In T lymphocytes, lipid rafts are preferred sites for signal transduction initiation and amplification. Many cell membrane receptors, such as the TCR, coreceptors, and accessory molecules associate within these microdomains upon cell activation. However, it is still unclear in most cases whether these receptors interact with rafts through lipid-based amino acid modifications or whether raft insertion is driven by protein-protein interactions. In murine T cells, a significant fraction of CD2 associates with membrane lipid rafts. We have addressed the mechanisms that control the localization of rat CD2 at the plasma membrane, and its redistribution within lipid rafts induced upon activation. Following incubation of rat CD2-expressing cells with radioactive-labeled palmitic acid, or using CD2 mutants with Cys226 and Cys228 replaced by alanine residues, we found no evidence that rat CD2 was subjected to lipid modifications that could favor the translocation to lipid rafts, discarding palmitoylation as the principal mechanism for raft addressing. In contrast, using Jurkat cells expressing different CD2 and Lck mutants, we show that the association of CD2 with the rafts fully correlates with CD2 capacity to bind to Lck. As CD2 physically interacts with both Lck and Fyn, preferentially inside lipid rafts, and reflecting the increase of CD2 in lipid rafts following activation, CD2 can mediate the interaction between the two kinases and the consequent boost in kinase activity in lipid rafts.
Cell surface molecule CD2 is a 45- to 58-kDa type I integral protein expressed on virtually all T lineage and NK cells (1). The extracellular domain of CD2 is composed by two Ig superfamily domains (2), of which the membrane-distal V-like domain is involved in the binding to the ligand (3). CD2 has different ligands in rodents (CD48) and humans (CD58), and the mechanisms of ligand binding are substantially different, as assessed by thermodynamic analysis (4, 5). Nevertheless, engagement of CD2 to CD58 in humans, or to CD48 in rodents, facilitates adhesion between T cells and APC and is proposed to promote the formation of an optimal intercellular membrane spacing (∼140 Å) suitable for TCR recognition of a peptide Ag bound to MHC (6).
Concomitantly with its function as an adhesion molecule, CD2 has a role in the process of signal transduction. It is known that stimulation of CD2 with combinations of mAb or with the physiological ligand induces T cell proliferation in rats and human (7, 8). Moreover, ligation of CD2 with CD58 augments the IL-12 responsiveness of activated T cells (9), and can also reverse T cell anergy induced by B7 blockage (10). In contrast, CD2 interacts with the inhibitory receptor CD5 (11, 12), and through this association is able to amplify modulatory signals at the T cell surface (11, 13).
The cytoplasmic domain is required for CD2-mediated activation (14, 15, 16, 17), but as it has no intrinsic enzymatic activity, it depends on physical interactions with other signaling molecules to propagate the stimulus received upon ligand binding. CD2 associates with the Src family tyrosine kinases Lck and Fyn and through these kinases couples to downstream signal transduction pathways (18, 19, 20). Proline-rich sequences of the cytoplasmic domain of CD2 have been shown to be involved in the interaction with the Src homology 3 (SH3)4domains of these kinases (20, 21, 22), also with the SH3 domain of the adaptor protein CD2AP (23), connecting CD2 to the cytoskeleton, and additionally with the SH3 domain of CD2BP1 (24), resulting in the down-regulation of activation-dependent CD2 adhesion. CD2 binds to the GYF domain of CD2BP2 as well, an interaction resulting in cytokine production (25).
The T cell plasma membrane contains regions of distinct lipid composition, enriched mainly in sphingolipids and cholesterol immersed in a phospholipid-rich environment, and termed lipid rafts (26). In resting cells, key signaling proteins such as the adaptors linker for activation of T cells (LAT) (27) or phosphoprotein associated with glycosphingolipid-enriched microdomain (PAG) (28), or the kinase Fyn (29), are resident in rafts, whereas most integral proteins are found outside these platforms. Localization and targeting of signaling molecules to lipid rafts is largely dependent on posttranslational acylation modification of proteins, one of the most important being membrane-proximal cysteine palmitoylation. These cysteine residues are usually present within a conserved motif consisting of cysteines and hydrophobic residues, CVRC in LAT and GCVC in Lck and Fyn. The composition of raft-associated proteins may, nevertheless, change upon cell stimulation. Membrane compartmentalization and partitioning of essential T cell-activating components in lipid rafts were shown to be involved in the initial stages of T cell activation (30). Several costimulatory molecules on the T cell surface, such as CD2 (31), CD5, CD9 (32), and CD28 (33) can up-regulate TCR signals by enhancing the association of the TCR with lipid rafts.
Human CD2 has been shown to translocate to lipid rafts upon CD2 Ab cross-linking or following binding to CD58 during conjugate formation (34), and the shift between phases may result in the replacement of CD2BP2 by raft-resident Fyn (22), which competes for the same proline-rich sequence of the cytoplasmic tail of CD2. In contrast to human CD2, which is reported to be entirely non-raft-resident in resting cells (34), mouse CD2 contains a CIC motif at the membrane-proximal region that can putatively address CD2 to lipid rafts (32). We have investigated the membrane localization of rat CD2 and its transition between different phases at the T cell surface, and show that a significant fraction of CD2 is constitutively present in lipid rafts, and that this proportion increases upon CD2 stimulation. However, membrane-proximal cysteine palmitoylation is not the decisive tag addressing rat CD2 to lipid rafts, but rather the interaction with the Src family kinase Lck.
