Radical‐free hyperpolarized MRI using endogenously occurring pyruvate analogues and UV‐induced nonpersistent radicals

It was recently demonstrated that nonpersistent radicals can be generated in frozen solutions of metabolites such as pyruvate by irradiation with UV light, enabling radical‐free dissolution dynamic nuclear polarization. Although pyruvate is endogenous, the presence of pyruvate may interfere with metabolic processes or the detection of pyruvate as a metabolic product, making it potentially unsuitable as a polarizing agent. Therefore, the aim of the current study was to characterize solutions containing endogenously occurring alternatives to pyruvate as UV‐induced nonpersistent radical precursors for in vivo hyperpolarized MRI. The metabolites alpha‐ketovalerate (αkV) and alpha‐ketobutyrate (αkB) are analogues of pyruvate and were chosen as potential radical precursors. Sample formulations containing αkV and αkB were studied with UV–visible spectroscopy, irradiated with UV light, and their nonpersistent radical yields were quantified with electron spin resonance and compared with pyruvate. The addition of 13C‐labeled substrates to the sample matrix altered the radical yield of the precursors. Using αkB increased the 13C‐labeled glucose liquid‐state polarization to 16.3% ± 1.3% compared with 13.3% ± 1.5% obtained with pyruvate, and 8.9% ± 2.1% with αkV. For [1–13C]butyric acid, polarization levels of 12.1% ± 1.1% for αkV, 12.9% ± 1.7% for αkB, 1.5% ± 0.2% for OX063 and 18.7% ± 0.7% for Finland trityl, were achieved. Hyperpolarized [1–13C]butyrate metabolism in the heart revealed label incorporation into [1–13C]acetylcarnitine, [1–13C]acetoacetate, [1–13C]butyrylcarnitine, [5‐13C]glutamate and [5‐13C]citrate. This study demonstrates the potential of αkV and αkB as endogenous polarizing agents for in vivo radical‐free hyperpolarized MRI. UV‐induced, nonpersistent radicals generated in endogenous metabolites enable high polarization without requiring radical filtration, thus simplifying the quality‐control tests in clinical applications.

It was recently demonstrated that nonpersistent radicals can be generated in frozen solutions of metabolites such as pyruvate by irradiation with UV light, enabling radical-free dissolution dynamic nuclear polarization. Although pyruvate is endogenous, the presence of pyruvate may interfere with metabolic processes or the detection of pyruvate as a metabolic product, making it potentially unsuitable as a polarizing agent. Therefore, the aim of the current study was to characterize solutions containing endogenously occurring alternatives to pyruvate as UV-induced nonpersistent radical precursors for in vivo hyperpolarized MRI. The metabolites alphaketovalerate (αkV) and alpha-ketobutyrate (αkB) are analogues of pyruvate and were chosen as potential radical precursors. Sample formulations containing αkV and αkB were studied with UV-visible spectroscopy, irradiated with UV light, and their nonpersistent radical yields were quantified with electron spin resonance and compared with pyruvate. The addition of 13 C-labeled substrates to the sample matrix altered the radical yield of the precursors. Using αkB increased the 13 C-labeled glucose liquid-state polarization to 16.3% ± 1.3% compared with 13.3% ± 1.5% obtained with pyruvate, and 8.9% ± 2.1% with αkV. For [1-13 C]butyric acid, polarization levels of 12.1% ± 1.1% for αkV, 12.9% ± 1.7% for αkB, 1.5% ± 0.2% for OX063 and 18.7% ± 0.7% for Finland trityl, were achieved. Hyperpolarized [1-13 C]butyrate metabolism in the heart revealed label incorporation into [1-13 C]acetylcarnitine, [1-13 C] acetoacetate, [1-13 C]butyrylcarnitine, [5-13 C]glutamate and [5-13 C]citrate. This study demonstrates the potential of αkV and αkB as endogenous polarizing agents for in vivo radical-free hyperpolarized MRI. UV-induced, nonpersistent radicals generated in endogenous metabolites enable high polarization without requiring radical filtration, thus simplifying the quality-control tests in clinical applications.

