Hyaluronan Binding to Link Module of TSG-6 and to G1 Domain of Aggrecan Is Differently Regulated by pH*

The physiological functions of hyaluronan (HA) in the extracellular matrix of vertebrate tissues involve a range of specific protein interactions. In this study, the interaction of HA with the Link module from TSG-6 (Link_TSG6) and G1 domain of aggrecan (G1), were investigated by a biophysical analysis of translational diffusion in dilute solution using confocal fluorescence recovery after photobleaching (confocal FRAP). Both Link_TSG6 and G1 were shown to bind to polymeric HA and these interactions could be competed with HA8 and HA10 oligosaccharides, respectively. Equilibrium experiments showed that the binding affinity of Link_TSG6 to HA was maximal at pH 6.0, and reduced dramatically above and below this pH. In contrast, G1 had maximum binding at pH 7.0–8.0 and moderate to strong binding affinity over a much broader pH range (5.5–8.0). The KD determined for Link_TSG6 binding to HA showed a 100-fold increase in binding affinity between pH 7.4 and 6.0, whereas G1 showed a 75-fold decrease in binding affinity over the same pH range. The sharp difference observed in their pH binding suggests that pH controls the physiological function of TSG-6, with a low affinity for HA at neutral pH, but with increased affinity as the pH falls below pH 7. TSG-6 and aggrecan interact with HA through structurally homologous domains and the difference in pH-dependent binding can be understood in terms of differences in the presence and topographical distribution of key regulatory amino acids in Link_TSG6 and in the related tandem Link domains in aggrecan G1.

Hyaluronan (HA) 3 is a high molecular mass (up to 10 7 Da) linear glycosaminoglycan (GAG), composed of repeating disaccharide units of N-acetyl-D-glucosamine (GlcNAc) and D-glu-curonic acid (GlcA) (1,2). It is a major component of the extracellular matrix (ECM) of vertebrate tissues acting as a scaffold for the association of various HA-binding proteins (hyaladherins (3)) leading to the formation of diverse molecular architectures (4,5). In cartilage HA has a profound influence on the structure and biomechanical properties of the tissue, being an integral part of the supramolecular aggregates formed by the proteoglycan aggrecan (6) and cartilage link protein (7)(8)(9); similar complexes may also be involved in the formation of perineronal nets during brain development (10). Aggrecan is the major proteoglycan of articular cartilage and consists of a protein core (225 kDa), to which are attached Ͼ100 chondroitin sulfate and keratan sulfate chains (7,11,12). It binds to HA via its N-terminal G1 domain, where this interaction is stabilized by cartilage link protein (11,13), which functions to anchor up to several hundred aggrecan molecules to each HA chain, thereby forming supramolecular aggregates (4 ϫ 10 7 to 4 ϫ 10 8 Da). The formation of aggregates and their immobilization within the fibrillar collagen network of the ECM is responsible for the load-bearing properties of articular cartilage (6,14).
Both aggrecan and cartilage link protein contain tandem pairs of Link modules within their HA-binding domains and it is these that mediate their highly specific binding to HA (3); this type of module is found in a family of hyaladherins. TSG-6, an HA-binding protein associated with inflammation (15,16), is another member of the Link module superfamily containing a single Link module, which binds to HA, but in contrast to other members it also binds to chondroitin 4-sulfate (C4S) (see (17)). TSG-6 is expressed in human cartilage (and synovium) during arthritis (18). In animal models of arthritis TSG-6 has been shown to have anti-inflammatory and chondroprotective properties in joint tissues (19 -24). This protective effect may be mediated by a number of mechanisms (25) including downregulation of the protease network (19,26) and inhibition of neutrophil migration (19,20). These properties of TSG-6 may not directly involve its binding to HA (20,26).
Previously we have determined the three-dimensional structure of the TSG-6 Link module by both NMR spectroscopy (16,27) and x-ray crystallography (28), which has defined the canonical fold for this type of domain. Determination of the structure of Link_TSG6 in the presence of an HA octasaccharide (HA 8 ) has allowed us to elucidate the bound conformation for this protein (27) and to model the HA⅐TSG-6 complex and other members of the Link module superfamily (29). With cartilage link protein and aggrecan, a speculative model was gen-erated of how these molecules may associate with HA to form a super-helical structure (29).
