An official website of the United States government
NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.
Noebels JL, Avoli M, Rogawski MA, et al., editors. Jasper's Basic Mechanisms of the Epilepsies. 5th edition. New York: Oxford University Press; 2024. doi: 10.1093/med/9780197549469.003.0014
This chapter highlights recent advances in our understanding of the bidirectional interactions between sleep and epilepsy. It describes underlying neuronal correlates of sleep oscillations and epileptic activity followed by an overview of how epileptic activity during the interictal and ictal state relates to sleep. It highlights effects of sleep microstructure and sleep stability on epileptic activity, with slow oscillations and sleep fragmentation enhancing epileptic activity, and phasic rapid-eye-movement sleep having suppressive properties. Moreover, it discusses evidence of the contribution of sleep studies for improved identification of the epileptogenic zone and prediction of seizure outcome. In addition to the effects of sleep, this chapter explains how epileptic activity is influenced by longer time intervals on a day, multiday, and year scale. It then discusses how the link between epileptic activity and sleep microstructure might affect physiological functioning beyond seizures by impairing the fine-tuned orchestration of slow oscillations, spindles, and ripples necessary for memory consolidation, and reviews research addressing this question in epilepsy. Finally, the chapter provides an overview of the relationship between sudden unexplained death in epilepsy, sleep, and seizures, and discusses cardio-respiratory dysfunction as one putative mechanism. The review concludes with discussing future directions necessary to bridge current gaps in understanding, and it provides suggestions for advancing this field.
The temporal relationship between the occurrence of seizures and sleep has been anecdotally observed since the time of Aristotle (Temkin, 1971). Despite this evidence, it is only in more recent years, and particularly since the introduction of advanced neurophysiology and computational methods, that its underlying correlates and mechanisms were further elucidated. This review chapter highlights recent advances in our understanding of the reciprocal interactions between sleep and epilepsy, and it demonstrates how cutting-edge neurophysiological techniques were able to contribute to our understanding of these relationships.
After reviewing neuronal underpinnings of sleep oscillations and epileptic activity, we discuss the relationship between epileptic activity and sleep by reviewing evidence for both sleep macrostructure and microstructure, and we discuss in particular how sleep slow oscillations during non–rapid eye movement (NREM) sleep as well as unstable sleep facilitate epileptic activity (e.g., Frauscher et al., 2015a; Peter-Derex et al., 2020), and how phasic elements of rapid eye movement (REM) sleep have strong suppressant effects on epileptic activity (e.g., Frauscher et al., 2016; Campana et al., 2017). Direct cortical recordings obtained with either subcortical grids/strips or depth electrodes were key to advance this understanding, as they provide a direct window into the brain at high spatiotemporal resolution. This information is required as the scalp electroencephalogram (EEG) is typically “blind” to epileptic activity arising from deep structures or small cortical generators (von Ellenrieder et al., 2014). We further discuss the localizing properties of sleep to better identify the epileptogenic zone and predict epilepsy trajectories. Sleep was shown to add valuable information for identification of the epileptogenic zone and prediction of the course of epilepsy (e.g., Klimes et al., 2019; review article of McLeod et al., 2020). Over the last years, collection of EEG data from the same individual across weeks, months, and even years has become possible due to the development of responsive neurostimulation devices (e.g., Baud et al., 2018; Chen et al., 2021; Leguia et al., 2021). This enabled research into the topic of circadian, multidien, and circannual rhythms. The study of the effects of epilepsy on sleep beyond seizures has received increasing attention over the past decade. Current evidence suggests that a fine-tuned coupling between cortical slow oscillations, spindles, and hippocampal ripples is needed for information transfer from the hippocampus to the neocortex in order to ensure consolidation of newly acquired memories into long-term memory (sleep-related memory consolidation) (Staresina et al., 2015; Maingret et al., 2016; see review of Amantantidis et al., 2019). We review studies attempting to disentangle the effect of epilepsy on physiological sleep-related functions and vice versa, and discuss the existing knowledge on how epilepsy is able to impair memory consolidation during sleep (Lambert et al., 2020; Kramer et al., 2021). Finally, the relationship among seizures, sleep, and sudden unexplained death in epilepsy (SUDEP) will be reviewed. Disturbance of cardiorespiratory function has received increasing attention as one important putative underlying mechanism (e.g., Latreille et al., 2017; Lacuey et al., 2019). This chapter concludes with a critical discussion of current needs in the field to further advance our knowledge on how sleep is able to shape epileptic networks and vice versa.
Both physiological and pathological brain functions are executed during three main brain states: wakefulness, NREM sleep, and REM sleep. Different neuronal oscillations mediate these brain states. Therefore, we start with a description of the neuronal activity that underlies the generation of the asleep and awake states.
