Hyaluronan Binding Properties of a CD44 Chimera Containing the Link Module of TSG-6*

CD44, a cell-surface receptor for the extracellular matrix glycosaminoglycan hyaluronan, can mediate leukocyte rolling on hyaluronan substrates and has been implicated in leukocyte migration to sites of inflammation. CD44-mediated binding to hyaluronan is of low affinity, and effective cell/matrix interaction depends on multiple interactions with the multivalent ligand. We replaced the Link module of CD44 with the homologous region of TSG-6, a hyaluronan-binding protein secreted in response to inflammation whose Link module has a higher affinity for ligand. Monoclonal antibodies raised against the CD44/TSG-6 chimera recognized recombinant human TSG-6 and native mouse TSG-6 and blocked hyaluronan binding to these proteins. Cells expressing the CD44/TSG-6 molecule bound hyaluronan with higher avidity than cells expressing CD44. This resulted in changes in the hyaluronan binding properties characteristic of cells expressing CD44 such as requirements for threshold levels of receptor expression and for hyaluronan of high molecular mass. In parallel plate flow assays used to model leukocyte rolling, cells expressing CD44/TSG-6

The cell-surface glycoprotein CD44 is a receptor for the extracellular matrix glycosaminoglycan hyaluronan (HA) 1 (1)(2)(3)(4)(5). CD44 can mediate leukocyte rolling on HA substrates, such as those presented on endothelial cells activated by inflammatory cytokines, and has been implicated in leukocyte migration to sites of inflammation and in tumor cell metastasis (6 -10). Rolling is an early step in the cascade of events leading to extravasation from the blood stream into tissues, employed by cells of the immune system (11,12) and perhaps also by metastatic cancer cells (13). Among cell adhesion receptors, only selectins (11,12,14,15), CD44 (16,17), and, in some instances, ␣ 4 integrins (18) have been shown to mediate leukocyte rolling under conditions that mimic blood flow. The same receptors also mediate transient tethering, observed as a short pause in cell movement prior to or during rolling (14,19). All three selectins (E-, P-, and L-) and CD44 recognize carbohydrates as adhesion ligands (11,12,16,17). Integrins, a class of cell adhesion receptors that recognize protein ligands, do not mediate cell adherence at all under physiologic flow unless activated (12,15,20). Ligand binding by activated integrins leads to sudden cell arrest (permanent tethering) and firm adhesion (15). Integrin-mediated attachment to endothelium is the second step of the adhesion cascade that precedes extravasation (11,21), but integrins are incapable of arresting cells under physiologic shear stresses unless rolling has been initiated first (12).
The HA-binding region of CD44 has been mapped by mutational analysis (22,23). It contains a "Link module," a domain of ϳ100 amino acids found in other HA-binding proteins (24 -26). Additionally, in CD44, N-and C-terminal sequences flanking the Link module are also necessary for proper folding and functional activity (23,27). CD44 interaction with ligand is of low affinity, and CD44-mediated cell adhesion is dependent on multiple interactions with a multivalent substrate (28,29). TSG-6 (the protein product of tumor necrosis factor-stimulated gene-6) is a protein that is secreted by a variety of cells in response to inflammatory mediators (30,31) and that also contains a Link module. In this case, the Link module forms an independently folded domain that can support HA binding without the need for additional flanking sequences (24,32). The structure of the recombinant Link module from human TSG-6 (denoted Link_TSG6) has been determined in solution (24,33), and its HA-binding site has been characterized by NMR and mutational analysis (32,34). The TSG-6 coordinates have been used to model other Link module structures, including the Link module from CD44 (22,24). Although the threedimensional structures of the Link modules of CD44 and TSG-6 are likely to be similar, there are many differences in the amino acid residues involved in interacting with HA (22,26,34), and the TSG-6 Link module has a much higher affinity for HA than does CD44 (32).
