Improving the Quality of Oriented Membrane Protein Spectra Using Heat-Compensated Separated Local Field Experiments

Abstract

Oriented sample solid-state NMR (OS-ssNMR) spectroscopy is a powerful technique to determine the topology of membrane proteins in oriented lipid bilayers. Separated local field (SLF) experiments are central to this technique as they provide first-order orientational restraints, i.e., dipolar couplings and anisotropic chemical shifts. Despite the use of low-E (or E-free) probes, the heat generated during the execution of 2D and 3D SLF pulse sequences causes sizeable line-shape distortions. Here, we propose a new heat-compensated SE-SAMPI4 (hcSE-SAMPI4) pulse sequence that holds the temperature constant for the duration of the experiment. This modification of the SE-SAMPI4 results in sharper and more intense resonances without line-shape distortions. The spectral improvements are even more apparent when paramagnetic relaxation agents are used to speed up data collection. We tested the hcSE-SAMPI4 pulse sequence on a single-span membrane protein, sarcolipin (SLN), reconstituted in magnetically aligned lipid bicelles. In addition to eliminating peak distortions, the hcSE-SAMPI4 experiment increased the average signal-to-noise ratio by 20% with respect to the original SE-SAMPI4.

INTRODUCTION

Solid-state NMR (ssNMR) spectroscopy has emerged as a powerful technique to characterize the structures, motions, and interactions of proteins ^1–5. This technique provides atomic-resolution information on amyloid fibrils, microcrystalline proteins, and membrane proteins reconstituted in lipids at the liquid-crystalline state. For membrane proteins, the synergistic use of magic angle spinning (MAS) and oriented sample (OS) ssNMR enables to obtain both structure and topology of membrane proteins ^6–9. While MAS-ssNMR provides isotropic chemical shifts, distances, and torsion angles ^10–13; OS-ssNMR offers topological information through the measurements of anisotropic parameters such as chemical shifts and dipolar couplings ^14–17. A few high-resolution structures have been obtained using a combination of these techniques ^18–20. However, both approaches suffer from low sensitivity and resolution, which hamper their routine application and throughput. One way to overcome the low sensitivity issue is the use of paramagnetic agents to accelerate data collection for multidimensional experiments ²¹. The unpaired electrons of the paramagnetic centers enhance the longitudinal relaxation times (T₁) of nuclei, making it possible to shorten the recycle delays between the RF transients and reducing the overall experimental time. Among the different paramagnetic agents, Cu²⁺- and Gd³⁺-chelated lipids have been employed in MAS-ssNMR to speed up data collection for microcrystalline, fibrillar as well as membrane proteins^22–26. More recently, Ramamoorthy and co-workers used Cu²⁺-chelated lipids to speed up the acquisition of 1D spectra of membrane proteins reconstituted in bicelles ^26,27, though these authors warned about the RF heating problem. Following this work, we recently obtained very promising results by doping phospholipid bicelles with 5% Cu²⁺-chelated lipids that enabled fast collection of 2D and 3D OS-ssNMR spectra ²⁸. However, for these multidimensional experiments short recycle delays together with high-power decoupling caused significant RF heating in spite of the use of low-E ²⁹ or E-free ³⁰ probes. Therefore, we were forced to use longer recycling delays to prevent line-shape distortions and sample deterioration.

To reduce the temperature variations during the execution of the experiments, we develop a heat-compensated version of SLF experiments and implement in SAMPI4 (selective averaging via magic sandwich pulses using π/4 flip pulses) experiment originally introduced by Nevzorov and Opella³¹ and later improved by our group with the sensitivity enhancement (SE) scheme³². We tested the new heat compensated SE-SAMPI4 (hcSE-SAMPI4) pulse sequence on sarcolipin (SLN) embedded in anisotropic lipid bicelles (Figure 2a). SLN is a 31-residue single-span polypeptide that regulates the activity of the sarcoplasmic reticulum Ca²⁺-ATPase (SERCA) ^33–36. By comparing the performances of SE-SAMPI4 and hcSE-SAMPI4 experiments, we found a significant reduction of line-shape distortions on the 2D spectra of SLN and an average increase of S/N ratio of 20%, with resonance specific enhancement up to 40%.

