Non-syndromic autism SCN2A variants selectively exert dominant-negative effects on Na v 1.2 channels

sodium channels; autism spectrum disorder; epilepsy; developmental epileptic encephalopathies; neurodevelopmental disorders; excitability. preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 2

The voltage -gated Na+ channel Na v1.2 has a key role for the initiation and propagation of action potentials and therefore in neuronal excitability in brain development and function. Genetic variants of the encoding gene SCN2A cause various neurodevelopmental phenotypes with infantile-childhood onset. Here, we investigated the functional impact on hNav1.2 function of 15 variants associated with pure non-syndromic ASD (nsASD), ASD with epileptic activity , developmental and epileptic encephalopathy or schizophrenia. Only nsASD variants caused a complete loss of function hNa v1.2 channels when expressed alone in tsA -201 cells and cultured neocortical neurons. Co-expression of the WT and mutant channels mimicking heterozygosis revealed that ASD mutants induce a dominant negative e ffect. Using different strategies to impair the domains of the channels that have been suggested to be implicated in the interaction of two NaV α subunits, we reversed the dominant negative effect of ASD mutants on WT channels. These findings identify in heterologous systems a mechanistically distinct class of SCN2A variants implicated in nsASD, defined by dominant -negative loss of Nav1.2 function, with potential utility as a biomarker for genetic counseling, patient stratification, and the development of precision therapeutic strategies. preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 3

The SCN2A gene encodes the voltage-gated Na+ channel α subunit NaV1.2, which is the main Na+ channel responsible for the initiation and propagation of action potentials of cortical excitatory neurons in the early postnatal period , while at a later age it is also involved in their dendritic excitability and probably in setting features of synaptic transmission1-3. Heterozygous p athogenic variants of SCN2A/NaV1.2 can cause a wide phenotypic spectrum, including mild epilepsy, different types of developmental and epileptic encephalopathies (DEEs) or neurodevelopmental disorders without epilepsy4,5. Some SCN2A/NaV1.2 variants cause phenotypes with late infantile or childhood onset, including infantile-childhood DEEs (ICDEEs) and different neurodevelopmental disorders6,7. ICDEE patients with onset between 3 months and 1 year of age often exhibit an infantile spasms syndrome phenotype, whereas ICDEE patients with onset after 1 year of age may have variable epilepsy phenotypes that cannot be classified within a specific epilepsy syndrome, developmental delay/intellectual disability and autistic traits. SCN2A/NaV1.2 neurodevelopmental disorders include autism spectrum disorder (ASD) and/or intellectual disability, as well as other neuropsychiatric conditions such as schizophrenia2,6,7. Although seizures have been observed in some of these patients after signs of neurodevelopmental dysfunctions became apparent , epilepsy is not a major feature of their phenotype. Large-scale human genetic studies have indicated that SCN2A/NaV1.2 variants are among the leading genetic causes of non-syndromic ASD (nsASD)8,9. Other SCN2A/NaV1.2 variants cause phenotypes with onset in the first 3 months of life, mostly in the neonatal period 6,7. Some of these patients exhibit self-limited neonatal/infantile epilepsy10, others exhibit severe neonatal-early infantile DEE (NEI DEE) phenotypes with drug-resistant seizures and intellectual disability. To disclose pathological mechanisms , stratify patients and identify specific therapies in a precision medicine framework, it is important to identify the functional effect s of the variants and correlate them to the phenotype. Functional studies in transfected cell lines ha ve provided some genotype- phenotype correlations, showing that gain-of-function (GoF) variants cause self-limited epilepsy or NEIDEE, which can respon d to treatment with Na + channel blockers, whereas loss -of-function (LoF) variants cause ICDEE or neurodevelopmental disorders without epilepsy , phenotypes that are generally worsen ed by Na+ channel blockers 1,2,7. However, it has been thus far more difficult to identify clear genotype-phenotype relationships within GoF or LoF variants1,5,11,12. preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 4 It has been proposed that SCN2A/NaV1.2 ASD variants , compared to ICDEE ones, cause larger LoF that lead to haploinsufficiency13. Haploinsufficient Scn2a+/- mice, especially at young age, show autistic-like features3,14-17, which are milder than those observed in mice with a larger reduction of Scn2a expression18-21. Thus, pathological mechanisms that can cause LoF larger than haploinsufficiency could be involved in SCN2A/NaV1.2 ASD. Although all these variants have been identified in heterozygous patients, functional studies have been performed thus far with conditions that mimic homozygosis. However, it has been proposed that α subunits, including NaV1.2, can interact and form dimers22. This interaction may generate dominant negative effects when wild-type and mutant proteins are co-expressed, as in conditions of heterozygosis, in which the mutant subunit may reduce the function of the wild-type one23. Here we performed functional studies in transfected cell lines and neurons of SCN2A/NaV1.2 variants that cause phenotypes with late infantile or childhood onset, in which we reproduced the conditions of heterozygosis . We investigated published variants and novel variants identified in a newly described cohort of patients. We identified dominant negative effects as a novel specific pathological mechanism for variants involved in nsASD. preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 5

Plasmid and Mutagenesis We used the cDNA of the human Na v1.2 channel α subunit ( hNaV1.2, GenBank accession no. NM_021007) provided by Dr. Jeff Clare (Glaxo-SmithKline, Stevenage, Herts, United Kingdom), which we subcloned into the pCDM8 vector to minimize rearrangements 24. The cDNA of the human Kir4.1 channel (GenBank accession no. NM_002241) was obtained from OriGene technologies Inc., USA (CAT#: SC118741). We introduced the mutations with the Quick-Change Lightning Kit (Stratagene) as already described 25,26; primers’ sequences are available on request. To isolate the Na+ currents generated by the transfected channels from the endogenous ones in the experiments in which we used neocortical neurons , we expressed hNav1.2 channels (WT or carrying a pathogenic mutation) resistant to the specific blocker tetrod otoxin (TTX), in which the phenylalanine at position 387 was replaced with a serine 24-26. The plasmid conta ining difopein (dimeric fourteen -three-three peptide inhibitor)27 fused to YFP (pEYFP-C1-difopein) was provided by Dr. Isabelle Deschênes (Case Western Reserve University, Cleveland, USA)22. Cell culture and transfection We used the cell line tsA -201, maintained and transiently transfected with CaPO 4 as already reported25,26,28. Neocortical neurons were prepared from E17 mouse embryos (Charles River) and maintained in primary culture as already described 25,29. Transfections of neurons were performed with Lipofectamin 2000 (Invitrogen) 5 days after the preparation and recorded 24-48h after the transfection. We co-transfected a plasmid expressing Yellow Fluo rescent Protein (pEYFP -N1; Clontech) to identify the transfected cells (tsA -cells 201 or neocortical neurons) for electrophysiological recordings. For recordings from cultured neurons, we selected cells with pyramidal morphology30. Electrophysiological recordings and analysis We used the whole -cell configuration of the patch -clamp technique to record Na+ currents, as previously described25. Recordings were realized at room temperature (20-24°C) with a Multiclamp 700A amplifier and pClamp 10.2 software (Axon Instruments/Molecular Devices). Signal s were filtered at 10 kHz and sampled at 50 kHz. Electrode capacitance and series resistance were compensated during the experiments. Pipette resistance was between 2 -2.5 MOhms and voltage error maintained under 2.5 mV. The P/4 subtraction paradigm was used to cancel the remaining preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 6 transient and leakage currents. Recording solutions for tsA-201 cells were (in mM): external solution 150 NaCl, 1 MgCl 2, 1.5 CaCl 2, KCl2 and 10 HEPES (pH 7.4 with NaOH); internal pipette solution 105 CsF, 35 NaCl, 10 EGTA, 10 HEPES and (pH 7.4 with CsOH). Recording solutions for neurons were (in mM): external solution 140 NaCl, 2 MgCl2, 2 CaCl2, 1 BaCl2, 1 CdCl2, 10 HEPES (pH 7.4 with NaOH) and TTX 1 µM; internal pipette solution (in mM): 130 CsF, 10 NaCl, 10 EGTA and 10 HEPES (pH 7.4 with CsOH). The recordings were started 5 minutes after reaching the whole-cell configuration to allow a complete dialysis of the cytoplasm . Voltage dependence of activation was studied applying test pulses of 100-ms from −110 mV to +60 mV from a holding potential at −120 mV. Voltage dependence of inactivation was studied with a 100-ms prepulse at different potentials followed by a test pulse at −10 mV. Conductance-voltage curves were derived from current-voltage (I–V) curves according to G = I/(V−Vr), where I is the peak current, V is the test voltage, and Vr is the apparent observed reversal potential. The voltage dependence of activation and the voltage dependence of inactivation were fit to Boltzmann relationships in the form y = 1/(1 + exp((V1 2 −V)/k)), where y is normalized GNa or INa, V1/2 is the voltage of half -maximal activation (Va) or inactivation (Vh) and k is a slope factor . Action potential clamp recordings were performed as already described26,31-33. The inter-sweep interval was 8s for all the protocols. Data were analyzed with pClamp v10.2 (Axon Instruments/Molecular Devices) and Origin2021 (OriginLab). Junction potential was not corrected. Binding assay The toxin AaHII, isolated from the venom of the scorpion Androctonus australis Hector was a generous gift of Dr. MF Eauclaire and D r. P. Bougis. It was radioiodinated an d used for binding experiments as described