Abstract

The cellular and network mechanisms of seizure emergence (ictogenesis) remain unknown. Recent studies brought new fundamental insights about the complex nature of seizure genesis and the dynamical principles that govern the shift of epileptic networks dynamics into the seizure state. In this chapter, we describe new perspectives on how to approach understanding ictogenesis. Primarily, we focus on the latest data about the dynamical pathways and pathophysiological processes that drive the transition to seizure. In addition, we review recent evidence about the existence of long-term fluctuations in seizure likelihood which represent a missing piece in the mosaic of ictogenesis. We demonstrate that combining the cellular and network mechanisms at multiple temporal scales (ranging from epileptogenesis to seizure initiation) is a crucial step to a unified theory of seizure genesis.

Introduction

A seizure is an abrupt change of brain state where a population of neurons fires excessively and rhythmically, and a much larger population of neurons is disrupted as a consequence of synaptic projections and nonsynaptic effects. Focal seizures are initiated in a confined area of the brain and may propagate from the focus, potentially developing into a bilateral tonic-clonic seizure (Scheffer et al., 2017). Acquired epilepsies are induced by traumatic, infectious, metabolic, or undiscovered initial precipitating events. Alterations at the molecular level impact the structure and function of neuronal networks in ways that progressively ready the network to become capable of generating spontaneous seizures, a process known as epileptogenesis. Once the epilepsy is established, seizures are separated by often prolonged interictal periods during which the abnormal epileptic network dynamics manifest by various forms of pathological activities ranging from interictal epileptiform discharges (de Curtis and Avanzini, 2001), pathological high-frequency oscillations (Jiruska et al., 2017; Zijlmans et al., 2012), to rhythmic epileptic bursts (Chang et al., 2018; Litt et al., 2001) or micro-seizures (Stead et al., 2010).

Seizure genesis (ictogenesis) can be entirely independent of epileptogenic processes, as is demonstrated in experiments using acute brain preparations with convulsive stimulants, where spontaneous seizures are present in the tissue but permanent structural modifications are unlikely to happen over the short time scales of these experiments. The recent observations on the long-term dynamics of epileptic networks in humans and in chronic animal models of epilepsy emphasize the importance of approaching the ictogenesis from a much more complex perspective. The emergence of seizures is not a simply repeated switching between interictal and seizure states. To understand the real nature of seizure genesis (especially in chronic epileptogenic tissue), we need to take into account various processes that are operating at multiple temporal and spatial scales. The interaction between them determines the probability/likelihood of seizure occurrence. From a simplified perspective and based on current knowledge, the ictogenic processes can be subdivided into stages (Fig. 9–1) based on their potential roles in, and temporal relationship to, seizures (Richardson and Jefferys, 2011).

Figure 9–1.

Processes that determine the nature of ictogenesis and the probability of seizure occurrence.

1.

Epileptogenesis is a slow and long-lasting process that results in the modification of neural networks’ structure and function, leading to the occurrence of spontaneous seizures (Dudek and Staley, 2011b). This process, however, continues beyond the first seizure and can contribute to epilepsy progression. The progressive nature may continuously change the landscape from which the seizures emerge (Dudek and Staley, 2011a; Williams et al., 2007; Crisp et al., 2020). Epileptogenesis processes range in duration from months to years.

2.

Fluctuations in seizure probability occur over the medium to long term: expressed over days, weeks, months, or even years and in some cases with a rhythmic nature (circadian, multidien, annual rhythms). The proictal state is the phase associated with an increased seizure probability. The proictal state can manifest also by seizures accumulating in time as seizure clusters. The mechanisms responsible for seizure emergence can differ during periods of low and high seizure probability (Maturana et al., 2020).

3.

Transition to seizure can be defined as a process of inevitable progression to seizures. The preictal period is the time before seizure during which the transition occurs. It can last from seconds to hours. Unlike the proictal period, during which multiple seizures can occur, the preictal period leads to one concrete seizure.

4.

Seizure initiation or onset is the moment when the seizure starts. The duration of processes that trigger the seizure ranges from milliseconds to seconds.

This chapter will discuss pro-ictogenic alterations arising from the proictal and preictal states that potentially take place in epileptic brains. We will discuss the underlying cellular and network mechanisms involved in various stages of seizure genesis and how dynamical approaches to epilepsy can advance our understanding of the complex nature of seizure genesis. Studying seizure emergence from the perspective of dynamic theories is currently trending. Epilepsy research oriented on the elucidation of the mechanisms of ictogenesis substantially benefits from past studies that presented seminal observations on the dynamical principles of seizure genesis, some of which aimed to predict seizures. Several of these studies made theoretical predictions about the possible existence of proictal and preictal states which were later confirmed experimentally in animal models and humans. Recent breakthroughs proposed that critical slowing, a universal feature preceding critical transitions throughout natural phenomena, can be observed ahead of seizures and marks the changes in epileptic neuronal resilience in animal models in vitro and in vivo, in computer models, and in people with epilepsy (Chang et al., 2018; Maturana et al., 2020).

Seizure Initiation

Transition to the Seizure and Preictal State

In its seminal work on epilepsy dynamics, the group of Lopes da Silva determined the three dynamic scenarios of how the brain can enter the seizure regime (Lopes da Silva et al., 2003a; Lopes da Silva et al., 2003b). In the first scenario, neural networks operate in a highly stable dynamic regime characterized by strong resilience to perturbations. To shift the dynamics to the seizure regime, therefore, requires an extreme perturbation. Experimentally, this dynamical scenario can be applied to seizures induced by maximal electroshock when a strong electrical stimulus must be used to induce a seizure. The second scenario assumes that the brain is resting in a highly unstable regime when even weak perturbation can shift the brain into a seizure. This dynamical principle probably operates in genetically determined generalized epilepsies or reflex epilepsies. Because the internal or external perturbations have a stochastic nature, spontaneous seizures occur randomly without any obvious preceding slow transition to seizure or warning. The third dynamical pathway is characterized by gradual changes in the brain’s stability, which slowly and inevitably shifts the brain dynamics toward the seizure initiation (Lopes da Silva et al., 2003a; Suffczynski et al., 2006; Scheffer, 2009; Scheffer et al., 2012). This pathway is characterized by a progressively decreasing stability of the brain and decreasing resilience to internal or external perturbations so that even a weak perturbation can flip the dynamics to the contrasting regime. If the third pathway of slow transition to seizure could be detected, in theory, seizure probability could be determined.

Seizure initiation is a dramatic and obvious event when the brain enters a completely different dynamical regime. Therefore, the seizure onset can be studied very well, and a large number of mechanisms responsible for seizure initiation have been identified. In contrast, the process of slow transition to seizure can display only modest changes in neuronal or network behavior that may be hard to detect. The existence of the preictal state and slow transition to seizure was extensively studied in the past. The central aim of these studies was to identify detectable changes in brain dynamics that would be highly informative of approaching seizure, so that the imminent seizure could be predicted or prevented. The motivation to search for, and to detect, the slow transition to seizure was supported by the existence of prodromal symptoms that a minority of people with epilepsy experienced (Schulze-Bonhage et al., 2006). The existence of a preictal state (although of short duration) was described in studies that examined changes in cerebral hemodynamics, regional oxygen level, BOLD in functional magnetic resonance imaging (fMRI), and heart rates. The majority of studies, however, oriented to the analysis of intracranial or scalp electroencephalogram (EEG), where preictal changes in the rate of focal spikes (Lange et al., 1983) or changes in linear or nonlinear parameters such as Lyapunov exponent, correlation density, Kolmogorov entropy, and other measures of complexity and synchronization preceded the seizure onset (Lehnertz et al., 2007; Mormann et al., 2007; Litt and Lehnertz, 2002; Kuhlmann et al., 2018).

These observations were exciting, but they lacked interpretation from the neuronal and network perspectives. Experimental studies which explored the existence of slow transition to seizure emerged later and were inspired by the studies that explored slow transition between dynamical regimes in various systems ranging from climate, ecosystem, human brain, or financial market (Scheffer, 2009).

The Phenomenon of Critical Slowing and Loss of Resilience

In dynamic systems theory, the state of the system is described by state variables (e.g., average firing rate of neurons in the brain; x-axis in Fig. 9–2). A displacement of the system (change of the state variables) by an external force (e.g., electrical stimulation of the brain) is called a perturbation. The behavior of the system depends on parameters (e.g., excitability of the brain; y-axis in Fig. 9–2). The state variables spontaneously tend to approach some of the stable fixed points (also known as stable equilibria; yellow solid line at the bottoms of the valleys in Fig. 9–2). The stable fixed point acts as an attractor and its neighborhood is a so-called basin of attraction (the darker areas, i.e., valleys, in Fig. 9–2). The basin of attraction in which the system resides defines the regime in which it operates (e.g., interictal regime or seizure). The basins of attraction are separated by unstable fixed points (dashed yellow line at the ridge in Fig. 9–2). If the parameters change smoothly, at a certain point a stable fixed point in whose neighborhood the system resides may disappear. The system then undergoes so-called critical transition or bifurcation to another regime, and the state variables change their values abruptly (Scheffer, 2009; Scheffer and Carpenter, 2003; Scheffer et al., 2012; Lopes da Silva et al., 2003a; Lopes et al., 2017; Strogatz, 2000). The parameter values at which the critical transition occurs are called the tipping point or bifurcation point (red points in Fig. 9–2). Predicting when the system will undergo the critical transition is extremely difficult. However, early warning signals of approaching critical transition have been discovered (Dakos et al., 2008; Scheffer et al., 2009; van de Leemput et al., 2014).

