Early genetic epidemiology studies suggested an appreciable heritability and genetic basis for the epilepsies (Berkovic et al., 1998; Berkovic and Scheffer, 1998; Hemminki et al., 2006; Peljto et al., 2014), though there are important differences in genetic architecture among the major classes of epilepsy. Genetic generalized epilepsy (GGE) and nonacquired focal epilepsy exhibit moderate to high heritability, but highly penetrant, causative pathogenic variants have been difficult to identify. The rare and severe developmental and epileptic encephalopathies (DEEs), on the other hand, are much less heritable and often due to highly penetrant de novo dominant, X-linked, or recessive pathogenic variants.

The ability to study the genetic basis of the epilepsies and many other human disorders accelerated with the introduction of technologies to detect both rare and common genetic and genomic variants. Two of the most important advances were chromosome arrays and next-generation sequencing (NGS) technology, both of which facilitated relatively unbiased, genome-wide investigation. Chromosome arrays include single nucleotide polymorphism (SNP) arrays and array comparative genomic hybridization (aCGH); both can be used to detect unbalanced copy number changes (deletions, duplications). SNP arrays, designed for genotyping, have been widely used for genome-wide association studies. Chromosome array studies in large cohorts of individuals with epilepsy led to the discovery of copy number changes that increase disease risk or are causative for epilepsy. NGS eliminated the need for linkage analysis and candidate gene prediction, further accelerating gene discovery across many human disorders. Unbiased sequence analysis of large sets of candidate genes or of the exome, especially when performed in a proband-parent trio fashion, highlighted the role of de novo variants in the DEEs, for example.

In this chapter, we present the current understanding of genetic contributions to the epilepsies followed by a discussion of approaches likely to further accelerate gene discovery in the next decade.

Genetic Generalized Epilepsy

Genetic generalized epilepsies, which include childhood absence epilepsy, juvenile myoclonic epilepsy, and generalized tonic-clonic seizures, account for approximately one-quarter of all epilepsies (Mullen et al., 2018). As the name implies, genetic contribution has been long suggested for GGE due to heritability estimates from family and twin studies. The recurrence risk of generalized epilepsy for a first-degree relative of an affected individual is 5 to 10 times higher than the background risk (Hemminki et al., 2006; Peljto et al., 2014). A high concordance rate among monozygotic compared to dizygotic twins (Berkovic et al., 1998; Kjeldsen et al., 2005) suggests high heritability. Early linkage studies followed by candidate gene sequencing in families with multiple affected individuals identified pathogenic variants in several causative genes, but the genes identified in these families are rarely responsible for GGE more broadly. Furthermore, families with a clear autosomal dominant inheritance pattern are unusual; most families with a history of GGE exhibit a more complex inheritance pattern likely due to a combination of genetic and nongenetic factors.

Though rare, single-gene causes of GGE have been identified. Heterozygous pathogenic variants in SLC2A1 disrupt the function of GLUT1, the sole transporter that moves glucose across the blood–brain barrier. The phenotype of autosomal dominant GLUT1 deficiency syndrome ranges from a severe encephalopathy with intractable epilepsy, ataxia, and cognitive impairment to a milder presentation with absence seizures and movement abnormalities (Wang et al., 1993). Notably, pathogenic variants in SLC2A1 are identified in up to 10% of individuals with early onset (<4 years old) absence epilepsy (Arsov et al., 2012). Much less frequently, pathogenic variants in GABRG2 (Wallace et al., 2001) and GABRA1 (Lachance-Touchette et al., 2011) have been identified as causative for rare cases of generalized epilepsy through candidate gene approaches.