in34,35. We selected this site‑3 toxin because it yields clean, highly sensitive binding, although some mutants could not be assessed, as they lack its binding site 36. Binding experiments were performed on intact transfected cells in binding buffer solution (Choline Chloride 130 mM, HEPES 50 mM, Glucose 5.5 mM, MgSO40.8 mM, KCl 5.4 mM, BSA 1 mg/ml, pH = 7.4 ) with the addition of 10 mg/ml gramicidin A for generating a more negative membrane potential. Cells were incubated in the presence of 125I-AaHII at 37 °C for 45 min and unbound toxin was removed by aspirating each well and rinsing 2 times using the washing buffer (Choline Chloride 163 mM, HEPES 5 mM, CaCl2 1.8 mM, MgSO4 0.8 mM, pH = 7.4). Adherent cells with bound toxin were dissolved in 1 ml of 0.4 N NaOH per well and total cell protein was determined using bovine serum albumin as a standard in a modified Lowry protein assay (Bio-Rad Protein Assay). preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 7 Clinical study Clinical assessment comprised standardized neurological, neuropsychological, and neuropsychiatric evaluations. Patients were followed longitudinally for a median duration of 10 years (range, 4 –14 years). Available brain MRI scans and EEG recordings obtained during wakefulness and sleep were systematically reviewed, with particular attention to background activity, physiological patterns and epileptiform discharges. Epilepsy syndromes were classified according to the International League Against Epilepsy (ILAE) criteria. Treatment efficacy was categorized based on longitudinal clinical records and the treating physician’s assessment. Neuropsychological and neuropsychiatric fe atures were evaluated using validated instruments when feasible, including the Autism Diagnostic Observation Schedule –Second Edition (ADOS -2), Autism Diagnostic Interview –Revised (ADI -R), Vineland Adaptive Behavior Scales, and Leiter International Performa nce Scale–Revised (Leiter-R). When formal testing was not possible, assessments were based on structured clinical observation focusing on communication abilities, adaptive functioning, and behavioral features. All clinical data were entered into a pseudonymized database and independently verified for accuracy. Phenotypic classification was determined based on neurological history, EEG findings, and developmental trajectory by investigators blinded to the functio nal data. Clinical and functional data were subsequently analyzed jointly. For each individual, genotype–phenotype correlations were explored by comparing in vitro functional data with the clinical presentation, with the aim of identifying potential associations between the functional mechanism of the variant (dominant -negative versus loss-of-function) and the observed phenotype. The study was approved by the Paediatric Ethics Committee of Meyer Children’s Hospital IRCCS (DECODEE project, Florence, Italy). Written informed consent for genetic testing and for the use of anonymized clinical data for research purposes was obtained from all participants’ parents or legal guardians. Statistical analysis. Datasets were tested for normal distribution with the Kolmogorov–Smirnov test; and homogeneity of variance with the Brown-Forsythe test. Groups with normal or Lognormal distribution were compared with Student’s t-test or one-way ANOVA followed by Dunnet’s T3 post hoc test . Groups not normally distributed or with n too small to test for normality were compared with the Kruskal- Wallis test followed by Mann-Whitney post -hoc test with Bonferroni correction for multiple comparisons. p < 0.05 was considered significant. Significance is indicated in the figures as: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. If not differently indicated, data are shown as means ± preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 8 SEM, ‘‘n’’ indicates the number of cells. Statistical tests were performed with Prism 10.6 (GraphPad) or Origin 2025 (OriginLab). preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 9

Clinical features of patients carrying SCN2A/Nav1.2 variants investigated in this study. We performed the functional study of 15 SCN2A/NaV1.2 variants identified in patients with different phenotypes ( Fig.1). Numerous variants are part of a cohort of patients of the Meyer Children’s Hospital IRCCS (Florence, Italy ), which includes variants that were not reported before (Supplementary Table 1). Some other variants that we studied were selected from the literature: R379H, R937H, C959X and G1013X had been identified in patients with nsASD9,37; R850P and V1289F had been identified in patients with schizophrenia 38; T1420M had been reported in a patient classified as nsASD, with no further detail available in the original publication39. However, based on the results of our functional analysis presented here, we can hypothesise a close similarity of the phenotype of the T1420M patient with that observed for the patient of the Meyer’s cohort carrying the R1635Q variant . We selected the medical records of patients carrying the L1314P and R1635Q variants for a detailed report. Additional information about these two patients and others of the Meyer’s cohort are reported in Supplementary Table 1. The patient carrying the L1314P variant was a 15-years-old girl, brought to neurological attention at age one year due to global developmental delay and microcephaly. Brain MRI performed at 18 months was unrevealing , except for the small brain size . Neuropsychological assessment at age 5 years revealed severe intellectual disability (ID) and absent speech. Since age 3 years the patient had manifested atypical attitudes in social interaction and communication skills, evaluated with ADOS -2 test (module 1), th at revealed severe autism spectrum disorder. No clinical seizures were observed or reported. The only EEG abnormality observed was an increased amplitude of vertex sharp waves during the wakefulness -to-sleep transition (Supplementary Fig.1). At age of 14 years a non -verbal scale (Leiter-R) was consistent with moderate ID. The patient carrying the R1635Q variant , a boy aged 12-years at the time of the study, was initially brought to medical attention at 3 years of age due to global developmental delay with absent speech and impaired communication skills. Brain MRI was normal. Subsequent evaluations highlighted autistic features with persistently absent speech, poor eye contact and social interaction , motor stereotypies, hypersensitivity to loud noises, no interest in interaction with peers. Although a training program with the applied behavior analysis (ABA) method was soon started with some progress, at age 6 years the ADOS-2 test (module 1) scores were consistent with severe autism spectrum disorder. Formal cognitive testing could not be administered. The child remained nonverbal and could only preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 10 express his requests through vocalizations and gestures. EEG recordings showed continuous bilateral centro-temporal spikes (Supplementary Fig. 2), which were greatly activated during sleep; no clinical seizures were observed. The R1635Q variant was previously identified in patients with epilepsy and neurodevelopmental disorders, in which epilepsy is a defining feature 40 (a phenotype aligning with the DEE spectru m), and in a patient described as autistic without observed clinical seizures 41, suggesting phenotypic similarities with the patient carrying this variant in our cohort. Other p atients in the cohort of the Meyer Children’s Hospital IRCCS carried the D284G, A896V, G659D, C1344Y and R1882X variants, and had ICDEE phenotypes , in which different seizure types, with variable outcomes, were associate to mild to severe intellectual disability (ID) (Supplementary Table 1). Some of these patients exhibited syndromic autistic phenotypes within a DEE framework. ASD-hNav1.2 mutants expressed in isolation in tsA-201 cells show complete/nearly complete LoF. To investigate genotype/phenotype relationships of SCN2A mutants, we introduced variants into the cDNA clone of the human Nav1.2 Na+ channel (hNav1.2) by site directed-mutagenesis. Using patch- clamp whole -cell recordings on transiently transfected tsA -201 cells, we analyzed the functional properties of mutants comparing with those of WT hNa v1.2. We first describe the variants that, considering available clinical data and our functional findings, we consider as occurring in n sASD patients: R379H, R937H, C959X, G1013X, L1314P, R1515X (Fig.1). Representative traces of Na + current elicited with a series of depolarizing steps in tsA-201 cells transfected with WT or ASD mutant channels expressed alone are shown in Fig .2a. Fig.2b displays the quantification of the maximal current density (to normalize for the size of the cells) and indicates that five of the six ASD mutants tested display no current , whereas one, R379H, shows strongly reduced (80%) current density. These data indicate nearly complete LoF for the 6 ASD mutants tested. To better disclose the overall impact of the variant R379H on the function of hNa v1.2 in a dynamic condition, we applied as voltage command a neuronal discharge, as we did in previous studies25,31,32, and recorded the corresponding Na + action currents quantified as current densities (Fig.2c). The comparison of the peak current density of the first action current and the mean peak current densities of the last 3 action currents (Fig.2d) indicates that the R379H induces 76% and 9 4% reduction, respectively, which results in a major LoF of hNav1.2. Therefore, our data confirm previous

