Accuracy and precision of protein structures determined by magic angle spinning NMR spectroscopy: for some 'with a little help from a friend'

Abstract

We present a systematic investigation into the attainable accuracy and precision of protein structures determined by heteronuclear magic angle spinning solid-state NMR for a set of four proteins of varied size and secondary structure content. Structures were calculated using synthetically generated random sets of C-C distances up to 7 Å at different degrees of completeness. For single-domain proteins, 9-15 restraints per residue are sufficient to derive an accurate model structure, while maximum accuracy and precision are reached with over 15 restraints per residue. For multi-domain proteins and protein assemblies, additional information on domain orientations, quaternary structure and/or protein shape is needed. As demonstrated for the HIV-1 capsid protein assembly, this can be accomplished by integrating MAS NMR with cryoEM data. In all cases, inclusion of TALOS-derived backbone torsion angles improves the accuracy for small number of restraints, while no further increases are noted for restraint completeness above 40%. In contrast, inclusion of TALOS-derived torsion angle restraints consistently increases the precision of the structural ensemble at all degrees of distance restraint completeness.

Keywords: Integrated structural biology; Magic angle spinning; Protein structure calculation.

Figure 1.

a) Generic polypeptide chain, illustrating select backbone dihedral angles and ¹H-¹H and ¹³C-¹³C distances. b) Distance dependence of the ¹H-¹H NOE and ¹³C-¹³C dipolar coupling. The NOE curve was calculated for τ_c=7.1 ns, corresponding to a spherical protein of 14.6 kDa molecular mass at T = 37 °C. c) Ribbon representations of dynactin’s CAP-Gly domain (PDBID: 2MPX), Oscilatoria aghardii agglutinin, OAA (PDBID: 3OB2), the carbohydrate binding domain (CBD) of galectin-3C (PDBID: 3ZSJ), and full-length chain of HIV-1 capsid protein (CA) in the assembled state (PDBID: 4XFX).

Figure 2.

Flow diagram of computational strategy for protein structure calculations on the basis of synthetic C-C distance restraints.

Figure 3.

Structure calculation for dynactin’s CAP-Gly domain. a) Accuracy as defined by atomic backbone RMSD with respect to the target (input) structure and b) Precision as defined by pairwise atomic backbone RMSD for ensemble members plotted vs. restraint completeness and number of restraints per residue. Data without (black symbols) or with (green symbols) backbone torsion angle restraints from TALOS-N using 200 experimental ¹³C and 71 ¹⁵N chemical shifts (BMRB 25005). The RMSD values for calculations using experimental C-C distance restraints are shown with open symbols. The horizontal dashed lines are the average values of those at 60, 80, and 100% restraint completeness. c) Experimental ¹³C and ¹⁵N chemical shifts used in the structure calculation listed along the amino acid sequence. Secondary structure elements are depicted below the sequence. d) Superposition C-C distances at 20% restraint completeness onto the CAP Gly structure. e) Top: Superposition of the experimental set of C-C distances onto the CAP-Gly structure. Bottom: Best-fit superpositions of the 10 lowest-energy structures calculated on the basis of the experimental distance restraints without (left) and with (right) TALOS-N derived backbone torsion angle restraints. f,g) Best-fit superpositions of the 10 lowest-energy structures calculated for different degrees of restraint completeness without (f) and with (g) TALOS-N derived backbone torsion angle restraints.

Figure 4.

Structure calculation for Oscilatoria aghardii agglutinin, OAA. a) Accuracy as defined by atomic backbone RMSD with respect to the target (input) structure and b) Precision as defined by pairwise atomic backbone RMSD for ensemble members plotted vs. restraint completeness and number of restraints per residue. Data without (black symbols) or with (green symbols) backbone torsion angle restraints from TALOS-N using 235 experimental ¹³C and 92 ¹⁵N chemical shifts. The dashed lines are the average values of those at 60, 80, and 100% restraint completeness. c) Experimental ¹³C and ¹⁵N chemical shifts used in the structure calculation listed along the amino acid sequence. Secondary structure elements are depicted below the sequence. d) Superposition of C-C distances at 20% restraint completeness onto the OAA structure. e, f) Best-fit superpositions of the 10 lowest-energy structures calculated for different degrees of restraint completeness without (e) and with (f) TALOS-N derived backbone torsion angle restraints.

Figure 5.

Structure calculation for the Galectin CBD. a) Accuracy as defined by atomic backbone RMSD with respect to the target (input) structure and b) Precision as defined by pairwise atomic backbone RMSD for ensemble members plotted vs. restraint completeness and number of restraints per residue. Data without (black symbols) or with (green symbols) backbone torsion angle restraints from TALOS-N using 385 experimental ¹³C and 132 ¹⁵N chemical shifts. The dashed lines are the average values of those at 60, 80, and 100% restraint completeness. c) Experimental ¹³C and ¹⁵N chemical shifts used in the structure calculation listed along the amino acid sequence. Secondary structure elements are depicted below the sequence. d) Superposition of C-C distances at 20% restraint completeness onto the galectin’s CBD structure. e, f) Best-fit superpositions of the 10 lowest-energy structures calculated for different degrees of restraint completeness without (e) and with (f) TALOS-N derived backbone torsion angle restraints.

Figure 6.

Structure calculation for the HIV-1 CA capsid protein (CA). a) Accuracy as defined by atomic backbone RMSD with respect to the target (input) structure and b) Precision as defined by pairwise atomic backbone RMSD for ensemble members plotted vs. restraint completeness and number of restraints per residue. Data without (black symbols) or with (green symbols) backbone torsion angle restraints from TALOS-N using 618 experimental ¹³C and 205 ¹⁵N chemical shifts. The dashed lines are the average values of those at 60, 80, and 100% restraint completeness. c) Experimental ¹³C and ¹⁵N chemical shifts used in the structure calculation listed along the amino acid sequence. Secondary structure elements are depicted below the sequence. d) Superposition of the C-C distances at 20% restraint completeness onto the CA structure. e,f) Best-fit superpositions of the 10 lowest-energy structures calculated for different degrees of restraint completeness without (e) and with (f) TALOS-N derived backbone torsion angle restraints.

Figure 7.

Atomic model of a hexamer unit of HIV-1 CA in tubular assemblies generated by combining MAS NMR-derived distances and cryoEM density. The NTD (residues 1–145) and CTD (residues 148–231) models were taken from the lowest-energy structure of a single chain CA calculated at 40% restraint completeness (see Figure 6). a) NTD and CTD domains were fit into the cryoEM map of CA hexamer by automated rigid-body docking. b) The position of one CTD domain was manually adjusted to improve the fit. c) CA hexamer after real-space refinement. NTDs and CTDs are shown in purple and cyan, respectively; the β-hairpin is colored yellow.