Dynamic recruitment of human CD2 into lipid rafts. Linkage to T cell signal transduction.

CD2 mediates T cell adhesion via its ectodomain and signal transduction utilizing its 117-amino acid cytoplasmic tail. Here we show that a significant fraction of human CD2 molecules is inducibly recruited into lipid rafts upon CD2 cross-linking by a specific pair of mitogenic anti-CD2 monoclonal antibodies (anti-T11(2) + anti-T11(3)) or during cellular conjugate formation by CD58, the physiologic ligand expressed on antigen-presenting cells. Translocation to lipid microdomains is independent of the T cell receptor (TCR) and, unlike inducible TCR-raft association, requires no tyrosine phosphorylation. Structural integrity of rafts is necessary for CD2-stimulated elevation of intracellular free calcium and tyrosine phosphorylation of cellular substrates. Whereas murine CD2 contains two membrane-proximal intracellular cysteines, partitioning CD2 into cholesterol-rich lipid rafts constitutively, human CD2 has no cytoplasmic cysteines. Mapping studies using CD2 point mutation, deletion, and chimeric molecules suggest that conformational change in the CD2 ectodomain participates in inducible raft association and excludes the membrane-proximal N-linked glycans, the transmembrane segment, and the CD2 cytoplasmic region (residues 8-117) as necessary for translocation. Translocation of CD2 into lipid rafts may reorganize the membrane into an activation-ready state prior to TCR engagement by a peptide associated with a major histocompatibility complex molecule, accounting for synergistic T cell stimulation by CD2 and the TCR.

T cell activation occurs following binding of T cell receptors (TCRs) 1 to peptides associated with major histocompatibility complex molecules (pMHC). This process is highly specific and sensitive despite the fact that the binding affinity between the TCR and pMHC complex is low, with relatively fast off-rates (1). Moreover, a small number of pMHC per antigen-presenting cell (APC) is sufficient to trigger and activate T cells (2). To maintain such a high degree of sensitivity and fidelity of T cell recognition, a number of co-stimulatory and accessory T cell surface molecules are coordinately engaged in the immune recognition process. These include CD2, CD4, CD8, CD28, and LFA-1 (3-6) among others. Furthermore, activation signals are induced, whereas signals inhibiting cell growth and effector function are suppressed.
Recently, plasma membrane compartmentation was shown to be involved in TCR signaling and the partitioning of essential T cell-activating components (reviewed in Refs. [7][8][9][10][11]. Certain constituents of the plasma membrane domain form an ordered liquid phase with reduced membrane fluidity (12), and these are termed lipid rafts, glycolipid-enriched membrane domains, or detergent-insoluble glycolipid-enriched microdomains (DIGs). These rafts maintain a distinct composition of lipid, sphingolipid, and cholesterol (12,13) and are resistant to mild detergent extraction, floating to the low density fraction on sucrose gradients (7,10), and can be easily isolated. They are thought to participate in protein sorting, membrane trafficking, and signal transduction (10) since numerous signaling molecules were found localized in DIGs (7,10). These include Src family protein-tyrosine kinases (14,15), heterotrimeric and monomeric Ras-like G proteins (16,17), molecules involved in Ca 2ϩ flux (18), as well as the small signaling molecule phosphatidylinositol bisphosphate (17,19). The co-localization of these signaling molecules with lipid rafts is largely due to post-translational acylation modification of proteins (7), the most important being membrane-proximal cysteine residue palmitoylation (reviewed in Ref. 20). Another post-translational modification that can bring cell surface proteins to the lipid rafts is glycosylphosphatidylinositol modification (21). Representatives of this class of cell surface molecule include Thy-1, CD48, CD58, Ly-6 ,and CD24. Although these structures lack both cytoplasmic and transmembrane domains, their cross-linking by suitable ectodomain-specific antibodies results in signal transduction and cellular responses.
The idea that lipid rafts can serve as a platform for lymphocyte signaling is supported by several recent reports demonstrating that multiple-chain immunoreceptors such as TCRs (17,22,23), B cell receptors (24,25), and high affinity IgE receptors (Fc⑀ receptor I) (26,27) are recruited to rafts and that raft integrity is required for effective signal transduction (reviewed in Ref. 9). A number of co-stimulatory molecules on the T cell surface can apparently up-regulate TCR signals by enhancing association of the TCR and lipid rafts. Representative molecules of the latter type include CD5 and CD9 (28) and CD28 (29).
The transmembrane T cell surface protein CD2 molecule is expressed on virtually all T cells, thymocytes, and natural killer cells (6). Human CD2 promotes the physical interaction of T and natural killer lineage cells with APCs, stromal ele-ments, and a variety of target cells bearing the ubiquitously expressed CD2 ligand, CD58 (3,6,30,31). CD2 functions to mediate T cell adhesion as well as to facilitate signal transduction. CD2 initiates T cell-APC contact even prior to TCR recognition of the pMHC (6,32,33). The ligation of CD2 by CD58 not only promotes the adhesion of T cells to APCs, but provides a suitable intercellular membrane spacing (ϳ140 Å) for TCR-pMHC recognition (34). Furthermore, the extremely fast onand off-rates of the CD2-CD58 interaction (35) enable a TCR to scan APC surfaces searching for the specific agonistic pMHC. Although the monomeric CD2-CD58 K d is weak (see references in Ref. 34), relatively strong binding of CD2-CD58 can be achieved by two-dimensional clustering at cell-cell junctions (32,36). These CD2 attributes can lower the physiologic TCR activation threshold (32,37,38).
Several observations indicate that CD2 participates in a distinct signaling pathway. First, binding of CD2 and CD58 augments the interleukin-12 (IL-12) responsiveness of activated T cells (39). Second, CD2-CD58 interaction plays a role in the reversal of T cell anergy (40). Third, although the CD2 tail is not responsible for its adhesion function (32,41,42), it is required for optimal CD2 augmentation of antigen-triggered T cell responses (33,(43)(44)(45)(46). In this regard, CD2 has a 117-amino acid cytoplasmic tail that is relatively conserved among different species (47,48). Unique pairs of CD2-specific monoclonal antibodies can transduce mitogenic signals to T cells and induce T cell proliferation and IL-2 production even in the absence of engagement of the TCR (33, 49 -51). Here we present data showing that CD2 can inducibly associate with lipid rafts and that raft integrity is important for CD2 signaling function. We also demonstrate that CD2-raft association is independent of tyrosine phosphorylation events as well as the CD2 cytoplasmic tail and TCR linkage to lipid rafts. Such inducible raft association accounts for many aspects of CD2-based signal transduction and explains the synergistic activation of T cells by CD2 and TCR molecules.