Here, we constructed a chimera in which the Link module of CD44 was replaced with the Link module of TSG-6 to determine how a change in HA affinity might affect cell functions involving CD44-dependent HA binding. As expected, cells expressing the chimeric CD44/TSG-6 molecule bound soluble HA more avidly than cells expressing CD44. These cells differed in several properties that are characteristic of CD44-mediated HA binding such as the necessity for "threshold" levels of receptor and a requirement for high molecular mass HA. In addition, cells expressing the CD44/TSG-6 chimera failed to roll on a HA substrate under conditions of flow, illustrating the importance of multiple low affinity interactions between adhesion receptors and their ligands to promote the transient cell adhesion/ dissociation events characteristic of rolling.
Construction of CD44/TSG-6 Chimera and Transfections-A cDNA clone encoding the standard form of mouse CD44.1 in pBluescript SK and a cDNA clone encoding human TSG-6 (40) were used to create a CD44/TSG-6 chimera by overlap extension PCR. The Link domain of mouse CD44 (Gly 32 -Ala 123 in the preprotein) (4) was replaced with the Link module of human TSG-6 (Gly 36 -Ala 132 in the preprotein) (41). These regions are equivalent, and the module boundaries were chosen on the basis of the structural alignment of Kohda et al. (33). The sequence of this construct was verified by the Salk Institute DNA Sequencing Facility, and it was subcloned into the expression vector p304SR␣. This expression vector was prepared as follows: The HindIII site of the vector pcDL-SR␣296 (42) was destroyed, and the PstI-KpnI region was excised and replaced by the PstI-KpnI portion of the multiple cloning site from the vector Litmus 28 (New England Biolabs Inc.). The ϳ1.4-kb SalI fragment, containing the promoter derived from pcDL-SR␣296 and the multiple cloning site, was excised from pcDL-SR␣296, blunted, and cloned into the blunted HindIII and SacI sites of the vector p304 (43).
TSG-6-specific Hybridomas-Hybridomas producing mAb against TSG-6 were produced using standard techniques (49) by fusion of mouse SP2/0 myeloma cells (50) with spleen cells from Sprague-Dawley rats immunized against CD44-negative mouse cell lines transfected with the CD44/TSG-6 chimera. TSG-6-specific hybridomas were selected by screening supernatants by flow cytometry for binding to CD44/TSG-6-transfected cell lines and for lack of binding to cell lines expressing wild-type CD44 and to CD44-negative cell lines. Monoclonal antibodies A6, A38, and A68 are from an immunization with a CTLL-2 cell line expressing the CD44/TSG-6 chimera, and mAb Q75 is from an immunization with the XJ(3)/cell line CD44 Ϫ expressing the CD44/ TSG-6 chimera.
Immunoblotting-Immunoblotting of cell lysates was done as described previously (51) using mAb IM7 or mAb A38 supernatants and anti-rat immunoglobulin conjugated with horseradish peroxidase. A polyclonal antibody against the cytoplasmic domain of CD44 was provided by James McCarthy (University of Minnesota) and was used in combination with horseradish peroxidase-conjugated goat anti-rabbit immunoglobulin. Link_TSG6 and mutants with single amino acid substitutions (with His 4 of Link_TSG6 replaced with lysine designated as H4K, etc.) have been described (34). 2,3 Recombinant human TSG-6 was prepared as described (40). Mouse cumulus cell-oocyte complexes containing TSG-6 (provided by Csaba Fulop, Cleveland Clinic Foundation, Cleveland, OH) have been described (52,53). Medium from CHO-K1 cells secreting recombinant mouse TSG-6 was provided by Tibor T. Glant (Rush University, Chicago) and has been described (54).
HA Binding Plate Assay-A HA 2,3 inding plate assay similar to that described (34) was used. Briefly, recombinant human TSG-6 was coated onto a microtiter plate at 0.3 g/ml; FL-HA and medium or supernatants of A38 and Q75 were added; and HA binding was detected with horseradish peroxidase-conjugated rabbit antibody specific for fluorescein.
