Voltage-gated Sodium Channels in Epilepsy


Address correspondence and reprint requests to Priv.-Doz. Dr. R. Köhling at Institut für Physiologie, Westfälische Wilhelms-Universität Münster, Robert-Koch-Str. 27a, 48149 Münster, Germany. E-mail: kohling@uni-muenster.de


Summary: Animal experiments, and particularly functional investigations on human chronically epileptic tissue as well as genetic studies in epilepsy patients and their families strongly suggest that some forms of epilepsy may share a pathogenetic mechanism: an alteration of voltage-gated sodium channels. This review summarizes recent data on changes of sodium channel expression, molecular structure and function associated with epilepsy, as well as on the interaction of new and established antiepileptic drugs with sodium currents. Although it remains to be determined precisely how and to what extent altered sodium-channel functions play a role in different epilepsy syndromes, future promising therapy approaches may include drugs modulating sodium currents, and particularly substances changing their inactivation characteristics.

Epilepsy is a common disease affecting ∼0.5–1% of the world population. The underlying etiology is multifold. It ranges from symptomatic seizures due to tumor, infection, and trauma, to cryptogenic forms without apparent cause. Increasing evidence suggests that—even though epilepsy is no single syndrome—common pathogenetic processes either as basic mechanisms or as final common factors may be responsible as etiologic causes. As neuronal excitability, and thus epileptogenicity, is critically governed by the interaction of voltage- and ligand-gated ion channels (1), it is not surprising that ion channel alterations may be pathogenetic causes, and indeed, at least for some syndromes, epilepsies are increasingly thought to be “channelopathies” either due to genetic mutations or as the end point of hitherto unknown pathologic processes (2–4). The clues come from different sources:

(a) animal experiments based on acute or chronic epilepsy models demonstrate that the density and distribution, molecular structure, and function of ion channels is altered after seizures or epileptiform activity; (b) genetic models of epilepsy in animals have been linked to ion-channel mutations; (c) ionic currents in human chronically epileptic tissue were found to have properties different from those in animal tissue, and furthermore, the degree of this difference was related to the degree of epilepsy-related neuropathologic changes. Most important, some specific epilepsy syndromes have recently been shown to be associated with distinct ion-channel mutations; (d) last, several antiepileptic drugs (AEDs) exert strong effects on ionic currents, and in many cases, this effect is thought to be their main mechanism of action. One channel for which all these clues point to a crucial involvement in epileptogenesis is the voltage-gated sodium channel. The aim of this article is to summarize recent developments regarding molecular and functional changes of sodium currents with epilepsy both in animal models and humans, to review specific genetic mutations associated with seizures, and to give an overview of the effects of different AEDs on sodium currents.


Molecular and functional subtypes of sodium currents

Sodium channels are voltage-operated transmembraneous proteins that become selectively permeable to sodium ions when the membrane depolarizes and that are generally closed at resting states.

From the molecular point of view, different types of sodium channels can be distinguished regarding their main, α subunits, several of which have been cloned. One such a unit is thought to constitute a fully functional channel (5,6)(Fig. 1), which, however, can underlie modulation by other subunits (see later). In the rat, human, and mouse CNS, the α subunits are encoded by at least five different genes. Of these, SCN1A, SCN2A, and SCN3A (corresponding to rat sodium channels type I, II, III in older nomenclature) are closely related and give rise to sodium currents carried by channel proteins Nav1.1 through 1.3. Additionally, the SCN8A gene (Nav 1.6, formerly called NaCh6), and a tetrodotoxin (TTX)-insensitive SCN5A (Nav 1.5) have been described in the CNS, the latter of which is particularly interesting, as it is localized primarily in the limbic system, which has a relatively low seizure threshold. Furthermore, a glia-specific channel (SCN6A or NaG) is present in the CNS (7–13). Of these channels, Nav1.2 and Nav 1.6 sodium channels compose most of the channels in the neocortex and forebrain, whereas Nav 1.1 is expressed at much lower levels. At least in the rat, Nav 1.3 channels are expressed predominantly only during early ontogenetic stages (14,15); in the human brain, Nav 1.3 also is apparently encountered during adulthood in hippocampus, frontal lobe, and cerebellum (16). The channel α subunits are often associated with auxiliary β subunits in brain (β1, β2, and β3; Fig. 1), as well as in heart and skeletal muscle (β1 and β2) (17–19). These subunits, although not essential for channel function, can accelerate channel activation and inactivation (or decelerate it in the case of the β3 subunit, which is found mainly in juvenile tissue), shift steady-state inactivation in a negative direction, and (at least β2) appear to facilitate incorporation of channels into the membrane (5,6,17,20,21).

Figure 1.

Putative structure of the α (encoded by SCN1A) and β (encoded by SCN1B) subunits of the voltage-gated sodium channel and mutations associated with generalized epilepsy with febrile seizures plus (GEFS+; solid circles) or severe myoclonic epilepsy of infancy (SMEI; grey squares). The channel consists of four consecutive domains (I–IV) with six transmembrane segments (1–6). Segment 4 acts as the voltage sensor (+). The domains associate to form a central pore. Mutations are shown as point mutations with exchange of amino acids at designated positions (e.g., T875M), insertions (ins), deletions (del), or splice site mutations (splice site). Electrophysiologic characterizations have been carried out on mutations T875M and R1648H, affecting the voltage sensors in domains II and IV, and C121W, affecting the β subunit, respectively (see text). The latter mutation prevents interaction with the α subunit. The only mutation so far found on another α subunit (encoded by SCN2A) consists of an exchange in R187W in domain I.

Much like other voltage-gated channels, apart from molecular differences, sodium currents also fall into classes with different properties. Thus both activation and particularly the type of inactivation (fast vs. slow or noninactivating currents) can differ considerably. These functional differences are probably not related to the different molecular channel subtypes (see section on ion channels); nevertheless, their impact on neuronal excitability is qualitatively different. Most studies addressing sodium currents and epilepsy do not specifically focus on the functional current subtypes but discuss mainly fast inactivating currents. However, the noninactivating sodium current has been given increasing attention in recent years, so that it appears helpful to differentiate between fast and noninactivating functional states of the sodium current, at least for some aspects of this review.

