Energetics of protein interactions in the membrane via computer simulations

Protein-lipid interactions underpin the biological activities in cell membranes. However, the energetic aspect of these interactions is notoriously difficult to study. Here, I performed a number of enhanced sampling simulations on biologically-interesting systems using different collective variables to provide converged estimates of protein binding affinities in the membrane. Replica exchange umbrella sampling (REUS) simulations were performed using a distancebased collective variable on two biologically relevant interactions: the ANT-Cardiolipin (CL) and Kir2.2-PIP2. Converged PMFs were obtained for both systems; for the Kir2.2 R186A channel mutant, the potential of mean force (PMF) well-depth is reduced, in qualitative agreement with experimental data, where the mutation reduces conductance. These results demonstrate that REUS methodology can be robustly used to estimate protein-lipid binding interactions with the coarse-grained model, on different binding pockets and lipid types. The REUS methodology was then extended to study prototypical transmembrane (TM) helix-helix interactions. A series of model helix-helix dimers was studied, experimentally known to span a range of affinities, including the Glycophorin A (GpA) TM fragment and a number of different Receptor Tyrosine Kinase (RTK) TM fragments. The coarsegrained MARTINI model correctly predicts the bound state to be most stable, for GpA. Unexpectedly the non-dimerizing GpA mutant was just as stable as the wild-type. Similarly, for RTKs, the calculated affinities were similar, in contrast to semi-quantitative TOXCAT data that suggests a range of affinities. Using the collective variable developed to study the coarse-grained helix-helix systems, an all-atom PMF was calculated for wild-type GpA in a 1-palmitoyl,2-oleoyl-sn-glycero-3- phosphocholine (POPC) bilayer. Unexpectedly, the dimer state was found to be metastable, with the CHARMM36 and AMBER force fields, with the unbound state being the global energy minimum. Two simple corrections were proposed for the CHARMM36 force field, which stabilize the bound state making the model consistent with experiment.