Imaging glutathione depletion in the rat brain using ascorbate-derived hyperpolarized MR and PET probes

Oxidative stress is a critical feature of several common neurologic disorders. The brain is well adapted to neutralize oxidative injury by maintaining a high steady-state concentration of small-molecule intracellular antioxidants including glutathione in astrocytes and ascorbic acid in neurons. Ascorbate-derived imaging probes for hyperpolarized 13C magnetic resonance spectroscopy and positron emission tomography have been used to study redox changes (antioxidant depletion and reactive oxygen species accumulation) in vivo. In this study, we applied these imaging probes to the normal rat brain and a rat model of glutathione depletion. We first studied hyperpolarized [1-13C]dehydroascorbate in the normal rat brain, demonstrating its robust conversion to [1-13C]vitamin C, consistent with rapid transport of the oxidized form across the blood-brain barrier. We next showed that the kinetic rate of this conversion decreased by nearly 50% after glutathione depletion by diethyl maleate treatment. Finally, we showed that dehydroascorbate labeled for positron emission tomography, namely [1-11C]dehydroascorbate, showed no change in brain signal accumulation after diethyl maleate treatment. These results suggest that hyperpolarized [1-13C]dehydroascorbate may be used to non-invasively detect oxidative stress in common disorders of the brain.

conversion is one of the motivations for studying ascorbate-derived imaging agents, whose interconversion depends on both the concentration of GSH and the presence or absence of reactive oxygen species 14,15 . In particular, [1-13 C]DHA, a 13 C enriched version of dehydroascorbic acid, has been studied with hyperpolarized (HP) 13 C magnetic resonance (MR) in preclinical models of cancer, diabetes, acute kidney injury, and fatty liver disease [15][16][17][18][19][20] . In addition, ascorbates labeled for positron emission tomography (PET), namely [1-11 C]DHA and [1-11 C]VitC, have recently been synthesized and studied in ROS-producing immune cells in vitro 21 . Together, these studies suggest that ascorbate-derived imaging methods may be able to image oxidative stress in vivo. In particular, DHA-derived imaging agents were designed to image intracellular redox capacity, typically defined as the concentration of intracellular reducing equivalents and the ability to regenerate them. We therefore compared these methodologies in an animal model of pharmacologic glutathione depletion, to determine if ascorbate-derived HP MR or PET probes could detect the modulation of redox capacity.
Although [1-13 C]DHA and [1-11 C]DHA are chemically and biochemically equivalent, the techniques used to study them in vivo are considerably different. HP 13 C MR via dissolution dynamic nuclear polarization (dDNP) is an emerging imaging technology that allows direct visualization of metabolic conversion of 13 C-enriched molecules, and has been applied extensively to study metabolism in cells, tissues, animal models, and patients [22][23][24] . dDNP transfers polarization from electron spins to 13 C nuclei spins at extreme low temperature (0.8-1.3 °K) through microwave irradiation, providing more than 10,000-fold MR signal enhancement of 13 C-labelled substrates 25 . The HP substrate can be subsequently injected into a living system and its conversion to other metabolites can be monitored by MR spectroscopy in a rapid, real-time manner. HP MR with [1-13 C]DHA takes advantage of the chemical shift between [1-13 C]DHA and [1-13 C]VitC (3.8 ppm) and the drastic signal enhancement (~10% polarization, >10 4 -fold at 3 Tesla) provided by dDNP to enable in vivo interrogation of redox capacity. Since the hyperpolarized signal lifetime is prescribed by the T 1 of the substrate (approximately 57 s for [1-13 C] DHA at 3 Tesla), HP 13 C signal can be observed in vivo for approximately 1 to 2 minutes at most 23 . Therefore, the real-time detection of [1-13 C]DHA to [1-13 C]VitC in vivo is primarily a kinetic measurement that depends on substrate transport and the availability of intracellular reducing equivalents.
