Somatic variants as a cause of drug‐resistant epilepsy including mesial temporal lobe epilepsy with hippocampal sclerosis

The contribution of somatic variants to epilepsy has recently been demonstrated, particularly in the etiology of malformations of cortical development. The aim of this study was to determine the diagnostic yield of somatic variants in genes that have been previously associated with a somatic or germline epilepsy model, ascertained from resected brain tissue from patients with multidrug‐resistant focal epilepsy.


| INTRODUCTION
The emergence of genomic technologies, including next generation sequencing (NGS), has accelerated our understanding of the genetic architecture of the epilepsies. 1 More than 180 genes have been reported to harbor genetic variants that cause monogenic epilepsy. 2Most pathogenic genetic variants affect single-ion channels, metabolic pathways, or signaling pathways, and this knowledge has helped provide targets for precision medicine in monogenic epilepsy. 3omatic variants (alterations in DNA that occur at the postzygotic stage) can arise during the course of prenatal brain development and are known to cause neurological disease, including specific forms of epilepsy. 4alformations of cortical development (MCDs), including focal cortical dysplasia (FCD) and hemimegalencephaly (HME), represent the most common cause of neocortical childhood onset seizures. 5MCDs are a major cause of severe pediatric refractory epilepsies and are often an indication for epilepsy brain surgery. 6oth germline and somatic variants have been reported to cause MCDs. 7[10][11][12] The mTOR-signaling cascade plays a key role in the integration of various environmental signals to regulate cell growth, proliferation, and metabolism. 7Somatic variants in mTOR have been identified in up to 37% of patients with FCD, 13 including more subtle FCDs as well as larger HME 5,[8][9][10][11] and tuberous sclerosis complex. 12rain somatic loss-of-function variants in SLC35A2 are responsible for 29% of mild MCD/FCD type I patient cases. 7FCD type II and HME have been shown to be caused by somatic and germline variants mostly identified in MTOR-pathway genes (gain-of-function variants in MTOR and its activators [AKT3, PIK3CA, RHEB] and loss-of-function variants in its repressors [DEPDC5, TSC1, TSC2, NPRL2, NPRL3, PTEN, STRADA], converging to a common phenotype 7 ).Although gene discovery in focal brain malformations has centered on the "mTORopathies," other genes are involved in epileptogenic brain malformations.Pathogenic variants in LIS1 and DCX have been observed in lissencephaly and subcortical band heterotopia. 14,15Pathogenic somatic variants in GLI3 and OFD1 have been reported in hypothalamic hamartomas, 16 and a recurrent somatic variant Significance: This study provides novel insights into the etiology of mesial temporal lobe epilepsy with hippocampal sclerosis, highlighting a potential pathogenic role of somatic variants in CBL and ALG13.We also report candidate diagnostic somatic variants in FLNA in focal cortical dysplasia, while providing further insight into the importance of MTOR and related genes in focal cortical dysplasia.This work demonstrates the potential molecular diagnostic value of variants in both germline and somatic epilepsy genes.

K E Y W O R D S
drug-resistant epilepsy, mesial temporal lobe epilepsy with hippocampal sclerosis, somatic variants

Key points
• Owing to advances in next generation sequencing, and the availability of appropriate tissue, we now appreciate the contribution of somatic pathogenic variants in the etiology of malformations of cortical development.in GNAQ has been found in patients with Sturge-Weber syndrome. 17esial temporal lobe epilepsy related to hippocampal sclerosis (MTLE-HS) is among the most drug-resistant types of focal epilepsy. 18It is a common indication for epilepsy surgery as, in drug-resistant patients, resection of the sclerotic hippocampus can be an effective treatment. 19The etiology, pathogenesis, and genetic architecture of MTLE-HS have proven to be largely complex and elusive. 20There has been a recent exciting development toward understanding the role of genetic factors in the etiology and surgical outcomes of MTLE-HS, as these cases carry a higher burden of rare, deleterious variants in constrained genes and in genes encoding voltage-gated cation channels. 21Hippocampal somatic variants, particularly those activating Ras/Raf/mitogen-activated protein kinase (MAPK) signaling, may also contribute to the pathogenesis of MTLE-HS. 22n light of the above, this study aimed to determine the diagnostic yield of somatic variants within genes with an established association with monogenic epilepsy (germline or somatic) using bulk tissue NGS of resected brain tissue and matched blood in patients with a variety of drug-refractory focal epilepsy.

