HETEROZYGOUS GABA_A RECEPTOR β3 SUBUNIT N110D KNOCK-IN MICE HAVE EPILEPTIC SPASMS

Abstract

Objective—

Infantile spasms are anis an epileptic encephalopathy of childhood, and its pathophysiology is largely unknown.We generated a heterozygous knock-in mouse with the human infantile spasms-associated de novo mutation GABRB3(c.A328G, p.N110D) to investigate its molecular mechanisms and to establish the Gabrb3^+/N110D knock-in mouse as a model of infantile spasm syndrome.

Methods—

We used electroencephalography (EEG) and video monitoring to characterize seizure types, and a suite of behavioral tests to identify neurological and behavioral impairment in Gabrb3^+/N110D knock-in mice. Miniature inhibitory postsynaptic currents were recorded from layer V/VI pyramidal neurons in somatosensory cortex, and extracellular multiunit recordings from the ventral basal nucleus of the thalamus in a horizontal thalamocortical slice were used to assess spontaneous thalamocortical oscillations.

Results—

The infantile spasms-associated human de novo mutation GABRB3(c.A328G, p.N110D) caused epileptic spasms early in development and multiple seizure types in adult Gabrb3^+/N110D knock-in mice. Signs of neurological impairment, anxiety, hyperactivity, social impairment, and deficits in spatial learning and memory were also observed. Gabrb3^+/N110D mice had reduced cortical miniature inhibitory postsynaptic currents and increased duration of spontaneous oscillatory firing in the somatosensory thalamocortical circuit.

Significance—

The Gabrb3^+/N110D knock-in mouse has epileptic spasms, seizures and other neurological impairments that are consistent with infantile spasm syndrome in patients. Multiple seizure types and abnormal behaviors indicative of neurological impairment both early and late in development suggest that Gabrb3^+/N110D mice can be used to study the pathophysiology of infantile spasms. Reduced cortical inhibition and increased duration of thalamocortical oscillatory firing suggest perturbations in thalamocortical circuits.

Introduction

Infantile spasm syndrome (ISS) are an epileptic encephalopathy of childhood with unknown pathophysiology. Clinical features include clusters of epileptic spasms, severe developmental delay with cognitive impairment, and often hypsarrhythmia, an interictal high amplitude, chaotic pattern seen on the electroencephalogram (EEG)¹, with age of onset between 3 and 12 months². Spasms are characterized as flexor, extensor, and mixed flexor-extensor involving proximal and truncal muscles and lasting 1–2 seconds³. Despite early diagnosis and treatment, prognosis for ISS is poor⁴, and most ISS patients develop other refractory epilepsy syndromes⁵. Neurodevelopmental delay is common in ISS (~75%), with about half of patients having cerebral palsy⁵. Other behavioral characteristics include attention deficit hyperactivity disorder, anxiety, depression, learning difficulties, and autism⁶. The majority (60–75%) of ISS cases are attributed to acquired factors⁷, and the remaining (25–40%) cases are most likely genetic, primarily arising from de novo mutations⁸.

A de novo mutation in the γ-aminobutyric acid receptor type A (GABA_A) receptor β3 subunit gene (GABRB3) (c.A328G, p.N110D) was found in an ISS patient⁹. GABA_A receptors are pentameric, ligand-gated ion channels assembled from a panel of 19 subunits, most commonly composed of 2α, 2β, and 1γ subunits¹⁰. GABA_A receptors mediate the majority of fast inhibitory neurotransmission, and dysfunction of these receptors is associated with several epilepsy syndromes¹¹. The GABA_A receptor β3 subunit is widely expressed throughout the brain and is abundant in embryonic and neonatal brains, especially in the cerebral cortex, hippocampus, and other regions involved in seizure generation¹². The β3 subunit plays an important role in neurodevelopment¹³, and mutations in GABRB3 have been associated with childhood absence epilepsy¹⁴ and epileptic encephalopathies including ISS and the Lennox-Gastaut syndrome ⁹.

The GABRB3(N110D) is a rare mutation and the N110 residue is highly conserved across GABA_A receptor subunits located in the N-terminal α2 helix, a structurally conserved region that stabilizes inter-subunit interfaces by a network of hydrogen bonds and salt bridges¹⁵. However, the impact of the mutation in in vivo is unknown. In this study, we generated the heterozygous Gabrb3^+/N110D KI mouse harboring the human de novo GABRB3 ISS mutation, c.A328G, p.N110D. The KI mouse recapitulated the seizures and behavioral abnormalities associated with IS due to loss of neuronal β3 subunit-mediated inhibition, making it useful for further investigation of mechanisms underlying its pathophysiology and development of new treatments for ISS.

