PHIP hyperpolarized [1-13C]pyruvate and [1-13C]acetate esters via PH-INEPT polarization transfer monitored by 13C NMR and MRI

Parahydrogen-induced polarization of 13C nuclei by side-arm hydrogenation (PHIP-SAH) for [1-13C]acetate and [1-13C]pyruvate esters with application of PH-INEPT-type pulse sequences for 1H to 13C polarization transfer is reported, and its efficiency is compared with that of polarization transfer based on magnetic field cycling (MFC). The pulse-sequence transfer approach may have its merits in some applications because the entire hyperpolarization procedure is implemented directly in an NMR or MRI instrument, whereas MFC requires a controlled field variation at low magnetic fields. Optimization of the PH-INEPT-type transfer sequences resulted in 13C polarization values of 0.66 ± 0.04% and 0.19 ± 0.02% for allyl [1-13C]pyruvate and ethyl [1-13C]acetate, respectively, which is lower than the corresponding polarization levels obtained with MFC for 1H to 13C polarization transfer (3.95 ± 0.05% and 0.65 ± 0.05% for allyl [1-13C]pyruvate and ethyl [1-13C]acetate, respectively). Nevertheless, a significant 13C NMR signal enhancement with respect to thermal polarization allowed us to perform 13C MR imaging of both biologically relevant hyperpolarized molecules which can be used to produce useful contrast agents for the in vivo imaging applications.

www.nature.com/scientificreports/ relaxation times are substantially longer compared to those of protons 20 , and therefore 13 C sites can act as excellent carriers of hyperpolarization. For example, under in vivo conditions the T 1 relaxation time of lactate methyl protons is only 1.7 s, while the carboxyl carbon T 1 time is 30 s for [1-13 C]pyruvate 21 . This provides a sufficient time window for the preparation, administration and circulation of 13 C-hyperpolarized metabolites and for monitoring their metabolic pathways. Currently, dissolution dynamic nuclear polarization (d-DNP) 22 is the leading hyperpolarization technique for in vivo applications, with the majority of all MRI studies utilizing hyperpolarized (HP) 13 C-labeled pyruvate and acetate as well as other metabolites being performed using d-DNP 23 . However, the associated high costs 24 of the d-DNP instrumentation and its operation is a major obstacle for a much wider use of this technique. Alternatively, the technique termed parahydrogen-induced polarization (PHIP) 25 is an emerging method for hyperpolarizing 13 C nuclei 26 . One of its variations, PHIP-SAH (PHIP using side-arm hydrogenation) 27 , is an elegant approach in which a side-arm fragment of an ester is hydrogenated via pairwise parahydrogen (p-H 2 ) addition while the carboxylic moiety (e.g., pyruvate or acetate) remains unchanged. Reineri et al. produced HP acetate by hydrogenation of vinyl acetate to ethyl acetate, followed by 1 H to 13 C polarization transfer via magnetic field cycling and subsequent hydrolysis of ethyl acetate with sodium hydroxide 28 . Subsequent studies extended this approach to lactate 29 and pyruvate 30,31 . Consequently, PHIP-SAH technique rekindled the interest in p-H 2based approaches for the production of HP contrast agents for in vivo MR metabolic studies.
Upon hydrogenation of a precursor molecule with p-H 2 , hyperpolarization is initially produced for 1 H nuclei and thus needs to be transferred to a 13 C nucleus. There are two general approaches to this task: magnetic field cycling (MFC) 19,32 and application of radio-frequency (RF) pulse sequences [32][33][34][35][36] . Both approaches have their merits. In particular, the pulsed RF approach allows one to perform polarization transfer directly in the MRI scanner used for 13 C MR imaging 37 . The generation of HP compounds directly in the detection area excludes polarization losses caused by T 1 relaxation and magnetic field variation during the transfer of a HP compound to the imaging instrument. Indeed, previous studies 37,38 demonstrated the possibility to perform polarization transfer step and ex vivo and in vivo MRI at the same location.
Many RF pulse sequences are suitable candidates for the polarization transfer step. Among these, PH-INEPT is one of the most widely used pulse sequences for polarization transfer from protons to heteronuclei 39 . This pulse sequence transforms the singlet spin order derived from p-H 2 into an anti-phase magnetization of a heteronucleus. An extension of this method, PH-INEPT-PLUS, allows one to refocus the anti-phase magnetization into the in-phase magnetization 40 , which is advantageous for imaging experiments.
In this work, biologically relevant [1-13 C]pyruvate and [1-13 C]acetate esters were hyperpolarized via the PHIP-SAH technique, followed by polarization transfer from p-H 2 -derived protons to 13 C nuclei by two different approaches, the optimized PH-INEPT-type pulse sequences and the MFC protocol, for their direct comparison under identical conditions. Based on these results, 13
Ethyl [1-13 C]acetate. First, to estimate an optimal magnetic field for polarization transfer, MFC experiments were performed by changing the field inside the magnetic shield mounted on top of a spectrometer using an automated MFC setup described elsewhere 44 . Experimental details and numerical calculation procedures can be found in Supplementary Information (SI). The substrates with natural isotopic abundance were used in these experiments (Fig. 2). All data is presented normalized to the maximum value of the experimental field dependence. The resulting field dependence of the intensity of the enhanced carboxyl 13 C NMR signal of ethyl acetate on the polarization transfer field shows a single maximum at 400 nT (Fig. 2c). As clearly seen in Fig. 2c, experimental data is in a general agreement with numerical calculations.
Next, MFC experiments were repeated using 13 C-labeled vinyl acetate. The maximum observed 13 C polarization P( 13 C) was 0.65 ± 0.05%, which corresponds to 1100-fold signal enhancement (SE) at 7 T (Fig. 2b). This allowed us to obtain a 13 C image with the signal-to-noise ratio (SNR) of 300 (Fig. 2d).
For RF polarization transfer based on the INEPT pulse sequence (Fig. 3), spin echo delays τ 1 and τ 2 were numerically optimized to maximize the calculated 13 C polarization levels P( 13 C) (the detailed information on the calculation of optimal delays in the PH-INEPT-PLUS pulse sequence can be found in "Methods" section). Experimental implementation was based on the PH-INEPT-PLUS pulse sequence ( Fig. 3a) with τ 1 = 0.057 s and τ 2 = 0.284 s (Fig. 3b). This optimization allowed us to obtain 0.19 ± 0.02% 13 C polarization and SNR = 100 for a 13 C MR image (Fig. 3e). In both cases, the initial concentration of vinyl [1-13 C]acetate was 80 mM and its conversion to ethyl [1-13 C]acetate was 40%; therefore, the concentration of HP ethyl [1-13 C]acetate was 32 mM.

