We examined a number of Drosophila mutants with increased susceptibility to seizures following mechanical or electrical stimulation to better understand the underlying factors that predispose neurons to aberrant activity. Several mutations in this class have been molecularly identified and suggest metabolic disruption as a possible source for increased seizure susceptibility. We mapped the bang-sensitive seizure mutation knockdown (kdn) to cytological position 5F3 and identified citrate synthase as the affected gene. These results further support a role for mitochondrial metabolism in controlling neuronal activity and seizure susceptibility. Biochemical analysis in bang-sensitive mutants revealed reductions in ATP levels consistent with disruption of mitochondrial energy production in these mutants. Electrophysiological analysis of mutants affecting mitochondrial proteins revealed an increased likelihood for a specific pattern of seizure activity. Our data implicate cellular metabolism in regulating seizure susceptibility and suggest that differential sensitivity of neuronal subtypes to metabolic changes underlies distinct types of seizure activity.
STUDIES in Drosophila over the past 25 years have led to the discovery of mutants exhibiting stress-sensitive seizures and paralysis (Benzer 1971; Grigliatti et al. 1973; Homyk et al. 1980; Ganetzky and Wu 1982; Pavlidis and Tanouye 1995; Kuebler and Tanouye 2000). Mutants that exhibit seizures and paralysis upon mechanical stimulation (“banging”), are termed bang-sensitive paralytics. Suppression of bang-sensitive mutant phenotypes following the disruption of voltage-gated Na+ channels demonstrated that seizure behavior likely involves neuronal hyperexcitability (Ganetzky and Wu 1982). Simultaneous electrophysiological and behavioral characterization of these mutants demonstrated that seizure behavior was paralleled by distinct states of neuronal activity at the electrophysiological level (Lee and Wu 2002). Electrical stimulation can elicit seizures in both wild-type and mutant animals, although the required stimulation voltage is much less in bang-sensitive mutants, demonstrating their increased seizure susceptibility (Kuebler and Tanouye 2000). Additionally, administration of valproate, phenytoin, gabapentin, and potassium bromide, all of which act as anticonvulsants in humans, suppresses seizures in bang-sensitive mutants (Kuebler and Tanouye 2002; Reynolds et al. 2004; Tan et al. 2004).
The affected loci of several of the bang-sensitive mutations have been identified, including two mitochondrial proteins, a mitochondrial ribosomal protein (technical knockout, tko), and an adenine nucleotide translocase (stress-sensitive B, sesB) (Royden et al. 1987; Zhang et al. 1999). Additionally, mutations in an ethanolamine kinase (easily shocked, eas) and an aminopeptidase (slamdance, sda) also result in bang sensitivity (Table 1) (Pavlidis et al. 1994; Zhang et al. 2002). Although molecular identification of two mutants with mitochondrial defects suggests that impaired aerobic metabolism may generate this seizure phenotype, the physiological mechanism remains unclear (Jan and Jan 1978; Ganetzky and Wu 1982; Royden et al. 1987; Schubiger et al. 1994; Zhang et al. 1999).
To further characterize the mechanisms underlying bang sensitivity in Drosophila and to test the hypothesis that these seizure phenotypes are due to metabolic disruption, we have characterized a new bang-sensitive mutant and performed additional analysis of previously studied mutants. We molecularly identified knockdown (kdn) and found that it encodes citrate synthase, an essential metabolic protein. Measurements of ATP levels revealed reductions in kdn as well as in several other bang-sensitive mutants. Electrophysiological analysis revealed alterations in neuronal firing patterns in the giant fiber (GF) pathway of these mutants. Our results demonstrate that metabolic perturbations can alter membrane excitability and increase seizure susceptibility.
MATERIALS AND METHODS
Flies were cultured on cornmeal–molasses agar medium at 22°–29°. Bang-sensitive mutant strains used in this study include kdn1, tko25t, eas1, bas1, bss1, and sesB1. Genomic deficiency lines in the 5D–F region—Df(1)dx81, Df(1)5D, Df(1)G4e[l]H24i[R], and Df(1)JF5 (diagrammed in Figure 1A)—were utilized for complementation analysis and mapping of kdn. P-element lines disrupting the kdn locus (CG3861) were isolated elsewhere [PG129 (Bourbon et al. 2002) and KG04873 (Berkeley Drosophila Genome Project)]. Aconitase lines p07054 and Df(2L)TW1 were used for genetic interaction experiments. Wild type and control refer to Canton-S, unless otherwise stated.
