Remarkable Levels of 15N Polarization Delivered through SABRE into Unlabeled Pyridine, Pyrazine, or Metronidazole Enable Single Scan NMR Quantification at the mM Level

While many drugs and metabolites contain nitrogen, harnessing their diagnostic 15N NMR signature for their characterization is underutilized because of inherent detection difficulties. Here, we demonstrate how precise ultralow field signal amplification by reversible exchange (±0.2 mG) in conjunction parahydrogen and an iridium precatalyst of the form IrCl(COD)(NHC) with the coligand d9-benzylamine allows the naturally abundant 15N NMR signatures of pyridine, pyrazine, metronidazole, and acetonitrile to be readily detected at 9.4 T in single NMR observations through >50% 15N polarization levels. These signals allow for rapid and precise reagent quantification via a response that varies linearly over the 2–70 mM concentration range.


■ INTRODUCTION
Hyperpolarization methods have been shown to dramatically improve the sensitivity of nuclear magnetic resonance (NMR) and magnetic resonance imaging (MRI) 1,2 in a process that involves increasing the purity of the magnetic states they detect. Signal amplification by reversible exchange (SABRE) reflects one such method. It harnesses the nuclear spin order of parahydrogen (p-H 2 ) 3−5 and is a consequence of the pioneering work of Weitekamp 6 and Eisenberg. 7 For SABRE to operate, the symmetry of p-H 2 is first broken by temporarily placing it in a metal complex so that the new hydride ligands couple distinctly to NMR active spins within the ligand sphere of the product. A process of reversible binding then allows a suitable substrate to become hyperpolarized through a catalytic process that transfers nuclear spin order within the complex rather than achieving a change in the chemical identity. 3,5 Typically, this process takes place in a specified magnetic field that is often called the polarization transfer field (PTF) and can be selected to optimize efficiency. 8,9 The selection of this field is made according to the chemical shift difference that exists between the interacting nuclear spins and their spin− spin couplings 10,11 in a process that has been accurately modeled. 12 As the active SABRE catalyst may break the symmetry of the two protons that were initially located in p-H 2 through chemical or magnetic inequivalence effects, the process of catalysis can be complex. 5,13 This is because for the spin order transfer from the p-H 2 derived hydride ligands to take place, the receiving ligand nuclei must exhibit different spin−spin couplings to these two protons.
Knowledge of this behavior has influenced the SABRE catalyst design, 14 and the resulting sensitization process has enabled the easy NMR detection of low-abundance inorganic species. 13 Other studies have used deuterated coligands to improve the spin-order yields in SABRE by reducing waste through the focusing of polarization transfer into fewer receptor sites. 15 When this is achieved in conjunction with 2 H labeling, the associated extension of the nuclear spin-order lifetime has proven to be particularly beneficial as decoherence within the SABRE catalyst reflects one route to reduce the overall processes efficiency. 16 These two effects combine to extend the duration, over which signals remain visible to NMR. As in classical terms, one T 1 period is associated with 63% destruction of the hard-won polarization level. Not surprisingly, the extended lifetimes associated with molecular singlet states 17−21 and their derivatives feature extensively in hyperpolarization research as one goal is often to study in vivo reactivity. 22 In further developments, Tessari et al. have shown how 1 H-SABRE can achieve precise analyte quantification at low substrate loadings by the involvement of a slow exchanging coligand. 23 25 and it was noted that sterically hindered amines, which failed to bind efficiently, benefited by the addition of smaller NCMe, which enables the formation of [Ir(H) 2 (IMes)(aniline) 2 (NCMe)]Cl. 26 The successful use of amines reflects an important boost to SABRE because the hyperpolarized NH response can be used to sensitize other molecules through proton exchange. 25 More recently developments of this ligand design route have enabled the hyperpolarization of pyruvate and acetate. 27,28 Normally, although the detection of 15 N by NMR is even more challenging than that of 1 H because of its 0.36% natural abundance and low magnetogyric ratio, 15 N detection is, however, needed for the characterization of important nucleobases, nucleosides, nucleotides, peptides, proteins, and transition metal complexes. In addition, as the T 1 of 15 N can exceed many minutes, magnetic state lifetimes can approach those of positron emission tomography. 29−33 It is not, therefore, surprising that 15 N hyperpolarization reflected an early target of both spontaneous 3 and radio frequency-driven SABRE. 34 Warren et al. refined these methods through SABRE SHEATH 35,36 to deliver 20% 15 N polarization in metronidazole. 37 Several alternative radio frequency strategies have also been exemplified 38−40 and given the goal of in vivo SABRE, water soluble SABRE catalysts have also been described 41,42 with the MRI detection of a 15 N response illustrated. 42 Here, we seek to demonstrate how amines as coligands can enable the highly efficient 15 N polarization of a range of target substrates (sub) via SABRE catalysis through [Ir(H) 2 (1)-(sub) 2 (BnNH 2 )]Cl (a) or [Ir(H) 2 (1)(sub)(BnNH 2 ) 2 ]Cl (b) of Scheme 1 in order to improve the potential of the SABRE approach to achieve in vivo MRI detection.

