SEARCH

SEARCH BY CITATION

Keywords:

  • epilepsy;
  • hippocampus;
  • ketogenic diet;
  • neuroprotection;
  • organotypic culture

Abstract

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

J. Neurochem. (2010) 113, 826–835.

Abstract

The ketogenic diet (KD), used successfully to treat a variety of epilepsy syndromes in humans and to attenuate seizures in different animal models, also provides powerful neuroprotection in various CNS injury models. Yet, a direct role for ketone bodies in limiting seizure and neuronal damage remains poorly understood. Using organotypic hippocampal slice cultures, we established an in vitro model of chronic ketosis for parallel studies of its neuroprotective and anti-convulsant effects. Chronic in vitro treatment with a ketone body, d-β-hydroxybutyrate, protected the cultures against chronic hypoglycemia, oxygen-glucose deprivation, and NMDA-induced excitotoxicity, but failed to suppress intrinsic and induced seizure-like activity, indicating improved neuroprotection is not directly translated into seizure control. However, chronic in vitro ketosis abolished hippocampal network hyperexcitability following a metabolic insult, hypoxia, demonstrating for the first time a direct link between metabolic resistance and better control of excessive, synchronous, abnormal electrical activity. These findings suggest that the KD and, possibly, exogenous ketone administration, can be more beneficial for the treatment of seizures associated with metabolic stress or underlying metabolic abnormalities, and can potentially be used to optimize clinical applications of the traditional KD or its variants.

Abbreviations used:
aCSF

artificial cerebro-spinal fluid

DβHB

d-β-hydroxybutyrate

KD

ketogenic diet

LG

low glucose

OGD

oxygen-glucose deprivation

PAD

primary afterdischarges

SLE

seizure-like event

The ketogenic diet (KD), a high fat, adequate protein, low carbohydrate diet, has been used for the treatment of difficult-to-control seizures in children since the 1920s (for reviews, see Keene 2006; Freeman et al. 2007; Hartman and Vining 2007), and its efficacy has been recently confirmed by long anticipated, randomized controlled clinical trials (Neal et al. 2008; Freeman et al. 2009). The KD was formulated to mimic fasting, with limited glucose supply and high fat availability favoring fatty acid oxidation, ketone body production, and their utilization by the brain as an alternative energy substrate. Despite a long history of clinical use and extensive animal research (reviewed in Bough and Rho 2007), the mechanisms underlying its anti-epileptic action remain poorly understood.

Most of clinical and experimental studies of the KD indicate that the establishment of ketosis by raising ketone bodies to some ‘threshold level’ is necessary for the anti-seizure effects. Moreover, ketone administration has been shown to be neuroprotective in various CNS injury animal models (for a review see Prins 2008), including in vitro models of Alzheimer’s (Kashiwaya et al. 2000) and Parkinson’s disease (Tieu et al. 2003) and in vitro glutamate toxicity (Massieu et al. 2003) and hypoxia (Masuda et al. 2005). Yet, a direct role for ketone bodies in limiting seizure activity has not been confirmed in in vitro seizure models (Wada et al. 1997; Thio et al. 2000; Donevan et al. 2003). The limiting factor of these studies was a relatively short treatment with the ketone bodies, whereas, in in vivo animal studies, seizure reduction was observed after several days and achieved maximally after several weeks (Appleton and DeVivo 1974; Rho et al. 1999; Bough et al. 2006), the time necessary for a metabolic switch from glucose to ketone bodies as the primary fuel source.

To achieve long-term in vitro treatment and observations, we used organotypic hippocampal slice cultures. Importantly, in contrast to dissociated neuronal cultures, organotypic hippocampal slices retain functional properties characteristic of the intact hippocampus, including neuronal network features such as seizure-like activity (McBain et al. 1989; Heinemann et al. 2006; Samoilova et al. 2008). Organotypic slice cultures also represent a feasible model for studies of the effects of metabolic insults on neuronal survival (for a review, see Noraberg et al. 2005) and allow for parallel studies of slice viability and function, and neuroprotective and anti-convulsant effects of chronic treatments.

