Characterization of complexes formed between TSG-6 and inter-alpha-inhibitor that act as intermediates in the covalent transfer of heavy chains onto hyaluronan.

The high molecular mass glycosaminoglycan hyaluronan (HA) can become modified by the covalent attachment of heavy chains (HCs) derived from the serum protein inter-alpha-inhibitor (IalphaI), which is composed of three subunits (HC1, HC2 and bikunin) linked together via a chondroitin sulfate moiety. The formation of HC.HA is likely to play an important role in the stabilization of HA-rich extracellular matrices in the context of inflammatory disease (e.g. arthritis) and ovulation. Here, we have characterized the complexes formed in vitro between purified human IalphaI and recombinant human TSG-6 (an inflammation-associated protein implicated previously in this process) and show that these complexes (i.e. TSG-6 x HC1 and TSG-6 x HC2) act as intermediates in the formation of HC x HA. This is likely to involve two transesterification reactions in which an ester bond linking an HC to chondroitin sulfate in intact IalphaI is transferred first onto TSG-6 and then onto HA. The formation of TSG-6 x HC1 and TSG-6 x C2 complexes was accompanied by the production of bikunin x HC2 and bikunin x HC1 by-products, respectively, which were observed to break down, releasing free bikunin and HCs. Both TSG-6 x HC formation and the subsequent HC transfer are metal ion-dependent processes; these reactions have a requirement for either Mg2+ or Mn2+ and are inhibited by Co2+. TSG-6, which is released upon the transfer of HCs from TSG-6 onto HA, was shown to combine with IalphaI to form new TSG-6 x HC complexes and thus be recycled. The finding that TSG-6 acts as cofactor and catalyst in the production of HC x HA complexes has important implications for our understanding of inflammatory and inflammation-like processes.

Hyaluronan (HA) 1 is a high molecular mass polysaccharide that plays an important role in extracellular matrix structure and is central to a wide variety of physiological and pathological processes such as ovulation and inflammatory disease (1,2). This glycosaminoglycan, which is composed solely of a repeating disaccharide of glucuronic acid and N-acetylglucosamine, can become covalently modified by the attachment of heavy chains (HCs) from the serum proteoglycans of the inter-␣-inhibitor (I␣I) family; this modification is likely to alter both its matrix and solution properties (3)(4)(5). For instance, the formation of HC⅐HA (also known as the serum-derived hyaluronan-associated protein⅐HA complex (5)) is critical for the stabilization of a nascent HA-rich matrix that is generated around the oocyte of placental mammals prior to ovulation (6 -8). The matrix expansion of the cumulus-oocyte complex (COC) is required for successful fertilization in vivo (9). HC⅐HA complexes are also formed under inflammatory conditions, e.g. in synovial fluids of arthritis patients (10,11), where this modification causes HA to become more aggregated (4,12,13), which is likely to alter its hydrodynamic size and rheological properties and makes it more resistant to degradation by oxygen free radicals (14).
I␣I is an unusual proteoglycan that contains three protein chains (HC1, HC2, and the serine protease inhibitor bikunin) held together via a chondroitin sulfate (CS) moiety (5). The CS chain is attached to Ser 10 of bikunin via a standard glycosaminoglycan attachment, whereas the HCs are linked through ester bonds between carboxylate groups of their C-terminal aspartic acid residues and the C-6 hydroxyls of internal Nacetylgalactosamines in the CS chain (15)(16)(17). The HCs appear to be transferred onto HA by a transesterification reaction because they become linked via an ester bond from their C termini to the C-6 hydroxylates of GlcNAc residues in HA (3), i.e. analogous to the HC ester bond to CS in I␣I. An intact I␣I protein is clearly required for this process because female mice lacking the bikunin gene, which express the HCs but are unable to assemble I␣I (or the related pre-␣-inhibitor (P␣I), which has a different heavy chain (HC3) linked to bikunin (18)), do not form any HC⅐HA in their cumulus matrix. As a consequence, they are infertile (6,19). Mixing of purified HA and I␣I in vitro does not give rise to the formation of HC⅐HA (20), indicating that other molecules are involved (21). Recent work has revealed that TSG-6 (the protein product of tumor necrosis factor-stimulated gene-6; also known as TNFIP6 (tumor necrosis factor-induced protein-6) (22)) is involved. TSG-6 is an in-flammation-associated HA-binding protein composed mainly of contiguous Link and CUB modules (23)(24)(25) and appears to have an essential role in the transfer of HCs from I␣I onto HA (7,26,27). Most importantly, Fü löp et al. (7) showed that Tsg-6 Ϫ/Ϫ female mice are infertile due to their inability to form the HA-rich extracellular matrix that is essential for cumulus expansion, a phenotype that correlates with the total absence of HC⅐HA complexes in the ovaries of these animals; the administration of murine TSG-6, either as a recombinant protein or as a transgene, rescues the fertility of Tsg-6 null mice. In this study, HC3 (a component of P␣I), in addition to HC1 and HC2, was missing from the cumulus matrix, indicating that TSG-6 is also necessary for the transfer of HC3 onto HA (7).
The expression of TSG-6 is up-regulated during COC expansion in the mouse and rat (27)(28)(29)(30)(31)(32)(33), where the protein has been shown to co-localize with HA and I␣I in the cumulus matrix (30,31). Western blot analyses revealed that TSG-6 is present as a free protein (ϳ35 kDa) and as a species of ϳ120 kDa that is immunoreactive with both anti-TSG-6 and anti-I␣I antibodies (30,31). Characterization of this ϳ120-kDa band by mass spectrometry demonstrated that it contains TSG-6, HC1, and HC2, but not bikunin. On the basis of their molecular masses (i.e. TSG-6 is ϳ35 kDa, and each HC is ϳ80 -85 kDa), the ϳ120-kDa band is thought likely to comprise a mixture of TSG-6⅐HC1 and TSG-6⅐HC2 complexes (31). These complexes are not sensitive to chondroitinase, indicating that the CS chain of I␣I is not involved in their linkage, but they are cleaved by mild NaOH treatment, consistent with the presence of ester bonds. Therefore, the TSG-6⅐HC complexes may act as intermediates in the formation of HC⅐HA, as has been suggested (7). However, prior to the present study, this had not been investigated directly. Species of ϳ120 kDa can also be formed when human recombinant TSG-6 and human purified I␣I are incubated together in vitro (27,34,35), and TSG-6⅐I␣I complexes of this size have been detected in the synovial fluids of arthritis patients (36). However, the composition of the ϳ120-kDa species formed in vitro from human components, which was reported to contain TSG-6, HC2, and bikunin held together via a chondroitinase-sensitive linkage (34), indicates that it might represent a different type of complex (with TSG-6 replacing an HC on the CS chain) compared with that formed during ovulation in the mouse.
In this study, we have characterized the complexes formed in vitro between purified human I␣I and recombinant human TSG-6 as TSG-6⅐HC1 and TSG-6⅐HC2 and have shown that they act as intermediates in the formation of HC⅐HA. This is accompanied by the production of bikunin⅐HC1 and bikunin⅐HC2 by-products, which break down to generate free bikunin and HCs. Both TSG-6⅐HC complex formation and subsequent HC transfer are metal ion-dependent processes, having a requirement for either Mg 2ϩ or Mn 2ϩ . TSG-6, which is released upon the transfer of HCs from TSG-6 onto HA, was shown to combine with I␣I to generate new TSG-6⅐HC complexes and thus acts as a true catalyst for the formation of HC⅐HA.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-Human full-length TSG-6 protein (Q allotype) was expressed in Drosophila Schneider-2 cells and purified to homogeneity as described (35), and the protein concentration was determined by amino acid analysis. I␣I was purified from human serum (37), and its concentration was determined as described previously (38).
