Metabolic flux from the Krebs cycle to glutamate transmission tunes a neural brake on seizure onset

Kohlschütter-Tönz syndrome (KTS) manifests as neurological dysfunctions, including early-onset seizures. Mutations in the citrate transporter SLC13A5 are associated with KTS, yet their underlying mechanisms remain elusive. Here, we report that a Drosophila SLC13A5 homolog, I’m not dead yet (Indy), constitutes a neurometabolic pathway that suppresses seizure. Loss of Indy function in glutamatergic neurons caused “bang-induced” seizure-like behaviors. In fact, glutamate biosynthesis from the citric acid cycle was limiting in Indy mutants for seizure-suppressing glutamate transmission. Oral administration of the rate-limiting α-ketoglutarate in the metabolic pathway rescued low glutamate levels in Indy mutants and ameliorated their seizure-like behaviors. This metabolic control of the seizure susceptibility was mapped to a pair of glutamatergic neurons, reversible by optogenetic controls of their activity, and further relayed onto fan-shaped body neurons via the ionotropic glutamate receptors. Accordingly, our findings reveal a micro-circuit that links neural metabolism to seizure, providing important clues to KTS-associated neurodevelopmental deficits.

Introduction TCA cycle in a specific set of glutamatergic neurons to the scale of their neural transmission to achieve metabolic control of seizure susceptibility. Given that early-onset seizure is one of the most prominent symptoms in KTS patients, our findings provide important insights into the neural mechanism responsible for the development of KTS.

Indy acts in glutamatergic neurons to suppress bang-induced seizure
The early-onset seizure is one of the most prominent neurological symptoms observed in KTS patients. We hypothesized that Drosophila mutants of KTS-associated genes might recapitulate genetic conditions in KTS patients and display seizure-like behaviors. After a mechanical stimulus (vortexing for 25 s), wild-type flies immediately recovered a normal posture and resumed their locomotion (Fig 1A). By contrast, Indy mutants homozygous or trans-heterozygous for loss-of-function alleles exhibited "bang-induced" seizure-like behaviors (e.g., severe wing-flapping, abdominal contractions, leg-twitching, or failure to stand upright), as reflected in a high seizure index and prolonged recovery time ( Fig 1B and S1 Movie). The bang-sensitive seizure (BSS) phenotypes in Indy mutants were comparable to those observed in other seizure mutants, such as easily shocked or slamdance but weaker than bang senseless mutants [35][36][37] [7][8][9]39]. Overexpression of INDY T245M in VGlut-expressing neurons similarly induced BSS in a wild-type Indy background ( Fig 1C). Moreover, transgenic expression of wild-type Indy cDNA in VGlut-expressing neurons was sufficient to rescue BSS in Indy mutants ( Fig 1D). These results suggest that Indy function in glutamatergic neurons is necessary and sufficient for seizure suppression. We further found that rogdi mutants displayed BSS comparably to Indy mutants (S4A Fig), and its seizure-suppressor function was similarly mapped to glutamatergic neurons (S4B- S4D Fig). Nevertheless, our subsequent analyses focused on elucidating Indy-dependent mechanisms of seizure control since the molecular function of ROGDI has been poorly defined.

Down-regulation of glutamate transmission induces BSS in Indy mutants
To determine if glutamate transmission actually contributes to BSS phenotypes in Indy mutants, we genetically manipulated the expression of VGLUT, a vesicular transporter that incorporates glutamate into synaptic vesicles [40], and examined subsequent effects on Indydependent BSS. The heterozygosity of VGlut did not induce seizure-like behaviors in wild-type flies (Fig 2A); however, it substantially increased seizure susceptibility in Indy heterozygous mutants. VGLUT overexpression in glutamatergic neurons partially but significantly rescued BSS phenotypes in INDY-depleted flies (Fig 2B). It has been shown that VGLUT overexpression increases synaptic vesicle size in larval motor neurons and their spontaneous release of  [40][41][42], whereas it could lead to neurodegeneration in a cell type-specific manner [42,43]. Nonetheless, no gross effects of VGLUT overexpression were detected on the seizure susceptibility in wild-type flies (Fig 2B). We further mapped two glutamate receptors, N-methyl-D-aspartic acid receptors (Nmdar1 and Nmdar2) and glutamate-gated chloride channel (GluClα), that mediate Indy-dependent BSS in transheterozygous mutants ( Fig 2C). However, hemizygous Nmdar2 mutants exhibited significant BSS even in a wild-type Indy background. These genetic interactions suggest that down-scaling of glutamate transmission via the ionotropic glutamate receptors may underlie BSS phenotypes in Indy mutants.

