The small molecule simufilam dose‐dependently attenuates the worsening of seizures in a mouse model of tuberous sclerosis complex

Abstract

Objective

Novel epilepsy treatments for patients with tuberous sclerosis complex (TSC) and focal cortical dysplasia type II (FCDII) are urgently needed. In these patients, mutations in the mechanistic target of rapamycin (mTOR) pathway genes lead to mTOR hyperactivity and focal cortical malformations that frequently cause intractable epilepsy and neurological sequelae. Recent evidence suggests that administration of simufilam, a small molecule thought to modulate the function of filamin A, reduces seizure activity independent of mTOR in a mouse model of FCDII. Here, we tested the hypothesis that simufilam treatment reduces seizure activity in a TSC mouse model and characterized its pharmacokinetic (PK) profile in wild‐type mice.

Methods

Seizure activity was recorded via electroencephalography (EEG) in juvenile Tsc1 conditional knockout (cKO) mice treated with simufilam or vehicle twice daily (BID). Simufilam plasma concentrations were measured at study termination using liquid chromatography–tandem mass spectrometry. PK analysis was performed in CD1 mice following daily administration of simufilam for 1 or 7 days at two different doses.

Results

Vehicle‐ and 5 mg/kg‐treated Tsc1 cKO mice showed a significant increase in seizure frequency and total ictal duration between the first and the last 5 days of EEG monitoring (Wilcoxon signed‐rank test, p < .05), whereas the 10‐ and 20‐mg/kg‐treated groups did not. In addition, simufilam treatment shifted the distribution of total ictal durations, with fewer mice exhibiting total ictal durations above the median of the vehicle‐treated group (Fisher's exact test, p = .021). Increasing simufilam dose and plasma concentration were associated with fewer daily seizures recorded during the last 5 days of EEG monitoring (p = .017 and p = .012, respectively). Every 100 ng/mL increase in plasma concentration corresponded to a decrease of ~.2 seizures per day during the last 5‐day period.

Significance

These findings suggest that simufilam treatment in a TSC mouse model prevented the worsening of seizure activity in a dose‐dependent manner and supports clinical investigation of simufilam as a potential therapeutic intervention for seizures in TSC.

Key points.

• Simufilam attenuated the worsening of seizures over time in Tsc1 conditional knockout mice.

• A positive correlation was found between improved seizure outcome and simufilam dose as well as plasma concentration.

• Simufilam, an oral small molecule, displayed rapid absorption, dose‐dependent pharmacokinetics, and low accumulation after repeated oral dosing in mice.

1. INTRODUCTION

Tuberous sclerosis complex (TSC) and focal cortical dysplasia type II (FCDII) are rare genetic disorders caused by somatic mutations in mechanistic target of rapamycin (mTOR) pathway genes (TSC1 or TSC2 in TSC and, for example, RHEB or mTOR in FCDII) leading to mTOR complex 1 (mTORC1) hyperactivity and focal malformations of cortical development. ¹ , ² , ³ , ⁴ , ⁵ , ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ In 80%–90% of patients, these abnormalities cause epileptic seizures, which often respond poorly to conventional anti‐epileptic drugs. ¹⁴ , ¹⁵ , ¹⁶ Despite these similarities between TSC and FCDII, most patients with TSC exhibit germline mutations (i.e., present in every cell of the body) as well as somatic mutations in multiple organs, whereas somatic mutations in patients with FCDII are restricted to the brain. As a result, TSC has an incidence of 1 in 6000–8000 live births ¹⁷ and patients can be genetically identified, making its patient population more suitable for targeted therapy than FCDII, which cannot be identified by genetic testing. In addition, patients with TSC exhibit a high incidence of neuropsychiatric deficits and non‐cancerous tumor growth in multiple organs, particularly the skin and kidneys. ¹⁷ , ¹⁸ Neurological manifestations, however, remain the most debilitating symptoms of TSC, particularly epilepsy often accompanied by difficult‐to‐manage comorbidities such as insomnia, intellectual disabilities, and behavioral issues (e.g., anxiety). ¹⁹ , ²⁰ Beyond low‐efficacious anti‐epileptic drugs, current therapeutic options for TSC include surgical resection of epileptogenic focal cortical malformations (applicable to only 10%–15% of patients ²¹ ), as well as the administration of the mTORC1 inhibitor, everolimus, and cannabidiol (Epidiolex). These treatments have several limitations: surgery is invasive and available drugs may be associated with adverse events and are not fully effective in all patients. ²² , ²³ There is an urgent need to improve epilepsy treatment strategies for patients with TSC, and one promising approach has been to identify molecules involved in disease pathogenesis.

