T-cell activation is initiated by the concerted engagement of the T-cell receptor and different co-stimulatory molecules, and requires cytoskeleton-dependent membrane dynamics. Here, we have studied the relationships between tetraspanins, cytoskeleton and raft microdomains, and their relevance in T-cell signaling. Localization studies and density-gradient flotation experiments indicate that part of tetraspanins localizes in raft microdomains linked to the actin cytoskeleton. First, partial coalescence of lipid raft is triggered by tetraspanin cross-linking and results in large caps in which F-actin also concentrates. Second, the amount of tetraspanins, which are recovered in the cholesterol-dependent insoluble fractions of low and intermediate density, and which appears to be membrane vesicles by electron microscopy, is under cytoskeletal influence. Disruption of actin filaments enhances the amount of tetraspanins recovered in typical raft fractions, whereas F-actin-stabilizing agents induce the opposite effect. Our data also reveal that CD82 constitutes a link between raft domains and the actin cytoskeleton, which is functionally relevant. First, tetraspanin signaling induces a selective translocation of CD82 from detergent-resistant membrane fractions to the cytoskeleton-associated pellet. Second, all functional effects linked to CD82 engagement, such as adhesion to culture plates, formation of actin bundles and early events of tyrosine phosphorylation, are abolished, or strongly reduced, by cholesterol depletion. We also show that dynamic relocalization of CD82 and F-actin at the periphery of the immune synapse is induced upon contact of T cells with antigen-presenting cells. This suggests that the tetraspanin web might participate in the membrane dynamics required for proper T-cell signaling. More generally, the interaction of tetraspanins with raft domains and with the actin cytoskeleton might relate with their role in many cellular functions as membrane organizers.
The tetraspanin family is composed of many four-transmembrane-segment proteins, which have been involved in various cellular functions such as activation, adhesion, proliferation, motility and cell fusion (for reviews, see Berditchevski, 2001; Boucheix and Rubinstein, 2001). Several tetraspanins, such as CD9, CD63 and CD82, have been shown to regulate the metastatic potential of various tumors. Proteins of this family are characterized by their ability to form large multimeric complexes involving a wide range of membrane and cytosolic proteins. Protein interactions common to several tetraspanins include those with other tetraspanins, a subset of β1 integrins, major histocompatibility complex (MHC) molecules and co-receptor molecules like CD4 (or CD8) in T cells and CD19/CD21 in B cells. Such properties have led to the hypothesis that tetraspanins might have a general role as organizing proteins allowing the formation and targeting of specific membrane complexes (Lagaudriere-Gesbert et al., 1998; Maecker et al., 1997; Seigneuret et al., 2001). Furthermore, the capability of mutual interactions between tetraspanins has suggested that these complexes could be organized in a membrane web involving a network of tetraspanin interactions (Horvath et al., 1998; Rubinstein et al., 1996).
This situation has recently prompted several authors to examine the relationships between tetraspanins and a better documented type of membrane microdomains, namely lipid rafts. Rafts are liquid-ordered small-size domains highly enriched in sphingolipids, cholesterol and saturated fatty acyl chains lipids that appear to be present in most plasma membranes and several internal membranes (Brown and London, 1997). Rafts are characterized by their relative insolubility in cold non-ionic detergent, which leads to their designation as detergent-resistant membrane (DRMs) or detergent-insoluble glycolipid-enriched membrane (DIGs) (Harder and Simons, 1997; Simons and Ikonen, 1997). Glycosylphosphatidylinositol (GPI)-anchored proteins and many acylated proteins involved in signal transduction, as well as some integral membrane proteins, localize preferentially to these raft domains.
These structures are highly dynamic. For example, distinct membrane receptors have been shown to associate with raft domains specifically after cell activation (Cerny et al., 1996; Montixi et al., 1998). Reciprocally, cell activation by various receptors has been shown to depend on the presence of rafts as revealed by inhibition by the use of cholesterol-depleting agents (Xavier et al., 1998). Altogether, these data emphasize that membrane organization plays a fundamental role in cell activation and lead to the concept that membrane rafts might function as signaling platforms (Kurzchalia and Parton, 1999; Xavier et al., 1998). Recently, several tetraspanins have been shown to be partly associated with lipid rafts in various cell types. These include CD81, CD9, CD63, CD82, CD151 and the tetraspanin-like protein ROM-1 (Boesze-Battaglia et al., 2002; Charrin et al., 2002; Claas et al., 2001; Horejsi et al., 1994; Yang et al., 2002). However, mutual tetraspanin interactions and their association with integrins or phosphatidylinositol-4-kinase are not mediated by the raft lipid phase and occur both inside and outside lipid rafts. Moreover, at least for two tetraspanins (CD9 and CD151), location in rafts is not dependent on palmitoylation, although palmitoylation appears to be involved in interactions between tetraspanins (Berditchevski et al., 2002; Charrin et al., 2002; Yang et al., 2002). It has been suggested that tetraspanins are present in lipid rafts but also occur in specific microdomains different from rafts (Kropshofer et al., 2002). In APC, both types of domain might be involved in the clustering of MHC class II and in the enrichment in specific MHC-class-II/peptide complexes (Hammond et al., 1998; Kropshofer et al., 2002; Vogt et al., 2002). Tetraspanin-containing lipid rafts have also been involved in the recruitment of membrane proteins to multivesicular bodies (Heijnen et al., 1999; Wubbolts et al., 2003).
T-cell activation is a cellular process in which an important role of lipid rafts, actin cytoskeleton and tetraspanins has been separately documented. Many lines of data indicate that, in T cells, specific targeting to rafts and raft dynamics are pivotal in the recruitment and signaling of the T-cell receptor (TcR) and of various co-stimulatory molecules, as well as in formation of the immunological synapse (IS) (Burack et al., 2002; Goebel et al., 2002; Langlet et al., 2000; Moran and Miceli, 1998; Xavier et al., 1998). By contrast, several tetraspanins (CD9, CD53, CD81 and CD82) are known as T-cell co-stimulatory molecules that function, at least for CD81 and CD9, independently of CD28 (Lebel-Binay et al., 1995; Maecker, 2003; Tai et al., 1996; Toyo-oka et al., 1997; Wack et al., 2001). The precise mechanisms responsible for T-cell co-stimulation by tetraspanins are currently unknown, although associations with the co-receptor CD4 or the adhesion molecule LFA-1 have been suggested to be involved (Mittelbrunn et al., 2002; Shibagaki et al., 1999). CD82 appears to play a particular role because, unlike other tetraspanins, its engagements leads to its association with the actin cytoskeleton (Lagaudriere-Gesbert et al., 1998). Activation mediated by CD82 has also been shown to depend on the guanine-nucleotide-exchange factor Vav-1 and on activity of the Rho GTPases (Delaguillaumie et al., 2002). Recently, redistribution of CD81 and colocalization with the TcR at the c-SMAC of the IS has been demonstrated (Mittelbrunn et al., 2002). In the present work, we have investigated, in T cells, the relationships between tetraspanins, lipid rafts and the cytoskeleton, as well as their role in tetraspanin signaling and during formation of the IS.
