¹H-Detected REDOR with Fast Magic-Angle Spinning of a Deuterated Protein

Abstract

Rotational echo double resonance (REDOR) is a highly successful method for heteronuclear distance determination in biological solid-state NMR, and ¹H detection methods have emerged in recent years as a powerful approach to improving sensitivity and resolution for small sample quantities by utilizing fast magic-angle spinning (>30 kHz) and deuteration strategies. In theory, involving ¹H as one of the spins for measuring REDOR effects can greatly increase the distance measurement range, but few experiments of this type have been reported. Here we introduce a pulse sequence that combines frequency selective REDOR (FSR) with ¹H detection. We demonstrate this method with applications to samples of uniformly ¹³C, ¹⁵N, ²H-labeled alanine and uniformly ¹³C, ²H, ¹⁵N-labeled GB1 protein, back-exchanged with 30% H₂O and 70% D₂O, employing a variety of frequency-selective ¹³C pulses to highlight unique spectral features. The resulting, robust REDOR effects provide (1) tools for resonance assignment, (2) restraints of secondary structure, (3) probes of tertiary structure, and (4) approaches to determine the preferred orientation of aromatic rings in the protein core.

TOC image

Introduction

Magic-angle spinning (MAS) solid-state NMR (SSNMR) spectroscopy is a powerful technique for elucidating the structures of insoluble proteins, especially amyloid fibrils and membrane proteins, and small molecule ligand-bound biological complexes^1–7. Three-dimensional (3D) structure determination of proteins generally relies upon large numbers of semi-quantitative distances and is greatly enhanced by the availability of quantitative distance restraint measurements. Rotational echo double resonance (REDOR) is a well-known SSNMR method for measuring heteronuclear distances^8–9, and it has been used extensively for studies of small molecules, proteins and bimolecular complexes^10–12. Analogous TEDOR approaches have been developed¹³ and applied to the determination of high-resolution protein structures^14–15. The measurable distance depends on the gyromagnetic ratio (γ) of the nuclei. Since ¹H has the highest gyromagnetic ratio of any stable isotope, the longest theoretical distance range for REDOR would be achievable with ¹H-¹⁹F pairs, as demonstrated by Hong and co-workers¹⁶ utilizing low-γ nucleus detection, followed by ¹H-³¹P and ¹H-¹³C.

Utilizing ¹H as the detected nucleus in such REDOR cases potentially adds another advantage of increased sensitivity. Significant technological advances in the field of SSNMR over the last decade have enabled the routine use of ¹H-detected experiments for biomolecular structure determination^17–20. The two main factors that have advanced ¹H detection are (ultra)-fast magic angle spinning (MAS), ranging from 30 to >110 kHz²¹ and complete or partial deuteration of the protein^21–22. The ¹H line width is inversely proportional to the spinning rate, and so even fully protonated samples yield sub-ppm line widths at 30 to 40 kHz MAS^17,23. Deuteration and partial back-exchange with H₂O further reduce the average ¹H-¹H dipolar couplings and in combination with fast MAS rates enhance the resolution of the ¹H-detected spectrum and transverse relaxation times (T₂)^24–25. State-of-the-art approaches now can often yield ~0.1 ppm or better amide ¹H linewidths^18,20–22.

Previous ¹H-X REDOR studies have emphasized spectral filtering based on coupling magnitude^26–28. The initial implementation, called medium and long distance heteronuclear correlation (MELODI-HETCOR), focused on suppressing the one-bond ¹H-¹³C correlation through dephasing of ¹H magnetization by REDOR π pulses on ¹³C, and detecting medium- and long-range ¹H-¹³C correlations in a HETCOR spectrum²⁶. This method was applied at moderate MAS rates (6 to 10 kHz), requiring a frequency-switched Lee-Goldburg (FSLG)²⁹ sequence to suppress the ¹H-¹H homonuclear couplings. An additional crucial requirement for this experiment is a partially ¹³C-labeled protein.

