The human SCN1A sequence cloned in the pCDM8 vector as described earlier (Hedrich et al., 2014) was corrected for two-point mutations, namely, Glu650Val and Ser1969Ala (Peters et al., 2016) by site-directed mutagenesis (Agilent Technologies, Santa Clara, CA, USA). The WT open reading frame included the canonical SCN1A adult isoform 2 (total length of 5997bp) equivalent to transcript NM_006920.5. The missense variant p.(Ala1783Val) was engineered into the human Na_V1.1 channel by site-directed mutagenesis using PCR with Pfu polymerase (Promega; mutagenesis primers 5′ to 3′; F: ATG TAC ATC GTG GTC ATC CTG GAG AAC TTC AGT, R: AGG ATG ACC ACG ATG TAC ATG TTC ACC ACA). The introduced variant was verified, and further variants were excluded by sequencing the whole SCN1A cDNA. Plasmid purification was performed using Escherichia coli One Shot TOP10/P3 (Thermo Scientific, Waltham, MA, USA).

tsA201 cells were cultured in the Dulbecco's modified Eagle nutrient medium (Thermo Scientific, Waltham, MA, USA) supplemented with pyruvate, 10% v/v fetal bovine serum (Pan Biotech, Aidenbach, Germany), and 2 mM l-glutamine (Merck Millipore, Burlington, MA, USA) at 37°C in a 5% CO₂ humidified atmosphere. For transfection, cells with a passage from P15 to P22 were used. Before transfection, 800,000–1,000,000 cells were split in 35-mm Petri dishes. Six hours later, cells were transfected by the following standard protocols: 4 μg of WT or mutant SCN1A cDNA, encoding the Na_V1.1 channel-α-subunit, and 0.4 μg each of the human β₁- and β₂-subunits of voltage-gated Na_V channels, which had been previously modified to express either green fluorescent protein (pCLH-hb1-EGFP) or a CD8 marker (pCLH-hb2-CD8) to label cells expressing both subunits (Liao et al., 2010), were added to 250 μl of Opti-MEM (Gibco) supplemented with 7.5 μl of Mirus TransIT reagent (Mirus Bio LLC, Madison, WI, USA). After 20 min, the transfection mixture was added to the cells.

Coronal brain slices from conditional B6(Cg)-Scn1a^tm1.1Dsf/J mice (JAX stock #026133) of either sex were obtained at P4-5. Mice were quickly decapitated, and cranium, brain stem, and hindbrain were removed. The forebrain was placed in ice-cold artificial cerebrospinal fluid (aCSF) bubbled with carbogen (95% O₂/5% CO₂). The aCSF contained the following [in mM]: 118 NaCl, 3 KCl, 1.5 CaCl₂, 1 MgCl₂, 25 NaHCO₃, 1 NaH₂PO₄, 30 glucose, and pH 7.4 with an osmolarity of 310–320 mOsm/kg. Forebrains were glued with the cerebellar end facing downward on an agar block, and 350 μm thick slices were cut using a vibratome (Microm, H650 V) in ice-cold aCSF. Slices were placed in 36°C warm aCSF (bubbled with carbogen) for 10 min. Coronal slices containing the somatosensory cortex were transferred to Millicell cell culture inserts (Merck Millipore, Burlington, MA, USA) floating on the slice culture medium (Minimum Essential Medium Eagle with 20% horse serum, 1 mM l-glutamine, 0.00125% ascorbic acid, 0.001 mg/ml insulin, 1 mM CaCl₂, 2 mM MgSO₄, 1% penicillin/streptomycin, 13 mM glucose, pH 7.28, and osmolarity 320 mOsm/kg).

