TSG-6 Modulates the Interaction between Hyaluronan and Cell Surface CD44* □ S

Interactions between CD44 and hyaluronan are implicated in the primary adhesion of lymphocytes to endo-thelium at inflammatory locations. Here we show that preincubation of hyaluronan with full-length recombinant TSG-6 or its Link module domain (Link_TSG6) enhances or induces the binding of hyaluronan to cell surface CD44 on constitutive and inducible cell back-grounds, respectively. These effects are blocked by CD44-specific antibodies and are absent in CD44-nega-tive cells. Enhancement of CD44-mediated interactions of lymphoid cells with hyaluronan by TSG-6 proteins was seen under conditions of flow at shear forces that occur in post-capillary venules. Increases in the number of rolling cells were observed on substrates comprising TSG-6-hyaluronan complexes as compared with a substrate containing hyaluronan alone. In ligand competition experiments, cell surface-bound TSG-6-hyaluronan complexes were more potent than hyaluronan alone in inhibiting cell adhesion to immobilized hyaluronan. Link_TSG6 mutants with impaired hyaluronan binding function had a reduced ability to modulate ligand

Interactions between CD44 and hyaluronan are implicated in the primary adhesion of lymphocytes to endothelium at inflammatory locations. Here we show that preincubation of hyaluronan with full-length recombinant TSG-6 or its Link module domain (Link_TSG6) enhances or induces the binding of hyaluronan to cell surface CD44 on constitutive and inducible cell backgrounds, respectively. These effects are blocked by CD44-specific antibodies and are absent in CD44-negative cells. Enhancement of CD44-mediated interactions of lymphoid cells with hyaluronan by TSG-6 proteins was seen under conditions of flow at shear forces that occur in post-capillary venules. Increases in the number of rolling cells were observed on substrates comprising TSG-6-hyaluronan complexes as compared with a substrate containing hyaluronan alone. In ligand competition experiments, cell surface-bound TSG-6-hyaluronan complexes were more potent than hyaluronan alone in inhibiting cell adhesion to immobilized hyaluronan. Link_TSG6 mutants with impaired hyaluronan binding function had a reduced ability to modulate ligand binding by cell surface CD44. However, some mutants that exhibited close to wild-type hyaluronan binding were found to have either reduced or increased activity, suggesting that some amino acid residues outside of the hyaluronan binding site might be involved in protein self-association, potentially leading to the formation of cross-linked hyaluronan fibers. In turn, cross-linked hyaluronan could increase the binding avidity of CD44 by inducing receptor clustering. The ability of TSG-6 to modulate the interaction of hyaluronan with CD44 has important implications for CD44-mediated cell activity at sites of inflammation, where TSG-6 is expressed.
Hyaluronan (HA), 1 a high M r glycosaminoglycan composed of repeating disaccharides of glucuronic acid and N-acetylglucosamine, is present in the extracellular compartment of most tissues and is a major component of cartilage and synovial fluid. The synthesis of HA is often up-regulated in response to inflammation, tissue damage, or invasion by tumor cells or pathogens (1)(2)(3). CD44, a major cell surface receptor for HA, is present on a wide variety of cell types. For instance, most hematopoietic cells express CD44, but their HA binding function is tightly regulated so that they do not constitutively bind HA (4,5). On T cells and B cells, HA binding can be activated in subsets of cells (5), where activating stimuli include phorbol esters and certain CD44-specific antibodies (6 -8). CD44 has been shown to participate in the migration of leukocytes to inflammatory sites (9 -12), where there is often an increase in CD44ϩ cells due to both the immigration of CD44ϩ leukocytes and to a greater degree of CD44 expression on resident cells (1). CD44 expression is increased on many cell types in inflamed arthritic joints (11,13), in which the synthesis and degradation of HA are also enhanced (14,15). TSG-6 (the secreted product of tumor necrosis factor-␣-stimulated gene-6) is an HA-binding protein, the expression of which is very tightly regulated (for reviews, see Refs. 16 -18). There is little or no constitutive expression of TSG-6 in adult tissues, but the protein is synthesized by fibroblasts, chondrocytes, monocytes, and vascular endothelial, and smooth muscle cells in response to stimulation with pro-inflammatory mediators or certain growth factors. TSG-6 protein has been detected at high levels in the synovial fluids of patients with arthritis and has been localized in articular cartilage and synovium from individuals with osteoarthritis and rheumatoid arthritis but is absent in normal joint tissues (16,19). Although exogenous TSG-6 has been shown to have anti-inflammatory activity in certain in vivo model systems (20 -24), the function of endogenously produced TSG-6 in the inflammatory milieu is unknown. It has been suggested that TSG-6 might modulate cellular interactions with HA at sites of inflammation (25).