Cell Culture-Human peripheral blood mononuclear cells were isolated from leukopaks of healthy blood donors by Ficoll gradient centrifugation using a nylon wool column. The purified peripheral blood mononuclear cells were cultured in RPMI 1640 medium containing 10% human AB serum (Nabi, Boca Raton, FL), 1% penicillin/streptomycin (Life Technologies, Inc.), and 2 mM glutamine (Life Technologies, Inc.) (complete medium) and used within 0 -2 days as resting human T cells. Activated T cells were obtained by culturing the peripheral blood mononuclear cells as described above plus a 1:200 dilution of 2Ad2 ascites and 25 ng/ml phorbol 12-myristate 13-acetate. After 48 h of activation, 94% of the cells were anti-CD2 mAb-reactive as shown by fluorescenceactivated cell sorting analysis (42). Cells were then diluted 1:5, maintained in culture, and used within 1-3 days. For Ca 2ϩ flux, human T cells were activated for 48 h, washed once with complete medium, and cultured in the same medium supplemented with 100 units/ml recombinant human IL-2 (Amgen, Thousand Oaks, CA), and the cells were used within 3-6 days.
The human Jurkat cell line J77 was maintained in RPMI 1640 medium and 10% fetal calf serum with 1% penicillin/streptomycin and 2 mM glutamine in 5% CO 2 at 37°C. Chinese hamster ovary cells (CHOP) without or with CD58 (CHO58) or a K87A variant (CHO58K87A) were cultured at 37°C in a 5% CO 2 incubator in Dulbecco's modified Eagle's medium/nutrient mixture F-12 with 10% fetal calf serum, 1% penicillin/streptomycin, and 2 mM glutamine and additionally supplemented with 10 mM HEPES (pH 7.5) and 0.2 mg/ml G418 (Life Technologies, Inc.). The 155.16 mouse T cell hybridoma cells and human CD2-transfected derivatives were cultured in the same medium as the J77 cells, but additionally supplemented with 0.5 mg/ml G418.
mAb Cross-linking, T Cell Lysis, and Cell Fractionation-CD2 crosslinking was achieved by adding either a combination of purified anti-T11 2 and anti-T11 3 mAbs (10 g/ml) or a 1:100 dilution of ascites. CD3 cross-linking was accomplished by addition of a 1:200 dilution of 2Ad2 ascites. Cell activation was performed at 37°C for 10 -15 min in the relevant complete medium. For pretreatment with various inhibitors, 1 ϫ 10 7 J77 cells were incubated with 0.1% (v/v) Me 2 SO for 2 h, 5 g/ml cytochalasin D for 2 h, 0.5 M colchicine for 17 h, 0.5 M paclitaxel for 17 h, or 20 mM methyl-␤-cyclodextrin (␤-MCD) for 30 min at 37°C in complete J77 medium. Cells were spun down at 1000 ϫ g and resuspended in medium with or without antibody stimulation. After 15 min at 37°C, cells were washed with ice-cold phosphate-buffered saline and lysed in 0.5 ml of lysate buffer containing 1ϫ TBS and 1% Nonidet P-40 supplemented with 1 mM phenylmethylsulfonyl fluoride, 0.35 trypsin inhibitor units/ml aprotinin (Sigma), and 5 g/ml leupeptin (Roche Molecular Biochemicals). Following incubation on ice for 30 min, samples were spun down at 12,000 ϫ g for 10 min. Cell pellets were resuspended in lysate buffer containing 0.1% SDS and 0.5% deoxycholate, followed by brief sonication. Supernatants were recovered as the insoluble fraction of the cell lysate and quantified with a Micro BCA kit (Pierce). Equal amounts of total protein were loaded per lane for Western blot detection of CD2.
Lipid Raft Separation-Sucrose gradient centrifugation was used to fractionate membrane fractions according to the protocol of Zhang et al. (53) with some modification. Briefly, 1 ϫ 10 8 J77 or activated human T cells or 2.5 ϫ 10 8 resting human T cells were individually harvested, resuspended in 1 ml of complete medium, and stimulated by crosslinking as described above. Non-cross-linked samples were treated identically, but without addition of mAbs. After cross-linking, cells were rinsed once with ice-cold phosphate-buffered saline and then resuspended in 1 ml of ice-cold 1ϫ MES-buffered saline (25 mM MES and 150 mM NaCl (pH 6.5)) and 1% Triton X-100 supplemented with 1 mM phenylmethylsulfonyl fluoride, 0.35 trypsin inhibitor units/ml aprotinin, and 5 g/ml leupeptin. Following a 30-min incubation on ice, the lysates were homogenized with 10 strokes in a Dounce homogenizer (Wheaton, Millville, NJ), gently mixed with an equal volume of ice-cold 80% (w/v) sucrose (Fisher) in 1ϫ MES-buffered saline, and loaded in the bottom of a centrifuge tube (Beckman Instruments). The sample was then overlaid with 2 ml of 30% (w/v) sucrose and 1 ml of 5% (w/v) sucrose, both prepared in 1ϫ MES-buffered saline. The sucrose gradient samples were spun at 39,000 rpm in a DuPont AH650 rotor at 4°C for 20 -21 h. DIGs were harvested by collecting 0.4-ml fractions, beginning at the top of the gradient.
Western Blotting-All cell lysates and gradient fractions were treated with peptide N-glycosidase F (New England Biolabs Inc., Beverly, MA) to deglycosylate human CD2 according to the manufacturer's protocol. Final samples were mixed with 5ϫ SDS loading buffer and resolved by SDS-PAGE under reducing conditions. Gels were transferred to polyvinylidene difluoride membranes (Bio-Rad) in transfer buffer consisting of 25 mM Tris-HCl, 200 mM glycine, and 20% methanol. Membranes were blocked in 5% skim milk in TBS/Tween (25 mM Tris, 150 mM NaCl, and 0.05% Tween 20) at 37°C for 30 min. M32B anti-CD2 polyclonal antiserum was diluted 1:1000 in 5% bovine serum albumin and TBS/ Tween and incubated with membranes at 4°C overnight. The membranes were then incubated with horseradish peroxidase-conjugated goat anti-rabbit antibody (Bio-Rad) at room temperature for 1 h and subsequently developed using a chemiluminescence reagent kit (PerkinElmer Life Sciences) on MR film (Eastman Kodak Co.). ECLdeveloped blots were scanned with a Microtek ScanMaker 5 and processed with Photoshop software. The intensity of anti-T11 2 and anti-T11 3 mAb cross-linking in the presence of the control (Me 2 SO or no other addition) was arbitrarily set to 100% using ImageQuant Version 1.1 (Molecular Dynamics, Inc.).
For phosphotyrosine blotting, activated human T cells were left untreated or were pretreated with ␤-MCD or PP2, followed by crosslinking of CD2 or CD3. Whole cell lysates were generated in lysis buffer containing 1ϫ TBS and 60 mM n-octyl ␤-D-glucopyranoside supple-mented with protease inhibitors. Equal amounts of lysate proteins were loaded on 4 -12% gradient SDS-polyacrylamide gels (Bio-Rad) and then transferred to polyvinylidene difluoride membranes as described above. The membranes were blocked in 5% bovine serum albumin and TBS/ Tween at 37°C for 30 min and subsequently incubated with antibody 4G10 at 4°C overnight. 4G10 signals were detected with horseradish peroxidase-conjugated goat anti-mouse IgG2b, followed by ECL. The membrane was then stripped with 1ϫ stripping buffer (Chemicon International, Inc., Temecula, CA) and reblotted with antibodies against Fyn, LAT, Vav, and phospholipase C␥1.
