Recent Advances in the Molecular Genetics of Epilepsy
Recent Advances in the Molecular Genetics of Epilepsy
Common inherited epilepsies such as the genetic generalised epilepsies (GGEs) including childhood absence epilepsy, juvenile absence epilepsy and juvenile myoclonic epilepsy show complex inheritance and, to date, only a small fraction of the susceptibility genes have been identified—these include rare variants in the calcium channel subunit CACNA1H and microdeletion 15q13.3. Until the genetic architecture of these disorders is better understood and the attributable risk of variants is known, clinical testing is not recommended. Although the variations are not yet clinically useful, it is likely they will be diagnostically screened in the future. A large number of association studies in epilepsy (n>165) have been performed but evidence for common variants influencing epilepsy risk has not emerged (CarpDB database—http://www.carpedb.ua.edu/; epiGAD database—http://www.epigad.org; July 2011). These studies have largely been underpowered and larger genome-wide association studies have only recently been reported.
Family-based linkage and heritability studies of GGEs suggest the existence of seizure-specific susceptibility loci in a heterogeneous genetic architecture (reviewed in ). Understanding this architecture is important because most epilepsy patients are sporadic with no family history of seizures and are thought to have a complex or polygenic basis. In these cases, a single susceptibility gene variant is typically not sufficient to cause epilepsy. Despite significant progress unravelling Mendelian forms of epilepsy it has proven problematic to apply this knowledge to sporadic cases. It has been postulated that identification of rare variants is likely to provide critical information on predisposition to epilepsy since such variants are more likely to be functionally significant than common variants. Characterisation of rare variants may also elucidate whether single variants of strong effect or a combination of multiple variants with weaker effects underlie susceptibility to epilepsy, an issue that needs resolving. Rare variants cannot be detected on tagSNP platforms that are used in association studies of common variants. Rather, they are identified by sequencing of candidate genes, entire exomes or genomes.
To this end, a recent report has provided insight into the complex inheritance of epilepsies of possible genetic origin, suggesting that the type and pattern of genetic variants, rather than the number of variants or whether they are common or rare, confer susceptibility to epilepsy. Since most known epilepsy genes encode ion channel proteins, a valid variant discovery approach is high-throughput sequencing of this functional group of genes in epilepsy patients. In the first study of this kind, Klassen and colleagues amplified and Sanger sequenced in parallel the exonic regions of 237 of the estimated 400 human channel protein genes in 152 patients diagnosed with idiopathic or cryptogenic epilepsies and 139 ethnicity- and age-matched control individuals. This sequencing detected a validated set of 3095 single nucleotide polymorphisms (SNPs) and revealed that an individual's channel phenotype is apparently unique as no two of the 291 individuals screened had the same ion channel variant complement.
Surprisingly, the distribution of the most functionally 'severe' variants between cases and controls did not provide evidence for an ion channel rare variant association in epilepsies of possible genetic aetiology. Even missense variants in known human epilepsy channel genes were commonly found in both cases and controls, supporting the complex inheritance of channelopathy phenotypes. The specific type and pattern of these variants, rather than the number (mutational load hypothesis), appeared to favour the disease phenotype. The authors concluded, based on their genotypic data and the assumed complexity of firing behaviour in brain circuits, that the pattern of genetic variation overrides the mutational load at the single cell level. They propose a variant pattern hypothesis for ion channelopathies stating that for a complex disease like the common genetic forms of epilepsy, a personalised channel SNP pathogenicity pattern must be defined in the context of all channel subunits and does not just depend on individual, functionally deleterious variants. If correct, and confirmed by other studies, this will provide a great challenge to individualising this information in the clinic.
