An allosteric site in the T-cell receptor Cβ domain plays a critical signalling role

The molecular mechanism through which the interaction of a clonotypic αβ T-cell receptor (TCR) with a peptide-loaded major histocompatibility complex (p/MHC) leads to T-cell activation is not yet fully understood. Here we exploit a high-affinity TCR (B4.2.3) to examine the structural changes that accompany binding to its p/MHC ligand (P18-I10/H2-Dd). In addition to conformational changes in complementarity-determining regions (CDRs) of the TCR seen in comparison of unliganded and bound X-ray structures, NMR characterization of the TCR β-chain dynamics reveals significant chemical shift effects in sites removed from the MHC-binding site. Remodelling of electrostatic interactions near the Cβ H3 helix at the membrane-proximal face of the TCR, a region implicated in interactions with the CD3 co-receptor, suggests a possible role for an allosteric mechanism in TCR signalling. The contribution of these TCR residues to signal transduction is supported by mutagenesis and T-cell functional assays.

A key step in T-cell-mediated adaptive immunity is the triggering of cell-surface ab T-cell receptors (TCR) by peptide-loaded major histocompatibility complex (p/MHC) proteins on target antigen presenting cells 1,2 . TCR-a and -b polypeptide chains are encoded by genes assembled by recombinatorial assortment of V-J and V-D-J gene segments, respectively, and non-templated nucleotides added at junctions of rearrangement during T-cell ontogeny in the thymus. Encounter of particular clonally expressed TCR with cognate p/MHC ligand triggers a signalling cascade leading to a variety of cellular programmes including thymic selection, proliferation, cytokine production and differentiation into effector and memory T cells 3 .
Whereas antigen specificity is dictated by the amino-terminal variable (V) domains of the ab-receptor, signalling function is mediated by the non-covalently associated co-receptor CD3ge, de and zz dimers, which bear cytoplasmic immunoreceptor tyrosine-based activation motifs (ITAMs) 4,5 . Ligand binding to the TCR/CD3 complex extracellularly initiates intracellular signalling through Src kinase-mediated phosphorylation of these ITAMs 6 . In addition to their signalling function, CD3 subunits are also required for stable cell-surface expression of the TCR/CD3 complex 7,8 . Mechanistic details concerning the transmission of signals from the extracellular domains of the TCR to the intracellular ITAMs are incomplete, and are the subject of considerable interest, the importance of which is highlighted by diseases associated with dysfunction of this cellular process 9 , the immunosuppressant role of therapeutic antibodies targeting the TCR/CD3 complex 10 and the potential of synthetic TCRs towards immunotherapeutic applications 11,12 . Efforts to understand the molecular basis of TCR-mediated signalling have relied largely on biophysical, structural and functional approaches 13 . Binding of p/MHC to the TCR induces structural changes at the cytoplasmic face of the TCR/CD3 complex, as evidenced by the accessibility of a polyproline sequence in the CD3e cytoplasmic tail 14 , and the repositioning of Tyr residues within the CD3 cytoplasmic ITAMs from a relatively inaccessible membrane-associated form to a cytoplasmically oriented, kinaseaccessible conformation 15 . However, the molecular mechanism by which p/MHC binding to the TCR is communicated to the associated CD3 subunits for signalling remains unknown.
To gain further insight into the dynamics of TCR/MHC interactions, we employ complementary biophysical methods to examine the high-affinity B4.2.3 TCR in both the liganded and unliganded states. X-ray structures indicate a large rearrangement of the complementarity-determining region 3 (CDR3) loops upon binding. In addition, chemical shift mapping utilizing complementary backbone amide and side-chain methyl NMR probes reveal several residues in the Cb domain of the TCR, distant from the ligand-binding interface and close to a putative CD3-binding site, that show significant perturbations upon ligand binding. Finally, mutational and functional analyses suggest a critical role of these allosteric sites in signal transduction. These results indicate a dynamic activation mechanism, where p/MHC recognition by the CDRs triggers conformational remodelling of interactions near the Cb H3 helix at the membrane-proximal face of the TCR.

Results
TCR binds to its pMHC ligand with high affinity. The B4.2.3 T-cell hybridoma, derived from a BALB/c mouse immunized with P18-I10 (RGPGRAFVTI), is sensitive to picomolar concentrations of peptide presented by the MHC-I molecule, H2-D d (refs 16,17). To probe the affinity and kinetics of the interaction between the TCR and p/MHC, we first employed surface plasmon resonance (SPR) where immobilized P18-I10/H2-D d was offered graded concentrations of the B4.2.3 TCR. The measured affinity (K D ) is B0.54 mM (Fig. 1a), and no binding of the TCR was detected to H2-D d displaying the motif peptide (MTF) AGPARAAAL, a negative control (Fig. 1b). As TCR affinities for p/MHC ligands span a wide range of K D from 0.5 to 4100 mM (ref. 18), this affinity is among the highest reported for a naturally occurring TCR.
