Dependence of T Cell Antigen Recognition on the Dimensions of an Accessory Receptor–Ligand Complex

The 2B4 T cell hybridoma and its derivatives 2B4.CD2 (2B4.mCD2-1.6; transfected with full-length mouse CD2) and 2B4.CD2trunc (2B4.mCD2.M3-2.1; transfected with cytoplasmic tail–deleted mouse CD2) have been described previously 26,32. They were cultured in RPMI-10 medium (RPMI 1640 [GIBCO BRL], 10% FCS [Sigma Chemical Co.], 100 U/ml penicillin, 100 μg/ml streptomycin [GIBCO BRL], 4 mM glutamine [GIBCO BRL], and 5 × 10⁻⁵ M 2-ME [Sigma Chemical Co.]). G418 (0.4 mg/ml; GIBCO BRL) was added to medium used for 2B4.CD2 and 2B4.CD2trunc cells in order to maintain expression of CD2. The Chinese hamster ovary (CHO)¹ cell line N1.A4 expressing mouse I-E^k has been described elsewhere 33,34. It was cultured in RPMI-10 medium containing 0.4 mg/ml G418 to maintain I-E^k expression.

The following mAbs were used: RM2-1, rat anti–mouse CD2, IgG2a 35; OX11, rat anti–rat κ chain, IgG2a 36,37; OX78, rat anti–mouse CD48, IgG2a 21; IE-D6, mouse anti–mouse I-E^k, IgG2a (Serotec); OX12, mouse anti–rat κ chain, IgG2a 36; 145-2C11, hamster anti–mouse-CD3 38; A2B4-2, mouse anti-2B4 TCR-α, IgG2a 39; and 30-H12, rat anti–mouse Thy1.2, IgG2b (PharMingen).

Four different mouse CD48 constructs were made: wild-type CD48 with a silent BssHII site NH₂-terminal to the glycosyl phosphatidylinositol (GPI) signal sequence for the insertion of additional domains; elongated molecules with two-domain human CD2 (CD48-CD2) or three-domain mouse CD22 (CD48-CD22) inserts; and a shortened CD48 construct (CD48d1) in which the membrane-proximal domain (d2) had been deleted (see Fig. 1 A). The wild-type CD48 construct was made by PCR in two steps using as template the vector pCD.MBCM-1, which contained the sequence for full-length mouse CD48 23. In the first PCR, the 5′ primer 1130 (tagtagtctagaccccatccgctcaagcaggccaccATGTGCTTCATAAAACAGGGATGGTG) encoded an XbaI site (restriction sites are underlined in all primers), the rat CD4 5′-untranslated region (in lower case), and the first 26 nucleotides of the CD48 sequence beginning with the start codon (upper case; compare with Wong et al. [23]). The 3′ primer GTTGTGACCACTAGCCAAGTTGCAGTCCAACATACTCCAGAAGAgCgcGC-TAGATCACAAGGTAG encoded for the 3′ end of the second domain of CD48, a silent restriction site (BssHII), and part of the GPI anchor signal sequence. The resulting product was elongated in a second PCR using the same 5′ primer (1130) with the 3′ primer ctactagGaTCcTCAGGTTAACAaGATCCTGTGAATGATGA-GTGTTGTGACCACTAGCCAAGTTGC. This primer overlaps the 3′ primer used in the first PCR and encodes for the remainder of the CD48 sequence followed by a BamHI site. In addition, a silent mutation (lower case “a”) was introduced to eliminate an internal BamHI site. The PCR product was digested with XbaI and BamHI and ligated into vector pEF-BOS-XB 40.

To obtain a DNA fragment encoding domains 1 and 2 of human CD2 which could be inserted into the silent BssHII site in the wild-type CD48 construct, a plasmid encoding the extracellular portion of human CD2 41 was used as template for a PCR with the 5′ primer tagtaggcgcgcGAGATTACGAATGCCTTGGAAACC, which encodes a BssHII site and the CD2 signal sequence 42, and the 3′ primer ctactagcgcgCTTTCTCTGGACAGCTGACAGG, which encodes the final 21 nucleotides of domain 2 42 flanked by a BssHII site.

