Dynamic allostery in the peptide/MHC complex enables TCR neoantigen selectivity

Abstract

The inherent antigen cross-reactivity of the T cell receptor (TCR) is balanced by high specificity. Surprisingly, TCR specificity often manifests in ways not easily interpreted from static structures. Here we show that TCR discrimination between an HLA-A*03:01 (HLA-A3)-restricted public neoantigen and its wild-type (WT) counterpart emerges from distinct motions within the HLA-A3 peptide binding groove that vary with the identity of the peptide's first primary anchor. These motions create a dynamic gate that, in the presence of the WT peptide, impedes a large conformational change required for TCR binding. The neoantigen is insusceptible to this limiting dynamic, and, with the gate open, upon TCR binding the central tryptophan can transit underneath the peptide backbone to the opposing side of the HLA-A3 peptide binding groove. Our findings thus reveal a novel mechanism driving TCR specificity for a cancer neoantigen that is rooted in the dynamic and allosteric nature of peptide/MHC-I binding grooves, with implications for resolving long-standing and often confounding questions about T cell specificity.

Fig. 1. TCRs distinguish between the PI3Kα neoantigen and WT peptide presented by HLA-A3 despite overlapping structures.

a Comparison of the WT peptide and a second crystallographic structure of the PI3Kα neoantigen in the HLA-A3 binding groove, demonstrating the closely overlapping structures. The neoantigen is cyan, the WT peptide is pink; this color scheme is maintained throughout all figures. Electron density of the neoantigen is from a 2F_o-F_c composite OMIT map calculated with simulated annealing, contoured at 1σ. Superimposition is by the Cα atoms of residues 1-180 of the HLA-A3 peptide binding grooves, yielding an all common atom RMSD for the peptides of 0.8 Å. b Measurement of T cell function via the degranulation marker CD107a. T cells expressing the neoantigen-specific receptor TCR4 were co-cultured with HLA-A3⁺ antigen presenting cells in the presence of increasing concentrations of neoantigen or WT peptide. Although neoantigen recognition is clear, there is no recognition of the WT peptide. Data are absolute frequencies; points and error bars are means and standard deviations from three replicates. c SPR experiments of TCR4 with the neoantigen or WT peptide/HLA-A3 complexes. Although neoantigen recognition was quantifiable, no binding was detected with the WT peptide. gp100/HLA-A2 is an irrelevant negative control complex of the peptide IMDQVPFSV presented by HLA-A2, for which no binding was also detected. d The neoantigen and WT peptide/HLA-A3 complexes bind the s3-4 scTv with identical affinities, indicating the WT sample is stable over the course of an SPR experiment, and our inability to detect TCR binding to the WT peptide/HLA-A3 in panel (c) is due to TCR discrimination. The gp100/HLA-A2 control was also wellrecognized. K_D and error values in panels (c) and (d) are the average and standard deviation of six and three replicates, respectively; titrations in both panels were performed at 4 °C for increased peptide/MHC stability.

Fig. 2. TCR binding to the neoantigen induces a large conformational change in the peptide that leads to new peptide-HLA interactions.

a Illustration of the peptide conformational change that occurs upon the binding of TCR3 and TCR4 to the neoantigen/HLA-A3 complex. The tryptophan at position 6 has flipped from aligning against the α1 helix to nestling between the peptide backbone and the α2 helix (see also Supplementary Fig. 1). b Distribution of conformational changes that occur in nonameric class I MHC complexes upon TCR binding as measured by bound vs. free RMSDs for Cα atoms and all peptide atoms. The RMSD values for the neoantigen upon recognition by TCR3 or TCR4 are shown as orange and yellow stars, respectively. The white squares in the box plots give the average, the gold boxes the interquartile ranges (IQR), lines in the gold boxes the medians, and the whiskers 1.5 × IQR. Although the changes for the backbone are only slightly above the mean, when considering all peptide atoms, the conformational change in the neoantigen is among the largest seen upon TCR recognition of nonamers. Data are from structural analysis of available PDB structures as described in the Methods. c In the TCR4 complex, the pTrp6 side chain is not contacted by the TCR. The indole nitrogen of the flipped conformation of pTrp6 forms a NH-π hydrogen bond with Trp147 of the HLA-A3 α2 helix. d In the complex with TCR3, the pTrp6 side chain is also not contacted by the TCR. The slight repositioning of the pTrp6 side chain in the complex with TCR3 distorts the interaction of pTrp6 with Trp147, although a structural water in close proximity bridges pTrp6 to Asp77 of the α1 helix.

