JLE

Epileptic Disorders

MENU

Pathogenesis of Lennox-Gastaut syndrome: Considerations and hypotheses Volume 3, issue 4, December 2001

Figures

See all figures

Lennox-Gastaut syndrome (LGS) has been defined as: 1) a childhood epileptic encephalopathy with multiple types of seizures, particularly tonic attacks, but also atypical absences and atonic seizures, and 2) interictal bilaterally synchronous slow spike-waves (SSW) on EEG [1-3]. Genton et al. [4] require that runs of bilateral paroxysmal fast activity, often associated with tonic seizures, must also be present to confirm Lennox-Gastaut syndrome.

Despite innumerable writings on Lennox-Gastaut syndrome (LGS), its pathogenesis remains obscure [5]. Three factors have limited our knowledge in this area: 1) lack of an animal model for the disorder, although models for its component epileptic seizures exist; 2) less progress in elucidating its genetics than for other generalised epilepsies, and 3) lack of a single specific pathology.

This article integrates clinical and experimental data relevant to LGS, to create a proposed scheme for its pathogenesis (figure 1).

"Lennox-Gastaut" is a syndrome

An epilepsy syndrome has been defined as a complex of signs and symptoms that defines a unique epilepsy condition [International League Against Epilepsy Task Force on Classification and Terminology, unpublished]. The following data support the concept that the defining features of Lennox-Gastaut syndrome occur together more frequently than they would by chance. Studies [6, 7] show that a medically-intractable generalised childhood onset seizure disorder is virtually always present among patients whose interictal, awake EEGs demonstrate slow spike-waves. Thus, 98% of the patients of the Blume et al. [7] SSW series had this type of epilepsy picture. Tonic seizures and atypical absences occurred in 56% and 44% of patients with slow spike-waves studied by Chevrie and Aicardi [6]. Among 20 patients whose EEGs demonstrated "generalised paroxysmal fast activity (GPFA)" studied by Brenner and Atkinson [8], all had a history of seizures, usually of multiple types. Eleven (55%) had tonic seizures, while absence (presumably atypical) occurred in 14 patients (70%). Significantly, 10 of the 20 patients with GPFA also had SSWs.

Conversely, Gastaut and Broughton [9] found SSW to be the most frequent interictal EEG accompaniment in patients with tonic seizures. Patients with atypical absence were found to have runs of 20-10 Hz fast activity or SSW [10]. Finally, Gastaut himself [11] defined Lennox-Gastaut syndrome as a "particular association of tonic seizures, atypical absences, and marked mental deficiency, plus interictal EEG records of irregular 1.5 to 2.5 CPS diffuse sharp and slow-wave complexes..." (Sharp and slow-wave complexes is a previous term for slow spike-waves.)

Associated conditions

Several lines of evidence implicate the cerebral cortex and the corpus callosum in the production of some of the phenomena of LGS. Antecedent conditions associated with LGS almost always involve the cerebral cortex. Among lesional causes, those involving both frontal lobes most commonly lead to the development of LGS [12-14]. However, any type of brain damage may be associated with LGS [13] and this lack of specificity has impeded the clarification of neurophysiological mechanisms. However, Roger and Gambarelli-Dubois [15] found at autopsy dysplastic lesions in sixteen out of thirty patients with LGS. Focal cortical malformations and other multifocal brain lesions are also listed as associative conditions by Aicardi [16] and by Genton et al. [4]. Bisynchronous spike-waves are commonly associated with multifocal lesions and epileptiform discharges, both clinically [17] and experimentally [18].

Some etiologies, such as perinatal anoxia and encephalitis, implicate multiple levels of the nervous system. Taken together, the clinical and basic science studies outlined below suggest that a caudal to cranial progression of epileptogenesis occurs with age of seizure onset and with chronological age. Chevrie and Aicardi [6], studying childhood epileptic encephalopathy with slow spike-waves, found that tonic seizures began at a mean age of 16.6 months whereas atypical absences began at 32 months, myoclonic attacks at 39 months, and clonic or tonic-clonic seizures at 42 months. Corresponding with these data are effects found by Velasco and Velasco [19], who obtained tonic-clonic seizures when perfusing the caudal mesencephalic brain stem with pentylenetetrazol, but obtained myoclonic seizures when perfusing its more rostral portion. Lesion studies indicate that the clonic component of tonic-clonic seizures is mediated by the forebrain and the tonic component by the mesencephalic reticular formation [20]. Therefore, the later onset of myoclonic and absence seizures in humans reflects increasing cortical participation in epileptogenesis [21]. Myoclonic-atonic seizures, usually occurring in patients older than those with tonic seizures, are likely consequent to corticofugal discharges descending to excite the pontine and medullary reticular formation. Brainstem descending fibres may excite or inhibit spinal motor neurones to produce myoclonic, tonic or atonic seizures [22-27]. A cortically-originating myoclonic event often preceeds an atonic event [28].

