TCR Scanning of Peptide/MHC through Complementary Matching of Receptor and Ligand Molecular Flexibility

T cells use the αβ TCR to recognize peptide Ags presented by MHC proteins (pMHC) on the surfaces of APCs. Crystallographic structures have demonstrated that TCR engagement of pMHC often proceeds with conformational changes in TCR CDR loops, peptides, and MHC proteins (reviewed in Ref. 1, see Refs. 2–6 for specific examples). However, beyond their influence on binding affinities and kinetics (7), the impact of these conformational changes is not fully understood. In general, protein conformational changes are associated with enhanced flexibility, as the lower energy barriers that facilitate structural alterations translate into faster rates of motion (8). For TCRs, CDR loop motion has been directly linked with receptor cross-reactivity, allowing the loops to optimize structural complementarity with different ligands (9). Although the magnitudes of conformational changes and the underlying motional properties can vary (4, 10, 11), TCR loop motion is believed to be an important component of the process through which TCRs scan for compatible ligands on the surfaces of APCs (12–14).

Peptide and MHC conformational changes have generally received less attention than those occurring in TCRs, but do occur, particularly for peptides (e.g., Refs. 5, 10, 11, 15). In some cases, conformational changes in both TCR and pMHC occur upon binding, a process we have termed conformational melding (6, 16). The occurrence of flexibility in both receptor and ligand adds to the complexity of the TCR–pMHC interaction and raises structural and energetic questions about how a TCR and pMHC can productively engage if regions of both molecules are moving and sampling conformations with varying degrees of compatibility (1).

Although a myriad of approaches have been used to study the motional properties of TCRs and pMHC complexes, including computation and various forms of spectroscopy (e.g., Refs. 5, 9, 17–21), nuclear magnetic resonance (NMR) is advantageous in that it can yield experimental insight into motion at atomic resolution without requiring the introduction of potentially interfering labels (22). When combined with structural information, NMR can yield information unattainable by other techniques.

To gain new insight into how receptor and ligand motion impacts TCR recognition of pMHC, in this study, we used NMR to examine the interaction between the murine 2C TCR and the QL9 peptide presented by H-2L^d (L^d). The 2C-QL9–L^d interaction is an archetypal TCR–pMHC interaction, studies of which have provided key details about the structural and physical nature of TCR recognition (23–27). Conformational changes occur in both the TCR and pMHC upon formation of the 2C-QL9–L^d complex (24, 28, 29), which, together with available immunological, biochemical, and biophysical data, make it an ideal system for investigating TCR and pMHC flexibility at an atomic level.

In examining 2C recognition of QL9–L^d, we found that the CDR loops of the 2C TCR generally undergo a reduction in flexibility upon binding, confirming prior inferences from various crystallographic and binding experiments. However, a surprising exception was seen for residues in the CDR3β loop, which retains significant mobility in the TCR–pMHC complex. Reciprocal analysis of the pMHC indicated the region of the peptide that interacts with CDR3β is similarly mobile in both the free pMHC as well as the TCR–pMHC complex. Remarkably, the rates at which CDR3β and the peptide move are similar in both the free proteins and the complex. The matching of receptor and ligand flexibility and its persistence in the TCR–pMHC complex provides a solution for the structural and energetic challenges posed when two flexible molecules engage and is evocative of behavior recently described in interfaces formed with other proteins that engage multiple targets (30). Considering the high incidence of TCR and pMHC structural rearrangements that have been observed crystallographically and their links to protein flexibility (8), we suggest that dynamic complementarity, or the matching of conformational exchange in TCR and pMHC, is an element of how TCRs finely discriminate among ligands for those that can bind with sufficient kinetic and thermodynamic stability to enable T cell signaling.

Unlabeled or labeled single-chain 2C (sc2C) and single-chain H-2L^d (scL^d) was refolded from bacterially expressed inclusion bodies grown with rich or ¹⁵N-labeled M9 minimal media as previously described (28, 31–33). Protein was purified via ion-exchange and size-exclusion chromatography. Unlabeled or labeled QL9 peptide was chemically synthesized (Genscript or Tufts University Core Facility). All samples for NMR were exchanged into a buffer containing 25 mM Bis-Tris (pH 7.1), 25 mM NaCl, 0.03% NaN₃, and 90% H₂O/10% D₂O.

