JLE

Epileptic Disorders

MENU

Seizure-induced hippocampal damage in the mature and immature brain Volume 4, issue 2, June 2002

Neurologists are keenly interested in the acute and chronic effects of brief and prolonged seizures in the human brain. These questions are of immediate clinical significance in judging how aggressively a physician should treat prolonged or recurrent seizures and in determining the prognosis following severe seizures. Although much has been learned from human and animal studies over the past quarter century, resulting in a consensus on many issues pertaining to seizure-induced damage, important controversies nevertheless remain. In this review, we will evaluate the evidence available from human and animal studies that forms the basis of our current understanding of seizure-induced damage in the adult and immature brain, and we will discuss those topics where additional investigation is necessary to resolve controversies in seizure-mediated damage in the hippocampus.

Temporal lobe epilepsy, characterized by recurrent complex partial seizures, is a common form of epilepsy in children and adults. The mesial structures of the temporal lobe, including the hippocampus and neighboring structures such as the amygdala and the piriform and entorhinal cortices are frequently the origin of medically refractory seizures [1]. For almost two centuries, we have known that mesial temporal structures, such as the amygdala and hippocampus, frequently appear damaged in patients with temporal lobe epilepsy. Boucher and Cazauvieilh, in 1825, noted hardening of the mesial temporal lobe in postmortem brains of epileptics [2]. In 1880, Sommer described the sclerotic changes in epileptic hippocampus, including the characteristic loss of hippocampal neurons. Later, Bratz observed that hippocampal sclerosis was not seen in all forms of epilepsy, and came to the conclusion that hippocampal sclerosis was a symptom rather than a cause of epilepsy [3]. On the other hand, more recent retrospective studies came to the conclusion that hippocampal sclerosis resulted from the effects of recurrent seizures [4]. As we will see, modern neuroimaging and current animal models indicate that severe seizures in the adult can often result in hippocampal sclerosis. Milder seizures may produce functional changes, or even alter synaptic connectivity without widespread neuronal damage. Similar unambiguous evidence regarding neuronal loss or synaptic changes is lacking for children or immature animals. Animal studies indicate that immature animals are largely protected against the more damaging consequences of seizures, though long-term functional changes may occur. The absence of persuasive prospective evidence connecting early life seizures and hippocampal sclerosis has led some to conclude that seizures are a symptom, not a cause, of hippocampal sclerosis [5, 6]. Nevertheless, early-life seizures may increase susceptibility to neuronal injury later in life. Gender, like age, may also influence the susceptibility to seizure-mediated damage, and alter the expression and consequences of seizures. Boys are more likely to develop epilepsy and seizures than girls, even after allowing for the increased male risk factors for epilepsy, such as head trauma and CNS infection [7]. Boys, for example, more frequently develop febrile seizures [8], and are 1.5-2 times more likely to develop multifocal epileptic syndromes than girls [9, 10].

Hippocampal anatomy and pathology

The hippocampus is a mesial temporal structure located posterior to the amygdala, and is comprised of three regions, the dentate gyrus, Ammon's horn, and the subicular complex [11]. Ammon's horn (Cornu Ammonis - CA - in Latin) is further divided into the CA1, CA2, and CA3 regions that are distinguished from one another by variations in the morphology of their pyramidal cells. In the past, the dentate hilus, the region that lies between the dentate gyrus and CA3 was classified as a fourth part of Ammon's horn, "CA4" [12]. Under current classification, the hilus region is grouped with the dentate gyrus rather than as part of Ammon's horn proper [13]. This discussion will use the term dentate hilus to include the "CA4" region. The hippocampus figures prominently in the understanding of epilepsy, by virtue of the interconnectivity between each of the three hippocampal components and between the three subregions of Ammon's horn. These intrahippocampal connections comprise circuits containing excitatory feedback loops capable of generating an epileptic state. Hippocampal afferent inputs arising from neocortical, subcortical, limbic, and precommissural septal and brainstem regions, synapse on the granule cells of the dentate gyrus [14]. Dentate granule cells receive a major input from the entorhinal cortex - which collects inputs from widespread cortical regions - in the form of the perforant path. Dentate granule cells, in turn project efferent connections to hilar neurons and CA3 pyramidal cells [11, 15]. These mossy fibers, as the dentate projections are known, provide excitatory glutamatergic input to hilar cells and CA3 neurons. Hilar neurons, called "mossy cells" - not to be confused with the mossy fibers arising from dentate granule cells - receive mossy fiber input and return axonal projections to the dentate granule cells to form recurrent synapses on dentate granule cell dendrites. In this manner, neuronal activity originating in dentate granule cells may "recur" or return to the granule cell layer polysynaptically via the hilus. Hilar neurons comprise at least two populations. Hilar neurons contain glutamate [16] and appear to form excitatory synapses by morphology, and produce excitation in dentate granule cells [17-19]. Some hilar neurons, however, are immunoreactive for GABA, and produce inhibition in target neurons [20-22]. A recurrent excitatory loop formed by interconnections from granule cell to hilar cell to granule cell could be epileptogenic [18]. Hilar "mossy" cell axons, however, also appear to synapse on inhibitory dentate interneurons, such as basket cells in and near the granule cell layer [23, 24]. These inhibitory interneurons can inhibit granule cell firing, and reduce the effects of recurrent excitation [25]. On the other hand, inhibitory neurons in the dentate gyrus, hilus and other regions of the hippocampus also synchronize the firing of excitatory neurons [26], and could contribute in this manner to epileptogenesis. The local circuitry of the dentate gyrus - in short - is complex and combines the effects of both positive and negative feedback circuitry [27]. Similar recurrent excitatory innervation can also be found in CA3, where the axons of CA3 pyramidal cells send local collaterals to strongly innervate neighboring neurons in CA3 [13]. It has been estimated that local axon collaterals are sufficiently dense that there is a 5% probability that any one pyramidal cell contacts another [28]. As in the dentate gyrus however, recurrent excitatory circuits coexist with local inhibitory interneurons in CA3 that also receive recurrent input from CA3 pyramidal neurons. In CA3, each pyramidal cell is surrounded by approximately 15 inhibitory interneurons [28], and is positioned to inhibit pyramidal cell output. Thus the presence of local excitatory feedback within the hippocampus in either the dentate gyrus or CA3 may provide a substrate for epileptogenesis, while activation of local inhibitory circuits may suppress seizures.