In contrast, PET is an established metabolic imaging technology whereby a positron-emitting isotope, such as 11 C or 18 F, is incorporated into a drug or metabolite of interest, and usually imaged following a significant delay depending on the half-life of the radioactive isotope. Therefore, the primary determinants of image contrast are the transport and retention/entrapment of the radiopharmaceutical, and the clearance of nonspecific signal. As shown in Fig. 1b, PET images of rat brains demonstrated the distinctive difference in accumulated signal between [1-11 C]DHA and [1-11 C]VitC, as previously reported, reflecting their differential transport across the blood-brain barrier and cellular membrane 21 . In the case of [1-11 C]DHA, the radiopharmaceutical is expected to be entrapped inside the cell once it is reduced to [1-11 C]VitC, a charged molecule at physiological pH, and unreacted [1-11 C] DHA may be washed out. We hypothesized that retention of [1-11 C]DHA would depend partially on intracellular reducing agents including GSH 26 , and therefore may also be modulated following glutathione-depletion.
In this study, we aimed to characterize ascorbate-derived HP 13 C MR and 11 C PET probes for brain imaging, and take advantage of their different capabilities to assess in vivo redox capacity of the rat brain. First, we studied the transport, compartmentalization, and biodistribution of both DHA and VitC in normal rat brain with HP 13 C Figure 1. Ascorbates as redox imaging probes. (a) Dehydroascorbate (DHA) can traverse the blood-brainbarrier, enter cells via glucose transporter (GLUT)1,3,4 and be rapidly reduced to ascorbic acid (vitamin C, VitC) by intracellular antioxidants including glutathione (GSH), which is coupled with the pentose phosphate pathway via NADPH. The conversion from hyperpolarized [1-13 C]DHA to [1-13 C]VitC can be monitored by MR spectroscopy, taking advantage of their unique chemical shifts. The C1 position of DHA and VitC can also be labelled with 11 C for PET and biodistribution studies. (b) In these PET images (adapted from Carroll et al. 21 ), the accumulation of [1-11 C]DHA in the normal rat brain is compared to that of [1-11 C]VitC in the normal rat brain. In normal rats, higher brain signal for [1-11 C]DHA is expected since it readily crosses the blood-brain barrier and is transported into cells via GLUT transporters. In contrast, transport of VitC into the brain via sodium-dependent vitamin C transporter-2 (SVCT2) is a slower process. Unlike measuring the real-time conversion rate by hyperpolarized MR, PET and biodistribution with [ 11 C]ascorbates measures the uptake and retention of 11  MR and 11 C PET probes. Then we modulated GSH content in rats by pharmacological treatment and studied the changes in DHA to VitC conversion in the brain using both HP 13 C MR and 11 C PET probes. Specifically, four groups of Sprague-Dawley rats were studied using: (1) two-dimensional (2D) chemical shift imaging (CSI) with HP [1-13 C]DHA and [1-13 C]VitC separately to determine the transport and compartmentalization of ascorbates in the head region (n = 6 for DHA, n = 3 for VitC); (2) one-dimensional (1D) slice-selective dynamic spectroscopy with HP [1-13 C]DHA, at 24 hours before and 2 hours after diethyl maleate (DEM) intraperitoneal injection (4.6 mmol/kg) 27 , to measure the real-time DHA to VitC conversion rate (n = 4); (3) biodistribution using [1-11 C]DHA and [1-11 C]VitC, respectively, to determine the accumulation of ascorbate-derived probes in the brain and other organs (n = 3 each); (4) biodistribution with [1-11 C]DHA in rats with DEM treatment as in group (2) to measure 11 C probe accumulation following glutathione depletion (n = 3). Additionally, 1D slice-selective dynamic spectroscopy with HP [1-13 C]DHA at 60% concentration of group (2) was performed to demonstrate DHA to VitC reaction order (n = 3), and an in vitro GSH assay was performed on brain tissues collected from normal and DEM-treated rats to validate the redox modulation by DEM (n = 7 each).