Whole-cell voltage-clamp recordings of tsA201 cells were performed 48 h after transfection from cells expressing all three sodium channel subunits indicated by an inward sodium current (α-subunit), anti-CD8 antibody-coated microbeads (Dynabeads M450, Dynal) (β₁-subunit), and green fluorescence (β₂-subunit). Cells were split in 35-mm Petri dishes 1 h prior to recordings, and each dish was used for up to 1 h after transfer to the patch-clamp setup. The extracellular bath solution contained the following [in mM]: 140 NaCl, 4 KCl, 1 MgCl₂, 2 CaCl₂, 5 HEPES, and 4 glucose. pH was adjusted with HCl to 7.4, and osmolarity was 300–305 mOsm/kg. Patch pipettes were pulled from borosilicate glass (Science Products GmbH) using a P97 Puller (Sutter Instruments), with resistances of 1.5–3.0 MΩ. Intracellular solutions for patch pipettes contained the following [in mM]: 5 NaCl, 2 MgCl₂, 5 EGTA, 10 HEPES, 130 CsF, pH 7.4 (adjusted with CsOH), and osmolarity 290–295 mOsm/kg.

Whole-cell current-clamp recordings of excitatory pyramidal cells (PCs) and FS-INs in the somatosensory cortex of organotypic cortical slice cultures of conditional or WT animals were performed 7–14 days after adeno-associated viral transduction with 1 μl/slice of AAV8-hSyn-Cre-GFP (1 × 10¹³ VG/ml; SignaGen Laboratories, Frederick, MD, USA). Slices were positioned in a submerged-type recording chamber (Luigs & Neumann, Ratingen, Germany), continuously superfused with oxygenated recording aCSF, and maintained at a temperature of 33 ± 1°C. Transduced neurons were visualized using an Axioskop 2FS microscope (Carl Zeiss, Oberkochen, Germany). Patch pipettes had resistances of 2.5–5.5 MΩ. Intracellular solutions contained the following [in mM]: 140 K-gluconate, 1 CaCl₂, 10 EGTA, 2 MgCl₂, 4 Na₂-ATP, 10 HEPES, 0.45% biocytin (Sigma-Aldrich, B4261), pH 7.2, and osmolarity of 300–310 mOsm/kg. PCs and FS-INs were used for current-clamp recordings. Cell identity was determined via morphological and electrophysiological properties. To block postsynaptic AMPA receptor-driven depolarization waves, 20 μM CNQX (Sigma-Aldrich, St. Louis, MO, USA) was added to the bath solution before recording.

Following recordings, slices were fixed for 24 h in 4% paraformaldehyde (Morphisto GmbH, Offenbach am Main, Germany), washed in DPBS (Thermo Fisher Scientific, Waltham, MA, USA, Cat. 12559069), and blocked with 1% normal goat serum with 0.2% Triton X-100 for 2 h. AAV-positive cells were stained with rabbit anti-GFP IgG primary antibody (dilution 1:1,000, Invitrogen, Cat. A-11122) overnight at 4°C. Subsequently, slices were washed in DPBS and counterstained with goat anti-rabbit IgG secondary antibody Alexa Fluor-488 (dilution 1:200, Invitrogen, Cat. A-11008) and Streptavidin-Cy3 (dilution 1:100, Sigma, Cat. S6402) for 3 h at room temperature. After washing, slices were mounted with DAPI Fluoromount-G (Southern Biotech, Birmingham, AL, USA, Cat. 011-20). Immunohistological images of stained slices were acquired on a Zeiss Axiovert 200M microscope (widefield images) and a Leica SP8 microscope (confocal images).

Simulations were run using custom Python 3.7 software. A single-compartment conductance-based model with a cylindrical shape with length L and diameter d was used. The model is based on Pospischil et al. (2008) and consists of two sodium currents (I_Na,wt and I_Na,mut), a delayed rectifier potassium current I_K, a M-type potassium current I_M, and a leak current I_leak:

with membrane capacitance Cm=1 μF/cm2, input current I_input, maximal conductance g_i, reversal potential E_i, gating variables m, h, n, and p with dynamics

and the slow inactivating gating variable s with

where ẋ_i is the derivative of the gating parameter with respect to time, α_i(V) is the opening rate, and β_i(V) is the closing rate of the respective gate x_i. Steady-state curves and time constants are given by x_{∞, i}(V) = α_i(V)/[α_i(V) + β_i(V)] and τ_i(V) = 1/[α_i(V) + β_i(V)]. The dynamics of the slow inactivating gating variable s followed

As suggested in the study by Pospischil et al. (2008), parameters for gating variables m, h, and n were taken from Traub and Miles (1991) and for gating variable p from Yamada (1989). We constructed the slow inactivation gate s to be similar to observed kinetics as suggested by Vilin and Ruben (2001). To simulate different neuron types, the conductivities of the ionic currents were adapted from the study by Pospischil et al. (2008). All parameters that differ from the study by Pospischil et al. (2008) are summarized in Table 1. Effects of variants were simulated with changes in the shifting parameter shift_i for the activation and slow inactivation sodium gates with shift_i = 0 mv unless stated otherwise.

