inhibitor, to hyaluronan.

We previously showed that serum-derived 85-kDa proteins (SHAPs, serum-derived hyaluronan associated proteins) are firmly bound to hyaluronan (HA) synthesized by cultured fibroblasts. SHAPs were then identified to be the heavy chains of inter-α-trypsin inhibitor (ITI) (Huang, L., Yoneda, M., and Kimata, K.(1993) J. Biol. Chem. 268, 26725-26730). In this study, the SHAP•HA complex was isolated from pathological synovial fluid from human arthritis patients. The SHAP•HA complex was digested with thermolysin, followed by CsCl gradient centrifugation. The HA-containing fragments thus obtained were further digested with chondroitinase AC II and subjected to TSK gel high performance liquid chromatography (HPLC). Peptide-HA disaccharide-containing fractions (the SHAP•HA binding regions) were further purified by reverse phase HPLC. Major peaks were analyzed by protein sequencing and mass spectrometry (electrospray ionization mass spectrometry and collision induced dissociation-MS/MS). By comparison with the reported C-terminal sequences of the human ITI family, the peptides were found to correspond to tetrapeptides derived from the C termini of heavy chains 1 of and 2 of inter-α-trypsin inhibitor (HC1 and HC2), and heavy chain 3 of pre-α-trypsin inhibitor (HC3), respectively, and a heptapeptide from HC1. Mass spectrometric analyses suggested that the C-terminal Asp of each heavy chain was esterified to the C6-hydroxyl group of an internal N-acetylglucosamine of HA chain. This report is the first demonstration to give evidence for the covalent binding of proteins to HA.

Hyaluronan (HA), 1 has been found as a ubiquitous compo-nent of the extracellular matrices of many tissues and in body fluids, including the vitreous body, synovial fluid, lymph, and blood (1)(2)(3)(4). It has been suggested that HA plays an important role in many biological processes, such as gamete maturation, tissue morphogenesis, cell migration, and cell proliferation (5)(6)(7)(8). HA is also involved in angiogenesis, wound healing, tumor invasion, and pathophysiological responses of tissues to inflammation (9 -12).
With regard to functional importance, a large number of HA-binding proteins have been reported, an important subset of which have highly homologous sequences for HA binding. These include link proteins (13), hyaluronectin (14), glial HAbinding protein (15), HA-binding proteoglycan such as aggrecan, PG-M/versican (16), and CD44 (17). These are the proteoglycan tandem repeat families of HA-binding proteins. CD44 is a typical example of the family. Variant forms of CD44 generated by alternative splicing may have individual functions such as lymphocyte homing and tumor cell metastasis (18,19). Tumor necrosis factor-stimulated gene-6, another new member of this family, is tumor necrosis factor or interleukin-1-inducible and was recently shown to bind covalently to inter-␣-trypsin inhibitor (ITI) (20).
We previously showed that serum-derived HA-associated proteins (SHAPs) appear to bind covalently to HA (21), and therefore to mediate the binding of HA to cell surface and other extracellular molecules, and might be one of the serum factors involved in the HA metabolism in cultured fibroblasts. Our recent study demonstrated that SHAPs are identical to the heavy chains of ITI, and the SHAP⅐HA complex could be formed by incubating serum with exogenous HA under physiological conditions (21,22).
ITI is a plasma protease inhibitor consisting of three genetically different peptides, a light chain (bikunin) and two heavy chains (HC1 and HC2) (23)(24)(25)(26)(27)(28). The three peptides are covalently cross-linked by a chondroitin sulfate chain (29 -35) in that each heavy chain is covalently bound to the chondroitin sulfate chain derived from the light chain by a unique ester bond between the carboxyl group of the C-terminal Asp of the peptides and C-6 hydroxyl group of an internal GalNAc of the chondroitin sulfate chain (34,35). Similar linkage structures exist in pre-␣-trypsin inhibitor (P␣I). P␣I is composed of one heavy chain (HC3) and one light chain which are covalently linked to each other by a chondroitin sulfate chain (30).
