Harnessing Water to Enhance Quadrupolar NMR Spectroscopy and Imaging

Abstract

¹⁷ O and ¹⁴ N are attractive targets for in vivo NMR spectroscopy and imaging, but low gyromagnetic ratios γ and fast spin relaxation complicate observations. This work explores indirect ways of detecting some of these sites with the help of proton-detected double resonance techniques. As standard coherence transfer methods are of limited use for such indirect detection, alternative routes for probing the quadrupolar spectra on ¹ H were tested. These centered on modulating the broadening effects imparted onto protons adjacent to the low-γ species through J couplings through either continuous wave or spin-echo double-resonance decoupling/recoupling sequences. As in all cases, the changes imparted by these double-resonance strategies were small due to the fast relaxation undergone by the quadrupoles, the sensitivity of these approaches was amplified by transferring their effects onto the abundant water ¹ H signal. These amplifications were mediated by the spontaneous exchanges that the labile ¹ Hs bound to ¹⁷ O or ¹⁴ N undergo with the water protons. In experiments designed on the basis of double-resonance spin echoes, these enhancements were imparted by looping the transverse encodings together with multiple longitudinal storage periods, leading to decoupling-recoupling with exchange (D-REX) sequences. In experiments designed on the basis of continuous on/off quadrupolar decoupling, these solvent exchanges were incorporated into chemical-exchange saturation transfer schemes, leading to decoupling-recoupling with saturation transfer (D-REST) sequences. Both of these variants harnessed sizable proportions of the easily detectable water signals, in order to characterize the NMR spectra and/or to image with atomic-site specificity the ¹⁷ O and ¹⁴ N species.

Keywords: chemical exchange; low-gamma MRI; quadrupolar NMR spectroscopy; sensitivity enhancement; water detection.

Figure 1

Pulse sequences considered in this work to highlight quadrupolar nuclei S bound to labile ¹Hs on the water resonance. a) Spin echo (SE) scheme based on the work of Ronen et al., in which S‐decoupling is/is not applied during an echo time TE and the difference on water is observed. b) D‐REX, a FLEX‐inspired scheme in which the S‐spin is/is not highlighted with either hard pulses or with S‐decoupling, and the whole module is repeated N times to port the S‐encoding thus imparted from the S‐bound proton to the water, as driven by exchanges happening during a mixing time t _mix. S‐encoding modes were based on either SEDOR, that is, on applying a series of π pulses i=1 times (as in BIRD ^[51] ), and i=4 or i=8 times using xy‐4 or xy‐8 phase cyclings. The S‐effects were also introduced by using a continuous decoupling (CW) scheme; this included an i=0 case representing a jump−return (JR) sequence in which τ comprises the full time in between π/2 pulses. c) D‐REST, a saturation transfer scheme in which an RF pulse is applied on protons for a selective saturation, while an S‐decoupling sequence is/is not applied on the S‐spin.

Figure 2

Analytical calculations of the signals resulting from the sequences introduced in Figure 1a, b for ¹⁷O in natural‐abundance water. T _Q=4.4 ms, J=91 Hz, k _sw=555 Hz, T _1,w 1.41 s, T ₂=41 ms (akin to gray matter ^[54] ) or 500 ms (as in CSF). a) Simulation expected for the spin‐echo ¹⁷O‐encoded signal. b) Expectations for the ¹⁷O‐encoded D‐REX sequence and Equation (4), for t _mix=10 ms, τ=1.6 ms, and optimization as in Figure S1. In (a) and (b), the responses account for the ¹⁷O natural abundance. c) Calculation of the D‐REST effect upon ¹⁴N‐encoding a dilute, 20 mM metabolite modeled on an amino acid,^[47,48] with an exchange rate of 200 Hz, T _Q=1 ms, J=62 Hz, T _1,w=1.41 s, T _2,H=60 ms, 99.6 % natural abundance. The dependence of this effect on the proton saturation field for a 2 s long CW pulse was explored, and the effect was combined with the overall CEST effect.

Figure 3

Optimization of the different parameters regulating ¹⁷O‐encoded D‐REX, conducted on a 50 : 50 mixture of water and [D₆]DMSO where water was enriched with 25 % ¹⁷O. All the experiments were acquired at 7 T and 24 °C with an ¹⁷O RF field of 846 Hz. a) Optimization of τ, with N=100, t _mix=15 ms, for the different D‐REX modes described. b) Optimization of t _mix, with τ=2 ms, and N=100. A t _mix of 20 ms was estimated as optimal. c) Optimization of N, with t _mix=20 ms and τ=2 ms. b) and c) were run with the JR‐based D‐REX, which in the best case achieved an ∼11 % contrast on the ¹H water signal. Also included in this panel is a simulation based on water's H₂ ¹⁷O quadrupolar parameters, an exchange rate k _sw=50 Hz, T _1,w=2 s, T _2,H=0.65 s, and other parameters as in the experiment itself.