Figure 9–2.

Dynamical model of the transition to the seizure. A. A system such as the brain can be conceptualized as a ball in a so-called potential landscape (here, the term potential has no connection to the electrical potential or cell membrane potential). The (more...)

Some of these early warning signals are a slowing of a recovery from a perturbation, increasing lag-1 autocorrelation, and increasing variance. Collectively, these markers are referred to as “critical slowing,” and they are intuitively easy to understand. If the system is on the stable fixed point, a small (i.e., within the basin of attraction) perturbation will cause a recovery reaction; that is, the system will return to the fixed point. If the system is far from the tipping point, the recovery is fast; that is, the system is very stable. In order to push the system to the contrasting basin of attraction, a large perturbation would be needed; that is, the system is highly resilient to perturbations (Fig. 9–2Ba). When the conditions of the system approach the tipping point (e.g., excitability increases; Fig. 9–2Bb), the rate of recovery decreases; that is, the system is less stable and a smaller perturbation is sufficient to push the system to the contrasting basin of attraction (i.e., the system is less resilient; Fig. 9–2Bc). At the tipping point, the rate of recovery reaches zero; that is, the system is at the margin of stability and an infinitesimally small perturbation is sufficient to push the system to the contrasting basin of attraction (i.e., the system lost the resilience; Fig. 9–2Bc). If we further continue the change of the parameter, this fixed point no longer exists (Fig. 9–2Bd), and the system inevitably transitions to the contrasting basin of attraction.

To track the recovery rate, one usually does not have to perturb the system artificially because most systems are under natural stochastic forcing. The recovery dynamics then act as a filter on this stochastic forcing. Intuitively, after each stochastic perturbation, the system returns to the fixed point at its recovery rate. If the recovery rate is slower, the subsequent measured value of the system’s state variable will be more similar to the previous one. In other words, the lag-1 autocorrelation function (measure of similarity of two consecutive signal points) of the state variable will get higher. Thus, the increase in lag-1 autocorrelation is one of the symptoms of the critical slowing down. The slower recovery rate will also cause each perturbation to have longer-lasting consequences which typically results in a higher variance of the state variable. Another marker of approaching critical transition can be the increasing skewness of the distribution of the state variable. When the boundary between the two basins of attraction approaches the current basin of attraction from one side (from the right in Fig. 9–2Bb), the basin of attraction gets shallower on that side and the system tends to recover more slowly from perturbations in that direction. Thus, the system spends relatively more time closer to the boundary, which leads to a skewed distribution of the state variable.

When the basins of attraction are shallow, strong enough perturbations can shift the system back and forth between the two contrasting basins of attraction. This phenomenon is known as flickering and is also a marker of approaching critical transition (Dakos et al., 2013; Scheffer et al., 2009). Statistically, flickering manifests by increased variance, skewness, and bimodality of the state variable distribution.

In more complex dynamic models, the attractors do not have to be only stable fixed points but can be represented also by limit cycles or chaotic attractors. Critical transitions in such systems can also be preceded by critical slowing, and the recovery from perturbation can be characterized by damped oscillations of the state variable. Also, important early warning signals of impending critical transition can be represented by specific changes in the spatial pattern (Scheffer et al., 2009). Namely, increased spatial correlation of neuronal activity can be used (Kramer et al., 2012).

Experimental and Empirical Evidence for a Critical Slowing and Loss of Stability in the Epileptic Brain

The presence of critical slowing down in human and experimental epilepsy is a matter of ongoing debate. The early warning signals of critical transition were examined in low-calcium and high-potassium models of acute seizures in brain slices (Chang et al., 2018; Jiruska et al., 2010). In both preparations, seizures were generated in the CA1 region, and the period between seizures was characterized by the presence of high-frequency activity (Jefferys and Haas, 1982; Konnerth et al., 1986; Traynelis and Dingledine, 1988; Jensen and Yaari, 1988; Chang et al., 2018; Jiruska et al., 2010). The early warning signals were determined from the properties of HFA. As a seizure approached, the autocorrelation, signal variance, and power increased, and the mean frequency of HFA decreased ahead of seizure (Fig. 9–3 and Table 9–1). The spatial expansion was assessed using various measures of synchronization: cross-correlation (Chang et al., 2018) and phase synchronization (Li et al., 2007). In both models, the synchronization progressively increased, suggesting that the population of neurons generating HFA spatially expanded (Table 9–1). At the cellular level, these preictal changes in early warning signals were accompanied by a net increase in neuronal firing. Analysis of individual principal cells and interneuronal action potential firing displayed heterogeneous behavior. While certain neurons increased firing, other neurons did not change their firing, and other neurons decreased firing rate. The presence of markers of critical slowing down ahead of seizures in both preparations suggests that the CA1 region became progressively less stable and less resilient to perturbations (Scheffer et al., 2009; Scheffer et al., 2012). To confirm these observations, that is, to determine the networks’ resilience (inverse of sensitivity to perturbation) and to evaluate delayed recovery from perturbations in advance of the seizure, CA1 networks were actively perturbed (Suffczynski et al., 2008; Kalitzin et al., 2005; Freestone et al., 2011). The analysis of the recovery from antidromic stimulation in the low-calcium model and from orthodromic stimulation of Schaffer’s collaterals in the high-potassium model revealed that as a seizure approached, the amplitude and duration of the responses increased, suggesting that it took much longer for the CA1 network to settle down to the stable fixed point after the perturbation; that is, the speed of recovery from the perturbation was becoming slower and slower with approaching seizure (Jiruska et al., 2013b; Suffczynski et al., 2008). To examine the decreased resilience to perturbations, external depolarizing electric fields (Bikson et al., 2004) of various intensities were applied to CA1 during the early and later stages of the interictal period (Jiruska et al., 2010). In a high potassium model, orthodromic stimuli of various intensities were applied in a similar manner (Chang et al., 2018). The results demonstrated that shortly before a predicted seizure, electric fields or orthodromic stimuli were sufficient to trigger seizures. The above-described changes were paralleled by a negative shift in DC fields and an increase in the concentration of extracellular potassium (Chang et al., 2018). Similar changes in network dynamics were also observed in other in vitro models of seizures (Avoli et al., 2013; Hamidi et al., 2014; Khosravani et al., 2005; Uva et al., 2017). Multichannel recording in CA1 slices demonstrated that the loss of resilience affects the entire CA1 (Chang et al., 2018; Jiruska et al., 2010). In low calcium, the seizure onset analysis revealed that seizure could initiate in any part of the CA1 region (Jiruska et al., 2010). When the entire network had low resilience, a small random perturbation occurring within a small area of the CA1 network could trigger seizure, which then rapidly spread to the rest of the network. In this scenario, one cannot predict the exact site within CA1 from where the seizure will emerge (Chang et al., 2018).

Figure 9–3.

Transition to seizures in vitro. A. Recurrent seizures recorded in the pyramidal layer of a rat hippocampal CA1 slice (removed CA3) perfused by high potassium artificial cerebrospinal fluid ([K⁺]_o >8 mM). B. A detail of the recording between (more...)

Table 9–1

A List of Early Warning Signals of Critical Transition and Experimental Evidence of Their Presence during In Vitro Ictogenesis in Low-Calcium and High-Potassium Models of Acute Seizures.

The exact cellular mechanisms responsible for the progressive loss of resilience are currently unknown. The transition to seizure may share similar mechanisms to seizure initiation, but these processes operate on a much slower time scale. The slow transition is probably a result of a cascade of discrete cellular events, which reciprocally act via positive feedback and contribute to the progressive loss of stability of the CA1 neuronal network. One of the most commonly cited candidate mechanisms is a vicious circle of intense neuronal activity, increased neuronal depolarization, an increase in extracellular potassium, leading to an alteration in the KCC2 co-transporter, which subsequently shifts the direction of the ion transport (Blauwblomme et al., 2014; de Curtis and Gnatkovsky, 2009; Jensen and Yaari, 1997). The consequent increased intracellular load of chloride and increase in extracellular potassium can be further aggravated by active inhibition. As the resilience and stability decline, other pathological phenomena such as ectopic action potential generation, conversion of neurons to burst firing, and increased nonsynaptic interactions can come into action and contribute to the vicious circle. In slice preparations, a pharmacological block of processes such as extracellular potassium transients (Bikson et al., 2002), altered chloride shift, the depolarizing effect of GABA (Perez Velazquez, 2003), or changing the size of extracellular space (Fox et al., 2007) suppressed in vitro ictogenesis.

Empirical Evidence for a Preictal Critical Slowing and Loss of Stability in Humans

The transition to seizure via critical slowing in humans was examined in several studies by analyzing intracranial data in patients who underwent presurgical evaluation and in patients implanted with chronic ambulatory seizure prediction devices (Chang et al., 2018; Maturana et al., 2020; Milanowski and Suffczynski, 2016; Wilkat et al., 2019; Cook et al., 2013). The temporal changes of early warning signals (lag-1 autocorrelation and variance) were examined in a cohort of patients with seizure prediction devices, and a total of 1,890 3-hour preictal periods were analyzed (Chang et al., 2018). One-third of patients presented an increase in lag-1 autocorrelation 30 minutes before the seizure, and in another third, seizures were preceded by a decrease. The variance was increasing only in two out of twelve patients.

