Experiments Optimized for Magic Angle Spinning and Oriented Sample Solid-State NMR of Proteins

Abstract

Structure determination by solid-state NMR of proteins is rapidly advancing as result of recent developments of samples, experimental methods, and calculations. There are a number of different solid-state NMR approaches that utilize stationary, aligned samples or magic angle spinning of unoriented ‘powder’ samples, and depending on the sample and the experimental method can emphasize the measurement of distances or angles, ideally both, as sources of structural constraints. Multi-dimensional correlation spectroscopy of low-gamma nuclei such as ¹⁵N and ¹³C is an important step for making resonance assignments and measurements of angular restraints in membrane proteins. However, the efficiency of coherence transfer predominantly depends upon the strength of dipole-dipole interaction, and this can vary from site to site and between sample alignments, for example, during the mixing of ¹³C and ¹⁵N magnetization in stationary aligned and in magic angle spinning samples. Here, we demonstrate that the efficiency of polarization transfer can be improved by using adiabatic demagnetization and remagnetization techniques on stationary aligned samples; and proton assisted insensitive nuclei cross-polarization in magic angle sample spinning samples. Adiabatic cross-polarization technique provides an alternative mechanism for spin-diffusion experiments correlating ¹⁵N/¹⁵N and ¹⁵N/¹³C chemical shifts over large distances. Improved efficiency in cross-polarization with 40% – 100% sensitivity enhancements are observed in proteins and single crystals, respectively. We describe solid-state NMR experimental techniques that are optimal for membrane proteins in liquid crystalline phospholipid bilayers under physiological conditions. The techniques are illustrated with data from both single crystals of peptides and of membrane proteins in phospholipid bilayers.

Introduction

Because membrane proteins reside in the chemically and dynamically heterogeneous liquid crystalline environment of phospholipid bilayers rather than a uniform environment, such as those of aqueous solutions or crystals, they require NMR experiments that take their unique properties into account. Notably, in “…-the so-called physiological state.”¹, membrane proteins undergo fast rotation about the bilayer normal and lateral diffusion in the plane of the bilayer, although they are immobilized on NMR timescales (<10⁴ Hz) in other dimensions^2–4. Taken together these limited, but rapid motions give membrane proteins mostly ‘solid-like’ spectroscopic characteristics; for example, their ¹H NMR spectra are very broad and barely detectable.

Samples that consist of protein-containing phospholipid bilayers (proteoliposomes) under physiological conditions of temperature and pH provide near-native conditions for the study of functional membrane proteins. This is generally not possible with competing methods. NMR methods optimized for studies of membrane proteins in liquid crystalline phospholipid bilayers are described in this article, which is complementary to our recent reviews on related topics that relied on earlier versions of these methods^{5, 6}.

Separated local field (SLF)^7–9 and homonuclear spin-exchange^10–12 experiments are among the mainstays of solid-state NMR of proteins. Two-dimensional versions of these experiments provide substantial spectroscopic resolution, and they can be readily expanded into three- and higher- dimensional experiments for larger proteins, and the measurement of additional frequencies, splittings, and powder patterns associated with individual sites. SLF experiments provide high spectral resolution due to the different orientational dependencies of the anisotropic heteronuclear dipolar and chemical shift interactions, and the observed frequencies provide measurements of the angles between bonds (and functional groups) and the direction of alignment. Homonuclear spin-exchange experiments provide an alternative approach to spectral resolution and semi-quantitative distance measurements, which, among other advantages, can provide assignments of resonances due to the correlation of resonances from proximate sites.

Oriented sample (OS) solid-state NMR is particularly well suited to the study stationary, uniaxially aligned samples⁹, such as membrane proteins in magnetically aligned bicelles¹⁰ or macro discs¹², or mechanically aligned bilayers on glass plates¹³. In OS solid-state NMR the basic SLF and spin-exchange experiments provide up to three orientationally-dependent frequencies for each isotopically labeled site (e.g., ¹H-¹⁵N dipolar coupling, ¹H chemical shift, and ¹⁵N chemical shift), and this not only resolves among many of the protein’s signals, even in crowded spectral regions, but also yields adequate input for structure calculations in aligned samples. In the past, the majority of experiments have relied on direct ¹⁵N-detection, and sensitivity has been a limitation because of ¹⁵N’s low gyromagnetic ratio; selectively or uniformly ¹⁵N labeled samples are essential because of the low natural abundance of the nucleus. Furthermore, while assignment schemes that work well with regular secondary structure, often by taking advantage of spectral patterns resulting from selective isotopic labelling, have been implemented in multiple membrane proteins^{5, 6}, it has been difficult to develop methods for efficiently assigning resonances in regions of proteins with irregular tertiary structure, such as inter-helical loops or the structured portions of the N- and C- termini.

Despite spectroscopic developments that enable the use of uniformly ¹³C and ¹⁵N labeled proteins, and the detection of nuclei with higher gyromagnetic ratios (and consequently sensitivity), i.e., ¹H and ¹³C, limitations remain for stationary samples due to the network of ¹³C-¹³C homonuclear dipolar couplings, which are difficult to decouple without compromising the signal-to-noise ratio during direct ¹³C-detection. Random dilute labelling enables some experiments¹⁴ because of the statistical isolation of the ¹³C nuclei, although it requires significant compromises including reduced sensitivity and difficulty in performing homonuclear spin-exchange experiments. Consequently, magic angle spinning (MAS) solid-state NMR methods, which average out anisotropic interactions, including homonuclear ¹³C-¹³C dipolar couplings, to first order, and yield effectively isotropic resonances in ‘solution-like’ NMR spectra¹⁵ play essential roles in NMR experiments on uniformly ¹⁵N and ¹³C labeled proteins. Notably, there are now pulse schemes that can be combined with MAS to selectively recouple the anisotropic interactions averaged out by the mechanical rotation¹⁶, which provides significant flexibility in experimental design.