We have reported previously that the binding of HA to the TSG-6 Link module is optimal at pH 5.5-6.0 with a major decrease in binding above or below this pH (30). This unusual pH dependence appeared to be in marked contrast to that for the G1 domain of aggrecan, where the highest affinity was reported at neutral pH 7.0 -8.0 (31). However, it has been suggested that the unusual pH dependence we reported for the HA-Link_TSG6 interaction may have resulted from an artifact of the solid phase assay used in the analysis (32). In recent work, isothermal titration calorimetry (ITC) was used to show that the Link_TSG6 had a higher affinity at pH 6.0 compared with pH 7.5 for short oligosaccharides of HA in solution phase and a structural basis for the pH dependence was identified using a combination of site-directed mutagenesis, NMR and ITC (33). However, the differences in affinity determined by ITC for the Link_TSG6 -HA interactions at pH 6.0 and 7.5 were fairly modest (ϳ2.5-and ϳ7-fold higher for HA 8 and HA 20 , respectively (33)) when compared with the dramatic changes apparent in the solid phase assays (30). As a large change in HA binding between pH 6.0 and 7.5 could have a major effect on the physiological function of TSG-6, it was important to gain more evidence of the pH dependence of the binding of Link_TSG6 to HA in solution, in particular, for polymeric HA, which has not been investigated to date.
In this study we used confocal fluorescence recovery after photobleaching (confocal FRAP) to characterize more fully the HA binding of the TSG-6 Link module and to compare it with the binding properties of the structurally related G1 domain of aggrecan. This technique determines the translational diffusion coefficients of fluorescently tagged molecules and provides a powerful method for characterizing the binding interactions of macromolecules in free solution (34). It is an equilibrium method and we have previously used it to characterize the interaction of aggrecan with HA (35). It thus provided an ideal experimental system to analyze the interaction of the TSG-6 Link module with polymeric HA.

EXPERIMENTAL PROCEDURES
Materials-Unless otherwise stated, all reagents, chemicals, and consumables were obtained from Sigma. HA (800 kDa) extracted from rooster comb was a kind gift from Seikagaku Corporation Inc. (Tokyo, Japan).
Preparation of Hyaluronan Oligosaccharides-HA (800 kDa) was digested by bovine testicular hyaluronidase (0.01 unit/mg) in sodium acetate buffer (0.2 M, pH 5.0) for 24 h at 37°C and purified by size exclusion chromatography on a Bio-Rad Bio-Gel P-30 column in phosphate-buffered saline at 25°C. The eluate was monitored at 206 nm and the oligosaccharide peaks were pooled separately, dialyzed against water, freeze-dried, and then rerun on the same column, to yield purified fractions of HA oligosaccharides (4-to 14-mers) as determined by polyacrylamide gel electrophoresis (36).
Preparation of Link_TSG6 and G1-Aggrecan-Recombinant TSG-6 Link module (Link_TSG6) was expressed in Escherichia coli, refolded, and purified to homogeneity as described previously (37,38). The G1 domain of aggrecan (G1) (65 kDa) was prepared from porcine cartilage as previously described and migrated on SDS-PAGE as a single protein band (39).
Fluorescein Isothiocyanate (FITC) Labeling of Proteins-To label the Link_TSG6 and the G1 domain of aggrecan with FITC, solutions at 0.2 mg/ml in bicarbonate buffer (0.05 M, pH 9) (1 ml) were mixed with FITC (20 l of 1 mg/ml) (40) and incubated for 2 h at 4°C, before running on a 5-ml Hi-Trap (GE Healthcare) desalting column. Fractions (250 l) collected containing the FITC-labeled protein were quantitated by a bicinchoninic acid protein assay kit (Sigma). For both the G1 and Link_TSG6 the labeling was ϳ2-3 FITC mol/mol, based on the fluorescence of a known concentration of protein in solution and free FITC standard. This level of FITC labeling was sufficient to obtain a strong signal for confocal FRAP bleach recovery experiments at the protein concentrations used and over the pH range investigated.
