Dipolar Assisted Assignment Protocol (DAAP) for MAS solid-state NMR of Rotationally Aligned Membrane Proteins in Phospholipid Bilayers

Abstract

A method for making resonance assignments in magic angle spinning solid-state NMR spectra of membrane proteins that utilizes the range of hetero-nuclear dipolar coupling frequencies in combination with conventional chemical shift based assignment methods is demonstrated. The dipolar assisted assignment protocol (DAAP) takes advantage of the rotational alignment of the membrane proteins in liquid crystalline phospholipid bilayers. Improved resolution is obtained by combining the magnetically inequivalent heteronuclear dipolar frequencies with isotropic chemical shift frequencies. Spectra with both dipolar and chemical shift frequency axes assist with resonance assignments. DAAP can be readily extended to three- and four- dimensional experiments and to include both backbone and side chain sites in proteins.

Introduction

The resolution and assignment of resonances is a requisite first step in protein structure determination by NMR spectroscopy. The resolution of individual resonances is generally accomplished with multidimensional NMR experiments that correlate chemical shift frequencies among directly bonded and proximate homo- and hetero- nuclei. Protocols for chemical shift-based resonance assignments are now well established for magic angle spinning (MAS) solid-state NMR [1–4]. However, in spite of recent advances in methods, it remains difficult to apply chemical shift-based assignment schemes to membrane proteins in phospholipid bilayers, whether in stationary aligned samples in oriented sample (OS) solid-state NMR [5] or in unoriented samples in MAS solid-state NMR [6]. This is largely because these proteins contain many similar hydrophobic residues in uniform helices whose signals have the same or very similar chemical shifts. Here we describe how the highly variable heteronuclear dipolar couplings observed in rotationally aligned (RA) solid-state NMR spectra of membrane proteins [7] can be used to assist in the resolution and assignment process by enabling one of the frequency dimensions in multidimensional spectra to be heteronuclear dipolar coupling frequencies rather than chemical shift frequencies. This is an alternative to isotope labeling based schemes that depend on the preparation of multiple samples [8].

Phospholipids and membrane proteins undergo global motions, in particular rapid rotational diffusion about the bilayer normal [9, 10] and translational diffusion in the plane of the bilayer [11, 12]. The rotational motion results in ‘rotational alignment’ of the proteins that yields frequencies equivalent to those obtained from mechanical or magnetic alignment [10, 13]. This was first demonstrated using ³¹P NMR of phospholipids by McLaughlin et al [14]. It was shown that a single-line chemical shift frequency observed from a uniaxially oriented bilayer sample and the parallel edge of a rotationally averaged powder pattern from an unoriented bilayer sample are equivalent. Later ¹⁵N NMR of single site ¹⁵N labeled polypeptide in lipid bilayer [15] also showed this to be the case for membrane proteins in addition to the phospholipids. An early experiment by Griffin and colleagues demonstrated that ¹³C′-labeled bacteriorhodopsin in unoriented phospholipid bilayers undergoes rotational diffusion around the bilayer normal and importantly that the motion can be switched on and off reversibly by varying the sample temperature [16]. The global motion is either frozen completely or slow enough at lower temperature (below the gel to liquid crystalline phase transition temperature of the phospholipids) to enable the static powder pattern to be observed. Similar results have been obtained for several expressed proteins [15, 17] including a seven transmembrane helix G-protein coupled receptor [2, 18, 19]. This enabled the measurement of orientationally-dependent frequencies werfrom unoriented samples of membrane proteins labeled with ¹⁵N and ¹³C nuclei [7, 19].

The global rotation of the protein occurs fast enough (τ_c ~ 10⁻⁶ s) to motionally average one-bond ¹H-¹³C and ¹H-¹⁵N heteronuclear dipolar coupling (DC) powder patterns in a predictable way. They are axially symmetric, and the span of the powder pattern is reduced by an amount that depends upon the angle between the bond axis and the direction of rotation, in this case the bilayer normal. This defines rotational alignment, a fundamental principle of solid-state NMR demonstrated in some of the earliest experiments for homonuclear dipolar couplings and chemical shift anisotropy [20, 21]. Of direct relevance for structural studies of membrane proteins, the alignment resulting from the rotational motion of the protein about the bilayer normal makes it possible to measure the equivalent angular constraints from unoriented proteoliposome sample as from stationary uniaxially aligned samples [19]. Importantly, in both cases, angles are measured relative to a single external axis (the bilayer normal in the case of RA solid-state NMR), which means that errors do not accumulate. The resulting angular constraints are a major source of input for the calculation of protein structures [7].

Experiments for dipolar assisted resolution and assignment of protein resonances are described in this article. The resolution and assignment method starts with conventional chemical shift based assignments with the addition of recoupling of hetero-nuclear dipolar frequencies in an additional frequency dimension. We show that much higher overall resolution is achievable from the combination of isotropic chemical shift frequencies and anisotropic heteronuclear dipolar coupling frequencies in rotationally aligned membrane proteins. Unambiguous resonance assignments are feasible with this approach for residues that are chemically equivalent but have different heteronuclear dipolar couplings due to their orientation relative to the bilayer normal. The method is demonstrated using an expressed 36-residue polypeptide that includes the residues corresponding to the single helix from the trans-membrane domain of the membrane protein Vpu from HIV-1 (Vpu TM). Even though this polypeptide is relatively small, it provides a rigorous test of the method because it has so many similar hydrophobic residues in its highly regular trans-membrane helix. It was difficult to assign using conventional approaches. Overall, the methods are highly effective in structural studies because in addition to spectral resolution and resonance assignments, structural constraints can be measured from the same three-dimensional spectra.

