John Libbey Eurotext

Prolonged seizures: what are the mechanisms that predispose or cease to be protective? A review of animal data Volume 16, numéro spécial 1, October 2014

Tableaux

Seizures occur upon changes in the brain and its function; this is a generally accepted fact. It is also widely accepted that some, if not most, of these changes impact on seizure susceptibility, or even on the mechanisms which determine seizure duration and termination. These changes have been the subject of a number of reviews on the mechanisms of seizure generation, particularly in the ontogenetically-immature brain (Holmes and Ben-Ari, 2001; Holmes et al., 2002; Heinemann et al., 2002; Avanzini and Franceschetti, 2003; Avoli et al., 2005; Brooks-Kayal, 2005; Löscher and Köhling, 2010); these cited reviews only represent a small sample. Apart from the general problem of differentiating between the chicken and the egg (i.e. causative changes and epilepsy, or vice versa), these changes may act in homeostatic or even protective ways. This review will focus on two main issues, ontogenetic particularities and pathophysiological conditions supporting evidence of susceptibility and seizure termination mechanisms in adult animal models, while only briefly touching on protective mechanisms and status epilepticus (SE)-associated functional alterations as predisposing factors to subsequent seizures.

Seizure-induced changes

There is a plethora of functional and structural changes associated with seizures, ranging from neuronal loss (from location- and neuron type-specific, to general), sprouting and network reorganisation, to alterations of voltage-gated or ligand-gated ion channels and receptors. Arguably, for many of these, it is difficult to differentiate whether they are a consequence or primary cause of seizures (other than those that are genetically determined), as many are considered to be detrimental by increasing seizure propensity or inflicting further functional damage (Sutula et al., 2003; Löscher and Brandt, 2010). In contrast, some may actually limit seizures or act in a neuroprotective fashion (Lado and Moshé, 2008). Below, the mechanisms involved in seizure predisposition or prevention will be reviewed.

Mechanisms that prevent seizures

In adult animal brain tissue, after single or prolonged seizures, a number of changes have been reported which either specifically affect changes in seizure threshold or more broadly result in preconditioning and neuroprotective actions.

Seizure threshold changes

While prolonged seizures, in particular SE, in tissues of adult animals, generally result in subsequent development of spontaneous recurrent seizures (as in pilocarpine- or kainate-induced SE), as first described by (Turski et al., 1983) and (Ben-Ari and Lagowska, 1978; Ben-Ari et al., 1979), single and brief seizures may have opposite effects, at least transiently. Increases in seizure threshold after previous seizures, known since the 1940s (Toman et al., 1946), have systematically been analysed by the group of Nutt et al. (1981); for an overview see Löscher and Köhling (2010). To summarise these reports briefly, in different models (kindling, maximal electroshock, and others), a previous, usually single seizure, or a series of mild seizures will at least transiently increase seizure-induction threshold for subsequent seizures, lasting from minutes to several hours, whereas this is not the case for subconvulsive attacks. As also reviewed by Löscher and Köhling (2010), changes in GABA transmission may play a role, albeit resulting in both up- (receptor density) and down-regulation (presynaptic release) (Tuff et al., 1983; Löscher and Frey, 1987; Swijsen et al., 2012). In the case of early-life seizures, these effects may even be long-lasting (Swijsen et al., 2012). Furthermore, evidence from transcriptome analyses demonstrate down-regulation of calcium signalling, including subunits of voltage-gated calcium channels, and neuronal excitability components, including NMDA- and AMPA-receptor subunits (Jimenez-Mateos et al., 2008), which could also account for an increase in seizure threshold. At least in immature animals, there are also conflicting reports suggesting an increase in excitability and a reduction in seizure threshold (Gashi et al., 2007). Summarising these findings, in mature tissue, single or mild seizures appear to have short- (hours) to longer-term (days) effects, resulting in transiently reduced excitability which is likely due to changes in inhibitory, as well as excitatory, transmission and voltage-gated channels. In immature tissue, some evidence points to increased excitability, an issue which will further be discussed below.

