We previously showed that the association of CD4 and GM3 ganglioside induced by CD4 ligand binding was required for the down-regulation of adhesion and that aggregation of ganglioside-enriched domains was accompanied by transient co-localization of LFA-1 (lymphocyte function-associated antigen-1), PI3K (phosphoinositide 3-kinase) and CD4. We also showed that these proteins co-localized with the GM1 ganglioside that partially co-localized with GM3 in these domains. In the present study, we show that CD4–p56lck association in CD4 signalling is required for the redistribution of p56lck, PI3K and LFA-1 in ganglioside-enriched domains, since ganglioside aggregation and recruitment of these proteins were not observed in a T-cell line (A201) expressing the mutant form of CD4 that does not bind p56lck. In addition, we show that although these proteins associated in different ways with GM1 and GM3, all of the associations were dependent on CD4–p56lck association. Gangliosides could associate with these proteins that differ in affinity binding and could be modified following CD4 signalling. Our results suggest that through these associations, gangliosides transiently sequestrate these proteins and consequently inhibit LFA-1-dependent adhesion. Furthermore, while structural diversity of gangliosides may allow association with distinct proteins, we show that the tyrosine phosphatase SHP-2 (Src homology 2 domain-containing protein tyrosine phosphatase 2), also required for the down-regulation of LFA-1-dependent adhesion, transiently and partially co-localized with PI3K and p56lck in detergent-insoluble membranes without association with GM1 or GM3. We propose that CD4 ligation and binding with p56lck and their interaction with GM3 and/or GM1 gangliosides induce recruitment of distinct proteins important for CD4 signalling to form a multimolecular signalling complex.
CD4 signalling, in the absence of TCR (T-cell receptor) engagement, is a critical event in the down-regulation of LFA-1 (lymphocyte function-associated antigen-1)-dependent adhesion between T- and B-cells. We have previously reported that CD4-induced down-regulation requires intermediary proteins between CD4 and LFA-1 such as CD4-associated tyrosine kinase p56lck, PI3K (phosphoinositide 3-kinase) and SHP-2 (Src homology 2 domain-containing protein tyrosine phosphatase 2) . In addition, given the rapid and transient complex of signalling proteins induced by CD4 signalling and signal inhibition by cholesterol removal, we suggested that these proteins were recruited into different membrane compartments . Lateral heterogeneity in the classic fluid-mosaic model of cell membranes indicates that the plasma membrane contains distinct microdomains that are enriched in cholesterol, GPI (glycosylphosphatidylinositol)-linked glycoproteins , and glycosphingolipids such as GM1 and GM3 which are involved in modulating signal transduction . Asymmetric distribution of T-cell rafts containing GM1 and GM3 glycosphingolipids indicates that rafts have distinct functions and play a role in several signalling processes involving receptors and integrins. Isolation of DRMs (detergent-resistant membranes) has proved to be a valuable tool for the analysis of lipid rafts and a useful starting point for defining membrane subdomains and their composition [5,6], even though DRMs do not necessarily represent all the rafts in living cells. However, although detergent extraction disrupts lipid–lipid interactions, and lipid–protein interaction, a minor fraction of cell membrane is preserved. Only those proteins that are strongly interacting with highly ordered domains retain their association with lipids and are recovered in DRMs. These DRMs or rafts allow lateral segregation of proteins and provide a mechanism for the compartmentalization of signalling components by concentrating or excluding certain components such as CD45 that inhibit T-cell signalling [7,8]. Thus lipid rafts may function as platforms to control the localization and function of proteins for the formation of multicomponent transduction complexes. The role of lipid rafts in T-cell signalling has also been emphasized given that disruption and/or displacement of signalling molecules abolishes TCR-mediated signalling events . In lymphocytes, GM3 and cholesterol, along with the Src-family kinase p56lck and a fraction of the CD4 pool, are acylated by saturated fatty acids and are selectively recovered in the rafts where they associate [10,11]. Mutant versions of p56lck and CD4 that are not palmitoylated and consequently not localized within rafts, do not function in T-cell activation [12,13]. Polarization and lipid raft recruitment to the receptor engagement site in immune cells following solid-phase engagement of the CD4 molecule alone, and subsequent redistribution of different proteins such as integrins, requires p56lck signalling . Raft localization of integrins appears to be correlated with integrin activity, since inactive integrins such as LFA-1 are tethered away from lipid rafts by cytoskeletal restraints . Other proteins such as SHP-2 and PI3K can affect both within raft-integrin localization and function. SHP-2 associates with adhesion molecules  and SHP-2 partitioning in raft domains triggers integrin-mediated signalling . SHP-2 also associates with F-actin  and plays a role in cytoskeletal organization, cell adhesion and migration . PI3K can also be recruited in raft domains following activation, and disruption of these domains inhibits the PI3K pathway . PI3K, through its different associations with protein kinase Cζ, RhoA or cytohesin-1, is also important for integrin regulation [21,22] as well as for the polar redistribution of LFA-1 induced by chemokines .
