Hyperpolarized 13C magnetic resonance imaging in neonatal hypoxic–ischemic encephalopathy: First investigations in a large animal model

Early biomarkers of cerebral damage are essential for accurate prognosis, timely intervention, and evaluation of new treatment modalities in newborn infants with hypoxia and ischemia at birth. Hyperpolarized 13C magnetic resonance imaging (MRI) is a novel method with which to quantify metabolism in vivo with unprecedented sensitivity. We aimed to investigate the applicability of hyperpolarized 13C MRI in a newborn piglet model and whether this method may identify early changes in cerebral metabolism after a standardized hypoxic–ischemic (HI) insult. Six piglets were anesthetized and subjected to a standardized HI insult. Imaging was performed prior to and 2 h after the insult on a 3‐T MR scanner. For 13C studies, [1‐13C]pyruvate was hyperpolarized in a commercial polarizer. Following intravenous injection, images were acquired using metabolic‐specific imaging. HI resulted in a metabolic shift with a decrease in pyruvate to bicarbonate metabolism and an increase in pyruvate to lactate metabolism (lactate/bicarbonate ratio, mean [SD]; 2.28 [0.36] vs. 3.96 [0.91]). This is the first study to show that hyperpolarized 13C MRI can be used in newborn piglets and applied to evaluate early changes in cerebral metabolism after an HI insult.

assessment of brain damage is also used to acquire a more detailed understanding of the pathophysiology involved in HIE and to evaluate novel neuroprotective interventions. 4 Early, valid, and clinically applicable biomarkers reflecting the degree of brain damage in HIE are thus warranted.
Magnetic resonance imaging (MRI) and magnetic resonance spectroscopy (MRS) are used as noninvasive tools to evaluate neural damage in the clinical and preclinical setting. 5,6In human neonatal HIE, the thalamic lactate to n-acetyl aspartate (Lac/NAA) ratio measured by 1 H-MRS at 4-15 days after birth has shown promise as a biomarker to predict long-term outcomes [6][7][8] The thalamic Lac/NAA ratio measured at 48 h after HIE has been shown to correlate with the degree of cell death assessed by histology in newborn piglets. 9However, 1 H-MRS is limited by relatively low signal-to-noise ratio (SNR) and the risk of contamination from lipid signals. 10 1H-MRS is also limited to metabolic profiling of few sampling points in the spatial dimension to maintain acceptable acquisition times.Furthermore, 1 H-MRS is unable to detect rapid kinetic changes of the metabolite quantified.
2][13] Hyperpolarized 13 C MRI amplifies the signal of 13 C-enriched molecules more than 10,000 times. 14 C MRI may therefore potentially supplement 1 H-MRS to improve the mapping of cerebral metabolism in evolving HIE even further. 13C MRI has been used in P10 mice to show changes in cerebral metabolism and an increase in Lac production after an HI insult. 15However, the application of 13 C MRI has yet to be investigated in a larger newborn animal model.
As a first step in the investigation of 13 C MRI as a potential novel biomarker of neural damage in HIE, we set out to investigate, first, whether 13 C MRI may be utilized in the newborn piglet model to quantify cerebral metabolism in vivo, and second, whether 13 C MRI can detect a change in the metabolic phenotype shortly after an HI insult.

| MATERIALS AND METHODS
This project was approved by the Danish Animal Inspectorate (Permission number 2016-15-0201-01052).This study is reported in accordance with the ARRIVE guidelines. 16The piglet model of HIE has been described in detail elsewhere. 17
Anesthesia was maintained by propofol 4-10 mg/kg/h and fentanyl 5-12 μg/kg/h.The level of sedation was adjusted to secure anesthesia and analgesia, and to minimize the use of medication.The ventilator was adjusted to maintain normoventilation assessed by end tidal CO 2 .sO 2 %, heart rate, mean arterial blood pressure (MABP), core temperature, and electrocardiogram (ECG) were continuously recorded (Datex Ohmeda S/5 Collect, Finland).A single-channel amplitude-integrated electroencephalogram (aEEG) was acquired by electrodes installed in the scalp (Natus Medical Incorporated, CA, USA).Prophylactic doses of gentamicin and ampicillin were given after intubation.Blood glucose and electrolytes were monitored and maintained within normal range by infusion of NeoKNaG (Na + , 15 mmol/L; K + , 10 mmol/L; Cl À , 25 mmol/L; glucose, 505 mmol/L).Deviations in glucose or electrolyte levels were corrected through changes in infusion fluid in accordance with local clinical guidelines used in the neonatal ward.We aimed to keep MABP above 40 mmHg.Hypotension was treated by a reduction of anesthetics and a bolus of normal NaCl (20 mL/kg).If this was insufficient, an infusion of noradrenaline 0.25-1.5 μg/kg/min and/or dopamine 5-15 μg/kg/min was initiated. 18

