Hyperpolarization of "Neat" Liquids by NMR Signal Amplification by Reversible Exchange.

We report NMR Signal Amplification by Reversible Exchange (SABRE) hyperpolarization of the rare isotopes in "neat" liquids, each composed only of an otherwise pure target compound with isotopic natural abundance (n.a.) and millimolar concentrations of dissolved catalyst. Pyridine (Py) or Py derivatives are studied at 0.4% isotopic natural abundance ¹⁵N, deuterated, ¹⁵N enriched, and in various combinations using the SABRE-SHEATH variant (microTesla magnetic fields to permit direct ¹⁵N polarization from parahydrogen via reversible binding and exchange with an Ir catalyst). We find that the dilute n.a. ¹⁵N spin bath in Py still channels spin order from parahydrogen to dilute ¹⁵N spins, without polarization losses due to the presence of ¹⁴N or ²H. We demonstrate P(15N) ≈ 1% (a gain of 2900 fold relative to thermal polarization at 9.4 T) at high substrate concentrations. This fundamental finding has a significant practical benefit for screening potentially hyperpolarizable contrast agents without labeling. The capability of screening at n.a. level of ¹⁵N is demonstrated on examples of mono- and dimethyl-substituted Py (picolines and lutidines previously identified as promising pH sensors), showing that the presence of a methyl group in the ortho position significantly decreases SABRE hyperpolarization.

Nuclear spin polarization can be temporarily enhanced by >4 orders of magnitude through the process of hyperpolarization with corresponding improvements in sensitivity or signal-to-noise ratio (SNR).[1−3] This significant sensitivity gain can also be used to decrease NMR acquisition time or analyte concentration,[4] but more importantly hyperpolarization techniques enable preparation of large batches of hyperpolarized (HP) compounds, which can be used as contrast agents[5] for various applications involving dynamic tracking in living organisms, including functional pulmonary imaging with HP 129Xe,[3,6] metabolic cancer imaging with HP 13C agents,[7,8] and many others.[9] Advances in hyperpolarization techniques have enabled production of HP agents with sufficient payload (the product of nuclear spin polarization P and concentration/quantity) for in vivo applications and ultimately for human clinical trials, as demonstrated using HP 13C-pyruvate for prostate cancer detection[10] in men and HP 129Xe for lung imaging.[6]

Signal Amplification by Reversible Exchange (SABRE)[11,12] is a particularly cost-efficient and fast hyperpolarization method that relies on exchange of a to-be-hyperpolarized substrate and parahydrogen (para-H2) on a catalyst. Similarly to conventional parahydrogen-induced polarization (PHIP),[13] SABRE utilizes para-H2 as the source of spin order,[14] and the entire hyperpolarization procedure can be completed in seconds; however, unlike PHIP, it does not require the irreversible chemical modification of the substrate.[11,12] Until recently, demonstrations achieving efficient hyperpolarization via SABRE were generally limited to protons; while in some cases the resulting 1H polarization values were relatively high (e.g., P ≈ 8%[15]), such nonequilibrium polarization is relatively short-lived (T1 of seconds). Recent approaches to extend SABRE to longer-lived (T1 ≈ 1 min) 15N hyperpolarization include LIGHT-SABRE (Low-Irradiation Generation of High Tesla-SABRE)[16] and SABRE-SHEATH (SABRE in SHield Enables Alignment Transfer to Heteronuclei)[17,18] using RF-irradiation-based and field-cycling-based approaches, respectively. SABRE-SHEATH[17−19] is an advantageous approach because it only requires that the exchange reaction with para-H2 be performed in a microTesla field. This condition can be created easily by shielding the Earth’s magnetic field using a mu-metal chamber and is therefore simple and inexpensive. 15N polarization levels of up to 10% were shown in the proof-of-principle demonstration.[17] However, such hyperpolarization was achieved in dilute (4–45 mM) alcohol solutions; that is, the payload (the product of concentration and polarization) of the HP agents was not optimized, and alcohol solutions have limited biocompatibility.

Here, we demonstrate the feasibility of SABRE-SHEATH hyperpolarization of “neat” liquids—each comprised only of an otherwise pure target compound and millimolar concentrations of dissolved catalyst, without any additional diluting solvent. In principle, such liquids could be used directly as hyperpolarized MRI contrast agents; the use of organic solvents is obviated, and we observe greater payload for the concentrated agents.