The complete seizure profile may persist into adulthood in almost half of patients with LGS [29]. In other patients, an evolution toward attacks representing those of a slightly more mature nervous system occur: brief tonic spasms become more prolonged; clonic components and tonic-clonic seizures may appear in addition to the classical tonic seizures. Thus, the time course in some patients supports the caudal-to-cranial evolution of seizure types as the brain matures.

Encephalopathies beginning in early life which are not associated with LGS may paradoxically suggest its pathogenesis. As the clinical and electrographic characteristics of LGS reflect widespread synchrony of epileptic discharge, disorders impeding such spread would not manifest as LGS. Patients with Aicardi's syndrome do not have LGS suggesting that the corpus callosum may be a significant vehicle for bisynchrony in some cases [30]. Improvement or cessation of LGS-associated drop attacks by callosotomy further supports this role [31]. Lissencephaly has plume-like bundles of myelinated fibres that punctuate the deep cortical neuronal layer into columns, impeding epileptic discharge propagation; it is rarely associated with LGS [30, 32, 33].

However, recurrent seizures are the first sign of central nervous system illness in the majority of patients with LGS [6, 7, 34], the disorder remaining cryptogenic in these cases. Although this implies a genetically-related pathogenesis, supporting data for this are meager (below).

Relationship with West syndrome

West syndrome evolves to LGS almost exclusively among the symptomatic West group [35, 36]. This may reflect presumed age of initial central nervous system insult: 97% of apparent causes of LGS occurs within the first twelve months of age [35] while idiopathic LGS begins at 1-5 years [37]. At time of writing, a clear physiological implication of these distinctions is lacking.

Physiological bases of electrophysiological features

No experimental paradigm has been created to represent the complete profile of LGS. Padjen [38] has stated: "There are no animal models of epilepsy - only models of pathological processes." Therefore, its pathological mechanisms must be inferred from several related studies. Several of the mechanisms proposed herein apply to other distinct syndromes that share clinical and/or electrographic features with LGS, e.g. absence for pathophysiology of bisynchronous spike-waves and focal epilepsy for cortical hyperexcitability. Future studies may reveal that an unusual coincidence and interaction of the several phenomena described below may be the defining pathophysiology of LGS.

Clinical data

Gloor et al. [39] found that patients with diffuse cortical or subcortical grey matter disorders have bilaterally synchronous discharges such as slow spike-waves. Slow spike-waves may also represent multifocal cortical epileptogenic foci [17]. Similarly, Mutani et al. [18] experimentally obtained bilaterally synchronous discharges from chronic asymmetrical foci as indicated above. An intense, uni-focal, cortical discharge could produce bisynchronous spike-waves as demonstrated in humans by Bickford and Klass [40]. Frontal lobe foci have a greater propensity for widely synchronous discharges and bilaterally synchronous epileptiform discharges than other areas [17, 41].

Spike-waves and slow spike-waves

The following reviews mechanisms involved in the generation of bisynchronous spike-waves and slow spike-waves. As these phenomena require the concerted participation of cortical and thalamic neurones, relationships among these players are reviewed.

Cortical events

The feline generalised penicillin epilepsy model [42] disclosed several properties of bisynchronous spike-wave discharges that were consistent with the aforementioned clinical data. Although ~ 2 Hz spike-waves can be generated by focal intra-cortical bicuculline in cats made athalamic ipsilaterally [43], bilateral participation of both the thalamus and cortex occurs in development of the spike-wave complex. Spike-waves produced by systemic injections of penicillin or bicuculline appear in the cortex before the thalamus [43-45]. Spontaneously occurring spike-waves begin in the cortex in awake monkeys [46] while multisite recordings show that those in anaesthetised cats also begin in the cortex before the thalamus [47, 48]. Moreover, Fisher and Prince [44] produced spike-wave discharges by bilateral cortical penicillin application but thalamic application did not elicit spike-waves. However, inactivation of the cat thalamus by potassium chloride can abolish spike-wave discharges [49].

Micro-electrode recordings revealed that the negative spikes of spike-wave complexes are associated with depolarising potentials resembling excitatory post-synaptic potentials (EPSPs) in upper cortical layers while the succeeding positive troughs are linked with excitation in lower cortical layers [50, 51]. In contrast, cortical action potential (AP) firing virtually ceases during the wave [52]. During the spike component of 2-4 Hz spike-waves in the cat, intracortical regular-spiking neurones [53] discharge sequences of APs while fast rhythmic bursting neurones (FRB) fire even more APs. During the wave, both classes of neurones become hyperpolarised and cease firing action potentials [54]. The first 40-80 ms of this hyperpolarisation may be chloride-dependent while the subsequent, longer-lasting component may reflect outward potassium current from spike AP-induced calcium influx [54]. This pattern of oscillation between a high incidence of AP discharge (spike and trough) and longer periods of near neuronal silence (wave) has been found in both cortex and the thalamus.

