Since the KD was originated over 85 years ago, several major hypotheses have been advanced, but none have been widely accepted. Several key aspects of the KD might ultimately result in seizure protection. Ketone bodies, free fatty acids (in particular, polyunsaturated fatty acids), or glucose restriction might each lead directly or indirectly to seizure control. While it is possible that any one of these KD-induced changes is responsible for the anticonvulsant action of the KD, available evidence suggests that improved seizure control, at a minimum, likely requires all three.
Role of ketone bodies
Beta-hydroxybutyrate (BHB) is the predominant ketone body measured in the blood, and as such, has been used as a clinical measure of KD implementation (Fig. 1). Accordingly, nearly all KD studies have attempted to establish a causative link between ketonemia and anticonvulsant efficacy. Although robust elevations in plasma BHB levels have been observed during KD treatment (Bough et al., 1999b; Thavendiranathan et al., 2000), there is no significant correlation between plasma BHB levels and seizure protection. Optimal seizure protection generally lags days to weeks behind the development of ketonemia, which occurs within hours of KD onset.
Figure 1. Metabolic pathways highlighting the production of ketone bodies fatty acids during fasting or treatment with the ketogenic diet (KD). Estimated fasting- or KD-induced concentrations of beta-hydroxybutyrate, acetoacetate, and acetone in blood are listed (large boxes). Measures of beta-hydroxybutyrate levels in blood are most commonly used as the clinical indicator of successful KD treatment. From Likhodii and Burnham (2004).
Download figure to PowerPoint
Nevertheless, there is some evidence that ketones other than BHB may possess anticonvulsant properties. When injected into animals, acetone and its parent acetoacetate (ACA), prevent acutely provoked seizures. Seminal work in the 1930s revealed that acute intraperitoneal administration of acetone or ethyl-acetoacetate protected rabbits from thujone-induced seizures (Helmholz and Keith, 1930; Keith, 1933). Thujone is the active constituent of wormword oil, and is an antagonist of GABAA receptors (Hold et al., 2000). More recent experimental studies have shown similar results in rodents. Acetone (Likhodii et al., 2003) and ACA (Rho et al., 2002)—but not BHB—were anticonvulsant in a variety of acute and chronic models of epilepsy, consistent with earlier observations (Helmholz and Keith, 1930; Yamashita et al., 1976; Vodickova et al., 1995). Clinically, acetone levels of up to 1 millimolar (mM) were detected in the brains of five of seven well-controlled epileptic patients following KD using magnetic resonance spectroscopic techniques (Seymour et al., 1999). Although acetone could not be detected in two other seizure-free patients, the authors concluded that acetone contributes to the anticonvulsant effect of the KD. Interestingly, the concept that a lipophilic solvent may potently block seizure activity is not new. The classic example of this is valproic acid, which was initially used as a solvent to dissolve investigational anticonvulsant compounds, but was serendipitously discovered to possess intrinsic anticonvulsant properties.
Whereas in vivo pharmacodynamic studies have suggested that both ACA and acetone may act as anticonvulsant agents, there is no evidence that ketone bodies can directly modulate synaptic transmission and/or neuronal excitability. In vitro cellular electrophysiological experiments have failed to demonstrate an effect on the principal ion channels that mediate neuronal excitability and inhibition. Specifically, neither L-BHB nor ACA were found to modulate GABAA receptors, AMPA receptors, or NMDA receptors in both hippocampus and neocortex of rats (Thio et al., 2000; Donevan et al., 2003). Despite these negative observations, it remained possible that ketone bodies might affect network activity or synchrony. However, in field potential recordings conducted in vitro, Thio et al. (2000) demonstrated clearly that neither ACA nor BHB modified evoked excitatory postsynaptic potentials (EPSPs) or population spikes in the CA1 subfield of the hippocampal tissue. In summary, there is no evidence for direct anticonvulsant effects for either ACA or BHB, and acetone has yet to be studied in neuronal (CNS) tissue. This may, in large part, reflect the technical difficulties in investigating a compound that is highly volatile and can react with perfusion systems ordinarily used in pharmacological in vitro experiments.
Recently, it has been suggested that ACA and/or its metabolic byproduct, acetone, may activate a novel class of potassium leak channels known as the two-pore domain or K2P channels (Vamecq et al., 2005). K2P channels represent a diverse superfamily of channels that generally hyperpolarize cell membranes, and regulate membrane excitability both pre- and postsynaptically (Lesage, 2003). These channels can be modulated by changes in pH, osmolality, temperature, mechanical pressure, and certain fatty acids (Franks and Honore, 2004). Links between KD-induced elevations in ketone bodies (and/or fatty acids, as discussed below) and K2P channels, however, have yet to be explored.
