Genome sequencing identifies rare tandem repeat expansions and copy number variants in Lennox–Gastaut syndrome

Cohort recruitment

Adults with DEEs (n = 30) were recruited from the Adult Epilepsy Genetics Program at Toronto Western Hospital. All patients had a prior negative or inconclusive genetic test. This study was approved by the University Health Network’s Research Ethics Board.

Whole-genome sequencing

Of the 30 patients, there were two pairs of identical twins. Parents were also sequenced where possible (n = 11 pairs). In brief, genomic DNA was extracted from whole blood, and then sequenced using the Illumina HiSeq X platform at The Centre for Applied Genomics. Sequence reads were aligned to the GRCh37 reference genome using the Burrows-Wheeler Aligner read alignment program.²⁰ Local realignment, base quality recalibration and removal of duplicate reads were carried out using the Broad Institute’s Genome Analysis Toolkit (GATK) package. The following algorithms were used for variant calling and genotyping: (i) GATK Haplotype Caller for single nucleotide variants (SNVs) and indels; (ii) ERDS (v1.1) and CNVnator (v0.3.2) for copy number variants (CNVs); and (iii) a combination of ExpansionHunter Denovo (EHdn, v0.7), Tandem Repeat Finder (TRF) and ExpansionHunter (EH, v3.0.2) for TRs.^21–23

Each variant call format file was annotated using a custom ANNOVAR pipeline.²⁴ To filter for only high quality SNVs, the following parameters were applied: (i) autosomal heterozygous variants met a genotype quality (GQ) cut-off value of ≥99, and an alternative allele fraction ≥0.3 and ≤0.7; (ii) homozygous or chromosome X variants had a GQ ≥ 25 and an alternative allele fraction >0.7; and (iii) all variants passed GATK pipeline filter constraints. DeNovoGear was used to detect de novo SNVs in patients sequenced in trio (n = 13, where two of the families have twins).²⁵ To filter for only robust CNVs, CNVs had to be called by both ERDS and CNVnator.²¹ For TRs, EHdn was used to estimate the size and location of TRs lengths in >60 known TR-associated disease loci, as previously described.^23,26 Ancestry and kinship analyses were carried out using PLINK (v1.9), as previously described.²⁷

Rare variant analysis

A rare variant was defined as being present with <1% frequency in the general population, using genomic databases (the 1000 Genomes project, Exome Aggregation Consortium, Genome Aggregation Database, DECIPHER and the Database of Genomic Variants). Rare SNVs and CNVs impacting both coding and non-coding regions of the genome were analysed, as previously described.^27,28 TR expansions were filtered based on the genomic content disrupted, biological relevance, known disease associations, and any overlap with reported literature or clinical reports in databases. Using the American College of Medical Genetics and Genomics’ (ACMG) guidelines, we classified rare SNVs and CNVs into pathogenic (i.e. disease-causing), likely pathogenic, variant of unknown significance (VUS), benign and likely benign categories.^29–32 If a VUS was identified in a gene without a validated association to the patient’s phenotype, it was considered a gene of uncertain significance.²⁹ As there are no ACMG standards to guide the interpretation of the pathogenicity of TR expansions, we defined a repeat expansion to be of interest if its size either (i) fell into the reported disease-causing range, or (ii) was larger than what is observed in control individuals in reported studies.

Variant validation

Where possible, variants were experimentally validated using either (i) Sanger Sequencing for SNVs; (ii) qPCR for CNVs; or (iii) Southern blotting and/or repeat-primed PCR (RP-PCR) for TR expansions. Specifically, southern blotting was used to validate the CGG expansion in DIP2B found in patient 4 as previously described.³³ In brief, genomic DNA was first digested using selected restriction-endonuclease enzymes, followed by the resolution of DNA fragments on an agarose gel. These DNA fragments were then transferred to a nylon membrane, and then subjected to hybridization analysis, where radioactively labelled probes were used to detect the sizes of the CGG alleles in DIP2B.

