Long-Lived ¹³C₂ Nuclear Spin States Hyperpolarized by Parahydrogen in Reversible Exchange at Micro-Tesla Fields

Abstract

Parahydrogen is an inexpensive and readily available source of hyperpolarization used to enhance magnetic resonance signals by up to 4 orders of magnitude above thermal signals obtained at ~10 T. A significant challenge for applications is fast signal decay after hyperpolarization. Here, we use parahydrogen based polarization transfer catalysis at micro-Tesla fields (first introduced as SABRE-SHEATH) to hyperpolarize ¹³C₂ spin pairs and find decay time constants of 12 s for magnetization at 0.3 mT, which are extended to 2 minutes at that same field, when long-lived singlet states are hyperpolarized instead. Enhancements over thermal at 8.5 T are between 30 and 170 fold (0.02% to 0.12% polarization). We control the spin dynamics of polarization transfer by choice of μT field allowing for deliberate hyperpolarization of either magnetization or long-lived singlet states. Density functional theory (DFT) calculations and experimental evidence identify two energetically close mechanisms for polarization transfer: First, a model that involves direct binding of the ¹³C₂ pair to the polarization transfer catalyst (PTC), and second, a model transferring polarization through auxiliary protons in substrates.

Graphical Abstract

Nuclear spin hyperpolarization is an intriguing research area, because of its ability to enhance nuclear magnetic resonance (NMR) and magnetic resonance imaging (MRI) signals by multiple orders of magnitude.^1–5 Hyperpolarization methods are particularly useful if they can enhance signals from heteronuclei such as ¹³C or ¹⁵N because they can be installed in a wide range of biomolecules, and they retain hyperpolarization on extended timescales.^6–14 At the same time, hyperpolarization of protons also has particular advantages, which stem from higher sensitivity and 100% natural abundance. A particularly simple hyperpolarization technique is para-H₂ induced polarization (PHIP).^15–16 Especially, when implemented as Signal Amplification By Reversible Exchange (SABRE) it allows for continuous and rapid hyperpolarization directly in solutions.^17–18 In the SABRE procedure, para-H₂ and the target (i.e. to-be-hyperpolarized) molecules bind reversibly with an iridium-based hexacoordinate catalyst¹⁹. At specific magnetic fields, polarization will transfer from para-H₂ to spins on the target molecule driven by J-coupling interactions, for example ~6.5 mT is ideal to hyperpolarize proton spins.^17–18 On the other hand, heteronuclei (e.g. ¹⁵N, ³¹P, ¹³C) are best magnetized in microTesla fields established in magnetically shielded environments, ^20–22 an approach that was coined SABRE-SHEATH (SABRE in Shield Enable Alignment Transfer to Heteronuclei).

However, if the goal is to hyperpolarize long-lived singlet states,^{6, 23–27} the picture changes slightly because the conditions for the transfer of scalar order have a different field dependence. For example, it has been shown that the singlet state of the ¹⁵N₂ spin pair of diazirines is hyperpolarized over a relatively wide range of magnetic fields between a few μT to about 100 mT.²⁸ These hyperpolarized nuclear spin singlet states of ¹⁵N₂ diazirines display relaxation time constants of above 20 minutes. Similarly, SABRE was used to hyperpolarize long-lived singlet states on ¹H₂-pairs,^29–30 where polarization decay time constants of above 4 min were observed.³¹ Such long hyperpolarization lifetime promises biomolecular tracking and imaging of low concentration analytes on significantly extended timescales. In this article, we use SABRE-SHEATH, to hyperpolarize magnetization as well as long-lived nuclear singlet states in carbon-13 spin pairs and find lifetime T₁ of 12 s for magnetization and T_S of 2 min for long-lived singlet states at 0.3 mT. Here it is important to note, that the current record of a long-lived singlet state is held by a ¹³C₂ spin pair (hyperpolarized by DNP, not SABRE) with lifetime, T_S, of more than one hour.⁶

