Atomic refinement with correlated solid-state NMR restraints.

The orientation data provided by solid-state NMR can provide a great deal of structural information about membrane proteins. The quality of the information provided is, however, somewhat degraded by sign degeneracies in measurements of the dipolar coupling tensor. This is reflected in the dipolar coupling penalty function used in atomic refinement, which is less capable of properly restraining atoms when dipolar sign degeneracies are present. In this report we generate simulated solid-state NMR data using a variety of procedures, including back-calculation from crystal structures of alpha-helical and beta-sheet membrane proteins. We demonstrate that a large fraction of the dipolar sign degeneracies are resolved if anisotropic dipolar coupling measurements are correlated with anisotropic chemical shift measurements, and that all sign degeneracies can be resolved if three data types are correlated. The advantages of correlating data are demonstrated with atomic refinement of two test membrane proteins. When refinement is performed using correlated dipolar couplings and chemical shifts, perturbed structures converge to conformations with a larger fraction of correct dipolar signs than when data are uncorrelated. In addition, the final structures are closer to the original unperturbed structures when correlated data are used in the refinement. Thus, refinement with correlated data leads to improved atomic structures. The software used to correlate dipolar coupling and chemical shift data and to set up energy functions and their derivatives for refinement, CNS-SS02, is available at our web site.