Proton-Detected 2D Radio Frequency Driven Recoupling Solid-state NMR Studies on Micelle-associated Cytochrome-b₅

Abstract

Solid-state NMR spectroscopy is increasingly used in the high-resolution structural studies of membrane-associated proteins and peptides. Most such studies necessitate isotopically labeled (¹³C, ¹⁵N and ²H) proteins/peptides, which is a limiting factor for some of the exciting membrane-bound proteins and aggregating peptides. In this study, we report the use of a proton-based slow magic angle spinning (MAS) solid-state NMR experiment that exploits the unaveraged ¹H-¹H dipolar couplings from a membrane-bound protein. We have shown that the difference in the buildup rates of cross peak intensities against the mixing time - obtained from 2D ¹H-¹H radio frequency-driven recoupling (RFDR) and nuclear Overhauser effect spectroscopy (NOESY) experiments on a 16.7-kDa micelle-associated full-length rabbit cytochrome-b₅ (cytb₅) - can provide insights into protein dynamics and could be useful to measure ¹H-¹H dipolar couplings. The experimental buildup curves compare well with theoretical simulations and are used to extract relaxation parameters. Our results show that due to fast exchange of amide protons with water in the soluble heme-containing domain of cyb₅, coherent ¹H-¹H dipolar interactions are averaged out for these protons while alpha and side chain protons show residual dipolar couplings that can be obtained from ¹H-¹H RFDR experiments. The appearance of resonances with distinct chemical shift values in ¹H-¹H RFDR spectra enabled the identification of residues (mostly from the transmembrane region) of cytb₅ that interact with micelles.

Introduction

Atomic-level characterization of structure and dynamics using solid-state nuclear magnetic resonance (NMR) spectroscopy can provide piercing insights into macromolecular systems, even though it is a challenging task due to low spectral resolution and sensitivity. Significant advances in the development of pulse sequences, ultrahigh field magnets, and a very fast (in the range of 30–40 kHz) to ultra-fast (≥ 60 kHz) magic angle spinning (MAS) techniques with a low radio-frequency decoupling power requirements have opened up new avenues to investigate a variety of biomolecules such as membrane proteins.^1–8 Particularly, the recoupling techniques^9,10 have facilitated an efficient measurement of internuclear distances from both small (uniformly labeled) and large (selectively labeled) biomolecules. Most of these recoupling-based structural studies utilize the homonuclear dipolar couplings among low-abundant nuclei (¹³C, ¹⁵N) due to the larger spread of their chemical shift frequency. On the other hand, homogeneously broadened ¹H resonances resulting from strong ¹H-¹H dipolar interactions have not been suitable to fully utilize the high sensitivity and abundance of protons. However, the availability of ultrafast MAS probes and the use of heavily deuterated samples have made the complete suppression of homonuclear ¹H-¹H dipolar couplings feasible; this has paved the way for the development of proton-detection based multi-dimensional experiments in the solid state.^5,11–22 Recent studies have also demonstrated the applicability of MAS experiments on soft solids like micelles and fluid lamellar phase bilayers which, unlike rigid solids, exhibit weak ¹H-¹H dipolar couplings due to their high molecular mobility.²³ In this regard, one particular approach is to employ Radio Frequency-Driven Recoupling (RFDR)^24,25 and Nuclear Overhauser Effect SpectroscopY (NOESY)²⁶ experiments under MAS to obtain 2D ¹H-¹H isotropic chemical shift correlation spectra via coherent ¹H-¹H dipolar coupling and NOE, respectively. This approach has been previously applied to investigate the motional characteristics of resin and membrane-bound peptides.^27,28 A recent study on a membrane-bound antimicrobial peptide (MSI-78, also commercially known as pexiganan) also highlights the importance of this approach for the measurement of ¹H-¹H residual dipolar couplings.²⁹ In this study, we extend this objective on a relatively large micelle-associated unlabeled cytochrome-b₅ (cytb₅) to gain better insights into its structure and dynamics.