A metabolic link between the TCA cycle and glutamate transmission underlies Indy-dependent seizure suppression
Two classes of presynaptic neurons-glutamatergic and GABAergic-are intrinsically coupled to astrocytes via the glutamate/GABA-glutamine cycle, mediating excitatory and inhibitory transmissions, respectively. An imbalance in their activity is associated with neurological disorders, including a seizure [44][45][46]. In contrast to the mammalian central nervous system, acetylcholine is the primary excitatory neurotransmitter in the fly brain whereas glutamate plays this role at the neuromuscular junction [47][48][49]. Nonetheless, our genetic evidence indicated that low glutamate transmission likely induced seizure-like behaviors in Indy mutants. We thus examined if the loss of Indy function impaired the biosynthesis of glutamate or GABA. Our quantitative assessment of free amino acids revealed that glutamate levels were substantially reduced in Indy mutants (Fig 3A). In contrast, no significant differences in GABA levels were detected between wild-type and Indy mutant flies. Considering that glutamate . n.s., not significant; � P < 0.05, �� P < 0.01, ��� P < 0.001, as determined by one-way ANOVA with Holm-Sidak's multiple comparisons test. (C) Heterozygosity of ionotropic glutamate receptor Nmdar or GluClα induces BSS in heterozygous Indy mutants. Significant Indy x glutamate receptor interaction effects on seizure index (P = 0.0047 for Nmdar1; P = 0.0023 for Nmdar2; P = 0.0009 for GluClα) and recovery time (P = 0.0136 for Nmdar2; P = 0.0049 for GluClα) were detected by two-way ANOVA. Data represent means ± SEM (seizure index, n = 60 flies in 3 independent experiments; recovery time, n = 2-37). � P < 0.05, �� P < 0.01, ��� P < 0.001, as determined by Holm-Sidak's multiple comparisons test.
To validate this hypothesis, we fed the GDH inhibitor, epigallocatechin gallate [50], to adult flies and tested its effects on BSS phenotypes. Pharmacological inhibition of GDH robustly increased both seizure index and recovery time after BSS in Indy mutants, but it did not induce BSS in wild-type flies (Fig 3C). Similar results were obtained using another GDH inhibitor (i.e., diethylstilbestrol), excluding possible off-target effects (S5 Fig). Accordingly, these data suggest that GDH-mediated metabolic flux is limiting for seizure suppression, particularly in the context of Indy deficiency. To further examine the involvement of the TCA cycle in seizure suppression, we genetically manipulated isocitrate dehydrogenase 3 (IDH3), a hetero- Indy-dependent BSS involves a metabolic link between the TCA cycle and glutamate transmission. (A) Indy mutants display low levels of free glutamate that are rescued by oral administration of α-ketoglutarate (AKG). Flies were fed control or AKG-containing food (20 mM) for 3 d before harvesting. Free amino acids in wholebody extracts were quantified using ion-exchange chromatography. Relative levels of glutamate and GABA were calculated by normalizing to glycine levels. A significant Indy x AKG interaction effect on glutamate levels (P = 0.0006) was detected by two-way ANOVA. Data represent means ± SEM (n = 3-4). n.s., not significant; ��� P < 0.001, as determined by Holm-Sidak's multiple comparisons test. (B) Glutamate biosynthesis via the TCA cycle. IDH, isocitrate dehydrogenase; GDH, glutamate dehydrogenase; GPT, glutamate-pyruvate transaminase; GOT, glutamate-oxaloacetate transaminase. (C) Oral administration of the GDH inhibitor epigallocatechin gallate (EGCG) exaggerates BSS phenotypes in Indy mutants. Flies were fed control or EGCG-containing food (5 mg/ml) for 3 d before the BSS assessment. Quantitative analyses of BSS in individual flies were performed as described in Fig 1. Two-way ANOVA detected a significant Indy x EGCG interaction effect on seizure index (P = 0.0025). Data represent means ± SEM (seizure index, n = 58-60 flies in 3 independent experiments; recovery time, n = 5-43 flies). n.s., not significant; �� P < 0.01, ��� P < 0.001, as determined by Holm-Sidak's multiple comparisons test. (D) Idh3g heterozygosity induces BSS phenotypes in heterozygous Indy mutants. Significant Indy x Idh3g interaction effects on seizure index (P = 0.0002 by two-way ANOVA) and recovery time (P = 0.0086 by Aligned ranks transformation ANOVA) were detected. Data represent means ± SEM (seizure index, n = 50-60 flies in 3 independent experiments; recovery time, n = 8-33 flies). ��� P < 0.001, as determined by Holm-Sidak's multiple comparisons test (seizure index) or by Wilcoxon rank sum test (recovery time). (E and F) Oral administration of AKG rescues BSS phenotypes in Indy mutants. Flies were fed control or AKG-containing food (20 mM) for 3 d before the BSS assessment. Data represent means ± SEM (seizure index, n = 50-119 flies in 3 independent experiments; recovery time, n = 10-56 flies). n.s., not significant; � P < 0.05, �� P < 0.01, as determined by one-way ANOVA with Holm-Sidak's multiple comparisons test (seizure index), by Welch's ANOVA with Dunnett's T3 multiple comparisons test (E, recovery time), or by Kruskal-Wallis test with Dunn's multiple comparisons (F, recovery time).
https://doi.org/10.1371/journal.pgen.1009871.g003 tetrameric enzyme that converts isocitrate into α-ketoglutarate in the rate-limiting step in the TCA cycle ( Fig 3B). Heterozygosity of the Idh3g mutant allele induced BSS in Indy heterozygous mutants, but not in wild-type flies ( Fig 3D). In addition, RNAi-mediated depletion of individual IDH3 subunit proteins in wild-type glutamatergic neurons alone was sufficient to induce BSS (S6 Fig). These lines of pharmacological and genetic evidence indicate that reducing the metabolic flux from the TCA cycle to glutamate biosynthesis may down-scale seizuresuppressing glutamate transmission, thereby causing BSS phenotypes in Indy mutants.
We further hypothesized that dietary supplements of TCA cycle intermediates should compensate for the genetic deficits in flies with loss of Indy function and rescue their BSS phenotypes. Oral administration of α-ketoglutarate indeed ameliorated seizure phenotypes in Indy mutants in a dose-dependent manner ( Fig 3E). Moreover, α-ketoglutarate supplementation restored glutamate levels in Indy mutants to wild-type levels ( Fig 3A). Unexpectedly, we found that α-ketoglutarate supplementation induced a dose-dependent increase in seizure index, but not recovery time, in wild-type flies (Fig 3F). A possible explanation for this observation is that an excess of α-ketoglutarate may lead to an imbalance in the metabolic flux between the TCA cycle and glutamate in wild-type flies, thereby lowing the threshold for seizure initiation. Alternatively, surplus α-ketoglutarate may deplete synaptic vesicles containing glutamate by promoting their fusion with the synaptic membrane [51] (see Discussion).

