|Home | About | Journals | Submit | Contact Us | Français|
Magic-angle spinning (MAS) solid-state NMR (SSNMR) spectroscopy of uniformly-13C,15N labeled protein samples provides insight into atomic-resolution chemistry and structure. Data collection efficiency has advanced remarkably in the last decade; however, the study of larger proteins is still challenged by relatively low resolution in comparison to solution NMR. In this study, we present a systematic analysis of SSNMR protein spectra acquired at 11.7, 17.6 and 21.1 Tesla (1H frequencies of 500, 750, and 900 MHz). For two protein systems—GB1, a 6 kDa nanocrystalline protein and DsbA, a 21 kDa nanocrystalline protein—line narrowing is demonstrated in all spectral regions with increasing field. Resolution enhancement is greatest in the aliphatic region, including methine, methylene and methyl sites. The resolution for GB1 increases markedly as a function of field, and for DsbA, resolution in the C-C region increases by 42%, according to the number of peaks that can be uniquely picked and integrated in the 900 MHz spectra when compared to the 500 MHz spectra. Additionally, chemical exchange is uniquely observed in the highest field spectra for at least two isoleucine Cδ1 sites in DsbA. These results further illustrate the benefits of high-field MAS SSNMR spectroscopy for protein structural studies.
Solid-state NMR (SSNMR) spectroscopy is a powerful tool for studying protein structure and function, uniquely able to address macroscopically disordered proteins. Insights from SSNMR include atomic-resolution structure, site-specific dynamics, metal center chemistry, and orientation of membrane proteins in bilayers (Smith and van Eck, 1999; Baldus, 2002; McDermott, 2004; Opella and Marassi, 2004). In recent years, methods have increasingly emphasized the use of uniformly-13C,15N-labeled samples to obtain site-specific information throughout an entire protein in each experiment, a capability for which high magnetic field has been demonstrated to be critical (McDermott et al., 2000; Pauli et al., 2001; Castellani et al., 2002; Bockmann et al., 2003; Igumenova et al., 2004a; Igumenova et al., 2004b). In principle, this approach greatly enhances throughput and precision of analysis for large peptides and proteins; in practice, however, spectra may be insufficiently resolved to permit resonance assignments and subsequent analysis. To enhance resolution, investigators have made advances in isotopic labeling schemes (LeMaster and Kushlan, 1996; Hong and Jakes, 1999), decoupling sequences (Bennett et al., 1995; Fung et al., 2000; Detken et al., 2002), and probe designs (Stringer et al., 2005; Doty et al., 2006; Dillmann et al., 2007).
Here we demonstrate systematically the benefits of using ultra-high magnetic fields for magic-angle spinning (MAS) spectra of proteins as nanocrystalline protein preparations. We demonstrate the experimental spectral resolution enhancement at ultra-high field with direct comparison of two proteins over a large range of resonance frequencies, including the highest magnetic field currently available for protein MAS studies (900 MHz). We observe a linear resolution benefit and find that chemical exchange is most prominent at ultra-high field.
DsbA has been successfully expressed in 100 mg quantities using an expression method described by Marley et al. (Marley et al., 2001) E. coli containing DsbA expression plasmid were grown in unlabeled rich medium (LB) to generate large cell mass and then transferred to one-fourth volume of isotopically labeled minimal media prior to induction with IPTG. The cells were lysed by osmotic shock to release the periplasmic proteins. The procedure is intended to wash the cells with sucrose and then lyse the outer membrane with water, but we found that collecting the wash and the water supernatants increased the yield. DsbA was then purified by anion exchange chromatography, and ultimately oxidized with 5 mM glutathione and crystallized in a solution of 30% PEG 8000, 50 mM cacodylate pH 6.5, and 1.5% 2-methyl-2,4-pentanediol. The nanocrystalline solid was packed into limited speed 3.2-mm rotors. The preparation for GB1 has been previously reported (Franks et al., 2005).