The previously developed setup for SABRE-SHEATH was utilized,[17,20] wherein para-H2 is bubbled through a liquid agent (e.g., pyridine (Py) or others) containing an activated catalyst (formed from the precursor: [IrCl(COD)(IMes)] (IMes =1,3-bis(2,4,6-trimethylphenyl)imidazol-2-ylidene; COD = cyclooctadiene)).[15,21] To date, this is the best catalyst for SABRE exchange processes involving para-H2 and the substrate (e.g., Py) shown schematically in Figure 1 (this figure also describes the sample manipulation during 15N SABRE-SHEATH process). Proton hyperpolarization via SABRE at milliTesla magnetic fields (6 ± 4 mT) was performed as described previously,[20,22] with para-H2 bubbling performed in the fringe field (6 ± 4 mT) of the 9.4 T magnet (instead of steps 1 and 2 shown in Figure 1A). See the Supporting Information (SI) for additional experimental details. We note that SABRE of protons is most efficient in the milliTesla regime, whereas SABRE of 15N spins is most efficient in microTesla fields,[17] as dictated by the respective matching conditions requiring that the frequency differences between the source and target spins be reduced to roughly match the scalar couplings between them.

Figure 1

(A) Schematic of the 15N SABRE-SHEATH hyperpolarization process conducted in a pressurized (1–7 atm) 5 mm NMR tube with para-H2 bubbling.[20] (B) Schematic of the SABRE exchange process, wherein parahydrogen and a substrate, for example, pyridine, exchange on activated Ir-hydride complexes.[15]

When neat natural abundance (n.a.) Py was used for 15N SABRE-SHEATH, large 15N signal enhancements (ε up to 2900) were observed, corresponding to P15N ≈ 1% (Figure 2A). Remarkably, 1H SABRE (conducted conventionally at ∼6 mT field[15]) yielded very small signal enhancement of ε ≈ 4 (Figure 2B). The 15N signal exhibited a strong, nearly linear dependence on the flow rate of para-H2 in the range studied (the flow-rate of 150 standard cubic centimeters (sccm) represents an experimental limitation of our setup at ∼7 atm), which was metered independently of the applied pressure and hence solution para-H2 concentration (Figure 2C). Note that the 15N signal enhancement was approximately independent of the para-H2 pressure (and solution concentration according to Henry’s law), indicating that the flux of the available para-H2 spin bath (the source of spin order) was indeed the limiting factor; that is, the potential possibility of exchanging more para-H2 per unit time would likely yield greater 15N signal enhancements. Larger para-H2 exposure can be attained by higher pressures and smaller bubbles/better gas-phase–liquid-phase mixing.

Figure 2

SABRE of “neat” natural abundance 15N (0.36%) pyridine (Py). (A) 15N SABRE-SHEATH hyperpolarized spectrum (red) and the corresponding thermally polarized reference spectrum (blue) after 192 signal averages. (B) 1H SABRE spectrum of hyperpolarized (red) sample in milliTesla magnetic field (∼6 mT) and the corresponding NMR spectrum using thermally polarized sample (blue). (C) Effect of the para-H2 flow rate (measured in standard cubic centimeters per minute or sccm) on 15N signal enhancement at ∼90 mM catalyst concentration under five para-H2 pressure values. (D) Effect of [Py] to [catalyst] ratio on 15N signal enhancement using 120 sccm flow rate under ∼7 atm of para-H2 pressure. (E) 15N SABRE-SHEATH dependence (modeled as exponential decay) as a function of the sample exposure to the microTesla magnetic field after stopping para-H2 bubbling time. (F) 15N T1 decay at 9.4 T. The experiments in panels E and F are conducted using ∼90 mM catalyst concentration (∼140:1 [Py] to [catalyst] ratio) at 120 sccm flow rate and ∼7 atm para-H2 pressure.

