Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
J Immunol. Author manuscript; available in PMC 2013 June 15.
Published in final edited form as:
PMCID: PMC3375328

Evidence for a dynamically driven T cell signaling mechanism1


T cells use the αβ T cell receptor (TCR) to bind peptides presented by major histocompatibility complex proteins (pMHC) on antigen presenting cells. Formation of a TCR-pMHC complex initiates T cell signaling via a poorly understood process, potentially involving changes in oligomeric state, altered interactions with CD3 subunits, and mechanical stress. These mechanisms could be facilitated by binding-induced changes in the TCR, but the nature and extent of any such alterations are unclear. Using hydrogen/deuterium exchange, we demonstrate that ligation globally rigidifies the TCR, which via entropic and packing effects will promote associations with neighboring proteins and enhance the stability of existing complexes. TCR regions implicated in lateral associations and signaling are particularly affected. Computational modeling demonstrated a high degree of dynamic coupling between the TCR constant and variable domains that is dampened upon ligation. These results raise the possibility that TCR triggering could involve a dynamically driven, allosteric mechanism.


T cells use the αβ T cell receptor (TCR) to bind antigenic peptides presented by major histocompatability complex proteins (pMHC) and initiate T cell signaling. Multiple mechanisms have been proposed to explain how TCR ligation triggers signaling, including TCR oligomerization, binding-induced conformational changes, and mechanical stress (1). Each of these mechanisms, alone or in combination, has been purported to alter the relationship between the TCR and its associated CD3 dimers and initiate signaling. Although results from solution measurements have been varied (2), there is evidence that ligand binding does lead to TCR oligomerization on T cell membranes, contributing to the development of higher order supramolecular complexes (3). Likewise, while large changes in TCR conformation outside of the antigen binding site have not been observed to occur upon pMHC binding, small conformational shifts have been seen in the constant domains (4). Lastly, mechanical forces augment T cell signaling in a manner that is dependent upon the association of the CD3 subunits with the TCR (5).

Oligomerization, stronger associations with neighboring subunits, and conformational changes can all be facilitated by changes in molecular flexibility. Global reductions in flexibility can help overcome the entropy barrier associated with binding and enhance complementarity at interfaces (6). More defined conformational changes that occur upon binding often take place at regions with heightened intrinsic flexibility (7). Recent work in a number of systems has demonstrated how ligand binding can alter protein flexibility across time and length scales, impacting binding behavior at distant sites in the absence of crystallographically-observed structural changes (810). This process, termed dynamically-driven allostery, has attracted attention as a unifying mechanism for the general phenomenon of allostery and is of particular interest in cases where signaling occurs in the absence of clear structural changes (reviewed in ref. 11).

As large conformational changes that could communicate pMHC binding have not been observed in the TCR, we wished to explore the potential for alterations in protein flexibility as a mechanism to communicate receptor engagement. In multiple systems and by both experiment and computation, we found that TCR ligation results in a global reduction in protein flexibility, extending from the variable into the constant domains. Regions in the TCR constant domains implicated in signaling and lateral associations on the T cell membrane were particularly affected. Altogether, our results raise the possibility that dynamically driven allostery could play a role in T cell signaling.

Materials and Methods

Protein expression and purification

Recombinant A6, DMF5, and HLA-A2 were refolded from inclusion bodies and purified as described previously (12). Constructs were stabilized either with an engineered disulfide bond (13) or a C-terminal leucine zipper (14). Synthetic peptide was purchased from Genscript.

Hydrogen/deuterium exchange

Exchange was initiated by diluting samples in at 25 °C in 25 mM HEPES, 50 mM NaCl, pH 7.4 10-fold with the same buffer made with 99.9% 2H2O. For analysis of free A6, DMF5, and Tax/HLA-A2, protein concentration after dilution was 10 μM. For analysis of bound A6, concentrations after dilution were 10 μM TCR, 40 μM pMHC, resulting in 94% occupancy. These concentrations were reversed for analysis of bound Tax/HLA-A2. For examining DMF5, concentrations after dilution were 20 μM TCR and 80 μM pMHC, resulting in 100% occupancy. For A6 with non-cognate MART-126-35/HLA-A2, concentrations after dilution were 10 μM TCR and 45 μM pMHC.

