Tiered modelling of a CACNA1A D1634N mutation linked to ataxia, epilepsy and cognitive deficits

Abstract

Missense mutations in CACNA1A, which encodes Ca_V2.1 voltage-gated calcium channels, underlie serious neurological diseases with inadequate treatment options. Development of new therapies for CACNA1A diseases is hindered by: (i) the unknown biophysical impact of many mutations; (ii) limited knowledge of genotype effects on channel function and behavioural phenotype; and (iii) dearth of animal models mimicking the human disease. Here, we focus on a CACNA1A missense mutation (D1634N) identified in a patient with ataxia, epilepsy, moderate intellectual disability and mild cerebellar loss.

Heterologously expressed D1634N channels displayed a significantly decreased whole-cell current amplitude, a +25 mV shift in the half-maximal voltage of activation (V_½) and diminished single-channel open probability, indicating a loss of function. Patient-specific cortical neurons derived from induced pluripotent stem cells displayed enhanced intrinsic excitability compared with isogenic controls. Heterozygous knock-in mice with the analogous mutation (D1585N) displayed ataxia, absence epilepsy and cognitive deficits, mirroring the human behavioural phenotype. Purkinje cell and cerebellar slice electrophysiology revealed a reduced whole-cell Ca_V2.1 current with a right-shifted V_½ of activation, altered Purkinje cell excitability and changes in presynaptic vesicle release probability at the parallel fibre to Purkinje neuron input.

The results introduce a new CACNA1A loss-of-function mutation and highlight the power of combined heterologous expression, induced pluripotent stem cell neurons and animal model studies to provide complementary insights into the molecular basis of disease, with implications for developing therapies.

Niu et al. characterize a loss-of-function CACNA1A calcium channel mutation that causes epilepsy and ataxia. Using heterologous systems, patient iPSC-derived neurons, and a knock-in mouse model, they link altered channel biophysics to neuronal hyperexcitability and disease-relevant behavioural phenotypes.

Introduction

CACNA1A encodes the pore-forming subunit (α_1A) of P/Q-type voltage-gated Ca²⁺ channels (Ca_V2.1), which are highly expressed in the cerebellum and broadly present in other brain regions. Ca²⁺ influx through presynaptic Ca_V2.1 controls neurotransmission across numerous CNS synapses and is critical to normal brain function.^1,2 Human CACNA1A mutations are associated with epilepsy, ataxia, intellectual disability and cognitive impairment, hemiplegic migraine, autism spectrum disorder and dystonia.^3-5 Hundreds of CACNA1A mutations are in the ClinVar database, with many classified as variants of unknown significance.⁶ Current treatments for CACNA1A diseases aim at alleviating symptoms and include anti-migraine and anti-seizure medications, physical and speech therapy, and lifestyle changes to avoid seizure triggers. However, many patients are refractory to available drugs, and treatments do not prevent the progressive ataxia and cerebellar degeneration observed with many genetic variants.⁶

The diverse mutations and broad symptoms experienced by CACNA1A patients emphasize personalized approaches to disease management. Crucial elements for progress include: (i) knowing how mutations affect function; (ii) understanding genotype-to-phenotype relationships; and (iii) development of model organisms reflecting human disease symptoms. Heterologous expression studies have illuminated biophysical impacts of some CACNA1A mutations and indicate that gain-of-function and loss-of-function (LoF) mutations yield distinctive disease phenotypes.^7-10 Two mutations, R192Q and S218L, in patients with mild and severe forms of familial hemiplegic migraine type 1, respectively, cause hyperpolarizing shifts in Ca_V2.1 activation, albeit to different extents.^11,12 In contrast, LoF mutations are associated with episodic ataxia type 2, absence epilepsy and developmental epileptic encephalopathy.^5,7 Several spontaneously arising mutant mice (e.g. tottering, leaner and rolling Nagoya) have provided valuable insights into distinctive behavioural impacts of CACNA1A LoF and gain-of-function mutations.^11-14 A significant limitation of these mouse models is that behavioural deficits are recessive, different from the autosomal dominant condition in humans.

Here, we generated multi-level models to investigate a de novo CACNA1A mutation (D1634N) identified in a patient with epilepsy, ataxia and intellectual disability. Heterologously expressed D1634N showed reduced whole-cell current, a +25 mV right shift in activation gating and diminished single-channel open probability (P_o). Human induced pluripotent stem cell-derived cortical neurons (iPSC neurons) generated from patient blood displayed increased intrinsic neuronal excitability compared with gene-edited isogenic controls. Finally, a heterozygous knock-in mouse with the equivalent mutation (D1585N^+/−) recapitulated behavioural features found in the patient.

Materials and methods

Detailed methods are provided in the Supplementary material.

Molecular biology

We used human Ca_V2.1 isoform 1 variant (NM_000068.4).¹⁵ Ca_V2.1 D1634N mutation and Venus insertion were accomplished by PCR amplification and restriction cloning.

Transfection of HEK293 cells

The calcium phosphate precipitation method was used to transfect plasmid DNA into HEK293 cells cultured on 60 mm dishes.

Purkinje neuron isolation

Purkinje neurons were isolated from mouse pups aged postnatal Days 11–20, as previously described.^16,17 Purkinje neurons isolated from each individual animal were used between 30 min and 5 h of isolation, and genotypes of animals were blinded until recordings were completed.¹⁷

Cerebellar and cortical slice preparation

Thin cerebellar slices (250 μm) were prepared from 2–12-month-old wild-type (WT) and D1585N^+/− mice. Procedures were as approved by Columbia University Institutional Animal Care and Use Committee (IACUC). Slices were cut in artificial CSF.¹⁸

Coronal brain slices (250–300 μm thick) containing the primary somatosensory cortex (S1) were prepared from young mice (postnatal Days 40–60).

Electrophysiology

Whole-cell recordings from HEK293 cells, iPSC-derived neurons and isolated Purkinje cells

Transfected HEK293 cells were split onto poly-D-lysine-coated glass coverslips, and patch-clamp experiments were performed using Axopatch 200A (Axon Instruments) or MultiClamp 700B (Molecular Devices) amplifiers. For iPSC neurons and Purkinje neurons, similar conditions to those for HEK293 cells were used, except that glass pipette resistances were increased to 6–12 MΩ and iPSC voltage-clamp measurements were performed with 5 mM BaCl₂ as the charge carrier.

Single-channel recordings

Approximately 14–16 h after transfection, single-channel recordings were performed at room temperature using an on-cell configuration.^19,20

Whole-cell recordings from Purkinje cells in cerebellar slices

Whole-cell recordings were obtained from Purkinje cells in cerebellar slices. Data were recorded and acquired using an AxoPatch 200B amplifier and pClamp 9 software (Molecular Devices).

iPSC cell-line derivation, characterization, maintenance and neuron differentiation

Columbia’s Stem Cell Core reprogrammed peripheral blood mononuclear cells isolated from the patient blood sample into iPSCs with Yamanaka factors through Sendai virus transduction.²¹ CRISPR gene editing was used to create a corrected control in the genetic background of the patient. Neural progenitor cells were differentiated into cortical glutamatergic neurons using the STEMdiff + SMADi neural induction kit (StemCell Technologies). iPSC cortical neurons were cultured for ∼70–85 days before experiments.