It has been demonstrated that angular measurements obtained from ¹H-¹⁵N and ¹H-¹³C dipolar couplings and from ¹⁵N and ¹³C chemical shift anisotropy are equivalent, whether made from unoriented samples in stationary or MAS solid-state experiments based on the angular dependence of the spin-interactions relative to the axis of rotational motion or the direction of alignment magnetically aligned samples using oriented sample (OS) solid-state NMR methods. Initial NMR experiments optimized for both MAS and OS approaches have been described, enabling the measurement of heteronuclear dipolar couplings and chemical shift frequencies as angular restraints for structure calculations^{17, 18}. In this article we describe several elaborations of solid-state NMR methodology designed specifically for samples of membrane proteins in liquid crystalline phospholipid bilayers with an emphasis on protein structure determination. The efficacy of adiabatic demagnetization and remagnetization pulse scheme in static OS solid-state NMR is demonstrated on single crystal and membrane protein samples. It is shown that higher efficiency in magnetization transfer among homo and hetero nuclei is achievable using a common dipolar bath. Mixing of magnetization between ¹³C and ¹⁵N nuclei is an essential feature of these pulse schemes that transfers magnetization efficiently even when the strength of dipolar interaction is weak. Pulse schemes for multi-dimensional experiments with the measurement of ¹H-¹⁵N dipolar couplings are demonstrated for magnetically aligned samples and magic angle spinning of rotationally aligned samples.

Experimental Methods

Sample Preparation

Uniformly ¹⁵N labeled Pf1 bacteriophage samples were obtained by following the previously published protocol¹⁹. The samples for the stationary solid-state NMR experiments were prepared with a protein concentration of ~40 mg/ml. The pH of the sample was adjusted to 8.0 using 5mM sodium borate buffer containing 0.1 mM sodium azide. A 5-mm thin wall glass tube containing ~150 μL of sample was used for all stationary sample experiments. Under these conditions, the bacteriophage particles align with their long axis parallel to the direction of the applied magnetic field.

For magic angle spinning solid-state NMR experiments, the unoriented proteoliposome samples were prepared following the protocols described in previous publications¹⁸. Pf1 coat proteins were extracted from bacteriophage particles by separating the DNA in a mixed organic solution (50% trifluoroethanol, 0.1% trifluoroacetic acid and 49.9% water). The precipitated DNA was pelleted by centrifugation. The supernatant was subjected to nitrogen gas flow for partial removal of solvent followed by lyophilization. In order to completely remove any residual trifluoroacetic acid, water was added and the lyophilization process was repeated twice. NMR samples were prepared by reconstituting the uniformly ¹³C, ¹⁵N labeled proteins in 1,2-dimyristoyl-sn-glycero-3-phosphocholin (DMPC) lipid bilayers following the previously described¹⁸ methods. ¹⁵N labeled or uniform ¹³C, ¹⁵N labeled mercury transporter protein (MerFt) were reconstituted in DMPC liposome vesicles at a lipid to protein ratio 5:1 (w/w) and nearly 30 μL of samples were packed in a 3.2 mm rotor. At this concentration, the two-transmembrane helix protein, MerFt, undergoes rapid rotational motion along the bilayer normal above the gel to liquid crystal phase transition temperature.

NMR experiments

¹⁵N detected multi-dimensional experiments on magnetically aligned samples were carried out on a Bruker Avance spectrometer with a ¹H resonance frequency of 700.1 MHz. ¹³C detected experiments were performed on a Varian Inova 500 MHz spectrometer. Both the spectrometers were equipped with homebuilt 5 mm modified Alderman Grant coil (MAGC) double- and triple-resonance probes^20–22.

For one-dimensional experiments on stationary samples, CP-MOIST²³ was employed for cross-polarization, typically with 50 kHz radio frequency (RF) applied to the ¹H and ¹⁵N- channels. SPINAL-16 modulation was used for ¹H decoupling²⁴ and continuous wave irradiation was used for either ¹³C or ¹⁵N decoupling. Two-dimensional high resolution separated local field spectra were obtained using either the SAMMY²⁵ or polarization inversion spin exchange at magic angle (PISEMA)²⁶ pulse sequences that correlates the ¹⁵N chemical shift with the associated ¹H-¹⁵N dipolar coupling, respectively. 41 kHz RF was used during magic sandwich and cross polarization pulses. ¹⁵N detected one-dimensional data in Figure 1 were collected with 2 and 64 scans on single crystal and bicelle samples respectively. Two- and three-dimensional SLF spectra with ¹⁵N/¹⁵N homonuclear correlation were obtained with 8–128 scans for single crystal and bacteriophage samples. 32 real data points with a dwell time of 120 μs and 40 μs were collected for both ¹⁵N chemical shift and ¹H-¹⁵N dipolar frequency evolution periods, respectively. All data were recorded with a 4s recycle delay. The aligned bacteriophage sample temperature was maintained at 0 ⁰C for optimal resolution²⁷. The ¹H carrier frequency was set following the optimum results observed for each of the samples of bicelle, single crystal and bacteriophages. During adiabatic cross-polarization, ¹H RF was varied from 50 kHz to 1 kHz using an amplitude modulated half-Gaussian shape pulse and ¹⁵N magnetization was locked at 12 kHz RF field. Half-Gaussian pulses used for adiabatic demagnetization and remagnetization were optimized to maximum RF amplitude of 20 kHz during mixing. For triple resonance experiments, the ¹³C carrier frequency was set at 55 ppm; and the ¹⁵N carrier frequency was set at 80 ppm. The spectra were acquired with 256 complex points and 40 μs dwell time in the direct dimensions, and 16 scans for signal averaging. 36 complex points were acquired in the indirect dimension with 200 μs dwells. The magnetization was transferred from ¹H to ¹⁵N with 1 ms CP-MOIST at 50kHz of B₁ field for both channels. 60 kHz SPINAL-16, 30 kHz SPINAL-16, and 50 kHz continuous wave (CW) RF were used for decoupling on ¹H, ¹³C, and ¹⁵N channels. Adiabatic passage ¹³C-¹⁵N transfer was made with 20 kHz B₁ field on ¹³C and ¹⁵N channels for 3.9 ms, and adiabatic ¹³C-¹⁵N transfer was made with 25 kHz B₁ field on ¹³C and ¹⁵N channels for 8ms. Amplitude modulated half-Gaussian pulses were used for adiabatic demagnetization and remagnetization.

