Mesial Temporal Lobe Epilepsy: Pathogenesis,Induced Rodent Models and Lesions

Hippocampal structures involved in epilepsy

Because a central role in epileptogenesis is attributed to the hippocampus, a brief review of anatomy, pathways and interaction of the hippocampus with other parts of the limbic system will facilitate understanding of the proposed mechanisms and pathologic changes of MTLE. An excellent review of hippocampal structures and pathways is provided by Goldensohn and Salazar (1986).

Important subanatomic locations affected in MTLE and its rodent models are included in Figure 1.

Briefly, the rodent hippocampus is a somewhat flattened, distinct, elongated structure between the mesial (medial) part of the temporal lobe and the underlying thalamus. In rats, the hippocampus extends caudally and ventrally from the dorsal midline just above the thalamus at Bregma -1.72 to the ventral subiculum at Bregma -6.84. Numerous fiber tracts connect the hippocampus with other parts of limbic system, via the circuit of Papez, as outlined in Figure 2. The hippocampus consists of two interlocking C-shaped areas called Ammon’s horn and the dentate gyrus, which are composed principally of pyramidal and granule cell neurons, respectively. Ammon’s horn has been divided into four sub-fields, CA1 through CA4. CA1 extends from the dorso-medial subiculum to a slightly thickened band, CA2. CA3 comprises the ventro-medial portion of the hippocampus. The terminal portion of CA3, as it courses into the dentate hilus, is sometimes referred to as CA4, although many authors simply refer to this terminal portion as CA3. CA3 neurons receive MF input from the dentate gyrus.

MFs are the axons of dentate granule cells, which normally synapse on the dendrites of CA3 pyramidal neurons. Dentate granule cells are oriented with their dendrites projected towards the molecular layer. A GABAergic population of neurons called dentate basket cells is intermingled with the dentate granule cells. Enclosed within the curvature of the dentate gyrus is the dentate hilus, which contains polymorphic neurons of CA4 extending from CA3 as well as the glutamatergic hilar mossy cells. Hippocampal pathways are outlined in Figure 1D. Several fiber tracts that connect the hippocampus with other parts of the limbic system, via the circuit of Papez are outlined in Figure 2.

Classification of seizures

This section reviews the types of seizure to facilitate understanding of the terminology used in subsequent sections. Epileptic seizures have been classified into two broad categories - partial (focal) and generalized, based on clinical signs. Partial seizures originate from a focus of neurons in one cerebral hemisphere and are subclassified as simple or complex. Simple partial seizures may be associated with sensory, somatosensory, autonomic, or psychic symptoms, but do not cause loss of consciousness and do not last more than one minute. In contrast, complex partial seizures are always associated with loss of consciousness and may last a few minutes. Both simple and complex partial seizures may gradually develop into secondary generalized seizures of no more than a few minutes’ duration. Primary generalized seizures result from paroxysmal discharges arising in both cerebral hemispheres that may manifest as tonic-clonic seizures with loss of consciousness (grand mal seizures) or generalized tonic or clonic seizures with or without loss of consciousness. Status epilepticus (SE), which can be either partial or generalized, is a state of continuous seizure activity for five minutes or longer. In this review, any prolonged seizure activity resulting from chemical or electrical injury or other stimuli will be classified as SE.

Proposed mechanisms of MTLE

In general, proposed mechanisms of MTLE are based on the unique anatomic relationship of the hippocampus to the limbic system and on receptor-based electrophysiological changes related to either recurrent inhibition or excitation.

The position of the hippocampus relative to other parts of the limbic system and its cross-talk with those parts may have a central role in seizure generation. The presence or absence of various surface receptors on neurons may impart resistance or susceptibility to selected hippocampal areas leading to neuronal degeneration and seizure generation.

As outlined in Figure 2, hippocampal circuitry is a part of a larger circuit known as the Papez circuit, which ultimately constitutes a part of the limbic system. It has been proposed that increased hippocampal activity goes out to other parts of the limbic system and returns as amplified input to the hippocampus via the Papez circuit, which can lead to generalized epileptic seizures. Although imaging studies have revealed abnormalities along the afferent and efferent pathways of the Papez circuit (Bronen et al., 1991; Oikawa et al., 2001), concrete evidence of a cause and effect relationship between the Papez circuit and generation of epileptic seizures is not available. In contrast, the role of excitatory and inhibitory receptors and/or their subunits is supported by various studies, as discussed next.

Shifts in the distribution of receptors on hippocampal neurons may also lead to increased hyperexcitability and/or neurodegeneration of selected neuronal populations. The excitatory neurotransmitter glutamate is released from presynaptic neurons and acts upon the ionotrophic glutamate receptors Kainate, AMPA and NMDA. Kainate and AMPA receptors cause fast depolarization of neurons by allowing influx of monovalent cations such as Na⁺ and K⁺, whereas NMDA receptors effect a slower depolarization by allowing influx of divalent cations, primarily Ca⁺⁺. Ionotropic glutamate receptors are composed of multiple combinations of subunits.