■ RESULTS
Hyperpolarization of the 15 N NMR Signal of Pyridine. We start by considering nonlabeled pyridine at 35 mM concentration because of its wide use in early SABRE research 3,4,13 in conjunction with the precatalyst [IrCl(COD)-(h 22 -1)] 43 (5 mM) of Scheme 1. Our experimental measurements involved examining an NMR tube containing methanold 4 solutions of these reagents under 3 bar (absolute) pressure of p-H 2 at 99% purity. p-H 2 gas is first dissolved by shaking the NMR tube while it is located in a preset magnetic field that lies between ±1 mG and ±70 G for ∼10 s (relative to the main NMR magnetic field orientation). Subsequently, the sample is placed in a 9.4 T magnet where the final NMR signal detection step occurs.
Under these conditions, the SABRE catalyst [Ir(H) 2 (h 22 -1)(py) 3 ]Cl forms and a 1 H NMR signal gain of 1452-fold can be seen for the ortho proton resonance of free pyridine that is present in the solution after being transferred from a 60 G field. This polarization transfer step takes 10 s to complete and the resulting polarization level (P H ) is 4.65% (P x reflects the percentage polarization associated with nuclei x). In this case, the catalyst breaks the symmetry of the two p-H 2 -derived protons through magnetic inequivalence effects, and hence, the spin order transfer flows optimally within the equatorial plane that contains the hydride ligands into bound pyridine. 44 3 ]Cl enables the efficient transfer of polarization at an approximate −1 mG field that is of the same sense to the main 9.4 T observation field. The consequence of this process is a 39200-fold (±2%) 15 N NMR signal gain, which means the corresponding P 15 N value is 12.9% (±2%). Hence, this unlabeled 35 mM sample of pyridine can be detected by 15 N NMR spectroscopy in a single scan NMR measurement at a magnetic field of 9.4 T with a signal to noise ratio of 11 using a routine inverse detection probe.
Establishment of Coligand Benzylamine As Beneficial to the Hyperpolarization of the 15 N NMR Signal of Pyridine. When the coligand d 7 -benzylamine (d 7 -BnNH 2 ) was added to such a sample, at an initial concentration of 17.5 mM, it proved to rapidly convert into its d 9 -benzylamine isotopologue. Consequently, we refer to d 9 -BnND 2 throughout this article even though d 7 -BnNH 2 is actually added to the samples. The resulting 1 H NMR spectra reveal that in addition to this labeling change, two new inorganic species are formed, which yield pairs of hydride ligand signals at δ −22.14 and −22.58, and δ −23.34 and −23.73, respectively. These hydride ligand signals arise from [Ir(H) 2 (h 22 -1)(d 9 -BnND 2 )(py) 2 ]Cl and [Ir(H) 2 (h 22 -1)(d 9 -BnND 2 ) 2 (py)]Cl, respectively, that are present in the solution in the ratio 2.6:1. The two complexes contain inequivalent hydride ligands that differ from one another according to the identity of the axial ligands in the complex, as detailed in Scheme 1 and the Supporting Information. Furthermore, as their proportions match the value seen when a similar sample is created by the initial addition of benzylamine and H 2 to [IrCl(COD)(h 22 -1)], but before pyridine addition takes place, it can be concluded that these two complexes are in equilibrium. Hence, the separation of their roles in the underlying SABRE process is impractical, but we note it would be expected that both will contribute to this process. In addition, it is important to recognize that both of these complexes contain chemically and magnetically distinct hydride ligands. The result of this change is that spin-order transfer can now proceed into ligands that lie trans and cis to hydride, which means that spin dilution, associated with polarization of the axial ligands, is expected and this will reduce the SABRE signal gains that are seen for the free substrate. 44 Hence, the involvement of polarization transfer protecting d 9 -BnND 2 , which limits spin-order wastage should be of significant benefit to the SABRE outcome.
