The conversion of atmospheric gases in organisms represents a crucial interface between the inanimate and animate nature, facilitated by biological catalysts called enzymes. These enzymes bear the potential to reduce air pollution emitted by a highly industrialized society. For this geoengineering purpose, the molybdenum-dependent formate dehydrogenase from Rhodobacter capsulatus might represent a promising candidate, due to its ability to reversibly convert formate into carbon dioxide by withdrawing two electrons. Withdrawn electrons are transferred via an electron transfer chain consisting of iron-sulfur (FeS) clusters within the enzyme to another protein domain, where an electron acceptor becomes reduced. However, the reaction mechanism at the molybdenum cofactor (MoCo) acting as the catalytic center and the electron transfer mechanism are not precisely described. A well-established method for investigating such metalloproteins is the electron paramagnetic resonance (EPR) spectroscopy, which is utilized in this work to examine the binding site of formate as a substrate and azide as a competitive inhibitor, potentially adopting similar orientations close to the MoCo. The application of pulse EPR spectroscopy revealed that carbon dioxide is already released in an azide-inhibited paramagnetic Mo(V) state, while the formate proton resides in close proximity to the MoCo. By employing a combined approach of density functional theory (DFT) and pulse EPR spectroscopy, the precise binding site of the formate proton was identified and the orientation of the azide inhibitor elucidated in proximity to the formate proton. Furthermore, the redox potentials of the electron transferring moieties were investigated using EPR-mediated redox titration elucidating the electron transfer path through the enzyme. This investigation along with the results from the analyses of the azide and formate proton positions lead to the proposal that MoCo acts as an electron transfer transducer, converting the two-electron transfer into a sequential one-electron transfer over the FeS clusters. This effect is reversed in the FdsGB domain by a further proposed electron transfer transducer. Additionally, it was demonstrated in this work that the investigation of such metalloproteins with multiple paramagnetic centers is only feasible after extensive prior analysis and assignment of the signals to their corresponding centers. To simplify such analyses for similar research subjects, a computational method based on deep learning was developed, separating EPR signals with respect to their associated lifetimes. This algorithm was extensively assessed using a defined test set to estimate its applicability.