Recent work has established the importance of the actin‐crosslinking and scaffolding molecule, filamin A (FLNA), in TSC and FCDII disease pathogenesis and has reported elevated FLNA expression in cortical malformations from patients with these disorders. ²⁴ , ²⁵ In addition, preclinical studies have reported that genetic normalization of FLNA expression led to substantial seizure reduction in a mouse model of FCDII with focal‐onset seizures. ²⁵ Similarly, simufilam (formerly known as PTI‐125)—a small molecule hypothesized to modulate FLNA function—improved neuronal connectivity and reduced seizure activity, independent of mTORC1 activity. ²⁵ However, it remained uncertain whether simufilam could reduce seizure activity in a mouse model of TSC. Therefore, we tested the efficacy of simufilam administered at several doses on seizure activity in Tsc1 conditional knockout (cKO) mice, followed by measurement of simufilam plasma concentration to assess exposure. In addition, we assessed the upregulation of FLNA in Tsc1 cKO mice and characterized the pharmacokinetic (PK) profile of simufilam in wild‐type (WT) CD1 mice.

2. MATERIALS AND METHODS

2.1. Animals

All procedures were in accordance with PsychoGenics and WuXi AppTec Institutional Animal Care and Use Committee regulations. WuXi AppTec purchased CD1 mice for the PK studies from Hilltop Lab Animals. Tsc1 ^flox/flox (Tsc1 WT) and Tsc1 ^flox/flox; human GFAP‐Cre conditional knockout (Tsc1 ^GFAP cKO) mice were bred at PsychoGenics as described previously. ²⁶ , ²⁷ These mouse lines were derived from a colony developed by Dr. M. Wong at Washington University in St. Louis. GFAP‐Cre induces recombination in both astrocytes and neurons throughout the brain. ²⁸ , ²⁹ Mice were housed at 20–23°C and 12/12 light/dark cycles on OptiMICE racks in an environment containing an igloo, bone, and nesting material. Food (Lab Diet, 5001), water, and NRadioimmunoprecipitation Assayutra‐Gel (Bio‐Serv, NBG‐2) were provided ad libitum. Pups were ear notched for genotyping identification on postnatal day 10 (P10). Animals were weighed once per week. Survival and health checks were performed once per day. Both male and female Tsc1 ^GFAP cKO mice (n = 18 per group) were generated for analyses of weight (to P56), survival (to P57, at the time of euthanasia), and seizure monitoring, although only 15 per group were ultimately included for EEG due to technical issues with the EEG implantation. Only CD1 male mice were included for the PK studies. We conducted simufilam PK studies in CD1 mice for several reasons: (1) to align with prior efficacy studies using this strain, ²⁵ (2) to inform dosing in the TSC1 model (WT background despite central nervous system [CNS] pathology), and (3) to correlate with PK data from outbred Sprague Dawley rats and healthy human volunteers obtained by Filana Therapeutics. As an outbred stock, CD1 mice better model human genetic variability than the inbred strains used for the TSC1 mouse model. Ultimately, these data, including TSC1 model exposure correlated with efficacy, can be used for modeling to estimate human dosing.

2.2. Lysate preparation and immunoblotting

Frozen brain tissues from three Tsc1 WT (P75) and three Tsc1 ^GFAP cKO (P57) mice were processed by homogenization in ice‐cold Radioimmunoprecipitation Assay (RIPA) buffer (ThermoFisher cat#89901) supplemented with protease and phosphatase inhibitors (ThermoFisher cat#78430, #78428). Standard protocol was used and is detailed in the Supporting Information.

2.3. Drug treatment and formulation

Simufilam was provided by Filana Therapeutics at 99.9% purity (lot #2021000033, Evonik Corp, Lafayette, IN). The molecular weight of simufilam (free base) is 259 g/mol and its formula weight (HCL salt) is 350.25 g/mol. A salt factor of 1.35, therefore, was applied to prepare dosing formulations using the HCL salt, and the reported simufilam doses refer to the non‐salt (free base) form. For CD1 mice (PK study), oral (per os; PO) simufilam was dissolved in .5 M citrate buffer pH 5.2 (±.1) to prepare a working solution that was used for the dosing week. For Tsc1 ^GFAP cKO mice, simufilam was dissolved in phosphate‐buffered saline (PBS; pH of 7) and administered twice daily via intraperitoneal (i.p.) injection to match our previous studies performed in a mouse model of FCDII. ²⁵ Simufilam was administered at 8–9 a.m. and 4–5 p.m., resulting in dosing intervals of ~8 h in the daytime and ~16 h overnight. Simufilam formulation for the Tsc1 ^GFAP cKO mice was prepared on a weekly basis and stored refrigerated at 4°C, shielded from light.

2.4. PK profiling of simufilam in CD1 mice

Sixteen CD1 male mice were randomly split into four groups (n = 4 per group). Groups 1 and 2: single oral administration (PO) dosing at 10 and 50 mg/kg; Groups 3 and 4: once per day (QD) PO dosing at 10 or 50 mg/kg for 7 days. For each group, blood was collected at eight time points up to 24 h post‐dosing (0, 30, and 60 min, and 2, 3, 6, 12, and 24 h). For Groups 3 and 4, sampling occurred after the Day 7 dose. Simufilam concentration was quantified in 128 mouse plasma samples (dipotassium ethylenediaminetetraacetic acid; EDTA‐K2) by a qualified Liquid chromatography–mass spectrometry (LC–MS/MS) method (WuXi AppTec) having a plasma lower‐limit of quantification (LLOQ) of 1 ng/mL. The method is detailed in the Supporting Information.

