ReviewEPILEPSY

Molecular Signaling Mechanisms Underlying Epileptogenesis

See allHide authors and affiliations

Science's STKE  10 Oct 2006:
Vol. 2006, Issue 356, pp. re12
DOI: 10.1126/stke.3562006re12

Abstract

Epilepsy, a disorder of recurrent seizures, is a common and frequently devastating neurological condition. Available therapy is only symptomatic and often ineffective. Understanding epileptogenesis, the process by which a normal brain becomes epileptic, may help identify molecular targets for drugs that could prevent epilepsy. A number of acquired and genetic causes of this disorder have been identified, and various in vivo and in vitro models of epileptogenesis have been established. Here, we review current insights into the molecular signaling mechanisms underlying epileptogenesis, focusing on limbic epileptogenesis. Study of different models reveals that activation of various receptors on the surface of neurons can promote epileptogenesis; these receptors include ionotropic and metabotropic glutamate receptors as well as the TrkB neurotrophin receptor. These receptors are all found in the membrane of a discrete signaling domain within a particular type of cortical neuron—the dendritic spine of principal neurons. Activation of any of these receptors results in an increase Ca2+concentration within the spine. Various Ca2+-regulated enzymes found in spines have been implicated in epileptogenesis; these include the nonreceptor protein tyrosine kinases Src and Fyn and a serine-threonine kinase [Ca2+-calmodulin–dependent protein kinase II (CaMKII)] and phosphatase (calcineurin). Cross-talk between astrocytes and neurons promotes increased dendritic Ca2+ and synchronous firing of neurons, a hallmark of epileptiform activity. The hypothesis is proposed that limbic epilepsy is a maladaptive consequence of homeostatic responses to increases of Ca2+concentration within dendritic spines induced by abnormal neuronal activity.

Introduction

The epilepsies, disorders of recurrent seizures, affect about 1% of the population worldwide. Available therapy is symptomatic in that drugs inhibit seizures but are not disease-modifying; that is, no effective pharmacological prevention or cure has been identified. The term epileptogenesis refers to the process by which a normal brain becomes epileptic (see Table 1). Understanding the cellular mechanisms of epileptogenesis in molecular terms may help identify molecular targets for which small-molecule therapeutics can be developed to prevent epilepsy in individuals at high risk.

About 60% of all epilepsies are partial in nature, meaning that they are thought to arise from some localized area of hyperexcitable cerebral cortex. A particularly problematic form of partial epilepsy is termed limbic epilepsy, because it arises from foci within the limbic system, particularly the temporal lobe (synonyms include complex partial epilepsy, temporal lobe epilepsy, psychomotor epilepsy). Limbic epilepsy is the single most common form of epilepsy, accounting for about 40% of all cases of adult epilepsy. Limbic seizures are often resistant to antiseizure drugs: 30% of adults experience recurrent limbic seizures despite state-of-the-art treatment (1). Limbic seizures induce impairment of consciousness, thereby severely limiting important activities of daily living (such as driving or maintaining employment) and leaving the individual susceptible to bodily injury.

This Review focuses on the molecular signaling mechanisms underlying partial epilepsies, with particular emphasis on limbic epileptogenesis. Most of the studies considered here center on what are likely relatively early events in epileptogenesis. Other important pathological events have been identified in limbic epileptogenesis. For example, death of neurons, although apparently not necessary for the occurrence of epilepsy, likely contributes to the progressive severity of the condition (2); analysis of the signaling pathways underlying death of neurons is beyond the scope of this Review and the reader is referred to other considerations of this topic (3). Likewise, the potential contribution of inflammatory signaling pathways is addressed in other reviews (4)

Genetic and Nongenetic Causes of Limbic Epilepsy

The partial epilepsies are broadly divided into symptomatic and idiopathic epilepsies. The term "symptomatic" implies that the epilepsy is a symptom or consequence of some underlying structural lesion of cerebral cortex. This inference is based on the association of lesions in specific regions of cortex with epilepsy, the temporal relation of the lesion to the emergence of epilepsy, and the cure of the epilepsy following excision of the lesion is some instances. Diverse structural lesions can cause limbic epilepsy; in adults, these most commonly arise as a consequence of head trauma or stroke. An increasingly appreciated cause of symptomatic partial epilepsy is "cortical dysgenesis," a defect of cortical development with distorted neural migration.

The term "idiopathic" implies that no overt cause has been detected, except for genetic causes (5). The past 15 years has witnessed the discovery of a bewildering array of "epilepsy genes" (68). Positional cloning of inherited epilepsies in humans, mice, and flies has identified tens of responsible mutant genes. The unexpected occurrence of an epileptic phenotype in mutant mice in which a given gene has been deleted or overexpressed has also revealed many epilepsy genes.

The most stunning feature of the inherited idiopathic epilepsies in humans is that each of the 13 distinct genes identified as an epilepsy gene encodes a subunit of a voltage- or ligand-gated ion channel. Proteins whose mutation has been implicated in the pathogenesis of epilepsy include subunits of voltage-gated sodium (9) and potassium (10) channels, as well as ion channels gated by ligands such as γ-aminobutyric acid (GABA) and acetylcholine (11). Historically, this era began with the positional cloning of the Shaker locus in Drosophila in the late 1980s, which revealed that a loss-of-function mutation of the Kv1.1 gene can cause hyperexcitability (1214). Engineering a null mutation of Kv1.1 in mouse caused a phenotype of limbic epilepsy (15), and a point mutation of Kv1.1 was subsequently identified in a human pedigree that manifested limbic epilepsy (16, 17).

Although collectively these 13 distinct genes account for a tiny fraction (much less than 1%) of all human epilepsies, discovery of these genes provides a defined molecular etiology for an epilepsy. In no instance is the chain of cellular and molecular events by which these mutant ion channels culminate in the emergence of recurrent seizures understood. However, approaching this question when the etiology is understood in molecular terms seems more straightforward than when the etiology is a more complicated event such as traumatic injury of cortex. Moreover, the discoveries of these epilepsy genes provide the opportunity to establish in vivo and in vitro models of epileptogenesis based on recapitulating an identified molecular etiology in humans, thereby enhancing the likelihood that such models are valid.

Models of Epileptogenesis

Various in vivo models of epilepsy (see Table 2) have been established in both flies and mammals, including mice, rats, and subhuman primates. Models of epilepsy have also been developed in several in vitro systems, including freshly isolated rat hippocampal slices, primary cultures of dissociated embryonic rat cortical neurons, and intact hippocampi that include the commissural connections. A brief consideration of a subset of these models will simplify discussing studies of the signals that underlie epileptogenesis; for detailed considerations of these and other models, the reader is referred to excellent reviews on the subject (18, 19). Because most of these models center on the hippocampal formation in particular, a schematic of the hippocampal circuitry is included (Fig. 1).

Fig. 1.

Structure of the hippocampal formation. A confocal image was taken from a coronal section (40 μM) of green fluorescent protein (GFP)-transgenic mouse (thy 1-GFP line M, provided by G. Feng and S. Zhao of the Department of Neurobiology of Duke University), in which GFP protein is expressed in a subset of hippocampal principal neurons. Axons of entorhinal cortical neurons (not pictured) provide excitatory input to the apical dendrites (A in figure) of the dentate granule cells (cell bodies denoted by B in figure). Mossy fiber axons (denoted by C in figure) of the dentate granule cells provide excitatory input to the apical dendrites of CA3 pyramidal cell (denoted by D in figure). Schaffer collateral axons (not pictured) of CA3 pyramidal cells provide excitatory input to CA1 pyramidal cells (denoted by E in figure). Axons (not pictured) of CA1 pyramidal cells provide excitatory input to neurons in subiculum (denoted by F in figure).

Status epilepticus models

"Status epilepticus" refers to a state of continuous seizure activity. Status epilepticus can be produced experimentally by chemical convulsants or by virtually continuous electrical stimulations administered through intracerebral electrodes, the common feature being continuous limbic and motor seizures manifest by tonic (the initial sustained contraction) and clonic (brief, rhythmic contractions) contractions of limb and facial muscles that persist for hours (18). In these experimental models, a transient episode of status epilepticus is followed by a seizure-free "latent period" of one to several weeks, after which spontaneous seizures emerge (20, 21). This mimics the human condition in which an otherwise normal individual undergoes an episode of status epilepticus and develops recurrent seizures following recovery. For example, intense febrile seizures in young children are sometimes followed by the emergence of recurrent seizures years later. The status epilepticus models are associated with death of hippocampal neurons, proliferation of astrocytes, and inflammation that involves microglial activation, reflecting aspects of the hippocampal sclerosis observed in many humans with severe complex partial epilepsy. The discovery that complicated febrile seizures can be followed by (and presumably can cause) hippocampal sclerosis in young children (22) establishes a commonality between these models and the situation in humans. In the status epilepticus animal models, pathologic neuronal activity in the form of electrographic seizure is a prerequisite for epileptogenesis, because abbreviating the duration of status prevents late-onset epilepsy (23). In contrast to the kindling model described below, the pathological neuronal activity in the status epilepticus models occurs continuously over several hours rather than being distributed into brief episodes lasting tens of seconds that are evoked at daily intervals. In the status models, epileptogenesis is verified by the emergence of spontaneous, recurrent seizures as assessed in long-term video-electroencephalographic (EEG) recordings of awake, freely moving rodents.

An in vitro model of epileptogenesis that mimics status epilepticus involves briefly incubating dissociated cultured rat hippocampal neurons in nominally magnesium (Mg2+)–free buffer (24). In this "low-magnesium" model, hippocampi are removed from rat pups from which dissociated neuronal cultures are prepared. These cultures are maintained for 14 to 16 days in vitro to permit synaptic connections to become established. Subsequently, the cultures are incubated in nominally Mg2+-free buffer for 3 hours and are then returned to normal medium. Intracellular or whole-cell patch-clamp recordings made during the Mg2+-free incubation period reveal that the synaptically connected network of neurons develops epileptiform activity that occurs synchronously in many neurons, activity that evolves into spontaneous recurrent seizures lasting tens of seconds. These seizures continue despite replacement of normal medium and persist for the life of the cultured neurons (days). This model is similar to the in vivo status epilepticus models in that a brief episode of chemically or electrically induced status epilepticus serves as an inducing event that transforms a normal brain into an epileptic brain. The relative simplicity and accessibility of the preparation are important advantages of the low-magnesium model, although disruption of the normal neuronal circuitry is a disadvantage.

The kindling model and related in vitro models

Discovered about 35 years ago (25), kindling is an in vivo model of epileptogenesis in which a brief, low-intensity electrical stimulation is periodically administered to an experimental animal through stereotaxically implanted intracerebral electrodes. Initially these low-intensity stimulations induce no change in the EEG recorded at the site of stimulation and no change in behavior. Eventually they lower the local seizure threshold and evoke brief focal seizures that can be detected by EEG but are accompanied by no overt change in behavior. Continued application of these initially subconvulsive electrical stimulations results in progressive intensification of induced seizures, with prolonged limbic and tonic-clonic seizures evoked after about 12 to 15 stimulations. Once such seizures are reliably elicited, this enhanced sensitivity to electrical stimulation persists for the life of the animal. Such animals have seizures evoked by electrical stimulation but do not have spontaneous seizures. The administration of about 80 additional stimulations, each of which evokes a limbic and tonic-clonic seizure, results in the emergence of recurrent spontaneous seizures (26, 27). The repeated elicitation of focal seizure activity—not electrical stimulation per se—is critical to epileptogenesis.

Epileptogenesis is quantified in this model by simply counting the number of stimulation-induced electrographic seizures (detected by EEG) required to induce the first class 2 (head nodding) or first class 5 (rearing and falling) behaviors, indices of the development of kindling. The kindling model affords the investigator control as to when stimulation will be administered, which simplifies examining the effects of a given pharmacological agent. Indeed, the ease of quantifying epileptogenesis and the effects of pharmacological and genetic perturbations have contributed to the extensive study of the kindling model.

