Address correspondence and reprint requests to Dr. C. Roch at Institut de Physique Biologique (UPRES-A 7004 ULP/CNRS), Faculté de Médecine, 4 rue Kirschleger, 67085 Strasbourg Cedex, France. E-mail: email@example.com
Summary: Purpose: In temporal lobe epilepsy, it remains to be clarified whether hippocampal sclerosis is the cause or the consequence of epilepsy. We studied the temporal evolution of the lesions in the lithium–pilocarpine model of epilepsy in the rat with magnetic resonance imaging (MRI) to determine the progressive morphologic changes occurring before the appearance of chronic epilepsy.
Methods: MRI was performed on an MR scanner operating at 4.7 T. We followed the evolution of lesions using T2- and T1-weighted sequences before and after the injection of gadolinium from 2 h to 9 weeks.
Results: At 2 h after status epilepticus (SE), a blood–brain barrier breakdown could be observed only in the thalamus; it had disappeared by 6 h. At 24 h after SE, edema was present in the amygdala and the piriform and entorhinal cortices together with extensive neuronal loss; it disappeared progressively over a 5-day period. During the chronic phase, a cortical signal reappeared in all animals; this signal corresponded to gliosis, which appeared on glial fibrillary acidic protein (GFAP) immunohistochemically stained sections as hypertrophic astrocytes with thickened processes. In the hippocampus, the correlation between histopathology and T2-weighted signal underscored the progressive constitution of atrophy and sclerosis, starting 2 days after SE.
Conclusions: These data show the reactivity of the cortex that characterizes the initial step leading to the development of epilepsy and the late gliosis that could result from the spontaneous seizures. Moreover, it appears that hippocampal sclerosis progressively worsened and could be both the cause and the consequence of epileptic activity.
The pilocarpine model, whether or not associated with lithium, is a well-studied model of temporal lobe epilepsy that reproduces most clinical and neuropathologic features of human temporal lobe epilepsy (1,2). In adult rats, pilocarpine leads to status epilepticus (SE), starting usually within 50–60 min after the injection, lasting for up to 8–12 h and followed by recurrent seizure episodes for up to 24–48 h. This acute seizure phase is followed by a “silent” seizure-free phase of a mean duration of 14–25 days. Thereafter, all animals exhibit spontaneous recurrent seizures (SRSs). This epilepsy lasts for the whole life of the animals (2,3).
In this model, neuronal damage is located mainly in the hippocampal formation, piriform and entorhinal cortices, septum, thalamus, amygdala, and neocortex (2,4). The extent of neuronal damage in the hippocampal formation correlates directly with the duration of the initial SE (5). The lithium–pilocarpine-induced neuronal injury occurs as a phenomenon of necrosis rather than apoptosis (6). Lesions in the hippocampus (CA1, CA3, and the hilus), amygdala, and piriform and entorhinal cortices begin to appear as soon as the first hour of SE and worsen or extend thereafter (7). During SE, brain areas where most damage occurs are those with highest expression of Fos (marker of cellular hyperactivation), acid fuchsin staining (marker of dying neurons), and largest metabolic activation (8–10).
An anatomic correlation has been found between large hypermetabolic levels measured during the second hour of seizures and neuronal damage assessed at 6 days after SE (10). At 14 days after SE (i.e. during the silent phase), cerebral metabolic rates for glucose (CMRglc) decrease in damaged forebrain areas that are involved in the circuitry of spontaneous seizures (cortical regions, hippocampal CA1 and CA3 areas, anterior and lateral thalamic nuclei) and increase in nondamaged brainstem areas involved in the remote control of epilepsy (substantia nigra reticulata, superior colliculus, red nucleus) (11). Later, the interictal period of the chronic phase of the epilepsy is characterized by normometabolic levels in the intact brainstem areas and hypometabolic levels in the damaged forebrain regions (12).