Preparation of Solutions for Confocal FRAP-To investigate the binding of Link_TSG6 to HA (800 kDa) under different pH conditions, solutions containing mixtures of FITC-Link_TSG6 (31 g/ml) with HA (1600 g/ml HA) were prepared in either acetate or phosphate buffers such that all experiments were performed at the same overall ionic strength (i.e. [100 mM Na ϩ ]; 100 mM sodium acetate buffer was used for pH 4.0 -6.0, whereas 50 mM phosphate buffer was employed for pH 6.0 -8.0. The dissociation constant (K D ) for Link_TSG6 binding to HA at pH 6.0 was determined using FITC-Link_TSG6 (50 g/ml) and HA (2.5-500 g/ml) equilibrated in 44 mM NaH 2 PO 4 , 6 mM Na 2 HPO 4 , 44 mM NaCl, pH 6.0. For the determination of the K D at pH 7.4, FITC-Link_TSG6 (50 g/ml) and HA (10 -1500 g/ml) were equilibrated in 9 mM NaH 2 PO 4 , 41 mM Na 2 HPO 4 , 9 mM NaCl, pH 7.4. Competition experiments contained FITC-Link_TSG6 (40 g/ml) and 800-kDa HA (100 g/ml), together with varying concentrations of HA decasaccharide (HA 10 ) (0 -400 g/ml) in the pH 6.0 buffer as above. The minimum size of HA oligomer required for binding to Link_TSG6 was determined using FITC-Link_TSG6 (40 g/ml, 4 M) and HA (50 g/ml), together with HA oligosaccharides of different sizes (4-to 14-mer; all at 21 M final concentration) in the pH 6.0 buffer (as above).
Binding of G1 domain to HA was investigated at the same ionic strength (i.e. 100 mM Na ϩ ) as for Link_TSG6 above. The affinity of G1-HA binding was determined at different pH using mixtures of FITC-G1 (100 g/ml) and HA (800 kDa, 100 g/ml), prepared in citrate buffers (containing 100 mM Na ϩ ion) at pH values ranging from pH 3.5 to 8.0. To determine the K D of G1 binding to HA at pH 6.0, FITC-G1 (80 g/ml) with varying concentrations of HA (800 kDa; 50 -400 g/ml) were equilibrated in 44 mM NaH 2 PO 4 , 6 mM Na 2 HPO 4 , 44 mM NaCl. For the determination of K D at pH 7.4, FITC-G1 (80 g/ml) with HA (5-200 g/ml) was equilibrated in 9 mM NaH 2 PO 4 , 41 mM Na 2 HPO 4 , 9 mM NaCl. Competition experiments were with mixtures of FITC-G1 (80 g/ml) and HA (100 g/ml), together with HA 10 (0 -400 g/ml) in the pH 7.4 buffer. The minimum size of HA oligomer required for binding to the G1 domain was determined with FITC-G1 (80 g/ml 1.2 M) and HA (25 g/ml), together with HA oligosaccharides of different sizes (HA 4 -HA 14 ) (all at 21 M) in the pH 7.4 buffer (as above).
Diffusion Measurements by Confocal FRAP-The confocal FRAP instrumentation and variance recovery analysis were as described previously (34,35,41). Solutions for equilibrium measurements were pre-mixed and left overnight at 4°C. Samples for analysis (30 l) were sealed in cavity slides and further equilibrated for 30 min at 37°C on a heated microscope stage (PE-60, Linkam Instruments, Tadworth, UK). Sample bleaching of FITC and recovery fluorescence monitoring was with an argon ion laser (100 milliwatts) in a Bio-Rad MRC-1000 confocal microscope (Bio-Rad, Hemel Hempstead, UK). Lateral selfand tracer-diffusion coefficients (D) were calculated from the time dependence of plots of the second moment of the radially averaged distribution of bleached fluorophores (42). For all samples the fluorescence recovery showed the expected linear dependence of the second moment of the radial distribution of bleached fluorophore with time and the kinetics were consistent with first-order behavior of a single component. All experimental data are based on the results from at least 5 replicate bleaches and are reported as the mean Ϯ S.D.