To validate the CTG repeat expansion in ATXN80S in both patient 6 and his mother (6A), a combination of primer sequences were used to amplify the CTG repeat locus in ATXNN8OS: a single forward fluorescently labelled primer (P1, 5′-6-FAM—CTGGGTCCTTCATGTTAGAAAACCT-3′), and a combination of two reverse primers in a 1:10 ratio (P3, 5′-TACGCATCCCAGTTTGAGACGC-3′, and P4, 5′-TACGCATCCCAGTTTGAGACGCAGCAGCAGCAGCAGCA-3′).^34,35 The reverse primer P4 anneals at different sites along the CTG repeat, resulting in a range of PCR product sizes up to the inclusion of the full repeat, depending on the amplification limit of the assay. These fluorescently labelled PCR products of different lengths were analysed using capillary electrophoresis and visualized using PeakScanner 2.

Statistical analysis: comparison of clinical characteristics

One-sided 2×2 Fisher Exact Tests were carried out to compare clinical characteristics between patients with an identified variant, and those with no identified variant, including sex, age of seizure onset, EEG, MRI, presence/absence of family history, developmental regression (DR) and intellectual disability (ID) severity. In each case, an odds-ratio (OR) and a P-value (P) were generated.

Data availability

The dataset(s) supporting the conclusions of this article is(are) included within the manuscript. Individual ancestries and kinship values are not included in order to preserve participants’ privacy.

WGS analysis

In total, 52 genomes were sequenced: 30 patients and 22 parents (Supplementary Table 1, Supplementary Fig. 1). There was an average read depth of 37.0 across the entire cohort. Overall, WGS detected an average of 4 535 791 indels and SNVs, and 706 CNVs per genome amongst patients. Regarding rare variants, patients had an average of 307 indels and SNVs, and 4.1 CNVs impacting coding regions per genome.

Ancestry analyses found that the cohort was largely European (n = 40, 76.9%), with the remainder of patients belonging to South Asian (n = 9, 17.3%), American (n = 2, 3.85%) and East Asian (n = 1, 1.92%) ancestry groups (Supplementary Fig. 2). We confirmed that patient 16’s parents (16A, 16B) were third-degree relatives, and that the two twin pairs were genetically identical. Only one individual in each twin pair was included in downstream analyses, resulting in a total of 28 patients’ WGS data for analysis.

Clinically relevant rare variant analysis

Upon rare variant analysis, we identified seven pathogenic or likely pathogenic SNVs, and two pathogenic CNVs, resulting in a molecular diagnosis in 9/28 (32.1%) patients (Tables 1 and 2).

In addition, we identified two de novo VUS SNVs, which do not meet clinical significance as per ACMG guidelines, but instead are potentially clinically relevant and merit further investigation. We also identified two TR expansions in DIP2B and ATXN8OS. Overall, patients with an identified variant had a younger age of seizure onset compared to the variant absent group (OR: 8.27, P = 0.042, one-sided Fisher’s Exact Test) (Table 3).

TR expansions

We identified two TR expansions in two patients with LGS: a CGG repeat expansion in the 5′ untranslated region (UTR) of DIP2B in patient 4 (Fig. 2), and a CTG repeat expansion (of unknown inheritance) in ATXN8OS in patient 6 (Fig. 3). There are no ACMG guidelines to interpret TR pathogenicity yet: therefore, these two variants were not classified, but have potential clinical implications.

Patient 4: At the age of one year, he had several febrile generalized tonic clonic seizures. At 19 months, he developed infantile spasms, and later atypical absence, myoclonic, tonic, and focal onset seizures, and EEG findings in keeping with LGS. He has significant ID, is non-verbal and not toilet trained, but can ambulate and eat unassisted. He continues to present intractable seizures of multiple types. His father had previously presented with seizures (Fig. 2A). Through WGS, a 5′UTR CGG TR expansion was identified in the brain-expressed DIP2B (disco-interacting protein 2 homolog B) gene. This TR expansion has previously been reported in two individuals with ID and/or seizures.³³

We confirmed the presence of an expanded allele in patient 4 using Southern blotting (Fig. 2B, Supplementary Fig. 3). Southern blotting estimated the allele sizes to be approximately 12 and 128 CGG repeat units in patient 4, which is larger than the initial estimates by our TR detection pipeline (12 and 99 units). Our TR detection pipeline estimated 12 and 15 CGG units in the mother, and 7 and 73 CGG units in the father.

