Phenotypic and genetic spectrum of epilepsy with myoclonic atonic seizures

We aimed to describe the extent of neurodevelopmental impairments and identify the genetic etiologies in a large cohort of patients with epilepsy with myoclonic atonic seizures (MAE).


| 997
TANG eT Al. the variable combination of seizure types and developmental profile prior to seizure onset is open to interpretation, such that the phenotypic and nosological boundaries for this disorder remain debated. 3 The spectrum of neurodevelopmental comorbidity in MAE has not been systemically described. Intellectual disability (ID) has been reported in 34%-60% of patients assessed with variable psychometric tools and different cognitive definitions. [2][3][4][5][6] Adaptive functioning, a collection of age-appropriate conceptual, social, and practical skills to enable a person to function in everyday life, has not been previously explored. Autism spectrum disorder (ASD) and attention-deficit/hyperactivity disorder (ADHD) symptoms have been documented in 5%-45% of very small case series and case reports. 7,8 The importance of genetic factors in MAE was recognized through family history and EEG studies in its first description. 2 Early twin studies and rare Mendelian pedigrees offered a glimpse of the genetic basis of MAE. 9,10 Subsequently, the era of gene discovery through next generation sequencing and the recognition of the role of de novo variants in epileptic encephalopathies (EE) and neurodevelopmental disorders has led to a multitude of gene associations for MAE.
Here, we set out to provide a multifaceted analysis of the condition and describe deep phenotyping and exome sequencing in a cohort of 101 MAE patients. We present assumed pathogenic variants in seven unpublished and five previously described cases from single gene discovery studies. [13][14][15]18,19,23 2 | MATERIALS AND METHODS

| Subjects
MAE patients fulfilling the criteria of (1) onset of myoclonic, myoclonic atonic, or atonic seizures between 7 months and 6 years; (2) presence of generalized spike or polyspike-wave EEG discharges; and (3) exclusion of other epilepsy syndrome 24, 25 were recruited through three cohorts. These were (1) the EuroEPINOMICS Rare Epilepsy Syndrome MAE cohort, 14 (2) an Italian cohort collected through a tertiary pediatric neurology center at the Meyer Children's Hospital, and (3) a UK cohort (award ref MR/J011231/1, ethics ref 09/ H0713/76). Written informed consent was obtained from all parents/legal guardians of participating patients.

| Phenotyping methods
Medical notes including seizure types, presence of febrile convulsions, and EEG reports were obtained from referring clinical collaborators for the UK and Italian cohorts. Details from the EuroEPINOMICS cohort were available through an online password-protected platform.
All UK cases were phenotyped for adaptive functioning skills, ASD, ADHD, and behavioral screening rated by parent/carer and teacher where indicated. Additionally, families that lived close to King's Health Partners, London were invited for deep phenotyping for cognition using the Wechsler Preschool and Primary Scale of Intelligence-Third UK Edition (WPPSI), Bayley Scales of Infant and Toddler Development-Third Edition (Bayleys), and Developmental, Dimensional and Diagnostic Interview (3di) rapid assessment. Deep phenotyping was assessed by a neuropsychologist or pediatric neurologist (S.T., A.S., M.A., D.K.P.).

| Adaptive behavior
Adaptive behavior was measured with the Adaptive Behaviour Assessment System (ABAS II) Parent Form. The ABAS II explores three domains: conceptual, social, and practical. The general adaptive composite score is derived from the sum of the three domains.

| Autism
Autism was measured with the Social Communication Questionnaire (SCQ) Autoscore Form: Lifetime and the 3di rapid assessment computerized interview.

| Behavior
Behavior was measured using Conners' Comprehensive Behavioural Rating Scale (CBRS). The presence of significant ADHD symptoms was assumed when a T score > 70 was surpassed in subscale N. We also used the Strength and Difficulties Questionnaire (SDQ) as a behavioral screening questionnaire (www.sdqin fo.com). The SDQ has an impact supplement that enquires about chronicity, distress, social impairment, and burden to others. The scores were classified as close to average, slightly raised, high, or very high.