Materials and methods

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Organotypic hippocampal slices were prepared from a 7-day-old Wistar rats (see Appendix S1 for details). All procedures were approved by the University Health Network Animal Resources Centre. At 5–6 days in vitro, the cultures (18–24 membrane units in each set) were divided into three groups. In the low glucose (LG) and the LG supplemented with d-β-hydroxybutyrate (DβHB) (LG + DβHB) group, the glucose concentration in the culture media was gradually (over 3 days) reduced by daily replacement of half of the normal culture media (containing 24 mmol/L glucose, the standard glucose concentration in culture media) with glucose-free media to achieve glucose concentration equal to 3 mmol/L. In the LG + DβHB group (in part mimicking the KD), glucose concentration lowering was accompanied by the addition of DβHB, to achieve a final clinically relevant concentration (10 mmol/L) in culture media. To achieve chronic treatment conditions, the slices were cultured under LG conditions with or without DβHB supplementation for at least 3 days before experimentation.

Cell viability was validated using propidium iodide fluorescence measurement. Neuronal network activity was recorded extracellularly in the pyramidal cell layers. The detailed procedures as well as in vitro seizure models and metabolic insult induction are described in Appendix S1.

Results

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Effects of chronic glucose deprivation and supplementation with DβHB on hippocampal slice viability and function

It has been previously demonstrated that 2–2.5 h exposure to glucose-free media resulted in selective propidium iodide staining of the CA1 area in organotypic hippocampal slices, with the CA3 area spared 24 h after the insult (Tasker et al. 1992; Newell et al. 1995), indicating selective vulnerability of CA1 neurons. In our experiments, while no marked cell death was observed in the media containing standard 24 mmol/L glucose (Fig. 1a), chronic hypoglycemia (reduced to 3 mmol/L but not removed glucose) resulted in marked cell death first detected in the CA1 region of the hippocampus after 2 days of LG exposure and later spreading to CA3 area and dentate gyrus (Fig. 1b). Supplementation of the medium with DβHB, added daily (chronic in vitro ketosis; LG + DβHB), prevented neuronal death (Fig. 1c) and supported the cultures for the period of observation (up to 7 days). In some (n = 7) experiments, DβHB supplementation was stopped. Two days after the last supplement of DβHB, a marked increase in neuronal death was observed (Fig. 1c) indicating that DβHB was utilized by neurons and protected them against chronic hypoglycemia.

image

Figure 1.  The effects of chronic glucose deprivation and supplementation with DβHB on hippocampal slice viability. Representative bright-field and gray-scale images of propidium iodide (PI) fluorescence were taken from (a) control slices cultured in the presence of 24 mmol/L glucose, (b) slices cultured under low (3 mmol/L) glucose conditions (low glucose, LG), and (c) slices cultured in low (3 mmol/L) glucose medium supplemented with 10 mmol/L DβHB (LG + DβHB; chronic in vitro ketosis) in four to six consecutive days; day 2 indicating the second day of the exposure to LG medium. The slices for each experimental condition were taken from the same culture set for uniform imaging. CA1 regions of the hippocampi that were found to be primarily sensitive to energy deprivation are delineated in the bright-field images. Scale bar: 1 mm. (d) The graph summarizes normalized PI fluorescence acquired from at least 42 slices for each experimental condition obtained from six different culture sets. For the period of observation, PI fluorescence did not significantly (p > 0.1) change in control slices, whereas chronic hypoglycemia (3 mmol/L glucose) resulted in significant (compared with the matched controls) time-dependent neuronal death which was prevented by DβHB supplement. PI fluorescence in LG + DβHB slices was not significantly (p > 0.1) different from that in control slices for 2–4 days. At day 5, some (significant compared to the matched control) damage was detected in LG + DβHB slices, however, being significantly (p < 0.001) lower than in the slices cultured in LG alone. ns, non-significant difference (p > 0.1); ***p < 0.001.