Generation of T Cell Transfectants Expressing Mutant CD2 Molecules-Wild-type CD2 transfectants as well as mutant CD2 molecules (including the cytoplasmic deletion CD2⌬25; the single mutation variants K167W, K167A, K170A, W10F, E131A, E8A, N117Q, N126Q, and H140A; the double mutant N117Q/N126Q; and the triple mutant K61E/ F63L/T67D) were generated as described (41). TfR 25 and TfR TM are two variants in which the CD2 transmembrane domain was swapped with that of the transferrin receptor (TfR). These mutants were produced using a one-step reverse transcription-polymerase chain reaction (Life Technologies, Inc.) to clone the cDNA encoding the transmembrane domain of the TfR, which was subsequently fused by additional polymerase chain reaction cloning with the extracellular domain of CD2 with (TfR 25 ) or without (TfR TM ) a cDNA sequence encoding 25 amino acids of the CD2 cytoplasmic tail. 2 Subsequently, the desired DNA sequence was verified, and the product was cloned into the expression vector pcDNA3.1(ϩ) (Invitrogen, Carlsbad, CA). The final expression vectors were linearized by SalI digestion and transfected into the 155.16 mouse T cell hybridoma cells by electroporation at 200 V and 1180 microfarads (Life Technologies, Inc.). Following neomycin selection (1.5 mg/ml), stable transfectant cell lines expressing CD2 mutants at levels comparable to wild-type CD2 were established by sorting four to five times on a MoFlo (Cytomation, Fort Collins, CO).
Ca 2ϩ Flux-1 ϫ 10 6 activated human T cells (see above) were loaded with 2 g/ml (ϳ2 nM) Indo-1 acetoxymethyl ester (Molecular Probes, Inc., Eugene, OR) in 1 ml of complete medium at 37°C for 30 min. Following three washes with RPMI 1640 medium, cells were resuspended in RPMI 1640 medium and prewarmed for 5-10 min at 37°C before analysis. Ca 2ϩ flux was measured on a FACS Vantage (Becton Dickinson, San Jose, CA). Samples were run for 30 s to obtain a base line, and then a combination of anti-T11 2 and anti-T11 3 ascites (1:100 dilution of each) was added. The ratio 405/525 nm was recorded for 8 min. 2 g of 4-bromo A23187 (calcium ionophore) (Molecular Probes, Inc.) was added subsequently to ensure the successful loading of Indo-1. For ␤-MCD treatment, different concentrations of ␤-MCD were added to the loading buffer concurrently with Indo-1.

Cross-linking-mediated Translocation of Human CD2 to the Detergent-insoluble Fraction-Previous
studies from our laboratory have indicated that CD2 cross-linking by mitogenic pairs of mAbs (anti-T11 2 and anti-T11 3 ) results in loss of a fraction of CD2 molecules from the 1% Triton X-100 detergentsoluble component of T cell lysates. 3 Hence, CD2 must become associated with detergent-insoluble cellular components such as nuclei, the cytoskeleton, and/or specialized domains of the plasma membrane. To investigate the basis of this phenomenon, we pretreated J77 cells prior to CD2 stimulation with chemicals known to modulate these components. These reagents included cytochalasin D, which blocks the formation of actin filaments; colchicine, which inhibits microtubule formation; paclitaxel (Taxol), which stabilizes microtubule structures; and ␤-MCD, known to specifically extract membrane cholesterol and therefore disrupt plasma membrane lipid rafts (54). Cell viability was not affected by any of these treatments as indicated by trypan blue exclusion (data not shown); following treatment, the cells were then stimulated with anti-T11 2 plus anti-T11 3 mAbs; cell pellets were generated; and proteins were deglycosylated with peptide N-glycosidase F under denaturing conditions as described under "Experimental Proce-dures." Deglycosylation facilitates quantitative analysis of the CD2 protein amount by collapsing the various species bearing distinct glycan adducts into a single band. Equivalent amounts of protein were resolved by SDS-PAGE, and CD2 was quantitated by Western blotting.
As shown in Fig. 1, without CD2 cross-linking, CD2 partitioned entirely in the detergent-soluble lysate; and hence, none was detected in the detergent-insoluble fraction (lane 7). By contrast, following CD2 cross-linking, some CD2 was found in the insoluble fraction (lane 1). Interestingly, stimulation of J77 cells with mitogenic anti-CD3⑀ mAb 2Ad2 (IgM) did not shift the CD2 into the insoluble cell pellet (lane 6), indicating that CD2 association with the detergent-insoluble fraction is independent of TCR cross-linking and the attendant cell activation induced via anti-CD3 mAb. Perturbation of the cytoskeleton by the inhibitors cytochalasin D and colchicine or, conversely, by the stabilizer paclitaxel did not affect the partitioning of CD2 into the insoluble cell pellet (lanes 2-4). On the other hand, ␤-MCD prevented Ͼ50% of the CD2 from associating with the insoluble fraction after CD2 cross-linking (lane 5). This finding suggests that some CD2 molecules become linked to lipid rafts upon anti-CD2 mAb cross-linking since the above cytoskeletal actin and microtubule modifiers were without effect. Although we interpret the influence of ␤-MCD on CD2 localization to reflect its disrupting effect on lipid rafts, the cholesterol-depleting activity of the compound might modify membranes elsewhere in the cell, making the conclusion tentative.
Inducible Association of CD2 with Lipid Rafts-Since plasma membrane lipid rafts have distinct lipid and protein contents, rafts are easily separated by sucrose gradient centrifugation (10). We took advantage of this methodology to directly and independently examine if CD2 can be recruited to lipid rafts. To this end, activated human peripheral blood T cells either were stimulated by the combination of anti-T11 2 plus anti-T11 3 mAbs or were left untreated at 37°C for 10 min. 2 Polymerase chain reaction primer sequences are available upon request. 3  . Subsequently, cells were lysed in 1% Triton X-100 and centrifuged, and cell pellets were resuspended in 0.1% SDS and 0.5% deoxycholate buffer and solubilized by brief sonication. Equal amounts of proteins from each condition were treated with peptide N-glycosidase F (PNGase F) and then resolved by SDS-PAGE. CD2 was visualized by M32B polyclonal antibodies, followed by horseradish peroxidase-labeled goat anti-rabbit antibody and ECL.
