In vivo investigation of hyperpolarized [1,3-13C2]acetoacetate as a metabolic probe in normal brain and in glioma

Dysregulation in NAD+/NADH levels is associated with increased cell division and elevated levels of reactive oxygen species in rapidly proliferating cancer cells. Conversion of the ketone body acetoacetate (AcAc) to β-hydroxybutyrate (β-HB) by the mitochondrial enzyme β-hydroxybutyrate dehydrogenase (BDH) depends upon NADH availability. The β-HB-to-AcAc ratio is therefore expected to reflect mitochondrial redox. Previous studies reported the potential of hyperpolarized 13C-AcAc to monitor mitochondrial redox in cells, perfused organs and in vivo. However, the ability of hyperpolarized 13C-AcAc to cross the blood brain barrier (BBB) and its potential to monitor brain metabolism remained unknown. Our goal was to assess the value of hyperpolarized [1,3-13C2]AcAc in healthy and tumor-bearing mice in vivo. Following hyperpolarized [1,3-13C2]AcAc injection, production of [1,3-13C2]β-HB was detected in normal and tumor-bearing mice. Significantly higher levels of [1-13C]AcAc and lower [1-13C]β-HB-to-[1-13C]AcAc ratios were observed in tumor-bearing mice. These results were consistent with decreased BDH activity in tumors and associated with increased total cellular NAD+/NADH. Our study confirmed that AcAc crosses the BBB and can be used for monitoring metabolism in the brain. It highlights the potential of AcAc for future clinical translation and its potential utility for monitoring metabolic changes associated with glioma, and other neurological disorders.

Evaluation of NAD + /NADH ratio and BDH activity in tissues. The enzymatic conversion of acetoacetate to β-hydroxybutyrate requires NADH as a co-factor and relies on BDH enzyme activity. To assess if changes in the in vivo ratio of [1-13 C]β-HB-to-[1-13 C]AcAc could be linked to alterations in cellular redox, we used a commercial spectrophotometric assay to evaluate the levels of NAD + and NADH. The hyperpolarized 13 C slab from which data were acquired contained both tumor and normal-appearing brain tissue for tumor-bearing mice, but only normal brain tissue for tumor-free mice. Therefore, tumor and normal-appearing brain tissues were collected and freeze-clamped for tumor-bearing mice, while only normal brain samples were collected for Hyperpolarized acquisitions were started 10 sec after the beginning of the hyperpolarized injection (350 μL of [1,3-13 C 2 ]AcAc injected over 12 sec). All dynamic data were summed. Resonances of [1-13 C]AcAc and [1-13 C]β-HB were fitted with a Lorentzian-Gaussian line shape using MestreNova for each animal and integrals of the fits normalized to SD of the noise were quantified. (C) Representative hyperpolarized 13 C data obtained from the sum of all dynamic data from a control mouse (left), U87wt-bearing mouse (middle) and U87mut-bearing mouse (right), showing that hyperpolarized [1-13 C]β-HB production could be detected in all three groups. (D) Quantification of [1-13 C]AcAc and [1-13 C]β-HB levels and ratio of [1-13 C]β-HB-to-[1-13 C]AcAc. A significant increase in [1-13 C]AcAc level in U87wt tumor-bearing mice and control mice and between U87wt and U87mut tumor-bearing mice was observed. A significant decrease in [1-13 C]β-HB-to-[1-13 C]AcAc ratio in both tumor models compared to normal mice was detected. No difference between U87wt and U87mut tumor-bearing mice was observed. SNR, signal to noise ratio; A.U., arbitrary units; AcAc, acetoacetate; β-HB, β-hydroxybutyrate. (2019) 9:3402 | https://doi.org/10.1038/s41598-019-39677-2 www.nature.com/scientificreports www.nature.com/scientificreports/ tumor-free mice (Fig. 5A). The NAD + /NADH ratios in U87wt tumor samples (n = 7), U87mut (n = 10) tumor samples, normal-appearing brain tissue from both tumor groups, and tissue from control tumor-free mice (n = 8) were, respectively, 7.8 ± 2.8, 5.1 ± 3.1, 2.9 ± 1.2, 2.3 ± 1.3 and 2.1 ± 0.6 ( Fig. 5B). The NAD + /NADH ratio was therefore significantly higher in the tumor tissue compared to normal-appearing brain tissue in both U87wtand U87-mut bearing mice as well as compared to normal brain tissue from control mice. We then investigated BDH enzyme activity in order to determine if changes in redox and conversion of hyperpolarized [1-13 C]AcAc to [1-13 C]β-HB in our tumor models could also be explained by alterations in the BDH enzyme. BDH enzyme activities in U87wt tumor samples (n = 7), U87mut (n = 8) tumor samples, normal-appearing brain tissue samples and tissue from controls (n = 7) were 0.05 ± 0.03, 0.37 ± 0.30, 0.04 ± 0.03, 0.27 ± 0.16 and 0.36 ± 0.15 μmol/ min/μg of protein respectively (Fig. 5C), thus BDH enzyme activity was significantly lower in tumors compared to normal-appearing brain of U87wt-and U87-mut bearing mice as well as compared to normal brain tissue of control animals with no tumors.