The command to generate various brain states originates in ascending neuromodulatory structures which release various neurotransmitters such as acetylcholine, serotonin, norepinephrine, orexin, and dopamine (Steriade and McCarley, 2005; Saper, 2006; Lee and Dan, 2012; Nevárez and de Lecea, 2019). In cortical and thalamic excitatory neurons, neuromodulators reduce K+ currents, which produce neuronal depolarization and increase neuronal firing; the opposite is true for inhibitory GABAergic cells (McCormick, 1992). Neuromodulatory activity also affects glutamatergic synaptic transmission (Gil et al., 1997; Chauvette et al., 2012). Upper brain structures, which generate sleep-wake oscillations, such as the neocortex, thalamus, or hippocampus, simply respond to a given neuromodulatory tone. Numerous studies have demonstrated that neuromodulatory systems are silent or almost silent (see Eschenko et al., 2012, for an example of exception) during slow-wave sleep. Multiple oscillations characterize slow-wave sleep with leading patterns of slow oscillations and spindles. Several studies suggest that slow-wave sleep is the default state of cortical networks (Lemieux et al., 2014; Sanchez-Vives and Mattia, 2014). Indeed, slow oscillations are present in chronically maintained cortical cultures (Hinard et al., 2012) or neocortical slices (Sanchez-Vives and McCormick, 2000). Slow oscillations also persist in isolated cortical slabs independent of the vigilance state of the animal (Lemieux et al., 2014). On the other hand, sleep deprivation leads to the generation of local slow waves in otherwise awake animals (Vyazovskiy et al., 2011). The elementary events underlying slow oscillations are slow waves. At the level of local field potentials (LFPs) and intracellular activity, slow waves are composed of a depth-positive wave, mediated by neuronal hyperpolarization and silence (down state) followed by a depth-negative wave, mediated by neuronal depolarization, a high level of synaptic activity, and neuronal firing (up state, Fig. 14–1a, c). During REM sleep or the awake state, LFPs are activated and the neurons are relatively depolarized, showing intense synaptic events and firing of action potentials (Fig. 14–1a). Detailed intracellular studies demonstrated that neuronal depolarization of cortical neurons is typically similar and in the order of –62 mV during slow-wave sleep up states, quiet wakefulness, and REM sleep (Steriade et al., 2001; Timofeev et al., 2001a; Chauvette et al., 2010).
Local field potentials (LFPs) and intracellular activity of the neocortex during slow-wave sleep, wake, and seizures. a. LFP and intracellular recording from a cortical neuron in transition from slow-wave sleep to wakefulness in a cat. b. LFP and intracellular (more...)
Another major type sleep activity is the sleep spindle. Sleep spindles are characterized by waxing and waning oscillations with a frequency between 10 and 16 Hz and a duration between 0.5 and 2 s (Loomis et al., 1935; Gibbs and Gibbs, 1950; Jankel and Niedermeyer, 1985). Spindles are generated in the thalamus with the reticular thalamic nucleus having a leading role (Steriade et al., 1985; Bazhenov et al., 1999). At the beginning stages of this activity, only the reticular thalamic nucleus fires. As thalamocortical cells start to fire rebound neuronal spike-bursts, spindle oscillations are transmitted to the cortex and can be detected at the cellular or field potential levels (Timofeev et al., 2001b). Although spindles can be generated in isolated thalamic preparations (von Krosigk et al., 1993; Timofeev and Steriade, 1996), cortico-thalamic feedback exerts a strong control on spindle onset and termination (Timofeev et al., 2001b; Bonjean et al., 2011). Manipulation of postsynaptic inhibition slows spindle oscillations and generates activity resembling electrographically absence seizures (Avoli et al., 1983; von Krosigk et al., 1993).
Ketamine-xylazine anesthesia induces slow oscillations. It is considered a good model of slow-wave sleep, thus allowing reliable intracellular recordings to be performed (Contreras and Steriade, 1995; Volgushev et al., 2006; Chauvette et al., 2010). Under this anesthesia, nearly all cortical neurons are involved in the slow oscillation and display regular up-down state transitions (Fig. 14–1d). The main differences between slow oscillations under ketamine-xylazine anesthesia and slow-wave sleep are higher rhythmicity, longer duration of down states, and more uniform distribution of slow-wave activity over the cortex in anesthetized animals (Chauvette et al., 2011). Cats anesthetized with ketamine xylazine anesthesia often develop electrographic seizures composed of spike-wave/polyspike-wave discharges at a frequency of 1.5–3 Hz and runs of fast EEG/LFP spikes (10–15 Hz, Fig. 14–1b, e) (Steriade et al., 1998; Timofeev and Steriade, 2004; Timofeev, 2010). In such patterns, runs of fast spikes lasting longer than 2 sec typically correspond to tonic seizures and spike/polyspike-wave complexes to clonic seizures (Niedermeyer, 2005).
