Sleep drive, not total sleep amount, increases seizure risk

Abstract

Sleep loss has been associated with increased seizure risk since antiquity. Using automated video detection of spontaneous seizures in Drosophila epilepsy models, we show that seizures worsen only when sleep restriction raises homeostatic "sleep drive," not simply when total sleep amount falls. This is supported by the paradoxical finding that acute activation of sleep-promoting circuits worsens seizures, because it increases sleep drive without changing sleep amount. Sleep-promoting circuits become hyperactive after sleep loss and are associated with increased whole-brain activity. During sleep restriction, optogenetic inhibition of sleep-promoting circuits to reduce sleep drive protects against seizures. Downregulation of the 5HT1A serotonin receptor in sleep-promoting cells mediates the effect of sleep drive on seizures, and we identify an FDA-approved 5HT1A agonist to mitigate seizures. Our findings demonstrate that while homeostatic sleep is needed to recoup lost sleep, sleep drive comes at the cost of increasing seizure susceptibility.

Fig. 1. Sleep loss leads to more severe induced seizures in bang-sensitive tko^25t and eas^pc80f mutant flies.

a Experimental protocol depicting seizure induction on vortexer followed by video recording of fly seizures. Seizures are quantified for atonic (“paralysis”), tonic/clonic (“convulsive”), and recovery (“postictal”) phases. “Seizure time” is atonic + tonic/clonic phases. “Total time” is “seizure time” + recovery phase. Created in BioRender. Sehgal, A. (2025) https://BioRender.com/p4jrytx. b–d tko^25t flies exhibit decreased mean sleep times with caffeine and increased mean sleep times with gaboxadol. n = 78 flies/condition. e tko^25t flies demonstrate prolonged mean seizure duration with caffeine treatment (n = 51 flies) as compared to control (n = 74 flies). No significant effect was seen with gaboxadol treatment (n = 67 flies). f–h eas^pc80f flies exhibit decreased mean nighttime sleep times with caffeine (n = 47 flies) and increased mean sleep times with gaboxadol (n = 62 flies) as compared to control (n = 59 flies). i eas^pc80f flies demonstrate prolonged mean seizure times with caffeine (n = 43 flies) and gaboxadol (n = 71 flies) as compared to control (n = 88 flies). j–l tko^25t;rye flies (n = 96 flies) exhibit decreased mean daytime and nighttime sleep times as compared to tko^25t and rye flies (n = 95 flies/condition). m tko^25t;rye flies (n = 33 flies) demonstrate prolonged mean seizure times as compared to tko^25t flies (n = 18 flies). n–p tko^25t;sss^p1 flies (n = 94 flies) and sss^p1 flies (n = 94 flies) exhibit decreased mean daytime and nighttime sleep times as compared to tko^25t flies (n = 96 flies). q tko^25t;sss^p1 flies (n = 33 flies) demonstrate prolonged mean seizure times as compared to tko^25t flies (n = 44 flies). Two-group two-tailed t-test or one-way ANOVA with Dunnett’s or Tukey’s multiple comparisons test was used. Data are presented as mean values ± SEM. Source data are provided as a Source Data file.

Fig. 2. Drosophila exhibit spontaneous seizures after sleep restriction.

a Flies are loaded into individual wells of a 24- or 48-well plate then video recorded for 96 h in a 12 h light:12 h dark cycle. EthoVision XT is used to detect fly position (yellow with red dot). An algorithm (“CynthiSeize”) was developed to detect movements consistent with hyperkinetic seizures. Red line tracing = fly position over previous 10 seconds; Blue arrow = fly with sample seizure (right). Created in BioRender. Sehgal, A. (2025) https://BioRender.com/sxslfsb. b tko^25t (blue), and eas^pc80f (pink), flies were fed vehicle or caffeine and sleep was measured over multiple day (white bar in x-axis) and night (dark bar in x-axis) cycles. Caffeine treatment led to less sleep across both models. n = 29–32 flies/condition. c Spontaneous seizures were counted across 48 30-minute time bins for each treatment condition in tko^25t (blue) and eas^pc80f (pink) flies. Sleep loss with caffeine treatment increased the number of spontaneous seizures (Σ) across both genotypes. n = 29-32 flies/condition. d Average number of seizures per day per fly in tko^25t (blue) and eas^pc80f (pink) flies is increased by caffeine across genotypes. n = 31 flies/condition for tko^25t flies and 32 flies/condition for eas^pc80f flies. e Average seizure duration trends upward in tko^25t (blue) and eas^pc80f (pink) flies. Statistical testing not possible due to only one fly with a seizure in tko^25t and eas^pc80f flies without caffeine. Error bars not depicted given only one seizure detected in conditions without caffeine. With caffeine, n = 10 seizures for tko^25t flies and 82 seizures for eas^pc80f flies. Violin plots depict a solid line at median and dotted lines at quartiles. Negative binomial model with Wald test for daily number of seizures was used. Data are presented as mean values ± SEM. Source data are provided as a Source Data file.

