Hyaluronan Anchoring and Regulation on the Surface of Vascular Endothelial Cells Is Mediated through the Functionally Active Form of CD44*

CD44 on lymphocytes binding to its carbohydrate ligand hyaluronan can mediate primary adhesion (rolling interactions) of lymphocytes on vascular endothelial cells. This adhesion pathway is utilized in the extravasation of activated T cells from the blood into sites of inflammation and therefore influences patterns of lymphocyte homing and inflammation. Hyaluronan is a glycosaminoglycan found in the extracellular matrix and is involved in a number of biological processes. We have shown that the expression of hyaluronan on the surface of endothelial cells is inducible by proinflammatory cytokines. However, the manner through which hyaluronan is anchored to the endothelial cell surface so that it can resist shear forces and the mechanism of the regulation of the level of hyaluronan on the cell surface has not been investigated. In order to characterize potential hyaluronan receptors on endothelial cells, we performed analyses of cell surface staining by flow cytometry on intact endothelial cells and ligand blotting assays using membrane fractions. Hyaluronan binding activity was detected as a major species corresponding to the size of CD44, and this was confirmed to be the same by Western blotting and immunoprecipitation. Moreover, alterations in the surface level of hyaluronan after tumor necrosis factor- a stimulation is regulated primarily by changes in the cell surface levels of the hyaluronan-binding form of CD44. In laminar flow assays, lymphoid cells specifically roll on hyaluronan anchored by purified CD44 coated on glass tubes, indicating


CD44 on lymphocytes binding to its carbohydrate ligand hyaluronan can mediate primary adhesion (rolling interactions) of lymphocytes on vascular endothelial cells. This adhesion pathway is utilized in the extravasation of activated T cells from the blood into sites of inflammation and therefore influences patterns of lymphocyte homing and inflammation. Hyaluronan is a glycosaminoglycan found in the extracellular matrix and is involved in a number of biological processes. We have
shown that the expression of hyaluronan on the surface of endothelial cells is inducible by proinflammatory cytokines. However, the manner through which hyaluronan is anchored to the endothelial cell surface so that it can resist shear forces and the mechanism of the regulation of the level of hyaluronan on the cell surface has not been investigated. In order to characterize potential hyaluronan receptors on endothelial cells, we performed analyses of cell surface staining by flow cytometry on intact endothelial cells and ligand blotting assays using membrane fractions. Hyaluronan binding activity was detected as a major species corresponding to the size of CD44, and this was confirmed to be the same by Western blotting and immunoprecipitation. Moreover, alterations in the surface level of hyaluronan after tumor necrosis factor-␣ stimulation is regulated primarily by changes in the cell surface levels of the hyaluronan-binding form of CD44. In laminar flow assays, lymphoid cells specifically roll on hyaluronan anchored by purified CD44 coated on glass tubes, indicating that the avidity of the endothelial CD44/hyaluronan interaction is sufficient to support rolling adhesions under conditions mimicking physiologic shear forces. Together these studies show that CD44 serves to anchor hyaluronan on endothelial cell surfaces, that activation of CD44 is a major regulator of endothelial surface hyaluronan expression, and that the non-covalent interaction between CD44 and hyaluronan is sufficient to provide resistance to shear under physiologic conditions and thereby support the initial steps of lymphocyte extravasation.
Hyaluronan (HA) 1 is a high molecular weight nonsulfated linear glycosaminoglycan comprised of a range of repeating disaccharide subunits, ␤1,3-N-acetyl-D-glucosamine in linkage to ␤1,4-D-glucuronic acid, and ranging in molecular mass up to 1 ϫ 10 7 Da. As a prominent ubiquitously expressed extracellular matrix component, HA can be produced by a wide variety of cell types and tissues and is most prevalent in soft connective tissues. HA has been suggested to play a key role in several biological processes including embryonic development (1), extracellular matrix organization and turnover (2)(3)(4), wound healing (5,6), tumor growth (7), and angiogenesis (8). It is generally thought to exert many of these functions by establishing a provisional matrix for supporting cellular migration and adherence (9). Moreover, studies have demonstrated that HA may function as a cellular signaling molecule under certain circumstances (10,11) and can deliver signals leading to or regulating cellular proliferation (2,(12)(13)(14).
Although a member of the family of glycosaminoglycans, HA is distinct from other members in that it is not found covalently linked to a peptide core (2). This is also in marked contrast to other carbohydrate ligands participating in extravasation, such as selectin ligands, which are generally glycoproteins anchored by transmembrane domains (15). In further contradistinction to other glycosaminoglycans and proteoglycans, HA is unique in its mode of synthesis. Rather than being confined to the Golgi apparatus or post-Golgi compartment, the assembly of HA by HA synthase is located at the plasma membrane where HA is synthesized and polymers are directly extruded into the extracellular space (16 -18). Although its extracellular organization and disposition in a number of solid tissue microenvironments has been examined, HA on endothelium at the interface with the bloodstream has not been extensively studied.