While there is little information to support the role of NMDA receptors or their subunits in causing increased hippocampal hyperexcitability or neurodegeneration, data do suggest that the absence or decreased expression of the GluR2 subunit in AMPA receptors may play a role in neuronal hyperexcitability and/or neurodegeneration. The presence of the GluR2 subunit in AMPA receptors controls the influx of Ca⁺⁺into the neurons. GluR2 knockdown caused by administration of antisense oligonucleotide probes in the dorsal hippocampus of adult rats caused degeneration of CA3 neurons (Friedman et al., 2003). Decreased expression of GluR2 subunits on CA3 neurons in a kainic acid (KA)-induced rat model of MTLE implicates their absence in CA3 neuronal degeneration (Sommer et al., 2001).

Similarly, decreased GluR2 expression by hippocampal CA1 neurons may lead to neuronal hyperexcitability in a hypoxia-induced perinatal rat model of epileptogenesis (Sanchez et al., 2001). Increased expression of kainate receptor subunit GluR6 by administration of HSV-1 vector bearing GluR6 into rat hippocampus caused spontaneous seizures, hyperexcitability in CA1 cells and loss of CA1, hilar, and CA3 neurons (During et al., 1993). The inhibitory receptors, GABA and adenosine A1 (A1Rs), are also implicated in some forms of seizure generation.

The principal inhibitory neurotransmitter in the brain, gamma amino butyric acid (GABA), acts on two types of receptors, namely GABA_A, the ligand-operated ion channels, and GABA_B, the G-protein-coupled metabotropic receptors. GABA_A receptors are responsible for fast inhibitory post-synaptic potential (IPSP), whereas GABA_B receptors cause slower IPSP. Altered expression of GABA _A receptor subunits has been reported in both MTLE models (Poulter, Brown et al., 1999; Peng, Huang et al., 2004) and human patients (Loup, Wieser et al., 2000). Altered GABA_A receptor subunits probably decrease inhibition resulting in neuronal hyperexcitability.

Presynaptic GABA_B receptors suppress neurotransmitter release and have anticonvulsant or proconvulsant activities depending on whether they are located on glutamatergic or GABAergic neurons. Although binding of GABA_B receptor antagonists to CA3 pyramidal cells made the depolarizing GABA response excitatory and proconvulsive (Kantrowitz, Francis et al., 2005), the exact role of these receptors in seizure generation is unclear.

In addition to GABA receptors, A1Rs also have an inhibitory effect on the hippocampus. A1Rs inhibit adenylate cyclase activity (Dunwiddie and Fredholm, 1989) leading to increased K⁺ conductance (Greene and Haas, 1985) and decreased Ca⁺⁺ influx (Wu and Saggau 1994). The presence of numerous receptors on CA2 neurons as compared to CA1 or CA3 neurons (Ochiishi et al., 1999) probably make CA2 neurons resistant to glutamate excitotoxicity (Mattson and Kater, 1989) and status epilepticus (Young and Dragunow, 1995).

Regardless of their origin, the partial-onset seizures that are typical of MTLE are recorded as focal interictal epileptiform spikes or sharp waves originating at or near the seizure focus. These spikes are caused by a paroxysmal depolarization shift (PDS) in neurons. During PDS, prolonged calcium-dependent conductance causes depolarization of the neuronal membrane, resulting in threshold voltage and activation of voltage-gated sodium channels. Influx of sodium reduces the resting potential and causes firing of multiple sodium-dependent action potentials, followed by hyperpolarization with active transport of sodium ions out of the neurons. Synchronous PDS discharges fired by millions of neurons create an electrophysiologically detectable focal interictal epileptiform spike that manifests as partial-onset seizures (Babb, Kupfer et al., 1991; Franck, Pokorny et al., 1995; Masukawa, O’Connor et al., 1995; Zhang and Houser, 1999).

The literature contains many hypotheses of MTLE. This section will focus on the recurrent excitation and inhibition hypotheses considered most tenable by the authors, based on review of the literature and their own research. Both hypotheses focus on the dentate gyrus as the nidus for epileptogenesis.

Recurrent excitation hypothesis

One of the most widely accepted explanations of epileptogenesis, the recurrent excitation hypothesis, states that spontaneous motor seizures in MTLE are a consequence of hyperexcitability of dentate granule cells, resulting from alterations of circuitry due to aberrant MF sprouting. These recurrent loops probably induce dentate granule cells to stimulate each other, leading to generation of spontaneous motor seizures from the hippocampus (Figure 3). Studies of aberrant MF sprouting have been reviewed in various animal models of MTLE, such as KA-, pilocarpine-, and kindling-induced models (White, 2002). Under normal circumstances, mossy fibers from dentate granule cells traverse the dentate hilus and stratum lucidum of Ammon’s horn and form synapses with hilar mossy cells and CA3 pyramidal neurons. In patients with MTLE, aberrant MF sprouts have been found in the granule cell layer and the inner molecular layer of the dentate gyrus (Okazaki, Molnar et al., 1999; Buckmaster, Zhang et al., 2002).