Cytb₅ is a 16.7-kDa electron transfer protein consisting of 134 amino acid residues. It is composed of three distinct regions - namely a structured water-soluble paramagnetic ferric low-spin heme-containing domain, a transmembrane helical domain, and an unstructured ~14-residue long linker region.³⁰ The NMR structure along with the amino acid sequence of the full-length membrane-bound rabbit cytb₅ (PDB # 2M33) is shown in Figure 1. The structure contains five α-helices (α1, L14-H20; α2, K39-E43; α3, E49-Q54; α4, T60-V66 and α5, T70-F79), five β-strands (β1, K10-Y12; β2, W27-L30; β3, K33-D36 and β4, G56-D58) and one 3₁₀ helix (P86-R89). The flexible linker domain (S90-D104) of cytb₅ lacks a defined secondary structure and connects the helical transmembrane and heme domains. Cytb₅, in combination with its redox partners such as cytochrome-P450 (cytP450) and cytP450 reductase (CPR), is essential for enzyme kinetics to metabolize 70% of the drugs presently available in the market.³¹ In addition, cytb₅ is vital to carryout important biochemical reactions such as the synthesis of testosterone and lipids which are important to preserve the cellular integrity.³¹ Previous solid-state NMR studies determined the topology of the transmembrane domains of cytb₅³² and cytP450³³, reported the drastic difference between the timescales of motions for the soluble and transmembrane domains,^34,35 and also characterized the dynamics of side chains.³² Further, solid-state NMR studies revealed the interaction between the transmembrane domains of cytb₅ and cytP450 by using magnetically-aligned bicelles.³⁶ Recently, we reported the first high resolution structure of a highly dynamic membrane-bound full-length cytb₅-cytP4502B4 electron transfer complex (~72 kDa) employing experimentally derived NMR constraints from solution and solid-state NMR techniques, and mutagenesis data.^30,32 Additionally, amino acid residues in the interacting interface of the cytb₅-cytP4502B4 complex and an electron transfer pathway were also identified.³⁰ These findings were well corroborated with our results based on backbone amide-¹⁵N chemical shift anisotropy (CSA) tensors derived from free and P450-bound cytb₅ using solution NMR and quantum chemical studies.^37–39

Figure 1.

Amino acid sequence and NMR structure (PDB # 2M33) of the full-length membrane-bound rabbit cytochrome-b₅.^30,40

In the present study, we restrict our focus on 2D ¹H-¹H correlation spectra obtained from the RFDR and NOESY experiments and analyze the intensity difference for the cross-peaks observed from these two MAS experiments. Experimental details and discussion of the NMR structure of micelle-associated cytb₅ (PDB # 2M33) are described elsewhere^30,40 and will not be discussed further. We believe that the RFDR experiment could be more sensitive compared to the routinely employed solution NMR NOESY experiment as the cross-peak intensities are higher due to additional dipolar recoupling from nearby spins. More importantly, this method can be potentially used to determine the membrane interacting residues, which otherwise may not be easily possible by solution NMR experiments due to their fast spin-spin relaxation.

Experimental

A micelle-associated form of full-length rabbit cytochrome-b₅ was expressed in E.Coli and purified as previously reported.³⁰ 2.5 mM unlabeled cytb₅ was incorporated into 45 mM perdeuterated dodecylphosphocholine (DPC) detergent micelles in 100 mM potassium phosphate (pH 7.4) containing 5% D₂O. 40 µL of this sample was packed into a 4 mm zirconium rotor. All NMR measurements were performed on a Varian VNMRS 600 MHz solid-state NMR spectrometer using a 4 mm double-resonance nanoprobe at a spinning speed of 2.7 kHz at 25 °C. Deuterium field-lock was used throughout the experiment and water suppression was achieved using a 1.5 s saturation pulse with a 5 Hz radio frequency power at the beginning of each scan. The proton carrier frequency was set at water proton peak for all the experiments. For assignment of resonances, a 2D ¹H-¹H TOCSY (TOtal Correlation Spectroscopy)⁴¹ experiment was performed under MAS at a spinning speed of 2.7 kHz. The isotropic mixing in the TOCSY pulse sequence was attained using the DIPSI (Decoupling In the Presence of Scalar Interactions)⁴² spin-lock scheme applied for a duration of 60.4 ms. A proton radio frequency field strength of 65.8 kHz was used for pulses in TOCSY, NOESY and RFDR experiments. 2D ¹H-¹H NOESY and RFDR spectra were recorded with an acquisition time of 1.6 s, 48 scans, 400 t₁ increments, 27028 t₂ complex points while the 2D ¹H-¹H TOCSY spectrum was recorded with an acquisition time of 200 ms, 124 scans, 512 t₁ increments, 3380 t₂ complex point and a spectral width of 14.1 ppm in both frequency dimensions. All spectra were processed using NMRpipe,⁴³ by applying phase-shifted sine bell multiplication prior to Fourier transformation with zero filling in both t₁ and t₂ dimensions. Resonance peak assignment and intensity measurements were carried out using Sparky.⁴⁴