To explore the TCR/MHC interaction in a complementary assay, we employed sedimentation velocity analytical ultracentrifugation, which permits assessment of the stoichiometry as well as the affinity of the interaction. The B4.2.3 TCR interacts with P18-I10/H2-D d strongly over a broad range of concentrations (0.1-20 mM of each component) displaying a characteristic concentration-dependent sedimentation coefficient distribution, indicative of a greater time average of molecules in complex as a function of concentration (Fig. 1c). The limiting sedimentation coefficient of B5S is consistent with a 1:1 stoichiometry, considering the size and shape of the individual components. The strong interaction of the TCR with the cognate P18-I10/H2-D d ligand contrasts sharply with mixtures of B4.2.3 TCR and MTF/H2-D d (Fig. 1d), both of which sediment at a concentration-independent velocity of 3.6S. To determine the binding affinity, we analysed the isotherm of signal weighted-average sedimentation coefficient (s w ) using the s w values determined from the c(s) distributions. A simple 1:1 association model with affinity (K D ) of 0.23 mM fits the data well (95% confidence interval: 0.14-0.40 mM). Thus, the B4.2.3 recombinant TCR interacts with P18-I10/H2-D d with high affinity, in two distinct biophysical assays, consistent with the high peptide sensitivity of the T-cell hybridoma in functional assays 16,17 .
TCR uses plasticity in the CDR3 loops to recognize pMHC. To elucidate the details of the ligand/receptor interaction, we determined the crystal structures of both free and p/MHC-bound states of the B4.2.3 TCR. The unliganded TCR crystallized in the P3 1 space group, with three heterodimers in the asymmetric unit (data collection and refinement statistics are provided in Table 1). The P18-I10/H2-D d /B4.2.3 complex formed crystals in the C2 space group and diffracted to 2.1 Å, revealing a structure with an overall orientation of the TCR on P18-I10/H2-D d that conforms to previously elucidated general principles [19][20][21][22] (Fig. 2a). The germline-encoded CDR1 and CDR2 loops of the TCR achain are positioned towards the C-terminal half of the H2-D d a2 helix, resulting in a diagonal orientation mode (Fig. 2b). The resulting crossing angle calculated as described 22 is 30°, within the range observed for the majority of stimulatory TCR/MHC complexes (22°-87°; ref. 22). The shape complementarity index 23 of the interface is 0.71 (1.0 is a perfect match), and is among the highest observed for any TCR/MHC complex 22 . A summary of the interactions between selected residues of the CDRs with the MHC a1 and a2 helices and peptide is shown in Fig. 2c. While aand b-subunits contribute almost equally to the 1,771 Å 2 area of interaction between B4.2.3 and H2-D d (calculated without peptide), nine of ten hydrogen bonds at the interface are provided by the b-chain CDRs. Thus, this pMHC/ TCR interface is dominated by Vb interactions within the broad range observed for many different pMHC/TCR complexes 22 .
Superposition of the aand b-chains of the liganded B4.2.3 TCR with their unliganded counterparts reveals marked changes in the disposition of the CDR3a and CDR3b loops, with little alteration discernible in CDR1 and CDR2 or in the C domains ( Supplementary Fig. 1a-d) as observed in other TCRs 22,24 . In the structures examined here, CDR3a, spanning Ala95 to Lys102, undergoes a large conformational change upon ligand binding, in particular a 9.1 Å displacement of the C a atom of Asp99 of the first molecule in the asymmetric unit of the unliganded TCR compared to the same atom in the liganded TCR ( Supplementary  Fig. 1b). The movement projects the CDR3a loop into the peptide-binding groove, and allows interactions of the Phe97 side chain with the peptide backbone of Gly2 and Gly4. This is illustrated by a 9.2 Å displacement of the C z atom of the Phe97 side chain from the free to bound form. CDR3b, from Ser92 to Val99, is displaced inwards on ligand binding by 3.1 Å as measured at the C a of His96. Thus, the X-ray structures indicate large movements of the CDR3a and -b loops upon p/MHC engagement that are critical for achieving a highly complementary interface. The extensive interface and hydrogen-bonding network between the aand b-CDR domains with the P18-I10 peptide suggests an enthalpic compensation for the apparent entropic loss due to loop rearrangement and provide a structural basis for the measured high affinity of the interaction, as demonstrated both functionally and biochemically.