In conclusion, although ketone bodies have been shown to possess anticonvulsant properties in vivo, there is no evidence to date that they mediate directly these effects. It is clear that some degree of sustained ketosis is required for clinical efficacy and that efficacy is maximized over a period of weeks versus days, despite a rapid onset of ketosis within hours. Whereas it is plausible that some dietary, pharmacokinetic factor(s) results in some level of seizure protection, the approximate 2-week time course for optimal seizure protection suggests a pharmacodynamic effect of the KD (e.g., parallel time course for changes in gene expression, mitochondrial proliferation, up-regulation of UCPs/transporters, etc) likely underlies the anticonvulsant nature of the diet. Thus, available data suggest that adaptations to, rather than a direct effect of, ketosis underlie the anticonvulsant nature of the KD.
Role of glucose restriction
Whereas most studies have suggested that persistent ketosis is essential to KD-induced seizure protection, others have posited that glucose restriction is the key feature (Greene et al., 2003). In addition to ketosis, it is clear that as ketonemia develops, another immediate consequence of CR and/or KDs is a ‘moderate’ reduction in blood glucose. Does caloric restriction simply act to limit gluconeogenic substrates that would otherwise reduce KD ratio and counter efficacy? Or, might glucose restriction result in another metabolic adaptation that helps quell aberrant hyperexcitability? Calorie restriction alone was sufficient to retard seizure susceptibility in juvenile and adult epileptic EL mice; and, blood glucose levels were inversely correlated with a decreased risk of seizures (Greene et al., 2001). Greene et al. (2003) hypothesized that CR reduces energy production through glycolysis, which limits a neuron's ability to reach (and maintain) high levels of synaptic activity necessary for seizure genesis.
Others have hypothesized that glucose restriction during KD treatment activates ATP-sensitive potassium (KATP) channels (Schwartzkroin, 1999; Vamecq et al., 2005). Interestingly, KATP channels are ligand-gated receptors broadly expressed throughout the central nervous system, in both neurons and glia (Thomzig et al., 2005). These channels act as metabolic sensors, linking cellular membrane excitability to fluctuating levels of ADP and ATP. Activation of this receptor by reduced ATP/ADP ratios opens the channel and leads to membrane hyperpolarization. When glucose is limited (e.g., during administration of a classic KD, which is typically CR by 25%), KATP channels might open to hyperpolarize the cell as the intracellular ATP concentrations fall. Conversely, when glucose is present and ATP concentrations rise, KATP channels close. As such, KATP channels may provide a measure of protection against a variety of metabolic stressors such as hypoxia, ischemia, and hypoglycemia, and are believed to regulate seizure threshold (Seino and Miki, 2003).
KATP channels are particularly abundant in the substantia nigra (Hicks et al., 1994), a region of the brain thought to act centrally in the propagation of seizure activity (Iadarola and Gale, 1982). KATP channels would therefore be ideally positioned to metabolically regulate the onset of several different seizure types, as does the KD. There is growing evidence that KATP channels may critically regulate seizure activity. Genetically engineered mice that overexpress the sulfonylurea (SUR) subunit of KATP channels were significantly more resistant to seizures induced by kainate, and showed no marked cell loss in hippocampus (Hernandez-Sanchez et al., 2001). Studies of KATP channel (Kir6.2-/-) knockout mice suggested that these channels help determine seizure threshold (Yamada et al., 2001). Following hypoxic challenge (∼5% O2), knockout mice exhibited myoclonic–tonic seizure activity, and, ultimately, death compared to controls who all recovered without sequelae.
Despite these observations, there is one important caveat in implicating KATP channels as mediators of a KD-induced anticonvulsant effect. Other studies have demonstrated an increase in energy reserves (specifically, ATP) after KD treatment (DeVivo et al., 1978; Pan et al., 1999). These data predict that KATP channels would remain closed, not open, during diet treatment, and would thus contribute to neuronal/glial cell membrane depolarization. Nevertheless, several findings are consistent with the notion that KATP channels are selectively activated during administration of a low-glucose, high-fat KD. First, KATP channels are regulated preferentially via glycolytic energy sources (Dubinsky et al., 1998). It has recently been shown that the glycolytic enzyme glyceraldehyde 3-phosphate dehydrogenase (GAPDH) serves as an accessory protein to KATP channels and regulates directly their activity (Dhar-Chowdhury et al., 2005; Jovanovic et al., 2005). The observed reduction in glycolytic processes after KD treatment (specifically, the concentration of fructose-1,6-bisphosphate, the key regulatory enzyme of glycolysis) is consistent with this notion (DeVivo et al., 1978; Puchowicz et al., 2005; Melo et al., 2006). Glycolytic flux may be further limited as a consequence of elevated ATP (DeVivo et al., 1978; Otani et al., 1984; Pan et al., 1999; Bough et al., 2006) and citrate (Yudkoff et al., 2001) levels on KD treatment; both ATP and citrate are feedback inhibitors of glycolysis.