Materials and Methods
Cell culture, antibodies and reagents
ATCC Jurkat cells variant expressing SV40 T antigen (Jurkat TAg cells) were maintained in RPMI 1640 glutamax (Life Technologies) supplemented with 7% fetal calf serum (FCS). The different T8.1 clones were maintained in Dulbecco's modified Eagle's medium (DMEM) with 10% FCS, 2 mM glutamine, 400 nM methotrexate, 1 mg ml–1 G418, 50 μM 2-mercaptoethanol and 1 μg ml–1 puromycin. Stable T8.1 clones expressing plasmids pEYFP-IA4, pEYFP-CD3ζ or pYFP were selected 48 hours after transfection by electroporation (950 μF, 250 V) in the presence of both the puromycin-resistance-gene-containing plasmid pSrα and pYFP-coupled vectors. Four different clones expressing different levels of yellow fluorescent protein (YFP)-tagged CD82 were used for relocalization studies. L625.7 cells were cultured in complete DMEM with 250 μg ml–1 G418 and 1 μg ml–1 puromycin. phytohemagglutinin (PHA) blasts were obtained by culturing nonadherent peripheral-blood mononuclear cells (PBMCs) from healthy donors in presence of PHA (1 μg ml–1) for 3 days in culture medium (RPMI 1640 glutamax + 7% FCS). After extensive washing, cells were cultured for 3-5 days more in culture medium alone before use.
The stimulating monoclonal antibodies (mAbs) used were: anti-CD82 mAb, purified γC11 (mouse IgG1) (Lebel-Binay et al., 1994); purified anti-CD3, OKT3 (mouse IgG2a, kind gift of Orthoclone); anti-CD9, purified syb1 (mouse IgG1) (Charrin et al., 2001); anti-CD81, TS81 (ascite) (Charrin et al., 2001); anti-CD53, TS53 (ascite) (Charrin et al., 2001); anti-CD63, 6H1 hybridoma supernatant (a gift from F. Berditchevsky, CR UK Institute for Cancer Studies, Birmingham, UK); anti-phosphorylated-tyrosine, 4G10 (Upstate Biotechnology); anti-LAT (a membrane adaptor protein), rabbit polyclonal IgG (Upstate Biotechnology); anti-CD81, JS64 (mouse monoclonal IgG2a; Immunotech). Sheep horseradish-peroxidase (HRP)-conjugated anti-mouse immunoglobulin (Ig) and donkey HRP-conjugated anti-rabbit Ig were purchased from Amersham. Goat phycoerythrin (PE)-labeled anti-mouse IgG, rabbit HRP-conjugated anti-sheep and goat HRP-conjugated anti-mouse IgG2b were from Southern Biotechnology. HRP-conjugated goat anti-mouse IgG1 and HRP-conjugated goat anti-mouse IgG2a were from Caltag. Biotin-, HRP- and fluorescein isothiocyanate (FITC)-labeled cholera toxin (CTx) were from Sigma. The anti-CD71 antibody (mouse IgG1) used in biochemical studies were from Zymed, and the one used for surface labeling was from Boehringer (clone B3.25)
Cholesterol depletion was induced by cell incubation for 30 minutes with 5 (or 10) mM methyl-β-cyclodextrin (MβCD; Sigma) at 37°C in RPMI supplemented with 2% bovine serum albumen (BSA) and 20 mM Hepes, followed by washing in RPMI. In order to check for possible effects of MβCD treatment on cell viability and functionality, the following controls were performed. Cell viability was assayed by classical trypan-blue exclusion. For simultaneous analysis of tetraspanin surface expression and cell viability, 5×105 cells (MβCD treated or not) were incubated for 30 minutes in 50 μl saturating dilution (50 mg μl–1) of anti-tetraspanin mAbs in PBS with 1% FCS. After three washes, cells were incubated for 30 minutes with FITC-labeled goat anti-mouse mAb, washed, transferred to Falcon tubes with 400 μl PBS and kept at 4°C. 15 minutes before cytometric analysis, propidium iodide (PI) was added to the cell suspension (20 μg ml–1 final concentration) and fluorescence at 520 nm (FITC) and 630 nm (PI) measured simultaneously. Neither control showed any detectable difference between treated and control cells. Inhibition of N-glycosylation was performed by overnight treatment with 1 μg ml–1 tunicamycin at 37°C. Inhibition of actin cytoskeleton was performed by a 2-hour treatment with 2 μg ml–1 latrunculin at 37°C.To induce F-actin stabilization, cells were electroporated (960 μF, 250 V, BioRad Gene Pulser II, 4 mm gap cuvette Bio-Rad) in presence of phalloidin 15 minutes before lysis.
Stimulation was performed by culturing cells on antibody-coated culture plates for various times. Coating of mAbs was performed as followed. mAbs diluted to the final concentration (as stated in figure legends) in coating buffer (11 mM Na2CO3, 35 mM NaHCO3, 3 mM NaN3) were incubated overnight at 4°C on various culture supports. To avoid variations in coating efficiency, oversaturating doses of mAbs were used. After one wash with PBS with 1% FCS, culture supports were saturated for 30 minutes at 37°C with PBS supplemented with 10% FCS. For tyrosine phosphorylation and adhesion assays, cells were plated at 2×105 cells per well in RPMI supplemented with 7% FCS, on antibody-coated 96-well microplates. For fractionation on sucrose gradient, cells were seeded with 2×107 cells in 5 ml culture medium on 10 cm Petri dishes. For cytoskeleton rearrangement assays, 1×106 cells in RPMI supplemented with 7% FCS were plated on antibody-coated Thermanox slides (Nunc, Naperville, IL) precleaned by three rinses in ethanol, in the bottom of four-well Multidishes sterilized Nunclon plates (Nunc). Protein concentrations were measured using MicroBC assays (Interchim).