When the resulting d 9 -BnND 2 solutions were examined for SABRE, the 1 H NMR response resulting from this mixture of catalysts proved to contain a free pyridine ortho proton   3 ]Cl. More notable is the fact that the corresponding 15 N NMR spectrum contains a signal that is indicative of a P 15 N value of 18% (53,300 ± 6000-fold) in conjunction with a PTF of approximately −1 mG ( Figure 1a). This reflects a 27% improvement in the SABRE efficiency when compared to that achieved by [Ir(H) 2 (h 22 -1)(py) 3 ]Cl and confirms that there is a benefit to using the coligand d 9benzylamine when seeking 15 N polarization. Upon changing to [IrCl(d 22 -1)(COD)], and completing a similar series of d 9 -BnND 2 promoted measurements, the levels of signal gain seen in the pyridine ortho proton 1 H NMR signal rises to 1324-fold, although the 15 N polarization level proved to be unaffected. Hence, while catalyst deuteration is not successful at improving SABRE 15 N activity, it is able to improve the level of 1 H signal gain because of reduced spin order wastage and improved 1 H relaxation. 16 This suggests that low-field 15 N-relaxation within the catalyst is not improved.
While it is well known that the optimum SABRE catalyst changes with the identity of the substrate, it has been clearly demonstrated here that there is also a further dependence on the efficiency of SABRE transfer within a given substrate according to whether 1 H or 15 N is the target. The optimum rate of ligand exchange for 1 H transfer has been proposed by Barskiy to be 4.5 s −1 in complexes of the type [Ir(H) 2 (h 22 -1)(py) 3 ]Cl. Consequently, the rate of pyridine substrate dissociation in [Ir(H) 2 (h 22 -1)(py) 2 (d 9 -BnND 2 )]Cl in methanol-d 4 solution was determined using the EXSY method and found to be 0.06 s −1 at 268 K. This value increases to 1.04 s −1 upon warming at 298 K, and 2.1 s −1 at 308 K. Our associated SABRE measurements reveal that the corresponding 1 H NMR signal gains change from 600-fold, through 4530-fold to 3550fold at the 308 K setting. Hence, it appears that a rate closer to 1.04 s −1 is optimal for 1 H transfer into pyridine using this catalyst. Our experiments also reveal that there is a 30% growth in efficiency of 15 N polarization for pyridine on moving from 268 to 298 K, and a further 22% improvement on moving to 308 from 298 K. Consequently, we can confirm that the two different nuclei are best served with different rates of ligand exchange.
Hyperpolarization of the 15 N NMR Signal of Acetonitrile. In order to develop this method further, acetonitrile was tested at a similar 35 mM concentration in conjunction with the SABRE catalyst [Ir(H) 2 (h 22 -1)(NCMe) 3 ]Cl. This catalyst also relies on magnetic inequivalence to break the symmetry of the hydride ligands and it yields a 1 H NMR signal gain of just 83-fold per methyl proton in the unbound acetonitrile present in the solution after transfer at 298 K from a 70 G field. The SABRE-derived 15 N NMR signal gain for CH 3 CN was found to be far more substantial, at 41,800 ± 6000-fold (14% polarization) after transfer from an approximate −1 mG field.
Acetonitrile hyperpolarization was then studied in conjunction with 3.6 equivalents of the coligand d 9 -benzylamine relative to 5.    48 The gain in the 1 H signal intensity relative to the situation with h 22 -1 is, however, consistent with a reduction in the polarization transfer into this ligand through deuteration and an extension of the hydride ligand relaxation times. 14 Hyperpolarization of the 15 N NMR Signal of Pyrazine. We next consider pyrazine (pz). This substrate was tested by taking 5.2 mM methanol-d 4 solutions of [IrCl(COD)(h 22 -1)] that contained a sevenfold excess of pz under 3 bar of p-H 2 . The resulting 1 H NMR signal gain for pz was now 900-fold per proton (2.9% polarization) and a P 15 N value of 16% (±2, per nitrogen used throughout) was observed after the transfer from −3 mG.