| INTRODUCTION
Hyperpolarization via dissolution dynamic nuclear polarization (DNP) can enhance the polarization of nuclear spins by several orders of magnitude. 1 Hyperpolarized 13 C-enriched probes have enabled real-time imaging of metabolic pathways in vivo. [2][3][4][5][6] More recently, a successful translation to human subjects was achieved, 7 demonstrating the potential of hyperpolarized MR for several clinical applications. [8][9][10][11][12] Hyperpolarization via DNP requires the presence of polarizing agents, in the form of free radicals, which will transfer their high spin order to the surrounding nuclei upon microwave irradiation at an appropriate frequency. Typically, these free radicals are persistent and added to the sample via chemical doping. Persistent radicals have proven to be highly efficient for dissolution DNP. 13,14 This poses a challenge for the clinical translation: free radicals may be toxic for living organisms and their presence shortens the brief measurement window of the imaging experiment. 3,15,16 Employing nonpersistent radicals generated via UV light irradiation of particular precursor molecules may address this challenge.
Nonpersistent radicals recombine into diamagnetic and biocompatible species at 190 K 17 and are thereby eliminated instantly during the dissolution process, resulting in radical-free hyperpolarized solutions. This obviates the need for filtration of the endogenous radical precursor.
Several candidates have been demonstrated as suitable precursors for UV-induced nonpersistent radicals, namely a mixture of butanol and phenol, 18 and several α-keto acids such as pyruvic acid (PA), [19][20][21][22] [2][3][4][5][6][7][8][9][10][11][12][13] C]PA, 23 phenylglyoxylic acid, 24 trimethylpyruvic acid 25 and α-ketoisocaproic acid. 26 It should be noted that trimethylpyruvic acid is not endogenous and showed great polarization capability only in its in-house synthesized deuterated form. 25 To date, a few in vivo studies have demonstrated the use of UV-induced nonpersistent radicals to measure in vivo metabolic processes. 19,21,24 While phenylglyoxylic acid was demonstrated to be beneficial as a polarizing agent for photosensitive metabolites, 24 the achievable polarization levels were relatively low, at least at 3.35 T and 1.25 K. The use of PA as a polarizing agent achieved relatively high polarization levels, [19][20][21][22] but its presence may interfere or even compete with metabolic processes that involve or are linked to pyruvate dehydrogenase (PDH) activity or the formation of acetyl-CoA. 6,[27][28][29] For example, while our group has demonstrated the copolarization of [1-13 C]butyrate, a short chain fatty acid, with [1-13 C]pyruvate, using UV-induced nonpersistent radicals, 21 a prior in vivo hyperpolarized 13 C MR study demonstrated that the metabolism of hyperpolarized [1-13 C]butyrate was altered in the presence of PA, 6 likely due to the acetyl-CoA produced by increased PDH flux.
For the above reasons, it is therefore of interest to find off-the-shelf endogenously occurring alternatives to PA for their use as nonpersistent radical precursors. Such alternatives may be advantageous for the measurement of hyperpolarized short-or medium-chain fatty acid metabolism. 6,[30][31][32][33][34][35][36] Additionally, such alternatives to PA may be beneficial for hyperpolarized MR studies, where the formation of pyruvate itself is a metabolic product of interest, 37-39 and can therefore not be coinjected as a radical precursor.
α-ketobutyrate (αkB) and α-ketovalerate (αkV) are two analogues of PA that naturally occur in human blood, 40,41 but which have different biochemical properties, with their longer side chains making them poorer substrates of the enzymes PDH and lactate dehydrogenase (LDH). [41][42][43][44] For example, αkB was approximately a 50% less efficient substrate of the enzyme PDH compared with PA, while αkV is neither a substrate nor inhibitor of PDH, 41 and unlikely to be a substrate for heart LDH. 43,44 Because it has been demonstrated that UV irradiation of α-keto acids creates radicals, 25 αkB and αkV may have high potential to be used as nonpersistent radical precursors.