Discussion
Our goal was to assess the value of hyperpolarized acetoacetate as a probe to monitor redox in the brain and to develop a non-invasive MR-based method to image metabolic alterations associated with brain disorders. We illustrate here the feasibility of using [1,3-13 C 2 ]AcAc and monitoring its conversion to [1,3-13 C 2 ]β-HB in healthy and tumor-bearing mouse brain in vivo. Our findings are associated with changes in BDH activity as well as alterations in overall cellular NAD + /NADH. AcAc and β-HB are two ketone bodies that can serve as the brain's main alternative fuels to glucose. They are rapidly transported via monocarboxylate transporters through the BBB into cells and mitochondria 32,33 . As such, we could readily detect the delivery of hyperpolarized AcAc into the brain of a normal mouse. Furthermore, no toxic effects were observed in our study following intravenous injection of AcAc, consistent with previous studies [26][27][28]30 . Beyond its biological utility, our study confirmed previous reports that [1,3-13 C 2 ]AcAc fulfills the technical requirements for dissolution DNP and is therefore a promising hyperpolarized probe [26][27][28] . Specifically, the liquid-state polarization levels (~20%) were sufficient to achieve adequate SNR for monitoring the fate of AcAc, and both labeled carbons (C1 and C3) have T1 relaxation times that are long enough for detection of metabolism (~30 sec). Our study was performed at relatively high magnetic fields (11.7 and 14T), and longer T1 values are expected at lower clinical magnetic fields for labeled carboxyl groups, as previously illustrated by a study at 3T 30 . Labeling of AcAc on C1 and C3 allows for potential detection of signal from C1 and C3 of β-HB following metabolic conversion by BDH enzyme. Importantly, there is sufficient chemical shift separation between substrate and product, i.e. 5.6 ppm between [1-13 C]AcAc and [1-13 C]β-HB and 144.6 ppm between [3-13 C]AcAc and [3-13 C] β-HB 26,30 . A recent study also reported a T1 value for [1-13 C]β-HB of ~20 sec at 11.7T 28 , which is sufficiently long to detect signal in vivo. But it should be noted that in the case of [3-13 C]β-HB, the directly bonded proton leads to a very short T1 that precluded detection of this second signal in vivo in our study at high field. A potential limitation of AcAc is that it is not commercially available because of its natural instability and spontaneous www.nature.com/scientificreports www.nature.com/scientificreports/ decarboxylation to acetone and carbon dioxide. Therefore, 13 C-AcAc needs to be synthesized prior to each set of hyperpolarized experiments.
Recent studies illustrated the potential of using hyperpolarized [1,3-13 C 2 ]AcAc as a biomarker of mitochondrial redox in vivo in the abdomen and in the heart 28,30 . Here we showed, to our knowledge for the first time, the feasibility of using hyperpolarized [1,3-13 C 2 ]AcAc to probe metabolism in the brain. Following intravenous injection of [1,3-13 C 2 ]AcAc, we were able to clearly monitor the production of [1-13 C]β-HB in healthy mice demonstrating the utility of this hyperpolarized agent for brain studies.
As previously mentioned, the IDH1 mutation is the most prevalent driver mutation in low-grade glioma and upgraded secondary glioblastoma 11,13 . It is responsible for the conversion of α-KG to 2-HG using NADPH as a cofactor 37,38 , and previous studies have shown that the mutation results in increased reactive oxygen species, and changes in NADPH 15,39,40 . Nonetheless, and although NADP + /NADPH and NAD + /NADH are interconnected 10 , we did not observe any difference in the NAD + /NADH ratio between the U87 IDH1 wild type and U87 IDH1 mutant tumor extracts (p value = 0.09). We also did not observe any difference in BDH enzyme activity in these tumor tissues (p value = 0.97). A possible reason is that our model was generated by genetically engineering glioblastoma cells to express mutant or wild type IDH1. As such, our cells might not fully recapitulate the complete metabolic signature associated with IDH1 mutant glioma. Indeed, a previous study in our lab did not find any differences in NADP + /NADPH between our two models (Fig. S3). Importantly however, consistent with the comparable NAD + /NADH ratios and BDH activities, we also observed that the ratio of hyperpolarized [1-13 C] β-HB-to-[1-13 C]AcAc was comparable between mice bearing U87wt tumors and mice bearing U87mut tumors (p value = 1).
