CACNA1I gain-of-function mutations differentially affect channel gating and cause neurodevelopmental disorders

Abstract T-type calcium channels (Cav3.1 to Cav3.3) regulate low-threshold calcium spikes, burst firing and rhythmic oscillations of neurons and are involved in sensory processing, sleep, and hormone and neurotransmitter release. Here, we examined four heterozygous missense variants in CACNA1I, encoding the Cav3.3 channel, in patients with variable neurodevelopmental phenotypes. The p.(Ile860Met) variant, affecting a residue in the putative channel gate at the cytoplasmic end of the IIS6 segment, was identified in three family members with variable cognitive impairment. The de novo p.(Ile860Asn) variant, changing the same amino acid residue, was detected in a patient with severe developmental delay and seizures. In two additional individuals with global developmental delay, hypotonia, and epilepsy, the variants p.(Ile1306Thr) and p.(Met1425Ile), substituting residues at the cytoplasmic ends of IIIS5 and IIIS6, respectively, were found. Because structure modelling indicated that the amino acid substitutions differentially affect the mobility of the channel gate, we analysed possible effects on Cav3.3 channel function using patch-clamp analysis in HEK293T cells. The mutations resulted in slowed kinetics of current activation, inactivation, and deactivation, and in hyperpolarizing shifts of the voltage-dependence of activation and inactivation, with Cav3.3-I860N showing the strongest and Cav3.3-I860M the weakest effect. Structure modelling suggests that by introducing stabilizing hydrogen bonds the mutations slow the kinetics of the channel gate and cause the gain-of-function effect in Cav3.3 channels. The gating defects left-shifted and increased the window currents, resulting in increased calcium influx during repetitive action potentials and even at resting membrane potentials. Thus, calcium toxicity in neurons expressing the Cav3.3 variants is one likely cause of the neurodevelopmental phenotype. Computer modelling of thalamic reticular nuclei neurons indicated that the altered gating properties of the Cav3.3 disease variants lower the threshold and increase the duration and frequency of action potential firing. Expressing the Cav3.3-I860N/M mutants in mouse chromaffin cells shifted the mode of firing from low-threshold spikes and rebound burst firing with wild-type Cav3.3 to slow oscillations with Cav3.3-I860N and an intermediate firing mode with Cav3.3-I860M, respectively. Such neuronal hyper-excitability could explain seizures in the patient with the p.(Ile860Asn) mutation. Thus, our study implicates CACNA1I gain-of-function mutations in neurodevelopmental disorders, with a phenotypic spectrum ranging from borderline intellectual functioning to a severe neurodevelopmental disorder with epilepsy.

present in the heterozygous state in the healthy parents (homozygous or compound heterozygous in patient). Exonic and splice variants were then prioritized by pathogenicity assessment using assessment using multiple in silico tools (CADD, REVEL, M-CAP, Human Splicing Finder 3.1, NetGene2-Server, and Berkeley Drosophila Genome Project-Database) (Harms et al., 2020;Schneeberger et al., 2020).
Sequence validation of the CACNA1I variant c.2579T>A (NM_021096.3) was performed in leukocyte-derived DNA of patient 1 and parents by Sanger sequencing and confirmed the missense variant to be de novo in the patient.

Patients 2, 3 and 4
We performed WES in the two affected sibs and their parents. WES libraries were prepared using SeqCap EZ MedExome (Roche Sequencing, Pleasanton, CA) and sequenced on a HiSeq2500 platform. Read alignment to GRCh37 (hg19) and variant calling were done with a pipeline based on BWA-MEM0.7 and GATK 3.3.0. The median coverage of the captured target region was at least 100x. Variant annotation and prioritizing were done using Cartagenia Bench Lab NGS (Agilent Technologies) as described previously (Bouman et al., 2018;Houweling et al., 2019).

Patient 5
Clinical exome sequencing was performed at the Division of Genomic Diagnostics at Children's Hospital of Philadelphia using DNA extracted from peripheral blood from patient 5 and unaffected biological parents. Library preparation was performed using Agilent's SureSelect XT protocol and target enrichment with Agilent's Clinical Research Exome version 1. Sequencing was performed using an Illumina HiSeq 2500 with 100 bp paired-end reads.
Alignment of sequencing reads was to the hg19 genome build. Sequencing data analysis was performed using a custom bioinformatics filtration pipeline as previously described (Gibson et al., 2018).

Patient 6
Using genomic DNA from patient 6 and mother, the exonic regions and flanking splice junctions of the genome were captured the IDT xGen Exome Research Panel v1.0. Massively parallel (NextGen) sequencing was done on an Illumina system with 100 bp or greater pairedend reads. Reads were aligned to human genome build GRCh37/UCSC hg19, and analyzed for sequence variants using a custom-developed analysis tool. Additional sequencing technology and variant interpretation protocol has been previously described (Retterer et al., 2016). The general assertion criteria for variant classification are publicly available on the GeneDx ClinVar submission page (http://www.ncbi.nlm.nih.gov/clinvar/submitters/26957/).