The recognition of endothelial cells by leukocytes and their subsequent extravasation through the blood vessel wall are based on a multistep pathway of sequential receptor engagement, in which a variety of molecular ligands participate (19 -21). We have described a novel primary (rolling) interaction between T cells and endothelial cells that is similar under laminar flow conditions to that mediated by selectins and also has as its basis a distinct protein-carbohydrate ligand interaction, namely that between the cartilage link protein family member CD44, a broadly distributed complex multifunctional family of cell surface glycoproteins (22), and its principal ligand HA (23). One well known consequence of antigen stimulation on T cells is increased surface levels of CD44 (22,24,25). However, elevated levels of CD44 do not necessarily correlate with increased HA binding, and thus the ability of CD44 to bind HA is not constitutive; rather, CD44 requires some form of structural alteration to engage this ligand (22,26,27). Although the mechanism by which CD44 is altered to bind HA remains to be completely elucidated, evidence has accumulated that T cell stimulation of normal lymphocytes in vitro or in vivo via signaling through the T cell receptor induces the activated form of CD44 and attendant rolling interactions on HA substrate (22,25,28). These observations have established the HA-binding form of CD44 as an early activation marker on T cells after T cell receptor stimulation and support a role for this interaction during the course of an immune response. In particular, CD44/HA interactions have been shown to be required for extravasation of superantigen-stimulated T cells into an inflamed site in a mouse model (28). CD44 has been prominently associated with human arthritis and with a model of collagen-induced murine arthritis (29 -32), and more recently with a murine model of multiple sclerosis (33). CD44 interactions with HA are also thought to be important in allogeneic graft rejection (34,35). Based on our observations, we have postulated that CD44 on lymphocytes interacts with HA on endothelium and participates in the well known preferential homing of activated lymphocytes to tertiary sites of inflammation. In support of this model, circulating lymphocytes bearing activated CD44 have been shown to be elevated during autoimmune exacerbations in both arthritis and systemic lupus erythematosus in humans (30).
A clear implication of this model is that regulation of HA on vascular endothelium in response to local pathophysiologic conditions should create a receptive site for leukocyte recruitment and therefore represents an important control point for extravasation. It has been reported that hyaluronan is found both in vitro and in vivo on endothelial cells (23, 36 -41). In addressing the potential regulation of HA on endothelial cells, we previously demonstrated that its expression on cultured endothelial cell lines and primary endothelial cultures is inducible by the proinflammatory cytokines TNF␣, interleukin-1␤, interleukin-15, and bacterial lipopolysaccharide (42,43). In addition, this inducibility appeared strikingly restricted to endothelial cells derived from microvascular but not large vessel sources, consistent with a role in the vessels where the majority of leukocyte trafficking occurs. The elevated HA levels induced by cytokines further resulted in increased adhesive interactions in both non-static shear and laminar flow adhesion assays. However, the changes in surface HA levels did not appear dependent either on described HA synthetic or degradative enzymes, suggesting other mechanisms for such regulation. These data added to the selectin and immunoglobulin gene families a new inducible endothelial adhesive molecule, hyaluronan, and helped to further our understanding of the potential physiologic roles of the CD44/HA interaction. However, the manner by which HA is anchored to the endothelial surface and the requirements for cell surface retention under conditions of physiologic shear forces has remained unclear.
A number of proteins have been characterized that can bind HA and anchor it in the tissues and on cell surfaces. The proteoglycans aggrecan and link protein are major binders of HA in cartilage and soft tissues (44). A specialized receptor for HA internalization is found on liver sinusoidal endothelial cells (45,46). Other cell surface HA receptors include the widely distributed CD44 molecule (22), and an unrelated protein, re-ceptor for hyaluronan-mediated motility (RHAMM) (47). In these studies we show that CD44 is the predominant cell surface molecule on endothelial surfaces responsible for HA binding and that it is the regulation of the activated form of CD44 that largely determines the level of surface HA expression. Furthermore, using isolated CD44 bound to HA, we establish that this interaction is sufficient to resist hemodynamic drag forces and support rolling adhesions of lymphocytes under physiologic laminar flow conditions.
Cell Culture-BW5147 murine T cells were maintained in RPMI 1640 containing 10% FCS, 100 mM sodium pyruvate, 200 mM L-glutamine, and 50 M ␤-mercaptoethanol. The murine lymph node endothelial cell line SVEC4-10 (56) was grown and maintained in DMEM containing 10% FCS and 200 mM L-glutamine. For TNF␣ stimulation, cell monolayers were grown to 60 -80% confluence, washed with DMEM, and stimulated with TNF␣ (10 ng/ml) for 4 h or as indicated in time course analyses. Single cell suspensions were made by incubation of monolayers in Versene (Life Technologies, Inc.) at 37°C for 10 min for staining or immunoprecipitation. For Fl-HA binding experiments, the cells were additionally treated with 20 units/ml hyaluronidase for 1 h at 37°C. Where indicated, trypsin was used at a final concentration of 0.5 mg/ml for the indicated times at 37°C. Trypsin was inactivated by the addition of 10% fetal bovine serum.
Primary cultures of human dermal microvascular endothelium (HD-MEC) were obtained from the Skin Center at Emory University (S. W. Caughman, Atlanta, GA) and maintained in Iscove's modified Dulbecco's medium supplemented with 20% human serum, 2.5 g/ml cAMP, 10 ng/ml epidermal growth factor, and 5 ng/ml hydrocortisone. After reaching initial confluence, cells were passaged and used directly or after one additional passage to fresh plates. Cells were stimulated with human TNF␣ at 10 ng/ml for 4 h.
Fluorescence-activated Cell Sorter Analysis (FACS)-5 ϫ 10 5 cells were stained with Fl-HA, anti-CD44-PE (IM7), anti-H-2, or bPG-biotin in 100 l of PBS, 5% FCS for 30 min on ice and then washed with 1 ml of PBS containing 5% FCS. Cells were incubated with Fl-HA for 10 min prior to addition of IM7-PE. Fluorochrome-labeled streptavidin or secondary antibody was added for 20 min as indicated, and cells were again washed before analysis. For blocking of Fl-HA staining, unlabeled blocking anti-CD44 antibody (KM81 or 515) was incubated with cells for 10 min prior to staining with Fl-HA. Data were collected using FACScan TM analytical instrument (Becton Dickinson, San Jose, CA) and analyzed using CellQuest TM software.