A number of genetic and genomic risk factors for GGE have been identified, including several large, recurrent deletions on chromosomes 15q11.2, 15q13.3, and 16p13.11 (Helbig et al., 2009; de Kovel et al., 2010; Heinzen et al., 2010; Muhle et al., 2011). The ~1.5 Mb recurrent deletion of 15q13.3 is one of the most prevalent genetic risk factors for GGE, with an odds ratio of 68 (29–181) (Dibbens et al., 2009). All three deletions also confer risk for intellectual disability, autism and related neurodevelopmental disorders. As risk factors for epilepsy, the deletions are neither necessary nor sufficient to cause disease. In addition, the risk-conferring deletions are more frequently found in individuals with GGE who also have intellectual disability (Mullen et al., 2013).

Sequence analysis has also revealed genetic risk factors for GGE. In one study using exome sequencing, ultra-rare sequence variants in known monogenic epilepsy genes such as KCNQ2, SCN1A, and GABRG2 were more likely to be found in affected individuals compared to population controls (Epi4K Consortium and Epilepsy Phenome/Genome Project, 2017). In a similar study using exome data, an excess of rare missense variants in genes encoding GABA _A receptor subunits was identified in individuals with familial GGE (May et al., 2018). However, segregation could not be tested in the first study, and variants did not always segregate with disease in the second study.

Genome-wide association studies, which identify common variants associated with (usually small) increase for a given disease, have also been performed (International League Against Epilepsy Consortium on Complex Epilepsies, 2014; International League Against Epilepsy Consortium on Complex Epilepsies, 2018). The largest study to date involved 3,769 individuals with GGE and 29,677 unaffected controls. The study identified 11 loci that reach genome-wide significance in association with GGE (International League Against Epilepsy Consortium on Complex Epilepsies, 2018), highlighting 13 candidate genes near the significant SNPs, including known epilepsy genes such as SCN1A. Analysis of subsyndromes of GGE revealed a novel association with juvenile myoclonic epilepsy and SNPs near the STX1B gene. At least three significant loci, including the locus near SCN1A, were shared between GGE and focal epilepsy in the same study.

Focal Epilepsy

Sequencing studies have identified causative genes for several familial forms of focal epilepsy including LGI1 in autosomal dominant epilepsy with auditory features (Kalachikov et al., 2002) and KCNT1 in autosomal dominant sleep-related hypermotor epilepsy (Heron et al., 2012). Pathogenic variants in DEPDC5, which encodes a member of the GATOR1 complex involved in mTOR regulation, are identified in various forms of familial focal epilepsy (Picard et al., 2014; Scheffer et al., 2014). Pathogenic variants in NPRL2 and NPRL3, two additional members of the GATOR1 complex, have also been identified in familial focal epilepsies (Ricos et al., 2016; Weckhuysen et al., 2016). Notably, pathogenic variants in the GATOR complex genes have been identified in both nonlesional focal epilepsy and focal epilepsy with focal cortical dysplasia (FCD).

Similar to GGE, a case-control study using exome sequencing found enrichment of ultra-rare sequence variants in five known epilepsy genes (DEPDC5, LGI1, PCDH19, SCN1A, GRIN2A) in individuals with familial focal epilepsy (Epi4K Consortium, Epilepsy Phenome/Genome Project, 2017). Segregation in families of those with ultra-rare variants was not performed. GWAS of 9,671 individuals with focal epilepsy and 29,677 unaffected controls revealed only one significant locus on 2q24 near the SCN1A gene. Two additional loci at 2p16 and 16q12 were significant when a combined analysis of focal and generalized epilepsies was performed (International League Against Epilepsy Consortium on Complex Epilepsies, 2018).

Somatic mosaic variants have been recognized as an important cause of some focal cortical dysplasias (FCDs) and brain malformations associated with epilepsy (Heinzen, 2020). Somatic variants—often restricted to brain tissue—have been identified in several genes in the mTOR pathway, including PIK3CA, PIK3R2, AKT3, MTOR, and CCND2 among others, with associated phenotypes ranging from FCD to hemimegalencephaly (Pavone et al., 1991; Mirzaa et al., 2004, 2012, 2014; Poduri et al., 2012; Riviere et al., 2012). Somatic mosaic variants in X-linked SLC35A2 have been identified in up to 15% of individuals undergoing surgery for radiographically nonlesional focal epilepsy (Sim et al., 2018; Winawer et al., 2018; Baldassari et al., 2019).

