■ INTRODUCTION

Hyperpolarized magnetic resonance imaging (MRI) is emerging as a technique to track biomolecular metabolism without radioactive labels or ionizing radiation. 1 Hyperpolarized (HP) MRI is currently under investigation in clinical trials to gain insights and diagnose metabolic disease states such as cancer, 1 diabetes, 2 or cardiovascular disease. 3,4 HP pyruvate is a leading candidate as a metabolic marker due to its safety and its central role in metabolism. 1 Through measuring pyruvate metabolism, striking advancements have been made in the detection of cancer cells in prostate, 5,6 breast, 7 and brain 8 tissues. 9 However, the leading method to hyperpolarize pyruvate, dissolution dynamic nuclear polarization (d-DNP), is limited in broad availability due to its high cost (≈$2.5M), long contrast agent production times (≈30 min or more), and instrument complexity. 10,11 In contrast, Signal Amplification By Reversible Exchange (SABRE) 12 is a fast (≈20 s), cheap (≈$25,000), and scalable hyperpolarization technique using parahydrogen (p-H 2 ) as a source of spin order to directly hyperpolarize small molecules in solutions, including pyruvate. [12][13][14][15] The hyperpolarization of heteronuclei (e.g., 13 C) is optimized in magnetic shields that establish microtesla magnetic fields, called SABRE in Shield Enables Alignment Transfer to Heteronuclei (SABRE-SHEATH). [16][17][18] Previous work has demonstrated 13 C pyruvate hyperpolarization with SABRE-SHEATH, but it remained limited in polarization relative to the high values of DNP (1.7% vs ≈60%). 13,14 Here, we present a combination of advances including the use of temperature cycling to overcome the in vivo polarization threshold of 10% with SABRE-SHEATH. The results of this study also indicate that further optimization is possible to maximize the critical molar polarization, defined as the product of concentration and polarization (introduced by Shchepin et al. 19 and Knecht et al. 20 ), ultimately the most important hyperpolarization parameter required for in vivo translation. 5,11,21 Previous demonstrations of parahydrogen induced polarization with side arm hydrogenation (PHIP-SAH) on pyruvate have shown the feasibility of in vivo studies. 22,23 These experiments demonstrate that an initial 13 C polarization of 10%, which was purified to give a 35 mM 3.5% polarization solution at the time of injection, is sufficient for in vivo chemical shift MRI. 22 PHIP-SAH involves synthesis of a propargyl pyruvate precursor, hydrogenation, complex spin transfer, hydrolysis, and phase transfer steps to obtain HP pyruvate. 22 In contrast, the facile nature of SABRE enables direct hyperpolarization of the 13 C spins in pyruvate with reduced complexity. Figure 1 highlights the catalytically active species originally described by Iali et al., where optimized hyperpolarization levels of [1-13 C]-pyruvate reached 0.96%, 13 substantially below the 13 C polarization achieved with DNP 6 or the PHIP-SAH methods. 22,24 In the present work, we highlight that sufficiently fast p-H 2 exchange still occurs in the complex at low temperatures to efficiently polarize bound pyruvate. Using this feature, we implement time-dependent temperature gradients with SABRE-SHEATH on [1-13 C]-pyruvate to reach P 13 C ( 13 C polarization) of 10.8% on free pyruvate in solution, which is over 6 times greater than previous optimized results (Figure 2A). Additionally, this figure is on par with the initial polarization achieved on allyl pyruvate with SAH-PHIP, 22 indicating that with simple purification methods 20,24 a viable biocompatible injectable for in vivo imaging could be produced. This is enabled by starting with P 13 C ≈ 20% on catalyst-bound pyruvate at lower temperatures. We also provide detailed insights and a kinetic model to describe exchange dynamics and relaxation processes during temperature gradients, which modulate substrate and hydride exchange rates. As detailed below, further optimization yields even greater molar polarization levels (concentration × %P) as needed for in vivo studies. 9,20,22 ■ RESULTS AND DISCUSSION