The mouse CD22 domain 1–3 insert was generated by PCR using as template a plasmid encoding a CD22(d1–3) Fc chimera with the mutation, R130A 43. The primers introduced flanking BssHII sites: 5′ primer, tagtaggcgcgcGATTGGACCGTTGACCATC; 3′ primer, ctactagcgcgcGTGCACCGTGAGTTCCAC. The CD2 and CD22 fragments were cloned into the BssHII site in the CD48 wild-type pEFBOS-XB plasmid. In the final construct, the junctions at the CD2 insert were TGTGATCTAGCGCGCgagattacg and ccagagaaaGCGCGCTCTTCT (CD48 in upper case, CD2 in lower case, and BssHII site underlined). Similarly, the junctions at the CD22 insert were GATCTAGCGCGCgattggacc and gtgcacGCGCGCTCTTCT.

The CD48d1 construct was obtained by PCR with the CD48 wild-type construct as template, using 1130 as 5′ primer and ctactagcgcgCAGGATCAAATACTTCCAG as 3′ primer. The latter encoded for the last nucleotides of the membrane-distal domain of CD48 and added a BssHII site. The PCR product was digested with XbaI and BssHII and inserted into the CD48 wild-type pEFBOS-XB plasmid in place of the fragment encoding CD48 d1 and 2. The boundary between domain 1 and the mCD48 GPI anchoring signal in the resulting construct was TTTGATCCTGCGCGCTCTTCT. All constructs were sequenced using an Applied Biosystems DNA sequencing system (model 373A).

For expression, the CD48 constructs were subcloned into the pEE14 expression vector using its XbaI and BclI sites 44,45. N1.A4 CHO cells were transfected and selected using the glutamine synthetase selection system 44,45. Stable lines were selected at 15–40 μM l-methyl sulfoximine (Sigma Chemical Co.) and recloned. Clones which gave comparable staining with CD48 mAb in cytometry analyses were chosen for functional assays. They were maintained in culture in CB2-10 medium (CB-2 [GIBCO BRL], 10% dialyzed FCS [First Link], 100 U/ml penicillin and 100 μg/ml streptomycin [GIBCO BRL], 0.4 mg/ml G418 [Sigma Chemical Co.], and 15–40 μM l-methyl sulfoximine). Revertants of the various clones were obtained by culturing cells for 8–12 wk in the absence of l-methyl sulfoximine and sorting for CD48⁻ cells on a FACSort™ cell sorter (Becton Dickinson).

Triton X-100 lysates of 10⁵ CHO cells per lane were separated by SDS-PAGE under reducing conditions and transferred to nitrocellulose (Hybond-C-super; Amersham Pharmacia Biotech). Membranes were probed with 1:7 dilutions of mAb OX78 and OX11 spent tissue culture supernatant, followed by a 1:2,000 dilution of peroxidase-conjugated rabbit anti–rat Ig serum (DAKO) and developed using the ECL+Plus system (Amersham Pharmacia Biotech). Purified soluble mouse CD48 was used as a positive control 46.

Standard procedures were used to stain cells with purified mAbs (15 μg/ml) or hybridoma culture supernatants. Binding of CD2 chimeric protein to CHO cells was assessed using a previously described method developed for detection of low-affinity interactions 47. In brief, 12.5 μl avidin-coated fluorescent beads (YFP-0552-5; Spherotech) per sample were rotated with 1.25 μg biotinylated anti–human Fcγ mAb SB2H2 47 and 40 μl of PBS/0.2% BSA for 45 min at 4°C, washed in PBS/0.02% BSA, and rotated with 3 μg/ml of human CD2 Fc 21 or human CTLA-4 Fc 48 for 2 h at 4°C in 1 ml of RPMI-10 medium. Beads were washed and resuspended in 12.5 μl PBS/BSA per sample, sonicated for 1 min, and added to 40 μl PBS/BSA containing 10⁵ CHO cells in a well of a flat-bottomed microtiter plate. Plates were centrifuged for 20 min at 1,000 rpm at 4°C and incubated at 0°C for 40 min before the cells were resuspended in 0.5 ml PBS/BSA and analyzed in a FACScan^® flow cytometer (Becton Dickinson).