Spectra were recorded at 298 K (25°C) on a Bruker Advance II 800 MHz spectrometer equipped with a TCI cryogenic probe (Bruker). Spectra were processed with Topspin 2.1 pI6 and analyzed with the Sparky and CcpNmr software packages (34). Spectra from Heteronuclear Single-Quantum Correlation (HSQC), HNCO, HNCA, CBCA(CO)NH, and HNCACB experiments (35–39) were used to sequentially assign the 2C TCR. Assignments for the TCR have been submitted to the Biological Magnetic Resonance Data Bank (identification number 19239).

Two NMR titrations were performed: [¹⁵N]-sc2C titrated with unlabeled QL9-scL^d and [¹⁵N]–QL9-scL^d titrated with unlabeled sc2C. [¹H, ¹⁵N]-HSQC spectra were recorded for each titration point at 298 K on a Bruker Advance 800 II MHz spectrometer (Bruker). In each titration, the concentration of ¹⁵N-labeled protein used was 300 μM. The concentration range of the unlabeled binding partner ranged between 0 and 1.5 times the concentration of the ¹⁵N-labeled protein, yielding >95% saturation at the highest concentration used. Peptide dissociation from scL^d was ignored as all concentrations of QL9-scL^d were several thousand–fold higher than the 7 nM K_D the QL9 peptide has for L^d (27).

Quantitative analysis of the NMR line shape data for Gly⁹⁷β of the TCR and Phe⁷ of the peptide was performed as described in Greenwood et al. (40). These residues were selected as they showed the greatest complexity in the titrations and were still compatible with the constraints required for line-shape analysis (well resolved or multiple peaks that shift in at least one dimension without the presence of intervening resonances). Briefly, one-dimensional ¹H line shapes were extracted from [¹H-¹⁵N]-HSQC spectra recorded during the titrations using the BiophysicsLab 1D NMR Python extension in Sparky and composited for fitting. For compositing, one-dimensional slices through the center of each two-dimensional peak were normalized by peak volume. Line shape fitting was performed using the BiophysicsLab software package (currently, Integrative Data Analysis Platform; http://kovrigin.chem.mu.edu/IDAP/) (41). Spectra were first normalized within their respective titration series. BiophysicsLab was then used with the Bloch-McConnell equations (42) modified to reflect the various binding models. The population matrix, which describes the populations of species present, was formulated based on the total concentration of TCR and pMHC and the equilibrium constants that describe conversion between the various states. The corresponding matrix of rate constants was formulated from the rate equations governing exchange between the states. Separate matrices were constructed for the chemical shifts for each state, the apparent transverse relaxation rate constants, and the frequency variables. Data were fit using a Newton interior point method, with convergence and fit quality evaluated by the χ² statistic.

For each fit, a global minimum was ensured through the use of a grid search, in which the initial guess for each parameter was systematically varied over a 10–20-fold range (8–10 initial guesses for each parameter). The overall K_D in each fit was constrained to the value of 2.20 μM previously determined using isothermal titration calorimetry (43). Parameter space was explored by refitting with individual parameters fixed at values near their best fit value, whereas other parameters were floated, and recording the χ² statistic for each new fit. Error space was further explored by varying the constraining K_D within the error range of the calorimetrically determined value and repeating the fitting procedure.

We began by titrating [¹⁵N]-labeled TCR with unlabeled pMHC using [¹H-¹⁵N]-HSQC NMR spectra collected as a function of pMHC concentration to monitor binding. Because the chemical shifts of the [¹H] and [¹⁵N] nuclei are sensitive to changes in molecular environment, the advantage of this experiment is that each backbone NH serves as a potential binding probe. This advantage was recently demonstrated through the use of chemical shift perturbation to identify residues within TCR–pMHC binding interfaces (31). However, because motion also alters molecular environment, careful analysis of [¹H-¹⁵N]-HSQC spectra can provide insight into flexibility and changes in flexibility that occur as binding proceeds. In ideal cases, quantitative analyses can provide information on the kinetics of conformational exchange and how this influences recognition (40–42, 44, 45).