In the present study, we prepared SHAP⅐HA complexes from human synovial fluid of arthritic patients. We analyzed the binding region by protein sequencing and electrospray ionization mass spectrometry (ESI-MS). We give evidence to show that SHAP binds covalently to HA.

EXPERIMENTAL PROCEDURES
Materials-Streptomyces hyaluronidase was obtained from Seikagaku Corp., Tokyo, Japan; protease-free chondroitinase AC II (from Arthrobacter aurescens) was a gift from Dr. K. Yoshida of Seikagaku Corp.; thermolysin (three times crystallized), ⑀-amino caproic acid, Nethylmaleimide, and phenylmethylsulfonyl fluoride were purchased from Nacalai Tesque, Kyoto, Japan; rabbit immunoglobulins to human ITI, and goat immunoglobulins to rabbit immunoglobulins were from DAKO Japan, Kyoto, Japan; DNase I was from Sigma; micro-BCA protein assay reagent was from Pierce; all reagents for protein sequence analysis were from Applied Biosystems, Chiba, Japan.
Preparation of the SHAP⅐HA Complex from Pathological Synovial Fluid-The method was essentially the same as described previously (21,22) with a slight modification. Briefly, to the pathological synovial fluid obtained from osteoarthritic patients (250 ml), the same volume of the extraction buffer was added. The extraction buffer contained 0.2 M Tris-HCl, pH 8.0, 8 M guanidine HCl, 10 mM EDTA, 10 mM amino caproic acid, 10 mM N-ethylmaleimide, and 2 mM phenylmethylsulfonyl fluoride. The extraction was performed at 4°C overnight with stirring, and the mixture was diluted and brought to a density of 1.35 g/ml with solid CsCl. A density gradient was established by centrifugation at 130,000 ϫ g, 10°C, for 48 h (Hitachi, 70P-72). The gradient was partitioned into 16 fractions, and the contents of HA and proteins in each fraction were determined by carbazole reaction and micro-BCA method, respectively (21,36). The lower one-fourth of the gradient which contained above 90% of the total HA was pooled. Furthermore, a second centrifugation with an initial density of 1.38 g/ml and subsequent assays for HA and proteins in partitioned fractions were performed as described above. The lower half of the gradient was pooled and subjected to a third centrifugation with an initial density of 1.45 g/ml, and subsequent assays for HA and proteins in partitioned fractions were as described above. The fractions from the bottom, nos. 4 -10, which contained most of the total HA, were pooled and subjected to ethanol precipitation to collect the SHAP⅐HA complex (21).
Preparation of the SHAP⅐HA Binding Region-SHAP⅐HA complex (160 mg) was dissolved in 10 mM Tris HCl, pH 7.5, containing 10 mM KCl and 3 mM MgCl 2 and treated with 300 units (100 g) of DNase I at 37°C overnight with agitation. After CaCl 2 was added (the final concentration 2 mM), the mixture was treated with 100 g of thermolysin at 37°C with agitation. After 6 h, another 100 g of thermolysin were added, and further incubation for 3 h was performed. The same volume of 8 M guanidine HCl (extraction buffer) was added to the digested mixture. After the density was brought to 1.35 g/ml by adding solid CsCl, centrifugation at 130,000 ϫ g for 48 h was performed. The HA content in the partitioned fractions was determined as described above.
The HA-containing fractions were precipitated with ethanol and then subjected to digestion with chondroitinase AC II (67 milliunits/mg of HA) in 50 mM sodium acetate, pH 6.0, at 37°C overnight. The digested mixture was further subjected to TSK gel HPLC. Two TSK G3000PWXL (7.8 mm ϫ 30 cm, Tosoh) columns were linearly connected for the better separation. The columns were eluted with 0.2 M NH 4 HCO 3 at a flow rate of 0.5 ml/min. The absorbance was measured at 214 nm. The peptide-containing peak was eluted just before the main HA disaccharide peak. The first peak was collected and concentrated by lyophilization. The second HPLC using the same columns was carried out to completely remove a trace of the HA disaccharide product. The fractions containing the SHAP⅐HA binding region were collected and lyophilized.
