Discussion

These examples demonstrate that single nuclei or very small anatomically connected networks may contain sufficient circuity to sustain SE.

The concept that SE may be highly restricted has clinical implications. Well recognized are focal convulsive manifestations, such as eye deviation or hand twitching in an alert patient with a frontal SE focus. Less well recognized by many clinicians is the fact that some nonconvulsive or deficit symptoms, such as aphasia (96), panic attacks (168), or nose wiping (23), may be a manifestation of focal cortical status. In view of the relative inaccessibility of the medial temporal lobe to EEG, the local occurrence of focal SE is probably underestimated, particularly in children. An interesting case report documented electrographic SE with 1-Hz sharp waves in the anterior hippocampus, involving tissue less than 1cm in diameter. Seizure activity occasionally spread to posterior hippocampus, amygdala, or temporobasal cortex, but was generally subclinical (52).

We reported an example of medial temporal lobe SE, associated with lethargy and periodic lateralized epileptiform discharges (PLEDs) on EEG, in which ¹⁸F-2-DG positron emission tomography (PET) revealed intense focal hypermetabolism (72). A focal increase in blood flow on single-photon emission computed tomography (SPECT) has been reported in a patient with complex partial SE and PLEDs (2) and in a series of 18 patients with PLEDs (5), supporting the interpretation that PLEDs signify partial SE. It may be noted that PLEDs are comparable to the slow periodic complexes associated with highly restricted or restricted patterns of SE in rat (e.g., Figure 19.8).

Recruitment from the basolateral amygdala

The basolateral amygdala nucleus projects to ventral subiculum, lateral entorhinal cortex, endopiriform nucleus, nucleus centralis of amygdala, prelimbic and infralimbic (areas 32 and 25) medial prefrontal cortex, insular cortex, perirhinal cortex, bed nucleus of stria terminalis, nucleus accumbens, and dorsomedial thalamus (102, 103).

2-DG studies of SE induced by electrical stimulation reveal that the bed nucleus of stria terminalis, dorsomedial thalamus, nucleus accumbens, medial prefrontal cortex, and nucleus centralis of amygdala are all prone to be activated early in restricted SE patterns. Thus, not all of the projections of basolateral amygdala are initially recruited, but with further amplification of seizure activity by deep olfactory cortical regions and hippocampal formation, the amygdala is well positioned to play an important role in seizure and SE propagation.

Recruitment from the hippocampal-entorhinal complex

The precisely organized anatomic relationship within Ammon's horn and subiculum (98) provides circuitry that is vulnerable to epileptogenesis and SE. Seizure discharges from medial entorhinal cortex do not propagate well to CA3/CA1 in slices from normal adult rats, but do so with recurrent discharges in entorhinal cortex if slices are taken from kindled rats (66). The recurrent ictal discharge may involve a projection from presubiculum, as interruption of the projection from presubiculum to entorhinal cortex has been reported to attenuate seizure-induced layer III damage in entorhinal cortex (173). Patients with temporal lobe epilepsy commonly display entorhinal cortical as well as hippocampal-amygdala atrophy on volumetric magnetic resonance imaging (MRI) (12).

It is thus not surprising that the hippocampus can be driven into SE by epileptiform activation of entorhinal afferents. Injection of kainate into the entorhinal cortex causes SE with consistent CA1 damage (130). Intermittent stimulation of the perforant path over 30 minutes reliably elicits self-sustained SE with hippocampal involvement (157), a model described more fully elsewhere in this book.

Conversely, seizures and SE readily propagate externally from hippocampus via projections of CA1 and subiculum. CA1 projects to entorhinal, perirhinal, and subicular cortices, as well as to lateral septum. The temporal CA1 also projects to amygdala, olfactory bulb, anterior olfactory nucleus, and nucleus accumbens (197). In addition to projections to entorhinal cortex, the subicular complex has extensive limbic projections to the anterior olfactory nucleus, the prelimbic cortex, retrosplenial cortex, hypothalamus (mamillary complex, dorsomedial, medial preoptic, and other nuclei), the anterior thalamic complex, nucleus reuniens, nucleus accumbens and portions of caudate, and amygdala (basomedial, basolateral, amygdalohippocampal, and other cortical amygdala structures) (20, 67, 87, 185, 214).

2-DG studies have amply documented that perirhinal/entorhinal cortex may be recruited by SE initiated in the hippocampus (72, 97, 206). As depicted in Figure 19.3, the initial spread of discharges from the hippocampal formation may propagate to entorhinal cortex and one or more specific first-order projection sites.

Figure 19.3.  

Bilateral hippocampal activation in a rat in immobile SE. Intense activation of the lateral septal nuclei (A), the hippocampus, and the right amydalohippocampal nucleus (B) is evident, as is partial activation of the entorhinal cortex (C).

Once fully engaged by ictus, the entorhinal cortex, with extrahippocampal projections that arise mainly in the deep layers of the lateral entorhinal cortex, is well-positioned to propagate seizure discharges to periamygdala cortex, basolateral and lateral amygdala, piriform cortex, endopiriform nucleus, anterior olfactory nucleus and olfactory bulb, medial prefrontal cortex, nucleus accumbens, and ventral pallidum, and to limbic thalamic nuclei (210, 211, 213). Entorhinal seizure discharges to these projection sites may be reinforced by similar projections from seizure-activated CA1, subiculum, and basolateral amygdala, as described earlier. Some of the distal sites, such as medial prefrontal cortex, nucleus reuniens, and anterior olfactory nucleus, have reciprocating projections, potentially further reinforcing seizure activity.