Western Blot Analysis-SVEC4-10 cells were grown in 80-mm tissue culture dishes to 60 -80% confluence in DMEM containing 10% FCS. The monolayer was washed with DMEM and stimulated with TNF␣ (10 ng/ml) for 4 h. The monolayer was again washed with DMEM, treated with 20 units/ml hyaluronidase for 1 h at 37°C, washed with chilled PBS, and cells scraped off the dish in lysis buffer (50 mM Tris, pH 8.0, 100 mM NaCl, 5 mM EDTA, 1 mM phenylmethylsulfonyl fluoride). BW5147 cells were grown in suspension, washed twice with chilled PBS, and resuspended in lysis buffer. The cell lysates were sonicated briefly, and membranes were prepared by centrifugation at 100,000 ϫ g (57). Pellets were resuspended in non-reducing sample buffer and resolved on a 10% polyacrylamide gel containing 0.1% SDS, after which proteins were transferred to nitrocellulose (Amersham Pharmacia Biotech). Ligand blotting was performed based on described methods (58). Membranes were incubated with Fl-HA/anti-fluorescein-biotin or IM-7-biotin. Fl-HA was used 1:100 diluted in a 50% Blocking solution (ECL kit, Roche Molecular Biochemicals) and was incubated 4 h at ambient temperature. Blots were washed with PBS, 0.1% Tween 20, 100 mM NaCl for 1 h three times. Biotinylated anti-fluorescein or anti-CD44 antibody was used at 1:1000 dilution and incubated for 2 h at room temperature and washed as above. Blots were developed with a chemiluminescence kit using streptavidin-POD according to manufacturer's instructions (Roche Molecular Biochemicals).
Immunoprecipitation-SVEC4-10 membrane lysates were prepared as described above for Western blots. The pellet was resuspended in immunoprecipitation buffer (50 mM Tris, pH 8.0, 5 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, and 1% Nonidet P-40). Fl-HA/anti-fluorescein-biotin or IM7-biotin was added to the above mixture and incubated at 4°C for 2 h on a rocking platform. Complexes were precipitated using streptavidin-Sepharose (Sigma). Bound material was eluted by lowering the pH to 3.0 with 100 mM glycine HCl, neutralized with 100 mM Tris-HCl, pH 8.0, and analyzed by SDS-PAGE on a 10% polyacrylamide gel under reducing conditions. Following transfer to nitrocellulose, precipitated material was probed with IM7-biotin and developed with chemiluminescence using streptavidin-conjugated peroxidase.
Reverse Transcription-Polymerase Chain Reaction (RT-PCR)-SVEC4-10 cells were grown to 60 -80% confluence and then stimulated with 10 ng/ml TNF␣ for 4 h. Total RNA was prepared according to manufacturer's instructions using Trizol reagent (Life Technologies, Inc.). The reverse transcriptase reaction and PCR amplification were performed as described (59). CD44 forward primer sequence was 5Ј-GCACACCTACCTTCCTAC-3Ј and reverse primer sequence was 5Ј-CTGGAATCTGAGGTCTCCTC-3Ј. ␤-Actin primers were used as control. PCR product was analyzed by agarose gel electrophoresis.
Adhesion Assay under Flow Conditions-Glass capillary tubes (1.41 mm inner diameter; Drummond Scientific, Broomall, PA) were coated with 1% BSA, soluble recombinant human CD44-Ig (350 g/ml), control human IgG (350 g/ml), soluble HA (1 mg/ml), or protein followed by sHA. To coat tubes with two substrates, tubes were first incubated with BSA, sCD44-Ig, or human IgG for 16 h at 4°C and then further incubated with 1 mg/ml HA solution for 3 h at room temperature. sHA binding to sCD44-Ig-coated tubes was blocked by preincubation with HA-blocking anti-CD44 antibody prior to the addition of sHA to the tubes. All tubes were additionally blocked with 1% BSA/PBS for 1 h at room temperature. BW5147 murine T cells were washed and resuspended in RPMI 1640 medium at a concentration of 2 ϫ 10 6 /ml. The medium and the flow chamber were equilibrated to 37°C. Medium containing BW5147 cells was continuously pulled through the capillary tube by means of a Harvard syringe pump at flow rates of between 0.5 and 5 ml/min, corresponding to shear stresses of 0.3-3.0 dynes/cm 2 (60). Interaction of the BW5147 cells with the wall of the capillary tubes (rolling) was monitored by the use of a Nikon Diaphot-TMD inverted phase contrast microscope connected to a video camera and recorder. Rolling numbers were determined by counting the number of cells/min crossing a fixed position perpendicular to the flow of cells on the video screen.

RESULTS
HA Is Anchored to the Endothelial Surface via a Trypsinsensitive Protein-Cell surface HA is often found non-covalently associated with membrane proteins, of which CD44 would be a prime candidate (22). To characterize the nature of the association of HA with the endothelial surface, the peripheral lymph node-derived microvascular EC line SVEC4-10 was used after TNF␣ activation to induce high HA expression as described (42). Cells were then stained with the HA-binding proteoglycan bPG to assess HA surface levels. To measure other surface markers, cells were stained simultaneously with anti-CD44 or anti-H-2 and then subjected to flow cytometric analysis before and after various intervals of treatment with trypsin. The ability of bPG to bind to this cell line was completely trypsin-sensitive (Fig. 1). Diminution of staining was evident within 5 min of trypsin treatment, and staining was 50% reduced by 10 min and completely ablated by 30 min. Staining with anti-CD44 IM7-PE diminished in parallel with bPG staining, correlating the trypsin sensitivity of HA expression with that of CD44, suggestive that this may be the HAanchoring protein. In contrast, staining with anti-H-2 antibody was unaffected by trypsin treatment, consistent with the trypsin insensitivity of the extracellular portion of mouse class I MHC molecules (61). Thus, HA is attached to endothelial surfaces through a trypsin-sensitive moiety that correlates with the cell surface levels of CD44.