Mouse generation and behaviour

Heterozygous knock-in mice were generated by Cyagen using CRISPR-Cas9 gene-editing tools to introduce the human patient mutation (D1634N) into the analogous position in the murine gene (D1585N) on the C57BL/6J genetic background. Mutations were verified via PCR and sequencing through Transnetyx. Behaviour tests were performed at Columbia University Mouse Neurobiology Core.

Beam test

Animals were tested on their ability to cross a 55-cm-long, 5-mm-wide elevated beam using Noldus’ Ethovision XT software.²²

Rotarod

Mice were probed for motor coordination and motor learning skills using a RotaRod (UGO Basile).²³

Catwalk

Mice gait patterns and weight distribution on a flat surface were assessed using a Catwalk apparatus (Noldus Information Technology).

Open field

The open field test was performed to evaluate locomotor activity in mice. Each mouse underwent a single 60 min trial in an open-field arena (27.9 cm × 27.9 cm).

Y-Maze

Mice were assessed for spatial working and reference memory by Y-maze tests, using Noldus’s Ethovision software for data recording and analysis, where each arm had a unique sticker for differentiation.^24,25

Fear conditioning

Mice were assessed for associated learning and memory with a contextual fear conditioning test.²⁶

Calcium imaging

Calcium imaging of HEK293 and GCaMP6f-expressing Purkinje cells in the cerebellar slices was performed using a Nikon Eclipse Ti inverted microscope equipped with Nikon 10×/0.25 Ph1 ADL, Plan Fluor ELWD 20×/0.45 Ph1 MD and Plan Fluor 40×/1.3 oil DICH/N2 objectives.

Immunohistochemistry

Cerebellar slices were fixed overnight in 4% paraformaldehyde, washed in PBS, incubated in 5% normal goat serum with Triton X-100 (NGST) blocking solution (5% normal goat serum, 0.1 M PBS and 0.3% Triton X-100) for 1 h, and subsequently incubated in 1% NGST buffer containing primary antibody for 12–36 h. Tissues were washed and incubated in 1% NGST containing secondary antibodies for 2 h, and embedded in Fluoromount-G.

Confocal imaging

Stained tissues were imaged using a Nikon Eclipse Ti A1-A laser scanning confocal microscope controlled by NIS-Elements AR 5.02.00 (64-bit) software.

Image analysis

Quantitative analysis of degenerated dendritic structures, referred to as ‘round or oval structures’, within cerebellar slices was performed using ImageJ software (National Institutes of Health).

EEG recording

Mice aged between 7 and 10 weeks were anaesthetized by isoflurane inhalation. Epidural electrodes were implanted as previously described.²⁷ The EEG data were acquired and visualized using Natus Neuroworks software (v.8.5.1). Spike wave discharge (SWD)-like events were detected and quantified with Assyst software.²⁸

Data analysis

Electrophysiology analysis

For whole-cell recordings, custom MATLAB (Mathworks, https://github.com/manubenjohny/WCDTY) or commercial pClamp software was used to determine peak currents. For single-channel recordings, patches were analysed by first fitting and subtracting the leak from each trace.

Sholl analysis

Induced PSC neurons were infected with adenovirus encoding green fluorescence protein (Ad5-GFP) and confocal images were captured at ×20. Images were processed in Fiji with the SNT module.²⁹

Statistics

Data were processed with Prism. The statistical tests chosen and the P-values are indicated where significance is referenced. Data are expressed as the mean ± standard error of the mean.

Results

Human phenotype of a de novo CACNA1A D1634N mutation

The proband is a 27-year-old female with a de novo c.4900G>A; p. D1634N CACNA1A variant identified on trio exome sequencing. She has a history of epilepsy and intellectual disabilities. Her mother noted that fetal movements appeared reduced during pregnancy compared with her previous pregnancies, although this observation was not quantified formally. She was delivered via vaginal delivery at 41 weeks, with a birth weight of 3.9 kg and Apgar scores of 9 and 9 at 1 and 5 min, respectively. Her prenatal and perinatal courses were otherwise unremarkable. Concerns were first noted at 6 weeks of age, when she was not holding her head unsupported and was a quiet baby who did not cry often. At 8 weeks she was not visually tracking. Her early milestones were delayed. She held her head unsupported at 6 months, sat unsupported at 10 months, crawled at 12 months and took her first steps at 22 months. She said her first words at 2 years and put words together at 3 years. She was noted to be hypotonic on physical examination at 6 months and started with physical therapy and occupational therapy. She had her first seizure at 9 months of age. Her seizures have been relatively well controlled on Keppra, with approximately two or three breakthrough seizures per year. Poor sleep can trigger seizures, which are generally brief and associated with a postictal period when she sleeps. Starting at age 14 years, approximately once a week she would become very dizzy and feel her head spinning with nystagmus and need to take a nap. She has frequent headaches and occasionally has tingling in her arms. EEGs have been abnormal, with frequent spikes and waves with bursting activity over the temporal lobes bilaterally. She has had mild cerebellar loss on MRI. She is currently treated with brivaracetam and topiramate with good effect. She has had intermittent difficulty with balance, has repeatedly fallen and sustained traumatic injuries, including bone fractures and a dislocated patella. In general, she is clumsy and has difficulty with coordination. On physical examination, heel to shin testing was clumsy, and tandem gait could not be performed without losing balance. Her receptive and expressive language is good, although her speech has deteriorated over time. She reads at a third-grade level and has a full-scale IQ of 79. Her memory is relatively good. She has greater challenges with executive function, visual–spatial processing, understanding complex information, processing speed and language, including verbal fluency and semantic fluency. She gets along well with others but has some limitations in engaging with peers socially.

Impact of CACNA1A D1634N mutation on Ca_V2.1 gating and ionic currents

D1634N occurs in Ca_V2.1 domain IV segment 3 (DIV S3) and provides a counter charge for movement of DIV S4 arginine residues during channel gating (Fig. 1).^30,31 To determine the biophysical impact of the mutation, we used site-directed mutagenesis to introduce D1634N into human Ca_V2.1 complementary DNA, and compared functional properties of mutant and WT Ca_V2.1 in transiently transfected HEK293 cells. Initially, we used whole-cell electrophysiology to characterize function. Test depolarizations of cells expressing WT Ca_V2.1 yielded robust inward currents that activated at a threshold of −20 mV and peaked at +10 mV (I_peak = −22.1 ± 4.7 pA/pF, n = 12; Fig. 2A and B). In comparison, whole-cell cell currents from cells expressing D1634N activated at a threshold of ∼−5 mV and displayed a significantly smaller maximum inward current (I_peak = −14.7 ± 2.8 pA/pF, n = 12, P < 0.05, Student’s unpaired t-test) that peaked at +30 mV (Fig. 2A and B). Tail-current activation curves confirmed the right shift in voltage dependence of D1634N activation compared with WT Ca_V2.1 (Fig. 2C). Unexpectedly, maximum tail current −90 mV following test pulse depolarization to +60 mV was larger for D1634N than WT Ca_V2.1 (Fig. 2C).

