Address correspondence and reprint requests to Dr. T.N. Seyfried at Boston College, Biology Department, Chestnut Hill, MA 02167, U.S.A. E-mail: email@example.com
Summary: Purpose: Caloric restriction (CR) involves underfeeding and has long been recognized as a dietary therapy that improves health and increases longevity. In contrast to severe fasting or starvation, CR reduces total food intake without causing nutritional deficiencies. Although fasting has been recognized as an effective antiseizure therapy since the time of the ancient Greeks, the mechanism by which fasting inhibits seizures remains obscure. The influence of CR on seizure susceptibility was investigated at both juvenile (30 days) and adult (70 days) ages in the EL mouse, a genetic model of multifactorial idiopathic epilepsy.
Methods: The juvenile EL mice were separated into two groups and fed standard lab chow either ad libitum (control, n = 18) or with a 15% CR diet (treated, n = 17). The adult EL mice were separated into three groups; control (n = 15), 15% CR (n = 6), and 30% CR (n = 3). Body weights, seizure susceptibility, and the levels of blood glucose and ketones (β-hydroxybutyrate) were measured over a 10-week treatment period. Simple linear regression and multiple logistic regression were used to analyze the relations among seizures, glucose, and ketones.
Results: CR delayed the onset and reduced the incidence of seizures at both juvenile and adult ages and was devoid of adverse side effects. Furthermore, mild CR (15%) had a greater antiepileptogenic effect than the well-established high-fat ketogenic diet in the juvenile mice. The CR-induced changes in blood glucose levels were predictive of both blood ketone levels and seizure susceptibility.
Conclusions: We propose that CR may reduce seizure susceptibility in EL mice by reducing brain glycolytic energy. Our preclinical findings suggest that CR may be an effective antiseizure dietary therapy for human seizure disorders.
Fasting has been recognized as an effective antiseizure therapy since the time of the ancient Greeks (1). Wilder (2) suggested that fasting-associated seizure protection may arise from the long-term persistence of ketonemia. Further studies showed that a high-fat, low-protein, low-carbohydrate ketogenic diet (KD) could produce ketonemia and protect children against seizures (1,3,4). Although several studies have since confirmed the antiseizure efficacy of the KD, it remains unclear how fasting or the KD control seizures or how they affect the epileptogenic process (5–9).
The role of ketone bodies (β-hydroxybutyrate and acetoacetate) and glucose in the seizure-protective effects of fasting or the KD remains unclear. Plasma ketone levels were associated with seizure protection in some studies, but were not associated in other studies (5,10–13). Because brain ketone levels can depend on the plasma levels of ketones, glucose, and other metabolites (14), associations between seizure protection and plasma ketone levels may be obscured. Lennox (1) reported that reductions in plasma glucose occurred with seizure decline during fasting and subsequent ketosis. Glucose levels were constant during KD treatment in some studies, but were reduced during the treatment in other studies (5,15–18). No studies, to our knowledge, have investigated the relations among ketones, glucose, and seizures susceptibility under conditions of prolonged caloric restriction.
Caloric restriction (CR) involves underfeeding and has long been recognized as a dietary therapy that slows aging, improves health, and extends longevity in rodents and other species (19–21). CR is produced from a total dietary restriction and differs from severe fasting or starvation in that CR reduces total food intake without causing deficiencies of any specific nutrients (22). Bough et al. (12) recently reported that mild CR significantly elevated the pentylenetetrazole (PTZ) seizure threshold in young rats and suggested that age and caloric restriction are important for implementing the ketogenic diet. It is not clear to what extent CR alone may contribute to the antiepileptogenic effects of the KD.