| Exome sequencing
Exome sequencing was performed at two centers. The EuroEPINOMICS cohort was sequenced at the Wellcome Trust Sanger Institute (Hinxton, Cambridgeshire, UK), and techniques have been described before. 28 BAM files from the EuroEPINOMICS cohort were converted back to FASTQ files, and variant calling was reperformed using the Guy's Genomics Facility pipeline consistent with the remainder of the cohort. The Italian and UK cohorts were sequenced at the Guys Genomics Facility (Guy's Hospital, London, UK). DNA libraries were prepared using the SureSelect Human All Exon 50 Mb Kit (Agilent Technologies). Samples were multiplexed (four samples on each lane), and 100-bp paired-end sequencing was performed on the Illumina HiSeq system. Sequencing reads were aligned using Novoalign (http://www.novoc raft. com). FastQC was used to perform quality control on FASTQ files. Variant calling was performed with SAMtools, and annotation was performed using ANNOVAR (http://annov ar.openb ioinf ormat ics.org). A read depth of 10 reads was the minimum cutoff, and heterozygous variants were called if the alternate allele was present in >20% of total reads. All genetic variants are reported in human genome build GRCh37 (hg19) coordinates.

| Gene and variant filtering
Variants were selected when not observed in 1000 Genome Project, Exome Variant Server, and ExAC and in genes overlapping with two gene lists: (1) the epilepsy-associated gene list (Table S1), consisting of 100 reported epilepsy-associated genes identified through various epilepsy sequencing studies; and (2) the neuropsychiatric gene list (Table S2), consisting of 2105 genes curated from de novo sequencing studies of ID, 29-31 developmental disorders, 32 EE, 33 and ASD. [34][35][36] For variants matching a gene in the neuropsychiatric gene list, variants were reviewed if they had more than one filtered variant per gene with a CADD score > 20. All selected variants were then assessed based on genotype-phenotype correlation through PubMed literature search, gene intolerance, and gene expression (as above), and in some instances by locus-specific databases.
Sanger sequencing using standard methods from polymerase chain reaction (PCR) fragments was used to confirm all variants and for inheritance studies. Individual variants were classified as likely benign, variant of uncertain significance (VUS), pathogenic, or candidate variants. Variants were considered pathogenic if they were not present in control population data, were nonsynonymous, were splice site altering, were nonsense or frameshift, predicted damaging by one or more prediction tools, had consistent phenotypegenotype characteristics for gene, and were de novo. Likely benign variants did not have supportive phenotype-genotype correlations. VUS had unclear phenotype-genotype correlations and/or were inherited from an unaffected parent. Candidate variants were variants identified in genes that are not known to be associated with an epilepsy phenotype. Variants were interrogated by a pediatric neurologist, a molecular geneticist, and a bioinformatics scientist (S.T., D.K.P., L.A.).

| Onset
The median age at seizure onset was 34 months (range = 6-72 months) based on 100 probands. Twenty (21.0%) of 95 patients had evidence of a developmental epileptic encephalopathy where developmental delay was reported prior to epilepsy onset; of these patients with prior developmental delay, nine reported isolated speech delay prior to seizure onset. Six patients had missing information about early development. Twenty-eight (38.3%) of 73 patients had a personal history of febrile convulsions.

| Family history
A family history of epilepsy was reported in 36 of 95 probands; a first-degree family member in 14 (of which one had MAE) and second-or higher-degree family member in 22. A family history of febrile seizures was reported in six probands; family history was missing in six.

| Seizures
Seizure types at onset were generalized tonic-clonic seizures (GTCS) in 52 patients, myoclonic atonic or atonic in 31, myoclonic in 13, and absence in four of 100 patients. Seizure type at onset was missing for one patient. During the epilepsy course, mainly generalized seizure types were reported (see Table 1). Clinical examination, available in 92 patients, was abnormal in 21 (22.8%), including tremor and/or ataxia in 15, pyramidal or motor signs in three, dysmorphism in two, and microcephaly in one. Dysmorphic features were described as prominent forehead, small eyes, large mouth, and syndactyly in one patient; and frontal balding, broad nasal bridge, thick alae nasi, broad and long philtrum, thin upper vermillion and thick lower vermillion, skin wrinkling, upturned nasal tip, and anterior projection of the upper lip over premaxilla in the other patient.