Methods

Gabrb3^+/N110D KI mice and neurobehavioral tests

The Gabrb3^+/N110D mouse line was generated with a targeted G328A mutation at the endogenous Gabrb3 locus using conventional homologous recombination in B6N embryonic stem (ES) cells (Primogenix, Inc) derived from the C57BL/6N mouse. The procedures for the KI mouse construction are the same as those for generation of the Gabrb3^+/D120N mice¹⁶ except the mutation was at G328A instead of G358A of exon 4. The procedures for all neurobehavioral tests including Elevated ZeroEZ maze, Open FieldOF and Barnes maze test as well as the concern of sex effect are detailed in Supplementary Method. The mice used for neurobehavioral tests and seizure experiments were from different cohorts.

Behavioral observations and Vigabatrin treatment in P14-P17 mice

Mice were observed for 20 minutes at a time from P10 until weaning. Twitches or spasms were recorded. .Mice were dosed with vigabatrin (100 mg/kg, i.p.) daily, on P13-P16 and observed for 20 minutes on P14-P17. We focused on P14-P17 because we did not observe obvious spasm in pups prior to P14 in our pilot study. Mice were injected after the observation period to minimize stress-induced effects. The tails of the mice were marked with a Sharpie in a unique pattern until the animals were old enough to be genotyped. The cohorts for observation and vigabatrin treatment were separate.

EEG electrode implantation, video-EEG recording, and analysis

Synchronized video-EEGs were recorded after electrode implantation, and video-EEG monitoring was performed with a monitoring system from Pinnacle Technology as previously described¹⁷. Briefly, KI and wild-type littermate mice (2 to 6-months old) were anesthetized using 2–4% isoflurane, and 5 mg/kg Ketofen (s.c.) was administered. For EEG headmount affixation, t As described in a previous study^18–20. Four (PMID: 35840120; PMID: 27340224; PMID: 31022638), fourFour holes for holding the electrode screws were drilled through the skull to the dura mater, with two placed 3.0 to 3.5 mm anterior to the bregma and two placed 3.0 to 3.5 mm posterior to the bregma and each being 1.5 mm lateral to the central sulcus. Although the exact location varies a bit based on user placement but typically EEG 1 covers the posterior region and EEG 2 spans frontal to parietal region. The EEG 1 lead typically includes more hippocampal activity as it is closer to the posterior hippocampus while the EEG 2 lead includes more delta activity as it incorporates frontal regions. They are designed to give an overall global view of cortical activity. The EMG(electromyography) electrodes were inserted into the nuchal muscles. Following a seven-day recovery period, the mice were moved to video-EEG recording chambers. The head mounts were connected to a pre-amplifier (#8202-SE), which connected to the 3-channel data acquisition and conditioning system (#8206-SE, Model 4100) to record from 2 EEG channels and 1 EMG channel. Video-EEG monitoring lasted for 24 h with freely moving mice in the chamber. The acquisition rate for these channels was 400 Hz. The pre-amplifier filtered at 1 Hz high pass and amplified the signal at a gain of 100X. Video-EEG-EMG data were analyzed offline with Sirenia® Seizure software (Supplementary Method). The detailed EEG analysis has been described in our previous study. Seizure events were automatically detected by the software but were manually confirmed with synchronized video monitoring to minimize detection artifact. The reviewer was blind to the genotypes.

Whole cell slice recording

Acute brain slices were prepared according to previously published methods²¹. Coronal brain slices (300 μm) were cut from brains from 30-day-old mice using a LEICA VT-1200S vibratome (Leica Inc, Buffalo Grove, IL, USA) with oxygenated dissection solution (mM: 2.5 KCl, 0.5 CaCl₂, 10 MgSO₄, 1.25 NaH₂PO₄, 24 NaHCO₃, 11 Glucose, 214 Sucrose). The brain slices were transferred to oxygenated ACSF (mM: 126 NaCl, 2.5 KCl, 2 CaCl₂, 2 MgCl₂, 26 NaHCO₃, 10 Glucose, pH 7.4) and recovered at room temperature for 1 hour before experiments.

Layer V pyramidal neurons were selected for recordings and were distinguished by their large soma size with apical dendrites, and layer VI pyramidal neurons were distinguished by their position right above white matter within coronal section brain slices. Whole-cell recordings of somatosensory cortex layer V/VI pyramidal neurons were obtained according to previously published methods²¹ (Supplementary method).

Thalamocortical oscillations

Horizontal thalamocortical brain slices (350–400 μm) were prepared²² from P30 mice and were put on top of a self-made nylon mesh interface. One unipolar tungsten electrode (MultiClamp 700B, current-clamp mode) was placed in the VBn of the thalamus to record spontaneous multi-unit activity, which was band-filtered between 100 Hz and 3 kHz. Experiments were performed at 31–32°C for 1–2 hours to collect sufficient spontaneous multi-unit activity. Data were analyzed with Clampfit (spike histogram and autocorrelation function) and Matlab to obtain auto-correlograms. Oscillation indices were calculated to compare spontaneous multi-unit activity in thalamocortical slices from wt and KI mice ²³.

