Role of the Extracellular and Cytoplasmic Domains of CD44 in the Rolling Interaction of Lymphoid Cells with Hyaluronan under Physiologic Flow*

CD44 can function as an adhesion receptor that mediates leukocyte rolling on hyaluronan (HA). To study the contributions of different domains of the standard isoform of CD44 to cell rolling, a CD44-negative mouse T lymphoma AKR1 was transfected with wild type (WT) or mutated cDNA constructs. A parallel flow chamber was used to study the rolling behavior of CD44 transfectants on immobilized HA. For CD44WT transfectants, the fraction of cells that rolled and the rolling velocity was inversely proportional to the amount of cell surface CD44. When the cytoplasmic domain distal to Gly305 or sequences that serve as binding sites for cytoskeletal linker proteins, were deleted or replaced with foreign sequences, no significant changes in the rolling behavior of mutant cells, compared with the transfectant expressing CD44WT, were observed. Transfectants lacking 64 amino acids of the cytoplasmic tail distal to Cys295 adhered to HA but showed enhanced rolling at low shear forces. When 83 amino acids from the “non-conserved” membrane-proximal region of the CD44 extracellular domain were deleted, cells adhered firmly to the HA substrate and did not roll at any fluid shear force tested. Unlike wild type cells that exhibited a nearly homogenous distribution of CD44 on a smooth cell surface, cells expressing the non-conserved region deletion mutant accumulated CD44 in membrane protrusions. Disruption of the actin cytoskeleton with cytochalasin B precluded the formation of membrane protrusions, however, treated cells still adhered firmly to HA and did not roll. We conclude that interaction between the cytoplasmic domain of CD44 and the cytoskeleton is not required for cell rolling on immobilized ligand. The strong effect of deletion of the non-conserved region of the extracellular domain argues for a critical role of this region in CD44-dependent rolling and adhesion interactions with HA under flow.

During inflammatory reactions, recruitment of leukocytes from circulating blood to the site of inflammation takes place in postcapillary venules. This multistep process begins with es-tablishment of relatively weak, transient adhesive interactions between the leukocyte and the vessel wall that result in cell rolling along the endothelium under blood flow (1). The rolling interaction is a prerequisite for the generation of high strength intermolecular bonds and firm adhesion of the leukocyte to the vessel wall via activation-induced binding of integrins to their ligands (2). The sequence of cell rolling, firm adhesion, and transendothelial migration requires cooperation among a number of adhesion receptors (1,3,4).
CD44 is a transmembrane glycoprotein expressed in a wide variety of cell types (reviewed in Refs. 5 and 6). The primary physiologic ligand of CD44 is hyaluronan (hyaluronic acid, HA) 1 (7), a component of the extracellular matrix. CD44 is believed to be one of the key adhesion molecules that direct the traffic of activated T cells to inflamed tissues (8 -10), in part, by its ability to mediate cell rolling on HA (9,11). Cell surface expression of CD44 is up-regulated in activated lymphocytes, and cell activation also facilitates HA binding (12,13). HA expression is also up-regulated on the endothelial cell surface upon pro-inflammatory stimuli (14,15), thus providing a high local ligand density that may enhance the CD44-mediated interactions of lymphocytes with HA.
CD44 has been implicated in metastasis formation by different types of malignant tumors that express high amounts of this receptor (16 -20). During the hematogenous spread of some types of cancer, the initial arrest of cancer cells, and the migration of these cells on endothelial HA or within an HA-rich stroma, could potentially be mediated via CD44 (16,18,21,22). It has not been demonstrated, however, whether tumor cell extravasation requires a rolling interaction with endothelial HA.
The cytoplasmic tail of CD44 has sites mediating interaction with ezrin (23,24), and a putative ankyrin binding site has also been reported (25). Intracellular association of CD44 with these cytoskeletal linker proteins could function in post-ligand binding events mediated by CD44. The cytoplasmic domain of CD44 may be phosphorylated at one or more serine residues (26,27), and phosphorylation has been implicated in mediating cell migration on HA (27) and in the interaction of CD44 with ezrin (26).
Whether specific sequence within the CD44 cytoplasmic domain or other regions of the standard form of CD44 are re-quired for or modulate cell rolling on an HA substrate remains unclear. Here we have used AKR1 lymphoma cells transfected with constructs encoding either WT or mutated mouse CD44 molecules to examine whether mutation of the CD44 cytoplasmic, transmembrane, or extracellular domains would alter the rolling behavior of transfected cells on an HA-coated substrate.
