T Cell Receptor Engagement Triggers Its CD3ε and CD3ζ Subunits to Adopt a Compact, Locked Conformation

How the T cell antigen receptor (TCR) discriminates between molecularly related peptide/Major Histocompatibility Complex (pMHC) ligands and converts this information into different possible signaling outcomes is still not understood. One current model proposes that strong pMHC ligands, but not weak ones, induce a conformational change in the TCR. Evidence supporting this comes from a pull-down assay that detects ligand-induced binding of the TCR to the N-terminal SH3 domain of the adapter protein Nck, and also from studies with a neoepitope-specific antibody. Both methods rely on the exposure of a polyproline sequence in the CD3ε subunit of the TCR, and neither indicates whether the conformational change is transmitted to other CD3 subunits. Using a protease-sensitivity assay, we now show that the cytoplasmic tails of CD3ε and CD3ζ subunits become fully protected from degradation upon TCR triggering. These results suggest that the TCR conformational change is transmitted to the tails of CD3ε and CD3ζ, and perhaps all CD3 subunits. Furthermore, the resistance to protease digestion suggests that CD3 cytoplasmic tails adopt a compact structure in the triggered TCR. These results are consistent with a model in which transduction of the conformational change induced upon TCR triggering promotes condensation and shielding of the CD3 cytoplasmic tails.


Introduction
Upon ligand binding, membrane receptors have to transmit information from their ectodomains to their cytoplasmic tails, and several mechanisms have been proposed to account for how this happens. One way would be to simply increase the local concentration of cytoplasmic tails by promoting aggregation of the ectodomains, which can be achieved by crosslinking with a multivalent ligand, or could even be induced by a monovalent ligand, if this binding promotes a conformational change which results in aggregation. The archetypal receptor aggregation mechanism is the activation of membrane receptor tyrosine kinases. Dimerization of these receptors results in activation of their cytoplasmic tyrosine kinase domains and auto-transphosphorylation of their tails [1,2]. A second possible mechanism involves a conformational change in the cytoplasmic tails themselves, somehow transmitted from the ectodomains. The clearest example of this is provided by G protein-coupled receptors [3,4]. The aggregation and conformational-change mechanisms are not mutually exclusive; indeed it has been suggested that membrane protein tyrosine kinase receptors also undergo a conformational change [5]. To date, much evidence has been gathered which strongly suggests that several receptors of immunological interest undergo a conformational change upon ligand binding. These include the erythropoietin receptor, the tumor necrosis factor receptors, Fas, the interleukin-6 receptor, and the B cell receptor [6][7][8][9][10]. In multichain membrane receptors the conformational change induced upon ligand binding could be a result of rearrangement of the quaternary structure of the complex.
The T cell receptor (TCR) complex is composed of ligandbinding subunits (TCRa and TCRb) and signal transducing subunits (CD3c,CD3d,CD3e and CD3f [CD247]) [11][12][13]. The ligand of the TCR consists of a peptide antigen bound to major histocompatibility complex (MHC) class I or class II molecules. Assembly studies, transfection and reconstitution experiments, and detergent dissociation studies suggest that the TCR complex components are organized as dimers [reviewed in [14]]. CD3e forms non-covalently-bound alternate dimers with CD3c and CD3d, TCRa forms disulfide-linked dimers with TCRb, and CD3f is expressed in the form of disulfide-linked homodimers.