Parallel Plate Flow Assay-Cell rolling on HA was analyzed using a parallel flow chamber system (GlycoTech, Rockville, MD). Plastic Petri dishes, previously coated with high molecular mass HA (0.2 mg/ml) and blocked with 1% bovine serum albumin, constituted the bottom of the chamber. All flow experiments were carried out using serum-free Dulbecco's modified Eagle's medium kept at 37°C. Cells were introduced into the chamber at a low fluid shear force (0.25 dynes/cm 2 ) using an automated syringe pump (Harvard Apparatus, Holliston, MA). At any given flow path width and syringe diameter, the fluid shear force was proportional to the flow rate (12,16,17); therefore, flow rates were programmed to achieve shear forces ranging from 0.25 to 20 dynes/cm 2 . Movement of the cells under flow was captured by streamline acquisition of images through a cooled CCD camera (RS Photometrics, Trenton, NJ) attached to a Nikon Diaphot inverted microscope and recorded for 120 s using a Metaview image analysis system (Universal Imaging Corp., West Chester, PA). Cell movement (at least 250 cells/ experiment) was analyzed by playback of digital images employing the object measurement options of the Metaview program. To illustrate the relative distances covered by the cells during the entire recording period, the cells in the first image (starting position) were color-coded with red, and in the last image (ending position) with green. Overlay of the two images produced red and green (for rolling cells that covered distances greater than a cell diameter), red only (cells that detached under flow), or yellow (cells that remained attached, but did not roll). Fig. 1, the HA-binding domain of CD44 is represented schematically and compared with the same region of the chimera, designated as CD44/TSG-6, in which the Link module of CD44 has been replaced by the Link module of TSG-6. The approximate locations of amino acid residues required for HA binding in both Link module domains are represented by closed and open circles (see Refs. 22,23,and 34). Four amino acid residues C-terminal and one residue N-terminal to the Link module have also been implicated in binding (23). N-Linked glycosylation sites of CD44, previously shown to be involved in the regulation of HA binding (37,(55)(56)(57)(58), are indicated by N1-N5. In the CD44/TSG-6 chimera, the four CD44 N-glycosylation sites that are in the Link domain (N2-N5) are eliminated. Only one glycosylation site is present in the TSG-6 Link module at a sequence that is close to the N4 site in CD44.

A CD44 Chimera Containing the Link Module of TSG-6 -In
HA Binding by Cells Expressing CD44/TSG-6 -Several CD44-negative cell lines were transfected with cDNA encoding the CD44/TSG-6 chimera. The AKR1 cell line is a CD44-negative T lymphoma that binds HA constitutively when transfected with wild-type CD44 (59). CTLL-2 is a CD44-negative cytotoxic T cell line that also binds HA constitutively when transfected with CD44 (39). XJ(3)/CD44 Ϫ is a CD44-negative variant of a pre-B cell line that, when transfected with wildtype CD44, shows an "inducible" HA binding phenotype (37). Wild-type CD44 ϩ transfectants of this cell line do not bind HA constitutively, but can be induced to bind by certain CD44specific inducing mAbs.
Transfectants were identified in flow cytometry by binding of FL-IM7, a CD44-specific mAb whose binding depends on CD44 Pro 125 , which is C-terminal to the chimeric Link module (see Fig. 1 and Ref. 60). To eliminate cells spontaneously expressing endogenous CD44, IM7-positive lines were also screened for failure to bind KM81, a mAb that recognizes residues within the CD44 Link module (60). Transfectants expressing the chimera were cloned, and several positive clones from each line were assayed for binding of FL-HA and FL-IM7. For comparison, parallel assays were done on clones of AKR1 cells transfected with wild-type CD44 (closed circles). HA binding and expression of the transfected constructs were determined by flow cytometry using FL-HA and FL-IM7, respectively. As we have shown previously (36,37), cells expressing wild-type CD44 did not exhibit detectable binding of FL-HA until a certain threshold level of CD44 expression was reached, in this case ϳ20 times background levels (determined using unlabeled cells and untransfected AKR1 cells). In contrast, AKR1 and CTLL-2 cells expressing the CD44/TSG-6 chimera (closed squares and open circles, respectively) showed significant FL-HA binding at the lowest amounts of CD44/TSG-6 expression detected with FL-IM7. Transfectants of the inducible cell line XJ(3)/CD44 Ϫ with the chimera bound FL-HA constitutively. Although the XJ(3)/CD44 Ϫ cells transfected with the chimera bound less FL-HA than AKR1 and CTLL-2 cells with comparable levels of chimera expression, they bound significantly more than AKR1 cells expressing wild-type CD44.