Flies were collected under CO2 at 0–3 days after eclosion and kept at 36 animals/vial for 1–2 days before behavioral analysis. Vials were mechanically stimulated by placement in a bench-top vortex for 15 sec at the maximum setting. The time for each fly to right itself after vortexing was recorded (Ganetzky and Wu 1982).
Deficiency chromosomes were made heterozygous with GFP-tagged balancer chromosomes and homozygous deficiency embryos were identified as lacking GFP expression. GFP-expressing embryos from these stocks served as positive controls. Genomic DNA was isolated from 10 embryos of each genotype and all experiments were done in duplicate. Breakpoint mapping primers for deficiency analysis in Figure 1 were as follows: set (A) 5′-TGACTCGACTGCTCCGCTTGACTG-3′ and 5′-GACGGCGAGGGCTGCTACTACCAC-3′, (B) 5′-GCGCTGCCCACCACTCACCTAT-3′ and 5′-AAAGCGGCAAATGTTCCTAT-3′, (C) 5′-AAAATGCGCAAACCACCGACAACC-3′ and 5′-TAGGCCAGACCCAGCACCCGTATG-3′, (D) 5′-CCCCGCCCCCAGCATTAGGT-3′ and 5′-CCGTTCGTTCAGGGCACTCAGGT-3′, (E) 5′-GCCGTCGGTTGTCAAGGTGGAGTC-3′ and 5′-AGATGATGGCGTGCTGCGGTGAA-3′, and (F) 5′-CGTGCTCAACTGCCCGTCGTA-3′ and 5′-CTGGCCGCCCCTTTTTGTAGC-3′.
Analysis of candidate gene expression in homozygous P-element mutants was performed using RNA collected from non-GFP expressing embryos (as above). RNA from embryos homogenized in 3 m LiCl/6 m urea was ultimately DNAse treated and purified through further phenol chloroform extraction. Reverse transcriptase reactions (Sigma-Aldrich, St. Louis) were performed on RNA from each genotype using a polyT primer. Subsequent PCR analysis of generated cDNA was performed using primers specific to the gene under study and primer pairs were from different constitutive exons where possible. To generate internal controls for these RT–PCR reactions, we introduced concentrated rp49 primers (32 μm) at the start of PCR cycle 15. Samples from PCR reactions were extracted at the end of cycles 30, 35, and 40.
ATP content was determined by using a Harta MicrolumiXS Microplate Luminometer with Lumiterm II acquisition software (Harta Instruments, Gaithersburg, MD) and an ATP determination kit A-22066 (Invitrogen, Carlsbad, CA). ATP extraction was performed using bodies from 15 adult females/genotype, collected at room temperature, and transferred to 1.5-ml Eppendorf tubes on ice containing 200 μl of 6 m guanidine hydrochloride. All samples were generated in duplicate. The preparations were homogenized and placed at 95° for 5 min. After incubation, samples were centrifuged at 14,000 rpm for 5 min and 120 μl of the supernatant was transferred into a new 1.5-ml Eppendorf tube. This extract was mixed, diluted 1000-fold with TE buffer at pH 8, and 20 μl was placed in triplicate in a 96-well, white-walled, white-bottom plate. Standard reaction solution (180 μl) was simultaneously added to each well and mixed using a multipetter. High-gain 1-sec exposure “glow” reads were obtained in triplicate at ∼15 min after reaction initiation. Sample values were corrected for protein content using DC protein assays (Bio-Rad Laboratories) and ATP concentrations were normalized as a percentage of wild-type control samples.