Total cell lysates were generated by extracting cells in a mild nonionic detergent (1% Triton X-100) at 4°C. Sucrose gradient fractionation and subsequent Western blotting analysis were performed as described under "Experimental Procedures." Consistent with previous reports, DIGs are found in sucrose gradient fractions 2-4 (22,53), where the palmitoylated LAT molecule is predominantly located (53). Other proteins including ␣-tubulin are not associated with lipid rafts (22,24), but rather are located in the soluble cellular fractions. As shown in Fig. 2, following CD2 cross-linking, substantial amounts of CD2 (ϳ32-fold increase) were recruited to the lipid raft fraction, whereas in the absence of CD2 cross-linking, little if any CD2 was localized to the rafts. This result indicates that CD2 association with lipid rafts is inducible. Note that although the majority of LAT is found in DIG as reported (53), the recruited CD2 component represents only 5-10% of total cellular CD2. Although these data were generated using activated T cells, similar results were obtained using resting human T cells, the human J77 tumor cell line, human CD2 transgenic primary mouse T cells (55), and human CD2 cDNA-transfected 155. 16  A Mitogenic Pair of Anti-CD2 mAbs Is Necessary for CD2-Lipid Raft Association-The above results indicate that CD2 association with lipid rafts is stimulated by the combination of anti-T11 2 plus anti-T11 3 mAbs. In this regard, it was previously shown that there are at least three distinct epitopes on the human CD2 molecule (49): the T11 1 epitope localized to the CD58-binding face; the T11 2 epitope adjacent to the CD58binding site; and the T11 3 epitope, termed CD2R, which is mainly restricted to activated human T cells, unlike the T11 1 and T11 2 epitopes, and involves the linker region between domains 1 and 2 of CD2 (41). Jurkat cells were treated with different combinations of antibodies, and then rafts were fractionated by sucrose gradient centrifugation. The various DIG fractions were analyzed by Western blotting as shown in Fig.  3A. As expected, anti-T11 2 plus anti-T11 3 mAb cross-linking recruited CD2 to lipid rafts (lane 1). Earlier studies demonstrated that bivalent anti-T11 2 mAb, but not monovalent Fab, induced expression of CD2R (41,49) and, furthermore, that anti-T11 2 Fab plus anti-T11 3 IgG was non-mitogenic (41). As shown in Fig. 3A (lane 2), when we tested anti-T11 2 Fab in combination with anti-T11 3 mAb, no CD2-raft association was induced, consistent with the inability of this combination to stimulate proliferation. Given that anti-T11 2 mAb alone was not capable of recruiting CD2 to DIG (lane 3), we can also exclude Fc receptor binding as the basis of CD2 translocation.
These results show that neither anti-T11 2 nor anti-T11 3 mAb alone is sufficient to translocate CD2 to lipid rafts.
As the binding of anti-T11 2 and anti-T11 3 mAbs is essentially orthogonal to one another, the combination of the two mAbs presumably supports extensive lattice formation (41). To examine whether other forms of significant CD2 cross-linking could recruit CD2, we bound biotinylated anti-T11 2 mAb to J77 cells and then cross-linked the CD2 by subsequent addition of avidin. Fig. 3A (lane 4) shows that there was no CD2 translocation to the raft fraction after such treatment, consistent with the notion that specific cross-linking of CD2 by the mitogenic combination of anti-T11 2 plus anti-T11 3 mAbs is required. The combination of anti-T11 1 plus anti-T11 2 mAbs failed to recruit FIG. 3. Specific anti-CD2 antibody requirements for induction of CD2-lipid raft association. A, 8 ϫ 10 7 J77 cells were treated with the indicated combinations of anti-CD2 mAbs (10 g/ml) at 37°C for 10 min. Cell lysates were then fractionated on sucrose gradients, and the raft component was collected and subjected to Western blot (WB) analysis with anti-CD2 heteroantisera. The membranes were stripped and reblotted (RB) with anti-LAT antisera to demonstrate equivalent loading of raft fractions. Band intensity Ͻ20% of the anti-T11 2 ϩ anti-T11 3 mAb-induced sample is considered as background in the mAb-triggered responses of wild-type CD2 molecules. B, 1 ϫ 10 8 155.16 mouse T hybridoma cells transfected with human CD2 bearing the indicated mutations were treated (ϩ) with anti-T11 2 ϩ anti-T11 3 mAbs (10 g/ml) or were untreated (Ϫ) and were processed as described for A. Note that only the raft fractions are shown. Bio, biotinylated. CD2 to lipid rafts as well, consistent with this view (lane 5).
Additional data supporting the selective CD2 translocationinducing action of the combination of anti-T11 2 plus anti-T11 3 mAbs derive from analysis of mouse T cell hybridomas transfected with variant CD2 molecules involving mutation at the T11 2 or T11 3 epitopes (41). As shown in Fig. 3B, for example, CD2 molecules harboring the triple mutant K61E/F63L/T67D, which disrupts anti-T11 2 binding while retaining anti-T11 1 and anti-T11 3 binding activity, failed to be recruited to lipid rafts upon anti-T11 2 plus anti-T11 3 mAb addition. Other CD2 mutants that disrupt anti-T11 3 binding (including K167W, K167A, K170A, W10F, E8A, and H140A) also failed to move into DIGs with the same mAb cross-linking pair. On the other hand, the CD2 double mutant N117Q/N126Q, lacking N-linked adducts through modification of the two glycan addition sites within the membrane-proximal domain 2 of CD2 (without perturbing T11 1 -, T11 2 -, or T11 3 -binding sites), and the CD2 variant E131A, both of which retain anti-T11 2 -and anti-T11 3binding sites, associated with rafts upon addition of the anti-T11 2 plus anti-T11 3 mAb combination. The above results illustrate that both anti-T11 2 and anti-T11 3 mAbs are necessary and sufficient to recruit wild-type CD2 molecules or CD2 variants bearing T11 2 and T11 3 epitopes to lipid rafts. The other combinations of these anti-CD2 mAbs tested with or without additional CD2 cross-linking are ineffective.
Lipid Rafts Play a Key Role in CD2 Signaling-The above results demonstrate that translocation of CD2 into lipid rafts requires a specific pair of anti-CD2 antibodies known to transduce mitogenic signals. Given that lipid rafts are rich in T cell signaling molecules, these findings raise the question as to whether lipid rafts contain critical signaling components for the CD2 transduction pathway. To this end, we examined whether CD2 signaling is altered by perturbation of raft structures. We utilized ␤-MCD, a cyclic oligosaccharide consisting of seven glucopyranose units that extracts cholesterol from the plasma membrane to disrupt the raft structures (54). Activated human T cells were treated with ␤-MCD or left untreated, and then cells were stimulated via the CD2 or CD3 pathways with anti-CD2 or anti-CD3 mAbs, respectively. Fig. 4A shows that upon CD2 cross-linking, multiple proteins were tyrosine-phosphorylated as identified by anti-phosphotyrosine mAb 4G10 in Western blotting (compare lanes 1 and 2). Following anti-CD3 mAb-mediated cross-linking, certain proteins (including CD3 and LAT) were more heavily phosphorylated than upon CD2 stimulation (compare lanes 2 and 3). Importantly, pretreatment of T cells with ␤-MCD reduced tyrosine phosphorylation (compare lane 4 and lanes 1 and 2). The strong Src family tyrosine kinase inhibitor PP2 was used as an additional control (lane 5) and virtually abrogated tyrosine phosphorylation triggered by anti-CD2 mAb, reducing phosphorylation to below the basal level of unstimulated cells. This result is consistent with previous reports showing that PP2 is able to block early T cell activation events (56).
