Overactive EAAT1 Cl⁻ channels impair GABAergic tonic inhibition in SLC1A3-associated episodic ataxia

Abstract

A missense variant in SLC1A3, which results in a proline to arginine substitution in the glial excitatory amino acid transporter 1 (EAAT1), causes a severe form of episodic ataxia type 6, characterized by recurrent attacks of ataxia and epilepsy. EAATs are dual function transport proteins, and the P290R variant reduces secondary active glutamate transport rates, while enhancing the anion channel activity. Here, we used complementary electrophysiological, imaging, biochemical and neuropathological techniques to characterize the cellular mechanisms underlying epileptic seizures in a mouse model of episodic ataxia type 6, the heterozygous Slc1a3^P290R/+ mouse. Video-EEG recordings demonstrate frequent and severe spontaneous epileptic seizures in vivo. “Gliosis only” phenotype in the cerebrum of Slc1a3^P290R/+ mice, which was restricted to the hippocampal formation, suggests that this brain structure may be involved in the development of epileptic seizure activity. Electrophysiological recordings from acute slices revealed a reduction in tonic GABAergic transmission in Slc1a3^P290R/+ dentate gyrus granule cells, and to a lesser extent in cortical layer 2/3 pyramidal neurons before seizure onset. Phasic GABAergic and glutamatergic transmission remained unaltered in juvenile animals of the same developmental stage. There was no difference in expression levels of the GABA_A receptor (GABA_AR) δ subunits, suggesting that impaired tonic inhibition results from decreased extrasynaptic GABA concentrations. We identified enhanced GABA uptake by hippocampal radial glia-like cells (RGLs), caused by elevated GABA transporter 3 (GAT-3) expression and by an increased GABA transport driving force—due to lower intracellular chloride levels [Cl^–]_int – as cellular basis of impaired tonic inhibition. Our study demonstrates how increased EAAT1 chloride channel activity of the P290R variant can cause hyperexcitability by modifying synaptic GABA concentrations, while impaired EAAT1 glutamate transport leaves glutamatergic synaptic transmission in Slc1a3^P290R/+ mice unaffected.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40478-026-02399-4.

Introduction

Excitatory amino acid transporters (EAATs) regulate glutamatergic synaptic activity by mediating glutamate uptake into neuronal and glial cells [11, 49, 59]. Beyond their transporter function, EAATs also operate as chloride channels, with channel opening and closing driven by conformational changes during the glutamate transport cycle [19, 20, 40]. In the past, variants of SLC1A3 (encoding EAAT1) were associated with episodic ataxia type 6, a disorder characterized by the co-occurrence of ataxia, hemiplegic migraine and epilepsy [7–9, 13, 27, 28]. One SLC1A3 variant predicts the p.Pro290Arg (P290R) substitution that markedly enhances EAAT1 anion channel activity while simultaneously impairing glutamate transport [26, 57, 63]. A knock-in mouse model (Slc1a3^P290R/+) carrying the heterozygous P290R variant in the rodent EAAT1 homologue, the glutamate aspartate transporter (GLAST), recapitulates the neurological symptoms observed in the corresponding human patient [37]. In contrast, homozygous knock-out animals (Slc1a3^−/−) exhibit only mild ataxia [56, 62], supporting the idea that epilepsy in episodic ataxia type 6 is triggered by a gain-of-function in EAAT1 anion channel activity, rather than by impaired glutamate transport. Ataxia in Slc1a3^P290R/+ mice was recently shown to be caused by Bergmann glia apoptosis, which is triggered by increased glutamate-activated chloride (Cl^–) efflux during infancy and leads to cerebellar degeneration [37]. However, cerebellar degeneration cannot account for the severe epileptic phenotype of Slc1a3^P290R/+ mice. Outside the cerebellum, EAAT1 and its rodent homologue GLAST are highly expressed in hippocampal radial glia-like cells (RGLs), a population of glial stem cells in the subgranular zone of the dentate gyrus (DG) that can differentiate into either granule neurons or astrocytes [4, 30]. Since RGLs are functionally and spatially coupled to granule cells [46], functional changes in EAAT1/GLAST might modify synaptic transmission in DG neurons and contribute to network dysfunction. To understand the cellular basis of epilepsy in Slc1a3^P290R/+ animals, we characterized the epileptic phenotype in vivo and studied neurotransmission in hippocampal and cortical neurons in vitro using electrophysiological, imaging, biochemical and histological approaches.

Material and methods

Animals

WT (wildtype) and heterozygous (Slc1a3^P290R/+) mice of both sexes from the 129 substrain 129S6/SvEvTac (Taconic Biosciences) were used for experiments conducted between P10 and P100. Animals were maintained under a 12-h light/dark cycle with ad libitum access to food and water. The common health status was monitored regularly 4 times per year. Each Slc1a3^P290R/+ mouse was assessed 2 × per week, starting from P15, with a scoring system and euthanasia criteria, which were proofed by the use commitees in North-Rhine Westphalia and Baden Württemberg. For more detail, please refer to Supplementary material.

In vivo video-EEG monitoring

Male Slc1a3^P290R/+ and WT animals were implanted with wireless EEG transmitters (ETA-F10 or HD-X02, Data Sciences International (DSI)) during two developmental time windows: P25-P32 (before seizure onset: N = 5/5) and P44-P53 (after seizure peak: N = 8/10, WT/Slc1a3^P290R/+). A minimum body weight of 14 g (juvenile group) and 20 g (older group) was required. Anaesthesia was induced either by a single intraperitoneal injection of fentanyl (5 mg/kg, Dechra Veterinary Products Deutschland GmbH), midazolam (5 mg/kg, hameln pharma Gmbh) and medetomidine (0.5 mg/kg, Zoetis Inc.), also referred to as 3-component-anaesthesia, or by 1.5% isoflurane (CP Pharma Handelsgesellschaft mbH) in O₂ via a face mask, with buprenorphine (0.1 mg/kg, CP Pharma Handelsgesellschaft mbH) administered for analgesia. Only the younger group received anaesthesia with buprenorphine. Carprofen (5 mg/kg, Zoetis Deutschland GmbH) was administered intra- and postoperatively. Animals were positioned in a stereotaxic frame equipped with a heating pad (Stoelting Co.). Small holes were drilled to expose the dura and electrodes were placed directly onto its surface (1 mm anterior to bregma, 1 mm lateral), then secured with dental cement (Supplementary Fig. S1a). The transmitter was positioned in a lateral subcutaneous pocket. Antagonisation of anaesthesia was achieved using naloxone (1.2 mg/kg, hameln pharma GmbH), atipamezole (2.5 mg/kg, Dechra Veterinary Products Deutschland GmbH) and, in older mice only, flumazenil (0.5 mg/kg, Fresenius Kabi Deutschland GmbH). Anaesthesia with buprenorphine was not antagonised.

Video-EEG monitoring was initiated 3–7 days post-surgery, or earlier if seizures were suspected, and recordings were performed between P25–P32 (juvenile group) and P44–P53 (older group). Monitoring was performed in a suitable undisturbed environment. In the younger group, the average EEG recording duration per heterozygous animal was 37.0 h for mice receiving the 3-component-anaesthesia and 46.9 h receiving the buprenorphine anaesthesia. In this group, the average EEG recording duration of the respective WT littermates was comparable. In the older group, the average recording duration per animal was 49.7 h for WT and 37.8 h for heterozygous mice. The difference in recording times reflects the higher mortality rate in the Slc1a3^P290R/+ group. Despite this, the total recording duration was similar between genotypes (WT: 347.6 h, Slc1a3^P290R/+: 378.4 h). One WT mouse died postoperatively, but the cause of death remained unclear. Mice were monitored in their home cages (PhysioTel receiver, Ponemah software, DSI; Noldus software, Noldus Information Technology BV), with sampling rates of 1.0 kHz (ETA-F10) or 0.5 kHz (HD-X02). EEG analysis was performed using Neuroscore software (DSI). Seizures were defined as rhythmic discharges lasting > 10 s and < 5 min, with or without behavioural manifestations. Status epilepticus was defined as seizure activity > 5 min or recurrent seizures without full recovery.