CD44 Is the HA-binding Protein on EC Surfaces-To establish further the nature of the EC surface HA-binding protein, we used a strategy of removing surface HA by digestion with hyaluronidase and then reintroducing fluorescein-labeled HA (Fl-HA) as a probe for rebinding to the cell surface. We were able to use this technique, in combination with monoclonal antibodies, as a means to inquire about the protein involved in the binding. Following hyaluronidase treatment, bPG staining of SVEC cells decreased essentially to background levels ( Fig.  2A). At the same time, there was a demonstrable increase in the ability of hyaluronidase-treated EC to bind Fl-HA (Fig. 2B). The Fl-HA binding was further shown to be HA-dependent by blocking with unlabeled HA (data not shown). In addition, the ability of hyaluronidase-treated EC to bind Fl-HA was blocked to background levels by incubation with the HA-blocking anti-CD44 antibody KM81 but not with non-HA-blocking anti-CD44 KM703 (Fig. 2C). This demonstrates that HA binding on the surface of intact EC is dependent on CD44 and in particular is localized to the HA-binding portion of the molecule, implying that it is the activated form of CD44 that is required. In addition, the data suggest that surface HA expression may depend more on the ability of cells to retain HA on the surface than to a change in the production level of HA. To generalize these results and also extend them to a more physiologic type of endothelium, a source of freshly isolated primary microvascular endothelial cells, human dermal microvascular EC (HDMEC), was used. We have previously shown that these cells respond to TNF␣ treatment with increased levels of surface HA (42). Results are shown in Fig. 2D. As with transformed endothelial lines, Fl-HA binding by hyaluronidase-treated EC was completely blocked by the HA-blocking anti-CD44 antibody 515 but not by a non-blocking anti-CD44 monoclonal antibody Hermes 3. Thus, the CD44 dependence of HA binding appears to maintain for primary cultures of microvascular endothelium as well as cell lines, and this occurs in human as well as murine EC.
Identification of CD44 as the HA-binding Membrane Protein by Ligand Blotting-To confirm further by a separate biochemical criterion that CD44 is the HA-presenting polypeptide on the EC surface, ligand blotting assays were performed. In addition to identifying the reactive species, since CD44 can potentially be expressed in a variety of isoforms resulting from both alternative splicing of variant exons and varied posttranslational modification, this analysis also has the potential to reveal any structural differences that would be seen as altered M r . In these experiments, the endothelial form of CD44 from SVEC4-10 was compared with that expressed on a murine T cell line, BW5147, that we have previously characterized (25). Western blot analysis of cell membrane preparations from these cells was performed. After transfer to nitrocellulose, samples were reacted with either anti-CD44 or Fl-HA. As shown in Fig. 3, both cell lines express a single discrete species of CD44 at the same molecular weight, which is consistent with the hematopoietic (CD44H) form of the protein (80 -90 kDa). By the criterion of alteration in size, no evidence of alternative splicing, variation in glycosylation, or other post-translational modification was seen at the level of resolution of this technique. In addition, Fl-HA detected a band of identical molecular weight in both SVEC and BW cells. Incubation of the nitrocellulose membrane with soluble, unlabeled HA prior to incubation with Fl-HA completely inhibits Fl-HA binding, demonstrating specificity and that the binding is specific for the HA portion of the probe. That this band represents CD44 was confirmed by depletion of CD44 from lysates with antibody prior to electrophoresis and blotting and development of membranes with Fl-HA. Such depletion completely removed the HA-binding species. Thus, the data from both cell surface staining of intact cells and Western blot analysis of cell membranes are consistent with CD44 being the HA-binding moiety on endothelial cells.
Cytokine Stimulation of Endothelial Cells Is Accompanied by Increased Capacity of CD44 to Bind HA-Whereas it has been previously shown that endothelial cell activation by proinflammatory stimuli results in increased expression of surface HA, this increase in HA expression did not appear associated with significant changes in the mRNA of any of its major metabolic or catabolic enzymes (42), suggesting that other mechanisms are more important for its surface expression. To determine whether increases in HA expression on activated EC result from alterations in the anchoring protein, in particular its expression level or ability to bind HA, we treated EC with TNF␣ followed by harvesting and staining at hourly intervals. At each interval, aliquots of cells were treated with hyaluronidase to remove bound HA and then restained with Fl-HA (Fig.  4) so that the HA binding capacity of these cells could be monitored. Cells were also stained with bPG, anti-CD44, or Fl-HA in similar manner to Fig. 2. As described previously, bPG staining increases with duration of TNF␣ treatment (42). When cells were pretreated with hyaluronidase, a concomitant gradual increase in the ability to bind Fl-HA over time is evident. In contrast, no change in Fl-HA binding is seen in the absence of hyaluronidase treatment (Fig. 4). Significantly, no  3. Western blotting further indicates that the HA-binding moiety on SVEC4-10 membranes is CD44. Membranes were prepared from confluent monolayers of SVEC4-10 and from BW5147 cells for Western blot analysis. Total membrane fractions were resuspended in non-reducing sample buffer, and 10 l of each was electrophoresed on a 10% polyacrylamide gel. The proteins were transferred to nitrocellulose membranes, and ligand blotting was performed using either biotinylated rat anti-mouse CD44 monoclonal antibody (IM7) or Fl-HA plus biotinylated anti-fluorescein, as indicated. Bound reagent was detected using a streptavidin-POD chemiluminescence system. In both SVEC and BW cell membrane preparations, CD44 migrates as a protein of 80 -90 kDa. In both extracts, Fl-HA also interacts with a protein of the same molecular weight. The binding of Fl-HA is abolished by preincubation of membranes with unlabeled soluble HA, and prior depletion (Depl) with anti-CD44 antibody removes Fl-HA-binding material from the membrane preparations, indicating that it is CD44 on both SVEC and BW cells that is responsible for Fl-HA binding. discernible increase in total CD44 expression was detected by anti-CD44 staining (Fig. 4). Since the surface Fl-HA binding we observe is CD44-dependent (Fig. 2), this is highly suggestive that the increase in HA surface expression results from activation of CD44 to its HA-binding form rather than an increase in the level of CD44 on the cell surface.
