Ion Channels in Genetic Epilepsy: From Genes and Mechanisms to Disease-Targeted Therapies

a. Clinical syndrome and molecular findings

Since first discovered in 2000, several hundred SCN1A mutations have been described in epilepsy, making it the most common known epilepsy gene (Oliva et al., 2012). Missense mutations are associated with genetic (previously generalized) epilepsy with febrile seizures plus (GEFS⁺). This is a familial syndrome in which affected individuals have a variety of epilepsy phenotypes, including febrile seizures, febrile seizures plus, and febrile seizures associated with absence, myoclonic, and even focal seizures. Seizures typically continue into later childhood or adolescence, where they usually resolve (Scheffer and Berkovic, 1997; Singh et al., 1999).

At the more severe end of the GEFS⁺ phenotypic spectrum, patients harboring SCN1A mutations suffer from Dravet syndrome (Claes et al., 2001; Harkin et al., 2007; Scheffer et al., 2009; Escayg and Goldin, 2010). This devastating disease begins with prolonged seizures with fever at about age 6 months, followed by multiple seizure types, developmental regression, and gait abnormalities (Scheffer, 2012; McTague et al., 2016). Dravet syndrome usually occurs as a sporadic disorder with de novo mutations. Approximately half the mutations are missense mutations, and half predict protein truncation with deletions of whole exons, multiple exons, or, more rarely, the whole gene (Claes et al., 2001; Ohmori et al., 2002, 2006; Fujiwara et al., 2003; Nabbout et al., 2003; Kearney et al., 2006; Madia et al., 2006; Mulley et al., 2006; Suls et al., 2006; Marini et al., 2011). Although several biophysical changes that suggest increased channel function have been reported for GEFS⁺ SCN1A missense mutations (Spampanato et al., 2001, 2004a,b; Cossette et al., 2003; Kahlig et al., 2006), the prevailing view is that loss-of-function is the likely basis of disease in most cases. Indeed, this is self-evident in cases with haploinsufficiency due to truncation variants.

b. Mechanisms and potential targeted therapies

The Na_V1.1 knockout mouse models the loss-of-function mutations suggested by molecular studies (Yu et al., 2006). In some genetic backgrounds, heterozygous mice develop spontaneous seizures and sporadic death, reflecting the severity of the disease in humans. Recordings from this mouse model indicated that, although Na⁺ currents were essentially unchanged in hippocampal excitatory pyramidal neurons, the current was substantially reduced in inhibitory GABAergic interneurons. This reduction is responsible for the collapse of action potentials at higher firing frequencies and occurs only in inhibitory neurons (Yu et al., 2006). Similarly, a pronounced action potential attenuation during continuous firing was seen in fast-spiking parvalbumin-positive inhibitory interneurons of heterozygous Na_V1.1 (R1407X) knock-in mice carrying a truncating mutation (Ogiwara et al., 2007). The Na_V1.1 (R1648H) mouse model based on a GEFS⁺ mutation also had impaired GABAergic interneuron function (Martin et al., 2010; Hedrich et al., 2014). Furthermore, using a conditional mouse model, it was possible to show that the selective deletion of Na_V1.1 in GABAergic inhibitory neurons was sufficient to cause a Dravet-like phenotype (Cheah et al., 2012). Even though early studies using human iPSC cells from Dravet patients produced mixed results (Higurashi et al., 2013; Jiao et al., 2013; Liu et al., 2013a), more recent work based on differentiating iPSCs into telencephalic excitatory neurons or medial ganglionic eminence-like inhibitory neurons demonstrates an inhibitory neuron deficit (Liu et al., 2016; Sun et al., 2016). Taken together, these data provide convincing evidence that an inability of GABAergic interneurons to fire robustly results in hyperexcitability, leading to seizures in Dravet syndrome.

The Na_V1.1 knockout-based Dravet mouse models recapitulate many of the pharmacosensitivity profiles seen in patients. A loss of GABAergic inhibitory neuron function may provide a basis for the seizure aggravation properties of Na⁺channel-blocking agents, such as lamotrigine, carbamazepine, and phenytoin, often observed in Dravet syndrome (Guerrini et al., 1998). In this study, the block by these drugs of already compromised Na⁺ channel function in GABAergic interneurons would be expected to further enhance network excitability. Consistent with this, Na⁺ channel blockers were not effective or exacerbated seizures in a Na_V1.1 knockout model of Dravet (Hawkins et al., 2017). The Na_V1.1 knockout-based Dravet mouse models also respond well to stiripentol and clobazam, which are commonly used as first drugs in this disease (Chiron and Dulac, 2011; Cao et al., 2012; Oakley et al., 2013; Hawkins et al., 2017). This provides confidence in the preclinical value of such mouse models.

Understanding the molecular, cellular, and network consequences of harboring SCN1A mutations allows predictive design of targeted therapies. The most obvious strategy would be to selectively enhance Na_V1.1 function. Peptides with these properties exist (Osteen et al., 2016). Small molecules (AA43279) with the desired properties have also been reported (Frederiksen et al., 2017). Another approach that restores Na_V1.1 activity is to take advantage of the recently discovered natural regulatory mechanism based on inhibitory long noncoding RNA. Oligonucleotides designed to block this site enhance Na_V1.1 expression and, at least partially, rescued the seizure phenotype in a mouse model of Dravet (Hsiao et al., 2016). Other strategies could include enhancing GABAergic inhibitory neuron function through secondary mechanisms. For example, the selective activation of K_V3.1 channels that underlie fast-spiking in certain GABAergic inhibitory neurons may help sustain activity and consequently reduce seizure susceptibility.

A number of strategies that are not disease-target based have been tried in mouse models of Dravet. For example, GS967, a drug that preferentially blocks persistent Na⁺ current, is very effective in reducing seizures and death in a Na_V1.1 knockout mouse model (Anderson et al., 2017). These mice could also provide a robust model on which to determine the mechanism of action of newer drugs used in the disease. For example, the amphetamine analog, fenfluramine, has been used with some success in a few Dravet patients, but the mechanistic basis of efficacy remains unknown (Schoonjans et al., 2017). These preclinical rodent models can also be used to inform already completed clinical studies. For instance, cannabidiol has recently been shown to be effective in Dravet syndrome, but its mechanism is unknown; interpretation is confounded by the fact that patients were already on several other antiepileptic drugs (Devinsky et al., 2017). Validated preclinical models provide a mechanism for testing drugs in isolation and/or in combination with other drugs (Hawkins et al., 2017), something that is difficult in humans. In summary, patients with SCN1A-based disease are well placed to significantly benefit from disease-targeted therapeutic strategies.

a. Clinical syndrome and molecular findings

Heterozygous mutations in SCN1B that are located mostly in the immunoglobulin-like extracellular domain of the protein have been described in families with GEFS⁺ (Wallace et al., 1998; Audenaert et al., 2003; Scheffer et al., 2007). Patients harboring SCN1B mutations seem to be at greater risk of developing temporal lobe epilepsy (Scheffer et al., 2007). Also, there are now several reports of homozygous SCN1B mutations causing an early infantile epileptic encephalopathy with features that strongly resemble Dravet syndrome (Patino et al., 2009; Ramadan et al., 2017).