Following the experiment, animals were euthanised by intraperitoneal injection of ketamine (150 mg/kg, Wirtschaftsgenossenschaft deutscher Tierärzte eG) and xylazine (15 mg/kg, Albrecht GmbH).

Neuropathology and immunostainings

For microscopic analyses, mouse brain tissues were fixed with paraformaldehyde overnight. For paraffin sections, brains were embedded into paraffin after dehydration with a series of increasing [EtOH] and Rotihistol (Sigma-Aldrich, Merck). 4 μm sections were used for both, hematoxylin–eosin (HandE) staining and immunohistochemistry by using standard protocols. Briefly, paraffin sections were de-paraffinised in xylene for 10 min followed by decreasing alcohol series (100–50%) for 2 min each. After washing, slides were subjected to a citric acid antigen retrieval procedure with inactivation of endogenous peroxidases by incubation in a solution containing (in mM): 21 C₆H₈O₇, 6 Na₂HPO₄, and 15% (v/v) H₂O₂ (15 min, RT). Antigen retrieval was performed by boiling the sections for 20 min in Tris–EDTA-buffer (2 Tris, 0.063 EDTA, pH 8.0). Subsequently, brain sections were rinsed for 2 h at 37 °C in PBS blocking buffer (10% normal goat serum). Slices were washed and incubated with antibodies against glial fibrillary acidic protein (GFAP, Sigma, 1:500) overnight at 4 °C. After washing, slices were incubated with respective biotinylated secondary antibodies (Vector Laboratories, USA, 1:200) in blocking buffer and incubated for 2 h at 37 °C. Slides were again washed and further processed according to manufacturer’s protocol (Sigma-Aldrich) and mounted with Mowiol 4–88 (Carl Roth GmbH).

Immunostainings of cryo-protected tissues for confocal microscopy were prepared as followed: Brains were immersion-fixed in 4% PFA in phosphate buffer (PB; in mM: 81 NaH₂PO₄·2H₂O, 9 Na₂HPO₄·H₂O, pH 7.4) for 30 min at 4 °C, followed by washing in PB for 30 min at RT. Cryoprotection was achieved by sequential incubation in PB containing 10% sucrose (1 h, RT) and 30% sucrose (overnight, 4 °C). Hemispheres were embedded in NEG-50 frozen section medium (Thermo Fisher Scientific) and sectioned sagittally at 14 μm using an HM560 cryotome (Thermo Fisher Scientific). Non-specific binding was blocked overnight at 4 °C in 1% (w/v) bovine serum albumin in PB. Frozen and paraffin sections were incubated overnight at 4 °C with primary antibodies against brain lipid-binding protein (BLBP), integrin α-M (CD11b), GABA transporter-3 (GAT-3), glial fibrillary acidic protein (GFAP), or Glutamate-Aspartate transporter (GLAST) (RRIDs: AB_10000325, AB_2650514, AB_304437, AB_1556315, AB_10829302), followed by incubation with cyanine dye-conjugated secondary antibodies (RRIDs: AB_2340370, AB_2340460, AB_2340612, AB_2340813) for 45 min at RT. All antibodies were diluted in PB supplemented with 5% ChemiBLOCKER (Merck Millipore) and 1% Triton X-100. RGLs were quantified by counting cells positive for three proteins (BLBP, GLAST, and GFAP) in individual confocal planes. GAT-3 fluorescence intensity was normalised to the BLBP fluorescence intensity per cell for analysis in Fig. 4j

Fig. 4.

Changes in intracellular ion concentrations in radial glia-like cells (RGLs) of Slc1a3^P290R/+ mice. a Confocal imaging of RGLs (DG) and astrocytes (CA1) confirm expression of EAAT1/GLAST (EAAT1/GLAST: yellow; BLBP—brain lipid binding protein (magenta); glial fibrillary acidic protein (GFAP): blue). EAAT1/GLAST-positive main processes of RGLs are indicated by arrowheads. Scale bar: 5 µm. b Multi-photon overview image showing a maximal projection of the Na⁺ sensitive indicator SBFI (green) and the glia cell exclusive staining with sulforhodamine 101 (SR101, red). White triangles highlighting radial glia like cells. Scale bar: 20 µm. c Boxplots summarizing SFBI recordings from WT and Slc1a3^P290R/+ animals between P22 and P27 (n = 34/40, N = 5/4, WT/Slc1a3^P290R/+, p = 0.09). d Representative fluorescence lifetime imaging microscopy (FLIM) recordings in the hippocampal subgranular zone (P25). Arrowheads mark SR101-positive RGLs; color-coded scale indicates [Cl⁻]_int. Scale bar: 25 µm. e Mean [Cl⁻]_int in RGLs from wildtype (WT) (grey circles) and Slc1a3^P290R/+ (orange circles) mice at P20–P30 (n = 622/616 (small circles) from N = 6/7 (big circles), WT/Slc1a3^P290R+, p = 0.004). Application of the EAAT1/GLAST-specific inhibitor UCPH-101 increased RGL [Cl⁻]_int in both genotypes (n = 423/272, N = 8/7, WT/Slc1a3^P290R/+, p ≤ 0.001/ ≤ 0.001, WT/Slc1a3^P290R+), although a significant difference between the genotypes remained (p = 0.0012, two-way ANOVA, Holm–Šidák post hoc tests). f Mean [Cl⁻]_int in CA1 astrocytes from individual WT (grey) and mutant (orange) mice at P20–P30 (n = 206/277, N = 9/8, WT/Slc1a3^P290R+, p = 0.53). Boxplots in (e) and (f) show distributions of [Cl⁻]_int from individual WT (grey) and Slc1a3^P290R/+ (orange) for animals (large circles, N) and single cells (small circles, n). g Western blot analysis of GLAST and GAT-3 expression in hippocampal tissue before seizure onset (b) at P10 (N = 4/4, p = 0.0247^GLAST/0.07^GAT−3) and during seizure onset (so) at P25 (N = 3/4, WT/Slc1a3^P290R+, p = 0.02^GLAST/0.09^GAT−3, two-tailed t-tests within age groups). h Quantification of western blot analyses in (g). i Representative confocal images of BLBP-positive (magenta) RGLs co-expressing GAT-3 (cyan; scale bar: 10 µm). j Mean GAT-3 fluorescence intensities normalised to BLBP signal at the single-cell level. Boxplots compare means (squares) and medians (lines) from individual animals (N = 3/4, WT/Slc1a3.^P290R+, p = 0.043, two-tailed t-test)

Whole-cell patch clamp recordings

For electrophysiological analyses, mice of both sexes, either prior to the paroxysmal period between P15 and 25, or after seizure onset at P40, were deeply anesthetized with isoflurane and rapidly decapitated. Brains were immediately transferred to ice-cold Ringer’s solution A (in mM: 125 NaCl, 2.5 KCl, 1.25 NaH₂PO₄, 26 NaHCO₃, 0.5 CaCl₂, 5 MgCl₂, 20 C₆H₁₂O₆; bubbled with 5% CO₂/95% O₂). Parasagittal 250 or 260 µm-thick slices and coronal 350 µm-thick slices were prepared using vibratomes (υ = 60/90 Hz, amplitudes = 1.0/0.9 mm; 7000 SMZ2 (Campden Instruments)/Microm HM650V (Thermo Fisher Scientific)). Slices were transferred to oxygenated Ringer’s solution B (125 NaCl, 2.5 KCl, 1.25 NaH₂PO₄, 26 NaHCO₃, 2 CaCl₂, 1 MgCl₂, 25 C₆H₁₂O₆) for recovery: 30 min at 37 °C followed by 90 min at RT for parasagittal slices and 60 min at 37 °C for coronal slices.

Whole-cell patch-clamp recordings from neurons in the hippocampus and in neocortical layer 2/3 were performed in acute brain slices, using either an EPC10-USB amplifier (RRID: SCR_018399) with PATCHMASTER software (HEKA Elektronik, RRID: SCR_000034) or a Multiclamp 700B amplifier (RRID: SCR_018455) and DigiData 1420 with pClamp 10.6 software (Molecular Devices, RRID: SCR_011323). Recordings were conducted at RT (hippocampus) or 34 °C (cortex) with constant perfusion of Ringer’s solution B supplemented with 50 μM (2R)-amino-5-phosphovaleric acid (AP-5) (Tocris Bioscience). Patch pipettes were filled with intracellular solution containing (in mM): 130 CsCl, 8 NaCl, 0.2 MgCl₂, 2 EGTA, 4 Mg-ATP, 0.3 Li-GTP, 1 QX-314, and 10 HEPES (pH 7.2, adjusted with CsOH). Cells were voltage-clamped at -70 mV.

