Quasi-Resonance Signal Amplification By Reversible Exchange

Abstract

Here we present the feasibility of NMR Signal Amplification by Reversible Exchange (SABRE) using radio-frequency irradiation at low magnetic field (0.05 T) in the regime, where the chemical shifts of free and catalyst bound species are similar. In SABRE, the ¹⁵N-containing substrate and parahydrogen perform simultaneous chemical exchange on an Iridium hexacoordinate complex. Shaped Spin-Lock Induced Crossing (SLIC) radio-frequency pulse sequence followed by a delay is applied at QUASi-Resonance (QUASR) condition of ¹⁵N spins of ¹⁵N-enriched substrate. As a result of this pulse sequence application, ¹⁵N z-magnetization is created from the spin-order of parahydrogen derived hyperpolarized hydrides. The repetition of the pulse-sequence block consisting of shaped radio-frequency pulse and the delay leads to the build-up of ¹⁵N magnetization. The modulation of this effect by the irradiation frequency, pulse duration and amplitude, delay duration, and the number of pumping cycles was demonstrated. Pyridine-¹⁵N, acetonitrile-¹⁵N, metronidazole-¹⁵N₂-¹³C₂ substrates were studied representing three classes of compounds (five- and six-membered heterocycles and nitrile) showing the wide applicability of the technique. Metronidazole-¹⁵N₂-¹³C₂ is an FDA-approved antibiotic that can be injected in large quantities promising non-invasive and accurate hypoxia sensing. The ¹⁵N hyperpolarization levels attained with QUASR-SABRE on metronidazole-¹⁵N₂-¹³C₂ were more than two-fold greater than with SABRE-SHEATH (SABRE in SHield Enables Alignment Transfer to Heteronuclei) demonstrating that QUASR-SABRE can deliver significantly more efficient means of SABRE hyperpolarization.

Table of Contents Graphic

Conventional NMR relies on equilibrium thermal nuclear spin polarization P dictated by the Boltzmann distribution among Zeeman energy levels in dependence of the applied static magnetic field B₀. Although P can be boosted significantly by applying stronger magnetic field (because P∝B₀), P is typically on the order of 10⁻⁵ to 10⁻⁶ for a conventional high-field NMR spectrometer (ca. 9.4 T) or MRI scanner (ca. 3 T) at room temperature, i.e. when the high temperature approximation holds. Hyperpolarization techniques increase P to the order of unity increasing NMR sensitivity increase by 4–5 orders of magnitude.^1–3

Several hyperpolarization techniques exist.^1–3 Signal Amplification by Reversible Exchange (SABRE) is one of more recent techniques pioneered by Duckett and co-workers in 2009.^4–7 SABRE relies on simultaneous chemical exchange of parahydrogen (p-H₂) and to-be-hyperpolarized substrate (Figure 1a).⁸ When the transient complex is formed, the parahydrogen symmetry is broken,⁹ and the network of spin-spin couplings can enable transfer polarization from parahydrogen-derived hydrides to the nuclear spins of the substrate.^4–7 Two major groups of approaches have been developed for SABRE polarization transfer: the first group employs a matching static magnetic field B_evo,^{4, 10–13} and the second group applies radio-frequency (RF) pulse sequences,¹⁴ to approach Level Anti-Crossings (LAC)^15–16 and induce polarization transfer. Both approaches have merit depending on the application. For biomedical applications, which represent the main driver for development of hyperpolarization technology,² the key is to achieve high degrees of polarization with the long lifetimes in a suitable biomolecular motif.¹⁷ So far, approaches relying on static magnetic fields such as SABRE-SHEATH (SABRE in SHield Enables Alignment Transfer to Heteronuclei)^18–20 have been the most efficient for preparation of long-lived ¹⁵N hyperpolarized spin states with exponential decay constant of more than 20 minutes²¹ and P_15N exceeding 30%.^22–23

Figure 1.

a) The diagram of molecular exchange with p-H₂ in SABRE hyperpolarization. b) the experimental setup for QUASR-SABRE, c) the RF pulse sequence for QUASR-SABRE, d) corresponding experimental setup for SABRE-SHEATH experiment.