Recordings from the GF escape circuit were performed as previously described (Kuebler and Tanouye 2000; Lee and Wu 2002). Briefly, the head and thorax of animals were tethered to a micropipette using nail polish under CO2 anesthesia and kept in a humidified chamber for at least 20 min to allow recovery. Glass microelectrodes were used to record potentials from the dorsal longitudinal indirect flight muscles (DLMs) and the reference electrode was placed in the abdomen. A sampling rate of 200 μsec was used with an Axopatch1D amplifier in current clamp = 0 configuration. An uninsulated electrolytically sharpened tungsten electrode was placed in each eye for GF circuit stimulation. Using a stimulation protocol optimized for eliciting initial discharge seizure activity (Lee and Wu 2002), we applied high-frequency electroconvulsive stimulation consisting of 0.2-msec pulses given at 200 Hz over a 300-msec train duration. Stimulation voltage was increased beyond the GF threshold until seizures were elicited. An interstimulus interval of at least 5 min was utilized to avoid the refractory state.
Mapping of kdn:
kdn homozygotes are developmentally delayed; time to eclosion is 4–7 days longer than wild type at room temperature. This delay is similar to that reported for sesB and tko as well as that for other mutants with deficits in cellular metabolism, including maggie, sluggishA, arginase, and the Minute mutants. Bang-sensitive paralysis in kdn homozygotes shows incomplete penetrance and variable expressivity that is independent of differences in age, rearing density, time of eclosion, or time of day.
On the basis of previous recombinational mapping (Ganetzky and Wu 1982), we performed deficiency mapping of kdn in the region between cytological positions 5A and 9F on the X chromosome. Df(1)dx81 failed to complement kdn, whereas Df(1)C149, Df(1)N73, Df(1)5D, and Df(1)G4e[L]H24i[R] did complement kdn, placing the gene in the region 5E–F. Cytological analysis of Df(1)5D by R. Kreber (personal communication) revealed breakpoints at 5D2–3;5E4–6, which defined a distal limit for kdn. Failure of complementation by Df(1)JF5, with reported breakpoints of 5E5;5E8, further refined the location of kdn. Bang-sensitive paralysis was stronger and less variable when kdn was heterozygous with a noncomplementing deficiency than in kdn homozygotes. Females carrying a single wild-type allele of kdn (Df/+) were completely wild type as were kdn/Dp(1;Y)dx+1 males, which carry X-chromosome region 5A;6D inserted into the Y chromosome.
Attempts to align the cytological map with the genomic sequence in the 5E interval indicated that the alignment given in FlyBase is incorrect. Primer pairs designed to cover cytological regions 5E2 through 5E8 (domains A–F in Figure 1A) gave positive PCR products for templates isolated from embryos homozygous for each of the deficiencies examined except Df(1)dx81. Since homozygous deficiency embryos should have completely lacked one or more of the genomic regions examined in these PCR experiments according to the previously assigned breakpoints, it was necessary to ascertain the correct breakpoints for these deficiencies at the molecular level. These corrected breakpoints are indicated diagrammatically in Figure 1A.
Two independently generated P-element insertion lines, PG129 and KG04873, fail to complement kdn. Both lines are lethal when homozygous and when heterozygous with Df(1)JF5. Flies heterozygous for kdn and either of these P elements displayed the same increase in severity of the bang-sensitive phenotype as in kdn/Df flies, suggesting that the insertions are kdn null alleles. Transposase-mediated precise excision of each P element resulted in viable revertants that complemented kdn (Figure 2). The P-element insertion sites of both PG129 and KG04873 are within the same large intron of CG3861 (Figure 1B), identifying it as the kdn locus. Consistent with this, RT–PCR analysis revealed that of the four genes neighboring these elements, only CG3861 showed a marked reduction in expression in embryos homozygous for the P-element insertions (Figure 1C).
Molecular identification of kdn:
CG3861 encodes citrate synthase, the first enzyme in the Krebs cycle. Sequence analysis of the original kdn mutant revealed an R95H substitution in the encoded protein (Figure 3). This change was not found in the wild-type background in which kdn was generated, nor were any sequence changes found in the neighboring gene, CG3446. The arginine residue altered in kdn is highly conserved in all species from yeast to humans (Figure 3) and corresponds to R94 in the human protein. This residue has not been implicated previously in enzymatic function (Wiegand and Remington 1986).