Download figure to PowerPoint

To evaluate synaptic network properties exhibited by slices undisturbed by perfusion, the activity was recorded from interface slices in CA1 and CA3 hippocampal areas simultaneously (Fig. 2). A single stimulus to the mossy fibers evoked a seizure-like event (SLE), a high amplitude (in average > 0.70 relative to the amplitude of the initial phase of the evoked response) long-lasting (tens of seconds) burst of spontaneous discharges, in most of control (12 of 15, 80%) (Fig. 2a) and LG + DβHB (8 of 9, 89%) (Fig. 2b) cultures. SLE duration was not significantly different and was 77.45 ± 14.86 s (n = 12) and 112.22 ± 24.24 s (n = 8) in control and LG + DβHB cultures, respectively. The amplitude of the initial phase of the responses did not significantly differ between the control (10.41 ± 1.16 mV in the CA1 region and 6.53 ± 0.73 mV in the CA3 region; Fig. 2, insert 1) and LG + DβHB cultures (7.70 ± 1.07 mV in the CA1 region and 9.39 ± 1.36 mV in the CA3 region; Fig. 2, insert 2). The amplitude of the spontaneous discharges during SLEs relative to the amplitude of the initial phase was not significantly different in control (0.72 ± 0.08 in the CA1 region and 0.77 ± 0.08 in the CA3 region) and LG + DβHB (0.86 ± 0.16 in the CA1 region and 0.68 ± 0.06 in the CA3 region) cultures. In the slices cultured in LG medium for at least 2 days, no evoked response in the CA1 region was observed, although the response in the CA3 region was preserved (Fig. 2, insert 3). As the cell death in CA1 region could affect slice properties in general and thus complicate data interpretation, further experiments were performed in control and LG + DβHB cultures.

image

Figure 2.  The effects of chronic glucose deprivation and supplementation with d-β-hydroxybutyrate (DβHB) on hippocampal slice function. Representative recordings of the evoked activity in the CA3 (upper traces) and CA1 (lower traces) regions of the hippocampi cultured in (a) the medium containing 24 mmol/L glucose (Control culture), (b) low (3 mmol/L) glucose medium supplemented with 10 mmol/L DβHB [low glucose (LG) + DβHB culture], and (c) in low (3 mmol/L) glucose. The activity was recorded under interface conditions. A single stimulus (indicated by arrows) to the mossy fibers evoked population spikes (inserts 1 and 2) followed by seizure-like events in 80% and 89% of control and LG + DβHB cultures, respectively. No evoked response (insert 3) was observed in the CA1 region of the hippocampus in the LG cultures, although the population spike in the CA3 region was preserved. The slices were cultured under LG conditions with or without DβHB supplementation for 3–4 days.

Download figure to PowerPoint

Induced seizure activity

To compare seizure susceptibility of control and LG + DβHB cultures, we used four in vitro models known to induce epileptiform synchronization in brain slices (for reviews, see Avoli et al. 2002, 2005; Bernard 2006; Heinemann et al. 2006), including bath application of drugs that reduce inhibition or enhance transmitter release or application of a medium containing lower (0.3 mmol/L) concentration of Mg2+ and stimulus train-evoked bursting (Fig. 3).

image

Figure 3.  Induced seizure activity. Representative recordings of the evoked (a–d) and spontaneous (e) activity in the CA1 region of the hippocampi cultured in [a(i)–e(i)] the medium containing 24 mmol/L glucose (control cultures) and [a(ii)–e(ii)] low (3 mmol/L) glucose medium supplemented with 10 mmol/L d-β-hydroxybutyrate (DβHB) [low glucose (LG) + DβHB]. The activity was recorded under submerged conditions. The activity evoked by a single stimulus (indicated by arrows) to Schaffer collaterals before (a, i and ii) and after (b, i and ii) 4-aminopyridine (50 μmol/L) application. The activity evoked by a train of stimuli (100 Hz, 2 s; indicated by arrows) to Schaffer collaterals in the slices perfused with artificial CSF containing 2 mmol/L Mg2+ (c, i and ii) and 0.3 mmol/L Mg2+ (d, i and ii). Recurrent spontaneous activity (e, i and ii) induced by bicuculline methiodide (10 μmol/L) application. The slices were cultured in LG medium supplemented with DβHB for at least 3 days.

