Introduction

The thalamocortical (TC) network comprises interconnected TC neurons from the ventral-basal (VB) nuclear complex of the sensory thalamus, reticular thalamic nucleus (TRN), and somatosensory cortex. This circuitry is characterized by rhythmic oscillations that are important for normal sensory processing, attention, and transitions from sleep and awake states, but are also implicated in some pathologic processes in the brain, such as absence seizures1,2,3. The central medial nucleus (CMT) of the thalamus is part of the intralaminar nuclear complex and has a critical role in arousal/attention and cognition4 with diffuse projections to the anterior and posterior regions of the cerebral cortex and many subcortical structures5. In contrast to the VB relay nuclei, which receive topographically organized sensory input, the thalamic intralaminar complex receives diverse inputs from various intra- and extra-thalamic sources6.

Various parts of the TC circuitry are implicated in pathologic hyperexcitable states during absence seizures that are readily induced with pharmacologic agents such as gamma-butyrolactone (GBL). Absence seizures in rodents are typically characterized by an impairment of consciousness and freezing behavior, accompanied by the appearance of 3–5 Hz spike-wave discharges (SWDs) in encephalographic (EEG) recordings that result from paroxysmal and synchronized firing in TC networks7. Previous studies demonstrated that low-voltage activated (T-type) calcium channels are crucial for the oscillatory behavior of the network necessary for generating absence seizures8. Different isoforms of these channels, such as CaV3.1, CaV3.2, and CaV3.3, are expressed within the TC network9 and in the CMT10. In contrast, CaV2.3 R-type channels are densely expressed in the cortex11 and TRN, but not in TC neurons12. We recently demonstrated the abundant expression of functional CaV2.3 channels in the CMT13. The role of CMT CaV2.3 in SWDs, however, is unclear. Interestingly, studies in CaV2.3 global knockout (KO) mice have produced different findings regarding the role of these channels in pharmacologically induced SWDs. For example, Zaman and colleagues found that CaV2.3 channels actively promote and sustain rebound burst firing in TRN neurons by activating small-conductance calcium-activated potassium type 2 channels, which results in hyperpolarization of the cell membrane causing repriming of T-type channels, and mice lacking these channels exhibit a markedly decreased duration of GBL-induced SWDs14. On the other hand, Weiergräber’s group demonstrated that mice lacking CaV2.3 R-type channels are more susceptible to absence seizures, suggesting that these channels have a role in controlling and suppressing SWD activity and synchronized oscillations of TC networks, possibly by depolarizing TRN neurons, facilitating tonic firing mode, and preventing the cells from transitioning into rebound burst firing15.

To investigate the potential role of R-type channels in modulating SWD activity in the thalamus, we recorded local field (LFP) potentials from the CMT and EEG potentials from the cortex in control WT and CaV2.3 KO mice. In addition, we compared cumulative SWD durations and absence seizure architectures, as well as the phase synchrony between the cortex and CMT.

Results

Our experiments were based on in vivo EEG/LFP recordings in freely moving animals in a commonly used rodent model of absence seizures induced by systemic administration of the GABAB agonist GBL, a prodrug of γ-hydroxybutyrate16. Baseline data were recorded during the quiet (passive) awake state for 15 min before intraperitoneal (i.p.) injection of GBL, and for 45 min following the injection. Using power spectral density (PSD) and phase-locking value (PLV) analyses, we examined and compared the EEG/LFP signal characteristics and functional connectivity of the CMT and somatosensory cortex between the WT and KO mouse groups. We further examined the cumulative duration and latency of SWDs following GBL administration. All SWD episodes were accompanied by behavioral motor arrest.