Genetic interactions with other mutants:
Our results suggest that altered cellular metabolism may underlie bang sensitivity in Drosophila. To further examine this possibility, we examined available adult viable mutants with affected metabolism, including Glycerol 3 phosphate dehydrogenase, arginase, and sluggishA, which encodes a proline oxidase, to determine if bang-sensitive phenotypes could be observed in mutants not previously analyzed for seizure behavior. None of the mutants in this limited sample exhibited bang sensitivity (data not shown). A genetic analysis to test the connection between metabolism and seizure behavior was performed by combining the kdn mutation, as well as other bang-sensitive mutations, with a deletion that removes the gene encoding aconitase, the enzyme directly following citrate synthase in the Krebs cycle. No significant alteration in the seizure phenotype was observed in animals heterozygous for this deletion and hemizygous for kdn, tko, eas, bas, or bss. Flies doubly heterozygous for either kdn or tko and the aconitase deletion also showed normal responses to mechanical shock. Similarly, heterozygotes for two or three of the bang-sensitive mutations (kdn, sesB, tko, and eas) appeared normal (Table 2), although increased bouts of flight-like wing activity were observed. These results indicate that partial reductions in these metabolic proteins do not have sufficient additive phenotypic effects to cause a bang-sensitive phenotype. However, both bss and sda mutations confer a partially dominant increase in seizure susceptibility (Ganetzky and Wu 1982; Kuebler and Tanouye 2000), suggesting that the cellular processes affected by these mutations are sensitive to gene dosage. We also tested for summation of seizure susceptibility in homozygous double mutant combinations. kdn sesB and sesB eas double mutants are small, weak, developmentally delayed, and exhibit dramatically increased sensitivity to mechanical stimulation (Table 2). The summation of bang sensitivity is similar to that reported previously for bss bas double mutants (Lee and Wu 2002) and suggests that all bang-sensitive mutations may ultimately impair the same neuronal mechanism.
Decreased ATP levels in bang-sensitive mutants:
To determine the effects of these mutations on energy levels, we performed quantitative ATP assays. ATP levels in control lines, including Canton-S, Oregon-R, and our w1118, were not significantly different from each other (Figure 4A). Homozygous kdn mutants had a mild reduction in ATP levels but this was not significant (P = 0.1, n = 8). Heterozygous kdn/PG129 animals displayed a significant reduction in ATP (P < 0.001, n = 5), which correlates with their significantly enhanced bang sensitivity relative to kdn homozygotes (Figure 2). Conversely, ATP levels were normal (P > 0.2; Figure 4A) in kdn/PG129R1 flies.
Other bang-sensitive mutants had similar decreases in ATP levels (Figure 4B). Although ATP levels were reduced in bas and bss mutants, the difference was not significant compared with wild type (P = 0.07 and P = 0.18, respectively). ATP levels were significantly reduced in eas, tko, and sesB mutants (P < 0.01 for each; Figure 4B). These results support the idea that bang-sensitive mutations are associated with defects in cellular metabolism.