Increased dynamics in regions of the TCR b-chain in solution.
To gain insight into the dynamics of the B4.2.3 TCR in solution in both its free and P18-I10/H2-D d -bound states, we prepared TCR samples for NMR labelled at the b-chain alone by in vitro assembly of unlabelled a-chain with triple-labelled ( 2 H, 13 C, 15 N) or AILV side-chain methyl-labelled b-chain. The resulting proteins showed well-dispersed spectra, indicating stable, properly conformed monomeric TCR, free of aggregation or degradation ( Fig. 3a and Supplementary Fig. 2).
Previous studies have established strategies for obtaining NMR assignments of TCR backbone atoms using TROSY-based methods under extensive perdeuteration of the a and b-chains 25,26 . Here we employed a multipronged approach for assignment and cross-validation of both backbone amide and side-chain methyl resonances of isotopically labelled b-chain TCR samples (outlined in Supplementary Fig. 2). First, backbone amide chemical shifts in the 2D 1 H-15 N TROSY-HSQC were assigned sequentially utilizing a combination of three-dimensional (3D) HNCA, 3D HN(CA)CB and 3D HNCO experiments recorded on a triple-labelled sample. Second, Ile 13 Cd1, Leu 13 Cd1/ 13 Cd2, Val 13 Cg1/ 13 Cg2 chemical shift assignments were obtained from 3D HMCM[CG]CBCA methyl out-and-back experiments 27 recorded on a selectively methyl-labelled sample 28 (Supplementary Fig. 3). Backbone amide assignments obtained using the J-correlated experiments were validated by acquiring amide-to-amide NOEs in TROSY-based 3D H N -NH N nuclear Ö verhauser enhancement spectroscopy (NOESY) experiments, while side-chain methyl assignments were validated with NOE connectivities obtained from 3D H M -C M H M and 3D C M -C M H M SOFAST NOESY experiments 29 . Close comparison of the NOE crosspeak intensities to their corresponding distances in the X-ray structure further permitted stereospecific disambiguation of the geminal Leu  29 . This approach allowed us to achieve backbone amide assignments for 90% of non-Pro residues and side-chain methyl assignment of 100% Iled1, Leud1/d2, Valg1/g2 and Alab methyls for b-chain labelled TCR. Examples of our sequential assignment and NOESY-based cross-validation strategy for two selected regions of the b-chain are presented in Supplementary Figs 4 and 5. Thus, with the addition of complete, stereospecific assignments of methyl groups to our backbone assignments we obtained a complementary network of probes towards mapping dynamics of the TCR b-chain in solution.
The backbone assignments and NOE connectivity patterns further confirm the structural features of the TCR b-chain, including the identification of all conserved Ig domain disulfide bonds in an oxidized form as evidenced by the 13 C a and 13 C b chemical shifts of the four Cys residues (22,90 in Vb and 141, 202 in Cb), as well as the observation of a single peak for all residues in the vicinity of the disulfides. Resonances from several residues within the b-chain CDR loops, CC 0 loop and FG loop ( Supplementary Fig. 6) were absent in the TROSY spectra, likely due to conformational exchange line-broadening, suggesting the sampling of alternative environments on a ms-ms timescale. This is consistent with the multiple crystallographic conformations observed for the same residues among the three TCR molecules in the asymmetric unit ( Supplementary Fig. 1d).
The NMR backbone chemical shifts are highly sensitive probes of the local environment and secondary structure of the molecule. On the basis of the assignments of backbone 1 H, 15 N H , 13 C a , 13 C b and 13 CO atoms, we used TALOS-N 30 to calculate the secondary structure index and found these predictions to be in excellent agreement with the DSSP 31 annotation of our X-ray structure of the free TCR (Fig. 3c,d). When provided with near-complete assignments, the detail of structural information contained in the chemical shifts analysis is highlighted by the robust prediction of the boundaries for the H3 and H4 a-helices in the Cb domain, as well as the 3 10 helical segment (residues 113-115) located in the linker between Vb and Cb (shown as red bars in Fig. 3c, and coloured red in the diagram of Fig. 3d). The chemical shift-derived order parameter (RCI-S 2 ), expressed in the range 0-1 with 0 indicating a random coil and 1 a fully ordered backbone structure, further reveals regions of increased disorder. As opposed to the more detailed model-free order parameters 32 derived from fitting 15 N relaxation rates and 15 N-{ 1 H} NOE ratios that probe ps-ns timescale motions directly, this analysis is based on a statistical comparison of backbone chemical shifts with database values for random coils using an empirical formula parameterized by comparison to molecular dynamics simulations 33 . Among regions of the TCR b-chain (lower plot in Fig. 3c), the b 2 -b 3 loop (residues 13-17) and the b 7 -b 8 loop (residues 68-72), not directly involved in p/MHC binding, as well as the CDR1 and CDR3 loops in the Vb domain, all show decreased order parameter values with correspondingly increased crystallographic B-factors. The most unstructured region of the b-chain of the TCR in solution is the FG loop located in Cb (residues 206-225), with RCI-S 2 values systematically below 0.75 for most residues and as low as 0.5 for Gly218. In contrast, the CC 0 loop (residues 157-166) shows only a minor reduction in backbone rigidity, with values in the range 0.8-0.9. However, the CC 0 loop may still undergo dynamic motions at a longer time scale window (ms-ms), as suggested by the missing resonances for three amides in that region due to conformational exchange linebroadening or increased solvent exchange rates.