Download figure to PowerPoint

To monitor the effects of drug application, the recordings were made from submerged slices in the CA1 area during perfusion. Under these conditions, the amplitudes of the population response evoked by single stimulus to the Schaffer collaterals were not significantly different in control (6.08 ± 0.44 mV, n = 38) and LG + DβHB (5.95 ± 0.69 mV, n = 18) cultures (Fig. 3a, inserts 1 and 2). No SLEs were recorded when the slices were perfused with artificial cerebro-spinal fluid (aCSF) containing no drugs (Fig. 3a), in contrast to the activity recorded from interface slices (Fig. 2). Reduced extracellular space volume under interface conditions may account for the increased excitability compared with submerged conditions (Schuchmann et al. 2002).

In the presence of 50 μmol/L 4-aminopyridine (potassium channel blocker), which enhances transmitter release at both glutamatergic and GABAergic terminals (Buckle and Haas 1982; Rutecki et al. 1987; Perreault and Avoli 1989, 1991), a single stimulus provoked long-lasting (hundred of seconds) epileptiform responses with multiple afterdischarges (Fig. 3b). The duration of the response varied from 104 to 535 s in different cultures and was 292 ± 70 s (n = 5) and 293 ± 90 s (n = 4) in control and LG + DβHB cultures, respectively. The maximal amplitude of the epileptiform discharges (Fig. 3b, inserts 4 and 6) relative to the amplitude of the initial phase of the evoked response (Fig. 3b, inserts 3 and 5) was 0.80 ± 0.12 (n = 5) and 0.52 ± 0.10 (n = 4) (the difference was not statistically significant, p = 0.13). The effects of 4-aminopyridine were developed very fast (3–5 min of application) and were reversible upon drug washout (data not shown).

A train of stimuli (100 Hz, 2 s) evoked seizure-like primary afterdischarges (PAD) lasting 21.4 ± 3.7 s (n = 37) in control and 33.5 ± 8.1 s (n = 18) in LG + DβHB cultures when perfused with aCSF containing 2 mmol/L Mg2+ (Fig. 3c). The amplitude of the spontaneous event during PAD was 2.92 ± 0.35 mV (n = 37) in control and 2.55 ± 0.41 mV (n = 18) in LG + DβHB cultures. The reduction of Mg2+ concentration in the aCSF from 2.0 to 0.3 mmol/L resulted in significant (p < 0.05) prolongation of the PAD both in control (4.3 ± 1.0 times in average, n = 8) and LG + DβHB (4.1 ± 1.9 times in average, n = 4) cultures (Fig. 3d). Low Mg2+-induced epileptiform activity has been demonstrated in rodent-combined hippocampus–entorhinal cortex slices in vitro (Walther et al. 1986; Jones and Heinemann 1988; Tancredi et al. 1988; Wilson et al. 1988). Under these experimental conditions, epileptiform synchronization may be caused by removal of the block of NMDA receptors caused by Mg2+ ions, loss of surface charge, or deterioration of inhibition (Traub et al. 1994; Whittington et al. 1995; Mangan and Kapur 2004). It has been previously reported that upon perfusion with Mg2+-free aCSF, organotypic hippocampal slice cultures develop spontaneous SLEs and tonic recurrent discharges (Kovacs et al. 1999). We did not observed spontaneous recurrent activity in low Mg2+ aCSF for the period of observation (20–30 min), possibly, because of the differences in culture conditions (serum-free medium in our experiments) or Mg2+ ion concentrations.

In agreement with previous in vitro studies (Schwartzkroin and Prince 1980; Gutnick et al. 1982; Jones and Lambert 1990), bath application of 10 μmol/L bicuculline methiodide (GABAA receptor antagonist) induced spontaneous activity: short (lasting 0.28–2.08 s) interictal-like events at frequency 0.09 ± 0.01/s (n = 5) and 0.08 ± 0.04/s (n = 3) in control and LG + DβHB cultures, respectively. The maximal amplitude of these events relative to the amplitude of the initial phase of the evoked response was 0.80 ± 0.09 (n = 5) and 0.66 ± 0.04 (n = 3) in control and LG + DβHB cultures, respectively (Fig. 3e).