For the presented experiments, we designed two molecules with various isotopic labeling schemes. We synthesized 1,2-(4-pyridyl) acetylene, with symmetric structure, and 1-phenyl-2-(4-pyridyl) acetylene, with asymmetric structure. For both, we consider isotopomers with naturally abundant ¹³C, as well as doubly ¹³C labeled substrates at the triple bond. The results presented in Figure 1 indicate that the acetylene carbon spins as well as the aromatic bridge carbon spins are hyperpolarized. The enhancements are between 30 to 170 fold (0.02% to 0.12% polarization), when compared to thermal signals acquired at 8.45 T. The molecules with ¹³C at natural abundance show 2–3 times higher enhancements compared to ¹³C enriched sites. This is likely due to faster T₁ relaxation in ¹³C₂ pairs as opposed to T₁ of isolated ¹³C spins. An additional cause may simply be the higher ratio of polarization source (p-H₂) to target spins in the naturally abundant case.³²

Figure 1.

¹³C spectra of naturally abundant (A&B) and ¹³C labelled (C&D) substrates used in experiments. A&C show results for the symmetrically substituted 1,2-2 pyridyl acetylene. B&D are from the asymmetrically substituted 1-phenyl-2-(4-pyridyl) acetylene. For the naturally abundant substrates the bridge carbons on the pyridyl rings (3, 4 in A, 3 in B) show significant enhancement, while the one on the benzene ring (4 in B) is only slightly hyperpolarized. The ¹³C-¹³C coupling, J_CC, read from the line-splitting in panel D is 185 Hz. (The SI also provides a thermal spectrum in Fig. S2.)

The hyperpolarization transfer from para-H₂ to these substrates occurs via iridium based polarization transfer catalysts (PTC’s). We used the standard precatalyst [IrCl(IMes)(COD)], (IMes = 1,3-bis(2,4,6-trimethylphenyl)imidazole-2-ylidene; COD = cyclooctadiene).^18–19 We used substrate concentrations of 30 mM or 160 mM, and catalyst concentrations of 2 mM or 10 mM for the symmetric and asymmetric compounds respectively. The solvent was methanol-d₄, and the pre-catalyst was activated by bubbling para-H₂ through the sample for 15 minutes at a pressure of 7 bar and a fractional parahydrogen enrichment of ~85%. Thereafter, hyperpolarization was performed according to the SABRE-SHEATH procedure:^{10, 20, 28} the sample is exposed to para-H₂ in a magnetically shielded environment outfitted with a small solenoid coil to obtain a controllable μT magnetic field. One minute of exposure to para-H₂ is sufficient to equilibrate polarization. Subsequently, the sample is transferred manually as quickly as possible (~ 8 s) to a Bruker 360 MHz (8.45T) magnet for read out. The manual transfer time of 8 s is relatively consistent, with variations of ~1 s.

The polarization transfer occurs in catalytically active PTC’s. Two possible, energetically low PTC species are depicted in Figure 2. The ground state energies were determined by density functional theory calculations using the all-electron FHI-aims code.³³ The geometries were optimized using the PBE parameterization of exchange and correlation³⁴ with a van der Waals correction³⁵ and the tier 2 basis sets^{33, 36}. Scalar relativity was handled in the atomic ZORA approximation. ³³ Additional possible configurations and the corresponding PTC-energy landscape are provided in the Supplemental Information (SI). Furthermore, we provide ¹H-NMR spectra of the hyperpolarized hydrides bound to the Iridium center demonstrating the presence of at least two catalytic species. In this first study, we were not able to detect hyperpolarized ¹³C signals from molecules bound to the Iridium molecules. Therefore, we rely on more indirect evidence coupled with ab initio calculations to determine likely PTC structures.

Figure 2.