Formation of TSG-6⅐I␣I Complexes under Standard Conditions-In the standard assay, recombinant full-length TSG-6 (80 g/ml final concentration; 2.7 M based on a molecular mass of 30 kDa) was incubated with I␣I (320 g/ml final concentration; 1.8 M based on a molecular mass of 180 kDa) in 20 mM HEPES-HCl (pH 7.5), 150 mM NaCl, and 5 mM MgCl 2 in a total volume of 25 l for 2 h at 4°C (i.e. on ice). The effects of protein concentration, temperature, ionic strength, pH, and metal ions on complex formation were investigated by varying these parameters individually while keeping all other conditions constant. Unless otherwise stated, 7.5 l of each sample was analyzed on 10% (w/v) Tris/Tricine/SDS-polyacrylamide gels following reduction with 5% (v/v) ␤-mercaptoethanol in SDS protein sample buffer (5 min at 100°C), and gels were stained with Coomassie Blue.
Effect of TSG-6 and I␣I Protein Concentrations on TSG-6⅐I␣I Complex Formation-The amount of TSG-6 in the assay was varied from 1 to 8 g (40 -320 g/ml final concentration) while keeping the amount of I␣I constant at 8 g (320 g/ml final concentration). Alternatively, between 8 and 32 g (320 -1280 g/ml final concentration) of I␣I was used in the presence of 1 g of TSG-6 (40 g/ml final concentration). As a control, TSG-6 (1 g) or I␣I (8 g) was incubated alone under standard assay conditions. These samples were analyzed by SDS-PAGE and by Western blotting using a rabbit anti-human polyclonal antibody raised against TSG-6 (39) as described (35).
Effect of Temperature, pH, and Ionic Strength on TSG-6⅐I␣I Complex Formation-TSG-6⅐I␣I complex formation was compared at 4 and 37°C; the reactions were carried out for 0.5, 2, 5, 15, 30, 60, and 120 min. Assays were performed at pH values ranging from 4.0 to 8.0. Sodium acetate buffer was used for pH 4.0 and 5.0; MES-HCl buffer for pH 6.0 and 6.5; and HEPES-HCl for pH 7.0, 7.5, and 8.0 (all at a 20 mM final concentration as in the standard assay described above). To investigate the effect of ionic strength on TSG-6⅐I␣I complex formation, the assay was run in 20 mM HEPES-HCl (pH 7.5) in the presence of 50, 100, 150, 200, 250, or 300 mM NaCl. In these experiments, control samples of TSG-6 or I␣I alone were incubated at the appropriate temperature for 2 h.
Effect of Divalent Metal Ions on TSG-6⅐I␣I Complex Formation-The requirement for divalent metal ions during formation of the TSG-6⅐I␣I complex was tested by incubating TSG-6 and I␣I with 1 mM EDTA in the absence or presence of 5 mM MgCl 2 , CaCl 2 , CoCl 2 , or MnCl 2 . Additionally, CaCl 2 and CoCl 2 were co-incubated with MgCl 2 in the absence or presence of EDTA (i.e. 5 mM MgCl 2 with 1 mM CaCl 2 or CoCl 2 ; 5 mM MgCl 2 with 5 mM CaCl 2 or CoCl 2 ; and 1 mM MgCl 2 and 1 mM EDTA with 5 mM CoCl 2 ). Complex formation was also investigated at a range of MgCl 2 concentrations (i.e. 0, 0.1, 0.5, 1.0, 5.0, 10.0, and 20.0 mM).
The effect of metal ions on TSG-6⅐I␣I complex formation was also examined using serum as the source of I␣I. Mouse serum (20 l; Rockland Immunochemicals) was incubated with 2 g of human recombinant TSG-6 in the presence or absence of 2 mM EDTA and 5 mM CaCl 2 , MgCl 2 , or CoCl 2 in a 25-l reaction volume at 37°C for 2 h. Samples (6 l) were run on 4 -20% precast gels (Invitrogen) and analyzed by Western blotting as described previously (31) using a rabbit anti-human polyclonal antibody raised against TSG-6 (39).
Effect of Chondroitinase and NaOH Treatment on TSG-6⅐I␣I Complex Stability-TSG-6⅐I␣I complexes were formed under standard conditions (i.e. 2 h at 4°C; see above). Aliquots of the reaction mixture (7.5 l) or of I␣I incubated in the absence of TSG-6 were then diluted with an equal volume of water. Either 1 l of chondroitinase ABC lyase (10 milliunits; Seikagaku Corp.) or 1.5 l of 1 M NaOH (0.1 M final concentration) was then added, followed by incubation for 2 h at 37°C or for 10 min at room temperature, respectively; 1.5 l of 1 M HCl was added to the NaOH-treated samples. To these (and a chondroitinaseonly control (10 milliunits in 15 l of H 2 O) and an untreated TSG-6⅐I␣I complex (7.5 l ϩ 7.5 l of H 2 O), both incubated for 2 h at 37°C) were added 15 l of 2ϫ SDS protein sample buffer, followed by SDS-PAGE (with the whole sample loaded) or Western blotting (with one-third of the sample loaded) as described above.
Characterization of TSG-6⅐I␣I Complexes by N-terminal Sequencing-Complexes were formed under standard conditions and run on 10% (w/v) Tris/Tricine/SDS-polyacrylamide gels with or without chondroitinase (10 milliunits) or NaOH (0.1 M final concentration) treatment essentially as described above, except that the "untreated" sample was not incubated at 37°C for 2 h, and in all cases, twice as much protein was loaded per lane. The gels were electroblotted onto Hybond-P membrane (Amersham Biosciences) in 10 mM CAPS (pH 11) and 5% (v/v) methanol at 100 V for 2 h. The membranes were stained with Coomassie Blue for 75 s, destained in 50% (v/v) methanol for 10 min, and air-dried. Bands were excised (with reference to an identical control gel) and subjected to protein sequencing on an Applied Biosystems Procise 494A protein sequencer using standard "pulsed liquid for polyvinylidene difluoride-bound peptides" sequencing cycles. Bands that did not yield any visible sequence were excised from an identical gel and subjected to in-gel digestion with trypsin, followed by mass spectrometric analysis as described previously (31).
Formation and Characterization of I␣I⅐HA Complexes-TSG-6⅐I␣I complexes were formed under standard conditions (2 h at 4°C), except that HA was included in the reaction; 1 g of medical grade low molecular mass polymeric HA (p-HA; ϳ120 kDa; Genzyme Corp.) or 1 g of HA 14-mer (HA 14 ; 2673 Da) prepared as described (40)) was added to the assay, and samples were analyzed by Tris/Tricine/ SDS-PAGE and protein sequencing as described above. For sequence analysis of the p-HA sample, three identical lanes were loaded (7.5 l of reaction mixture), followed by transfer onto Hybond-P (in 10 mM CAPS (pH 11) for 3 h at 100 V) and staining/destaining as described above. Equivalent bands (at the interface of the stacking and resolving gels) were excised from the three lanes and combined for sequencing. Samples containing p-HA (or I␣I alone) were also treated with Streptomyces hyaluronidase (Seikagaku Corp.) prior to SDS-PAGE analysis. The reaction mixture (7.5 l) was diluted with an equal volume of water to which 1 l of enzyme (10 milliunits) was added, followed by incubation at 37°C for 2 h. I␣I treated with NaOH as described above was included as a control.
