Skip to main content
Browse Subject Areas

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Next-generation sequencing in childhood-onset epilepsies: Diagnostic yield and impact on neuronal ceroid lipofuscinosis type 2 (CLN2) disease diagnosis


Epilepsy is one of the most common childhood-onset neurological conditions with a genetic etiology. Genetic diagnosis provides potential for etiologically-based management and treatment. Existing research has focused on early-onset (<24 months) epilepsies; data regarding later-onset epilepsies is limited. The goal of this study was to determine the diagnostic yield of a clinically available epilepsy panel in a selected pediatric epilepsy cohort with epilepsy onset between 24–60 months of life and evaluate whether this approach decreases the age of diagnosis of neuronal ceroid lipofuscinosis type 2 (CLN2). Next-generation sequencing (NGS)-based epilepsy panels, including genes associated with epileptic encephalopathies and inborn errors of metabolism (IEMs) that present with epilepsy, were used. Copy-number variant (CNV) detection from NGS data was included. Variant interpretation was performed per American College of Medical Genetics and Genomics (ACMG) guidelines. Results are reported from 211 consecutive patients with the following inclusion criteria: 24–60 months of age at the time of enrollment, first unprovoked seizure at/after 24 months, and at least one additional finding such as EEG/MRI abnormalities, speech delay, or motor symptoms. Median age was 42 months at testing and 30 months at first seizure onset; the mean delay from first seizure to comprehensive genetic testing was 10.3 months. A genetic diagnosis was established in 43 patients (20.4%). CNVs were reported in 25.6% diagnosed patients; 27.3% of CNVs identified were intragenic. Within the diagnosed cohort, 11 (25.6%) patients were diagnosed with an IEM. The predominant molecular diagnosis was CLN2 (14% of diagnosed patients). For these patients, diagnosis was achieved 12–24 months earlier than reported by natural history of the disease. This study supports comprehensive genetic testing for patients whose first seizure occurs ≥ 24 months of age. It also supports early application of testing in this age group, as the identified diagnoses can have significant impact on patient management and outcome.


Neurologic and metabolic disorders that include epileptic seizures are among the most common genetic disorders presenting in early childhood. Major improvements in sequencing technologies during the last two decades have significantly improved our understanding of the molecular genetic origins of epilepsies and sparked a new era in the genetic diagnosis of epilepsy. The etiology of epileptic seizures can be difficult to discern. Although epilepsy can result from trauma, tumors or infections, a significant number of patients with epilepsy are thought to have a causative genetic factor [1,2].

The spectrum of epilepsies can vary from self-limited and treatable to severe, progressive, drug-resistant forms. In addition to channelopathies and neurotransmitter receptor and transporter defects as causes of epilepsy, >300 inborn errors of metabolism (IEM) can present with epileptic seizures, adding a further layer to the differential diagnosis [3]. Given the complexity of the disease spectrum, the limitations of traditional genetic diagnostics, the lack of recently updated recommendations on genetic testing in pediatric epilepsy, and limited access to comprehensive genetic testing, it is not surprising that patients often experience years of delay before a genetic diagnosis is established and the underlying etiology is understood. Nonetheless, early molecular diagnosis is essential for the personalized management and targeted therapy of epilepsy and improved outcomes [4].

Several studies have addressed the utility of next-generation sequencing (NGS) in the diagnosis of epilepsy. These studies varied in terms of the patient cohort size, patient demographics and phenotypic spectrum, the testing strategy, the variant interpretation approach, and the quality and performance of the NGS technology. These factors contribute to the variability observed in the diagnostic yields and conclusions. The majority of these studies have focused on patients with early-onset seizures (birth to two years) or have analyzed unselected cohorts of patients with large variability in age of onset, severity, and outcome. Studies with cohorts of >100 patients [511] have generally demonstrated a lower diagnostic yield when compared to smaller studies [1217]. Several studies have provided evidence that the molecular diagnostic yield is strongly correlated to the age of onset and epileptic syndrome, with the highest yield being among patients with developmental and epileptic encephalopathies (DEE) and seizure onset in the neonatal period.

As many NGS studies have focused on early-onset and severely affected patients, our understanding about the genetics of patients with seizure onset after two years of age is limited. Only a few studies have specifically addressed this age group, and those had small patient numbers. Oates et al reported a 4% diagnostic yield in 46 patients with seizure onset after two years of age and suggested that NGS panels are not cost-effective in this patient group [18]. In another study, Moller et al demonstrated a 14% diagnostic yield in a cohort of 29 patients with seizure onset between two and nine years [19]. Both studies demonstrated lower diagnostic yields when seizure onset was after two years of age. In summary, studies addressing the clinical utility of genetic diagnostics for patients with seizure onset after the neonatal period, and especially after two years of age, are lacking.

The goal of this study was to evaluate the clinical utility and diagnostic yield of NGS-based genetic testing in a patient age group that has not been studied extensively in the past. Patients with ages ranging from 24 to 60 months of life with their first unprovoked epileptic seizure between 24 and 60 months of age were eligible for the program. We also evaluated whether this genetic diagnostic approach could decrease the age of diagnosis of neurodegenerative diseases such as neuronal ceroid lipofuscinosis (CLN2).