The aim of this study was to further expand the field of UV-induced nonpersistent radicals by characterizing the endogenous pyruvate analogues αkV and αkB following UV irradiation and to quantify their potential as endogenous polarizing agents for dissolution DNP. Effects of matrix composition on radical yields and polarization levels of 13 C-labeled butyric acid (BA) and glucose (Glc) were quantified and a comparison was made with PA. In a proof-of-concept in vivo study it was investigated whether αkV and αkB could be used to measure cardiac metabolism of 13 C-labeled BA.

| Sample formulation and preparation
All chemicals were ordered from Sigma-Aldrich (Buchs, SG, Switzerland). Different sample formulations were used depending on the type of experiment.
Ultraviolet-visible (UV-Vis) spectroscopy experiments were performed at room temperature on 100 mM of αkV, αkB or PA in glycerol-water for characterizing UV light absorption of the radical precursors.
Electron spin resonance (ESR) was used to characterize the ESR line-shape and to quantify the concentration of the photo-induced radicals.
ESR experiments were performed with 5 M solutions of αkV, αkB or PA in glycerol: water (1:1, v/v). The ESR signal intensity was calibrated using six glycerol-water solutions with known TEMPOL concentrations between 0 and 100 mM ( Figure S1). In a second series of experiments, 2 M unlabeled Glc or 5.7 M BA was added to glycerol-water and the amount of radical precursor was empirically optimized to obtain 40 mM of nonpersistent radicals after 200 s of UV irradiation, to ensure comparability of the DNP experiments (see the supporting information for more details on the empirical optimization procedure). Setting the target radical concentration to 40 mM was a choice made based on previous experience with broad line-width radicals used to hyperpolarize 13 C-labeled nuclei at 7 T. 37,45 Hyperpolarized 13 C MRS was performed on samples containing fully deuterated, fully 13 C-labeled glucose ([U-13 C 6 , U-2 H 7 ]Glc) or [1-13 C] butyric acid ([1-13 C]BA). After optimization on samples prepared with nonlabeled compounds, the amount of radical precursor was set to 5.7 M αkV, 4.1 M αkB and 1.6 M PA for samples containing [U-13 C 6 , U-2 H 7 ]Glc. Conversely, the radical precursor was set to 2.4 M αkV and 4.0 M αkB for samples containing [1-13 C]BA. The latter were also used for in vivo measurements.
To compare liquid-state polarization levels obtained with persistent trityl radicals, [1-13 C]BA was hyperpolarized using OX063 by adjusting a previously published recipe 31,34 by increasing the radical concentration to 25 mM without adding a Gd-based contrast agent. Additionally, based on the poor performance of OX063 and to compare with a matrix formulation used for the UV-induced nonpersistent radical measurements, Finland trityl acid was added to a final concentration of 25 mM to the [1-13 C]BA + αkB sample formulation (without performing UV irradiation).

| Creating nonpersistent radicals on metabolites using UV irradiation
To create nonpersistent radicals, the sample formulations described in the previous section were sonicated and degassed at 50 C for 20 min prior to pipetting 6-μl droplets and freezing them in liquid nitrogen to create a solid pellet. The pellets were transferred to a quartz Dewar (Magnettech, Freiberg Instruments, Germany) filled with liquid nitrogen and irradiated with a broadband UV lamp (Dymax BlueWave 200, Torrington, CT, USA) at maximum power (40 Wcm À2 ) for a maximum of 200 s using a home-built irradiation setup. 22 In experiments aiming to investigate the time course of radical generation for αkV, αkB or PA (n = 3, each precursor), frozen beads were irradiated for a set duration (i.e. 20, 45, 70, 110, 150 and 200 s), and ESR was measured at the end of each step.

