On the relevance of prolonged convulsive status epilepticus in animals to the etiology and neurobiology of human temporal lobe epilepsy

Abstract

Convulsive status epilepticus (SE) in animals has been an invaluable experimental method for elucidating the role of excitatory synaptic activity in seizure-associated neuronal death (Meldrum et al., 1973; Olney et al., 1974; Nadler et al., 1978; Schwarcz et al., 1978; Ben-Ari et al., 1980; Schwob et al., 1980). Pioneering studies by Meldrum and colleagues established that prolonged convulsive seizures produced brain damage even when systemic physiological factors were normalized (Meldrum et al., 1973). Subsequent studies using the glutamate receptor agonist kainic acid (Olney et al., 1974) led to the hypothesis that excessive glutamatergic presynaptic discharge per se was the cause of “excitotoxic” postsynaptic neuronal death (Schwob et al., 1980). Studies using a variety of chemoconvulsants or electrical stimulation finally provided conclusive evidence that prolonged presynaptic excitation, regardless of how it is initiated, can be neurotoxic to postsynaptic target cells (Nadler et al., 1978; Ben-Ari et al., 1980; Olney, 1981; Sloviter and Damiano, 1981; McIntyre et al., 1982). Early anecdotal observations that spontaneous seizures developed in rats that survived convulsive SE led to suggestions that these rats might be useful models of human temporal lobe epilepsy (TLE; Pisa et al., 1980; Nadler, 1981; Tremblay and Ben-Ari, 1984). Although SE-induced brain damage in animals is obviously an excellent model of SE-induced brain damage in humans, recent data indicate that convulsive SE fails to reproduce the main features of human TLE (Harvey and Sloviter, 2005). Characteristic features of human refractory TLE include extensive hippocampal atrophy and limited extrahippocampal damage (Margerison and Corsellis, 1966), as well as seizure onset from the hippocampus and closely related structures (Spencer and Spencer, 1994). However, most animals exhibit the reverse pattern of pathology after convulsive SE in rats, i.e., extensive extrahippocampal damage, but limited hippocampal atrophy (Schwob et al., 1980; Harvey and Sloviter, 2005). In addition, spontaneous seizures in these animals appear to arise from outside the hippocampal formation, possibly as a result of extensive damage to thalamic and extratemporal cortical structures (Bertram et al., 2001; Harvey and Sloviter, 2005). The only depth recordings in awake epileptic animals of which we are aware suggest that the hippocampus may be only secondarily involved, if it is involved at all (Harvey and Sloviter, 2005). Failure to make the distinction between epileptogenesis occurring somewhere within the brain, and the assumption that the hippocampus becomes “epileptic,” or has undergone “hippocampal epileptogenesis” following convulsive SE, has tended to confuse the issue (Tauck and Nadler, 1985; Buckmaster and Dudek, 1997; Esclapez et al., 1997). In fact, Olney and colleagues noted in their original histological study of SE-induced brain damage that prolonged convulsive SE produced far more extrahippocampal than hippocampal damage, and questioned whether such extensive brain damage could constitute a useful model of human TLE (Schwob et al., 1980). We were motivated to revisit this issue because virtually all of the hippocampal changes that we produced by controlled, focal activation of the hippocampus, including extensive hilar neuron loss, granule cell disinhibition, and principal cell hyperexcitability (Sloviter, 1991), have been difficult to reproduce in any post-SE animal model (Milgram et al., 1991; Buckmaster and Dudek, 1997; Esclapez et al., 1997; Ratzliff et al., 2002; Zappone and Sloviter, 2004). The question for reconsideration, therefore, is what state does prolonged convulsive SE actually model? That is, do post-SE rats bear any significant resemblance to TLE patients, who appear relatively normal on neurological examination, and who exhibit temporal lobe-onset seizures as their primary or sole clinically detectable neurological defect? Or, do post-SE rats more closely resemble patients with severe brain damage, who exhibit obvious behavioral abnormalities and have frequent generalized seizures as part of a larger constellation of clinical abnormalities? In their hallmark study of epileptic brains at autopsy, Margerison and Corsellis noted that, “gross structural abnormalities were virtually confined to the severely subnormal…” (Margerison and Corsellis, 1966). Thus, it would appear that extensive brain damage in humans and animals has little resemblance to the relatively subtle features of human TLE (Lehericy et al., 1997; Van Paesschen et al., 1997; Liu et al., 2002; Salmenpera et al., 2005). To address these issues, and develop animal models that reproduce some of the main features of human TLE, we focused on trying to understand