15N Hyperpolarization by Reversible Exchange Using SABRE-SHEATH.

NMR signal amplification by reversible exchange (SABRE) is a NMR hyperpolarization technique that enables nuclear spin polarization enhancement of molecules via concurrent chemical exchange of a target substrate and parahydrogen (the source of spin order) on an iridium catalyst. Recently, we demonstrated that conducting SABRE in microtesla fields provided by a magnetic shield enables up to 10% 15N-polarization (Theis, T.; et al. J. Am. Chem. Soc.2015, 137, 1404). Hyperpolarization on 15N (and heteronuclei in general) may be advantageous because of the long-lived nature of the hyperpolarization on 15N relative to the short-lived hyperpolarization of protons conventionally hyperpolarized by SABRE, in addition to wider chemical shift dispersion and absence of background signal. Here we show that these unprecedented polarization levels enable 15N magnetic resonance imaging. We also present a theoretical model for the hyperpolarization transfer to heteronuclei, and detail key parameters that should be optimized for efficient 15N-hyperpolarization. The effects of parahydrogen pressure, flow rate, sample temperature, catalyst-to-substrate ratio, relaxation time (T1), and reversible oxygen quenching are studied on a test system of 15N-pyridine in methanol-d4. Moreover, we demonstrate the first proof-of-principle 13C-hyperpolarization using this method. This simple hyperpolarization scheme only requires access to parahydrogen and a magnetic shield, and it provides large enough signal gains to enable one of the first 15N images (2 × 2 mm2 resolution). Importantly, this method enables hyperpolarization of molecular sites with NMR T1 relaxation times suitable for biomedical imaging and spectroscopy.

Introduction

Nuclear magnetic resonance (NMR) sensitivity can be enhanced through hyperpolarization by temporarily increasing the relatively low nuclear spin polarization (P = 10–6–10–5)—in some cases approaching unity—effectively providing 104–105-fold NMR signal enhancement.[1,2] Despite the short-lived nature of hyperpolarized (HP) spin states, with typical lifetimes on the order of seconds for 1H or minutes for heteronuclei (15N, 13C), the considerable sensitivity gain has led to many biomedical applications where a given HP compound serves as injectable or inhalable contrast agent.[3−5] Not surprisingly, biomedical applications have become the greatest driver of the field of HP MR, with the first clinical trial of an HP injectable 13C agent already completed in 2013.[6]

Currently the leading hyperpolarization method for preparation of HP contrast agents is dissolution dynamic nuclear polarization (d-DNP).[2] d-DNP enables the production of highly polarized (P > 0.2) 13C contrast agents used for in vivo real-time metabolic imaging,[7] disease grading,[8] monitoring response to treatment,[9] and other biomedical applications.[3,10] The two major disadvantages of d-DNP hyperpolarization are its relatively high cost and the complexity of the hyperpolarizer hardware (comparable to that of a PET cyclotron). Additionally, concerns persist about the relatively slow throughput of HP agents and the technique’s ultimate scalability.[4] An alternate hyperpolarization strategy of focus here uses the nuclear singlet state of parahydrogen (para-H2), which can be unlocked to yield up to near unity nuclear spin polarization—as demonstrated by Bowers and Weitekamp in their seminal work.[11] Originally, this method relied on the pairwise addition of para-H2 to an unsaturated chemical bond (frequently C=C or C≡C bonds); if the symmetry of the nascent parahydrogen protons is broken, the singlet state of para-H2 is transformed into observable magnetization.[12] Further advances in “ultrafast” hydrogenation (on the time scale of seconds) catalysis and synthetic chemistry led to the first demonstration of HP 13C in a contrast agent.[1] In such demonstrations, hyperpolarization was transferred to a 13C nucleus adjacent to the hydrogenated site via the network of spin–spin (J) couplings.[13−17] The HP 13C molecules produced by this approach have been efficiently used as HP contrast agents in vivo for plaque imaging,[18] TCA cycle fluxomics,[19] cancer imaging,[20−22] and detection of elevated glycolysis.[23−26] This method of hydrogenative parahydrogen-induced polarization (PHIP)[27] has significant advantages, including relatively low cost and high contrast agent production throughput.[4] Despite these advantages and other recent demonstrations,[1,14] traditional PHIP requires a sophisticated precursor molecule for pairwise para-H2 addition to yield a useful molecule. This limitation represents a major impediment for its widespread application in biomedicine.[3]

In 2009, Duckett, Green, and co-workers pioneered signal amplification by reversible exchange (SABRE).[28,29] Rather than irreversible pairwise addition of para-H2 (as required in traditional PHIP), SABRE involves rapid and reversible exchange of both substrate and para-H2 with sites on an Ir hexacoordinate complex. When this exchange is performed in a low magnetic field (e.g. 5–7 mT),[28−30] the spin order of the para-H2 singlet state can be spontaneously transferred into nuclear spin polarization of the Ir hydride protons and—more importantly—transferred all the way to the protons of a substrate molecule such as pyridine (Py). Although currently limited in the scope of applicable substrates, SABRE does not require hydrogenative chemical reaction and therefore can potentially be expanded to a broad range of molecular substrates. The SABRE technique has been applied to several biomedically relevant molecules using the established Ir catalyst precursor [IrCl(COD)(IMes)] (IMes = 1,3-bis(2,4,6-trimethylphenyl)imidazol-2-ylidene; COD = cyclooctadiene)),[31,32] including nicotinamide,[33,34] pyrazinamide,[35] isoniazid,[35] and others.[30,36] Furthermore, this exchange reaction leading to HP substrate has been demonstrated in 100% aqueous media[37] and with heterogeneous catalysts.[38,39] While further catalyst/substrate optimization can lead to biomedical magnetic resonance imaging (MRI) experiments[33,40−42] using aqueous solutions with sufficient payload (defined as the concentration of agent multiplied by P(43)), current SABRE approaches are only optimized for hyperpolarization transfer to protons;[34,41,44,45] however, the HP state of protons is relatively short-lived compared to that of 13C and 15N nuclei.[3,4] As a result, even if a relatively high payload of biocompatible 1H SABRE is generated, the agents can only be tracked for a short time because of significant losses during preparation, in vivo administration, circulation, tissue penetration, etc. Therefore, some recent efforts in the SABRE field are focused on developing polarization-transfer methods from HP proton sites in SABRE contrast agents to 13C[33,34] as well as the development of radiofrequency (RF)-based approaches for direct polarization of 15N sites with SABRE directly within the NMR or MRI scanner.[46,47] While these methods may potentially lead to high levels of hyperpolarization on long-lived heteronuclear sites in SABRE HP contrast agents, thus far they have yielded only relatively low %P ≪ 1%.[33,34,46]