Separation of membrane fractions by sucrose density gradient
2×107 cells washed in PBS were lysed in 800 μl MNE buffer [25 mM 2-morpholinoethanesulfonic acid (MES), 150 mM NaCl, 2 mM EDTA] supplemented with 1 mM Na3VO4, 1 mM NaF and protease inhibitors (EDTA-free tablets; Boehringer Mannheim) in the presence of 0.5% Triton X-100, or 1% Brij-58 when stated in the text. After 1 hour on ice, cell lysates were homogenized with 15 strokes of a loose-fitting Dounce homogenizer. After adjustment of protein concentration in each samples (MicroBC assay, Interchim), 700 μl of cell lysates were mixed with an equal volume of 85% sucrose (w/v in MNE) and placed in the bottom of SW55Ti ultracentrifuge tubes. The samples were then overlaid with 1.4 ml 30% sucrose and 700 μl 5% sucrose in MNE supplemented with 1 mM Na3Vo4, 1 mM NaF and protease inhibitor tablets. The gradient was centrifuged for 16 hours at 200,000 g at 4°C. Membrane domains were harvested by collecting 12 fractions of 280 μl from the top of the gradient. Insoluble pellets were collected, washed once in MNE and solubilized in sodium-dodecyl-sulfate (SDS) sample buffer (50 μl). The influence of cell concentration was tested by performing experiments with a five times lower cell concentration at lysis, which yielded similar proportions of insoluble material.
Proteins were resolved by SDS-PAGE in non-reducing conditions and then transferred to polyvylidene difluoride (PVDF) membranes (Dupont-New England Nuclear). Immunoblotting was performed with anti-tetraspanin, anti-CD71 or anti-LAT antibodies followed by HRP-conjugated sheep anti-mouse IgG (tetraspanins and CD71) or HRP-conjugated donkey anti-rabbit (LAT) antibodies. Immunoreactive proteins were visualized using the ECL Western Blotting Detection Kit (Amersham) and Kodak films.
Biotin labeling and immunoprecipitation assays
For surface biotinylation, cells were washed three times in Hanks' buffered saline and resuspended at 3×107 ml–1 in 10 mM Hepes, pH 7.3, 150 mM NaCl, 0.2 mM CaCl2, 0.2 mM MgCl2, 0.25 mg ml–1 sulfo-NHS-LC-biotin (Pierce, Rockford, IL). After a 30 minute incubation at 4°C, cells were washed three times in TBS (20 mM Tris-HCl, pH 7.4, 137 mM NaCl, 0.2 mM CaCl2, 0.2 mM MgCl2) to remove free biotin and to inhibit the reactive group.
After cell lysis in Triton X-100 and sucrose fractionation, each fraction was diluted in MacDougal buffer (20 mM Tris-HCl, pH 7.8, 120 mM NaCl, 200 μM EGTA, 0.2 μM NaF, 0.2% sodium deoxycholate, 0.5% Nonidet P-40) and biotin-labeled proteins were collected using avidin-Sepharose beads and washed five times in MacDougal buffer. After addition of 60 μl SDS sample buffer and 10 minutes of boiling, 30 μl of each sample were separated using SDS-PAGE on 15% non-reducing gels. Proteins were transferred on PVDF membranes. Immunoblotting with syb1 (anti-CD9) or JS64 (anti-CD81) was followed by HRP-conjugated sheep anti-mouse IgG1 or HRP-conjugated goat anti-mouse IgG2a antibodies.
Cell adhesion assays
2×105 cells were plated on anti-CD82 or anti-CD3 coated plates in RPMI supplemented with 7% FCS at 37°C. After 1 hour of culture, cells were washed once with PBS, fixed and stained by adding 100 μl per well 0.5% violet crystal in 20% methanol. After 10 minutes, plates were extensively washed with distilled water and violet crystal was extracted by cell lysis in 1% SDS. Optical density at 520 nm was measured in an enzyme-linked immunosorbent assay (ELISA) reader (Dynatech).
Cytoskeleton rearrangement assays
Jurkat cells (106 ml–1) in RPMI supplemented with 7% FCS were layered over the coated Thermanox slides and incubated for various times at 37°C. Non-adherent cells were removed by gentle washing with PBS. Adherent cells were fixed with 4% paraformaldehyde, 2% sucrose in PBS for 30 minutes at room temperature. After permeabilization by a 1 minute incubation at 0°C with permeabilization buffer (0.1% Triton X-100, 20 mM Hepes, pH 7.4, 300 mM sucrose, 1 mM MgCl2, 1 mM CaCl2, 150 mM NaCl), F-actin was labeled with 2 U ml–1 rhodamine-conjugated phalloidin (Molecular Probes, Eugene, OR) for 20 minutes at room temperature. Slides were mounted in Fluoromount (Southern Biotechnology) to prevent photobleaching. Cells were analysed by confocal fluorescence microscopy (Bio-Rad MRC1000; Bio-Rad Laboratories, Hercules, CA). Digitized pictures were analysed with Confocal Assistant software (Todd Clarke Brejle) and processed using Adobe Photoshop.
For direct tyrosine-phosphorylation assays, cells (pretreated or not) were cultured on antibody-coated 96-well microplates. After incubation at 37°C for various times, cells were lysed by adding 50 μl of a 2× stock of reducing SDS sample buffer. Samples were boiled for 5 minutes and proteins separated by SDS-PAGE on 10% gels. Anti-phosphotyrosine immunoblotting was performed using 4G10 (UBI) followed by HRP-conjugated goat anti-mouse IgG2b (GaM-IgG2b; Southern Biotechnology).
Jurkat cells (5×105) were washed in PBS supplemented with 2% FCS (PBS-2%) and incubated with 10 μg ml–1 biotin-labeled CTx for 30 minutes at 4°C. After two washes in PBS-2%, GM1 gangliosides were revealed by FITC-labeled avidin (30 minute incubation at 4°C). After two washes in PBS-2%, capping was induced by incubation of 30 minutes at 37°C. Cells were fixed (15 minute incubation in PBS supplemented with 4% paraformaldehyde, followed by a saturation step of a 15 minute incubation at room temperature in PBS supplemented with 0.1 M glycine) and washed twice in PBS-2%. Surface proteins were visualized by a 30 minute incubation at 4°C with saturating concentration of specific mAbs, two washes in PBS-2% and a 30 minute incubation at 4°C with Cy5-labelled goat anti-mouse antibody (Amersham Pharmacia Biotech). Reciprocal experiments were performed by a first incubation with primary mAbs against surface proteins. Capping was induced by secondary mAbs (30 minutes at 4°C, two washes, and further incubation of 30 minutes at 37°C). After fixation, GM1 labeling was achieved with FITC-labeled CTx. Under both sets of conditions, cells were washed twice in PBS and then permeabilized (1 minute incubation at 0°C with permeabilization buffer), and F-actin was revealed by labeling with 2 U ml–1 rhodamine-conjugated phalloidin.
