Solid-State NMR of a Large Membrane Protein by Paramagnetic Relaxation Enhancement

Abstract

Membrane proteins play an important role in many biological functions. Solid-state NMR spectroscopy is uniquely suited for studying structure and dynamics of membrane proteins in a membranous environment. The major challenge to obtain high quality solid-state NMR spectra of membrane proteins is sensitivity, due to limited quantities of labeled high-molecular-weight proteins. Here we demonstrate the incorporation of paramagnetic metal (Cu²⁺) ions, through either EDTA or a chelator lipid, into membrane protein samples for rapid data collection under fast magic-angle spinning (MAS) and low power ¹H decoupling. Spectral sensitivity of DsbB (20 kDa), an integral membrane protein, more than doubles in the same experimental time due to ¹H T₁ relaxation enhancement by Cu²⁺ ions, with DsbB native fold and active site intact. This technique can be implemented to acquire multidimensional solid-state NMR spectra for chemical shift assignments and structure elucidation of large membrane proteins with small sample quantities.

Solid-state NMR spectroscopy (SSNMR) is a powerful and versatile tool for studying structural biology and biophysics of membrane proteins in a membranous environment.^1–14 A variety of structural and dynamic information, such as internuclear distances, dihedral angles, tensor orientations and order parameters, can be extracted by rational design of spin labeling and pulse sequences. The rich information can then be used to elucidate the functional mechanisms of membrane proteins.¹⁵ However, a major hurdle preventing SSNMR from fast acquirement of structural information is the low sensitivity of membrane proteins, because the molar quantity of isotopically labeled protein that can be packed into magic-angle spinning rotors is limited by the high molecular weight and the presence of required amount of lipids to mimic the membranous environment. New methods under development in recent years have great potential to significantly improve the sensitivity of SSNMR samples. For example, paramagnetic relaxation enhancement (PRE) effect of paramagnetic metal ions can reduce ¹H spin-lattice relaxation times (T₁), which govern experimental times of most SSNMR measurements.¹⁶ Previous demonstrations have shown SSNMR samples with paramagnetic metal ions to speed up the data collection, such as Cu(II)-EDTA for fibrillar protein,¹⁶ a Co(II) replaced metal center for metalloprotein¹⁷ and covalently bound paramagnetic tags for nanocrystalline protein mutants.¹⁸ All those experiments were carried out under fast (> 35 kHz) magic-angle spinning (MAS) to take advantage of low power ¹H decoupling to minimize the radio-frequency induced sample heating, which allows for significant reduction of the recycle delay. The fast MAS SSNMR rotors also require smaller amounts of labeled protein, which are especially beneficial for the low molar quantities of large membrane proteins.

Recently, a novel copper chelated lipid was demonstrated to shorten ¹H T₁ of an antimicrobial membrane peptide in magnetically aligned bicelles with enhanced signals for static SSNMR experiments.¹⁹ The usage of paramagnetic metal ions in large membrane protein samples will significantly facilitate SSNMR data collection. Nevertheless, the effort of incorporating paramagnetic metal ions in large membrane protein samples is still nontrivial, and the effects of the metal ions and fast MAS on spectral characteristics and structures of membrane proteins remain unexplored.

Here we present the sensitivity enhancement of an integral membrane protein DsbB (20 kDa) under fast MAS conditions by incorporating Cu²⁺ ions through either EDTA or a chelator lipid. DsbB belongs to a disulfide bond generating enzymatic system in E. coli, including periplasmic protein DsbA, membrane protein DsbB and cofactor ubiquinone, which is responsible for forming disulfide bonds in substrate proteins.²⁰ The combination of short ¹H T₁ and low power ¹H decoupling under fast MAS resulted in five-fold reduction of experimental times of two-dimensional ¹³C-¹³C correlation spectra. The spectra show the same correlations as the spectra collected on the DsbB samples that do not contain copper. These results demonstrate that the PRE effect of paramagnetic metal ions can be applied to large membrane proteins, without disrupting the native fold and the active site, to enhance the sensitivity of fast MAS SSNMR experiments for chemical shift assignments and distance measurements, and thus accelerate structure determination of membrane proteins.