We have previously reported that lipid raft integrity is required for the down-regulation of Ag-independent LFA-1-dependent adhesion induced by CD4 ligands, and more precisely by an anti-CD4 Ab (antibody; 13B8.2) that specifically binds the N-terminal CD4 domain (D1) . The fact that CD4 triggering induces aggregation of GM1- and GM3-enriched raft domains that are partially co-localized suggests that GM1 and GM3 may also be localized in distinct domains and may associate with distinct proteins. Raft involvement in CD4-triggered events has been strengthened by the results showing that the CD4–GM3 association induced by CD4 ligand binding was required for the down-regulation of LFA-1-mediated adhesion . Raft aggregation was accompanied by transient co-localization of LFA-1, PI3K and CD4 in these domains. We investigate here whether LFA-1, PI3K and SHP-2 co-localize and associate with gangliosides after CD4 ligation and whether gangliosides act as intermediary partners for the sequential association of these different proteins. In addition, since p56lck associated with CD4 was also required for the down-regulation of Ag-independent LFA-1-dependent adhesion induced by CD4 ligands , we also investigate whether CD4–p56lck association is also required for the LFA-1, PI3K and SHP-2 co-localization and association with gangliosides.
MATERIALS AND METHODS
Abs and reagents
The following Abs were used: 13B8.2, 25.3 [IgG1, anti-CD4 and anti-LFA-1α mAbs (monoclonal Abs) respectively], anti-HLA (human leucocyte antigen) class I Ab from Immunotech, anti-p85-PI3K polyclonal Ab (Upstate Biotechnology), anti-p56lck polyclonal Ab and anti-SHP-2 mAb (Santa Cruz Biotechnology). F(ab)′2 GAMIg (goat anti-mouse Ig; Jackson ImmunoResearch Laboratories) was used for cross-linking experiments. For confocal immunofluorescence experiments, the following Abs were used: TRITC (tetramethylrhodamine β-isothiocyanate)-conjugted GAMIg and FITC-conjugated sheep anti-rabbit IgG (Jackson ImmunoResearch Laboratories), and anti-GM1 and anti-GM3 mAbs (Seikagaku).
The A201 T-cell line was a CEM-derived T-cell line, transfected with wild-type CD4 cDNAs (A201-CD4), or with a mutated form of CD4 cDNAs (A201-2C>A) in which the two cysteine residues required for association with p56lck were replaced by two alanine residues . These T-cell lines normally express CD4 and equivalent amounts of active p56lck at the cell surface . No association between the mutated form of CD4 and p56lck was found in A201-2C>A. Cells were cultured in RPMI 1640 medium (Life Technologies) supplemented with 10% (v/v) fetal calf serum, 2 mM L-glutamine and 0.5 mg/ml of G418 (Life Technologies).
