Transfer of Inter-α-inhibitor Heavy Chains to Hyaluronan by Surface-linked Hyaluronan-TSG-6 Complexes*

Inter-α-inhibitor, TSG-6, and hyaluronan have important functions in fertility and inflammation. Two subunits of inter-α-inhibitor, the heavy chains, form covalent bonds with TSG-6 or hyaluronan in vitro. TSG-6-heavy chain complexes serve as intermediates in the transfer of heavy chains from inter-α-inhibitor to hyaluronan. In vivo, in addition to these complexes, stable ternary complexes of hyaluronan with both TSG-6 and heavy chains have been demonstrated in the ovulatory cumulus oophorus. In our ongoing efforts to characterize the multiple interactions between hyaluronan, TSG-6 and inter-α-inhibitor, we recently characterized the formation of highly stable complexes of TSG-6 with hyaluronan that had been tethered to a solid surface. Here we show that these hyaluronan-TSG-6 complexes are functionally active and transfer heavy chain subunits from inter-α-inhibitor to either free or surface-bound hyaluronan. Transitional hyaluronan-TSG-6-heavy chain complexes do not accumulate in vitro. Our data show the capability for heavy chain transfer by both free TSG-6 and preformed hyaluronan-TSG-6 complexes, suggesting that both might contribute to hyaluronan modification in vivo. Transfer of heavy chains to surface-tethered hyaluronan by either free TSG-6 or surface-tethered hyaluronan-TSG-6 complexes did not affect the CD 44-mediated binding of BW 5147 cells in vitro. We show how TSG-6 and hyaluronan together can deplete inter-α-inhibitor and generate bikunin, as has been observed in sepsis, and discuss the role of TSG-6 in the generation of hyaluronan-heavy chain complexes associated with ovulation, arthritis, and sepsis.

Tumor necrosis factor-stimulated gene 6 (TSG-6) 2 encodes a glycoprotein of ϳ35 kDa that is commonly referred to as TSG-6 protein (1-3). Expression and function of TSG-6 have been associated with inflammation and fertility. TSG-6 protein consists of two domains, the N-terminal link module and the C-terminal CUB domain. The N-terminal domain of TSG-6, a so-called link module (4), identifies TSG-6 as a hyaluronan (HA)-binding protein.
In addition to HA, TSG-6 also interacts with the plasma protein inter-␣-inhibitor (I␣I) in a HA-independent fashion (5). I␣I is a protein-polysaccharide complex of unique structure consisting of three polypeptide chains linked by a glycosaminoglycan bridge (6,7). The smallest of the three polypeptides, the serine protease inhibitor bikunin, carries a single chondroitin 4-sulfate chain attached via a classical proteoglycan linkage group, whereas the two closely homologous heavy chains (HC) 1 and 2 are linked to hydroxyl groups of the chondroitin sulfate via alkali-sensitive ester bonds formed by their C-terminal aspartic acid residues (6,8,9). TSG-6 interacts with I␣I, resulting in the transfer of either one of the HCs from I␣I to TSG-6, resulting in a covalent TSG-6⅐HC complex (5, 10 -14). Although both HC1 and HC2 can be transferred, the presence of HC2 is essential for any HC transfer (14).
In the presence of HA the HC is further transferred from TSG-6⅐HC to HA, resulting in the formation of covalent HA⅐HC complexes and the release of free TSG-6 (11). HA⅐HC complexes in synovial fluid of patients with rheumatoid arthritis have been known for decades, although they were assumed to be complexes of HA and complete I␣I (15). These complexes are also known as serum-derived HA-associated protein (SHAP) (16). As in I␣I, the HCs are coupled to HA by an ester bond formed by its C-terminal aspartic acid residue (17).
The currently accepted pathway of HC transfer from I␣I to TSG-6 and subsequently to HA may not be the only pathway to occur in vivo. We have been interested in an alternate pathway that utilizes preformed HA⅐TSG-6 complexes. We have previously studied the interactions of TSG-6 with HA that had been covalently tethered to a surface by chemical cross-linking (18). HA tethered to a surface can be considered a model for HA associated with cell surfaces before release from hyaluronan synthase or by binding to cellular receptors. We found that TSG-6, in a strictly temperature-dependent fashion, formed complexes with surface-tethered HA that were resistant to stringent dissociating and reducing agents (18). However, nei-ther we 3 nor others (10,11) found evidence that such highly stable HA⅐TSG-6 complexes are formed with HA free in solution.
In vivo, stable complexes consisting of HA, TSG-6, and one HC have been demonstrated in the HA-rich extracellular matrix of ovulatory cumulus cell-oocyte complexes (COCs) (19). These complexes formed in vivo are clearly different from all HA⅐HC complexes that have been generated in vitro in that they incorporate TSG-6 in a stable complex with HA. In contrast, when HA in solution was used as a substrate for HC transfer in vitro, free TSG-6 was released and recycled (11). In the current study we provide evidence that stable complexes formed between TSG-6 and tethered HA are useful models for the complexes observed in vivo and that these complexes themselves can function to catalyze the subsequent transfer of HCs to HA.
