Sensitivity Enhancement in ¹³C Solid-State NMR of Protein Microcrystals by Use of Paramagnetic Metal Ions for Optimizing ¹H T₁ Relaxation

Abstract

We discuss a simple approach to enhance sensitivity for ¹³C high-resolution solid-state NMR for proteins in microcrystals by reducing ¹H T₁ relaxation times with paramagnetic relaxation reagents. It was shown that ¹H T₁ values can be reduced from 0.4-0.8 s to 60-70 ms for ubiquitin and lysozyme in D₂O in the presence of 10 mM Cu(II)Na₂EDTA without substantial degradation of the resolution in ¹³C CPMAS spectra. Faster signal accumulation using the shorter ¹H T₁ attained by paramagnetic doping provided sensitivity enhancements of 1.4-2.9 for these proteins, reducing the experimental time for a given signal-to-noise ratio by a factor of 2.0-8.4. This approach presented here is likely to be applicable to various other proteins in order to enhance sensitivity in ¹³C high-resolution solid-state NMR spectroscopy.

Introduction

Over the past years, significant progress has been achieved in solid-state NMR (SSNMR) spectroscopy of biomolecules such as peptides and proteins.[1-16] Particularly, use of protein micro-/nano-crystals[17] has significantly improved resolution in high-resolution SSNMR of dilute spins such as ¹³C and ¹⁵N, permitting signal assignment and structural determination of various uniformly ¹³C and/or ¹⁵N-labeled proteins by SSNMR.[18-25] However, restricted sensitivity in ¹³C and ¹⁵N SSNMR has been still one of the major limiting factors in SSNMR analysis of proteins. In an experimental time required for ¹³C SSNMR of proteins, more than 95 % is typically consumed for recycle delays to retrieve spin polarization by ¹H T₁ relaxation, to protect a probe from arcing due to RF irradiation, or to avoid sample degradation due to heating. The latter two problems have been addressed by improving designs of MAS probes to minimize sample heating[26] and tolerate handling of relatively high RF power,[27] or by using low-power ¹H decoupling.[28; 29] Doping with paramagnetic metal ions such as Cu(II) has been utilized to reduce ¹H T₁ relaxation times in ¹³C CPMAS of proteins/peptides in cryogenic conditions.[30] However, the effects of paramagnetic doping on resolution and T₁ relaxation have not been fully examined for biomolecular SSNMR, in particular, for ¹³C CPMAS of protein micro-crystals, which generally provides excellent resolution.

In this study, we experimentally investigate the effects of paramagnetic ion doping to reduce ¹H T₁ values in 1D ¹³C CPMAS for microcrystals of two model proteins: ubiquitin and lysozyme using a Cu(II)Na₂EDTA complex (Cu-EDTA) as a relaxation reagent. In addition to ¹H T₁ values, we examine resolution and line positions in ¹³C CPMAS spectra for these proteins in the presence of Cu-EDTA. Motivations and prospects of this approach for sensitivity enhancements in biomolecular SSNMR are presented.

Materials and Methods

D₂O was purchased from Cambridge isotope (Andover, MA). Ubiquitin from bovin red blood cells (ubiquitin), lysozyme from chicken egg white (lysozyme), and other chemicals were purchased from Sigma-Aldrich (St. Louis, MO). Purified water (double deionized and distilled) was prepared using a High-Q 103 S water still system (High-Q Corp., Wilmette, IL). The purified water was used for preparation of all the protein microcrystals in H₂O.