Cortical-thalamic interactions

The neuronal aggregates participating in the spike-wave phenomenon include: cortico-thalamic neurones and thalamo-cortical relay cells (TCR) (excitatory), and adjacent cortical and thalamic GABAergic interneurones (inhibitory). Cortical pyramidal cells and thalamocortical cells have mutually excitatory connections [55]. Thalamic neurones normally operate with two modes of firing: relay and oscillatory. Acting in the alert state, the relay mode conveys detailed afferent sensory data to the cortex. Relatively depolarised, thalamic neurones fire tonically as a sequence of single APs in the relay mode. In contrast, the oscillatory mode is involved in cortical synchronisation in drowsiness or sleep, but also serves to detect the occurrence of novel stimuli while awake [56]. Thalamic neurones are hyperpolarised and fire in rhythmic bursts [57]. The oscillatory firing mode underlies sleep spindle generation and spike-wave discharges [48, 56]. It appears that "generalised" seizures (e.g. spike-waves) emerge "without discontinuity from normally synchronised brain electrical activity" (e.g. spindles and delta activity) [48].

Thalamic neurones that do not project to the cortex are mostly GABAergic of two principal subtypes: local circuit interneurones and thalamic reticular neurones. Such thalamocortical excitation is thereby regulated through local circuit GABAergic interneurones within the thalamus and the cortex and in the thalamic reticular nucleus. Local circuit interneurones mediate feed-forward and feedback local circuit inhibition whereas thalamic reticular neurones produce a more diffuse and strong inhibition [55, 57]. The thalamic reticular nucleus consists entirely of GABAergic neurones projecting to most other thalamic neurones, most notably to thalamocortical relay (TCR) nuclei [58]. They receive excitatory input from thalamocortical and corticothalamic cells.

Both TCR and thalamic reticular neurones possess a low-threshold calcium current that is de-inactivated by membrane hyperpolarisation. This calcium current switches the TCR firing mode from tonic to bursting [55, 58].

The thalamic reticular nucleus has been identified as the pacemaker for the iniation of sleep spindle oscillations [59]. A rhythmic burst output of the inhibitory thalamic reticular neurones produces rhythmic inhibitory post-synaptic potentials (IPSPs) on thalamo-cortical neurones, hyperpolarising them toward the voltage range of inward calcium current de-inactivation and production of the low threshold calcium "spike" (AP) in some thalamocortical neurones [57, 60, 61]. Such rebound activation depends principally (but not exclusively) on GABAB receptor-mediated IPSPs in thalamocortical cells [62].

Dual intracellular recordings in thalamic reticular and thalamocortical cells in the ferret indicate that a burst of 2-6 APs in thalamic reticular cells activates IPSPs in thalamocortical cells principally through GABAA receptors, similar to those underlying spindles. Increasing the train of action potentials to > 10 activates GABAB receptors and consequently a more prolonged IPSP [55]. Similarly, Kim et al. [63] and Blumenfeld and McCormick [64] found that brief thalamic reticular cell bursts produced 100-150 msec IPSPs in thalamocortical cells mediated by GABAA receptors, while more sustained bursts elicited 300 msec GABAB-mediated IPSPs in addition to GABAA IPSPs. Blumenfeld and McCormick [64] demonstrated that increased firing in the corticothalamic pathway can actually transform the thalamic reticular cell discharge from brief to sustained bursts thereby effecting conversion to GABAB IPSPs in thalamocortical cells. The longer duration GABAB IPSPs slow the rate of reverberatory activity between thalamic reticular and thalamocortical neurones. This relationship suggests that increased cortical excitability at least partially underlies the slower rate of spike-wave discharges (3 Hz) than spindles (14 Hz) and that slow spike-waves (2 Hz) reflect even greater excitability.

Further data support the pivotal role of thalamic reticular cells in producing spike-waves and slow spike-waves. Blocking GABAA receptors in the thalamic reticular nucleus increases action potentials in thalamic reticular neurones, presumably by preventing inhibition by other reticular neurones [65]. In the gamma-hydroxybutyric (GHB) rat absence model, GABA inhibition in thalamic reticular neurones is reduced, facilitating burst firing in these GABAergic cells [66]. The low voltage-activated calcium current of thalamic reticular cells is increased in the Genetic Absence Epilepsy Rat from Strasbourg [58]. As this increase occurs after post-natal day eleven, it coincides with enhanced epileptogenicity in other models (see below). However, other factors likely create spike-waves in this model as they only become fully developed by day 30-40. Lesions of the thalamic reticular nucleus abolish spike-waves in this model [58]. A single thalamic reticular neurone may simultaneously, but transiently inhibit >= 100 thalamocortical cells. The subsequent low threshold calcium-driven action potential bursts in thalamocortical cells may then excite an even larger portion of neocortex (see below); the consequent augmented cortical-driven excitation of the thalamus completes and accentuates the oscillatory activity [67]. However, only ~ 40% of thalamo-cortical cells participate in this process. The majority of TC cells remain hyperpolarised with phasic IPSPs, and do not participate in burst firing [48].