Methods

Sample preparation

The expression and purification of Vpu TM were performed as previously described [22]. The pET31b(+) vector containing the sequence of Vpu TM fused to that of keto steroid isomerase (KSI) was expressed in C43(DE3) competent cells in isotopically enriched M9 medium suitable for ¹³C/¹⁵N double labeling of proteins. Cells were harvested 5 hrs after Isopropyl β-D-1-thiogalactopyranoside induction, lysed by sonication, and then centrifuged to isolate the inclusion bodies containing the Vpu TM fusion protein. Purification was performed by nickel chelating affinity chromatography followed by cyanogen bromide cleavage to separate the Vpu TM target protein from the KSI fusion partner. Lyophilized protein powder was dissolved in 1:4 (v/v) hexafluoro-isopropanol: dichloromethane and sonicated to precipitate the KSI. The organic solution was filtered through a 0.2um PTFE filter and dried to a film with nitrogen gas. Final pure protein powder was obtained by another filtration of the sample in 1:0.998:0.002(v/v) water: hexafluoro-isopropanol: trifluoroacetic acid followed by solvent removal by lyophilization.

To prepare the proteoliposomes samples used in the solid-state NMR experiments, 2 mg of ¹³C/¹⁵N labeled Vpu TM and 10mg 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC) were dissolved in 1 ml of 0.5% sodium dodecyl sulfate (SDS) buffer. The mixture was shaken for 2 hr., dialyzed against water for 24 hr., and then dialyzed against a solution containing 20 mM potassium chloride (KCl) for 12 hr. to remove any residual SDS. The sample pH was adjusted to 4.0 followed by ultra-centrifugation at 65,000 rpm for 12 hrs. The pellet was then transferred to a 3.2 mm OD MAS rotor.

The sequence of the polypeptide contains 36 residues; those marked in bold and underlined constitute the trans-membrane helix of Vpu and are analyzed in Table 1 and Figure 8: QPIQIAIVA₁₀LVVAI₁₅IIAIV₂₀VWSIV₂₅IIEGRGGKKKK

Table 1.

Residue	15N shift (ppm)	13CA shift (ppm)	13CO shift (ppm)	15N-1H coupling (kHz)	13Cα-1Hα coupling (kHz)
I6	124.3	64.9	175.6	2.4	20.7
A7	120.3	53.7	177.0	3.6	14.0
I8	120.6	63.9	175.2	16.2	21.5
V9	121.6	65.9	175.2	13.6	19.6
A10	122.6	53.7	176.8	6.6
L11	121.4	56.0	176.0	12.0	4.4
V12	122.2	65.7	175.1	18.2	21.0
V13	123.0	66.0	175.3	9.6	7.4
A14	122.5	53.8	175.0	8.0	20.0
I15	120.8	64.0	175.1	16.6	16.3
I16	122.6	64.5	175.5	15.2	19.1
I17	121.3	64.9	175.3	8.7	8.6
A18	122.6	53.8	176.8	10.4	14.5
I19	121.1	64.3	175.3	16.6	2.2
V20	121.8	65.9	176.1	14.0	14.0
V21	121.0	65.8	175.8	9.0	15.0
W22	122.7	58.0	178.1	10.6	2.4
S23		61.7	175.3	15.6	19.3
I24	124.2	64.4	176.2	10.2	19.0
V25	121.7	65.9	175.7	9.2	18.1

Figure 8.

Dipolar wave plot for ¹H-¹⁵N dipolar frequencies as a function of residue type. Comparison between the dipolar waves of normalized dipolar couplings (x) with order parameter S = 0.8 obtained from 14-O-PC/6-O-PC bicelles (q = 3.2) at 42°C and those of dipolar couplings (o) obtained from DMPC liposomes at a spinning rate of 11.111 kHz and 25°C.

NMR Spectroscopy

The experimental data were acquired using the pulse sequences diagrammed in Figure 1. In all the experiments, swept frequency two-pulse phase modulation (SW_f-TPPM) [23] with 100 kHz radio frequency (RF) strength was used for ¹H decoupling and 2.5 kHz RF irradiation for ¹⁵N decoupling with WALTZ-16 [24] modulation. Cross-polarization (CP) [25] from ¹H to either ¹⁵N or ¹³C was optimized using (50%) amplitude-modulated RF irradiation on the ¹H frequency channel. Subsequently, 30 kHz and 60 kHz RF pulses were applied on the ¹⁵N and ¹³C frequency channels, respectively. Varying the CP contact times between 0.1 and 1.0 ms, the carrier frequency offsets, and the decoupling irradiations optimized the sensitivity of individual experiments. Double cross-polarization (DCP) from ¹⁵N to ¹³C was accomplished using spectrally induced filtering in combination with cross-polarization (SPECIFIC-CP) [26]. Adiabatic tangential pulses on the ¹³C channel were optimized for maximum polarization transfer with RF field strengths of 27 kHz for ¹⁵N, 16 kHz for ¹³C and 38 kHz for ¹³CO. Magnetization from ¹³C to ¹⁵N was transferred using z-filter transferred-echo double-resonance (z-TEDOR) experiments [27]. In TEDOR-based experiments, 62.5 kHz and 50 kHz RF irradiations were applied on the ¹³C and ¹⁵N frequency channels, respectively. Sixteen rotor periods with XY8 phase cycling were used for recoupling of heteronuclear dipolar couplings. Homonuclear ¹³C/¹³C correlation in two and three-dimensional experiments was performed using either a 50 ms mixing period under proton driven spin diffusion (PDSD)[28–31] or a 20 ms mixing period with dipolar assisted rotational resonance (DARR) [32].