Recently it has become increasingly clear that de novo mutations contribute to epilepsy pathogenesis. Although the size of this contribution is not yet known, a substantial contribution to the common epilepsies is suspected since de novo mutations have already been strongly implicated in severe epileptic encephalopathies and other neuropsychiatric disorders including schizophrenia, autism and mental retardation. In fact, de novo mutations may explain the common occurrence of these disorders despite their association with severely reduced fecundity. The increasing recognition of the role of de novo mutations in genetic disease is perhaps not surprising given recent reports indicating that the per generation mutation rate, estimated at between 7.6×10 to 2.2×10 in humans, is exceptionally high. The introduction of exome sequencing has established de novo paradigms for intellectual disability and autism spectrum disorders. In these studies rapid detection of de novo point mutations in case-parent trios was made possible by MPS technology.
In epilepsies, de novo mutagenesis has been shown for sequence variation including CNVs associated with both Mendelian and polygenic epilepsies. It is likely that de novo mutagenesis plays a greater role than previously appreciated in both mild and severe epilepsies. The application of MPS approaches in the coming years promises to elucidate the size of this contribution. The contribution of de novo mutations to epilepsy is exemplified by the discoveries involving the neuronal sodium channel. In pedigrees with mild genetic (formerly generalised) epilepsy with febrile seizures plus (GEFS+), familial mutations in the accessory β subunit and the pore-forming α subunit genes (SCN1B and SCN1A) were first described by traditional mapping and positional candidate methods. Then, in 2001, de novo SCN1A mutations were found in Dravet syndrome (formerly known as severe myoclonic epilepsy of infancy) which was regarded as the most severe phenotype of the GEFS+ spectrum. This observation has been repeatedly confirmed; 70%–80% of cases with Dravet syndrome have SCN1A mutations with more than 800 patients reported and genetic testing being widely used in clinical practice. Strikingly, in 90% of Dravet syndrome cases the SCN1A mutations identified are de novo.
This high rate of de novo mutation in Dravet syndrome demonstrates that, even in the absence of a family history, a genetic cause should be considered in the aetiology of epilepsy. This is particularly so in the infantile epileptic encephalopathies—a complex group of epilepsies characterised by severe epilepsy, intellectual disability and often regression; these are typically sporadic disorders. In addition to the typical occurrence of de novo SCN1A mutations in Dravet syndrome, other examples of de novo mutagenesis in epileptic encephalopathies include mutations in STXBP1, PCDH19, CDKL5, ARX, KCNQ2, SCN2A, PLCB1, SLC25A22, SPTAN1, SCN8A and SLC19A3. Distinctive epilepsy syndromes are emerging with mutations in particular genes, but unexpected heterogeneity and overlap have also been observed. Indeed, even for the paradigm of Dravet syndrome and SCN1A, other genes such as PCDH19 cause phenotypes that resemble Dravet syndrome, while de novo SCN1A mutations are observed in other syndromes including infantile spasms and migrating partial seizures of infancy. Milder epilepsy syndromes, including those where the individual may have their own children, are also associated with de novo mutations, but the scope of this has yet to be established. Examples include the CHRNA4 gene for nocturnal frontal lobe epilepsy, and the LGI1 gene for autosomal dominant partial epilepsy with auditory features, and even very mild self-limited epilepsies such as benign familial infantile epilepsy due to PRRT2 mutations.
The origin of de novo mutations is of considerable interest for counselling and understanding the biology. It appears that the contribution of the male germline is higher than that of the female germline due to the increased number of cell divisions. Recent studies comparing mutation rates between parent–offspring trios by genome sequencing have indicated that there actually exists considerable variability in genome-wide mutation frequency within and between human families. The authors investigated two families with disparate findings—in one family 92% of germline de novo mutations were of paternal origin, whereas in the other 64% were of maternal origin, with the caveat that maternal and paternal age at the time of conception was not known. Mutation rate has now even been successfully estimated within a founder population. Despite this apparent genome-wide variability, for at least one epilepsy locus a clear bias for paternally-derived de novo mutations has been demonstrated. As outlined above, in the majority of patients with Dravet syndrome de novo mutations in the SCN1A gene are the underlying cause. Genetic analysis of 44 patients carrying de novo mutations in SCN1A revealed paternal origin of the de novo mutation in 33 (75%) of cases. This selection for mutational events during spermatogenesis appears genuine since the average age of the parents at the time of conception did not differ from the general population.
