Methods to accelerate protein structure determination by solution NMR

National Institute Of Diabetes And Digestive And Kidney Diseases

Investigators

Linked publications & trials

Abstract

Structural analysis of meta-stable proteins is challenging because measured parameters often reflect a population weighted average over the different states dynamically sampled by such systems. New three- and four-dimensional NMR experiments have been developed that provide access to the structure and dynamics of systems that are large in comparison to proteins normally studied by NMR spectroscopy, such as the homodimeric Main protease of the SARS-CoV-2 virus. The novel methods combined with 900 MHz high-field measurements provided sufficient data for a detailed structural characterization that reveals subtle but statistically quite significant differences relative to all X-ray structures available so far. A characterization of the backbone motions shows a strong effect of substrate binding on the dynamics of the protein backbone in the active site region. Specifically, new schemes to acquire and process triple-resonance data for measurement of residual dipolar couplings (RDCs) in combination with non-uniform sampling have been developed that increase the precision at which such RDCs can be extracted from the data. The TROSY-HNCO 3D NMR experiments was extended in a manner similar to that of the original amplitude-modulated RDC TROSY Spectroscopy (ARTSY) experiment to obtain vastly improved spectral resolution, which allowed the measurement of a nearly complete set of data for the SARS-CoV-2 main protease, and to assess its structure in solution and to compare it with the many X-ray structures deposited in the protein databank (PDB). Results show that although, on average, solution measurements are in good agreement with the X-ray data, in several localized regions the solution data fall outside the range observed in the nearly 200 X-ray structures. Substantial differences appear in the active site region, and refinement of the solution NMR structure of this homodimeric structure indicates that many of the residues near the active site exist in a microsecond timescale equilibrium between multiple states. None of the X-ray structures, provides a good fit to the solution NMR RDCs, although some improvement in agreement is observed when comparing the NMR data to ensemble models developed by the PHENIX method to fit X-ray diffraction data to ensembles of structures. In related work, we have developed methods that permit NMR detection of sulfhydryl resonances in proteins. Ten out of 11 Cys thiol resonances proved observable, but their chemical shifts mostly disagreed with those of AlphaFold2 models, indicating that the proton placement in such models remains far from optimal. Good agreement with experimental chemical shifts is observed when using quantum-chemical calculation to define the energetic minima for the positions of these hydrogen atoms, revealing substantially different hydrogen bonding networks.