Tetraspanin CD82 regulates compartmentalisation and ligand-induced dimerization of EGFR

The metastasis suppressor tetraspanin CD82/KAI-1 is a member of the family of ubiquitously expressed fourtransmembrane domain proteins. Although the biochemical function of these proteins is not well established various tetraspanins have been implicated in cell migration, invasion, proliferation and cell-cell fusion (Berditchevski, 2001; Boucheix and Rubinstein, 2001; Hemler, 2001). On the cell surface CD82 is associated with a number of transmembrane receptors and co-stimulatory molecules and regulates their biological activities (Shibagaki et al., 1998; Odintsova et al., 2000). In addition, CD82 forms complexes with other tetraspanins and is assembled in large multiprotein aggregates/clusters (or tetraspanin-enriched microdomains, TERM), which also include various unrelated transmembrane proteins (e.g. integrins, receptor type tyrosine kinases, transmembrane precursors of growth factors). The cellular mechanisms that control segregation of tetraspanins into TERM may involve protein palmitoylation (Charrin et al., 2002; Yang et al., 2002; Berditchevski et al., 2002) and glycosylation (Ono et al., 1999). Recent evidence suggests that tetraspanins play a pivotal role in recruiting the associated receptors in TERM (Berditchevski et al., 2002).

The receptor type tyrosine kinases of the ErbB family (EGFR/ErbB1, ErbB2/neu/HER2, ErbB3 and ErbB4) control various aspects of embryonic development, tissue differentiation and maintenance, and are implicated in tumour progression (Olayioye et al., 2000; Waterman and Yarden, 2001; Yarden and Sliwkowski, 2001). The ErbB receptors are activated by soluble and membrane-anchored ligands of EGF family, which bind to the extracellular part of the proteins and induce activation of the cytoplasmic kinase domain of the receptors. The diversity of signalling events and cellular responses induced by ErbB proteins are regulated at various levels including receptor compartmentalisation in lipid microdomains (Miljan and Bremer, 2002), ligand-induced dimerization (Olayioye et al., 2000; Yarden and Sliwkowski, 2001) and intracellular transport of activated receptors (Waterman and Yarden, 2001). Recent findings strongly suggested that there are multiple ErbB-containing signalling compartments within the plasma membrane. Furthermore, distribution between various surface compartments (controlled by the content of cholesterol and gangliosides in the plasma membrane) may have a dramatic effect on the early events in the ErbB-induced signalling, including ligand binding, ligandinduced dimerization and recruitment of the activated receptors to the clathrin-coated pits (Miljan and Bremer, 2002; Chen and Resh, 2002; Roepstorff et al., 2002; Ringerike et al., 2002). However, the cellular mechanisms that regulate compartmentalisation of the ErbB proteins on the cell surface remain unknown.

We have previously shown that tetraspanin CD82 is associated with EGFR and attenuates EGF-induced signalling (Odintsova et al., 2000). In this report we present evidence that identifies CD82 as a regulator of compartmentalisation and dimerization of the ErbB receptors. Furthermore, our data suggest that ganglioside GD1a is a mediator of the attenuating function of CD82 towards EGFR.

Human mammary epithelial cells HB2/ZEO and HB2/CD82 (Odintsova et al., 2000) were maintained in DME medium (Sigma) supplemented with 10% foetal calf serum (FCS), 10 μg/ml of hydrocortisone and 10 μg/ml of insulin. The human breast carcinoma cell line MCF-7 was grown in RPMI (Sigma) medium supplemented with 10% FCS and minimal essential amino acids solution (Sigma). T47D and 293T cells were maintained in DMEM medium containing 10% FCS. MCF-7/ZEO and MCF-7/CD82 cell lines were generated after transfection of the pZeoSV and pZeoSV-CD82 plasmids (Odintsova et al., 2000) into MCF-7 cells using FuGene 6 (Roche Molecular Biochemicals). Positive clones selected by culturing transfected cells in medium containing 100 μg/ml Zeocin (Invitrogen), were pooled and sorted for positive cells by flow cytometry using anti-CD82 mAb M104. The anti-CD82 mAbs M104 and C33 were kindly provided by Dr O. Yoshie. The anti-CD82 mAbs γC11, γC12 and IA4 were kindly provided by Dr H. Conjeaud. The anti-CD82 mAb TS82b was kindly provided by Dr E. Rubinstein. All anti-EGFR, anti-ErbB2 and anti-ErbB3 mAbs were purchased from Neomarkers. Rabbit polyclonal anti-EGFR and anti-ErbB3 Ab were purchased from Autogen Bioclear. Rabbit polyclonal Ab to caveolin-1 was from Transduction Laboratories. Mouse anti-EMMPRIN mAb 8G6 has been described previously (Berditchevski et al., 1997). Mouse mAbs to gangliosides GD1a and GT1b were kindly provided by Dr R. Schnaar. Mouse mAb to ganglioside GM3 (DH2) was purchased from Glycotech. FITC-conjugated cholera toxin was purchased from Sigma.