The EL mouse has been studied as a natural model of multifactorial idiopathic epilepsy (23,24). Although several El epilepsy genes have been mapped in EL mice, their chromosomal location and influence on seizure susceptibility is markedly influenced by environmental factors (24,25). The seizures in EL mice originate in or near the parietal lobe and then spread quickly (generalize) to the hippocampus and to other brain regions (26,27). EEG abnormalities (synchronized spike–wave complexes at three to four per second) accompany the seizures in addition to clinical abnormalities involving vocalization (squeaking), incontinence, loss of postural equilibrium, excessive salivation, and head, limb, and chewing automatisms (24,28-31). Although no significant neurostructural abnormalities have been detected in the EL brain, an intense hippocampal gliosis is associated with seizure progression, as is often observed with complex partial seizures in humans (32,33). Phenytoin (PHT) and phenobarbital (PB), anticonvulsant drugs (AEDs) used for treatment of human partial epilepsies, also inhibit the seizures in EL mice (34–36). Based on these observations, the EL mouse is considered a genetic model for human complex partial seizures with secondary generalization.
In this report, we show that mild and moderate CR significantly inhibits epileptogenesis in juvenile and adult epileptic EL mice, and that plasma glucose levels are predictive of plasma ketone levels and seizure susceptibility.
MATERIALS AND METHODS
The inbred EL/Suz (EL) mice were originally obtained from J. Suzuki (Tokyo Institute of Psychiatry). The mice were maintained in the Boston College Animal Care Facility as an inbred strain by brother and sister mating. The mice were housed individually in plastic cages with Sani-chip bedding (P.J. Murphy Forest Products Corp., Montville, NJ, U.S.A.) and kept on a 12-h light/dark cycle at ∼22°C. Cotton nesting pads were provided for warmth. All cages and water bottles were changed once per week after seizure testing. The mice were weaned at age 25–27 days, and both females and males were used. Litters were split at weaning, so each group consisted of littermates. The mice were weighed each week after seizure testing for a total of 10 weeks. All procedures for mice use were in strict accordance with the NIH Guide for the Care and Use of Laboratory Animals and were approved by the Institutional Animal Care Committee.
All mice received a regular chow diet (PROLAB; Agway Inc.), which contained a balance of mouse nutritional ingredients and delivered 4.4 Kcal/g of gross energy. CR was administered by implementing a total dietary restriction. This involved restricting the amount of food a mouse normally consumed per day under ad libitum (AL) conditions. The average daily food intake (in grams) was measured under AL conditions in individual male and female mice at both juvenile and adult ages. The mice were given a known amount of food (∼50 g), and the difference in chow weight was recorded the following day at approximately the same time (11 a.m.–1 p.m.).
The food-restricted groups received 85% (15% CR) of the AL amount at juvenile ages. At adult ages, the mice received 85% and 70% (30% CR) of the AL amount. CR was initiated in the juvenile mice and adult mice at ages 30 and 70 days, respectively. The quantity of chow given to the CR mice was adjusted every 2 days according to the intake of the corresponding AL mice.
Seizure testing paradigm
Our recently developed seizure testing protocol was used to determine the handling induced seizure susceptibility of EL mice (24). The testing procedure included repetitive handling and simulated the stress normally associated with weekly cage changing (i.e., picking the mouse up by the tail for short intervals and transferring it to a clean cage with fresh bedding). The test included two trials that were separated by 30 min. In each trial, a mouse was held by the tail for 30 s, ∼10–15 cm above the bedding of its home cage. After 30 s, the mouse was placed into a clean cage with fresh bedding for 2 min. The mouse was then held again for 15 s before being returned to its home cage. Trial 2 was performed even if the mouse experienced a seizure in trial 1. The epileptic seizures commenced during holding or soon after the mice were placed on the clean bedding. Mice were tested each week for a total of 10 weeks by using this method. All mice were seizure naive at the start of treatment and were not stimulated or tested before the start of the study. Mice were undisturbed between testing phases (no cage changing), and testing was performed between 12 and 3 p.m. All mice were tested for seizures beginning 1 week after diet initiation.
Mice were designated seizure susceptible if they experienced a generalized seizure during seizure testing. Generalized seizures in EL mice involve loss of postural equilibrium and consciousness, together with excessive salivation, head, limb, and chewing/swallowing automatisms. An erect forward-arching Straub tail, indicative of spinal cord activation, was also seen in most mice having generalized seizures. Mice that displayed only vocalization and twitching without progression to generalized seizure were not considered seizure susceptible.