Results and discussion

One-dimensional ¹H NMR spectra of micelle-associated cytb₅

Proteins embedded in a membrane mimetic environment undergo slow tumbling on the NMR timescale and as a result the static NMR spectrum can suffer from significant line broadening caused by the anisotropic interactions, such as the chemical shift anisotropy and the dipolar couplings. In order to enhance the resolution and sensitivity for such macromolecular systems, performing experiments under high magnetic field strengths has always been an obvious choice. However, comparable resolution and sensitivity could also be obtained by performing experiments at a low magnetic field strength under MAS, which narrows the spectral lines by averaging out the orientation dependent anisotropic interactions. To validate this fact, we obtained 1D ¹H NMR spectra of full-length cytb₅ incorporated in DPC detergent micelles under static and MAS conditions (Figure 2). It is obvious from Figure 2 that the MAS spectrum of cytb₅ shows additional resonances in the 2.1 to 4.6 ppm region in comparison to the static spectrum. On the basis of these results from 1D experiments, we carried out 2D ¹H-¹H RFDR and NOESY experiments under MAS to get additional insights into high-resolution structural characterization of micelle-associated cytb₅.

Figure 2.

One-dimensional ¹H NMR spectra of micelle-associated cytb₅ under (A) static condition at 900 MHz and (C) 2.7 kHz magic angle spinning at 600 MHz. (B) Peaks from the amide and aromatic proton chemical shift regions are shown with a magnified intensity scale. Circled regions show additional resonances observed under MAS, while the asterisk represents the spinning sidebands. DSS (4,4-dimethyl-4-silapentane-1-sulfonic acid) (at 0 ppm) was used as a reference for ¹H chemical shift.

RFDR and NOESY experiments under MAS

2D RFDR and NOESY pulse sequences are depicted in Figure S1 of the supporting information. These two pulse sequences differ only in the mixing period: the RFDR pulse sequence is composed of a series of rotor-synchronised 180° pulses in the mixing period, whereas the NOESY pulse sequence is devoid of such pulses. The XY16 phase cycling scheme was used for the 180° pulses in the finite-pulse-RFDR to suppress the effects of pulse imperfections like RF field inhomogeneity and offset.⁴⁵ The rotor-synchronized 180° pulses in the RFDR experiment reintroduces the homonuclear dipolar couplings under MAS. Consequently, the cross-peak intensities in the 2D ¹H-¹H correlation RFDR spectrum arise from the combined effect of NOE cross-correlation and recoupled ¹H-¹H dipolar interactions. Hence, the difference in peak intensities observed in the RFDR and NOESY spectra is due to residual (or unaveraged and recoupled by MAS) dipolar couplings that can be used to determine interatomic distances.

The time evolution of the longitudinal proton magnetization is represented by the rate equation, $\raisebox{1ex}{$dM$}\!\left/ \!\raisebox{-1ex}{$dt$}\right.=[R+L]M$ ; where R and L are the kinetic matrices.^46–48 The diagonal elements (R_ii) of the R matrix represent the auto-relaxation rate while the off-diagonal elements $(R_{\text{ij}}=-a_{\mathrm{H}}/r_{\text{ij}}^6(6J_2-J_0)$ represent the cross-relaxation rate between spins i and j; where a_H=5.6965×10¹⁰ Å⁶s⁻¹, r_ij is the internuclear distance and J_n represents the spectral density function. In the above equation, L_ij is the magnetization transfer between spins calculated using the expression $L_{\text{ij}}=D_{\text{ij}}/r_{\text{ij}}^2$ where D and r are the diffusion constant and distance between two spins, respectively. The application of rotor-synchronized 180° pulses in the RFDR experiment allows for the magnetization transfer via the recoupled homonuclear dipolar coupling (~1/r³)^49,50 in addition to the NOE (~1/r⁶), and thus resulting in a faster buildup rate along with the increased intensity for the cross peaks measured as a function of the mixing time in comparison to the NOESY experiment. The RFDR and NOESY buildup rates were calculated using a diagonalization scheme by implementing the above rate equation in an in-house Matlab script. The buildup curves for the RFDR and NOESY experiments were fitted using three parameters; relaxation time T₁*, buildup rate R and the peak intensity I₀. T₁* represents the combined contributions from the spin-lattice relaxation, T₁, of protons and the imperfection of 180° pulses in RFDR. The buildup curves plotted against the mixing time have both exponential growth and decay fit parameters.