Neuroprotection

It is highly conceivable that a reduction in excitability, as discussed above, can also lead to a protection from cell death, i.e. neuroprotection, as calcium signalling is also linked to excitotoxicity. This injury protection (coined “epileptic tolerance”) was proven in a variety of models, as reviewed by Jimenez-Mateos and Henshall (2013), including the kindling model (Kelly and McIntyre, 1994; Andre et al., 2000; Penner et al., 2001) and electroshock (Kondratyev et al., 2001) and kainic acid-induced seizures (Blondeau et al., 2000). Again, these findings are not undisputed; anti-apoptotic protection may be provided, however, it may also fail, even after single seizures, and induce damage (Bengzon et al., 1997; Andre et al., 2000). It remains to be tested whether distinct models (electroshock vs. amygdala kindling; i.e. deleterious vs. beneficial; [Andre et al., 2000]) or kindling positions (amygdala vs. hippocampal; i.e. beneficial vs. deleterious; [Bengzon et al., 1997; Andre et al., 2000 respectively]) actually account for these disparate results.

Predisposing mechanisms

In contrast to the main tenor of the previous section, the consequences of seizures may also result in increased seizure predisposition and excitability, in particular, as long-term (days and months) rather than short-term effects of initial, and severe, insults on the one hand, and under specific circumstances in juvenile tissue on the other. These findings will briefly be summarised.

During ontogenesis

Most findings in neonatal and infant tissue (in rats this age ranges from postnatal day 0 to 17 [P0-P17]) suggest that early-life seizures, or more specifically SE, have little impact on subsequent seizure susceptibility, or even neuronal injury or network changes (as reviewed by Scantlebury et al. [2007]), even though SE can be more severe than in adults. In prepubescent rodents (P18-P30), this relative protection against subsequent damage declines, and full vulnerability is reached in adults (Scantlebury et al., 2007). Under certain circumstances, however, subsequent increases in seizure susceptibility have also been reported in immature animals. Whether increased excitability is induced or not may either depend on the epilepsy model used or an underlying pathology. Regarding underlying pathologies, the conversion of relatively resistant tissue to tissue that becomes more seizure-susceptible, and more excitable, appears to result from both artificially induced migration disorders or cortical dysplasia and neonatal hypoxia, possibly due to interference with GABAergic inhibition or voltage-gated currents, such as Ih, a hyperpolarisation-activated inward current which is down-regulated after early-life hypoxia (Jensen et al., 1992; Germano and Sperber, 1997; Germano et al., 1998; Jensen et al., 1998; Scantlebury et al., 2004; Zhang et al., 2006). Regarding the epilepsy model, febrile seizures appear to be another exception: If rodents experience these early in life (P8-P11), they will develop increased hippocampal excitability, again due to increased Ih (Chen et al., 2001; Brewster et al., 2002), possibly mediated via hyperthermia-induced hyperventilation and associated alkalinisation, as well as cannabinoid receptor up-regulation (Chen et al., 2001; Brewster et al., 2002; Schuchmann et al., 2006). In addition, a reduction of GABA-mediated inhibition (Liebregts et al., 2002; Swijsen et al., 2012) may be involved. The effect is dependent on duration of the febrile condition; more severe effects are observed if the temperature rise lasts for more than an hour, compared to around 25-30 minutes. Under prolonged febrile conditions, inflammation, as evidenced by increases in interleukin-1ß, may also play a role (Dube et al., 2010). Even at later developmental stages, i.e. at P21, a febrile seizure episode increases responsiveness to epileptogenic agents, as well as induced cognitive dysfunction (Wilhelm et al., 2012).

Importantly, although SE early in life does not increase sensitivity to convulsants (Nehlig et al., 2002), other models of infantile epilepsy (at age P6-13; PTZ-induced recurrent seizures, kainate SE) do induce changes in receptor expression and function later in life. Such changes can be interpreted as increases in net excitability induced by early-in-life seizures. Thus, adult up-regulation of NMDA-receptors (Gashi et al., 2007), down-regulation of the GluR2 subunit (which would convey a reduced calcium permeability to the AMPA receptor if present in the receptor molecule) (Zhang et al., 2004), and even gender-specific effects on GABAergic signalling (GABA-reversal potential) (Galanopoulou, 2008) were observed.

