address: Mariko Osawa, National Epilepsy Center, Shizuoka Higashi Hospital, Shizuoka Prefecture, Urushiyama 886, Japan. Email: email@example.com
Abstract Repeated electrical stimulation of limbic structures has been reported to produce the kindling effect together with morphological changes in the hippocampus such as mossy fiber sprouting and/or neuronal loss. However, to argue against a causal role of these neuropathological changes in the development of kindling-associated seizures, we examined mossy fiber sprouting in amygdala (AM)-kindled rats using Timm histochemical staining, and evaluated the hippocampal neuronal degeneration in AM-kindled rats by terminal deoxynucleotidyl transferase-mediated digoxigenin-11-dUTP nick end labelling (TUNEL). Amygdala kindling was established by 10.3 ± 0.7 electrical stimulations, and no increase in Timm granules (neuronal sprouting) was observed up to the time of acquisition of a fully kindled state. However, the density and distribution of Timm granules increased significantly in the dentate gyrus compared with unkindled rats after 29 after-discharges or more than 10 kindled convulsions. In addition, no significant increase in TUNEL-positive cells was found in the hilar polymorphic neurons or in CA3 pyramidal neurons of the kindled rats that had fewer than 29 after-discharges. However, a significant increase of TUNEL-positive cells was found in the granule cell layer in the dentate gyrus of the stimulated side after 18 after-discharges or 10 kindled convulsions. Our result show that AM kindling develops without evidence of mossy fiber sprouting, and that mossy fiber sprouting may appear after repeated kindled convulsions, following death of the granule cells in the dentate gyrus.
Mesial temporal sclerosis is the most commonly detected lesion in surgically resected specimens from patients with intractable temporal lobe epilepsy. It is characterized by neuronal loss and gliosis involving the hippocampus and other mesial temporal structures. Recent studies have demonstrated aberrant axonal sprouting of the granule cells (mossy fiber sprouting, MFS) both in animal epilepsy models1–6 and in patients with temporal lobe epilepsy.6–8
Kindling is an experimental epilepsy model. In this model, progressive and long-lasting functional epileptogenic changes are produced by repeated subthreshold electrical stimulation in a certain cerebral region,9 without accompanying non-specific tissue damage. The merit of using kindling as an epilepsy model is that neuronal plasticity associated with epileptogenicity can be followed with time.
Recent studies showed that kindling induces lasting histological changes in the brain, including MFS in the dentate gyrus (DG) or CA3 area and a decrease in hilar neuronal density.4,10–12 The overall role of MFS in the kindling effect is not clear. Some reports proposed that MFS might modify the sequence of synaptic events in the DG and induce recurrent excitation to generate epileptic activities,13–15 while others suggested that MFS is not essential for epileptogenesis or the induction of kindling.16–19 Nevertheless, these findings did not clarify whether MFS is indispensable for the development of kindling and whether MFS is a consequence of kindling-induced neuronal loss in the hippocampus.
Two hypotheses regarding the mechanism of development of MFS have been proposed. One is that MFS is a secondary consequence of axonal degeneration induced by neuronal loss in the hilum,10,14,20 and the other is that the aberrant axonal outgrowth probably arises from newly generated granule cells as a result of abnormal neurogenesis induced by amygdala (AM) kindling.21
Hippocampal neuronal death in the kindled rats has been suggested by several studies. Cavazos et al. reported a progressive decrease of neuronal density in the limbic structures.11 Moreover, recent studies showed that hippocampal neuronal death had the morphological feature of apoptosis in animal models of epilepsy.22,23 Sloviter et al. also reported death of dentate hilar neurons and pyramidal neurons in the hippocampus with consequent necrotic changes resulting from repetitive stimulation of the perforant path.24
Using terminal deoxynucleotidyl transferase-mediated digoxigenin-11-dUTP nick end labelling (TUNEL) and silver impregnation methods, Pretel et al. reported that rapid electrical stimulation of the entorhinal cortex induced hippocampal neuronal death mainly as apoptosis and partially as necrosis.25 However, they used a status epilepticus model and not a kindling model; moreover, the time of cell death was not reported.
In the present study, we used AM-kindled rats to examine the following: (i) relationship of MFS with the development of kindling, (ii) relationship of neuronal death with the development of kindling and (iii) relationship of MFS with dentate neuronal loss.