After dilution into 2H2O, 5 μL aliquots were removed at time points ranging from five seconds to 10 hours. For each aliquot, exchange was quenched at 0 °C with 100 μL of 0.1% trifluoracetic acid at pH 2.4, conditions under which the rate of hydrogen/deuterium exchange is minimal (15). Pepsin (8 μLat 1 mg/mL) was added and digestion performed for 5 minutes before freezing in liquid nitrogen. Samples were stored at −80 °C until analysis.

Identification of peptic fragments of TCR and HLA-A2

Peptic fragments were identified by LC/MS/MS. 500 ng of the pepsin digest was loaded onto a 75 μm × 15 cm C12 column. Peptides were eluted using an Eksigent nano Ultra LC into the nanospray source of an AB Sciex 5500 QTrap spectrometer. An individual survey scan was performed and up to six dependent MS/MS acquisitions were collected every cycle. Data were searched using the Paragon algorithm in thorough search mode with no enzyme specificity (16). The sequences of the TCR, HLA-A2, β2m, and potential contaminants (keratins, pepsin, and the E. coli genome)were searched. False detection rates were determined using a decoy-database strategy (17). In all cases, the relevant proteins were the top hit and all peptides passed on for further analysis identified at a 99% CI. An example LC/MS/MS spectrum is shown in Supplemental Fig. 1.

MALDI mass spectrometry

A matrix solution of 2,5-dihydroxybenzoic acid in 50:50 0.1% trifluoroacetic acid (pH 2.4) and acetonitrile was prepared and mixed 1:1 with the acidified protein on a chilled MALDI target. The target was dried under vacuum to minimize back exchange; experiments in which fully deuterated fragments were processed the same way indicated back exchange never exceeded 5%. Spectra were acquired on a BrukerAutoflex III Smartbeam MALDI-TOF mass spectrometer. The instrument was calibrated with a standard peptide mixture prior to collection of each dataset.

MALDI data processing

The centroid (mass weighted average)of the isotopic envelope for each peptide was calculated via a spreadsheet after data export. The level of backbone deuterium incorporation for each peptide was calculated as:


where <m> is the centroid mass, m0% is the mass of the undeuterated sample, and m100% is the mass of the peptide in which the backbone is fully deuterated. In all cases examined, exchange was complete by 10 minutes as shown in Fig. 1B and Supplemental Fig. 1; thus percent deuteration at 10 minutes was used for comparisons.

Figure 1
Hydrogen/deuterium exchange reveals that the A6 TCR undergoes a global reduction in flexibility upon pMHC binding

Surface plasmon resonance

Surface plasmon resonance experiments with the wild-type and high affinity variant of DMF5 were performed using a steady state assay with a Biacore 3000 instrument as previously described (12). Solution conditions were 10 mM HEPES, 150 mM NaCl, 3 mM EDTA, 0.005% surfactant P-20, pH 7.4, 25 °C.

Anisotropic Network Model calculations

Anisotropic Network Model calculations were performed using the Anisotropic Network Model server with a cutoff for interactions between α carbon atoms of 15 Å and a distance weight for interactions between α carbon atoms of 2.5 (18). Calculations were performed on the separately solved structures of A6 and its complex with pMHC (14, 19).

Structure-based computational design of a high affinity DMF5 variant

Design of the high affinity DMF5 variant was performed using the ZAFFI algorithm (20) and the structure of wild-type DMF5 in complex with the decameric MART-1 epitope (21). Briefly, mutations were simulated using the interface protocol of Rosetta 2.3 (22), using gradient-based minimization to refine backbone and torsion angles and rigid-body position before and after mutation (this was superior for reproducing binding affinities compared to the fixed backbone protocol used originally). From this analysis, the threshold of the ZAFFI filter function was raised to 0.3. Computational mutagenesis of 31 DMF5 residues resulted in 589 mutants. Ten of the highest scoring mutants were evaluated experimentally, and two that led to enhanced binding (αD26Y, βL98W) combined to generate the high affinity variant.

