The concept of sleep homeostasis emerged from studies on sleep regulation by Borbély [1]. This on-demand model connected sleep regulation with activity from the preceding day, thereby confirming an exponential relationship between the duration of wakefulness and the power of sleep slow waves (0.75–4.0 Hz) in the subsequent sleep period. In other words, the longer the previous waking time, the greater the sleep slow wave power. The term “duration of wakefulness” includes daytime activities, i.e., synaptic use. This model is supported by studies showing that sleep deprivation exponentially increased slow wave power in the subsequent sleep period [2, 3].

This model involved the notion of “use-dependent” regulation [4], which is best seen in the realms of slow oscillation (SO) on the descending slopes of the first sleep cycles- and prevailing in the frontal lobes and the left hemisphere. This SO sweeps posteriorly from the frontal to the occipital lobes [5].

Studies in anesthetized animals revealed an additional cortical SO ($\le$1 Hz) characterized by peculiar “ups and downs” on electroencephalography (EEG) and reflecting the synchronized membrane polarization changes of pyramidal cells and interneurons [6, 7, 8]. The up-state (depolarization) envelops several faster rhythms- sleep spindles and ripples- and rich synaptic traffic near the waking level. The hyperpolarized (down) state does not contain any unit or synaptic activity and is a “black hole”, also termed “disfacilitation” [9]. Similar up and down states of SO were also found in naturally sleeping animals [10]. This rhythmicity is assumed to have an essential role in sleep plasticity functions.

There is abundant evidence for the importance of slow wave activity. However, what is the function of slow waves that are strongly protected by homeostatic forces?

Tononi and Cirelli [11, 12] have elaborated upon their popular “synaptic homeostatic hypothesis”, which postulates that synaptic facilitation builds up (upregulates) throughout the day and is then downregulated during sleep. An alternative concept is that cortical synapses are depressed (downregulated) by waking activities and upregulated by sleep SO via silent (down) states [13].

Another level of non-rapid eye movement (NREM) sleep oscillation is an EEG micro-cyclicity termed the cyclic alternating pattern (CAP) [14]. The CAP has recently been linked to brain structures [15].

Activity within the frame of CAP reflects an ongoing spontaneous cortical activity during NREM sleep. This is characterized by periodic EEG activity (CAP and non-CAP periods) that recurs with a frequency of up to one minute. The existence of CAP reflects the sleeping brain’s flexible adaptation to external and internal conditions. CAP patterns are consistent with micro-arousals manifesting as two types, depending on and indicative of the actual balance between sleep-promoting and arousal forces. A well-known arousal pattern (A3 in CAP terms) emerges at low homeostatic pressure and is characterized by the attenuation of EEG activity with fast rhythms, muscle and autonomic changes. When the homeostatic pressure is high, such as in the deepening periods (descending slopes) of the first NREM sleep cycle, a paradoxical (anti-arousal) slow wave response (CAP A1) occurs [16] (Fig. 1).

Fig. 1.

Schematic representation of the distribution of different cyclic alternating pattern (CAP) responses for the descending (D) and ascending (A) slopes of sleep cyclicity. On the D slope, a phasic auditory input evokes phasic slow wave responses (CAP A1 phases, upward arrows) that provide instant homeostatic “delta injections”. On the A slope, the response is similar to the known electroencephalography (EEG) + autonomic arousal response (A2 and A3 phases). The upper insert shows the distribution of CAP phases during a typical night’s sleep. Black bars on the horizontal line represent rapid eye movement (REM) sleep phases. (A1 = slow wave EEG response. CAP A2–3 response = traditional arousal) Note the exponential decay of A1 responses that follow the decrease in homeostatic pressure, in contrast to A2 and A3.

Reactive CAP A1 slow waves provide immediate homeostatic protection [17]. The continuous slow waves of deep slow wave sleep show a prolonged and delayed homeostatic response to synaptic up-scaling (exhaustion) depending on previous use. In contrast, the phasic slow wave reactions of CAP A1 occur instantly as slow wave “injections”. Both types reimburse slow waves’ amount and prevail over the frontal lobes.

Plasticity is the general capability of the brain’s neural elements, from synapses to networks, to modify (up- or down-regulate) their response in reaction to previous activity in the form of long-term potentiation (LTP) or depression (LTD). LTP is defined as a persistent increase in signal transmission between two neurons by synapses, as observed by recent patterns of activity [18], whereas LTD is the opposite. In other words, as defined by Steriade [9], an “activity-dependent alteration in the strength of connections among neurons, through which information is stored”.

Plasticity and epilepsy are closely related but differ in their effects on the brain. Plasticity is a building force, whereas epilepsy is a disfigured and exaggerated caricature that distorts the affected functions.