Neurons in the giant fiber pathway have differential sensitivities to seizure induction in metabolic mutants:
The GF pathway in Drosophila is a neural circuit that mediates an escape reflex by initiating flight behavior in response to various sensory stimuli (Tanouye and Wyman 1980). By recording intracellularly from the DLMs, electrical activity in the GF pathway can be assayed. In bang-sensitive mutants this circuit exhibits physiological seizure activity that correlates with the behavioral seizure phenotype (Lee and Wu 2002). Physiological and behavioral seizures can be elicited in wild-type as well as in mutant animals by high-frequency electrical brain stimulation, but stimuli of longer duration or greater intensity are required in the former (Figure 5A) (Kuebler and Tanouye 2000; Lee and Wu 2002). Absence of seizure activity recorded from outputs of the GF pathway other than DLMs suggests that the aberrant activity originates in the peripherally synapsing interneuron (PSI) and/or DLM motor neuron rather than in the giant neuron itself (Pavlidis and Tanouye 1995; Lee and Wu 2002) (Figure 5G). This seizure activity begins with an initial neuronal burst or discharge, two classes of which have been distinguished in bang-sensitive mutants (Pavlidis and Tanouye 1995). Type I seizures are exemplified by trains at ∼10–30 Hz lacking any clear firing pattern that terminate abruptly (Figure 5A). Type II seizures exhibit increasing firing frequency with a concomitant decrease in transmission amplitude. Type II seizures predominate in bss mutants (Figure 5B) and most likely result from motor neuron dysfunction because reductions in amplitude of DLM responses evoked by action potentials should not result from defects in a neuron upstream of the motor neuron. kdn and tko display predominantly type I initial discharge seizures (Pavlidis and Tanouye 1995) (Figure 5C). sesB mutants also have type I initial discharge seizures (Figure 5E). Type II activity is most common in eas, although type I seizures were frequently observed as well (Figure 5, D and F) (Pavlidis and Tanouye 1995). Thus, it appears that type I seizures, which are most likely to originate upstream of the motor neuron, are the most readily evoked since they predominate in the mutants showing the weakest bang-sensitive behavior (Table 1). Conversely, type II seizures, which involve uncontrolled motor neuron activity, are most often observed in mutants with the strongest bang-sensitive behavioral phenotype, suggesting that the motor neuron may be able to tolerate metabolic perturbations better than other neurons (Figure 5G).
Mutations that cause seizures in response to mechanical stimulation include defects in genes encoding a mitochondrial ribosomal protein (Royden et al. 1987), an ethanolamine kinase (Pavlidis et al. 1994), the ADP/ATP translocase (Zhang et al. 1999), and an aminopeptidase (Zhang et al. 2002). The identification of another bang-sensitive mutation in a gene encoding a key component of the Krebs cycle supports the hypothesis that cellular energy, the primary product of mitochondrial function, is closely connected with regulation of seizure susceptibility. In fact, kdn is the third Drosophila seizure mutation associated with a mitochondrial protein (Figure 6). Following molecular identification of the Drosophila tko gene, the first human epilepsy-associated gene to be identified was also found to encode a gene involved in mitochondrial protein synthesis (Shoffner et al. 1990), suggesting the existence of conserved mechanisms. Unlike tko and sesB mutations, which have potentially broad effects on mitochondria and cellular function, disruption of citrate synthase by the kdn mutation most likely results in a specific impairment of aerobic metabolism. This finding provides a direct link between citrate synthase and seizure susceptibility.
Previous evidence links metabolic disruption with seizure susceptibility. These disruptions may be transient, as in hypoglycemia and ischemia, or constant, as caused by mutations that disrupt mitochondrial function (DiMauro et al. 2002; Nordli and De Vivo 2002; Vigevano and Bartuli 2002). Mutations affecting the two enzymes upstream of citrate synthase in the Krebs cycle, pyruvate carboxylase and pyruvate dehydrogenase, as well as less common mutations perturbing various steps in the TCA cycle, have been linked with human seizure disorders (De Vivo 1993; De Meirleir 2002). The phenotypes observed in the Drosophila mutants investigated here are strikingly similar to those seen in the related human syndromes. These similarities include developmental delay, inactivity, reduced viability, decreased life span, and seizures. Although mutation of a glycolytic enzyme in Drosophila can confer seizures and paralysis upon exposure to elevated temperature (Wang et al. 2004), it does not exhibit bang-sensitive seizures. Thus, bang-sensitive seizure susceptibility may involve a specific block in mitochondrial metabolism. Furthermore, the existence of lethal alleles in every bang-sensitive gene identified in Drosophila suggests that reduction in metabolism to a particular degree increases seizure susceptibility, but further disruption of metabolism is not compatible with life.