In the following study in the same cohort of patients, Maturana et al. analyzed in greater detail the transition to seizure and took into account long-term seizure distribution and fluctuation in seizure probability (Maturana et al., 2020). They observed markers of critical transitions and critical slowing down in most patients. Interestingly, the markers not always preceded the clinically marked seizure onset. In agreement with the highly complex nature of epileptic seizures, several types of critical transitions could even occur within a single seizure (e.g., transition from tonic to clonic phase or similar). When only lead seizures (the first seizure of a seizure cluster) were taken into account, a gradual increase of autocorrelation function width or signal variance was found in 10 of 14 patients up to 30 seconds prior to the seizure.

The results of Maturana et al. (2020) are in contradiction with the analysis of intracranial EEG data which found consistent changes in autocorrelation, variance, or skewness only in 3 of 26 patients (Milanowski and Suffczynski, 2016). Interestingly, when only complex partial seizures were analyzed, a consistent decrease of autocorrelation, variance, and skewness was observed, which is the opposite of what is expected in the critical slowing down. Meanwhile, seizures of unclassified type (i.e., other than complex partial, simple partial, and secondarily generalized) presented a consistent increase of variance but no change in the other markers.

In a recent study, Wilkat et al. analyzed presurgical iEEG recorded in a total of 1,647 brain sites of 28 patients and found statistically significant changes of autocorrelation or variance in only one-seventh of the sites (Wilkat et al., 2019). Moreover, decreases in the markers were found in the majority of cases which contradicts the critical slowing. The authors concluded that they found indications of critical slowing down only in two to three patients in about 1%–2% of their sampled brain sites.

In a study by Litt et al. (2001), the authors analyzed intracranial recordings from presurgical evaluation of pharmacoresistant patients. The authors found occasional bursts of high-energy activity lasting several minutes. The frequency of occurrence of these bursts increased with the approaching seizure and could be detected up to 7 hours before the seizure onset. However, another observation from this study is significant for increases in signal variance up to 50 minutes in advance of the seizure onset. Interestingly, this increase was independent of the high-energy bursts and can be viewed as a marker of the critical slowing. Unfortunately, in their study, the authors did not analyze the lag-1 autocorrelation.

The discrepancy between these studies could be attributed to a different nature of the recordings (long-term ambulatory monitoring vs. presurgical evaluation likely with tapered drug dosing) or by the selection of lead seizures in case of seizure clusters. Another reason could be different statistical approaches. Chang et al. and Maturana et al. first averaged the autocorrelation or variance curves over the preictal periods and then fitted a straight line to the resulting average curve and determined the slope and significance of the fit (Chang et al., 2018; Maturana et al., 2020). In contrast, Milanowski and Suffczynski fitted the straight lines to each preictal period and determined the statistical difference of the slopes from zero by a binomial test which may give more conservative results (Milanowski and Suffczynski, 2016). Wilkat et al. used the area under the ROC curve as a metric of separability of the distributions of autocorrelation during interictal and preictal periods and surrogate data to determine the statistical significance (Wilkat et al., 2019).

Seizure genesis is a complex phenomenon, and processes operating on longer temporal scales may influence the dynamical pathway of the transition to seizure and its initiation (Maturana et al., 2020; Baud et al., 2018). There is growing evidence that characteristics of seizures undergo long-term evolution within an individual subject (Crisp et al., 2020), and mechanisms underlying the ictogenesis may be influenced by circadian and multidien cycles (Kudlacek et al., 2021; Proix et al., 2021). Combining the information about the epileptic network dynamics at multiple temporal scales can thus provide meaningful information about ictogenesis.

Proictal States, Seizure Probability Fluctuation, and Markers of Resilience

In the last decade, long-term implantable intracranial EEG (iEEG) devices emerged which enabled studies on long-term dynamics of seizures in people with pharmacoresistant epilepsy who could be monitored for months and years. Two research projects dominated this field. The NeuroPace project aimed to perform closed-loop responsive neurostimulation to abort seizures, and the chronically implanted device allowed recording of intermittent short epochs of iEEG (Baud et al., 2018). The NeuroVista study aimed to test the feasibility of continuous long-term iEEG recording to improve seizure forecasting (Cook et al., 2013). Both studies brought intriguing and groundbreaking information about the long-term fluctuations in seizure probability and substantially advanced our understanding of how seizures emerge. In the NeuroPace recorded data, Baud et al. (2018) analyzed temporal profiles of interictal epileptiform activity and demonstrated the existence of circadian as well as multidien rhythmic fluctuations in IED rate. In each patient, multiple rhythms of interictal activity could coexist. Surprisingly, seizures tend to occur on the rising phases of these fluctuations. These observations were supported by analogous results obtained in experimental models of chronic epilepsy induced by pilocarpine and kainate in rats (Baud et al., 2019). In data recorded during the NeuroVista data, Maturana et al. (2020) analyzed the features of critical slowing, also at multiple temporal scales. They observed that the autocorrelation function width (which conveys similar information to lag-1 autocorrelation) and signal variance fluctuated in similar rhythms together with the IED rate. Again, seizures emerged during the rising phases of these rhythms. Interestingly, they could also use the fluctuations of the markers of critical slowing down for seizure forecasting. Experimentally, these human observations were supported by Chang et al., who examined the transition between seizure clusters in the intrahippocampal tetanus toxin model of temporal lobe epilepsy (Chang et al., 2018). By extracting the early warning signals from the epileptic bursts, they demonstrated that the transition to the next cluster and period of high seizure probability displayed features of critical slowing.

The mechanisms of seizure probability fluctuations are not fully understood, but certain theories exist. In the case of the circadian fluctuation, one possible factor could be the hormone melatonin, which has highest levels during the night in humans (Claustrat and Leston, 2015). Agomelatine, an agonist of MT1 and MT2 melatonin receptors, was shown to have an antiseizure effect (Aguiar et al., 2012; Dastgheib and Moezi, 2014), which is consistent with lower probability of seizures in the mesial temporal lobe epilepsy patients at night (Leite Goes Gitai et al., 2019; Mirzoev et al., 2012). Expression of several genes showing circadian pattern (e.g., mTOR, mechanistic target of rapamycin) was shown to influence the seizure probability (Leite Goes Gitai et al., 2019).

A common cause of multidien fluctuations of seizure probability is the menstrual cycle, in which case we talk about catamenial epilepsy. Several patterns were observed, one of which is perimenstrual seizure exacerbation. This may be caused by the withdrawal of progesterone during the premenstrual period which results in a decrease of allopregnanolone, a positive allosteric modulator of GABAA receptors (Herzog, 2015). However, often the fluctuation of seizure probability is not related to the obvious biological cycles or environmental changes.

The cause of the seizure probability fluctuation and seizure clustering can then be the seizures per se. For example, Kudlacek et al. (2021) speculated that the focal seizures occurring at the first part of the cluster facilitate generalization of later seizures via kindling mechanism (Racine, 1972; Loscher and Kohling, 2010). Once the generalized seizures appear, they have rather anticonvulsive effect (Mucha and Pinel, 1977; Handforth, 1982; Fink, 1978; Green, 1960; Shapira et al., 1996), so the seizures get less and less frequent and finally disappear for some time.

The molecular mechanism of the seizure-induced inhibition may be similar to those responsible for seizure termination and postictal depression (Loscher and Kohling, 2010). Energy failure, metabolic disturbances, depletion of glutamate, or increase in GABA concentration, as well as changes in pH, ion concentrations, or gap junction function, seem rather unlikely causes of seizure arrest and subsequent inhibition (Loscher and Kohling, 2010).

The most probable mechanisms of postictal depression and possibly long-term fluctuations in seizure probability are neuromodulators (Loscher and Kohling, 2010). Endogenous opioids were shown to be released in the epileptic foci of patients with reading-induced seizures (Koepp et al., 1998). Endogenous cannabinoids may also have antiseizure effect, although their release during seizures was not confirmed so far (Loscher and Kohling, 2010).

An important inhibitory neuromodulator is adenosine (Boison, 2005). Antiepileptic effects of adenosine or its analog 2-chloroadenosine were shown in numerous studies in vitro and in vivo (Dragunow et al., 1985; Cavalheiro et al., 1987; Fukuda et al., 2010; Kostopoulos et al., 1989). Since adenosine is released during seizures (Berman et al., 2000; During and Spencer, 1992), it may act as a negative feedback responsible for the seizure arrest and postictal depression. It is not clear if it could be responsible for the slower fluctuations in seizure probability. During and Spencer reported that adenosine levels in the seizure focus of human patients remained elevated for at least 18 minutes after the seizure, but longer measurements are missing (During and Spencer, 1992).

Another extensively studied neuromodulator is neuropeptide Y (NPY). NPY targets five classes of receptors (Y1–Y5). In the hippocampus, the most abundant one is Y2, which reduces glutamate release and, thus, excitatory neurotransmission (Lado and Moshe, 2008; Vezzani and Sperk, 2004). Antiepileptic action of NPY through Y2 receptor was demonstrated by numerous studies (Melin et al., 2019; Woldbye et al., 1996; Vezzani and Sperk, 2004). NPY is strongly released during excessive neuronal activity. Afterward, it is degraded and must be resynthesized, which explains why tissue levels of NPY may be decreased for 2–3 days after severe seizures. This may, however, increase the seizure probability within that period. Therefore, NPY depletion could be one of the mechanisms of seizure clustering. To replenish the depleted NPY, its expression increases after the seizures. In the kindling model of epilepsy it was shown to remain increased up to 2 weeks after the last seizure. Also Y2 receptor expression increases for up to 10 days after kinate-induced status epilepticus (Vezzani and Sperk, 2004). This slow time course makes NPY a possible driver of the multidien fluctuations in seizure probability.