The spectrum and results shown in Figure 2A are the maximum achieved single-shot polarization. These data show 14.3% polarization on bound pyruvate (after warm-up) and 10.8% on free pyruvate, corresponding to a total polarization of 11.8%. The calculation of %P uses the reference signal displayed in Figure 2B and is detailed in the Supporting Information. This result is enabled by (a) the use of a high catalyst to substrate ratio (5 equiv of pyruvate, 3.3 equiv of DMSO), as done in previous work; 14,25 (b) the use of [1-13 C]pyruvate (both [2-13 C]-pyruvate and [1,2-13 C 2 ]-pyruvate give lower polarizations under identical conditions); and (c) precooling to slow exchange followed by bubbling at elevated temperature causing a time-dependent temperature gradient. To ensure reproducibility, we conducted the same experiment five times on different days and obtained an average of 10 ± 1% polarization on free pyruvate (see the Supporting Information). On bound pyruvate, polarization levels approaching 20% are observed when bubbling at even lower temperatures as detailed below.

Using the high polarization on [1-13 C]pyruvate, we acquired a 13 C image, shown in Figure 2C, utilizing a fast spin echo sequence at 1.5 T of a cryogen-free MRI system that can be operated at any field between 5 mT and 3 T. At the clinically relevant field of 1.5 T, we imaged the sample directly in an NMR tube with submillimeter resolution. The HP signal enables 3D multislice 13 C-imaging of the 3.45 mm crosssectional area of the NMR tube (full details are provided in the Supporting Information). As can be seen in the images, even the small, submillimeter sized capillary can be resolved in the images.

To characterize the temperature dependence of the hyperpolarization, we conducted the experiments depicted in Figure 3A-E. We used a pneumatic shuttle, 26 where the sample is first cooled in the probe and subsequently shuttled out of the cooled atmosphere into magnetic shields for SABRE-SHEATH. 13,27,28 Figure 3E shows the change in sample temperature as a function of bubbling time when starting at a sample temperature of 0 °C. The temperature was assessed with an internal methanol thermometer (see Supporting Information).

At low initial temperature, the slower exchange promotes efficient polarization buildup on the catalyst-bound pyruvate. This effect is evidenced by up to 20% polarization on the catalyst-bound pyruvate achieved by starting at the lowest temperature of -10 °C (see Figure 3C). With only 15 s of bubbling (Figure 3B), the polarization remains almost exclusively on the bound species 3b. In contrast to previous work, 14 our data suggest that, at low temperatures, efficient hydrogen exchange still occurs on 3b and 3a species, yet at a sufficiently slow rate to allow the weak hydride- 13 C couplings to pump large degrees of polarization onto bound 13 C pyruvate, which barely exchanges. As the sample warms during the bubbling period, [1-13 C]-pyruvate can exchange off the catalyst more rapidly while SABRE continues, albeit with reduced efficiency, ultimately leading to high polarization on free pyruvate. As is evident from Figure 3B,C, at even further elevated temperatures the free and bound polarization numbers equilibrate due to efficiently exchanging polarization pools. To unequivocally confirm that experiments with a temperature gradient give higher polarization than experiments with constant temperature, we conducted the study shown in Figure 3F. For this direct comparison, we had to use manual sample transfer experiments with a 1.1 T benchtop NMR spectrometer. In these experiments, the sample is either bubbled in a water bath at constant temperature in a magnetic shield (purple, Figure 3F) or first precooled in a water bath at a set temperature and then bubbled in the shield at ambient temperature (green, Figure 3F). The constant temperature experiments consistently stay below the experiments with temperature gradients: lower relative polarization values in these data, compared to automated shuttling, are due to unavoidable inconsistencies in slow manual sample transfer. Additionally, the temperature gradient experienced in this setup is different from that in shuttling experiments. In the shuttling system, the sample is bubbled in an atmosphere at ≈14 °C (see Figure 3E), while in the manual transfer experiments the sample is just moved into a room temperature (≈23 °C) atmosphere.

Relaxation and polarization buildup data shown in Figure 4 additionally support and characterize the described dynamics. We fit the data to a two-state (bound-free) model, which we developed inspired by previous work. 29 First, for relaxation dynamics in the absence of a pumping term (Figure 4A), the model takes into account chemical exchange and relaxation of the bound and free pyruvate species.