Recruitment from rostral limbic regions

Induction of SE from limbic prefrontal regions has not been well studied. Given the prelimbic/infralimbic early activation with restricted SE patterns associated with amygdala or hippocampal induction and the rich connections of the medial prefrontal region with insular, entorhinal, and perirhinal cortex, ventral hippocampal formation, amygdala, and dorsomedial and reuniens thalamus (31), the medial prefrontal cortex should exhibit proclivity for status induction. The left column of Figure 19.2 shows images from a rat with SE induced by electrical stimulation of this region. All activated areas are known to receive projections from prelimbic or infralimbic cortex (162). The activation pattern overlaps that seen with restricted patterns from posterior limbic regions. In addition, this example displays activation of deep insular layers and endopiriform nucleus that courses in a continuous band into entorhinal cortex, suggesting that this zone is prone to SE activity.

Araki et al. (4) described sequential recruitment of structures, as shown with 2-DG after injection of kainate into the olfactory bulb. After initial activation of the olfactory bulb and anterior olfactory nucleus, additional structures to show hypermetabolism included the entorhinal cortex, endopiriform nucleus, nucleus accumbens, and ventral pallidum. In the most advanced pattern, the dorsomedial, centromedian, and ventromedial thalamus, entopeduncular nucleus (equivalent to internal globus pallidus), substantia nigra reticulata, and sensory cortex were activated.

Discussion

Studies of the highly restricted and less restricted patterns of SE activation suggest that within a network, one nucleus may serve as a generator, which then activates other nuclei via efferent projections. This is suggested when a small network of hypermetabolic nuclei are known to receive projections from a single nucleus, particularly when the highly restricted patterns have shown that nucleus to be activated in virtual isolation. The interpretation that a specific structure is an SE generator is additionally supported by observations that that nucleus is the one most activated metabolically, whereas other nuclei are only partly activated. Examples encountered earlier in this chapter included the basolateral amygdala, hippocampus, entorhinal cortex, medial prefrontal cortex, and olfactory bulb.

Another inference that can be made from studies of SE 2-DG patterns is that some structures are more prone to engage in SE than others. In the electrogenic limbic SE model in rat, although there is clearly some nonoverlap between restricted activation patterns corresponding to focal SE induced from different sites, the basolateral nucleus of amygdala, AHA, and the ventral hippocampal formation are highly prone to be involved, even when SE is evoked from amygdala, hippocampus, or medial prefrontal cortex (72). The tendency of temporal lobe seizures to spread to frontal lobe regions is well known to clinicians. For example, a case study showed that when a young epileptic patient with a medial temporal onset engaged in disturbed affective behavior, a seizure network involved the amygdala, orbitofrontal cortex, and frontal operculum in coupled activity (6).

One potential clinical implication is that, after SE has originated and spread from one site, such as a neocortical focus, other nuclei, for example amygdaloid or hippocampal, may become intensely engaged and thereby liable to damage.

From amygdala

The basolateral amygdala is connected with hippocampus, entorhinal cortex, perirhinal cortex, insular, prefrontal, and piriform/endopiriform regions, all of which might play a role in seizure generalization and propagation. In an interesting study by Imamura et al. (84), intra-amygdala kainate injection induced SE with mastication and clonic convulsions; histologic examination showed extensive amygdala and hippocampal damage. If a cut was made to separate the dorsal and ventral hippocampus, a less severe form of SE was induced by intra-amygdala kainite, and histologic examination revealed pronounced damage to the amygdala but not the hippocampus. This study suggests that the intact hippocampus plays an important role in promoting SE spread from the amygdala. Similarly, lidocaine injection of CA1 has been reported to reduce seizure activity and retard kindling from the amygdala (139).

From perirhinal cortex and adjacent structures

Perirhinal cortex. McIntyre et al. (124) demonstrated that layer V projects dense efferents to the frontal cortex. In addition, the anterior perirhinal cortex projects to insular cortex, claustrum, entorhinal and posterior piriform cortex, infralimbic cortex, basal amygdala, and nucleus accumbens. Convulsive seizures can be kindled very rapidly from the anterior perirhinal cortex (94, 144). Kainate injections of this structure trigger motor seizures (85). Lesions of the perirhinal cortex attenuate intra-amygdala kainate-induced mastication, facial twitching, and forelimb clonus (57), but lesions or local microinjection with procaine or glutamate blockers do not interfere with amygdala kindling (99, 170). Nonetheless, the available observations suggest that the perirhinal cortex could contribute to the spread of seizure activity from limbic regions to convulsive networks in SE.