The activity of CA3 pyramidal neurons reaches the pyramidal neurons of the CA1 region through axon collaterals, called Schaeffer collaterals, to the pyramidal cells of CA1, where the activity is relayed directly and via the subiculum back to the entorhinal cortex, completing the loop of excitatory activity that began with the input of perforant path activation to the dentate gyrus. From the entorhinal cortex, hippocampal activity is distributed to cortical regions.

Temporal lobe epilepsy is associated with characteristic hippocampal pathology that may often be visualized radiographically by MRI during the presurgical evaluation of an epilepsy patient. An early surgical series found hippocampal sclerosis, consisting of neuronal loss in the CA1, CA3 subregions with relative sparing of CA2, in 50-70% of pathological specimens following temporal lobectomy [29]. The dentate gyrus, including the hilus, and the prosubiculum also show evidence of neuronal loss [30-32].

Consequences of seizures
in the adult hippocampus

Three questions concern the neurologist when considering the effects of seizures on the adult hippocampus. First, do severe seizures, such as status epilepticus, damage the hippocampus, and how long must the seizure last, before damage accrues? Second, do repetitive brief seizures produce the same hippocampal pathology as status epilepticus or are there differences? And third, what are the effects of a single brief seizure.

Effects of status epilepticus in the adult hippocampus

Prolonged seizures, such as generalized convulsive status epilepticus, cause damage in the adult brain, at least at the level of the hippocampus when evaluated radiographically in a case study [33], or pathologically in a small case series [34]. Prompt treatment of status in a study of eight patients, however, can prevent the radiographical development of hippocampal damage [35]. A review of three patients also found that prolonged focal motor status could result in hippocampal damage [36]. Human studies on the effects of seizures, however, are necessarily limited in examining the relationship between seizure severity or frequency and features of hippocampal sclerosis. Instead, our understanding of the effects of status and less severe seizures on the hippocampus derives from numerous animal studies. Several animal models have been used to investigate both the pathogenesis of temporal lobe seizures as well as the role of prolonged or recurrently provoked seizures on the development of epilepsy.

Meldrum and coworkers [37, 38] produced the first evidence that severe seizures caused excitotoxic cell death, from paralyzed and ventilated baboons that developed neuronal loss following status epilepticus, despite adequate metabolic support. Similar work in ventilated and paralyzed rats found that neuronal injury occurred despite adequate oxygenation [39]. Neuronal injury is present in adult rats exposed for 24 hours to purely electrical seizures without convulsions [40-42]. Kainic acid delivered either systemically or injected directly into the mesial temporal lobe produces seizures and characteristic chronic changes that resemble those of hippocampal sclerosis [43, 44]. Histological changes resulting from kainic acid status epilepticus include loss of neurons in the CA1 and CA3 hippocampal regions and in the dentate hilus [45-47], as well as neuronal loss in other regions [48]. Hippocampal damage varies in severity, ranging from gliosis and a small reduction in neuron number to frank necrosis with destruction of the hippocampus [46]. After a variable latent period of several weeks, spontaneous seizures originating in the amygdala and hippocampus may develop in adult rats after exposure to kainic acid [49, 50]. Hippocampal neuronal loss and spontaneous recurrent seizures have also resulted from seizures produced by other methods. Pilocarpine, a cholinergic agonist, produces seizures and loss of neurons in CA1, CA3 and the hilus when given systemically [51-54]. Relatively brief stimulation of the hippocampus [55, 56] or the perforant pathway [57-59] produces self-sustained status epilepticus that will result in neuronal injury and mossy fiber sprouting.

Evidence of altered synaptic connectivity within the hippocampus coexists with the classical findings of neuronal loss in hippocampal sclerosis. Evidence of aberrant mossy fiber collaterals, consisting of the abnormal growth of axonal terminals from mossy fibers into the supragranular layer of the dentate gyrus, can be found in the surgically resected hippocampi of patients with chronic, medically refractory epilepsy. Similar changes in collateral growth and synaptic changes are seen in adult rats following status epilepticus induced by kainic acid adminstration. In these animals, new mossy fiber collaterals are found in two regions, CA3 [60, 61], and the supragranular layer of the dentate gyrus; the region containing the dendrite of the dentate granule cells [62, 63]. Supragranular sprouting is present in hippocampi removed from epileptic patients, but CA3 sprouting was not observed [15]. Mossy fiber "sprouting", as the creation of these collaterals is known [15, 64-66], results in the the creation of new synapses within the dentate gyrus. Mossy fiber collaterals carrying the excitatory output of dentate granule cells synapse in the dendritic arbor of the dentate gyrus, and may create or strengthen recurrent excitatory circuits within the hippocampus [63, 67-70]. Some have suggested, however, that the net effect of mossy fiber sprouting may be inhibitory rather than excitatory because some of the aberrant mossy fiber collaterals target GABAergic inhibitory interneurons within the granule cell layer [71]. The role of these synaptic rearrangements in epileptogenesis is unclear. Administration of cycloheximide, an inhibitor of protein synthesis, to rats receiving intrahippocampal or systemic kainate did not prevent the long-term development of recurrent seizures [72, 73]. Moreover, emerging data indicate that seizure-induced hippocampal changes also include changes in synaptic receptor function, and neuronal phenotype [74]. Thus, the precise functional consequences of the mossy fiber sprouting remain unknown and the subject of controversy.