Most de novo mutations appear to occur in the parental gametes, more often the father, and the recurrence risk is low. However, even when the mutation is absent in peripheral blood of the parents, recurrence is possible if there is gonadal mosaicism which is reported in Dravet syndrome. Similarly, somatic mosaicism, which in the case of SCN1A may manifest as a mild seizure disorder in the parent and Dravet syndrome in the child, has important counselling implications.
The timing of de novo mutational events can also be determined. For example, in one study of three twin pairs, heterozygous, protein truncating, de novo mutations in the SCN1A gene were detected. In one discordant monozygotic twin pair different embryonic tissue lineages were examined to determine the timing of the SCN1A mutation in the affected twin. Genotyping of skin fibroblasts, lymphocytes, hair follicles, buccal cells and cell lines derived from the olfactory neuroepithelium was positive for the SCN1A mutation in the affected twin but not in the unaffected twin, their parents or older sibling. Somatic mosaicism was ruled out in this twin pair and their parents by allele-specific and semiquantitative PCR. Based on these results the authors concluded the mutation most likely occurred in the premorula embryo. This demonstrates that de novo mutagenesis can occur at any stage of the sexual cycle—usually in the parental gametes, but also the very early embryo or later in embryonic or fetal development.
Somatic mosaicism confined to the brain or neuroectodermal tissues has recently been shown for certain sporadic brain malformation syndromes presenting with seizures, further enlarging the spectrum of genetic epilepsies. Discovery was possible by directly examining brain tissue specimens removed at surgery, but in some cases may also be detectable using cheek cells. If mutation confined to the neuroectodermal tissues is more widespread in the aetiology of commoner epilepsies, screening investigations may need to include tissues other than traditional peripheral blood samples.
Detection of de novo epilepsy-causing mutations also leads to unexpected insights and clinical benefit, as in the case of 'vaccine encephalopathy'. The so-called 'vaccine encephalopathy' is where infants suffer refractory seizures and developmental slowing shortly after vaccination. The recognition that cases resembled Dravet syndrome led to the hypothesis that SCN1A mutations may underlie these cases. Pathogenic SCN1A mutations were revealed in 11 of 14 patients, and at least nine mutations arose de novo showing that the underlying cause of the 'vaccine encephalopathy' was genetic. This dispelled the view that the vaccine caused the devastating neurological disorder; rather it was a disorder that the infant was destined to have and the de novo mutation explained the absence of a family history of seizures. A follow-up study showed that, although vaccination was associated with the onset of Dravet syndrome due to an SCN1A mutation and was a possible trigger to the first seizure, it did not change the outcome of the disorder. These findings were recently confirmed in an independent study.
Large-scale variations in human genomic DNA sequence are common, even in healthy individuals. Variations of at least 1 kb in length are defined as CNVs, and it has been estimated that there are around 1500 regions of variable copy number spread across 360 Mb (~12%) of the human genome. The average CNV is 250 kb in length and often includes many known genes. CNVs occur at specific chromosomal loci that are sites of interspersed segmental duplications sensitised to copy number variation by unequal crossing over during meiosis. Traditionally CNVs have been challenging to identify and map as they cannot be detected by routine karyotyping. Higher resolution molecular techniques including long-range PCR, genotyping arrays, SNP assays, multiplex ligation-dependent probe amplification and array comparative genomic hybridisation can effectively identify these variations. The development of these approaches has led to an increase in the number of CNVs linked to neurological and neuropsychiatric diseases including epilepsy. Surprisingly, CNVs can cause both monogenic disease (familial or de novo) or act as risk alleles. Recent high-profile studies have linked large deletions and duplications to mental retardation, schizophrenia and autism. Remarkably, a subset of these deletions (at 15q13.3 and 16p13.11) has now been associated with GGE, a relatively mild and common form of epilepsy.
