All animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC) of Tel Aviv University. Mice used in this study were housed in a standard animal facility at the Goldschleger Eye Institute at a constant (22°C) temperature, on 12-h light/dark cycles, with ad libitum access to food and water.

DS mice harboring the global Scn1a ^A1783V/WT mutation on pure C57BL/6J background were generated as described before (Fadila et al., 2020; Almog et al., 2021), by crossing conditional Scn1a ^A1783V males (The Jackson Laboratory, stock #026133) with CMV-Cre females (The Jackson Laboratory, stock #006054).

Electrode implantation was done on P19-P25, as previously described (Fadila et al., 2020). Briefly, the mice were anesthetized (ketamine/xylazine, 191/4.25 mg/kg), and carprofen (5 mg/kg) was given for analgesia prior to surgery and 24 h post-op. A midline incision was made above the skull, and fine silver wire electrodes (130 μm diameter bare; 180 μm diameter coated) were placed at visually identified locations: bilaterally above the somatosensory cortex; a reference electrode was placed on the cerebellum midline; ground and electromyography (EMG) electrodes were placed subcutaneously over the left and right shoulders, respectively. The electrodes were secured using dental cement and connected to a micro-connector system. Mice were allowed to recover for at least 48 h before recording.

The drugs that were used were: valproic acid (VA, 300 mg/kg in saline; Sigma-Aldrich), clobazam (CLB, 10 mg/kg in sesame oil; EDQM), stiripentol (STP, 150 mg/kg in sesame oil; Angene Chemical), a combination of clobazam and stiripentol (CLB + STP, 5 mg/kg + 100 mg/kg in sesame oil), carbamazepine (CBZ, 20 mg/kg in 30% polyethylene glycol 400; Alomone Labs) and lamotrigine (LTG, 10 mg/kg in 30% polyethylene glycol 400; Alomone Labs). Drug dosages were chosen to reach therapeutic-relevant concentrations following acute administration (Hawkins et al., 2017). All drugs were administered as a single IP injection in a volume of 10 ml/kg.

Video-ECoG recordings were obtained from freely behaving mice at P21-P27, as described before (Fadila et al., 2020). The recordings lasted for 4 h, between 10 a.m. and 5 p.m. After 2 hours of recording, one of the drugs was administered. The first 30 min of the recording and the first 30 min post-drug administration were considered acclimation periods and were not taken for analysis. The EMG signal was denoised and smoothed using a custom-written Python script, based on methods described before (Solnik et al., 2010). Briefly, the raw EMG signal was denoised with the Teager-Kaiser Energy Operator (TKEO), and the output was rectified and smoothed across a 3 Hz cut-off low pass filter. Next, to calculate spectral properties, the recording was divided into 5 s epochs with a 2.5 s sliding window. Power spectral densities of the ECoG were computed for each epoch using Welch’s method with a 50% overlap Hann window. The power bands were defined as: delta, 0.9–3.9 Hz; theta, 4.8–7.8 Hz; alpha, 8.7–11.7 Hz; beta 12.6–29.2 Hz; gamma, 30.2–99.6 Hz. We used a custom-written threshold algorithm in python to extract artifact-free epochs where the mice were awake and not moving. An epoch was taken for analysis if all conditions were met: i) the delta power of the epoch was lower than the median delta power; ii) the theta to delta ratio was lower than the 75% quantile of the entire 1.5 h recording block; iii) the enveloped EMG was less than two times the harmonic mean throughout the entire epoch (Weber et al., 2015; Barger et al., 2019). Finally, all automatically selected epochs were concatenated and manually inspected. For each mouse, at least 100 epochs were analyzed. Then, the data were averaged across the different epochs. Thus, only one data point before drug administration, and one data point after ASM administration, was considered for statistical analysis. To quantify phase synchrony between the left and right somatosensory electrodes, we calculated the interhemispheric coherence coefficient (coherence) using a custom-written Python script, based on the coherence coefficient equation as described before (Bastos and Schoffelen, 2016).

Custom-written codes are available at https://github.com/shirquinn/Rubinstein_Lab.git.

To quantify the frequency of interictal spikes, we applied a bandpass filter between 0.5 and 60 Hz and used the spike histogram module in LabChart 8 software (ADInstruments). The threshold was set to 4–5 times the standard deviation of the filtered signal, with a 50 ms pretrigger interval and 250 ms maximal total duration. The spikes were then manually inspected in the scope view window, and artifacts were rejected.

Thermal induction of seizures was performed as previously described (Almog et al., 2021). Briefly, 30 min after injecting one of the selected drugs, the body temperature was increased by 0.5°C every 2 min using a heat lamp attached to a feedback temperature controller (TCAT-2DF, Physitemp Instruments), until a generalized tonic-clonic seizure (GTC) was provoked. Drugs that were used for assessing thermal induction (TI) of seizures were: i) VA (300 mg/kg); ii) CLB + STP (5 mg/kg + 100 mg/kg). Control DS mice were injected with the appropriate vehicle control.