why most patients with refractory TLE exhibit extensive hippocampal atrophy and limited extrahippocampal damage, whereas post-SE animals exhibit minor hippocampal atrophy and extensive extrahippocampal damage. We hypothesized that human-pattern hippocampal sclerosis is the result of less than maximally intense excitation that activates the hippocampus effectively, but stays sequestered within temporal seizure circuits. According to this view, convulsive SE evokes seizure activity predominantly in extrahippocampal pathways (Schwob et al., 1980), or is lethal before the hippocampus receives the duration of excitation needed to produce classical hippocampal sclerosis. In this presentation, two newly developed animal models are described for the first time. Both models avoid prolonged convulsive SE, and closely approximate the pathological and epileptogenic features of the two main TLE patient populations. These patient populations consist of: 1) the majority of TLE patients, which presents with no detectable structural abnormalities on routine imaging (Van Paesschen et al., 1997; Liu et al., 2002; Salmenpera et al., 2005); and 2) a smaller population, most of which present with refractory TLE, hippocampal atrophy, and other relatively subtle structural abnormalities (Lehericy et al., 1997). Both models involve perforant pathway stimulation for the duration of the experiment, which drives the hippocampus to discharge throughout the initial insult episode. This approach is different from methods that use chemoconvulsants or electrical stimulation to initiate convulsive SE, but which then permit seizures to continue in a “self-sustained” manner (Lothman et al., 1989). Self-sustained SE results in sporadic hippocampal involvement as seizures take different routes and involve different brain structures in different animals (Sloviter et al., 2003). Both of the models we have developed have the advantages of minimal variability, minimal lethality, spontaneous hippocampal-onset seizures verified by hippocampal depth recordings in the awake state, and human-pattern hippocampal sclerosis (endfolium sclerosis or classical hippocampal sclerosis), all without involving convulsive SE. Although patients often exhibit asymmetrical brain damage (Margerison and Corsellis, 1966), we used bilateral stimulation for several reasons. First, our goal has been to minimize variability in animal models, and unilateral stimulation produces variable contralateral injury (Sloviter, 1991). Second, the presence of one relatively undamaged hippocampus may delay clinical seizures, or impede seizure spread, possibly contributing to variability in both the extent of the initial injury and the rate of spontaneous seizures. Third, and perhaps most importantly, the presence of strong commissural connections in rodents (Sloviter, 1991) means that bilateral stimulation activates each hippocampus more strongly and uniformly than unilateral stimulation, because each hippocampus is doubly activated by both the ipsilateral and the contralateral inputs. The first animal model involves bilateral, intermittent perforant pathway stimulation for 24 h under urethane anesthesia, which prevents behavioral SE, as previously described for unilateral stimulation (Sloviter and Damiano, 1981; Sloviter, 1991). These animals receive 10-s trains of 20-Hz stimuli once per min plus continuous 2-Hz paired-pulse stimulation throughout the stimulation period. We now report that all of these bilaterally stimulated animals exhibit spontaneous, hippocampal-onset, Stage 3–5 behavioral seizures. Although the ultimate source of seizures is unknown, we call them “hippocampal-onset” seizures because live depth recording shows that dentate granule cell and hippocampal pyramidal cell epileptiform discharges reliably occur prior to the first detectable behavioral seizure manifestations. This model exhibits focal hilar neuron loss and incomplete hippocampal CA3 pyramidal neuron loss (together called “endfolium sclerosis” in humans; Fig. 1B), as well as a pattern of relatively subtle pathology in the amygdala, thalamus, and entorhinal cortex similar to that observed by Margerison and Corsellis in epilepsy patients postmortem (Margerison and Corsellis, 1966). The observation that these animals with relatively subtle brain damage are nonetheless chronically epileptic suggests that this model may represent the large TLE patient population without extensive hippocampal atrophy (Van Paesschen et al., 1997; Liu et al., 2002; Salmenpera et al., 2005). Human-pattern endfolium sclerosis or classical hippocampal sclerosis after prolonged nonconvulsive status epilepticus (SE). (A): Sham control hippocampus 241 days post-electrode implantation; NeuN immunostaining, which marks hippocampal neurons. dg: dentate gyrus; h: hilus; CA1, CA2, CA3: pyramidal cell layer subregions. (B): Endfolium sclerosis 227 days after bilateral