Figure 1.

Predicted D1634N location and structural implications in Ca_V2.1. (A) Cartoon schematic depicting the location of the patient mutation, D1634N, in domain IV transmembrane segment 3 as a counter charge to the positive charges in transmembrane segment 4. (B) AlphaFold prediction of the structure of Ca_V2.1, showing (inset) the predicted interaction between K1670 in the voltage domain and D1634.

Figure 2.

Biophysical impact of D1634N mutation on heterologously expressed Ca_V2.1 channels. (A) Top: Schematic diagram of whole-cell patch clamp of HEK293 cell and voltage pulse protocol. Bottom: Exemplar whole-cell current traces evoked by test depolarizations (from −50 to +60 mV in +10 mV increments) in HEK293 cells expressing either wild-type (WT) Ca_V2.1 (α_1A/β₄/α₂d-1; black) or D1634N Ca_V2.1 (red). (B) Population data peak current–voltage (I–V) relationships for WT Ca_V2.1 (black squares, n = 12) and D1634N (red squares, n = 12). (C) Population tail current–voltage relationships WT Ca_V2.1 (black circles, n = 12) and D1634N (red circles, n = 12). (D) Left: Exemplar whole-cell current traces (bottom) evoked with action potential waveforms (APWs) in HEK293 cells expressing either WT Ca_V2.1 (black trace, n = 12) or D1634N (red trace, n = 12). Right: Summary of APW-evoked peak currents with WT and D1634N Ca_V2.1 (*P < 0.01, Student’s unpaired t-test). (E) Exemplar single-channel recordings evoked with a voltage ramp (from −50 to +40 mV) in cells expressing either WT Ca_V2.1 (black traces) or D1634N (red traces). (F) Ensemble averages of single-channel recordings for WT Ca_V2.1 (black, n = 5) and D1634N (red, n = 6). (G) Images of fura2-loaded HEK293 cells (top) and exemplar traces of fura2-reported Ca²⁺ transients evoked with high-K⁺ stimulation (bottom) for untransfected HEK293 cells (blue trace, left) or cells expressing either WT Ca_V2.1 (black trace, middle) or D1634N (red trace, right). Population Ca²⁺ transient amplitude data for untransfected cells (blue, n = 69), WT Ca_V2.1 (black, n = 39) and D1634N (red, n = 35). (H) Radial plot of key parameters from different assays for WT (black) and D1634N (red) channels.

Next, we used a Purkinje neuron action potential waveform (APW) as the command potential to elicit inward currents from heterologously expressed WT Ca_V2.1 and D1634N, respectively. WT Ca_V2.1 yielded a transient inward current that peaked during the falling phase of the Purkinje neuron APW (I_peak,APW = −3.0 ± 0.6 pA/pF; Fig. 2D). Purkinje neuron APW-evoked inward currents from D1634N were significantly smaller compared with WT Ca_V2.1 (I_peak,APW = −1.5 ± 0.2 pA/pF, P < 0.01, Student’s unpaired t-test), confirming that the mutation would be expected to result in decreased Ca²⁺ influx in neurons (Fig. 2D).

To gain deeper insights into microscopic mechanisms underlying decreased D1634N whole-cell current, we turned to single-channel recordings. A voltage-ramp protocol (from −80 to +80 mV) in cell-attached patch-clamp mode was used to elicit currents from WT Ca_V2.1 and D1634N single channels. Exemplar recordings from WT Ca_V2.1 show stochastic inward deflections to the predicted Goldman–Hodgkin–Katz equation curve (Fig. 2E, left). Ensemble averages obtained from multiple sweeps yielded a P_o versus voltage relationship with a half-maximal voltage of activation (V_½) = +13.5 mV and a maximum P_o (P_o,max) of 0.32 (Fig. 2F). In contrast, D1634N displayed right-shifted unitary currents (V_½ = +23 mV) and reduced P_o,max = 0.14 (Fig. 2E and F).

Finally, we measured Ca²⁺ influx using the ratiometric Ca²⁺-sensitive indicator fura-2 in HEK293 cells expressing either WT Ca_V2.1 or D1634N together with the inwardly rectifying potassium channel, Kir2.1. Exposure to high K⁺ (140 mM) depolarized the membrane and yielded a robust Ca²⁺ transient in cells expressing WT Ca_V2.1 (Fig. 2G). In contrast, untransfected cells and D1634N-expressing cells displayed markedly lower responses to high-K⁺ stimulation (Fig. 2G).

A radial plot comparing various parameters provided a holistic signature for WT Ca_V2.1 and D1634N; in comparison to WT, mutant channels displayed a similar level of expression (F_Venus); reduced current density, single-channel P_o and Ca²⁺ influx; increased maximal tail current; and right-shifted voltage dependence of activation (Fig. 2H).

Assessing functional impact of Ca_V2.1 D1634N in human iPSC cortical neurons

We next turned to iPSC neurons to determine the impact of D1634N in human neurons. Patient blood was reprogrammed into iPSCs with Yamanaka factors using Sendai virus transduction.²¹ Patient iPSCs displayed no chromosomal abnormalities upon karyotyping, and the presence of the heterozygous D1634N variant was confirmed by sequencing (Fig. 3A). We used CRISPR-Cas9 gene editing to create an isogenic control, in which D1634N was corrected to WT (Fig. 3A). For the CRISPR-edited line, we introduced a silent mutation into the protospacer adjacent motif (PAM) sequence we targeted in the mutant allele. Sequencing of the gene-edited iPSC line shows the presence of both the WT allele (GCC) and the silent PAM mutation (GCG), indicating that both alleles are present (Fig. 3A). iPSCs were converted to neural progenitor cells, then differentiated into cortical neurons ∼70–85 days before they were used for subsequent experiments. The iPSC cortical neurons stained positive for MAP2 and β3 tubulin (Supplementary Fig. 1A and B). Copy number variance analysis was performed to check for potential insertions, deletions and duplications and indicated no chromosomal aberrations or large deviations between the patient and CRISPR gene-edited iPSC cell lines (Supplementary Fig. 1C). Furthermore, we performed off-target sequencing on the top four possible sites determined by the cutting frequency determination (CFD) score calculated by CRISPOR and found no differences at these sites between patient and gene-edited iPSC cell lines (Supplementary Fig. 1D). Sholl analysis indicated D1634N^+/− iPSC neurons trended towards an increased average dendritic length (patient = 22.3 ± 1.3 μm, gene-edited = 17.9 ± 1.8 μm, P = 0.05, Student’s t-test), and reduced total dendritic length (patient = 826.2 ± 91.0 μm, gene-edited = 1103.1 ± 181.2 μm, P = 0.14, Student’s t-test; Fig. 3B–D).