Fig. 3. TCR recognition of the neoantigen presented by HLA-A3 is critically dependent on the flip of the tryptophan at position 6.

a The tryptophan analog Bta replaces the indole NH with a sulfur atom, eliminating the capacity of the tryptophan side chain to serve as a hydrogen bond donor. b In the structure of the free neoantigen/HLA-A3 complex, the pTrp6 side chain remains accessible to solvent. The solvent accessible surface of the indole nitrogen is blue; the surface of the carbon atoms is cyan. c Substitution of pTrp6 with Bta does not alter peptide binding to HLA-A3 as indicated by differential scanning fluorimetry. Datapoints indicate the temperature derivative of the fluorescence ratio; only every 5th datapoint is shown for clarity. T_m and error values are the average and standard deviation of four replicates. d Substitution of pTrp6 with Bta does not alter the structural properties of the neoantigen in the HLA-A3 binding grooves as shown by the crystallographic structure of the Bta6-neoantigen/HLA-A3 complex. The Bta-substituted peptide is superimposed on the unsubstituted neoantigen from the replicate structure determined here. Superimposition is via the Cα atoms of residues 1-180 of the HLA-A3 peptide binding groove, yielding an all common atom RMSD for the peptides of 0.8 Å. The orange surface shows the solvent accessibility of the Bta sulfur atom to compare with that of the indole nitrogen in panel (b). Electron density of the Bta-substituted neoantigen is from a 2F_o-F_c composite OMIT map calculated with simulated annealing, contoured at 1σ. e SPR experiments show no detectable binding of TCR4 or TCR3 to the Bta-substituted neoantigen/HLA-A3 complex, although binding to the non-substituted neoantigen complex was quantifiable. Experiments were performed at 25 °C; K_D and error values are the average and standard deviation of three replicates. f Control SPR experiments confirming binding of the Bta-substituted and non-substituted neoantigen/HLA-A3 complexes to the peptide-independent s3-4 scTv. Experiments were performed at 25 °C; K_D and error values are the average and standard deviation of four replicates. g Measurement of T cell function via production of the cytokine TNFα. T cells expressing either TCR4 or TCR3 were co-cultured with HLA-A3⁺ antigen presenting cells in the presence of increasing concentrations of Bta6-substituted or unmodified neoantigen. Although unmodified neoantigen recognition is clear, there is no recognition of the Bta6-modified version. The unmodified and Bta6-modified WT peptides at the highest concentration were included as controls. Data are absolute frequencies; points and error bars represent means and standard deviations from three replicates.

Fig. 4. Conformational sampling differs in the neoantigen and WT peptide/HLA-A3 complexes.

a Mass-weighted RMS fluctuations for each amino acid of the neoantigen and WT peptide in the HLA-A3 binding groove from 2 μs of unrestrained, fully atomistic molecular dynamics simulations. The central regions of both peptides are mobile, as is the N-terminal half of the WT peptide but not the neoantigen, consistent with the neoantigen’s more optimal position 2 anchor. b Despite high mobility, neither peptide samples the TCR-bound conformation, as shown by the RMSD of the pTrp6 amino acid relative to its position in the ternary complex with TCR4 during the simulations (data were smoothed using LOWESS; see Supplementary Fig. 13 for unsmoothed data). The value of 7.1 Å from superimposition of the TCR4-bound and free structures is indicated by the arrow. c Conformational sampling differs across the centers of the peptide backbones of the neoantigen and WT peptide as indicated by a D-score analysis comparing ϕ/ψ bond angles derived from the average peptide conformations during the two simulations. d Conformational space occupied by the pTrp6 side chain during the simulation with the WT peptide. Color density reflects degree of sampling (voxels sampled <10% of the time are excluded), values give volumes of sampled space. Note the tendency for the pTrp6 side chain to move above the peptide backbone: although it reaches over, it does not flip to the TCR-bound state as indicated by panel (b). e As in panel (d), except volume occupied by pTrp6 during the simulation with the neoantigen. Note the tendency for the pTrp6 side chain to move under the peptide backbone. f Solvent accessible surface areas of the pTrp6 side chain during the WT and neoantigen simulations. Although the average values are similar, the WT peptide samples a much wider range of values, reflecting the volumetric analyses in panels (d) and (e). White circles in the violin plots show averages, black rectangles the IQR, and vertical lines 1.5 × IQR.