It appears that an intrinsic aim of cortical networks is to achieve a slow wave “state”. Interestingly, isolated slabs and even cortical cell cultures follow this rule, thus hinting at homeostatic regulation [19, 20].

Cortical deafferentation has been shown to lead to a compensating (homeostatic) upregulation of neural excitability. Greater damage induces stronger upregulation that may result in epileptic excitation. Thus, while homeostatic force can neutralize a small amount of harm, greater damage needs stronger compensation. This may result in derailment of paroxysmal activity, as seen in post-traumatic epilepsy [21, 22].

The isolated cortex and hippocampus “operate extremely close to the transition point between a quiescent state and an abnormally active epileptic state” [9, 23, 24]. Wherever plastic processes occur, an epileptic shift is close by.

Goddard and Douglas [25] first proposed that the plastic process of the memory trace (engram)-formation is similar to epileptogenesis.

LTP is the elementary model of long-term plasticity and the basis of kindling, with daily electrical stimulation resulting in hyperexcitation and the development of an epileptic process leading to spontaneous seizures [26, 27]. Similarly, experimental epileptic foci that bombard distant regions with spikes may establish secondary spike foci that subsequently become independent. Such secondary cortical spots “learn to be epileptic” due to the plastic and potentially progressive runs induced by recurrent interictal epileptiform discharges (IEDs) [28, 29].

The concept of system epilepsy considers epilepsy to be a network disease. This concept may overcome the untenable dichotomy between “focal” and “generalized” epilepsies (Fig. 2). and reveal that epileptogenesis has a common molecular and electrophysiological mechanism that originates from the derailment of normal brain functioning in systems that are most liable to plasticity [30, 31, 32, 33, 34, 35]. According to this concept, the differences between epilepsies lie only in the localization and specificity of the affected systems. Concerning seizure-triggers, besides accepted reflex epilepsies we incorporated seizures initiated by the activation of the affected physiological brain system (like falling asleep, or arousal).

Fig. 2.

Schematic illustration of the system epilepsy concept, wherein the epilepsy is built upon physiological systems that share a common process of epileptogenesis.

Medial temporal lobe epilepsy (MTLE) may be considered system epilepsy since it affects the bilateral episodic memory system with the hippocampi and manifests abundant interictal and ictal memory disturbances. NREM sleep is essential in the memory process, and the unstable hippocampal memory engrams are replayed and transmitted to the frontal lobe during NREM sleep [37, 38, 39].

Any injury to certain hippocampal sectors (a first hit) may induce a 7–7.5 year-long “rewiring” process in synaptic structure and connectivity [40]. The discharges from a damaged hippocampus to produce nonsense information, rather than useful memory fragments, may obstruct the hippocampo-frontal memory correspondence [41, 42]. Memory consolidation is accomplished by the interplay of sleep slow waves, sharp-wave ripples, hippocampal spindles, and ripples [36]. Slow waves orchestrate this teamwork by placing it under homeostatic regulation [43, 44, 45, 46].

An essential step in epileptic transformation leading to MTLE is the metamorphosis of hippocampal sharp-wave-ripples to an epileptic pattern, epileptic spikes, and pathological high frequency oscillations (HFOs) [36]. The mechanism underpinning this transformation was described recently [47].

During the process of epileptic transformation, the number of evolving hippocampal spikes correlates inversely with the number of spindles [42], with the decrease in spindles also contributing to the memory disturbances of MTLE patients.

The role of the homeostatic process is supported by the circadian and night-time distribution of evolving interictal spikes. These prevail during high homeostatic pressure periods such as the first part of night sleep and the evening-afternoon time of day, as compared to the morning hours [48].

Both the interictal and ictal symptoms of MTLE are interwoven with memory disturbances [49, 50]. The insidiously evolving interictal memory impairment remained undetected for a long time because it can only be shown by neuropsychological follow-up studies. In ictal symptomatology, pure amnestic seizures, déjá- and jamais vu, as well as psychic or intellectual auras (dreamy states) are consistent with memory disturbances.

Disturbances of consciousness by focal onset temporal lobe seizures may also be thought of as memory-knockouts that evolve because of hippocampal involvement, i.e., a hippocampal “knockout” by seizure, as revealed by deep electrodes. Patients lose contact but not their muscle tone. They do not fall, but cannot understand speech during the seizure. This symptomatology can be understood as an acute memory loss, or a transient loss of the capacity to recognize things, language, or people.

The memories evoked by stimulation of the exposed temporal cortex of epileptic patients [51] also support the association between epilepsy and the memory system.

The corticothalamic network hosts the function of falling asleep. Absence epilepsy (AE) is linked to this network and is thus considered to be a corticothalamic system epilepsy [52].