GABAergic mIPSCs and tonic GABA currents were recorded in the presence of 1 μM tetrodotoxin (TTX) (Tocris Bioscience) and 10 μM cyanquixaline (CNQX) (Tocris Bioscience). Tonic GABA_AR-mediated conductance was quantified by comparing the mean holding current over 30 s before and after 3 min application of 50 µM PTX. α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor-mediated mEPSCs were recorded in the presence of 1 µM TTX and 100 μM picrotoxin (PTX) (Tocris Bioscience). Whole-cell patch-clamp recordings of electrogenic GABA-uptake currents in DG RGLs were performed in acute slices (260 µm) from P19-P23 mice. To isolate GABA-uptake currents, Ringer B solution was supplemented with 100 µM PTX and 1 µM 3-(1-(3,4-dichlorophenyl)ethyl)amino-2-hydroxypropyl-phenyl-methyl-phosphinicacid-hydrochloride (CGP55845) to block GABA_A and GABA_B receptors, respectively. Patch pipettes were filled with two different intracellular solutions containing different Cl^– concentrations ([Cl^–]) based on the mean values obtained from FLIM microscopy of DG RGLs (WT: 20.8 mM KCl; P290R/ + : 18 mM KCl). Equimolar substitution was performed to achieve a final potassium concentration ([K⁺]) of 140 mM using potassium methanesulfonate (K-Met-SO₄) (Sigma-Aldrich) (in mM, WT: 119.2 K-Met-SO₄/20.8 KCl; P290R/ + : 122 K-Met-SO₄/18 KCl; both genotypes: 2 Mg-gluconate, 1 Ca-gluconate, 3 Mg-ATP, 2 Na-GTP, 10 HEPES–KOH, pH 7.4). Acute slices were incubated with 2 µM sulforhodamine 101 (SR101) (Sigma-Aldrich) for 20–30 min at 37 °C to label RGLs, followed by a 20 min incubation in dye-free Ringer B solution, and visualised using infrared microscopy (Slicescope, Scientifica). GABA uptake responses of RGLs held at -70 mV were evoked by brief puff applications (5 s) of Ringer B solution containing 0.5 mM GABA, GABA_A- and GABA_B blockers, using a pneumatic drug ejection system (PDES-DXH, npi electronic GmbH). Maximum response amplitudes were recorded in individual cells, averaged per cell and then pooled across animals (N) for statistical analysis.

Sodium imaging of RGLs

Sodium imaging was performed on brain slices from male and female mice aged P22 to P27 that had no prior history of overt seizures. Detailed informations about intracellular sodium imaging and the technical equipment is provided in Supplementary material.

Fluorescence lifetime imaging microscopy

We measured [Cl^–]_int in hippocampal astrocytes and RGLs in acute brain slices from animals between P20 and P30 (prior and at the beginning of overt seizure onsets), using fluorescence lifetime imaging (FLIM) and the Cl^– sensitive quinolinium dye 1-(ethoxycarbonylmethyl)-6-methoxyquinolinium bromide (MQAE) (Sigma-Aldrich). Detailed information about the used technical equipment is provided in the Supplementary material. The mean fluorescence lifetimes were determined using multidimensional time-correlated single-photon counting (TCSPC, Simple-Tau 152; Becker and Hickl) with a volume of 0.08 mm³ per individual pixel, resulting in a three-dimensional resolution of approximately 1.3 mm in the z-axis and 0.35 mm in the x- and y-axis [67]. Prior to imaging, acute brain slices were incubated in oxygenated Ringer’s solution A containing 2 μM SR101 for 20 min at 37 °C, followed by a ten-minute incubation in SR101-free oxygenated Ringer’s solution B to stain astrocytes and RGLs. After 30 min at RT, slices were incubated with MQAE (3.5 mM, Sigma-Aldrich, Merck) in Ringer's solution B for 30 min. MQAE fluorescence is collisionally quenched by Cl⁻ ions, resulting in a linear relationship between the inverse fluorescence lifetime and [Cl^–]_int [32, 60]:

where τ is the MQAE fluorescence lifetime at a given [Cl⁻]_int, τ₀ is the MQAE fluorescence lifetime in the absence of Cl^–, and K_SV is the cell type-specific Stern–Volmer constant. After calibration in astrocytes and RGLs [18] using the two-ionophore calibration method [32, 38], the [Cl^–]_int for individual cell somas was calculated from the mean fluorescence lifetime of all pixels within the region of interest. To block EAAT1/GLAST activity, slices were incubated in Ringer’s solution B supplemented with 20 μM of the EAAT1/GLAST blocker UCPH-101 (Abcam) for 20 min.

Western blotting

For details on Western blotting procedures and used antibodies, please refer to Supplementary Material.

Quantitative receptor autoradiography

Quantitative receptor autoradiography was performed as previously described [66] using animals of both sexes after the paroxysmal period at P50. For further details, please refer to Supplementary material.

Dendritic spine quantification

Golgi–Cox staining was performed following established protocols [48]. The analysis was done with the open-source software Synapse Web Reconstruct (RRID: SCR_002716). For details on analysis, please refer to Supplementary material.

Statistical analysis

Statistical analyses were performed using SigmaPlot (Systat Software GmbH, RRID: SCR_003210), Microcal-Origin (OriginLab; RRID: SCR_002815) or GraphPad Prism 9.5.1 (GraphPad Software, RRID: SCR_002798). Unless stated otherwise, data are presented as boxplots, where the box represents the interquartile range (IQR: 25th to 75th percentile), the horizontal line indicates the median and the black squares indicate the data means. Whiskers span the minima and maxima of means within animals (N). No outliers were excluded from the analysis. For patch-clamp recording and [Cl^–]_int data, individual cells (n) are plotted as small symbols, while animal-level averages (N) are indicated by larger symbols. In the main text, [Cl^–]_int are reported as means ± SD, calculated from individual animal averages. Datasets with normal distributions were compared using one or two-way ANOVA with Holm–Šidák, Bonferroni or Tukey post hoc tests or in case of single dual datasets with two-tailed t-tests. Non-normally distributed data, as determined by Shapiro–Wilk or Kolmogorov–Smirnov-tests, were analysed using Kruskal–Wallis ANOVA on ranks or Mann–Whitney U-tests. P-values ≤ 0.05 were considered statistically significant. Wherever feasible, experiments were performed in a blinded manner. However, due to the overt phenotype, particularly in experiments involving older animals (> P15), which had to be monitored regularly, or the use of different genotype-dependent pipette solutions in electrophysiological recordings, blinding of investigators was not always possible. Data supporting this study are freely available at: https://github.com/peterkovermann/episodic_ataxia_6_II/.

Results

Slc1a3^P290R/+ mice exhibit severe epilepsy

Slc1a3^P290R/+ mice exhibit severe epilepsy [37], with disease onset typically occurring during the third postnatal week, and peak seizure susceptibility between three and four weeks. Seizures typically begin with rapid tail movements and head turning, progressing to whole-body stretching and clonic activity, while severe episodes involve tonic–clonic seizures with loss of startle reflexes (Fig. 1a, Supplementary video SV1). Postictal behaviour includes immobility and reduced responsiveness. Due to the severity of the phenotype, many Slc1a3^P290R/+ animals reached predefined euthanasia criteria, primarily due to ongoing seizures or weight loss, preventing accurate mortality quantification. Seizure susceptibility in Slc1a3^P290R/+ animals was associated with significant reductions in body weight (Fig. 1b).

Fig. 1.

