Bringing Molecular Dynamics and Ion-Mobility Spectrometry Closer Together: Shape Correlations, Structure-Based Predictors, and Dissociation
Résumé
Unfolding of proteins gives detailed information about their structure and energetics and can be probed as a response to a change of experimental conditions. Ion mobility coupled to native mass spectrometry is a gas-phase technique that can observe such unfolding in the gas phase by monitoring the collision cross section (CCS) after applying an activation, for example, by collisions (collision-induced unfolding, CIU). The structural assignments needed to interpret the experiments can profit from dedicated modeling strategies. While predictions of ion-mobility data for well-defined and structurally characterized systems is straightforward, systematic free-energy calculations or biased molecular dynamics simulations that employ IMS data are still limited. The methods with which CCS values are calculated so far do not allow for analytical gradients needed in biased molecular dynamics (MD), and further, explicit CCS calculations still can pose computational bottleneck—when integrated into MD-bioinformatics workflows. These limitations motivate one to revisit known correlations of the CCS with the aim to find computationally cheap and versatile but still at least semiquantitative descriptions of the CCS by pure structural descriptors. We have therefore investigated the correlation of CCS with the key structural parameter often used in computational unfolding studies—the gyration radius—for several small monomeric and dimeric proteins. We work out the challenges and caveats of the combinations of the configurational sampling method and the CCS-calculation algorithm. The correlations were found to be sensitive to the generation conditions and additionally to the system topology. To reduce the amount of fitting to be undertaken, we devise a simple structural model for the CCS that shares some commonalities with the hard-sphere model and the projection algorithm but is designed to take unfolding into account. With this model, we suggest a two-point interpolating function rather than fitting a large data set, at only little deterioration of the predictive power. We further proceed to a model with composition and structure dependence that builds only upon the gyration radius and the chemical formula to apply the found CCS scaling behavior—the scaled macroscopic sphere (sMS) predictor. We demonstrate its applicability to describe unfolding and also its transferability for a larger set of structures from the RSCPDB. As we have found for the dimeric systems, that shape correlations with one global descriptor qualitatively break down, we finally suggest a recipe to switch between global and fragment-based CCS prediction, that takes up the ideas of coarse-graining protein complexes. The presented models and approaches might provide a basis to boost the integration of structural modeling with multistage IMS experiments, especially in the field of large-scale bioinformatics or “on-the-fly” biasing of MD, where computational efficiency is critical.