Patient 6 has a history of seizures, with initial onset at four years of age. A brain MRI scan was normal. He underwent resective surgery, guided by cortical EEG with partial resections of the left frontal, anterior parietal and left mesial temporal lobes. Pathology identified a mild disorganization of neuronal lamination with clustering of neurons. Despite this extensive resection, this patient evolved with classic LGS, exhibiting typical clinical and electrophysiological markers. Later, he also had a partial corpus callosotomy, but still has drop seizures. He does not have ataxia. Despite upper limb ataxia, his brain MRI shows a normal cerebellum, without any evidence of volume loss. Through WGS, we identified a CTG expansion in ATXN80S: specifically, our TR detection pipeline estimated 19 and 119 CTG repeat units in ATXN80S.

Using RP-PCR and capillary electrophoresis, we confirmed the presence of the expanded CTG repeats in ATXN80S in the proband: the repeat size is estimated to be greater than 135 CTG repeat units—the exact number cannot be determined as it is beyond the limit of this assay (Fig. 3B). We also verified the presence of a shorter expansion (∼95 repeat units) in the mother (6A) (Fig. 3C). Previously, CTG TR expansions between 80 and 250 units in ATXN80S have been reported in spinocerebellar ataxia type 8 (SCA8).³⁷

Novel TR expansions found in patients with LGS

For patient 6, both a CMA and gene panel did not find a genetic cause for his LGS diagnosis. Through WGS, a trinucleotide CTG repeat expansion in ATXN8OS in the pathogenic range was identified. ATXN8OS repeat expansions are associated with SCA8: a progressive ataxia exhibiting reduced penetrance, where seizures are rare, but have been reported.³⁷

Upon identifying the ATXN8OS repeat expansion, this patient was evaluated in the ataxia clinic to determine if the falls were due to ataxia. Despite upper limb ataxia, he does not have gait ataxia nor other hallmarks of SCA8. It is unclear if the dearth of SCA8 symptoms in patient 6 is caused by low penetrance, or if he will eventually develop other symptoms of SCA8 as he ages. However, it is also possible that repeat expansions in ATXN8OS may manifest as LGS. This would not be completely unexpected, as both SCA8 and some other forms of SCAs also manifest with seizures and ID. For example, Swaminathan⁴⁸ reported a 22-year-old female patient with typical SCA8 who developed recurrent seizures and episodes of status epilepticus, which was successfully treated using antiepileptic therapies. In general, seizures and epilepsy are uncommon features in patients with SCAs: epilepsy as an inherent part of the phenotype has only been consistently described in SCA10 and SCA17, although an anecdotal association has also been reported in SCA2, SCA8, SCA13, SCA19/22 and SCA48.^48–52

We found that a second LGS patient, patient 4, harboured 12 and 128 CGG repeat units in DIP2B, compared to 6–23 units found in controls.^33,53 This expansion has previously been reported to be associated with the fragile site FRA12A in two individuals with ID and/or seizures.^33,53 The expansion in patient 4 exceeds the length of a normal allele by over 350 base pairs (>140 excess repeat units). Interestingly, DIP2B is highly expressed in the brain, mainly in the frontal and parietal cortices, making it a candidate gene for an epileptic disorder.³³ The discovery of this CGG repeat expansion at 5′UTR of DIP2B in patient 4 adds to the limited cases in the literature, collectively suggesting that this expansion may contribute to ID and seizure-related phenotypes.³³

To date, TR expansions have been reported as a disease-causing mechanism in over 40 disorders, including spinocerebellar ataxias.⁵³ To the best of our knowledge, repeat expansions have not previously been reported in LGS patients. TR expansions have been reported in patients with BAFMEs, which is associated with intronic (TTTTA)_n(TTTCA)_n repeat expansions in SAMD12, YEATS2, STARD7 and MARCH6.^16–19 We did not identify any such expansions in our cohort, suggesting that this may be a rare occurrence in patients with epilepsy, or is specific to BAFMEs. Interestingly, we investigated the genome-wide characteristics of TRs in 17 231 genomes of individuals with ASD, their families and control individuals, finding that TR expansions contribute a total of 2.6% risk to ASD.²⁶ This suggests that TR expansions are not only responsible for movement disorders, and may be a more common mechanism in neurological disorders than currently reported.