Mouse CD44 cDNA Construction and Cell Transfection-cDNA constructs, encoding wild type and mutated forms of the standard isoform (also known as hemopoietic form, CD44H) of mouse CD44, were generated using PCR and site-directed mutagenesis as described previously (29, 34 -36). The sequences of the PCR products were verified on an automated DNA sequencer, and the constructs were subcloned into expression vectors pRC/RSV, pH␤APr-1, or p304SR␣ (34 -37). Schematic structures of CD44 proteins, encoded by these constructs, are shown in Fig. 1.
Flow Cytometry and Confocal Microscopy-AKR1 transfectants expressing WT and mutated CD44 molecules were assayed for cell surface CD44 density and for binding of FL-HA from solution. CD44 was detected using biotinylated IRAWB14 mAb (shown in Fig. 2A) or IM7 (not shown), both mAb giving essentially identical results for each transfectant. Following incubation with streptavidin-PE and washing, fluorescence was analyzed using a FACScan flow cytometer and CellQuest software (BD Biosciences, San Jose, CA). For CD44 immunostaining, cells (with or without pretreatment with 40 M cytochalasin B at 37°C for 3 h) were incubated on HA-coated culture dishes (see below) for 10 min. Following washing and fixation with 2% formaldehyde, cells that adhered to immobilized HA were incubated with Alexa Fluor 594-conjugated IM7 mAb. Because the epitope recognized by IM7 is outside of the HA binding domain of CD44 (30), binding of this antibody to the epitope is not expected to be affected by ligand occupancy at the membrane-substrate interface. For actin staining, HAadherent, formaldehyde-fixed cells were permeabilized with 0.2% Triton X-100 and incubated with TRITC-phalloidin (32). The distribution of CD44 or filamentous actin was examined using an Eclipse TE200 confocal microscope (Nikon, Garden City, NJ) and Metavue imaging software (Universal Imaging, West Chester, PA). For three-dimensional reconstruction of images, Z-series of 300-nm-thick slices were used.
Cell Rolling on HA under Fluid Flow in a Parallel Flow Chamber-The experimental conditions for assessment of cell rolling on immobilized HA have been described in detail before (37). In brief, HA (1 mg/ml) was adsorbed onto the bottom of 60-mm plastic Petri dishes that constituted the bottom of a parallel plate flow chamber (GlycoTech, Rockville, MD). The chamber was perfused with the cell suspension using a programmable syringe pump (Harvard Apparatus, Holliston, MA) in withdrawal mode. Cells were accumulated at a shear force of 0.25 dyn/cm 2 , followed by exposure to increasing levels of fluid shear (0.25, 0.5, 1, 2, 4, and 8 dyn/cm 2 , each for 2 min). Cell movement was recorded by streamline acquisition of images using a digital camera (RS Photometrics, Trenton, NJ) attached to a Nikon Diaphot inverted phase-contrast microscope. To assess the strength of cell adhesion, the number of adherent cells was determined at the end of each 2-min flow period and expressed as a percentage of the number of adherent cells at the starting shear force (0.25 dyn/cm 2 ). The ratio of rolling cells was expressed as the percentage of adherent cells that moved parallel with the fluid flow on the HA substrate (37). The rolling velocity (microns/s) was also determined at each level of shear force. Images were collected and analyzed using a Metaview image analysis system (Universal Imaging). At least three fields of microscopic view and, for velocity measurements, at least 100 cells were analyzed at each shear force.
In initial experiments, both mammalian (umbilical cord) and avian (Healon, prepared from rooster comb) HA was used to coat culture dishes that constituted the bottom of the flow chamber. Although both HA preparations supported rolling, the CD44 transfectants attached more evenly to, and rolled more smoothly on, the surfaces coated with umbilical cord HA than on those coated with Healon. For this study, therefore, HA of mammalian origin was chosen as a substrate for the cell rolling experiments. To rule out the possibility that the rolling interactions were mediated by contaminating proteins or chondroitin sulfate, which could be present in this preparation, specificity experiments were conducted. Attachment and subsequent rolling of CD44 transfectants on immobilized mammalian HA were completely abolished by digestion of the substrate with Streptomyces hyaluronidase (10 units/ml at 37°C for 30 min) and were inhibited in a dose-dependent manner by a soluble HA competitor (either mammalian HA or Healon), but not by chondroitin sulfate. For example, at a concentration of 200 g/ml, Healon reduced the fraction of cells that bound to immobilized mammalian HA to 15%, whereas 99% of the cells attached to, and rolled on, the same substrate when chondroitin sulfate was present at 200 g/ml. These results indicate that the rolling interactions are dependent on HA and not on contaminants in the immobilized substrate.