Animals and surgical procedures
Ninety-two male Sprague-Dawley rats (Funabashi Co., Japan) weighing 270–330 g at the time of surgery were used. A rat under sodium pentobarbital anesthesia (50 mg/kg, i.p.) was placed in a stereotaxic holder. The electrodes were tripolar stainless steel (diameter 0.15 mm) insulated with Teflon, and were implanted in the left basolateral AM (0.6 mm posterior and 4.8 mm lateral to the bregma, and 7.8 mm ventral to the dura)26 for electrical stimulation and electroencephalographic (EEG) recording. The animals were allowed 1 week postsurgical recovery before the start of kindling sessions.
The rats were stimulated for 1 s with a train consisting of 60 Hz, 1 msec positive and 1 msec negative biphasic square waves at an intensity of 200 μA, once daily to induce after-discharges (AD). The behavioral seizures were classified according to Racine’s criteria:27 stage 1, mouth and facial twitching; stage 2, head nodding; stage 3, forelimb clonus; stage 4, rearing; stage 5, rearing and falling.
Kindling indexes were evaluated by the total number of AD, total AD duration, number of all convulsions, number of partial convulsions, number of generalized convulsions, and number of stage 5 seizures. These were used to examine the relationship between kindling and MFS. Kindling was judged complete when all the animals manifested stage 5 seizures after stimulation on 3 consecutive days.
The rats were sampled for Timm and TUNEL studies after various days of stimulation, and were grouped according to the total number of AD recorded. Timm histochemistry was examined after five, 11, 18 and 29 AD [five AD (n = 8), 11 AD (n = 5), 18 AD (n = 6) and 29 AD (n = 6) groups]. TUNEL immunohistochemistry was examined after seven, 18 and 29 AD [seven AD (n = 6), 18 AD (n = 14) and 29 AD (n = 7) groups]. Two additional groups of rats that manifested 10 (n = 7) and 30 stage 5 seizures (n = 9) were also studied by TUNEL immunohistochemistry.
Control rats received electrode implantation. Some were not stimulated (Control 1; n = 6 for Timm, n = 9 for TUNEL) and some were stimulated with a 1-s train of 3 Hz, 20 msec positive and 20 msec biphasic negative square waves at an intensity of 200 μA once daily (Control 2; n = 4 for Timm, n = 5 for TUNEL). The latter stimulation is insufficient to induce AD, but this stimulation parameter corresponds to the same amount of current for high-frequency stimulation, making it possible to obviate the effect of electrical stimulation.
Timm histological procedure
Mossy fiber reorganization was evaluated by Timm staining, a histochemical technique that labels the terminals of mossy fibers that have a high content of Zn ions.27,28
Two weeks after the last stimulation, the rats were deeply anesthetized with sodium pentobarbital (70 mg/kg) and perfused transcardially with 200 mL of 0.16% Na2S in 0.1 mol/L phosphate buffer (PB) solution, pH 7.3, followed by 400 mL of 3% glutaraldehyde and 0.16% Na2S in 0.1 mol/L PB, and then 300 mL of 15% sucrose and 0.16% Na2S in 0.1 mol/L PB. The brains were removed, immersed for 2 h in the last perfusion solution for cryoprotection, frozen in dry ice, and cut into 20 μm horizontal sections with a microtome. The sections were mounted on gelatin-coated slides and developed in a solution containing 75 mL of 20% gum arabic, 15 mL of a 2% hydroquinone and 3% citric acid mixture, and 1.5 mL of 10% silver nitrate, in the dark for about 1 h. The reaction was stopped by immersing in 5% sodium thiosulphate.
