Adult loss of Cacna1a in mice recapitulates childhood absence epilepsy by distinct thalamic bursting mechanisms

Inborn errors of CACNA1A-encoded P/Q-type calcium channels give rise to childhood absence epilepsy, ataxia and dystonia. Miao et al. show that adult-onset ablation of the P/Q channel alpha subunit in mice replicates the inborn loss-of-function phenotype, and that distinct thalamic bursting mechanisms can generate similar patterns of cortical hypersynchrony.

Abstract

Inborn errors of CACNA1A-encoded P/Q-type calcium channels impair synaptic transmission, producing early and lifelong neurological deficits, including childhood absence epilepsy, ataxia and dystonia. Whether these impairments owe their pathologies to defective channel function during the critical period for thalamic network stabilization in immature brain remains unclear. Here we show that mice with tamoxifen-induced adult-onset ablation of P/Q channel alpha subunit (iKO^p/q) display identical patterns of dysfunction, replicating the inborn loss-of-function phenotypes and, therefore demonstrate that these neurological defects do not rely upon developmental abnormality. Unexpectedly, unlike the inborn model, the adult-onset pattern of excitability changes believed to be pathogenic within the thalamic network is non-canonical. Specifically, adult ablation of P/Q channels does not promote Cacna1g-mediated burst firing or T-type calcium current (I_T) in the thalamocortical relay neurons; however, burst firing in thalamocortical relay neurons remains essential as iKO^p/q mice generated on a Cacna1g deleted background show substantially diminished seizure generation. Moreover, in thalamic reticular nucleus neurons, burst firing is impaired accompanied by attenuated I_T. Interestingly, inborn deletion of thalamic reticular nucleus-enriched, human childhood absence epilepsy-linked gene Cacna1h in iKO^p/q mice reduces thalamic reticular nucleus burst firing and promotes rather than reduces seizure, indicating an epileptogenic role for loss-of-function Cacna1h gene variants reported in human childhood absence epilepsy cases. Together, our results demonstrate that P/Q channels remain critical for maintaining normal thalamocortical oscillations and motor control in the adult brain, and suggest that the developmental plasticity of membrane currents regulating pathological rhythmicity is both degenerate and age-dependent.

Introduction

Synaptic input during early postnatal development is essential to refine adult thalamocortical (Toulmin et al., 2015) and cerebellar function (Kano et al., 2013), and depends upon the fidelity of the release process (Kakizawa et al., 2000) as well as low threshold T-type calcium channel activity (Yoshimura et al., 2008). Disruption of normal synaptic transmission in sensory and motor control pathways during this critical period leads to long-lasting deficits (Kano et al., 2013; Hubener and Bonhoeffer, 2014), but the effects of aberrant transmission initiated after this early period are less understood.

Childhood absence epilepsy (CAE) is a major form of early onset epilepsy that involves abnormal synchronization of thalamocortical networks. It is the most common type of paediatric epilepsy and is clinically diagnosed by a sudden, brief impairment of consciousness, behavioural arrest, and spike-and-wave discharges (SWDs) on EEG (Glauser et al., 2010). The mutant mouse ‘tottering’ was the first model to reveal a monogenic cause for this disorder (Noebels and Sidman, 1979), which is due to reduced high-voltage activated calcium current (Ca_v2.1) through single P/Q-type channels (Fletcher et al., 1996; Wakamori et al., 1998). Loss of function of the orthologous human gene, CACNA1A, causes CAE with cognitive deficits and ataxia as well (Rajakulendran et al., 2010; Damaj et al., 2015). Mice with null mutation of Cacna1a develop a similar but rapidly progressive neurological deficit including absence epilepsy, ataxia and dystonia and die 3–4 weeks after birth (Jun et al., 1999). However, little is known about whether the distinctive spike-wave seizure and ataxia emerge from developmental defects that would require early therapeutic intervention, or whether they can also arise in a mature network. If the latter is true, it is important to also determine whether mechanisms underlying the spike-wave seizure generation depend upon the same pattern of intrinsic excitability alterations that determine rebound burst firing within the thalamocortical circuit. As absence epilepsy is genetically and physiologically heterogeneous (Maheshwari and Noebels, 2014; Tenney et al., 2014), further insight into the range of pathogenic rebound burst firing properties that sustain spike-wave seizures within this network can provide new insight into human CAE and help develop improved therapies for the disorder.

Here, we examined the consequences of ablating P/Q-type calcium channels in adult mice by tamoxifen-induced genomic recombination. Adult-onset P/Q channel induced knockout mice (iKO^p/q) showed EEG activity indistinguishable from their control loxP littermates before tamoxifen induction, but rapidly developed the characteristic pattern of ataxia, dystonia, and absence epilepsy within weeks following deletion of P/Q channels upon tamoxifen induction. We found that, in sharp contrast to the inborn models, burst spiking and Cacna1g-mediated T-type calcium current (I_T) in thalamocortical relay neurons were not altered in iKO^p/q mice. Interestingly, similar to inborn models, loss of Cacna1g in iKO^p/q mice substantially suppressed SWD generation. Moreover, thalamic reticular nucleus (TRN) burst firing in iKO^p/q mice was impaired and accompanied by decreased I_T, which is reversed from inborn models. Deletion of the TRN-enriched and human CAE-linked gene Cacna1h in iKO^p/q mice slightly reduced TRN burst firing and promoted rather than blocked seizures in these mice. Together, these results demonstrate that P/Q-type calcium channels, which share control of calcium-mediated exocytosis with N- and R-type calcium channels at most central synapses, retain an essential role in normal adult thalamocortical network synchrony, and indicate that inherited childhood absence and ataxia syndromes do not necessarily arise from an irreversible structural perturbation of synaptogenesis during early brain development. The alternative thalamic patterns of pathogenic T-type currents, believed to be a major target of clinical antiepileptic therapy for this seizure type, suggest a possible basis for the treatment failure reported in over 50% of children with this disorder (Glauser et al., 2013).