A pair of glutamatergic neurons mediates Indy-dependent seizure suppression
Genetic and biochemical analyses in Indy mutants revealed the contribution of low glutamate transmission to their seizure phenotypes. We found that daily locomotor activity was reduced in Indy mutants, but their waking activity (i.e., activity count per minute awake) was indistinguishable from wild-type control (S7A Fig). Moreover, wild-type and Indy mutants displayed similar climbing activities in a negative geotaxis assay (S7B Fig). We thus reasoned that the glutamate transmission in Indy mutants might not be limiting for general motor function as reported previously [13,16,[52][53][54], but a subset of the glutamatergic neurons would be somehow sensitized to loss of Indy function for seizure control. Our transgenic mapping of the Indy RNAi phenotypes actually identified a very small group of neurons expressing the neuropeptide leucokinin [55] (hereafter, LK neurons) as a neural locus important for Indy-dependent seizure suppression (Figs 4A and S2B). Introduction of a temperature-sensitive Gal80 transgene allowed us to turn off the Gal4-driven expression of the Indy RNAi transgene at low temperature (i.e., 21˚C) but silence endogenous Indy expression in LK neurons at high temperature (i.e., 29˚C). The conditional RNAi confirmed that Indy was necessary in adult LK neurons for seizure suppression (S8 Fig). Conversely, transgenic expression of the wild-type Indy cDNA in LK neurons was sufficient to rescue BSS in Indy mutants ( Fig 4B).
LK neurons can be divided into three groups based on their neuroanatomical positions in the adult brain and ventral nerve cord [56]. These include lateral horn LK (LHLK), subesophageal ganglion LK (SELK), and abdominal LK (ABLK) neurons ( Fig 4A). To determine which subset of LK neurons contributed to Indy-dependent seizure suppression, we employed additional Gal80 transgenes that were constitutively expressed in specific subgroups of LK neurons and inhibited Gal4 activity to restrict their expression of the Indy RNAi transgene. This intersectional strategy revealed that ABLK neurons in the ventral nerve cord, targeted by tsh-Gal80 [57], were dispensable for BSS control (Fig 4A and 4C). In contrast, INDY depletion in a single pair of LHLK neurons, specifically targeted by a glutamatergic VGlut-Gal80 transgene, was necessary to induce BSS (Fig 4A and 4C). Data represent means ± SEM. n.s., not significant; � P < 0.05, �� P < 0.01, ��� P < 0.001, as determined by two-way ANOVA with Holm-Sidak's multiple comparisons test (seizure index, n = 60 flies in 3 independent experiments) or by Aligned ranks transformation ANOVA with Wilcoxon rank sum test (recovery time, n = 9-29). (C) INDY depletion in a pair of LHLK neurons is necessary for BSS phenotypes in Indy RNAi flies. LK neuron-specific expression of the Indy RNAi transgene was inhibited specifically in LHLK neurons by the VGlut-Gal80 transgene. Data represent means ± SEM (seizure index, n = 55-60 flies in 3 independent experiments; recovery time, n = 3-28 flies). n.s., not significant; �� P < 0.01, ��� P < 0.001, as determined by one-way ANOVA with Holm-Sidak's multiple comparisons test. https://doi.org/10.1371/journal.pgen.1009871.g004