Conclusions and Future Research

The evidence herein demonstrates that shifting the attention from seizure initiation and transition also to seizure probability fluctuations and rhythms is beneficial for understanding the ictogenesis. The markers of the critical slowing together with the IED rate could then be used for forecasting seizure probability. Forecasting should not be confused with seizure prediction. Whereas seizure prediction aims at predicting individual seizures minutes or hours in advance, forecasting rather informs about longer periods (e.g., a day) of increased seizure probability without the ambition to predict the exact number and timing of the seizures. The forecasting algorithm that utilized early warning signals and IED rate was developed by Maturana et al. (2020). The algorithm correctly classified 77 ± 8% seizures into the high-risk category. The percentage of time spent in the high-risk category was 9 ± 8%. The performance of the forecaster was better than all previous seizure predictors tested on the data set. The authors noted that their forecaster was the first to use the whole long-term continuous recordings and speculated that this might be the reason for the good performance of their forecaster. Recently, Proix et al. published a development and validation study of a seizure forecaster based on the rhythmic fluctuations of the IED rate. In the retrospective study, their forecasts for the next calendar day were significantly better than random forecasts in 66% of the validation cohort (Proix et al., 2021). In contrast to seizure prediction, seizure forecasting may be relevant also in epilepsies where seizures are inherently unpredictable, such as absence epilepsy. Although it is impossible to predict individual seizures, it may still be possible to track parameters related to mean seizure frequency and get information on relative seizure probability (Suffczynski et al., 2004).

Most importantly, combining information about the processes at multiple time scales together is fundamental to understanding how seizures really emerge. Furthermore, identifying the mechanisms involved in those processes and dynamical pathways would bring new therapeutic targets to improve the control of seizures. From the perspective of therapeutic neurostimulation for epilepsy, knowledge of the dynamical principles can improve the optimization of stimulation therapy with respect to the current dynamical state of the brain and improve the stimulation efficacy (Baud and Rao, 2018; Loddenkemper, 2012). The ability to determine the seizure probability in each patient would be beneficial for optimizing the dose of available antiseizure medication during the periods of low or high seizure probability, so-called chronopharmacology. However, methods for seizure forecasting need to be rigorously tested prospectively rather than retrospectively before they can really become useful. There is no doubt that future research into mechanisms of seizure emergence will benefit from a close interdisciplinary collaboration between experimental epileptologists and experts in the dynamics of complex systems. Great attention should be paid to the ictogenic processes operating on long temporal scales.