Structure modelling
Homology modeling was performed using Rosetta and MOE (Molecular Operating Environment, version 2019.08, Molecular Computing Group Inc., Montreal, Canada).
Additionally, ab initio Rosetta (Rohl et al., 2004) was used to generate structures for loops that were not resolved in the original Cav3.1 α1-subunit template (Zhao et al., 2019). The structures for the mutants were derived from the WT model by replacing the mutated residue and carrying out a local energy minimization using MOE. The C-terminal and N-terminal parts of each domain were capped with acetylamide (ACE) and N-methylamide to avoid perturbations by free charged functional groups. The structure model was embedded in a plasma membrane consisting of POPC (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine) and cholesterol in a 3:1 ratio, using the CHARMM-GUI Membrane Builder (Jo et al., 2009;Lee et al., 2019). Water molecules and 0.15 M KCl were included in the simulation box. Energy minimizations of WT and mutant structures in the membrane environment were performed. The topology was generated with the LEaP tool of the AmberTools20 (Case et al., 2018), using force fields for proteins and lipids, ff14SBonlysc and Lipid14 (Dickson et al., 2014), respectively. The WT and mutant structures were gradually heated from 0 to 300 K, keeping the lipids fixed, and then equilibrated over 1 ns. Then molecular dynamics simulations were performed for 5 ns, with time steps of 2 fs, at 300 K and in anisotropic pressure scaling conditions. Van der Waals and short-range electrostatic interactions were cut off at 10 Å, whereas long-range electrostatics were calculated by the Particle Mesh Ewald (PME) method. MOE and Pymol (The PyMOL Molecular Graphics System, Version 2.0 Schrödinger, LLC.) was used to visualize relevant interactions and point out differences in the WT and mutant structures.

Expression plasmids
The coding sequence of the human CaV3.3 subunit (Genebank ID AF393329) was obtained from a1Ic-HE3-pcDNA3, which was a gift from Edward Perez-Reyes (Addgene plasmid # 45810; http://n2t.net/addgene:45810; RRID:Addgene_45810) (Gomora et al., 2002). The coding sequence of CaV3.3 was inserted downstream of the GFP in the GFP-CaV1.1 vector (Grabner et al., 1998) (nt 527-2723 for I860N and I860M or nt 2567-4824 for I1306T and M1425I) was amplified by PCR with overlapping primers introducing each mutation in separate PCR reactions. For each mutation, the two separate PCR products were then used as templates for a final PCR reaction with flanking primers to connect the nucleotide sequences. These fragments were then digested with BamHI/AvrII (for I860N and I860M) or with AvrII/HindIII (for I1306T and M1425I) and ligated into the corresponding sites of GFP-CaV3.3, yielding GFP-CaV3.3-I860N, I860M, I1306T and M1425I. Note that the numbering for the reported I1306T and M1425I mutations differ from the actual position in the CaV3.3 splice variant used to construct the expression plasmids (I1271T and M1390I), which contains a shorter alternatively spliced exon in the I-II linker. For consistency we designated the constructs according to the numbering used for the disease variants I1306T and M1425I. Sequence integrity of all newly generated constructs was confirmed by sequencing (MWG Biotech, Martinsried, Germany).

Cell culture and transfections
HEK293T cells were cultured in Dulbecco´s modified Eagle medium (DMEM, Thermo Fisher Scientific) supplemented with 10% (v/v) fetal bovine serum (FBS; Merck) and penicillinstreptomycin (100 U/ml and 100 μg/ml, respectively; Thermo Fisher Scientific) and incubated at 37°C in a humidified atmosphere with 5% CO2. HEK293T cells were transfected with 1 μg of plasmid DNA on the day of plating, using FuGENE-HD transfection reagent (Promega) and cultured in DMEM overnight. The cells were used for experiments on the first or second day after transfection.
Chromaffin cells from 6 to 8-week-old male mice were obtained as described previously with slight modifications (Marcantoni et al., 2010;Calorio et al., 2019) and transfected by electroporation with plasmid DNA using Mouse Neuron Nucleofector Kit (Cat# VAPG-1001, Lonza Group Ltd) (Courel et al., 2008). To this end, the cell pellet was re-suspended in 300 µl of Supplemented Mouse Neuron Nucleofector Solution (MNNS) pre-warmed to RT. Per construct (CaV3.3 WT, I860M or I860N), 100 µl of the MNNS-cell suspension was used for nucleofection with 3 µg of DNA and program A-33 on Nucleofector TM 1 device (Amaxa). After nucleofection cells were immediately transferred into polystyrene tubes (Cat# 120160, Fisher Scientific) containing 500 μl of pre-conditioned RPMI 1640 medium (Cat# 42401018, Thermo Fisher Scientific) and allowed to recover for 20 min in a humidified incubator (37°C, 95% air and 5% CO2). 100 µl of cell suspension was placed into the center of a pre-coated 35 mm culture dish (Cat# 353001, Fisher Scientific) containing 200 µl of pre-conditioned culture medium. 1-2 h after plating, culture medium was filled up to 2 ml and medium was completely changed after 3 h to get rid of debris. Cells were then maintained in a humidified incubator (37°C, 95% air and 5% CO2) and used for patch-clamp experiments within 1-2 d after plating.