b-chain residues involved in the a/b-interface in solution. To characterize the a/b-interface of the B4.2.3 TCR in solution, we *Asterisked numbers correspond to the last resolution shell.
where Ii(h) and oI(h)4 are the ith and mean measurement of the intensity of reflection h. zA pseudo-merohedral twin fraction was estimated by Xtriage in PHENIX.
where Fobs (h) and Fcalc (h) are the observed and calculated structure factors, respectively. No I/s(I) cutoff was applied. ||Rfree is the R value obtained for a test set of reflections consisting of a randomly selected 5% subset of the data set excluded from refinement.
utilized a TROSY-based cross-saturation transfer experiment 34 in which TCR was prepared with 15 N, 2 H-labelled b-chain and unlabelled (protonated) a-chain. Here selective irradiation of the aliphatic protons of the a-chain is expected to transfer to b-chain amides located at the interface with the a-chain through crossrelaxation. Peak intensity ratio analysis of b-chain amide resonances from saturated relative to non-saturated control experiments revealed cross-saturation transfer effects occurring from the a-chain to b-chain along discrete surfaces. These include a patch of residues on Vb near the a/b interface (Gly41, Leu42, Gln43, Cys90, Phe101) and a more extended surface area on Cb that includes residues near the H3/H4 helix regions (Ala137, Val140, Arg187, Val188; Fig. 4a,b). Importantly, b-chain residues distal from the a/b interface are not affected by a-chain saturation as expected (Fig. 4a,b). Mapping of affected b-chain residues on the X-ray structure of free B4.2.3 TCR suggests that the a/b-interface in solution is in good agreement with the crystallographic interface, and is dominated primarily by contacts between the constant domains (Fig. 4b). These NMR data are consistent with analysis of the X-ray structures of the TCR, which reveal that the Ca/Cb interface has nearly twice the surface area as that of Va/Vb (1216 versus 698 Å 2 ). Examination of a number of different TCR X-ray structures indicates that this is a generally observed phenomenon.
NMR reveals pMHC-binding effects on the TCR b-chain domain.
The completeness of the backbone assignments of unliganded TCR b-chain ( Fig. 3a) allowed us to probe local conformational changes upon binding to its p/MHC ligand. The complex between the TCR ectodomains and P18-I10/H2-D d is 94.7 kDa and is therefore difficult to characterize by standard NMR methods.
Here the use of extensive deuteration of side-chain protons to improve 13 C relaxation and the application of TROSY methods at high magnetic fields to improve 15 N relaxation allowed us to obtain high-quality NMR spectra of the bound state at room temperature ( Fig. 3b). In agreement with the previous SPR and AUC measurements (Fig. 1b,c), B4.2.3 TCR formed a tight complex with P18-I10/H-2D d under the NMR sample conditions. The exchange between free and bound states of the TCR was slow on the chemical shift timescale, as indicated by a single set of peaks for the complex, with large chemical shift changes relative to the free state (up to 0.5 p.p.m. when scaled relative to the 1 H field, as shown along the b-chain sequence in Fig. 5c). Analysis of the changes in peak positions and intensities revealed two types of effects on the TCR b-chain amide resonances: (1) chemical shift perturbations (Fig. 5c, black bars), indicating a change in the local magnetic environment and (2) conformational exchange-induced line-broadening in the bound state (Fig. 5c, red bars), suggesting a perturbation in ms-ms timescale dynamics or increased solvent exchange rates. Several of the observed chemical shift changes correlate with the displacement of residues in Vb domain loops seen in the X-ray structures, such as Ser27 in CDR1b, Asp52 in CDR2b and several residues in CDR3b ( Fig. 5a, right panel). In addition, the peak intensities of residues 49-53 located in the CDR2b loop were significantly attenuated in the bound form (Fig. 