NMR experiments were facilitated through the use of recently developed single-chain, minimized variants of 2C and L^d (referred to as sc2C and scL^d, respectively), which, because of their smaller size, reduced experimental complexity and improved the signal compared with the full-length molecules (28, 31). The single-chain variants and their full-length counterparts interact with similar affinities, kinetics, and structural topologies (23, 24, 28), indicating that minimization of the proteins has not significantly altered the recognition properties of either TCR or pMHC. HSQC spectra were well-dispersed and consistent with a stable, well-folded protein (Supplemental Fig. 1A). NMR assignments were performed using standard sequential strategies as described in the 2Materials and Methods. Concentrations in the titrations were varied to permit collection of data over a range of saturation, from zero to >95%.

We focused our studies on residues in the various CDR loops of the 2C TCR, following changes in [¹H-¹⁵N] cross-peaks as the TCR was titrated with pMHC. In most cases, the resonances for the residues of the CDR loops of the unliganded TCR were broad, coalescing into sharper, more defined peaks in the TCR–pMHC complex. Data illustrating this behavior for each TCR residue that contacts peptide or MHC are shown in Fig. 1 and Supplemental Fig. 1B. NMR peak broadening can be attributed to conformational exchange, in which intrinsic molecular flexibility moves the backbone NH between different environments. Peak broadening is also expected upon the formation of a larger complex as the rate of overall molecular tumbling decreases. As our observations are to the contrary, we interpret our results to indicate that the backbones of the various CDR loops are structurally loose (or mobile) in the free TCR and that this mobility is restricted in the TCR–pMHC complex. A role for loop motion in Ag recognition by the 2C TCR has been inferred from biophysical binding data (27), as well as the various crystal structures of the free and bound receptor (2, 24, 29), which demonstrate that shifts in backbone conformation as large as 7 Å are required for the TCR to engage ligand (Supplemental Fig. 2A) (note that reduced amide proton exchange with solvent could also lead to sharper peaks in the complex; however, the solvent accessible surface areas of the backbone amides of the various residues in Fig. 1 and Supplemental Fig. 1 are not always reduced, and in some cases even increase upon binding).

Unlike other positions, the triple-glycine motif of the sc2C CDR3β loop (Gly⁹⁶, Gly⁹⁷, and Gly⁹⁸; see Supplemental Fig. 1B) showed greater complexity in the NMR titrations. For these residues, the broad resonances in the free TCR evolved into additional cross-peaks upon titration with pMHC, with discrete peaks persisting even when the TCR was saturated. This behavior was most evident for the backbone nitrogen of Gly⁹⁷ at the apex of CDR3β (Fig. 2A, 2B). The simplest interpretation of these data is that at least two distinct bound states are formed with Gly⁹⁷β, with the tip of the CDR3β loop moving between the different conformations on a slow timescale (as aromatic ring currents can alter proton chemical shifts, the existence of multiple peaks for Gly⁹⁷β in the TCR–pMHC complex could also arise from motion of the side chain of Phe⁷ of the QL9 peptide, which is adjacent to Gly⁹⁷ of CDR3β, as shown in Fig. 2A. However, given the packing in the 2C-QL9-L^d interface, any Phe⁷ ring flips or substantial motion would still necessitate compensatory motion in CDR3β to avoid steric clashes).

To gain further insight into the behavior of Gly⁹⁷β, we extracted one-dimensional ¹H line shapes from the [¹H-¹⁵N]-HSQC titration data for Gly⁹⁷β. By extracting the rate constants that underlie the chemical shift data, quantitative analysis of such NMR line shapes can yield the kinetics that govern the movement of NMR-visible nuclei between bound, free, and other conformational states (40–42, 44, 45). NMR line-shape analysis has been recently used, for example, to examine the isomer-selective binding mechanism of the Pin1 cis-trans prolyl isomerase (45) and identify the site-specific mechanisms through which the Src homology 2 domain of PI3K binds flexible peptides (44). In determining kinetics, the line shape data as a function of concentration are fit to modified Bloch equations that incorporate the mass action laws and corresponding rate constants for the model under investigation.

The simplest model consistent with the spectra for Gly⁹⁷β is one in which the backbone can move between two conformations in the free state of the TCR, both of which are competent to bind ligand and remain accessible in the TCR–pMHC complex (Fig. 2C). The line shape data as a function of concentration were well fit by this model (Fig. 2D). The kinetic parameters indicated that movement between the two conformations in the free TCR occurred with rate constants near 0.2 s⁻¹, with the two conformations equally populated (Table I). The rates of motion were elevated in the complex, but neither conformation was significantly favored, indicating that neither of the states can be considered a binding intermediate on the path toward a final complex. The rates of TCR binding and dissociation were similar for both conformations and close to those determined previously by surface plasmon resonance (23, 28). Overall, the results suggest that Gly⁹⁷β moves slowly in both the free and bound TCR, sampling discrete conformations without significantly impacting TCR binding affinity or kinetics.