| HI insult
A global HI insult was induced over 45 min.The global insult was chosen to mimic the clinical setting where a neonate will experience the combination of a brain-specific event with hypoxia and ischemia combined with the systemic effect with multiorgan dysfunction and metabolic acidosis. 17 2 was titrated by aEEG and MABP to ensure survival, but with a clinically relevant insult severity.The target was an aEEG trace below 7 μV, combined with hypotension (MABP < 70% of baseline) for at least 10 min.To ensure sufficient hypotension, FiO 2 was further reduced to achieve a MABP less than 30 mmHg at some point during the 10-min period.FiO 2 was shortly increased if MABP was less than 50% of baseline, HR less than 80 min À1 , or aEEG less than 3 μV, to ensure survival.After 45 min the piglet was resuscitated with room air.If needed, FiO 2 was increased to keep SatO 2 at more than 90%.Arterial blood gas was measured at 15, 30, and 45 min during the insult.The core temperature of all animals was kept stable during the insult.

| MRI
Imaging was performed at baseline (before HI) and 2 h after the HI insult on a 3-T MR system (GE Healthcare MR750) equipped with a 1 H/ 13 C dual-tuned volume coil for rodent imaging (Rapid Biomedical, Germany).T1, T2, and diffusion-weighted images were acquired following in-house routine protocols.For 13 C studies, 127 mg [1-13 C]pyruvate was hyperpolarized in the SpinLab polarizer (GE Healthcare) with AH111501 (15 mM) as the radical. 19After hyperpolarization, the pyruvate was dissolved into neutralization buffer, yielding hyperpolarized [1-13 C]pyruvate with a concentration of $125 mM and polarization more than 40%.Following rapid injection (6 mL) into an ear vein, images were acquired using a single-slice spectro-spatial sequence with spiral readout (16-mm slice thickness, real pixel size 5 Â 5 mm, TR 500 ms). 20Pyruvate (flip angle 8 ) was excited interleaved between metabolites (flip angle 90 ).The transmit gain and center frequency were determined by a Lac phantom placed on the scalp of the piglet using the Bloch-Siegert shift approach (BLOSI). 21Correction for polarization at the time of injection was not performed as this is done inherently by normalization to the pyruvate signal in the region of interest.Enzymatic kinetics and conversion of pyruvate to metabolite were estimated by means of a two-sided metabolic exchange model and the area under the curve approach. 22,23The regions of interest were the whole brain, cortex, and deep gray matter in the thalamus (Figure S1).

| Experimental protocol
A total of six piglets were included.After anesthesia and monitoring were established, the animals were allowed a rest for 30-60 min.They were then subjected to a baseline scan, an HI insult, and scanned again 2 h after the insult.After the last scan animals were euthanized by means of an iv injection of pentobarbital (Figure S2).

| Statistics
Vital signs, blood-gas values, animal weight, and measures representing insult-severity are presented as median and range.MRI results are presented as mean and standard deviation (SD).Use of inotropes is reported as an average amount by hours used, and hours used after the HI insult.Change in Lac and bicarbonate production were tested with a paired Student's t-test.Changes in pyruvate signal in time series data were tested with a repeated measures analysis of variance (ANOVA) test.A two-sided p value of less than 0.05 was considered statically significant.13 C MRI has not previously been used in piglets with HIE.Before the start of the current study, pilot scans were performed on two healthy piglets.
The number of animals included was an estimate based on data from these pilot animals.

| Survival and insult severity
All animals survived until euthanasia was planned.The HI insult resulted in severe aEEG suppression, hypotension, and metabolic acidosis (Tables 1 and 2).Norepinephrine was used to treat hypotension after the HI insult in four of six piglets (average dose of 0.65 ug/kg/min).Apparent diffusion coefficient (ADC) maps showed restricted diffusion after the HI insult compared with the baseline scan (Figure 1).
T A B L E 1 Insult severity in six piglets subjected to a global hypoxic-ischemic insult.Hypotension: mean arterial blood pressure lower than 70% of baseline values.Data are median with range.Abbreviations: aEEG, amplitude-integrated electroencephalogram; MABP, mean arterial blood pressure; min, minutes.