Figure 1.

Multidimensional pulse schemes for dipolar assisted assignment protocol experiments. (A) Intra or inter residue correlation of ¹H-¹⁵N DC/¹⁵N(Cα)¹³CX or ¹H-¹⁵N DC/¹⁵N(C)¹³CX. (B) ¹H-¹⁵N DC/¹³Cα¹⁵N¹³CO. (C) ¹Hα-¹³Cα DC/¹³C CS/¹³C CS. (D) ¹H-¹⁵N DC/¹Hα-¹³Cα DC/¹³Cα CS. (E) ¹Hα-¹³Cα DC/¹⁵N CS/¹³Cα CS. Thin and wide solid lines are for 90° and 180° pulses. Delta symbol represents the delay for z-filter. CP and DCP represents cross polarization and double cross polarization under SPECIFIC CP condition. CS stands for chemical shift ϕ represents the phase change under States mode.

Recoupling of the hetero-nuclear dipolar coupling frequencies in MAS experiments was performed using symmetry-based R18⁷₁ and R18⁵₂ pulse schemes [33]. A pair of 180° pulses with a phase modulation of 70° (π₇₀-π₇₀) was employed for the R⁷₁ scheme. The scaling factors for the pulse sequences were measured experimentally with ¹³C and ¹⁵N detection using a uniformly ¹³C, ¹⁵N labeled N-acetyl leucine powder sample. The measured dipolar splittings of 6.7 kHz for ¹H-¹³C and 3.4 kHz for ¹H-¹⁵N correspond to a scaling factor of 0.16. Two- and three-dimensional separated local field (SLF) [34] experiments were performed using direct ¹³C-detection with ¹⁵N filtering. ¹H-¹³C and ¹H-¹⁵N dipolar coupling frequencies were correlated to the isotropic chemical shift frequencies from ¹³C and ¹⁵N sites.

Three-dimensional SLF experiments correlating ¹H-¹⁵N DCs with ¹³C and ¹⁵N isotropic chemical shifts were performed using the pulse scheme shown in Figure 1A and B. In Figure 1A, the ¹H magnetization was transferred to ¹⁵N under Hartmann-Hahn CP followed by the dipolar frequency evolution under R18⁷₁ recoupling pulses. The ¹H-¹⁵N dipolar coupling frequencies are recoupled in a constant time evolution period and are encoded in a third dimension. The ¹⁵N magnetization is then flipped back to the laboratory frame using a hard 90° pulse. A z-filter is then employed to dephase remaining magnetization. Low power RF under DARR [32] condition is applied on ¹H channel during the z-filter. ¹⁵N transverse magnetization is created using a flip back 90° pulse followed by the ¹⁵N isotropic chemical shifts evolution encoded in the 2^nd dimension. The phase sensitive chemical shifts are recorded in States mode by varying the phase of the flip back 90° pulse. Heteronuclear decoupling during the evolution is achieved with a 180° hard pulse on ¹³C and SW_f-TPPM pulses on ¹H. ¹⁵N magnetization was then transferred selectively to either ¹³C or ¹³CO using DCP followed by the ¹³C/¹³C spin diffusion mixing period and direct ¹³C detection under heteronuclear decoupling. Three-dimensional data were acquired correlating ¹H-¹⁵N DC frequencies with either N(CA)CX or N(CO)CX two-dimensional planes. The pulse scheme provides an opportunity to correlate intra and inter residues in a sequential manner.

In a similar fashion, simultaneous correlation in intra and inter residues is achieved using the pulse scheme shown in Figure 1B. In this projected three-dimensional experiment, the ¹H-¹⁵N DCs are correlated with ¹⁵N, ¹³C chemical shift frequencies within a residue, and the ¹³CO shift from the preceding residue. The correlation is established following the steps of coherence transfer pathway. 1.) The ¹H magnetization is transferred to ¹³C using Hartmann-Hahn CP. 2.) Magnetization from ¹³C is transferred to ¹⁵N using TEDOR mixing followed by ¹H-¹⁵N dipolar coupling frequency evolution under R18⁷₁ recoupling pulses. 3.) ¹³C magnetization is flipped back to the Z-direction followed by a short z-filter period. 4.) ¹⁵N transverse magnetization is created using a hard 90° pulse followed by ¹H-¹⁵N dipolar evolution under recoupling pulses. 5.) Following the dipolar evolution, ¹⁵N chemical shifts are encoded under ¹H and ¹³C decoupling. Phase sensitive data are collected by varying the phase of the flip back 90° pulse under States mode followed by the second z-filter. 5.) ¹⁵N magnetization is then transferred back to ¹³C using a second TEDOR mixing followed by a third z-filter. Each of the z-filters is synchronized with two-rotor periods of ¹H irradiation under DARR. 6.) In the final step, both ¹³C and inter residue ¹³CO signals are acquired directly in the presence of ¹H and ¹⁵N heteronuclear decoupling.