Normalization of CD44 Expression-The capacity of cells to bind HA is dependent on the amount of CD44 on the cell surface (34). Because a change in the rolling properties of the cells transfected with mutant CD44 constructs could be caused by either the specific mutation or a change in cell surface CD44 density, the amount of CD44 expressed by each mutant cell line had to be "normalized" to CD44 expression in a wild type transfectant. This normalization was done for each set of experiments by pairing each mutant with a wild type counterpart that expressed a similar amount of CD44. Two strategies were employed to achieve this: 1) AKR1 cells transfected with mutant CD44 constructs were cultured on immobilized HA for several days prior to the rolling assays. Regular removal of non-adherent cells from the HA-coated dishes resulted in enrichment for a homogenous population of HAbinding cells with relatively high densities of CD44, comparable with CD44 expression in CD44WT transfectants. 2) Using the limiting dilution method, several subclones of AKR1/CD44WT cells were generated, each with a different amount of CD44 on the cell surface (see Fig. 2). Subclones with lower amounts of CD44WT served as "controls" for mutant lines in which high CD44 expression could not be achieved.
For flow chamber experiments, transfectants were characterized for CD44 expression, and the results of rolling assays for each mutant line were compared with those of an AKR1/CD44WT clone with similar CD44 expression. All experiments were repeated at least three times under similar conditions. Although differences between mutant and wild type lines were consistently observed, the measured values showed a relatively large variation from experiment to experiment that could be ascribed to fluctuations in CD44 expression of wild type and mutant transfectants. Some measurements also had a statistically non-normal (left-skewed) distribution. Therefore, several statistical tests, including the paired Student's t test, repeated measures analysis of variance and the non-parametric Friedman test (SPSS, Chicago, IL), were employed to determine significant differences between mutant and wild type cells in their rolling parameters. Fig. 1 shows the schematic structure of WT CD44 (standard isoform) and seven CD44 mutants. "ERM" below the first boxed area in the cytoplasmic (CY) domain of the WT schematic in Fig. 1 indicates the location of sequences (Arg 292 -Lys 300 ) required for binding ERM (ezrin, radixin, moesin) proteins (23,24), and "Ank" below the second box indicates the location of sequences (Asn 304 -Leu 318 ) reported to mediate ankyrin binding (25). P after Ank indicates the location of Ser 325 , which is constitutively phosphorylated and whose phosphorylation is required for CD44mediated cell migration on HA (27). The numbering of amino acids is according to that of a previous study (38).

Structure and HA Binding Properties of WT and Mutant CD44 Expressed in AKR1 Lymphoma Cells-
Several deletion and substitution mutants (34, 36) were used to examine the importance of specific regions of the CD44 cytoplasmic domain in mediating cell rolling on HA. CD44G305 contains a cytoplasmic domain truncated at Gly 305 and lacks the putative ankyrin binding motif (25). The CD44⌬CY ("tailless") mutant lacks all amino acid residues distal to Cys 295 , resulting in loss of the ankyrin binding sequence together with one of the two clusters of positively charged amino acids reported to be necessary for ezrin (ERM) binding (24). In CD44⌬292-301, the entire ezrin binding site (24) has been deleted. CD44SG5N has 5 Ser-Gly repeats inserted between the transmembrane and cytoplasmic domains, followed by the 6 membrane-proximal residues of the cytoplasmic domain (amino acids 290 -295). In CD44␤5TMϩCY the entire transmembrane and cytoplasmic regions of CD44 have been replaced by homologous domains of the human ␤ 5 integrin. In summary, the ERM binding region is absent or disrupted in all of the cytoplasmic domain mutants except CD44G305, the putative ankyrin binding region is absent in all of these transfectants except CD44⌬292-301, and all cytoplasmic tail mutants, except CD44⌬292-301, lack Ser 325 . Two mutants were used to examine the importance of features of the CD44 extracellular domain. CD44mArg41 has the Arg 41 residue, which is located in the ligand binding domain and is critical for HA binding (40,41), mutated to Ala. In CD44⌬161-244, 83 amino acids of the non-conserved membrane-proximal region of the CD44 ectodomain have been deleted (35).