The presence of neuronal loss and alterations in synaptic connectivity are well established through animal models and clinical case studies. Less well understood is the minimum duration seizures must be present to produce damage. Studies in the rat receiving focal hippocampal and systemic injections of kainic acid found that hippocampal and rhinal cortex neurons were undamaged after 30 min of convulsive status epilepticus, but that DNA fragmentation - a marker of apoptotic cell death - was present in both regions after 60 minutes of onset of status epilepticus [75]. Tuunanen et al. examined the extent of damage in the amygdala using DNA fragmentation as a marker of neuronal injury following systemic status epilepticus [76]. These investigators looked for damage 1, 2, 4 or more hours after administration of kainic acid, and found that the earliest damage occurred at 4 h. Since convulsive status epilepticus began about one hour after kainic acid administration, amygdala damage was present after approximately 3 h of convulsive motor seizures. Ventilated and paralyzed rats exposed to flurothyl, a gas producing generalized clonic and tonic convulsions, developed status epilepticus-related neuronal loss after short intervals of status [77].

Excitotoxicity and calcium-mediated damage

The delay in onset of neuronal damage following onset of status epilepticus reflects the accumulation of damage at the neuronal level that reaches a threshold where changes become irreversible. Unraveling the sequence of events resulting in neuronal death may uncover opportunities to halt or prevent the damage accrued by prolonged or frequent seizures. An early step in the evolution of status epilepticus is the increased release of excitatory neurotransmitters. Increased levels of glutamate - the major excitatory neurotransmitter in the mammalian brain - and sometimes aspartate have been shown to occur during seizures induced in animals [78-80] and at the onset of complex partial seizures in humans [81, 82]. The postsynaptic effects of glutamate are mediated through four major receptor subtypes: the N-methyl D-aspartate (NMDA) receptor, the alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionate (AMPA) receptor, the kainate receptor [83], and the metabotropic glutamate receptor [84, 85]. The distribution of glutamate binding sites in the rat are greatest in the CA1 and CA3 regions of the hippocampus [86], and correspond to the regions of greatest damage following status epilepticus in rats [78]. CA1 and CA3 may undergo greater injury during status epilepticus possibly because the higher concentration of glutamate receptors in the CA1 and CA3 regions may render these areas more susceptible to glutamate-mediated excitotoxic damage.

The majority of excitotoxic neuronal damage is most likely mediated at the cellular level by the excessive entry of calcium ions into the cell during seizures [41]. Intracellular calcium increases as a result of passage through the ionic channel of the NMDA receptor [87], or through voltage-gated calcium channels that open in response to membrane depolarization produced by synaptic excitation [88]. High intracellular calcium levels trigger a sequence of intracellular events that culminate either in necrotic - i.e. metabolic - cell death or in apoptotic - i.e. programmed - cell death. High intracellular calcium leveling activate nitric oxide synthetase which interferes with oxidative metabolism, leading to generation of free-radicals that damage the neuronal membrane. High intracellular calcium levels also aggregate and activate procaspase proteins that in turn activate caspase proteins. Caspase proteins initiate apoptosis [89, 90]. Although the major mechanism of neuronal death in the hippocampus during status epilepticus is believed to be apoptosis [42, 91-94], some have presented evidence, that neurons undergoing necrotic degeneration also express some of the markers used to detect apoptotic degeneration, such as DNA laddering on gel electrophoresis or in situ DNA nicking by TUNEL staining [95].

Direct imaging of intracellular calcium levels during epileptic activity has revealed dynamic changes in calcium regulation that may underlie the eventual neuronal loss and synaptic changes associated with hippocampal sclerosis. Imaging of intracellular calcium in cultured neurons maintained in vitro shows that calcium levels revert to normal following brief epileptic discharges. After 60 min or longer of epileptic activity, however, the intracellular calcium concentration remains elevated and no longer returns to baseline [96]. Calcium levels were also noted to increase within the nucleus of the neuron, where they may influence gene expression. The changes in calcium regulation are chronic following status epilepticus. Intracellular calcium levels remained elevated in neurons taken from rats one year following status epilepticus [97]. In summary, in vivo evidence from animal models of status epilepticus and in vitro evidence of intracellular calcium regulation suggest that irreversible changes occur as early as one hour after onset of convulsive status epilepticus. Changes in intracellular calcium are believed to trigger the apoptotic and necrotic processes in some neurons, while in other neurons the elevation of calcium may initiate plastic changes in synapses.

Effects of frequent recurrent seizures
in the adult hippocampus

Next, we consider the question whether brief transient seizures cause hippocampal neuronal loss or changes in synaptic connectivity or function. Kindling, an epilepsy model whereby seizures are produced by repeated electrical or chemical limbic stimulation, produces hippocampal cell loss, although the abnormalities are generally less severe than the changes produced by kainic acid administration [98, 99]. Stereological neuronal counting methods sensitive to small differences in neuronal counts were applied to the amygdala in amygdala-kindled rats and found no change in the number of amygdala neurons in kindled rats compared to unkindled controls [100]. The effects of hippocampal kindling on neuronal damage, however, are complex. Hippocampal kindling prior to the induction of status epilepticus resulted in more severe seizures during status epilepticus, but reduced the status-induced damage in the hippocampus. The recurrent brief seizures produced by kindling had a neuroprotective effect against more severe damage related to status epilepticus [101]. Injection of a small amount of kainic acid unilaterally into the hippocampus in most animals produced non-convulsive seizures lasting 35-45 min that did not result in neuronal damage. A minority of rats, however, developed status epilepticus following injection of low-dose kainic acid and showed evidence of early damage after periods of 30-60 min of status epilepticus [102].

