Multi-gene panel testing in Korean patients with common genetic generalized epilepsy syndromes

Genetic heterogeneity of common genetic generalized epilepsy syndromes is frequently considered. The present study conducted a focused analysis of potential candidate or susceptibility genes for common genetic generalized epilepsy syndromes using multi-gene panel testing with next-generation sequencing. This study included patients with juvenile myoclonic epilepsy, juvenile absence epilepsy, and epilepsy with generalized tonic-clonic seizures alone. We identified pathogenic variants according to the American College of Medical Genetics and Genomics guidelines and identified susceptibility variants using case-control association analyses and family analyses for familial cases. A total of 57 patients were enrolled, including 51 sporadic cases and 6 familial cases. Twenty-two pathogenic and likely pathogenic variants of 16 different genes were identified. CACNA1H was the most frequently observed single gene. Variants of voltage-gated Ca2+ channel genes, including CACNA1A, CACNA1G, and CACNA1H were observed in 32% of variants (n = 7/22). Analyses to identify susceptibility variants using case-control association analysis indicated that KCNMA1 c.400G>C was associated with common genetic generalized epilepsy syndromes. Only 1 family (family A) exhibited a candidate pathogenic variant p.(Arg788His) on CACNA1H, as determined via family analyses. This study identified candidate genetic variants in about a quarter of patients (n = 16/57) and an average of 2.8 variants was identified in each patient. The results reinforced the polygenic disorder with very high locus and allelic heterogeneity of common GGE syndromes. Further, voltage-gated Ca2+ channels are suggested as important contributors to common genetic generalized epilepsy syndromes. This study extends our comprehensive understanding of common genetic generalized epilepsy syndromes.


Introduction
A proposal written by the 2010 International League Against Epilepsy (ILAE) recommends changing the term "idiopathic" to "genetic" when describing epilepsy syndromes [1]. Idiopathic/genetic generalized epilepsy (IGE/CGE) is classically categorized into several PLOS  Study. In the present study, control group were from public genetic databases of 1100 Korean individuals from the Ansan-Ansung cohorts. The whole genome of 1100 Korean individuals were sequenced using an Illumina Hiseq2000 sequencer. The raw sequences were mapped to reference genome hg19 from the University of California Santa Cruz (original GRCh37 from the National Center for Biotechnology Information, Feb. 2009) via the Burrows-Wheeler alignment tool, and the mapped sequences were analyzed using sequence alignment map (SAM) tools. To ensure sequencing accuracy, the average sequencing coverage depth was at least 30X per sample.

Epilepsy multi-gene panel testing design
A literature review yielded 111 genes, including 50 potential candidate genes for common GGE syndromes and 61 causative genes for other genetic epilepsies with absence, myoclonic, and generalized tonic-clonic seizures.
The 50 potential genes included 6 candidate Mendelian GGE genes, such as CACNB4, CASR, GABRA1, GABRD, CLCN2, and EFHC1; 9 candidate genes based on rare copy-number variants, including CHRNA7(15q13. 3 Detailed features of the 111 genes are described in S1 Table. The size of this region was 344.061 kbp, and the number of probes totaled 6,831, with a size of 474.708 kbp. This panel of genes includes coding exon regions as well as exon-intron boundaries of the region extensions, with 10 bases from the 3 0 end and 10 bases from the 5 0 end.

Epilepsy multi-gene panel testing and data generation
Genomic DNA was extracted from peripheral blood leukocytes. The Agilent SureSelect Target Enrichment protocol from the Illumina paired-end sequencing library (v. 2.0.1, May 2010), with 1 μg input DNA, was used to generate a standard exome capture library. Sequencing was performed using the HiSeq™ 2000 platform (Illumina, San Diego, USA) at Macrogen (Seoul, Republic of Korea). The quality of the whole genome sequencing reads from all samples was assessed using FastQC, and a raw data quality control process was initiated with Trimmomatic. Alignment of the sequence reads, indexing of the reference genome hg19 from the University of California Santa Cruz (original GRCh37 from the National Center for Biotechnology Information, Feb. 2009), and variant calling with a pipeline were performed using the Genome Analysis Tool Kit best practice guidelines. Alignment was performed with the Burrows-Wheeler alignment tool (version 0.7.12) and duplicate reads were marked with Picard (version 1.130, http://broadinstitute.github.io/picard/). Local alignment, base quality recalibration, and variant calling were performed using the Genome Analysis Tool Kit (version 3.4.0), and annotation was conducted with SnpEff (version 4.1 g) at the Bioinformatics Institute of Macrogen. Low-quality variants were manually filtered. Variants with an allele depth < 10 were excluded, and multi-alleles at the same chromosomal position were removed. The alternate read ratio was calculated as alternative allele depth/total depth. Therefore, a boundary of allele ratios between 0.4 and 0.6 was established, within which single-nucleotide polymorphisms (SNPs) would be considered heterozygous for alternate reads. Allele ratios of 1 were considered homozygous for alternate reads.

