Ultrahigh Resolution in Proton Solid‐State NMR Spectroscopy at High Levels of Deuteration

Abstract

Slim peaks: Using a perdeuterated protein recrystallized from a 10:90 H2O:D2O mixture in magic-angle spinning (MAS) solid-state NMR spectroscopy experiments gives small 1H line widths at moderate spinning frequencies without application of homonuclear decoupling. This labeling strategy opens new perspectives for assignment of large protein spin systems. Structure investigations of biological solids by high- resolution magic-angle spinning (MAS) solid-state NMR spectroscopy has rapidly progressed in the last few years and resulted in complete structure elucidation of several peptides and small proteins.1–4 Successful spectral assignment and determination of structural constraints in isotopically enriched materials (mostly 13C, 15N) is, however, still limited by resolution and sensitivity. A gain in sensitivity in solid-state NMR (ssNMR) experiments can in principle be achieved using direct proton detection. This technique makes use of the high gyromagnetic ratio γ of protons, a property which however, leads to broad resonance lines. Several approaches have been suggested to achieve line narrowing. Application of windowed homonuclear decoupling schemes5, 6 yield a rescaled 1H line width on the order of 140–400 Hz, but require large receiver bandwidths, which allows radio-frequency (RF) noise to fold into the spectral region which finally compromises overall sensitivity. In addition, the applied pulse sequences scale the 1H chemical shift. In recent years, high-speed (35–60 kHz) MAS instrumentation has become available.7–9 However, even at these high spinning rates, fully protonated samples still have homogeneously broadened lines (>500 Hz). Alternatively, 1H line narrowing could be achieved by isotopic spin dilution at moderate (10–20 kHz) MAS frequencies.10–14 Dilution is achieved by perdeuteration of the sample and subsequent back-exchange of deuterons by protons. In these experiments, the 1H line width of most of the resonances is typically on the order 150–250 Hz or 80–150 Hz in the absence and in the presence of homonuclear 1H,1H decoupling, respectively. This labeling strategy allows, in addition the determination of long-range HN–HN distances,12, 15, 16 detection of dynamic water molecules in the protein structure16, 17 and the characterization of protein side-chain dynamics.18, 19 Herein, we demonstrate that a further increase in the degree of deuteration by using a 10:90 H2O:D2O mixture for recrystallization results in significant narrowing of the proton line width without loss in sensitivity. A 1H line width on the order of 17–35 Hz can be achieved at moderate spinning frequencies (8–24 kHz) without application of homonuclear decoupling. The experiments are carried out using a perdeuterated, 15N-enriched microcrystalline sample of the SH3 domain from chicken α-spectrin. To our knowledge, this strategy yields the most dispersed 1H correlation spectra in the solid state reported to date. Figure 1 represents a comparison of the 2D-1H,15N correlation spectra for the reference sample (Figure 1 B, D1) and the HN dilute sample (Figure 1 A, C1). The 1H-detected experiments (Figure 1 A at 400 MHz, Figure 1 C at 600 MHz) and the 15N-detected experiments (Figure 1 B at 400 MHz, Figure 1 D at 600 MHz) were acquired using the NMR pulse sequence shown in Figure 4 A and Figure 4 B, respectively (see also the Experimental Section). The spinning frequency for the 1H experiments was set to 13 kHz and 10 kHz for the 15N experiments. For clarity, Figure 1 B and D1 are displayed such that the proton dimension appears on the horizontal axis. The proton line width is not sensitive to moderate changes in the spinning rate under homonuclear dipolar decoupling,6, 20 if broadening conditions which arise from interference between MAS and the periodicity of the decoupling sequence are avoided.20 Clearly, the resolution in the proton dimension is clearly improved for the HN dilute sample yielding a fully resolved 15N–1H correlation spectrum of the protein even at 400 MHz.Figure 2 shows the 1H line-width dependence as a function of the rotor period for selected residues. We find that the line width is inversely proportional to the spinning rate. This result is in agreement with previous studies which show that in the fast-spinning regime the residual dipolar line width depends linearly on the rotational frequency.7, 21 In an investigation of protonated alanine which was embedded in a deuterated alanine matrix, Rienstra and co-workers found that the slope of the MAS-dependent 1H line width depends only on the average proton concentration in the sample.22 The slope varies from 6728 Hz ms−1 to 970 Hz ms−1 depending on the degree of protonation. We observe in our studies a slope of 80.8 Hz ms−1 and of 144 Hz ms−1 for G51HN and A56HN, respectively. In spin systems that behave purely inhomogeneous (based on the criteria of Maricq and Waugh23), the line width should not depend on the MAS rotational frequency, if ωr is comparable or larger than the size of the interaction. Non-zero slopes of the spinning-frequency dependence of the line width indicate therefore that 1H,1H dipolar couplings are not totally suppressed despite the high level of deuteration. The fitted lines in Figure 2 have non-vanishing y-intercepts, implying that line broadening cannot be removed even at infinite MAS rates. The inherent transverse relaxation time T2, sample heterogeneity, static-field inhomogenity (ca. 7 Hz, based on the 1H line width of a water sample which was used for shimming the probe) might be responsible for this observation. Given the