Functional analysis has uniformly ascribed a loss-of-function to mutations in SCN1B associated with epilepsy (reviewed in Reid et al., 2009; O’Malley and Isom, 2015). However, given that the β1-subunit is not pore-forming, the consequent impact on neuronal function is more difficult to interpret. Coexpression with other pore-forming α-subunits suggests that the loss of modulatory action of the β1-subunit can increase excitability through various biophysical mechanisms, including slowing of inactivation, increased availability of channels at hyperpolarized potentials, and reduction in channel rundown during high-frequency activation (Wallace et al., 1998; Meadows et al., 2002; Thomas et al., 2007; Xu et al., 2007). Protein dysfunction is not limited to altered channel kinetics, with a cell adhesion assay indicting a disruption in the ability of mutated β-subunits to mediate protein–protein interactions (Kruger et al., 2016).

b. Mechanisms and potential targeted therapies

The SCN1B (C121W) knock-in mouse model is based on a GEFS⁺ mutation and recapitulates the febrile seizure phenotype seen in patients (Wimmer et al., 2010; Kruger et al., 2016). A temperature-sensitive increase in axon initial segment excitability was observed, consistent with the exclusion of the mutant protein from this neuronal compartment (Wimmer et al., 2010). It is important to note that a more pronounced febrile seizure phenotype is observed in SCN1B (C121W) mice when compared with SCN1B knockout mice. This suggests that the SCN1B (C121W) epilepsy mutation has an impact over and above simple haploinsufficiency (loss of one allele) that may be related to the β1 protein’s adhesion role (Kruger et al., 2016). The multiple biologic roles of β1 make it difficult to devise a therapeutic strategy that would best target heterozygous disease.

Homozygous SCN1B variants are associated with severe epilepsy phenotypes (Patino et al., 2009; Ramadan et al., 2017). This is seen in both the homozygous SCN1B knockout (Chen et al., 2004) and the SCN1B (C121W) mouse models (Reid et al., 2014), with both having a severe phenotype similar to Dravet syndrome. The homozygous SCN1B (C121W) mouse shows a significant reduction in dendritic branching consistent with a defect in β1-mediated cell–cell adhesion (Chen et al., 2004; Davis et al., 2004; Reid et al., 2014). The contraction of the dendritic arbor increases the electrical compactness of the hippocampal neurons, which results in an increase in the voltage depolarization caused by excitatory synaptic input, increasing both the likelihood of action potential firing and consequently network excitability (Reid et al., 2014). Although it is not feasible to correct the morphologic deficit in the SCN1B (C121W) mouse, the addition of the K⁺ channel activator, retigabine, is able to reverse the impact on electrical compactness, providing a potential targeted therapy. In keeping with this, retigabine is very effective at reducing febrile seizure susceptibility in the homozygous SCN1B (C121W) mouse model of Dravet (Reid et al., 2014).

The severe seizure and comorbid phenotypes observed in both Na_V1.1 and homozygous SCN1B (C121W) knock-in mouse models recapitulate several features of Dravet syndrome, suggesting that both are good models of the disease (Yu et al., 2006; Reid et al., 2014). Comparisons between these models of Dravet highlight an important concept in genetic epilepsy. Genetic heterogeneity is a phenomenon in which a single disease phenotype may be caused by mutations in different genes. As discussed above, mutations in SCN1A are the most common genetic cause of Dravet syndrome in which an inhibitory neuron deficit is the well-established disease mechanism. In contrast, no inhibitory neuron deficit is seen in the homozygous SCN1B (C121W) mouse model, indicating that two very different cellular mechanisms can underlie the same epilepsy syndrome. This has clear implications for targeted therapy, with different strategies possibly required for the same syndrome depending on the causative gene.

a. Clinical syndrome and molecular findings

For a long time, SCN2A mutations were thought to only associate with benign familial neonatal–infantile seizures, a mild, self-limited epilepsy syndrome of the first year of life (Heron et al., 2002; Berkovic et al., 2004; Herlenius et al., 2007). More recently, SCN2A has emerged as one of the most prominent epilepsy genes associated with a wide spectrum of seizure and neurodevelopmental phenotypes. These include early-onset epileptic encephalopathies with reported epilepsy syndromes, including Ohtahara syndrome, epilepsy of infancy with migrating focal seizures, infantile spasms (West syndrome), Lennox-Gastaut syndrome, Dravet-like syndrome, as well as unclassified epileptic encephalopathies (Howell et al., 2015; Wolff et al., 2017). Interestingly, SCN2A has also emerged as one of the most prominent genes associated with neurodevelopmental disorders, including autism, intellectual disability, and schizophrenia, with most patients carrying truncation mutations not having recognized seizures (Jiang et al., 2013; Iossifov et al., 2014; Sanders et al., 2015; Carroll et al., 2016; Wang et al., 2016; Ben-Shalom et al., 2017). The genotype-phenotype relationship in epilepsy is more complex with both loss-of-function and gain-of-function changes observed. A recent retrospective study that reverse-phenotyped epilepsy patients with SCN2A mutations began to dissect this conundrum. A phenotypic classification based on seizure onset, cognitive outcome, and movement disorders revealed that patients carrying gain-of-function mutations show an earlier onset of the disease, more severe seizure phenotype, and better responsiveness to Na⁺ channel blockers than the carriers of loss-of-function mutations (Ben-Shalom et al., 2017; Wolff et al., 2017). It is possible that the biophysical measures indicating both loss-of-function and gain-of-function for similar phenotypes are not targeting the critical cellular mechanism of Na_V1.2 in these disorders. Clarification of this has significant implications for diagnosis, treatment strategies, and mechanistic understanding in diseases associated with SCN2A mutations.

b. Mechanisms and potential targeted therapies

The classification of SCN2A-related epilepsy into either a gain-of-function or a loss-of-function grouping already has significant implications for targeted therapy. As suggested above, one simple prediction of this finding is that antiepileptic drugs that act as Na⁺ channel blockers should have a differential impact on the two populations. Consistent with this, Wolff et al. (2017) show that phenytoin is reasonably effective in ameliorating disease caused by gain-of-function mutations, but less effective, or possibly even exacerbates seizures in patients with loss-of-function mutations (Wolff et al., 2017). These findings strongly argue that biophysical characterization should be part of a complete evaluation of subjects with SCN2A mutations. With time, a comprehensive database will develop, meaning that many mutations will be ascribed a functional class. This will significantly aid diagnosis and help devise the best treatment strategy for a given patient.

The finding that both loss and gain of Na_V1.2 function can cause disease adds complexity to the development of targeted therapy (Ben-Shalom et al., 2017; Wolff et al., 2017). For gain-of-function SCN2A mutations, it is reasonable to assume that molecular strategies that reduce protein expression or protein function may be effective. However, in the context of loss-of-function SCN2A mutations, these would be expected to increase seizure susceptibility, as well as exacerbating neurodevelopmental symptoms. A converse argument for strategies based on loss-of-function mutations can be made. This suggests that any approach that directly targets the specific SCN2A deficit is likely to have narrow therapeutic windows and will rely heavily on accurate functional diagnosis.

Mouse models, in which emerging pathologies can be discovered, may provide alternative targets based on disease mechanism. The Na_V1.2 (Q54) transgenic mouse expresses a gain-of-function channel mutation that results in a progressive epilepsy phenotype beginning with spontaneous focal motor seizures that evolve to include bilateral convulsive seizures with age (Kearney et al., 2001). Recordings from hippocampal pyramidal neurons exhibit an enhanced persistent Na⁺ current (Kearney et al., 2001). The Na_V1.2 (A263V) knock-in mouse line, which is homozygous for a mild disease gain-of-function mutation, presents with frequent seizures and increased mortality (Schattling et al., 2016). At the cellular level, current clamp recordings in slices reveal increased excitability in hippocampal pyramidal neurons (Schattling et al., 2016). These mouse lines provide preclinical models of gain-of-function disease. Comparing the genetically different models may also reveal points of mechanistic convergence that could be targeted. For example, if the increase in persistent Na⁺ current was evident in both, this would provide a clear target for which strategies already exist (see SCN8A).