5c). Notably, residues 100-105 located at   19 , is 30°. (c) Contact map illustrates interactions between TCR and peptide residues (in gold), and between TCR and MHC helices (in red and blue). Data collection and refinement statistics are provided in Table 1. Structural changes between the free and bound forms of the TCR are outlined in Supplementary Fig. 1. the Va/Vb domain interface also showed above average chemical shift perturbations (Fig. 5c), indicating that p/MHC-bindinginduced conformational changes in the V domains are not restricted to the CDR loops. Strikingly, the NMR data revealed long-range effects on the Cb of the TCR, at the membrane proximal face of the molecule. These changes are highly localized, and cluster near the H3 helix of Cb (Glu130, Thr138) at the interface with the Ca domain (Fig. 5c). In addition to the shifted peaks, the resonances of several amides in the region were significantly attenuated in the bound form, suggesting further changes in dynamics at or near the H3 and H4 regions. These include Ser127, Lys134, Ser183, Arg187 and Val188, all located at the interface with Ca (highlighted with red asterisks in Fig. 5c and shown as red bars in Fig. 5a, left panel). Since amide chemical shifts can be influenced by the local backbone conformation and hydrogenbonding geometry, these results point to discrete structural changes in the H3 and H4 regions of the b-chain of the B4.2.3 TCR upon p/MHC binding. The structural changes observed in solution in the region of the H3 helix prompted us to compare available liganded and unliganded X-ray structures of MHC-I-restricted TCR 19 . Although there are a limited number of X-ray structures of pMHC-I/TCR complexes for which comparison with the unliganded TCR may be made, careful inspection of 10 pairs of superposed structures in the H3 helix revealed no differences in backbone configuration and no consistent changes in side-chain orientations. Notably, the H3 helix does not participate in any crystallographic contacts in our study. We additionally carried out NMR titrations of B4.2.3 TCR selectively methyl-labelled in its b-chain with p/MHC to complement our amide-mapping results. Methyl groups, such as Ala Cb, Ile Cd1, Leu Cd1/Cd2 and Val Cg1/Cg2, are highly sensitive probes of side-chain packing within hydrophobic cores of proteins and are less influenced by the molecular size of the system under investigation because of their favourable relaxation properties 35 . The quality of the 2D 1 H-13 C HMQC spectrum of fully assigned AILV-methyl labelled b-chain TCR, refolded with unlabelled a-chain (Fig. 6a, red), allowed us to probe dynamic changes that occur in the P18-I10/H2-D d -bound state (Fig. 6b, blue). To transfer the methyl assignments to the bound state, we recorded a 3D H M -C M H M SOFAST NOESY data set on the P18-I10/H2-D d /B4.2.3 complex and compared it to a similar experiment recorded for the free state ( Supplementary Fig. 4e). Similar to the 15 N-TROSY results obtained for b-chain backbone amides (Fig. 5c), we measured highly reproducible slow-timescale changes in peak positions (Fig. 6a, arrows) and intensities (Fig. 6a, asterisks). In particular, we observed significant effects for the methyls of Leu42, Ile45 and Ile47, located on the b-strand of the Vb domain leading to CDR2b, as well as Val51 located on CDR2b (Fig. 6b, top panel). This corroborates the changes observed in the X-ray structure of the p/MHC-bound TCR relative to the apo structure ( Supplementary Fig. 1). Smaller but statistically significant changes were also observed for the resonances of Leu75, Leu77 and Val87 located near the core of the Vb domain (Fig. 6b, top panel). Likewise, the change in peak intensity between the free and bound states revealed effects: (1) near and on the CDRs including Val3, Ile47, Val51, Leu94, Val99; (2) near the interface between the Vb and Cb domains including Ala80, Val116; and (3) in the H3 and H4 helix regions including Ala132 and Ala190 (Fig. 6b, bottom panel, Fig. 6c).