Identification of candidate or susceptibility genes or variants from sequence data
Identification of pathogenic variants. Disease-associated single rare variants with pathogenic effects were identified according to the American College of Medical Genetics and Genomics (ACMG) guidelines [12]. In the present study, rare variants were defined as those with minor allele frequencies (MAFs) 0.05. Rare variants were selected according to the population frequency recorded in public databases, including 1000Gp3 (1000 Genomes Project Phase 3) and ExAC (Exome Aggregation Consortium). Considering differences in allele frequencies between ethnic groups, allele frequencies were also determined via the KRGDB. To assess the variants found in patients with disease and pathogenicity, we evaluated the ClinVar and Human Gene Mutation Database (HGMD) professional version 2017. For each single rare variant, we compared differences in the frequency between the patient and control group, using Fisher's exact test. A p value < 0.05 was considered significant. To predict whether an amino acid substitution in a protein affects protein function, missense variants were annotated using the following protein function prediction tools: SIFT, PolyPhen-2 HDIV, PolyPhen-2 HVAR, LRT, MutationTaster, MutationAssessor, FATHMM, PROVEA, MetaSVM, and MetaLR. Four nucleotide conservation prediction tools including PhastCons, GERP, PhyloP, and SiPhy contributed to the predictions.
Identification of susceptibility variants using case control association analyses. Non-Mendelian genetics are commonly used to explain GGE syndromes, wherein multiple susceptibility alleles exert small effects on the occurrence of disease. To determine susceptibility alleles, case control genetic association analyses were performed. We selected all coding singlenucleotide variants (SNVs). The allele frequency of coding SNVs was compared between the disease group and 1100 control group from KRGDB. The p values were calculated using Fisher's exact tests for SNVs and control group. The obtained p values for each SNV were then summarized via Manhattan plots. The y-axis of a Manhattan plot typically represents the negative log (base 10) of the p values obtained for the association tests applied. The x-axis is typically organized by chromosome (chromosome 1 to 22, X). Each dot on the Manhattan plot signifies the mean for each SNV. We identified SNVs with p cut-off values of 0.05 (-log 10 p = 1.3) and corrected p-values using several variants.
Family analyses in familial cases. Genes with rare variations that segregate in association with Mendelian common GGE syndromes were identified for each familial case via family analyses. All affected family member samples underwent multi-gene panel testing. Rare variant candidates were selected using ACMG guidelines. Sanger confirmations were conducted for identified variants in families.

Clinical description
A total of 57 patients with common IGE/GGE syndromes were enrolled in the present study, including 51 patients with sporadic disease and 6 with familial disease. Age at onset of seizures ranged from 5-34 years. Table 1 presents clinical data for the 57 patients. The pedigrees of the 6 familial groups are presented in Fig 1. Additionally, 8 affected family members from the 6 familial groups were included in the analyses. Thus, epilepsy multi-gene panel testing was performed for a total of 65 patients.

Statistical results from sequence data
An average of 13 million reads per patient were generated. Each read was approximately 101 bp long; the average total targeted region size was 340,412 bp. Approximately 99.8% of the mapped reads included the targeted regions. The mean depth was 1342 X. Table 2

Identification of susceptibility variants using case-control association analyses
In total, 231 variants that were predicted to disrupt protein coding sequencing in 57 patients.
The p values and -log 10