The Na_V1.2 knockout mouse model is of interest in light of evidence implicating loss-of-function SCN2A mutations in both epilepsy and other neurodevelopmental disease. Initial observations report that heterozygous Na_V1.2 knockout mice appear normal (Planells-Cases et al., 2000). This suggests that overt seizures are not part of the heterozygous phenotype, although a full behavioral analysis that includes probing neurodevelopmental deficits is clearly needed. Homozygous Na_V1.2 knockout mice die within 1–2 days of birth. It will be important to develop new knock-in mouse models of loss-of-function disease based on human mutations, which may have impacts over and above simple haploinsufficiency.

a. Clinical syndrome and molecular findings

Heterozygous mutations in SCN8A are associated with an epileptic encephalopathy characterized by developmental delay, seizure onset within the first 18 months of life, and intractable epilepsy. These patients have multiple seizure types, including infantile spasms, generalized tonic-clonic seizures, absences, and focal seizures (Veeramah et al., 2012; Allen et al., 2013; Carvill et al., 2013a; Vaher et al., 2014; Blanchard et al., 2015; Larsen et al., 2015; Takahashi et al., 2015; Wagnon et al., 2015a; Boerma et al., 2016). Mirroring the situation with SCN1A and SCN2A, where there are both mild and severe phenotypes, very recently there were two reports of familial benign infantile seizures, without cognitive impairment, and some with paroxysmal dyskinesias with missense variants in SCN8A (Anand et al., 2016; Gardella et al., 2016).

The majority of mutations in SCN8A associated with epileptic enecephalopathy are missense mutations. Functional analysis has predominantly identified changes predicting gain-of-function that are likely to increase neuronal excitability. These include incomplete channel inactivation, depolarizing shift in the voltage dependence of steady-state fast inactivation, and hyperpolarizing shifts in the voltage dependence of activation (Veeramah et al., 2012; de Kovel et al., 2014; Estacion et al., 2014; Blanchard et al., 2015; Wagnon et al., 2015a; Barker et al., 2016). One common finding for SCN8A variants tested is an increase in persistent Na⁺ current (Estacion et al., 2014), a biophysical property that has been implicated in several epileptic states (Stafstrom, 2007).

b. Mechanisms and potential targeted therapies

The heterozygous Na_V1.6 (N1768D/+) knock-in mouse model recapitulates the seizures, ataxia, and sudden death seen in patients with the mutation, providing a robust preclinical model (Wagnon et al., 2015b; Lopez-Santiago et al., 2017; Sprissler et al., 2017). Consistent with functional analysis in heterologous assays, recordings from excitatory and inhibitory neurons reveal an increase in persistent Na⁺ current. CA1 excitatory neurons also have unusual depolarizing events that are proposed to be due to a change in the function of the Na⁺/Ca²⁺ exchanger caused by excessive internal Na⁺ concentrations (Lopez-Santiago et al., 2017).

Although the human genetic evidence is not strong, there is good evidence from animal models that loss of Na_v1.6 function can also result in nonconvulsive seizures (Papale et al., 2009). Elegant recent work has demonstrated reduced function of inhibitory neurons in the thalamus as the likely basis of increased susceptibility to spike-and-wave discharge seizures (Makinson et al., 2017). Specifically, a reduction in the ability of thalamic reticular nucleus GABAergic neurons to sustain tonic firing results in the disruption of intrathalamic reticular nucleus inhibition, leading to an increase in oscillatory behavior in the thalamus. Interestingly, a pyramidal neuron-specific loss of Na_v1.6 function is protective against convulsive seizures (Makinson et al., 2017). Reduced Na_V1.6 function was also able to ameliorate seizure severity in a Scn1a mouse model of Dravet syndrome, suggesting that it could act as a modifier of disease (Martin et al., 2007). This study highlights the importance of understanding the cell-specific impact of epilepsy mutations, allowing insight into not only pathogenic mechanisms, but also the potential adverse impact of any therapeutic intervention.

Mechanistic insights provide a number of potential targeted therapeutic options. In particular, targeting the increased persistent Na⁺ current makes sense. Significantly, GS967 (PRAX-330), a specific blocker of persistent Na⁺ current, extends survival of the Na_V1.6 (N1768D/+) mouse (Anderson et al., 2014). A further possibility is the use of the persistent Na⁺ current blocker, riluzole, an approved drug that is used to treat amyotrophic lateral sclerosis. Riluzole is effective in blocking the early depolarization events seen in CA1 pyramidal neurons (Lopez-Santiago et al., 2017), but as yet there is no published evidence of its efficacy in the Na_V1.6 (N1768D/+) mouse. Patel et al. (2016) have also demonstrated that increased persistent current of Na_v1.6 mutant channels can be preferentially reversed with cannabidiol. Another strategy could be to block the Na⁺/Ca²⁺ exchanger that is implicated in the disease mechanism (Lopez-Santiago et al., 2017). Approved drugs, including amiodarone (Watanabe and Kimura, 2000), bepridil (Watanabe and Kimura, 2001), aprindine (Watanabe et al., 2002), and cibenzoline (Yamakawa et al., 2012), have been found to have inhibitory actions on the Na⁺/Ca²⁺ exchanger, potentially providing repurposing opportunities. Finally, given the gain-of-function observed with the majority of SCN8A epilepsy mutations, developing molecular knockdown strategies may be beneficial. It is worth noting that these knockdown strategies could potentially result in an increase in susceptibility to nonconvulsive spike-and-wave seizures (Makinson et al., 2017).

a. Clinical syndrome and molecular findings

The majority of human KCNA1 mutations are associated with episodic ataxia type 1 (Browne et al., 1994, 1995; Zuberi et al., 1999; Eunson et al., 2000; Rajakulendran et al., 2007). Patients with episodic ataxia type 1 are ten times more likely to have seizures than the general population, strongly implicating mutations in this gene as a cause of epilepsy (Rajakulendran et al., 2007). Another common feature of KCN1A disease is myokymia, characterized by involuntary twitching of facial and limb muscles (Browne et al., 1994, 1995). Furthermore, this gene has been suggested as one of the contributing genes to sudden unexpected death in epilepsy in a 3-year-old proband with pharmacoresistant epileptic encephalopathy (Klassen et al., 2014). The majority of functional studies of KCNA1 mutations revealed loss-of-function through a variety of mechanisms, including changes to the voltage threshold and gating properties of K_V1.1-containing K⁺ channels, reduction in peak K⁺ current amplitude, and channel trafficking defects (Rajakulendran et al., 2007; D’Adamo et al., 2015).

b. Mechanisms and potential targeted therapies

K_v1.1 knockout mice display similar phenotypes to human patients, including frequent spontaneous seizures and in some cases sudden unexpected death, suggesting that it is a good preclinical model of disease (Smart et al., 1998; Lopantsev et al., 2003; Wenzel et al., 2007; Glasscock et al., 2010). The heterozygous K_v1.1 (V408A) knock-in mouse model based on a human mutation linked to episodic ataxia type 1 has stress–fear responses and induced motor dysfunctions that recapitulate some disease phenotypes, but no spontaneous seizures (Herson et al., 2003).

The cellular basis of morbidity is not clear but may involve changes in neurotransmitter release. It is well established that blocking K_V1 channels broadens the action potential, thus increasing the flux of Ca²⁺ into the presynaptic terminal and as a consequence increasing triggered neurotransmitter release (Shu et al., 2006; Foust et al., 2011). Consistent with this, the overexpression of a human K_V1.1 epilepsy mutation in cultured rat hippocampal neurons enhances neurotransmitter release probability (Heeroma et al., 2009). K_V1.1 is localized in the axon initial segment of fast-spiking neocortical GABAergic interneurons and has a critical role in regulating the excitability of these cells (Goldberg et al., 2008). Functional studies have reported an increase in GABAergic inhibitory synaptic transmission in both the K_v1.1 knockout and K_v1.1 (V408A) knock-in mouse models (van Brederode et al., 2001; Herson et al., 2003). At a neuronal network level, multielectrode array analysis of hippocampal slices from K_v1.1 knockout mice indicates an increase in spontaneous spike-wave and high-frequency ripples that are associated with hyperexcitability (Simeone et al., 2013). This is proposed to be due to increased synaptic release within the CA3 region, decreasing precision of CA3 principal cell spike timing, which in turn leads to the altered hippocampal network oscillatory behavior (Simeone et al., 2013).

