Introduction

Epilepsy is a one of the commonest neurological diseases, with a lifetime risk of 3 % [1]. Genetics is thought to play a role in at least 70 % of patients with epilepsy [2], both through the action of multiple genetic risk factors in common epilepsies, such as childhood absence epilepsy or juvenile myoclonic epilepsy, and through single gene mutations in rare monogenic epilepsy syndromes. The list of epilepsy genes is rapidly growing, and several of these genes can now be readily screened in clinical practice.

In the last few years, major advances have been made in the field of monogenic epilepsy syndromes. The rate of gene discovery in these syndromes has mainly been driven by the availability of genomic sequencing techniques. At the time of the description of the first epilepsy gene, CHRNA4 in 1995 [3], genetic studies were mainly restricted to large families with inherited (mild) forms of epilepsy. With use of linkage analysis, the genetic cause of familial epilepsy syndromes, such as autosomal dominant nocturnal frontal lobe epilepsy (ADNFLE) and genetic epilepsy with febrile seizures plus (GEFS+), was described [4, 5]. Most of these gene findings supported the still prevailing concept of epilepsies being channelopathies.

Linkage analysis did, however, not allow gene discovery in isolated patients with epilepsy. Genetic factors in patients with epileptic encephalopathies (EE), severe forms of epilepsies associated with developmental stagnation or regression, therefore remained poorly studied for a long time. Candidate gene screening has been a successful approach to tackle this problem. The best known example is the detection of mutations in SCN1A, a gene initially identified in GEFS+ families, in Dravet syndrome patients [6]. This candidate gene approach is still leading to major breakthroughs, as shown by the recent elucidation of the cause of some of the most devastating epilepsies, such as malignant migrating partial seizures of infancy (MMPSI) [7] and other forms of neonatal EE [8••]. Nevertheless, this approach creates a bias towards already known genes and pathways. A first step towards an unbiased gene discovery approach has been the advent of microarray techniques allowing genome-wide detection of microdeletions and duplications, and leading to the discovery of EE genes such as STXBP1 and GRIN2A [9, 10••, 11••, 12••]. It was, however, the introduction of massive parallel sequencing technology, used to rapidly and cost-effectively screen a whole exome or a whole genome, that has led to a real explosion of genetic discoveries in epilepsy. This technology has finally enabled us to tackle the genetics of both monogenic familial and sporadic epilepsies, without the need to have prior assumptions. The rapid speed with which new genes are being described lately confronts clinicians with the difficult task of keeping up to date with the continuous supply of new publications. This article aims to discuss some of the genes that were recently discovered in monogenic familial epilepsy syndromes or EE for which an underlying cause remained unknown for a long time.

Neonatal Epileptic Encephalopathy and KCNQ2

Early in 2012, de novo missense mutations in KCNQ2 were identified in 10 % of patients with neonatal EE with a presumed genetic cause [8••]. Since this report, several studies have confirmed this high frequency of mutations in patients with neonatal EE, and have further delineated the phenotype associated with KCNQ2 encephalopathy [13–15].

This gene was, however, not unfamiliar to pediatric neurologists. KCNQ2 and KCNQ3, encoding the voltage-gated potassium channel subunits Kv7.2 and Kv7.3, were long known to be present in 60-70 % of families with benign familial neonatal epilepsy (BFNE) [16]. Children with this benign autosomal dominant epilepsy syndrome experience seizures in the first week of life. The seizures often begin with a tonic component followed by a range of autonomic and motor changes, and are often accompanied by apnea. The seizures are brief but can occur up to 30 times a day. Prognosis is, however, favorable: the seizures remit between 1 and 12 months of age, and the children develop normally.

Children with KCNQ2 encephalopathy have a similar early neonatal onset, and present with similar seizure symptoms. In contrast to the EEG in BFNE, however, the EEG is grossly abnormal, showing a burst suppression pattern or multifocal epileptic activity, and neonates soon show delayed development with axial hypotonia and lack of visual tracking. The seizures are difficult to treat, although some patients respond well to sodium channel blockers such as phenytoin and carbamazepine. Two thirds of patients become seizure-free some months or years after seizure onset, but all children show developmental delay, ranging from mild to profound intellectual disability. Older children often have axial hypotonia, sometimes associated with limb spasticity, and most patients are not ambulatory.

In over half of patients, early MRI of the brain shows hyperintensities in basal ganglia and thalamus that later resolve. At a later stage, cerebral atrophy and a thin corpus callosum can be seen, correlating with the severity of developmental delay.