Fig. 3. Sleep loss associated with increased sleep drive exacerbates seizures.

a Experimental protocol depicting seizure induction on vortexer followed by video recording of fly seizures. Seizures are quantified for previously detailed seizure phases. Created in BioRender. Sehgal, A. (2025) https://BioRender.com/p4jrytx. b–d tko^25t; c584-Gal4>UAS-TrpA1 flies, the “sleep rebound group”, were maintained at 18 °C, shifted to 30 °C overnight for 12 h, then recovered in 18 °C. Thermogenetic activation of the “sleep rebound group” at 30 °C led to sleep loss that recovered back to baseline at 18 °C. After returning to 18 °C, night sleep was increased as compared to baseline night sleep, consistent with homeostatic sleep rebound. ZT = zeitgeber time. e Change in p(Doze) was calculated at ZT 0-3 and ZT 9-12 and reveals higher levels in the “sleep rebound group” consistent with increased sleep drive. For (b–e), n = 43 for tko^25t flies and n = 40 flies for tko^25t;UAS-TrpA1 and tko^25t; c584-Gal4>UAS-TrpA1 each. f After thermogenetic activation of the “sleep rebound group” for 24 h, total seizure times are prolonged. n = 30 tko^25t flies, n = 45 tko^25t; UAS-TrpA1 flies, and n = 38 tko^25t; c584-Gal4>UAS-TrpA1 flies. g–itko^25t; Tdc2-Gal4>UAS-TrpA1 flies, the “no sleep rebound group”, were maintained at 18 °C, shifted to 30 °C overnight for 12 h, then recovered in 18 °C. Thermogenetic activation of the “no sleep rebound group” at 30 °C led to sleep loss that persisted even after flies were restored to 18 °C. There was no homeostatic sleep rebound after temperature was returned to 18 °C. j Change in p(Doze) was calculated for the “no sleep rebound group” and demonstrates no significant changes, indicating no change in sleep drive with sleep loss. For g-j, n = 45 for tko^25t flies, n = 46 for tko^25t;UAS-TrpA1 flies, and n = 41 for tko^25t;Tdc2-Gal4>UAS-TrpA1 flies. k After thermogenetic activation of the “no sleep rebound group” for 24 h, there was no significant difference in total seizure times. n = 45 for tko^25t flies, n = 22 for tko^25t;UAS-TrpA1 flies, and n = 32 for tko^25t;Tdc2-Gal4>UAS-TrpA1 flies. One-way ANOVA with Tukey’s multiple comparisons test was used. Data are presented as mean values ± SEM. Source data are provided as a Source Data file.

Fig. 4. A Gal4 screen reveals activation sleep-promoting circuits worsens seizures.

a Experimental protocol depicting flies in the tko^25t background were raised at 18 °C, placed at 25 °C for 4 min for TrpA1 activation, vortexed for seizure induction, and video recorded for seizures at 25 °C. Atonic, tonic/clonic, and recovery phases of seizures were quantified. To drive TrpA1, Gal4 drivers were selected for peptidergic (C929-Gal4), serotonergic/dopaminergic (Ddc-Gal4), octopaminergic (Tdc2-Gal4), dopaminergic (TH-Gal4), serotonergic (trh-Gal4), dorsal fan-shaped body (104y-Gal4), wake-promoting (11H05-Gal4, 60D04-Gal4, c584-Gal4), mushroom body (201y-Gal4), pars intercerebralis (kurs58-Gal4), and ellipsoid body (R58H05-Gal4) neurons. Created in BioRender. Sehgal, A. (2025) https://BioRender.com/pt2kiq2. b, c Tonic/clonic times were prolonged upon driving of TrpA1 with 104y-Gal4, 201y-Gal4, kurs58-Gal4, and improved with c584-Gal4. d, e Total seizure times demonstrate worsening of seizures upon driving of TrpA1 with 104y-Gal4, 201y-Gal4, kurs58-Gal4, R58H05-Gal4, and trh-Gal4. c, e The x values represent experimental seizure durations minus the seizure durations in the UAS-TrpA1 controls. The y values represent experimental seizure durations minus seizure durations in the Gal4 genetic controls. Significant values (p<0.05) values are red with an asterisk. The number flies assessed for each condition is given in (): tko^25t (52), TrpA1 (52), C929 (4), C929>TrpA1 (23), ddc (12), ddc>TrpA1 (26), tdc (17), tdc>TrpA1 (69), TH (19), TH>TrpA1 (39), trh (12), trh>TrpA1 (68), 104 y (4), 104 y>TrpA1 (17), 11H05 (22), 11H05>TrpA1 (51), 201 y (12), 201 y>TrpA1 (20), 60D04 (12), 60D04>TrpA1 (28), c584>TrpA1 (8), kurs58 (41), kurs58>TrpA1 (77), R58H05 (5), and R58H05>TrpA1 (20). Two-group two-tailed t-test. Data are presented as mean values ± SEM. Source data are provided as a Source Data file.

Fig. 5. Activity is increased across whole brain, including sleep-promoting circuits, after sleep restriction.