Acknowledgments

Disclosure Statement

References

1. Aguiar CC, Almeida AB, Araújo PV, Vasconcelos GS, Chaves EM, do Vale OC, et al. Anticonvulsant effects of agomelatine in mice.  Epilepsy Behav.  2012;24(3):324–8. Epub 2012/06/05. doi: 10.1016/j.yebeh.2012.04.134. PubMed PMID: 22658946. [PubMed: 22658946]
2. Ahmed OJ, John TT. A Straw Can Break a Neural Network’s Back and Lead to SeizuresBut Only When Delivered at the Right Time.  Epilepsy Curr.  2019;19(2):115–6. doi: 10.1177/1535759719835349. PubMed PMID: 30955435. [PMC free article: PMC6610409] [PubMed: 30955435]
3. Alonso-Nanclares L, Garbelli R, Sola RG, Pastor J, Tassi L, Spreafico R, et al. Microanatomy of the dysplastic neocortex from epileptic patients.  Brain.  2005;128(1):158–73. doi: 10.1093/brain/awh331. [PubMed: 15548558]
4. Arellano JI, Muñoz A, Ballesteros-Yáñez I, Sola RG, DeFelipe J. Histopathology and reorganization of chandelier cells in the human epileptic sclerotic hippocampus.  Brain.  2004;127(1):45–64. doi: 10.1093/brain/awh004. [PubMed: 14534159]
5. Avoli M, Bernasconi A, Mattia D, Olivier A, Hwa GG. Epileptiform discharges in the human dysplastic neocortex: in vitro physiology and pharmacology.  Ann Neurol.  1999;46(6):816–26. Epub 1999/12/10. doi: 10.1002/1531-8249(199912)46:6<816::aid-ana3>3.0.co;2-o. PubMed PMID: 10589533. [PubMed: 10589533]
6. Avoli M, de Curtis M. GABAergic synchronization in the limbic system and its role in the generation of epileptiform activity.  Prog Neurobiol.  2011;95(2):104–32. doi: 10.1016/j.pneurobio.2011.07.003. PubMed PMID: WOS:000295949700002. [PMC free article: PMC4878907] [PubMed: 21802488]
7. Avoli M, Jefferys JGR. Models of drug-induced epileptiform synchronization in vitro.  J Neurosci Methods.  2016;260:26–32. doi: https://doi​.org/10.1016/j​.jneumeth.2015.10.006. [PMC free article: PMC4878885] [PubMed: 26484784]
8. Avoli M, Panuccio G, Herrington R, D’Antuono M, de Guzman P, Lévesque M. Two different interictal spike patterns anticipate ictal activity in vitro.  Neurobiol Dis.  2013;52:168–76. Epub 2012/12/29. doi: 10.1016/j.nbd.2012.12.004. PubMed PMID: 23270790. [PMC free article: PMC4880481] [PubMed: 23270790]
9. Baud MO, Ghestem A, Benoliel JJ, Becker C, Bernard C. Endogenous multidien rhythm of epilepsy in rats.  Exp Neurol.  2019;315:82–7. Epub 2019/02/19. doi: 10.1016/j.Exp neurol.2019.02.006. PubMed PMID: 30776337. [PubMed: 30776337]
10. Baud MO, Kleen JK, Mirro EA, Andrechak JC, King-Stephens D, Chang EF, et al. Multi-day rhythms modulate seizure risk in epilepsy.  Nat Commun.  2018;9(1):88. doi: 10.1038/s41467-017-02577-y. PubMed PMID: 29311566. [PMC free article: PMC5758806] [PubMed: 29311566]
11. Baud MO, Rao VR.  Gauging seizure risk.  Neurology.  2018;91(21):967–73. Epub 2018/10/26. doi: 10.1212/wnl.0000000000006548. PubMed PMID: 30355701. [PubMed: 30355701]
12. Berman RF, Fredholm BB, Aden U, O’Connor WT. Evidence for increased dorsal hippocampal adenosine release and metabolism during pharmacologically induced seizures in rats.  Brain Res.  2000;872(1-2):44–53. Epub 2000/08/05. doi: 10.1016/s0006-8993(00)02441-0. PubMed PMID: 10924674. [PubMed: 10924674]
13. Bikson M, Id BR, Vreugdenhil M, Kohling R, Fox JE, Jefferys JG. Quinine suppresses extracellular potassium transients and ictal epileptiform activity without decreasing neuronal excitability in vitro.  Neuroscience.  2002;115(1):251–61. [PubMed: 12401338]
14. Bikson M, Inoue M, Akiyama H, Deans JK, Fox JE, Miyakawa H, et al. Effects of uniform extracellular DC electric fields on excitability in rat hippocampal slices in vitro.  J Physiol.  2004;557(Pt 1):175–90. [PMC free article: PMC1665051] [PubMed: 14978199]
15. Blauwblomme T, Jiruska P, Huberfeld G. Mechanisms of ictogenesis.  Int Rev Neurobiol.  2014;114:155–85. doi: 10.1016/B978-0-12-418693-4.00007-8. PubMed PMID: 25078502. [PubMed: 25078502]
16. Boison D. Adenosine and epilepsy: from therapeutic rationale to new therapeutic strategies.  Neuroscientist.  2005;11(1):25–36. Epub 2005/01/06. doi: 10.1177/1073858404269112. PubMed PMID: 15632276. [PubMed: 15632276]
17. Cavalheiro EA, Calderazzo Filho LS, Bortolotto ZA, Mello L, Turski L. Anticonvulsant role of adenosine.  Pol J Pharmacol Pharm.  1987;39(5):537–43. Epub 1987/09/01. PubMed PMID: 3506733. [PubMed: 3506733]
18. Chang WC, Kudlacek J, Hlinka J, Chvojka J, Hadrava M, Kumpost V, et al. Loss of neuronal network resilience precedes seizures and determines the ictogenic nature of interictal synaptic perturbations.  Nat Neurosci.  2018;21(12):1742–52. doi: 10.1038/s41593-018-0278-y. PubMed PMID: 30482946. [PubMed: 30482946]
19. Claustrat B, Leston J. Melatonin: Physiological effects in humans.  Neurochirurgie.  2015;61(2-3):77–84. doi: 10.1016/j.neuchi.2015.03.002. PubMed PMID: 25908646. [PubMed: 25908646]
20. Cohen I, Navarro V, Clemenceau S, Baulac M, Miles R. On the origin of interictal activity in human temporal lobe epilepsy in vitro.  Science.  2002;298(5597):1418–21. Epub 2002/11/16. doi: 10.1126/science.1076510. PubMed PMID: 12434059. [PubMed: 12434059]
21. Cook MJ, O’Brien TJ, Berkovic SF, Murphy M, Morokoff A, Fabinyi G, et al. Prediction of seizure likelihood with a long-term, implanted seizure advisory system in patients with drug-resistant epilepsy: a first-in-man study.  Lancet Neurosci.  2013;12(6):563–71. doi: 10.1016/S1474-4422(13)70075-9. PubMed PMID: 23642342. [PubMed: 23642342]
22. Cossart R, Dinocourt C, Hirsch JC, Merchan-Perez A, De Felipe J, Ben-Ari Y, et al. Dendritic but not somatic GABAergic inhibition is decreased in experimental epilepsy.  Nature Neurosci.  2001;4(1):52–62. doi: 10.1038/82900. [PubMed: 11135645]
23. Crisp DN, Cheung W, Gliske SV, Lai AL, Freestone DR, Grayden DB, et al. Quantifying epileptogenesis in rats with spontaneous and responsive brain state dynamics.  Brain Commun.  2020;2(1):fcaa048. doi: ARTN 48 10.1093/braincomms/fcaa048. PubMed PMID: 32671339. [PMC free article: PMC7331126] [PubMed: 32671339]
24. Dakos V, Scheffer M, van Nes EH, Brovkin V, Petoukhov V, Held H. Slowing down as an early warning signal for abrupt climate change.  Proc Natl Acad Sci USA.  2008;105(38):14308–12. doi: 10.1073/pnas. 0802430105. PubMed PMID: 18787119. [PMC free article: PMC2567225] [PubMed: 18787119]
25. Dakos V, van Nes EH, Scheffer M. Flickering as an early warning signal.  Theor Ecol-Neth.  2013;6(3):309–17. doi: 10.1007/s12080-013-0186-4. WOS:000322022400007.
26. D’Antuono M, Louvel J, Kohling R, Mattia D, Bernasconi A, Olivier A, Turak B, Devaux A, Pumain R, Avoli M. GABA_A receptor-dependent synchronization leads to ictogenesis in the human dysplastic cortex. Brain.  2004;127(7):1626-40. doi: 10.1093/brain/awh181 [PubMed: 15175227]
27. Dastgheib M, Moezi L. Acute and chronic effects of agomelatine on intravenous penthylenetetrazol-induced seizure in mice and the probable role of nitric oxide.  Eur J Pharmacol.  2014;736:10–5. Epub 2014/05/08. doi: 10.1016/j.ejphar.2014.04.039. PubMed PMID: 24803306. [PubMed: 24803306]
28. de Curtis M, Avanzini G. Interictal spikes in focal epileptogenesis.  Prog Neurobiol.  2001;63(5):541–67. [PubMed: 11164621]
29. de Curtis M, Avoli M. GABAergic networks jump-start focal seizures.  Epilepsia.  2016;57(5):679–87. doi: 10.1111/epi.13370. PubMed PMID: WOS:000380140700004. [PMC free article: PMC4878883] [PubMed: 27061793]
30. de Curtis M, Gnatkovsky V. Reevaluating the mechanisms of focal ictogenesis: The role of low-voltage fast activity.  Epilepsia.  2009;50(12):2514–25. [PubMed: 19674056]
31. Dragunow M, Goddard GV, Laverty R. Is adenosine an endogenous anticonvulsant? Epilepsia.  1985;26(5):480–7. Epub 1985/09/01. doi: 10.1111/j.1528-1157.1985.tb05684.x. PubMed PMID: 4043018. [PubMed: 4043018]
32. Dudek FE, Staley KJ. Seizure probability in animal models of acquired epilepsy: a perspective on the concept of the preictal state.  Epilepsy Res.  2011a;97(3):324–31. [PubMed: 22094446]
33. Dudek FE, Staley KJ. The time course of acquired epilepsy: implications for therapeutic intervention to suppress epileptogenesis.  Neurosci Lett.  2011b;497(3):240–6. [PubMed: 21458536]
34. During MJ, Spencer DD. Adenosine: a potential mediator of seizure arrest and postictal refractoriness.  Ann Neurol.  1992;32(5):618–24. Epub 1992/11/01. doi: 10.1002/ana.410320504. PubMed PMID: 1449242. [PubMed: 1449242]
35. During MJ, Spencer DD. Extracellular hippocampal glutamate and spontaneous seizure in the conscious human brain.  Lancet.  1993;341(8861):1607–10. doi: https://doi​.org/10.1016​/0140-6736(93)90754-5. [PubMed: 8099987]
36. Dzhala VI, Staley KJ. Transition from Interictal to Ictal Activity in Limbic Networks <em>In Vitro</em&gt.  J Neurosci.  2003;23(21):7873-80. doi: 10.1523/ JNEUROSCI. 23-21-07873.2003. [PMC free article: PMC6740615] [PubMed: 12944517]
