SHAP Potentiates the CD44-mediated Leukocyte Adhesion to the Hyaluronan Substratum*

CD44-hyaluronan (HA) interaction is involved in diverse physiological and pathological processes. Regulation of interacting avidity is well studied on CD44 but rarely on HA. We discovered a unique covalent modification of HA with a protein, SHAP, that corresponds to the heavy chains of inter-α-trypsin inhibitor family molecules circulating in blood. Formation of the SHAP·HA complex is often associated with inflammation, a well known process involving the CD44-HA interaction. We therefore examined the effect of SHAP on the CD44-HA interaction-mediated lymphocyte adhesion. Under both static and flowing conditions, Hut78 cells (CD44-positive) and CD44-transfected Jurkat cells (originally CD44-negative) adhered preferentially to the immobilized SHAP·HA complex than to HA. The enhanced adhesion is exclusively mediated by the CD44-HA interaction, because it was inhibited by HA, but not IαI, and was completely abolished by pretreating the cells with anti-CD44 antibodies. SHAP appears to potentiate the interaction by increasing the avidity of HA to CD44 and altering their distribution on cell surfaces. Large amounts of the SHAP·HA complex accumulate in the hyperplastic synovium of rheumatoid arthritis patients. Leukocytes infiltrated to the synovium were strongly positive for HA, SHAP, and CD44 on their surfaces, suggesting a role for the adhesion-enhancing effect of SHAP in pathogenesis.

Leukocyte cell-surface adhesion molecules play important roles in hematopoiesis, trafficking, recruitment, and the activation of leukocytes. Efforts toward identifying the adhesion molecules mediating the leukocyte-endothelial interaction have led to two families of adhesion molecules: the glycan-recognizing selectin family, which participates in the primary adhesion and rolling of leukocytes on the endothelium, and the integrin family members, which recognize the immunoglobulin superfamily molecules that are involved in the subsequent firm adhesion. The participation of these adhesion molecules has been well documented in leukocyte extravasation and function in diverse systems. Nevertheless, it is certain that there are more adhesion molecules involved in the regulation of these processes.
A promising candidate is the CD44-hyaluronan (HA) 2 interaction that has attracted much attention since the late 1980s. CD44, first cloned as a lymphocyte surface antigen recognized by lymphocyte homing-interfering monoclonal antibodies (1), is a single polypeptide type I transmembrane protein encoded by a single gene. The standard form of CD44 consists of an N-terminal HA-binding domain, a stem region, a transmembrane domain, and a cytoplasmic domain. There are many CD44 variant forms with insertions at the stem region resulting from the alternative splicing of the exon 6 -15 (or exon V1-V10) (2,3). CD44 shares a highly conserved HA-binding domain, the link module, with HA-binding large proteoglycans (aggrecan, PG-M/versican, brevican, and neurocan), the link protein, and tumor necrosis factor-stimulated gene 6 product (TSG6) (1,4). CD44 is also expressed in many kinds of cells other than leukocytes. Although CD44 interacts with a number of cell-surface or extracellular molecules, such as fibronectin, collagen, selectin, osteopontin, and chondroitin sulfate proteoglycans, its principal ligand is believed to be HA (3). The roles of CD44-HA interaction in lymphocyte extravasation (5,6) and stromal interaction (7,8) have become evident not only in cell adhesion systems in vitro with cell monolayer or tissue sections, but also in systems in vivo using specific antibodies or CD44null mice. These studies implicated that the CD44-HA interaction may underlie the pathogenesis of a wide range of disorders such as the autoimmune diseases, including arthritis (9 -11), encephalomyelitis (12), crescentic glomerulonephritis (13), and uveoretinitis (14); allergic diseases, including asthma (7) and contact dermatitis (15); inflammatory bowel disease (8,16); graft rejection (17); atherosclerosis (18); hepatitis (19); and lung inflammation (20). Depending on the diseases, the consequences of the interaction may either be inflammatory, for example in the autoimmune diseases, where its role in the recruitment and activation of circulating leukocytes is predominant, or anti-inflammatory, e.g. in hepatitis and lung inflamma-* This work was supported in part by a grant-in-aid for Scientific Research (B) from Japan Society for the Promotion of Science, by a grant from the CREST of Japan Science & Technology Agency, and by a special research fund from Seikagaku Corp. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1 To whom correspondence should be addressed. tion, where its principal role seems to be the induction of apoptosis of leukocytes. Given the diversity of the targeting tissue and effects of CD44-HA interaction, it is not surprising that the interaction is under the control of extremely complex mechanisms. It is well known that most CD44-bearing leukocytes are actually not adhesive to HA until they are activated. Extensive studies on the activation-associated molecular modification of CD44 have revealed that it is multifactorial, involving transcriptional upregulation (21), the expression of new variant forms (22), glycosylation (N-glycan, O-glycan, and glycosaminoglycan) (23,24), sialylation (25), sulfation (26), phosphorylation (27), palmitoylation (28), and dimerization/oligomerization (29,30). The regulation is strongly event-and cell type-dependent. Even the same kind of modification may lead to opposite outcomes under different circumstances; for example, the blockage of N-glycosylation with tunicamycin or by amino acid residue substitution enhances the CD44-HA interaction in pre-B cells, fibroblasts, Chinese hamster ovary cells, and melanoma cells (31)(32)(33) but reduces it in lymphoma cells, ovarian tumor cells, and melanoma cells (34,35).