Figure 1.

¹⁵N NMR spectra of a peptide crystal and a protein obtained by different methods of cross-polarization. A. Spectrum of a single crystal of ¹⁵N-Acetyl Leucine (NAL) at an arbitrary orientation obtained using CP-MOIST. B. Same as in Panel A, except that the spectrum was obtained using CP-ADRF. C. CP-MOIST spectrum of uniformly ¹⁵N labeled Pf1 coat protein in a magnetically aligned unflipped q=3.2 DMPC/DHPC bicelles with the bilayer normal perpendicular to the direction of the magnetic field. D. CP-ADRF spectrum of Pf1 bicelle sample. All spectra were obtained for cross-polarization contact periods between 1ms (A, C and D) and 3 ms for B. All of the spectra were obtained on a Bruker Avance spectrometer with a ¹H resonance frequency of 700 MHz using a home-built double-resonance probe. 50 kHz RF amplitude SPINAL modulated decoupling was applied during data acquisition. Spectra were recorded at room temperature for single crystal and 42 ºC for a bicelle sample.

Magic angle spinning (MAS) experiments were performed on a 750 MHz Bruker Avance spectrometer equipped with a Bruker low-E ¹H/¹³C/¹⁵N triple-resonance 3.2 mm MAS probe. The spinning rate was controlled to 11.11 kHz ± 2 Hz using a Bruker MAS controller. During ¹³C detection, 100 kHz RF irradiation was used for ¹H decoupling with swept frequency two-pulse phase modulation (SW-TPPM)²⁸, and 2.5 kHz RF irradiation with WALTZ-16 was used for ¹⁵N decoupling. Cross-polarization (CP) from ¹H to ¹⁵N was optimized using amplitude-modulated RF irradiation (100-50%) applied on the ¹H channel; 50 kHz RF irradiation on the ¹H and ~27 kHz RF irradiation on the ¹⁵N channels were applied during the 1 ms contact time. ¹⁵N to ¹³C polarization transfer was accomplished using either proton-assisted insensitive nuclei CP (PAIN-CP)²⁹ or by double CP pulse schemes. Double CP was accomplished by transferring magnetization from ¹⁵N to ¹³CA or ¹³C’ using spectrally induced filtering in combination with cross-polarization (SPECIFIC-CP)³⁰. Adiabatic tangential pulses were applied on the ¹³C channel. Typical RF irradiations of 27 kHz were applied on the ¹⁵N channel and ~16 kHz on the ¹³C channel for band-selective polarization transfer. SW-TPPM decoupling with 100 kHz RF irradiation was applied during the double CP transfer. 50 kHz, 39 kHz and 55 kHz CW RF were applied on the ¹³C, ¹⁵N and ¹H channels for spin locking during PAIN-CP. Sample temperature was calibrated externally using ethylene glycol and water peak in proteoliposome samples. Experiments were carried out at sample temperature between 25–30 ºC.

¹H-¹⁵N dipolar recoupling experiments were carried out using either polarization inversion time averaged nutation spin exchange at magic angle (PITANSEMA)³¹ or symmetry based pulse sequence³² at 11.111 kHz sample spinning. A nutation frequency of 62.5 kHz at a resonance-offset frequency of 44.2 kHz was used during the frequency switched Lee-Goldburg (LG) cycles. An effective nutation frequency of 76.5 kHz was calculated for Lee-Goldburg (LG) pulses with a dwell time of 13.07 μs. A time modulation of 15.4 μs and 7.7 μs for +LG and −LG period was used to average the nutation to one third. The Hartmann-Hahn match to ¹⁵N was accomplished with an RF field match to the first order sidebands. Amplitude modulation on the ¹⁵N channel was used at nutation frequencies of 36.5 kHz and 14.3 kHz during the phase alternated LG irradiations. The experimentally determined scaling factor of 0.37 was taken into account in data processing.

Two and three-dimensional experiments with R18⁷₁ symmetry based pulse scheme were recorded at a spinning speed of 11.111 kHz. Two consecutive π pulses of 5 μs duration each and phase shifted by 70º and −70º were used in the R18⁷₁ pulse scheme³². Dipolar frequency measurement with ¹⁵N detection was carried out with a dwell time of one rotor period for a total dipolar evolution period of 5 ms, 4s recycle delay and 512 scans. 1.8 and 3.6 ms mixing periods were used for PAIN-CP and SPECIFIC-CP. 10 ms chemical shift evolution under proton irradiation were used to record signals from ¹⁵N and ¹³C nuclei.

The chemical shift frequencies were referenced externally to the adamantane methylene ¹³C resonance at 38.48 ppm and the ammonium sulfate ¹⁵N resonance at 26.8 ppm.