Isolation of DRMs by flotation experiments
Optiprep gradient analysis was performed according to a previously described method with some modifications . Briefly, as previously reported , A201-CD4 T-cell lines were starved overnight in serum-free medium; 25×106 cells were then incubated with anti-CD4 Ab for 20 min at 4 °C, washed and incubated for the indicated time at 37 °C with a GAMIg. Cells were then lysed on ice for 15 min in 900 μl of lysate buffer A (50 mM Tris/HCl, pH 7.4, 110 mM NaCl, 10 mM EGTA, 1 μg/ml leupeptin, 1 μg/ml pepstatin, 2 μg/ml aprotinin and 1 mM PMSF) without detergent and sonicated gently (5 s bursts, 5 W; Branson sonifier 250). After centrifugation at 800 g for 5 min at 4 °C, the PNS (post-nuclear supernatant) was incubated with 1% Brij 58 at 4 °C for 1 h. DRMs were isolated by ultracentrifugation at 28000 rev./min for 4 h at 4 °C in a SW41 rotor (Beckman Instruments), in a 40%/30%/5% Optiprep density gradient (Sigma). Seven fractions were collected from the top of the tube. F1 corresponds to the top of the gradient. The low-density fractions 2, 3 and 4 contained detergent-insoluble raft fractions enriched in gangliosides. As expected, transferrin receptor that does not reside in rafts was only detected in fractions 6 and 7 that contained detergent-soluble membrane. Normalized protein amounts for each fraction were determined using the Bio-Rad kit with BSA as the standard. Each fraction was then immunoprecipitated with specific Abs and analysed by SDS/8% PAGE and Western blotting as previously described . Proteins were visualized using an ECL® (enhanced chemiluminescence) detection system (Amersham Biosciences) with an anti-rabbit or anti-mouse Ig coupled with HRP (horseradish peroxidase) as the secondary Ab (Amersham Biosciences). GM1 and GM3 gangliosides were detected using HPTLC (high-performance TLC) , as reported below. The ganglioside extract was split into two aliquots. The first one was run on silica gel 60 HPTLC plates (Merck) and stained with resorcinol to detect GM3. The second one was run on HPTLC aluminium-backed silica gel 60 (20×20) plates (Merck). The plates were immunostained with 0.5 μg/ml of CTxB (cholera toxin)–HRP for 1 h to detect GM1 at room temperature (25 °C). Immunoreactivity was assessed by chemiluminescence.
Ganglioside detection by HPTLC
A201-CD4 or A201-2C>A T-cell lines were incubated overnight at 37 °C in serum-free medium before incubation with anti-CD4 Abs. Cells (40×106) were incubated for 20 min at 4 °C with anti-CD4 Ab and then washed and incubated for the indicated time at 37 °C with a GAMIg. Cells were then lysed on ice for 20 min in 900 μl of lysate buffer A supplemented with 1% Nonidet P40, and clarified by centrifugation at 12000 g for 15 min. The same amount of each PNS (4 mg) was immunoprecipitated with specific Abs and analysed for the presence of the gangliosides by HPTLC. Gangliosides were extracted twice in chloroform/methanol/water (4:8:3, by vol.) and subjected to Folch partition by the addition of water, resulting in a final chloroform/methanol/water ratio of 1:2:1.4. The upper phase, containing polar glycosphingolipids, was purified of salts and low-molecular-mass contaminants using Bond Elut C18 columns (Superchrom). The eluted glycosphingolipids were dried and separated by HPTLC using silica gel 60 HPTLC plates (Merck). Chromatography was performed in chloroform/methanol/aq. 0.25% KCl (5:4:1, by vol.). The plates were stained with resorcinol (ganglioside-specific stain) or with cholera toxin for GM1, and anti-GM3 Ab . Quantification was carried out by densitometric scanning analysis using a Mac OS 9.1 (Apple Computer International) and NIH (National Institutes of Health) Image 1.61 software. The amount of immunoprecipitated protein was checked by control Western blotting with the Abs used for specific immunoprecipitation.
Co-localization experiments using scanning confocal microscopy
As described previously , after activation with cross-linked anti-CD4 Ab and saturation of the free sites with GAMIg incubation at 4 °C, cell suspensions containing 8×104 cells/slip were layered on to poly(L-lysine)-coated coverslips for 45 min at room temperature. Immunofluorescence staining of cell surface molecules was performed with the appropriate mAbs in the absence of permeabilizing agent, followed by FITC- or TRITC-anti-mouse IgG1 or anti-rabbit secondary Abs. Intracellular proteins were stained for 1 min in 0.05% saponin-permeabilized cells using the appropriate mAbs. We have previously checked that labelled GAMIg does not reveal anti-CD4 Ab . Confocal microscopy was performed on a Zeiss LSM-510 confocal microscope. Images were acquired using the maximum signal detection setting below the saturation limit of the detector.