| Quantification and characterization of nonpersistent radicals
UV-Vis spectroscopy was performed to measure the light absorbance of the different radical precursors using a single beam UV-3100PC spectrophotometer (VWR International) and a 1-mm pathlength quartz cuvette. UV light absorption of 100 mM αkV, αkB or PA in glycerol-water samples was measured from 280 to 600 nm in steps of 0.5. Measurements were performed at room temperature because there was no significant difference between the absorbance spectra of PA solutions acquired at room and liquid-nitrogen temperatures. 22 ESR was used to determine the nonpersistent radical concentration generated in the samples after UV irradiation at liquid-nitrogen temperature. X-band ESR was performed at 77 K as well using a MiniScope MS 400 spectrometer (Magnettech GmbH, Germany). Spectrometer parameters were chosen to ensure that saturation of the ESR signal was avoided over the entire range of radical concentrations and kept constant throughout all experiments. Parameters were set to: 20 s sweep time, 20 mT magnetic field range, 0.2 mT magnetic field modulation amplitude and 30 dB power attenuation. The ESR experiments were performed on two 6-μl beads for each sample formulation. Subsequently, the beads were extracted from the quartz Dewar and transferred to a preweighed microcentrifuge tube, which was then weighed to determine their exact volume and correct the concentration calibration.

| Hyperpolarization via DNP
All DNP experiments were performed in a 7-T custom-built polarizer. Nuclear spins were hyperpolarized using a millimeter-wave source centered at 197 GHz, with digital control for frequency modulation (FM), a 55-mW output power and a 1-GHz tuning range (Elva-1 VCOM-06/197/1.0/55-DD). Experiments were performed to determine the optimal hyperpolarization conditions in terms of microwave irradiation frequency and to quantify polarization build-up times. The microwave frequency sweeps were performed with and without microwave FM to find the optimal DNP conditions and to quantify the effect of microwave FM on the enhancement. The sample cup was filled with 12 frozen UV-irradiated beads containing [U-13 C, U-2 H]Glc. Microwave frequency-modulated profiles were acquired at 4.2 K using a constant 40 MHz modulation amplitude and 5 kHz modulation rate for each microwave frequency step. The step size was 40 MHz for beads containing αkV or αkB, and 20 MHz for beads containing PA. For each frequency step, the sample was irradiated for 40 min and the polarization build-up was monitored using hard 2 RF excitation pulses every 5 min. FM was not used, either for Finland trityl or for Ox063 samples.
The optimal conditions found in the previous experiments for a given set of FM parameters were used to hyperpolarize [U-13 C, U-2 H]Glc as well as [1-13 C]BA for quantification of 13 C polarization levels in liquid state after dissolution and for the in vivo experiments. The 13 C-labeled metabolites were hyperpolarized at 1.05 ± 0.02 K for 2.5 h, with frequency-modulated (40 MHz amplitude at a rate of 5 kHz) microwave irradiation at 196.69 GHz center frequency for αkV and αkB, and at 196.65 GHz center frequency for PA. Their polarization build-up was monitored using 2 RF excitation pulses every 5 min.

| Dissolution DNP, hyperpolarized 13 C MRS in phantoms and in vivo
Frozen hyperpolarized beads were dissolved using either 5.5 ml of D 2 O or a phosphate buffered saline solution for liquid-state in vitro and in vivo experiments, respectively. 45,46 The dissolved sample was automatically transferred to a separator/infusion pump located in a 9.4-T horizontal bore magnet (Agilent, Palo Alto, CA, USA). 47 Hyperpolarized 13 C MR spectra were recorded within the pump using a dual 1 H/ 13 C volume coil starting 3 s after dissolution and using a 5 RF excitation pulse with 3 s repetition time (TR). After a complete decay of the hyperpolarized magnetization, a thermal equilibrium 13 C spectrum of the sample was acquired using a 90 RF excitation pulse, with a TR of 60 s and 64 averages.
The enhancement ϵ was calculated as the ratio of the hyperpolarized and thermal signal peak integral, taking into account a correction for the RF excitation angle and the number of averages. The remaining 13 C hyperpolarization after dissolution and transfer was calculated as , where the hyperbolic tangent represents the 13 C thermal equilibrium polarization at 293 K and 9.4 T.
In vivo experiments were performed on two male Wistar rats (one injection each) to demonstrate the feasibility of the novel polarizing agents to measure the cardiac metabolism of hyperpolarized [1-13 C]BA, and to obtain preliminary information on in vivo chemical shifts. The feasibility study was conducted according to federal ethical guidelines and was approved by the local regulatory body. Anesthesia protocols and physiological monitoring were described previously. 6 points were acquired to sample a bandwidth range of 20.5 kHz. The first spectra acquired following injection, in which metabolic products were absent, were used to confirm the resonances identified in phantom experiments ( Figure S2).