In the 2C-QL9-L^d ternary complex, Phe⁷ of the QL9 peptide interacts with Gly⁹⁷β and together with Asp⁸ forms the majority of the peptide contacts with the TCR (Fig. 2A). To complement the Gly⁹⁷β data, we examined Phe⁷ using a site-specifically ¹⁵N-labeled QL9 peptide, titrating unlabeled sc2C TCR into a solution of ¹⁵N-labeled QL9–scL^d. In the spectrum of the unliganded QL9–scL^d complex, two cross-peaks were observed for the amide nitrogen of Phe⁷, indicating the presence of two distinct peptide conformations (Fig. 3A). The detection of two conformations is consistent with the crystallographic structure of QL9-L^d (43), in which the two molecules in the asymmetric unit presented the peptide differently (Supplemental Fig. 2B). One peptide conformation was close to that observed in the 2C-QL9-L^d ternary complex (24). Although the quality of the electron density was poorer, the second conformation would require complementary adjustments in CDR3β to avoid steric clashes and electrostatic repulsion, as shown in Supplemental Fig. 2C (as noted above, due to the potential influence of the Phe⁷ aromatic ring, the NMR experiments cannot clearly differentiate between Phe⁷ backbone and side-chain motion, although, as indicated by the crystallographic data, both may be involved).

Upon titration with the 2C TCR, both cross-peaks for Phe⁷ shifted, mimicking the behavior of Gly⁹⁷ of CDR3β and suggesting that, as with CDR3β, Phe⁷ remains mobile in the TCR–pMHC complex (Fig. 3A). We thus performed a similar NMR line-shape analysis for the Phe⁷ data as was performed for Gly⁹⁷β. We used the same two-state conformational exchange model as used for Gly⁹⁷β, except that motion was located in the peptide rather than TCR (Fig. 3B). Intriguingly, this analysis yielded values similar to those determined for Gly⁹⁷β (Fig. 3C and Table I). The concordance between the data for Gly⁹⁷β and Phe⁷ suggests that the two peptide conformations seen during titration with TCR reflect slow movement of the peptide and the CDR3β loop in the TCR–pMHC complex. As modeled in Fig. 3E, this could involve synchronous movement between the various CDR3β and QL9 peptide conformations that have been observed crystallographically.

A more complex spectrum consisting of at least five cross-peaks was seen for the unliganded QL9–scL^d complex when the nitrogen of Asp⁸ of the peptide was labeled (Fig. 3D), consistent with the peptide sampling multiple conformations on a slow timescale. Upon titration with the TCR, however, the complexity was reduced to two resonances of nearly equal volume. The complexity of the spectrum in the free QL9–scL^d complex prohibited NMR line-shape analysis. However, the behavior is indicative of a relatively mobile peptide whose motions are reduced but not eliminated in the TCR–pMHC complex.

Conformational flexibility in TCR CDR loops has been linked to Ag identification and engagement, with loop motions in the free TCR permitting the optimization of structural and chemical complementarity with target pMHC complexes. This has been most clearly demonstrated with the human TCR A6, for which the mobility of the hypervariable loops, and CDR3β in particular, allows the receptor to engage a myriad of targets that differ in the center of the peptide (9). Considered alongside the variation in CDR loop conformation seen between the free and various bound structures of the 2C TCR, our observation that the majority of the CDR loops of the 2C TCR undergo a reduction in backbone flexibility is consistent with a general role for TCR loop flexibility in facilitating Ag recognition.

However, the observation of residual flexibility for the 2C CDR3β loop in the TCR–pMHC complex was unanticipated. Particularly striking was our finding that at least two slowly exchanging and essentially equally populated conformations exist for Gly⁹⁷β of CDR3β. Although the presence of flexibility might be expected from a loop enriched in glycine, in this case, the observation was puzzling, as in the 2C-QL9-L^d crystal structure the majority of contacts to the QL9 peptide are formed by this region of CDR3β. Mutations in this region and substitutions to contacted peptide residues have moderate to significant effects on recognition (46, 47). How can a key region of the TCR binding site maintain substantial mobility within a TCR–pMHC complex?