The CD4–p56lck association is required for the aggregation of GM1(+) and GM3(+) DRMs induced by anti-CD4 Ab
We have previously shown that CD4 ligand binding specifically induces CD4 aggregation and partial co-localization of GM1- and GM3-enriched domains . Integrity of raft domains and the CD4–p56lck association were required for the down-regulation of LFA-1-dependent adhesion induced by CD4 ligand binding . We have therefore investigated whether the CD4 ligand induces ganglioside-enriched domain aggregation in a CD4–p56lck-dependent manner. We observed that GM1- and GM3-enriched domain aggregation, detected following CD4 ligation (Figure 1A), was dependent on CD4–p56lck association, since it was not observed in A201-2C>A, an A201-T-cell line transfected with a mutated form of CD4 that does not bind p56lck (Figure 1B). Patching was not observed with a control Ab specific to transferrin receptor (Figures 1A and 1B, Ct). HPTLC showed that both T-cell lines expressed the same amount of ganglioside under the tested conditions (Figure 1C).
Role of CD4–p56lck association in GM1 and GM3 distribution following CD4 ligation
CD4 ligation induces the redistribution of LFA-1 and PI3K with GM1 and GM3 in a CD4–p56lck association-dependent manner
We have previously reported that PI3K and LFA-1 partially co-localized with GM1 and GM3 in DRMs following CD4 triggering . In the present study, we investigated whether CD4–p56lck association was also required for these co-localizations. Distribution of LFA-1 and PI3K, which is uneven and punctuated over the plasma membrane following CD4 cross-linking, appeared distributed all over the membrane without detectable clustering in the A201-2C>A T-cell line; this was also observed in both T-cell lines in the absence of CD4 triggering (NA) and in the control Ab-treated cells (Ct; Figures 2A and 3A). These immunofluorescence experiments also showed that the co-localizations of GM1, LFA-1 and PI3K respectively and of GM3 and PI3K were partial but stable over a 20 min period. Co-localization of LFA-1 with GM3 was more dominant after a short period of incubation with anti-CD4 Ab (2 min) than following longer incubation (Figure 2A, left panel).
Role of CD4–p56lck association in the co-localization and association of LFA-1 with GM1 and GM3 gangliosides following CD4 ligation
Role of CD4–p56lck association in the co-localization and association of PI3K with GM1 and GM3 gangliosides following CD4 ligation
To determine whether LFA-1 and PI3K interact with GM1 and GM3 gangliosides following CD4 cross-linking, cell-free lysates from anti-CD4-treated cells were immunoprecipitated with an anti-LFA-1 mAb (Figure 2B, and Supplementary Figure S1A at http://www.BiochemJ.org/bj/402/bj4020471add.htm) or an anti-PI3K Ab (Figure 3B and Supplementary Figure S1B), and gangliosides were detected by HPTLC as previously reported . Immunoprecipitation with an anti-LFA-1 mAb (Figure 2B and Supplementary Figure S1A) and gangliosides detected by HPTLC showed a faint GM3 band but no GM1 band following control incubation with GAMIg (Figure 2B). In contrast, after 2 min of CD4 cross-linking, two main resorcinol-positive bands co-migrating with GM3 [82±10 arbitrary units (AU); Supplementary Figure S1A] and, to a lesser extent, with GM1 (57±5 AU; Supplementary Figure S1A) were detected. Co-precipitation of GM3 with LFA-1 rapidly decreased (30±5 and 24±2 AU after 10 and 20 min respectively; Supplementary Figure S1A). In contrast, co-precipitation of GM1 with LFA-1 was lower but stable and decreased more slowly (Supplementary Figure S1A). Confocal microscopy analysis also showed that the LFA-1–GM1 interaction was more stable than the LFA-1–GM3 interaction. Co-localization in patches (Figure 2A) and co-precipitation of LFA-1 and gangliosides (Figure 2B) both required CD4–p56lck association since neither was detected in the A201-2C>A T-cell line. The same amount of LFA-1 was immunoprecipitated under the different conditions used in A201-CD4 and A201-2C>A T-cell lines as shown with the anti-LFA-1 immunoblot (Figure 2B).