For electron microscopy of fractions, aliquots of pool of fractions 2-3, 4-5, 6-7, 8-9 and 10-12 were diluted in PBS and pelleted by 100,000 g centrifugation for 2 hours. Pellets were placed on Formvar carbon-coated electron-microscopy grids, fixed by incubation in a mixture of 2% paraformaldehyde in 0.2 M phosphate buffer, pH 7.4, 0.125% glutaraldehyde for 2 hours at room temperature. The sections were contrast stained, embedded in a mixture of methylcellulose and uranyl acetate, and viewed under a CM120 Twin Phillips electron microscope (Phillips, Eindhoven, The Netherlands).
CD82 re-localization and changes in intracellular calcium
In order to express a CD82-YFP fusion, the CD82-coding sequence was excised from pCMVIA4pa2 (Pique et al., 2000) by HindIII and ligated into with EYFP-N1 in frame to the N-terminus of the enhanced yellow fluorescent protein. The sequence of the resulting CD82-YFP fusion was confirmed by sequencing. To check the functionality of the tagged protein, phenotypes and tyrosine phosphorylations induced by stimulation with anti-CD3 and anti-CD82 mAbs were compared from cells overexpressing wild-type CD82 or CD82-YFP. Antigen-presenting cells (APCs; 3×105 L625.7) were loaded with tetanus toxin peptide tt (830-843) overnight at 37°C on Thermanox or glass slides. 3×105 T8-1 clones expressing YFP-CD82, YFP-TcRζ or YFP were incubated with 1 μg fura2-AM (Molecular Probes) for 20 minutes at 37°C and then added to the APCs. Cells were visualized using a Nikon Eclipse TE 300 inverted microscope (Nikon, Melville, NY) and an IMSTAR imaging system as described previously (Donnadieu et al., 1994). Images of phase-contrast microscopy, Fura-2 fluorescence ratio (380/340) and YFP fluorescence (525 nm) were analysed using Metafluor or Metavue software (Universal Imaging, West Chester, PA).
Tetraspanins molecules co-cap with GM1 gangliosides together with F-actin
To visualize in detail the relationships between tetraspanins, rafts and the actin cytoskeleton, co-localization studies were performed using antibodies, CTx and phalloidin as probes, respectively, for tetraspanins, GM1 gangliosides and F-actin on the Jurkat leukemic T-cell line. Flow-cytometry analysis indicated that these cells express comparable levels of the three tetraspanins CD9, CD81 and CD82 (relative fluorescence intensities were 10, 16.2 and 16, respectively). On untreated Jurkat cells, owing to the very small size of raft domains, all individual markers led to uniform membrane staining without specific colocalizations (Fig. 1Ae-h). However, formation of large rafts, detected as GM1 patches, was induced by streptavidin cross-linking of biotin-labeled CTx (Fig. 1Ab). A large proportion of all assayed tetraspanins (CD9, CD81, CD82) localized with these large GM1 patches in most cells (respectively, 60%, 75% and 75% of all cells). Dense F-actin structures were often observed in these patches (respectively, in 50%, 65% and 65% of all cells). In Fig. 1Aa-d,Ba-c, co-localization of the three molecules appears in white and is indicated by arrows. Interestingly, CD9, CD81 or CD82 cross-linking also induced partial raft coalescence (respectively, in 75%, 75% and 85% of all cells). Large patches of tetraspanins were formed in which GM1 and F-actin were frequently, but not systematically, observed (60% of cells in the case of CD82; Fig. 1Be-g). Control experiments indicated that the raft-excluded molecules CD71 (transferrin receptor; Fig. 1Bd) and the co-stimulatory protein CD28 (not shown), did not colocalize with GM1. Consistently, cross-linking of CD71 induced only its own patching in membrane areas in which neither GM1 nor actin colocalized with CD71. These data indicate that a proportion of T-cell tetraspanins associates with lipid rafts together with F-actin.
Tetraspanins partially localize to fractions of low density
To analyse in more details the link between tetraspanins and rafts, and considering the relative insolubility of rafts in various detergents (Schroeder et al., 1994), flotation experiments were performed using sucrose-density gradients. In a first approach, we compared the sucrose-density distribution of CD82 to the one observed for raft-associated and raft-excluded proteins in two extraction detergents (Triton X-100 or Brij 58). Cell lysis using Brij 58 led to large amount of CD82 (Fig. 2A, left) in the lightest fractions (1-4). However, only Triton X-100 (0.5%) was able to solubilize properly the raft excluded membrane protein CD71 (Fig. 2A, right). These results suggest that Brij 58 poorly solubilizes some T-cell membrane proteins although these are excluded from raft domains. Therefore, all further studies were performed using Triton X-100. As expected, the membrane adaptor protein LAT (Fig. 2B), the tyrosine kinase Lck (not shown) and GM1 gangliosides (not shown) were mostly detected in the lightest fractions of the sucrose gradient (1-4). The cytosolic protein ZAP-70 was almost completely recovered in the dense fractions (9-12) together with solubilized proteins (not shown).
From the 12 fractions obtained by sucrose gradient, the first four (1-4), and the last (9-12) sequential fractions, were termed DRM and heavy fractions, respectively. Indeed, the acknowledged raft marker is mainly located in DRM (1-4), while most of the soluble and solubilized material (ZAP70 and CD71) is recovered in fractions (9-12). The remaining fractions (5-8) define what we called intermediate-density fractions. Although this definition is approximate because of possible overlap with the heavy fractions (i.e. CD71 occasionally appears in fraction 8, as for example in Fig 2B), it allows to monitor the tetraspanin density pattern.