Allyl [1-13 C]pyruvate.
For allyl pyruvate (Fig. 4), the field dependence has a more complex appearance due to a large number of different J-couplings in the product molecule. The calculated optimal field for the polarization transfer is 200 nT (Fig. 4c), however, the experimental maximum is located at a slightly higher field. It should be noted that any such simulation is necessarily semi-quantitative at best. The fine details of the www.nature.com/scientificreports/ calculated curve may depend significantly on numerous parameters which are very difficult to match precisely between the experiment and simulation. Therefore, the simulation results provide the general guidance, while the optimum parameters are further refined in the experiments. MFC experiments were then performed using propargyl [1-13 C]pyruvate. The maximum observed P( 13 C) was 3.95 ± 0.05% (SE = 6900 at 7 T), which allowed us to obtain a 13 C image with a SNR of 2200 (Fig. 4d). The polarization level is presented for the hemiketal form because it is prevalent in a CD 3 OD solution. When vinyl moiety is used as a side-arm, two p-H 2 -derived protons are added to the carbon-carbon double bond at three to four bonds away from the carboxyl carbon, while with propargyl moiety the two protons are added at four to five bonds away from the target 13 C nucleus. Therefore, implementing the PH-INEPT-PLUS pulse sequence to hyperpolarize carboxyl 13 C atom in allyl pyruvate is problematic because of extremely small J-coupling constants between the p-H 2 -derived protons and the 13 C nucleus (Fig. 1). In this case, numerical optimization yields τ 1 = 0.786 s and τ 2 = 2.046 s, which in practice would result in a poor performance because of proton polarization decay induced by T 2 relaxation and diffusion of polarized molecules during polarization transfer. To deal with this issue, we added an additional spin-echo block to the PH-INEPT-PLUS sequence (Fig. 5a). The resulting pulse sequence, denoted as PH-ECHO-INEPT-PLUS, allows one to transfer proton polarization through an intermediate group of nuclei. In allyl [1-13 C]pyruvate the role of these nuclei is fulfilled by protons in the methylene group (-CH 2 -) of the allyl fragment, which have reasonable J-couplings to both p-H 2 -derived protons and the target carbon ( Fig. 1). Optimization of the delays was performed in the same way as for PH-INEPT-PLUS, giving the following values: τ 1 = 0.115 s, τ 2 = 0.045 s and τ 3 = 0.035 s (Fig. 5b). Using an optimized PH-ECHO-INEPT-PLUS pulse sequence (Fig. 5c), we observed a maximum P( 13 C) of 0.66 ± 0.04% (Fig. 5d), which allowed us to acquire a 13 C MR image with the SNR of 300 (Fig. 5e). The conversion of propargyl [1-13 C]pyruvate to allyl [1-13 C] pyruvate was 80-90%, i.e., higher than that obtained for [1-13 C]acetate esters (40%). The initial concentration of propargyl [1-13 C]pyruvate was 80 mM; therefore, the concentration of HP allyl [1-13 C]pyruvate was 64-72 mM.