Claustrum. Mohapel et al. (144) kindled the perirhinal cortex, adjacent insular cortex, or posterior claustrum and concluded that all three sites were highly epileptogenic, with very rapid kindling and short latency to convulsions. Zhang et al. (219) similarly described the anterior claustrum as highly kindling sensitive. Lesions of the anterior and posterior claustrum delay amygdala kindling to clonic convulsions (143) but do not prevent convulsions that have already been kindled (174). The anterior claustrum is connected to the frontal and motor cortex, amygdala, endopiriform nucleus, midline thalamus, nucleus accumbens, and substantia nigra. It is postulated that the claustrum participates in the development of limbic-onset generalized seizures (174, 219).

From piriform cortex and endopiriform nucleus

Piriform cortex. Kindling studies have shown that the piriform cortex kindles very rapidly, is recruited early when kindling is conducted at other sites, and is prone to interictal discharges. Injections of GABA-promoting drugs or NMDA antagonists into the piriform cortex block amygdala-kindled seizures (108). In 2-DG studies of SE propagation from amygdala, we have observed transitional forms between restricted and extensive limbic patterns in which deep piriform layers are selectively activated. Consistent with this observation, it has been shown in vitro and in vivo that epileptiform potentials can be evoked from deep piriform layers on stimulation of the basolateral amygdala (108). Bicuculline microinjections into the deep layers of piriform cortex have revealed that all regions of the piriform cortex are highly susceptible to seizures (46).

Central piriform cortex. Bilateral lesions of the central piriform cortex, but not of the anterior or of the posterior piriform cortex, retard kindling of convulsive seizures from the amygdala (169). Injection of vigabatrin into the central piriform cortex retards amygdala kindling (172) and is more successful than anterior or posterior piriform injections in blocking kindled convulsions (171). GABA-secreting cells implanted in the central piriform cortex increase the latency to convulsions but do not retard amygdala kindling (63). It is speculated that the central piriform cortex is especially important among the piriform areas in generalizing amygdala-onset seizures owing to its denser connections with all piriform regions, including contralaterally, orbitofrontal cortex, and rich connections with the entorhinal cortex and amygdala (171).

Endopiriform nucleus. Studies of the endopiriform nucleus indicate that this structure is highly epileptiform (reviewed by Behan and Haberly [7]). The endopiriform nucleus lies deep to the piriform cortex along its rostral-caudal axis. Figure 19.2 shows involvement of the endopiriform nucleus by SE. The endopiriform nucleus has dense intrinsic connections and projects to the piriform, insular, and amygdaloid cortical regions. It provides a massive caudal to rostral pathway that is lacking in piriform cortex. This structure is thus well positioned to promote seizure spread within olfactory cortex and related regions (7).

Deep anterior piriform cortex (area tempestas). Piredda and Gale (160) observed that injection of a low dose of bicuculline or other proconvulsant into the deep layers of a small region of the rostral piriform cortex readily produced limbic-originated clonic convulsions. The area of greatest predilection for producing convulsions was found to be quite circumscribed. Seizure activity elicited in the deep layer readily propagates to the cell body layer of overlying piriform cortex, which then participates in seizure propagation (43). Injection of a non-NMDA glutamate antagonist into the anterior piriform cortex reduces CA3 cell loss in systemic kainate-induced SE (93).

Examination of this area with standard histologic stains does not suggest why this region should differ functionally from other regions of piriform cortex that, as discussed above, are also seizure prone. However, Ekstrand et al. (49) have provided connectional and immunohistochemical evidence that this region is indeed distinctive. They term the cortex the rostroventral anterior piriform cortex (APCrv) and the underlying deep layer the pre-endopiriform nucleus (pEn). The pEn corresponds to the zone of greatest sensitivity to the seizure-producing effect of bicuculline. The APCrv directly and the pEn indirectly project to the ventrolateral orbital cortex. The pEn also projects to the submedial thalamic nucleus. Immunostaining indicates a paucity of GABA terminals on axon initial segments.

The model of forebrain seizures elicited by focal chemoconvulsants injected into the deep anterior piriform cortex has provided considerable utility for the study of anatomic networks regulating seizure spread, as noted in other sections of this chapter.

Potential role of thalamic nuclei in SE spread

2-DG studies of restricted and extensive limbic activation patterns associated with SE display activation of parataenial, dorsomedial, reuniens, and rhomboid nuclei (Figures 19.1 to 19.5). In addition, the medial portions of the anterior thalamus and the ventromedial thalamus may be involved.

Dorsomedial thalamus. In rat, topographic afferents to dorsomedial subnuclei arrive from medial prefrontal, orbital, cingulate, and agranular insular cortex regions, amygdala, substantia innominata, hypothalamus, and midbrain (32). Portions of the dorsomedial thalamus and the adjacent parataenial nuclei project topographically to medial precentral, cingulate, prelimbic, medial and lateral orbital, dorsal and ventral agranular insular cortical regions and to the basolateral amygdala nucleus (101). In the primate, the dorsomedial thalamus is reciprocally and topographically connected with prefrontal fields (64) and has received considerable attention for a potential role in memory functions and neurobehavioral disorders. The dorsomedial thalamus projects to the nucleus accumbens and ventral striatum, which also receive afferents from numerous structures connected with the dorsomedial thalamus. Thus the dorsomedial thalamus, as part of the limbic thalamus, is interconnected with selective regions of limbic/prefrontal cortex and basolateral amygdala, with these structures sending converging projections to the limbic striatum. The limbic striatum in turn projects to the ventral pallidum.