During the adult state

While there appears to be some consensus that in juvenile epilepsies at least two conditions, a second underlying pathology and febrile seizures, predispose to subsequent seizures in adulthood, it is much less clear which of the manifold changes associated with chronic epilepsy are actually causal for subsequent seizures (and therefore support the disease and its progression), and which are perhaps even homeostatic reactions to the seizures.

Obviously, based on genetic models of epilepsy, we can deduce that some functional changes cause subsequent seizures, with the general cautionary note that usually the whole transcriptome of such animals is not known, and consequently it remains uncertain whether compensatory processes might play pivotal roles in addition. Such defects (and their relation to human “mutants”) have been extensively reviewed by Noebels (2003) and Lerche et al. (2013). In short, they include mutations in: voltage-gated sodium channels (prolonged opening, reduced inactivation, but also exclusive down-regulation in inter-neuronal populations), voltage-gated potassium channels (again with channel defects in principal or interneurones, or ontogenetically bound), voltage-gated calcium channel compartments (pre-synaptic down-regulation resulting in selective transmitter release changes, favouring rhythmic discharges and increases in thalamic currents), transmitter release machinery (reduction of inhibitory transmitter release, sustained release, or even alterations of co-release of e.g. Zn2+), GABAergic function (GABA synthesis, receptor down-regulation or changes in deactivation kinetics, reduction in late GABA responses, and alterations in chloride distribution leading to depolarising GABA), glutamatergic function (increased calcium permeability, impaired glutamate reuptake), cholinergic and serotonin receptors, and finally proteins controlling proliferation or migration etc. (all reviewed inNoebels [2003], Lerche et al. [2013], and Lerche et al. [2001]). Not all of the mutations generated in animals that result in epileptic phenotypes, however, are applicable to humans.

Reports on transcriptional or post-translational changes that play a role in disease progression in animal models usually focus on two conditions: febrile seizures early in life and post-SE mesial temporal lobe epilepsy (TLE) (Lerche et al., 2013), which can therefore be considered as the two main factors favouring the emergence of increased seizure susceptibility and disease progression, as reported in animal models (McCloskey and Scharfman, 2011) and TLE patients (Walker et al., 2002). Since febrile seizures have already been discussed in the previous section, in the following, the focus will be on post-SE TLE models.

In post-SE models, long-term changes in expression and post-transcriptional changes of a variety of genes have been described, among them up-regulation of juvenile forms of NMDA-receptor subunits, differential expression and splicing of AMPA-receptor subunits, down-regulation of GABA-receptors or differential GABAA-receptor subunit expression, dysregulation of chloride transporters and GABA-synthesizing enzymes (glutamate decarboxylase), down-regulation of calcium-activated potassium (SK) channels, up-regulation of Ih, down-regulation of sodium channel subunits, up-regulation of calcium channels mediating T-type currents, up-regulation of synaptotagmin subtypes, nitric oxide (NO)-synthase and nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB), and DNA methylation changes, which either influence net excitability (e.g. neuronal bursting, synaptic synchronisation) or promote cell death (Babity et al., 1997; DeLorenzo and Morris, 1999; Gilby et al., 2005; Porter et al., 2006; Becker et al., 2008; Chuang et al., 2010; Barmashenko et al., 2011; Schulz et al., 2012; Müller et al., 2013; Ryley et al., 2013). The cited papers (ibid) and other studies (Macdonald and Kapur, 1999; Shao and Dudek, 2004; Shao and Dudek, 2006; Chen et al., 2011) also generally provide evidence of functional sequelae of these genetic changes, such as: faster desensitization, depolarisation, or reduction in the frequency of GABA currents (regarding mini IPSC); reduced SK-mediated after-hyperpolarisations; and increased sodium-channel, Ih and T-type calcium channel-dependent bursting, etc. Beyond this, there is a discussion as to whether cell death (Andre et al., 2000) and stem cell proliferation (Parent et al., 1997; Parent et al., 2006) are major consequences. Moreover, these may also be epileptogenic factors at least for TLE, and further give rise to deafferentation of interneurons (dormant basket cells; a controversial subject) (Sloviter, 1991; Bernard et al., 1998), a reorganisation of interneuronal wiring (Andre et al., 2001), and, importantly, a reduction of intrinsic antiepileptic mechanisms such as control of adenosine-mediated excitability via increased adenosine degradation (Fedele et al., 2005) or receptor desensitization (Hamil et al., 2012).