Materials and Methods
Monoclonal Ab recognizing rat CD2 were OX-54, OX-55 (8), and OX-34 (35). None of these mAb cross-reacted with human CD2 as confirmed by FACS (data not shown). The anti-phosphotyrosine mAb 4G10 HRP-conjugated was from Upstate Biotechnology. Polyclonal Abs and conjugates were: rabbit anti-LAT, from Upstate Biotechnology; rabbit anti-Lck, raised against a peptide of aa 39–64 of murine Lck, a gift from J. Borst (The Netherlands Cancer Institute, Amsterdam, The Netherlands); BL90, polyclonal anti-Fyn, a gift from J. Bolen and M. Tomlinson (DNAX Research Institute, Palo Alto, CA); goat anti-mouse peroxidase conjugate, purchased from Molecular Probes; and goat anti-rabbit peroxidase conjugate, from Zymed Laboratories.
Cells, cDNA cloning, plasmids, and transfections
Splenocytes were obtained by maceration of spleens from 3-mo old Wistar male rats in ice-cold PBS. The rat thymoma cell line W/FU(C58NT)D (36), referred to in this study as C58 cells, was provided by A. Conzelmann (University of Fribourg, Fribourg, Switzerland).
Jurkat E6.1 cells (37) were obtained from A. Weiss (University of California, San Francisco, CA). E6.1 cells expressing full-length rat CD2, E6.1-CD2(CY whole) (referred to in this text as E6.1-CD2) or expressing CD2 cytoplasmic deletion mutants, E6.1-CD2(CY 6), (CY 40), (CY 66), (CY 81), and (CY 97), which possess respectively the first 6, 40, 66, 81, and 97 amino acids of the cytoplasmic tail (17, 38), were provided by M. Puklavec (University of Oxford, Oxford, U.K.).
Jurkat E6.1 cells expressing a CD2 mutant with cytoplasmic cysteine residues substituted by alanines, or CD2 with a deleted sequence between aa 7 and 40 of the cytoplasmic tail (rat CD2(Δ7–40)), were produced as follows: pKG5-CD2–2C, a plasmid for which the codons for Cys226 and Cys228 are mutated to encode alanine residues, was constructed by PCR using pRCD2–11 (17), a plasmid containing the cDNA sequence of full-length rat CD2, as template, with the forward primer 5′-GGGGCGCTG TTTATTTTCGCTATCGCCAAGAGGAAAAAACGGAAC-3′ and the reverse primer 5′-GTTCCGTTTTTTCCTCTTGGCGATAGCGAAAATAAACAGCGCCCC-3′. The changed codons are underlined. The PCR product was digested with DpnI and used to transform competent cells. Positive mutants were digested with BamHI and replaced the equivalent restriction fragment containing wild-type CD2 in pRCD2–11. To prepare the vector encoding rat CD2(Δ7–40), we performed a PCR using pRCD2–11 as template, with the forward primer 5′-CTGCAAGAGGAAAAAACGGAAC/CCAGTGGCTTCCCAAGCT-3′ and a reverse primer with the complementary sequence. The primers encode a CD2 sequence in which the codon of aa 6 of the cytoplasmic tail is followed (separated by a slash) by the codon of aa 41, thus excluding aa 7–40. The amino acid sequence at the junction is KRKKRN/NPVASQ. The PCR product was digested with DpnI to digest the parental cDNA without the deletion, and used to transform competent cells.
Plasmids pKG5-CD2–2C and pKG5-CD2(Δ7–40) were used for transfecting E6.1 cells to handle E6.1-CD2(CIC/AIA) and E6.1- CD2(Δ7–40), respectively. Briefly, cells at 1.0 × 107 cells/ml in ice-cold PBS were transferred into 0.4-cm electroporation cuvettes, mixed at 4°C with plasmid at a final concentration of 0.1 mg/ml, and pulsed with 0.62 kV and 25 μF using a Gene Pulser II electroporator from Bio-Rad. Cells were transferred to culture flasks containing RPMI 1640 with 10% FCS, 1 mM sodium pyruvate, 2 mM l
The Lck-deficient J.CaM1 cell line (39) was transfected with pRCD2–11 or pKG5-CD2/PAG/CD2 (see below), and G418-resistant cells that expressed rat CD2 were selected and named J.CaM1-CD2 and J.CaM1-CD2/PAG, respectively. To generate J.CaM1/Lck(KD)-CD2 cells, the newly obtained J.CaM1-CD2 cells were cotransfected with pSRα-Lck-KD, a plasmid containing Lck cDNA carrying a mutation at the ATP binding site (K273A) (40). J.CaM1 cells reconstituted with wild-type Lck, J.CaM1/Lck(WT), and with a Lck molecule having Cys3 mutated to Ala3, J.CaM1/Lck(C3A) (41) were obtained from P. Kabouridis (Queen Mary, University of London, London U.K.), and transfected with pRCD2–11, to handle J.CaM1/Lck(WT)-CD2 and J.CaM1/Lck(C3A)-CD2 cells, respectively.
A chimeric pKG5-CD2/PAG/CD2 was produced (GenScript), in which the sequences coding for the transmembrane region and six amino acids of the cytoplasmic domain of rat CD2 were substituted by the corresponding sequences encoding the transmembrane region and nine amino acids of the cytoplasmic tail of PAG, including the palmitoylation sequence. The sites of junctions of the chimeric protein are shown by slashes: PEKGLP/LWGSLA…. CDREKK/RRKGEE. The chimeric cDNA obtained was subcloned into the BamHI restriction site of the expression vector pKG5. For simplicity, the mutant molecule is referred to in this study as CD2/PAG.