| Ethics statement
This project was approved by Beaumont Hospital Ethics Committee under study protocol REC 13/75 and REC 14/44.All study participants provided written informed consent.

| Study patients
Patients were recruited from Beaumont Hospital, Dublin, Ireland.All patients included in this study had multidrugresistant focal epilepsy treated with epilepsy surgery through the National Epilepsy Surgery Programme, Beaumont Hospital.We followed the International League Against Epilepsy definition of refractory epilepsy, whereby there had been a failure of adequate trials of two tolerated and appropriately chosen and used antiseizure medication schedules (whether as monotherapies or in combination) to achieve sustained seizure freedom. 23Focal epilepsy patients with neoplasia (primary or metastatic) were excluded.Paired tissue and blood samples were collected for each participant.Clinical phenotypes were extracted from the epilepsy electronic patient record at Beaumont Hospital, with radiology (magnetic resonance imaging [MRI]) and video-electroencephalography (EEG) analyzed and discussed at our multidisciplinary epilepsy surgery review meeting.Patients were subdivided into three categories based on the histopathology of the resected brain tissue, as well as MRI and EEG findings as follows: (1) MCDs, (2) mesial temporal sclerosis (or hippocampal sclerosis), and (3) nonlesional focal epilepsy.

| DNA extraction
DNA was isolated from peripheral blood samples using Tris/EDTA and extracted using Qiagen blood DNA extraction kits using a QIAcube.Brain tissue samples (37.8 mg average size, range = 12-113 mg) were homogenized using the Qiagen Tissue Lyser LT (40 s at 30 Hz) using the manufacturer's protocols.A maximum of 50 mg of tissue was used per column in an elution volume of 100 μL.DNA was extracted from the lysed tissue using Qiagen DNeasy Tissue DNA extraction kits on the Qiagen QIAcube.The proteinase K digestion was omitted from DNA extractions from both blood and tissue, as it is incompatible with downstream NGS protocols.

| Sequencing
The genes included for sequencing using a customdesigned targeted panel were selected as follows: (1) genes previously associated with any somatic neurological disorder, (2) highly expressed genes from the PI3K-AKT-mTOR phosphoinositide-3-kinase-protein kinase B (Akt)-mammalian target of rapamycin (pathway) in the cortex and hippocampus (determined using data from RNA-seq data consortiums: the Human Protein Atlas and the Genotype Tissue Expression project), and (3) genes with a high-frequency yield of putatively damaging variants from a large-scale review of epilepsy and neurological disorder gene panels. 24NA was prepared for sequencing using SeqCap EZ (Roche) library custom targeted panel preparation reagents according to the manufacturer's instructions.Two hundred nanograms of tissue-derived DNA and 500 ng of blood-derived DNA were used for library preparation.All blood-derived samples were pooled together for hybridization, whereas tissue-derived samples were pooled into five groups of seven samples.The custom targeted panel (see Table S1) was a SeqCap EZ Choice library, which captured 229 genes (.72 Mb).All 84 matched samples from 42 patients were sequenced via a service provider (Novogene) on a single NextSeq 600 sequencing run.Resulting unaligned reads were processed at Royal College of Surgeons in Ireland, Dublin.