Almost all reported patients with KCNQ2 encephalopathy had seizure onset in the first week of life, but one patient with onset at 5 months has been reported. A similar spectrum of onset age is also seen for KCNQ2 mutations in benign infantile epilepsies: a recent report found KCNQ2 mutations in patients with onset of seizures at up to 6 months of age.

Whereas individuals with BFNE develop normally and pass mutations to the next generation, mutations in KCNQ2 encephalopathy occur de novo. Mutations in BFNS range from missense mutations, over frameshift, splicing, and nonsense mutations to whole-gene deletions. Loss of function has been proven for several missense mutations, and whole-gene deletions show that even a 50 % reduction of the amount of protein leads to the benign phenotype [17]. All mutations reported in KCNQ2 encephalopathy so far are novel missense mutations, and a dominant negative effect has been shown for most of the mutations studied [18, 19].

Retigabine, a recently marketed enhancer of Kv7 opening, has been shown to reverse some of the functional effects of these severe mutations [18]. Further clinical studies are warranted to investigate whether the early use of this drug in KCNQ2 encephalopathy can positively influence neurodevelopmental outcome.

Neonatal and Infantile Epileptic Encephalopathy and SCN2A

KCNQ2 was not the only gene involved in a benign infantile epilepsy syndrome that has now been implicated in EE. Mutations in SCN2A were described in 2002 in a subset of families with benign neonatal and infantile epilepsy [20]. Individuals with this benign syndrome have a phenotype intermediate between BFNE and benign familial infantile epilepsy (BFIE). The mean age of seizure onset is 11 weeks, but within a single family both neonatal and infantile seizure onset is seen. Rare SCN2A mutations have also been identified in GEFS+ families [21], and subsequently de novo mutations were identified in a few patients with Dravet syndrome [22]. The use of unbiased next-generation sequencing (NGS) techniques has led to the identification of SCN2A mutations in a wide range of EE and to identification of the genetic entity SCN2A encephalopathy [23–26]. De novo SCN2A mutations are reported to be present in 13 % of patients with Ohtahara syndrome, and a minority of patients with West syndrome, Dravet syndrome, and unclassified EE with seizure onset from birth to the age of 1 year. Cognitive outcome ranges from mild to severe intellectual disability, and this does not always seems to correlate with seizure duration and severity. Patients can be microcephalic, and brain MRI shows cerebral atrophy and/or a thin corpus callosum in severely affected patients. Detailed phenotypical descriptions of larger cohorts of patients with SCN2A encephalopathy are warranted to show us what the common phenotypic characteristics are.

Remarkably, whole-exome studies have also revealed de novo SCN2A mutations in patients with intellectual disability or autism without seizures [27–30]. These findings show that the mutation affects neurodevelopment independently from its effect on seizure susceptibility.

Phenotype–genotype correlations remain speculative so far. Both gain of function and loss of function have been described for mutant channels. In general, the mutations identified in EE tend to alter the channel properties to a greater extent than the mutations seen in benign neonatal and infantile epilepsy. Truncating SCN2A mutations have only been described in patients with EE, autism, or intellectual disability, and both whole-gene deletions and whole-gene duplications lead to a more severe phenotype. Also, mutations linked to more severe phenotypes tend to be located more frequently in the linker regions between the transmembrane segments of the protein, or in the voltage sensor domain. Genetically engineered animal models of specific SCN2A mutations will help us to understand better the underlying pathogenetic mechanisms and the role of SCN2A in neurodevelopment and epileptogenesis.

Infantile Convulsions and PRRT2

Mutations in PRRT2, a gene located on chromosome bands 16p11.2 and 16q11.2 and coding for proline-rich transmembrane protein 2 (PRRT2), were initially discovered by whole-exome sequencing in eight Han Chinese families with a history of paroxysmal kinesigenic dyskinesia (PKD) [31]. Around the same time, PRRT2 mutations were reported in 14 of 17 families (82 %) with BFIE, and in five of six families (83 %) with infantile convulsions with choreoathetosis (ICCA) [32]. Linkage analysis studies had previously linked these two conditions to the same pericentromeric region of chromosome 16, also found to be involved in PKD [33, 34].

BFIE and ICCA share the presence of clusters of self-limited focal seizures with onset between 3 and 12 months, but in ICCA, PKD appears in late childhood or adolescence. Within the same family, individual family members can have infantile seizures, PKD, or a combination of both [33].