Family analyses in 6 familial cases
Family A. An 18-year-old woman (ID: SMJ) was first assessed at 13 years of age for unprovoked GTCSs upon waking. The pedigree is illustrated in Fig 1. Her mother first experienced seizures at 25 years of age and received antiepileptic treatment with lamotrigine. Neither the proband nor her mother experienced any febrile generalized seizures. Morning myoclonic and absence seizures were not reported in either subject. The common GGE subtype diagnosis was EGTCS. Considering the autosomal dominant inheritance pattern and filtered heterozygous variants, the missense variant c.2363G>A (rs532296550) p.(Arg788His) in CACNA1H was strongly supported as the disease-causing variant, on the basis of the reference sequence NM_021098.2. Characteristics of the CACNA1H p.(Arg788His) variant are summarized in S2  Table and Table 4. The proband also had a heterozygous variant, c.2362C>T p.(Arg788Cys) in CACNA1H, which is known as a functional polymorphism that alters T-type calcium channels. The results of Sanger sequencing are presented in Fig 1 [13].
Family B. A 15-year-old girl (ID: LJY) presented with unprovoked GTCSs. She experienced unprovoked seizures since the age of 14 years that were characterized by brief episodes of myoclonic jerks, especially in the upper limbs, and GTCSs upon waking. A family history indicated epilepsy in her mother, who was diagnosed with JME. The mother was 40 years and had received treatment with an antiepileptic drug for approximately 3 years in her late teens and early twenties. The proband and her mother had no history of febrile convulsions. Given the autosomal dominant inheritance pattern and filtered heterozygous variants, no likely candidates were identified.
Family C. A 17-year-old boy (ID: E-026-P) had a history of recurrent myoclonic jerks, predominantly upon awakening, and GTCSs that first appeared at 8 years of age. He was diagnosed with JME. His mother, dizygotic twin brother, and younger brother had similar clinical seizures, were also diagnosed with familial JME. There was no family history of febrile seizures.    The proband's father was unaffected, and his DNA was obtained for control sequencing. Autosomal dominant inheritance from the mother was suspected; therefore, we filtered the variants. However, no candidate variants were identified. Family D. A 20-year-old man (ID: E-047-P) was confirmed to have JME. He first experienced myoclonic jerks and GTCSs upon waking at 8 years of age. His younger brother also developed similar seizures (myoclonic jerks and GTCSs) at 12 years of age, was also diagnosed with JME. Both patients did not have any history of febrile seizures. Considering the autosomal recessive inheritance pattern, homozygous variants were filtered; however, no candidate variants were identified. Although it did not fit the inheritance pattern, a rare heterozygous variant JRK p.(Ser388Leu) was observed in both affected family members. The characteristics of the JRK p.(Ser388Leu) variant are described in S3 Table and Table 4. The proband also had the rare variant, CACNA1A p.(Ile239Phefs Ã 5), which was suspected as possibly pathogenic; however, his sibling did not have this variant S2 Table and Table 4.
Family E. A 19-year-old man (ID: E-087-P) was diagnosed with JME. He first experienced myoclonic jerks and GTCSs upon waking at 17 years of age. His younger brother was also diagnosed with JME, which developed at 12 years of age. Neither patient had a history of febrile seizures. He was treated with levetiracetam and his younger brother was treated with valproate. Considering the autosomal recessive inheritance pattern, homozygous variants were filtered; however, no candidate variants were identified. Although it did not fit the inheritance pattern, the rare heterozygous variant GPR98 p.(Gly4447Asp), which is suspected of potentially causing the disease, was observed in both affected members. Characteristics of the GPR98 p. (Gly4447Asp) are illustrated in S3 Table and Table 4.
Family F. A 23-year-old man (ID: E-184-P) was diagnosed with JME. He first experienced myoclonic jerks and GTCSs upon waking at 13 years old. His younger sister had similar seizures that developed at 10 years old. Neither had a history of febrile seizures. Considering the autosomal recessive inheritance pattern, we filtered homozygous variants and did not identify any candidates. Further, heterozygous rare variants or possible single variants were not detected.