Figure 2. Effects of RF heating on the SLN resonances.

a) Amino acid sequence of SLN and its secondary structure. b) Temperature effect on the SE-SAMPI4 spectra of SLN reconstituted in magnetically aligned bicelles. The blue, green, and red SE-SAMPI4 spectra were acquired at 20 °C, 27 °C and 35 °C respectively. The arrows indicate the direction of the resonance changes toward the isotropic chemical shift values of the amide groups as well as the direction of the low dipolar couplings as the temperature increases.

MATERIALS AND METHODS

Recombinant [U-¹⁵N] SLN was produced as previously described ³⁷. For the SLN samples reconstituted in bicelles with Cu²⁺-chelated lipids, we prepared two different preparations. For the first sample, we used CHCl₃ to solubilize 29.0 mg 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC) and 8.2 mg 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC); while for the second sample, we used 6.1 mg 1,2-dihexanoyl-sn-glycero-3-phosphocholine (DHPC) and 2.4 mg 1,2-dimyristoyl-sn-glycero-3-phosphoethanolamine-N-diethylenetriaminepentaacetic acid copper salt (Cu²⁺-DMPE-DTPA). To eliminate any trace of organic solvents, we dried all lipids under a flux of N₂ gas and lyophilized for at least 2 hours. We dissolved DMPC and POPC in Octaethylene glycol monododecyl ether (C₁₂E₈) (detergent/lipid ratio of 1:1 w/w) to form a clear suspension of mixed micelles. We then solubilized SLN (1.5 mg) separately with 1% C₁₂E₈ solution and then added to the mixed micelle preparation. We adjusted the sample volume to 20 mL using the NMR buffer consisting of 20 mM HEPES, 100 mM KCl, 5.0 mM MgCl₂, 0.25 mM DDT, 0.02% NaN₃, 2.5% Glycerol. We incubated the mixture containing SLN, long-chain lipids, and C₁₂E₈ with Biobeads® (1:20 w/w ratio for detergent/Biobeads®) for 3 hours to completely remove the detergent. The cloudiness of the suspension confirmed the formation of the DMPC/POPC vesicles containing SLN. After Biobeads® removal, we adjusted the pH of the sample to 7.0 and centrifuged the DMPC/POPC vesicles containing SLN at a speed of 50,000 g for 50 min. The DHPC/ Cu²⁺-DMPE-DTPA lipids were dissolved in 100 μL of NMR buffer and pH to 7.0, then added to the vesicles. The protein/lipid mixture was vortexed for a few minutes at 4°C to ensure the suspension was homogenous. The bicelles were formed by carrying out 3–5 freezing/thawing cycles. Finally, the sample was concentrated to a volume of 150 μL prior to loading it into a 5 mm flat-bottom glass tube for spectroscopy.

All the NMR experiments were performed on Varian VNMRS spectrometer operating at a ¹H frequency of 600 MHz equipped with low-E static bicelle probe built by the RF program at the National High Magnetic Field Laboratory (NHMFL) in Florida²⁹. The VT cooling gas rate was set as 20 L/min. The accuracy of VT gas temperature was confirmed using CH₃OH ³⁸. For SE-SAMPI4 experiment shown in Figure 3b ³², ¹⁵N spin polarization was prepared with adiabatic cross polarization (CP) using a linear ramped pulse at the ¹⁵N channel and a rectangular pulse at the ¹H channel. The ¹⁵N RF field strength was swept from 38 kHz to 62 kHz with the average RF field at 50 kHz; while the ¹H RF field amplitude was kept constant at 50 kHz. The CP contact time was 750 µs. During the t₁ evolution, the homonuclear ¹H decoupling was achieved by applying 62.5 kHz magic sandwich pulses on ¹H and simultaneous phase switched spin lock pulses on ¹⁵N with RF amplitude set to 62.5 kHz, matching the Hartmann-Hahn condition. The spin-lock pulse durations were 14 µs and 24 µs for ¹H and ¹⁵N channels, respectively. For the ¹H channel, two 4 µs 90° pulses were applied between two spin-lock pulses with a 12 µs interval. The dwell time of the t₁ acquisition was 96 µs. After the t₁ acquisition, two 90° pulses were applied on ¹H and ¹⁵N channels simultaneously, followed by a 35.3° pulse on the ¹H channel only. For the sensitivity enhancement (SE) sequence, a 75 µs τ delay of the SE scheme was used, corresponding to two cycles of phase-modulated Lee-Goldberg (PMLG) with effective RF amplitude of 80 kHz on the ¹H channel. After the τ delay, a 180° inversion pulse was used on the ¹⁵N channel followed by another τ delay period with a 50 kHz SPINAL decoupling on the ¹H channel. A final 90° pulse was applied on the ¹⁵N channel to prepare the ¹⁵N polarization for the t₂ detection. The t₂ acquisition was set to 10 ms for all the NMR experiments. For the hcSE-SAMPI4 experiment, an extra heat-compensation block was added after the t₂ acquisition, which mirrors the homonuclear decoupling applied during the t₁ acquisition period. Note that the length of the heat-compensation scheme is complementary with the length of the t₁ acquisition to ensure that their sum gives a constant time, which equals to the maximum t₁ acquisition time.