Density analysis revealed that 40-55% of CD82 was recovered in the heavy fractions in which about 75% of total proteins and 87% of the raft-excluded CD71 were recovered (Fig. 2B, Table 1). Similar distribution was observed for other tetraspanins (CD9 and CD81). The broader band observed for CD82 in western blots reflects its heterogeneous N-glycosylation leading to a smear in SDS-PAGE (Lebel-Binay et al., 1994). A small proportion (5-10%) of tetraspanins migrated in the lighter fractions characteristic of classical rafts and in which 85% of the DRM-associated protein LAT was recovered. Strikingly, a large proportion (40-50%) of CD82 and other tetraspanins was recovered in fractions of intermediate density (particularly in fractions 6 and 7) in which only 15% of total proteins, 10% of CD71 and 5% of LAT were recovered. As shown in Fig. 2C, fractionation of PHA-activated T cells from healthy donors (3 days, 1 μg ml–1) after Triton X-100 solubilization also yielded a low-density insoluble fraction for CD82, which is the main tetraspanin expressed. Because CD9 and CD81 are expressed in PHA blasts at very low levels compared with CD82, no conclusion can be reached concerning their possible occurrence in low-density insoluble fractions.
|.||Light fractions (1-4) .||Intermediate fractions (5-8) .||Soluble fractions (9-12) .|
|.||Light fractions (1-4) .||Intermediate fractions (5-8) .||Soluble fractions (9-12) .|
2×107 Jurkat cells were lysed and fractionated by sucrose gradient. An equal volume of a pool of four sequential fractions (1-4, 5-8 and 9-12) was resolved by SDS-PAGE and immunoblotted using antibodies against CD82, LAT or CD71, or analysed for total protein contents (Bradford assay). Different exposures of Kodak films were analysed and quantified by visible illumination on a Bioprofil gel analyser. Mean values from three independent experiments are shown.
Tetraspanin density distribution in sucrose-density gradient depends on the presence of cholesterol but not of tetraspanin post-translational modifications
The major role played by sterol molecules in DRM formation (Harder and Simons, 1997) led us to analyse how cholesterol depletion would modify the tetraspanin distribution in a sucrose gradient. This was achieved by addition of a cholesterol-specific detergent (0.1% saponin) to the T-cell lysis buffer. Increased solubilization of raft-associated proteins was checked by the decreased amount of LAT recovered within low-density fractions (not shown). Interestingly, a strong decrease of the amount of tetraspanin recovered in the low- and intermediate-density fractions was induced by cholesterol solubilization with saponin (Fig. 3A). This confirms that tetraspanins associated with the lightest fractions are indeed associated with classical rafts. This also indicates that fractions of intermediate density that are enriched in tetraspanins are dependent upon membrane cholesterol, as are typical DRMs.
Various post-translational modifications have been suggested to participate in protein interactions with raft domains. Because CD82 is highly N-glycosylated, we have analysed whether its glycosylation state participates in its insolubility. Culturing Jurkat cells in the presence of tunicamycin for 24 hours inhibited N-glycosylation of newly synthesized CD82. As shown in Fig. 3B, the strong reduction of the apparent molecular weight of CD82 induced by tunicamycin did not correlate with any alteration of its distribution on a sucrose-density gradient. Similar negative effects were observed for the density distribution of the weakly glycosylated tetraspanin CD9 (not shown). Similarly, palmitoylation has been shown to be important for raft localization of various integral proteins such as LAT (Zhang et al., 1998). However, although wild-type CD9 was shown to be highly palmitoylated at cysteine residue located at juxtamembrane domains, mutated CD9 [which failed to incorporate palmitic acid (Charrin et al., 2002)] migrated with similar distribution on a sucrose-density gradient as did wild-type CD9 (Fig. 3C). These results indicate that post-translational modifications of tetraspanins (glycosylation and palmitoylation) do not contribute significantly to their presence within typical DRMs as well as within intermediate fractions.
Low- and intermediate-density fractions originate from poorly solubilized microdomains of the plasma membrane
To ascertain the origin of tetraspanins that migrate within low- and intermediate-density fractions, we compared the distribution in sucrose-density gradient of total and surface-membrane proteins by specific labeling of cell-surface proteins before cell lysis and sucrose fractionation. Indeed, tetraspanins are known to accumulate in multivesicular bodies derived from the endocytic machinery, which can be secreted to the external cell medium as exosomes (Escola et al., 1998; Kropshofer et al., 2002; Raposo et al., 1996; Wubbolts et al., 2003). Recently, the lipid composition of exosomes has been found to be highly enriched in cholesterol and sphingomyelin, and to display similar distribution in sucrose-density gradient to those usually attributed to plasma membrane rafts (Stoorvogel et al., 2002). As shown in Fig. 4A (left), both low- and intermediate-density fractions contain biotin-labeled proteins indicating that insoluble proteins originate in part from the plasma membrane. Moreover, after precipitation with avidin-coated beads, biotin-labeled CD9 (Fig. 4A, right) and biotin-labeled CD81 (not shown) display similar distributions on sucrose-density gradients to those of total CD9 and CD81 (CD82 cannot be biotin labeled and therefore could not be analysed).
To confirm that low- and intermediate-density fractions are indeed DRMs, they were analysed by electron microscopy. In two independent experiments, it was found that light fractions (2 and 3) and intermediate fractions (6 and 7) were mainly composed of small membrane vesicles of 50-100 nm, typical of cholesterol-enriched insoluble membranes (Montixi et al., 1998), whereas fractions 4 and 5 or 8 and 9 were mainly devoid of such vesicles (Fig. 4B). Fractions 6 and 7 were also enriched in larger membranes than fractions 2-3.
Altogether, this suggest that the specific tetraspanin distributions in both low- and intermediate-density fractions of sucrose gradient relates to their ability to associate with cholesterol-dependent insoluble membrane domains (i.e. rafts) that originate, at least in part, from the plasma membrane.
Specific tetraspanin density distribution on a sucrose-density gradient is highly dependent on actin polymerization state
Previous studies, as well as the present one (see above), have shown that actin accumulates in raft patches upon cross-linking of raft proteins (Harder and Simons, 1999). Furthermore, tetraspanins and especially CD82 have been shown to display specific interactions with the actin cytoskeleton (Lagaudriere-Gesbert et al., 1998). To investigate potential links between tetraspanins located in raft microdomains and the actin cytoskeleton, we studied the effects of actin-modifying agents on the detergent insolubility of T-cell tetraspanins. Jurkat cells were treated either with F-actin-depolymerizing (latrunculin) or stabilizing (phalloidin) compounds. Latrunculin pretreatment (Fig. 5A) triggered the enrichment of all the tetraspanins tested (CD9, CD81 and CD82) into low-density typical DRM fractions. Interestingly, the F-actin-stabilizing agent phalloidin induced an opposite effect on CD82 (Fig. 5B) and other tetraspanins (not shown). Although the proportions of CD82 recovered in the low, intermediate and dense fractions were, respectively, 6%, 47% and 47% in untreated cells, they changed upon latrunculin or phalloidin treatments to, respectively, 14%, 44% and 42% (Fig. 5A), and 1%, 35% and 64% (Fig. 5B). Similar results were observed for the other tetraspanins, whereas latrunculin did not significantly affect the proportions of LAT and CD71 recovered in each fraction [see the overexposed CD71-WB (Fig. 5A), which allowed us to detect the very small proportion of CD71, less than 3% of total, present in low-density fractions].
