SABRE-Relay: A Versatile Route to Hyperpolarization

Signal Amplification by Reversible Exchange (SABRE) is used to switch on the latent singlet spin order of para-hydrogen (p-H2) so that it can hyperpolarize a substrate (sub = nicotinamide, nicotinate, niacin, pyrimidine, and pyrazine). The substrate then reacts reversibly with [Pt(OTf)2(bis-diphenylphosphinopropane)] by displacing OTf– to form [Pt(OTf)(sub)(bis-diphenylphosphinopropane)]OTf. The 31P NMR signals of these metal complexes prove to be enhanced when the substrate possesses an accessible singlet state or long-lived Zeeman polarization. In the case of pyrazine, the corresponding 31P signal was 105 ± 8 times larger than expected, which equated to an 8 h reduction in total scan time for an equivalent signal-to-noise ratio under normal acquisition conditions. Hence, p-H2 derived spin order is successfully relayed into a second metal complex via a suitable polarization carrier (sub). When fully developed, we expect this route involving a second catalyst to successfully hyperpolarize many classes of substrates that are not amenable to the original SABRE method.

T he huge sensitivity improvement that is provided by hyperpolarization has significantly extended the scope of magnetic resonance-based in vivo study. 1−3 While several hyperpolarization methods are available to achieve this result, the para-hydrogen-induced hyperpolarization (PHIP) route is popular because it is fast, simple, and relatively low-cost. 4,5 Classically, PHIP, however, needs to chemically modify the target substrate, and this makes it unsuitable for some agents. 6 This limitation has partially been overcome through the variant of PHIP known as Signal Amplification by Reversible Exchange, termed as SABRE, which no longer relies on active hydrogenation to deliver hyperpolarized material in seconds. 7 Since its inception, SABRE has become highly successful in delivering huge sensitivity enhancements to a wide range of molecular systems that are clinically relevant. 8−11 It has been shown to work for NMR-active nuclei such as 1 H, 13 C, 15 N, 19 F, and 31 P that feature in substances such as nicotinamide, nicotinate, pyridazine, diazirine, and imidazole and achieves net polarization levels as high as 50% for 1 H and 20% for 15 N. 10,12−15 Recently, SABRE has been combined with long-lived states such that the hyperpolarized signals that it creates remain visible for up to 30 min. 16−21 This exciting development reflects one route to overcome the normal relaxation time scale of NMR that limits many methods. Such advancements are beginning to feature in human metabolomics where the creation of tools for the diagnosis of disease is possible. 1−3 Despite the general success of the SABRE hyperpolarization technique, the signal enhancements achieved by this process are currently limited to resonances that originate in ligands that were previously bound to the polarization transfer catalyst (M 1 ). Although several types of spin system have been shown to perform well with SABRE, there is a need to make this approach both robust and more generally applicable. In this study, we show how it is possible to sensitize a second metal complex (M 2 −S 1 ) via the relay of hyperpolarization from the substrate (S 1 ) in a process we term SABRE-Relay. Given that M 2 −S 1 no longer needs to react with H 2 and that with development it could contain labile ligands, we expect this approach to widen the range of substrates that can be hyperpolarized by this type of approach. Scheme 1 illustrates the basis of this effect.