The other important effect limiting the maximum achievable hyperpolarization is spin–lattice relaxation. The 15N spin–lattice relaxation time is significantly shorter in microTesla fields than at high field (9.4 T), 5.5 ± 0.5 versus 60.8 ± 0.6 s, respectively (Figures 2E,F), and such efficient relaxation results in SABRE-SHEATH 15N enhancements reaching significantly lower steady-state levels after the hyperpolarization procedure. More importantly, the supply of para-H2 is limited because only ∼0.1 mmol/s pass through the tube at the maximum flow rate of 150 sccm, whereas 90 mM catalyst (in ∼0.4 mL volume) alone is capable of exchanging of ∼0.2 to 0.4 mmol/s of H2 because the hydrogen exchange rate is ∼5–10 per s.[23] However, Ir-hydride protons do not have 100% exchange efficiency with para-H2 gas. Instead, this exchange is further constricted by at least two major bottlenecks: (i) exchange of H2 between gas and liquid phases and (ii) exchange of dissolved para-H2 with Ir-hydride. Note that equilibrium H2 concentration in organic solvents is <4 mM/atm;[24−26] that is, even at the maximum para-H2 pressure used (∼7 atm), para-H2 concentration is <30 mM, that is, at least three times lower than that of the Ir-hydride catalyst at 90 mM concentration. Moreover, when para-H2 singlet spin order is transferred to Py via SABRE, para-H2 becomes ortho-H2, manifesting as an HP byproduct,[22] and this resulting ortho-H2 can no longer serve as a source of hyperpolarization in conventional ex situ SABRE. Furthermore, hydride proton exchange rates are on the order of 10 per second;[23] therefore, each para-H2 molecule on average experiences >30 exchanges per second under these conditions ([catalyst] of ∼90 mM results in the total of ∼900 para-H2 exchanges per second for <30 mM [para-H2] dissolved). The main implication of the above two bottlenecks, the fast hydrogen exchange and the limited flux of para-H2 gas, is that [ortho-H2] ≫ [para-H2]. Furthermore, the additional feature of the complex interplay of microTesla 15N effective T1 and limited access to para-H2 is that it should imply the existence of an optimal catalyst concentration and an optimal ratio of Py to catalyst concentrations. Taken together, these results indicate that greater signal enhancements are potentially feasible, provided that engineering issues limiting para-H2 access for SABRE hyperpolarization are solved.

The additional evidence that the finite para-H2 spin bath is limiting the SABRE processes is also seen when n.a. Py (ε ∼ 2900) was replaced by 99% 15N enriched Py (15N-Py, ε ≈ 33), Table 1. 15N signal enhancement decreases by nearly 2 orders of magnitude (88-fold), while the concentration of 15N spins is increased by 278 fold (= 1/0.0036); however, note that the total Py concentration and quantity is maintained the same. In summary, the observed signal (given by the product of [15N] and ε) only decreases by 3 fold when working with n.a. Py. Another important aspect in this context is that 15N microTesla effective T1 of 15N-Py (10.2 ± 1.1 s) is actually longer than that of n.a. Py (5.5 ± 0.5 s); see Table 1.

Table 1

Summary of Experimental Results with Natural Abundance (n.a.) Pyridine (Py), Py-d5, 15N-Py, and Their Mixturesa

	[15N] (mM)	15N ε @ 9.4T	15N effective T1 μT (s)	15N T1 9.4T (s)	ortho-[1H] (mM)	1H ε @9.4T	[catalyst] (mM)
1) Py (n.a.)b	∼45	∼−2900	5.5(0.5)	60.8(0.6)	∼25000	∼−4.2	∼90
2) Py-d5 (99.5% d)	∼45	∼−850	2.2(0.1)	74.3(2.9)	∼125	∼−60	∼90
3) 15N-Py	∼12500	∼−33	10.2(1.1)	66.8(0.5)	∼25000	∼−0.3	∼90
4) catalyst activated with 15N-Py, then Py-d5 added	∼2000	∼−520	10.1(0.8)	69.9(0.3)	∼4000	∼−2.6	∼90
5) catalyst activated with Py-d5, then 15N-Py is added	∼1800	∼−400	15.1(2.3)	73.2(0.3)	∼3600	∼−2.7	∼90
6) catalyst activated with 15N-Py, then in n.a. Py is added	∼1800	∼−450	9.9(1.1)	70.0(0.3)	∼3600	∼−1.0	∼90
7) catalyst activated with n.a. Py, then 15N-Py is added	∼1800	∼−380	8.2(1.1)	69.9(0.3)	∼3600	∼−0.6	∼90

Note the activation sequence, which determines the axial non-exchangeable ligand of the activated Ir-hydride catalyst complex. See the SI for details.