Neuroprotective properties of the in vitro ketosis

In a good agreement with previously published data (Newell et al. 1990, 1995; Pringle et al. 1997), oxygen-glucose deprivation (OGD) in control cultures (Fig. 4a) induced marked neuronal death clearly detected 24 h after the insult, which increased with time. In LG + DβHB cultures (Fig. 4b), the same insult resulted in strongly attenuated neuronal death. The difference was statistically significant in all three culture set tested. The pooled data are shown in Fig. 4(c).

image

Figure 4.  Neuroprotective properties of chronic in vitro ketosis. Representative gray-scale images of propidium iodide (PI) fluorescence were taken from (a and d) control slices cultured in the presence of 24 mmol/L glucose and (b and e) slices cultured in low (3 mmol/L) glucose medium supplemented with 10 mmol/L DβHB [low glucose (LG) + DβHB] in three consecutive days, before, 24,  and 48 h after the insult. (a and b) Oxygen-glucose deprivation (OGD). (c) The graph summarizes normalized PI fluorescence acquired from 15 control slices (white bars) and 35 slices cultured LG + DβHB (black bars) obtained from three different culture sets. The slices were cultured in LG medium supplemented with DβHB for 3 days. Note that OGD induced significant time-dependent neuronal death in both control and LG + DβHB cultures; however, in LG + DβHB cultures, the same insult resulted in significantly attenuated damage. (d and e) Treatment with 1 mmol/L NMDA. (f) The graph summarizes normalized PI fluorescence acquired from 38 control slices (white bars) and 42 slices cultured LG + DβHB (black bars) obtained from six different culture sets. The slices were cultured in LG medium supplemented with DβHB for 5 days. Note that NMDA induced significant neuronal death in both control and LG + DβHB cultures; however, in LG + DβHB cultures, the same insult resulted in significantly attenuated damage despite some (significant compared to the matched controls) initial damage detected in these culture prior to NMDA treatment. CA1 regions of the hippocampi are delineated. Scale bar: 1 mm. ns, non-significant difference (p > 0.1); **p < 0.01; ***p < 0.001.

Download figure to PowerPoint

The cell death process induced by ischemia is complex and involves a variety of mechanisms that differ depending on the stage of the process (Lipton 1999). The first, the induction stage, includes several changes initiated by ischemia and reperfusion including over-activation of NMDA receptors, leading to massive Ca2+ influx and consequent free-radical formation. In organotypic hippocampal cultures, NMDA receptor-mediated neurotoxicity is an important component of injury when neurons are deprived of oxygen and glucose, as MK-801, an NMDA receptor channel blocker, is strongly neuroprotective (Newell et al. 1990, 1995; Pringle et al. 1997). In our experiment, pre-treatment with 10 μmol/L MK-801 provided complete protection of the cultures against the effects of OGD (data not shown).

To further analyze the neuroprotective features of chronic in vitro ketosis, the control and LG + DβHB cultures were treated with a neurotoxic NMDA dose (1 mmol/L) (Kristensen et al. 2001). NMDA severely damaged control cultures (Fig. 4d) but its effect was strongly reduced in LG + DβHB cultures (Fig. 4e). The difference was statistically significant in each culture set tested (n = 6). The pooled data are shown in Fig. 4(f).

Metabolic insult-induced hyperexcitability

Metabolic insults evoke complex alterations in energy dependent processes, involving neurotransmitter systems and a variety of ion channels (for a review, see Krnjevic 2008). During the initial phase of the insult, the neuronal activity is rapidly blocked because of adenosine-dependent reduction of glutamatergic transmission (Fowler 1989; Gribkoff and Bauman 1992; Katchman and Hershkowitz 1993; Doolette and Kerr 1995; Khazipov et al. 1995; Zhu and Krnjevic 1997) and neuronal hyperpolarisation resulting from an activation of potassium conductances (Hansen et al. 1982; Fujiwara et al. 1987; Leblond and Krnjevic 1989). This energy saving mode of network activity representing an intrinsic neuroprotective mechanism can be interrupted by transient recovery of the evoked glutamatergic responses during the insult (Sick et al. 1987; Fowler 1989; Zhu and Krnjevic 1999). Alternatively, neuronal hyperexcitability may appear during the period of reperfusion (Schiff and Somjen 1985; Doolette and Kerr 1995). Importantly, in immature tissue, increased excitability in response to metabolic insults (deprivation of both oxygen and glucose and deprivation of either glucose or oxygen alone) is often manifested by spontaneous seizure-like activity (Jensen and Wang 1996; Dzhala et al. 2000; Abdelmalik et al. 2007) that may promote the damage.