The first major site containing CNVs that predispose to GGE was reported by Helbig and colleagues in 2009. Based on a previous study linking microdeletion at 15q13.3 to disease in 0.2%–0.3% of patients with a mental retardation and epilepsy syndrome, and the mapping of known GGE loci to 15q13–14, the authors genotyped two cohorts of GGE cases with European ancestry using different screening methods. 15q13.3 Microdeletions were identified in 12 of 1223 (0.98%) GGE cases but none of 3699 matched controls. Surprisingly deletions but not duplications at this locus were associated with epilepsy since the same number of GGE cases had duplications as controls (23/3699). The association of 15q13.3 microdeletions with GGE was confirmed in a study of 539 Australian and European GGE cases in which 7 (1.3%) were shown to carry deletions absent in 3777 matched controls. Subsequent investigations in independent populations have revealed a further six microdeletions in GGE cases, bringing the total number of 15q13.3 microdeletions implicated in epilepsy to 25 (Table 2).
These early findings led to additional screens that have revealed a large number of CNVs associated with epilepsies at other genomic loci including some hotspots for recurrent deletions. The most common site of recurrent microdeletion has been detected at 16p13.11 in patients exhibiting a diverse set of epilepsy syndromes. In these three separate studies consisting of a combined 5563 cases and 6814 controls of European or North American origin, 16p13.11 microdeletions were significantly overrepresented in cases compared with controls being detected in a total of 34 cases (Table 2). In the same set of studies recurrent microdeletions at 15q11.2 were also found to be associated with GGEs. These deletions were significantly more frequent in cases than controls and in total 19 cases were reported (Table 2).
Many rare CNVs at other genomic loci have also been revealed in these extensive screens including for epileptic encephalopathies. Since most of these loci do not contain known epilepsy genes these variable copy number genes represent new candidates. In some of these rare CNVs known genes do reside. For example, in one syndromic patient with mental retardation, neurobehavioural abnormalities and infantile seizures, a ~110 kb deletion at 2q24.3 was identified that involves part of the SCN2A and SCN3A genes. These genes are in a cluster of five brain-expressed sodium channel subunit genes and their mutation has been implicated in childhood epilepsies including benign familial neonatal–infantile seizures (SCN2A) and childhood onset focal epilepsy (SCN3A). Recently a micro-duplication syndrome of the 2q24.3 region including the SCN1A, SCN2A and SCN3A genes has been recognised in cases and families with neonatal seizures. In addition, mutations that delete or truncate the SCN1A gene and lead to haploinsufficiency are associated with Dravet syndrome in the GEFS+ spectrum.
These converging studies indicate that there are two broad classes of CNVs that should be considered when determining epilepsy aetiology. The first is recurrent CNVs that are either inherited or arise de novo and confer risk to common epilepsies. The second is non-recurrent, typically large CNVs that arise de novo and cause severe epilepsies including the epileptic encephalopathies. Both are significant causes of epilepsy but the former class has been better studied at the molecular level. The most common locus for CNVs in epilepsy patients so far reported is 16p13.11 (~ 0.6% of cases), although 15p13.3 microdeletions remain the most frequently observed CNVs in GGE patients (~1% of cases). Numerous rare CNVs like those at the 2q24.3 locus have already been reported. The recent introduction of high-throughput massively parallel sequencers that can readily generate paired-end sequence of genomic targets is certain to increase the detection of CNVs in epilepsy patients. This approach exploits the principles of paired-end mapping, where the distance between nucleotides at either end of random genomic fragments relative to a reference genome are used to map deletions or insertions. The application of this technology to epilepsy is discussed in the next section.