However, all the studies described above give only information about the nature of the regions injured or involved in the circuitry of the epilepsy. They do not provide information about the nature of the regions responsible for the induction of this epilepsy and about the changes that are the consequence of recurrent seizures. Moreover, all autoradiographic and histologic methods used in previous studies need to kill the animals, and therefore they represent either the SE, the silent phase, or the chronic epilepsy, but they do not allow any progressive study of the lesions and related phenomena occurring in diverse brain regions. For this reason, we chose to use magnetic resonance imaging (MRI), which allows a follow-up of the animals, thus giving a better insight into the lesional and plastic mechanisms underlying the establishment of temporal lobe epilepsy.
In the present study, we explored the temporal evolution of the MRI signal in adult rats from 2 h after the onset of lithium–pilocarpine-induced SE up to the occurrence of the SRSs. This procedure allowed a detailed follow-up of the progressive morphologic and lesional changes occurring during the acute, silent, and chronic phases that allow the establishment of the chronic epilepsy.
Animals and treatment
For the study, 35 adult male Sprague–Dawley rats (Janvier Breeding Center, Le Genest-St-Isle, France) weighing 300–350 g were used. All animals were maintained under standard laboratory conditions on a 12/12-h light/dark cycle.
Lithium chloride (3 mEq/kg; Sigma, St. Louis, MO, U.S.A.) was administered intraperitoneally to all rats 18 to 20 h before the subcutaneous injection of pilocarpine (25 mg/kg, Sigma) to the experimental group (30 rats) or saline to the control group (five rats). The animals received 1 mg/kg methylscopolamine (Sigma) 30 min before the convulsant to reduce the peripheral consequences of pilocarpine administration.
To improve survival, a first group of rats (n = 26) was put on a fixed standardized protocol of diazepam (DZP) injections: these animals received 2 mg/kg DZP (Valium; Roche, France) at 2 h after the onset of SE and 1 mg/kg at 6 h and 24 h after the onset of SE. A second group of animals (n = 4) received, as previously described in the literature, a protocol of DZP injections based on behavioral signs. This protocol consisted of a first deep intramuscular injection of 2 mg/kg DZP at 2 h after the onset of SE, followed by 1 mg/kg every 2–3 h as necessary (8,13–15).
Anesthesia for MRI was induced by an intramuscular injection of 37 mg/kg of ketamine (Ketalar; Parke-Davis, U.S.A.) and 5.5 mg/kg of xylazine (Rompun; Bayer, Germany). A 60-mm field of view, 256 × 256-pixel matrix, 1-mm slice thickness, and two excitations were used as imaging parameters. Two imaging sequences were used: first, a T2-weighted, spin-echo fast imaging method sequence (3,800/80) to detect the lesions, and second, a T1-weighted spin-echo method (500/22) 15 min after the injection of 0.5 mmol/kg gadolinium (Omniscan; Nycomed, Oslo, Norway) to assess the blood–brain barrier (BBB) breakdown. For the two protocols, the brain was scanned from the olfactory bulb to the brainstem by using consecutive 1-mm-thick coronal slices.
MRI was performed on a scanner operating at 4.7 T (SMIS) from all animals. Among them, 20 were killed for histologic analysis at different times: seven after the 24-h scan (five pilocarpine and two controls), seven after the 14-day scan (five pilocarpine and two controls), and six after the 9-week scan (five pilocarpine and one control). The five control animals underwent a first scan before the injection of lithium and saline and a second scan before killing. The 30 rats undergoing SE were observed by MRI before the lithium–pilocarpine injection and during
• the acute phase: 6 h (n = 30), 24 h (n = 30), and 30 h (n = 25) after SE;
• the silent phase: 2 (n = 25), 3 (n = 25), 7 (n = 25), 14 (n = 25), and 21 (n = 20) days after SE;
• the chronic phase: 5 (n = 20), 7 (n = 20), and 9 weeks (n = 20) after SE.
Only three animals in the whole group underwent a scan at 2 h after SE.
Quantification of neuronal damage
Determination of neuronal damage was performed at three different times after SE: 24 h (n = 5), 14 days (n = 5), and 9 weeks [i.e., at the end of the study (n = 5)]. The data were compared with one group of five control rats that received lithium and saline and were killed either at 24 h, 14 days, or 9 weeks after saline. Brains were removed, immediately frozen in isopentane, and stored at –30°C. Serial 20-μm sections were cut in a cryostat and allowed to dry for a few days before thionine staining. Quantification of cell density was performed with a 10 × 10 boxes 1-cm microscopic grid on coronal sections according to the stereotaxic coordinates of the rat brain atlas (16). Cell counts were performed twice in a blind manner and were the average of at least three values from two adjacent sections in each individual animal. Neurons touching the inferior and right edges of the grid were not counted. In thionine-stained sections, counts involved only neurons with cell bodies larger than 10 μm. Cells with small cell bodies were considered glial cells and were not counted.