Results
Behavior of HP 13 C DHA and VitC in vivo is consistent with known features of blood-brain barrier penetration and transport. Both HP [1-13 C]DHA and HP [1-13 C]VitC were administered separately to normal rats (group 1). Following HP [1-13 C]DHA injection, its reduction to [1-13 C]VitC was observed in the brain, demonstrated by the presence of a new downfield resonance (approximate 3.8ppm) in the 13 C MR spectrum (n = 6). For HP [1-13 C]DHA study, the DHA resonance was observed in most tissue voxels including brain and its surrounding tissues (muscle and vasculature), while the VitC metabolite was only observed in the brain voxels ( Fig. 2a) with VitC/(DHA + VitC) = 0.48 ± 0.15 (estimated from 18 brain voxels), consistent with penetration of DHA across the blood-brain barrier. In contrast, when HP VitC is injected, the 2D CSI (n = 3) found drastically lower VitC signal in the brain (mostly below quantification limit) than in surrounding tissues and no oxidation to DHA (Fig. 2b). These data are consistent with the slower transport of VitC across the blood-brain barrier, and the high reducing capacity of the normal brain 8 .

Glutathione depletion via diethyl maleate results in a lower rate of DHA-to-VitC conversion.
Using a 1D dynamic HP [1-13 C]DHA study, we investigated whether the DHA to VitC conversion rate is correlated with GSH content in the rat brain (group 2). We adopted a variable flip angle scheme to account for the non-renewable nature of hyperpolarized signal and to utilize signal-to-noise ratio (SNR) efficiently since the T 1 of [1-13 C]VitC is relatively short (29.1 s, in vitro at 3 T) 16 . Dynamic spectra of the rat brain showed that the [1-13 C] ]DHA, with conversion to VitC in the normal rat brain. The VitC resonance is only observed in voxels corresponding to the brain as shown in the VitC signal intensity map and slab spectra, while the DHA resonance is observed both in the brain and surrounding tissue. A representative voxel corresponding to brain tissue shows both a resonance corresponding to the introduced hyperpolarized [1-13 C]DHA and its metabolite [1-13 C]VitC. (b) Injection of hyperpolarized [1-13 C]VitC in normal rats. Lower (in this case negligible) VitC signal is seen in voxels corresponding to brain as shown in the VitC signal intensity map and slab spectra, and no oxidation to DHA is observed. A representative voxel corresponding to tissues in the neck show a resonance corresponding to the introduced hyperpolarized [1-13 C]VitC without evidence of metabolic conversion. VitC signal gradually increases during the acquisition window used (Fig. 3), suggesting VitC production rate is greater than signal decay rate mediated by T 1 relaxation. Because VitC signal is mainly from in the brain while DHA signal is from both the brain and surrounding tissues in the 1D slab spectra, we focused our analysis on the kinetic rates of VitC production as opposed to including both DHA and VitC in the kinetic model or using ratiometric approach.
To validate our kinetic modeling of the DHA to VitC conversion, dynamic 1D spectroscopy was performed on a separate cohort of rats (n = 3) with 18 μmoles HP DHA (approximately 60% dose of the control/baseline group). We found there was no statistically significant difference between the control/baseline group and lower dose group in k app (4.94 ± 1.04 vs. 4.38 ± 0.70 s −1 , p = 0.678), or k (11.22 ± 1.74 vs. 0.94 ± 1.86 s −1 , p = 0.917) as well as the sum of VitC signal (510.3 ± 65.06 vs. 507.6 ± 73.94 A.U., p = 0.9792) (Fig. 4d-g), suggesting that DHA dose used in our experimental set-up was in excess and not a rate limiting factor for VitC production.

