Introduction

Glucose is the major fuel for the brain in normally fed humans. However, during periods of fasting, ketones provide a significant contribution to cerebral energetics, up to 70 % in prolonged starvation [1]. Furthermore, during times of severe stress or cerebral injury, ketones become a significant source of energy. In fact, there is evidence to suggest preferential uptake of ketones over glucose by the injured brain [2]. Several studies in animals have confirmed the benefit of exogenous ketone supplementation in terms of contusion size, cerebral blood flow and levels of adenosine tri-phosphate in traumatic brain injury and stroke models [3–6]. There is also evidence that ketosis induced via diet modification is protective in a number of neurological states including epilepsy and Alzheimer’s disease [7–9].

As the blood brain barrier (BBB) is relatively impermeable to ketones, cerebral uptake is via monocarboxylate transporter (MCT) proteins and increases with prolonged fasting [10, 11]. Cerebral levels are therefore largely determined by plasma ketone concentration. Most studies of supplementation have examined a single bolus or a short term infusion of enteric or intravenous ketones on cerebrospinal fluid (CSF) and brain metabolism [12–15].

Previous attempts to use ketones have been complicated by problems of hyperosmolarity and acid–base disturbances [12, 14]. Acetoacetate and β-hydroxybutyrate (BHB) are strong anions and sodium salts of these when administered as a hypertonic IV infusion will tend to lower the strong ion difference of the plasma and may result in an initial metabolic acidosis, particularly if the rate of reduction of strong ion difference overwhelms the metabolic capacity of the liver to metabolise these anions. With time, metabolic clearance of these anions will result in an increase in plasma strong ion difference resulting in a metabolic alkalosis. For ketone supplementation to be a potentially therapeutic agent in patients with traumatic brain injury (TBI), a few challenges remain: achieving a formulation which can be administered as a prolonged infusion, which rapidly increases ketone levels in the brain, achieves a target hyperosmolarity and maintains neutral acid–base status.

Fig. 1
figure 1

Plasma beta-hydroxybutyrate profiles. BHB beta-hydroxybutyrate, BHB 40, 3 % NaCL with 40 mmol/l, BHB 120 3 % NaCL with 120 mmol/l. In both groups, change over time was significant: values at 6 h significantly > 1 h significantly > baseline for both plasma and CSF. There was a significant difference in plasma BHB concentrations between BHB 120 and BHB 40 at 1 and 6 h

Full size image

Hypertonic Saline (HTS) has been extensively studied in animal and human models of TBI and may be beneficial for increasing cerebral perfusion pressure and cerebral blood flow and decreasing intracranial pressure while maintaining hemodynamic stability [16–18]. As such, a solution containing both hypertonic saline and ketones would theoretically provide a number of advantages over current solutions. As no published data exist on the effect of prolonged infusions of hypertonic NaBHB, the purpose of this study is to investigate hypertonic sodium ketone solutions with the aim of assessing the metabolic effects in general and specifically, on the brain. We hope to produce a formulation of hypertonic NaBHB (sodium D-3-hydroxybutyrate) which will increase serum levels of BHB without major metabolic side effects.

Materials and methods

This study was approved by the animal ethics committee of the University of Queensland (PAH/029/09/UQ). Adult Sprague–Dawley rats weighing between 420 and 570 g were utilized in the study. All animals were housed at the Biological Research Facility of the University of Queensland and were maintained according to the Standard Operating Procedures for the humane care of animals by the facility. Animals were provided with free access to food and water up to the time of the experiment.

BHB solution

Animals were divided into three groups of five, one control and two study arms. The control group received an intravenous infusion of 3 % NaCl at 5 ml/kg/h. The animals in the two study arms were assigned to receive one of the two formulations of ketone solutions. The composition of the solutions is listed in Table 1.