Recent Advances in the Molecular Genetics of Epilepsies: Contribution of Susceptibility Alleles, De Novo Mutations and CNV
Susceptibility Alleles
Common inherited epilepsies such as the genetic generalised epilepsies (GGEs) including childhood absence epilepsy, juvenile absence epilepsy and juvenile myoclonic epilepsy show complex inheritance and, to date, only a small fraction of the susceptibility genes have been identified—these include rare variants in the calcium channel subunit CACNA1H and microdeletion 15q13.3. Until the genetic architecture of these disorders is better understood and the attributable risk of variants is known, clinical testing is not recommended. Although the variations are not yet clinically useful, it is likely they will be diagnostically screened in the future. A large number of association studies in epilepsy (n>165) have been performed but evidence for common variants influencing epilepsy risk has not emerged (CarpDB database—http://www.carpedb.ua.edu/; epiGAD database—http://www.epigad.org; July 2011). These studies have largely been underpowered and larger genome-wide association studies have only recently been reported.
Family-based linkage and heritability studies of GGEs suggest the existence of seizure-specific susceptibility loci in a heterogeneous genetic architecture (reviewed in ). Understanding this architecture is important because most epilepsy patients are sporadic with no family history of seizures and are thought to have a complex or polygenic basis. In these cases, a single susceptibility gene variant is typically not sufficient to cause epilepsy. Despite significant progress unravelling Mendelian forms of epilepsy it has proven problematic to apply this knowledge to sporadic cases. It has been postulated that identification of rare variants is likely to provide critical information on predisposition to epilepsy since such variants are more likely to be functionally significant than common variants. Characterisation of rare variants may also elucidate whether single variants of strong effect or a combination of multiple variants with weaker effects underlie susceptibility to epilepsy, an issue that needs resolving. Rare variants cannot be detected on tagSNP platforms that are used in association studies of common variants. Rather, they are identified by sequencing of candidate genes, entire exomes or genomes.
To this end, a recent report has provided insight into the complex inheritance of epilepsies of possible genetic origin, suggesting that the type and pattern of genetic variants, rather than the number of variants or whether they are common or rare, confer susceptibility to epilepsy. Since most known epilepsy genes encode ion channel proteins, a valid variant discovery approach is high-throughput sequencing of this functional group of genes in epilepsy patients. In the first study of this kind, Klassen and colleagues amplified and Sanger sequenced in parallel the exonic regions of 237 of the estimated 400 human channel protein genes in 152 patients diagnosed with idiopathic or cryptogenic epilepsies and 139 ethnicity- and age-matched control individuals. This sequencing detected a validated set of 3095 single nucleotide polymorphisms (SNPs) and revealed that an individual's channel phenotype is apparently unique as no two of the 291 individuals screened had the same ion channel variant complement.
Surprisingly, the distribution of the most functionally 'severe' variants between cases and controls did not provide evidence for an ion channel rare variant association in epilepsies of possible genetic aetiology. Even missense variants in known human epilepsy channel genes were commonly found in both cases and controls, supporting the complex inheritance of channelopathy phenotypes. The specific type and pattern of these variants, rather than the number (mutational load hypothesis), appeared to favour the disease phenotype. The authors concluded, based on their genotypic data and the assumed complexity of firing behaviour in brain circuits, that the pattern of genetic variation overrides the mutational load at the single cell level. They propose a variant pattern hypothesis for ion channelopathies stating that for a complex disease like the common genetic forms of epilepsy, a personalised channel SNP pathogenicity pattern must be defined in the context of all channel subunits and does not just depend on individual, functionally deleterious variants. If correct, and confirmed by other studies, this will provide a great challenge to individualising this information in the clinic.