The measurement of ¹H-¹³C dipolar frequencies within a residue was carried out using the pulse scheme shown in Figure 1C and E. In Figure 1C, the ¹H-¹³C DC is correlated to a two-dimensional ¹³C/¹³C homo-nuclear chemical shift correlation. The three-dimensional experiment is carried out following the steps: 1.) ¹³C magnetization was initially generated by CP from ¹H. 2.) ¹H-¹³C dipolar frequencies are recoupled using the symmetry pulse scheme R18⁷₁ in a constant time manner. The dipolar frequencies are encoded in a third dimension. 3.) Following the dipolar evolution, phase sensitive isotropic ¹³C chemical shifts are encoded with hetero-nuclear decoupling. A hard 180° pulse on ¹⁵N and SW_f–TPPM pulses for ¹H are used for ¹⁵N and ¹H decoupling. Homonuclear correlation is established using spin diffusion with DAAR or PDSD mixing followed by the direct ¹³C detection under ¹H decoupling.

The pulse sequence diagrammed in Figure 1E correlates ¹Hα-¹³Cα dipolar frequencies with ¹⁵N and ¹³Cα isotropic shifts. This is equivalent to the pulse scheme shown in Figure 1D except that ¹⁵N chemical shift is encoded during t1 evolution. ¹⁵N shifts are measured in phase sensitive mode by varying the phase of the ¹⁵N CP pulse. A 10 s hard pulse for ¹⁵N and a 180 s soft pulse with Gaussian shape for ¹³C nuclei are used for refocusing during the spin echo.

¹H-¹⁵N and ¹Hα-¹³Cα dipolar frequency correlation within a residue is obtained using the pulse schemes shown in Figure1D. ¹⁵N edited ¹³Cα shifts are correlated with the dipolar frequencies in three-dimensions. The pulse scheme begins with the ¹H to ¹⁵N CP followed by the ¹H-¹⁵N dipolar recoupling during t1 evolution. Magnetization is then transferred to ¹³Cα nuclei using SPECIFIC-CP. ¹Hα-¹³Cα dipolar frequencies are encoded in two orthogonal dimensions following SPECIFIC-CP. Finally; ¹³Cα signals are detected directly under SW_f -TPPM ¹H decoupling.

The experiments were performed on a 750 MHz spectrometer equipped with Bruker Avance console and a 3.2 mm low-E ¹H/¹³C/¹⁵N triple-resonance MAS probe. The spinning rate was controlled at 11.111 kHz ± 2 Hz. Temperature calibration under MAS was carried out using ethylene glycol as an external reference. The ¹H chemical shift of water was used to monitor the heating of the protein-containing phospholipid bilayer samples in the presence of strong radiofrequency irradiation. It also served as an internal reference frequency. The chemical shift frequencies were referenced externally to solid samples with the methylene ¹³C resonance at 38.48 ppm in adamantane and the ammonium sulfate ¹⁵N resonance at 26.8 ppm.

Results

Rotational diffusion in phospholipid bilayers

Membrane proteins undergo rotational diffusion about the bilayer normal in liquid crystalline phospholipid bilayers at a rate faster than the NMR interaction timescale (τ_c ~ 10⁻⁶ s) [13]. The partial averaging of the powder spectrum depends on the angle between the principal axis of the spin-interaction tensor and the axis of the rotational motion. The most notable effects in the spectra of membrane proteins undergoing fast rotational diffusion are the dramatic reduction in the frequency span of the carbonyl carbon chemical shift powder pattern, and its transformation to axial symmetry [16].

The presence or absence of fast rotational diffusion of Vpu TM reconstituted in DMPC bilayers was monitored as a function of temperature. Here we measure the breadth of the ¹³CO chemical shift powder using simple one-dimensional ¹³C detection with 5 kHz “slow” magic angle spinning. As shown in Figure 2A, a family of sidebands is observed at temperatures below ~15 °C; in contrast, essential no side band intensity can be observed above ~25 °C. The sideband intensities are indicative of the breadth and shape of the underlying chemical shift anisotropy powder patterns of the ¹³CO sites. At low temperatures, the family of side bands represents the rigid lattice powder pattern [35]. On the other hand, the near-total absence of sidebands at higher temperatures confirms that the protein undergoes rotational diffusion. This effect, as plotted in Figure 2B, has also observed in a wide variety of other membrane proteins, including those with seven trans-membrane helices as well as β-barrel membrane proteins.

Figure 2.

Rotational alignment of uniformly ¹³C,¹⁵N labeled single trans-membrane helix from Vpu in DMPC proteoliposome. (A) ¹³C detected one-dimensional spectra recorded at various temperatures ranging from 5 °C to 25 °C. Data were collected using ¹H to ¹³C cross-polarization under 5 kHz magic angle sample spinning. (B) Graphical plot showing the variation in intensity ratios (central peak to the first sideband on the left) as a function of temperature. Central peak is marked with an arrow in A and asterisks denote spinning side bands. One-dimensional data were acquired with 100 kHz spectral width, 15 ms acquisition time, and 2 s recycle delay for 16 scans (5 °C), 32 scans (10 °C, 15 °C, 20 °C) and 64 scans (25 °C).

