Abstract

Long-range pseudo-contact NMR shifts (PCSs) provide important restraints for the structure refinement of proteins when a paramagnetic metal center is present, either naturally or introduced artificially. Here we show that ab initio quantum-chemical methods and a modern version of the Kurland–McGarvey approach for paramagnetic NMR (pNMR) shifts in the presence of zero-field splitting (ZFS) together provide accurate predictions of all PCSs in a metalloprotein (high-spin cobalt-substituted MMP-12 as a test case). Computations of 314 ¹³C PCSs using g- and ZFS tensors based on multi-reference methods provide a reliable bridge between EPR-parameter- and susceptibility-based pNMR formalisms. Due to the high sensitivity of PCSs to even small structural differences, local structures based either on X-ray diffraction or on various DFT optimizations could be evaluated critically by comparing computed and experimental PCSs. Many DFT functionals provide insufficiently accurate structures. We also found the available 1RMZ PDB X-ray structure to exhibit deficiencies related to binding of a hydroxamate inhibitor. This has led to a newly refined PDB structure for MMP-12 (5LAB) that provides a more accurate coordination arrangement and PCSs.