Measurement of plasma glucose and β-hydroxybutyrate
Blood was collected every week ∼1 h after seizure testing. The mice were anesthetized with Isovet (Schering Plough), and the blood was collected in heparinized tubes by puncture of the retroorbital sinus. The blood was centrifuged at 5,000 g for 10 min, and the plasma was collected and stored at -80°C until analysis. Plasma glucose and β-hydroxybutyrate (β-OHB) concentrations were measured spectrophotometrically using the Trinder assay and a UV enzymatic assay, respectively (Sigma, St. Louis, MO, U.S.A.). At least three separate mice per group were used at each time point except in the adult 30% CR group, in which blood was collected from two mice.
A two-tailed t test was used to evaluate the significance of differences between the AL and CR groups for body weights. A χ2 test with a 2 × 2 contingency table was used to evaluate the significance of differences between the AL and CR mice in the number of mice seizing each week. Simple linear regression was used to determine the relation between plasma glucose and β-OHB, and multiple logistic regression was used to assess the influence of plasma glucose and trend (2-week interval of treatment) on seizure susceptibility (37). In each figure, n designates the number of individual mice analyzed.
Tolerance of caloric restriction
No adverse side effects were seen in the mice receiving either the 15% CR diet (at juvenile ages) or the 30% CR diet (at adult ages). Despite reductions in total body weight (discussed later), the CR-fed mice appeared healthy and were more active than the AL-fed mice, as assessed by ambulatory and grooming behavior. No signs of vitamin or mineral deficiency were observed in the CR-fed mice. These findings are consistent with the well-recognized health benefits of mild to moderate diet restriction in rodents (21).
Influence of caloric restriction on body weight in juvenile and adult EL mice
Juvenile CR-fed and AL-fed mice were matched for age and body weight before the treatment (Fig. 1A). The total energy intake of the AL-fed mice increased from ∼20 Kcal/day to about 29 Kcal/day over the 10-week treatment period. The total energy intake of the CR-fed mice was adjusted to 85% of the AL energy intake. Although body weight was ∼3 g greater in the juvenile AL-fed males than in AL-fed females, body weights for males and females were pooled in each dietary group because the effects of CR on body weight were similar in both sexes. The mice from the 15% CR juvenile group lost ∼8% of their body weight during the first week of treatment, and their weights remained significantly lower than those of the AL group throughout the study. Despite this initial body-weight loss, the rate of body-weight gain beyond week 2 was similar in the AL-fed and CR-fed mice. These findings together with the overall healthy appearance of the mice indicate that 15% CR did not significantly alter global body development.
The adult AL-fed and CR-fed mice also were matched for age and body weight before treatment (Fig. 1B). The total energy intake of the AL-fed mice was ∼30 Kcal/day and remained relatively constant over the 10-week treatment period. The total energy intake of the CR-fed mice was adjusted to 85% (15% CR) or 70% (30% CR) of the AL energy intake. As at juvenile ages, body weights were pooled for males and females because the effects of CR were similar in both sexes. In contrast to the juvenile AL-fed mice, the body weights of the adult AL-fed mice remained relatively constant over the 10-week testing period. The body weights in the 15% CR and 30% CR groups were ∼12% and 15% lower, respectively, than the weights of the AL-fed group and remained stable over the testing period (Fig. 1B). The CR-induced reductions in body weights were correlated with the degree of CR.
Influence of caloric restriction on seizure susceptibility in juvenile and adult EL mice
Because seizure susceptibility is similar in repetitively tested male and female EL mice (24), seizure susceptibility was pooled for males and females in each dietary group at both juvenile and adult ages. CR significantly reduced the incidence and delayed the onset (weeks after diet initiation) of seizures in the juvenile mice (Fig. 2A). Whereas all (18 of 18) of the AL-fed mice had seizures at week 10, only 71% (12 of 17) of the CR-fed mice had seizures. The seizure-protective effect of CR in adult mice was correlated with the degree of CR, where the delay in seizure onset and the reduction in seizure susceptibility was greater in the 30% CR group than in the 15% CR group (Fig. 2B). The 15% CR also was less effective in reducing seizure susceptibility in the adult mice than in the juvenile mice (Fig. 2A and B). These findings indicate that CR delayed epileptogenesis in EL mice at both juvenile and adult ages and that the antiepileptogenic effect of CR at adult ages was proportional to the degree of CR.