As described under "Experimental Procedures," for each series of rolling assays, AKR1/CD44WT cells were subcloned, and a WT clone was chosen as a control for each mutant line on the basis of similar amounts of CD44 expression. Binding of FL-HA was determined by flow cytometry, and the transfectants expressing mutant CD44 were compared with WT controls matched for CD44 expression (data not shown). With the exception of AKR1/CD44⌬CY, all of the cell lines expressing CD44 with mutations in the cytoplasmic domain bound FL-HA similarly to WT controls, consistent with previous reports (34,36). As reported previously (29), the line expressing ⌬CY (tailless) CD44, despite fairly high CD44 density on the cell surface, bound FL-HA very poorly (data not shown). Cells expressing CD44⌬161-244 with a deletion in the ectodomain bound FL-HA similarly to CD44WT controls, as reported before (35). The mArg41 mutant did not bind HA at all, a finding that is also consistent with previous reports (40,41). Although binding of FL-HA to the tailless transfectant increased dramatically in the presence of "inducing" mAb IRAWB14 (29,42), no ligand binding to the mArg41 mutant could be induced by IRAWB14 treatment (results not shown).
Relationship between Cell Surface CD44 Density and Rolling Behavior of AKR1/CD44WT Cells-The cellular avidity of AKR1/CD44WT transfectants for soluble HA increases with increasing expression of the receptor on the cell surface (34). To determine whether CD44 density also affects rolling behavior, we examined the rolling of CD44 WT transfectants expressing different amounts of CD44. Fig. 2A shows CD44 expression and FL-HA binding (mean fluorescence intensity at saturating concentrations of mAb and FL-HA, respectively) for five AKR1/ CD44WT subclones with different levels of CD44 expression. As shown previously (34), the greater the amount of CD44 on the surface, the greater is the capacity of the cells to bind FL-HA. When these clones were exposed to increasing fluid shear forces, the fraction of AKR1/CD44WT cells that remained adherent to HA (Fig. 2B) was roughly proportional to the amount of CD44 on the cell surface. At any given shear force, fewer cells with high than low CD44 density detached from the substrate. The percentage of adherent cells that rolled on HA (Fig. 2C) and the rolling velocity (Fig. 2D) were inversely related to the level of receptor expression. Cells expressing the greatest amounts of CD44 only began to roll on HA at higher shear forces (at or above 1 dyn/cm 2 ) and exhibited lower rolling velocities.
Rolling of AKR1 CD44 Cytoplasmic/Transmembrane Domain Deletion/Replacement Mutants on HA-Cells expressing the CD44G305 truncation mutant rolled in a manner similar to CD44WT transfectants at all shear forces tested (0.25-8 dyn/ cm 2 ) (Fig. 3A). Two transfectants expressing CD44 with a cytoplasmic tail truncated more distally (Thr 345 and Ser 323 (34)) also displayed rolling kinetics similar to cells expressing WT CD44 (results not shown).
Although AKR1/CD44⌬CY cells bound soluble FL-HA very  (38); exon nomenclature is as in Ref. 39). ERM below the WT cytoplasmic domain indicates the location of clusters of basic aa residues (Arg 292 -Lys 300 ) required for ezrin binding (24), and Ank indicates the location of the putative ankyrin binding region (Asn 304 -Leu 318 ) (25). P denotes the position of the constitutively phosphorylated Ser 325 residue (27). Compared with CD44WT, cytoplasmic domain truncation mutants ⌬CY and G305 lack all but 6 and 16 aa residues, respectively, of the 72-aa-long cytoplasmic tail. ⌬292-301 lacks the binding site for the cytoskeletal linker protein ezrin. SG5N has a tandem insertion of 5 Ser-Gly dimers following the transmembrane domain, followed by the 6 membrane-proximal aa of the cytoplasmic tail. In the ␤5TMϩCY construct both the transmembrane and cytoplasmic domains of mouse CD44 are replaced by the homologous domains of the human ␤ 5 integrin. The mArg41 mutant has a single aa Arg 41 within the HA binding domain mutated to Ala ("X"). The region of the CD44 extracellular domain (between aa 161 and 244) that is relatively non-conserved among species has been removed in ⌬161-244 by a loop-out deletion between nucleotides 540 and 790 of the CD44 cDNA (35). poorly (Refs. 29, 34, and 36 and data not shown), cells with WT and tailless CD44 rolled with similar kinetics on immobilized HA (Fig. 3B), except for one consistent difference. At low shear forces, the fraction of AKR1/CD44⌬CY cells that rolled was greater than the fraction of AKR1/CD44WT cells that rolled (Fig. 3B, middle panel), however, the rolling velocity of the mutant at any given shear force was not appreciably greater (Fig. 3B, bottom panel).