PLOS GENETICS
A previous study suggested that LK neurons are unlikely glutamatergic [56]. However, we found additional evidence supporting that LHLK neurons indeed express the glutamatergic marker VGlut. LK neuron-specific expression of B3 recombinase led to the genomic excision of VGlut coding sequence flanked by the B3 recombination target sites from the genomeedited VGlut allele (B3RT-VGlut-B3RT-LexA) [58], thereby driving the downstream expression of a transgenic LexA driver only in VGlut-expressing LK neurons (S9A Fig). Assessment of the transgene expression in the adult fly brain revealed that LHLK neurons, but not SELK neurons, expressed the VGlut-derived LexA (S9B Fig), indicating VGlut expression only in LHLK neurons. Given these observations, we asked whether Indy controls the glutamate transmission from LHLK neurons for seizure suppression. A transgenic fluorescence sensor for detecting synaptic release of glutamate [59] validated that INDY depletion in LHLK neurons lowered the levels of glutamate release ( Fig 5A). Transgenic overexpression of wild-type VGLUT not only rescued the glutamate transmission in INDY-depleted LHLK neurons ( Fig  5A) but also suppressed their BSS phenotypes ( Fig 5B). These observations indicate that the glutamate transmission from INDY-depleted LHLK neurons is likely limiting for seizure suppression.
To validate the Indy-relevant link between the TCA cycle and glutamate transmission in seizure-suppressing LHLK neurons, we performed two additional experiments. First, oral administration of α-ketoglutarate partially but significantly restored the glutamate transmission in INDY-depleted LHLK neurons ( Fig 5C) and suppressed the BSS phenotypes induced by loss of Indy function in LK neurons ( Fig 5D). Second, IDH3 depletion in LK neurons was sufficient to induce BSS (S10 Fig), phenocopying the seizure induction by the pan-