Two possible polarization transfer catalysts (PTC’s). Top: 3D models obtained after energy minimization in the all electron code FHI-aims. Bottom: Structural formulas of the PTCs for clarity. A) The substrate is bound to all Ir binding sites via nitrogen. B) One of the molecules’ triple bond binds to the iridium catalyst, which has a higher energy than the structure in A. DFT calculations reveal that the energy difference between the two proposed complexes is relatively small (0.09 eV). Other possible complexes (with higher energies) are discussed in the SI.

In the first PTC model (Figure 2A), all substrate molecules bind to the Ir center via nitrogen. This is the energetically lowest PTC species identified by us. Here, polarization transfers from para-H₂ to the pyridyl protons first and finally arrives at the acetylenic carbons. In the second PTC model (Figure 2B), the catalyst binds with the triple bond and polarization is transferred directly to ¹³C sites.

The spectra displayed in Figure 1 could quickly lead to the conclusion, that the active PTC must be the directly binding model (Figure 2B), because we do not observe hyperpolarization from the ring carbons, other than from those in the bridge to the acetylene bond. Moreover, we observe hydrogenation, which most certainly requires binding of the triple bond to the iridium center. Hydrogenation rates depend on the ratio of substrate to catalyst: at 3:1 hydrogenation completes in less than 30 minutes, however at above 15:1, hydrogenation takes more than 12 h. In a single SABRE experiment (with 1 min of bubbling) we estimate significantly less than 1% hydrogenation at the 15:1 ratio, which was used for most experiments. We even observe hyperpolarized hydrogenation products that display typical ALTADENA type enhancements due to incorporation of para-H₂ (spectra provided in the SI), still, keeping in mind that the displayed spectra of Fig. 1 must all result from SABRE as they are uniquely associated with the intact, non-hydrogenated substrates.

However, these conclusions may be premature. First, the PTC on of Figure 2A is energetically lower. Furthermore, notice that all non-detected ¹³C spins are directly bound to protons. This leads to much faster ¹³C relaxation (a typical T₁ relaxation time for aromatic ¹³C directly bound to a proton is ~5s see Ref.³⁷, whereas T₁ relaxation constants of the bridge carbons are found to be 11(1) s and T₁ of the acetylenic carbons is 12(0.5) s at low fields) with two important consequences. First, the hyperpolarization buildup at these ¹³C sites will be much less efficient, and second, a small amount of hyperpolarization may quickly relax during the ~8 s sample transfer from polarization region into the magnet. In addition, we performed SABRE under optimized condition for ¹H polarization transfer at 6.5 mT, and this resulted in strong enhancement of the pyridyl ring protons, while enhancement of ¹³C were negligible and ¹H enhancements on the distant phenyl ring were much smaller. Though bound species are never observed from the ¹³C spectra, the hydride peaks are available in the SI. We observe a small chemical shift difference of the hydride peaks (~0.2 ppm, which would be much larger for the binding mode in Figure 2B based on DFT calculation). Finally, we attempted to hyperpolarize diphenyl-¹³C₂-acetylene (no ring nitrogens) in the SABRE-SHEATH mode and did not observe enhancements. All these considerations point to a strong contribution of the PTC shown in Figure 2A.

To investigate this in more detail, we performed a careful characterization of hyperpolarization transfer as function of micro-Tesla field using the doubly ¹³C labeled molecules. As depicted in Figure 3, we varied the magnetic field between −12 and +12 μT, accompanied by simulations of the hyperpolarization transfer process.

Figure 3.

Field dependent hyperpolarization for the two substrates. Panel A shows (1) a hyperpolarized magnetization spectrum hyperpolarized at 0.17 μT (and acquired at 8.45 T) for the symmetric substrate and (2) its field dependence in the μT range. Panel B shows the experimental and simulated results of creating magnetization and singlet order for the asymmetric substrate, as function of magnetic field; (1) Magnetization spectrum hyperpolarized at 0.28 μT. (2) Experimental and (3) simulated field dependence for magnetization. (4) Singlet spectrum hyperpolarized at 6.2 μT. (5) Experimental and (6) simulated field dependence for singlet order. In (B3) and (B6), the highlighted points are the local maxima for polarization transfer labeled by analytically derived resonance conditions from careful inspection of the nuclear-spin Hamiltonian provided in the SI.