For both the AKR1/CD44⌬292-301 mutant with an internal ("loop-out") deletion that encompasses the entire sequence of the ezrin binding motif, and the AKR1/CD44SG5N mutant, the interaction between CD44 and ezrin should be impaired (24). CD44SG5N lacks, in addition, the putative ankyrin-binding motif (25). The CD44␤5TMϩCY chimera, with all of the CD44 transmembrane and cytoplasmic domain sequences replaced by the corresponding sequences of human ␤ 5 integrin, should not be able to interact with intracellular linker, adaptor, or signaling molecules in a CD44-specific manner. This chimera lacks all CD44-specific phosphorylation sites (26) as well as Cys 286 , which has been reported to have the potential to "bypass" a requirement for the CD44 cytoplasmic domain in binding HA by mediating CD44 dimerization via the transmembrane region (43,44). When matched with AKR1/CD44WT cells showing the same level of receptor expression, however, transfectants for all three of these mutants showed rolling behavior that was indistinguishable from that of the CD44WT transfectants (Fig. 4).
Rolling of AKR1 CD44 Extracellular Domain Replacement/ Deletion Mutants on HA-AKR1/CD44mArg41 cells expressing CD44 with the Arg 41 /Ala mutation did not bind FL-HA and were also unable to attach to, and roll on, immobilized HA (results not shown). The membrane-proximal region of the CD44 extracellular domain, which is poorly conserved across species, has no established function in ligand binding (35). AKR1/CD44⌬161-244 cells, expressing CD44 in which 83 amino acids of this non-conserved region were deleted, did not roll on immobilized HA like the control CD44WT transfectant. Instead, AKR1 cells with the CD44⌬161-244 mutation remained firmly adherent to the substrate under fluid flow. These mutant cells failed to detach from or roll on the HAcoated surface under a wide range of shear forces (Fig. 5A).
To determine whether AKR1/CD44⌬161-244 cells lose firm adhesion to HA at lower densities of the receptor on the cell surface, we generated subclones of the mutant line that expressed varying amounts of CD44 and compared their adhesion/rolling behavior on HA. When exposed to a shear force of 2 dyn/cm 2 , the fraction of firmly adherent cells increased and the fraction of rolling cells decreased roughly in proportion to receptor density for the AKR1/CD44WT clones, whereas the majority of AKR1/CD44⌬161-244 cells remained firmly adherent and failed to roll on HA at all levels of CD44 expression (Fig.  5B). We then asked whether the extreme "stickiness" of AKR1/ CD44⌬161-244 cells was a consequence of secondary mechanisms that enhanced the interaction of CD44 with the actin cytoskeleton. However, similarly to AKR1/CD44WT cells, treatment of the AKR1/CD44⌬161-244 transfectant with the cytoskeleton-disrupting drug cytochalasin B for 3 h before the rolling assay did not result in any change in the adhesion or rolling behavior of the cells (Fig. 5C), although, as discussed below, this treatment resulted in the collapse of the actin cytoskeleton.