perforant pathway stimulation for 24 h (10-s trains once per min plus continuous 2-Hz paired-pulse stimulation) under urethane anesthesia. Note loss of neurons in the hilus and CA3 region. (C): Classical hippocampal sclerosis 53 days after bilateral perforant pathway stimulation for 8 h (10-s trains once per min) in awake rats, which did not produce convulsive SE. Note virtually total loss of pyramidal layer neurons except for surviving CA2 pyramidal cells (asterisk). The second model involves an entirely new approach in awake rats, and is based on the realization that convulsive SE, once begun, is an often lethal, “all-or-none” phenomenon that cannot be modulated or controlled by the experimenter. Therefore, we tested the hypothesis that decreasing the intensity of excitation in awake animals would: 1) avoid convulsive SE and frequent mortality; 2) permit a longer duration of focal temporal lobe excitation to be tolerated; and 3) reproducibly result in classical hippocampal sclerosis not produced by 2–4 h of convulsive SE (Harvey and Sloviter, 2005; Sloviter, 2005). In this model, awake, chronically implanted rats received 10-s trains of 20-Hz stimuli once per min for 8 h, which reliably evoked focal hippocampal discharges that did not spread sufficiently to cause convulsive SE. By forcing the hippocampus to discharge intermittently throughout the initial injury episode, and by stimulating at a moderated intensity that apparently keeps seizure activity sequestered within the temporal “seizure circuit,” brain damage remains relatively focal, convulsive SE is avoided, death does not occur, and there is minimal interanimal variability. All animals exhibit spontaneous hippocampal-onset, Stage 3–5 seizures confirmed by continuous (24/7) depth recording and video monitoring, and the pathology reliably reproduces classical hippocampal sclerosis (Fig. 1C). This model may represent the smaller patient population with refractory TLE and classical hippocampal sclerosis (Lehericy et al., 1997). These results have implications for epilepsy model development, and suggest an epileptogenic role of nonconvulsive status epilepticus (NCSE) in humans. First, the observation that human-pattern hippocampal sclerosis and hippocampal-onset epilepsy are uniquely produced by moderate, rather than severe, excitation suggests that human hippocampal sclerosis, when it does occur, may be the result of prolonged, focal excitation that does not progress to generalized convulsive SE (VanLandingham et al., 1998), and may often be cryptic clinically. Second, when convulsive SE is initiated chemically, or by electrical stimulation, and allowed to continue for hours in a “self-sustained” manner (Lothman et al., 1989), seizure activity propagates in different pathways in different animals (Schwob et al., 1980), frequently bypasses the hippocampus (Sloviter et al., 2003), and results in significant inter-animal variability in the location and extent of brain damage (Brandt et al., 2003; Sloviter et al., 2003). Conversely, focal excitation of the perforant pathway throughout the experimental period tightly controls the magnitude, duration, and route of focal excitation, consistently resulting in both hippocampal injury and hippocampal epileptogenesis. Third, the availability of new in vivo models that closely resemble the human neurological condition provides an opportunity to investigate epileptogenic mechanisms that have been difficult to study in animals subjected to convulsive SE. Clearly, pinpointing the epileptogenic regions and elucidating epileptogenic mechanisms should be more feasible in epileptic animals with highly restricted, human-pattern pathology than it is in animals with severe brain damage and often excessively frequent spontaneous seizures of unknown origin (Harvey and Sloviter, 2005). Finally, minimal variability in experimental animals may make studies of neuroprotection and antiepileptogenesis feasible for the first time, since the variability seen after convulsive SE makes it difficult to show that any treatment has a statistically significant protective effect. In summary, prolonged focal, nonconvulsive SE produces animals that resemble the human neurological disorder more closely than animals subjected to prolonged convulsive SE, suggesting that the pathology of human TLE is caused by less than maximally intense excitation that usually stays sequestered within temporal lobe seizure circuits. The use of these in vivo experimental methods avoids the variability, lethality, and the confounds of interpretation that chemoconvulsants and self-sustained convulsive SE inevitably involve, which should make mechanistic studies of epileptogenesis, neuroprotection, and antiepileptogenesis more feasible and interpretable than currently possible. This work was supported by NIH grant NS18201 from the National Institute of Neurological Disorders and Stroke.