Electrophysiologic properties of INa

The voltage-dependent fast inactivating sodium current (INa) is the principal current responsible for the depolarizing phase of the action potential and thus is the essential current for neuronal excitation in general. Consequently it can be considered to be indispensable also for the generation of epileptiform activity. Indeed, signal transmission in neuronal networks, and therefore synchronization among neurons as the characteristic process of epileptogenesis, are unthinkable without the existence of sodium currents. Computer models of neuronal networks generating epileptiform activity generally include INa(1,22–24). Further, albeit indirect, evidence for the crucial role of voltage-gated sodium channels for epileptogenesis can seen in the observation that several epileptogenic agents appear to exert their effect by enhancing sodium currents. Thus the convulsant drug pentylenetetrazol (PTZ) prolongs sodium-carried action potentials, and prolonged cocaine treatment, which induces seizures in animals, increases peak sodium current conductance (25,26), actions that might, however, also be attributed to interference with other channels. Unlike the noninactivating INa, the fast INa under normal circumstances is unlikely to affect excitability in general. Its properties (in comparison to INaP) are illustrated in Fig. 2. It shows fast, submillisecond activation, and usually fast inactivation with millisecond time constants will allow it only to code short signals as needed for information processing. After the signal, INa automatically falls silent. It substantially (∼80%) recovers only with a time constant of milliseconds. A further “safety factor” prevents INa from becoming fully available too fast: Prolonged inactivation of INa leading to recovery time constants of several seconds can be observed after short trains of, or even single, action potentials (27–29). Such prolonged inactivation (also termed “slow” by many authors) indeed regulates and limits backpropagation of action potentials in dendrites (30). As this phenomenon depends on frequency and duration of the previous depolarizations, this will particularly lead to a dampenening of INa with strong (epileptic) activity. One can speculate that loss of this slow inactivation would increase excitability and thus epileptogenicity, and future studies should investigate changes of slow inactivation with respect to epilepsy. Conversely, AEDs that strengthen or restore prolonged inactivation might be particularly helpful. Indeed, in one study on hamster striatal neurons, lamotrigine (LTG) appeared to prolong recovery from inactivation in this sense (31,32). A peculiar property of sodium currents—in a way opposing slow inactivation—is the resurgent current, which has been described only in cerebellar Purkinje neurons (Fig. 2). It consists of a pronounced inward current after moderate repolarization (e.g., to –30 mV) after strong depolarization (e.g., to 30 mV) (33). If such a current could be observed not only in the cerebellum (which lacks epileptogenicity) but also in seizure-prone areas such as the hippocampus (where it is not present, at least in nonepileptic tissue) (33), it could increase excitability and contribute to seizure generation. Again, future studies should clarify whether (in chronic epilepsy models) such sodium-channel behavior can be seen. Further, a conversion of otherwise fast-inactivation kinetics to slow ones would likewise support epileptogenesis. Several epileptogenic substances seem to act primarily by slowing the otherwise fast-inactivation kinetics of sodium channels and thereby increasing their open probability. Among these are the sodium-channel opener veratridine (34), the insecticides pyrethrin and DDT (35–38), and the venom of the scorpion Centruruides sculpturatus(39). An enhancement of INa, however, is probably not the main pathogenetic mechanism of epileptogenesis in general, because in some models (e.g., with strychnine and PTZ application in molluscan neurons), INa was found to be decreased (40,41).

Figure 2.

Fast inactivating (INa), noninactivating (INaP), and resurgent (INa resurgent) sodium currents in Purkinje neurons of normal and mutant (null mutation of Nav1.6) mice. Currents were elicited by voltage steps, as given in diagram. Current–amplitude scaling is relative to peak amplitude of INa. Note that INa composes ∼98% of total current in normal mouse neurons, and that it decays to ∼10% of its peak amplitude within ∼5 ms. By contrast, 2% of the current does not inactivate, thus remaining as INaP. As a special feature of cerebellar Purkinje neurons, a sizeable current larger than the remaining INaP can be reelicited after repolarization and subsequent strong depolarization on stepping back to –30 mV. In a mutant mouse, where Nav1.6 is not functional, INa is normal, whereas INaP and particularly the resurgent current is reduced. Reprinted from ref. 33, with permission.

Whereas INa is the current responsible for neuronal activity per se, there is compelling evidence that a persistent, noninactivating sodium current (INaP) is particularly instrumental in finely modulating neuronal excitability and thus possibly determining epileptogenicity, as it lacks the “safety factor” of automatic inactivation. The properties of INaP and its role in epileptogenesis are discussed in the following section.

Noninactivating sodium current (INaP)

Electrophysiological properties of INaP

The existence of INaP was first suspected to underlie the appearance of a long-lasting low-threshold depolarization and the inwardly rectifying voltage response on depolarizing current injection in hippocampal pyramidal neurons (42), cerebellar Purkinje cells (43), and neocortical layer V neurons (44). Since then, it has been shown in a variety of neurons of other structures, including the squid axon (45), thalamus (46), striatum (47,48), and entorhinal cortex (49). Noninactivating sodium currents also have been reported in human neocortical neurons (50).

Persistent sodium currents usually make up only a small fraction (1–3%) of total sodium peak current in neurons (Fig. 2)(50,51). They fail to inactivate (52–54) or at least have exceedingly long inactivation time constants in comparison to the fast inactivating INa(55,56). In steady state, the persistent sodium current activates in the region of –60 mV and reaches its peak amplitude between –40 and –35 mV (57) with the steady-state activation curve showing half-maximal activation at around –50 mV (58,59).

Ion channels generating INaP

It is still a matter of debate whether the persistent sodium current originates from a voltage-gated channel different from the channel responsible for the transient sodium current. Indeed, three alternative mechanisms may account for the appearance of INaP.

First, the fact that activation and inactivation of INa overlap in a certain membrane-potential range could result in a “window current” within this voltage overlap, which gives the appearance of a slowly inactivating current (58). However, the described activation kinetics of INaP would not fully be in line with this hypothesis.

Second, INaP might be generated by a specific subtype of sodium channel. In favor of this hypothesis is the observation that INaP apparently has an ∼10 mV lower activation threshold than INa(53,56,59,60), and that INaP can strongly be reduced at least in the cerebellum if a specific channel (e.g., Nav 1.6) is not functional (Fig. 2)(33). Against this theory stand the pharmacologic properties of INaP. Thus sustained sodium currents are generally blocked by the same agents that block INaP[i.e., by TTX and by intracellular application of the membrane-impermeable local anesthetic QX 314 (51,58)], with rare exceptions in which a TTX-insensitive INaP was described (48). Such a TTX resistance, however, may simply be conveyed on the INa by single point mutations to otherwise identical channels (61). Additionally, INaP, like INa, is largely insensitive to inorganic Ca2+ channel blockers such as Cd2+, Co2+, or Mn2+. The more negative activation threshold of INaP as compared with INa can be assumed to be a direct consequence of the lack of inactivation. Thus the steady-state activation curve of INa is shifted to more negative potentials if inactivation is removed by manipulations such as enzymatic treatment with papain (62).

