CD3δ Establishes a Functional Link between the T Cell Receptor and CD8*

T cells expressing T cell receptor (TCR) complexes that lack CD3δ, either due to deletion of the CD3δgene, or by replacement of the connecting peptide of the TCRα chain, exhibit severely impaired positive selection and TCR-mediated activation of CD8 single-positive T cells. Because the same defects have been observed in mice expressing no CD8β or tailless CD8β, we examined whether CD3δ serves to couple TCR·CD3 with CD8. To this end we used T cell hybridomas and transgenic mice expressing the T1 TCR, which recognizes a photoreactive derivative of the PbCS 252–260 peptide in the context of H-2Kd. We report that, in thymocytes and hybridomas expressing the T1 TCR·CD3 complex, CD8αβ associates with the TCR. This association was not observed on T1 hybridomas expressing only CD8αα or a CD3δ− variant of the T1 TCR. CD3δ was selectively co-immunoprecipitated with anti-CD8 antibodies, indicating an avid association of CD8 with CD3δ. Because CD8αβ is a raft constituent, due to this association a fraction of TCR·CD3 is raft-associated. Cross-linking of these TCR-CD8 adducts results in extensive TCR aggregate formation and intracellular calcium mobilization. Thus, CD3δ couples TCR·CD3 with raft-associated CD8, which is required for effective activation and positive selection of CD8+ T cells.

The differentiation of CD4 CD8 double-positive (DP) 1 thymocytes into CD8 single-positive (SP) T cells requires appropriate signals from the TCR and the coreceptor CD8 (1,2). DP thymocytes and CD8 SP peripheral T cells express TCR␣␤ that are associated with three signal-transducing units, namely CD3␦⑀ and CD3␥⑀ heterodimers and a disulfide-linked chain homodimer (3)(4)(5). The CD3␥␦ and ⑀ subunits contain in their cytoplasmic tail a single immunoreceptor tyrosine-based activation motif, whereas the tail of the chain harbors three immunoreceptor tyrosine-based activation motifs. For surface expression of TCR␣␤, their association with CD3⑀, ␥, and but not with CD3␦ is required (6 -9). Accordingly, knockout of CD3⑀, ␥, and chain arrests T cell development at early stages (6 -11). By contrast, in CD3␦ knockout mice T cell development proceeds to the DP stage, but positive selection of CD8 (and CD4) SP T cells is severely compromised (9,12). During TCR␣␤ assembly the TCR␤ chain first associates with CD3⑀␥ and the TCR␣ chain with CD3⑀␦, and the resulting trimers then associate and the TCR␣␤ disulfide bond is formed (13). Although CD3␦ is physically associated with the pre-TCR complex, it is not required for pre-TCR signaling, which is essential for the transition of double-negative (DN) to DP thymocytes (10,14,15).
A conserved motif in the TCR␣ chain-connecting peptide domain, which connects the transmembrane and the Ig domains, referred to as ␣CPM, plays a crucial role in positive selection of CD8 and CD4 SP T cells (16 -18). The ␣CPM consists of seven highly conserved amino acids (FETDXNLN) and is present in TCR␣␤ but not in TCR␥␦ (16). In mice expressing TCR in which the ␣CPM is replaced by the corresponding sequence of the TCR␦ chain, positive selection of SP T cells is greatly impaired, whereas negative selection is normal (16 -18). These variant TCRs, referred to as ␣IV/␤III (16) TCR, exhibit impaired association with CD3␦, -chain phosphorylation, defective activation of p59 Fyn and extracellular signalregulated kinase, impaired phosphorylation and recruitment of ZAP-70, p56 lck (Lck), and LAT to lipid rafts (17)(18)(19). Very similar findings were obtained in CD3␦ knockout mice (12,15), arguing that the defects observed for the ␣CPM variant TCR are mainly accounted for by their impaired association with CD3␦.