GFAP and MAP2 immunohistochemistry
Immunohistochemical methods for the localization of GFAP (glial fibrillary acidic protein) and MAP2 (microtubule-associated protein), in sections adjacent to the ones used in the Nissl-stained procedure, were used to determine the location and extent of astrocytic and neuronal changes, respectively. The sections were incubated in primary antiserum to GFAP (rabbit antibody, diluted 1:500; Dako, Denmark) for 2 h or antiserum to MAP2 (rabbit antibody, diluted 1/10,000; personal gift) for 2.5 h and rinsed twice in phosphate-buffered saline (PBS). They were incubated for 1 h with the secondary antibody, biotinylated (goat anti-rabbit, dilution 1:400; Biosys BA 1000, Vector Laboratories, Burlingame, CA, U.S.A.), rinsed twice, and covered with the ABC reagent (1:50, Vectastain Kit; Vector) for 45 min. Sections were then rinsed twice and incubated for 5–8 min in a mixture of 0.02% diaminobenzidine, 0.5% nickel chloride, and 0.05% hydrogen peroxide in PBS. Thereafter, sections were dehydrated in ethanol and coverslipped.
One-way analysis of variance (ANOVA), followed by post hoc Fisher's LSD test, was used to compare neuronal cell counts in the control and three experimental groups.
Within 5 min after the injection of pilocarpine, rats developed diarrhea, piloerection, and other signs of cholinergic stimulation. During the following 15–20 min, they exhibited head bobbing, scratching, masticatory automatisms, wet-dog shakes, and exploratory behavior. Recurrent seizures started ∼20–25 min after pilocarpine administration. These seizures, which associated episodes of head and bilateral forelimb myoclonus with rearing and falling, progressed to SE at ∼50 min after pilocarpine, as previously described (2,4,17). After the initial phase of long-lasting, sustained seizure activity, recurrent seizures occurred less and less frequently but could still be recorded up to 24–48 h after the induction of SE by pilocarpine. In the rats that were on a fixed schedule of DZP injections, the silent phase lasted for a mean duration of 11 ± 4 days (8–19 days after SE), after which all animals started to exhibit SRSs. The average seizure frequency was two to three per week.
In the second group of rats to which diazepam was given according to the severity of behavioral signs, two rats that had received fewer injections of DZP entered the chronic phase earlier (6 and 7 days after SE); two other ones that had received more injections of DZP entered the chronic phase after a longer latent period (25 and 28 days).
Magnetic resonance imaging
None of the lithium–saline-treated rats showed any change on the T2- or T1-weighted signal at any time of the study.
During the acute phase
At 2 h after the onset of the SE, a BBB breakdown was observed only in the thalamus; this signal had disappeared by 6 h after SE (Fig. 1). At 24 h after SE, a transient but marked increase in the T2-weighted signal appeared in the thalamus, the amygdala, and the piriform and entorhinal cortices (Fig. 2A). This hypersignal detected at 24 h occurred in the absence of any obvious BBB breakdown, which could not be detected in any of the lithium–pilocarpine rats on the T1-weighted image after the injection of gadolinium (Fig. 1). The signal intensity was unchanged in the hippocampus at 6 h after the onset of SE and faintly increased by 24 h (Fig. 2A and B).
During the acute period, the intensity of the signal differed in the first and the second group of rats under different DZP regimens: in the group of rats receiving DZP injection according to their behavioral state, the hyperintensity on the T2-weighted images was inversely proportional to the number of injections of DZP the rat had received (i.e., the two rats that needed fewer injections of DZP presented a more intense signal).