Neuron loss in the dorsomedial nucleus is prominent after prolonged SE in humans (56). Experimentally, Bertram et al. (14) found that afterdischarges propagate early to midline thalamic nuclei during kindling and that the dorsomedial, reuniens, and rhomboid nuclei show neuronal loss during kindling. Lidocaine microinjections shortened afterdischarge duration.

Noting that dorsomedial thalamus is damaged in SE induced from anterior piriform cortex (area tempestas), Cassidy and Gale found that intradorsomedial thalamic injection of the GABA agonist muscimol, the AMPA antagonist NBQX, or the GABA transaminase inhibitor vigabatrin protected against seizures induced by intrapiriform bicuculline injection (22). Injection of an NMDA antagonist into the dorsomedial thalamus had no effect. Cassidy and Gale (22) concluded that the dorsomedial thalamus plays a critical role in seizure generalization from the piriform cortex, and that AMPA and GABA receptors are crucial in this process.

Anterior thalamus. The subnuclei of the anterior thalamus receive topographically organized projections from the mammillary complex nuclei (180). The anteromedial subnucleus has complex topographic projections to area 2 of frontal lobe, anterior cingulate, and entorhinal cortex in the rat (181). The anteroventral and anterodorsal nuclei project to the retrosplenial cortex and parts of the subicular complex (198). From these connections it may be surmised that the anterior nucleus may serve as a conduit of hippocampal formation activity to recruit via excitatory projections important regions of the frontal lobe and limbic cortex, as well as amplifying limbic seizure activity by its projections back to the subicular complex and entorhinal cortex.

The anterior thalamus has received considerable recent interest as a therapeutic target. Mirski and Ferrendelli reported that muscimol microinjection into the anterior thalamus blocked pentylenetetrazol (PTZ)-induced seizure activity (140). Lesions of the mamillothalamic tracts also protected against PTZ-induced convulsions, while lesions of the fornices or mamillary bodies exerted efficacy but were not as effective (141). Miller et al. (137) found that injection of vigabatrin into the anteromedial thalamus blocked PTZ seizures without producing behavioral effects. These observations provided justification for preliminary clinical trials of deep brain stimulation of the anterior thalamus in the United States and Canada (80, 95). Preliminary observations include reductions in total seizures, convulsions, and seizure-related falls, with good tolerance. Indeed, patients could not tell whether they were receiving stimulation.

In an interesting experiment, Hamani et al. (68) report that rats receiving bilateral anterior thalamus DBS had delayed onset of SE induced by pilocarpine. However bilateral anterior thalamotomy was more effective, preventing entry into SE. If this finding is confirmed in other models, a potential therapeutic role of anterior thalamotomy might warrant consideration.

Anterior midline thalamus. Miller et al. (137) reported the interesting observation that microinjection of muscimol into anterior midline nuclei, namely, the interanteromedial, intermediodorsal, centromedial, and paraventricular, produced sedation and enhanced susceptibility to PTZ seizures. This finding suggests that normally this system promotes alertness and resistance to seizures. Conceivably, encephalopathies that impair this system could thereby promote susceptibility to seizures and SE.

Ventromedial thalamus. The principal ventromedial thalamic nucleus receives afferents from prefrontal cortex, rostral agranular cortex, and premotor cortex. It also receives significant extrapyramidal inputs from the internal globus pallidus, substantia nigra reticulata, and deep cerebellar nuclei (89). The ventromedial thalamic complex projects to the granular insular cortex and to the frontal cortical pole (101). Few studies have examined the role of ventromedial thalamus in seizures. Moshe et al. (145) found that lesions of this nucleus did not affect susceptibility to flurothyl-induced seizures.

Discussion

An interesting feature of the unilateral extensive pattern is that despite its capture of a large cerebral territory, the associated behavior is merely exploratory-like. Thus, even when SE is nonconvulsive, large limbic territories may be engaged by seizure activity. A comparable and not infrequent clinical situation is the EEG finding of extensive continuous ictus over temporal lobe regions, often bilateral, in complex partial SE. Such patients may have moment-to-moment cycling or continuous alteration in behavior. The altered behavior may range from frank psychomotor automatisms to subtle deficits in conversational patients. The latter has been termed subtle complex partial status by Treiman. (The “nonconvulsive SE” of complex partial SE must not be confused with that of absence SE or with the “nonconvulsive SE” of moribund patients in subtle generalized convulsive SE.)

Another interesting concept is that even extensive patterns of ictal involvement may depend on the involvement of one or a few nuclei acting as generators to drive the entire network. Few experiments have been performed to address this issue in limbic-onset SE; thus the experimental evidence for this concept is tentative. White and Price (209) showed that inactivation of the basolateral amygdala with focal lidocaine microinjection during limbic SE in the rat promptly abolishes unilateral extensive limbic SE. Lidocaine injection within anterior piriform cortex or hippocampus did not have this effect. Microinjection of the amygdala with the GABA agonist muscimol was also ineffective, likely because the GABA receptor becomes unresponsive in SE. In a comparable study, Hirsch et al. (79) found that lidocaine injections of the amygdala prevent seizure spread from brain stem to forebrain, whereas hippocampal injections are ineffective.

