The patients were recruited at the Epilepsy Center of the Second Affiliated Hospital of Guangzhou Medical University and the Affiliated Yuebei People’s Hospital of Shantou University Medical College. Clinical phenotypes of epileptic seizures and epilepsy syndromes were assessed following the criteria of the Commission on Classification and Terminology of the International League Against Epilepsy (ILAE) (1981, 1989, 2001, 2010, 2017). Partial epilepsy was used to denote cases with partial seizures and EEG features of idiopathic epilepsy including shift, bilateral or multiple focal discharge. Generalized epilepsy was diagnosed on the basis of typical generalized seizures, such as absence, myoclonic, atonic, and generalized tonic-clonic seizures, supported by the results of generalized discharges on EEG. Participants with acquired causes like brain malformation, infection or metabolic disorders were excluded. We collected the comprehensive clinical materials, including gender, current age, seizure onset age, seizure type and frequency, outcome, response to antiepileptic drugs (AEDs), family history, and results from general and neurological examinations. Brain CT or MRI scans were performed to detect abnormalities in brain structure. Long-term (24 h) video EEGs that included open-close eyes test, hyperventilation, intermittent photic stimulation and sleep recording, were performed and the results were double-reviewed by two qualified researchers. All individuals enrolled were unrelated ethnic Han Chinese with four Han Chinese grandparents, and were born to non-consanguineous Chinese parents. All of subjects were followed up for at least 1 year at epilepsy centers. A total of 468 cases were recruited, including 318 cases with partial epilepsy and 150 cases with generalized epilepsy. Additionally, we recruited 296 healthy Chinese volunteers as a normal control group as our previous report (Wang et al., 2018, 2020, 2021).

All procedures in this study involving human participants have been approved by the ethics committee of the Second Affiliated Hospital of Guangzhou Medical University. Written informed consents have been obtained from all participants or their parents/legal guardians in the case of child or those with intellectual disability.

The genomic DNAs were extracted from the peripheral blood samples of the probands, their parents, and available family members using the FlexiGene DNA kit (Qiagen). Trio-based whole-exome sequencing was performed on an Illumina HiSeq 2000 sequencing platform as previously reported (Shi et al., 2019; Wang et al., 2021). To obtain high-quality reads, the massive parallel sequencing was performed with more than 125 times average depth and more than 98% coverage in the capture region of the chip. The original read data were aligned to the Genome Reference Consortium Human Genome build 37 (GRCh37) using Burrows-Wheeler alignment (BWA) with default parameters. Variant calling and quality filtration were conducted using the Genome Analysis Toolkit (DePristo et al., 2011).

To derive the whole candidate pathogenic variants in each trio, we adopted a case-by-case analytical approach as previously described (Zhou et al., 2018; Wang et al., 2021). Initially, we removed the common variants presenting a minor allele frequency ≥ 0.005 in the Genome Aggregation Database (gnomAD).¹ We then prioritized potentially pathogenic variants, including frameshift, nonsense, canonical splice site, initiation codon, and missense variants predicted as being damaging in silico tools (VarCards).² We screened CACNA1A mutations with origination of explainable for genetic diseases, including de novo mutation, mutation with segregations, and homozygous/compound heterozygous mutation. Additionally, I-Mutant 3.0 program was applied³ to predict the effect of CACNA1A missense variants on protein stability, which was indicated by free energy change (DDG). Negative DDG value means that the mutated protein possesses less stability and vice versa. Eventually, the pathogenicity of the identified CACNA1A mutations was evaluated by American College of Medical Genetics and Genomics (ACMG) scoring (Richards et al., 2015). Polymerase chain reaction and sanger sequencing was performed to validate the identified potential pathogenic variations by using ABI 3730 sequencing platform (Applied Biosystems, Foster City, CA, United States). All CACNA1A mutations identified in this study were annotated to reference transcript NM_001127222.