It is well accepted that certain types of neocortical focal epilepsy primarily occur nocturnally (Proserpio et al., 2011; Timofeev et al., 2014; Gibbs et al., 2019). Neocortical seizures evolve from slow oscillations without discontinuity (Timofeev and Steriade, 2004). As shown in Figure 14–1b, slow oscillations progressively transform into polyspike-wave discharges. Spike-wave/polyspike-wave discharges share several features with cortical slow oscillations. Similar to sleep slow waves or slow waves induced by ketamine-xylazine anesthesia, during paroxysmal waves, cortical neurons are hyperpolarized and do not show synaptic noise, while during depth-negative waves of sleep/anesthesia and EEG-LFP neuronal firing, neurons are depolarized and fire action potentials (compare c, d, and e in Fig. 14–1). The main difference is that during seizures, the depolarizing component is much stronger, often resulting in depolarizing block and thus inactivation of neuronal firing. Because major neuronal states during slow-wave sleep, anesthesia, and spike-wave components of seizures are similar, the major behavioral manifestation, namely absence of conscious perception, is similar across all three conditions.
There is a tight relationship between epileptic activity in the interictal and ictal state and the various stages of vigilance. A meta-analysis comprising 42 EEG-based studies analyzed a total of 1458 patients (Ng & Pavlova, 2013). Indexed to duration, the authors showed that the highest rates of epileptic activity are present during NREM sleep, whereas lowest rates were found during REM sleep. The study further subdivided distribution of epileptic activity based on the epilepsy type (generalized versus focal). Similar distributions were found between people with generalized and focal epilepsy, with highest rates of epileptic activity occurring during NREM sleep and lowest rates during REM sleep. Temporal lobe epilepsy showed enhanced epileptic activity during slow-wave sleep (Fürbass et al., 2021). In generalized epilepsies, interictal activity was absent during REM sleep, whereas in focal epilepsy it was rare (Ng & Pavlova, 2013). Experimental research showed that this increase is linked to hypersynchronization during NREM sleep that is achieved via the thalamocortical circuitry as a facilitator of epileptic activity mediated by GABAergic mechanisms (see review of Steriade, 2005). In contrast, the decrease in epileptic activity during REM sleep is thought to be due to EEG desynchronization mainly mediated by cholinergic neurotransmission (Shouse et al., 1989). In a feline epilepsy model, atropine, an antagonist of muscarinic cholinoreceptors, was shown to abolish EEG desynchronization during REM sleep and resulted in an increase in interictal epileptic discharges and seizures (Shouse et al., 2000). As well as being less frequent during REM sleep, epileptic activity was shown to be more focally restricted during REM sleep. In contrast, it is more widespread during NREM sleep, where additional foci can become apparent (Sammaritano et al., 1991). An interesting study in the pentylenetetrazol rat epilepsy model further underlined these mechanisms; it showed that deep brain stimulation of the unilateral anterior thalamic nucleus increased the seizure threshold by augmenting the proportion of REM sleep and decreasing the progressive enhancement of delta power during NREM sleep (Tseng et al., 2020).
Despite these general tendencies, more recent work has demonstrated the existence of a location-dependent effect regarding the modulation of epileptic activity by sleep. A recent invasive intracranial EEG study showed that the modulation of epileptic activity by sleep is particularly prominent in the mesiotemporal lobe (Lambert et al., 2018). Moreover, this increase in interictal epileptic activity in the mesiotemporal lobe is locally confined, whereas neocortical structures show a clear propagation effect of epileptic activity during NREM sleep. This may be related to the fact that connections between the dorsal midline thalamic nuclei sending widespread thalamo-cortical afferents and the mesiotemporal structures are more sparse, and not reciprocally connected with the reticular nucleus (Jones, 2007). Interestingly, the modulation of epileptic activity was similar for structures inside and outside the seizure-onset zone, with higher rates of epileptic activity inside than outside the seizure-onset zone irrespective of the stage of vigilance (Bagshaw et al., 2009; Lambert et al., 2018). Furthermore, sleep was shown to be an important modulator of pathological EEG patterns such as those present in focal cortical dysplasia, an example of a highly epileptogenic lesion (Menezes Cordeiro et al., 2015). Indeed, despite the presence of an almost continuous discharge, the epileptic EEG patterns of focal cortical dysplasia were influenced by the thalamo-cortical control mechanisms involved in the generation of sleep.
Seizures show a specific distribution across the sleep-wake cycle that is dependent on the type of epilepsy (generalized vs. focal) and in case of focal epilepsy the location of the epileptic focus (see review of Frauscher & Gotman, 2019). Regarding generalized epilepsies, it is particularly epilepsy with generalized tonic-clonic seizures alone (former epilepsy with grand mal seizures at awakening) and juvenile myoclonic epilepsy that are tightly linked to sleep, with lack of sleep as a clear provoking factor (David, 1955), and seizures occurring usually in the first 2 hours after awakening in the morning (Trenité et al., 2013). Regarding focal epilepsies, studies showed that depending on the lobe of origin, focal seizures occur preferentially at different times of the day: frontal lobe seizures occur mostly out of sleep, with an early morning peak; mesiotemporal seizures were shown to have two diurnal peaks, morning and late afternoon (also true when evaluated in a noncircadian environment); while occipital seizures peak in the early evening and rarely occur during sleep. Peaks of parietal seizures are less clear (Herman et al., 2001; Mirzoev et al., 2012; Durazzo et al., 2008; Karafin et al., 2010; Nzwalo et al., 2016; Spencer et al., 2016). Similar to humans, circadian clustering of seizures has been observed in animal models of epilepsy such as the pilocarpine mesiotemporal lobe epilepsy mouse model (Pitsch et al., 2017).