Results

Oriented sample solid-state NMR

An advantage of direct spin manipulations and direct-detection of low γ, dilute nuclei, such as ¹⁵N, is that their weak homonuclear dipole-dipole interactions do not require decoupling at any stage of the experiments. However, many experimental procedures, including those that lead to assignments of resonances among proximate nuclei and the semi-quantitative measurement of inter-nuclear distances, are based on spin-exchange among like nuclei, and these are generally inefficient for low γ, dilute nuclei. Several techniques have been proposed to enhance spin-exchange among dilute like-nuclei, e.g., ¹⁵N, in applications of correlation spectroscopy. Notably, proton driven spin diffusion (PDSD)³³, cross relaxation driven spin diffusion (CRDSD)³⁴, and Hartmann-Hahn mismatch mixing (HHMM)³⁵ schemes have been shown to be useful methods to correlate ¹⁵N sites in oriented sample (OS) solid-state NMR. Using these methods, sequential resonance assignments have been made in membrane proteins.³⁶. The performance of the above spin diffusion techniques has been compared by Traaseth et al³⁷. Additionally, spin diffusion through cross-relaxation via local fields can correlate dilute nuclei, e.g., ¹³C or ¹⁵N that are separated by relatively large distances.

Adiabatic cross-polarization via an intermediate dipolar-ordered state can significantly increase polarization of dilute nuclei compared to conventional cross-polarization (CP) techniques such as Hartmann-Hahn spin-lock matching and through-bond polarization transfer³⁸. Adiabatic demagnetization/remagnetization in the rotating frame (ADRF/ARRF) converts the Zeeman order of the abundant nuclei to dipolar order, and then, into Zeeman order of the dilute nuclei for detection. Several adiabatic cross-polarization schemes have been proposed that polarize the dilute nuclei to an extent that is many times greater than possible through direct polarization^{38, 39}. It has been shown that CP-ADRF in combination with polarization inversion spin exchange at magic angle (PISEMA²⁶) can provide very high sensitivity in two-dimensional SLF experiments⁴⁰. ADRF/ARRF has also been shown to facilitate spin diffusion.⁴¹

Adiabatic cross-polarization from ¹H to ¹⁵N in single crystal and aligned protein samples is demonstrated with the spectra in Figure 1. Signals obtained with conventional Hartmann-Hahn cross-polarization under phase alternated pulses, CP-MOIST (cross-polarization mismatch optimized IS transfer)²³, are shown in Figure 1A. CP-MOIST is an alternative to conventional continuous-wave spin-lock cross-polarization under matched Hartmann-Hahn conditions⁴² that is more tolerant to RF amplitude mismatch. In practice, CP-MOIST is less tolerant towards ¹H irradiation frequency offset and as a result sensitivity is compromised (data not shown here). Moreover, equal or superior performance is observed in CP-ADRF (Figure 1B and D) with low power RF irradiation resulting in less sample heating of electrically lossy samples, such as hydrated membrane proteins in phospholipid bilayers. Reduction in sample heating is essential for maintaining the stability of proteins at high resonance frequencies²⁰. Sensitivity enhancement resulting from the use of CP-ADRF is demonstrated by the comparison of intensities in the spectra in Figure 1A and B. In a single crystal, increase in intensities of up to a factor of two can be observed. In membrane proteins sensitivity enhancements in the 20%–40% range are typically observed, reduced from those seen in crystalline compounds largely due to the short relaxation time (T_1D) of the dipolar reservoir that is a consequence of the protein’s rotational diffusion about the bilayer normal. Nonetheless, the enhancement factor along with low RF sample heating and broad excitation profile are very helpful in the execution of OS solid-state NMR studies of membrane proteins.

The spectra of crystals and proteins in Figure 1 were obtained with initial preparation of ¹H transverse magnetization using a hard 90º pulse. Magnetization was then transferred from Zeeman order to dipolar order adiabatically using an amplitude modulated half-Gaussian pulse. The ¹⁵N nuclei were polarized with low power continuous wave RF irradiation. It is noteworthy that a simultaneous adiabatic demagnetization pulse applied on ¹H and remagnetization pulse on ¹⁵N channel can be used to create the polarization for ¹⁵N nuclei. The dipolar reservoir can also be used to propagate spin diffusion if contact with the rare spins can be established. Figure 2 contains diagrams for a selection of pulse schemes that utilize adiabatic demagnetization and remagnetization pulses to facilitate spin diffusion among dilute nuclei. Homonuclear correlation spectra with ¹⁵N detection were acquired using either of the pulse schemes shown in Figure 2A and B.

Figure 2.

Timing diagrams for two- and three- dimensional correlation experiments on stationary and magic angle sample spinning. A. and B. ¹⁵N detected homonuclear correlation spectroscopy with ADRF-CP and adiabatic mixing with demagnetization-remagnetization of ¹⁵N nuclei in static oriented samples. C. and D. ADRF-CP and adiabatic passage CP (APCP) in ¹³C detected ¹³C/¹⁵N two-dimensional correlation in oriented samples respectively. E. Three-dimensional separated local field spectroscopy correlating ¹H-¹⁵N/¹⁵N/¹⁵N frequencies. Half-Gauss line shapes with amplitude modulation are used for demagnetization and remagnetization. G. and H. ¹³C detected three-dimensional separated local field spectroscopy for magic angle spinning experiments correlating ¹H-¹⁵N dipolar coupling with ¹⁵N/¹³C shift correlation. G, Adiabatic passage CP mixing under spectrally induced filtering in combination with cross-polarization (SPECIFIC-CP). H, Proton assisted insensitive nuclei CP (PAIN-CP) mixing. Thin and wider dark lines represent the hard 90º and 180º pulses. Rectangular boxes represent either continuous wave or SPINAL modulated RF pulses. Tangential line shapes are used for adiabatic passage cross polarization (APCP). The phase cycling for G and H are as follows; φ₁= 1 3, φ₂= 0 0 2 2, φ₃= 1 1 3 3, φ₄= 3, φ₅= 1, φ₆= 8(0), φ₇= 4(0) 4(2), φ₈= 4(0) 4(1) 4(2) 4(3), φ₉=2(0) 2(2), φ₁₀= 4(0) 4(2) and receivers at φ_DCP=φ₁₊φ₇₊φ₈ and φ_PAIN=φ₁₊φ₉₊φ₁₀.