Data Analysis of Binding Studies-Because of the large difference in the free diffusion of HA (800 kDa) and the FITC-Link_TSG6 (ϳ10 kDa) and FITC-G1 (ϳ65 kDa)), the binding of these proteins results in a large decrease in their apparent diffusion coefficients as their diffusion becomes the same as the HA. The translational diffusion coefficient determined by confocal FRAP is a long time-based analysis (over many seconds). Thus, if there is a fast exchange between bound and unbound forms (i.e. microseconds), then what is measured by confocal FRAP is an "average" diffusion coefficient of the bound and unbound fractions. The diffusion coefficient determined by the method thus provides a measure of the fractions of ligand that are bound and free (Equation 1).
Where D average is the measured diffusion, D free is the free diffusion coefficient of FITC-protein, D bound is the diffusion coefficient of HA, and F bound is the fraction of FITC-protein bound to HA. Values of the association (K A ) and dissociation (K D ) constants can then be calculated by measuring the fractions of FITC-protein bound at varying HA concentrations and using a Scatchard (Equation 2) or reverse Scatchard (Equation 3) analysis (35).
Where [HA] bs is the molarity of potential binding sites on HA, and n is the moles of FITC-protein bound per mol of binding site. Based on experimental data from this study, a 10-mer binding site was assumed for FITC-Link_TSG6, whereas a 12-mer binding site was assumed for FITC-G1.

RESULTS
Effect of pH on the Binding of the TSG-6 Link Module to HA-To investigate the binding of FITC-Link_TSG6 to polymeric HA in solution between pH 4.0 and 8.0 we used confocal FRAP to determine the translational diffusion coefficient. We had previously shown that the translational diffusion of HA was unchanged between pH 4.0 and 8.0 (41,43). The binding of FITC-Link_TSG6 to polymeric HA was thus easily detected, as the lateral translational free diffusion coefficient (D o ) of HA was ϳ10-fold less than that for FITC-Link_TSG6 and D 0 of FITC-Link_TSG6 (ϳ10 kDa) (1.25 ϫ 10 Ϫ6 cm 2 s Ϫ1 ) was similar at pH 6.0 and 7.4 and independent of concentration up to 100 g/ml (data not shown).
In the presence of HA (Fig. 1) the diffusion coefficient of FITC-Link_TSG6 was dramatically reduced and the pH profile suggested maximal binding to HA at pH 6.0. On the basis of the results at each pH, the fraction of FITC-Link_TSG6 bound to HA (Fig. 1) was calculated (Equation 1). The experiment was carried out with a large excess of HA and from the concentration of HA and FITC-Link_TSG6 the maximum binding was 1.6 Link_TSG6 molecules per HA chain (at pH 6.0). As the binding of Link_TSG6 to HA in this experiment would increase by only 2% the mass of the HA-FITC-Link_TSG6 complex, this was unlikely to have any significant effect on the diffusion coefficient of HA. With this assumption it was possible to estimate (using Equation 4) the K D for the binding of Link_TSG6 with HA at different pH values as shown in Equation 4.
. Effect of pH on the binding of Link_TSG6 and G1 to HA. Translational diffusion coefficients of FITC-Link_TSG6 (31 g/ml with 1600 g/ml 800-kDa HA) and FITC-G1 (100 g/ml with 100 g/ml 800-kDa HA) were determined by confocal FRAP between pH 3.5 and 8.0. The numerical values of the diffusion coefficients were used to calculate the proportion of FITClabeled proteins bound to HA, using Equation 1; data are shown as the mean of 5 replicate readings Ϯ S.D. The lateral diffusion coefficients of HA at 100 and 1600 g/ml were assumed to be 4.39 ϫ 10 Ϫ8 and 2.76 ϫ 10 Ϫ8 cm 2 s Ϫ1 , respectively (based on measurements with fluorescein-labeled HA of the same molecular weight at the corresponding concentrations), whereas that of free FITC-G1 and FITC-Link_TSG6 were determined to be 33 ϫ 10 Ϫ8 and 125 ϫ 10 Ϫ8 cm 2 s Ϫ1 , respectively.