Three samples of uniformly labeled DsbB were prepared to investigate the PRE effect on membrane proteins: (1) DsbB without copper as a control, (2) DsbB with Cu²⁺ in the chelator lipid – 14:0 PE-DTPA (DsbB/CuDMPE-DTPA), (3) DsbB with Cu²⁺ in EDTA (DsbB/CuEDTA). The ¹H T₁ of the three samples were measured under the same condition with 36 kHz MAS rate and the sample temperature ~ −15 °C to achieve the optimal resolution and sensitivity as determined in the previous study.²¹ Table 1 summarizes the ¹H T₁ values of DsbB, DsbB/CuDMPE-DTPA and DsbB/CuEDTA. Clearly, the Cu²⁺ ions shorten the ¹H T₁ significantly by 10 and 6 times for CuDMPE-DTPA and CuEDTA, respectively. In principle, the recycle delay of the cross-polarization (CP) experiments from proton can be as short as three times the ¹H T₁, i.e. 0.15 s. However, in practice, the recycle delays of 0.3~0.5 s were used considering the high-power radio-frequency irradiation during CP and recoupling periods to keep the duty cycle under 3% in combination with fast MAS and low power decoupling during acquisitions. Even with this limitation the experimental time was shortened by up to 10 times with Cu²⁺ doping compared to the conventional experiments with the recycle delays of 1.5~3 s.

Table 1.

¹H T₁ values of DsbB, DsbB/CuDMPE-DTPA and DsbB/CuEDTA under 36 kHz MAS.

Samples	DsbB	DsbB CuDMPEDTPA	DsbB CuEDTA
1H T1/ms	490	50	80

The samples of DsbB with and without Cu²⁺ ions were characterized by SSNMR two-dimensional (2D) ¹³C-¹³C correlation spectra, from which one can assess the PRE effects of Cu²⁺ ions on the resolution, sensitivity and secondary structure distribution of DsbB samples. Figure 1 shows the 2D ¹³C-¹³C correlation spectra of DsbB/CuDMPE-DTPA and DsbB with the similar experimental times (~40 h) and similar amounts of labeled proteins (~3 mg in 1.6 mm rotors). The difference between the two experiments lies in the recycle delays. The spectrum collected of DsbB with CuDMPE-DTPA was collected with a five-fold reduction in recycle delay (from 1.5 s to 0.3 s). Between these two samples, the observed chemical shift resonances remain the same, as shown in Table 2, indicating that the chelator lipids (DMPE-DTPA) did not disturb the secondary structure of DsbB in the membrane.

Figure 1.

2D ¹³C-¹³C correlation spectra of (a) DsbB/CuDMPE-DTPA (recycle delay 0.3 s, total time 46 h) and (b) DsbB (recycle delay 1.5 s, total time 43 h) under 36 kHz MAS with RFDR 7.1 ms mixing (B₀ = 17.6 T, T_sample = −15 °C, maximum t₁ = 15.36 ms, maximum t₂ = 20.48 ms). The crosspeaks of residues that are resolved in both spectra are marked. Both data were processed with 30 and 30 Hz net line broadening in each dimension and plotted at the same contour levels, with the lowest contour at 5σ.

Table 2.

Comparison of ¹³C chemical shifts of resolved resonances in 2D ¹³C-¹³C correlation spectra of DsbB, DsbB/CuDMPE-DTPA and DsbB/CuEDTA.

Chemical shift (ppm)	DsbB	DsbB CuDMPE-DTPA	DsbB CuEDTA
G11 Cα	43.8	43.3	43.4
G11 C	172.4	172.3	172.4
T21 Cγ2	22.6	22.3	22.4
T21 Cβ	69.0	69.0	68.9
G61 Cα	45.7	45.3	45.3
G61 C	173.6	173.6	173.6
A64 Cβ	19.2	18.9	19.0
A64 Cα	53.3	53.1	53.1
T67 Cγ2	21.9	21.6	21.5
T67 Cβ	69.9	69.8	70.0
T88 Cγ2	24.8	24.8	24.7
T88 Cβ	68.8	68.8	68.7
G148 Cα	47.9	47.6	47.6
G148 C	174.5	174.6	174.8
A168 Cβ	18.9	18.9	18.8
A168 Cα	51.2	51.1	51.1