Statistical analysis

Data were analyzed with GraphPad Prism 8 or the statistical package for the social sciences (SPSS) 23.0 software. Two-way ANOVA, independent-samples t-tests or a chi-square (χ²) test were used for the comparisons between genotypes when appropriate. Normality and equal variances were tested. For those non normally distributed data, a nonparametric Wilcoxon rank sum test was used. All analyses used an alpha level of 0.05 to determine statistical significance. Data were presented as Mean ± SEM.

Results

Gabrb3^+/N110D KI pups had epileptic spasms consistent with ISS.

We generated the Gabrb3^+/N110D KI mice by using conventional homologous recombination. The correctly targeted ES cell clones were identified by long-range PCR with one of the primers outside of the homologous arms to exclude the clones with random vector insertion (Fig. 1A). PCR fragments were digested with the Hind III (H3) enzyme to further confirm correctly targeted recombination (Fig. 1B). Chimeric mice were bred with Flpe transgenic mice to screen for germline transmission and to excise the PGK-Neo selection cassette. The leftover FRT site allowed convenient genotyping (Fig. 1C). The presence of the G358A mutation was further confirmed by PCR/sequencing. Semi-quantitative RT-PCR showed that the mutant allele did not affect Gabrb3 gene expression as demonstrated by glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as a loading control (Fig. 1D, E). The mutation was confirmed by sequencing (Fig. 1F).

Figure 1. Construction of the Gabrb3^+/N110D knockin mouse model of infantile spasms.

A. Scheme for targeted KI of the Gabrb3 locus. A vector was constructed to replace Gabrb3 exon 4 genomic DNA with the positive selection cassette (PGK-Neo) and an exon 4 containing the A328G mutation. B. A long-range PCR strategy was used to identify the correctly targeted clones, G5 and E10. 6 kb and 4.6 kb fragments were amplified from correctly targeted clones using primers A/B and C/D, respectively. The PCR fragments were digested with the Hind III restriction enzyme (to generate the predicted 4.2 and 1.8 kb fragments for the 5’ end, and 3.6 and 1 kb fragments for the 3’ end) to further confirm correct homologous recombination. C. The KI mice were genotyped. PCR amplified a 392 bp fragment from the wild-type β3 subunit allele and a 480 bp fragment from the β3 subunit KI allele due to the insertion of FRT sequences and other modifications. D. Semi-quantitative RT-PCR showed that the KI allele did not affect Gabrb3 gene expression, using glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as a loading control. E. Band intensities of the RT-PCR products were first normalized to GAPDH for quantification, and then to wild-type levels. F. Sequencing chromatogram of RT-PCR derived from total brain RNA showed the presence of the G328A mutation. G. The number of twitches or spasms were counted in the wild-type and the heterozygous KI pups. (N=11 for both wild-type and the KI heterozygous pups). H. The number of twitches or spasms was counted in untreated KI pups (n = 16) and vigabatrin-treated KI pups (n = 13). In G, H, A nonparametric Wilcoxon rank sum test and A (G) Two-way ANOVA with post hoc Turkey test for multiple comparison (H). In G, *P=0.0068 for P14; ****P=0.0001 for P15 and P16. In H, *p < 0.05 KI vs KI+vegabatrin.

We observed the gross development of the KI mice and determined whether KI mice exhibited an epilepsy phenotype by systematic home cage observation and video-EEG analyses. KI mice developed spasms as early as P14 in their home cage. Pilot study indicated that there was no obvious spasm observed in the KI pups prior to P14. The spasms were very brief and occurred in clusters (Supplemental Video), and the magnitude varied from dramatic truncal flexion and extension to subtle head drops. The wildtype littermates had minimal amounts oftwitches or spasms while the KI pups had more obvious and frequent twitches or spasms during all the observed days (Figure 1G and H, supplementary video). The KI pups also had reduced viability and reduced body weight compared with the wild-type littermates (Supplementary Fig 1). However, the Gabrb3^+/N110D mice had no difference in GABA_A receptor subunits α1, β3, and γ2 in the cortex, cerebellum, hippocampus, and thalamus (Supplemental Fig. 2).

Epileptic spasms in Gabrb3^+/N110D KI mice responded to vigabatrin.

Vigabatrin is an antiseizure medication that is commonly used to treat IS patients, and in about 56.9% of patients, it effectively controls seizures²⁴. It is an irreversible inhibitor of γ-aminobutyric acid transaminase (GABA-T), thus reducing metabolism of GABA and increasing GABA levels in GABAergic synapses²⁵. We tested vigabatrin in wt and KI mice to assess its efficacy to control epileptic spasms from P14 through P17 and found that the average number of spasms observed in KI mice on P15 and P16 were significantly lower in vigabatrin-treated KI mice. (Fig. 1H). Both the wildtype and the KI mice were slightly sedated during vigabatrin treatment.