TUNEL immunohistological procedures
Neuronal death was evaluated by the TUNEL immunohistochemical method.29 In a typical experiment, kindled rats and Control 2 rats were deeply anesthetized with sodium pentobarbital (70 mg/kg) 18 h after the last stimulation, and perfused transcardially with 300 mL of 0.01 mol/L PB solution containing 0.9% NaCl (PBS), pH 7.4, followed by 300 mL of a fixative containing 0.2% picric acid, 4% paraformaldehyde and 0.26% glutaraldehyde in 0.1 mol/L PBS. After perfusion, the brains were removed and immersed overnight in the same fixative without glutaraldehyde at 4°C. After cryoprotection by overnight immersion in 0.1 mol/L PBS containing 15% sucrose, the tissues were frozen with dry ice and cut into 20 μm coronal sections. These sections were then placed in 0.1 mol/L PBS and preserved at 4°C. The free-floating sections were first digested with 0.1 U/mL papain (Sigma, St. Louis, MO, USA) in 0.1 mol/L PBS for 10 min at room temperature. Endogenous peroxidase activity was quenched by 0.5% H2O2 in 0.1 mol/L PBS for 30 min at room temperature. Sections were first incubated with a tailing buffer (0.1 mol/L sodium cacodylate, 1 mmol/L CoCl2, and 0.1 mmol/L dithiothreitol in distilled water) for 20 min at room temperature, and then with terminal deoxynucleotidyl transferase (TdT) buffer [TdT (0.4 U/mL, Takara, Japan), digoxigenin-11-dUTP (0.01 nmol/mL, Boehringer Mannheim, Indianapolis, IN, USA) and 10 × tailing buffer in distilled water] at 37°C for 60 min in a humidified chamber. The TdT buffer reaction was stopped by three washes in 1 × sodium/sodium citrate buffer (0.3 mol/L NaCl, 0.03 mol/L sodium citrate, 10 min) followed by one wash in 0.1 mol/L PBST (0.3% Triton X-100 in 0.1 mol/L PBS, pH 7.4, 10 min). The sections were incubated overnight with mouse antidigoxigenin monoclonal antibody (1 : 2000 dilution, Boehringer Mannheim Biomedica, Mannheim, Germany) at 4°C. The sections were then incubated with biotinylated antimouse IgG-rat adsorbed antibody (1 : 1000 dilution, Vector Laboratories, Burlingame, CA, USA) at room temperature for 1 h, washed three times in 0.1 mol/L PBST (10 min), incubated in avidin-biotin peroxidase (Vectastain ABC Elite, Vector Laboratories) at room temperature for 1 h, and processed with diaminobenzidine (DAB) solution [0.02% DAB (Sigma), 0.045% H202 and 0.3% nickel ammonium sulphate in 50 mmol/L Tris-HCl buffer, pH 7.6]. The sections were mounted on gelatin-coated slides. Finally, the sections were counterstained with methyl-green.
Negative control sections were incubated with only the tailing buffer, and positive control sections were first incubated with DNase 1 (Takara Co., Kusatu, Japan) for 10 min at room temperature and then in TdT buffer as described above. As positive controls for TUNEL, a deoxyribonuclease 1-treated section and a section of a 3-month-old rat killed 7 days after kainic acid injection (12 mg/kg in 0.9% NaCl, i.p), were used to demonstrate the morphological features of TUNEL-positive neurons in the hilus and to confirm the validity of the method.
From the 18-AD group, rats were processed 2 h (n = 4), 6 h (n = 4) and 18 h (n = 6) after the last stimulation to examine the effect of sampling time on the results of neuronal death.
Scoring for Timm histochemistry
The distribution of Timm granules was scored in the DG (Fig. 1) according to the criteria of Cavazos et al.30 as follows: 0, no granules between the tips and crest of the DG; 1, sparse granules in the supragranular layer in a patchy distribution between the tips and crest of the DG; 2, more numerous granules in the supragranular region in a continuous distribution between the tips and crest of the DG; 3, prominent granules in the supragranular layer in a continuous pattern between tips and crest with occasional patches of confluent granules between the tips and crest of the DG; 4, prominent granules in the supragranular layer that form a confluent, dense, laminar band between the tips and crest of the DG; 5, confluent, dense, laminar band of granules in the supragranular layer that extends into the inner molecular layer.
Quantification of TUNEL-positive neurons
From the dorsal hippocampus of each animal, sections of the granular cell layer, the hilar polymorphic neurons and the CA3 pyramidal neurons in the DG were processed for TUNEL immunohistochemistry, and were examined at a magnification of × 200 (Fig. 2). TUNEL-positive cells observed in the border zone (between the granule cell layer and hilar polymorphic layer or between the hilar polymorphic layer and CA3 pyramidal cell layer) were counted as one cell each for the two adjacent layers.
Statistical analysis for Timm score and TUNEL-positive cells
Quantitative assessments of Timm scores and TUNEL-positive cells were performed independently by three observers who were blind to the identity and seizure history of each case. The mean Timm scores and the quantified TUNEL-positive neurons were statistically analyzed by one-way analysis of variance. A post-hoc comparison using Fisher’s protected least significant difference or Mann–Whitney’s U-test was done and the correlation between kindling development and Timm score was determined by Spearman’s correlation coefficient (Spearman’s correlation coefficient by rank; rs) using StatView software (Abacus Concepts Inc., Berkeley, CA, USA). Differences were considered significant at P < 0.05.
Table 1 shows the total number (mean ± SEM) of AD and total number of stage 5 seizures observed in the kindling-induced groups. Kindling was established when the total AD number averaged 10.3 ± 0.7. Therefore, the kindling-induced groups (5–29 AD) were examined in the time course covering the process of kindling development and after attainment of a fully kindled state.