Capping and immunofluorescence microscopy
All procedures were performed at 4°C unless otherwise described. Cells were washed in RPMI 1640, resuspended at 2 × 106 cells/ml in RPMI 1640/5% FCS, and incubated for 10 min with the mAb OX-34-FITC conjugate (Caltag Laboratories) at 5 μg/ml. Cells were washed twice with PBS containing 0.2% BSA, resuspended in RPMI 1640/5% FCS, and incubated for 15 min at 37°C to allow for Ab-induced capping. Cells were then fixed with 4% paraformaldehyde in PBS for 30 min at room temperature, and then incubated with saponin (0.4% w/v) for 20 min. Anti-LAT mAb (20 μg/ml) was added for 10 min to the preparations, followed by a PBS wash and incubation for 10 min with polyclonal goat anti-rabbit Alexa Fluor 568 conjugate (Molecular Probes) at 10 μg/ml, and again washed. In parallel, cells without CD2 cross-linking were fixed, permeabilized, and posteriorly labeled with OX-34-FITC and anti-LAT and goat anti-rabbit Alexa Fluor 568.
Cells were plated onto glass coverslips and mounted in Vectashield medium (Vector Laboratories). Stained preparations were observed with an AxioImager Z1 microscope (Carl Zeiss), and images acquired with Axiocam MR v.3.0 camera (Carl Zeiss). Images were processed with Photoshop 6.0 (Adobe Systems). The percentage of colocalization represents the counts of CD2 caps that also colocalize with LAT, quantified with blind scoring, counting a minimum of 200 caps in each of two experiments. Each experiment was observed by two independent examiners.
Cells were washed and resuspended in PBS containing 0.2% BSA and 0.1% NaN3 (PBS/BSA/NaN3), at a concentration of 1 × 106 cells/ml. Staining was performed by incubation of 5 × 105 cells/well with mAbs (20 μg/ml) for 15 min on ice, in 96-well round-bottom plates (Greiner). Cytometric analysis was as previously described (11).
Sucrose gradient centrifugation
Approximately 1.5 × 108 cells were used per sample. Cells were washed, resuspended in 1 ml of RPMI 1640 medium, and maintained at 37°C for 15 min. Cells were then washed twice with ice-cold PBS and lysed for 30 min on ice in 1 ml of MBS buffer (25 mM MES, 150 mM NaCl (pH 6.5), containing 1% Triton X-100, 1 mM PMSF) and cocktail protease inhibitors (1 mM AEBSF, 0.8 μM aprotinin, 50 μM bestatin, 15 μM E-64, 20 μM leupeptin and 10 μM pepstatin A; Calbiochem). The lysates were homogenized by brief sonication for 10 pulses on ice, using a Heat Systems/Ultrasonics sonicator (model W-375) equipped with a microtip and set to 50% duty cycle, output 3. To obtain the rafts fraction, cell lysates were mixed with an equal volume of 85% sucrose in MBS buffer and transferred to the bottom of Sorvall ultracentrifuge tubes. The samples were then overlaid with 6 ml of 35% sucrose followed by 2 ml of 5% sucrose. After centrifugation at 200,000 × g for 17 h at 4°C in a TST41.14 swing-out rotor (Sorvall), 10 fractions of 1 ml were collected from the bottom of the tube and analyzed by Western blotting. The fractions are labeled from the top of the gradient.
Metabolic labeling of cellular proteins with [3H]palmitate
Aliquots of 1.5 × 107 cells were collected from culture, resuspended in 1.5 ml of RPMI 1640 complete medium supplemented with 300 μCi/ml [3H]palmitic acid, and incubated for 16 h at 37°C. Cells were harvested and lysed in lysis buffer (10 mM Tris Cl (pH 7.4), 150 mM NaCl, 1 mM EDTA, 1 mM PMSF, and 1% (v/v) Triton X-100). The nuclear pellet was removed by centrifugation at 12,000 × g for 10 min at 4°C, and the lysate run on SDS-PAGE under nonreducing conditions. Immunoprecipitations were performed by incubating the lysate with Abs (5 μg of IgG or 2 μl of antiserum) and 100 μl of 10% protein A-Sepharose CL-4B beads (Amersham Biosciences), for 90 min at 4°C by end-over-end rotation. The beads containing the immune complexes were washed three times in 1 ml of lysis buffer, denatured, and loaded on an SDS-PAGE gel. The gel was soaked for 30 min in fixing solution (25/65/10 isopropanol/H2O/acetic acid) and additionally treated for 30 min with “Amplify” (Amersham Biosciences), dried, and exposed to film for 6 wk.
Proteins were separated by SDS-PAGE under reducing or nonreducing conditions and then transferred to Hybond C-extra membranes by electroblotting. Membranes were blocked in 0.1% TBST, containing 5% (w/v) nonfat dried milk, probed with unconjugated primary Ab for 1 h and revealed with HRP-conjugated goat anti-mouse or goat anti-rabbit IgG (1/20,000 dilution). For phosphotyrosine detection, SDS-PAGE was performed under reducing conditions, and detection was done with incubation with 4G10 HRP-conjugated. Immunoblots were developed using ECL or ECL-plus (Amersham Biosciences) and exposed to CL-XPosure films (Pierce).
Densitometric analyses were performed on autoradiographs of the SDS-PAGE gels, using a GS-800 Densitometer (Bio-Rad) and Quantity One software. All densitometry values obtained were calculated from nonsaturated signals.
For lipid rafts analysis, ∼1.7 × 108 cells were used per sample. Cells were washed and resuspended in 1 ml of RPMI 1640 medium containing OX-34 at a 1/100 dilution of ascitic fluid. After 5 min of incubation on ice, cells were cross-linked with rabbit anti-mouse at 20 μg/ml and maintained at 37°C for 15 min. Cells were then washed twice with ice-cold PBS, lysed, and fractionated by sucrose gradient centrifugation as described.