A number of in vitro models have been developed that likely involve some mechanisms similar to those of the kindling model. In particular, an acute hippocampal slice model involving repeated tetanic stimulation of Schaffer collaterals (28) results in evoked epileptiform activity that resembles "interictal" (between seizure) spikes, asymptomatic events recorded in the EEGs of humans and experimental animals with limbic epilepsy. The intracellular correlate of the interictal spike is a "paroxysmal depolarizing shift" (PDS), a brief (typically 250 ms or less) massive membrane depolarization with an accompanying burst of action potentials (29). Repeated tetanic stimulation of the Schaffer collateral afferents in hippocampal slices acutely isolated from adult guinea pigs resulted in stimulation-evoked PDSs in CA1 pyramidal cells as detected by intracellular recordings (28), a model subsequently referred to as "Schaffer collateral model." The preservation of some physiological synaptic circuitry and identified neurons (for instance, both principal neurons and interneurons) and their glial interactions, together with experimental accessibility, is an important advantage of this model.

A recently developed in vitro model uses two intact hippocampi connected by the hippocampal commissure isolated from postnatal day 6 (P6) to P7 rats and maintained in vitro for 24 hours or so (30). This is a technical tour de force in which the hippocampus from each side of the brain and the connecting commissure each reside in separate experimental chambers, permitting each hippocampus and the commissure to be exposed to distinct solutions. Repeated applications of the convulsant kainic acid to one hippocampus (the "primary" hippocampus) induce repeated seizures in the contralateral ("secondary") hippocampus, which is bathed in normal medium. This procedure eventually results in the emergence of spontaneous recurrent seizures in the secondary hippocampus despite no further treatment of the primary hippocampus with kainic acid. Application of the sodium channel blocker, tetrodotoxin (TTX), to the commissural chamber prevents propagation of seizures and the emergence of spontaneous seizures in the secondary hippocampus. This "three-chamber" model retains the circuitry of the intact hippocampi, yet permits perturbation of each hippocampus and of the commissure independently, because each is maintained in a distinct chamber.

Other models

Description of other models relevant to a single section of this review will be presented in the appropriate section below. Numerous additional models have been developed, including mouse models in which engineered mutations produce epilepsy mimicking that observed in some rare Mendelian forms in humans. These incorporate genetic models of both "idiopathic" and "symptomatic" epilepsy (15, 31). Models of symptomatic epilepsies induced by injury, cortical dysgenesis, and others have also been developed (3234). These will not considered here, because only limited study of the role of signaling pathways in epileptogenesis has been undertaken in these models to date.

Ionotropic Glutamate Receptors and Epileptogenesis

Ionotropic glutamate receptors mediate the vast majority of excitatory neurotransmission in the central nervous system (CNS). The molecular cloning in the early 1990s of cDNAs encoding glutamate receptor subunits clarified earlier classification systems based on the pharmacological and electrophysiological properties of these receptors (35). The ionotropic glutamate receptors can be divided into three major classes: (i) α-amino-3-hydroxy-5-methyl-4-isoxazole propionate (AMPA)–type receptors, (ii) N-methyl D-aspartate (NMDA)–type receptors, and (iii) 2-carboxy-3-carboxymethyl-4-isopropenylpyrrolidine (kainate)–type receptors. With respect to epileptogenesis, the principal focus has been on the role of the NMDA receptor and to a lesser extent the AMPA receptor and the signal transduction pathways initiated by their activation.

NMDA Receptors and Epileptogenesis

Pharmacological studies

The ease of quantifying epileptogenesis in the kindling model led to studies examining the role of NMDA receptors in the development of kindling. Structurally distinct antagonists of different regions of the NMDA receptor (including the glutamate or glycine recognition sites or the channel itself) administered just before an electrical stimulation powerfully inhibit epileptogenesis, as evident in increased numbers of stimulations required to induce kindling (3638). When rats were treated with high doses of NMDA antagonists (with marked unwanted behavioral effects) (37, 38), two to three times as many stimulations were required than for vehicle-treated controls. In contrast, treatment with the AMPA receptor antagonist NBQX [6-nitro-7-sulfamoylbenzo[f]quinoxaline-2,3-dione] did not inhibit development of kindling, despite the use of doses that effectively inhibit seizures evoked after induction of kindling (39). Moreover, once epileptogenesis had been induced in vehicle-treated animals, treatment with NMDA receptor antagonists was relatively ineffective in inhibiting seizures (38).

NMDA receptor antagonists are also effective antiepileptogenic agents in various additional models. Treatment with the uncompetitive antagonist of NMDA receptors, MK-801, prevented late-onset spontaneous, recurrent seizures in the pilocarpine model even without modifying the duration of the status epilepticus (40). Treatment with MK-801 administered after the first hour of status epilepticus prevented late-onset spontaneous recurrent seizures in the electrical stimulation–induced status epilepticus model (23), which implies that a given duration of NMDA receptor activation is required for epileptogenesis. NMDA receptor antagonists also inhibit epileptogenesis in diverse in vitro models, including the three-chamber model in which bathing the "target" or secondary hippocampus with an NMDA receptor antagonist (APV) prevents emergence of spontaneous seizures. Likewise, inclusion of NMDA receptor antagonists in the low-magnesium model prevents emergence of spontaneous, recurrent seizures (24). Likewise, NMDA receptor antagonists inhibit stimulation-induced bursting of CA1 pyramidal cells in the Schaffer collateral model in isolated slices (28, 41).

Genetic studies

Genetic perturbations enable assessment of the effects on epileptogenesis of specific NMDA receptor subunits. However, the identity of the particular subunit combinations required for epileptogenesis remains incompletely understood. Determining the role of either of these subunits in epileptogenesis is precluded by the neonatal deaths of mice carrying null mutations of either the NR1 or NR2B subunits. Genetic studies implicate NR2A-containing NMDA receptors in epileptogenesis in that gene-targeted mice lacking the large carboxy-terminal domain of the NR2A subunit exhibited impaired epileptogenesis (42) similar to that induced by NMDA receptor antagonists. Mice lacking the carboxy-terminal domain of the NR2B subunit died soon after birth, which precludes addressing the contribution of NR2B-containing NMDA receptors (42). Pharmacologic experiments examining putative NR2B-selective antagonists have yielded conflicting results (43, 44). The lack of highly selective NR2A antagonists has prevented assessment of these receptors.

In contrast to the deletion experiments described above, transgenic overexpression of NR2D resulted in a marked impairment of kindling (45). Although it was not definitively shown that the overexpressed NR2D subunits were incorporated into the NMDA receptor complex, the most likely explanation for the decreased rate of kindling development in these animals is the reduced conductance of the NR2D-containing NMDA receptors (46). These data suggest that NMDA receptor subunit composition is a key variable in the rate, and possibly occurrence, of epileptogenesis.

Putative cellular and molecular mechanisms

The pharmacological and genetic studies of NMDA receptors in diverse in vivo and in vitro models provide a remarkable consensus that NMDA receptor activation promotes limbic epileptogenesis. Given that NMDA receptors are undergoing activation during the epileptogenic stimuli in these models, these findings raise the questions of what plastic cellular changes resulting from NMDA receptor activation might underlie epileptogenesis and, in turn, what are the molecular signaling mechanisms by which NMDA receptor activation effects these plastic changes. Three of the most likely mechanisms for NMDA receptor–mediated epileptogenesis are considered here, namely, decreased synaptic inhibition, enhanced synaptic excitation, or a structural reorganization that is itself proepileptogenic. Note that these mechanisms are not mutually exclusive. Our consideration of signaling mechanisms underlying plastic changes in synaptic function centers on posttranslational modifications, whereas our consideration of pathways underlying long-term structural plasticity centers on regulation of gene expression (Fig. 2).

Fig. 2.

NMDA receptor (NMDAR) activation promotes epileptogenesis. Neuronal activity triggers the synaptic release of glutamate, which binds to and activates NMDA receptors at postysynaptic sites on dendritic spines. Activation of NMDA receptors leads to Ca2+ influx. The increase in intracellular Ca2+ activates CaMKII and calcineurin. Activated CaMKII promotes the surface insertion of GluR1 of AMPA receptors at synaptic sites and thus contributes to LTP of excitatory synapses. Activated calcineurin binds to and dephosphorylates the β2/3 or γ2 subunit of GABAA receptors, resulting in accelerated internalization of GABAA receptors and depression of GABA-mediated synaptic inhibition. An alternative mechanism of reducing GABA-mediated synaptic inhibition involves the reduced function or expression (or both) of KCC2, which results in a shift of the chloride equilibrium potential, transforming inhibitory to excitatory neurotransmission. The increase in intracellular Ca2+ concentration also enhances Ca2+-dependent gene expression and promotes mossy fiber sprouting. We suggest that excessive NMDA receptor activation and the resulting sustained increase of Ca2+ might induce a homeostatic reduction of CaMKII activity and expression, leading to epileptogenesis by unknown cellular mechanisms.
A subset of these same pathways may contribute to epileptogenesis after neonatal seizures. Here, the increased intracellular Ca2+ resulting from activation of Ca2+-permeable AMPA receptors activates calcineurin and leads to dephosphorylation of the β2/3 subunits of GABAA receptors (GABAARs) and reduction of synaptic inhibition.

The possibility that reduced synaptic inhibition underlies the hyperexcitability found in epilepsy has long dominated consideration of the mechanisms of the epilepsies (47). This hypothesis leads to the question of whether impaired GABAA receptor–mediated inhibition can be demonstrated in these models of epileptogenesis, and if so, whether impaired GABAA receptor–mediated inhibition is a consequence of NMDA receptor activation.

Impairments of GABAA receptor–mediated synaptic inhibition have been identified in three distinct in vitro models of epileptogenesis and one in vivo model. Intracellular recordings of CA1 pyramidal cells in the Schaffer collateral model revealed reductions in the amplitude of spontaneous and evoked inhibitory postsynaptic potentials (IPSPs) accompanied by reduced responsiveness to iontophoretically applied GABA (28). Likewise, whole-cell patch-clamp analyses in the low-magnesium model revealed impaired synaptic inhibition, as evidenced by reduced amplitude of miniature IPSCs (48). Reduction of inhibitory GABA synaptic function accompanied emergence of spontaneous seizures in the three-chamber model (30). The results of these three distinct in vitro models are reinforced by study of an in vivo model. Field potential recordings of dentate granule cells in an electrical stimulation–induced status epilepticus model revealed impaired paired-pulse inhibition, a finding consistent with weakened GABAA receptor–mediated synaptic inhibition (49). In each of these four models, NMDA receptor antagonists prevented the suppression of GABA-mediated inhibition, which implies that disinhibition is one mechanism by which NMDA receptor activation could promote epileptogenesis.

What is the mechanism underlying the impairment of GABA-mediated inhibition? The reduced responsiveness of CA1 pyramidal cells to iontophoretic GABA (28) indicates that the locale of the reduction is at least in part postsynaptic. Analyses of the low-magnesium model revealed that the reduced amplitude of the miniature IPSCs was accompanied by increased intracellular accumulation of GABAA receptor subunit β2/3, a finding consistent with removal of synaptic GABAA receptors due to an accelerated rate of endocytosis (48, 50).

What are the signaling events triggered by NMDA receptors that lead to impaired GABA-mediated inhibition? Studies of NMDA receptor–mediated reduction of GABA responsiveness in a simple model provide an interesting possibility. Whole-cell recordings of hippocampal neurons removed from a slice acutely isolated from an adult rat (51) demonstrated that a brief (several-second) application of NMDA induced the abrupt onset (within seconds) of a brief (lasting 12 s) reduction of GABA-mediated responses. Addition of okadaic acid, a phosphatase inhibitor, prevented the NMDA-induced suppression of the GABA response. Likewise, inclusion of an autoinhibitory fragment of calcineurin, the Ca2+-calmodulin-activated protein phosphatase, in the intracellular perfusing solution also reduced NMDA-triggered suppression of the GABA response. This finding supports the interpretation that NMDA receptor activation induced a Ca2+-dependent dephosphorylation process that suppressed the GABA response. This interpretation found support in analyses of CA1 pyramidal cells after high-frequency stimulation of the Schaffer collateral afferents (52). These tetanic stimuli of the Schaffer collaterals simultaneously induce both long-term potentiation (LTP) of the CA3 to CA1 excitatory synapse and long-term depression (LTD) of inhibitory synapses on CA1 pyramidal neurons. Induction of LTD of the inhibitory synapses was accompanied by recruitment of activated calcineurin to the second intracellular domain of the γ2 subunit of the GABAA receptor (52). Whether activation of calcineurin results in accelerated receptor internalization, reduced function of cell surface receptors, or both is uncertain, but this regulation of calcineurin, nonetheless, provides a plausible signaling mechanism by which NMDA receptor activation reduces responsiveness to synaptically released GABA; the net effect is epileptiform activity.