Table 1 Formulations of intravenous solutions
Full size table

The rats in the control arm received 3 % NaCl only while the two intervention groups received a 3 % NaCl solution containing 40 and 120 mmol of BHB respectively at 5 ml/kg/h. The D-3-hydroxybutyrate acid was manufactured by Sigma-Aldrich Pty. Ltd. (Castle Hill, NSW, Australia). The infusions were maintained for 6 h. Bloods and CSF were sampled at baseline, 1 and 6 h. At the end of the 6 h period, the animals were euthanized (Table 2).

Table 2 Baseline data (mean ± SD)
Full size table

The solutions were prepared as follows. For the 40 mmol/l solution, 5.0436 g/L of D-3-hydroxybutyrate Acid Sodium Salt was combined with 27.70 g/l NaCl (474 mmol/l) to create a 3 % solution (514 mmol/l). Similarly, for the 120 mmol/l solution, the relative concentrations were 15.13 g/l D-3-hydroxybutyrate Acid Sodium Salt and 23.03 g/l NaCL. The appropriate amounts were weighted for the particular volume required. The solution was mixed thoroughly until dissolved and prepared the same day of the experiments.

Anaesthesia and vascular access

The animals were anaesthetized using a facemask and 4 % isoflurane in 100 % oxygen. The isoflurane was subsequently decreased to 0.5–1.5 % for maintenance. Intraperitoneal xylazine 3 mg/kg was also administered. Body temperature was maintained with an animal heating blanket. A 24 gauge BD Insyte IV catheter (Franklin Lakes, NJ, USA) was inserted into a tail vein for continuous drug administration. A midline incision was made in the neck and a single common carotid artery was exposed and separated from surrounding tissues. A 24 GA BD Insyte catheter was inserted into the artery to monitor blood pressure and to collect blood samples. In order to maintain adequate oxygenation and control ventilation, a tracheostomy was performed. The trachea was dissected out and a small incision made between the rings. A 14 gauge BD Insyte IV catheter was shortened and inserted into the trachea and secured. The animals were attached to a Harvard Small Animal Ventilator (Holliston, Massachusetts USA) using tidal volumes of between 1.5 and 2.0 ml per breath and rates from 80 to 120 bpm. Heart rate and oxygen saturations were monitored with the Nellcor N-65 pulse oximeter (Covidien, Boulder, CO, USA).

Specimen collection

Arterial blood specimens were taken at baseline, 1 h and prior to euthanasia. Tests included arterial blood gas with acid/base status, oxygenation, serum sodium and chloride and BHB levels. The CSF specimens were obtained by performing a puncture into the cisterna magna using a 27 GA VanishPoint insulin syringe (Little Elm, Texas USA) and aspirating approximately 20 μl of CSF. The CSF specimens were tested for BHB levels. At completion of the study, the animals were euthanized using pentobarbital sodium at a dose of 15 mg/kg. Following euthanasia, the animals were decapitated and the brains were removed immediately, and stored at −80 degrees Celcius for later analysis.

Assays

Beta-hydroxybutyrate

Serum and CSF specimens were measured using the Stanbio Laboratory β-Hydroxybutyrate LiquiColor Assay Kit (Boerne, Texas, USA).

Brain BHB

Brain tissue was homogenised into ice-cold saline and the clarified extracts were appropriately diluted for use in the BHB assay kit. Brain BHB levels were measured using the Cayman Chemical—β-Hydroxybutyrate (Ketone Body) Assay Kit (Ann Arbor, Michigan, USA). Briefly, the method for BHB determination is based upon the oxidation of D-3-Hydroxybutyrate to acetoacetate by the enzyme 3-hydroxybutyrate dehydrogenase. Concomitant with this oxidation, the cofactor NAD+ is reduced to NADH. In the presence of diaphorase, NADH reacts with the colorimetric detector WST-1 to produce a formazan dye with an absorbance maximum at 445–455 nm. The absorbance of the dye is directly proportional to the BHB concentration. The range for the assay is 0–0.5 mmol/l. All samples were diluted to fall within this range.

