Involvement of CD44 in cytoskeleton rearrangement and raft reorganization in T cells

T cells continuously change from a mobile to a sessile state. They leave the blood either to home into lymph nodes or to migrate into sites of infection (Brown, 1997; Butcher and Picker, 1996). Furthermore, T cell activation as well as effector functions are initiated by intimate contact between T cells and antigen presenting cells (APC) and target cells, respectively (Geiger et al., 1982; Ryser et al., 1982). For T cell maturation, too, an intimate contact with stromal cells is important (Kobayashi et al., 1994). The changes between the mobile and the sessile state are accompanied by adhesion, flattening and spreading (Reinhold et al., 1999; Rosenman et al., 1993). These processes require at the initial state the engagement of adhesion molecules (Dustin and Springer, 1989; Kaga et al., 1998), which trigger signal transducing molecules involved in the reorganization of the cytoskeleton (Shaw and Dustin, 1997; Wülfing and Davis, 1998). The rearranged cytoskeleton may then serve as a scaffold for cascades of signal transducing molecules inducing gene activation and transcription (Penninger and Crabtree, 1999).

The cytoskeleton is a highly dynamic structure that reorganizes when cells respond to extracellular stimuli by division and/or changes in shape or activity. The key player for the cellular shape is the actin cytoskeleton (Howard and Watts, 1994). Changes in plasma membrane morphology are due to actin polymerization and rearrangement of the underlying cortical actin filaments. Signaling molecules involved in polymerization are tyrosine kinases, small GTPases of the Rho subfamily and membrane phospholipids (Harder and Simons, 1999; Janmey, 1998; Mackay and Hall, 1998). As explored in fibroblasts, activation of the Rho subfamily members, Rho, Rac and Cdc42, trigger formation of actin stress fibers and focal adhesion complexes (Rho), actin polymerization at the plasma membrane to produce lamellipodia and membrane ruffles (Rac), filopodial protrusions and microspikes (Cdc42) (Kozma et al., 1995; Ridley et al., 1992). Target molecules of the Rho subfamily include WASP and p95vav (Holsinger et al., 1998; Miki et al., 1996), the latter being found exclusively in hematopoietic cells. Notably, the involvement of the actin cytoskeleton in the process of T cell activation could be convincingly demonstrated in vav and WASP deficient T cells, which poorly respond to an antigenic stimulus (Holsinger et al., 1998; Kong et al., 1998; Snapper et al., 1998). The same accounts for T cells treated with cytochalasin, an agent which disrupts the actin cytoskeleton (Holsinger et al., 1998).

The adhesion molecule CD44 has been implicated in all processes associated with changes in T cell morphology. Originally it has been described as an adhesion molecule mediating lymphocyte homing (Jalkanen et al., 1988). Meanwhile it is known that CD44 is also involved in the extravasation of activated lymphocytes (DeGrendele et al., 1997), in lytic effector functions (Galandrini et al., 1994; Matsumoto et al., 1998) as well as in T cell activation. Engagement of CD44 has been described to modulate T cell proliferation and cytokine production (Galluzzo et al., 1995; Huet et al., 1989; Sommer et al., 1995; Toyama-Sorimachi et al., 1995), has been implicated in enhanced binding of dendritic cells to T cells (St John et al., 1990) and has been suggested to function as a costimulus in allogeneic and mitogenic T cell responses by binding to the chondroitin sulfate form of the invariant chain (Naujokas et al., 1993). Cross-linking of CD44 has been noted to prolong the Ca²⁺ influx (Galandrini et al., 1994; Galluzzo et al., 1995; Dianzani et al., 1999). Furthermore, it has been described that CD44 associates with src-family kinases, an interaction which occurs in GEMs (Ilangumaran et al., 1998; Oliferenko et al., 1999; Taher et al., 1996). In this context it should be remembered that CD44 is known to interact with several cytoskeletal proteins like actin, ankyrin and members of the ERM family (Bourguignon and Jin, 1995; Tsukita et al., 1994). So far, functional importance of the interaction with the cytoskeleton only has been reported for HA binding, which depends on the association with ankyrin (Bourguignon et al., 1993) and for tumor cell migration and invasion (Bourguignon et al., 1998). We recently have described that CD44 exerts costimulatory function under physiological conditions of antigenic T cell activation (Föger et al., 2000). We could show that cross-linking of CD44 significantly enhances signal transduction via the TCR/CD3 complex. We interpreted our findings in the sense that the costimulatory function of CD44 relied on its cooperativity with the TCR. Since CD44 was found to be constitutively associated with p56^lck and p59^fyn, it became tempting to speculate that it may be the recruitment of phosphokinases which significantly lowered the threshold for initiation of signal transduction via the TCR. Here we show that, indeed, cross-linking of CD44 leads to rearrangement of the cytoskeleton, cap formation and the recruitment of the CD44-associated phosphokinases into lipid rafts.