Materials and methods


The cohort included 211 consecutive patients referred for genetic testing as part of the BioMarin Pharmaceuticals Inc. sponsored testing program in 2018 and 2019 ( The testing program was offered in European and Middle Eastern countries (68 countries). Thirty-four patients (16.1%) were from Middle Eastern countries (Saudi Arabia, Oman, Kuwait and United Arab Emirates), 19 patients (9.0%) were from Israel, and 158 (74.9%) were from Europe. Patients had to be between 24 and 60 months of age and had to have their first unprovoked seizure between 24 and 60 months of age. Additionally, patients had to have at least one of the following findings: abnormal MRI (such as, but not limited to, cerebellar atrophy, cerebral atrophy, periventricular white matter hyperintensity) abnormal EEG, history of language delay or regression, motor impairment or regression. In addition, the requisition form completed by the referring physician also allowed them to specify additional findings in their patient such as developmental delay, behavioral abnormalities, sleep disturbances, visual impairment and other sensory impairment. These features were not defined as they were not required for inclusion in the study nor were they mandatory fields for completion of the order form. Detailed clinical information regarding the patients was not available to the authors.

Referring physicians were responsible for determining that their patient met eligibility criteria and were required to sign a statement to this effect. Testing did not proceed unless the patient met eligibility criteria and the signed statement was received. The study did not stipulate specific exclusion criteria, however, sites were encouraged to enroll patients with a suspected genetic etiology for seizures to ensure maximum benefit for those tested.

Informed parental consent was obtained by the ordering health care provider and attested to in writing for all patients who participated in this testing program. This work was reviewed by the Western Institutional Review Board (IRB) and received an excemption determination (WCG IRB Work Order #1-1379311-1).

Next-generation sequencing

The study was initiated with a 194-gene panel (S1 Table); 25 patients (11.8%) were analyzed with this panel. The 194-gene panel analysis was done using the OS-Seq™ (oligonucleotide-selective sequencing) NGS method on the NextSeq sequencing system (Illumina) [20]. In this analysis, the mean sequencing depth was >170x and 99.6% of target nucleotides were covered with >20x sequencing depth. The remaining 186 patients (88.2%) were analyzed with a 283-gene epilepsy panel (S1 Table). The 283-gene panel was carved out of an in-house tailored IDT-based whole-exome sequencing platform performed on the NovaSeq sequencing system (Illumina). The mean sequencing depth was >200x and 99.6% of target nucleotides were covered with >20x sequencing depth. The platform included custom oligonucleotides targeting 70 non-coding/deep intronic variants associated with epilepsy (S2 Table).

Sequencing-derived raw image files were processed using a base-calling software (Illumina) and the sequence data was transformed into FASTQ format. Clean sequence reads of each sample were mapped to the human reference genome (GRCh37/hg19). Burrows-Wheeler Aligner (BWA-MEM) software was used for read alignment. Duplicate read marking, local realignment around indels, base quality score recalibration and variant calling were performed using GATK algorithms (Sentieon).

Sanger sequencing

Bi-directional Sanger sequencing was used to confirm likely pathogenic (LP) and pathogenic (P) sequencing variants. All primers are available upon request. The sequence variant analysis pipeline was validated in the CLIA- and CAP-accredited Blueprint Genetics diagnostic laboratory.

Copy-number variant analysis

Copy-number variant (CNV) analysis was performed bioinformatically concurrently for all patients from the NGS data using a commercially available bioinformatic pipeline CNVkit [21] and an in-house developed deletion caller based on read-depth to improve the detection of small CNVs. Further information about this proprietary deletion caller is available upon request. All heterozygous CNVs affecting < 10 target exons and hemizygous deletions < 3 target exons were confirmed using quantitative-PCR assays. The CNV analysis pipeline was validated in the CLIA- and CAP-accredited Blueprint Genetics diagnostic laboratory. The sensitivity to detect single exon deletions was validated to be 71.5% (NextSeq OS-Seq assay) and 92.3% (NovaSeq WES assay).

Variants were classified according to a point-based, modified adaptation of the Association for Molecular Pathology/American College of Molecular Genetics and Genomics guidelines [22] with evidence from population and gene-/disease-specific databases, in silico prediction tools (including PolyPhen [23], SIFT [24] and Mutaster [25]), our in-house variant database, multiple publicly and commercially available mutation databases and appropriate scientific literature as the foundation for scoring) as outlined in the Blueprint Genetics website ( The Blueprint Genetics classification scheme is similar to the ACMG/AMP guidelines in that multiple independent lines of evidence must be met (e.g. rare in population databases, predicted deleterious by in silico software, segregation with disease, de novo in a patient with no family history, damaging impact shown in well-established functional studies, etc.) to achieve a likely pathogenic or pathogenic classification, and several of these criteria are equivalent in both schemes. A test result was considered diagnostic when the patient was found to have one or two pathogenic or likely pathogenic variants in a single gene, depending on the mode of inheritance. Highly suspicious variants of uncertain significance (VUS) are listed in Table 4. These VUS were considered suspicious based on 1) a strong and specific correlation between the gene and patient’s phenotype, 2) the variant being novel or extremely rare in the Genome Aggregation Database (gnomAD) control cohorts, and 3) in silico predictions supporting pathogenicity or the amino acid position in question being highly conserved in mammals and evolutionary more distant species, suggesting that the position does not tolerate variation. All reported variants were de-identified and shared in the ClinVar database.