rs-LECL-CD82 and rs-EMMPRIN were produced in 293T cells. pIg-CD82-Fc and pIg-EMMPRIN-Fc plasmids were constructed using a standard overlapping PCR protocol. The large extracellular loop of CD82 was `fused' in frame with the murine Ig kappa-chain secretion signal at the N terminus and cloned (HindIII-BamHI) into pIG-1 vector (a gift from C. Buckley, University of Birmingham, UK). The extracellular domain of EMMPRIN (17) containing the leader sequence was amplified and cloned (HindIII-BamHI) into pIG-1 vector. The cells were transfected with the pIG-CD82-Fc and pIg-EMMPRIN-Fc plasmids and the growth medium conditioned by the cells was collected 7-10 days later for purification. The recombinant soluble proteins were purified from the conditioned medium by affinity purification on protein A conjugated to agarose beads.

The proteins were solubilised into the immunoprecipitation buffer containing 1% Brij98/PBS (or 1% Triton X-100/PBS), 2 mM phenylmethylsulphonyl fluoride, 10 μg/ml aprotinin, 10 μg/ml leupeptin for 2 hours at 4°C. The insoluble material was pelleted at 7000 g for 10 minutes. The cell lysates were then precleared by incubation for 2 hours at 4°C with agarose beads conjugated with goat anti-mouse antibodies (mIgG-beads; Sigma). Immune complexes were collected using appropriate mAbs prebound to mIgG-beads and washed four times with the immunoprecipitation buffer. The complexes were eluted from the beads with Laemmli sample buffer. Proteins were resolved in 6-10% SDS-PAGE, transferred to a nitrocellulose membrane and developed with the appropriate Ab. Protein bands were visualised using horseradish peroxidaseconjugated secondary antibodies (Sigma) and Enhanced Chemiluminescence reagent (Amersham Pharmacia Biochem).

Cells grown on two 100 mm² tissue culture plates (∼7-8×10⁶ cells) were washed three times with ice-cold PBS and incubated with ∼50-100 ng/ml ¹²⁵I-ligand (EGF, TGFα, HRG) in serum-free DMEM/0.1% BSA for 2 hours at 4°C with regular rocking. Subsequently, the medium was discarded and cells were carefully washed with ice-cold PBS, to remove non-bound ligand. Then the cells were incubated with 0.5 mM non-permeable cross-linking reagent, bis(sulphosuccinimidyl) suberate [BS³(Pierce)], at 4°C for 1 hour. The reaction was quenched by further incubation with 50 mM Tris-HCl, pH 7.5. The cells were washed with ice-cold PBS and lysed overnight in lysing buffer based on 1% Triton X-100. Cell lysates were precleared by incubation for 2 hours at 4°C with mIgG-beads. Immune complexes were collected on protein A- or mIgG-beads that were preincubated with monoclonal or polyclonal antibodies against respective ErbB receptors. After washing with the immunoprecipitation buffer the complexes were eluted from the beads with Laemmli sample buffer, loaded onto a 6% polyacrylamide gel and subjected to SDS-PAGE. The gel was dried and exposed to X-ray film (Kodak) at –70°C.