Resonances in 2D ¹H-¹H RFDR and NOESY spectra for micelle-associated cytb₅ recorded under MAS were assigned using a combination of 2D ¹H-¹H NOESY and 3D HSQC-NOESY spectra with a 100 ms mixing time obtained on a 900 MHz solution NMR spectrometer, and a 2D ¹H-¹H TOCSY spectrum at 60.4 ms mixing time recorded on a 600 MHz solid-state NMR spectrometer under 2.7 kHz MAS (Figure 3). Isotropic chemical shift values for resonances obtained from 3D NOESY are reported elsewhere.⁴⁰

Figure 3.

Representative 2D ¹H-¹H TOCSY spectral regions of cytb₅ incorporated in DPC micelles recorded using a mixing time of 60.4 ms under 2.7 kHz MAS showing chemical shift correlations for (A) amide ¹H-alpha and side chain protons, (B) aromatic protons, (C) ¹H_β-side chain protons and (D) ¹H_α-side chain protons.

Residual ¹H-¹H dipolar couplings observed in the H_α and side chain regions of cytb₅

Superimposed 2D ¹H-¹H RFDR and NOESY spectra in Figure 4, showing the chemical shift correlations for ¹H_β-side chain protons (Figure 4A) and ¹H_α-side chain protons (Figure 4B), were obtained from an unlabeled cytb₅ incorporated in perdeuterated DPC micelles. A comparison is made at two different mixing times (50 ms for RFDR and 200 ms for NOESY) because the maximum intensity for most of the cross-peaks could only be obtained at these values. Substantial differences in the cross-peak intensities are observed in 2D ¹H-¹H RFDR and NOESY spectra. The variation in the cross-peak intensity could arise from many factors: such as the spin-spin coupling (both through space (or dipolar) and scalar through-bond (or J) couplings) among neighboring protons, electrostatic interactions and chemical shift exchange effects. In Figure 4, it can be seen that the cross-peaks in the RFDR spectrum corresponding to K7H_β1−H_β2, K10H_β−H_ε and K10H_γ−H_ε correlations have much higher intensities as compared to the corresponding peaks in the NOESY spectrum. This observation indicates that side chains of residues K7 and K10 are associated with strong electrostatic interactions possibly due to the formation of a salt bridge between lysine and aspartate residues (D6/D8) located in the N-terminal region of cytb₅ that could prevent the complete averaging of ¹H-¹H dipolar couplings. Surprisingly, few cross-peaks arising from the side chains of the linker region residues (M96 and E97) of cytb₅ show higher peak intensities and hence these residues are associated with ¹H-¹H dipolar couplings. This could be attributed to the restricted isotropic motion of these residues due to their interaction with surrounding DPC molecules. This finding is vital as our earlier solution NMR investigation suggested the linker region of cytb₅ to be highly flexible due to solvent exposure and lacks any definite secondary structure based on the absence of both inter- and intra-residue NOEs for most residues in the linker region.⁴⁰

Figure 4.

Representative regions of 2D ¹H-¹H RFDR (black) and NOESY (red) spectra of cytb₅ incorporated in perdeuterated DPC micelles showing the chemical shift correlations for ¹H_β-side chain protons (panel A) and ¹H_α-side chain protons (panel B). RFDR and NOESY spectra were recorded at a mixing time of 50 and 200 ms, respectively, under 2.7 kHz MAS condition. Many additional cross-peaks are observed in the RFDR spectrum, which mostly belong to the side chains of the residues located in the transmembrane region of cytb₅.

In order to provide a detailed discussion on the observed cross-peak intensity differences and to quantify residual dipolar couplings to extract ¹H-¹H distances, we have plotted experimentally measured cross-peak intensities against the mixing time (the so called buildup curves) for various residues of cytb₅ (Figure 5). The experimental intensities were simulated using the rate equation discussed above and the best fitting parameters are reported in Table 1. The difference in cross-peak intensities measured from the RFDR and NOESY spectra can be used to extract ¹H-¹H dipolar coupling constraints for the structural studies of the protein. It can be seen from Figure 5 that the simulated buildup curves for different residues of cytb₅ are in close agreement with the experimental data. Figure 5 also shows that the cross-peak intensity for H_α and side chain regions reaches a maximum within a short mixing time of ~50–100 ms in the RFDR experiment, which is in contrast to the NOESY experiment where the maximum intensity is obtained at longer mixing times. As discussed above, the transfer of magnetization is mediated through coherent ¹H-¹H dipolar interactions recovered from the RFDR experiment in addition to the magnetization transfer via an incoherent cross-relaxation from NOE. The T₁* of the RFDR cross-peaks are usually much shorter in comparison to that of the NOESY cross-peaks due to the applied rotor-synchronized 180° pulses in the RFDR sequence. The imperfection of 180° pulses in the RFDR causes attenuation of the cross-peak intensity for longer mixing times, due to which a significant difference in the buildup curves obtained from the RFDR and NOESY experiments is observed.