The first important finding is that we can directly choose to polarize different states of the ¹³C pair: magnetization or singlet, which are easily distinguishable by their spectra. Magnetization is easily detected from both molecules (Figure 3 A1, B1), whereas singlet-order can only be detected immediately from the asymmetric 1-phenyl-2-(4-pyridyl) acetylene because the acetylenic carbons have a chemical shift difference (Figure 1C). For this asymmetric compound, the acetylenic carbons are strongly coupled at low fields (J_CC is ~185 Hz, whereas their chemical shift difference Δν_C is less than 0.5 mHz). Upon transfer to the high field in the magnet (8.45 T) for read out, the chemical shift difference becomes significantly larger than the J_CC coupling (Δν_C ~770 Hz), the carbons are now weakly coupled, and the singlet state is no longer an eigenstate. The sample transfer from low to high field transforms I₁·I₂ singlet order into detectable (I_1z – I_2z) which gives antiphase signals in a pulse acquire experiment, as shown in Figure 3 (B4) (Full analysis of singlet order transfer is provided in the SI). However, for the symmetric molecule, since the two carbons will remain symmetric at high field, the singlet state cannot be accessed immediately. In principle, access to the singlet can be accomplished by specialized pulse sequences such as singlet-to-magnetization (S2M)^38–39 or SLIC,^40–41 yet this is beyond the scope of the present work.

In order to understand the polarization transfer dynamics at micro-Tesla fields in detail, we consider resonance conditions dictated by the Hamiltonian of the doubly ¹³C labeled molecule. As detailed in the SI, at low fields of <0.6 μT, we encounter a resonance condition to polarize magnetization, given as

$$\mid {\nu }_H-{\nu }_C\mid =\pm J_{\text{HH}},$$ (Eq.1)

where ν_H and ν_C are the frequencies of protons and carbons and J_HH is the J-coupling between the two para-H₂ derived hydrides on the iridium. When solved for the magnetic field using ν = −γB we obtain the magnetization transfer field as

$$B_{\text{trans}}=\pm J_{\text{HH}}/({\gamma }_H-{\gamma }_C),$$ (Eq. 2)

where γ_H=42.577 Hz/μT and γ_C=10.705 Hz/μT. When the field is increased to a few μT, additional resonance conditions to create magnetization and/or singlet are encountered. The Hamiltonian reveals overlapping conditions to create magnetization and singlet given as

$$\mid {\nu }_H-{\nu }_C\mid =\pm (J_{\text{CC}}\pm J_{\text{HH}}),$$ (Eq. 3)

where J_CC is the acetylenic ¹³C J-coupling. Again, solving for the transfer field we obtain:

$$B_{\text{trans}}=\pm (J_{\text{CC}}\pm J_{\text{HH}})/({\gamma }_H-{\gamma }_C)$$ (Eq. 4)

Equations (1,2) and (3,4) fully encompass the behavior observed in Figure 3. In the low field region, maximum magnetization transfer is observed at ~ ± 0.34(0.1) μT, whereas there is negligible singlet buildup. At slightly elevated fields, both magnetization and singlet have local maxima/minima at ~ ±5.6(0.2) μT and ~ ±6.4(0.2) μT (see Fig. 3B). These values are consistent with J_HH ~ 10(3) Hz, and J_CC ~190(5) Hz. J_CC can also be estimated from the hyperpolarized NMR spectrum of the free form where we find J_CC = 185 Hz (see Fig. 1 panel D).