Materials and methods

Animals

All procedures to maintain and study these mice were conducted with approval of the Institutional Animal Care and Use Committee at Baylor College of Medicine. The Cacna1a^{Citrine(C57BL/6 129/SvJae)} (Cacna1a^flox/flox for brevity) strain was described in Mark et al. (2011). These mice are viable, fertile and display normal P/Q channel function (Mark et al., 2011). The CAGGcre-ER^TM [B6.Cg-Tg(CAG-cre/Esr1*)5Amc/J; Stock No: 004682] (CAG-ER-Cre for brevity) strain was obtained from the Jackson Laboratory. CAGGcre-ER^TM mice express a fusion protein consisting of Cre recombinase and a modified ligand-binding domain of the oestrogen receptor (ER) under the control of the ubiquitous promoter CAG (CMV enhancer, chicken β-actin promoter, rabbit β-globin polyA). The fusion protein only moves into the nucleus upon exposure to tamoxifen, which induces deletion of floxed sequences via Cre-mediated recombination. To prevent Cre leakage, male but not female mice carrying the Cre were used as breeders. In Cacna1a^flox/flox mice, a floxed green fluorescent protein derivative Citrine is in the first exon of Cacna1a, and exon 1 deletion generates a frame shift to disrupt Cacna1a expression. To induce Cre expression, adult mice (>6 week) were injected with a single dose of tamoxifen [dissolved in corn oil, 75 mg/kg body weight, intraperitoneally (i.p.)] per day for five consecutive days. Heterozygous Cre and homozygous loxP mice were used in the experimental group, while homozygous loxP littermates were used as controls unless otherwise stated. Rare Cre leakage was observed in CAG-ER-Cre; Cacna1a^flox/flox mice older than 4 months indicated by mild ataxia without tamoxifen induction. These mice were excluded in the study. Cacna1g knockout mice were generated by breeding heterozygous female mice carrying the Sox2-Cre gene [B6.Cg-Edil3Tg^{(Sox2-cre)1Amc}/J; Stock No: 008454] to Cacna1g-flox mice [B6(129S4)-Cacna1g^tm1Stl/J; Stock No: 021932] (Anderson et al., 2005). Both mouse lines were obtained from the Jackson Laboratory. In female Sox2-cre mice, Cre recombinase is active in the female germline. Offspring arising from a hemizygous transgenic female exhibit Cre recombinase activity, regardless of genotype. This maternal inheritance effect enables a rapid and efficient generation of null offspring with loxP animals (Hayashi et al., 2002). Cacna1h knockout (B6;129-Cacna1^htm1Kcam/J; Stock No: 013770) mice were obtained from the Jackson Laboratory. Cacna1g^−/− and Cacna1h^−/− mice were crossed with Cacna1A^floxflox;CAG-Cre-ER mice to generate Cacna1A^floxflox;CAG-Cre-ER/Cacna1g^−/− and Cacna1A^floxflox;CAG-Cre-ER/Cacna1h^−/−, respectively. We used the same tamoxifen induction protocol for these mice. Both male and female mice were studied in all experiments. The number of animals used in each experiment is stated in figure legends.

Polymerase chain reaction

The genetic background of the mice was determined by PCR of genomic DNA from tail or ear biopsy. The following primer pairs were used: (i) Cacna1a^Citrine forward 5′-TATATCATGGCCGACAAGCA-3′, reverse 5′-TTCGGTCTTCACAAGGAACC-3′; wild-type (WT) 5′-forward GGGGTCTGACTTCTGATGGA-3′, reverse 5′-AAGTTGCACACAGGGCTTCT-3′; (ii) Cre recombinase forward 5′-GCGGTCTGGCAGTAAAAACTATC-3′, 5′-GTGAAACAGCATTGCTGTCACTT-3′; (iii) Cacna1g^tm1Stl forward 5′-TTCTGCAGCTCTTTCAATGC-3′, reverse 5′-TGTTCCTACCCAAGGTCTGG-3′; and (iv) Cacna1h^tm1Kcam mutant forward 5′-GCTAAAGCGCATGCTCCAGACTG-3′, wild-type forward 5′-ATTCAAGGGCTTCCACAGGGTA-3′, common reverse 5′-CATCTCAGGGCCTCTGGACCAC-3′.

EEG recordings

Mice were anaesthetized with isoflurane (1.5–3% in oxygen, Matrox Vip300 ventilator), and silver wire electrodes (0.005″ diameter) soldered to a connector were surgically implanted bilaterally into the subdural space over frontal and parietal cortex. Mice were allowed to recover for at least 7 days before recording. Simultaneous video-EEG and behavioural monitoring (Labchart 8.0, ADI Systems) was performed in adult (>6 weeks) mice of either sex. EEG was recorded for 4 h while mice moved freely in the test cage on serial days post-first injection (DP1i) (DP1i7, DP1i14, DP1i21 and DP1i28–29) unless otherwise noted. To quantify epileptic activity, the number and duration of SWDs (>0.5-s duration) were visually analysed during the first 30 min of EEG recorded in each mouse. A single dose of ethosuximide (200 mg/kg, i.p.) was injected into iKO^p/q mice at DP1i28–29 to test for blockade by this drug. All EEG signals were amplified by a g®.BSAMP biosignal amplifier (Austria), digitized by PowerLab with a 0.5 Hz high-pass and 50 Hz low-pass filter (ADInstruments) and acquired via Labchart 8.0 (ADInstruments). SWDs were identified manually with amplitude larger than twice the root mean square of baseline EEG signal. Power spectral density (PSD) analyses were performed in the Labchart with a FFT (fast Fourier transform) size of 128k (unless otherwise noted) and Hann (cosine-bell) window 50% overlapped.

Rotarod test

After habituation (30 min) in the test room, motor coordination was measured using an accelerating rotarod apparatus (TSE Systems). Mice were tested daily four times (before tamoxifen injection, and at DP1i2, DP1i4 and DP1i7), three trials each, with a rest interval of 60 min between trials. Each trial lasted for a maximum of 10 min, and the rod accelerated from 4 to 40 rpm in the first 5 min. The time that it took for each mouse to fall from the rod (latency to fall) was recorded.

