Temporal and Potential Predictive Relationships between Sleep Spindle Density and Spike-and-Wave Discharges

Spike-and-wave discharges (SWDs) and sleep spindles are characteristic electroencephalographic (EEG) hallmarks of absence seizures and nonrapid eye movement sleep, respectively. They are commonly generated by the cortico–thalamo–cortical network including the thalamic reticular nucleus (TRN). It has been reported that SWD development is accompanied by a decrease in sleep spindle density in absence seizure patients and animal models. However, whether the decrease in sleep spindle density precedes, coincides with, or follows, the SWD development remains unknown. To clarify this, we exploited Pvalb-tetracycline transactivator (tTA)::tetO-ArchT (PV-ArchT) double-transgenic mouse, which can induce an absence seizure phenotype in a time-controllable manner by expressing ArchT in PV neurons of the TRN. In these mice, EEG recordings demonstrated that a decrease in sleep spindle density occurred 1 week before the onset of typical SWDs, with the expression of ArchT. To confirm such temporal relationship observed in these genetic model mice, we used a gamma-butyrolactone (GBL) pharmacological model of SWDs. Prior to GBL administration, we administered caffeine to wild-type mice for 3 consecutive days to induce a decrease in sleep spindle density. We then administered low-dose GBL, which cannot induce SWDs in normally conditioned mice but led to the occurrence of SWDs in caffeine-conditioned mice. These findings indicate a temporal relationship in which the decrease in sleep spindle density consistently precedes SWD development. Furthermore, the decrease in sleep spindle activity may have a role in facilitating the development of SWDs. Our findings suggest that sleep spindle reductions could serve as early indicators of seizure susceptibility.


Introduction
The relationship between spike-and-wave discharges (SWDs), an electroencephalographic (EEG) hallmark of absence seizures, and sleep spindles, a prominent characteristic of nonrapid eye movement (NREM) sleep, has been a topic of interest in the field of epilepsy.Several studies have investigated the structural relationship between these two types of oscillations (Gloor, 1968;Kostopoulos et al., 1981;Sitnikova, 2010;Kozák et al., 2020).These oscillations are believed to originate from the same cortico-thalamo-cortical (CTC) networks, which include the corticothalamic and thalamocortical circuits and the thalamic reticular nucleus (TRN) (Steriade et al., 1993;Avoli, 2012;Lüthi, 2014;Fernandez and Lüthi, 2020).Previous studies have indicated a link between SWD development and alteration in sleep spindle patterns.Specifically, children with SWDs exhibited lower sleep spindle density during NREM sleep compared with healthy children (Zhang et al., 2023).Similar findings have been observed in Wistar Albino Glaxo Rijswijk (WAG/Rij) rats, a validated genetic animal model of absence seizures, and in some Long-Evans rats (Akman et al., 2010;Sitnikova et al., 2014;Kozák et al., 2020;Sitnikova, 2021).Studies on these rats have highlighted an age-dependent increase in epileptic activity alongside changes in the intrinsic dynamics of sleep spindle patterns.These investigations suggested that the occurrence of SWDs is associated with a reduction in sleep spindle density.However, the temporal relationship-whether the decrease in sleep spindle density precedes, coincides with, or follows the onset of SWDs-remains unclear.
This study aimed to investigate this temporal relationship between the development of SWDs and sleep spindle density.Specifically, it aimed to determine whether a decrease in sleep spindle density precedes or follows the onset of SWDs.To achieve this, a doubletransgenic mouse [Pvalb-tetracycline transactivator (tTA)::tetO-ArchT] was used, which selectively expresses ArchT, an inhibitory opsin, in parvalbumin-positive (PV) cells in the TRN.The expression of ArchT in TRN-PV is sufficient to induce spontaneous, cortical 7-11 Hz SWDs (Abdelaal et al., 2022).Previous findings indicated that the PV-ArchT mice fulfilled the face and pharmacological validities of the two existing genetic rat models of absence seizures: WAG/Rij and Genetic Absence Epilepsy Rat from Strasbourg (GAERS) rats (Akman et al., 2010;Kandratavicius et al., 2014;Abdelaal et al., 2022).A unique aspect of the PV-ArchT model is its ability to consistently and controllably induce SWDs through the regulated expression of ArchT, dependent on the presence or absence of doxycycline (DOX) in the diet.In this study, a DOX-controllable approach was employed in PV-ArchT mice to investigate the temporal relationship between the development of SWDs and alterations in sleep spindle patterns.
SWDs in rat models have a similar pharmacological profile as humans.Drugs that enhance GABAergic inhibition have been found to exacerbate SWDs and decrease sleep spindles, while drugs that promote sleep spindles have been shown to decrease SWDs (Hirshkowitz et al., 1982;Van Luijtelaar, 1997;Sitnikova and van Luijtelaar, 2005).However, the link between reduced sleep spindle and the likelihood of SWD development is not yet fully understood.Therefore, a pharmacological mouse model was used to confirm the temporal relationship between sleep spindle alteration and SWD development and to explore the role of sleep spindle reduction on SWD development.We observed that administering a low dose (50 mg/kg) of gamma-butyrolactone (GBL), which is known to be insufficient to induce SWDs on its own, induced SWDs in wild-type mice whose sleep spindle density had been reduced by repeated prior caffeine treatment (Ishige et al., 1996;Venzi et al., 2015).These confirm the temporal relationship that sleep spindle reduction plays a predictive role in the development of SWDs.Also, sleep spindle reduction may have a role in facilitating the development of SWDs.Our findings suggest that sleep spindle alterations could serve as early indicators of absence seizures.