We have recently presented a simple but effective strategy dubbed “SABRE-SHEATH” (Signal Amplification by Reversible Exchange in SHield Enables Alignment Transfer to Heteronuclei) for direct hyperpolarization of 15N during the SABRE process.[48] This method has a significant advantage compared to RF-based hyperpolarization methods because it does not require a strong magnetic field or sophisticated RF probes; i.e., the engineering and cost requirements are significantly reduced for biomedical applications where the goal is to cheaply and reliably produce batches of high-payload agents for in vivo administration. SABRE-SHEATH uses a simple magnetic shield[48] that enables spontaneous polarization transfer from hydride HP protons to 15N of molecules already amenable to SABRE.[33,35,37,48] In our first proof-of-principle experiments, P values of ∼10% have already been demonstrated on long-lived 15N sites of Py and ∼7% for nicotinamide (vitamin B3 amide).

In this article, we detail the theoretical underpinnings of the hyperpolarization transfer process and analyze the key experimental parameters influencing SABRE-SHEATH hyperpolarization, including para-H2 pressure and flow rate, sample temperature, and 15N T1 in microtesla magnetic fields. Our results strongly suggest that further improvements of the experimental setup, particularly those allowing higher para-H2 partial pressures, may further increase SABRE-SHEATH efficiency and improve the resulting heteronuclear polarization enhancement—potentially even approaching the theoretical limit of 100%. Finally, we demonstrate SABRE-SHEATH to produce one of the first 15N images (hyperpolarization of the same agent via d-DNP and subsequent imaging was also reported elsewhere during the review process[49]).

Experimental Methods

SABRE Catalyst and Sample Preparation

The SABRE samples were prepared by the addition of an aliquot of 15N-enriched or natural-abundance Py to a solution of catalyst precursor, resulting in desired concentrations of Py and catalyst. The SABRE catalyst employed for this study was created using the precursor [IrCl(COD)(IMes)],[31,32] synthesized previously.[40,44] The catalyst precursor activation was monitored via in situ detection of HP intermediate Ir-hydride species within a 9.4 T NMR spectrometer by proton NMR spectroscopy using the SABRE effect,[44] as described earlier.[37] Once fully activated, SABRE experiments in the magnetic shield (microtesla) or in the low magnetic field (fringe field of the 9.4 T magnet) of 6 ± 4 mT (Figure 1) were performed. Some samples were prepared by taking an aliquot of already-activated sample (Py with catalyst) and performing a series of dilutions (one to three) to achieve the desired concentrations. Perdeuterated methanol (methanol-d4) was used as a solvent for all experiments. >90% para-H2, prepared and bottled as described earlier,[50] was used for all experiments.

Figure 1

Diagram of the experimental setup. A) SABRE-SHEATH setup using a medium-wall 5 mm NMR tube and the magnetic shield; para-H2 gas is supplied from a high-pressure tank, regulated (via a flow meter), and bubbled through 1/16 in. tubing placed inside the 5 mm NMR tube under 1–7 atm para-H2 pressure. “Used” para-H2 gas leaves the NMR sample via an exhaust line. In situ SABRE (9.4 T) and low-field (ex situ) SABRE are performed using a 9.4 T NMR magnet and within its fringe field, respectively (B); all NMR detection was performed at 9.4 T. (C) The sequence of events for SABRE-SHEATH hyperpolarization.

Experimental SABRE Setup at 9.4 T

This setup was described in detail previously.[37,48] Briefly, a freshly prepared sample containing the Ir precursor catalyst and Py in CD3OD is placed inside a 5 mm medium-wall NMR tube (3.43 mm i.d.) for SABRE hyperpolarization. Normal H2 gas or para-H2 gas was bubbled through the methanol-d4 solution via 1/16 in. o.d. (1/32 in. i.d.) tubing inside the NMR tube as shown in Figure 1. A flow-meter[37] was used to regulate the gas flow at the rate of ∼1 mL·atm·s–1. The bubbling time and para-H2 pressure varied from ∼1 to 60 s and from 1 to ∼7 atm, respectively, for SABRE experiments. Activation by H2 bubbling used longer time periods and 1 atm pressure as described earlier.[37] For the in situ SABRE experiments at 9.4 T, para-H2 bubbling occurred inside the NMR magnet, and the signal was acquired (3 ± 2 s) after the bubbling was stopped. For low-field SABRE experiments and SABRE-SHEATH experiments, para-H2 bubbling occurred outside of the NMR detector (in the 6 ± 4 mT fringe field of the NMR magnet) or inside the magnetic shield, respectively (Figure 1). The sample was manually shuttled to the NMR magnet (9.4 T) into the detection coil, and the signal was recorded with typical transit times of 5 ± 2 s. The experimental setup was modified in studies of 15N SABRE-SHEATH signal dependence on the flow rate at four different pressure values by replacing the flow meter (Figure 1A) with a mass flow controller (Sierra Instruments, Monterey, CA, model no. C100L-DD-OV1-SV1-PV2-V1-S0-C0).