Effect of Divalent Metal Ions on I␣I⅐HA Complex Formation-TSG-6⅐I␣I complexes were preformed under essentially standard conditions in the presence of 0.109 mM MgCl 2 for 2 h at 4°C, followed by the addition of ϳ1 mM EDTA and incubation for an additional 30 min. Divalent metal ions (MgCl 2 , MnCl 2 , CaCl 2 , and CoCl 2 ) and 1 g of HA 14 were then added (20 mM HEPES-HCl (pH 7.5), 150 mM NaCl, 0.1 mM MgCl 2 , 1 mM EDTA, 5 mM M 2ϩ , and 40 g/ml HA 14 (final concentrations); 25-l final volume) and incubated for 2 h at 4°C. Control experiments were also performed in the absence of HA, EDTA, or metal ions. Alternatively, TSG-6⅐I␣I complexes were preformed in the presence of 5.4 mM MgCl 2 (5 mM final concentration), and then CoCl 2 was added (1 or 5 mM final concentration) and incubated for 30 min at 4°C before adding 1 g of HA 14 and incubating for 2 h at 4°C. In another experiment, TSG-6, I␣I, and HA 14 were incubated together for 2 h at 4°C in the absence and presence of 5 mM metal ion (MgCl 2 , MnCl 2 , CaCl 2 , or CoCl 2 ), but without the inclusion of EDTA; co-incubation of 5 mM MgCl 2 , MnCl 2 , or CaCl 2 with 5 mM CoCl 2 was also performed. Gel samples were prepared from 7.5-l reaction volumes as described above.
The effect of metal ions on the formation of I␣I⅐HA complexes was also examined using serum as the source of I␣I. Mouse serum (5 l), 5 g of high molecular mass HA (Healon GV, Pharmacia-Upjohn), and 250 ng of human recombinant TSG-6 were incubated for 24 h at 37°C in the presence or absence of 2 mM EDTA and 5 mM CaCl 2 , MgCl 2 , or CoCl 2 in 50 l of phosphate-buffered saline. Aliquots of each reaction mixture (10 l) were treated with 200 milliunits of Streptomyces hyaluronidase for 1 h at 37°C. Hyaluronidase-digested and untreated samples were run on 4 -20% precast gels and analyzed by Western blotting as described previously (31) using anti-I␣I polyclonal antibody (Dako Corp.).
Requirement of TSG-6⅐I␣I Complexes as Precursors in the Formation of I␣I⅐HA-TSG-6⅐I␣I complexes were preformed under essentially standard conditions (2 h at 4°C) and then incubated at 4°C for a range of times (0, 2, 4, and 22 h) before the addition of 1 g of HA 14 (1 l) or 1 l of water as a control and incubation for an additional 2 h. In a separate experiment, I␣I (8 g) and HA 14 (1 g) were incubated for 2, 4, or 22 h under standard conditions with various concentrations of TSG-6 (0, 0.04, 0.2, and 2 g) at 4 or 37°C. Samples were analyzed by SDS-PAGE as described above.

Formation of TSG-6⅐I␣I Complexes in Vitro-
We have shown previously that incubation of recombinant human full-length TSG-6 (expressed in Drosophila Schneider-2 cells and purified to homogeneity) with human I␣I (purified from serum) leads to the production of a stable complex of ϳ120 kDa recognized by an anti-TSG-6 antibody (35). Analysis here of essentially identical TSG-6/I␣I incubation mixtures by SDS-PAGE and Coomassie Blue staining revealed that three novel protein species were formed during this reaction: i.e. a diffuse ϳ130-kDa band (labeled A in Fig. 1) and an ϳ120-kDa doublet (labeled B/C). Of these, only the ϳ120-kDa doublet was found to be immunoreactive with the anti-TSG-6 antiserum (Supplemental Fig. S1), indicating that bands B and C correspond to the TSG-6⅐I␣I complex. This was subsequently confirmed by protein sequence analysis (see below). Fig. 1 shows that these three species formed at both 4 and 37°C and that bands A and C were clearly visible after as little as 30 s. Maximal amounts of the B/C doublet were formed after ϳ15 min at 37°C and after ϳ60 min at 4°C. At the higher temperature, an additional band of ϳ80 kDa (labeled I in Fig.  1) accumulated over time. This species ran at an identical position compared with HC1, and N-terminal sequencing revealed that it contained both HC1 and HC2 at an ϳ5:1 ratio (data not shown). Band I could therefore result from breakdown of the TSG-6⅐I␣I complex or of other HC-containing species (see below). In this regard, an ϳ80-kDa band (albeit much fainter) and a band of the same apparent molecular mass as species A were also seen when I␣I was incubated alone for 2 h at 37°C, indicating that there was some breakdown of I␣I at this temperature. Additionally, at 37°C, a high molecular mass doublet (near the top of the gel) also accumulated over time during the incubation of TSG-6 with I␣I, but significant amounts of this were not formed at 4°C; this doublet was recently identified as a high molecular mass species of I␣I containing additional HCs (41). Therefore, in subsequent experiments, incubations were generally carried out for 2 h at 4°C to maximize I␣I⅐TSG-6 complex formation while minimizing any production of high molecular I␣I species or of degradative reactions.
Determining the Optimal Conditions for TSG-6⅐I␣I Complex Formation-The effect of varying the protein concentration on the formation of the TSG-6⅐I␣I complex (i.e. B/C doublet) was investigated as described under "Experimental Procedures." Similar amounts of the B/C doublet were formed when 8 g of I␣I was incubated with between 2 and 8 g of TSG-6, whereas less was seen under the other conditions tested (i.e. 1 g of FIG. 1. TSG-6⅐I␣I complexes form rapidly at 4 and 37°C. TSG-6 (Q allotype) and I␣I were incubated together for various times at 4°C (upper panel) or 37°C (lower panel) and analyzed on Tris/Tricine/SDSpolyacrylamide gels stained with Coomassie Blue. Comparison of these reaction mixtures with TSG-6 or I␣I proteins incubated alone for 2 h revealed that they contained three novel species: an ϳ130-kDa band (designated species A) and an ϳ120-kDa doublet, where species B and C are the upper and lower bands, respectively. At 37°C, a band of ϳ80 kDa (corresponding to the running position of HC1) was also seen and is labeled I. Equivalent experiments done using the R allotype of TSG-6 (35) gave identical results (not shown). TSG-6 in the presence of 8 -32 g of I␣I) (Supplemental Fig.  S1). Assays conducted at different protein concentrations (e.g. 2 or 4 g of TSG-6 with 16 g of I␣I) (data not shown) did not lead to significantly increased levels of the B/C doublet. Therefore, 2 g of TSG-6 and 8 g of I␣I (i.e. the lowest concentrations that gave close to maximal complex formation) were used as the standard experimental conditions so as to minimize the amount of protein used in each assay.
From Fig. 2, it is clear that species A-C did not form when TSG-6 and I␣I were incubated at pH 4.0 or 5.0 (under otherwise standard conditions). There also appeared to be somewhat less complex formation at pH 6.0 compared with pH 6.5-8.0, which all led to similar levels of the B/C doublet. Experiments at pH 7.5 in which the ionic strength was varied revealed that there was a reduction in the amount of TSG-6⅐I␣I generated in 50 mM NaCl, whereas similar amounts of complex were formed between 100 and 300 mM NaCl (Fig. 2). Therefore, 150 mM NaCl and pH 7.5 (i.e. physiological serum conditions), which supported optimal complex formation, were used as the standard.
Preliminary experiments demonstrated that, in the absence of added metal ion, neither band A nor the B/C doublet was formed (Supplemental Fig. S2). However, these species were seen at similar levels when MgCl 2 was included in the reaction mixture at a wide range of concentrations (0.1-20 mM). MgCl 2 at the nominal concentration of 5 mM was chosen as the standard.
The above experiments therefore allowed us to determine the conditions under which optimal amounts of TSG-6⅐I␣I complex could be formed. Under the standard conditions chosen (i.e. 80 g/ml TSG-6 and 320 g/ml I␣I in 20 mM HEPES-HCl (pH 7.5), 150 mM NaCl, and 5 mM MgCl 2 ), which are close to physiological, I␣I was at a concentration similar to that found in normal human serum (i.e. 400 -500 g/ml (42)).