The singlet state concept is central to the SABRE process. In a pair of coupled spin-1/2 nuclei (e.g., H 2 ), the term singlet state relates to their antisymmetric spin eigenstate. 22 The PHIP concept 5 then harnesses this singlet state to facilitate enhanced NMR detection. Remarkably, in 2004, it was shown by Levitt and co-workers that such singlet spin order can be created by suitable radio frequency (rf) pulses in many ordinary molecules and the resulting nuclear spin lifetime can extend beyond the more usual T 1 boundary. 23 The long-lived nature of these singlet states can be traced back to the fact that they are immune to one of the major relaxation-causing mechanisms, the intrapair dipole−dipole process. 22 Long-lived singlet states (LLSs) are now being used for hyperpolarization storage, to obtain molecular structure information, and to follow slow molecular processes. 24−28 Warren and co-workers developed a related technique to access LLSs in chemically equivalent spin systems by exploiting magnetic inequivalence. 29,30 In previous work, it has been shown that SABRE derived hyperpolarized LLSs can be formed and accessed for several minutes after storing either in high-or low-magnetic fields. 18−21 In these experiments, SABRE is first used to hyperpolarize the exchangeable substrate S 1 of Scheme 2 via its scalarcoupling framework. When these are two-spin substrates, the hyperpolarization is associated with both Zeeman and singlet spin order, as detailed in Scheme 3. Alternatively, the resulting Zeeman polarization can be turned into singlet spin order via rf excitation, as detailed. 16−21 S 1 is binding reversibly to M 2 −S 1 throughout this process according to Scheme 3. If S 1 were to have singlet spin order and its symmetry were to be broken through the J-coupling network of M 2 −S 1 , its latent polarization should be unlocked and transferred further, in this case into 31 P. 31 This strategy reflects the relayed transfer of polarization from p-H 2 into a second agent that never actually comes into contact with H 2 , and hence, the key requirement is that hyperpolarized S 1 has a long lifetime. This process is depicted in Scheme 3 for [Pt-(S 1 ) 2 (dppp)](OTf) 2 (M 2 −S 1 ). Here, M 2 −S 1 rapidly forms from [Pt(OTf) 2 (dppp)], where the bidentate phosphine controls the lability of the triflate (OTf − ) and the potential for any substrate (S 1 ) to bind (see Supporting Information sections S3 and S4 for more details). Transfer would also be possible via chemical and magnetic inequivalence through For context, it has been shown experimentally that the addition of p-H 2 leads to sensitization of the 31 P signature of [Ru(H) 2 (dppp)(PPh 3 )(CO)]. This effect was explained theoretically on the basis of the coherent evolution of the zero quantum (ZQ) coherence associated with the p-H 2 singlet state under 1 H− 31 P spin−spin coupling. 32−34 Furthermore, the complex Ir(H) 2 Cl(PPh 3 ) 3 has been shown to hyperpolarize its bound 31 P responses in addition to that of PPh 3 via SABRE. 35,36 In contrast, the transfer of single-spin-based 129 Xe hyperpolarization into a second agent has been shown to proceed via the incoherent spin polarization-induced nuclear Overhauser effect (SPINOE). 37,38 There is also the possibility of coherent Zeeman order transfer. 39,40 Hence, there are several welldefined pathways for hyperpolarization transfer between diamagnetic materials that might operate here.
In this section, we demonstrate the experimental viability of SABRE-Relay. We employ the efficient SABRE precursor [IrCl(IMes)(COD)] (M 1 ) or its 2H labeled form. Standard SABRE methods are used to polarize the carrier substrates (S 1 ) of Scheme 3 in the presence of M 1 . Platinum-based M 2 contains bis-diphenylphosphinopropane (dppp), bis-diphenylphosphinomethane (dppm), and bis-diphenylphospinoethane (dppe) and a pair of weakly bound triflate ligands. Synthetic details for the formation of M 2 and its reactions to form M 2 −S 1 can be found in the Supporting Information (sections S3 and S4). Samples were prepared in deuterated methanol and contained a 1:1 ratio of M 1 and M 2 , with each having a concentration of 5 mM and substrate loading (S 1 ) of 50 mM. These solutions were then degassed prior to activation with p-H 2 . This led to the formation of [Ir(H) 2 (IMes)(S 1 ) 3 ]Cl. Subsequently, SABRE transfer was undertaken at a range of magnetic mixing fields, and a series of 1 H and 31 P NMR measurements was made according to the process detailed in Figure 1. After bubbling with p-H 2 at two different magnetic fields, the solution was then transferred into the magnet for NMR measurement, which took place after the application of simultaneous 90°pulses to both channels. The magnetic field cycling illustrated is needed to optimally polarize the corresponding 1 H and 31 P responses of S 1 and M 2 −S 1 , respectively, and relates to 1 H− 15 N transfer, as exemplified by Theis et al. 16 The formation of a singlet state within the carrier substrate (S 1 ) was realized either naturally (protocol-1: low-field hyperpolarization) or by a suitable rf based magnetization-to-singlet (M2S) pulse sequence (protocol-2). 23 We note that hyperpolarized Zeeman derived spin order remains present when the sample enters the high-field magnet for observation because the SABRE process is ongoing.