Here we introduce a modified pulse sequence that produces a ¹⁵N-¹H correlation spectrum in which the peak intensities depend on selective REDOR dephasing from ¹³C. Optimal resolution is obtained in a deuterated protein with fast MAS, negating the need for multiple pulse decoupling approaches for removing ¹H-¹H homonuclear couplings. Further, due to the increased ¹H T₂ relaxation times, the ¹H magnetization can be dephased for a longer time (on the order of 15–20 ms), in principle enabling longer ¹H-X distances to be detected with this pulse sequence (up to 15–20 Å). But since the coupling magnitudes are so large (>1 kHz) for many structurally informative distances (vide infra), even REDOR applied for ~1 ms is more than sufficient to reveal a variety of unique effects that depend on the details of protein structure. Band-selectivity of the REDOR dephasing³⁰ is achieved with frequency-selective soft π-pulses applied to the ¹³C nuclei. In addition to backbone (Cα and C’) selective versions for spectral filtering, we also evaluate cases in which the π-pulses are applied to the methyl and aromatic regions. These results provide insights into resonance assignment, secondary and tertiary structure, and preferred sidechain orientations.

Experimental Methods

Preparation of uniformly ¹³C, ¹⁵N, ²H-labeled Ala

¹³C₃, D₄, ¹⁵N labeled L-Ala was purchased from Cambridge Isotope Laboratories (Andover, MA), exchanged into 99.9% D₂O, lyophilized and then re-crystallized by slow evaporation of a saturated (~250 mg/mL) solution at room temperature. The crystals were crushed to make a fine powder sample and 4 mg was packed in a 1.6 mm FastMAS rotor (Revolution NMR, LLC, Fort Collins, CO).

Preparation of uniformly ¹³C, ¹⁵N, ²H-labeled GB1

Uniformly ¹³C, ¹⁵N, ²H labeled (U-CDN) GB1 protein was expressed in E. coli cells and purified according to Ref.³¹. The purified protein was dialyzed into 30:70 H₂O:D₂O buffer, concentrated to ~25 mg/mL and precipitated with 2-methyl-2,4-pentanediol (MPD) and isopropanol (IPA) solution according to Ref.³². Microcrystals were ultracentrifuged and ~4 mg packed into a 1.6 mm FastMAS rotor (Revolution NMR, LLC, Fort Collins, CO).

¹H-Detected frequency selective REDOR (FSR) pulse sequence

The pulse sequence (Fig. 1) starts with adiabatic cross-polarization (CP)³³ from ¹H to ¹⁵N, followed by evolution in t₁ with low power TPPM ¹H decoupling^34–35. After evolution, the ¹⁵N magnetization is stored along the z-axis in a constant time manner, and MISSISSIPPI solvent suppression is applied³⁶, followed by ¹⁵N-to-¹H CP. The FSR³⁰ period starts with in-phase ¹H magnetization, which evolves under REDOR by applying two π-pulses per rotor period on the dephasing channel (¹³C in this case). The ¹³C-¹³C dipolar interactions present in a fully ¹³C labeled protein will be reintroduced by the application of the π-pulse trains on ¹³C and hence will lead to some additional signal decay. We also attempted versions with one π pulse per rotor period on each channel (Fig. S1) or all the π pulses on ¹H; both resulted in less satisfactory performance due to ¹H-¹H recoupling effects. A simultaneous hard π-pulse on the ¹H channel and a soft frequency selective π-pulse on the dephasing channel are applied in the middle of the REDOR period (after the first train of REDOR π-pulses, with the hard π-pulse centered on the soft π-pulse), with extra rotor periods before and after the REDOR to ensure an in-phase echo for detection with minimal baseline distortion. The soft pulse on ¹³C will also lead to partial removal of the recoupled ¹³C-¹³C interaction. The S₀ experiment is performed with no REDOR π-pulse train on the dephasing channel, and the S experiment is performed with all the REDOR π-pulses. The ¹H magnetization is then detected in the direct dimension with WALTZ decoupling³⁷ applied to both ¹⁵N and ¹³C.

Fig. 1.

