De novo high-resolution protein structure determination from sparse spin-labeling EPR data

Abstract

As many key proteins evade crystallization and remain too large for nuclear magnetic resonance spectroscopy, electron paramagnetic resonance (EPR) spectroscopy combined with site-directed spin labeling offers an alternative approach for obtaining structural information. Such information must be translated into geometric restraints to be used in computer simulations. Here, distances between spin labels are converted into distance ranges between beta carbons by using a "motion-on-a-cone" model, and a linear-correlation model links spin-label accessibility to the number of neighboring residues. This approach was tested on T4-lysozyme and alphaA-crystallin with the de novo structure prediction algorithm Rosetta. The results demonstrate the feasibility of obtaining highly accurate, atomic-detail models from EPR data by yielding 1.0 A and 2.6 A full-atom models, respectively. Distance restraints between amino acids far apart in sequence but close in space are most valuable for structure determination. The approach can be extended to other experimental techniques such as fluorescence spectroscopy, substituted cysteine accessibility method, or mutational studies.

Figure 1

Rational for translating d_SL into d_Cβ for use as a restraint. A) Chemical structure of a nitroxide spin label side chain with the distance from the Cβ atom to the spin label atom indicated (Borbat, Mchaourab et al. 2002). B) Illustration of how the maximum distance from Cβ to spin label, SL, is reduced to an effective distance, SL^eff (depicted by a double line). C) d_SL is a starting point for the upper estimate of d_Cβ, and subtracting the effective distance of 6Å twice from d_SL gives a starting point for the lower estimate of d_Cβ. D) A histogram compares T4-lysozyme crystal structure (black bars, left y-axis, bottom x-axis) and αA-crystallin comparative model (white bars, left y-axis, bottom x-axis) d_SL–d_Cβ values with those obtained from the simple cone model (circles and line, right y-axis, top x-axis).

Figure 2

Map of the EPR restraints on the T4-lysozyme crystal structure (A-D) and on the αA-crystallin comparative model (E-H). A and E) Red dotted lines show d_Cβ distances, which are restrained by respective d_SL. B and F) Residues for which accessibilities e_SL were measured are depicted as space-filling models. C and G) Diagram shows d_SL (blue circle), the range of the derived distance restraints (blue), and the corresponding crystal/comparative model d_Cβ (red bar). D and H) Diagram illustrating the correlation of e_SL with e_Cβ. The lines indicate the consensus model fit ±3•σ_Cβ, where σ_Cβ, was recalculated based on the consensus fit to be 1.70Å. In B, F, D, and H the residues are color-coded with decreasing e_SL from blue – cyan – yellow – orange – red; black indicates amino acids in αA-crystallin that show reduced experimental accessibility due to intermolecular contacts with other αA-crystallin units in the oligomeric protein.

Figure 3

Illustration of the value of the experimental restraints in de novo protein folding for T4-lysozyme (A - H) and αA-crystallin (I - P). The backbone RMSD distribution of 10,000 T4-lysozyme de novo models created A) without the use of EPR restraints, B) with only the use of EPR distance restraints, C) with only the use of EPR accessibility restraints, D) with the use of EPR distance and accessibility restraints. E) The backbone RMSD distribution of 10,000 T4-lysozyme de novo models created with the use of 1/3 of the EPR distance restraints: top bar) those with the largest information content; second bar) those between amino acids furthest apart in sequence. The third and fourth black and white bars denote the sum percent of information content of the restraints used for the top and second bars, respectively. The width of the blocks comprising the black and white bars denotes the information content of individual restraints. F) Same as for E) but using the distance restraints with the lowest information content (top bar) and nearest in sequence (second bar). G) Same as for E) but using 2/3 of the total distance restraints. H) Same as for F) but using 2/3 of the total distance restraints. I – P) Same as A - H but for αA-crystallin.

Figure 4

Correlation of de novo models' accuracy with the energy of the de novo models. A) and C) The non-loop RMSD versus Rosetta energy for T4-lysozyme and αA-crystallin models, respectively. B) and D) The percentage of incorrectly built side chain conformations versus Rosetta energy for T4-lysozyme and αA-crystallin models, respectively. In all diagrams the minimized crystal structure or comparative model is depicted as a circle; the lowest energy model is shown as a square.

Figure 5

Overlay of lowest energy de novo models on crystal structure or comparative model. A) and B) For T4-lysozyme and αA-crystallin, respectively, superimposition of the lowest energy model (rainbow colored) with the crystal structure or comparative model (gray). The backbone is given as a ribbon diagram. Side chains of T4-lysozyme and of the β-sandwich of αA-crystallin are shown as stick models without hydrogen atoms.

Figure 6

Distance measurements at room temperature and in the solid state between spin labels using EPR. A) Representative reference EPR in the absence of dipolar coupling obtained from the digital sum of the corresponding single mutant spectra. B) Spectra of double mutants along with the non-linear least squares fit obtained by the convolution method as described in the experimental methods section. C) Distance distributions obtained from CW-EPR spectra. D) Distance measurements by DEER for representative double mutants. E) Raw DEER signals were background corrected and then fit using Tikhonov regularization to obtain F) average distances and distance distributions.