This notion is based on studies reporting that spike wave discharges (SWD) emerge from the same circuit that normally produces sleep spindles [53]. Using ablation experiments, Steriade et al. [54] showed that “SWD originate in the neocortex and are disseminated through mono-, oligo- and multi-synaptic intracortical circuits, and exhibit generalized features”.

Experiments by Meeren et al. [55, 56] confirmed the focal onset of seizures deemed “primary generalized”. In addition, several clinical and EEG reports have highlighted the focal features of apparently generalized SWD. However, caution is needed when using the term “focal” in the original sense (as in traditional focal epilepsies) because absences build up (enrolled) step-by-step, mostly from a fronto-cortical network in each seizure, in contrast to the permanent epileptogenic zones of focal epilepsies [57]. It is important to note that a genetic background allowing epileptic transformation in those critical periods has been shown both in animals [58] and humans.

The regulation of sleep spindling and other sleep rhythms appears to be controlled by the thalamic reticular nucleus [59] via mutual interaction between its GABAergic neurons, thereby regulating the level of output inhibition exerted on thalamic relay cells [60]. Based on studies with animal models, SWD of absence seizures also appear to originate in the thalamus by a mechanism like spindle generation [61] and which is based on reduced GABAergic inhibition [62].

Absences are linked to transitional periods between wakefulness and N2 sleep, and are referred to as “critical” vigilance levels. Whereas absences seemingly occur in wakefulness, they actually emerge during falls of vigilance. This vigilance level-dependence explains their distribution throughout the 24-h sleep/wake cycle [63].

When arousals in sleep induce absences, EEG microstructural analysis can reveal the actual link with reactive sleep-like (slow wave) anti-arousal responses because there is a strong positive correlation between CAP A1 and SWD. Additionally, SWD is prevalent in the first sleep cycles and declines later with the decay of the delta power from evening to morning [64]. Regarding the deepening slopes of sleep dominated by subtype A1, the CAP-related activation of SWD increases three-fold compared to ascending slopes containing more A2 and A3 events [65].

Further supporting the nature of “falling asleep epilepsy”, more SWD occurs during fluctuations toward NREM sleep than toward wakefulness or REM sleep [66].

The vigilance-level dependence of absences is also supported by neuroimaging showing increased connectivity in the anterior thalamic structures during the falling asleep period of genetic juvenile myoclonic epilepsy patients [67].

The slow wave-dependence of SWD links absences to homeostatic regulation and supports the notion of a cortico-thalamic, falling-asleep system epilepsy underpinned by a genetic predisposition, thus making AE another network disease.

The anatomical regions around the Sylvian fissure are termed the perisylvian network (PN). The PN covers the lower lateral surface of the frontal lobe from the prefrontal areas including the Broca area, through the opercular structures and the parieto-temporal areas with the Wernicke field, the first and second temporal gyri, and partially the occipital cortex. The abundance and complexity of the involved regions occupying an important central part of the cortex reflect the high phylogenetic advance of human communication. This network organizes writing, reading, speaking, and calculation with the auxiliary machinery of hearing, sound production, articulating, and even vision [90] (Fig. 4, Ref. [90]).

Fig. 4.

The human perisylvian language network. M1, primary motor area of speech expression (flesh-colored); PM, premotor area; PF, prefrontal area working memory (pink); A1, AB, PB, auditory areas. The purple arrows represent interconnections among areas that exist only in humans. Reproduced with permission from Schomers MR, Neurocomputational Consequences of Evolutionary Connectivity Changes in Perisylvian Language Cortex, 2017 [90].

The spectrum of self-limited focal epilepsies (SeLFE) comprises the most frequent childhood epilepsies that can be considered system epilepsies, i.e., network diseases of the PN. The focus here is only on the part of the spectrum where the spectral cohesion is clear.

In this paper involved conditions include: (1) non-epileptic pattern-carriers of centrotemporal spikes (CTS), involving mainly the relatives of SeLFE children; (2) typical SeLFE with centrotemporal spikes (SeLECTS; Rolandic epilepsy) and presenting with rare focal sensory-motor seizures and CTS during NREM sleep; (3) atypical SeLECTS characterized by more seizures and frequent CTS over both hemispheres, as well as early onset, more prominent cognitive deficits, and occasional opercular status epilepticus in sleep; (4) developmental epileptic encephalopathies developmental/epileptic encephalopathy with spike wave activation in sleep (DEE-SWAS) where spikes and slow waves cover widespread regions of both hemispheres more or less continuously in slow wave sleep, as well as electrical status epilepticus in sleep (ESES) and Landau-Kleffner syndrome, with a regional ESES-like EEG pattern.