Gabrb3^+/N110D KI mice had multiple seizure semiologies and EEG patterns consistent with IS.

We observed multiple types of spontaneous seizures in adult KI mice (Supplemental Video), but not in wild-type littermates (except for infrequent, brief, typical absence-like seizures, as is expected in C57BL6 mice)²⁶. Myoclonic seizures were observed most frequently (Fig. 2A), lasting approximately 1 s. They started with a sudden, brief muscle contraction accompanied by a single high amplitude sharp spike in the EEG and EMG channels (Fig. 2B). Both atypical and typical absence seizures were observed. Atypical absence seizures were associated with sudden behavioral arrest that was not always time-locked to the spike-wave discharges at seizure onset and with diffuse, slow, irregular continuous slow spike-wave discharges (3–5 Hz) that lasted up to several minutes (Fig. 2C). During the atypical absence seizures, brief, small movements were occasionally observed. Absence seizures were rare and consisted of time-locked sudden behavioral arrest and symmetrical, high-amplitude spike-wave discharges with a frequency between 4–8 Hz and a few seconds duration (Fig. 2D). Generalized tonic-clonic seizures (GTCSs) were observed relatively rarely (Fig. 2E). GTCSs are characterized by a sudden seizure onset with the loss of consciousness, a high-amplitude high-frequency tonic phase and followed a lower frequency clonic phase characterized by myoclonic and clonic jerking. In humans, tonic seizures are characterized by unilateral or bilateral sustained contraction of one or more muscle groups, lasting for a few seconds up to a minute²⁷.

Figure 2. Spontaneous seizures and increased PTZ sensitivity were observed in adult Gabrb3^+/N110D KI mice.

A. The average numbers of seizure that occurred in adult KI mice in a 24 h period was plotted. Myoclonic seizures were the dominant seizure type, while absence (atypical and typical) seizures were second most frequent. Generalized tonic-clonic and tonic seizures both occurred more rarely. B-F. Representative EEG traces for myoclonic (B), atypical absence (C), typical absence (D), generalized tonic-clonic (E) seizures were presented. In C, purple boxed segment was expanded. The traces in B, D, E and F were from channel 1. The supplementary video shows examples of each seizure type. PTZ was used to test seizure susceptibility. F. KI mice (11 of 11, gray line) had myoclonic seizures within 10 minutes of administration of 30 mg/kg PTZ, and 0 of 9 wt mice (black line) had myoclonic seizures (Mantel-Cox p < 0.0001). G. Generalized tonic-clonic seizures were observed in 8 of 11 KI mice (gray line) within 20 minutes of drug administration, and 0 of 9 wt mice (black line) had GTCS (Mantel-Cox p = 0.0015). H. Analysis of mean relative power of interictal EEGs showed significantly higher amplitude in multiple frequency ranges for KI (n = 6) compared to wt (n = 6) mice. Mean relative power and standard deviations were plotted for delta (0.5–3 Hz) p = 0.022, theta (4−-8 Hz) p = 0.022, alpha (8–12 Hz) p = 0.022. Beta (13–20 Hz) was not significantly different (Kolmogorov-Smirnov test).

Gabrb3^+/N110D KI mice had increased sensitivity to PTZ.

PTZ is a non-competitive GABA_A receptor antagonist that is widely used to provoke seizures as a method to assess CNS network excitability and seizure susceptibility²⁸. KI mice and wild-type littermates (between 1–3 months old) were injected intraperitoneally with PTZ, and the latencies to onset of seizures were recorded. At a dose of 30 mg/kg, KI mice had myoclonic seizures (11 of 11 within 10 min; Fig. 2F) and generalized tonic-clonic seizures (8 of 11 within 20 min; Fig. 2G). By contrast, the wildtype mice had no myoclonic or generalized tonic-clonic seizures when dosed with 30 mg/kg of PTZ. These results indicated that mutant β3(N110D) subunits affected the level of neuronal inhibition in the brain and substantially lowered PTZ seizure threshold.

Gabrb3^+/N110D KI mice had abnormal interictal EEGs.

EEG background consists of low-amplitude, mixed-frequency desynchronized rhythms. Analysis of interictal EEGs from KI and wild-type littermates showed significantly higher amplitude in multiple frequency ranges for KI mice (Fig. 2H). By contrast, the wildtype littermates only had minimal absence-like spike wave discharges at baseline level and higher seizure threshold upon PTZ seizure induction. This is consistent with our previous studies on other epilepsy mouse models¹⁹.

Gabrb3^+/N110D KI mice had hyperactivity, abnormal social interactions, anxiety, and cognitive impairment.

Children with IS often manifest long-term neurological impairment beyond epilepsy, including autism, cognitive impairment, and hyperactivity²⁹. We performed a suite of tests to compare behavioral features of KI and littermate mice.