Mutagenesis suggests a role for the TCR Cb H3 helix. Our observation that the resonances of residues in and near the Cb H3 helix of the B4.2.3 TCR undergo significant changes in the bound state led us to examine the functional role of this putative allosteric site in cell-surface expression and signalling. We targeted Cb residues that demonstrate significant NMR effects, that is, Ser127, Glu130, Asn133, Lys134 and Thr138 (Fig. 5a,c). We examined single Ala substitution mutants of each of these residues for their effects on cell-surface expression when paired with the parental a-chain. As shown in Fig. 7a, Ala mutation of Glu130 or Thr138 abolishes cell-surface expression as shown by anti-Va2 antibody staining that was indistinguishable from untransfected controls. Notably, Ala mutations at the same residues in soluble b-chain constructs result in inability to refold into functional TCR heterodimers in vitro, consistent with their placement in a critical region at the interface with the a-subunit of the TCR. The remaining three mutants show levels of surface expression identical to the parental receptor and were further analysed for p/MHC binding and TCR activation in stimulation assays. All three mutant transfectants show decreased interleukin (IL)-2 production when compared to the wild-type transfected cells with the Asn133 Ala mutant showing the greatest reduction (Fig. 7b). Comparison of the amount of IL-2 elicited by 1 mM peptide reveals a reduction from 587±45 pg in the parental transfectant to only 8 ± 4 pg in the Asn133A mutant, corresponding to a 98% decrease in IL-2 levels. Similarly, at a 1 mM peptide dose, the Lys134A mutant shows a decrease of 84% from the wild-type IL-2 levels, while the Ser127A mutation exerts a less severe effect on IL-2 levels, revealing a reduction of only 34% compared to the wild type. The severely attenuated signalling by the Asn133A and Lys134A mutants is not caused by defects in antigen recognition as all three mutants bind P18-I10/H2-D d tetramers with equal avidity and show the same rate of dissociation as the parental transfectant (Fig. 7c). These results emphasize the functional importance of TCR Cb residues Asn133 and Lys134 in p/MHC-dependent signalling and highlight the role of Glu130 and Thr138 for correct assembly, association with the CD3 co-receptor and subsequent surface expression.
To examine further whether the Asn133A substitution affected the affinity of the mutant TCR for P18-I10/H2-D d , or the stability of the TCR heterodimer itself, we prepared recombinant Asn133A TCR and compared it with the parental B4.2.3 TCR for binding to immobilized p/MHC in an SPR assay ( Supplementary Fig. 7 and Supplementary Table 1). The thermal stability of the mutated TCR was assessed using differential scanning fluorimetry, and was found to be identical to the parental protein ( Supplementary Fig. 8). The SPR-binding curves, analysed both kinetically and at steady state, revealed little or no difference in the kinetic association or dissociation rate constants (k a and k d values) as well as the calculated or steady-state determined equilibrium constants (K D values), indicating that the functional effect of the mutation is not the result of a p/MHCbinding defect, but rather reflects the putative allosteric site.

Discussion
A fundamental question in T-cell immunity is the mechanism by which p/MHC engagement by the TCR is relayed to the associated CD3 subunits to initiate intracellular signalling.  conformational changes in the signalling domains of the CD3zz homodimer 42 and the CD3e subunit 38 .
Here we use NMR to elucidate changes in TCR dynamics upon binding to the p/MHC ligand. Using a complementary combination of backbone amide and side-chain methyl probes, we extensively map ligand-induced chemical shift perturbations along the TCR b-chain. Consistent with our X-ray structures of the free and p/MHC-bound forms of the receptor, the Vb CDR loops that directly engage the p/MHC ligand show large changes in our NMR spectra. In particular, we report significant chemical shift perturbations for the amide resonances of Ser27 on CDR1b, Asp52 on CDR2b and several CDR3b residues, as well as the methyl resonances of Val51 on CDR2b and Leu94, Val99 on the hypervariable loop region. Notably, we also observe significant NMR effects on sites located near the H3 and H4 helices of the Cb domain, distal to the p/MHC recognition site. These include  ARTICLE the amides of Glu130 and Thr138 and the methyls of Ala132 and Ala190. The importance of the H3 helix for signalling is further corroborated by functional data obtained for selected mutants that inhibit signalling without affecting the expression levels of the TCR, its stability or its ability to recognize the p/MHC. The potential of TCR self-association having an impact on the observed chemical shift and intensity changes is ruled out by several lines of evidence. First, our AUC data are consistent with a fully monomeric TCR sample. Second, the 15 N and 13 C linewidths are consistent with a predominantly monomeric form of the heterodimer in solution, and remain very similar over a range of concentrations from 70 to 350 mM. Finally, the NMR spectra show the same slow-exchange process and changes over a range of TCR concentrations from 70 to 250 mM. These results conclusively support the view that the significant distal changes observed by NMR arise from an allosteric communication mechanism, as opposed to a secondary binding site or transient dimerization.