Figure 5.

The simulated buildup curves for various cross-peaks observed in H_α and side chain regions for selected residues of cytb₅ obtained from the RFDR (Black) and NOESY (Red) experiments on cytb₅ incorporated in perdeuterated DPC micelles. Though the intensity scale was arbitrarily chosen, the intensities of data points in all panels are directly comparable. RFDR and NOESY spectra were collected for six different mixing times, 25, 50 100, 150, 200 and 300 ms.

Table 1.

Relaxation parameters obtained from the simulation of experimentally measured crosspeak intensities from cytb₅ incorporated in DPC micelles.

Residues	RFDR			NOESY
	1/T1* (s−1)	R (s−1)	I0	1/T1* (s−1)	R (s−1)	I0
M96*Hα-Hβ1	4.7	5	18	2.5	4	7
M96Hα-Hβ1	3	10	25	2.5	2	5
K39Hε-Hδ	3.7	35	48	2	10	8
K39Hε-Hβ	4.3	30	38	1	10	7
R52Hδ−Q54Hγ	8	20	10	0.8	8	6
E53Hα-Hγ2	6	3	12	0.2	3	5
D71Hα-Hβ2	8	10	14	1.6	2	10
D134Hα-Hβ2	6	18	18	1	3	5
W27	13.5	27	8.8	4	6	5.5
Y12	15	22	21	7	7	11
F79Hε-Hδ	14	25	42	4	19	21
F40	5.5	12	25	5	3	50
Y11	14.8	16	30	4	4.5	8
D8Hα-H	6	8	3	2.5	8	5
K77Hα-I80H	8	10	8	0.8	12	18
V9Hγ1-H	2	14	1.2	2.5	3.5	6.2
D134Hα-H	8.9	4.7	17	3	12	9
D133Hα-D134H	3.3	5.8	2.2	3.2	4.0	3.9
Q54Hα-H	4.8	15	3.5	2.7	7	3
D6Hα-Hβ1	8	14	10
D6Hα-Hβ2	8	14	12
W109/110/113Hα-Hβ (ω1/ω2:4.57/3.01ppm)	7	6	7
W109/110/113Hα-Hβ (ω1/ω2:4.57/2.87ppm)	7	5	7
N106Hα-Hβ1	7	14	13
N106Hα-Hβ2	6	10	15

It is apparent from the buildup curves in Figure 5 that the side chains of a residue, M96 and its conformer (M96H_α-H_β2 and M96*H_α-H_β), belonging to the linker region are associated with unaveraged dipolar couplings, which could be because of their interaction with DPC molecules as discussed earlier. A stronger dipolar coupling and/or higher buildup rate is also observed for K39 (K39H_ε-H_δ and K39H_ε-H_β) that is located in the helical region of cytb₅ and, as a consequence, it may be rigid due to H-bonding interactions and therefore could avoid the averaging of the dipolar couplings associated with this residue. Additionally, it can undergo restricted motion due to electrostatic interactions. Also, the side chains of residues such as R52, E53 and D71 (R52H_δ−Q54H_γ2, E53H_α-H_γ2 and D71H_α-H_β2) in a alpha-helix are shown to have weak dipolar interactions possibly due to weak electrostatic interactions. Interestingly, a stronger dipolar coupling is observed for the side chain of a C-terminal residue (D134) in the presence of micelles. This could be due to its interaction with micelles that results in a hindered motion of its side chain.