When comparing tumor-bearing mice to controls, an increase in the hyperpolarized [1,3-13 C 2 ]AcAc signal was observed in the U87wt-tumor mice in the dynamic acquisitions. Similarly, in the 90° acquisitions, we observed an increase in the [1,3-13 C 2 ]AcAc signal in the U87wt-bearing mice compared to control mice, and an increase in [1-13 C]AcAc and a trend towards an increase in [3-13 C]AcAc in U87mut-bearing mice (p value = 0.15). The small Figure 5. NAD + /NADH ratio and BDH activity quantified by spectrophotometric assays. (A) 1D slab 13 C acquisitions were 10 mm-thick and therefore covered the whole brain. In tumor-bearing mice, the slab contained both tumor and normal-appearing brain whereas in tumor-free (control) mice, it contained only normal brain. At sacrifice, both normal-appearing and tumor tissues from tumor-bearing mice and normal brain tissues from control mice were freeze-clamped and kept at −80 °C for spectrophotometric assays. (B) NAD + and NADH levels were measured and NAD + /NADH ratios were quantified. No differences were observed in NAD + /NADH ratio between control normal brain tissue and tumor-bearing mouse normalappearing brain tissue and between U87wt and U87mut tumor tissue. A significant increase in NAD + /NADH ratio was observed between controls normal brain and U87wt/U87mut tumor tissues as well as between tumorbearing normal-appearing brain tissues and U87wt/U87mut tumor tissues. (C) BDH activity from isolated mitochondria was quantified. No differences were observed between controls normal brain tissues and tumorbearing mice normal-appearing brain tissues. A significant decrease in BDH activity in tumor tissues compared to normal-appearing brain tissues was observed for both tumor models. A significant decrease was measured between U87wt and U87mut tumor tissues and normal brain tissues from healthy controls. N.B., Normal Brain; N-A.B., Normal-Appearing Brain; T., Tumor. (2019) 9:3402 | https://doi.org/10.1038/s41598-019-39677-2 www.nature.com/scientificreports www.nature.com/scientificreports/ differences between dynamic and 90° acquisitions are likely due to differences in SNR. The higher [1,3-13 C 2 ]AcAc in tumor-bearing mice is likely due to an increase in its delivery to the tumor region as a result of breakdown of the BBB and increased vascular permeability [41][42][43] . However, although [1,3-13 C 2 ]AcAc signal was increased in tumor-bearing mice, no changes in the [1-13 C]β-HB levels were detected, pointing to reduced metabolism in the tumor tissue. As a result, there was a significantly lower (~38%, average for U87wt and U87mut tumors) [1-13 C] β-HB-to-[1-13 C]AcAc ratio in the brain of tumor-bearing mice compared to tumor-free healthy controls.
To assess the association between lower conversion of [1-13 C]AcAc to [1-13 C]β-HB and cellular metabolism we determined cellular NAD + /NADH and mitochondrial BDH enzyme activity using spectrophotometric assays. The cellular NAD + /NADH ratios reported here were within the range of values found in the literature 18,44 . The increase observed in our tumors was consistent with one previous study 45 but most previous reports suggest that the ratio of NAD + /NADH should decrease in cancer cells 46 . However, it should be noted that we are comparing tumor tissue of human origin to normal tissue from the rodent. Thus the primary value of this data is in assessing the significance of our hyperpolarized information. Our studies showed a significant increase (~60-70%) in NAD + /NADH in the tumor tissues compared to normal brain, which is consistent with the hyperpolarized observations. Nonetheless, as mentioned above, the metabolism of AcAc occurs in the mitochondrial compartment, whereas our ex vivo NAD + /NADH measurement represents an average of the mitochondrial and cytosolic pools (the kit used does not allow differentiation between the NAD + and NADH pools (both free and bound forms) of the cytosol and mitochondria). Therefore, the increase in NAD + /NADH ratio measured here in tumor-bearing mice cannot directly be linked to changes in redox at the level of the mitochondria. Another limitation of our study is that, due to insufficient SNR, kinetic analysis of the hyperpolarized data could not be conducted, and therefore we cannot ascertain that equilibrium between AcAc and β-HB has been reached. As a result we cannot directly infer the mitochondrial NAD + /NADH ratio from our hyperpolarized measurement. In contrast however, our findings with regard to the activity of the BDH enzyme identify a likely link between hyperpolarized AcAc metabolism and enzyme activity that were both significantly reduced in tumor tissues compared to normal brain. Lower BDH activity is also consistent with previous reports showing a decrease in BDH activity in a rat hepatoma cell lines compared to normal rat hepatocytes 47 . This finding is also in good agreement with published data reporting lower BDH-1 expression in glioma cells compared to normal brain 34,35 . Collectively, our findings therefore support the conclusion that the hyperpolarized [1-13 C]β-HB-to-[1-13 C]AcAc ratio is informative of mitochondrial BDH activity and could be indicative of cellular NAD + /NADH ratio.