Seizure burden in Slc1a3^P290R/+ mice changes across development. a Sequential stages of seizures, beginning with tail lifting (1) and lowering (2), followed by body extension (3) and clonic movements (4). Severe episodes include tonic–clonic activity and transient loss of startle reflexes (5). Reduced responsiveness occurs postictally (6). b Age-dependent changes in body weight of wildtype (WT, grey) (N = 25) and Slc1a3^P290R/+ (orange, N = 42) animals; individual mice are shown as small symbols. Shading indicates the seizure peak. c Timeline of transmitter implantations and video-EEG monitoring in animals before/during (P25-P36) and after the peak seizure period (P44-P60). d Representative EEG trace showing rhythmical spike-and-wave discharges in a P29 mutant animal after implantation under 3-component anaesthesia during seizure peak, with a magnified 10 s segment of the ictal event (cut-out location indicated). e Representative EEG trace showing rhythmical spike-and-wave discharges in a mutant animal during seizure peak under buprenorphine-based anaesthesia, with a magnified 10 s segment of the ictal event (cut-out location indicated). f Percentage of animals that suffered from seizures/status or died/euthanised in EEG-monitored Slc1a3^P290R/+ mice during seizure peak (3-components and buprenorphine-anaesthesia cohorts) and post-peak period (young group: N = 5 Slc1a3^P290R/+ animals; older group: N = 10 Slc1a3.^P290R/+ animals). g Representative EEG trace from a mutant animal recorded during a later stage of epileptogenesis, with a gradual increase in spike-and-wave discharge amplitude. Insets display 10 s and 5 s magnifications of the ictal event (cut-out locations indicated)

To further characterize the epileptic phenotype, video-EEG monitoring was performed in heterozygous and WT mice during and after peak seizure susceptibility in young (P25-P32) and older (P45-P60) cohorts (young group: N = 5/5 mice per subgroup; old group: N = 8/10 mice, WT/Slc1a3^P290R/+). The young cohort consisted of two subgroups implanted under either 3-component- or buprenorphine-based anaesthesia (Fig. 1c, Supplementary Fig. 1a), as the initial 3-component regimen resulted in 100% incidence of fatal seizures/status epilepticus. In young Slc1a3^P290R/+ mice, EEG monitoring was initiated within 24 h post-implantation and revealed frequent ictal activity across both anaesthesia regimens during the peak susceptibility window (P25-P32), with differences in frequency and severity (Fig. 1d, e). In the subgroup implanted under 3-component-anaesthesia, all Slc1a3^P290R/+ mice exhibited severe epileptic activity with recurrent seizures and no recovery of consciousness between events. EEG recordings showed profoundly suppressed interictal activity, consistent with continuous status epilepticus (Fig. 1d, Supplementary Fig. 1b). Four Slc1a3^P290R/+ animals reached euthanasia criteria, and one animal died overnight during monitoring, highlighting a high mortality rate during the peak seizure period (Fig. 1f). In contrast, young Slc1a3^P290R/+ mice implanted under buprenorphine-based anaesthesia displayed a milder and more heterogeneous phenotype. Two mice developed prolonged near-continuous status epilepticus, whereas three showed only isolated seizures and/or status events (Fig. 1e, f). Overall, seizure burden and mortality were reduced compared to the 3-component-anaesthesia group, with only two animals reaching endpoint criteria (Fig. 1f). The animals exhibited one to four ictal events (seizures and status epilepticus) per mouse. Individual seizures were rare (one to two per animal; 12 s and 3.6 min, mean 1.5 min), while status epilepticus occurred as discrete or near-continuous episodes (zero to three per animal) lasting 0.26 to 47.70 h.

In the post-peak group (P45-P60), seven out of ten mice exhibited at least one seizure or status epilepticus, while three were seizure-free; six animals reached euthanasia criteria or died during monitoring (Fig. 1f). Compared to younger mice, older animals showed a less severe epileptic phenotype with intermittent seizure or status epilepticus events on EEG (Fig. 1g). The number of ictal events per Slc1a3^P290R/+ animal ranged from one to 25 ictal events, with all affected mice experiencing at least one episode of status epilepticus. Seizures lasted on average 48.8 s (range: 12 s to 2.4 min) and occurred one to 15 times per day (mean five). Status epileptici lasted 0.1 to 10.8 h (mean 2.2 h) and were accompanied by prolonged periods of altered consciousness with intermittent convulsive behaviours.

Tonic GABAergic inhibition is impaired in Slc1a3^P290R/+ hippocampal granule cells

To investigate potential mechanisms underlying increased seizure susceptibility in Slc1a3^P290R/+ mice, we examined synaptic transmission in the hippocampus and also in the cortex – two brain regions frequently implicated in the pathogenesis of genetic epilepsies [2]. Epileptic activity is commonly associated with an imbalance between excitation and inhibition, often resulting from impaired GABA_AR-mediated inhibition [64]. We therefore performed whole-cell patch-clamp recordings in acute hippocampal and coronal brain slices from WT and heterozygous mice, targeting DG granule cells, CA1 pyramidal neurons, and layer 2/3 pyramidal neurons in the somatosensory cortex. Tonic GABAergic inhibition arises from the persistent activation of extrasynaptic GABA_ARs and plays a critical role in controlling neuronal excitability through Cl⁻-mediated shunting inhibition and hyperpolarisation [64]. Even subtle alterations in tonic current amplitude can have a substantial impact on neuronal excitability [45]. Tonic GABAergic currents were pharmacologically isolated using the GABA_AR antagonist PTX and quantified as the difference in holding current before and after PTX application (Fig. 2a). In DG granule cells of P20 Slc1a3^P290R/+ mice—prior to the onset of spontaneous seizures—tonic GABA_AR-mediated currents were significantly reduced compared to WT controls, by approximately 63% (P ≤ 0.001; two-way ANOVA, Holm–Šidák post hoc tests; Fig. 2b). This deficit was reversed by pharmacological inhibition of the glial glutamate transporter EAAT1/GLAST with UCPH-101 [1], restoring tonic current amplitudes to WT levels (p = 0.80; Fig. 3b). Tonic inhibition declined with age in DG granule cells of WT animals, in line with previous reports [25], but this age-dependent reduction was absent in Slc1a3^P290R/+ mice, suggesting altered homeostatic regulation of GABA levels in the mutant genotype (Fig. 2b). Tonic GABAergic currents in layer 2/3 cortical pyramidal neurons of WT mice were approximately 42% smaller compared to DG neurons, but significantly reduced in Slc1a3^P290R/+ mice before the onset of seizures (P15-P20) (p = 0.043; two-tailed t-test; Fig. 2c).

Fig. 2.

GABA_AR-mediated tonic inhibition is reduced in Slc1a3^P290R/+ dentate gyrus (DG) granule cells. a Representative tonic currents recorded from juvenile (P20) WT (top) or Slc1a3^P290R/+ (bottom) DG granule cells with the start of the PTX (50 µM) application marked by an arrow. b Mean amplitudes of tonic currents at P20 (n = 11/18 Slc1a3^P290R/+/WT cells, N = 6/9 animals; p ≤ 0.001) and P40 (n = 11/8, N = 3/3, WT/Slc1a3^P290R/+, p = 0.99, two-way ANOVA, Holm–Šidák post hoc tests). UCPH-101 treatment increases tonic current amplitudes in mutant DG granule cells at P20 (n = 17, N = 5, Slc1a3^P290R/+, p = 0.044). Comparison of WT controls and the Slc1a3^P290R/+  + UCPH-101 group revealed no significant difference (p = 0.80). c Mean tonic current amplitude of L2/3 cortical pyramidal neurons at P15-P20 (n = 14/15, N = 7/10, p = 0.043; two-tailed t-test). d Western blots of GABA_AR δ subunit expression in hippocampal and cortical tissues at P20. Na⁺/K⁺-ATPase and vinculin served as loading controls. The graphs show the ratios of GABA_AR δ subunit expression in Slc1a3^P290R/+ mutant tissues by WT control values for the hippocampus (d’, N = 4/4^P20, 3/3^P40, p = 0.195^P20/0.888^P40) and for the cortex (d’’, N = 12/11^P20, WT/Slc1a3^P290R/+, p = 0.115, two-tailed t-tests). e Representative current responses from WT (top) or Slc1a3^P290R/+ (bottom) RGLs to puffs of 0.5 mM GABA for 5 s (holding potential of -70 mV). f Mean GABA uptake current amplitudes of WT and Slc1a3^P290R/+. Shown are individual responses to GABA applications (small symbols) and means of animals (N = 8/12, WT/Slc1a3.^P290R/+, p = 0.021, two-tailed t-test -Welch corrected)

Fig. 3.