It was of further interest to determine whether this result extended to the primary EC source, HDMEC, used above (Fig.  2D). When primary human dermal EC cultures were similarly treated with TNF␣ for 4 h, results similar to those observed for SVEC4-10 cells were obtained. HA levels increased in response to TNF␣ (Fig. 5A), and the ability to bind Fl-HA following hyaluronidase treatment was increased as the result of TNF␣ treatment (Fig. 5B). As with the SVEC4-10 cells, CD44 levels remained constant (Fig. 5C).

Biochemical Analysis Substantiates That the HA-binding Form of CD44 on EC Membranes Is Increased after Cytokine
Induction-We proceeded to more directly assess the activated state of CD44 on EC membranes by using direct immunoprecipitation followed by SDS-PAGE analysis. We had determined that the denaturation of CD44 as the result of SDS-PAGE analysis and transfer to nitrocellulose membranes done in a standard Western analysis (Fig. 3) renders the CD44 in an active HA-binding conformation, even for cell populations that do not bear active CD44 on the surface (data not shown). Therefore, direct Western blotting does not necessarily reflect the original CD44 conformation with respect to HA binding on the cell surface. We therefore used an alternative means of identifying the amount of active form of CD44 for SDS gel analysis by immunoprecipitating with either Fl-HA or anti-CD44 from isolated intact EC membrane preparations before and after TNF␣ treatment and detecting precipitated material with anti-CD44 following polyacrylamide gel electrophoresis and transfer to nitrocellulose. The relative level of Fl-HA binding material in the membrane fraction compared with IM7 binding material was examined. Resting and TNF␣-activated cells were first treated with hyaluronidase to remove surface HA as described above. Cells were lysed by sonication, and membrane fractions were immunoprecipitated with either IM7-biotin or Fl-HA/anti-Fl-biotin, followed by streptavidinagarose. Precipitated, eluted material was electrophoresed on a 10% reduced SDS-polyacrylamide gel, transferred to nitrocellulose, and probed with IM7-biotin plus streptavidin-conjugated peroxidase. The portion of the gel below 55 kDa was removed to eliminate signal from the biotinylated reagents used in the immunoprecipitation. As seen in Fig. 6A, Fl-HA and IM7 precipitate bands of identical molecular weight. Notably, although there was no significant difference in the intensity of the IM7 precipitated band before and after TNF␣ treatment, consistent with the results of cell surface staining (Fig. 5), there is a dramatic increase in the amount of Fl-HA precipitable material in membranes of activated EC. This result suggests that whereas overall CD44 levels are not increased following TNF␣ treatment, TNF␣ activation does result in CD44 being expressed in its conformationally active, HA-binding form. All of the HA-precipitable material is removed by prior depletion of the samples with anti-CD44 antibody. The lighter band at Ͼ97 kDa represents an intermittent artifact at the interface of the stacking and resolving gels, which is preferentially removed with stringent washing. The detection of only a single, discrete band in either IM7 or Fl-HA precipitates also indicates little or no change in CD44 glycosylation or variant exon expression as a result of TNF␣ activation.
Although the protein analysis provided above did not show variation in the size of CD44 on transition to its activated state, to ascertain further whether alternative exon usage occurs on cytokine-stimulated EC, PCR analysis of mRNA from TNF␣treated SVECs was performed to determine whether HA induction is associated with alternatively sized transcripts. By using oligonucleotide primers specific for the 5Ј and 3Ј invariant exons flanking the variant exon insertion region (62)(63)(64), no detectable difference in the amount of PCR product was found at any time point when compared with product from untreated cells (Fig. 6B), consistent with the cell surface staining (Fig. 4). In addition, no shift in the 218-base pair (no variant exons) size of PCR product or bands of increased size were seen. The PCR data are thus consistent with the CD44 immunoprecipitation data, together suggesting variant exon usage does not explain the transition of CD44 to its active form on activated EC.
