Structural analysis of intrinsically disordered proteins: computer atomistic simulation

Intrinsically disordered proteins (IDPs) are biomolecules that do not have a definite 3D structure; their role in the biochemical network of a cell relates to their ability to switch rapidly among different secondary and tertiary structures. For this reason, applying a simulation computer program to their structural study turns out to be problematic, as their dynamical simulation cannot start from a known list of atomistic positions, as is the case for globular proteins that do crystallize and that one can analyse by X-ray spectroscopy to determine their structure. We have established a method to perform a computer simulation of these proteins, apt to gather statistically significant data on their transient structures. The only required input to start the procedure is the primary sequence of the disordered domains of the protein, and the 3D structure of the ordered domains, if any. For a fully disordered protein the method is as follows: (a) The first step is the creation of a multi-rod-like configuration of the molecule, derived from its primary sequence. This structure evolves dynamically in vacuo or in an implicit model of solvent, until its gyration radius - or any other measure of the overall configuration of the molecule - reaches the experimental average value; at this point, one may follow two different paths. (b1) If the study focuses on transient secondary structures of the molecule, one puts the structure obtained at the end of the first step in a box containing solvent molecules in explicit implementation, and a standard molecular dynamics simulation follows. (b2) If the study focuses on the tertiary structure of the molecule, a larger sampling of the phase space is required, with the molecule moving in very large and diverse regions of the phase space. To this end, the structure of the IDP is let evolve dynamically in an implicit solvent using metadynamics, an algorithm that keeps track of the regions of the phase space already sampled, and forces the system to wander in further regions of the phase space. (c) One can increase the accuracy of the statistical information gathered in both cases by fitting, where available, experimental data of the protein. In this step one extracts an ensemble of ’best’ conformers from the pool of all configurations produced in the simulated dynamics. One derives this ensemble by means of an ensemble optimization method, implementing a genetic algorithm. We have applied this procedure to the simulation of tau, one of the largest fully disordered proteins, which is involved in the development of Alzheimer’s disease and of other neurodegenerative diseases. We have combined the results of our simulation with small-angle X-ray scattering experimental data to extract from the dynamics an optimized ensemble of most probable conformers of tau. The method can be easily adapted to IDPs entailing ordered domains.