Dynamic Nuclear Polarization of Oxygen-17

Abstract

Oxygen-17 detected DNP NMR of a water/glycerol glass enabled an 80-fold enhancement of signal intensities at 82 K, using the biradical TOTAPOL. The >6,000-fold savings in acquisition time enables ¹⁷O-¹H distance measurements and heteronuclear correlation experiments. These experiments are the initial demonstration of the feasibility of DNP NMR on quadrupolar ¹⁷O.

Recent years have seen an avalanche of new magic angle spinning (MAS) NMR methods developed to determine structures of macromolecular systems that are not amenable to study by either solution NMR or diffraction techniques, the two principle tools of structural chemistry. In particular, with MAS NMR there are now established methods to assign spectra and to perform distance and torsion angle measurements for the I=1/2 species ¹H, ¹³C, ¹⁵N and ³¹P.^1,2 This in-turn has enabled studies of protein and nucleic structures and dynamics. Largely missing from the biomolecular NMR repertoire is oxygen, a key element of water and a variety of other chemically and biologically important functional groups. The difficulty is that the most abundant oxygen isotope, ¹⁶O, has no magnetic moment and the next most abundant, ¹⁷O, has a quadrupolar nucleus (I=5/2), with attendant spectral complications. On the other hand, ¹⁷O chemical shifts span ~1000 ppm with the potential for chemically significant spectral resolution. Two recent examples showed that the ¹⁷O shifts of carboxyl groups are dispersed over 60 ppm³ and that the –C¹⁷O₂ shift tensor elements are sensitive to H-bonding and span ~550-600 ppm⁴.

Progress in ¹⁷O NMR has been stymied primarily by the low sensitivity resulting from the broad 2^nd order powder patterns and low resolution of the spectra. This problem is further exacerbated by the low efficiency of the techniques used to observe ¹⁷O. In more detail, ¹⁷O MAS spectra display residual 2^nd order broadening characteristic of a 4^th rank tensor that is not averaged to zero by MAS ^5,6. To observe isotropic chemical shifts in the presence of and convolved with the 2^nd order interaction requires either special instrumentation, as in case of double rotation (DOR) and dynamic-angle spinning (DAS)^7,8, or special spectroscopic techniques, as in case of multiple-quantum magic angle spinning (MQMAS)⁹ and or satellite-transition magic angle spinning (STMAS)¹⁰. However, when the quadrupole coupling is large (> 5 MHz) the excitation efficiency of these approaches drops dramatically; in the case of MQMAS spectra, to about ~5% ^11,12. Thus, although there are a number of exciting MQMAS studies of ¹⁷O labeled biological samples, the experimental results are clearly limited by signal-to-noise^3,13-20. In order to enable ¹⁷O NMR as an important spectroscopic technique, a dramatic increase in sensitivity is required.

Recently, high frequency dynamic nuclear polarization (DNP) has provided immense gains in NMR sensitivity via microwave-induced transfer of polarization from paramagnetic centers to nuclei ²¹. In general, ¹H's are polarized directly and then cross-polarization is used to transfer the enhanced polarization to other nuclei (e.g., ¹³C, ¹⁵N, etc.), all at cryogenic temperatures (~85 K).^22-28 This approach has been successfully applied to membrane proteins^29-31 , peptides^32,33, amyloid fibrils³⁴ as well as surfaces^35,36. Most of these studies focus on I=1/2 nuclei (e.g., ¹³C, ³¹P, ²⁹Si, etc.) and a few on the I=1 and 5/2 quadrupolar species ^35-37. The magnetization transfer yields enhancements ε = 30-250, and the low temperature yields another gain of ~3.5 in Boltzmann polarization relative to ambient temperature. In combination, ε^†~105-875, which could dramatically improve the prospects of performing ¹⁷O experiments in biological systems.

Here we demonstrate the feasibility of this approach in the simple case of H₂¹⁷O. We observe for the first time an enhancement of ε~80 or ε^†~280 for ¹⁷O using high field DNP NMR and the polarizing agent TOTAPOL in a glassy glycerol/D₂O/H₂O matrix. In particular we are able to polarize ¹⁷O for both echo and Carr-Purcell-Meiboom-Gill (CPMG) ³⁸ experiments and to use spin-echo double resonance (SEDOR) ^39,40 experiments to measure ¹H-¹⁷O distances. The dramatic savings in acquisition time demonstrates a potentially important approach to ¹⁷O spectroscopy that can be extended to studies of H₂¹⁷O in other chemical systems and of ¹⁷O in other functional groups.