Purified CD44 Is Sufficient to Present HA and Serve as Substrate for Primary Adhesion of Lymphoid Cells under Conditions of Laminar Flow-Whereas carbohydrates are clearly involved in selectin interactions that result in the capture and rolling of leukocytes in vitro and in vivo, these are typically found in covalent association with transmembrane protein cores (15). The noncovalent association of HA with CD44 raises the question of whether this interaction is sufficiently avid to resist the hemodynamic drag forces delivered by attaching and rolling leukocytes under conditions of dynamic flow. To address this, an in vitro expressed soluble CD44-immunoglobulin fusion protein (sCD44-Ig), previously shown to bind soluble HA (26,40), was produced. We made use of this protein to test the capacity of activated CD44 to present HA as a substrate for primary adhesion. Glass capillary tubes were coated with sHA, sCD44-Ig, or sCD44-Ig plus sHA, along with control substrates. The T cell lymphoma line, BW5147, which expresses constitutively activated CD44 and exhibits rolling under laminar flow (23), was analyzed under laminar flow in tubes coated with various substrates at 1.0 dynes/cm 2 (60). Interactions with substrates were then measured. Although soluble HA does adhere to polystyrene cell culture dishes and has been used directly for rolling on this surface (23,25,28), it is our experience that HA does not adhere directly to glass capillaries tubes, and thus no rolling is seen in tubes coated with sHA alone (Fig.  7A). Likewise, no rolling was observed in tubes coated with sCD44-Ig alone. However, when tubes were coated with sCD44-Ig followed by sHA, rolling interactions were striking. Neither BSA nor human IgG controls with or without HA supported this activity. The capacity for sCD44-Ig to bind and present sHA for rolling was inhibited by incubation with blocking anti-CD44 prior to the addition of sHA to the tubes. Thus, an active form of CD44 can bind soluble HA in sufficient quantity and with sufficient avidity to support CD44/HA-mediated primary adhesion under shear stress. To analyze further the  6. Analysis by immunoprecipitation shows increases in the HA-binding form of CD44 after cytokine induction. A, untreated and TNF␣-treated (4 h) SVEC4-10 cells were treated with hyaluronidase, and membrane fractions were prepared, as described under "Experimental Procedures." Biotinylated anti-CD44 monoclonal antibody (IM7) or Fl-HA plus biotinylated anti-fluorescein monoclonal antibody were used together with streptavidin-Sepharose for immunoprecipitation. The bound material was eluted by lowering the pH to 3.0, and aliquots of the sample were resolved on 10% polyacrylamide under reducing conditions, transferred to nitrocellulose membranes, and proteins detected with anti-CD44-biotin plus streptavidin-POD. Anti-CD44 detected bands of identical size in both anti-CD44 and Fl-HA immunoprecipitates. No change was observed in the amount of anti-CD44 precipitable material in the presence or absence of TNF␣. However, the HA-binding fraction was considerably increased in TNF␣stimulated cells, suggesting CD44 activation rather than increased CD44 synthesis. Depletion with anti-CD44 prior to precipitation with Fl-HA removes essentially all HA-precipitable material. B, RT-PCR amplification of RNA from untreated and TNF␣-treated SVEC4-10 cells. Amplification was done using primers specific for the 3Ј-and 5Ј-flanking regions of the variant exon insertion site. Cycling conditions were 95°C/60 s, 55°C/90 s, and 72°C/120 s. Amplification of ␤-actin RNA under the same conditions from the same samples was carried out for each reaction set, and the resulting product was run in parallel with the CD44 reaction products. Results of ␤-actin amplification are shown in the bottom panel. RT-PCR amplification indicates that untreated and TNF␣-treated EC make equivalent levels of CD44 mRNA at all time points. In addition, no products other than the 218-base pair (bp) band deriving from the CD44H isoform were seen, indicating no significant role for expression of variant isoforms by activated EC.
ability of this interaction to support rolling over a range of wall shear stresses, cells were introduced at various pump speeds to affect varied wall shear stress, and the number of interacting cells was counted. As shown in Fig. 7B, significant numbers of cells interact with HA presented by CD44 over a range of WSS between 0.3 and 3.0 dynes/cm 2 . This shear response curve is similar to that seen between selectins and their ligands, is within the range of forces found in venules in which trafficking occurs, and is also similar to WSS responses seen previously with BW5147 on intact EC (23). Thus, CD44 interactions with HA anchored by sCD44 appear to mimic closely the flow characteristics observed on EC, and therefore this interaction has the capacity to represent the major mechanism for supporting rolling physiologically. DISCUSSION Investigations on the inducibility of endothelial cell surface molecules involved in extravasation have centered primarily on cell surface glycoproteins, including members of the selectin family, as well as various leukocyte integrin ligands such as the ICAMs and VCAM-1. Our prior demonstration that activated CD44 on leukocytes can mediate primary adhesion on its carbohydrate ligand, HA, on endothelium added a new class of molecule, glycosaminoglycans, as a novel target for the regulation of extravasation. Like endothelial glycoprotein adhesion receptors, HA surface levels can be modulated by proinflammatory stimuli (42,43), supporting the idea that such regulation may have direct relevance during immune responses and inflammatory processes. However, in contrast to such glycoproteins, which are generally firmly anchored in the plasma membrane by virtue of transmembrane insertion, HA is a simple linear polysaccharide that must rely on other mechanisms to adhere to the endothelial surface with sufficient avidity to resist shear forces. Our current studies have addressed this issue as well as the mechanism for the regulation of the level of HA on the endothelial cell surface.