Developmental and Epileptic Encephalopathy

The most significant advances in understanding the genetic etiology of epilepsy have been made in the developmental and epileptic encephalopathies (DEEs). The DEEs are the most severe group of epilepsies, characterized by intractable seizures and developmental delays or regression. Because DEEs often occur without any family history, these conditions were long thought to be sporadic and nongenetic.

An early clue to the role of de novo genetic variants in DEE came with the discovery of SCN1A as a cause of Dravet syndrome (Claes et al., 2001) through candidate gene sequencing. Candidate gene sequencing, which often focused on brain-expressed ion channels, identified rare pathogenic variants in patients with early-onset forms of epilepsy in genes, including GABRG2, KCNQ2, and KCNQ3 among others. NGS enabled rapid and relatively inexpensive sequencing of many or all genes (the exome) simultaneously to identify variants without the need to perform linkage analysis or to select candidate genes. Bypassing linkage and candidate gene selection was particularly important for DEE. As discussed below, many disease-causing variants are de novo, which makes linkage analysis impossible and explains the “sporadic” occurrence of DEE. Furthermore, unbiased sequencing of all genes highlighted novel classes of (non-ion channel) genes in the etiology of DEE.

The most efficient use of NGS for gene discovery has been widespread application of exome sequencing in trios of an affected child and both unaffected parents. This approach allows rapid filtering for variants that are present in the affected child but not inherited from either parent (de novo); on average, there is one de novo variant per exome (Acuna-Hidalgo et al., 2016). Sequencing parent-child trios is also efficient for identifying potentially pathogenic compound heterozygous, homozygous, and X-linked variants. Over the past decade, numerous studies using the trio exome approach confirmed that de novo, highly penetrant, single-gene variants are one of the most important causes of DEE (Rauch et al., 2012; Epi4K & EPGP Investigators, 2013; Suls et al., 2013; Euro et al., 2014). Dozens of genes in which de novo pathogenic variants cause DEE have been identified, along with numerous genes that cause X-linked or autosomal recessive DEE (Calhoun and Carvill, 2018). While many of the new genes identified involved ion channels, important new classes of genes in which pathogenic variants can cause DEE include genes involved in transcriptional regulation (Carvill et al., 2013; Nakajima et al., 2014), intracellular signaling (Nakamura et al., 2013; Muir et al., 2021), synaptic vesicle trafficking (Rohena et al., 2013), and others (Myers and Mefford 2015; Muona et al., 2016).

We have made remarkable progress in understanding the genetic architecture of the major classes of epilepsy. For DEE, genetic testing in the clinical setting yields a diagnosis in up to 40%–50% of affected individuals when exome sequencing and chromosome array is applied. Yet 50% or more remain without a genetic diagnosis after the same tests. Furthermore, much of the heritability remains to be explained in GGE and focal epilepsy. Potential strategies to fill the gaps exist or are emerging and include both novel analytic approaches and the implementation of new or different genome-wide technologies to detect variants. Avenues of research in the next decade will include genome sequencing, epigenetic studies, approaches to identify low-level mosaicism, and studying alternative inheritance models, including oligogenic inheritance and polygenic risk (Table 41.1).

Table 41.1

Detection of Broad Categories of Genomic Variation.

Genome Sequencing

Exome sequencing interrogates the ~1% of the genome that is protein-coding, and exome sequencing studies in families and in cohorts of affected individuals has had a significant impact on our understanding of epilepsy genetics, especially for DEE. Limiting sequence generation and analysis to 1% of the genome is practical for two reasons: cost and interpretation. When NGS was introduced, the cost of exome sequencing was much less than genome sequencing, though the cost differential has decreased significantly in recent years. Sequencing the exome to identify disease-causing variants has also been a sensible approach, because interpretation of variants in the exome is relatively straightforward with knowledge of the amino acid code and canonical splice site sequences. Furthermore, the vast majority of pathogenic variants identified to date are in protein-coding regions of the genome (Chong et al., 2015).