Second, it is hypothesized that the accumulation of free fatty acids over the course of KD administration (Dekaban, 1966) may boost KATP channel activation (Vamecq et al., 2005). Whereas PUFAs freely cross the BBB, saturated free fatty acids are transported across the BBB via carrier-mediated processes (Avellini et al., 1994). Fatty acids that intercalate within neuronal cell membranes have been shown to interact potently with KATP channels, specifically reducing their affinity for (and inhibition by) ATP (Shyng and Nichols, 1998). Overall, these findings suggest that the unique nature of low-glucose, high-fat KDs promotes KATP channel activation, despite observed enhancements in oxidative energy production.
Recent experiments involving 2-deoxyglucose (2-DG) provide further support for a glucose-restriction hypothesis of KD action. Two-deoxyglucose is a glucose analogue, which inhibits phosphoglucose isomerase and, hence, glycolysis. Stafstrom et al. (2005) reported that the addition of 1 mM 2-DG decreased epileptiform burst frequency to 25–80% of baseline in rat hippocampal slices exposed to elevated extracellular potassium. More significantly, the same group also showed that 2-DG (250 mg/kg, i.p) elevated the after-discharge threshold in olfactory bulb of perforant-path kindled rats, markedly reduced the progression of kindling, and limited the expression of BDNF and its cognate receptor, trkB (Garriga-Canut et al., 2006).
Interestingly, there are a number of anticonvulsant parallels between 2-DG (Stafstrom et al., 2005; Garriga-Canut et al., 2006) and KD treatment (Bough et al., 2003). First, both 2-DG and KD elevated electrographic seizure threshold in vivo; second, both 2-DG and KD potently retarded the progression of epileptogenesis in kindling models of epilepsy in vivo; and, third, both 2-DG (in vitro) and KD (in vivo) diminished measures of hippocampal hyperexcitability. These results collectively suggest that the anticonvulsant actions of KD may work, in large part, via an inhibition of glycolysis. Importantly, because 2-DG is fairly well tolerated when administered orally (Pelicano et al., 2006), this compound may represent a novel treatment strategy for epilepsy.
Role of fatty acids
Polyunsaturated fatty acids (PUFAs) such as docosahexanoic acid (DHA, C22:6ω3), arachidonic acid (AA, C20:4ω6), or eicosapentanoic acid (EPA, C20:5ω3) are believed to affect profoundly cardiovascular function and health (Leaf and Kang, 1996; Nordoy, 1999; Leaf et al., 2003). In cardiac myocytes, PUFAs inhibited fast, voltage-gated sodium channels (Xiao et al., 1998) and L-type calcium channels (Xiao et al., 1997). Similar findings have been observed in neuronal tissue. For example, DHA and EPA diminished neuronal excitability and bursting in hippocampus (Xiao and Li, 1999).
It is not surprising then that PUFAs are becoming an increasingly popular focus of KD research. After KD treatment, specific PUFAs (i.e., AA and DHA) were found to be elevated in both serum (Cunnane et al., 2002; Fraser et al., 2003) and brain (Taha et al., 2005) of patients and animals. Importantly, one report documented that the rise (or drop) in total fatty acids during KD treatment closely paralleled clinical improvement (or loss) of seizure control (Dekaban, 1966). An additional study found that dietary supplementation with 5 g of (65%) n-3 PUFAs once per day produced a marked reduction in seizure frequency and intensity in a few epileptic patients (Schlanger et al., 2002). These findings suggest that KD-induced elevations in PUFAs such as DHA and/or AA might act directly to limit neuronal excitability and dampen seizure activity.