The HA binding domains of TSG-6 and CD44 both contain a "Link module," a structural unit of about 100 amino acids that is present in many other members of the hyaladherin family including the proteoglycan aggrecan and cartilage link protein (17,18,26,27). Although the HA-interaction surfaces in CD44 and TSG-6 appear to be similar (27)(28)(29)(30), there are clearly significant differences in the molecular details of HA binding in these proteins (29). For example, recent structural studies show that in the case of CD44, N-and C-terminal sequences flanking the Link module are required to form a folded and functionally active domain (27). Although TSG-6 binds HA with high affinity (28), the binding between CD44 and HA is much weaker (31,32) and seems to depend on multivalent interactions (33)(34)(35). Therefore, extensive occupancy of HA with TSG-6 might be expected to inhibit its interaction with CD44 by reducing the number of sites available for binding.
In this study we have asked how the presence of TSG-6 influences the HA binding function of CD44ϩ cells. We have found that recombinant full-length human TSG-6 as well as its isolated Link module domain (Link_TSG6) are potent modulators of HA binding to CD44 on lymphoid cell lines. This finding has important implications for the regulation of CD44-mediated leukocyte migration to sites of inflammation.

EXPERIMENTAL PROCEDURES
Cell Lines-AKR1 is a CD44-negative mouse T lymphoma (isolated from an AKR/J mouse) that does not bind HA. AKR1/CD44ϩ is a transfectant of AKR1 with cDNA encoding the CD44.1 allele. AKR1/ CD44ϩ cells bind HA constitutively (36,37). AKR1/CD44⌬CY is a transfectant of AKR1 that expresses a mutant CD44 molecule whose truncated cytoplasmic domain contains only the six most membraneproximal amino acids and which binds soluble HA poorly without induction (36 -38). EL4 is a T lymphoma that expresses CD44 at a level similar to AKR1/CD44ϩ but does not bind HA constitutively (39). HA binding to both AKR1/CD44⌬CY and EL4 cells can be induced by the CD44-specific mAb IRAWB14 (5, 36, 39 -41). AKR1/CD44R41A is a transfectant of AKR1 that expresses a mutant CD44 with an Arg 3 Ala substitution at a critical position (Arg-41) in the HA-binding site of CD44 (27,30,31). The AKR1/CD44R41A transfectant does not bind HA at all, and in contrast to AKR1/CD44⌬CY or EL4 cells, HA binding to this mutant cannot be induced by IRAWB14 treatment (36 -38). All of the CD44ϩ cell lines used in this study express the hematopoietic (H) or standard isoform of CD44 that contains none of the variant exon products (5,42).
Preparation and Characterization of Full-length Human TSG-6 and Link_TSG6 -Full-length human TSG-6 (allelic variant TSG-6Q with Gln at amino acid position 144 in the pre-protein) was expressed in Drosophila Schneider 2 cells and purified by ion exchange chromatography and reverse-phase high performance liquid chromatography as described previously (43).
The Link module corresponds to residues 36 -133 in the TSG-6 preprotein (25). Cloning and overexpression of this module (Link_TSG6) in Escherichia coli has been reported previously (44,45). Link_TSG6 mutants H4K, E6K, Y12F, K13E, F70V, and Y78F were produced as described in Getting et al. (24) and Mahoney et al. (29). The additional H29K mutant was made in an identical manner and was analyzed by electrospray ionization mass spectrometry and one-dimensional NMR spectroscopy as before (29); H29K had an experimental mass within 0.8 Da of its theoretical mass and gave a one-dimensional NMR spectrum essentially identical to wild-type (WT) protein (data not shown). The affinity of the interaction between H29K and an octasaccharide of HA (HA 8 ) was determined by isothermal titration calorimetry on a Micro-Cal VP-ITC instrument at 25°C in 5 mM Na-MES, pH 6.0, exactly as described for the other mutants (24). In this regard, HA 8 has been shown to bind to Link_TSG6 with optimum affinity (26).
Analysis of HA Binding to Cell Surface CD44 -HA binding to cells was determined by flow cytometry using HA (from rooster comb, protein-free; Sigma) conjugated with fluorescein (FL-HA) as described before (39). The cells were labeled (at FL-HA concentrations indicated in Figs. 1-4 and Table I), and the fluorescence of ϳ10,000 cells in each sample was measured. FL-HA binding by each cell population is expressed as the mean fluorescence intensity (MFI) of cells incubated with FL-HA (with or without TSG-6 proteins) relative to the background fluorescence of the same cells incubated with diluent alone (MFI ϭ 1.0). FL-HA and TSG-6 proteins were added to the cells directly or after preincubation (see "Results").
Parkar et al. (46) observed that Link_TSG6 binding to HA was optimal at pH 6.0 and significantly reduced at pH 7.0. In preliminary experiments we found that preincubation of HA and Link_TSG6 at either pH 6.0 or 7.0 resulted in similar binding to CD44. However, binding of HA to CD44ϩ cells was reduced at pH 6.0 whether or not Link_TSG6 was present. Therefore, labeling of cells was performed at pH ϳ7, whereas preincubation was at pH 6.0.