For phosphotyrosine detection, the procedure was as described (42), using a combination of OX-54 plus OX-55 at 10 μg/ml each. Cell activation for kinase assays was performed as previously described (11), using OX-54 plus OX-55 at 10 μg/ml each.
Immunoprecipitations and in vitro kinase assays
Immunoprecipitations, in vitro kinase assays, and reprecipitation of phosphorylated substrates were performed as previously described (19).
A significant proportion of CD2 associates with plasma membrane lipid rafts in resting rat T cells and in a rat thymoma cell line
Previous work has shown that in human resting T lymphocytes the transmembrane glycoprotein CD2 is virtually absent from membrane lipid rafts, as defined based on a Triton X-100 insolubility criterion (34). However, in mouse cells a considerable fraction of CD2 molecules are raft-resident (32). We have investigated the membrane localization of the rat homolog of CD2 in resting primary cells as well as in the rat thymoma cell line C58. Nonactivated splenocytes or C58 cells were solubilized in a lysis buffer containing 1% Triton X-100, and the lysates were subjected to sucrose gradient centrifugation. The localization of the molecules within or outside lipid rafts was evaluated by immunoblotting with the mAb OX-55, using as marker for the rafts fractions immunoblottings of the resident palmitoylated molecule LAT (27). As measured by densitometry of immunoblots in different experiments, between 7 and 15% of rat CD2 is constitutively confined to the rafts in unstimulated splenocytes, and the majority of CD2 molecules are recovered from the “soluble” phase (Fig. 1⇓). Using nonstimulated C58 cells, we could observe that CD2 is over-represented in the rafts (∼75%), compared with the non-rafts fractions. The increased amount of rat CD2 in the rafts of the thymoma cell line, compared with the resting physiological ex vivo cells, probably reflects the level of permanent activation of the cycling cells.
Rat CD2 targeting to lipid rafts is not fully determined by palmitoylation of Cys226 and Cys228
The difference in raft localization between human and mouse CD2 has been attributed to two membrane-juxtaposed cysteine residues present in the mouse amino acid sequence and absent from the human molecule, which could be palmitoylated and thus address the molecule to lipid rafts. We examined whether the membrane-juxtaposed cysteine residues Cys226 and Cys228, also present in rat CD2, might have a role in targeting the molecule to the rafts, using a mutant for which these residues were replaced by alanines. The mutant CD2(CIC/AIA) was stably expressed in E6.1 Jurkat cells, and surface expression levels were confirmed by flow cytometry (data not shown). E6.1-CD2(CIC/AIA) cells, as well as E6.1 cells expressing full-length rat CD2, E6.1-CD2, were lysed in 1% Triton X-100-based lysis buffer, lysates were subjected to sucrose gradient separation and analyzed by immunoblotting. As can be seen in Fig. 2⇓, wild-type rat CD2 is largely concentrated in lipid rafts (74% as measured by densitometry); however, rat CD2(CIC/AIA) is still able to localize very significantly to the rafts (49%), indicating that the Cys226 and Cys228 are not fully accountable for the lipid raft-targeting of rat CD2.
We next examined whether rat CD2 could in fact be palmitoylated in the cysteine residues. In these studies we used E6.1-CD2 cells, and cells expressing a cytoplasmic deletion mutant, E6.1-CD2(CY 6), containing only the first six amino acids of the cytoplasmic domain, but still retaining the two cysteine residues. Cells were incubated in medium containing [3H]palmitate, for 16 h at 37°C. Following cell lysis, rat CD2 was immunoprecipitated from E6.1-CD2 and E6.1-CD2(CY 6) cells, and samples from the lysates as well as from the immunoprecipitates were run on SDS-PAGE under nonreducing conditions because the presence of 2-ME is recognized to interfere with S-ester and hydroxyester linkages, as can be the case for the covalent linkage of palmitate with the cysteine residues (43). As a positive control for the experiment, we immunoprecipitated Lck, previously shown to incorporate [3H]palmitate (44).
Fluorography analysis of radiolabeled products in SDS-PAGE showed that a number of cellular proteins did incorporate [3H]palmitate, as can be seen in the lysate lanes (Fig. 3⇓A). Lck was easily detected in the respective lane. However, we were not able to confirm that rat CD2 could be subjected to lipid modification, as no labeled protein bands could be perceived in the lanes of rat CD2. The failure to detect palmitoylated rat CD2 could not be attributable to a low cellular expression of the molecule or to a low affinity of the Abs. As seen in Fig. 3⇓B, rat CD2 immunoprecipitated from E6.1-CD2 and E6.1-CD2(CY 6) cells is clearly evident by immunoblotting. Moreover, rat CD2 was well expressed at the cell surface as confirmed by FACS analysis (Fig. 3⇓C).
Raft localization of rat CD2 mutants correlates with their association with Lck
To determine the molecular basis for the constitutive presence of a large fraction of rat CD2 in the rafts, we focused on the cytoplasmic domain, determining which segments could play a role. We used previously characterized E6.1 Jurkat clones stably expressing cytoplasmic domain truncations of rat CD2 retaining 97, 81, 66, 40, and 6 aa of the cytoplasmic tail (12, 17, 19, 38). All CD2 variants were expressed at similar levels at the membrane (data not shown). The location of each mutant was analyzed following sucrose gradient centrifugation and immunoblotting of rat CD2 with OX-55. As shown in Fig. 4⇓, the deletion mutants CD2(CY 97), CD2(CY 81), CD2(CY 66), and CD2(CY 40), like the full-length CD2 molecule, are mostly raft-resident, whereas rat CD2(CY 6) is totally excluded from lipid rafts and found in the soluble fractions. This finding suggests that the first 40 N-terminal amino acids of CD2 cytoplasmic tail are sufficient for the localization of CD2 in lipid rafts.