Ethics statement
All animal procedures were conducted according to the National Institutes of Health Guide for the Care and Use of Laboratory Animals and approved by the Animal Research Committee of Keio University School of Medicine.

Stereotaxic surgery
EEG/electromyography (EMG) electrode implantation was performed as described in a previous study (Abdelaal et al., 2022).For anesthesia, mice received a mixture of ketamine and xylazine (100 and 10 mg/kg, respectively) intraperitoneally.The cortical EEG electrodes were located above the parietal area (anteroposterior −1.5 mm, mediolateral +2.0 mm from the bregma).For EMG recording, silver wires (AS633; Cooner Wire) were inserted bilaterally into the trapezius muscles.The mice were allowed to recover for 1 week before recording.

Pharmacological intervention
Ethosuximide (Tokyo Chemical Industry) was dissolved in normal saline and injected intraperitoneally into mice at a dose of 100 mg/kg.GBL (Sigma-Aldrich) was mixed with normal saline and injected intraperitoneally into mice at 50 and 100 mg/kg.Caffeine (Sigma-Aldrich) was dissolved in normal saline and injected intraperitoneally into mice at 12 mg/kg.

EEG/EMG recordings
EEG/EMG recordings were performed on ad libitum moving mice in their home cages placed in a soundproof box.During the DOX-off or pharmacological experimental period, mice were repeatedly connected to EEG/EMG cables and monitored in their home cages for >24 h at a time, including both 12 h light and dark phases.EEG/EMG recordings were consistently synchronized with video monitoring.To acclimate mice to the recording procedures, they underwent three separate habituation sessions for EEG/EMG recordings prior to the commencement of the experiment.
The EEG and EMG signals were amplified 500 times and bandpass-filtered (0.1-1,000 Hz for EEG; 10-1,000 Hz for EMG; Model 3000, A-M Systems; DAM50, World Precision Instruments).The analysis was performed using in-house software developed in MATLAB R2020 (MathWorks).The EEG signals were subjected to Fast Fourier transform analysis from 1 to 60 Hz.
Detection of typical SWDs in PV-ArchT mice.SWDs were visually identified using EEG/EMG and video recordings.A previous study (Abdelaal et al., 2022) described the criteria for SWD events in PV-ArchT mice as repetitive spike-and-wave complexes associated with behavioral arrest lasting >0.5 s and ranging from 7 to 10 Hz.The average power of SWDs was double that of the EEG background.
Detection of other epileptiform discharges in PV-ArchT mice.Epileptiform discharges other than typical SWDs were visually identified from the EEG/EMG data based on specific criteria: spikes (sharp deflections) and spike-wave complexes that did not exhibit the typical features characteristic of SWDs (Abdelaal et al., 2022;Fig. 1B).
Detection of SWDs in GBL-treated mice.Since GBL-treated animals are known to display SWDs with lower frequencies compared with other animal models with robust behavioral arrest (Ishige et al., 1996), we used the following criteria for the SWD detection: repetitive spike-and-wave complexes associated with behavioral arrest lasting >0.5 s and ranging from 4 to 6 Hz (Ban et al., 1967;Venzi et al., 2015).
Detection of behavioral arrest during SWDs.Epileptic behavioral arrest was identified through simultaneous EEG/EMG and video monitoring.During the appearance of SWDs in the EEG, the mice abruptly and temporarily ceased their ongoing behavior, accompanied by very low EMG amplitudes.Upon termination of the SWDs, the mice's behavior resumed to normal, along with normal EMG amplitudes.
Detection of wake-sleep stages.EEG/EMG recordings with simultaneous video monitoring were conducted continuously for >24 h at a time.Sleep-wakefulness stages-wakefulness, NREM sleep, and REM sleep-were categorized from the EEG/EMG recordings through off-line analysis of 1 s epochs, following established methodologies described in previous studies (Ahnaou and Drinkenburg, 2011;Funato et al., 2016;Kato et al., 2022;Rayan et al., 2022).The EEG power spectra were averaged over the three frequency bands: delta (0.5-4.0 Hz), theta (6-9 Hz), and sigma (10-16 Hz).Wakefulness was scored based on low amplitude, fast EEG activity, and high-amplitude EMG signals.NREM sleep was identified based on high-power delta band and low-amplitude EMG signals.REM sleep was identified based on highpower theta band (6-9 Hz) signals and EMG atonia.The duration of each sleep-wake stage was aggregated every 3 h and utilized for the 24 h timeline assessment, encompassing both the 12 h light and dark phases.
Detection of sleep spindles.For sleep spindle analysis during NREM sleep, NREM sleep periods were identified in each mouse beforehand, as described previously.For each mouse, NREM sleep periods of 30 min were extracted from 2 to 4 h recordings, during which sleep spindles were detected and used for analysis.The sleep spindle detection method was based on previous studies (Fernandez and Lüthi, 2020;Osorio-Forero et al., 2021).That is, the raw data were filtered from the cortical EEG signal using a sigma band (10-16 Hz) filter, squared the resulting signal, and applied a threshold of 1.5 times the standard deviation above the mean values observed during NREM sleep.All peaks crossing this threshold were detected, and the endpoints of the events were extended to the nearest zero-crossing points before and after the threshold.Once sleep spindles were detected in NREM sleep periods, the following features were extracted: amplitude, which corresponds to the maximum value of the filtered signal within the spindle events; density, which is defined as the number of spindles divided by the unit of time; and duration, which is calculated as the time between the beginning and end of the spindle event (Table 1).

Statistical analysis
All experiments were analyzed using MATLAB R2020 (MathWorks), Excel (Microsoft 365), and IBM SPSS Statistics 23.Mean values are reported for all numerical data, with error bars showing the standard error of the mean unless specified otherwise.Parametric tests, such as paired and two-tailed Student's t tests, one-way analysis of variance (ANOVA), and repeated-measure ANOVA followed by a Tukey-Kramer post hoc test, were used for the analysis.A p value of <0.05 was considered statistically significant (*p ≤ 0.05, **p ≤ 0.01, and ***p ≤ 0.001).

Data availability
The corresponding author can provide the datasets upon reasonable request to support this study.