Here P B and P F are the free and bound polarizations, k is the pyruvate exchange rate, and ρ B and ρ F are the relaxation rates of the free and bound pyruvate species. Solving the system of differential equations yields a fitting function for the bound and free spin relaxations. The full derivation of the fitting functions is given in the Supporting Information.

After solving these differential equations, we use the resulting model to fit the relaxation data in Figure 4A. At t = 0 the bound polarization exceeds the free polarization. The difference of the two is illustrated by the purple curve in Figure 4A. Initially, bound polarization decreases quickly because of exchange and relaxation. In contrast, the free polarization only experiences a very slow initial decrease because of the exchange with the highly polarized bound species. After about 8 s, the free polarization surpasses the bound polarization due to faster relaxation of the bound species.

A similar model is used to fit the polarization buildup data displayed in Figure 4B. The only difference is that we introduce a temperature (i.e., bubbling time) dependent polarization pumping rate, Γ,

where eq 4 is identical to eq 2. The present model for Γ is purely empirical with fit parameters b, a, and τ. After this new set of differential equations is solved, only k, b, a, and τ are used as fit parameters. ρ B and ρ F are used as extracted from the relaxation data (see the Supporting Information for details). With this model, the fits explain the rapid initial buildup of bound polarization where pumping is efficient yet pyruvate exchange is inefficient. We point out that, without a temperature-dependent pumping rate Γ, the resulting models cannot represent the data in any reasonable way even if a temperature-dependent k is used. It appears that Γ has a larger temperature dependence than the pyruvate exchange k. In forthcoming work, we will examine this question and characterize the activation parameters of both hydrogen and pyruvate exchange in full. In the current absence of activation enthalpy and entropy and without knowledge of the exact Jcoupling values that drive polarization transfer from p-H 2 to bound [1-13 C]-pyruvate, the empirical model for Γ gives valuable information, showing that hydrogen exchange becomes too fast at elevated temperatures to effectively drive SABRE. Therefore, temperature cycling solves the conundrum of having to optimize both hydrogen exchange and substrate exchange simultaneously.

Finally, in Figure 4C we illustrate that p-H 2 is not the limiting substrate at pressures above 75 psi for the investigated sample composition of 6 mM Ir-IMes catalyst, 20 mM DMSO, and 30 mM [1-13 C]-pyruvate. This graph implies that at higher substrate and catalyst concentrations the p-H 2 pressure can be increased to maintain the same polarization levels while boosting the ultimately important molar polarization. This insight is further stressed by the results displayed in Figure 4D, which demonstrate the scalability of 13 C pyruvate polarization from 30 mM [1-13 C]-pyruvate (where all the previously discussed results were obtained) to 60 mM [1-13 C]-pyruvate (maintaining the same ratios of catalyst and DMSO). Doubling of the concentration actually leads to an increase in polarization, more than doubling the molar polarization, indicating that 60 mM [1-13 C]-pyruvate may be the ideal concentration for future studies. Shifting to an even higher concentration yielded slightly reduced polarization; however, these higher concentrated samples are likely to be parahydrogen limited (see Figure 4C), so higher p-H 2 pressures may return similar polarization levels while boosting molar polarization.

■ CONCLUSION

In summary, we demonstrated a high (11.8% weighted average) total polarization for [1-13 C]-pyruvate and 10.8% free polarization, paving the way for further optimization and significantly enhancing the feasibility of in vivo work. Specifically, the facile and robust nature of SABRE hyperpolarization relative to other hyperpolarization methods makes it an easily scalable technology. In these results, we emphasize the role that spin system, sample composition, and temperature gradients play in achieving high polarization levels. A kinetic model of differential equations was used to rationalize the high polarization levels. We used these high polarization levels to acquire multislice HP 13 C images with a cryogen-free MRI system operated at 1.5 T. This achievement indicates that it is possible to combine low-cost hyperpolarization with lowcost MRI to achieve high-sensitivity molecular imaging. Future work will focus on partnering these methods with previous demonstrations of catalyst extraction 30 or phase switching 31