Pulse sequence for ¹HN-detected frequency selective REDOR spectroscopy. The 1H magnetization is transferred to ¹⁵N for indirect evolution and then transferred back to ¹H for detection. A REDOR period is added before detection to dephase the ¹H magnetization with nearby ¹³C atoms (more details in text). The phase cycles were as follows: ϕ1 = x,x,-x,-x; ϕ2 = x,x,x,x,-x,-x,-x,-x; ϕ3 = y,-y; ϕ4 = y,y,-x,-x; ϕ5 = y,y,x,x,x,x,y,y,-x,-x,y,y,y,y,-x,-x,-y,-y,-x,-x,-x,-x,-y,-y,x,x,y,y,y,y,x,x; ϕ6 = x,y,x,y,-x,-y,-x,-y; ϕ7 = x,-x,-y,y,x,-x,-y,y,-x,x,y,-y,-x,x,y,-y

Solid State NMR (SSNMR) Spectroscopy

Experiments on U-CDN Ala were carried out at 11.7 T (500 MHz ¹H frequency) on a Varian (Walnut Creek, CA and Loveland, CO) VNMRS spectrometer with a 1.6 mm HCDN FastMAS probe. Experiments on U-CDN GB1 (30% H₂O back exchanged) were carried out at 17.6 T (750 MHz ¹H frequency) on Varian VNMRS spectrometer with a 1.6 mm HFXY FastMAS probe configured in HFCN mode. On both the instruments, the spinning was controlled at 33333 ± 10 Hz by a Varian MAS controller. The variable temperature gas was maintained at 0 °C resulting in an approximate sample temperature of 15–20 °C. ¹³C chemical shifts were referenced externally with adamantane at 40.48 ppm for the methylene signal³⁸.

¹H detected FSR on U-CDN Ala

The π/2 pulse widths were 1.4 μs for ¹H, 3.6 μs for ¹⁵N, 1.4 μs for ¹³C. The contact time for ¹H-to-¹⁵N CP transfer and ¹⁵N-to-¹H CP transfer was 1.6 ms. The indirect dimension was not digitized in this experiment (since there is only one ¹⁵N frequency). The soft (Gaussian) π pulse was applied for 600 μs Gaussian pulse with Gaussian shape and a maximum nutation frequency of ~5.2 kHz power on the ¹³C channel. The REDOR π-pulse widths were 2.8 μs. The ¹³C channel transmitter frequency was set to three different values for exciting different ¹³C atoms in Ala, ¹³Cα (50 ppm), ¹³CO (175 ppm) and ¹³CH₃ (20 ppm). The dephasing time was varied from 0 to 1.2 ms. 64 scans were collected for each experiment.

¹H site-specific T₂ measurements on U-CDN GB1

The ¹⁵N-¹H 2D experiment with a ¹H refocusing π-pulse before detection was used to measure the ¹H T₂ for each residue. The echo time was varied in 18 steps from 60 μs to 21.6 ms. The contact time for ¹H-to-¹⁵N CP transfer was 900 μs and ¹⁵N-to-¹H CP transfer was 360 μs. The indirect ¹⁵N dimension was digitized to 23 ms and the ¹H acquisition time was set to 20 ms. ~10 kHz of decoupling power through TPPM decoupling was applied on ¹H during indirect evolution on ¹⁵N.

¹H detected FSR on U-CDN GB1

The contact time for ¹H-to-¹⁵N CP transfer was set to 900 μs and ¹⁵N-to-¹H CP transfer was set to 360 μs. The indirect ¹⁵N dimension was digitized to 23 ms. TPPM decoupling (~10 kHz, 50° phase, 50 μs pulse width) was applied on ¹H during indirect evolution on ¹⁵N. The S₀ experiment was performed with a refocusing hard π-pulse on ¹H and a soft Gaussian π pulse (600 μs duration) on the ¹³C channel. The S experiment was performed with additional REDOR π-pulse (4 μs pulse width) trains on the ¹³C channel. Four versions of the experiment were performed, each with the ¹³C channel transmitter frequency set to a particular region: ¹³Cα (50 ppm), ¹³CO (175 ppm) and ¹³CH₃ (20 ppm) and aromatics (130 ppm). The bandwidth of the soft pulse was ~15 ppm. ¹⁵N-¹H 2D S₀ and S spectra for each transmitter frequency were collected with REDOR times ranging from 240 μs to 1.2 ms.