| [1-13 C]pyruvate and products
One baseline scan in one piglet was lost because of scanner problems resulting from frequency miscalibration.A second scan was lost due to lack of signal in the brain after pyruvate injection, despite proper iv access and injection.Metabolic profiles of all piglets before and after the HI insult are visualized on a time scale (Figure 2).The time series showed a tendency to a reduction in pyruvate delivery and a reduction in bicarbonate production 2 h after HI (Figure 2).

| Metabolic phenotype and enzyme-kinetics analysis
The Lac/bicarbonate ratio showed a shift in metabolic phenotype in the thalamus after the HI insult (2.28 [0.36] vs. 3.96 [0.91], p = 0.04) (Figure 3).The same tendency was observed in the whole brain (2.12 [0.26] vs.  4. Conversion of pyruvate to Lac increased after the HI insult in all three regions of the brain (Figure 3).Conversion of pyruvate to bicarbonate was reduced in the whole brain (Figure 3).To the best of our knowledge, this is the first study to show that 13 C MRI can be utilized in a larger newborn animal model resembling human physiology and that 13 C MRI can identify changes in the metabolic phenotype shortly after a global HI insult.
There are currently few published studies on 13 C MRI in applied preclinical newborn models.In 2016, Chen et al. introduced the application of hyperpolarized 13 C MRI to the developing brain. 24They investigated the change in metabolic phenotype in healthy mice 18 days after delivery (P18 mice), a timing equivalent to the developmental stage of the human brain in early childhood. 24They found that Lac levels and conversion of pyruvate to Lac (k PL ) decreased with age. 24The authors proposed further investigations of the metabolic profile of the injured developing brain.
Mikrogeorgiou et al. investigated the change in metabolic phenotype in P10 mice (equivalent to term human newborns) subjected to an HI insult with 13 C MRI performed 2-5 h after the HI insult, at P17, and P31 days of age. 15In the scans acquired 2-5 h after the HI insult, they found a reduction in pyruvate delivery and an increased Lac/pyruvate ratio compared with prior to the insult. 15However, these changes reverted over time and HI animals showed similar values to sham animals at P31. 15 These two studies provide important information on the 13 C MRI metabolic profile of the developing brain under healthy and pathological conditions, and also indicate that timing of the scan is crucial.The results from P10 scans acquired 2-5 h after the HI insult were similar to our results in the gyrencephalic piglet model, which is more translatable to the human newborn brain.In accordance with Chen et al., we found a tendency of reduced pyruvate delivery after HI and a change in metabolic phenotype, with a shift from mitochondria-dependent metabolism with bicarbonate production towards pyruvate conversion to Lac.
We also found a change in metabolic phenotype with an increase in k PL and a decrease in k PB .This finding is in agreement with several known pathomechanisms of HIE that need to be considered.In the acute phase of injury, a lack of oxygen will inhibit oxidative phosphorylation in the mitochondria and result in anaerobe metabolism with Lac production. 25Puka-Sundvall et al. found that, despite sufficient delivery of oxygen and Examples of enzymatic kinetic conversion of pyruvate to lactate and bicarbonate in the whole brain at (A) Baseline, and (B) 2 h after the hypoxia ischemia (HI) insult.
substrates, mitochondrial impairment with condensation, edema, and calcium accumulation occurred 3 h after hypoxia in 7-day-old rats. 26Lac production has also been proposed as an energy substrate in addition to glucose. 27Lac produced in astrocytes is shuttled to neurons by monocarboxylate transporter (MCT)-2 in the astrocyte membrane via the MCT-4 in the neuronal cell membrane. 28MCT-2 and MCT-4 have been found to be upregulated at 2 h after an HI insult in piglets. 29Another source of Lac production is monocytes and macrophages. 30A known mechanism contributing to HIE is neuroinflammation, with activation and migration of microglia a few hours after the HI insult. 31,32Sriram et al.
showed increased Lac production in lipopolysaccharide (LPS)-stimulated macrophages in vitro, quantified by 13 C MRI. 33 They found increased lactate dehydrogenase (LDH) activity after LPS exposure and were able to differentiate between LPS-stimulated (M1 state) and LPS + indomethacin-stimulated (M2 state) macrophages. 33However, Lac production due to microglial infiltration or astrocyte delivery will not result in reduced bicarbonate production.This is consistent with our results as we found a more pronounced change in k PL than in k PB .Furthermore, the decrease in k PB was minor and only detected in the whole brain region of interest.One possible reason for this is the timing of the scan, as apoptosis and mitochondrial failure are expected to transpire in the secondary phase 24-48 h after the HI insult. 34 this study, we used a well-established model for studies in HIE and neuroimaging with MRI/MRS. 9,17,35Newborn piglets have not previously been used for neuroimaging with 13 C MRI.Previous 13 C MRI studies in adult anesthetized healthy pigs showed low SNR in the brain, which was improved by disruption of the blood-brain barrier (BBB) by mannitol infusion. 36In this study, 13 C MRI scans showed high SNR before and after HI.Compared with adult piglets, the pyruvate dosage relative to the animal's weight was higher in this study.Another reason for the higher SNR could be because of increased expression of the MCT-1 pyruvate transporter in the newborn brain. 37Pyruvate delivery to the cerebral tissue may also be increased during the acute phase of HIE as the HI insult will result in disruption of the BBB. 38The use of anesthetics has been shown to alter neural metabolism and influence 13 C MRI signal output. 39The piglets in this study were anesthetized by propofol and fentanyl.This needs to be taken into consideration, as the model with global HI may alter drug metabolism due to multiorgan failure, similar to what is seen with clinical HIE.To avoid the risk of drug accumulation and toxic effects we carefully adjusted the level of drugs to a minimum while securing anesthesia and pain relief. 40Despite these factors, we acquired 13 C MRI images with high SNR from both the healthy baseline scan and the post-HI scan.
The major limitation of this study is the missing two scans in two different animals.One scan was lost due to frequency miscalibration.In the other scan, pyruvate was correctly injected, but no signal was detected in the brain.A likely cause of the missing brain signal after pyruvate injection was reduced cerebral perfusion after the HI insult.We had no measures on perfusion in this study.However, in a previous study using the same piglet model, we showed that cerebral perfusion, measured by arterial spin labeling, can be severely reduced 6 h after the HI insult. 41cordingly, the potential lack of pyruvate delivery in piglets with severe brain damage needs to be considered when interpreting the results.
Whether reduced or absent cerebral blood flow is the cause of the compromised pyruvate delivery will be investigated in future studies.Pyruvate delivery will depend on cerebral perfusion, which has been shown to decrease in the hours following HI. 42In fetal sheep, hypoperfusion, increased oxygenation, and reduced metabolism was consistent with suppression of cerebral metabolism in the first hours after the HI insult. 43In a study of human neonates with HIE, arterial spin labeling showed reduced blood flow on Day 0-1 compared with healthy controls. 44Another limitation of this study is the lack of 1 H-MRS data, as this would allow for a comparison of 1 H-MRS with 13 C MRI.Although performed, 1 H-MRS data were lost postprocess due to technical reasons during data transfer to back-up drives.As the Lac/NAA peak metabolic ratio has been shown to correlate with neural cell death by histology from the same animal model, 1 H-MRS or histology measures would have allowed for a comparison of changes in metabolic phenotype measured by 13 C MRI and cell death. 9In the current study, piglets were scanned 2 h after the HI insult during the acute phase of injury.As neural damage after an HI insult is a dynamic process, future studies should focus on later scans to study the development of the metabolic changes, and to compare 13 C MRI with the better described 1 H-MRS and histological data, which mostly have resulted from later scans.Scans performed at a later time point will also allow for an evaluation of the changes in cerebral perfusion and pyruvate delivery to further elucidate this issue.