In sharp contrast to CD44, little information is available about the regulation of CD44-HA interaction on HA side. The major reason is due to the structural simplicity of HA: the linearly repeating N-acetylglucosamine-glucuronic acid disaccharide units. In the absence of a core protein modification, sulfation, and epimerization, the chain length appears to be the sole molecular variation of HA. In recent years, there are accumulating data indicating that the degradation of HA macromolecules occurs during some inflammatory disorders and that the resulting low molecular weight HA fragments exhibit different biological activities from their precursors (36,37). The size-dependent avidity of HA to cell-surface CD44 has been examined using HA oligomers. Monovalent binding was observed in 6-to 18-sugar-oligomers, and divalent binding with greatly and progressively increased avidity in 20-to 38-sugar-oligomers (38). The relationship of such a mechanism to the pathological role of naturally occurring HA fragments, which are usually longer than the oligomers examined, remains to be further examined. Additionally, TSG6 has been recently shown to modulate the CD44-HA interaction, thus providing a new insight into the functional regulation of CD44-HA interaction on HA side (39).
We have found that HA, ever thought to be a free molecule, could also be covalently modified by a protein. The protein, designated SHAP (serum-derived HA-associated protein), is bound to HA via a unique ester bond between the carboxyl of its C-terminal aspartate and the C-6 hydroxyl group of an internal N-acetylglucosamine of HA (40,41). SHAP is identical to the heavy chains of inter-␣ (trypsin) inhibitor (I␣I) family molecules (42), which link via the same type of ester bond to the chondroitin sulfate chain of the I␣I light chain, the bikunin proteoglycan (43). The formation of the SHAP⅐HA complex is believed to be a transesterification reaction, where HA substitutes for chondroitin sulfate to form an ester bond with HC/SHAP. I␣I family molecules, principally synthesized and secreted by the liver, occur constitutively at high concentrations in the circulation. The SHAP⅐HA complex is most likely formed between the extravasated I␣I and the local HA, as best exemplified by rheumatoid arthritis (RA) and ovulation process (44,45). The reaction is indispensable for the construction of HA-rich extracellular matrix of the expanding cumulus oophorus, oocyte transportation, and successful oocyte-sperm interaction (fertilization) (45). In addition, the complex is present in large amounts in the synovial fluid of RA patients. Patients with various inflammatory disorders often have an elevated plasma level with the complex (44,46). In this study, we examined the effect of SHAP on CD44-HA interaction and found that SHAP significantly increase the avidity of HA to the cell-surface CD44.
Purification of the SHAP⅐HA Complex from RA Synovial Fluid-The purification and characterization of the synovial SHAP⅐HA complex has been described in detail elsewhere (47). Briefly, the synovial fluid was subjected to sequential cesium chloride isopycnic centrifugation in the presence of 4 M guanidine hydrochloride to separate hyaluronan from the non-covalently associated proteins. The final highly purified prepara-tion contains HA molecules with an average molecular weight of 1.8 million, about half of which bears SHAP at an average ratio of three to five SHAPs per HA chain. To cleave the ester linkage between SHAP and HA, the SHAP⅐HA complex solution was supplemented with NaOH to reach a final concentration of 0.1 M and incubated at room temperature for 30 min, followed by neutralization with HCl solution. In control tubes, NaOH solution and HCl solution were mixed before adding to the SHAP⅐HA complex solution.
Purification of the SHAP⅐HA Complex from RA Hyperplastic Synovium-The hyperplastic synovium was obtained from RA patients receiving synovectomy. Synovium (6 g of wet weight) was homogenized with a glass grinder in 20 ml of phosphatebuffered saline (PBS) buffer containing 5 mM deferoxamine mesylate, 10 mM EDTA, and 6% protease inhibitor cocktail. The homogenates was mixed with an equal volume of 8 M guanidine hydrochloride and then stirred overnight at 4°C. After removing the insoluble materials by centrifuging at 10,000 rpm for 30 min, the density of the solution was adjust to 1.38 g/ml with solid cesium chloride. The solution was then subjected to isopycnic centrifugation as described above. The resulting density gradient was partitioned into 20 fractions, which were then precipitated twice with 3 volume of 95% ethanol containing 1.3% potassium acetate and finally dissolved in 0.4 ml of MilliQ water. The protein and uronate contents in each fraction were measured by using MicroBCA kit and carbazole reaction, respectively.
Purification of I␣I-Mouse I␣I was purified from pooled serum from C57BL/6 mice as described before (45). The highly purified human I␣I was a gift from the Chemo-Sero-Therapeutic Research Institute, Kumamoto, Japan.
