The determination of a representative set of protein structures is a chief aim in structural genomics. Solid-state NMR may have a crucial role in structural investigations of those proteins that do not easily form crystals or are not accessible to solution NMR, such as amyloid systems or membrane proteins. Here we present a protein structure determined by solid-state magic-angle-spinning (MAS) NMR. Almost complete (13)C and (15)N resonance assignments for a micro-crystalline preparation of the alpha-spectrin Src-homology 3 (SH3) domain formed the basis for the extraction of a set of distance restraints. These restraints were derived from proton-driven spin diffusion (PDSD) spectra of biosynthetically site-directed, labelled samples obtained from bacteria grown using [1,3-(13)C]glycerol or [2-(13)C]glycerol as carbon sources. This allowed the observation of long-range distance correlations up to approximately 7 A. The calculated global fold of the alpha-spectrin SH3 domain is based on 286 inter-residue (13)C-(13)C and six (15)N-(15)N restraints, all self-consistently obtained by solid-state MAS NMR. This MAS NMR procedure should be widely applicable to small membrane proteins that can be expressed in bacteria.

The assignment of nonexchanging protons of a small microcrystalline protein, the alpha-spectrin SH3 domain (7.2 kDa, 62 residues), was achieved by means of three-dimensional (3D) heteronuclear (1H-13C-13C) magic-angle spinning (MAS) NMR dipolar correlation spectroscopy. With the favorable combination of a high B(0)-field, a moderately high spinning frequency, and frequency-switched Lee-Goldburg irradiation applied during 1H evolution, a proton linewidth < or =0.5 ppm at 17.6 Tesla was achieved for the particular protein preparation used. A comparison of the solid-state 1H chemical shifts with the shifts found in solution shows a remarkable similarity, which reflects the identical protein structures in solution and in the solid. Significant differences between the MAS solid- and liquid-state 1H chemical shifts are only observed for residues that are located at the surface of the protein and that exhibit contacts between different SH3 molecules. In two cases, aromatic residues of neighboring SH3 molecules induce pronounced upfield ring-current shifts for protons in the contact area.

In this paper, a three-dimensional (3D) NMR-based approach for the determination of the fold of moderately sized proteins by solid-state magic-angle spinning (MAS) NMR is presented and applied to the R-spectrin SH3 domain. This methodology includes the measurement of multiple 13 C- 13 C distance restraints on biosynthetically site-directed 13 C-enriched samples, obtained by growing bacteria on [2-13 C]glycerol and [1,3-13 C]glycerol. 3D 15 N-13 C-13 C dipolar correlation experiments were applied to resolve overlap of signals, in particular in the region where backbone carbon-carbon correlations of the C R -C R , CO-CO, C R -CO, and CO-C R type appear. Additional restraints for confining the structure were obtained from φ and ψ backbone torsion angles of 29 residues derived from C R , C , CO, NH, and H R chemical shifts. Using both distance and angular restraints, a refined structure was calculated with a backbone root-mean-square deviation of 0.7 Å with respect to the average structure.Many biological systems, such as membrane proteins and amyloid fibrils, remain a challenge in structural biology because of difficulties with crystallization and solubility. In the past years, solid-state NMR 1 has become a promising method for obtaining structural information about these systems, via the measurement of accurate distances (1-7), φ and backbone torsion angles (8-10), and chemical shift anisotropy (11,12). In these studies, samples labeled only in the positions of interest were investigated. For the determination of the complete protein folds, however, a different approach that allows the collection of a large number of structural restraints from a small number of samples has to be followed. The quality of the structures increases with the number of restraints, and the more that are measured, the lower the accuracy of the individual restraints may be. From a close analysis of the topology of helical and -sheet structures, it transpires that carboncarbon distances are very important in defining the fold of a protein. For example, distances between backbone carbons, i.e., R-carbons and carbonyl carbons, define the secondary structure of a protein (Table 1). Distances between backbone and side chain carbons or between side chain and side chain carbons provide information about the tertiary structure. The detection of structure-defining long-range carbon-carbon restraints is only possible when so-called dipolar truncation effects are suppressed (13,14). This can be accomplished by employing a reduced labeling scheme, in which chemically bonded carbons are not simultaneously labeled and hence the number of strong dipolar couplings between connected nuclei is reduced. For proteins expressed in bacterial systems, this can be achieved by using [2-13 C]glycerol or [1,3-13 C]glycerol as the only carbon source in the media (15)(16)(17). In combination with this labeling pattern, long-range 13 C-13 C distance restraints may be collected by using a broad-band recoupling method like the proton-driven spindiffusion (PDSD...

In this communication, we demonstrate the feasibility of 1H detection in MAS solid-state NMR for a microcrystalline, uniformly 2H,15N-labeled sample of a SH3 domain of chicken alpha-spectrin, using pulsed field gradients for suppression of water magnetization. Today, B0 gradients are employed routinely in solution-state NMR for coherence order selection and solvent suppression. We suggest to use gradients to purge water magnetization which cannot be suppressed using conventional water suppression schemes. The achievable gain in sensitivity for 1H detection is in the order of 5 compared to the 15N detected version of the experiment (at a MAS rotation frequency of 13.5 kHz). We expect that this labeling concept which achieves high sensitivity due to 1H detection, in combination with the possibility to measure long range 1H-1H distances as we have shown previously, to be a useful tool for the determination of protein structures in the solid state.