Cortical hyperexcitability

The functional consequence of burst generation in thalamocortical neurones is amplification of the effect of such excitatory input by mechanisms intrinsic to the neocortex. High frequency excitatory post-synaptic potentials will summate and may increase synaptic efficiency [68, 69]. Excitation of burst-generating neocortical neurones, particularly intrinsic-bursting layer V pyramidal cells, will excite other such neurones through axone collaterals creating a powerful recurrent excitation in this network of neurones [70]. Evidence of a particular predominance of excitation in the immature cortex is presented in a subsequent section.

Animal models of generalised spike-wave seizures have disclosed an increased effectiveness of cortical glutamatergic neurotransmission mediated by AMPA receptors that may adopt characteristics of an NMDA-receptor-like response [58]. Additionally, Pumain et al. [71] found NMDA responses to be more widely distributed in absence rats than in controls.

Further evidence of the predominance of excitation over inhibition in the cortex during generalised epilepsy comes from the gamma-hydroxybutyric (GHB) absence seizure model in the rat: GABA, but not glutamate, is decreased in laminae I-III of frontal cortex at the onset of absence seizures [66]. This suggested to these authors a presynaptically-mediated GHB-induced impairment of GABAergic transmission in this model. Additionally, a possible decrease in the density of GABAA receptors in the cortex may make remaining GABA available to GABAB receptors and therefore enhance longer-duration GABAB-mediated inhibitory post-synaptic potentials (IPSPs) [58] as indicated earlier. This accentuated prolongation of IPSPs could convert a spike-wave generating system into one producing slow spike-waves.

The predominance of excitation over inhibition in the cortex increases the discharge of corticothalamic neurones, creating a substantial phasic excitation of thalamic reticular, thalamocortical and local GABAergic interneurones [55]. Activation of the corticothalamic pathway may directly excite and depolarise some thalamocortical cells. Others hyperpolarise through disynaptic inhibitory routes via thalamic reticular and local inhibitory neurones. Hyperpolarisation may generate low threshold calcium "spikes", thus action potential bursts, feeding the oscillatory cycle.

Therefore, two principal mechanisms for spike-wave, and presumably slow spike-wave generation appear: 1) abnormally strong activation of thalamic GABAergic neurones by corticothalamic afferents [72, 73]; and 2) loss of GABA receptor-mediated inhibition among thalamic reticular cells facilitating action potential bursts [66, 74].

Cortical slow spike-waves have occurred in the feline generalised penicillin epilepsy model when thalamocortical interaction has been destroyed or grossly interrupted, either by severing thalamic input to the cortex [75] or through systemic hypoxia [76]. Additionally, spike-waves as slow as 2 Hz have been produced by several feline generalised epilepsy paradigms [43, 48, 54]. However, prolongation of the wave of the slow spike-wave may result from lengthy corticothalamic bursts emanating from an immature cortex possessing the excitatory characteristics to be described below.

Fast rhythmic waves

Bursts of fast rhythmic waves or polyspikes, at 10-20 Hz, known also as the "epileptic recruiting rhythm", characterise and define Lennox-Gastaut Syndrome [9]. This activity may be associated with tonic seizures or absence with tonic features [77, 78]. The experimental correlate of this fast activity appears to be "fast runs". During such activity, most cortical cells are tonically depolarised [55].

These high frequency bursts ("fast runs)" are preceded by an enhanced amplitude of ~ 100 Hz activity ("ripples"), constituting the pattern of fast rhythmic burst (FRB) neurones [54]. When depolarised, these recently described layer V-VI pyramidal-shaped cells produce 300-600 Hz action potential bursts recurring rhythmically at 20-40 Hz [79]. Such behaviour probably results from cortical and thalamic excitatory afferents as well as properties intrinsic to these neurones [80, 81]. During "fast runs", regular spiking neurones are tonically depolarised and discharge single or double-action potentials [79].

The high frequency action potential rate within bursts augments synaptic "efficiency" [69] and favours rapid propagation of excitation within the neocortex via horizontal axones of FRB neurones [79, 82]. Such excitation can induce long-lasting depolarisation of post-synaptic neurones either directly or, through local interneurones, by inhibiting other inhibitory neurones [83]. Thus, cortical excitation overwhelms inhibition during such bursts of "fast runs."

Several data suggest that "fast runs" can be entirely generated within the cortex including: 1) reflection of fast runs as short latency EPSPs in related thalamic neurones suggesting activation via corticothalamic fibres, and 2) their elicitation by electrical stimulation of isolated cortical slabs [55, 80].