37. Eid T, Thomas MJ, Spencer DD, Rundén-Pran E, Lai JCK, Malthankar GV, et al. Loss of glutamine synthetase in the human epileptogenic hippocampus: possible mechanism for raised extracellular glutamate in mesial temporal lobe epilepsy.  Lancet.  2004;363(9402):28–37. doi: https://doi​.org/10.1016​/S0140-6736(03)15166-5. [PubMed: 14723991]
38. Elahian B, Lado NE, Mankin E, Vangala S, Misra A, Moxon K, et al. Low-voltage fast seizures in humans begin with increased interneuron firing.  Ann Neurol.  2018;84(4):588–600. doi: https://doi​.org/10.1002/ana.25325. [PMC free article: PMC6814155] [PubMed: 30179277]
39. Falcón-Moya R, Sihra TS, Rodríguez-Moreno A. Kainate Receptors: Role in Epilepsy.  Front Mol Neurosci.  2018;11:217–. doi: 10.3389/fnmol.2018.00217. PubMed PMID: 29988380. [PMC free article: PMC6023982] [PubMed: 29988380]
40. Fink M. Efficacy and safety of induced seizures (EST) in man.  Compr Psychiatry.  1978;19(1):1–18. Epub 1978/01/01. PubMed PMID: 620522. [PubMed: 620522]
41. Fox JE, Bikson M, Jefferys JG. The effect of neuronal population size on the development of epileptiform discharges in the low calcium model of epilepsy.  Neurosci Lett.  2007;411(2):158–61. [PubMed: 17097230]
42. Freestone DR, Kuhlmann L, Grayden DB, Burkitt AN, Lai A, Nelson TS, et al. Electrical probing of cortical excitability in patients with epilepsy.  Epilepsy Behav.  2011;22 Suppl 1:S110–8. doi: 10.1016/j.yebeh.2011.09.005. PubMed PMID: 22078511. [PubMed: 22078511]
43. Fritsch B, Reis J, Gasior M, Kaminski RM, Rogawski MA. Role of GluK1 kainate receptors in seizures, epileptic discharges, and epileptogenesis.  J Neurosci.  2014;34(17):5765–75. doi: 10.1523/JNEUROSCI.5307-13.2014. PubMed PMID: 24760837. [PMC free article: PMC3996208] [PubMed: 24760837]
44. Fukuda M, Suzuki Y, Hino H, Kuzume K, Morimoto T, Ishii E. Adenosine A1 receptor blockage mediates theophylline-associated seizures. Epilepsia.  2010;51(3):483–7. Epub 2009/10/23. doi: 10.1111/j.1528-1167.2009.02382.x. PubMed PMID: 19845735. [PubMed: 19845735]
45. Gnatkovsky V, Librizzi L, Trombin F, de CM.  Fast activity at seizure onset is mediated by inhibitory circuits in the entorhinal cortex in vitro.  Ann Neurol.  2008;64(6):674–86. [PubMed: 19107991]
46. Green MA. Relation between threshold and duration of seizures and electrographic change during convulsive therapy.  J Nerv Ment Dis.  1960;131:117–20. Epub 1960/08/01. PubMed PMID: 13708197. [PubMed: 13708197]
47. Hamidi S, Lévesque M, Avoli M. Epileptiform synchronization and high-frequency oscillations in brain slices comprising piriform and entorhinal cortices.  Neuroscience.  2014;281:258–68. Epub 2014/10/08. doi: 10.1016/j.neuroscience.2014.09.065. PubMed PMID: 25290016. [PMC free article: PMC4929005] [PubMed: 25290016]
48. Handforth A. Postseizure inhibition of kindled seizures by electroconvulsive shock.  Exp Neurol.  1982;78(3):483–91. Epub 1982/12/01. PubMed PMID: 7173367. [PubMed: 7173367]
49. Herzog AG. Catamenial epilepsy: Update on prevalence, pathophysiology and treatment from the findings of the NIH Progesterone Treatment Trial. Seizure.  2015;28:18–25. Epub 2015/03/15. doi: 10.1016/j.seizure.2015.02.024. PubMed PMID: 25770028. [PubMed: 25770028]
50. Huberfeld G, Menendez de la Prida L, Pallud J, Cohen I, Le Van Quyen M, Adam C, et al. Glutamatergic pre-ictal discharges emerge at the transition to seizure in human epilepsy.  Nat Neurosci.  2011;14(5):627–34. doi: 10.1038/nn.2790. [PubMed: 21460834]
51. Huberfeld G, Wittner L, Clemenceau S, Baulac M, Kaila K, Miles R, et al. Perturbed chloride homeostasis and GABAergic signaling in human temporal lobe epilepsy.  J Neurosci.  2007;27(37):9866–73. Epub 2007/09/15. doi: 10.1523/jneurosci.2761-07.2007. PubMed PMID: 17855601. [PMC free article: PMC6672644] [PubMed: 17855601]
52. Jefferys JG, Haas HL. Synchronized bursting of CA1 hippocampal pyramidal cells in the absence of synaptic transmission.  Nature.  1982;300(5891):448–50. [PubMed: 6292731]
53. Jensen MS, Yaari Y. Role of Intrinsic Burst Firing, Potassium Accumulation, and Electrical Coupling in the Elevated Potassium Model of Hippocampal Epilepsy.  J Neurophysiol.  1997;77(3):1224–33. doi: 10.1152/jn.1997.77.3.1224. [PubMed: 9084592]
54. Jensen MS, Yaari Y. The relationship between interictal and ictal paroxysms in an in vitro model of focal hippocampal epilepsy.  Ann Neurol.  1988;24(5):591–8. [PubMed: 2849367]
55. Jiruska P, Alvarado-Rojas C, Schevon CA, Staba R, Stacey W, Wendling F, et al. Update on the mechanisms and roles of high-frequency oscillations in seizures and epileptic disorders.  Epilepsia.  2017;58(8):1330–9. doi: 10.1111/epi.13830. PubMed PMID: 28681378. [PMC free article: PMC5554080] [PubMed: 28681378]
56. Jiruska P, Csicsvari J, Powell AD, Fox JE, Chang WC, Vreugdenhil M, et al. High-frequency network activity, global increase in neuronal activity, and synchrony expansion precede epileptic seizures in vitro.  J Neurosci.  2010;30(16):5690–701. Epub 2010/04/23. doi: 10.1523/jneurosci.0535-10.2010. PubMed PMID: 20410121; PubMed Central PMCID: PMC6632330. [PMC free article: PMC6632330] [PubMed: 20410121]
57. Jiruska P, de Curtis M, Jefferys JG, Schevon CA, Schiff SJ, Schindler K. Synchronization and desynchronization in epilepsy: controversies and hypotheses.  J Physiol.  2013a;591(Pt 4):787–97. doi: 10.1113/jphysiol.2012.239590. PubMed PMID: 23184516. [PMC free article: PMC3591697] [PubMed: 23184516]
58. Jiruska P, Mormann F, Jefferys JGR. Neuronal and Network Dynamics Preceding Experimental Seizures.  Recent Advances in Predicting and Preventing Epileptic Seizures. In “Recent Advances in Predicting and Preventing Epileptic Seizures” Eds Tetzlaff R, Elger CE, Lenhertz K. Singapore: World Scientific 2013b:16–29. WOS:000337570200002.
59. Kalitzin S, Velis D, Suffczynski P, Parra J, da Silva FL. Electrical brain-stimulation paradigm for estimating the seizure onset site and the time to ictal transition in temporal lobe epilepsy.  Clin Neurophysiol.  2005;116(3):718–28. doi: 10.1016/j.clinph.2004.08.021. PubMed PMID: 15721087. [PubMed: 15721087]
60. Khosravani H, Pinnegar CR, Mitchell JR, Bardakjian BL, Federico P, Carlen PL. Increased High-frequency Oscillations Precede in vitro Low-Mg²⁺ Seizures. Epilepsia.  2005;46(8):1188–97. doi: https://doi​.org/10.1111/j​.1528-1167.2005.65604.x. [PubMed: 16060927]
61. Koepp MJ, Richardson MP, Brooks DJ, Duncan JS. Focal cortical release of endogenous opioids during reading-induced seizures.  Lancet.  1998;352(9132):952–5. Epub 1998/09/30. doi: 10.1016/s0140-6736(97)09077-6. PubMed PMID: 9752818. [PubMed: 9752818]
62. Kohling R, D’Antuono M, Benini R, de Guzman P, Avoli M. Hypersynchronous ictal onset in the perirhinal cortex results from dynamic weakening in inhibition.  Neurobiol Dis.  2016;87:1–10. PubMed PMID: 26699817. [PMC free article: PMC4878890] [PubMed: 26699817]
63. Kohling R, D’Antuono M, Benini R, de Guzman P, Avoli M. Hypersynchronous ictal onset in the perirhinal cortex results from dynamic weakening in inhibition.  Neurobiol Dis.  2016;87:1–10. Epub 2015/12/25. doi: 10.1016/j.nbd.2015.12.002. PubMed PMID: 26699817. [PMC free article: PMC4878890] [PubMed: 26699817]
64. Konnerth A, Heinemann U, Yaari Y. Nonsynaptic epileptogenesis in the mammalian hippocampus in vitro. I. Development of seizurelike activity in low extracellular calcium.  J Neurophysiol.  1986;56(2):409–23. [PubMed: 3760928]
65. Kostopoulos G, Drapeau C, Avoli M, Olivier A, Villemeure JG. Endogenous adenosine can reduce epileptiform activity in the human epileptogenic cortex maintained in vitro.  Neurosci Lett.  1989;106(1):119–24. doi: https://doi​.org/10.1016​/0304-3940(89)90212-7. [PubMed: 2586817]
66. Kramer MA, Truccolo W, Eden UT, Lepage KQ, Hochberg LR, Eskandar EN, et al. Human seizures self-terminate across spatial scales via a critical transition.  Proc Natl Acad Sci USA.  2012;109(51):21116–21. doi: 10.1073/pnas.1210047110. [PMC free article: PMC3529091] [PubMed: 23213262]
67. Kudlacek J, Chvojka J, Kumpost V, Hermanovska B, Posusta A, Jefferys JGR, et al. Long-term seizure dynamics are determined by the nature of seizures and the mutual interactions between them.  Neurobiol Dis.  2021;154:105347. Epub 2021/03/28. doi: 10.1016/j.nbd.2021.105347. PubMed PMID: 33771663. [PubMed: 33771663]
68. Kuhlmann L, Lehnertz K, Richardson MP, Schelter B, Zaveri HP. Seizure prediction - ready for a new era.  Nature Rev Neurol.  2018;14(10):618–30. Epub 2018/08/23. doi: 10.1038/s41582-018-0055-2. PubMed PMID: 30131521. [PubMed: 30131521]