PV-ArchT mice exhibited SWDs during both wakefulness and sleep stages
The PV-ArchT mice were previously reported to display an absence seizure phenotype and fulfilled the criteria for face and pharmacological validities (Abdelaal et al., 2022).When the PV-ArchT mice were fed normal food (without DOX; Fig. 1A), the inhibitory opsin ArchT was highly expressed in the PV neurons of the TRN.Adult PV-ArchT mice fed normal food from birth exhibited cortical SWDs and other epileptiform discharges (Fig. 1B).The typical SWDs were characterized by 7-10 Hz repetitive activity lasting >0.5 s (Table 1).The power of SWDs exceeded that of the EEG background (n = 5 mice; 7.3 ± 1.0 vs 3.5 ± 0.7 mV 2 ; t (4) = 6.2; p = 0.002).With age, both the duration and frequency of SWDs increased (Abdelaal et al., 2022).In this study, the distribution of SWD events during both awake and sleep stages was examined in PV-ArchT mice fed normal food (Fig. 1C).Vigilance stages were classified into wakefulness, NREM sleep, and REM sleep, and the number of SWDs in each stage was quantified according to the criteria described in the Materials and Methods section (Table 1).SWDs predominantly occurred during wakefulness and the transition from awake to NREM sleep stages.The occurrence of SWD events was 64.3% during the awake stage, 31.2% during the transition from awake to NREM sleep (categorized as NREM sleep), and 4.51% during REM sleep.The frequency of SWDs during wake, NREM sleep, and REM sleep remained consistent at 7-10 Hz.However, the duration of SWDs during REM sleep was shorter than that during wakefulness or NREM sleep (wake, 2.9 ± 0.7 s; NREM sleep, 3.0 ± 0.3 s; REM sleep, 1.9 ± 0.5 s).Furthermore, the number of SWDs increased significantly with age in the awake and NREM sleep stages (Fig. 2A, Table 2).
Next, whether PV-ArchT mice exhibited changes in the incidence of SWDs during the light/dark phases was investigated.EEG/EMG signals were recorded for 24 h, and the number of SWDs during the awake, NREM sleep, and REM sleep stages was quantified.No significant diurnal change was observed in the number of SWDs (Fig. 2B).Additionally, whether the SWD development in PV-ArchT mice caused changes in sleep duration was investigated, as it is well known that seizures disrupt sleep regulation (Sitnikova, 2021;Lehner et al., 2022), and 24 h EEG/EMG recordings were used to quantify the periods of wakefulness, NREM sleep, and REM sleep.The diurnal variations in wakefulness, NREM sleep, and REM sleep duration in PV-ArchT mice were not different from those in control mice (Fig. 2C).These results indicate that SWDs in PV-ArchT mice occurred in all sleep-wake stages regardless of light/dark phases.

SWD development is associated with sleep spindle alterations in PV-ArchT mice
The CTC network, including TRN neurons, has been implicated in the development of SWDs, as well as in physiological oscillations like sleep spindles (Steriade et al., 1993;Snead, 1995;McCormick and Bal, 1997;Pinault and O'Brien, 2005).This study aimed to investigate whether SWDs are linked to sleep spindle dysfunction, given their established connection.Specifically, the alteration of sleep spindle patterns was investigated in PV-ArchT mice during the development of SWDs compared with that in control mice.Our control group consisted of PV-ArchT mice fed with DOX-containing food from birth (Fig. 3A).Our results revealed a reduction in sleep spindle density in PV-ArchT mice compared with the control group.Specifically, when examining changes in sleep spindle density between Postnatal Day (P)60 and P120, a decline was observed (Fig. 3B, Table 3).However, the duration and power of the sleep spindle did not differ from those of the control group and remained unchanged with age (Fig. 3B, Table 3).