Neoneurogenesis in the adult hippocampus
following seizures

One reason that neuronal loss is not striking following kindled seizures may be that kindling also promotes neoneurogenesis - the birth of neurons in the adult brain - and increases the introduction of new neurons into the dentate granule cell layer of the hippocampus [103, 104]. Neoneurogenesis has also been observed following status epilepticus induced by systemic convulsants, such as kainic acid [105] and pilocarpine [106], and in electroconvulsive seizures - a model of human electroconvulsive therapy (ECT) [107, 108]. In normal adult rats, new dentate neurons develop continuously from the subgranular proliferative zone in the dentate gyrus and migrate into the granule cell layer of the dentate gyrus. Following status epilepticus, there is an increase in the production of new granule cells that migrate aberrantly and are found in the hilus and in the inner molecular region of the dentate gyrus [106]. These new granule cells also make aberrant connections in CA3 and the inner molecular layer of the dentate. Inhibition of granule cell proliferation by brain irradiation [109] did not prevent the development of mossy fiber sprouting, indicating that mature granule cells are capable of forming these novel synaptic connections [110]. The nature of the connections made by these new neurons has not been determined, however. The pro- or anticonvulsant effects of seizure-induced neurogenesis are unknown.

Effects of single brief seizure in the adult hippocampus

In view of the evidence indicating that either a single prolonged seizure or repetitive brief seizures induce alterations in the neuronal pool and synaptic connections in the hippocampus, one may also ask whether a single, brief seizure produces lasting changes in hippocampal structure or function. To date, this question has not been systematically studied, but initial investigations are highlighting areas that hold promise for understanding how seizures bring about hippocampal alterations. Blumenfeld and colleagues have studied electroconvulsive therapy (ECT) in a clinical population as a model of human epilepsy [111]. These investigators used single photon emission computed tomography (SPECT) perfusion studies in patients undergoing treatment of refractory depression to identify brain regions activated during the seizure. SPECT injections performed 30-60 s after application of unilateral ECT showed increases in perfusion in the ipsilateral frontotemporal region, which included the hippocampus. Bifrontal ECT produced increased perfusion in the frontotemporal regions bilaterally. Application of the same SPECT methods to patients with temporal lobe epilepsy found preferential activation of the medial thalamus early in the seizure and in the contralateral medial thalamus after seizure termination [112]. Activity and seizure-related damage in the medial thalamus are correlated with epileptic discharges in the hippocampus [113, 114]. Although the SPECT study of patients with temporal lobe epilepsy identified specific perfusion changes following one seizure, these patients already had epilepsy, and a history of recurrent seizures. Kindled seizures in rats have been used to study the changes to c-fos - an intermediate-early gene - activation following single or repeated seizures. Kindled seizures increase expression of c-fos in the dentate gyrus [115], and dorsal hippocampus [116], and in other regions [117, 118]. Partial seizures produced in amygdala-kindled rats increased c-fos expression either in the ipsilateral amygdala and piriform and entorhinal cortices [117, 118], or in the hippocampus, but not both [118]. Generalized seizures kindled from the amygdala produced additional widespread bilateral expression of c-fos in the hippocampus and dentate gyrus, the contralateral entorhinal and piriform cortices and amygdala, and throughout neocortex [117, 118].

The induction of c-fos expression results from the epileptic discharge itself rather than from the kindling process. An afterdischarge duration greater than 30 s following hippocampal stimulation in either kindled or unkindled rats increased expression of c-fos, while shorter afterdischarges in either group did not [119]. Naïve rats exposed acutely to cocaine-induced seizures expressed increased c-fos in the dentate gyrus [120]. Rats exposed to one flurothyl-induced generalized seizure showed similar patterns of c-fos expression as rats with generalized seizures kindled after repeated flurothyl exposure [121]. The threshold for induction of c-fos may also be site-dependent. Briefer afterdischarges in the amygdala, compared to the hippocampus, may increase c-fos expression [122]. In fact, electrical stimulation alone in the amygdala, without accompanying afterdischarge may even induce c-fos expression, and c-fos expression may increase in the contralateral amygdala even without a local afterdischarge [123]. These observations have led some investigators to conclude that c-fos induction results from neuronal activation of any cause and not only from seizures [123].

Although the activation of c-fos may not be specific to seizures, c-fos regulation may be important for the subsequent development or prevention of epilepsy. Mice carrying a null mutation of c-fos developed kindled seizures more slowly and underwent less mossy fiber sprouting than wild-type mice, suggesting that c-fos expression plays a role in kindling [124]. Other investigators, however, have found that focal administration of c-fos antisense oligonucleotides, inhibited expression of c-fos and accelerated amygdala kindling, implying that c-fos expression is important in retarding the development of kindled seizures [125, 126]. Single, brief seizures produce acute changes in c-fos expression both in the naïve and in the kindled rat. The effects of increased c-fos expression, and the role of c-fos in the development or prevention of epileptogenesis, however, remain unclear.

