A Wide Spectrum of Genetic Disorders Causing Severe Childhood Epilepsy in Taiwan: A Case Series of Ultrarare Genetic Cause and Novel Mutation Analysis in a Pilot Study

Background: Pediatric epileptic encephalopathy and severe neurological disorders comprise a group of heterogenous diseases. We used whole-exome sequencing (WES) to identify genetic defects in pediatric patients. Methods: Patients with refractory seizures using ≥2 antiepileptic drugs (AEDs) receiving one AED and having neurodevelopmental regression or having severe neurological or neuromuscular disorders with unidentified causes were enrolled, of which 54 patients fulfilled the inclusion criteria, were enrolled, and underwent WES. Results: Genetic diagnoses were confirmed in 24 patients. In the seizure group, KCNQ2, SCN1A, TBCID 24, GRIN1, IRF2BPL, MECP2, OSGEP, PACS1, PIGA, PPP1CB, SMARCA4, SUOX, SZT2, UBE3A, 16p13.11 microdeletion, [4p16.3p16.1(68,345–7,739,782)X1, 17q25.1q25.3(73,608,322–81,041,938)X3], and LAMA2 were identified. In the nonseizure group, SCN2A, SPTBN2, DMD, and FBN1 were identified. Ten novel mutations were identified. The recurrent genes included SCN1A, KCNQ2, and TBCID24. Male pediatric patients had a significantly higher (57% vs. 29%; p < 0.05, odds ratio = 3.18) yield than their female counterparts. Seventeen genes were identified from the seizure groups, of which 82% were rare genetic etiologies for childhood seizure and did not appear recurrently in the case series. Conclusions: Wide genetic variation was identified for severe childhood seizures by WES. WES had a high yield, particularly in male infantile patients.


Introduction
In children, genetic disorders cause severe neurological disease, congenital malformation, inborn errors of metabolism, and developmental epileptic encephalopathy (DEE). DEEs refer to a group of ictal and interictal epileptiform anomalies (clinical and encephalographic) associated with severe cognitive and behavioral impairments according to the classification and terminology criteria of the International League against Epilepsy (ILAE) [1,2]. DEEs are age-specific and of diverse etiologies. Increasing evidence suggests that genetics play a pivotal role in pediatric DEEs and severe neurological disorders [3][4][5]. Although the incidence of each disease is low, the combined incidence is not adequately estimated and unknown. DEEs are highly heterogeneous genetically, but genetic etiologies have A genomic DNA purification kit (Gentra Puregene Buccal Cell Kit; Qiagen Taiwan, Taipei City, Taiwan) was used to extract genomic DNA from peripheral blood. Three volumes of RBC lysis buffers were added to blood sample and mixed by inverting 30 times, incubated for 10 min at room temperature, and centrifuged 3000 rpm for 5 min. The supernatant was discarded. Next, we added 3 volumes of cell lysis buffers (3 mL) to the pellet and vortexed vigorously for 10 s to lyse the cell and centrifuged 3000 rpm for 5 min, followed by transferring the supernatant into a new 15 mL tube. Then, we added 3 mL isopropanol into a 15 mL tube, mixed by inverting 50 times, and centrifuged 3000 rpm for 5 min. The supernatant was carefully discarded and the pellet was air dried for 10 to 20 min. The dried pellet was resuspended in 300 µL nuclease-free water and frozen at −20 • C or −80 • C for storage. DNA were also extracted from their parents.

WES
We used the SureSelect XT HS2 DNA reagent kit (Agilent Technologies, Inc., Santa Clara, CA, USA) protocol for the Illumina Hiseq paired-end sequencing library (catalog# G9611A), and raw reads were mapped to the human genome assembly GRCh37 (also known as hg19) using Burrows-Wheeler Aligner software (version 0.6.1; Intel Corporation, Santa Clara, CA, United States). The SureSelectXT Human All Exon Version 6 (51 Mb) probe (Agilent Technologies, Inc., Santa Clara, CA, USA) set was used. For Library Preparation, 50 ng genomic DNA was used. The adapter-ligated DNA sample was purified using Agencourt Ampure Xp Pcr Purification Beads and analyzed using an Agilent DNA Kit. From the purified sample, the hybridization between DNA libraries (750 ng) and baits was carried out and purified using Agencourt Ampure Xp Pcr Purification Beads. The Agilent protocol was used for adding tags by Post-Hybridization Amplification. Finally, the sample was sequenced on Illumina NextSeq500 using the generated reads of 2 × 150 bp. Every analyte passed all the quality control requirements, and 99% of targeted nucleotides were covered at more than 20×.