Using control cultures and electrophysiological recordings, we conducted a series of experiments to find an appropriate in vitro model of a metabolic insult capable of inducing hyperexcitability in the CA1 region of the cultured hippocampal slices. The slices were subjected to a metabolic insult while the population responses evoked by a single stimulus to the Schaffer collaterals and spontaneous network activity were recorded. When control slices were perfused with aCSF without glucose and saturated with 95% N2 + 5% CO2 to induce OGD, rapid decline in the amplitude of the evoked responses was observed in all five slices studied. The effect was irreversible upon reperfusion in four out five slices, and no signs of increased excitability (evoked and/or spontaneous) were recorded. Glucose deprivation (reduction from standard 10 mmol/L to 0 mmol/L) resulted in a gradual reduction of the peak amplitude of the evoked responses in the CA1 region. After 40–60 min in 0 mmol/L glucose, the amplitude was reduced by 45 ± 7% (n = 7) without manifestations of hyperexcitability.

When cultured hippocampal slices were deprived of oxygen (aCSF with 10 mmol/L and saturated with 95% N2 + 5% CO2), synaptic depression developed that was reversible upon reperfusion (Fig. 5a and c). Moreover, the amplitude of the recovered evoked responses in four of six slices exceeded that prior to the insult by 29 ± 8% before return to the control level (Fig. 5c), demonstrating an increased evoked excitability. In addition, during the initial period (2–4 min) of reperfusion, five of six control cultures transiently exhibited recurrent spontaneous activity (Fig. 5d) (event duration 1569.4 ± 684.79 ms; 2–10 events per min; amplitude relative to that of the initial phase of the evoked response: 0.60 ± 0.22, n = 5). The data indicate that O2 deprivation provides a reliable in vitro model for studying metabolic insult-induced hyperexcitability.

image

Figure 5.  Oxygen deprivation-induced hyperexcitability. Representative extracellular recordings of the synaptic responses evoked by single stimuli to Schaffer collaterals in the CA1 region of the slices from control (a, i–iii) and low glucose (LG) + d-β-hydroxybutyrate (DβHB) (b, i–iii) cultures before [a(i) and b(i)], during oxygen deprivation [a(ii) and b(ii)] and upon reperfusion [a(iii) and b(iii)]. (c) The graph summarized the data obtained from six control and six LG + DβHB slices. (d and e) Spontaneous activity recorded during first 2–3 min of the reperfusion. The slices were cultured in LG medium supplemented with DβHB for 3–4 days.

Download figure to PowerPoint

In LG + DβHB cultures (Fig. 5b, ii), synaptic failure during O2 deprivation was very similar to that in control cultures (Fig. 5a, ii), with the dynamics of the synaptic depression and the recovery upon reperfusion being alike in both control and LG + DβHB cultures (Fig. 5c). However, in contrast to control cultures, the amplitude of the recovered evoked responses never exceeded that prior to the insult. Moreover, during reperfusion, no spontaneous epileptiform activity was recorded from the LG + DβHB cultures (Fig. 5e) (n = 6).