Data Processing

The ¹H-detected FSR 1D spectra collected on Ala were processed in Mnova NMR software (MESTRELAB 2014), with polynomial baseline correction. 2D spectra were processed using NMRPipe³⁹ with back linear prediction and polynomial baseline correction applied to the direct dimension. Lorentzian-to-Gaussian apodization, phase-shifted sine bells and zero filling were applied to both dimensions.

Microsoft Excel for Mac 2011 (version 14.7.1) software was used to plot and fit the ¹H T₂ values assuming a single exponential decay. Error analysis was performed by fitting the natural log of the intensities with the Excel ‘LINEST’ function to extract the standard error associated with the slope of the linear fit (slope is equivalent to 1/T₂). The standard error was then converted to standard deviation and propagated to determine the standard deviation associated with T₂ of each residue.

Results and Discussion

REDOR effects on residual amino protons in U-CDN Ala

We initially tested the pulse sequence with U-CDN Ala, a small model compound with ~0.1% residual amino group protons from the crystallization. The low protonation level greatly reduces the average ¹H-¹H dipolar couplings, resulting in line widths of 0.5 ppm or less at 33.333 kHz MAS rate. Thus, we obtained site-specific ¹³C-dephased ¹H_N spectra (Fig. 2), allowing us to evaluate the specificity of the REDOR effect with selective pulses. The ¹³Cα is an average of ~2.2 Å from the amino ¹H (assuming three-site hopping in the fast limit), the shortest ¹³C-¹H distance available in this sample, and so with the ¹³Cα frequency selective pulse, the ¹H_N signal shows almost complete dephasing (ΔS/S₀ = 0.7) already at 360 μs (Fig. 2a). The ¹³CH₃ and the ¹³CO are each effectively ~2.8 Å from the amino (taking motionally averaging into account), and in these cases a slightly smaller dephasing (ΔS/S₀ = 0.5) is observed when the REDOR π pulses are on the ¹³CH₃ and ¹³CO region (Fig. 2b,c).

Fig. 2.

Application of ¹H detected FSR pulse sequence on uniformly ²H, ¹³C, ¹⁵N labeled Ala (residual protonation level is 1%). The ¹H_N 1D spectra are shown when the REDOR π pulse trains are on (a) ¹³Cα region, (b) 13CO region, and (c) ¹³CH₃ region. S₀ spectrum is in black (collected without REDOR π pulses on ¹³C channel) and S spectrum is in red (collected with REDOR π pulses on ¹³C channel)

¹H-¹⁵N correlation spectrum of GB1 at 30% protonation level

The ¹H line width in fully protonated and 100% back-exchanged GB1 protein at 40 kHz MAS was reported as ~1 ppm¹⁷ and at least 0.2 ppm²⁴ respectively, with partial overlap especially in central region of the spectrum (115-126 ppm ¹⁵N and 8.0-9.5 ppm ¹H), as well as in the Thr residues which were systematically broader due to the high level of sidechain protonation. Akbey et al.²² demonstrated that the optimal combination of ¹H resolution and sensitivity at ~30 kHz MAS is in the range of 30% to 40% H₂O back-exchange. We therefore utilized a 30% protonation level for the U-CDN GB1 sample, resulting in a ¹⁵N-¹H spectrum (Fig. 3) that is almost entirely resolved. The ¹H line widths are typically 0.1 ppm (60 to 80 Hz at 750 MHz ¹H frequency); roughly 20 Hz of this line width is contributed from the magnetic field inhomogeneity (based on the ¹³C line width in adamantane of ~5 Hz). Even the central region is mostly resolved, with only a few partially overlapped peaks (like K31/D36 and K4/A23/A34). The fractional back-exchange has an especially large benefit for Thr residues, and the individual amine protons of the Asn/Gln side chains are resolved.

Fig. 3.

¹⁵N-¹H ²D spectrum of uniformly ¹³C, ¹⁵N, ²H labeled GB1 protein, back-exchanged with 30% H₂O. The high level of deuteration of GB1 allows optimal resolution in ¹H dimension.