Cells were lysed for 10 minutes in ice-cold 1% Brij98/MES (25 mM, pH 6.5) buffer supplemented with 100 μM Na₃VO₄, 5 mM NaF, 10 mM Na₄P₂O₇, 10 μg/ml aprotinin, 10 μg/ml leupeptin and 2 mM PMSF. Lysates (derived from 9×10⁶ cells) were passed successively twenty times through the 25G hypodermic needles and 800 μl of lysate was mixed with two volumes of 2 M sucrose solution. This layer was overlaid with 200 μl layers of decreasing concentration of sucrose – from 0.9 M to 0.2 M. Samples were centrifuged at 100,000 g for 16-18 hours at 4°C in a Beckman SW60 rotor, and 9 fractions each of 200 μl were then collected from the top of gradient. The pellet was suspended in 200 μl of the lysis buffer. Equal amounts of each fraction were mixed with 4× Laemmli loading buffer and the proteins were resolved in 10% SDS-PAGE.

Cells were grown on glass coverslips in complete media for 24-36 hours. Spread cells were fixed with 2% paraformaldehyde/PBS for 10-15 minutes. Staining with primary and fluorochrome-conjugated secondary antibodies was carried out as previously described (Berditchevski and Odintsova, 1999). Staining was analysed using a Nikon Eclipse E600 microscope. Images were acquired using a Leica DC200 digital camera and subsequently processed using a DC200 image processing programme.

Cells were incubated with saturating concentrations of primary mouse mAbs for 45 minutes at 4°C, washed twice and then labelled with fluorescein isothiocyanate (FITC)-conjugated goat anti-mouse immunoglobulin. Stained cells were analysed using Coulter Epics programme (Becton Dickinson).

Previously, we found that tetraspanin protein KAI1/CD82 is associated with EGF receptor in various cell types and attenuates its signalling by facilitating ligand-induced endocytosis of the receptor (Odintsova et al., 2000). In that report we proposed that CD82 could be involved in the regulation of steps that precede internalisation of EGFR. Ligand-induced dimerization of EGFR is an early event leading to auto- and cross-phosphorylation of the receptors and further propagation of signalling. To investigate whether CD82 affects ligand-induced formation of ErbB dimers we first compared the dimerization of the receptors in mammary epithelial cells HB2/ZEO and HB2/CD82 (Odintsova et al., 2000) in response to EGF. (We have previously shown that the expression level of CD82 in HB2/CD82 cells exceeded that in HB2/ZEO cells by ∼15 fold and was comparable to the expression found in other CD82-positive cells, e.g. primary human keratinocytes.) The results of western blotting indicate that HB2/ZEO and HB2/CD82 cells express comparable numbers of EGFR, ErbB2 and ErbB3 receptors (Fig. 1A). The cells were incubated with ¹²⁵I-EGF at 4°C and formed ErbB/¹²⁵I-EGF complexes that were stabilised further after treatment with a membrane-impermeable chemical crosslinker, BS³. Further immunoprecipitation experiments have shown that ligand binding induced formation of homodimers (EGFR/EGFR) and heterodimers (EGFR/ErbB2) in both HB2/CD82 and HB2/ZEO cells (Fig. 1B, top panel, lanes 1, 2, 4 and 5). In contrast, although anti-ErbB3 Ab precipitated the receptor from both cell lines (Fig. 1A, bottom panel) no radioactive signal was detected in the immunoprecipitates (Fig. 1B, top panel, lanes 3 and 6). It is probable that in HB2 cells, which express high levels of ErbB2 [a preferable partner for EGFR (Graus-Porta et al., 1997; Ogiso et al., 2002; Garrett et al., 2002)], the number of EGFR-ErbB3 dimers formed was insignificant and below the sensitivity of the assay. Notably, we observed that the number of dimers was significantly lower in CD82-expressing cells than in the control cells (Fig. 1B, top panel, compare lanes 1 and 4 or 2 and 5). The differences were particularly pronounced when dimers were immunoprecipitated using anti-ErbB2 antibodies (Fig. 1B, top panel, lanes 2 and 5). The densitometric analysis has shown that ErbB2-containing dimers (a protein band of 350 kDa in the ErbB2 immunoprecipitates) were more than nine times more abundant in HB2/ZEO cells than in HB2/CD82 cells. In contrast, the differences between the cell lines in the EGFR homodimers were only 1.4 fold (Fig. 1B, top panel, lanes 1 and 4). Similar differences were observed when cells were stimulated with ¹²⁵I-TGFα. Specifically, the amount of dimers pulled down by anti-EGFR, and particularly by anti-ErbB2, Abs was significantly diminished in HB2/CD82 cells: a 1.25-fold decrease in the EGFR immunoprecipitate and an 5-fold decrease in the ErbB2-immunoprecipitate (Fig. 1B, lower panel, lanes 2 and 5). These results clearly indicate that CD82 has a negative effect on the ligand-induced dimerization of EGFR.