One likely explanation as to why RF-based methods are lagging behind in the context of SABRE is the reliance on conventional high-field NMR spectrometers with small coils that only encompass a small fraction of the liquid sample in a 5 mm NMR tube,¹⁴ which is continuously bubbled with p-H₂.²⁴ These factors result in major RF-inhomogeneities. In contrast, a previous approach used in hydrogenative Parahydrogen Induced Polarization (PHIP^{9, 25–26}), employs low-field (ca. 5–50 mT) magnets and RF excitation coils,^27–32 which encompass the entire sample volume, sometimes in excess of 50 mL.^33–35 Moreover, the hardware behind such low-field devices is significantly less complex and less costly compared to that of the high-field NMR spectrometers.^{17, 31–32, 35–36} Here it is demonstrated that these advantageous features can also be translated to SABRE.

RF-based polarization transfer such as Low-Irradiation Generation of High Tesla (LIGHT)-SABRE employs RF irradiation of the catalyst-bound substrate (Figure 1A), which typically has a chemical shift difference of 30–50 ppm with respect to the free substrate.^{14, 37} Irradiation of the catalyst bound species allows for polarization transfer from p-H₂-derived hydrides.¹⁴ When the complex dissociates, the ¹⁵N nuclear spin polarization is preserved in the free substrate, because it is not affected by the frequency-selective RF pulses.¹⁴ Achieving this frequency selective irradiation at high magnetic fields is trivial due to large chemical shift dispersion. For example, 50 ppm difference equals to ~2 kHz at 9.4 T. However, this difference vanishes at low magnetic fields: for example, 50 ppm difference equals 10 Hz at 0.05 T (the magnetic field employed in this Letter), and selective RF-excitation becomes challenging. We demonstrate that this challenge can be overcome through the use of quasi-resonance (QUASR) spin-lock induced crossing (SLIC)³⁸ irradiation to polarize ¹⁵N spins from p-H₂-derived hydrides.

The QUASR-SABRE experiment (Figure 1B) is performed at 0.05 T magnetic field using previously described p-H₂ bubbling setup in a medium-walled 5-mm NMR tube.^{24, 39–40} During p-H₂ bubbling a triangular-shaped pulse is applied for duration τ_PULSE followed by a delay period τ_DELAY. The process is repeated, and the net z-magnetization increases during this “rf-pumping” process. The resulting magnetization can be conveniently assessed by applying a broad-band excitation (π/2) RF pulse, Figure 1c. We compare the performance of this QUASR-SABRE approach with SABRE-SHEATH approach (Figure 1d), which has been employed previously to obtain record-high ¹⁵N polarization in excess of 30%.^22–23

The previously described SABRE-SHEATH setup,^{39–40 15}N RF coil and 0.05 T magnet have been employed here.⁴¹ Samples of three substrates and IrIMes catalyst precursor in perdeuterated methanol were prepared as follows: pyridine-¹⁵N/catalyst, ~20 mM/~1 mM, acetonitrile-¹⁵N/catalyst, ~40 mM/~1–2 mM, metronidazole-¹⁵N₂-¹³C₂/catalyst, ~20 mM/~1 mM. All ¹⁵N-enriched compounds were purchased from Isotec. The RF pulse sequences was coded and applied on Kea-2 NMR spectrometer (Magritek, New Zeeland) using Tomco RF amplifier. The employed Kea-2 spectrometer was operated in the signal averaging mode, where the signal is averaged during multi-scan acquisitions version being added. As a result, the signal integral value from multi-scan spectrum is similar to that acquired using 1-scan acquisition, e.g. the integral values of spectra shown in corresponding displays in Figures 2–4 can be compared directly without any scaling even though the spectra were recorded using different numbers of scans. A Three-layered mu-metal magnetic shield was employed (Magnetic Shield Corp., Bensenville, IL, P/N ZG-206). All data was acquired employing 75–80% parahydrogen prepared using home-built parahydrogen generator based on Sunpower cryo-chiller at a flow rate of 150 standard cubic centimeters per minute (sccm). For the SABRE experiments, the following conditions were used: 75 psi back-pressure and 70 sccm para-H₂ flow rate (Figure 1b and Figure 1d). The flow rate was maintained by via a mass flow controller (MFC, Sierra Instruments, Monterey, CA, P/N C100L-DD-OV1-SV1-PV2-V1-S0-C0).