Further Purification of the SHAP⅐HA Binding Region by Reverse Phase HPLC-The SHAP⅐HA binding region was further purified by reverse phase HPLC. The column (TSK ODS-120T, 4.6 ϫ150 mm, Tosoh) was equilibrated with 0.06% (v/v) trifluoroacetic acid in water. The peptides were eluted by an acetonitrile gradient 0 to 13% acetonitrile in 0.06% (v/v) trifluoroacetic acid in water. The flow rate was 0.5 ml/min, and the absorbance was measured at 214 nm.
Protein Sequence Analyses-Automated Edman degradation was carried out in an Applied Biosystems model 476A sequencer with online phenylthiohydantoin analysis using an Applied Biosystems model 120A HPLC apparatus. The instruments were operated as recommended in the user bulletins and manuals distributed by the manufacturer.
ESI-MS and Collision Induced Dissociation (CID) MS/MS Analyses-Samples were dissolved in 1%-trifluoroacetic acid aqueous solution admixed with acetonitrile (3:7, v/v), typically at the concentration of 10 pmol/l, and introduced by a mechanical infusion through a microsyringe into the electrospray needle at a flow rate of 1 l/min. Negative ion ESI mass spectrometry was performed using a Finnigan MAT TSQ 700 triple stage quadrupole mass spectrometer equipped with an electrospray ion source (Analytica of Branford). A potential difference of 3 kV was applied against the ground potential of the needle. Hot nitrogen gas was used as the counter gas to desolve sample droplets in the ion source.
Argon gas was used at 0.13 Pa as the collision gas at 20 eV for CID-MS/MS spectra.

Isolation of SHAP⅐HA Complex from Human Synovial Flu-
id-Synovial fluid from knee joints of osteoarthritic patients was used as a source of the complex. Since the synovial fluid was highly viscous and contained a variety of proteins at high concentrations, dissociative conditions (in 4 M guanidine HCl solution) were used for purification. A series of three isopycnic centrifugations with initial densities of 1.35, 1.38, and 1.45 g/ml, respectively, were performed. The density and the contents of HA and protein of each fraction after the third centrifugation suggested the firm association of some proteins with HA ( Fig. 1). Fraction nos. 4 -10 which contained a constant ratio of protein to HA (about 0.06) were pooled, precipitated with ethanol, and stored as the SHAP⅐HA complex preparation.
SDS-PAGE and immunoblotting of the SHAP⅐HA complex thus prepared from human synovial fluid were performed to confirm that the complex was identical with the one prepared from the incubation mixture of human serum with HA as described previously (22) (Fig. 2). Proteins in the preparation from the synovial fluid were retained in the starting gels of SDS-PAGE, indicating that the complex in the synovial fluid preparation remained undissociated under both dissociative conditions of 4 M guanidine HCl during the centrifugation and 1% (w/v) SDS during the PAGE ( Fig. 2A). This was also the case with the SHAP⅐HA complex prepared from the incubation mixture of human serum with HA as described previously (22). Treatment of both complex preparations with HA-degrading enzymes, such as protease-free Streptomyces hyaluronidase or chondroitinase AC II released the proteins that corresponded to HC1 and HC2 of ITI (two bands with the higher mobility and the lower mobility, respectively), judging from both the immunoreactivities and the mobilities of proteins in the gels (Fig. 2,   FIG. 1. Isolation of the SHAP⅐HA complex from pathological synovial fluid of human arthritis patients. Guanidine HCl extract of pathological synovial fluid was subjected to a series of CsCl density gradient centrifugations as described under "Experimental Procedures." After the third centrifugation (initial density, 0 ϭ 1.45 g/ml), the gradient was partitioned into 16 fractions. Contents of HA and proteins in each fraction were determined by carbazole reaction and micro-BCA method, respectively. B and C). Treatment with alkali (0.02 M NaOH) also caused their release from both the SHAP⅐HA complex preparations (Fig. 3). Therefore, synovial fluid of human arthritic patients contained a SHAP⅐HA complex that was identical in both molecular structure and properties to the complex produced by incubation of HA with serum. In addition, when the preparation of the SHAP⅐HA complex from the pathological synovial fluid was digested with V 8 -protease and then subjected to SDS-PAGE, two major peptide bands resulted which have partial N-amino acid sequences identical with HC1 and HC2 of human ITI, respectively (data not shown). Taken together, it is very likely that the protein-HA complex obtained from human synovial fluid corresponds to the SHAP⅐HA complex that we had described previously (22).