It was not possible to obtain transfectants expressing the CD44/TSG-6 chimera at very high levels (above 20 times background levels). Cells expressing large amounts of the chimera formed large, tight clumps, and cells in the middle of the clumps died. In addition, these lines were not stable, and expression drifted downward.
Monoclonal Antibodies Specific for the CD44/TSG-6 Molecule-Although the chimeric CD44/TSG-6 molecule could be detected on the cell surface by the CD44-specific mAb IM7 and by several other CD44-specific mAbs dependent on the same epitope (see Ref. 60), these antibodies did not recognize the chimera on Western blots and did not immunoprecipitate it from detergent lysates. Therefore, we immunized rats with cells expressing the CD44/TSG-6 chimera and made mouse ϫ rat hybridomas with spleen cells from the rats. Hybridomas were screened for antibodies specific for the chimera or CD44 as described under "Experimental Procedures." Several hybridomas made mAbs that recognized the cellsurface chimeric CD44/TSG-6 molecule. Fig. 3 shows proteins from equal numbers of AKR1 cells expressing CD44 (lanes A and C) and CD44/TSG-6 (lanes B and D) immunoprecipitated with CD44-specific mAb (lanes A and B) or with one of the chimera-specific mAbs (lanes C and D). The immunoprecipitates were immunoblotted with polyclonal rabbit antiserum specific for the CD44 cytoplasmic domain. Only CD44 was precipitated by the CD44-specific mAb IM7 (lane A), and only the chimera was precipitated by the CD44/TSG-6-specific mAb A38 (lane D), but both molecules were recognized by the antiserum specific for the CD44 cytoplasmic domain. This result confirms the chimeric nature of the CD44/TSG-6 molecule. The Western blot (Fig. 3) also indicates that the CD44/TSG-6 chimera was expressed on AKR1 cells at 5-10-fold lower levels than wild-type CD44, which is consistent with the data from flow cytometry (Fig. 2). From Fig. 3, it can be seen that there was a difference in migration upon SDS-PAGE between CD44 and CD44/TSG-6 expressed in the same cell line, which is consistent with the presence of two N-glycosylation sites in the chimera, rather than five in the wild-type protein (see Fig. 1).
Four of the chimera-reactive mAbs showed specificity for the Link module of TSG-6 in Western blotting and enzyme-linked immunosorbent assay plate assays. Fig. 4 demonstrates that mAb A38 recognized Link_TSG6 (lane A) and recombinant human full-length TSG-6 expressed in insect cells (lane B). mAb A38 also recognized native mouse TSG-6 from cumulus cell-oocyte complexes digested with hyaluronidase to release proteins from the extracellular coat (lane C) and mouse TSG-6 in medium from Chinese hamster ovary cells expressing transfected recombinant mouse TSG-6 (lane D), but not in medium alone (lane E). The mouse preparations showed a band at ϳ37 kDa, representing fully glycosylated TSG-6, and a band at ϳ120 kDa (indicated by arrows). The 120-kDa band is a complex of TSG-6 covalently associated with another protein recognized by antiserum raised against serum protein inter-␣inhibitor (data not shown). This TSG-6⅐inter-␣-inhibitor complex has been reported previously in mouse cumulus cell-oocyte complexes (53,61). mAb A38 also immunoprecipitated recombinant human TSG-6 (data not shown) and mouse TSG-6 (lane D) as well as the CD44/TSG-6 chimera (shown in Fig. 3).