During the silent and chronic phases
Two days after pilocarpine, a slight increase in the T2-weighted signal was seen in the dorsal and ventral hippocampus, and this signal began to be really intense by the end of the first week. The signal hyperintensity in the hippocampus and periventricular regions intensified progressively until the end of the study (Figs. 2 and 3). In spite of the difference in the intensity of the signal at 24 h after SE between the groups of rats under fixed or behaviorally related DZP injection protocol, the intensity of the signal in the hippocampus was similar in both groups of animals.
In the piriform and entorhinal cortices, and in the amygdala, the hyperintensity on the T2-weighted images disappeared gradually from ∼5 days after SE. It subsequently reappeared in these structures in almost all animals at the beginning of the chronic phase (i.e., at an average of 3 weeks after the SE; Figs. 2A and B and 3). In the thalamus, the hypersignal disappeared after 2 days and did not reappear during the chronic phase (Fig. 2A and B).
In areas CA1 and CA3a and b of the hippocampus, there was a progressive neuronal loss (Figs. 3 and 4). At 24 h, neuronal loss ranged from 13 to 35% compared with control levels but was significant only in CA3a. At 14 days, neuronal loss reached 30–57%. This decrease was significant in CA1 and CA3a compared with the control level and in CA1 compared with the loss at 24 h. At 9 weeks, the number of neurons was significantly decreased by 52–64% in CA1 and CA3a and b compared with the control level and also was significantly lower in CA1 and CA3a compared with the value at 24 h.
In the hilus of the dentate gyrus, the number of neurons decreased more rapidly than in the other hippocampal subregions. Neuronal loss reached 50% at 24 h and 70% at 14 days and 9 weeks. All values were significantly different from control levels, and the data at 14 days and 9 weeks were significantly lower than at 24 h (Figs. 3 and 4).
In the piriform and entorhinal cortices, the profile of neuronal loss was identical. There was a significant loss of neurons at 24 h after SE compared with the control group. This loss did not change with time, except in layers III/IV of the piriform cortex, where it worsened at 14 days. Neuronal loss was more dramatic in the piriform cortex (58% in layer II and 69% in layers III/IV) than in the entorhinal cortex (39% in layer II and 59% in layers III/IV; Figs. 3 and 5).
In the thalamus and amygdala, there also was a significant loss of neurons at 24 h after SE compared with the control group. This loss did not change with time in the thalamus (45%) but decreased significantly at 9 weeks in the amygdala compared with the value at 14 days (neuronal loss of 36–57%; Fig. 6).
We did not see any change at 24 h after SE in GFAP immunohistochemistry. At 14 days and 9 weeks, we observed morphologic changes in astrocytes, especially at the level of the cortex. These changes were mainly a hypertrophy of the cell bodies and cell processes of the astrocytes near the lesions rather than an increase in cell number (Fig. 7).
MAP2 immunohistochemistry showed the progressive loss of neurons at the level of CA1 and CA3 (Fig. 8). The loss was not very prominent at 24 h after SE. At this time, the dendritic tree appeared more injured than the neuronal cell bodies, mainly at the level of CA1. Then 14 days after SE, in CA1 as in CA3, the MAP2 immunohistochemical reaction appeared very low, which is characteristic of extensive neuronal loss.
Conversely, the MAP2 immunohistochemical reaction was impossible to assess at the level of the piriform and entorhinal cortices. Indeed, the very rapid necrosis of the tissue that was already extensive at 24 h after SE hindered the quality of the MAP2 immunohistochemical reactivity.
It has been well established that MRI is the most sensitive in vivo technique permitting the diagnosis of structural abnormalities that underlie seizure disorders. This great sensitivity is because the signal amplitude is related to the water environment of tissues (i.e., the concentration and mobility of water), but MRI fails to provide detailed information about the histopathologic nature of lesions. In almost all structural changes, including edema, gliosis, and neuronal loss, the concentration of free water is increased in the tissue. In the present study, in agreement with previously published data (2,6,12,18), the nature of lesions at different stages of the lithium–pilocarpine model of temporal lobe epilepsy has been clearly established in the 15 rats that underwent histopathologic examination (Fig. 3). In this context, MRI is a powerful technique allowing assessment of the temporal evolution of the different lesions.