A murine CD4⁺ T cell clone, IP12-7, specific for influenza virus hemagglutinin (HA317-329) has kindly been provided by Dr E. Rajnavolgyi, L. Eotvos University, God, Hungary (Rajnavolgyi et al., 1994). The clone was maintained in RPMI 1640, supplemented with 10% FCS and antibiotics. The T helper line D10.G4.1 has been obtained from the American Type Culture Collection (TIB-224) and has been maintained according to the suppliers suggestion. Where indicated IP12-7 cells were transiently transfected with myc epitope-tagged dominant negative mutant forms of RhoA or Rac1 by electroporation. The eucaryotic expression vector encoding myc epitope tagged dominant negative mutant forms of RhoA (myc-RhoN19) and Rac1 (myc-RacN17) were kindly provided by P. Boquet, INSERM U452, Nice, France (Qui et al., 1995). Aliquots of 5×10⁶ IP12-7 cells were resuspended in 400 μl ice-cold serum free RPMI 1640 and 40 μg of DNA was added. Electroporation was performed using an EUROGENTEC electroporation apparatus at a setting of 250 V and 900 μF. Expression of negative mutant forms was controlled by staining with a myc-specific antibody. In one experiment lymph node cells of BALB/c mice have been used. T cells were enriched by passage over a nylon wool column and were kept in culture overnight in the presence of 10 U/ml IL-2.

IM-7 (anti-CD44, rIgG2b) and KM81 (anti-CD44, rIgG2a) have been obtained from the American Type Culture Collection. 9E10 (anti-myc, mIgG1) has been obtained for the European Collection of Animal Cell Cultures. Antibodies were purified from culture supernatants by passage over Protein G-Sepharose. The eluates were concentrated and filter sterilized. The following monoclonal antibodies were obtained commercially: Anti-phopsphotyrosine, anti-fyn, anti-Rac1, anti-RhoA, anti-Cdc42 (all Santa Cruz Biotechnology), FITC-conjugated goat anti-rat IgG2b (Bethyl Laboratories), horseradish peroxidase (HRP)-conjugated sheep anti-rat IgG (Amersham Buchler), Cy2-and HRP-conjugated donkey anti-mouse IgG, Texas red-and HRP-conjugated donkey anti-rabbit IgG (Dianova).

CD44 receptor caps were induced by incubating 1×10⁵ IP12-7 cells for 15 minutes at 37°C with IM7 (10 μg/ml). After washing, the primary antibody was cross-linked using FITC-conjugated goat anti-rat IgG2b (10 μg/ml, 30 minutes, 37°C). Cap formation was stopped by 2 washes in ice-cold PBS.

Cell spreading was induced by layering IP12-7 cells (1×10⁵ ml in RPMI 1640, 10% fetal calf serum) on Labtek chamber slides, which had been precoated with 10 μg/ml IM7. Poly-L-lysine coated chamber slides served as control. Cells were incubated at 37°C for various times. Where indicated cells had been pretreated for 20 minutes at 37°C with the following agents: 50 mM 2-deoxyglucose plus 0.04% sodium azid (DOG/azid), 1 μg/ml cytochalasin B, 1 μM nocodazole, 80 μM pp2, 100 nM wortmannin, 50 μM Ly294002, 0.5 μM ocadaic acid, 12 mM methyl-β-cyclodextrin or DMSO (carrier).

Flow cytometry followed routine procedures using 3-5×10⁵ IP12-7 cells per sample. Samples were analyzed by a FACSCalibur (Becton Dickinson, Heidelberg, Germany). After cap formation, cells were transferred onto adhesion slides and were incubated for 15 minutes in PBS, the slides were rinsed once with PBS and cells were fixed for 30 minutes in 4% paraformaldehyde (w/v in PBS).