The NMR signal reference samples for 13C and 15N were loaded in standard 5 mm (4.14 mm i.d.) NMR tubes. All NMR experiments were conducted with a single-scan acquisition (90° excitation RF pulse) using 400 MHz Bruker Avance III spectrometer unless noted otherwise.

Calculations of NMR Polarization Enhancements at 9.4 T

Calculation of P enhancement (ε) and the %P were performed as follows: ε(1H) was calculated as the ratio of NMR signals from HP signal (SHP) vs thermally polarized equilibrium signal from the same sample at 9.4 T (STHER): ε(1H) = SHP/STHER. The equilibrium signal intensities for 15N and 13C samples were too low, and signal averaging was impractical due to excessively long (≥1 min) T1 values. Therefore, external signal reference samples of 12.5 M 15N-Py and neat methanol (24 M with ∼1.1% natural abundance of 13C isotope) were employed instead. Heteronuclear enhancement values were thus calculated as follows: ε(15N or 13C) = (SHP/SREF)(CREF/CHP)(AREF/AHP), where CREF and CHP are concentrations of reference and HP samples, respectively, SHP and SREF are integrated NMR signals of HP and references samples, respectively, and AREF and AHP are the corresponding cross-sectional areas of these solutions. The AREF/AHP ratio was ∼1.85, computed as 4.142/(3.432 – 1.62), where 4.14 mm is the inner diameter of the standard 5 mm NMR tubes used for NMR signal referencing samples, 3.43 mm is the inner diameter of the medium-pressure tubes used for SABRE samples, and 1.6 mm is the outer diameter of the 1/16 in. PTFE tubing inserted into the medium-wall NMR tube for para-H2 bubbling (note that (3.432 – 1.62) mm2 corresponds to the effective solution cross-section in the medium-wall NMR tubes used for SABRE experiments, in contrast to 4.142 mm2 used for signal reference samples). P was calculated as the following product: εPTHER, where PTHER is the thermal equilibrium nuclear spin polarization of 1H, 13C, or 15N nuclei at 9.4 T and 300 K (3.2 × 10–5, 8.1 × 10–6, and 3.3 × 10–6, respectively).

Theoretical Background

Previous work demonstrates transfer of hyperpolarization from para-H2 to heteronuclei via hydrogenative PHIP[11,27] at low magnetic fields[13,15−17,51,52] or by field cycling through low magnetic fields.[1,53,54] Recently, it has also become clear that similarly for the non-hydrogenative SABRE approach, RF-based methods or field cycling to very low magnetic fields can be employed to transfer hyperpolarization from para-H2 to heteronuclei.[28,29,46,47,55,100] Loosely speaking, for efficient hyperpolarization transfer, frequency differences between the para-H2-derived hydride protons and the to-be-polarized target nuclei have to match the J-coupling interactions that connect the polarization source and target nuclei.

In our case, efficient transfer of hyperpolarization from nascent parahydrogen protons of Ir-hydride to 15N-Py in a SABRE experiment is possible if the frequency difference between those Ir-hydride protons and 15N on the complex matches specific J-coupling terms, as displayed in Figure 2. As derived below, both of the following resonance conditions enable hyperpolarization transfer:[48,54]andwhere ΔνHN = νH – νN is the frequency difference between Ir-hydride protons and catalyst-bound 15N, and the J-couplings are as depicted in Figure 2.

Figure 2

Schematic representation of AA′BB′ spin systems. (A) Generalized representation of AA′BB′ showing relevant spin–spin couplings. (B) AA′BB′ spin system formed by two Ir-hydride protons and two 15N sites of two exchangeable 15N-pyridines shown in the structural diagram of the activated Ir-IMes catalyst using 15N-Py substrate.

The two hydride protons and two 15N nuclei within the two exchangeable substrates form an AA′BB′ spin system. Within this AA′BB′ system, the polarization transfer takes place. In general, the Hamiltonian of an AA′BB′ system is given asAll of its matrix elements were first listed by Pople et al. in 1957,[56] and the resonance conditions of eqs 1 and 2 have already been derived, in the context of hydrogenative PHIP.[54] The fundamental difference in our work will come from the role of exchange, which effectively broadens the matching condition.

To gain intuition about the polarization transfer dynamics, it is useful to embrace the vector representation of spin rotations. For example, in the traditional (one spin-1/2) vector representation, the position along +z corresponds to the spin state |α⟩ and the position along −z corresponds to spin state |β⟩. Thus, an initial spin state |α⟩ could be rotated by a Hamiltonian pointing along x as depicted in Figure 3a. This matrix element connects |α⟩ and |β⟩, transferring population in |α⟩ to population in |β⟩. The corresponding density matrix representation of initial state, ρi with population in |α⟩, x-Hamiltonian, , and final state, ρf with population in |β⟩, arewhere the important features are that populations are represented by real on-diagonal elements and the x-Hamiltonian is represented by real off-diagonal elements without contributions on the diagonal.

Figure 3

(a) Illustration of an initial state in the z-direction (population on the diagonal element) rotating about a Hamiltonian along x. This Hamiltonian has real positive off-diagonal and zero diagonal elements. (b,c) Illustrations of the SABRE-SHEATH hyperpolarization process. (b) Hyperpolarization transfer dictated by eq 5a. Here, the off-diagonal elements, −ΔJAB/2, are real and negative (isomorphic with −σ); hence, this part of the Hamiltonian is depicted along −x. The initial population of |S0AS0B⟩ on the diagonal is represented by a vector along +z. This is then rotated by the J-coupling term into a vector along −z representing population of the targeted state |T–AT+B⟩. (c) Hyperpolarization transfer dictated by eq 5b according to the same principles: Initial |S0AT–B⟩ population on the diagonal along +z is rotated into a population of |T–AS0B⟩ along −z by a Hamiltonian with real and positive off-diagonal elements, ΔJAB/2, represented along +x. In the diagrams, initial and final states are represented by the most faded and most solid vectors, respectively.