Brain slice preparation

Acute sagittal slices were prepared as previously described (Miao et al., 2016). In brief, the brain was dissected from mice under tribromoethanol (Avertin^TM) anaesthesia (250 mg/kg body weight, i.p.), and placed in ice-cold artificial CSF containing (in mM) 125 NaCl, 3 KCl, 2 CaCl₂, 2 MgSO₄, 1.25 NaH₂PO₄, 1.3 Na⁺-ascorbate, 0.6 Na⁺-pyruvate, 26 NaHCO₃, and 11 glucose (pH 7.4). Sagittal slices (300-µm thick) were cut with a vibratome (VT-1200S, Leica) and incubated in oxygenated protective recovery NMDG (N-methyl-d-glucamine) artificial CSF at 34°C for 10–15 min, and then transferred to normal artificial CSF at room temperature (20–25°C) for >45 min before recordings. NMDG artificial CSF contained (in mM) 110 NMDG-Cl, 3 KCl, 0.5 CaCl₂, 6 MgSO₄, 1.25 NaH₂PO₄, 1.3 Na⁺-ascorbate 3, 0.6 Na⁺-pyruvate, 26 NaHCO₃, and 11 d-glucose (pH 7.4). All recordings were performed in artificial CSF at 32 ± 1°C. Artificial CSF and NMDG artificial CSF were bubbled with 95% O₂/5% CO₂.

Whole-cell recordings

Whole-cell recordings of neurons in the TRN, laterodorsal nucleus (LD), and ventroposterior medial nucleus (VPM) were made with a MultiClamp 700B amplifier (Molecular Devices) as previously described (Miao et al., 2016). LD, VPM and TRN were identified by reference to the Allen Mouse Brain Atlas through an Olympus microscope (5× and 40× objectives, BX-51WIF, Olympus) equipped with infrared video camera and differential interference contrast optics. Signals were filtered at 3 kHz using a Bessel filter and digitized at 10–20 kHz using Digidata 1332A (Molecular Devices). Liquid junction potential was not corrected.

Low-voltage activated transient calcium current (I_T)

I _T was recorded in the presence of tetrodotoxin (TTX, 1 µM, Tocris). A Cs-based internal solution containing (in mM) 125 Cs-gluconate, 5 TEA-Cl, 2 CsCl, 1 EGTA, 10 HEPES, 4 Mg₂ATP, 0.3 Na₂GTP, 10 phosphocreatine (pH 7.2–7.4, 270–310 mOsm) was used (Sun et al., 2010). T current in TRN neurons was recorded in the presence of 10 mM TEA-Cl and 0.5 mM 4-AP in artificial CSF to make cells more electrically compact since TRN neurons have long dendritic processes where T-type calcium channels reside (Joksovic et al., 2005). The perfusion solution contained (in mM) 118 NaCl, 3 KCl, 2 CaCl₂, 2 MgSO₄, 1.25 NaH₂PO₄, 1.3 Na⁺-ascorbate, 0.6 Na⁺-pyruvate, 26 NaHCO₃, 10 TEA-Cl and 0.5 4-AP and 11 glucose (pH 7.4).

Rebound spiking and membrane excitability

To assess rebound bursting/spiking and firing-input relationships (F-I curve) in thalamocortical neurons, pipettes were filled with an internal solution containing (in mM) 130 K-gluconate, 20 KCl, 10 HEPES, 0.2 EGTA, 4 Mg₂ATP, 0.3 Na₂GTP, and 10 Na₂-phosphocreatine (pH 7.3, 290–310 mOsm). For TRN neurons, the internal solution contained (in mM) 120 K-gluconate, 11 KCl, 1 MgCl₂, 1 CaCl₂, 10 HEPES, and 1 EGTA, (pH 7.4, 270–290 mOsm) (Clemente-Perez et al., 2017).

Spontaneous inhibitory and excitatory postsynaptic currents

A high chloride internal solution containing (in mM) 94 K-gluconate, 60 KCl, 10 HEPES, 0.2 EGTA, 4 Mg₂ATP, 0.3 Na₂GTP, and 10 Na₂-phosphocreatine (pH 7.3, 290–310 mOsm) was used. To record spontaneous excitatory postsynaptic currents (sEPSCs), neurons were voltage-clamped at −70 mV, with a pipette (resistance of 3–4 MΩ) in the presence of picrotoxin (100 µM) to block GABAergic transmission. To record spontaneous inhibitory postsynaptic currents (sIPSCs), neurons were held at −70 mV in the presence of DNQX (20 µM) and d-AP5 (50 µM) to block AMPARs and NMDARs respectively. Recordings had a typical access resistance of <20 MΩ, and those changed >30% during recordings were excluded from analyses. Access resistance did not significantly differ between experimental and control groups. sEPSCs and sIPSCs were analysed with MiniAnalysis software (Synaptosoft, Fort Lee, NJ) with a detection threshold of 10 pA followed by manual examination.

Immunohistochemistry and image analysis

Anaesthetized mice [tribromoethanol (Avertin^TM), 250 mg/kg body weight, i.p.] were perfused with phosphate-buffered saline (PBS, pH 7.4), followed by 4% paraformaldehyde (PFA, w/v) in PBS. Brains were removed and post-fixed in 4% PFA at 4°C for 2 h, then cryoprotected in 30% sucrose (w/v) in PBS for at least 1 day at 4°C. Coronal or sagittal sections (40 µm) were cut on a freezing microtome (Leica) from the fixed brain embedded in optimal cutting temperature (OCT) medium. Slides were thawed at room temperature for 1 h, washed in PBS three times, blocked in a solution containing 10% bovine serum albumin and 0.3% Triton™ X-100 in PBS for 1 h and then overnight with the primary antibody in blocking solution. Slides were washed in blocking solution three times, incubated with the secondary antibody for 2 h, then washed once in blocking solution and three times in PBS before mounting with Antifade mounting medium with DAPI (VECTASHIELD®, Vector Laboratories). The primary antibodies were: rabbit anti-GFP (1:500, Abcam, Cat# ab6556), rabbit anti-Calcium Channel Ca_v2.1 (α_1A subunit) (1:100, Sigma, Cat# C1353), rabbit anti-Calcium Channel Ca_v3.1 (α_1G Subunit) (1:300, Alomone, Cat# ACC-021), mouse anti-Tyrosine Hydroxylase (TH, 1:500, Millipore MAB318), anti-Parvalbumin (1:1000, Millipore MAB1572). The secondary antibodies were: Goat anti-Mouse Alexa Fluor® 594 (1:1000 except for 1:500 for TH staining, Thermo Fisher Scientific, Cat# A-21125), Goat anti-Rabbit Alexa Fluor® 488 (1:1000, Thermo Fisher Scientific, Cat# A-11034). Images were acquired on a Nikon C2 laser-scanning confocal microscope with a Plan Apo 20× DIC M (NA = 0.75) or Plan Fluor 10× Ph1 DL objective (NA = 1.4), at 1024 × 1024 pixels. We immunostained GFP to examine the expression of P/Q channels since the α1A subunit is tagged with Citrine (a GFP variant) in the Cacna1a floxed mice (Mark et al., 2011). Similar CACNA1A loss in iKO^p/q mice was observed with a specific anti-Calcium Channel Ca_v2.1 (α_1A subunit) antibody (Sigma, Cat# C1353) (data not shown). Immunofluorescence images were captured using the same settings for laser power and detector gain for the same brain region in experimental and corresponding control littermate groups.