Table 1. Results of kindling
Total no. AD
Total no. St-5s
Value represents mean ± SD. AD, after-discharges; St-5, stage-5 seizures.
5-AD (n = 8)
0.1 ± 0.4
7-AD (n = 6)
0.5 ± 0.8
11-AD (n = 5)
1.5 ± 1.7
18-AD (n = 6 for Timm n = 14 for
9.6 ± 4.4
29-AD (n = 6 for Timm n = 7 for
15.3 ± 7.5
10 St-5 (n = 7)
23.2 ± 6.1
30 St-5 (n = 9)
37.4 ± 3.1
Timm score and progression of kindling
Kindling was established after 10.3 ± 0.7 (mean ± SEM) AD. Aberrant reorganization of mossy fibers in the supragranular layer of the DG was detected by Timm staining and quantified by Timm score (Table 2, Fig. 3). To relate the Timm scores with seizures, the 25 rats examined for Timm histochemistry were also stratified into three groups according to the number of stage 5 seizures manifested [0 (n = 11), < 10 (n = 8), and ≥ 10 (n = 6) stage 5 seizure groups] (Table 3).
Table 2. Timm score
Value representes mean ± SEM. Cont-1, unstimulated control; Cont-2, non-kindling stimulated control; AD, after-discharges. *P < 0.05 vs Cont-2, P < 0.01 vs Cont-1, 5-AD, 11-AD and 18-AD groups. **P < 0.05 vs Cont-2 and 11-AD groups, P < 0.01 vs Cont-1, 5-AD and 18-AD groups.
The mean Timm scores increased significantly on both the ipsilateral and contralateral sides of electrode implantation in the 29 AD group compared to the control (Table 2) and in the group that manifested more than 10 stage 5 seizures (Table 3).
No difference was found between the control groups and the five AD, 11 AD or 18 AD group or between the seizure groups that manifested less than 10 stage 5 seizures.
Figure 3 shows a high magnification of the Timm-stained horizontal section on the ipsilateral side of electrical stimulation in the 18 AD and 29 AD groups. Sparse Timm granules were observed in the rat that had 18 AD and eight stage 5 seizures (Fig. 3a). The rat in the 29 AD group that had manifested 21 stage 5 seizures (Fig. 3b) showed prominent Timm granules as a dense laminar band in the supragranular layer. In this experiment, the Timm score after attainment of a fully kindled state was significantly higher than the control score.
Correlation between kindling indexes and Timm score
At least 29 AD or more than 10 stage 5 seizures were required to induce MFS. Table 4 shows the relationship between the kindling indexes and MFS.
Table 4. Correlation between Timm score and kindling indexes
A positive correlation was observed between the number of convulsions and Timm score on the ipsilateral side (rs = 0.499), and between the number of stage 5 seizures and Timm score on the ipsilateral (rs = 0.525) as well as that on the contralateral sides (rs = 0.436). Weak correlation was found between Timm score and the total number of AD or AD duration.
TUNEL immunoreactions were determined in three regions of the DG; the hilar polymorphic cell layer, the CA3 pyramidal cell layer and the granular cell layer in the DG (Fig. 2). TUNEL-positive neurons were calculated in the kindled groups and controls (Fig. 4). There was no difference in the number of TUNEL-positive cells between Control 1 and Control 2 rats. While no difference in the number of TUNEL-positive cells (Fig. 4) was observed in the hilar region or CA3 area in any of the kindling-induced groups compared with the controls, significant increases in TUNEL-positive cells were found in the dentate granule cell layer in the 18 AD group sampled 18 h after stimulation and in the group that had 10 stage 5 seizures.
Figure 5 shows representative examples of the TUNEL-stained DG of the dorsal hippocampus. There were markedly more TUNEL-positive cells in the dentate granular layers in the rat that had 18 AD sampled 18 h after stimulation (Fig. 5b) group than in the Control 2 (Fig. 5a) rat.
The purpose of the present study was to examine the following two possibilities: whether kindling develops without evidence of MFS in the DG, and whether MFS is induced without evidence of neuronal death in the DG. To examine these possibilities, we discuss the results of the present study in respect to three areas: first to consider the causal role of MFS in the development of kindling, next to examine the timing of neuronal death in the DG, and last to consider the relationship between MFS and neuronal death in the DG.
No study to date has examined the temporal relationship between the development of kindling and the evidence of MFS. The present study reconfirmed previous reports that kindling induces MFS in the supragranular layer of the DG in fully kindled animals.4,30,31 No increase in MFS in the DG was observed following electrical stimulation that does not induce AD. Although AM kindling was completed after an average of 10.3 ± 0.7 AD, MFS was only observed in animals that had accumulated more than 29 AD and had manifested over 10 stage 5 seizures of kindled generalized convulsions. From these results, it seems reasonable to conclude that MFS is not associated with the development of AM kindling or the acquisition of susceptibility to lasting and generalized seizure.