These notions are easily generalized to understand the hyperpolarization transfer process in AA′BB′ systems by choosing the right basis set. We follow the original work by Pople and co-workers,[56] and use the singlet–triplet basis applied to both the A spin pair and the B spin pair:

Combining the A states and the B states results in 4 × 4 = 16 total states (for example, the “singlet A–singlet B” state |S0AS0B⟩). Para-H2 is the prototypical singlet state and populates all four states that contain S0A. SABRE-SHEATH experiments work with |S0AS0B⟩ and |S0AT–B⟩, which are coupled, by the AA′BB′ Hamiltonian, to |T–AT+B⟩ and |T–AS0B⟩, respectively:where ΔJAB = JAB – JAB′ and ∑JAB = JAB + JAB′.

The form of eq 5a implies that population can be transferred from |S0AS0B⟩ to |T–T+B⟩ when the difference between the diagonal elements in that part of the Hamiltonian becomes small. For example, when the diagonal elements are equalized as −(JAA + JBB) = ∑JAB/2 – (νA – νB), then the off-diagonal elements can take full effect and rotate population from |S0AS0B⟩ to |T–AT+B⟩, as depicted in Figure 3b: The off diagonal elements −ΔJAB/2 are real and negative (isomorphic with −σ); hence, we depict them as a vector along −x. This process forms hyperpolarization in the T+ state of the targeted (B) spins corresponding to detectable magnetization. Hence, from eq 5a we deduce the first resonance condition given in eq 1 (by equalizing the diagonal elements). Similarly, eq 5b shows that hyperpolarization can be transferred from the |S0AT–B⟩ diagonal element to |T–AS0B⟩ when the diagonal elements in the Hamiltonian are equalized −νB – J =−νA – JBB, as illustrated in Figure 3c, establishing the second resonance condition given in eq 2. Here, hyperpolarization is observed because T–B is depleted, in effect creating overpopulation in T+B, just as predicted by the first resonance condition as well. Overpopulation in T+B corresponds to alignment with the main magnetic field only for species with positive γ (e.g., 13C); for species with negative γ (such as 15N), overpopulation in T+B corresponds to anti-alignment with the main magnetic field—in accordance with the experimental observations detailed in subsequent sections. Finally, in the probable scenario that JHH (∼9 Hz) dominates the J-coupling terms, all resonance conditions (i.e., those of eqs 1 and 2) are satisfied simultaneously, and we can estimate the optimal hyperpolarization transfer field as B0–transfer ≈ JHH/(γA – γB) ≈ 0.26 μT, assuming a J-coupling term (i.e., JHH) of ∼9 Hz (Figure 2).

In contrast to the PHIP case, the resonance conditions need not be met precisely, because continuous exchange of para-H2 and substrate reduce the residence times typically to about 0.2 s. For short times (relative to any resonance condition mismatch), the mismatch has only a modest effect on the population transfer. This fact implies that the effect of multiple exchanges will tend to equilibrate the populations of the states |S0AS0B⟩ and |T–AT+B⟩, and of the states |S0AT–B⟩ and |T–AS0B⟩. This assertion is also backed by our experimental observation that the specified resonance conditions are fairly loose. In other words, the specified matching conditions do not have to be met exactly; instead, if the magnetic field has the adequate order of magnitude we observe the desired effect. In this sense, the magnetic field simply has to be low enough, however “true” zero field would likely not produce the observed effects because a sufficient difference between T+ and T– states must prevail in order to create alignment along the residual magnetic field.

Results and Discussion

Catalyst Activation and Effect of O2

The Ir-IMes catalytic complex with Py substrate requires an initial activation with H2 to eliminate the COD moiety and other changes leading to formation of the hexacoordinate Ir-hydride complex. This catalyst activation was monitored by in situ SABRE at 9.4 T, i.e., by the in situ detection of intermediate hydride species during the activation process followed by the disappearance of these intermediates.[44] Completion of the catalyst activation process usually takes 10–20 min.[37] Figure 4A,B demonstrates the conventional low-field (achieved via para-H2 exchange at 6 ± 4 mT) SABRE proton NMR spectroscopy and corresponding NMR spectrum of the thermal reference. As indicated previously, 1H hyperpolarization levels are usually suppressed when using 15N-Py vs natural abundance Py.[48] Nevertheless, signal enhancements of ∼140 for ortho-protons of 15N-Py can be seen under the present conditions.

Figure 4

1H and 15N NMR spectroscopy of SABRE catalyst activation and 15N SABRE-SHEATH build-up. (A) 1H thermal NMR spectrum of 2 mM activated Ir-IMes catalyst solution with 48 mM 15N-pyridine. (B) 1H spectrum of hyperpolarized 15N-Py via conventional low-field (6 ± 4 mT) SABRE. The resonances labeled with dashed lines correspond to catalyst-associated Py.[58] (C–F) 15N NMR spectra of 15N-Py hyperpolarized by SABRE-SHEATH. (C) NMR spectrum of 15N-Py (εfree ∼ 300) sample corresponding to completely activated catalyst solution (as validated by 1H NMR using conventional low-field SABRE through achieving efficient enhancement of Py proton polarization, and also validated through in situ detection of the disappearance of SABRE hyperpolarized Ir-hydride intermediate species). (D) 15N NMR spectrum of 15N-Py sample corresponding to maximum SABRE-SHEATH signal intensity (εfree ≈ 3600) achieved with ∼20 min of para-H2 bubbling (with a ∼20% duty cycle, see text—para-H2 bubbling at this step was used for sample-degassing purposes; actual para-H2 bubbling for SABRE-SHEATH was only ∼30 s) after acquisition of the spectrum shown in C (but with the same para-H2 bubbling time of ∼30 s for SABRE-SHEATH hyperpolarization). (E) 15N NMR spectrum (εfree ≈ 185) of 15N-Py sample after it was exposed to air; the spectrum is recorded ∼23 min after spectrum shown in C. (F) 15N NMR spectrum of 15N-Py sample after SABRE-SHEATH intensities (εfree ≈ 3600) fully recovered from exposure to air; the spectrum was recorded ∼31 min after the spectrum shown in C.