Effect of Divalent Metal Ions on TSG-6⅐I␣I Complex Formation-As described above, TSG-6⅐I␣I complexes could be formed in the presence of MgCl 2 , but did not form in the absence of added metal ion (Supplemental Fig. S2). Consistent with this, there was no complex formation in 0.1-50 mM EDTA (data not shown). Therefore, the effects of other divalent cations were investigated; TSG-6 and I␣I were incubated under standard conditions with 5 mM M 2ϩ (i.e. MgCl 2 , CaCl 2 , CoCl 2 , or MnCl 2 ) and 1 mM EDTA, which was added to chelate any metal ion impurities. Fig. 3 shows that complex formation occurred with MgCl 2 or MnCl 2 , but the B/C doublet was not seen in the presence of the other metal ions (e.g. Ca 2ϩ ). It should be noted that experiments conducted in the absence of any EDTA also demonstrated that either Mg 2ϩ or Mn 2ϩ (but not Ca 2ϩ or Co 2ϩ ) ions could support complex formation (data not shown). Interestingly, co-incubation of 5 mM CoCl 2 with 1 mM MgCl 2 , which was sufficient for optimal complex formation (Supplemental Fig. S2), was found to completely inhibit production of TSG-6⅐I␣I (Fig. 3). Assays including both CaCl 2 (1 or 5 mM) and MgCl 2 (5 mM) showed that Ca 2ϩ ions had no such inhibitory effect regardless of whether the experiments were done at 4 or 37°C (data not shown).
Experiments conducted with mouse serum as the source of I␣I showed that TSG-6⅐I␣I formed without any requirement for additional metal ions (Supplemental Fig. S3). When EDTA (2 mM) was included in the reaction mixture, complex formation was completely inhibited, which is consistent with the results obtained with purified components (see above). Addition of 5 mM MgCl 2 to these assays rescued the formation of TSG-6⅐I␣I as expected. Interestingly, 5 mM CaCl 2 or CoCl 2 in the presence of 2 mM EDTA also led to complex formation (Supplemental Fig. S3). The likely explanation for these results is that the addition of Ca 2ϩ or Co 2ϩ releases sufficient Mg 2ϩ from the EDTA (or serum proteins) to allow the reaction to proceed; this is consistent with the log K 1 values for the binding of these metal ions to EDTA.
Characterization of the TSG-6⅐I␣I Complex-As described above, the incubation of TSG-6 and I␣I under standard conditions led to the formation of three main species, i.e. the B/C doublet of ϳ120 kDa (band 2), which was immunoreactive with anti-TSG-6 antibody, and the ϳ130-kDa species A (band 1) (Fig. 4). These species were characterized by N-terminal sequencing (Supplemental Fig. S4). This analysis revealed that Approximately 50 or 10% complex formation was seen for 5 mM MgCl 2 co-incubated with 1 or 5 mM CoCl 2 , respectively (data not shown). species A (band 1) contained the three protein chains of I␣I (i.e. bikunin, HC1, and HC2) (Table I). However, no TSG-6 was detected, consistent with its lack of immunoreactivity with anti-TSG-6 antibody. Band 2, which was cut into upper and lower portions corresponding to species B and C, respectively, contained TSG-6 in addition to both HCs (but no bikunin). As shown in Table I, the upper band contained TSG-6 and HC2 in similar amounts (as based on the initial sequencing yields), with about half as much HC1 present. The converse was observed for the lower band, which gave equivalent initial yields of TSG-6 and HC1, but less HC2. Therefore, from this analysis and considering the molecular masses of TSG-6 and the three chains of I␣I, it seems likely that species B and C correspond to complexes of TSG-6 with HC2 and HC1, respectively, whereas species A comprises a mixture of bikunin linked to one or the other of the HCs; given the proximity of species B and C (Supplemental Fig. S4), it is perhaps not surprising that both HC1 and HC2 were detected in these bands.
Effect of Chondroitinase and NaOH on TSG-6⅐I␣I Complex Stability-The TSG-6⅐I␣I complexes were formed under standard conditions and then treated with either chondroitinase ABC lyase or NaOH. As shown in the Western blot in Fig. 4 (lower panel), species B/C (i.e. the TSG-6⅐HC complexes), which was detectable by an anti-TSG-6 antibody, disappeared completely upon mild NaOH treatment, but the band was not altered by chondroitinase. SDS-PAGE analysis showed that species A (band 1) and intact I␣I (either that remaining in the TSG-6/I␣I reaction mixture or an I␣I control) were degraded by both chondroitinase and NaOH (Fig. 4, upper panel). It is therefore likely that the HCs in these bikunin⅐HC complexes are linked via ester bonds to the chondroitin sulfate chain attached to bikunin, as they are in the intact parent molecule (15,16). Both chondroitinase and NaOH treatments led to the appearance of bands with apparent molecular masses of ϳ80 and ϳ85 kDa. N-terminal sequencing revealed that the upper of these (i.e. bands 6 and 8 in Fig. 4 and Supplemental Fig. S4) correspond predominantly to HC2, whereas the lower (bands 7 and 9) are essentially HC1 (Table I), with only a small amount (between 4 and 11%) of the alternative HC being detected in each of these bands. These HC1 and HC2 species were also seen upon treatment of I␣I alone (Fig. 4). A faint band of ϳ160 kDa also appeared following chondroitinase digestion of TSG-6/I␣I mixtures (band 3) or of I␣I alone, and this was shown by mass spectrometry to contain HC1 and HC2 (14 and 18 peptides identified, respectively) (data not shown), but did not contain any TSG-6 or bikunin. It is likely that this species represents a partial degradation product of I␣I in which the two HCs remain linked by CS.
Although chondroitinase treatment clearly did not degrade the B/C doublet (Fig. 4, lower panel), it did alter the separation of these species upon SDS-PAGE. As shown in Fig. 4  (upper panel), after treatment, the tight doublet (band 2) ran FIG. 4. TSG-6⅐I␣I complexes are degraded by NaOH, but are insensitive to chondroitinase treatment. TSG-6⅐I␣I complexes were formed under standard conditions and then treated with either chondroitinase ABC lyase (Ch'ase) or NaOH. These samples (along with untreated complex, I␣I, and chondroitinase ABC lyase controls) were analyzed by SDS-PAGE (upper panel) or Western blotting with anti-TSG-6 antibody (lower panel). Numbered bands (bracketed) indicate species that were characterized by protein sequencing following electroblotting of an essentially equivalent gel onto a Hybond-P membrane (Supplemental Fig. S4). The chondroitinase enzyme (species 10, which was positively identified by mass spectrometry (59 peptides) (data not shown), but did not give any detectable N-terminal sequence) ran at a position between species 4 and 5 (i.e. the TSG-6⅐HC2 and TSG-6⅐HC1 complexes, respectively).
as the more widely separated bands 4 and 5 (with chondroitinase running between them). N-terminal sequence analysis of these species (Table I) showed that bands 4 and 5 correspond to complexes of TSG-6 with HC2 and HC1, respectively (i.e. these are equivalent to the upper and lower bands of species 2); in bands 4 and 5, there was only a relatively small amount of the alternative HC detected (i.e. 9 and 22%, respectively), presumably due to the slightly better separation of these bands (Supplemental Fig. S4). The reason underlying this improved separation is not clear, but it could result simply from the removal of higher molecular mass species (i.e. intact I␣I and species A) or the presence of the chondroitinase enzyme itself, altering the mobility of species B and C, rather than necessarily being due to removal of chondroitin sulfate from the TSG-6⅐HC complexes.