A causal relationship between disease progression and these changes is epistemologically impossible, but it is tempting to speculate that such prolonged changes are somehow involved in progression of seizure severity in animal models (McCloskey and Scharfman, 2011). Even beyond the discussion of the chicken and the egg, it is, however, improbable that any single mechanism is key to chronification of the epileptic condition, especially since studies so far have identified a large number of divergent mechanisms.

Seizure termination mechanisms

When addressing the question of predisposing mechanisms to prolonged seizures, rather than looking at processes which aggravate seizures (as in the previous section B), one may also ask which factors lead to seizure termination, and consequently fail once a seizure does not stop but continues to develop into SE. Studies focusing specifically on seizure termination mechanisms, interestingly, are rare, and hence most of the hints from animal studies are indirect. Nevertheless, numerous hypotheses have been put forward, which will be discussed in this section, and an overview of which is given in table 1.

Theoretical considerations are perhaps useful in setting the scene. In one study in which the question of why some seizures stop and others evolve into SE was decidedly addressed, human EEG, electrocorticogram (ECoG), and local field potential and multi-unit recordings were actually used (Kramer et al., 2012). In this study, three distinctive features were used to characterise the phases of seizure termination: a decrease in power (and frequency), an increase in temporal correlation as well as spatial correlation, and flickering (alternating high or low variances of spectral power). The authors concluded that all seizures terminated as a result of a discontinuous critical transition from one attractor to another, i.e. ictal to post-ictal state. All data from SE, in turn, displayed repetitive periods of strong correlation and anti-correlation of spectral power and temporal correlation measures; in other words, with SE, the system approaches but fails to cross the boundary condition repetitively (Kramer et al., 2012). This suggests that seizure termination is a sudden, and likely time-dependent, and perhaps even deterministic, rather than gradual process, as already corroborated by other modelling studies using animal absence model data (Suffczynski et al., 2006). The studies also suggest that it is unlikely that a single mechanism, such as intracellular calcium accumulation, is responsible, as speculated in another modelling study (Kudela et al., 2003). Rather, the theoretical considerations suggest that different mechanisms can converge into the same transition border state (Kramer et al., 2012).

One tempting hypothesis to explain seizure arrest is a proposed depletion of resources, under the presumption that e.g. oxygen and/or glucose supply would drop under continued seizure activity, and hence also intracellular ATP (Doman and Pelligra, 2004). Indeed, in both chronically epileptic human tissue, as well as tissue from post-SE rats, induced seizure-like activity results in reduced NAD(P)H recovery, suggesting mitochondrial respiratory chain or glycolysis failure (Kann et al., 2005). While this explains ictal hypometabolism, it is unlikely, however, that it actually is instrumental in stopping seizures, precisely because, in these models, seizure-like activity actually progresses. Indeed, the authors speculate that the reduction of NADH production could be instrumental in developing pharmacoresistance (Heinemann et al., 2002). Furthermore, although hypoxia (and particularly re-oxygenation after hypoxia) can induce seizures, and likewise severe hypoglycaemic conditions (while moderate reductions are actually pro-convulsant) (Kirchner et al., 2006), local cerebral glucose utilisation is generally reduced immediately postictally in kindled rats, which suggests a lowered rather than increased glucose demand at the end of a seizure (Namba et al., 1989).

Loss of neuronal synchronisation via loss of excitatory drive or increasing impact of inhibitory mechanisms or differential function of electrical coupling is another attractive hypothesis to explain seizure termination. At first sight, experiments using an in vitro model of status-like activity (high K+), suggesting a progressive exhaustion of presynaptic glutamate release during epileptic bursts, favour this hypothesis (Staley et al., 1998). However, again, in this model, inter-burst intervals successively increase due to this mechanism, but the activity, as such, persists, and burst duration actually increases with reduced and desynchronized glutamate release (Jones et al., 2007). The role of glutamatergic failure in seizure termination is thus uncertain. Increasing inhibitory restraint might also play a role in seizure arrest.