However, variants outside of coding regions also play an important role in disease etiology. Examples include pathogenic variants in promoter regions (Martyn et al., 2019), 5′ and 3′ untranslated regions (Johnston et al., 2019; Zhang et al., 2020), poison exons (Carvill et al., 2018), cis-regulatory elements (Weedon et al., 2014), and deep intronic regions (Kapoor et al., 2008). While many of the examples cited were identified by sequencing candidate genes, genome sequencing would have identified all of the variants and is increasingly used to investigate the role of noncoding variants across many human disorders. A handful of studies have applied trio genome sequencing to DEE, though the number of individuals studied is modest (n = 6, 14, 30, and 197) (Martin et al., 2014; Hamdan et al., 2017; Ostrander et al., 2018; Qaiser et al., 2021), and analysis in each study focused almost exclusively on the coding exome. Genome sequencing of additional studies of large cohorts of unsolved DEE are likely to reveal pathogenic, noncoding variants.

A major challenge for genome sequencing is variant interpretation. The number of variants, even ultra-rare variants, is much greater in a genome than in an exome, and variant interpretation in individuals who do not have European ancestry can be more difficult due to current lack of diverse population frequency data. Even filtering for de novo variants leaves on the order of 100 variants per individual to evaluate compared to approximately one de novo variant per exome (Acuna-Hidalgo et al., 2016). In addition, the impact of most exonic variants can be predicted using the amino acid code, and knowledge of splicing mechanisms can be used to predict the impact of nearby intronic variants. Predicting the effect of variants outside of the exome remains a significant challenge, and experimental validation is usually required to confirm the effect of a noncoding variant on gene expression or splicing. Major efforts to understand which regions of the genome are important for regulation of gene expression are ongoing (Gasperini et al., 2019). Other efforts to characterize a complete and more diverse reference genome through the Telomere-to-Telomere (T2T) (Nurk et al., 2022) and Human Pangenome Reference Consortium (Lia et al., 2023) projects will also improve variant interpretation.

Long-Read Sequencing

The most frequently employed genome sequencing technology is short-read genome sequencing, in which billions of short (100–200 bp) fragments of DNA are simultaneously sequenced. Short-read genome sequencing identifies sequence variants and some copy number variants efficiently, but it misses other types of variants that can be important causes of disease, including epilepsy (Cen et al., 2018; Ishiura et al., 2018). These include structural variants, simple repeat expansions, and epigenetic variants. Long-read sequencing technology, which has emerged more recently, enables sequencing of much longer stretches up DNA that range from several kilobases to several megabases depending on the technology; the longer reads may encompass an entire structural change, providing breakpoints, or full repeat expansions, with both sequence content and size. Finally, some long-read technologies also generate data about DNA modifications such as methylation. Long-read genome sequencing of six proband-parent trios with previously unsolved neurodevelopmental disorders revealed a pathogenic L1-mediated insertion in one case and a complex structural variant in another case, both of which were missed by short-read genome sequencing (Hiatt et al., 2021). In another study, targeted use of long-read sequencing was used to identify the “missing variant” in 2/2 individuals with an X-linked phenotype without a known variant and in 6/8 individuals who had a phenotype suggestive of a recessive disorder but only one pathogenic variant detected by prior testing (Miller et al., 2021). Variants detected by long-read technology included mobile element insertions, a repeat expansion, and an inversion, among others. In the same study, targeted long-read analysis was able to simultaneously determine repeat expansion length, sequence content, and methylation status for a previously identified repeat expansion (LaCroix et al., 2019; Miller et al., 2021). These pilot studies suggest that targeted or whole-genome sequencing with long-read technologies will identify novel variants that have been missed by other platforms.