The crystal structure of ligand-bound EGFR indicates that the extracellular domain plays a critical role in homodimerization of the receptor (Ogiso et al., 2002; Garrett et al., 2002). The structure revealed that ligand binding generates a conformational change in the receptor extracellular domain and leads to the exposure of a `dimerization arm' that subsequently interacts with a similar part on a dimerization partner (Ogiso et al., 2002; Garrett et al., 2002). Although, the proposed model does not offer a mechanism for the formation of heterodimers involving the ligand-less ErbB2, it has been recently established that the dimerization loop of this protein is maintained in the constitutively `open' configuration thus making it a preferable dimerization partner for other ErbB receptors (Garrett et al., 2003). Given the crucial role played by the tetraspanin large extracellular loop (LECL) in regulation of the biological activities of the associated protein partners (Baudoux et al., 2000; Berditchevski, 2001; Boucheix and Rubinstein, 2001; Hemler, 2001) we hypothesised that this part of the protein may be important in conferring the negative effect of CD82 on the EGFR-ErbB2 dimerization. To test this hypothesis we carried out dimerization experiments in the presence of a recombinant soluble protein containing LECL of CD82 (rs-LECL-CD82). Firstly, we established that rs-LECLCD82 (produced in 293T cells) could be immunoprecipitated with a panel of four anti-CD82 mAbs (Fig. 1C). This suggests that overall folding of LECL-CD82 is similar to that of a native protein. For the control experiments we also produced a recombinant soluble protein containing the extracellular domain of EMMPRIN (rs-EMMPRIN), an unrelated transmembrane protein (Biswas et al., 1995; Berditchevski et al., 1997). Cells were pre-incubated with rs-LECL-CD82 (or rs-EMMPRIN) and subsequently stimulated with ¹²⁵I-EGF in the presence of the recombinant-soluble proteins (Fig. 1D). As expected, in the control experiments (i.e. no recombinant soluble protein added) a number of EGFR-ErbB2 dimers formed in HB2/CD82 cells was significantly lower than in HB2/ZEO cells (Fig. 1D, compare lanes 1 and 5 in top and bottom panels). However, we also found that the presence of rs-LECL-CD82 (or rs-EMMPRIN) did not affect the formation of EGFR-ErbB2 dimers in either HB2/ZEO or HB2/CD82 cells (Fig. 1D, top and bottom panels, lanes 2-4 and 6-8). Although these results do not rule out completely a possible involvement of LECL-CD82 in the dimerization they indicate that the other part(s) of the protein may be critical for interference with the dimerization process.

There is evidence that in some cells ErbB proteins undergo spontaneous dimerization in the absence of ligand (Yu et al., 2002). This can prime the receptors to more efficient ligand binding and, in turn, facilitate ligand-induced dimerization. To investigate whether expression of CD82 affects ligandindependent dimerization, serum-starved HB2/ZEO and HB2/CD82 cells were pre-treated with a membraneimpermeable chemical crosslinker and EGFR- and ErbB2-containing complexes were purified using an immunoprecipitation protocol. Formation of EGFR-ErbB2 dimers was subsequently examined by western blotting. As illustrated in Fig. 2 there was a small number of the pre-formed ErbB homodimers in both HB2/ZEO and HB2/CD82 cells (detected as proteins of a higher molecular mass in the EGFR and ErbB2 immunoprecipitates; lanes 1, 2 and 7, 8). However, the densitometric analysis of the signals showed no apparent differences between the cell lines. Furthermore, no pre-formed EGFR-ErbB2 heterodimers were detected in either HB2/ZEO or HB2/CD82 cells (Fig. 2, lanes 3, 4 and 5, 6). These data indicate that CD82 only affects ligand-induced dimerization.