Statistics

Data are presented as mean ± standard error of the mean (SEM) with n indicating the number of neurons for whole-cell patch-clamp recordings, and the number of animals for immunohistochemistry and EEG datasets. No statistics were used to predetermine sample size. Statistical methods are indicated when used. All analyses were performed using Labchart (8.0), MATLAB (MathWorks), NIS Elements (4.0) and Origin. No method of randomization was used in any of the experiments. Experimenters were not blind to group allocation in behavioural experiments. All animals that finished the entire behavioural testing were included in analysis. All statistical tests were two-sided. For both paired and unpaired tests, we ensured that the variances of the data were similar between the compared groups. For Student t-tests, we verified that the data were normally distributed using Jarque-Bera test for normality.

Data availability

Data supporting the findings of this study are available within the article and its Supplementary material. All supporting data in this study are available from the corresponding author on request.

Results

Adult ablation of Cacna1a leads to absence seizures and ataxia in mice

To assess the impact of P/Q channel dysfunction beginning in adulthood, we created conditional mice by crossing Cacna1a^flox/flox mice with tamoxifen inducible cre line CAGGcre-ER^TM (CAG-Cre-ER here after) mice (Fig. 1A) . The resulting Cacna1a^flox/flox;CAG-Cre-ER offspring are viable, fertile, and indistinguishable from their Cacna1a^flox/flox littermates (Supplementary Video 1). Cacna1a is expressed normally in these mice until tamoxifen injection, when a metabolite binds to ER, activating Cre and deleting the first exon. We injected tamoxifen [75 mg/kg, i.p. (intraperitoneal injection)] in Cacna1a^flox/flox;CAG-Cre-ER (iKO^p/q) mice and their control Cacna1a^flox/flox (loxP) littermates at 6–10 weeks of age and probed CACNA1A expression 3 weeks later (Fig. 1B). At this time point, CACNA1A protein was greatly reduced throughout the iKO^p/q brain, including neocortex, hippocampus, thalamus, and cerebellum compared to control (Fig. 1C) and uninjected Cacna1a^flox/flox;CAG-Cre-ER mice (Supplementary Fig. 1). We then assessed how adult loss of the Cacna1a gene alters brain rhythms by comparing video-EEG recordings in freely moving animals before, and at DP1i7, DP1i14, DP1i21 and DP1i28. iKO^p/q mice showed EEG activity indistinguishable from their control loxP littermates before tamoxifen induction, but progressively developed SWDs with concomitant behavioural arrest within 2 weeks following tamoxifen induction (Fig. 1D, E and Supplementary Video 2) and at later stages (Supplementary Video 3). To exclude potential bias in the manual detection of SWDs, we performed PSD analysis on recorded EEG signals. The appearance of SWD dominant peak frequency (4–6 Hz) was revealed in iKO^p/q mice after tamoxifen induction (Fig. 1F and Supplementary Fig. 2A), but not in vehicle-treated Cacna1a^flox/flox;CAG-Cre-ER mice (Supplementary Fig. 2B) or tamoxifen induced Cacna1a^flox/flox (Supplementary Fig. 2C) or CAG-Cre-ER mice (Supplementary Fig. 2D). Similar absence seizures were induced in iKO^p/q mice older than 4 months and 9 months, respectively (Supplementary Fig. 3). These seizures reproduced those described in spontaneous Cacna1a mutants (Noebels and Sidman, 1979). As in inborn P/Q channel genomic mutants, SWDs in all iKO^p/q mice (n = 6) were rapidly abolished by systemic injection of ethosuximide, a first choice anti-absence drug (Glauser et al., 2013) (Fig. 1G). Along with absence seizures, iKO^p/q mice also developed severe ataxia (Supplementary Video 4). To quantitatively assess the motor deficit, we evaluated their coordination performance on an accelerating rotarod and found a rapidly progressive decline in latency to fall (iKO^p/q, n = 13; Control, n = 17) (Fig. 1H). We also observed dystonia characteristic of the inborn genomic mutant, accompanied by a similar abnormality in cerebellar cortex of elevated ectopic expression of tyrosine hydroxylase (TH), a distinctive biomarker of Purkinje cell dysfunction (Hess and Wilson, 1991) (Fig. 1I). Together, these results demonstrate that adult loss of P/Q-type calcium channels fully recapitulates the major forebrain and cerebellar inborn neurological phenotypes.

Figure 1.

Adult Cacna1a^flox/flox;CAG-ER-Cre mice lose CACNA1A after tamoxifen induction and develop spike-wave seizure and ataxia. (A) Breeding diagram of Cacna1a^flox/flox;CAG-ER-Cre mice and control loxP (Cacna1a^flox/flox) littermates. (B) Experimental timeline for (C). DP1i21, 28, 21 and 28. (C) Left: Confocal images showing loss of CACNA1A in an Cacna1a^flox/flox;CAG-ER-Cre mouse (iKO^p/q) compared with control littermate (Ctrl) after tamoxifen induction (DP1i25). Scale bars = 1 mm. Right: Magnified images show the loss in the cerebellum cortex, thalamic LD and TRN, hippocampus, and cerebral cortex. Nuclei were labelled with 4′,6-diamidino-2-phenylindole (DAPI). Scale bars = 20 μm. (D) Representative EEG traces from control and iKO^p/q mice before (pre-tamo) and after tamoxifen induction. Right: Expanded traces show the three boxed areas. Note the development of spontaneous SWDs in iKO^p/q mice. (E) Summary graphs for the total number (left) and total duration (right) of SWDs in 30 min for iKO^p/q mice. Kolmogorov–Smirnov test, ***P < 0.001. (F) PSD (solid line, averaged; shadow, SEM) shows the dominant frequency (4–6 Hz) of SWDs in iKO^p/q mice (n = 7) at DP1i21 compared with that before tamoxifen induction. PSD was normalized to total PSD under 50 Hz. (G) Representative EEG traces show that ethosuximide (200 mg/kg body weight, i.p.) abolished SWDs from iKO^p/q mice (n = 6). (H) iKO^p/q mice (n = 13), but not control (n = 17) show a progressively shorter latency to fall on a rotarod after tamoxifen induction. Two-way (genotype × trial) ANOVA with repeated measures, ***P < 0.001. (I) Confocal images show enhanced expression of TH in the vermis of cerebellum from iKO^p/q mice compared with control. Scale bars = 20 μm. Mean ± SEM.