A total of 211 patients were included in this study. The mean age was 3.5 years (42 months [range 24–60 months]) (Table 1). Males represented 58.8% of the group. The mean age at epilepsy onset was 2.5 years (30 months [(range 24–57 months]). The mean time from seizure onset to taking part in this genetic testing program was 10.3 months. At the time of inclusion, language delay was reported in 70.1% of patients, motor disturbance in 55.9%, and developmental delay in 48.3%. In addition, 42.7% of patients had abnormal EEG and 30.8% had abnormal MRI findings. Behavioral abnormalities were described in 41.7% of patients, sleep disturbances in 25.6%, and vision impairment in 10.4%. A family history of seizures was reported in 16.6% of cases.

In this study, we established a diagnosis in 43 patients, resulting in a diagnostic yield of 20.4%.(Table 1). We detected a highly suspicious VUS in an additional 17 (8.1%) cases. These are variants that would be reclassified as likely pathogenic if they are shown to be de novo in a proband with autosomal dominant disease or if in trans with a known disease-causing variant in a proband with autosomal recessive disease. The demographic and clinical characteristics of all categories (total cohort, diagnosis, and suspicious VUS) are listed in Table 1.

Among the 211 patients tested, our NGS technology detected a pathogenic or likely pathogenic CNV in 11 patients (5.2%) (Table 2). In the diagnosed cohort, 25.6% of patients were identified to have a disease-causing CNV (Fig 1). The deletions ranged in size from a 242bp intragenic deletion to a 9Mb deletion affecting a number of genes. The smallest deletion was a 242bp deletion in CACNA1A. It was identified in a child (ID 01) with myoclonic and atonic seizures, hemiparesis, ataxia, and delayed speech. A heterozygous single exon deletion of similar size was detected in PPT1 together with a pathogenic nonsense variant p.(Arg151*) in a patient (ID 10) originally suspected to have Angelman syndrome who suffered from atonic seizures, motor disturbances and developmental delay. This was the only patient who was found to have both a disease-causing CNV and SNV. A 766bp hemizygous deletion in SLC6A8, affecting exons 10–13, was detected in a child (ID 02) with speech delay and focal seizures starting at the age 36 months. A 122,226bp deletion in MEF2C, affecting the full protein coding region of the gene, was found in a child (ID 09) with language delay, abnormal MRI and EEG findings, and seizure onset at 26 months. A heterozygous 2.1Mb deletion involving the short arm of chromosome 1 was detected in a child (ID 03) with treatment-resistant epilepsy with onset at 27 months, severe intellectual disability, global developmental delay, scoliosis, microcephaly, and dysmorphic features. This deletion partially overlaps the 1p36 deletion syndrome and includes 37 genes, including seven OMIM Morbid genes: CEP104, GABRD, GNB1, PEX10, PRDM16, SKI, and SMN1. Mowat-Wilson syndrome, caused by a 9Mbp deletion in chromosome two including all of ZEB2, was identified in a four-year-old girl (ID 07) with developmental delay, microcephaly, and seizure onset at 28 months. Three patients (IDs 04, 05, and 06) harbored heterozygous chromosome 15 deletions, ranging from 4.87 to 6.32Mbp, involving the Angelman/Prader-Willi syndrome critical region. All of these patients had developmental delay and seizure onset at 24 months of age. Interestingly, a patient (ID 08) with developmental delay and seizures was diagnosed with a 72Mbp duplication (1q25.2-q44) involving 786 genes on chromosome one and a 40Mbp deletion (Xq23-q28) involving 480 genes on the X chromosome, suspicious for an unbalanced translocation. Lastly, one patient (ID 11) was found to have a 772kb deletion on chromosome 16 consistent with 16p11.2 microdeletion syndrome (Table 2).

Fig 1. The contribution of CNVs, SNVs and Indels in patient’s receiving a diagnosis.