Discussion
Non-invasive in vivo evaluation of brain redox capacity and oxidative stress could provide insights into neurological disease progression and therapeutic response. In this study, we used ascorbate-based imaging probes to study brain redox capacity leveraging both HP-MR and PET modalities. We first demonstrated that HP [1-13 C]DHA is readily converted to HP [1-13 C]VitC in the normal brain, evidenced by 2D chemical shift imaging revealing both DHA and VitC resonances in voxels corresponding to brain tissues following [1-13 C]DHA injection. This finding is consistent with the penetration of DHA through BBB and its rapid entry into cells. In contrast, when HP [1-13 C] VitC was administered, lower hyperpolarized signal was seen in brain voxels and no oxidation to HP [1-13 C]DHA was observed. These findings reflect the slow cellular transport process of VitC via SVCT2 against a concentration Both apparent VitC production rate (k app ) and kinetic rate (k) of VitC production in the brain decreased significantly after DEM treatment (a,b), consistent with significantly lower brain GSH content in DEM-treated rats compared to control rats (c). To demonstrate DHA to VitC reaction order, 1D spectroscopy was performed on a separate cohort of rats with HP [1-13 C]DHA at 60% of the usual dose. The average signal of [1-13 C]DHA and [1-13 C]VitC at each time point were plotted (d), and VitC signal appears unchanged with decreased DHA dose. The resulting k app , k, and total 13 C VitC SNR are not significantly different from the control group (e,f,g). (*p < 0.05, ****p < 0.0001).

Figure 5. Biodistribution studies reveal different accumulation of [1-11 C]dehydroascorbate (DHA) and [1-11 C]
vitamin C (VitC) in normal rat brain, but no effect of diethyl maleate (DEM) treatment. Gamma counting of harvested brains showed a significantly higher signal accumulation for [1-11 C]DHA than [1-11 C]VitC in the normal brain 1 hour following radiopharmaceutical administration. In addition, there was significantly higher uptake/retention for [1-11 C]VitC versus [1-11 C]DHA in the lungs and liver in normal rats. In the DEM-treated rats with [1-11 C]DHA, no significant differences were found in 11 C radiopharmaceutical uptake/retention between normal and DEM-treated rats in major organs, including the brain. (*p < 0.05, ***p < 0.001).
SCientiFiC REPoRTS | (2018) 8:7928 | DOI:10.1038/s41598-018-26296-6 gradient, and the role of VitC as an antioxidant reservoir that is not easily oxidized. These features represent the major shortcomings of [1-13 C]VitC as a HP MR probe for imaging the brain.
After characterizing the transport and compartmentalization of DHA and VitC, we modulated GSH content with DEM treatment which depletes GSH by direct conjugation 27 , and studied the changes in DHA to VitC kinetic rates using dynamic HP 13 C MR spectroscopy. 2D CSI with radiometric quantification has lower reproducibility due to sensitivity to small differences in acquisition timing and substrate delivery, demonstrated by the variabilities in VitC/DHA ratios in 6 normal rats (reported in Supporting Information). Therefore, we adopted 1D slab dynamic acquisition approach with variable flip angle scheme for efficient SNR usage, which is important given the short T 1 of [1-13 C]VitC, to study the effect of redox modulation. In dynamic spectra of normal brain, there was observable real-time production of VitC; we also observed that DHA signal is about 5 times higher than VitC signal, suggesting that cellular transport and/or enzyme and cofactor turnover rate are likely rate-limiting factors for the observed VitC production. In other words, these data suggested that the DHA dose was in excess and not a rate-limiting factor in our experimental set-up. To demonstrate the reaction order, we performed dynamic spectroscopy with lower (60%) doses of [1-13 C]DHA and found that the kinetic rates of VitC production and sum of VitC SNR are independent of DHA concentration, which validated the pseudo-zero-order kinetic model. To quantify in vivo reaction kinetics, we generated two kinetic parameters (reported in Methods section): apparent VitC production rate (k app ) is a straightforward parameter that reflects the rate of product accumulation, assuming T 1 relaxation equally affects the control and DEM-treated groups; whereas kinetic rate (k) was generated by non-linear fitting and takes account of T 1 relaxation. We found that DHA to VitC conversion rate correlated with the brain GSH content, as evidenced by significantly reduced kinetic parameters (k app and k) after DEM treatment that depletes GSH by more than 50%.