For preincubation of FL-HA with TSG-6Q or Link_TSG6, a small aliquot of FL-HA (at ϳ800 g/ml) was added to the stock solution of each protein (at 250 g/ml or 1 mg/ml in 50 mM Na-HEPES, pH 6.0) and incubated undiluted for 30 min at room temperature. The FL-HA/ protein mixture was then diluted in DMEM (buffered with Na-HEPES, pH 7.2) or in phosphate-buffered saline (pH 7.3) to the highest concentration to be used for labeling cells, and serial dilutions were made from that dilution. Control samples of FL-HA without TSG-6Q or Link_TSG6 were preincubated in HEPES alone. Kinetics of dissociation of FL-HA from the cell surface and blocking of FL-HA binding with unlabeled HA were determined as described in Lesley et al. (34). Low M r FL-HA fragments were prepared by digestion with Streptomyces hyaluronidase (Fluka, Buchs, Switzerland) and separated by filtration through Centricon filters (Millipore, Billerica, MA) of various molecular weight cut-off as described previously (35).
Determination of the enhancing/inducing activity of mutant Link_TSG6 proteins relative to the activity of WT Link_TSG6 was performed as follows. WT and mutant proteins preincubated with FL-HA at high concentrations (at a w/w ratio of 5:1) were diluted to 0.5 g/ml FL-HA and assayed for binding to AKR1/CD44ϩ cells by flow cytometry. The MFI of FL-HA binding in the presence of each mutant protein was expressed as the percentage of FL-HA bound in the presence of WT Link_TSG6.
Flow Chamber Experiments-The influence of TSG-6Q or Link_TSG6 (WT and mutant) proteins on HA-dependent cell rolling and adhesion was determined using a parallel flow chamber (GlycoTech, Rockville, MD) as described previously (32,37). To allow for TSG-6⅐HA complex formation, HA was preincubated with TSG-6 proteins (or with diluent only) in 50 mM HEPES, pH 6.0, for 30 min at room temperature, as described for the FL-HA binding experiments. The volume was adjusted to 10 l with phosphate-buffered saline. This substrate (containing HA or protein-HA complexes) was placed as a single spot in the center of a 60-mm plastic Petri dish, dried onto the surface overnight, washed extensively with phosphate-buffered saline, and blocked with DMEM containing 10% ultra-pure bovine serum albumin (Sigma). The flow chamber was placed over the bottom of the dish with the flow path covering the substrate-coated area. The chamber was perfused with the cell suspension using a programmable syringe pump (Harvard Apparatus, Holliston, MA). Cells were accumulated at a shear force of 0.2 dyn/cm 2 followed by exposure to increasing levels of fluid shear (0.5, 1, 2, 4, and 8 dyn/cm 2 , each for 2 min). Cell movement was recorded by streamline acquisition of images using a digital camera (RS Photometrics, Trenton, NJ) attached to a Nikon Diaphot inverted phase contrast microscope. Images were collected and analyzed using a Metaview image processing system (Universal Imaging). The number of rolling cells was determined by counting the cells that moved on the substrate parallel with the fluid flow (32,37) at each level of shear force between 0.5 and 8 dyn/cm 2 . To assess firm cell adhesion, the number of immobile cells on the substrate-coated area was determined at the end of each 2-min flow period. Cells in at least three fields of microscopic view were counted at each shear force, and the average number of rolling or adherent cells was calculated. The enhancement of HA-dependent cell rolling or firm adherence by TSG-6 proteins was expressed as the ratio of the number of rolling or adherent cells on the protein/HA substrate to the number of rolling or adherent cells on the HA substrate. To determine the efficiency of Link_TSG6 mutants (relative to WT) to support cell rolling, WT and mutant proteins were preincubated with HA in a similar manner before immobilization in the flow chamber. The activity of substrates containing Link_TSG6 mutants was expressed as the percentage of the activity of WT Link_TSG6.
For competition experiments, cells were pretreated in suspension with HA or HA complexed with TSG-6 proteins before the rolling assays. First, HA and TSG-6 proteins (or diluent) were preincubated (1 g/ml HA, 5 g/ml protein) as described above. The mixture was then diluted with 1 ml of DMEM and incubated with the cells (4 ϫ 10 6 /ml) for 30 min at 37°C. The final volume was adjusted to 4 ml (without washing the cells) with DMEM before the flow chamber experiment. Cell rolling and tight adherence under flow (at 2 dyn/cm 2 shear force) in the chamber (coated with HA alone) were assayed as described above, and the effects of pretreatment with soluble reagents were expressed as percent inhibition of rolling or adhesion interactions with immobilized HA.

RESULTS
Influence of Link_TSG6 on the Binding of Soluble HA to Cell Surface CD44 -The HA binding activity of TSG-6 has been shown to reside within its Link module (24,28,29,47); therefore, this well characterized domain (Link_TSG6) was used first to examine the effects of TSG-6 on the interaction of HA with cell surface CD44. Because HA 20 is the minimum length of HA that can accommodate 2 Link_TSG6 molecules with optimal affinity 2 (indicating that each Link module occupies ϳ10 sugar units), a 5:1 (w/w) ratio of Link_TSG6 (M r ϭ 10.9) to HA would be expected to saturate the majority of protein-binding sites on an HA chain. Therefore, this approximately equimolar ratio of Link_TSG6 to HA (when expressed as 10-mer eq) was used for initial titration experiments.