Brains dissected from nsyb-Gal4>CaLexA flies treated with vehicle or caffeine. a Representative images in the anterior, medial, and posterior brain revealing the mushroom body (MB), ellipsoid body (EB), and dorsal fan-shaped body (dFSB). White lines outline the region of interest used for subregion quantification. Images are pseudocolored with the “fire” lookup table with warmer colors indicating increased intensity. b–e Sleep restriction with caffeine in nsyb-Gal4>CaLexA flies leads to increased whole brain, MB, EB, and dFSB CaLexA (GFP:RFP) signal. n = 10-21 flies/condition. Two-group two-tailed t-test. Data are presented as mean values ± SEM. Source data are provided as a Source Data file.

Fig. 6. Inhibition of sleep-promoting circuits protects against seizures in the setting of sleep loss.

a Experimental protocol showing flies were raised in darkness, entrained in red light, then placed into 24- or 48-well plates for chronic video monitoring with green light stimulation. Fly positions over time were converted into XY coordinates, and a “CynthiSeize” algorithm was developed to identify seizures. Created in BioRender. Sehgal, A. (2025) https://BioRender.com/9siy0zu. b, d, e Sleep quantification after 23E10-Gal4, 201y-Gal4 > GtACR1 activation. Caffeine and 23E10-Gal4, 201y-Gal4 > GtACR1 decrease sleep. n = 34 flies/condition. c, f, g Seizure quantification after 23E10-Gal4, 201y-Gal4 > GtACR1 activation. Caffeine increases seizure frequency and this is blocked with 23E10-Gal4, 201y-Gal4 > GtACR1 activation. n = 34 flies/condition. One-way ANOVA with Dunnett’s T3 multiple comparisons adjustment, Kruskal-Wallis test with Dunn’s multiple comparisons adjustment, negative binomial model with Wald test (for daily number of seizures), or mixed effects model (for seizure durations) was used. Data are presented as mean values ± SEM. Source data are provided as a Source Data file.

Fig. 7. 5HT1A is downregulated after sleep loss, and this is sleep-promoting.

a Transcriptomic analysis of the dorsal fan-shaped body after sleep-restriction reveals downregulation of 5HT1A, but not other neurotransmitter receptors or rate-limiting enzymes. N = 3 biological replicates. b, d Sleep quantification after addition of 8-OH-DPAT, a selective 5HT1A receptor agonist, reveals decreased sleep, consistent with the inhibitory effects of 5HT1A on the activity of sleep-promoting centers. Conversely, 23E10-Gal4, 201y-Gal4 > 5HT1A RNAi-mediated downregulation of 5HT1A in sleep-promoting centers leads to increased sleep. n = 46 flies/condition. e The effect of a 5HT1A agonist on daily sleep loss is decreased after downregulation of 5HT1A in sleep-promoting centers. n = 7 biological replicates. c, f, g Seizure severity is reduced after both 5HT1A agonism and 5HT1A downregulation. For daily seizure quantification, n = 36 flies/condition. For seizure duration, n = 76 seizures for ‘23E10, 201 y + veh’, n = 29 seizures for ‘23E10, 201 y + 8OHDPAT’, n = 30 seizures for ‘23E10, 201 y > 5HT1A RNAi + veh’, and n = 6 seizures for ‘23E10, 201 y > 5HT1A RNAi + 8OHDPAT’. One-way ANOVA with Dunnett’s T3 multiple comparisons test, paired two-tailed t-test, negative binomial model with Wald test (for daily number of seizures), or mixed effects model (for seizure durations) was used. Data are presented as mean values ± SEM. Source data are provided as a Source Data file.

Fig. 8. Increasing 5HT1A activity with the FDA-approved drug buspirone can protect against seizures induced by sleep loss.

a, c, d Sleep quantification after administration of caffeine demonstrates decreased sleep, but no significant change with administration of buspirone, an FDA-approved 5HT1A agonist. n = 36 flies/condition. b, e, f Seizure severity is increased with caffeine, but reduced to baseline levels with co- administration of buspirone. n = 36 flies/condition. One-way ANOVA with Dunnett’s T3 multiple comparisons test, paired two-tailed t-test, negative binomial model with Wald test (for daily number of seizures), or mixed effects model (for seizure durations) was used. Data are presented as mean values ± SEM.

Fig. 9. Schematic depicting how homeostatic sleep drive drives increased seizure risk.

Sleep drive leads to increased activity of sleep-promoting circuits, and we find that this, in part, is driven by decreased expression of 5HT1A in the dorsal fan-shaped body (dFB). Increased activity of sleep-promoting circuits then leads to increased sleep drive, which allows flies to reach the required amount of sleep. This process is known as sleep homeostasis. A “price” of sleep homeostasis is that sustained activation sleep-promoting circuits leads to increased “sleepiness” and increased seizure activity. On the other hand, if the activity of sleep-promoting circuits is decreased (e.g., experimentally with GtACR1 activation or pharmacologically with 5HT1A agonism), seizures are less likely to occur.