Similar MF sprouts were detected in animal models of MTLE (Buckmaster, Zhang et al., 2002). These aberrant MF sprouts formed synapses with dendrites of granule cells leading to recurrent excitatory circuitry (Lothman, Stringer et al., 1992). In comparison to 61% of granule cells from rats exhibiting aberrant MF sprouting following pilocarpine-induced status epilepticus, only 13–16% of granule cells in control rats exhibited excitatory postsynaptic current (EPSC). This implicated MF sprouts in seizure generation (Obenaus, Esclapez et al., 1993; Houser and Esclapez, 1996). Histopathologic and ultrastructural findings in a pilocarpine-induced status epilepticus model showed that 93–96% of synapses of aberrant mossy fibers were formed with GABA-negative granule cells and fewer than 1 of 20 new synapses were formed with GABAergic interneurons. The resulting circuits were predominantly excitatory (Ribak, 1985; Robert, 1994; Sloviter, 1996; Menks and Sankar, 2002). Because dentate granule cells seldom have spontaneous epileptiform discharges, they form a barrier to the spread of seizures to other parts of the brain (Menks and Sankar, 2002); however, it is proposed that this recurrent excitatory circuitry overcomes this barrier.

Recurrent inhibition hypothesis

Results of animal models and human studies also support the hypothesis that dentate granule cells form a nidus for epileptogenesis secondary to loss of inhibitory neurons (Menks and Sankar, 2002), such as hilar mossy cells or dentate basket cells. Mossy cells are glutamatergic neurons in the dentate hilus, whereas dentate basket cells are GABAergic neurons of the dentate gyrus. The perforant pathway from the entorhinal cortex (EC) provides stimulatory input to dentate granule cells as well as to hilar mossy cells and dentate basket cells. Stimulation of hilar mossy cells and dentate basket cells causes feed-forward inhibition of the granule cells (Ratzliff, Howard et al., 2004). Glutamatergic mossy cells are stimulatory, whereas the glutamic acid decarboxylases (GAD)-positive inhibitory GABAergic neurons, also known as basket cells, are inhibitory, especially to dentate granule cells. In addition to innervating dendrites of the CA3 pyramidal cells, MF collaterals also innervate the hilar mossy cells.

Hence, stimulation of granule cells via the perforant path causes stimulation of hilar mossy cells, which in turn stimulate the GABAergic basket cells leading to inhibition of dentate granule cells, a phenomenon also termed feedback inhibition (Obenaus, Esclapez et al., 1993). Ablation of hilar basket cells caused dentate granule cell hyperexcitability in a trauma-induced rat model of TLE (Robert, 1994; Sloviter, 1996; Zappone and Sloviter, 2004). In situ hybridization detection of GAD mRNA revealed a marked reduction in the number of GAD-positive hilar neurons in a rat model of TLE (Buckmaster, Wenzel et al., 1996). However, this study could not detect GABAergic neurons in this cell population and indicated that basket cells may be resistant to epileptogenesis.

The “dormant basket cell hypothesis” is based on a loss of hilar mossy cells. In the normal hippocampus, hilar mossy cells induce inhibition via the basket cells. Loss of hilar mossy cells causes dormancy in inhibitory basket cells and loss of inhibition of the dentate granule cells (Figure 3) (Ratzliff, Howard et al., 2004). In contrast, others (Houser and Esclapez, 1996; Schwarzer, Tsunashima et al., 1997; Brooks-Kayal, Shumate et al., 1998) have found that mossy cells innervate granule cells rather than inhibitory basket cells and some investigators (Esclapez and Houser, 1999; Smith and Dudek 2001; Bausch and McNamara, 2004; Perez-Mendes, Cinini et al., 2005) suggested that loss of mossy cells causes decreased excitability of granule cells following perforant path stimulation. Because of these discrepancies, the recurrent excitation hypothesis is favored over the dormant basket cell hypothesis.

Background, Dosing Information, and Pathological Alterations in Rodent Models of MTLE

Many rodent models have been developed to study the pathogenesis of MTLE (Coulter, McIntyre et al., 2002; White, 2002; Buckmaster, 2004). Examples include chemoconvulsant and electrical stimulation-induced post-status epilepticus, kindling, tetanus toxin, hyperthermia, post-traumatic epilepsy, and perinatal hypoxia/ischemia (HI) models. These models have certain similarities to, and dissimilarities from, MTLE that impact their utility or relevance. Our literature search failed to find reports of genetic models of MTLE. Information on genetic models of other types of epilepsies (Consroe and Edmonds, 1979; Loscher, 1984; Lason, 1998) are available to readers interested in non-MTLE models of epilepsy. Selected protocols and hippocampal lesions in various models of MTLE are listed in Table 1.