In the PI3K immunoprecipitates (Figure 3B and Supplementary Figure S1B), resorcinol-positive bands were not detected following control activation (Ct), whereas GM1 and, to a lesser extent, GM3 were detectable following activation with the anti-CD4 Ab. Although co-precipitation was stable for 20 min, GM3 co-precipitation with PI3K was always lower than with GM1 (Figure 3B and Supplementary Figure S1B). No ganglioside bands were detected in the extracts from the PI3K immunoprecipitates at the surface of the A201-2C>A T-cell line (Figure 3B), indicating that PI3K interaction with lipid rafts also requires CD4–p56lck association. The same amount of PI3K was immunoprecipitated under the different conditions used in A201-CD4 and A201-2C>A T-cell lines as shown with the anti-PI3K immunoblot (Figure 3B).
Anti-CD4 Ab induces redistribution of p56lck and SHP-2 with GM1 and GM3 in a CD4–p56lck association-dependent manner
The above results indicated that CD4-associated-p56lck is required for the aggregation of GM1 and GM3, and the co-localization of PI3K and LFA-1 after CD4 signalling, and that p56lck may also be redistributed with these gangliosides. Scanning confocal microscopy revealed p56lck clustered in patches co-stained with anti-GM1 and anti-GM3 Abs (Figure 4A). All of the GM3-containing domains but only some of the GM1-containing domains co-localized with p56lck (Figure 4A). Aggregation of p56lck induced by CD4 ligand binding was not detected in the A201-2C>A T-cell line, indicating that aggregation was dependent on CD4–p56lck association (Figure 4A). Patching was not detected in either of the T-cell lines following cross-linking with a control Ab (Figure 4A, Ct-2). However, CD4 cross-linking led to a transient increase in co-precipitation of GM3 with p56lck, which was also dependent on CD4–p56lck association since it was not detected in the A201-2C>A T-cell line (Figure 4B). TLC immunostaining analysis (a more sensitive method than resorcinol staining) using CTxB that specifically binds GM1 and anti-GM3Ab for GM3 detection (Figure 4C) revealed co-precipitation between p56lck and GM1. Ganglioside association was not detected following incubation with a control Ab (Figure 4B, Ct). The same amount of p56lck was immunoprecipitated in the HPTLC (Figure 4B) and TLC (Figure 4C) analyses in both T-cell lines.
Co-localization and association of p56lck with GM1 and GM3 induced by CD4 ligation are dependent on CD4–p56lck association
SHP-2 tyrosine phosphatase was previously reported to be required for the formation of a multi-protein complex with PI3K induced by CD4 ligand binding . In the present study, CD4 triggering induced very rapid redistribution of SHP-2 with both GM1 and GM3 (Figure 5A). After CD4 cross-linking for 20 min, SHP-2 was distributed predominantly in the cytoplasm, and was less aggregated in patches on the plasma membrane, indicating that SHP-2 only transiently localized in these ganglioside-enriched domains. In contrast, GM1 and GM3 aggregation was maintained, indicating that the delocalization of SHP-2 from these domains was not due to disruption of the ganglioside-enriched domains. The rapid and transient localization of SHP-2 in GM1- and GM3-enriched domains was also dependent on CD4–p56lck association, since SHP-2 aggregation was not detected in the A201-2C>A T-cell line (Figure 5A). Patching was not detected following incubation with a control Ab (Figure 5A, Ct-2). However, co-precipitation between SHP-2 and GM1 and GM3 was not detected, suggesting that either SHP-2 does not interact directly with these gangliosides or, alternatively, the affinity of these interactions was too low to be detected under immunoprecipitation conditions (Figure 5B).