pH Controls HA Binding to Aggrecan and TSG-6
Where F unbound is the fraction of FITC-Link_TSG6 that is unbound, and [HA] is the concentration of the binding sites on HA. Estimates of the K D at different pH showed there was maximal binding of Link_TSG6 to HA at pH 6.0 with much lower affinities both above and below this pH (Fig. 2). Isotherms for the Binding of Link_TSG6 to HA at pH 6.0 and 7.4-To obtain a more accurate determination of the K D of Link_TSG6 binding to HA, diffusion coefficients and the fraction bound to HA (F bound ) were determined at a range of HA concentrations (Figs. 3A and 4). At pH 6.0 and 7.4, the lateral diffusion coefficient of FITC-Link_TSG6 decreased as the concentration of HA increased and from these results the fraction bound was determined at each HA concentration (Fig. 4) and the results were analyzed by Scatchard analysis (Fig. 3, B and C). This showed a much stronger binding affinity of Link_TSG6 to HA at the pH 6.0 than at pH 7.4, with K D values of 3.0 ϫ 10 Ϫ7 and 3.0 ϫ 10 Ϫ5 M, respectively. The TSG-6 Link module thus had a 100-fold higher affinity for HA at pH 6.0 compared with pH 7.4 under the conditions tested in this solution phase analysis. In the Scatchard analysis it was assumed that Link_TSG6 binds to a HA 10 , because with this as the basis for representing HA concentration, the calculated stiochiometry of binding at both pH values was 1:1 (1.00 and 0.97 at pH 6.0 and 7.4, respectively). This was also guided by our previous studies on the stoichiometry of the Link_TSG6-HA interaction and the finding that HA 20 is the minimum length of oligomer that binds 2 Link_TSG6 molecules (27,33).
Competition of Link_TSG6 Binding to HA by HA Oligosaccharides-To determine the minimum length of HA oligosaccharide capable of binding a single Link_TSG6 molecule with high affinity, a range of HA oligomers (HA 4 -HA 14 ) were used to compete the interaction with the high molecular mass HA (800 kDa) at pH 6.0 (Fig. 5). The results showed that an HA 8 was the shortest to give strong competition, although there was a small increase with the longer oligosaccharides. This result was consistent with our previous studies using ITC (27). Because TSG-6 also binds to C4S (44), similar competition experiments were conducted with C4S oligosaccharides and indicated that C4S 10 was the minimum length that could compete strongly for HA binding (see supplemental Fig. S1).
To compare the affinity of Link_TSG6 binding to high molecular mass HA (polyvalent) and an HA oligosaccharide (monovalent), the concentration dependence of competition by HA 10 (ϳ2 kDa) for the interaction of FITC-Link_TSG6 with polymeric HA was determined at pH 6.0. This analysis (Fig. 6A) showed that the HA 10 competed equally with HA (800 kDa) when they were present at equimolar concentrations (with the HA polymer concentration calculated as 10-mer units).
Based on this result the high molecular weight HA and the HA 10 have similar binding affinities for Link_TSG6. In an equivalent analysis C4S 12 was also found to bind to Link_TSG6 with similar affinity to HA (data not shown).
Effect of pH on the Binding of the G1 Domain of Aggrecan with HA-Previous results using viscometry (7) suggested a large increase in the binding affinity of aggrecan to HA as the pH increased from 4.0 to pH 8.0. Similarly, platebased assays indicated that there was an increase in HA binding affin-  ity of the G1 domain between pH 4.0 and 6.0, with close to maximal binding between pH 6.0 and 8.0 (30). The binding of FITC-G1 to HA (800 kDa) was thus analyzed by confocal FRAP, over the pH range of 3.5 to 8.0. Initial measurements showed that the D o of FITC-G1 (ϳ75 kDa) was similar (3.3 ϫ 10 Ϫ7 cm 2 s Ϫ1 ) at both pH 3.5 and 8.0 and was independent of the concentration up to 100 g/ml (data not shown). In the presence of HA there was a large decrease in the diffusion of G1 over a broad pH range between pH 4.0 and 8.0, with the greatest effect seen at pH 7.0 -8.0 (Fig. 1). From these data the K D values were determined using Equation 4 for FITC-G1 binding to HA and showed that, in contrast to Link_TSG6, the affinity was highest between pH 7.0 and 8.0, with lower affinity at pH 4.0 -6.0 (Fig. 2).