The mechanism underlying reduced responsiveness to GABA in the intact hippocampi model differs from that observed in the low-magnesium model (51). That is, the reduction of inhibitory GABA synaptic function identified in the three-chamber model (30) was due to a shift of the chloride equilibrium potential; the net result was a transformation of GABA responses from mainly inhibitory to mainly excitatory. Reduced activity of the K+-Cl cotransporter KCC2 would provide a plausible molecular explanation for this shift. How NMDA receptor activation might effect this change is unclear.

In each instance described above, NMDA receptor–mediated GABA disinhibition has been identified in an acute model of epileptogenesis, with three of the four representing status epilepticus models analyzed within minutes of onset of the status epilepticus (30). Paradoxically, evidence of increased GABAA receptor–mediated synaptic inhibition has been identified in some long-term models of epilepsy, like kindling (53, 54). Because increased GABAA receptor–mediated synaptic inhibition of dentate granule cells seen in long-term models would be expected to reduce the likelihood of a seizure, this may represent a homeostatic response to recurrent seizures rather than a cause of epileptogenesis. Such findings in a long-term in vivo model in no way minimize the significance of the reduced inhibition common to an early stage of epileptogenesis in these other models.

One feature of epileptogenesis in the kindling model is that repeated application of an initially subconvulsive stimulus induces seizures that involve progressively more widespread populations of neurons within the limbic system, as evidenced by the increased duration and intensity of limbic seizures. This same progression of behavioral seizure intensity is evident during the earliest stages of seizures in the status epilepticus models. These populations of neurons are almost certainly connected mainly by excitatory synapses, most of which use glutamate as a neurotransmitter. The fact that sustained, high-frequency firing of neurons promotes more widespread invasion of synaptically connected populations of neurons led to the idea that potentiation of the efficacy of excitatory synapses similar to that of LTP contributed to the cellular mechanisms of epileptogenesis (55). Indeed, repeated LTP-inducing stimuli facilitate epileptogenesis in the kindling model (56). Moreover, a stimulation-induced focal seizure is sufficient to induce LTP (57). Enhanced function of excitatory synapses has also been identified in recordings from CA1 pyramidal cells in the Schaffer collateral model (28). Intracellular recordings of granule cell responses to perforant path stimuli in acutely isolated hippocampal slices revealed an enhanced NMDA receptor component in the evoked excitatory postsynaptic potentials (EPSPs) following induction of kindling, providing evidence for enhanced function of an excitatory synapse (58). Together, these findings raise the possibility that one mechanism of epileptogenesis involves enhanced function of excitatory synapses.

In addition to the plastic changes in excitatory and inhibitory synaptic function addressed in the preceding sections, structural remodeling of neuronal circuits may also promote epileptogenesis. The most extensively studied population of neurons that undergo structural changes and network reorganization during epileptogenesis is the dentate granule cells of hippocampus, the reorganization of which is evident in many animal models, as well as in humans (59). Sprouting of the mossy fiber axons of the granule cells during epileptogenesis leads to formation of aberrant synaptic connections between granule cells, resulting in an expansion in the number of recurrent excitatory synapses among the granule cells (6063). In addition, seizures increase granule cell neurogenesis, and some of these newborn neurons migrate to ectopic locations, notably the hilar region of the dentate gyrus (64, 65). Many granule cells of the epileptic brain exhibit not only axonal sprouting, but also a basal dendrite that provides a new surface for innervation by recurrent mossy fibers (66, 67). The circuitry of the dentate gyrus of the epileptic brain thus shows a reduced threshold for synchronous firing of granule cells both in humans (68, 69) and animal models (7073), and it seems likely that the aberrant circuitry contributes to this increased excitability. The extent of mossy fiber sprouting is greatest in animal models in which death of hilar neurons has occurred, which suggests that the sprouting is induced in part by reinnervation of synapses denervated as a consequence of death of hilar neurons. Yet mossy fiber sprouting has also been identified in early stages of the kindling model at a time when no loss of hilar neurons is detectable (74); this reorganization is reminiscent of the reorganization of microcircuitry in adaptive plasticities such as that of the inferior colliculus of the barn owl induced by raising the animal with prisms that displace the visual field (75). Formation of functional, aberrant recurrent excitatory synapses among neurons of neocortex was identified in a model of trauma-induced epileptogenesis (76), raising the possibility that such structural reorganizations may be common to various types of partial epilepsy.

The molecular mechanisms underlying the structural plasticity of the dentate granule cells are incompletely understood. Interestingly, pharmacological blockade of NMDA receptors not only impairs epileptogenesis in the kindling model, but also impairs induction of mossy fiber sprouting in response to repeated evoked seizures (77). Pretreatment with the NMDA receptor antagonist MK801 resulted in reduced synaptic reorganization of the mossy fiber pathway, as compared with vehicle-treated kindled animals experiencing an equivalent number of seizures. MK801 also reduces mossy fiber sprouting in kainic acid–treated animals (78). The fact that several days appear to be required for development of these structural changes, together with their lifelong persistence, raises the possibility that the molecular mechanism involves NMDA receptor–dependent regulation of gene expression. One gene that might be involved is the immediate early gene, c-fos, which encodes a transcription factor. Seizures induce dramatic increases in expression of c-fos mRNA in the dentate granule cells, increases mediated in large part by activation of NMDA receptors (79). Moreover, the extent of mossy fiber sprouting is reduced in c-fos null mutant compared with wild-type mice after kindling (74). An additional candidate is the gene encoding brain-derived neurotrophic factor (BDNF); NMDA receptor activation mediates much of the dramatic increases of BDNF expression induced by seizures (80). Evidence supporting a role for BDNF in structural plasticity of the granule cells is addressed in the section on neurotrophins.

A number of additional candidate epilepsy genes have emerged in the past several years based on the criteria that expression of these genes is regulated during epileptogenesis and that they encode proteins that have been implicated in regulation of axonal growth. Regulation of axonal growth is mediated through three general classes of genes and the proteins they encode: (i) myelin-associated proteins, (ii) extracellular matrix proteins, and (iii) developmentally associated growth cone repellants. Expression of at least one member of each of these classes is regulated in a model of epileptogenesis. First, the number of transcripts of the myelin-associated protein NogoA are increased in the kainate model at a time that correlates with the initiation of mossy fiber sprouting (81). Second, expression of multiple extracellular matrix proteins, including several proteoglycans (82, 83) and the adhesion-like molecule tenascin-C (8385), is regulated in status epilepticus models. Indeed, the increased expression of tenascin-C RNA and protein after kainate status epilepticus is postulated to guide migrating neurites during mossy fiber sprouting. Finally, mRNA content of developmentally associated growth cone inhibitors such as molecules of the semaphorin classes is reduced in the kainate status epilepticus model (86). Whether the altered expression of any of these genes is a consequence of NMDA receptor activation remains unclear.

Perspective

NMDA receptor activation is a key step in epileptogenesis in various models. Given the critical role of Ca2+ in the diverse plastic changes triggered by NMDA receptors, it seems likely that flux of Ca2+ through the NMDA receptor initiates the molecular signaling cascades that culminate in epileptogenesis.

AMPA Receptor Activation and Epileptogenesis

Pharmacological and genetic evidence

Neonatal seizures are associated with an increased risk of epilepsy later in life (87), which suggests that these seizures themselves may be epileptogenic. The most common cause of neonatal seizures is hypoxia. A brief hypoxic episode in the neonatal period is sufficient to evoke seizures in rats, particularly when hypoxia is administered at age P10 to P12 rather than in younger or older animals (88). Hypoxic seizures induced at P10 to P12 result in increased susceptibility to seizures evoked later in life (89), demonstrating a proepileptogenic effect of these early life seizures.

The hypothesis emerged that AMPA receptor activation may be critical to the proepileptogenic effects of these early life seizures. This hypothesis derived from the observation that the time of peak sensitivity to hypoxic seizures coincides with the time of peak expression of Ca2+-permeable AMPA receptors. That is, a relatively lower expression of the GluR2 subunit compared with other AMPA receptor subunits in neocortex and hippocampus occurs at this time in development (90, 91). Consistent with the hypothesized role of AMPA receptors in epileptogenesis, treatment with the AMPA receptor antagonist NBQX immediately after seizures induced at P10 to P12 prevents an enhanced sensitivity to seizure-induced cell death later in life and results in a nonsignificant trend toward reversing the persistent reduction of seizure threshold.

Putative cellular and molecular mechanisms

Remarkably, hippocampal slices isolated at times ranging from 10 min to hours after a brief hypoxic episode in P10 rats exhibited increased excitability, thereby providing the opportunity to examine the underlying mechanisms. For example, hippocampal slices from animals subjected to low oxygen levels compared with controls exhibit a greater likelihood of low Mg2+–evoked seizures in CA1 pyramidal cells (92). Whole-cell patch-clamp recordings of CA1 pyramidal cells from slices isolated 1 hour after hypoxia revealed reduced frequency and amplitude of spontaneous IPSCs and reduced amplitude of evoked IPSCs (93), findings consistent with reduced GABA-mediated synaptic inhibition of the pyramidal cells. As described in the above section on NMDA receptors, enhanced activation of calcineurin can reduce GABAA receptor responsiveness (51, 52), which suggests that calcineurin might play a role in the disinhibition induced by hypoxic seizures (93) (Fig. 2). In comparison with normal controls, increased activity of calcineurin was detected in homogenates of the CA1 region in slices isolated 1 hour after in vivo hypoxia. The enhanced activation of calcineurin correlated with reduced serine-threonine phosphorylation of the β2/3 subunits of the GABAA receptor, as well as with reductions of spontaneous IPSCs in slices after hypoxia in vivo. Incubation of slices isolated from rats under hypoxic conditions with a calcineurin inhibitor, FK-506, or a calcineurin autoinhibitory peptide increased spontaneous IPSCs recorded in CA1 pyramidal cells. Likewise, incubation of slices from rats exposed to low oxygen levels with the noncompetitive antagonist of AMPA receptors, GYKI 52466, prevented the increased calcineurin activity and also increased frequency and amplitude of sIPSCs in CA1 pyramidal cells. Similarly, Joro spider toxin (JSTx), an antagonist of Ca2+-permeable AMPA receptors, increased the frequency and amplitude of sIPSCs in slices from animals given inadequate oxygen. Together, these findings are consistent with a model (Fig. 2) in which hypoxic seizures result in increased activation of Ca2+-permeable AMPA receptors by glutamate during and after hypoxia-induced seizures in P10 rat pups. The enhanced activation of these Ca2+-permeable AMPA receptors leads to activation of calcineurin, dephosphorylation of the β2/3 subunits of the GABAA receptor, and reduced GABA-mediated synaptic inhibition of CA1 pyramidal cells. This sequence of events is proposed to promote epileptogenesis, as evident in a long-term reduction of seizure threshold. Consistent with this notion is the fact that treatment of animals with the calcineurin inhibitor, FK-506, immediately following the hypoxic seizures prevented the enhanced seizure frequency and duration induced by a second exposure to hypoxia, 24 hours later.

