Resolving Membrane-Protein Complexes with Joint Neutron Reflectivity and Molecular Simulation Refinement

Abstract

The structural characterization of membrane-associated proteins on fluid lipid bilayers remains a challenge in structural biology. Neutron reflection (NR) from engineered planar interfaces has emerged as a method that provides sub-nanometer resolution of protein-membrane complexes in a functional lipid environment. Interpretation of NR data gives a scattering length density (SLD) envelope, which describes the distribution of protein density normal to the membrane. A full interpretation of the SLD profile requires knowledge of the internal structure of the protein. However, for intrinsically disordered protein regions, multiple conformational states contribute to an SLD density profile that cannot be mapped back to a single structure. A complete molecular interpretation then requires sophisticated modeling approaches. Molecular Dynamics (MD) simulations are a well-established tool for studying the structure and dynamics of biomolecular systems, including protein-membrane complexes. We previously used MD simulations to refine NR results beyond the limits of scattering resolution. However, such studies are often plagued by long equilibration times. To better interpret the limited structural information from NR, we designed an algorithm that implements a potential to bias MD simulations toward configurations that reproduce the experimental results. The potentials are determined by comparing experimental SLD distributions with time-averaged profiles from the molecular simulations. These are converted into steering forces that become weaker as the resulting simulation profiles match the experimental results more closely. This results in guiding the structure toward the desired configuration, rather than rigidly confining it to the experimental density. Here we show applications of this method to model peptides and small protein systems.