C NMR/MRI on Bruker BioSpec 7 T instrument.
During the experiments on an animal MRI scanner, the parahydrogen fraction of only 50% was provided by a home-built parahydrogen generator operating at liquid nitrogen temperature. The concentrations of the substrates were increased from 0.08 M to 0.8 M for vinyl [1-13 C]acetate and to 0.18 M for propargyl [1-13 C]pyruvate. The additional reason to increase the concentration of the to-be-polarized substrate is a longer time for sample transfer to the animal imaging scanner compared to conventional vertical bore instruments. The resulting 13

Discussion
MFC experiments. Polarization transfer in the PHIP-SAH approach using MFC has been reported previously 28,43,45,46 . For example, in the study of Cavallari and coauthors 46 , two concentric μ-metal cylinders were used to shield the laboratory's magnetic field. The carboxyl carbon-13 polarization of 2.3% was observed. Subsequent studies with an optimized MFC profile showed 4.8% 13 C polarization for allyl lactate 29 and 5-6% for allyl pyruvate 27,30 . In another study 48 , the polarization levels obtained for ethyl [1-13 C]acetate and allyl [1-13 C] pyruvate were 4.4% and 5.4%, respectively. The difference between the values obtained in our work and the other studies can be explained by different polarization transfer protocols. In our experiments, p-H 2 was bubbled through the solution in the μ-metal shield and then transferred to the high magnetic field of the spectrometer (see "Methods" section for more details). In the other studies, the sample was hydrogenated at the Earth's magnetic field with subsequent transfer to the μ-metal shield and then to the spectrometer. In addition, in one study 48 the polarization for allyl [1-13 C]pyruvate was evaluated using the sum of the signal intensities of both pyruvate forms, while we used only the hemiketal form for this. We also note that the polarization vs. magnetic field dependences obtained in this work are different from those reported earlier 28,43,45,46 . In the MFC procedure employed in this work, the maximum polarization corresponds to the evolution of the spin system in the magnetic field which puts it the level anticrosing (LAC) 47 region, whereas an adiabatic sweep through the LAC is more efficient if started at a much lower magnetic field. Thus, evolution of the spin system in our work occurred for a substantial period of time (hydrogenation for 20 s in the experiments) in the LAC region 47 instead of an adiabatic passage through this region employed in a more conventional MFC polarization transfer procedure. While the latter can be somewhat more efficient, the MFC procedure used in our work was deemed advantageous as its experimental optimization was carried out on an automated setup capable of achieving a non-adiabatic field jump. This facilitates the comparison of the experimental results with numerical modeling of the field dependence because it avoids the variability of the conventional adiabatic MFC performed manually, and in addition simplifies the mathematical modeling of the process.