LK neuron activity gates the behavioral output from seizure initiation
LK signaling has been implicated in neural physiology and behaviors relevant to metabolism in Drosophila [55,57,[60][61][62][63][64]. It has also been shown that Indy mutant flies physiologically mimic calorie-restricted animals and exhibit low fat levels [16], raising the possibility that energy deficiency might underlie Indy mutant seizure. Nonetheless, we found some evidence against this idea. First, a lipid-rich diet did not rescue Indy mutant seizures while significantly suppressing BSS phenotypes in other seizure mutants [37] (S11A Fig). We note, however, the lipid-rich diet did not rescue low glucose levels in Indy mutants (S11B Fig). Second, LK neuron-specific loss of Indy function did not affect the baseline levels of triglyceride and glucose (S11C Fig) and the corresponding seizure was insensitive to the lipid-rich supplement (S11D Fig). Finally, hypomorphic mutants of Lk and Lk receptor (Lkr) genes did not exhibit any detectable phenotypes in seizure index under our experimental conditions (S12 Fig). We thus reasoned that LK neuron activity, but not the LK signaling or metabolism per se, might be critical for Indy-dependent seizure suppression. Live-brain imaging revealed that INDY depletion reduced the baseline levels of intracellular Ca 2+ in LHLK neurons (Fig 6A), possibly reflecting their low activity. We asked whether transgenic manipulations of LK neuron activity could either suppress or mimic INDY-depletion phenotypes in BSS.
To this end, optogenetic transgenes were combined with different genetic tools to excite or silence LK neuron activity only during seizure induction for the assessment of their effects on BSS (Fig 6B). Transgenic flies expressing CsChrimson [65] were exposed to red light to transiently excite LK neurons during a mechanical stimulus. This manipulation significantly suppressed BSS phenotypes in Indy RNAi flies (Fig 6C), whereas the light transition per se negligibly affected Indy-dependent seizure. These observations validate that Indy RNAi phenotypes are readily reversible in adult LK neurons within a range of seconds, excluding the possibility that any long-term effects of INDY depletion on neuronal development or metabolic stress contribute to the seizure control. We consistently found that amber light-dependent silencing of LHLK neurons by the eNpHR transgene [66] was sufficient to induce BSS, even in a wild-type background ( Fig 6D). These results demonstrate that LK neuron activity inversely correlates with seizure susceptibility. Nonetheless, the optogenetic inhibition of LK neuron activity alone did not trigger seizure-like behaviors in the absence of a mechanical stimulus (S13 Fig). We reason that this neural locus may not initiate seizure-like behaviors but inhibit the behavioral output from an as-yet-unmapped seizure-initiating locus in the adult Drosophila brain.

NMDA receptors in dorsal fan-shaped body neurons act downstream of LHLK neurons to suppress BSS
Although the implication of LK-LKR signaling in Indy-dependent seizure was not evident, we reasoned that LKR-expressing neurons could still act downstream of LHLK neurons via the seizure-suppressing glutamate transmission. It has been shown that LKR neurons are present in distinct parts of the adult fly brain, including the pars intercerebralis, ellipsoid body, and fan-shaped body (FSB) [55,57,60]. Also, our transgenic expression of the GFP-fused synaptotagmin 1 revealed LHLK projections in the LH and FSB (Fig 7A). We thus employed a dorsal FSB (dFSB)-specific Gal4 driver (i.e., R23E10-Gal4) to examine whether glutamate receptors expressed in dFSB neurons contribute to the susceptibility of BSS. We indeed found that dFSB-specific depletion of NMDAR2 was sufficient to cause BSS phenotypes, comparable to the high seizure index and long recovery time observed in Nmdar2 mutants (Fig 7B). Neither electrical silencing of dFSB neurons by the inwardly rectifying Kir2.1 channel [67] nor blocking their synaptic transmission by tetanus toxin light chain (TNT) [68] significantly affected seizure index; however, both the transgenic manipulations lengthened the recovery time in BSS-positive animals ( Fig 7C). These data together map the metabolic control of seizure onset to the glutamate transmission between LHLK and dFSB neurons and suggest that the subsequent output of dFSB neurons may govern the duration of seizure-like behaviors.

