Evaluating the usefulness of metadynamics simulations to model hydrogen-deuterium exchange in proteins
Abstract
We want to use Molecular Dynamics (MD) simulation to interpret data acquired from Hydrogen-Deuterium eXchange Mass Spectrometry (HDX-MS). In HDX-MS, a protein is exposed to a buffer containing high levels of deuterium for different lengths of time during which the amide hydrogens can exchange with the deutriums, and the rate of exchange is a function of several factors including how buried the amide hydrogen is and it hydrogen bonding pattern. HDX-MS analysis gives protection factors per residue in a protein that is correlated to solvent accessibility, structural features and hydrogen bonding. It also gives information on how these protection factors change when the protein is subjected to different conditions, for example, addition of a ligand. Although extremely powerful, HDX-MS cannot exactly underpin the changes that occur on an atomistic level when the system goes from one state to another. Therefore, we envision to use molecular dynamics to observe the conformational changes that occur during this transition. We have picked bovine calmodulin as our model protein as it has been shown to undergo a conformational change upon binding four calcium ions per molecule, and each conformation has been separately crystallised. It is also relatively small in size (19kDa), which makes the HDX-MS analysis quicker. The HDX-MS data has been acquired for both the apo (no Ca2+) and holo (50 mM Ca2+) states, for four different labelling times (10 s, 1 min, 10 min and 1 hour) and four replicates for each time. To model the structural changes that give rise to the HDX data, we want to use an enhanced sampling method, known as metadynamics. Metadynamics encourages the system to explore more regions of the phase space by filling explored minima with Gaussians, essentially adding a penalty term for visiting an already visited spot on the potential energy landscape. To perform metadynamics, one has to carefully choose a low-dimensional collective variable (CV). As calmodulin has already been crystallised in its apo and holo states, one choice of CV was the root-mean-squared deviation (RMSD) between the two reference structures, where the protein would move from the apo to the holo structures, along the RMSD coordinate. To account for a more general case, where the final structure of the protein is unknown, we wish to explore other collective variables such as Solvent-Accessible Surface Area (SASA) as our CV, as it is a good proxy for how accessible certain amide hydrogens are for deuterium. We want to test different collective variables and metadynamics parameters to be able to show that structural changes that give rise to certain HDX uptake patterns can be reproduced using molecular dynamics.