Link_TSG6 was tested for its ability to modify the binding of FL-HA to cell surface CD44 in three ways as follows. 1) Link_TSG6 (ϳ1.5 M final concentration) and FL-HA (3 g/ml; ϳ1.5 M 10-mer eq final concentration) were diluted separately and added to the cell suspension. 2) Link_TSG6 (ϳ20 M final concentration) was preincubated for 30 min at room temperature with FL-HA (ϳ20 M 10-mer eq final concentration) before dilution of the mixture and addition to cells. 3) Link_TSG6 was preincubated with CD44ϩ cells (up to 20 g/ml final concentration for 30 min) before FL-HA was added. Fig. 1A shows the results obtained with AKR1/CD44ϩ cells that express the active form of CD44 (i.e. bind HA constitutively). When Link_TSG6 and FL-HA were added to the cells according to protocol 1 (diluted before mixing) there was no difference in the FL-HA binding to CD44ϩ cells (filled circles) compared with that seen in the absence of protein (open triangles). Unexpectedly, HA binding was enhanced when FL-HA was preincubated with Link_TSG6 before dilution and incubation with the cells (protocol 2, open circles in Fig. 1A). Preincubation of AKR1/CD44ϩ cells with Link_TSG6 followed by the addition of FL-HA (according to protocol 3 above), had no effect on FL-HA binding to the cells (results not shown). We found, however, that once the FL-HA⅐Link_TSG6 complex was formed by preincubation at high concentrations, the complex remained active after overnight incubation at 4°C even though it had been diluted in cell-labeling buffer (data not shown).
For the experiments shown in Fig. 1A the ratio of Link_TSG6 to FL-HA was constant. The influence of Link_TSG6 present at a fixed concentration throughout the titration of FL-HA was also examined, and the ratio of Link_TSG6 to HA increased as the concentration of FL-HA was reduced (Fig. 1B). These results are expressed as "enhancement ratio" i.e. MFI of FL-HA bound in the presence of Link_TSG6 relative to MFI of FL-HA bound alone (in Fig. 1B (y axis) a ratio greater than 1.0 indicates enhancement, whereas a ratio of less than 1.0 represents inhibition of FL-HA binding). Without preincubation, there was no enhancement of FL-HA binding in the presence of Link_TSG6 at either 2 g/ml (open triangles) or 10 g/ml (open circles). However, enhanced binding was observed (filled circles) after preincubation of Link_TSG6 (10 g/ml) with FL-HA when FL-HA was present at concentrations greater than 0.3 g/ml. At very low concentrations of FL-HA (Ͻ0.3 g/ml) and a high molar excess of Link_TSG6 (Ͼ30:1, protein:HA 10-mer eq), there was inhibition of FL-HA binding regardless whether or not Link_TSG6 and HA were preincubated (Fig. 1B), although preincubation resulted in greater inhibition of binding than did the direct addition of Link_TSG-6 and FL-HA. Binding of FL-HA to the cells from the preincubation mixture containing 10 g/ml Link_TSG6 was uniformly prevented at all concentrations of FL-HA by the CD44-specific mAb KM81 (Fig.  1B) that inhibits HA binding to mouse CD44 (40). ml (open triangles) throughout the titration of FL-HA. At 10 g/ml Link_TSG6, FL-HA was either preincubated with Link_TSG6 at a ratio of 3.3:1 (w/w) before dilution (filled circles) or was diluted and added to 10 g/ml Link_TSG6 without preincubation (open circles). Filled triangles represent FL-HA binding after preincubation with 10 g/ml Link_TSG6 to the cells in the presence of the CD44-specific mAb KM81 that blocks the HA-binding site of CD44 (40). The results are presented as a ratio (y axis) of MFI values for FL-HA bound in the presence or absence of Link_TSG6. When there is no effect of Link_TSG6 on FL-HA binding, the ratio is 1.0 (dotted line). Enhancement gives a ratio greater than 1.0, and inhibition gives a ratio less than 1.0. This is an example from two separate experiments with comparable results.

TSG-6 Modulates Hyaluronan Binding by CD44
cence of unlabeled cells), FL-HA binding to EL4 cells overlapped the background. Panels D-F of Fig. 2 show binding when FL-HA was preincubated with Link_TSG6 (using protocol 2). Binding to AKR1/CD44ϩ cells (panel E) was increased as indicated by the shift of fluorescence histogram to the right. EL4 cells bound FL-HA complexed with Link_TSG6 (Link_TSG6⅐FL-HA, labeled as Link_TSG6/FL-HA; panel F) significantly above background. Binding to both cell lines was further enhanced when FL-HA was complexed with full-length TSG-6 (TSG-6Q⅐FL-HA, labeled as TSG-6Q/FL-HA; bottom panels). However, FL-HA binding could not be induced by TSG-6 proteins in untransfected (CD44-negative) AKR1 cells (panels D and G). This observation together with the inhibition of Link_TSG6/FL-HA binding to AKR1/CD44ϩ cells in the presence of the KM81 mAb (see Fig. 1B) confirmed that enhancement/induction of HA binding by TSG-6 was mediated through cell surface CD44.