observed resolution, major hardware improvements, especially in view of the 2H lock system will be required in the future. We attribute differences in the line widths, y intercepts, and slopes for different residues to site-to-site variations in the local proton density and to local backbone dynamics.24, 25 A,C) 1H-detected 2D 15N–1H correlation spectra of the HN dilute sample (deuterated SH3, 10 % 1H at labile proton positions) recorded at 400 MHz (A) and 600 MHz (C). (64 scans per increment; t1max(15N)=26.4 ms; t2max(1H)=100.0 ms; total experimental time=3.8 h.) B,D) 15N-detected 2D 15N–1H correlation experiments for the reference sample (deuterated SH3, 100 % 1H at labile proton positions) recorded at 400 MHz (B) 600 MHz (D). (8 scans per increment; t1max(1H)=17.2 ms; t2max(15N)=37.0 ms; total experimental time=0.6 h). Acquisition of more increments in the indirect 1H dimension did not result in a higher resolution in the 1H dimension. All spectra were apodized using a 5 Hz lorentzian broadening in both dimensions. Dependence of the 1H line width on the inverse MAS spinning frequency for selected residues in perdeuterated HN dilute SH3. While the 1H line width for the HN dilute sample is on the order of 17–35 Hz, the effective line width of the reference sample under phase-modulated Lee–Goldberg (PMLG) conditions amounts to 80–150 Hz taking chemical shift scaling resulting from PMLG into account. The resolution in the proton dimension is therefore improved by a factor of 4–5. The 15N line width in both experiments is on the order of 20–30 Hz, and is limited by the acquisition time employed. In case of the 15N detected experiment, two pulsed phase modulation (TPPM) decoupling is applied during 30–37 ms in each scan. This irradiation induces significant sample heating, which reduces the life time of the sample even under good cooling conditions in case of short repetition delays.26 The 1H-detected NMR experiments presented herein for the HN dilute sample do not require homonuclear or heteronuclear decoupling, thus, reducing sample heating, set-up time, and possible experimental missettings. In addition, the proposed scheme does not require rescaling of the proton chemical shift, which is often problematic when spectra are acquired with homonuclear decoupling. Figure 3 shows the experimental data for the 1H T1 measurements. 1H T1 times were found to be equal to 0.98 s (reference sample) and 1.76 s (HN dilute sample), respectively. The HN dilute sample has an unexpectedly short inversion recovery time T1, allowing a recycle delay of 2.2 s, while for the reference sample a repetition delay of 2.3 s was used to allow for dissipation of heat and to avoid sample degradation. Experimental data for a 1H inversion recovery experiment of 1HN bulk signal of the HN dilute (▪; 1H T1=1.76 s) and the reference sample (•; 1H T1=0.98 s). The improvement in 1H line width is achieved by dilution of the proton spin density by a factor of 10. This results in a decrease of the Boltzman magnetization in the same proportion and in an increase of the longitudinal 1H T1 relaxation time. These drawbacks in sensitivity are compensated by using protons for detection which yields a gain in sensitivity by a factor of 5–9.21 In addition, the 1H line width is decreased by approximately a factor of 4. The signal-to-noise ratio for the 15N- and 1H-detected experiments amount to 10.4:1 and 25.3:1, respectively (determined for a 1H cross section through the cross peak G51, as indicated in Figure 1 C and D1). Taking into account the different amount of material and experimental time, the normalized signal-to-noise ratio for the HN dilute sample is reduced by only a factor of 1.2 compared to the reference sample. The experiments therefore demonstrate that a high degree of deuteration leads to ultra-high-resolution proton spectra, which could not be achieved to date despite many attempts to improve homonuclear decoupling schemes and the development of fast MAS technologies. We expect that this labeling approach will enable a straight forward assignment strategy, making an additional backbone nucleus available for resonance dispersion. The employed 2D pulse scheme using proton detection is illustrated in Figure 4 A. Effective suppression of the dominant water resonance was achieved by modification of the constant time (CT) experiment suggested by Zilm and co-workers.14 After magnetization transfer from 1H to 15N, polarization is stored along the z-axis during a variable delay (τ−t1/2), which precedes and follows the 15N evolution period t1. Two variable delays are required to achieve J decoupling in the indirect dimension and to keep the experiment constant time with respect to water magnetization. The fixed delay τw (60–120 ms) which follows the CT period, is optimized for water signal suppression. After back-transfer of magnetization to 1H, magnetization is acquired using Waltz-16 (ω1=1.6 kHz) for heteronuclear scalar decoupling.14 Pulse sequences employed for A) 1H-detected and B) 15N-detected 15N–1H correlation experiments. CP=cross polarization. Figure 4 B shows the pulse sequence which was employed for the 15N-detected 1H–15N correlation spectra. PMLG-927 was implemented in the indirect 1H evolution period to achieve 1H–1H dipolar decoupling (ω1=81 kHz). A 180° pulse on the 15N channel is applied in the center of t1 for heteronuclear J decoupling. During 15N detection, TPPM proton decoupling was applied using a RF field of ω1=90 kHz. Supporting information for this article is available on the WWW under http://www.wiley-vch.de/contents/jc_2002/2006/z600328_s.pdf or from the author. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.