There are several important considerations and limitations of our dynamic [1-13 C]DHA MR study. First, each animal was subjected to two intravenous injections of [1-13 C]DHA, an oxidant that could deplete GSH, 24 hours apart between baseline and post-DEM-treatment imaging 12 . However, Weber et al. have demonstrated that rat brain GSH content returns to normal 24 hours after DEM treatment, a prototypical GSH deleting agent that directly conjugates with GSH and is more potent than DHA 27 . Furthermore, depletion of GSH by DHA could induce increases in other intracellular reducing equivalents (such as thioredoxins) whose pool size would be expected to impact the rate of DHA to VitC reduction. Timm et al. have shown that DHA administration resulted in a rapid increase in GSSG/GSH ratio and pentose phosphate pathway (PPP) flux in cells and tumors 20 . It can be argued that the initial DHA administration for baseline imaging could upregulate PPP flux and other reducing equivalents, which would be a confounding factor for our study. Nevertheless, the DEM treatment seemed to offset the increase in PPP flux and other reducing equivalents as the VitC production rate decreased significantly. Second, the supraphysiologic concentration of administered HP [1-13 C]DHA could saturate the transporter or relevant enzymes and create a rate-limiting condition, or result in substrate inhibition and decrease of enzymatic flux. Third, unlike the first-order reaction model, the kinetic parameters generated from pseudo-zero-order model are highly influenced by the numerical scale (i.e. SNR), which could potentially explain the inter-subject variability. For this reason, our dynamic acquisition and analysis scheme could only detect the relative changes in GSH content, but not the absolute concentration.
Another important consideration is the physiologic effect of DEM treatment in rats. We noted that rats became lethargic after DEM treatment. Numerous studies have explored the potential toxic effects of DEM in the heart and lung, and the dose we used was approximately 40% of the LD 50 reported for oral administration to rats [28][29][30][31] . We also observed that the total 13 C SNR was approximately 35% lower on average in DEM-treated than normal rats despite similar DNP solid state buildup, suggesting less HP 13 C substrate was delivered to the imaging slab. Although we have demonstrated that the rate of VitC production is independent of DHA dose administered in our supraphysiologic dose range, it is still possible that lower DHA substrate delivery contributed to the lower VitC production rate.
In addition to ascorbate-derived probes designed for HP MR, we also leverage 11 C radiopharmaceuticals, namely [1-11 C]DHA and [1-11 C]VitC, to study the uptake and biodistribution of ascorbates. [1-11 C]VitC showed lower accumulation than [1-11 C]DHA in the brain, consistent with HP 13 C MR findings and previously reported PET imaging data 21 , but higher retention in the lung and liver 1 hour after administration, possibly due to high expression of sodium-depended vitamin C transporters (SVCT) in these tissues [32][33][34] . [1-11 C]DHA reduction to [1-11 C]VitC represents a potential trapping mechanism 21,26 , with unreduced [1-11 C]DHA likely washed out of the cell. In the DEM redox modulation study using [1-11 C]DHA, we initially hypothesized that the lower brain GSH content in DEM-treated cannot adequately reduce [1-11 C]DHA to [1-11 C]VitC, and un-reacted [1-11 C]DHA would be washed out resulting less [1-11 C]VitC retention (lower gamma counting). However, there was no difference in 11 C radiopharmaceutical retention rate in major organs between the normal and DEM-treated groups.
There are several important differences between HP 13 C MR and 11 C biodistribution studies that could explain the inconsistency between their findings. As discussed in Methods section, 11 C radiopharmaceuticals were administrated at a considerably lower pharmacologic dose compared to HP 13 C probes (0.16μmol vs. 30μmol). It is possible that residual GSH in DEM-treated rats is adequate to reduce the lower pharmacologic dose of DHA used in 11 C studies. More importantly, the two technologies differ significantly in their mechanisms of generating image contrast and timing, making direct comparison of data difficult. Biodistribution studies were performed at 1 hour after DHA administration as compared to the 1 minute kinetic analysis applied to HP MR studies. Hence, it appears that although [1-11 C]VitC was likely generated at a slower rate in GSH-depleted animals, there was insufficient washout of unreacted [1-11 C]DHA to yield observable differences in 11 C accumulation. It is possible that dynamic PET imaging or biodistribution studies by gamma counting at earlier time points could reveal subtle differences in the radiopharmaceutical accumulation rate in the brain. Overall, our results suggested that the [1-11 C]DHA was insensitive to changes in redox capacity in this pharmacologic GSH depletion model using our experiment set-up. This highlights the potential value in studying real-time substrate conversion by HP MR as opposed to probe retention rate by biodistribution of radiopharmaceutical or PET imaging analysis.