All CACNA1A mutations and related phenotypes were systematically retrieved from the professional edition of Human Gene Mutation Database (HGMD)⁴ and the PubMed database⁵ up to December 2021. To explore the relationship between genotype and phenotype, we divided the CACNA1A mutations into two categories, destructive (null) and missense mutations. Null mutations were those causing gross malformation of the gene/protein and leading to loss of function and haploinsufficiency (Richards et al., 2015; Liu et al., 2020), including truncating mutations (non-sense and frameshifting), splice site mutations (canonical ± 1 or 2), and mutations at initiation codon or with single/multi exon deletion. To facilitate analyzing the effect of molecular sub-regional on epileptic phenotypes, we collected the information on the distribution of CACNA1A missense mutations in various regions of the Cav2.1 channel. Considering that epileptic phenotype severity may be associated with mutation origin (de novo or inherited), we also collected the inheritance information of epilepsy associated CACNA1A mutations.

Statistical analyses were performed in GraphPad Prism version 8.00. A two-tailed Fisher’s exact test was used to compare the frequencies of null mutation, missense mutation, and de novo mutation between different phenotype groups. The recessive CACNA1A variants burden was also analyzed according to a recent research (Martin et al., 2018). The cutoff value for statistical significance is 0.05.

Twelve CACNA1A mutations were identified in ten unrelated cases of epilepsy, including four null mutations (c.2963_2964insG/p.Gly989Argfs*78, c.3089 + 1G > A, c.4755 + 1G > T, and c.6340-1G > A) and eight missense mutations (c.203G > T/p.Arg68Leu, c.3233C > T/p.Ser1078Leu, c.3965G > A/p.Gly1322Glu, c.4891A > G/p.Ile1631Val, c.5032C > T/p.Arg1678Cys, c.5393C > T/p.Ser1798Leu, c.5978C > T/p.Pro1993Leu, and c.6061G > A/p.Glu2021Lys). The Ser1798Leu mutation has previously been described as a de novo mutation in a case of EA2 (Ohba et al., 2013), whereas the remained 11 mutations have not been reported and were novel findings. Four of the missense mutations (c.3233C > T/p.Ser1078Leu&c.6061G > A/p.Glu2021Lys and c.4891A > G/p.Ile1631Val&c.5978C > T/p.Pro1993Leu) constituted two pairs of compound heterozygous mutations; the remaining four missense mutations and four null mutations were de novo (Figures 1, 2 and Table 1). The eight de novo mutations were neither in gnomAD populations nor in our 296 normal control subjects and were evaluated as pathogenic or likely pathogenic mutations according to the criteria of ACMG (Table 2). The two pairs of the compound heterozygous mutations were absent in our 296 normal control subjects and present in gnomAD with an extremely low frequency (Table 2). When the recessive variants burden was analyzed, a statistically significant difference of the compound heterozygous CACNA1A mutations in this cohort was observed comparing the excepted number by chance in the controls of East Asian and all populations in the Exome Aggregation Consortium (P = 7.30 ×10^–4, P = 2.53 × 10^–4) (Martin et al., 2018).

The c.2963_2964insG/p.Gly989Argfs*78 mutation was considered to potentially pathogenic by yielding a truncated transcript that gave rise to a non-functional Cav2.1 protein or haploinsufficiency. The three canonical splice site mutations (c.3089 + 1G > A, c.4755 + 1G > T, and c.6340-1G > A) could destroy the original splice donor or acceptor site that generally resulted in the skipping of the single exon or multiexon with consequent translational frameshift. All of the eight missense mutations were predicted to be damaging by at least one of the commonly used in silico prediction tools (Table 2). The amino acid sequence alignment indicated that Arg68Leu, Gly1322Glu, Arg1678Cys, Ser1798Leu, Pro1993Leu, and Glu2021Lys were located at residues that are highly conserved in various species; Ile1631Val are highly conserved in vertebrates but less so in lower animals (Figure 2B). The Ser1078Leu was located at a less conserved site but was predicted to be conserved by GERP (score = 5.17), phyloP (score = 5.429), and SiPhy (score = 17.443). Furthermore, I-Mutant 3.0 program showed that Arg68Leu, Gly1322Glu, Ile1631Val, and Arg1678Cys mutants have a strong influence on protein stability (Table 2).