Figure 1 shows a series of central transition (1/2 → -1/2) 1D spectra demonstrating DNP enhancements of ¹⁷O. Spectra acquired with a cross polarization (CP) echo experiment with and without microwave irradiation are shown in Figure 1b and 1e, respectively. The wave on-signal was acquired with ~20 minutes of signal averaging and the wave off-signal was acquired over a period of 14 hours with acquisition parameters identical to those for the on-signal. Comparison of the two yields ε~80.

Figure 1.

Central transition ¹⁷O-¹H cross-polarization-echo of a static glycerol/D₂O/H₂¹⁷O (60/30/10 % by weight) recorded at T = 82 K in which the water was originally 74% enriched in H₂¹⁷O. The final concentration of H₂¹⁷O was ~7.4%: (a) lineshape simulation, (b) wave on with ε~80 (128 scans), (c) μwave on using a CPMG (16 scans), (d) wave off scaled by 10 and (e) wave off (4864 scans).

The shape of the static powder pattern shown in Figure 1b is characteristic of a second-order quadrupolar coupling pattern ^41-43. Chemical shielding anisotropy can also influence the lineshape, although, for water at 5 T, this effect is negligible compared to that of the quadru-polar interaction. The quadrupolar coupling constant, C_Q = 6.8 ± 0.2 MHz, the asymmetry parameter, • = 0.95 ± 0.05 and the isotropic chemical shift, δ_iso = 0 ± 50 ppm were determined by simulating the non-spinning powder pattern (Figure 1a). The sizeable quadrupolar coupling is reasonable for a frozen water environment. ⁴⁴ The ~85 kHz breadth of the central transition limits the ability to use magic-angle spinning (MAS). In particular, the water/glycerol oxygen site at 5 T would require ω_r/2π > 30 kHz to isolate the central transition from a series of rotational sidebands. Current spinning frequencies with 4 mm rotors at cryogenic temperatures are limited to ~7 kHz due to the density of the N₂ drive and bearing gases being near the liquefaction point (T = 77 K). As DNP NMR instruments are developed for higher fields, lower spinning frequencies could suffice due to the inverse relationship between the breadth of the second-order quadrupolar interaction and B₀. For example, at 14 T, ω_r/2π = 10 kHz would be sufficient for ¹⁷O MAS DNP experiments. Simulations illustrating this point are included in the Supporting Information.

CPMG experiments can provide higher signal-to-noise ratios by splitting the broad quadrupolar powder pattern into a series of sharp spikes,^38,45-49 and Figure 1c shows that DNP-enhanced CPMG can further accelerate the acquisition of ¹⁷O spectra. A standard echo provided sufficient signal-to-noise with enrichment on the order of 10% or more; however CPMG proves to be useful for natural abundance (0.038%) H₂O/D₂O/glycerol glasses (Figure S1). This may enable future studies using low-levels of enrichment. DNP and CPMG might usefully be coupled to other sequences for quadrupolar nuclei and species that show large chemical shielding anisotropies if T₂'s are sufficiently long. Care must be taken when multiple sites are present, as different T₂'s could complicate spectral interpretation. Radical concentration can also compromise the CPMG method by paramagnetic relaxation. In Figure 1c, a train of 128 echoes, provided a signal beyond 11 ms. This indicates that, in this system, 20 mM of the biradical TOTAPOL^50,51 (or 40 mM electrons) had little effect on the spectrum (i.e., the breadth of the central transition is dominated by the quadrupolar line-shape). The build-up time constant (τ_B) for the water/glycerol glass was experimentally determined to be 4.1 s.

The low sensitivity of ¹⁷O NMR often requires weeks of acquisition time, rendering multidimensional structural experiments impractical. In consequence, most distances measured between ¹⁷O and either ¹H or ¹³C by NMR are indirect, applying a dephasing pulse on ¹⁷O and observing ¹H or ¹³C. ^47,48,52,53 In Figure 2 we show the results of a SEDOR experiment used to measure the average ¹H-¹⁷O distance in the glycerol/water glass. A 1D stepped-SEDOR experiment was used to acquire a ¹⁷O⁵⁴ (1-S/S_o) curve characteristic of the dipolar coupling (Figure 2a).

Figure 2.

Central transition ¹⁷O-¹H SEDOR experiment recorded at T=85 K using 74% enriched in ¹⁷O: (a) SEDOR curve with SIMPSON simulation and (b) 2D SEDOR spectrum. The sample composition is as in Figure 1.