| Bioinformatic analysis
Sequencing data were processed and variants identified using a Genome Analysis ToolKit (GATK4; v4.1.8.1) pipeline. 25Briefly, reads were aligned to human reference genome GRCh37 using Burrows-Wheeler aligner.A panel of normals (PoN) consisting of 100 780 variants was created using unaffected parents of children with epilepsy from a previous project. 26GATK4 Mutect2 was used to identify variants, applying a germline (blood-derived sample) reference and PoN for somatic variant screening. 25,27his process identified and removed germline variant calls showing concordance between brain-and bloodderived samples.The remaining variants were screened for common human genetic polymorphisms by eliminating overlap between the variant call format (.vcf) and databases comprising of >12 million common single nucleotide polymorphisms.These included a gnomAD 120 k exome variant database sourced from the Broad Institute (as recommended by GATK4 Mutect2 best practices) and dbSNP version 138.Following recommended quality control filtering, high-confidence somatic variant calls were annotated using ANNOVAR. 28Qualifying variants were selected that (1) were either exonic or splicing variants, (2) matched the expected inheritance pattern (e.g., a heterozygous variant in a gene typically associated with autosomal dominant disease), (3) had a variant allele frequency of >.05%, (4) had a minor allele frequency of 0% in the gnomAD population database, 29 and (5) passed manual visual inspection of read alignments using Integrative Genome Viewer (IGV) version 2.3.97 (this included visual screening for low-quality base calls, read-end artifacts, and erroneous alignments in low-complexity regions). 30emaining qualifying variants were crossed-referenced using Online Mendelian Inheritance in Man 31 to assess the phenotypes associated with the gene of interest and ClinVar 32 to identify previous reports of the variants in disease.In silico prediction of somatic variant pathogenicity data is included in Table S2.

| Variant confirmation
Qualifying variants were resequenced from the same DNA sample using targeted amplicon sequencing.Qualifying variants were polymerase chain reaction-amplified in 150-500-bp fragments using custom primers (Table S3).Amplified fragments were sequenced using Amplicon-EZ sequencing via service provider GENEWIZ/Azenta Life Sciences (approximately 50 000 reads per sample).The resulting sequencing reads were aligned to human reference genome GRCh38 using Burrows-Wheeler aligner. 33Aligned sequencing reads were visualized using IGV version 2.3.97 to identify the qualifying variants of interest and manually inspect sequencing quality.

| RESULTS
In total, 42 patients were recruited to the project.These included patients with MCDs (n = 10), nonlesional focal epilepsy (n = 9), and MTLE-HS (n = 23).There were 34 temporal lobectomies, three frontal lobe resections, one parietal lobe resection, one occipital lobe resection, one hemispherotomy, one disconnective procedure, and one corpus callosotomy.Detailed phenotype information on participants is provided in Table S4.One of the 42 pairs of samples failed sequencing (MTLE-HS).In the remaining samples, the average sequencing coverage in the germline (blood) DNA was 585×, whereas the average sequencing coverage in the brain tissue was 1360×.
Following variant identification and filtering (see Materials and Methods), 13 qualifying variants were detected in nine patients (Table S5) and sent for validation using high-coverage amplicon sequencing.A total of five variants from five patients were successfully validated (Table 1).One of these variants (NM_004975:c.2206delC:p.R736fs) was later rejected as a technical artifact based on further review in IGV (homopolymer region at which both deletions and insertions were called; see Figure S1).The overall diagnostic yield in the cohort was 10% (4/41).The diagnostic yield in MCD patients was 20% (2/10) and in mesial temporal sclerosis patients was 9% (2/22).No somatic molecular diagnosis was made in the nonlesional focal epilepsy group.

| CBL
A missense variant in CBL (NM_005188:c.1103A> G:p.Y368C) with variant allele fraction (VAF) 1.4% was identified in Patient SE024, a 27-year-old male patient who has a mild learning disability, with right temporal lobe epilepsy with MRI showing hippocampal sclerosis.He underwent a right temporal lobectomy, which resulted in a 50% reduction in seizures postoperatively (with histology shown in Figure 1).Pathogenic germline variants in the CBL gene are associated with autosomal dominant Noonan syndrome-like disorder with or without juvenile myelomonocytic leukemia.To our knowledge, CBL has not been associated with epilepsy under a somatic model.According to UniProt (www.unipr ot.org/ ), this variant lies within the linker region of the CBL protein.This protein is an E3 ubiquitin ligase that acts as a negative regulator of kinase signaling pathways.Although Patient SE024 does not have the typical features of patients with germline Noonan syndrome (short stature, intellectual disability, and delayed puberty), this is the first instance of a somatic CBL variant identified in brain tissue, and so the resulting clinical features of somatic CBL variants have not been established.We hypothesize that this variant is a plausible candidate to explain Patient SE024's temporal lobe epilepsy, although functional evidence is required to confirm this association.