We prepared protein microcrystals in D₂O or H₂O following the protocols by Martin and Zilm for preparing protein nanocrystals[17] with minor modifications. Unless otherwise mentioned, all the data were collected for samples prepared in D₂O for potential applications to partially deuterated proteins in D₂O. As will be discussed, protein microcrystals prepared in D₂O have longer ¹H T₁ values than those in H₂O. Thus, the effects of paramagnetic doping are more notable for proteins in D₂O, although this approach is also effective for samples prepared in H₂O. An equal volume mixture of a protein stock solution (25 mg protein/ml) and a crystallization solution was concentrated to approximately half the starting volume by using a SpeedVac concentrator (Savant, Farmingdale, NY).[17] For preparation of the protein stock solution, lysozyme was dissolved in a 100 mM sodium acetate buffer (pH 4.5) while ubiquitin was dissolved in pure D₂O/H₂O. The crystallization solution for ubiquitin contained 25% w/v PEG 8000 and 200 mM cadmium acetate in a 50 mM sodium HEPES buffer (pH 7); the solution for lysozyme contained 12.5% w/v PEG 2000 and 75 mM sodium chloride in a 100 mM sodium acetate buffer (pH 4.5). The concentrated protein solution of ∼0.5 mL was kept in a microtube at 4 °C for 10-12 hours to produce protein crystals. Then, the solution containing crystals was centrifuged at 1.5 × 10³ g for 5 min using an Eppendorf 5414D micro-centrifuge (Eppendorf, Westbury, NY).

A Cu-EDTA complex was selected as a relaxation reagent that minimizes undesired interactions between metal ions and proteins.[31] To prepare protein crystal samples containing Cu-EDTA, about 0.4 mL of the mother liquor was separated as a supernatant from the protein crystals after the centrifugation. Then, 1.0-14.9 mg of Cu-EDTA was dissolved in the mother liquor. This solution was kept at 4 °C for 2-4 hours, and centrifuged to remove any precipitated proteins due to the salts. After the pH was adjusted to an appropriate value for the particular protein, the mother liquor containing Cu-EDTA was reintroduced to the protein crystals, and left at 4 °C for another 10-12 hours to dope Cu-EDTA into protein crystals. The samples which do not contain Cu-EDTA were prepared in the same manner for a control, but without the addition of Cu-EDTA. Then, the sample was centrifuged for 5 min, and the collected protein crystals were packed into a MAS rotor by centrifugation. The concentration of Cu-EDTA was estimated from the amount of Cu-EDTA used and the total volume of the mother liquor and the protein microcrystals. Formation of the protein crystals were confirmed under an optical microscope. The images of the crystals were obtained at ×32 magnification using a CCD camera (CoolSnap, Roper, Trenton, NJ) attached to a Carl Zeiss Axiovert 25 inverted microscope (Carl Zeiss MicoImaging, Thornwood, NY).

SSNMR experiments were performed at 9.4 T (¹H NMR frequency of 400.2 MHz) with a Varian InfinityPlus 400 NMR spectrometer. For experiments at the spinning of 10 kHz, a Varian T3 3.2-mm MAS double-resonance NMR probe or a home-built 2.5-mm MAS double-resonance probe was used. The signals were collected during an acquisition period of 10 ms at the spinning speed of 10,000 ± 5 Hz with cooling air at −10 °C supplied through a Varian VT stack at a flow rate of ∼140 standard-cubic-feet per hour (scfh). For experiments at the spinning of 40 kHz, we used a 2.0-mm MAS double-resonance probe developed in Dr. Samoson's lab.[32-34] Unless otherwise mentioned, the signals were collected during an acquisition period of 20 ms at the spinning speed of 40,000 ± 10 Hz with cooling air at −5 °C supplied through the Varian VT stack at a flow rate of ∼140 scfh and cooled bearing air (1 °C). The data were processed with Varian Spinsight software. The spectra in Fig 2 were processed with Gaussian line broadening of 15 Hz; other spectra were processed with Gaussian line broadening of 25 Hz. ¹H T₁ values were calculated from the data collected by ¹H inversion recovery experiments detected by ¹³C CPMAS, where a π-pulse to ¹H spins and the following inversion recovery delay were added prior to the conventional CPMAS sequence with ramped CP[35] and TPPM decoupling[36]. The signal intensities were measured for the highest signals in the ¹³CO (160-190 ppm), ¹³Cα (40-65 ppm), ¹³CH (10-30 ppm) regions to estimate ¹H T₁; the average of the ¹H T₁ values estimated for the three regions was used as ¹H T₁ of the sample. For proteins without Cu-EDTA, some variations in ¹H T₁ values (∼10 %) were observed from batch to batch. We also noticed that ¹H T₁ is gradually reduced (10-25 %) over the course of experiments after a week for the proteins without Cu-EDTA. The values for ¹H T₁ in Table 1 were measured within a week after the sample was packed in a rotor.