Close inspection of interactions near the H3 region in the crystal structures of the free and p/MHC-bound forms of the TCR suggests a possible explanation for the observed NMR chemical shift perturbations at this site. The interface between the Ca and Cb domains of the TCR is composed primarily of polar and charged residues that form a network of electrostatic interactions 22 . In particular, Tyr125a in the unliganded TCR interacts with Asn133b and Lys134b in one of the three molecules in the asymmetric unit (Fig. 8a) and has long-range contacts with these residues in the other two molecules. However, a displacement of the Ca domain upon p/MHC-binding promotes the formation of a Glu121a-Lys134b salt-bridge in the bound state (Fig. 8b). Notably, the side chains of Arg187b and Asp142a interact in both the free and the bound structures (Fig. 8a,b). These structural displacements are consistent with the measured chemical shift perturbations for the Glu130b, Thr138b and the broadening of the Lys134b, Arg187b backbone amide resonances (Fig. 5c), as well as the broadening of the Ala132b (on H3), Ala190b (on H4) methyl resonances (Fig. 6b). Thus, solution NMR mapping and X-ray structural data are consistent with a systematic reorganization of the Ca/Cb interface involving a remodelling of electrostatic interactions near the H3 and H4 regions of Cb. Taken together, our results provide a link between p/MHC binding at the CDR loops and dynamic changes in the TCR Cb domain that are critical for signalling 43 .
The interface between the Ca and the Cb domains plays an essential role for the stability and assembly of the TCR heterodimer, as shown both in vitro and in vivo in a recent study 44 . In accordance, residues at the Ca/Cb domain interface near the H3 helix are highly conserved across aand b-chain sequences from different species, which could result from their role in stabilizing the TCR itself as well as communicating an activation signal to CD3 (Fig. 8c). While Asn133b and Lys134b are not 100% conserved, residues with potential for hydrogen bond formation are retained at these positions, suggesting that the interface interactions with Tyr125a and Glu121a are crucial for TCR stability and CD3 signalling.
A plausible structural mechanism for the transmission of conformational changes from the CDRs to the Cb distal sites is suggested by the observation of affected sites in the Vb/Cb linker region, in particular Val116b, which shows significant line broadening in our NMR spectra. It is therefore conceivable that dynamically driven p/MHC-induced allosteric changes in the constant regions of the TCR, supported both by NMR data presented here and previous hydrogen/deuterium exchange experiments 45 , could potentiate interactions with CD3ge and de, which, according to recent NMR studies, interact only weakly with the free TCR 25,26 .
The role of the a-chain in mediating these structural changes cannot be directly addressed with the b-chain labelling scheme used here. However, similar effects on the a-chain structure are implied by the fact that most sites that undergo conformational transitions are localized near the a/b-interface. This is consistent with an earlier crystallographic study of a human TCR that revealed extensive conformational rearrangement of the BC and DE turns on Ca (ref. 46) as well as recent fluorescence 47 and hydrogen/deuterium exchange studies 45 , which showed that binding of the cognate p/MHC ligand induces long-range effects on the Ca A-B loop region. The structural basis of such changes at the TCR a/b-interface, and their impact on interactions with the CD3 co-receptor should be investigated in future studies.

Methods
Protein expression and purification. The expression and purification of P18-I10/H2-D d was performed using standardized protocols 48,49 . Briefly, cDNAs encoding the entire coding sequence of the B4.2.3 TCR aand b-chains were amplified by RT-PCR from RNA isolated from the B4.2.3 hybridoma 16 . DNAs encoding the extracellular portions of the TCR chains up to, but not including, the membrane-proximal cysteine, were cloned into pET21b (Novagen) and then expressed in LB-broth at 37°C after transformation of Escherichia coli BL21(DE3) (Novagen). (All reference to TCR V chain numbering is sequential beginning with the mature TCR chains).  Table 1. Molecular replacement solutions were readily found for the TCR and H2-D d with Phaser 52 using the murine AHIII 12.2 TCR (PDB 1LP9) and the previously determined structure of H2-D d (PDB 1DDH) as search models. For the unliganded B4.2.3 TCR, the best solution was in the P3 1 space group, but analysis of the data by Xtriage indicated pseudo-merohedral twinning with a twin fraction of 0.45, necessitating the application of twin law h, -h-k, -l during refinement in Phenix 53 . This improved R-free to 0.268 along with better electron density. The electron density map of the third heterodimer was less clear compared to the other two suggesting partial occupancy. The occupancy of chains E and F was adjusted to 0.6. Further manual increase in geometrical restraints led to improved refinement. All structure refinements were carried out in Phenix 53  double-sector charcoal-filled epon centrepieces with 12-or 3-mm path length and sapphire windows. The sedimentation process of the protein molecules was monitored using both Rayleigh interference and ultraviolet absorbance at 280 nm detection at 20°C and 50,000 r.p.m. The acquired sedimentation velocity data were analysed with SEDFIT using the c(s) sedimentation coefficient distribution approach 58 , from which the signal weighted-average sedimentation coefficient (s w ) was obtained by integration. To determine the binding affinity, the isotherm of s w as a function of macromolecular concentrations was fitted with the 1:1 hetero-dimerization model: where c A,tot and c B,tot denote the total molar concentration for A and B, c indicates the molar concentration of the free component, s denotes the sedimentation coefficient, e denotes the extinction coefficient and K AB (K D ¼ 1/K AB ) is the equilibrium association-binding constant. In the analysis, s A and s B were fixed at the experimentally determined values, while K AB and s AB were subject to optimization through nonlinear regression. The error surface projection analysis was exploited to determine the error intervals of the best-fit K D values at a 95% confidence level.