In addition to the larger CNVs observed in 11 patients, we identified a clinically relevant single-nucleotide variant (SNV) or small deletion explaining the symptoms in 32 patients (Table 3). Interestingly, nine (21%) of the diagnosed patients had lysosomal storage diseases, specifically neuronal ceroid lipofuscinoses (NCLs). Six patients (14.0%; IDs 36–41) were diagnosed with neuronal ceroid lipofuscinosis type 2 (CLN2) caused by variants in Tripeptidyl peptidase 1 (TPP1)/(CLN2); five patients were homozygous for the c.622C>T p.(Arg208*) nonsense variant, while the sixth patient was homozygous for the c.533del p.(Pro178Glnfs*5) frameshift variant. The onset of seizures in the CLN2 cohort varied from 35 months to 44 months of age (Table 3), and genetic testing was carried out at 45 months to 56 months. All of these patients also had language delay and motor disturbances. One (2.3%) patient had a CNV and a nonsense variant in the PPT1 gene (Table 2). Variants in this gene are associated with infantile, juvenile, and adult onset NCL (CLN1 disease). Lastly, in the lysosomal storage cohort, testing also identified two (4.7%; IDs 23 and 24) patients with pathogenic variants in MFSD8 associated with neuronal ceroid lipofuscinosis type 7 (CLN7 disease). Both patients were male, and seizure onset was at 42 months age in one patient and 57 months in the other (Table 3).

Additionally, we identified four (9.3%; IDs 19–22) patients with MECP2 variants associated with Rett syndrome (Table 3). Age of seizure onset varied from 30 to 42 months and all of these patients had developmental delay and motor disturbances. Three (7%; IDs 27–29) patients were found to have SCN1A variants; they experienced their first unprovoked seizures between the age of 24–28 months of life, and all had developmental delay. Detailed phenotypic information was not available for these patients, however, one of these patients was noted to have a provoked (febrile) seizure at 11 months of age. KCNA2 variants were identified in two patients (IDs 30 & 31) whose first unprovoked seizures were at 25 and 32 months, respectively. In addition, both had developmental delay, ataxia, and MRI abnormalities. Two (4.7%; IDs 13 & 14) patients with seizure onset at 35 and 36 months who exhibited ataxia, developmental delay, feeding, and behavioral problems were found to have DEE-associated variants in CHD2. The cohort also included one (2.3%; ID 15) patient with lissencephaly caused by a heterozygous splice site variant in DCX. This patient’s first unprovoked seizure occurred at the age of 25 months and they had severe developmental delay and MRI abnormalities. One patient (ID 18) with a heterozygous KIF1A variant experienced their first unprovoked seizure at the age of 44 months and also had developmental delay. A heterozygous variant in PCDH19 gene was identified in one female (ID 25) with MRI abnormalities, behavioral abnormalities, and motor difficulty. This patient’s first unprovoked seizure occurred at the age of 24 months. A homozygous missense variant was observed in PIGT in one patient (ID 26) with seizure onset at 27 months and with vision impairment, language delay, ataxia, and motor difficulty. Two patients (IDs 33 & 34) with STXBP1 variants had their first unprovoked seizure at 24 and 35 months, respectively. In addition, both suffered from developmental delay. One patient (ID 43) with a TSC1 duplication with seizure onset at 25 months also had mild global developmental delay, microcephaly, and hypopigmented macules. A four-nucleotide duplication was detected in SYNGAP1 causing seizure onset at 41 months, vision impairment, motor impairment, language delay, and behavioral abnormalities (ID 35). A child (ID 42) with a homozygous missense variant in the Aicardi-Goutieres syndrome-associated gene, RNASEH2B, presented with optic atrophy, developmental delay, abnormal MRI and seizure onset at 36 months. Finally, a homozygous missense variant in SLC19A3, associated with biotin-thiamine-responsive basal ganglia disease, was identified. This patient (ID 32) suffered from developmental delay, ataxia, and seizure onset at 48 months (Table 3).

Among the 32 patients with SNVs or small deletions and duplications, missense variants were the most common variant type making up 38.4% of variants. Nonsense variants accounted for 34.1%, frameshift for 13.6% and splice site variants for 9.1% of all diagnostic variants (Fig 2). Indels and synonymous variants were both found with a frequency of 2.3%. Fifty-six percent of patients had variants consistent with autosomal dominant inheritance, 28% with autosomal recessive inheritance, and 16% with X-linked inheritance (Fig 3). Interestingly, 11 patients (25.6%) were diagnosed with an IEM; nine patients with NCL, one with creatinine deficiency syndrome (SLC6A8), and one with thiamine metabolism deficiency (SLC19A3). Among the patients with an IEM, the inheritance pattern was autosomal recessive in 91% and X-linked in 9%. This was markedly different among the non-IEM cases, where only 6.3% of variants were autosomal recessive, 75% were autosomal dominant, and 18.7% were X-linked. Disease-causing variants analyzed in this work were submitted to the ClinVar repository (SUB7819159).

Fig 2. Type of likely pathogenic and pathogenic sequencing variants.