As expected, based on the overall improvements in the resolution of the ¹H-¹⁵N spectrum, the T₂ values were also greatly improved relative to 100% back-exchanged sample studied previously²⁴. We measured the bulk ¹H_N T₂ of GB1 to be around 13 ms (measured with a Hahn echo directly before acquisition. with ¹⁵N-filtering). We also measured site-specific T₂ values with a series of 2D ¹H-¹⁵N correlation spectra with incremented Hahn echo times prior to detection (Fig. S2). In this manner, we were able to determine site-specific T₂ values for 43 backbone amide sites, which range from 3 to 83 ms (Fig. 4). Even the fastest relaxing sites (e.g., T25, Y45) had more than sufficient echo lifetimes to enable application of the REDOR experiment directly on the ¹H transverse magnetization. In the following sections we demonstrate the magnitude of these REDOR effects and the utility for assignment and conformational analysis.

Fig. 4.

Site-specific T₂ measurements of backbone amide ¹H in uniformly ²H, ¹³C, ¹⁵N labeled GB1 protein, back exchanged to 30% amide ¹H. The transverse relaxation time varies from 3 ms (T25) to 83 ms (D40). The residues that are not observable or overlapped in a hNH 2D spectrum are not shown here. The error bars represent standard deviation calculated using linear regression model (described in text in details)

¹H{¹³C} REDOR effects in GB1 spectra

Dephasing by ¹³Cα emphasizes backbone correlations

Three versions of the ¹³C-dephased, ¹H-detected REDOR experiment are illustrated in Fig. 5, for different soft pulse bandwidths and mixing times, presented as difference spectra (S₀-S). First we examined the ¹³Cα-dephased version, which results in strong REDOR effects at only 240 (Fig. 5a) to 360 μs (Fig. 5b), as expected due to the short ¹H_N to ¹³Cα distance (~2.1-2.2 Å) and corresponding dipolar coupling of ~3 kHz. All the residues show significant REDOR effects, with roughly a third (18) exhibiting >50% dephasing (ΔS/S₀ > 0.5) at 240 μs, and most of the remaining (32 additional) sites reaching this threshold by 360 μs. These spectra are effective for providing a filtered spectrum including only the backbone, since the sidechain amine sites show significant dephasing only at 600 μs (Fig. 5c). For example, the N37-Hδ1 (ΔS/S₀ = 0.3), N37-Hδ2 (ΔS/S₀ = 0.5) and W43-Hε (ΔS/S₀ = 0.3) signals appear with significant intensity. In general, helical Asn/Gln residues have shorter distances from the sidechain amines to the backbone ¹³Cα atoms (3.2-4.2 Å distance between amine ¹H and intraresidue Cα or i±1 residue Cα) than the sidechain amine of those residues in β-strands (4-5.5 Å distance between amine ¹H and intraresidue ¹³Cα). In GB1, Q2 and N8 are on β-strands, whereas Q32, N35 and N37 are on the helix. In particular, the intraresidue distances to Cα are 3.2 Å for Q32-Hε1, 3.4 Å for N37-Hδ1, and 4.2 Å for N35-Hδ2. In addition to the short intraresidue distances, helical Asn/Gln residues can have relatively short interresidue Cα distances; both D36 (4.2 Å) and A34 (4.3 Å) Cα sites contribute to the N37δ1 dephasing, which is the strongest among all the sidechain signals here.

Fig. 5.

¹³C dephased ¹⁵N-¹H REDOR difference spectra (S₀-S) collected on uniformly ²H, ¹³C, ¹⁵N labeled GB1 protein, back exchanged with 30% ¹H_N. (a–c) Spectra with ¹³Cα dephasing at 240 μs, 360 μs and 600 μs respectively. (d–f) Spectra with 13C(methyl) dephasing at 240 μs, 600 μs and 960 μs. (g–i) 13C(aromatic) dephasing at 240 μs, 600 μs and 840 μs.