Intact ITI and metabolic products of ITI were also present in synovial fluid (data not shown). In our purification procedures, the SHAP⅐HA complex was recovered in the bottom fractions after the first and second CsCl isopycnic centrifugation, while ITI, P␣I, and most of bikunin were in the upper fractions (data not shown). Complete separation from those proteoglycans were appraised by SDS-PAGE of the purified preparation before and after the treatment with either Streptomyces hyaluronidase or chondroitinase AC II and subsequent immunoblotting with anti-human ITI (Fig. 2). Since Streptomyces hyaluronidase degrades HA only while the chondroitinase degrades both HA and chondroitin sulfate, the fact that no distinctive difference was observed in the immunostaining pattern of protein bands between the Streptomyces hyaluronidaseand chondroitinase AC II-treated samples indicated that the purified preparation from the synovial fluid was neither contaminated with ITI nor with P␣I. The slight difference in mobility of the stained bands between the two samples could be explained by the difference in size of the products between the two enzymes. It is known that major digestion products of HA with Streptomyces hyaluronidase are tetramers and hexamers while the chondroitinase AC II digestion of HA yields dimers. In addition, there were also some differences between the Coomassie Blue staining and immunostaining patterns (Fig. 2, A  and B). These differences could be explained by the apparent differences in the immunoreactivity of the anti-human ITI antibodies between HC1 and HC2.
Preparation of the SHAP⅐HA Binding Region from SHAP⅐HA Complex-Since DNA has a density similar to HA in the CsCl centrifugation, the possibility that the SHAP⅐HA complex preparation was contaminated with DNA was not excluded. If this be the case, the DNA might disturb the subsequent processes. Therefore, the preparation was first subjected to DNase I digestion. The SHAP⅐HA complex was then digested with thermolysin. It was also possible that the bound HA protects SHAP from proteolysis (21). Therefore, a high enzyme/substrate ratio (1:40 in weight) and a long incubation time (9 h) were used to ensure complete digestion. The HA-containing products were purified by CsCl isopycnic centrifugations. The products were further digested with chondroitinase AC II overnight. In our preliminary test Streptomyces hyaluronidase digestion yielded not only different sizes of HA oligosaccharides (tetramers and hexamers) but also the SHAP⅐HA binding regions with these HA oligosaccharides, which caused difficulties in subsequent separation and purification of the binding regions from free HA oligosaccharides and those from each other. The digestion with chondroitinase AC II, instead, yielded the SHAP⅐HA binding regions containing only unsaturated disaccharide of HA. Subsequent HPLC on linearly connected two TSK gel columns gave a successful resolution of the binding regions from the HA disaccharide (Fig. 4A, peak I for the binding regions and peak II . The samples were electrophoresed on SDS gel (9% gel, under nonreducing conditions) and the gels were stained with Coomassie Blue (A). About 1/20 of the same set of the samples were electrotransferred to nitrocellulose membrane after SDS-PAGE, and immunoblotted with antibodies to human ITI. The immune complexes were visualized by enhanced chemiluminescence assay (B). Aliquots of SHAP⅐HA complex prepared from the incubation mixture of human serum with HA were treated, and immunoblotted the same way as described above (C). Since the antibodies used in this experiment were more reactive to HC2 of ITI than the other heavy chains, the observed differences between the Coomassie Blue staining and immunostaining patterns could be explained by such differences in the immunoreactivity. for the HA disaccharide plus salt). The binding regions were rechromatographed onto the same column to ensure the complete separation from the HA disaccharide (Fig. 4B, peak 1 for the binding regions, peak 2 for the residual HA disaccharide, and peak 3 for buffer salt).