Discussion

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

The neuroprotective potential of the KD and exogenous ketone administration is widely recognized (for reviews, see Guzmán and Blázquez 2004; Gasior et al. 2006; Acharya et al. 2008; Prins 2008; Maalouf et al. 2009). In particular, the KD or ketone body (DβHB and/or acetoacetate) exposure have been shown to protect neurons against metabolic insults (such as hypoglycemia, hypoxia or ischemia) and excitotoxic damage (i.e. induced by glutamate) both in vivo (Suzuki et al. 2001, 2002; Massieu et al. 2003; Noh et al. 2003; Yamada et al. 2005; Mejía-Toiber et al. 2006) and in vitro (Izumi et al. 1998; Massieu et al. 2003; Masuda et al. 2005; Bough et al. 2006; Noh et al. 2006; Maalouf et al. 2007). It should be noted that these previous in vitro studies utilized dissociated neuronal cultures. Using organotypic hippocampal slice cultures that retain much of the anatomy, synaptic circuitry, and neurotransmitter receptors as the intact hippocampus, we developed a novel model of chronic in vitro ketosis demonstrating that chronic DβHB exposure provides potent protection against neuronal damage induced by chronic hypoglycemia (exposure to LG concentrations, Fig. 1), in vitro ischemia (OGD, Fig. 4a–c), and NMDA receptor over-activation (Fig. 4d–f). The mechanisms of the neuroprotective effects of chronic DβHB exposure require further studies but inability of DβHB to directly affect hippocampal synaptic transmission (Wada et al. 1997; Izumi et al. 1998; Thio et al. 2000), NMDA receptor function (Donevan et al. 2003), or hippocampal synaptic network properties (present study) suggests the involvement of the neurotoxic cascade triggered by NMDA receptor activation rather than direct targeting of the NMDA receptor itself. In the ketotic brain, more efficient removal of glutamate through increased astrocytic metabolism (reviewed in Yudkoff et al. 2007) may result in attenuated glutamate release in response to metabolic stress, thus contributing to the neuroprotective features of the KD. Importantly, organotypic hippocampal slice cultures, enriched in astrocytes, may serve as a reliable in vitro model for future studies of the proposed role of astrocytes (Guzmán and Blázquez 2004; Meløet al. 2006) in the beneficial effects of the KD.

It has been proposed that a switch to an alternative metabolic fuel, ketone bodies, may improve mitochondrial function (Appleton and DeVivo 1974; DeVivo et al. 1978) and provide a more efficient (compare to glucose) source of energy for brain per unit oxygen (Veech et al. 2001). Recent studies demonstrated that, in the hippocampus, the KD increases glutathione peroxidase levels (Ziegler et al. 2003) and mitochondrial uncoupling protein activity (Sullivan et al. 2004). The KD (Sullivan et al. 2004) and exogenously applied ketone bodies (Noh et al. 2006; Kim et al. 2007; Maalouf et al. 2007; Haces et al. 2008) decrease reactive oxygen species production by brain mitochondria, thus preventing oxidative injury. Gene-expression analysis demonstrated increased expression of genes in mitochondrial metabolic pathways (Noh et al. 2004; Bough et al. 2006). In addition, an increase in the density of mitochondria in the hippocampus and alterations in the production of energy metabolites indicated the enhancement in brain energy reserve after the KD (Bough et al. 2006) that may account for the neuroprotective features of the KD.

It has been proposed (Bough et al. 2006) that improved ability to sustain the ATP levels after the KD may also stabilize the membrane potential in neurons and, thus, elevate the seizure threshold, explaining the anti-convulsant effects of the KD. The effect can be mediated by increased activity of ATP-sensitive or two-pore domain potassium channels (Vamecq et al. 2005; Ma et al. 2007) leading to neuronal membrane hyperpolarization and excitability depression. Another hypothetical mechanism (Masino and Geiger 2008) implicates depression of synaptic transmission through the activation of purinoreceptors by proposed increased levels of adenosine after the KD. Indeed, dysregulation of the adenosine system is implicated in epileptogenesis and cell therapies have been developed to locally augment adenosine in an approach to prevent seizures (for a review see Boison 2008). However, the contribution of the adenosine system to the anti-epileptic of the KD as well as other mechanisms remains to be elucidated.