Post-Status Epilepticus Models

Post-status epilepticus models closely mimic the clinical manifestations of MTLE in humans, in which an acute triggering process is frequently followed by a latency period with subsequent development of spontaneous motor seizures. An acute episode of seizures, also known as status epilepticus, is the triggering process in these animal models. The models are created by systemic or local administration of chemoconvulsants or electrical stimulation. In status epilepticus models, seizures are induced in rats or mice for 1 to 2 hours before status epilepticus is interrupted by an anticonvulsant. Rehydration or supplemental nutrition may be required to restore the rodents to a healthy state. After a few weeks, many will start exhibiting spontaneous, secondary generalized seizures.

Chemoconvulsants

KA (Tauck and Nadler, 1985; Okazaki, Molnar et al., 1999) and pilocarpine (Turski, Cavalheiro et al., 1984; Okazaki, Molnar et al., 1999) are the two of the most commonly used chemoconvulsants used to create status epilepticus models of MTLE. Hippocampal lesions in these models are similar to the HS observed in humans with MTLE (Turski, Cavalheiro et al., 1984; Ben-Ari, 1985; Cavalheiro, Leite et al., 1991; Sloviter 1996). In addition to loss of specific cell populations in Ammon’s horn, hippocampal alterations in these animal models and human patients include aberrant MF sprouting (Turski, Cavalheiro et al., 1984; Cronin and Dudek, 1988; Sutula, Cascino et al., 1989; Babb, Kupfer et al., 1991; Mello, Cavalheiro et al., 1993; Obenaus, Esclapez et al., 1993; Okazaki, Evenson et al., 1995) and other circuit rearrangements (Okazaki, Molnar et al., 1999).

Electrical Stimulation

Compared to KA and pilocarpine, electrical stimulation is used less often as a model of MTLE. However, this model produces clinical signs and lesions comparable to those of the KA and pilocarpine models. Electrical pulse trains have been used to stimulate different regions of the brain to produce self-sustaining seizures without prior kindling or administration of chemoconvulsants. Repetitive tetanic stimulation of hippocampal afferents such as the perforant path (Schwob, Fuller et al., 1980; Vicedomini and Nadler 1987; Sloviter 1996; Ribak, Tran et al., 2000; Gorter, van Vliet et al., 2001; Mazarati, Lu et al., 2004), hippocampus (Racine 1972; Lothman and Collins 1981; Vezzani, Conti et al., 1999; Bastlund, Jennum et al., 2005), or amygdala (Racine 1972; Cavazos and Sutula 1990; Tilelli, Del Vecchio et al., 2005) has been used to produce status epilepticus. Of the various protocols for electrical stimulation, the method reported by Vicedomini (1987) is the prototype. Unanesthetized rats were stimulated with 0.2 to 0.4 millisecond monophasic rectangular pulses at 20Hz with 10 seconds train duration and 30 seconds inter-train interval through chronically implanted electrodes in angular bundles or fimbria. Electrical stimulation was stopped when 10 consecutive trains each produced 30 seconds of hippocampal after-discharge. Rats were then monitored for self-sustained electroencephalographic seizure activity; 85% exhibited status epilepticus within 7 hrs.

Sloviter (1996) described the neuropathologic changes following electrical stimulation of the brain. Unilateral sustained electrical stimulation of the perforant path in anesthetized rats caused necrosis of CA1 and CA3 pyramidal cells and hilar neurons, whereas CA2 neurons were generally unaffected. Based on unilaterally or bilaterally evoked granule cell discharges, the neuronal damage was likewise unilateral or bilateral (Tilelli, Del Vecchio et al., 2005). In addition to the hippocampal lesions, electrical stimulation of the amygdala caused neuronal necrosis in the piriform cortex (Gorter, van Vliet et al., 2001). Electrical stimulation of the angular bundle caused extensive aberrant MF sprouting in the inner molecular layer of the dentate gyrus (Brandt, Glien et al., 2003). Self-sustained status epilepticus induced by high intensity (700 μA) pulsed-train electrical stimulation of the basolateral nucleus of the amygdala produced neuronal necrosis in the ipsilateral amygdala, piriform cortex, entorhinal cortex, endopiriform nucleus, and mediodorsal thalamus in rats. Aberrant mossy fiber sprouting was also present in the inner molecular layer of the dentate gyrus (McIntyre, Nathanson et al. 1982). Depending upon the area of the brain and the intensity of pulse train electrical stimuli, these models show minor variations in the site and the severity of brain lesions. However, all protocols were generally successful in producing the hippocampal lesions of MTLE.

Histopathologic findings are comparable to those in KA and pilocarpine-induced models and MTLE patients. However, electrode implantation is cumbersome and labor-intensive, which reduces the desirability of this model relative to chemoconvulsant models.