Additional experiments exhibited an unexpected dependence on the presence of residual oxygen in the sample. When a SABRE-SHEATH experiment was attempted on a freshly (and fully) activated catalyst/15N-Py mixture (Figure 4C), only relatively modest 15N P enhancements (ε ≈ 300) were observed. Note that equilibrium 15N P is ∼10 times lower than that of 1H due to the large difference in their gyromagnetic ratios. Therefore, 15N P values achieved initially in SABRE-SHEATH experiments were significantly lower than the corresponding 1H P in conventional low-field SABRE experiments (e.g., Figure 4B)—highlighting the practical challenges of observing efficient SABRE-SHEATH hyperpolarization. Subsequent experiments showed that this reduced efficiency was related to the sample’s exposure to air—affording an opportunity to greatly improve the resulting 15N polarization. Indeed, additional para-H2 bubbling during the next 20 min (with a ∼20% duty cycle, i.e., with bubbling “on” for ∼30 s and “off” for ∼2 min at ∼6 atm of para-H2 at a flow rate of ∼1 mL·atm·s–1) resulted in significantly improved efficiency of the SABRE-SHEATH experiment. For example, Figure 4D shows a 15N NMR spectrum taken from a SABRE-SHEATH HP sample, with 15N P ε ≈ 3600 for this representative sample. Yet subsequent (re)exposure of this sample to air (via air bubbling for ∼5 s) resulted in significant reduction of SABRE-SHEATH efficiency to ε ≈ 185, Figure 4E. Nevertheless, renewed para-H2 bubbling (repeating the cycle described above a few times) resulted in a full recovery of the SABRE-SHEATH hyperpolarization efficiency, as depicted in Figure 4F. However, it should be pointed out that bubbling with para-H2 for an extended period of time does not appear to change the expected 15N hyperpolarization level for any reason other than removing oxygen: In fact, once the oxygen has been removed, the subsequent 15N SABRE-SHEATH experiments generate a reproducible level of 15N hyperpolarization, regardless of the magnitude of the delay (during which para-H2 bubbling was stopped) between the individual experiments (even when the delay was changed from 1 min to 1 h). Taken together, these results likely indicate that residual O2 in the catalyst/15N-Py methanol-d4 solution is significantly more detrimental to the efficiency of SABRE-SHEATH (performed in microtesla fields) compared to that of conventional low-field (1H, millitesla) SABRE. O2 is a well-known paramagnetic relaxation source that has been shown to significantly reduce T1 of HP 129Xe, particularly in low magnetic fields.[57]

Effect of Para-H2 Pressure and Flow Rate

The maximum achieved 15N P of 15N-Py in the presented experiments was ∼10%, corresponding to ε ≈ 30,000 at the time of signal detection in the 400 MHz NMR spectrometer (Figure 5A–C). The actual initial 15N polarization created within the shield prior to sample transfer is likely even higher, because of T1 relaxation losses suffered in transit (requiring ∼5 s). The respective effects of para-H2 pressure and flow rate could not be reliably discriminated using the flow meter[37] because its throughput (in mL·atm·s–1) is proportional to the product of volume and pressure per unit time. As a result, the same setting of the flow meter at two different pressures would result in two different mass flow rates measured in standard cubic centimeters per minutes (sccm). Therefore, a digital mass-flow controller regulating flow irrespective of bubbling pressure (see Experimental Methods) was utilized instead for these measurements, where the flow rate was varied at different pressures resulting in isobar curves, Figure 5D.

Figure 5

(A) Schematic of SABRE showing 15N-Py and para-H2 exchange on the activated Ir-IMes catalyst producing efficient 15N hyperpolarization. (B) 15N spectrum of HP 4 mM 15N-Py (0.24 mM catalyst) via SABRE-SHEATH procedure using ∼6 atm of para-H2 pressure.[48] (C) Corresponding 15N reference signal from neat 15N-Py. (D) 15N SABRE-SHEATH signal dependence on the para-H2 flow rate (sccm) at various para-H2 pressures: 1.0, 2.7, 5.1, and 7.1 atm. (E) 15N SABRE-SHEATH signal dependence on temperature at ∼6 atm para-H2 pressure. The data acquired in D utilized [catalyst]/[15N-Py] = 4 mM/96 mM. The data acquired in E utilized [catalyst]/[15N-Py] = 2 mM/48 mM. (F) 15N T1 measurements at three different magnetic field strengths: μT regime ([catalyst]/[15N-Py] = 0.2 mM/20 mM) in the magnetic shield, ∼6 mT fringe field ([catalyst]/[15N-Py] = 6 mM/63 mM), and 9.4 T ([catalyst]/[15N-Py] = 0.2 mM/20 mM) field inside the NMR spectrometer.