As the pattern of association of rat CD2 truncation mutants with lipid rafts correlated entirely with the pattern of binding of these same mutants to the tyrosine kinase Lck (19), we raised the possibility that the interaction between rat CD2 and lipid rafts could be mediated via Lck. Moreover, as seen from Fig. 4⇑, all CD2 mutants except CD2(CY 6) are in general confined to the same fractions as Lck. We tested further this hypothesis by constructing additional CD2 mutants that would have their association with Lck disrupted while retaining most of the cytoplasmic tail. CD2(CY 6) does not associate with Lck (19) and lies outside lipid rafts, whereas CD2(CY 40) retains both those features. We therefore constructed and tested a mutant CD2 molecule excluding precisely the sequence between aa 7 and 40 of the cytoplasmic domain. This new mutant, CD2(Δ7–40), was expressed at the surface of E6.1 cells, as detected by flow cytometry (Fig. 5⇓A), and had an apparent molecular mass compatible with its slightly shorter tail (Fig. 5⇓B). Strikingly, when we assessed the membrane localization of CD2(Δ7–40), we determined that this mutant does not localize within lipid rafts, but rather in the soluble fractions (Fig. 5⇓C).
We then tested whether CD2(Δ7–40) was incompetent for interacting with Lck, performing immunoprecipitations followed by kinase assays of the immune complexes of CD2 from E6.1-CD2(Δ7–40) cells, as well as from E6.1-CD2 cells for comparison. Among the proteins coprecipitated with CD2 from E6.1-CD2 cell lysates, a prominent phosphoprotein of 56 kDa was clearly visible, whereas no corresponding protein band was detected in the immune complexes from E6.1-CD2(Δ7–40) cells, although a number of other phosphoproteins with similar sizes are present (Fig. 5⇑D). To confirm that the putative 56-kDa protein was indeed phosphorylated Lck, we performed a reprecipitation of Lck using a polyclonal Ab and confirmed the identity of p56 as Lck. No protein was detected in the reprecipitation of Lck from E6.1-CD2(Δ7–40) cells, suggesting that CD2(Δ7–40) looses the ability to bind to Lck, therefore strengthening our assumption that CD2 is trapped in lipid rafts fractions due to its association with the kinase.
We checked whether having CD2 associated with Lck in rafts, as presented in E6.1-CD2 cells, would result in an increase of the Lck activity associated with CD2 in activated cells. In Fig. 5⇑E, stimulation of CD2 with mAb triggers a sharp increase of the kinase activity associated with CD2 as shown by the increased autophosphorylation of Lck. By contrast, triggering CD2 in E6.1-CD2(Δ7–40) cells did not induce any detectable changes.
A physical association of rat CD2 with Lck is sufficient for the raft targeting of CD2
To obtain definitive proof that Lck is the key determinant in the addressing of rat CD2 to lipid rafts, we resorted to the Lck-deficient J.CaM1 cellular model. Rat CD2 was stably expressed in J.CaM1 cells and its inclusion in the membrane fractions analyzed. Supporting our prediction, rat CD2 expressed in J.CaM1 cells localizes to the soluble fractions, whereas the kinase is absent in these cells as expected (Fig. 6⇓A). Further evidence that Lck could couple rat CD2 to lipid rafts was obtained through co-capping experiments in which we induced capping of CD2. Following fixation of the cells, we analyzed whether rafts would colocalize with polarized CD2 that would demonstrate a functional association between rafts and CD2. Rat CD2 labeled with OX-34-FITC could be visualized (Fig. 6⇓B, green), and the localization of rafts was detected by Alexa Fluor 568 (Fig. 6⇓B, red) conjugated to Abs recognizing raft-resident LAT. Whereas in untreated cells no polarization of CD2 or rafts were induced (Fig. 6⇓B, E6.1-CD2 cells, upper row; J.CaM1-CD2 cells, data not shown), in E6.1-CD2 cells, capping of CD2 induced very clear co-capping of LAT-labeled rafts in the majority of cells, nearly 80% (Fig. 6⇓, B and C). By contrast, in the Lck-deficient J.CaM1-CD2 cells, capping of CD2 induced by CD2 Abs was not always accompanied by co-capping of LAT. In the fewer cells (31%) where we could observe LAT polarization, colocalization of LAT with CD2 was less well defined. Very similar results were obtained using co-capping of CD2 together with cholera toxin subunit B that binds to raft-embedded GM1 (data not shown).
Our hypothesis that Lck couples CD2 to the rafts was further strengthened when J.CaM1-CD2 cells were reconstituted with Lck, and targeting of CD2 to the lipid rafts fractions, colocalizing with the re-expressed Lck, was by large restored (Fig. 7⇓). To determine whether a functional Lck kinase was required for addressing CD2 to the rafts, we transfected J.CaM1-CD2 cells with an Lck cDNA encoding a mutation at the ATP binding site (K273A), and so referred to as kinase defective. J.CaM1/Lck(KD)-CD2 cells were lysed and subjected to sucrose gradients, SDS-PAGE, and immunoblotting with OX-55. As can be seen in the respective panels in Fig. 7⇓, both Lck(KD) as well as CD2 still localize in lipid rafts, discarding the Lck kinase activity as required for the lipid raft-CD2 localization.