Acid–Base status and sodium

Arterial blood gases and electrolytes were measured using the EC8+ cartridge with the Abbott Point of Care system, iSTAT (Abbott Park, Il, USA). The device was calibrated daily as per manufacturer's instructions.

Statistical analysis

Analysis of variance was used for analysis of BHB for blood, CSF and brain, as well as base excess (BE) and sodium (Na). Factors assessed were group (control, BHB 40, and BHB 120), time (baseline, 1 h, and 6 h) as well as their interaction for all variables except the brain which was only measured at the end; hence, only the group factor was included. A statistically significant interaction implies that changes over time differed between groups. Tukey’s method was used for multiple comparisons of significant (p < 0.05) factors. There was an extreme outlying observation for BHB for blood and CSF that was removed prior to ANOVA analysis. Spearman’s correlation was used to assess the correlation between BHB for blood and CSF at the three time-points as well as brain at the final time-point. The SAS version 9.1 for Windows and Stata Version 10.1 for Windows were used for the analysis.

Results

Plasma BHB

Blood levels of BHB demonstrated strong evidence of a change over time (p < 0.0001) with least squares means (standard error) at baseline: 0.19 mmol/l (0.03), 1 h: 0.29 mmol/l (0.03) and end: 0.47 mmol/l (0.03) (Fig. 1). Multiple comparisons showed all these means were statistically different (p < 0.05). There was also strong evidence of a difference between groups (p < 0.0001) where the group least squares means (standard errors) were control: 0.21 mmol/l (0.03), BHB 40: 0.26 mmol/l (0.03), and BHB 120: 0.48 mmol/l (0.03). The BHB 120 mean was statistically larger than that for BHB 40 (p < 0.05) and control (p < 0.05) but there was no evidence of a difference between control and BHB 40. There was no evidence of an interaction between group and time (p = 0.071) (Table 3).

Table 3 BHB data for 3 time points (mean ± SD)
Full size table

CSF BHB

The BHB levels in CSF showed strong evidence of a change over time (p < 0.0001) with least squares means (standard error) at baseline: 0.15 mmol/l (0.02), 1 h: 0.23 mmol/l (0.02) and end: 0.36 mmol/l (0.02). Multiple comparisons showed all these means were statistically different (p < 0.05). There was no difference between groups (p = 0.46) however there was evidence of an interaction between group and time (p = 0.009) where there was an increase in BHB CSF over time in the BHB 40 and 120 groups but not the control group. To investigate the change in BHB CSF over time, separate analyses were conducted on the three groups. Results showed evidence of change over time for BHB 40 group (p = 0.001) and for the BHB 120 group (p = 0.0005) but not for the control group (p = 0.21) (see table 3).

Brain BHB

For BHB levels there was strong evidence of a difference between groups (p < 0.0001) where the group least squares means (standard errors) were control: 0.15 mmol/l (0.01), BHB 40: 0.19 mmol/l (0.01), and BHB 120: 0.28 mmol/l (0.01). Multiple comparisons showed all these means were statistically different (p < 0.05).

There was no differences over time (p = 0.31) or between groups (p = 0.33) or an interaction between groups and time (p = 0.47) for base excess (see Table 4). An analysis of just the final time-point also showed no difference between groups for base excess (p = 0.24). For Na there was evidence of a difference over time (p < 0.0001) but not between groups (p = 0.24) or interaction between groups and time (p = 0.15). Least squares means (standard error) were, at baseline: 137 mmol/l (1.5), 1 h: 143 mmol/l (1.5) and end: 154 mmol/l (1.5). Multiple comparisons showed all these means were statistically different (p < 0.05).

Table 4 Acid base and electrolyte data for 3 time points (mean ± SD)
Full size table

In a final analysis there was evidence of a correlation between BHB blood and CSF at the end (correlation = 0.81, p < 0.0001) but not at baseline (correlation = 0.0, p = 0.99) or 1 h (correlation = 0.31, p = 0.26). There was evidence of a correlation between BHB blood and brain at the end (correlation = 0.67, p = 0.012) as well as between BHB CSF and the brain at the end (correlation = 0.62, p = 0.023).