Phasic GABAergic transmission is not altered in Slc1a3^P290R/+ mice. a Representative miniature inhibitory postsynaptic current (mIPSC) traces recorded from dentate gyrus (DG) granule cells of wildtype (WT) and Slc1a3^P290R/+ mice at P20. b-d Mean mIPSC amplitudes (I), frequencies (ν) and decay times (τ) recorded from DG granule cells at P20 (before seizure onset; n = 19/26 cells from N = 7/8 animals, p = 0.82^I/0.42^ν/0.95^τ, WT/Slc1a3^P290R+) and P40 (after seizure onset; n = 14/13, N = 3/3, WT/Slc1a3^P290R+, p = 0.86^I/0.62^ν/0.43^τ). e Representative mIPSC traces recorded from layer 2/3 (L2/3) pyramidal neurons of the somatosensory cortex from WT and Slc1a3^P290R/+ mice (before seizure onset). f–h Mean mIPSC amplitudes, frequencies and decay times recorded from L2/3 cortical pyramidal cells at P15-P20 (before seizure onset; n = 11/18, N = 4/10, WT/Slc1a3^P290R+, p = 0.74^I/ 0.24^ν/0.13^τ). Patch clamp data were analysed with one-way ANOVAs within ages and Holm–Šidák post hoc tests

Tonic GABAergic currents are predominantly mediated by δ subunit-containing GABA_ARs [64], and we therefore analysed hippocampal and cortical tissue lysates for changes in δ subunit expression by western blotting. Hippocampal and cortical δ subunit levels did not differ between Slc1a3^P290R/+ and WT samples (Fig. 2d, Hippocampus: (d ‘), p = 0.195^P20/0.888^P40, Cortex: (d ‘‘), p = 0.115, two-tailed t-tests), suggesting that the observed reduction in tonic inhibition was not due to altered receptor abundance. Notably, only male mice were used for the analysis to avoid confounding effects related to δ subunit expression fluctuations during the ovarian cycle [43]. These findings point toward a decrease in ambient GABA levels, rather than receptor expression, as the likely cause of diminished tonic inhibition in the DG and somatosensory cortex.

In the hippocampus, EAAT1/GLAST is expressed in astrocytes as well as in RGLs of the DG [4, 30]. RGLs are spatially associated with DG granule cells [46], suggesting a role in regulating the synaptic and extrasynaptic neurotransmitter environment of these neurons. Changes in GABA clearance by RGLs may thus contribute to the reduction of tonic inhibition in Slc1a3^P290R/+ mice. RGLs take up GABA via the Na⁺-dependent GAT-3 transporter, and the electrogenic GABA uptake can be quantified by electrophysiological approaches [5, 33]. We evoked GAT-3-mediated currents by pressure GABA puffs in whole-cell patch clamp recordings from RGL cells at holding potentials of -70 mV (Fig. 2e). To isolate GAT-3 currents from other current components, GABA_A and GABA_B receptors were blocked with PTX (100 µM) and CGP55845 (1 µM), respectively [12, 51]. We found GAT-3 currents in mutant RGLs to be approximately twofold increased as compared to WT (p = 0.021, Fig. 2f, two-tailed t-test, Welch-corrected).

We next assessed phasic inhibition by recording GABA_AR-mediated miniature inhibitory postsynaptic currents (mIPSCs) in DG granule cells (Fig. 3a) and layer 2/3 cortical pyramidal neurons (Fig. 3e). The amplitude, frequency, and decay kinetics of mIPSCs did not differ between Slc1a3^P290R/+ mice and their WT littermates, indicating preserved synaptic GABAergic transmission in both hippocampal and cortical circuits during early postnatal development (Fig. 3b–d, f–h). Both, tonic and phasic inhibitory current amplitudes in CA1 pyramidal neurons remained unaffected at this developmental stage (p = 0.67^tonic/0.39^phasic, Supplementary Fig. 2). The regional specificity reinforces the DG as a key site of early tonic inhibitory dysfunction in this epilepsy model, but also suggests a potential, though less significant, contribution from the cortex to epileptogenesis. After seizure onset, both phasic and tonic GABAergic currents in DG granule cells were comparable between genotypes (Figs. 2b, 3b–d), suggesting a transient and developmentally specific deficit in inhibition. Our observations suggest that enhanced GAT-3 mediated GABA uptake reduce extrasynaptic GABA levels and thereby contribute to diminished tonic inhibition.

Intracellular [Cl⁻] is reduced in radial glia-like cells, but not in hippocampal astrocytes of Slc1a3^P290R/+ mice

EAAT1/GLAST is expressed in hippocampal RGLs and CA1 astrocytes (Fig. 4a). GABA transport into glial cells is driven by Na⁺ and Cl^– gradients [41], and changes in EAAT1 function may modify glial [Na⁺]_int—via coupled Na⁺/H⁺/glutamate: K⁺ transport – as well as glial [Cl^–]_int [18, 58] via EAAT anion channel function. Ratiometric measurements using the Na⁺ indicator SBFI-AM [39] revealed no differences in [Na⁺]_int between WT and mutant RGLs (Fig. 4b, c; p = 0.09, N = 8/7 slices). Using fluorescence lifetime imaging microscopy (FLIM) and the Cl⁻-sensitive dye MQAE, we quantified [Cl^–]_int in RGLs (Fig. 4d, e) and CA1 astrocytes (Fig. 4f) from P20–P30 mice of both sexes [18, 58]. RGLs from Slc1a3^P290R/+ mice displayed significantly reduced [Cl^–]_int compared to WT RGLs (Fig. 4e; 20.8 ± 1.0/18.0 ± 0.8 mM, ± CI, N = 6/7 mice, WT/Slc1a3^P290R/+; p = 0.004; one-way ANOVA, Holm–Šidák post hoc test). Incubation with the specific EAAT1/GLAST blocker UCPH-101 (20 μM) restored [Cl^–]_int in mutant RGLs to levels close to those of untreated WT cells (Fig. 4e; p = 0.471, N = 6/7, WT/Slc1a3^P290R/+). In the presence of UCPH-101, [Cl^–]_int was lower in Slc1a3^P290R/+ (Fig. 4e) than in WT RGLs (Fig. 4e; 24.4 ± 1.6 mM / 21.4 ± 0.7 mM, p = 0.0012, N = 8/7, WT/Slc1a3^P290R/+; one-way ANOVA, Holm–Šidák post hoc test), suggesting a compensatory upregulation of alternative Cl⁻-extruding mechanisms in RGLs of Slc1a3^P290R/+ animals [32]. In contrast, [Cl^–]_int in CA1 astrocytes did not differ between genotypes under baseline conditions (Fig. 4f; 20.8 ± 1.3/20.2 ± 1.3 mM, N = 9/8 mice, WT/Slc1a3^P290R/+; p = 0.53; one-way ANOVA, Holm–Šidák post hoc test).

Western blot analysis of hippocampal lysates showed a non-significant tendency for GAT-3 upregulation in Slc1a3^P290R/+ mice before (b, P10) and at the age of seizure onset (so, P25) (Fig. 4g, h; p = 0.068^b/0.086^so; two-tailed t-tests). EAAT1/GLAST protein levels were reduced in hippocampal lysates of heterozygous mice—by 29% in the second postnatal week and by 56% at seizure onset (Fig. 4h, i; P10^b: p = 0.025; P25^so: p = 0.021; two-tailed t-tests). Co-immunostaining for GAT-3 and BLBP confirmed their co-localisation in RGL somata and their vertically oriented processes extending into the DG granule cell layer (Fig. 4i). Direct quantification of fluorescence intensities by fluorescence microscopy (Fig. 4i, j) demonstrated a significantly increased GAT-3 signal within RGLs of P25 Slc1a3^P290R/+ mice compared to WT controls (N = 3/4, WT/Slc1a3^P290R/+, p = 0.043, two-tailed t-test).