Resolution in two-dimensional spectra

In order to identify signals associated with the types of amino acid residues in Vpu TM, two-dimensional ¹³C/¹³C homo-nuclear correlation and ¹⁵N/¹³C hetero-nuclear correlation spectra were obtained on a uniformly ¹³C, ¹⁵N labeled sample. The results are shown in Figure 3. These experiments were carried out at 15 °C where the polypeptide is immobile on the relevant NMR timescales (as shown in Figure 2). The ¹³C/¹³C homo-nuclear correlation spectrum obtained with a 50 ms PDSD mixing time shows high intensity cross-peaks that correlate isotropic chemical shift signals from proximate ¹³C sites. Resonances from the Ser, Leu, Trp, and Pro residues are labeled in Figure 3A. Dashed lines mark the correlations of ¹³C resonances in backbone and side chains for the few distinguishable residues in the polypeptide. The overlapped isotropic chemical shifts for the Ala, Val, and Ile residues are also marked; four Ala residues contribute to a single resonance frequency for ¹³Cα and ¹³Cβ, and only two distinct frequencies in the ¹³CO region. The highly overlapped signals from six Val residues show a single frequency for ¹³Cα, ¹³CO, as well as side chain sites. The signals from the 10 Ile residues are significantly overlapped with only three distinguishable ¹³CO resonances, and essentially no resolution among the ¹³Cα and side chain carbon frequencies. The two-dimensional ¹⁵N/¹³C correlation spectrum in Figure 3B displays somewhat better resolution. Nonetheless, in this two-dimensional N(CA)CX spectrum only ~13 individual signals are distinguishable. Two signals from the Ala and Val residues are well resolved, and ~ 8 signals are recognizable from the 10 Ile residues.

Figure 3.

¹³C detected two-dimensional correlation spectra of uniformly ¹³C, ¹⁵N labeled Vpu-TM. (A) ¹³C/¹³C correlation spectrum obtained from 50 ms proton driven spin diffusion (PDSD) mixing. Single resonance assignment for ¹³C shifts in P3, L11and S23 are labeled. Chemical shift dispersion for other residues such as Ala, Val, Ile, Glu, Gln, Trp, and Lys are also been labeled. (B) Two-dimensional N(CA)CX correlation spectra. 20 ms DARR mixing was used for carbon spin exchange in B. Residue numbers are marked following the sequential resonance assignment. Homo- and hetero- nuclear correlation two-dimensional spectra were acquired with 128 scans (¹³C/¹³C) and 512 scans (¹³C/¹⁵N) with a 2 s recycle delay. Experiments were carried out with 242 ppm (¹³C), 80 ppm (¹³C) and 32 ppm (¹⁵N) spectral widths for direct and indirect acquisition. The acquisition periods were 12 ms for direct and 4 ms (¹³C)/6 ms (¹⁵N) for indirect detection. 100 μs, 500 μs and 3000 μs contact times were used for ¹H to ¹³C, ¹H to ¹⁵N and ¹⁵N to ¹³C cross-polarization, respectively.

By themselves, three-dimensional experiments correlating resonances from back bone and side chains in a CCC type experiment do little to resolve the severe degeneracies in the chemical shifts of the hydrophobic residues. However, three-dimensional correlation experiments such as NCACX and NCOCX used in sequential assignment show somewhat better resolution [36]. Two-dimensional planes correlating ¹³C signals from backbone and side chains extracted at ¹⁵N shift frequencies are shown in Figure 4. Irrespective of the resolution gain in three-dimensional experiments, unambiguous assignment is not possible due to the overlap of many chemical shift frequencies, for example for V9, V13, I16, I17, I26, I27. However, as shown infra the large variation among heteronuclear dipolar coupling frequencies in rotationally aligned samples [7, 19] can be used to increase the resolution and assist in making unambiguous assignments of resonances.

Figure 4.

Strip-plots for resonance assignment for residues V9–V12 in uniformly¹³C,¹⁵N labeled Vpu TM. Two-dimensional ¹³C/¹³C correlation plot from NCACX (red) and NCOCX (blue) data extracted at ¹⁵N shifts. ¹³C shifts are marked according to their positions in backbone and side chains. Three-dimensional data were acquired with 256 scans (NCACX) and 512 scans (NCOCX). The experiments were carried out under similar conditions to those described in Figure 3 except that the homonuclear spin diffusion was obtained with 40 ms DARR mixing, and 40 ppm and 20 ppm ¹³C spectral widths for NCACX and NCOCX, respectively.