Discussion
Human genetic studies have previously identified several loss-of-function alleles of ROGDI or SLC13A5 among KTS-associated loci. In the current study, we behaviorally assessed the susceptibility to a specific form of seizure-like activity in Drosophila mutants of the KTS- PLOS GENETICS associated genes. We defined genetic, biochemical, and neural pathways of SLC13A5/Indy-relevant seizure in the context of the citric acid cycle, glutamate metabolism, and glutamatergic transmission (Fig 7D). We further mapped the reversible and scalable functions of the seizure suppressor gene to a surprisingly small group of glutamatergic neurons (i.e., a pair of LHLK neurons) in the adult fly brain. In a broad sense, these findings demonstrate a specific microneural circuit that gates both initiation and duration of seizure-like behaviors.
Given the citrate transporter activity of SLC13A5/INDY, it has been suggested that neuronal energy failure may cause SLC13A5-associated epilepsy [7,8]. Citrate acts as a precursor of fatty acid biosynthesis while also serving as a key intermediate in the TCA cycle. Accordingly, an Indy mutation mimics calorie-restricted conditions and thereby causes metabolic and longevity phenotypes in Drosophila and mouse models, including alterations in lipid metabolism (i.e., low-fat storage), insulin signaling, and mitochondrial biogenesis [16,21]. Perturbations of the TCA cycle and altered levels of extracellular TCA metabolites are consistently observed in SLC13A5-associated KTS patients [69]. In fact, genetic losses of other metabolic enzymes involved in the TCA cycle (e.g., IDH3, malate dehydrogenase, fumarase hydratase) have been implicated in several neurological disorders, including early-onset seizure and developmental delay [70][71][72][73][74]. Nonetheless, there are conflicting reports on the effects of a ketogenic diet on KTS-associated epilepsy [8].
Genetic or pharmacological manipulations of metabolic enzymes and relevant genes likely impact on general cell physiology. The long-term metabolic stress may thus lead to pleiotropic effects and poor phenotypic outcomes, possibly explaining the seizure phenotypes observed in our genetic models. Nonetheless, several lines of our evidence support a more specific mechanism for Indy-dependent seizure suppression. First, Indy mutants did not display any gross motor defects, while their seizure phenotypes were mapped very specifically to a small group of the glutamatergic LK neurons among other broadly defined groups of neurons (e.g., GABAergic or cholinergic neurons). Second, Indy displayed specific genetic interactions with Idh3, VGlut, and select glutamate receptors on BSS in trans-heterozygous conditions, whereas BSS were not detected in individual heterozygous mutants. Third, genetic enhancement of the glutamate transmission by VGLUT overexpression in LK neurons was sufficient to suppress BSS in INDY-depleted flies. Finally, the INDY-depletion phenotypes were readily reversible in the adult LK neurons (i.e., AKG administration, conditional RNAi) and the optogenetic excitation of INDY-depleted LK neurons only during a mechanical stimulus was sufficient to suppress BSS although the possible off-target effects of the Indy RNAi transgene were not completely excluded. These observations together support our model that impairment of the metabolic flux between the TCA cycle and glutamate biosynthesis-specifically, the conversion of α-ketoglutarate to glutamate-down-scales the seizure-suppressing transmission from a specific subset of glutamatergic neurons, resulting in BSS phenotypes.
Previous studies in animal models have actually implicated low levels of TCA metabolites and glutamate in epilepsy [73,[75][76][77][78][79]. Moreover, co-injection of α-ketoglutarate suppresses chemically induced epilepsy in mice [80,81], consistent with our results. Intriguingly, the recent observation that α-ketoglutarate promotes the interaction between synaptotagmin 1 and phospholipids demonstrates an unexpected role of α-ketoglutarate in synaptic vesicle fusion [51]. Thus, we reasoned that an excess of α-ketoglutarate might deplete synaptic vesicle pools and thereby interfere with their transmission. This explains why oral administration of α-ketoglutarate to wild-type flies induces rather than suppresses BSS under our experimental conditions.
Although Drosophila genetic studies have led to the isolation and characterization of a number of seizure-related mutants, much less is known about the neural loci responsible for inducing seizure and sustaining seizure-like activity [82]. For instance, the seizure-suppressing function of the potassium-chloride symporter kazachoc and seizure-inducing neural locus in a BSS mutant of the voltage-gated sodium channel paralytic have been mapped to the mushroom body in the adult brain [83,84]. Seizure suppression by the product of the transmembrane domain gene, julius seizure, was more broadly mapped to cholinergic or GABAergic neurons but not to glutamatergic neurons [85]. The emergence of a seizure has been thought to involve three distinct components of neural circuits that mediate seizure initiation, seizure buildup, and seizure spread, respectively [86]. Modulatory neurons likely regulate the excitability of these neural components and thus shape seizure intensity or duration. Based on this model, we propose that glutamatergic LK neurons mediate a seizure-inhibitory pathway such that their conditional silencing de-represses both seizure onset and duration. Accordingly, down-regulation of the glutamatergic transmission from LK neurons induces BSS, as opposed to the general involvement of excitatory glutamate in seizure induction [44][45][46].
Our data suggest that LK signaling is not directly involved in controlling seizure susceptibility, which makes sense considering the temporal scale of neuropeptide signaling in general. Nonetheless, we found that the metabolic flux of the TCA cycle in LK neurons and their glutamate transmission onto dFSB neurons might be closely linked to the control of BSS. Evidence for the involvement of LK neurons in feeding, metabolism, and associated physiology (e.g., circadian behaviors, sleep, memory) is abundant [55,57,[60][61][62][63][64]. The activity of LHLK neurons is also sensitive to the metabolic state and the time of day [57,60,63]. Thus, we hypothesize that LK neurons act as a neural sensor that integrates internal cues (e.g., circadian clocks, sleep state, and metabolic state) while engaging different postsynaptic partners and divergent pathways for relevant physiological outputs through distinct signaling molecules. In particular, dFSB neurons are a key sleep-promoting locus in the adult fly brain, analogous to the sleeppromoting ventrolateral preoptic nucleus (VLPO) in mammals [87,88]. Given the implication of VLPO in consciousness loss during epileptogenesis [89], the LK-dFSB pathway may reveal the conserved neural principles underlying intimate interactions among metabolism, epilepsy, and sleep [90][91][92].
Drosophila models of human genetic disorders have proven to be valuable tools for elucidating underlying pathogenic mechanisms. Our findings enrich this body of knowledge by providing a genetic, biochemical, and neural map of the seizure suppressor pathway related to KTS. The complexity of the mammalian brain and species-specific organization of excitatory neurotransmitters in the nervous systems may limit the direct relevance of our working model to KTS pathogenesis at neural-circuit levels. Nonetheless, given the strong conservation of SLC13A5/Indy homologs and their physiological function among different species, we propose that similar cellular mechanisms may underlie early-onset seizures in KTS patients and explain their relative resistance to generic antiepileptic drugs.