We conclude that the P290R-associated gain-of-function in EAAT1/GLAST anion channel activity reduces the [Cl^–]_int in hippocampal RGLs and simultaneously increases GAT-3 expression in RGLs. The resulting increase in the driving force for GAT-3-mediated GABA uptake, together with elevated RGL-specific GAT-3 expression, may reduce extracellular GABA levels and consequently tonic inhibition in the DG of Slc1a3^P290R/+ mice.

Hippocampal RGLs are reduced in Slc1a3^P290R/+ mice during the postictal phase

In the cerebellum of Slc1a3^P290R/+ animals, apoptosis of Bergmann glial cells is triggered by the aberrantly increased anion channel activity caused by the P290R variant [37]. To test whether a similar glial pathology occurs in the hippocampus of Slc1a3^P290R/+ mice, we quantified RGLs in the DG from WT and mutant mice of both sexes at stages preceding and following seizure onset using immunohistochemistry. Confocal imaging of hippocampal section immunolabeled for the glial markers BLBP, GFAP, and GLAST revealed comparable RGL densities in WT and mutant animals in the subgranular zone of the DG prior to seizure onset (P12-15; Fig. 5a, b; p = 0.3, two-tailed t-test). In contrast, a significant reduction in RGL density was observed in Slc1a3^P290R/+ mice older than P40 (Fig. 5a, b; p = 0.021, two-tailed t-test). The overall density of granule cells remained unchanged in mutant mice after epilepsy onset (Fig. 5c; p = 0.7, two-tailed t-test). At P27, we detected CD11b-positive microglia containing GFAP-positive cell fragments in the subgranular zone of Slc1a3^P290R/+ mice, consistent with active phagocytosis of astroglial cells (Fig. 5d). This was accompanied by pronounced hippocampal astrocytosis, characterized by hypertrophy of astrocytic somata and processes, and microgliosis, indicated by increased CD11b expression (Fig. 5e, f).

Fig. 5.

Structural and cellular alterations in the hippocampus of adult Slc1a3^P290R/+ mice. a Representative confocal images of the dentate gyrus (DG) region from wildtype (WT) (P27, top) or Slc1a3^P290R/+ mice before (P20) and after (P40) age of seizure onset, immunolabeled for brain lipid binding protein (BLBP) (magenta), EAAT1/GLAST (yellow) and GFAP (blue). A marked reduction in radial glia-like cells (RGLs) is evident in mutants following seizure onset (middle: P23^bfo, bottom: P27^aso, scale bar: 100 µm). b RGL densities in WT and mutant animals at P20 and P40: P20^bfo, N = 4/4, WT/Slc1a3^P290R+, p = 0.29; P40^aso, N = 7/7; p = 0.006) c Comparison of granule cell (GC) numbers from WT and Slc1a3^P290R/+ mice at ages after seizure onsets (aso) shows no significant difference between genotypes (p = 0.7, N = 5/4, WT/Slc1a3^P290R+, two-tailed t-test). d Confocal images of the subgranular zone in a mutant DG (P27) depicting phagocytic events. Glial fibrillary acidic protein (GFAP)-positive (blue) astrocytes display condensed apoptotic nuclei (TOPRO-3, grey) surrounded by CD11b-positive (cyan) microglial processes. Scale bar: 10 µm. e Overview of hippocampal sections stained for GFAP and CD11b, with nuclear counterstain (TOPRO-3). f Higher magnification of the stratum lacunosum-moleculare reveals hypertrophic GFAP-positive astrocytic processes and increased microglial branching in mutants, consistent with reactive astrocytosis and microgliosis after seizures (P27; scale bars in (d) and (e): 100 µm). g Western blot analysis of doublecortin X (DCX) expression in hippocampal tissue at P10 (N = 4/4), P20 (N = 5/5) and P40 (N = 3/3, WT/Slc1a3^P290R+). Na⁺/K⁺-ATPase served as loading control. The graph shows DCX expression in mutants relative to WT levels (dashed line); p = 0.0025^P10/0.0072^P20/0.1741^P40)

To assess whether impaired RGL survival affects neurogenesis, we examined expression of doublecortin X (DCX), a marker of immature granule cells, by western blot analysis. DCX protein levels were reduced in Slc1a3^P290R/+ hippocampi prior to seizure onset but returned to control levels at P40 (Fig. 5g; p = 0.0025^P10/0.0072^P20/0.174^P40; two-tailed t-tests), indicating a transient impairment of neurogenesis during early postnatal development. Since RGLs serve as neural progenitors that give rise to granule cells in the DG, their loss is likely to directly impact the generation of new neurons. Taken together, these observations indicate that expression of the P290R EAAT1/GLAST mutant in RGLs disrupts hippocampal neurogenesis through a mechanism independent of seizure activity.

Slc1a3^P290R/+ mice show an abnormal reactive astrogliosis in the hippocampus

Hippocampal formations from Slc1a3^P290R/+ mice displayed a neuroanatomically regular architecture (Fig. 6), without significant neuronal rarefication, particularly no segmental loss of neurons at different time points after birth, including P5, P10 and P35. The density of DG granule cells appeared unaltered at all examined ages, and granule cell dispersion was absent. At P5, the pyramidal band and DG showed no relevant infiltration by reactive astrocytes, whereas by P35, astroglial cell elements were frequently found adjacent to neuronal elements, as revealed by GFAP immunohistochemistry (Fig. 6). This cellular reactive astrogliosis affecting the hippocampal formation emerged at P10 and was neuropathologically most pronounced at P35. Age-matched WT animals do not show astrogliosis (Fig. 6). More chronic fibrillary astrogliotic changes were not observed. The changes observed in Slc1a3^P290R/+ resemble the ‘gliosis only’ phenotype of temporal lobe epilepsy (TLE) [23, 24], while hippocampal sclerosis, as the most frequently pathological change observed in TLE [3], was absent.

Fig. 6.

Slc1a3^P290R/+ hippocampi show a ´gliosis only´ phenotype. Hippocampal formations of brains isolated from Slc1a3^P290R/+ at days 5 and 10 after birth (P5/10, far ahead from paroxysmal period) and 35 (P35, within paroxysmal period) and WT mice (P35) stained with hematoxylin and eosin (HE, upper row) and subjected to immunohistochemistry with antibodies against glial fibrillary acidic protein (GFAP, second row). Below, higher power magnifications of CA1 (from purple squares in ‘GFAP’) and dentate gyrus (DG; from light blue squares in ‘GFAP’; all scale-bars: 200 µm)

Excitatory transmission is unaltered prior to seizure onset of Slc1a3^P290R/+

Ionotropic glutamate receptors, including amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors (AMPARs) and N-methyl-D-aspartate ionotropic glutamate receptors (NMDARs), play distinct roles in epileptiform activity, with AMPARs being particularly important for the initiation of epileptic discharges [5]. To evaluate AMPAR-mediated synaptic transmission in the hippocampus, we recorded miniature excitatory postsynaptic currents (mEPSCs) in DG granule cells and CA1 pyramidal neurons from both juvenile (P20) and adult (P40) Slc1a3^P290R/+ and WT animals of both sexes (Fig. 7a, e). Recordings at P20 revealed similar mEPSC amplitudes, frequencies, and decay kinetics in both genotypes and both tested regions (Fig. 7b–d, f–h). In CA1 neurons, mEPSC amplitudes were slightly increased in P40 Slc1a3^P290R/+ mice compared to WT controls (p = 0.023; Fig. 7f), whereas the frequency and decay times remained unchanged (Fig. 7g–h). In contrast, mEPSC characteristics in DG granule cells remained unaltered between Slc1a3^P290R/+ and WT mice at P40 (Fig. 7b–d). AMPAR-mediated mEPSCs in layer 2/3 pyramidal neurons of the somatosensory cortex at a juvenile, pre-seizure stage (P15–20) (Fig. 7i) neither differed in mEPSC amplitudes, frequencies nor decay times between juvenile heterozygous and WT mice (Fig. 7j–l).

Fig. 7.