Further work is needed to acquire 2D images of [1-13 C]β-HB production and distribution. For instance, a multiband pulse sequence that applies a small flip angle to 13 C-AcAc to preserve its magnetization while a larger flip angle is applied to [ 13 C]β-HB to increase its SNR, would improve the likelihood of detecting metabolism at a higher spatial resolution and performing kinetic analysis. Nonetheless, the low toxicity of AcAc and its ability to cross the BBB are encouraging for translation of this hyperpolarized agent to the clinic. In addition, as mentioned above, the lower magnetic field used in the clinic will increase the T1 of both [1-13 C]AcAc and [1-13 C]β-HB, facilitating the detection of metabolism.
Beyond cancer, previous studies have reported changes in the ratio of AcAc-to-β-HB in the immature rat brain under cerebral metabolic stress (hypoxia, ischemia, anoxia) and in the aging rat brain 48,49 . In addition, other reports illustrate the importance of BDH in oxidative stress models, and of ketone bodies in parkinson disease and traumatic brain injury 32,50,51 . These investigations point to the potential of AcAc for investigating a range of neurological diseases.
In summary, this study shows for the first time the feasibility of using hyperpolarized [1,3-13 C 2 ]AcAc to monitor the conversion of AcAc to β-HB in the brain and in glioma in vivo. The reduced β-HB production in tumor models in vivo was consistent with lower BDH activity and in agreement with higher NAD + /NADH. Our findings thus demonstrate the potential of this hyperpolarized agent to measure metabolism in the normal and diseased brain.
Relaxation and polarization levels. Following beginning of dissolution, 2 mL of hyperpolarized [1,3-13 C 2 ]AcAc was transferred within ~20-25 sec into a 10 mm NMR tube at 37 °C. Dynamic 13 C spectra were then acquired immediately using a non-localized single pulse acquisition (TR = 3 sec, FA = 5°, SW = 20 kHz, np = 20 k, NT = 100, number of repeats (n) = 5) on a 11.7T INOVA spectrometer (Agilent Technologies, USA) using a 10 mm broadband probe. Following total decay of the hyperpolarized signal, a thermal equilibrium spectrum was acquired using FA = 90°, TR = 300 sec, NT = 16 and other parameters identical to the ones mentioned above (n = 3). T 1 of hyperpolarized [1-13 C]AcAc and [3-13 C]AcAc were determined by quantifying peak integrals using MestreNova, correcting for flip angle, and fitting the signal decay with a mono-exponential curve. The level of polarization in solution was calculated by comparing the signal on the first hyperpolarized spectrum of the dynamic set to the corresponding signal in the thermal equilibrium spectrum, after correction for flip angle and number of transients.
Animal models. All studies were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committee (IACUC) of the University of California (San Francisco, CA). A total of 28 6-week old female athymic nu/nu mice were used. 8 mice served as tumor-free controls and 20 mice were intracranially injected with 3 × 10 5 tumor cells. 8 mice were implanted with IDH1 wild-type (U87wt) U87 cells and 12 with U87 cells genetically-engineered to express mutant IDH1 (U87mut) 37 . 4 U87mut and 1 U87wt mice underwent the hyperpolarized study prior to tumor implantation and served as additional tumor-free controls. Prior to implantation, cell lines were maintained between passages 15 and 30 in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum, 2 mM glutamine, and 100 U/ml penicillin and streptomycin under normoxic conditions 37 . Both cell lines were routinely tested for mycoplasma contamination and authenticated by short tandem repeat fingerprinting (Cell Line Genetics) within 6 months of any study.