As demonstrated in Figure 2, above 25 °C the protein undergoes rotational diffusion about the bilayer normal resulting in partial averaging of both chemical shift and dipolar coupling powder patterns, which varies depending upon the alignment of the tensors in the molecular frame relative to the bilayer normal. Using symmetry based pulse schemes [33, 37], ¹H-¹⁵N dipolar frequencies were recoupled and correlated to ¹³C isotropic chemical shifts in two-dimensional experiments. ¹³C-detected SLF spectra of Vpu TM at 25 °C are shown in Figure 5. The spectra were recorded using R18⁷₁ pulses for hetero-nuclear dipolar recoupling under 11 kHz MAS. ¹H-¹⁵N dipolar coupling and ¹H-¹³C dipolar coupling frequencies were encoded in a constant time spin echo period with refocusing of chemical shifts. Dipolar frequencies were measured for evolution periods of 4.68 ms for ¹H-¹⁵N DC and 2.88 ms for ¹H-¹³C DC. For ¹H-¹⁵N/¹³Cα correlations, DCP was used to transfer coherence from ¹⁵N to ¹³Cα sites using the pulse sequence shown in Figure 1A without the inclusion of ¹⁵N shift evolution and spin diffusion periods. In a similar fashion, ¹H-¹³C dipolar coupling frequencies were measured using the pulse sequence shown in Figure 1C without incorporation of spin-diffusion period and direct detection of ¹³C shifts. As can be seen from the figure several distinguishable cross-peaks are observed for a single chemical shift. This is mainly due to the magnetic in-equivalence in the helix caused by the topological arrangement relative to the bilayer normal. For example, in Figure 5A, a single chemical shift at 53.5 ppm has at least three resolved cross peaks in the ¹H-¹⁵N dipolar frequency dimension. Similar resolution is also observed for ¹H-¹³C dipolar coupling frequencies (Figure 4B). Even though the introduction of the dipolar coupling frequencies is a step forward in resolution enhancement for residues with very similar or identical chemical shifts, complete resolution is not achievable for Val and Ile residues using two-dimensional experiments.

Figure 5.

¹³C-detected two-dimensional separated local field NMR spectra. (A) ¹⁵N edited spectrum correlating ¹³Cα chemical shifts and ¹H-¹⁵N dipolar couplings. The spectrum was obtained using the pulse scheme shown in Figure 1A without the ¹⁵N chemical shift evolution and spin diffusion mixing periods. (B) ¹³C CS/¹H-¹³C DC correlation spectrum. All spectra were recorded under 11.111 kHz MAS at 25 °C. Two-dimensional data were acquired with 512 scans (A) and 128 scans (B) using a 2 s recycle delay. Dwell times of 90 μs for A and 30 μs for B were used in the indirect dimensions to record the spectra. A 12 ms acquisition time was used for direct ¹³C detection.

Resolution in three-dimensional spectra

Superior resolution is observed in three-dimensional separated local field experiments with the incorporation of ¹⁵N chemical shift frequencies. Three-dimensional SLF spectra correlating ¹Hα-¹³Cα and ¹H-¹⁵N DCs with ¹⁵N and ¹³Cα shifts were acquired using the pulse schemes diagrammed in Figure 1A and D. The spectra were recorded with 2.16 ms ¹H-¹³C and 3.76 ms ¹H-¹⁵N dipolar evolution, and 6 ms ¹⁵N and 10 ms ¹³C evolution for indirect and direct detection. Two-dimensional SLF planes correlating ¹H-¹⁵N/¹³Cα and ¹Hα-¹³Cα/¹³Cα at ¹⁵N shift frequencies are shown in Figure 6A–D. Figure 6A and C represent the ¹H-¹⁵N/¹³Cα and ¹Hα-¹³Cα/¹³Cα planes at the 121.6 ppm ¹⁵N chemical shift frequency. Three distinguishable ¹H-¹³C dipolar frequencies are observable compared to a single ¹H-¹⁵N dipolar frequency at a single resonance for valine. The two-dimensional SLF planes shown in Figure 6B and D were extracted at the 122.8 ppm ¹⁵N chemical shift frequency. Isotropic shifts attributed to Ala residues show three well-resolved contours for ¹H-¹⁵N dipolar frequencies compared to two signals with ¹H-¹³C dipolar frequencies. Somewhat better resolution was achieved in a three-dimensional experiment correlating ¹H-¹⁵N and ¹Hα-¹³Cα dipolar frequencies with ¹³Cα isotropic shifts using the pulse sequence shown in Figure 1D. The two-dimensional SLF planes correlating ¹H-¹⁵N DCs with ¹Hα-¹³Cα DCs extracted at 65.7 ppm and 53.5 ppm ¹³Cα shifts are shown in Figure 6E and F. The dipolar frequencies were encoded in constant spin echo periods for 5.12 ms ¹H-¹⁵N and 2.88 ms ¹H-¹³C dipolar evolutions, respectively.

Figure 6.

¹³C detected three-dimensional separated local field spectra of uniformly ¹³C,¹⁵N labeled Vpu TM. (A) and (B) Two-dimensional SLF planes correlating ¹³Cα CS/¹H-¹⁵N DC extracted from a three-dimensional SLF data set for ¹⁵N shifts at 121.6 ppm and 122.8 ppm, respectively. The spectrum was recorded using the pulse scheme shown in Figure 1A without incorporating the spin diffusion mixing. (C) and (D) Two-dimensional SLF planes correlating ¹³Cα CS/¹Hα-¹³Cα DC for ¹⁵N shifts at 121.6 ppm and 122.8 ppm, respectively. The experiment correlating ¹³Cα CS/¹⁵N CS/¹Hα-¹³Cα DC was acquired using the pulse scheme in Figure 1E. (E) and (F) Two-dimensional SLF planes correlating ¹H-¹⁵N/¹Hα-¹³Cα dipolar frequencies extracted from a three-dimensional SLF data set for ¹³C shifts at 65.7 ppm and 53.5 ppm, respectively. The three-dimensional data was collected using the pulse scheme in Figure 1D. (G) ¹³C/¹³C two-dimensional plane obtained at 9.9 kHz ¹Hα-¹³Cα DC. Three-dimensional SLF data correlating ¹Hα-¹³Cα DC/¹³C CS/¹³C CS recorded using the pulse scheme shown in Figure 1C. All spectra were collected at a spinning frequency of 11.111 kHz and 25 °C sample temperature. t1 noise from lipid signals is denoted with asterisks. The three-dimensional data in A–F were collected with 512 scans (A–F) or128 scans (G). A 2 s recycle delay and 90 μs dwell time were used in the experiments.