The existence of stable complexes consisting of HA and both TSG-6 and HCs in ovulatory COCs points to a physiological role, most likely in the stabilization of the HA-rich extracellular matrix of COCs (19). The significance of the complex interactions between TSG-6, HA, and I␣I is most clearly demonstrated by the infertility of female mice deficient in either TSG-6 (20) or I␣I (21,22), resulting from the instability of ovulatory COCs. Stable HA⅐TSG-6 complexes may also form in other hyaluronan-rich tissues, e.g. cartilage, where accumulation of TSG-6 has been shown in both arthritic humans and in mice (23,24). They may form a local reservoir of functional TSG-6, as has been suggested by the persistence of the anti-inflammatory effects of recombinant TSG-6 in experimental models of arthritis (25,26) and even in the absence of detectable TSG-6 in plasma (26).
The functional activity of stable HA⅐TSG-6 complexes formed in vitro (18) has not yet been investigated but is the main subject of this study. We characterized simultaneous TSG-6 binding and HC transfer to surface-tethered HA and transfer of HCs by preformed HA⅐TSG-6 complexes. In vitro, intermediates of this transfer did not accumulate, resulting in HA⅐HC, but not HA⅐TSG-6⅐HC complexes, as the final products of this reaction. We suggest that HA⅐TSG-6 complexes like the ones generated here in vitro may be related to stable ternary complexes of HA, TSG-6, and HCs that have been described earlier in vivo (19).

EXPERIMENTAL PROCEDURES
Reagents-Covalink NH(Cov-NH) plates were purchased from Nunc, sulfo-NHS was from Pierce, and EDC was purchased from Sigma. HA from rooster comb was purchased from Sigma, and bacterial HA was obtained from Lifecore Biomedical; both were used for coupling to Cov-NH. Equivalent results were obtained with Cov-HA prepared with rooster comb HA and bacterial HA. EAH-Sepharose was purchased from GE Healthcare. Hyaluronidase from Streptomyces hyalurolyticus and biotinylated hyaluronan-binding protein (HABP-bio) were purchased from Associates of Cape Cod. SelectHA TM 30 kDa was purchased from Hyalose L.L.C. Human plasma was obtained from the blood bank of the NYU Medical Center. Rabbit anti-human I␣I was from Dako. Although this antibody was raised against I␣I, it did not detect bikunin in immunoblotting and enzyme-linked immunosorbent assay and is, therefore, referred to as HC-specific. The rabbit anti-TSG-6 antibody was raised against native recombinant TSG-6 and has been described earlier (27). The rabbit anti-bikunin antibody has been described and characterized earlier (9). Biotinylated goat anti-rabbit Ig was from Dako, and streptavidin-alkaline phosphatase conjugate was from Invitrogen. The alkaline phosphatase substrate was p-nitrophenyl phosphate (Sigma) for solution assays and 5-bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium from Bio-Rad for immunoblots. Polyvinylidene difluoride membranes for immunoblotting were from Millipore (Immobilon P).
Purified I␣I-I␣I was purified from human plasma as described previously (29).
Coupling of HA to Covalink-NH Plates and EAH-Sepharose-HA was coupled to Covalink-NH plates as described (18,30). In brief, HA solution containing sulfo-NHS was added to the wells of Cov-NH plates. After the addition of EDC, the plates were incubated for 2 h at room temperature and overnight at 4°C. Thereafter, the plates were washed extensively with 2 M NaCl and blocked with 0.2% casein in TTBS (20 mM Tris, pH 7.5, 500 mM NaCl, 0.1% Tween 20). Coupling of HA to EAH-Sepharose was carried out similarly to the coupling of HA to Cov-NH. HA at 0.55 mg/ml in a 22.2% (v/v) suspension of EAH-Sepharose was incubated with 0.89 mM sulfo-NHS and 0.36 mM EDC in H 2 O for 2 h at room temperature and overnight at 4°C under constant movement. Thereafter, the HA-Sepharose was washed with an at least 10-fold volume of 2 M NaCl and blocked with 0.2% casein in TTBS.
TSG-6 Binding Assay and HC Transfer Assay Using HA Immobilized on Cov-HA-The assay for TSG-6 binding to Cov-HA was carried out as described (18). To determine HC transfer by free TSG-6, TSG-6 at the indicated concentration was mixed with human plasma at the indicated dilution, usually 1:1000, in TTBS and co-incubated in wells of Cov-HA plates for 2 h at 37°C (100 l/well). After washing 3 times with 200 l of TTBS, the wells were incubated with a 1:1000 dilution (in TTBS) of a rabbit anti-I␣I antiserum that is HC-specific, i.e. it does not detect bikunin, for 1 h at 37°C followed by incubation with biotinylated goat anti-rabbit Ig (1:1000 in TTBS, 1 h at 37°C). After incubation with a streptavidin-alkaline phosphatase conjugate (1:1000 in TTBS, 1 h at 37°C) and with p-nitrophenyl phosphate (2 mg/ml in 50 mM Tris, pH 9.5, 2 mM MgCl 2 , ϳ5-30 min at 37°C), the amount of dephosphorylated substrate was determined by measuring the absorbance at 410 nm using 800 nm as a reference wavelength. Each set of experimental conditions included appropriate positive and negative controls, but absolute absorbance values were not standardized across different experiments.