Third, the same channel that carries INa also could be responsible for INaP. Several findings support this hypothesis. In one study on neocortical single sodium channels (63), these channels were found to open not only immediately, but also with a delay. These late openings consisted of clustered channel openings (minibursts) or long-lasting single openings, all with similar unitary amplitudes and slope conductances. In entorhinal cortex neurons, similar persistent activity was found, although with higher conductances than transient one (64). In summary, INaP seems to be carried by the same channel molecules as INa, and to be brought about by a particular gating, the slow-gating mode of these channels. From molecular biologic studies, there also is no evidence for different channel types possessing different inactivation kinetics; although the different CNS sodium channels possess different primary structures, only a few indications of functional differences with regard to gating kinetics were found (15,16,65,66). Rather, electrophysiologic data indicate that the different subtypes of sodium channels do not convey specific gating behaviors but can account for both fast and noninactivating functional states. Thus the human Nav1.3 channel (expressed in kidney cell lines) shows both slow and fast gating modes, albeit with a slightly higher tendency to persisting gating than Nav1.2 (16). Similar results were obtained in the Xenopus oocyte expression system (67), although in this system, sodium-channel kinetics are generally different from those found in mammalian cell lines. Whereas in the latter, fast gating kinetics are usually observed, in oocytes, slow inactivation is seen, which can be converted to fast modes only when low-molecular-weight proteins, later identified as β1 subunits, are coexpressed (68–70). Thus results reported in Xenopus oocytes do not necessarily correspond to those with mammalian cells. These considerations notwithstanding, the results suggest that the conversion of fast-inactivating to slow-inactivating channels, rather than being based on differences in channel structure, may be linked to β subunits, or a G protein–dependent phosphorylation of the channels (52,71). Notable exceptions are sodium channels found in dorsal root ganglion cells and peripheral nervous system (PNS) neurons. These channels (SCN9A, SCN10A, and SCN11A, corresponding to SNS, NaN, and hNe-Na) give rise to persistent sodium currents, are resistant to TTX (SCN9A, SCN10A), or can be enhanced by Cd2+ (SCN11A) (72). However, in the CNS, the distinction between INa and INaP thus appears to be mainly a functional rather than a molecular one.

Neuronal excitability and INaP

As INaP contributes only a small fraction to the total sodium current, it is particularly vulnerable to (pharmacologic) modulation. Further, its properties such as the low activation threshold and, above all, the lack of inactivation predispose it to substantially modify neuronal behavior, even with just small changes and at membrane potentials close to resting levels, at which most other voltage-gated currents are inactive. Slow inactivation kinetics of sodium currents have been shown to be responsible for pathologic states of excitable cells. In muscle cells, for example, point mutations in sodium channels lead to prolonged inactivation, and the resulting INaP may cause excessive depolarizations, causing the symptoms of hyperkalemic periodic paralysis (73,74). Similarly, modeling and experimental studies suggest that INaP, particularly under conditions of increased excitability (due to, e.g., extracellular K+ increases) can critically act in synergy with synaptic or other voltage-gated currents to induce seizures (75,76). Furthermore, INaP is speculated to underlie a number of bioelectric phenomena leading to increased neuronal excitability and that may therefore be of functional significance for epileptogenesis. Among these are depolarizing afterpotentials, burst firing, modulation of synaptic transmission, and subthreshold membrane potential oscillations.

A number of neurons display spike afterdepolarizations. This afterdepolarization, in hippocampal neurons, is blocked by application of TTX in a more sensitive fashion than action-potential blockade and therefore is speculated to be carried by INaP(77,78). Likewise, rebound action potentials in layer III entorhinal cortex neurons after hyperpolarizations appear to be mediated by INaP, as is the sustained tonic firing of these neurons, which occurs even at resting membrane potential (79). Especially the latter is thought to be essential for a high-fidelity transfer of signals to the postsynaptic targets of these cells, which also have been shown to be particularly involved in epileptogenesis in in vitro models (79).

Another neuronal behavior that has been associated with increased excitability is burst firing (80). It is held responsible, at least in part, for the initiation of synchronous network activities such as epilepsy (81). This burst firing has been shown to depend on INaP in both hippocampal CA1 (77,82,83) and neocortical layer V neurons (84). Likewise, bursting in layer II/III neocortical neurons appears to be supported by INaP, although by a TTX-resistant type (85). Isolated hippocampal neurons with excitatory autapses grown in microcultures display epileptiform activity after sustained treatment with a blocker of glutamatergic transmission, kynurenic acid, which is subsequently withdrawn. They were demonstrated to generate spontaneously either interictal-like single epileptiform bursts or paroxysmal depolarization shifts (PDSs) or runs of PDSs in the way of ictal-like activity (86). Under conditions of synaptic blockade in 0 Ca2+/high-Mg2+ medium, those neurons discharging single PDSs fell silent, whereas those displaying barrages of PDSs were tonically depolarized and continued to generate bursts. The tonic depolarization of these neurons was found to be TTX sensitive and probably dependent on INaP. On blockade of this tonic depolarization by TTX, the burst discharges ceased as well. Such intrinsic bursts were thus instrumental in maintaining seizure-like activity, and were shown to depend on the noninactivating sodium current (86).

Apart from firing patterns, INaP appears to modulate synaptic responses. Thus dendritic excitatory postsynaptic potentials (EPSPs) are boosted by dendritic sustained sodium currents, consequently enhancing excitatory current flow to the soma way beyond the level that sole electrotonic conduction could achieve (87–89).

Neurons can display a further interesting activity pattern, an oscillatory fluctuation of the membrane potential. Although subthreshold, such phenomena are thought to underlie rhythmic synchronized neuronal activities and to determine the pacing of an entire neuronal network (90). Membrane-potential oscillations are thought to be crucially dependent on, or at least amplified by, INaP in neurons from a variety of structures such as neocortex (91–93), entorhinal cortex (94,95), and tegmentum (96).

In summary, the slow or noninactivating gating function of sodium channels is able to finely tune neuronal excitability. This property appears to rest in all sodium-channel subtypes, which may, however, be speculated to be differently affected in an epileptic condition. Some recently described genetic channel mutations associated with epilepsy in humans (see section, Epilepsy in humans) favor slow inactivation gating. In the future, it may be particularly promising to target AED discovery to compounds reducing INaP, and particularly to focus on the interaction of potential AEDs with known ion-channel mutations.

Expression, mutations, and functions of sodium channels in chronic epilepsy

Experimental epilepsy models offer the possibility of investigating sodium-channel expression and function in tissue that has already been epileptically active for more or less extended periods (i.e., with chronic or acute seizure activity). In the following, results of such investigations are reviewed. In most of these studies, no distinction between INaP and INa is made; and in the case of, for example, molecular investigations, such a distinction is indeed probably of little help, because functional states rather than structural differences appear to underlie INa and INaP (see earlier).

Although an inherent difficulty of such studies is to differentiate between cause and effect, many changes of expression, molecular characteristics, and functions of INa have been found in a variety of experimental preparations, including human epileptic brain tissue from epilepsy surgery, which suggest that altered sodium-channel expression or function may be one of the intimate processes underlying epileptogenesis. This holds true particularly for defined genetic mutations in some specific epilepsy syndromes (see section, Epilepsy in humans). Other evidence, obtained from studies on both human neurons and animal models, is less direct in that the pathogenetic mechanisms underlying sodium-channel alterations may be multifold, and the sodium-current changes associated with epilepsy would thus be only secondary to unknown defects of second messengers, mitochondrial function, phosphorylation states, and so on.