It has to be noted that cytoskeleton disruption only marginally reduced the amount of tetraspanins recovered in intermediate fractions. However, owing to the large amount of tetraspanins in such fractions, changes of the same absolute magnitude than those found for low-density fractions would have been difficult to detect.
These results suggest that actin polymerization partially control the location of tetraspanins into typical cholesterol dependent DRMs of low density.
Engagement of CD82 modifies its repartition on sucrose-density gradients
T-cell activation is known to modify both the localization of various molecules within raft domains (Miceli et al., 2001; Xavier et al., 1998), and their binding with the actin-cytoskeleton (Pardi et al., 1992; Rozdzial et al., 1995). Although raft dynamics and cytoskeleton mobilization are both required for efficient T-cell activation, the relationship between these two processes is not well known. The observation that cross-linking of CD82, or other tetraspanins, with antibodies is sufficient to induce the formation of GM1 patches in which polymerized actin also concentrated, suggest that tetraspanins might be involved in cytoskeleton-driven raft dynamics. Moreover, the fractionation results described above also confirm that tetraspanins partially localize to detergent insoluble membranes in an actin-dependent fashion. Because our previous study have revealed that CD82 engagement promotes its own translocation to actin-associated insoluble fractions (Lagaudriere-Gesbert et al., 1998), we analysed how tetraspanin engagement, triggered by culturing Jurkat cells for 15 minutes on anti-tetraspanin-antibody-coated plates, alters their location in cholesterol-dependent DRMs. As shown in Fig. 6A, CD82 engagement significantly reduced the amount of CD82 recovered in low (8% and 5%, respectively, on untreated and stimulated cells) and intermediate-sucrose-density fractions (40% and 22%, respectively, on untreated and stimulated cells), whereas its presence in the soluble protein pool (high-density fractions) was only significantly altered in the last fraction (Fig. 6A). Analysis of the insoluble pellet, in which cytoskeleton components concentrate, revealed that CD82 engagement triggered a clear translocation of CD82 from the membrane-insoluble fraction (low and intermediate densities, 50% decrease of the total amount of CD82) towards the actin-associated insoluble pellets (Fig. 6B, top right, lane CD82).
We have also used oversaturating doses of high-affinity antibodies to study how the engagement of several tetraspanins modifies their own distribution on sucrose-density gradients and those of other tetraspanins. In contrast to the common abilities of CD9, CD81 and CD82 to co-stimulate T cells (Lagaudriere-Gesbert et al., 1997), triggering CD9 or CD81 (expressed at similar levels than CD82) did not alter their own distribution pattern, as was found for CD82 (Fig. 6B, middle and bottom, lanes 2 and 5). By contrast, a weak increase in the amount of CD82 recovered in the insoluble pellet (∼10% of the amount induced by CD82) was observed after engagement of CD9 (Fig. 6B, top right, lane CD9). CD82 engagement did not alter the density distribution of CD9 or CD81. Because the translocation of CD82 within the insoluble pellet is inhibited by actin-depolymerizing agents (Lagaudriere-Gesbert et al., 1998) and is insensible to cholesterol extraction (not shown), it suggests a specific ability of DRM-associated CD82 to associate with condensed F-actin.
MβCD treatment inhibits all CD82 signaling events
Membrane organization in raft domains and cytoskeleton integrity have been shown to be essential for T-cell signaling (Acuto and Cantrell, 2000; Valitutti et al., 1995; van Leeuwen and Samelson, 1999). We therefore examined how tetraspanin signaling depends on the integrity of raft microdomains using Jurkat cells treated with the selective surface-active cholesterol-depleting agent MβCD (Ilangumaran and Hoessli, 1998). Before stimulation by immobilized antibodies, cells were treated for 30 minutes with 5 mM or 10 mM MβCD, doses that lead to partial cholesterol depletion (Xavier et al., 1998) but do not alter cell viability. Such treatment strongly inhibited the CD82-induced adhesion to culture plates (Fig. 7A) whether or not suboptimal doses of anti-CD3, known to increase CD82-induced adhesion, were added (Lagaudriere-Gesbert et al., 1998). Moreover, the few remaining adherent cells failed to develop any membrane extensions, in contrast to the many filopodia and lamellipodia that were observed in untreated cells (Fig. 7B). Similarly, MβCD strongly reduced the ability of CD82 to induce the tyrosine phosphorylation of various proteins (Fig. 7C, left, lanes 2 and 3) and to potentiate TcR signaling (Fig. 7C, right, lane 7-8). As previously described (Xavier et al., 1998), cholesterol depletion also partially inhibited optimal TcR activation (Fig. 7C, lane 3-4). However, upon the extensive TcR cross-linking used in these experiments (oversaturating doses of immobilized anti-CD3), substantial phosphorylations of early TcR intermediates were still observed upon MβCD treatments. Although, as recently reviewed (Munro, 2003), MβCD treatment might have non-specific effects, these results suggest that the amount of CD82 located in classical rafts and intermediate-density cholesterol-dependent DRMs is functionally important for CD82 signaling.
CD82 concentrate with F-actin upon CD82 cross-linking and in the IS upon contact with APCs
Functional effects linked to CD82 have been suggested to mimic the physiological engagement of either CD82 (by an unknown ligand) or of any of its associated partners during T-cell activation. The specific features of CD82 described above prompted us to examine directly how physiological activation of T cells by contact with antigen-pulsed APCs might alter its localization. Indeed, contact with APCs induces the formation of a highly organized contact zone, the IS, which is known to depend strongly upon rafts and cytoskeleton organization. To address this question, redistribution of CD82 and polymerized actin was studied upon contacts between a murine T-cell hybridoma (T8.1) expressing a human-mouse TcR chimera specific for a tetanus toxin peptide (tt830–843) restricted by HLA-DRB1*1102 (Blank et al., 1993) and HLA-DR-expressing murine fibroblasts L625.7 used as APCs (Donnadieu et al., 2001; Michel et al., 1998). Stable expression of YFP-tagged human CD82 in the murine T8.1 cells allowed its visualization. T-cell activation was measured by following intracellular Ca2+ levels. Before contacts, CD82 was either uniformly distributed in resting T cells (Fig. 8A, cell 1) or concentrated in both the front and the back poles of crawling T cells (Fig. 8A, cell 2). Following antigen recognition and establishment of stable contacts with APCs, T cells underwent long-lasting intracellular calcium increases, as previously observed (Donnadieu et al., 1994). Soon after the onset of the calcium response, YFP-CD82 reorganized and concentrated in the IS (Fig. 8B), whereas YFP-free proteins remained diffusely localized in the cytoplasm (not shown). Quantification of the responses indicated that CD82 relocalization was observed in 74±4% of the T cells in contact with APCs (analysis of more than 500 T cells in ten different experiments on living or fixed cells). This proportion is similar to that found for TcRζ relocalization observed under identical conditions in T cells expressing YFP-tagged TcRζ chain (78±2%, data not shown). Consistent with the requirement for the T-cell response, both YFP-CD82 and YFP-TcRζ relocalization were antigen-dependent processes.