| ALG13
In Patient SE026, a frameshift deletion variant with VAF 1.1% was identified in ALG13 (NM_001099922:c.573delT:p.A191fs).Patient SE026 is a 39-year-old female patient with left temporal lobe epilepsy with seizure onset at age 12 years without intellectual disability.She underwent left temporal lobectomy, which resulted in complete seizure freedom with histology confirming hippocampal sclerosis (Figure 2).Pathogenic germline variants in the ALG13 gene are associated with dominant developmental and epileptic encephalopathy.To our knowledge, ALG13 has not been associated with epilepsy under a somatic model.The majority of patients with pathogenic germline variants in ALG13 are females, suggesting an X-linked dominant (XLD) inheritance pattern, although there have been rare reports of affected males with a similar phenotype.This is the first loss-of-function ALG13 variant reported in an epilepsy patient, as well as the first somatic variant.We hypothesize that this is a possible cause of Patient SE206's temporal lobe epilepsy, although without functional evidence we cannot conclusively confirm this association.Affected female patients with periventricular heterotopia often present with epilepsy without intellectual disability, as in Patient SE038 (although this patient did not have periventricular heterotopia).The variant identified in Patient SE038 has not been reported previously in the germline or somatic state.To our knowledge, this is the first time a somatic variant in FLNA has been associated with FCD type II.Functional evidence is required to confirm the pathogenicity of this variant.

| MTOR
A variant (NM_004958:c.4385_4402del:p.1462_1468del)with VAF .8% in MTOR was identified in Patient SE028.Patient SE028 had a seizure onset at 13 years, with right temporal lobe epilepsy.Although MRI was normal, his EEG indicated focal epilepsy arising from the right temporal region.The patient underwent a right temporal lobectomy and achieved complete seizure freedom as a result of that surgery.The resected tissue was classified as cortical dysplasia Palmini grade IIb (Figure 4).Although variant has not been previously reported as a germline or somatic variant, MTOR is a well-established cause of somatic FCD type II.The expected phenotype for patients with somatic variants matches well with the phenotype of Patient SE028.The variant is absent from control populations (gnomAD), is located in the kinase domain (amino acids 1382-1982), and is present at a detectable allele fraction in an affected tissue sample (brain) but is absent from another tissue (blood).This evidence supports the view that this variant is the cause of Patient SE028's epilepsy.Additional functional work would confirm the pathogenicity of this variant.

| KCNB1
Finally, a KCNB1 frameshift deletion variant (NM_004975:c.2206delC:p.R736fs) with VAF 1.4% was identified in Patient SE004 (a 52-year-old female patient who underwent left temporal lobectomy for drugrefractory epilepsy with histology showing hippocampal sclerosis) from both targeted panel and replication amplicon sequencing, but was later classified as a technical artifact (on the basis of manual read inspection of both primary and replication sequencing data using IGV).