RFDR reveals the membrane interaction of residues in the transmembrane region of cytb₅

The alpha-helical structure of the transmembrane region of cytb₅ incorporated in magnetically-aligned DMPC/DHPC bicelles and its orientation in the membrane were previously determined using a 2D HIMSELF (Heteronuclear Isotropic Mixing leading to Spin Exchange via the Local Field) or HERSELF (HEteronuclear Rotating frame Spin Exchange via the Local Field)⁵¹ static solid-state NMR experiment.^33,52 In our earlier report on the micelle-associated structure of rabbit cytb₅ (PDB # 2M33), the tryptophan residues (W109/110/113) were predicted to be at the edge of transmembrane domain based on the observed broad indole side chain NH resonances along with additional peaks due to chemical exchange in the ¹H-¹⁵N TROSY-HSQC (Transverse Relaxation-Optimized SpectroscopY - Heteronuclear Single Quantum Coherence) spectrum.^30,40 However, no cross-peaks were observed for these residues in 2D ¹H-¹H NOESY and 3D HSQC-NOESY spectra recorded at 900 MHz. Analogous to this, we did not observe these resonances in the 2D NOESY spectrum obtained under MAS at 600 MHz (Figure 4). In contrast, some well-resolved resonances from the side chains of the transmembrane residues (Figure 6) were observed in the RFDR experiment. Resonance assignment based on the isotropic chemical shifts of AMX spin system indicated that these cross-peaks correspond to the side chains of D104, S105, N106, S107, S108, W109, W110, N112 and W113 located at the end of the linker region and/or the beginning of the transmembrane region of cytb₅. These residues may interact with DPC molecules that could motionally-restrict the side chains to result in the observed dipolar couplings. Furthermore, this study also confirms our earlier prediction that tryptophan rings (W109/110/113) are in fact at the edge of the transmembrane domain of cytb₅ in DPC micelles. This finding is cross validated further by comparing with the RFDR measurement on cytb₅ in the absence of DPC micelles (Figure 6). Since the presence of DPC micelles restricts the motion of the transmembrane region, it should be highly mobile in the absence of the DPC molecules as indicated by the almost negligible dipolar recoupling observed for most of the resonances (Figure 6). Hence, in the present study, the RFDR experiment provides a direct evidence for the interaction of transmembrane residues of cytb₅ with DPC micelles. It can be seen from the buildup curves (Figure 6B) that the observed dipolar couplings for these residues are weaker in comparison to that of the H_α and side chain regions. The dynamic parameters for these residues are extracted from the theoretical simulation and are given in Table 1. It is also interesting to note that the RFDR cross-peaks from the side chains of N-terminal (D6) and C-terminal (D134) residues appear both in the presence and absence of DPC micelles; however, their intensities are weaker in the absence of DPC micelles. This observation clearly suggests that the C-terminal residue is comparatively more rigid in the presence of micelles as discussed above. Similarly, the isotropic motion of the side chain of D6 is restricted probably due to electrostatic interactions with nearby charged residues as pointed out earlier.

Figure 6.

(A) Superimposed 2D ¹H-¹H RFDR spectra of cytb₅ in the presence (black) and absence (blue) of DPC micelles recorded at a mixing time of 50 ms. (B) Simulated buildup curves for the RFDR cross-peak intensities measured from the side chain resonances of residues in the transmembrane region of cytb₅ in the presence of DPC micelles. All other experimental details are as given in Figure 5 caption.

To better understand the sources of unaveraged ¹H-¹H dipolar couplings present in the micelle-associated cytb₅, it is important to discuss the interactions between the protein and DPC micelles used in this study. It is known that there are approximately 50 DPC molecules per micelle (molecular weight ~19.5 kDa),⁵³ which gives a total molecular weight of ~36.2 kDa for the micelle-associated cytb₅ system. On the basis of molecular size, the micelle-associated protein sample that does not form any aggregates should be suitable for structural studies using solution NMR spectroscopy. However, depending on the folding of the protein and its interaction with micelles, the difference in the time scale of motion for different domains of the protein could result in a difference in their transverse relaxation rates. This is indeed the case as demonstrated by our previous study on a well-folded micelle-associated cytb₅.³⁰ As an example, the transmembrane domain of micelle-associated cytb₅ undergoes slow motion (ms or slower) in comparison to the soluble domain that undergoes fast motion (µs or faster). While the high-resolution structure of the soluble domain of cytb₅ incorporated in DPC micelles was solved by traditional solution NMR experiments, the residues in the transmembrane domain of the protein were not observable due to a significant line broadening caused by their restricted motion in the hydrophobic region of micelles.³⁰ On the other hand, as demonstrated in this study, MAS overcomes the line broadening effects and thus enables the observation of resonances from the transmembrane residues. While the motionally-unaveraged ¹H-¹H dipolar couplings, that are not completely suppressed due to the restricted motion of cytb₅ in DPC micelles under static conditions, are suppressed by MAS, and are recoupled by the 2D RFDR experiment as shown in this study. It is worth mentioning here that that ¹H-¹H RFDR cross-peaks for some of the residues were not detected from the transmembrane region of micelle-associated cytb₅, which could be due to the stronger ¹H-¹H dipolar couplings. RFDR experiments performed at shorter mixing times and/or higher spinning speeds might be suitable to address this issue.