Figure 2.

Pyridine-¹⁵N data. The other experimental conditions were as follows: 20 mM pyridine-¹⁵N, 1 mM catalyst in CD₃OD. Note the width of the signal in displays b and d. a) the ¹⁵N spectrum obtained after performing SABRE-SHEATH; b) The ¹⁵N spectrum recorded using 90-degree excitation pulse when bubbling p-H₂ in situ of the 0.05 T magnet; c) the ¹⁵N spectrum obtained after performing QUASR-SABRE; d) ¹⁵N QUASR-SABRE signal dependence on the applied radio frequency offset from the actual resonance condition; e) the build-up of ¹⁵N QUASR-SABRE signal as a function of the number of pumping cycles. Note the individual spectra employed for figures in displays d and e were auto-phased, and the data is presented in the magnitude mode.

Figure 4.

Metronidazole-¹⁵N₂-¹³C₂ data. The other experimental conditions were as follows: ~20 mM metronidazole-¹⁵N₂-¹³C₂, 1 or 2 mM catalyst in CD₃OD. Note the width of the signal in displays a and c is nearly the same as opposed to pyridine-¹⁵N case. a) the ¹⁵N spectrum obtained after performing SABRE-SHEATH; b) The ¹⁵N spectrum recorded using 90-degree excitation pulse when bubbling p-H₂ in situ of the 0.05 T magnet; c) the ¹⁵N spectrum obtained after performing QUASR-SABRE; d) ¹⁵N QUASR-SABRE signal dependence on the frequency of the applied RF shaped pulse; e) the build-up of ¹⁵N QUASR-SABRE signal as a function of the number of pumping cycles N. Note the individual spectra employed for figures in displays d and e were auto-phased, and the data is presented in the magnitude mode.

Low-field detection of hyperpolarized (HP) compounds offers sufficient detection sensitivity in the context of NMR detection of HP states.⁴² However, detection of thermal polarization is challenging due to low P even at high concentrations (see Supporting Information (SI) for details). As a result, the quantification of ¹⁵N enhancements (ε_15N) and polarization (P_15N) relied on signal to noise measurements (see SI for details) in order to determine the minimum values achieved.

¹⁵N NMR spectra with high SNR were obtained for all three studied molecules: pyridine-¹⁵N, acetonitrile-¹⁵N, and metronidazole-¹⁵N₂-¹³C₂ using SABRE-SHEATH (Figure 2a, Figure 3a, and Figure 4a respectively) and QUASR-SABRE (Figure 2c, Figure 3c, and Figure 4c respectively). We note that while p-H₂ bubbling through the sample placed in 0.05 T leads to ¹⁵N signal even without RF pulses (Figure 2b, Figure 3b and Figure 4b respectively), this ¹⁵N signal is distinctly anti-phase,^{14, 41, 43} and has significantly lower intensity compared to those obtained using SABRE-SHEATH and QUASR-SABRE protocols.

Figure 3.

acetonitrile-¹⁵N data. The other experimental conditions were as follows: 40 mM acetonitrile-¹⁵N, 1 or 2 mM catalyst in CD₃OD. Note the width of the signal in displays a and c is nearly the same as opposed to pyridine-¹⁵N case. a) the ¹⁵N spectrum obtained after performing SABRE-SHEATH; b) The ¹⁵N spectrum recorded using 90-degree excitation pulse when bubbling p-H₂ in situ of the 0.05 T magnet; c) the ¹⁵N spectrum obtained after performing QUASR-SABRE; d) the build-up of ¹⁵N QUASR-SABRE signal as a function of the number of pumping cycles; Note the individual spectra employed for figures in display d were auto-phased, and the data is presented in the magnitude mode.

