Direct Hyperpolarization of Nitrogen-15 in Aqueous Media with Parahydrogen in Reversible Exchange

Abstract

Signal Amplification By Reversible Exchange (SABRE) is an inexpensive, fast, and even continuous hyperpolarization technique that uses para-hydrogen as hyperpolarization source. However, current SABRE faces a number of stumbling blocks for translation to biochemical and clinical settings. Difficulties include inefficient polarization in in water, relatively short lived ¹H-polarization, and relatively limited substrate scope. Here we use a water soluble polarization transfer catalyst to hyperpolarize nitrogen-15 in a variety of molecules with SABRE-SHEATH (SABRE in Shield Enables Alignment Transfer to Heteronuclei). This strategy works in pure H₂O or D₂O solutions, on substrates that could not be hyperpolarized in traditional ¹H-SABRE experiments, and we record ¹⁵N T₁ relaxation times of up to 2 min.

Graphical abstract

Introduction

NMR and MRI are non-destructive methods to obtain information about molecular structure and spatial morphology. However, magnetic resonance is restricted mainly because of the inherently low sensitivity as a result of low thermal polarization levels. For example, NMR spectroscopy and clinical MRI predominantly use highly abundant ¹H nuclei. Even so observation of low concentration analytes remains challenging. Hyperpolarization methods (e.g. DNP, PHIP, SABRE, SEOP)^1–7 enhance MR signals by 4–5 orders of magnitude and overcome inherent sensitivity limitations.^8–11

Traditionally, hyperpolarization methods require extensive optimization. Usually methods and optimization are associated with high experimental complexity and cost. In this regard, Signal Amplification By Reversible Exchange (SABRE) stands out because it is simple, fast and continuously repeatable.^{4, 12} SABRE uses readily available para-hydrogen (p-H₂) as source of polarization. The transfer occurs in reversibly formed substrate-hydrogen adducts in a transition metal complex. The magnetic evolution field B_evo must be sufficiently low to mix energy levels between hydride-¹H and the target nucleus to establish a path for polarization transfer.^{7, 13} While protons in the substrate are targeted at magnetic fields around 65 G,¹⁴ transfer to heteronuclei (e.g. ¹⁵N, ¹³C, ³¹P) occurs at µT magnetic fields using a technique termed SABRE in Shield Enables Alignment Transfer to Heteronuclei (SABRE-SHEATH).⁷ As shown in Scheme 1, the required hardware is relatively simple.

Scheme 1.

A sample is hyperpolarized via SABRE-SHEATH for an evolution time t_evo at optimized matching field B_evo of ~0.5 µT established by a small solenoid coil in a magnetic shield that attenuates the Earth’s magnetic field. The sample is transferred into a benchtop (1 T) NMR spectrometer or conventional high-field (8.45 T) spectrometer for detection after hyperpolarization.

As a result of experimental simplicity and its promise, SABRE and SABRE-SHEATH are now attracting an increasing number of research groups contributing to its rapid development.^{4, 7, 14–20} A milestone for SABRE was the transition from organic solvents to aqueous solutions, which was recently achieved for ¹H-SABRE.^21–24

Still, for ¹H spin lattice relaxation times are relatively short and the substrate scope is limited. Direct polarization transfer to heteronuclei has not been demonstrated in aqueous environment. Hyperpolarizing nitrogen-15 via SABRE-SHEATH allows a wider range of structural motives and relaxation times are characteristically larger.

SABRE-SHEATH with ¹⁵N targets is made accessible with the water soluble [IrCl(IDEG)(COD)] precatalyst (1a). As shown in Scheme 2 the precatalyst is converted to the catalytically active species (1) in presence of substrates under a hydrogen atmosphere.²¹ At µT magnetic field hydride and ¹⁵N energy levels match and the spin system coherently evolves with a rate given by J_NH-into ¹⁵N-polarization on substrates.²⁵

Scheme 2.

The precatalyst [IrCl(IDEG)(COD)] (1a) is transformed into the active species Ir(IDEG)(H)₂Sub₃ (Sub = Substrate) (1) in the presence of a substrate of choice (2–6) under a hydrogen atmosphere. Reversible exchange leads to polarization buildup on ¹⁵N within 30–120 s. The polarization transfer is primarily driven by the J_NH-coupling through the bonds that form a 180° angle. The N-H coupling through bonds forming a 90° angle is close to zero.