Several publications have confirmed these findings, and have identified mutations in sporadic epilepsy patients with self-limited infantile seizures. Asymptomatic carriers are frequent owing to the reduced penetrance of PRRT2 mutations [35•]. PRRT2 was also shown to play a role in various other paroxysmal disorders, such as hemiplegic migraine, paroxysmal torticollis, and episodic ataxia [36–38]. Mutations were identified in rare individuals with febrile seizures [37, 39], although their direct role in fever sensitivity has been debated [40]. Recessive PRRT2 mutations have been described in a consanguineous family with intellectual disability and in two siblings of a consanguineous family with intellectual disability, absences, and episodic ataxia in addition to BFIE and PKD [41]. This suggests not only that recessive mutations may considerably worsen the phenotype, but also that PRRT2 dysfunction influences neurodevelopment. Most PRRT2 mutations described so far are nonsense, frameshift, and splice site mutations predicted to lead to protein truncation and thus loss of function. Several missense mutations have been described too. The same recurrent frameshift mutation, c.649-650insC; p.Arg217Profs*7, has been identified in families and patients of different ethnicity [35•]. There is a surprising lack of genotype–phenotype correlation as every single mutation described can be found in individuals with any of the three main phenotypes (PKD, BFIE, ICCA). This is consistent with the phenotypic heterogeneity seen within ICCA families. Most probably, the expression of the phenotype is influenced by other, still to be determined, genetic or environmental factors.

The exact function of PRRT2 remains to be clarified, but it has been shown to interact with synaptosomal-associated protein 25 kDa (SNAP25), a protein involved in neurotransmitter release from synaptic vesicles [42]. Interestingly, the expression of PRRT2 in the brain seems particularly important in the cerebral cortex and basal ganglia, two regions likely to be specifically involved in the main semiological features of BFIE and ICCA [32].

Malignant Migrating Partial Seizures in Infancy and KCNT1, SCN1A, PLCB1, SCL25A22, and TBC1D24

MMPSI is a rare but severe EE characterized by migrating polymorphic focal seizures and progressive neurological deterioration with onset in the first 6 months of life. The seizures are pharmacoresistant and arise from different areas in both hemispheres. The “migrating” pattern of seizures on ictal EEG is the main feature of this syndrome. Seizures gradually become more frequent in the first year of life, often becoming almost continuous for several days. Brain MRI is normal at the onset but signal abnormalities may appear during the course of the disease [43]. A genetic cause has long been suspected, and several groups have now identified five different genes implicated in this devastating syndrome.

KCNT1

A recent study identified de novo missense mutations in the gene KCNT1 in half of sporadic epilepsy patients with MMPSI [44••]. At the very same time, another group identified missense mutations in the same gene in three families and one sporadic epilepsy patient with ADNFLE [45••]. Compared with ADNFLE families with mutations in nicotinic acetylcholine receptor subunit genes such as CHNRA4, patients with KCNT1 mutations have an earlier age of seizure onset and frequently have comorbidities of intellectual disability and psychiatric features. More recently, one patient with an infantile EE with leukoencephalopathy and a de novo missense mutation and one patient with Ohtahara syndrome and a homozygous missense mutation in KCNT1 have been reported [46, 47]. The latter suggests that both dominant and recessive inheritance patterns can be seen in disorders linked to this gene.

KCNT1 encodes a sodium-activated potassium channel that contributes to the slow hyperpolarization that follows repetitive firing. Missense mutations in MMPSI have been proven to lead to a gain of function of the channel [44••]. The large differences in phenotypic expression of mutations might depend on their influence on non-ion-conductance pathways of the protein. Indeed, the C-terminus interacts with developmental proteins such as fragile X mental retardation protein (a messenger RNA binding protein), and mutations altering this developmental signaling pathway might lead to more severe neurodevelopmental disorders.

SCN1A

SCN1A mutations are associated with a wide spectrum of epilepsies, ranging from the benign febrile seizures in families with GEFS+, to intractable epilepsy and intellectual disability in Dravet syndrome [48]. De novo mutations in SCN1A are found in 70-80 % of patients with Dravet syndrome. Two MMPSI patients with a de novo mutation in SCN1A have now been described, making MMPSI the most severe end of the spectrum of epilepsies associated with sodium channel gene defects [7].

PLCB1, SLC25A22, and TBC1D24

Rare recessive mutations in three different genes were identified in isolated families with MMPSI. One affected sib pair carried a recessive mutation in the mitochondrial glutamate carrier SLC25A22 [49], and a second sib pair had a recessive mutation in the gene TBC1D24 [50]. A sporadic epilepsy patient with consanguineous parents carried a homozygous deletion disrupting PLCB1, encoding phospholipase C β1 [51]. Follow-up screening of these genes in additional patients with MMPSI did not reveal any further patients, showing that these are rare causes of MMPSI.