There are no reported effective targeted therapies based on our understanding of cellular mechanisms. However, the preclinical knockout mouse model has provided some insights. Interestingly, homozygous K_v1.1 knockout mice have disrupted sleep, and seizures peak during times of light (Wright et al., 2016). Almorexant, a dual orexin receptor antagonist used in sleeping disorders, improves sleep and reduces seizure severity, suggesting that it may be useful for patients with mutations in KCNA1 (Roundtree et al., 2016). Treatment of homozygous K_v1.1 knockout mice with a ketogenic diet also reduces seizure frequency (Fenoglio-Simeone et al., 2009) and has been shown to successfully extend life span (Simeone et al., 2016). Finally, homozygous K_v1.1 knockout mice with partial genetic ablation of Na_V1.2 exhibit reduced duration of spontaneous seizures, and a significant improvement of survival rates argues that blockers of this Na⁺ channel may have some treatment value (Mishra et al., 2017).

a. Clinical syndrome and molecular findings

De novo mutations in KCNA2 have been identified in cases of early infantile epileptic encephalopathy (Pena and Coimbra, 2015; Syrbe et al., 2015; Allen et al., 2016; Hundallah et al., 2016). Seizure onset is between 5 and 17 months, with a phenotypic spectrum including febrile and afebrile, hemiclonic, myoclonic, myoclonic-atonic, absence, focal dyscognitive, focal, and generalized seizures; mild to moderate intellectual disability; delayed speech development; and severe ataxia (Pena and Coimbra, 2015; Syrbe et al., 2015). The KCNA2 spectrum also encompasses milder familial epilepsy (Corbett et al., 2016). De novo KCNA2 mutations in patients with neurodegenerative hereditary spastic paraplegia and ataxia have also been reported (Helbig et al., 2016; Manole et al., 2017). In heterologous expression systems, de novo mutations in KCNA2 can cause both loss- and gain-of-function, all with a dominant effect (Syrbe et al., 2015).

b. Mechanisms and potential targeted therapies

The idea that loss-of-function can cause disease is consistent with findings from animal models. Homozygous K_V1.2 knockout mice display severe seizures and die early (Brew et al., 2007). Heterozygous K_V1.2 knockout mice do not display spontaneous seizures but are more sensitive to a proconvulsant challenge compared with wild-type littermates (Brew et al., 2007). The Pingu mutant mouse that carries a loss-of-function mutation in K_v1.2, induced through N-ethyl-N-nitrosourea mutagenesis, provides another potential model of disease (Xie et al., 2010). This mouse carries a mutation in the S6 segment, which is in close proximity to the loss-of-function mutation found in patients with epileptic encephalopathy, including ataxia (Syrbe et al., 2015). Both heterozygous and homozygous mutant mice display significant gait abnormalities (Xie et al., 2010). No mouse models with a gain-of-function KCNA2 variant have been reported to date.

The cellular basis of hyperexcitability due to either loss- or gain-of-function KCNA2 mutations is not clear. This is in part because K_v1.2 is found in both excitatory and inhibitory neurons (Wang et al., 1994; Lorincz and Nusser, 2008). It is also due to the fact that K_v1.2 channels potentially play dual roles in defining cellular excitability. By rapidly repolarizing neurons, they can help sustain rapid firing rates. In addition, they play a critical role in defining membrane potential. Cellular studies in mouse models highlight this complexity by showing contrasting changes in firing patterns. Recordings from inhibitory cerebellar basket cells of the Pingu mutant mouse show an increased firing rate (Xie et al., 2010). In contrast, firing is reduced in K_V1.2 knockout glycinergic neurons (Brew et al., 2007).

Despite the limited understanding of the cellular basis of disease, therapeutic strategies based on molecular findings are possible. The 4-aminopyridine is an approved K⁺ blocker already being trialed in patients carrying gain-of-function mutations (H. Lerche, personal communication). The use of molecular knockdown strategies may also prove to be useful in this patient cohort. There are currently no obvious mechanism-based therapeutic strategies for loss-of-function disease. However, Xie et al. (2010) have tested a nontargeted approach and report that the carbonic anhydrase inhibitor, acetazolamide, is capable of rescuing the motor incoordination in Pingu mice. This led to a successful clinical outcome in a single patient with ataxia and myoclonic epilepsy caused by a KCNA2 mutation, which highlights the potential utility of the preclinical mouse model (Pena and Coimbra, 2015).

a. Clinical syndrome and molecular findings

Several de novo mutations in KCNB1 have been identified in cases of infantile epileptic encephalopathy (Torkamani et al., 2014; Saitsu et al., 2015; Thiffault et al., 2015; Allen et al., 2016). The phenotypic spectrum includes cognitive and motor dysfunction, severe infantile generalized seizures with high-amplitude spike-and-wave discharges, and prominent stimulus and photosensitive epilepsy (Saitsu et al., 2015; Allen et al., 2016). A range of functional impacts has been reported for epilepsy-associated KCNB1 mutations. For example, mutations located within the pore domain of K_v2.1 were associated with a loss of K⁺ selectivity. This could lead to an increased inward conductance of cations, resulting in neuronal depolarization (Torkamani et al., 2014; Thiffault et al., 2015). Other biophysical findings include changes in gating, with mutant channels showing less voltage dependence and an ability to be constitutively open (Torkamani et al., 2014). Expression of a de novo missense mutation in heterologous cells revealed significant changes in protein expression and localization (Thiffault et al., 2015).

b. Mechanisms and potential targeted therapies

There are currently no good rodent models for KCNB1 disease. A homozygous K_V2.1 knockout mouse model exhibits no spontaneous seizures, but does exhibit a reduced threshold for induced seizures, in addition to reduced spatial learning and hyperactivity (Speca et al., 2014). This mimics loss-of-function aspects of disease and argues that it can cause some level of hyperexcitability. However, given the biophysical changes seen, including altered ion selectivity and changes in voltage sensitivity, it is not entirely surprising that simple loss-of-function does not recapitulate more severe disease. Saitsu et al. (2015) have investigated the impact of two different human epilepsy mutations on neuronal excitability in an overexpression culture model. They report that expression of both mutant channels slows action potential firing in excitatory pyramidal neurons by reducing repolarization. How this translates to increased network excitability in epilepsy remains unclear. Knock-in mutant mice based on human KCNB1 mutations are more likely to serve as better preclinical models. Given the lack of functional insight, it is difficult to propose potential targeted therapy in KCNB1 disease.

a. Clinical syndrome and molecular findings

A recurrent de novo mutation in KCNC1 has been identified as a major cause of progressive myoclonus epilepsy (Muona et al., 2015; Oliver et al., 2017). This is a distinctive epilepsy syndrome characterized by myoclonus, generalized tonic-clonic seizures, and progressive neurologic deterioration. An interesting clinical trait is the transient clinical improvement that occurs with fever in some patients (Oliver et al., 2017). In vitro analysis has revealed two key biophysical changes resulting from the recurrent KCNC1 mutation: a dominant-negative impact that significantly reduces currents, and a concomitant hyperpolarizing shift in the activation voltage (Muona et al., 2015; Oliver et al., 2017). A nonsense mutation in KCNC1 has also been described in one family with intellectual disability, but no epilepsy (Poirier et al., 2017).

b. Mechanisms and potential targeted therapies

The cellular mechanism underlying KCNC1-mediated disease is not understood. One possibility is that the reduced current levels lead to an inability of fast-spiking GABAergic interneurons to sustain firing, with a consequent increase in network excitability caused by disinhibition (Muona et al., 2015; Oliver et al., 2017). However, the hyperpolarizing shift in activation would act to increase channel function, tempering this reduced current. Interestingly, in vitro analysis demonstrates that the hyperpolarizing shift is enhanced at febrile temperatures, and this may explain the clinical improvement seen with fever (Oliver et al., 2017). The homozygous K_V3.1 knockout mouse displays only a mild phenotype that includes impaired coordinated motor skills, but no spontaneous seizures, suggesting that this mouse is not a good preclinical model of the disease (Ho et al., 1997). RE1, a K_V3.1 channel opener, provides a potential therapeutic option. Importantly, RE1 partially restores impaired firing in fast-spiking neurons from a mouse model of 15q13.3 microdeletion syndrome (Thelin et al., 2017). However, the utility of this and other K_V3.1 channel agonists is yet to be tested in KCNC1 disease models.