Enzyme-linked Immunosorbent Assay-The SHAP⅐HA complex and HA were quantified by enzyme-linked immunosorbent assay as previously described (44,47). Briefly, the SHAP⅐HA complex or HA in the sample solutions were captured on HABP (2 g/ml in 0.1 M NaHCO 3 , pH ϳ9.5)-coated wells of MaxiSorp plates, and then detected with rabbit anti-I␣I antibody and then HRP-conjugated goat anti-rabbit IgG antibody, or biotinylated HABP and then HRP-conjugated streptavidin, respectively. The color reaction was developed with 3,3Ј,5,5Ј-tetramethylbenzidine substrate and stopped with 1 N HCl. The absorbance value was read at 450 nm.
Cell Adhesion Assay-The cells were propagated in RPMI 1640 medium supplemented with 10% fetal calf serum, 10 g/ml streptomycin, and 6.25 g/ml penicillin G. The floating cells were collected by centrifugation, washed with PBS for three times, counted, and then resuspended at proper densities. For the adhesion assay under static conditions, plastic dishes (Falcon, #1008) or cover glasses were coated with HABP by incubating with 10-l drops of HABP solution (4 g/ml in 0.1 M NaHCO 3 , pH 9.5) overnight at 4°C. After removing the HABP solutions followed by washing with PBS, 10 l of the SHAP⅐HA or HA solution at indicated concentrations were added to the HABP-coated spots and incubated at room temperature for 1 h to immobilize the SHAP⅐HA complex or HA. After removing the drops with pipettes, the spots were washed extensively with PBS, overlaid with cell suspensions (1 ϫ 10 6 cells/ml), and then incubated at room temperature for 30 min. For inhibition experiments, the cells were resuspended in PBS containing HA or I␣I at indicated concentrations instead of PBS only. For antibody blocking experiment, the cells were preincubated with 5 g/ml antibodies on ice for 30 min followed by washing with PBS. After washing with PBS to remove the unadhered cells, the central areas of the spots were photographed with an Olympus charge-coupled device digital camera mounted on an Olympus inverted microscope (100ϫ magnification) to count the number of adhered cells. The strength of cell adhesion was represented by the density of adhered cells (number of cells per unit area).
For the adhesion assay under flow shearing conditions, microscope slides were coated with HABP and then SHAP⅐HA complex (20 or 100 g/ml) or HA (250 or 1000 g/ml) as mentioned above and assembled in the parallel flow chamber designed by Lawrence et al. (48). The SHAP⅐HA or HA-coated microscope slides were attached onto a transparent plastic plate with a silicon rubber gasket in parallel so that between the two surfaces a chamber with a thickness of 320 m was formed, to which the inlet and outlet are connected. Rolling assays with the chamber were performed as described previously (49). Briefly the chamber was mounted on the stage of an inverted phase-contrast microscope (Model IX70, Olympus Products, Tokyo). Hut78 cells were introduced into the chamber at a concentration of 5 ϫ 10 5 /ml at decreasing levels of fluid shear (1.13, 0.85, 0.57, 0.42, and 0.28 dyne/cm 2 ). The shear stress in the chamber was controlled using a syringe pump (Model 944, Harvard Apparatus, South Natick, MA). The numbers of rolling or attached cells were determined from digital images recorded with a Model CS-220 video camera system (Olympus). The flow chambers were maintained at 37°C with an air-curtain incubator (Olympus Products, Model IX-IBM) throughout the experiments. The wall shear stress (T) was calculated from the equation, T ϭ 3Q/2a 2 b, where ϭ coefficient of viscosity (1.0 centipoise), Q ϭ volumetric flow rate (cm 3 /s), 2a ϭ channel height (3.2 ϫ 10 Ϫ2 cm), and b ϭ channel width (1.3 cm).
Immunolabeling and Fluorescence Microscopy-Hut78 cell suspension was laid on the SHAP⅐HA or HA-coated cover glass (prepared as mentioned above), followed by incubating at room temperature for 20 min or 1 h. After washing, the adhered cells were fixed with 4% buffered neutral formaldehyde. The cover glass was blocked with 10% goat serum (Nichirei, Tokyo, Japan), incubated with PBS-T (0.1% Tween 20 in PBS) containing Ab4 antibody (1 g/ml) and biotinylated HABP (0.5 g/ml), and then with PBS-T containing Alexa 594-rabbit anti-mouse IgG (1:1,000 dilution) and Alexa 488-streptavidin (1:1,000 dilution) at room temperature for 1 h each. Finally, the cover glass was washed with PBS-T for 5 times, and then mounted on a glass slide with Gelmount (CosmoBio, Tokyo, Japan). The cells were examined and photographed with a Zeiss LSM5 PASCAL confocal laser microscope.