HC Transfer to HA-Sepharose-100 l of HA-Sepharose (sediment) was incubated with 200 l of 400 nM TSG-6 in TTBS for 2 h at 37°C. After extensive washing with TTBS (3 times 400 l) the HA-Sepharose was incubated with a 1:100 dilution of human plasma in PBS (2 h at 37°C). After extensive washing as above, the HA-Sepharose was treated with 100 l of 10 units/ml Streptomyces hyaluronidase (in PBS, 2 h at 50°C). 100 l of 2ϫ SDS-PAGE sample buffer was added to the digested HA-Sepharose, and after vortexing, the liquid phase was recovered by centrifugation in a Spin-X column (Corning Inc., Corning, NY). The recovered filtrate was analyzed by immunoblotting for the presence of TSG-6, HCs, TSG-6⅐HC complexes etc. using HCand TSG-6-specific antibodies. Controls were treated either only with TSG-6 or only with plasma or without TSG-6 and plasma. To discern between HC transfer by free TSG-6 versus HA⅐TSG-6 complexes, the incubation with plasma was also carried out in TTBS instead of PBS (see "Results" for the differential effects of these buffers).
HC Transfer by Solid-phase HA⅐TSG-6 Complexes to HA in Solution-TSG-6 was bound to HA-Sepharose as described above. 100 l of HA-Sepharose (sediment) was incubated with 200 l of 100 nM TSG-6 in TTBS for 2 h at 37°C. After extensive washing with TTBS (3 times 400 l) the HA-Sepharose was incubated with 100 l of a 1:600 dilution of human plasma in PBS containing 100 g/ml SelectHA TM of 30 kDa (2 h at 37°C). 100 l of 2ϫ SDS-PAGE sample buffer was added to the HA-Sepharose and, after vortexing, the liquid phase was recovered by centrifugation in a Spin-X tube (Costar). The recovered filtrate was analyzed by immunoblotting for the presence of HA⅐HC complexes using HC-specific antibodies and a HA-specific HABP probe. Controls were treated either only with TSG-6 or only with plasma or without TSG-6 and plasma. To differentiate between HC transfer by free TSG-6 versus stable HA⅐TSG-6 complexes, the incubation with plasma was also carried out in TTBS instead of PBS (see "Results" for the differential effects of these buffers).
Formation of HA⅐HC Complexes in Solution-20 nM TSG-6, 100 g/ml SelectHA TM , and human plasma diluted 1:600 were co-incubated for 2 h at 37°C in PBS, and the reaction products were analyzed by immunoblotting for the presence of TSG-6, I␣I, HCs, TSG-6⅐HC complexes, HA⅐HC complexes, and free bikunin using HC-, TSG-6-, or bikunin-specific antibodies or the HA-specific HABP probe. Controls contained either one or two of the three components (TSG-6, plasma, HA).
Binding of BW 5147 Cells to Immobilized HA-This assay was carried out as previously described (18) with minor modifications. BW 5147 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% of fetal bovine serum. To evaluate their binding to immobilized HA, 3 ϫ 10 5 BW 5147 cells were plated into wells of Cov-HA or Cov-NH plates after treatment of the plates with TSG-6 and/or plasma to facilitate either TSG-6 binding or HC transfer. After incubation for 2 h at 37°C, the medium of the wells was aspirated, and the wells were washed 3 times with 200 l of PBS. Adherent cells were fixed for 15 min in 10% formalin (in 0.1 M sodium acetate buffer, pH 3.5) and stained for 30 min in 0.05% Naphtol Blue Black in the same buffer. Thereafter, the wells were washed 3 times with distilled water, and the dye was eluted from the stained cells with 150 l of 50 mM NaOH. Finally, the concentration of the eluted dye was determined by measuring the absorbance at 615 nm. The wave length of 800 nm was used as a reference wave length. The concentration of the eluted dye provides a measure of the number of cells adhering to immobilized HA.

TSG-6 Mediates Transfer of HCs from I␣I to Immobilized HA-
TSG-6-mediated transfer of HCs from I␣I to HA in solution has been studied previously (10,31). We reported earlier that the binding of TSG-6 to HA covalently tethered to a surface resulted in the formation of stable HA⅐TSG-6 complexes (18), whereas there was no evidence for the formation of such complexes with HA in solution (11). Because of these differences between the binding of TSG-6 to HA free in solution and to HA tethered to a surface, we first characterized TSG-6-mediated HC transfer from I␣I to HA that was covalently tethered to Covalink. We used purified I␣I to test whether recombinant TSG-6 is necessary and sufficient for the transfer of HCs from I␣I to Cov-HA. To minimize nonspecific binding, the reaction was carried out in TTBS buffer (18). HCs transferred to Cov-HA were detected using a rabbit anti-HC antibody that reacts with I␣I and HCs but did not detect bikunin. As shown in Fig. 1A, there was virtually no HC transfer from purified human I␣I or binding of purified I␣I to immobilized HA in the absence of TSG-6. HC binding to HA was readily detectable if purified TSG-6 was added along with purified I␣I. This assay does not distinguish between bound HCs or complete I␣I, and this issue will be addressed further below.