a. Clinical syndrome and molecular findings

A mutation in KCNMA1 was first reported in a large family with autosomal dominant generalized epilepsy and paroxysmal nonkinesigenic dyskinesia (Du et al., 2005). Two de novo mutations and an autosomal recessive homozygous frameshift duplication have also been described (Zhang et al., 2015b). Functional analysis provided in the first study showed that the mutant BK channel had a greater macroscopic current, likely due to an increase in Ca²⁺ sensitivity. In contrast, the homozygous frameshift duplication is likely to result in loss-of-function.

b. Mechanisms and potential targeted therapies

The activation of BK channels can both increase and decrease firing activity in neurons depending on context. Du et al. (2005) propose that enhancement of BK channels leads to increased excitability by allowing rapid repolarizing action potentials, resulting in increased firing rates. A reduction in BK activity has been implicated in causing neuronal depolarization at rest that consequently increases excitability by taking the neuron closer to the firing potential (Verma-Ahuja et al., 1995). Therefore, predicting how a given mutation impacts excitability is difficult. The KCNMA1 knockout mouse exhibits intention tremor and abnormal gait consistent with cerebellar ataxia but is seizure-free and is therefore not a good model of epilepsy (Sausbier et al., 2004). There are a range of small organic molecules and peptides that can both activate and block BK channels (Bentzen et al., 2014; Yu et al., 2016). These provide a rich source of pharmacological tools that could be used to probe efficacy once a good model of disease is developed.

a. Clinical syndrome and molecular findings

Mutations in both KCNQ2 and KCNQ3 cause an autosomal-dominant self-limited epilepsy known as benign familial neonatal epilepsy (Biervert et al., 1998; Charlier et al., 1998; Schroeder et al., 1998; Singh et al., 1998; Lerche et al., 1999; Maljevic and Lerche, 2014). More recently, screening of patients with neonatal onset epileptic encephalopathy revealed many cases of de novo mutations in KCNQ2 (Weckhuysen et al., 2012, 2013; Maljevic and Lerche, 2014), with fewer mutations reported in KCNQ3 (Miceli et al., 2015a,b). Approximately half of KCNQ2 mutations are predicted to truncate the protein, with the other half being missense mutations (Maljevic and Lerche, 2014). Functional analysis revealed a loss-of-function caused by a different molecular mechanism (Maljevic and Lerche, 2014). It should be noted that disease-related gain-of-function mutations have been more rarely reported for both KCNQ2 and KCNQ3 (Miceli et al., 2015b; Millichap et al., 2016). A genotype–phenotype correlation has been suggested, with mutations in severe epileptic encephalopathies more likely to have a dominant-negative impact (Orhan et al., 2014).

b. Mechanisms and potential targeted therapies

K_V7.2 and K_V7.3 colocalize at the axon initial segment and/or nodes of Ranvier (Devaux et al., 2004; Schwarz et al., 2006), where they play an integral role in defining excitability, including reducing spiking frequency during trains of activity (spike-frequency adaption) (Schwarz et al., 2006). A reduction in the M-current will reduce adaptation and is the likely basis of neuronal hyperexcitability in loss-of-function KCNQ2 and KCNQ3 disease. Support for this idea comes from a conditional transgenic mouse that has spontaneous seizures and carries a dominant-negative pore mutation (Peters et al., 2005). CA1 pyramidal neurons have a smaller medium after-hyperpolarization current, and as a consequence a reduced ability to adapt action potential firing (Peters et al., 2005). CA1 pyramidal neurons from a spontaneous K_V7.2 mutant mouse also have a reduced ability to adapt action potential firing (Otto et al., 2006). The mechanism of disease for gain-of-function mutations is less clear. Both K_V7.2 and K_V7.3 proteins are found in parvalbumin-positive hippocampal interneurons (Lawrence et al., 2006; Nieto-Gonzalez and Jensen, 2013; Grigorov et al., 2014) with a reduction in spike adaption in these cells possibly increasing network excitability. The K_V7.2 (A306T) and K_V7.3(G311V) knock-in mouse models are both based on loss-of-function mutations found in benign familial neonatal convulsions (Singh et al., 2008; Otto et al., 2009). In both cases, homozygous mice have spontaneous seizures, whereas heterozygous mice have increased seizure susceptibility, suggesting that they may act as reasonable models of mild disease (Singh et al., 2008; Otto et al., 2009). No mouse models with epileptic encephalopathy have been reported to date.

One striking feature of benign familial neonatal seizures is that, within a few weeks or months of birth, symptoms spontaneously remit in the vast majority of patients. The developmental pattern of K_V7.2 and K_V7.3 protein expression seems unlikely to explain remission, as the expression of both subunits increases during maturation (Weber et al., 2006; Maljevic and Lerche, 2014). A plausible hypothesis is that remission coincides with a switch of the inhibitory neurotransmitter GABA, from a predominantly excitatory transmitter in early development to a more traditional inhibitory role in later development (Rivera et al., 1999).

From a therapeutic perspective, retigabine, a selective opener of the M-current (Rundfeldt and Netzer, 2000; Gunthorpe et al., 2012), is an obvious drug choice for loss-of-function KCNQ2- and KCNQ3-mediated disease. More broadly, retigabine has been used in the treatment of epileptic encephalopathy patients, with studies revealing significant improvement in some patients (Weckhuysen et al., 2013; Millichap et al., 2016). Unfortunately, the side effects seen in some patients have led to the removal from the market of this drug (Tompson et al., 2016). Ongoing efforts to develop similar but safer M-current activator are required, and these would be predicted to be particularly efficacious in epilepsy caused by most mutations in KCNQ2 and KCNQ3.

a. Clinical syndrome and molecular findings

Mutations in KCNT1 have been identified in familial and sporadic cases with autosomal dominant nocturnal frontal lobe epilepsy (ADNFLE). It is also the cause of about 40% of cases with a quite different syndrome, epilepsy of infancy with migrating focal seizures (Wilmshurst et al., 2000; Barcia et al., 2012; Heron et al., 2012; Lim et al., 2016). Other epileptic encephalopathies, such as Ohtahara syndrome, have also been linked to this gene (Møller et al., 2015). All mutations in KCNT1 described to date are missense mutations (Lim et al., 2016). Notably, they all result in significantly increased K⁺ current, which argues strongly that this is the basis of disease. The extent of the gain-of-function may correlate with seizure phenotype severity (Milligan et al., 2014). However, the fact that two mutations have been described causing both nocturnal frontal lobe epilepsy and epilepsy of infancy with migrating focal seizures suggests a more complex genotype–phenotype correlation (Lim et al., 2016). KCNT1-mediated epilepsy has become an example of how quickly targeted therapy might enter the clinic through the repurposing of already approved drugs (see below).

b. Mechanisms and potential targeted therapies

The exact mechanisms underlying disease caused by mutation in KCNT1 are still unclear. Rapid elevations of intracellular Na⁺ concentrations in neurons are sensed by Na⁺-activated K⁺ channels influencing firing patterns. For example, rapid repolarization by K⁺ channels can sustain rapid action potential firing, with an increase in function potentially enabling excitatory neurons to fire at aberrantly high rates (Bhattacharjee and Kaczmarek, 2005). Alternatively, during sustained activity, slow activation of such channels can lead to the cessation of firing that could cause inhibitory dysfunction similar to that observed in SCN1A disease (Bhattacharjee and Kaczmarek, 2005). Given that the expression pattern of K_Na1.1 is largely unknown, it is difficult to make a prediction as to the likely mechanism. To date, no mouse model based on gain-of-function KCNT1 mutations has been reported.