Experimental epilepsy models

One of the most commonly used chronic models of epilepsy is the kindling model. Here epilepsy is induced in animals by repetitive subconvulsive application of chemical or electrical stimuli. Hippocampal electrical kindling in rats has been found to alter sodium-current properties substantially. Thus peak sodium current is markedly enhanced, specifically in the long term (5 weeks after the last generalized seizure) (97). Furthermore, both as short-term (1 week after last seizure) and long-term effects of the kindling procedure, the voltage dependence of inactivation of INa is shifted to more depolarized potentials, resulting in an increased availability of sodium channels at resting membrane potential (97). The AED valproate (VPA) is able to reverse this shift in kindled animals, whereas carbamazepine (CBZ) is much less effective in doing so shortly after the last seizure, possibly reflecting therapy resistance of the model, and shows only comparable effectiveness several weeks after the last seizure (98) (see also later). Elevations of sodium-channel current on kindling epileptogenesis such as these may be species specific. Although electrophysiologic data may diverge from binding studies in that channel densities may not be strictly correlated to peak current amplitudes (99), data from frogs show a downregulation of sodium-channel density and thus suggest a reduction also of sodium currents with kindling in this species, presumably as a compensatory process counteracting hyperexcitability (100).

Other models of epilepsy include the systemic or local application of convulsive substances either in vivo or in vitro. One of these models is based on the intraperitoneal injection of kainate in rats, resulting in limbic seizures mimicking, at least in part, human temporal lobe seizures (101). After kainate injection, levels of messenger RNA (mRNA) encoding Nav1.2 and 1.3 subunits increase substantially in the hippocampus of rats, but not in neocortical tissue (102). This overall increase is accompanied by a shift from adult to neonatal isoforms of these mRNAs (103). Electrophysiologic data suggest that sodium currents carried by juvenile channels apparently possess electrophysiologic properties different from adult ones, as their activation and steady-state inactivation curves are shifted to depolarized potentials (50). Both the absolute increase in sodium channel mRNA—provided that it is followed by translation of functional channels—and the increased presence of juvenile forms might cause an increase in excitability due to both larger peak currents and higher availability of sodium channels around threshold. However, the processes after kainate injection possibly result in more complex reactions than a mere increase in sodium channels, as the β subunits, in contrast to the α subunits, are downregulated (104).

Rather than affecting sodium channels per se, chronic epileptic activity also may induce the production of molecules that interact with sodium channels and influence current properties. Thus Onozuka et al. (105) isolated a 70-kDa protein from neocortical tissue of rats in which focal epileptic activity had been induced by topical application of cobalt. This protein, when injected into the motor cortex of control rats, induced seizures in these animals, and likewise epileptiform bursting in snail neurons when applied intracellularly. In the latter neurons, the main bioelectric effect appeared to be a pronounced increase of inward current carried by sodium (105). Last, chronic epilepsy also may affect glial cell physiology: In kainate-injected rats, reactive and presumably immature astrocytes lose INa that they normally express. This change is attributed to the loss of neuronal elements due to kainate toxicity, suggesting that normal glial-channel outfit is maintained only in intact neuronal tissue (106).

Genetic models

Common genetic epilepsy models include several mouse strains such as the tottering (tg/tg) or epileptic (El) mouse, as well as rat strains and inbred populations of other species, including monkeys such as Papio papio. Possibly less conspicuous are certain strains of the fruit fly Drosophila, which generally are deemed to display some behavioral features at least analogous to epileptic fits.

Among Drosophila melanogaster, there are flies that carry mutations in gene loci termed “shaker” and “seizure.” Some of the previously mentioned animal strains appear to be affected by altered sodium-channel distribution or function. Thus the Drosophila“seizure” mutation shows convulsive seizures and consecutive transient paralysis when exposed to temperatures >38°C (107). Mutations in the “seizure” locus, two of which have been characterized (seits-1 and seits-2), result in alterations of voltage-gated sodium channels consisting either of a decrease in both sodium-channel and peak sodium current densities (107,108) or sodium channel structure, as reflected in a decreased saxitoxin-binding affinity (109). How these changes might eventually bring about the phenotype of the “seizure” mutation still remains unclear (40,41). One possible explanation may be that current density changes may simply affect only specific subpopulations of neurons, resulting in overall increased network excitability (108). Changes in channel structure, in turn, might be responsible for alterations in current kinetics, although no indications for this have been found (108,109).

One vertebrate strain of genetically epilepsy-prone animals is the epileptic “El” mouse. In these animals, neuronal sodium channel alterations consist of an overexpression of Nav 1.2 mRNA (110,111), an increased density of sodium channels determined by saxitoxin binding, particularly in the neocortex, and increased synaptosomal influx of 22Na+ induced by the sodium channel opener veratridine (111–113). As in “El” mice, in the mutant mouse tottering (tg/tg), apart from alterations of the P/Q calcium channel, modifications of sodium channel function have been found. In these animals, an increase of total 22Na+ influx through sodium channels in synaptosomes of neocortical neurons was demonstrated, although the sodium channel density, as determined by saxitoxin binding, appeared to be downregulated (99). In view of the calcium channel modifications, these changes could, however, be secondary. Last, myelin-deficient rat mutants, which among other symptoms show seizures, display clusters of putative voltage-gated sodium channels along neuronal axons, which might be held responsible for the hyperexcitability of these animals (114,115).

Epilepsy in humans

Epilepsy in humans is in many cases cryptogenic or is considered to have a multifactorial etiology, often with complex inheritance (116). Nevertheless, one can speculate that some structural correlates must account for the seizure susceptibility of epilepsy patients. For certain types of epilepsy, genetic linkage analyses have indeed yielded associations with specific gene mutations. One of these, a mutation on chromosome 21q, is coupled to progressive myoclonus epilepsy of the Unverricht–Lundborg type. The affected gene codes for a protein with homologies to sodium channel proteins, which may indicate that for this syndrome, sodium channel mutations may play a role (117).