In spite of significant advances in understanding how the TCR signal is propagated within the cell, little is known about the mechanisms that initiate TCR signaling. A number of models have been proposed, including oligomerization of the TCR complex [15,16], conformational changes occurring within a single TCRa/ b heterodimer or within the complete TCR complex [17,18], geometrical rearrangements within a multivalent TCR complex [19], and segregation of tyrosine kinases and phosphatases from the TCR complex [20,21]. The poor ability of monovalent (Fab) anti-CD3 antibodies to stimulate T cells compared with bivalent antibodies has long suggested ligand-induced oligomerization as a necessary component of the activation mechanism [22][23][24]. This model has been reinforced by experiments comparing monomeric with an open box, and the foreign Flag epitope, appended to the N-terminus of CD3e, with a dotted square. The relative positions of the ITAMs within the cytoplasmic tails are marked. (B) Addition of an anti-CD3 antibody that induces TCR conformational change protects the tail of CD3e from trypsin proteolysis. The indicated concentrations of anti-CD3 (OKT3) or anti-Flag antibody were added to a detergent lysate of Jurkat cells transfected with Flag-CD3e before incubation with trypsin. The bands corresponding to CD3e and its degradation products were detected by immunoblotting first and oligomeric forms of soluble ectodomains of MHC complexed with antigen peptide [25][26][27]. Nevertheless, there is growing evidence that TCR signaling also involves a conformational change. Early experiments showed that monovalent forms of certain clonotypic (anti-TCR) antibodies induce cocapping with CD4, in what has been taken as the first evidence of a ligandinduced conformational change in the TCR [28]. Further, the isolated tail of CD3f is converted from a phospholipid-bound helical form to a random coil upon tyrosine phosphorylation [29]. By using monovalent and multivalent TCR ligands, we have recently shown that a multivalent engagement is required for induction of the conformational change, and that TCR crosslinking and conformational change are both required for full T cell activation [30]. Indeed, the conformational changes associated with the induced fit of the complementarity-determining region loops have been proposed as a mechanism that contributes to ligand discrimination [31]. If conformational changes in the whole TCR complex are considered, and not only those in the complementarity-determining regions, this mechanism can be generalized to all TCR-pMHC interactions [32].
We have identified Nck as a ligand for the polyproline sequence (PPS) of CD3e [17]. Within the TCR complex, CD3e is in a nonbinding state in non-stimulated T cells, but upon ligand (pMHC, superantigen or antibody) engagement, the CD3e tail undergoes a conformational change that exposes the PPS for Nck binding. To understand the mechanisms that allow the transmission of the conformational change from TCR-complex ectodomains across the membrane to the CD3 cytoplasmic tails, we have now investigated whether the conformational change is transmitted to CD3 subunits other than CD3e. Using a protease-sensitivity assay, we show that the tails of CD3e and CD3f become protected from degradation upon induction of the conformational change. These results suggest that the conformational change in the TCR is transmitted to the tails, not only of CD3e but of CD3f as well. In addition, our results suggest that contrary to our initial models [17,33], the CD3 tails are not converted from a resting ''closed'' conformation into an active ''open'' conformation, but rather are converted from a loose, protease-sensitive conformation to a compact, protease-resistant conformation. We propose that both conformations be renamed as loose and locked, respectively.

Results
The CD3 tails adopt a compact conformation upon TCR engagement We have previously reported that in a pull-down assay immobilized glutathione S-transferase (GST)-Nck binds to stimulated TCR complex but not to the non-stimulated complex [17]. Since Nck binds through its N-terminal SH3 domain to the PPS of CD3e, the pull-down assay revealed a rearrangement of the CD3e tail. The assay indicated that the non-stimulated TCR complex was in a non-binding conformation, which we called closed-CD3 [33]. This form was converted upon stimulation into a binding conformation, which we called open-CD3. This conformational change was also suggested by positive immunostaining of stimulated cells and tissues with the CD3e's PPS-specific monoclonal antibody APA1/1 [34,35]. Thus, the pull-down and APA1/1 staining assays both demonstrated that the PPS of CD3e becomes exposed after TCR engagement.
To further define ligand-induced conformational change in TCR cytoplasmic tails, and also to study whether CD3 subunits other than CD3e are affected, we performed a series of proteasesensitivity assays. Since the tail of CD3e is rich in lysine and arginine residues that are recognized as cleavage sites by trypsin (Fig. 1A), we chose this protease for our studies. To allow examination of the products of C-terminal-end digestion, experiments were done in Jurkat T cells stably expressing a human CD3e chain labeled at its N-terminus with a Flag epitope (fe-Jk cells). Trypsin digestion of detergent lysates of fe-Jk cells generated a CD3e partial-digestion product of 19 kDa, representing a loss of 4 kDa, which is most of the CD3e tail (Fig. 1B). The partial digestion product was not recognized on immunoblots by the PPS-specific antibody APA1/1, suggesting that this sequence had been degraded. Digestion was, however, partially inhibited by addition to the lysate of the stimulatory anti-CD3 antibody OKT3. In contrast, the anti-Flag antibody did not inhibit digestion. Anti-Flag stimulation of fe-Jk cells was previously demonstrated, using the pull-down assay, to be a poor inducer of the conformational change in the TCR, even though it induced tyrosine phosphorylation (Fig. S1). The OKT3-protected CD3e band was recognized by APA1/1, suggesting that the PPS was protected (Fig. 1B). These results indicate that binding of a conformational-change-inducing antibody to the TCR complex renders the CD3e resistant to trypsin digestion, providing further support that the tail of CD3e undergoes a conformational change in response to engagement by a stimulatory antibody.