PK parameters were calculated using Phoenix WinNonlin version 8.3. Noncompartmental analysis was performed with Model 200, appropriate for extravascular drug administration. The linear/log trapezoidal rule was used to estimate the area under the plasma concentration‐–time curve (AUC), utilizing linear interpolation on ascending concentrations and logarithmic interpolation for declining concentrations. All PK parameters were estimated using nominal dose levels and nominal sampling times relative to dose administration, ensuring standardized analysis based on planned treatment and sampling schedules.

2.5. Simufilam plasma concentration at EEG study termination

Mouse EDTA‐K2 plasma samples were analyzed for simufilam, according to the Worldwide Clinical Trials non‐GLP qualitative assay on a UPLC SiL40 with an API5000 instrument. The method is detailed in the Supporting Information.

2.6. EEG headmount implantation

Mice were implanted between P19 and P20 with an 8201‐DEP EEG Headmount (Pinnacle Technology, Inc., Lawrence, KS), consisting of four supradural screw electrodes and an indwelling insulated, stainless steel wire electrode. The method is detailed in the Supporting Information.

2.7. EEG recording (P35–P54) and analysis

Tsc1 ^GFAP cKO male and female mice (15 mice per group) were split randomly into four groups that received vehicle, 5, 10, or 20 mg/kg/dose simufilam twice daily (BID) from P22 to P56. Mice were implanted with EEG electrodes at P19–P20 and EEG monitored from P35–P54 (13 days of treatment before 20 days of EEG with treatment). Blood was collected 1 h after the last dosing to quantify simufilam plasma concentration (i.e., exposure). Only animals that were fully recovered from surgery (healthy‐looking and displaying normal behaviors such as eating, grooming, exploring, and nesting) were used. Animals were housed individually from the time they underwent surgery and throughout EEG recordings (tethering at P35) in cylinder cages. Spontaneous EEG was recorded continuously using the Pinnacle Technology 8206 or 8401 data conditioning and acquisition system (DCAS), which performed secondary amplification and filtering before sending data to Pinnacle's Sirenia Acquisition software. The DCAS had 14‐bit resolution, user‐selectable gain (5000–100 000×), adjustable sampling rate (1000 Hz), and software configurable filters (low‐pass: 500 Hz; high‐pass: .5 Hz). EEG was recorded using 10x gain pre‐amplifiers. The pre‐amplified tether was connected to a low‐torque commutator (swivel) mounted above the cage allowing unencumbered freedom of movement.

EEG data were collected from frontal cortex, parietal cortex, and hippocampus. Analysis was performed for tonic–clonic seizures. Only data from the frontal cortex were analyzed, as these were sufficient for evaluating seizure activity. Moreover, frontal cortical recordings are considered the most translational from rodent to human in terms of anatomic electrode placements. ³⁰ Seizure count and duration were identical in all recorded channels. After exporting the data into an EDF file, data were read into the EEG analysis software for visualization, processing, and analyses. Electrographic seizures were identified by their characteristic pattern of discrete periods of rhythmic spike discharges that evolved in frequency and amplitude lasting >10 s, typically ending with repetitive burst discharges and voltage suppression. These were defined as having very stereotypical beginning, end, and evolution in the middle, starting with low‐amplitude fast activity and high‐frequency spikes (tonic), which gradually evolved into a slower, bursting (clonic) phase followed by a severe voltage suppression often with rhythmic artifact superimposed, which represents respirations and is clearly visible because the EEG itself was suppressed. Seizures were detected manually by an observer that was blind to the treatment groups. Most typical seizures in Tsc1 ^GFAP cKO mice lasted 30–40 s. These criteria were developed based on published literature. ³¹ In the current study, some animals exhibited large hippocampal events that were either isolated from or mixed in with the more common bilateral tonic–clonic seizures present in all three channels. These events typically had large amplitude spikes in the hippocampal channel that were much larger than in the cortical channels. Such events, confirmed by video in a few animals to be non‐convulsive, were not scored. Individual seizure events were counted only if the EEG returned to normal interictal baseline activity for ≥30 s between events. This standard criterion was not applied to clustered seizures, which may represent status epilepticus‐like events and often preceded animal death. These clustered seizures were counted as separate events regardless of inter‐event interval. Seizure incidence on each day from P35–P39 and P45–P54 was tabulated for each mouse. Seizure data are reported as the number of seizures per 24 h.

2.8. Survival and body weight

Survival was tracked throughout the study from enrollment through tissue collection at P56 with twice‐daily survival checks performed throughout this period. Survival curves are represented as the percentage survival of each treatment group over the course of the study. Body weight was assessed once weekly throughout the study period.