Fig. 5. The extensive motions of pHis2 in the WT peptide/HLA-A3 complex lead to more interatomic interactions with pTrp6 as it moves in the binding groove.

a Volume occupied by pHis2 during the simulations with the WT peptide (left) or pLeu2 during the simulations with the neoantigen (right). Color density reflects degree of sampling (voxels sampled <10% of the time are excluded), values give volumes of sampled space. Note the greater mobility and sampled volume of pHis2 in the WT peptide. The TCR4-bound conformation of pTrp6 is colored yellow. b In the simulation with the WT peptide, the movement of pHis2 is associated with an alternate conformation for Tyr99 of HLA-A3 and formation of contacts (red dashes) between the altered Tyr99 conformer and pTrp6 in the WT peptide. Structural snapshot is representative of cluster 1 in Supplementary Fig. 4B. c The χ1 torsion angles of Tyr99 of HLA-A3 and pHis2 of the peptide during the WT peptide/HLA-A3 simulation. During the first half of the simulation, the rotation in pHis2 induces a rotation in Tyr99 (in the latter half of the simulation, the peptide N-terminus has become less recessed in the binding groove, decoupling pHis2/Tyr99 motion). d The χ1 torsion angles of pLeu2 and Tyr99 remain fixed in the simulation with the neoantigen. e Other conformations of pHis2 in the simulation with the WT peptide show contacts (red dashes) between the side chains of the histidine and pTrp6. Structural snapshot is representative of cluster 2 in Supplementary Fig. 4B. f Average distance between the centers of mass of the side chains of the position 2 amino acid, pTrp6, and Tyr99 during the simulations with the neoantigen and WT peptide. Average distances are all closer in the WT simulation. Error bars are SEM, calculated from the 2000 1 ns frames of the 2 μs simulations. Dotted lines indicate the distribution of distances calculated from the 2000 datapoints. **** = p < 0.0001. g Average counts of side chain contacts between the side chains of the position 2 amino acid, pTrp6, and Tyr99 during the simulations with the neoantigen and WT peptide. Contacts are all higher in the WT simulation. Error bars are SEM, calculated from the 2000 1 ns frames of the 2 μs simulations. Dotted lines indicate the distribution of distances calculated from the 2000 datapoints. **** = p < 0.0001. p values in panels (f) and (g) are from an unpaired t test.

Fig. 6. The flip in the neoantigen occurs via an under-peptide motional pathway resembling a limbo dance.

a Energy diagram describing the peptide conformational change as a transition from a lower to higher energy flipped state, resulting in a high energy barrier in the forward direction and a low barrier in the reverse direction. b Under-peptide motional pathway illuminated by the reverse WEMD simulations beginning with the conformation in the ternary complex with TCR4, showing pTrp6 of the neoantigen moving underneath the peptide backbone. The three under-peptide conformations are extracted from three roughly equally spaced time points of a single successful transition. c Solvent accessible surface area of the pTrp6 side chain from each frame of the 109 successful reverse WEMD transitions. The solid cyan-to-yellow curve was generated from LOWESS smoothing of the data. The value expected from an exposed side chain if pTrp6 transitioned by moving over the top of the peptide (~120 Ų) is indicated by the arrow. d RMSD of pTrp6 from the TCR4-bound conformation in SMD simulations of the peptide/HLA-A3 complexes forcing under- or over-peptide rotations. Data are shown for both the neoantigen and the WT peptide as a function of progressively larger spring constants. Only the neoantigen in an under-peptide rotation reaches the bound state, as indicated by the top row. e Number of unfavored ϕ/ψ torsion angles for all non-terminal amino acids of the neoantigen in the under- or over-peptide SMD simulations. Attempting to force an over-peptide rotation by increasing the spring constant results in greater torsional resistance. f Distance between the center of mass (COM) of the pTrp6 side chain and the position 2 side chain in the WT peptide (left) or neoantigen (right) under-peptide SMD simulations, colored by degree of van der Waals (vdW) overlap between the side chains. In the neoantigen simulation, the side chains remain distant, with little to no atomic overlap. In the WT peptide simulation, the side chains come in close proximity, with substantial overlap as the simulation progresses. Data are from the simulations with the 100 kJ/mol/mm² spring constant. Bracketed values give the average atomic overlap in Å.