Discussion

Under normal conditions, glucose is the main source of energy for the brain. However, hyperglycemia has repeatedly been shown to be deleterious to the injured or ischemic brain [19]. It therefore becomes a clinical challenge to provide sufficient substrate without utilisation of glucose containing solutions. It has been clearly shown that ketones can provide a significant proportion of the cerebral energy requirements and may be preferred to glucose following acute brain injury [6, 15, 20]. There are however a number of questions as to the correct formulation and metabolic effects of intravenous BHB administration. We were able to demonstrate that a continuous infusion of BHB in nonfasted rats led to steadily increasing plasma and CSF concentrations of BHB. Despite this, there were no significant effects on acid–base metabolism at the concentrations used. Furthermore, the solution achieved a desirable level of hypernatremia necessary for managing intracranial hypertension.

Comparison with previously published data

Prior studies have only examined the effects of short term BHB infusions on BHB levels. Suzuki et al. [6] used infusions of BHB to examine effects in acute brain injury, but blood levels were not reported. Katayam et al. [21] administered an isotonic solution of BHB at 30 umol/kg/min for 2 h in a hemorrhagic model. This led to BHB levels of approx. 1.5 mmol. However, the treatment animals demonstrated a significant alkalosis with final pH 7.54. Prins et al. [15], while examining the cerebral uptake of BHB in rat model of TBI, administered 30 mg/kg/h for 3 h. There was a modest increase in BHB levels, the peak of which reached 0.41 mmol/l with no alteration in pH. As part of a study examining effects of BHB on cerebral blood flow and cerebral oxygen delivery, Linde et al. [14] examined the effect of administering 37.5 mg/kg/min to rats over a period of 2 h. They demonstrated a rapid increase in pH to 7.52 after 2 h which loosely correlated with the increase in serum BHB levels to 6.08 mmol/l.

Some authors have suggested that high plasma levels of BHB are necessary to produce a beneficial effect on cerebral energetics. As such, many studies have administered large concentrations of BHB over short periods of time to rapidly increase blood BHB [12]. This is not necessarily true as relatively small increases in BHB may be sufficient to improve cerebral metabolism. Although a correlation exists between blood BHB level and cerebral uptake there is period of adaptation as the MCT transfer molecules are upregulated over time [22, 23] Therefore, cerebral uptake of BHB is not only dependent on plasma level but also increases following prolonged hyperketonemia [12, 13]. This was demonstrated by Linde et al. [14] following an acute increase in peripheral BHB levels to greater then 6 mmol/l. They observed no decrease in glucose metabolism and no substitution of glucose by ketone bodies for oxidate metabolism, suggesting adaptation had not yet occurred. Pan et al. [13] demonstrated a small, but statistically significant, increase in cerebral ketones (0.24 mmol/l ± 0.04) following a 75 min infusion of BHB, but this was considerably lower in comparison with previous data acquired from fasted adults. These findings suggest the way the brain handles BHB is different following acute as compared to chronic hyperketonemia and likely represents transport upregulation.

We demonstrated that ketones could be readily measured in CSF and that there was a progressive increase in CSF concentration. We also found an increase in CSF levels related to the concentration of solution administered. This is despite the fact that the rats were not starved before the study. Furthermore, the effect was likely related to BHB infusion as CSF levels did not change significantly in the control group. Similarly, there was a significant increase in brain tissue BHB levels which depended on the amount of BHB administered. Pan et al. [13], using H + magnetic resonance spectroscopy, were able to show an increase in occipital lobe concentrations of BHB in healthy human subjects following a 75 min infusion of BHB. Although Pan reached blood levels of 2.12 mmol/l the cerebral BHB level was only 0.24 mmol/l whereas we achieved a significant increase by administering a lower concentration of BHB over a prolonged period. Furthermore, there was no comment on the effects of these levels on pH. Pan previously demonstrated the effects of prolonged ketone exposure on cerebral levels with significant increases in cerebral ketones after 48 and 72 h starvation [12].