Figure 3.

Morphological and electrical characterization of patient induced pluripotent stem cell-derived cortical neurons. (A) Top: Karyotype of patient induced pluripotent stem cells (iPSCs; left) and gene-edited iPSCs (right). Bottom: Chromatogram of gene sequencing for D1634N variant (left) and gene-corrected sequence (right), where the variant is highlighted in red and the silent PAM mutation for gene editing is highlighted in blue. (B) Exemplar Sholl analysis of patient (top) and gene-edited (bottom) iPSC-derived cortical neurons. Each concentric circle increases in diameter by 10 μm. (C and D) Population statistics for total dendritic length (P = 0.14, Student’s unpaired t-test) (C) and average dendritic length (P = 0.05, Student’s unpaired t-test) (D) shows no significant difference between patient (red) and gene-edited (blue) neurons. (E) Exemplar total Ca²⁺ current trace elicited at −30 mV (left), ω-agatoxin IVA (ω-Aga IVA)-insensitive Ca²⁺ current (middle) and ω-Aga IVA-sensitive Ca²⁺ current determined from subtraction, with expanded time scale (right and inset). (F) Population data for peak ω-Aga IVA-sensitive Ca²⁺ current at −30 mV (P = 0.12, Student’s unpaired t-test). (G) Population I–V relationships for patient and gene-edited ω-Aga IVA-sensitive Ca²⁺ currents. (H) Fitted V_½ for the I–V curves (P = 0.02, Student’s unpaired t-test). (I) Resting membrane potentials for patient and gene-edited neurons (P = 0.56, Student’s t-test). (J) Threshold potential for action potential firing for patient and gene-edited neurons (P = 0.04, Student’s unpaired t-test). (K) Exemplar action potential waveforms (APWs) for patient (left, red) and gene-edited (right, blue) neurons evoked with +30 pA injected. (L) Frequency of action potential spikes for patient (red, n = 19) and gene-edited (blue, n = 11) neurons (P = 0.03 by Student’s unpaired t-test). *P < 0.05, Student’s two-tailed unpaired t-test. ns = not significant; WT = wild-type.

In whole-cell recordings with solutions designed to isolate Ca_V channel currents, a 0 mV test pulse elicited inward currents with two components: a sharp inward current that decayed rapidly to a plateau level (Fig. 3E). To identify the current component carried by Ca_V2.1, we exposed voltage-clamped iPSC neurons to 500 nM ω-agatoxin IVA (ω-Aga IVA), a selective blocker of Ca_V2.1.^32,33 Surprisingly, the rapidly decaying component of current was eliminated selectively by ω-Aga IVA, identifying it as being carried through Ca_V2.1. Current–voltage (I–V) curves indicated that ω-Aga IVA-sensitive current from D1634N^+/− iPSC neurons trended towards reduced amplitude at all test voltages (I_peak = −14.6 ± 4.0 pA/pF) compared with gene-edited neurons (I_peak = −22.6 ± 3.1 pA/pF, P = 0.12, Student’s unpaired t-test; Fig. 3F). Fits of I–V relationships (Fig. 3G) to a modified Boltzmann equation indicated a significantly right-shifted V_½ for D1634N^+/− iPSC neurons (V_½ = −19.9 ± 3.0 mV) compared with gene-edited controls (V_½ = −28.3 ± 1.8 mV, P = 0.02, Student’s unpaired t-test; Fig. 3H).

In current-clamp experiments, there was no difference in resting membrane potential between patient and gene-corrected neurons (V_rest,patient = −57.0 ± 1.4 mV, V_rest,edited = −55.4 ± 2.7 mV, P = 0.56, Student’s t-test; Fig. 3I). Both patient and gene-corrected neurons displayed action potential (AP) trains when injected with +30 pA current pulse for 1 s (Fig. 3K). The threshold potential for AP firing was significantly lower for patient (V_threshold = −37.2 ± 1.0 mV) versus gene-corrected neurons (V_threshold = −33.7 ± 1.0 mV, P = 0.04, Student’s unpaired t-test; Fig. 3J). Injection of a family of 1 s current pulses (10–50 pA in 10 pA increments) in D1634N^+/− neurons yielded AP trains with spike frequency increasing monotonically, reaching a peak of 13.2 ± 1.1 spikes/s (n = 19). In contrast, gene-corrected neurons displayed a significantly reduced maximum spike frequency (8.1 ± 1.9 spikes/s, n = 11, P = 0.03 by Student’s unpaired t-test; Fig. 3L). Finally, we measured miniature excitatory postsynaptic currents (mEPSCs) with whole-cell patch clamp in the presence of tetrodotoxin and picrotoxin (Supplementary Fig. 2). We observed no significant difference in either the frequency or the amplitude of mEPSCs, indicating no major variation in vesicle release probability. Overall, the results show that D1634N^+/− neurons display moderately decreased Ca_V2.1 current and increased intrinsic excitability, but no gross defects in cortical neuron development. A previous study of human iPSC cortical neurons with CACNA1A haploinsufficiency likewise found an increase in intrinsic excitability, potentially attributable, in part, to altered gene expression of particular K⁺ channels.³⁴ Ca_V2.1 is known to couple functionally to calcium-activated small- and large-conductance K⁺ channels (SK and BK channels, respectively), potentially providing a mechanism by which mutant channels impact neuronal excitability.^35,36

Heterozygous D1585N knock-in mice recapitulate patient symptoms of ataxia and cognitive deficits

To model disease in a whole organism, we generated a knock-in carrying Ca_V2.1 D1585N mutation (homologous to human D1634N) in C57BL/6J mice (Supplementary Fig. 3). Animals were crossbred with C57BL/6J WT mice to diversify the gene pool and expand the heterozygous (HET) colony. Progeny from breeding heterozygous D1585N mice were expected to follow Mendelian ratios. However, homozygous D1585N^+/+ mice exhibited lethality within 24 h of birth, resulting in 57% D1585N^+/− and 43% WT surviving to postnatal Day 21 (male WT:HET = 30%:70%; female WT:HET = 59%:41%). Progeny from crossing WT and D1585N heterozygous mice were born and matured to postnatal Day 21 at 56% WT and 44% D1585N^+/− (male WT:HET = 55%:45%; female WT:HET = 57%:43%).

We compared WT and heterozygous littermates in a panel of behavioural tests. Initially, we performed a beam test to assess balance and motor coordination.²² WT Ca_V2.1 animals crossed the beam faster (velocity = 10.85 ± 1.04 cm/s, n = 11) than D1585N^+/− mice (velocity = 6.61 ± 0.55 cm/s, n = 12, P = 0.0026, Student’s unpaired t-test; Fig. 4A).