None of the 10 patients had pathogenic or likely pathogenic mutations in the genes known to be associated with epileptic phenotypes (Wang et al., 2017) except CACNA1A mutations.

In this study, we identified CACNA1A mutations in 10 unrelated cases. The seizure onset age of the ten cases ranged from 3 months to 11 years old, with a median age of onset of 3.5 years. Eight of the cases were diagnosed as partial epilepsy, including six cases with de novo missense/null mutations and two with compound heterozygous missense mutations. They had simple/complex partial seizures or secondarily generalized tonic-clonic seizures. They all had focal epileptic discharges with normal backgrounds or trends of generalization especially during sleep (Figures 3A,C–E). The remaining two cases were diagnosed as childhood absence epilepsy and carried two de novo canonical splice site mutations. They experienced frequent absence seizures and detected ictal or interictal generalized 3 HZ spike and slow waves on EEG recordings (Figure 3B). A patient (case 2) also carried a canonical splice site mutation (c.3089 + 1G > A) and was diagnosed as partial epilepsy, her EEGs presented bilateral occipital epileptic discharges with features of idiopathic epilepsies (Figure 3A). These findings indicated that splice site mutations of CACNA1A were potentially associated with generalized epilepsies or idiopathic epilepsies. All of the ten cases were ultimately seizure-free after antiepileptic treatment, although frequent epileptic seizures were observed in four cases (Cases 3, 4, 6, and 9; Table 1). One patient (Case 3) had antecedent febrile seizures. Three patients had mild intellectual abnormalities (Cases 5, 6, and 8; Table 1). All the ten cases were born by normal delivery, and the brain MRI findings were normal. No ataxia or migraine were observed in any of them.

To explore the correlation between genotype and phenotype, we systematically reviewed all reported CACNA1A mutations. Previously, 312 mutations have been reported, including 115 null mutations, 183 missense mutations, 10 in-frame insertion/deletion mutations, and 4 (CAG)n dynamic mutations. These mutations were associated with a variety of clinical phenotypes that included EA2, FHM1, SCA6, CSVD (cerebral small vessel disease), and epilepsies. EA2 group present a significantly higher frequency of null mutation than the groups of epilepsy (P = 7.92 × 10^–5), FHM1 (P = 2.85 × 10^–5), SCA6 (P = 5.69 × 10^–3), or CSVD (P = 3.77 × 10^–6) (Figure 4A).

Cav2.1 encoded by CACNA1A, contains four homologous domains (DI-DIV) with six helical transmembrane segments (S1–S6). The S4 segments of each repeat serve as actual voltage sensors while S5 and S6 segments together with S5-S6 loop of each repeat form the channel pore (Figure 2A; Striessnig, 2021). In the present study, de novo missense mutations, except the Arg68Leu, were all located at pore region or near the voltage sensor region while compound heterozygous missense mutations were mainly located at linker region or C-terminal. We analyzed the data together with that from literature (Figure 2B). Previous studies have shown CACNA1A mutations potentially have an association with developmental abnormalities (Allen et al., 2013; Damaj et al., 2015; Epi4k Consortium, 2016). In this cohort, three cases of partial epilepsy also have ID. We then analyzed the epilepsies with ID and those without ID (Supplementary Table 1). No statistical difference in the frequency of missense was observed between the two epilepsy sub-groups (Figure 4B). However, it was found that missense mutations in the epilepsy with ID were more frequently located in the pore region than those in the epilepsy without ID (P = 1.67 × 10^–4) (Figure 4C), suggesting a molecular sub-region effect. Moreover, the cases in the epilepsy with ID group had a higher percentage of de novo mutations than those in the epilepsy without ID (P = 1.92 × 10^–3) (Figure 4D), suggesting a potential correlation between epileptic phenotype severity and mutation origins.