| EEG
EEG background was slow during the epilepsy course in 16 of 70 (22.8%) patients. EEG background data were not available for the EuroEPINOMICS cohort. One hundred patients had abnormal generalized epileptiform activity of spikewave discharges. Additional EEG features comprised polyspike and wave in 31 and focal epileptiform activity in seven.

| Neuroimaging
Computed tomography and/or magnetic resonance imaging reports were available for 78 patients. This was reported normal in 72 patients, with generalized cerebral volume loss in two, nonspecific focal signal alteration in two, and immature myelination and benign hydrocephalus in one patient each.

| Remission
Data regarding seizure remission, where seizure freedom was achieved for >2 years with or without antiepileptic drugs, were obtained for 72 patients within the UK and Italian cohorts. Twenty-four of 72 (33.3%) patients fulfilled this definition of seizure remission. There was no statistical difference in the median age of seizure onset in both groups (32.5 months in remission group compared with 33.5 months in nonremission group, Mann-Whitney U test, P = .93). The most common antiepileptic drugs used This cohort was ascertained to assess treatment responsiveness of myoclonic atonic or atonic seizures, and therefore all patients had this seizure type.
in both groups were sodium valproate and clobazam, suggesting patients likely to go into remission would respond to these medications.

| Adaptive behavior
Forty-six of 56 (response rate = 81%) ABAS II Parent Forms in the UK cohort were returned. Extremely low adaptive scores, which indicated a percentile rank ≤ 2nd, were seen in 27 (58.6%) patients for the conceptual, 15 (32.6%) for the social, and 31 (67.3%) for the practical domain. Correspondingly, 32 (69.5%) patients had extremely low general adaptive composite scores. Figure S1 shows the distribution of individual domain scores of the ABAS II.

| Autism
Fifty (38 males, 12 females) of 57 (response rate 87%) SCQ questionnaires were returned in the UK cohort. Fifteen (30%) patients reached the threshold for suspecting ASD with a score of ≥15. The mean score of 10.1 (SD = 7.41) was significantly higher in both males and females when compared with the ALSPAC cohort (P < .0001; see Table S3). In addition, 3di interviews were conducted with one or both parents of 19 patients. Based on the results of the 3di interview, eight patients had an ASD diagnosis. MAE patients reported the most difficulties in the social reciprocity domain, with a mean score of 8.43 (SD = 6.89) within the abnormal range for that subscale. Based on clinician report, two (25%) of eight patients in the Italian cohort and three (12%) of 25 patients in the EuroEPINOMICS cohort reported a diagnosis of ASD. Considering all these measures, 20 (24.1%) of 83 patients reported either a diagnosis or symptoms of ASD.

| Behavior
Fifty of 56 SDQs (response rate = 89%) were returned in the UK cohort. High scores were returned in seven (14%) with emotional problems, 17 (34%) with conduct problems, 19 (38%) with hyperactivity problems, 21 (42%) with peer problems, 18 (36%) with prosocial problems, and 31 (62%) reporting these problems to significantly impact the family and child. The scores were significantly higher (P < .0001) in all domains except emotional symptoms when compared with a normative sample of British children, but were not significant when compared with children with established epilepsy excluding MAE (see Table S4). 26,27 ADHD symptoms were ascertained through clinician reports in the Italian and EuroEPINOMICS cohorts and through the Parent and Teacher CBRS in the UK cohort. Two (25%) of eight patients in the Italian cohort, 13 (50%) of 26 patients in the EuroEPINOMICS cohort, and 13 (32.5%) of 40 UK patients reported ADHD symptoms. Therefore, the estimated prevalence of ADHD symptoms in the entire cohort was 28 (37.8%) of 74 patients. Both parent and teacher CBRS scores were available in 34 UK patients. Table S5 details the results of the parent and teacher CBRS scores.

| Multiple neurodevelopmental comorbidities
Patients for whom data for ID, ASD, and ADHD were available were analyzed for multimorbidity. Fifty of 70 (71.4%) patients had at least one comorbidity, and 22 of 70 (31.4%) had two or more comorbidities (Figure 1).