Figure 4.

Heterozygous D1585N^+/− mice display a range of behavioural deficits. (A) Beam test measure of velocity of traversing a 6.35-mm-wide beam for wild-type (WT; black, n = 11) and D1585N^+/− mice (red, n = 12) (P = 0.0026, Student’s unpaired t-test). (B) Latency to fall off a rotating rotarod for WT (black, n = 11) and D1585N^+/− mice (red, n = 12) (two-way ANOVA, Sidak’s multiple comparison test: trial 3, P = 0.0091; trial 4, P = 0.0127; trial 5, P = 0.0017; and trial 6, P = 0.0008). (C) Catwalk experiments for gait measurements for WT (black, n = 11) and D1585N^+/− mice (red, n = 12). (P < 0.001, two-way ANOVA). (D) Novel arm Y-maze test for WT (black, n = 11) and D1585N^+/− mice (red, n = 12) (P < 0.0005, two-way ANOVA). (E) Fear conditioning test for WT (black, n = 11) and D1585N^+/− mice (red, n = 12). (F) Top: Image of a D1585N^+/− mouse implanted with an electrode. Middle: Exemplar EEG trace for a D1634N^+/− mouse, showing prevalence of spike wave discharges (SWDs). Bottom: Occurrence and incidence of SWDs in WT (one of six animals) and D1585N^+/− mice (seven of eight animals; SWD frequency = 3–20/h; SWD duration = 0.5–2 s). ^#P < 0.05 and *P < 0.0005, Student’s two-tailed unpaired t-test.

We next used a rotarod to probe motor coordination and motor learning skills.²³ Animals were put on a rotating rod, which increased in speed linearly from 5 to 40 rpm over 5 min. Latency to fall was recorded for each mouse over 2 days. Both WT and D1585N^+/− mice displayed similar latency to fall in the first two trials, as animals adapted to the rotarod (Fig. 4B). However, the groups diverged from trials 3 to 6 (two-way ANOVA, Sidak’s multiple comparison test: trial 3, P = 0.0091; trial 4, P = 0.0127; trial 5, P = 0.0017; and trial 6, P = 0.0008), wherein WT animals displayed an increased latency to fall, whereas D1585N^+/− did not, reflecting a significant deficit in motor learning.

In open field tests, WT and D1585N^+/− mice showed no apparent differences in movement patterns, deficits in walking speed, or anxiety behaviour. There were several disparities in gait between WT and D1585N^+/− mice as determined by Catwalk assessment; WT mice had a significantly shorter stride length (P < 0.0001, two-way ANOVA) but faster swing speed compared with D1585N^+/− mice (Fig. 4C).

We used a Y-maze test to compare short-term memory between WT and D1585N^+/− mice.²⁴ When permitted to alternate spontaneously among the three arms of the Y-maze, neither WT nor D1585N^+/− mice displayed a preference for any arm. The lack of difference in spontaneous alternation in the Y-maze test suggests that D1585N^+/− mice have an intact working memory and, therefore, relatively unimpaired prefrontal cortical functions.²⁵ To probe specifically for short-term spatial reference memory, we applied the novel arm test, in which one arm of the Y-maze was closed off during a 10 min training period.²⁴ Both WT and D1585N^+/− mice displayed a preference for the ‘novel’ arm that was initially closed off, indicating an unimpaired spatial memory and suggesting adequate hippocampal function in mutant mice (Fig. 4D).

We also assessed associated learning and memory in mice using a contextual fear conditioning test.²⁶ In the pretraining period, D1585N^+/− mice displayed a significantly higher percentage of baseline freezing compared with WT (P = 0.0201; Fig. 4E). Fear conditioning training was induced by playing a pure tone (5 kHz, 80 dB) for 30 s (conditioning stimulus; CS), with the last 2 s accompanied by a shock to the feet. Both WT and D1585N^+/− mice displayed a significant increase in freezing in response to CS alone post-training (Fig. 4E). The next day, both WT and D1585N^+/− mice, when placed in the chamber with the same context as the previous day, still exhibited a similarly elevated percentage of freezing even without the CS or shock (Fig. 4E). On the final day, the features of the chamber were masked in order to change the context. WT mice placed in the masked chamber displayed a significantly lower baseline (pre-cue) percentage freezing (P = 0.0002). In contrast, D1585N^+/− mice a maintained an increased level of baseline freezing that was similar to when they were placed in the undisguised chamber, suggesting heightened anxiety (Fig. 4E). In the disguised chamber, both WT and D1585N^+/− mice responded to CS with a substantial increase in percentage freezing behaviour (Fig. 4E).

Finally, we implanted electrodes in six WT and eight D1585N^+/− mice and measured EEGs over a period of 24 h. Seven of eight D1585N^+/− mice displayed SWDs, albeit at frequencies ranging from 3 to 20 events/h, and lasting 0.5–2 s (Fig. 4F). Animals showed motion arrest with SWDs consistent with prevalence of absence seizures. In contrast, we observed only one of six WT mice showing SWDs or freezing behaviour over the recording period (P < 0.01, χ² test).

D1585N^+/− mice display abnormal electrophysiological properties in cerebellar Purkinje neurons

Ca_V2.1 is dominant in the cerebellum, which is crucial for motor control and learned movement. Accordingly, we compared electrophysiological properties of cerebellar Purkinje neurons from WT and D1585N^+/− mice. Ca_V2.1 immunostaining between WT and mutant cerebellar slices was similar, suggesting that overall protein expression levels are unchanged (Supplementary Fig. 4). We used whole-cell patch clamp to compare biophysical properties of Ca_V2.1 in acutely dissociated Purkinje neurons isolated from 11–20-day-old WT and D1585N^+/− mice, respectively (Fig. 5A). Purkinje neurons were easily identifiable by their pear shape with a short stump.¹⁶ WT mouse Purkinje neurons displayed robust whole-cell currents that were almost completely blocked by 500 nM ω-agatoxin, indicating that they are carried predominantly by Ca_V2.1 (Fig. 5B). The I–V curves indicated that peak whole-cell current density (I_peak = −146.4 ± 13.2 pA/pF, n = 13) occurred at −15 mV test pulse (Fig. 5C). Ca_V2.1 currents from D1585N^+/− mouse Purkinje neurons displayed a lower amplitude compared with WT (Fig. 5B and C); Ca_V2.1 I–V curves from D1585N^+/− mouse Purkinje neurons exhibited a significantly reduced peak current (I_peak = −102.3 ± 14.9 pA/pF, n = 18, P < 0.05, Mann–Whitney U-test) and right-shifted voltage dependence of activation (V_½ = −17.2 ± 2.6 mV for D1585N^+/− versus V_½ = −28.1 ± 2.0 mV for WT, P < 0.05, Mann–Whitney U-test; Fig. 5C).