The plasma membrane receptor CD44 is frequently the molecule responsible for binding and organizing HA around intact stationary and adherent cells (22). Under some circumstances, RHAMM, a molecule associated with cell motility and transformation, can also fulfill this function (11,47). However, unlike CD44, RHAMM has recently been shown to be decreased following activation of lymphoid cells (66). In the studies herein, virtually all HA binding on endothelial cell surfaces could be explained by activated CD44. This was established by specific blocking of Fl-HA binding to intact cells by anti-CD44 monoclonal antibody after removing surface HA with hyaluronidase (Fig. 2, C and D) and by direct Fl-HA blotting of electrophoresed endothelial membrane preparations (Fig. 3). Whereas other HA-binding species could exist in this system, their association with HA would need to be hyaluronidase-resistant (for cell surface staining) and/or their binding activity lost upon gel electrophoresis (ligand blotting). Hyaladherins generally require a minimum decasaccharide unit for binding (44), larger than the average fragment length resulting from hyaluronidase digestion. This is consistent with our direct experience with the proteoglycan we have used for HA detection, bPG, for which HA binding is completely hyaluronidase-sensitive ( Fig.  2) (23). The concordance of both cell surface staining of intact EC and ligand blotting of total EC membranes represents two independent and disparate criteria strongly suggesting that CD44 is indeed the major surface HA receptor on endothelial cells. HA synthase 1 (42) and HA synthase 2 2 expression in these ECs suggests other possible sources of HA binding material (16). For example, it has been suggested, in contrast to these studies, that although about 50% of HA binding in a model of keratinocyte development could be attributed to CD44, the remaining more diffusely distributed HA could not be and might instead be due to HA synthase association (67). Although HA has been shown to be associated with the HA FIG. 7. Soluble CD44 can present HA for primary adhesion under physiological shear stress. A, rolling interactions of BW5147 cells on soluble HA presented by sCD44-Ig using a capillary flow apparatus. BW5147 cells at a concentration of 2 ϫ 10 6 per ml were applied to the feed solution and pulled continuously through a glass capillary tube coated as indicated at a fixed wall shear stress of 1.0 dynes/cm 2 . Cells were allowed to equilibrate, and the total number of cells interacting were obtained by counting the number of cells/min passing a virtual line perpendicular to the flow of cells. Primary adhesion (rolling) was only observed on those tubes coated with sCD44-Ig followed by sHA. Binding of sHA to sCD44-Ig-coated tubes could be blocked by HA-blocking mouse anti-human CD44 monoclonal antibody 515, and no rolling was seen in blocked tubes. B, adhesion of BW5147 a glass capillary tube coated with sCD44-Ig plus by sHA was assessed at a variety of wall shear stresses. Cells were applied to feed solution already equilibrated under flow at an initial WSS of 0.3 dynes/cm 2 . The flow rate of the feed solution was incrementally adjusted by increasing the outlet pump speed to effect altered WSS, as indicated and as previously reported (80). The wall shear stress was calculated using the Poiseuille's Law for Newtonian fluid given by the equation T ϭ (V) (8/D) (u), where T is the wall shear stress in dynes/cm 2 , V is the velocity in mm/s, u is the viscosity in poise, and D is the diameter in mm. Cells were allowed to equilibrate for 2 min at each flow rate, and the total number of cells interacting were obtained by counting the number of cells/min passing a virtual line perpendicular to the flow of cells. The number of cells/min rolling across the line was determined for each WSS. The data shown are representative of at least three independent experiments. synthase of intact cells and membrane preparations (16), the topological orientation of intracellular and extracellular domains has not been experimentally determined, and it is unclear whether exogenous intact HA would in fact remain bound to this enzyme. It has also been suggested that HA synthase loses binding activity after HA chain elongation is complete (68). Thus, the nature and requirements for the tethering of nascent HA chains to HA synthases has not been well characterized, and it is uncertain whether binding of large HA polymers would be detected at the outer surface of the plasma membrane under conditions used in these studies.
In cells of mesodermal origin as well as some epidermal cells such as keratinocytes, substantial surface coats of HA can be observed that can occur to thicknesses of up to several microns (67,69). Our observations herein that CD44 is the primary molecule responsible for binding HA to the surface of endothelial cells suggests that, rather than serving to enhance homotypic aggregation and organization of cells in solid tissues, the anchoring of HA to CD44 on EC results in the capture of heterologous activated lymphocytes during extravasation. This function clearly places additional requirements on the interaction between CD44 and HA in several ways. The CD44 avidity to HA must be sufficient to resist the fluid shear forces encountered in the vasculature, as well as the additional forces generated by the hydrodynamic drag of circulating cells on HA. The HA filaments, while bound to CD44 on the endothelial surface, must additionally present unexposed HA-binding sites to the activated CD44 on circulating leukocytes in order for them to be captured in appropriate vascular beds.
It has been demonstrated using physicochemical means that HA molecules in solution form a three-dimensional network in which there are prominent interchain interactions (70). It would be expected that the conformation and organization of such HA filaments would vary depending on factors such as HA chain length and physiologic circumstances. Thus the conformation that HA assumes under postcapillary venular shear forces may be considerably different than that which it maintains in solid tissues. In some systems, shear stress has indeed been shown to alter macromolecular conformation. For example, shear flow has been shown to elongate mucin molecules in a manner so as to presumably expose relevant carbohydrate epitopes for selectin recognition (71), and activities of von Willibrand factor have been shown to be dependent on a conformation that is in turn dependent on shear forces (72). Significant lateral aggregation forces of HA chains have been previously inferred from atomic force microscopic studies (73). Since it appears from our studies that access to CD44-binding sites on HA must be available both at the endothelial as well as at the blood flow surfaces, it will be of interest to examine the organization and conformation of HA under shear conditions and determine in particular whether interchain HA interactions have sufficient stability to support rolling interactions or whether direct CD44 anchoring of all participating HA chains is necessary to support this function. It is also possible that there is selection of a particular molecular mass or mass range of HA on endothelial surfaces, since it has been reported that HA molecules of less than 5 ϫ 10 5 Da do not organize into complex structures (73), another issue meriting future study in this system.