The three-dimensional spectrum correlating ¹H-¹³C DCs with ¹³C/¹³C shifts was acquired using the pulse scheme shown in Figure 1C. The spectrum was recorded with a 2.16 ms dipolar evolution, and 5 ms/10 ms ¹³C evolution for indirect/direct detection. Homonuclear correlation was established using 50 ms PDSD mixing and 100 μs cross-polarization mixing during preparation. The two-dimensional ¹³C/¹³C correlation plane correlated to 9.9 kHz ¹H-¹³C DC is shown in Figure 6G. Each cross peak in the two-dimensional planes corresponds to a single site resolution.

Dipolar Assisted Assignment Protocol (DAAP)

Irrespective of single site resolution in three-dimensional SLF experiments, measurement of dipolar frequencies is still ambiguous due to the degeneracy in ¹⁵N and ¹³Cα chemical shits. Here we illustrate how the new assignment protocol assists in the assignment of chemical shift and dipolar frequencies. The three-dimensional experiments correlating dipolar coupling and chemical shift frequencies are used for establishing sequential assignments in backbone and side chain atoms similar to chemical shift assignment experiments such as NCACO, NCOCA and CANCO. A three-dimensional correlation ¹H-¹⁵N DC/¹⁵N CS/¹³CX CS spectrum is shown in Figure 7. The experiments correlating ¹H-¹⁵N DC with N(CA)CX or N(CO)CX heteronuclear correlation were obtained using the pulse sequence diagrammed in Figure 1A. Cross-polarization from ¹⁵N to either ¹³Cα or ¹³CO was performed using 3 ms DCP mixing. Homo-nuclear polarization transfer was carried out with 20 ms DAAR. A two-dimensional plane correlating ¹⁵N/¹³C chemical shifts obtained at a ¹H-¹⁵N dipolar coupling frequency of 2.4 kHz is shown in Figure 7A. The intra-residue correlation spectrum was assigned to Val residue (V13) in the sequence. Similarly, a two-dimensional plane correlating ¹³CO and ¹³Cα chemical shifts of V13 to the ¹⁵N chemical shift of A14 was obtained for ¹H-¹⁵N dipolar coupling frequency at 2.55 kHz. Both intra- and inter- residue shift correlation with ¹H-¹⁵N dipolar frequencies was obtained using the pulse sequence shown in Figure 1B. In this experiment, the ¹³CO chemical shift of the preceding residue (i-1) is correlated with ¹H-¹⁵N dipolar coupling and the ¹⁵N, ¹³Cα chemical shift frequencies of the same residue (i). Figure 7C represents the two-dimensional plane correlating the ¹³CO chemical shift of V13 to the ¹⁵N and ¹³Cα chemical shifts of A14. The two-dimensional plane was obtained for 2.55 kHz ¹H-¹⁵N dipolar frequency. Following the chemical shifts and dipolar correlation, ¹H-¹⁵N dipolar frequencies at 2.4 kHz and 2.55 kHz were assigned to V13 and A14 respectively.

Figure 7.

¹³C detected three-dimensional SLF data correlating ¹³C and ¹⁵N isotropic chemical shifts with ¹H-¹⁵N dipolar frequencies. (A) N(CA)CX two-dimensional plane obtained at 2.4 kHz ¹H-¹⁵N dipolar coupling (DC) frequency. The three-dimensional data were collected using the pulse scheme shown in Figure 1A with a 20 ms DARR (dipolar assisted rotational resonance) mixing. (B) N(CO)CX 2D-plane at 2.55 kHz ¹H-¹⁵N DC. The three-dimensional data were collected using the pulse scheme shown in Figure 1A with 40 ms DARR mixing. (C) CONCA two-dimensional correlation plane obtained for 2.55 kHz ¹H-¹⁵N DC. The three-dimensional data were collected using the pulse scheme shown in Figure 1B with 1.44 ms TEDOR mixing. Dashed lines illustrate the connectivity for residues A14 and V13 for ¹⁵N and ¹³C shifts. (D) ¹³C/¹³C two-dimensional plane obtained at 7.4 kHz ¹Hα-¹³Cα DC. Three-dimensional SLF data correlating ¹Hα-¹³Cα DC/¹³C CS/¹³C CS recorded using the pulse scheme shown in Figure 1C. The three-dimensional data were collected with 1024 scans (A–C), 128 scans (D), 2 s recycle delay and 90 μs dwell time.