It was apparent that in Jurkat cell lines, the gross majority of rat CD2 molecules would localize with Lck in lipid rafts when a physical interaction could be established. We thus investigated whether CD2 would still localize in the rafts if Lck was excluded, and for this experiment we used J.CaM1 cells reconstituted with an Lck mutant that has had Cys3 replaced by an alanine residue. Lck contains a double palmitoylation site at Cys3 and Cys5 (41). In contrast to a Lck mutant C5A or the double mutant C3.5A, which are mainly or exclusively cytoplasmic, respectively, Lck(C3A) confines to the plasma membrane. However, as can be seen in Fig. 7⇑ (lower blots), it no longer addresses to the lipid raft fractions, neither does exogenous rat CD2 expressed in these cells, thus proving that CD2 homing to lipid rafts is dependent on a physical association with Lck.
The association between CD2 and Lck is held preferentially within lipid rafts, but it is not inhibited in the soluble phase
It did not become clear, however, whether in J.CaM1/Lck(C3A)-CD2 cells rat CD2 could associate with soluble phase Lck(C3A), and through this association was dragged out of lipid rafts, or whether it was free and kept out from the rafts by default, with the theoretical possibility of being recruited through the association of another raft-resident molecule. We tested these hypotheses through the analysis of molecular associations of rat CD2, expressed in J.CaM1/Lck(WT)-CD2 as well as in J.CaM1/Lck(C3A)-CD2 cells, thus addressing the interaction with raft-resident Lck vs raft-excluded Lck molecules.
Nonactivated cells were lysed in Triton X-100 lysis buffer, Lck or CD2 were immunoprecipitated from the lysates, and immune complexes were subjected to in vitro kinase assays. CD2 immunoprecipitates from J.CaM1/Lck(C3A)-CD2 cells displayed substantially less kinase activity when compared with CD2 immunoprecipitates from J.CaM1/Lck(WT)-CD2 (Fig. 8⇓A). Because raft-resident Lck(WT) and raft-excluded Lck(C3A) displayed comparable kinase activities and autophosphorylation efficiency, this suggested that the association between CD2 and Lck is preferentially held in the rafts, and that association in the soluble phase is much reduced, albeit still detectable.
Given that the T cell-specific Src-like kinase Fyn has also been shown to contribute to some of the CD2-associated kinase activity, we tested whether Fyn could also associate with CD2 in J.CaM1/Lck(WT)-CD2 and J.CaM1/Lck(C3A)-CD2 cells. The primary CD2 immune complexes were disrupted by heat and SDS, and Fyn as well as Lck were reprecipitated. As can be seen from the lower panels of Fig. 8⇑A, both Lck and Fyn are clearly present in CD2 complexes from J.CaM1/Lck(WT)-CD2 cells, but mostly absent from CD2 immunoprecipitates from J.CaM1/Lck(C3A)-CD2 cells, indicating that in nonactivated cells CD2 associates with both kinases mostly within lipid rafts.
Using the set of Lck mutants expressed in J.CaM1 cells, we tested the signaling ability of CD2 molecules dispersed in the soluble phase of the membrane vs CD2 molecules in rafts. J.CaM1/Lck(WT)-CD2 and J.CaM1/Lck(C3A)-CD2 cells, as well as the control J.CaM1-CD2, were stimulated with the CD2 mAb OX-54 and OX-55. At different time points, reactions were stopped and lysates analyzed by Western blotting for tyrosine phosphorylation (Fig. 8⇑B). J.CaM1/Lck(C3A)-CD2 cells had actually a faster response, peaking at 1 min, but a response that declined fast as well, whereas J.CaM1/Lck(WT)-CD2 displayed the peak of phosphorylation of a larger number of substrates at 2 min, and with a more sustained response (Fig. 8⇑B).
Given that in J.CaM1/Lck(C3A) cells Lck is excluded from rafts, we could use this cell line to address the raft function of CD2 in the absence of Lck, if we could set up the conditions to artificially address CD2 to the membrane microdomains. We thus engineered a CD2 mutant containing the transmembrane and membrane-proximal intracellular sequences of the raft-resident adaptor PAG (28). Unfortunately, this mutant did not partition to lipid rafts in J.CaM1 cells, as shown in Fig. 9⇓C. Conversely, this result further strengthens our claim that the CD2-Lck association is central to deliver CD2 to lipid rafts.
The traffic of CD2 in the membrane of rat T cells upon CD2 activation
We established that CD2 localizes to rafts dependently of its association with Lck and that within rafts also interacts with Fyn, both protein tyrosine kinases deeply involved in signal transduction in T cells. Thus, we investigated whether upon cell activation CD2 could shift through different membrane microenvironments, and what would be the role of CD2 reallocation in the regulation of kinase activity in the different membrane phases.
Splenocytes from 3-mo-old Wistar rats were stimulated through CD2 cross-linking for 15 min at 37°C or left unstimulated. Cells were lysed in Triton X-100-based lysis buffer and lysates subjected to sucrose gradient. The different recovered fractions were analyzed for the presence of CD2, Lck, and Fyn. In resting cells, the majority of CD2 molecules, but also of Lck, are outside lipid rafts, whereas virtually all Fyn resides in the rafts (Fig. 10⇓, A and B). Following CD2 stimulation, the proportion of Lck and Fyn in and outside rafts remained unchanged. However, a significant fraction of CD2 was translocated to lipid rafts. Quantification of the immunoblots by densitometry indicates that the percentage of CD2 in the rafts increased over 4-fold from 8 to 36% (Fig. 10⇓B).