Analysis with further magnification was performed by confocal microscopy on cells fixed, permeabilized and co-stained with rhodamine-phalloidin after 20 minutes of contact between T cells and APCs. It revealed that, in 96±4% of the cells, polymerized actin also concentrated also in the cell-to-cell contact area together with CD82. Analysis of many T-cell/APC contact areas perpendicular to the analysis plane suggests that actin and CD82 patches were mainly observed in the periphery of the IS or pSMAC (Fig. 8C), although precise localization would require 3D reconstruction.
Upon TcR-CD3 engagement, the T-cell response depends largely on the co-ligation of various co-receptors. Although CD28 has been recognized as the main co-stimulatory receptor, various molecules, including CD4, integrins, MHCs and different tetraspanins, have been shown to share this co-stimulatory activity (Watts and DeBenedette, 1999). However, the mechanisms underlying these processes are still poorly understood. Various co-stimulators (Miceli et al., 2001; Viola et al., 1999), including the tetraspanin CD9 (Yashiro-Ohtani et al., 2000), have been suggested to be involved in the raft aggregation processes that play a prominent role in T-cell activation. This prompted us to analyse whether tetraspanins could be part of detergent-resistant membrane domains (rafts) in T cells and whether their signaling properties could be modulated by cholesterol depletion.
In this instance, the observation of a colocalization of tetraspanin with coalesced rafts, and flotation experiments indicate that the tetraspanins CD9, CD81 and CD82 are partially located in rafts in resting T cells. A similar partial location in rafts has already been reported by Claas and co-workers for CD151 in fibroblasts (Claas et al., 2001). However, our results suggest that the amount of tetraspanin associated within cholesterol-dependent insoluble domains might be higher than previously thought. After standard isolation procedures that do not alter the actin cytoskeleton, we found more than half of the tetraspanins within insoluble membrane fractions of a sucrose gradient. Only 10% correspond to the classical low-density DRM/rafts from which typical proteins such as LAT are recovered. Most DRM-associated tetraspanins are found in fractions identified as insoluble, poorly solubilized plasma-membrane vesicles of intermediate density, which, in many aspects (sensitivity to cholesterol depletion and tetraspanin engagement), behave like the classical DRMs. This suggests that such intermediate-density fractions correspond to particular type of raft of higher density, presumably owing to higher protein content than classical rafts. Although the occurrence of such tetraspanin-containing high-density rafts has not been previously emphasized, large amounts of CD9 (Claas et al., 2001) and CD151 (Yang et al., 2002) also partition within fractions of intermediate density from lymph-node T cells and epidermoid carcinoma cell lines, respectively. These suggest that partial localization to high-density rafts might be a general property of tetraspanins.
Interestingly similar intermediate-density rafts have recently been reported. Indeed, recent work by Lindwasser and Resh (Lindwasser and Resh, 2001) revealed that, after infection of Jurkat cells, part of the Gag protein of the immunodeficiency virus type 1 (HIV-1) was localized to raft-like domains (`barges') of higher density than typical rafts. The authors suggested that the higher density was linked to the formation of oligomeric Gag-Gag assembly complexes. More recently, Yasui (Yasui et al., 2004) found that, during infection of Burkett's-lymphoma cells with Epstein-Barr virus, the LMP1 viral transmembrane protein is located in similar barges. Previous studies have indicated that tetraspanin complexes than can be precipitated from both light and dense fractions are similar. This led to the suggestion that their ability to associate does not relate to their affinity for raft-like domains. However, only tetraspanin complexes with other proteins (e.g. integrins) have been tested in this regard (Claas et al., 2001). Recent data indicate that, at least for CD9, the detergent conditions that disrupt tetraspanin-tetraspanin interactions abolishes their partition in DRMs (Charrin et al., 2003b). Therefore, different association states of tetraspanin complexes could be responsible for the occurrence of different types of raft domains. The occurrence of such intermediate density DRMs could also be due to the tetraspanin enrichment in raft microdomains linked to cytoskeletal fragments of significant size and therefore of higher density. Indeed, important effects on the tetraspanin density distribution in sucrose gradient were induced by actin-modifying agents. Actin depolymerization significantly increased the amount of tetraspanins recovered within typical DRMs, whereas F-actin stabilization had opposite effects. Although the effect of cytoskeletal alteration on tetraspanin partition was mainly observed in typical rafts, modification of a similar extent would have been difficult to detect in the intermediate-density fractions. Moreover, the presence of polymerized actin in tetraspanin raft domains is also suggested from colocalization studies. Indeed, coalescence of lipid rafts could be triggered not only by cross-linking GM1 gangliosides but also by any of the tetraspanins, and resulted in partial colocalization of GM1, tetraspanins and F-actin. Interestingly, specific effects of another ganglioside, GM3, was found on the distribution of CD82 in low-density membrane domains that are reported to migrate similarly to caveolin and Src kinases (typical raft markers) (Ono et al., 2000; Kawakami et al., 2002).