Insight was provided by the observations on the pMHC ligand, which revealed that the region of the peptide that interfaces with CDR3β is also mobile, both in the free pMHC as well as the complex. This was particularly notable, as motion occurring within pMHC complexes has not been widely considered in structural immunology. Remarkably, though, Phe⁷ of the peptide was found to move with kinetics that closely match those of Gly⁹⁷β. The data suggest that in the TCR–pMHC complex, the CDR3β loop and the peptide synchronously interconvert between compatible conformations as modeled in Fig. 3E (although correlated interconversion between structurally compatible conformational states as shown in Fig. 3E is the simplest interpretation of our data, our experiments cannot rule out the existence of other states, although given the proximity of Gly⁹⁷β and Phe⁷, such states would be less stable due to steric and electrostatic repulsion, as shown in Supplemental Fig. 2C).

Intriguingly, our observations in the sc2C-QL9–scL^d complex are not wholly unique. Mobility in interfaces formed by flexible proteins has been observed in other systems, most notably in complexes formed by proteins that engage multiple targets (30). This behavior is believed to serve two purposes. First, it reduces the entropic penalty associated with fixing mobile regions. Second, if regions that interact possess complementary motions within the interface, structural and chemical complementarity can be maintained as the proteins move, preventing the loss of favorable contacts and/or the introduction of unfavorable ones. Our observations thus highlight a mechanism of reciprocal, dynamic complementarity between the receptor and ligand that enhances the stability of what would otherwise be a weak TCR–pMHC complex.

It is further striking that the rates at which CDR3β and the peptide move are similar in the free TCR and free pMHC. We suggest this is not coincidental: the fact that the peptide and CDR3β move with similar rates, likely sampling compatible conformations as shown in Fig. 3E, indicates that the two molecules are ideally poised to interact from both a structural and dynamic perspective. As discussed above, ligands with mismatched motional properties are likely to be a poorer ligand for the receptor. The matching of conformational exchange between receptor and ligand thus provides a mechanism for facile engagement of QL9-L^d by the 2C TCR. Put more simply, as the dynamic 2C TCR scans for ligands, those that can match both its structural and motional properties are likely to bind better.

The influence of matched structural and motional properties on receptor/ligand selectivity is diagrammed schematically in Fig. 4. Fig. 4A illustrates our observations with sc2C and QL9–scL^d: the TCR and pMHC move between compatible conformations at similar rates and retain their mobility in the TCR–pMHC complex. Due to the structural complementarity between the different receptor and ligand configurations, there are two ways in which the complex can form. Fig. 4B illustrates formation of a complex from molecules possessing mismatched motional properties and for which motion is restricted in the TCR–pMHC complex. Compared to the case in Fig. 4A, the resulting complex is of weaker thermodynamic stability. This results from the need for structural adjustments to optimize complementarity, as well as a greater entropic penalty due to the restriction of motion in the complex. The reduced stability of the complex results in both weaker binding and faster dissociation.

Should we expect these observations to be unique to 2C and QL9-L^d? A number of studies have demonstrated mobility of both peptides in MHC binding grooves and TCR CDR loops (5, 9, 17, 21, 48–50), and other studies have directly linked crystallographically observed conformational changes to TCR and pMHC motion (5, 9). Moreover, conformational melding, or the observation of conformational changes occurring in both pMHC and TCR upon binding, has been observed in multiple cases. We thus propose that the matching of structural fluctuations between receptor and ligand is one element of the process through which TCRs scan for and identify the most compatible ligands from a myriad of potential targets. As a range of structural changes and motions over different timescales have been observed for TCRs and pMHC complexes (4, 9, 10, 17–19, 51), this strategy would certainly be used to varied extents with different interactions. For example, more rigid TCRs may rely less on matched motions. Further, flexibility in a free receptor does not necessarily require retained motion in the complex, as entropic penalties associated with reducing motions could be offset by energetic gains elsewhere in an interface. Nonetheless, the strategy identified in this study is generalizable and readily superimposable onto other mechanisms that influence how TCRs discriminate among ligands for those that can bind sufficient well to enable T cell signaling.

We thank Kristina Davis of the Notre Dame Center for Research Computing for assistance in generating Fig. 4.

The authors have no financial conflicts of interest.