Complex partial SE is recognized as a cause of neuronal damage to temporal lobe structures such as the hippocampus in patients (50, 105). Fujikawa et al. (56) studied the brains of three patients who presented with focal motor SE and had EEG evidence of epileptiform activity in one or both temporal lobes during SE. None of the patients had systemic complications that could cause neuronal loss. SE lasted 9 hours to 3 days, and death occurred 11–27 days after SE onset. Neuronal loss occurred in the hippocampus, amygdala, dorsomedial thalamus, piriform and entorhinal cortex, and in Purkinje cells. These findings are comparable to the regions of damage occurring after electrogenic limbic SE in the rat (82).

Thalamic modulation of seizure susceptibility

Because of the role of the anterior thalamus in influencing frontal lobe and hippocampal formation activity, bilateral inactivation of this nucleus may be expected to impede limbic propagation of seizures and the recruitment of frontal cortex for convulsive activity.

The centromedian-parafascicular thalamic complex is of interest, as it projects extensively to the neocortex, the entire striatum, the globus pallidus, subthalamus, and substantia nigra. The centromedian nucleus in particular projects to motor regions (163, 165).

Velasco's group has described the beneficial effects of centromedian deep brain stimulation on seizure control and EEG discharges in patients with primarily and secondarily generalized convulsions, and atypical absences. However, they concluded that complex partial seizures and temporal lobe EEG discharges were not suppressed by centromedian stimulation (200, 201). Fisher et al. (55) performed a controlled clinical study of centromedian stimulation (2 hours per day) and found a trend toward reduction of tonic-clonic convulsions that was not statistically significant. On open follow-up, stimulation 24 hours per day appeared to be more effective. Further controlled studies are required to assess the efficacy of centromedian deep brain stimulation, but these preliminary results suggest that generalized convulsions, which require motor cortical involvement, may be ameliorated, whereas the complex partial seizures of limbic origination are not suppressed.

Basal ganglia modulation of seizure susceptibility

In a simplified version of basal ganglia circuitry, the substantia nigra reticulata and the globus pallidus interna (GPi, corresponding to the entopeduncular nucleus in rodents) are output structures, with inhibitory projections to the thalamus and brain stem. These nuclei receive excitatory afferents from the subthalamus that release glutamate, and GABA-containing afferents from the striatum. They also receive afferents from the globus pallidus externa. The dorsal and limbic striatum project respectively to the globus pallidus and ventral pallidus, which send GABA-containing efferents to the subthalamus. The striatum receives topographically organized convergent excitatory inputs from cortical, thalamic, and limbic regions. This circuit scheme is greatly simplified and does not explain well some problems, such as why GPi lesions ameliorate dyskinesias or dystonia. Nonetheless, it has proved useful in the development of surgical and pharmacologic therapies for movement disorders.

Substantia nigra reticulata. The discovery two decades ago that lesions or muscimol injection into the substantia nigra reticulata suppressed full generalization of seizures in maximal electroshock and chemoconvulsant models in the rat (59, 83) spurred great interest. Lesions or muscimol injections into substantia nigra also block kindled seizures (127) and convulsions elicited by bicuculline injected into the anterior piriform cortex (115). Since these original observations, it has become recognized that the effect of nigral manipulations is a function of age, sex, and nigral topography (62, 179, 203). In addition, it has been reported that although GABAergic inhibition of substantia nigra reticulata suppresses absence and clonic seizures in rats, which depend on forebrain circuits, this intervention fails to suppress audiogenic seizure-induced tonic seizures, which depend on brain stem circuits (36).

Bonhaus et al. (16) determined that nigral cells fire in bursts during generalized seizures in kindled rats but not in amygdala-induced seizures in unkindled rats. This observation corresponds with 2-DG evidence that the substantia nigra is not activated in highly restricted limbic SE patterns but becomes involved as seizure activity becomes anatomically more extensive (see Figure 19.4). Deransart et al. (34) similarly recorded substantia nigra reticulata bursting in association with cortical spike-wave discharges during absence seizures in GAERS rats. Chronically increased firing of posterior but not anterior reticulata neurons has been reported after amygdala kindling in response to generalized convulsions (61). Such chronic and seizure-associated nigral reticulata increases in firing would be expected to promote seizure generalization.

Intranigral blockade of glutamate receptors also blocks generalized seizures in a genetic model (37), and electroshock- and pilocarpine-induced seizures (33). Intranigral fluoxetine also suppresses anterior piriform cortex-elicited seizures, an action that depends on endogenous serotonin and is exerted through multiple serotonin receptor subtypes (153).

The above evidence is compatible with the notion that reticular nigral neurons normally fire tonically, with the downstream effect of facilitating forebrain generalized seizures. During seizures, nigral cells fire more, tending to promote seizure propagation. Maneuvers that suppress nigral neuronal firing, such as injection of GABA agonists or blockade of glutamate receptors, have an anticonvulsant effect. Recently, innovative approaches have shown seizure inhibition with intranigral GABA-releasing implants (109, 187) and deep brain stimulation in the rat (202) to suppress evoked or spontaneous seizures (188).