Unaltered burst firing in thalamocortical relay neurons

Burst firing in thalamocortical relay neurons has been widely considered to be essential for thalamocortical rhythmogenesis and SWDs (Huntsman et al., 1999; Kim et al., 2001; McCormick and Contreras, 2001, but see McCafferty et al., 2018). Thalamocortical relay neurons show augmented burst firing and/or low-threshold T-type calcium current (I_T) in various models (Zhang et al., 2002, 2004; Song et al., 2004; Cheong et al., 2009; Ernst et al., 2009; Bomben et al., 2016). We first examined rebound burst responses to membrane hyperpolarization in thalamocortical relay neurons in brain slices from iKO^p/q mice and control loxP littermates. LD and VPM thalamic nuclei, were selected because LD projects to frontal cortical regions in which cortical SWDs in inborn mutants predominate (Zhang et al., 2002), and SWDs are thought to be initiated in the somatosensory cortex targeted by VPM (Meeren et al., 2002; Polack et al., 2007). In all neurons tested, a hyperpolarization was followed by rebound excitation characterized by a burst of action potentials riding on a low-threshold calcium depolarization/spike (LTS). However, there was no significant difference in the number of burst spikes or the peak amplitude of the LTS compared to control mice in either the LD (Fig. 2A–C and Supplementary Fig. 4A and B) or VPM (Supplementary Fig. 4E). No differences were detected in the injected current, membrane capacitance or resistance (Fig. 4D and E). We then directly measured I_T via whole-cell voltage-clamp recordings and found that this current was unaffected (Fig. 2I and J). These results demonstrate that burst firing mediated by I_T in thalamocortical relay neurons remains unchanged following adult P/Q channel excision. No change in the tonic firing of LD neurons was found in iKO^p/q mice either (Supplementary Fig. 4C and D). Interestingly, we found a reduced depolarizing sag in iKO^p/q thalamocortical realy neurons (Fig. 2F, G and Supplementary Fig. 4F), indicating downregulation of hyperpolarization-activated current (I_h) mediated by hyperpolarization-activated cyclic nucleotide-gated (HCN) channels. Consistent with this finding, the resting membrane potential in these cells was significantly shifted toward more hyperpolarized value by 3.1 mV on average (Fig. 2H).

Figure 2.

Adult Cacna1a loss reduces voltage sag and hyperpolarizes resting membrane potential in thalamic laterodorsal neurons. (A) Schematic of whole-cell recording experiments. Rec. = recording electrode. (B) Representative rebound spiking of thalamocortical neurons from tamoxifen-induced control loxP and iKO^p/q mice in response to a hyperpolarizing current step. (C) Summary graph for the number of rebound spikes in the thalamic LD neurons of control and iKO^p/q mice. On average, the number of rebound spikes is similar (control: n = 13; iKO^p/q: n = 18). Kolmogorov–Smirnov test, P = 0.99. (D and E) current injected, membrane capacitance (Cm) and resistance (Rm) are similar. (F) Representative responses (solid line, averaged; shadow, repeats) to a hyperpolarizing current step into LD neurons (160 and 100 pA, respectively). (G and H) Summary graphs for sag ratio (G) and resting membrane potential (H). Depolarizing sag is smaller and the average resting membrane potential is more hyperpolarized in LD neurons from iKO^p/q mice compared with control. Student’s t-test. (I) T-type calcium current (I_T) recorded from LD neurons in response to test potentials ranging from −70 to −45 mV after being hyperpolarized (2–3 s) to −90 mV. Traces are aligned to the mean of the last 50 ms. (J) Summary graph of the peak amplitude of I_T (control: n = 10; iKO^p/q: n = 10). The average amplitude of I_T in iKO^p/q mice is similar to that in control mice. Student’s t-test. P = 0.99. *P < 0.05, **P < 0.01. n.s. = not significant. Mean ± SEM. TC = thalamocortical relay neuron.

Figure 4.

Adult Cacna1a ablation results in reduced rebound spiking and membrane excitability of neurons in TRN. (A) Schematic of whole-cell recording experiments. (B) Representative rebound spiking of TRN neurons from tamoxifen induced control loxP and iKO^p/q mice in response to a hyperpolarizing current step. (C) Summary graph of the number of rebound spikes (control: n = 22; iKO^p/q: n = 23). The average number of TRN rebound spikes in iKO^p/q mice is smaller than that for control mice. Mann–Whitney U-test. **P < 0.01. (D and E) Current injected, membrane capacitance and resistance are similar. (F) As in B, but for tonic firing in response to a series of depolarizing current steps. (G) Summary graph shows firing frequency versus the amplitude of injected current (individual neuron: grey and light red lines; average: black filled square and red filled circle) for TRN neurons (control: n = 21; iKO^p/q: n = 22). N-way ANOVA. ***P < 0.001. (H) T-type calcium current (I_T) recorded from TRN neurons in response to test potentials ranging from −60 to −30 mV after being hyperpolarized (2–3 s) to −90 mV. Traces are aligned to the mean of the last 50 ms. (I) Summary graph of the peak amplitude of I_T (control: n = 18; iKO^p/q: n = 17). On average, the amplitude of I_T in iKO^p/q mice is smaller compared with control. Student’s t-test. *P < 0.05. Mean ± SEM. TC = thalamocortical relay neuron.