Spike-wave discharge characteristics

To analyze the excitability of TC networks, we quantified individual and cumulative SWD durations from both the cortex and CMT in the first 30 min after GBL administration in WT and CaV2.3 KO mice. In in vivo recordings, the cumulative SWD duration was significantly decreased in both the cortex and CMT in CaV2.3 KO mice compared with the WT mice (Fig. 1). Representative traces from the cortical and CMT recordings in WT and CaV2.3 KO mice are depicted in Fig. 1a and b, respectively. The cumulative SWD duration in the cortex during the first 30 min after GBL administration was significantly decreased (~ 20%) in the CaV2.3 KO group compared with the WT group (Fig. 1c left). Further, the cumulative SWD duration in the CMT observed during the first 30 min after GBL administration was significantly decreased (~ 32%) in the CaV2.3 KO group compared with the WT group (Fig. 1c right). In contrast, the latencies in both the cortex and CMT were not significantly different between the WT and CaV2.3 KO groups (data not shown). As expected, cumulative duration of SWDs was not different when we compared cortex and CMT recordings in the WT and CaV2.3 KO cohorts separately (Fig. 1d). Representative traces of individual SWDs on extended time scale from the cortex (blue trace) and CMT (red trace) are depicted on Supplementary Fig. 1a. Interestingly, the average frequency during 30 min of recording following GBL injections and duration of individual SWDs in cortex and CMT was not different between WT and mutant mice (Supplementary Fig. 1b and c). However, the average duration of inter-ictal (non-SWD) period observed in the first 30 min after the i.p. injection of GBL from cortex was significantly increased in the CaV2.3 KO group compared to the WT group, but not in CMT group (Supplementary Fig. 1d).

Fig. 1
figure 1

CaV2.3 KO animals demonstrate significantly decreased duration of SWD in both cortex and CMT compared to the WT animals on EEG recordings in vivo. Representative EEG traces recorded simultaneously from CMT and Cortex in the WT (a) and CaV2.3 KO (b) mice. Arrowheads indicate the timepoint of GBL injection. Inset below panel (a) shows EEG trace from cortex in WT mouse post-GBL injecton on an extended time scale. C, left Cumulative SWD duration observed in the first 30 min after the i.p. injection of GBL from cortex was significantly decreased in the CaV2.3 KO group (n = 6) compared to the WT group (n = 6) by ~ 20% (*, p < 0.05, unpaired t-test). c, right Cumulative SWD duration observed in the first 30 min after the i.p. injection of GBL from CMT was significantly decreased in the CaV2.3 KO (n = 5) group compared to the WT group (n = 6) by ~ 32% (**, p < 0.01, unpaired t-test). Note that on the graphs of panel C red symbols denote WT animals, black symbols denote WT littermates. d, Cumulative SWD duration observed in the first 30 min after the i.p. injection of GBL from the cortex and CMT in the WT (d left) and CaV2.3 KO (d right) animals was not different (ns, p > 0.05, unpaired t-test).

Full size image

Power spectral density analyses

  • Spectral analysis is a standard method used to quantify EEG/LFP signals. The PSD reflects the distribution of signal power over frequency. We used total and relative power analyses of the EEG signal from the CMT and cortex in WT and CaV2.3 KO mice before and after GBL administration. In this pharmacologic model of absence seizures, GBL administration induces synchronous oscillations in the TC network, leading to rebound burst firing in the VB and TRN. This is reflected on EEG as high-amplitude SWDs (3–5 Hz), with power shifting predominantly to the delta (δ) 0.5–4 Hz and theta (θ) 4–8 Hz bands, which become the dominant frequencies. Hence, in this study, we focused on changes in the delta and theta bands. Representative heat maps with representative LFP traces recorded from the CMT are depicted for a WT mouse on (Fig. 2a) and a CaV2.3 KO mouse (Fig. 2b). Representative heat maps with representative EEG traces recorded from the cortex are depicted for a WT mouse on (Fig. 3a) and a CaV2.3 KO mouse (Fig. 3b). As expected, total power analyses in WT animals showed significantly increased power densities in both the CMT and cortex in the delta band (Figs. 2c and 3c) and theta band (Figs. 2c and 3c). These changes were also summarized in Supplemental Fig. 2. The power of the gamma (γ) > 30 Hz band in the cortex also significantly decreased in these animals. Relative power showed a large and significant shift toward the delta band in both the CMT and cortex (Figs. 2d and 3d), as well as an increase in the theta band in the CMT (Fig. 2d), and a concomitant decrease in relative powers of other frequency bands: alpha, beta, and gamma (Figs. 2d and 3d). In the CaV2.3 KO group, however, total power analyses demonstrated different changes compared with the control group. In the CMT, there were no significant changes in the delta band, while there were significant decreases in the theta, alpha, and gamma bands (Fig. 2e). On the other hand, power density changes in the cortex were more similar between the CaV2.3 KO and WT groups: a significant increase in the delta band (Fig. 3e), but no significant change in the theta band. The power in the alpha and gamma bands also significantly decreased (Fig. 3e). Relative power from the CMT showed no statistically significant shift at all, although there was a trending increase in the delta band (Fig. 2f). Relative power from the cortex showed a significant increase in the delta band (Fig. 3f).