Figure 5.

Impact of heterozygous D1585N mutation on electrical properties in Purkinje neurons and cerebellum slices. (A) Schematic of Purkinje cell whole-cell patch clamp. (B) Top: Exemplar whole-cell Ba²⁺ current traces evoked with 300 ms voltage steps from −50 to +60 mV in +10 mV increments from Purkinje neurons acutely dissociated from wild-type (WT; black) and D1585N^+/− (HET; red) mouse cerebellum. Bottom: Recordings from the same cells after addition of 500 nM ω-agatoxin to block Ca_V2.1 currents. (C) Population I–V curves for WT (black circles, n = 13) and D1585N^+/− (red circles, n = 18) Purkinje cells. (*P < 0.05, Mann–Whitney U-test). (D) Schematic diagram of current-clamp recordings of Purkinje cells in cerebellum slices (bottom) and the voltage-ramp protocol (top). (E) Top: Exemplar types of action potential (AP) trains from Purkinje cells in response to voltage-ramp depolarization protocol: Na⁺-only APs (left), Ca²⁺-only APs (middle) and mixed Na⁺ and Ca²⁺ APs (right). Bottom: Same APs as on the top on an expanded time scale to facilitate visual discrimination of individual spikes. (F) Frequency distribution of types of ramp-evoked AP trains in cerebellum slices from WT and D1585N^+/− mice (P < 0.05 for χ² test). (G) Exemplar miniature excitatory postsynaptic currents (mEPSCs) recorded from Purkinje neurons in cerebellum slices from WT (black trace) and D1585N^+/− (red trace) mice (P < 0.05, Student’s unpaired t-test). (H) Frequency of mEPSCs (P < 0.05, Student’s unpaired t-test with Welch’s correction. (I) Amplitude of mEPSCs (P = 0.62, Student’s unpaired t-test). (J) Schematic diagram of two-electrode protocol. Purkinje cells were clamped with a patch pipette, and a stimulating electrode was used to excite parallel fibres. (K) Exemplar paired pulse facilitation recordings from Purkinje neurons in cerebellum slices from WT (black trace) and D1585N^+/− (red trace) mice. (L) Population paired pulse ratio for WT (black) and D1585N^+/− (red). *P < 0.05, Student’s unpaired t-test.

Next, we compared electrical properties of Purkinje neurons in cerebellar slices obtained from WT and D1585N^+/− mice. Cerebellar slice Purkinje neurons were patched in whole-cell configuration, and APs were elicited in current-clamp mode using a ramp protocol (Fig. 5D). We observed three types of excitability behaviours: neurons that displayed persistent APs with an upstroke generated solely by Na_V channels (Na AP); those that showed exclusively slow action potentials with an upstroke generated by Ca_V2.1 (Ca AP); and neurons that exhibited both Na⁺ and Ca²⁺ APs (Na/Ca AP) (Fig. 5E and F). Purkinje neurons from WT and D1585N^+/− mice displayed striking differences in relative prevalence of the three types of excitability patterns (P < 0.05, χ² test; Fig. 5F). In D1585N^+/− cerebellar slices, 31% of Purkinje neurons displayed only Na Aps, whereas this pattern was absent in WT neurons (0 of 11 neurons; Fig. 5F). In contrast, the majority of WT Purkinje neurons (64%) exhibited solely Ca APs, whereas this pattern was the nadir (19%) for D1585N^+/− Purkinje neurons. Finally, WT and D1585N^+/− Purkinje neurons displayed a comparable fraction of mixed Na/Ca APs at 36% and 50%, respectively (Fig. 5F). In a complementary current-clamp protocol, Purkinje cells were initially hyperpolarized to −90 mV to remove inactivation of voltage-gated Na⁺ and Ca²⁺ channels, then depolarized. Among cells capable of generating high-frequency Na⁺ APs, 75% of WT neurons exhibited Ca²⁺ APs, compared with only 14% of D1585N^+/− neurons (P < 0.05, χ² test; Supplementary Fig. 5).

We next measured mEPSCs in Purkinje cells in the presence of tetrodotoxin and bicuculline (Fig. 5G). Compared with WT, D1585N^+/− mouse Purkinje neurons displayed a higher frequency of mEPSCs (17 ± 2/min, n = 14 versus 31 ± 5/min, n = 16, P < 0.05, Student’s t-test) with no change in amplitude (16.9 ± 0.2 pA, n = 17 versus 16.6 ± 0.2 pA, n = 16, P = 0.62, Student’s t-test; Fig. 5H–I), indicating an increased probability of spontaneous presynaptic vesicle release. We subsequently used a paired pulse protocol to assess synaptic properties of parallel fibre input to Purkinje neurons in WT and D1585N^+/− mice. We recorded from Purkinje cells in cerebellar slices in whole-cell mode while using a second electrode to stimulate parallel fibres at the molecular layer (Fig. 5J). Using a paired pulse stimulus protocol, in which two stimuli of similar amplitude, P1 and P2, are given 20 ms apart, WT mice displayed a larger EPSC in response to P2 compared with P1, a signature of paired pulse facilitation. D1585N^+/− mice displayed an augmented paired pulse facilitation compared with WT mice (2.9 ± 0.1, n = 10 versus 2.3 ± 0.1, n = 6, P = 0.016, Student’s t-test; Fig. 5K and L), suggesting a decrease in evoked presynaptic vesicle release probability at parallel fibre terminals to Purkinje cells in mutant mice. To determine whether the readily releasable pool of presynaptic vesicles at parallel fibre terminals is altered in D1585N^+/− cerebellum, we applied a 50 Hz stimulus train to evoke 30 EPSCs in Purkinje cells. Compared with WT slices, D1585N^+/− slices exhibited a delayed peak response (number of stimuli to reach peak response, WT: 5.4 ± 0.3, n = 5 versus HET: 9.4 ± 0.4, n = 5, P = 0.0042, Student’s t-test) and greater facilitation of normalized EPSC area under the curve (WT: 123 ± 4, n = 5 versus HET: 171 ± 8, n = 5, P = 0.028; Supplementary Fig. 6), suggesting that the readily releasable pool of synaptic vesicles is not impaired in mutant mice.³⁷

We also performed electrophysiological recordings in WT and D1585N^+/− cortical slices to enable similar cell type comparison to the human iPSC neuron recordings. Similar to the iPSC neurons, D1585N^+/− cortical slice neurons showed some evidence of a higher intrinsic excitability, such as a decreased threshold for firing APs, and also a higher AP frequency in response to small injected current amplitudes (<100 pA). However, at higher levels of input current (>100 pA) the WT cortical slice neurons attained higher AP firing frequencies compared with D1585N^+/− mice, indicating a biphasic effect (Supplementary Fig. 7).