ISS is linked to increased risk of developing autism³². One of the criteria for an autism diagnosis is impaired social skills. To assess sociability of KI mice, we used the three-chamber socialization test³³. KI mice spent substantially less time exploring novel mice (Fig. 3A), indicating decreased sociability. Given the opportunity to interact with a novel or familiar mouse, wild-type mice showed a strong preference for the novel mouse. In contrast, KI mice did not show a significant preference for the novel mouse or the familiar mouse (Fig. 3B). These results all indicate that the KI mice had reduced sociability.

Figure 3. Gabrb3^+/N110D KI mice exhibited social deficits, hyperactivity, and anxiety.

A, B. The three-chamber socialization test showed social deficits in KI mice. A. Sociability: KI mice spent significantly less time than wild-type mice investigating a novel mouse (Student’s t-test, p = 0.0059). As expected, wild-type mice spent significantly more time investigating the novel mouse (Student’s t-test, p < 0.001), and the KI mice showed no preference for the novel mouse or novel object. Two-way ANOVA showed a significant genotype effect. F(1,36)=31.61, 26.82; P<0.0001 for wt vs KI; F(1,36)=26.82, P<0.0001 for novel mouse vs novel object. B. Social novelty: KI mice showed no preference for a novel mouse or a familiar mouse. As anticipated, wild-type mice preferred to investigate the novel mouse instead of the familiar mouse (Student’s t-test, p < 0.001). The KI mice spent significantly less time investigating the novel mouse compared to the wild-type mice (Student’s t-test, p = 0.0003). Two-way ANOVA showed a significant genotype effect (p < 0.01). wt n = 9, KI n = 10. C, D. Hyperactivity was assessed using the number of open field test for 60 mins. C. KI mice reared more than wt mice (wt 375.4 ± 41.48, n = 10; KI 575.4 ± 72.05, n = 9; Student’s t-test, p = 0.0214). D. KI mice traveled significantly greater distances (wt 5355 ± 522.5 cm, n = 10; KI 11573 ± 1271 cm, n = 9; Student’s t-test, p < 0.0001). E, F. KI mice displayed anxiety-like behaviors in both the open field test and the elevated zero maze. E. In Open Field test, KI mice spent significantly less time in the center of the open field compared to wild-type littermates during the first 10-min in the test chamber (wt 330.2.8 ± 40.76 s, n = 10; KI 182.7 ± 30.9s n = 9, unpaired two-tailed Student’s t-test, p < 0.0115). F. In Elevated Zero Maze test, KI mice spent significantly less time in the open arms of the elevated zero maze (wt 123.4 ± 8.801 s, n = 10; KI 80.46 ± 7.481 s, n = 9, unpaired t-test, p < 0.0005).

To determine if KI mice were hyperactive and anxious, we used the open field test³⁴. We used a 60-min time frame to determine the activity exploration during the whole hour, but the data were binned at a 5-min interval based on the standard protocol at Vanderbilt neurobehavioral core. During the 60-minute test, the number of times KI mice reared was significantly greater than wild-type littermates (Fig. 3C), and KI mice travelled significantly farther than wild-type littermates (Fig. 3D).During open field test, mice tend to accommodate with reduced activity after initial exploration. We thus compared the time spent in the center during the first 10-min. KI mice spent less time in the center of the open field relative to wild-type littermates (Fig. 3E), suggesting increased anxiety in KI mice. In addition, KI mice spent substantially less time in the open arms of the elevated zero maze than wild-type littermates (Fig. 3F). Together, these data indicated KI mice had hyperactivity and anxiety.

To assess whether KI mice had cognitive deficits, their spatial learning and memory were assayed with the Barnes maze test³⁵. The latency to find the target hole was significantly greater in the first three training days for KI mice compared to their wild-type littermates, which indicated KI mice have a learning deficit. The delayed learning was reflected by a steeper learning curve in the KI mice (Fig. 4A), although the mice caught up eventually. In addition, KI mice committed a greater number of errors by entering non-target hole zones (Fig. 4B), suggesting slower acquisition in spatial learning for KI mice. During the probe trial, the KI mice spent much less time in the target hole area compared to wild-type littermates (Fig. 4C). KI mice committed increased number of entries into the non-target holes than wild-type littermates (Fig. 4D). This suggested that KI mice had impaired spatial memory.

Figure 4. Gabrb3^+/N110D KI mice had impaired spatial learning and memory.