Fig. 2
figure 2

Observed changes in the LFP signal recorded from CMT in WT and CaV2.3 KO animals before and after intraperitoneal injection of GBL (Pre-GBL and Post-GBL). a, b, Representative heat maps from a WT mouse (a) and a CaV2.3 KO mouse (b) with representative LFP traces recorded from the CMT (WT, black; CaV2.3 KO, green). c, Total power from WT mice across various frequency bands before and after GBL administration (δ: ****, p < 0.0001; θ: **, p < 0.01; two-way ANOVA). d, Relative power from WT mice across various frequency bands before and after GBL administration (δ:****, p < 0.0001; θ: *, p < 0.05; α: ****, p < 0.0001; β: **, p < 0.01; γ: ****, p < 0.0001; two-way ANOVA). e, Total power from CaV2.3 KO mice across various frequency bands before and after GBL administration (θ: *, p < 0.05; α: *, p < 0.05; γ: **, p < 0.01; two-way ANOVA). f, Relative power from CaV2.3 KO mice across various frequency bands before and after GBL administration. Data was collected from 6 WT and 6 mutant mice.

Full size image
Fig. 3
figure 3

The observed changes in the EEG signal recorded from cortex in WT (n = 6) and CaV2.3 KO (n = 6) mice before and after intraperitoneal injection of GBL (Pre-GBL and Post-GBL). (a, b), Representative heat maps from a WT mouse (a) and a CaV2.3 KO mouse (b) with representative EEG traces recorded from the cortex (WT, black; CaV2.3 KO, green). (c), Total power from WT animals across various frequency bands before and after GBL administration (δ: ****, p < 0.0001; θ: ****, p < 0.0001; γ: **, p < 0.01; two-way ANOVA). (d), Relative power from WT mice across various frequency bands before and after GBL administration (δ:****, p < 0.0001; α: ****, p < 0.0001; β: **, p < 0.01; γ: ****, p < 0.0001; two-way ANOVA). e, Total power from CaV2.3 KO animals across various frequency bands before and after GBL administration (δ: ***, p < 0.001; α: ***, p < 0.001; γ: *, p < 0.05; two-way ANOVA). f, Relative power from CaV2.3 KO animals across various frequency bands before and after GBL administration (δ: ****, p < 0.0001; α: **, p < 0.01; two-way ANOVA).