To analyze the kinase activity associated with CD2 in the different fractions, we selected fraction 2 as representative of rafts, fraction 9 as a soluble fraction, and also a control fraction 6 (Fig. 10⇑A, arrowheads), where no CD2, Lck, or Fyn are present. CD2 was immunoprecipitated from fractions 2, 6 and 9, both from resting as well as from activated cells, and immune complexes were subjected to in vitro kinase assays. Immune complexes of Lck were processed in parallel. As can be seen in Fig. 10⇑C, phosphoproteins of 55–60 kDa were present in CD2 immune complexes already in resting cells, and were better defined in the rafts than in the soluble fraction, with no signal in the control fraction 6. Upon cellular activation, increased phosphorylation of the CD2-associated 55- to 60-kDa proteins was detected in the rafts and the soluble fractions. These phosphoproteins corresponded mostly to phosphorylated Lck and also Fyn, as can be seen in the reprecipitations of the kinases from the original CD2 immunoprecipitates.
In resting cells, the kinase activity of Lck reflects the proportion of the protein in and out of rafts. However there is a clear enhancement of the activity of raft-based Lck upon CD2 stimulation that does not require any translocation of Lck to the rafts (Fig. 10⇑D, refer to Fig. 10⇑, A and B). Interestingly, Fyn, which is not known for associating specifically to Lck, coprecipitates with Lck only in rafts, and displays higher activity following CD2 stimulation (Fig. 10⇑D, bottom blots). This suggests that Lck and Fyn may interact in lipid rafts of activated cells, and that this association may be mediated by CD2.
It is increasingly clear that lipid rafts play an important role in T cell signaling because the assembly of the TCR signaling machinery takes place within this membrane microenvironment enriched in glycosphingolipids, sphingomyelin, and cholesterol (30, 45, 46). Moreover, cross-linking the TCR with accessory proteins such as CD2, CD5, CD9, and CD28 readily increases the association of the TCR with lipid rafts (31, 32, 33). The conventional view is that the TCR and other signaling molecules translocate to these platforms that facilitate signal transduction mechanisms. However, given that MHC-engaging T cell receptors that originate activation and signaling, together with accessory molecules binding to the respective counter receptors, are polarized toward the sites of cell contact with APCs, a fairer assessment is that lipid rafts coalesce into the interface of TCR-Ag recognition (47, 48).
Therefore, rather than considering individual TCRs or other signaling molecules being cumulatively trapped by the raft, depending on certain characteristics such as intrinsic affinity for the raft lipids or interactions with raft-peripheral proteins, a TCR-centered perspective should consider instead how the whole of the raft is captured at the sites of Ag recognition. There are currently many open theoretical possibilities, but a simple one advances that lipid rafts need not be fairly large or complex structures: homotypic FRET studies have shown that some individual rafts can form high-density clusters of nanometer size (4–5 nm), containing as few as four GPI-anchored molecules (49). Moreover, Douglass and Vale (50) showed that the raft-associated molecule LAT, as well as Lck, had a much larger diffusion coefficient than non-raft CD2, suggesting that lipid rafts can be highly mobile and diffuse to the TCR activation spots. The study also suggests that, as Lck and LAT change from a highly mobile to a motionless state when they encounter a signaling cluster, it is possible that protein-protein interactions, rather than any sort of raft-addressing labels, can play a major role in establishing specific interactions between raft and non-raft proteins. We demonstrate that the association of rat CD2 with membrane lipid rafts is in fact mostly determined by the physical interaction established with the protein tyrosine kinase Lck.
CD2 has been shown to associate to both Lck and Fyn protein tyrosine kinases (20, 21), and deletion of the cytoplasmic tail eliminates these interactions and abolishes CD2-mediated signaling (17, 51). Given that our results, using several mutant molecules of CD2 and Lck and different cells lines, show an absolute correlation between the capacity of rat CD2 to interact with raft-resident Lck and its ability to be found in lipid rafts, we can conclude that it is Lck, and not Fyn, that retains CD2 at the rafts in nonactivated cells. Moreover, the levels of Fyn expression in the J.CaM1 cells used in this study were fair (data not shown), so it should not be attributed to Fyn a relevant role in the primary addressing of CD2 to lipid rafts, although it may play a part in retaining CD2 within rafts following cell stimulation.
The interaction between Lck and CD2 has been extensively characterized, and it has been shown that a GST/Lck-SH3 fusion protein could bind, in solution, to peptides containing different proline-rich sequences of the cytoplasmic domain of rat CD2 (21). However, SH3-domain/proline-rich contacts do not fully account for the total of the CD2-Lck association, as the first 40 amino acids of the CD2 tail are enough for coprecipitating Lck (19). It was nevertheless unexpected that removal of these amino acids would result in such a sharp decrease in the level of association (Fig. 5⇑, D and E), given that CD2(Δ7–40) still retains all proline-rich sequences. It is possible that the stretch of aa 6–40 of the CD2 cytoplasmic tail is crucial to set up the appropriate distance of the proline sequences to the SH3 domain of Lck, or to correctly orient the relevant modules. However, it is also plausible that this sequence, although functionally coupling CD2 to Lck, does not associate directly with Lck, and that other proteins or protein complexes mediate the CD2-Lck interaction at the membrane-proximal level and help directing CD2 localization. Thus, a more detailed analysis of the CD2 membrane-proximal sequence is required for establishing the key features and the functional consequences of the molecular interactions between CD2, Lck, and lipid rafts.