The chromatography of peak II in Fig. 4A on SAX 10 column (37) indicated that none of any disaccharide products derived from chondroitin, chondroitin 4-sulfate and chondroitin 6-sulfate other than the HA disaccharide was detected in this fraction, judging from the elution positions of the standard disaccharides (data not shown). The result further confirmed no contamination with ITI and P␣I in the SHAP⅐HA complex prepared from the human synovial fluid.
Further purification of the SHAP⅐HA binding regions was performed by applying the peak 1 from the second TSK gel HPLC onto C 18 reverse phase HPLC. Elution with an acetonitrile gradient as described in the experimental procedures yielded 12 major peaks detected by the absorption at 214 nm and numbered by numerical order (Fig. 5).
Peptide Sequences of the SHAP⅐HA Binding Regions-Amino acid sequences of peptides, if any, in each peak of the reverse phase HPLC were determined by an automated amino acid sequencer. Peaks 1, 2, 8, 9, and 12 did not contain any detectable amino acid residues. The amino acid sequences determined for peaks 3, 4, 5, 6, 7, 10, and 11 are shown in Table I. The fourth cycle for the shorter peptides and the seventh cycle for the longer peptides gave no peak corresponding to any standard amino acid residue under the employed conditions. Although there were other possibilities, the result suggested a modification of the C-terminal amino acid residues of those peptides. The residues might be covalently linked to hyaluronan in a similar way to the binding of the heavy chains to the chondroitin sulfate chain of bikunin in ITI family molecules (30,34). In comparison with the C-terminal sequences of the heavy chains of human ITI and P␣I, the C-terminal amino acid residue of each peptide might be Asp. In such a case, the SHAP⅐HA binding regions in those peaks would be expected to contain three different C-terminal tetrapeptides derived from HC1, HC2, and HC3, respectively, and a longer heptapeptide derived from HC1 (Table II). As a whole, of the SHAP⅐HA complexes in human pathological synovial fluid, SHAPs were considered to be derived mostly from HC1 and HC2 of ITI and less from HC3 of P␣I (Table II).

ESI-MS and CID-MS/MS Analyses of the SHAP⅐HA Binding
Regions-Structures of SHAP⅐HA binding regions were studied by ESI-MS and ESI CID-MS/MS in the negative ion mode. Peaks 3, 4, 5, 6, 7, 10, and 11 of C 18 reverse phase HPLC (see Fig. 5) were subjected to the analyses. ESI-MS was tried first in the positive ion mode, but clear spectra were not obtained. On the other hand, the negative ion mode provided prominent molecular ions at m/z 835 for peak 4, at m/z 808 for peak 5, at m/z 835 and 764 for peak 6, at m/z 808 for peak 7, and at m/z 1065 for peaks 10 and 11, respectively (data not shown). No significant molecular ion was obtained around the molecular ion region at m/z 600-1100 for peak 3, suggesting that major molecules in peak 3 may not be related to the expected molecules. Molecular masses thus obtained were all 361 atomic mass units higher than the calculated molecular masses of (M Ϫ H) Ϫ for respective peptides, and this increment in the molecular masses appeared to correspond to the covalent glycosylation by a dehydrohexulonosyl N-acetylhexosamine (see Table II). Further structural studies were carried out by CID-MS/MS having the (M Ϫ H) Ϫ as precursors for samples of peak 6 (m/z 835) and peak 7 (m/z 808) (Figs. 6 and 7, respectively). Daughter ions at m/z 227, 341, 358, 456, and 474 for peak 6 represented the tetrapeptide sequences as (VE)-N-D, while ions at m/z 212, 313, and 429 for peak 7 similarly indicated the structure as (VD)-T-D. Lacking a negative charge center, the N-terminal valine did not give a negative daughter ion