Model of TSG-6⅐HC Complex Formation-The protein sequence analysis of the TSG-6/I␣I reaction products and characterization of the species formed following treatment with chondroitinase or NaOH described above have allowed us to generate a model for TSG-6⅐I␣I complex formation. As shown in Fig. 5, incubation of TSG-6 and I␣I in the presence of metal ions (Mg 2ϩ or Mn 2ϩ ) leads to the formation of either a TSG-6⅐HC1 or TSG-6⅐HC2 complex (i.e. species C or B, respectively). This is likely to occur via a transesterification reaction in which the ester bond linking one or the other HC to CS is transferred onto TSG-6. Consistent with this, the TSG-6⅐HC complexes are degraded by mild NaOH treatment (to yield free HCs), whereas they are refractory to digestion with chondroitin ABC lyase. Via this mechanism, formation of TSG-6⅐HC1 would leave HC2 still attached to bikunin via the CS chain (i.e. bikunin⅐HC2) as a by-product, whereas if the TSG-6⅐HC2 complex were formed, then bikunin⅐HC1 would be left over. In this regard, species A (band 1 in Fig. 4) is likely to represent a mixture of the bikunin⅐HC1 and bikunin⅐HC2 by-products because sequencing revealed that this species contained a similar amount of bikunin (2.19 pmol) compared with the combined HCs (2.35 pmol) and that it was susceptible to cleavage with chondroitinase. In this model, there would be no cleavage of the CS chain upon transfer of an HC onto TSG-6 and consequently no CS "stub" left attached to either of the TSG-6⅐HC complexes. Overall, this model is similar to that proposed recently by Sanggaard et al. (41).
TSG-6 Mediates the Transfer of HCs onto HA-Recently, it was found that covalent HC⅐HA complexes do not form in the COCs of Tsg-6 null mice (7), indicating that TSG-6 is required for the covalent transfer of HCs from I␣I onto HA. In addition, HC transfer onto HA has been observed in vitro using mouse serum and human recombinant TSG-6 (7, 8). We investigated whether a mixture containing only purified recombinant TSG-6, purified human I␣I, and medical grade HA is sufficient to form HC⅐HA complexes in vitro. As shown in Fig. 6, when 1 g of p-HA was included in the standard assay (i.e. containing TSG-6 and I␣I), an intense additional species (band 11) at the top of the gel was visible; no such band appeared when I␣I and HA were incubated together in the absence of TSG-6 (data not shown). N-terminal sequencing of equivalent reactions (transferred onto Hybond-P membranes) revealed that band 11 contained both HCs in approximately equal amounts (Table I). Furthermore, treatment of such reaction mixtures with Streptomyces hyaluronidase (an enzyme specific for HA) released bands that ran at the same positions as the free HCs (Fig. 6,  right panel). Hyaluronidase treatment of I␣I alone under identical reaction conditions did not lead to its degradation. Therefore, these data indicate that complexes between the HCs and HA can form in vitro in the presence of TSG-6. To investigate this further, a defined oligosaccharide of HA (a 14-mer) was used instead of p-HA, and this led to the appearance of a species of ϳ85 kDa (band 12 in Fig. 6), which N-terminal  Fig. S4 (equivalent to the SDS-polyacrylamide gel in Fig. 4) and bands equivalent to species 11 and 12 on Fig. 6 For each band, multiple sequences are shown in the order of their decreasing initial yields. Where the same residue is found in two sequences at a particular cycle of Edman degradation, it is included in both sequences if this is consistent with the amount of the amino acid detected. Dashes are shown for residues that were not detected due to low yield. sequencing showed to contain mainly HC1, but also a small amount of HC2 (Table I). Mass spectrometric analysis confirmed that both species were present, perhaps with more HC2 than indicated by sequence analysis (36 and 12 peptides of HC1 and HC2 identified, respectively) (data not shown). Given the different mobilities of HC1 and HC2 upon SDS-PAGE, it is surprising that both HC⅐HA 14 complexes should run at identical positions. However, when a longer HA oligomer (i.e. HA 32-mer) was used, this also gave rise to a single novel species of ϳ90 kDa (i.e. a higher apparent molecular mass than that of either HC1 or HC2). Mass spectrometry of this band identified 36 and 15 peptides of HC1 and HC2, respectively (data not shown), whereas amino acid sequencing indicated that there was much more HC1 (93%) than HC2 (7%). Importantly, the finding that the size of the HA oligomer used has a significant effect on the apparent molecular mass of the released HCs is consistent with the formation of stable complexes between the HCs and HA. It therefore seems likely that both HC1 and HC2 can become covalently linked to HA oligosaccharides and that this assay can be used to visualize the TSG-6-mediated transfer of HCs onto HA. In this regard, co-incubation of different w/w ratios of HA 14 and p-HA with a TSG-6/I␣I mixture revealed that this oligosaccharide is likely to be as good a substrate for HC transfer as the ϳ120-kDa HA preparation (Fig. 7), and therefore, HA 14 was used in all subsequent "transfer" assays. . When equal quantities of p-HA and HA 14 were used, both a high molecular mass HC⅐HA complex and an HC⅐HA 14 complex were formed in approximately equal amounts (i.e. based on intensity of Coomassie Blue staining). When p-HA was present in a w/w 10-fold excess over HA 14 , only the HC⅐HA complex was seen, and conversely, when the oligosaccharide was in excess, only the HC⅐HA 14 complex was generated. This indicates that p-HA and HA 14 are likely to be similarly efficient substrates for HC transfer, i.e. the high molecular mass preparation did not lead to preferential transfer of HCs onto HA. The left panel shows an SDS-polyacrylamide gel on which TSG-6 and I␣I were incubated under standard assay conditions, but with the additional inclusion of low molecular mass p-HA or a defined oligosaccharide (HA 14 ). The novel bands (i.e. bands 11 and 12) were identified by Nterminal sequencing and/or mass spectrometry (Table I) as a high molecular mass HC⅐HA complex (containing both HCs) and an HC⅐HA 14 complex, respectively. Species II (ϳ160 kDa), which formed in the presence of HA 14 , was shown by mass spectrometry to contain both HC1 and HC2 (19 and 22 peptides, respectively) and is therefore likely to represent a complex in which both HCs are attached to the same HA oligomer. The right panel shows the results from SDS-PAGE analysis of TSG-6/I␣I/p-HA reaction mixtures or an I␣I control following treatment with (ϩ) and without (Ϫ) Streptomyces hyaluronidase (H'ase). I␣I degraded with NaOH was used to show the positions of HC1 and HC2.