showing that nsASD SCN2A variants cause a severe LoF13. preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 11 ASD hNav1.2 mutants are not rescued in neocortical neurons and show dominant negative effects. To determine whether a neuronal cell background rescues plasma membrane targeting of hNav1.2 mutants, as we have previously shown for other Na+ channel mutants 25,29, we expressed ASD - associated mutants in primary neocortical neurons. Using TTX -resistant (F385S) constructs in the presence of 1µM TTX to isolate exogenous currents25,29, we found that five mutants failed to generate measurable current, while R379H reduced current density by 69% (Fig.3a). These results indicate that the LoF observed in tsA-201 cells persists in a neuronal context, showing that the cellular background does not influence the functional deficit. Then, to mimic the heterozygous condition observed in patients, we co -expressed WT and mutant hNaV1.2 channels in neocortical neurons (1:1 molar ratio, with 50% of the standard WT cDNA) (Figure 3b). As a control, to evaluate the effect of co-expression of a membrane protein that does not directly interact with hNav1.2 but can compete with it for protein synthesis and trafficking, we co-transfected neurons with WT hNav1.2 and hKir4.1, a non-interacting K⁺ channel, which did not affect WT current density (Fig.3b). Strikingly, co -expression with any of the ASD mutants significantly reduced WT current density: R379H by 38%, L1314P by 52%, R937H and R1515X by 53%, and C959X and G1013X by 58% (Fig.3b). Notably, even R379H, which retained partial function when expressed alone (Fig.3A), showed a dominant-negative effect in the presence of WT channels. These data demonstrate that all the ASD -associated hNa V1.2 mutants we tested impair WT channel function, consistent with a dominant-negative mechanism. The negative dominant effect of the nsASD NaV1.2 mutants depends on regions that have been implicated in the interaction between Na+ channel α-subunits. To investigate the molecular basis of the dominant-negative effects produced by nsASD-associated SCN2A variants, we tested whether these could arise from physical interactions between Na V α-subunits. Prior work, largely on the cardiac channel Na V1.5, suggested that NaV α-subunits form functional dimers through binding of 14-3-3 to two sites within the DI –DII intracellular linker, and a second region in the same linker that mediates direct α -subunit interactions (residues 493–517in NaV1.5) independently of 14-3-322 (Fig.4a). Inhibiting 14-3-3 action, either using the peptide blocker difopein or introducing the S460A mutation in NaV1.5 (which disrupts 14 -3-3 binding ), prevented dimer formation22. We tested whether the dominant -negative effects of nsASD-associated SCN2A variants require regions implicated in Na V α-subunit dimerization by: (i) inhibiting 14 -3-3 through difopein preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 12 co-expression, (ii) introducing the S487A mutation in hNa V1.2 (analogous to S460A in Na V1.5), and (iii) generating the Δ523 –554 deletion in hNav1.2, corresponding to the direct α –α interaction site identified in NaV1.5. Control experiments evaluating the features of WT-hNav1.2 co-expressed with difopein, of hNav1.2- S487A, or of hNav1.2-Δ523–554 showed no changes in current density or in the voltage dependence of activation or fast inactivation compared to WT -hNav1.2 (Fig .4b–c), indicating th at these manipulations did not alter main channel properties. We then assessed their impact on the dominant -negative effects of nsASD mutants (Fig .3d–f). Inhibition of 14 -3-3 with difopein abolished the reduction in current density normally produced by ASD mutants, and the S487A mutation produced the same rescue (Fig .4e). Thus, preventing 14 -3-3 binding, which has been implicated in α-subunit dimerization, eliminates dominant-negative effects. Likewise, ASD mutants failed to reduce current density when co -expressed with Δ523 –554, which lacks the site proposed to be implicated in direct α–α interaction (Fig.4f). Together, these results support a model in which dimerization of NaV α-subunits are required for nsASD-associated dominant-negative effects. The LoF and the negative dominant effects of nsASD mutants are caused by decreased plasma membrane targeting It has been proposed that L oF of Nav1.2 nsASD mutants is caused by a functional impairment (non- conductive mutants) of channels that are efficiently targeted to the plasma membrane 13. However, our data are consistent with a trafficking defect of WT-nsASD mutant  subunit dimers , with the reduction of plasma membrane targeting of the dimer due to increased retention in the endoplasmic reticulum and subsequent degradation, as also proposed for NaV1.5 mutants23. To investigate if dominant negative effects observed evaluating the current density correlate with defects in cell surface expression, we performed binding experiments on intact transfected tsA201- cells expressing WT or ASD mutant channels using the radiolabeled -scorpion toxin AaHII (125I- AaHII), which binds specifically with nanomolar affinity to the receptor site 3 of Na v  subunits (the extracellular loop between S3 -S4 transmembrane segments o f domain IV)36. These binding experiments allow a sensitive and specific quantification of Na+ channel surface expression that is not contaminated by intracellular signal s34,35. Our b inding d ata show a tight correlation between the reduction of cell surface expression and the reduction of cur rent density (Fig.5a). In fact, for the mutant R379H, both the current density and specific 125I-AaHII binding were reduced by 70 -80%, preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 13 whereas f or R937H and L1314P, which do not generate current, no binding was observed. The truncating variants C959X, G1013X and R1515X do not retain the AaHII binding site and, as expected, show no signal in binding experiments (not shown). Thus, the reduction or absence of current that we observed for the ASD mutants in patch-clamp experiments are caused by a reduction or absence of plasma membrane targeting. Next, we performed binding experiments on intact tsA -201 cells co-transfected with WT and nsASD mutants to reproduce heterozygous conditions. In this assay, binding reflects only hNav1.2-WT channels when co-transfected mutants lack the AaHII binding site . Our data show that the channels targeted at the cell surface was reduced by >50% when the nsASD mutants are co-expressed (Fig.5b). Therefore, our binding data correlate with the results of patch-clamp experiments and demonstrate that the targeting to the membrane of hNav1.2-WT is drastically reduced when nsASD mutants are co-expressed. Overall, our patch -clamp and binding data show that the ASD mutants induce dominant negative effects by reducing the targeting to the cell surface of WT -channels, which is mediated by the interaction between WT and nsASD mutant  subunits. hNav1.2 mutants causing other phenotypes with infantile-childhood onset do not show dominant negative effects We compared the mechanism of action of ASD mutants with that of other LoF hNav1.2 mutants. We studied hNav1.2 mutations causing other phenotypes with clinical onset in the infantile-childhood period or later , such as ICDEE (D284G, G659D, A896V, C1344Y and R1882X), schizophrenia (R850P and V1289F), and ASD with severe epileptiform EEG activity without clinical seizures (T1420M and R1635Q; although for T1420M we inferred it considering the functional study ). Fig .6a shows representative Na + current traces elicited with a series of depolarizing steps in tsA -201 cells transfected with WT or mutant channels. The analysis of the maximal current density is shown in Fig.6b and indicates that all the mutants studied (except D284G) induced a significant reduction of current, although none of them show complete LoF. To find out if the cellular background could influence the expression at the cell surface , we transfected the m in neocortical neurons and performed patch -clamp experiments to measure the maximal current density , as we did for ASD mutants (Fig.6c), observing that R1635Q was rescued by the neuronal cellular background. Therefore, preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 14 differently from ASD mutants, all the se mutants can be targeted to the plasma membrane and can generate currents; some of them are rescued by the neuronal cellular background. To better characterize the functional impact of R1635Q, which is rescued in neurons, as well as of D284G, which does not show a reduction of current density in either tsA-201 cells or in neurons, we studied their voltage dependence of activation and inactivation in transfected neocortical neurons (Fig.6d). The R1635Q variant induced a large 10.8 -mV negative shift of the voltage dependence of inactivation, indicating a clear LoF. The voltage dependence of activation was less steep for D284G possibly indicating a mild LoF, although the half maximal potential was not modified. We also studied the effect induced by the co-expression in neurons of hNav1.2-WT with the mutants involved in DEE, schizophrenia and ASD with epileptiform EEG activities without clinical seizures . None of the mutants modified the current density of the co-expressed WT channel (Fig.6e), not even those exhibiting a dramatic reduction in current density when expressed alone (R850P, T1420M, G659D, C1344Y and R1992X). Although the analysis of current density and voltage dependenc ies was consistent with LoF for all mutants, these parameters represent only a subset of functional properties . A comprehensive assessment of gating properties is labor-intensive, and integrating multiple functional alterations into a single, interpretable functional outcome is often not straightforward. To disclose the overall impact of these variants, we therefo re recorded in transfected neurons action Na + currents elicited in response to a neuronal discharge , as performed in Fig.2c. However, because the action potential - evoked Na+ currents were already small in WT hNav1.2 under our recording conditions, potential LoF effects of some of the variants could not be reliably resolved. To overcome this limitation, we enhanced the current by adding the sea anemone toxin ATX ‑II (10 nM) to the extracellular solution. ATX‑II slows fast inactivation of voltage ‑gated Na+ channels, thereby increasing the current during repetitive action potential firing , while exerting only small effects on the kinetics of activation. Representative traces of hN av1.2-WT with or without 10 nM ATX II are shown in Fig .7a, and action currents obtained in response to an action potential discharge, displayed as mean current densities, are shown in Fig.7b. Using these conditions, we studied the overall functional effect of the variants implicated in schizophrenia, ASD with epileptiform EEG activity and DEE, as well as of the only nsASD mutant that showed residual current when expressed alone (R379H). The analysis of the peak current density of the first action current is shown in Fig .7c and indicates that its amplitude is significantly reduced for all the mutants tested , except for R1635Q and D284G. However, comparing the mean peak amplitude of the last three action currents in the discharge, we found that all mutants induced preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 15 a significant decrease in comparison to WT (Fig.7d). These data point to LoF as the general effect of all the mutants studied, which is consistent with neuronal hypoexcitability. Only nsASD variants exhibited negative dominant effects, a pattern that was statistically significant (Fisher’s exact test, p = 0.0002). This indicates that the mechanism of nsASD variants observed in heterologous systems is distinct from that of other LoF variants. preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 16