RNA Extraction and Reverse Transcription-PCR-Total RNA was prepared from cells using TRIzol reagent (Invitrogen) and quantified by measuring the UV absorbance. The first strand cDNAs were synthesized from 4 g of total RNA using Superscript II RNase H Ϫ reverse transcriptase and oligodT [12][13][14][15][16][17][18] primer. The CD44 transcripts were amplified with AmpliTaq DNA polymerase and the primer set CTGGCGCAGATC-GATTTGAA (forward, corresponding to the boundary of exon 1 and 2) and GGCGTACGTATAGGACCAGAGGTTGT (reverse, corresponding to exon 17, with a SplI site). The cycling conditions were: 95°C for 3 min, 30 cycles of 95°C for 30 s, 56°C for 45 s, and 72°C for 90 s, and finally 72°C for 5 min.
Cloning of CD44 Variants and Cell Transfection-The fulllength cDNA of human standard form CD44 (sCD44 or CD44H), including ϳ50 bp of the 5Ј-untranslated region as well as 15 bp of the 3Ј-untranslated region, was cloned into the EcoRI site of pCDNA3.1/Myc-His(Ϫ) B vector (Invitrogen). The stop codon TAA was mutated to GAA, which, together with the 15-bp 3Ј-untranslated region, generates an extra peptide sequence EHLHHY between the CD44 polypeptides and the vector-derived C-terminal tags. The above sCD44 construct was used to generate the CD44 variant form (CD44v) constructs by inserting proper variable exons (exon 6 to exon 15, or exon V1 to V10) between exons 5 and 16. The HpaI site (GTTAAC) present in exon 5 and a SplI site generated by silent nucleotide mutation (from AGGACA to CGTACG) in exon 17 were selected for cloning, to which PCR-amplified CD44v transcripts were inserted. All CD44 constructs were verified by sequencing with ABI 310 sequencer (PerkinElmer Life Sciences). For transfection of Jurkat cells with Bio-Rad Gene Pul-serII, 11 g of PvuI-linearized CD44 constructs were mixed with 0.75 ml of cell suspension in PBS (1 ϫ 10 7 cells) in a 0.4-cm cuvette and incubated on ice for 5 min both before and after the electroporation. The voltage strength was 1 kV/cm, the capacitator was 960 microfarads, and the pulse lengths were between 12 and 15 ms. The cells were then cultured in RPMI 1640 medium supplemented with 10% fetal calf serum, 10 g/ml streptomycin, and 6.25 g/ml penicillin G and subjected to G418 selection (0.5 mg/ml) 24 h later. After 2 weeks, the cell pools were labeled with Ab4 antibody (1 g/ml) followed by fluorescein isothiocyanatesheep anti-mouse IgG antibody, and subjected to flow cytometry sorting with FACSVantage S.E. (BD Biosciences) to isolate the CD44-positive cells. To induce the HA avidity, the CD44-transfected Jurkat cells were cultured in the presence of 100 ng/ml phorbol 12-myristate 13-acetate overnight (30).
Generation of Recombinant Human TSG6 Protein-Human TSG6 cDNA was amplified from the total RNA preparation of human hepatoma cell line, HLF, with the primers TCGCTAGCTCTGAGAATTTGTGAGC and CAAGATCT-TAAGTGGCTAAATCTTCCAG and a reverse transcription-PCR protocol as described for CD44 transcript. The resulting 0.9kbp fragment was digested with NheI and BglII, and then cloned into the XbaI and BamHI sites of p3ϫFLAG-CMV-14 vector. The sequence-verified construct, which encodes a fusion protein of human TSG6 and a 3ϫFLAG tag at C terminus, was introduced into HLF cells using Fugene 6 transfection reagent. Stable transfectants obtained by G418 selection were cultured in a serum-free medium, Cosmedium 001. The recombinant TSG6 protein was then purified from the pooled conditioned medium with the anti-FLAG M2 affinity gel by following the manufacturer's instruction.
Western Blot-Samples were mixed with the loading buffer (1% SDS, 50 mM Tris-HCl, pH 6.8, 4% 2-mercaptoethanol, and 20% glycerol), followed by boiling for 5 min. After electrophoresis, the proteins on the gel were electrotransblotted to a Hybond extra C membrane. The membrane was blocked with PBS-T containing 10% skim milk at 37°C for 2 h. For SHAP immunostaining, the membrane was incubated successively with rabbit anti-human I␣I antibody (1:2000 in PBS-T) and HRP-conjugated goat anti rabbit immunoglobulin antibody (1:3000 in PBS-T). For TSG6 immunostaining, the membrane was incubated successively with goat anti-human TSG6 antibody (0.4 g/ml in PBS-T containing 1% skim milk) and HRP-conjugated rabbit anti goat immunoglobulin antibody (1:2000 in PBS-T containing 1% skim milk). Finally, the immune complexes were visualized using enhanced chemiluminescence reagent and exposed to Hyperfilm ECL.
Immunohistological Study-The synovium samples were obtained surgically from patients having a diagnosis of typical RA at Steinbrocker's stage III-IV and receiving a synovectomy at Nagoya University Hospital with informed consent. The tissues were fixed and embedded in paraffin to prepare 4-m sections. The sections were subjected to immunostaining with Ab4, anti-I␣I antibody, anti-UTI antibody, and biotinylated HABP, and then the respective HRP-conjugated second antibodies (Dako, Carpinteria, CA). Finally, the Liquid DAB Substrate-chromogen system was used to develop a color (Dako).