In the examples of spectra of a single crystal of ¹⁵N-acetyl-leucine (NAL) at an arbitrary orientation and uniformly ¹⁵N labeled, magnetically aligned Pf1 bacteriophage in Figure 3, the ¹⁵N nuclei were polarized using ADRF prior to chemical shift evolution in the indirect dimension. The SLF spectra in Figure 3A and F show the heteronuclear ¹H-¹⁵N dipolar couplings associated with the chemical shift frequencies. Figure 3 also contains homonuclear spin-exchange correlation spectra. The single crystal spectra with signals from four unique sites are in the top row and the bacteriophage coat protein spectra with signals from >40 unique amide sites are in the bottom row. Following cross polarization, a flip back 90º pulse was applied to select cosine and sine modulated shift coherences. A z-filter period was used to eliminate any remaining transverse magnetization while preserving the signals of interest. During the mixing period, Zeeman order of the ¹⁵N nuclei was transferred to dipolar order using adiabatic demagnetization, spin lock, or adiabatic remagnetization-demagnetization pulses to propagate the dipolar order through ¹H/¹H spin diffusion. The transfer of ¹⁵N dipolar order back to Zeeman order was achieved using adiabatic remagnetization. The spin diffusion among ¹H can be partially or fully suppressed using almost any homonuclear decoupling pulse scheme, such as Lee-Goldburg⁴³, or heteronuclear decoupling irradiations⁴⁴. There is no correlation spectrum (only diagonal peaks) observed in the presence of heteronuclear decoupling, which confirms that spin-diffusion occurs via dipolar order. The spectrum in Figure 3H was obtained with ¹H decoupling during the mixing period while adiabatic demagnetization-remagnetization of the dilute ¹⁵N nuclei was active. No cross peaks were observed, as expected, since dipolar order cannot be transferred since heteronuclear decoupling removes any contact with the local field reservoir.

Figure 3.

Experimental two-dimensional ¹⁵N-detected spectra. Top row: A, B., C., D., and E. are of a single crystal of NAL at an arbitrary orientation. Bottom row: F., G., H., I, and J. are of a magnetically aligned sample of uniformly ¹⁵N labeled Pf1 bacteriophage at 0 ºC. A. SLF spectrum obtained with the SAMPI4 pulse sequence. B. ¹⁵N/¹⁵N correlation spectrum obtained from a 4s proton driven spin diffusion mixing time. C. Same as Panel B but obtained using the pulse scheme shown in Figure 2A and a total adiabatic mixing time of 6 ms. D. and E. are two-dimensional planes extracted from a three-dimensional spectrum at dipolar frequencies of 5 kHz and 3 kHz, respectively. F. Same as Panel A. but obtained with 2ms adiabatic mixing using the pulse scheme shown in Figure 2E and without the incorporation of ¹⁵N shift evolution period (t₁). G. same as Panel C. but obtained with 2 ms mixing. H. same as G. but obtained under heteronuclear decoupling pulses applied on the ¹H channel and simultaneous adiabatic mixing on the ¹⁵N channel. I. and J. Same as in D. and E. but extracted at dipolar frequencies of 9.7 kHz and 6.7 kHz, respectively. Resonance assignments are marked.

PDSD with a 4 sec mixing period results in correlation among the four uniquely oriented ¹⁵N sites in the unit cell of the NAL crystal (Figure 3B). However, the transfer of magnetization is relatively weak, which is why the cross-peak intensities are much lower than those of the diagonal signals. In contrast, with adiabatic mixing, uniform polarization transfer occurs among all four signals (Figure 3C). This may result from all of the nuclei being connected to a common dipolar reservoir enabling transfer of dipolar order to Zeeman order to occur equally. ¹⁵N/¹⁵N shift correlation spectrum for a uniform ¹⁵N labeled Pf1 bacteriophage sample is shown in Figure 3G. A number of high intense cross-peaks are easily distinguishable in this two-dimensional spectrum.

The two-dimensional SLF spectrum of the NAL single crystal is shown in Figure 3A. The spectrum was recorded using the high resolution SLF experiment, SAMPI4²⁵. The equivalent SLF spectrum obtained for the coat protein in Pf1 bacteriophage particles with adiabatic mixing is quite crowded (Figure 3F). The spectrum was recorded using the pulse scheme shown in Figure 2E in a two-dimensional fashion without ¹⁵N shift evolution. The data represent the cross-correlation among dipolar-coupled systems connected to a common dipolar bath. The resolution of the protein spectrum is greatly enhanced in three-dimensional SLF/correlation experiments that result in a correlation of ¹H-¹⁵N dipolar coupling frequencies and the ¹⁵N/¹⁵N chemical shift frequencies in three orthogonal dimensions. The pulse scheme shown in Figure 2E was used to obtain the three-dimensional spectra. In this experiment, ¹⁵N magnetization was polarized using the CP-MOIST technique prior to spin exchange under frequency switched Lee-Goldburg decoupling. Following the spin exchange period, ¹H decoupled ¹⁵N chemical shifts were encoded in the first indirect dimension. A 90º flip back pulse was applied to select cosine and sine modulated shift coherences. A z-filter period was used to eliminate any remaining transverse magnetization while preserving the signals of interest. Spin-diffusion between ¹⁵N nuclei was carried out under adiabatic demagnetization/remagnetization mixing followed by the direct detection of free chemical shift evolution under proton decoupling. Dipolar evolution during the spin exchange period was encoded in a third dimension. Two-dimensional planes correlating ¹⁵N/¹⁵N shifts extracted at 5 kHz and 3 kHz ¹H-¹⁵N dipolar coupling frequencies from the three-dimensional spectrum of the single crystal are shown in Figure 3D and E, respectively. The two-dimensional planes shown in Figure 3I and J were extracted at 9.7 kHz and 6.7 kHz ¹H-¹⁵N dipolar coupling frequencies, respectively, from the three-dimensional spectrum of the protein. Several resonance assignments are marked following the previously published assignments by Thiriot et al^{19, 27}.