To investigate whether the lack of HC transfer from purified I␣I to Cov-HA is an artifact resulting from partial denaturation of this complex protein during its purification, we examined the transfer of HCs from I␣I in plasma to immobilized HA. Fig. 1B shows that the transfer of HCs from I␣I in 1:1000-diluted human plasma to Cov-HA is also completely dependent on the presence of added TSG-6 protein. As little as 16 pM TSG-6 is sufficient to mediate detectable transfer of HCs from I␣I in human plasma to Cov-HA (Fig. 1B). The average concentration of I␣I in human plasma is about 700 mg/liter (32). We tested how the plasma concentration affects the transfer of HCs to HA (Fig. 1C). With respect to HC transfer to Cov-HA in the presence of TSG-6, the I␣I concentration in human plasma is saturating, and a 1:100 dilution of human plasma did not result in any decrease of HC transfer. Even at a 1:4000 dilution of plasma the transfer of HCs to HA was only decreased by about 50%, and HC transfer was still detectable at a plasma dilution of 1:16000. This means that concentrations of unpurified I␣I of as little as 200 pM are sufficient for detectable HC transfer to HA. Interestingly, some HC transfer to Cov-HA could be detected in the absence of exogenous TSG-6 at high plasma concentrations (Fig. 1D). However, a plasma dilution of only 1:100 completely abolished any detectable transfer.
Stability of the Bond between HCs and Immobilized HA-We reported earlier that TSG-6 forms a remarkably stable bond to immobilized HA in a strictly temperature-dependent fashion. Using the same assay system, we found that I␣I HCs transferred to immobilized HA in the presence of 50 nM TSG-6 are also almost completely resistant to treatment with 6 M guanidine HCl, 6 M guanidine HCl containing 8% lauryl sulfobetain, or SDS-PAGE sample buffer containing 2-mercaptoethanol ( Fig.  2A). However, they were sensitive to treatment with hyaluronidase from S. hyalurolyticus and, in contrast to HA⅐TSG-6 complexes, they were also sensitive to treatment with 50 mM NaOH, in agreement with the presence of an ester bond between HA and HC in HA⅐HC complexes (17) (Fig. 2B). This provides evidence that the HC transfer we observed on Cov-HA yields HA⅐HC complexes similar to those observed with HA in solution.
Simultaneous Versus Sequential TSG-6 Binding and HC Transfer to Cov-HA-Because TSG-6 itself binds to Cov-HA under the conditions used to evaluate HC transfer to Cov-HA, as shown previously (18), we investigated if both TSG-6 binding and HC transfer to Cov-HA can occur simultaneously. Earlier experiments demonstrating TSG-6-dependent HC transfer to HA in solution indicated that TSG-6 did not form complexes with HA; instead, it was recycled (11). In contrast, we show here that after co-incubation of TSG-6 and diluted plasma in Cov-HA, both TSG-6 and HCs could be detected attached to the surface-bound HA (Fig. 3A).
To exclude the possibility that the formation of stable HA⅐TSG-6 or HA⅐HC complexes is an artifact caused by the unlikely presence of residual active intermediates of the chemical coupling reaction used to tether HA, we showed that immunoglobulin, a protein that can be coupled efficiently by the EDC/ sulfo-NHS coupling procedure, did not bind or couple to Cov-HA under the conditions of our binding experiments (supplemental Fig. S1) and that stable bond formation seems to be unique for TSG-6 among the proteins sharing homologues of the link module; HABP, consisting of isolated link modules of both aggrecan and link protein, did bind to Cov-HA but did not form a stable bond (supplemental Fig. S2). In addition, the formation of the stable bond between TSG-6 and surfacebound HA is completely prevented at 4°C (18), whereas the EDC/sulfo-NHS-mediated coupling works well at this temperature.
We also investigated if preformed HA⅐TSG-6 complexes are able to mediate the subsequent transfer of HCs from I␣I to Cov-HA in a twostep reaction. Fig. 3B shows that TSG-6, after binding to Cov-HA under conditions that result in the formation of stable HA⅐TSG-6 complexes, is able to mediate the subsequent transfer of HCs to the Cov-HA substrate after all TSG-6 in solution has been removed by extensive washing. Because this binding assay cannot distinguish between HCs bound directly to HA, i.e. HA⅐HC complexes, and HCs bound to TSG-6, i.e. HA⅐TSG-6⅐HC complexes, we carried out additional experiments to address this question, described below.