Role of the BBB breakdown in the genesis of the lesions
Many observations in epilepsy patients (19–24), confirmed by animal studies (25–28), demonstrated that SE provokes transient BBB breakdown and vasogenic edema in the area of the seizure focus. It is commonly proposed that local (relative hypoxia secondary to increased ictal metabolism with an increase in local Pco2 and lactate) and systemic (increase in blood pressure and decrease in blood pH) physiologic changes induce a vasodilatation, resulting in increased regional blood flow and BBB breakdown (25,26,29).
In the present study, we observed a BBB breakdown during SE only at the level of the thalamus. The role of this phenomenon in the genesis of the lesions is not clear. However, some hypotheses can be advanced. First, the known anatomic connections of the thalamus to the limbic system and its excitatory input to these structures indicate its central role in limbic seizures. Indeed, thalamic stimulation yields short-latency excitatory responses in the entorhinal cortex and hippocampus (30–32). Moreover, in a number of studies using different models for limbic SE, the 2-deoxyglucose mapping of metabolic activity has underscored clearly the involvement of the midline thalamic nuclei (33–36). Second, we did not analyze BBB breakdown before 2 h or between 2 and 6 h after the onset of SE. The thalamus may be the structure most sensitive to SE and the only one with a BBB breakdown, but it is also possible that other structures may present a BBB breakdown but at a time that we did not explore. Third, MRI may be not sensitive enough or the contrast medium, gadolinium, too large to detect a small increase in permeability of the BBB.
Despite the presence of marked edema in the amygdala and piriform and entorhinal cortices at 24 and 48 h, we did not see any obvious BBB breakdown in those regions. These MRI observations were associated with early selective neuronal loss in the amygdala and the cerebral cortex. Of the various theories for the pathophysiology of the neuronal damage in SE, one suggests that excessive presynaptic neuronal activity causes postsynaptic neuronal damage (37,38). The excitotoxic phenomenon responsible for the excessive presynaptic activity is presumed to be the release of normal endogenous neurotransmitters (e.g., glutamate and aspartate) in excessive, neurotoxic amounts (37,39). This concept is supported by studies with kainate (an excitotoxic analogue of glutamate), which causes excessive neuronal excitation at distant postsynaptic sites, leading to neuronal damage (38).
Key role of the cerebral cortex in epileptogenesis
We saw a hypersignal during the acute and the beginning of the silent phase (i.e., at 24 h to 5 days after the SE) in the piriform and entorhinal cortices and in the amygdala. This signal disappeared by 5 days and reappeared at the onset of the chronic phase. We associated this hypersignal with a edema that is present at the time at which marked loss of neurons is found on thionine-stained sections in the same regions. In fact, most neuronal loss in those regions, mainly in the cortex, was complete as soon as 24 h after SE. As neuronal injury, in this model, has been shown to be a phenomenon of necrosis (6), the neurons of the piriform and entorhinal cortices undergo swelling during the first 24 h after SE, which increased the amount of free water. Because the signal disappeared thereafter, it means that the T2-weighted signal reflected edema rather than only the neuronal loss. In the future, it will be interesting to couple the MRI scans with magnetic resonance spectroscopy to distinguish neuronal damage and edema (40,41). In previous studies on kainate-induced SE, diffusion-weighted MRI allowed detection of early and minimal brain damage, whereas no such changes could be seen on T2-weighted images (42,43). The present data confirm this observation. Indeed, an early T2-weighted signal was detected in regions like the cerebral cortices and amygdala in which neuronal damage is rapid and massive. Conversely, in regions like the hippocampus, where the damage is more limited and progressive, the signal intensity does not change over the first 24 h, and in the latter case, diffusion-weighted MRI would probably be more appropriate to detect early changes.