Co-localization without association of GM1 and GM3 with SHP-2 in a CD4–p56lck association-dependent manner
Anti-CD4 Ab induces distinct interactions between p56lck, SHP-2 and PI3K in DRMs
We then investigated whether p56lck co-localized and interacted with SHP-2 and PI3K in DRMs. Immunofluorescence analysis showed that p56lck did not co-localize with SHP-2 and PI3K in the same manner. Indeed, CD4 cross-linking induced a transient co-localization of p56lck with SHP-2 (Figure 6A), whereas the co-localization of p56lck with PI3K appeared stable for up to 20 min (Figure 7A). These observations were consistent with those described above, showing the transient localization of SHP-2 in raft domains (Figure 5A), in contrast with PI3K (Figure 3A) and p56lck (Figure 4A). Partial co-localization between SHP-2 and PI3K was observed (Figure 6B). Patching was not observed following incubation with a control Ab (Figures 6A, 6B and 7A).
Distribution and association of SHP-2 with p56lck and PI3K in GM1- and GM3-containing microdomains following anti-CD4 Ab incubation
Distribution and association of p56lck with PI3K in GM1- and GM3-containing microdomains following anti-CD4 Ab incubation
The GM1- and GM3-enriched DRMs containing the isolates from activated and unstimulated T-cells obtained by Optiprep density gradient centrifugation and ultracentrifugation were mainly detected in fractions F2–F4 (Figure 6C, top panel). F1 corresponds to the top of the gradient, and the detergent-soluble membrane was detected in fractions F6 and F7. Neither GM1 nor GM3 were detected in F1, F6 and F7 (Figure 6C). p56lck (Figure 6C, bottom panel, and Supplementary Figure S2A at http://www.BiochemJ.org/bj/402/bj4020471add.htm) was detected in the immunoprecipitates from DRM fractions F2–F4 and one of the detergent-soluble fractions (F6) in the absence of, or following, CD4 cross-linking. In contrast, SHP-2 was mainly detected in detergent-soluble fraction (F6) and DRM fraction 4 in the absence of CD4 cross-linking (Figure 6D and Supplementary Figure S2B). Following CD4 cross-linking for 2 min, SHP-2 was predominantly detected in F3 [RN (relative number)=4, Figure 6C and Supplementary Figure S2B] and F4 DRM-containing fractions (RN=2.5, Figure 6C; RN=5, Supplementary Figure S2B). Following longer CD4 ligation, SHP-2 was detected in all of the tested fractions (Supplementary Figure S2B). However, SHP-2 did not co-precipitate with p56lck in any of the DRM fractions with the exception of the detergent-soluble fraction F6 (Figure 6C, and Supplementary Figure S3A at http://www.BiochemJ.org/bj/402/bj4020471add.htm). Similarly, SHP-2 only co-precipitated with PI3K (Figure 6D and Supplementary Figure S3B) in detergent-soluble membrane fraction F6, as previously reported under other conditions . Co-precipitation between SHP-2 and PI3K was not detected in any of the DRM fractions (Figure 6D and Supplementary Figure S3B). In contrast, PI3K was detected in all of the fractions following CD4 ligation (Supplementary Figures S2C and S3B) as expected, and was also co-precipitated with p56lck in all of the fractions (Figure 7B and Supplementary Figure S3C).