Figure 3. Implementation of the heat compensations in the SAMPI4 experiment.

a) SAMPI4, b) SE-SAMPI4 and c) hcSE-SAMPI4 pulse sequences. The color-shaded units represent the different blocks in the pulse sequences. The black rectangle indicates 90° pulses, the grey rectangle indicates 180° pulse, the yellow rectangle indicates CP with constant amplitude, the yellow trapezoid indicates CP with linear ramp, the blue rectangle indicates continuous-wave pulse for homonuclear decoupling during t₁ acquisition; the red rectangle indicates 35.3° pulse, the green rectangle indicates Lee-Goldberg decoupling; the purple rectangle indicates SPINAL decoupling during the t₂ acquisition. Single phase is used for most of the pulses. For the remaining pulses the following phase cycles were implemented: ϕ₁ = y, −y, y, −y; ϕ₂ = x, x, −x, −x; ϕ_rec = y, −y, y, −y. The heat-compensation (blue-shaded) of hcSE-SAMPI4 after the t₂ acquisition is implemented as a constant time.

RESULTS

To quantitate the extent of sample heating, we first performed a calibration curve, monitoring the ¹H chemical shift of H₂O in bicelle samples as a function of the temperature using a one-pulse experiment (Figure 1a) ^39,40. Then, we acquired two different ¹H one-pulse spectra of H₂O in a SLN bicelle sample immediately before and after performing dummy scans with the SE-SAMPI4 pulse sequence. The resulting H₂O chemical shift difference between two ¹H spectra indicates the rise in sample temperature caused by the SE-SAMPI4 pulse sequence. Subsequently, we estimated the temperature variations caused by RF heating by projecting the measured ¹H chemical shift of H₂O on the calibration curve. These experiments were repeated for different recycle delays, keeping the t₁ acquisition period at 3.1 ms and the t₂ acquisition time at 10 ms. These parameters are typically used for SLF experiments ²⁸. As shown in Figure 1b, the actual temperature of the sample increased during the execution of the SE-SAMPI4 experiment and reached a thermal equilibrium after 2 min, which corresponds to a total number of scans ranging from 40 to 80. The temperature changes were 9.6, 7.3, 6.1, and 5.2 °C for recycle delays of 1.5, 2.0, 2.5, and 3.0 s, respectively, confirming that the heat generation is rather significant in spite the use of low-E probes²⁹. Surprisingly, RF heating is quite significant even for longer recycle delays.

Figure 1. Quantification of sample heating by SLF experiments.

a) Calibration curve using H₂O in the bicelle sample at different temperatures. The ¹H chemical shift of H₂O is measured using a single pulse experiment. b) Temperature increase caused by SE-SAMPI4 experiment measured at different recycle delays. The red, orange, green, and blue colors represent the data points acquired with 1.5 s, 2.0 s, 2.5 s, and 3.0 s recycle delays. For these experiments, 3.1 and 10 ms acquisition times for t₁ and t₂ were used. The VT gas temperature was set as 25 °C. c) Temperature change caused by homonuclear decoupling during the t₁ acquisition period for the SE-SAMPI4 experiment. The red and green colors used for circles and dashed lines represent the experimental data and the fitting curve for SE-SAMPI4 experiments acquired with 1.5 and 3.0 recycle delays, respectively. The VT gas temperature was set as 25 °C.