In the first measurement, S 1 was H 1 -nicotinamide S 1a of Scheme 3. The resulting 1 H NMR signals for S 1a showed a 370 ± 20-fold Zeeman-based signal enhancement after SABRE (see Table 1) at 6 mT, but no 31 P NMR signal was detected for M 2 −S 1a after rapid transfer into the high-field spectrometer for observation according to protocol 1. In this case, S 1a contains a single proton, and hence, there is no possibility to create singlet order in S 1a alone. A further control experiment was performed

The Journal of Physical Chemistry Letters
Letter without p-H 2 , and again, no 31 P NMR signal was seen for M 2 − S 1a in a single scan measurement, although its formation was confirmed after appropriate signal averaging. Hence, we conclude that the presence of M 2 and M 2 −S 1a does not stop SABRE from operating with M 1 and that substrate S 1a is unable to relay polarization into M 2 −S 1a . Furthermore, we note that M 2 −S 1 is not hyperpolarized as a consequence of the p-H 2 that is in solution. We interpret these results to suggest that any single spin Zeeman derived hyperpolarization or SPINOE transfer is at best weak for this material and note that the short 1 H-relaxation times might account for this. 41 We then examined the three variants of nicotinamide, S 1b , S 1c , and S 1d , where we have previously demonstrated that both Zeeman and singlet order can be created in their aromatic protons via SABRE and rf-driven transfer. 19 All three of these substrates now yield SABRE hyperpolarized 1 H NMR signals with signal gains lying between 200-and 250-fold per proton for their Zeeman polarizations after transfer at 6 mT when observed at 9.4 T. A 31 P NMR measurement was then made on each sample according to protocol 1 of Figure 1. This resulted in the detection of a 31 P signal for both of the chemically inequivalent phosphorus centers in the corresponding mono substituted M 2 −S 1 complexes. These signals were enhanced over their thermally polarized levels by 65-, 88-, and 32-fold, respectively. The corresponding enhancement values were determined by comparison with signal-averaged 31 P NMR spectra of M 2 −S 1 . When protocol 2 for S 1b and S 1c was applied, the resulting 31 P signal gains were 42-fold and 50-fold, respectively. It is likely that the weaker 31 P signal gains result from the longer experiment time. We note that the similarity in the single-proton relaxation times of S 1a and S 1b suggests again that any SPINOE or Zeeman derived contribution would be weak.
These observations can be explained if the singlet symmetry of S 1 is broken upon binding and polarization transfer into the Figure 1. Experimental scheme for SABRE-Relay, showing timings, magnetic field variance, and rf sequence. First, the sample is mixed with enriched p-H 2 at low magnetic field (∼6 mT and ∼1−10 μT) for the durations of τ LF1 and τ LF2 before moving to high field (τ tr ) for NMR observation. A simultaneous 90°pulse is applied to 1 H and 31 P prior to acquiring the 31 P signal with 1 H decoupling (protocol 1). In a second variant, protocol 2, an M2S sequence, 23 is applied between τ LF1 and τ LF2 .