INEPT-based experiments. The efficiency of polarization transfer in the RF-based methods depends on
the J-couplings between the p-H 2 -derived protons and the heteronucleus of interest in the product molecule, with generally lower efficiency of polarization transfer for smaller heteronuclear J-coupling values 48 . Thus, a straightforward implementation of PH-INEPT for polarization transfer in acetate and pyruvate derivatives is encumbered because J-coupling values between the carboxyl 13 C nucleus and the p-H 2 -derived protons in the hydrogenated side-arm moiety are relatively small (Fig. 1). The efficient spin order transfer to heteronuclei via relayed INEPT chains (ESOTHERIC) 49 method allows one to surpass this problem by using a set of sequential INEPT steps on a specially designed precursor molecule isotopically labeled with 13 C and 2 H nuclei. For example, in the study of Korchak et al. 50 13 C polarization of ethyl acetate-d 6 obtained using ESOTH-ERIC pulse sequence was ca. 59% at the natural abundance of the 13 C isotope. Better efficiency than the one achieved in this study can be explained by a higher conversion of vinyl acetate to ethyl acetate and a higher pH 2 concentration than in our study. In addition, the substrate molecule was deuterated. In another study, Dagys et al. 51 showed 7.6% of 13 C polarization during hydrogenation of 3-butyn-2-ol-d 5 , also using the ESOTHERIC pulse sequence and a deuterated substrate. The pulse sequence contained three INEPT blocks and transferred hyperpolarization from parahydrogen protons to a relayed proton, with further transfer of hyperpolarization to the heteronucleus. This method modification was called PR-SAH (proton-relayed SAH). The authors also demonstrated the hydrolysis and phase separation, which led to 4.3% 13 C hyperpolarization of [1-13 C]pyruvate in aqueous solution. While these studies achieved better results, the need for such a specific labeling is a clear disadvantage as it drastically increases the synthetic complexity and the cost of the PHIP precursor [52][53][54] . Moreover, the introduction of deuterium atoms may have deleterious effects on polarization transfer 55 . In addition, in the case of singly 13 C-labeled molecules, it is often preferable to make the RF pulse sequence as simple as possible to reduce the undesired coherent evolution in a complex spin system. Thus, PH-INEPT-PLUS and its modifications remain more relevant for polarization transfer in molecules hyperpolarized by PHIP-SAH than the complex ESOTHERIC pulse sequence.
Potential future improvements. In the future experiments, we expect to achieve an improved catalytic conversion using high-pressure spray injection hyperpolarizer equipped with RF pulsing and in situ detection capability 56,57 , and additionally improve the polarization yields using a larger p-H 2 enrichment fraction 58,59 and an optimized sample transfer procedure. For in vivo applications, the use of aqueous medium is required instead of toxic CD 3 OD. There are several studies devoted to the observation of PHIP effects using aqueous media 60,61 . Therefore, the future improvements mentioned above and the use of aqueous medium with water-soluble catalysts or the aqueous extraction procedures will certainly provide better results and will be compatible with in vivo MRI.

Conclusions
In this work, two different approaches to polarize 13 C nuclei by the PHIP-SAH method in molecules possessing an acetate or a pyruvate moiety are reported and compared. The first one is based on the magnetic field cycling, which is the most popular approach used in the previous PHIP-SAH studies. The resulting 13 C polarization levels were 3.95 ± 0.05% for [1-13 C]pyruvate and 0.65 ± 0.05% for [1-13 C]acetate. The second approach is based on the INEPT-type pulse sequences to transfer polarization from 1 H to 13 C in a high magnetic field. It significantly simplifies the in vitro imaging procedure because no sample shuttling is required. The observed polarization levels were 0.66 ± 0.04% for [1-13 C]pyruvate and 0.19 ± 0.02% for [1-13 C]acetate. The achieved polarization levels allowed us to perform 13 C MRI on a standard research scanner at 9.4 T and on a 7 T animal imaging scanner. www.nature.com/scientificreports/ The results demonstrate the possibility to observe polarization levels that are high enough for 13 C MRI on two biologically relevant molecules using MFC and PH-INEPT-type pulse sequences in combination with the PHIP-SAH approach.

Methods
Experimental details. In the experiments, an unsaturated substrate and a catalyst were dissolved in methanol-d 4 . Vinyl acetate with natural 13 C abundance (Sigma-Aldrich) was used as received; propargyl pyruvate with natural 13 C abundance and 13 C-labeled vinyl [1-13 C]acetate and propargyl [1-13 C]pyruvate were synthesized according to the procedures reported elsewhere 42 . In most experiments, hydrogen gas was enriched with paraisomer using Bruker Parahydrogen Generator BPHG-90 operating at 43 K, resulting in 85% p-H 2 fraction. The magnetic field cycling experiments were carried out using a three-layer μ-metal shield (Magnetic Shield Corp., USA), with an additional solenoid inside it to adjust the magnetic field in the shield. NMR spectra were acquired on a 7 T Bruker AV 300 NMR spectrometer. Acquisition of the NMR spectra during MFC experiments was started 0.3 s after the placement of the sample into the NMR spectrometer using a 90-degree pulse without any signal averaging. During the experiments on the BioSpec 7 T scanner, hydrogen gas was enriched with paraisomer using a home-built setup operating at liquid nitrogen temperature, which delivers ca.  Table S1. NMR tube was placed inside the μ-metal shield with p-H 2 pressure of 2.8-3 bar. Parahydrogen was bubbled through the solution at the low rate of 30 sccm for 20 s. After that, the sample was transferred from the low magnetic field to the Earth's field (ca. 1 s), and then to the high magnetic field of the detection area. Further details on the MFC procedure are provided in the next section. In the experiments with INEPT-type pulse sequences, the hydrogenation reaction and the acquisition of spectra or images were carried out at the same high magnetic field. During the optimization of experimental conditions for NMR, multiple experiments were performed to ensure the reproducibility of the results. The reported signal enhancements are the mean values of five independent experimental results; the variations did not exceed 10%.