Consequences of seizures
in the immature hippocampus

In humans, the greatest incidence of epilepsy occurs in the immature brain as children mature through the first years of life. The models of adult epilepsy discussed to this point clearly establish that severe or repetitive seizures in the mature brain produce changes in neuronal populations and rearrangements of synaptic circuits that are ultimately epileptogenic and may result in spontaneous recurrent seizures. The effects of prolonged or repetitive seizures in the immature brain, however, differ significantly from the effects in the adult brain. To understand epileptogenesis - which is largely a process occurring in childhood - one must consider a series of incremental steps that lower seizures threshold and eventually result in epilepsy. The first seizure in childhood results from the combined interaction of genetics, age and environmental initiating events that promote proconvulsive tendencies while suppressing endogenous seizure suppressing mechanisms. The progression from the first seizure to recurrent spontaneous seizures and hippocampal - the hallmarks of mesial temporal lobe epilepsy - may result from several possible distinct mechanisms. The first possible mechanism is that the first seizure produces hippocampal injury and synaptic rearrangement, as has been observed in the adult brain. These changes may be clinically apparent immediately after the first seizure - especially if the seizure is severe and prolonged - or a long silent interval may separate the first seizure from the emergence of epilepsy. Another possibility is that damage is present in the hippocampus before the first seizure. The preexistent hippocampal injury lowers the seizure threshold for proconvulsant stimuli, such as fever. The first seizure may cause no damage and is a symptom - not a cause - of neurological injury. The final possibility combines the first two. Preexistent damage to the hippocampus lowers the threshold for the first convulsion. The first seizure sets up a cycle whereby each seizure produces additional hippocampal injury and lowers the threshold for the next seizure. Some have argued that epilepsy is a progressive disease [32, 127-129]. Engel favors early aggressive intervention to prevent the cycle of seizures from producing additional proconvulsant injury to the brain [129].

Do febrile seizures cause hippocampal sclerosis?

The question of whether seizures beget hippocampal sclerosis or whether preexistent hippocampal sclerosis begets seizures has been studied in children experiencing febrile convulsions. An MRI investigation of infants with complex febrile convulsions found that six out of fifteen infants with lateralizing febrile seizures had underlying structural abnormalities in the brain [130]. Of the six infants with structural abnormalities, two had preexistent hippocampal atrophy, four had increased hippocampal volume and T2-weighted signal, indicative of acute hippocampal edema. Two of the infants with acute edema - one with temporal seizure activity confirmed by EEG and the other with a choroid fissure cyst adjacent to the hippocampus - went on to develop atrophy. A study of three pairs of identical twins where one twin had temporal lobe epilepsy while the other was asymptomatic, found hippocampal sclerosis in the affected child, but not in the unaffected twin and concluded that hippocampal sclerosis is an acquired, postnatal lesion [131]. Two of the three unaffected twins, however, also had febrile seizures, but did not go on to develop epilepsy or hippocampal sclerosis. A study of two families with a total 23 members - 13 with a history of febrile seizures and one from each family with temporal lobe epilepsy - found that all family members with a history of febrile or temporal lobe seizures had hippocampal sclerosis. Hippocampal asymmetry, however, was also present in six out of ten family members without a history of febrile seizure or epilepsy. The authors of the study concluded that hippocampal abnormalities predated the onset of febrile seizure and may have predisposed to later convulsions and hippocampal sclerosis [132].

The need to better understand the relationship between seizures, epileptogenesis and development requires investigation of animal models of childhood epilepsy. These models must take into account the altered excitability of immature neurons, the relative increase in myelination from birth to adulthood, and the hormonal changes that accompany growth and sexual maturation. The rate of development and periods of relative vulnerability to neurological injury vary among species, and the optimal choice for laboratory work will compress the period of maturation while expressing similar developmental progression as humans. The rat, which matures post-natally, has been the subject of many studies of epilepsy. The developmental stage of rat pups at eight days of age is similar to the full-term human infant [133]. The maturation of rats, which reach puberty between 33 and 38 days of age, between the ages of 8 to 33 days parallels the human stages of infancy and childhood. Moshé [134] has suggested that increased seizure susceptibility seen in rats in the third week of life corresponds to human seizure susceptibility between the ages of six months to six years.

Effects of status epilepticus
in the immature hippocampus

The extent of hippocampal neuronal loss following kainate-induced status epilepticus depends on age, despite the fact that seizures may be more severe. For example, unlike adult rats, juvenile rats between 14-15 days old largely avoid hippocampal damage following kainate administration, despite the fact that the behavioral seizures are more severe in the younger animals [135, 136]. Following kainate administration, juvenile rats developed both generalized and lethal seizures at lower equivalent doses compared to adult rats. The resistance to neuronal damage is not limited to kainate-induced status, but can also be found in other seizure models such as kindling [137], pilocarpine-induced status epilepticus and flurothyl-induced status epilepticus [138]. The resistance to hippocampal damage may be relative, and not absolute, however, as some investigators have suggested that lithium-pilocarpine-induced status epilepticus can produce damage to neurons [139, 140] and reduced hippocampal neuronal density [141]. Though, in the latter instance, the reduction in neuronal density may have resulted from increased volume of the post-status hippocampus. A model of febrile seizures that used hyperthermia-provoked convulsions in immature rats produced transient changes in hippocampal and amygdala neurons that persisted for two weeks and then returned to normal. Markers of neuronal death, such as DNA fragmentation, were not found [142, 143]. Combined application of hyperthermia and 45 min of electrical stimulation to the hippocampus also found no evidence of neuronal loss in 20-day-old rats [144]. The authors of this study concluded that rats are resistant to hyperthermia-induced damage and questioned the role of febrile convulsions in the development of hippocampal sclerosis.