69. Lado FA, Moshe SL. How do seizures stop? Epilepsia.  2008;49(10):1651–64. doi: DOI 10.1111/j.1528-1167.2008.01669.x. PubMed PMID: 18503563. [PMC free article: PMC2738747] [PubMed: 18503563]
70. Lagarde S, Bonini F, McGonigal A, Chauvel P, Gavaret M, Scavarda D, et al. Seizure-onset patterns in focal cortical dysplasia and neurodevelopmental tumors: Relationship with surgical prognosis and neuropathologic subtypes.  Epilepsia.  2016;57(9):1426–35. doi: 10.1111/epi.13464. PubMed PMID: 27406939. [PubMed: 27406939]
71. Lehnertz K, Mormann F, Osterhage H, Muller A, Prusseit J, Chernihovskyi A, et al. State-of-the-art of seizure prediction.  J Clin Neurophysiol.  2007;24(2):147–53. [PubMed: 17414970]
72. Leite Goes Gitai D, de Andrade TG, Dos Santos YDR, Attaluri S, Shetty AK. Chronobiology of limbic seizures: Potential mechanisms and prospects of chronotherapy for mesial temporal lobe epilepsy.  Neurosci Biobehav Rev.  2019;98:122–34. doi: 10.1016/j.neubiorev.2019.01.004. PubMed PMID: 30629979. [PMC free article: PMC7023906] [PubMed: 30629979]
73. Li X, Cui D, Jiruska P, Fox JE, Yao X, Jefferys JG. Synchronization measurement of multiple neuronal populations.  J Neurophysiol.  2007;98(6):3341–8. doi: 10.1152/jn.00977.2007. PubMed PMID: 17913983. [PubMed: 17913983]
74. Litt B, Esteller R, Echauz J, D’Alessandro M, Shor R, Henry T, et al. Epileptic seizures may begin hours in advance of clinical onset: a report of five patients. Neuron.  2001;30(1):51–64. PubMed PMID: 11343644. [PubMed: 11343644]
75. Litt B, Lehnertz K. Seizure prediction and the preseizure period.  Curr Opin Neurosl.  2002;15(2):173–7. [PubMed: 11923631]
76. Loddenkemper T. Chrono-epileptology: time to reconsider seizure timing.  Seizure.  2012;21(6):411. doi: 10.1016/j.seizure.2012.05.013. PubMed PMID: 22698973. [PubMed: 22698973]
77. Lopes da Silva FH, Blanes W, Kalitzin SN, Parra J, Suffczynski P, Velis DN. Dynamical diseases of brain systems: Different routes to epileptic seizures.  IEEE Trans Biomed Eng.  2003a;50(5):540–8. doi: 10.1109/Tbme.2003.810703. PubMed PMID: 12769430. [PubMed: 12769430]
78. Lopes da Silva FH, Blanes W, Kalitzin SN, Parra J, Suffczynski P, Velis DN. Epilepsies as dynamical diseases of brain systems: Basic models of the transition between normal and epileptic activity.  Epilepsia.  2003b;44:72–83. PubMed PMID: 14641563. [PubMed: 14641563]
79. Lopes MA, Lee KE, Goltsev AV. Neuronal network model of interictal and recurrent ictal activity.  Phys Rev E.  2017;96(6-1):062412. PubMed PMID: 29347379. [PubMed: 29347379]
80. Loscher W, Kohling R. Functional, metabolic, and synaptic changes after seizures as potential targets for antiepileptic therapy.  Epilepsy Behav.  2010;19(2):105–13. doi: 10.1016/j.yebeh.2010.06.035. PubMed PMID: 20705520. [PubMed: 20705520]
81. Loscher W, Puskarjov M, Kaila K. Cation-chloride cotransporters NKCC1 and KCC2 as potential targets for novel antiepileptic and antiepileptogenic treatments.  Neuropharmacology.  2013;69:62–74. doi: https://doi​.org/10.1016/j​.neuropharm.2012.05.045. [PubMed: 22705273]
82. Matsumoto H, Marsan CA. Cortical cellular phenomena in experimental epilepsy: interictal manifestations.  Exp Neurol.  1964;9:286–304. Epub 1964/04/01. doi: 10.1016/0014-4886(64)90025-1. PubMed PMID: 14145629. [PubMed: 14145629]
83. Mattia D, Olivier A, Avoli M. Seizure-like discharges recorded in human dysplastic neocortex maintained in vitro.  Neurology.  1995;45(7):1391–5. Epub 1995/07/01. doi: 10.1212/wnl.45.7.1391. PubMed PMID: 7617202. [PubMed: 7617202]
84. Maturana MI, Meisel C, Dell K, Karoly PJ, D’Souza W, Grayden DB, et al. Critical slowing down as a biomarker for seizure susceptibility.  Nat Commun.  2020;11(1):2172. doi: 10.1038/s41467-020-15908-3. [PMC free article: PMC7195436] [PubMed: 32358560]
85. McCormick DA, Contreras D. On the cellular and network bases of epileptic seizures.  Ann Rev Physiol.  2001;63:815–46. Epub 2001/02/22. doi: 10.1146/annurev.physiol.63.1.815. PubMed PMID: 11181977. [PubMed: 11181977]
86. McNamara JO.  Cellular and molecular basis of epilepsy.  J Neurosci.  1994;14(6):3413–25. Epub 1994/06/01. doi: 10.1523/jneurosci.14-06-03413.1994. PubMed PMID: 8207463. [PMC free article: PMC6576925] [PubMed: 8207463]
87. Melin E, Nanobashvili A, Avdic U, Gøtzsche CR, Andersson M, Woldbye DPD, et al. Disease Modification by Combinatorial Single Vector Gene Therapy: A Preclinical Translational Study in Epilepsy.  Mol Ther Methods Clin Dev.  2019;15:179–93. Epub 2019/10/30. doi: 10.1016/j.omtm.2019.09.004. PubMed PMID: 31660420. [PMC free article: PMC6807261] [PubMed: 31660420]
88. Milanowski P, Suffczynski P. Seizures Start without Common Signatures of Critical Transition.  Int J Neural Syst.  2016;26(08):1650053. doi: 10.1142/s0129065716500532. PubMed PMID: 27633895. [PubMed: 27633895]
89. Mirzoev A, Bercovici E, Stewart LS, Cortez MA, Snead OC, 3rd, Desrocher M.  Circadian profiles of focal epileptic seizures: a need for reappraisal.  Seizure.  2012;21(6):412–6. doi: 10.1016/j.seizure.2012.03.014. PubMed PMID: 22537415. [PubMed: 22537415]
90. Mormann F, Andrzejak RG, Elger CE, Lehnertz K. Seizure prediction: the long and winding road.  Brain.  2007;130(Pt 2):314–33. [PubMed: 17008335]
91. Mucha RF, Pinel PJ. Postseizure inhibition of kindled seizures.  Exp Neurol.  1977;54(2):266–82. Epub 1977/02/01. PubMed PMID: 838017. [PubMed: 838017]
92. Muñoz A, Méndez P, DeFelipe J, Alvarez-Leefmans FJ. Cation-Chloride Cotransporters and GABA-ergic Innervation in the Human Epileptic Hippocampus.  Epilepsia.  2007;48(4):663–73. doi: https://doi​.org/10.1111/j​.1528-1167.2005.65604.x. [PubMed: 17319917]
93. Ortinski PI, Dong J, Mungenast A, Yue C, Takano H, Watson DJ, et al. Selective induction of astrocytic gliosis generates deficits in neuronal inhibition.  Nature Neurosci.  2010;13(5):584–91. doi: 10.1038/nn.2535. [PMC free article: PMC3225960] [PubMed: 20418874]
94. Parrish RR, Codadu NK, Mackenzie-Gray Scott C, Trevelyan AJ. Feedforward inhibition ahead of ictal wavefronts is provided by both parvalbumin- and somatostatin-expressing interneurons.  J Physiol.  2019;597(8):2297–314. doi: https://doi​.org/10.1113/jp277749. [PMC free article: PMC6462485] [PubMed: 30784081]
95. Perez EL, Lauritzen F, Wang Y, Lee T-SW, Kang D, Zaveri HP, et al. Evidence for astrocytes as a potential source of the glutamate excess in temporal lobe epilepsy.  Neurobiol Dis.  2012;47(3):331–7. doi: https://doi​.org/10.1016/j​.nbd.2012.05.010. [PMC free article: PMC3392431] [PubMed: 22659305]
96. Perez Velazquez JL. Bicarbonate-dependent depolarizing potentials in pyramidal cells and interneurons during epileptiform activity.  Eur J Neurosci.  2003;18(5):1337–42. PubMed PMID: 12956733. [PubMed: 12956733]
97. Perucca P, Dubeau F, Gotman J. Intracranial electroencephalographic seizure-onset patterns: effect of underlying pathology.  Brain.  2014;137:183–96. PubMed PMID: 24176980. [PubMed: 24176980]
98. Proix T, Truccolo W, Leguia MG, Tcheng TK, King-Stephens D, Rao VR, et al. Forecasting seizure risk in adults with focal epilepsy: a development and validation study.  Lancet Neurol.  2021;20(2):127–35. doi: 10.1016/s1474-4422(20)30396-3. [PMC free article: PMC7968722] [PubMed: 33341149]
99. Racine RJ. Modification of seizure activity by electrical stimulation. II. Motor seizure.  Electroencephalogr Clin Neurophysiol.  1972;32(3):281–94. Epub 1972/03/01. PubMed PMID: 4110397. [PubMed: 4110397]
100. Richardson MP, Jefferys JGR. Introduction—Epilepsy Research UK Workshop 2010 on “Preictal Phenomena”. Epilepsy Res. 2011;97(3):229–30. doi: https://doi​.org/10.1016/j​.eplepsyres.2011.10.027. [PubMed: 22094447]
101. Rogawski MA.  Common Mechanisms in Migraine and Epilepsy: A Comment Reply.  Arch Neurol-Chicago.  2008;65(11):1546–7. doi: DOI 10.1001/archneur.65.11.1546-c. WOS:000260797200028. [PubMed: 19001180]
102. Rowley NM, Madsen KK, Schousboe A, Steve White H. Glutamate and GABA synthesis, release, transport and metabolism as targets for seizure control.  Neurochem Int.  2012;61(4):546–58. Epub 2012/03/01. doi: 10.1016/j.neuint.2012.02.013. PubMed PMID: 22365921. [PubMed: 22365921]
103. Salami P, Levesque M, Gotman J, Avoli M. Distinct EEG seizure patterns reflect different seizure generation mechanisms.  J Neurophysiol.  2015;113(7):2840–4. PubMed PMID: 25652916. [PMC free article: PMC4416632] [PubMed: 25652916]
104. Scheffer IE, Berkovic S, Capovilla G, Connolly MB, French J, Guilhoto L, et al. ILAE classification of the epilepsies: Position paper of the ILAE Commission for Classification and Terminology.  Epilepsia. 2017;58(4):512–21. Epub 2017/03/10. doi: 10.1111/epi.13709. PubMed PMID: 28276062. [PMC free article: PMC5386840] [PubMed: 28276062]