In addition to the 43 patients with genetic diagnoses, our analysis revealed 17 additional patients (8.1%) with a highly suspicious VUS (Table 4). In these cases, the identified variants were novel and absent in reference databases, were predicted deleterious by bioinformatic predictions and showed a strong and specific correlation between the gene and the patient’s phenotype. The genes with a suspicious VUS were ATRX, GABRA1 GRIA3, GRIN1, KCNA2, KCNQ2, KCNQ3, KDM5C, SCN1A, SCN8A, SLC6A1, SLC6A8, SLC9A6, STXBP1, UNC80, and ZEB2. Further analyses, such as parental testing or functional assays, would be needed in order to reclassify the VUS detected in these genes to likely pathogenic. Interestingly, at least two patients (11.8%) within this group had variants in genes responsible for an IEM with associated neurological morbidity (SLC6A1 and SLC6A8).

Table 4. Highly suspicious variants of uncertain significance (VUS).

In this cohort, none of the patients who received a diagnosis were found to have one or more of the custom-added 70 non-coding/deep intronic variants associated with epilepsy.


To our knowledge, this represents one of the first studies in patients with epilepsy onset between 24 to 60 months of life with comprehensive genetic testing [26]. The inclusion criteria were selected according to common early findings of CLN2 disease: unprovoked seizure after 2 years of age and at least one of motor disturbance, speech delay, MRI abnormalities and abnormal EEG. Indeed, the group of patients molecularly diagnosed with this panel mostly presented with motor disturbance (79.1%) and developmental delay (65.1%). In this patient cohort, we demonstrated a molecular diagnostic yield of 20.4% [21], which is higher than the 4–14% presented in previous studies of patients with epilepsy onset after two years of age [8,18]. One of the reasons for the difference could be the characteristics of the cohort. In this study, only patients with seizures and at least one additional neurological finding were included and thus this cohort may be more severely affected. Further, patients in the Moller et al [8] study ranged in age from 2 years– 9 years. As the likelihood of a genetic diagnosis decreases with increasing age of seizure oneset, this may also contribute to the difference in diagnoses. The diagnostic yield may also be affected by the quality and performance of the sequencing and bioinformatics pipeline in addition to the number of genes included in the analysis. In our study, 186 (88.2%) of the patients were analyzed with a panel of 283 genes while 25 (11.8%) were analyzed with a 195-gene panel. Both panels had >20x coverage in >99.5% of target nucleotides. In previous studies, the panel size varied from 29 to 102 genes and the quality metrics indicated that 3–5% of target regions were not covered [8,18]. These metrics represent a significant difference between studies and are likely to contribute to the differences in yield. In addition, several genes that yielded diagnostic results in our study, such as the NCL genes, are not commonly covered by standard epilepsy panels and were also missing from the analyses by Moller et al. and Oates et al [8,18].

A panel testing approach was selected for this study as we wanted to evaluate the clinical utility and diagnostic yield of panel-based testing within this age group. While we do acknowledge that a high-quality WES testing strategy may have increased the diagnostic yield in this cohort, a panel-based approach remains the standard of care for the evaluation of patients with epilepsy.

A significant contributor to yield was the high prevalence of CNVs among diagnosed patients. In this study, >5% of all patients and 25.6% of diagnosed patients harbored a clinically significant CNV. Small intragenic deletions represented 27.2% of all CNVs. Several studies have previously addressed the role of CNVs in patients with epilepsy. Mefford et al. [27] used a genome wide array CGH with average probe spanning 38kb to study a cohort of 517 mixed type epilepsies. They identified a rare, potentially causative CNV in 8.9% of tested patients. That study was published in 2010 and thus lacked access to reference databases to assist in interpretation of identified CNVs, and the technology utilized did not have adequate performance for the detection of smaller intragenic CNVs [23]. Niestroj et al [28] studied a cohort of >10,000 subjects with one of four different epilepsy phenotypes with a genome wide SNP array and identified a likely pathogenic CNV is 1.15%-2.88% of patients, depending on the epilepsy phenotype. The authors point out that especially small structural variants were not detectable with the available genotyping platform. A recent study of >8500 patients with epilepsy used an exon-level array CGH with enhanced performance of 70 selected epilepsy genes [10]. The authors reported that <2% of epilepsy patients carried a clinically relevant CNV, contributing to 10% of diagnosed cases [10]. Strikingly, approximately 60% of detected CNVs were intragenic with one- to two-exon CNVs (37.0%).

The detection of CNVs has shifted towards molecular genetic technologies given the recent advances in techniques. While conventionally available chromosome microarrays and SNP arrays have CNV detection capabilities of ~20kb– 200kb, whole exome sequencing has been reported to have CNV detection capabilities of 100bp—~150kb [29]. The assays used in this study were validated to detect single exon deletions with a sensitivity of 71.5% (NextSeq OS-Seq assay) and 92.3% (NovaSeq WES assay) and detected CNVs ranging in size from 242bp—>70Mbp. The application of molecular techniques for the detection of CNVs has the potential to identify both exon-level CNVs in addition to large microdeletions and microduplications, making it a powerful tool for the simultaneous evaluation of sequence variants and CNVs.

Despite the significant difference in the patient cohorts and technologies used, our study, in addition to the Lindy et al study [10], strongly support the role of incorporating high-resolution, molecular-based CNV detection in the genetic testing of patients with epilepsy. Considering the high prevalence of small intragenic CNVs, typical and commonly used chromosomal array CGH tests may not be sufficient for epilepsy patients. Laboratories that put a high premium on sequencing coverage, sequencing uniformity, and bioinformatic algorithm development can offer similar, or better, CNV detection than CGH arrays [30].