Isotherms for the Binding of G1 to HA at pH 6.0 and 7.4-To obtain a more accurate determination of the K D of G1 binding to HA at pH 6.0 and 7.4, the fraction of FITC-G1 bound to HA (F bound ) was determined at a range of HA concentrations (Fig.  4). From Scatchard analysis (Equation 2), the K D values at pH  pH Controls HA Binding to Aggrecan and TSG-6 6.0 (Fig. 7A) and 7.4 (Fig. 7B) were calculated as 3.0 ϫ 10 Ϫ6 and 4.0 ϫ 10 Ϫ8 M, respectively. It was assumed that G1 binds to a HA 12-mer, because with this as the basis for representing HA concentration, the calculated stiochiometry of binding at both pH values was 1:1 (1.01 and 0.92 at pH 6.0 and 7.4, respectively). On the basis of this analysis, there was a 75-fold stronger binding at pH 7.4 compared with pH 6.0 for the interaction of G1 with HA. The affinity of G1 for HA thus increased between pH 6.0 and 7.4, whereas that of Link_TSG6 greatly decreased. Competition of G1 Binding to Polymer HA by HA Oligosaccharides-Competition of HA oligosaccharides (4-to 14-mer) on the binding of FITC-G1 to high molecular mass HA (800 kDa) was determined at pH 7.4. This analysis (Fig. 5) showed that a HA 10 was the minimum length of HA able to bind to the G1 domain with high affinity. In further competition experiments with a range of HA 10 concentrations the fraction of FITC-G1 bound to the polymer HA at each concentration was calculated, based on assumptions of either a 10-or a 12-mer binding site on polymeric HA. The results were in closest agreement to the curve predicted assuming a 12-mer binding site (Fig. 6B). Interestingly, in previous studies HA 12 was seen to be a slightly better competitor than HA 10 in binding to native aggrecan (45) or recombinant G1 (13), although HA 10 was the shortest HA oligosaccharide able to show strong binding to both. Therefore, it seems likely that HA 10 is the minimum length of HA able to occupy the HA-binding site on G1, but that HA 12 has slightly tighter binding. Earlier estimates of the footprint of aggrecan on HA suggested that at maximal packing density it occupied a longer stretch of HA (ϳHA 18 ) than was part of the minimal binding site (31) and more recent experiments comparing the binding of recombinant G1 to HA 32 and HA 40 also suggested a longer site occupied, which may result from negative cooperativity (13). Alternatively, this may be caused by the spatial constraints of how close G1 domains (which each have 2 Link modules and an Ig fold) can pack when bound to HA. Given the low protein:HA ratios in the present study (i.e. ϳ0.5 G1 molecules binding per HA polymer chain) the results would be independent of any affects of negative cooperativity or steric hindrance.

DISCUSSION
In this study confocal FRAP was used to investigate the interactions of the HA-binding domains from TSG-6 and aggrecan with polymeric HA. This solution-based approach, which measures equilibrium binding, avoids the criticisms of previous solid phase analysis (32). In particular, the results with aggrecan G1 domain on the pH dependence of binding (7,46) and the affinity at pH 7.4 (47,48) were in good agreement with those previously published and validated the technique with a well characterized system. The results also confirmed that the FITC labeling of G1 had no effect on its binding properties.
Analysis of the binding of the TSG-6 Link module to HA by confocal FRAP confirmed the unusual pH dependence of the interaction previously seen in plate assays (30,33) with maximal binding at pH 6.0 and a sharp decrease in the affinity above or below this pH (see Fig. 2). This pH dependence was thus strikingly different from that determined for the G1-HA interaction, where the affinity was low at pH 6.0 and was much higher between pH 7.0 and 8.0. More precise determinations of the K D values for the binding of Link_TSG6 to HA demonstrated a hundred-fold decrease in the affinity between pH 6.0 and 7.4 (K D 3.0 ϫ 10 Ϫ7 and 3.0 ϫ 10 Ϫ5 M, respectively), whereas there was a 75-fold increase in the affinity of the G1-HA interaction (K D 3.0 ϫ 10 Ϫ6 and 4.0 ϫ 10 Ϫ8 M, respectively). Thus, this study showed for the first time the dramatic effect of pH on the relative affinities of TSG-6 and aggrecan for polymeric HA in solution determined at close to physiological ionic strengths and at 37°C.