RESULTS
CD44-positive Cells Adhere More Strongly to the SHAP⅐HA Complex than to HA-We isolated the SHAP⅐HA complex from synovial fluids of RA patients by sequential isopycnic centrifugation. The final preparation was a mixture of the SHAP⅐HA complex and the free HA chains (the ratio of protein to HA in weight was 0.1) but was free of other proteins and proteoglycan contaminants (47). For the cell adhesion assay, the SHAP⅐HA complex thus obtained or SHAP-free HA was immobilized on the HABP-coated plastic dish or cover glass. Hut78 cells, a human cutaneous T cell lymphoma cell line bearing the activated form of CD44, and Jurkat, a CD44-negative acute T cell leukemia cell line, were used in this study. As expected, Hut78 cells, but not Jurkat cells (data not shown), adhered to the SHAP⅐HA substratum (both plastic and glass surfaces) (Fig. 1). No cell adhesion to the surface coated with only HABP was observed (Figs. 1 and 2A). In addition, digestion of the SHAP⅐HA complex with Streptomyces hyaluronidase before immobilization abrogated the adhesion of Hut78 cells (Fig. 1). These observations confirmed that the cell adhesion was mediated by CD44-HA interaction.
Interestingly, an alkali pretreatment that releases the SHAP from HA (47) also abrogated the adhesion of Hut78 cells (Fig.  1), suggesting a significant involvement of SHAP in mediating the cell adhesion. We thus compared in detail the adhesion strength of Hut78 cells between the substrata of SHAP⅐HA complex and free HA. As shown in Fig. 2, the Hut78 cells exhibited a dose-dependent adhesion to both the SHAP⅐HA complex and HA substrata. A concentration of immobilizing solution, ϳ20 -30 times higher, was required for HA to trap as many cells as that on the SHAP⅐HA substratum (Fig. 2B), although the amounts of immobilized HA were comparable between both substrata (Fig. 2C). The results clearly demonstrate an enhancing effect of SHAP on HA-mediated cell adhesion. Because the SHAP⅐HA complex at a concentration of 20 -50 g/ml was enough to adhere a large number of cells, the condition was adopted in the following experiments.
In addition to the static adhesion assay, we also performed the adhesion assay under flow shearing conditions (Fig. 3). Similarly, the cells showed a stronger avidity to the SHAP⅐HA complex than to HA, as manifested by the significantly increased number of cells either firmly adhering to or rolling on the SHAP⅐HA substratum. Intriguingly, a considerable population of cells exhibited a very low rolling speed, even stopping, on the SHAP⅐HA complex, in sharp contrast to those mediated by the selectin-ligand interaction that we described previously (49).
Although there were firm adherent cells at 1.13 dyne/cm 2 to the SHAP⅐HA substratum with a coating concentration of 20 g/ml, those to HA substratum with a coating concentration of 1 mg/ml were barely observed even at 0.28 dyne/cm 2 . The maximum cell rolling was observed at 0.42 dyne/cm 2 or below on both the SHAP⅐HA complex and HA substrata, with 4.9 ϫ 10 3 cells/mm 2 /min on the SHAP⅐HA substratum (coating concentration at 100 g/ml) and 1.3 ϫ 10 3 cells/mm 2 /min on HA substratum (coating concentration at 1 mg/ml) in average.

CD44-HA Interaction Exclusively Mediates the Strong Adhesion of Hut78
Cells to the SHAP⅐HA Complex-We next tried to define the mechanism for the enhancing effect of SHAP. First we examined the effect of the presence of SHAP on the immobilization of HA on the HABP-coated surface using an enzymelinked immunosorbent assay system. The results shown in Fig.  2C indicate that the immobilization of the SHAP⅐HA complex, HA dissociated from the SHAP⅐HA complex by alkali treatment, and free HA on the HABP-coated surface are comparable. Therefore, we concluded that the enhancing effect of SHAP on cell adhesion is not a result of the increased immobilizing efficiency of HA.
We then examined the possibility of a direct interaction between SHAP and cell-surface molecules, including CD44 that shares a conserved link module domain with HABP (1). Hyaluronan, but not purified I␣I, significantly inhibited the adhesion of Hut78 cells to the SHAP⅐HA complex, suggesting that SHAP did not mediate the adhesion of cells directly (Fig. 4). Consistently, Hut78 cells did not adhere to I␣I-coated surface, even with a high coating concentration (0.2 mg/ml, data not shown). On the other hand, pretreatment of cells with anti-CD44 monoclonal antibody (Ab-4 or IM7) completely abrogated the cell adhesion, whereas neither the unrelated anti-I␣I nor another anti-CD44 monoclonal antibody, KM81, exhibited any inhibitory effect (Fig. 5). Flow cytometry analysis showed that KM81 did not recognize human CD44 on the surface of Hut78 cell (Fig. 5). These data provide convincing evidence for the exclusive role of CD44-HA interaction in mediating the strong adhesion of Hut78 cells, implying an indirect role of SHAP in enhancing the cell adhesion, i.e. by increasing the HA avidity to CD44.