Research and innovations in the field of rare diseases, including epilepsies, are improving the clinical utility of genetic testing. A genetic diagnosis can, in many cases, be used for targeting and optimizing therapy. Moreover, the declining cost of sequencing (80–90% over the past 10 years) have increased the use of genetic testing in diagnosing patients [31]. In our study, one of the striking findings was the high prevalence of CLN2 disease cases among these patients (14% of diagnosed patients). This is especially interesting since a targeted enzyme replace therapy, cerliponase alfa, has been approved by FDA and EMA to treat patients with CLN2 disease [32]. The treatment, by intraventricular infusion of cerliponase alfa in patients with CLN2, resulted in less decline in motor and language function than in historical controls [32]. CLN2 disease is caused by deficiency of the lysosomal enzyme TPP1 and is characterized by rapid psychomotor decline and epilepsy. Seizures are one of the first symptoms noted in patients with CLN2 disease typically occurring after two years of age (median age 35 months), often preceded by history of language delay. Subsequently, substantial loss of language and motor skills happen by the age of four to five years and death between six and twelve years of age. Unfortunately, due to low awareness of the disease, non-specific early signs and limited access to specific testing in some regions [33,34] the diagnosis of CLN2 disease is usually reached late at a mean age of five years (median age 54 months). Remarkably, patients with CLN2 disease in this study were diagnosed at a mean age of 38.7 months with a mean time from seizure onset to diagnosis of 10 months, a significantly shorter time compared to natural history diagnostic data [34], demonstrating that use of a comprehensive epilepsy gene panel can effectively and timely diagnose patients with CLN2 disease. Interestingly, consistent with our study, three other reports have recently shown TPP1 (CLN2) pathogenic or likely pathogenic variants and CLN2 diagnosis being a relatively common finding when using NGS panels for diagnostics of pediatric seizure patients [10,26]. These findings stress the importance of the genes included in the selected panel when evaluating the diagnostic yield and appropriateness of the panel.

Other treatable disorders within our cohort included biotin-thiamine-responsive basal ganglia disease caused by variants in SLC19A3 (Table 3). Without biotin and thiamine replacement therapy, the outcome for these patients can be lethal [35]. Recently, the targeted therapy everolimus was approved for the treatment of TSC1 related disease [36]. SLC6A8 creatinine transporter deficiency can be effectively treated with L-arginine, glycine and creatinine supplements [37]. In addition, the EMA and FDA have approved stiripentol and cannabidiol for the treatment of Dravet syndrome, mainly caused by variants in the SCN1A gene [3840]. Moreover, it has been suggested that patients with Dravet syndrome and patients with SCN1A variants causing loss of function of the Na+ channel should avoid Na+ channel blocking anti-epileptic drugs. Studies on existing anti-epileptic medications have provided evidence on the optimal treatment combinations for epileptic disorders related to PCDH19 and STXBP1 [3134].

In addition to existing targeted management strategies, there were several diagnoses with ongoing interventional clinical trials. These included disorders caused by variants in CACNA1A, CDKL5, MECP2, and SYGNAP1. In brief, at least 27 (63%) patients were diagnosed with a disorder for which there is targeted treatment, evidence for optimizing pharmaceutical treatment, or on-going clinical trials available (S3 Table).

A potential limitation of the study is that the patients were referred to us as part of a no-cost testing program. The program had strict eligibility criteria, which were outlined on the requisition form, however, patients were assessed and referred for testing by their own health care provider. Referring physicians were required to sign a statement confirming that their patient met the eligibility criteria but systematic evaluation by a pediatric neurologist or neuroradiologist was not required. Misapplication of this no-cost testing program could lead to atypical or misleading correlations between genotype and phenotype. One area of interest was the age of the first unprovoked seizure. For most of the molecular diagnoses, the age of seizure onset between two and five years is consistent with the described disease phenotype. However, in some cases, the findings may suggest a novel or rare/atypical form of the disease. Variants in CDKL5 are typically associated with early onset epileptic encephalopathy with seizures usually starting between three and six months of age [30,35,36]. According to the referral, our CDKL5 patient had her first unprovoked seizure at the age of 25 months, which represents an atypical age of onset (Table 3). One patient with a homozygous PIGT variant had their first unprovoked seizure at the age of 27 months (Table 3), which is later in onset compared to the published cases with PIGT-related multiple congenital anomalies hypotonia-seizure syndrome 3 (MCAHS3) [41]. Interestingly, our cohort included three SCN1A cases and two SCNA2 cases (Table 3). The reported age of unprovoked seizure onset varied from 24 to 28 months. One of the patients with an SCN1A variant was reported to have their first provoked (febrile) seizure at 11 months of age, consistent with the natural history of SCN1A-related seizures [3840]. Detailed medical records regarding the other two patients were not available to the authors so it is not clear if they had provoked seizures prior to 24 months of age or if they represent a broadening of they phenotype associated with SCN1A variants. Our findings support the hypothesis that genes classically associated with early onset epilepsy should be included as part of the routine genetic diagnostic process in children with onset of seizures after two years of age, and especially when seizure symptoms are associated with other neurological abnormalities.