Interestingly, in the present study, which was performed with high molecular mass HA (800 kDa), there was a much larger (100-fold) change in affinity detected over the pH 6.0 -7.4 range than was previously determined for Link_TSG6 binding to short HA oligosaccharides at pH 6.0 and 7.5 in ITC experiments (33); in these experiments the corresponding difference in binding constant for HA 20 was ϳ7-fold and for HA 8 ϳ2.5-fold. Although, the ITC experiments were conducted under lower ionic strength conditions (5 mM Na ϩ ) than the confocal FRAP (100 mM Na ϩ ), NMR studies (27,29,49) suggest that neither the structure of Link_TSG6, nor the position of its HA interaction surface are significantly affected by ionic strength and both ITC (27) and confocal FRAP (Fig. 5) showed the same length dependence for HA oligosaccharide binding, indicating that the mode of interaction is similar at both low and high ionic strengths. However, the present results show that with HA in dilute solution, there is a much higher affinity of Link_TSG6 for polymeric HA at pH 6.0 than at pH 7.4. The physiological con-sequences of this large increase in affinity as the pH is lowered toward pH 6.0 will be to reduce TSG-6 mobility in tissue ECM, thereby increasing its local concentration and promoting the formation of TSG-6⅐HA complexes.
On the basis of our previous work (33), we hypothesized that some form of positive cooperativity between Link_TSG6 molecules bound on the same HA chain may contribute to the pH dependence. However, in the competition experiments analyzed in this study by confocal FRAP, the affinity of Link_TSG6 for HA 10 , which can only bind one Link_TSG6 molecule (27), was the same as its affinity for polymeric HA, which can bind many. There was, thus, no evidence for any cooperative mechanism of binding in the dilute solutions and at the low protein to HA stoichiometries used in this confocal FRAP analysis.
As noted in the Introduction we have recently identified a molecular mechanism that could explain the pH-dependent binding of HA to Link_TSG6 (33). This involves 4 key amino acid residues, His 4 , Tyr 12 , His 45 , and Asp 89 , where it is the loss of protonation on the His 45 and His 4 that leads to the increase in affinity between pH 3.5 and 6.0 and the decrease in binding constant between pH 6.0 and 8.0, respectively. These four amino acids are found invariantly in all TSG-6 species sequenced to date, indicating that the pH-dependent binding of HA is likely to be a conserved feature of this protein (33). To understand how this mechanism may relate to aggrecan binding to HA, it is important to consider that the G1 domain contains two Link modules arranged in tandem. The HA-binding site on G1 is reported to involve both Link modules (48) and is therefore unlikely to have an identical topography to the corresponding binding groove on Link_TSG6 (13). Of the residues important for mediating the pH dependence of HA binding in Link_TSG6, the equivalent amino acids are also present (and highly conserved) in the first Link module of G1. However, they are not conserved in the second Link module, with the exception of Asp 89 (29), which is, however, buried in the hydrophobic core of the domain and is highly conserved across the Link module superfamily as a whole. Thus, although the first Link module of aggrecan apparently has the "molecular machinery" needed for pH dependence, there are important differences compared with TSG-6. In particular, whereas His 4 and His 45 are solvent exposed in the TSG-6 Link module (27) they are likely to be much less solvent accessible in the G1 domain. Based on our modeling of the pair of Link modules in human G1 (essentially as described for cartilage link protein in Ref. 29), it is apparent that these amino acids form part of the buried interface between the two Link modules. These residues in the first Link module of G1 are therefore likely to have pK a values that differ significantly from those of His 4 and His 45 in Link_TSG6 (33). Thus, the lack of solvent accessibility for these histidine residues in the first Link module and the absence of histidines at equivalent sequence positions in the second Link module of the G1 domain may provide a structural basis to explain why the interactions of TSG-6 and aggrecan with HA have pH dependences that are so different. The results also imply that, although the HA-binding domains in these proteins are based on a common structural unit (i.e. the Link module), they have evolved to have key differences in function. For example, it may be important that aggrecan binds to HA at neutral pH to sup-port the formation of supramolecular aggregates in cartilage, whereas for TSG-6, the high affinity for HA at pH 6.0 and low affinity at pH 7.4 are likely to have evolved as a necessary part of its physiological role. The two proteins have also evolved to have other distinct properties, e.g. Link_TSG6 binds to a wide range of glycosaminoglycans and proteins that do not interact with the G1 domain of aggrecan (e.g. C4S, heparin, bikunin, pentraxin-3, thrombospondin-1 (17,25). Although the effect of pH on these interactions has been less well studied, increased tissue acidity in inflammation, could promote through binding to polymeric HA, the generation of multiple arrays of TSG-6, which could serve to modulate its ligand-binding functions.