Influence of caloric restriction on plasma glucose and β-hydroxybutyrate levels
CR caused plasma glucose levels to decrease significantly after 1 week in the 15% CR-fed juvenile mice (Fig. 3A). These levels remained significantly lower than those in the AL-fed mice throughout the study, but increased gradually in the CR-fed mice after week 5. In contrast to glucose levels, 15% CR caused plasma β-OHB levels to increase significantly in the juvenile mice (Fig. 3B). These levels peaked from weeks 3 to 7, and then decreased in association with the increase in glucose (Fig. 3A and B). In the adult EL mice, the decrease in plasma glucose levels was proportional to the degree of CR (Fig. 3C). As observed in the juvenile mice, the changes in plasma glucose levels in the adults were inversely related to the changes in plasma β-OHB levels (Fig. 3D). We also noted a gradual increase in glucose levels and a decrease in β-OHB levels during the later weeks in the juvenile and adult CR groups. These findings indicate that reductions in caloric intake cause inverse changes in plasma glucose and β-OHB levels in EL mice.
Statistical analysis of plasma glucose levels, β-hydroxybutyrate levels, and seizure susceptibility
To determine whether blood glucose levels were predictive of blood β-OHB levels and seizure susceptibility, we analyzed the data using both simple linear regression and multiple logistic regression. These statistical analyses also included those mice considered outliers (i.e., mice that experienced CR), but showed no changes in body weight, plasma glucose, or β-OHB levels. One outlier was found in the juvenile CR group, and three outliers were found in the adult CR groups (two in the 15% CR group and one in the 30% CR group). The outliers were considered nonrestricted for caloric intake and similar to the mice in the AL-fed groups. Although the outliers were excluded from the data presented in Figs. 1–3, their data were included in the regression analyses to assess the relations between glucose, β-OHB, trend (2-week interval of treatment), and seizures.
Simple linear regression analysis was used to examine the relation between plasma glucose levels (the independent or explanatory X variable) and β-OHB levels (the dependent or response Y variable) in both the juvenile (Fig. 4A) and adult mice (Fig. 4B). These variables were identified based on physiologic and neurochemical studies, which indicated that plasma glucose levels determine plasma β-OHB levels during periods of fasting (38). The assumptions of simple linear regression were met according to established criteria (37). The slopes of the regression lines were highly significant and showed that the plasma β-OHB levels increased as glucose levels decreased in both the juvenile mice [slope = -0.14; 95% CI (confidence interval) = -0.16 to −0.12; t = -5.8; Y = 1.9 - 0.14X] and in the adult mice (slope = −0.16; 95% CI = −0.17 to −0.15; t = -14.4; Y = 2.3 - 0.16X). The coefficient of determination (r2) was greater in the adult mice than in the juvenile mice because two CR dosages (15 and 30%) were used in the adults. These findings indicate that plasma β-OHB levels are dependent on plasma glucose levels and that glucose levels are predictive of β-OHB levels in all mouse groups.
Multiple logistic regression was used to quantify the risk of seizure based on changes in plasma glucose levels in the juvenile and the adult mice. Trend was included in the analysis to assess the effects of repetitive seizure testing on seizure susceptibility. Because the simple linear regression indicated that β-OHB levels are dependent on glucose levels, the β-OHB levels could not be included in the logistic regression analysis (37).
In the juvenile and adult mice, trend and glucose were significant in predicting seizures (Table 1). The data show that for every decrease in 1 mM of glucose, the odds of having a seizure decreased by 2.6 and 1.9 in the juvenile and adult mice, respectively. Likewise, the odds of having a seizure increased by 12.8 and 4.1 in juvenile and adult mice, respectively, for every increase in the 2-week interval of treatment. This positive correlation between seizure frequency and the number of weeks tested is consistent with the finding of Todorova et al. (24). Our findings indicate that plasma glucose levels are predictive of seizure susceptibility in EL mice at both juvenile and adult ages.