To examine whether CD82 influences ligand-induced dimerization of the ErbB receptors, other than EGFR, we carried out the dimerization assay using a pair of newly established cell lines, MCF-7/CD82 and MCF-7/ZEO. Parental MCF-7 cells express high levels of ErbB2 and ErbB3 but no EGFR or CD82. Thus, by stimulating the transfectants with ¹²⁵I-HRG, a well-established ligand for ErbB3, we were able to examine the effect of the tetraspanin on formation of the ErbB2-ErbB3 heterodimers. MCF-7/ZEO and MCF-7/CD82 cells pre-incubated with the labelled ligand were treated with BS³ and ErbB complexes were recovered from the cellular lysates using either anti-ErbB2 or anti-ErbB3 Abs (expression levels of both receptors are illustrated in Fig. 3A). As illustrated in Fig. 3 formation of ¹²⁵I-HRG-ErbB dimer complexes was detected in both cell lines (the upper protein band in the ErbB2 and ErbB3 immunoprecipitates). However, the amount of immunoprecipitated dimers was similar for MCF/ZEO and MCF-7/CD82 cells (Fig. 3B, compare lanes 1 and 2, and 3 and 4). These results indicated that although CD82 has a negative role in formation of EGFR-ErbB2 dimers, ligand-induced dimerization of ErbB3 with ErbB2 is not affected by this tetraspanin.

To examine the mechanisms underlying functional selectivity of CD82 towards ligand-induced formation of various ErbB heterodimers we analysed the association of CD82 with ErbB2 and ErbB3 in HB2/CD82 and MCF7/CD82 cells. Fig. 4 illustrates that ErbB3 could be coimmunoprecipitated by anti-CD82 antibodies from protein lysates prepared from both cell lines (Fig. 4B, lanes 2, 8). In addition, we found that CD82 forms complexes with ErbB3 in T47D cells, which express both of these proteins endogenously (Fig. 4B, lane 5). In contrast, the association of CD82 with ErbB2 was cell-type specific (Fig. 4B): it could be detected in MCF-7/CD82 cells but not in HB2/CD82 or T47D cells (Fig. 4B, top panels, lanes 2, 8 and 5, respectively). In the control experiments we showed that anti-CD82 mAb efficiently precipitated the protein from all cell lines (Fig. 4B, lower panels, lanes 2, 5, 8). Furthermore, the level of expression of ErbB2 in T47D and HB2/CD82 is similar or higher of that detected in MCF-7/CD82 cells (Fig. 4A). These results indicate that the assembly of CD82-ErbB2 and CD82-ErbB1 complexes is controlled by distinct mechanisms. Although the molecular mechanisms that underlie these differences remain to be established one possibility is that EGFR exerts a negative effect on the formation of the CD82-ErbB2 complex (as observed for HB2/CD82 and T47D cells). Previously we reported that CD82/EGFR complex was destabilised in the presence of EGF (Odintsova et al., 2000). In contrast, we have found that stability of ErbB2-CD82 and ErbB3-CD82 complexes in MCF-7/CD82 cells is not affected after HRG treatment (Fig. 4C). Similar amounts of ErbB3 and ErbB2 were precipitated by anti-CD82 mAb from unstimulated or HRG-treated MCF-7/CD82 cells (Fig. 4C, compare lanes 1 and 2, 4 and 5, 6). These data demonstrate that ligand binding has a differential effect on the stability of various CD82-ErbB complexes. Taken together, our results demonstrate that the association itself does not predicate the negative effect of CD82 on ligand-induced dimerization of ErbB receptors.