High-frequency oscillations (HFOs), comprised of ripples (80–250 Hz) and fast ripples (>250 Hz), are a promising biomarker of the epileptogenic zone (see review of Frauscher et al., 2017). Despite this potential, ripples have also been reported in normal cortical areas not affected by epilepsy (Axmacher et al., 2008; Nagasawa et al., 2012; Melani et al., 2013; von Ellenrieder et al., 2016; Frauscher et al., 2018; Vaz et al., 2019). Under physiological conditions, HFOs have been related to memory consolidation, motor, somatosensory and visual processes (see review of Thomschewski et al., 2019), and memory retrieval (Vaz et al., 2019). HFOs are likely generated by multiple, possibly not exclusive, mechanisms occurring at the cellular and network level, with interneurons playing a complex role (see reviews of Jefferys et al., 2012 and Jiruska et al., 2017). HFOs are modulated across the sleep-wake cycle: rates of HFOs are highest during NREM sleep and lowest during REM sleep in keeping with what we know from epileptic spikes (Grenier et al., 2001; Staba et al., 2004; Bagshaw et al., 2009; Dümpelmann et al., 2015; von Ellenrieder et al., 2017; Al-Bakri et al., 2018). HFO rates in seizure-onset contacts differed from non-seizure-onset contacts independent of the sleep-wake cycle (Bagshaw et al., 2009; Dümpelmann et al., 2015). Moreover, similar to region-specific differences in HFO rates (Frauscher et al., 2018), it was shown that sleep had a different modulating impact depending on brain anatomy; there were no changes observed across the sleep-wake state for the frontal lobe in contrast to the temporal, parietal, and occipital lobes (Dümpelmann et al., 2015). Additionally, HFOs were more widespread during NREM sleep and more focally restricted during REM sleep (von Ellenrieder et al., 2017). Finally, the ratio of suppression of HFOs during REM sleep as opposed to NREM sleep has been shown to be related to localization of the epileptogenic zone and seizure outcome; better outcome was demonstrated when areas with less suppression of ripples during REM sleep versus NREM sleep were included in the postsurgical resection cavity (Sakuraba et al., 2016). This latter might be explained by the fact that the more epileptogenic the tissue is, the less it follows what is expected for physiological activity during REM sleep. Indeed, recent work showed that HFOs generated in the normal cortex were higher in phasic compared to tonic REM sleep, whereas HFOs generated in the epileptic cortex showed the inverse, with lowest values in phasic compared to tonic REM sleep (Frauscher et al., 2016, 2020). A similar pattern was also observed for fast ripples that are predominantly epileptic in nature (Staba et al., 2004; Frauscher et al., 2016).
Specific epilepsy syndromes are tightly linked to sleep. For the relation of childhood-related epilepsy syndromes and sleep, such as epilepsy with continuous spike and waves during slow-wave sleep, self-limiting epilepsy with centro-temporal spikes, Panyotopoulos syndrome, please refer to the clinical review of Wu et al. (2021). Here we will briefly discuss sleep-related hypermotor epilepsy (SHE), formerly nocturnal frontal lobe epilepsy, as research into this syndrome underwent the most developments in the last decade (Provini et al. 1999; Brain 1999; Gibbs et al., 2019). SHE is a focal epilepsy syndrome characterized by hypermotor seizures occurring predominantly in clusters during non-REM sleep (Tinuper et al., 2016). The definition of this entity was modified at a Consensus Workshop in 2016 (Tinuper et al., 2016). In approximately 70% of cases, seizures arise from the frontal lobe and in the remaining 30% from the extra-frontal lobe, including areas such as the insula, temporal lobe, or parietal cortex (Gibbs et al., 2019). Interestingly, the originally described underlying genetic mutation is known to affect the nicotinic acetylcholine receptor (Steinlein et al., 1995). Activation of the frontal cortex with acetylcholine is essential to generate arousals, and with SHE, an overactive cholinergic system might result in abnormal arousals as a possible mechanism for triggering sleep-related seizures in this syndrome. In this context it is noteworthy that patients with SHE also have frequent comorbid NREM parasomnias as a disorder of the sleep-wake regulation (Bisulli et al., 2010). A systematic investigation of the presence of NREM arousal parasomnias in the personal and family histories of subjects with SHE found this comorbidity to be present in 34% and 39% of cases (Provini et al., 1999). In addition, significant polysomnographic alterations such as increased REM latency, sleep fragmentation, and increased cyclic alternating pattern rate as measures of unstable sleep were observed. The authors concluded that within NREM sleep, unstable sleep represents a powerful predisposing condition for the occurrence of nocturnal motor seizures (Parrino et al., 2012).