We further evaluated the various ways of introducing Cu²⁺ ions into membrane protein samples: (1) CuDMPE-DTPA in the membrane; (2) CuEDTA in solution. The concentrations of Cu²⁺ ions in DsbB/CuDMPE-DTPA and DsbB/CuEDTA are ~0.6 mM and 2.5 mM, respectively. Figure 2 shows the comparison of 2D ¹³C-¹³C correlation spectra of DsbB/CuDMPE-DTPA and DsbB/CuEDTA. Both spectra exhibit the same chemical shift resonances of DsbB as shown in Table 2, indicating that CuEDTA does not affect DsbB structure either. The resolved crosspeak signal-to-noise ratios (S/N) of the residues (G11, T21, G61, A64, T67, T88, G148, A168) in DsbB/CuDMPE-DTPA and DsbB/CuEDTA samples are summarized in Figure 3. The sensitivity of DsbB/CuDMPE-DTPA and DsbB/CuEDTA are consistently enhanced (1.1 ~ 3.6) compared to DsbB without copper due to PRE effect and reduced recycle delays. Interestingly, G11, T67 and A168 in the loop regions of DsbB have less sensitivity enhancement compared to the residues in the transmembrane helices. This is probably because the residues in the loops or termini are more mobile than the transmembrane residues, and thus have slower ¹H spin diffusion rate, which is the mechanism for ¹H magnetization transfer through the samples.¹⁶ Hence, the transmembrane residues are more rigid and spin-diffuse faster to obtain the ¹H magnetization by PRE from Cu²⁺ ions. Another explanation would be that the residues in the loops or termini might be close to Cu²⁺ ions to have shorter T_1ρs than transmembrane residues, and thus have less cross-polarization intensities.

Figure 2.

2D ¹³C-¹³C correlation spectra of (a) DsbB/CuDMPE-DTPA (recycle delay 0.3 s) and (b) DsbB/CuEDTA (recycle delay 0.5 s) under 36 kHz MAS with RFDR 3.6 ms mixing (B₀ = 11.7 T, T_sample = −15 °C, maximum t₁ = 15.36 ms, maximum t₂ = 20.48 ms). The crosspeak pattern is the same between the two samples, indicating both CuDMPE-DTPA and CuEDTA are effective in shortening ¹H T₁ of membrane protein DsbB. Both data were processed with 40 and 40 Hz net line broadening in each dimension and plotted at the same contour levels, with the lowest contour at 5σ.

Figure 3.

(a) Relative signal to noise ratios (S/N) of DsbB/CuDMPE-DTPA (blue) and DsbB/CuEDTA (red) in the selected residues that are clearly resolved in all three 2D ¹³C-¹³C spectra. All S/N of corresponding crosspeaks are normalized to those in 2D ¹³C-¹³C spectrum of DsbB without copper, considering the same sample quantities and experimental times. (b) Schematic respresentation of DsbB topology in the membrane. Residues selected in (a) are marked to indicate their locations relative to the membrane. T21, G61, A64, T88 and G148 are in the transmembrane helices, while G11, T67 and A168 are in the loops or termini.

All these 2D ¹³C-¹³C correlation experiments used the radio-frequency field-driven dipolar recoupling^22–24 (RFDR) for ¹³C-¹³C mixing. This mixing scheme was chosen because the commonly utilized dipolar recoupling schemes involving protons, like proton-driven spin diffusion (PDSD),²⁵ dipolar-assisted rotational resonance (DARR)²⁶ and radio-frequency-assisted diffusion (RAD),²⁷ do not work well under fast MAS conditions. Additionally, RFDR takes advantage of the high power ¹³C B₁ field without requiring high power ¹H heteronuclear decoupling during the mixing to yield the efficient ¹³C-¹³C homonuclear recoupling of DsbB with rapid data collection, similar to the experiments performed on microcrystalline and fibrillar protein samples that were reported before.^{16, 28}