Uniform ¹³C labeling plays a crucial role in complete sequential assignments of protein backbones. Hence, heteronuclear correlation of ¹³C and ¹⁵N signals is an essential building block for making resonance assignments. In OS solid state NMR, direct ¹³C detection improves sensitivity compared to ¹⁵N detection. However, it is not practical in all cases because, among other reasons, homonuclear dipolar coupling enhances the relaxation effect due to motions on inconvenient timescales, and therefore broadens the line width resulting in degradation of resolution. The performance of adiabatic pulse schemes is demonstrated on a ¹³Cα/¹⁵N labeled single crystal in Figure 4 with heteronuclear correlation spectra obtained using adiabatic passage and demagnetization/remagnetization techniques.

Figure 4.

¹³C detected two-dimensional ¹³C/¹⁵N correlation spectra obtained on a single crystal of ¹³Cα/¹⁵N labeled N-Acetyl Leucine (NAL). A. NMR spectrum obtained using the pulse scheme in Figure 2C. B. Spectrum obtained using the pulse scheme in Figure 2D. One-dimensional slices taken at the frequencies marked with dotted lines are shown along the sides of the two-dimensional plots.

The spectrum shown in Figure 4A was obtained with adiabatic demagnetization of the ¹⁵N nuclei and simultaneous remagnetization of the ¹³C nuclei. This was carried out using the pulse scheme shown in Figure 2C. Magnetization transfer takes place when both of these dilute nuclei are connected to the dipolar bath. Nearly half of the magnetization was recovered compared to the use of ¹H/¹³C CP-MOIST. Similarly, the ¹³Cα/¹⁵N correlation spectrum using the adiabatic passage cross polarization scheme during mixing is shown in Figure 4B. It was obtained using the pulse sequence diagrammed in Figure 2D. A tangential pulse was used for ¹³C during the cross polarization from ¹⁵N to ¹³C. Three out of four molecules in a unit cell have relatively strong ¹³Cα-¹⁵N dipolar couplings, and transfer 80%–90% of the magnetization. However, the signal at 64 ppm is relatively weak due to the fact that ¹³Cα-¹⁵N dipolar coupling at this orientation of the crystal is relatively weak. Polarization transfer via adiabatic passage shows only one bond correlation. Notably, the spectrum in Figure 4A shows uniform polarization transfer among all four molecules in the unit cell, and is an important demonstration that magnetization can be transferred across relatively long molecular distances via dipolar order propagation.

Magic angle spinning solid-state NMR

Magic angle spinning solid-state NMR spectroscopy enables high-resolution spectra to be obtained from unoriented ‘powder’ samples of proteins. However, it has the disadvantage of strongly attenuating the anisotropic properties of the nuclear spin interactions, such as the chemical shift and dipole-dipole couplings between neighbouring nuclei that are relied on for the experiments on stationary samples, as discussed above. Recent developments in decoupling and recoupling pulse schemes suspend the averaging effects of magic angle spinning and retain the geometrical information without sacrificing the high sensitivity and resolution provided by MAS. Heteronuclear dipolar coupling frequencies are recoupled by a wide variety of pulse schemes based on the Hartmann-Hahn match condition under MAS^{31, 45, 46}. There are numerous applications that use heteronuclear dipolar coupling measurements to characterize the structure and dynamics of small organic molecules⁴⁷ and biomolecules^{48, 49}. Symmetry-based pulse sequences are of particular importance in this approach⁵⁰. Applied on one or two channels they can restore the averaging effect on the other nuclei such as ¹³C and ¹⁵N during detection^{32, 51}. In particular, ¹H-¹⁵N dipolar coupling measurements with direct ¹³C-detection is a crucial development for increasing the sensitivity for applications to proteins and other biomolecules. Symmetry based pulse design offers a broad range of experiments that are applicable to moderate, fast and ultra-fast MAS techniques^{18, 32, 52}. In our recent studies of uniformly ¹³C and ¹⁵N labeled proteins, we used symmetry based pulse schemes to measure heteronuclear dipolar frequencies in membrane proteins under MAS¹⁸.

Recently developed pulse sequences³¹ are quite efficient in recoupling powder line shapes such as Pake doublets under MAS for uniformly labeled proteins. The measurements are carried out in two ways, either by proton evolved local field⁵³ or by SLF. PISEMA, originally developed for oriented solid state NMR, can also be applied in MAS experiments with the Hartmann-Hahn match to the first and second sidebands⁴⁵. This provides a maximum theoretical scaling factor of 0.577. Time averaged nutation incorporated into PISEMA (PITANSEMA) under MAS reduces the required RF power on the low γ nuclei³¹. The recoupled spectrum for a ¹H-¹⁵N dipolar frequency in a uniformly ¹³C/¹⁵N labeled NAL crystal is shown in Figure 5C.