Results and Discussion

Ligation by pMHC globally rigidifies the A6 TCR

Although NMR remains the most powerful method for investigating protein motions, complications resulting from the size, complexity, and stability of TCRs and their complexes have limited NMR studies of TCR-pMHC interactions, requiring the use of simplified systems (23). To circumvent these challenges, we utilized hydrogen/deuterium exchange monitored by proteolysis and mass spectrometry (HDX-MS) as a tool to probe flexibility (reviewed in ref. 24). In HDX-MS, protein prepared in aqueous solution is incubated in 2H2O, digested into fragments, and backbone deuteron incorporation for each fragment determined via mass spectrometry. Increased levels of deuteron incorporation indicate that backbone amides become more accessible to exchange though greater sampling of more open, solvent-exposed states, while lower levels of incorporation indicate that amides are protected from exchange through less frequent excursions into solvent exposed states. HDX-MS has been utilized in a number of cases to explore protein motions and their changes, and has been particularly useful in characterizing low-frequency motions in samples not easily adaptable to NMR (24).

We began by performing HDX-MS on the ectodomain of the well-characterized A6 TCR in the absence and presence of its cognate pMHC ligand, the Tax peptide presented by the class I MHC HLA-A*0201 (HLA-A2). The A6 TCR is particularly amenable to this analysis, as its high affinity towards Tax/HLA-A2 allows full receptor occupancy to be achieved without requiring concentrations of pMHC that would hinder detection of TCR signals in the analysis of the exchange reactions (12). As shown in Fig. 1A, the A6 TCR showed decreased levels of deuteration upon ligation. Data for each peptide fragment were collected as a function of time out to several hours, but in all cases examined, exchange was essentially complete within 10 minutes as shown in Fig. 1B and Supplemental Fig. 1. The average deuteration of the free TCR after 10 minutes of exchange was 35%, whereas in the complex the average deuteration after 10 minutes of exchange was 18% (Fig. 1C). These values were reproducible, with a repeat experiment yielding free and bound levels of 34% and 16%, respectively. As shown in Fig. 1C, the reduced deuteration was distributed across the TCR, indicating that TCR ligation results in a global reduction in structural fluctuations, extending from the variable to the constant domains, although the extent varies throughout the molecule. Similar global reductions in flexibility upon ligand binding have been seen in other protein systems, such as the DNA binding protein BirA (25).

As expected, peptide fragments covering the A6 CDR loops showed reduced deuteration. Beyond the CDR loops, reduced deuteration was also seen for other key regions of the TCR, including a more than three-fold reduction in deuteration for a fragment spanning the β chain FG loop, a prominent structural feature that forms part of a cavity that has been proposed as a docking site for CD3εγ (26). A molecular dynamics simulation has shown that the CD3εδ subunit also rigidifies upon binding (in this case a monoclonal antibody)(27). It is not known if CD3εγ likewise rigidifies upon ligand binding; however, stronger binding of one or both of the CD3ε subunits to the TCR resulting from pMHC-induced rigidification could play a role in T cell triggering.

Although the FG loop is the most prominent feature impacted in the TCR constant domains, the reduction in flexibility includes other elements also implicated in either TCR oligomerization or interactions with CD3 subunits. This includes the α chain AB loop, which has been suggested to play a role in mediating interactions between adjacent TCRs on T cell membranes (3). Reduced AB loop flexibility could underlie a prior observation of a conformational change occurring upon TCR ligation (4), which is observed to a small extent when comparing the bound and free structures of A6 (19).

The ectodomain of the A6 TCR utilized for the data in Fig. 1 included an engineered disulfide bond near the base of the TCR constant domains to ensure heterodimer stability and facilitate proper chain pairing during protein refolding (13). An experiment with an alternate construct stabilized instead by a C-terminal heterodimeric coiled coil (14) also showed reduced deuteration upon binding at regions distant from the antigen binding site (Supplemental Fig. 1), indicating that the results in Fig. 1 are not a consequence of the method used to stabilize the receptor. A control experiment in which HDX-MS was performed with A6 before and after mixing with non-cognate ligand did not show reduced deuteration (Supplemental Fig. 1).

Ligation by pMHC globally rigidifies the DMF5 TCR

To ask if the reduction in exchange occurring upon binding was general or specific to A6, we repeated the HDX-MS experiments using the DMF5 TCR, which recognizes the MART-126-35 peptide presented by HLA-A2 (21). However, DMF5 binds more weakly than A6, and achieving high receptor occupancy required excess ligand concentrations that negatively impacted detection of TCR signals by MALDI mass spectrometry. We thus utilized a high affinity DMF5 variant, produced using the recently developed structure-guided TCR design algorithm ZAFFI (20). Surface plasmon resonance analysis of 10 DMF5 point mutants predicted to improve affinity identified two that when combined yielded a variant that bound with an affinity of 43nM, more than 250-fold tighter than wild-type (Fig. 2A).