Recent findings suggest that there are multiple EGFR-containing signalling compartments within the plasma membrane (Mineo et al., 1999; Waugh et al., 1999; Miljan and Bremer, 2002). Furthermore, membrane compartmentalisation has a significant effect on the ligand-induced dimerization and signalling potential of activated EGFR. To investigate whether expression of CD82 changes the membrane microenvironment of EGFR and ErbB2 we fractionated HB2/CD82 and HB2/ZEO cellular lysates in a sucrose gradient. The distribution of proteins in the gradient fractions was assessed by immunoblotting. Our data have shown that EGFR from HB2/ZEO cells is distributed almost evenly between the gradient fractions 5, 6, 7, 8 and 9 (Fig. 5): densitometric measurements of the blots indicate that the combined amount of EGFR in these fractions is over 85% of the total amount of the protein detected in all fractions of the gradient. The amount of the receptor detected in fraction 4 is only 8.5% of the total protein and no signal was detected in the first three fractions of the gradient. However, we have consistently observed (in three independent experiments) that in HB2/CD82 cells the distribution of EGFR has been shifted to the left towards light fractions of the gradient. Over 50% of EGFR was concentrated in fractions 4, 5 and 6 (Fig. 5), with fraction 4 containing the largest amount of the receptor (over 20%). Furthermore, approximately 7% of the total protein could be detected in the third fraction of the gradient. CD82 has a similar effect on the fractional distribution of ErbB2: the protein from HB2/ZEO cells was abundant in fractions 5 and 6 (∼24% and ∼22% of the total amount, respectively). No signal was detected in the gradient fraction 3 and less than 9% of the protein was found in fraction 4. In contrast, similar amounts (∼20%) of ErbB2 from HB2/CD82 cells were detected in both fraction 4 and 5 of the gradient. In addition, more than 8% of the protein was found in fraction 3 of the gradient. Given the fact that ErbB2-CD82 complex is not detectable in HB2/CD82 cells (Fig. 4B), we concluded that the alteration of floating characteristics of ErbB2 was independent of its association with the tetraspanin. In fact, no CD82 was detected in the fraction 3 and less than 5% of the protein was found in fraction 4 of the gradient (Fig. 5). Thus, we propose that the expression of CD82 causes global changes in physico-chemical characteristics of the plasma membrane, which has a general impact on the organisation of various microdomains. Indeed, the fractional distribution of caveolin, a cholesterol-associated membrane protein that is not linked to tetraspanin microdomains, was also affected by CD82 (Fig. 5). In the control experiments we found that the fractional distribution of EMMPRIN was not influenced by the expression of CD82 (Fig. 5). These results suggest that CD82 only affects membrane compartmentalisation of proteins that are associated with lipid rafts.

To further illustrate differences in membrane compartmentalisation of EGFR we analysed surface distribution of the receptor by immunofluorescence staining. In the majority of HB2/ZEO cells (>70%) fine clusters of EGFR were evenly distributed over the cell surface (Fig. 6A). In contrast, in more than 80% of HB2/CD82 cells EGFR was more concentrated towards the cell periphery (Fig. 6B). Gangliosides play an important role in regulation of signalling through EGFR (Miljan and Bremer, 2002). We examined whether the effect of CD82 on compartmentalisation and dimerization of EGFR correlates with surface distribution of gangliosides. Flow cytometry experiments have shown that out of four gangliosides that are known to influence EGF-dependent signalling (i.e. GM3, GM1, GD1a and GT1b) (Zurita et al., 2001; Miljan and Bremer, 2002; Mirkin et al., 2002), only GM1 and GD1a could be detected on the surface of HB2/ZEO and HB2/CD82 cells (Table 1). Notably, expression of both GM1 and GD1a was significantly higher in HB2/CD82 cells with more dramatic difference observed for GD1a (∼2.8-fold increase). This observation was confirmed further in immunofluorescence experiments (Fig. 6C,D). Interestingly, in HB2/CD82 cells a significant proportion of GD1a clusters re-localised towards cell margins and contained CD82 (Fig. 6D,G,H) and EGFR (Fig. 6I,J). In contrast, cellular distribution of ganglioside GM1 was similar in HB2/ZEO and HB2/CD82 cells (Fig. 6E,F). We observed similar changes in the surface expression (∼2.5 fold) and distribution of GD1a in MCF-7/CD82 cells (Fig. 7 compare A and B). Furthermore, a significant proportion of GD1a in these cells was co-localised with CD82 (Fig. 7B-D). Taken together these results confirm that CD82 causes re-distribution of EGFR and GD1a, and suggest that ganglioside GD1a may function as mediators of activity of the tetraspanin towards the receptor.