After spreading, slides were gently washed, cells were fixed in 4% paraformaldehyde (w/v in PBS) and were permeabilized by incubation for 4 minutes in 0.1% (v/v) Triton X-100. After washing and blocking non-specific binding sites, cells were incubated with the primary antibody at pretested concentrations (5-10 μg/ml) in PBS/BSA for 60 minutes. Slides were rinsed and subsequently incubated for 60 minutes with a fluorochrome-conjugated secondary antibody. F-actin was stained by an additional 60 minutes incubation with phalloidin-TRITC (1 μg/ml). After washing 3 times in PBS and once in H₂O, slides were mounted in Elvanol. Digitized images were generated using a confocal laser scanning microscope (TS NT, Leica, Germany). For the evaluation of two-color experiments digital images were overlaid electronically.

IP12-7 cells (5×10⁶) were seeded into Petri dishes precoated with either anti-CD44 (IM7) or control IgG and incubated for 15 minutes at 37°C. Stimulation was terminated by transferring the dishes to ice and immediate lysis of the cells. IP12-7 cells (5×10⁶) were lysed for 30 minutes in 1 ml of ice-cold TNE-buffer (20 mM Tris-HCl, pH 7.4, 150 ml NaCl, 2 mM EDTA, 10 mM NaF, 1 mM Na₃VO₄) containing 0.5% Triton X-100 and a protease inhibitor cocktail (Boehringer Mannheim). The lysate was adjusted to 40% (w/v) sucrose by mixing with 1 ml 80% (w/v) sucrose made with TNE-buffer. After transfer of the lysate to the centrifuge tube, 2 ml 30% (w/v) and 1 ml 5% (w/v) sucrose in TNE was overlaid. Samples were centrifuged for 16-18 hours at 200,000 g at 4°C. Gradient fractions (0.4 ml) were collected from the top. Fractions were analyzed by SDS-PAGE and western blotting. GEM-associated fyn was solubilized in 1% Triton X-100 plus 0.2% saponin in TNE-buffer at 4°C before immunoprecipitation by anti-fyn mAb.

IP12-cells (5×10⁶) were lysed in ice-cold TNE-buffer (20 mM Tris-HCl, pH 7.4, 150 ml NaCl, 2 mM EDTA, 10 mM NaF, 1 mM Na₃VO₄) containing either 1% (v/v) Triton X-100 or 1% (v/v) Triton X-100 plus 0.2% (w/v) saponin. All lysis buffers contained a protease inhibitor cocktail (Boehringer Mannheim). After clarification by centrifugation for 10 minutes at 10,000 g cell lysates were subjected to immunoprecipitation.

Lysates were precleared by the addition of 5 μg control antibody for 60 minutes followed by incubation with 1/10 volume Protein A-Sepharose for 2 hours at 4°C. Precleared lysates were incubated for 60 minutes at 4°C with 2 μg of anti-fyn or control IgG. Protein A-Sepharose was added for an additional 60 minutes. Immune complexes were washed 4 times with lysis buffer. Immunoprecipitated proteins were analyzed by SDS-PAGE, followed by western blotting.

Lysates were resolved on 7.5% SDS-PAGE under reducing or non-reducing conditions and the proteins transferred to Immobilon P at 90 V for 1 hour. After blocking the membranes with 3% BSA, immunoblotting was performed by using the indicated antibodies, followed by donkey anti-mouse HRP or donkey anti-rabbit HRP. Blots were developed with the enhanced chemiluminescence detection system. When the same blot was revealed with different probes, antibody stripping was performed according to the manufacturer’s recommendations.