Most studies focused predominantly on the relationship between epileptic activity and sleep on a macrostructural level. Findings remain ambiguous with evidence of a general trend of more awake time in temporal lobe epilepsy patients (see meta-analysis of Sudbrack-Oliveira et al., 2019). Only more recently, studies were undertaken to investigate the relationship between sleep microstructure and epileptic activity. It was shown that it is in fact sleep slow oscillations and not NREM sleep itself that is associated with increased rates of interictal epileptic activity and high-frequency oscillations (Frauscher et al., 2015a; von Ellenrieder et al., 2016). This is likely explained by the presence of hypersynchronization during these states. In contrast, it was shown that it is REM sleep with rapid eye movements (phasic REM sleep) as opposed to REM sleep without rapid eye movements (tonic REM sleep) which shows the most suppression of epileptic activity during sleep (Frauscher et al., 2016; Campana et al., 2017). Both studies point to the importance of EEG desynchronization as assessed with both power and measures of synchrony, and they confirm findings from the animal model that this desynchronization is mainly the result of cholinergic mechanisms that are most pronounced during phasic REM sleep (Shouse et al., 2000). In line with these findings, cholinergic transmission is higher during phasic than tonic REM sleep (Siegel, 1979; Szymusiak et al., 1989; Kim and Jeong, 1999). When looking at the phase of the slow wave with respect to epileptic activity, it was found to be associated with the transitions from the up to the down state, whereas that of physiological ripples was found at the transition from the down to the up state (Frauscher et al., 2015a; von Ellenrieder et al., 2016). Similar results were found in the following years using various approaches such as analysis of the modulation index or phase amplitude coupling between high- and low-frequency oscillations (Nonoda et al., 2016; Song et al., 2017; Iimura et al., 2018; Motoi et al., 2018; Samiee et al., 2018; Dickey et al., 2022). Amiri et al. (2016) also found that phase-amplitude coupling between high and low frequencies was highest in stage N3 and lowest in REM sleep, and it was higher in the seizure-onset zone than in other regions.
Spindles are distinct EEG events which are the hallmark of NREM sleep stage 2. Recent research points to the importance of spindles for memory consolidation, cortical development, and sleep stability (Khazipov et al., 2004; Schabus et al., 2004; Dang-Vu et al., 2010a; Fogel and Smith, 2011; review of Rasch and Born, 2013). For the mesiotemporal lobe it was shown that there is an inverse relationship between the rate of hippocampal spikes and sleep spindles, with a higher degree of spiking being associated with lower rates of sleep spindles (Frauscher et al., 2015b). One possible explanation for this is Pierre Gloor’s hypothesis, which suggests that spike-wave discharges of generalized absence seizures are generated by the same thalamocortical circuits as physiological spindles but occur in the presence of hyperexcitable and hyperresponsive cortical neurons (Gloor, 1978; von Krosnigk et al., 1993; review of Kostopoulos, 2000). Similarly, hippocampal spikes could be facilitated by thalamo-hippocampal activity which in the normal hippocampus triggers spindles. Alternatively, it might be possible that spikes have an occluding effect on spindle activity. With respect to this latter possibility, there are also studies showing that epileptic spikes in the hippocampus represent a pathological transformation of the physiological sharp-wave-ripple complexes (see review of Gulyas and Freund, 2015). Furthermore, it was shown that hippocampal spikes evoke remote sleep spindles in the prefrontal cortex, and that interictal epileptiform discharges shape large-scale intercortical communication (Gelinas et al., 2016; Dahal et al., 2019). Alternatively, it may also be the case that slow waves evoke both interictal epileptiform discharges and sleep spindles (von Ellenrieder et al., 2020).
The direct influence of epileptic activity on sleep stability remains poorly understood. What is known from studies in both focal and generalized epilepsies in the scalp is that epileptic discharges are linked to more unstable sleep as expressed by the cyclic alternating pattern, an EEG measure of sleep stability (see review of Parrino et al., 2006). Studies in the scalp EEG showed that epileptic activity was predominantly coupled to unstable sleep. Indeed, a recent study using invasive intracranial EEG showed in addition to this finding that it is not only seizures (including paucisymptomatic or pure EEG seizures) that result in arousals but also isolated interictal epileptic discharges (Peter-Derex et al., 2020). Systems important for wakefulness are the reticular activating system, the basal forebrain, and the hypothalamus (review of Saper et al., 2010). These structures receive descending projections from the frontotemporal cortices presenting a cortical feedback system (see book chapter of Afifi and Bergmann, 2005). In addition, epileptic discharges are associated with a release of excitatory neurotransmitters, which exert an additional activating effect (Devinsky et al., 1992).
