Sampling Bottleneck in Validating Membrane Dynamics

Abstract

ABSTRACTMolecular dynamics (MD) simulations have become increasingly impactful in membrane biophysics because they offer atomistic resolution into the atomistic fluctuations of lipid assemblies. Validation of the simulation trajectories with experimental data is crucial for interpretation and application of MD results. As an ideal benchmarking technique, NMR spectroscopy delivers order parameters of the carbon−deuterium bond fluctuations along the lipid chains. Additionally, NMR relaxation can access lipid dynamics providing yet another point for validation of simulation force fields. Here we performed short resampling simulations of membrane trajectories to investigate the lipid CH bond fluctuations on sub-40-ps timescales to explore the local fast dynamics. We recently established a robust framework for analysis of NMR relaxation rates from MD simulations, which improves upon current approaches and shows excellent agreement of experimental and theoretical results. The calculation of relaxation rates from simulations presents a universal challenge that we addressed by hypothesizing the existence of fast CH bond dynamics that evade the analysis of simulation data with temporal resolution of 40 ps (or lower). Indeed, our results support this hypothesis confirming the validity of our solution to the sampling problem. Furthermore, we show that the fast CH bond dynamics occur on timescales at which carbon−carbon bond conformations appear nearly stationary and unaffected by cholesterol. Lastly, we discuss the correspondence to the CH bond dynamics of liquid hydrocarbons and relate their existence to the apparent microviscosity of the bilayer hydrocarbon core.STATEMENT OF SIGNIFICANCENuclear magnetic resonance data have been historically used to validate membrane simulations through the average order parameters of the lipid chains. However, the bond dynamics that give rise to this equilibrium bilayer structure have rarely been compared between in vitro and in silico systems despite the availability of substantial experimental data. Here we investigate the logarithmic timescales sampled by the lipid chain motions and confirm a recently developed computational protocol that creates a dynamics-based bridge between simulations and NMR spectroscopy. Our results establish the foundations for validating a relatively unexplored dimension of bilayer behavior and thus have far-reaching applications in membrane biophysics.