Overall, these results indicate that CD4 ligand binding induces co-localization and aggregation of several proteins involved in CD4 signalling in detergent-insoluble membrane domains. Lateral compartmentalization of the plasma membrane into raft domains is a key feature of immune cell activation and subsequent functions . In previous studies, we have shown that CD4 ligation induced both aggregation of GM1 and GM3 gangliosides and co-localization of LFA-1 and PI3K required for the down-regulation of LFA-1-mediated T-cell–B-cell adhesion . In the present study, we show that CD4 signalling also induces a strong protein–ganglioside association within lipid rafts and we provide evidence that the CD4-associated tyrosine kinase p56lck required for the down-regulation of LFA-1-dependent adhesion induced by CD4 ligation  is also localized in ganglioside-enriched domains. In addition, given the absence of aggregation and protein recruitment when a mutant T-cell line expressing CD4 unable to bind p56lck was tested, we propose that p56lck is required for the aggregation of GM1 and GM3, as well as for the recruitment and association of LFA-1 and PI3K with these gangliosides. These results appear to contrast with those reported by Fragoso et al.  showing that association between CD4 and p56lck is not essential for CD4-induced lipid raft aggregation. The most likely explanation for this difference could be the use of an anti-CD4 Ab (OKT4) that binds to the region of CD4 (D4) proximal to the transmembrane domain by Fragoso's team, whereas we used the 13B8.2, which recognizes an epitope located within the first N-terminal domain (D1). The signal induced by these two Abs differs because only anti-CD4 Abs binding to the D1 CD4 domain (such as Leu3a, OKT4a and 13B8.2) inhibit Ag-independent LFA-1-dependent T-cell adhesion on B-cells . In our study, inhibition was not detected with other anti-CD4 Abs, such as OKT4 (F. Mazerolles, unpublished work). In addition, the pattern of tyrosine-phosphorylated proteins following binding of either OKT4- or D1 domain-specific ligands is distinct [29,30]. Fragoso et al.  observed enhanced tyrosine phosphorylation of several proteins induced by OKT4 binding in the absence of association between CD4 and p56lck in these mutant lines, showing that this setting differs from ours. Moreover, as OKT4a and Leu3a epitopes are required for GM1 signalling in contrast with OKT4 , it appears that cross-linking of distinct CD4 epitopes induces different signalling and aggregation of raft domains in a CD4–p56lck association-dependent or -independent manner. Since 13B8.2 binds to the HLA class II interactory domain of CD4, it may be a more appropriate physiological model for inducing CD4 aggregation than OKT4. Furthermore, our results are consistent with reports showing that in the absence of p56lck, lipid rafts do not form caps following CD4 engagement, so CD4 ligation may provide a sufficient signal to induce aggregation of lipid rafts in p56lck (+) Jurkat cells in the absence of TCR engagement .
We also show that p56lck totally co-localized with GM3-containing domains in a CD4–p56lck association-dependent manner, while interaction between p56lck and GM1 was less intense. Since CD4 associates with GM3 , CD4 appears to be the main intermediary between p56lck and GM3. On the contrary, as CD4 does not associate with GM1 in a similar fashion, additional partners are likely to be inserted between GM1 and the CD4–p56lck complex (as proposed in the model depicted in Figure 8). Interestingly, the association between p56lck and GM1 and GM3 gangliosides did not persist after prolonged CD4 cross-linking (20 min), whereas p56lck remained localized in the DRM domains, indicating that p56lck can persist in rafts without association with gangliosides. The localization of p56lck in lipid rafts in the absence of TCR activation is controversial. In unstimulated cells, most of the p56lck co-localizes with CD4 outside of lipid rafts and translocates into raft domains upon antigen stimulation . Raft integrity has been shown to be important for p56lck activity . On the other hand, others have reported that p56lck localizes mainly in lipid rafts [32,34]. Our results support the view that p56lck can localize within rafts in the absence of antigen presentation, but is further recruited and clustered in these domains following CD4 engagement. They also confirm that CD4 ligation plays a critical role in the recruitment of signalling proteins to rafts and may facilitate cellular reorganization by recruiting adhesion and signalling molecules .
