Towards Ligand Design Against Pathological Blood Clotting: A Multiscale Simulation Approach

Methods

We applied a computational strategy combining atomistic molecular simulations, machine learning-based dimensionality reduction and enhanced sampling. First, unbiased molecular dynamics simulations of wild-type prothrombin and selected mutant systems (E466A and G548A mutants) were used to describe the structural ensembles associated with the open and closed conformations. From these trajectories, structural descriptors based on inter-domain distances and contacts were extracted and used to train a Deep Linear Discriminant Analysis (Deep-LDA) model, which provided a reaction coordinate, namely collective variable (CV), able to discriminate the two conformational states. This collective variable, together with a second coordinate related to structural displacement from a reference configuration, was then employed in enhanced-sampling simulations using OPES Explore to reconstruct the free-energy landscape of the open-to-closed transition.

Results

The simulations of wild-type prothrombin revealed a free-energy landscape dominated by two major basins corresponding to the open and closed conformations, connected by a relatively well-defined transition pathway. By integrating the Boltzmann-weighted probability over the relevant regions of the DeepLDA CV space, we estimated equilibrium populations of about 33% for the open state and 67% for the closed state, in quantitative agreement with available smFRET data reporting an approximate 30% vs 70% open/closed ratio. We also reconstructed the free-energy profile using the experimental FRET-like distance and it also reproduced this trend, further supporting the validity of the simulation protocol. When the same analysis was applied to the E466A and G548A mutants, both systems showed a clear shift toward open-like conformations, with the effect being strongest for E466A. In particular, E466A displayed an estimated openstate population of 76% and closed-state population of 24%, whereas G548A showed 63% open and 37% closed. Both mutants also exhibited a more heterogeneous and fragmented conformational landscape than wild type, including additional metastable intermediates that were not prominent in the wild-type ensemble. The project was extended by three months because the enhanced-sampling strategy evolved during the course of the work. In the original proposal, the conformational transition of prothrombin was planned to be studied using well-tempered metadynamics. During the project, however, the computational protocol was refined and OPES Explore was adopted for the production enhanced-sampling simulations, combined with the machine learning DeepLDA-based CV as it provided a more robust framework describing the open/closed transition of prothrombin. This change required additional time for the setup, testing, and validation on the wild-type system first and the extended period was then used to complete the OPES Explore simulations of the clinically relevant mutant systems. We could so extract their free-energy profiles, which made it possible to quantify how E466A and G548A shift the equilibrium toward open-like conformations relative to the wild type.

Discussion

Overall, the results support the idea that the conformational equilibrium of prothrombin is a direct molecular determinant of coagulation efficiency and should be considered explicitly in multiscale models of thrombus formation. The good agreement between the wild-type simulations and experimental open/closed population ratios indicates that the chosen enhanced-sampling framework captures the essential modes of the transition. The mutant simulations further show that even single amino-acid substitutions near the kringle-1/protease interface can reshape the free-energy landscape, increase conformational heterogeneity, and stabilize open-like states that are expected to alter activation kinetics. In particular, the stronger effect observed for E466A suggests that this residue plays an important role in maintaining the stability of the closed conformation. The appearance of additional metastable intermediates in the mutant systems is also relevant because it suggests that pathogenic effects may arise not only from a simple population shift between two end states, but also from changes in the accessibility of transient conformations. As an outlook, these molecular findings would be then transferred into a convection-diffusion-reaction thrombosis model in which prothrombin would be represented as two transported species, corresponding to the open and closed conformations, so that molecular-scale conformational changes could influence thrombin generation at the fluid-dynamics macroscopic level.

Additional Project Information

DFG classification: 201-02 Biophysics
Software: Gromacs
Cluster: CLAIX