Voltage-clamp experiments
In the HEK293T cells calcium currents were recorded with the whole-cell patch-clamp technique in voltage-clamp mode using an Axopatch 200A amplifier (Axon Instruments). Patch pipettes (borosilicate glass; Science Products) had resistances between 1.8 and 4.5 MΩ when filled with (mM) 135 CsCl, 1 MgCl2, 10 HEPES, and 10 EGTA (pH 7.4 with CsOH). The extracellular bath solution contained (mM) 2 CaCl2, 165 choline-chloride, 10 HEPES, and 1 Mg-Cl2 (pH 7.4 with CsOH). Data acquisition and command potentials were controlled by pCLAMP software (Clampex version 10.2; Axon Instruments); analysis was performed using Clampfit 10.7 (Axon Instruments) and SigmaPlot 12.0 (SPSS Science) software. The currentvoltage dependence was fitted according to where Gmax is the maximum conductance of the L-type calcium currents, Vrev is the extrapolated reversal potential of the calcium current, V1/2 is the potential for half maximal conductance, and k is the slope. The conductance was calculated using G = (− I * 1000)/(Vrev − V), and its voltage dependence was fitted according to a Boltzmann distribution: Steady-state inactivation curves were fitted using a modified Boltzmann equation: where V1/2 is the half-maximal inactivation voltage and k is the inactivation slope factor.

Current-clamp experiments
The whole cell current-clamp recordings of isolated mouse chromaffin cells were performed in perforated-patch mode. The patch pipettes had a resistance between 1.8-4 MΩ when filled with (in mM): 10 NaCl, 10 KCl, 76 K2SO4, 1 MgCl2, 5 HEPES, adjusted to pH 7.35 with KOH and supplemented with amphotericin B (240 μg/m; Cat# Y0000005, Merck KGaA). To facilitate sealing, glass pipettes were dipped in amphotericin-free intracellular solution before being back-filled with solution containing amphotericin, which was light-protected and kept cold (0-4°C) during the experiment. The external bath solution contained (in mM): 140 NaCl, 3.6 KCl, 2 NaHCO3, 0.5 NaH2PO4, 0.5 MgSO4, 2.5 CaCl2, 5 HEPES, 5 glucose, adjusted to pH 7.4 with NaOH. Recordings were performed at room temperature (20-24 °C) using an EPC 10 amplifier (HEKA Elektronik) controlled by PatchMaster software (version 2.80, HEKA Elektronik). Data were sampled at 10 kHz and low-pass filtered at 2.9 kHz. Action potentials were recorded without injecting any current (spontaneous activity) and after clamping the resting membrane potential to -70 mV. Offline data analysis of AP parameters (Fig. 7) was performed using Clampfit software (version 10.7, Axon Instruments). Firing modes of chromaffin cells were characterized according to T-type channel mediated firing patterns of thalamic neurons (Cain and Snutch, 2010).

Computer model
Modelling was performed in the NEURON simulation environment (Hines and Carnevale, 1997) using the model for thalamic relay neurons (Destexhe et al., 1996) from the model database at Yale University (https://senselab.med.yale.edu/modeldb/). The electrophysiological properties of the Cav3.3 channels were modelled using Hodgkin-Huxley equations as described previously (Huguenard and Prince, 1992;Destexhe et al., 1996).  The functional impact of the identified variants was predicted by the Combined Annotation Dependent Depletion (CADD) tool, the Rare Exome Variant Ensemble Learner (REVEL) scoring system, and the Mendelian Clinically Applicable Pathogenicity (M-CAP) Score. CADD is a framework that integrates multiple annotations in one metric by contrasting variants that survived natural selection with simulated mutations. Reported CADD scores are phred-like rank scores based on the rank of that variant's score among all possible single nucleotide variants of hg19, with 10 corresponding to the top 10%, 20 at the top 1%, and 30 at the top 0.1%. The larger the score the more likely the variant has deleterious effects; the score range observed here is strongly supportive of pathogenicity, with all observed variants ranking above ~99% of all variants in a typical genome and scoring similarly to variants reported in ClinVar as pathogenic (~85% of which score >15) (Kircher et al., 2014). REVEL is an ensemble method predicting the pathogenicity of missense variants with a strength for distinguishing pathogenic from rare neutral variants with a score ranging from 0-1. The higher the score the more likely the variant is pathogenic (Ioannidis et al., 2016). M-CAP is a classifier for rare missense variants in the human genome, which combines previous pathogenicity scores (including SIFT, Polyphen-2, and CADD), amino acid conservation features and computed scores trained on mutations linked to Mendelian diseases. The recommended pathogenicity threshold is >0.025 (Jagadeesh et al., 2016). Genetic tolerance at the affected amino acid position in the protein was predicted by MetaDome (Wiel et al., 2019).