Several prior studies have demonstrated the development of a significant metabolic alkalosis following BHB infusion [24]. Blood pH reached a maximum of 7.6 following BHB infusion during the study by Linde et al. Concentration, rate of administration and length of administration are likely to be important factors in determining acid/base response. Metabolic alkalosis has several potential deleterious effects on cerebral metabolism [25–27]. These include a decrease in CBF, lethargy and confusion and if severe, coma and seizures. More specifically, Giffard et al. [28] found an alkaline pH exacerbated glutamate induce excitotoxic neuronal damage and appeared to sensitize neurones to ischemic injury and potentiate reperfusion injury. Alkalosis has been shown to decrease CBF in a number of animal studies by both Arvidsson et al. and Pannier et al. [25, 27]. Furthermore, Schrock et al. [26] demonstrated a decrease in CBF and glucose utilization during severe hypochloremic alkalosis.

Bourdeaux et al. [29] on the other hand, found infusions of 8.4 % sodium bicarbonate effective at reducing intracranial pressure although whether the hypertonic nature of the solution or alkalinizing effects were responsible is not clear. We found that concentrations of BHB up to 120 mmol/l were well tolerated with no significant change in base excess over time. Others have found NaBHB to be alkalinizing at high doses given over short periods [14]. This is likely due to the insufficient time for acid/base homeostasis to occur. The clinical effects of this have not been investigated.

As one of the potential uses of this solution includes the management of intracranial hypertension in patients with traumatic brain injury, a key outcome was to demonstrate the solution to be capable of producing hyperosmolar conditions. All solutions were able to achieve this and the concentration of BHB did not alter the outcome. Hyperosmotic therapy has been the corner stone of the management of intracranial hypertension for a number of years. In many units, hypertonic saline has replaced mannitol as the principle solution [30]. The HTS has a number of potential benefits including a high coefficient of reflection, maintenance of BBB integrity, modulation of inflammatory response and augmentation of volume resuscitation. Furthermore, a number of animal and human studies have confirmed it as an effective treatment for intracranial hypertension [31, 32]. A combined solution of hypertonic saline and BHB could provide the benefits of hyperosmolar therapy with improved cerebral energetics [33].

Limitations

There are several limitations of our study. Although the infusion lasted 6 h, longer than most prior studies, this is still relatively short. The effects of longer duration infusions are unknown although a number of studies have examined prolonged hyperketonemia induced by either diet or starvation. Secondly, we did not measure clinical endpoints such as intracranial pressure, cerebral perfusion pressure or CBF. Nor did we look at cerebral metabolic neuroprotective endpoints such as reductions in reactive oxygen species and the mitochondrial permeability transition complex although prior studies have demonstrated improved brain energetics from ketones as compared to glucose. And thirdly, we used healthy rats. The pharmacokinetic and pharmacodynamic effects cannot necessarily be extrapolated to head injury models as the blood brain barrier permeability and cerebral metabolism cannot be predicted. Experiments in acute brain injury models appear to improve the brain's ability to utilize ketones leading to decreased contusion size.

Conclusion

We have demonstrated that IV infusions of hypertonic saline/BHB are possible and lead to increased plasma and CSF BHB levels in healthy rats and that increases in brain levels of BHB are dependent on plasma concentrations. Solutions containing 40 and 120 mmol/l of hypertonic NaBHB are not alkalinizing and are capable of producing adequate hypernatremia. This study only provides a proof of principle and will need to be replicated in animal models of brain injury with concomitant measurement of cerebral hemodynamics and appropriate biochemical assays to demonstrate their utility as a neuroprotective agent.