Kindling Models

Kindling triggers epileptic seizures by repeated small electrical or chemical stimulation to the brain. Kindling models differ from post-status epilepticus models, as repeated subconvulsant levels of either electrical stimulation or chemoconvulsants do not produce SE immediately following the treatments. Moreover, once the animal is fully kindled, “evoked” seizures can be incited, rather than the spontaneous motor seizures observed in MTLE patients and post-status epilepticus models. The resulting lesions, therefore, differ from those in the post-status epilepticus models and MTLE patients and may vary among protocols.

Kindling was used to stimulate the hippocampus, amygdala (McIntyre, Nathanson et al., 1982; Lehmann, Ebert et al., 1998; Brandt, Ebert et al., 2004), olfactory regions (Sutula, Cascino et al., 1989) or other areas of the brain (Cavazos and Sutula, 1990; Inoue, Morimoto et al., 1992) repeatedly over days to weeks to induce seizures in rats. Subsequent kindling treatments elicited progressively increasing electroencephalographic seizure after-discharge and behavioral seizures. Several months after kindling, animals responded with evoked seizures to subconvulsant stimulation. Among the chemoconvulsants, pentylenetetrazole (PTZ) a GABA receptor antagonist (Sutula, Cavazos et al., 1992) and the GABA_A ergic antagonist, bicuculline methiodide (Uemura and Kimura, 1988) have been used for kindling.

Electrical stimulation is more commonly used than chemoconvulsants to induce kindling and allows focused stimulation of key neuroanatomic sites via electrode placement. Cavazos et al., (1994) stereotaxically implanted stainless steel bipolar electrodes in the perforant path near the angular bundle, amygdala, or olfactory bulb of anesthetized rats. After 2 weeks’ recovery, unanesthetized rats were kindled with a 1-sec train of 62 Hz biphasic constant current 1.0 msec square wave pulses to produce stage 5 seizures. Once the rats demonstrated stage 5 evoked seizures, they were deemed fully kindled. The amygdala is by far the most responsive region to kindling. No differences were observed in generalized seizure threshold or after-discharge durations in two groups of rats kindled in cortical versus basolateral amygdala; however, rats stimulated in the cortical amygdala took longer to attain stage 5 seizures (Gilbert, Gillis et al., 1984).

Seizures induced by kindling have been classified into five stages based on clinical signs. This scale was later adopted to describe seizures in other animal models of epilepsy (Racine, 1972). However, in the kindling models, seizures are evoked and spontaneous motor seizures are rare (McNamara 1984). Thus, kindling models are not representative of MTLE seizures, but some of the induced lesions may be useful for study.

Amygdaloid kindling caused neuronal necrosis and loss with massive gliosis in the ipsilateral hemisphere extending from the medial olfactory bulb through the amygdala-piriform cortex to the ventral hippocampus, especially CA1 fields, and to midline thalamic nuclei (Inoue, Morimoto et al., 1992). Unilateral neuronal loss in ipsilateral CA3 and bilateral loss in CA1 were observed after status epilepticus induced by kindling-like electrical stimulation of the deep pre-piriform cortex in rats (Cavazos and Sutula, 1990; Cavazos, Das et al., 1994). Represa and Ben-Ari (1992) counted neurons utilizing quantitative stereological methods following kindling by perforant path stimulation. Neuronal loss increased in the dentate hilus with increasing numbers of kindled generalized tonic clonic seizures; however, the lesions bore only slight resemblance to those reported in MTLE.

Sutula et al., (1992) described aberrant MF sprouting in the inner molecular layer of the dentate gyrus with no other hippocampal damage following electrical kindling. Similar sprouting occurred after generalized tonic-clonic seizures in rats following systemic administration of pentylenetetrazole (Golarai et al., 1992).

The primary advantage of kindling is that specific areas of the brain can be stimulated in a controlled manner. This, and the lack of HS, makes kindling a useful model for investigating the mechanisms underlying paradoxical TLE or TLE without HS. The disadvantages of kindling as a model of MTLE are the cumbersome nature of stereotactic electrode implantation or chemical inoculation into the brain, the long time needed for induction, the lack of spontaneous seizures, and the dissimilarity of lesions compared with MTLE.

Tetanus Toxin

Compared to post-status epilepticus and kindling models, less information is available about tetanus toxin models, which suggests lesser acceptance of these models. Tetanus toxin blocks neurotransmitter release in the brain. Stereotactic inoculation of tetanus toxin into the brain has been used to reproduce some features of MTLE; however, the resultant seizures are either weak or temporary, and lesions differ from those in the post-status epilepticus models or human MTLE patients.