TSG-6-mediated Transfer of HCs onto HA Is Metal Ion-dependent-Preliminary
experiments demonstrated that, when TSG-6, I␣I, and HA 14 were incubated in the presence of EDTA under otherwise standard conditions, neither TSG-6⅐HC (species B/C) nor HC⅐HA 14 was formed (data not shown). Therefore, to test the metal ion dependence of HC transfer onto HA, it was necessary to preform TSG-6⅐HC complexes, which was done in the presence of 0.109 mM MgCl 2 , prior to the subsequent addition of 1 mM EDTA, followed by the addition of HA 14 in the absence and presence of various metal ions (see "Experimental Procedures"). In control experiments in which HA 14 was added to preformed TSG-6⅐HC complexes in the absence of EDTA (i.e. in reactions containing a final concentration of 0.1 mM MgCl 2 ), a band corresponding to HC⅐HA 14 was seen (Fig. 8, upper panel,  lane 1), whereas in the presence of 1 mM EDTA, no HC⅐HA 14 was formed (lane 3). This clearly shows that the formation of HC⅐HA 14 complexes requires the presence of metal ions and is inhibited by EDTA. As shown in Fig. 8 (upper panel), when HA 14 was added in the presence of MgCl 2 , MnCl 2 , or CaCl 2 , the HC⅐HA 14 complex formed. In a separate experiment in which TSG-6, I␣I, and HA 14 were all incubated together under standard conditions with 5 mM MgCl 2 , MnCl 2 , CaCl 2 , or CoCl 2 in the absence of EDTA, the HC⅐HA 14 complex (and the TSG-6⅐HC complex) formed only in the presence of Mg 2ϩ or Mn 2ϩ ions (Fig. 8, lower right panel). This demonstrates that Mg 2ϩ or Mn 2ϩ can support both TSG-6⅐HC complex formation and HC transfer, whereas Ca 2ϩ or Co 2ϩ alone does not give rise to either of these products. Therefore, in the experiment shown in Fig. 8  (upper panel), it is likely that transfer (i.e. formation of HC⅐HA 14 ) occurred in the presence of Ca 2ϩ due to its displacement of Mg 2ϩ ions from the EDTA rather than having a direct effect on the reaction; this is not surprising given the much larger log K 1 for the binding of Ca 2ϩ to EDTA compared with Mg 2ϩ . However, the possibility that Ca 2ϩ , although not supporting the formation of the TSG-6⅐HC complex (Fig. 3), could be involved in subsequent HC transfer cannot be entirely ruled out.
Addition of CoCl 2 to preformed TSG-6⅐HC complexes inhibited the subsequent formation of HC⅐HA 14 (Fig. 8, upper panel). Interestingly, significant inhibition of the transfer reaction by CoCl 2 was seen even when MgCl 2 was present at a 5-fold higher concentration (i.e. 1 mM CoCl 2 and 5 mM MgCl 2 ) (Fig. 8,  lower left panel). Co-incubation of TSG-6, I␣I, and HA 14 with 5 mM MgCl 2 or MnCl 2 in the presence of an equimolar concentration of CoCl 2 did not lead to the formation of HC⅐HA 14 or TSG-6⅐HC complexes (data not shown). These data indicate that Co 2ϩ ions are potent inhibitors of TSG-6-mediated transfer of I␣I HCs onto HA, even in the presence of Mg 2ϩ , presumably due to their tighter binding to the metal ion center involved in this reaction.
The requirement for divalent metal ions during HC transfer was also examined using mouse serum as the source of I␣I and high molecular mass HA by Western blotting with anti-I␣I antibody to visualize the species formed with or without treatment with Streptomyces hyaluronidase (Fig. 9). This showed that, in the absence of EDTA or additional divalent cations (i.e. with just the metal ions present in serum), the only species detected was either a high molecular mass smear corresponding to HC⅐HA complexes of polydispersed molecular masses (Fig. 9A) or a free HC (Fig. 9B) depending on whether the samples were untreated or hyaluronidase-digested, respec- tively. Furthermore, there were no bands corresponding to I␣I or P␣I, indicating that all of these proteins had been converted (in a TSG-6-mediated manner) into an HC⅐HA complex, as shown previously (7). Conversely, the presence of EDTA completely inhibited HC⅐HA complex formation, and intense bands for I␣I and P␣I were present (Fig. 9), indicating that the TSG-6-mediated transfer of HCs onto HA from both I␣I and P␣I is metal ion-dependent. The inclusion of either Ca 2ϩ or Mg 2ϩ led to the formation of HC⅐HA complexes with the concomitant amounts of I␣I and P␣I being greatly diminished. However, the presence of Co 2ϩ ions caused a significant inhibition of I␣I-dependent transfer (a strong band for I␣I was seen), but did not inhibit the consumption of P␣I. Consequently, HC⅐HA complexes were still detected. This demonstrates that the mechanisms underlying TSG-6-mediated formation of HC⅐HA complexes from I␣I and P␣I are distinct. In this regard, experiments in which purified human P␣I was incubated with recombinant human TSG-6 (with and without HA) under a range of conditions in our in vitro assay did not lead to forma-tion of TSG-6⅐HC3 complexes or the transfer of HC3 onto HA (data not shown), indicating that, in this case, TSG-6 alone is not sufficient for the formation of HC3⅐HA and that another factor (e.g. in serum) is also required.
Formation of TSG-6⅐HC Complexes Is Necessary for HC Transfer-As shown in Fig. 8, in the reactions in which HC⅐HA 14 complexes were formed, there appeared to be significantly less B/C doublet (i.e. TSG-6⅐HC complexes) visible compared with assays in which transfer was inhibited, whereas the level of species A (i.e. bikunin⅐HC by-products) was unaffected. In addition, in transfer experiments in which high concentrations of HA (e.g. 10 g of p-HA or HA 14 ) were incubated with TSG-6 and I␣I under otherwise standard conditions, very low amounts of TSG-6⅐HC complexes were visible. Again, the bikunin⅐HC complexes were present in normal amounts (Fig.  7). These data demonstrate that the TSG-6⅐HC complexes are consumed upon the formation of HC⅐HA, i.e. they are likely to act as intermediates in the transfer of HCs onto HA. To investigate this further and to determine the stability of the TSG-6⅐HC complexes, these were formed under standard conditions and then incubated for a range of times (i.e. 0, 2, 4, and 22 h) before the addition of HA 14 (or water) and an additional incubation of 2 h. As shown in Fig. 10, in the absence of HA, there was no reduction in the amount of TSG-6⅐HC complex (B/C doublet) even after the maximal time of incubation (26 h in total); in fact, species B/C was more intense than at the shorter time points. This shows that the TSG-6⅐HC complexes are stable over relatively long time scales. However, a band of ϳ80 kDa (labeled I in Fig. 10) became more intense over time, and this is likely to be the same species as the ϳ80-kDa band seen in reactions mixtures of TSG-6 and I␣I incubated at 37°C (Fig.  1, lower panel). Given the stability of the TSG-6⅐HC complexes, it seems probable that this species results from the breakdown of I␣I and the bikunin⅐HC by-products. Upon sequencing, this band, which was also present in the HA-containing samples (Fig. 10), was found to contain HC1 and HC2 at an ϳ2:1 ratio (data not shown). The molecular mass of this band and the fact that it was formed in both the absence and presence of HA indicate that it is likely to correspond to free HCs, although it is not clear why HC1 and HC2 should migrate together; in both FIG. 9. Metal ion dependence of TSG-6-mediated transfer of HCs onto HA using mouse serum as the source of I␣I. Mouse serum, high molecular mass HA, and human recombinant TSG-6 were incubated at 37°C in the presence or absence of 2 mM EDTA and 5 mM Ca 2ϩ , Mg 2ϩ , or Co 2ϩ . Untreated samples of these reaction mixtures (A) or those digested with Streptomyces hyaluronidase (H'ase) (B) were run on SDS-polyacrylamide gels and analyzed by Western blotting using a polyclonal antibody raised against I␣I. In the latter, sera either left untreated or digested with chondroitin ABC lyase (Ch'ase) were run as controls. In both A and B, the presence of a band for intact I␣I (and the absence of HC⅐HA complexes in A or free HC in B) indicates that I␣I-dependent transfer of HCs had been inhibited, whereas the absence of an I␣I band shows that transfer had occurred. In the presence of TSG-6, P␣I can also lead to the formation of HC⅐HA complexes (7), and the occurrence of a strong band for this protein indicates that this process had been inhibited, whereas the presence of a weaker band indicates that it had taken place.