The SEDOR curve was simulated using SIMPSON ⁵² and SPINEVOLUTION ⁵⁵ software and is compared in Figure 2a with the experimental curve determined from the peak areas. Various spin models included a single oxygen site with 2-5 ¹H's between 0.8 and 1.2 Å and <HOH angles between 100⁰ to 110⁰ (values were chosen based on upper and lower limits of the experimental data). These inputs were adjusted until a reasonable model of the glass was obtained. A three-spin model (one - O and two – H's) indicated an average H---O bond distance of 1.1 ± 0.1 Å, a value that agrees well with X-ray and neutron diffraction data from various crystalline ice structures where H--O distances are found between 0.86 and 1.15 Å. ^56-58

A second Fourier transformation of the evolution dimension of the same SEDOR experiment yields a 2D dipole-quadrupolar correlation spectrum (Figure 2b) similar to those reported previously ^59,60. This spectrum is sensitive to the relative orientation of the dipole and quadrupolar tensors, and this information, in principle, can be extracted by fitting the experimental data. ⁵⁹

We found it to be difficult to achieve acceptable agreement between the experimental and simulated spectra. This difficulty can be attributed to the following two reasons. First, the samples used for all experiments were in the glassy state. Therefore, there must be a certain degree of structural disorder present (e.g., distance, angle and possibly coordination). Second, mobile protons were significantly diluted with deuterons, which come from deuterated glycerol and D₂O used for sample preparation. Therefore, the observed 1D and 2D experiments have contributions from ¹⁷O sites with varying number of neighboring protons. In the 2D SEDOR, the ¹⁷O-¹H heteronuclear dipolar coupling can be estimated from the dipolar dimension⁶¹. The observed splitting of 16.3 ± 0.3 kHz, corresponds to a one bond distance of ~1 Å.

DNP NMR enables SEDOR experiments to measure single ¹⁷O bond lengths within hours. This could be extended to ¹⁷O-¹³C or ¹⁷O-¹⁵N bond distances with appropriate hardware and to other sequences, such as rotational echo double resonance (REDOR) and transferred echo double resonance (TEDOR), with higher field DNP. Experimental times for the identical experiments without DNP are projected to require >1.4 years.

Multidimensional experiments also enable the extraction of structural information by correlating resonance(s) from one NMR nucleus to another, through either the dipolar or J-coupling interactions. Figure 3 shows a ¹⁷O-¹H heteronuclear correlation spectrum (HETCOR) for a water/glycerol glass. A projection along the ¹⁷O dimension yields the quadrupolar pattern similar to Figure 1b. The indirect dimension (¹H) exhibits a broad resonance dominated by ¹H-¹H homonuclear dipolar coupling. ¹H chemical shielding and heteronuclear dipolar coupling are also present, although they contribute minimally to the observed lineshape.

Figure 3.

Central transition ¹⁷O-¹H HETCOR-CPMG spectrum of water/glycerol glass using solid-state DNP NMR, T=85 K. The sample composition is as in Figure 1.

For quadrupolar NMR, a 16-phase cycle is often used during acquisition of one- and two-dimensional data often leading to improved quadrupolar lineshapes. With the sensitivity boost from DNP and CPMG, a HETCORCPMG spectrum of the water/glycerol glass was also acquired using a 8 scan, 8-phase cycle acquisition scheme (not shown). The reduced phase cycle (versus 16) provided similar results, with only minor variations within the discontinuities. With the use of hard pulses, increased field (B_o) strength and studying systems with lineshapes below 80 kHz, this could provide an alternative for multidimensional experiments. Further reducing acquisition time for broad resonances and low sensitivity.

In summary, DNP in the simple case of H₂¹⁷O provided an enhancement of 80. Water is of great importance in both biological and chemical processes, but is challenging to study using conventional ¹⁷O solid state NMR. The ability to polarize protons by microwave irradiation followed by cross-polarization to oxygen has enabled a suite of multi-dimensional NMR experiments of this low-gamma, quadrupolar nucleus within hours. ¹⁷O-detected multidimensional spectra were acquired between three and 10 hours using DNP NMR. By enhancing the signal-to-noise ratio, DNP provides an ~6,000-fold gain in time at cryogenic temperatures, and, in the absence of relaxation, a >18,000-fold gain in time compared to room temperature (298 K). These gains should be transferable to other difficult quadrupolar nuclei, providing a savings in acquisition time. The significant gain in sensitivity and reduction of experimental time from months to hours is extremely beneficial and provides new possibilities for further exploration of heteronuclear correlations and distances in multidimensional experiments. Extensions to MAS experiments for site assignments are currently in progress ⁶².