Loss of burst spiking in the thalamocortical relay neurons diminishes seizure generation

Burst firing in thalamocortical relay neurons is mediated by low threshold-activated T-type Ca²⁺ channels Ca_v3.1 (Kim et al., 2001). Its pore-forming α subunit is encoded by Cacna1g. To determine directly whether the generation of spontaneous SWDs in our adult-onset model depends on burst firing in thalamocortical neurons, we eliminated Cacna1g-mediated currents in iKO^p/q mice by generating Cacna1a^flox/flox;CAG-Cre-ER;Cacna1g^−/− mice (iKO^p/q+g). Deletion of α1G was confirmed by immunohistochemistry using its specific antibody (Supplementary Fig. 5). In iKO^p/q+g mice, I_T and burst firing were abolished in thalamocortical relay neurons (Fig. 3A–C). Interestingly, a similar decrease in the sag response of LD neurons (sag ratio: 0.157 ± 0.019; Student’s t-test; P = 0.032) was found in tamoxifen induced iKO^p/q+g mice compared with control Cacna1a loxP mice. We recorded vEEG in adult iKO^p/q+g mice (n = 6) before (averaged age: 52.2 ± 1.1 days) and after tamoxifen induction, and found that after tamoxifen injection, all iKO^p/q+g mice displayed similar SWDs to those observed in iKO^p/q mice (Fig. 3D–F), although with a greatly reduced occurrence (DP1i21, SWDs number and total duration in 30 min: 37.7 ± 25.4; 102.5 ± 74.8 s, Kolmogorov–Smirnov test, P = 0.0087 compared with iKO^p/q mice at DP1i21, 169.1 ± 20.0; 454.0 ± 113.8 s; Fig. 1E). iKO^p/q+g mice also developed severe ataxia after tamoxifen induction as iKO^p/q mice (data not shown). In addition, Cacna1g knockout mice (Cacna1g^−/−) did not exhibit spontaneous SWD before or after tamoxifen induction (data not shown). These results demonstrate that Cacna1g is not strictly necessary for the adult generation of absence seizure, but still remains critical for the generation of SWDs as in inborn models.

Figure 3.

α1G deficiency suppresses but does not prevent absence seizure in Cacna1a adult-ablated mice. (A) PCR results for genotyping Cacna1a ^flox/flox;CAG-ER-Cre;Cacna1G^−/− (iKO^p/q+g) mice. The Cre primers generated a 100-bp product. The Cacna1a primers generated a 300-bp band for the wild-type or two bands (400 bp and 700 bp) for the loxP-flanked allele. The Cacna1g primers produced a 175-bp band for the wild-type, a 300-bp product for the floxed allele or a 700-bp product for knockout allele. (B) Representative responses to a hyperpolarizing current (200 pA) and a depolarizing current (100 pA) in LD neurons from iKO^p/q and iKO^p/q+g mice. No rebound spike was evoked in LD neurons (n = 9) from iKO^p/q+g mice as shown in bar graph. Student’s t-test. ***P <0.001. (C) Representative T-type calcium current (I_T) recorded from LD neurons in response to test potentials from −90 mV in tamoxifen induced iKO^p/q (DP1i23) and iKO^p/q+g (DP1i31) mice. Bottom right: Summary graph of the peak amplitude of I_T (iKO^p/q+g: n = 5). On average, the amplitude of I_T in iKO^p/q+g mice is 9.91 ± 7.51 pA. Student’s t-test. *P < 0.001. (D) Representative EEG traces (two channels, L and R) show spontaneous spike-wave-discharges (SWDs) from a tamoxifen induced iKO^p/q+g mouse (DP1i21). Expanded traces show the boxed area. (E) PSD analysis shows that unlike iKO^p/q mice, a similar PSD distribution in the iKO^p/q+g mouse at DP1i21 to that before tamoxifen induction (Pre-tamo). (F) Summary graph for PSD before and after tamoxifen induction for iKO^p/q+g mice. Mean ± SEM.

Impaired burst firing in the thalamic reticular nucleus

Thalamocortical relay neurons and TRN neurons form a reciprocal excitatory-inhibitory network. Enhanced burst firing in the TRN is widely believed to be essential for the generation of abnormal thalamocortical oscillations underlying absence epilepsy in multiple genetic models (Tsakiridou et al., 1995; Broicher et al., 2008; McCafferty et al., 2018). To examine the burst firing of TRN neurons, we performed in vitro whole-cell recordings. Unexpectedly, we found a significant impairment of burst firing (Fig. 4A–E and Supplementary Fig. 6A–C) along with reduced membrane excitability (Rheobase: Control, 47.6 ± 7.6 pA; iKO^p/q, 57.3 ± 7.7 pA; Kolmogorov–Smirnov test. P = 0.66; Fig. 4F and G) in these cells in iKO^p/q mice compared with control loxP littermates. TRN neuron resting membrane potential was increased in iKO^p/q mice (Supplementary Fig. 6D). A consistent decrease in I_T was found in TRN neurons (Fig. 4H and G). Interestingly, in tamoxifen induced iKO^p/q+g mice, a similar decrease in the burst spiking of TRN neurons was found (1.33 ± 0.76 rebound spikes; n = 14; Mann–Whitney U-test P = 0.0056 compared with control Cacna1a loxP mice, 8.27 ± 2.30 rebound spikes; n = 22). We also found an increase in the frequency of sIPSCs, rather than sEPSCs onto TRN neurons in iKO^p/q mice (Supplementary Fig. 7A and B), This finding implicates increased inhibition from GABAergic neurons located in other brain regions that target TRN [as adult mice lack intra-TRN GABAergic connections (Hou et al., 2016)], or possibly new formation of intra-TRN GABAergic synapses induced by P/Q channel ablation. In contrast, the frequency and amplitude of sIPSCs and sEPSCs onto thalamic relay neurons remained unchanged (Supplementary Fig. 7C and D). Together, these results demonstrate that burst firing in TRN is impaired in mice with adult-onset P/Q channel ablation and suggest that suppressing, rather than promoting TRN excitability might be essential for generating adult-onset absence seizures.