FIG. 10. SDS-PAGE analysis of HC⅐HA 14 complexes formed from preformed TSG-6⅐HC complexes that had been incubated for various times before the addition of HA. TSG-6⅐HC complexes were preformed under standard conditions (i.e. 2 h at 4°C) and then incubated at 4°C for 0, 2, 4, or 22 h before the addition of HA 14 (ϩ) or water (Ϫ), followed by an additional incubation on ice for 2 h. In reactions containing HA 14 , a band of ϳ85 kDa (corresponding to HC⅐HA 14 ) was formed. Additionally, two species of ϳ160 and ϳ90 kDa were present only in the HA-containing samples (labeled II and III, respectively). A band of ϳ80 kDa (labeled I) was seen in all reactions whether or not they contained HA. cases, the expected N-terminal sequence was present.
When HA 14 was added to preformed TSG-6⅐HC complexes, transfer took place even after they had been incubated at 4°C for a total of 26 h (i.e. the 22-h time point in Fig. 10); the HC⅐HA 14 complex was seen at all incubation times, along with a decrease in the intensity of the B/C doublet compared with reactions containing no oligosaccharide. In this regard, N-terminal sequencing revealed that the HC⅐HA 14 band contained both HC1 and HC2 (94 and 6%, respectively), which is consistent with earlier results (Fig. 6 and Table I).
Additional species of ϳ160 and ϳ90 kDa were visible in the samples containing HA 14 . As noted above (Fig. 6), the ϳ160-kDa band (labeled II) most likely corresponds to an oligomer with both HC1 and HC2 attached. The ϳ90-kDa band (labeled III) was shown by N-terminal sequence analysis to also contain both HCs (with HC1 and HC2 at a 2:1 ratio) (data not shown). This species was formed only in the presence of HA 14 and is likely therefore to represent an HC⅐HA 14 complex, but why this should run at a slightly higher molecular mass is not clear.
On the basis of the above data, it therefore seems likely that TSG-6⅐HC in the presence of HA converts into a HC⅐HA complex. Consistent with this, there was a clear increase in the intensity of the TSG-6 band in samples in which the formation of the HC⅐HA 14 complex had taken place, i.e. TSG-6 was released upon the transfer of HCs onto HA.
TSG-6 Is a Catalyst for HC Transfer-The observation that TSG-6 may be released upon the transfer of HCs from TSG-6⅐HC onto HA led to the hypothesis that such molecules of TSG-6 may be able to form new complexes with I␣I and thus can be recycled. If this were the case, then it would be expected that suboptimal concentrations of TSG-6 should support, over time, the formation of significant amounts of HC⅐HA, given that neither I␣I or HA is limiting. To test this possibility, standard amounts of I␣I (ϳ45 pmol of 1.8 M) and HA 14 (15 M) were incubated with various concentrations of TSG-6 for 2, 4, or 22 h. Experiments conducted at 4°C showed that, when TSG-6 was present at 0.27 or 0.054 M (i.e. 10-and 50-fold lower concentrations than in the standard assay), very little, if any, HC⅐HA 14 complex was formed after 2 or 4 h (data not shown); some transfer occurred with 0.27 M TSG-6 (but not 0.054 M) after 22 h. However, at 37°C, a small amount of HC⅐HA 14 complex was formed after 2 h even in the presence of the lowest concentration of TSG-6 (0.054 M); and after 4 or 22 h, a significant amount of HC⅐HA 14 was apparent (Fig. 11). Fig. 11 shows that, although there was a relatively constant amount of HC⅐HA 14 generated between 2 and 22 h in reactions containing 2.7 M TSG-6, when TSG-6 was present at lower concentrations, the HC⅐HA 14 complex accumulated over time, with the concomitant disappearance of I␣I. Importantly, in the samples containing 0.27 or 0.054 M TSG-6, no TSG-6⅐HC complexes (i.e. B/C doublet) were seen. It is clear, however, that TSG-6⅐HC complexes had been formed because species A, corresponding to the bikunin⅐HC by-products of this reaction (Fig. 5), were present. These data indicate that any TSG-6⅐HC complexes made are converted into HC⅐HA 14 and that the TSG-6 released is indeed recycled. If all of the I␣I (45 pmol at 1.8 M) was consumed, then a maximum of 45 pmol at HC⅐HA 14 could be generated (i.e. one molecule of I␣I is converted into one molecule of either HC1⅐HA 14 or HC2⅐HA 14 ), which would require 45 pmol of TSG-6⅐HC1/TSG-6⅐HC2 complexes to be formed and converted. Although not all of the I␣I was consumed in any of these experiments, Fig. 11 shows that ϳ50% of the I␣I disappeared after 22 h in the presence of 0.054 M TSG-6 (1.3 pmol). This corresponds to the formation of Ͼ20 pmol of HC⅐HA 14 , which is Ͼ15-fold the amount of TSG-6 present. Therefore, the above data show that TSG-6 can be recycled. Thus, it not only acts as an essential cofactor for HC transfer, but is also a true catalyst for this reaction, as illustrated in Fig. 12. DISCUSSION We have shown here that the incubation of human recombinant TSG-6 with purified human I␣I led to the formation of TSG-6⅐HC1 and TSG-6⅐HC2 complexes with the generation of the corresponding bikunin⅐HC2 and bikunin⅐HC1 by-products in a pH-, salt strength-, and metal ion-dependent manner. Our characterization of the TSG-6⅐HC complexes and bikunin⅐HC by-products and the mechanism of complex formation we have proposed (Fig. 5) (43) agree well with a recent study (41). Earlier work from Wisniewski et al. (34) suggested that the TSG-6⅐I␣I complexes formed in vitro consist of TSG-6⅐bikunin⅐HC2 and that this complex is susceptible to degradation by chondroitinase, which does not agree with our data. Furthermore, Sanggaard et al. (41) demonstrated that, in the TSG-6⅐HC2 complex at least, TSG-6 is linked directly to the C-terminal aspartic acid of HC2, providing definitive evidence that TSG-6⅐HC complexes are covalent in nature, as was suspected (31). Given our previous finding that monoclonal antibody A38, which recognizes an epitope in the TSG-6 Link module (44), can inhibit the formation of the TSG-6⅐HC complexes (27), it seems likely that this domain is directly involved in complex formation and may be the site of covalent attachment (as shown in Fig. 5). However, the Link module alone is unable to form a covalent complex with I␣I (45), indicating that other regions of TSG-6 are also necessary for this process (see below).