Epigenetics

Epigenetic changes are often studied as a consequence of disease (or aging or environment), but epigenetic changes can also be etiologic. A classic example of a disorder due to aberrant methylation is Fragile X syndrome, where hypermethylation of an expanded CGG repeat in the 5′UTR of FMR1 prevents gene expression. Although the aberrant methylation is due to an underlying DNA change (CGG repeat expansion), the repeat is difficult to detect using standard sequence analysis. Similarly, Prader-Willi syndrome is an imprinting disorder that may arise due to one of several DNA changes (deletion, uniparental disomy, mutation), all of which exhibit the same methylation difference. Genome-wide methylation studies in a cohort of individuals with developmental delays or congenital anomalies suggests that de novo, constitutional changes in methylation may contribute to the development of neurodevelopmental disorders in a subset of individuals without a clear genetic cause (Barbosa et al., 2018). In that study, over 20% of individuals had a rare, differentially methylated region; in a subset where parents were available for testing, nearly half occurred de novo, suggesting the methylation change may be related to the phenotype. Similarly, epigenetic variants may contribute to DEE, explaining a subset of unexplained cases, and should be investigated.

Genome-wide methylation studies have also been used to identify distinct methylation “signatures” associated with specific genetic syndromes and can be used as a diagnostic tool (Kerkhof et al., 2021; Sadikovic et al., 2021). Even when a pathogenic sequence variant has not been identified, detection of a gene-specific methylation signature is suggestive of a genetic diagnosis and can inform analysis or interpretation of variants of uncertain significance or noncoding variants, for example. Identifying methylation signatures specific to genetic DEE syndromes will help guide diagnosis and may also help resolve uncertain variants in known epilepsy genes. Assigning unsolved cases to a methylation signature may guide analysis of genome data to identify causative noncoding or epigenetic variants in some affected individuals.

Multiomics

Beyond methylation studies, other -omics and advanced computational approaches are instrumental in the discovery and understanding of novel causative mechanisms in epilepsy. Functional studies in multiple cell types, organoids, and animal models provide insight into underlying disease etiology. As an example, one transcriptomics study performed single-nucleus transcriptomics of >110,000 neuronal transcriptomes from temporal cortex in samples from multiple individuals with temporal lobe epilepsy and controls and identified differences in epilepsy-associated neuronal subtypes and gene expression underlying epileptogenesis (Pfisterer et al., 2020). Metabolomics has been used to characterize metabolic pathways in epilepsy; linking gene expression and histologic data with metabolomic data can help identify specific biomarkers and therapeutic targets (Wu et al., 2017; Lai et al., 2022). Comparative proteomic studies have the potential to identify protein presence, abundance, and dynamics (do Canto et al., 2020). Performing comparative proteomic research using mouse models has led to identification of proteomic signatures for Dravet syndrome in Scn1a-A1783V mouse model (Miljanovic et al., 2021), and surgical and postmortem brain tissue research in humans is ongoing in epilepsy and neurodegenerative diseases (do Canto et al., 2020). Repositories for phenotypic data like the Enhancing Neuroimaging Genetics Through Meta-Analysis (ENIGMA), the NIH Epilepsy Phenome/Genome Project (EPGP), and the EuroEPINOMICS-RES Consortium have been established to create large data sets that can be interrogated by computational methods (Lhatoo et al., 2020). Integration of transcriptomics, proteomics, metabolomics, other functional omics, and Big Data science (including artificial intelligence) with genomic and rich phenotypic data will lead to a more holistic and complete understanding of epilepsy disease mechanisms.