Figure 4

Comparisons between the enhancements afforded by ¹⁷O‐encoded SE experiments at different echo times TE (a and c), and by ¹⁷O‐encoded D‐REX for different N loop numbers (b and d). For the latter, a JR version of D‐REX was employed. Experiments were conducted on the same sample as in Figure 3, at 7 T and 37 °C with the indicated parameters.

Figure 5

Indirect chemical shift mapping of the ¹⁷O NMR spectrum for a) the sample introduced in Figure 3 and b) 10 % natural abundance water in 90 % D₂O. Both spectra involved a point‐by‐point scanning of the ¹⁷O offset over 70 ppm in increments of 1 ppm, where each point corresponds to one scan, normalized by the first point (which is considered as off‐resonance for ¹⁷O). A JR ¹⁷O‐encoded D‐REX sequence was used in both cases with the indicated parameters, in ∼10 min each (7 T field, 24 °C).

Figure 6

Images resulting from xy‐8‐based ¹⁷O‐encoded D‐REX, with the final pulse in Figure 1b replaced by a RARE detection (RARE‐factor=8). a) Experiments performed on the sample introduced in Figure 3, with τ=4 ms, t _mix=20 ms, N=80, slice thickness=3 mm, in‐plane resolution=150x300 μm², 64 phase encodes, 1 scan per phase encode, acquisition time of 3 min 12 s. b) Experiments on an extracted brain immersed in Fluorinert, conducted with τ=0.6 ms, t _mix=20 ms, N=40, slice thickness=5 mm, in‐plane resolution=313x313 μm², 64 phase encodes, 8 scans per phase encode, acquisition time of 18 min 8 s. Both experiments employed a decoupling field of 424 Hz and were performed at 7 T and 24 °C.

Figure 7

¹⁴N‐encoded D‐REX experiments on a 1.86 M, pH 1 glycine solution in 10 % D₂O at 7 T and 24 °C, with a JR‐CW version of D‐REX. N was optimized with on‐resonance decoupling at 10 ppm with a field of 300 Hz; a) the two individual spectra and b) the difference. Experiments were performed with optimized t _mix=15 ms and τ=5.2 ms, with a ∼5 % contrast on the ¹H water signal being observed.

Figure 8

D‐REST experiments detecting a low‐γ ¹⁴N nucleus on the abundant ¹H signal of water (set to 0 ppm). a) Comparison of on‐ and off‐resonance CW decoupling on the CEST spectral mapping of a 200 mM urea sample at pH=7.1; an RF field of 10 Hz was applied for 1 s for the proton saturation. The on and off experiments involved ¹⁴N decoupling at 0 and at 450 ppm from the urea ¹⁴N resonance, respectively. b) Idem for a 1 M lysine sample at pH 2.0, with a proton ω₁ of 10 Hz applied for 1 s for saturation, and ¹⁴N decoupling applied at 21 and 471 ppm for the on/off experiments (lysine's ¹⁴N^ϵ resonating was the only one identified at 20.6 ppm) All experiments were performed at 7 T and 24 °C and employed decoupling nutation fields of 1.4 kHz, 1 scan per ¹H offset, and a ¹H offset incremented in 0.1 ppm steps.

Figure 9

a) ¹⁴N chemical shift mapping in the urea sample introduced in Figure 8, arising from stepping the ¹⁴N decoupling frequency in a D‐REST experiment over 250 ppm in 2 ppm steps. For all steps, the urea protons were saturated with a 10 Hz field at +1 ppm downfield from water, and a ¹H water spectrum was collected thereafter. b) D‐REST experiment targeting a urea/lysine/alanine phantom, incorporating a RARE block as signal acquisition. Lysine and urea were as in Figure 8; alanine was present at 100 mM and pH 2.2. The NH₂ ^{ϵ 1}H resonance of lysine was saturated with a field of 16 Hz on resonance at +2.6 ppm, and performed as a difference experiment between ¹⁴N on‐resonance decoupling at 28.8 ppm and an off‐resonance decoupling spectrum, slice thickness=2 mm, in‐plane resolution=117×117 μm², 128 phase encodes, 1 scan per phase encode, RARE factor of 8, acquisition time of 1 min 4 s. All remaining conditions were as in Figure 8.