Figure 2.

(a-d) ¹³C CPMAS spectra of protein micro crystals for (a, b) lysozyme and (c, d) ubiquitin prepared in D₂O obtained (a, c) with and (b, d) without 10 mM Cu-EDTA at ¹³C NMR frequency of 100.6 MHz. The recycle delays were (a) 0.18 s, (b) 1.50 s, (c) 0.18 s, and (d) 2.50 s. The spectra were acquired at a spinning speed of 40 kHz with π-pulse-train decoupling. In the decoupling scheme, a π-pulse with the width of 2.5 μs was rotor-synchronously applied at the end of every rotor cycle (25 μs) with the XY-8 phase cycle.[51] Signals were accumulated during acquisition periods of 20 ms with (a) 36,000, (b) 4,800, (c) 72,000, and (d) 5,680 scans in a common total experimental time of (a, b) 2 hours or (c, d) 4 hours. During the CP period of 1.0 ms, the ¹³C RF field was swept from 53 kHz to 71 kHz, while the ¹H RF field was kept at 102 kHz. All the spectra were processed with Gaussian line broadening of 15 Hz.

Table 1.

¹H T₁ values of ubiquitin and lysozyme in the presence and absence of Cu-EDTA with their approximate crystal sizes measured from microscopic images.

Protein	Crystal size (μm)	1H T1D2O (ms) a	1H T1H2O (ms) b	With Cu-EDTA in D2O
				[Cu-EDTA] (mM)	1H T1 (ms)c
Ubiquitin	1-5	820 ± 10	307 ± 19	10	73 ± 1
Lysozyme	5-20	350 ± 4	280 ± 5	10	59 ± 1

^a1H T₁ for the sample prepared in D₂O used for Fig. 1 (b, d). ¹H T₁ values were 500 and 820 ms for Fig. 2(b), (d), respectively.

^b1H T₁ for the sample prepared in H₂O.

^c1H T₁ for the sample used for Fig. 1 (a, c). ¹H T₁ values were 60 ms for both Fig. 2(a) and (c).

Results and Discussion

In Fig. 1(a-d), we demonstrate sensitivity enhancements in ¹³C CPMAS by faster signal accumulation using short ¹H T₁ optimized with Cu-EDTA doping for (a, b) lysozyme and (c, d) ubiquitin microcrystals. These spectra were collected (a, c) with 10 mM Cu-EDTA and (b, d) without Cu-EDTA in a common experimental time of 4 hours. Inspections of the microscopic images in Fig. 1(e-h) for the crystals (e, g) with and (f, h) without Cu-EDTA revealed that there were no noticeable changes in the crystal shapes or sizes due to the introduction of Cu-EDTA for (e, f) lysozyme and (g, h) ubiquitin. Hence, it is not likely that the addition of Cu-EDTA damaged the protein crystals. The recycle delays for the lysozyme samples were (a) 0.5 s (¹H T₁ = 59 ms) and (b) 1.1 s (¹H T₁ = 350 ms) while the delays for the ubiquitin samples were (c) 1.0 s (¹H T₁ = 73 ms) and (d) 2.5 s (¹H T₁ = 820 ms). As will be discussed later, increasing the Cu-EDTA concentration further reduces ¹H T₁ values. For comparison, the spectra were scaled so that the spectra for the same protein display a common noise level. Faster signal accumulation using the shorter ¹H T₁ in the presence of Cu-EDTA clearly yielded considerable sensitivity enhancements by factors of 1.4 and 1.6 in Fig. 1(a) and (c), respectively. It is also clear that no major changes in the line positions or line-widths were induced by the presence of 10 mM Cu-EDTA. Since line positions in ¹³C CPMAS spectra were not altered by the Cu-EDTA doping, major structural changes in the proteins due to Cu-EDTA doping are not likely. ¹H T₁ values were successfully reduced by a factor of 6-11. On the other hand, we were able to speed up the experiment by only up to 2.5 times to protect a probe from arcing under high-power ¹H RF decoupling. We also tested decoupling by fast MAS at 20-30 kHz, which was successfully employed in our recent studies for paramagnetic systems.[37-39] However, the resolution without ¹H RF decoupling is considerably limited for the diamagnetic poly-crystalline proteins.