By reblotting the above membrane with antibodies specific for individual signaling molecules, we were able to examine the status of each after ␤-MCD treatment. The Src family tyrosine kinase p59 fyn was shown to be critical in the CD2 signaling, so in fyn Ϫ/Ϫ mice, murine T cells demonstrated impaired CD2 signaling in terms of Ca 2ϩ flux, T cell proliferation, and tyrosine phosphorylation (55). The current results show that ␤-MCD treatment decreased phosphorylation of p59 fyn 2-fold (Fig. 4A, lane 2 versus lane 4), reflecting reduced activity of the p59 fyn kinase for CD2 stimulation. The LAT scaffold protein, a substrate of the ZAP70 family of kinases, plays a central role in T cell activation (57,58). Although there is a severalfold increase in phospho-LAT upon CD2 stimulation, ␤-MCD treat-ment diminished this phosphorylation by 50%. Another intracellular signal molecule affected by ␤-MCD treatment is Vav. Vav is known to be involved in both T cell development and antigen receptor-mediated T cell activation due to its ability to trigger activation of ERK and NFAT as well as NF-B (59). Vav has also been shown to be essential for actin cytoskeletal reorganization and TCR capping (60,61). As shown in Fig. 4A, Vav was phosphorylated upon CD2 or CD3 stimulation, whereas virtually no phospho-Vav was detected following ␤-MCD pretreatment of anti-CD2 mAb-triggered T cells. This result is consistent with our recent report that the Vav signaling pathway is important for CD2-triggered T cell activation (55). The tyrosine phosphorylation of phospholipase C␥1 was also severely impaired after ␤-MCD treatment (Fig. 4A), suggesting that the enzymatic activity of phospholipase C␥1 is likewise decreased (62).
Since phospholipase C␥1 plays a critical role in calcium mobilization, CD2-triggered Ca 2ϩ flux could be affected by disruption of CD2 association with lipid rafts. To test this possibility, we loaded activated human T cells with Indo-1 and, in some cases, pretreated them with different concentrations of ␤-MCD. Cells were then analyzed by fluorescence-activated cell sorting and subjected to CD2 stimulation by anti-T11 2 plus anti-T11 3 mAb cross-linking. The ratio 405/525 nm was recorded as an indication of Ca 2ϩ flux. Fig. 4B shows two typical histograms of CD2 stimulation of T cells with or without ␤-MCD (2 mM) treatment. After ␤-MCD treatment, the Ca 2ϩ flux was decreased relative to the untreated T cell control. Subsequent addition of Ca 2ϩ ionophore indicated that both ␤-MCD-treated and untreated cells were loaded with Indo-1 to a similar extent. This decrease in CD2-triggered Ca 2ϩ flux was ␤-MCD dose-dependent, as shown in Fig. 4C. A decrease in calcium flux was evident at a ␤-MCD concentration as low as 0.5 mM and was almost completely inhibited at 8 mM. Thus, two early T cell activation events, tyrosine phosphorylation and a rise in intracellular-free calcium, were both impaired upon CD2 stimulation following perturbation of lipid rafts by ␤-MCD. These results indicate that the integrity of the lipid raft structure is important for CD2 signal transduction. Collectively, these data show that, in addition to TCR signaling, CD2 signaling requires a functional lipid raft membrane microdomain.
Independent Recruitment of CD2 and CD3 Molecules to Lipid Rafts-TCR cross-linking by mAbs or engagement with the pMHC was previously reported to recruit TCRs to lipid rafts (17,22). Co-stimulatory molecules such as CD28 (29), CD5, and CD9 exert their co-stimulatory functions, at least in part, by contributing to enhanced association between the TCR and lipid rafts (28). To determine whether CD2 mediates its costimulatory function by a similar mechanism, we stimulated activated human T cells with either anti-CD2 or anti-CD3 mAb cross-linking, separated raft fractions by sucrose gradient centrifugation, and analyzed them for inducible association with CD2 and/or CD3 by Western blotting. Consistent with previous reports, we failed to observe CD3 redistribution to the lipid raft fraction in the presence of 1% Triton X-100 (data not shown). This lack of association is probably due to the weak interaction between the TCR components and lipid rafts (8,22) since, by lowering the concentration of Triton X-100 to 0.2%, we were able to observe that significant amounts of CD3⑀ were recruited to DIGs (Fig. 5A). Upon CD2 cross-linking, however, there was no concurrent translocation of CD3 to the lipid raft fraction, although CD2 could be readily found associated with DIG. Conversely, upon anti-CD3⑀ mAb cross-linking, there was no increase in CD2 in the DIG fraction. The equivalent loading of raft fractions was evident from the reanalysis with anti-LAT antibody. This result strongly suggests that CD2-and CD3lipid raft associations are independent of one another. Additional evidence supporting this view comes from the observation shown in Fig. 1 (lane 6) that anti-CD3⑀ mAb did not induce association of CD2 with the detergent-insoluble fraction.
Since TCR recruitment to rafts requires tyrosine phosphorylation events (22), it was of interest to determine if, by analogy, similar phosphorylation was required for CD2-DIG association. In this regard, the CD2 tail can bind to Src family kinases, including Fyn (63) and Lck (64), which are known to be localized in rafts, as well as phosphatidylinositol 3-kinase (65), which is recruited to rafts upon T cell activation (17). As shown in Fig. 5B, PP2 treatment of cells could not block CD2 recruitment to lipid rafts, even though it virtually eliminated CD2triggered protein tyrosine phosphorylation (Fig. 4A, lane 5). Thus, tyrosine phosphorylation and CD2 translocation to DIGs are separable events. Similar phosphorylation-independent CD2 relocation to lipid rafts was observed in freshly isolated resting human T cells and in primary murine T cells from human CD2 transgenic mice with or without an accompanying fyn gene knockout (data not shown).