We recently have described that CD44 functions as a costimulatory molecule likely by the recruitment of lck and fyn into the vicinity of the T cell receptor (Föger et al., 2000). Furthermore, it is known that F-actin accumulates at the interface between the T cell and an antigen presenting cell, which stabilizes this interaction and provides a scaffold for signaling components required for T cell activation (Penninger and Crabtree, 1999). Since CD44 has been repeatedly reported to associate with elements of the cytoskeleton, it could well have been that engagement of CD44 supports this rearrangement of the cytoskeleton. To test the hypothesis we first evaluated whether CD44 is involved in cap formation. When IP12-7 cells, a TH line which expresses CD44 at a high level (Rajnavolgyi et al., 1994), were incubated with anti-CD44 (IM7) and a secondary FITC-labeled antibody for cross-linking, we observed capping of CD44. Cap formation was associated with a strong and polarized accumulation of F-actin (Fig. 1A). Furthermore, Rac1, a member of the Rho family of small GTPases, clearly colocalized with CD44 (Fig. 1B). Cap formation was also induced by KM81, a CD44-specific antibody blocking the hyaluronan binding site. Although Rac1 was enriched in the KM81-induced cap, there remained a considerable proportion of free Rac1, i.e. KM81 may not be an activating antibody (data not shown). Neither Rho nor Cdc42 accumulated at the site of the caps.

Cross-linking of CD44, in addition, initiated adhesion, flattening and spreading of IP12-7 cells. This has been observed when seeding the cells on IM7-coated plates (Fig. 2A). It has not been observed when seeding the cells on KM81-coated plates, KM81 recognizing a distinct CD44 epitope in the HA binding domain. On IM7-coated plates the cells adhered within minutes and spread over a period of 1 hour as evidenced by the formation of F-actin bundles. The phenomenon has not been restricted to IP12-7 cells, which express very high levels of CD44, but was also observed with another T cell line (D10-G4.1) as well as with T cell blasts (Fig. 2B). Spreading depended on a functional cytoskeleton, because cells did neither spread in the presence of the actin disruption agent cytochalasin B nor in the presence of nocodazole, which disrupts the microtubule system (Fig. 2A). The process was energy dependent. It did not take place at 4°C nor after pretreatment of the cells with 2-deoxyglucose-azide (Fig. 2A). Finally, as described for cap formation, spreading was accompanied by formation of punctuated spots of Rac1, while Rho and Cdc42 remained distributed throughout the cytoplasm (Fig. 2C).

PI3 kinase and kinases of the src family have been implicated in cell adhesion and spreading (Han et al., 1998). In fact, when IP12-7 cells were pretreated with the src kinase inhibitor pp2, CD44-mediated spreading was inhibited in a dose dependent manner (Fig. 3A). On the other hand, there was no evidence for an involvement of PI3 kinase, i.e. spreading appeared unaltered when cells had been preincubated with LY294002 or wortmannin, two inhibitors of PI3-kinase (Fig. 3B). Spreading was also independent of serine phosphatase type 1 and 2 as evidenced by spreading after pretreatment with the inhibitor ocadaic acid.

Because CD44-mediated cell spreading was accompanied by targeting Rac1 to cell protrusions, we next explored whether Rac1 was essential for the CD44-induced rearrangement of the cytoskeleton. IP12-7 cells were transiently transfected with myc-tagged dominant negative mutant forms of Rac1 (myc-RacN17) and Rho (myc-RhoN19). Expression of the mutant proteins was confirmed by immunoblotting with anti-myc (data not shown). When myc-Rac1N17 transfected cells, which had been cultured for 1 hour on anti-CD44-coated dishes were stained with phalloidin-TRITC (and counterstained with the anti-myc antibody) no formation of F-actin fibers could be seen (Fig. 4). In addition, cells remained round and did not spread at all. Overexpression of myc-RhoN19 did not interfere with spreading and the formation of actin bundles was, if at all, only slightly reduced. Thus, Rac1 is required for the CD44-mediated cytoskeletal rearrangement, which leads to cell spreading.

Since CD44 functions as a costimulatory molecule likely by the recruitment of lck and fyn into the vicinity of the T cell receptor (Föger et al., 2000), it became tempting to speculate that the CD44-induced rearrangement of the cytoskeleton may be associated with a redistribution of CD44 into lipid-rich rafts. A first hint was obtained by seeding IP12-7 cells on IM7-coated plates after treatment with methyl-β-cyclodextrin, which leads to redistribution of rafts components by extracting cholesterol (Xavier et al., 1998). In fact, methyl-β-cyclodextrin-treated IP12-7 cells did not spread at all neither on substrate (fibronectin) nor on IM7-coated plates (Fig. 5A). Furthermore, when Triton X-100 lysates of IP12-7 cells, derived from suspension cultures, were immunoprecipitated with fyn, only a small amount of coprecipitated CD44 could be detected (Fig. 5B). Instead, when rafts were solubilized by the addition of 0.2% saponin, significantly more p59^fyn could be immunoprecipitated and the amount of coprecipitated CD44 was strongly increased. Control precipitates did neither contain p59^fyn nor CD44. The data suggest that in T cells the fraction of CD44 which constitutively associates with p59^fyn is mainly located in the rafts.