2.9. Statistics

The graphs of experimental results were generated using GraphPad 10 Prism. p‐Values are reported to two significant digits (e.g., p = .0116 reported as p = .012). The p‐values smaller than .001 are reported as p < .001. A p‐value < .05 was considered statistically significant. Data normality was determined using the D'Agostino & Pearson test and the Shapiro–Wilk test. Statistical analyses of the western blot, survival, and weight data were also performed in GraphPad 10 Prism. A logrank (Mantel–Cox) test and logrank test for trend were performed for survival data with treatment as a factor. Summary survival data are reported as percentage of survival across days. Body weight was analyzed by a mixed‐effects model in which time was a repeated measure. Western blot data were tested with Welch's t‐test. All other analyses were conducted in R (version 4.3.3). For EEG analyses, descriptive statistics were used to summarize measurements across treatment groups. For each group of EEG data, the mean, median, standard deviation (SD), minimum, and maximum were reported (Tables S1, S2, S5, and S8). A two‐tailed paired Wilcoxon signed‐rank test with continuity correction was applied when comparing repeated measures of seizure frequency and total ictal duration per day (first vs last 5 days of EEG recordings) within each treatment group (vehicle and treated groups). Total ictal durations per recording day were log₂‐transformed [log₂(count + 1)] prior to visualization to accommodate the wide dynamic range while handling zero values. A multivariable mixed‐effects linear regression with random intercepts was used to assess the relationship between dose, time, and (i) seizure number and (ii) seizure duration. For seizure number, the outcome was the natural log–transformed daily seizure count, ln(Seizure + 1), and for seizure duration, the outcome was ln(Duration + 1), where Duration represents the total ictal duration (cumulative seizure duration) normalized by EEG days for a given mouse. In both models, the fixed‐effects structure was:

$$\begin{array}{c}\ln \left(\mathrm{Y}+1\right)={\mathrm{\beta }}_{0,\mathrm{I}}^{\prime }+{\mathrm{\beta }}_{0,\mathrm{I}}+{\mathrm{\beta }}_1\mathrm{L}5\mathrm{D}+\text{Dose5}\left({\mathrm{\beta }}_2+{\mathrm{\beta }}_3\mathrm{L}5\mathrm{D}\right)+\hfill \\ \text{Dose}10\left({\mathrm{\beta }}_4-{\mathrm{\beta }}_5\mathrm{L}5\mathrm{D}\right)+\text{Dose}20\left({\mathrm{\beta }}_6-{\mathrm{\beta }}_7\mathrm{L}5\mathrm{D}\right)\hfill \end{array}$$

with Y= seizure duration or frequency. β_0,i is a random intercept for each mouse, β'_0,i is a random intercept per day within each time category, L5D is a binary variable which is zero if the measurement was in the first 5 days and one if the measurement was in the last 5 days, and Dose5, Dose10, and Dose20 are binary variables indicating if the mouse subject was given doses of 5, 10, or 20 mg/kg simufilam. The given transformation was chosen because of increased linearity in a normal quantile–quantile plot. This model was chosen in order to account for both the repeated measures of each mouse subject and day and missing data due to mouse death. As indicated, interaction between dose and time as well as individual dose and time effects were considered. Analyses were performed in R Studio 2025.09.1 Build 401 using the lmer, lmerTest, and modelsummary libraries. Total ictal duration per day and condition was analyzed with a Kruskal–Wallis test for not normally distributed data followed by Dunn's post hoc tests. Fisher's exact test was used to assess the association between the ictal duration being above or below vehicle median and the dose of simufilam. Finally, for dose–response analyses, the relationship between treatment dose and seizure frequency during the last 5 days of EEG recordings was assessed using Spearman's rank correlation (robust to monotonic but non‐linear associations). In addition, a simple linear regression model (seizure frequency as the dependent variable and dose as the predictor) was used to provide descriptive estimates of the average change in seizure frequency per unit dose within the tested range and should not be interpreted as evidence of perfect linearity. Correlation coefficients, regression estimates, and associated p‐values were reported.

3. RESULTS

3.1. Simufilam PK

The PK of simufilam was analyzed following either a single administration at 10 or 50 mg/kg, or multiple administrations at 10 or 50 mg/kg once daily for 7 days via oral gavage. Figure 1 illustrates simufilam concentration as a function of the eight collection time points up to 24 h using linear and logarithmic plots. The maximal concentration was reached at ~30 min under all conditions, and the half‐life was about 1.7 h and 4.0–4.5 h for the 10 and 50 mg/kg doses, respectively, independent of treatment duration. The mean peak concentration (C_max, ±SD) was 846 (±75) ng/mL after a single 10 mg/kg administration and 1170 (±209) ng/mL after 7 days of administration. The mean area under the concentration–time curve (AUC_0–last) was 1590 (±180) ng·h/mL and 1814 (±191) ng·h/mL after a single 10 mg/kg administration and 7 days of administration, respectively. There was an ~5.5‐fold increase in both C_max and AUC when the dose was increased from 10 mg/kg to 50 mg/kg. Using the AUC_0–12 values, there was an ~12% accumulation at 10 mg/kg and 9% accumulation at 50 mg/kg over 7 days of administration.