It is, however, unlikely that Lck is responsible for the actual cotransport of CD2 to the rafts. CD2 and Lck do not seem to interact extensively outside lipid rafts in resting cells, as concluded from experiments in which mutated Lck was excluded from the rafts (Fig. 8⇑A), as well as from the analysis of associations of the different fractions of physiological cells (Fig. 10⇑). Moreover, whereas there is an increase of 4.5-fold of CD2 into the rafts upon CD2-induced activation, there is no net shift of Lck to the rafts (Fig. 10⇑, A and B). Furthermore, Yang and Reinherz (34) have shown that the recruitment of activated CD2 to the rafts does not require the cytoplasmic tail and, hence, any association with Src-like kinases. Therefore, it is plausible that following TCR-Ag recognition and CD2 engagement to its ligand, and upon an encounter with a lipid raft, transitory interactions mediated through the CD2 ectodomain with a raft-resident receptor may allow for a transient overlap of the raft with CD2 complexes and induce raft-resident Lck to dock to the cytoplasmic domain of CD2.
Analysis of the kinase assays from Fig. 10⇑C suggested that CD2 could precipitate more active Lck and Fyn from the rafts fraction than the soluble fraction. Whereas more Fyn is expected to associate with CD2 in the rafts, given that 99% of Fyn is raft-resident, the fact that more active Lck was also coprecipitated with CD2 from the rafts fraction clearly confirms that the association of CD2 and Lck is favored within lipid rafts. This confirmation is despite the fact that the majority of CD2 and Lck are not resident in lipid rafts, neither in resting nor in activated cells. Nevertheless, some level of interaction between CD2 and Lck, and even with Fyn, can be perceived in the soluble fractions in activated cells (Fig. 10⇑C). The assembly of signaling complexes outside rafts is, albeit reduced, still possible and could explain the level of protein phosphorylation observed upon activation of J.CaM1/Lck(C3A)-CD2 cells, where neither Lck nor CD2 localize to lipid rafts (Fig. 8⇑B).
CD2 stimulation clearly induces a functional association between Lck and Fyn (Fig. 10⇑D). Lck and Fyn bind to nonoverlapping sequences of CD2 (21, 22), but other CD2 partners compete for these sequences. The CD2 proline-rich motif PPPPGHR is complexed with the GYF domain of CD2BP2, but only outside lipid rafts, as upon translocation, the SH3 domain of Fyn displaces CD2BP2 (22). The penultimate CD2 proline motif PPLPRPR has been reported to associate to the SH3 domains of both Lck and CD2BP1 (21, 52), and so it is possible that the association between CD2 and Lck may be similarly regulated and confined to lipid rafts as well. As the amounts of Lck and Fyn, in and outside the rafts, remain unchanged following activation, the increase of the phospho-Fyn signal coprecipitated with Lck (Fig. 10⇑C, bottom blot) has to be due to an increase in the amount of Fyn associating with Lck or to the phosphorylation of Fyn induced by Lck (29). Conversely, the augmented Lck signal in the raft fraction of activated cells can be caused by an increase in the phosphorylation of Lck, possibly catalyzed by Fyn (Fig. 10⇑D, bottom). Therefore, a central role of CD2 in the reorganization of lipid rafts upon cellular activation could be the facilitation of the association between Lck and Fyn.
Finally, what can then be the role of the two membrane proximal cysteine residues of rat CD2? It is intriguing that even using a chimeric CD2 molecule where we have inserted the relevant sequences of PAG, known to address very effectively the adaptor to lipid rafts (53), we failed to link CD2 to lipid rafts (Fig. 9⇑). Palmitoylation of integral proteins can be a very dynamic and reversible process, sometimes dependent on activation signals (54). For different transmembrane receptors, palmitoylation can occur at the plasma membrane or already in the endoplasmic reticulum (55, 56), so it is possible that CD2 is not available to lipid modifications depending on its own biosynthetic pathway, or that CD2 palmitoylation may depend on very specific activation signals, or even that the ectodomain may interact with other proteins that in some form can prevent the assembly of the complexes involved in lipid modifications to integral proteins. Alternatively, CD2 cysteines may perhaps help establishing protein-protein interaction with raft-resident molecules, although it is unlikely that they mediate the actual association with Lck. The cysteine residues of Lck known to interact with CD4 and CD8 are localized further away from the membrane (57) and thus unable to contact to the CD2 membrane-juxtaposed residues. The determination of the function of these two amino acid residues will no doubt require further investigation.
We thank Simon Davis (The Weatherall Institute of Molecular Medicine, University of Oxford) for valuable comments and discussion of the manuscript.
The authors have no financial conflict of interest.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
↵1 This work was funded by European Regional Development Fund FEDER, by Programa Operacional Ciência, Tecnologia, Inovação (POCTI) and Programa Operacional Ciência e Inovação 2010 from the Fundação para a Ciência e a Tecnologia (FCT), and support from the Programa Pessoa (cooperation Grices/Egide). R.J.N. and M.B. are recipients of studentships from FCT-POCTI. M.A.A.C. is supported by a Postdoctoral Fellowship from FCT-Programa Operacional Sociedade da Informação.
↵2 Current address: Lymphocyte Development Group, Medical Research Council, Clinical Sciences Centre, Imperial College School of Medicine, Hammersmith Hospital, Du Cane Road, London W12 0NN, U.K.
↵3 Address correspondence and reprint requests to Dr. Alexandre M. Carmo, Group of Cell Activation and Gene Expression, Instituto de Biologia Molecular e Celular, Universidade do Porto, 4150-180 Porto, Portugal. E-mail address:
↵4 Abbreviations used in this paper: SH3, Src homology 3; PAG, phosphoprotein associated with glycosphingolipid-enriched microdomains; LAT, linker for activation of T cell.
- Received November 5, 2007.
- Accepted November 5, 2007.
- Copyright © 2008 by The American Association of Immunologists