To simulate the effects of heterozygous variants in SCN1A on the firing behavior of PCs and interneurons, different ratios of Na_V1.1 and Na_V1.6 expression were modeled in the respective neuron types with the assumption of equal gating properties in both WT sodium channels (g_Na,wt = g_NaV1.1/2 + g_NaV1.6) and potentially altered Na_V1.1 channels (g_Na,mut = g_NaV1.1/2). The ratio of assumed sodium conductance was g_NaV1.6 = 5g_NaV1.1 for pyramidal neurons and g_NaV1.6 = 0.8g_NaV1.1 for FS-INs based on expression data from cortical mouse neurons (Yao et al., 2020). To investigate how each observed change in gating parameters contributes to the observed change in firing behavior, we constructed several variant models, one combined variant model with all parameter changes, and for each adapted parameter, one model with only one respective change.

The excitability of these models was computed with the f−I curve. We simulated square pulse stimuli between 0 and 500 pA with increments of 5 pA. Action potential (AP) frequency of each step was computed by dividing the number of measured spikes, reduced by 1, with the time difference between the first and the last spike:

All data are displayed as mean ± SEM. Statistical testing was performed via unpaired, two-tailed t-test for normally distributed data and Mann–Whitney U test for non-normally distributed data and n indicates the number of cells. Data normality was tested by using the D'Agostino and Pearson normality test. Mean-, n- and p-values are shown in Supplementary Tables 1–3.

All data were analyzed using pClamp 10.703 (Molecular Devices, LLC, San Jose, CA, USA), Microsoft Excel (Microsoft Corporation, Redmond, WA, USA). Statistics were performed using GraphPad 7 (GraphPad Prism, San Diego, CA, USA).

Wild-type or mutant Na_V1.1 channels were transfected into tsA201 cells, and whole-cell patch recordings were performed. Representative raw current traces are shown in Figure 1A. While peak current density was not changed for mutant channels in comparison to WT (Figure 1B), activation and inactivation properties were markedly altered. We found a significant right shift of the half-maximal conductance indicating that mutant channels open and reach their maximal activation at more depolarized potentials (Figure 1C). Fast inactivation was unaffected as revealed by comparable voltage-dependence of steady-state fast inactivation (Figure 1C), time constant τ (Figure 1E), and recovery from fast inactivation (Figure 1D) for WT and Na_V1.1^A1783V channels. In contrast, entry into slow inactivation was accelerated (Figure 1F), and voltage-dependence was shifted to more hyperpolarized potentials (Figure 1G). Overall, Na_V1.1^A1783V channel LOF became very obvious on rapid successive depolarizing stimulations at 40 Hz revealing a markedly increased use-dependence with consecutively pronounced “rundown” of the sodium current (Figure 1H).

Full haploinsufficiency is considered as the genetic mechanism underlying the majority of SCN1A Dravet variants. About half of the described variants lead to protein truncation with reduced protein expression in affected neurons and likely markedly reduced conductance (Marini et al., 2011). Our recordings revealed clear biophysical LOF changes of Na_V1.1^A1783V compared with WT, however no reduction of peak current density. Then, we asked how these combined functional alterations translate to the neuronal level and how they compare to the heterozygous knockout condition. To address these questions, we built a single-compartment conductance-based model with the combined biophysical alterations reflecting intrinsic and firing properties of either cortical FS-INs or excitatory PCs neurons. Differential Na_V1.1–Na_V1.6 expression ratios of FS-IN and PCs and either the WT (Na_V1.1^+/+), the mutant (Na_V1.1^+/A1783V), or the full haploinsufficiency (Na_V1.1^+/−) condition were implemented. The parameters for the different model types are summarized in Table 1. Additionally, we simulated altered voltage-dependence of activation as well as altered voltage-dependence and kinetics of slow inactivation separately to dissect their joint or exclusive impact on neuron action potential firing in comparison to the WT and the full haploinsufficiency condition (Figure 2A).