All the SCN2A variants tested induce a loss of function of hNa v1.2 channels in both cell lines and cultured neurons. All SCN2A variants characterized in our study, whether identified from the literature or newly reported, induced NaV1.2 LoF. This shared functional effect was independent of the associated clinical phenotype, which was within the late infantile or childhood onset spectrum for all. The LoF resulted from reduced peak Na+ currents or altered biophysical properties, as observed for D284G, A896V, and R1635Q. Variants associated with nsASD exhibited complete LoF, with the exception of R379H. The R379H variant recapitulates the effect of the homologous R367H variant in NaV1.5, which causes Brugada syndrome and shows a large reduction in peak current density42. Our functional data show prominent LoF for nsASD variants, consistent with other studies 13. Two hypotheses have been proposed to explain the absence of measurable current in these variants: (1) the mutant channels reach the plasma membrane but fail to conduct, or (2) trafficking to the plasma membrane is impaired. Immunohistochemistry and TIRF microscopy have been used to propose that ASD mutants localize to the plasma membrane but are non -conducting13. However, these measurements can be contaminated by signal from the endoplasmic reticulum. To further address this question, we used a different approach: binding assays with a radio-labelled scorpion toxin on intact transfected cells, which is a highly quantitative, sensitive, and specific measure of the density of functional channels at the plasma membrane, offering a more direct readout of specific surface expression than other methods. Our results showed no detectable binding for the nsASD mutants tested, but for R379H that showed reduced binding, consistent with the reduction in current density observed in patch -clamp recordings . These findings suggest that nsASD mutants fail to traffic properly to the plasma membrane, rather than inducing impaired conductance. Heterozygous conditions highlight different mechanisms of action between nsASD SCN2A variants and SCN2A variants underlying other phenotypes To better model the heterozygous pathophysiological state, we co-expressed WT and mutant NaV1.2 channels. The nsASD-associated mutants consistently diminished WT channel function by approximately 50%, as evidenced by reductions in both Na+ current density, measured via patch - clamp recordings, and surface expression, assessed through binding assays. These findings indicate a dominant-negative mechanism. By contrast, variants linked to other phenotypes did not show such dominant-negative eff ects. This finding aligns with the more pronounced autistic phenotype preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 17 reported in mouse models exhibiting greater reductions in Scn2a expression18-21, relative to haploinsufficient Scn2a+/- mice3,14-17. The R1635Q variant identified in our study was initially classified as nsASD, but it does not exert a dominant-negative effect. However, u pon completion of the clinical evaluation, this variant was found to be associated with a more complex phenotype, including global cognitive regression, absent speech, and continuous centrotemporal spikes during sleep , without clinical seizures , indicating a presentation inconsistent with nsASD. Similarly, the T1420M variant, previously reported in association with nsASD39, shows functional properties comparable to R1635Q. Given the lack of detailed phenotypic data for the T1420M carrier, we propose reclassifying this variant alongside R1635Q. The patient carrying the dominant -negative L1314P variant, who underwent the same comprehensive clinical assessment, did not exhibit significant EEG abnormalities. Collectively, these findings suggest that a dominant -negative effect correlates with pure nsASD phenotypes, while its absence is associated with different clinical presentations. Domains implicated in the d imerization of hN av1.2 channels are necessary for the dominant negative effect of ASD mutations Previous studies have proposed that dominant -negative effects of the cardiac Na + channel Na V1.5 can arise from interactions between wild -type (WT) and mutant Na+ channel α-subunits22,23,43. The proposed interaction site lies within the IS6–IIS1 linker, which contains both a 14-3-3 protein binding motif and a region between amino acids 523–554 mediating a direct α-subunit interaction. We sought to directly test the dimerization hypothesis as a mechanism underlying dominant-negative effects of Na V1.2 mutants . We employed several complementary strategies: inhibition of 14-3-3 binding using difopein, mutation of serine 487 to alanine (S487A) in hNa V1.2 to block 14-3-3 association, and deletion of the region proposed to mediate direct α subunit interactions (Δ523–554). Each of these conditions abolished the dominant -negative effect in our experiments, providing strong evidence that this phenomenon depends on the previously proposed α -subunit interaction site. Moreover, co-transfection of WT channels with nsASD-associated variants revealed that these mutants reduce trafficking of the WT -mutant complex to the plasma membrane, as quantified by radioligand binding assays. Together, these findings indicate that nsASD-associated mutations exert their effects through a dominant-negative mechanism that requires the proteins and domains that have been implicated in interaction between α -subunits. Our binding studies further reveal that, unlike correctly folded channels that are trafficked to the plasma membrane, the reduced preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 18 targeting of WT -nsASD or nsASD-nsASD complexes to the plasma membrane results in the intracellular retention of the dimer, which is probably subsequently degraded (Fig.8). Despite reduced current density , d ominant-negative effect s were not observed for variants associated with DEE, schizophrenia and ASD with continuous epileptiform EEG discharges during sleep. Overall, our data reveal a specific mode of action for ASD -associated SCN2A variants and highlight potential avenues for targeted therapeutic interventions. Our findings for SCN2A variants are part of a broader pattern observed in voltage-gated ion channels, in which certain truncating or missense mutations can exert dominant-negative effects on wild-type subunits, involving aberrant interactions between truncated or misfolded subunits and wild -type channels that lead to retention in the endoplasmic reticulum and subsequent degradation. This mechanism has been proposed for the hom ologous cardiac Na V1.5 channel, where Brugada syndrome-associated truncations reduce wild -type channel function both in vitro and in vivo 44. Similarly, truncating mutations in Cav2.1 (P/Q-type) channels associated with episodic ataxia type 2 (EA2) can suppress wild-type channel currents through similar mechanisms of ER retention/impaired trafficking, and in vivo studies in EA2 knock -in mice support the pat hological relevance of these interactions, consistent with a dominant negative effect45,46. Some of the SCN2A mutants showing dominant negative effects are truncated proteins. Although premature termination codons located far upstream of exon–exon junctions are generally expected to trigger nonsense -mediated mRNA decay (NMD), incomplete mRNA degradation can allow synthesis of truncated mutants47,48, which may exert dominant -negative effects by interacting with wild-type proteins in the endoplasmic reticulum. Such effects have been observed both in vitro and in vivo for NaV1.5 and Ca V2.1 truncations, where misfolding, ER retention, and proteasomal degradation can reduce wild-type channel function 23,43,44,49. These findings suggest that even small amounts of truncated channel subunits can interfere with trafficking or function of wild -type proteins, providing a mechanism for dominant -negative pathogenicity across Na+ and Ca2+ channelopathies. Notably, while dominant-negative effects are robustly observed in heterologous expression systems for multiple channel families, in vivo evidence is more nuanced . These observations highlight that dominant-negative effects can be context-dependent and may be modulated by variant position, NMD efficiency, ER quality control features, and expression levels47,48. Overall, the mechanism we identified for nsASD variants is consistent with the broader paradigm of dominant-negative effects observed across Na+ and Ca2+ channelopathies, reinforcing the preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 19 importance of evaluating each variant in pathophysiological relevant systems to understand its pathogenic potential. Clinical relevance of the functional study From a clinical perspective, functional data have potential direct implications for stratification and management of patients . Determining whether a variant acts through haploinsufficiency or a dominant-negative mechanism can be important not only for mechanistic interpretation but also for precision medicine. Our findings support an association between molecular mechanisms and clinical phenotypes. We observed that patients carrying dominant-negative SCN2A variants present with ASD