The CD2 Extracellular Domain Is Required for Lipid Raft Recruitment-To next determine the domain(s) of CD2 responsible for raft association, we tested a series of mutant and truncated CD2 molecules, including CD2⌬25 and CD2⌬7 (41), which retain only 25 and 7 amino acids, respectively, of the 117 CD2 tail residues (Fig. 6A). Surprisingly, CD2⌬25 and CD2⌬7 translocated to the DIG fraction after anti-CD2 mAb crosslinking (Fig. 6B), even though these truncations eliminate the components of the CD2 tail that interact with known signaling molecules (43,44). Thus, the CD2 tail is not responsible for the inducible DIG association, raising the possibility that either the CD2 extracellular domain or transmembrane region might be necessary. We therefore created chimeric molecules consisting of the CD2 extracellular segment fused with the transmembrane region of a non-raft resident molecule, CD71 (TfR). Prior studies showed that CD71 was excluded from lipid rafts using biochemical methods (17,22) as well as microscopic observations (23). To this end, we cloned the TfR transmembrane domain by reverse transcription-polymerase chain reaction and fused it to a human CD2 extracellular domain, with (TfR 25 ) or without (TfR TM ) appending the membrane-proximal 25 amino acid residues of the CD2 cytoplasmic tail. Subsequently, stable transfectant cell lines were established using the 155.16 mouse T cell hybridoma cells as a recipient. As shown in Fig. 6C, the CD2 transmembrane region was not responsible for CD2-DIG translocation: TfR 25 and wild-type CD2 (W33) molecules were equivalently translocated to DIG upon anti-CD2 mAb crosslinking. In the case of the TfR TM variant, translocation was also observed, although the reduced surface CD2 expression of the TfR TM variant yielded only a 3.0-fold increase compared with TfR 25 (9.2-fold increase). As an independent means of verifying these results, we made chimeric molecules using a mutant LAT transmembrane domain in lieu of that of CD71. In this mutant LAT segment, two cysteines were mutated to serine to abolish the palmitoylation of the LAT transmembrane domain, responsible for its localization in lipid rafts (53). These results also showed that the lipid raft association of the CD2 chimeric molecule upon CD2 cross-linking was not impaired (data not shown). Therefore, we conclude that the CD2 transmembrane domain is not required for its relocation to lipid rafts.
CD2 Translocation to DIG Is Independent of the Membraneproximal Domain Glycans-In non-lymphoid cells, it has been reported that N-glycans on membrane proteins can mediate raft association through binding to resident lectin-like molecules within lipid rafts (66,67). To assess whether this might be the case for CD2 in lymphoid cells, we used three mutant CD2 molecules in which one or the other complex N-linked CD2 glycans (N117Q and N126Q) or both (N117Q/N126Q) were eliminated (41). Under these circumstances, recruitment of mutant CD2 molecules into lipid rafts was still observed after antibody cross-linking (Fig. 6D), implying that the membraneproximal glycans are not required to induce CD2 association with DIG.
CD58 Ligation Is Sufficient to Recruit CD2 to Lipid Rafts-Although the above experiments identified a role for a specific pair of anti-CD2 mAbs in inducing CD2 translocation to DIG, it remained to be determined whether a similar translocation might be induced by the physiologic human CD2 ligand, CD58. To address this question, we utilized CHO cells as a simplified version of antigen-presenting cells. A CHO cell transfected with the expression vector alone (CHOP) was used as a negative control, whereas CHO cells transfected with the glycosylphosphatidylinositol-linked form of human CD58 (CHO58) were used as the ligand-positive control. The CHO cells were mixed with J77 cells at a 1:5 cell/cell ratio, spun down, and resuspended in a small volume of warm medium to facilitate cell-cell conjugate formation. The cell-cell conjugates were formed at 37°C for 5 min, and then cells were quickly collected by spinning, followed by raft isolation as described above. As shown in Fig. 7A, there was a 2.3-fold increase in the amount of CD2 localized to rafts upon CD58 ligation in comparison with the negative control CHOP cells. These findings indicate that CD2 binding to CD58 is able to induce translocation of CD2 into lipid rafts. That comparable amounts of raft fractions were loaded from T cells contacting CHOP and CHO58 APCs was verified by reblotting the raft fraction with anti-LAT antibody (Fig. 7A).
We further confirmed the above result using a CHO cell transfectant bearing a single point mutation (K87A) of CD58 and defective in CD2 binding (68). To demonstrate that these observations were not restricted to J77 cells, we employed activated human T cells. As shown in Fig. 7B, compared with CHO58K87A cells, there was a 2.5-fold increase in CD2 recruitment to rafts by T cells when conjugated with CHO58 cells. By comparing the ratio of CD2 in the raft fraction versus CD2 in the soluble fraction, we can conclude that the extent of raft recruitment induced by CD58 is about an order of magnitude weaker compared with antibody-induced cross-linking (compare Figs. 2 and 7). This result is not surprising given that the affinity of the CD2-CD58 interaction is 100-fold less than that of the CD2-mAb interaction.

DISCUSSION
Inducible Nature of CD2-Raft Association-In this study, we show that human CD2 is not constitutively associated with membrane rafts, consistent with earlier reports (69). More importantly, we demonstrate, for the first time, that CD2 is inducibly recruited to lipid rafts after CD2 cross-linking by specific anti-CD2 mAbs or CD58-bearing APCs. Most membrane raft-associated proteins are constitutive residents because of glycosylphosphatidylinositol modification or through palmitoylation (7,10). Included among the latter are G proteins (16,17), the Src family protein-tyrosine kinases (p56 lck and p59 fyn ) (14,15), and CD4 and LAT (53). Each of these proteins is palmitoylated at membrane-proximal cysteine residue(s) (reviewed in Ref. 20). By contrast, the human CD2 protein lacks any cysteine residues in its cytoplasmic domain and hence cannot be palmitoylated. In this regard, it is particularly significant that there are two membrane-proximal cysteines in the FIG. 5. CD2-lipid raft association is independent of CD3-raft recruitment and TCR signaling. A, 1 ϫ 10 8 activated human T cells were stimulated with either a combination of anti-T11 2 ϩ anti-T11 3 mAbs (CD2XL) or anti-CD3⑀ mAb 2Ad2 (CD3XL) at 37°C for 10 min. Total cell lysates were prepared using 0.2% Triton X-100 (TX-100) and subjected to sucrose gradient separation of lipid rafts. Only raft fractions were run on SDS-polyacrylamide gel and Western-blotted (WB) for CD3⑀. Subsequently, the membranes were stripped and reblotted (RB) with anti-CD2 or anti-LAT polyclonal antibodies. B, 8 ϫ 10 7 J77 cells were pretreated with PP2 at 37°C for 10 min or left untreated (Ϫ). Anti-T11 2 ϩ anti-T11 3 antibodies were used to stimulate both cells at 37°C for 15 min. Total cell lysates were then fractionated on sucrose gradients, and raft fractions were subjected to Western blotting with anti-CD2 mAbs and reblotting with anti-LAT antibodies. murine and rat CD2 tails that are absent from horse and human homologs (Fig. 6A) (47,48). This species difference is noteworthy since it was previously reported that substantial amounts of mouse CD2 are raft-associated in resting murine lymphocytes (28). The evolution of CD2 membrane partitioning to a raft-inducible rather than constitutive association mechanism may offer a partial explanation for any functional distinc-tions in the role of CD2 in man and mouse, respectively. In addition, mouse CD48 is a weaker ligand for mouse CD2 than human CD58 is for human CD2 (by at least an order of magnitude) and has a more restricted distribution (35,70). Perhaps the constitutive membrane partitioning of mouse CD2 compensates for this lower affinity ligand interaction.