To determine whether the CD3f subunit also undergoes conformational change, trypsin-digested Jurkat T cell lysates were immunoblotted with antibody 448, which is specific for the Cterminal-most 34 amino acids of CD3f. This showed that the CD3f dimer was digested by trypsin upstream of the sequence recognized by the antibody (Fig. 1C). Incubation with the stimulatory antibody OKT3, but not with an irrelevant isotypematched antibody, partially prevented the loss of the 448 epitope. Furthermore, reprobing the membrane with a polyclonal antibody raised against the whole tail of CD3f detected intermediate-sized partial-digestion products only in cell lysates that had been incubated with OKT3 ( Fig. 1C, WB anti-fcit).
Since TCR triggering with the anti-CD3 antibody OKT3 was performed after lysis of the cells in detergent, it was unlikely that the protective effect on the CD3e and CD3f tails was due to shielding caused by tyrosine phosphorylation of the immune receptor tyrosine-based activation motifs (ITAMs) or to recruitment of signalling proteins such as ZAP70. Nevertheless, in order with anti-Flag antibody and second with the PPS-specific antibody APA1/1. The asterisk indicates the presence of a protein fragment in lane 1 that corresponds to the partial proteolysis of Flag-CD3e, probably caused by cellular proteases contained in the cell lysate. The balance sheet indicates the percentages of protected CD3e molecules, calculated by densitometry of the anti-Flag and APA1/1 immunoblots. (C) The CD3f tails becomes protected from proteolytic cleavage after TCR triggering. A Jurkat cell lysate was incubated with 10 mg/ml of either OKT3 or an isotypic control antibody (OKT8) before digestion with trypsin. Immunoblotting was performed with anti-CD3f antibodies reacting with the C-terminus (448) or the whole tail (anti-fcit). (D) Tyrosine phosphorylation does not affect the sensitivity of CD3e to trypsin proteolysis. A Jurkat cell lysate was incubated with 10 mg/ml of OKT3 in the presence of 20 mM of the src kinase inhibitor PP2 for 15 min at room temperature, before digestion with trypsin. Immunoblotting was performed with antibody APA1/1. (E) Protection against trypsin proteolysis is independent of TCR aggregation. A Jurkat cell lysate was preincubated with 50 mg/ml of Fab fragments of OKT3 or the isotypic control antibody 12CA5 (anti-HA), before digestion for 15 min at 37uC with the indicated concentrations of trypsin. Immunoblotting was performed with APA1/1 followed by reprobing with the 448 antiserum. The presence of immunoglobulin light chain (L) from the anti-HA Fab fragment in the APA1/1 immunoblot is indicated. doi:10.1371/journal.pone.0001747.g001 to exclude this possibility, we included the potent Src family kinase inhibitor PP2 [36] in the protease sensitivity assay. The result showed that, even in the presence of PP2, the triggered TCR acquired resistance to trypsin digestion (Fig. 1D), suggesting that the protection effect of TCR triggering was not due to a post-lysis modification of the TCR. Another important control was to demonstrate that the acquisition of resistance to trypsin was not due to the formation of aggregates that are poorly accessible to trypsin upon crosslinking of the TCR with the bivalent anti-CD3 antibody. Engagement of the TCR complex in Jurkat cell lysates with a monovalent Fab fragment of OKT3, but not the incubation with an isotype-matched irrelevant antibody, increased the resistance of the CD3e and CD3f tails to trypsin digestion (Fig. 1E). This result indicates that the acquisition of resistance to trypsin digestion is not caused by aggregation of the TCR, and is consequent with previous evidence showing, with the pull-down assay, that a monovalent anti-CD3 antibody induces the conformational change [17]. The results shown in Figures 1D  and 1E exclude alternate explanations, and suggest that the acquisition of resistance to trypsin digestion is caused by a conformational change transmitted to the CD3 tails.
