Structure Determination of Biomineral-Associated Proteins Using Combined Structure Prediction and Solid-State NMR

Protein-biomineral interactions are paramount to materials fabrication in biology. Unfortunately, the structure of biomineral-associated proteins cannot be determined by X-ray crystallography or solution NMR. Solid-state NMR (ssNMR) remains the only method for determining the relative distance between atoms at the protein-biomineral interface. However, the amount of data acquired during protein structure determination by solution NMR, typically 10-15 measurements at each residue, is not tractable by ssNMR methods. We have previously reported a computational structure-prediction method, called RosettaSurface, which is capable of generating ensembles of candidate structures of protein adsorbed states. Here we report a method for determining the structure of biomineral-associated proteins by combining ssNMR and RosettaSurface in an iterative fashion. Naïve structure predictions can guide the selection of sites for radioisotope labeling for measurement by ssNMR. These measurements are then used as soft constraints in a subsequent round of structure prediction. The cycle can be iterated. Measurements in high-confidence regions are used to validate or invalidate the model, and measurements in low-confidence regions are used to remove ambiguities and restrict the available model space. In addition, the algorithm is able to identify lattice geometries most compatible with ssNMR constraints. We demonstrate the approach with two case studies, human salivary statherin and the leucine-rich amelogenin protein, both adsorbed to hydroxyapatite (the primary mineral component of mammalian skeletal and dental tissue). The combined method has the potential to determine high-resolution protein structure at interfaces for the first time.