Spectra were acquired at B0 fields of 11.7, 17.6, and 21.1 Tesla on a 500 MHz (1H frequency) Varian Infinity Plus spectrometer, a 750 MHz (1H frequency) Varian Unity Inova, and a 900 MHz (1H frequency) Varian Unity Inova. MAS rates were 11.111 kHz at 500 MHz, 12.500 kHz at 750 MHz, and 13.333 kHz at 900 MHz. The 500 MHz and 750 MHz data were acquired with Varian Balun™ 1H-13C-15N 3.2 mm probes, and the 900 MHz data were acquired with a Varian BioMAS™ 1H-13C-15N 3.2 mm probe in double resonance (1H-13C) mode (Stringer et al., 2005). The DsbA data were acquired on a 3.2 mm standard wall rotor with ~6 mg protein on the 750 MHz spectrometer, and on a 3.2 mm limited speed rotor with ~18 mg protein on the 500 and 900 MHz spectrometers. The GB1 data were acquired on three different 3.2 mm standard wall rotors with ~24 mg protein on the 750 and ~10 mg protein on the 500 MHz, and ~12 mg protein on the 900 MHz spectrometers. All pulse sequences utilized tangent ramped cross polarization (Hediger et al., 1994) with TPPM decoupling (Bennett et al., 1995) applied during the acquisition and evolution periods. The variable temperature gas was maintained at −10°C for DsbA with 100 scfh flow. The temperatures for GB1 varied at each field such that GB1 was maintained at 0°C for the 500 MHz experiments and −10°C for the 750 MHz experiments, and the 900 MHz experiments. Chemical shifts were referenced externally with adamantane (Morcombe and Zilm, 2003). Data were processed with nmrPipe (Delaglio et al., 1995) with back linear prediction and polynomial baseline (frequency domain) correction applied to the direct dimension. Zero filling to 16,384 in the direct dimension and 8,192 in the indirect dimension was applied and Lorentzian-to-Gaussian apodization was employed for each dimension before Fourier transformation; further details with specifics for each spectrum are found in the respective figure captions.
To minimize variations arising from technical details such as spectrometer electronics, probe design and sample preparations, we employed highly homologous instrument configurations and standard pulse sequences, applied to replica samples throughout this study. We prepared two proteins: (1) the 56 residue, 6 kDa immunoglobulin binding domain B1 of protein G, GB1 (Gronenborn et al., 1991; Franks et al., 2005) and (2) the 21 kDa soluble disulfide bond forming enzyme, DsbA. Each sample was packed in a limited speed 3.2-mm MAS rotor (Varian, Inc., Palo Alto, CA and Fort Collins, CO). Spectra were collected on three similar spectrometers and triple resonance 1H-13C-15N probes (500 MHz Infinity Plus and 750 MHz Inova with Balun™ 1H-13C-15N probe at the University of Illinois at Urbana-Champaign; 900 MHz Inova with BioMAS™ at the Pacific Northwest National Laboratories, Richland, WA), employing standard pulse sequences, including 1H-13C CP-MAS and two-dimensional (2D) 13C-13C with DARR mixing (Takegoshi et al., 2001; Morcombe et al., 2004) and carefully adjusted TPPM decoupling (Bennett et al., 1995). One can envision further improvements upon application of more recently developed pulse sequences.
GB1 was used initially to calibrate the instruments and assess the mixing time dependence of cross peak intensities (Fig. 1). We found the polarization transfer rates (at approximately the same MAS rates) to scale with the inverse square of magnetic field. Thus subsequent 2D 13C-13C data sets were acquired with DARR mixing times of 50, 100, and 200 ms at 500, 750, and 900 Mhz (1H frequencies) respectively, to produce similar cross peak intensity patterns. Not only the rate but also the overall efficiency of mixing at higher field is decreased (based on the ratio of cross peak to diagonal intensity); therefore sensitivity in the 2D spectra cannot be accurately compared, and we focus subsequent discussion on the relative resolution of the spectra. Spectra were acquired with a minimum of 6 ms evolution in the indirect dimension (t1) and 15 ms in the direct dimension (t2), with appropriate apodization applied (net broadening was applied using a 1:2 Lorentzian-to-Gaussian function), prior to zero filling (8k and 16k points respectively), Fourier transformation, and baseline correction. Complete processing details and spectra for each protein are included in the figure captions.
Resolution enhancement can be easily observed in the GB1 spectra where at 500 MHz (1H frequency) (Fig. 1a) there is some overlap of resonances, predominantly in the methyl region (Fig. 2a) close to the diagonal and in the Cα-Cβ region where peaks can not be baseline resolved from one another. However moving up in magnetic field (Fig. 1b–c, Fig. 2b–c) the resolution is improved to the point that every peak in the 900 MHz (1H frequency) spectra can be individually baseline resolved. The methyl region specifically shows amazing improvement since even the peaks that are close to the diagonal become resolved, not only from other cross peaks, but also from the diagonal itself.
This trend can also be seen in larger proteins, like DsbA, which shows the resolution enhancement with increasing magnetic field, just as GB1 does (Fig. 3). Although DsbA has a high degree of diversity of secondary structure elements (Martin et al., 1993; Guddat et al., 1998; Schirra et al., 1998) giving peaks within specific amino acids types greater chemical shift dispersion (Wishart and Sykes, 1994; Cornilescu et al., 1999; Oldfield, 2002), there are still many overlapped regions in the spectra.