PUFAs could ultimately block seizure activity in a number of ways (Fig. 2). First, PUFAs may inhibit directly ion channel activity. Omega-3 (ω-3) PUFAs have been shown to: (1) inhibit both voltage-gated Na+ and Ca2+ channels, (2) increase the resistance to bursting induced by bicuculline, zero Mg2+, pentylenetetrazole or glutamate, and (3) prolong the recovery time from inactivation in hippocampal neurons (Vreugdenhil et al., 1996; Xiao and Li, 1999; Young et al., 2000). Second, in conjunction with ketone bodies, PUFAs may activate a lipid-sensitive class of K2P potassium channels (Vamecq et al., 2005). And, third, PUFAs may enhance the activity of the Na+/K+-ATPase (sodium pump). Elevated ω-3 and diminished ω-6 PUFAs levels in plasma membranes significantly increased sodium pump function (Wu et al., 2004). These findings indicate that elevations in brain levels of PUFAs after KD treatment (Taha et al., 2005) could help reduce neuronal hyperexcitability via a variety of direct mechanisms.
Figure 2. Potential pathways through which polyunsaturated fatty acids (PUFAs) may limit hyperexcitability in the brain. Acting directly, PUFAs such as arachidonic acid (AA), docosahexanoic acid (DHA), and/or eicosapentanoic acid (EPA) might inhibit both voltage-gated Na+ and Ca2+ channels, activate a lipid-sensitive class of K2P potassium channels, and enhance the activity of the Na+/K+-ATPase to limit neuronal excitability and dampen seizure activity. Acting indirectly, PUFAs might induce the expression and activity of uncoupling proteins (UCPs) to diminish reactive oxygen species (ROS), reduce neuronal dysfunction and induce a neuroprotective effect. Finally, PUFAs are expected to activate PPARα and induce a coordinate up-regulation of energy transcripts leading to enhanced energy reserves, stabilized synaptic function and limited hyperexcitability.
Download figure to PowerPoint
In addition to their direct actions on neuronal excitability, PUFAs may also act indirectly to limit excitotoxicity and neurodegeneration. PUFAs regulate the expression of numerous genes in brain via transcription factors such as PPARα (peroxisome proliferator-activated receptor-α; Sampath and Ntambi, 2004). Through induction of PPARα and its coactivator PGC-1, PUFAs induce the expression of mitochondrial uncoupling proteins (UCPs) and activate these proteins directly as well (Jaburek et al., 1999; Diano et al., 2003). Recent evidence suggests that PUFAs are required for mitochondrial UCP activity (Garlid et al., 2001).
Uncoupling proteins are homodimers that span the inner mitochondrial membrane and allow a proton leak from the intermembrane space to the mitochondrial matrix. There are three major isoforms that have been identified in the brain, UCP2, UCP4 and UCP5 (a.k.a., BMCP-1 or brain mitochondrial carrier protein-1). UCP proteins are increasingly implicated in the regulation of neuronal excitability and survival (Andrews et al., 2005). The uncoupling effect, albeit of small magnitude, reduces the proton-motive force, disassociates or ‘uncouples’ electron transport from ATP production, and indirectly decreases the production of reactive oxygen species (ROS). Although it would seem that increased levels of UCP proteins would diminish cellular energy production, Diano et al. (2003) showed that chronic overexpression of UCP2 in neuronal tissue increased cellular ATP and ADP levels by triggering mitochondrial biogenesis. KD appears to do the same; that is, studies show that the KD induces UCP expression, stimulates mitochondrial biogenesis, and enhances energy production (see also below). Seizures, by comparison, increase ROS generation and/or mitochondrial dysfunction, which can lead to neuronal dysfunction and excitotoxicity (Layton and Pazdernik, 1999; Kovacs et al., 2001; Kovacs et al., 2002; Sullivan et al., 2003). Interestingly, UCP2 is up-regulated after seizures (Diano et al., 2003). The protective role of UCPs was recently highlighted by Sullivan et al. (2003) who demonstrated that dietary enhancement of UCP expression and function in immature rats protected against kainate-induced excitotoxicity, most likely by decreasing ROS generation (Andrews et al., 2005). Further work demonstrated that mice maintained on a high-fat KD demonstrated an increase in the hippocampal expression and activity of all three mitochondrial UCPs and exhibited a significant reduction in ROS generation in mitochondria isolated from the same brain region (Sullivan et al., 2004). In conjunction with reports that ketone bodies potently decrease ROS generation (Veech et al., 2001; Veech, 2004), these reports suggest that the KD compensates for seizure-induced elevations in ROS generation and neuronal dysfunction to provide a neuroprotective effect.