Chronological order of reduction in sleep spindle density and SWD onset
As the presence of SWDs was associated with a decrease in sleep spindle density, we aimed to determine whether the decrease in sleep spindle density precedes or follows the emergence of SWDs.To assess temporal changes, EEG/EMG monitoring of PV-ArchT mice before and after the withdrawal of DOX-containing food (DOX-off) was conducted.This manipulation induced the expression of ArchT in PV-TRN neurons, resulting in the induction of SWDs (Abdelaal et al., 2022).By initially providing DOX-containing food to PV-ArchT mice from birth and subsequently switching to normal food (Fig. 4A), the SWD's development with EEG/EMG/video monitoring over time was evaluated.The occurrence of typical SWDs, epileptiform discharges other than SWDs, and sleep spindle alterations before and after DOX discontinuation was examined.The same parameters were used to quantify the SWDs, as previously described in the Materials and Methods section.Initially, during Week 2 after DOX discontinuation, epileptiform discharges instead of the typical SWDs that occurred during the NREM sleep was observed.By Week 3, these epileptiform discharges had spread to the awake stage (Fig. 4B).It was only during Week 3 that typical SWDs were detected (Fig. 4C).Over time, the number of epileptiform discharges and typical SWD events increased with high ArchT expression (Fig. 4B,C; Tables 4, 5, respectively).Notably, sleep spindle density began to decrease in Week 2 after DOX discontinuation, which coincided with the occurrence of epileptiform discharges and preceded the onset of typical SWDs in Week 3 (Fig. 4D).However, from Weeks 2 to 6, sleep spindle density remained consistently reduced without further decrease (Fig. 4D, Table 6).Additionally, no significant changes in the duration of sleep spindles before and after discontinuing DOX were noted (Fig. 4D).These findings suggest a specific temporal relationship in which a reduction in sleep spindle density precedes the onset of typical SWDs.

Temporal relationship between sleep spindles and SWDs in pharmacological absence seizure models
Next, whether the temporal relationship between SWD development and sleep spindle alterations observed in our genetic model of absence seizures (PV-ArchT) could also be observed in other pharmacological models was investigated.GBL was first administered to wild-type mice at a dose of 100 mg/kg during the light phase (Hodor et al., 2015;Venzi et al., 2015; Fig. 5A), which successfully induced SWDs with a frequency of 4-6 Hz (n = 4 mice; SWD events, 73.4 ± 6.82 SWDs/h; p = 0.001; Fig. 5B).The SWDs began to appear 4.2 ± 0.5 min after the injection of GBL.The average duration of these SWDs was 3.3 ± 0.1 s.Although the GBL has sedative/hypnotic effects, a dose below 200 mg/kg does not significantly increase sleep time during the light phase; it only increases the power of slow waves (Van Sassenbroeck et al., 2001;Felmlee et al., 2010;Spano et al., 2022).The changes in sleep spindle patterns during NREM sleep before and after the GBL injection were examined.A significant reduction was noted in sleep spindle density during the first 4 h after the GBL administration (n = 4 mice; density, 11.6 ± 0.59 vs 6.98 ± 0.73 spindle events/min; t (3) = 3.82; p = 0.01; Fig. 5C).No changes were observed in the power and duration of sleep spindles before and after the GBL injection (n = 4 mice; duration, 0.37 ± 0.04 vs 0.43 ± 0.03 s; t (3) = −1.89;p = 0.08; power, 4.26 ± 0.05 vs 4.15 ± 0.13 mV 2 ; t (3) = 0.76; p = 0.25; Fig. 5C).These results are consistent with the findings that in PV-ArchT mice, the SWD development was associated with a decrease in sleep spindle density.
Then, caffeine at a dose of 12 mg/kg was repeatedly administered to wild-type mice to induce a reduction in sleep spindles (Fig. 6C).This dose is known to promote wakefulness and decrease sleep duration in mice (Lazarus et al., 2011).On Day 1, the mice were unable to sleep after the caffeine injection at 9:00 A.M. in the light phase.However, after injecting caffeine for 3 consecutive days, the mice were able to sleep even after caffeine administration.

Discussion
This study provides insights into the temporal relationship between sleep spindle alterations and the occurrence of SWDs in both genetic and pharmacological animal models of absence seizures: a reduction in sleep spindle density preceded the occurrence of SWDs.This suggests that the changes in the sleep spindle may have potential predictive value for the occurrence of SWDs.Furthermore, these changes may also play a role in the development of SWDs.