The acquisition of resistance to trypsin digestion by CD3e and CD3f was also noted when OKT3 was used to stimulate intact Jurkat T cells before lysis, but not when an irrelevant antibody (anti-CD4) was used ( Fig. 2A). Furthermore, the protected CD3e and CD3f bands correspond to the full-length proteins, since they were recognized by M20 and 448, respectively, two polyclonal antibodies specific for the C-terminal ends. This result suggests that the conformational change involves a rearrangement of the whole tails.
The protective effect of TCR stimulation against trypsin digestion of CD3e and CD3f tails was seen not only with anti-CD3 dimer antibodies (e.g. OKT3), but also with stimulatory antibodies for the TCRa/b heterodimer. Thus, stimulation of intact Jurkat cells with antibodies C305 and BV8, specific for the variable Vb region, partly prevented the degradation of the CD3e and CD3f tails by trypsin (Fig. 2B). This effect was, however, not seen with the anti-Cb antibody Jovi.1, which is a poor inducer of the conformational change according to the GST-Nck pull-down assay ( [17] and Fig. 2C). Isotypic differences cannot explain the differential effect of the anti-TCR antibodies, at least for antibodies OKT3, Jovi.1 and HP2/6 (irrelevant antibody in Fig. 2C, lane 2) which are of the same isotype (IgG2a). These results strongly suggest that TCR engagement induces a conformational change that is transmitted to the cytoplasmic tails of CD3e and f.
From the mobilities of the digestion products ( Fig. 1A and 1C) it appeared that the cytoplasmic tail of CD3e from non-triggered TCRs was completely digested, whereas the tail of CD3f was only partly accessible to trypsin. The estimated loss of relative mass in the CD3f 2 dimer after digestion was 16 kDa (from 32 to 16 kDa), which is below the 24 kDa loss that would be expected if all the CD3f tail were digested (Supplemental material Fig. S2). Upon stimulation, partial proteolytic products were detected at 22 and 27 kDa (Fig. 1C). The sequences that separate the three Immune receptor Tyrosine-based Activation Motifs (ITAM) in CD3f are particularly rich in basic amino acids, and therefore in potential trypsin-cleavage sites (Fig. 1A, 3A). Inspection of ITAM distribution in CD3f suggests that the 27 kDa fragment could derive from a cleavage between ITAMs B and C, and the 22 kDa fragment from cleavage within ITAM B (Suppl. Fig. S2). The 16 kDa product resulting from digestion of CD3f in resting TCRs could derive from cleavage between ITAMs A and B. If these calculations are correct, they would indicate that compared with the more membrane-distal ITAMs, ITAM A might be constitutively protected from trypsin attack in the resting TCR. No antibody specific for ITAM A was available, so to test this we generated a truncated CD3f mutant with a Flag epitope appended immediately after ITAM A (Fig. 3A, construct fAflag). Transfection of this construct into Jurkat cells generated disulfide-linked fAflag homodimers and heterodimers of fAflag and endogenous CD3f (Fig. 3B, total lysates, TL). TCR complexes containing either one (ffAflag) or two fAflag constructs (fAflag2) underwent the conformational change after TCR triggering, as indicated by a positive reaction in the GST-Nck pull-down assay (Fig. 3B). However, fAflag was completely resistant to trypsin digestion even in non-triggered TCRs, whereas the Flag epitope was completely digested when appended at the C-terminal end of full-length CD3f (Fig. 3C). These results suggest that the CD3f ITAM closest to the membrane is permanently protected from digestion and that the conformational change in the TCR modifies the exposure of the second and third ITAMs of CD3f to trypsin cleavage.