On the other hand, the ␤ chain of CD8 plays a key role for positive selection of CD8 SP T cells. Although CD8 can be readily expressed as the CD8␣␣ homodimer, the number of CD8 SP T cells in CD8␤ knockout mice is greatly reduced (20,21). In a milder form, positive selection is also compromised in mice overexpressing tailless CD8␤, and activation is impaired in CD8 ϩ T cells expressing tailless CD8␤ (22,23). Given the similarity in impaired activation and positive selection of CD8 SP T cells in mice lacking CD8␤ or CD3␦, we examined here whether this is accounted for by the same mechanism, i.e. whether CD3␦ couples TCR with CD8␣␤. Several studies indicated that CD8 (and CD4) associates with TCR⅐CD3. For example, the proximity between CD8 (and CD4) and the TCR has been demonstrated on cells by using fluorescence resonance * This work was supported by grants from the Swiss National Foundation (Grant 31-61946.00) and the Sandoz Foundation. 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.
To investigate whether and how CD3␦ establishes a functional link between the TCR⅐CD3 and CD8, we used thymocytes and T cell hybridomas expressing the T1 TCR. This TCR recognizes the Plasmodium berghei circumsporozoite (PbCS) peptide 252-260 (SYIPSAEKI) containing photoreactive 4-azidobenzoic acid on Lys-259 (PbCS(ABA)) in the context of K d (30). Photoactivation of the ABA group results in crosslinking of the T1 TCR with K d -PbCS(ABA), which permits direct assessment of TCR-ligand interactions by TCR photoaffinity labeling (30 -32). Using TCR photoaffinity labeling, FRET, co-immunoprecipitation, and confocal microscopy, we find that CD8␣␤, but not CD8␣␣, associates with the TCR via CD3␦ and that this is required for efficient T cell activation.
FACS and FRET Analyses-T cell hybridomas and T1 thymocytes were washed and resuspended at 2 ϫ 10 6 cells/ml in serum-free Opti-MEM medium (Invitrogen, Merebeke, Belgium) containing 1% BSA and 0.02% NaN 3 . For FACS analysis aliquots of 25-l cell suspension were incubated in a 96-well plates with fluorescence-labeled antibodies (5 g/ml) for 20 min at 4°C. After two washes in the same medium, fluorescence associated with live cells was measured on a FACSCalibur (BD Biosciences, Erembodegen, Belgium). For FRET analysis, 50-l aliquots of cells (2 ϫ 10 6 cells/ml) were incubated in the same medium in 96-well plates with 5 g/ml Cy5-labeled anti-CD8␣ 53.6.72, anti-CD8␤ KT112, anti-CD4 GK 1.5, or anti-Thy-1 antibody; PE-labeled anti-CD3⑀ 17A2 mAb; or 50 nM of PE-coupled K d -PbCS(ABA) multimers. After 45 min of incubation at 4°C, the cells were washed and fixed with paraformaldehyde (3% w/v in PBS) for 10 min at room temperature, and cell-associated fluorescence was assessed on a FACSCalibur at 580 nm upon excitation at 488 nm (E1), at 670 nm after excitation at 630 nm (E2), and at 670 nm after excitation at 488 nm (E3). The transfer of fluorescence was calculated as FRET units as follows: The different fluorescence values (E) were measured on unlabeled cells (E none ), or cells labeled with PE (E PE ), Cy5 (E Cy5 ), or Cy5 and PE (E both ).
Isolation of Lipid Rafts-T1 TCR hybridomas or T1 TCR thymocytes (5 ϫ 10 7 cells) photoaffinity-labeled with K d -125 "IASA"-YIPSAEK-(ABA)I were lysed in 1 ml of MN buffer (25 mM MES, pH 6.5, 150 mM NaCl) containing 0.5% Brij96 (Sigma) and protein inhibitors (Roche Molecular Biochemicals) for 30 min on ice. The lysates were homogenized with a Dounce homogenizer (10 strokes) and fractionated on sucrose density gradients as described (29,38). The sucrose gradients were fractionated from the top in ten fractions of 500 l. Aliquots of the fractions were resolved on SDS-PAGE (10%, reducing) and either analyzed by phosphorimaging or Western blotted using antibodies specific for Thy-1 or CD45.