The level of HA on endothelial cells is clearly regulated, and we have suggested that regulation occurs in response to microenvironmental perturbations within the underlying tissues such as provided under conditions of inflammation (42). However, a biochemical basis for this regulation had not previously been provided. Our prior studies did not support a major role for either HA synthase 1 or major HA degradative enzymes, including hyaluronidase, ␤-D-glucuronidase, nor ␤-N-acetyl-Dhexosaminidase (2,74), at least at the level of transcription of these genes. Of course, activity of these enzymes could be regulated other than transcriptionally, e.g. translational events such as enzyme modification, stabilization of these proteins intracellularly, or ability of these enzymes to retain newly synthesized HA could also play a role. Whereas HA synthetic and degradative enzymes must clearly play a role in the overall metabolism of HA in EC, a striking finding in the current studies is that a major determinant of the amount of HA expressed at the cell surface is the control of the level of the active HA-binding form of CD44 on the cell surface. TNF␣ stimulation was associated with significant elevation of HA surface expression without a change in overall levels of CD44 expression (Figs. 4 and 5). Nonetheless, after hyaluronidase digestion and restaining with Fl-HA, virtually all of the increase in Fl-HA binding could be accounted for by increases in the HA-binding form of CD44, suggesting that the regulation of CD44 activity is a key control point for the level of surface HA expression on vascular endothelium. Since HA synthase is anchored to the plasma membrane and HA is directly extruded to the outside, the relationship of CD44 to HA synthase on the surface is of keen interest. One model suggests that each molecule of HA synthase loses activity after HA chain elongation is complete (68), which could then be directly released for capture by CD44. It will be of interest to determine if the balance of HA secretion versus cell surface expression is controlled by the active form of CD44, whether the HA-binding state of CD44 is established intracellularly or at the cell surface, and at what stage of elongation the HA chain is delivered to the activated CD44. Answers to these questions may also provide clues as to the biochemical basis for the transition between inactive and HA-binding forms of CD44.
The ability of CD44 to bind HA is clearly not constitutive, requiring conformational or other structural alteration to engage this ligand. A variety of transcriptional and post-translational mechanisms have been suggested to explain the transition between active and inactive forms of CD44, including variant exon usage, phosphorylation, oligomerization, and particularly glycosylation (22,26,27,36,(75)(76)(77). Our PCR analysis (Fig. 6B) does not indicate alternative isoform usage by SVEC cells, and direct immunoprecipitations before and after TNF␣ (Fig. 6A) further do not support either isoform or glycosylation changes as detectable at the level of SDS gel resolution. This is similar to previous conclusions regarding activation of CD44 on stimulated normal T cell populations (25). However, a prior study stimulating EC with basic fibroblast growth factor in a model of angiogenesis did find alternative exon usage (78). This discrepancy may be due both to the mode of stimulation and the type of EC used, human umbilical vein endothelial cell, which do not up-regulate HA expression in response to proinflammatory cytokines (42). The structural basis for the activation of CD44 likely varies depending on the cell type examined and the physiologic circumstances, and the biochemical basis for activation on these endothelial cells will require further investigation.
A major distinction of the CD44/HA-initiated adhesion pathway is that HA is not an integral membrane protein as are the other EC receptors that participate in adhesion and for which the ability to resist shear forces is assumed to be due to the anchorage of the glycoproteins in the plasma membrane. Our results using an isolated recombinant form of CD44 indicate that the association of CD44 with HA is in itself sufficient to support rolling under laminar flow (Fig. 7), further supporting our other evidence that CD44 is in fact the relevant HA-binding protein on the surface of EC. The behavior of rolling under shear stress in capillary tubes coated with sCD44-Ig plus sHA is similar to CD44/HA-dependent shear responses we have observed on endothelial cell lines (23), with interactions essentially disappearing at about 3 dynes/cm 2 . The similarity of interactions on isolated ligands and on EC further suggests that the CD44/HA interaction on the endothelial surface is sufficient to account for much if not all of the observed rolling, and that this CD44-mediated rolling would most likely operate, as do selectins, at common postcapillary venular wall shear stresses (1-4 dynes/cm 2 ). These shear stress responses are similar to those reported for polymorphonucleocytes interacting with cytokine-stimulated human umbilical vein endothelial cell, where increasing wall shear stresses to 4 dynes/cm 2 also resulted in virtual loss of adhesive interactions (79,80). In addition, the shear stress responses presented closely duplicate those observed between the selectin family of adhesion molecules and their endothelial carbohydrate ligands (80 -86). Since random coating of a glass surface with a standard (hematopoietic) form of CD44 in conjunction with a high molecular weight form of HA was sufficient for rolling, the requirements for this interaction in terms of cell surface localization or HA deposition appear to have limited stringency. Thus, the avidity provided by CD44 interactions with large molecular weight HA is sufficient to provide significant shear resistance. The affinity of cartilage aggrecan for HA has been reported at a K d of 2 ϫ 10 Ϫ8 M (87), and when stabilized with link protein, dissociation was not detectable (88). For CD44 on the surface of 3T3 cells, the estimated K d value was higher at 1-2 ϫ 10 Ϫ9 M (65). The data provided here give the first demonstration of a mechanism whereby an endothelial adhesion receptor can serve to capture and support rolling in which direct transmembrane anchoring in the plasma membrane is not required.
In summary, the results presented in these studies establish that the structural basis for HA retention on the endothelial surface is CD44, and that it is in large part the regulation of the activated HA-binding form of CD44 that determines the level of surface HA expression. Moreover, interactions between CD44 and HA are sufficient to support rolling adhesions under physiologic laminar flow conditions, thereby defining a novel mechanism for glycosaminoglycan participation in pathways of extravasation.