Blocking mAbs Recognize an Epitope in the Link_TSG6 HAbinding Site-Three of the Link_TSG6-specific mAbs (A6, A38, and A68) blocked binding of FL-HA to cells expressing the CD44/TSG-6 chimera (data not shown) and to recombinant human TSG-6. Fig. 5A shows that mAb A38 blocked HA binding to recombinant human TSG-6 immobilized on enzymelinked immunosorbent assay plates, whereas another chimera/ TSG-6-specific mAb (Q75) did not. Link_TSG6 proteins containing single amino acid mutations were used to probe the epitope recognized by mAb specific for the TSG-6 Link module. Two of these mutations result in a significant reduction of HA binding activity (K13E and Y78F) (34). 3 As shown in Fig. 5B, mAb A38 did not Western blot the Link_TSG6 protein in which Tyr 78 of the Link module was mutated to Phe; this mutation does not affect the Link module fold (34). This residue (indicated by an open circle in the Link module of TSG-6 in Fig. 1) has a critical role in HA binding and is likely to make direct interactions with the sugar (34). Other mutant Link_TSG6 proteins were recognized by mAb A38, although binding to H4K was weaker than to the other mutants. mAb Q75 recognized all of the mutant proteins (data not shown), so its binding epitope remains unknown. Two other mAbs that blocked HA binding (A6 and A68) also failed to blot Y78F, but their pattern of binding to the other mutants differed from that for mAb A38 (data not shown).
Although mAb A38 bound to the HA interaction site of the CD44/TSG-6 chimera, binding of the mAb was not readily inhibited by HA; 2 mg/ml unlabeled HA was insufficient to block 50% of the binding of fluorescein-conjugated mAb A38 (1 g/ml) (data not shown). This is probably due to a much higher affinity of mAb A38 for the TSG-6 Link module relative to the affinity of HA.
HA Binding Properties of Transfectants Expressing the CD44/TSG-6 Chimera-To verify that HA binding by cells expressing the chimera was of higher avidity than binding by cells expressing wild-type CD44, as suggested by the data shown in Fig. 2, we compared several other properties of these cell lines. We could not compare lines expressing similar levels of the transfected molecules because cells expressing large amounts of the chimera could not be obtained and because cells expressing low amounts of CD44 did not bind FL-HA detectably (see Fig. 2). Therefore, comparisons were made between lines that bound similar levels of high molecular mass FL-HA. This was achieved using cell lines in which the chimera was expressed at 10 -20% of the level of wild-type CD44, based on the binding of FL-IM7 (see Fig. 2). These relative levels of  D) were immunoprecipitated (i.p.) with the CD44-specific mAb IM7 (lanes A and B) and the chimera-specific mAb A38 (lanes C and D). Precipitates were subjected to 10% SDS-PAGE, transferred to membrane, and immunoblotted with polyclonal rabbit antiserum specific for the CD44 cytoplasmic domain.

FIG. 4. Immunoblots of TSG-6. Lanes A and B, recombinant human
Link_TSG6 (40 ng) and recombinant human TSG-6 (9 ng), respectively, were run on a 12% SDS gel and blotted with mAb Q75 (anti-Link_TSG6). The same result was obtained with mAbs A6, A38, and A68 (data not shown). Lane C, shown are extracts from mouse cumulus cell-oocyte complexes blotted with mAb A38. Lanes D and E, medium from Chinese hamster ovary cells secreting recombinant mouse TSG-6 and medium only, respectively, were immunoprecipitated with mAb A38, blotted with fluorescein-conjugated mAb A38, and detected with horseradish peroxidase-conjugated anti-fluorescein antibody. Arrows to the right of lanes D and E indicate the positions of intact TSG-6 and TSG-6 complexed with inter-␣-inhibitor (see "Results"). In all lanes, sample buffer was nonreduced and included 50 mM iodoacetamide. receptor expression were confirmed using other fluoresceinconjugated mAbs that recognized both CD44 and the CD44/ TSG-6 chimera (data not shown).
We showed previously that FL-HA fragments below ϳ100 kDa (obtained by digestion of high molecular mass FL-HA with hyaluronidase) do not bind stably to CD44 ϩ cells (29). This indicates that many linked HA-binding sites have to be engaged to retain the FL-HA molecule at the cell surface. Fig. 6A shows that such fragments did bind significantly to cells expressing the CD44/TSG-6 chimera.