Figure 1.

(a-d) ¹³C CPMAS spectra of protein crystals for (a, b) lysozyme and (c, d) ubiquitin prepared in D₂O obtained (a, c) with and (b, d) without 10mM Cu-EDTA at ¹³C NMR frequency of 100.6 MHz. The recycle delays were set to (a) 0.5 s, (b) 1.1 s, (c) 1.0 s, and (d) 2.5 s. The spectra were acquired at a spinning speed of 10 kHz under ¹H TPPM decoupling of (a, b) 71 kHz and (c, d) 90 kHz. Signals were accumulated during acquisition periods of 10 ms with (a) 28160, (b) 12800, (c) 14336, and (d) 5736 scans in a common total experimental time of 4 hours. During the CP period of 1.0 ms, ¹³C RF field was swept (a, b) from 56 kHz to 76 kHz and (c, d) from 52 kHz to 72 kHz, while ¹H RF field was kept at (a, b) 76 kHz and (c, d) 72 kHz. All the spectra were processed with Gaussian line broadening of 25 Hz. At the right of the spectra in (a-d), microscope images of the corresponding protein micro/nano-crystals used for the NMR experiments are displayed in (e-h). The images were obtained at 32× magnification.

To overcome this problem, we employed a rotor-synchronous π-pulse train decoupling under ultra-fast MAS condition (40 kHz), which was recently proposed by Hafner and coworkers at Varian.[40] We found that this decoupling permits narrowing comparable to TPPM decoupling of ¹H RF irradiation at 200 kHz under MAS at 40 kHz. An equivalent decoupling sequence was successfully applied for ¹H decoupling for paramagnetic systems[39] and ¹⁹F decoupling for fluoro-polymers[41] under fast MAS (≥ 20 kHz). Figure 2 shows CPMAS spectra at a spinning speed of 40 kHz for (a, b) lysozyme and (c, d) ubiquitin microcrystals (a, c) with and (b, d) without 10 mM Cu-EDTA doping. The recycle delays for the lysozyme samples were (a) 0.18 s (¹H T₁ = 60 ms) and (b) 1.5 s (¹H T₁ = 500 ms) while the delays for the ubiquitin samples were (c) 180 ms (¹H T₁ = 60 ms) and (d) 2.5 s (¹H T₁ = 820 ms). We confirmed that faster signal accumulation with the low-duty-facor ¹H decoupling sequence at a spinning speed of 40 kHz further enhanced sensitivity in the ¹³C CPMAS spectra by a factor of 2.7-2.9, which speeds up our experiments 7-8 fold. The resolution of the CPMAS spectra obtained with fast recycling in the presence of Cu-EDTA was comparable to that without Cu-EDTA. It is noteworthy that the effective RF duty factor due to ¹H decoupling is only ∼1.1 % in the experiment with a recycle delay of 180 ms and an acquisition period of 20 ms. Because of the low RF duty factor and high efficiency of the ¹H RF circuit of the fast MAS probe, we could collect the signals up to 30 ms of acquisition periods without any arcing problems. Thus, significant sensitivity enhancements with uncompromised resolution in this approach are possible under the fast MAS condition.