Binding of CD48 to T cells was measured in a similar assay but using a chimeric protein consisting of mouse CD48 fused to rat CD4 domains 3 plus 4, followed by a short peptide tag which can be enzymatically biotinylated 40. To create this construct, an XhoI site was inserted 5′ of the His tag in a mouse sCD48his encoding plasmid 46 to enable the sCD48 portion to be excised with XbaI and XhoI and cloned into the XbaI/SalI sites of a vector encoding rat CD4d3+4 and a consensus peptide sequence recognized by the Escherichia coli biotin holoenzyme synthetase, BirA 40. The CD48-CD4 boundary in the resulting chimera was CTAGCCcgctcgacATCCATC. A rat CD5-CD4d3+4–biotin construct was made in an analogous way. Both constructs were expressed and biotinylated as described 40. The protein-coated beads were prepared by rotating 12.5 μl of avidin-coated fluorescent beads with 1 μg biotinylated protein in 350 μl PBS/BSA for 1 h at 4°C, followed by washing, sonication, and cell labeling as described above.

In each well of a V-bottomed microtiter plate (Bibby Sterilin Ltd.), 5 × 10⁴ 2B4 cells were mixed with 5 × 10⁴ irradiated (3,000 rad) CHO cell APCs and different concentrations of HPLC-purified moth cytochrome c (MCC) peptide 88–103 (ANERADLIAYLKQATK; made by Dr. N.P. Groome, Oxford Brookes University, Oxford, UK) and cultured in RPMI-10 medium for 18 h. In some experiments, CHO cells were preincubated with 10 μg/ml of mouse CD48 mAb OX78. This concentration of mAb was maintained during the 2B4/APC/peptide coculture. Subsequently, supernatants were harvested, and 2 × 10⁴ per well of IL-2–responsive CTLL-2 cells (American Type Culture Collection) were cultured with serial dilutions of the supernatants in RPMI-10 medium for 18 h. Purified recombinant mouse IL-2 was used as the standard (rec. mIL-2; Boehringer Mannheim). Wells were pulsed with 0.5 μCi/well of [³H]thymidine (Amersham Pharmacia Biotech) and harvested after an additional 7 h. Linear regression (Cricket Graph; Computer Associates International) was used to generate a linear fit, which was used to calculate the unknown IL-2 concentrations. Probably as a consequence of clonal variation, the different 2B4 cells exhibited somewhat different levels of antigen-induced IL-2 production even in the absence of CD2 ligation. This effect had been noted previously in a study with T cells transfected with different forms of human CD2 49. To allow comparison of the effects of CD48 expression on APCs between different 2B4 clones, the results in some experiments were normalized, letting 100% equal maximal peptide-stimulated IL-2 secretion in the presence of CD48⁻ CHO cells.

For phorbol ester/calcium ionophore stimulation, peptide was replaced with 50 ng/ml PMA (Sigma Chemical Co.) and 500 ng/ml ionomycin (Sigma Chemical Co.). For CD3 mAb stimulation, flat-bottomed microtiter plates (Falcon) were incubated with serum-free tissue culture supernatant containing CD3 mAb 145-2C11 for 2 h and washed with RPMI-10 before cells were added and cultured as described above.

10⁵ CHO cell APCs were incubated with MCC peptide 88–103 for 2 h in round-bottomed microtiter plates before the addition of 5 × 10⁴ 2B4.CD2 cells. After coculture for 3 h at 37°C, the cells were resuspended in PBS containing 0.5 mM EDTA and stained with anti-Thy1.2 and anti-2B4 TCR mAbs, followed by PE-conjugated goat anti–rat IgG (Serotec) and FITC-conjugated goat anti–mouse IgG2a (Southern Biotechnology Associates). TCR expression on the Thy1.2⁺ cells (i.e., excluding the APCs) was measured using a FACScan^® flow cytometer.