The idea that enhanced activation of Ca2+-permeable AMPA receptors may contribute to epileptogenesis is supported by an entirely distinct line of evidence (94). The arginine (R) residue at position 586 of the GluR2 subunit renders heteromeric AMPA receptor channels impermeable to calcium (9598). This single amino acid change results from RNA editing of a genomically encrypted glutamine (Q) codon (99). If editing of the Q codon is abolished by knockout of the editing enzyme ADAR2 (adenosine deaminases that act on RNA) (100) or by removing the cis-acting exon complementary sequence (ECS) in intron 11 of the GluR2 gene (94), a substantial portion of the AMPA channels switch from low to high Ca2+ permeability. One of the phenotypes of the heterozygous mice engineered by gene targeting to harbor an editing-incompetent GluR2 allele in principal neurons and interneurons was the development of seizures and death by 3 weeks of age (101). However, when the AMPA receptor switch to high Ca2+ permeability was attenuated by inducing it in only a few forebrain areas during juvenile development, early death could be avoided, and most mice lived longer than 2 months (101). These mice exhibited limbic seizure activity, as evident in video monitoring and EEG recordings. Depth electrode recordings revealed the first detectable seizures to be in hippocampus. Collectively, these observations demonstrate that enhanced expression of Ca2+-permeable AMPA receptors, in particular, expression of the unedited GluR2(Q) subunit itself, is sufficient to induce limbic epilepsy.

Perspective

As with the NMDA receptor, the flux of Ca2+ through the Ca2+ -permeable AMPA receptor is almost certainly a signal that initiates the molecular signaling cascade that culminates in epileptogenesis. Also, as does the signal mediated by the NMDA receptor, a synaptic disinhibition mediated by endocytosis of GABAA receptors likely contributes to the increased excitability. In contrast to the key role of NMDA receptors in epileptogenesis in the mature brain, the Ca2+ -permeable AMPA receptors promote epileptogenesis during a brief stage of postnatal development marked by the high expression of these receptors.

Metabotropic Glutamate Receptors and Epileptogenesis

In addition to the ionotropic receptors considered above, glutamate also signals through metabotropic receptors (mGluRs), a family of heterotrimeric guanine nucleotide–binding protein (G protein)–coupled receptors (GPCRs). The family of mGluRs includes 8 subtypes that have been implicated in distinct aspects of CNS physiology and pathology. Group 1 mGluRs (mGluR1 and 5) are mainly coupled through Gq/11 to phospholipase C, whereas the group 2 (mGluR 2 and 3) and group 3 (mGluR4, 6, 7, and 8) mGluRs are mainly coupled through Gi/Go family G proteins to inhibition of adenylyl cyclase activity. With respect to epileptogenesis, studies have focused on group 1 mGluRs, which are localized to the peripheral parts of postsynaptic densities and regulate synaptic plasticity (102). Evidence of mGlu5 localization to astrocytes has also emerged (103).

Pharmacological and genetic evidence

The key evidence linking group 1 mGluRs to epileptogenesis is the finding that bath application of a group I-specific mGluR agonist, dihydroxyphenylglycine (DHPG), induces epileptogenesis in acutely isolated hippocampal slices. Intracellular recordings of CA3 pyramidal neurons revealed that the GABAA receptor antagonist picrotoxin induced PDSs, typically lasting 250 ms or less (104). Addition of DHPG for 30 to 60 min transforms these PDSs into recurrent episodes of electrographic seizures lasting one to several seconds as detected in intracellular recordings of CA3 pyramidal neurons. These seizures consist of a massive depolarization associated with high-frequency firing of action potentials (104). The recurrent seizure activity persists for hours despite removal of DHPG from the bath, which implies that the recurrent seizures were not simply due to persistence of a convulsant agent, but rather that epileptogenesis had been induced by a 30- to 60-min incubation with DHPG.

The development of mGluR subtype–specific antagonists permitted elucidation of the contribution of distinct mGluR subtypes to epileptogenesis in this model. Although DHPG is selective for group 1 mGluRs, DHPG is an agonist for both mGluR1 and mGluR5. An mGluR5-selective antagonist [2-methyl-6-(phenylethynyl)-pyridine (MPEP)] prevented DHPG-induced epileptogenesis, whereas an mGluR1 selective antagonist (LY367385) did not (105). Once epileptogenesis had occurred and recurrent seizures were expressed (despite removal of DHPG), incubation with selective antagonists of either mGluR1 or mGluR5 partially inhibited the seizures (although the mGluR1 antagonist was more effective); moreover, the antiseizure effects of antagonists of mGluR1 and mGluR5 receptors were additive. If the assumptions regarding specificity of these drugs are correct, then activation of mGluR5 is required for epileptogenesis in this model, whereas activation of both mGluR1 and mGluR5 is required for full expression of the seizures.

One limitation of this model is that it relies on activation of mGluRs by bath application of an agonist rather than on synaptically released glutamate. This limitation has been addressed in part in recent studies of hippocampal slices acutely isolated from mice carrying null mutations of the fragile X mental retardation protein (FMRP) (106). As described above, incubation of slices from wild-type mice in a GABAA receptor antagonist resulted in PDS lasting 250 ms or less; in contrast, incubation of slices from FMRP-null mutant mice in a GABAA receptor antagonist resulted in brief bursts that subsequently evolved into electrographic seizures similar to those induced by the mGluR agonist, DHPG, in slices from wild-type littermates. Once epileptogenesis had occurred, antagonists of either mGluR1 or mGluR5 suppressed these recurrent seizures, just as was found with DHPG-evoked seizures in slices from normal animals. It seems plausible that epileptogenesis in the slices from the FMRP-null mutants was mediated by synaptically released glutamate-mediated activation of group 1 mGluRs; however, experiments addressing whether mGluR 1 or 5 antagonists prevented epileptogenesis were not reported.

It remains to be determined whether activation of group 1 mGluRs contributes to epileptogenesis in in vivo models of epileptogenesis. The striking changes found in group 1 mGluR-evoked hydrolysis of phosphoinositide in animal models of epileptogenesis heighten the interest in this question. For example, a single electrographic seizure evoked by a brief, low-intensity electrical stimulation of rat hippocampus is sufficient to induce dramatic increases in mGluR-induced hydrolysis of phosphoinositide in slices isolated from hippocampus (107). This increased mGluR-induced phosphoinositide hydrolysis is associated with increased mRNA and protein expression of mGluR1, but not mGluR5, in both the kindling and kainate models (108). Indeed, increased expression of mGluR1 is evident in the molecular layer of the dentate gyrus in a pattern corresponding to the dendrites of the granule cells in the kindling and kainate status epilepticus models, as well as in hippocampi excised from patients with refractory epilepsy (109). A preliminary report using antisense oligonucleotides raises the possibility that reduced expression of mGluR1 in hippocampus results in a slower rate of hippocampal kindling (110). That said, detailed studies examining epileptogenesis in the in vivo models, by using subtype-selective antagonists of mGluR1 and mGluR5 and also study of mice carrying null mutations of mGluR1 and mGluR5, are needed to clarify the putative roles of these receptors in limbic epileptogenesis in vivo.

Putative cellular and molecular mechanisms

The molecular signaling cascades activated by mGluR1 and mGluR5 are well understood. Both of these GPCRs act through Gαq/11 and phospholipase C–β1 (PLCβ1). Agonist binding of mGluR1 and mGluR5 leads to dissociation of the heterotrimeric G protein into the Gα and Gβγ subunits. The Gα subunit then activates PLCβ1 and thereby elicits signaling cascades mediated by both inositol 1,4,5-trisphosphate (IP3) with Ca2+ and diacyglycerol–protein kinase C (PKC). The extracellular signal-regulated kinase (ERK) mitogen-activated protein kinase (MAPK) is also activated by mGluR1 and mGluR5. In addition, activation of mGluR1 and mGluR5 stimulates G protein–independent signaling cascades. mGluR1 signals through a G protein–independent signaling pathway mediated by Src family kinase activation (111), and JNK is activated by group 1 mGluR through transactivation of epidermal growth factor receptors (EGFRs) in primary cultures of striatal neurons (112).

This knowledge of the mGluR1 and mGluR5 signaling pathways provides a valuable context with which to dissect the cascade of molecular events underlying DHPG-evoked epileptogenesis (3). In addition, the in vitro nature of this model, together with its localization to a well-understood local circuit in the CA3 region of hippocampus, provides powerful advantages for elucidating the cellular consequences of activation of this signaling pathway. Application of either of two structurally distinct inhibitors of mRNA translation, cycloheximide and anisomycin, prevents the appearance of DHPG-evoked seizures altogether or permits the appearance of DHPG-evoked seizures for a short time, followed by their disappearance (106). Application of two structurally distinct MEK inhibitors also blocks epileptogenesis; by contrast, application of these inhibitors after induction of epileptogenesis fails to inhibit seizures. Neither DHPG-induced ERK activation nor DHPG-induced epileptogenesis was altered by inhibition of PKC activity in hippocampal slices, which suggests that the PKC signaling pathway (which is activated by mGluR) is not required for epileptogenesis in this model. In contrast, the tyrosine kinase inhibitor genistein—but not its inactive analog, gensitin—prevented both DHPG-induced ERK activation and epileptogenesis. Among tyrosine kinases, Src family kinases in particular may be required for epileptogenesis, because PP2 (4-amino-5-(4-chlorophenyl)-7-(t-butyl)pyrazolo[3,4-D]pyrimidine), an inhibitor of these kinases, prevented epileptogenesis. Finally, prevention of epileptogenesis by a PLC inhibitor and the inability of DHPG to induce epileptogenesis in slices from PLCβ1 null mutant mice (PLCβ1−/− mutant mice) implicate PLCβ1 signaling as a critical downstream component of the mGluR pathway (106)

Additional experiments sought to elucidate how the group 1 mGluR–induced molecular signaling cascade (Fig. 2) might explain the cellular events that underlie DHPG-induced epileptogenesis in this model. Analyses of the effects of bath-applied DHPG using intracellular recordings of CA3 pyramidal cells in slices acutely isolated from wild-type and mutant mice provide the lynchpin of an appealing hypothesis (113). Bath application of DHPG induces recurrent depolarizations in CA3 pyramidal cells despite inhibition of most excitatory and inhibitory synaptic inputs; these depolarizations, which last several seconds and are accompanied by high-frequency firing of action potentials, are interspersed with intervals of equivalent duration in which the neuron is silent. This pattern of firing in an individual CA3 pyramidal cell is strikingly reminiscent of the pattern of firing in populations of CA3 pyramidal cells during seizures after DHPG-evoked epileptogenesis. A cellular event that appears critical to generation of this firing pattern is a DHPG-evoked voltage-dependent cationic conductance named ImGluR(V) for type I mGluRs of layer V pyramidal neurons. The DHPG-induction of ImGluR(V), the recurrent depolarizations with high-frequency firing in single cells, and the seizures in populations are all prevented by a PLC inhibitor and are not seen in cells from PLCβ1−/− mutant mice. Together, the data support the conclusion that DHPG-evoked ImGluR(V) in CA3 pyramidal cells is the elementary event underlying the seizures (Fig. 3). The occurrence of this event in individual CA3 pyramidal cells combined with the extensive recurrent excitatory synaptic connections among these neurons provides a parsimonious and appealing explanation for DHPG-evoked epileptogenesis in populations of CA3 pyramidal cells.

Fig. 3.

Group I mGluR activation promotes epileptogenesis in CA3 pyramidal neurons. Glutamate released from glutamatergic afferents is mimicked by bath-applied DHPG, a specific agonist for group I mGluRs, which activates mGluR1/5 on dendritic spines of CA3 pyramidal neurons and leads to dissociation of the heterotrimeric G protein into Gα and Gβγ subunits. The Gα subunit activates PLCβ, which hydrolyzes phosphatidylinositol 4,5-bisphosphate (PIP2) into IP3 and diacyglycerol (DAG), which results in release of Ca2+ from ER stores and increased activity of PKC, respectively. Src and ERK1/2 MAPK are also activated upon mGluR1/5 activation. Gβγ leads to reduction of background K+ conductance after mGluR1/5 activation. Activation of PLCβ contributes to the activation of ImGluR(V). Whether ERK1/2 and Src are required for generation of ImGluR(V) is not clear. Reduced background K+ conductance and increased voltage-dependent inward current ImGluR(V) enhance intrinsic neuronal excitability, which is a plausible mechanism underlying mGluR-induced epileptogenesis. The mGluR-induced protein synthesis is also implicated in this model.

