Long-range structural restraints in spin-labeled proteins probed by solid-state nuclear magnetic resonance spectroscopy.

Magic-angle spinning (MAS) solid-state nuclear magnetic resonance (SSNMR) spectroscopy is rapidly developing as a technique for the atomic-level characterization of structure and dynamics of biomacromolecules not amenable to analysis by X-ray crystallography or solution NMR.1-5 While nearly complete resonance assignments have been achieved for multiple 13C,15N-enriched proteins up to ∼100 aa,1,2 enabling the determination of relatively high-resolution 3D protein structures in several cases,6-8 studies of this type are generally hampered by the availability of a limited number of long range (>5 Å) structural restraints. Here we investigate the possibility of deriving long range (∼10 to 20 Å) restraints from MAS SSNMR spectra of 13C,15N-enriched proteins containing a covalently attached paramagnetic moiety. In general, the presence of unpaired electrons leads to electron-nucleus distance-dependent NMR chemical shift changes and enhanced longitudinal and transverse relaxation rates,9,10 and these effects have been successfully exploited in solution-state NMR studies of macromolecular structure.11,12 On the other hand, the majority of MAS NMR studies of paramagnetic solids to date have been carried out on metal coordination complexes,13-17 and only very recently have the initial applications to paramagnetic metalloproteins been reported.18-20 We focus here on proteins containing a bound nitroxide spin label. Nitroxide radicals, characterized by relatively large electronic relaxation time constants (T1e, T2e g ∼100 ns) and small g-anisotropy,10,11,21 are expected to significantly enhance the transverse relaxation of the neighboring nuclei in immobilized proteins (with rate constant R2 ∝ γI/r, where γI is the gyromagnetic ratio of the nuclear spin I and r is the electron-nucleus distance), while generating negligible pseudocontact shifts.10 A nitroxide sidechain (R1) (or its diamagnetic analogue, R1′, used here as a negative control) can be incorporated into proteins using the site-directed spin-labeling approach developed by Hubbell and co-workers22 (Figure S1, Supporting Information), where a cysteine residue is introduced at the desired position in the protein using site-directed mutagenesis followed by the specific reaction of the thiol group with a suitable reagent. A model 56 aa protein, B1 immunoglobulin-binding domain of protein G (GB1), was used in this study. GB1, which contains no native cysteines, has been extensively studied using biophysical and spectroscopic techniques, and detailed information about its structure, dynamics, and folding is available, including 3D solution23 and crystal24 structures and the complete 13C and 15N resonance assignments in the solid state.25 R1 and R1′ side-chains were incorporated at solvent-exposed sites in the R-helix (residue 28) and â4-strand (residue 53) as described in the Supporting Information (for brevity the proteins are named 28R1, 28R1′, 53R1, and 53R1′). Solution and SSNMR chemical shifts, and solution transverse relaxation enhancements reveal that these GB1 analogues retain the wild-type fold (Figures S2-S5). SSNMR measurements were performed on (i) microcrystalline 13C,15N-labeled diamagnetic proteins (28R1′ and 53R1′) and (ii) 13C,15N-28R1 (53R1) diluted in a diamagnetic matrix by cocrystallization with natural abundance 28R1′ (53R1′) in ∼1:3 molar ratio (to minimize intermolecular electron-nuclear dipolar couplings). 2D 15N-13CR spectra of 53R1/53R1′ and 28R1/28R1′ acquired at ∼11 kHz MAS rate are shown in Figure 1A,D. Backbone 15N and 13C assignments for 28R1′ and 53R1′ (Figure S5) were obtained using 2D 15N-(13CR)-13CX, 15N-(13C′)-13CX, and 13C-13C experiments, and relatively well-resolved correlations (∼50% of residues) are indicated. Notably, ∼25 to 50% of cross-peaks exhibit significantly reduced intensities in the R1 spectra relative to R1′. In addition, 15N-13CR correlations, which are detected in both spectra, display only minor linebroadening for R1 (∼5 to 30 Hz for 13C and ∼2 to 10 Hz for 15N), and essentially identical resonance frequencies indicating negligible pseudocontact shifts. Given that 53R1′ and 28R1′ adopt the GB1 fold, the reduced cross-peak intensities in R1 spectra are found to be highly correlated with the proximity of the corresponding nuclei to the spin label. For example, T25 and V29 (R-helix) are among the least affected correlations in the 53R1 spectrum, whereas I6 (â1-strand) and T49 (loop between â3 and â4) peaks are effectively suppressed. While the precise conformation of R1 (and hence the spin-label location) in 53R1 is currently unknown, the 1HN, 15N, and 13CR atoms are likely to be within ∼10 Å of the electron for I6 and T49, and ∼20 Å away for T25 and V29 (Figure S7). This spin topology is roughly reversed in 28R1 (T25/V29 and I6/T49 are ∼5 to 10 Å and ∼15 to 20 Å from the radical, respectively), resulting in T25/V29 (I6/T49) correlations being among those most (least) suppressed. The modulation of peak intensities, based on each residue’s proximity to the electron spin (Figure S7), persists throughout both 53R1 and 28R1. Figure 1 shows the relative cross-peak intensities (heights) in R1/R1′ spectra as a function of residue location in the primary (B,E) and tertiary (C,F) protein structure. For 53R1, the relaxation effects due to the spin label are largest for residues in the â1-â4 strands and connecting loops, while for 28R1 the most strongly affected residues are found in the R-helix and adjacent loops. Note that these data are in qualitative agreement with the solution-state paramagnetic relaxation enhancements for 28R1 and 53R1 shown in Figure S4. Dipolar contributions to 1H, 13C, and 15N transverse relaxation rates due to the spin label were estimated using the SolomonBloembergen equation,9,10 assuming an electron correlation time of 100 ns (Figures S8 and S9). These calculations indicate that the reduced cross-peak intensities in R1 spectra result primarily from the decay of transverse 1H and 13C coherences during 1H-15N and 15N-13CR cross-polarization (CP) steps, respectively, and are further supported by measurements of magnetization decay during spinlock pulses (Figure S10). Under our experimental conditions (0.15 ms 1H-15N CP, 3 ms 15N-13CR CP), cross-peaks arising from nuclei within ∼10 Å of the spin label are expected to be at most ∼20% as intense for R1 relative to R1′, even in the absence of additional † On leave from the Department of Chemical and Environmental Sciences, University of Limerick, Limerick, Ireland. Published on Web 05/27/2007