Although normal immature animals may be resistant to some of the damage caused by seizures at older ages, the same protection is not present for animals with preexisting neurological abnormalities. In humans, a history of febrile convulsions and a preexisting neurological abnormality significantly increased the prospective risk of recurrent status epilepticus [145]. A history of febrile convulsions in neurologically impaired children also significantly increased risk prospectively for recurrent, unprovoked seizures and the development of epilepsy [146]. Though suggestive, these studies however cannot determine whether seizures induce neuronal injury in the presence of preexistent neurological abnormalities even in the immature brain or whether the presence of neurological abnormalities reduces the seizure threshold and increases the risk of developing additional seizures, status epilepticus and epilepsy. In the latter case, febrile convulsions are a marker for the preexistent condition and not a cause of the sequelae.

The hypothesis that the neurologically abnormal immature brains are more likely to manifest seizure-mediated damage than normal controls has been tested using the methylaxozymethanol (MAM) model of cortical dysplasia in the immature rat. Intraperitoneal MAM administration to pregnant rats crosses the placenta to reach the developing embryos and induces a variety of dysplasias, abnormal migration patterns and cortical disorganization, including cortical heterotopias consisting of malformed and abnormally oriented neurons [147-149]. MAM-exposed pups have an increased susceptibility to seizures produced by hyperthermia [150], kainic acid [151], and kindling [152] compared to controls. Furthermore, seizures produced by each of these methods resulted in hippocampal damage. Although greater than the damage seen in controls, the extent of neuronal injury in MAM-treated pups was still less than the extent of hippocampal damage seen in comparably treated adult rats [150-152]. The MAM-treated rats were also more sensitive to hyperthermia-induced hippocampal damage than controls [150].

Neoneurogenesis in the immature hippocampus

As with adult rats, repeated severe seizures can affect neurogenesis, but the data are not yet clear. Lithium-pilocarpine status epilepticus in two- and three-week-old rat pups increased neurogenesis in the dentate gyrus [153]. An initial report of neurogenesis following exposure to 25 flurothyl seizures in the first four days of life, also found neurogenesis increased in the dentate gyrus following the last seizure. As adults, rats exposed to flurothyl were found to have increase granule neurons compared to controls. However, in a subsequent study using the same experimental method of 25 flurothyl seizures in the first four days of life, dentate neurogenesis decreased for at least six days following the last seizure [154].

Age-dependent mechanisms
of seizure-mediated synaptic reorganization

Just as the extent of hippocampal damage following neuronal loss depends on the developmental age, so does the extent of mossy fiber sprouting. Although seizures in 15-day-old rat pups induced by kainate or lithium-pilocarpine are more severe than in adult rats [155], mossy fiber sprouting is not found in the immature pups [135]. Only under extreme conditions - such as undergoing 15 generalized convulsions daily for 15 days, has mossy fiber sprouting occurred in the first two weeks of life [156]. Holmes and McCabe have hypothesized that the immature brain resists mossy fiber sprouting because synaptogenesis and sprouting require the presence of adult levels of GAP-43 expression, which do not reach full expression until day 25 of life [157]. Ribak and Navetta [158] have reported that mossy fiber innervation approaches the adult pattern between day 21-25 of life, consistent with the hypothesis of Holmes and McCabe. Contradictory evidence however, has also been put forward by Holmes et al. [156] indicating that the normal adult mossy fiber pattern is present by day 10 of life, at a time mossy fiber sprouting is not seen. The differences in synaptic function - tested by "paired-pulse inhibition" in vitro in pups and adults [49, 135, 159] - parallel the difference seen in mossy fiber sprouting. Seizures in adult rats produced by kainate, flurothyl, and kindling all decreased inhibition acutely and increased inhibition chronically, while similar treatment does not produce any change in developing animals [67, 71, 135, 138, 141]. On the other hand, hyperthermia-induced seizures - but not kainic acid seizures - in immature rats produced a persistent increase in inhibition in the CA1 region of the hippocampus [143]. An increase in inhibitory post-synaptic currents lasted as long as 10 weeks after the hyperthermia-induced seizures. The increased inhibition depended on the activation of protein kinase A and originated presynaptically. Despite the chronic increase in CA1 inhibition, immature rats exposed to hyperthermia had a lower holdthreshold to kainic acid-induced seizures in adulthood [160]. The significance of the persistent increase in inhibition, and whether the change is detrimental or beneficial over the long term is unknown. The mechanism whereby immature rats resist the changes in mossy fiber connectivity is also unknown.

Why is the immature brain resistant to excitotoxic
and calcium-mediated damage

What are the mechanisms through which the immature brain acquires resistance against recurrent seizures? The degree of calcium entry into the neurons of CA1 following exposure to glutamate appears to vary with postnatal age. On postnatal days 1-3, minimal calcium entry was seen in the neurons of CA1 [161]. In contrast, morphological changes in neurons accompanied by marked entry of calcium were both present following exposure on the 25th day of life. Some have hypothesized that the increase of susceptibility to seizure-induced damage with age may result from changes in the quantity or type of glutamate receptors as the animal develops [162-165]. By two weeks of age, a full complement of glutamate receptors for kainate and NMDA are present, even though seizure-mediated damage does not occur. It is possible, however, that changes in the level of expression of glutamate receptor subunits are responsible for the increase in seizure-mediated damage that accompanies maturation. Developing rats express GluR2 and GluR3 receptors at adult levels. Expression of GluR1 subunits, in contrast, is much higher in immature rats compared to adults [166]. Maturation also produces similar changes in GluR6 subunit expression, as well as NR1 mRNA and mGluR mRNA expression [167-169]. Importantly, when assembled with other glutamate receptor subunits, GluR2 subunits block calcium ions from entering the neuron. Seizures may down-regulate GluR2 levels in adult rats in regions associated with seizure-induced neuronal loss, such as CA3, but not in regions without cell loss, such as the dentate gyrus. Developing rats, in contrast, do not show changes in GluR2 or GluR3 levels following seizures [170].