Studies with added h 7 -BnND 2 resulted in a 1 H NMR signal gain of 566-fold (0.8%) and a 15 N signal gain of 12% due to the associated spin dilution effects. However, when d 9 -BnND 2 and h 22 -1 were used with a PTF of 60 G, radiation damping resulted with 1 H signal detection. In order to aid the analysis, this artifact could be suppressed if a less efficient PTF of 120 G was used. Analysis under these conditions was used to deduce that the corresponding P H level is 13.5% (±0.6) per proton for a 60 G measurement while for 15 N it was 38% (per nitrogen). The 1 H NMR signal gain grew further to 30.9% (±0.7) when [IrCl(d 22 -1)(COD)] was used, but the corresponding 15 N signal response fell in intensity meaning that the scaler relaxation of the second kind is again important. We also tested the related SIMes containing precatalyst [IrCl(COD)(2)] 49 with pyrazine and discovered that a P 15 N value of 15.8% could be achieved without a coligand. Samples containing both the d 7 -benzylamine and pyrazine yield [Ir(H) 2 (pz) 2 (d 9 -BnND 2 )-(h 22 -2)]Cl and [Ir(H) 2 (d 9 -BnND 2 ) 2 (pz)(h 22 -2)]Cl in the ratio 2:1 and a P 15 N value of 44.2% via PTF from an approximate −1.9 mG field (Figure 1c). This falls to 31.8% with d 22 -SIMes in agreement with a role for 2 H-drive relation in the SABRE catalyst at low field. Figure 1d shows that the sign of the PTF, relative to that of the main observation field, affects the measured 15 N pz signal gains. This is because upon moving the sample slowly between the points of polarization transfer and measurement, if it experiences a zero-field point, there is a loss in the spin order due to relaxation at this point.
The rate of pyrazine dissociation from [Ir(H) 2 (pz) 2 (d 9 -BnND 2 )(h 22 -2)]Cl was determined using the EXSY method to be 0.33 s −1 at 268 K when the 1 H NMR signal gain is 660-fold. This rate increases to 1.8 s −1 at 298 K where the 1 H signal gain is 2200-fold. Our experiments reveal that the 20% growth in the efficiency of 15 N polarization on moving from 268 to 298 K for pyrazine is a consequence of this rate increase, which is faster than that of pyridine loss in the related complex [Ir(H) 2 (h 22 -1)(py) 2 (d 9 -BnND 2 )]Cl. This kinetic difference is consistent with the relative 15  Data were now collected on the d 22 -2 system to demonstrate that the PTF value can be used to control which of the two substrates present in solution receives polarization. This effect serves to illustrate how selectivity can be introduced into the analysis of mixtures if peak overlap is an issue (see the Supporting Information). Furthermore, a catalyst change to [IrCl(COD)(d 34 -4)] increased the N 1 value to 51% for metronidazole with 4% polarization being achieved on N 2 and 1% on d 9 -benzylamine (Table 1).
The rates of metronidazole dissociation from the resulting complex [Ir(H) 2 (mtz) 2 (d 9 -BnND 2 )(d 34 -4)]Cl were determined in methanol-d 4 solution at 268, 298, and 308 K by the EXSY method as being 0.80, 2.37, and 5.5 s −1 , respectively. For the 1 H signal gain, 298 K proved to be the best, yielding an enhancement of 856-fold. We now see an 80% growth in the efficiency of 15 N polarization on moving from 268 to 298 K, but the P 15 N values falls to just 18% at 308 K. Hence, increasing the ligand exchange rate beyond 2.4 s −1 seems detrimental.
Usage of Higher Proportions of p-H 2 to Improve the NMR Signal Gain. A series of measurements were then completed on metronidazole using a 10 mm NMR tube to deploy a larger excess of p-H 2 in conjunction with [IrCl-(COD)(d 34 -4)] and d 9 -benzylamine. A slight increase in the The Journal of Physical Chemistry B pubs.acs.org/JPCB Article 15 N polarization level to 54% results alongside a reduction in response variability to 2%. Consequently, as shown in Figure  1g, a −3.6 mG PTF can be deduced as being optimal. Similar 10 mm measurements were then made for pyridine with [IrCl(COD)(h 22 -1), acetonitrile with [IrCl(COD)(h 22 -1), and pyrazine with [IrCl(COD)(h 22 -2) in the presence of d 9benzylamine. These studies saw the P 15 N level for pyridine increase to 48% at 4 bar p-H 2 pressure. When acetonitrile was examined, a 30.7% P 15 N level was reached, but for pyrazine it became 59.4% per nitrogen. Further increases in the pyrazine % P 15 N level can be achieved through reagent dilution such that when an initial 5 mM solution of [IrCl(COD)(h 22 -2)] with a 3.6-fold excess of d 9 -benzylamine and sevenfold excess of pyrazine based on iridium is diluted 10 fold, the P 15 N value increases to 79%; the S/N ratio in this case is 11.3. In this case, the effect is directly analogous to increasing the volume of p-H 2 available.