Extent of methyl-¹³C dephasing depends on secondary structure

The second version of the experiment was performed with methyl-specific dephasing. It is notable that even for small methyl-bearing residues (Ala, Thr and Val), the ¹H_N and ¹³C(methyl) are separated by at least three bonds and therefore the internuclear distance depends on the backbone ϕ angle (for all cases) and additionally on the χ1 angle for Thr and Val. This is notable and unlike the ¹⁵N-¹³Cβ distance in Ala, for example, which is ~2.5 Å and invariant to dihedral angles, or the ¹⁵N-¹³Cγ distances in Thr and Val which depend only on χ1. As a result of the ¹H_N-¹³C(methyl) conformation dependence, at 240 μs REDOR time (Fig. 5d) the methyl-bearing helical residues (A23/A34, A24, T25, A26 and V29) appear with greatest intensity (ΔS/S₀ > 0.2), along with some loop (V21 and T49) and residues neighboring methyls (F30). No β-strand residues show up at this short REDOR time; thus the effect in this case begins to provide a restraint on secondary structure. At increasing mixing time (Fig. 5e) the magnitude of the REDOR effect increases (ΔS/S₀ > 0.5) and additional methyl-bearing residues (I6, T16, T18, A20, V39, A48 and T55) show up prominently. In addition there are sites that begin to report on longer-range interactions, such as K13 and Q32. After 960 μs of dephasing (Fig. 5f), almost all the backbone amide ¹H start showing strong modulation in intensity (ΔS/S₀ > 0.5). Thus the most informative mixing times for the methyl-dephased version of the experiment are in the range 240 to 600 μs.

Aromatic dephasing reveals tertiary contacts

Performing the aromatic-selective version selects for ¹³C sites that are at least (for the ¹³Cγ) four bonds from the intraresidue ¹H_N and often (for the remainder of the ring) closer to the amides of neighboring or long-range residues. This observation is immediately apparent in Fig. 5g, since GB1 has six aromatic residues (Y3, F30, Y33, W43, Y45 and F52), yet roughly 15 peaks show similar intensities at short REDOR time, and none of the aromatic residues show strong attenuation in intensity of their own backbone amide ¹H. In fact, the two strongest REDOR effects are observed to W43-Hε (which has two aromatic ¹³C sites, ¹³Cδ1 and ¹³Cδ2, within ~2.2 Å bonds of the Hε) and, surprisingly, D46. (We will discuss this observation further in the following section). Within 600 μs or aromatic REDOR ¹³C dephasing (Fig. 5h), the backbone amide ¹H_N signals of the aromatic residues appear, and Y3, W43, Y45 and F52 exhibit ΔS/S₀ > 0.4. Few other non-aromatic amino acids like V21/E27, N37 (sidechain) also show substantial REDOR effects at this time point, and the magnitude continues to increase as expected at 840 μs (Fig. 5i).

Quantitation of REDOR effects and structural interpretation

In order to investigate the more nuanced site-specific details of the REDOR effects observed above, we examined the intensity changes (ΔS/S₀) of the ¹⁵N-¹H correlations by residue number at 600 μs for the methyl and aromatic versions of the experiment (Fig. 6). The plot of the methyl-dephased data (Fig. 6a) illustrates the extent of backbone and sidechain ¹H_N sites that have been dephased. To examine the detailed variations in REDOR effects and their relationships to the structural elements, we modeled the ¹H_N positions using WHATIF web server (Add Protons to the Structure)⁴⁰ and the high-resolution crystal structure (PDB 2QMT). Then we considered the distances from ¹H_N to methyl ¹³C sites. Almost all backbone ¹H_N sites are within 5 Å of some methyl groups, and so the average ΔS/S₀ value is ~0.4, but some sites show significantly larger effects (ΔS/S₀ > 0.6). For example, helical Ala residues exhibit nearly the shortest possible intraresidue H_N-C(methyl) distance, which is 2.5 Å because a ϕ angle of -60° results in an eclipsed H_N-Cβ (H_N-N-Cα-Cβ dihedral of ~0°). Therefore A23 and A34 exhibit a REDOR effect of >0.6 at 600 μs, and even larger effects are observed for A24 and A26, which have additional interresidue HN-C(methyl) distances of < 3.5 Å. Specifically, the A24-H_N to A23-Cβ distance is 3.3 Å and the A26-H_N to A20-Cβ distance is 3.2 Å (Fig. 7a).

Fig. 6.

Plot of S/S₀ intensity vs backbone amide (left) and sidechain (right) amino protons in uniformly labeled ²H, ¹³C, ¹⁵N labeled GB1 protein, back-exchanged with 30% amide protons. (a) shows the plot for ¹³CH₃ dephasing for 600 us, and (b) shows the plot for aromatic ¹³C dephasing for 600 us. The residues showing maximum dephasing have been labeled. Red asterisks indicate that the corresponding residues have overlapped peaks in the ¹⁵N-¹H spectrum.