We investigate different molecular motifs found in medical drugs, biomolecules and molecular tags. Structural motifs could be readily translated from the established [IrCl(IMes)(COD)] system.^{15, 26}

Pyridine (2), the canonical SABRE substrate,⁴ was a logical first choice. Next, nitriles are often encountered in drugs,²⁷ polarize consistently well, tolerate complex backbones, and show large ¹⁵N-SABRE-SHEATH enhancements, despite little to no ¹H-SABRE.^{26, 28} We selected benzonitrile (3) and α-cyano-4-hydroxycinnamic acid (4) (CHCA, buffered with NaOD to pH 7.5). Diazirines, which also do not exhibit ¹H enhancements, are common biomolecular tags that can replace CH₂ groups in many classes of biomolecules.²⁹ Here we use 2-cyano-3-(D₃-methyl-¹⁵N₂-diazirine)-propanoic acid (5). Lastly, we focus on nicotinamide (6), the amide of vitamin B₃, which could be tolerated in vivo at detectable concentrations and is a potential option for translation to biomedical studies.^{19, 30}

For these substrates we detail hyperpolarization levels, carefully characterize temperature and magnetic field dependencies, consider the effect of deuterated vs protonated solvents (D₂O vs H₂O), and measure relaxation time constants at various magnetic fields.

Results & Discussion

In Figure 1 we show a comparison between single scan spectra originating from compounds directly SABRE-SHEATH hyperpolarized in aqueous medium, referenced to thermally polarized neat ¹⁵N-pyridine at 8.45 T. Concentrations of investigated compounds are different as a result of solubility as well as sample loss phenomena for benzonitrile and pyridine. Both pyridine and benzonitrile were initially prepared as 100 mM solutions but after activation by H₂ bubbling the concentrations were significantly reduced.

Figure 1.

¹⁵N spectra of A) thermally polarized reference at 8.45 T and B–F) hyperpolarized compounds (in D₂O unless denoted otherwise). A) neat ¹⁵N-pyridine. B) ¹⁵N-Pyridine, C) ¹⁵N-Benzonitrile, D) ¹⁵N-CHCA, E) ¹⁵N2-Diazirine F) ¹⁵N-Nicotinamide (in H₂O).

A synopsis of experimental results and conditions is given in Table 1 (experimental details provided in Materials and Methods). Spectra are acquired at 1 T and 8.45 T (see Scheme 1.) to study the field dependence of T₁ relaxation as detailed below. The 1 T measurements also demonstrate the feasibility of high sensitivity single scan ¹⁵N detection with a benchtop NMR system. Furthermore, to determine the effect of proton containing solvents, nicotinamide was investigated in H₂O.

Table 1.

Synopsis of experimental conditions, enhancements and polarization levels. Substrates are ¹⁵N labelled, solvent is D₂O unless otherwise specified. Concentrations of liquid substrates are determined at the time of the experiment using ¹H spectroscopy. T = 75 °C.

Substrate	Activation time [h]	csubstrate /ccatalyst	cSubstrate [mM]	Enhancement 8.45 T (P [%])	Enhancement 1 T (P [%])	Enhancement ratio (1 T/8.45 T)
(2) pyridine	12	3.3	16.5[b]	1100	-[c]	-[c]
(3) benzonitrile	0.25	20	50[a,b]	90	-[c]	-[c]
(4) CHCA	12	10	50	440 (0.13)	3700 (0.13)	8.4
(5) diazirine	6	10	16.3	560 (0.16)	4700 (0.17)	8.4
(6) nicotinamide	12	20	100	520(0.16)	6100 (0.21)	8.9
(6) nicotinamide H2O	12	20	100	630 (0.2)	10500 (0.37)	28

^[a]Signal maximum obtained 30 min. after activation, subsequent substrate loss due to evaporation.

^[b]Initial concentration 100 mM.

^[c]Insufficient SNR.

We find that polarization levels in deuterated solvents are largely independent of the detection field i.e. enhancements simply scale with the thermal polarization. In contrast, for nicotinamide in H₂O (Table 1, Entry 6), we observe lower apparent polarization levels at 8.45 T. This is caused by relaxation losses during transfer because it takes much longer to transfer the sample into the high field magnet (~8 s) than into the benchtop device sitting right next to the magnetic shields (~2 s). The solvent protons (and deuterons) are in chemical exchange with the ¹⁵N-substrate where they cause spin-dipole relaxation. This relaxation mechanism scales with the distance between the relaxation partners r_ij⁻⁶ as well as the gyromagnetic ratio, which is 6.5 times smaller for deuterium,^31,32 explaining the observed differences between solvents.