Influence of Full-length TSG-6 and Link_TSG6 on FL-HA Binding to the Cell Surface CD44 of Constitutively Active and Inducible Cell Lines-The effects of preincubating FL-HA with Link_TSG6 or the full-length protein (TSG-6Q) on the binding of HA to "constitutively active" AKR1/CD44ϩ cells and "inducible" EL4 cells were examined. TSG-6Q, like Link_TSG6, was able to enhance FL-HA binding to AKR1/CD44ϩ cells (see Fig.   FIG. 2. FL-HA binding in the absence and presence of Link_TSG6 or full-length TSG-6. These flow cytometry panels show examples of FL-HA binding to constitutive and inducible CD44-positive cells in the absence or presence of TSG-6 proteins and the lack of HA binding to the CD44-negative AKR1 cell line under these conditions. Untransfected AKR1 cells (left column), AKR1 cells transfected with wild-type CD44 (AKR1/CD44ϩ, middle column) that bind HA constitutively, and inducible EL4 cells (right column) were labeled with FL-HA (0.5 g/ml) alone (panels A-C) or FL-HA preincubated at 5:1 (w/w) ratio with Link_TSG6 (panels D-F) or full-length TSG-6 (TSG-6Q; panels G-I). Ten thousand cells were analyzed by flow cytometry, and dead cells (less than 5%) were gated out on the basis of staining with propidium iodide. Shaded histograms represent FL-HA fluorescence, and unshaded histograms show background fluorescence of unlabeled cells. The x axis indicates the fluorescence intensity of the FL-HA-labeled cells relative to cells "stained" with diluent alone, whose MFI was normalized to 1.0. CD44 expression levels were determined using fluorescein-conjugated CD44-specific mAb IM7 (FL-IM7). Relative MFIs of CD44 expression on AKR1, AKR1/CD44ϩ, and EL4 were Ͻ2, 61, and 57, respectively (not shown).

TSG-6 Modulates Hyaluronan Binding by CD44
2). Surprisingly, these proteins also caused a large induction of FL-HA binding to EL4 cells that in the absence of TSG-6Q or Link_TSG6 did not interact with FL-HA to any significant extent (see Fig. 2, right panels). Fig. 3 shows that at equimolar ratios to HA, the full-length protein (filled circles) was more effective than Link_TSG6 (open circles) in enhancing FL-HA binding by both types of cells. From these experiments it is also clear that the ratio of enhancement of FL-HA binding by either full-length TSG-6 or Link_TSG6 was greater for the inducible EL4 cells than the constitutive AKR1/CD44ϩ cells (compare y axis scales in Fig. 3, A and B).
EL4 cells and AKR1/CD44⌬CY (expressing a "tailless" deletion mutant of CD44 lacking all but the six most membraneproximal amino acids of the cytoplasmic domain) do not constitutively interact with FL-HA but can be induced to bind by treatment with the CD44-specific mAb IRAWB14 (5,36). Fig. 4 shows a titration of TSG-6Q and Link_TSG6 on EL4 and AKR1/CD44⌬CY cells and compares the enhancement of FL-HA binding with that induced by a constant amount of IRAWB14. In these experiments the ratio of TSG-6Q or Link_TSG6 in the preincubation mixture was 5:1 (w/w), whereas FL-HA was serially diluted from 4 to 0.05 g/ml. HA binding to both EL4 (Fig. 4A) and AKR1/CD44⌬CY (Fig. 4B) cells was greatly enhanced by both TSG-6Q (filled circles) and Link_TSG6 (open circles), particularly at higher FL-HA concentrations (Ͼ1 g/ml) where the effect of IRAWB14 (open squares) was more modest.
The Link_TSG6⅐FL-HA Complex Has a High Avidity for Cell Surface CD44 and Allows Stable Binding of Low M r HA Fragments-The relative avidity of FL-HA and Link_TSG6⅐FL-HA complexes for CD44 on AKR1/CD44ϩ cells can be compared by their sensitivity to blocking by unlabeled HA in a kinetic competition assay described in an earlier study (34). Binding of FL-HA at a defined concentration (0.3 g/ml) alone or preincubated with Link_TSG6 (at a w/w ratio of 5:1) was blocked with Results are expressed as enhancement ratio, as defined in Fig. 3. Binding of FL-HA alone to non-induced cells was equal to unstained cells (MFI ϭ 1.0) except at the highest FL-HA concentration (4 g/ml), when it was Յ1.5. Binding of FL-HA from the Link_TSG6⅐FL-HA complex to both EL4 and AKR1/CD44⌬CY cells was inhibited by the KM81 mAb, i.e. at the highest concentration of the components (4 g/ml FL-HA and 20 g/ml Link_TSG6) in the preincubation mixture, the enhancement ratio was Ͻ 2.0 for both cell lines in the presence of KM81.