Full size image

Functional connectivity analyses

Phase locking value (PLV) provides insight into how different brain regions synchronize their activity. A higher PLV (ranging from 0 to 1) suggests more significant coordination of the neural activity between two regions. In our study, we compared levels of TC synchronization (between the cortex and CMT) and corticocortical (CC) synchronization in WT and CaV2.3 KO mice across various frequency bands. Additionally, we examined how functional delta band connectivity evolved over time. We analyzed EEG/LFP signals during a quiet awake baseline period lasting 5 min before GBL administration, as well as during a post-injection period spanning 30 min, divided into six 5-min segments to capture changes over time. The following results are presented on Figs. 4 and 5. During resting baseline quiet awake activity, we observed stronger TC phase synchronization in the delta band in the CaV2.3 KO mice compared with the WT mice (Fig. 4a). No statistically significant differences were observed in TC synchronization of activity 10–15 min after GBL administration (Fig. 4b). Additionally, there was increased CC gamma synchronization in CaV2.3 KO animals during the quiet awake state (Fig. 5a), with no differences in CC synchronization after GBL administration across all analyzed frequency bands (Fig. 5b) between WT and CaV2.3 KO mice. TC/CC delta band synchronization in WT animals increased from baseline as expected, but the TC delta band PLVs decreased from baseline in CaV2.3 KO mice (Figs. 4c and 5c). Evaluation of changes in TC/CC PLV within the delta band over time showed approximately 2-fold higher phase synchronization during baseline pre-GBL activity in CaV2.3 KO mice compared with WT mice (Fig. 4d). Following GBL administration, this disparity ceased to be statistically significant at any of the measured time points (Fig. 4d). Analysis of CC delta PLVs over time showed statistically higher values in CaV2.3 KO mice at a time point of 5 min after GBL injections (Fig. 5d).

Fig. 4
figure 4

Functional connectivity measured by phase-locking values (PLV) between the cortex and CMT. a, Comparison of PLVs between WT and CaV2.3 KO animals in different frequency bands measured during the 5-min time period before GBL administration. The difference in PLV in the delta band was statistically significant (*, p < 0.05, two-way ANOVA) between the two groups. b, Comparison of PLVs between WT and CaV2.3 KO animals in different frequency bands measured from the 10-15 min after GBL administration. The differences in the PLVs between groups were not statistically significant. c, Change in the functional connectivity in different frequency bands between WT and CaV2.3 KO groups after GBL administration, measured as the difference from the baseline (pre-GBL) PLV values. The differences between groups were not statistically significant. d, Time course showing changes in the connectivity in the delta band in WT and CaV2.3 KO groups acquired in 5-min intervals before and after GBL administration (Pre-GBL and Post-GBL). The pre-GBL time point is indicated by the 0 on the X-axis. The only statistically significant difference was observed before GBL administration (**, p < 0.005, two-way ANOVA). Data was collected from 6 WT and 6 mutant mice.

Full size image
Fig. 5
figure 5

Functional connectivity measured by phase-locking values (PLV) between the left (L Ctx) and right cortex (R Ctx). a, Comparison of PLVs between WT and CaV2.3 KO groups in different frequency bands measured in the 5-min time period before GBL administration. The difference in the gamma band PLV was statistically significant (*, p < 0.05, two-way ANOVA) between the two groups. b, Comparison of PLVs between WT and CaV2.3 KO animals in different frequency bands measured from the 10-15 min after GBL administration. The differences in the PLVs between groups were not statistically significant. c, Change in the functional connectivity in different frequency bands between WT and CaV2.3 KO mice after GBL administration, measured as the difference from the baseline (pre-GBL) PLV values. The differences between groups were not statistically significant. d Time course showing changes in the connectivity in the delta band in WT and CaV2.3 KO mice acquired at 5-min intervals before and after GBL administration (Pre-GBL and Post-GBL). The pre-GBL time point is indicated by the 0 on the X-axis. The only statistically significant difference was observed during the first 5 min after GBL application (*, p < 0.05, two-way ANOVA). Data was collected from 6 WT and 6 mutant mice.

Full size image

Discussion

Absence epilepsy is a generalized non-convulsive epilepsy characterized by a sudden, brief impairment of consciousness; it is often difficult to diagnose and mainly affects children17. Our study highlights the significance of CaV2.3 channels in the context of pharmacologically induced absence seizures within the TC network. Additionally, we present novel evidence of SWD activity in the CMT of the intralaminar thalamus. Further analyses of the EEG and LFP signals pointed to increased functional connectivity between the CMT and cortex in the baseline activity within the delta band in CaV2.3 KO mice.