SABRE-SHEATH in water gives rise to a new set of challenges. Water has significantly higher viscosity and surface tension than methanol, and at room temperature the solubility of hydrogen in water is five times lower.^33–34 We observed that some samples, specifically non-polar liquid state substrates (e.g. benzonitrile and pyridine) are extracted from the solvent when bubbling with hydrogen during the polarization buildup. Nicotinamide and CHCA, both crystalline solids when isolated, were used for systematic studies as substrate loss did not occur.¹⁹

Of particular interest are the dependence of the ¹⁵N polarization on temperature and magnetic evolution field B_evo. Figure 2 contrasts the established [IrCl(IMes)(COD)] in methanol and catalyst (1) in H₂O/D₂O as a function of these variables (T, B_evo).

Figure 2.

Comparison of ¹⁵N polarization as a function of temperature in A) H₂O, B) D₂O and C) methanol-d₄ at B_evo = 0.5 µT. D) Hyperpolarized signals as function of µT field at the temperature corresponding to maximum polarization in the respective solvents: 22 °C for ¹⁵N-acetonitrile in MeOH-d₄ and 72 °C for ¹⁵N-CHCA in D₂O (blue: 5 mM [IrCl(IMes)(COD)], 30 mM pyridine, 100 mM ¹⁵N-CH₃CN, methanol-d4; magenta: 5mM [IrCl(IDEG)(COD)], 30 mM pyridine, 50 mM ¹⁵N-CHCA, D₂O).

The temperature dependence was studied using a 100 mM nicotinamide sample. For catalyst system (1) in H₂O (Fig. 2A) and D₂O (Fig. 2B) the ¹⁵N polarization increases with temperature. In contrast, in methanol (Fig. 2C, [IrCl(IMes)(COD)] precursor) the largest polarization is recorded at room temperature.

The magnetic field dependence is shown in Figure 2D. We compare normalized data (max. ¹⁵N polarization: 0.13% in D₂O, 1.7% in MeOH-d4) of two nitrile/solvent systems: first, in blue: ¹⁵N-acetonitrile in MeOH-d₄ with [IrCl(IMes)(COD)] and second, in magenta, ¹⁵N-CHCA in D₂O with[IrCl(IDEG)(COD)] (1a).

We note that nitriles are better suited for this study than nicotinamide, as enhancements are more robust and reproducible. Additionally, they exhibit inversion of the NMR signal upon inversion of B_evo. Variation of the temperature changes the dissociation rate constants of substrate and catalyst bound H₂.^{13, 15, 35} Optimal polarization transfer efficiency is expected when the exchange rate k_diss is on the order of the ¹⁵N-to-hydride J_NH-coupling across the iridium center (see scheme 2).^{13, 35} Figures 1A–C show that the IMes catalyst in methanol yields largest ¹⁵N-polarization at room temperature, whereas catalyst (1) requires significantly elevated temperatures to achieve comparable exchange rates leading to maximum polarization. Based on these insights it is reasonable to expect ¹⁵N polarization in water to decrease at even higher temperatures in analogy to methanol, as shown in Fig. 2C.

As seen in Fig. 2D, the methanol and water systems show very similar responses to B_evo at their respective optimized temperatures (22 °C and 72 °C). The response curves originate from two distinct matching conditions associated with overpopulation in ¹⁵N-α or ¹⁵N-β, giving either positive or negative NMR signal with identical polarization levels.²⁶ The matching conditions are given by⁷

$$B_{\text{evo}}=\pm \frac{J_{\text{HH}}+J_{\text{NH}}/2}{{\gamma }_{\mathrm{H}}-{\gamma }_{\mathrm{N}}}$$ (Eq. 1)

where J_HH is the hydride-to-hydride J-coupling (~ 10 Hz) and J_NH the hydride to ¹⁵N coupling (~ 20 Hz) in (1). Experimentally, we observe maxima at B_evo ≈ ±0.5 µT which is slightly higher than the ±0.3 µT predicted from Eq. 1, as the limited lifetime broadens the matching conditions.

Taken together, the observations of Fig. 1 (A–D) suggest, that the activation energy of substrate dissociation from (1) is significantly larger than for the established [IrCl(IMes)(COD)]-methanol systems. This is also supported by the fact that catalyst (1) in methanol at RT did not yield any enhancement.

The absolute polarization level in D₂O is about one order of magnitude smaller than for the methanol system, when compared at their respective optimized temperatures (¹⁵N-CHCA in D₂O, P(¹⁵N) = 0.13 %, ¹⁵N-CH₃CN in d4-MeOH, P(¹⁵N) = 1.7 %). Interestingly, this difference in hyperpolarization level can simply be attributed to the difference in hydrogen solubility (factor 5) and the difference in solvent concentration (c(H₂O) = 55 mol/L, c(MeOH) = 28 mol/L, factor 2).