| Data processing and analysis
The ESR concentration calibration curve was obtained from fitting the second integrals of the ESR signal intensity linearly as a function of the radical concentration ( Figure S1). To calculate the build-up rate of nonpersistent radical formation, the second integral of the ESR signal intensity was corrected for bead volume variation prior to performing an intensity calibration to quantify absolute concentrations.
Following methods previously described, 22 the radical generation time course was fit to a monoexponential function to extract the radical generation rate constant and plateau value.
Regarding the DNP sweeps, the y-axis value at each frequency point corresponded to the polarization plateau measured at 4.2 K. The nonmodulated sweeps were normalized to 1 and the relative increase in DNP enhancement due to microwave modulation was then calculated accordingly. The microwave frequencies at which the highest polarization levels are achieved are referred to as the negative DNP maximum and the positive DNP maximum.
For each sample, the polarization time course at the best DNP condition was fit to a monoexponential function to extract the build-up time constant and the polarization plateau, as is customary. Accordingly, each sample was irradiated for at least three time constants prior to dissolution.
Statistically significant differences between polarization levels obtained using αkB or αkV compared with PA were tested via an unpaired, two-tail t-test assuming equal variance, with p less than 0.05 considered significant.
In vivo spectra were postprocessed in VnmrJ 3.2 (Agilent, Palo Alto, CA, USA) using 20 Hz line-broadening, baseline correction and drift correction. The signal-to-noise ratio (SNR) was calculated as the ratio of the highest signal intensity after phasing and the standard deviation of the noise over a region without metabolic or injected substrate resonances. All in vivo spectra where the [1-13 C]acetylcarnitine resonance was visible were summed, corresponding to 10 consecutive spectra in the case of αkV, and 12 spectra in the case of αkB. Chemical shifts were assigned using [1-13 C]acetylcarnitine as the reference peak resonating at 173.9 ppm and assigning all other metabolites as indicated before. 6

| RESULTS
To characterize structural changes and absorption characteristics of UV-light irradiation on the metabolites αkV, αkB and PA, UV-Vis spectroscopy and ESR measurements were performed: UV-Vis absorption spectra showed an $ 1.7-fold higher absorbance for αkV compared with αkB or PA ( Figure 1A). αkB and PA showed nearly identical absorbance maxima in the UV range of 300-400 nm ( Figure 1A). Absorbance of all three metabolites peaked at around a wavelength of 320 nm.
ESR performed on frozen samples prior to UV irradiation indicated the initial absence of unpaired electron spins in the matrixes of glycerolwater mixed with 5 M of αkV, αkB or PA. ESR spectra acquired after 200 s of UV irradiation demonstrated that free radicals were generated within the frozen samples ( Figure 1B-D). ESR spectra of αkV and αkB showed a nearly identical shape ( Figure 1B,C), but were distinct from the PA spectrum ( Figure 1D). The production of nonpersistent radicals as a function of UV-irradiation time followed a near monoexponential build-up ( Figure 1E) with a characteristic time constant of 30.9 ± 5.1 s for αkV, 37.0 ± 5.2 s for αkB and 46.5 ± 1.4 s for PA ( Table 1) Abbreviations: αkB, alpha-ketobutyric acid; αkV, alpha-ketovaleric acid; PA, pyruvic acid. nonpersistent radical concentrations were 41.6 ± 0.6 mM for αkV, 56.1 ± 2.7 mM for αkB and 55.0 ± 1.9 mM for PA ( Figure 1E). Irradiating the samples for 200 s resulted in plateauing the radical concentration while avoiding pulverization of the beads due to excessive UV irradiation.