FIGURE 1.

Simufilam pharmacokinetics (PK) electroencephalogram in CD1 male mice. (A, B) Simufilam concentrations (linear [A] and semi‐log graphs [B]) as a function of time of collection (hours) for the different treatments. n = 4 mice per group.

3.2. FLNA expression was elevated in Tsc1 ^GFAP cKO mice

Before performing electroencephalogram (EEG) recordings and testing the effect of simufilam on seizure activity, we examined whether FLNA expression was increased in Tsc1 ^GFAP cKO mice compared to Tsc1 WT mice. Immunoblotting for FLNA identified a significant 2.3‐fold increase in full‐length FLNA (280 kDa, p < .039, Welch's t‐test) and a 1.6‐fold increase in the 110 kDa FLNA fragment (p < .046, Figure 2A,B).

FIGURE 2.

FLNA levels and simufilam effects on weight and survival in Tsc1 ^GFAP cKO mice. (A) Immunoblot of FLNA and β‐tubulin on the same blot for Tsc1 WT and Tsc1 ^GFAP cKO mice (n = 3 per group). The ladder is on the left of the blot. The FLNA blot was stripped and probed for β‐tubulin. (B) Quantification of the blots. Each circle represents a mouse. Welch's t test. (C) Weight of the mice (N = 68 male and female) throughout vehicle or simufilam treatments. Body weight is expressed as group mean ± SEM. Mixed‐effects analysis with row factor (weight) p < .0001 F(1.92, 102.5) = 521, column factor (treatment) p = .78 F(3, 66) = .35, and row by column factor p = .27, F(5.75, 102.5) = 1.28. (D) Percent of surviving animals at each time point from the beginning of treatment. Deaths were in vehicle 7 of 17; 5 mg/kg, 5 of 18; 10 mg/kg, 5 of 17; and 20 mg/kg, 6 of 18 (cKO, conditional knockout; GFAP, Glial‐Fibrillary Acidic Protein; WT, wild‐type)

3.3. Simufilam did not affect weight gain or mortality but prevented the worsening of seizure activity in Tsc1 ^GFAP cKO

For these experiments, Tsc1 ^GFAP cKO mice known to exhibit spontaneous seizures around P21 were utilized. ²⁸ , ²⁹ This TSC mouse model is considered severe, as ~50% mortality occurred by P60 (Figure 2D). These mice were treated with either vehicle (PBS) or simufilam at 5, 10, or 20 mg/kg twice daily (BID) for 35 days, starting at P22 until P56 (Figure 2C). Of note, simufilam administration at either dose did not affect weight gain in male and female mice combined (statistical values are provided in figure legends and in the figures, Figure 2C). In addition, simufilam did not significantly improve mouse survival, although fewer mice died in the treated groups (mouse number in figure legend, Figure 2D). When male and female mice were analyzed separately, male (N = 35) body weight increased over time across all groups with similar trajectories, consistent with no effect of simufilam on weight (Figure S1A). In female mice (N = 35), body weight also increased over time across all groups with broadly similar trajectories (Figure S1A). Although small separations in mean curves are visible at some time points (but not the final time point), the study is underpowered to detect small‐to‐moderate differences in body weight trajectories. Thus, modest treatment effects of simufilam on female body weight cannot be excluded. Mortality occurred in both sexes across dose groups except at 5 mg/kg (0/7 female vs 5/11 male). No statistically significant differences in survival were detected between vehicle and simufilam‐treated groups within either sex (Figure S1B). However, the study was underpowered, particularly for female mice, to detect small‐to‐moderate differences in mortality within sex and to formally assess sex differences in survival.

Following 13 days of treatment, EEG recordings were performed for 20 days (from P35 to P54, Figure 3A). A representative EEG trace of a tonic–clonic seizure is illustrated in Figure 3B. Using heatmap plots, a striking observation was that vehicle‐treated and 5 mg/kg‐treated Tsc1 ^GFAP cKO mice displayed a clear increase in seizure frequency over time (Figure 3C; descriptive values are in Tables S1 and S2). In contrast, this increase was less pronounced in the 10‐ and 20‐mg/kg‐treated mice. In addition, 3 of 14 mice in the 20‐mg/kg‐treated group reached seizure freedom in the last 5–8 days. To quantify these observations, we compared seizure frequency during the first vs last 5 days of recording, which captured the period when the worsening of seizures was most evident in vehicle‐treated mice, as visualized in the heatmap representation in Figure 3C. A paired analysis showed that the increase in seizure frequency was significant in both the vehicle‐treated and 5‐mg/kg‐treated mice (p < .02), but not in the 10‐mg/kg or 20‐mg/kg‐treated groups (Wilcoxon signed‐rank test, statistical values in Figure 3D). In addition, although no mice in the control group displayed a decrease in seizure activity and most showed an increase (9/11, 82%), in the 20‐mg/kg‐treated group, 4 of 11 mice showed a reduction in seizure activity (orange lines) and only 5 of 11 (45%) exhibited an increase. The orange highlighting is provided for visualization only and is not intended to denote a statistically defined responder subgroup. With respect to sex, due to the low number of mice per sex and high within‐group variability in seizure frequency, the study is underpowered to assess sex‐specific effects of simufilam on seizure activity.