A further limitation of this study was the lack of segregation analysis. Variants classified as pathogenic or likely pathogenic had sufficient evidence for classification without parental testing. However, 17 patients (8.1%) in this cohort were found to have a highly suspicious VUS. Testing of the parents in these patients would have likely been sufficient for classification as likely pathogenic, resulting in a significant addition to diagnostic yield. Thus, when possible, testing of parents should be considered to resolve the classification of suspicious VUS.

In conclusion, in children with epilepsy who had their first unprovoked seizure at 24 to 60 months of age and at least one other neurological finding, NGS testing demonstrated a diagnostic yield of 20.4%. Our findings further the importance of the early use of genetic testing, including high-resolution CNV detection, in this age group to efficiently identify severe disorders with targeted management available such as CLN2. Considering the high prevalence of CLN2, this study supports the addition of TPP1 (CLN2) and other genes linked to NCLs in diagnostic NGS-based epilepsy panels.

Supporting information

S1 Table. Gene content of panels used in the analysis.


S2 Table. Non-coding variants that are covered by the NGS analysis.


S3 Table. Next best action after molecular diagnosis.



We thank all the referring health care providers and patients for participating in the program and enabling this study. We acknowledge the support of past and present members of Blueprint Genetics and BioMarin Pharmaceuticals, especially Nicole Miller, that have contributed to the study.