Confocal Microscopy-T1 TCR hybridomas or T1 thymocytes were washed with DMEM and incubated with 10 mM methyl-␤-D-cyclodextrin at room temperature, washed again with medium, and incubated in PBS containing 1% BSA for 30 min at room temperature or for 40 min at 4°C with PE-labeled K d -PbCS(ABA) multimers (50 nM). Following two washes with PBS, cells were analyzed using an LSM510 Zeiss confocal microscope. Median sections of cells were recorded. Alternatively, TCR hybridomas were incubated for 20 min at room temperature with anti-CD8␤ mAb KT112 and anti-TCR mAb H57 mAb in PBS containing 1% BSA, washed, and incubated with FITC-labeled anti-rat IgG antibody, and the distribution of CD8␣␤ was examined.
Modeling of CD3⑀␦-An homology model of the CD3⑀␦ complex was built based on the CD3⑀␥ 3D structure and CD3␦ sequence alignment (5), using the MODELLER program (39). The conformations of the connecting loops of the immunoglobulin fold were refined using an ab initio method based on simulated annealing (40).

CD8␣␤ and Wild Type TCR ␣CPM Are Required for Efficient
Intracellular Calcium Mobilization-To examine the role of CD8 and the TCR ␣CPM for T cell activation we first assessed intracellular calcium mobilization in T cell hybridomas expressing CD8 and wild type T1 TCR or T1 TCR in which the ␣CPM was replaced with the corresponding sequence of TCR␦ (T1 TCR ␣IV/␤III). As shown in Fig. 1A, hybridomas expressing the wt TCR and CD8␣␤ exhibited strong calcium mobilization upon incubation with P815 cells pulsed with 10 Ϫ6 to 10 Ϫ5 M IASA-YIPSAEK(ABA)I. This calcium flux was stable over the assayed period of 15 min. By contrast, no significant calcium mobilization was observed in the presence of the CD8␤ blocking antibody H35. Similarly, on hybridomas expressing the T1 TCR ␣IV/␤III, only scant calcium mobilization was observed, which was reduced to background levels in the presence of mAb H35 (Fig. 1B).
Using the same method we next examined thymocytes from T1 TCR transgenic mice. These mainly DP cells exhibited strong and stable intracellular calcium mobilization upon incubation with IASA-YIPSAEK(ABA)I-pulsed P815 cells (Fig.  1C). Maximal response was observed at 10 Ϫ7 M IASA-YIP-SAEK(ABA)I. The stronger calcium responses observed on T1 thymocytes, as compared with CD8 ϩ T1 T cell hybridomas, is explained, at least in part, by the higher surface expression of TCR and CD8 (see "Experimental Procedures"). In the presence of mAb H35, no marked calcium mobilization was observed, indicating that CD8 was required for this response (Fig. 1C). Taken together these results indicate that, for efficient calcium mobilization, CD8␣␤ and CD3␦ ϩ TCR are required, which is in accordance with previously studies showing that CD8 and the ␣CPM are crucial for efficient TCR signaling and positive selection of CD8 SP T cells (16 -22).
CD8 Increases MHC-peptide Binding on Cells Expressing Wild Type but Not ␣CPM Variant TCR-TCR photoaffinity labeling with soluble monomeric K d -125 "IASA"-YIPSAEK-(ABA)I complexes allows direct assessment of TCR-ligand binding and its dependence on CD8 (30 -32). Using this technique we compared TCR-ligand binding on T cell hybridomas expressing the wild type T1 TCR or the T1 TCR ␣IV/␤III. As shown in Fig. 2A, TCR photoaffinity labeling was reduced by over 6-fold in the presence of FabЈ fragments of anti-K d ␣3 mAb SF1-1.1.1, which block CD8 binding to K d (32). A slightly larger inhibition was observed on T1 thymocytes and on cloned T1 CTL (Fig. 2B and Ref. 32). The same reductions were observed upon blocking of CD8 with anti-CD8␤ mAb H35 (data not shown).