Distribution of SWDs across wake-sleep stages
The occurrence of SWDs demonstrates a clear link with alertness levels, occurring more frequently during wakefulness and NREM sleep in absence seizure patients (Posner, 2013).SWDs are commonly observed during the N2 stage of NREM sleep.These events are characterized by a 3 Hz frequency and a duration of >2 s, similar to patterns observed during wakefulness, but often display greater disorganization (Chen et al., 2022).The occurrence of SWDs during sleep, especially in children, may indicate drug resistance or pose diagnostic challenges (Sadleir et al., 2011).Notably, SWDs are rarely detected during REM sleep.However, clinical studies on SWDs during sleep are limited, prompting researchers to rely on animal models.The distribution of SWDs in relation to the vigilance stage has been studied in genetic rat models, such as WAG/Rij rats and GAERS, feline models, and other mouse models that present spontaneous or paroxysmal SWDs.In these animals, SWDs were most prevalent during wakefulness and NREM sleep, with minimal occurrences during REM sleep (Lannes et al., 1988;Marescaux et al., 1992;Sitnikova et al., 2014;Funato et al., 2016;Sitnikova, 2021).Our PV-ArchT mice shared similarities with the previous models.In the PV-ArchT mice, SWDs were observed across all vigilance stages but were more frequent during wakefulness and transition from awake to NREM sleep.SWDs were rarely observed during REM sleep.SWDs have a different appearance from theta waves during phasic REM, characterized by increased frequency and amplitude, in that SWDs have spikes.Although they are easily distinguished visually, it might be difficult to completely exclude phasic REM from SWDs during REM sleep.Phasic REM sleep is typically identified by measuring REMs (Simor et al., 2020).However, simultaneous recording of eye movements in mice is technically challenging during continuous 24 h freely moving monitoring.Future studies could benefit from advanced recording techniques or algorithms to perfectly differentiate between these states.In summary, SWDs in both human and animal models exhibit vigilance-dependent characteristics and are influenced by the wake-sleep cycle.