  1. 1. Dyment D. A. et al. Whole-exome sequencing broadens the phenotypic spectrum of rare pediatric epilepsy: a retrospective study. Clin Genet 88, 34–40 (2015). pmid:25046240
  2. 2. Ostrander B. E. P. et al. Whole-genome analysis for effective clinical diagnosis and gene discovery in early infantile epileptic encephalopathy. Npj Genom Medicine 3, 22 (2018). pmid:30109124
  3. 3. Campistol J. Epilepsy in Inborn Errors of Metabolism With Therapeutic Options. Semin Pediatr Neurol 23, 321–331 (2016). pmid:28284393
  4. 4. Leach J. P. Treatment of epilepsy—towards precision. F1000research 7, F1000 Faculty Rev-1932 (2018). pmid:30613377
  5. 5. Carvill G. L. et al. Targeted resequencing in epileptic encephalopathies identifies de novo mutations in CHD2 and SYNGAP1. Nature genetics 45, 825–830 (2013). pmid:23708187
  6. 6. Mercimek-Mahmutoglu S. et al. Diagnostic yield of genetic testing in epileptic encephalopathy in childhood. Epilepsia 56, 707–716 (2015). pmid:25818041
  7. 7. Trump N. et al. Improving diagnosis and broadening the phenotypes in early-onset seizure and severe developmental delay disorders through gene panel analysis. Journal of Medical Genetics 53, 310–317 (2016). pmid:26993267
  8. 8. Møller R. S. et al. Gene Panel Testing in Epileptic Encephalopathies and Familial Epilepsies. Molecular Syndromology 7, 210–219 (2016). pmid:27781031
  9. 9. Helbig K. L. et al. Diagnostic exome sequencing provides a molecular diagnosis for a significant proportion of patients with epilepsy. Genetics in Medicine 18, 898–905 (2016). pmid:26795593
  10. 10. Lindy A. S. et al. Diagnostic outcomes for genetic testing of 70 genes in 8565 patients with epilepsy and neurodevelopmental disorders. Epilepsia 59, 1062–1071 (2018). pmid:29655203
  11. 11. Berg A. T. et al. Early-Life Epilepsies and the Emerging Role of Genetic Testing. JAMA Pediatrics 171, 863–9 (2017). pmid:28759667
  12. 12. Lemke J. R. et al. Targeted next generation sequencing as a diagnostic tool in epileptic disorders. Epilepsia 53, 1387–1398 (2012). pmid:22612257
  13. 13. Kodera H. et al. Targeted capture and sequencing for detection of mutations causing early onset epileptic encephalopathy. Epilepsia 54, 1262–1269 (2013). pmid:23662938
  14. 14. Mina E. D. et al. Improving molecular diagnosis in epilepsy by a dedicated high-throughput sequencing platform. European Journal of Human Genetics 23, 354–362 (2014). pmid:24848745
  15. 15. MD E. S. et al. Diagnostic Yield of Epilepsy Panels in Children With Medication-Refractory Epilepsy. Pediatric Neurology 64, 66–71 (2016). pmid:27726903
  16. 16. Fung C.-W., Kwong A. K.-Y. & Wong V. C.-N. Gene panel analysis for nonsyndromic cryptogenic neonatal/infantile epileptic encephalopathy. Epilepsia Open 2, 236–243 (2017). pmid:29588952
  17. 17. Demos M. et al. Diagnostic Yield And Treatment Impact Of Targeted Exome Sequencing In Early-Onset Epilepsy. 1–23 (2017)
  18. 18. Oates S. et al. Incorporating epilepsy genetics into clinical practice: a 360° evaluation. npj Genomic Medicine 3, 1–11 (2018).
  19. 19. Møller R. S. et al. Gene Panel Testing in Epileptic Encephalopathies and Familial Epilepsies. Molecular Syndromology 7, 210–219 (2016). pmid:27781031
  20. 20. Myllykangas S., Buenrostro J. D., Natsoulis G., Bell J. M. & Ji H. P. Efficient targeted resequencing of human germline and cancer genomes by oligonucleotide-selective sequencing. Nat Biotechnol 29, 1024–1027 (2011). pmid:22020387
  21. 21. Talevich E. et al. CNVkit: Genome-Wide Copy Number Detection and Visualization from Targeted DNA Sequencing. PLoS Comput Biol 12, 1–18 (2016). pmid:27100738
  22. 22. Richards S. et al. Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology. Genet Med 17, 405–423 (2015). pmid:25741868
  23. 23. Adzhubei IA. et al. Nat Methods 7, 248–249 (2010). pmid:20354512
  24. 24. SIFT missense predictions for genomes. Nat Protocols 11, 1–9 (2016). pmid:26633127
  25. 25. Schwarz J. et al. MutationTaster2: mutation prediction for the deep-sequencing age. Nat Methods 11, 361–362 (2014). pmid:24681721
  26. 26. Amadori E. et al. Targeted re-sequencing for early diagnosis of genetic causes of childhood epilepsy: the Italian experience from the ‘beyond epilepsy’ project. Ital J Pediatr 46, 92 (2020). pmid:32631363
  27. 27. Mefford H. C. & Mulley J. C. Genetically complex epilepsies, copy number variants and syndrome constellations. Genome Med 2, 71 (2010). pmid:20923578
  28. 28. Niestroj LM. et al. Epilepsy subtype-specific copy number burden observed in a genome-wide study of 17 458 subjects. Brain 143, 2106–2118 (2020). pmid:32568404
  29. 29. Silva M. et al. European guidelines for constitutional cytogenomic analysis. Eur J Hum Genet 27, 1–16 (2019). pmid:30275486
  30. 30. Russo C. D. et al. Comparative study of aCGH and Next Generation Sequencing (NGS) for chromosomal microdeletion and microduplication screening. J Prenat Medicine 8, 57–69 (2014). pmid:26266003
  31. 31. Hamdan F. F. et al. De novo STXBP1 mutations in mental retardation and nonsyndromic epilepsy. Ann Neurol 65, 748–53 (2009). pmid:19557857
  32. 32. Lotte J. et al. Effectiveness of antiepileptic therapy in patients with PCDH19 mutations. Seizure 35, 106–110 (2016). pmid:26820223
  33. 33. Trivisano M., Specchio N. & Vigevano F. Extending the use of stiripentol to other epileptic syndromes: a case of PCDH19-related epilepsy. European J Paediatr Neurology Ejpn Official J European Paediatr Neurology Soc 19, 248–50 (2014). pmid:25510386
  34. 34. T Y. et al. Loss-of-function mutations of STXBP1 in patients with epileptic encephalopathy. Brain Dev. 3, 280–4 (2016).
  35. 35. Alfadhel M. et al. Biotin-responsive basal ganglia disease should be renamed biotin-thiamine-responsive basal ganglia disease: a retrospective review of the clinical, radiological and molecular findings of 18 new cases. Orphanet Journal of Rare Diseases 8, 1–1 (2013).
  36. 36. Franz D. N. et al. Everolimus for treatment-refractory seizures in TSC. Neurology: Clinical Practice 8, 412–420 (2018). pmid:30564495
  37. 37. S J. et al. Treatment of Creatine Transporter (SLC6A8) Deficiency With Oral S-Adenosyl Methionine as Adjunct to L-arginine, Glycine, and Creatine Supplements. Pediatr Neurol. 53, 360–363 (2015). pmid:26205312
  38. 38. Chen J. W., Borgelt L. M. & Blackmer A. B. Cannabidiol: A New Hope for Patients With Dravet or Lennox-Gastaut Syndromes. Ann Pharmacother 53, 603–611 (2019). pmid:30616356
  39. 39. O D. et al. Long-term cannabidiol treatment in patients with Dravet syndrome: An open-label extension trial. 2, 294–302 (2019).
  40. 40. Knupp K. G. & Wirrell E. C. Treatment Strategies for Dravet Syndrome. Cns Drugs 32, 335–350 (2018). pmid:29594870
  41. 41. Bayat A. et al. PIGT-CDG, a disorder of the glycosylphosphatidylinositol anchor: description of 13 novel patients and expansion of the clinical characteristics. Genet Med 21, 2216–2223 (2019). pmid:30976099