Transfer of I␣I HCs To Immobilized HA Is Strictly Dependent on Metal Ions-It was established earlier that the transfer of HCs to HA is dependent on the presence of divalent metal ions and could be prevented by the addition of EDTA (11,33,34). Fig. 4A confirms that the TSG-6-dependent transfer of HCs to Cov-HA is also completely prevented by the presence of 10 mM EDTA, a treatment that does not prevent the binding of TSG-6 to immobilized HA (18). We also determined the role of metal ions on the individual steps of the 2-step reaction described above, i.e. the transfer of HCs to Cov-HA by preformed Cov-HA⅐TSG-6 complexes. Although the presence of EDTA during the binding of TSG-6 to Cov-HA has a modest effect on the transfer of HCs in the following step, the presence of EDTA during the second step, the interaction of preformed HA⅐TSG-6 complexes with I␣I, completely prevented any HC transfer (Fig. 4B). The partial inhibitory effect of EDTA during the binding of TSG-6 to Cov-HA on the subsequent transfer of HCs is in good agreement with a similar effect of EDTA on the binding of TSG-6 to Cov-HA itself (18).  (Figs. 1, A-D,  and 5A), the sequential transfer of HCs by preformed HA⅐TSG-6 complexes required PBS and was completely sup-pressed in TTBS (Fig. 5A). To determine whether the failure of sequential HC transfer was the result of the difference in ionic strength or the result of the presence of the nonionic detergent, Tween 20, in TTBS, or the different ions in the two buffer systems, we carried out the second step of the sequential HC transfer, i.e. the transfer of HCs by preformed HA⅐TSG-6 complexes in either PBS, Tris buffer pH 7.5 containing 150 mM NaCl, Tris buffer pH 7.5 containing 500 mM NaCl, Tris buffer pH 7.5 containing 150 mM NaCl and 0.1% Tween 20, or Tris buffer pH 7.5 containing 500 mM NaCl and 0.1% Tween 20 (TTBS). As can be seen in Fig. 5B, the second step of the sequential HC transfer is completely inhibited in the presence of 500 mM NaCl but not by the presence of Tween 20 or by the different buffers used. This differential ionic strength requirement for the simultaneous and the two-step, or sequential transfer of HCs can, therefore, be used to differentiate between HC transfer by free TSG-6 and HC transfer by preformed HA⅐TSG-6 complexes. Although the two-step transfer of HCs by preformed HA⅐TSG-6 complexes is completely inhibited in 500 mM NaCl, HC transfer by TSG-6 in solution can still take place.
HC Transfer by Preformed HA⅐ TSG-6 Complexes Bound to Sepharose Results in HA⅐HC Complexes-Preformed HA⅐TSG-6 complexes interact with I␣I, resulting in the transfer of HCs to the HA substrate. Presumably, the first step of this transfer is the formation of a HA⅐TSG-6⅐HC complex, in analogy to the formation of TSG-6⅐HC complexes with free TSG-6. These HA⅐TSG-6⅐HC complexes may transfer HC in an additional step to HA, resulting in HA⅐HC complexes as final products, in analogy to HA⅐HC complexes formed with HA free in solution. Alternatively, HA⅐TSG-6⅐HC complexes could be the final product. These alternatives raise an important question because they would result in structurally different modifications of surface-bound HA versus HA free in solution.
To address this question, we employed TSG-6 bound to HA-Sepharose under conditions resulting in the formation of stable HA⅐TSG-6 complexes (18). After stringent washing in TTBS to remove all free TSG-6 protein, the HA⅐TSG-6 complexes were incubated with diluted plasma in PBS, as a source of I␣I, washed again with TTBS, and then treated with hyaluronidase from S. hyalurolyticus to release HA fragments and attached proteins from the Sepharose. The recovered products were then analyzed by immunoblotting for the presence of either free HCs or TSG-6⅐HC complexes. To exclude the possibility that the HC transfer is due to the presence of residual free TSG-6 instead of HA⅐TSG-6 complexes, the incubation with diluted plasma was carried out either in TTBS or in PBS, as described above. Fig. 6 demonstrates that hyaluronidase treatment releases TSG-6 (left panel) and HCs (right panel). Only trace amounts of TSG-6⅐HC complexes could be detected and only under non-permissive conditions, i.e. in TTBS. If the reaction between preformed HA⅐TSG-6 and plasma was carried out in TTBS, only trace amounts of HCs were transferred to the HA-Sepharose (right panel), and the hyaluronidase treatment releases only the bound TSG-6 (left panel). This demonstrates that there was no significant HC transfer to HA by free TSG-6, i.e. TSG-6 released from the HA during the incubation in TTBS. Therefore, virtually all observed HC transfer to HA in PBS was actually mediated by preformed HA⅐TSG-6 complexes that are unable to mediate the transfer in the presence of TTBS. The diffuse nature of the HC bands in Fig. 6 (right panel) may be a consequence of residual HA fragments remaining attached to the HCs after hyaluronidase digestion. The almost complete absence of TSG-6⅐HC complexes after hyaluronidase digestion strongly suggests that HA⅐TSG-6⅐HC complexes are not a final product of this reaction, but they can still be postulated to occur as a short-lived intermediate. The data also allow further characterization of the TTBS-sensitive step in the reaction sequence. If HA⅐TSG-6 complexes were able to accept HCs from I␣I in TTBS but were not able to transfer them to HA (in agreement with the lack of HA⅐HC complex formation in TTBS), HA⅐TSG-6⅐HC complexes would accumulate on the Sepharose-HA in TTBS. Therefore, the almost complete absence of TSG-6⅐HC complexes in Fig. 6 strongly suggests that in TTBS HA⅐TSG-6 complexes are unable to accept HCs from I␣I.