Cortex, thalamus and synchrony

Although both electrographic signatures of Lennox-Gastaut syndrome appear to be generated principally in the neocortex, the thalamus is significantly involved. Thalamic reticular and thalamocortical cells are depolarised and emit trains of action potentials during 10-15 Hz bursts of "fast runs" [48, 80]. In contrast, spike-wave or polyspike-wave complexes synchronously elicit bursts of action potentials in the GABAergic thalamic reticular cells and thus in thalamocortical cells. The pronounced inhibitory mechanisms described above in cortical, thalamic reticular and thalamocortical cells appear to underlie the synchronous activity of these structures during non-REM sleep rhythms and "spike-wave epilepsy". A greater preponderance of cortical excitation may slow the spike-wave repetition rate to that of slow spike-waves, and also would lead to a more severe epilepsy. As rebound action potential bursts follow inhibitory periods within a narrow time window, such inhibition facilitates the synchrony and incidence of these bursts [48]. The propensity of FRB cells to fire high frequency action potential bursts enables them also to play a crucial role in synchronisation and possible initiation of Lennox-Gastaut seizures [54].

Seizures and the immature brain

As the intractable generalised epilepsies usually begin in early childhood, a review of factors underlying heightened epileptogenicity of the developing nervous system may disclose mechanisms of LGS.

Lower seizure threshold

Several experimental models have demonstrated that rat pups have a lower threshold for seizures then do adult rats [84]. The most pronounced seizure susceptibility occurs in weeks 2 and 3 of post-natal life in both the hippocampus [85] and in the neocortex [86, 87]. This period of principal seizure susceptibility roughly corresponds to a full term neonate in humans [88].

Mechanics of excitability

A developmental window of enhanced neocortical seizure susceptibility appears to open at this time during a particular balance of effective excitation and inhibition. A greater resistance of the neocortex to convulsant-induced seizure activity before this may be due in part to a higher threshold for evoking EPSPs and their depression at higher stimulus frequencies [89]. However, evoked NMDA and non-NMDA excitatory synaptic currents occur in the frontal neocortex within the first week of post-natal life in the rat, whereas GABAA and GABAB IPSPs do not appear until days 9-14 [89, 90]. Density of both NMDA and AMPA receptors on hippocampal and neocortical neurones also increases during this period of enhanced seizure susceptibility [91]. GABAergic inhibition appears to develop earlier in the hippocampal CA1 region [84].

Non-NMDA-gated channels, which in the adult admit little calcium, may admit greater quantities of calcium in the immature cortex [92]. Glutamate transmission is increased as NMDA receptor-mediated activity is less sensitive to magnesium block, and the NMDA receptor-mediated current is more prolonged [93]. Moreover, immature neurones show lower levels of sodium-potassium ATPase activity than mature cells; this may account for a less effective maintenance of extracellular potassium concentrations in immature tissue [94, 95]. The higher extracellular potassium levels lower the seizure threshold. Gap junctions may exist in greater quantity in the immature nervous system as revealed by intracellular dye injections [96]; this may enhance epileptic discharge propagation (see below).

Significantly, hippocampal slices of adult rats show evidence of enhanced excitability if these rats had seizures in infancy: afterdischarges from the GABAA receptor blocker picrotoxin are prolonged [84]. A similar relationship between some types of infantile epilepsy and adult neocortical epilepsy has not been sought (Swann, JW, personal communication). Experimental febrile seizures in the developing rat brain also evoke epileptogenicity in the adult rat [97, 98].

Epileptic pathways

Normally, in several sensory systems of the brain, use-dependent and NMDA receptor-dependent synaptic remodelling of central nervous system connections occur in ontogenesis such as in development of ocular dominance columns [99, 100] and in the inferior colliculus [101]. In an immature brain afflicted with seizures, similar mechanisms may allow abnormal persistence of epileptogenic pathways. Thus, lithium-pilocarpine injections induce seizures at all ages of an experimental animal but brain-derived neurotrophin factor is markedly increased in immature rats and this appears to be elevated in most cells whose firing rate is increased with seizure activity [102]. As neurotrophins enhance neuronal differentiation, it is possible that seizures in the immature brain may lead to abnormal patterns of innervation which could contribute to epileptogenesis and abnormal brain development [103, 104]. However, other mechanisms may well apply.

Several additional studies clarify how neuronal activity influences the formation of synaptic connections that largely determine the properties of neural networks. Synaptic connections are developed in two stages. The first is extension of axones via chemotropic factors and non-diffusable molecules to specific synaptic targets [105, 106]. The second is remodeling of early patterns of synaptic connectivity. Seizure activity in the immature brain appears to principally influence this second stage by one or more of three ways. Such seizures may "freeze" developing networks at immature patterns of connectivity. Experimental epileptogenic lesions, by creating excessive and abnormally directed activity, may lead to a failure of "pruning" and therefore permanently abnormal connectivity [107], a situation termed "the hyperconnected cortex". Secondly, ictal discharges could enhance the growth of recurrent axone arbours that occurs in the second post-natal week of the rat, resulting in augmented profusion of local networks. To examine this, Swann's group studied morphology of CA3 apical and basal dendrites of adult rats whose hippocampi were rendered chronically hyperexcitable by intrahippocampal tetanus toxin injections while immature (post-natal day 10). Instead of profuse axonal and dendritic branching, a prominent decrease in dendritic spine density and simpler axonal branching were found [108-110]. These findings are more consistent with a third postulated mechanism of epilepsy-induced synaptic modification, similar to that found at the neuromuscular junction: elimination of some axones and their synaptic connections with enhanced development of arbours of the remaining axones [111]. This remodeling pattern, shown to depend on neuronal activity, has also been found in the developing visual system [112].