Figure 5.

One dimensional Pake doublets for ¹H-¹⁵N dipole-dipole interactions in a poly crystalline sample of uniform ¹³C, ¹⁵N labeled NAL. A. Simulated Pake doublet for a dipolar coupling constant of 10 kHz. B. Experimental measurement of recoupled powder spectrum using R18⁷₁ symmetry pulse scheme under 11 kHz MAS. C. Experimental, Pake doublet extracted from a SLF-2D experiment acquired using PITANSEMA under 11 kHz MAS.

A series of symmetry pulses can also be used for recoupling of hetero-nuclear dipolar couplings³². R18⁷₁ and R18⁵₂ pulse schemes are particularly useful under moderate sample spinning speeds. Several other pulse sequences based on the symmetry principle with rotor synchronization have also been reported⁵² for ultra-fast spinning speed. R18⁷₁ consists of a pair of 180₇₀180₋₇₀ pulse pairs, where each 180° pulse occupies exactly 1/18 of a rotor period. A total of 9 repetitions of pulse pairs are applied over one rotor period and requires RF field strength of 9 times the spinning frequency. However, R18⁷1 irradiation leads to recoupling of ¹H–¹⁵N dipolar couplings and ¹H chemical shift anisotropy when applied on the ¹H channel. All other interactions notably homo-nuclear dipolar interactions among the ¹H spins are suppressed. This provides a theoretical scaling factor of 0.36 for R18⁷₁. A two-pulse phase modulation with 70° and −70° phases for each pulse length of 5 μs were used in R18⁷₁ cycle under 11.111 kHz spinning. A recoupled dipolar spectrum for a polycrystalline sample of NAL is shown in Figure 5B. Both of the experimental spectra in the Figure were rescaled according to the theoretical scaling factor. As can be seen the symmetry based pulse scheme selectively recouples the perpendicular edge of the Pake doublet with higher efficiency than other orientations in the crystals. SLF experiments in MAS solid-state NMR correlate the isotropic shifts with recoupled dipolar coupling (DC) frequencies, in contrast to OS solid-state NMR, where anisotropic shifts are correlated to heteronuclear DC frequencies.

Two-dimensional ¹³C or ¹⁵N detected SLF experiments that correlate ¹H-¹⁵N dipolar coupling frequencies with ¹³C or ¹⁵N isotropic chemical shifts are performed in two steps using the pulse scheme shown in Figure 2F. First, ¹³C or ¹⁵N magnetization is generated by cross-polarization from ¹H prior to dipolar evolution. Second, ¹H-¹⁵N dipolar evolution was carried out under R18⁷₁ symmetry pulse sequences to assist in the simultaneous recoupling of hetero-nuclear and decoupling of homonuclear dipolar couplings. ¹H-¹⁵N DC frequencies were encoded in a constant time manner with refocusing of ¹³C or ¹⁵N chemical shifts. The evolution of the isotropic chemical shift frequencies is measured under hetero-nuclear decoupling of ¹H in the direct dimension. In a similar fashion, the SLF experiment can be extended to ¹⁵N edited ¹³C detection. This can be achieved by adding a third step to the SLF pulse scheme. After dipolar evolution, ¹⁵N magnetization is transferred to either ¹³Cα or ¹³C′ using spectrally induced filtering in combination with cross polarization (SPECIFIC-CP)⁵⁴. The ¹³C signals are enhanced using adiabatic passage cross-polarization from ¹⁵N, which is then detected under ¹H decoupling. For comparison, two-dimensional ¹⁵N detected ¹H-¹⁵N/¹⁵N or ¹⁵N edited ¹H-¹⁵N/¹³C correlation spectra are shown in Figure 6. Single site resolution for ¹⁵N and ¹³C chemical shifts correlating ¹H-¹⁵N dipolar frequency in a uniform ¹³C, ¹⁵N labeled NAL polycrystalline sample is shown in Figure 6A, B and E respectively. As expected, similar resolution in dipolar slices but higher sensitivity in case of ¹³C detection are observed. The two-dimensional SLF spectra with ¹³C detection were obtained using the pulse scheme shown in Figure 2F and G without the incorporation of ¹⁵N shift evolution. Similar experiments were performed on a proteoliposome sample containing a 60-residue two-transmembrane helices mercury transporter protein (MerFt). ¹⁵N and ¹³C detected SLF-2D spectra recorded at 25 ºC are shown in Figure 6C and D respectively. At this temperature the protein undergoes fast rotational diffusion along the bilayer normal¹⁸.

Figure 6.

Two-dimensional SLF ¹⁵N and ¹³C detected ¹H-¹⁵N dipolar frequency measurements. Experiments were carried out using the pulse schemes shown in Figure 2F– H. A. ¹H-¹⁵N DC/¹⁵N shift correlation spectra on a single crystal sample of uniformly ¹³C/¹⁵N labeled NAL. Spectrum was recorded using the pulse scheme shown in Figure 2F. C. Same as Panel A. except for a 60 residue membrane protein, MerFt, in unoriented DMPC bilayers at 25 ºC. B. and D., same as A. and C. respectively but with ¹⁵N edited ¹³C detection. ¹H-¹⁵N DC/¹³Cα spectra were recorded using the pulse scheme shown in Figure 2G without the ¹⁵N shift evolution. E., same as Panel B. but acquired under PAIN-CP mixing using the pulse scheme shown in Figure 2H. Note that the full dipolar spectrum is scaled to show the perpendicular edges of the dipolar Pake pattern. 11 kHz MAS with 4s recycle delay was used to obtain the two-dimensional spectra.