D1585N^+/− mice exhibit elevated aberrations in cerebellar Purkinje cell morphology

Given the mild decrease in cerebellar size in the patient, we probed the size of the cerebellum and morphology of Purkinje cells in WT and mutant mice. We stained WT and D1585N^+/− mouse cerebellar slices and computed Purkinje cell dendritic area from images (Fig. 6A). We found no difference between WT and D1585N^+/− mice on this metric (1.00 ± 0.13 versus 0.97 ± 0.08, P = 0.87, Student’s unpaired t-test; Fig. 6B), suggesting minimal impact on overall cerebellum size.

Figure 6.

Purkinje cells of D1585N HET mice show dendritic degeneration. (A) Representative cerebellar slices immunostained for calbindin D-28K and GCaMP6f from control (Pcp2^Cre;Rosa26^GCaMP6f/+) and heterozygous (HET) mutant (Pcp2^Cre;Rosa26^GCaMP6f/+;CACNA1A^D1585N/+) mice. (B) Population total Purkinje cell dendritic area for wild-type (WT) and D1585N^+/− cerebella. (C and D) Representative images of biocytin-filled Purkinje cells, illustrating morphological differences between WT (C) and D1585N^+/− (D) mice. (E and F) Enlarged images demonstrating normal dendritic morphology in WT (E) and dendritic degeneration in D1585N^+/− (F) Purkinje cells. The magenta puncta in the overlay indicate round structures that are potentially degenerated dendrites. (G) Prevalence of abnormal dendritic segments in WT and D1585N^+/− Purkinje cells (P < 0.005, Student’s unpaired t-test). PC = Purkinje cells.

To gain more precise insights on cellular morphology, we evaluated the architecture of single Purkinje cells filled with biocytin dye via the patch pipette (Fig. 6C and D). WT Purkinje cells illuminated by this method displayed the characteristic intricate dendritic field morphology, with abundant branching and spines, typically associated with this cell type (Fig. 6C). In D1585N^+/− mice, many Purkinje cells showed a morphology that was mostly indistinguishable from WT cells (Fig. 6D, left). However, we observed a significant fraction of cells that displayed clear-cut signs of neurodegeneration, characterized by diminished branching and emergence of globular structures in the dendritic field (Fig. 6D, right). For a more global perspective on putative neurodegeneration, we generated Pcp2^Cre;Rosa26^GCaMP6f/+ and Pcp2^Cre;Rosa26^GCaMP6f/+;CACNA1A^D1585N/+ mice, which express calcium-sensitive GCaMP6f protein in cerebellum Purkinje cells. Cerebellar slices from control Pcp2^Cre;Rosa26^GCaMP6f/+ mice stained with calbindin D-28K antibody displayed healthy-looking Purkinje cells, with minimal globular structures in the dendritic field (Fig. 6E). In contrast, Pcp2^Cre;Rosa26^GCaMP6f/+;CACNA1A^D1585N/+ mice displayed a significant increase in the prevalence of globular structures in the dendritic field, indicative of neurodegenerating Purkinje cells (482 ± 105, n = 6 versus 1091 ± 103/mm², n = 8, P = 0.0015, Student’s unpaired t-test; Fig. 6F and G). To assess potential differences in spontaneous firing activity of Purkinje cells between WT and D1585N^+/− cerebella globally, we performed GCaMP6f calcium imaging in cerebellar slices. Purkinje cell activity was categorized into three groups: continuous firing; intermediate firing; and resting. No significant difference in cell population distribution was found between WT and D1585N^+/− cerebella (Supplementary Fig. 8).

Discussion

Mutations in CACNA1A are associated with a broad spectrum of neurological symptoms. To develop effective new treatments, it is crucial to understand how distinct CACNA1A mutations affect Ca_V2.1 function, understand genotype–phenotype relationships and develop models that recapitulate essential features of the human disease. Here, we focus on a single de novo CACNA1A variant, D1634N, identified in a patient with ataxia, intellectual disability and epilepsy. We performed a series of experiments involving: (i) assessing the biophysical impact of the mutation on heterologously expressed Ca_V2.1; (ii) modelling disease in vitro using patient-derived iPSC neurons; and (iii) evaluating an equivalent heterozygous D1585N knock-in mouse.

Comparison of D1634N biophysical deficits with previously studied CACNA1A loss-of-function mutations

D1634N mutation shifted Ca_V2.1 activation gating by +25 mV and reduced single-channel P_o. Located in Ca_V2.1 DIV S3, D1634 provides a counter charge important for movement of DIV S4 gating charges (R residues) in response to membrane potential changes that control channel opening and closing.^30,31 Voltage-clamp fluorometry experiments indicate non-equivalence of the four Ca_V2.1 voltage sensor domains (VSDs) in channel gating: VSD1 showed two distinct voltage-dependent steps, with the second activation step coinciding with channel opening; VSD2 was voltage independent; and VSD3 and VSD4 both activated at negative potentials near the physiological resting potential, with VSD4 displaying a shallower voltage dependence than VSD3.³⁸ The right-shifted opening of D1634N suggests that movement of VSD1, which is directly linked to channel opening, is not independent of VSD4. We speculate that D1634N impairs voltage-dependent movement of VSD4, which adversely affects the second phase of VSD1 activation to right-shift voltage dependence of opening and depress single-channel P_o.

Because D1634N results in reduced whole-cell current, it is considered a LoF mutation. The LoF defect in D1634N is attributable to a right shift in voltage dependence of activation and decreased single-channel P_o. These characteristics are also evident in Ca_V2.1 currents recorded in D1585N^+/− mouse Purkinje neurons. How does D1634N compare to other previously characterized CACNA1A LoF mutations? Tottering (P601L) and leaner (splice site mutation) mouse mutations cause 40% and 60% decreases, respectively, in whole-cell current mediated by a reduced single-channel P_o, but no change in voltage dependence of activation.^39-41 Heterologous expression of rolling Nagoya mutation (R1262G), yields ionic currents with a strongly depressed amplitude (72% decrease) and a 10 mV right shift in voltage dependence of activation.⁴² Thus, the biophysical deficit in rolling Nagoya is most functionally similar to D1634N. Curiously, however, homozygous rolling Nagoya Purkinje cells display a relatively modest 24% decrease in Ca_V2.1 current amplitude.⁴² Heterologous expression of two CACNA1A mutations, G230V and I1357S, associated with developmental epileptic encephalopathy in humans indicated LoF mutations in whole-cell current amplitude (76% and 36% reduction, respectively) apparently mediated by decreased channel surface density.⁷

Comparison of D1585N^+/− Purkinje neurons with WT neurons

D1585N^+/− Purkinje cells displayed pronounced shifts in intrinsic excitability, including an increased reliance on Na⁺-mediated APs and substantially reduced Ca²⁺-dependent excitability during ramp- and step-induced depolarization. Similar altered responses were also reported in rolling Nagoya mutation (R1262G).⁴² Interestingly, spontaneous vesicle release probability was increased, whereas evoked release probability at parallel fibre–Purkinje cell synapses was reduced in mutant mice, as evidenced by enhanced paired pulse facilitation, as previously reported in rolling Nagoya (R1262G) and leaner mice.^43,44 This paradoxical finding indicates either that climbing fibre terminals innervating Purkinje cells contribute to the total increased vesicle release probability or that mEPSCs and evoked EPSCs triggered by parallel fibre activation originate from different synaptic vesicle pools. Morphological analyses also revealed clear neurodegenerative features. These morphological abnormalities, together with altered synaptic responses and cell firing properties, might contribute collectively to behavioural deficits in these mutant mice.