Recently, a few amplitude- or phase-modulated spin diffusion techniques have been developed for fast MAS, such as mixed rotational and rotary-resonance (MIRROR),²⁹ phase-alternated recoupling irradiation scheme (PARIS)³⁰ and rotor-synchronized symmetry-based recoupling (R2₂¹).³¹ We also utilized the R2₂¹ scheme to collect 2D ¹³C-¹³C correlation spectra to compare with RFDR. As shown in Figure 4, the 2D ¹³C-¹³C correlation spectra of DsbB/CuDMPE-DTPA with 7.1 ms of RFDR mixing and 300 ms of R2₂¹ mixing have similar correlations. In RFDR mixing, only one-bond or two-bond correlations were observed, while in R2₂¹ mixing there were additional three-bond correlations present, such as Ile Cα-Cδ1 and Leu Cα-Cδ. This is most likely due to R2₂¹ suffering less from dipolar truncation than RFDR.^{23–24, 31} However, R2₂¹ requires much longer mixing time than RFDR (300 ms vs. 7.1 ms) to yield similar crosspeak intensities due to its low power irradiation and reduced recoupling efficiency at fast MAS, which greatly increases the experimental time to collect the equal-quality spectra. Further, paramagnetic Cu²⁺ ions might reduce ¹³C T₁ as well, causing lower sensitivity for prolonged ¹³C-¹³C longitudinal mixing time for R2₂¹. Due to short mixing time requirement, easy experimental setup and sequence robustness, RFDR is advantageous for fast MAS and rapid data collection.

Figure 4.

2D ¹³C-¹³C correlation spectra of DsbB/CuDMPE-DTPA (a) with RFDR 7.1 ms mixing and (b) with R2₂¹ 300 ms mixing under 36 kHz MAS (B₀ = 11.7 T, T_sample = −15 °C, recycle delay 0.3 s, maximum t₁ = 15.36 ms, maximum t₂ = 20.48 ms). The dashed circle indicates Ile Cα-Cδ1 correlations that are more prominent in R2₂¹ 300 ms mixing. Both data were processed with 30 and 30 Hz net line broadening in each dimension and plotted at the same contour levels, with the lowest contour at 5σ.

In addition, the sensitivity enhancement from rapid data collection and the broadband recoupling from RFDR provide good crosspeak intensities for the cofactor ubiquinone-8 (UQ-8) (Figure 5) for ~ 40 h, which required long time signal averaging (~ 70 h) to observe previously at moderate MAS frequencies (10–20 kHz) with the twice amount of labeled protein.²¹ The chemical shifts obtained from DsbB/CuDMPE-DTPA are the same as those assigned from the previous study,²¹ as listed in Table 3, indicating that Cu²⁺ ions in chelated forms do not perturb the active site of charge-transfer complex between DsbB and UQ-8.²⁰ This is also supported by the identical UV-Vis spectra of three DsbB samples with and without copper (Figure 6) showing the signature 500 nm absorbance corresponding to the charge-transfer complex.

Figure 5.

(a) UQ-8 structure with IUPAC nomenclature. (b) Expansion of UQ-8 region of 2D ¹³C-¹³C correlation spectra of DsbB/CuDMPE-DTPA under 36 kHz MAS with RFDR 7.1 ms mixing (B₀ = 11.7 T, T_sample = −15 °C, recycle delay 0.3 s, maximum t₁ = 15.36 ms, maximum t₂ = 20.48 ms). The UQ-8 C5′-C5 peak was marked and dashed circles highlighted the UQ-8 correlations. The data were processed with 30 and 30 Hz net line broadening in each dimension.

Table 3.

Comparison of ¹³C chemical shifts of UQ-8 in DsbB/CuDMPE-DTPA and DsbB. Chemical shifts of UQ-8 in DsbB are from the previous study.²¹

Chemical Shift (ppm)	UQ-8 in DsbB CuDMPE-DTPA	UQ-8 in DsbB21
CH3 (5′)	15.6	15.9
CH3 (9′~37′)	20.5~17.8	20.7~17.9
CH3 (38)	28.0	28.2
CH2 (11~35)	30.4~27.5	30.3~27.3
CH2 (10~34)	43.8~41.5	43.4~41.6
CH3O (2′, 3′)	63.2, 63.7	63.6, 64.0
C=C (12~36)	126.0~123.0	126.6~123.6
C=C (13~37)	138.6~135.0	139.0~135.2
C=C (9)	140.2	140.2
C=O (1, 4)	154.5	154.7

Figure 6.

UV-Vis spectra of DsbB, DsbB/CuEDTA and DsbB/CuDMPE-DTPA.