During t₁ dipolar coupling evolution of the SE-SAMPI4 experiment, ¹H and ¹⁵N RF amplitudes are set to 62.5 kHz that corresponds to a dwell time of 64 µs, and the number of t₁ increments is typically set between 25 and 40. To determine the heat generation during the variable t₁ increments of SE-SAMPI4 pulse sequence, we performed a series of SE-SAMPI4 experiments with different t₁ evolution periods. Each SE-SAMPI4 experiment was executed in between two ¹H one-pulse experiments for 5 min to allow the sample to reach a temperature plateau. The t₂ acquisition time was set to 10 ms and two recycle delays, 3 and 1.5 s, were tested. When the acquisition time was set to zero, the temperature increase was significantly lower, i.e., 1.2 and 2.7 °C for 3 and 1.5 s recycle delays, respectively (Figure 1c). On the other hand, a dramatic temperature increase was detected by lengthening the t₁ acquisition period for both recycle delays, supporting that the homonuclear decoupling during the t₁ acquisition is the major source of RF heating. Although the t₁ is much shorter than the t₂ acquisition time, the simultaneous high-amplitude irradiation on both ¹H and ¹⁵N channels generates significant RF heating. These results also reveal that the temperature during the course of the SE-SAMPI4 experiment is not constant but increases with t₁ increments. Since the degree of alignment for bicelles samples is highly sensitive to temperature changes^41–43, these temperature variations affect both sample stability and spectral quality. To determine the latter effects, we acquired a series of SE-SAMPI4 spectra for SLN in DMPC/POPC/DHPC bicelles at 20, 27, and 35 °C. To minimize the effects of the temperature fluctuation, the recycle delay was set to 5 s. As shown in Figure 2b, the spectra of SLN display the typical PISA wheel observed both in bicellar preparations and mechanically aligned plates ^44,45. However, the amide resonances of the fingerprint shifted toward the isotropic values of both ¹⁵N chemical shifts and ¹⁵N-¹H dipolar couplings, indicating that as the temperature increases the bicelle/protein complex undergoes faster rotational diffusion^46,47. Therefore, similar spectral changes are expected for RF heating.

To overcome these temperature variations, we implemented a heat compensation element in the original SE-SAMPI4 experiment (hcSE-SAMPI4) by adding a t₁ pulse sequence block of length (T-t₁) after t₂ acquisition period. As the time of t₁ evolution increases, the (T-t₁) period decreases accordingly and keeps the total time to constant ‘T’ for all t₁ increments (Figure 3c) ⁴⁸. This constant-time element holds the temperature constant throughout the course of the experiment, while making the irradiation time the same for the different t₁ increments. Wang and Bax used a similar strategy to minimize the effects of RF heating in multidimensional solution NMR experiments ⁴⁸. In practice, it is not necessary to build the entire calibration curve to set up the experiment at a specific temperature. For our experiments, we fixed the variable temperature (VT) unit at the target temperature and acquired a 1D ¹H spectrum of H₂O as a reference. Then, we decreased the temperature of the VT unit by a few degrees and executed several dummy scans for approximately 5 min using the hcSE-SAMPI4 pulse sequence. Immediately after, we acquired a second ¹H spectrum and compared the ¹H chemical shift with the reference spectrum and adjusted the VT temperature in a stepwise manner until we matched the values of the reference H₂O chemical shift. Figure 4 shows the comparison of the SE-SAMPI4 and hcSE-SAMPI4 spectra of SLN recorded with a t₁ acquisition period of 3.1 ms and a recycle delay of 1.5 s. Figure 4a shows the overall SE-SAMPI4 spectrum is affected by the temperature variations. Due to the combination of the chemical shifts and dipolar coupling changes, several peaks more susceptible to temperature changes resemble asymmetric ellipses. This effect is accentuated for the most flexible residues of SLN. For instance, we observed the clear splitting of the resonances corresponding to Leu8 and Ser28 located at the N- and C-terminus portions of the polypeptide, respectively. In contrast, the corresponding hcSE-SAMPI4 spectrum acquired with identical parameters did not show any distortion of the resonance line-shapes (Figure 4b). Importantly, we observed an average increase of 20% in the S/N ratio, with up to 40% increase for the most dynamic residues (Figure 4d).