Remarkably, on hybridomas expressing T1 TCR ␣IV/␤III, K d -PbCS(ABA) binding was over 4-fold lower than on cells expressing wild type T1 TCR and blocking of CD8 binding to K d caused only a small reduction. The nonspecific labeling, as seen in the presence of anti-K d ␣1 mAb 20 -8-4S, which blocks binding of K d to TCR (32), was in the range of 3% on the hybridomas and below 1% on thymocytes. Upon blocking of CD8 binding to K d , TCR photoaffinity labeling was slightly lower on hybridomas expressing T1 TCR ␣IV/␤III, as compared with hybridomas expressing the wild type T1 TCR ( Fig. 2A). Because the TCR expression is slightly lower on the former as compared with the latter cells (see "Experimental Procedures"), it seems that both TCR bind K d -PbCS(ABA) with very comparable efficiency. This is consistent with the fact that both TCR have the same variable and constant domains and argues that the poor TCR photoaffinity labeling observed on T1 TCR ␣IV/␤III is accounted for by inefficient participation of CD8 in TCR ligand binding. Because CD8-mediated increase in TCR-ligand binding relies on association of CD8 with TCR⅐CD3 (29), this argues that the T1 TCR ␣IV/␤III associates poorly with CD8.
Wild Type but Not ␣CPM Variant TCR Associates with CD8 on Intact Cells-To validate this conclusion, we assessed the proximity of TCR⅐CD3 and CD8 by FRET. To this end we stained hybridomas expressing CD8 and wild type T1 TCR in the cold with Cy5-labeled anti-CD8␤ mAb KT112 and PElabeled K d -PbCS(ABA) multimers and measured the FRET value from PE to Cy5 by FACS. As shown in Fig. 3A, on CD8 ϩ hybridomas expressing the wild type T1 TCR, FRET was 2.2 units but only 0.4 unit on hybridomas expressing the T1 TCR ␣IV/␤III. The nonspecific signal, as recorded on the corresponding CD8 Ϫ hybridomas, was about 0.03 unit. When using PElabeled anti-CD3⑀ mAb 17A2, slightly less efficient FRET was observed (1.5 units, Fig. 3B). Remarkably, this FRET value was enhanced very little when soluble K d -PbCS(ABA) monomers were present in the incubation at saturating concentration, indicating that the proximity of CD8 and TCR⅐CD3 on these hybridomas was not induced by MHC-peptide, i.e. it was constitutive.
Strong FRET was observed on T1 thymocytes following staining with PE-labeled K d -PbCS(ABA) multimers and Cy5labeled anti-CD8␤ mAb KT112 (37 units, Fig. 3D). About onethird lower FRET (25 units) was recorded when using Cy5labeled anti-CD8␣ mAb 53.6.72. Because thymocytes express only CD8␣␤, but CD8-transfected T cell hybridomas always express high levels of CD8␣␣ (see "Experimental Procedures" and Refs. 29 and 33), the over 4-fold reduced FRET observed on CD8␣␣ ϩ T cell hybridomas indicates that CD8␣␤ couples with TCR⅐CD3 more extensively than does CD8␣␣ (Fig. 3C). This is consistent with the finding that CD8␣␤ is co-immunoprecipitated with the TCR more efficiently as compared with CD8␣␣ (29). In the presence of anti-CD8␤ mAb H35, this FRET was reduced to 6.2 units, indicating that mAb H35 impedes the association of TCR and CD8. Moreover, faint FRET (5 units) was observed when using Cy5-labeled anti-CD4 mAb GK1.5 as acceptor, which was about 2-fold above background, as recorded when using Cy5-labeled anti-Thy-1 antibody (Fig. 3D). This is consistent with the observation that CD4 also associates with TCR⅐CD3 (25,26). This FRET was not reduced in the presence of mAb H35, indicating that this antibody does not impair the PE K d -PbCS(ABA) multimer staining. The differences in FRET values observed in the different experiments in Fig. 3 are accounted for in part by variations in TCR and CD8 expression of the different cells (see "Experimental Procedures"). Together these results indicate that CD8 