Conducted with >90% para-H2, while the rest of the data is collected using 65–75% para-H2, resulting in ∼30–40% lower signal enhancements compared with those shown in row 1. Note that the data for pairs 2 and 3, 4 and 5, and 6 and 7 were respectively collected on the same day at the same level of para-H2 enrichment and stored in a pressurized aluminum cylinder as previously described (and thus should be directly comparable).[40]

Furthermore, achieving such significantly greater (by 88-fold) 15N ε in n.a. Py with respect to 15N-Py under the conditions of limited access to para-H2 has a major significance for the mechanistic understanding of the SABRE-SHEATH phenomenon. In particular, this result indicates that the hyperpolarization para-H2 spin bath is not depleted when the exchanging substrate on Ir-hydride catalyst is 14N-Py. If no interaction between para-state of hydride and 15N-Py occurs (i.e., the exchanging partner is 14N-Py), para-state of hydride should exchange back into para-H2 with preservation of the para-H2 hyperpolarization pool (Figure 4B). This is an important conclusion because the spin order residing in the entire pool of para-H2 can be selectively channeled to hyperpolarize 15N nuclei of the exchangeable substrate (e.g., n.a. Py) rather than (say) being depleted by rapidly relaxing 14N sites acting as hyperpolarization sinks. This finding fundamentally enables achieving relatively high levels of 15N hyperpolarization (e.g., P15N ≈ 1%), even when performing SABRE-SHEATH in the high substrate concentration regime encountered with effectively neat solutions and when the supply and transport of para-H2 are restricted. The 14N species likely do not deplete the para-H2 state because the quadrupolar relaxation rate of the 14N spins is faster than the J-coupling interactions that would otherwise transfer hyperpolarization to the target spins; hence, the 14N spins are effectively (self-)decoupled from the bound para-H2.

Figure 4

Diagrams of para-H2 exchange and 15N SABRE-SHEATH hyperpolarization in the absence (A) and in the presence (B) of 14N-Py excess. Note that the exchange with 14N-Py does not cause a significant reduction in the spin order of the para-H2 pool. It should also be noted that both equatorial pyridines of the active complex undergo the chemical exchange with free Py in solution, while the axial pyridine (labeled as “Py”) is not exchangeable.

Our previous theoretical model of SABRE-SHEATH, while appropriate for 15N-enriched substrates, no longer applies for n.a. Py, and hence we present an amended theoretical model to describe the polarization transfer in the n.a. case. The original model invokes an AA′BB′ four spin system, where AA′ represents the parahydrogen-derived hydrides and BB′ represents the equatorial (exchangeable) 15N spins depicted in Figure 3A. For this case, we originally derived that the magnetic field must be chosen such that at least one of the following resonance conditions are met

Figure 3

Spin systems used for analytical derivation of the resonance conditions for (A) 15N-Py solutions and (B) n.a. Py solutions. In panel A, in addition to the displayed couplings, JHN = JH′N′ and JHN′ = JH′N. Couplings to spins in axial positions are ignored because they generally are smaller than equatorial couplings and play a subordinate role. (Additionally, this site does not exchange with free substrate.)

When these resonance conditions are met, then the N–H J couplings drive the hyperpolarization transfer; specifically, the term (JHN – JHN′)/2 determines the rate of hyperpolarization transfer.

However, in the n.a. Py case, this spin system has to be adjusted because in 99.64% (= 100 – 0.36%) of species that contain one 15N spin the adjacent equatorial species is a 14N spin, not 15N; therefore, we change our model to an AA′B three-spin system, where AA′ represents the parahydrogen-derived hydrides (as before) and B represents the 15N spin. The 14N spin can be ignored because as previously mentioned the strong quadrupolar interaction decouples the 14N spin from the depicted spin systems.

In the SI we show that for the three-spin system the resonance condition is

In the three-spin system it is also the NH-J couplings that drive the hyperpolarization transfer; here it is specifically the term (J – J)/(2√2), which determines the rate of hyperpolarization transfer. Full derivation and additional details are given in the SI.