| DISCUSSION
In this study, we characterized somatic variants within genes with an established association with monogenic epilepsy (germline or somatic) in resected brain tissue and matched blood in 41 patients with a variety of drugrefractory focal epilepsies.We report a diagnostic yield of 20% in patients with MCD and 9% in patients with MTLE-HS.The diagnostic yield of 20% in the MCD group is consistent with previous reports of yields between 13% and 46%. 7,10,11,41,42The 9% diagnostic yield we observed in the MTLE-HS patients is consistent with the yield of 11% found in a recent important publication, which for the first time provided insight into the potential pathogenesis of MTLE-HS, indicating a possible link to Ras/Raf/ MAPK signaling genes. 22The lack of genetic diagnosis in the nonlesional focal epilepsy group is also consistent with previous efforts. 43,44Candidate disease-causing somatic variants (requiring further functional validation to confirm pathogenicity) were identified for the first time in three genes (CBL, ALG13, and FLNA), previously associated only with germline models of pathogenicity.
CBL is an established cause of autosomal dominant Noonan syndrome-like disorder, with comorbid developmental delay.It is noteworthy that CBL is part of the Ras signaling pathway, and recent publications have reported candidate pathogenic variants in this pathway in patients with epilepsy. 22,34Somatic variants in proto-oncogene CBL are also associated with cancers, including juvenile  myelomonocytic leukemia. 31The CBL variant we fied in a patient with MTLE-HS (Patient SE024) has previously been reported as a somatic variant in a mature pineal gland teratoma 35 and a myeloid neoplasm. 36The variant is currently absent from germline databases.This being a gene with a known seizure phenotype when observed in the germline, 37 and a gene associated with cancer in the somatic state, 35,36 we hypothesize that this gene may cause epilepsy under somatic as well as germline models, albeit with a different epilepsy phenotype than reported to date for germline CBL variants. 37The reported germline phenotype includes developmental delay as in our patient, but with focal epilepsy secondary to moyamoya phenomenon and an abnormal MRI scan as well as an atypical hemolytic uremic syndrome (not observed in our patient). 37LG13 is an established cause of developmental and epileptic encephalopathy under a germline model.ALG13 is a glycosylation gene that inhibits N-glycosylation, a similar action to SLC35A2. 38Patients with pathogenic ALG13 variants present with early onset seizures that are typically refractory to treatment.The variant we observed (NM_001099922:c.573delT:p.A191fs) in an individual with MTLE-HS (Patient SE026) is absent from the ClinVar (germline) and COSMIC (somatic) databases and, at the time of writing, previously unreported.There have been no previous reports of pathogenic germline protein lossof-function (pLoF) ALG13 variants, and we speculate that such variants may be incompatible with life.There are seven pLoF variants listed in ALG13 in gnomAD v2.1.1, of which six are flagged for dubious quality or annotation indicating that these may be sequencing artifacts.The gene overall has a gnomAD probability of being loss-offunction intolerant score of 1 (where ≥.9 is considered extremely pLoF intolerant).This would be consistent with this first report of a pLoF ALG13 variant being somatic and causal, although functional studies are warranted to confirm this association.
Pathogenic germline variants in the X-linked FLNA gene are associated with a range of neurological conditions, including periventricular heterotopia and FG syndrome. 31Patients with periventricular heterotopia as a result of variation in FLNA tend to present with epilepsy and normal intelligence. 31Patients with FG syndrome fall within a wide phenotypic spectrum but can have clinical features including congenital hypotonia and macrocephaly. 31Filamin A protein expression has been reported to be increased in resected cortical tissue responsible for seizures in FCD. 39In a mouse model, Zhang and colleagues demonstrated that FLNA knockdown reduced seizure frequency in mice independent of mTOR signaling. 39We identified a somatic FLNA variant in an individual with MCD (Patient SE038), which is a plausible candidate for this patient's epilepsy.
A candidate pathogenic variant was identified in MTOR in Patient SE028.mTORopathy-associated genes are a well-established cause of FCD, accounting for >60% of diagnoses in patients with FCD. 7 The VAF for this MTOR variant was .8%, a low fraction compared to other established pathogenic, somatic MTOR variants in the literature. 11We hypothesize that patients with higher VAF somatic variants may present with a more severe phenotype.This hypothesis was investigated by Miller et al., who demonstrated that somatic SLC35A2 VAF in different tissue types correlated with severity of electrophysiologic and radiographic findings. 38We acknowledge that pathogenic MTOR variants cause disease through a gain-of-function mechanism and that this variant is a nonframeshift deletion.Although this variant type is typically loss of function, it is possible that it may result in gain of function (although this is an uncommon consequence of this variant type). 40We also note that both nonsense and frameshift (presumably loss of function) variants listed as disease-causing have been reported in MTOR since 2022 in ClinVar.
Previous studies, including targeted panel and whole exome sequencing (WES), limited analysis to a subset of genes of interest such as those in the mTOR, PI3K, and AKT pathways, 7,10,11,42,44 omitting established germline epilepsy genes that have provided potential pathogenic somatic diagnoses in our patients.This suggests that a larger panel or WES approach (without a virtual bioinformatics panel) is warranted for somatic variant identification in diagnosis of drug-refractory epilepsy.WES was used by Zhang et al., who studied a cohort of 17 pediatric patients with FCD. 45 These authors included variants from all brain-expressed genes, providing a genetic diagnosis in 35% cases, including variants in five genes not previously associated with cortical malformations. 45Similarly, Lai et al. implemented a WES approach, which identified a high rate of diagnostic yield in both hemimegalencephaly (75%) and FCD (30%). 46These results support our observation that expanding the list of genes in which variants are called may reveal novel somatic causes of epilepsy.The VAF of all candidate pathogenic variants reported here are low (many studies set a threshold of >1% VAF 8 ).It is established that variants with VAF of <1% are sufficient to result in intractable epilepsy. 46These results demonstrate the value of high read depth sequencing (>1000× coverage), which is required to detect variants at this threshold.
There are several limitations to our study.We did not include an analysis of copy number variants, another cause of malformations of cortical development.Our sequence capture design omitted intergenic variants, which may explain pathogenicity in a proportion of cases.Functional studies were included as part of this study, but would provide extra evidence for pathogenicity of the somatic variants we have discussed.
Although all the candidate pathogenic variants we present are within established epilepsy or neurodevelopmental disorder genes and have been stringently filtered for pathogenicity criteria, interpreting the clinical consequences of these variants is challenging.Detailed variant interpretation guidelines are available for germline variants, 47 and somatic cancer variants, 48 but guidelines are lacking for the interpretation of noncancer somatic variants.We suggest that standardized guidelines for the interpretation of noncancer somatic variants, and perhaps guidelines for somatic variants in epilepsy specifically, are warranted.
If expanded upon and replicated, these findings may have potential therapeutic implications.The CBL gene is involved in the tyrosine kinase pathway, and a series of tyrosine kinase inhibitors are in trial phase for multiple sclerosis (suggesting potential therapeutic target implications for CBL). 49For the MTOR variant, inhibitors of the MTOR pathway such as everolimus may have additional therapeutic potential. 50his study proposes a potential somatic component to the genetic etiology of MTLE-HS and provides further evidence to the recent work of Khoshkhoo et al., who highlighted that hippocampal somatic variants may contribute to the pathogenesis of sporadic, drugresistant MTLE-HS. 22Neither of the two genes in the MTLE-HS group (CBL or ALG13) has previously been reported under somatic models, nor has FLNA for MCDs (FCD II).Our results also provide further insights into the importance of MTOR and related genes as a cause of FCD.Further understanding of the role of somatic variation in drug-resistant epilepsy may soon be provided by emerging high-resolution sequencing techniques such as single-cell sequencing. 45Our study also underlines the importance of continued genetic and molecular studies of resected brain tissue of subjects who have undergone epilepsy surgery to increase our knowledge and understanding of the epilepsies in patients who have undergone epilepsy surgery.