and TCR⅐CD3 are in close proximity in T1 thymocytes and in T1 T cell hybridomas, given they express CD3␦ ϩ TCR and CD8␣␤. Wild Type but Not ␣CPM Mutant T1 TCR Docks to Raftassociated CD8 -Because wild type T1 TCR, but not variant T1 TCR ␣IV/␤III, associates with CD8␣␤ (Figs. 2 and 3) and CD8␣␤ is a raft constituent (33, 36), we examined the raft association of these two TCR. To this end we TCR photoaffinity-labeled CD8␣␤ ϩ T cell hybridomas expressing wild type or T1 TCR ␣IV/␤III, lysed the cells in Brij96, and fractionated the detergent-soluble and -insoluble components on sucrose gradients. A significantly larger fraction of K d -125 "IASA"-YIPSAEK-(ABA)I-labeled T1 TCR was found in the detergent-insoluble light fractions 2-4 on hybridomas expressing the wild type T1 TCR as compared with hybridomas expressing the T1 TCR ␣IV/ ␤III (Fig. 4, A and B). To verify that our fractionation procedure was correct, we assessed the distribution of CD45 and Thy-1, which are known markers for the detergent-soluble and detergent-insoluble raft fractions, respectively (41,42). Indeed CD45 was found exclusively in the detergent-soluble dense gradient fractions (fractions 6 -10), whereas glycosylphosphatidylinositollinked Thy-1 was detected only in the detergent-insoluble light fractions (fractions 2-4) (Fig. 4, E and F).
In T1 thymocytes an appreciable fraction of photoaffinity-labeled TCR was found in the light fractions (Fig. 4C). This fraction was greatly diminished when TCR photoaffinity labeling was performed in the presence of the CD8␤-blocking mAb H35 (Fig. 4D). These findings indicate that a fraction of TCR is raft-associated due to association of TCR⅐CD3 with raft-resident CD8. This is consistent with the findings that ␣CPM variant TCR (18) or TCR from CD3␦ knockout mice (12) exhibit impaired raft-association and argues that CD3␦ couples TCR⅐CD3 with CD8␣␤ and hence with lipid rafts. Cross-linking of TCR⅐CD8 Adducts Results in the Formation of Large TCR⅐CD8 Aggregates-We next investigated what consequences TCR⅐CD3 cross-linking has on TCR aggregation. As assessed by confocal microscopy, incubation of T cell hybridomas expressing CD8␣␤ and wild type T1 TCR with PE-labeled K d -PbCS(ABA) multimers at room temperature resulted in extensive patch formation and internalization (Fig. 5A). The aggregate formation, but not the internalization, also took place when the incubation was performed in the cold (Fig. 5B), suggesting that it does not require cell activation. Strong TCR⅐CD8 aggregate formation was also seen on T1 thymocytes upon incubation with K d -PbCS(ABA) multimers and on CD8␣␤ ϩ , T1 TCR ϩ hybridomas after incubation with anti-TCR, and anti-CD8␤ antibodies (Fig. 5, A and C). By contrast, on hybridomas expressing T1 TCR ␣IV/␤III, aggregate formation and internalization were greatly reduced. The same diffuse, mainly surface staining, was also observed when CD8␣␤ ϩ , T1 TCR ϩ hybridomas, or T1 thymocytes were pretreated with methyl-␤-cyclodextrin, which destabilizes lipid rafts (Fig. 5A). Taken collectively, these findings demonstrate After UV irradiation the washed cells were lysed in Brij96 (0.5%) and the lysates were fractionated on sucrose gradients. Ten fractions were collected from the top, aliquots were immunoprecipitated with anti-TCR mAb H57, and the precipitates were resolved on SDS-PAGE (10%, reducing) and evaluated by phosphorimaging (A-D). Aliquots of the fractions from A were analyzed by SDS-PAGE (10%, reducing) and Western blotted with antibodies specific for CD45 (E) or Thy-1 (F). Detergent-insoluble components were mainly found in fractions 2-4, and detergent-soluble components were in fractions 6 -10. Each experiment was repeated at least once.
