Experimental protocol

Sixteen neonatal female Yorkshire swine (3.8 to 5.8 kg; post-natal age: 6 to 10 days of life) were acclimated to their surroundings and bottle fed 5 times daily by research staff for 6 days. On the day of experimentation, swine were anesthetized by intramuscular injections of tiletamine (Telazol), xylazine and atropine, and tracheally intubated. Swine were then sedated by using inhaled isoflurane (0.25% to 2%). Neuromuscular blockade (cisatracurium) was administered upon anesthetic induction and then again prior to sternotomy incision. A right femoral arterial (3-F sheath) catheter and a right internal jugular venous (4-F, 5 cm) catheter were placed and continuously transduced. Esophageal and rectal temperature probes were placed. A median sternotomy was performed, a subtotal thymectomy performed, and the pericardium opened. A sterile human infant-sized CPB circuit (S5 infant perfusion pack, Sorin Group, Arvada, Colorado) was primed with blood from an adult donor swine sacrificed on the previous day. The right atrium and ascending aorta were cannulated (18-F DLP malleable single-stage venous and 10-F arterial cannulas, Medtronic-Biomedicus, Eden Prairie, Minnesota), and full-flow CPB was instituted. As is our institution’s clinical practice, a dose of methylprednisolone (30 mg/kg intravenous [IV]) was administered upon initiation of CPB. Swine were then cooled to 25°C (measured rectally) over 30 min, using a pH-stat management strategy (carbon dioxide added). Following cooling, the aorta was cross-clamped, and a solution of cold blood, potassium, magnesium, and lidocaine were administered into the aortic root, and cardioplegia was induced (del Nido Cardioplegia Solution [24]), causing prompt diastolic arrest. Circulation was then discontinued for a 75-min period of circulatory arrest. Rectal temperature was maintained as close to 25°C as possible by using surface cooling as needed. Following circulatory arrest, circulation was restored, and swine were warmed to 37°C over 60 min. Swine were then weaned from CPB using inotropic support as needed to maintain systolic blood pressure >80 mm Hg. Cannulas were removed, hemostasis ensured, and the sternum closed.

Survival period

Sedation was then transitioned to infusions of propofol (1 to 3 mg/kg/h) and fentanyl (1 to 2 μg/kg/h), and inhalational isoflurane was discontinued for an 18-h period of regimented intensive care staffed by intensive care nursing staff. During this time, esophageal temperature was continuously monitored and maintained below 38°C by using a cooling blanket. Blood pressure was maintained with an infusion of dopamine (3 to 5 μg/kg/min, titrated to systolic blood pressure >70 mm Hg). Diuresis was achieved by using furosemide (1 mg/kg every 12 h). Mechanical ventilation was continued by using synchronized, intermittent mandatory ventilation with a fraction of inspired oxygen, required to maintain pulse oximetry saturation of 95% and target tidal volumes of 8 ml/kg. Animals were continuously monitored for clinical seizure activity. Seizures lasting longer than 2 min were treated according to a protocol of lorazepam (0.1 mg/kg IV every 5 min up to 3 doses), then phenobarbital (20 mg/kg IV every 20 min for 2 doses), then fosphenytoin (20 mg/kg IV every 20 min for 2 doses), then an increase in the rate of propofol infusion (up to 10 mg/kg/h). Inotrope score was calculated as: [dopamine (μg/kg/min) + dobutamine (μg/kg/min) + 100× epinephrine (μg/kg/min)] (25).

After the animals underwent 18 h of intensive care, the arterial catheter, thoracic drain, and tracheal tube were removed. Swine were then observed for 3 days, with quantification of neurologic status by daily neurologic examinations. Blood drawn prior to and daily after the injury was analyzed for complete blood count, chemistry profile, hepatic transaminases, and venous blood gas analysis. Glial fibrillatory acidic protein (GFAP) was assessed using an electrochemiluminescent sandwich immunoassay (Meso Scale Diagnostics, Rockville, Maryland), with a detection range of 0.001 to 40.0 ng/ml (26).