Distinct TRN activity involved in sleep spindle and SWD generation
The TRN, within the CTC network, regulates sleep spindles and SWDs through two distinct firing patterns: tonic and burst firing (Steriade et al., 1993;McCormick and Bal, 1997;Schönauer and Pöhlchen, 2018).These firing activities in the TRN are triggered by T-type Ca 2+ channels, specifically the Cav3.2 and Cav3.3 isoforms.Notably, Cav3.3 and its Table 3. Results of the average sleep spindle patterns (density, duration, and power, respectively), with age in PV-ArchT and control mice for the data in Figure 3B Density Duration Power P60 6.44 ± 0.55 vs 11.9 ± 1.84 0.39 ± 0.02 vs 0.37 ± 0.002 3.39 ± 0.42 vs 3 ± 1.03 p = 0.01** p = 0.33 p = 0.68 P90 6.20 ± 0.27 vs 12.0 ± 1.97 0.43 ± 0.03 vs 0.38 ± 0.03 3.77 ± 0.38 vs 3.13 ± 0.82 p = 0.01** p = 0.24 p = 0.44 P120 4.94 ± 0.33 vs 10.9 ± 2.26 0.36 ± 0.009 vs 0. associated burst firing, which are essential for sleep spindle initiation, may not be critical for SWD development (Crunelli and Leresche, 2002;Halassa et al., 2011).In Cav3.2 and Cav3.3 double-knock-out mice, the complete loss of burst firing and increase in tonic firing suppressed sleep spindles and induced SWD following systemic GBL injection (J. Lee et al., 2013;S. E. Lee et al., 2014;Lory et al., 2020).Based on these findings, Lee's group proposed that high tonic firing in 2,13) = 0.12; p = 0.89).Note that, for Days 0 and 3, data from NREM sleep between 9:00 and 13:00 were used for analysis, whereas for Day 4, data monogenic mutant models of absence seizures, such as stargazer, tottering, and lethargic, is lacking.The only available data on this relationship come from the WAG/Rij rat model and Long-Evans rats, which showed that the development of SWDs is associated with changes in sleep spindle patterns (Leresche et al., 2012;Sitnikova et al., 2014Sitnikova et al., , 2023;;Kozák et al., 2020).These previous data support the idea that the development of SWDs is associated with a reduction in sleep spindle density.Building on the findings from PV-ArchT mice, this study contributes a novel perspective by uncovering the temporal relationship between SWD development and changes in sleep spindles.A decrease in sleep spindle density was noted 1 week before the onset of SWDs.These findings suggest that identifying absence seizures involves not only recognizing pathological oscillations such as spikes followed by waves but also noting the absence of physiological oscillations.Moreover, in the PV-ArchT mouse model, a chronological sequence of typical SWDs and other epileptiform discharges was observed.Our findings indicate that these other epileptiform discharges appear 1 week before the onset of typical SWDs, providing two significant insights.First, the occurrence of these epileptiform discharges before the SWD development could aid in early diagnosis and treatment.Second, investigating whether these epileptiform discharges and SWDs correlate with the level of ArchT expression in PV-TRN could elucidate whether they stem from abnormalities in the CTC network, including TRN neurons.This approach could deepen our understanding of the pathophysiology of absence seizures and illuminate the underlying mechanisms of these events.