One of the scientifically most exciting discoveries is the observation that distinct epileptic syndromes are definitely linked to sodium channel mutations (3,118). Thus recently, a syndrome of febrile convulsions in childhood followed by generalized seizures in adulthood [generalized epilepsy with febrile seizures plus (GEFS+) type 1] has been linked to a mutation of the β1 sodium channel subunit gene SCN1b on chromosome 19q (Fig. 1)(119). This subunit, as mentioned in section on ion channels, plays an important role in accelerating activation and inactivation kinetics. Consequently, sodium channels with an amino acid substitution corresponding to this mutation, and expressed in Xenopus oocytes, yielded currents with much longer inactivation kinetics than normal channels (119). Additional sodium channel mutations with GEFS+ have been discovered since on the α subunit SCN1a gene on chromosome 2q (GEFS+ type 2) (Fig. 1)(120–122). Electrophysiologic evidence from the Xenopus oocyte expression system, in which sodium channels with two of these mutations were introduced (T875M and R1648H; Fig. 1), suggests that inactivation of INa is slowed, or recovery from inactivation is accelerated, resulting in an overall greater availability of INa(123). Data from a mammalian expression system only partially confirm this finding: Although in one mutation (R1648H), acceleration of recovery from inactivation was likewise observed, in the other (T875M), inactivation of INa was not slowed but was faster (124,125). This seemingly paradoxical finding would still lead to hyperexcitability if this defect were functionally important only in inhibitory interneurons; their high firing rates would indeed predispose them to be more affected. Apart from SCN1, SCN2 or SCN3 α subunit genes also are possibly linked to GEFS+(126); in their study, the authors found changes in a chromosomal region containing SCN1A, 2A, and 3A genes, any of which might be a candidate gene. Recently a GEFS+ patient with a mutation in the SCN2A gene was reported (127). Experimentally introducing this mutation in rat SCN2A and expressing it in a mammalian cell line, again slowed inactivation was observed. Similarly, a mutation in the mouse SCN2A channel resulted in spontaneous seizures of these transgenic animals. This was associated with gliosis and neuronal loss in the hippocampus and an increased INaP sodium current in hippocampal CA1 neurons (128). Apart from causing the relatively mild GEFS+ syndrome, diverse mutations of the SCN1A gene were recently reported to be associated with severe myoclonic epilepsy of infancy (SMEI) (129)(Fig. 1). In contrast to GEFS+ displaying autosomal dominant inheritance, all seven mutations described with SMEI were de novo mutations. Alterations of the β2 subunits do not appear to play a key role in idiopathic generalized epilepsy syndromes (130,131). In summary, there is thus increasing and important evidence that at least some epilepsy syndromes constitute channelopathies.

Studies on brain tissue obtained during epilepsy surgery suggest also that focal epilepsy syndromes may be associated with changes of sodium channel distribution. Thus with complex partial seizures, the ratio between SCN1 and SCN2 sodium channel densities was significantly increased in hippocampal and neocortical temporal lobe tissue as compared with control autoptic samples (132). This effect may be argued to be [like the relative abundance of the two sodium channel classes in general (133)] region specific, because in frontal lobe epilepsy, a decrease of this ratio was observed (132).

Epilepsy surgery further offers the intriguing possibility of studying bioelectric properties of living neurons in human tissue. Thus in recent studies, sodium currents could be characterized in human hippocampal (134,135) and neocortical (50,135,136) neurons. The functional properties of these sodium currents do not diverge from animal preparations regarding both hippocampal and neocortical neurons (50,97,134), although current densities were found to be rather large, particularly in epileptic hippocampal neurons (134) and, interestingly, strongly and significantly so regarding persistent sodium current in subicular neurons from epilepsy patients (136). AEDs such as CBZ and VPA induce a shift of the steady-state inactivation curve in the hyperpolarizing direction, resulting in an overall decrease of sodium channel availability (135,137). In hippocampal neurons originating from hippocampal resectates showing typical signs of Ammon's horn sclerosis, CBZ had a far smaller effect than it did in neocortical neurons or cells from hippocampi without sclerosis (135) (see later). These findings match observations found in the kindling model (see earlier).

Not only neurons but also glial cells were investigated in human tissue resectates. As an unusual finding, glia cells from patients with oligodendrogliomas or astrocytomas show neuron-like physiologic properties in being able to generate action potentials and in displaying large-amplitude voltage-gated TTX-sensitive sodium currents (138–140). Similarly, astrocytes obtained from chronically epileptic human tissue also display INa. Channel density appears to be sufficiently high to support action-potential generation only in cells from the epileptic focus (141,142). One can speculate that the expression of such currents in glial tissue may facilitate glial Ca2+ waves and thereby widespread glial neurotransmitter release, causing an overall increase of network excitability (143,144). This in turn may contribute to the high incidence of seizures associated with brain tumors, although—at least for gliomas—a direct correlation between “spiking” glial cells and EEG abnormalities could not be established (145).

In conclusion, experimental evidence from widely differing models and preparations shows that epileptic activity can go along with different, and (depending on the model) sometimes opposing changes of sodium channel distribution, molecular structure, or function. Whether and precisely how these apparently static changes bring about paroxysmal events still must be clarified.

Antiepileptic drugs

Experimental evidence gathered during the past 15 to 20 years suggests that the sodium channel is a primary target for a variety of AEDs, the most effective of which are phenytoin (PHT), CBZ, and LTG where sodium channel blockade is considered to be the main mechanism of action (Table 1). Additionally, a host of other AEDs, including new and potential AEDs, whose main mechanism of action is often unrelated to sodium currents, have been shown to interact with voltage-gated sodium currents or INa-dependent processes. Some of these reports merit a note of caution: Whereas voltage-clamp experiments on mammalian neurons can specifically disclose sodium current alterations in nerve cells, analysis of spike shape, repetitive firing properties, toxin binding, or sodium uptake only indirectly reflects sodium channel involvement. Thus action-potential shape and repetitive firing properties are critically determined not only by sodium inward, but also by several outward currents as well. Drugs affecting potassium currents, for example, will automatically cause changes in these parameters. Studies in nonmammalian expression systems such as Xenopus oocytes may be difficult to interpret (see section on ion channels). Last, some drugs were shown to have effects on sodium currents or other sodium current–dependent processes, but at concentrations beyond therapeutic levels, and thus the mechanism of action cannot be considered to consist of sodium current modulation. For these reasons, Table 1 summarizes the data obtained in voltage-clamp experiments on mammalian neurons. It reveals that of the drugs discussed later, only CBZ, PHT, LTG, and felbamate (FBM) have been demonstrated to affect sodium currents specifically, with ambiguous data on VPA and topiramate (TPM). For the sake of completeness, all other substances are discussed as well. Special emphasis is given to voltage-clamp experiments whenever available, which are listed first in the different sections.

Table 1. 
  1. Overview of the effects on INa of antiepileptic drugs (AED). AED listed are those which have been shown to act on sodium-current dependent processes in different animal models irrespective of dosage (abbreviations of AED names as given in the text). A specific reduction of sodium currents in voltage-clamp experiments on mammalian neurons within the therapeutic concentration range could only be demonstrated for some AED (++; names underlined), whereas for others, reports are ambiguous (+/−) or even negative (−−), even though repetitive firing, sodium spikes or sodium-uptake were affected. For some AED, no voltage-clamp data are available (n.a.); in the case of OCBZ, a mechanism of action similar to CBZ is likely. GBP is listed with question marks due to its possibly indirect action observed only after very prolonged exposure. STM and LSG are not listed due to paucity of data. Approximate therapeutic concentrations in CSF or brain tissue and plasma (*) were extrapolated from appropriate chapters dealing with the above drugs from Wyllie et al. (146) (a), Engel and Pedley (147) (c), and microdialytic measurements of Rambeck et al. (148) (b).