In summary, we have demonstrated the sensitivity enhancement of membrane protein samples under fast MAS through the incorporation of paramagnetic metal ions (CuDMPE-DTPA and CuEDTA). The combination of fast T₁ relaxation enhanced by Cu²⁺ ions and low power ¹H decoupling enabled by fast MAS facilitated rapid data collection for a large membrane protein (20 kDa). The experimental times are reduced by more than five times. 2D ¹³C-¹³C spectra of DsbB with CuDMPE-DTPA or CuEDTA resulted in the same correlations as the normal DsbB sample without copper, indicating that the copper dopants did not affect the structure of DsbB, especially for the cofactor UQ-8 in the active site. The ¹³C-¹³C mixing scheme of RFDR has been found to be optimal for chemical shift assignments of intraresidual correlations considering the recoupling efficiency and the economy of overall experimental time. In principle, other mixing schemes for interresidue correlations can be implemented to obtain long-range distances, such as proton assisted recoupling (PAR)³² and CHHC.³³ The ¹H-detected experiments^34–35 of perdeuterated membrane proteins can also benefit significantly from the sensitivity enhancement of Cu²⁺ ions given the low yield of perdeuterated membrane protein expression. Therefore, this widely applicable technique of sensitivity enhancement will greatly facilitate chemical shift assignments and structure determinations of membrane proteins.

EXPERIMENTAL SECTION

Sample preparation

Uniformly-labeled ¹³C, ¹⁵N DsbB C41S was expressed in E. coli C43 (DE3) using modified Studier Medium P, containing 10 ml/L ¹³C, ¹⁵N Bioexpress (Cambridge Isotope Laboratories, Andover, MA) as described previously.²¹ Cell membranes were isolated, and the solubilized DsbB was purified on a Talon cobalt column.²¹ Fractions containing DsbB were pooled, concentrated, and dialyzed against 25 mM HEPES, pH 7.8. The His-tag was removed by incubation with Factor Xa overnight at 23 °C. DsbB was then centrifuged to remove precipitated contaminants and further concentrated. The concentration of dodecylmaltoside (DDM) was determined to be 2% using a colorimetric assay.³⁶ CuEDTA was added to 2.5 mM to half of the purified DsbB. To the other aliquot of DsbB, CuDMPE-DTPA (Avanti Polar Lipids, Inc., Alabaster, AL) in 5% DDM was added in an approximate ratio of 3 CuDMPE-DPTA:DsbB, resulting in a final concentration of 0.6 mM CuDMPE-DPTA. This solution was stirred 24 h at 23 °C to facilitate exchange of the CuDMPE-DPTA with the endogenous lipids bound to the solubilized DsbB. For both samples, addition of the chelated Cu²⁺ ions had no effect on the visible absorbance bands of DsbB C41S (Figure 6). The samples were then precipitated by removal of DDM over several days by addition of methyl-β-cyclodextrin and dialysis,²¹ and packed into 1.6 mm rotors for SSNMR. The UV-Vis spectra of DsbB(C41S) were acquired on an Evolution 600 Spectrophotometer (Thermo Scientific, Waltham, MA).

NMR spectroscopy

SSNMR experiments were performed on a 750 MHz Varian VNMRS spectrometer and a 500 MHz Varian VNMRS spectrometer (¹H frequency). All spectrometers were equipped with 1.6 mm fast MAS probes. Typical π/2 pulse widths were 1.5~1.7 μs for ¹H and 1.6 μs for ¹³C. The spinning speed was 36 kHz. All experiments utilized tangent ramped cross polarization (CP)³⁷ with low power ¹H SPINAL³⁸ decoupling (~15 kHz) applied during acquisition and evolution periods. ¹H T₁s were measured by ¹H 1D inversion recovery experiments. The ¹³C B₁ field for RFDR was 140 kHz and no ¹H decoupling was applied during RFDR. The ¹H RF field for R2₂¹ was 18 kHz as half of the spinning speed. Detailed parameters are given in the figure captions. Sample temperatures were determined by ethylene glycol calibration.³⁹ Chemical shifts were referenced to DSS, using adamantane as a secondary standard.⁴⁰ Spectra were processed with NMRPipe⁴¹ and peak picking and assignments were performed in Sparky (T.D. Goddard and D.G. Kneller, University of California, San Francisco). Back linear prediction and polynomial baseline correction were applied to the direct dimension. Zero filling and Lorentzian-to-Gaussian apodization were used for each dimensions before Fourier transformation.