The in vivo KD has been proven to effectively protect against seizures in a variety of animal models (for reviews see Stafstrom 1999; Bough and Rho 2007). Electrophysiological recordings from the hippocampal slices of the KD-fed rats revealed reduced excitability compared with control animals (Stafstrom et al. 1999), which is likely mediated through augmentation of inhibition (Bough et al. 2003). In the present study, the in vitro chronic ketosis (LG supplemented with DβHB) was not sufficient to attenuate epileptiform activity (Fig. 3) which depends on hippocampal network properties, including glutamatergic and GABAergic synaptic transmission, ephaptic mechanisms, and gap junctional communications (for reviews, see Perez Velazquez and Carlen 2000; Avoli et al. 2002; Jefferys 2003; Nakase and Naus 2004; Traub et al. 2004; Heinemann et al. 2006; Kohling and Avoli 2006; Isomura et al. 2008). It is possible that the choice of DβHB as primary ketone body in our experiments may limit the anti-convulsant efficacy of the in vitro ketosis. Indeed, exogenously administered acetoacetate and acetone, but not DβHB, have anti-convulsant properties in several animal models of seizures in vivo (Likhodii and Burnham 2002; Rho et al. 2002; Likhodii et al. 2003; Gasior et al. 2007). Taking into account metabolic pathways of ketone body formation and utilization to support brain function (DβHB is first oxidized to acetoacetate for convertion to acetyl CoA in the mitochondria) (Sokoloff 1973), in particular, the conversion between DβHB and acetoacetate, the ability of DβHB to maintain long-term hippocampal culture viability (Fig. 1c) indicates metabolic processing of DβHB with the production of acetoacetate and acetone under our experimental conditions. At the same time, the well known chemical instability of acetoacetate, because of spontaneous decarboxylation, especially at physiological temperatures, and acetone’s volatility, obviously, prevent their accumulation under in vitro conditions. The relative chemical stability of DβHB compared with that of acetoacetate and acetone was the main reason it was chosen as a primary ketone to model the KD in vitro. Although technically challenging, the studies of acute and chronic effects of acetoacetate and acetone are needed to confirm possible differences in anti-convulsant efficacy of different ketone bodies.

The data suggest that DβHB, although being a constituent of the KD, is not anti-convulsant despite the chronic exposure. The same exposure, however, protects organotypic hippocampal cultures from the metabolic and excitotoxic insults (Figs 1 and 4) and reduced post-hypoxic hyperexcitability (Fig. 5), indicating the efficiency of the treatment (sufficient concentration and exposure duration) and pointing toward the differences in the mechanisms of the two major effects of the KD, manifested clinically by seizure cessation and neuroprotection. However, it could be that the neuroprotective aspects of the KD, which is classically administered to cases of severe refractory epilepsy, which are associated with repeated and intense epileptiform activity causing significant metabolic stress, is what makes the KD anti-convulsant and possibly ‘anti-epileptogenic’ in these cases. Using the immature intact hippocampus, we have recently shown that pre-treatment with DβHB reduced the occurrence of seizures induced by acute hypoglycemia in vitro (Abdelmalik et al. 2007). In the present study, the in vitro ketosis abolished hyperexcitability exhibited by control slices during the recovery after oxygen deprivation (Fig. 5). Although being transient, this hyperexcitability may lead to the remodeling of the synaptic network and eventually, when persisted, to epileptogenesis (Epsztein et al. 2008), especially in the immature brain, which is more resistant to metabolic insults but generates seizure more readily than mature brain (Ben-Ari and Holmes 2006). The efficiency of DβHB in preventing this activity suggests that exogenous ketone administration, or treatment with ketone body precursors, such as 1,3-butanediol (Gueldry et al. 1990; Sims and Heward 1994; Veech 2004), or using a medium-chain triglyceride diet (Bach and Babayan 1982; Liu 2008; Neal et al. 2009), that provide simple and safe methods to induce elevated plasma levels of ketone bodies, may be preferable for the treatment of seizures related to the metabolic abnormalities or increased susceptibility to metabolic insults. These findings give research insight into the neuroprotective and anti-convulsant properties of the KD and can potentially be used to optimize clinical applications of the traditional KD or its variants.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

This work was supported by the Hospital for Sick Children Research Foundation, Epilepsy Canada, and Canadian Institutes of Health Research (CIHR). The authors declared no conflict of interest. We thank Y. Adamchik for organotypic slice preparation and F. Vidic for technical assistance. Imaging experiments were performed using the equipment at the Wright Imaging Facilities at Toronto Western Research Institute.

References

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Appendix S1 Supplementary Materials and methods.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer-reviewed and may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

FilenameFormatSizeDescription
JNC_6645_sm_SupportingInformation.doc53KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.