The biochemical functional data presented here also suggest that CD82 might provide a specific link between tetraspanin raft microdomains with the actin cytoskeleton. Although modifications of the F-actin/G-actin equilibrium affect all tetraspanins similarly, tetraspanin engagement induces a specific alteration of the density of CD82: a large amount of CD82 was translocated from membrane-insoluble fractions to the F-actin-insoluble pellet. Although sharing co-stimulation activity with CD82, neither CD9 nor CD81 were recovered within the insoluble pellet, whatever the stimulation. This suggests a specific ability of CD82 among the tetraspanin family to associate with the actin cytoskeleton, albeit a greater detection sensitivity of CD82 could not be totally excluded. The ability of each tetraspanin to associate with specific partners as well as with each others has led to the concept of the tetraspanin web. Specific interactions have already been established for CD151 with α3β1 integrins (Yauch et al., 2000), and for CD9 and CD81 with EWI-1 and EWI-2 (Charrin et al., 2003a; Charrin et al., 2001; Stipp et al., 2001a; Stipp et al., 2001b). Recent work indicates that CD82 also associates with EWI-2 (Zhang et al., 2003). Although translocation of CD82 to the polymerized actin cytoskeleton has been reported previously (Lagaudriere-Gesbert et al., 1998), one of the new pieces of information contained in the present study is that it is mainly the fraction of CD82 associated with both classical and higher-density rafts that is involved in this process. Moreover, we show that the translocation of the raft-associated CD82 towards the actin-cytoskeleton can be driven not only through CD82 engagement but also, to a lesser extent, by the engagement of other tetraspanins, namely CD9 or CD81. However, CD9 and CD81 themselves do not translocate upon their own engagement, although, as for CD82, they are associated with both classical and higher-density rafts in an actin-polymerization-sensitive fashion. Our results are the first demonstration of an association of rafts and actin that involves the tetraspanin web and, more specifically, CD82 as an intermediate, although the mechanisms involved are still poorly understood. Analysis of the tetraspanin signaling pathway, which is in progress in our laboratory, suggests the weakly polymerized actin constitutively associated with the tetraspanins might help in the formation of signaling platforms and lead to further polymerization or nucleation of the actin cytoskeleton closed to CD82.
The initial observation that interactions of tetraspanins with molecular partners are not dependent upon the raft environment has obscured a possible role of such domains on the tetraspanin functions. Here, we have established for the first time that CD82-induced processes in T cells are critically dependent upon membrane organization. Indeed, all functional effects linked to CD82 engagement, adhesion to culture plates, formation of actin bundles and early events of tyrosine phosphorylations were abolished or strongly reduced by MβCD treatment. Similarly, the ability of CD82 to potentiate and stabilize TcR/CD3 signaling appears to be critically dependent on raft integrity. This suggests that, at least in T cells, tetraspanin function is dependent on membrane organization, possibly owing to their partial localization to organized cholesterol-dependent microdomains. However, rafts do not appear to be the sole determinant of tetraspanin function because, as shown here, different tetraspanins have different behaviors upon activation, in spite of a similar cytoskeleton-modulated location in classical and higher-density rafts in resting cells. This presumably reflects the occurrence of other molecular partners different for each tetraspanins.
Our previous observations showing that CD82 functional effects are also critically dependent on an intact cytoskeleton further support the concept that CD82 might constitute a site of direct or indirect attachment of lipid microdomains with the cytoskeleton, which might be involved in raft reorganization during T-cell activation. To investigate this possibility in more detail, we have analysed whether contact of T cells with APCs might concentrate CD82, as has been shown for lipid rafts. We show that YFP-CD82 concentrates in the IS together with the actin cytoskeleton. The fact that relocalization of CD82 can be triggered by T-cell stimulation with anti-CD3-antibody-coated beads (Delaguillaumie et al., 2002) suggests that raft and cytoskeleton dynamics induced by TcR signaling are sufficient to relocalize CD82. Mittelbrunn et al. (Mittelbrunn et al., 2002) recently found that CD81 also redistributes within the IS. Interestingly, this study revealed a preferential localization of CD81 to the central zone of the synapse (c-SMAC), to which the co-receptors CD4 and CD8, known to interact with CD81, also redistribute. Our results suggest that CD82 localizes to the periphery of the IS (p-SMAC). This indicates that, even if interaction with CD4/CD8 participates in the relocalization of CD81, it does not occur for CD82, even though it is also described to interact with CD4/CD8. These differences might have the same origin as those found upon engagement of the two tetraspanins. In any case, this confirms the idea that, in spite of similar cytoskeleton-modulated location in classical and higher-density rafts in resting cells, tetraspanin might have distinct roles in activated cells.
The specific CD82 dynamics associated with lymphocyte activation is possibly linked to its association with another specific molecular partner. In this hypothesis, our results might involve the interaction of CD82 with the β2 integrin LFA-1. Indeed, both LFA-1 and CD82 have been shown to co-precipitate in Jurkat T-cells and to colocalize upon T-cell stimulation (Shibagaki et al., 1999). LFA-1 and CD82 clustering can be triggered by TcR engagement through cytoskeleton-dependent processes. Moreover, recent data indicate that, similarly to what is found here for CD82, LFA-1 signaling is inhibited by cholesterol depletion (Marwali et al., 2003). Moreover, F-actin-destabilizing agents promote the association of LFA-1 with rafts (Leitinger and Hogg, 2002). The current view of the role of LFA-1 in the IS formation suggest a dynamic link with the actin cytoskeleton (Ardouin et al., 2003; Krawczyk et al., 2002). It has to be noted that, although LFA-1 is supposed to be anchored with the actin cytoskeleton in resting cells, it is not mainly found within the actin-associated pellet. Whether CD82 and LFA-1 dynamics and signaling depend on their reciprocal association and ability to interact with both raft microdomains and cytoskeletal components is under investigation.
In conclusion, our data indicate that tetraspanin molecules partition to the membrane both within and outside raft microdomains that are dynamically linked to the actin cytoskeleton through CD82. It is tempting to suggest that, once tetraspanin dynamics is turned on, it should stabilize the anchorage of CD82 with the actin cytoskeleton. In turn, these novel links between CD82 and polymerized actin might favor the formation of signaling complexes involving Vav and SLP-76, which might drive new actin polymerization through the Rho-GTPase cascade, thereby amplifying the coalescence of lipid rafts and cell signaling. The ability of tetraspanins to interact with many partners offers various possibilities to trigger tetraspanin dynamics: ligand binding on any partner of the tetraspanin web; inside-out signaling through cytoskeleton interaction; or direct binding, as shown recently by the effects induced on CD81 by E2, the envelope protein of hepatitis C virus (HCV) (Crotta et al., 2002; Mittelbrunn et al., 2002; Soldaini et al., 2003; Tseng and Klimpel, 2002). These might explain why these molecules are involved in a large array of cellular functions.
We thank G. Raposo for her help in electron microscopy studies and M. Nedelec for technical support. This work was supported by grants from the Association pour la Recherche contre le Cancer and the Fondation pour la Recherche Médicale.