Variations in dipolar frequency that range between 0 kHz to +/− 5 kHz (maximum for the parallel edge of the ¹H-¹⁵N Pake powder pattern) are observed due to the differential rotational alignment of sites in the membrane proteins along the bilayer normal. A ¹³C detected SLF spectrum shows better resolution and sensitivity compared to that obtained using ¹⁵N detection. A few assigned signals with dipolar and shift frequencies are noted following the previous publication¹⁸. Adiabatic passage cross-polarization at this temperature showed efficient polarization transfer for strong ¹³C-¹⁵N dipolar coupled spins but relatively weaker signals for small dipolar-coupled spins. This is a practical limitation to rotationally aligned samples where the ¹³C-¹⁵N-dipolar frequencies may vary over a range of 0–1000 Hz. On the other hand, PAIN-CP utilizes a dipolar coupled ¹H bath to propagate the dipolar order and transfer magnetization efficiently over long range distances even for spin systems with negligible ¹³C-¹⁵N dipole-dipole interaction²⁹. This is demonstrated on a uniform ¹³C, ¹⁵N labeled NAL poly crystals (Figure 6E) and uniformly ¹³C, ¹⁵N labeled Pf1 coat protein in DMPC proteoliposome sample (Figure 7). The 2D-spectrum in Figure 6E correlates ¹H-¹⁵N dipolar frequency with ¹³C shifts. Note that all the carbons within the molecule are correlated with uniform intensity. Two-dimensional ¹³C/¹⁵N shift correlation spectra in Figure 7 were recorded at 25 ºC with adiabatic passage/SPECIFIC CP (Figure 7A) and PAIN-CP (Figure 7B) mixing respectively. Adiabatic mixing shows selective transfer of magnetization to ¹³Cα shifts. Where as a number of carbon atoms within the same residue and nearby amino acids are correlated to a single ¹⁵N shift with PAIN-CP mixing. This is an important demonstration of distribution of ¹⁵N magnetization to distant ¹³C nuclei with negligible direct dipole-dipole interaction. Henceforth, PAIN-CP may be a method of choice in the study of membrane proteins. Higher resolution is achievable through incorporation of a ¹⁵N chemical shift evolution period as the third dimension. The first two-dimensional plane correlating ¹³C/¹⁵N shifts in MerFt is shown in Figure 8A. ¹⁵N shifts for residues 69, 31 and 14 are marked. Corresponding two-dimensional planes correlating ¹H-¹⁵N DC/¹³C shifts at the ¹⁵N shift frequencies for D69, L31 and G14 residues are shown in Figure 8B, C, and D, respectively.

Figure 7.

¹³C detected two-dimensional ¹³C/¹⁵N correlation spectra obtained on a uniform ¹³Cα/¹⁵N labeled Pf1 coat protein in DMPC bilayer. A. SPECIFIC CP mixing. B. PAIN-CP mixing. Both spectra were recorded at 25 ºC and 11 kHz MAS. 3.6 and 1.8 ms ¹³C/¹⁵N mixing periods were used to record spectra A. and B. respectively.

Figure 8.

Two-dimensional planes from a three-dimensional SLF experiment correlating ¹H-¹⁵N DC/¹⁵N/¹³C shifts. The experiment was carried out using the pulse scheme shown in Figure 2G. A. First ¹⁵N/¹³Cα correlation spectrum. B., C., and D. Planes correlating ¹H-¹⁵N DC/¹³Cα shifts at various ¹⁵N shifts. 12 ms, 6 ms and 5 ms evolution periods were used to obtain the ¹³C, ¹⁵N shifts and ¹H-¹⁵N dipolar coupling. Equal numbers of data points were linear predicted for the domains yielding the ¹⁵N shifts.

Conclusions

Although adiabatic methods have been available since the early days of high resolution solid-state NMR, they provided little in the way of specific advantages for single- and poly-crystalline samples. However, with the growth of solid-state NMR of proteins, their advantages become crucial components of the design and implementation of new experiments. They help deal with limitations imposed by both the dynamics of supramolecular assemblies, for example the rotational diffusion of virus particles and of membrane proteins in phospholipid bilayers, and the fundamental sensitivity limitations of dealing with low γ, dilute nuclei.

The use of adiabatic and other refinements are illustrated here for the two principal classes of solid-state NMR experiments. These are for stationary samples, where the full static dipolar and chemical shift interactions must be dealt with, and for samples undergoing magic angle spinning, where the primary objective is to recover some or all of the same spin interactions that have been averaged out by the sample rotation. Adiabatic cross-polarization method is shown to provide higher sensitivity and tolerance towards a number of experimental errors such as Hartmann-Hahn mismatch, RF inhomogeneity and carrier frequency offsets. It is also shown that adiabatic mixing distributes magnetization efficiently and in a fast order than PDSD scheme (5ms adiabatic mixing period compared to 4s in PDSD). Additionally, correlation spectroscopy is shown for low γ nuclei separated over large distances and negligible dipole-dipole interaction. This is demonstrated in Figure 3 and 4. Figure 4 illustrates the substantial gains in spin-exchange in a stationary sample by invoking adiabatic methods. However, the method may be limited by certain systems with short relaxation time T_1D of dipolar bath. The results shown in this article provide an alternative pathway to further development of pulse schemes for sequential assignment in OS using the adiabatic mixing method. The benefits of multi-dimensional experiments that exploit recoupling during magic angle spinning are shown in Figures 6–8. The flexibility in choice of pulse sequences enables physiological conditions of pH and temperature to be used and to avoid the addition of cryoprotectants and other agents needed for low temperature and sensitivity enhancement experiments. In combination these methods increase the feasibility of high-resolution NMR studies of proteins in the variety of situations needed to prepare functionally active samples of proteins in supramolecular assemblies.