Differential tors (including CD44) for ligand depends to a considerable extent on their distribution and presentation on the cell surface (45)(46)(47). Both the cytoplasmic and extracellular domains of adhesion molecules have been found to play a role in determining membrane localization (45,48,49). Drastic truncation of the CD44 cytoplasmic tail in AKR1/CD44⌬CY and deletion of a large region of the extracellular domain in AKR1/CD44⌬161-244 could affect the membrane topology of the receptor, thus contributing to the HA binding and rolling properties of these mutants. We used confocal microscopy to examine and compare the cell surface localization of CD44 (visualized by immunofluorescence) in the WT control with that in the tailless and ⌬161-244 mutants. In AKR1/ CD44WT cells that attached to HA, CD44 was present in a small circular area in the focal plane of membrane contact with immobilized ligand (Fig. 6A), was evenly distributed along the membrane as seen from the middle section of the cell (Fig. 6B), and was nearly homogenously dispersed on the entire cell surface (Fig. 6, C and D). In AKR1/CD44⌬CY cells (Fig. 6, E-H), the membrane attachment site was larger than in the CD44WT control (Fig. 6, compare A with E, and D with H) and contained CD44 dispersed in small clusters (Fig. 6E). Also, CD44 showed a patchy distribution over the entire cell surface (Fig. 6, F-H). HA-adherent AKR1/CD44⌬161-244 cells demonstrated a fundamentally different distribution of CD44 (Fig. 6, I-L). The area of membrane attachment in this mutant was much larger than the contact area in the CD44WT (Fig. 6, compare I with A) and markedly larger than that in the tailless transfectant (Fig. 6, compare I with E). This difference was also obvious from the side view of the cells (Fig. 6, compare L with D and H). Furthermore, unlike in the other two transfectants where CD44 showed a homogenous or patchy distribution within the planar cell membrane, in the AKR1/CD44⌬161-244 mutant CD44 was organized into thin filamentous structures reminiscent of microvilli, pro-truding from virtually the entire cell surface (Fig. 6, I-L). Staining of the actin cytoskeleton with TRITC-phalloidin revealed the presence of actin filaments in the membrane projections of AKR1/ CD44⌬161-244 (Fig. 7, C and D), confirming that they were, indeed, microvilli. These microvilli were not present in the wild type transfectant (Fig. 7, A and B). Following incubation with cytochalasin B, the cortical cytoskeleton collapsed (Fig. 7, F and  H), and the microvilli disappeared from the membrane contact site (Fig. 7G) and from the cell surface (Fig. 7H) in the AKR1/ CD44⌬161-244 mutant. CD44 showed an even distribution in the planar plasma membrane of both the wild type transfectant (Fig. 7, I and J) and the CD44⌬161-244 mutant (Fig. 7, K and L) following treatment with cytochalasin B.

DISCUSSION
The roles of the cytoplasmic, transmembrane, and extracellular domains of CD44 in HA binding and membrane localization have been studied by several laboratories using site-directed mutagenesis and domain swapping techniques (24, 25, 29, 34 -36, 40, 41, 43, 45, 50). In this study, we utilized a number of well characterized mutants of the standard form of mouse CD44, expressed in AKR1 lymphoma cells (which, when transfected with CD44WT, bind HA constitutively), to ask how specific mutations within these functional domains influence the ability of CD44 to mediate cell rolling on HA under flow.
The amount of CD44 expressed on the cell surface is one of the most important factors regulating ligand binding function (34,36,37). Here we have demonstrated that cell surface receptor density is also a critical factor for CD44-dependent cell rolling on HA. As shown in Fig. 2, each of the rolling parameters (adhesion strength, rolling fraction, and rolling velocity) are affected by the net amount of CD44 expressed. At low levels of receptor expression (where the cells are virtually unable to FIG. 3. HA-mediated rolling of AKR1 CD44 cytoplasmic domain truncation mutants G305 and ⌬CY. A, HA-mediated rolling of AKR1/CD44G305 mutants versus AKR1/CD44WT cells. B, rolling of AKR1/CD44⌬CY cells in comparison with the control CD44WT transfectant. The parameters measured are the same as described in Fig. 2. Starting net cell numbers in the experiment shown here were: 254/field for G305, 248/field for the expression-matched WT control; 281/ field for ⌬CY, 269/field for the paired WT control. Statistical analysis of data from three experiments for each pair did not reveal significant differences between the mutant and WT lines in rolling behavior, except for AKR1/CD44⌬CY, where the difference from wild type in the percentage of cells rolling at or below 1 dyn/cm 2 shear force approached significance (p ϭ 0.083; Friedman test).
bind HA from solution), the capture of AKR1/CD44WT cells on immobilized HA is poor. Almost all cells that remain attached, however, engage in relatively high speed rolling interactions with the substrate. AKR1 cells with high CD44 densities adhere strongly to HA under flow, and they begin to roll with low velocity only at or above 1 dyn/cm 2 (Fig. 2). Thus, depending on the density on the cell surface, standard CD44 (the most abundant isoform) is capable of mediating low, moderate, and high strength adhesion to HA in a cellular environment that is permissive for constitutive ligand binding.