Statistical analyses were performed using Prism 9 (GraphPad Software). Data are reported as mean ± SE. Statistical analysis utilized Student’s t-test or the non-parametric Mann-Whitney test (for data with normal distribution or data that did not distribute normally). For comparison between wild-type (WT) and DS mice and the effect of drugs on thermal-induced seizures, we used the Student’s t-test (for data with normal distribution) or the non-parametric Mann-Whitney test (for data that did not distribute normally). To compare WT or DS mice before and after drug administration, we used paired Student’s t-test (for data with normal distribution), or the Wilcoxon matched-pairs signed rank test (for data that did not distribute normally). To compare the effect on normalized power spectral density (PSD) or coherence, we used Two-Way repeated measures ANOVA followed by Sidak posthoc analysis. For the correlation analysis, the Spearman rank-order correlation coefficient was used. Statistical analysis of pie charts was performed using Fisher’s exact test. We considered p < 0.05 as statistically significant.

Global neuronal activity in juvenile WT and DS littermates (P21-P27) was examined using simultaneous video-ECoG recordings, followed by quantitative background activity spectral properties analyses. In accordance with previous reports (Fadila et al., 2020; Gerbatin et al., 2022), a global reduction in total power was seen in DS mice at this age, observed in all frequency bands up to 30 Hz (Figures 1A–D). Extracting the relative power for each band revealed a higher contribution of the gamma band and a reduced contribution of theta and alpha bands (Figure 1E).

While seizure activity is characterized by synchronized activity (Supplementary Figure S1), altered interhemispheric phase synchrony, or coherence, of background non-epileptic segments was found in various brain disorders, including epilepsy and autism (Bowyer, 2016). Here, we found reduced coherence in DS mice compared to WT controls, mainly in the alpha and beta frequency bands (Figures 1F, G), indicating deficits in synchrony between frequency components of background ECoG. Together, quantitative analyses of background ECoG recordings from DS mice demonstrated reduced power and coherence.

Thermally-induced seizures are a hallmark phenotype of Dravet, and multiple studies examined the effect of ASMs on the temperature threshold of these seizures (Hawkins et al., 2017; Pernici et al., 2021). However, this measurement provides a unidimensional description of a transition into a convulsive seizure and cannot be adopted for clinical use. We reasoned that analyses of the properties of background oscillations might provide additional insights into the effect of ASMs in Dravet. From these recordings, we also examined the effect of these ASMs on epilepsy, quantifying the change in the frequency of interictal epileptic spikes. Interictal spikes reflect synchronous neuronal firing. They are strongly associated with epilepsy and were shown to have adverse effects on brain development and contribute to cognitive impairment (Holmes, 2013). Although the relationship between interictal spikes and spontaneous seizures is not clear (Avoli et al., 2006), quantification of their frequency is commonly used to evaluate the severity of epilepsy and drug response in animal models. We focused on the acute effect of ASMs due to challenges with long-term ECoG recording and prolonged drug administration in immature mice.

Valproic acid (VA) is a broad-spectrum antiseizure medication and the drug of choice as a first-line treatment for Dravet. Specifically, VA was shown to reduce the frequency of spontaneous seizures in 50%–70% of patients (Shi et al., 2016; Hawkins et al., 2017). In DS mice, chronic treatment with VA protected from spontaneous convulsive seizures (Hawkins et al., 2017), but its acute effects on background ECoG activity were not assessed.

When given acutely, VA has several molecular targets: i) increasing GABA level by reducing its degradation via inhibition of GABA transaminase and succinic semi-aldehyde dehydrogenase; ii) potentiation of the activity of GABA_A receptors; iii) direct inhibition of voltage-gated sodium channels and T-type voltage-gated calcium channels; iv) inhibition of NMDAR. In addition, long-term VA administration, which was not tested here, results in histone deacetylase inhibition, positive modulation of M currents, and activation of the ERK and JNK pathways (Rosenberg, 2007; Zanatta et al., 2019).