To obtain transparent glassy beads that remained intact upon irradiation with UV light, 2 M Glc was dissolved with glycerol-water and admixed with 5.7 M αkV, 4.1 M αkB or 1.6 M PA. These formulations yielded a nonpersistent radical concentration of 41.5 ± 2.5 mM in αkV, 39.5 ± 2.3 mM in αkB and 41.7 ± 2.0 mM in PA after 200 s of UV irradiation (Figure 2A).
To assess the effect of FM of the microwave irradiation, the 13 C nuclear polarization was measured as a function of the microwave frequency, which for the αkV and αkB sweeps showed a broadening when FM was applied. FM improved the DNP performance in terms of signal enhancement of hyperpolarized 13 C-labeled Glc by 100%, 50% and 30% for αkV, αkB and PA, respectively ( Figure 3, Table 2 Table 2).
To determine the build-up rates of nuclear magnetization, the latter was measured as a function of microwave irradiation duration. The solid state build-up times of hyperpolarized [U-13 C, U-2 H]Glc were 1.7 ± 0.5, 1.3 ± 0.5 and 1.6 ± 0.5 ks for αkV, αkB and PA, respectively (Table 3).
Following dissolution, the liquid-state polarization for the C 2-5 group of [U-13 C, U-2 H]Glc was 8.9% ± 2.1%, 16.3% ± 1.3% and 13.1% ± 1.5% for αkV, αkB and PA, respectively (Table 3, Figure 4). While [U-13 C, U-2 H]Glc hyperpolarized with αkB showed significantly higher liquid-state polarization than that hyperpolarized using PA (p = 0.048), there was no significant difference between PA and αkV (p = 0.09). The quantification of polarization levels on the C 1 and C 6 resonances of Glc did not alter the results (Table 3).   Figure 5). Note that pH was not neutralized in these experiments, and the chemical shifts were different in the pH-neutralized in vivo experiments (Table 5).
T A B L E 2 Microwave center frequencies at which dynamic nuclear polarization (DNP) maxima and minima occur at 7 T. Microwave frequency sweeps ( Figure 3) were conducted on [U-13 C, U-2 H]Glc in glycerol-water samples. The microwave frequency was swept using either monochromatic microwave irradiation or frequency modulation with 40 MHz modulation amplitude at a 5 kHz rate    T A B L E 4 Solid state build-up times at 7 T, 1.05 K are reported for [1-13 C]butyrate samples containing alpha-ketovaleric acid (αkV) or alphaketobutyric acid (αkB) as nonpersistent radical precursors, and OX063 and Finland trityl as persistent radicals. Room-temperature liquid-state enhancements and liquid-state polarization were calculated after sample dissolution and transfer to a 9.4-T MR scanner. Mean and average values are obtained from n datasets. Exemplary spectra for the nonpersistent radical samples are shown in Figure 5 Radical F I G U R E 5 Liquid-state 13 C MR spectra of [1-13 C]butyrate hyperpolarized using alpha-ketovaleric acid (αkV) (top) and alpha-ketobutyric acid (αkB) (bottom) displaying the 13 C-labeled substrate and precursor resonances. Spectra were acquired 3 s postdissolution and were line-broadened with 2 Hz. Solutions were acidic because pH neutralization was not performed in these experiments T A B L E 5 Chemical shifts of observed metabolites in the heart after injection of pH-neutralized hyperpolarized [1-13 C]butyrate in vivo using either alpha-ketovaleric acid (αkV) or alpha-ketobutyric acid (αkB) as a polarizing agent (in vivo spectra are displayed in Figure 6). [1-13 C] acetylcarnitine was used as the reference peak (*) and assigned to 173.9 ppm. The resonance of natural abundance [1-13 C]αkV-hydrate was not detected in these experiments (see also Figure S2) in vivo, further evidenced by its absence immediately after injection ( Figure S2). Maximum in vivo SNR on 13 C BA was observed 12 s after dissolution, with an SNR of 1370 for αkV and 1780 for αkB. Cardiac metabolism of hyperpolarized [1-13 C]BA ( Figure 6, Table 5) resulted in 13

| DISCUSSION
This study shows that αkV and αkB can be used to generate nonpersistent radicals for hyperpolarizing 13  of each compound to UV light such that a specific sample formulation was required to achieve the targeted nonpersistent radical concentration of 40 mM. As has been observed before, 22 the relationship between nonpersistent radical yield and the concentration of the precursor was nonlinear. This posed a challenge during sample formulation optimization. Even although αkV, αkB and PA differ from each other by additional methylene units on their aliphatic side chain, the results from our experiments illustrate that sample formulation requires a careful optimization in terms of UV-induced radical yield and polarization level for each 13 C-labeled metabolic substrate, which is an empirical and nontrivial process.