The present results showed that the G1 domain of aggrecan and Link_TSG6 have very different pH dependences of HA binding between pH 6.0 and 7.4, which is over a pH range highly relevant to tissue physiology and pathology. At pH 7.4 the affinity for the interaction of G1 with HA is 750-fold higher than that for Link_TSG6, whereas at pH 6.0 the TSG-6 Link module has an affinity that is 10-fold higher than the aggrecan G1 domain (see Fig. 2). In particular, as the pH is reduced from 7.4 to pH 6.5 the relative increase in the HA binding affinity for Link_TSG6 compared with G1 is about 35-fold (ϳ4-fold increase and ϳ9-fold decrease in affinity respectively), whereas a decrease in pH from 6.5 to 6.0 results in an even larger (ϳ340fold) relative increase in HA binding. Thus it is apparent that small changes in pH could have a significant effect. In this regard, the presence of pH gradients in healthy tissues and the more acidic tissue pH that is associated with inflammation, could serve to differently regulate the functions of TSG-6 and aggrecan (30,33). For instance, in cartilage the production of lactic acid by anaerobic glycolysis in chondrocytes, which is enhanced in inflammation, combined with the Gibbs-Donnan equilibrium, could lead to pH values in the tissues that are as much as 1 pH unit lower than the equilibrating synovial fluid (51), where the pH of this fluid may be as low as pH 6.8 in inflammatory arthritis (52). Therefore, it is conceivable that TSG-6 and aggrecan could be present in joint tissues across a wide range of pH values from pH 6.0 to 7.4, where there is 7500-fold overall relative difference in their affinities for HA. In tissue microenvironments in cartilage with "high" pH it would be expected that aggrecan-HA complexes would predominate, whereas at "lower" pH values there would be a much greater tendency for TSG-6⅐HA complexes to form. Although TSG-6 may not actively displace aggrecan from link protein-stabilized complexes with HA, which are not freely reversible (7,31), it may dominate the interaction with newly synthesized HA in the pericellular environment. In the context of cartilage, where TSG-6 is expressed by chondrocytes in response to inflammatory mediators (17), the preferential binding of TSG-6 to HA at sites of "low" pH (e.g. in pericellular regions), and its competition with the aggrecan-HA interaction (30), could facilitate the diffusion of de novo synthesized aggrecan out to the inter-territorial matrix (where the pH is likely to be higher) and thus promote the production/maintenance of inter-territorial aggrecan/HA/link protein aggregates. This could potentially contribute to the chondroprotective activity of TSG-6 seen in animal models of arthritis (24). Although further work is needed to test this hypothesis, the results presented here are consistent with a potential role for TSG-6 in matrix remodeling. In this regard, the local pH could dictate what kind of ECM networks are produced at a given tissue site. For example, we have recently shown that TSG-6 can interact with fibronectin via its CUB_C domain and thereby enhance the binding of fibronectin to thombospondin-1, a ligand that interacts with the Link module domain of TSG-6 (50). It is possible therefore that TSG-6 could bridge between fibronectin and other of its Link module ligands such as HA, effectively linking together fibronectin-based and HA-based ECM networks.
In this study we have demonstrated that the HA binding activities of both TSG-6 and aggrecan are quite differently dependent on pH and the results indicate that pH may play a crucial role in the regulation of HA-protein complexes and determine the type of ECM that is formed in particular tissue microenvironments. Furthermore, we have identified that pH is likely to be a major regulator of TSG-6 and its functions are likely to be exquisitely sensitive to the pH gradients found at sites of inflammation.