Current experimental data and theoretical considerations indicate that SABRE polarization levels are limited by the exchange of hydrides on the iridium center and the exchange kinetics of other ligand types (substrate/solvent), as well as both pressure and flow rate of para-hydrogen. Exchange of hydrogen restores the polarization source to the active complex species and process proceeds via the mixed classical non-classical hydride [Ir(H)₂(η-H₂)(IMes)L₂], with arbitrary ligands L.^{13, 15} Formation of this species requires collision between a 16-electron complex and a hydrogen molecule, where collision with a para-hydrogen molecule may refresh the active species. As a result, the polarization is proportional to the concentration of para-hydrogen in solution, not the saturation concentration of hydrogen (ortho + para). Accordingly pressure dependence of polarizations is relatively weak, whereas dependence on the flow rate is significant. Depending on system composition a linear or exponential dependence of ¹⁵N polarization on the flow rate was reported.^{26, 36–37} We conclude the para-hydrogen enrichment in solution is limited by the exchange at the gas-liquid interface.

Let us now consider the substrate exchange process. The rates of ligand dissociation k_diss and association k_asso determine not only the lifetime of the complex where polarization transfer from the hydrides to the target nuclei occurs, but also the concentration of the 16-electron species required for the hydride exchange.¹³ As a result ¹⁵N polarization depends directly on the concentration of the 16-electron species. Accordingly, largest polarizations are observed at relatively low catalyst concentrations and high catalyst loadings. It is noteworthy that an exponential dependence of polarization on the substrate concentrations has been observed by Appleby et al.³⁸

We point out that all reported polarization levels are not optimized with respect to sample composition, concentrations, hydrogen pressure or flow rate. Optimization of catalyst concentration and loading afforded an 8–10-fold increase of ¹⁵N polarization level for the methanol system. Maximum polarizations are recorded at low catalyst concentrations and high catalyst loadings (¹⁵N-nicotinamide P(¹⁵N) = 7 % ¹⁵N-benzonitrile P(¹⁵N) = 16 %, metronidazole at natural abundance P(¹⁵N) = 20%).^{7, 13, 17, 36, 39} We conclude, that ¹⁵N polarization can be increased by at least a factor 10 by using low substrate concentrations and high catalyst loading. Further improvements are expected by modifications to the experimental setup to allow for more effective mixing of hydrogen and solvent at higher pressures.

¹⁵N Relaxation times in water

Of particular importance for hyperpolarization applications is the spin lattice relaxation time T₁, which defines the viable time delay between preparation of hyperpolarization and detection. We examined the T₁ lifetime for ¹⁵N-Nicotinamide³⁹ and ¹⁵N-CHCA, which constitute biocompatible compounds and contain ¹⁵N in chemically different environments.^39–41 Table 2 shows the ¹⁵N-T₁ relaxation times in D₂O, which at 1 T exceed 1 min for both compounds.

Table 2.

¹⁵N T₁ times of 100 mM Nicotinamide and 50 mM CHCA in D₂O at different detection fields.

	T1 [s] 1 T	T1 [s] 8.45 T
15N-CHCA	71 ± 15[a]	24 ± 3
15N-Nicotinamide	116 ± 10	32[b] ± 4

^[a]detected and stored at 1 T. Control by detection at 8.45 T: T₁ = 68 ± 2 s.

^[b]In H₂O: 32 ± 5.5 s

For ¹⁵N-Nicotinamide at 8.45 T we find the effect of proton containing solvent (H₂O) on the T₁ time to be negligible. It should be noted that the ¹⁵N T₁ time of nicotinamide at 8.45 T and room temperature is close to the T₁ reported for ¹³C in the ¹³C(1)-pyruvate markers currently in clinical use for prostate cancer diagnostics (T₁ = 29.2 s in-vivo, T₁ = 60 s, ex-vivo, 3 T).^8–9 It is noteworthy, that the ¹³C T₁ values in-vivo are smaller than ex-vivo, characteristic for diffusion in constricted environments.

To elucidate this field dependence in more detail we hyperpolarized ¹⁵N-CHCA and held the sample at different fields for variable times prior to detection. The results are shown in Fig. 3 displaying ¹⁵N-relaxation time of CHCA (50 mM, pH 7.5, D₂O) at with different magnetic fields. For this compound relatively low magnetic fields of about 0.2 T give the longest relaxation times. This is an intriguing finding in the context of low-field approaches to NMR and MRI^42–43, which could be coupled with SABRE to establish low-cost spectroscopy and molecular imaging.

