Modeling of spin–spin distance distributions for nitroxide labeled biomacromolecules

Abstract

Electron Paramagnetic Resonance (EPR) spectroscopy is a powerful method for unraveling structures and dynamics of biomolecules. Out of the EPR tool box, Pulsed Electron–Electron Double Resonance spectroscopy (PELDOR or DEER) enables one to resolve such structures by providing distances between spin centers on the nanometer scale. Most commonly, both spin centers are spin labels or one is a spin label and the other is a paramagnetic metal ion, cluster, amino acid or cofactor radical. Often, the translation of the measured distances into structures is complicated by the long and flexible linker connecting the spin center of the spin label with the biomolecule. Nowadays, this challenge is overcome by computational methods but the currently available approaches have a rather large mean error of roughly 2–3 Å. Here, the new GFN-FF general force-field is combined with the fully automated Conformer Rotamer Ensemble Search Tool (CREST) [P. Pracht et al., Phys. Chem. Chem. Phys., 2020, 22, 7169–7192] to generate conformer ensembles of the R1 side chain (methanthiosulfonate spin label (MTSL) covalently bound to a cysteine) in several cysteine mutants of azurin and T4 lysozyme. In order to determine the Cu²⁺–R1 and R1–R1 distance distributions, GFN-FF based MD simulations were carried out starting from the most probable R1 conformers found by CREST. The deviation between theory and experiment in mean inter-spin distances was 0.98 Å on average for the mutants of azurin (1.84 Å for T4 lysozyme) and the right modality was obtained. The error of the most probable distances for each mode was only 0.76 Å in the case of azurin. This CREST/MD procedure does thus enable precise distance-to-structure translations and provides a means to disentangle label from protein conformers.