The key distinct feature of QUASR-SABRE phenomenon is strong RF offset frequency dependence: Figure 2d and Figure 4d demonstrate the dependence of the ¹⁵N QUASR-SABRE signal for pyridine-¹⁵N and metronidazole-¹⁵N₂-¹³C₂ respectively. Note that very small signal is obtained at the resonance frequency. The maximum intensity of ¹⁵N QUASR-SABRE was compared to that of SABRE-SHEATH and is reported here as the ratio of ¹⁵N QUASR-SABRE and SABRE-SHEATH signal, η. We find η of ~1.0 for pyridine-¹⁵N, η of ~0.44 for acetonitrile-¹⁵N, and η of (at least) ~2.4 for metronidazole-¹⁵N₂-¹³C₂. The relatively low η value for acetonitrile-¹⁵N is in part explained by the fact that frequency optimization was not performed, and the data was recorded using the frequency offset parameter optimized for pyridine-¹⁵N, which unfortunately did not provide a fair comparison. Based on the range of on the frequency optimization data for pyridine-¹⁵N (Figure 2e) and for metronidazole-¹⁵N₂-¹³C₂ (Figure 4d), we estimate that this optimization may potentially yield an improvement of up ten-fold. We note that η value for metronidazole-¹⁵N₂-¹³C₂ assumes that both ¹⁵N sites are hyperpolarized via SABRE-SHEATH and QUASR-SABRE protocols. While SABRE-SHEATH indeed yields hyperpolarization of both ¹⁵N sites due to spin-relay of polarization at very low magnetic fields (the Earth’s field (ca. 50 μT) and below), QUASR-SABRE may yield hyperpolarization of only one (N₃) site, Figure 4. If that is the case, then η would be effectively doubled to 4.8, because only one ¹⁵N site contributes to the QUASR-SABRE signal (Figure 4c) versus two ¹⁵N sites to the SABRE-SHEATH NMR signal (Figure 4a). We note that the Full Width at the Half-Height (FWHH) was approximately the same for SABRE-SHEATH and QUASR-SABRE NMR spectra for acetonitrile-¹⁵N and metronidazole-¹⁵N₂-¹³C₂: Figure 3 and Figure 4 respectively, whereas QUASR-SABRE spectrum FWHH was approximately half of that for SABRE-SHEATH spectrum for pyridine-¹⁵N (Figure 2c and Figure 2a respectively). The latter observation may in part explain pyridine-¹⁵N η value, which is significantly lower than that of metronidazole-¹⁵N₂-¹³C₂.

All three molecules studies exhibited a clear and strong dependence of ¹⁵N QUASR-SABRE signal on the duration of the pulse (τ_PULSE) and the duration of the delay (τ_DELAY): Figure S2a, Figure S3a and Figure S3b, Figure S4a and Figure S2b, Figure S3c, Figure S4b respectively. This strong dependence is likely due to the dynamics and kinetics of the substrate and p-H₂ exchange on the catalyst. When the τ_PULSE and τ_DELAY are in sync with chemistry of exchange, the maximum QUASR-SABRE signal may be achieved. However, we note that QUASR-SABRE signal has very complex dependence on τ_PULSE and τ_DELAY. For example, the τ_PULSE-curves for acetonictrile-¹⁵N are vastly different at τ_DELAY of 2 ms (Figure S3a) and τ_DELAY of 33 ms (Figure S3b). Future theoretical work is certainly warranted to study this complex behavior of QUASR-SABRE effect, which is outside the scope of this pioneering phenomenological report.

The second key feature of QUASR-SABRE effect allows for continuous “RF-pumping” of ¹⁵N z-magnetization: Figure 2e, Figure 3d, and Figure 4e for the three studied compounds respectively. This fitting of the exponential dependence of the build-up processed yielded T_b of 0.5±0.05 s for acetonitrile-¹⁵N, T_b of 1.26±0.02 s for pyridine-¹⁵N, and T_b of 2.17±0.09 s for metronidazole-¹⁵N₂-¹³C₂. The T_b values correlate well with η values with R² of 0.93, Figure 5, suggesting that the build-up rate may be affecting the efficiency of QUASR-SABRE polarization process. A likely explanation of this observation in the contribution of the polarization destruction processes to T_b: when the destruction due to RF pulses is significant, T_b is reduced resulting in lower ¹⁵N signals in a manner analogous to that of batch-mode Spin Exchange Optical Pumping (SEOP).^44–47 Further experimental and theoretical studies are certainly warranted in the future to maximize the efficiency of QUASR-SABRE approach described here.