FIGURE 3.

Simufilam significantly prevented the increase in seizure number in Tsc1 ^GFAP cKO mice. (A) Experimental design. (B) A representative EEG trace of a tonic–clonic seizure. (C) Heatmap of the daily number of seizures per mouse during EEG recordings in each condition. EEG monitoring spanned 20 days, but seizure counts were quantified for the first 5 and last 10 days, and comparative analysis was performed between the first and last 5 days. Gray squares: Mice had died. (D) Top: Mean daily number of seizures for all the mice in each condition. Plot of the mean number of daily seizures during the first and last 5 days of EEG recording for each condition. p‐Values are listed at the top of each graph and obtained using a Wilcoxon signed‐rank test. The numbers of mice were vehicle, 11; 5 mg/kg, 12; 10 mg/kg, 13; and 20 mg/kg, 11. The orange and blue highlighting are provided for visualization only. (E) Plot of the mean number of seizures during the first and last 5 days and per group. A mixed‐effects analysis provided the p‐values listed on the plot. The numbers of mice from the first 5 days to the last 5 days were for vehicle (15 to 11), 5 mg/kg (14 to 13), 10 mg/kg (15 to 13), and 20 mg/kg (14 to 11) (cKO, conditional knockout; EEG, electroencephalogram; GFAP, Glial‐Fibrillary Acidic Protein; WT, wild‐type)

To utilize data from all mice, including those that died before study completion, we performed a mixed‐effects analysis comparing seizure activity during the first and last 5 days (L5D) of EEG recordings (Figure 3E). The results of the mixed‐effects linear regression model for the natural log of one plus seizure number are shown in Tables S3 and S4. The model attained a conditional $R^2$ of .31. The significant predictors of this model were $L5D$ and the interaction term $\mathit{\text{Dose}}20*L5D$. All else being equal, there was a significant average increase in one plus the seizure number from the first 5D to the L5D of the study by a factor of 1.87 (p < .0001; 95% confidence interval [CI] 1.43–2.46). In the last 5‐day epoch, the 20 mg/kg dose significantly reduced one plus the seizure number by a factor of .57 (p = .0015; 95% CI .40–.80).

3.4. Simufilam reduced the total ictal duration by the end of treatment

Next we measured the cumulative time spent experiencing seizures, referred to as the total ictal duration, per mouse during the first and last 5 days of recordings (Figure 4). Mice with at least 2 days of EEG were included. Heatmap plots revealed that vehicle‐treated and 5‐mg/kg‐treated Tsc1 ^GFAP cKO mice displayed a clear increase in total ictal duration over time (Figure 4A). Considering that several mice died before reaching the end of the treatment phase, we divided the total ictal duration by the number of EEG recording days and compared the resulting values across groups. A paired analysis comparing the first vs last 5 days of EEG monitoring showed that the increase in total ictal duration was significant in vehicle‐ and 5‐mg/kg‐treated mice (p < .02), but not in the 10‐ or 20‐mg/kg‐treated groups (Wilcoxon signed‐rank test, statistical values in Figure 4B, descriptive values in Table S5). Although one mouse in the 5 mg/kg group had a particularly large total ictal duration of 71 min (or 6.17 as log₂(count + 1) vs a mean of 4.96 for all the other mice in this group), excluding this data point did not change the results of the analysis (p = .012 with and p = .024 without). To account for all the mice, including those that died prior to the last 5 days of recordings, we performed a mixed‐effects linear regression model for the natural log of one plus ictal duration over EEG days (Figure 4C); the results are shown in Tables S6 and S7. The model attained a conditional $R^2$ of .2306. The significant predictors of this model were $L5D$ and the interaction term $\mathit{\text{Dose}}20*L5D$. All else being equal, there was a significant average increase in one plus ictal duration over EEG days from the first 5 days to the last 5 days by a factor of 5.24 (p = .0011; 95% CI 2.07–13.26). In the last 5‐day epoch, the 20 mg/kg dose significantly reduced one plus ictal curation over EEG days by a factor of .17 (p = .0032; 95% CI .05–.55).

FIGURE 4.