Donald Hebb [113] postulated the following concerning synaptic modification:

When an axon of cell A is near enough to excite cell B or repeatedly or consistently takes part in firing it, some growth or metabolic change takes place in one or both cells such as that A's efficiency, as one of the cells firing B, is increased.

Note that cell B must respond to the input from cell A. This postulate for associative learning suggested to Constantine-Paton et al. [112], a two part rule for use-dependent modification of synapses in the developing nervous system:

1. Synaptic contacts between synchronously active pre- and post-synaptic neurones are selectively reinforced.

2. Synaptic contacts between asynchronously active pre- and post-synaptic neurones are selectively depressed or eliminated.

This suggests that synchronous or near synchronous inputs from converging synapses will be summated, i.e. reinforced, at the post-synaptic membrane and therefore will more likely activate the post-synaptic cell than will asynchronous input whose synapses will be functionally eliminated [114]. For example, temporarily shutting one eye in kittens results in the adult visual cortex neurones receiving virtually all their afferents from the non-deprived eye, the only eye providing synchronous input to the cortex at a critical stage of development. Under this principle, synapses involved in epileptic discharge will more likely be reinforced than those receiving less synchronous and lower frequency afferents.

Recall that Hebb's postulate stipulated that cell B had to respond to input from cell A. Timing of these events in cells A and B has been found to be critical for synaptic modification. An action potential generated at the axon hillock of cell B may back-propagate to dendrites, informing the synapse whether and when the post-synaptic cell has fired. This feature affects synaptic modification in both the neocortex and the hippocampus [115, 116]. Thus long term potentiation (LTP) was induced when action potentials were induced by current injection 10-20 msec after EPSPs, whereas long term depression (LTD) occurred if the order were reversed [115, 117]. Action potentials occurring after synaptic input helps to open N-methyl-D-aspartate receptors (NMDARs), allowing high-level calcium influx and LTP while the reverse sequence allows low-level calcium influx through voltage-gated calcium channels, then through NMDARs, resulting in LTD. Additional mechanisms underlie this relationship [118]. Arrival of a second, coincident EPSP, depolarising the membrane further, also permits substantial intracellular calcium influx [110]. Transient high level intracellular calcium elevation may lead to activation of certain protein kinases and LTP while sustained low level calcium elevation may activate phosphatases and LTD [119-121].

It is thought that coincident synaptic contacts underlying LTP contribute to synaptic maintenance while non-coincidentally activated synapses or synaptic timing producing LTD are eliminated [110, 112, 122]. It is not the level but the pattern of neuronal activity that determines synaptic survival. Therefore the pattern of epileptic discharge in the immature brain, by nourishing some synapses and eliminating others, may permanently imprint an epileptogenic synaptic network on the central nervous system.

Theoretical analyses suggest that synchronous action potentials in multilayer networks may retain their precise action potential timing while propagating from one neuronal aggregate to the next [123, 124], thereby enhancing synaptic efficiency and maintenance at sequential points of propagation. This property may facilitate propagation and synchronisation of epileptic discharges, both features of the LGS.

Gap junctions

Alterations in glial and neuronal gap junctions probably effect and affect epileptogenesis. Data relevant to the Lennox-Gastaut syndrome follow.

Neuronal gap junctions contribute to neuronal action potential synchrony and thereby to the generation and maintenance of epileptic seizures [125]. Thus, in aplysia, epileptogenic agents increase electrotonic coupling, a gap junction-related phenomenon, whereas blocking gap junction channels can reverse seizures created by GABAA blockers in the guinea pig brain preparation [126, 127].

Neuronal gap junction communication (GJC) is particularly abundant in the immature brain where there is transient and extensive coupling between groups of neurones [128-130]; non-synaptic epileptiform activity is consequently more common [125]. Although GJC normally declines with maturation, epileptic activity in the developing brain may preserve this extensive neuronal coupling, thereby contributing to the epileptogenetic network described above.

In the CA1 region of the hippocampus, GJC increases spontaneous and synchronous activity whereas blocking GJC may abolish such events [131]. Moreover, encasing coupling in neuronal models both synchronises and slows neuronal activity [132, 133]. GJC may increase both high frequency and low frequency neuronal oscillatory activity [125]. Importantly, only weak coupling is required to convert "spiking" neurones into bursting cells as found by both experimental evidence and neuronal modeling [134, 135]. This conversion may promote propagation of epileptic discharges as it is the intrinsic bursting cells of layer V of the cortex which are particularly involved in intracortical epileptic propagation [136]. Moreover, coupling may modify intrinsic neuronal characteristics which would favour synchronisation of "spike" bursts [137]. Additionally, changes in mRNA translation can effect the number of gap junctions present in neurones and therefore effect long-term changes in GJC [138]. Conceivably, early seizure activity could produce this effect.