Again, the phenotypic spectrum of mutations in all these genes seems broader than pure MMPSI. Recessive mutations in SLC25A22 were also identified in patients with neonatal EE and a suppression burst pattern [52], and another homozygous deletion in PLCB1 was identified in a patient with infantile EE with infantile spasms [53]. The gene TBC1D24 has the most puzzling phenotypic variability. Recessive mutations were identified in a family with benign familial infantile myoclonic epilepsy [54], in a consanguineous family with focal epilepsy, intellectual disability, and cortical thickening [55], and in a family with severe myoclonic EE with dystonia and progressive neurodegeneration [56]. Finally, mutations seem to be an important cause of the rare DOORS syndrome, consisting of deafness, onychodystrophy, osteodystrophy, intellectual disability, and seizures [57]. The seizure types in individuals with TBC1D24 mutations included generalized tonic–clonic, focal dyscognitive, or clonic seizures, and infantile spasms. Seizures and deafness were not consistent features.

TBC1D24 is a gene that binds GTPases such as ADP-ribosylation factor 6, a protein regulating dendritic branching, spine formation, and axonal elongation. The reason why some mutations cause DOORS syndrome and other mutations cause only epilepsy might depend on the way the mutations affect the interaction with different protein partners.

In summary, when dealing with patients with MMPSI, one should first screen for KCNT1, followed by the other reported genes. In consanguineous families, or in families with multiple affected sibs where recessive inheritance is suspected, screening for the three recessive genes should be prioritized.

Myoclonic Epilepsies (Including Myoclonic–Atonic Epilepsy and Dravet Syndrome) and CHD2

In the past year, CHD2 was identified as a novel gene for EE using two different NGS techniques, whole-exome sequencing and a targeted gene panel. Ten patients with de novo CHD2 mutations have been described to date [58•, 59••, 60••]. All patients had epilepsy onset between 6 months and 3 years of life. Seizure types were mostly generalized with a prominent myoclonic component, but also included tonic–clonic, absence, and atonic seizures. Several patients shared features with Dravet syndrome or myoclonic atonic epilepsy, and had been diagnosed as such. Cognitive outcome ranged from mild to severe intellectual disability, and some patients showed developmental delay prior to epilepsy onset. MRI was normal and EEG showed generalized (poly)spike wave complexes and/or multifocal discharges.

Both missense and truncating mutations have been described, and a loss-of-function mechanism has been proposed. A knockdown zebra fish model showed spontaneous seizures. CHD2 encodes chromodomain-helicase-DNA-binding protein 2. Chromodomain-helicase-DNA-binding proteins are assumed to modify gene transcription by affecting chromatin structure through helicase function. It is not known how dysfunction of a gene with such a crucial function produces a purely neurological phenotype. Further studies will determine the role of chromatin remodelers in the etiopathogenesis of EE, and will provide new avenues for epilepsy treatment.

Idiopathic Focal Epilepsy with Rolandic Spikes and GRIN2A

Idiopathic focal epilepsy with rolandic spikes is considered a spectrum that includes benign epilepsy with centrotemporal spikes (BECTS), Landau–Kleffner syndrome (LKS), and epilepsy with continuous spikes and waves during sleep (CSWS) [61]. BECTS, the most frequent of all epilepsies in childhood, manifests itself at school age by rare focal and mostly nocturnal motor seizures in otherwise normal children and spikes in the centrotemporal region on EEG. This frequent epilepsy syndrome remits by puberty, and treatment is most often not necessary. LKS and epilepsy with CSWS are more severe conditions, in which epileptic EEG abnormalities increase during sleep so that spike wave complexes occupy more than 85 % of non-REM sleep, and cognitive sequelae are the rule.

The genetics of these epilepsies is controversial. Family studies suggest that BECTS has a complex inheritance [62], and that centrotemporal spikes, the EEG hallmark of this disorder, are a marker of one of the BECTS genes with an autosomal dominant mode of inheritance with high but incomplete inheritance [61, 63]. For long, even less was known about the genetics of epilepsy with CSWS and LKS, until the recent simultaneous reports of three studies on this topic.