Inhaled hydrogen therapy

Animals were randomized to treatments as described above with or without inhaled hydrogen (2.40%) for a 24-h period during and after the ischemic injury (n = 8 per group). At the beginning of the study, we created a table that dictated the treatment allocation for each experiment in random order, according to which patients were treated. Due to environmental hazards and logistical considerations, members of the veterinary, perfusion, and overnight nursing staff were not blinded to treatment group allocation, whereas surgical staff, neurologists, and histopathologists were blinded to treatment allocation. Premixed, certified hydrogen gas blends containing 2.40 ± 0.05% of grade-6 purity (99.9999%) hydrogen gas with either balance medical air or medical oxygen (Praxair Distribution, Inc., Jessup, Maryland) were obtained and received as nonflammable gas mixtures. These tanks were fitted with a 50-psi regulator and a flash arrestor and then attached directly to the air and oxygen (respectively) inlets of the anesthesia machine (Dräger Apollo, Coppell, Texas) during the experimental period and to the mechanical ventilator (Servo I model, Maquet, Gothenburg, Sweden) during the survival period (Supplemental Figure S1). Additional hydrogenated “carbogen” mixtures were made, including 0%, 4%, 6%, and 8% carbon dioxide, 2.40% hydrogen, and balance oxygen for use during hypothermic perfusion. Ambient hydrogen concentrations were measured continuously (Eagle 2 model, RKI Instruments, Union City, California).

Swine neurologic deficit scores

Swine were evaluated prior to and following each 24-h period after the injury using a previously described swine neurologic deficit score (SNDS) (27) by 2 research technicians, present for each examination (unblinded to treatment allocation), and 2 neurologists (by review of videotaped examinations, blinded to treatment allocation). The mean of the 4 scores was taken at each time point. The examination assessed cranial nerve function, respiratory pattern, motor and sensory function, level of consciousness, and behavior, each assigned a total of 100 points. Points were assigned based on specific abnormal neurologic findings (Supplemental Table 1), such that a score of 0 was normal, and a score of 500 represented brain death. The presence or absence of clinical seizures was not part of the scoring system.

Brain magnetic resonance imaging

On day 3 post-injury, swine were anesthetized for brain magnetic resonance imaging (MRI) (3-T Skyra model scanner, 64-channel head and neck coil, Siemens, Corp., Munich, Germany). High-resolution T1, T2, fluid-attenuated inversion recovery, and diffusion-weighted images were obtained. Areas of enhancement on axial and coronal T2 and apparent diffusion coefficient images were manually assessed on a voxel-per-voxel basis and outlined (itk-SNAP software application, Penn Image Computing and Science Laboratory, University of Pennsylvania, Philadelphia, Pennsylvania, and Scientific Computing and Imaging Institute, University of Utah, Salt Lake City, Utah) by a radiology technician (Mr. Abdelhakim Ouaalam, Department of Radiology, Boston Children’s Hospital, Boston, Massachusetts), a radiologist (E.Y.), and a clinical neurologist (J.N.), all of whom were blinded to treatment allocation. From these values, total volumes of cranial injuries per swine were calculated using in-house software normalized to brain volume, and the total volumes of injuries were compared between groups by using the Mann-Whitney U test.