De Novo Mutations
Recently it has become increasingly clear that de novo mutations contribute to epilepsy pathogenesis. Although the size of this contribution is not yet known, a substantial contribution to the common epilepsies is suspected since de novo mutations have already been strongly implicated in severe epileptic encephalopathies and other neuropsychiatric disorders including schizophrenia, autism and mental retardation. In fact, de novo mutations may explain the common occurrence of these disorders despite their association with severely reduced fecundity. The increasing recognition of the role of de novo mutations in genetic disease is perhaps not surprising given recent reports indicating that the per generation mutation rate, estimated at between 7.6×10 to 2.2×10 in humans, is exceptionally high. The introduction of exome sequencing has established de novo paradigms for intellectual disability and autism spectrum disorders. In these studies rapid detection of de novo point mutations in case-parent trios was made possible by MPS technology.
In epilepsies, de novo mutagenesis has been shown for sequence variation including CNVs associated with both Mendelian and polygenic epilepsies. It is likely that de novo mutagenesis plays a greater role than previously appreciated in both mild and severe epilepsies. The application of MPS approaches in the coming years promises to elucidate the size of this contribution. The contribution of de novo mutations to epilepsy is exemplified by the discoveries involving the neuronal sodium channel. In pedigrees with mild genetic (formerly generalised) epilepsy with febrile seizures plus (GEFS+), familial mutations in the accessory β subunit and the pore-forming α subunit genes (SCN1B and SCN1A) were first described by traditional mapping and positional candidate methods. Then, in 2001, de novo SCN1A mutations were found in Dravet syndrome (formerly known as severe myoclonic epilepsy of infancy) which was regarded as the most severe phenotype of the GEFS+ spectrum. This observation has been repeatedly confirmed; 70%–80% of cases with Dravet syndrome have SCN1A mutations with more than 800 patients reported and genetic testing being widely used in clinical practice. Strikingly, in 90% of Dravet syndrome cases the SCN1A mutations identified are de novo.
This high rate of de novo mutation in Dravet syndrome demonstrates that, even in the absence of a family history, a genetic cause should be considered in the aetiology of epilepsy. This is particularly so in the infantile epileptic encephalopathies—a complex group of epilepsies characterised by severe epilepsy, intellectual disability and often regression; these are typically sporadic disorders. In addition to the typical occurrence of de novo SCN1A mutations in Dravet syndrome, other examples of de novo mutagenesis in epileptic encephalopathies include mutations in STXBP1, PCDH19, CDKL5, ARX, KCNQ2, SCN2A, PLCB1, SLC25A22, SPTAN1, SCN8A and SLC19A3. Distinctive epilepsy syndromes are emerging with mutations in particular genes, but unexpected heterogeneity and overlap have also been observed. Indeed, even for the paradigm of Dravet syndrome and SCN1A, other genes such as PCDH19 cause phenotypes that resemble Dravet syndrome, while de novo SCN1A mutations are observed in other syndromes including infantile spasms and migrating partial seizures of infancy. Milder epilepsy syndromes, including those where the individual may have their own children, are also associated with de novo mutations, but the scope of this has yet to be established. Examples include the CHRNA4 gene for nocturnal frontal lobe epilepsy, and the LGI1 gene for autosomal dominant partial epilepsy with auditory features, and even very mild self-limited epilepsies such as benign familial infantile epilepsy due to PRRT2 mutations.
The origin of de novo mutations is of considerable interest for counselling and understanding the biology. It appears that the contribution of the male germline is higher than that of the female germline due to the increased number of cell divisions. Recent studies comparing mutation rates between parent–offspring trios by genome sequencing have indicated that there actually exists considerable variability in genome-wide mutation frequency within and between human families. The authors investigated two families with disparate findings—in one family 92% of germline de novo mutations were of paternal origin, whereas in the other 64% were of maternal origin, with the caveat that maternal and paternal age at the time of conception was not known. Mutation rate has now even been successfully estimated within a founder population. Despite this apparent genome-wide variability, for at least one epilepsy locus a clear bias for paternally-derived de novo mutations has been demonstrated. As outlined above, in the majority of patients with Dravet syndrome de novo mutations in the SCN1A gene are the underlying cause. Genetic analysis of 44 patients carrying de novo mutations in SCN1A revealed paternal origin of the de novo mutation in 33 (75%) of cases. This selection for mutational events during spermatogenesis appears genuine since the average age of the parents at the time of conception did not differ from the general population.