We prepared crystals in D₂O and H₂O without Cu-EDTA to examine solvent effects on ¹H T₁ values. Considerably longer ¹H T₁ was observed in D₂O for ubiquitin and, to a lesser degree, for lysozyme, as shown in Table 1. The sensitivity enhancements for the proteins in H₂O are less, particularly for ubiquitin. Nevertheless, considerable sensitivity gains (∼2) are still expected under the fast MAS and decoupling condition used for Fig. 2. Martin and Zilm reported that ¹H T₁ of unlabeled ubiquitin nanocrystals in H₂O is 0.5 s.[17] Zilm and coworkers more recently reported average ¹H T₁ values of 300-400 ms for uniformly ²D- and ¹⁵N-labeled ubiquitin nanocrystal samples for which amide protons are back-exchanged in H₂O, as well as for unlabeled ubiquitin.[42] Thus, the long ¹H T₁ observed for our ubiquitin sample in D₂O cannot be simply explained by lower ¹H density. Understanding this solvent dependence of ¹H T₁ requires more systematic work, and a fuller study is outside of the scope of this study.

Figure 3 shows ¹³C CPMAS spectra of ubiquitin microcrystals acquired at different ¹H inversion recovery delays (a) without and (b) with 10 mM Cu-EDTA. This result clearly shows that the ¹H T₁ value of lysozyme is reduced by the introduction of Cu-EDTA in a uniform manner for different chemical groups and residues. After 50 ms, the ¹³C CPMAS spectra in (b) display a null signal in the presence of Cu-EDTA, while the spectra in (a) show a signal close to the null only after 800 ms without Cu-EDTA. ¹H T₁ values obtained from the experiments are (a) 830 ms and (b) 73 ms. In a recent study on uniformly ¹³C labeled ubiquitin by Igumenova et. al., it has been reported that a 2D ¹³C/¹³C chemical-shift correlation experiment at ¹H frequency of 800 MHz required about 36 hours with recycle delay of 1.5 s.[43] Assuming that the RF-duty factor and the sample stability are not the limiting factor, it is possible to speed up this experiment up to seven times with a shorter ¹H T₁ value of ∼70 ms in the paramagnetic doping approach.

Figure 3.

¹H inversion recovery delay (τ_IR) dependence of ¹³C CPMAS spectra of ubiquitin microcrystals prepared in D₂O (a) without and (b) with 10 mM Cu-EDTA obtained at 10 kHz MAS. The values of τ_IR used in the experiments are indicated in the figure.

Figure 4 shows Cu-EDTA concentration dependence of the longitudinal relaxation rate 1/T₁ for the lysozyme sample. Figure 4 clearly demonstrates that 1/T₁ is not linearly proportional to the concentration of Cu-EDTA. The slope is the largest at the lower Cu-EDTA concentration (5-10 mM). At higher concentration, the relaxation rate increases more slowly for a given increase in the Cu-EDTA concentration. This is probably because ¹H T₁ relaxation due to the paramagnetic reagents is mediated by a ¹H-¹H spin diffusion mechanism.[44] Since Cu-EDTA is hydrophilic, it is most likely that ¹H polarization is retrieved more quickly by paramagnetic T₁ relaxation around the water-accessible protein surface, where Cu-EDTA is easily accessible. The recovered polarization can be transferred across the molecule by ¹H-¹H spin diffusion. We noticed that there is minor solvent dependence of ¹H T₁ even in the presence of Cu-EDTA. Slightly lower ¹H 1/T₁ rates in D₂O may be attributed to slower ¹H-¹H spin diffusion in D₂O, in which amide hydrogens are exchanged for ²D. In spite of the difference, we found that in both H₂O and D₂O solvents, ¹H T₁ for the protein microcrystal samples can be reduced to 20-30 ms in the presence of 75 mM Cu-EDTA.

Figure 4.

Cu-EDTA concentration dependence of the ¹H relaxation rate (1/T₁) for lysozyme microcrystals prepared in D₂O (triangle) and H₂O (square). The ¹H T₁ values were measured by the inversion recovery experiment shown in Fig. 3. The error bars indicate standard errors in the measurements of 1/T₁.