In naïve rats, following perforant path stimulation, dentate granule cells filter and block any seizure discharges from spreading to other parts of the temporal lobe. Tetanus toxin blocks GABA_B receptors, and diminishes the ability of the dentate granule cells to block the spread of seizure discharges (Finnerty, Whittington et al., 2001). The reported neurobehavioral alterations and histopathologic findings in this animal model of epilepsy vary widely. When tetanus toxin was injected into the hippocampus of rats, (Mellanby, Hawkins et al., 1984), the induced spontaneous motor seizures were reversible and without associated lesions, but the rats had memory and learning deficits.

Jefferys and Williams (1987) reported that the evoked responses from CA3 pyramidal cells were depressed for several weeks after tetanus toxin administration. Spatial reference memory tasks in these rats were also impaired. Tetanus toxin injection into the parietal neocortex of rats caused excessive synchronization of neuronal activity in parietal and temporal areas leading to epileptic activity (Brener, Amitai et al., 1991). Although the seizures produced by tetanus toxin were spontaneous motor seizures, the lack of a pronounced motor component made it difficult to score the epileptic activity (Jefferys, Borck et al., 1995).

In tetanus toxin models, the limited distribution and intensity of reported lesions correlated with the brief duration of seizures (Toth, Yan et al., 1998; Jiang, Duong et al., 1999; Bender, Dube et al., 2003). Lesions were often confined to areas within and immediately adjacent to the injection site. The extent of reported neuronal damage varied from none to as much as 30% of CA1 pyramidal neurons, although the latter represented an area near the injection site (Bagetta, Corasaniti et al., 1990; Bagetta, Corasaniti et al., 1991). When injected into the dentate gyrus, tetanus toxin-induced lesions were limited to a marked loss of granule cells in the ipsilateral dentate gyrus (Jefferys, Evans et al., 1992). The intense neurodegeneration reported in ipsilateral dentate gyrus was probably a direct effect of tetanus toxin, because dentate granule cells generally exhibited only subtle degenerative change in other models of MTLE.

The low intensity and transient nature of spontaneous motor seizures, along with the dissimilarity of lesions to those in MTLE, limits the usefulness of the tetanus toxin model. Nonetheless, this model may facilitate study of the role of GABA_B receptors in epileptogenesis (see Section 4).

Hyperthermia

Many adults with TLE had a childhood history of febrile convulsions (Vestergaard et al., 2007). Whether prolonged febrile seizures changed the excitability of the limbic system and caused subsequent TLE remains to be determined. It is proposed that the early febrile seizures damage the hippocampus leading to HS and consequent development of MTLE in adulthood (Harvey et al., 1995). Since brain development in rats 1–2 weeks of age is similar to that of human infants, hyperthermia is used to model MTLE.

Hyperthermia is especially useful for studying the culmination of febrile seizures in MTLE (Holtzman, Obana et al., 1981; Hjeresen, Guy et al., 1983; Hjeresen and Diaz, 1988; Baram, Gerth et al., 1997; Lemmens, Lubbers et al., 2005; Scantlebury, Gibbs et al., 2005). Ambient hyperthermia for six- and 10-day-old rat pups produced seizures that caused electrocortical paroxysmal discharges (Holtzman, Obana et al., 1981). Microwave energy (Hjeresen, Guy et al., 1983) also increased the core temperatures of 13- and 17-day-old rats to cause convulsions similar to febrile convulsions in human infants. A regulated stream of mildly heated air induced hyperthermia in 10- to 11-day-old rats (Baram, Gerth et al., 1997); about 93% of the rats developed stereotypical seizures. Although hyperthermic treatment caused immediate seizures, spontaneous motor seizures, as reported in post status epilepticus animal models and in human MTLE patients, did not occur in this model.

There are few reports of the neuropathologic changes of febrile seizure model rats. Rats were placed in a water-bath at 45°C for 4 minutes (Jiang, Duong et al., 1999) resulting in both visual and electroencephalographic seizures ranging in duration from 30 s to 6 min; seizure duration increased with the number of seizures. In most rats, no neurodegeneration was detected, although there was aberrant MF sprouting of granule cell collaterals into the inner molecular layer of the dentate gyrus. A few rats with evolution of short seizures into status epilepticus had neuronal degeneration in the hippocampus, temporal cortex, and mediodorsal thalamus. Findings were similar in another model in which hyperthermia was induced using a warmed air stream above the animals (Toth, Yan et al., 1998).

Neuronal argyrophilia, indicating physicochemical changes, and scattered apoptotic neurons were present in the central nucleus of the amygdala and CA1 and CA3 subfields of the hippocampus up to 2 weeks following seizures (Lowenstein, Thomas et al., 1992; Golarai, Greenwood et al., 2001; Santhakumar, Ratzliff et al., 2001; D’Ambrosio, Fender et al., 2005). Bender et al., (2004) performed serial MRIs at several time-points on immature rats exposed to hyperthermia and recorded altered T2 values in the dorsal hippocampus, the piriform cortex, and amygdala, without any evidence of neuronal degeneration, as assessed using the Fluoro-Jade histologic staining method. It was concluded that, although neuronal degeneration was not present, the affected areas may have neuronal alterations that produced seizures. The lesions have limited similarities to those reported in post-status epilepticus models and humans.