Next, we also performed conventional homonuclear 1H-SABRE experiments. The 1H signal enhancements, which are optimized in the milliTesla regime (Table 1) followed the general trend seen for 15N SABRE-SHEATH, with signal enhancements being greater when the proton spin bath of to-be-hyperpolarized substrate was reduced. For example, ε ≈ (−)60 was observed for Py-d5 versus ε ≈ (−)4.2 for n.a. Py, which is in agreement with previous studies.[15]

Because 14N and other quadrupolar nuclei could have the potential to act as direct or indirect hyperpolarization sinks (e.g., polarization transfer from Ir-hydride protons to 14N, D, etc. or from 15N (after hyperpolarization transfer from para-H2) to 14N, D, etc.) at such low magnetic fields (analogous to interaction between 129Xe and 131Xe in xenon lattices[27]), and because the local molecular environment can significantly alter the 15N effective T1 in the microTesla field regime, 15N SABRE-SHEATH of deuterated Py (Py-d5) was studied as well as various mixtures of 15N-Py and Py-d5 with 15N-Py and n.a. Py (Table 1). Note that the Py type (i.e., n.a. Py, Py-d5, or 15N-Py) used during the activation period determines the spin configuration of Py in the axial nonexchangeable position of the hexacoordinate Ir-hydride complex, whereas the abundance of the Py type in the mixture determines the most probable type of exchangeable Py in the two equatorial positions. (The corresponding most-probable configurations are summarized in the SI.) Deuteration of to-be-polarized 15N-substrate had the most detrimental effect on microTesla 15N effective T1, a decrease from 5.5 ± 0.5 to 2.2 ± 0.1 s for n.a. Py versus Py-d5 (row 1 vs row 2 of Table 1). A similar but slightly larger decrease (from ε ≈ (−)2900 to (−)850) was observed for the corresponding SABRE-SHEATH 15N enhancement values, indicating that the majority of deuterium-induced depolarization is due to indirect transfer, for example, from 15N to 2H.[28] However, the direct depolarization losses are likely to have a significant contribution as well. For example, in cases when nondeuterated 15N-Py was used in combination with Py-d5, microTesla 15N effective T1 is actually greater when the catalyst is first activated with Py-d5 versus that when catalyst is first activated with 15N-Py, 15.1 ± 2.3 versus 10.1 ± 0.8 s (remembering that the activation order defines the nonexchangeable ligand in the axial position), but the 15N signal enhancements were somewhat lower, ε ≈ (−)400 vs (−)520, indicating that at least some polarization losses occurred on the hyperpolarized Ir-hydride due to the presence of deuterium in the catalyst structure. This particular finding at first seems to contradict recent studies where deuterium was incorporated in both catalyst[29] and exchangeable substrate[15,30] in the original homonuclear SABRE but can be explained by the different field regimes (microTesla vs several milliTesla) involved and by the fact that the previous efforts involved hyperpolarization of protons in exchangeable substrate instead of the present focus on 15N nuclei polarized using 15N SABRE-SHEATH.[17]

The effect of 14N presence in the catalyst structure as a potential relaxation or polarization sink was studied by comparing two samples prepared using a mixture of 15N-Py and n.a. Py (consisting mostly of 14N-Py), rows 6 and 7 of Table 1. Activation of SABRE catalyst with 15N-Py versus n.a. Py resulted in a slight increase in the microTesla 15N effective T1 (9.9 ± 1.1 s vs 8.2 ± 1.1 s) as well as the 15N signal enhancement (ε ≈ (−)450 vs (−)380), indicating that 14N presence indeed can act as a weak relaxation or polarization sink, likely through contributions from both mechanisms; that is, direct transfer from hyperpolarized Ir-hydrides and from exchangeable 15N-Py. To summarize, the above evidence advocates for avoiding the utilization of quadrupolar nuclei (e.g., deuterium and 14N studied here) for 15N SABRE-SHEATH hyperpolarization processes, whose presence can result in reduced hyperpolarization in microTesla fields.

As previously described, the 15N SABRE-SHEATH of neat liquids is an advantageous tool for efficient hyperpolarization of 15N spins, particularly at their low natural abundance level. One potential use is for rapid compound screening, which we demonstrate here on a series of picolines and lutidines shown in Figure 5. We find that the presence of a methyl group in position 2 or 6 results in no detectable 15N hyperpolarization via SABRE-SHEATH, whereas the substituents in other positions result in 15N signal enhancements levels similar to those of Py. The obvious explanation is that steric hindrance induced by the presence of methyl groups in ortho positions significantly alters the time scale of the SABRE exchange process or reduces the association constant.