Loss of Cacna1h promotes seizure generation

Burst firing in TRN is mediated by two subtypes of T-type calcium channels, Ca_v3.2 and Ca_v3.3 (encoded by Cacna1h and Cacna1i, respectively) (Talley et al., 1999). The Ca_v3.2-encoding gene CACNA1H has been associated with human CAE in genetic studies (Chen et al., 2003b; Heron et al., 2007). The significance of CACNA1H in CAE is further heightened by the fact that nearly half of patients with CAE failed in treatment with ethosuximide, a front line anti-CAE drug (Coulter et al., 1989; Glauser et al., 2010; Wang et al., 2015). How CACNA1H mutations contribute to CAE is, however, unclear. In addition, unlike Ca_v3.1 and Ca_v3.3, the role of Ca_v3.2 in SWDs has not been explored by genetic deletion (Lee et al., 2014). To address this question, we generated Cacna1a^flox/flox;CAG-Cre-ER;Cacna1h^−/−mice (iKO^p/q+h) to examine the effect of adult P/Q deletion on a Cacna1h^−/− background (Chen et al., 2003a) (Fig. 5A). The additional knockout of Cacna1h slightly reduced burst spiking and membrane excitability in TRN neurons to a level comparable to tamoxifen induced iKO^p/q mice (Fig. 5B–D). We recorded EEG activity from these mice before and after tamoxifen induction. Surprisingly, we found that tamoxifen reliably induced SWDs in all iKO^p/q+h mice, similar to iKO^p/q mice (Fig. 5E and F); indeed, SWD duration, which purportedly depends upon rebound burst firing, was even longer (Fig. 5G). All iKO^p/q+h mice also developed severe ataxia after tamoxifen induction as seen in iKO^p/q mice (data not shown). Together, these results indicate a potential downstream role of Cacna1h reduction in favouring the generation of absence epilepsy due to adult-onset P/Q calcium channel dysfunction. Consistent with our findings, further suppression of rebound burst firing in Cacna1i-deficient TRN neurons by Cacna1h deletion did not abolish, but instead enhanced pharmacologically-induced absence seizures in adult mice (Lee et al., 2014).

Figure 5.

Cacna1h null mutation results in reduced TRN excitability and prolonged SWDs in Cacna1a adult-ablated mice. (A) PCR results for genotyping Cacna1a^flox/flox;CAG-ER-Cre;Cacna1h^−/− (iKO^p/q+h) mice. The Cre primers generated a 100-bp product. The Cacna1a primers generated a 300-bp band for the wild-type or two bands (400 bp and 700 bp) for the loxP-flanked allele. The Cacna1h primers produced a 480-bp band for the wild-type or a 330-bp product for the knockout allele. (B) Left: Representative responses to a hyperpolarizing step current injected into TRN neurons in iKO^p/q+h mice without tamoxifen induction. Right: Summary graph shows fewer rebound spikes in TRN neurons from iKO^p/q+h mice (n = 12) compared with Cacna1a^flox/flox mice (Ctrl, dash bar). Mann–Whitney U-test. *P < 0.05. (C) As in B, but for tonic firing in response to a series of depolarizing current steps. (D) Summary graph shows firing frequency versus the amplitude of current injected (individual neuron: light magenta lines; averaged: magenta filled circle and grey line) for TRN neurons (iKO^p/q+h: n = 7). TRN neuron from iKO^p/q+h mice is less excitable compared with Cacna1a^flox/flox mice (Ctrl, grey line). N-way ANOVA. ***P < 0.001. (E) Representative EEG traces show spontaneous SWDs from an iKO^p/q+h mouse induced by tamoxifen injection. Expanded traces show the boxed areas. (F) PSD analysis shows the dominant frequency (4–6 Hz) of SWDs in iKO^p/q+h mice (n = 5) at DP1i21 compared with that before tamoxifen induction (Pre-tamo). (G) Bar graph shows that the average duration of SWDs from iKO^p/q+h mice is longer compared with iKO^p/q mice. Student’s t-test. *P < 0.05, **P < 0.01. Mean ± SEM.

Discussion

We established an adult-onset animal model through tamoxifen induced ablation of P/Q-type calcium channels for childhood absence epilepsy, and determined that the thalamocortical synchronization disorder in the mature brain arises from distinct thalamic circuit dynamics compared with inborn models.

Adult-onset P/Q channel deletion in mice recapitulates childhood absence epilepsy

Previous studies demonstrated that genomic loss of P/Q calcium channels shifts the reliance of excitation-release coupling at central synapses to more loosely coupled N- and R-type channels (Midorikawa et al., 2014), and when limited to a single presynaptic corticothalamic input, trans-synaptically elevates T-type calcium current in postsynaptic thalamic neurons, leading to absence epilepsy (Bomben et al., 2016). Here, through loss-of-function experiments, we demonstrate that P/Q calcium channels remain essential for regulating brain rhythomogenesis and motor functions in the adult animal, and show that a key cerebellar endophenotype (ectopic expression of TH) is also recapitulated in our model. Together, these results demonstrate a development-independent role of P/Q calcium channels in the generation of epilepsy and ataxia due to mutation of Cacna1a.

The role of Ca_v3.1-mediated burst firing in absence epilepsy

Numerous studies have supported the idea that repetitive burst firing of thalamocortical relay neurons plays a key role in the generation of SWDs, manifest as hyper-synchronized episodic oscillation activities in the thalamocortical network. However, there is now increasing controversy on this point (Huguenard, 2019). Burst firing in the thalamocortical relay neurons is mediated by Cacna1g (Ca_v3.1 channels). Despite the fact that overexpression of Cacna1g promotes SWD (Ernst et al., 2009), we found, as have others (Song et al., 2004), that Ca_v3.1 deletion suppressed SWDs in Cacna1a mutants to a significant extent, even in the adult-onset model. Our results support the notion that Ca_v3.1-mediated thalamocortical burst firing is essential for the generation of SWDs, although it is possible that loss of Ca_v3.1 in brain regions other than thalamus, such as neocortex, is responsible for the suppression. Region specific Ca_v3.1 channel manipulation is required to test this idea in the future. It is noteworthy that Ca_v3.1 null mutation completely abolished SWDs in inborn α1A knockout mice (although these mice died around 3–4 weeks after birth) (Jun et al., 1999; Song et al., 2004), whereas loss of Ca_v3.1 does not completely block SWDs in our adult-onset P/Q channel deleted mice (iKO^p/q+g), similar to the results found after removing Ca_v3.1 in three other mutants with absence seizures, i.e. tottering (α1A^tg/tg), lethargic (β4^lh/lh), or stargazer (γ2^stg/stg) (Song et al., 2004).