Dorsal midbrain and pontine reticular formation. On anatomic grounds, nigral efferents to the thalamus appeared to be the logical route of downstream effects on seizures. However, lesioning experiments indicated this was not the case (60, 145). Instead, the nigral target for seizure modulation was within the dorsal midbrain (60, 177). The nigrotectal target zone comprises an anticonvulsant zone that includes the deep superior colliculus, the adjacent mesencephalic reticular formation, and the intercolliculus. Bicuculline injections into this region protect against tonic hindleg extension in electroshock seizures (177) and against clonic convulsions elicited by focal injections of bicuculline into the deep anterior piriform cortex (58). Electrical stimulation of the superior colliculus suppresses absense seizures in the GAERS rat, as do picrotoxin injections into the superficial and intermediate caudal superior colliculus (147). Knife cuts of efferent pathways from this region suggest that the descending pathway, which projects to pontine reticular structures, is essential for the dorsal midbrain anticonvulsant effect (175, 178). It was further shown that bicuculline microinjection into the ventrolateral pontine reticular formation, a projection site of the dorsal midbrain anticonvulsant zone, protected against electroshock-induced hindleg extension (176).

Entopeduncular nucleus/globus pallidus interna. The GPi (entopeduncular nucleus) has similar connections as substantia nigra reticulata. Injection of muscimol or a glutamate blocker suppresses electroshock- and pilocarpine-induced seizures in rat (33, 81, 154). Some investigators have found microinjections in the entopeduncular nucleus to be less effective than in the substantia nigra (33, 81).

Subthalamus. Intrasubthalamic injections of muscimol suppress generalized seizures in the genetic absence model in the rat, motor seizures elicited by focal bicuculline injected into the anterior piriform cortex, flurothyl-induced convulsions, and convulsive seizures, but not afterdischarges in amygdala-kindled rats (35, 37, 45, 204). The therapeutic application of subthalamic deep brain stimulation for refractory clinical epilepsy has been examined in pilot studies (8, 24, 42, 107). The premise of subthalamic deep brain stimulation is that it may suppress its pro-seizure excitatory output to the substantia nigra reticulata and GPi, and also may block epileptic discharge excitation of the subthalamus, propagated by direct corticosubthalamic projections (42).

Globus pallidus externa and ventral pallidum. The globus pallidus receives afferents from sensorimotor striatum, whereas the ventral pallidum receives afferents from the limbic striatum, including nucleus accumbens. Tonically firing neurons of these pallidal structures project inhibitory fibers to the subthalamus. Disinhibition of pallidal neurons with local injection of a GABA antagonist, which causes these neurons to fire more, suppresses absence seizures in a genetic rat model, presumably by inhibiting the subthalamus and preventing it from exerting its normal pro-seizure effect on substantia nigra reticulata and entopeduncular/GPi nucleus. Indeed, this seizure suppression is associated with a fall in extracellular glutamate levels in the substantia nigra, as expected. Moreover, injection of a GABA agonist into pallidal structures aggravates absence seizures, as predicted (38).

Despite the coherence of these results, other evidence suggests that labeling the globus pallidus externa as an antiseizure nucleus may be an oversimplification. Chen et al. (26) found no effect of local zolpidem on systemic PTZ-induced tonic seizures and mortality in rats, but intrapallidal injection of the GABA_B receptor agonist baclofen completely blocked tonic seizures and mortality. Intrapallidal injections of picrotoxin cause lethal seizures that are prevented by injecting amphetamine into the striatum (215). Injection of kainate into the rat globus pallidus externa induces multiple seizures over hours, including clonic convulsions. The 2-DG shows mainly unilateral activation of limbic, cortical, basal ganglia, and thalamic structures (167).

Striatum. Striatal output neurons project GABAergic efferents directly to the GPi/entopeduncular nucleus and substantia nigra; activation of this direct pathway would be expected to suppress seizures. On the other hand, the striatum also projects to these output nuclei indirectly via GABAergic projections to the globus pallidus and ventral pallidum, which inhibit the subthalamus. Activation of the striatum's effect on the indirect pathway would be expected to be proconvulsant, as the pro-seizure effect of the subthalamus would be disinhibited. Experimentation indicates that of these two outcomes, the action of the direct pathway generally prevails. Injection of NMDA or bicuculline into caudate-putamen in rat suppresses pilocarpine-induced seizures (194, 195), and similar findings have been reported in other models (39).

Discussion of basal ganglia role. The preponderance of experimental evidence indicates that activated striatal neurons inhibit GPi and substantia nigra reticulata via a direct projection, thereby disinhibiting antiseizure neurons in the dorsal midbrain, which may act by exciting the ventrolateral pontine reticular formation. Activation of the subthalamus, by striatal inhibition of GPe or ventral pallidum, thereby causing subthalamus disinhibition, or direct excitation by afferents from cortex or thalamus has a pro-seizure effect, mediated by increased inhibitory outflow from nigral or GPi neurons. Although this body of evidence is already leading to exploration of new therapies, such as GABAergic implants or deep brain stimulation, there are several gaps in our knowledge. Because the work has been conducted with rats, it is not clear whether the GPi may be more or less important than substantia nigra for seizure control in primates. Much needs to be learned about the downstream effector pathways.