Despite the lack of progress on the underlying cellular basis of disease, the robust gain-of-function seen in all KCNT1 disease provides a clear target. It has been known for some time that the approved antiarrhythmic drug, quinidine, is able to block rodent K_Na1.1 channels (Santi et al., 2006; Yang et al., 2006). Milligan et al. (2014) showed that quinidine was able to also block human K_Na1.1 channels, including those expressing epilepsy mutations. This has motivated clinical studies in which quinidine was repurposed and trialed in patients carrying KCNT1 mutations. Several case reports suggested efficacy, particularly in very young patients (Bearden et al., 2014; Mikati et al., 2015; Fukuoka et al., 2017), whereas others have shown limited or no efficacy (Mikati et al., 2015; Chong et al., 2016; Fukuoka et al., 2017). Quinidine is not a safe drug, with severe life-threatening cardiac liability, and a small clinical trial in adults in ADNFLE failed to show efficacy, and doses were limited by cardiac toxicity (Mullen et al., 2017). The antiarrhythmic drug, clofilium, which also blocks K_Na1.1 channels, potentially provides another repurposing opportunity, although it too has significant cardiac side effects (Castle, 1991). These drugs do form good lead compounds for the development of more specific and better tolerated drugs targeted against K_Na1.1. The consistent gain-of-function seen for all disease mutations in KCNT1 argues that molecular strategies in which the channel is knocked down are likely to be effective. Importantly, the homozygous KCNT1 knockout mouse has only minor phenotypes, suggesting that there may be a large therapeutic window for such strategies (Bausch et al., 2015).

a. Clinical syndrome and molecular findings

Three members of a family with progressive myoclonic epilepsy were reported to harbor a homozygous mutation in KCTD7 (Van Bogaert et al., 2007). Progressive myoclonic epilepsy is a disorder characterized by myoclonic and tonic-clonic seizures, ataxia, and cognitive regression (see KCNC1). Numerous other cases of progressive myoclonic epilepsy caused by mutations in KCTD7 have since been reported (Blumkin et al., 2012; Kousi et al., 2012; Krabichler et al., 2012; Staropoli et al., 2012; Farhan et al., 2014).

b. Mechanisms and potential targeted therapies

As KCTD7 channels have no clearly defined role, it is difficult to come up with a unifying mechanistic basis of disease. The KCTD7 channel is widely expressed throughout the brain, with immunohistochemical analysis identifying expression in cortical neurons, in granular and pyramidal cell layers of the hippocampus, and in cerebellar Purkinje cells (Kousi et al., 2012). KCTD7 mutations have been recently reported to reduce the ability of KCTD7 channels to sustain a K⁺ conductance either through a trafficking or biophysical deficit (Moen et al., 2016). This suggests the reduction in function is likely to result in a more depolarized membrane potential of excitatory neurons, leading to hyperexcitability. K⁺ channel openers, such as retigabine, may be expected to be beneficial based on this mechanism. However, the impact of mutations on the SAT2 neuronal glutamine transporter or other cellular roles as a basis of excitability cannot be ruled out. No good animal models of KCTD7-mediated disease have been reported to date.

a. Clinical syndrome and molecular findings

Early evidence supported the idea that loss-of-function CACNA1A mutations were associated with GGE (Jouvenceau et al., 2001; Imbrici et al., 2004). Microdeletions that encompass CACNA1A and a single truncating mutation have been associated with severe epileptic encephalopathies that include infantile spasms and West syndrome (Auvin et al., 2009; Damaj et al., 2015; Hino-Fukuyo et al., 2015). More recently, de novo missense mutations have been convincingly shown to cause severe epileptic encephalopathies with seizure types that typically include focal, tonic, and tonic-clonic seizures; severe intellectual disability; and motor impairment (Epi4K Consortium, 2016; Reinson et al., 2016). Although the biophysical impacts of all mutations are yet to be reported, the presence of deletions and truncating mutations argues that a loss-of-channel function is the primary underlying molecular deficit.

b. Mechanisms and potential targeted therapies

Homozygous Cacna1a knockout mice have progressive neurologic deficits that result in ataxia and dystonia, with mice dying within 4 weeks of birth (Jun et al., 1999; Fletcher et al., 2001). Similar phenotypes are seen in a range of spontaneous Cacna1a gene mutant mice with dystonia, ataxia, premature death, and epilepsy all observed in leaner, tottering, rolling, and rocker mice (Pietrobon, 2005). These potentially provide good preclinical mouse models of CACNA1A disease. Most reports investigating cellular mechanism in these mouse models implicate a reduction in excitatory synapse transmitter release, consistent with the integral role of these channels in this process. For example, reduced evoked postsynaptic currents have been measured at the Calyx of Held in the homozygous Cacna1a knockout mouse (Caddick et al., 1999). Similarly, a deficit has been reported at the cerebellar parallel fiber–Purkinje cell synapses and excitatory synapses within the ventrobasal nuclei in the leaner mouse model (Caddick et al., 1999). Recently, analysis of a conditional Cacna1a knockout mouse, in which deletion is limited to layer VI, reports that reduced excitatory neurotransmission at cortico-thalamic synapses is sufficient to generate generalized seizures (Bomben et al., 2016). How this change resulted in increased network excitability is not known. This lack of a robust cellular mechanism makes it difficult to develop disease-targeted strategies. However, mouse models may be predictive of pharmacosensitivity. For example, high-power low-frequency oscillations, which are thought to be a marker of cortical excitability in the tottering mouse, were markedly decreased by acetazolamide and 4-aminopyridine, drugs that are treatment options for episodic ataxia 2 (Cramer et al., 2015). Of note is the finding that spontaneous seizures in the Cacna1a knockout mouse can be abolished by knocking out Cacna1g (Song et al., 2004). This suggests that T-type Ca²⁺ channel blockers, including ethosuximide, may be good therapeutic options (Zamponi, 2016).

a. Clinical syndrome and molecular findings

Both heterozygous and homozygous GRIN1 mutations have been implicated in a spectrum of disease, including epileptic encephalopathies and intellectual disability. De novo heterozygous mutations seem to associate with profound developmental delay, with a lack of speech as a dominant feature. Associated phenotypes include seizures, hyperkinetic movements, involuntary movement, muscular hypotonia, oculogyric crises, and cortical blindness (Allen et al., 2013; Ohba et al., 2015; Lemke et al., 2016). Heterozygous GRIN1 mutations have also been identified in autosomal-dominant mental retardation (Hamdan et al., 2011; Redin et al., 2014; Zhu et al., 2015; Zehavi et al., 2017). Functional analysis of pathogenic heterozygous de novo GRIN1 mutations suggests they cause a loss-of-function (Lemke et al., 2016). Different mutations in GRIN1 have been shown to exhibit dominant-negative interactions of varying degrees, but with no clear genotype–phenotype correlation (Lemke et al., 2016). Patients with homozygous GRIN1 truncation mutations experience continuous seizure activity and early death (Lemke et al., 2016).

b. Mechanisms and potential targeted therapies

Numerous studies use mouse lines that allow the deletion of Grin1 in a subset of neurons potentially modeling loss-of-function disease. Specific forebrain Grin1 knockout results in impaired synaptic plasticity and memory storage (Nakazawa et al., 2002; Cui et al., 2004; Hasan et al., 2013). Seizures are not observed in any of these rodent models. This said, the impaired synaptic plasticity and memory deficits recapitulate the intellectual disability and motor disorders caused by mutations in GRIN1. The homozygous Grin1 knockout mice die as neonates, but have no seizures, suggesting that they also only partially recapitulate the severe clinical phenotype seen in humans harboring homozygous GRIN1 truncation mutations (Forrest et al., 1994). Therefore, although genetic mouse models exist, they have not faithfully recapitulated key features of GRIN1 disease.