High molecular mass FL-HA dissociates fairly rapidly from CD44-expressing cells in the presence of excess unlabeled HA (28,29). As shown in Fig. 6B, dissociation of FL-HA from cells expressing CD44/TSG-6 was significantly slower (t1 ⁄2 ϳ 120 min) than dissociation from wild-type CD44 ϩ cells (t1 ⁄2 ϳ 30 min) in the presence of competing unlabeled HA. Also, the low molecular mass FL-HA fragments, whose binding is shown in Fig. 6A, dissociated very rapidly from cells expressing the chimera (t1 ⁄2 ϳ 2 min) when unlabeled high molecular mass HA was present. Together, these results indicate that CD44/TSG-6 mediates cell-surface binding of HA with a higher avidity than CD44, although both interactions are likely to involve multiple binding events along the repetitive HA chain.
CD44/TSG-6 Binding Site Size-We (28) and others (62, 63) have found that a HA oligosaccharide of six sugar residues is the minimum size that can compete for binding of high molecular mass HA to cell-surface CD44. We also found that an oligosaccharide of 10 sugars blocks better than ones composed of six or eight residues. No further increase in blocking efficiency was observed with longer HA chains until 20-sugar chains were used. We concluded that, at chain lengths of 20 sugar residues (and above), two CD44 molecules can bind a single HA molecule, thus increasing the avidity and the corresponding blocking efficiency (28). Fig. 7 shows a similar experiment using unlabeled HA with chain lengths of 4, 6, 8, 10, 12, 14, 16, 18 -20, and 22-24 sugars to block binding of FL-HA to cells expressing CD44/TSG-6. In this case, a lower molecular mass FL-HA fragment (ϳ50 kDa) (see Fig. 6A) was used to detect binding because blocking of the binding of high molecular mass FL-HA to cells expressing CD44/TSG-6 required prohibitively high concentrations of the oligosaccharides. Efficient blocking was first seen with HA 6 , but improved slightly with HA 8 and more so with HA 10 -14 . As with cells expressing CD44, HA 4 was not an effective blocker of FL-HA binding. With HA 18 -20 and HA [22][23][24] , there was an ϳ3-fold decrease in the concentration required to block (50%) compared with HA 10 , HA 12 , and HA 14 , whereas HA 16 was intermediate in blocking activity. These data show that HA oligosaccharides of between 18 and 24 sugars are more effective inhibitors than the shorter oligomers. In this regard, we have reported previously that Link_TSG6 binds in a monovalent fashion with oligomers ranging from HA 6 to HA 16 and that the ligand-binding site on the Link module can accommodate an oligosaccharide of between six and eight sugar residues (32). Recent studies have indicated that an 18-mer of HA is the minimum size that can support binding of two molecules of Link_TSG6. 4 It seems therefore likely that HA oligosaccharides of between 18 and 22 sugar monomers can cross-link two CD44/TSG-6 molecules, indicating that the HA-binding site of the chimera is similar in size to that of CD44. Oligomers of HA 10 -16 are more effective inhibitors of FL-HA binding to CD44/TSG-6 (compared with HA 8 ) than would have been expected based on experiments on the isolated TSG-6 Link module (see Ref. 32), indicating that the HA binding properties of Link_TSG6 have been affected somewhat by its insertion into cell-surface CD44.