Because the current work focused on the feasibility of αkV and αkB as nonpersistent radical precursors, optimization of the sample formulation did not go beyond obtaining the targeted concentration of 40 mM nonpersistent radicals, resulting in relatively large concentrations of precursor compared with PA in the 13 C-labeled sample formulations. It is important to point out that the samples were not specifically optimized to maximize radical yield, to minimize precursor concentrations or to maximize achievable polarization levels, which is a limitation of our study. The empirical approach of optimizing sample formulations, plus the observation that radical quantum yields were comparable in glycerol-water mixtures (Figure 1), show that there is significant room for improvement. Our experience with UV-irradiated nonpersistent radical precursors and published work 21,22 indicate that precursor volumes can be reduced when boosting radical yield or UV-light penetration, which may be achieved by using different glassing agents such as ethanol, by deuteration of the radical precursors, by changing the relative ratios of precursor to glassing agent, by adapting the UV-irradiation time, or by changing the bead size.
The nuclear polarization levels increased when applying microwave FM, which is a consequence of the relatively broad radical ESR spectrum and short electron T 1 characterizing the UV-radical family ($ 100 ms). [50][51][52] By contrast, microwave FM has little effect on narrow linewidth radicals and radicals with long electron T 1 such as OX063 at 6.7 T. 53 Similarly, for polarizer systems operating at low field ($3.35 T), microwave FM may be less effective because of longer radical T1e and more narrow spectral width (less g-anisotropy). Microwave FM is a feature that is available on commercially available polarizers working at high field (e.g. the Spin Aligner polarizer; Polarize, Copenhagen, Denmark). Polarization levels in the current study were of the order of 9% to 16% for Glc and 13% for BA, close to the polarization of 18.7% measured when adding Finland trityl to a non-UV-irradiated αkB BA mixture. The recipe for OX063 did not perform as expected on our system based on the low polarization ($2%) that was obtained. This can be a consequence of the apolar nature of the BA sample formulation. Indeed, Ox063 is very soluble in polar solvents such as water (log P of À0.5 and À1.38, respectively). Differently, BA has a log P of 0.79. Most likely, OX063 in the presence of BA made a suspension rather than a solution. The opposite reasoning applies to Finland trityl. In previous work using similar experimental conditions, UVirradiated metabolite mixtures containing 13 C-labeled PA and BA were hyperpolarized to significantly lower levels (i.e. 3.3% ± 0.5%À5.2% ± 0.5% for 13 C BA). 21 The latter was a consequence of a different UV-light source, yielding a third of the radical concentration measured in this study, and the absence of microwave FM. 21 In the current study, a broadband UV-light source was used, delivering 40 times more power, and glycerolwater was added to the mixtures to improve the DNP sample matrix. Although performed under different experimental conditions and using different methods to calculate the liquid-state polarization, previous studies reported polarization levels of between 22.2% ± 2.1% 54 and 30.1% ± 1.8% 25 for Glc and of between 7% ± 2% 31 and 28% ± 4% 34 for BA. The current results are within a similar range and thus are promising. Although a dedicated optimization in terms of polarization level was not performed in the current study, polarization levels may be further increased by optimization of the sample formulation (i.e. components, concentrations and bead size), or by a detailed investigation of the microwave FM. However, previously reported improvements in DNP performance may conflict with the UV performance and nonpersistent radical generation: for example, using dimethyl sulfoxide as the glassing agent increased the DNP performance by 18-fold when polarizing 13 C-labeled BA. 34 However, UV-Vis spectroscopy and ESR measurements have shown the photosensitive nature of dimethyl sulfoxide and its unsuitability as a glassing agent for UV-induced nonpersistent radical formulations using a broadband UV source. 24 Hyperpolarized BA has been used before as a probe to study short-chain fatty acid cardiac metabolism, 6,21,31,34,35 with different results in terms of the amount of observed metabolites. In the current study, hyperpolarization of BA with αkV and αkB reached sufficient SNR to observe cardiac metabolism, with a similar metabolic profile observed in previous studies using persistent radicals at high magnetic field. 6,31 High magnetic field allows for a better spectral resolution and thus may facilitate the resolved detection of metabolites such as 13 C-labeled citrate and glutamate.