To further confirm the above observations, we measured the buildup rates for the RFDR cross-peaks detected in the absence of DPC micelles (Figure 7); superimposed 2D RFDR spectra displaying alpha-side-chain ¹H-¹H correlations obtained at different mixing times are shown in Figure S2 of the Supporting Information. It can be seen from Figure 7 that the maximum intensity for these cross-peaks could only be obtained at longer mixing times resulting in slower buildup rates in comparison to that observed from the micelle-associated cytb₅ (refer to the relaxation parameters in Table S1 of the Supporting Information). This observation suggests that cytb₅ in the absence of DPC micelles is more flexible than in association with DPC micelles. Importantly, this observation (in Figure 7) is also similar to what is observed from NOESY buildup curves for micelle-associated cytb₅ wherein the ¹H-¹H cross-peaks arise only due to incoherent NOE; which is in contrast to the observations in RFDR where the magnetization exchange occurs via both incoherent NOE effect and coherent (recoupled) ¹H-¹H dipolar interactions. In other words, these results suggest that the ¹H-¹H dipolar couplings are suppressed by the isotropic tumbling of cytb₅ in the absence of DPC micelles and therefore the magnetization exchange is mostly due to incoherent NOE effect. Notably, this also accounts for the reason why a recoupling RFDR sequence should work on micelle-associated molecules.

Figure 7.

The simulated buildup curves for the RFDR ¹H-¹H cross-peak intensities observed in H_α and side chain regions for selected residues of cytb₅ in the absence of DPC micelles.

Dipolar recoupling for aromatic protons

Since an aromatic side chain could undergo a restricted motion depending on its location in the protein sequence and secondary structure, we investigated the ability of RFDR experiments to recouple dipolar couplings associated with aromatic protons. Superimposed 2D ¹H-¹H RFDR and NOESY spectra for aromatic side chain resonances are shown in Figure 8. Results from Figure 8 indicate that the residues containing aromatic groups such as Y11, Y12, W27, Y35 and F79 exhibit higher cross-peak intensity in the RFDR spectrum than that in the NOESY spectrum. In contrast, residue F40 seems to have negligible intensity difference from the two experiments.

Figure 8.

(A) Superimposed 2D ¹H-¹H RFDR (black) and NOESY (red) spectra of cytb₅ incorporated in DPC micelles recorded at a mixing time of 50 ms and 200 ms, respectively, showing the chemical shift correlation of aromatic protons. (B) Simulated buildup curves and experimentally measured cross-peaks intensities for aromatic side chains obtained from RFDR (black) and NOESY (red) experiments. All other experimental details are as given in Figure 5 caption.

In addition, observed dipolar couplings are weaker for most of the aromatic protons in comparison to H_α and side chain regions (Figure 8B). It can be seen from the buildup rates that aromatic protons of F40 are associated with a smaller dipolar coupling in comparison to Y11, Y12, W27 and F79 residues. On the basis of our recently reported backbone amide-¹⁵N chemical shift anisotropy (CSA) measurements, residue F40 is associated with a smaller CSA span than Y11, Y12, Y35, W27 and F79.³⁹ In other words, F40 was found to be more flexible than Y11, Y12, Y35, W27 and F79. This finding corroborates well with the present result wherein ¹H-¹H dipolar coupling for aromatic protons of F40 is averaged out due to flexibility while other residues exhibit unaveraged dipolar couplings. The buildup curve for aromatic protons of Y35 is not shown in Figure 8B as its RFDR cross peak intensity reached the maximum at a mixing time of 50 ms and decayed entirely with a further increase in the mixing time due to fast relaxation. Superimposed 2D ¹H-¹H RFDR and NOESY spectra showing aromatic ¹H-¹H correlations obtained using different mixing times (25–300 ms) are shown in Figure S3 of the supporting information.

Dipolar couplings associated with amide-protons cannot be recoupled

An overlay of 2D amide ¹H-¹H RFDR and NOESY spectra for cytb₅ incorporated in DPC micelles is shown in Figure 9. It is interesting that the 2D NOESY spectrum reveals more cross-peaks with higher intensities in comparison to the RFDR spectrum.

Figure 9.