The effect of the pressure is negligible at flow rates of ≤10 sccm in the range of the para-H2 pressures studied, suggesting that in this regime the quantity of delivered para-H2 per unit of time was the limiting factor—not the para-H2 pressure per se (or by extension, the para-H2 concentration as dictated by Henry’s law). 15N SABRE-SHEATH (and thus 15N Pmax) generally rose with increasing flow rate (except for the abnormal trend at 2.7 atm at elevated flow rates discussed below). The growth of the 15N signal with increasing flow rate (and the quantity per unit time of delivered para-H2 acting as source of nuclear spin hyperpolarization) was most significant in the regime of low flow rates (∼20 sccm and 439 below, Figure 5D). Further increase in the flow rate (above ∼20 sccm) has a diminishing return, likely because hydrogen mass transport across the gas–liquid interface becomes a limiting factor; i.e., the introduced para-H2 gas is exchanging with the liquid less efficiently. This problem is especially well-illustrated in the 2.7 atm isobar, where the increase in the flow rate eventually leads to the abnormal decrease of the 15N NMR signal. It is likely that this behavior is largely caused by the gas bubble dynamics in the NMR tube, because the volume of the gas at low pressure (at the same flow rate in sccm) is greater, which resulted in larger bubbles and inefficient mixing of para-H2 gas and its transfer into the liquid phase. Indeed, the isobars at 1.0 and 2.7 atm produced lower signal at higher flow rates when compared to those at 5.1 and 7.1 atm. Thus, higher pressure is still more desirable to maximize 15N signal (and 15N P), but from the perspective of more efficient gas mixing and para-H2 delivery (moles per unit of time) to the catalyst (in this setup using para-H2 bubbling in the NMR tube). Therefore, future improvements in such bubbling SABRE-SHEATH hyperpolarization setups would likely benefit from operation at significantly higher para-H2 pressures to maximize the amount of para-H2 available for SABRE processes. Furthermore, generally speaking access of the Ir-hydride complex to para-H2 (i.e., delivery of sufficient quantity of para-H2 to Ir-hydride centers to maintain their HP state for further polarization transfer to heteronuclear sites of substrates) is one of the key determining factors for increasing the efficiency of SABRE-SHEATH and maximizing the payload of HP agent.

Effect of Temperature

The reaction temperature modulates the residence time of both the Py substrate and para-H2; i.e., it alters exchange rates. Correspondingly, previous studies of conventional low-field SABRE have observed temperature dependences of SABRE hyperpolarization levels.[30,35] Figure 5E demonstrates an explicit temperature dependence of 15N SABRE-SHEATH polarization, with greater signal enhancements measured at lower temperatures. While others have identified that for 1H SABRE the optimal performance for this catalyst peaks at >40 °C in methanol-d4,[35] the discrepancy likely can be explained by several factors, including that (i) higher para-H2 pressures (and consequently solution concentrations) were used here and (ii) the interactions—i.e., the relevant spin–spin couplings—leading to NMR hyperpolarization in homonuclear (for proton SABRE hyperpolarization) and heteronuclear (for SABRE-SHEATH 15N hyperpolarization) are different (i.e., J2HN and J3HH), possibly giving rise to correspondingly different optimal residency times for SABRE hyperpolarization.

Effect of Magnetic Field on 15N T1

T1 measurements were performed by inducing 15N SABRE-SHEATH hyperpolarization in the magnetic shield, which was followed by a variable delay for polarization decay in the magnetic shield (microtesla), in the fringe field of the main magnet (∼6 mT), or in the 9.4 T field of the magnet. Representative corresponding data sets showing the dependence of the NMR signal on the delay time at these fields are respectively provided in Figure 5F, exhibiting the overall trend: 15N T19.4T > 15N T16 mT > 15N T1μT.

Effect of 15N T1 in Microtesla Fields on %P

As indicated above the 15N T1 values were lowest in the μT field, which is important because T1 potentially modulates the build-up rate and the maximum attainable polarization. The latter is indeed the case, as can be seen in a dilution series (6 mM/100 mM, 1.2 mM/20 mM, and 0.24 mM/4 mM for the fixed [catalyst]/[15N-Py] ratio) shown in Figure 6A, where μT T1 gains are correlated with gains in Pmax (defined as 15N P after SABRE-SHEATH polarization accumulation for a period of time greater than 3T1, Figure 6E). Note that while further dilution leads to additional T1 gain (0.048 mM/0.8 mM), there are clearly other parameters (such as the residence times) that are affected by this extreme dilution, which results in significantly lower Pmax (Figure 6A). Figure 6B provides additional evidence showing the trends for Pmax and T1 for solutions with fixed [catalyst] and variable [15N-Py] (i.e., [catalyst]/[15N-Py] ratio was varied). While the μT T1 remains at approximately the same level of ∼10 s, the increase in [15N-Py] resulted in the corresponding decrease of Pmax (Figure 6F). Figure 6C,D shows the trends of Pmax and T1 as a function of [catalyst] at fixed concentrations of [15N-Py] of 100 mM and 20 mM, respectively. Here, multiple effects (T1 ∝ [catalyst]−1, Pmax ∝ [catalyst]/[15N-Py], and potentially others) cause changes in Pmax in a complex fashion. These examples are important, because they highlight the underlying factors that must be considered when attempting to maximize the hyperpolarization level Pmax, which is a key deliverable for biomedical applications using HP contrast agents.

Figure 6

Summary of 15N relaxation times T1 at the microtesla (μT) field inside the magnetic shield (T1μT) and at 9.4 T (T19T) and percentage 15N polarization (%P) for 15N SABRE-SHEATH of 15N-Py for various Ir-IMes catalyst and 15N-Py concentrations. The data are tabulated in Table S1. (A,E) Dilution series corresponding to ∼1:16 catalyst to 15N-Py ratio. (B,F) Series of catalyst/15N-Py solutions with fixed catalyst concentration. (C,D) Series of catalyst/15N-Py solutions with fixed 15N-Py concentrations at 100 mM and 20 mM, respectively. Note that data for the relaxation at 9.4 T were not measured for the first sample shown in A.