Table 1. Multiple logistic regression analysis of plasma glucose and trend on seizure susceptibility in juvenile and adult EL micea
Plasma glucose levels and seizure susceptibility were analyzed over a period of 10 weeks.
Variables include the intercept, a mathematical constant with no clinical interpretation; trend, a 2-week interval of treatment; glucose, the plasma glucose concentration (mM).
The mathematical weighting of each variable in the model.
The estimated error of the mathematical weighting, indicating the precision of the estimated coefficient.
e The Wald test statistic was calculated from the data compared by using the χ2 distribution with 1 degree of freedom. The test statistic is used to determine the p value.
The probability value indicating the trend and glucose are significantly predictive of seizures.
Controlling for other variables in the model, for every unit increase in trend and decrease in glucose, the odds of having a seizure increased by 12.8 and decreased by 2.6 for juvenile mice, respectively, and increased by 4.1 and decreased by 1.9 for adult mice, respectively.
Juvenile (n = 30)
Adult (n = 50)
Comparison of caloric restriction and the ketogenic diet on epileptogenesis in EL mice
We next compared the effects of CR with those of the KD on seizure susceptibility in juvenile EL mice by using our data and the previously published data of Todorova et al. (5), in which the KD was fed ad libitum. All the mice were matched for age, sex (male), and body weight before the analysis (Fig. 5). The 15% CR diet was more effective than the KD in delaying seizure onset and in reducing seizure susceptibility (Fig. 5). The analysis of only male mice in this comparison accounts for the slight difference in the percentage of mice seizing between the data presented in Figs. 2A and 5. Our comparative analysis of CR and the KD was restricted to juvenile mice because the KD does not inhibit seizure susceptibility in adult EL mice (Todorova, unpublished data). In contrast to CR, which lowers blood glucose levels (Fig. 3A), the KD has no significant effect on blood glucose levels in the juvenile EL mice (24). Our findings indicate that mild CR has a greater antiepileptogenic effect than the KD in the juvenile EL mice.
We found that mild to moderate CR was an effective dietary therapy for preventing seizures in epileptic EL mice. The antiepileptogenic effect of CR was significantly correlated with reductions in plasma glucose levels, and with elevations in plasma ketone body levels. Although similar changes in plasma glucose and ketone levels occur during acute fasting (1), CR is less severe than fasting because CR reduces total food intake without producing nutritional deficiency or physical discomfort. The CR-induced body weight reductions did not produce developmental delay in the juvenile mice. Furthermore, the CR-fed mice appeared more active and healthy than the AL-fed mice at both juvenile and adult ages. These findings concur with the observations of others that the ad libitum feeding of sedentary rodents is a form of overfeeding and can be associated with adverse metabolic events (21). We also found for the first time that changes in plasma glucose levels could predict seizure susceptibility at both juvenile and adult ages in a natural epilepsy model. Because CR reduces the amount rather than the types of calories ingested, our findings indicate that changes in energy intake can influence seizure susceptibility in EL mice.
The antiepileptogenic efficacy of CR was both age and dose dependent in the EL mice. The 15% CR diet was more effective in preventing seizures at juvenile ages than at adult ages. This may arise from the generally higher ketotic state at younger than older ages. In the adult mice, conversely, the 30% CR diet was more effective in preventing seizures than the 15% diet. It is important to note that the antiepileptogenic effects of CR waned in both the juvenile and adult mice in the later weeks of the experiment. This was associated with a gradual increase in glucose levels and decrease in β-OHB levels, suggesting a physiologic adaptation to the effects of CR over time. Variation in food intake among individual mice also could influence the antiseizure effects of the diet. This was most apparent in the outlier mice, in which CR had no noticeable effect on seizure susceptibility, body weights, or plasma glucose and ketone levels. These findings indicate that food intake must be carefully adjusted for each mouse to maintain the antiepileptogenic effectiveness of CR.