Discussion
Limited proteolysis has become an established tool for the study of conformational changes [37,38]. For instance, the atrial natriuretic peptide receptor becomes susceptible to cleavage by exogenously added protease when bound to its lignad [38]. In this paper we have used limited proteolysis to study the induction of conformational changes in the TCR complex. Our results show that the cytoplasmic tails of CD3e and CD3f subunits extracted from resting cells are almost completely digested by added trypsin, but become protected upon TCR triggering. Up to now, the strongest evidence for a conformational change in the TCR has come from studies showing ligand-induced exposure of the PPS in the cytoplasmic tail of CD3e [17]. One biochemical assay is based on pull-down with immobilized GST-Nck, and has served to demonstrate that certain stimulatory anti-CD3 antibodies [17] and a panel of pMHC ligands [33] induce a conformational change in the TCR. A second assay to detect these conformational changes is based on exposure of a neo-epitope recognized by the monoclonal antibody APA1/1. This epitope coincides with the PPS in CD3e and reveals, like the pull-down assay, a conformational change transmitted to the cytoplasmic tail of CD3e [34]. Through the use of APA1/1 it has been possible to confirm that the TCR undergoes conformational changes during antigen recognition in vivo, and that the TCR complexes undergoing the conformational change are located in the immune synapse; these experiments moreover showed that the conformational change is elicited by full but not by partial agonist/antagonist peptides [34]. Unfortunately, the pull-down and APA1/1 recognition assays reflect the same molecular event (rearrangement of the PPS in the tail of CD3e) and thus did not provide evidence of conformational change in TCR subunits other than CD3e.
Previous evidence for conformational change in the cytoplasmic tail of CD3f was based on in vitro biochemical studies. Isolated CD3f cytoplasmic tails change from a lipid-bound helical structure to an unfolded conformation upon tyrosine phosphorylation [29]. These results support the earlier finding that a synthetic peptide corresponding to the third ITAM of CD3f adopts an a-helical structure in the non-phosphorylated form and a b strand conformation when phosphorylated [39]. However, both these studies were performed in cell-free systems with synthetic peptides and recombinant proteins, in the absence of other TCR subunits. The limited proteolysis studies reported in this paper therefore provide the first evidence that CD3f undergoes a conformational change within the TCR complex in response to ligand binding. Due to the lack of appropriate antibodies, we have not yet been able to study the effect of limited proteolysis on CD3c and CD3d, but a model could be proposed in which a conformational change is transmitted from the ligandbinding ectodomains of the TCRa/b to the cytoplasmic tails of all CD3 subunits.
The conformational change induced in CD3f, however, shows distinct features from those confirmed for CD3e. Whereas the whole cytoplasmic tail of CD3e is susceptible to proteolytic degradation in the non-triggered TCR and is protected upon stimulation, in CD3f ITAM A is protected even in non-triggered TCRs, and stimulation extends this protection to ITAMs B and C. Interestingly, CD3f is sequentially phosphorylated on its three ITAMs during T cell activation (for a review see [40]). In some cases ITAMs B and C are phosphorylated constitutively, producing the p21 tyrosine phosphorylated form of CD3f. In contrast ITAM A is phosphorylated only upon TCR triggering, yielding the p23 form. Although the detailed functional significance of the p21 and p23 forms of CD3f is not clear, it is wellestablished that phosphorylation of ITAM A defines the difference between triggered and non-triggered TCRs. In light of our limited proteolysis results, we suggest that, in a resting TCR, ITAM A is in a compact conformation that is not accessible to tyrosine phosphorylation by the priming src kinases [41], and that the conformational change not only compacts ITAMs B and C, but also reorients ITAM A into a conformation more susceptible to phosphorylation.