Regarding isolated interictal epileptic discharges, a bidirectional relationship was observed with increased rates of spikes in the prearousal period that persisted during the beginning of the arousal in the neocortex but not the mesiotemporal lobe before returning to baseline. This region-specific effect of arousal patterns could explain the strong activating effect of frontal discharges during arousals, given the prominence of slow waves during arousals in the frontal lobe (Peter-Derex et al., 2015). Indeed, the increase in spiking was directly related to the amount of delta power present during the arousal itself (Fig. 14–2). Whether this presents a causal relationship or an effect of an unknown underlying third factor awaits further investigation possible only by directly manipulating sleep stability.
Peri-arousal and intra-arousal dynamics of EEG epileptic spiking activity. (A) Intra-arousal dynamics of epileptic spikes. The arousal is preceded by an increase in epileptic activity in the orbitofrontal cortex (spikes and low voltage fast activity, (more...)
Although interictal epileptic discharges are rare in REM sleep (Ng & Pavlova, 2013), REM sleep has emerged as a promising sleep-wake state to identify the seizure focus, as epileptic phenomena captured in REM may spatially correspond more closely to the epileptogenic zone (Lieb et al., 1980; Montplaisir et al., 1987; Sammaritano et al., 1991; Ochi et al., 2011; Yuan & Sun, 2020). It has been postulated that, given the desynchronization present during this stage of vigilance (Shouse et al., 2000), it requires a more pathological tissue to overcome the suppressing properties of this stage of sleep on epileptic activity (McLeod et al., 2020). However, there are a few studies that show that false localization can occur during REM sleep (Singh et al., 2014), and it is well known that many patients have no interictal epileptiform discharges during REM sleep as assessed with scalp EEG (Ng & Pavlova, 2013). Therefore, the utility of REM sleep for identification of the epileptic focus is limited to those patients exhibiting interictal epileptiform discharges during REM sleep. There is a single study using a quantitative multi-feature supervised machine-learning approach for investigation of which state of vigilance is best suited for identifying the epileptic focus. Klimes et al. used a multi-feature approach for identification of the epileptogenic zone in 30 consecutive patients undergoing invasive intracranial EEG; the authors found that NREM sleep was the state of vigilance that allowed best identification of the epileptogenic zone (Klimes et al., 2019). Whether this apparently contradictory finding might in fact be related to the total number of detected interictal epileptic events that are evidently highest in NREM sleep and statistical considerations awaits further confirmation.
Apart from the potential of sleep to be useful for localizing the epileptic focus, there is a first study showing that the coupling of HFOs to slow waves is able to predict the seizure-onset pattern in mesiotemporal lobe epilepsy (Amiri et al., 2019). Coupling of HFOs to slow waves had good discriminatory properties to predict either low-voltage fast activity or low-frequency high-amplitude periodic spiking. In the case of low-voltage fast activity, HFOs occurred mostly at the peak or the transition of the peak to trough, whereas in low-frequency, high-amplitude periodic spiking, HFOs occurred mostly during the transition of the trough to peak (Amiri et al., 2019).
Finally, recent work showed that the presence of epileptic spikes in REM sleep during scalp video-EEG monitoring is a predictor of poor outcome trajectories (McKenzie et al., 2020). Controlling for confounding variables, the authors found that for every 1% increase in REM spike burden, the peak seizure frequency increased by 1.69 more seizures per month. This may be explained by a more severe underlying epilepsy whose ictal and interictal activity more often breaks through the typically suppressive effect of REM against epileptic activity (McKenzie et al., 2020). Whether or not this also has implications on surgical outcome awaits further investigation.
Cyclic occurrence patterns of seizures have been observed via anecdotal observation, chart studies obtained from institutionalized epilepsy patients, and seizure diaries, but it is only more recently that technical advances have enabled characterization of cycles of seizure occurrence with direct brain recordings in both humans and animals over prolonged time periods (see review of Karoly et al., 2021). Results of this work provided evidence of the existence of cycles of epileptic brain activity (seizures, interictal epileptic discharges, HFOs) over diverse time scales following a daily (circadian), multiday (multidien), and even yearly (circannual) rhythmicity across various species (Anderson et al., 2015; Karoly et al., 2016; Spencer et al., 2016; Baud et al., 2018, 2019; Gliske et al., 2018; Chen et al., 2021, Leguia et al., 2021). Capitalizing on data of the Neuropace trial, Baud et al. showed that multidien periodicities, most commonly 20–30 days in duration, are robust and relatively stable for up to 10 years (Baud et al., 2018). A longitudinal study in 222 subjects analyzing results of the same long-term study found circadian, multidien, and circannual rhythms in 89%, 60%, and 12% of study participants, respectively, with a stronger signal for circadian and multidien rhythms compared to circannual rhythms (Leguia et al., 2021). The authors concluded that circannual cycles likely play a relatively minor role compared to circadian or multidien cycles. Using data from 15 patients participating in the NeuroVista trial, Cook et al. (2013) analyzed the behavior of HFOs and epileptiform spikes; they found that rates of HFOs and interictal epileptiform discharges recorded immediately after implantation of electrodes do not reflect the long-term behavior after recovery from surgery. Moreover, they showed that HFO rates fluctuate with periodicities of a duration similar to that of seizures (Chen et al., 2021). A recent retrospective study provided proof of principle of the clinical significance of this finding; scheduling monitoring times based on personalized seizure risk forecasts was shown to improve the yield of hospitalizations for video-EEG monitoring for presurgical workup (Karoly et al., 2021).