We examined the effect of varying the dimensions of the CD2/ligand complex using an in vitro T cell antigen recognition assay. The responder cells were the 2B4 murine MHC class II–restricted T cell hybridoma 26,32,50. The 2B4 TCR recognizes a peptide fragment (88–103) of MCC (MCC 88–103) bound to I-E^k 51. As APCs, we used CHO cells expressing I-E^k 33. These I-E^k+ CHO APCs were transfected with wild-type, elongated, or shortened forms of the mouse CD2 ligand, CD48 (Fig. 1 A). The elongated forms of CD48 had either two (CD48-CD2) or three (CD48-CD22) Ig domains inserted as spacers between the membrane-proximal Ig domain of CD48 and its GPI signal sequence (Fig. 1 A). The two-domain insert consisted of the ectodomain of human CD2, whereas the three-domain insert comprised the three membrane-distal Ig domains of mouse CD22. Both these inserts are functionally inert: human CD2 does not bind to mouse CD48 46,52, and the CD22 insert has a point mutation (R130A) that abolishes binding to its sialoglycoconjugate ligands 43. The short form of CD48 (CD48d1) was produced by deleting the membrane-proximal Ig domain (Fig. 1 A).

A Western blot of transfected I-E^k+ CHO cells using an anti-CD48 mAb showed that the various CD48 constructs migrated on SDS-PAGE at positions consistent with their predicted molecular mass (Fig. 1 B). Transfected I-E^k+ CHO clones that expressed equivalent levels of CD48 and H2-E^k were selected for further analysis (Fig. 2 A). The exception was CD48d1 I-E^k+ CHO cells, which expressed CD48 at 50% of the level of CD48 I-E^k+ CHO cells (Fig. 2 B). Beads coated with mouse CD2 Fc 21 bound at least as well to I-E^k+ CHO cells expressing elongated forms of CD48 as to cells expressing wild-type CD48, indicating that the CD48 remained functional (Fig. 2 A). In contrast, binding of these beads to CD48d1 I-E^k+ CHO cells was reduced by ∼90% (Fig. 2 B).

Although the interaction of human CD2 with its ligand, CD58, has been shown to enhance T cell antigen recognition, such an effect has not been formally demonstrated for the equivalent murine interaction between CD2 and CD48. This is an important point because CD2-deficient mice have surprisingly mild alterations in T cell function 27,29, and the mouse and rat CD2/CD48 interactions have an ∼6-fold lower solution affinity 46,53 and ∼40-fold lower membrane affinity 54 than the human CD2/CD58 interaction 41. Therefore, we examined the effect that CD48 expression on I-E^k+ CHO cells had on antigen recognition by 2B4 T cells. Because the original 2B4 hybridoma expressed only very low levels of CD2 (Fig. 2 C), we used cells that had been stably transfected with full-length mouse CD2 (2B4.CD2; reference 26). T cell antigen recognition by 2B4.CD2 cells, as measured by IL-2 secretion, was substantially enhanced by the expression of CD48 on I-E^k+ CHO APCs (Fig. 3 A). This enhancement was inhibited by incubating the APCs with a blocking CD48 mAb (Fig. 4 C), and was lost in CD48⁻ revertants of these APCs (Fig. 3 A). In contrast, the CD48 mAb had no effect on the 2B4 cell response to CD48⁻ I-E^k+ CHO cells (Fig. 4 C) nor on the CTLL line used in the IL-2 assay (data not shown), indicating that the inhibitory effect was not the result of binding to 2B4 or CTLL-2 cells (which also express CD48). The CD48 mAb consistently inhibited IL-2 secretion to below the levels seen with CD48⁻ I-E^k+ CHO cells (Fig. 4 C). This may be because mAb treatment killed up to 30% of CD48⁺ I-E^k+ CHO cells (data not shown).

Recently, the CD2-like 2B4 molecule has been shown to be a second ligand for CD48 40, raising the question as to whether the CD48 on I-E^k+ CHO cells was interacting with the 2B4 molecules on 2B4 cells. It was not possible to use a blocking CD2 mAb to test this because ligation of CD2 with mAb appears to transduce an inhibitory signal in 2B4 cells 26. However, two observations indicate that the enhancement by CD48 involves binding to CD2 alone. First, the enhancement was much reduced in 2B4 cells expressing very low levels of CD2 (Fig. 3 B), and second, the binding of CD48-coated beads to 2B4 cells is completely blocked by anti-CD2 mAbs (Fig. 3 D).