Sanggaard et al. (41) also observed and characterized high molecular mass species that are likely to correspond to I␣I with one or two additional HCs attached. These were present only in very small amounts in our experiments conducted at 4°C. Therefore, it seems likely that these species result from a side reaction not required for the formation of the TSG-6⅐HC1 and TSG-6⅐HC complexes. Free bikunin was also visualized by a trypsin inhibition counterstaining technique (41), whereas no band corresponding to bikunin was seen on our SDS-polyacrylamide gels, which were stained with Coomassie Blue. However, N-terminal sequence analysis of the ϳ35-kDa "TSG-6" band from gels equivalent to those shown in Fig. 1 (lower panel) (i.e. a time course of TSG-6⅐HC complex formation at 37°C) detected bikunin (in addition to TSG-6) after 30 and 120 min, but no bikunin was detected at 5 min (data not shown). The amounts of bikunin observed at the 5-, 30-, and 120-min time points (0, 0.6, and 0.8 pmol, respectively) correlate reasonably well with the appearance of species I (i.e. free HCs). The absence of detectable bikunin after 5 min, at which point there was already significant TSG-6⅐HC (B/C doublet) present, indicates that this arises from the subsequent breakdown of the bikunin⅐HC complexes (Fig. 12), i.e. bikunin is not released as part of the mechanism of TSG-6⅐HC complex formation. The means by which bikunin⅐HC complexes are degraded is not yet clear; however, a role of TSG-6 in this process cannot be ruled out. The formation of bikunin in this way is likely to be of biological relevance because the inhibition of certain serine proteases occurs only when this inhibitor is in its free state (e.g. tissue kallikrein in the context of asthma). 2 We have demonstrated here that TSG-6⅐HC complexes act as intermediates in the formation of HC⅐HA. It is likely that this also involves a transesterification reaction in which the ester bond linking a HC to TSG-6 is transferred onto HA; it has been shown previously that HCs become attached to HA via an ester bond between the carboxylates of their C-terminal aspartic acid residues and the C-6 hydroxyls of GlcNAc residues (3). In other words, the TSG-6-mediated formation of HC⅐HA complexes requires two sequential transesterification reactions involving the "same" ester bond at the C terminus of the HC. Given that, in our assays, the only protein components present were TSG-6 and I␣I, it is probable that this transesterase activity is located in one or the other (or both) of these molecules. It is very unlikely that this activity could be due to a contaminating enzyme because we have used many different preparations of these proteins and have seen no variability in the formation of either TSG-6⅐HC or HC⅐HA. In this regard, our results indicate that TSG-6⅐HC complex formation and HC transfer onto HA are metal iondependent processes with a requirement for either Mg 2ϩ or Mn 2ϩ and are inhibited by cobalt. The similarity of their metal ion dependences (and inhibition) suggests that the same metal ion center may be involved in both reactions. Divalent metal ion-binding sites are likely to be present in the HCs because they each have a single von Willebrand factor A domain (Figs. 5 and 12) containing the metal ion-dependent adhesion site consensus sequence known to support the binding of Mg 2ϩ /Mn 2ϩ in other proteins (46). Interestingly, TSG-6 is also likely to bind metal ions. Recent structural studies on complement C1s have revealed that its CUB1 module contains a Ca 2ϩ /Mg 2ϩ -binding site, which would also be expected to accommodate Mn 2ϩ , where the residues involved in metal ion coordination are also conserved in the CUB module of TSG-6 (47). Further work will be necessary to determine whether these metal ion-binding sites in I␣I and TSG-6 are involved in the transesterification processes. It should be noted that, although our data clearly show that Ca 2ϩ ions do not support the formation of TSG-6⅐HC complexes and are not a requirement for HC transfer, we cannot completely rule out their involvement in the second transesterification step.
It has been reported previously that Ca 2ϩ ions are in fact essential for the coupling of I␣I to HA and that Mg 2ϩ ions are not involved (48), which is at odds with our findings. The reason for this is not clear, but may relate to the fact that the previous study was conducted in synovial fluid containing EDTA, where it is likely that the addition of Ca 2ϩ ions may release sufficient Mg 2ϩ from the EDTA or the various proteins present to allow the reaction to proceed. We also found that the formation of TSG-6⅐HC complexes in serum/EDTA could be rescued by the addition of CaCl 2 (Supplemental Fig. S3), even though it is clear that Ca 2ϩ does not support this reaction in assays containing just the purified proteins and no additional metal ions (Fig. 3).
A major issue relating to our proposed model of HC⅐HA formation is the question of where the energy that drives these reactions comes from. Although the generation of both TSG-6⅐HC and HC⅐HA complexes occurred rapidly even at 4°C (e.g. HC⅐HA was visible after only 30 s) (data not shown), these FIG. 12. TSG-6 is released upon HC transfer onto HA and can be recycled to generate more TSG-6⅐HC complexes. Shown is a schematic model of TSG-6⅐HC complex formation and the subsequent transfer of the HC onto HA. A generic TSG-6⅐HC complex is illustrated for simplicity. TSG-6 released during transfer can interact with a new I␣I molecule, thus acting as both a catalyst and a cofactor in the formation of HC⅐HA. In the transfer product, the HC is shown linked via its C terminus (an aspartic acid) to a GlcNAc of HA, as has been shown previously (3). Both the processes of TSG-6⅐HC complex formation and HC transfer are Mg 2ϩ /Mn 2ϩ -dependent, and a divalent metal ion binding-site is present both within the I␣I von Willebrand factor A domains (denoted by red asterisks in the active molecules) and within the TSG-6 CUB module (denoted by yellow asterisks). The bikunin⅐HC by-products of TSG-6⅐HC complex formation (illustrated for bikunin⅐HC1) break down over time (dotted arrow), generating free bikunin and free HCs; the mechanism of breakdown and the fate of the CS chain are currently not known. The above model and that in Fig. 5  processes are temperature-dependent because they did not occur efficiently at 4°C when the amount of TSG-6 was limiting, unlike the situation at 37°C (Fig. 11). However, although heat energy is required, the major driving force underlying the formation of a new ester bond must be derived from the fission of the existing covalent linkage. In other words, the energy of these processes is stored in the CS/HC linkages of the I␣I molecule (15,16), which are formed intracellularly during its biosynthesis (49). Consistent with this, we have found that incubating individual HC1 or HC2 (alone or together) with TSG-6 did not lead to either TSG-6⅐HC or HC⅐HA complex formation in the absence or presence of HA (data not shown).
The composition of TSG-6⅐I␣I complexes (i.e. TSG-6⅐HC1 and TSG-6⅐HC2) made from human proteins in vitro and their sensitivity to NaOH, but not chondroitinase, indicate that they are likely to be the same as those characterized previously from murine COCs in vivo (31). Therefore, the same processes of cumulus matrix expansion may occur in both mice and humans, i.e. the cross-linking of the HA-rich COC matrix via the generation of HC⅐HA (7,8,27,43). In this regard, the human recombinant TSG-6 used in this study can rescue the expansion of COCs from Tsg-6 Ϫ/Ϫ mice (7). Furthermore, recent work has indicated that HC⅐HA is formed during cumulus expansion in the pig, where this may also be mediated by TSG-6 (50), making it possible that this is a mechanism common to eutherian (placental) mammals.
Given the rapidity of TSG-6⅐HC and HC⅐HA complex formation in our assay system, it seems likely that these complexes would form in vivo wherever TSG-6, I␣I, and HA meet; TSG-6⅐HC can be formed under a wide range of pH (6.0 -8.0) and salt strength (50 -300 mM) conditions (Fig. 2), including acid pH values that are often associated with inflammation. For example, TSG-6⅐I␣I complexes of ϳ120 kDa, which are likely to correspond to TSG-6⅐HC, and HC⅐HA have been detected in a variety of inflammatory processes (e.g. rheumatoid arthritis and osteoarthritis (4,10,11,13,36) and asthma) 2 and inflammation-like settings (e.g. ovulation (reviewed in Ref. 43) and cervical ripening (39)). In this regard, TSG-6 expression (24) and HA synthesis (51) are up-regulated in response to inflammatory mediators, and I␣I can ingress into tissue spaces from serum due to either vasodilation (10,13) or regulated changes to specific blood-tissue barriers, such as in the ovarian follicle (52,53). The stability of the TSG-6⅐HC complexes would further increase the likelihood of HC⅐HA being produced once an HA molecule is encountered. Consistent with this notion, it has been reported that the synthesis of HA is the rate-limiting step in the formation of the cumulus matrix (54). In other cases where there is a high concentration of HA (e.g. synovial fluid), it would be expected that any TSG-6⅐HC formed would be converted quickly into HC⅐HA, and thus, there would not be an accumulation of the TSG-6⅐HC complex. For example, in synovial fluids from arthritis patients, there is only a small amount of ϳ120-kDa complex detected relative to the amount of free TSG-6 (36). Importantly, the finding that TSG-6 acts as a true catalyst in the transfer of HCs onto HA is likely to mean that low levels of TSG-6 expression could lead to a significant production of HC⅐HA over time, which may be of relevance to diseases that take many years to develop (e.g. atherosclerosis). Further research is clearly necessary to understand the biological functions of HC⅐HA, to identify the locations of TSG-6⅐HC/ HC⅐HA in physiological and pathological processes, and to determine the precise molecular details of their formation.