We recently described that both TCR crosslinking and conformational change are required for full tyrosine phosphorylation of different intracellular effectors [30]. In this regard, the acquisition of protease resistance by the tail of CD3f shown in the present study suggests that the conformational change may also affect the phosphorylation of CD3f and, therefore, the recruitment of ZAP70 and the subsequent phosphorylation of downstream effectors. This requirement of the conformational change for tyrosine phosphorylation appears to be contradicted by the activation of tyrosine phosphorylation with the anti-flag antibody, a poor inducer of the conformational change in Flag-CD3eexpressing Jurkat cells (Supplemental Figure S1). Several explanations can be given. The most simple is that the anti-flag antibody induces the conformational change to a level that is sufficient to pass a threshold for activation of tyrosine kinases. A second possibility, is that TCR crosslinking in the absence of conformational change is sufficient for the activation of tyrosine kinases but not for a normal pattern of phosphorylation, i.e. TCRassociated tyrosine kinases could be activated without the TCR undergoing a conformational change, but the access to their potential substrates (ITAMs and downstream effectors) would be limited. Finally, the number of tools that we have to study the conformational change, or conformational changes, in the TCR is limited. The exposure of the PPS in CD3e, revealed by the GST-Nck pull-down assay and APA1/1 epitope display [17,34], and the trypsin sensitivity assay shown in the present study, might be only coarse methods to understand the fine tuning of signal transduction by the TCR.
The acquisition of resistance to trypsin digestion upon TCR stimulation suggests that, contrary to our prior prediction, stimulation does not shift the cytoplasmic tails of the TCR from a ''closed'' to an ''open'' conformation [17,33]. Our present data indicate instead that the cytoplasmic tails in the non-engaged TCR are in a loose conformation that makes them accesible to trypsin digestion. We now prefer to name this conformation the Nck nonbinding or loose conformation. Upon antibody stimulation, a conformational change is transmitted from the TCR ectodomains to the CD3 cytoplasmic tails, which become packed into a more compact structure with reduced accesibility for trypsin. This is the Nck-binding or locked conformation. The reduced exposure of the CD3 cytoplasmic tails to trypsin must occur simultaneously with an increased exposure of the CD3e PPS to Nck perhaps by fixing it into a conformation adequate for binding. We do not know the ultimate causes that explain why a trypsin-sensitive loose conformation in CD3e is incompatible with a Nck-binding conformation of the PPS. We have however structural information (obtained by NMR) on how the PPS binds the SH3.1 domain of Nck (Borroto and Alarcón, in preparation). The CD3e polypeptide makes an extensive fingerprint on the SH3.1 domain, where not only the central proline residues participate, but also upstream and downstream charged amino acids. This explains why CD3e binds with abnormally high affinity for a SH3-ligand interaction (in the order of 0.1 mM), and may also indicate structural requirements for the interaction. In the loose conformation the sequences upstream and/or downstream of the PPS-and not necessarily the central prolines themselves-might be in non-binding conformation. The compaction of the CD3 tails resulting in the locked Figure 4. Model of conformational change in the cytoplasmic tails of CD3 subunits after TCR engagement. In non-stimulated cells, the CD3 subunits in the TCR are in a loose conformation that is accessible to trypsin digestion. The PPS of CD3e (pink square) is in a non-binding conformation for Nck. After TCR engagement with pMHC or with stimulatory antibodies, the CD3 ectodomains adopt an active conformation (symbolized with rectangular forms) that is transmitted to the cytoplasmic tails of the CD3 subunits via a rotation and/or scissor movement. The cytoplasmic tails close up to form a compact structure that is less sensitive to proteolytic attack. This conformation permits Nck-binding by CD3e. doi:10.1371/journal.pone.0001747.g004 conformation may bring the Nck-binding sequence into the appropriate conformation.
In this study we have used anti-CD3 antibodies to elicit the conformational change that results in the trypsin-protected or locked conformation in both CD3e and CD3f. The locked conformation was also induced when antibodies that recognize the variable domain of the TCRb ectodomain were used for stimulation. We can therefore propose that a conformational change transmitted from the ectodomains of the TCRa/b heterodimer to the CD3 tails induces their locked conformation. We have attempted to demonstrate that a similar rearrangement takes place upon antigen stimulation. However these experiments have failed, probably due to the fact that continued pMHC-TCR interaction after lysis of the cells is necessary to preserve the conformational change [30].