Most de novo mutations appear to occur in the parental gametes, more often the father, and the recurrence risk is low. However, even when the mutation is absent in peripheral blood of the parents, recurrence is possible if there is gonadal mosaicism which is reported in Dravet syndrome. Similarly, somatic mosaicism, which in the case of SCN1A may manifest as a mild seizure disorder in the parent and Dravet syndrome in the child, has important counselling implications.
The timing of de novo mutational events can also be determined. For example, in one study of three twin pairs, heterozygous, protein truncating, de novo mutations in the SCN1A gene were detected. In one discordant monozygotic twin pair different embryonic tissue lineages were examined to determine the timing of the SCN1A mutation in the affected twin. Genotyping of skin fibroblasts, lymphocytes, hair follicles, buccal cells and cell lines derived from the olfactory neuroepithelium was positive for the SCN1A mutation in the affected twin but not in the unaffected twin, their parents or older sibling. Somatic mosaicism was ruled out in this twin pair and their parents by allele-specific and semiquantitative PCR. Based on these results the authors concluded the mutation most likely occurred in the premorula embryo. This demonstrates that de novo mutagenesis can occur at any stage of the sexual cycle—usually in the parental gametes, but also the very early embryo or later in embryonic or fetal development.
Somatic mosaicism confined to the brain or neuroectodermal tissues has recently been shown for certain sporadic brain malformation syndromes presenting with seizures, further enlarging the spectrum of genetic epilepsies. Discovery was possible by directly examining brain tissue specimens removed at surgery, but in some cases may also be detectable using cheek cells. If mutation confined to the neuroectodermal tissues is more widespread in the aetiology of commoner epilepsies, screening investigations may need to include tissues other than traditional peripheral blood samples.
Detection of de novo epilepsy-causing mutations also leads to unexpected insights and clinical benefit, as in the case of 'vaccine encephalopathy'. The so-called 'vaccine encephalopathy' is where infants suffer refractory seizures and developmental slowing shortly after vaccination. The recognition that cases resembled Dravet syndrome led to the hypothesis that SCN1A mutations may underlie these cases. Pathogenic SCN1A mutations were revealed in 11 of 14 patients, and at least nine mutations arose de novo showing that the underlying cause of the 'vaccine encephalopathy' was genetic. This dispelled the view that the vaccine caused the devastating neurological disorder; rather it was a disorder that the infant was destined to have and the de novo mutation explained the absence of a family history of seizures. A follow-up study showed that, although vaccination was associated with the onset of Dravet syndrome due to an SCN1A mutation and was a possible trigger to the first seizure, it did not change the outcome of the disorder. These findings were recently confirmed in an independent study.
Copy Number Variations
Large-scale variations in human genomic DNA sequence are common, even in healthy individuals. Variations of at least 1 kb in length are defined as CNVs, and it has been estimated that there are around 1500 regions of variable copy number spread across 360 Mb (~12%) of the human genome. The average CNV is 250 kb in length and often includes many known genes. CNVs occur at specific chromosomal loci that are sites of interspersed segmental duplications sensitised to copy number variation by unequal crossing over during meiosis. Traditionally CNVs have been challenging to identify and map as they cannot be detected by routine karyotyping. Higher resolution molecular techniques including long-range PCR, genotyping arrays, SNP assays, multiplex ligation-dependent probe amplification and array comparative genomic hybridisation can effectively identify these variations. The development of these approaches has led to an increase in the number of CNVs linked to neurological and neuropsychiatric diseases including epilepsy. Surprisingly, CNVs can cause both monogenic disease (familial or de novo) or act as risk alleles. Recent high-profile studies have linked large deletions and duplications to mental retardation, schizophrenia and autism. Remarkably, a subset of these deletions (at 15q13.3 and 16p13.11) has now been associated with GGE, a relatively mild and common form of epilepsy.