To further validate the conclusion, we transfected the non-adhesive Jurkat cells with CD44 cDNAs and examined their adhesion to the SHAP⅐HA complex. Reverse transcription-PCR study revealed transcripts with various lengths (Fig. 6), indicating the expression of both sCD44 and CD44v in Hut78 cells. Cloning and sequencing of the PCR products indicated that the major bands were derived from CD44H, and CD44E, the epithelial form containing variable exons V8 -10, whereas the minor one was a mixture of PCR products derived from two variant forms containing V8 -9 and V10, respectively. All these CD44 variant forms have been reported (3). All four CD44 transcripts were cloned and introduced into Jurkat cells. The CD44-transfected Jurkat cells exhibited no adhesion to HA substratum (data not shown) until they were activated by culturing in the presence of phorbol 12-myristate 13-acetate, which was reported to induce the dimerization of CD44 and thus increase the avidity of CD44 to HA (30). Again, the activated Jurkat CD44 transfectants exhibited stronger adhesion to the SHAP⅐HA complex than to HA (Fig. 7). The results provide further evidence for the above conclusion.
Since it has been reported that the RA synovial fluid contains TSG6⅐I␣I complex, and that TSG6 modulates the CD44-HA interaction (39), it is important to examine whether the adhesion-enhancing effect of the purified SHAP⅐HA complex is due to a contaminating TSG6. We, therefore, generated the recombinant human TSG6 protein in cultured human hepatoma cells. The recombinant protein was efficiently purified via its C-terminal 3ϫFLAG tag, as demonstrated by the single band in a sliverstained SDS-PAGE gel (Fig. 8). An anti-human TSG6 polyclonal antibody, which was able to detect nanogram amounts of the recombinant TSG6 protein, failed to show any TSG6 immunoreactivity at 38 kDa in the SHAP⅐HA preparation containing 1 g of protein (Fig. 8). The faint bands with higher molecular weights in lane 4 are background reaction of anti-TSG6 antibody with SHAP proteins, because TSG⅐I␣I complex is known to dissociate after an alkali treatment. The results rule out the possibility of a contaminating activity of TSG6.
Immunofluorescence Study on the Cell Adhesion-mediating CD44 and HA Molecules-We further investigated the molecular interactions upon cell adhesion using immunofluores-  cence microscopy. Hut78 cells attaching to the substratum after 20-min incubation were strongly positive for CD44 staining on their surfaces (Fig. 9, A and B, the flank surface and bottom of the cells, respectively). No apparent difference in the distribution of CD44 was found on the bottom surfaces between the cells adhering to the SHAP⅐HA substratum and to the HA substratum, but more diffused distributions were observed on the flank cell surfaces. Although both SHAP⅐HA and HA substrata exhibited an even distribution of fluorescent signals of HA staining, stronger signals were found on the surfaces of adhered cells, indicating that HA molecules condensed on the cell surface upon CD44-mediated cell adhesion. The substratum origin of HA was also supported by the negative result in staining cell-surface HA by flow cytometry analysis (data not shown). Interestingly, the condensation was more fre-quently found on the SHAP⅐HA substratum than on the HA substratum, especially on the flank surfaces (Fig. 9A), suggesting that the enhancing effect of SHAP may be related to the accelerated condensation of HA on the surfaces of CD44-bearing cells. The cells incubated for 1 h exhibited similar patterns, which excludes the possibility that the difference was due to some metabolic reactions, such as the uptake and degradation of HA.
Formation of the SHAP⅐HA Complex in the RA Hyperplastic Synovium-We further examined the formation and distribution of the SHAP⅐HA complex in the inflammatory hyperplastic synovium of RA patients. Glycosaminoglycans were extracted from hyperplastic synovium with 4 M guanidine hydrochloride and separated from proteins by cesium chloride isopycnic centrifugation (Fig. 10A). Western blot assay demonstrated the release of SHAP from the synovium membrane-derived glyco- saminoglycan by alkali treatment and Streptomyces hyaluronidase digestion, which specifically degrades hyaluronan (Fig.  10B). The larger molecular weights of hyaluronidase digestionderived SHAPs are attributed to HA remnants due to the steric hindrance of SHAP proteins and probably alkali treatment-induced loss of O-glycan as well. The results indicated clearly the formation of the SHAP⅐HA complex in the hyperplastic synovium of RA patients. Histological study showed strong staining of hyaluronan by HABP (Fig. 11), consistent with the previous findings that the hyperplastic synovium extensively produces hyaluronan. Anti-I␣I antibody also gave strong staining results, whereas anti-bikunin antibody gave negative results, suggesting that the immunoreactivity corresponds to the SHAP⅐HA complex rather than I␣I. The distribution of SHAP is patchier than the smearing pattern of HA. Interestingly, the CD44-positive lymphocytes in the tissues were strongly positive for both SHAP and HA, implying an interaction of these cells with the SHAP⅐HA complex in hyperplastic synovium.