The concept of surround inhibition playing a role in spatial containment of seizures was already put forward by David Prince and Joe Wilder in the 1960s (Prince and Wilder, 1967). Direct recordings from interneurones from such foci demonstrate inhibitory cells to be very active during discharges (Domann et al., 1991). Irrespective of this initial restraint (Trevelyan et al., 2006; Trevelyan et al., 2007; Schevon et al., 2012), however, there are strong indications that, with the progression of seizures, or more precisely with the spatial spread of activity, the inhibitory restraint fails, as reviewed by Trevelyan and Schevon (2013).

There may be several reasons for this, including a depolarising block, presynaptic inhibition of GABA release, GABAergic vesicular depletion or postsynaptic desensitisation, and in fact also development of a depolarising drive for GABA as chloride accumulates (Dzhala et al., 2010); reviewed by Trevelyan and Schevon (2013). Hence, at least in acute models of epilepsy, the role of inhibition is probably a restraining one at first, but then changes to one which likely merely shapes the structure of an ictal event (i.e. support of tonic-like phases vs. clonic-like ones) without, however, playing a role in determining their duration (Swartzwelder et al., 1988). Indeed, GABA, by being rhythmically released during seizure-like activity, may actually subserve network synchronisation by pacing bursts, both in acute in vivo animal models (Khazipov and Holmes, 2003), as well as in human epileptic tissue (Köhling et al., 1998). Beyond this, in chronically epileptic tissue, the aforementioned chloride accumulation may chronify as well, as GABAergic activity, at least in the hippocampus in TLE models and in human tissue, is often depolarising, likely due to a dysregulation in chloride transporter expression (Köhling et al., 2000; Cohen et al., 2002; Pathak et al., 2007; Bragin et al., 2009; Barmashenko et al., 2011). In the worst case, GABAergic activity can finally also contribute to the expansion and chronification of the epileptic condition itself, since GABAergic fast oscillations are apparently able to establish mirror foci under certain conditions (Khalilov et al., 2003; Le Van Quyen et al., 2006). Phasic GABAA transmission is hence not a realistic candidate mechanism limiting seizure duration.

The case is less clear for tonic GABAergic activity, which is stable or even increased in chronically epileptic tissue (Walker and Kullmann, 2013). Whether this up-regulation of tonic current is mediated via neurosteroids, as in non-epileptic tissue (Stell et al., 2003), remains to be elucidated, but can be disputed on the grounds that neurosteroid-sensitivity in epileptic animals in vivo is reduced (Lawrence et al., 2010). At any rate, a reduction in the inhibitory action of phasic GABA release, combined with an up-regulation of a tonic action, is actually likely to increase the gain of neurons, and hence their responsiveness to excitatory input (Walker and Kullmann, 2013).

With respect to neuronal and/or glial gap junctions, the studies so far favour the notion that inter-neuron coupling supports synchronisation, while inter-glia coupling reduces it. The evidence comes from experiments using pharmacological or genetic blockade, which either leads to increased activity in the case of glial gap junction knockout (due to loss of spatial potassium buffering and increased extracellular potassium) (Wallraff et al., 2006) or to a decrease in activity when gap junctions are pharmacologically blocked (putatively then also, or even mainly, neuronal ones) (de Curtis et al., 1998; Köhling et al., 2001; Gigout et al., 2006; Roopun et al., 2010). It would be critical to demonstrate an activity-dependent loss of neuronal gap junctional coupling or increase of glial coupling to support gap-junction involvement in shaping seizure duration. Although changes in the level of intracellular pH and calcium are activity-dependent (de Curtis et al., 1998) and could well induce functional changes in gap junctions, a direct demonstration of these effects remains to be confirmed. Hence, the impact of gap-junctional coupling in seizure termination remains uncertain. Regarding ephaptic interactions, i.e. transmembranous currents induced by extracellular currents due to resistance changes in extracellular vs. membrane compartments, these could also possibly influence synchronisation (Köhling et al., 2000). However, as there is cell swelling during seizures (Lux et al., 1986), it is likely to increase, rather than to decrease at the end of seizures, and hence can probably be ruled out as a termination mechanism.