Burst firing in thalamocortical relay neurons is regulated by various ion channels, including HCN channels. Homo- or heterotetrameric channels are formed by four HCN channel subunits (HCN1–4) that display a distinct pattern of distribution in the nervous system (Santoro et al., 2000; Notomi and Shigemoto, 2004). Loss of HCN1 subunits promotes epileptogenesis in mice and HCN1-deficient rat and HCN2-deficient mice exhibit spontaneous absence seizures (Ludwig et al., 2003; Chung et al., 2009; Huang et al., 2009; Nishitani et al., 2019), while pharmacological blockade or local antisense reduction of HCN channels in thalamocortical relay neurons suppress SWDs in two genetic models (Cain et al., 2015; David et al., 2018). Although the number of spikes in a rebound burst evoked by hyperpolarization was not altered in the thalamocortical relay neurons of iKO^p/q mice, we found a significant reduction in the depolarizing sag response that is likely mediated by HCN channels. How this change affects thalamocortical relay neuron burst firing in vivo and contributes to the generation of SWDs remains to be investigated. Further studies are also warranted to investigate how HCN channels are regulated. It is noteworthy that HCN channels could form a complex with Ca_v3.2 channels (Fan et al., 2017), implying that the HCN and Ca_v3.2 channels might be co-regulated by the same signalling pathway in iKO^p/q mice upon P/Q channel ablation.

Epileptogenic regulation of absence epilepsy by the CACNA1H gene

Altered low-threshold bursting in TRN neurons has been hypothesized to drive thalamocortical hyper-synchrony underlying SWDs (Tsakiridou et al., 1995; Cain et al., 2018; McCafferty et al., 2018). In contrast, complete loss of burst firing in TRN neurons does not prevent but rather promote pharmacologically induced SWDs in mice with compound deletion of both Ca_v3.2 and Ca_v3.3 (Lee et al., 2014). Similar results were found in Ca_v3.3 knockout mice in the same study. How loss of Ca_v3.2 alone affects SWD generation has not previously been examined in a genetic absence seizure model. Some mutations found in human CAE patients were shown to be gain-of-function in heterologous cells; however, the majority had no effect on the biophysical properties of Ca_v3.2 channels (Khosravani et al., 2004, 2005; Peloquin et al., 2006), suggesting that epilepsy-related mutations in CACNA1H might affect seizure threshold via regulating channel trafficking or RNA splicing rather than directly mediating biophysical changes in pore dynamics the channels (Khosravani and Zamponi, 2006; Proft et al., 2017). In one study, absence epilepsy was not found in Ca_v3.2 knockout mice, although it was reported that Ca_v3.2 knockout mice had longer slow-wave (non-rapid-eye-movement) sleep bouts (Pellegrini et al., 2016). Nonetheless, our results support a seizure promoting role for Cacna1h gene loss-of-function in a genetic absence seizure model. This ancillary role may bear on the current controversy over the contribution of TRN and thalamocortical neuron bursting underlying SWD generation (Huguenard, 2019).

It has been recently reported that local infusion of a T-type calcium channels blocker, TTA-P2 in ventrobasal thalamus had no effect on absence seizures (McCafferty et al., 2018). One explanation for this result might be that the pharmacological action of this agent on Ca_v3.1 and Ca_v3.2 obscures the individual contribution of each T-type channel, since TTA-P2 blocks all three T-type calcium channels (Shipe et al., 2008). In addition, other than contributing to bursting properties of TRN neurons, Ca_v3.2 is also expressed at axon terminals and regulates synaptic transmission (Huang et al., 2011; Wang et al., 2015), although whether Ca_v3.2 is expressed at TRN presynaptic terminals targeting thalamocortical relay neurons and its role in synaptic transmission remains to be determined. A similar confounding effect might explain the ability of ethosuximide to block SWD generation in our model, as intraperitoneal ethosuximide acts throughout the thalamocortical circuit. Without minimizing the contribution of intracortical circuitry to spike-wave seizure generation, our analysis focuses on the role of thalamic rhythm generation.

Circuit heterogeneity and network burst disparity underlying corticothalamic dysrhythmia

Based on prior studies and our results, we propose that excitability changes in thalamocortical circuits generating SWDs are heterogeneous and that some SWDs, at least in mouse genetic models, can be generated by variable sets of conductance distributed within these circuits. In support of this view, circuit heterogeneity is well established in human disorders featuring SWDs (Blumenfeld, 2005; Tenney et al., 2014). Our findings provide direct support for degenerate current profiles underlying pathogenic rhythmic behaviour in the mouse thalamocortical circuitry. Namely, the precisely regulated burst patterns within this tightly-coupled network involving neocortex, TRN, and thalamo-cortical relay neurons, may arise from flexible sets of underlying conductances. This seminal construct, supported by analysis of a tri-cellular circuit in a crustacean ganglion and in silico modelling (Gutierrez and Marder, 2014; O'Leary et al., 2014), predicts not only a spectrum of homeostatic membrane excitability properties driving stable oscillatory rhythms at various ages, but also multiple, and perhaps age-dependent therapeutic solutions for restoring normal circuit behaviour. Our evidence demonstrates that P/Q-type calcium channels, which share control of calcium-mediated exocytosis with N- and R-type channels at most central synapses, retain an essential role in normal adult cortical network synchrony, and indicate that inherited childhood absence and ataxia syndromes do not necessarily arise from an irreversible structural perturbation of synaptogenesis during early brain development. Adult deletion of P/Q channels also provides novel insight into the persistent plasticity of downstream thalamic T-type currents in adult mice, revealing cell and age-specific patterns of homeostatic remodelling that may serve as new targets for reversing neurological disorders mediated by the P/Q-type calcium channel.

Glossary

Abbreviations

CAE =

childhood absence epilepsy

DP1i =

days post-first injection

HCN =

hyperpolarization-activated cyclic nucleotide-gated

LD =

laterodorsal nucleus

PSD =

power spectral density

SWD =

spike-and-wave discharge

TRN =

thalamic reticular nucleus