| CONCLUSION
Hyperpolarized 13 C MRI can be applied in the newborn piglet model to map changes in metabolic phenotype under healthy conditions and in the acute phase after an HI insult.Further studies are needed to investigate cerebral perfusion and pyruvate delivery at later time points and to com- Hansen, and Christoffer Laustsen performed data analysis.Ted C. K. Andelius drafted the first manuscript.All authors have critically reviewed the drafted manuscript.All authors have approved the final manuscript and agree to be held accountable for all aspects of the work. FiO

F I G U R E 1
Cerebral edema measured through ADC mapping on diffusion-weighted imaging showed reduced ADC values after the hypoxia ischemia (HI) insult.(A) ADC values at baseline and after HI. (B) ADC maps in a healthy piglet.(C) ADC maps 2 h after HI.Paired t-test in five animals.*p < 0.05.ADC, apparent diffusion coefficient; HIE, hypoxic-ischemic encephalopathy.F I G U R E 2 Time series 120 s after injection of [1-13 C]pyruvate in all animals (n = 5 baseline, n = 5 HI scan) before and 2 h after the HI insult.Data are mean and standard deviation.Data were tested by analysis of variance (ANOVA).HI, hypoxia ischemia.I G U R E 3 Lactate/bicarbonate ratios and enzymatic kinetic analyses at baseline and 2 h after a hypoxic-ischemic insult.Data are scatterplots paired from baseline to HI scan.*p < 0.05.Baseline scan (n = 5) and HI scan (n = 5).Paired t-test in four animals.HI, hypoxia ischemia.