Mice were tested with Barnes maze for learning and memory. Mice were pretrained on the first day of the learning trial and trained for 5 days with one trial per day before the probe trial. A, B. Five days of learning trials demonstrated a profound spatial learning deficit in KI mice. A. The time it took each animal (latency) to find the target hole and B. the number of non-target hole zones entered (errors) were quantified for each day. Two-way ANOVA for repeated measures with Bonferroni post-test, In A, F(1,14)=7.506, P<0.0160 for wt vs KI; F(4,56)=18.62, 1,098; P<0.0001 for different days. In B, F(1,70)=75.88, P<0.0001 for wt vs KI; F(4,70)=13.06, 0.5770; P<0.0001 for different days.(.wt n = 8, KI n = 8). C, D. Five-minute probe trial for spatial memory showed impaired memory in KI mice. The target hole was now covered and appeared identical to the other 11 holes. C. Time spent near the target hole (KI 29.46 ± 2.23 s, n = 8; wt 40.06 ± 2.45 s, n = 8) and D. the number of non-target hole zones entered (errors) were counted (wt 51.94 ± 2.432, n = 8; KI 78.65 ± 7.33, n = 8). Unpaired two-tailed Student’s t-test, *p < 0.05, **p < 0.01.

Gabrb3^+/N110D KI mice had reduced mIPSC amplitudes in somatosensory cortical layers V/VI pyramidal neurons.

We investigated whether mutant β3(N110D) subunits affected the function of inhibitory synapses in somatosensory cortical layer V/VI pyramidal neurons, which are known to participate in the thalamocortical feedback loop³⁶. Whole-cell patch-clamp recordings in acute coronal slices from KI mice showed decreased GABA_A receptor-mediated mIPSC amplitude (Fig. 5A to E), unchanged mIPSC frequency and decay time (data not shown). The reduced current amplitudes of mutant GABA_A receptors containing β3(N110D) subunits in somatosensory cortical layer V/VI pyramidal neurons from KI mice could decrease interneuron phasic GABAergic inhibition and lower the seizure threshold by promoting neuronal hyperexcitability.

Figure 5: Gabrb3^+/N110D KI mice had reduced amplitude of mIPSCs.

Representative recordings from A. wild-type and B. KI somatosensory cortical layer V/VI pyramidal neurons. C. Normalized mIPSC events from wild-type (gray) and KI (black) showed the frequency of mIPSCs in each amplitude bin. D. Normalized cumulative distribution was plotted for wild-type (gray) and KI (black) mIPSCs. E. Amplitude of mIPSCs from the KI mice was significantly decreased (wt 40.60 ± 3.27 pA, n = 9 neurons, 6 mice; KI 28.16 ± 3.18 pA, n = 9 neurons, 5 mice; Student’s t-test, p = 0.014).

Gabrb3^+/N110D KI mice had increased spontaneous thalamocortical network oscillations.

The balance in the ratio of excitation to inhibition critically regulates neural network activity and brain function. Impaired inhibitory neurotransmission would disrupt the balance of excitation and inhibition and lead to disinhibition and hyperexcitability in the brain, resulting in seizures and neurobehavioral disorders³⁷. Thalamocortical network oscillatory activity was recorded in VBn in a horizontal slice that retained thalamocortical circuitry. KI mice showed prolonged and high amplitude spontaneous network oscillations compared to the very short and low amplitude spontaneous network bursts in slices from wild-type littermates (Fig. 6A to C), which indicated that mutant β3(N110D) subunits caused excessive synchronization and hyperexcitability in the thalamocortical loop, producing generalized seizures.

Figure 6: Gabrb3^+/N110D KI mice had prolonged network oscillations.

Representative recordings from A. wild-type and B. KI horizontal brain slices showing spontaneous thalamocortical brief burst and prolonged oscillations. One short burst from a wild-type littermate mouse and one oscillation from a KI mouse were expanded to show the multiple spikes in the burst or oscillation. C. Average duration of spontaneous bursts from wild-type littermate mice (0.24 ± 0.16 s, 6 slices, 4 mice) and oscillations from P42 KI mice (2.55 ± 0.95 s, 6 slices, 5 mice) were plotted (Student’s t-test, p = 0.039).

Discussion

ISS is an epileptic encephalopathy that can be genetic or acquired with etiologies including structural, metabolic, and other unknown factors and is associated with seizures and multiple neuropsychiatric comorbidities and related behavioral traits⁴. Substantial progress has been made in understanding the genetic basis of IS. The Epi4K consortium studied trios by whole exome sequencing identified a de novo mutation, A328G, p.N110D, in GABRB3 in a patient with IS. We generated a mouse model of IS, the heterozygous Gabrb3^+/N110D KI mouse, to permit study of the effects of mutant β3(N110D) subunits in vivo. We demonstrated that the KI mouse successfully recapitulated seizures and behavioral phenotypes in patients with IS. Although the N110D mutation was only identified in one patient, the pathophysiology identified in the mouse could provide critical insights into other GABRB3 mutations and IS.

Gabrb3^+/N110D KI mice have epileptic spasms during infancy.