The versatility of standard CD44 in its ability to support either rolling or strong cell adhesion has not been fully appreciated, although a large body of literature suggests that this molecule plays important roles in a variety of pathologic processes (12, 18, 20 -22, 31, 51-54). As shown in Fig. 2, the density of cell surface CD44 is an important factor regulating HA-dependent cell adhesion and rolling in vitro. It seems likely that CD44 density is also important in regulating these processes in vivo. Activated leukocytes may represent the cells, which, at moderate levels of CD44 expression, can participate in rolling interactions with endothelial HA prior to extravasation into injured tissues in acute inflammation and inflammatory disease (8,12,14,20). Metastasis formation by cancer cells, which often express CD44 at high densities, represents a pathologic process in which CD44 is thought to play a detrimental role. By mediating attachment (not necessarily rolling) of tumor cells to HA on endothelium, CD44 can facilitate tumor cell extravasation, thus contributing to the hematogenous dissemination of several types of cancer (18, 20 -22, 53).
The experiments summarized in Figs. 3 and 4 demonstrate that mutations in the cytoplasmic domain have little or no effect on CD44-mediated adhesion and rolling interactions with immobilized HA. AKR1 cells expressing the CD44 truncation mutant Gly 305 , the internal deletion mutant ⌬292-301, the CD44SG5N chimera, and the CD44␤5TMϩCY chimera all showed rolling interactions with immobilized ligand that were similar to those of AKR1 cells transfected with wild type CD44. These results indicate that CD44-mediated rolling of AKR1 cells on HA is not regulated via associations between the cytoplasmic tail and the cytoskeletal linker proteins ezrin or ankyrin and that specific phosphorylated serine residues (26,27) are unlikely to be involved in mediating rolling. It is important to note that, in contrast to cell attachment or rolling under flow, there is evidence that cytoskeletal signaling and coordination are required for HA-dependent cell migration under static conditions (16,26,27,55,56).
In this study, we found only three mutations within the CD44 molecule that resulted in significant deviations from the normal (WT-like) rolling behavior of AKR1 transfectants on immobilized ligand: 1) truncation of the cytoplasmic tail to 6 amino acids (⌬CY), 2) substitution of Arg 41 within the HA binding domain with Ala (mArg41), and 3) deletion of 83 amino acids from the non-conserved membrane-proximal region of the extracellular domain (⌬161-244). Because the mArg41 mutation affects a critical amino acid in the Link module (40), the failure of AKR1/CD44mArg41 cells to bind to immobilized HA (and thus to roll) was expected.
For the CD44⌬CY mutant, a greater fraction of AKR1 cells transfected with this construct rolled on HA compared with transfectants expressing similar amounts of CD44WT. In this respect, the tailless mutant behaved like a WT transfectant with a lower level of receptor expression, consistent with the known properties of this mutant, extremely inefficient binding of soluble HA and a reduced, but still reasonably efficient, binding to immobilized HA (16,29,34,36). An altered distribution of tailless, relative to wild type, CD44 on the cell surface might change the rolling behavior on HA. In this connection, confocal microscopy of HA-adherent cells revealed that the distribution of CD44 in the ⌬CY mutant was, indeed, different from that in cells expressing CD44WT (Fig. 6). AKR1/CD44WT cells showed a nearly homogenous distribution of CD44 on the cell membrane, including the contact site with the immobilized substrate. In contrast, AKR1/CD44⌬CY cells demonstrated a patchy distribution of the receptor over the cell body as well as in the substrate contact site. It is possible that cell contact with immobilized HA, but not soluble HA, stabilizes weak receptorligand interactions by promoting clustering of tailless CD44. This receptor clustering could increase cellular avidity for ligand sufficiently to enable adhesion to and rolling on the immobilized substrate.
The ⌬161-244 deletion (35) does not involve the Link module, which contains residues responsible for HA recognition (6,40,41). Instead, this mutation removes a portion of the extracellular domain that is relatively non-conserved among species, including the site where the products of alternatively spliced exons can be inserted, giving rise to variant isoforms of CD44 in certain types of cells (6). Deletion of this region results in no significant change in soluble HA binding (Ref. 35 and data not shown). Unexpectedly, AKR1/CD44⌬161-244 cells adhered to HA much more strongly than the corresponding AKR1/ CD44WT transfectant, because the mutant cells did not detach from or roll on the HA substrate at any level of shear force tested (Fig. 5A).