Finally, many of the foregoing modifications could promote "saltatory conduction" between adjacent groups of even transiently coupled neurones [139]. As coupled groups fire synchronously, such synchronous firing could produce larger electrical and ionic gradients as well as a larger synaptic drive, all favouring epileptic propagation [125].

As glial cells help to regulate extracellular potassium levels, their gap junctions aid in potassium movement within the brain. Conceivably, abnormalities of glial gap junctions could cause fluctuating and higher extracellular potassium levels which could contribute to epileptogenesis [140].

The foregoing suggests that modifications of gap junction communication in the immature brain may contribute to epileptogenesis in Lennox-Gastaut syndrome, through its significant role in neuronal synchronisation and propagation.

Direction of epileptic discharge propagation

By circumscribing or undercutting penicillin-induced focal cortical epileptogenic lesions in newborn and pubescent (24 months) Macaca mulatta, Caveness et al. [141] found that with maturation the direction of seizure propagation shifted from subcortical (presumably thalamic) to contiguous cortical. These data are consistent with Caveness' earlier study comparing intact Macaca mulatta: seizures spread first to homotopic areas in the newborn, but first to ipsilateral cortical regions in the pubescent monkey [142]. As the newborn corpus callosum was minimally myelinated, Caveness concluded that seizures in the newborn spread homotopically via the thalamus or the midbrain reticular formation.

These data suggest that occurrence of an epileptic system in the immature brain will preferentially create a cortical-thalamic engram, en route to bisynchronous electrographic and clinical phenomena.

Genetics, incidence and prevalence

In a substantial percentage of patients, no antecedent factor explains the development of the LGS [143]. Moreover, not all patients with some neurological antecedents, even genetically related diseases such as tuberous sclerosis, proceed to LGS. These circumstances raise the possibility of a genetic predisposition for this syndrome, sometimes superimposed upon a genetically-related disease. Unfortunately, less progress has been made in this area for LGS than for other generalised epilepsies [144, 145]. A strong genetic component has been established for myoclonic-atonic (myoclonic-astatic) epilepsy which shares some features of LGS, but remains distinct from it [144]. The incidence of a family history of epilepsy in patients with LGS has varied from 2.5 to 48% [5, 6, 146]. Single case reports of a parent with a primary generalised epilepsy may represent a chance recurrence and do not contribute. Several authors [144, 147, 148] believe that the generalised epilepsies may have a complex inheritance involving more than one gene: one gene or set of genes may predispose to idiopathic generalised epilepsy and other genes would specify the subsyndrome. This double requisite may underline the rarity of LGS: in a major referral centre such as Centre St.-Paul, only 6.6% of patients whose seizures began before 10 years of age had LGS [5]. The overall prevalence and incidence of LGS is quite low, i.e. ~ 1% of childhood epilepsies and 0.5/100,000/year among children < 10 years of age [149].

A further hint that genetic factors may participate in the development of LGS is the only occasional occurrence in patients with childhood onset epileptogenic diseases. LGS occurred in five out of ten patients with progressive encephalopathy, edema, hypsarrhythmia and optic atrophy (PEHO) syndrome [150] and in two out of ten with the 'double cortex' syndrome [151]. Patients with Angelman syndrome have slow spike-waves but minimal tonic seizures and no paroxysmal fast activity [152].

A genetic predisposition to LGS is suggested by the greater susceptibility of Long-Evans hooded rats to develop slow spike-waves in Snead's AY model than Wistar or Sprague-Dawley rats [153]. Unfortunately, genetic investigations of this phenomenon in rat have not been conclusive [153].

Pathophysiology of antecedent conditions

Do data about the pathophysiology induced by models of the more common antecedents of LGS clarify, or at least harmonise with above-described experimental findings? The following are relevant data from these principal entities.

Dysplasia

About 50% of an autopsied series of patients with LGS demonstrated diffuse dysplastic lesions [15], while clinical series list "diffuse brain malformation" among principal aetiologies [37, 143, 154]. Experimental models of such lesions might clarify the pathophysiology of their epileptogenesis and that of LGS.

Prince [155], using freeze lesions, has developed a model for human cortical microgyria, thus resembling somewhat focal cortical malformations commonly seen in association with LGS [15, 30]. A substantial circuit reorganisation occurred surrounding such experimental lesions and extending far beyond the lesion itself. A similar lesion, produced by methylazoxymethanol (MAM) exposure has been associated with an increase in burst firing [156]. Unlike normal rat cortex in which stimulus-evoked EPSPs involved only AMPA receptors, in rat dysplastic cortex such EPSPs utilised both AMPA and NMDA receptors [157]. The participation of NMDA receptors may underlie the greater evoked and spontaneous excitability seen in this and other studies [158, 159]. In a rat model of microgyria, Hablitz and Defazio [160] found evidence of continued expression of an NMDA subunit normally seen only in the early post-natal period. Along with alterations in local recurrent excitatory pathways, activation of NMDA receptors at normal or even hyperpolarised membrane potentials suggests molecular changes in receptor subunit composition in dysplastic cortex [157]. Such NMDA receptor activation at near resting membrane potentials with dysplasia and in immature cortex may account for the prominent role of NMDA receptors in creating an epileptogenic synaptic network in these circumstances [89].