Home cage observations of KI and wild-type littermates were performed on P10 to P20 mice. KI pups displayed clusters of twitches or spasms only from P14 to P17. These flexion-extension spasms (Supplemental video), typically consisting of a head drop, flexed tail and subtle limb extension or flexion with brief behavioral arrest, were similar to spasms reported in models of infantile spasms³⁸. Rarely were the spasms severe enough to cause the animal to lose its posture, and while twitches were observed in wild-type and KI mice, the wild-type mice did not have any flexion-extension spasms during the observation period. This suggests that the observed twitches or spasms are specific due to the mutational effect of N110D and during a specific time window as the similar epileptic spasms were not observed in the wildtype nor adult N110D KI mice

GABA_A receptors play an essential role in pruning synapses and establishing neuronal circuitry, and impaired inhibition can negatively affect the activity-dependent construction of neural circuits³⁹. The β3 subunits are the predominant β subunits in embryonic and neonatal development in most parts of the brain. Expression of β3 subunits remains high during adulthood in regions associated with seizure generation, including the cortex, thalamus, hippocampus, septum, and basal forebrain⁴⁰. In addition, the β3 subunit knockout mice have been shown to develop seizures¹³. Therefore, it is reasonable to speculate that mutant β3(N110D) subunits, impairing the function of GABA_A receptors, may hamper the early formation of cortical, thalamic, and hippocampal circuitry, resulting in a hyperexcitable state that precipitates development of epileptic spasms in KI pups and a variety of seizure types later in development.

Adult Gabrb3^+/N110D KI mice exhibit multiple seizure types.

ISS is a disease of childhood, and while the characteristic spasms usually remit by 3 to 4 years of age, many patients develop other epilepsy syndromes such as Lennox-Gastaut syndrome. Patients with Lennox-Gastaut syndrome have multiple seizure types, the predominant ones being tonic, atonic, and atypical absence seizures. Myoclonic jerks and generalized tonic-clonic seizures have also been reported in Lennox-Gastaut syndrome patients. To assess the effects of β3(N110D) in adult mice, we conducted video-EEG monitoring to characterize seizure semiologies. Myoclonic seizures were observed most frequently, and atypical absence, absence, generalized tonic-clonic and tonic seizures were also observed. In addition to recapitulating the spasms seen in infancy, the KI mouse is a promising model to study the evolution from ISS to other seizure types.

Gabrb3^+/N110D KI mice display features of ISS.

GABA_A receptor β3 subunits are widely expressed throughout the brain during embryonic and early postnatal stages⁴⁰ and are involved in neuronal growth and differentiation⁴¹. Expression decreases throughout development, but remains high in the cortex, reticular nucleus of the thalamus, hippocampus, and hypothalamus⁴². Sensory processing requires many brain regions and neural circuits, but nearly all sensory information is routed through the thalamus, particularly the VBn⁴³, which filters and directs information to the cortex. β3 subunits are a major component of GABA_A receptors in the reticular nucleus, and thus, mutant β3 subunits could impact the function of the VBn to filter inputs to the cortex, thereby affecting somatosensory processing and function⁴⁴. Impaired somatosensory function is common in autism, which suggests that mutant β3 subunits could cause seizures and behaviors associated with autism. Patients with IS usually suffer from a wide spectrum of behavioral phenotypes that negatively impact their social and physical health. Many IS patients have significant learning and memory deficits, social impairment, hyperactivity, and anxiety. An animal model of IS that replicates both seizures and behavioral phenotypes could be a powerful tool for use in testing new therapies. With that aim, the KI mice were tested for IS-associated behaviors.

Aberrant interictal activity is considered to be strongly associated with cognitive difficulties and neurodevelopmental regression⁴⁵, which is supported by the fact that patients with well-controlled seizures by AEDs often show reduced interictal activity and a regain of cognitive functions⁴⁶. The interictal EEG activity of KI mice were significantly different than wild-type mice, it came as no surprise that the behavioral assessments showed significant differences. In addition to hyperactivity and increased anxiety in KI mice, we found that the KI mice had significant impairment in socialization.

Taken together, these data suggest that the KI mice have many of the behavioral traits found in patients with IS, which opens the door to further study of the effects of pharmaceutical therapies on behavioral correlates, not just seizures.

Gabrb3^+/N110D KI mice had impaired cortical synaptic inhibition and increased spontaneous oscillations.