Discussion
Based on significant advances in our knowledge regarding the genetics of epileptogenic mechanisms in the recent decades, the present study was designed to test the hypothesis that known epilepsy genes, rather than unrecognized new genes, are responsible for common GGE syndromes. We aimed to conduct a focused analysis of potential candidate genes for common GGE syndromes using multi-gene panel testing with higher coverage depth and reduced burden of data analysis compared to WES. Analyses to determine pathogenic variants indicated very high genetic heterogeneity. Three pathogenic variants and 19 likely pathogenic variants on 16 different genes were identified. The diagnostic yield in common GGE syndromes was 28% (n = 16/57). Of the 22 candidate variants, 73% (n = 16/22) on 10 genes, including GABRD, SCN1A, GRM4, CPA6, CHRNB3, CACNA1H, NDE1, CACNA1G, CACNA1A, and LGI4, were sites for 50 candidate variants for common GGE syndromes. Interestingly, 27% of variants (n = 6/22) on 6 genes (SZT2, DOCK7, GPR98, GATM, POLG, and ARX) belonged to 61 causative genes known to be associated with other forms of genetic epilepsy. The results reinforced the hypothesis that common GGE syndromes are polygenic disorders with extreme locus and allelic heterogeneity. Variants in voltage-gated Ca 2+ channel genes including CACNA1A, CACNA1G, and CAC-NA1H were also frequently observed and accounted for 32% (n = 7/22) of the candidate variants. Voltage-sensitive calcium channels mediate the entry of Ca 2+ into excitable cells and are also involved in a variety of Ca 2+ -dependent processes [14,15]. Voltage-gated Ca 2+ channel genes are known as susceptibility factors for GAE with typical absence seizures [15]. In the present study, pathogenic variants and likely pathogenic variants on CACNA1A, CACNA1G, and CACNA1H were found in 4 patients with JME and familial EGTCS, as shown in Table 3. We suggested that voltage-gated Ca 2+ channels are important to common GGE syndromes, as well as typical absence epilepsy.
CACNA1H variants (n = 4/22, 18%) were most frequently observed. The alpha-1H isoform produces T-type Ca 2+ currents, and T-type Ca 2+ channels belong to the low voltage-activated group. Particular characteristics of this type of channel include opening at highly negative potentials and voltage-dependent inactivation [14]. T-type channels serve pace-making functions in both central neurons and cardiac nodal cells and support calcium signaling in secretory cells and vascular smooth muscles [14]. They may also be involved in modulating neuronal firing patterns, which are crucial for information processing and cell growth processes [14]. CACNA1H activation produces T-type Ca 2+ channels, and functional variations in CACNA1H are increased susceptibility to GGE [13,16]. Previous studies have reported associations between common GGE syndromes and CACNA1H, especially GAE (CAE/JAE), primarily in the Chinese Han population [15,[17][18][19]. Eckle et al. [14] suggested that gain-offunction variants in CACNA1H directly increase seizure susceptibility by altering neuronal electrical properties, and indirectly increase seizure susceptibility by changing gene expression. Results of the present study suggest that CACNA1H c.2363G>A p.(Arg788His) is a pathogenic variant responsible for autosomal dominant EGTCS in family A. This missense change occurred at an amino acid residue in which a CACNA1H c.2362C>T (p.Arg816Cys) variant has been reported previously [13], and the p.Arg788Cys variant has been proposed as a functional polymorphism that occurs more frequently in people with IGEs than in the general population [13]. As such, the present study suggests that arginine 788 in CACNA1H is an important contributor to epilepsy.
Analyses to identify susceptibility variants using case-control association analyses indicated that 1 variant of KCNMA1 c.400G>C is significantly associated when the corrected p value was considered. The KCNMA1 gene underlies Ca 2+ activated K + channels, which are fundamental to smooth muscle tone and neuronal excitability. Diseases associated with KCNMA1 include autism, cerebellar atrophy, and generalized epilepsy and paroxysmal dyskinesia [MIM 609446]. Further, KCNMA1 c.400G>C may be the site of susceptibility variants in common GGE syndromes. However, results of the case-control association analyses are controversial, attributable to study limitations resulting from multi-gene panel testing and the relatively small sample size.
Although no meta-analysis study has been conducted, <1% of common GGEs are suspected to be heritable Mendelian monogenic diseases [7]. In the present study, 6 patients had familial disease and underwent additional family segregation analyses. Of the 6 families, only 1 (family A) had the candidate variant p.(Arg788His) on the CACNA1H gene. The other 5 familial cases (B-F) did not exhibit any causative gene that was consistent with the inheritance pattern via linkage analysis. Interestingly, the present study did not find any variants that were consistent with expected Mendelian JME inheritance, such as variants of CACNB4, CASR, GABRA1, GABRD, CLCN2, or EFHC1. This unexpected result in familial analysis could have several causes. The possibility of monogenic disorders with variations in penetrance and expressivity of causative genes makes it difficult to identify the pathogenic variants. Despite high coverage and depth, challenges remain in the complete coverage of targeted regions, and it is impossible to identify copy number variations (CNVs) in multi-gene panel sequencing data. For recessive disorders such as those observed for families D-F, parental testing is particularly important for analyses of pathogenic variants. Unfortunately, we did not obtain parental DNA samples from families D-F. Each proband had several rare candidate variants; however, variants were not co-segregated in each family. In particular, family D and E exhibited JRK p. (Ser388Leu) and GPR98 p.(Gly4447Asp), which are rare heterozygous candidate variants and were not consistent with the expected autosomal recessive inherited pattern. Recently, several genetic studies of familial GGE showed that no individual genes were significantly associated with familial GGE; however, several known epilepsy genes and variants were significantly enriched in a group of familial GGE [16,20]. Therefore, familial cases may result from multigenic disorders where interactions between multiple causative and susceptibility genes or variants with modest or strong effects produce sporadic cases.
In summary, we identified candidate genetic variants causing epilepsy susceptibility in about a quarter of the patients (28%, n = 16/57) in this study. An average of 2.8 variants was identified in each of the 57 patients. The results of this study reconfirm the polygenic disorder with very high locus/allelic heterogeneity of common GGE syndromes and suggest that voltage-gated Ca 2+ channels are possibly important contributors to common GGE syndromes. Furthermore, these data suggest that common familial GGE syndromes are polygenic disorders with low penetrance and variable expressivity. Considering the possibility of ethnic and racial differences in the distribution of epilepsy susceptibility genes and an effect of genetic background on the epilepsy phenotype, it is worth noting that this study is the first analysis of a single ethnic group of Korean patients with common GGE syndromes using multi-gene panel testing. This study extends our comprehensive understanding of common GGEs.