Figure 2
The DMF5 TCR also undergoes a global reduction in flexibility upon pMHC binding

HDX-MS experiments with the high affinity DMF5 variant showed a reduction in hydrogen/deuterium exchange upon ligation that closely resembled that observed with A6 (Fig. 2B and Supplemental Fig. 1). The average deuteration of DMF5 before and after binding was 37% and 19%, respectively, with a repeat experiment yielding values of 35% and 18%. As with A6, the reduction was distributed across the TCR, and included the FG loop and the AB loop, as well as other regions implicated in TCR oligomerization or association with CD3 subunits(3). The reduction in exchange behavior for both DMF5 and A6 mapped to the structures of the TCRs is shown in Fig. 3.

Figure 3
Percent change in deuteration upon binding of each A6 and DMF5 fragment mapped to the TCR structure

Highly correlated internal dynamics of the TCR are dampened upon ligation

To further investigate changes in TCR dynamics upon pMHC binding, we performed anisotropic network model (ANM) dynamics simulations, which report on low-frequency protein motions captured by HDX-MS (reviewed in ref. 28). In ANM simulations, slow motions inaccessible to traditional molecular dynamics simulations are sampled by modeling protein structures as connected nodes whose motions are regulated by empirically-determined force constants. ANM simulations have been used to study the dynamics of complex systems as large as ribosomes, and have been particularly useful in exploring how ligand-induced changes in flexibility can give rise to allosteric effects. Several comparisons have shown good agreement between ANM data and experimental assessments of low-frequency protein motions, including data from hydrogen/deuterium exchange and NMR (28).

ANM simulations of the free and complexed A6 TCR reproduced the general features of the HDX-MS analysis, showing the global dampening of dynamics associated with ligand binding (Fig. 4A). The motions of the binding loops were extensively impacted, as were those of the FG loop and other elements of the constant domains. We next performed a cross-correlation analysis of the data to examine the dynamic coupling between regions of the TCR. In its free state, a high degree of local, fragment-level correlated motion was observed throughout the TCR, including both within and across the Vα, Vβ, Cα, and Cβ domains (Fig. 4B). These correlated local motions were significantly reduced in the ligated state. Rather than correlated motion occurring between individual elements such as the CDR and constant domains loops, in the bound state the domains themselves moved in correlated, en bloc motions (Fig. 4C). This switch from fragment to domain level correlated motion suggests the TCR is mechanically optimized to dynamically respond to pMHC ligation. The interconnected nature of the immunoglobulin fold could underlie this behavior. Inter-domain dynamics and their reduction upon binding could be communicated through the domain linkers, which in other modular proteins have been shown to be key elements in propagating signals from one domain to another (29). Another contributor could be the inherent flexibility in the relative positioning of the TCR variable domains (30), the reduction of which upon binding could impact the motions within and between the constant domains.

Figure 4
Anisotropic network model simulations indicate the TCR possesses a high degree of correlated internal motion that is reduced upon ligand binding

TCR ligation also rigidifies the HLA-A2 heavy chain and β2-microgloublin

To determine if TCR ligation alters the dynamics of the MHC, we performed HDX-MS on Tax/HLA-A2. TCR ligation resulted in an overall reduction in deuteration of the HLA-A2 heavy chain, from 27% to 16% (Supplemental Fig. 2). As expected, the α1 and α2 helices had decreased incorporation of deuterons when ligated to the TCR. However, distant regions (including sites which contact the CD8 coreceptor) also had lower levels of deuteration, indicating that, similar to the TCR, the MHC also rigidifies upon binding. β2-microglobulin also rigidified, showing a reduction in deuteration from 24% to 18%. The latter observation confirms that binding-induced changes in flexibility can be communicated across interfaces in non-covalently associated complexes.

Concluding remarks

In conclusion, we demonstrate that formation of a TCR-pMHC complex results in global TCR rigidification. Via entropic and packing effects, rigidification will promote associations with other neighboring proteins and strengthen the stability of existing complexes. New or strengthened assemblages could impact the density, positions, and accessibility of cytoplasmic domains and their associated kinases, as well as prime a complex to communicate a mechanical stress. Overall, our data raise the possibility that the TCR and its pMHC ligand are components of a dynamically-modulated, allosteric T cell signaling system.