Although these results clearly demonstrate that an interaction between CD2 on T cells and CD48 on APCs can substantially enhance T cell antigen recognition, they do not reveal the mechanism of this effect. In addition to being an adhesion molecule, there is evidence that CD2 may transduce signals that activate T cells in the human 55 and the rat 30 or inhibit T cells in the mouse 26. Signaling through CD2 can be abolished by truncation of its cytoplasmic domain 12,25,26,30. We found that CD48 is also able to enhance antigen recognition by 2B4 cells that express a truncated form of mouse CD2 lacking all but 19 of the 116 amino acids normally present in the cytoplasmic domain (Fig. 3 C). Studies using truncated forms of mouse, rat, and human CD2 indicate that this truncation abolishes antibody-induced signaling 12,25,26,30. These results indicate that the enhancing effect of the CD2/CD48 interaction is at least partly a consequence of improved adhesion. Indeed, recent studies 17,56 suggest that the weaker enhancing effect observed with truncated CD2 may reflect a role for the cytoplasmic domain in enhancing adhesion rather than in transducing a signal.

In striking contrast to the effect of wild-type CD48, transfection of I-E^k+ CHO APCs with long forms of CD48 inhibited T cell antigen recognition (Fig. 4 A). It is notable that the slightly longer form of CD48, CD48-CD22, showed greater inhibition than CD48-CD2 (Fig. 4 A). CD48-CD22 had the larger inhibitory effect in seven out of eight independent antigen response assays (data not shown). The single exception is shown in Fig. 4 D, where CD48-CD2 and CD48-CD22 inhibit to a similar extent. 2B4 cells expressing very low levels of CD2 exhibited only very weak inhibition (Fig. 4 B), consistent with an effect mediated by CD2 ligation. Blocking anti-CD48 antibodies reversed the inhibitory effect of elongated CD48 (Fig. 4 C). This reversal was partial, possibly because the anti-CD48 antibody killed a proportion of the CD48-expressing I-E^k+ CHO cells (not shown). Finally, the inhibitory effect was completely lost in CD48⁻ revertants derived from the CD48-CD2– and CD48-CD22–transfected I-E^k+ CHO cells (Fig. 4 D).

These findings strongly suggest that the inhibitory effect is a consequence of CD2 ligation by the elongated forms of CD48, but do not indicate the mechanism of the inhibition. Since CD2 has been reported to transduce inhibitory signals when cross-linked with mAbs 26, it was important to exclude that inhibitory signaling through CD2. 2B4 cells expressing the truncated form of CD2 (2B4.CD2trunc), which does not transduce an inhibitory signal 26, was also inhibited by expression of elongated CD48 (Fig. 4 B). We also examined whether long forms of CD48 are inhibitory when T cells are activated by mechanisms that do not require peptide–MHC ligation. When 2B4.CD2trunc cells were activated with a phorbol ester plus ionomycin (Fig. 5 A) or an immobilized anti-CD3 mAb (Fig. 5 B), CHO cells expressing long forms of CD48 had no inhibitory effect. Similar results were obtained with 2B4 cells expressing full-length CD2 (not shown). Thus, the inhibitory effect of elongated CD48 was only evident when T cells were activated by TCR engagement of peptide–MHC. Taken together, these results indicate that the inhibitory effect of elongated CD48 is not a consequence of negative signaling through CD2.

We next examined the effect of shortening CD48. Because CD48 has only two Ig domains, it was only possible to shorten it by a single domain (∼3.5 nm). In contrast to the long forms of CD48, transfection of I-E^k+ CHO with a shortened form of CD48 (CD48d1; Fig. 1 A) enhanced peptide–MHC-induced activation of 2B4 (Fig. 6). Enhancement by CD48d1 was observed with 2B4 cells expressing both full-length (Fig. 6 A) and truncated (Fig. 6 B) CD2, but was somewhat reduced compared with wild-type CD48 (Fig. 6). Because the surface expression of CD48d1 was only half that of CD48 (Fig. 2 B), and CD2 binding to these cells was also substantially reduced (Fig. 2 B), it is not possible to deduce from these data how much less effective CD48d1 is than full-length CD48 at enhancing T cell antigen recognition. Nevertheless, it is clear that shortening of CD48 does not have the same inhibitory effect as elongation. This is an important control, since both shortening and elongation of CD48 would be expected to disrupt cis-interactions with other surface molecules. It follows that elongation of CD48 cannot be inhibiting by disrupting such cis-interactions.