Unanswered questions

The mGluR model of epileptogenesis provides a detailed molecular signaling cascade, together with a plausible cellular and network mechanism. The great promise of this model in turn necessitates answers to a series of questions. (i) Can epileptogenesis in this local circuit be induced by endogenously released glutamate, rather than bath application of the group 1 mGluR agonist, DHPG? (ii) Does epileptogenesis in any in vivo model utilize this mechanism? The striking increase in mGluR activation of phosphoinositide hydrolysis in brain slices acutely isolated from kindled animals strengthens the likelihood that this signaling pathway may contribute to epileptogenesis in the kindling model. Yet the finding that ERK MAPK is required for epileptogenesis in the DHPG model, but not in the in vivo pilocarpine model (114), raises concerns as to the in vivo relevance of the DHPG model. The paradoxical finding that PLCβ1−/− mutant mice that are resistant to DHPG-induced epileptogenesis are themselves epileptic raises additional concerns. (iii) What is the cellular localization of mGluR1 and mGluR5 in this local circuit within the CA3 region? Although simple by CNS standards, the anatomy of this local circuit is nonetheless complicated, because it consists of extensive recurrent excitatory synaptic connections among the principal (CA3 pyramidal) cells and various interneurons mediating both feedforward and feedback inhibition. In addition, the potential contribution of astrocytes, particularly in view of their expression of group 1 mGluRs, must be considered. Understanding the DHPG model requires knowing on which cell population(s) the mGluR1 and mGluR5 critical for epileptogenesis reside and what is their subcellular locale. (iv) What is the molecular identity of the channel(s) underlying ImGluR(V)? And what is its cellular and subcellular locale, and what are the signaling mechanisms by which mGluR5 activation leads to its activation?

Perspective

Like epileptogenesis triggered by activation of the NMDA and AMPA receptors, this in vitro model of epileptogenesis induced by exogenous application of a group 1 mGluR agonist likely requires an increase of intracellular Ca2+. In contrast to the ionotropic glutamate receptors in which Ca2+ enters the cell through the receptor itself, here, the increased intracellular Ca2+ is likely due to release from stores in the endoplasmic reticulum (ER) in response to IP3, which is generated by activated PLC.

Neurotrophin Receptors and Epileptogenesis

The mammalian neurotrophins consist of four small secreted proteins, nerve growth factor (NGF), brain-derived neurotrophic factor (BDNF), and neurotrophins 3 and 4 (NT-3 and NT-4, respectively). The genes encoding these four proteins share sequence similarity, a consequence of their sharing a common ancestral gene. The neurotrophins activate three different members of the tropomyosin-related kinase (Trk) family of receptor tyrosine kinases. These three receptors exhibit distinct affinities for different neurotrophins, with NGF activating TrkA, BDNF and NT-4 activating TrkB, and NT-3 predominantly activating TrkC (115). Neurotrophin binding to Trk receptors results in receptor dimerization and kinase activation. There are several conserved tyrosines in the intracellular domain of each Trk receptor, some of which are present in the autoregulatory loop of the kinase domain. Phosphorylation of the tyrosines in the autoregulatory loop activates kinase activity, whereas phosphorylation of the other residues promotes signaling by creating docking sites for adaptor proteins that couple these receptors to intracellular signaling cascades. These cascades include the Ras-ERK pathway, the phosphatidylinositol-3-OH kinase (PI 3-kinase)-Akt pathway, and the PLCγ1 pathway. The following section reviews the role of neurotrophins in epileptogenesis.

Pharmacological and genetic evidence

The discovery that limbic seizures induce a striking increase in the amount of NGF mRNA in the dentate granule cells of hippocampus of adult rats (116) suggested that this neurotrophin may serve some activity-regulated function in the mature nervous system. Unlike NGF, BDNF is widely expressed in neurons of the normal adult nervous system (117). However, like NGF, the expression of BDNF is increased as a consequence of seizures (118120). Together with emerging evidence implicating BDNF expression in structural and functional neuronal plasticity (121), this provided a strong rationale for probing the effects of BDNF and TrkB in epileptogenesis. Highly specific small-molecule antagonists of TrkB (or TrkA or TrkC) are not available, necessitating study of genetic perturbations. Neonatal mice carrying null mutations of either BDNF or TrkB die, which further complicates determining the role of these proteins in the adult nervous system. This led to studies of epileptogenesis in mice heterozygous null for BDNF that revealed a marked reduction in the rate of kindling development; this finding provides the first evidence that endogenous BDNF promoted epileptogenesis (122). Results of this genetic perturbation were confirmed and extended by a cumbersome but specific pharmacological approach involving intraventricular infusion of Trk receptor bodies, recombinant proteins containing the ligand-binding domain of distinct Trk receptors fused to the Fc portion of human immunoglobulin IgG1; these proteins can sequester and limit the action of distinct neurotrophins. Intraventricular infusion of TrkB-Fc inhibited development of kindling, whereas infusion of TrkA-Fc or TrkC-Fc did not (123). This implied that signaling through the TrkB receptor promoted epileptogenesis in this system and that the effects of the BDNF heterozygous null mutant mice could not be due solely to a consequence of reduced BDNF expression in the developing nervous system. These studies further underscored the selectivity of the TrkB receptor relative to the TrkA and TrkC receptors. The limited penetration of brain parenchyma by the TrkB receptor body (a protein of ~100 kD), together with uncertainty as to how much endogenous BDNF was actually scavenged, led to further examination of the role of BDNF and TrkB signaling in epileptogenesis (124). To circumvent the neonatal lethality of TrkB null mutations, these studies (124) used a conditional approach, in which either BDNF or TrkB alleles were selectively eliminated in CNS neurons. In the kindling model, only a modest inhibition of epileptogenesis was detected in mice homozygous null for BDNF, whereas behavioral evidence of epileptogenesis was eliminated altogether in mice homozygous null for TrkB (124). Although epileptogenesis was eliminated, seizures of equivalent duration and intensity could be evoked by electroshock in mice heterozygous and homozygous null for TrkB in comparison with wild-type mice. Thus, the plastic response of epileptogenesis was selectively eliminated in the TrkB null mutant mice, which implies that TrkB was an absolute requirement for epileptogenesis in the kindling model.

The elimination of epileptogenesis in the TrkB conditional null mutants implies that, in the kindling model, a signaling pathway or pathways activated by TrkB promote epileptogenesis. The development of mice carrying a point mutation of the Shc binding site in the TrkB gene provided an opportunity to test this hypothesis. In this mutation, substitution of phenylalanine for tyrosine at residue 515 of TrkB (Y515F) disrupts the binding of the Shc adaptor protein to TrkB and abolishes Shc site–mediated downstream signaling events (125). Epileptogenesis, as assessed by the development of kindling, did not differ between wild-type and TrkBshc mutant mice, which demonstrated that signaling through the Shc pathway downstream from TrkB is not required for epileptogenesis (126). The potential contribution of PI 3-kinase and PLCγ pathways remains to be elucidated.

Together, the data from genetic and pharmacological studies designed to limit TrkB activation are mutually reinforcing and provide consistent, strong support for the conclusion that limiting activation of TrkB inhibits epileptogenesis in the kindling model. In contrast, perturbations aimed at increasing TrkB activation have yielded conflicting results with respect to epileptogenesis. Transgenic overexpression of BDNF results in spontaneous seizures and enhanced sensitivity to kainate-evoked seizures, which, together, suggest that enhanced activation of TrkB is sufficient to induce epileptogenesis in mice (127). Consistent with these findings, intrahippocampal infusion of BDNF-stimulated limbic seizures in a subset of rats (128). By contrast, long-term infusion of BDNF into hippocampus inhibited the rate of hippocampal kindling development and reduced the duration of electrographic seizures (129). Likewise, intraventricular infusion of BDNF partially inhibited development of kindling (130). One potential limitation of these pharmacological experiments may lie in the difficulty of accurately replicating the spatiotemporal pattern of release of endogenous BDNF with these methods of BDNF application. Indeed, one explanation for these paradoxical findings may lie in the fact that prolonged exposure to high concentrations of BDNF can suppress TrkB receptor signaling and can reduce TrkB mRNA and protein levels in vitro (131), which raises the possibility that infusion of BDNF might actually result in a net reduction of TrkB signaling. Consistent with this suggestion, a 6-day infusion of BDNF into adult hippocampus decreases the abundance of full-length TrkB receptor by 80% (131). If the long-term infusion of BDNF in the kindling studies cited above also led to reduced TrkB content or responsiveness to exogenous and endogenous BDNF, then the observed reduction of the rate of kindling development is consistent with the genetic perturbations of BDNF and TrkB and with the pharmacology of using TrkB receptor bodies.

Putative cellular and molecular mechanisms

A clue to some of the cellular and molecular mechanisms by which TrkB activation may promote limbic epileptogenesis emerged from evidence that enhanced activation of TrkB occurred in the hippocampus during limbic epileptogenesis. Activation of Trk receptors involves phosphorylation of specific tyrosine residues in the cytoplasmic domain; therefore, the availability of antibodies that selectively recognize the phosphorylated form of Trk receptors (pTrk) (132) permits an immunohistochemical, as well as a biochemical, assessment of Trk receptor activation. This led to the discovery of increased pTrk immunoreactivity in the mossy fiber pathway of the hippocampus in both rats and mice in several in vivo models of limbic epileptogenesis (including kainate, kindling, and others) (126, 133, 134). This increase begins about 12 hours after occurrence of seizure, peaks at 24 hours, and returns to normal by 1 week thereafter (133). The development of mice carrying a point mutation of the Shc-binding site (Y515F) in the TrkB gene (TrkBshc) described above provided an opportunity to test the hypothesis that TrkB is the neurotrophin receptor undergoing phosphorylation. The epileptogenesis-associated increase of pTrk immunoreactivity was eliminated in a gene dose–dependent fashion in TrkBshc mutant mice, which implies that TrkB in particular was the neurotrophin receptor exhibiting increased phosphorylation (126). This conclusion was reinforced by immunoprecipitation and Western blot experiments that demonstrated increased pTrkB content in hippocampus; no change in the amount of TrkB itself was found, which implies that the increased pTrkB represented a posttranslational modification of TrkB (124). Together with the genetic and pharmacological evidence implying that TrkB is necessary for limbic epileptogenesis in the kindling model, this surrogate measure of TrkB activation raises the possibility that enhanced TrkB activation is a key signaling event mediating limbic epileptogenesis.

Earlier biochemical and immunohistochemical experiments revealed a striking increase in the amount of BDNF protein in the mossy fiber pathway in multiple models of limbic epileptogenesis, and the timing of the peak increase in BDNF coincided with the timing of increased amounts of pTrkB (119, 135). From this spatial and temporal concordance came the hypothesis that increased amounts of BDNF released from mossy fiber terminals bound to and activated TrkB, which led to the increased pTrkB detected in the mossy fiber pathway. Contrary to this hypothesis, increased pTrkB was observed in the mossy fiber pathway during epileptogenesis in conditional BDNF null mutant mice, mice in which elimination of BDNF from the dentate granule and CA3 pyramidal cells was virtually complete (124). The mechanism mediating TrkB activation in the BDNF null mutants is not clear. Both NT-4 and a modest (40%) homeostatic increase of hippocampal NT-3 that occurred in these mice may have contributed, although the epileptogenesis-induced rise in pTrkB occurs despite elimination of NT-4 (136). Additional possibilities include residual BDNF that may have escaped detection or transactivation of the TrkB receptor through a G protein–coupled receptor such as an adenosine 2A receptor (137).