First, we tested the effect of VA on the threshold of heat-induced seizures. As shown in Figure 2A, mice treated with VA (300 mg/kg) tended to convulse at higher temperatures, but this difference did not reach statistical significance. Previous results reported similar observations using the same mouse model (Scn1a ^A1783V on the pure C57BL/6J background, (Pernici et al., 2021)). Conversely, others that used DS mice with a mixed genetic background demonstrated a significantly increased seizure threshold (Hawkins et al., 2017; Ho et al., 2021). Therefore, we repeated these experiments in Scn1a ^A1783V mice on a mixed C57BL/6J:129x1/SvJ background. VA significantly reduced the susceptibility to thermal seizures in these mice, increasing the average threshold temperature by 1.69°C ± 0.72°C (Supplementary Figure S2). C57BL/6J mice carry a variant of the Gabra2 gene that results in reduced expression of the encoded GABA_A α2 subunits. This gene was shown to be a genetic modifier in Dravet (Hawkins et al., 2021). As VA also acts via the potentiation of GABA_A, this genetic variant in C57BL/6J may be related to the reduced effect of VA on heat-induced seizures. Thus, genetic background affects the ability of VA to protect from febrile seizures. However, as DS mice on a mixed background were reported before to have milder epileptic phenotypes (Yu et al., 2006; Mistry et al., 2014; Rubinstein et al., 2015), the severely affected DS mice on the pure C57BL/6J background were used for the rest of the study.

In a different cohort of mice on the C57BL/6J background, we tested the acute effect of VA on the frequency of interictal spikes and spectral properties of background ECoG. VA had a variable impact on spike frequency, reducing the frequency by over 50% in 57% of the tested mice, with a smaller effect in 28% of the mice, and increased spike frequency in one mouse (Figures 2B, C). Focusing on quantitative analyses of background activity, VA reduced the total power spectral density (PSD) in all the tested DS mice (Figures 2D–F), in agreement with previous reports in humans (Clemens et al., 2006; Moon et al., 2022). When inspecting the relative power of each frequency band, we observed an increase in the contribution of the beta band (Figure 2G). Despite the global effect on ECoG power, VA did not affect the coherence patterns in DS mice in any specific band. Yet, a slight overall reduction in phase coherence was observed (Figure 2H). The acute effect of VA was not specific to DS mice, and the reduction in total power was also seen in WT mice (Figure 2I–L, Supplementary Figure S3A). Together, although VA had variable effects on the frequency of interictal spikes in DS mice, it reduced the power and modulated the spectral properties of background ECoG activity.

Clobazam (CLB), frequently in combination with stiripentol (STP), is another commonly used standard treatment in Dravet (Strzelczyk and Schubert-Bast, 2022; Wirrell et al., 2022). As a monotherapy, clobazam was reported to be effective in 25%–50% of patients (Shi et al., 2016), while the addition of STP increased responsiveness. CLB is a long-acting 1,5-benzodiazepine that acts as a positive allosteric modulator of ionotropic GABA_A receptors (Hammer et al., 2015). CLB as monotherapy in mice was shown to protect from thermally induced seizures (Hawkins et al., 2017; Pernici et al., 2021). Here, CLB (10 mg/kg) reduced the frequency of interictal spikes in 80% of the mice (Figures 3A, B), had no effect on the total power or coherence (Figures 3C–E, G), but resulted in an increased contribution of the beta frequency band (Figure 3F).

STP is often administered as an adjunct drug. It enhances the effective concentration of CLB by inhibiting CYP450 isoenzymes. In addition, STP also directly modulates inhibitory GABAergic transmission via allosteric modulation of the GABA_A receptors, which enhances the positive effect of CLB on these channels. Moreover, STP also inhibits voltage-gated T-type calcium channels and lactate dehydrogenase (LDH), which was shown to suppress seizures (Fisher, 2011).

Acute administration of CLB + STP (5 mg/kg, 100 mg/kg) significantly reduced the susceptibility to thermally induced seizures, elevating the average temperature of convulsive seizures by ∼ 1.4°C (Figure 4A). The effect of this drug combination was further tested using ECoG recordings. The frequency of interictal spikes decreased by over 50% in 33% of the mice, and by 20%–50% in another 33%, while no change or an increase in spike frequency was observed in the rest of the mice (Figures 4B, C). Moreover, we did not observe a significant effect on the total power of background activity in DS mice (Figures 4D–F). Nevertheless, acute administration of CLB + STP altered the relative power of the delta, theta, and beta frequency bands, reducing the delta contribution and increasing that of theta and beta (Figure 4G). The interhemispheric coherence did not change after CLB + STP acute administration (Figure 4H).

The effect of CLB + STP on WT mice was similar to that of DS mice, with no effect on the total power or coherence (Figures 4I–L), but with an enhancement of the relative power of the beta frequency band (Supplementary Figure S3B). Together, CLB + STP reduced the susceptibility to thermally induced seizures, reduced the frequency of interictal spikes in most of the tested mice, and modulated the spectral properties of background ECoG activity.

STP as monotherapy was shown to reduce seizure burden in patients (Hawkins et al., 2017). Neverthelss, it is usually administered as adjunct therapy (Wirrell et al., 2022), and was not shown to be protective against thermally induced seizures in mice (Hawkins et al., 2017; Pernici et al., 2021). With its direct effect on GABAergic transmission (Fisher, 2011), we wondered if it would also affect the spectral properties of background ECoG activity.