TSG-6 Modulates Hyaluronan Binding by CD44
serial dilutions of unlabeled HA. As can be seen from Table I, row 1, ϳ4ϫ more unlabeled HA was required to block 50% of the binding of Link_TSG6-associated FL-HA to AKR1/CD44ϩ cells than for blocking the binding of FL-HA alone. This suggests that the Link_TSG6⅐FL-HA complex has a higher avidity for cell surface CD44 than does FL-HA. Significantly, unlabeled HA (at concentrations greater than 500 g/ml) was able to block the binding of the Link_TSG6⅐FL-HA complex completely (data not shown). These data together with the findings that TSG-6⅐FL-HA complexes do not bind to constitutive or inducible cell lines in the presence of the blocking mAb KM81 (Fig. 1B and data not shown) nor bind to the AKR1/ CD44Arg41Ala transfectant that expresses a mutant CD44 molecule unable to bind HA (see Supplemental Fig. S1) indicate that the HA-binding site of CD44 is responsible for interaction with the complex.
FL-HA bound to cell surface CD44 dissociates in the presence of excess unlabeled HA (34). Table I (row 2) shows that dissociation of the Link_TSG6⅐FL-HA complex was greatly retarded compared with FL-HA alone (a half-life of 100 and 18 min, respectively), suggesting a greater avidity of the complex for cell surface CD44.
The FL-HA used for analysis of binding to CD44ϩ cells is prepared from rooster comb HA with an average M r of ϳ2 ϫ 10 6 (48). Stable binding of FL-HA to CD44ϩ cells is not observed with HA of molecular mass below 100 kDa (35). As shown in Table I, FL-HA fragments with estimated M r of ϳ30 (row 3) and ϳ50 kDa (row 4) bound poorly to AKR1/CD44ϩ cells on their own but exhibited significant binding (over 2-fold increase in MFI) if they were preincubated with Link_TSG6 at a 5:1 (w/w) ratio.
TSG-6 Enhances CD44-dependent Cell Rolling on an HA Substrate-CD44 mediates lymphocyte rolling on HA in parallel plate flow chamber experiments that mimic the conditions of blood flow in post-capillary venules (37,49). As shown in Fig.  5A, when TSG-6Q (open circles) or Link_TSG6 (open triangles) was included in the HA substrate (each preincubated with HA at a 5:1 w/w ratio) there was a significant increase in the number of AKR1/CD44ϩ cells that rolled compared with the number of cells that rolled on immobilized HA alone (filled circles). Enhancement of HA-dependent rolling by TSG-6 proteins was seen even at a high shear stress (8 dyn/cm 2 ), although the number of rolling cells decreased at increasing fluid shear on all substrates. Thus, although cell capture and rolling were increased on the substrates containing TSG-6⅐HA complexes at each level of shear force, the cells exhibited an overall trend of increased detachment at higher shear forces as they did on the substrate containing HA alone (Fig. 5A).
A small fraction of AKR1/CD44ϩ cells, particularly those with very high CD44 expression, adhere tightly to HA and do not roll under flow (37). As compared with HA alone, there was a modest enhancement of tight adhesion to the HA substrate containing TSG-6Q but no significant enhancement of adherence to the Link_TSG6⅐HA substrate (Fig. 5B). When HA was preincubated with bovine serum albumin or another HA-binding protein (HA-BP from cartilage), no enhancement of rolling or firm adhesion was observed. 3 The AKR1/CD44⌬CY transfectant, which requires induction to bind FL-HA (36,38), is capable of attaching to and rolling on immobilized HA (37), as is the EL4-inducible cell line. The number of AKR1/CD44⌬CY cells and EL4 cells that rolled on substrates containing TSG-6 proteins was enhanced relative to HA alone, although in these cases there was no significant enhancement of firm adhesion (Fig. 6).
Binding of TSG-6⅐HA Complexes to CD44 Antagonizes Cell Adhesion to Immobilized HA-HA, bound from solution to constitutive CD44ϩ cells, can inhibit the rolling and firm adhesion interactions of the cells with immobilized HA in a concentrationdependent manner. 4 Therefore, it might be expected that HA, bound to the surface of AKR1/CD44ϩ cells in complex with full-length TSG-6 or Link_TSG6, would be more efficient than HA alone in blocking the rolling and adhesion interactions of the cells with the immobilized HA substrate. As shown in Fig.  7, cells pretreated with HA, TSG-6Q⅐HA, or Link_TSG6⅐HA exhibited reduced rolling on and adhesion to immobilized HA. The protein-HA complexes (bound to cells at w/w ratios of 5:1 to HA) had greater inhibitory effects on these interactions than HA alone. Both TSG-6Q⅐HA and Link_TSG6⅐HA as well as HA alone reduced firm adhesion more efficiently than rolling (compare Figs. 7, A and B).