Data Analysis and Interpretation
Variant calling was performed using the recommended best practices of GATK version1.0.5506 (Broad Institute). Variant annotation and prioritization were performed using a well-developed pipeline called wANNOVAR [10], which is used for functional annotations, including various gene annotations, alternative allele frequency in the 1000 Genomes Project, conserved element annotation, dbSNP annotation, deleteriousness prediction scores for nonsynonymous variants, ClinVar variant annotation, and genome-wide association study (GWAS) variant annotation [11].
We implemented a variant-reduction pipeline based on commonly used filters and disease models to select nonsynonymous variants and splice variants, rare or novel variants in the 1000 Genomes Project database, and predicted deleterious variants. Thus, synonymous variants, variants with variant frequency <10%, and variants with allele frequency >1% were removed. The remaining variants were annotated with reference to ClinVar, a freely accessible human variation and phenotype database hosted by the National Center for Biotechnology Information (NCBI) [12]. Pathogenicity prediction programs, including PolyPhen2 [13], SIFT [14], and Combined Annotation Dependent Depletion (CADD) [15], were used. All the variants identified were further confirmed by Sanger sequencing. The corresponding gene contexts were evaluated according to Online Mendelian Inheritance in Man ® [16] with the individual phenotypes (clinical, laboratory, and imaging data). Segregation analysis was performed to select de novo or compound heterozygous variants. The identified variants were classified "pathogenic," "likely pathogenic," or "of uncertain significance" according to the American College of Medical Genetics (ACMG) standards and guidelines [17].

Demographic Data of the Enrolled Patients
A total of 54 patients were enrolled, of which 45 had a history of seizure, including 10 patients with seizure onset before 1 month of age, 31 patients with seizure onset between 1 month and 2 years of age, and 4 patients with seizure onset after 2 years of age. Of the patients, 10 received 1 drug and 35 received at least ≥2 AEDs. Nine patients had no history of seizures and were suspected to have severe neurological or neuromuscular disease with regression of symptoms. The demographic data are shown in Table 1. The detailed clinical information including antiepileptic drugs, magnetic resonance imaging (MRI) findings, and outcomes are demonstrated in Table 2.

Diagnostic Yield
The overall diagnostic rate was 44.4% (24/54). We compared patients with identified and nonidentified genetic etiologies through WES, which indicated a significant difference (p < 0.05) in sex between the two groups. In the identified genetic group (N = 24), the sex ratio was 18 males to 6 females. The inheritance patterns were 3 X-linked, 6 autosomal recessive, and 15 autosomal dominant. Male pediatric patients had a significantly higher yield than their female counterparts (57% vs. 29%, odds ratio = 3.18 (95% confidence interval = 1.04 to 9.70)). We found no difference in the number of epileptic drugs taken and the time of seizure onset between the identified genetic and nonidentified genetic groups (Table 1). Among 45 patients with seizures, the yield was 44% (20/45). Considerable genetic variability was found in the seizure group, as recurrent genetic etiologies included KCNQ2, TBCID24, and SCN1A in the cases series, with two patients carrying each gene. Other genes were nonrecurrent for each case in the case series. Seventeen genetic etiologies were identified in the seizure group, of which 82% (14/17) were rare and nonrecurrent.

Genotype
The clinical genotype and phenotype in genetically positive cases are shown in Table 2. Clinical and molecular data for 24 patients are listed. In 54 patients, 10 novel mutations were identified (Table 3). Among 24 patients with identified genetic etiologies, there were 13 with DEE, 6 patients had seizures with neurodevelopmental regressive symptoms, 2 patients with muscle disease (1 patient with muscle dystrophy carrying DMD gene and 1 patient with congenital muscle dystrophy with seizures), 1 with autism with SCN2A mutation, 1 with spinocerebellar ataxia, and 1 patient with Marfan syndrome. Eleven patients had a single autosomal-dominant disorder and all showed de novo variants. Of these, 10 mutations were novel ( Table 3). The diagnosis in four patients was changed after WES, including one with migraine changed to spinocerebellar ataxia, one with Leigh-like syndrome changed to sulfite oxidase deficiency, and two which remained undiagnosed but were suspected of carrying copy number variation after read depth was computed from WES. Thereafter, a final diagnosis was given through multiplex ligation-dependent probe amplification (MLPA) or microarray-based comparative genomic hybridization (aCGH) (patients 9 and 17). The 10 novel mutations include 2 missense, 2 splicing, 3 nonsense, 2 indel, and 1 frameshift mutations. They were considered pathogenic or likely pathogenic mutations by ACMG guidelines (Table 3).