but no seizures or severe epileptiform activity, whereas variants causing haploinsufficiency or non-dominant LoF are more frequently associated with epilep sy, often configuring a DEE, or other phenotypes, although there was no cl ear correlation between amount of LoF and phenotype for these variants . Our data underscore how functional stratification of SCN2A variants could refine clinical classification and prognosis, helping to distinguish between pure neurodevelopmental disorders and epilep sy syndromes within the same genetic continuum. A recent study demonstrated that upregulating the remaining functional SCN2A allele using CRISPR activation (CRISPRa) can rescue neurological phenotypes in Scn2a haploinsufficient mice, highlighting a potential therapeutic strategy for neurodevelopmental disorders associated with SCN2A haploinsufficiency50. However, in the presence of dominant -negative variants, the concomitant increased expression of muta nt channels could counteract the benefits of such approaches. Consequently, future therapeutic strategies that enhance Na+ channel expression should integrate detailed clinical phenotyping with functional data, to guide variant-specific interventions and a void potential adverse effects in dominant-negative conditions. The translational relevance of our findings depends on validating the functional consequences of SCN2A variants in patients. Although our experiments in heterologous cells and rodent cortical neurons in culture reveal clear mechanistic distinctions between haploinsufficient and dominant-negative variants, the extent to which these mechanisms operate in human neurons remains to be fully defined. While patient -derived induced pluripotent stem cell (iPSC) –based neurons and cortical organoids may offer compleme ntary insight, variability in differentiation state, cellular maturity, and line-to-line heterogeneity can limit the reproducibility and interpretability of

obtained from these systems51, which should be considered as informative but not definitive. preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 20 In vivo studies in animal models are warranted for establishing the relevance of dominant -negative effects and determining how SCN2A nsASD variants specifically influence neuronal circuits and behaviour within an intact organism during different stages of development. Although additional studies are needed to better elucidate detailed molecular mechanisms underlying dominant-negative effects, and clarify why such effects are not observed in variants not associated with nsASD, our findings indicate that assessing negative dominant effects in heterologous systems may serve as a useful biomarker for predicting the clinical phenotype of LoF SCN2A variants.

We analyzed SCN2A variants associated with infantile and childhood phenotypes and confirmed that all produce NaV1.2 loss of function. The nsASD‑linked variants caused the most severe deficits, largely due to impaired trafficking, and uniquely showed dominant‑negative effects when co‑expressed with wild‑type channels, consistent with interactions between mutant and WT α ‑subunits. This mechanism distinguishes ASD‑related variants and may offer an in vitro biomarker, which should be validated in larger cohorts. Our findings un derscore the importance of integrating functional data into diagnostic workflows for precision medicine in neurodevelopmental channelopathies. preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 21 Funding This work was supported by funding from the French government, through the France 2030 investment plan managed by the Agence Nationale de la Recherche (ANR), as part of the Université Côte d'Azur's Initiative of Excellence (IdEx) Jedi (ANR -15-IDEX-01), by the Laboratory of Excellence “Ion Channel Sc ience and Therapeutics” (LabEx ICST ANR -11-LABX-0015-01), by the ANR Nav1.2RESCUE (ANR-21-CE18-0042) and by the Fondation Jérôme Lejeune (France) to MM. The study was also supported, in part, by funds from the ‘2024 and 2025 Current Research Annual Funding ’ of the Italian Ministry of Health, by #NEXTGENERATIONEU (NGEU), by the Ministry of University and Research (MUR) and the National Recovery and Resilience Plan (NRRP), project MNESYS (PE0000006)—A Multiscale integrated approach to the study of the nervous system in health and disease (DN. 1553 11.10.2022). preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 22

1. Mantegazza M, Cestele S, Catterall W. Sodium Channelopathies of Skeletal Muscle and Brain. Physiol Rev. Mar 26 2021;doi:10.1152/physrev.00025.2020 2. Sanders SJ, Campbell AJ, Cottrell JR , et al. Progress in Understanding and Treating SCN2A -Mediated Disorders. Trends Neurosci. Jul 2018;41(7):442-456. doi:10.1016/j.tins.2018.03.011 3. Spratt PWE, Ben -Shalom R, Keeshe n CM , et al . The Autism -Associated Gene Scn2a Contributes to Dendritic Excitability and Synaptic Function in the Prefrontal Cortex. Neuron. Aug 21 2019;103(4):673-685 e5. doi:10.1016/j.neuron.2019.05.037 4. Guerrini R, Conti V, Mantegazza M, Balestrini S, Galanopoulou AS, Benfenati F. Developmental and epileptic encephalopathies: from genetic heterogeneity to phenotypic continuum. Physiol Rev . Jan 1 2023;103(1):433-513. doi:10.1152/physrev.00063.2021 5. Rusina E, Simonti M, Duprat F, Cestele S, Mantegazza M . Voltage-gated sodium channels in genetic epilepsy: up and down of excitability. J Neurochem. Aug 31 2023;doi:10.1111/jnc.15947 6. Wolff M, Brunklaus A, Zuberi SM. Phenotypic spectrum and genetics of SCN2A -related disorders, treatment options, and outcome s in epilepsy and beyond. Epilepsia. Dec 2019;60 Suppl 3:S59 -S67. doi:10.1111/epi.14935 7. Wolff M, Johannesen KM, Hedrich UBS , et al . Genetic and phenotypic heterogeneity suggest therapeutic implications in SCN2A -related disorders. Brain. May 1 2017;140(5 ):1316-1336. doi:10.1093/brain/awx054 8. Satterstrom FK, Kosmicki JA, Wang J , et al . Large -Scale Exome Sequencing Study Implicates Both Developmental and Functional Changes in the Neurobiology of Autism. Cell. Feb 6 2020;180(3):568-584 e23. doi:10.1016/j.cell.2019.12.036 9. Sanders SJ, Murtha MT, Gupta AR, et al. De novo mutations revealed by whole-exome sequencing are strongly associated with autism. Nature. Apr 4 2012;485(7397):237-41. doi:10.1038/nature10945 10. Heron SE, Crossland KM, Andermann E , et al. Sodium-channel defects in benign familial neonatal - infantile seizures. Lancet. Sep 14 2002;360(9336):851-2. doi:10.1016/S0140-6736(02)09968-3 11. Tan M, Goad BS, Allen M , et al . Mechanisms of SCN2A loss of function do not predict presence or phenotype of epilepsy. Epilepsia. Feb 5 2026;doi:10.1002/epi.70100 12. DiStefano N, Cooper JN, Elisha DH, et al. Decoding SCN2A Variants: Bridging Genetics and Phenotypes in Autism Spectrum Disorder. J Clin Med. May 28 2025;14(11)doi:10.3390/jcm14113790 13. Ben-Shalom R, Keeshen CM, Berrios KN, An JY, Sanders SJ, Bender KJ. Opposing Effects on NaV1.2 Function Underlie Differences Between SCN2A Variants Observed in Individuals With Autism Spectrum Disorder or Infantile Seizures. Biol Psychiatry. Jan 27 2017;doi:10.1016/j.biopsych.2017.01.009 14. Léna I, Mantegazza M. NaV1.2 haploinsufficiency in Scn2a knock -out mice causes an autistic -like phenotype attenuated with age. Scientific Reports. 2019/09/09 2019;9(1):12886. doi:10.1038/s41598 -019- 49392-7 15. Shin W, Kweon H, Kang R, et al. Scn2a Haploinsufficiency in Mice Suppresses Hippocampal Neuronal Excitability, Excitatory Synaptic Drive, and Long -Term Potentiation, and Spatial Learning and Memory. Front Mol Neurosci. 2019;12:145. doi:10.3389/fnmol.2019.00145 16. Tatsukawa T, Raveau M, Ogiwara I , et al . Scn2a haploinsufficient mice display a spectrum of phenotypes affecting anxiety, sociability, memory flexibility and ampakine CX516 rescues their hyperactivity. journal article. Molecular Autism. March 28 2019;10(1):15. doi:10.1186/s13229-019-0265-5 17. Marcantonio W, Simonti M, Lena I, Mantegazza M. Sex -Specific Behavioral Features of Juvenile and Adult Haploinsufficient Scn2a(+/-) Female Mice, Model of Autism Spectrum Disorder. Genes Brain Behav. Oct 2025;24(5):e70034. doi:10.1111/gbb.70034 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 23 18. Spratt PWE, Alexander RPD, Ben-Shalom R, et al. Paradoxical hyperexcitability from Na(V)1.2 sodium channel loss in neocortical pyramidal cells. Cell Rep . Aug 3 2021;36(5):109483. doi:10.1016/j.celrep.2021.109483 19. Wang HG, Bavley CC, Li A , et al . Scn2a severe hypomorphic mutation decreases excitatory synaptic input and causes autism-associated behaviors. JCI Insight. Aug 9 2021;6(15)doi:10.1172/jci.insight.150698 20. Zhang J, Che n X, Eaton M , et al . Severe deficiency of the voltage -gated sodium channel Na(V)1.2 elevates neuronal excitability in adult mice. Cell Rep . Aug 3 2021;36(5):109495. doi:10.1016/j.celrep.2021.109495 21. Eaton M, Zhang J, Ma Z , et al. Generation and basic characterization of a gene-trap knockout mouse model of Scn2a with a substantial reduction of voltage -gated sodium channel Na(v) 1.2 expression. Genes Brain Behav. Apr 2021;20(4):e12725. doi:10.1111/gbb.12725 22. Clatot J, Hoshi M, Wan X , et al . Voltage-gated sodium channels assemble and gate as dimers. Nat Commun. Dec 12 2017;8(1):2077. doi:10.1038/s41467-017-02262-0 23. Sottas V, Abriel H. Negative -dominance phenomenon with genetic variants of the cardiac sodium channel Nav1.5. Biochim Biophys Acta. Jul 2016;1863(7 Pt B):1791-8. doi:10.1016/j.bbamcr.2016.02.013 24. Bechi G, Scalmani P, Schiavon E, Rusconi R, Franceschetti S, Mantegazza M. Pure haploinsufficiency for Dravet syndrome Na(V)1.1 (SCN1A) sodium channel truncating mutations. Epilepsia. Jan 2012;53( 1):87- 100. doi:10.1111/j.1528-1167.2011.03346.x 25. Dhifallah S, Lancaster E, Merrill S, Leroudier N, Mantegazza M, Cestele S. Gain of Function for the SCN1A/hNav1.1-L1670W Mutation Responsible for Familial Hemiplegic Migraine. Front Mol Neurosci . 2018;11:232. doi:10.3389/fnmol.2018.00232 26. Cestele S, Scalmani P, Rusconi R, Terragni B, Franceschetti S, Mantegazza M. Self -limited hyperexcitability: functional effect of a familial hemiplegic migraine mutation of the Nav1.1 (SCN1A) Na+ channel. JNeurosci. 2008;28(29):7273-7283. 