Traditionally, sleep-related seizures are interpreted to be less relevant and problematic than seizures arising during wakefulness. This notion is, however, incorrect if one considers the impairment of important physiological functions of sleep. Recent work suggested that even pauci-symptomatic or subclinical pure EEG seizures have deleterious effects on sleep structure, resulting in arousals or awakenings (Peter-Derex et al., 2020). Moreover, there is some evidence showing that spikes are, in fact, not innocuous (Kleen et al., 2013), and a few studies suggest that isolated interictal epileptiform discharges have deleterious effects on sleep structure and function (Lambert et al., 2020; Peter-Derex et al., 2020), in contrast to what was expected. This is interesting as one important function of sleep is to stabilize memories (memory consolidation) (Clemens et al., 2007; Fogel et al., 2011; review of Buzsaki and Silva, 2012; review of Rasch and Born, 2013; Staresina et al., 2015; review of Amantatidis et al., 2019). Many patients with epilepsy indeed also suffer from memory dysfunction (Hermann et al, 2006; Holmes, 2015); hence, it is tempting to speculate that this could be at least in part explained by alterations of physiological sleep rhythms by epileptic activity. Indeed, several recent studies based both on scalp and intracranial EEG are pointing in this direction (Gelinas et al., 2016; Lambert et al., 2020, Kramer et al., 2021; Lachner-Piza et al., 2021). A first study in drug-resistant focal epilepsy patients capitalizing on invasive intracranial EEG showed that the rate of epileptic spikes and seizures during sleep is inversely correlated to long-term memory potentiation (Lambert et al., 2020). Potential explanations could be disruption of physiological sleep rhythms by epileptic activity. From experimental research, it is well known that a precise spatiotemporal coupling between slow waves, sleep spindles, and ripples in the hippocampus is required for memory consolidation to occur during sleep (Amantatidis et al., 2019). Several studies in both adults and children with epilepsy showed that reduced sleep spindle rates are associated with decreased memory performance (Kramer et al., 2021). Prior work has observed that the thalamocortical circuits that generate spindles can be hijacked to support epileptiform spikes as one potential explanation of the above-described finding (Steriade, 2005; Beenhakker and Huguenard, 2009; Clemente-Perez et al., 2017). Nevertheless, in humans, only a few studies prospectively assessed sleep-related memory consolidation, with conflicting findings. Some authors could confirm impaired sleep-related memory consolidation in epilepsy (Urbain et al., 2011; Galer et al., 2015), whereas others found no effect of sleep (Deak et al., 2011; Fitzgerald et al., 2013; Moroni et al., 2014; Sud et al., 2014; Storz et al., 2020), and other groups even showed a benefit of sleep similar to what is known in healthy subjects (Atherton et al., 2014, 2016; Sarkis et al., 2016; Chan et al., 2017). These conflicting results are probably best explained by differences in methodology and heterogeneous patient samples therefore precluding any definitive conclusions on the alteration of sleep-related memory consolidation in epilepsy.