CBZ60 μmol/l(a,b)++
DZP2 μmol/l*(a)−−
ETX100 μmol/l(a)−−
FBM350 μmol/l*(c)++
LTG20 μmol/l(b,c)++
OCBZ125 μmol/l(c)n.a.
PHB100 μmol/l*(a)−−
PHT200 μmol/l(a,c)++
REM3 μmol/l*(c)n.a.
TPM100 μmol/l*(c)+/−
VPA150 μmol/l(a)+/−
ZSM60 μmol/l*(c)n.a.

Most of the drugs act on sodium channels in similar ways (149–151). As an example, Fig. 3 shows the action of LTG on sodium currents in accutely isolated striatal neurons. LTG not only reduces sodium currents elicited by depolarizing the neurons to various potentials, but more important, it shifts the steady-state inactivation curve to the left. As a consequence, at hyperpolarized membrane potential, the blocking effect is much less pronounced than at depolarized potentials. Further, recovery from inactivation is much prolonged, so that availability of sodium channels at high activity states (depolarization, high-frequency discharge) is reduced. Thus one recurring theme of AED action on sodium currents is that their effects are often voltage and frequency dependent, with a pronounced action at depolarized potentials and—as a consequence—at high discharge frequencies. This general property makes the sodium channel block a particularly well-suited antiepileptic mechanism, as it would mainly occur in hyperexcitable states, but not under “normal” conditions. This review summarizes data on established, new, and potential AEDs suggesting sodium channel modulation, regarding both INa and INaP.

Figure 3.

Actions of the antiepileptic drug lamotrigine (LTG; 100 μM) on fast sodium current INa in accutely isolated hamster striatal neurons. A: Original traces show reduction of inward currents by LTG. B: Current–voltage relation of INa before (CTRL) and after LTG, showing that INa starts to activate at membrane potentials of approximately –40 mV and reaches peak currents at approximately –20 mV. Voltage protocol as in A. C: Steady-state inactivation curves of INa representing availability of INa at different membrane potentials. Under CTRL, at a resting membrane potential of approximately –80 mV, ∼50% of the channels are available for action-potential generation. The left shift of the steady-state inactivation curve caused by LTG reduces the availability of INa to ∼10%. Such an effect is not visible at –120 mV, showing that LTG block is voltage dependent. D: Prolongation of recovery from inactivation due to LTG. Note that, with LTG, even after 50 ms, INa has not recovered fully, thus limiting availability for high-frequency firing. Corresponding voltage protocols are shown as insets. Modified from refs. 31 and 32, with permission.

Actions of anticonvulsants on INa


Direct evidence of an interaction of CBZ with voltage-gated sodium currents has been gathered in a variety of neuronal preparations, and CBZ may share an extracellular binding site in sodium channels with PHT and LTG, whose actions on INa are discussed later (152). In accutely isolated hippocampal neurons, obtained during epilepsy surgery from patients with intractable epilepsy, or from rats, CBZ reduces sodium inward currents (98,137,153). This effect is primarily caused by a shift of the steady-state inactivation curve to hyperpolarizing potentials, thereby reducing availability of sodium channels at resting membrane potential levels, whereas CBZ does not modify activation parameters (98,137,153). The suppression of INa is very much voltage dependent [i.e., the effect was pronounced at depolarized membrane potential, with the apparent median inhibitory concentration (IC50) of 900 μM when eliciting currents from –100-mV holding potential reduced to 20 μM when stepping from –70 mV, indicating that CBZ binds primarily to the inactivated sodium channel](153). In comparison to PHT tested in the same preparation (see later), CBZ block occurs ∼5 times faster, but with lower affinity, suggesting that antiepileptic effects might be expected to be more pronounced with short-duration discharges (137,153). Additionally, CBZ is less potent in shifting the steady-state inactivation curve in neurons from chronically epileptic rats (98). A blocking effect of CBZ on INa also was reported for neuroblastoma cells maintained in culture. CBZ reduced sodium currents with half-maximal inhibition at ∼30–140 μM, and the block was both voltage and use dependent (i.e., it increased with repetitive stimulations) (154–156). Investigations in rat dorsal root ganglion neurons suggest that the action of CBZ may be dependent on the sodium channel subtype. Thus action-potential amplitudes as well as sodium currents were reduced potently by CBZ at a concentration of 100 μM in cells that displayed TTX-sensitive currents, whereas in TTX-resistant neurons, even CBZ concentrations of 300 μM were much less effective (157). This appears to hold particularly true for TTX-resistant currents displaying steady-state inactivation only at relatively positive potentials (158) and may explain why no effect of CBZ on sodium currents in dorsal root ganglion cells has been reported when the substance was used at a low concentration of 50 μM(159). It is interesting to note a similarity of findings in dorsal root ganglion cells (158) and CA1 neurons from kindled rats (97,98). Both show inactivation at relative depolarized potentials, and CBZ in these neurons is less effective in modulating INa.

Other evidence of a suppressive action of CBZ on INa was obtained in mammalian cell line transfected with Nav1.2 (160), mammalian and amphibian nerve fibers (161,162), and snail neurons (163), in essence confirming a voltage- and use-dependent block and negative shift of the steady-state inactivation curve. The reduction of sodium current under voltage-clamp conditions is reflected in the amplitude reduction of action potentials, and more important, a suppression of high-frequency repetitive firing as a functional consequence of prolonged binding of CBZ to sodium channels. This has been confirmed in both nerve fibers and neurons (165–167). At least at low concentrations of 30 μM, regular synaptic transmission, however, as judged by the effect of CBZ on evoked field potentials, is not impeded (168). The interaction of CBZ with sodium channels has been investigated by studying the competition of CBZ with binding of batrachotoxinin (BTX), a molecule adhering specifically to a site on the channel molecule responsible for the activation of sodium current, and by competition with veratridine or veratrine, two sodium channel openers. CBZ displaces BTX binding and reduces veratridine effects in rat brain synaptosomes and neocortical slices (168–171). Furthermore, it may act by modulating synaptic transmission due to blockade of sodium influx, as measured by 22Na+ uptake in synaptosomes or adrenal medullary cells homologous with sympathetic postganglionic neurons (172,173).

Diazepam (DZP).

Although its main mechanism of action is generally thought to involve the γ-aminobutyric acid (GABA)ergic inhibitory system, some experiments suggest sodium current blockade as one effect of DZP. In isolated rat hippocampal neurons, as well as in mouse neuroblastoma cells, DZP reduces INa by 40% from a holding potential of –80 mV, however, at concentrations way beyond therapeutic levels. This reduction is voltage dependent and caused by a negative shift of the inactivation curve, which appears to be less pronounced than the one obtained with PHT and CBZ (154,174).

DZP, in nanomolar concentrations, reduces repetitive firing in spinal cord neurons in culture, as do nitrazepam (NZP), lorazepam (LZP), and clonazepam (CZP) (175). In higher concentrations of 100 μM, DZP reduces compound action potentials in rat optic nerves, suggesting that, as for other AEDs, repetitive firing is much more vulnerable to DZP than are single action potentials (166).