Distinctions between CD2-and TCR-inducible Membrane Raft Association-Inducible association with lipid rafts upon ligand binding or antibody cross-linking has been demonstrated for several other surface molecules. The first studies involved Fc⑀ receptor I (26,27), a heterotetrameric complex expressed on basophils and mast cells. Ligand-engaged IgE receptors, but not unaggregated receptors, partition into lipid rafts (26). This interaction was found to be very sensitive to detergent extraction, so only in 0.025% Triton X-100 could this association be observed (26). A similar behavior was identified for the TCR in T lymphocytes (17,22). Upon TCR ligation or antibody-mediated cross-linking, an increased amount of TCR was found in lipid raft fractions (22). Moreover, phosphorylated -chain as well as phosphorylated ZAP70 and phospholipase C␥1 were also observed in rafts (17), showing that the inducible TCR-raft translocation correlates with co-association of known signaling molecules. As with Fc⑀ receptor I, the association was sensitive to the specific detergent and its concentration (8,22). Thus, weaker detergents such as Brij 58 preserved TCR-raft interaction, whereas 1% Triton X-100 or Nonidet P-40 did not (22). Similar to the TCR, the B cell receptor was also shown to translocate into lipid rafts upon antibody cross-linking (24,25).
The CD2-raft association observed here is different from the above immunoreceptor complexes based on its stability under various extraction conditions. In particular, the CD2-raft inter-FIG. 6. Domains of CD2 responsible for lipid raft recruitment. 155.16 T cells transfected with wild-type (W33) (41) or mutant CD2 molecules were either cross-linked by a combination of anti-T11 2 ϩ anti-T11 3 ascites (ϩ) or left untreated (Ϫ). Lipid raft fractions were separated on sucrose gradients, and equal amounts were loaded on SDS-polyacrylamide gel and then subjected to Western blot (WB) analysis with anti-CD2 mAbs. The same membrane was stripped and reblotted (RB) with anti-LAT antibodies to ensure equivalent loading of each raft fraction. A, sequence alignment of the CD2 transmembrane and cytoplasmic membrane-proximal regions. The thick line represents the transmembrane domain (TM). The two arrows indicate the unique membrane-proximal cysteine residues in rodent CD2. The broken lines represent the constructions used in B. Residues that are identical in at least three of the homologs are boxed. B, analysis of wild-type CD2 (W33), CD2 truncated after residue 25 in its tail (⌬25), and CD2 truncated after residue 7 in its tail (⌬7). The results shown are representative of four experiments. C, analysis of CD2 mutants in which the transmembrane segment was swapped with the transferrin receptor transmembrane region with (TfR 25 ) or without (TfR TM ) 25 amino acids of the CD2 tail appended. For quantitation purposes, the relative intensity of unstimulated cells is set at 1.0. D, analysis of glycan-deficient CD2 mutants. CD2 molecules with a single N117Q or N126Q mutation or a double N117Q/N126Q glycan mutation are shown. The experiment was repeated three times with similar results.
FIG. 7. CD2 interaction with CD58 during cell-cell conjugation is sufficient to recruit CD2 to lipid rafts. A, 1 ϫ 10 8 J77 cells were mixed with CHOP or CHO58 cells (5:1) and incubated at 37°C for 5 min. The cell lysates were made and subjected to sucrose gradient separation. Lipid raft fractions (fraction 3 (Frac3)) and soluble fractions (fraction 10 (Frac10)); one-fifth the amount of fraction 3) were resolved by SDS-PAGE and Western-blotted (WB) with anti-CD2 M32B polyclonal antisera. The membrane was stripped and reblotted (RB) with anti-LAT antibodies to ensure the equal loading of raft fractions. Numbers under the upper panel represent the relative intensity of each band. Shown is a blot representative of four individual experiments. B, 1 ϫ 10 8 activated human T cells were treated identically as described for A, except that the CHO transfectant bearing the mutant CD58 molecule K87A known to disrupt CD2 binding (CHO58K87A) (68) was used as a negative control. This experiment was repeated three times. action is more robust than the TCR, B cell receptor, or Fc⑀ receptor I, being stable even using 1% Triton X-100 extraction conditions. This distinction in raft association strength implies that the mechanisms involved in their respective association could be different. Interestingly, another monomeric lymphocyte surface protein, CD20, which is expressed on B cells, shows very similar properties to CD2 in terms of inducible association following antibody cross-linking and detergent stability (71). Notwithstanding, CD20 and CD2 are structurally unrelated.
The recruitment of CD2 to lipid rafts is independent of the cytoplasmic tail segments (amino acids 8 -117) known to be involved in CD2 signaling function. In fact, this very segment has been shown to bind to important signaling and/or adaptor molecules, including the Src family kinases p59 fyn and p56 lck (63,64); phosphatidylinositol 3-kinase (65); and the CD2 tailbinding molecules CD2BP1 (42), CD2BP2 (72), and CD2AP (73). Such a result is distinct from the TCR behavior given that 1) recruitment of the TCR to lipid rafts is Src family kinasedependent; 2) TCR-induced raft redistribution can be inhibited by Src-type tyrosine kinase inhibitors such as PP1 (22); and 3) these events are also blocked by wortmannin, an inhibitor of phosphatidylinositol 3-kinase. Triggering through the TCR by anti-CD3⑀ mAb cross-linking activates multiple proximal TCR signaling molecules such as ZAP70 and p56 lck as evident by their phosphorylation. Activated p56 lck and LAT, a substrate of ZAP70, are both resident lipid raft proteins. p56 lck and ZAP70 bind to immune receptor tyrosine-based activation motifs in the cytoplasmic tail of the CD3␥, CD3␦, CD3⑀, and CD3 subunits of the TCR. These interactions can recruit the TCR to rafts. In support of the notion that signal transduction is responsible for TCR relocation to rafts, it was reported that the phospho-TCR -chain can interact with the raft resident p59 fyn SH2 domain, suggesting that localization of Src family kinases is also important (15).
These data indicate that TCR interactions with rafts are downstream of prerequisite tyrosine phosphorylation events. In contrast, pretreatment of activated human T cells with tyrosine kinase inhibitors does not affect CD2-raft redistribution (Fig. 5B). Moreover, all known CD2 tail associations with signaling pathways (43,44) are eliminated by the CD2⌬7 truncation, and yet this molecule redistributes to rafts like wildtype CD2. Stimulating cells with anti-CD3⑀ mAb cross-linking fails to recruit CD2 to rafts (Fig. 5A), further indicating that the mechanisms behind TCR-and CD2-raft recruitment are different.
Inducibility of the CD2R epitope in the CD2 ectodomain upon specific anti-CD2 mAb cross-linking as well as by the CD58 ligand (41,49), both stimulators of CD2 translocation to lipid rafts, suggests that ectodomain conformational change upon ligation is responsible, at least in part, for the CD2-DIG translocation. We cannot exclude a role for the membrane-proximal seven CD2 cytoplasmic residues since deletion of these residues reduces surface CD2 expression to a level too low to accurately measure translocation quantitatively. Likewise, although the sialic acid-containing membrane-proximal domain glycans at Asn 117 and Asn 126 were excluded by mutagenesis as important for raft translocation, the high mannose glycan in domain 1 was not eliminated given that removal of the latter would disrupt the native fold of CD2 domain 1 (41). Interaction of a resident DIG lectin-like protein with this domain 1 glycan might theoretically offer a basis for translocation (66,67), but as this simple glycan is exposed on resting T cells, such a docking could not explain the inducible CD2 translocation behavior observed.