as a single unit. Ring-opening daughter ions at m/z 501 and 559 shown in Fig. 6 for peak 6 suggested that the peptide VENDwas covalently linked to the C6-hydroxyl group of the N-acetylhexosamine. In addition, the above-mentioned carboxylate ion at m/z 474 for peak 6 suggested that the linkage was an ester between one of the carboxylates of Asp and the C6-hydroxyl group of the N-acetylhexosamine, rather than an amide (see Fig. 6, inset). On the other hand, peak 7 did not give the FIG. 4. TSK gel HPLC of thermolysin-and chondroitinase AC II-digested products of SHAP⅐HA complex. SHAP⅐HA complex was treated with thermolysin and subjected to CsCl gradient centrifugation. HA-containing fractions were pooled and treated with chondroitinase AC II as described under "Experimental Procedures." The digested mixture was loaded onto two linearly connected TSK G3000 HPLC columns and eluted with 0.2 M NH 4 HCO 3 (A). The peptide-containing fraction (peak I) was collected and rechromatographed on the same columns (B). The peptide-containing fraction (peak 1) was collected and lyophilized for the next purification step.
FIG. 5. Reverse phase HPLC of the TSK gel fraction for the purification of the SHAP⅐HA binding region. The lyophilized peak 1 of the second TSK gel HPLC was applied onto a C 18 reverse phase HPLC, eluted by an acetonitrile gradient in 0.06% (v/v) trifluoroacetic acid as described under "Experimental Procedures." The absorbance was measured at 214 nm. The separated peaks were designated as peaks 1-12 in numerical order. Each peak was collected and lyophilized, respectively, for protein sequencing and mass spectrometry analysis.
carboxylate ion at the ester linkage. However, the ion at m/z 489 in Fig. 7 again indicated a structure in which the peptide was linked to the C6-hydroxyl group of the N-acetylhexosamine. Daughter ions at m/z 659 for peak 6 (Fig. 6) and at m/z 632 (with m/z 614) for peak 7 (Fig. 7) indicated an elimination of the terminal dehydroxyhexulonic acid, which suggests that the sugar unit directly linked to the peptides is N-acetylhexosamine in both samples. Thus, the negative charge was found localized in an acidic side chains in the peptides, and the sugar side negative ions were not observed.
CID MS/MS analysis of (M Ϫ H) Ϫ at m/z 764 in peak 6 suggested that the same linkage occurs between HC3 and the HA disaccharide (data not shown). However, its limited sample amount prevented us from carrying out detailed structural studies.
Taken together, the analyses demonstrated linkages be-   Table I) with cDNA sequences of heavy chains of ITI and P␣I from gene bank. The amino acids in brackets were predicted from cDNA sequences, and the presence was confirmed by mass spectrometry. Percentage of each peptide was calculated from the yields of amino acids in the sequence data.

Peaks
Starting sample 4 5 6 7 10 11 Peptides tween the C-6 of the reducing terminal N-acetylhexosamine residue of HA disaccharide and a carboxyl group of the Cterminal aspartic acid of the peptides derived from the heavy chains, HC1, HC2, and HC3 of ITI and P␣I. DISCUSSION Enghild et al. (30,34) showed that three different heavy chains (HC1, HC2, and HC3) were covalently linked to bikunin by chondroitin sulfate chain in ITI and P␣I. We have shown in this study that a similar linkage structure exists between SHAPs (HC1, HC2, and HC3) and HA. Mass spectrometic analyses revealed esterification of the carboxyl group of the C-terminal Asp with C-6 hydroxyl group of reducing terminal N-acetylhexosamine of the unsaturated HA disaccharide. Thus, we conclude that SHAPs (heavy chains of ITI and P␣I) are covalently linked to HA by the esterification with the C6-hydroxyl groups of N-acetylglucosamine of HA via the C-terminal Asp.