To control whether cross-linking of CD44 is associated with a redistribution of CD44 and CD44-associated p59^fyn into lipid rafts, IP12-7 cells were stimulated on plates coated with anti-CD44 or control IgG. After solubilization and density sucrose gradient ultracentrifugation to separate the low density GEM fractions, it was in particular a tyrosine phosphorylated protein with a molecular mass of approximately 60 kDa, which was augmented in the raft fractions after stimulation by anti-CD44 (Fig. 6A). Stripping the blot and reprobing with anti-fyn revealed the same pattern (Fig. 6B), i.e. quantification by densitometry revealed a ratio of 2.3:1 pixel in lysates of anti-CD44 stimulated cells as compared to control lysates in the GEM fractions. Since pp60 and fyn displayed the same mobility in the polyacrylamide gel, it is very likely that pp60 is identical to fyn. This was controlled by immunoprecipitating the pooled GEM and Triton X-100 soluble fractions with anti-fyn. After SDS-PAGE of the immunoprecipitates they were blotted with anti-phosphotyrosine and anti-fyn (Fig. 6C). In fact, pp60 was recovered in the anti-fyn precipitate, i.e. is identical to p59^fyn. Thus, engagement of CD44 leads to a redistribution of fyn into the rafts. However, it should be noted that targeting of fyn into the lipid rafts was not associated with an increase in phosphorylation, i.e. the increase in tyrosine phosphorylation correlated with the increase in the amount of fyn. Besides of fyn, p56^lck also was found to be redistributed to the rafts (data not shown). Importantly, CD44, too, was consistently found to be enriched in the rafts (ratio of densitometry values of GEM fractions of unstimulated cells to GEM fractions of stimulated cells: 1:5.4) (Fig. 6D). To control for a direct association of fyn with CD44 in the lipid rafts, pooled Triton solubilized fractions and GEM fractions were immunoprecipitated with anti-CD44 and anti-fyn. After SDS-PAGE the immunoprecipitates were blotted with anti-CD44 and anti-fyn. In the Triton soluble fraction only a small amount of fyn coprecipitated with CD44, while in the GEM fraction comparable amounts of fyn were precipitated by anti-fyn and anti-CD44, i.e. in the GEM fraction the vast majority of fyn, indeed, was found to be associated with CD44 (Fig. 6E).

Taken together, engagement of CD44 initiates F-actin bundle formation accompanied by a redistribution of CD44 and the associated tyrosine kinases into the rafts, i.e. into the neighborhood of the peptide-MHC engaged TCR.

Efficient activation of T cells requires the engagement of costimulatory molecules (Chambers and Allison, 1997; Croft and Dubey, 1997). Recently, it has been debated whether these costimulatory molecules exert their function by stabilizing the TCR-MHC interaction and/or by recruiting signal transduction molecules towards the TCR/CD3 complex (Shaw and Dustin, 1997; Wülfing and Davis, 1998; Viola et al., 1999). Our previous analysis of a costimulatory function of CD44 in T cell activation supported the latter hypothesis, i.e. costimulation by the recruitment of the src-family kinases p59^fyn and p56^lck (Föger et al., 2000). This finding raised the question as to the mechanism by which CD44 strengthened signaling via the TCR/CD3 complex. There were at least 3 possible explanations: (i) CD44 functions as an adhesion molecule which allows for a prolonged contact between the TCR and its ligand; (ii) cross-linking of CD44 initiates reorganization of the cytoskeleton towards a focal actin-scaffold, which is required for effective T cell activation (Wülfing and Davis, 1998; Howard and Watts, 1994; Dustin et al., 1998; Valitutti et al., 1995); (iii) cross-linking of CD44 leads to its redistribution in the membrane in such a way that it neighbors the TCR/CD3 complex, which would facilitate an efficient delivery of the constitutively CD44-associated PTKs fyn and lck. In the latter case it became of special interest, whether CD44 would be driven towards glycolipid-enriched microdomains, which also gather the TCR during the activation process (Wülfing and Davis, 1998; Xavier et al., 1998; Viola et al., 1999; Kosugi et al., 1999; Montixi et al., 1998). We now report on a CD44-mediated rearrangement of the actin cytoskeleton and a redistribution of CD44 as well as of the associated kinases p59^fyn and p56^lck into lipid rafts.