The first major site containing CNVs that predispose to GGE was reported by Helbig and colleagues in 2009. Based on a previous study linking microdeletion at 15q13.3 to disease in 0.2%–0.3% of patients with a mental retardation and epilepsy syndrome, and the mapping of known GGE loci to 15q13–14, the authors genotyped two cohorts of GGE cases with European ancestry using different screening methods. 15q13.3 Microdeletions were identified in 12 of 1223 (0.98%) GGE cases but none of 3699 matched controls. Surprisingly deletions but not duplications at this locus were associated with epilepsy since the same number of GGE cases had duplications as controls (23/3699). The association of 15q13.3 microdeletions with GGE was confirmed in a study of 539 Australian and European GGE cases in which 7 (1.3%) were shown to carry deletions absent in 3777 matched controls. Subsequent investigations in independent populations have revealed a further six microdeletions in GGE cases, bringing the total number of 15q13.3 microdeletions implicated in epilepsy to 25 (Table 2).
These early findings led to additional screens that have revealed a large number of CNVs associated with epilepsies at other genomic loci including some hotspots for recurrent deletions. The most common site of recurrent microdeletion has been detected at 16p13.11 in patients exhibiting a diverse set of epilepsy syndromes. In these three separate studies consisting of a combined 5563 cases and 6814 controls of European or North American origin, 16p13.11 microdeletions were significantly overrepresented in cases compared with controls being detected in a total of 34 cases (Table 2). In the same set of studies recurrent microdeletions at 15q11.2 were also found to be associated with GGEs. These deletions were significantly more frequent in cases than controls and in total 19 cases were reported (Table 2).
Many rare CNVs at other genomic loci have also been revealed in these extensive screens including for epileptic encephalopathies. Since most of these loci do not contain known epilepsy genes these variable copy number genes represent new candidates. In some of these rare CNVs known genes do reside. For example, in one syndromic patient with mental retardation, neurobehavioural abnormalities and infantile seizures, a ~110 kb deletion at 2q24.3 was identified that involves part of the SCN2A and SCN3A genes. These genes are in a cluster of five brain-expressed sodium channel subunit genes and their mutation has been implicated in childhood epilepsies including benign familial neonatal–infantile seizures (SCN2A) and childhood onset focal epilepsy (SCN3A). Recently a micro-duplication syndrome of the 2q24.3 region including the SCN1A, SCN2A and SCN3A genes has been recognised in cases and families with neonatal seizures. In addition, mutations that delete or truncate the SCN1A gene and lead to haploinsufficiency are associated with Dravet syndrome in the GEFS+ spectrum.
These converging studies indicate that there are two broad classes of CNVs that should be considered when determining epilepsy aetiology. The first is recurrent CNVs that are either inherited or arise de novo and confer risk to common epilepsies. The second is non-recurrent, typically large CNVs that arise de novo and cause severe epilepsies including the epileptic encephalopathies. Both are significant causes of epilepsy but the former class has been better studied at the molecular level. The most common locus for CNVs in epilepsy patients so far reported is 16p13.11 (~ 0.6% of cases), although 15p13.3 microdeletions remain the most frequently observed CNVs in GGE patients (~1% of cases). Numerous rare CNVs like those at the 2q24.3 locus have already been reported. The recent introduction of high-throughput massively parallel sequencers that can readily generate paired-end sequence of genomic targets is certain to increase the detection of CNVs in epilepsy patients. This approach exploits the principles of paired-end mapping, where the distance between nucleotides at either end of random genomic fragments relative to a reference genome are used to map deletions or insertions. The application of this technology to epilepsy is discussed in the next section.