Can seizures early in life augment
later hippocampal injury?

Immature animals largely avoid the worst immediate detrimental effects of seizures that are manifested in in adults [137, 152, 171, 172]. The possibility exists, however, that seizures in the immature animal induced more subtle changes that may alter the resistance to seizure-induced damage in adulthood [173]. Okada et al. [174] found that rats that experienced kainic acid status epilepticus early in life were no more susceptible to kindling or to kindling-induced neuronal damage than normal controls. Koh et al. [175] found that the amount of hippocampal injury caused by kainic acid-induced status epilepticus at 45 days of life was greater in animals that underwent an episode of status epilepticus at 15 days compared to animals that only experienced a single episode of status epilepticus in adulthood. Seizures early in liflife, in other words, sensitize the brain to the effects of later seizures, and reduce the resistance to seizure-mediated neuronal damage.

Are seizures epileptogenic?

Whether seizures are epileptogenic, i.e. cause changes in neurons and circuits that result in spontaneous seizures, has not yet been answered through clinical or laboratory studies. Following status epilepticus produced by kainic acid [176] or pilocarpine [50, 177], or after extensive kindling [127, 178], rats may develop spontaneous seizures. Status epilepticus in adult rats can also accelerate the development of kindled seizures, suggesting that latent proepileptic changes may be present following status [179, 180]. Kainic acid status epilepticus in one and fifteen-day-old rat pups, however, does not accelerate kindling in adulthood [127, 174]. Indeed, the likelihood of developing spontaneous seizures or cognitive impairments following lithium-pilocarpine status epilepticus increased with the pup's age at the time of status epilepticus, with the first differences noted at 20 days of age [181]. Pentylenetetrazol - a convulsant drug - administration on day one of life resulted in slower kindling in adulthood [182]. In two-week-old pups, on the other hand, kindling early in life accelerates repeated kindling in adulthood [183], even when performed in the hemisphere contralateral to the initial site of kindling [184]. Hyperthermic seizures on day 10 of life increase the susceptibility and severity of kainic acid status epilepticus in adulthood [160]. Thus, some types of seizures - such as amygdala-kindled afterdischarges or hyperthermic seizures may induce secondary epileptogenesis in immature rats, while other seizures, such as kainic acid status epilepticus, do not [174, 184]. On the whole, the immature brain resists epileptogenic changes more readily than the adult brain, yet early life seizures may still facilitate the development of epilepsy even in the absence of clear structural or synaptic rearrangements.

Sex hormones and neuroprotection

Clinical experience with sex hormone-sensitive seizures in women, i.e. catamenial epilepsy, most often supports the conclusion that estrogen lowers seizure threshold and is proconvulsant. Progesterone, on the other hand, may be anticonvulsant in some women [185]. Laboratory studies of the effects of sex hormones on seizures largely confirm this clinical impression [186-189]. In addition to genomically-mediated excitatory effects, estrogen directly reduces inhibition mediated by the GABA-A receptor and acts as an agonist at the NMDA receptor [190]. Progesterone, in contrast, directly enhanced GABAergic inhibition, while antagonizing NMDA-mediated excitation. Through genomic mechanisms, progesterone also increased synthesis of GABA and increased the number of GABA-A receptor subunits [190]. The effects of sex hormones however, may vary depending on brain region. Velisek and colleagues [191] found that in vivo progesterone treatment in ovariectomized rats prior to sacrifice decreased the latency to onset of in vitro epileptiform activity in the CA1 region, but increased the latency to onset of epileptiform activity in entorhinal cortex. Pretreatment with estrogen produced milder but similar effects as progesterone; however, the differences were not statistically significant. The effects of sex hormones may also vary by seizure type. In a series of 15 women with primary generalized epilepsy, Jacono and Roberston [192] found that estrogen peaks were associated with decreased seizure frequency. Along the similar lines, Schwartz-Giblin [193] found that 75% of ovariectomized rats exposed to picrotoxin experienced a generalized seizure, while ovariectomized rats receiving estradiol implants did not experience a seizure after treatment with picrotoxin. Low dose estradiol pretreatment prior to kainic acid exposure, significantly delayed onset of the first clonic seizure, but not of status epilepticus, indicating a possible anticonvulsant benefit that may be dose-dependent [194].

Despite clinical and laboratory evidence that estrogen may alter seizure threshold, there is also growing evidence that sex-hormone, particularly estrogen, may have neuroprotective effects. Pretreatment with estrogen protected against glutamate excitotoxicity in neurons in vitro, and the effect could be blocked with tamoxifen, a specific antagonist of the estrogen receptor [195]. Similar neuroprotection did not result from progesterone, dihydrotestosterone, dexamethasone, or cholesterol treatment. Neuroprotective effects of estrogen pretreatment have been observed in multiple models of brain injury in vivo, including oxidative-stress injury of mesencephalic dopamine neurons in vitro and in vivo [196, 197]; focal ischemia [198]; traumatic brain injury [199]; and status epilepticus [194, 200]. Particularly, estrogen at low or moderate doses, can significantly modulate neuronal damage produced by prolonged seizures [194, 200]. Reibel and colleagues found that pretreatment with moderate-dose estradiol reduced hippocampal cells loss in CA1, CA3 and in the hilus following kainic acid status epilepticus in ovariectomized rats. Estradiol treatment on the day of status epilepticus was less effective than pretreatment, but still reduced cell loss in CA3. Veliskova and colleagues [194] found that pretreatment with low-dose estradiol significantly reduced hippocampal cell loss compared to rats treated with vehicle. Estradiol treatment after status epilepticus, however, did not protect against seizure-induced damage. Treatment with tamoxifen reversed the neuroprotective effects of estradiol.