HC Transfer by Surface-bound HA⅐TSG-6 Complexes to HA Chains in Solution-To determine whether preformed HA⅐TSG-6 complexes tethered to the surface of Sepharose are able to transfer HCs from I␣I to HA molecules in solution, i.e. to a different HA chain than the one binding TSG-6, we first incubated TSG-6 with HA-Sepharose as described above. In the second step, we incubated the HA⅐TSG-6 complexes with diluted plasma plus HA in solution using an HA preparation with a narrow molecular mass distribution (SelectHA TM , 30 kDa). After the incubation, we analyzed the liquid phase of the reaction mixture by immunoblotting for the presence of HA⅐HC complexes. To exclude the possibility that the HC transfer is due to the presence of contaminating free TSG-6 instead of HA⅐TSG-6 complexes, the incubation with diluted plasma and SelectHA TM was carried out either in TTBS or in PBS, as described above (Fig. 5, A and B). Fig. 7 shows that HA⅐HC complexes with SelectHA TM are indeed formed in PBS and only trace amounts in TTBS, demonstrating that the Sepharose-HA⅐TSG-6 complex is able to transfer HCs to HA molecules free in solution. The formation of trace amounts of HA⅐HC complexes during incubation in TTBS may be due to small amounts of residual TSG-6 in solution. The significance of these data is that they indicate that TSG-6 bound to one HA chain is able to transfer HCs to a different HA chain.

TSG-6-mediated HC Transfer can Deplete I␣I and Generate Free
Bikunin-When TSG-6 and plasma, as a source of I␣I, are co-incubated in the absence of HA, TSG-6⅐HC complexes are the major species detected by immunoblotting (lane 6 in Fig. 8, A and C). However, when SelectHA TM is included in the reaction mixture, HA⅐HC complexes are the major species detected (lane 7 in Fig. 8, A and B). HA⅐HC complexes are detected both by an anti-HC antibody (Fig. 8A, lane 7) and by the HA-detecting HABP-bio probe (Fig. 8B, lane 7). The lack of detection of free HA by this probe is due to the failure of free HA to blot to the PVDF membrane; it can be detected by an HABP-bio affinity probe after transfer to positively charged blotting membranes (data not shown).
Interestingly, whereas we observed earlier that the interaction of TSG-6 with immobilized HA resulted in the formation of stable HA⅐TSG-6 complexes (18), there is no indication for the generation of stable HA⅐TSG-6 complexes resistant to dissociating and reducing agents with HA when free in solution (Fig. 8C, lanes 4 and 7). However, because the transfer of HCs to HA in solution was strictly TSG-6-dependent, the interaction of TSG-6⅐HC complexes with HA has to be considered a precondition for the transfer of HCs to HA. Notably, the presence of both TSG-6 and SelectHA TM in the reaction mixture resulted in significant depletion of the available I␣I (Fig. 8A, lane 7) and the generation of free bikunin (Fig. 8D,  lane 7). When only TSG-6 was added to plasma, the I␣I depletion was far less pronounced (Fig. 8A, lane 6 versus lane 7), suggesting that the presence of HA as acceptor for HC transfer contributes to I␣I depletion. The same is less evident for the generation of free bikunin (Fig. 8D, lane 6 versus lane 7), although the generation of free bikunin is the direct result of the depletion of I␣I.

Effect of TSG-6 Binding and HC Transfer on the CD44-mediated
Binding of BW 5147 Cells to HA-BW 5147 cells have been reported to constitutively express the active form of the cellular HA receptor CD 44 (35). BW 5147 cells, therefore, bind readily to Cov-HA but not to Cov-NH (Fig. 9, A and B). Modification of the surface-bound HA of Cov-HA by either binding of TSG-6, HC transfer mediated by TSG-6 in solution, or HC transfer by TSG-6 bound to Cov-HA did not significantly affect the binding of BW 5147 cells to the modified substrate ( Fig.  9, C, D, and E). Obviously, the binding of BW 5147 cells (Fig. 9E) does not correlate with either the binding of TSG-6 ( Fig. 9C) or the presence of bound HCs (Fig. 9D) but is clearly dependent on the presence of HA. Similar experiments carried out with Cov-HA plates that had been modified with either a 5-fold higher or 5-fold lower HA concentration compared with the HA concentration used for the experiments depicted in Fig. 9, produced essentially the same pattern of BW 5147 cell binding or HC transfer (supplemental Figs. S3-S6).

DISCUSSION
The complex interactions between TSG-6, I␣I, and HA have physiological significance in fertility and in inflammation. Lack of either TSG-6 or I␣I results in infertility in mice, and TSG-6 has been shown to have potent anti-inflammatory and chondroprotective activities in experimental models of acute inflammation and arthritis.
Although earlier work demonstrated the transfer of HCs to HA resulting in HA⅐HC complexes, but not in HA⅐TSG-6 complexes, our data obtained with surface-associated HA resulted in the generation of both HA⅐TSG-6 and HA⅐HC complexes. Our data further show that the HA⅐TSG-6 complexes themselves are able to interact with I␣I and transfer HCs to HA, thereby forming an alternative pathway for the generation of HA⅐HC complexes. This alternative pathway of HA⅐HC complex formation may be active in hyaluronan-rich tissues in parallel to HC transfer by TSG-6 in solution. HC transfer by HA⅐TSG-6 complexes may become particularly important in instances when the local expression of TSG-6 pre- cedes the influx of I␣I from the vascular compartment. Although TSG-6 is expressed locally both in inflammation and in ovulation, access of I␣I from plasma depends on increased permeability of the local blood vessels (36).