The present study using ESI CID-MS/MS technique did not provide information to show which of the two carboxylate groups in the C-terminal Asp participated in the ester linkage formation. This is because the low energy collision-induced dissociation in the quadrupole instrument under our conditions failed to cleave C-C bonds in the aspartic acid. We assume tentatively that the ␣-carboxylate is the one which covalently bonds to the C6-hydroxyl group of the N-acetylhexosamine by an ester linkage, by analogy to the previously reported structure for the interchain linkage between the heavy chains and the light chain (bikunin core protein) via a chondroitin sulfate chain originating from the light chain in ITI or P␣I (30,34). This suggests that SHAP⅐HA complex may be formed simply by transferring the heavy chains from the chondroitin sulfate to HA (substitution reaction of HA for the chondroitin sulfate). However, the mechanism for the reaction has not been determined yet.
We noted that two fractions with different retention time in reverse phase HPLC contained peptides with the identical amino acid sequence (for example, peaks 4 and 6 in Tables I  and II). Since there was no significant difference observed between those two fractions by peptide sequencing or by ESI CID-MS/MS, this could be due to difference of the anomeric configuration of hydroxyl group at the reducing end (␣ and ␤) of the HA disaccharide.
In mouse ovulation, preovulatory synthesis of hyaluronan within the cumulus mass plays an important role for cumulus expansion (8). Recently, Chen et al. (38,39) reported that the cumulus extracellular matrix stabilizing factor in fetal bovine serum is a member of the ITI family, and that stabilizing ability is achieved through its direct binding to HA, which is sensitive to ionic strength and has a dissociation constant of 1.9 ϫ 10 Ϫ8 M at pH 7.2. Therefore, the properties of the interaction of cumulus extracellular matrix stabilizing factor with HA appear to be different from those for the formation of SHAP⅐HA complex.
The tight binding of ITI to HA in human pathological synovial fluid was firstly reported in 1965 (40). The present results show that the formation of a covalent linkage is involved in this binding. TSG-6, a 35-kDa glycoprotein of the proteoglycan tanden repeat HA-binding family, also exists in pathological synovial fluid of patients with arthritis (41). Recently, Wisniewski et al. (20) have shown that TSG-6 forms a covalently bound complex (120 kDa) with serum ITI. This TSG-6⅐ITI complex was formed rapidly even in the apparent absence of other proteins at 37°C, but not at 4°C. TSG-6 appeared to form a direct covalent bond to the chondroitin 4-sulfate chain of ITI for the stability of the TSG⅐ITI complex (20). In our study, we have not detected any peptides derived from TSG-6 in the SHAP⅐HA binding regions prepared from human pathological synovial fluid. Therefore, TSG-6 may not be involved in the formation of SHAP⅐HA complex in synovial fluid.
It is interesting to note that the formation of the SHAP⅐HA complex from HA and ITI or P␣I is accompanied by the release of bikunin. Bikunin contains two tandem repeats of Kunitztype domains, and the trypsin inhibitor activity of ITI is localized in this part (23,42). Bikunin is also identified in urine as UTI (urinary trypsin inhibitor), which was shown to be a urine proteoglycan with molecular mass ranging from 40 to 45 kDa (43)(44)(45). Some tumor cells have UTI receptors (46). UTI and fragments derived from UTI by limited proteolysis efficiently inhibit tumor cell invasion and metastasis (46 -49). Therefore, the formation of SHAP⅐HA complex might be related to some defense mechanism from proteolysis.