Polyunsaturated fatty acids additionally regulate the transcription of numerous genes linked to energy metabolism (Sampath and Ntambi, 2005) through activation of PPARα, a scenario in which the KD is thought to re-program cellular metabolism (Cullingford, 2004). Indeed, numerous studies have described changes consistent with an enhancement in energy production following KD treatment. First, microarray expression studies demonstrated that KD induces a coordinated up-regulation of several dozen metabolic genes associated with oxidative phosphorylation after KD (Noh et al., 2004; Bough et al., 2006). Second, KD treatment stimulated mitochondrial biogenesis, resulting in a striking 46% increase in the number of mitochondria in the hilar/dentate gyrus region of rat hippocampus (Bough et al., 2006). And, third, levels of energy metabolites were increased after KD. Brain glycogen and ATP concentrations were boosted throughout rodent brain (DeVivo et al., 1978; Otani et al., 1984) and there was an elevation in the phosphocreatine-to-creatine (PCr:Cr) energy-reserve ratio in both animals (Bough et al., 2006) and humans (Pan et al., 1999). These findings are consistent with results that show ketones (4 mM BHB + 1 mM ACA) increased hydraulic work by 14% and improved energy status in perfused heart tissue (Sato et al., 1995). Further, there is an overall increased metabolic efficiency (DeVivo et al., 1978; Bough et al., 2006), decreased respiratory quotient (Bough et al., 2000b), and maximal mitochondrial respiratory rate in rodents following the KD (Sullivan et al., 2004). Collectively, these data provide compelling evidence that the KD enhances oxidative energy production by activating a variety of transcriptional, translational, and biochemical mechanisms in a concerted fashion.
Metabolic dysfunction has been identified in regions of hyperexcitability within the brain and is associated with several epileptic conditions. Impairment of mitochondrial function has been observed in the seizure foci of both human and experimental epilepsies (Kunz et al., 2000). Severe metabolic dysfunction occurred in both human and rat hippocampal tissue during periods of heightened neuronal activity (Kann et al., 2005). Kudin et al. (2002) demonstrated that seizure activity down-regulated mitochondrial enzymes involved in oxidative phosphorylation. In an earlier study, the same group demonstrated a specific deficiency in complex I activity and mitochondrial ultrastructural abnormalities within the hippocampal CA3 region of epileptic tissue resected from 57 human patients (Kunz et al., 2000). In view of previous studies demonstrating impaired oxidative phosphorylation capacity in pilocarpine-treated rats (Kudin et al., 2002) and in patients with epilepsy (Antozzi et al., 1995; Kunz et al., 2000), a KD-induced augmentation in oxidative phosphorylation and energy reserves seems likely to counter energetic deficiencies in epileptic tissue, making neuronal tissue more resilient to aberrant neuronal activity and, in this way, contributing to the diet's anticonvulsant actions.
Stabilized synaptic function
Intriguing as this argument may be, how exactly would enhanced energy reserves lead to stabilized synaptic function and diminished seizures? One possibility is via the sodium pump. ATP is primarily used to maintain ionic gradients, especially through actions of the transmembrane sodium pump (Hulbert and Else, 2000). Schwartzkroin originally hypothesized that KD-induced elevations in ATP concentrations might enhance and/or prolong the activation of the Na+/K+-ATPase, perhaps via an increase in the delta-G′ of ATP hydrolysis (Veech et al., 2001; Veech, 2004). In neurons, increased sodium pump activity might hyperpolarize the cell and/or reduce the resting membrane potential to diminish firing probability. Enhanced Na+/K+-ATPase function in neurons might also preserve normal neuronal functioning and/or delay a pathological buildup of high external K+ (Xiong and Stringer, 2000). In glia, increased activation of the Na+/K+-ATPase might slow glial depolarization and allow for prolonged uptake of extracellular K+ during periods of intense neuronal activity (e.g., high-frequency bursting). Increases in neuronal and/or glial action of the sodium pump would be expected to limit hyperexcitability and increase the resistance to seizures, as is noted after treatment with KD.
Although no studies have tested this sodium-pump hypothesis directly, a recent report suggests that KD tissue is more resistant to metabolic stress. When challenged with mild hypoglycemia, synaptic transmission within the dentate gyrus was maintained for approximately 60% longer in tissue from KD-fed animals compared to controls (Bough et al., 2006). These data suggest that the KD stabilizes synaptic transmission (both excitatory and inhibitory) for prolonged periods of time during metabolic stress such as during seizure activity. Hence, it seems likely that the KD induces seizure protection in part by preventing neuronal dysfunction (diminution of ROS/enhancement of energy reserves) and stabilizing synaptic transmission (enhancement in energy reserves).