One goal of our study was to determine whether the final products of the interaction of I␣I with HA⅐TSG-6 complexes would be HA⅐TSG-6⅐HC complexes or HA⅐HC complexes, i.e. if the putative HA⅐TSG-6⅐HC complexes are final products of the interaction of HA⅐TSG-6 complexes with I␣I or intermediates that actually transfer the HCs to HA. We demonstrated that the final product of this alternative HC transfer is the same as after HC transfer by free TSG-6, i.e. a HC directly bound to HA (Fig.  6A). This is consistent with the structure of the HA⅐HC complexes identified in synovial fluid of patients with rheumatoid arthritis and in ovulatory COCs (17,19). Interestingly, the HA⅐TSG-6⅐HC intermediate complex was detected only in trace amounts and mostly at high ionic strength (in TTBS), when the formation of HA⅐HC complexes was suppressed. The detection of only trace amounts of the HA⅐TSG-6⅐HC interme-diate in TTBS and its absence in PBS is consistent with the inhibition of both formation and depletion of the intermediary complex in TTBS. An obvious reason for the lack of accumulation of the HA⅐TSG-6⅐HC intermediary complex in PBS, i.e. under permissive conditions, is that HA is simultaneously part of the HC donor complex and final HC acceptor.
We also demonstrated that HA⅐ TSG-6 complexes can transfer HCs to an HA chain different from the one forming the HA⅐TSG-6 complex. This is mechanistically interesting because it requires that the TSG-6 in the HA⅐TSG-6 complex simultaneously interacts with a second HA chain. Assuming that the interaction between the HA⅐TSG-6 complex and the second HA chain is mediated by the HA binding site on the link module of TSG-6, we tentatively conclude that the HA chain forming the stable HA⅐TSG-6 complex does not block the HA binding site. Further work, outside the scope of this study, will be necessary to identify the nature of the stable bond between TSG-6 and HA and its location on the link module of TSG-6.
In the case of HC transfer to HA in solution there is no evidence for the formation of stable HA⅐TSG-6 complexes (Fig. 8, B and C). This is in agreement with published data (11), but it is strikingly different from what we find with HA tethered to a surface. In addition, the results of the interaction of TSG-6, I␣I, and HA in solution cannot explain the generation of stable complexes comprising HA, HC, and TSG-6 in addition to stable HA⅐HC complexes in ovulatory COCs (19). The presence of an alkalisensitive ester bond between TSG-6 and the HC in the stable ternary complexes present in COCs (19) provides an important clue for the relative positions of the three components. HCs form ester bonds in I␣I, HA⅐HC, and TSG-6⅐HC complexes, and in each case the ester bond involves the ␣-carboxyl group of the C-terminal aspartic acid residue of the HC (6,12,17), making it impossible for HCs to form ester bonds simultaneously with HA and TSG-6. Therefore, the stable ternary complexes present in ovulatory COCs (19) may have the structure HA⅐TSG-6⅐HC, identical with the intermediary complex for HC transfer by HA⅐TSG-6 complexes. In the presence of HA, these complexes should transfer HCs and generate HA⅐HC complexes, which were indeed present in the same location (19). FIGURE 6. Transfer of HCs by HA⅐TSG-6 complexes. 100 nM TSG-6 was incubated with 100 l of HA-Sepharose for 2 h at 37°C. After washing with TTBS, the HA-Sepharose was incubated with human plasma diluted 1:600 in either PBS or TTBS. Controls were incubated either with TTBS, TSG-6, or plasma only. All samples were treated with 10 units/ml hyaluronidase (HAse from S. hyalurolyticus. Released material was analyzed by immunoblotting using either rabbit anti-TSG-6 or rabbit anti-HC, as indicated. To provide controls for the various I␣I fragments, purified I␣I (I␣I) was diluted 1:2 with 50 mM NaCl and incubated for 30 min at ambient temperature (I␣I/NaOH). a, bikunin-HC1/HC2; b, HC2; c, HC1. Trace amounts of TSG-6⅐HC complexes are visible in both panels in the right lane at ϳ120 kDa. FIGURE 7. Transfer of HCs by HA⅐TSG-6 complexes to a different HA strand. 100 nM TSG-6 was incubated with 100 l HA-Sepharose for 2 h at 37°C. After washing with TTBS, the HA-Sepharose was incubated with human plasma diluted 1:600 and 100 g/ml SelectHA TM (molecular mass 30 kDa) in either PBS or TTBS for 2 h at 37°C. Controls were incubated either with TTBS, TSG-6 without plasma, or plasma without TSG-6 with SelectHA TM . Supernatants containing SelectHA TM but not the immobilized HA were analyzed by immunoblotting using either rabbit anti-HC or HABP-bio.