Activity of Mutant Link_TSG6 Proteins-The HA binding properties of several mutant Link_TSG6 proteins with single amino acid substitutions are shown in Table II. The H29K mutant generated here was shown by NMR spectroscopy to have a WT fold, as is the case for the other mutants that were analyzed previously (24,29). Isothermal titration calorimetry experiments with the H29K mutant revealed that the alteration of His at amino acid 29 to Lys has no effect on the interaction of Link_TSG6 with HA (Table II). H4K has a slightly reduced affinity for HA, whereas Y12F, K13E, F70V, and Y78F have significantly lower HA binding activities (between 1 and 30% of WT); E6K, on the other hand, has an increased affinity for HA (24). Previously, we have shown that five amino acids of Link_TSG6 (Lys-11, Tyr-12, Tyr-59, Phe-70, and Tyr-78) play an important role in HA binding, and these are all clustered on one face of the Link module (29). From Fig.  8 it can be seen that His-4 and Glu-6 are on the face adjacent to the HA-binding site. As noted previously, the reduction in HA binding affinity seen with K13E (next to Tyr-12) is likely to be due to perturbation of a salt bridge between Lys-11 and HA (24). Conversely, the introduction of a positive charge at posi- a FL-HA (0.3 g/ml) or Link_TSG6⅐FL-HA complex was added to the cells in the presence of increasing concentrations of unlabeled HA. The concentrations of unlabeled HA at which the binding of FL-HA or Link_TSG6⅐FL-HA was blocked by 50% are indicated.
b The time at which 50% of FL-HA (at 0.3 g/ml alone or in complex with Link_TSG6 dissociated from the cell surface in the presence of 250 g/ml unlabeled HA. c Binding of low M r fragments of FL-HA to the cell surface (at 0.6 g/ml) alone or in complex with Link_TSG6, expressed as MFI.
tion 6 may allow the formation of an additional ionic interaction between HA and the protein (26). The effects of Link_TSG6 mutants on the interactions between HA and CD44 on AKR1/CD44ϩ cells (i.e. binding of FL-HA from solution and rolling on immobilized HA) were compared with that of the WT protein (Table II). It can be seen from these results that mutants with a significant reduction in HA binding activity (i.e. Y12F, K13E, F70V, Y78F) were also deficient in their ability to enhance the HA binding function of CD44. This supports the conclusion, from the data above, that enhancing/inducing activity depends on the interaction of Link_TSG6 with HA. However, the H4K mutant, which only has a slightly reduced HA binding affinity, exhibits significantly reduced ability to modulate CD44 function. In addition, H29K, which is on a face of the Link module opposite from the HA-binding site (Fig. 8), exhibited slightly increased enhancement of both FL-HA binding and cell rolling (ϳ160 and ϳ115% of WT activity, respectively; Table II) while having WT binding to HA 8 . Clearly, the residues involved in TSG-6-mediated enhancement of CD44 HA binding activity (i.e. His-4, Tyr-12, Lys-13, His-29, Phe-70, and Tyr-78) are widely spaced on the Link module. The results indicate that a large surface area that extends beyond the structural boundaries of the HA-binding site of the TSG-6 Link module is involved in mediating enhancement of HA binding to CD44.

DISCUSSION
In this study we have examined the effects of recombinant full-length human TSG-6 (TSG-6Q) and its Link module (Link_TSG6, a high affinity HA binding domain (28,29,47)) on HA binding by lymphoid cells expressing CD44. We found that complexes formed between HA and these proteins have CD44 binding properties very different from those of HA alone. Depending on the molar ratio of TSG-6Q or Link_TSG6 to FL-HA and the absolute concentration of the labeled HA in the complex, binding to CD44 (on a constitutively active cell background) is either inhibited or enhanced as compared with assays done in the absence of protein (Fig. 1). Inhibition is only seen when TSG-6 is present at high molar excess (at Ͼ30:1 w/w ratios to HA). Given the fact that the Link module of TSG-6 binds to HA with higher affinity than CD44 (32), it is perhaps not surprising that TSG-6 effectively competes with CD44 when HA-binding sites are limiting. At higher concentrations of HA and lower protein:HA ratios, binding of the TSG-6⅐HA complex to CD44 is enhanced compared with the binding of HA alone. This enhancement of HA binding by CD44 was also found to occur under physiologically relevant shear forces, where the presence of TSG-6 proteins in the immobilized HA substrate led to an increase in the number of rolling cells and, to a lesser extent, in the number of tightly adherent cells (Figs. 5 and 6). Reciprocally, binding of soluble HA or TSG-6⅐HA complexes to cell surface CD44 inhibited the subsequent interaction of cells with immobilized HA, where the protein-HA complex was a more potent antagonist of CD44-mediated cell adhesion than HA alone (Fig. 7).