There is a strong correlation between MFS and the number of kindled convulsions, especially stage 5 generalized convulsions. Since MFS also occurred in a model of transient ischemia,32 MFS observed in kindling may be a secondary consequence of brain ischemia induced by repeated generalized seizures. The possibility of MFS playing a role in inducing spontaneous seizures after the kindled convulsions requires further investigation.
Recently, Cavazos et al. reported a progressive decrease in the density of hilar neurons in the DG of the kindling model, which may be the result of neuronal death.11 Sloviter et al. showed intermittent stimulation of the perforant path for 24 h induced apoptosis in the CA3 pyramidal neurons and granule cells in the DG, and the polymorphic neurons in the hilus became necrotic after 12 h of intermittent stimulation.24 Pretel et al. also reported apoptosis and necrosis in the hippocampus and other limbic structures after 5–85 stage 5 seizures induced by perforant path kindling.25 However, the temporal relation between induction of kindling and neuronal death in the DG is unclear.
In the present study, neuronal death was examined using TUNEL staining, which allows visualization of the 3′-OH ends of fragmented DNA.29 Recent findings indicate that large base-pair DNA fragments are generated in necrosis and apoptosis,33 and that the TUNEL immunostaining method can detect the fragmented DNA in dying neurons not only in apoptosis but also in necrosis. Using this method, we found no neuronal death in the hilar polymorphic layer and CA3 area of DG even after inducing 29 AD, which was sufficient to produce MFS, although there was neuronal death in the granule cell layer on the ipsilateral side after inducing 18 AD or 10 stage 5 kindled convulsions, and on the contralateral side after inducing seven AD or 10 stage 5 convulsions. These results showed that neuronal death in the DG did not occur during the process of development of kindling, but neuronal death was induced in the granule cell layer after repeated occurrence of 9–10 kindled convulsions when the kindling effect was fully established.
Two hypotheses of the relationship between neuronal death and MFS in the DG have been proposed. First, Cavazos and Sutula reported that progressive decrease of neuronal density (presumably reflecting neuronal death) in limbic kindling resulted in denervation, and that might promote mossy fiber axonal sprouting.34 Second, Parent et al. reported that although cell division of dentate granule cells is normally continued up to adulthood in rats, kindling induced abnormal cell division in the granule cell layer, and the newly generated cells sent aberrant axons into the molecular layer, which was MFS.35 They also demonstrated that abnormal cell division in the DG was induced only after nine or more kindled generalized convulsions. According to Gould and McEwen, dentate granule cell progenitors required a signal from dead cells to initiate cell division.36 Our results showed that neuronal death in the DG occurred after 9–10 repeated stage 5 kindled generalized convulsions, suggesting that the neuronal death may induce a signal for proliferation of the dentate granule cells.
From these findings, the two phenomena may be related as follows. Following neuronal death, neurogenesis of the granule cells is induced and the axon of the newly generated cell forms aberrant synaptic connection that contributes to MFS.
Taken together, we propose that it may be suggested that when kindling is induced in rats using electrical stimulation, the kindling effect develops, seizures become generalized, and the animal acquires the susceptibility of lasting seizures, all in the absence of MFS; and that MFS may be the secondary consequence of neuronal death of the granule cell layer in the DG.
The results of neuronal death in the present study contradict those of previous reports,2,3,10,11,13,22,23,25,34,37 for two possible reasons. One is the timing of examining neuronal death by the TUNEL method. The earliest time examined was 2 h after stimulation. If neuronal death occurred earlier than 2 h after stimulation, then our method would have missed detecting the cell death. Another possibility is that prolonged seizure activities such as epileptic status may be required to induce neuronal death. In the literature, a kainate model with a few hours of convulsive status,3,13,23 an over-kindling model with many stage 5 convulsions,25 or the use of intermittent or prolonged electrical stimulation for several hours22,25 was apparently required to induce neuronal death.
This work was supported by Grant-in-Aid for Scientific Research (C) 07671044 and (B) 09470158 from the Japan Ministry of Education, Science, Sports and Culture and from Research Grant for Nervous and Mental Disorders. The work was also supported by ‘A collaborate study on intractability factors, prognosis and treatment for intractable epilepsy (7 A-1)’ from the Japan Ministry of Health and Welfare. We thank Eiji Nakagawa for assistance in TUNEL immunohistology.