Quantitative analysis confirmed that bang-sensitive mutants reduce ATP levels with the reduction greatest in mutants with defects in mitochondrial proteins. Although comparisons between kdn/kdn and kdn/PG129 animals show a good correlation between the strength of the bang-sensitivity phenotype and the reduction in ATP levels, this correlation does not hold for the other bang-sensitive mutants examined here. Thus, it is possible that increased seizure susceptibility of some bang-sensitive mutants involves some factor(s) in addition to or instead of metabolic perturbations. Nonetheless, the enrichment for mutations affecting mitochondrial proteins with resultant decreases in ATP levels among bang-sensitive mutants and the identification of kdn as yet another mutation in this category strongly suggest that there is an important link between seizure susceptibility and metabolic perturbations.
Although we cannot rule out secondary developmental defects in the nervous system, it appears that there is a threshold for metabolic function below which increased seizure susceptibility occurs. Application of mitochondrial toxins, such as 3-nitropropionic acid, results in seizures in vertebrate models (Ludolph et al. 1992; Urbanska et al. 1998), suggesting that some of the key underlying mechanisms are conserved. Such summation was also observed between sesB and eas mutants, which exhibit different types of seizure activity, further suggesting that both mutations ultimately impair the same processes. Summation of seizure susceptibility in bang-sensitive double mutants may help explain why manifestation of various epilepsies in humans varies with the background genotype.
Initial seizure activity observed in bang-sensitive mutants following mechanical or electrical stimulation may reflect unchecked neuronal firing due to prolonged membrane depolarization. Such depolarization may be enhanced in cells lacking energy since it has been estimated that 75–90% of neuronal ATP is expended solely to maintain ionic gradients across the cell membrane (Beal et al. 1993; Lees 1993; Attwell and Laughlin 2001). Although specific mutations in the largest consumer of neuronal ATP, the Na+/K+ ATPase, result in temperature-sensitive paralysis and mild bang sensitivity, no mutations affecting the ATPase are known to cause the stereotypic behavioral and electrophysiological phenotypes seen in the class of bang-sensitive mutants (Schubiger et al. 1994; Pavlidis and Tanouye 1995). Thus, it remains unclear if reduced Na+/K+ pump function is closely connected with increased seizure susceptibility in bang-sensitive mutants. Alternative mechanisms, such as defects in Ca2+ buffering and/or synaptic function due to mitochondrial impairment, may exist (Royden et al. 1987; Pavlidis et al. 1994; Trotta et al. 2004).
Independent of the cellular mechanism, our results suggest that the initial seizure-like neurotransmission observed at the DLM in these convulsive mutants results from aberrant activity of either the PSI interneuron or the DLM motor neuron. The occurrence of type II seizure activity in the more extreme bang-sensitive mutants most likely results from uncontrolled motor neuron firing. In addition to the ethanolamine kinase mutant eas, the only other identified bang-sensitive mutant exhibiting type II seizure activity is sda, which encodes an aminopeptidase (Zhang et al. 2002). The mechanistic link, if any, between these two gene products is unknown. Isolation and characterization of additional mutations in this class may prove informative. We suggest that type I seizures prevail in animals exhibiting milder seizure susceptibility because the PSI is more sensitive than the motor neuron to metabolic perturbations.
These results link increased seizure susceptibility with animals defective in cellular metabolism. Moreover, these results suggest that neurons have differential susceptibilities to seizures following metabolic impairment. It will be informative to identify alterations in the most critical human metabolic pathways and investigate their correlation with predisposition toward epilepsy. Current studies are aimed at detailed dissection of the neuronal circuits mediating the bang-sensitive phenotype in flies and physiological analysis of seizure onset. It will also ultimately be useful to compare metabolic profiles of these neurons to determine which factors are associated with increased seizure susceptibility.
We thank Seymour Benzer, Marie-Laure Samson, and the Centre for Biologie du Developpement in Toulouse for fly stocks; Dan Kuebler and Mark Tanouye for technical advice; Robert Kreber for polytene chromosome analysis; Julie Simpson for helpful comments; and Al Laughon for luminometer use. This research was supported by NIH grants NS-44722 to T.F. and NS-15390 to B.G. This is paper no. 3628 from the Laboratory of Genetics.
Communicating editor: R. S. Hawley
- Received February 21, 2006.
- Accepted April 18, 2006.
- Copyright © 2006 by the Genetics Society of America