Uncertainty as to the mechanism mediating the increased pTrkB during epileptogenesis notwithstanding, its localization to the mossy fiber pathway provides an anatomic locale for investigating functional and structural changes resulting from TrkB activation that may be instructive for limbic epileptogenesis. Identifying such plastic changes is more difficult, because both kindling and kainate status epilepticus (the models in which the increased pTrkB has been identified) are in vivo models. Moreover, identifying the relevant plastic changes in the in vivo models requires first elucidating the cellular and subcellular locale of the epileptogenesis-induced increase of pTrkB within the mossy fiber pathway; this has yet to be accomplished.

Bath application of BDNF combined with low-frequency stimulation of the dentate hilus results in multiple population spikes and spreading depression of CA3 pyramidal cells in acutely isolated rat hippocampal slices, demonstrating that BDNF can induce increased excitability of CA3 pyramidal cells, either directly or indirectly, and that this could be one consequence of increased pTrkB in the mossy fiber pathway (138).

The uncertainties as to the precise cellular and subcellular locale of the increased pTrk immunoreactivity notwithstanding, it seems reasonable to consider some potential epileptogenesis-related plasticities in this region that might be a consequence of enhanced activation of TrkB (Fig. 4). TrkB activation has been implicated in LTP of excitatory synapses at both the Schaffer collateral–CA1 synapse and perforant path–dentate granule cell synapse (139142). Whether TrkB is required for LTP of the mossy fiber–CA3 pyramidal cell synapse has not been reported; if TrkB activation does promote LTP of this synapse, then perhaps the epileptogenesis-induced TrkB activation induces LTP of the mossy fiber–CA3 pyramidal cell synapse. High-frequency stimulation of mossy fibers induced LTP of the mossy fiber–CA3 pyramidal cell synapse in hippocampal slices isolated from vehicle-treated rats but not after status epilepticus (143). Nevertheless, high-frequency stimulation of associational-commissural afferents to CA3 pyramidal cells induced LTP of these synapses to an equivalent extent in slices from rats after status epilepticus and from vehicle-treated rats. These findings are consistent with the idea that the mossy fiber–CA3 pyramidal cell synapse undergoes LTP in vivo after status epilepticus and that this occludes induction of LTP of this synapse in the isolated slices. Given the evidence implicating TrkB in LTP of excitatory synapses elsewhere in the hippocampus, this is one potential consequence of activation of TrkB during limbic epileptogenesis. If so, this likely involves TrkB-mediated activation of phospholipase Cγ (PLCγ) (144).

Fig. 4.

TrkB signaling promotes epileptogenesis in kindling. Upon binding of ligands BDNF and NT-4, TrkB receptors undergo dimerization and increased intrinsic kinase activity. TrkB activation induces tyrosine autophosphorylation within the intracellular domain, including Y515 and Y816, providing docking sites for signaling proteins Shc and PLCγ1, respectively. Subsequently, Ras-Raf-ERK1/2 MAPK, PI3K/Akt, and IP3/Ca2+ cascades are activated. ERK1/2 MAPK is not required for LTP or for kindling. Activation of PLCγ is essential for CA1 LTP, but its role in epileptogenesis has yet to be tested. TrkB activation can reduce GABA-mediated synaptic inhibition by reducing expression of KCC2 and also by increasing the activity of protein phosphatase PP2A, thereby enhancing internalization of GABAA receptors (GABAARs). TrkB activation may also promote mossy fiber sprouting of dentate granule cells.

In addition to the possibility of enhanced efficacy of excitatory synapses, TrkB activation can limit GABA-mediated synaptic inhibition; in particular, it reduces GABAA receptor–mediated responses of CA1 pyramidal cells (145, 146). One mechanism by which TrkB activation could compromise GABAA receptor–mediated responses is through reduced expression of the neuron-specific K+-Cl cotransporter, KCC2. That is, KCC2 function is required to maintain the Cl gradient so that GABAA receptor activation mediates hyperpolarizing (inhibitory) responses. BDNF-mediated activation of TrkB activation reduces KCC2 mRNA and protein expression in principal neurons of rat hippocampal explant cultures, a reduction associated with impaired neuronal Cl extrusion capacity (147). Moreover, reduced expression of KCC2 mRNA and protein is evident in hippocampal neurons after induction of kindling in vivo (147) Together, these findings raise the possibility that activation of TrkB during epileptogenesis in vivo may result in reduced expression of KCC2, which in turn impairs GABAA receptor–mediated synaptic inhibition.

In addition to regulating synaptic plasticity, enhanced activation of TrkB in the mossy fiber pathway could modify the intrinsic properties of CA3 pyramidal cells. As noted above, one signaling cascade activated by TrkB is PLCγ, the activation of which leads to generation of diacyglycerol and IP3, and thereby activation of PKC. Indeed, the dendrites of CA1 pyramidal cells in hippocampal slices isolated from rats rendered epileptic in the pilocarpine model exhibit increased excitability (148). Decreased availability of A-type K+ channels, apparently due in part to enhanced PKC activity, contributed to the increased dendritic excitability. Thus, plasticity of intrinsic neuronal excitability, leading to increased excitability of these neurons, must be considered a potential consequence of TrkB activation.

Apart from plasticity of synapses and intrinsic properties of neurons, enhanced activation of TrkB might contribute to some of the structural plasticities of the dentate granule cells observed in limbic epileptogenesis. Indeed, overexpression of BDNF, but not NGF, in dentate granule cells of hippocampal explant cultures is sufficient to induce increased axonal branching and sprouting of basilar dendrites in a way that could be inhibited by the neurotrophin receptor antagonist K252a (149), which implies that enhanced activation of TrkB is sufficient to induce these changes. The signaling pathway downstream from TrkB mediating these plasticities is unknown. Whether it will be proved that TrkB activation is required for induction of these plasticities in vivo remains to be seen.

Perspective

Given the pivotal role of Ca2+ in signaling underlying epileptogenesis triggered by activation of glutamate receptors, it seems likely a rise of intracellular Ca2+ underlies epileptogenesis mediated by activation of TrkB. This suggests that PLCγ signaling downstream of TrkB mediates the pro-epileptogenic consequences of TrkB activation. If correct, as for the mGluRs, the source of the increased intracellular Ca2+ is likely to be the ER.

TrkA and TrkC and p75 and limbic epileptogenesis

The inhibitory effects of intraventricular infusion of TrkB-Fc, but not TrkA-Fc, suggest that neither NGF nor TrkA contributes to epileptogenesis in the kindling model (123). This conclusion conflicts with previous studies in which intraventricular infusion of NGF antisera reduced the rate of kindling development in rats (150, 151). Likewise, infusion of a peptide mimic of NGF designed to prevent binding to TrkA partially inhibited development of kindling (152). Interpretation of these results must be tempered by uncertainty as to the specificity of these reagents. For example, the antisera in these studies were raised against the entire NGF molecule; because neurotrophin family members share about 50% sequence identity, such antisera may contain antibodies against other neurotrophins, including BDNF. Indeed, the NGF antisera and peptides were shown to cross-react to some extent with BDNF and NT-3 in in vitro assays (152, 151), yet these cross-reacting antibodies were not removed by preabsorption with BDNF before use in vivo.

The lack of effect of TrkC-Fc (123) argues against the contribution of either NT-3 or TrkC to epileptogenesis in the kindling model. This result contrasts with the reduced rate of kindling development reported in NT-3 heterozygous null mutant mice. One explanation for these divergent findings might relate to the developmental consequences of reduced levels of NT-3 protein in the NT-3 heterozygous null mutant mice. The authors also found altered seizure-induced BDNF, TrkB, and TrkC gene expression in the NT-3 heterozygous null mutant mice, so it is possible that the observed effects were an indirect consequence of the reduction in NT-3. To our knowledge, there are no reports examining effects of intraventricular infusion of antibodies against NT-3.

The role of an additional neurotrophin receptor, p75, in limbic epileptogenesis is incompletely understood. The neurotrophins are synthesized as precursors, termed "pro-neurotrophins," which undergo proteolytic cleavage into mature neurotrophins, the forms of neurotrophin referred to in the preceding paragraphs. The pro-neurotrophins exhibit high affinity for p75, a member of the TNFα receptor family, but exhibit low affinity for Trks A, B, and C (153). By contrast, the mature forms of the neurotrophins exhibit high affinity for the respective Trk receptor, but low affinity for p75. The expression of p75 is increased in select populations of hippocampal and entorhinal cortical neurons in the pilocarpine model (154); the colocalization of the increased p75 immunoreactivity with Tunel reactivity suggests that increased p75 promotes the apoptotic death of these neurons following status epilepticus. This suggestion is consistent with the presence of a type 2 death domain in p75 and evidence that p75 activation promotes apoptosis (155). That said, whether and how p75 contributes to epileptogenesis per se awaits study of epileptogenesis in genetically modified mice.

Ca2+-Regulated Enzymes and Epileptogenesis

Calcium–calmodulin kinase II

Increased intracellular Ca2+ is one of the key steps in initiating neuronal signal transduction cascades (156). One consequence of increased intracellular Ca2+ can be activation of CaMKII. CaMKII activation depends on conformational changes induced by Ca2+-calmodulin binding. These changes activate CaMKII and increase autophosphorylation of Thr286 (157), which in turn makes the kinase partially Ca2+-calmodulin independent (158). However, CaMKII activation is defined not only by an enzymatic (autophosphorylation) step, but also by steps involving translocation and localization within a neuron. Upon NMDA receptor activation and Ca2+ entry, CaMKII is activated and translocates from the cytosol to excitatory synapses (159). This particular translocation event serves to bring the kinase closer to specific substrates within the postsynaptic density (160).

Of the Ca2+-dependent enzymes studied to date, CaMKII is the best understood as far as its role in epileptogenesis. Mice carrying a null mutation of the α subunit of CaMKII exhibit limbic epilepsy (161). The increased excitability appears to depend on gene dose, in that heterozygotes exhibit a trend toward increased duration of stimulus-evoked seizures in comparison with wild-type littermates. This observation has been supported by studies of cultured hippocampal neurons treated with antisense oligonucleotides to α-CaMKII (162). Again, by reducing the levels of α-CaMKII, spontaneous epileptic discharges emerge in these cultures. Together, these studies provide strong evidence that reduced expression of α-CaMKII is sufficient to induce limbic epilepsy.

The genetic and pharmacological evidence cited above raises the question as to whether reductions of α-CaMKII can be identified in models of limbic epilepsy in wild-type mice or rats. Interestingly, decreased activity of CaMKII has been a consistent finding in many models. Reductions of Ca2+-calmodulin–stimulated phosphorylation of synaptic membrane proteins were identified in hippocampus of kindled rats, reductions that persisted 8 weeks after induction of kindling (163). Reductions of CaMKII activity have also been identified after electroconvulsive seizures (164) and repeated audiogenic seizures (165). Likewise, decreases of CaMKII activity, as measured by Thr286 phosphorylation, phosphorylation of exogenous substrates, or both, have been identified in in vivo models of status epilepticus induced with either pilocarpine or kainate (162, 166168). This decrease in enzymatic activity has also been observed in the low-magnesium model in vitro (162); here, the reduced CaMKII activity can be prevented by NMDA receptor antagonists that prevent epileptogenesis. Finally, it has been demonstrated that CaMKII undergoes a change in its subcellular distribution after epileptiform activity. In these studies, seizures induced by pentylenetetrazol (169) or ECS (168) caused a translocation of CaMKII protein from the synaptic to the cytosolic fraction. The redistribution of CaMKII would effectively cause a cessation of its signaling role at the postsynaptic density (PSD) component of the excitatory synapses and distance it from key synaptic substrates (159).

The biochemical evidence of reduced CaMKII activity, together with the compelling genetic and pharmacological evidence, supports the idea that reduced CaMKII activity per se is sufficient to effect molecular and cellular changes that render the brain epileptic. In addition, its translocation from synapse to cytosol could further reduce its effective enzymatic activity by distancing the enzyme from its substrates. Moreover, reductions of CaMKII content may result in loss of a pivotal PSD scaffold. The structural role of CaMKII is suggested by its ability to self-associate and to bind to a number of postsynaptic proteins (159). Another possibility is that decreased abundance of CaMKII protein permits increased concentrations of calmodulin available for other calmodulin-dependent enzymes such as calcineurin.