However, B 0 inhomogeneities also increase, which may complicate shimming procedures, especially around the moving heart, which may result in distorted line shapes and difficult-to-phase spectra. In the current study, the citrate resonance could not be reliably detected due to the shoulder of the neighboring [1-13 C]αkB-hydrate resonance. Conversely, the absence of [1-13 C]αkV-hydrate ( Figure 6) enabled citrate detection when using αkV. Note that the relative amount of αkV in the BA sample formulation was much smaller compared with the amount of αkB, which contributed to increased signal intensities of natural abundance αkB resonances compared with those of αkV. This can be appreciated when observing the relative signal ratio of [1-13 C]αkV or [1-13 C]αkB with [1-13 C]acetylcarnitine ( Figure 6). Furthermore, the detection of [1-13 C]αkV-hydrate in the non-pH-neutralized liquid-state experiments compared with its absence in the pH-neutralized in vivo experiment, as well as the increased signal intensity of the [1-13 C]αkB-hydrate resonance in vivo compared with the non-pH-neutralized liquid-state experiment, demonstrate the sensitivity of the proposed polarizing agents to pH and illustrate the importance of pH optimization.
The use of endogenous αkV and αkB is a potential alternative to PA as nonpersistent radical precursors. Based on the location of their natural abundance 13 C resonances, the potential detection of pyruvate as metabolic product would not be disturbed. It is, however, a limitation of the current study that interferences of αkV and αkB with in vivo metabolic processes were not investigated. Reportedly, PDH, as well as LDH, have reduced or even absent affinities for αkB and αkV, respectively. [41][42][43][44] This was partially confirmed in recent in vivo experiments using hyperpolarized 13 C-labeled αkB that showed a decreased 13 C-labeling of bicarbonate compared with PA. 33 Also, αkB is converted by PDH to propionyl-CoA, avoiding any potential competing perturbation of the acetyl-CoA pool, as seen with PA. 6 In the former study the in vitro LDH response to 13 C-labeled αkB varied depending on the LDH isoform. 33 While αkV is not anticipated to be appreciably converted by PDH and LDH, 41,43,44 determining their in vivo response to hyperpolarized 13 C-labeled αkV, and elucidating any potential metabolic interference by it or αkB, including dose dependencies, remains to be established. Based on the aforementioned evidence of αkV not being a significant substrate of either LDH or PDH, and the lower αkV hydrate signal observed in our study, we hypothesize that αkV may be the preferred choice over αkB of nonpersistent radical precursor.
Nevertheless, radical-free dissolution DNP via the use of endogenous nonpersistent radicals provides a benefit at low cost and may increase the duration of the hyperpolarized state by avoiding one filtration step in clinical applications. In addition, UV-induced radicals of PA quench with increasing temperature 18 and recombine at a threshold of 190 K, 17 providing an additional benefit of UV-induced radicals in αkV and αkB to produce transportable hyperpolarized 13 C-labeled substrates.

| CONCLUSION
We conclude that the endogenous α-keto acids αkV and αkB can be used as efficient endogenous nonpersistent radicals following irradiation with UV light, achieving similar or higher 13 C polarization compared with PA and thus are potential candidates for translational clinical hyperpolarized MRI, enabling high polarization without requiring radical filtration. Cardiac metabolism of 13 C-labeled butyrate hyperpolarized with αkV or αkB demonstrated label propagation in a wide range of metabolites, demonstrating their potential as endogenous polarizing agents for in vivo radicalfree hyperpolarized MRI.