2-DG autoradiographs reveal intense metabolic activation of parts of striatum, the globus pallidus, the ventral pallidum, and the substantia nigra reticulata when extensive regions of cerebrum are involved in SE. Damage to the substantia nigra reticulata is a well-known consequence of experimental SE, likely reflecting excessive glutamate release from subthalamic afferents. Interesting questions are whether the dominant effect of striatal activation by seizure activity shifts toward the pro-seizure indirect pathway during the development of SE, and whether the subthalamus loses its inhibitory response to direct striatal afferents while retaining excitatory responses to cortical afferents.

Brain stem and basal forebrain modulation of forebrain generalized seizures

A number of brain stem structures have been studied with regard to their role in modulating forebrain seizures, and thus may be important in influencing SE. In addition, brain stem structures themselves may be involved by seizure activity. The latter issue is considered in a separate section.

Periaqueductal gray. The PAG receives afferents from the lateral reticular, oral and caudal pontine reticular, raphe magnus and pallidus, zona incerta, thalamic parafascicular, and several hypothalamic nuclei (116). The dorsolateral PAG projects efferents mainly to the locus ceruleus, subceruleus, A5 cell group, parts of the paragigantocellular and gigantocellular nuclei, the centrolateral and paraventricular thalamus, and anterior hypothalamus. The ventrolateral PAG projects mainly to the nucleus magnus, caudal paragigantocellularis, rostroventrolateral reticular formation, centromedian and parafascicular thalamus, lateral hypothalamus, and lateral bed nucleus of stria terminalis (18, 19). Peterson et al. (158) elicited seizures of the forebrain type on carbachol microinjection into the PAG, associated with forebrain electrographic discharges. Thus the PAG may, on excitation by forebrain or brain stem afferents, promote forebrain seizure activity.

Cerebellum and deep cerebellar nuclei. Electrical stimulation of the cerebellum cortical surface was initially felt to be effective for intractable epilepsy in open treatment, but it did not prove to be effective in a controlled trial (212).

Studies of the role of cerebellar regions in relation to seizures have yielded conflicting results. Here we cite several studies and do not attempt to review this field, which has not received thorough investigation with modern techniques. Electrolytic lesions of the fastigial or of the dentate nucleus lead to accelerated amygdaloid kindling, suggesting that these nuclei inhibit seizures (138). Such lesions may transect axons of passage, but these results partly confirm the findings of Miller et al. (135), who employed the microinjection technique, which avoids affecting passing axons. Injections of GABA agonists into the fastigial nuclei increase susceptibility to systemic bicuculline-induced seizures in the rat, suggesting that the fastigial nucleus outflow normally is seizure-inhibitory. Injection of GABA agonists into the dentate nuclei had no effect (135).

On the other hand, transection of the cerebellar peduncles retards amygdaloid kindling, suggesting that the cerebellar output is normally pro-convulsive (155). In partial support of this notion, Rubio et al. (164) stimulated the superior cerebellar peduncle with 100-Hz pulses during amygdala kindling in the rat and found facilitation of the early stages of limbic kindling, but shortening of the kindled clonic convulsion. Kandel and Buzsaki (90) recorded unit bursting in cerebellar cortex and deep nuclei in synchrony with spike-and-wave discharges in a rat model of absence seizures and suggested that the cerebellum contributes to generalized spike-and-wave activity of absence seizures. Clinical case reports have described motor seizures and simple partial motor SE in patients with isolated cerebellar lesions, with propagation of discharges to the cerebral cortex (129, 196). In the case described by Mesiwala et al., ictal activity was recorded from a cerebellar ganglioglioma, and all seizures ceased with resection of the cerebellar lesion.

Laterodorsal tegmental reticular nucleus. The brain stem contains structures with neurons that release specific neurotransmitters within the forebrain. Miller and colleagues studied the laterodorsal tegmental nucleus of the pons/midbrain, which contains cholinergic neurons, and found that microinjection of a GAB_A or GABA_B agonist lowered the threshold to clonic but not tonic seizures produced by systemic bicuculline or PTZ in the rat (132). In another experiment it was determined that microinjection of a neurotoxin specific for cholinergic neurons resulted in a lower threshold for PTZ-induced face and forelimb clonic seizures, which are forebrain-mediated. It was thus suggested that the laterodorsal tegmental nucleus exerts an important antiseizure effect on the forebrain (134). Further studies indicated that the target site is the centromedial thalamic nucleus, which receives cholinergic projection from the laterodorsal tegmental nucleus. Injection of inhibitory substances into either site suppresses behavioral arousal and exacerbates forebrain convulsive seizures (133, 136).

Cholinergic basal forebrain. Ferencz et al. (53) lesioned select regions of cholinergic neurons by microinjecting an immunotoxin into basal forebrain structures and were able to show that depletion of cholinergic neurons in the nucleus basalis facilitates the late stages of amygdala kindling. By contrast, depletion of cholinergic neurons in the medial septum and vertical limb of the diagonal band facilitates the early stages of hippocampal kindling (54). Depletion of basal forebrain cholinergic neurons with saporin also shortens the latency to generalized convulsions in rats exposed to flurothyl or PTZ (182). These results suggest that basal forebrain cholinergic neurons exert suppressive effects on seizure spread and generalization.