FIG. 5. CD3␦ and CD8␣␤ are required for cross-linking-induced formation of large TCR⅐CD8 aggregates. A, T cell hybridomas expressing CD8␣␤ and wild type (wt), T1 TCR (␣IV/␤III), or thymocytes from T1 TCR transgenic mice, pretreated or not with methyl-␤-D-cyclodextrin (MCD), were incubated at room temperature for 20 min with PE-labeled K d -PbCS(ABA) multimers and examined using confocal microscopy. B, alternatively, the staining of the hybridomas was performed for 40 min at 4°C. C, the hybridomas were incubated for 20 min at room temperature with anti-CD8␤ mAb KT112, anti-TCR mAb H57, and FITC-labeled rabbit anti-rat IgG antibody and analyzed by confocal microscopy. Representative images are shown from over ten cells analyzed per condition and from at least two different experiments. that co-cross-linking of TCR⅐CD3 and CD8 results in formation of large TCR⅐CD8 aggregates and that for this to occur CD8␣␤, lipid rafts, and CD3␦ ϩ TCR are required.
CD3␦ Associates with CD8 -We next examined whether CD3␦ co-immunoprecipitates with CD8. To this end, we lysed T1 thymocytes in Triton X-100 and analyzed their lysate as well as TCR and CD8 immunoprecipitates by SDS-PAGE and Western blotting. In the CD8 immunoprecipitate, CD3␦, but neither CD3⑀ nor chains, was highly enriched; especially when compared with the TCR immunoprecipitates, the preferential co-precipitation of CD3␦ was striking (Fig. 6A and data  not shown).
To obtain further information on the association of CD3␦ with CD3⑀ and CD8, we washed the CD8 immunoprecipitates twice with different buffers and assessed the amount of coprecipitated CD3␦ and CD3⑀ by Western blotting. The association of CD3␦ with CD8 was substantially reduced (to 36%) upon washing with ethylmaleinimide, which alkylates free cysteines (Fig. 6B). A smaller reduction (to 55%) was observed upon washing with EDTA-containing buffer, which chelates divalent cations. About 40% reduction was noted upon washing with n-octylglucoside, which disrupts association of transmembrane proteins (41). Strikingly, washing with 0.5 M NaCl had no marked effect, suggesting that the association of CD3␦ with CD8 is not ionic in nature. By contrast, washing with this buffer removed most of the CD3⑀ from the immunoprecipitates. The other buffers affected the co-precipitation of CD3⑀ in the same way as the co-precipitation of CD3␦.