(A) 2D ¹H-¹H RFDR (black) and NOESY (red) spectra for the amide-side chain proton region of cytb₅ incorporated in DPC micelles recorded using a mixing time of 50 ms (RFDR) and 200 ms (NOESY) under 2.7 kHz MAS. Spinning sidebands are denoted by SSB. (B) Simulation of the variation of cross-peak intensities measured for amide-protons resonances from the RFDR (black) and NOESY (red) spectra. All other details are as given in Figure 5 caption.

This observation is attributed to a constant exchange of amide protons with water that average out coherent dipolar interactions associated with amide protons. But, if this is the only explanation for this observation, then RFDR intensities should be very similar to that of NOESY as the magnetization transfer related to amide protons in the RFDR experiment is only due to incoherent cross-relaxation effect. As can be seen from Figure 9, this is obviously not the case.

The amide-proton cross-peak intensities measured from the RFDR spectra are, in fact, lower than that obtained from the NOESY spectra. It is probably due to the loss of amide-proton magnetization to surrounding spins via solvent exchange. This finding is not surprising since the major contribution to the RFDR cross-peak intensity comes from the spin diffusion driven process originating from the coherent dipolar interaction among protons. Therefore, in cases where solvent exchange of protons is highly plausible, 2D NOESY measurement could be a better option than RFDR to obtain ¹H-¹H correlations; however, there could be few exceptions to this. For example, the RFDR intensity for a helical residue, Q54, is slightly higher than the NOESY intensity despite that its amide proton is most likely to be exchanging with water (Figure 9B); but the overall intensity is considerably lower than that observed for protons in the side chain region.

Conclusion

There is considerable interest in the development of proton-based solid-state NMR experiments to study the structure and dynamics of membrane-bound proteins. Though solution NMR experiments are routinely used to solve high-resolution structures of relatively small-size membrane proteins embedded in isotropic phases like detergent micelles or isotropic bicelles, the use of MAS experiments on these samples to recouple motionally-unaveraged ¹H-¹H dipolar couplings could provide additional valuable constraints (long-range distance) to improve the accuracy of structures solved by NMR spectroscopy. In addition, such residual dipolar couplings measured by recoupling MAS solid-state NMR techniques could be used to characterize the slow motions and the aggregation process that may not be easily determined from solution NMR experiments. To demonstrate the feasibility of recoupling dipolar couplings and using proton-based solid-state MAS experiments, we successfully carried out 2D ¹H-¹H RFDR and NOESY experiments under MAS on an unlabeled cytb₅ incorporated in perdeuterated DPC micelles.

The variations of experimentally measured cross-peak intensities with the mixing time compare very well with our theoretical simulations and enabled the extraction of NMR parameters to understand the dynamics of micelle-associated cytb₅. The longitudinal relaxation time (T₁*) for the RFDR cross-peaks are found to be much shorter than that for the NOESY cross-peaks, thus accounting for an accelerated magnetization transfer via the recoupled ¹H-¹H dipolar couplings. Our results demonstrate that the residual ¹H-¹H dipolar couplings associated with H_α and side chain protons could be easily recoupled by RFDR and their measurement could be valuable in structural studies of micelle-associated proteins. Results from the RFDR measurement provides a direct evidence for the interaction between the side chains of the transmembrane residues with DPC detergent molecules, which is observed through the appearance of additional resonances in the RFDR spectra that is otherwise not possible due to their fast spin-spin relaxation. Our results also show that the coherent dipolar interactions associated with amide-protons are suppressed by the fast exchange between amide and water protons. While the results presented here can provide qualitative information about the structural constraints in the form of ¹H-¹H dipolar couplings or distances, further studies are needed to obtain quantitative estimation of the constraints. One of the major challenges to overcome is to accurately model the multiple ¹H spin network to extract structural constraints. Although our measurements were on a micelle-associated cytb₅ from a 600 MHz spectrometer under slow MAS, we strongly believe that multidimensional MAS solid-state NMR experiments utilizing RFDR performed at higher magnetic fields and higher spinning speeds can be of greater applicability over commonly employed methods for 3D structure determination of unlabeled membrane proteins and amyloid fibrils.

Highlights.

• 2D ¹H/¹H RFDR MAS spectra of an unlabeled cytochrome-b₅ are presented.

• Dipolar couplings are recoupled by RFDR-MAS for aromatic and side chain protons.

• Residues interacting with membrane are revealed.

• Dipolar couplings with amide-protons are suppressed due to exchange with water.

• T₁ values of ¹H resonances observed in RFDR are shorter than that in NOESY.