In the interneuron model, solely shifting the activation curve to more depolarized membrane potentials caused the most severe reductions of the firing rates in comparison to the WT, closely followed by the Na_V1.1^+/A1783V and Na_V1.1^+/− conditions. Shifting the slow inactivation curve to more hyperpolarized potentials reduced firing rates to a lesser extent (Figure 2B, left). Although the faster slow inactivation model was the only simulation yielding an accelerated rundown of sodium current amplitudes as had been observed in tsA201 cells expressing Na_V1.1^A1783V (Supplementary Figure 1C), it showed almost no impact on firing rates in comparison to WT (Supplementary Figures 1A,B). The firing properties of all pyramidal models remained largely unaffected (Figure 2B, right; Supplementary Figure 1B).

To better understand why shifting the activation curve for half of the Na_V1.1. channels (to mimic the heterozygous patient situation) had a stronger effect on neuronal firing than the simulation of the heterozygous protein knockout, we analyzed the current dynamics of the first action potential in more detail (Figures 2C–G). Although on a first glance, the action potential waveforms seemed to be rather similar (Figure 2C), the sodium activation gate (Figure 2D) of the model with a shifted activation curve (orange) opened later and closed earlier compared with the WT (black) and the Na_V1.1^+/− models (blue). The resulting sodium current was reduced during AP initiation and rising phase of the action potential compared to the WT simulation but was similar during the falling phase of the action potential (AP). Only at the very end of the AP, the sodium current appeared again slightly reduced. In contrast, the Na_V1.1^+/− model depicted a generally reduced sodium current (Figure 2E), which was followed by a pronounced reduction of the delayed rectifier potassium current. This effect on the potassium current was almost absent when modeling a right-shifted activation curve of Na_V1.1. (Figure 2F). In combination with the observed effects on the sodium current, the right-shifted activation curve model showed a greater reduction of the sodium-to-potassium current ratio in the final phase of the afterhyperpolarization than the Na_V1.1^+/− model (Figure 2G). This increased after-hyperpolarization lead then to a delay of the next action potential.

To confirm the simulated effects of Na_V1.1^A1783V on neuronal firing, we performed whole-cell patch-clamp recordings in cortical brain slice cultures derived from P4-5 heterozygous B6(Cg)-Scn1a^tm1.1Dsf/J mice and WT littermates. Cultures were virally transduced at day 1 in vitro with AAV8 Cre-recombinase-GFP under the human synapsin promoter to induce pan-neuronal recombination and subsequent expression of the Na_V1.1^+/A1783V variant. Since Scn1a is upregulated in the postnatal period only from P11 onward (Cheah et al., 2013), this early expression of Cre-recombinase allowed for recombination of transduced neurons carrying the floxed Scn1a^A1783V allele before endogenous upregulation of the Scn1a gene. This approach ensured activation of the variant following the endogenous expression time course of the Na_V1.1 channel. Notably, 7–14 days after transduction, GFP+ neurons were patched (Supplementary Figure 2). While the firing of PCs was not altered (Figures 3A,B), we found a LOF in FS-INs (Figures 3C,D). Na_V1.1^+/A1783V expressing FS-INs displayed a significantly increased rheobase, reduced maximum firing frequency, and reduced input resistance compared to WT controls (Supplementary Table 2). Similarly, we could detect a trend for a reduction of the input resistance of PCs (Supplementary Table 3). However, the firing properties of PCs were not significantly altered. AP threshold, AP amplitude, AP rise time, AP half-width, and afterhyperpolarization amplitude within evoked AP trains were indistinguishable between neurons (FS-INs and PCs) recorded from slices of WT and mutant animals. We applied the observed changes in input resistance to the in silico model (Figure 2, brown). These simulations combining changed gating and membrane properties predicted an even stronger LOF in FS-INs while only tentatively affecting the firing of simulated PCs.