STP (150 mg/kg) had variable effects on the frequency of interictal spikes, reducing their frequency by 25%–53% in three mice and increasing their frequency by 35%–90% in two other DS mice (Figures 5A, B). Moreover, while no significant effects were observed on the total power or coherence (Figures 5C–E, G), STP increased the contribution of the beta frequency band (Figure 5F). Together, Dravet-prescribed drugs, VA, CLB, CLB + STP, or STP alone, demonstrated variable, subject-dependent changes in the frequency of interictal epileptic spikes but with significant modulation of the spectral properties of background activity, and a common increase in the contribution of the beta frequency band.

ASMs that act mainly via inhibition of voltage-gated sodium channels are contraindicated in Dravet. These drugs were shown to aggravate the epileptic phenotypes and adversely affect the cognitive outcome (de Lange et al., 2018; Wirrell et al., 2022). Carbamazepine (CBZ) mainly targets voltage-gated sodium channels. It has an increased affinity to the inactivated state, resulting in a voltage or frequency-dependent inhibition of these channels. However, CBZ was also demonstrated to inhibit the N-type and L-type voltage-gated calcium channels and increase the secretion of serotonin (Ambrósio et al., 2002).

CBZ was reported to exacerbate the susceptibility to thermally induced seizures in DS mice (Pernici et al., 2021), but its effect on ECoG and background activity was not reported. Here, among the mice treated with CBZ (20 mg/kg), three out of four mice demonstrated an increased frequency of epileptic spikes by 2-9 folds, and a marginal (3%) decrease in spike frequency was seen in the fourth mouse (Figures 6A, B). Background spectral properties analysis showed that CBZ did not modify the power, spectral profiles, or coherence (Figures 6C–G). Similarly, in WT mice, CBZ did not modify the properties of background oscillations (Figures 6H–K).

Lamotrigine (LTG) is another voltage-gated sodium channel blocker avoided in Dravet (Wirrell et al., 2022). Additional mechanisms of action include inhibition of voltage-gated calcium channels and reduced presynaptic glutamate release (Prabhavalkar et al., 2015). In DS mice, acute administration of LTG (10 mg/kg) resulted in increased interictal spike frequency (Figures 7A, B). Moreover, LTG did not affect the power or spectral properties of background activity (Figures 7C–G). Together, acute administration of either CBZ or LTG increased the frequency of interictal spikes in DS mice without affecting the properties of background oscillations.

We wondered if comprehensive analyses of the various drug-induced changes would highlight parameters that can differentiate between Dravet-prescribed and contraindicated drugs and correlate with the effect on the epileptic phenotypes as measured here by the frequency of interictal spikes.

First, we quantified the overall effect of ASMs on spike frequency. VA, CLB, or CLB + STP reduced the spike frequency in ∼80% of the mice (Figures 2–4), while STP lowered their frequency in 60% of the tested mice (Figure 5). Overall, Dravet-recommended ASMs reduced spike frequency by over 50% in 30% of the mice and by over 10% in 66% of the mice. Conversely, contraindicated drugs increased the spike frequency by more than 50% in 87.5% of the tested mice (Figures 8A, B).

Next, we examined the relationship between background spectral changes and the effect on interictal spikes (see Supplementary Figure S5 for the complete correlation analyses of spectral changes). Focusing on the total power shift, we did not see a statistical difference between the effect of Dravet-recommended and contraindicated drugs (Figure 8C). However, interestingly, the level of total power reduction correlated with a decrease in spike frequency (Figures 8D, E). This correlation was also significant for the cohort of mice treated with CLB + STP (Supplementary Figure S4). Examination of the spectral power redistribution showed an increase in the relative beta contribution among the majority of mice treated with Dravet-prescribed ASMs and a less profound change after CBZ and LTG administration (Figure 8C). Nevertheless, the relative drug-induced change in the beta band contribution did not correlate with the reduction in spike frequency (Figure 8D). Thus, we wondered if power redistribution across multiple bands would correlate with the effect on interictal spikes. The ratio between theta and beta has been extensively investigated in relation to attention processes and is FDA-approved for clinical confirmation in attention deficit disorder (Clarke et al., 2020). Interestingly, a reduction in this ratio correlated with a reduction in spike frequency. A similar trend was observed when we calculated the ratio of theta to beta and gamma combined (Figures 8C–E). Together, examination of multidimensional readouts for the effect of acute antiseizure treatment on background activity and epileptic interictal spikes revealed a correlation between the frequency of interictal spikes, background power, and the relative contribution of theta, beta and gamma frequency bands.