Bang-sensitive seizure analysis
To avoid any genetic-background effects on bang-sensitive seizure, Indy mutant stocks were outcrossed six times to the wild-type control (w 1118 ) for isogenization. All the behavioral tests for heterozygous, homozygous, and trans-heterozygous mutants were conducted with the isogenized Indy lines and w 1118 control. For transgenic lines, seizure behaviors in trans-heterozygous animals were compared to those in all appropriate heterozygous controls. Where applicable, additional transgenic combinations or independent transgenes were tested for validation. Unless otherwise indicated, BSS was assessed in male flies. Three to seven-day-old flies were fed cornmeal-yeast-agar medium in LD cycles at 25˚C for 3 d. They were CO 2 -anesthetized and harvested into fresh vials containing the same food >3 h before a group of flies (n = 5 for non-optogenetic experiments; n = 5-10 for optogenetic experiments) was transferred to an empty vial using an aspirator. Each vial was vortexed at a maximum speed for 25 s using Vortex-Genie 2 (Scientific Industries, Inc.) and then video-recorded for >30 s using a cellular phone (LG G5 Pro or Samsung Galaxy Note 8). BSS-positive flies were defined if they displayed initial paralysis upon mechanical stimulus, followed by uncoordinated movements such as wing-flapping, abdominal contractions, or leg-twitching before their recovery of a normal posture. BSS was quantitatively assessed by three parameters. A seizure index was calculated as the ratio of the number of BSS-positive flies to the total number of flies tested in each experiment (n =~20 flies per genotype or condition) and averaged from three independent experiments. Recovery time was calculated individually for BSS-positive flies as the latency to normal posture after vortexing and was averaged for each genotype or condition. Percent seizure was calculated as the percentage of BSS-positive flies at each second after vortexing (n = 60 flies per genotype or condition from all the three experiments). To determine the length of a refractory period in Indy mutants, the first mechanical stimulus was given to each fly and then the second mechanical stimulus was given only to BSS-positive animals at the indicated time after recovery from their first BSS. The seizure index was calculated in each experiment (n = 10 flies per time point) and averaged from three independent experiments.