The TCR has a relatively low affinity for its peptide-MHC ligand (Davis et al., 1998). Furthermore, T cells and APCs have a tendency to repel each other owing to their net negative surface charge (Springer, 1994). On the other hand, T cell activation requires a sustained contact (Davis et al., 1998). The prime candidates to provide the required attractive forces are adhesion molecules (Dustin and Springer, 1989; Croft and Dubey, 1997; Valitutti et al., 1995). Thus, CD44, a well characterized adhesion molecule, could well function as a costimulatory molecule by strengthening the contact between T cell and APC either directly or by potentiating integrin activity as observed after triggering of CD44 on T cells (Koopman et al., 1990). However, CD44 capping, spreading and, as already reported, costimulation (Föger et al., 2000) depended on the epitope recognized by the cross-linking antibody. These findings argue against ligand binding as the exclusive function of CD44.

Recently it has been demonstrated that TCR clustering requires actin polymerization in the T cell (Wülfing and Davis, 1998; Penninger and Crabtree, 1999; Dustin et al., 1998; Valitutti et al., 1995). It has been suggested that in the first instance signals have to be delivered which initiate actin polymerization and accumulation of cortical actin at the contact area, which could serve as a scaffold for the reorganization of membrane microdomains as well as for the formation of the multimolecular signaling complexes required for T cell activation (Wülfing and Davis, 1998; Penninger and Crabtree, 1999). The importance of actin reorganization for receptor clustering and T cell activation has been demonstrated by the treatment of T cells with cytochalasins, which disrupt actin filaments: TCR capping is inhibited, T cells do not change shape and activation is significantly impaired (Holsinger et al., 1998; Kong et al., 1998; Valitutti et al., 1995). Cross-linking of CD44 via IM7 (but not via KM81) induced profound morphological changes, like increased adhesion, flattening and spreading accompanied by the formation of actin bundles. Accordingly, capping of CD44 by IM7 was accompanied by accumulation of F-actin at the site of the cap. Taking these features, it becomes very likely that CD44 is a potent candidate for mediating the cytoskeletal rearrangements required for the initiation of T cell activation. Our finding also provided the basis for a detailed analysis of the involved signaling elements. Although this analysis has yet to be completed, a variety of candidate molecules could be identified or excluded by the use of specific inhibitors. Thus, we could demonstrate the involvement of src-family kinases, which also have been described to be involved in integrin mediated spreading and migration (Price et al., 1998). Since CD44 is constitutively associated with fyn and lck, it is likely that these kinases may be involved in the very proximal signaling leading to cytoskeletal restructuring. PI-3 kinase has been reported to influence actin dynamics following engagement of CD2 and CD28 (Shimizu et al., 1995). However, it apparently is not involved in CD44 mediated actin reorganization. The Rho subfamily of small GTPases are considered as key regulators of the actin cytoskeleton (Mackay and Hall, 1998; Hall, 1998). By immunofluorescence as well as by the use of dominant negative mutants we could clearly demonstrate the involvement of Rac1 in CD44-induced actin polymerization, which recently has also been described for hyaluronic acid binding of CD44 (Oliferenko et al., 2000). Besides of its well characterized role in mediating actin cytoskeletal changes in fibroblasts (Mackay and Hall, 1998; Hall, 1998), Rac has also been implicated in integrin-mediated adhesion and spreading of T lymphocytes (D’Souza et al., 1998). Two downstream targets of Rac which likely play a central role in actin polymerization are phophatidyl-4-phosphate kinase and p65PAK (Hartwig et al., 1995). The product of phophatidyl-4-phosphate kinase, phophatidyl-4,5-biphosphate is known to affect actin filament assembly (Hartwig et al., 1995), p65PAK is supposed to interact with a molecular complex that controls actin polymerization (Price et al., 1998). Whether p65PAK is involved directly in the activation of the JNK pathway is still a matter of debate (Brown et al., 1996; Tapon et al., 1998). With respect to cross-linking of CD44 which induced Rac1 activation, we did not observe c-jun phosphorylation. This argues against CD44 affecting T cell activation directly via a Rac1 – p65PAK – JNK signaling pathway. We also found no evidence for an involvement of ERM proteins, which have been described to anchor actin to cell surface receptors as a prerequisite for Rho and Rac to induce cytoskeletal changes (Mackay et al., 1997). We could, however, neither detect ezrin nor RhoGDI in CD44 immunoprecipitates of T cells (data not shown). Finally, it should be mentioned that the CD44-induced morphological changes also involved the microtubular system. This is not surprising, since in many systems the cortical actin and microtubules act in concert. It has been described that microtubule growth, activation of Rac1 and actin polymerization functionally cooperate (Waterman-Storer et al., 1999), an observation which is fully in line with our findings. Although we have discussed the CD44-induced cytoskeletal reorganization particularly in view of T cell activation, it should be mentioned that T cell adhesion, flattening and spreading may also be important for homing, rolling, extravasation and migration through the space of the extracellular matrix. CD44 is known to be involved in all of these processes (Borland et al., 1998).