Neurohistopathology

Following brain MRI of the swine on post-injury day 3, both carotid arteries and jugular veins were then cannulated by cutdown and perfused with normal saline (2 l), followed by 4% paraformaldehyde (4 l). The heads of swine were fixed in 10% formaldehyde for 24 h and then removed and paraffin embedded and then stained for hematoxylin and eosin. Hypoxic-ischemic injuries of the frontal cortex, temporal cortex, hippocampus, dentate gyrus, caudate nucleus, and thalamus were graded according to a previously defined scale by a pathologist (H.G.W.L.) blinded to treatment allocation. Briefly, histologic injury was evaluated by using both light and fluorescence microscopy (28) on a scale of 0 through 5 for each of 6 regions, as follows: 0 = normal, no injury; 1 = rare hypereosinophilic neurons; 2 = small clusters of hypereosinophilic neurons; 3 = majority of neurons (>50%) are hypereosinophilic; 4 = significant damage to neurons; and 5 = cavitated infarction with histologic necrosis. Animals that did not survive for 3 days (due to refractory status epilepticus) were assigned the median histologic score for animals in the control group.

Clinical outcomes

All animals were successfully weaned from CPB. The degrees to which hypothermia was achieved were similar between the groups (mean rectal temperature: 27.4 ± 1.8°C vs. 26.5 ± 1.9°C in controls and hydrogen-treated groups, respectively; p = 0.2387) (Supplemental Figure S2). Cerebral and somatic near infrared spectroscopy values were also similar between groups, frequently reaching a nadir of <20 during the deep hypothermic circulatory arrest period (Supplemental Figure S3). Two swine in the control group exhibited refractory status epilepticus and were sacrificed at 32 and 36 h post-injury following a failed trial of extubation; no hydrogen-treated animals exhibited seizures. Survival to 3 days was similar between groups (log rank test: p = 0.1435). Hydrogen-treated swine exhibited a higher rate of neurologically intact survival, defined as an SNDS of ≤120 at the time of death (log-rank test; p = 0.0035) (Figure 3A). SNDSs were significantly improved in hydrogen-treated swine in the postoperative period (p < 0.0001) (Figure 3B). Interobserver reliability was excellent among in-person scorers (i.e., unblinded research team members) and videotaped reviewers (i.e., blinded neurologists) (Pearson correlation coefficient: 0.895). The increase in serum GFAP concentrations relative to those at baseline was significantly higher in controls than in H₂-treated swine at 60 min post-injury (p = 0.0068) (Supplemental Figure S4).

Relative to controls, H₂-treated swine exhibited a significantly lower level of serum creatinine during the survival period (p = 0.0152), an average of 0.38 mg/dl lower by postoperative day 3. There were no differences in serum markers of hepatic injury or function (Supplemental Figure S5). There were no significant differences in acute hemodynamics, and inotrope scores were similar between the groups (Supplemental Figure S6). In the postoperative period, there were no differences in PaO₂/FiO₂ ratios as a marker of lung function (p = 0.92), nor were there significant differences in blood gas values during the postoperative period (Supplemental Figure S7). Similarly, there were no significant differences in hematologic endpoints between groups during the survival period (Supplemental Figure S8).

Neuroradiology

Swine in both groups exhibited a radiographic predominance of frontal and temporal lobe injuries. However, H₂-treated swine exhibited significantly lower volumes of white matter injury on T2 imaging than did controls (median: 134 mm³ [interquartile range [IQR]: 84 to 200 mm³] in H₂-treated swine vs. 383 mm³ [IQR: 77 to 639 mm³] in controls; p = 0.0460) (Figure 4).

The authors thank veterinary staff Cara Pimental, Madeleine Woomer, and Hugh Simonds; clinical perfusion staff Greg Matte, Kevin Connor, Natalie Toutenel, and Molly Bryant; overnight nursing staff Jay Hartford, Danielle Healey, and Stephanie Pietrafitta; clinical radiology staff Peter Morriss and Joseph Zmuda; and feeding volunteers Abigail Moore, Lindsay Thomson, Katherine Black, Andrew Lock, Jemima Lamothe, Yifeng Peng, Raymond Seekell, and Cameron Russell. The authors also thank Jie Zhu for performing GFAP assays; institutional safety officer Chad Pires, Boston Fire Department; and Centers for Disease Control and Prevention consultant Isaac Zlochower for assistance with the technical implementation of this work.