Although Ferrer et al. [161, 162] found loss of two subpopulations of inhibitory neurones (parvalbumin- and calbindin-immunoreactive) with human dysplasia as did Prince [155] with freeze lesions, Luhmann et al. [157] found no loss of parvalbumin inhibitory interneurones in rat dysplasia. However, Hablitz and Defazio [160] found persistence of an immature GABA receptor in a rat microgyria model. The decreased responsiveness of this GABA receptor to a benzodiazepine receptor agonist suggests impaired inhibitory function in dysplastic cortex.

These findings would further support the concept that increased cortical excitability in infancy underlies the two characteristic EEG phenomena of the Lennox-Gastaut syndrome: slow spike-waves and paroxysmal fast activity.

Anoxia

Anoxic episodes in early life are the aetiology of 41% of patients who develop LGS from West syndrome [154]. Hypoxia in rats at post-natal days 10-12, corresponding physiologically to full term human neonates, induced seizures acutely, and lowered the seizure threshold in adulthood. These effects of hypoxia did not occur at younger or older ages [88, 163]. Both non-NMDA and NMDA mechanisms appear to be involved in these effects [164, 165]. Long term potentiation (LTP) increased in CA1 hippocampal slices in these animals [165], further evidence of NMDA receptor participation. Moreover, sodium channel excitability to acute stress chronically increases after moderately depriving immature rats of oxygen [166].

Hemorrhage

About 20% of patients progressing from West to Lennox-Gastaut syndrome had neonatal intracranial hemorrhage [154].

Experimental cortical injection of blood or iron produces iron-filled macrophages, ferruginated neurones, neuronal loss and astrogliosis [167, 168]. Unilateral cortical injection of iron induces focal, homotopic and bilateral epileptiform activity including bisynchronous spike-waves [169, 170], with propagation to the thalamus and substantia nigra [171].

Three consequences of hemorrhage-induced iron deposition in the brain may contribute to epileptogenesis. Singh and Pathak [172] studied ferric chloride-induced epilepsy in the rat and showed that acute and chronic increases in lipid peroxidation correlated with epileptiform discharges. Secondly, glial glutamate transport was impaired in this model; this correlated with increased epileptogenesis [173, 174]. Thirdly, nitric oxide synthase activity was decreased in a rat ferric chloride epileptogenic focus [175]. Inhibitors of nitric oxide synthase augment spontaneous epileptiform bursts in hippocampal slices [176]. How these components interrelate to promote hemorrhagic-induced intractable generalised epilepsy in children remains unclear.

Encephalitis

A paucity of data meaningfully relates neonatal or infantile encephalitis with LGS. Putatively, micro-infarcts and hemorrhages may occur inducing mechanisms discussed above under anoxia and hemorrhage [177, 178]. Astrocytic and microglial proliferation could impair glutamate uptake or alter gap junctions. Nitric oxide production is inhibited, diminishing its antiepileptic effect [176, 178].

Synthesis

Evidence has been presented that increased cortical excitability underlies several aspects of Lennox-Gastaut syndrome including its clinical and EEG features, as well its intractability.

Identifiable central nervous system conditions, such as dysplasia, anoxia and hemorrhage augment cortical excitation involving both NMDA and AMPA receptors in at least two of these examples (figure 1). Both clinical and experimental data have disclosed enhanced epileptogenicity in the immature nervous system. Therefore, if known and unknown excitatory factors coincide with this window of excitability, an enduring epileptogenic system will develop. Coincident EPSPs, as occurs with epileptic discharge, will generate long-term potentiation (LTP) and synaptic maintenance while non-coincidently activated synapses, producing long-term depression (LTD), would be eliminated. Propagation of epileptic discharge in the immature nervous system may more greatly involve gap junction communication. Thus, enduring synaptic and non-synaptic epileptic systems would form. In the immature primate, greater homotopic and thalamic epileptic discharge propagation than contiguous cortical spread would favour the bisynchronous epileptic discharge pathways that constitute the Lennox-Gastaut syndrome. The excess bilateral cortical excitability would abnormally augment the oscillatory system of the cortex and thalamus. Epileptogenic networks created during this heightened period can become engrammed, afflicting the nervous system with a chronic intractable epilepsy. These effects would create the defining clinical and electrographic signatures of Lennox-Gastaut syndrome.

CONCLUSION

Acknowledgements:

Dr Stan Leung reviewed the manuscript and commented helpfully. Dr C. Toth helped to produce the figure. Mrs. Maria Raffa carefully prepared the manuscript.

Received May 21, 2001 Accepted November 19, 2001