Structural Ensembles of Intrinsically Disordered Proteins using Molecular Dynamics Simulation

Abstract

Intrinsically disordered proteins (IDPs) fulfill many biological roles and are important drug targets. However, they are poorly understood relative to the wealth of structural information available for globular proteins. Their structural characterization presents a formidable challenge to both theory and experiment: the structure of an IDP must be described as a structural ensemble of many interconverting conformations. Here, we use molecular dynamics simulations to obtain structural ensembles of two IDPs: (1) a 79-residue FG-nucleoporin peptide, which is responsible for the selectivity of the nuclear pore complex and (2) a 24-residue RS-repeat peptide derived from serine/arginine-rich-splicing-factor-1, which is crucial in RNA splicing. Because force fields for polypeptides have been developed primarily to study folded proteins, it is not clear how accurately they can model disordered states. To address this issue, we performed simulations using four force fields: amber99sb∗-ildn, amber ff03w, CHARMM22∗, and CHARMM36 using replica exchange (RE) for a total of 150 microseconds per force field. The structural ensembles we obtain for both the FG and RS peptides differ markedly between force fields with respect to hydrogen bonding, radius of gyration, and secondary structure, and are sufficiently converged to make such a comparison. Importantly, secondary structure content differs more on average between force fields than between the two peptide sequences. Thus, disordered peptides appear to be particularly sensitive to force field selection, much more so than globular proteins. We compared the structural ensembles obtained with each force field to both NMR and small angle x-ray scattering data for the RS peptide by computing ensemble averages of the experimental observables. The CHARMM22∗ force field provides the most accurate description of the force fields tested. In addition, we find that CHARMM22∗ provides the fastest sampling of configuration space.