Excitatory transmission is unaltered before seizure onset in Slc1a3^P290R/+ mice. a Representative miniature inhibitory postsynaptic current (mEPSC) traces recorded from dentate gyrus (DG) granule cells at P20 (before seizure onset) and P40 (after seizure onset). b–d mEPSC amplitudes (I), frequencies (ν) and decay times (τ) of mEPSCs in DG granule cells at P20 (n = 17/28^I,ν,29^τ WT/Slc1a3^P290R/+ cells, N = 10/6 mice, p = 0.44^I/0.32^ν/0.9^τ) and P40 (n = 20/33, N = 8/11, p = 0.21^I/0.33^ν/0.14^τ). e Representative mEPSCs recorded from CA1 pyramidal neurons at P40. f–h Amplitudes, frequencies and decay times of mEPSCs in CA1 pyramidal neurons at P20 (n = 18/14, N = 7/4, WT/Slc1a3^P290R/+, p = 0.43^I/0.72^ν/0.46^τ) and P40 (n = 16/25, N = 6/10^I,11^ν,τ, p = 0.02^I/0.48^ν/0.36^τ). i Representative mEPSC traces recorded from layer 2/3 (L2/3) pyramidal neurons of the somatosensory cortex at P15-P20. j–l Amplitudes, frequencies and decay times of mEPSCs in L2/3 pyramidal cells at P15-P20 (n = 14/15, N = 6/6, p = 0.17^I/0.12^ν/0.65^τ). Patch-clamp data were analysed using one-way ANOVAs within ages and Holm–Šidák post hoc tests

To test whether the increased mEPSC amplitude in CA1 pyramidal neurons after seizure onset is associated with altered neurotransmitter receptor expression, we used quantitative autoradiography to measure AMPAR, GABA_BR, and metabotropic glutamate 2/3 receptor (mGlu2/3R) densities in WT and mutant mice from both sexes [66]. We found increased AMPAR densities in the caudatum putamen, motor and sensory cortices, and the hippocampal region of Slc1a3^P290R/+ mice at P45 (Fig. 8a–c). The AMPAR density was elevated in the CA1 region, consistent with the increased mEPSC amplitude detected in CA1 pyramidal cells (Fig. 7f, 8c). In contrast, the AMPAR density was reduced in the cerebellum of Slc1a3^P290R/+ animals (Fig. 8d), likely reflecting the progressive glial and neuronal degeneration in this brain region [37]. The cerebellar downregulation was also observed for other tested receptors. Densities of all other tested receptors were unaffected in the cerebrum of adult Slc1a3^P290R/+ mice, except for mGlu2/3 receptors, which were decreased by approximately 37% in the CA1 region (data not shown as figure, for raw data refer to raw data file at https://github.com/peterkovermann/episodic_ataxia_6_II/). We conclude that AMPAR-mediated excitatory synaptic transmission in the hippocampus and neocortex is largely unaltered prior to seizure onset in Slc1a3^P290R/+ mice (Fig. 7), suggesting that glutamatergic signalling is not a driving factor in epileptogenesis.

Fig. 8.

Hippocampal AMPAR densities and neuronal spine density are altered in adult Slc1a3^P290R/+ mice. a Contrast-enhanced color-coded (inset) AMPAR densities in the brain of wild-type (left) and Slc1a3^P290R/+ mice (right). Three cutting planes (I–III) show I, the caudate putamen (Cpu), the motor (M) and sensory (S), cortex; II, the hippocampus (Hip); and III, the cerebellum (Cb). The assigned color scale represents equally spaced density ranges in pmol/mg protein. Scale bar: 1000 µm b–d AMPAR densities from the indicated brain regions (Cpu: p = 0.042^Cpu/0.0086 ^M/0.0046^S/0.0013^CA1/0.0015^CA2/3/0.0056^DG/0.0021^Cb, two-tailed t-tests, N = 10/10 animals, WT/Slc1a3^P290R/+). e Representative Golgi–Cox-stained secondary and tertiary dendrites of WT and Slc1a3^P290R/+ granule cells (> P70, scale bar: 5 µm). f Mean spine lengths in DG and CA1 neurons (DG: 1.13 ± 0.12 /0.72 ± 0.15 µm; CA1: 0.95 ± 0.2/0.75 ± 0.09 µm, means ± SD, WT/Slc1a3^P290R+, p = 0.004^DG/0.16^CA1, two-tailed t-tests). g Mean spine density measured in WT and mutant animals in DG and CA1 neurons (DG: 1.73 ± 0.18/1.37 ± 0.07 spines/µm; CA1: 1.20 ± 0.44/1.12 ± 0.7 spines/µm, p = 0.045^DG/0.884^CA1, WT/Slc1a3^P290R+, two-tailed t-tests). (f) and (g) present data for dendritic spines in 10 μm dendritic segments of DG granule cells and CA1 pyramidal cells (large symbols: N = 3/3, WT/Slc1a3.^P290R+, small symbols: means from individual slices)

To identify possible activity-dependent structural adaptations that may dampen hippocampal excitability and restrict epileptic activity, we analysed dendritic spine morphology in Golgi-Cox-stained DG granule cells and CA1 pyramidal cells from Slc1a3^P290R/+ mice after the peak-seizure period. Spine lengths were significantly reduced in secondary and tertiary dendrites of DG granule cells, but not in CA1 pyramidal neurons (Fig. 8e-f; p = 0.004/0.16, two-tailed t-tests). Additionally, spine density was selectively decreased in DG neurons (Fig. 8g; p = 0.035, two-tailed t-test). These structural alterations, including spine retraction, may reflect a compensatory form of synaptic pruning aimed at restraining network hyperexcitability in the epileptic hippocampus.

Discussion

We here investigated the cellular pathophysiology of a severe form of episodic ataxia type 6 associated with the P290R variant in the human SLC1A3 gene [28] using the Slc1a3^P290R/+ knock-in mouse model that recapitulates key features of the human phenotype [37]. Slc1a3^P290R/+ animals exhibit a pronounced epileptic phenotype, particularly during a critical time window in early postnatal development (Fig. 1). Video-EEG recordings revealed epileptic seizures and status epileptici at juvenile and later stages with increased seizure severity and high seizure-associated mortality, in juvenile mice (Fig. 1). Histological analysis revealed a pronounced astrogliosis in the hippocampus (Fig. 6), whereas other brain areas, including the cortex, showed no comparable gliotic response (data not shown). Regionally confined astrogliosis is a well-established marker of chronic neuronal hyperexcitability and is frequently observed at epileptogenic foci, particularly in hippocampal and limbic epilepsies [54]. The presence of hippocampal gliosis supports the hippocampal formation, rather than the neocortex, as the origin of seizure generation in Slc1a3^P290R/+ mice. In acute slices from mice, we did not detect any differences in the amplitude, frequency or decay kinetics of AMPA receptor-mediated mEPSCs in CA1 pyramidal neurons, DG granule cells or layer 2/3 cortical pyramidal neurons between WT and Slc1a3^P290R/+ animals at P20 (Fig. 7), indicating intact excitatory synaptic transmission at this developmental stage. Likewise, phasic GABAergic inhibitory transmission in DG granule cells and layer 2/3 cortical pyramidal neurons remains unaltered (Fig. 3). However, tonic GABAergic currents were markedly reduced in DG granule cells (Fig. 2), and only slightly attenuated in layer 2/3 cortical pyramidal neurons. Since EAAT1/GLAST is exclusively expressed in glial cells and not in neurons [31, 50], with a high expression in DG RGLs [15], a direct impact on DG granule cell excitability is not expected. Taken together, these findings indicate that glutamate transporter dysfunction contributes to the observed epilepsy in Slc1a3^P290R/+ mice by impairing tonic inhibition in the hippocampus, without significantly affecting glutamatergic synaptic transmission.

The DG plays a key role in regulating hippocampal excitability, and tonic GABAergic inhibition of granule cells is particularly important for modulating hippocampal network activity [10]. Disruption of tonic GABAergic inhibition is a well-established mechanism in epileptogenesis. Tonic GABAergic currents are mediated by GABA_ARs containing δ subunits, which confer high sensitivity to ambient GABA and are typically located extrasynaptically [21]. Deletion of the δ subunit in mice leads to spontaneous seizures [52, 55], and variants that impair either the open probability or membrane trafficking of GABA_ARs are associated with human epilepsy syndromes [22]. Moreover, δ subunit abundance varies across the ovarian cycle and correlates with fluctuations in seizure susceptibility, further highlighting the critical role of tonic inhibition in modulating network excitability [43].