Lipid Raft Association Versus Mitogenicity-The mAb requirement for CD2 translocation to lipid rafts identically parallels the requirement for antibody-induced mitogenicity. Hence, to induce CD2 translocation or T cell proliferation, binding of the anti-T11 2 plus anti-T11 3 mAb pair is required. This observation suggests that the two processes are linked, particularly given the requirement for intact lipid rafts to mediate CD2-triggered activation function (Fig. 4) and the known raft association of substrate and enzyme components involved in Ca 2ϩ flux, IL-2 gene activation, and T cell proliferation. On the other hand, we failed to observe mitogenicity upon exposure of T cells to CD58-expressing antigen-presenting cells, although such cell-cell conjugates trigger CD2 translocation to DIG (Fig. 7). Such a disparity between anti-CD2 mAb-and CD58-triggered translocation of CD2 may be a consequence of differences both in the duration of the interaction and in the extent of raft association. Rapid k on and k off rates of CD58 (35) compared with anti-CD2 mAbs imply that the CD2-ligated state is short-lived in the case of the natural ligand. In fact, the dynamic binding of the CD2-CD58 co-receptor pair is important; this characteristic facilitates adhesion (micromolar K d ), but does not impede diffusion of TCRs into the abutting T cell-APC membranes in a given cell-cell conjugate. In turn, and possibly as a consequence of the binding kinetics, the amount of CD2 associated with lipid rafts following CD58 ligation of CD2 represents only 10% of the CD2 molecules induced to translocate to DIG by anti-CD2 mAbs (compare Figs. 2 and 7). The CD2 association with rafts is transient, being substantially reduced by 30 min following ligation with CHO58 cells (data not shown). In this regard, prior studies of T cell proliferation have shown that a potent stimulating signal needs to be sustained for hours to maintain gene transcription and to promote T cell cycle progression, with the cumulative amount of signal received determining the fate of T cell activation (74). Such transient interaction between CD2 and rafts upon CD58 ligation noted herein may provide initial signals for activation, but not, by itself, achieve the threshold required to trigger T cell proliferation. Nevertheless, CD58 ligation is known to activate itk (75), to stimulate IL-12 responsiveness (39), and to reverse anergy (40).
The Specific Nature of CD2 Cross-linking Stimuli Required for DIG Translocation and Mitogenicity-As described in Fig.  3, the anti-CD2 mAb requirement for CD2 translocation to DIG and mitogenicity is highly specific. Cross-linking of CD2 molecules per se is not sufficient either for translocation or mitogenicity. Hence, anti-T11 2 or anti-T11 3 mAb alone or with secondary anti-mouse Ig cross-linking, cross-linking of a biotinylated anti-T11 2 mAb with avidin, and cross-linking by rabbit anti-human CD2 heteroantisera were all insufficient to induce raft translocation or T cell proliferation despite extensive capping of CD2 observed microscopically in the latter two cases (Fig. 3, Ref. 41, and data not shown). These findings suggest that cross-linking of CD2 must be achieved to generate a molecular lattice with CD2 molecules in a specified conformation and/or distance from other CD2 molecules (41). The importance of one of the two mAbs in a mitogenic anti-CD2 mAb pair mapping to Phe 54 of CD2 domain 1 is consistent with this notion (76). Such a precedent for defined cross-linking has been observed in the erythropoietin (EPO) receptor system (77,78). The EPO receptor is a dimer even in the absence of its ligand EPO (77). Upon EPO binding, however, the receptor undergoes a conformational change resulting in rearrangement of the signaling tail such that JAK2 can be autophosphorylated and transduce the EPO signal (78). The EPO-mimetic peptides that function as antagonists permit the receptor to dimerize, but in a different orientation from that of EPO. This orienta-tional difference precludes productive signaling through the receptor; and hence, EPO receptor-based signaling is inhibited. Assuming that there is a specific set of CD2 conformational and/or distance requirements for CD2 lattice formation resulting in signaling upon CD2 cross-linking, then the combination of anti-T11 2 plus anti-T11 3 mAbs, but not random forms of antibody-mediated cross-linking, will result in raft association or T cell mitogenicity. Given that there are other combinations of anti-CD2R plus anti-CD2 mAbs that can trigger a mitogenic response (antibodies 9.1 and 9.6, respectively) (51) as well as CD2 translocation to DIG (data not shown), these reagents may induce a cross-linking lattice equivalent to that stimulated by anti-T11 2 plus anti-T11 3 mAbs and distinct from other antibody pairs.
CD2-based Membrane Reorganization and Raft Coupling Prior to pMHC Complex Recognition: A Mechanism to Facilitate TCR Signaling-The movement of TCRs to DIG requires tyrosine phosphorylation by kinases that are resident within the lipid rafts (22). In turn, the integrity of the lipid rafts is essential for TCR signal transduction (17,22). Thus, the question remains as to how a TCR initially becomes linked to DIG. That CD2 molecules redistribute to the membrane cell-cell conjugation surface (32,41) and, upon CD58 ligation, couple to lipid rafts in the absence of phosphorylation events offers one explanation. CD2-based adhesion is pMHC-independent and occurs prior to and without any requirement for antigen-triggered immune responses (6,32,33). The CD2-enriched, DIGcoupled cell surface membrane patch is ideally designed to facilitate TCR contact with the pMHC at the intercellular contact zone by promoting dynamic adhesion between abutting membranes with a spacing suitable for TCR-pMHC interaction (Ref. 34 and references therein). Recent microscopic analysis of murine T cell recognition processes has shown that cell surface molecules are redistributed into supramolecular activation clusters (79) with a central accumulation of TCR-pMHC complexes surrounded by CD2-CD48 complexes and a more peripheral ring of LFA-1/ICAM-1, also referred to as an "immunologic synapse" (80,81). Given the aforementioned species differences, it remains to be determined whether a similar supramolecular activation cluster morphology is observed during human T cell-APC interaction. Note that this supramolecular organization occurs as the consequence of TCR triggering rather than being the basis of such activation (79,80).
Because of the inducible human CD2 coupling to lipid rafts, p56 lck and p59 fyn kinases are pre-configured to participate in the initial TCR activation process. These enzymes as well as their adaptors (endogenous or associated) may facilitate subsequent recruitment of TCRs to DIGs in the T cell-APC contact zone, hence favoring further activation events. The ability of CD2 cross-linking to synergistically facilitate triggering through the TCR (51) may be reflective of this "shared" raft coupling phenomenon involving distinct translocation mechanisms. Further studies in this regard are clearly warranted.