Fig. 7.

(a–f): GB1 structure illustrating origins of REDOR effects for the methyl bearing and aromatic residues. (a–b) shows a few methyl bearing residues Ala (a) and Thr,Val (b), (c) shows Y45 and D46, (d) shows F52 and E27, (e) shows Y3, K31 and W43, (f) shows Y33 and N37. The black dotted lines indicate the distances between ¹HN and C(methyl) (a–b) and C(aromatics) (c–f). Figure f shows distances to sidechain amino 1H of N37. All the figures have been rendered in VMD.

For Thr and Val residues, the H_N-C(methyl) distances range from 2.6 to 3.0 Å for eclipsed or gauche conformations. For example, T49 (despite being present in a turn), has a H_N-Cγ2 methyl distance of 3 Å, as well as an interresidue T49H_N-A48Cβ distance of 3.3 Å, leading to leading to a strong REDOR effect of ~0.65 at 600 μs. V29, within the helix, has ΔS/S₀ > 0.6, arising from the H_N-Cγ1 distance of 2.9 Å (the H_N-Cγ2 distance is 4 Å). T25 is another helical residue that demonstrates a large ΔS/S₀ (>0.7), in this case due to the short interresidue distance of 3.0 Å from T25-H_N to A24-Cβ, which is significantly shorter than the T25H_N-Cγ2 intraresidue distance of 4 Å (Fig. 7b). Thus the strongest methyl-¹³C-dephased REDOR effects (ΔS/S₀ between 0.6 and 0.8 at ~600 μs) arise from Ala residues in helical conformations or from some Thr/Val residues with favorable conformations. A number of other methyl-bearing residues or neighbors of methyl-bearing residues show REDOR effects are between 0.4-0.6.

Dynamics plays an important role in determining the REDOR effects shown by methyl groups. The C3 axes of the methyl groups can undergo libration, thereby attenuating the ¹H_N-¹³C dipolar couplings. The methyl bearing residues whose side chains are extended in the solvent and not buried within the protein core will show such librational motions. One particular example in GB1 protein consists of V21 and V29. V21 is in a turn, with the ¹H_N-Cγ1 distance as 2.6 Å and ¹H_N-Cγ2 distance as 3 Å. Hence, V21 should show larger REDOR effect than V29. But as mentioned in Shi et al.³², the side chain of V21 is more mobilized than that of V29. This should explain the similar REDOR effects shown by V21 and V29.

Fig 6b plots the S/S₀ intensity versus the residue number for the backbone ¹H_N and sidechain amino group ¹H for aromatic residues at 600 μs. The strong ¹³C-¹³C dipolar interactions present in a fully ¹³C labeled protein results in an average ΔS/S₀ of 0.2. The residue showing the biggest REDOR effect (ΔS/S₀ ~0.6) at 600 μs is surprisingly not an aromatic residue, but a neighboring one, D46, which we attribute to the Y45 aromatic ring being trans relative to its intraresidue ¹H_N (distances >4.3 Å) but in much closer proximity to D46H_N; the distances from D46H_N to Y45Cδ2, Y45Cγ and Y45Cδ1 are 3 Å, 3.3 Å and 4.1 Å respectively (Fig. 7c). Hence, the REDOR effect shown by Y45 is ~0.45. F52 shows the strongest intraresidue REDOR effects among the aromatic residues (ΔS/S0 ~ 0.5), given a gauche conformation and F52H_N-Cδ1 distance of 3 Å, as well as other aromatic carbons within 4.5 Å of the amide. The F52 aromatic ring is also responsible for dephasing of the E27H_N which is ~4.3 Å from the F52Cε (Fig. 7d); the E27 REDOR effect of 0.4 is second strongest among the non-aromatic residues. Y3 and W43 are the two other aromatic residues that have intraresidue REDOR effects > 0.4. Both Y3 and W43 aromatic rings are in a gauche conformation with respect to the H_N (Fig 7e). Y3 ¹H_N- Cγ distance is 3.4 Å. W43 ¹H_N-Cδ1 distance is 3.5 Å. A surprising observation among the sidechain amines is N37δ1, which shows a REDOR effect of ~0.4, due to its orientation relative to the Y33 aromatic ring (Fig. 7f); the N37Hδ1 to Y33Cδ distance is 3.3 Å.