Refractory Seizure Cases
Of the 45 patients with seizure, a definitive genetic etiology was identified in 20 (44%). Of the 10 patients with seizure onset before the first month, etiology was confirmed through WES in 5 (50%) patients. These included two patients with KCNQ2 and one patient each with PACS1, OSGEP, and PIGA ( Table 2). All patients had poor neurodevelopmental outcomes. One patient with homozygous OSGEP mutations died before 4 months of age due to severe pleural effusions. One patient with PACS1 has survived and is 15 years old.
Of the 31 patients with seizure onset between 1 month and 2 years of age, 13 (42%) received a definitive diagnosis, including 2 patients each with TBCID24 and SCN1A and 1 patient each with    [12]. The sequence data of each patient were checked against the GenBank reference sequence and version number of genes.  PACS1 has survived and is 15 years old.

Recurrent Mutations
The recurrent mutations involved SCN1A, KCNQ2, and TBC1D24. Six cases were attributed to these genes, with seizure onset before 2 years of age. The recurrence of these mutations in the case series indicates that these genes are more commonly involved in childhood DEE, particularly in cases where seizure onset occurs before 2 years of age. Two patients with mutations in KCNQ2 had severe neurological outcomes, and they could not speak at 3 years of age or walk. Two patients with mutations in TBCID24 had persistent refractory seizures with an attention deficit. Two patients with mutations in SCN1A had severe neurological outcomes and refractory seizures.

Genes Accounting for Epilepsy Demonstrated by the Age of Seizure Onset
The genes accounting for epileptic patients in the study and in a literature review [18][19][20][21][22][23] (Table 4) are demonstrated in the order of age of seizure onset from birth to 1 month old ( Figure 4A), 1 month to 24 months old ( Figure 4B), and 24 months to adulthood ( Figure 4C).

Nonseizure Group
In the nonseizure group, the yield was 44% (4/9), including one with SCN2A with autism, one FBN1 with Marfan syndrome, one DMD with muscle dystrophy, and one SPTBN2 with spinocerebellar ataxia.

Discussion
Of 54 patients with severe neurological disorders, we identified variants corresponding with the disease in 24 patients. In neonatal infants with seizures, genetic etiologies varied. SCN1A, KCNQ2, and TBCID24 had recurrent mutations in the case series. Mutations in other genes were sporadic and not recurrent. Seventeen genetic etiologies were identified in the seizure group. Of those, 82% (14/17) responsible for childhood seizure were rare and nonrecurrent, indicating that the genetic spectrum for neonatal infantile seizures is wide. Thus, the genes selected for the epileptic panel may not represent all etiologies.
The diagnostic yield of WES was 44%, which is not significantly different from that in other studies, which varied from 30% to 70% [19]. However, results vary greatly and depend on the disease

Nonseizure Group
In the nonseizure group, the yield was 44% (4/9), including one with SCN2A with autism, one FBN1 with Marfan syndrome, one DMD with muscle dystrophy, and one SPTBN2 with spinocerebellar ataxia.