CD44 has been shown to localize to the planar cell membrane (as seen in AKR1/CD44WT) and, unlike L-selectin or ␣ 4 integrin, is excluded from the tips of the microvilli in leukocytes that possess such protrusions (45,49). Adhesion receptors such as L-selectin and ␣ 4 integrins that localize to the tips of microvilli can initiate attachment to and rolling on immobilized ligands under flow, whereas several other adhesion receptors, localized to the planar cell body (with the notable exception of CD44), are unable to do so (48,49). Also, concentration of L-selectin on the tips of microvilli has been shown to enhance the resistance of leukocytes to shear stress upon initial attachment to native ligand under flow (45,48). In AKR1/CD44⌬161-244 cells, whereas CD44 was also present on the cell body, a large proportion of receptors was localized to microvillus-like membrane protrusions (Fig. 6). The questions arise as to what mechanism is responsible for directing CD44 (with a shortened ectodomain) to microvilli, and whether this particular distribution of CD44 molecules contributes to the increased strength of cell adhesion to HA, characteristic of this mutant. Treatment with cytochalasin B resulted in a WT-like homogenous distribution of CD44⌬161-244 on the cell surface (Fig. 7), yet, this treatment did not result in weakening of the adhesive interactions of mutant cells with HA (Fig. 5). These observations suggest that, although cytoskeletal integrity is required for the formation of microvilli in this particular transfectant (Fig. 7), neither the microvilli nor the actin cytoskeleton appear to contribute to the stability of the bond between CD44⌬161-244 and immobilized HA under shear stress.
The extracellular domain of CD44 is substituted with carbo- hydrate side chains, often including sulfated glycosaminoglycans (GAGs) (20), that may confer negative electrostatic charge and rigidity to the receptor. Because the shortened core protein of CD44⌬161-244 lacks a number of these carbohydrate attachment sites, the intermolecular repulsion afforded by the GAG chains (57) might be reduced in the mutant. It is tempting to speculate that poorly glycosylated CD44⌬161-244 molecules would readily undergo clustering with each other or with other adhesion receptors at the ligand contact site, thus increasing the strength of adhesion. However, treatment of AKR1/ CD44WT cells with sodium chlorate (an inhibitor of sulfation (58)), chondroitinase ABC (which removes the GAG chondroitin sulfate), or p-nitrophenyl-␤-D-xylopyranoside (which prevents addition of GAG side chains to the CD44 core protein (59)) under conditions that have been reported to affect soluble HA binding to CD44 (59), failed to increase adhesion of wild type cells to immobilized HA. 2 In addition, although AKR1 cells express significant amounts of the leukocyte integrin ␣ 4 ␤ 1 (VLA-4), which could potentially amplify the strength of CD44-initiated cell adhesion (60 -62), we found no evidence of coclustering or other type of direct interaction between CD44⌬161-244 (or CD44WT) and the VLA-4 adhesion molecule in the transfected cells upon exposure to immobilized HA. 3 Although enhanced receptor clustering is not definitely ruled out by these experiments, the results suggest that the increased stability of binding between the mutant receptor and immobilized ligand could reflect a conformational change in the ectodomain of CD44, caused by deletion of 83 amino acids. Changes in ectodomain length, frequently seen under pathologic conditions, can occur by insertion of variant sequences (by inclusion of one or more of ten alternatively spliced exons between exons 5 and 16) in the region deleted in CD44⌬161-244 (35,63). We did not investigate whether insertion of variant sequences in the non-conserved region of the extracellular domain would have the same or opposite effect on cell rolling as the deletion of this region. It is known that modification in CD44 structure by insertion of variant exon products within the non-conserved region can have either negative or positive 2 I. Gá l, T. T. Glant, and K. Mikecz, unpublished data. 3 I. Gá l and K. Mikecz, unpublished data.
FIG. 6. Membrane distribution of CD44 on AKR1 cells expressing CD44WT, CD44⌬CY, and CD44⌬161-244 during adhesion to HA. To visualize surface distribution of CD44 in cells that attached to immobilized HA, the substrate-bound cells were fixed with formaldehyde and then stained with Alexa Fluor 594-conjugated IM7 mAb. Using confocal microscopy, the cells were optically sectioned A, E, and I, at the plane of membrane attachment, and B, F, and J, in a position midway between the bottom and the top (middle sections). Three-dimensional view of the cells in C, G, and K, from the bottom, and in D, H, and L, from the side, was reconstructed from Z-series of optical sections. One typical cell of each CD44 transfectant is shown. influence on receptor-ligand interactions (18,35,(63)(64)(65). Our results extend these findings by demonstrating that the nonconserved region of CD44 ectodomain is critically involved in the regulation of CD44-mediated cell rolling under flow.