Supplementary Material


1Supported by grants GM067079, GM084884, and RR025761 from the NIH; the Walther Cancer Research Foundation; and grant PF-11-143-01 from the American Cancer Society.


1. van der Merwe PA, Dushek O. Mechanisms for T cell receptor triggering. Nat Rev Immunol. 2011;11:47–55. [PubMed]
2. Deng L, Langley RJ, Brown PH, Xu G, Teng L, Wang Q, Gonzales MI, Callender GG, Nishimura MI, Topalian SL, Mariuzza RA. Structural basis for the recognition of mutant self by a tumor-specific, MHC class II-restricted T cell receptor. Nat Immunol. 2007;8:398–408. [PubMed]
3. Kuhns MS, Girvin AT, Klein LO, Chen R, Jensen KDC, Newell EW, Huppa JB, Lillemeier BF, Huse M, Chien Y-h, Garcia KC, Davis MM. Evidence for a functional sidedness to the αβTCR. Proceedings of the National Academy of Sciences. 2010;107:5094–5099. [PubMed]
4. Beddoe T, Chen Z, Clements CS, Ely LK, Bushell SR, Vivian JP, Kjer-Nielsen L, Pang SS, Dunstone MA, Liu YC, Macdonald WA, Perugini MA, Wilce MCJ, Burrows SR, Purcell AW, Tiganis T, Bottomley SP, McCluskey J, Rossjohn J. Antigen Ligation Triggers a Conformational Change within the Constant Domain of the αβ T Cell Receptor. Immunity. 2009;30:777–788. [PubMed]
5. Kim ST, Takeuchi K, Sun ZYJ, Touma M, Castro CE, Fahmy A, Lang MJ, Wagner G, Reinherz EL. The αβ T Cell Receptor Is an Anisotropic Mechanosensor. Journal of Biological Chemistry. 2009;284:31028–31037. [PMC free article] [PubMed]
6. Frederick KK, Marlow MS, Valentine KG, Wand AJ. Conformational entropy in molecular recognition by proteins. Nature. 2007;448:325–329. [PubMed]
7. Tobi D, Bahar I. Structural changes involved in protein binding correlate with intrinsic motions of proteins in the unbound state. Proceedings of the National Academy of Sciences. 2005;102:18908–18913. [PubMed]
8. Zhang J, Chalmers MJ, Stayrook KR, Burris LL, Wang Y, Busby SA, Pascal BD, Garcia-Ordonez RD, Bruning JB, Istrate MA, Kojetin DJ, Dodge JA, Burris TP, Griffin PR. DNA binding alters coactivator interaction surfaces of the intact VDR–RXR complex. Nat Struct Mol Biol. 2011;18:556–563. [PMC free article] [PubMed]
9. Popovych N, Sun S, Ebright RH, Kalodimos CG. Dynamically driven protein allostery. Nat Struct Mol Biol. 2006;13:831–838. [PMC free article] [PubMed]
10. Tzeng SR, Kalodimos CG. Dynamic activation of an allosteric regulatory protein. Nature. 2009;462:368–372. [PubMed]
11. Smock RG, Gierasch LM. Sending Signals Dynamically. Science. 2009;324:198–203. [PMC free article] [PubMed]
12. Davis-Harrison RL, Armstrong KM, Baker BM. Two Different T Cell Receptors use Different Thermodynamic Strategies to Recognize the Same Peptide/MHC Ligand. Journal of Molecular Biology. 2005;346:533–550. [PubMed]
13. Boulter JM, Glick M, Todorov PT, Baston E, Sami M, Rizkallah P, Jakobsen BK. Stable, soluble T-cell receptor molecules for crystallization and therapeutics. Protein Eng. 2003;16:707–711. [PubMed]
14. Ding YH, Baker BM, Garboczi DN, Biddison WE, Wiley DC. Four A6-TCR/peptide/HLA-A2 structures that generate very different T cell signals are nearly identical. Immunity. 1999;11:45–56. [PubMed]
15. Englander SW, Sosnick TR, Englander JJ, Mayne L. Mechanisms and uses of hydrogen exchange. Current Opinion in Structural Biology. 1996;6:18–23. [PMC free article] [PubMed]