GBL administration causes electrographic and behavioral events similar to generalized absence seizures18. This is primarily attributed to its actions on GABAB receptors located on neurons within the TC network. Hyperpolarization of membrane potentials induced by the activation of GABAB receptors evokes rebound burst discharges in TC neurons19. At hyperpolarizing membrane potentials, low-voltage activated T-type calcium channels recover from inactivation, which allows them to be activated by small depolarization steps. This results in the generation of low-threshold calcium spikes superimposed with bursts of action potentials20. In addition to CaV3.1 T-type channels, CaV2.3 R-type calcium channels may have an important role in generating SWDs15,21. Here, we demonstrated that global deletion of CaV2.3 channels reduces the susceptibility to pharmacologically induced SWDs recorded from the somatosensory cortex and CMT. This finding supports the observations of Zaman and colleagues14 that CaV2.3 channels in TRN neurons are key drivers of GBL-induced hypersynchronous oscillations. In contrast, Weiergräber’s group reported that mutant CaV2.3 KO mice are more susceptible to absence seizures15. An apparent reason for this discrepancy is not immediately obvious, but it is worth mentioning all of these studies used the mice with global deletion of CaV2.3 channels. Future studies with region-specific deletion or knock-down of these channels would be useful in delineating specific contributions of CMT, sensory thalamus and cortex in SWDs.

Although we used a mouse model with global deletion of CaV2.3 channels, we found significant differences between CMT and sensory cortex. For example, we did not observe the expected increase in delta power in the CMT following GBL injection in the CaV2.3 KO group (Supplemental Fig. 2D), which is contrary to the findings from both the cortex of these animals and the CMT/cortical recordings in the control WT mice. We previously showed the presence of CaV2.3 R-type calcium channels in the CMT together with the dominant CaV3.1 T-type calcium channels13, and R-type calcium channels are expressed in the somatosensory cortex12. However, the CMT receives inputs from the brainstem and is thus considered part of the ascending reticular activating system and an important contributor to wake cortical activity, regardless of the presence or absence of sensory stimulation22. In a rat genetic model of SWDs, Seidenbecher and Pape showed that SWD-related bursts were delayed in the intralaminar thalamic complex compared with those in the TRN, relay TC nuclei, and cortex23. Although possible differences between GBL-induced absence seizures and those seen in different genetic rodent models should be taken into consideration, this finding suggests that the role of the CMT in the mechanisms underlying SWD-related activity in the intralaminar thalamic complex differs from that in the sensory thalamus or cortex. This could possibly arise from the different anatomic CMT-cortical connectivity compared with the sensory cortex and sensory thalamus. For example, projections from layer V of the cortex24 and GABAergic input from the TRN25 provide direct anatomic connections between the TC network and CMT. However, the CMT receives inhibitory inputs from the substantia nigra pars reticulata, habenula, zona incerta, and the external segment of the globus pallidus26. Importantly, it appears that inhibitory inputs from the basal ganglia27, which project to the CMT but not to the sensory thalamus, are particularly important for GBL-induced SWDs.

In rodents, waking behavior can be segregated into an active state associated with alertness and active exploration and a quiet state devoid of locomotion. During the active wake state, the EEG shows desynchronized activity dominated by fast frequencies, whereas the quiet wake state shows mixed frequencies with spontaneous activity in the lower frequencies28. During absence seizures, generalized rhythmic SWDs form and represent a pathologic neuronal hypersynchrony with dominant high-amplitude delta waves29. Given that synchronization is a fundamental feature of epilepsy, many studies have analyzed the epileptic brain through measures of functional synchronization, such as PLV30. Importantly, there is an increasing interest in developing new ways to characterize and detect absence seizures in children, and researchers are investigating the potential for PLV extracted from EEG recordings to serve as a biomarker for identifying absence seizure onset31.