DISCUSSION
We showed in this study that the covalent association of SHAP with HA greatly enhanced its avidity to the cell-surface CD44 and thus provided a new insight into the complex regulatory mechanism for the CD44-HA interaction-mediated cell adhesion; that is, the regulation can also occur on the HA side. Consistently, Lesley et al. (39) showed that TSG6 modulates the interaction between HA and CD44 on the cell surface. How-ever, there are some differences in the regulation pattern between SHAP and TSG6. For example, TSG6⅐HA complex is capable of mediating adhesion of cells without pre-activation of CD44, whereas the SHAP⅐HA complex is not.
Many studies have indicated that the CD44-HA interaction could support the rolling and initial adhesion of leukocytes onto the vascular endothelium under physiological shearing force, and could be a new mechanism other than the selectin-glycan interaction for the capture of circulating leukocytes. The present results support the previous findings, and further imply an enhancing role of SHAP in the molecular interaction. Once the endothelial cells deposit HA on their luminal surfaces, the association of SHAP to these HA molecules is very likely, because the blood contains a large amount of I␣I and the enzymatic activity, and in fact, the simple incubation of plasma with HA could result in the complex formation (45). Although it is not yet clear what events are triggered by such a mechanism at present, we noticed an interesting observation made by Milinkovic et al. (17) in acute graft-versushost disease, one of the major limitations for the curative therapy of aplastic anemia and hematological malignancy with allogeneic hematopoietic stem cell transplantation. Acute graft-versus-host disease is characterized by dermal lymphocytic infiltration following hematopoietic stem cell transplantation. A lymphocyte adhesion study revealed that the papillary dermal vessels in acute graftversus-host disease, but not those in most other posthematopoietic stem cell transplantation skin eruptions, supported the shear-resistant adherence of lymphocytes via a CD44-HA interaction (17,50). Only the HA deposits on the vascular endothelium, but neither the comparable HA deposits at the dermalepidermal border and within glandular elements of the reticular dermis, nor HA immobilized on plastic, supported the lymphocyte adhesion, suggesting that the endothelial HA deposits have special properties (17). It would be interesting to examine the formation of the SHAP⅐HA complex under this condition.
It is noteworthy that the range of shear force where an enhancing effect of SHAP was observed (Ͻ1 dyne/mm 2 ) is lower than the physiological levels (at 1-2 dynes/mm 2 ) (Fig. 3), making the physiological relevance of the present finding in a microcirculation environment uncertain. However, as discussed below, many other molecules may participate in the regulation of CD44-HA interaction in vivo, making SHAP effective at higher ranges of shear force as well.
On the other hand, CD44-HA interaction also occurs in extravascular tissues such as the joint cavity and various stromal extracellular matrices with low shear forces. HA is one of the most important components of extracellular matrix supporting in general the active cell propagation, migration, and differentiation. Processes such as inflammation, wound healing, and tumor malignancy are well known to be associated with up-regulated HA production and deposition and the CD44-HA interaction between extravasated leukocytes/metastatic cancer cells and the stromal extracellular matrix. Participation of SHAP in these interactions has been suggested by de la Motte et al. (8,51) in inflammatory bowel diseases. The main pathological changes of IBD include an increase in intestinal mucosal mononuclear leukocytes and a dramatic hyperplasia of the muscularis mucosae. The interaction between recruited leukocytes and mesenchymal smooth muscle cells is thought to be important in the development and propaga- Liposarcoma that expresses predominantly as the standard form of CD44 was used as a control. In addition to the CD44H-derived product, two variant form-derived products were found in Hut78 cells. By the cloning and sequencing of the PCR products, it was found that the longer one was derived from CD44E (containing variable exons V8, V9, and V10), whereas the shorter one was a mixture of fragments derived from two variant forms, one containing V8 and V9, and the other containing V10.