Behavioural impact of D1585N compared with other CACNA1A LoF mouse models

In humans, CACNA1A LoF mutations are typically associated with episodic ataxia type 2.^5,45 Nevertheless, CACNA1A haploinsufficiency has also been associated with cognitive impairment, autism and epileptic encephalopathy in humans.⁴⁶ Homozygous D1585N knock-in mice die within 24 h of birth, and D1585N^+/− mice display a constellation of behavioural deficits, including impaired motor skills, altered gait and absence epilepsy, that mimic symptoms and autosomal dominant nature of the human disease. This aspect of the D1585N knock-in mouse model differs importantly from several other well-studied CACNA1A LoF animal models, which do not reliably recapitulate the autosomal dominant nature of human disease. For example, patients carrying the CACNA1A F1406C mutation exhibited symptoms of episodic ataxia type 2; however, homozygous F1406C knock-in mice did not display a deficit in motor skills on rotarod test despite having a ∼60% decline in Ca_V2.1 current amplitude in isolated Purkinje cells (with no change in V_½ of activation).⁴⁷ Spontaneously arising CACNA1A LoF models (tottering, leaner, rolling Nagoya and rocker) have been widely studied and invaluable for linking diminished Ca_V2.1 dysfunction to a spectrum of ataxic and epileptic symptoms.^14,48 Similar to D1634N, tottering is attributable to a CACNA1A missense mutation and displays a phenotype of paroxysmal motor attacks, absence epilepsy and ataxia.¹⁴ However, in contrast to D1634N, the tottering symptoms are inherited in an autosomal recessive manner. Likewise, leaner, Rolling Nagoya and rocker mouse symptoms are also inherited in an autosomal recessive manner.⁴⁸ In other mouse models, conditional knockout of CACNA1A in distinct brain regions, including Purkinje cells (purky), rhombic-lip-derived neurons (quirky) and layer VI cortical neurons, have been used elegantly to show sufficiency of Ca_V2.1 elimination in discrete brain regions to recapitulate different disease symptoms, including ataxia and absence epilepsy phenotypes.^49-51 Nevertheless, these experimental models are also different from the autosomal dominant human phenotype.

Our findings indicate that D1634N increases intrinsic excitability in human iPSC-derived cortical neurons, murine cortical slice neurons with low current injections, and Purkinje neurons. Previous studies of leaner and tottering mice indicated that loss of Ca_V2.1 function leads to impaired precision of Purkinje cell firing attributable to compromised coupling to Ca²⁺-activated K⁺ channels (K_Ca).^52,53 Accordingly, either chronic perfusion of K_Ca activator 1-ethyl-2-benzimidazolinone in cerebellum, or oral administration of a US Food and Drug Administration-approved structurally related compound, chlorpromazine, moderately improved baseline ataxia in tottering mice.^52,53 It will be important to test whether K_Ca activation can improve motor symptoms in D1585N^+/− mice in future experiments.

Implications for developing Ca_V2.1-targeted therapies for CACNA1A LoF diseases

CACNA1A LoF mutations can manifest in myriad ways including: (i) nonsense protein production (owing to deletions, frameshifts or premature stop codons); (ii) impaired trafficking; (iii) gating deficits (decreased P_o, enhanced inactivation and right-shifted activation); or (iv) diminished permeation (decreased conductance). Whole-cell ionic current (I) is defined by I = N × i × P_o, where N is total number of surface channels, i is unitary current amplitude, and P_o is single-channel open probability. Understanding which parameters a CACNA1A mutation downregulates to decrease I is crucial for rational development of Ca_V2.1-targeted therapies.

The simplest case might involve nonsense mutations, where disease is caused by CACNA1A haploinsufficiency (decrease in N). Here, approaches that double WT channel abundance would be most appropriate. This could potentially be accomplished by increasing gene expression or translation or by prolonging protein half-life. Targeted protein stabilization approaches reliant on selectively stripping ubiquitin from a target protein offer a potential approach to enhance WT Ca_V2.1 abundance.^54,55 Circumstances in which CACNA1A disease is associated with heterozygous genotypes, as with D1634N, present added complexity, in that mutant channel is present and must be either corrected or eliminated. A solution for D1634N would be to identify a molecule that normalizes V_½ of activation but is inert on the WT channel. An alternative strategy would be to eliminate mutant protein while boosting WT channel expression. Possible methods to eliminate mutant channels include allele-specific small interfering RNA knockdown,⁵⁶ or antisense oligonucleotides that force exon switching in cases where the mutation is present on an alternatively spliced exon.⁵⁷ Because both LoF and gain-of-function mutations in Ca_V2.1 cause disease, solutions for correcting or normalizing activity of a mutant should ideally have no effect on the WT channel. For CACNA1A mutations, this circumstance emphasizes why it is important to have heterozygous animal models for use in preclinical testing of potential therapeutics that could be translated into humans.

Our longitudinal studies also emphasize unique and complementary advantages of the three model systems, which together provide comprehensive insights into the molecular, cellular and in vivo neurological basis and manifestation of a CACNA1A mutation. Heterologous expression enabled in-depth study of the biophysical impact of D1634N and its classification as a LoF mutation. Patient-specific iPSC neurons and isogenic controls yielded the insight that the LoF CACNA1A mutation leads to a neuronal hyperexcitability. Both systems are amenable to high-throughput discovery of corrective molecules. Finally, knock-in mice provide information on how the impact of the mutation across multiple brain regions integrates into behavioural abnormalities. Such animal models are necessary for testing whether potential corrective molecules can prevent and/or reverse disease manifestations.

Data availability

We declare that all data supporting the findings of this study are available within the paper and its Supplementary material.

Funding

This work was supported by grants R01-NS126850, RO1-HL121253 and P01-HL164319 from the National Institutes of Health (to H.M.C.); and a grant from the CACNA1A Foundation (to H.M.C.). We are grateful to the family for funding the generation of the knock-in mouse model and for support that enabled initiation of this project. Confocal images were collected in the Columbia University HICCC Confocal and Specialized Microscopy Shared Resource, supported by NIH (P30 CA013696).

Competing interests

H.M.C. is a scientific co-founder and on the Scientific Advisory Board of two start-ups, Stablix, Inc. and Flux Therapeutics, pursuing targeted protein stabilization therapeutics.

Supplementary material

Supplementary material is available at Brain online.