Our previous in vitro studies indicated that mutant β3(N110D) subunits slowed the activation and accelerated the deactivation of currents from β3(N110D) subunit-containing GABA_A receptors, which could lead to altered IPSCs and reduced currents and inhibition. We showed that KI mice had significantly increased sensitivity to PTZ, suggesting that mutant β3(N110D) subunits negatively affected the level of neuronal inhibition in the brain. As previously discussed, GABRB3 is widely expressed throughout the brain during early development⁴⁰ and are involved in neuronal growth and differentiation⁴¹. Homozygous and heterozygous GABRB3 KO mice have increased seizure susceptibility and EEG abnormalities⁴⁷. As major components of GABA_A receptors in the thalamic reticular nucleus and the somatosensory cortex, β3 subunits play an important role in the thalamocortical circuit. Indeed, analysis of KI mice revealed two major mechanisms that have been shown to lead to epilepsy, dysfunction of GABAergic-signaling and excessive network oscillations. In KI mice, the amplitudes of mIPSCs recorded from somatosensory cortical layer V/VI pyramidal cells in coronal brain slices were reduced about 30% with no change in frequency or decay, suggesting alteration of postsynaptic, but not presynaptic, function ⁴⁸. Consistent with the notion that excessive thalamocortical synchronization and hyperexcitability occur during seizures, we found that KI mice displayed significantly prolonged, high amplitude spontaneous thalamocortical network oscillatory activity recorded in the VBn in horizontal slices that retained some of the thalamocortical circuitry.

Many studies have examined the effects of impaired GABAergic inhibition in the thalamocortical circuitry on seizure generation⁴⁹. The decreased amplitude of cortical mIPSCs and the increased thalamocortical oscillations suggests that this circuitry is impacted in the KI mice and that the β3(N110D) subunits affect the function of the brain by reducing inter-neuronal inhibition. Impaired thalamocortical circuitry has been implicated in the propagation of absence seizures, and increased cortical activity is involved in generalized tonic clonic seizures. The neural networks involved in tonic and myoclonic seizures are less well understood. Our model of IS offers a new opportunity to study how a mutation in the widely expressed GABA_A receptor β3 subunit can cause a variety of seizure types.

Gabrb3^+/N110D KI mice will aid in the development of therapeutic strategies for infantile spasms.

The findings presented here identify in vivo functional effects of the mutant β3(N110D) subunit, including decreased mIPSC amplitude in cortical layer V/VI pyramidal neurons and increased spontaneous oscillations recorded from the thalamus. The multitude of seizure phenotypes observed in the adult KI mice and the behavioral traits suggest that the mutant β3(N110D) subunit affects the function of multiple neural circuits and brain regions, including the somatosensory thalamocortical circuit. Future studies will aim to characterize additional functional effects. The Gabrb3^+/N110D KI mice are a useful tool in continued study of IS pathogenesis and pathophysiology. This model will hopefully aid in the development of effective therapies by addressing the epilepsy phenotype and the concomitant encephalopathies.

Supplementary Material

supinfo2. Supplemental Figure 1: Gabrb3^+/N110D KI mice had reduced viability and reduced body weight.

A. KI mice had reduced viability. All offspring from wt x KI het breeding were genotyped at weaning. Out of 432 genotyped weanlings, 243 were wild type and 189 were KI, suggesting that KI mice had reduced viability. B. The body weight of KI and their littermate mice at weaning were plotted and normalized to the wildtype littermates (n = 22 for wt and 18 for KI).

supinfo3. Supplemental Figure 2: Expression and trafficking of receptors containing mutant β3 subunits and normal partnering subunits in Gabrb3^+/N110D KI mice were not affected by the N110D substitution.

The N110D amino acid substitution did not alter GABA_A receptor subunit expression, surface trafficking, or synaptosomal distribution. A. Whole cell lysates from different brain regions [cortex (co), cerebellum (ce), hippocampus (hi) and thalamus (Th)] were collected from KI and WT littermates and subjected to SDS-PAGE and immunoblotted with anti-α1, -β3, and -γ2 subunit antibodies. GAPDH served as a loading control. B. Band intensities of α1, β3, and γ2 subunits were normalized to the GAPDH signal, then KI was compared to WT (n = 3 pairs of mice). C. Brain plasma membrane proteins isolated from mouse brain were analyzed by SDS-PAGE and immunoblotted with anti-α1, -β3, and -γ2 subunit antibodies. Na+/K+ ATPase served as a loading control. D. Band intensities of α1, β3, and γ2 subunits were normalized to the ATPase signal, then did relative comparison (n = 3 pairs of mice). E. Brain synaptosomal proteins were analyzed by SDS-PAGE and immunoblotted with anti-α1, -β3, and -γ2 subunit and anti-gephyrin antibodies. Na+/K+ ATPase served as a loading control. F. Band intensities of α1, β3, and γ2 subunits and gephyrin were normalized to the ATPase signal, then presented as a relative comparison (n = 3 pairs of mice). Data were presented as mean ± SEM. Student’s t-tests were used in each panel.

Key points.

• Gabrb3^+/N110D mice have epileptic spasms from P14-P17 and display multiple seizure types in adulthood.

• Gabrb3^+/N110D mice have increased spontaneous oscillations recorded from the VBn and decreased inhibition in the cortex, indicating impaired thalamocortical circuitry.

• Gabrb3^+/N110D mice have deficits in social interaction, cognitive impairment, increased anxiety, and hyperactivity.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.