To further investigate the mechanism by which long forms of CD48 inhibited T cell antigen recognition, we investigated the effect of elongated CD48 on TCR downmodulation. Recent studies suggest that TCR downmodulation is a consequence of TCR engagement 57,58, and is relatively independent of costimulation and downstream signaling events. Although wild-type CD48 enhanced TCR downmodulation induced by peptide-loaded I-E^k+ CHO cells, CD48-CD22 strongly inhibited downmodulation (Fig. 7). This suggests that wild-type CD48 enhances, whereas elongated CD48 inhibits, TCR engagement of peptide–MHC.

The key observation in this study is that whereas expression of wild-type CD48 on APCs enhances T cell antigen recognition, CD48 elongated by two or three Ig-like domains is strongly inhibitory. The most likely explanation for these results, strongly supported by measurements of TCR downmodulation (Fig. 7), is that binding of CD2 on T cells to elongated forms of CD48 on APCs positions the plasma membranes at a distance too far apart for efficient TCR engagement of peptide–MHC (Fig. 8, right). Conversely, the binding of CD2 to wild-type CD48 promotes TCR/peptide–MHC engagement through a favorable positioning of the plasma membranes (Fig. 8, left). Control experiments have excluded several other explanations. Thus, the inhibitory effect is not the result of negative signaling through CD2, nor is it likely to be a consequence of the disruption of cis-interactions between CD48 and other molecules. Although our data indicate that TCR/peptide–MHC engagement is inhibited, we cannot formally exclude the possibility that elongated forms of CD48 also inhibit T cell antigen recognition by forcing the segregation of CD2 from the TCR, thereby sequestering CD2-associated molecules required for TCR signaling.

The human CD2 fragment inserted into the CD48-CD2 chimera is ∼7 nm long 15. Since the length of the CD2/CD48 complex is ∼14 nm 8, the CD2/CD48-CD2 complex will span 14–21 nm. If the proposed explanation for the inhibitory effects of elongated CD48 is correct (Fig. 8), our data imply that optimal TCR engagement of peptide–MHC requires that the plasma membranes of the T cell and APC have to be <21 nm apart. It follows that molecules larger than this (e.g., CD45 and CD43), and molecular interactions that span greater distances (e.g., LFA-1/intercellular adhesion molecule [ICAM]-1), would need to be excluded from the immediate vicinity of the TCR when it engages peptide–MHC.

Our results thus provide support for a model proposed by Davis and van der Merwe (henceforth called the “kinetic-segregation” model) for T cell antigen recognition 13. This model postulates that molecules in the contact area between T cells and APC/target cells will segregate according to size, and that this segregation is a prerequisite for TCR triggering. The concept of segregation of T cell surface molecules according to size was first introduced by Springer 10. A similar model has been also been proposed recently by Shaw and Dustin 9. The kinetic-segregation model postulates that small molecules such as CD2, CD4, CD8, CD28, and CD154 will localize, together with their extracellular and cytoplasmic ligands, in zones of especially intimate contact (∼15 nm between membranes) called “close-contact zones,” and that larger molecules such as integrins, CD43, and CD45 would be excluded from these zones. It is proposed that TCR binding to peptide–MHC leads to TCR triggering by trapping the TCR/CD3 complex within the tyrosine kinase–enriched (and tyrosine phosphatase–deficient) close-contact zone for a period of time sufficient for tyrosine phosphorylation of TCR/CD3-associated signaling molecules. Several lines of evidence support this model. First, intermembrane distances of 12–15 nm have been measured in the contact area between cytotoxic T cells and target cells 59. Second, Kupfer and colleagues 60 have shown that TCR/CD3 and the integrin LFA-1 are segregated into different regions at the T cell/APC interface. Third, Dustin et al. 17 have shown that when Jurkat cells are allowed to adhere to lipid-anchored CD48 and ICAM-1 (CD54, an LFA-1 ligand) inserted into glass-supported planar lipid bilayers, the ICAM-1 and CD48 segregate from each other in the contact area. Fourth, Sperling et al. 61 have reported that CD43 is completely excluded from the binding interface between T cells and APCs. Interestingly, CD45 was not excluded from the binding interface, and was sometimes polarized towards it 61. Higher resolution imaging studies will be required to determine whether, like LFA-1 17,60, CD45 segregates from the TCR within the interface. Finally, our finding that elongation of CD2/CD48 complexes to the approximate size of an integrin molecule (∼21 nm) inhibits T cell antigen recognition provides experimental support for the notion that segregation is a prerequisite for TCR engagement of peptide–MHC.