A possible seizure termination mechanism could also be a dynamic change in intrinsic neuronal excitability or transmembranous currents. The main class of transmembranous currents which reduce neuronal excitability consists of various potassium currents. Of the multitude of these, those currents which show activity-dependent activation would be particularly interesting in this context. Indeed, at least in an in vivo model of spike-and-wave discharges, in particular, calcium-activated potassium currents appear to limit activity duration (Timofeev et al., 2004). As calcium is known to accumulate intracellularly during seizures, this mechanism could be a plausible one. Interestingly, in chronically epileptic tissue (post-SE TLE), this current appears to be critically reduced, both regarding function and expression. This predisposes the tissue to prolonged discharges (Schulz et al., 2012). Input-resistance changes have been implicated in determining epileptogenicity, as they decrease in animal models and human epileptic tissue (Isokawa, 1996; Stegen et al., 2009). In effect, this means that synaptic currents need to be much larger to change neuronal membrane potential, and hence, excitatory drive will be less efficient. In turn, however, this also means that firing frequency will be increased as a consequence of a decrease in time constant (remember that τ=RC) (Bikson et al., 2003a). More importantly, input resistance often dynamically decreases rather than increases in the course of a seizure, since further channels open and the cells become leaky, making this too a possible mechanism to stop the seizure (Timofeev et al., 2004).

Among the different dynamic changes during seizures and epileptiform activity, alterations in ionic microenvironment, including pH, are well documented. A comprehensive review by Hans-Dieter Lux, Uwe Heinemann and Irmgard Dietzel summarises these phenomena (Lux et al., 1986). During activity, in the focus, extracellular potassium rises to a ceiling of 12-14 mM; further increase is prevented by glia buffering. This trans-glial potassium flux is compensated, albeit not fully, by sodium, resulting in focal extracellular sodium increases, but a decrease in osmolarity, and hence cell swelling. Calcium is reduced focally to 0.6 mM due to influx into neurons, while chloride follows the potassium buffering flux and is hence reduced in the focus. What does this mean for excitability? In particular, the increase in potassium will lead to a depolarisation of all cells within the focal area by an estimated 15-20 mV, which could actually lead to a depolarisation block (inactivation of sodium currents), and is speculated to support activity termination at least in acute in vitro models (Lux et al., 1986; Bikson et al., 2003a; Pinto et al., 2005). Although in chronically epileptic tissue, such potassium increases are generally lower than in healthy tissue (Köhling et al., 1995) and potassium levels are mainly lower at the end of a seizure-like event than at the start (Avoli et al., 1996a), it is likely that these changes do exert some influence on seizure duration, particularly should buffering be compromised. There are indeed indications in this direction. In human tissue from epilepsy patients, induced changes in extracellular potassium are not modified by Ba2+, a blocker of inwardly-rectifying astrocytic potassium-currents (KIR), and the expression of KIR-channels was found to be down-regulated (hence at the first prerequisite for potassium buffering, astrocytic potassium uptake appears to be compromised). Furthermore, in addition, astocytic connexins in rodent brain (the second prerequisite for potassium buffering), challenged with blood-brain-barrier breakdown, are equally down-regulated (Heinemann et al., 2012). A critical experiment would now be to show that the duration of seizures actually is inversely correlated to the extent of potassium increases.

In contrast, the decrease in extracellular calcium plays an ambiguous role; it increases excitability by reducing surface charge, leading to neuronal depolarisation (Bikson et al., 2003b) (something which is taken advantage of in the low-calcium epilepsy model), but also decreases epileptogenicity by reducing synaptic transmitter release (Cohen and Fields, 2004). A drop in extracellular chloride, in turn, is likely to increase excitability, since this will shift its equilibrium potential to positive, depolarising values. Having identified potassium as one possible factor of seizure termination, pH is another critical one. Thus, during seizure-like activity, the extracellular space acidifies (Somjen, 1984). Furthermore, activity duration and extracellular (Lux et al., 1986), as well as intracellular (Xiong et al., 2000), acidification are related, and artificial acidification (also via CO2-ventilation) stops activity both in vitro and in vivo (Caspers and Speckmann, 1972; Velisek et al., 1994), likely via several mechanisms, including activation of acid-sensing channels or interference with glutamatergic synaptic transmission (Velisek, 1998; Ziemann et al., 2008). Any condition supporting alkalinisation, such as hyperventilation, in turn, will prolong and exacerbate seizures (Schuchmann et al., 2006). Activity-dependent acidification is hence, together with extracellular potassium accumulation, probably a critical candidate mechanism controlling seizure duration.