3. 1
.3 | FLNA A missense variant with VAF .4% in FLNA was identified in Patient SE038 (NM_001110556:c.1604A> G:p.D535G).Patient SE038 is a 29-year-old female with right frontal lobe epilepsy and seizure onset at age 11 years.The patient achieved complete seizure freedom following a right anterior frontal resection with histology confirming cortical dysplasia Palmini grade IIb (Figure 3A,B).Variants in FLNA are associated with a range of XLD epilepsy syndromes, including XLD periventricular heterotopia, XLD frontometaphyseal dysplasia, and XLD FG syndrome.

F I G U R E 1 F I G U R E 2
Histopathology slide for Patient SE024 (gene: CBL) with high-power magnification showing a gliotic and sclerotic hippocampus with focal neuronal loss.Histopathology slide for Patient SE026 (gene: ALG13), which shows hippocampal sclerosis with dentate fascia (blue arrow) granular neuronal depletion and dispersion.

F
I G U R E 3 (A) Histopathology slide for Patient SE038 (gene: FLNA), showing clustering of maloriented neurons with dysmorphic neurons (blue arrow) and balloon cells (B), the pathological hallmark of focal cortical dysplasia type IIb.(B) Histopathology slide for Patient SE038 (gene: FLNA), showing balloon cells (yellow arrows).

F I G U R E 4
Histopathology slide for Patient SE028 (gene: MTOR), showing dysmorphic neurons (blue arrow) and balloon cells (yellow arrow), the pathological hallmark of focal cortical dysplasia type IIb.
Qualifying variants confirmed using amplicon sequencing.Four of the five variants listed were considered as candidate pathogenic variants.One variant (Patient SE004, gene KCNB1) was excluded as a candidate pathogenic variant as it is suspected to be a technical artifact following visual inspection in Integrative Genome Viewer.