Figure 4. Comparison of SAMPI4 experiments with and without heat-compensation.

a) SE-SAMPI4 spectrum, and b) hcSE-SAMPI4 spectrum of SLN acquired at 20 °C and 17 °C respectively with 1.5 s recycle delay. The actual sample temperature changes from 22 °C to 30 °C for the SE-SAMPI4 experiment as the number of t₁ point increases. In contrast, the sample temperature is constant at 26 °C during the hcSE-SAMPI4 experiment. The arrows in the SE-SAMPI4 spectrum indicate the most affected residues. c) Overlap of the 1D projections of the 2D SE-SAMPI4 (blue) and 2D hcSE-SAMPI4 (red) spectra to visualize the overall improvement of the sensitivity. d) Comparison of signal-to-noise (S/N) ratios of SE-SAMPI4 (blue bars) and hcSE-SAMPI4 (red bars) spectra for each residue. The S/N ratios for the overlapping peaks in the 2D spectra are not shown. The average S/N ratios for SE-SAMPI4 and hcSE-SAMPI4 spectra are 3.9 and 4.7, respectively.

DISCUSSION

While for ultra-fast MAS it is possible to use low RF power (~10 kHz) for heteronuclear dipolar decoupling ⁴⁹, MAS-ssNMR at moderate speeds and OS-ssNMR experiments still require high-power decoupling in order to achieve line narrowing. However, high-power decoupling gives rise to significant RF heating in spite of the use of E-free or Low-E probes. For OS-ssNMR, the RF heating effect is strongly amplified due to the simultaneous irradiation on both ¹H and ¹⁵N channels during t₁ homonuclear decoupling. Therefore, it is necessary to lengthen the recycle delays for RF heat dissipation up to 3–5 s. Nonetheless, the RF heating is not homogeneous throughout the pulse sequences and these temperature variations cause significant distortions in the line-shapes. This is especially true for membrane proteins reconstituted in lipid bicelles. Notably, the CP scheme, which also irradiates on both ¹H and ¹⁵N channels simultaneously, is another source of heat generation in the pulse sequence ⁵⁰. However, it does not introduce any temperature variation because the same amplitude and duration are used for all t₁ increments. Low power PISEMA experiments (PITANSEMA) developed by Ramamoorthy and co-workers can be a good alternative for studying liquid crystals with dramatic heat reduction ⁵¹. However, these pulse sequences lead to poor dipolar line widths for membrane proteins. The RF heating problem is exacerbated when bicelle preparations of membrane proteins are doped with Cu²⁺-chelated lipids that are used to speed up the longitudinal relaxation times and reduce the total experimental time. In fact, Cu²⁺-chelated lipids increase the ionic strength and contribute to heat generation as reported by Ramamoorthy and co-workers²⁶. Because of the thermotropic nature of the bicelles ^52,53, their magnetic alignments and dynamics are significantly influenced by temperature changes, resulting in distorted line-shapes that reflect chemical shift and dipolar coupling variations during the acquisition of the SLF experiment. Heat-compensation schemes have been widely used in solution NMR and constitute an efficient method to eliminate the thermal inhomogeneity particularly for triple-resonance and Carr-Purcell-Meiboom-Gill (CPMG) experiments ^48,54. The extension of heat-compensation elements to the OS-ssNMR pulse sequences achieves similar effects and provides a reliable temperature control. In our hands, this implementation is critical to obtain higher quality and more sensitive SLF spectra of membrane proteins without resonance distortions. The latter made it possible to gain an average S/N ratio of 20%, with an enhancement for individual resonances up to 40%. Although the sensitivity improvement is expected to be less for OS-ssNMR experiments with shorter t₁ acquisition time or longer recycle delays, the heat-compensated SLF are still recommended for sample stability. Furthermore, the temperature variations are sample-dependent as the magnitude of the heat generated depends on the experimental factors such as salt, concentration of lipids, concentration of the protein, etc., thus we suggest to perform a calibration for each bicelle sample to estimate the need and extent of the temperature corrections to use.

CONCLUSIONS

In conclusion, we propose a heat-compensated version of the SE-SAMPI4 pulse sequence to mitigate temperature fluctuations during the execution of the OS-ssNMR experiment. The new experiment eliminates lineshape distortions and improves the sensitivity of the chemical shift anisotropy/dipolar coupling correlated spectra. Similar modifications can be easily implemented for all the SLF experiments such as polarization inversion spin exchange at magic angle (PISEMA) and heteronuclear isotropic mixing separated-local-field (HIMSELF) experiments, as well as 3D SE-SLF-HETCOR and SE-SLF-PDSD. The combination of PRE-assisted fast collection of NMR data with heat-compensated SLF experiment not only speeds up the acquisition of 2D and 3D OS-ssNMR experiments, but also provides a constant temperature environment thereby improving sample stability and spectral quality.