Dynamics of Protein–Ligand Interactions

Informacje na temat projektu

DynaPLIX

Identyfikator umowy o grant: 101071843

DOI

10.3030/101071843

Data podpisania przez KE 17 Lutego 2023

Data rozpoczęcia 1 Maja 2023

Data zakończenia 30 Kwietnia 2029

Finansowanie w ramach

• European Research Council (ERC)

Koszt całkowity

€ 8 721 625,00

Wkład UE

€ 8 721 625,00

Koordynowany przez

LUNDS UNIVERSITET
Sweden

Periodic Reporting for period 2 - DynaPLIX (Dynamics of Protein–Ligand Interactions)

Okres sprawozdawczy: 2024-11-01 do 2026-04-30

Podsumowanie kontekstu i ogólnych celów projektu

Proteins are vital to all processes of life. Understanding the functions of proteins has great scientific and commercial value: proteins are used as industrial enzymes, as pharmaceutical treatments, and many proteins are the targets of drugs. Current knowledge of protein function is primarily based on static structuresstructures, which have provided great insights into structure-function relationships that form the basis for protein science and protein engineering. However, proteins are not static molecules, but undergo spontaneous transitions between alternative structural states, some of which are transient conformations essentially invisible to traditional methods. These dynamical properties are known to be critically important for protein function, but high-resolution studies of dynamics have so far been conducted merely as an “add-on” following structural studies.

To change the situation, we aim to establish “integrative biomolecular dynamics” by developing methods that integrate time-resolved X-ray crystallography, nuclear magnetic resonance spectroscopy, and molecular simulations to study the motions of proteins while they carry out their function. We focus on the challenging problem of molecular recognition because it represents a poorly understood frontier in molecular science where advances are expected to have great impact. Specifically, we will address the question of how proteins bind ligands by describing with atomic resolution the entire dynamic process to reach a consistent kinetic, thermodynamic, and structural view.

We are at a point where it will be possible to develop the individual techniques required for our integrative biomolecular dynamics approach. As a team we can leverage ongoing developments in hardware and methods, while ensuring the tight integration between methods that is needed to study complex dynamical systems. We thus aim to move structural biology into a new era of protein dynamics.

Prace wykonane od początku projektu do końca okresu sprawozdawczego oraz najważniejsze dotychczasowe rezultaty