Rolling of CD44/TSG-6-expressing Cells on a HA Substrate-CD44 can mediate rolling of leukocytes on a substrate coated with HA under conditions of flow (17). We compared the behavior of CTLL-2 transfectants expressing wild-type CD44 (FL-IM7 binding 20-fold above background levels) with that of cells expressing the CD44/TSG-6 chimera (FL-IM7 binding 9-fold above background levels) in a parallel plate flow assay. Fig. 8A illustrates cell behavior on the HA-coated substrate under physiologic flow (2 dynes/cm 2 ). The overlay of color-coded images of a flow chamber experiment (where the starting position of cells is denoted in red, and the final position in green) indicates at a glance that CTLL-2 cells expressing CD44 (left panel) traveled to significant distances by rolling on HA, whereas untransfected cells (CD44-negative) (right panel) were swept away by the fluid flow. CTLL-2 cells expressing the CD44/TSG-6 chimera (center panel) showed attachment to HA that was strong enough to withstand fluid shear. As a consequence, these cells remained firmly stuck (or permanently tethered) to the HA substrate. Pretreatment of cells expressing the chimera with saturating amounts of the blocking mAb A38 inhibited all binding of these cells to immobilized HA (data not shown), giving pictures similar to those of the untransfected cells (right panel). Fig. 8 (B and C) summarizes the results of observations made over a range of fluid shear forces. Fig. 8B shows that a much higher percentage of cells expressing CD44 rolled compared with cells expressing the CD44/TSG-6 chimera at a wide range of fluid shear forces. In addition, the rolling velocity of cells expressing CD44 increased with increasing shear force, whereas the rolling velocity of cells expressing the chimera did not change (Fig. 8C). These results indicate that the cells expressing the CD44/TSG-6 chimera generally failed to roll. Similar results were obtained with AKR1 transfectants: cells expressing wild-type CD44 rolled on HA substrates under fluid shear force, whereas cells expressing the chimera remained "tethered" (data not shown).
We have not explicitly examined whether transfectants expressing very low levels of the CD44/TSG-6 chimera have a sufficiently low avidity for ligand to allow rolling on a HA substrate or whether transfectants expressing very high levels of wild-type CD44 might exhibit such a high avidity for ligand that they remain tethered at physiologic shear forces. We have observed that AKR1 transfectants expressing wild-type CD44 at 190-fold over background levels still exhibited rolling behavior, although the percentage of rolling cells and the rolling velocity at a high physiologic shear force (ϳ10 dynes/cm 2 ) were significantly reduced compared with transfectants expressing lower amounts of CD44 on their cell surface. 5 We have also found that a small percentage of cells expressing the CD44/ TSG-6 chimera demonstrated rolling on HA in the presence of very low concentrations of the blocking mAb A38, suggesting that reduction of cellular avidity for HA (by reducing the number of receptors available for interaction with ligand) would allow the cells to roll. 5

DISCUSSION
Like other adhesion receptors that mediate cell rolling, the interaction of CD44 with ligand is of low affinity, and CD44mediated cell adhesion is dependent on multiple interactions with a multivalent substrate (28,29). We have calculated the affinity of the interaction between HA 10 and the HA-binding domain of CD44 to be between 10 and 100 M, 6 which is consistent with the estimates of others (58). On the other hand, the K d for Link_TSG6 binding to HA 10 was determined (under equivalent conditions) to be 0.28 M (32). Therefore, the TSG-6 Link module has between ϳ35 and ϳ350 times higher affinity than CD44 for monomeric HA. We investigated what biological consequences would result from increasing the affinity of CD44 for HA by constructing and expressing a chimera with the Link module of CD44 replaced by that of TSG-6.
Initial measurements (Fig. 2) indicated that the TSG-6 Link module was functional in the context of CD44 and therefore was likely to be correctly folded; we have shown previously that, under conditions in which TSG-6 is unfolded, it does not bind to HA (64). These experiments also demonstrated that the chimeric receptor has dramatically increased HA binding compared with the wild-type protein. Cells expressing the chimera retained HA at the cell surface with fewer binding sites engaged, as shown by the high levels of FL-HA bound with fewer  7. Relationship between oligosaccharide size and inhibition of FL-HA binding. The micromolar concentration of low molecular mass HA oligosaccharide needed for 50% inhibition of binding of a FL-HA fragment of ϳ50 kDa to AKR1 cells expressing the CD44/TSG-6 chimera (y axis) is plotted against the size of the HA oligomer (number of sugar residues; x axis). In the case of HA oligosaccharide preparations that contained a mixture of sizes, e.g. HA 18 -20 and HA [22][23][24] , the data point is plotted at the average, and bars indicate the range of sizes. The 50% inhibiting concentration was determined from blocking curves such as those shown previously (28), where serial dilutions of unlabeled HA oligosaccharides were used to compete with binding of a fixed concentration of the FL-HA fragment (0.5 g/ml). All the blocking curves were determined in the same experiment using the same CD44/TSG-6 expressing cells and the same dilution of FL-HA. receptors than were required with CD44 (Fig. 2), by more stable binding of low molecular mass HA fragments (Fig. 6A), and by a reduced rate of dissociation of bound FL-HA (Fig. 6B). However, HA binding by the chimera probably retains the dependence on multiple receptor/ligand interactions, as suggested by the more rapid dissociation of low compared with high molecular mass HA (Fig. 6B).