Quantification of Reagent Concentrations at the mM Level through a SABRE-Enhanced 15 N Signal. Once we had ascertained how to achieve these polarization levels, we tested how the magnitude of the pyridine, pyrazine, and metronidazole response varied as a function of substrate concentrations between 2.2 and 70 mM. These solutions were made up by simply diluting a stock solution with an initial catalyst, d 7 -benzylamine, and substrate concentration of 10, 36, and 70 mM, respectively. We discovered that there was a linear variation in the signal response in each case, as shown in Figure  2.
In the second series of studies, we maintained a constant iridium and coligand concentration while changing the pyrazine concentration. A linear change in the 15 N signal intensity was again observed (Figure 3 Previous studies have established that using deuterated NHC ligands (d 22 -1 and d 22 -2) improve SABRE hyperpolarization transfer efficiency into methylnicotinate. This improvement is based on an extension of the hydride ligand relaxation times. 14 Studies here confirm that higher P 1 H values result in all cases in support of this benefit. However, deuteration is not beneficial for the 15 N transfer in pyridine, pyrazine, and acetonitrile. Barskiy's observations that in microtesla transfer fields, scaler relaxation of the second kind 47 associated with the quadrupolar 14 N− 13 C interaction limits the level of 13 C polarization under SABRE offer a route to explain this view. 48 For metronidazole, however, an improved value of 54% on N 1 results with d 9 -benzylamine and [IrCl(COD)(d 34 -4)], compared to that seen with precatalyst [IrCl(COD)(h 34 -4)]. Hence, 2 H labeling of the catalyst can also be of significant benefit to P 15 N .
The rates of ligand exchange were also assessed alongside the collection of variable temperature SABRE data. It was found that the rate of optimum ligand exchange could slower than that found for 1 H transfer, despite the larger 1 H− 15 N  The Journal of Physical Chemistry B pubs.acs.org/JPCB Article transfer coupling. We are currently exploring this behavior in more detail. Data were also presented that was collected from larger 10 mm NMR tubes using a 4 bar pressure of p-H 2 . This acted to increase the relative excess of the hyperpolarization fuel p-H 2 relative to the substrate and proved to result in greatly improved response reproducibility. Consequently, results demonstrated that a PTF precision of ±0.2 mG is needed for optimal 15 N transfer. In addition, ∼50% 15 N polarization levels could now be achieved in pyrazine, pyridine, or metronidazole, which makes them all highly detectable even at low concentrations.
In order to demonstrate an analytical use for these 15 N signals, results were presented to demonstrate that the magnitude of the resulting NMR response scales linearly with concentration over the range 2.2−70 mM. This means that such SABRE-derived data can be used to quantify their amount in the solution when set against a suitable reference trace. Tessari have completed a growing range of studies, which demonstrate that 1 H detection levels can be linked to both speciation and quantity, 23,24 while we have described how 13 C signals in glucose can be linked to amount. 56 These studies employed a methylated triazol coligand to simplify the exchange kinetics in order to produce the necessary linear response. We were unable to benchmark our data with that of the triazol coligand as it is not commercially available. We did, however, test d 6 -DMSO, which is finding widespread use as a coligand for the sensitization of weakly binding substrates as an alternative. As detailed in the Supporting Information, the corresponding SABRE performance was degraded.
It is therefore clear that SABRE offers a simple and yet efficient route to analyte quantification by 15 N NMR spectroscopy. Not surprisingly, we predict these results will, therefore, be of benefit if you wish to use 15 N NMR as a characterization tool, or simply to quantify precise, and yet low, levels of nitrogen-containing drugs that are present in solution or to collect 15  The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

Funding
Financial support from the Wellcome Trust (Grants 092506 and 098335), the MRC (MR/M008991/1) and the University of York is gratefully acknowledged.

Notes
The authors declare no competing financial interest.

■ ACKNOWLEDGMENTS
We thank Peter Rayner for providing some of the complexes used in this work.