In contrast to the definitive spontaneous motor seizures in adult MTLE human patients, spontaneous limbic seizures do not occur during adulthood in the hyperthermic-seizure models (Dube, Chen et al., 2000), which casts doubt on their merit. Moreover, lesions in hyperthermic seizure models have limited resemblance to those in MTLE or post status epilepticus models. Therefore, the utility of this model relative to MTLE may be limited.

Post-Traumatic Epilepsy Model

Traumatic brain injury, with associated reduction in hippocampal volume (Bigler, Blatter et al., 1997; Tate and Bigler, 2000), is one of the most frequent causes of human TLE (Salazar, Jabbari et al., 1985; Annegers, Hauser et al., 1998). Trauma-associated MTLE has been partially replicated in post-traumatic epilepsy animal models. These models have produced spontaneous motor seizures as reported in MTLE. However, in contrast to the mesial temporal cortical origin of spontaneous motor seizures in MTLE and post-status epilepticus models, the seizures in post-traumatic epilepsy models originate in the frontal-parietal cortex and progress to the mesial temporal cortex.

In post-traumatic models of TLE, brain trauma has been produced by fluid percussion injury (FPI) (Lowenstein, Thomas et al., 1992; Golarai, Greenwood et al., 2001; D’Ambrosio, Fender et al., 2005) or cortical weight drop injury (Weisend and Feeney 1994; Golarai, Greenwood et al., 2001). Santhakumar et al., (2001) used electrophysiology in the rat FPI model to reveal that a low-frequency, single-shock stimulation of the perforant path led to early granule cell hyperexcitability with return to normal by 1 month. However, a persistent decrease in threshold to induction of seizure-like electrical activity was observed in response to high-frequency tetanic hippocampal stimulation in those rats. A brief (10 ms) pressure pulse of 3.75–4 atm was delivered directly to the intact dura through a 3-mm wide burr hole in the skull (D’Ambrosio, Fairbanks et al., 2004). Transient dural compression by a single episode of severe FPI resembled human cases of closed head injury and led to post-traumatic epilepsy. In this model, chronic electrocorticography revealed partial spontaneous motor seizures that originated from the neocortex at the injury site, deteriorated progressively, and spread to other parts of the neocortex. The same laboratory subsequently reported seizure origin from the frontal-parietal cortex that progressed to the mesial temporal cortex (D’Ambrosio, Fender et al., 2005). The seizures originated only from the mesial temporal neocortex in the post status epilepticus models of MTLE.

There are several descriptions of pathologic alterations in post-traumatic rat models of epilepsy (Chiba 1985; Jensen, Applegate et al., 1991; Jensen, Holmes et al., 1992; Lowenstein, Thomas et al., 1992; Jensen, Blume et al., 1995; Jensen, Wang et al., 1998; D’Ambrosio, Fender et al., 2005). The number of dentate hilar neurons was markedly decreased without abnormality of CA neurons 1 week post-trauma in a FPI model (Golarai, Greenwood et al., 2001). These lesions were confirmed by another study (Santhakumar, Ratzliff et al., 2001), which also described marked ipsilateral necrosis of pyramidal neurons in CA3 with milder necrosis in CA1 neurons and granule cells after weight drop contusion. Timm’s stain-positive MF sprouting was recorded in the dentate gyrus in the FPI model (D’Ambrosio, Fender et al., 2005). Two to four weeks post-FPI, remarkable neuronal loss and calcification were present in the ipsilateral thalamus; CA1 and CA3 hippocampal subfields exhibited loss of pyramidal neurons similar to that in human patients suffering from FPI (Chiba 1985; Jensen, Applegate et al., 1991; Jensen, Holmes et al., 1992; Jensen, Blume et al., 1995; Jensen, Wang et al., 1998). Gliosis was present in the ipsilateral hippocampus and temporal cortex.

Post-traumatic models have been useful in studying trauma-associated MTLE in humans. These protocols have produced spontaneous motor seizures as reported in MTLE; however, in contrast to the mesial temporal cortical origin of spontaneous motor seizures in MTLE and post-status epilepticus models, the seizures originated in the frontal-parietal cortex and progressed to the mesial temporal cortex in the post-traumatic epilepsy model (D’Ambrosio, Fender et al., 2005). The difference in seizure origin, the inconsistent development of hippocampal lesions that resemble those of post-status epilepticus models and MTLE, the high morbidity and mortality, and the labor-intensive nature of the procedures make post-traumatic models less desirable than post-status models for replicating MTLE.