The role of sleep spindle in the SWD dynamics (pharmacological insights) Acute pharmacological experiments have been performed to explore the reciprocal connection between sleep spindles and SWDs in WAG/Rij rats (Van Luijtelaar, 1997;Sitnikova and van Luijtelaar, 2005).For instance, barbiturates and benzodiazepines have been shown to enhance sleep spindles and concurrently suppress SWDs.Conversely, clonidine has been shown to aggravate SWDs in a dose-dependent manner and has been associated with a reduction in sleep spindles (Hirshkowitz et al., 1982;Van Luijtelaar, 1997;Sitnikova and van Luijtelaar, 2005).These findings provide evidence suggesting a shared involvement of CTC circuits in both sleep spindles and SWDs.The association between decreased sleep spindle density and the occurrence of SWDs was investigated.Notably, SWDs were induced in mice with lower-than-normal sleep spindle density using low-dose GBL.Previous studies have confirmed that the induction of SWDs by GBL is dose-dependent, with doses below 70 mg/kg being insufficient to induce SWDs in mice.Our observations indicate that mice with reduced sleep spindle density due to caffeine administration exhibited SWD upon being injected with a non-SWD-inducing dose of GBL (Ishige et al., 1996;Venzi et al., 2015).These findings underscore the importance of sleep spindles in SWD dynamics.However, there are limitations to our approach of using pharmacologically induced SWDs and alterations in the sleep spindles.First, we did not use a control group with saline injections.Even saline injections can affect sleep architecture and spindle activity due to the restraint stress with injection.However, we have not from NREM sleep between 11:00 and 13:00 were used for analysis.Repeated-measure ANOVA followed by the Tukey-Kramer post hoc test, and the results of statistical tests are shown in Tables 7 and 8  addressed this possibility in the present study.Second, caffeine exhibits dual effects on sleep spindles and overall sleep duration, complicating the interpretation of our results.While our study highlights a temporal relationship between reduced sleep spindle density and SWD onset, whether this reduction is the primary driver of SWDs or if changes in sleep duration also play a significant role cannot be definitively concluded.Third, the artificial induction of SWDs using GBL may not fully replicate the complexity of naturally occurring SWDs.Drug-drug interactions, such as those between caffeine and GBL, could also introduce confounding factors, potentially influencing SWD generation.Finally, our study primarily focused on short-term effects, and the long-term consequences or reversibility of changes in sleep spindle density following GBL administration were not explored.Future research should incorporate more naturalistic models and delve into the underlying neurobiological mechanisms.Direct manipulation of sleep spindles, such as through optogenetic techniques to trigger or suppress sleep spindles, could provide conclusive evidence on the prevention or induction of SWDs.Such investigations will be crucial for comprehensively understanding the bidirectional relationship between SWD development and sleep spindle density.Enhanced understanding of the role of sleep spindle in SWD developments in the CTC circuit will help improve the treatment of absence seizure.Our finding of the crucial temporal link between sleep spindle reduction and the onset of SWDs paves the way for new approaches to early detection and intervention in absence seizures.