The resolution enhancement is most prominently observed in the aliphatic Cα-Cβ region (Fig. 4) and the methyl region (Fig. 5) of the 2D 13C-13C DsbA spectra acquired at 500 MHz (Fig. 4a--5a),5a), 750 MHz (Fig. 4b--5b),5b), and 900 MHz (Fig. 4c--5c).5c). Despite the presence of 189 residues in this protein, the majority of aliphatic correlations (55–65 ppm in F1, 25–40 ppm in F2) are uniquely resolved at 900 MHz (Fig. 4c). Although some chemical shift patterns (Ala, Thr, Val) are sufficiently resolved even at 500 MHz, the more congested methyl region (20–26 ppm in F2) illustrates a particularly significant improvement in resolution (Fig. 5). In the 900 MHz spectrum of DsbA, a total of 234 methyl peaks can be uniquely identified and integrated, compared with only 99 peaks in the 500 MHz spectrum.
To quantify the narrowing of all three proteins, comparisons of line widths for all resolved aliphatic resonances were made (limited by the number of sites resolved in the lower field spectra). Linewidths were determined in Sparky (Goddard and Kneller) using Gaussian peak integration methods. Each protein was analyzed with the same protocol to compare line widths among data sets acquired at different magnetic fields. Peaks were chosen for analysis based on three prerequisite criteria: (1) signal-to-noise ratio greater than 10:1, (2) resolved in all three spectra, and (3) integrated with an experiment v. fit RMSD of 10% or less. The same set of peaks (resolution-limited by the lowest field data) was chosen for analysis among the three data sets acquired at 500, 750, and 900 MHz (1H frequencies). Multiplet structure (arising from 13C-13C scalar couplings) was generally not observed due to the apodization functions employed; therefore the total line width includes contributions from homogeneous relaxation, multiplet structure and the line broadening added. These factors together result in overall line shapes that are well approximated by a Gaussian fitting function (as evidenced by the RMSD fit values of 10% or better in all cases, and typically better than 2%). The line widths are reported as the fitted line width minus the line broadening used. Plotting the line widths as a function of B0 field (Fig. 6), we find a nearly linear improvement in resolution, arising from the approximately constant absolute line widths. This result indicates that the absolute homogeneous contributions to the linewidth (such as finite 1H decoupling and residual 13C-13C dipolar couplings) are to first approximation invariant with magnetic field, and scalar couplings determine the absolute linewidth. In general all three spectral regions, including the methyl (CH3; Fig. 6a), methane (CH; Fig. 6b), and methylene (CH2; Fig. 6c) signals, show similar behavior for the two proteins (GB1, DsbA) with enhanced benefit from increased magnetic field.
In addition to the resolution enhancement, we observe signatures of chemical exchange solely at ultra-high magnetic fields. For example, in the methyl region of the 900 MHz DsbA spectrum, we observed two sets of chemical shifts for the side chain 13C resonances (Cβ, Cδ1, Cγ1, and Cγ2) of I102 and I113 (Fig. 7c; Table 1). These intermediate-to-slow exchange events are more prominently observed at high field, where the chemical shift difference significantly exceeds the exchange rate; conversely, slowing the exchange by decreasing temperature did not provide sufficiently well resolved spectra to identify the multiple conformations of Ile residues.
In conclusion, our experimental results illustrate that systematic and substantial resolution benefits are observed at ultra-high magnetic field. Here we evaluated these effects directly by comparison of practically identical samples at three magnetic fields, where most relevant hardware performance issues were highly similar. The improved resolution at ultra-high field promises to accelerate interpretation of MAS SSNMR data on large, macroscopically disordered proteins of broad interest, when used in combination with 3D experiments to establish 15N correlations along the peptide backbone.
The authors thank the National Institute of Heath for funding through NIGMS (GM073770), NIGMS/Roadmap Initiative (GM075937) and Molecular Biophysics Training Grant (to LJS and AJN), David Hoyt, Jesse Sears, and Paul Ellis at the Environmental Molecular Science Laboratory (a national scientific user facility sponsored by the Department of Energy Office of Biological and Environmental Research) located at Pacific Northwest National Laboratory and operated for DOE by Batelle for their assistance in acquiring the 900 MHz data, Dr. Donghua Zhou for pulse sequence code, Dr. Trent Franks and Benjamin Fisher of the VOICE NMR Facility for technical assistance.