The Journal of Physical Chemistry Letters
Letter 31 P nuclei of M 2 −S 1 occurs. In the case of S 1b , a five-bond 1 H− 31 P coupling would be involved in this step, while for S 1c and S 1d , larger four-bond couplings would operate. The reduced enhancement level seen with S 1d is predicted to reflect the rapid relaxation of its singlet state in the reaction mixture ( Table 1). The corresponding values for S 1b and S 1c are longer but still smaller than their 34 and 39 s values in the absence of the catalyst. Hence, in accordance with this hypothesis, the presence of M 1 and M 2 −S 1 is seen to impact directly on the T LLS lifetimes. Figure 2 shows the corresponding 31 P NMR traces after the one-shot SABRE-Relay process with S 1a , S 1b , S 1c , and S 1d to illustrate this behavior.
An improvement in the level of 31 P NMR signal gain, from 32-fold to 56-fold, is observed with M 2 −S 1d when protocol 2 is used. This again suggests the involvement of a coherent spin order transfer mechanism leading to the hyperpolarized 31 P NMR signal. Similar experimental strategies were then employed to examine related S 1e and S 1f . In the case of S 1e , while very strong 1 H SABRE results, the relayed 31 P signal gains are lower. This is again likely to be a consequence of their relatively short magnetic state lifetimes.
When substrates S 1g (pyrimidine) and S 1h (pyrazine) were examined, the observed 31 P NMR signal enhancements increased to over 100-fold (Table 1). Figure 3 shows the corresponding 31 P NMR spectra of M 2 −S 1g and M 2 −S 1h when a 50 mM substrate concentration was employed with M 2 at the 5 mM level. This improvement confirms that the identity and properties of S 1 are important in controlling the visibility of M 2 −S 1 . In these two cases, slow exchange is predicted, which confirms that catalyst lifetime plays a role in this process.
In a final refinement, we studied two other platinum metal complexes, [Pt(OTf) 2 (dppm)] and [Pt(OTf) 2 (dppe)], and similar SABRE-Relay experiments were performed. While 31 Psignal enhancements were again observed, they were significantly lower than those seen for the dppp complex. The decrease in enhancement factor reflects a combination of residence time and relaxation effects. We are currently studying these effects in greater detail. Nonetheless, the associated signal enhancements confirm that several metal complexes are active for SABRE-Relay. The experimental details and results for [Pt(OTf) 2 (dppm)] and [Pt(OTf) 2 (dppe)] are presented in Supporting Information sections S3−S5.
In summary, we have demonstrated how SABRE can be cascaded into a second metal complex via a coherent transfer pathway involving a series of hyperpolarized substrates. This involved the detection of enhanced 31 P NMR responses in a metal complex that does not interact directly with H 2 . Hence, we have presented a route to overcome one of the key SABRE limitations, associated with p-H 2 being the singlet carrier. Nheterocycles can bind to many metal complexes, which in turn may contain other labile ligands. We expect to be able to use SABRE-Relay to enhance new classes of agents that are not amenable to the traditional SABRE hyperpolarization route. The mechanism of transfer is likely to be based on a coherent spin order route such as that involved in singlet state evolution via S 1 , which is directly analogous to the original SABRE concept. However, we note that this process simply involves propagation of the low-field-created ZQ coherence, but other routes involving coherent polarization transfer from Zeeman order under these low-field conditions may contribute and we are now seeking to differentiate their contributions. 40,42 Regardless of the pathway, we take advantage here of what would be expected to be relatively large 31 P couplings to propagate these effects and therefore expect an efficient second step. Given the interest in hyperpolarized MRI, the potential of this approach to improve magnetic resonance sensitivity may be significant, and we are currently working on optimization of this technique.   (structures above the NMR spectra) that form from M 2 and S 1g or S 1h , respectively. In both cases, the upper 31 P NMR spectrum is the control, which involved 128 transients, while the lower NMR spectrum was acquired by SABRE-Relay through process 1 and associated with a single detection pulse according to Figure 2.

Notes
The authors declare no competing financial interest.

■ ACKNOWLEDGMENTS
This work was supported by the Wellcome Trust (Grant Numbers 092506 and 098335). We also thank the EU for funding the EUROPOL network (award 642773, Elizabeth Fear). We thank Dr. Richard John for experimental support.