Likely benign variants and variants of uncertain significance
Six different variants from six different genes were classified as likely benign, and 15 variants in 12 genes were classified as VUSs (see Tables S6 and S7 and Appendix S1 for further discussion of each gene).

| Gene filtering using neuropsychiatric genes
Twenty-one genes from the neuropsychiatric gene set had nonsynonymous variants with CADD > 20 in the MAE cohort. There were no recurrent variant matches with the MAE cohort and the gene set variants. Eighteen genes were deprioritized due to RVIS > 25th percentile and negative ExAC Z score; or conflicting gene function; or inadequate nervous system expression; or unsupportive heterozygosity types or lack of segregation. Three genes, ASH1L, CHD4, and SMARCA2, remained as possible candidate genes (Table 3).
ASH1L (absent, small or homeotic disc 1 like histone lysine methyltransferase) encodes histone methyltransferase, which is involved in histone and chromatin modification and gene regulation. Ash1L is enriched in the brain in mice, and a role in epigenetic modification in brain functioning was implicated when in Ash1L knockout mice the activity-dependent repression of neurexin 1α, a presynaptic adhesion molecule required for synaptic formation, was completely abolished. 39 CHD4 (chromodomain helicase DNA binding protein 4) is an adenosine triphosphate (ATP)-dependent chromatin remodeler involved in epigenetic regulation of gene transcription, DNA repair, and cell cycle progression. 40 It is also a paralogue of CHD2, which is associated with MAE. 12 SMARCA2 mutations are associated with Nicolaides-Baraitser syndrome. It is one of six genes that encode the SWItch/sucrose nonfermentable like chromatin remodeling complex and alters chromatin structure through ATP hydrolysis. Subject 3003 301, with a de novo p.Gln1241Glu SMARCA2 variant, was subsequently identified as also having clinical features compatible with a diagnosis of Nicolaides-Baraitser syndrome along with MAE. 23