Fig. 7. Experimental confirmation of differential peptide dynamics in the HLA-A3 binding groove through ¹⁹F NMR.

a One-dimensional ¹⁹F NMR spectra of the free neoantigen and the neoantigen/HLA-A3 complex. Whereas the spectrum of the free peptide shows a single sharp peak, the spectrum of the complex shows the presence of multiple states, with broad linewidths as expected for the 44 kD complex. Note that the resonance frequency of the free peptide and the major form of the neoantigen/HLA-A3 complex do not coincide. b Zoomed in view of the neoantigen/HLA-A3 complex, with relative peak areas as determined by line shape fitting (dashed lines). c Conformational Exchange Saturation Transfer (CEST) experiments demonstrating that the multiple peaks in the neoantigen/HLA-A3 complex result from the ¹⁹F spin experiencing a slow dynamic exchange between at least three different conformations. All four traces were acquired with the same receiver gain, and positions of selective irradiation are shown by red arrows. Irradiation of each of the major resonances significantly reduced the intensity of the two other peaks indicating that they arise from the same ¹⁹F spin dynamically switching between distinct environments. d Two-dimensional Exchange Spectroscopy (EXSY) experiments independently confirm all three peaks of the neoantigen complex are in conformational exchange. Black rectangles indicate positions of cross-peaks expected if the diagonal peaks represented alternative environments of ¹⁹F in dynamic equilibrium in a slow exchange regime. The mixing time of the experiment was 50 ms. e As is in panel (a) but for the free WT peptide and the WT peptide/HLA-A3 complex. Note the coinciding but different widths of the resonance of the free peptide and one of the two major peaks for the peptide/HLA-A3 complex, as well as the more complex pattern of additional peaks for the WT complex compared to the neoantigen complex shown in panel (a). f As in panel (b) but zoomed in for the WT peptide/HLA-A3 complex. Line shape fitting reveals at least eight peaks, compared to the three with the neoantigen complex in panel (b). Excluding the sharper peak at −125.12 ppm, the linewidths of these peaks are similar to those of the neoantigen complex and consistent with a 44 kD complex. g As in panel (c) but for the WT peptide/HLA-A3 complex. Selective irradiation at each resonance reduced the intensity of the others except for the resonance at −125.12 ppm. Likewise, selective irradiation at −125.12 ppm saturated this resonance but did not alter the intensity of the other peaks. The multiple broad peaks thus represent states in slow conformational exchange on the NMR time scale with exception of the –125.12 ppm resonance, which corresponds to a distinct, non-exchanging but still protein-bound population of the peptide. h As in panel (d), but for the WT peptide/HLA-A3 complex, showing cross-peaks for major resonances except for that at −125.12 ppm. The lower signal-to-noise ratio in the WT peptide/HLA-A3 sample led to poorer detection of some cross-peaks, yet still confirms conformational exchange between at least three conformational states as well as the non-exchanging character of the −125.12 ppm resonance.

Fig. 8. TCR binding rigidifies the flipped pTrp6 of the neoantigen.

a One dimensional ¹⁹F NMR spectra of the neoantigen in the TCR3-neoantigen/HLA-A3 ternary complex, along with the data from the unbound neoantigen/HLA-A3 complex for comparison (see Fig. 7a). The peak for the ternary complex is extensively broadened, as expected from its slower tumbling rate. A single shoulder aligns with the major peak of the unbound complex, reflecting the small percentage of neoantigen/HLA-A3 that is unbound in the sample. The presence of a single resonance at a different position for the TCR-bound state indicates binding-associated rigidification of the flipped pTrp6 side chain. b Molecular dynamics simulations of the TCR-neoantigen/HLA-A3 ternary complexes illustrate the rigidification of the flipped pTrp6 upon TCR binding. Results show conformational space occupied by pTrp6 during 1 μs simulations of the ternary complexes with TCR3 and TCR4. Color density reflects degree of sampling (voxels sampled <10% of the time are excluded), values give volumes of sampled space.