The results published within this manuscript are part of a larger series of experiments that tested the effects of alternative energy substrates in the context of CA (protocol number 704N), and data from the control group have therefore been reported previously [22], in accordance with the 3Rs principle (Replacement, Reduction and Refinement) to reduce the total number of animals included. All experimental procedures were approved by the Institutional Review Board for Animal Care of the Free University of Brussels (Belgium) (number of Ethical Committee approval: 704 N) and followed ARRIVE (Animal Research: Reporting in Vivo Experiments, see Supplementary material online, Table S1) guidelines. Care and handling of the animals were in accord with National Institutes of Health guidelines (Institute of Laboratory Animal Resources).

A detailed description of the model has been published previously [21, 22]. On the day of the experiment, the selected adult swine was sedated in the cage with a combined intramuscular injection of midazolam (1 mg/kg) and ketamine (10 mg/kg), in the neck and placed on the operating table. After obtaining peripheral venous access, a continuous infusion of sufentanyl citrate (0.2–1 µg/kg/h) was started, and a catheter was inserted in the femoral artery for invasive arterial pressure monitoring, orotracheal intubation was performed and ventilation started in controlled volume mode with a tidal volume of 8 ml/kg, 5 cmH2O of positive end-expiratory pressure (PEEP), fraction of inspired oxygen (FiO2) of 1, and inspiratory to expiratory ratio of 2, with a square wave flow pattern. A 1% mixture of inspired sevoflurane was started, and ventilation parameters were progressively adjusted to ensure a PaCO2 between 35 and 45 mmHg and a PaO2 > 70 mmHg, with the minimally required FiO2. A continuous infusion of rocuronium (1- 4 mg/kg/h) and sufentanyl citrate (3.5 µg/kg/h) was maintained for the entire experiment and 2 g of amoxicillin-clavulanate (Sandoz, Basel, Switzerland) administered as slow intravenous bolus. A Foley catheter was then surgically placed in the bladder, followed by a three-lumen central venous catheter in the external jugular vein, a pulmonary artery catheter, and an upward internal jugular single-lumen catheter to sample venous cerebral outflow. The animal was then proned and multimodal neuromonitoring placed. Neuromonitoring consisted of two cerebral microdialysis (CMD) catheters, a multifunctional probe measuring intracranial pressure (ICP), cerebral temperature, and brain tissue oxygen tension (PbtO2), a laser Doppler probe, and two stereoelectroencephalography (sEEG) wires (one for each parietal lobe).

The animal was returned to the supine position and ventricular fibrillation induced electrically using a pacemaker wire. No treatment was given for 10 min, and cardiopulmonary resuscitation (CPR) was then started, using an automated chest compressor at the rate of 100/minute, and continued for 5 min; ventilation was resumed. After one minute of CPR, an intravenous dose of epinephrine was administered, and at the end of the 5 min CPR period, a biphasic electric countershock was delivered. If there was no return of spontaneous circulation (ROSC), an additional minute of CPR was given followed by an electric countershock, and an additional dose of epinephrine was given after 7 min of CPR. ROSC was defined as the presence of a cardiac rhythm along with a mean arterial pressure (MAP) > 65 mmHg for more than 20 min.

After ROSC, the animal was proned and observed for 12 h. All animals were treated with targeted temperature management at 34 °C during the observation phase, at the end of which the animals were sacrificed with potassium chloride injection and brain tissue samples harvested. Death was confirmed by ventricular fibrillation waves on the EKG trace and a concomitant decrease in arterial blood pressure.

On the day of the experiment, animals were randomly assigned (simple randomization, following a predetermined schedule) to receive either a bolus of 20 mL of NaCl 0.9% at the start of CPR followed by a continuous infusion of 0.18 g/Kg/h of sodium DL-3-ß-hydroxybutyrate (SBHB, GoldBio, St Louis, MO, US) during the observation period (Ketone Body group), or a bolus of 20 mL of NaCl 0.9% at CPR initiation followed by a continuous infusion of balanced crystalloids during the observation period (Control group) (Fig. 1). The SBHB solution had an expected osmolarity of 1189 mOsm/l and a concentration of 0.075 g/ml. Because of the expected associated changes in plasma sodium concentrations, blinding was not feasible. In the KB animals, specific safety limits were pre-established to standardize the administration of SBHB and reproduce conditions translatable to clinical practice: every hour the infusion rate was reduced by 20% if the arterial Na+ exceeded 155 mEq/l or plasma osmolarity was ≥ 320 mOsm/l.