The project has established the main experimental and computational platforms required for this goal. Recombinant expression and purification have been developed for model systems including CTX-M-14 beta-lactamase, T4 lysozyme L99A, Helicobacter pylori MurI and HIV protease. Isotopically labelled CTX-M-14, T4L and MurI have been produced for NMR studies. Crystallization protocols have been optimized for CTX-M-14, T4L, and MurI, yielding crystals suitable for serial and time-resolved crystallography.
High-resolution static reference structures have been determined for CTX-M-14 and T4L in apo and ligand-bound states, and for MurI in the latter state, which define endpoints for interpreting ligand-binding trajectories. CTX-M-14 structures revealed previously unreported interactions with some ligands, while room-temperature structures of acyl-enzyme intermediates report on active-site flexibility. MurI structures provided high-quality models of the dimeric enzyme with D-glutamate in both active sites.
Time-resolved serial synchrotron crystallography (TR-SSX) and cryo-trapping have been implemented for CTX-M-14 and T4L. For T4L L99A, indole binding has been followed using endpoint structures, concentration titrations and time series from 100 ms to 2 min, revealing ligand-induced backbone changes including movements of the F-helix. Major progress has also been made in TR-SSX analysis. Unit-cell parameters from T4L L99A time series revealed coexistence of lattice populations during ligand binding, enabling population-specific processing and refinement. Ligand-geometry refinement has been improved to follow reaction-coordinate descriptors in time-resolved data. The MEROS workflow for multi-state ensemble refinement and occupancy statistics has been developed into a publishable software package that disentangles mixed-state structures, tracks binding and reaction progress, and provides error estimates.
NMR spectroscopy has provided complementary solution information. Backbone assignments were validated for T4L L99A and CTX-M-14, and CPMG relaxation-dispersion data have been acquired at two magnetic field strengths and multiple temperatures. For CTX-M-14, global analysis identified residues undergoing exchange and provided estimates of minor-state populations and exchange rates. For T4L L99A, relaxation and titration data enabled analysis of binding affinity and exchange dynamics. Nearly complete backbone assignments were obtained for indole-bound T4L, and initial encounter-complex experiments indicate weakly preferential ligand interactions with specific regions of the T4L surface.
New NMR methods and analysis tools have been developed. A new CEST experiment can measure an entire offset profile from a single 2D spectrum plus a reference experiment, greatly reducing acquisition time compared with conventional approaches. Numerical Bloch-McConnell software has been written for fitting CPMG relaxation-dispersion data and can be extended to multi-state exchange. Additional density-matrix approaches are being developed to model finite pulse lengths, power levels and intrinsic relaxation during the full CPMG period.
Computational work has advanced on several fronts. Enhanced-sampling simulations, free-energy-surface construction and Langevin dynamics are being combined to predict NMR relaxation-dispersion observables from molecular simulations. A related forward model is being developed for time-resolved crystallography, in which molecular simulations generate conformational ensembles from which electron-density maps can be calculated and compared directly with TR-SSX data. This pipeline has been accelerated from approximately 36 hours to about 30 minutes for a full analysis and made compatible with GPU acceleration.
Molecular simulations and machine learning have also been used to study ligand-binding pathways. Atomistic, biased, unbiased and coarse-grained simulations of ligand binding to T4L and CTX-M-14 have been performed. A workflow has been developed that embeds trajectories in a learned dynamical space, allowing intermediate states, pathway probabilities, kinetics and transition-state ensembles to be identified and compared across resolutions.

Innowacyjność oraz oczekiwany potencjalny wpływ (w tym dotychczasowe znaczenie społeczno-gospodarcze i szersze implikacje społeczne projektu)

The project has produced advances beyond the state of the art in experimental and computational analysis of protein dynamics. In NMR, we demonstrated that relaxation-dispersion data acquired as a function of ligand concentration can resolve induced-fit and conformational-selection pathways of protein-ligand complex formation. This yielded an essentially complete set of rate constants for a four-state binding equilibrium and showed that binding flux can be dominated by induced fit even when the initial binding affinity along the conformational-selection route is much stronger.
The new single-spectrum CEST approach is a substantial methodological advance. By measuring an entire CEST profile from one 2D spectrum plus a reference experiment, it can reduce experimental time by more than an order of magnitude compared with conventional CEST experiments. This should make CEST measurements feasible for systems with limited stability, large ligand-titration or temperature series, and experiments requiring rapid acquisition.
In time-resolved crystallography, the project has moved beyond endpoint structure determination towards quantitative descriptions of ligand binding. Population-resolved processing, ligand-geometry analysis and multi-state occupancy refinement allow heterogeneous TR-SSX data to be interpreted in terms of structural populations, binding progress and reaction coordinates. The MEROS workflow provides a general route to extracting kinetic and structural information from mixed-state crystallographic data.
Computationally, the project has developed forward models that connect enhanced-sampling simulations directly to dynamic NMR and time-resolved crystallographic observables. These models incorporate conformational heterogeneity and non-equilibrium dynamics explicitly, going beyond standard comparisons between simulations and static structures. The ability to calculate relaxation-dispersion observables and electron-density maps from simulated ensembles is a key step towards integrative modelling of protein motions during function.
The ligand-pathway classification framework also advances the state of the art. By combining coarse-grained, biased atomistic and unbiased atomistic simulations in a machine-learned embedding space, the method can identify intermediate states, assign pathways, estimate relative weights and fluxes, and compare mechanisms across molecular resolutions. This one-shot analysis provides a transferable framework for interpreting ligand recognition and for future integration of simulation-derived pathways with time-resolved structural data.