In CD8␣␤ ϩ , T1 TCR ϩ T cell hybridomas, similar as with T1 thymocytes (Fig. 6A), anti-CD8␤ antibody efficiently co-immunoprecipitated CD3␦ but little CD3⑀ (Fig. 6C). In contrast to the T1 thymocytes, there was less CD8 co-immunoprecipitated with the TCR, especially when compared with the amount of CD8 in the lysate. However, although thymocytes express only CD8␣␤, the hybridomas express CD8␣␤ heterodimers and CD8␣␣ homodimers (29,33), which accounts for the large amount of CD8␣ detected in the lysate. Taken together these results indicate that association of CD3␦ with CD8 is remarkably strong and resists washing with n-octylglucoside and high salt, which combined effectively disrupt the association of CD3␦ with CD3⑀. On the other hand, the association of CD8 with CD3␦ (and CD3⑀) is sensitive to alkylation or chelating of divalent cations, suggesting that it involves free cysteines and chelate complexes. DISCUSSION A key finding of the present study is that CD3␦ mediates a functional link between TCR⅐CD3 and CD8 and that this is crucial for efficient TCR triggering and activation of CD8 ϩ T cells. CD3␦ knockout mice or mice expressing an ␣CPM variant TCR, which lacks the ␦ chain in their CD3 complex, exhibit strongly impaired positive selection and TCR-mediated activation of CD8 SP T cells ( Fig. 1 and Refs. 12,15,18). The same findings were obtained on CD8␤ knockout mice (20,21) and in a milder form, on mice expressing tailless CD8␤ (22,23,29). The present study shows that these signaling defects are explained by the same molecular mechanism, namely that CD3␦ couples the TCR with the coreceptor CD8. Several observations support this conclusion. Using TCR photoaffinity labeling with soluble monomeric K d -PbCS(ABA) complexes, we find that on cells expressing the ␣CPM variant T1 TCR, CD8 fails to markedly increase TCR-ligand binding (Fig. 2). It is known that CD8 increases the avidity of TCR-ligand interactions by binding to TCR-associated MHC complexes and that this coordinate binding requires association of TCR and CD8 (29,32,43). For example, CD8␣␣ or CD8␣␤ lacking the tail of CD8␤ (CD8␣␤Ј), poorly associate with TCR⅐CD3 and therefore inefficiently increase TCR-ligand binding (29). Because T1 TCR ␣IV/␤III lack the ␦ chain of their CD3 complexes (17), our TCR photoaffinity labeling experiments argue that CD3␦ mediates association of TCR⅐CD3 with CD8. Consistent with this are our FRET data showing that in T cell hybridomas wild type T1 TCR is in close proximity to CD8 whereas T1 TCR ␣IV/␤III is not (Fig. 3).
How does CD3␦ associate with the coreceptor ? Because CD8␣␣ poorly associates with TCR⅐CD3 ( Fig. 3 and Ref. 29), CD8␤ is important for this interaction. It has been shown that the tail of CD8␤ is involved in coupling CD8 with TCR⅐CD3 (29), but it is unclear what other portions of CD8 are involved in this interaction and in what way. Moreover, CD3␦ can also be selectively immunoprecipitated with CD4 (26), which is surprising, given the striking structural differences between CD4 and CD8. Because CD4 and CD8 have in common that they associate with Lck and LAT (44), the question arises whether these may be involved in coupling the TCR with the coreceptor. This possibility, however, seems unlikely because mice expressing tailless CD3␦ exhibit nearly normal positive selective of SP T cells (12), arguing that the tail of CD3␦ is not required for its interaction with the coreceptor. Because Lck is intracellular and LAT has only nine extracellular residues, it seem inconceivable that they could interact with tailless CD3␦. Moreover, D3 and D4 of CD4 have been shown to be critical for coupling the TCR with CD4 (25).
Our co-immunoprecipitation experiments indicate that the association of CD8␣␤ with CD3␦ is sensitive to EDTA and even more to alkylation of cysteines by ethylmaleinimide (Fig. 6B). It is interesting to note that CD3␥, ␦, and ⑀ and CD8␣ and ␤ all have two free cysteines in their extracellular membrane proximal regions (Fig. 7A). It has been shown that those of the CD3 chains are engaged in chelate complexes with Zn 2ϩ , which contribute to their dimer formation (4,5). Although for CD8 two of these cysteines form an interchain disulfide bond, the other two may participate in such chelate complexes and thus strengthen its association with CD3␦ (Fig. 7A). The transmembrane portions of CD8 and CD3␦ may also contribute to their association, as suggested by the destabilizing effect of n-octylglucoside (Fig. 6B), which disrupts the association of spanning proteins (41). Further studies are clearly needed to elucidate how CD3␦ associates with the coreceptor, in particular what role its Ig domain plays. Based on the 3D structure of CD3⑀␥ (5), straightforward homology modeling of the Ig domain of CD3␦ is possible. Such modeling suggests that the outer surfaces of CD3␦ and CD3␥ (opposite the interface with CD3⑀), which are most likely to interact with the coreceptor, are strikingly different. The outer surface of CD3␦ is flatter and much less polar than the one for CD3␥, which is consistent with the finding that the association of CD3␦ with CD8 is only slightly ionic in nature (Figs. 6B and 7B).