With the onset of spontaneous recurrent seizures, the T2-weighted signal reappeared in the cortex. This can be correlated with the results obtained by GFAP immunohistochemistry, which shows swollen, reactive astrocytes, and underlines the glial nature of the late MRI signal in the cortex. The common reaction of astrocytes to injury in central nervous system is the striking hypertrophy and occasional proliferation of these cells (44–47). The hypertrophy involves an increase in the size of the soma as well as an increase in the length and width of astrocytic stellae, with a hallmark accumulation of intermediate filaments such as GFAP and an increase of free water. Studies on animals model of epilepsy show that an apparent gliosis is almost always present in brain regions that exhibit epileptiform activity because of neuronal injury as well as seizure activity per se (48–53). In kindled seizures (a nonlesional model), it has been shown that astrocytes present a prominent hypertrophy that is accompanied by a reorganization of astrocytic cytoskeleton. The changes in their morphology appear to be seizure-intensity dependent, occur early in the kindling process, and persist for weeks after the last seizure (51–53).
In the group of rats that received a number of DZP injections related to the behavioral state, we noticed that the intensity of the cortical edema seen at 24 h after the injection of pilocarpine was inversely proportional to the number of injections of DZP a given rat received. Moreover, it appeared that the rat that presented the most prominent edema at 24 h after SE had received fewer injections of DZP and entered the chronic phase, and thus presented SRSs earlier than rats that received a higher number of DZP injections. However, the more or less marked severity of the edema at 24 h after SE did not influence the volume of lesions in the hippocampus, neither in the group of rats with a fixed schedule of DZP injections nor in the group receiving DZP injections according to their behavioral state. In both cases, the piriform and entorhinal cortices were always the first structures affected, and the hippocampus reacted always after some delay. The early MRI signal and lesional process of the cortex compared with the later and prolonged neuropathology of the hippocampus also was reported recently in adult rats subjected to pilocarpine SE (54). Thus the present results suggest the predictive character of the early involvement of the cortex as well as the key role of the piriform and entorhinal cortices in epileptogenesis induced by the administration of lithium and pilocarpine. Conversely, the secondary involvement of the hippocampus seemed to be rather independent of the intensity of the cortical signal.
These results are in agreement with previous studies from our group showing that the rapid destruction of the entorhinal cortex induced by a pretreatment consisting of a series of maximal electroshocks before lithium–pilocarpine SE prevents the occurrence of SRSs (17,55), whereas the mechanical lesion of the same structure 4 days after SE does not prevent the occurrence of SRSs in adult rats (56). The prominent role of the entorhinal cortex in the primary step of the process of epileptogenesis appears also in our metabolic studies. Indeed, during SE, the entorhinal cortex is the structure the most largely activated area in adult rats, whereas in P10 rats, which do not become epileptic after lithium–pilocarpine SE, it is the least activated within the forebrain (10). Our data also are in accordance with a study performed in the rat hippocampal–entorhinal slice preparation in which an epileptiform activity can be induced by pilocarpine (57). This study indicates that epileptiform discharges originate in the entorhinal cortex, from where they project to the hippocampus proper via the perforant path.
Progressive setup of the hippocampal sclerosis
At the level of the hippocampus, the T2-weighed signal appeared ∼2 days after SE and gradually intensified until the end of the study, fully correlating with the progressive loss of neurons seen in the thionine-stained sections. Indeed, although neuronal loss was apparent in the hilus as soon as 24 h after SE, it became significant only at 14 days in CA1 and CA3a. CA3b was significantly injured only later, at 9 weeks after SE. Thus the T2-weighted signal observed on the MRI scan corresponds to the progressive constitution of atrophy and gliosis in the hippocampus.
Finally, our data confirm the results of a previous study (58) using the kainate model of epilepsy in mice, which leads to hippocampal sclerosis in two steps. The first period is characterized by an edema provoked by the neurotoxin underlying a period of limited histologic damage. The second phase is characterized by the extension of the damage, leading to a progressive and complete degeneration of the hippocampal CA1 area, partly spreading to CA3. This late phase is accompanied by the onset of the occurrence of SRSs and gliosis. Therefore, it appears that in these two models of human temporal lobe epilepsy, the development of hippocampal sclerosis is the consequence of the primary excitotoxic event, the SE, and the ongoing seizure activity during the chronic phase of SRSs.
Acknowledgment: We are very grateful to B. Guignard and A. Ferrandon for skillful technical assistance. This work was supported by a grant from the Hôpitaux Universitaires de Strasbourg, the Université Louis Pasteur, the Institut National de la Santé et de la Recherche Médicale, and the Fondation pour la Recherche Médicale.