Feasibility of 13C SABRE-SHEATH

Hyperpolarization of aromatic 13C sites of 15N-Py via SABRE in general may offer some advantages compared to 15N including (i) greater 13C natural abundance vs 15N, (ii) more readily available detection hardware, (iii) better detection sensitivity, and others. However, since SABRE-SHEATH relies on the J-coupling between exchangeable protons of Ir-hydride and the target nucleus, its efficiency may be reduced because the requisite long-range (three-, four-, and five-bond) J-couplings are weak, Figure 2. Figure 7 shows the comparison between conventional (low-field) and SABRE-SHEATH hyperpolarization processes for 13C demonstrating that the expected antiphase signatures seen in low-field SABRE[34] can indeed be collapsed into in-phase NMR peaks (split by J1CH spin–spin couplings) by SABRE-SHEATH. However, the achieved 13C signal enhancements were marginal, i.e., ε = 7 or below, and by orders of magnitude lower than those for 15N. Furthermore, the enhancement values for three magnetically inequivalent carbon sites of the Py molecule were in accord with the range of J-coupling interactions between aromatic carbons and Ir-hydride protons, i.e., ε(J3) > ε(J4) > ε(J5) for ortho-, meta-, and para-positions, respectively. Thus far, 13C SABRE-SHEATH did not result in significant (≫10-fold) 13C P enhancements suitable for imaging and biomedical applications; however we are hopeful that an adequate field-cycling scheme for increasing 13C hyperpolarization levels may be found, and anticipate that double-resonance experiments (for polarization transfer from 1H or 15N to 13C) may also work to more efficiently hyperpolarize 13C spins via SABRE.

Figure 7

Comparison of 13C and 15N SABRE signal enhancements. (A) 15N SABRE using SABRE-SHEATH at μT field and (B) at ∼6 mT for a 63 mM 15N-Py sample with 6 mM of Ir-IMes catalyst. (C) Thermally polarized reference spectrum of 12.5 M 15N-Py used as the polarization enhancement reference for 15N SABRE. The intensity scale for the spectrum corresponding to the conventional (low-field) SABRE at ∼6 mT (shown in B) is zoomed in to twice the level of the μT SABRE 15N spectrum (shown in A), while the 15N-Py reference (shown in C) spectrum is zoomed in 12-fold. 13C SABRE was also conducted on the same sample at μT field (D) and at ∼6 mT (E). Neat methanol (24 M at ∼1.1% natural abundance of 13C) is used as the 13C polarization/signal reference (F). All of the 13C SABRE spectra are plotted on the same intensity scale. The polarization enhancements (ε) for selected peaks are shown for their respective spectra.

Feasibility of HP 15N MRI

The demonstrated %P of 15N by the SABRE-SHEATH method is better than or at least comparable to 15N hyperpolarization achieved by d-DNP[59−61] and by PHIP methods,[62] yet requires only seconds vs tens of minutes to hours of hyperpolarization time. The relatively easy access to 15N hyperpolarization by SABRE-SHEATH motivated a feasibility study of HP 15N MRI. A slice-selective 2D 15N MR imaging experiment was performed using a preclinical 4.7 T MRI scanner (Figure 8B). A modified setup was used to hyperpolarize ∼1.2 mL of solution containing 20 mM 15N-Py, corresponding to ∼24 μmol of 15N Py hyperpolarized to P of ∼1% at the time of the detection after a ∼30 s long transfer from the magnetic shield to the bore of the 38 mm i.d. 15N volume coil of a triple-resonance (1H/15N/31P) RF probe (Figure 8A). The modified setup utilized a high-pressure HPLC column (Western Analytical Products, Lake Elsinore, CA, USA, item no. 006SCC-06-15-FF; rated to ∼30 atm); with the bottom of the column directly connected to the para-H2 supply from the flow meter via 1/8 in. o.d. PTFE tubing. The top (exit) port was connected to the gas exhaust without any modification of the HPLC column. Para-H2 was bubbled through the standard unmodified filter of this HPLC column from the bottom port, and used H2 gas exited through the top of the HPCL column. It was possible to detect both free and exchangeable 15N Py resonances after >20 min long catalyst activation using ∼9.5 atm of para-H2 pressure. Moreover, an axial 2D 15N MR image with 2 × 2 mm2 spatial resolution was successfully acquired showing HP liquid placed inside a 6.6 mm i.d. high-pressure HPLC column. Note the partial volume effects in the voxels adjacent to the column wall, because the entire column i.d. was less than 4 pixels wide. To the best of our knowledge this image represents the first example of 15N MRI of HP compounds by parahydrogen-based hyperpolarization method (and indeed, we have been unable to find any previous example of 15N MRI in the literature, hyperpolarized or otherwise, except for a very recent d-DNP hyperpolarization study of 15N-Py[49]). Furthermore, this image was acquired in ∼0.4 s, demonstrating the feasibility of subsecond 15N MRI of HP contrast agents. To enable future in vivo experiments, optimization may lead to further gains in %P at higher substrate concentrations (i.e., improving the payload of net magnetization from a 24 μmol dose at %P ≈ 1%). We also note that the detection sensitivity of 15N HP compounds can be further enhanced by implementation of MRI pulse sequences yielding more SNR (e.g., balanced steady-state free precession (bSSFP)[21,63,64]). The long-lived 15N hyperpolarization can potentially also be transferred to protons for more efficient indirect detection.[65,66] This practice could boost sensitivity by approximately 10-fold, because the detection sensitivity is directly proportional to the gyromagnetic ratio γ and γ(1H) ≈ 10γ(15N). While indirect proton imaging may be challenging due to the background signal arising from water protons in vivo,[67] low-field HP MR, which can be more sensitive than high-field MRI of HP agents,[68] does not suffer from this shortcoming because at very low fields (e.g., 47 mT) the proton background signal is significantly attenuated.[41] Additionally, polarization transfer schemes that enable indirect proton detection of 15N HP contrast agents[69,70] can enable partial polarization transfer from 15N (storage nucleus with high T1) to 1H (detection nucleus with best read-out sensitivity) to enable acquisition of multiple ultrafast (∼<1 s) images to trace not only the distribution of the contrast agents, but also the kinetics of uptake, metabolism, and other in vivo processes.