SUDEP is the leading cause of death in patients with refractory epilepsy (Devinsky et al., 2016). It is defined as the sudden, unexpected death of a person with epilepsy with or without evidence of a seizure and in whom postmortem examination does not reveal a structural, infectious, or toxicological explanation for the death (Nashef et al., 2012). Suspected pathophysiological mechanisms include seizure-induced cardiac and respiratory dysregulation, suggesting seizure spread to the brainstem regions. SUDEP frequently occurs at night and has been linked to sleep. A recent meta-analysis showed that 69% of 1025 cases gathered from 67 studies occurred during sleep (Ali et al., 2017). Sleep in the prone position was identified to be an important risk factor similar to sudden infant death syndrome (Liebenthal et al., 2015). The importance of respiratory compromise during sleep is supported by the occurrence of acute peri-ictal apnea. The MORTEMUS study evaluated cardiorespiratory arrests encountered in 147 epilepsy monitoring units worldwide (Ryvlin et al., 2013). Cardiorespiratory data, available for 10 cases of SUDEP, showed a consistent and previously unrecognized pattern where rapid breathing developed after secondary generalized tonic-clonic seizures, followed within 3 min by transient or terminal cardiorespiratory dysfunction (Ryvlin et al., 2013). Interestingly, in contrast to seizures occurring out of wakefulness, sleep-related seizures were more frequently associated with more severe and longer hypoxemic events, and more frequently followed by paroxysmal generalized EEG suppression (Latreille et al., 2017). The amygdala might be a key anatomical structure, as electrical stimulation applied to awake patients with epilepsy demonstrated that electrical stimulation of the amygdala causes cessation of respiration that went unnoticed by the patients (Nobis et al., 2018; Lacuey et al., 2019). Similar to humans, experimental models of epilepsy in which animals experience high rates of death have comorbid sleep or circadian disruption (Li & Buchanan, 2019). In light of this, the growing literature exploring the genetic components of the biological clock shows that clock abnormalities are common in experimental epileptic animals (Li & Buchanan, 2019). Experimental research showed that reducing clock expression in excitatory neurons increases behavioral activity and leads to spontaneous seizures (Li et al., 2017). This might at least in part explain the association between SUDEP, seizures, and sleep. Finally, although this is rarer, there is evidence that seizures during REM sleep may confer disproportionate mortality. Acute seizures induced via maximal electroshock are associated with seizure-induced respiratory arrest; all seizures induced during REM sleep were fatal in this model (Hajek et al., 2016; Purnell et al., 2017). Other postulated mechanisms of SUDEP include brainstem spreading depolarization (Aiba and Noebels, 2015), peri-ictal breathing dysfunction (Lacuey et al., 2018; Villela et al., 2019, 2019), structural changes of the brain (Wandschneider et al., 2015; Allen et al., 2017, 2019; Ogren et al., 2018), and brainstem serotonergic neuronal depletion (Patodia et al., 2018).
Results from sleep studies have shown that it seems to be hypersynchronization that triggers epileptic activity, and desynchronization that suppresses epileptic activity. Hence it is intriguing to speculate whether new therapeutic agents for epilepsy should, in fact, capitalize on these mechanisms. In addition, there is the question of whether sleep consolidation could result in reduction of epileptic activity by stabilizing sleep microstructure. Recent research suggests that it is not only seizures but also isolated interictal epileptic discharges that can result in arousals, and conversely that arousals themselves seem to evoke epileptic activity (Peter-Derex et al., 2020). In order to investigate causality of these relationships, prospective interventional studies are needed to evoke arousals and hence study whether epileptic activity will indeed increase when inducing arousals. Also, drugs promoting sleep stabilization or enhancing REM sleep bear the promise to reduce epileptic activity. In this context, dual hypocretin receptor antagonists approved for the treatment of primary insomnia might be an interesting therapeutic agent, given their sleep-stabilizing properties and REM-enhancing effects (Ng, 2017; Rosenberg et al., 2019). Chronic intracranial EEG recordings over prolonged time periods revealed modulation of seizure activity across multiple time scales. A renaissance of chronotherapy in epilepsy is hence another interesting avenue of research, given that seizures seem to be associated with certain moments in time. Uncovering the mechanistic basis for seizure cycles, particularly the factors that govern multidien periodicity, will be a major focus of future work. Animal models will certainly be helpful to answer this question. Although there are studies pointing to the presence of impaired sleep-related memory consolidation in epilepsy (Urbain et al., 2011; Galer et al., 2015), the evidence is still conflicting, and future work in this area will ultimately answer this question and elucidate whether sleep-related memory consolidation is indeed impaired in epilepsy. To do so, studies will have to be prospective, well powered, use sleep-sensitive memory tasks, and assess the load of epileptic activity during learning, sleep, and testing. Furthermore, preliminary data are now available from human studies, which suggests that impaired orchestration between slow waves, spindles, and ripples might present one important factor for impaired sleep-related memory consolidation (Gelinas et al., 2016). Further research will be needed to investigate if treatment of epileptic activity is able to counteract the hijacking of slow-wave-spindle-ripple coupling and if noninvasive stimulation procedures that enhance sleep patterns might be able to improve memory consolidation in epilepsy. Finally, it will be important to improve our understanding of the mechanisms of SUDEP in order to identify prevention strategies. It will be key to further explore the role of dysregulated breathing during sleep in its generation.
The authors would like to thank Sana Hannan, PhD, postdoctoral fellow in the ANPHY lab, for her help with carefully proofreading the manuscript. B. F.’s salary is supported by a salary award (“Chercheur-boursier clinicien Senior”) for 2021–2023 by the Fonds de Recherche du Québec—Santé. B. F.’s laboratory was supported by NSERC, CIHR, CFI, start-up funding from the Neuro, and the Hewitt Foundation. I. T.’s laboratory is currently supported by NSERC and NIH.
The authors declare no relevant conflict related to this work.
This is an open access publication, available online and distributed under the terms of a Creative Commons Attribution-Non Commercial-No Derivatives 4.0 International licence (CC BY-NC-ND 4.0), a copy of which is available at https://creativecommons.org/licenses/by-nc-nd/4.0/. Subject to this license, all rights are reserved.
Your browsing activity is empty.
Activity recording is turned off.
See more...