The transmission of TCR conformational change across the plasma membrane presents a conceptual challenge. The structure of the CD3c-CD3e dimer ectodomains led to the proposal that the transmembrane domains of the dimer undergo a piston-like movement. In this way, the paired G beta strands of CD3e and CD3c (as well as those of CD3e and CD3d) would form a rigid rod-like connector that would displace the transmembrane helices [42]. On the other hand, monovalent Fab fragments of the anti-CD3 antibody OKT3 bind to CD3e in a side-on orientation [43], and this interaction induces a conformational change in the TCR [17]. Considering this in conjunction with the electrostatic properties of CD3e, it has been proposed that the transmission of the conformational change might require a rotational or scissorlike movement of both the transmembrane domains and the cytoplasmic tails of the CD3 subunits [43]. The protease-sensitivity data reported in the present paper support a model in which ligand binding to the TCRa/b heterodimers somehow promotes a combination of rotational and closing-scissor movements that simultaneously condense and shield the cytoplasmic tails of the CD3 subunits, while exposing key features, such as the PPS of CD3e and ITAM A of CD3f. Studies with purified recombinant proteins have shown that all ITAM-bearing cytoplasmic tails studied (including those of the TCR and BCR) form oligomers in solution [44], although the interaction is weak (Kd in the order of 10 mM). The conformational change in the TCR might exploit this natural propensity of ITAM-bearing cytoplasmic tails to dimerize or oligomerize, by forcing the local effective ITAM concentration above the dissociation constant. Although the capacity of the three ITAMs of CD3f to homodimerize has not been measured on a one-by-one basis, it is tempting to speculate that the protease resistance of ITAM A could be caused by a higher propensity of this ITAM to dimerize in the resting TCR, compared to ITAMs B and C. Alternatively, protease resistance of ITAM A in the resting TCR could be due to the position of CD3f in the TCR complex, i.e. the tail of CD3f could occupy an internal position within the TCR complex, where the tails of the other CD3 subunits could shield the membrane most proximal region of CD3f's tail.
In our current model, the conformational change is initiated after binding of a cluster of two or more pMHC agonists to two or more TCRa/b heterodimers within a multivalent TCR complex [30]. This would generate a torque on the TCRa/b heterodimers that could be transmitted to the CD3 dimers. In turn, this binding would induce the rotation or sliding of the transmembrane domains of the CD3 dimer with respect to those of the TCRa/b heterodimers, and an ensuing transmission of this movement to the CD3 cytoplasmic tails. Crystal structures of TCRa/b ectodomains bound to pMHC complexes almost universally show that the orientation of TCRa/b is approximately diagonal to the MHC peptide-binding groove (for a review see [45]). We hypothesize that in the context of a multivalent TCR, the diagonal orientation imposed by pMHC binding on the two TCRa/b heterodimers may be responsible for the torque transmitted to the CD3 subunits.

Plasmids
The pGEX-4T1 derivative GST-SH3.1, containing the aminoterminal SH3 domain of Ncka, was kindly provided by Dr. R. Geha (Children's Hospital, Harvard Medical School, Boston). The pSRa-CD3f-Flag plasmid encoding human CD3f was generated by PCR. The truncated CD3fA construct (pSRa-CD3fA-Flag) expresses a protein truncated immediately after the first ITAM of human CD3f. Both constructs encode the Flag peptide fused to the C-terminus of the protein.

Cells
The human T cell line Jurkat was maintained in complete RPMI 1640 supplemented with 10% fetal bovine serum (Sigma). The Jurkat cell clone fe-Jk expressing Flag-CD3e has been described previously [17]. The JkfAF and JkfF cell lines were generated by stable transfection into Jurkat of the pSRa-CD3fA-Flag and pSRa-CD3f-Flag, respectively.
Trypsin digestion 10 6 cells per point were stimulated with 10 mg/ml soluble antibody for 5 min at 37uC and disrupted in lysis buffer without leupeptin, aprotinin or PMSF. A total of 1 mg/ml trypsin (Sigma) was added to the lysate and the reaction mix was incubated for 15 min at 37uC. Afterwards, 36 Laemmli's sample buffer was added, and samples boiled, to stop the enzyme activity. Alternatively, postnuclear cell lysates of unstimulated cells were incubated with the anti-CD3 antibody before trypsin digestion under the same conditions.

Supporting Information
Supplemental Figure S1