BPS2026 – Molecular dynamics simulations as an aid in prediction of T cell receptor efficacy

We used AlphaFold3 and enhanced-sampling Gaussian accelerated molecular dynamics simulations to investigate the series of 11 class II MHC-restricted T cell receptors (TCRs) in a murine cancer model (Wolf et al., 2024. Sci. Immun. , https://doi.org/10.1126/sciimmunol.adp6529). These TCRs react to an immunodominant antigen, mL9, in a MHC molecule, I-Ek. They were classified into three groups according to their ability to induce tumor remission: effective (H6, H9, H12, and H13), intermediate (H11, H14, H15, and H16), and ineffective (H7, H8, and H10). Both α- and β-chains of effective TCRs demonstrate durable, long-lived (>80% of simulation time) contacts and have highly correlated motion (correlation coefficient > 0.5) with all chains of the p-MHC. In contrast, at least one, and in some cases both, chains of the ineffective TCRs lack stable contacts with the p-MHC, particularly the peptide. To examine the basis of these durable contacts, we performed H-bond and salt-bridge analysis. Both chains of effective TCRs form persistent H-bonds (>50% of simulation time) with both chains of the MHC. By comparison, none or only one chain of ineffective TCRs forms stable H-bonds with MHC. We also observed stable salt bridges in effective TCR-p-MHC complexes that were not present for ineffective TCRs. These interactions lent TCR-p-MHC domains a coherence that the ineffective TCRs lacked. Our findings suggest that a combination of MD simulations and analysis may be useful in distinguishing therapeutically effective from intermediate and ineffective TCR clones.