The cell-specific post-translational modifications, which confer the different "activation states" on CD44, are evident in the reduced HA binding activity of the chimera expressed in the inducible cell line XJ(3)/CD44 Ϫ compared with expression in the "constitutively active" cell lines AKR1 and CTLL-2 (Fig. 2). This is likely to result from carbohydrate modifications at the N1 site in CD44 (which is N-terminal to the Link module) (see Fig. 1) that may lead to steric hindrance of HA binding. Prior studies have shown that elimination of glycosylation at this site by substitution of serine for Asn 25 converts CD44 on the cell surface and as a receptor-globulin (CD44-Ig) from an inducible to a constitutive HA binding phenotype (37). Clearly, the post-translational modifications occurring in the XJ(3)/ CD44 Ϫ cell line are not sufficient to completely suppress constitutive HA binding by the chimera. This is perhaps not surprising, as it seems unlikely that the Asn-linked carbohydrate in the N-terminal flanking sequence occurs in the same spatial orientation relative to the TSG-6 Link module in the chimera compared with the native protein. It should be noted that this N-terminal region is believed to form an integral part of the CD44 HA-binding domain (25, 27) and is therefore not expected to have a wild-type fold in the context of the chimera.
Several mAbs raised against the chimera recognized a residue within the binding site of Link_TSG6 (Tyr 78 ) and competitively inhibited HA binding by TSG-6 ( Fig. 5A). Although 50% inhibition of FL-HA binding to the cell-surface CD44/TSG-6 chimera was achieved with mAb A38 concentrations below 1 g/ml, milligram/ml concentrations of HA were required for inhibition of mAb A38 binding to the chimera (data not shown). This suggests that the affinity of mAb A38 for the Link module of TSG-6 would be sufficient to block TSG-6/HA interactions at concentrations that could be achieved in vivo, which may be useful in investigating the function of this protein.
The Link module has been found to have a fold very similar to that of the C-type lectin module (24,33). This structure is found in the carbohydrate-binding domain of selectins, the first cell-surface receptors shown to mediate leukocyte rolling on microvascular endothelium (12,21). Through multiple low affinity interactions with ligand, CD44 supports cell rolling on HA in cells that express the active (constitutive ligand-binding) form of the receptor (Refs. 16 and 17 and this study) in a manner similar to selectin-mediated rolling on carbohydrate substrates. CD44 that binds HA constitutively and functions to support rolling is found on T cells that have been induced by antigen stimulation (8,65,66), and constitutively active receptors are found on circulating T cells in patients with chronic inflammation (8,67). Furthermore, in a mouse model, activated T cells can use the CD44/HA interaction to home into a site of inflammation (8). The results shown here demonstrate that strong adhesion to a carbohydrate substrate, achieved through an increase in the intrinsic affinity of the ligandbinding domain of CD44 for HA, is incompatible with leukocyte rolling and transient tethering. The Link module of TSG-6 confers integrin-like characteristics to CD44 in the sense that cells expressing CD44/TSG-6 chimeric receptors become firmly adherent or permanently tethered to HA. These results suggest that the low avidity of an individual CD44 receptor for its ligand and the reversible and multivalent nature of the CD44/HA interaction are essential for the function of CD44 in mediating initial steps of cell migration into an inflammatory site.