The dependence of P( 13 C) on magnetic field in the MFC experiment.
To measure the magnetic field dependence of the PHIP transfer efficiency, we have used a 9.4 T Bruker Avance 400 NMR spectrometer equipped with an automated field cycling setup 44 . A gas bubbling system based on computer controlled TTL valves allowed us to perform repetitive bubbling with high reproducibility 62 . The sample was bubbled for 20 s in the magnetic shield placed on top of the spectrometer magnet. The magnetic field inside the shield was controlled by adjusting the current in the electromagnetic coils positioned inside it. After bubbling, the sample was transferred to the high magnetic field of the detection area, where 13 C NMR spectrum was acquired. Before the sample transfer, a field jump of 25 μT was applied to make the field change non-adiabatic. The total delay between bubbling and detection was 1.2 s, including 0.6 s for sample transfer and 0.5 and 0.1 s waiting times with the sample in the shield and in the probe, respectively. The entire field dependence was measured with a single sample. The depletion of the substrate was taken into account by adding reference points measured for a predefined magnetic field (400 nT for vinyl acetate, 200 nT for propargyl pyruvate) in between every five points of the field dependence. This allowed us to assess the amount of PHIP produced in every measurement of the field dependence data points and to exclude the effect of chemical transformation of the substrate on the measured field dependence of signal enhancement. The temperature in the rf probe of the spectrometer was maintained at 45 °C. The parahydrogen bubbling pressure was set at 3 bar.
Calculation of optimal delays for the PH-INEPT-PLUS sequence. The simulation was carried out for the spin density matrix of two protons in a singlet state and all other nuclei being non-polarized. All nondiagonal elements of the density matrix written in the high field eigenbasis were removed reflecting the coherence damping due to averaging over hydrogenation time. The resulting matrix was subjected to a set of rotations and free evolution delays corresponding to the PH-INEPT-PLUS pulse sequence. Transverse magnetization of 13 C nuclei was calculated for various τ 1 and τ 2 delays to estimate their optimal values (Fig. 3b).
MRI on a 9.4 T vertical bore Bruker Avance III spectrometer. MR images were acquired on a 9.4 T vertical bore Bruker Avance III spectrometer equipped with microimaging accessories using commercially available 25-mm (for [1-13 C]acetate ester) and 30-mm (for [1-13 C]pyruvate ester) ID 13 C/ 1 H RF coils. All MRI experiments were performed using gradient echo pulse sequence (fast low angle shot, FLASH) 63 with one scan (no signal averaging). The total acquisition time was less than 1 s with the following parameters for FLASH MRI: repetition time (TR) of 3.1 ms, echo time (TE) of 1.5 ms, 16 phase encoding steps, 64 readout points, acquisition bandwidth 20 kHz, Cartesian spatial encoding. The durations of the phase encoding and readout gradients were 400 μs and 2.1 ms, respectively; the strength of both gradients was set at 6 G/cm. The FOV was 3.07 × 3.07 cm 2 . The initial 64 × 16 matrix was zero-filled to 64 × 64 for better presentation so that the digital in-plane resolution is similar in both dimensions. The resulting images were zoomed to eliminate the uninformative parts. As a result, 13 C images were obtained with the spatial resolution of 0. 48  www.nature.com/scientificreports/ In vitro MRI on a Bruker BioSpec 7 T instrument. Experiments on a Bruker BioSpec 7 T imager equipped with a 40-mm ID 13 C/ 1 H RF volume coil were done in the following way: first, the pressurized sample was placed in a three-layer μ-metal shield where it was heated using a heat gun. In some experiments, the samples were additionally preheated before placing them in the shield (see Supplementary Table S1). The hydrogenation was achieved by bubbling parahydrogen through the solution with the sample residing in the shield. After the hydrogenation, the sample was de-pressurized and transferred to the animal imaging coil with the subsequent transfer to the scanner. The sample transfer time from the μ-metal shield into the Earth's field took ca.