CD44ϩ cells that do not constitutively bind HA, such as EL4 and AKR1/CD44⌬CY, were found to exhibit significant ligand binding function after HA preincubation with TSG-6 ( Figs.  2-4). In other words, these TSG-6-HA complexes are capable of switching CD44 into an active state, reminiscent of the induc-tion seen with certain antibodies (e.g. IRAWB14). TSG-6, like IRAWB14 (34), retards the dissociation of HA from the cell surface and facilitates binding of low M r HA fragments ( Table  I). The inducing antibody enhances HA binding by direct interaction with CD44 (34,36). Recent structural studies reveal that the epitope for IRAWB14 is on the opposite face of CD44 from its HA-binding site (27), which is consistent with a mechanism of activation involving receptor cross-linking/clustering. TSG-6, which does not appear to bind directly with CD44 or other cell surface molecules (see below and Supplemental Fig.  S2), is likely, therefore, to exert its enhancing/inducing effect by interacting with HA. A requirement for pre-formation of the TSG-6-HA complex using high concentrations of the reagents, and the stability of the preformed complex in dilute solution indicates that a structural rearrangement of the HA polymer occurs upon TSG-6 occupancy. It is possible that through selfassociation TSG-6 molecules cross-link multiple HA chains to form an "extended fiber" that exhibits increased receptor binding properties and might also facilitate CD44 clustering on the cell surface. Therefore, although the mechanisms by which TSG-6 and IRAWB14 influence the CD44-HA interaction are likely to be distinct, each of these reagents appears to promote coordinated binding of HA by multiple CD44 receptors, resulting in increased avidity. It should be noted that we have not found evidence for the direct binding of TSG-6 to CD44ϩ cells, as no cell surface-associated TSG-6 could be detected in the absence of HA (Supplemental Fig. S2). However, TSG-6 is retained on the cell surface in the presence of HA (Supplemental Fig. S2). This most likely occurs through the interaction of CD44 with HA in the TSG-6⅐HA complex, although we cannot completely rule out the possibility that this also involves a direct association between TSG-6 and CD44 (i.e. the formation of a ternary complex). In this regard, the molecular details of this interaction and the possibility of subsequent CD44-mediated signaling will be explored in further studies.
Cells that express CD44 in an inducible state bind soluble HA poorly without induction but are able to recognize immo-  bilized HA as an adhesion substrate in the absence of inducing stimuli. For example, AKR1/CD44⌬CY cells that fail to bind FL-HA, roll on an HA substrate under flow (37). Similarly, myeloid leukocytes from mouse bone marrow that are virtually unable to bind soluble HA were reported to exhibit attachment to immobilized HA (50). The absence of a requirement for inducing stimuli for CD44-mediated rolling suggests that HA in immobilized form is more favorable for interaction with cell surface CD44 than HA in soluble form. However, in this study we found that significantly more cells rolled on immobilized TSG-6⅐HA complexes than on HA alone regardless of the constitutive or inducible state of CD44 they expressed. Thus, TSG-6 bound to HA in the immobilized substrate can enhance the CD44-mediated rolling of both constitutive and inducible cells.
Results from mutagenesis studies indicate that the HA binding activity of Link_TSG6 is required for the enhancement of ligand binding to CD44ϩ cells, which is consistent with the observation that formation of a TSG-6⅐HA complex is necessary for enhancement. In addition, amino acids in Link_TSG6 that are not directly involved in HA binding are also implicated in the TSG-6-mediated modulation of CD44 function (i.e. His-4 and His-29; Fig. 8 and Table II). It is possible that these residues (and other amino acids in their vicinity) form a part of a self-association site in the Link module domain, leading to the formation of cross-linked HA fibers.
We have shown here that CD44-expressing cells that do not constitutively bind HA (as is the case for many peripheral leukocytes (6,8)) are able to bind TSG-6⅐HA complexes (Figs. [2][3][4] and exhibit enhanced rolling on substrates comprising these complexes. This observation suggests a mechanism whereby circulating leukocytes might become adhesive in an inflammatory milieu where TSG-6⅐HA complexes could be present. Because the expression/synthesis of HA and TSG-6 are up-regulated in blood vessels during inflammation (1-3, 18, 19, 51, 52), the inflammatory environment may provide the proper conditions for activation of "dormant" CD44 by creating a ligand consisting of HA modified by TSG-6. Indeed, CD44-mediated, HA-dependent rolling of T lymphocytes on an endothelial substrate was enhanced when the endothelial cells were stimulated with pro-inflammatory cytokines, which increased the retention of HA on the endothelial monolayer (53). Cytokine treatment could also induce TSG-6 expression in endothelial cells (52).
Although we demonstrate that the interaction of TSG-6 with HA can alter the HA binding function of CD44ϩ cells in both positive and negative ways, we cannot predict how these effects might influence inflammatory cell function in, for example, an arthritic joint. Are these functions of TSG-6 likely to be proinflammatory or anti-inflammatory? It is possible that enhanced cell binding to TSG-6⅐HA complexes (on the surface of endothelium) facilitates CD44-mediated recruitment of inflammatory cells whose CD44 is in an inducible state, thus exerting a pro-inflammatory effect. On the other hand, a large excess of TSG-6 in the presence of low HA concentrations (Fig. 1B) or binding of soluble TSG-6⅐HA complexes to the surface of leukocytes (Fig. 7) could inhibit the adhesive interactions of circulating cells with endothelium, thus having an anti-inflammatory outcome.
A further question is how HA signaling through CD44 might be influenced by the presence of TSG-6. It has been reported that CD44 signaling can be initiated by low M r HA (48,51,54,55), and we found that TSG-6 can promote binding of relatively small HA fragments to CD44 (Table I). Although CD44 has been shown to be involved in the migration of leukocytes into inflammatory sites in several in vivo model systems (9 -12), further studies will be needed to determine the role of TSG-6 in the local recruitment of CD44ϩ cells in inflammation and in other CD44-mediated cell functions.