In Fig. 5, we examined resolution of ¹³C CPMAS spectra for (a, b) lysozyme and (d, e) ubiquitin microcrystals (a, d) without and (b, e) with 10 mM Cu-EDTA, where only the aliphatic region of the spectra are displayed for clarity. The spectra in (c, f) show the difference between the spectra without and with Cu-EDTA for (c) lysozyme and (f) ubiquitin. As discussed above, in (b, e), no major changes were observed by the addition of 10 mM Cu-EDTA. Although ¹H T₁ values were reduced by factors of (b) 8.3 and (e) 13.7, the line broadening is only subtle (10-20 %). However, we observed that a few resonances observed in Fig. 5(a, d) are substantially reduced in intensity in the spectra in (b, e) (indicated by arrows). The quenched signals may be assigned to residues exposed to the protein surface, at which ¹³C spins can be subject to much faster paramagnetic T₂ relaxation due to Cu-EDTA in water phase. It is known that paramagnetic ¹³C T₂ relaxation rates are proportional to 1/R⁶, where R is the distance between a ¹³C spin and a paramagnetic ion.[45; 46] Hong et al. reported that the paramagnetic quenching can be utilized to measure distances of ¹³C sites in membrane bound peptides from the membrane surface with ranging Mn(II) concentration in water phase.[46] Thus, it is probable that signals for residues exposed to the protein surface are selectively quenched by the paramagnetic effects, while the majority of signals for other residues are unaffected. Further studies are needed to examine the possibilities of using the effects for structural analysis. The difference spectra in (c, f) also suggest that the effects of paramagnetic quenching are relatively minor; the integral intensities of (c) and (f) are only 2 and 11 % of those for (a) and (d), respectively. Considering that the amount of the proteins may differ by 5-10 % between (d) and (e) (or (a) and (b)), the overall signal quenching due to the Cu(II) addition is less than the significant level. We also found that the resolution in ¹³C CPMAS spectrum of lysozyme was not substantially degraded by the addition of 75 mM Cu-EDTA (data not shown). The ¹H T₁ value in this condition is about 27 ms, which is one tenth of the ¹H T₁ for the same protein without Cu-EDTA. Therefore, for the majority of the signals, further sensitivity enhancement is possible with more optimized pulse sequences for this purpose.

Figure 5.

¹³C CPMAS spectra of (a, b) lysozyme and (d, e) ubiquitin protein crystals with Cu-EDTA concentrations of (a, d) 0 mM, (b, e) 10 mM. The spectra in (c, f) are the difference between the spectra obtained with and without Cu-EDTA for the same sample. The spectra were obtained at ¹³C NMR frequency of 100.6 MHz at a spinning speed of 40 kHz with ¹H π pulse-train decoupling during the acquisition periods of 10 ms. The shorter acquisition periods were adopted to minimize the noise level for comparison. The recycle delays were (a) 1.50 s, (b) 0.18 s, (d) 2.50 s, and (e) 0.18 s. All the spectra were processed with Gaussian line broadening of 25 Hz.

Conclusion

In this study, we demonstrated that ¹H T₁ values of the two model proteins, lysozyme and ubiquitin, in microcrystals can be reduced to ∼60 ms by Cu-EDTA doping without major degradation in the resolution of their ¹³C CPMAS spectra. We also demonstrated that significant sensitivity enhancements of 1D ¹³C CPMAS spectra were attained for these proteins by faster signal repetitions using the reduced ¹H T₁ values under fast MAS. Although the full potential of this approach for sensitivity enhancement is still restricted only in the fast MAS condition, our study presented a new opportunity to gain significant sensitivity enhancements using paramagnetic doping in biomolecular SSNMR. In this communication, we focused on testing this approach in 1D ¹³C CPMAS for unlabeled ubiquitin and lysozyme microcrystals. It is probable that the present approach can be adopted in CPMAS or static experiments of other insoluble proteins. The successful reduction of ¹H T₁ relaxation times in the present experiments will also open new possibilities of sensitivity enhancements in more advanced experiments such as multi-dimensional ¹³C SSNMR of uniformly ¹³C-labeled proteins[18-25] and various distance measurements[3; 11; 13; 47-50] for selectively ¹³C-labeled protein samples in our future studies.