CR was more effective than the KD in preventing seizures in juvenile EL mice and, in contrast to the KD, also was effective in preventing seizures in adult EL mice. The KD was developed originally to mimic the positive antiseizure effects of starvation or fasting by producing ketosis without nutritional deficiency (1,2,9). However, controversy exists as to whether ketosis alone can account for the antiseizure effects of the KD (6–8,18,39,40). We also found no association between seizure protection and blood ketone levels in juvenile EL mice fed the KD ad libitum(5). We suggest that some of the ambiguity associated with these studies may arise from the administration of the KD under either restricted or nonrestricted conditions.
Our findings in mice together with those in humans indicate that CR, like fasting, lowers blood glucose levels while inducing ketosis (1,41–44). This contrasts with studies of the KD, in which blood glucose levels are not reduced in association with ketosis (5,15). It is interesting to note that the antiseizure effect of the KD was greater when it was administered under restricted than under ad libitum conditions (12), suggesting that reduced blood glucose levels may enhance the efficacy of the KD. Despite evidence for an inverse relation between blood glucose and ketone levels in normal humans and humans with epilepsy under fasting or the KD (45), little attention has been given to the possibility that these dietary therapies prevent seizures through an effect on blood glucose levels. From previous neurochemical studies and from our statistical analyses, we show that blood glucose levels determine both blood ketone levels and seizure susceptibility in EL mice and emphasize the importance of blood glucose as a predictor of epileptogenesis in this epilepsy model.
We propose that CR may reduce seizure susceptibility in EL mice by reducing brain energy production through glycolysis. Previous studies indicate that glycolytic energy is crucial for maintaining synaptic activity and that the energy for a seizure is derived largely through glycolysis (46–50). Furthermore, recent studies indicate that CR downregulates glycolytic gene expression in the mouse brain (51). Although CR may not reduce the total energy available to the brain, it will reduce the energy available from glycolysis.
Reduced glycolytic energy would be expected to deplete the reserves of immediately available energy necessary for seizure initiation and spread. Although glucose is used exclusively for brain energy metabolism under normal physiologic conditions, the mammalian brain will metabolize ketone bodies for energy when blood glucose levels decrease, as during fasting or CR (38). Because ketone bodies are metabolized directly to acetyl-coenzyme A in the mitochondria, they bypass cytoplasmic glycolysis and provide brain energy directly through the Krebs cycle (14,52). Although they provide adequate energy for most brain activities under normal conditions, ketones alone are unable to deliver the immediate and large amount of energy necessary for seizure initiation or maintenance. Thus, a shift in energy metabolism from glucose utilization to ketone utilization may underlie the mechanism by which CR inhibits seizure susceptibility in EL mice.
Whereas we favor the glycolytic energy hypothesis for the antiepileptogenic mechanism of CR in EL mice, we are aware of the conflicting data regarding whether decreases or increases in bioenergetic substrates are favorable for seizure control (53). Consequently, we do not exclude the possibility that CR may reduce seizure susceptibility in EL mice through other mechanisms. Indeed, CR is known to influence neuroendocrine systems that might also alter seizure susceptibility (54). Further studies in the EL and other epilepsy models will be necessary to identify the antiepileptogenic effects of CR.
It is of interest that reduced cerebral glucose metabolism has been observed in association with the use of several common AEDs [e.g., PHT, carbamazepine, valproate, PB, and primidone (55–59)]. These findings suggest that changes in cerebral energy metabolism may be common to epilepsy diets and medications. The association between AED efficacy and cerebral energy metabolism may require additional study.
In summary, we have demonstrated for the first time that CR alone can inhibit seizure susceptibility in a genetic model of idiopathic epilepsy. Moreover, the antiepileptogenic action of CR may operate by reducing blood glucose, which affects cerebral energy metabolism. CR was more effective than the KD in reducing seizure susceptibility in juvenile EL mice and could also prevent seizures in adult EL mice. Because CR is easy to administer and is devoid of adverse side effects, our preclinical findings suggest that CR may be an effective dietary therapy for some human epilepsies.
Acknowledgment: This research was supported by the Boston College Research Fund and NIH grant (HD39722).