The pattern of ligand-induced dimerization of ErbB receptors dictates the physiological responses of cells to stimulation with EGF or related growth factors (Olayioye et al., 2000; Waterman and Yarden, 2001; Yarden and Sliwkowski, 2001). Here we show that CD82 has a specific negative effect on ligand-induced dimerization of EGFR. Thus, our results ascribe to CD82 a crucial and specific function in regulation of signalling via ErbB receptors. Furthermore, the finding that CD82 has a prominent negative effect on ligand-induced dimerization of EGFR with ErbB2 provides a mechanistic explanation for our previous observation of the increased rate of internalisation of ligand-activated EGFR in HB2/CD82 cells (Odintsova et al., 2000). In this regard, earlier study has shown that the association with ErbB2 decreases endocytosis of EGFR (Wang et al., 1999). Our data clearly demonstrate that the association with ErbB receptors does not warrant the effect of CD82 on ligand-induced dimerization. Indeed, although CD82 is associated with ErbB2 and ErbB3 in MCF-7 cells and modifies heregulin-induced signalling (E.O. and F.B., unpublished results), it does not affect ErbB2-ErbB3 dimerization upon stimulation with HRG. These results suggest that the mechanism of action of CD82 does not involve sterical interference and that CD82 has a more active role in the dimerization process.

The effect of CD82 on dimerization correlates with redistribution of EGFR and ErbB2 in cells (detected by sucrose gradient fractionation and immunofluorescence staining). We propose that this observation is indicative of a novel function of CD82 as a regulator of compartmentalisation of proteins associated with the plasma membrane. Furthermore, our results show that the effect of CD82 on compartmentalisation is not limited to the tetraspanin-associated proteins (e.g. compartmentalisation of caveolin is also affected in HB2/CD82 cells). Not only does this observation distinguish CD82 from CD151 (a tetraspanin that specifically regulates compartmentalisation of the integrin α3β1) (Berditchevski et al., 2002), but it also suggests that the elevated expression of CD82 (and consequent changes in compartmentalisation) may significantly affect the function of a broad range of transmembrane receptors. In this regard, it has been previously reported that the elevated expression of CD82 in T cells stimulated integrin-mediated cell-cell adhesion (Shibagaki et al., 1999). Although the effect of the tetraspanin on compartmentalisation of integrins has not been analysed, recent results clearly demonstrated that localisation in lipid rafts has a key role in regulation of integrin functions (Leitinger and Hogg, 2002).

Compartmentalisation of EGFR in microdomains on the surface membrane has a significant impact on the early events associated with the receptor activation. These include ligandbinding, dimerization and re-localisation of activated receptors into clathrin-coated pits (Mineo et al., 1999; Carpenter, 2000). Although the mechanistic link between signalling via EGFR and its localisation in microdomains is not established, one possibility is that localisation in microdomains affects its spatial orientation relative to other ErbB proteins, an important factor in ligand-induced dimerization of the receptors. Recent results clearly established that lipid microenvironment plays an important role in signalling via EGFR (Miljan and Bremer, 2002; Chen and Resh, 2002; Pike and Casey, 2002; Roepstorff et al., 2002; Ringerike et al., 2002). For example, EGFR has been found in the cholesterol-enriched microdomains, and pharmacological depletion of cholesterol profoundly increased ligand-induced dimerization and activation of the receptor (Chen and Resh, 2002; Pike and Casey, 2002; Roepstorff et al., 2002; Ringerike et al., 2002). EGF-induced signalling is also affected by gangliosides (reviewed by Miljan and Bremer, 2002). Various gangliosides were shown to interact with the extracellular domain of EGFR and either inhibit or stimulate ligand-induced activation of the receptor (Miljan et al., 2002). This association is dependent on glycosylation of EGFR (Wang et al., 2001b), which in turn, regulates the conformation of the extracellular domain and ligand-induced dimerization of the receptor (Bishayee, 2000; Tsuda et al., 2000; Fernandes et al., 2001). In this regard, we found that expression of CD82 had a marked effect on surface expression of gangliosides GM1 and GD1a (Table 1 and Fig. 6). It remains to be established how CD82 affects expression of gangliosides. One possibility is that CD82 targets one of the glycosyltransferases responsible for biosynthesis of gangliosides. In cells the activity of various glycosyltransferases, residents of Golgi apparatus can be regulated at the transcriptional (Kolter et al., 2002) and posttranscriptional levels (Yu and Bieberich, 2001). For example, both N-acetylgalactosaminyltransferase (GalNAcT) and sialyltransferase IV (SAT IV), an enzyme that converts GM1 into GD1a, are regulated by phosphorylation (Yu and Bieberich, 2001). Furthermore, it has been established that the activity of GalNAcT is regulated via the cAMP-dependent mechanisms. However, SAT IV is associated with a 33kD isoform of 14-3-3 proteins (Gao et al., 1996), well established participants of various signalling pathways (Yaffe, 2002). Alternatively, CD82 may affect the activity (or expression level) of membrane sialidases. Interestingly, it has been reported that elevated expression of Neu3, a lipid raftassociated sialidase, enhanced signalling via EGFR (Wang et al., 2001a). Finally, it is possible that CD82 stabilises GD1a and GM1 on the cell surface by suppressing shedding of these gangliosides from the plasma membrane or their internalisation (Dolo et al., 2000).