After the identification of microdeletions including the gene GRIN2A in patients with a peculiar phenotype associating developmental delay, dysmorphic features, and rolandic seizures [64], GRIN2A was sequenced in a series of patients affected by idiopathic focal epilepsies of various subtypes [10••]. Heterozygous mutations were identified in 27 of 359 individuals (7.5 %), and, more specifically, in 12 of 245 patients (4.9 %) with BECTS and in nine of 51 patients (17.6 %) with epilepsy with CSWS. Heterozygous GRIN2A deletions were found in 1 % of the cohort. In two parallel studies, GRIN2A mutations were found in 20 % of familial and sporadic cases of LKS, epilepsy with CSWS, or atypical rolandic epilepsy with verbal dysphasia or dyspraxia [12••], and in 9 % of individuals with an EE in the epilepsy–aphasia spectrum [11••]. A considerable interfamilial and intrafamilial phenotypic variability and incomplete penetrance with frequent asymptomatic carriers was seen in these studies. Although it seems that some mutations might act as risk factors rather than causal mutations, these studies clearly emphasize the important role of GRIN2A in the spectrum of rolandic epilepsy.

GRIN2A encodes the α2 subunit of the NMDA glutamatergic receptor, and functional analysis of some of the missense mutations identified showed an increased activation of the channel [10••, 12••]. These findings may soon open up therapeutic possibilities.

Autosomal Dominant Focal Epilepsies and DEPDC5

Familial focal epilepsy with variable foci (FFEVF) is an autosomal dominant epilepsy in which various members of a same family exhibit seizures originating from different cortical regions. Seizure onset in the temporal or frontal lobe is most often seen. Affected individuals typically develop normally, although family members with intellectual disability, psychiatric disturbances, or autism spectrum disorders have been described [65]. Cerebral MRI studies are unremarkable. In a recent study, exome sequencing identified DEPDC5 mutations in seven of eight large families with FFEVF, thereby establishing the gene as the major causal factor involved in this disorder [66••]. This was confirmed by a simultaneous study that identified DEPDC5 mutations in three of five families with FFEVF [67••].

Both studies also identified mutations in DEPDC5 in 12-27 % of smaller families with familial focal epilepsies, including small families with only frontal or temporal lobe epilepsy. Asymptomatic carriers are frequently present in families, and one de novo mutation was identified in a sporadic epilepsy patient with temporal lobe epilepsy. Future studies will determine the frequency of DEPDC5 mutations in sporadic epilepsy patients with nonlesional focal epilepsies, an important patient population of the adult epilepsy clinic.

Both truncating and missense mutations are seen, and loss of function of DEPDC5 is thus the likely cause of the genesis of seizures. The exact function of the DEPDC5 protein, however, remains to be elucidated. An interaction with the mammalian target of rapamycin pathway, which is involved in cell-growth regulation, is suspected [68].

Conclusion

NGS technologies have tremendously increased the speed of gene discovery in monogenic epilepsies, enabling us to identify a genetic cause in an increasing proportion of patients. Some of the most important recent findings have been summarized in this article. Whereas ion channel genes were long considered to be the main players in genetic epilepsy, a growing number of non-ion-channel genes are now being implicated, especially in epilepsies associated with neurodevelopmental problems. It is exactly in the group of EE that most of the genetic discoveries were made in recent years, and this has had a great impact on our understanding of these disorders. The term “epileptic encephalopathy” (EE) reflects the idea that the epileptic activity itself contributes to the cognitive and behavioral impairment. In its revised version of the terminology in epilepsies, the International League Against Epilepsy acknowledged the fact that the term should be viewed as a concept and as a description of what is observed clinically, but that the actual source of the encephalopathy is usually unknown. It may be the product of the underlying cause, the result of an epileptic process, or a combination of both [69]. Genetic knowledge has now shown us that probably very few patients have a “true” EE according to the literal definition of the term. De novo mutations in genes that are widely accepted to be a cause of EE, such as STXBP1, SCN1A, CHD2, and SCN2A, have now also been identified in individuals with intellectual disability or autism without seizures [28, 70, 71]. This clearly shows that mutations in these genes directly influence neurodevelopment, independently of seizure activity. This is also shown by the fact that some patients with mutations in EE genes show developmental delay prior to seizure onset, or that outcome does not necessarily correlate well with seizure severity. This does not mean that seizure activity does not play a role, and early control of seizures should remain the main goal. Nevertheless, future research should focus on the development of novel treatment strategies targeting the underlying cause rather than the symptom.

Lastly, NGS techniques increasingly confront us with the extent of genetic and phenotypic heterogeneity in epilepsies. Mutations in one specific gene can give rise to a broad spectrum of phenotypes, and conversely, several different genes are implicated in one specific epilepsy syndrome. Although specific clinical features might still direct the clinician in his or her choice of genetic tests, it can be difficult to directly identify the gene responsible in a specific individual. Genetic discoveries, therefore, support the evolution in clinical practice towards diagnostic screening using large targeted gene panels and eventually exome sequencing.