Locus ceruleus. Neurons of this nucleus project norepinephrine-releasing terminals to widespread regions of forebrain. Depletion of norepinephrine within the amygdala by local injection with 6-hydroxydopamine accelerates amygdala kindling (120). Because the locus ceruleus (LC) projects fibers to the amygdala and other parts of the limbic system, this finding implies a role for LC in controlling seizures. Transection of ascending noradrenergic pathways accelerates kindling (48), as does administration of 6-hydroxydopamine or DSP-4 so as to deplete norepinephrine terminals or LC cell bodies (21, 54, 122, 151). These lesions promote both focal and generalized forebrain seizures. Mishra et al. (142) demonstrated that lesions of LC with a selective neurotoxin lowered the threshold to corneal electroshock seizures in rat. LC lesions also increase the frequency of absence seizures in the genetic rat model, but the effect is transient (106). In addition, LC lesions confer greater susceptibility to audiogenic seizures, supporting a role in regulating brain stem seizure susceptibility (88).

Krahl et al. (100) demonstrated that the LC is an essential relay in vagus nerve stimulation seizure suppression. When the LC was lesioned, vagus nerve stimulation no longer suppressed electroshock-elicited seizures.

Attempted kindling at short intervals is blocked by endogenous seizure-induced seizure-suppressive mechanisms (69, 126). McIntyre et al. (126) showed that short-term suppression could be ameliorated by norepinephrine depletion, suggesting that norepinephrine plays a role in seizure-activated seizure suppression. This notion is compatible with evidence that in reaction to activation by a seizure, the LC markedly increases forebrain norepinephrine release; a response lasting approximately 8 minutes in the rat (11). These characteristics of the LC system would thus suggest that it is important for preventing repeated discrete seizures from occurring and merging into continuous SE. Contrary to expectations, McIntyre and Edson (123) did not find that norepinephrine depletion affected entry into SE after 1 hour of amygdala stimulation. However, Giorgi et al. (65) described experimental results that suggest that LC-released norepinephrine may play an important role in preventing the conversion from discrete seizures to continuous SE. They found that when low-dose bicuculline is injected into the anterior piriform cortex, the normal response is self-limited sporadic seizures. If the animal previously received systemic DSP-4 to lesion LC neurons, the same dose of bicuculline elicited self-sustained, continuous, long-lasting SE.

Nucleus of solitary tract. The vagus nerve terminates predominantly on the nucleus of the solitary tract. Muscimol injection of the caudal medial subdivision, but not of other subdivisions, suppresses seizures elicited by bicuculline methiodide injection into anterior piriform cortex, or by systemic bicuculline or PTZ. Microinjection of kynurenate, a glutamate receptor blocker, or the local anesthetic procaine has similar effects. Injection of bicuculline methiodide into the medial nucleus of the solitary tract has no effect. These results were interpreted as suggesting that inhibition of medial nucleus of solitary tract outflow leads to resistance to forebrain limbic motor seizures (205).

Magdaleno-Madrigal et al. (114) electrically stimulated the medial nucleus of solitary tract with 30-Hz pulses just prior to amygdala kindling stimuli in cats and observed prevention of kindling to the stages of secondary seizure generalization. Stimulation of the nearby nucleus intercalatus or the lateral tegmental field had no effect. The effect of solitary nucleus stimulation on preventing kindling progression was striking, and the authors speculated that the LC may in part mediate this action.

Dorsal raphe nucleus. Racine and Coscina (161) found that lesions of the dorsal and median raphe facilitated amygdala kindling, but concluded the effect was not robust. Neuman and Thompson (148) found that the suppressive effects of peripheral noxious stimulation on a penicillin cortical focus could be abolished by drugs that lowered the firing rate of raphe neurons. Injection of drugs directly into the dorsal raphe, which reduces neuronal firing, also blocks the effect of peripheral stimuli in producing cortical desynchronization and suppression of focal cortical epileptiform activity (189). Parallel observations were made by Moyanova et al. (146) after lesions of the dorsal raphe. These observations suggest that the suppressive effect of serotoninergic dorsal raphe neurons on seizure spread is more apparent when the dorsal raphe is activated.

Discussion

Examination of 2-DG autoradiographs indicates that once limbic-onset SE has captured extensive domains of the cerebrum, the originating SE site can no longer be identified, because the anatomic SE substrate and behavioral correlate become the same for different originating sites.

Experimentally, the same 2-DG activation pattern and ictal behavior can occur with a variety of originating SE sites in rats with exploratory SE, minor convulsive SE, or clonic convulsive SE, so that an observer would be unable to determine from where within the forebrain the SE had originated. Similarly, the clinician may be unable to assess where seizure activity originated in a patient presenting with secondarily generalized convulsive SE.

In this section we provided a survey of thalamic, basal ganglia, and brain stem structures that could potentially modulate the occurrence of generalized convulsive SE. This evidence is based mainly on animal seizure models that range from single seizures to SE. Kindled rats are not usually epileptic, yet prior kindling facilitates SE entry in rat (70, 125). Few studies of SE have been performed in epileptic animals, although many patients with clinical SE have prior epilepsy. Prior epilepsy may alter the ease of SE entry, the severity of SE, and the portions of cerebrum affected.