A model for the distinct membrane associations between LFA-1, PI3K, p56lck and SHP-2 induced by CD4 ligation, in a CD4–p56lck-dependent manner
We also provide evidence that CD4–p56lck association is required for the recruitment of LFA-1, PI3K and also SHP-2, which is another partner involved in the down-regulation of LFA-1-dependent adhesion induced by CD4 ligand binding . This result supports the hypothesis that CD4-associated p56lck could induce the recruitment of adhesion and signalling molecules in DRMs that are likely to concentrate these proteins in signalling platforms . Our finding that LFA-1 and PI3K were not always co-precipitated with the same gangliosides during CD4–p56lck-induced signalling suggests that these proteins move to different domains containing GM1 and/or GM3 gangliosides which are not always co-localized. However, spatial resolution would be necessary to differentiate GM1- or GM3-containing domains. It is possible that prolonged CD4 ligation induced a modification of affinities between proteins and gangliosides and cannot be excluded. Indeed, PI3K did not appear to move from the DRMs during CD4 ligation. It is likely that PI3K indirectly associates with GM3 via the p56lck–CD4 complex because CD4 cross-linking also induced association between PI3K and p56lck in the GM3-enriched fractions. The fact that PI3K was less associated with GM3 than with GM1 suggests that the PI3K isoforms (α and β, indistinguishable from the Ab we used) have a distinct affinity for these gangliosides. On the other hand, LFA-1 partially and stably co-localized with GM1, without notable modification after longer CD4 ligation, while the co-precipitation of LFA-1 with GM1 only slowly decreased. In addition, LFA-1 totally co-localized with GM3 after a short period of CD4 ligation, followed by a decrease in co-localization. This decrease was correlated with a rapid decrease in LFA-1/GM3 co-precipitation, suggesting a possible modification of affinity between these molecules. This modification can also induce a modification of LFA-1 localization in distinct ganglioside-enriched-domains. A similar raft segregation of membrane-receptor redistribution has also been described in migrating cells [35,36]. Integrins co-localized with GM1-enriched rafts redistributed to the uropod, whereas PI3K co-localized with GM3-enriched rafts redistributed to the leading edge. As the GM3–CD4 association is required for CD4 signalling [32,37] leading to the down-regulation of the LFA-1-dependent adhesion induced by CD4 ligation , GM3 could be a transducer of transmembrane signalling and could play a role in cell adhesion regulation by triggering partners in a multi-protein complex formation . The molecular basis of these ganglioside–protein interactions remains to be determined. Indeed, isoforms of PI3K and LFA-1 are not acylated by saturated fatty acids such as p56lck or CD4 to partition into rafts. Thus we suggest that, in addition to p56lck, other unidentified proteins moving in lipid rafts (X and Y in the model) are also involved in these interactions . LFA-1 may bind cytohesin-1, which binds D3 phospholipids that are enriched in rafts, and other partners participating in the down-regulation of LFA-1 adhesion . This localization in rafts would allow LFA-1 to have a stronger affinity for its ligand and enhance its stability, and increase the activity of PI3K. SHP-2, which associates with PI3K outside of rafts following CD4 binding , may be another intermediate regulator of Src kinase activation due to its localization in raft domains . However, although CD4 ligation induces rapid and transient SHP-2 localization in a CD4–p56lck association-dependent manner, direct interaction between SHP-2 and gangliosides or with p56lck and PI3K was not detected. Nevertheless, SHP-2 was co-precipitated with PI3K and p56lck after CD4 ligation in detergent-soluble membrane. The absence of co-precipitation in rafts suggests a possible alternative role for SHP-2 and other partners. We recently observed that outside of raft domains, SHP-2 can associate with the serine kinase PDK1 (phosphoinositide-dependent kinase-1), a specific effector of PI3K that also binds D3 phospholipids (F. Mazerolles and M. Trucy, unpublished work). Thus it is possible that PDK1 also migrates to rafts with SHP-2 and could be the intermediary between PI3K and SHP-2. This hypothesis is under investigation.
In conclusion, these results further support our previous findings of the key role of CD4–p56lck association in CD4 signalling. Indeed, this association, which is important for p56lck and PI3K activities induced by CD4 ligation [25,29], appears to be required for the redistribution of these kinases in ganglioside-enriched domains and co-localized with LFA-1 and SHP-2 that is involved in the regulation of adhesion. Our results also suggest that gangliosides associate with distinct proteins that probably have different affinity binding, and can be modified following CD4 signalling. Through these associations, gangliosides could transiently sequestrate these proteins and consequently inhibit LFA-1-dependent adhesion.
We thank J. Lipecka for invaluable help with image acquisition with the confocal microscope, and R. Fagard for the helpful discussions. M.T. was supported by a doctoral fellowship from La Ligue Nationale Contre le Cancer. This work was supported by Inserm.
goat anti-mouse Ig
human leucocyte antigen
lymphocyte function-associated antigen-1
Src homology 2 domain-containing protein tyrosine phosphatase 2
These authors contributed equally to this work.