Mosaicism

Somatic mosaic variants are clearly important for some focal epilepsies, specifically for FCDs. In many respects, FCDs are similar to tumors, so the role of somatic mutation makes intuitive sense; furthermore, surgical resection of tissue in some cases facilitates sequencing in the affected tissue and allows detection of variants that are present only in the affected tissue. It is plausible that brain-restricted, somatic, mosaic variants could also cause DEE that is indistinguishable from DEE due to heterozygous de novo variants in the same gene; however, detecting such variants is challenging since most individuals with DEE do not undergo surgery. Methods under investigation include retrieval of tissue from invasive EEG electrodes and analysis of cell-free DNA from cerebral spinal fluid (CSF) (Ye et al., 2021a, 2021b), both of which are highly promising. However, with current methods, the yield of DNA is extremely low, which only allows for highly targeted testing. While targeted testing is likely to be useful for some phenotypes in which there are a few, recurrent, hotspot variants, technological advances will be required for these approaches to be efficient and clinically useful if expanded sequencing (e.g., gene panel or exome) is required. If the variant is not restricted to the brain, (ultra-)deep sequencing of peripheral tissue such and blood or saliva may detect low-level, mosaic, pathogenic variants. There are numerous examples using the approach to detect both probands and parents (Myers et al., 2018; Moller et al., 2019) who harbor mosaic pathogenic variants.

Oligogenic and Polygenic Risk

Oligogenic and polygenic risk refer to genetic risk from a few or many different genes or variants, respectively, and should be explored in the epilepsies. GWAS in GGE and focal epilepsy have identified multiple common risk variants, each of which confers a very small risk of disease. The recent development of polygenic risk scores (PRSs) allows an estimate of disease risk that combines the small effects of risk variants across the human genome. At this time, clinical utility of PRS is limited by access to large samples sizes, particularly in rare disease, and a lack of diverse population control data. Historically, the vast majority of GWAS and PRS studies have been performed in European ancestry majority populations, though work to expand the diversity of PRS studies is ongoing (Popejoy and Fullerton, 2016). Large data sets from repositories like the All of Us Research Program and the UK Biobank will help provide additional, robust population genotyping data from a broader and more diverse cohort.

Efforts to develop PRSs for generalized and focal epilepsy are ongoing (Gramm et al., 2020; Leu et al., 2020), and with increased cohort sizes will be refined. PRSs will certainly be important for GGE and focal epilepsies, which are more common and exhibit complex inheritance, but exploration of the contribution of common variants to as-yet-unexplained DEE should be pursued as well. In addition, although the focus of gene discovery in DEE has been on highly penetrant, single-gene variants, another potential model to explore for the rare DEEs is the inheritance of a small number of variants of moderate impact (oligogenic model). PRS may be useful in understanding disease modifiers, including providing some explanation for the variation in penetrance between individuals with the same known single-gene cause for their epilepsy (Leu et al., 2019).

The Importance of Collaboration

The epilepsy genetics community has been highly collaborative over the past decade. In addition to individual labs collaborating across institutions, efforts such as Epi4K, Epi25, EuroEPINOMICS, the International League Against Epilepsy, and others have brought together very large cohorts of affected individuals for analysis, no doubt accelerating discovery. Gene-specific rare disease foundations and family groups—the majority of which were created by motivated parents of genetically diagnosed children with epilepsy—have been instrumental in building patient cohorts, connecting patients and families with experts, driving innovation in research guided by scientific advisory boards, and facilitating partnerships across institutions and with industry. Programs and tools such as GeneMatcher, Matchmaker Exchange, and MyGene2 connect patients with rare and uncertain genetic findings with clinicians, researchers, and other patients with interest in or known genetic findings in the same gene. National and international open data sharing policies are becoming more widespread and are also opening more opportunities for shared knowledge and discovery. Continued discovery will require continued collaboration. For GGE and focal epilepsies, large numbers of individuals from diverse backgrounds will be required to update GWAS and refine PRS calculation. For the rarer DEEs, each new genetic cause is likely to be a (very) rare cause of disease, so large, diverse cohorts will be required to identify multiple affected individuals with the same etiology.