Calcineurin

Calcineurin is a Ca2+-calmodulin–regulated serine-threonine phosphatase consisting of a catalytic (CnA) and a regulatory (CnB) subunit. Both calcineurin A and B subunits comprise several isoforms coded by different genes or generated by alternative splicing. Three separate genes encode the catalytic subunit of calcineurin in mammalian cells, CnAα, CnAβ, and CnAγ.

When activated by Ca2+-calmodulin binding, calcineurin dephosphorylates various substrates. The Ca2+-calmodulin binding domain is on CnA; Ca2+-calmodulin binding relieves the autoinhibitory domain, which binds to the catalytic site in the absence of Ca2+-calmodulin and inhibits the enzyme. Calcineurin activity can be blocked in vitro or in vivo by the immunosuppressive drugs cyclosporin A (CsA) and FK506.CsA and FK506 form complexes with endogenous cyclophilin and FKBP12 (FK506 binding protein 12), respectively, and these complexes bind the catalytic subunit of calcineurin.

The effects of FK506 on epileptogenesis in the kindling model in rats have been known for over a decade (170). These studies demonstrate that FK506-treated animals develop kindling at a slower rate than controls (170, 171) and do not display greater than a class 3 seizure even after 15 stimulations (171). These limited pharmacological data suggest that calcineurin activity is required for epileptogenesis.

Several biochemical studies have demonstrated activation and translocation of calcineurin during epileptogenesis in vivo (170, 172), raising the possibility that enhanced calcineurin signaling could promote epileptogenesis. In the pilocarpine model (172), there is an increase in calcineurin activity 1 hour after status epilepticus that is not due to increased calcineurin protein, but rather to an increase in enzyme catalysis and substrate affinity. NMDA receptor activation is critical to the increase in calcineurin signaling. Increased calcineurin protein content has also been identified in a membrane fraction in the kindling model (170).

There is evidence of increased calcineurin activity and protein in the neonatal hypoxic seizure model, an effect dependent on activation of Ca2+-permeable AMPA receptors (93). Under these conditions, one of the molecular mechanisms by which calcineurin affects seizure susceptibility may be through the dephosphorylation and subsequent endocytosis of GABAA receptors. Another mechanism by which calcineurin might regulate neuronal excitability is through K+ channel dephosphorylation. An increase in intracellular Ca2+, either through activation of ionotropic glutamate receptors or through membrane depolarization, is sufficient to induce dephosphorylation and translocation of the Kv2.1–type K+ channel from the cell membrane. A decrease in the phosphorylated form of Kv2.1 is also evident in membrane fractions prepared from the brains of rats subjected to kainate status epilepticus. These Ca2+-dependent effects on Kv2.1 phosphorylation and localization appear to be mediated by calcineurin-dependent dephosphorylation (173). The predicted consequence of this Ca2+-dependent signaling cascade is increased intrinsic excitability of neurons.

Nonreceptor Tyrosine Kinases

Evidence implicating some nonreceptor protein tyrosine kinases in synaptic plasticity provides the rationale for studying their role in epileptogenesis. The abundance of the Src family kinases, c-Src and c-Fyn, in cortical and hippocampal neurons (174) contributes to extensive study of these enzymes in particular. Because Src family kinases share a common mechanism of activation and can phosphorylate many of the same substrates in neurons, there is potential for compensation among these protein tyrosine kinases (PTKs), potentially complicating interpretation of targeted mutations in mice. PTKs contain many regulatory domains in addition to a catalytic domain. Regulatory signals, such as phosphorylation of certain residues in the regulatory or catalytic domain, often modulate kinase activity by affecting domain-domain interactions. On activation, these PTKs are recruited to the plasma membrane or PSD, where they act on key substrates at the synapse.

Pharmacological and genetic evidence

The best-studied member of the Src family with respect to epileptogenesis is Fyn. Mice with a targeted deletion of fyn exhibit impaired epileptogenesis, as evidenced by a slower rate of kindling development (175). Complementary studies showed that transgenic mice expressing a constitutively active form of Fyn exhibit spontaneous seizures and are prone to sudden death (176). Mice overexpressing native Fyn do not exhibit spontaneous seizures, but they do show a faster rate of kindling. Furthermore, the Src kinase inhibitor PP2 reduces the frequency of epileptiform activity in a hippocampal slice model of epileptogenesis (177).

Putative cellular and molecular mechanisms

Analyses of both in vitro and in vivo models of epileptogenesis have provided direct evidence for activation of Src family kinases. Induction of epileptiform activity in hippocampal slices in vitro results in increased Src kinase activity as detected by increased catalysis of an exogenous substrate or increased phosphorylation of Tyr418 (Y418), a measure of activated Src (177). Analyses of status epilepticus in vivo revealed evidence of increased amounts of Fyn and activated Src in a PSD fraction of hippocampus (178180). The increases of Fyn and Src during epileptogenesis, together with the genetic and pharmacological evidence cited above, support the idea that enhanced activity of these Src family kinases may promote epileptogenesis.

The substrates of these Src family kinases that contribute to epileptogenesis are unclear. Interestingly, kainate-status epilepticus results in a marked increase of tyrosine phosphorylation of many hippocampal proteins (181). Transient increases of tyrosine phosphorylation of NR2A and NR2B in particular have been identified following status epilepticus induced by kainate or pilocarpine (178, 182). Given the enhanced NMDA receptor currents induced by Src-mediated tyrosine phosphorylation (183), the tyrosine phosphorylation of these subunits of the NMDA receptor would be expected to promote epileptogenesis.

Astrocytes and Epileptiform Activity

To this point, consideration of signaling pathways underlying epileptogenesis has centered on neurons and on functional and structural plasticities intrinsic to neurons. Although this reflects the focus of investigation over the past decade, investigation of the potential importance of astrocytes to epilepsy has a rich history in epilepsy research. Interest in astrocytes has resulted not only from the fact that astrocytes greatly outnumber neurons, but that astrogliosis—abnormal shape and perhaps increased numbers of astrocytes—is a prominent feature of Ammon’s horn sclerosis, the hippocampal pathology identified in 70% of patients with medically refractory temporal lobe epilepsy. This observation led to early studies investigating whether abnormalities of potassium homeostasis in gliotic brain may contribute to expression of seizures (184). Moreover, interest in the potential role of astrocytes in epileptogenesis has been increased by discoveries of the past decade, demonstrating that astrocytes dynamically contribute to signaling and information processing in the brain (185). In particular, the discoveries that glutamate released from neurons can activate metabotropic glutamate receptors on astrocytes, which in turn leads to release of glutamate from the astrocyte through a calcium-dependent process, provides a mechanism for dynamic cross-talk between neurons and astrocytes (186) (Fig. 5). Interest in astrocytes in epileptogenesis was further heightened by the discovery that uncaging calcium in a single astrocyte can activate NMDA receptor–mediated increases in intracellular Ca2+ concentration simultaneously in many neighboring neurons, a finding that provides a mechanism for synchronous activation of neurons (187). This is important because the synchrony of firing of neurons is a key feature of a seizure, and this discovery provides a heretofore unrecognized potential mechanism of effecting synchronous neuronal firing.

Fig. 5.

Cross-talk between neurons and astrocytes promotes synchronous epileptiform activity in CA1 pyramidal neurons. Neuronal activity (for instance, in response to picrotoxin) induces glutamate release from the presynaptic terminal of the neuron on the left. The excessive glutamate released escapes the synaptic cleft and activates group I mGluRs (mGluR5 in particular) of astrocytes, resulting in PLCβ activation, increased intracellular Ca2+ released from ER, and in turn, release of glutamate from astrocytes. Glutamate diffuses to neighboring neurons, where it activates extrasynaptic NMDA receptors and AMPA receptors (red), resulting in synchronized PDS of multiple CA1 pyramidal cells.

Together, these findings provided a strong rationale for studies aimed at determining whether glutamate released from astrocytes might evoke epileptiform activity in neighboring neurons (188). These studies combined whole-cell patch-clamp and field potential recordings of CA1 pyramidal cells with two-photon imaging and uncaging of calcium in acutely isolated rat hippocampal slices. The epileptiform activity consisted of a PDS lasting 250 ms or less that was recorded from CA1 pyramidal cells (Fig. 5). Perfusion of slices with each of four different convulsant agents induced PDS in neurons, thereby replicating findings by countless investigators. The amplitude and frequency of the PDS was dramatically reduced by APV and to a lesser extent by CNQX (6-cyano-7-nitroquinoxaline-2,3-dione), findings consistent with the notion that the PDS was triggered by activation of NMDA and AMPA receptors on the CA1 pyramidal cell, again replicating earlier work (189). Surprisingly, including TTX, which blocks sodium channel–dependent action potentials, with the convulsant agents only minimally inhibited the PDS, raising the possibility that a nonneuronal source of glutamate evoked the PDS. Indeed, laser-evoked release of caged Ca2+ in astrocytes was sufficient to induce the massive membrane depolarization component of a PDS in neighboring neurons, despite inclusion of TTX in the bath.

Taken together, these findings provide direct evidence that glutamate released from an astrocyte is sufficient to trigger a PDS in neighboring neurons. These findings suggest a positive-feedback model (Fig. 5) in which glutamate released by firing of neurons activates mGluRs intrinsic to astrocytes and leads to Ca2+-dependent release of glutamate and the ensuing NMDA receptor–mediated PDS of these neurons. Thus, dynamic signaling between neurons and astrocytes not only may promote PDS in vivo, but may contribute to additional epileptiform events, including actual seizures. This also provides a novel mechanism for the synchronization of neuronal firing in a seizure, a seminal feature of this pathologic event. Given the powerful role of NMDA receptors in epileptogenesis, it seems plausible that glutamate released from astrocytes may contribute to epileptogenesis in some settings. Indeed, it will be interesting to determine whether the enhanced mGluR-evoked phosphatidylinositol hydrolysis evident in brain slices isolated from in vivo models of epileptogenesis may reflect, in part, plastic changes intrinsic to astrocytes that promote enhanced mGluR-evoked glutamate release.

Perspective

Stages of epileptogenesis

Analyses of diverse models of epileptogenesis suggest that the temporal evolution of this process can be arbitrarily subdivided into two stages and that neuronal activity is important for both stages. One stage is "preseizure" and a second stage is "postseizure." The majority of studies of signaling underlying epileptogenesis have centered on models that use focal seizures as the inciting agent for epileptogenesis; these include all of the status epilepticus, kindling, and neonatal hypoxic seizure models in vivo, as well as the three-chamber model in vitro (Table 2). For example, the occurrence of focal seizures is critical to epileptogenesis in the kindling model (190, 191). Preventing seizure invasion of the secondary hippocampus by application of TTX to the connecting commissure in the three-chamber model prevents emergence of spontaneous, recurrent seizures (30). The causal role of seizure activity in epileptogenesis is also evident in status epilepticus models, because limiting the duration of evoked seizures prevents emergence of late-onset recurrent spontaneous seizures in the pilocarpine model (20, 21).

Table 1.

Basic terminology used in discussing epilepsy.

Table 2.

Models used to study epileptogenesis. Various in vitro and in vivo model systems have been used to study epileptogenesis. Table 2 provides a reference for the models discussed in the text, giving the names by which the models are identified and a brief description of the different systems.