105. Scheffer M, Bascompte J, Brock WA, Brovkin V, Carpenter SR, Dakos V, et al. Early-warning signals for critical transitions.  Nature.  2009;461(7260):53–9. doi: 10.1038/nature08227. PubMed PMID: 19727193. [PubMed: 19727193]
106. Scheffer M, Carpenter SR, Lenton TM, Bascompte J, Brock W, Dakos V, et al. Anticipating critical transitions.  Science.  2012;338(6105):344–8. doi: 10.1126/science.1225244. PubMed PMID: 23087241. [PubMed: 23087241]
107. Scheffer M, Carpenter SR. Catastrophic regime shifts in ecosystems: linking theory to observation.  Trends Ecol Evol.  2003;18(12):648–56. doi: 10.1016/j.tree.2003.09.002. WOS:000187223100011.
108. Scheffer M. Critical transitions in nature and society. Princeton: Princeton University Press; 2009. xi. 384 p. p.
109. Schevon CA, Tobochnik S, Eissa T, Merricks E, Gill B, Parrish RR, et al. Multiscale recordings reveal the dynamic spatial structure of human seizures.  Neurobiol Dis.  2019;127:303–11. doi: https://doi​.org/10.1016/j​.nbd.2019.03.015. [PMC free article: PMC6588430] [PubMed: 30898669]
110. Schevon CA, Weiss SA, McKhann G, Jr., Goodman RR, Yuste R, Emerson RG, et al. Evidence of an inhibitory restraint of seizure activity in humans.  Nature Commun.  2012;3:1060. doi: 10.1038/ncomms2056. PubMed PMID: 22968706. [PMC free article: PMC3658011] [PubMed: 22968706]
111. Schulze-Bonhage A, Kurth C, Carius A, Steinhoff BJ, Mayer T. Seizure anticipation by patients with focal and generalized epilepsy: A multicentre assessment of premonitory symptoms.  Epilepsy Res.  2006;70(1): 83–8. doi: 10.1016/j.eplepsyres.2006.02.001. PubMed PMID: 16531010. [PubMed: 16531010]
112. Shapira B, Lidsky D, Gorfine M, Lerer B. Electroconvulsive therapy and resistant depression: clinical implications of seizure threshold.  J Clin Psychiatry.  1996;57(1):32–8. Epub 1996/01/01. PubMed PMID: 8543545. [PubMed: 8543545]
113. Shiri Z, Manseau F, Levesque M, Williams S, Avoli M. Activation of Specific Neuronal Networks Leads to Different Seizure Onset Types.  Ann Neurol.  2016;79(3):354–65. PubMed PMID: 26605509. [PMC free article: PMC4878884] [PubMed: 26605509]
114. Shiri Z, Manseau F, Levesque M, Williams S, Avoli M. Interneuron Activity Leads to Initiation of Low-Voltage Fast-Onset Seizures.  Ann Neurol.  2015;77(3):541–6. PubMed PMID: 25546300. [PMC free article: PMC4880461] [PubMed: 25546300]
115. Singh S, Sandy S, Wiebe S. Ictal onset on intracranial EEG: Do we know it when we see it? State of the evidence.  Epilepsia.  2015;56(10):1629–38. PubMed PMID: 26293970. [PubMed: 26293970]
116. Spencer SS, Guimaraes P, Katz A, Kim J, Spencer D. Morphological Patterns of Seizures Recorded Intracranially.  Epilepsia.  1992;33(3):537–45. PubMed PMID: 1592034. [PubMed: 1592034]
117. Stead M, Bower M, Brinkmann BH, Lee K, Marsh WR, Meyer FB, et al. Microseizures and the spatiotemporal scales of human partial epilepsy.  Brain.  2010;133(9):2789–97. [PMC free article: PMC2929333] [PubMed: 20685804]
118. Strogatz S.  Nonlinear Dynamics and Chaos. Cambridge MA: Westview Press. 2000.
119. Suffczynski P, Kalitzin S, da Silva F, Parra J, Velis D, Wendling F. Active paradigms of seizure anticipation: Computer model evidence for necessity of stimulation.  Phys Rev E.  2008;78(5): 051917. doi: 10.1103/PhysRevE.78.051917. [PubMed: 19113165]
120. Suffczynski P, Kalitzin S, Lopes Da Silva FH. Dynamics of non-convulsive epileptic phenomena modeled by a bistable neuronal network.  Neuroscience.  2004;126(2):467–84. doi: 10.1016/j.neuroscience.2004.03.014. PubMed PMID: 15207365. [PubMed: 15207365]
121. Suffczynski P, Lopes da Silva FH, Parra J, Velis DN, Bouwman BM, van Rijn CM, et al. Dynamics of epileptic phenomena determined from statistics of ictal transitions.  IEEE Trans Biomed Eng.  2006;53(3):524–32. PubMed PMID: 16532779. [PubMed: 16532779]
122. Tóth K, Erőss L, Vajda J, Halász P, Freund TF, Maglóczky Z. Loss and reorganization of calretinin-containing interneurons in the epileptic human hippocampus.  Brain.  2010;133(9):2763–77. doi: 10.1093/brain/awq149. [PMC free article: PMC2948815] [PubMed: 20576695]
123. Traub RD, Wong RK. Cellular mechanism of neuronal synchronization in epilepsy.  Science.  1982;216(4547):745–7. PubMed PMID: 7079735. [PubMed: 7079735]
124. Traynelis SF, Dingledine R. Potassium-induced spontaneous electrographic seizures in the rat hippocampal slice.  JNeurophysiol.  1988;59(1):259–76. [PubMed: 3343603]
125. Trevelyan AJ, Schevon CA. How inhibition influences seizure propagation.  Neuropharmacology.  2013;69:45–54. doi: https://doi​.org/10.1016/j​.neuropharm.2012.06.015. [PubMed: 22722026]
126. Trevelyan AJ, Sussillo D, Watson BO, Yuste R. Modular propagation of epileptiform activity: evidence for an inhibitory veto in neocortex.  J Neurosci.  2006;26(48):12447–55. doi: 10.1523/JNEUROSCI.2787-06.2006. PubMed PMID: 17135406. [PMC free article: PMC6674895] [PubMed: 17135406]
127. Truccolo W, Donoghue JA, Hochberg LR, Eskandar EN, Madsen JR, Anderson WS, et al. Single-neuron dynamics in human focal epilepsy.  Nature Neuroscience.  2011;14(5):635–41. doi: 10.1038/nn.2782. [PMC free article: PMC3134302] [PubMed: 21441925]
128. Ure J, Baudry M, Perassolo M. Metabotropic glutamate receptors and epilepsy.  Journal of the Neurological Sciences.  2006;247(1):1–9. doi: https://doi​.org/10.1016/j​.jns.2006.03.018. [PubMed: 16697014]
129. Uva L, Boido D, Avoli M, de Curtis M, Lévesque M. High-frequency oscillations and seizure-like discharges in the entorhinal cortex of the in vitro isolated guinea pig brain.  Epilepsy Res.  2017;130:21–6. Epub 2017/01/21. doi: 10.1016/j.eplepsyres.2017.01.001. PubMed PMID: 28107659. [PubMed: 28107659]
130. Uva L, Breschi GL, Gnatkovsky V, Taverna S, de Curtis M. Synchronous Inhibitory Potentials Precede Seizure-Like Events in Acute Models of Focal Limbic Seizures.  J Neurosci.  2015;35(7):3048–55. doi: 10.1523/Jneurosci.3692-14.2015. PubMed PMID: 25698742. [PMC free article: PMC6605586] [PubMed: 25698742]
131. van de Leemput IA, Wichers M, Cramer AO, Borsboom D, Tuerlinckx F, Kuppens P, et al. Critical slowing down as early warning for the onset and termination of depression.  Proc Natl Acad Sci USA.  2014;111(1):87–92. doi: 10.1073/pnas.1312114110. PubMed PMID: 24324144. [PMC free article: PMC3890822] [PubMed: 24324144]
132. Vezzani A, Sperk G. Overexpression of NPY and Y2 receptors in epileptic brain tissue: an endogenous neuroprotective mechanism in temporal lobe epilepsy? Neuropeptides.  2004;38(4):245–52. Epub 2004/09/01. doi: 10.1016/j.npep.2004.05.004. PubMed PMID: 15337376. [PubMed: 15337376]
133. Virtanen MA, Uvarov P, Mavrovic M, Poncer JC, Kaila K. The Multifaceted Roles of KCC2 in Cortical Development.  Trends Neurosci.  2021;44(5):378–92. Epub 2021/03/01. doi: 10.1016/j.tins.2021.01.004. PubMed PMID: 33640193. [PubMed: 33640193]
134. Weiss SA, Alvarado-Rojas C, Bragin A, Behnke E, Fields T, Fried I, et al. Ictal onset patterns of local field potentials, high frequency oscillations, and unit activity in human mesial temporal lobe epilepsy.  Epilepsia.  2016;57(1):111–21. doi: https://doi​.org/10.1111/epi.13251. [PMC free article: PMC4862409] [PubMed: 26611159]
135. Wilkat T, Rings T, Lehnertz K. No evidence for critical slowing down prior to human epileptic seizures.  Chaos.  2019;29(9):091104. Epub 2019/10/03. doi: 10.1063/1.5122759. PubMed PMID: 31575122. [PubMed: 31575122]
136. Williams PA, Hellier JL, White AM, Staley KJ, Dudek FE. Development of spontaneous seizures after experimental status epilepticus: implications for understanding epileptogenesis.  Epilepsia.  2007;48 Suppl 5:157–63. [PubMed: 17910596]
137. Wittner L, Erőss L, Czirják S, Halász P, Freund TF, Maglóczky Z. Surviving CA1 pyramidal cells receive intact perisomatic inhibitory input in the human epileptic hippocampus.  Brain.  2005;128(1):138–52. doi: 10.1093/brain/awh339. [PubMed: 15548550]
138. Woldbye DP, Madsen TM, Larsen PJ, Mikkelsen JD, Bolwig TG. Neuropeptide Y inhibits hippocampal seizures and wet dog shakes.  Brain Res.  1996;737(1-2):162–8. Epub 1996/10/21. PubMed PMID: 8930362. [PubMed: 8930362]
139. Yekhlef L, Breschi GL, Lagostena L, Russo G, Taverna S. Selective activation of parvalbumin- or somatostatin-expressing interneurons triggers epileptic seizurelike activity in mouse medial entorhinal cortex.  J Neurophysiol.  2015;113(5):1616–30. PubMed PMID: 25505119. [PubMed: 25505119]
140. Zijlmans M, Jiruska P, Zelmann R, Leijten FSS, Jefferys JGR, Gotman J. High-frequency oscillations as a new biomarker in epilepsy.  Ann Neurol.  2012;71(2):169–78. doi: 10.1002/ana.22548. PubMed PMID: 22367988. [PMC free article: PMC3754947] [PubMed: 22367988]