Ethosuximide (ESM).

ESM, commonly known as a T-type calcium channel blocker, appears to block sodium currents in some preparations. In squid axons, ESM reduces sodium peak currents, although at very large and clinically irrelevant concentrations of 60 mM(177). Interestingly, the block is voltage dependent when applied internally, and voltage independent when perfused externally, suggesting two different blocking mechanisms, the former of which should primarily influence repetitive firing (177). In cat and rat thalamic neurons, ESM was reported to suppress primarily INaP without affecting INa(178) (see earlier). Both T-type calcium currents, as well as INaP, have been speculated to be especially suited to convey burst firing in neurons, a firing pattern deemed to pace epileptic spike-and-wave discharges during absence seizures by thalamocortical interplay. Correspondingly, ESM does reduce burst firing in thalamocortical neurons (178).

Felbamate (FBM).

In isolated rat striatal neurons, FBM reduced INa with an IC50 of 28 μM, inducing (like other AEDs) a shift of the inactivation curve to negative potentials (179).

Repetitive firing was likewise attenuated, although the first action potential remained unchanged, as were evoked excitatory postsynaptic or field potentials in striatal slice preparations (179). In a voltage-clamp study on Xenopus oocytes in which human or rat brain sodium channels had been expressed, FBM was effective in reducing INa only when applied intracellularly in a high concentration of 1 mM, again inducing a negative shift of the inactivation curve (180). FBM has been reported to reduce repetitive firing in cultured mouse spinal cord neurons (181).

Gabapentin (GBP).

GBP, which displays interactions particularly with synaptic systems, also may act by modulation of INa, but the mechanisms of this action appear to be less straightforward than those with other AEDs. Thus GBP reduces repetitive firing in mouse central neurons in culture, but this effect strongly depends on the duration of exposure to the drug and is maximal only after 12–48 h of incubation (176). Strikingly, the IC50 decreases substantially from ∼130 to 4 μM when the exposure time is increased from 60 s to 48 h (176), suggesting that an effect on sodium channels is exerted only after slow uptake or by metabolic channel modifications. This exceedingly long latency required to show full efficacy may explain why generally negative findings were reported for GBP. Thus with shorter application times or lower concentrations, GBP did not block repetitive firing in neuronal cultures of spinal cord (182), nor INa in cell lines expressing rat brain sodium channel (183) or snail neurons (184). In line with the notion that GBP does not bind directly to sodium channels is the observation that GBP does not reduce BTX binding (185).

Lamotrigine (LTG).

The interactions of LTG with sodium channels appear to be very similar to those found for PHT. Thus in voltage-clamp studies on neurons, accutely isolated or in culture, or neuroblastoma cells in culture, LTG potently blocked INa(32,155,186–189)(Fig. 3). This action was strongly voltage and use dependent, with IC50 ranging from 500 to 60 μM at holding potentials of –90 to –60 mV and ∼20% reduction of INa at concentrations as low as 1 μM(32,155,186–188), and associated with a negative shift of the steady-state inactivation curve (32,187–189) and a prolonged recovery from inactivation (Fig. 3)(31,32), reducing the availability of sodium channels and suggesting a block, particularly in the inactivated state. Like PHT, LTG is suggested to bind to sodium channels relatively slowly (187).

Repetitive firing was found to be reduced by LTG in concentrations of 10 to 100 μM in hippocampal and spinal cord neurons, without effects on single action potentials (190,191). Furthermore, LTG inhibits BTX binding in rat brain synaptosomes, as well as veratrine- or veratridine-induced release of NO and glutamate in rat brain both in vivo and in vitro (169,192,193).

Losigamone (LSG).

Data on the mechanism of action of LSG are scarce. Two reports support the notion that LSG may modulate INa. Thus LSG reduced the frequency of spontaneously occurring action potentials in cultured hippocampal neurons, as well as repetitive firing in the same preparation (194), confirming that LSG increases spike frequency habituation in entorhinal cortex neurons (195).

Oxcarbazepine (OCBZ).

OCBZ, which structurally only differs slightly from CBZ, and which is rapidly metabolized to an active 10-monohydroxy compound (MHD), effectively blocks repetitive firing of spinal cord neurons in culture, as does MHD. The IC50 for this effect was very low for both drugs: 50 nM for OCBZ, and 20 nM for MHD (196). Furthermore, like CBZ, OCBZ and MHD appear to act frequency dependently (i.e., with increasing stimulation frequencies, action potentials are elicited with decreasing probability) (196).

Phenobarbital (PB) and pentobarbital (PTB).

Evidence that barbiturates—generally acting on the GABAergic system—like PB or PTB might influence INa is scarce and is derived primarily from experiments on amphibian nerve fibers as well as squid and mammalian axons. Under voltage-clamp conditions, sodium inward currents are completely blocked by 3 mM PTB internally applied in squid axons (i.e., in a concentration nearly two orders of magnitude larger than therapeutic ones) (197). Similar findings have been reported in frog nerve fibers, with an IC50 close to 600 μM with extracellular application (161). Current flowing through human brain sodium channels incorporated into lipid bilayer membranes was voltage dependently diminished by PTB (198). Likewise, PB reduced sodium spikes at concentrations of ∼5 mM, and repetitive firing ∼1 mM, in rat preganglionic sympathetic nerves (165). PTB inhibits veratridine-stimulated 22Na+ uptake into synaptosomes, an effect precluded by prior long-term PB treatment, suggesting that, as with VPA (see later), some kind of adaptive process is initiated by prolonged sodium channel blockade (199).

Phenytoin (PHT).

As for CBZ, there is compelling evidence that PHT acts primarily by reducing INa, with only scanty reports explicitly denying sodium current modulation (200). Thus in neocortical neurons from rats, PHT in concentrations as low as 1 μM diminishes sodium currents by ∼20%, with an IC50 of 50 μM(187,201). Again, as with CBZ, this effect is use dependent and caused by a shift of the voltage dependence of inactivation in the hyperpolarizing direction (187,201). The effects of PHT on sodium currents have been studied very extensively in hippocampal neurons and neuroblastoma cell lines, yielding subtle differences in PHT action as compared with that of CBZ. Although the effect of PHT is equally voltage dependent, with IC50 ranging from 600 μM to <10 μM, in a holding-potential range of –90 to –50 mV (153–156,174,202,203), binding of CBZ is much slower and requires long depolarizations (up to seconds) to produce blocking effects (153,202,203) (see earlier). Consistent with this observation, single channel analysis revealed that late channel openings, rather than those immediately after the voltage step, are preferentially diminished by PHT (204,205), which may explain why plateau depolarizations underlying bursts as well as the persistent sodium current INaP are blocked, whereas action-potential generation is left intact (see later) (86,203,204). Different sensitivities of TTX-sensitive and TTX-resistant neurons, as have been observed for CBZ, also are present with PHT (157,158).