Activity-dependent release of neuromodulators, i.e. substances released as non-classic transmitters with metabotropic action, which are capable of influencing synaptic transmission, is another candidate group of seizure-terminating mechanisms. Among these, adenosine and neuropeptide Y (NPY) are most interesting. Regarding adenosine, it was shown early on that it is released endogenously in an activity-dependent fashion during seizures in animal models (Lewin and Bleck, 1981) and patients (During and Spencer, 1992). Furthermore, prolonged seizures were speculated to result from loss of adenosine function (Young and Dragunow, 1994), which is generally accepted to be net inhibitory (Dunwiddie and Masino, 2001). It is not surprising that novel therapeutic strategies are being considered on the basis of adenosine delivery (Boison, 2005). Regarding NPY, this molecule is also released during and particularly after seizures (Marksteiner et al., 1989), to inhibit excitatory synaptic transmission (Tu et al., 2005). In chronic epilepsy models, its release and receptor expression are up-regulated, which is interpreted as an intrinsic antiepileptic compensatory reaction (reviewed in Vezzani et al. [1999]).

Other potential modulators also include cytokines, endogenous opioids, and cannabinoids, however, the role of these requires further elucidation, suffice to say that cytokines are usually considered to be pro-epileptic. Yet, in chronic models, an endogenous receptor blocker of interleukin 1 receptors was also reported to be up-regulated (Vezzani et al., 2002), thus one might speculate that this may also be activity-dependent. Supporting evidence for this hypothesis, however, is still lacking. Likewise, endogenous dynorphin was shown to control seizures as long as releasing fibres were conserved (Wasterlain et al., 2002), presumably via κ receptors (Loacker et al., 2007), a finding which is corroborated also in human tissue (Koepp et al., 1998; Hammers et al., 2007). The ambiguous nature of opioids, however, becomes evident when considering that these negatively control GABA release (Avoli et al., 1996b) via μ receptors, and that activation of this receptor eventually leads to induction of seizures (via brain-derived neurotrophic factor [BDNF] expression) (Zhang and Ko, 2009). Endogenously-released cannabinoids, likewise, are ambiguous in their action, being both anti- and pro-epileptic (Wada et al., 1973; Wada et al., 1975; Karler et al., 1986). They share with opioids a negative control of GABA release, and beyond this they enhance glutamate release, since they cross-react also with TRPV1 channels (Isokawa and Alger, 2005; Bhaskaran and Smith, 2010). Again, more investigations are needed to gauge their net role in seizure termination.

A factor clearly supporting sustained SE apparently is Substance P (Wasterlain et al., 2000), which was shown to be released during seizures, and its receptors were up-regulated in the chronic epileptic condition. The unfortunate combination of release and receptor up-regulation may allow for a vicious cycle to be initiated, which is speculated to maintain SE.

Outlook

In an attempt to simplify the interpretation of the findings discussed in this review, which probably reflect a broad consensus, the main factors derived from animal experiments which predispose to prolonged seizures are febrile seizures early in life (and possibly also later), as well as migration disorders/cortical dysplasia, or neonatal hypoxia. This emphasises the point that ontogenetic factors are paramount. The only other factor which apparently also plays an important role in determining seizure susceptibility, and perhaps also prolonged seizures, is a history of SE itself, and perhaps traumatic brain injury (Holmes et al., 2002), again, mainly in an ontogenetic context. Of the underlying mechanisms determining seizure duration, extracellular increases in potassium, acidification, adenosine, NPY, and substance P are interesting candidates, and future key experiments should establish whether defects in these mechanisms can be found both in animal models and the clinical context which specifically relates to prolonged seizures.

Acknowledgements and disclosures

Supported by: BMBF grant ROSAge 31P66662, BfArM grant V-14415/68502/2011-2014.

The author has no conflicts of interests to declare.