Our data showed the presence of higher baseline TC delta frequency band PLVs in CaV2.3 KO mice with no change after GBL injection. Similarly, we observed higher CC synchronization in the gamma band in CaV2.3 KO mice during a quite wakeful state before GBL injection with no difference between WT and CaV2.3 KO mice after GBL injection. As expected, after GBL injection, TC delta band PLVs increased from baseline in WT mice while they decreased from baseline in CaV2.3 KO mice; in contrast, CC delta PLVs increased from baseline following GBL injection in both WT and CaV2.3 KO mice. The observed baseline delta TC and gamma CC hypersynchronization in CaV2.3 KO mice shows how the lack of CaV2.3 channels affects overall neuronal synchronization. Moreover, we observed higher delta and theta power in the cortex in CaV2.3 KO mice during a quiet awake state, as previously described by Siwek and colleagues32. Although the CaV2.3 KO animals seem protected from GBL-induced seizures (no increase in thalamic delta oscillations, inability to induce an increase in TC delta synchronization, and fewer cumulative SWDs), the very high baseline delta total power and high delta TC PLVs can partially explain the inability of GBL to induce additional spectral changes in the CMT. Keeping in mind the important role of voltage-gated T-type calcium channels in the generation of absence seizures33, we cannot exclude the possibility that animals lacking CaV2.3 calcium channels have increased function of T-type calcium channels as a compensatory mechanism. Although this compensatory increase in T-type calcium channels in CaV2.3 KO animals could explain the existence of higher delta band power and delta TC hypersynchronization during the awake state, it would not explain the inability of GBL to induce thalamic spectral and behavioral changes. Interestingly, a previous study using quantitative polymerase chain reaction for CaV3.1-3.3 isoforms, as well as CaV1.2, CaV1.3, CaV2.1 and CaV2.2 channels with thalamic RNA preparations from CaV2.3+/+, CaV2.3+/−, and CaV2.3−/− mice showed no compensatory alterations in the expression of these voltage-gated calcium channels15,32. However, functional studies to support this notion are lacking. At the same time with disrupted functional synchronization, altered power in the delta/theta bands in CaV2.3 KO mice before GBL administration implies that the deletion of R-type calcium channels disrupts normal brain activity and can decrease the ability of GBL to induce SWDs.

In conclusion, our data strongly support the importance of thalamic CaV2.3 channels in promoting and sustaining synchronized TC oscillations. Furthermore, our data are consistent with the pro-convulsant role of CMT and cortical CaV2.3 channels in a rodent model of GBL-induced absence seizures manifested as SWDs. Thus, the potential involvement of CaV2.3 R-type channels in the physiologic and pathologic functioning of TC circuits warrants further investigation.

Methods

Animals

CaV2.3-deficient mice backcrossed into C57Bl/6 were generated and described previously34. For this study, adult male CaV2.3-deficient animals (n = 6) and male control mice (n = 6) were used. Two of the six control animals were WT littermates of the CaV2.3-deficient mice, while the other four were obtained from the Jackson Laboratory as C57Bl/6 WT mice. As we did not observe differences in recordings from WT littermates and C57BL/6 WT mice, these data were grouped together. All the animals were 4–6 months of age.

Drugs

GBL (MilliporeSigma, St. Louis, MO, USA), a photosensitive compound, was kept in the dark until use, then dissolved in sterile saline to prepare stock solutions of 50 mg/ml for injections at 70 mg/kg15. Injections were administered i.p. in a sterile manner.