MRI and hyperpolarized 13 C MRS studies in vivo.
Studies were performed on a 14.1T vertical MR system (Agilent technologies). Animals were anesthetized using isoflurane (1-2% in O 2 , 1 L/min). Axial anatomical T 2 -weighted images were acquired using a multi-slice spin-echo sequence (TE = 20 ms, TR = 1200 ms, in-plane resolution = 0.12 × 0.12 mm 3 , slice thickness (thk) = 1 mm, number of acquisition (NA) = 2) and a single channel volume 1 H coil in order to monitor tumor size. Once tumors reached a volume of ~0.24 cm 3 , hyperpolarized experiments were performed on tumor-bearing animals using a dual-tune volume 1 H-13 C coil. Prior to hyperpolarized studies, a 27-gauge catheter was secured in the tail vein for injection of hyperpolarized material. T 2 -weighted images were acquired using a multi-slice spin-echo sequence (same parameters as above, except thk = 1.5 mm) to position a 10 mm-thick axial 13 C slab through the brain for hyperpolarized acquisitions. Each mouse was injected with 350 μL of hyperpolarized [1,3-13 C 2 ]AcAc solution in the tail-vein over 12 sec. Dynamic 13 C slice-localized spectra (FA = 20°, NT = 10) were acquired every 4 sec from the 10 mm-thick slab, starting at the end of the injection, followed by a 90° 13 C slice-localized acquisition (NT = 1). Tumor-free mice (controls) underwent the same hyperpolarized experiments as tumor-bearing mice.
In vivo MR data analysis. Tumor size was measured by manually contouring the tumor area in each slice and summing all the areas multiplied by slice thickness using in-house MR software (MRSC Image). Tumor size was quantified for each animal. The hyperpolarized experiments were performed when the tumor reached a volume of ~0.24 cm 3 . Hyperpolarized 13 C MRS data were analyzed as follows using MestreNova. Each 13 C spectrum from the dynamic set was apodized (line broadening (lb) = 20 Hz) and phased. All dynamic data were then summed. 13 C spectra from 90° acquisitions were apodized (lb = 40 Hz) and phased. Resonances of [1-13 C]AcAc, [3-13 C]AcAc and [1-13 C]β-HB were fitted with a Lorentzian-Gaussian line shape. Finally, integrals of the fits were normalized to standard deviation (SD) of the noise and ratios of [1-13 C]β-HB-to-[1-13 C]AcAc were quantified for each animal.
Spectrophotometric assays. Following hyperpolarized experiments, animals were euthanized and tissue collected for spectrophotometric assays. Hyperpolarized signal was recorded from a 10 mm-thick slab that included both tumor and normal brain tissues for tumor-bearing mice, and normal brain only for controls. Therefore, both tumor and normal-appearing brain tissues were collected and freeze-clamped for tumor-bearing mice, while normal brain tissue was only collected for tumor-free mice. All tissue samples were kept at −80 °C until they were processed for the spectrophotometric assays. NAD + and NADH levels were quantified for each sample (~25 mg, n = 7 U87wt samples, n = 10 U87mut samples, n = 8 tumor-free samples) using a commercial colorimetric kit (BioVision, USA) and following manufacturer instructions. BDH enzyme activity was quantified (~25 mg, n = 7 U87wt samples, n = 8 U87mut samples, n = 7 tumor-free samples) using previously described methods 55 . The collected tissue was first used to conduct the NAD + /NADH assay, therefore for some samples, not enough tissue remained for the BDH activity assays resulting in less repeats for the BDH assay. For both assays and all samples, tissues were homogenized using a tissue lyser (Qiagen, USA) and further processed as follows. NAD + /NADH ratio: Briefly, NADt (NAD + +NADH) was extracted following manufacturer instructions. For each of the extracted samples, half of the sample was heated to 60 °C for 30 min to decompose NAD + while keeping NADH intact. Both NADt and NADH samples were mixed with NAD cycling enzyme and absorbance was measured at 450 nm using an Infinite 300 m200 spectrophotometer (Tecan Systems, Inc., USA). NADt and NADH were quantified by comparing with NADH standard curve and normalizing to mg of protein (determined by a Bradford protein assay). Finally, ratios of NAD + (NADt -NADH) to NADH were calculated. BDH enzyme