Small CNV in a patient with LGS

We also identified a CNV that was too small (28 273 bp) to be detected by a CMA in patient 10 with LGS. Individuals with damaging genetic variants impacting MBD5 are collectively referred to as MBD5-associated neurodevelopmental disorders (MANDs). They exhibit DD and/or regression, ID, seizures, ASD-like behaviours, speech impairment, motor delays and hypotonia, which corresponds to patient 10’s phenotype.⁵⁴ In addition, ORC4 is implicated in the autosomal recessive Meier-Gorlin syndrome, involving short stature and microcephaly.⁵⁴ While patient 10 does not harbour a second variant impacting ORC4, her height and head circumference are lower than expected, suggesting that the ORC4 disruption may be modifying the patient’s overall core phenotype (which is due to MBD5 haploinsufficiency).

Pathogenic variation in KCNA2 associated with abnormal cerebellar findings

We identified three patients with pathogenic or likely pathogenic variants in KCNA2. KCNA2 encodes a voltage-gated delayed rectifier potassium channel, and is associated with various DEEs.^55,56 Using either targeted gene sequencing or WES, Masnada et al.⁵⁶ identified variants in KCNA2 in a cohort of 23 patients with epileptic encephalopathies, and found abnormal cerebellar findings in patients with KCNA2 with gain-of-function (GoF) pathogenic variants.

In our cohort, out of the three carriers of variants in KCNA2, two had significant cerebellar signs (Patients 27 and 3). Patient 27 also harboured a VUS in UBR5 (a ubiquitin-protein ligase component), which has previously been reported as potentially related to epilepsy, ASD and ID.^6,57–59 Previously, a missense p. Arg1907His variant was reported in a family with BAFME, which also features cerebellar dysfunction.⁵⁷ Since not all patients in the Masnada et al., 2017 cohort had undergone WES, it may be possible that some may have a variant impacting a second gene affecting cerebellar function, including those with GoF pathogenic variants in KCNA2. An alternative explanation is that the pathogenic variant in KCNA2 in our patient may be the only one responsible for patient 27’s phenotype, and that UBR5 remains a gene of uncertain significance.

Patient 3 had seizures, ID, dystonia and ataxia. In addition to the GoF p. R297Q variant in KCNA2 (which might explain the ataxia and dystonia), she also had a de novo VUS in KCNIP4.^29,55,56 KCNIP4 encodes a potassium channel-interacting protein expressed in all regions of the brain, and is involved in regulating the frequency of slow repetitive firing and back-propagation of action potentials.^60,61 It has been suggested as a candidate gene for ADHD, epilepsy and personality disorders, though there is only a single case study and one genome-wide association study to support this.^60,61 Jenkins et al.⁶² recently reported that the canine KCNIP4 is implicated in two Norwegian Buhund dogs with progressive cerebellar ataxia, suggesting that it may be a potential candidate for cerebellar ataxia in humans and other species. Taken altogether, we propose that these additional de novo variants in KCNIP4 and UBR5 genes may contribute to the more severe ID and movement disorders seen only in patients 3 and 27, but not on the other KCNA2 variant carrier, patient 19 (who presented with epilepsy, mild ID, mild ASD and passed away from SUDEP). Sequencing KCNIP4 and UBR5 in larger, well-characterized cohorts will be necessary to determine their association with epilepsy, ID, and movement disorders.

Limitations

While we achieved a 32.1% diagnostic yield, it is important to note that our patients were previously screened by various next generation sequencing tests, and the overall yield would likely have been higher if there was an unbiased patient recruitment. It is also important to note that the nine rare genetic variants we identified are either disease-causing (pathogenic) or have a high likelihood to be disease-causing (likely pathogenic) for epilepsy, but may not be the sole variants contributing to the patient’s phenotype (i.e. there may be other variants contributing to the patient’s phenotype either independently or collectively). We did not evaluate variants with lower effect sizes (e.g. common variants) as this is a small cohort. We also did not analyse somatic variation present in resected tissues. This approach has previously identified a ‘two-hit’ phenomenon, where a combination of a germline and a second somatic variant can lead to focal epilepsies.⁸ Finally, the two reported TR expansions cannot be classified as there are no guidelines to facilitate pathogenicity interpretation, although these expansions are in the disease range size for other diseases.⁵³