Perinatal Hypoxia/ischemia

Perinatal HI has been associated with increased risk of epilepsy during early childhood or later (Watanabe, Kuroyanagi et al., 1982; Bergamasco, Benna et al., 1984). However, the relationship of perinatal HI to the development of seizures is poorly understood. Perinatal HI has been used in rats to investigate one of the initiating injuries causing epileptogenesis in MTLE. However, this treatment is inconsistent at producing spontaneous motor seizures or histopathologic lesions.

In a rat model of perinatal hypoxia, brief, moderate global hypoxia (3–4% O₂) at postnatal day (P) 10–12 led to spontaneous tonic clonic seizures. Hypoxia induced at younger or older ages did not lead to seizures, suggesting the age-dependent epileptogenic potential of hypoxia (Jensen, Applegate et al., 1991). Rat pups exposed to hypoxia during P10-12 exhibited minimal to no histopathologic lesions, but had increased susceptibility to convulsant-induced seizures during adulthood (Jensen, Applegate et al., 1991; Jensen, Holmes et al., 1992; Lowenstein, Thomas et al., 1992). Global hypoxia caused an acute increase in excitability in the immature brain and hypoxia during postnatal day 10 led to persistent hippocampal susceptibility to seizures. The age-dependent epileptogenic effects of hypoxia were partly due to a direct and permanent effect on neuronal excitability in the hippocampus (Jensen, Wang et al., 1998). The increased susceptibility of CA1 pyramidal neurons is likely related to the presence of greater numbers of glutamate receptors on these neurons (McDonald and Johnston 1990). These findings are corroborated by another observation that the non-N-methyl-D-aspartate (NMDA) antagonist 6-nitro-7-sulfamoylbenzo (f) quinoxaline-2,3-dione (NBQX) blocks the acute and chronic epileptogenic effects of global hypoxia in a perinatal HI model (Jensen, Blume et al. 1995).

Because hypoxia alone cannot induce spontaneous motor seizures, it has been combined with ischemia (Williams, Dou et al., 2004). In one study utilizing combined hypoxia/ischemia, approximately 50% of 7-day-old rats exhibited spontaneous motor seizures and 20% died without exhibiting any seizures, after they were kept in a chamber filled with 8% oxygen following ligature of the right common carotid artery.

Although there are several reports about electrophysiological and behavioral studies in the perinatal hypoxia/ischemia model of epilepsy (Lowenstein, Thomas et al., 1992; Towfighi and Mauger, 1998; Williams, Dou et al., 2004), the literature characterizing the morphological changes is minimal. The reported lesions, although similar to those of MTLE patients and post-status epilepticus models, are inconsistently produced (Towfighi and Mauger 1998). Neuronal necrosis was observed in the CA1 through CA4 regions of the hippocampus, lateral and medial nuclei of the dorsal thalamus, nucleus reticularis of the thalamus, striatum, medial habenular nucleus, and olfactory tubercle (Williams, Dou et al., 2004). These lesions, which were limited to the ipsilateral hippocampus following ligation of the right common carotid artery and subsequent hypoxemia, decreased with age and were no longer present in pups older than 13 days. Increased Timm’s staining (evidence for mossy fiber sprouting) was observed in the inner molecular layer of the dentate gyrus in about half of the 7-day-old rats that exhibited spontaneous motor seizures following a similar technique of hypoxia-ischemia induction (Liu, Mikati et al., 1995).

Opportunities to investigate the initiating epileptogenic injury in MTLE and the ease of induction are advantages of the perinatal HI model. However, high mortality, limited induction of spontaneous seizures and the inconsistent and less extensive MF sprouting as compared to the status epilepticus models and changes in human MTLE reduce the acceptability of this model.

Influence of Age/Strain Selection on Response to Epileptogenic Treatments

In rodents, both age and strain affected the response to epileptogenic treatments. Immature rats were more susceptible to SE than were adult rats. In immature (18-day-old) rats, KA and kindling caused relatively mild and no hippocampal neuronal degeneration, respectively. Neither treatment caused aberrant MF sprouting. In contrast, in mature rats, both KA and kindling caused milder seizures, but more severe hippocampal neuronal degeneration and aberrant MF sprouting (Haas et al., 2001). Despite apparent age differences, no strain differences in susceptibility to epileptogenic agents have been documented for rats. In a study in progress, the authors did observe more severe neurodegeneration in CA1 as compared to CA3 of F-344 rats than those previously reported in SD rats; however, differences may have been due to slight differences in KA exposure. In contrast, mice exhibited strain-based variations in susceptibility to epileptogenic treatments, as evidenced by development of hippocampal neuronal degeneration and aberrant MF sprouting in 129/SvEMS mice and lack of such lesions in ICR mice following KA administration (McNamara et al., 1996; Schauwecker and Steward 1997; Cantallops and Routtenberg 2000). The relatively severe hippocampal neuronal degeneration and altered synaptic physiology may sensitize adult rats and susceptible mouse strains to epileptogenic stimuli and development of spontaneous seizures.