Structural modification of surface-associated HA by TSG-6 binding and HC transfer could be significant for its function as a receptor for lymphocytes and leukocytes. HA serves as a primary receptor for T cell extravasation in vivo (37)(38)(39) and for the infiltration of neutrophils in the liver of lipopolysaccharidetreated mice (40), with lymphocyte binding to HA being mediated by the cellular HA receptor CD 44 (38,(41)(42)(43)(44)(45). Therefore, we tested the possibility that modification of surface-bound HA by TSG-6-mediated HC transfer modulates the binding of BW 5147 cells that express fully activated CD 44 and bind readily to both HA in solution and to surface-bound HA (18,35). We report here that the stable modification of surface-bound HA by TSG-6 binding and HC transfer did not affect the subsequent binding of BW 5147 cells, independent of the mode of HC transfer by either TSG-6 in solution or by HA⅐TSG-6 complexes. Therefore, earlier reports of a correlation between the presence of HA⅐HC complexes and increased cell adhesion (46,47) may not apply to all cell types. However, there are also basic differences between the HA used here and in the earlier studies; the HA cables used by de la Motte et al. (46) were produced by and associated with cells in culture, whereas the HA⅐HC (SHAP (serum-derived HA-associated protein)) complexes employed by Zhuo et al. (47) had been isolated from synovium of patients with rheumatoid arthritis. It should be stressed that our data and published reports are not mutually exclusive and that HCs may affect cell adhesion in context with other factors that have yet to be identified.
Our data show that in the absence of TSG-6 no HC transfer to surface-bound HA occurs with purified I␣I or with highly diluted plasma, confirming data obtained with purified I␣I and HA in solution (10,11,33). However, added TSG-6 in picomolar concentrations is sufficient for HC transfer, demonstrating the high efficiency of TSG-6 ( Fig. 1, A and B). However, and in agreement with other published data (16), we found that modest HC transfer to HA can take place in plasma in the absence of added TSG-6, provided that the plasma is not diluted more than about 1:10. Although we cannot definitively explain the discrepancy in the literature, the high efficiency of TSG-6-mediated HC transfer demonstrated by our data makes it feasible that the modest HC transfer observed in the absence of added TSG-6 and exclusively with undiluted or modestly diluted plasma (1:10) is the result of trace amounts of TSG-6 that are below the detection limit of currently available immunoassays. Alternatively, it cannot be ruled out that native I␣I has a limited capacity to transfer HCs to HA in the absence of TSG-6 and that this activity is lost during the purification of I␣I. The essential involvement of HC2 in any HC transfer (14) points to a catalytic function of this chain and supports such a possibility. However, significant generation of HA⅐HC complexes in vivo has only been associated with rheumatoid arthritis (15,48,49) and with the ovulatory COC (10,19,20). In both cases the expression of TSG-6 has been demonstrated (20, 50 -52).
In the presence of limited amounts of TSG-6 and sufficient HA, I␣I can be almost completely depleted, and HA⅐HC complexes and free bikunin are the final products (Fig. 8, A, B, and  D). A decrease of I␣I in plasma or serum has been associated with infectious disease and sepsis (32,(53)(54)(55). In lipopolysaccharide-treated mice, a model of sepsis, hepatic accumulation of HA⅐HC complexes has been observed (40), and HA⅐HC com-plexes have also been found in sera of patients with chronic liver disease associated with hepatitis virus infection (56). It is very likely that the decrease of I␣I concentrations in sepsis and the simultaneous generation of HA⅐HC complexes are related, and TSG-6 may be at least partially responsible for the HC transfer. We have reported earlier the presence of TSG-6 in sera of sepsis patients, making TSG-6-mediated HA⅐HC generation in sepsis feasible (57).
Although TSG-6 has consistently shown anti-inflammatory and tissue-protective effects in models of acute inflammation and arthritis (24 -26, 29, 58), the role of HA⅐HC complexes is far from clear. HA⅐HC complexes in synovial fluid of patients with rheumatoid arthritis correlated with the severity of the disease (49), and hepatic accumulation of HA⅐HC complexes was observed in lipopolysaccharide-treated mice suffering liver damage (40). However, low serum concentrations of I␣I in sepsis patients correlated with mortality (54), and more importantly, administration of I␣I reduced mortality in experimental sepsis in rats (59,60), strongly suggesting that I␣I has a protective role in sepsis. If both TSG-6 and I␣I play protective roles in the pathology of inflammation and if they interact with HA to form HA⅐HC complexes, it is intriguing to speculate that HA⅐HC complexes may not cause tissue damage but may have a FIGURE 9. HA-specific binding of BW 5147 cells is not affected by TSG-6 binding and HC transfer. BW 5147 (shown after staining with naphthol blue black) cells bind to untreated Cov-HA (A) but not to Cov-NH (B). The bar in A represents the equivalent of 29.5 m. C-E, Cov-HA plates were treated with 10 nM TSG-6, human plasma diluted 1:1000, or 10 nM TSG-6 plus human plasma 1:1000 (all in TTBS) for 2 h at 37°C as indicated. In addition, Cov-HA was treated with 10 nM TSG-6 in the same way followed by incubation with human plasma 1:1000 (in PBS) for 2 h at 37°C. Thereafter, TSG-6 binding, HC transfer, and BW 5147 cell binding to the plates was determined as described under "Experimental Procedures." C shows the binding of TSG-6, D depicts the transfer of HCs, and E shows the binding of BW 5147 cells. Each data point represents the mean of 6 wells Ϯ S.E. yet unknown protective function, too. However, whether the protective effect of I␣I is mediated by HA⅐HC complexes, bikunin generation, or by a completely unrelated mechanism has yet to be demonstrated.