Protein structure determination from NMR chemical shifts

Abstract

NMR spectroscopy plays a major role in the determination of the structures and dynamics of proteins and other biological macromolecules. Chemical shifts are the most readily and accurately measurable NMR parameters, and they reflect with great specificity the conformations of native and nonnative states of proteins. We show, using 11 examples of proteins representative of the major structural classes and containing up to 123 residues, that it is possible to use chemical shifts as structural restraints in combination with a conventional molecular mechanics force field to determine the conformations of proteins at a resolution of 2 angstroms or better. This strategy should be widely applicable and, subject to further development, will enable quantitative structural analysis to be carried out to address a range of complex biological problems not accessible to current structural techniques.

Fig. 1.

Schematic illustration of the molecular fragment replacement procedure implemented for chemical shifts in the CHESHIRE procedure. The protein shown is ubiquitin, and fragments are generated with main-chain dihedral angles compatible with the information contained in the chemical shifts. The fragments are then assembled in a combinatorial manner to produce an ensemble of trial structures that are subsequently refined by exploiting the information about tertiary structure contained in the chemical shifts.

Fig. 2.

Comparison of the structures, also showing side chains in the hydrophobic cores, determined from chemical-shift information using the CHESHIRE procedure and those determined by standard x-ray or NMR methods. (A) Ubiquitin (blue) and PDB entry 1UBQ (pink). (B) FF domain (blue) and PDB entry 1UZC (pink). (C) Calbindin (blue) and PDB entry 4ICB (pink). (D) HPr (blue) and PDB entry 1POH (pink).

Fig. 3.

Landscapes of the CHESHIRE scores (E) for four of the proteins analyzed in this work. The landscapes report the E scores as a function of the RMSD from the reference structures (see Table 1) or the structures of minimal CHESHIRE scores (Insets). The proteins are those shown in Fig. 2. In all cases, the Z scores of the best structures are below −3, indicating that the landscapes are funneled toward the native structure. The averages and standard deviations are shown by red horizontal lines. The E_pred energies are also shown as horizontal green lines. (a) Ubiquitin. (b) FF domain. (c) Calbindin. (d) HPr. See also Fig. 2.