27. Masters SC, Fu H. 14 -3-3 proteins mediate an essential anti -apoptotic signal. J Biol Chem . Nov 30 2001;276(48):45193-200. doi:10.1074/jbc.M105971200 28. Mantegazza M, Cestele S. Beta-scorpion toxin effects suggest electrostatic interactions in domain II of voltage-dependent sodium channels. JPhysiol. 2005;568(Pt 1):13-30. 29. Cestele S, Schiavon E, Rusconi R, Franceschetti S, Mantegazza M. Nonfunctional NaV1.1 familial hemiplegic migraine mutant transformed into gain of function b y partial rescue of folding defects. Proc Natl Acad Sci U S A. Oct 22 2013;110(43):17546-51. doi:1309827110 [pii] 10.1073/pnas.1309827110 30. Scalmani P, Rusconi R, Armatura E , et al. Effects in neocortical neurons of mutations of the Na(v)1.2 Na+ channel causing benign familial neonatal-infantile seizures. JNeurosci. 2006;26(40):10100-10109. 31. Matricardi S, Cestele S, Trivisano M, et al. Gain of function SCN1A disease-causing variants: Expanding the phenotypic spectrum and functional studies guiding the choice of effective antiseizure medication. Epilepsia. May 2023;64(5):1331-1347. doi:10.1111/epi.17509 32. Brunklaus A, Brunger T, Feng T , et al. The gain of function SCN1A disorder spectrum: novel epilepsy phenotypes and therapeutic implications. Brain. Nov 21 2022;145(11):3816-3831. doi:10.1093/brain/awac210 33. Cestele S, Labate A, Rusconi R, et al. Divergent effects of the T1174S SCN1A mutation associated with seizures and hemiplegic migraine. Epilepsia. May 2013;54(5):927-935. doi:10.1111/epi.12123 34. Cestele S, Ben Khalifa RB, Pelhate M, Rochat H, Gordon D. Alpha -scorpion toxins binding on rat brain and insect sodium channels reveal divergent allosteric modulations by brevetoxin and veratridine. J Biol Chem. Jun 23 1995;270(25):15153-61. 35. Bechi G, Rusconi R, Cestele S, Striano P, Franceschetti S, Mantegazza M. Rescuable folding defective Na1.1 (SCN1A) mutants in epilepsy: Properties, occurrence, and novel rescuing strategy with peptides targeted to the endoplasmic reticulum. Neurobiol Dis. Jan 6 2015;75C:100-114. doi:S0969-9961(14)00399-4 [pii] 10.1016/j.nbd.2014.12.028 36. Cestele S, Catterall WA. Molecular mechanisms of neurotoxin action on voltage -gated sodium channels. Biochimie. Sep-Oct 2000;82(9-10):883-92. doi:S0300-9084(00)01174-3 [pii] preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 24 37. De Rubeis S, He X, Goldberg AP , et al . Synaptic, transcriptional and chromatin genes disrupted in autism. Nature. Nov 13 2014;515(7526):209-15. doi:10.1038/nature13772 38. Carroll LS, Woolf R, Ibrahim Y, et al. Mutation screening of SCN2A in schizophrenia and identification of a novel loss -of-function mutation. Psychiatr Genet . Apr 2016;26(2):60 -5. doi:10.1097/YPG.0000000000000110 39. Iossifov I, O'Roak BJ, Sanders SJ , et al . The contribution of de novo coding mu tations to autism spectrum disorder. Nature. Nov 13 2014;515(7526):216-21. doi:10.1038/nature13908 40. Lindy AS, Stosser MB, Butler E , et al . Diagnostic outcomes for genetic testing of 70 genes in 8565 patients with epilepsy and neurodevelopmental disorder s. Epilepsia. May 2018;59(5):1062 -1071. doi:10.1111/epi.14074 41. Guo H, Wang T, Wu H, et al. Inherited and multiple de novo mutations in autism/developmental delay risk genes suggest a multifactorial model. Mol Autism. 2018;9:64. doi:10.1186/s13229-018-0247-z 42. Selga E, Sendfeld F, Martinez -Moreno R, et al . Sodium channel current loss of function in induced pluripotent stem cell -derived cardiomyocytes from a Brugada syndrome patient. J Mol Cell Cardiol . Jan 2018;114:10-19. doi:10.1016/j.yjmcc.2017.10.002 43. Clatot J, Ziyadeh -Isleem A, Maugenre S , et al . Dominant -negative effect of SCN5A N -terminal mutations through the interaction of Na(v)1.5 alpha -subunits. Cardiovasc Res . Oct 1 2012;96(1):53 -63. doi:10.1093/cvr/cvs211 44. Doisne N, Grauso M, Mougenot N , et al . In vivo Dominant -Negative Effect of an SCN5A Brugada Syndrome Variant. Front Physiol. 2021;12:661413. doi:10.3389/fphys.2021.661413 45. Rose SJ, Kriener LH, Heinzer AK , et al . The first knockin mouse model of episodic ataxia type 2. Exp Neurol. Nov 2014;261:553-62. doi:10.1016/j.expneurol.2014.08.001 46. Raike RS, Kordasiewicz HB, Thompson RM, Gomez CM. Dominant -negative suppression of Cav2.1 currents by alpha(1)2.1 truncations requires the conserved interaction domain for beta subunits. Mol Cell Neurosci. Feb 2007;34(2):168-77. doi:10.1016/j.mcn.2006.10.011 47. Lindeboom RGH, Vermeulen M, Lehner B, Supek F. The impact of nonsense-mediated mRNA decay on genetic disease, gene editing and cancer immunotherapy. Nat Genet . Nov 2019;51(11):1645 -1651. doi:10.1038/s41588-019-0517-5 48. Sato H, Singer RH. Cellular variability of nonsense -mediated mRNA decay. Nat Commun . Dec 10 2021;12(1):7203. doi:10.1038/s41467-021-27423-0 49. Mezghrani A, Monteil A, Watschinger K , et al . A destructive interaction mechanism accounts for dominant-negative effects of misfolded mutants of voltage -gated calcium channels. J Neurosci . Apr 23 2008;28(17):4501-11. doi:10.1523/JNEUROSCI.2844-07.2008 50. Tamura S, Nelson AD, Spratt PWE , et al . CRISPR activation for SCN2A -related neurodevelopmental disorders. Nature. Sep 17 2025;doi:10.1038/s41586-025-09522-w 51. Mantegazza M, Broccoli V. SCN1A/NaV 1.1 channelopathies: Mechanisms in expression systems, animal models, and human iPSC models. Epilepsia. Dec 2019;60 Suppl 3:S25-S38. doi:10.1111/epi.14700 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 25 FIGURES Figure 1: Localization on the hNav1.2 protein of the 15 genetic variants that we have characterized. The α subunit consists of four homologous domains (I-IV) containing 6 transmembrane segments (S1- S6) connected by extra- and intra-cellular loops. The voltage sensor is present in the S4 segments of each domain and the pore region is formed by S5 -S6 segments. nsASD: non -syndromic Autis m Spectrum Disorder; DEE: Developmental and Epileptic Encephalopathy. preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 26 WT WT b R379H R379H d R937H WT R379HR937HC959XG1013XL1314PR1515X 0 100 200 300CD (pA/pF) **** **** ******** **** **** **** *** a c WT R379H 0 90 180CD First Peak (pA/pF) WT R379H 0 10 20 Mean last 3 peaks (pA/pF) −50 0 50 Figure 2: ASD-hNav1.2 mutants expressed in tsA-201 cells show complete/nearly complete loss-of- function. (a) Representative whole -cell Na + current families for hNa v1.2 WT and ASD mutants recorded with depolarizing voltage steps from -80 to +60 mV in 5 mV increments form a holding potential of -100 mV. Among the mutants with no current, only R937H is shown; scale bars 1nA, 4ms. (b) Maximal current density calculated for tsA-201 cells transfected with the WT or the ASD mutants. (c) Action Na+ currents recorded using a voltage stimulus an action potential discharge (first panel) for WT or R379H channels (data are expressed as mean of current density, error bars are not shown for clarity); scale bars 20pA/pF, 20ms. (d) Comparison of the current density of the first action current and of the mean of the last three action currents in the discharge . Data shown as mean ±SEM. See Supplementary material (Statistical Tables) for values, n and statistical tests. preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 27 Figure 3: ASD hNav1.2 mutants are not rescued in neocortical neurons and show dominant negative effects. Maximal current density obtained from recordings of neocortical neurons in primary culture transfected with WT hNav1.2 or the ASD mutant channels (a) or co-transfected with the WT and the mutant channels (b), which reduce the current density of the WT (the arrow highlights the negative dominant effects). The data on hNa v1.2-WT channels in control for panel b were obtained from a novel series of transfections performed in parallel with those of the other conditions displayed in the panel. Data shown as mean ±SEM. See Supplementary material (Statistical Tables) for values, n and statistical tests. preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 28 0 40 80 CD (pA/pF) b a d e f WT Subunit direct interaction site 523-554 c S487AS487A Difopein 14.3.3 protein WT WT+R379HWT+R937HWT+C959XWT+G1013XWT+L1314PWT+R1515X 0 25 50 75 100CD (pA/pF) + Difopein WT + Difopein S487A Del 523-554 50:50 S487A S487A+R379HS487A+R937HS487A+C959XS487A+G1013XS487A+L1314PS487A-R1515X 0 30 60 90 120 del del+R379Hdel+R937Hdel+C959Xdel+G1013Xdel+L1314Pdel+R1515X 0 20 40 60 80 -80 -40 0 0.0 0.5 1.0Normalized conductance V (mV) WT WT+Difopein S487A Del 523-554 Figure 4: Domains that have been implicated in the interaction between two α subunits are necessary for the dominant negative effect of ASD NaV1.2 mutants. (a) Schematic illustration of the interaction between two sodium -channel α-subunits, in which the DI–DII intracellular linker engages 14 -3-3 proteins to form an inter-subunit bridge while a defined interaction domain mediates direct contact and promotes subunit association (modified from