tion of inflammatory bowel disease. By infecting the cultured colon smooth muscle cells with virus, an etiological factor of inflammatory bowel disease, the authors observed an up-regulation of HA production and an extracellular deposition of these HA molecules into the pericellular "coat" structure and the "cable" structure spanning several cell lengths, in contrast to the small patchy structure in unstimulated cells. The formation of the special HA structures was accompanied by a dramatically increased CD44-mediated adhesion of mononuclear leukocytes. Interestingly, the study indicated that I␣I heavy chains are required for the formation of the HA structures (51). In the present study, an immunofluorescence staining study failed to reveal a different microstructure of the immobilized SHAP⅐HA complex from that of free HA in the absence of cells. The most likely reason was that the SHAP⅐HA complex used in this study was highly purified, lacking the other components found in de la Motte's cable structure. In relation to this, our previous study showed that versican/PG-M interacts with both HA and the heavy chains of I␣I to facilitate the complex formation (52); mesenchymal smooth muscle cells synthesize a large amount of versican/PG-M and deposit it around them (53). In addition, several recent studies have shown that TSG6 is able to replace the chondroitin sulfate of I␣I to form a covalent complex with the heavy chain and then transfer it to HA (54,55). The TSG6⅐HC complex has been found in synovial fluids from RA patients (56) and the extracellular matrix of expanded cumulus oophorus (57). The absence of either TGS6 or SHAP in the cumulus extracellular matrix causes severe functional impairment (45,59). However, it is still likely that a novel type of fine microstructure of HA was formed in the presence of SHAP because of the aggregating property of the SHAP⅐HA complex (47) and the increased avidity to CD44 observed in this study. Relating to this possibility, the confocal microscopic observations on the cell-surface distributions of HA and CD44 revealed somewhat different distributions of both molecules: their co-distributed population on the flank surfaces on the SHAP⅐HA substratum (see the yellow color of the merged photos in Fig. 9). The question remains to be answered using more powerful techniques, for example, the rotary shadowing technique (60,61) or atomic force microscopy (62), both having been applied to the study of HA microstructure and/or conformational change. Taken together, the present results argue strongly for a role of SHAP modification of HA in the pathogenesis of related diseases, although the presence of other HA-binding molecules, for instance, TSG6 and PG-M/versican, may be further required to form the special HA structures exhibiting high avidity to CD44-bearing leukocytes.
We also examined immunohistologically the involvement of these molecular interactions in synovitis of RA patients. It is known that the synovial fluid of RA patients includes mas-sively the SHAP⅐HA complex, which was speculated to largely derive from the inflammatory synovium, a site with active HA production and serum protein efflux. As predicted, the results revealed the presence of a large amount of the SHAP⅐HA complex in the hyperplasic synovium. Notably, most of the infiltrated leukocytes were positive on their surface for the complex, in addition to CD44, suggesting molecular interactions among these molecules on the surfaces of synovial infiltrating leukocytes. Studies in mice with CD44-blocking antibodies (10,11) or with CD44 knock-out mice (9) have revealed a positive role of CD44 in the development of collagen-induced arthritis, a murine model for human RA. Consistently, we observed that the mice incapable of SHAP⅐HA complex formation were more resistant to the collagen-induced arthritis than their normal littermates (46). Therefore, it is likely that SHAP, by potentiating the CD44-HA interaction, plays a positive role in the leukocyte infiltration and activation in the hyperplasic synovium of RA joints.  The present results suggest that the enhancing effect of SHAP is indirect. Although further studies are required to define the exact mechanism, it is most likely related to the aggregating property of the SHAP⅐HA complex that we observed recently (47). The huge linear HA molecule adopts a random coil configuration with a certain degree of stiffness. The size of the random coils is so large that at concentrations higher than 1 mg/ml the HA coils entangle to form a continuous flexible molecular network and exhibit unique physicochemical and biological features (63). Given the oligomerization of CD44 upon activation (29,30) and the higher avidity of multivalent HA to CD44 (38), we hypothe-size that the CD44-HA interaction-enhancing effect of SHAP is due to the accelerated aggregation of HA as a consequence of the selfassociation of SHAP molecules. In contrast, we found no such effect on the HABP-HA interaction in this study, although the HABP and CD44 share the conserved HA-binding link module domain. The exact molecular mechanism remains to be further investigated. A noteworthy fact is that both aggrecan and link protein included in the HABP consist of two tandem link module domains forming an extremely stable ternary complex with HA, whereas CD44 has only one, of which the dimerization/oligomerization has been shown to be an important mechanism of the regulation of HA avidity.
The signaling cascades initiated by HA-CD44 interaction have been extensively studied in recent years (64,65). In leukocytes, the mitogenactivated protein kinase and NF-Bmediated pathways were found to be important (58). In general, cell response to HA varies depending on the HA amount, HA chain length, and cell background. Although experimental evidence is lacking, it may be expected that the HA-binding proteins (e.g. SHAP and TSG6) become an additional factor modifying HA signaling. The question of how SHAP alters HA signaling, quantitatively or qualitatively, also remains to be answered. However, considering the results of this study, it is likely that SHAP strengthens the signal intensity of HA but does not evoke a novel pathway unresponsive to free HA.   11. Representative immunohistology of the hyperplastic synovium of RA patients (Steinbrocker's stage III). The synovium is extensively stained by HABP and anti-I␣I antibody (counterstained with methyl green). The negative staining by anti-bikunin antibody (counterstained with hematoxylin) indicates that the immunoreactivity corresponds to the SHAP⅐HA complex rather than I␣I. Compared with the distribution pattern of HA, that of SHAP is patchier. The infiltrated CD44-positive leukocytes (counterstained with methyl green) are strongly positive on their surfaces for HA and SHAP staining. The arrow and arrowhead indicate the leukocyte and synoviocyte, respectively. Similar staining patterns were obtained with the samples at stage IV. (Magnification: ϫ400.)