Our findings raise the question as to why large cell–cell adhesion molecules such as the integrin LFA-1 do not inhibit T cell antigen recognition through the same mechanism as postulated here for the elongated forms of CD48. Indeed, LFA-1 interactions appear to enhance TCR engagement of peptide–MHC 62. One possible explanation, strongly supported by recent data 60, is that T cells have evolved specific mechanisms, perhaps cytoskeletally driven 63, to actively segregate LFA-1 from the TCR within the contact zone. This would enable LFA-1 to participate in T cell–APC conjugate formation without physically interfering with TCR binding to peptide–MHC. In contrast, there is evidence that CD2 may be actively clustered within the contact zone into the immediate vicinity of the TCR 17. This would be expected to prevent the segregation of elongated CD2/CD48 complexes from the site of TCR engagement.

Since CD48d1 is missing a single Ig domain (∼3.5 nm), it can reasonably be predicted that the CD2/CD48d1 complex will be ∼3.5 nm shorter than the CD2/CD48 complex (i.e., ∼10.5 nm). Our finding that CD48d1 can still enhance T cell antigen recognition implies that some reduction in the size of accessory molecules (and thus intermembrane distance) can be tolerated. Mechanisms by which the TCR/peptide–MHC interaction could adjust to closer membrane approximation include compression of long stalk regions of the TCR, which lack secondary structure, or tilting of the TCR/peptide–MHC complex relative to the plane of the membrane. Some limited flexibility in the size of accessory molecule and TCR/peptide–MHC complexes may be important because it is unlikely that all molecular interactions involved in T cell antigen recognition will span exactly the same distance.

The kinetic-segregation model postulates that small accessory molecules contribute to T cell antigen recognition in part by driving the formation of intimate close-contact zones. While the CD2/ligand interaction may contribute to the formation of close-contact zones, any accessory molecule interaction with approximately the same dimensions could fulfill the same function. This is supported by the observation that CD2-deficient mice have very mild abnormalities in T cell function 27,29. Candidate accessory molecules include CD28/ligand interactions and interactions involving other members of the CD2 family. The domain composition of CD28 and its ligands CD80 and CD86 suggests that these interactions will span 10–15 nm 7. Although CD28 can transduce activation signals through its cytoplasmic region 64, the recent observation that a form of CD28 lacking a cytoplasmic domain is very effective at enhancing T cell antigen recognition 65 provides strong support for an important adhesion role. Structural and mutagenesis studies suggest that interactions between other CD2 family members will span the same distance as the CD2/CD48 complex 13,46, making them suitable for membrane approximation in T cell antigen recognition. Potential candidates include SLAM 66, which is expressed on T cells, and 2B4, which is expressed on NK cells and a subpopulation of T cells 67,68, and has recently been shown to bind CD48 40.

In conclusion, our results demonstrate the importance of accessory molecule size, and suggest that close membrane approximation is required for T cell antigen recognition. This and other recent studies 17,60,61 support models of T cell antigen recognition 9,13 which propose that the cell surface molecules segregate according to size at the T cell/APC interface, and that this segregation is required for TCR engagement and triggering.

We thank Mark Davis (Howard Hughes Medical Institute, Stanford University, Stanford, CA) and Phil Reay (Nuffield Department of Clinical Medicine, Oxford University, Oxford, UK) for providing us with cell lines and antibodies. We thank Michael Puklavec, Vanessa Grimwade, Elizabeth Adams, Steve Cobbold, and Paul Fairchild (all from the Sir William Dunn School of Pathology, Oxford University) for providing antibodies and technical assistance. We are indebted to Hideo Yagita (Juntendo University School of Medicine, Tokyo, Japan) for providing CD2 Fc and antibodies, and Stefan Meuer (Institute for Immunology, Heidelberg, Germany) for antibodies. We also thank Neil Barclay, Don Mason, Michael Dustin, Simon Davis, Hans Reiser, and Christina Piperi for helpful advice.

This study was supported by the Medical Research Council and a Wellcome Trust Travelling Research Fellowship (no. 046136) to M.K. Wild.