Mechanical versus thermal free energy landscape analysis of a coarse-grained protein

Abstract

Coarse-grained models have become a valuable tool for biomolecule modeling and the understandin of biological processes. The sacrifice of some degrees of freedom by gathering groups of atoms allows the simulation of larger systems during longer timescales, compared to the more detailed all-atom simulations. In particular, proteins become a very suitable system for coarse-grained modeling, giving rise to a large variety of models that pretend to describe its complex behavior. Perhaps, the central problem concerning proteins is that of protein folding. Huge effort has been devoted on understanding the mechanisms that drive this key process, either by means of massive simulations or by single molecule experiments, where a single protein is mechanically unfolded by means of an external force. In any case general questions arise, such as the choice of the correct reaction coordinate or the comparison between thermal unfolding (entropy based) and mechanical unfolding, where both states have similar entropy. Here we use an off-lattice coarse-grained model where each aminoacid is reduced to a bead centered in the alpha-carbon and explicit interaction terms are considered. The model accounts also for three different “flavors”: neutral, hydrophobic and hydrophilic, according to the behavior of the aminoacids in water. This model is used on a toy sequence that successfully folds into a stable structure constituted by four beta strands and whose thermal behavior has been correctly characterized. We simulate the protein by means of Langevin equation at different temperatures and applying a constant force, mimicking force-clamp experiments. The output of the simulations is analyzed by choosing different reaction coordinates, such as the native contacts, or the radius of gyration. More sophisticated techniques are also applied, mainly Principal Component Analysis and Conformational Markov Networks. Two main questions are explored. Firstly, the adequacy of each of the description levels to the specific problems we are studying. Secondly, thermal and mechanical landscapes are compared in order to find common features. Even though it is known that both processes do not have to share common characteristics, we find similar unfolding patterns in both unfolding mechanisms that can be correctly characterized by the construction of a Conformational Markov Network and its division into the basins of attraction.