By reorganization of the cytoskeleton, which would allow for an increase in the contact area between the T cell and the APC, CD44 could strengthen signaling via the TCR/CD3 complex. There remains the question whether the reorganization of the cytoskeleton is accompanied by the recruitment of receptors and/or signaling molecules to the zone of contact, as has been described for the TCR (Shaw and Dustin, 1997; Wülfing and Davis, 1998; Xavier et al., 1998; Kosugi et al., 1999; Montixi et al., 1998; Grakoui et al., 1999; Moran and Miceli, 1999). Several lines of evidences strongly support the hypothesis. It has been described for the human system (Dianzani et al., 1999) and we have described for the murine system, too, that CD44 is constitutively associated with lck and fyn, which phosphorylate multiple ITAM motifs of the TCR/CD3 complex at the initiation of activation. We also could identify a binding motif of CD44 which is responsible for the association with src-kinases (Rozsnyay, 1999). An efficient phosphorylation of both the TCR/CD3 complex and ZAP70 by lck and fyn has also been described for the costimulatory molecules CD4, CD2 and CD28 (Davis and van der Merwe, 1996; Holdorf et al., 1999). Furthermore, lck and fyn have also be shown to be essential as adaptor proteins, important for the assembly of a transduction-competent signaling complex at latter stages of T cell activation. It also has been suggested that cellular stimulation via CD44 may proceed through the signaling machinery of GEMs, because CD44 has been found to selectively associate with active src-family protein tyrosine kinases in these microdomains (Ilangumaran et al., 1998). The hypothesis is fully in line with our observations that CD44-induced spreading is prohibited by sequestration of cholesterol, that CD44 associated fyn is particularly enriched in GEMs and that cross-linking of CD44 is associated with a redistribution of CD44, fyn and lck into lipid rafts. Accordingly, it has been described that costimulation by CD28 or LFA-1 initiates an actin-myosin driven directional transport of protein and lipid domains to the T cell – APC contact zone (Wülfing and Davis, 1998), coengagement of CD28 and TCR/CD3 complex recruiting all GEMs to the contact area (Viola et al., 1999).

Taken together, cross-linking of CD44 induces a reorganization of the actin cytoskeleton, which involves the microtubule system and leads to T cell adhesion, flattening and spreading. The reorganization of the cytoskeleton is accompanied by a redistribution of CD44 and the associated PTKs fyn and lck into lipid rafts. Because of these features and the findings that i. engagement of the TCR also leads to its redistribution in lipid rafts, (ii) CD44 functions as a costimulatory molecule by lowering the threshold for signal transduction via the TCR/CD3 complex, we interpret our data in the sense that it is the CD44-induced reorganization of the cytoskeleton and the associated redistribution into GEMs which provide the basis for the costimulatory function of CD44.

This work was supported by the Deutsche Forschungsgemeinschaft, grant no Zo40/5-3 (MZ). We cordially thank Dr E. Rajnavolgyi. L. Eotvos University, God, Hungary, for the generous supply of the T cell line IP12-7 and Prof. P. Boquet, INSERM U452, Nice, France, for kindly providing us with expression vectors encoding myc epitope tagged dominant negative mutant forms of RhoA and Rac1.