Blocking effects of PHT on INa were likewise reported in other preparations, such as Xenopus oocytes in which human brain sodium channels were expressed (205), a mammalian cell line transfected with rat Nav 1.2 (160), and nerve fibers (161,162,209). As functional consequences of slow sodium channel blockade, PHT limits repetitive firing without affecting single action potentials in neurons or evoked potentials in neuronal populations, at least in lower concentrations (154,166,168,210,211). Like CBZ, PHT itself binds to and reduces BTX binding to these channels (170,171,212,213). Mutations of the rat SCN2a sodium channel suggest that PHT binds to the inner lumen of the channel pore and shares this binding site with lidocaine and other local anesthetic drugs (150). Furthermore, PHT, like CBZ, reduces effects of the sodium channel opener veratridine (169,214,215).

Remacemide (REM).

In hippocampal slice preparations, REM, as well as its metabolite ARL12495AA reduces sustained repetitive firing of single neurons, with their IC50 values being practically identical at ∼60 μM(216). Furthermore, they also reduce action-potential amplitudes in higher concentrations of 120–400 μM(216). Regarding the racemic mixture of the metabolite, the respective stereoisomers are equally effective (217). Even lower IC50 values for the reduction of repetitive firing were reported for mouse spinal cord neurons in culture (i.e., 8 μM for REM, and 1 μM for its metabolite FPL12495AA), with the effect being voltage and use dependent (218).

Sulthiame (STM).

Regarding STM, only one report on isolated hippocampal neurons confirms an action on INa(219). In this study, both peak current amplitudes and repetitive spiking were reduced by STM in concentrations of 35 μM.

Topiramate (TPM).

The action of TPM on INa does not appear to consist of a straightforward block in all preparations: In rat cerebellar granule cells in culture, TPM was effective and depressed INa with a shift of the inactivation curve to hyperpolarized potentials with an IC50 of 50 μM(220).

Furthermore, TPM has been reported to attenuate repetitive firing in hippocampal and spinal cord neurons (221–223), although possibly with a distinctly delayed or even only transient action (210). Correspondingly, extracellular field potentials in hippocampal slices were decreased by TPM in a frequency-dependent fashion (223). In isolated neocortical neurons, however, TPM was reported to be ineffective in blocking INa at low concentrations (25–30 μM), and only at 100 μM did it cause a reduction associated with a left shift of the inactivation curve.

Valproate (VPA).

In rat neocortical neurons in cell culture, extracellularly applied VPA (0.2–2 mM) reduced sodium currents by ∼25%(224). The same holds true for human hippocampal neurons, which were obtained during epilepsy surgery; despite the therapy resistance of the patients, VPA led to a negative shift of the steady-state inactivation curve of INa, thus reducing excitability, with an IC50 of 1.4 mM(137). Similar results were obtained in hippocampal neurons of kindled and sham-operated rats (98).

In voltage-clamp experiments on axons of the squid, VPA led to a slowing of the activation of sodium currents and to a voltage-dependent reduction of peak sodium currents, notably only with internal application, suggesting an intracellular site of action (177). Likewise, VPA reduces INa in frog nerve fibers (225) and axons of the annelid Myxicola, shifting the inactivation curve to hyperpolarized potentials (226,227). The notion that VPA exerts its antiepileptic effect by sodium channel blockade is not unchallenged. In one study on hippocampal slices, evoked repetitive (doublet) firing was only slightly affected at physiologic temperatures, and an effect became evident only at <30°C (228). Paradoxic reactions have been reported as well. Prolonged treatment with VPA leads to an increase in sodium channel density and veratridine-induced current (229). Matching these data, in Helix neurons, Altrup et al. (230) observed a decrease and successive reincrease of action-potential duration with extracellular application of VPA (10 mM), and indeed, in the sodium current–dependent veratridine epilepsy model, VPA was reported to be proepileptic (231). The biphasic action of VPA as seen in the snail experiments was shown to be dependent on VPA first acting extracellularly, and then intracellularly (230). Apparently VPA slowly diffused into the cells, and only there, according to the authors, it exerts its main long-term antiepileptic effect, mediated in this case through potassium current modulation. An intracellular action, together with slow diffusion into the cells, would explain the delay of VPA action observed in clinical practice.

Zonisamide (ZNS).

The action of ZNS on sodium currents has been investigated only in annelid (Myxicola) axons. In this preparation, the compound reduces INa, with a concentration as low as 12 μM yielding half-maximal effects, which consist mainly, as for other AEDs, in a negative shift of the inactivation curve (227).

Potential AEDs.

Apart from the previously mentioned AEDs, a host of other compounds display anticonvulsant properties in experimental epilepsy models, and at the same time modify sodium current or sodium current–dependent processes such as repetitive neuronal firing. To list these findings extensively would go beyond the scope of this review. For the sake of completeness, some of the drug classes should, however, be mentioned. Among them are, not surprisingly, local anesthetic drugs such as lidocaine, whose prime action is considered to be exerted on INa(232), unspecific calcium channel modulators such as flunarizine, which also act on sodium channels (233,234), kava pyrones (235,236), aconitines (237,238), the candidate AED ralitoline (239,240), and the neuroprotective drug riluzole (187,241–244). In light of a renaissance of the ketogenic diet, it is of interest to note that polyunsaturated fatty acids (abundantly found in this diet), like AEDs discussed earlier, also cause a left shift of the inactivation curve of INa in micromolar concentrations (245).

Actions of anticonvulsants on INaP

As already stated, the persistent sodium current lacks the safety factor of inactivation the fast current possesses. Further, it is active also at resting membrane potential and does not need activation by depolarization. Thus small changes of INaP may have profound effects on neuronal excitability. On the basis of these findings, in addition to a large number of investigations on INa (see earlier), some studies have been carried out to test the effect of AEDs explicitly on INaP. In these studies, INaP was usually depressed by the AEDs. Thus PHT (50–200 μM) was shown to reduce neuronal excitability by blocking repetitive firing (207) and strongly diminishing INaP both in isolated cells and neurons in neocortical slices (206,207) and in isolated hippocampal neurons in microcultures (205). In lower concentrations of 8 μM, persistent TTX-sensitive plateau potentials were likewise blocked, whereas repetitive firing remained unaffected, suggesting that PHT reduces INaP, while sufficiently many sodium channels can still be recruited to generate action potentials (86). In neurons of the snail Lymnaea stagnalis, INaP induced by the convulsant PTZ was equally reduced by CBZ, although in rather high and clinically irrelevant concentrations of 1 mM(164). VPA (10–100 μMl) likewise decreased repetitive neuronal firing and epileptiform activity in rat amygdala slices and abolished the inwardly rectifying response to depolarizing current injection brought about by INaP(246). Likewise, in neocortical neurons, VPA was reported to block INaP selectively at concentrations of 10–30 μM(247). TPM showed some depressive effect on INaP in this preparation, which was, however, much less pronounced (248). ESM (750–1,000 μM) decreased INaP in thalamic neurons, notably without affecting the fast sodium current (178).