| DISCUSSION
This paper uses a large cohort of patients with MAE (n = 101) to describe the phenotypic variability in seizures and a high incidence of neurodevelopmental comorbidities. We report a genetic cause in 14% with some associated motor symptoms. The findings pose questions for the nosological boundaries of MAE and overlap with genetic generalized epilepsies (GGEs) and developmental epileptic encephalopathies (DEEs).
Although the most frequent seizure types in MAE are as expected (myoclonic atonic or atonic seizure, 100%; GTCS, 72.7%; and myoclonic seizures, 68.2%), 3,5,6,41 a minority of MAE patients have focal seizures, some (6.9%) with additional focal epileptiform activity on EEG. Focal EEG features have been reported in up to 39% of MAE cohorts 40 and have been proposed as a marker of poor prognosis in MAE. 42 However, we were not able to confirm this hypothesis in this cohort, as two of seven patients with focal EEG abnormalities achieved seizure remission, compared to 22 of 70 patients without focal abnormalities.
Approximately 20% of patients had evidence of a DEE with developmental or speech delay prior to onset of seizures, a much higher frequency of antecedent impairments than was previously appreciated. ID was reported in 61 (62.8%) of 97 patients, and the impact of these deficits is reflected in extremely low adaptive functioning in 32 (69.5%) of 46 patients. ASD symptoms occurred in 24.1%, possibly due to predisposing factors such as ID, early onset of seizures, and EE. ADHD symptoms, predominantly inattentive, were identified in 37.8%, in keeping with other pediatric epilepsies.
The variation in phenotype overlaps with other epilepsy syndromes, and we recognize that the current concept of MAE provides a phenotypic bridge between GGEs and DEEs, including Lennox-Gastaut syndrome. For example, antecedent neurodevelopmental impairments and autistic and cognitive deficits are more characteristic of DEE. 43  De novo 0 (P), 1 (P), 32 switching. 40 However, we believe that the MAE concept remains broadly useful to guide treatment and prognosis. Unsurprisingly, cases with neurodevelopmental symptoms or neurological signs were more likely to have an identified genetic etiology. ID was reported in 11 of 12 (3/3 candidate) patients, ASD in five of 12 (2/3 candidate) patients, and ADHD in three of 12 (1/3 candidate) patients. ID, ASD, and epilepsy share causative genes and biological pathways including gene transcription regulation, neurotransmission, and maintenance of synaptic structure. These neurodevelopmental comorbidities may be a primary feature of the genetic disease rather than secondary to disturbance of brain function due to excessive epileptiform activity (an epileptic encephalopathy). Abnormal neurological findings were reported in 21 of 92 patients, higher than the 12% in previous smaller MAE series. 2,41 Ataxia and tremor have been recognized in MAE patients with pathogenic variants in SLC2A1, 11 STX1B, 15 and SLC6A1 16 and occurred in our probands with pathogenic variants in SYNGAP1 (n = 3) and MECP2 (n = 1). These motor signs may be useful in suggesting specific genotype correlation in MAE.
Gene-specific features were observed. Our subjects with SYNGAP1 variants all had seizure onset in infancy, ID, ataxia, and hypotonia consistent with a SYNGAP1 encephalopathy. 13 The two female subjects with p.Leu421fs and p.Arg322* KIAA2022 variants both have MAE comorbid with ID. Subject 291J with a de novo MECP2 p.Pro-225Thr variant has overlapping clinical features found in classic Rett syndrome of cerebellar signs and an EEG with a disorganized background. De novo MECP2 variants are recognized as causative for MAE and other epilepsies. 21,44 A potential clinical implication is that children with Rett syndrome have an increased risk of life-threatening arrhythmias associated with prolonged QT c interval and should avoid drugs known to prolong QT c interval. It is unknown whether this also applies to patients with an MECP2-associated epilepsy.
Atypical EEG features should prompt a search for a genetic etiological link as demonstrated in our cases with KCNA2 and KCNB1. Moreover, subject 00595 with an SLC6A1 variant had an unusual EEG feature of posterior eye closure sensitivity with 3-to 4-Hz spike-wave complexes on the posterior third of the head provoked by eye closure, similarly identified in one of seven other MAE cases with SLC6A1 variants. 16 Finally, we identified ASH1L and CHD4 as candidate genes in this paper. Seven other de novo ASH1L variants have been reported in ID/ASD patients. 45,46 Only one case has been described with an epilepsy phenotype; the patient had a missense ASH1L variant of unknown inheritance, leaving its pathogenic role uncertain. 46 The highly conserved p.Arg1342* variant identified here is located between two annotated protein domains, like other reported mutations. 46 A de novo CHD4 variant has been reported in a single patient with EE, and individuals with overlapping phenotypes of developmental delay, hearing loss, macrocephaly, distinct facial dysmorphisms, palatal abnormalities, ventriculomegaly, and hypogonadism but no epilepsy. 33,47 The variant p.His896Arg identified in subject 00526 is located within the ATPase/helicase domain and may disrupt the ATPase activity of CHD4. Further studies are required to understand the phenotypic variability of CHD4 mutations.
There were several limitations in this study. The ascertainment of patients may have been biased toward more severe phenotypes. Patients were recruited from different European countries, which may have contributed to phenotypic heterogeneity. Many cases were phenotyped through screening questionnaires, which does not substitute for direct neuropsychological testing. Whole exome sequencing has poor capacity to identify structural variants and somatic mosaic variants, and this was not investigated. Additionally, in silico prediction tools were available only for missense variants, and the role of indels and frameshift variants may be underestimated.
Several lines of enquiry remain for the unsolved MAE cases. While there are remaining undiscovered genes within the exome space, the impact of variants in noncoding regions, regulatory regions, and microRNAs remain to be interrogated and may be revealed with whole genome sequencing. Copy number variations (CNVs) in this cohort were not investigated, but from previous studies we would expect a frequency of pathogenic CNVs in 5% of cases. 48,49 Somatic mosaic variants are increasingly recognized in early onset genetic disease and remain challenging to investigate. 50 Susceptibility genetic factors and rare variants contributing to specific phenotypic features such as seizure types and EEG features remain undetermined.
In summary, we demonstrated that MAE is associated with significant neurodevelopmental comorbidity and the yield of identifying a possible monogenetic etiology was about 14% in this cohort. An identifiable genetic etiology was more likely in patients with associated neurodevelopmental disorders and brings the possibility of personalized medicine closer, such as the use of the ketogenic diet in patients with GLUT1 deficiency, sodium channel blockers in patients with