Several lines of evidence suggest that GD1a plays a major role in the attenuating activity of CD82 towards EGFR in HB2 cells. Firstly, expression of CD82 specifically affected surface distribution of GD1a whereas distribution of GM1 has not changed. Secondly, whereas a significant proportion of GD1a clusters also contained EGFR, co-localisation of EGFR with GM1 was less apparent. Finally, it has been recently demonstrated that GD1a and not GM1 can inhibit EGF-induced signalling in intact cells (Mirkin et al., 2002). Further experiments will be needed to establish the connection between the GD1a-enriched microdomains and a modulatory effect of CD82 on other transmembrane receptors (e.g. integrins, T-cell receptor complex). In this regard it is important to emphasize that localisation of transmembrane receptors in lipid-ordered microdomains may have opposing effects on the receptors' activity. Whereas disruption of the cholesterol-enriched microdomains potentiates signalling via receptor tyrosine kinases (Chen and Resh, 2002; Pike and Casey, 2002; Roepstorff et al., 2002; Ringerike et al., 2002), it inhibits the signalling function of the FAS/CD95 complex (Hueber et al., 2002). Similarly, in various cell types both GD1a and CD82 can either negate (Van Brocklyn et al., 1993; Odintsova et al., 2000; Mirkin et al., 2002) or potentiate (Zuberbier et al., 1995; Shibagaki et al., 1999; Lang et al., 2001) transmembrane signalling triggered by different types of the receptors.

How can CD82 influence membrane compartmentalisation of GD1a and EGFR? Firstly, CD82 may directly regulate the distribution of these molecules. Being palmitoylated on multiple sites (Charrin et al., 2002; Yang et al., 2002; Berditchevski et al., 2002) CD82 has an inherent tendency to coalesce in the lipid-ordered microdomains assembled within the inner leaflet of the plasma membrane. Inevitably, this would have an affect on distribution and dynamics of CD82-associated molecules (e.g. EGFR). Interestingly, it has been reported that when added exogenously to colorectal cancer cells ganglioside GM3 could form complexes with the tetraspanin CD9 (Ono et al., 1999). Furthermore, it has been proposed that this association is critical for anti-migratory function of CD9 (Ono et al., 2001). Although physical interaction between CD82 and GD1a is yet to be established, our data show that this ganglioside is present in the CD82-containing clusters on the cell surface. This observation suggests that CD82 may function as a molecular linker that physically connects GD1a-enriched microdomains of the outer leaflet with lipid-ordered domains of the inner leaflet of the membrane. Secondly, CD82 is known to associate with various protein kinase C isoforms (Zhang et al., 2001), enzymes that have a short-term influence on the dynamics of glycolipid domains in the plasma membrane (Pitto et al., 1999) and regulate recruitment of cellular proteins into lipid-ordered microdomains (Parolini et al., 1999). Finally, CD82 is associated with and may modulate the activity of CD36 (E.O. and F.B., unpublished results), a transmembrane protein that functions as a fatty acid and cholesterol transporter (Febbraio et al., 2001). This may have a general impact on the lipid content of the plasma membrane and, consequently, distribution of GD1a and EGFR on the cell surface.

In summary, we have shown that CD82 attenuates the activity of EGFR by redistributing the receptor into the GD1a-containing microdomains on the cell surface. Not only do these results link the function of CD82 with a particular ganglioside, but they also provide a new mechanistic insight into the role of the tetraspanin in a signalling process.

We are grateful to Drs. O. Yoshie, E. Rubinstein, R. Schnaar and H. Conjeaud for providing mAbs and human CD82 cDNA. Work was supported by a grant from the Breast Cancer Campaign (to E.O. and F.B.).