Molecular studies report that most mutations in GRIN1 result in a loss-of-function. It is not immediately clear how this would lead to increased excitability, and it is possible that emerging pathologies are to blame. Correcting the loss-of-function deficit may benefit patients harboring loss-of-function mutations. However, activators of NMDA receptors are likely to increase neuronal activity, arguing that such drugs would have a narrow therapeutic window. The lack of good preclinical models and, as a consequence, a lack of understanding of the mechanisms behind the GRIN1 disease are currently hampering progress in developing targeted therapies.

a. Clinical syndrome and molecular findings

GRIN2A mutations have been associated with epilepsy and speech disorders (also known as epilepsy–aphasia spectrum disorders) that present as a broad clinical spectrum. At the less severe end of the spectrum, seizure phenotypes include typical or atypical rolandic epilepsy (also known as childhood epilepsy with centrotemporal spikes) (Lemke et al., 2013; Lesca et al., 2013). More severe phenotypes include epileptic encephalopathies such as Landau-Kleffner syndrome and epilepsy with continuous spikes and waves during slow-wave sleep (Endele et al., 2010; Carvill et al., 2013b; Lemke et al., 2013; Lesca et al., 2013; Turner et al., 2015; Gao et al., 2017). Speech impairment is seen in all individuals with GRIN2A-related disorders and includes acquired aphasia, auditory agnosia, dysarthria, speech dyspraxia, impaired intelligibility and language delay, and regression (Turner et al., 2015; Myers and Scheffer, 2016). Intellectual disability occurs in up to 70% of individuals (Lemke et al., 2013; Lesca et al., 2013).

GRIN2A mutations include missense, truncating, splice, and copy number variants and can result in both gain- and loss-of-function. To date, no clear genotype–phenotype correlation has emerged (Strehlow et al., 2015). Causes of gain-of-function include reduction of Zn²⁺-mediated inhibition (Lemke et al., 2013; Serraz et al., 2016), increased glutamate binding (Pierson et al., 2014), altered duration of channel open and closed states (Lesca et al., 2013), and reduced Mg²⁺ block (Endele et al., 2010). Causes of loss-of-function for missense mutations include reduced agonist potency, decreased total protein levels, and reduced cell surface expression (Addis et al., 2017). Mutations predicted to cause truncation and deletion mutations further support the idea that loss-of-function can result in disease (Reutlinger et al., 2010; Carvill et al., 2013b; Lesca et al., 2013).

b. Mechanisms and potential targeted therapies

How both loss-of-function and gain-of-function mutations in GRIN2A cause disease is not clear. Gain-of-function mutations in NMDA receptor activation will increase hyperexcitability in brain networks. Blockers of NMDA receptors are therefore a potential effective targeted therapy. Pierson et al. (2014) have shown that the approved use-dependent NMDA receptor antagonist, memantine, can reduce currents back to normal levels for a gain-of-function mutation. Promisingly, memantine has been shown to reduce seizure activity in a patient with this mutation, although cognitive function showed no improvement. The mechanism underlying pathology in loss-of-function mutations is less intuitive and likely to be part of an emerging pathology. In this case, a targeted therapy would enhance GluN2A-containing receptor function. Recently, a positive allosteric modulator of GluN2A-containing receptors (compound 275) has been reported to rescue loss-of-function missense mutations providing a preclinical pharmacologic tool, although this is yet to be tested in rodent models (Yu et al., 2015; Addis et al., 2017). With strategies for dealing with both gain- and loss-of-function available, epilepsy due to mutations in GRIN2A should benefit from targeted therapy led by genetic diagnostics and functional screening.

a. Clinical syndrome and molecular findings

Mutations in GRIN2B have been associated with early infantile epileptic encephalopathies, including West syndrome with severe developmental delay, Lennox-Gastaut syndrome, and early-onset seizures with intellectual disability (Allen et al., 2013; Lemke et al., 2014; Zhang et al., 2015a; Smigiel et al., 2016). The mutated gene has also been associated with milder seizure syndromes, including early-onset focal epilepsy (Lemke et al., 2014). Intellectual disability associated with GRIN2B mutations can occur independently of epilepsy (Endele et al., 2010; Hamdan et al., 2014).

Initial functional assessment showed a decrease of Mg²⁺ block and higher Ca²⁺ permeability that correlated with the clinical phenotype severity (Lemke et al., 2014). These effects were seen for mutations located in the extracellular glutamate-binding domain region, as well as those variants affecting the re-entrant pore-forming loop implicated in Mg²⁺ block. Early-onset epileptic encephalopathy has also been associated with a de novo GRIN2B splice-site mutation, providing evidence for GRIN2B loss-of-function as a cause of epileptic encephalopathies. Thus, both gain and loss of NMDA receptor function can cause similar disease phenotypes (Smigiel et al., 2016).

b. Mechanisms and potential targeted therapies

It is yet to be understood how both gain- and loss-of-GluN2B function leads to a similar disease phenotype. The use-dependent NMDA receptor blocker, memantine, would be predicted to be useful in gain-of-function disease. However, despite some efficacy in vitro, memantine had limited impact in patients with gain-of-function GRIN2B mutations (Platzer et al., 2017). For loss-of-function GRIN2B mutations, a strategy based on selective antagonism may be beneficial, but has not been tested. Again, a lack of good preclinical models is hampering progress with heterozygous GluN2B knockout mice not displaying a seizure phenotype. Homozygous Grin2b knockout mice die shortly after birth, but this is thought to be due to a defect in the suckling response (Kutsuwada et al., 1996).

a. Clinical syndrome and molecular findings

A recurrent de novo mutation in GRIN2D has been identified in patients with early infantile epileptic encephalopathy, with phenotypes that include severe developmental delay, intellectual disability, and movement disorders (Li et al., 2016). The missense mutation found in GRIN2D occurs in the M3 transmembrane domain and results in a gain-of-function through slower deactivation, increased channel open probability, and increased glutamate and glycine potency (Li et al., 2016).

b. Mechanisms and potential targeted therapies

Transfection of cortical neurons with mutant GluN2D subunits leads to cell death. This is consistent with NMDA receptor overactivation and is blocked by the NMDA receptor blocker, memantine (Li et al., 2016). A clinical study using oral memantine decreased seizures in patients in which conventional antiepileptic drugs were unsuccessful (Li et al., 2016). One patient, who developed status epilepticus, improved with treatment of ketamine and magnesium, both NMDA receptor blockers (Li et al., 2016).

GluN2D knockout mice partially recapitulate some of the clinical features of human disease, with a reduction in spontaneous locomotor activity similar to that seen in patients (Ikeda et al., 1995). However, no seizures have been reported in heterozygous or homozygous GluN2D knockout mice (Ikeda et al., 1995). There is a clear need for a knock-in mouse model to provide further insights into the disease mechanism. Given the importance of this subunit in early development, the engineering of conditional mouse models will be important in dissecting out developmental versus acute impact of the mutation in defining seizure susceptibility (Perszyk et al., 2016).

a. Clinical syndrome and molecular findings

GABRA1 mutations have been reported to associate with GGE (Cossette et al., 2002; Maljevic et al., 2006; Lachance-Touchette et al., 2011; Johannesen et al., 2016). Epilepsy syndromes include JME, CAE, and febrile seizures (Cossette et al., 2002; Maljevic et al., 2006; Lachance-Touchette et al., 2011; Johannesen et al., 2016). More recent evidence implicates GABRA1 mutations in early infantile epileptic encephalopathies, including phenotypes of Dravet, Ohtahara, or West syndromes (Carvill et al., 2014; Johannesen et al., 2016; Kodera et al., 2016). Both GGE and the severe encephalopathies share common clinical features, including myoclonic seizures, tonic-clonic seizures, and photosensitivity (Cossette et al., 2002; Lachance-Touchette et al., 2011; Johannesen et al., 2016). Functional analysis has unanimously shown loss-of-function, including reduced surface expression of the assembled GABA_A receptor (Lachance-Touchette et al., 2011). This can be caused by increased endoplasmic degradation of the misfolded protein, as noted for the GABRA1 (A322D) variant (Gallagher et al., 2004, 2005). A reduction in the sensitivity to GABA has also been reported (Krampfl et al., 2005; Carvill et al., 2014). Interestingly, there is a correlation between GABRA1 mutations causing severe epilepsy and a significant loss of the GABA sensitivity of the receptor (Johannesen et al., 2016).

b. Mechanisms and potential targeted therapies

Heterozygous Gabra1 knockout mice have spontaneous electrographic spike-wave discharges and behavioral absence-like seizures (Arain et al., 2012), recapitulating some of the milder GGE phenotypes. The GABRA1 (A322D) knock-in mouse model based on a JME mutation (Cossette et al., 2002) also develops a spike-and-wave electrographic phenotype (Arain et al., 2015). A direct comparison with the Gabra1 knockout indicated that both equally expressed the spike-and-wave discharge phenotype, suggesting that simple haploinsufficiency was the basis of disease (Arain et al., 2015). Both mutant mice develop myoclonic seizures later in life, which may model JME (Arain et al., 2015). These mice lines therefore are good preclinical models of GGE, providing an opportunity to devise and test new therapeutic strategies. New models that recapitulate more severe disease are still required.