Figure 3.

¹⁵N T₁ time constant of CHCA as a function of the magnetic field. The sample is hyperpolarized and stored at a given field for an incremented delay time and detected at 8.45 T.

The scaling of signal-to-noise with magnetic field strongly depends on the exact experimental conditions. For traditional thermal NMR, signal is proportional to polarization and the induction. Both terms are proportional to B₀, thus the signal scales with B₀².^{31, 44} In NMR, coil noise is typically dominant, which scales as B₀^1/4, hence signal-to-noise (S/N) is proportional to B₀^7/4.^44–46 However, with a hyperpolarized sample spin polarization is independent of B₀ and thus, S/N scales with B₀^3/4.

Another scenario arises for human MRI. Here, dielectric losses dominate, which are proportional to B₀. Thus S/N only increases proportional to B₀ for thermal MRI experiments.^{45, 47} Therefore, S/N is expected to be independent of B₀ for hyperpolarized human MRI.^48–49 MRI in low magnetic fields has significant advantages, as magnet and RF-circuit design are flexible, easy to construct, and relatively inexpensive.^49–50 For example, high performance ¹H-MRI at 6.5 mT with thermal magnetization has already been reported.⁴² It is noteworthy that recent advances in the low field domain, such as “External High-Quality-factor-Enhanced NMR” (EHQE-NMR)⁵¹ and others⁵² lead S/N independent of B₀ even for spectroscopic applications.

Materials and Methods

Solutions of substrates in D₂O/H₂O were added to [IrCl(IDEG)(COD)] (IDEG = 1,3-bis-(3,4,5-tris(diethyleneglycol)benzyl))imidazole-2-ylidene), COD = 1,5-cyclooctadiene), stirred until a homogeneous solution of known concentration in catalyst is obtained, and transferred to a 5 mm medium wall pressure NMR tube (Wilmad 524-PV-7). The typical sample volume was 350 µL. The solution was bubbled with argon for 30 minutes, pressurized with 10 bar of para-H₂ and hydrogen flow adjusted to obtain adequate bubbling. Catalyst activation times were 0.25–12 h depending on substrate, solvent (deuterated solvents require longer activation times), and temperature. Catalyst activation can be sped up significantly by raising temperature. For SABRE SHEATH experiments para-H₂ (Bruker BPHG 090, 38 K, 90%) was bubbled through a sample placed in a µT magnetic field. Hyperpolarization buildup is achieved in 0.5–2 min. The µT field is generated by a small solenoid inside a magnetic shield (see Scheme 1). The sample temperature was controlled with a water bath inside the magnetic shields. Measurements were performed with a Bruker Avance DX 360 (8.45 T) or Magritek Spinsolve ¹H/¹⁵N Spectrometer (1 T). Enhancements are calculated relative to neat ¹⁵N labeled pyridine. The concentration in the samples was monitored by ¹H spectroscopy.

Conclusions

We have demonstrated SABRE SHEATH hyperpolarization of 15N in aqueous media at moderate temperatures (20 – 80 °C) and achieve up to 1000-fold enhancements over thermal measurements at 8.45 T. We applied SABRE-SHEATH in water to biocompatible marker groups in different molecules (CHCA, nicotinamide, diazirine-moieties). Hyperpolarization of ¹⁵N-nitrile and the ¹⁵N2-diazirine exemplifies how SABRE-SHEATH is amendable to more substrate classes because ¹⁵N is closer to the hyperpolarization source than protons in the molecular backbone.

Furthermore, we demonstrated T₁ times comparable to, or exceeding, clinically used DNP tracers.^8–9 For example, nicotinamide in D₂O exhibits a ¹⁵N relaxation time of 2 min, which is significantly longer than typical ¹H-T₁ (seconds) of traditional ¹H-SABRE substrates. Still, recent advances have demonstrated long lived ¹H singlet states with decay times of up to 4.5 min.³⁰ When such strategies are translated to ¹⁵N, lifetimes in excess of 20 min become available.⁵³

Imaging applications of SABRE hyperpolarized protons²⁴ as well as nitrogen-15 have already been reported.³⁶ Hyperpolarized heteronuclei are beneficial as they are background free and have a large chemical shift range which allows for easy chemical identification. Future developments may be expected to advance SABRE to in vivo molecular imaging complementing DNP-hyperpolarized ¹³C tracers, which have quickly become an essential and routine tool giving detailed and fundamental insight into in vivo metabolism and biochemistry.^{8–9, 54–58}