Figure 1 .
Figure 1.Detection of SWDs during awake-sleep stages in PV-ArchT mice.A, Left, Schematic representation of PV-tTA::tetO-ArchT transgenic mice fed with normal food from birth; right, a timeline of ArchT expression and EEG/EMG recordings.B, Types of epileptiform discharges in PV-ArchT mice.C, Example of EEG recording on P120.From top to bottom, Cortical EEG (red and blue traces illustrate typical SWDs and other epileptiform discharges, respectively), EEG spectrogram, EMG, hypnogram, and powers of the delta, theta, and sigma bands.

Figure 3 .
Figure 3. Relationship between sleep spindle alteration and SWD development in PV-ArchT mice.A, Method for detecting sleep spindles in PV-ArchT (left) and control (right, PV-ArchT mice fed with DOX-containing food from birth) mice.From top to bottom, Raw EEG data recording, filtered sigma signals, wavelet analysis, and spindle power.The red lines represent the threshold value.Red-shaded areas indicate sleep spindle events.B, Summary graphs from left to right showing the sleep spindle patterns (density, duration, and power, respectively), comparing PV-ArchT mice (n = 5 mice for each group) with control mice (n = 3 mice for each group; PV-ArchT mice fed with DOX-containing food all life were used as controls).The results of statistical tests are shown in Table 3. C, Summary graphs from left to right show the number of SWDs/h and the sleep spindle patterns (density, duration, and power) before and after ethosuximide injection (n = 5 mice).
PV-ArchT mice fed with DOX-containing food from birth; n = 3 mice.PV-ArchT mice fed with normal food from birth; n = 5 mice.The bold values indicate statistically significant p values.**p ≤ 0.01.

Figure 5 .
Figure 5. Drug-induced SWDs are associated with a reduction in sleep spindle density.A, Timeline illustrating the administration of GBL (100 mg/kg) to wild-type mice.B, Example of an EEG recording before (left) and after (right) GBL injection.From top to bottom, Cortical EEG, EEG spectrogram, and EMG.Red traces indicate SWDs.C, Summary graphs from left to right show the number of SWDs/h and the sleep spindle patterns (density, duration, and power) before and after GBL administration (n = 4 mice).

Figure 6 .
Figure 6.Temporal relationship between sleep spindle and SWDs in the GBL model of absence seizure.A, Timeline illustrating the administration of GBL (50 mg/kg) to wild-type mice.B, Summary graphs from left to right show the number of SWDs/h and the sleep spindle patterns (density, duration, and power) before and after GBL administration (n = 4 mice).C, The panels illustrate the daily protocols for caffeine and GBL injections.D, Summary graphs from left to right show the time course of NREM sleep duration per hour (one-way ANOVA from 11:00 to 13:00,F(2,9) = 12.2; p = 0.002) and sleep spindle patterns (density, duration, and power) on Days 0, 3, and 4 (one-way ANOVA; density, F (2,13) = 20.3;p = 0.00009; duration, F (2,13) = 4.45; p = 0.33; power, F (2,13) = 0.12; p = 0.89).Note that, for Days 0 and 3, data from NREM sleep between 9:00 and 13:00 were used for analysis, whereas for Day 4, data . E, Left, Example of an EEG after GBL injection.From top to bottom, Cortical EEG, EEG spectrogram, and EMG.Red traces indicate SWDs.Right, Dot plot showing the number of SWDs before and after GBL injection (n = 6 mice).

Table 7 .NREM
Results of Tukey-Kramer post hoc test for the changes in time course of NREM sleep duration per hour from 11:00 to 13:00 for the data in Figure 6D indicate statistically significant p values.**p ≤ 0.01.Table 8. Results of Tukey-Kramer post hoc test for the changes in sleep spindle density on Days 0, 3, and 4 for the data caffeine (n = 6 mice), caffeine + GBL (n = 4 mice).The bold values indicate statistically significant p values.***p ≤ 0.001.

Table 1 .
Basic features of SWDs and sleep spindles

Table 2 .
Results of Tukey-Kramer post hoc test for the number of SWDs per hour with age during awake and NREM sleep for the data in Figure2A

Table 6 .
Results of Tukey-Kramer post hoc test for the changes in sleep spindle density for the data in Figure4D The bold values indicate statistically significant p values.***p ≤ 0.001.