Although these diverse postseizure models clearly establish a causal role for seizure activity in epileptogenesis, a key question relevant to the human condition is, what are the earlier events that lead to focal hyperexcitability that culminates in emergence of a focal seizure in the first place? Stated differently, what underlies the occurrence of the initial seizure? Given the diverse structural lesions and genetic causes of limbic epilepsy, numerous factors almost certainly contribute. One of these factors is likely to be some form of neuronal activity that is not actual seizure activity. For example, deafferentation of cerebral cortex is thought to be one factor that results in epilepsy after head injury, which has led to studies of the effects of surgical transection of axons connecting cortical and subcortical structures of rats while preserving pial circulation—–the "undercut cortex" model (Table 2) (192). In vitro electrophysiological analyses of cortical slices isolated weeks to months after surgical transection revealed that stimulation at the junction of white matter and layer 6 evoked interictal, epileptiform field potentials in undercut but not sham-operated controls. Neither spontaneous nor stimulation-evoked seizures were observed. Surprisingly, elimination of action potentials by application of TTX-soaked Elvax overlying undercut cortex prevented development of the hyperexcitability in slices (193), which therely implicated neuronal activity in this prefocal seizure model of epileptogenesis. With respect to the nature of the neuronal activity, studies of the Schaffer collateral model in vitro (28) provide a clue. They demonstrated that repeated tetanic stimulation of the Schaffer collateral afferents is sufficient to induce evoked-PDS or interictal bursts in the CA1 pyramidal cells (28). Together, these models examining mechanisms underlying early stages of epileptogenesis before emergence of focal seizures implicate some form of neuronal activity, likely including high-frequency firing to induce focal cortical hyperexcitability.

Ca2+ within dendritic spines and epileptogenesis

Both circumstantial and direct evidence support the idea that increased Ca2+within dendritic spines may serve as the mechanism by which abnormal neuronal activity initiates epileptogenesis. Increases of intracellular Ca2+ represent critical signaling events induced by activation of each of the neuronal receptors implicated in epileptogenesis, namely, the NMDA receptor, the Ca2+-permeable AMPA receptor, mGluRs 1 and 5, and TrkB. Direct evidence implicating a causal role for increased Ca2+ emerged from analyses of epileptogenesis in the model in which cultured hippocampal neurons are incubated for 3 hours in a Mg2+-free buffer. Fluorescent measurements revealed increased intracellular Ca2+ in these neurons during the 3-hour incubation in Mg2+-free buffer. Importantly, inclusion of the extracellular Ca2+ chelator, BAPTA, in the Mg2+-free buffer prevented epileptogenesis. Likewise, the fact that genetic or pharmacological inhibition of PLC prevented epileptogenesis in the mGluR model suggests a requirement for increased intracellular Ca2+; that is, PLC activation results in synthesis of inositol trisphosphate (IP3), which in turn activates IP3 receptors located on ER, resulting in release of Ca2+ from ER into the cytosol.

Dendritic spines represent a particularly attractive locale within which the pathological consequences of increased intracellular Ca2+may be initiated. The dendritic spine is a protrusion apposed to the presynaptic terminals of most excitatory synapses in the brain. These spines are predominantly localized to principal neurons. Each of the receptors implicated in epileptogenesis is richly expressed in PSDs, a complex of many proteins within dendritic spines that serves to localize receptors to the sites of neurotransmitter release. One consequence of the abnormal neuronal activity implicated in epileptogenesis is the synaptic release of endogenous ligands for the relevant cell surface receptors, namely, glutamate and the neurotrophins, BDNF and NT-3, the net effect of which should be elevated Ca2+ within the spine. Notably, the source of Ca2+ differs in these models and involves influx through ionotropic receptors in the case of NMDA and AMPA receptors, in contrast to presumed release from ER within a spine after mGluR- or TrkB-mediated activation of PLC and subsequent activation of IP3 receptors in ER. It seems likely that these distinct sources result in differences in the precise locale within the spine at which the rise of Ca2+ occurs, and the absolute amounts and kinetics of the increased Ca2+ may also differ. This, in turn, suggests that distinct signaling cascades might be engaged, perhaps converging on some common final pathways. One class of cellular mechanisms on which these signaling mechanisms within spines might converge is plasticity of both excitatory and inhibitory synapses.

Plasticity of excitatory and inhibitory synapses and epileptogenesis

Analyses of the Schaffer collateral model (28) advance LTP of excitatory synapses as an early cellular mechanism contributing to focal hyperexcitability. Not only did the excitatory synapses undergo LTP, but there was an associated depression of inhibitory synapses on the CA1 pyramidal cells, the latter associated with reduced responsiveness to applied GABA (28). Analyses of the cellular and molecular mechanisms in LTP paradigms have provided a framework for understanding the molecular mechanisms underlying these early events of epileptogenesis (see Fig. 1). In vivo studies provide support for the conclusion that LTP of excitatory synapses contributes to focal hyperexcitability before the emergence of focal seizures. For example, repeated, low-intensity tetanic stimulation of perforant path afferents, stimuli sufficient to induce LTP but not focal seizures, facilitated epileptogenesis in the kindling model (56), which suggests that LTP of the perforant path–granule cell synapse is one mechanism underlying focal hyperexcitability before the first seizure. A key question raised by these findings is what differences exist in the cellular and molecular events underlying LTP compared with epileptogenesis.

Calcium-regulated enzymes within dendritic spines and epileptogenesis

A diversity of enzymes associated with the PSD of dendritic spines can transduce the effects of Ca2+ into downstream signaling pathways. It seems plausible that these same enzymes may be effectively hijacked by pathological levels of intracellular Ca2+, with the net effect of epileptogenesis. Four such enzymes that have been implicated in epileptogenesis are CaMKII, calcineurin, and the nonreceptor tyrosine kinases Src and Fyn.

An intriguing picture has emerged with respect to the roles of CaMKII and calcineurin in particular. Each of these enzymes is regulated by both Ca2+ and calmodulin. Each of these enzymes is located at the PSD of excitatory synapses. Moreover, analyses of "postfocal seizure" models with genetic and pharmacologic tools reveal that reductions of CaMKII expression are sufficient to induce epileptogenesis, and biochemical studies reveal reduced expression of CaMKII during epileptogenesis. Conversely, analyses of postfocal seizure models reveal that pharmacologic inhibition of calcineurin activity impairs epileptogenesis, and biochemical studies reveal increased activity of calcineurin in these models. Together with their residing jointly in PSDs of dendritic spines, these findings raise the possibility of some common substrates for these enzymes and suggest that tipping the balance of their phosphorylation versus dephosphorylation in favor of net dephosphorylation of one or more of these substrates is sufficient to induce epileptogenesis (194). One attractive candidate is the β2/3 subunit of the GABAA receptor, in which dephosphorylation promotes endocytosis, removal of the receptor from the cell membrane, and resultant disinhibition (195).

It is noteworthy that the proposed role for CaMKII in epileptogenesis runs precisely opposite to its proposed role in LTP of the Schaffer collateral–CA1 synapse. That is, extensive evidence supports the conclusion that Ca2+ influx through the NMDA receptor in a dendritic spine of a CA1 pyramidal cell promotes activation of CaMKII, which in turn leads to insertion of GluR1 into the synaptic membrane and LTP of this synapse (196, 197). By contrast, phosphatase (here PP1) activity promotes LTD (198). Yet the occurrence of limbic epilepsy in mice carrying null mutations of α-CaMKII demonstrates that the absence of CaMKII is sufficient to cause limbic epilepsy. How might these apparently conflicting observations be reconciled? We suggest that activation of CaMKII does occur as a result of activity-driven activation of NMDA receptors and the consequent rise of intracellular Ca2+ during early (prefocal seizure) stages of epileptogenesis and that LTP ensues at synapses like the Schaffer collateral–CA1 synapse. In contrast to the preseizure activity, the greater increases and duration of increased intracellular Ca2+ that almost certainly accompany focal seizure activity may lead to a homeostatic reduction of CaMKII activity and expression, a maladaptive consequence of which is limbic epilepsy. Thus, different amounts of intraspinous Ca2+ arising from distinct forms of neuronal activity at different stages of epileptogenesis may have opposing consequences on activity of Ca2+-regulated enzymes and the phosphorylation state of key substrates within dendritic spines. Elucidating the key substrate(s) of these calcium-regulated enzymes and the signaling consequences of their phosphorylation state will facilitate testing this proposal.

Questions Arising

This review of the molecular signaling pathways underlying limbic epileptogenesis highlights the fact that most of the current insights have been derived from study of postseizure models. The question arises as to what aspect of limbic epilepsy in humans is addressed by these models. As noted above, insights derived from study of these models may not shed light on the mechanisms underlying emergence of the initial seizure, but rather on some of the events transpiring after the initial and subsequent seizures. Indeed, about one-third of individuals who experience a limbic seizure become unresponsive to available antiseizure medications. Because induction of repeated seizures in experimental animals alone is sufficient to result in progressive severity of the epileptic condition, including emergence of fatal seizures (26), it seems likely that seizures themselves represent one factor contributing to the progressive severity of the condition evident in the medically refractory cases clinically. If so, identification of the underlying molecular mechanisms may lead to small-molecule therapeutics that could limit progression and, thereby, maintain a condition in which the seizures can be controlled by available medications.

The evidence implicating the diverse signaling pathways in these models of limbic epileptogenesis raises the question as to the relative importance of each of them in the clinical condition. In part, this undoubtedly reflects the developmental stage; for example, the key role of Ca2+-permeable AMPA receptors is evident in the neonatal hypoxic seizure model, which in turn reflects susceptibility at a discrete time within postnatal development but not in an adult. With respect to the remaining pathways, implicating the group 1 mGluRs in limbic epileptogenesis awaits additional study of in vivo models with selective pharmacological and genetic approaches. With respect to the NMDA (199) and TrkB (200) receptor pathways, a number of interactions have been elucidated, but whether and how these interactions underlie limbic epileptogenesis remains to be determined.

References

  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
  41. 41.
  42. 42.
  43. 43.
  44. 44.
  45. 45.
  46. 46.
  47. 47.
  48. 48.
  49. 49.
  50. 50.
  51. 51.
  52. 52.
  53. 53.
  54. 54.
  55. 55.
  56. 56.
  57. 57.
  58. 58.
  59. 59.
  60. 60.
  61. 61.
  62. 62.
  63. 63.
  64. 64.
  65. 65.
  66. 66.
  67. 67.
  68. 68.
  69. 69.
  70. 70.
  71. 71.
  72. 72.
  73. 73.
  74. 74.
  75. 75.
  76. 76.
  77. 77.
  78. 78.
  79. 79.
  80. 80.
  81. 81.
  82. 82.
  83. 83.
  84. 84.
  85. 85.
  86. 86.
  87. 87.
  88. 88.
  89. 89.
  90. 90.
  91. 91.
  92. 92.
  93. 93.
  94. 94.
  95. 95.
  96. 96.
  97. 97.
  98. 98.
  99. 99.
  100. 100.
  101. 101.
  102. 102.
  103. 103.
  104. 104.
  105. 105.
  106. 106.
  107. 107.
  108. 108.
  109. 109.
  110. 110.
  111. 111.
  112. 112.
  113. 113.
  114. 114.
  115. 115.
  116. 116.
  117. 117.
  118. 118.
  119. 119.
  120. 120.
  121. 121.
  122. 122.
  123. 123.
  124. 124.
  125. 125.
  126. 126.
  127. 127.
  128. 128.
  129. 129.
  130. 130.
  131. 131.
  132. 132.
  133. 133.
  134. 134.
  135. 135.
  136. 136.
  137. 137.
  138. 138.
  139. 139.
  140. 140.
  141. 141.
  142. 142.
  143. 143.
  144. 144.
  145. 145.
  146. 146.
  147. 147.
  148. 148.
  149. 149.
  150. 150.
  151. 151.
  152. 152.
  153. 153.
  154. 154.
  155. 155.
  156. 156.
  157. 157.
  158. 158.
  159. 159.
  160. 160.
  161. 161.
  162. 162.
  163. 163.
  164. 164.
  165. 165.
  166. 166.
  167. 167.
  168. 168.
  169. 169.
  170. 170.
  171. 171.
  172. 172.
  173. 173.
  174. 174.
  175. 175.
  176. 176.
  177. 177.
  178. 178.
  179. 179.
  180. 180.
  181. 181.
  182. 182.
  183. 183.
  184. 184.
  185. 185.
  186. 186.
  187. 187.
  188. 188.
  189. 189.
  190. 190.
  191. 191.
  192. 192.
  193. 193.
  194. 194.
  195. 195.
  196. 196.
  197. 197.
  198. 198.
  199. 199.
  200. 200.
  201. 201.
View Abstract

Navigate This Article