EEG Electrode implantation and recording

To record EEG signals, mice were surgically implanted with a stainless-steel insulated wire electrode in the CMT (anteroposterior [AP] = -1.35 mm from bregma, mediolateral [ML] = 0 mm [on the midline], and dorsoventral [DV] = 3.75 mm below the skull surface) and two screw cortical EEG electrodes (AP = − 1.0 mm, ML = ± 3.0 mm from midline) under isoflurane anesthesia (0.5–2%). Screw electrodes placed behind lambda on each side of the midline served as ground (right) and reference (left). The electrodes were fixed to the skull using dental acrylic. The mice were treated postoperatively with an analgesic (flunixin, Merck, Rahway, NJ, USA), immediately after the procedure (0.3 mL of 5% solution). To ensure adequate postoperative recovery, we recorded synchronized, time-locked video and EEG signals from mice using the Sirenia Acquisition system (Pinnacle Technology Inc., Lawrence, KS, USA) at least 1 week after the surgery. Acquired EEG signals were amplified 10 times, digitized at a sampling frequency rate of 2000 Hz (with a 0.5-Hz high-pass filter and a 500-Hz low-pass filter), and stored on a hard disk for offline analysis. EEG recordings were obtained for 15 min before and for 45 min after GBL administration. After completion of the experiments, the mice were anesthetized with ketamine (100 mg/kg, i.p.) and electrolytic lesions were made by passing a 5 µA current through the depth electrode for 1 s (5 times). Mice were subsequently anesthetized with isoflurane and perfused with ice-cold 0.1 M phosphate buffer containing 1% potassium-ferrocyanide. Next, brains were extracted, kept in 4% formalin for 1–2 days, and sliced in the coronal plane (100 μm) using a vibratome (Leica VT 1200 S). Images of slices with electrode locations were obtained using a stereo zoom microscope (Zeiss Stemi 508).

Power spectral analysis

We analyzed the EEG/LFP recording data points at 10 min before and 30 min after GBL administration using power spectral analysis (Fast Fourier Transform). The EEG frequency spectrum was divided into the following frequency bands: delta (0.5–4 Hz), theta (4–8 Hz), alpha (8–12 Hz), beta (13–30 Hz), and gamma (> 30 Hz with exclusion of 60 Hz frequency). Power density, total power, and relative power spectral analysis were calculated using LabChart 8 software (ADInstruments Inc., Colorado Springs, CO, USA). Power spectrograms were computed using the Brainstorm software package implemented in MATLAB.

Spike and wave discharge analysis

GBL serves as a prodrug of γ-hydroxybutyrate that reliably induces absence seizures in experimental animals. In mice, these seizures are characterized by a sudden behavioral arrest with occasional vibrissal twitching35. The EEG recordings show a typical bilateral synchronous SWD pattern. We manually selected SWD events and identified the true onset and offset of each SWD by locating the time point of the first and last peak of the event (as defined by sections of the recording that were two times the background baseline). SWD duration was defined as the duration between the first and last peak. SWD frequency was quantified by computing a fast Fourier transform (FFT) on the event to confirm typical 3–5 Hz frequency. SWDs with voltage amplitudes of at least twice the background EEG and a minimum duration of 0.7 s were considered separate events if they were separated by > 1 s14 (see representative traces of individual SWDs on Supplementary Fig. 1a). Inter-ictal duration (non-SWD period) was considered as time between two separate SWD events. CMT recording from one mutant mouse was noisy and was excluded from the analysis because we could not score SWDs (Fig. 1and Supplementary Figs. 1 and 2).

Functional connectivity analysis

We measured phase synchrony between the cortex and CMT using PLV. This method is well-suited for connectivity analysis because it provides a measure of neural signal temporal relationships independent of their signal amplitude36. PLVs of the EEG signals were measured in 5-min intervals: once before the injection of GBL (baseline) and 6 times (a total of 30 min) after the injection of GBL, starting immediately after the injection. Analyses were performed using the Brainstorm software package implemented in MATLAB.

Data analysis

Statistical analyses were conducted using two-way ANOVA with multiple comparisons to analyze differences in the total and relative power of EEG signals between frequency bands, as well as the differences in PLV values across different frequencies. For SWD duration, frequency, inter-ictal period and latency, an unpaired t-test was used to compare data among different animal groups, and a paired t-test was used to compare data within the same animals. Significance was accepted with p-values < 0.05. The statistical and graphical analysis was performed using GraphPad Prism 9.2.0 software (GraphPad Software) and GIMP 2.10.4 (GNU Image Manipulation Program). For all experiments, power was calculated using standard algorithms with reference to previously published studies12,15,32.