Figure 5.

Correlation plot of T_b and η for the three studied compounds.

All experiments were performed at room temperature (ca. 298 K). We note that the rf pumping of QUASR-SABRE process needs to be effectively matched to the chemical exchange dynamics in order to maximize the polarization transfer efficiency to yield the highest P_15N value. Therefore, it is expected that the optimal values of τ_PULSE and τ_DELAY (Figure S2, Figure S3, and Figure S4) would exhibit temperature dependence, because temperature modulates the rates of substrate and p-H₂ exchange. Moreover, it is entirely possible that optimum temperature (i.e. yielding the highest value of P_15N) may be significantly different from room temperature and may be different from optimal temperature of the SABRE-SHEATH process (note that the optimal SABRE-SHEATH temperature for pyridine-¹⁵N and for metronidazole-¹⁵N₂-¹³C₂ corresponds to approximately room temperature, i.e. SABRE-SHEATH experiments were optimized with respect to temperature). As a result, some additional improvement in the maximum value of P_15N may be potentially expected for QUASR-SABRE process.

We have also investigated the dependence of the QUASR-SABRE signal on the amplitude of SLIC power amplitude. Figure S4c exhibits a plateau (with a range of approximately 6 decibels) with relatively steep slopes on both sides. This trend is expected, because LAC conditions are usually created in a relatively narrow power range.^{15, 48}

Optimization of P_15N was not the goal of this report. Moreover, due to lack of direct ¹⁵N signal reference (due to insufficient thermal equilibrium signal), we can only report the low limit of ε_15N and P_15N values. Metronidazole-¹⁵N₂-¹³C₂ QUASR-SABRE estimates were ε_15N ~9*10⁵ and P_15N ~1.5% (these values are doubled if only N₃ site is hyperpolarized). Metronidazole-¹⁵N₂-¹³C₂ SABRE-SHEATH estimates were ε_15N ~3.7*10⁵ and P_15N ~0.6%. Note these lower-limit estimates are in good agreement with P_15N~1.5% reported for SABRE-SHEATH under similar conditions using detection provided by high-resolution 9.4 T NMR spectrometer.^39–40 Pyridine-¹⁵N lower-limit estimates were ε_15N ~ 6.6*10⁵ and P_15N ~ 1.1% for both QUASR-SABRE and SABRE-SHEATH—in line with previous SABRE-SHEATH studies.^18–19 Acetonitrile-¹⁵N lower-limit estimates were ε_15N ~ 5.3*10⁴ and P_15N ~ 0.09% for QUASR-SABRE and ε_15N ~ 1.2*10⁵ and P_15N ~ 0.2% for SABRE-SHEATH respectively. See SI for details.

With regards to the limitations of the QUASR-SABRE method, it remains to be seen if QUASR-SABRE is capable of hyperpolarization of long-range spin sites in the substrate compounds. Moreover, future systematic experimental and theoretical studies are certainly needed to further optimize the efficiency of QUASR-SABRE technique. For example, more advanced shaped forms (e.g. sine, exponential, trapezoid, etc.) and strategies (adiabatic pulses) can be envisioned.

In summary, radio-frequency based polarization transfer approach has been presented for polarization of ¹⁵N sites. At least in some compounds, this method appears to be more efficient than the SABRE-SHEATH approach, which has already been shown to yield more than 30% ¹⁵N polarization.²³ This is remarkable because in all previous demonstrations RF-SABRE approaches yielded significantly less polarization than static field matching / field cycling approaches. We hope that QUASR-SABRE may ultimately yield ¹⁵N polarization of the order of unity for a wide range of biomolecules. The employed pulse-sequence is a shaped variant of SLIC pulse sequence,³⁸ which has the benefit of using low power levels. The applicability of this technique has been explored for three different types of compounds (six- and five-membered N-heterocycles and acetonitrile) including the antibiotic metronidazole. Metronidazole is an antibiotic that can be administered in large doses,⁴⁹ and contains the nitroimidazole moiety, which has been exploited in a wide range of molecular contrast agents for hypoxia sensing using position emission tomography (PET).^50–55 Therefore, metronidazole is a promising candidate as molecular probe for hypoxia imaging using HP MRI.⁴⁰