Timeline of the experiment. The Control group is represented in mauve, and the Ketone Body group in green. VF: ventricular fibrillation; CPR: cardiopulmonary resuscitation; ROSC: return of spontaneous circulation; T0-3: blood sampling timepoints

Throughout the experiment, a MAP target of > 65 mmHg was used in both groups and achieved with norepinephrine administration whenever necessary. All animals received a continuous infusion of balanced crystalloids (5–10 ml/Kg/h), and any additional fluid administration was titrated to keep the pulse pressure variation (PPV) < 14%, without surpassing the baseline pulmonary capillary wedge pressure. Hyperglycemia was left uncorrected if blood glucose concentration was < 300 mg/dL within the first hour after ROSC or < 250 mg/dL later during the study period and otherwise corrected with an infusion of 10 Units of insulin (Actrapid, Novo Nordisk, Bagsværd, Denmark).

Arterial blood samples were obtained prior to CA induction (T0), 20 min after ROSC (T1), and 6 h (T2) and 12 (T3) hours later. Jugular vein blood gas analyses were carried out before and after endotracheal intubation, at T0 and T1, and then hourly. The arterial-jugular venous differences in glucose and lactate at each time point were used as proxies for cerebral uptake of glucose and lactate, respectively. Central venous mixed blood was collected at T0, T1, and then every three hours to allow instrument calibration. CMD samples were taken from the same catheter at T0 and then every hour.

ß-hydroxybutyrate concentrations were measured in treated animals and in a representative group of controls (n = 4) using a point-of-care approach with a dedicated veterinary tool (Nova Vet, Nova Medical, Waltham, MA, US) at T0, T1, T2, and T3. Plasma levels of the brain injury biomarkers, glial fibrillary acid protein (GFAP, #102336), neurofilament light chain (NFL, #103400), and neuron specific enolase (NSE, #102475), were measured on an SR-X analyzer according to the manufacturer’s instructions (Quanterix, Billerica, MA). A single batch of reagents was used for each analyte.

After animal sacrifice, two cortical sections of about 0.5 cm3 of brain parenchyma were harvested from each parietal lobe and immediately frozen in liquid nitrogen. Gene expression analysis was subsequently carried out on a selected number of genes representative of different injury-response related pathways (i.e., inflammation, structural integrity, apoptosis, endothelial function, and oxidative stress) [21, 22].

Offline analysis of the filtered EEG signal (1–15 Hz; 4th order Butterworth bandpass filter, filtfilt function in Matlab) enabled extraction of the signal amplitude (Hilbert function in Matlab). All computing was carried out using a sliding 1 min window with 50% overlap. Two independent neurophysiologist experts in reading EEGs who were blinded to the group assignment assessed the presence of suppressed background or burst suppression at T3 [21, 22].

Given the absence of similar protocols in the literature, the relatively small sample size in other protocols investigating the impact of ketosis in CA, and an expected survival rate of > 90% [21], an a priori sample size of 8 animals in the KB group was considered adequate for study purposes. Most recorded variables (i.e., blood pressure, heart rate, ICP, PbtO2, brain temperature and cerebral blood flow [CBF]) were recorded continuously with a sampling frequency between 1 and 100 Hz. Data were extracted as means over periods of 60 s and successively reduced to means over 10 min.

Continuous variables are expressed as means with standard deviation or medians with interquartile range, and discrete variables as percentages with 95% confidence intervals. Categorical variables were compared using Fisher's exact test or a Chi-square test, as appropriate. For multiple group comparisons, ANOVA, Welch's, or Mann–Whitney tests were used as appropriate. For the evolution of continuous variables over time, a linear mixed-effect model fitted for restricted maximum likelihood estimation (REML) was used. EEG data were analyzed as aggregates over the observation period; outliers were detected and eliminated using the ROUT method with a Q = 1% (where Q is the maximum desired false discovery rate). For variation in gene expression compared to the Control group, a one sample t-test or Wilcoxon test was used, as appropriate. A value of p < 0.05 was considered statistically significant.

Data analyses were performed using GraphPad Prism (version 10.1.1 for Macintosh, GraphPad Software, La Jolla, CA, US) and Matlab (version 9.7, R2019b update 9, The MathWorks Inc., Natick, MA, US).