Given that GABRA1 mutations impair GABAergic inhibition, it is not surprising that they increase neuronal excitability and consequently result in seizure phenotypes. Drugs that specifically increase GABA_A function are therefore likely to be highly effective in patients carrying GABRA1 mutations. Consistent with this, vigabatrin and sodium valproate, both of which have been shown to maintain high GABA concentrations in the synaptic cleft (Connelly, 1993; Johannessen, 2000), successfully control seizures of patients with GABRA1 mutations (Kodera et al., 2016). Unfortunately, barbiturates and benzodiazepines that specifically target phasic GABA_A inhibition have significant side effects (Kwan et al., 2001).

a. Clinical syndrome and molecular findings

The Epi4K Consortium and Epilepsy Phenome/Genome Project were the first to provide strong evidence of GABRB3 mutations in early infantile epileptic encephalopathy (Allen et al., 2013). Several more recent studies have confirmed this original finding (Allen et al., 2013; Hamdan et al., 2014; Zhang et al., 2015a; Papandreou et al., 2016; Møller et al., 2017). The affected patients display a broad spectrum of seizure types, which includes absence, tonic, myoclonic, and generalized tonic-clonic seizures. Some of them also show delayed development, and mild to severe intellectual disabilities (Allen et al., 2013; Papandreou et al., 2016; Møller et al., 2017). Epilepsy syndromes reported include Lennox-Gastaut, infantile spasms, West, and Dravet syndromes (Allen et al., 2013; Møller et al., 2017). Milder syndromes caused by mutations in GABRB3 include GEFS⁺, early-onset absence epilepsy, epilepsy with myoclonic-atonic seizures, and multifocal epilepsy (Møller et al., 2017). Mutations in GABRB3, typically whole-gene deletions, have also been associated with Angelman syndrome (Tanaka et al., 2012).

Functional analysis uniformly suggests loss-of-function of GABRB3 variants through a variety of biophysical and trafficking deficits (Janve et al., 2016; Møller et al., 2017). As seen for GABRA1 mutations, some GABRB3 mutations causing more severe epilepsy show a significant loss of GABA sensitivity, suggesting a potential genotype–phenotype correlation (Møller et al., 2017). Similarly, mutations associated with Lennox-Gastaut syndrome have reduced GABA-evoked current amplitudes, whereas infantile seizure variants show changes in current kinetics (Janve et al., 2016).

b. Mechanisms and potential targeted therapies

Given the role of GABRB3 in neuronal migration, synaptogenesis, and overall brain development, mutations in GABRB3 are likely to be linked to abnormal development of neuronal networks (Noebels et al., 2012; Tanaka et al., 2012). Therefore, targeting therapies in early development may be required to rescue seizure and neurodevelopmental symptoms in these patients. Homozygous Gabrb3 knockout mice show increased mortality and seizures and impaired social behavior, cognition, and motor coordination, whereas heterozygous mice have increased epileptiform EEG activity and seizure susceptibility (Homanics et al., 1997; DeLorey et al., 1998, 2008). These features recapitulate some of the phenotypes seen in GABRB3 disease, including Angelman syndrome.

a. Clinical syndrome and molecular findings

Mutations in GABRG2 have been associated with GEFS⁺, Dravet syndrome, CAE, familial febrile seizures, and epileptic encephalopathy (Baulac et al., 2001; Wallace et al., 2001; Harkin et al., 2002; Kananura et al., 2002; Audenaert et al., 2006; Shen et al., 2016; Hernandez et al., 2017). In vitro studies suggest a range of mutation-mediated functional deficits, including changes in benzodiazepine sensitivity, receptor kinetics, assembly, trafficking, and cell surface expression (Baulac et al., 2001; Bianchi et al., 2002; Bowser et al., 2002; Sancar and Czajkowski, 2004; Hales et al., 2005; Krampfl et al., 2005; Kang et al., 2006; Eugène et al., 2007; Kang and Macdonald, 2016; Hernandez et al., 2017). In general, these functional analyses suggest that epilepsy-causing mutations result in a reduction in GABA_A receptor-mediated inhibition, albeit through varying mechanisms. A consequent reduction in both phasic and/or tonic inhibition is proposed to underlie increases in neuronal hyperexcitability, resulting in seizures.

b. Mechanisms and potential targeted therapies

The GABRG2 (R43Q) knock-in mouse model was one of the first syndrome-specific models developed for genetic epilepsy, and faithfully recapitulates both the absence epilepsy and febrile seizures seen in the family (Tan et al., 2007). The GABRG2 (R43Q) mutation mouse model also recapitulates some of the genetic complexity seen in patients. Namely, the GABRG2 (R43Q) mice only display an absence phenotype in a spike-and-wave prone mouse strain (DBA), whereas the febrile seizure phenotype is robust in all mouse strains tested. This observation matches the penetrance seen in humans, which is low for absence seizures, whereas febrile seizures segregate as a highly penetrant autosomal dominant trait (Wallace et al., 2001). This implies that genetic principles that occur in human populations can be recapitulated in mouse models, providing the tools to interrogate complex heritability. The GABRG2 (R43Q) mouse model also has an expected pharmacosensitivity profile responding to first-line anti-absence drugs, ethosuximide and sodium valproate (Tan et al., 2007; Kim et al., 2015). The GABRG2 (Q390X) knock-in mouse model recapitulates the more severe seizure phenotype seen in patients with truncation mutations (Warner et al., 2016). These models provide an excellent opportunity to identify cellular mechanisms, and act as good preclinical models of disease.

A comparison between GABRG2 knock-in and knockout mice reveals that the molecular mechanisms for different seizure phenotypes may result from distinct molecular deficits. A spike-and-wave discharge phenotype in both the GABRG2 (R43Q) and heterozygous Gabrg2 knockout mice suggested that haploinsufficiency was sufficient to cause absence seizures (Reid et al., 2013; Warner et al., 2016). In contrast, only the GABRG2 (R43Q) model displayed a thermogenic seizure phenotype, suggesting a mechanism over and above haploinsufficiency was responsible for febrile seizures. In the GABRG2 (Q390X) mouse, which models a more severe epilepsy phenotype, a reduction in the expression of the GABA_A γ2-subunit is greater than that seen in the Gabrg2 knockout, suggesting a dominant-negative effect of the mutant on the wild-type subunits (Warner et al., 2016). Whether disease caused by these distinct molecular mechanisms will require different targeted therapeutic strategies is yet to be determined.

The conditional GABRG2 (R43Q) mouse model provides an opportunity to explore the impact of the mutation in early development (Chiu et al., 2008). Mice harboring the GABRG2 (R43Q) mutation from birth were reported to be more susceptible to seizures (Chiu et al., 2008). This argues that mutation-mediated dysfunction can trigger a cascade of events (e.g., morphologic and/or transcriptional changes) that define long-term network stability. Therefore, targeted treatment early in development may be necessary to fully rescue patients harboring epilepsy-causing mutations.

In summary, mouse models of GABRG2-associated epilepsy have provided key insights into disease mechanisms. They suggest that phenotype severity may depend on the extent of dominant-negative impact of mutations and that different therapeutic strategies may be required for different seizure phenotypes. These models also highlight the potential need to treat early in development. Clearly, activators of the GABA_A receptors, such as benzodiazepines, are likely to benefit these patients, with the early use of this class potentially overcoming the developmental consequences of harboring GABRG2 mutations (Haas et al., 2013).