16. Shilov IV, Seymour SL, Patel AA, Loboda A, Tang WH, Keating SP, Hunter CL, Nuwaysir LM, Schaeffer DA. The Paragon Algorithm, a Next Generation Search Engine That Uses Sequence Temperature Values and Feature Probabilities to Identify Peptides from Tandem Mass Spectra. Molecular & Cellular Proteomics. 2007;6:1638–1655. [PubMed]
17. Balgley BM, Laudeman T, Yang L, Song T, Lee CS. Comparative Evaluation of Tandem MS Search Algorithms Using a Target-Decoy Search Strategy. Molecular & Cellular Proteomics. 2007;6:1599–1608. [PubMed]
18. Eyal E, Yang LW, Bahar I. Anisotropic network model: systematic evaluation and a new web interface. Bioinformatics. 2006;22:2619–2627. [PubMed]
19. Scott DR, Borbulevych OY, Piepenbrink KH, Corcelli SA, Baker BM. Disparate Degrees of Hypervariable Loop Flexibility Control T-Cell Receptor Cross-Reactivity, Specificity, and Binding Mechanism. Journal of Molecular Biology. 2011;414:385–400. [PMC free article] [PubMed]
20. Haidar JN, Pierce B, Yu Y, Tong W, Li M, Weng Z. Structure-based design of a T-cell receptor leads to nearly 100-fold improvement in binding affinity for pepMHC. Proteins: Structure, Function, and Bioinformatics. 2009;74:948–960. [PMC free article] [PubMed]
21. Borbulevych OY, Santhanagopolan SM, Hossain M, Baker BM. TCRs Used in Cancer Gene Therapy Cross-React with MART-1/Melan-A Tumor Antigens via Distinct Mechanisms. J Immunol. 2011;187:2453–2463. [PMC free article] [PubMed]
22. Kortemme T, Baker D. A simple physical model for binding energy hot spots in protein-protein complexes. Proc Natl Acad Sci U S A. 2002;99:14116–14121. [PubMed]
23. Varani L, Bankovich AJ, Liu CW, Colf LA, Jones LL, Kranz DM, Puglisi JD, Garcia KC. Solution mapping of T cell receptor docking footprints on peptide-MHC. Proceedings of the National Academy of Sciences. 2007;104:13080–13085. [PubMed]
24. Marcsisin SR, Engen JR. Hydrogen exchange mass spectrometry: what is it and what can it tell us? Analytical and bioanalytical chemistry. 2010;397:967–972. [PMC free article] [PubMed]
25. Laine O, Streaker ED, Nabavi M, Fenselau CC, Beckett D. Allosteric Signaling in the Biotin Repressor Occurs via Local Folding Coupled to Global Dampening of Protein Dynamics. Journal of Molecular Biology. 2008;381:89–101. [PMC free article] [PubMed]
26. Kim ST, Touma M, Takeuchi K, Sun ZYJ, Dave VP, Kappes DJ, Wagner G, Reinherz EL. Distinctive CD3 Heterodimeric Ectodomain Topologies Maximize Antigen-Triggered Activation of αβ T Cell Receptors. The Journal of Immunology. 2010;185:2951–2959. [PMC free article] [PubMed]
27. Martinez-Martin N, Risueno RM, Morreale A, Zaldivar I, Fernandez-Arenas E, Herranz F, Ortiz AR, Alarcon B. Cooperativity Between T Cell Receptor Complexes Revealed by Conformational Mutants of CD3{varepsilon} Sci Signal. 2009;2:ra43. [PubMed]
28. Bahar I, Lezon TR, Yang LW, Eyal E. Global Dynamics of Proteins: Bridging Between Structure and Function. Annual Review of Biophysics. 2010;39:23–42. [PMC free article] [PubMed]
29. Ma B, Tsai C-J, Haliloğlu T, Nussinov R. Dynamic Allostery: Linkers Are Not Merely Flexible. Structure. 2011;19:907–917. [PubMed]
30. Gagnon SJ, Borbulevych OY, Davis-Harrison RL, Turner RV, Damirjian M, Wojnarowicz A, Biddison WE, Baker BM. T Cell Receptor Recognition via Cooperative Conformational Plasticity. Journal of Molecular Biology. 2006;363:228–243. [PubMed]