Figure 5

Maximum 15N SABRE-SHEATH signal enhancements obtained for pyridine, picolines, and lutidines in neat liquids using ∼45 mM catalyst concentration and naturally abundant levels of 15N (∼0.35%) under ∼7 atm of para-H2 pressure and flow rate of 100–120 sccm. Note that the value labeled with a single asterisk (*) corresponds to the optimized catalyst concentration of ∼90 mM, the values labeled with double asterisks (**) correspond to the experiments conducted at 5 atm of para-H2 and the flow rate of 60 sccm, and n.d. stands for none detected.

We chose to examine picolines and lutidines because it was previously shown that pH-mediated protonation of N-heterocylic compounds can be useful for in vivo pH imaging using conventional proton-based nonhyperpolarized sensing, where the difference in 15N chemical shift induced by the agent protonation can be useful for pH imaging provided that the agent’s pKa is in the physiologically relevant range.[31,32]15N centers of the Py class screened here were also identified as promising hyperpolarized pH sensors with potential biomedical application to noninvasively image local variances in tissue pH.[33] Unlike previously demonstrated pH imaging with hyperpolarized H13CO3–/13CO2 that relies on the measurement of the ratio of two exchanging species,[34] pH imaging using hyperpolarized 15N heterocycles relies on the modulation of 15N chemical shift, which changes by up to 100 ppm between protonated and deprotonated states.[33,35] This feature offers a significant sensitivity advantage because only one species requires detection (i.e., ratiometric measurements are not needed), and low signal-to-noise ratio would not affect the accuracy of the measurement because the chemical shift reports on the pH. Moreover, hyperpolarized 15N sites have significantly longer T1 in aqueous media (>30 s)[33] compared with 13C bicarbonate (∼10 s),[34] which can also be a significant advantage for in vivo applications (especially relevant for applications involving cancer, given the known hallmarks of elevated glycolysis and mildly acidic microenvironments).[36,37] We note that the 15N signal enhancements reported in Figure 5 are obtained in a nonoptimal setup, and thus they could potentially be increased through improved apparatus design, allowing for better access to the hyperpolarization source of para-H2 (as well as reduced transit times to high field for detection). Moreover, the combination of heterogeneous SABRE[38] catalysts with the method presented here may allow preparation of pure hyperpolarized liquids because such solid-phase catalysts can be separated[38] and recycled.[39] Nevertheless, the reported 15N signal enhancement values are already comparable to 15N enhancements previously reported using dissolution DNP technology and a commercial DNP hyperpolarizer.[33] However, the method reported here achieves the steady-state maximum hyperpolarization level in <1 min without sophisticated equipment, versus ∼2 h using expensive DNP hyperpolarizers.[33] It should be noted that unlike PHIP or DNP technologies, which have been successfully tested in vivo using relevant biomolecules, SABRE technology has not yet been demonstrated for in vivo use. SABRE for hyperpolarization of 15N pH sensors can in fact directly lead to promising in vivo applications because the 15N SABRE-SHEATH procedure is a relatively simple process and because in vivo pH sensors address an important metabolic biomedical question.

In conclusion, 15N SABRE-SHEATH of neat liquids was successfully demonstrated on a model Py molecule and applied as a screening technique for mono- and dimethyl-substituted Py compounds shown to be promising for minimally invasive pH imaging.[33] Catalyst access to para-H2 was found to be the limiting factor for achieving 15N polarization levels beyond 1%. The 15N SABRE-SHEATH process was found to be selective for utilizing the spin order of the para-H2 spin bath for hyperpolarization of 15N versus 14N, enabling efficient hyperpolarization of neat liquids containing a naturally abundant level of 15N. Deuterium and 14N nuclei can act as direct and indirect hyperpolarization sinks and should be avoided or minimized where possible. While the NMR signal and polarization enhancements are relatively modest for 1H SABRE of n.a. Py, (Py in milliTesla regime, ε ≈ 4) and 15N SABRE (15N-Py in microTesla regime ε ≈ 33) due to the finite capacity of the parahydrogen/Ir-hydride spin bath, the resulting payload of 15N hyperpolarization (the product of agent concentration and its hyperpolarization) is more than doubled in this proof-of-principle demonstration compared with previous demonstrations of 15N SABRE in methanol-d4 solutions.