In Slc1a3^P290R/+ DG granule cells, we observed reduced GABAergic tonic inhibition despite unchanged GABA_AR δ subunit expression levels (Fig. 2). These findings indicate that the attenuation of tonic inhibition is not due to altered receptor availability, but is likely driven by decreased ambient GABA levels in the microenvironment of granule cells. Decreased ambient GABA levels were associated with enhanced GABA uptake by mutant RGLs (Fig. 2e, f). GABA transporters (GATs) are driven by the electrochemical gradients of Na⁺ and Cl^– [41], and we detected diminished [Cl^–]_int in RGLs from Slc1a3^P290R/+ mice, which will increase the driving force for glial GABA uptake. Application of the EAAT1/GLAST-specific inhibitor UCPH-101 restored both tonic GABAergic inhibition (Fig. 2b) and [Cl^–]_int in RGLs to WT levels (Fig. 4e), indicating a mechanistic link between these alterations and P290R EAAT1/GLAST gain-of-anion channel function. RGL [Cl^–]_int was only reduced by about 15% as compared to WT values, a rather small change to account for the observed threefold increase of GAT-3 currents. However, we can only determine [Cl^–]_int in the RGL soma, so that larger differences appear possible in glial extensions close to extrasynaptic localizations of granule cells. Moreover, we observed increased GAT-3 expression levels in young animals (Fig. 4g–j). [Na⁺]_int were not different between WT and mutant RGLs (Fig. 4b, c). It is tempting to speculate that enhanced GAT-3-mediated Na⁺ inward transport in microdomains near granule cells could locally stimulate Ca²⁺ uptake via increased Na⁺/Ca²⁺ exchange [52], which may promote GAT-3 upregulation [53]. These mechanisms could further reinforce GABA clearance from the extracellular space, thereby exacerbating the deficit in tonic inhibition.

We observed astrocytic and microglial activation in Slc1a3^P290R/+ animals (Fig. 5a–f), which may result from seizures [16, 61, 68] and may further exacerbate epileptic activity. After the fifth postnatal week, increased amplitudes of AMPAR-mediated mEPSCs were recorded in CA1 pyramidal cells (Fig. 7), and autoradiography experiments revealed increased AMPAR densities in the hippocampus of Slc1a3^P290R/+ animals (Fig. 8). The release of tumor necrosis factor alpha (TNF-α) by activated microglia, in conjunction with astrocytes [44], has been shown to enhance synaptic AMPAR number [47]. Moreover, epileptic discharges may directly potentiate AMPAR-mediated synaptic transmission in hippocampal neurons [14, 29], leading to excessive glutamatergic excitation and thus epileptic synchronisation. EEG recordings in Slc1a3^P290R/+ animals during the post-peak phase revealed persistent seizures and status epileptici, albeit with notably reduced severity and frequency. Interestingly, in some older animals, seizures were absent during the recording period. At this later stage, spine density in the DG was reduced, and dendritic spines in both DG and CA1 showed signs of retraction (Fig. 8f, g). Seizure activity may have led to activity-dependent elimination of synaptic connections at later time points, potentially serving a protective role for hippocampal neurons in these animals [44, 47].

Whereas Slc1a3^P290R/+ Bergmann glia undergo apoptosis triggered by cell shrinkage at the onset of glutamatergic synaptic transmission [37], we observed only a modest reduction in the number of hippocampal RGLs (Fig. 5b). Apoptotic events in the hippocampal subgranular zone of Slc1a3^P290R/+ mice (Fig. 5d), along with downregulation of DCX (Fig. 5g), a marker of immature neurons, suggest that RGLs undergo apoptosis, but do not prematurely differentiate, thus preventing depletion of the RGL pool [17]. These findings imply that RGLs are less prone to apoptosis compared to Bergmann glia. A potential explanation for the differential responses between the two cell types may lie in the lower [Cl^–]_int in RGLs, as opposed to Bergmann glial cells [18, 58].

Our results suggest that impaired tonic GABAergic inhibition in hippocampal granule cells, resulting from the dysfunction of the EAAT1/GLAST-associated anion channel, triggers the epileptic phenotype. The increased Cl^– efflux through the mutant EAAT1/GLAST leads to a reduction in [Cl^–]_int in RGLs, facilitating GABA uptake via glial GAT-3. This depletion of ambient GABA around DG granule cells diminishes tonic GABAergic currents, which in turn induces hyperexcitability in the DG and contributes to the onset of epilepsy. Enhanced AMPAR-mediated synaptic transmission does not trigger but may contribute to the maintenance of epileptic discharges and facilitate prolonged seizures in older animals. We speculate that repetitive epileptic activity contributes to RGL loss, diminishes spine density and triggers dendritic spine retraction in principal hippocampal neurons, potentially helping to limit seizure frequency and severity at later stages (Fig. 1).

In recent years, an increasing number of neurological conditions has been linked to variants in SLC1A2 or SLC1A3. Contrary to earlier assumptions, not the impairment of glutamate transport, but rather changes in EAAT1/2 anion channel function seem to trigger the disease [7, 34, 36, 37, 63, 65]. In certain cases of epileptic encephalopathies, SLC1A2 mutations impair the selectivity of EAAT2 anion channels, resulting in glutamate conduction and presynaptic glutamate release [34, 36]. However, most pathogenic SLC1A2 or SLC1A3 variants leave anion channel selectivity intact and solely enhance anion channel activity [7, 34].

We, here link gain-of-anion channel function of P290R EAAT1 to increased GABA uptake and impaired tonic GABAergic inhibition as basis of hippocampal hyperexcitability. Inhibition of EAAT1/GLAST with the specific EAAT1 inhibitor UCPH-101 [1] restores the impaired tonic inhibition to WT levels (Fig. 2b), thus identifying selective EAAT1-block as promising therapeutic approach for SLC1A3-associated episodic ataxia. We hypothesize that GABAergic and glutamatergic neurotransmitter systems are functionally linked via the two EAAT transport functions, i.e. secondary-active glutamate transport and anion channel function [6, 19, 35, 42, 58]. Glutamate release from synapses activates EAAT1 anion channels, which reduce glial [Cl^–]_int and subsequently enhance GABA uptake, leading to decreased tonic GABA_AR-mediated inhibition. This mechanism may contribute to adjusting the balance between excitatory and inhibitory synaptic transmission under physiological conditions. The severe epileptic phenotype in Slc1a3^P290R/+ mice represents a pathophysiological consequence of disrupted interaction between glutamatergic and GABAergic signalling arising from glial EAAT1/GLAST dysfunction.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Y.K., D.B-M., T.G., J.M., C.R., A.B., U.H., Ch.F and P.K. designed the research; Y.K., D.B-M., M.E., J.M., J.B., S.Be, S.Bl, and P.K. performed data aquisition and analysis. Y.K., D.B-M., J.B., J.M., A.B., U.H., Ch.F. and P.K. drafted the figures and the manuscript.

Funding

Open Access funding enabled and organized by Projekt DEAL. The project was supported from the Federal Ministry of Education and Research (E-RARE network Treat-ION: 01GM1907C to Ch.F. and 01GM1907G to U.H.; Project SynGluCross to C.R.R.) and the German Research Foundation (FOR 2795; FA 301/13–1 to Ch.F.; Project #461542557, RO 2327/13–2,14–2 to C.R.R.).

Data availability

All data supporting this study are publicly available on GitHub at the following URL: https://github.com/peterkovermann/episodic_ataxia_6_II/

Declarations

Ethics approval and consent to participate

All animal procedures were performed in accordance with the Animal Research guidelines (ARRIVE): Reporting of In Vitro and In Vivo experiments was approved by the appropriate animal care and use committees [LANUV (State Agency for Nature, Environment and Consumer Protection) of North Rhine-Westphalia (reference numbers: 84-02.04.2014.A334, 81-04.04.2020.A441), Regierungspräsidium Tübingen of Baden-Württemberg (reference number: N18/20G)]. Subsets of the data presented in this manuscript have been previously included in the doctoral theses from the two first authors YK and DBM.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.