Discussion
Of 54 patients with severe neurological disorders, we identified variants corresponding with the disease in 24 patients. In neonatal infants with seizures, genetic etiologies varied. SCN1A, KCNQ2, and TBCID24 had recurrent mutations in the case series. Mutations in other genes were sporadic and not recurrent. Seventeen genetic etiologies were identified in the seizure group. Of those, 82% (14/17) responsible for childhood seizure were rare and nonrecurrent, indicating that the genetic spectrum for neonatal infantile seizures is wide. Thus, the genes selected for the epileptic panel may not represent all etiologies.
The diagnostic yield of WES was 44%, which is not significantly different from that in other studies, which varied from 30% to 70% [19]. However, results vary greatly and depend on the disease groups selected. Studies have reported diagnostic rates of 19% [24], 41% [23], and 49.1% [25]. Neonatal or infantile DEE was the most commonly found syndrome in this study. Studies have used comprehensive epilepsy gene panel analysis, with a success rate ranging from 10% to 40% [20,21] depending on gene selection and number of cases. However, previous studies have not performed gene panel analysis, which explains why WES is more effective in diagnosing genetic etiologies.
The finding that the WES yield was significant in male pediatric patients provides a viable treatment strategy. The finding needs further explanation: One factor is X-linked inheritance. The sex ratio was 18 males to 6 females. However, only 3 were X-linked. The second factor is that the majority (76%) of 54 patients were aged below 2 years old. In the identified genetic group, seizures onset before 1 year old were found in 18 (75%) patients. Demos et al. [18] found that 12 (44.4%) of 27 males and 9 (35%) of 26 females with epileptic diagnosis <6 months had a genetic diagnosis by WES in 360 epileptic cases with an average seizure onset at ≤5 years (Table 4). That indicated a trend of higher yield in male newborn infants with a younger age of seizure onset. Genetic counseling to prevent next offspring was carried out in all patients, even in the autosomal dominant patients with de novo mutation. In our cohort, 12 patients with autosomal dominant inheritance had de novo mutations. The next offspring were explained to prevent unnecessary concern. In addition, we changed antiepileptic drug selection after detecting a causative KCNQ2 variant in patients 4 and 5, which resulted in more effective seizure control. A patient was identified with a mutation in KCNMA1, which encodes the α-subunit of the large conductance calcium-sensitive potassium channel. Mutations in KCNMA1 increase Ca 2+ sensitivity of the channel by three-to five-fold, resulting in generalized epilepsy and paroxysmal dyskinesia [24]. We chose to administer levetiracetam (LEV; 20 mg/kg/day) because LEV can limit epileptogenesis by inhibiting Ca 2+ elevation following seizures, thus exerting neuroprotective and antiepileptogenic effects [26,27]. The administration of LEV resulted in no episodes of seizure and low frequency of paroxysmal dyskinesia in the patient. Thus, WES data can support precision therapy and provide information for managing pediatric patients with epilepsy.
SCN1A is studied in the context of genotype-phenotype correlations in the GEFS+ spectrum and DS [20,28,29]. Pathological mutations in structural and functional proteins correlate with specific clinical presentations. Modifier genes also contribute to modulating the phenotype. Most phenotypes that cause severe disability are a result of de novo mutations [20,30]. However, the hypothesis does not account for mosaicism [31]. Some studies have proposed that epigenetic factors contribute to determining phenotypes [20,30,31]. In SCN1A channelopathy, DS is the most severe phenotype. Approximately 70-80% of patients with DS [20,30] have SCN1A mutations, which are mostly de novo, as in our patients. Most patients with DS have mutations in the "ion pore" of SCN1A [32][33][34][35]. However, one of our patients with a mutation in the ion pore region had an identical unaffected twin sister. Although studies have reported that patients with pathogenic SCN1A mutations may remain unaffected, the mechanism is unknown. In a large family with familial pathogenic SCN1A missense mutations, clustering of three unaffected carriers was observed in the same generation [35]. Genetic or epigenetic changes were proposed as the causative factors but not established, because the effects of other confounding factors could not be excluded. KCNQ2 encodes a voltage-gated potassium channel that is expressed in the brain and is involved in the etiology of epileptic encephalopathy, early infantile (EIEE7, phenotype MIM# 613720), and benign familial neonatal seizures-1 (BFNS1, phenotype MIM#121200) [36,37]. Two (patients 4 and 5) de novo heterozygous mutations in KCNQ2, namely c.740C > T (p.Ser247Leu) and c.740C > T (p.Pro285Thr), are highly pathogenic and located in the critical pore domain [38]. This finding helps us select suitable AEDs, such as oxcarbazepine, to control seizures.
The diagnostic methods focusing on the genotype-phenotype correlation followed by specialized biochemical tests and Sanger sequencing for suspected genes are time-consuming and costly [39]. Moreover, only "typical" diseases can be diagnosed, and even when the same causative mutations are present, atypical diseases can be overlooked. In summary, we provide a perspective for diagnosing severe neurological disorders in children using WES. The study presents important findings on the genetic causes of severe neurological disorders in childhood, particularly in epilepsy and neurodegenerative diseases.
In the clinical setting, if the clinical course is thought to be benign, including benign epileptic syndromes [40] in childhood, WES is not performed. However, if, after treatment with AEDs, the response is unfavorable and seizures are refractory, to obtain the best time of seizure control and to avoid sacrificing the moment of children's brain development, as in our case with UBE3A mutation whose seizure was far preceded to her developmental regression, WES should be performed. Therefore, we recommend WES for assessing all childhood seizures with poor response to AEDs without definitive etiology to save time in selecting appropriate AEDs.

Conclusions
The findings on the genetic causes of severe childhood neurological disorders are significant. A scientific and efficient WES analysis is particularly important and should be emphasized because the affected gene may not be present in the panel. This findings of this study highlight the diagnostic relevance of WES for childhood patients with epilepsy in clinical practice.