Simufilam prevented the increase in total ictal duration in Tsc1 ^GFAP cKO mice. (A) Heatmap of the total ictal duration per day and per mouse. Asterisks mark 2 days during which the mouse had total ictal durations of 218 and 104 min. Gray squares: Mice had died. Mice with a minimum of 2 days of EEG were included. (B) Plot of the mean number of total ictal duration per day during the first and last 5 days of EEG recording for each condition. p‐Values are listed at the top of each graph and obtained using a Wilcoxon signed‐rank test. The numbers of mice were 11 for vehicle, 12 for 5 mg/kg, 13 for 10 mg/kg, and 12 for 20 mg/kg. Orange line: Values going down, blue line: Unchanged value, and black lines: Values going up. (C) Plot of the mean ictal duration per day during the first and last 5 days and per group. A mixed‐effects analysis provided the p‐values listed on the plot. The numbers of mice from the first 5 days to last 5 days were for vehicle (15 to 11), 5 mg/kg (14 to 13), 5 mg/kg (15 to 13), and 20 mg/kg (14 to 11). (D) Total ictal duration per day and per condition for the last 5 days of EEG recording. Data were not normally distributed (D'Agostino and Pearson test); therefore, we used a Kruskal–Wallis test followed by a post hoc Dunn's test. No significant differences were found with the latter. Violin plots: Median and quartiles. Circles represent mice (number as in C). (E) Graphs showing the number of mice in each dose group with ictal durations below or above the vehicle median. Vehicle median indicated the reference threshold. Fisher's exact test (cKO, conditional knockout; EEG, electroencephalogram; GFAP, Glial‐Fibrillary Acidic Protein; WT, wild‐type)

When comparing the last 5 days across groups, a non‐parametric Kruskal–Wallis test was statistically significant (p = .019, Figure 4D, values in Table S8), suggesting that there is an overall difference despite no significant pairwise differences. The lack of significance between groups with Dunn's post hoc test suggests that the sample size or variability might limit the power of pairwise tests. Upon examination of the ictal duration graphs (Figure 4D), we observed that most mice in the 10‐mg/kg‐ and 20‐mg/kg‐treated groups displayed total ictal durations below the vehicle median. We therefore performed a Fisher's exact test using the vehicle median as the reference threshold. As simufilam dose increased, the proportion of mice with ictal durations exceeding the vehicle‐group median decreased, indicating a significant dose‐related shift toward shorter ictal durations (Fisher's exact test, p = .02, Figure 4E).

3.5. Higher doses and plasma concentrations of simufilam are associated with greater seizure reduction

The previous data suggest that simufilam at 5 mg/kg did not have a significant effect on seizure activity, whereas the 10 and 20 mg/kg did. We thus plotted the mean number of seizures per mouse (last 5 days of EEG) as a function of simufilam dose and found a significant Spearman correlation between the number of seizures and dose level (correlation coefficient of −.34, p = .017, Figure 5A). Based on a linear regression model, a 10 mg/kg increase in dose was associated with a decrease of 1.35 seizures per day during the last 5‐day period. To evaluate exposure in this mouse model, we collected blood samples 1 h after the last dosing and measured the plasma concentration of simufilam. The mean exposures (±SD) were 150 (±85), 448 (±190), and 1042 (±598) ng/mL for 5, 10, and 20 mg/kg, respectively. As for the simufilam doses, we found a significant Spearman correlation between the number of seizures and plasma concentration (correlation coefficient of −.39, p = .012, Figure 5B). Based on a linear regression model, a 100 ng/mL increase in plasma concentration was associated with a decrease of .2 seizures per day during the last 5‐day period.

FIGURE 5.

Higher simufilam doses and plasma levels are associated with reduced seizure activity in Tsc1 ^GFAP cKO mice. (A, B) Plot of the mean number of seizures measured during the last 5 days of EEG recordings against simufilam doses (A) and plasma concentrations (B). The colors represent dose groups: Gray (vehicle, 11 mice (A) and (B)), orange (5 mg/kg, 13 (A) and 10 (B) mice), blue (10 mg/kg, 13 (A) and 12 (B) mice), and green (20 mg/kg, 11 (A) and 9 (B) mice). Circles represent mice (cKO, conditional knockout; EEG, electroencephalogram; GFAP, Glial‐Fibrillary Acidic Protein; WT, wild‐type)

As reported for the other parameters (e.g., weight and survival), sex‐based comparison is limited due to the small sample size. However, at the highest dose, a subset of females showed higher plasma exposures than males, an observation that warrants further evaluation in a larger cohort (Figure S2).