22). (b) Maximal current density of hNav1.2-WT expressed in neocortical neurons in culture compared to that obtained in conditions that should inhibit the interaction ( hNav1.2-WT co- expressed with difopein, hNav1.2-S487A and hNav1.2-Del523-554). The data on hNav1.2-WT channels in control were obtained from a novel series of transfections performed in parallel with those of the other conditions displayed in the figure . (c) Mean voltage dependence of activation and fast inactivation; lines are Boltzmann curves generated using the mean parameters of the fits of the single cells. (d) Maximal current density obtained in neocortical neurons co-transfected with hNav1.2-WT and ASD mutant channels in the presence of difopein, compared with hNav1.2-WT alone. (e) Maximal current density of neocortical neurons co-transfected with hNav1.2-S487A and ASD mutant channels. (f) Maximal current density of neocortical neurons co-transfected with hNav1.2-Del523-554 and ASD mutant channels. Data shown as mean ±SEM. See Supplementary material (Statistical Tables) for values, n and statistical tests. preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 29 Figure 5: ASD mutants have decreased/lack of plasma membrane targeting and decrease membrane targeting of hNav1.2-WT when co-expressed. (a) Specific binding of 125I-AaHII-scorpion toxin to intact tsA-201 cells transfected with hNav1.2-WT or the mutants that retain the binding site of the toxin. (b) Specific binding of 125I-AaHII-scorpion toxin to intact tsA -201 cells co -transfected with hNa v1.2-WT and the ASD mutants, compared with cells transfected with hNav1.2-WT alone (50% cDNA in the transfection); in this assay, binding reflects only hNav1.2-WT channels when co-transfected mutants lack the AaHII binding site. The arrow highlights the negative dominant effect. Data shown as mean ±SEM. See Supplementary material (Statistical Tables) for values, n and statistical tests. preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 30 −80 −60 −40 −20 0 20 0.0 0.5 1.0 WT R850PV1289FT1420MR1635QG959DD284GA896VC1344YR1882X 0 200 400 CD (pA/pF) **** ** **** **** ** **** **** **** h1.2-WT h1.2-R850P h1.2-V1289F h1.2-T1420M h1.2-R1635Q h1.2-G659D h1.2-D284G c d e h1.2-A896V h1.2-C1344Y h1.2-R1882X b 50:50 WT R850PV1289FT1420MR1635QG959DD284GA896VC1344YR1882X 0 60 120CD (pA/pF) * * *** ** ****** Rescue WT R1635Q D284G Normalized conductance Voltage (mV) a WT WT+R850PWT+V1289FWT+T1420MWT+R1635QWT+G659DWT+D284GWT+A896VWT+C1344YWT+R1882X 0 40 80 CD (pA/pF) Figure 6: Functional analysis of hNa v1.2 mutants responsible for other phenotypes with infantile- childhood onset. (a) Representative whole-cell Na+ current families for hNav1.2 WT and hNav1.2 mutants responsible for schizophrenia (R850P and V1289F), ASD with epileptiform EEG activity without clinical seizures (T1420M and R1635Q) and DEE (D284G, G659D, A896V, C1344Y and R1882X) expressed in tsA -201 cells. Maximal current density of the WT and the different mutants expressed in tsA -201 cells (b) or expressed in neocortical neurons in culture ( c). ( d) Voltage dependence of the hNa v1.2-WT and mutants. (e) Maximal current density of the hNav1.2-WT co-expressed with the different mutants in neocortical neurons. Data shown as mean±SEM. See Supplementary material (Statistical Tables) for values, n and statistical tests. preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 31 WT R379HR850PV1289FT1420MR1635QG659DD284GA896VC1344YR1882X 0 350 700 WT R379H T1420M C1344Y D284G R1882X V1289FR850P a b d A896V G959DR1635Q WT-CONTROL WT-ATX II WT R379HR850PV1289FT1420MR1635QG659DD284GA896VC1344YR1882X 0 100 200 Mean last 3 peaks (pA/pF) * * * * * * ** *** c First peak (pA/pF) **** * * ** * Figure 7: Overall effects of hNav1.2 variants evaluated with action potential discharge-clamp in the presence of ATX II. (a) Representative normalized hNav1.2 Na+ current traces elicited by a step depolarization at the maximal peak current density in control or in the presence of 10 nM ATX II; scale bar 20ms. The inset shows the same traces in a 7-ms window. (b) Action Na+ currents in the presence of ATX II displayed as mean of current densities for hNav1.2-WT, hNav1.2-R379H (nsASD with residual current), hNav1.2- R850P and hNav1.2-V1289F (schizophrenia), hNa v1.2-T1420M and hNav1.2-R1635Q (ASD with epileptiform EEG discharges), hNav1.2-G659D, hNav1.2-D284G, hNav1.2-A896V, hNav1.2-C1344Y and hNav1.2-R1882X (infantile-childhood DEE) ; sca le bars 50pA/pF, 25ms . Comparison of the current density of the first action current (c) and of the mean of the last three action currents in the discharge (d). Data are shown as mean ± SEM. See Supplementary material (Statistical Tables)for values, n and statistical tests. preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint 32 Figure 8: Proposed m echanism of action of non-syndromic ASD (nsASD) hNav1.2 mutants. We propose that nsASD-linked mutations act through a dominant-negative mechanism that depends on interactions between α-subunits. WT –nsASD and nsASD –nsASD dimers exhibit reduced surface trafficking: they are retained intracellularly, where they are likely targeted for degradation. preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint SUPPLEMENTARY MATERIAL Non-syndromic autism –associated SCN2A variants selectively exert dominant -negative effects on NaV1.2 channels. Sandrine Cestèle 1,2,3, Renzo Guerrini 4,5 Sandra Difhallah 1,2,3, Davide Mei 4, Natalie Leroudier1,2,3, Marialuisa Ricci4, Simona Balestrini4,5 and Massimo Mantegazza1,2,3,*. 1 University Cote d’Azur, Valbonne-Sophia Antipolis, France 2 CNRS UMR 7275, Institute of Molecular and Cellular Pharmacology (IPMC), Valbonne - Sophia Antipolis, France 3 Inserm U1323, Valbonne-Sophia Antipolis, France 4 Neuroscience Department, Meyer Children’s Hospital, Florence, Italy 5 University of Florence, Florence, Italy Correspondence to: Massimo Mantegazza Institute of Molecular and Cellular Pharmacology (IPMC) University Côte d'Azur-CNRS UMR7275-Inserm U1323 660 Route des Lucioles, 06560 Valbonne-Sophia Antipolis, France E-mail [email protected] preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint Cestèle et al 2026 Supplementary material 2 Mutation Phenotype NM HGMD SCN2A variant Analysis

Inherithance Prediction HGMD, Phenotype (Reference) Age at follow- up Age at seizure onset Seizure types at onset Seizures during follow-up EEG interictal- ictal Brain MRI (age) Epilepsy syndrome ASMs / efficacy Epilepsy outcome Autism Spectrum Disorder: yes/no Autism diagnostic tools Cognitive function / Neuropsychology Additional findings Overall effect Negative Dominant effect L1314P nsASD NM_021007.2 c.3941T>C, p.(Leu1314Pro) Single-gene sequencing De novo Deleterious No 15 y - - - Normal Normal (18 m) - - - Yes ADOS-2, Module 1: social affect score=17; restricted repetitive behaviors score=8; total score=25; severity score=10 (severe). ADI-R: reciprocal social interaction score=22; communication score=12; restrictive and repetitive behaviors score=8; development score=3. Moderate ID/Leiter-R scale: fluid reasoning =48; non-verbal IQ=36. VABS-II: communication domain=42; daily living skills domain=38, socialization domain=43; total=39. Cystic fibrosis, microcephaly Loss of function Yes D284G ICDEE NM_021007.2 c.851A>G, p.(Asp284Gly) Gene panel analysis De novo Deleterious Yes, ASD (Retterer et al., 2016, 1) 7 y 3 m Myoclonic and GTC Myoclonic and GTC Multifocal and generalized discharges - Myoclonic and GTC seizures Mild atrophy with anterior predominance (2 y); diffuse atrophy (6 y) DEE LEV/-, CLN/-, TPM/-, KD/-, VPA/+, CLB/+ Drug-resistant Yes NA Severe ID (non- verbal) Quadriparesis, axial hypotonia, dysphagia Loss of function No C1344Y ICDEE NM_021007.2 c.4031G>A, p.(Cys1344Tyr) Gene panel analysis De novo Deleterious Yes DEE, early onset (Parrini et al. 2017, 2) 12 y 18 m Spasms Drug-resistant spasms; developmental regression; seizure-free on polytherapy since age 3 y Slow and monomorphic

- Right parieto- temporal discharges - Epileptic spasms Normal (16 m, 20 m, 44 m) IESS VGB/+, ACTH/+, CBZ/+++ Seizure free Yes NA Severe ID (non- verbal) - Loss of function No R1635Q ICDEE NM_021007.2 c.4904G>A, p.(Arg1635Gln) WES De novo Deleterious Yes ASD (Guo et al., 2018, 3); Epilepsy and neurodevelopment al disorders (Lindy et al. 2018, 4) 12 y NA - - Centro- temporal discharges, almost continuous during sleep Normal (3 y) - - - Yes ADOS-2: social affect total score=17; restricted and repetitive behavior total score=14; social affect and restricted and repetitive behavior total score=31 (severe). Severe ID (non- verbal) - Loss of function No R1882X ASD with continuous interictal EEG discharges NM_021007.2 c.5644C>T, p.(Arg1882X) Gene panel analysis Inherited from father with epilepsy Deleterious Yes DEE (Chen et al. 2022, 5); Microcephaly, Global developmental delay (Canavati et al. 2024, 6). 11 y 3 y 9 m GTC GTC Generalized spike- wave discharges, anterior slow waves Normal (4 y) DEE VPA/+ ++ Seizure free No - Mild ID/Griffiths-III scale: DQ=71 - Loss of function No G659D Absences (childhood) NM_021007.2 c.1976G>A, p.(Gly659Asp) WES Inherited from father with epilepsy Deleterious No 14 y 2 y Eyelid myoclonia with absences Eyelid myoclonia with absences Multifocal and generalized spike- wave discharges NA Eyelid myoclonia with absences VPA/++, ETS/++ Improvement but absences with eyelid myoclonia persist No - Mild ID/WISC-IV: IQ=71 - Loss of function No A896V ICDEE NM_021007.2 c.2687C>T, p.(Ala896Val) Gene panel analysis De novo Deleterious Yes DEE (Parrini et al. 2017, 2) 4 y 3 d Tonic Tonic Multifocal discharges - Tonic seizures Global atrophy, involving brain, brainstem and cerebellum; thin corpus callosum DEE PB/-, VGB/-, HCT/-, PHT/- Drug-resistant No - Severe ID (non- verbal) Severe quadriplegia. Deceased at 6 y due to respiratory infection Loss of function No Supplementary Table 1. Clinical features of the patients from the cohort of the Meyer Children’s Hospital. Abbreviations: ACTH, adrenocorticotropic hormone; ADI-R, autism diagnostic preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint Cestèle et al 2026 Supplementary material 3 interview–revised; ADOS, autism diagnostic observation schedule; ASD, autism spectrum disorder; ASMs, anti-seizure medications; CBZ, carbamazepine; CLB, clobazam; CLN, clonazepam; d, days; DEE, epileptic and developmental encephalopathy; DQ, developmental quotient; EEG, electroencephalogram; ETS, ethosuximide; GAI, general ability index; GTC, generalized tonic -clonic; ICDEE, infantile-childhood epileptic and developmental encephalopathy; HCT, hydrocortisone; HGMD, human gene mutation database; KD, ketogenic diet; ID, intellectual disability; IESS, infantile epileptic spasms syndrome; IQ, intelligence quotient; LEV, levetiracetam; m, months; NA, not available; nsASD, non-syndromic autism spectrum disorder; PB, phenobarbital; PHT, phenytoin; TPM, topiramate; VABS-II, vineland adaptive behavior scales second edition; VGB, vigabatrin; VPA, valproic acid; WES, Whole Exome Sequencing; WISC-IV, Wechsler intelligence scale for children-fourth edition; y, years; +, mildly effective; ++, effective; +++, very effective; -, ineffective.