To summarize, for aromatic residues with gauche conformations, the H_N REDOR effects are between 0.4 and 0.6 at 600 μs. This is observed for Y3, W43 and F52 in GB1. Substantially weaker intraresidue REDOR effects are observed for trans conformations (Y33 and Y45), and in these cases the neighboring (i+1) residues can show larger effects, as we observed for A34 and D46. In some cases, long-range effects are detected involving the sidechain protons, such as in the case of N37.

We note that for the aromatic version of the experiment utilized here, the soft pulse was centered among the aromatic signals at ~130 ppm, with a bandwidth of ~15-16 ppm, thus exciting the region from ~122 to ~138 ppm. This region encompasses the majority of aromatic frequencies, but with some notable exceptions. For example, we might expect some attenuation of REDOR effects for Tyr Cε sites with upfield shifts (~120 ppm), and clearly the Tyr Cζ at ~160 ppm is outside of the range noted here. For Trp, the Cδ atoms were inverted by the soft pulses here, but Cγ (110 ppm) was outside of the bandwidth and other carbons would be near the edge of the soft pulse (Cε at 120 ppm or 140 ppm and Cζ at 120 ppm). Finally the Phe Cγ atom is near the edge at 140 ppm. It is apparent that the effects are sensitive to this range of offset, and given this large range of possible excitable frequencies, further selectivity could be achieved with soft pulses centered at ~160 ppm or ~110 ppm in order to detect backbone correlations in a residue-specific manner. For example, narrower bandwidth pulses could be used to detect the W43Cζ long-range distances, such as the K31H_N (Fig 7e). There are many additional opportunities for more selectivity and given the long T₂ values observed, detection of at least 10 Å distances may be feasible.

Conclusions

The ¹H-detected FSR pulse sequence introduced here is a broadly applicable technique for measuring long-range distances in SSNMR and leverages the high sensitivity of ¹H detection in combination with fast MAS. The REDOR difference spectra are collected as 2D ¹⁵N-¹H planes, so majority of the residues in the protein can be site-specifically resolved. Involving ¹H as one of the nuclei in the dipolar coupled spin pair also enhances the distance measurement range. No multiple pulse sequences are required for effective ¹H-¹H dipolar decoupling, given the fast MAS rates and deuteration. Another advantage of deuteration is enhancement of echo lifetimes of the detected nucleus (¹H), thereby increasing potential distance measurement range.

This method can be used with backbone ¹³C dephasing for purposes of spectral filtering, and the methyl- and aromatic-selective versions give insights into the secondary and tertiary structure of the protein. Specifically, the dephasing by the methyl bearing residues is stronger in general for helical residues, particularly Ala, and Thr or Val residues helical/loop regions can also yield strong REDOR effects (ΔS/S₀ > 0.6) at 600 μs dephasing time. The aromatic-specific version gives some initial insights into tertiary structure, since aromatic rings can be significantly closer to neighboring or long-range amides than intraresidue amides (e.g., Y33, Y45). This information can also restrain the side chain orientation of the aromatic ring with respect to its own backbone amide proton, based on the relative REDOR magnitude.

We envision that this approach can be extended to incorporate ³¹P or ¹⁹F dephasing in the case of phospholipids or fluorinated drugs bound to proteins. Notably, the detection range for ¹H{¹⁹F} REDOR effects with >10 ms ¹H T₂ values would be well in excess of 15 Å. We also anticipate that the large REDOR effects observed here could facilitate 3D TEDOR studies¹³ on sparsely ¹³C-isotopically labeled samples¹⁵. Sparse labeling in this case would be necessary in order to minimize signal loss due to recoupling of ¹³C-¹³C interactions from directly attached pairs. Using these approaches, ¹H-detected FSR has excellent potential for improving biomolecular SSNMR-based structure determination with enhanced sensitivity, resolution and distance measurement range.