In conclusion, we demonstrated the transport and biodistribution features of DHA and VitC in the brain by both HP 13 C MR and 11 C PET probes, and found that the kinetic rates of [1-13 C]DHA to [1-13 C]VitC conversion correlated with brain GSH content. These studies suggest that hyperpolarized [1-13 C]DHA can assess brain redox capacity non-invasively, and potentially allow visualization of oxidative stress in numerous brain pathologies. 11 C radiopharmaceutical synthesis. [1-11 C]VitC were synthesized from L-xylosone and [ 11 C]HCN precursors using a GE PETtrace cyclotron at our on-site radiopharmaceutical facility as previously described 21 and reported in the supplementary information. Overall, [1-11 C]VitC was obtained with a decay-corrected radiochemical yield of 35.8 ± 18% and a specific activity of 1.7 ± 0.4 mCi/μmol (n = 3). [1-11 C]DHA was obtained by oxidation of [1-11 C]VitC with a decay-corrected radiochemical yield of 25.8 ± 2.6% and a specific activity of 1.85 ± 0.5 mCi/μmol (n = 6).

MR studies.
All animal experiments were approved by our Institutional Animal Care and Use Committee (IACUC) and performed in accordance with IACUC protocols. Rats were placed under 1.5-2% isoflurane anesthesia and administered 30 μmoles HP [1-13 C]DHA or 100 μmoles HP [1-13 C]VitC solution via tail-vein catheter for both 2D CSI studies and dynamic 1D spectroscopy. For group (1) rats (n = 6 for DHA and n = 3 for VitC), two-dimensional single time-point CSI was performed on a GE Signa 3 Tesla MR scanner using a dual-tuned 1 H-13 C coil with the following parameters: voxel size = 5 × 5 × 20 mm, TE/TR = 135/210 ms, flip angle = 20°, spectral bandwidth/resolution = 5000/9.77 Hz, and 25 s acquisition delay after the start of injection. For group (2) rats, one-dimensional slice-selective dynamic spectroscopy with HP [1-13 C]DHA was performed using a similar set-up on a GE Signa 3 Telsa (n = 2 before and after DEM treatment) and a Bruker Biospec 3 Tesla MR scanner (n = 3 before DEM treatment, n = 2 after DEM treatment, n = 3 for 60% dose experiments), with the following parameters: spectral bandwidth/resolution = 25 kHz/12.2 Hz, slice thickness = 10 mm, TR = 2 s, 10 repetitions, RF-compensated variable flip angle scheme 35,36 , and 25 s acquisition delay after the start of injection. Axial and sagittal T 2 -weighted anatomic images were acquired with a fast spin echo sequence prior to HP 13 C studies and used to prescribe the spectroscopy slab. Rats were fasted for 8 hours before HP 13 C MR studies to reduce glucose competition with DHA for GLUTs as previously reported 14,17,37 . unknown parameter to estimate apparent in vivo T 1 of VitC, which was subsequently incorporated into the kinetic model as a fixed parameter to generate k assuming T 1 is the same across the subjects. Apparent kinetic rate (k app ) reflects the rate of observed VitC signal change caused by both T 1 relaxation and VitC production from DHA, while k reflects only the VitC production rate. A paired Student's t-test was used to compare kinetic parameters before and after DEM treatment. An unpaired Student's t-test was used to compare the GSH content of rat brains with and without DEM treatment and biodistribution of [1-11 C]ascorbates. All statistical analyses were performed using GraphPad Prism 7d (La Jolla, CA). All quantitative results are reported and graphed as mean ± standard deviation; p values < 0.05 were considered statistically significant, and our p values for biodistribution analysis were not adjusted for multiple comparisons.
Data availability. The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.