NMR sample preparation and backbone assignments. U-[ 15 N, 13 C, 2 H]-labelled b-chain B4.2.3 TCR samples for NMR were prepared by substituting the TCR b-chain growth medium with M9 minimal media in 2 H 2 O containing 2 g l À 1 13 C, 2 H glucose (Sigma #552151) and 1 g l À 1 15 NH 4 Cl. To promote overexpression, we added 1 g l À 1 2 H, 15   A control experiment without saturation of the a-chain was obtained using the same parameters with the saturation pulse turned off. The change in peak intensity was determined by calculating the ratio of I saturated /I non-saturated for each assigned b-chain amide resonance and then the ratio was normalized to 1 based on the most-intense NMR peak (T30).
ILV* and AILV-methyl sample preparation and methyl assignments.
Methyl-labelled samples of the TCR b-chain were prepared according to standardized protocols 28 using selectively labelled precursors in two distinct 13 13 Cg2 methyl resonances in the HMQC spectra recorded using the AILV and CT-HMQC spectra recorded using the ILV* sample, respectively, confirmed that the TCR structure and monomeric state were identical in the two samples.
To assign Ile, Leu and Val side-chain methyls, a 3D HMCM[CG]CBCA methyl out-and-back experiment 27 was recorded on 225 mM ILV*-labelled b-chain B4.2.3 TCR at 800 MHz, 25°C. The use of the ILV*-labelling scheme successfully generates a linear spin system needed for these experiments. Acquisition parameters were 80, 80, 1,536 complex points in the 13 C aliphatic , 13 29 . Here the use of the AILVlabelling scheme outlined above allowed the acquisition of well-resolved 13 C spectra, without the need for constant-time evolution in the indirect 13 Table 2. Full-length parental or mutant B4.2.3 TCR aand b-chain, linked via a 2A sequence 66 was cloned into the pMXs retroviral vector 67 and transfected into the Phoenix-E ecotropic packaging line using X-treme GENE 9 DNA transfection reagent (Roche). Supernatants were collected after 48 h and used to infect logarithmically growing cultures of 58a À b À , a variant of the DO11.10.7 mouse T-cell hybridoma that lacks a functional TCR 68 . Transductants expressing the B4.2.3 TCR or its mutants were stained with phycoerythrin (PE)-conjugated anti-Va2 (BD Pharmingen cat. no. 553289 used at 1/100 dilution) and magnetically enriched to 490% with anti-PE Microbeads (Miltenyi Biotech). For stimulation assays, 5 Â 10 4 parental 58a À b À or transductants expressing parental or Cb mutant B4.2.3 TCR were stimulated for 16 h with graded concentrations of P18-I10 peptide in the presence of an equal number of the BALB/c-derived B lymphoma line A20 as presenting cells in 96-well flat-bottom plates (Costar). Supernatants were diluted 1:20 for measurements of secreted IL-2 levels by ELISA (BD-Pharmingen) following the manufacturer's instructions. The results of three independent experiments were combined to obtain the means and their s.d.'s. For tetramer binding and dissociation assays, 10 6 58a À b À cells expressing parental or Cb mutant B4.2.3 TCR were incubated with 200 ng of PE-labelled P18-I10/H2-D d tetramer in a volume of 100 ml for 1 h on ice, washed with buffer (PBS containing 2% fetal calf serum and 0.1% sodium azide) resuspended in 0.5 ml buffer containing 4 mg anti-H2-D d mAb 34-5-8S to block rebinding of dissociated tetramer. At various time points during incubation at room temperature, cells were analysed by flow cytometry for residual bound tetramer.
Data availability. The refined coordinates and structure factors for the X-ray structures of free B4.2.3 TCR and P18-I10/H2-D d -bound B4.2.3 TCR have been deposited in the Protein Data Bank (www.rcsb.org) with PDB IDs 5IW1 and 5IVX, respectively. NMR assignments for the backbone and side-chain methyl chemical shifts of the b-chain of the B4.2.3 TCR have been deposited into the Biological Magnetic Resonance Data Bank (http://www.bmrb.wisc.edu) under accession number 26977. All other data are available from the corresponding authors upon reasonable request.