1. Retterer K, Juusola J, Cho MT , et al . Clinical application of whole -exome sequencing across clinical indications. Genet Med . Jul 2016;18(7):696 -704. doi:10.1038/gim.2015.148 2. Parrini E, Marini C, Mei D , et al. Diagnostic Targeted Resequencing in 349 Patients with Drug -Resistant Pediatric Epilepsies Identifies Causative Mutations in 30 Different Genes. Hum Mutat. Feb 2017;38(2):216-225. doi:10.1002/humu.23149 3. Guo H, Wang T, Wu H, et al. Inherited and multiple de novo mutations in autism/developmental delay risk genes suggest a multifactorial model. Mol Autism. 2018;9:64. doi:10.1186/s13229-018-0247-z 4. Lindy AS, Stosser MB, Butler E, et al. Diagnostic outcomes for genetic testing of 70 genes in 8565 patients with epilepsy and neurodevelopmental disorders. Epilepsia. May 2018;59(5):1062-1071. doi:10.1111/epi.14074 5. Chen S, Xiong J, Chen B, et al. Autism spectrum disorder and comorbid neurodevelopmental disorders (ASD-NDDs): Clinical and genetic profile of a pediatric cohort. Clin Chim Acta. Jan 1 2022;524:179-186. doi:10.1016/j.cca.2021.11.014 6. Canavati C, Sherill -Rofe D, Kamal L , et al . Using multi -scale genomics to associate poorly annotated genes with rare diseases. Genome Med . Jan 4 2024;16(1):4. doi:10.1186/s13073-023-01276-2 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint Cestèle et al 2026 Supplementary material 4 Supplementary Figure 1. Patient carrying the L1314P variant. EEG recording while falling asleep shows sharp wave discharges over the vertex and central regions that are more pronounced than the physiological expected vertex sharp waves accompanying wakefulness-sleep transition. preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint Cestèle et al 2026 Supplementary material 5 Supplementary Figure 2. EEG recording during sleep of the patient carrying the R1635Q variant. There is activation of very frequent bilaterally synch ronous and asynchronous spikes and sharp waves over the central regions and the vertex area. preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint Cestèle et al 2026 Supplementary material 6 Statistical tables: values, n and statistical tests for main figures Panel 2a Current density (pA/pF). Mean ± SEM WT: 213.1 ± 18.4 (n=16) (control) R379H: 33.1 ± 3.8 (n=10) R937H: 0.9 ± 0.1 (n=9) C959X: 0.95 ± 0.08 (n=7) G1013X: 1.0 ± 0.1 (n=8) L1314P: 1.1 ± 0.1 (n=10) R1515X: 1.0 ± 0.1 (n=9) Logarithmic transformation; One-Way ANOVA (F=398, DFb=6, DFw=62) p<0.0001 (****) Dunnett’s post-hoc test P<0.0001 (****) p<0.0001 (****) p<0.0001 (****) p<0.0001 (****) p<0.0001 (****) p<0.0001 (****) Panel 2b CD first peak (pA/pF). Mean ± SEM WT: 87.7 ± 9.2 (n=16) (control) R379H: 20.9 ± 1.4 (n=10) t-test Welch corrected t=7.143, df=15.67 P<0.0001 (****) CD mean last 3 peaks (pA/pF). Mean ± SEM WT: 6.6 ± 1.3 (n=16) (control) R379H: 0.4 ± 0.06 (n=10) t-test Welch corrected t=4.622, df=15.07 P=0.0003 (***) Figure 3 Panel 3a Current density (pA/pF). Mean ± SEM WT: 45.4 ± 8.2 (n=14) (control) R379H: 14.1 ± 2.4 (n=9) R937H: 1.2 ± 0.1 (n=10) C959X: 1.10 ± 0.08 (n=8) G1013X: 1.08 ± 0.09 (n=9) L1314P: 1.03 ± 0.06 (n=11) R1515X: 1.01 ± 0.05 (n=9) Logarithmic transformation; Welch’s One-Way ANOVA (W=108.4, DFb = 6.0, DFw = 26.98), p<0.0001 (****) Dunnett’s post-hoc test p=0.0007 (***) p<0.0001 (****) p<0.0001 (****) p<0.0001 (****) p<0.0001 (****) p<0.0001 (****) Panel 3b Current density (pA/pF). Mean ± SEM WT50%: 37.1±4.6 (n=15) (control) WT50% + Kir4.1: 33.3 ± 3.4 (n=18) WT50% + R379H: 23.4 ± 1.6 (n=14) WT50% + R937H: 17.8 ± 1.7 (n=17) WT50% + C959X: 15.8 ± 2.5 (n=10) WT50% + G1013X: 15.9 ± 1.9 (n=14) WT50% + L1314P: 18.1 ± 1.3 (n=13) Logarithmic transformation; One-Way ANOVA (F=11.46, DFb = 7, DFw = 103), p<0.0001 (****) Dunnett’s post-hoc test p=0.99 p=0.035 (*) p<0.0001 (****) p<0.0001 (****) p<0.0001 (****) p<0.0001 (****) preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint Cestèle et al 2026 Supplementary material 7 WT50% + R1515X: 17.5 ± 2.2 (n=9) p=0.0002 (***) Figure 4 Panel 4b Current density (pA/pF). Mean ± SEM WT50%: 39.9±4.8 (n=15) (control) WT50% + Difopein: 32.7±3.2 (n=25) hNaV1.2-S487A50%: 39.9±5.2 (n=10) hNaV1.2-Δ523-53350%: 30.7±6.7 (n=8) Logarithmic transformation; One-Way ANOVA (F=1.38, DFb=3, DFw=54), p=0.26 Panel 4c Va (mV) Mean ± SEM WT50%: -13.6±1.1 (n=15) (control) WT50% + Difopein: -13.3±0.6 (n=25) hNaV1.2-S487A50%: -12.7±1.0 (n=10) hNaV1.2-Δ523-53350%: -11.0±1.2 (n=8) One-Way ANOVA (F=1.07, DFb=3, DFw=54), p=0.37 Ka (mV) Mean ± SEM WT50%: 6.0±0.4 (n=15) (control) WT50% + Difopein: 6.3±0.2 (n=25) hNaV1.2-S487A50%: 7.4±0.6 (n=10) hNaV1.2-Δ523-53350%: 6.4±0.4 (n=8) One-Way ANOVA (F=2.19, DFb=3, DFw=54), p=0.10 Vh (mV) Mean ± SEM WT50%: -49.8±1.3 (n=15) (control) WT50% + Difopein: -50.1±1.2 (n=22) hNaV1.2-S487A50%: -46.7±1.9 (n=10) hNaV1.2-Δ523-53350%: -50.4±1.1 (n=8) One-Way ANOVA (F=1.13, DFb=3, DFw=51), p=0.34 Kh Mean ± SEM WT50%: 6.6±0.4 (n=15) (control) WT50% + Difopein: 5.7±0.2 (n=25) hNaV1.2-S487A50%: 6.2±0.4 (n=10) hNaV1.2-Δ523-53350%: 6.7±0.6 (n=8) One-Way ANOVA (F=1.53, DFb=3, DFw=51), p=0.22 Panel 4d Current density (pA/pF). Mean ± SEM WT50%: 39.9±4.8 (n=15) (control) Difopein + … WT50% + R379H: 29.7 ± 4.7 (n=9) WT50% + R937H: 32.2 ± 5.1 (n=8) WT50% + C959X: 33.5 ± 3.6 (n=10) WT50% + G1013X: 42.1 ± 9.5 (n=10) WT50% + L1314P: 30.9 ± 5.6 (n=10) WT50% + R1515X: 43.8 ± 8.9 (n=6) Logarithmic transformation; One-Way ANOVA (F=0.54, DFb=6 , DFw=61), p=0.77 Panel 4e Current density (pA/pF). Mean ± SEM S487A 50%: 39.9±5.2 (n=10) (control) S487A 50% + R379H: 40.7 ± 9.9 (n=7) S487A 50% + R937H: 50.5 ± 12.4 (n=8) S487A 50% + C959X: 34.5 ± 4.6 (n=8) S487A 50% + G1013X: 48.0 ± 10.6 (n=8) S487A 50% + L1314P: 36.7 ± 4.8 (n=7) S487A 50% + R1515X: 41.6 ± 4.8 (n=5) Logarithmic transformation; One-Way ANOVA (F=0.31, DFb=6, DFw=46), p=0.93 Panel 4f Current density (pA/pF). Mean ± SEM Δ523-53450%: 30.7±6.7 (n=8) (control) Δ523-53450% + R379H: 26.7 ± 3.9 (n=7) Δ523-53450% + R937H: 22.0 ± 4.1 (n=5) Δ523-53450% + C959X: 30.4 ± 8.2 (n=7) Δ523-53450% + G1013X: 29.8 ± 5.2 (n=5) Δ523-534 50% + L1314P: 43.0 ± 8.5 (n=6) Δ523-534 50% + R1515X: 35.0 ± 7.7 (n=6) Logarithmic transformation; One-Way ANOVA (F=0.80, DFb=6, DFw=37), p=0.58 preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint Cestèle et al 2026 Supplementary material 8 Figure 5 Panel 5a Specific binding (fmol/mg) Mean ± SEM Median WT: 9.0 ± 1.6 (n=7) (control) 7.1 R379H: 1.7 ± 0.6 (n=4) 1.4 R937H: 0.06 ± 0.02 (n=4) 0.05 L1314P: 0.01 ± 0.004 (n=4) 0.01 Kruskal-Wallis test, pU) with Bonferroni correction: p=0.009 (**) p=0.009 (**) p=0.009 (**) Panel 5b Specific binding (fmol/mg) Mean ± SEM Median Mean ± SEM Median WT50%: 6.9 ± 0.9 (n=7) (control) 6.1 WT50% + R379H: 3.3 ± 0.1 (n=3) 3.3 WT50% + R937H: 2.1 ± 0.2 (n=3) 2.0 WT50% + C959X: 1.5 ± 0.2 (n=3) 1.5 WT50% + G1013X: 1.7 ± 0.1 (n=3) 1.8 WT50% + L1314P: 2.3 ± 0.1 (n=3) 2.4 WT50% + R1515X: 2.3 ± 0.1 (n=3) 2.2 Kruskal-Wallis test, p=0.001 Mann-Whitney post -hoc test (exact probability >U) with Bonferroni correction: p=0.0049 (*) p=0.0049 (*) p=0.0049 (*) p=0.0049 (*) p=0.0049 (*) p=0.0049 (*) Figure 6 Panel 6b Current density (pA/pF). Mean ± SEM WT: 216.2 ± 20.6 (n=14) (control) R850P: 78.8 ± 10.4 (n=16) V1289F: 102.6 ± 9.1 (n=19) T1420M: 57.7 ± 6.2 (n=12) R1635Q: 37.3 ± 5.1 (n=9) G959D: 83.4 ± 17.1 (n=8) D284G: 158.5 ± 20.1 (n=24) A896V: 21.6 ± 2.9 (n=14) C1344Y: 68.2 ± 8.3 (n=21) R1882X: 23.8 ± 4.0 (n=11) Logarithmic transformation; One-Way ANOVA (F=21.4, DFb=9, DFw=138) p<0.0001 (****) Dunnett’s post-hoc test P<0.0001 (****) P=0.004 (**) p<0.0001 (****) p<0.0001 (****) p<0.001 (**) p=0.13 p<0.0001 (****) p<0.0001 (****) p<0.0001 (****) Panel 6c preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint Cestèle et al 2026 Supplementary material 9 Current density (pA/pF). Mean ± SEM WT: 50.5 ± 6.8 (n=15) (control) R850P: 16.5 ± 1.7 (n=10) V1289F: 26.9 ± 3.8 (n=11) T1420M: 24.6 ± 2.8 (n=8) R1635Q: 47.9 ± 9.7 (n=8) G959D: 17.2 ± 2.4 (n=6) D284G: 35.6 ± 5.3 (n=16) A896V: 27.1 ± 4.1 (n=14) C1344Y: 20.6 ± 2.1 (n=8) R1882X: 16.0 ± 2.1 (n=7) Logarithmic transformation; One-Way ANOVA (F=6.1, DFb=9, DFw=93) p<0.0001 (****); Dunnett’s post-hoc test P0.99 p=0.0004 (***) p=0.13 p=0.004 (**) p=0.003 (**) p<0.0001 (****) Panel 6d Va (mV) WT: -17.5 ± 0.9 (n=15) (control) D284G: -15.4 ± 0.8 (n=16) R1635Q: -17.3 ± 0.9 (n=8) One-Way ANOVA (F=1.75, DFb=2, DFw=36), p=0.19 Ka (mV) WT: 5.8 ± 0.3 (n=15) (control) D284G: 8.1 ± 0.4 (n=16) R1635Q: 6.2 ± 0.3 (n=8) One-Way ANOVA (F=12.3, DFb=2, DFw=36), p<0.0001 (****); Dunnett’s post-hoc test p<0.0001 (****) p=0.68 Vh (mV) WT: -50.1 ± 1.2 (n=15) (control) D284G: -53.2 ± 1.8 (n=16) R1635Q: -62.9 ± 0.9 (n=8) One-Way ANOVA (F=13.6, DFb=2, DFw=36), p<0.0001 (****); Dunnett’s post-hoc test P=0.23 p<0.0001 (****) Kh (mV) WT: 6.3 ± 0.2 (n=15) (control) D284G: 5.9 ± 0.1 (n=16) R1635Q: 5.4 ± 0.1 (n=8) One-Way ANOVA (F=6.9, DFb=2, DFw=36), p=0.003 (**); Dunnett’s post-hoc test P=0.08 p=0.002 (**) preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint Cestèle et al 2026 Supplementary material 10 Panel 6e Current density (pA/pF). Mean ± SEM WT50%: 31.1±3.0 (n=13) (control) WT50% + R850P: 34.8 ± 4.6 (n=8) WT50% + V1289F: 45.7 ± 3.4 (n=6) WT50% + T1420M: 34.8 ± 7.6 (n=8) WT50% + R1635Q: 43.9 ± 3.4 (n=12) WT50% + G959D: 36.1 ± 6.9 (n=5) WT50% + D284G: 44.7 ± 4.7 (n=8) WT50% + A896V: 45.7 ± 5.5 (n=8) WT50% + C1344Y: 35.2 ± 3.5 (n=13) WT50% + R1882X: 45.8 ± 2.7 (n=5) Logarithmic transformation; One-Way ANOVA (F=2.2, DFb=9, DFw=76) p=0.03 (*); Dunnett’s post-hoc test p=0.99 p=0.11 p=0.99 p=0.07 p=0.99 p=0.12 p=0.10 p=0.97 p=0.14 Figure 7 Panel 7c Current density (pA/pF). Mean ± SEM WT: 374.5 ± 70.9 (n=9) (control) R379H: 41.6 ± 4.0 (n=6) R850P: 111.9 ± 23.2 (n=7) V1289F: 111.1 ± 18.1 (n=12) T1420M: 70.7 ± 6.9 (n=12) R1635Q: 165.9 ± 36.4 (n=7) G959D: 98.9 ± 15.0 (n=10) D284G: 205.3 ± 29.8 (n=8) A896V: 27.1 ± 4.1 (n=9) C1344Y: 14.0 ± 2.3 (n=12) R1882X: 54.8 ± 4.7 (n=11) One-Way Welch ANOVA (W=20.5, DFn=10.0, DFd=32.6) p<0.0001 (****); Dunnett’s post-hoc test p=0.013 (*) p=0.046 (*) p=0.047 (*) p=0.022 (*) p=0.17 p=0.035 (*) p=0.34 p=0.010 (*) p=0.008 (**) p=0.016 (*) Panel 7d Current density (pA/pF). Mean ± SEM WT: 122.0 ± 24.1 (n=9) (control) R379H: 7.4 ± 1.1 (n=6) R850P: 25.2 ± 3.1 (n=7) V1289F: 25.7 ± 2.4 (n=12) T1420M: 22.0 ± 4.3 (n=12) R1635Q: 18.8 ± 3.7 (n=7) G959D: 8.7 ± 0.9 (n=10) D284G: 33.0 ± 6.0 (n=8) A896V: 4.1 ± 0.9 (n=9) C1344Y: 2.4 ± 0.3 (n=12) R1882X: 10.8 ± 1.1 (n=11) One-Way Welch ANOVA (W=25.5, DFn=10.0, DFd=32.1) p<0.0001 (****); Dunnett’s post-hoc test p=0.012 (*) p=0.032 (*) p=0.033 (*) p=0.023 (*) p=0.023 (*) p=0.013 (*) p=0.048 (*) p=0.010 (*) p=0.009 (**) p=0.014 (*) preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.24.707547doi: bioRxiv preprint