Three-dimensional data correlating ¹H-¹³C DC/¹³C CS/¹³C CS frequencies were obtained using the pulse scheme shown in Figure 1C. A two-dimensional plane with ¹³C/¹³C shift correlation at a dipolar coupling frequency of 7.4 kHz is shown in Figure 6D. The correlation of ¹³Cα with ¹³CO and side chain carbons is observed at a selected ¹Hα-¹³Cα dipolar frequency. The other advantage of this experiment is that side chain dipolar frequencies are also observable, opening up a promising path to resolution, assignment, and structure determination of the side chains. Thus, DAAP is applicable to both backbone and side chain sites. However, data for side chains are not included here because this topic is well beyond the scope of the current discussion. Following the method of dipolar coupling and chemical shift correlation spectroscopy supplemented by the dipolar assisted assignment protocol, the dipolar frequencies and chemical shifts of all residues in Vpu TM were assigned unambiguously. These are collated in Table 1 and Figure 8.

Rotationally aligned solid-state NMR and structure determination of a transmembrane helix

¹H–¹⁵N dipolar coupling values measured for each residue in Vpu TM are plotted as a function of the residue number in Figure 8A. A characteristic Dipolar Wave pattern is observed that enables helical segments of the protein to be readily identified and characterized with respect to the helix length, helix orientation in the membrane, and presence of any kinks [38]. Notably, the ¹H–¹⁵N DC values measured here for Vpu TM in proteoliposomes (open circles) are very similar to the values measured previously for the same protein in mechanically aligned bilayers (cross marks) and magnetically aligned bilayers (q = 3.2 bicelles), demonstrating that information obtained from motion averaged powder patterns of proteins in liposomes, recoupled under MAS, is the same as that from single line OS solid-state NMR spectra of uniaxially aligned proteins.

The experimental measurements from an aligned, stationary bicelle sample were taken from an earlier publication. Although there is some variation in the ¹H-¹⁵N dipolar coupling values that are plotted, most points are nearly identical, and they all fall within the 10% estimated margin of error and reproducibility. The original measurements of 1H-15N dipolar coupling in DMPC bilayers on glass plates yielded a value of 27°, and in DMPC bicelles a value of 30°. Here the calculation in DMPC proteoliposomes include both 1H-15N and 1H-13C dipolar couplings, and yield a value of 28°. Taking into account the somewhat different sample conditions and different experimental methods, we consider all of these values to be consistent [39] [22].

Summary

The ambiguities in resonance assignments of membrane proteins are mainly due to the poor chemical shift dispersion. However, rotational alignment resulting from the global motion of the protein provides variation in the dipolar frequency for residues those are chemically equivalent but magnetically inequivalent. Here, we show experimental methods that correlate dipolar coupling frequencies with isotropic chemical shifts to enhance spectral resolution and aid in making resonance assignments. We refer to this as a dipolar assisted assignment protocol. In DAAP, ¹³C chemical shift resonance frequencies, mainly ¹³Cα and ¹³CO, within a residue are correlated to the ¹⁵N chemical shift resonances and ¹H-¹⁵N dipolar coupling frequencies of adjacent residues. The correlations were established using the three-dimensional experiments diagrammed in Figure 1. In these experiments, the ¹H-¹⁵N dipolar coupling frequencies are correlated to the resonances observed in two-dimensional heteronuclear N(CA)CX, N(CO)CX and CANCO correlation spectra. The results are shown in Figures 3 – 7 for a uniform ¹³C, ¹⁵N labeled membrane protein. Single site resolution for nearly all resonances is observed. Variation in dipolar frequencies adds further resolution to the severely crowded spectra obtained from chemical shift correlation alone. Unambiguous assignment of residues is made possible due to the gain in resolution. These pulse schemes can also be used for isotope labeled and partial isotope labeled samples. As shown in the results, the method also benefits with accurate measurement of dipolar frequencies for residues with similar ¹³C and ¹⁵N chemical shifts. It is now possible to assign and measure dipolar frequencies in a sequential manner. Dipolar waves are compared for results obtained from two sample preparation methods for Vpu TM in phospholipid bilayers. Very similar dipolar waves confirm the similar topological arrangement for the protein whether made from uniaxially oriented samples from mechanical alignment in phospholipid bilayers or magnetically aligned bicelles or rotationally aligned unoriented proteoliposome. Equivalent angular constraints are measured from the unoriented proteoliposome samples using the pulse sequences reported in this article. The pulse schemes can also be extended to higher dimensions to obtain superior resolution. For instance, Figure 1A can readily be extended to a four-dimensions by adding ¹³Cα or ¹³CO chemical shift evolutions after DCP transfer. In a similar fashion Figure 1B can be extended to a pseudo 4D or 5D experiments to measure ¹³C-¹⁵N dipolar couplings. This can be achieved by adding a variable mixing period for TEDOR transfer and incorporating ¹³C shift evolution after CP. In the future, it should also be possible to perform ¹H NMR analogs of these experiments [40]. These methods can be readily extended to side chain resonances, enabling complete structure determinations of membrane proteins.

Highlights.

• Equivalent angular constraints are obtained for MAS of rotationally aligned Vpu TM in phospholipid bilayers.

• Pulse sequences for dipolar assisted assignment protocol are presented.

• Multi-dimensional experiments are proposed to measure chemical shifts and dipolar frequencies in rotationally aligned membrane proteins.

• Chemical shift and dipolar frequency correlation provide high resolution and assist in unambiguous resonance assignment.