CONCLUSION

... and directions for future research

Although areas of controversies exist, there is emerging consensus on several points regarding the effects of seizures at progressive stages of development. In most instances, there is no convincing evidence that either febrile convulsions in normal children or provoked seizures in normal, immature animals produce hippocampal sclerosis. There is, however, evidence from human and animal studies that in a small number of cases, hippocampal alterations and even damage may result. Why a few children develop hippocampal injury is an important question for future research, especially since the observed changes may not be typical of mesial temporal sclerosis seen in adults. In some cases, the proximate causes of seizures, such as encephalitis or hypoxia-ischemia, may themselves produce brain injury. Furthermore, genetic or acquired abnormalities may lower the protection against seizure-induced changes or injury present early in development. In children with neurological impairment, and in congenitally-lesioned rats, seizures early in development can indeed result in neuronal loss and gliosis, although it is less severe than that seen in adults [130, 145, 146, 152]. With time, the hippocampal injury may become epileptogenic. Finally, seizure-induced neuronal damage and synaptic rearrangements increase with progressive maturation [136, 176]. It is possible that recurrent seizures, especially as an animal matures, may produce a cycle of progressive brain injury that in turn lowers resistance to future seizures. Over time, the progressive damage to the brain and concomitant lowering of seizure threshold establishes an epileptic state characterized by spontaneous recurrent unprovoked seizures. Moreover, recurrent seizures may also kindle epileptogenic areas of the brain, progressively increasing the likelihood and severity of future seizures. For the present, it is generally believed that kindling is uncommon in humans, but conclusive evidence is lacking [201].

Little is known of the physiological and cellular functions accounting for the developmentally regulated change in susceptibility to seizures and seizure-induced damage. The specific nature of seizure-induced synaptic rearrangement and neuronal injury are increasingly understood, at least in the hippocampus, but the specific mechanisms mediating these changes are unknown. For example, it is possible that differences in rates of maturation between cortical and subcortical regions may alter the balance of excitation and inhibition. Alternatively, developmentally regulated changes in gene expression may affect the excitability of neuronal populations and contribute to potential epileptogenesis. Coulter [202] has presented preliminary evidence that seizures alter receptor function in neurons. Research is needed to determine how such seizure-induced and developmentally - provoked changes alter gene expression, and which changes in gene expression contribute to establishment of the epileptic state.

The ultimate goal of individuals with epilepsy and of their physicians is the prevention of epileptogenesis and of seizure-induced brain injury. To this end, considerable work remains to be done. To begin with, the causal connection between early-life seizures - especially status epilepticus - and hippocampal sclerosis and temporal lobe epilepsy could be resolved with a prospective multicenter study. In this manner, it would be possible to determine whether seizures and status epilepticus cause acute hippocampal changes on MRI that predict the evolution of hippocampal sclerosis. Greater insight is also needed into the mechanisms underlying prolonged status epilepticus, in order to develop therapies to prevent or curtail these severe seizures. Finally, additional investigation into neuroprotective agents may permit the development of drugs that will reduce seizure-induced injury acutely, since the greatest injury occurs shortly after seizures [203, 204]. Furthermore, neuroprotective strategies may forestall epileptogenesis by preventing the structural and synaptic changes that likely contribute to epileptogenesis. Neuroprotection is currently an area of interest in almost all conditions that involve neuronal injury, and potential neuroprotective therapies for epilepsy overlap with neuroprotection strategies developed for other neurological disorders, such as stroke and neurodegenerative diseases [205]. Prevention of calcium entry through blockade of ligand- or voltage-gated calcium channels [91, 93], or by blockade of glutamate receptors [206] or by blockade of sodium channels to reduce neuronal excitability [207], may prevent the calcium-triggered cascade of events that culminate in cell death. Other methods of protecting neurons against the effects of seizures many include the use of antioxidants and inhibitors of nitric oxide [208, 209], neurotrophic factors [210], and apoptosis-related protease inhibitors [94]. As discussed earlier, hormonal influences [211, 212] may also attenuate seizure-induced damage, and are currently under investigation.

The benefits of neuroprotective strategies should ultimately be tested not only as regards the prevention of neuronal death, but also in the prevention of recurrent unprovoked seizures, i.e. epilepsy. Identification of effective agents in the prevention of epileptogenesis will require longitudinal laboratory and clinical studies tracking the occurrence of seizures after neurological injury. Recently Prasad et al. [213] found that MK801, a specific blocker of the NMDA-subtype of the glutamate receptor, and phenobarbital prevented the development of chronic epilepsy in rats, while phenytoin did not. Such laboratory investigations will play an important role in the near future, in testing compounds for the prevention of human epilepsy following neurological injury as in traumatic brain injury, stroke, or status-epilepticus. Rapid advances in the identification of genes contributing to epilepsy, along with traditional clinical methods detailing family and individual risk factors for recurrent seizures may, in the near future identify, those individuals at greatest risk of the development of epilepsy before seizures emerge. Future neuroprotective strategies hold promise for benefiting individuals at increased risk of developing epilepsy before permanent epileptogenic changes in the brain take hold.

Acknowledgements:

Supported by NIH grants 1K08-NS41340 (FAL) and NS-20253 (SLM). Dr Moshé is also the recipient of a Martin A. and Emily L. Fisher fellowship in Neurology and Pediatrics. The authors also wish to thank Dr Jana Veliskova for helpful discussions and comments on the manuscript.

Received March 13, 2002 / Accepted April 12, 2002