Locomotor behavior analysis
Locomotor activities in individual male flies were measured in 1-min bins using the Drosophila Activity Monitor system (TriKinetics). Daily locomotor activity (total activity count per 12-hour light: 12-hour dark cycle) and waking activity (averaged activity count per minute awake) were analyzed as described previously [95,96]. For the climbing activity measurement, a group of 10 male flies was kept in the climbing chamber. Flies were allowed to climb for 5 seconds after gentle tapping-down. Climbing distance in each fly was measured as the highest position during the recording.

Oral administration of chemicals
Each chemical was dissolved in 5% sucrose and 2% agar food (behavior food) at the indicated concentration and fed to flies for 3 d before the BSS assessment. The chemicals tested included epigallocatechin gallate (EGCG, Sigma), diethylstilbestrol (DES, Sigma), α-ketoglutarate (Sigma), and KetoCal 4: 1 (fat: carbohydrate plus protein, Nutricia).

Optogenetics
Transgenic flies were grown in light-tight vials and kept in constant dark. They were loaded on to the behavior food containing 0.1 mM all-trans retinal (Sigma) and entrained at 25˚C for 3 d. A light source with the indicated wavelength was turned on during the mechanical stimulation. A white LED was covered with a 585-nm or 470-nm filter to generate amber or blue light, respectively, while the 630-nm LED provided red light.

Immunofluorescence imaging
Immunostaining of whole-mount adult brains was performed as described previously [97]. Dissected brains were fixed with 3.7% formaldehyde in PBS, blocked with PBS containing 0.3% Triton X-100 and 5% normal goat serum, and incubated with anti-GFP antibody (Invitrogen, A-11122; diluted at 1:1000) or anti-mCherry antibody (Developmental Studies Hybridoma Bank, DSHB-mCherry-3A11; diluted at 1:20) at 4˚C overnight. After washing in PBS containing Triton X-100, immunostained samples were further incubated with species-specific Alexa Fluor secondary antibodies (Jackson ImmunoResearch) at 4˚C overnight, washed with PBS containing Triton X-100, and then mounted using VECTASHIELD mounting medium (Vector Laboratories). Confocal images were obtained using an FV1000 microscope (Olympus) and analyzed using ImageJ software.

Live-brain imaging
Transgenic flies were loaded on to the behavior food containing 0 mM (control) or 20 mM αketoglutarate (Sigma), entrained at 25˚C for 3 d, and then anesthetized in ice. Whole brains were dissected out in hemolymph-like HL3 solution (5 mM HEPES pH 7.2, 70 mM NaCl, 5 mM KCl, 1.5 mM CaCl 2 , 20 mM MgCl 2 , 10 mM NaHCO 3 , 5 mM trehalose, 115 mM sucrose) and then transferred on cover glass in a magnetic imaging chamber (Chamlide CMB, Live Cell Instrument) filled with HL3 buffer. Fluorescence images from live brains were recorded at room temperature using photoactivated localization microscopy (ELYRA P.1, Carl Zeiss) with a C-Apochromat 40x/1.20 W Korr M27 at a pixel resolution of 512 x 512. Fluorescence signals were quantified by background subtraction and analyzed using ZEN software (Carl Zeiss).

Amino acid profiling
Male flies were loaded on to behavior food containing either 0 mM (control) or 20 mM αketoglutarate and then entrained in LD cycles at 25˚C for 3 d before harvest. Whole-body extracts were prepared from 50 flies per condition, and the levels of free amino acids were quantitatively measured using ion-exchange chromatography as described previously [95,96].