What are the implications of CD3␦-mediated TCR⅐CD8 coupling for TCR signaling? TCR⅐CD3, lacking CD3␦, exhibits defective signaling such as impaired activation of kinases like Lck, p59 fyn , ZAP-70, and Erk and reduced phosphorylation of LAT (12,18,19). This has been attributed to their poor association with lipid rafts, which by concentrating kinases and their substrates and by excluding phosphatases are privileged sites for the induction of TCR signaling (12,18,36,41,42). Our results demonstrate that raft association of TCR is mediated by binding of CD3␦ to raft-resident CD8 or, more precisely, with CD8/Lck, because CD8 associates with Lck in rafts (29,33). Although they are small in resting cells rafts, they dramatically increase in size upon TCR triggering, which greatly increases the separation of kinases and phosphatases and hence the efficiency of TCR signaling (41,42). Our confocal studies show that co-cross-linking of TCR and CD8 by soluble MHCpeptide multimers or anti-TCR⅐CD3 and CD8 antibodies results in the formation of large aggregates of TCR and CD8 (Fig.  5). This was also observed under conditions where cell activation is prevented, e.g. in the cold or in the presence of Src kinase inhibitors ( Fig. 5 and Ref. 45). 2 This aggregate formation was also inhibited by methyl-␤-cyclodextrin or similar agents, which disrupt lipid rafts, and was not observed on cells expressing the CD3␦ Ϫ ␣CPM variant TCR (Fig. 5). Taken together these findings argue that cross-linking of raft-associated TCR⅐CD3 adducts with CD8/Lck results in strong TCR aggregation and the formation of large rafts and that this is essential for efficient TCR signal induction. Consistent with this is the observation that disruption of rafts greatly diminishes multimer staining of CD8 ϩ CTL, because TCR and CD8 aggregate formation increases the binding of soluble MHC-peptide complexes (45). Furthermore, in cells lacking CD8␣␤ or upon blocking of CD8, raft association of TCR is diminished and TCR cross-linking-mediated TCR aggregation is strongly impaired, just as it is in cells expressing CD3␦ Ϫ TCR⅐CD3 or upon disruption of rafts ( Fig. 5 and Refs. 12, 18, 29, 36, and 45).
In conclusion, the present study shows that CD3␦ serves to Cysteines are shown in boldface, basic residues in black boxes, and acidic ones are in ovals. B, the electrostatic potential of the outer surface (opposite to CD3⑀) of the CD3␥ structure (left) and the CD3␦ model (right) shown from top to bottom. Acidic domains are shown in red, and basic ones are in blue. The images were produced using the software GRASP (49). establish a functional link between the TCR and the coreceptor CD8 and that this is essential for efficient TCR signaling, which in turn is needed for activation and positive selection of CD8 SP T cells. A similar conclusion was reached in a related study using a different approach (50). Even though strikingly different in structure, CD4 seems also to associate with CD3␦, thus forming a similar link with the TCR (25,26). Indeed, mice expressing CD3␦ Ϫ TCR ␣IV/␤III also exhibit impaired positive selection of CD4 SP T cells (16,17). In accordance with this is the finding that Lck plays a crucial role in T cell development and that for positive selection of CD4 and CD8 SP T cells, Lck must be associated with the coreceptors CD4 and CD8, respectively (1,2,46). Finally, it has been shown that the negative selection of CD8 (and CD4) SP T cells is normal in mice expressing CD3␦ Ϫ TCR (16 -18). It is interesting to note that TCR-mediated apoptosis of CD8 ϩ cells is CD8-independent, i.e. in contrast to cell activation and positive selection it is not impaired by the lack of CD8 co-engagement (47,48).