Recent neuropsychological research has revealed mild cognitive deficits even in “benign” SeLECTS children, mainly affecting their language functions. The language deficit is more important in atypical forms and culminates at the encephalopathic end of the spectrum. CTS-activity that accumulates in NREM sleep appears to increase from asymptomatic CTS carrier relatives, through SeLECTS and atypical SeLECTS children, to DEE-SWAS patients (Fig. 5, Ref. [91]). The degree of CTS-activity parallels that of HFO (ripples), considered to be the best markers of epileptogenicity [92]. Thus, no ripples associate with CTS in non-epileptic CTS carriers, and HFO activity parallels the severity of epilepsies in the spectrum. The mechanism of metamorphosis from sparse or frequent spiking to the ESES EEG pattern is still unclear. In ESES, a bilateral tsunami of spikes floods both hemispheres and tends to be continuous during NREM sleep. IEDs are rare during waking, and REM sleep is free of discharges. Anti-seizure medication does not always work, and a permanent mental handicap may occur if the ESES duration exceeds 18 months [93].

Fig. 5.

Three forms (A, B, C) of the centrotemporal spikes (CTS) across the self-limited focal childhood epilepsy spectrum. Left (A), CTS without epileptic seizures; middle (B), self-limited childhood focal epilepsy with centrotemporal spikes (SeLECTS) with ripples superimposed on CTS (spectrogram: top, averaged spikes; middle, spectral frequencies superimposed on spikes (red arrow)). Right (C), the electrical status epilepsy in sleep (ESES) discharge pattern assumed to be consistent with augmented CTS with pathological ripples (not shown). Reproduced with permission from Kobayashi K, Epilepsy with electrical status epilepticus during slow sleep and secondary bilateral synchrony. 1994 [91]. LKS, Landau-Kleffner syndrome.

Although it is assumed that the electrical pattern of electrical status epilepticus in sleep/Landau-Kleffner syndrome is consistent with exaggerated CTS discharges, this requires further study. A large body of literature supports the notion of a focal origin and secondary bilateral synchronization behind the bilateral manifestations [94, 95, 96, 97, 98].

Bölsterli et al. [99, 100, 101] reported a deficit of sleep slow wave downscaling in ESES patients. This finding suggests that a sleep homeostatic disorder underpins the cognitive loss related to the extraordinary increase in epileptic EEG manifestations. The remaining PN epilepsies are also shown to link with sleep homeostasis, even bi-directionally.

The concept of system-epilepsy links epilepsies to a derailment of physiological brain functions, i.e., seizures manifest the disfigured features of the hosting brain-systems. This concept also appears to be valid for SeLFE. Such epilepsies can be considered network diseases of the PN, where the high-level plasticity required for communication is sensitive to epileptic derailment.

Post-traumatic epilepsies are well-known complications of brain injuries. Acute symptomatic seizures often evolve shortly after the trauma, and chronic epilepsy may develop later. Between 20–60% of epilepsies originate from traumatic injuries [106], while 86% of early acute symptomatic seizures are followed by epilepsy within two years [107].

The homeostatic origin of trauma-induced epileptogenesis has been revealed by cortical undercut models. The partial deafferentation of the cortex caused by the trauma results in the inability to maintain neuronal activity, as reflected by the prolongation of silent (down) states during slow-wave sleep, as well as in waking and REM sleep. This requires a homeostatic recovery with the potential for epileptic derailment [22, 108, ].

The cortical undercut is an accepted model for penetrating wound epileptogenesis and works well in rats, mice, and cats. It reproduces the paroxysmal pattern seen in brain injuries, with acute symptomatic seizures first being triggered by the trauma, followed by a latent period, and then spontaneous and unprovoked seizures evolving later [109].

This theory proposes that an increase in overall network-silence would increase excitability. Indeed, the duration of silent states was shown to correlate with the instantaneous firing rates [110], with the intrinsic and synaptic excitability increasing within and around the undercut cortex (Fig. 6, Ref. [22]).

Fig. 6.

The cortical undercut model of penetrating wound epileptogenesis. (A) Cat brain depicting localization of the undercut (arrows). Left: global view; middle: frontal section; right: middle and sagittal sections. (B) Intracellular and field potential recordings during different states of vigilance. The epochs indicated by horizontal bars are expanded below. Note the presence of large amplitude hyperpolarizing potentials (indicated by the shadowed area) during NREM and REM sleep, as well as in the waking state. (C) Intracellular neuronal recording in the intact cortex (Reproduced with permission from Timofeev I, Posttraumatic epilepsy: the roles of synaptic plasticity. 2010 [22]). NREM, non-rapid eye movement; REM, rapid eye movement.