By numerical simulations of the spin dynamics we confirm that the general behavior is largely independent of the polarization transfer mechanism (direct to ¹³C₂ (Fig. 2B) vs. indirect via auxiliary protons (Figure 2A); see SI for details). However, the numerical value of J_CC strongly depends on the exact nature of the PTC. We have performed first principles calculations of the relevant J-couplings using the FHI-aims code.³³ We used the PBE³⁴ parameterization for exchange and correlation and the fully uncontracted cc-Pv5Z⁴² basis sets (tier 2 for iridium³³). The ab initio calculations predict a J_CC of ~191 Hz for substrate bound via nitrogen (Figure 2A) vs. a J_CC of 120 Hz for substrate bound directly via the acetylenic bond (see full details in the SI). Based on the measurements shown in Figure 3, we can now conclude with more confidence that the primary PTC is the energetically favored species shown in Figure 2A because for the PTC in 2B we would expect efficient hyperpolarization at significantly lower fields of 3.5±0.3 μT, which is not observed.

Finally, since the asymmetric molecule allows for easy read out of, both, singlet state and magnetization, we can measure their lifetimes T_S or T₁. As displayed in Figure 4, we measured T_S at 0.3 mT and 50 mT, and fit with exponential decay constants of 117(7) s and 69(4) s respectively. For comparison, we measured T₁ at the field where it has longer T_S (0.3 mT) and find that magnetization decays much more rapidly with exponential decay constant T₁ of 12(.5) s. The T₁ lifetime of the ¹³C₂ pair at 8.45T is measured as 8(0.4) s.

Figure 4.

T₁ and T_S measurements of 1-phenyl-2-(4-pyridyl) acetylene. For all measurements, the sample was first hyperpolarized in the shield using 0.4 μT (polarize magnetization) / 6 μT (polarize singlet order) then positioned at 0.3 mT or 50 mT. After varying delay times the sample was transferred to the magnet quickly to measure the remaining signal. The data points were sampled randomly to eliminate the effect of the slow triple bond hydrogenation, and the lifetime constants were obtained using single exponential fit.

In conclusion, we demonstrated that, both magnetization and long-lived singlet order can be induced on ¹³C₂ using SABRE-SHEATH. Hyperpolarization lifetime is extended to ~2 minutes, or 10 times T₁. Furthermore, we describe direct hyperpolarization of long-lived singlet order by SABRE-SHEATH when the J-coupling in the targeted spin pair is much larger than the J_HH coupling between the hydrides. This is in contrast to the first demonstration, of heteronuclear (¹⁵N₂) long-lived states hyperpolarized by SABRE, where J_NN and J_HH were comparable in size leading to a resonance condition that is matched at a broad range of fields,²⁸ raising the question if long-lived states could be hyperpolarized when J_NN or J_CC are much larger. Here we have shown that specific μT-fields work in that case. Hyperpolarization levels and enhancements remained relatively low in this first demonstration. A likely culprit are the quadrupolar ¹⁴N nuclei, as we are finding that quadrupoles act as highly efficient polarization sinks at μT fields. Therefore, we could likely boost hyperpolarization by additional ¹⁵N labeling of our substrates and other strategies detailed in the literature.^43–45 Finally, we have also assembled clear evidence for at least two potential PTC species that simultaneously exist in solution and we presented arguments that lead us to believe that polarization transfer is primarily mediated indirectly via protons in the substrates. Overall, the presented results illustrate an avenue towards simple and fast hyperpolarization of long-lived ¹³C hyperpolarization with potential applications in biomolecular MRI or the observation of slower processes by hyperpolarized NMR. The presented advances can be translated to biomolecules already shown to be amenable to heteronuclear SABRE hyperpolarization including nicotinamide^{20, 46}, in vivo pH sensor imidazole⁴⁷, hypoxia sensor metronidazole⁴³ and others.^{10, 48} While the current work was performed in methanol solutions, recent advances in heterogeneous^49–50 and water-soluble^51–56 SABRE catalysis may lead to in vivo translation of the presented approach for fast hyperpolarization of long-lived ¹³C molecular probes.