Figure 8

15N NMR spectroscopy and MR imaging of HP 15N-Py at 4.7 T. (A) 15N single-scan NMR spectroscopy of a thermal reference sample of 15NH4Cl in an aqueous medium and HP 15N-Py at 20.3 MHz using the following acquisition parameters: RF pulse width (pw) = 128 μs (90°), spectra width (sw) = 19 840 Hz, acquisition time (acq) = 0.5 s. (B) 15N 2D projection gradient echo (GRE) MRI using the following acquisition parameters: slice thickness = 60 mm, pulse width = 500 μs (∼15°), field of view = 64 × 64 mm2, imaging matrix size = 32 × 32 pixels, pixel size (spatial resolution) = 2 × 2 mm2, repetition time (TR) = 13 ms, echo time (TE) = 6.4 ms, acq = 10.6 ms, sw = 3005 Hz, and total scan time ∼0.4 s. The image was post-processed with zero filling to 256 × 256 points for enhanced presentation.

Outlook for Biomedical Translation

Py and other aromatic N-heterocycles, which are already amenable to SABRE hyperpolarization,[33−35] represent the fundamental molecular frameworks for many classes of biologically relevant compounds: DNA and RNA bases, vitamins, and numerous drugs and drug building blocks. Therefore, a number of potential HP contrast agents can be envisioned, where N sites amenable to 15N SABRE-SHEATH can serve as hyperpolarization storage sites for imaging in vivo processes. For example, nicotinamide is linked to many diseases including Alzheimer’s disease, cancer, and anxiety, and therefore can be a potentially useful HP probe for these diseases. Importantly, Py-based HP 15N agents have been shown useful for pH sensing using the d-DNP hyperpolarization method.[49] This capability is enabled by the large 15N chemical shift change (>70 ppm) induced by protonation,[49,71] which causes the 15N chemical shift to have a straightforward dependence on pH.[49] Non-HP agents using the same principle have already been successfully demonstrated in vivo, with the most prominent application in cancer imaging, because many types of tumors are known to be slightly acidic.[72−74] Therefore, non-invasive pH sensing and mapping are potential biomedical applications for 15N SABRE-SHEATH in the context of molecular in vivo imaging, although clearly much work remains to expand this method to other agents beyond 15N-Py (studied here) and the recently demonstrated 15N-nicotinamide.[48] Nevertheless, 15N-SABRE-SHEATH has already been shown to produce 15N-Py and 15N-nicotinamide[48] with ∼6 times greater hyperpolarization levels compared to those achieved by d-DNP[49] with a much more rapid hyperpolarization process (∼1 min vs ∼2 h),[49] highlighting the practical advantages of 15N SABRE-SHEATH for this class of compounds.

In contrast to 13C HP agents (e.g., the leading agent 1-13C-pyruvate) and in contrast to 15N-choline, the 15N-isotopic enrichment of heterocycles based on Py is fairly straightforward, starting from relatively inexpensive 15NH4Cl (<$20/g).[75−78] Therefore, the 15N isotopic enrichment required for 15N SABRE-SHEATH is likely to produce relatively inexpensive contrast agents. Combined with the very simple setup and instrumentation required for SABRE-SHEATH, it may enable fast, high-throughout, scalable, and low-cost production of HP 15N contrast agents.

In vivo use of HP contrast agents, including 15N agents,[60] typically requires their administration in aqueous media free from catalysts and activation byproducts. Relevant to this work, the recent reports of heterogeneous SABRE catalysts,[38,39] continuous SABRE hyperpolarization[42,79] and aqueous SABRE catalysis[37] are certainly highly synergistic with SABRE-SHEATH for production of HP contrast agents on demand with suitable in vivo administration. Moreover, heterogeneous SABRE catalysts[38,39] provide an additional benefit of potential catalyst recycling to further minimize the costs associated with catalyst preparation and waste disposal.

While 15N detection hardware and MRI pulse sequences are not commonplace in clinical settings, indirect proton detection can be used instead for the detection of 15N HP contrast agents produced by SABRE-SHEATH.[65−67,80] Furthermore, more advanced indirect-detection methods do not require any heteronuclear hardware potentially allowing RF excitation and detection to be performed only on the proton channels commonly available on all MRI scanners.[81] The latter characteristic thus obviates the requirement for dedicated 15N channels on MRI scanners and can potentially enable widespread use of 15N HP contrast agents on conventional MRI scanners—requiring only a software upgrade.

Conclusions

15N P of 10% was successfully demonstrated by SABRE-SHEATH in 15N-Py, corresponding to signal gains of 30,000-fold at 9.4 T. Key parameters affecting the efficiency of 15N SABRE-SHEATH hyperpolarization processes were studied. Avoiding reversible O2 contamination of the activated Ir-based SABRE catalyst was found to be critical for establishing maximum SABRE-SHEATH HP in a microtesla magnetic field. 15N T1, para-H2 pressure and flow rate, catalyst-to-Py ratio, and Py and catalyst concentrations strongly influence the level of 15N polarization achieved by SABRE-SHEATH, but can be optimized to maximize payload (the product of 15N P and Py concentration) of 15N hyperpolarization for biomedical and other applications. While the feasibility of 13C SABRE-SHEATH was demonstrated, the signal enhancements remain low at this point (i.e., ε < 10 at 9.4 T) likely because of weak scalar couplings between the relevant nuclei and the absence of an adequate field-cycling route.

15N SABRE-SHEATH not only allows for preparation of very high heteronuclear P levels, but the manifested enhanced 15N NMR signals are in-phase—and therefore suitable for MRI imaging. The feasibility of high-resolution sub-second 2D 15N MRI was demonstrated, here using a preclinical MRI scanner and RF coil to yield 2 × 2 mm2 spatial resolution with a 20 mM 15N-Py solution and modest hyperpolarization of ∼1%. This detailed study of SABRE-SHEATH can potentially enable hyperpolarization and molecular imaging of other SABRE-amenable compounds, including nicotinamide[48]—paving the way to biomedical and other uses of SABRE-SHEATH 15N hyperpolarization.