4. DISCUSSION

In the current study, we show that simufilam treatment effectively prevented the worsening of seizure activity, and that higher simufilam doses and exposure led to improved seizure outcomes in a TSC model characterized by diffuse cortical malformation. These findings are consistent with data obtained in a mouse model of FCDII exhibiting focal cortical malformations due to the expression of a constitutively active Ras homolog enriched in brain protein (RHEB), the canonical mTOR activator. ²⁵ FCDII and TSC are mTORopathies exhibiting similar pathologies, in particular, the presence of focal cortical malformations responsible for seizure activity. ³² , ³³ In the Tsc1 ^GFAP cKO mice, simufilam was administered twice daily, considering its relatively short half‐life of about 1.7 h in mice. Simufilam had no effect on mouse weight as previously reported, ²⁵ and fewer mice died in the simufilam‐treated group. Regarding seizure activity, the effect of simufilam was progressive, as shown on the heatmap illustrating the number of seizures over time per mouse. Notably, the heatmap was plotted after the mice had already received 13 days of simufilam treatment, at which point the seizure frequency was similar across both the vehicle and treated groups. After 25–30 days of treatment, simufilam at 10 and 20 mg/kg prevented the progressive increase in seizure activity observed in the vehicle‐treated Tsc1 cKO group. However, when we restricted the analysis to the final 5 days of EEG monitoring and compared dose groups to vehicle, the differences in seizure activity and total ictal duration did not reach statistical significance at the predefined α, likely reflecting variability and reduced power due to mortality in the vehicle group. In terms of dose effect, the lowest dose of 5 mg/kg did not prevent the worsening of seizure activity, whereas the 10 and 20 mg/kg doses did. Two non‐mutually exclusive factors may contribute to this threshold effect. First, at a sub‐therapeutic dose, simufilam may improve structural abnormalities (e.g., reduced dendritic overgrowth observed at 6 mg/kg in our mTOR‐related RHEB model), but this may be insufficient to suppress network hyperexcitability. Second, systemic exposure increases nonlinearly across doses in this range, such that 10 mg/kg achieves substantially higher exposure than 5 mg/kg (threefold higher, 448 ± 190 ng/mL vs 150 ± 85 ng/mL), which may be required for more consistent network normalization. Together, these considerations support the conclusion that effective seizure control requires higher exposures (10–20 mg/kg) to restore circuit stability and avoid a threshold response in which partial remodeling is insufficient to shift the network to a stable state. In addition, there was a significant correlation between the dose and the number of seizures, with higher simufilam doses leading to better seizure outcomes. This correlation was even stronger when using the plasma concentrations of simufilam (a covariate in the analysis) measured in animals that survived to the end of the EEG study. These exposure data are important for future modeling studies to estimate the simufilam exposure required to potentially reduce seizure activity in humans.

The mechanism of action of simufilam remains to be fully elucidated. Simufilam has been proposed to interact with the actin‐crosslinking protein, FLNA, as recent work provides indirect evidence showing that restoring proper FLNA levels in the FCDII mouse model produces similar effects on neuronal morphology and seizure activity as simufilam. ²⁵ Both manipulations reduced neuronal abnormalities, including soma overgrowth and increased dendritic complexities of mutant neurons, leading to a restoration of network connectivity. The exact molecular mechanism of action remains unclear, but since simufilam was shown to not modify FLNA levels or mTOR activity, ²⁵ it may act by affecting the binding of FLNA to one or more of its numerous binding partners. In Tsc1 ^GFAP cKO mice used here, FLNA levels were also elevated, as reported in other mouse models of TSC and FCDII, as well as in vitro conditions. ²⁴ , ²⁵ , ³⁴ Whether restoring FLNA levels in Tsc1 ^GFAP cKO mice would reduce seizure activity could be a subject of future studies.

Our study has several limitations: (1) Because pharmacokinetic profiling was performed only in male CD1 mice, potential sex‐dependent differences in simufilam disposition may not be fully captured. In our study, a subset of females at the highest dose showed higher plasma exposures than males, but this trend did not reach significance due to limited power, and should be evaluated in a larger, sex‐balanced cohort. (2) Although the study design was to quantify seizure burden across two standardized 5‐day epochs, a subset of mice died during the recording interval, leading to truncated datasets (≥2 days in some cases). Because seizure counts can vary day‐to‐day, and because death may be related to disease severity, unequal recording duration could skew group summaries and complicate interpretation of early‐ vs late‐phase comparisons. To mitigate this issue and retain information from all available animals, we included mice with at least 2 days of EEG and analyzed daily seizure counts using a mixed‐effects framework with random effects to account for within‐mouse repeated measures and missing data attributable to death. In parallel, total ictal duration was normalized by the number of recording days to address unequal exposure time. Although these approaches reduce bias, the possibility of residual survival‐related bias cannot be fully excluded, and future studies designed to minimize monitoring loss or increase animal number will strengthen confidence in the magnitude of treatment effects. (3) Additional behavioral phenotyping, spike analyses, or testing in additional TSC models would be valuable, but we believe that these are beyond the scope of the present work and are better suited to future follow‐up studies. (4) Given the low number of mice per sex and the high within‐group variability in seizure frequency, this study was not powered to detect sex‐specific effects of simufilam on seizure activity.

In conclusion, our findings show that simufilam treatment effectively improves seizure outcome in a mouse model of TSC, as was shown previously in a mouse model of FCDII. The treatment effect of simufilam in an established TSC mouse model supports further clinical investigation of this potentially novel therapeutic for treating seizures in patients with TSC.

CONFLICT OF INTEREST STATEMENT

Drs. Thornton, Stansley, and Bordey are employees of Filana Therapeutics, Inc. Drs. de Kater, Santos, and Yuan are consultants to Filana Therapeutics, Inc.

ETHICS STATEMENT

All procedures were performed in accordance with PsychoGenics and WuXi AppTec Institutional Animal Care and Use Committee regulations. We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

Supporting information

Figure S1:

DATA AVAILABILITY STATEMENT

The data that support the findings of this study will be provided upon request.