To sleep or not to sleep – Effects on memory in normal aging and disease

Abstract

Sleep behavior undergoes significant changes across the lifespan, and aging is associated with marked alterations in sleep amounts and quality. The primary sleep changes in healthy older adults include a shift in sleep timing, reduced slow-wave sleep, and impaired sleep maintenance. However, neurodegenerative and psychiatric disorders are more common among the elderly, which further worsen their sleep health. Irrespective of the cause, insufficient sleep adversely affects various bodily functions including energy metabolism, mood, and cognition. In this review, we will focus on the cognitive changes associated with inadequate sleep during normal aging and the underlying neural mechanisms.

Introduction

We spend about one-third of our lives asleep. This altered state of consciousness has fascinated humankind for centuries. Despite several decades of research in humans and animals, the simple question – why do we sleep – has remained mysterious. Although many functions, including energy conservation, learning and memory, immune support, and toxin clearance from the brain, have been proposed, no consensus has been reached regarding its primary role [1], [2], [3], [4], [5]. Nevertheless, it is undeniable that sleep is absolutely necessary for general physical and mental well-being, and it cannot be substituted with any other processes.

Mammalian sleep can be divided into two major states including rapid eye movement sleep (REM sleep) and non-rapid eye movement sleep (NREM sleep), which are accompanied by specific alterations in brain activity, causing changes in physiology and behavior. NREM sleep is identified electrophysiologically by the presence of a “synchronized” electroencephalogram (EEG) consisting of high-voltage slow waves with a dominant delta rhythm (0.5–4 Hz) and a decrease in muscle tone. In humans, NREM sleep is subdivided into four states (N1-4) based on EEG characteristics, but stages 3 and 4 have recently been combined and referred to as stage 3 (N3) or slow-wave sleep (SWS) [6], [7]. In laboratory rodents, it is common practice to consider NREM sleep as a single state, but some researchers divide this stage into two states based on the magnitude of synchronization and the presence of isolated or merged sleep spindles (0.5–2 s waxing-and-waning oscillations of 10–15 Hz sigma activity in the EEG) [8], [9], [10], [11]. In addition to EEG slowing and muscle relaxation, NREM sleep is accompanied by decreases in body temperature as well as reduced heart rate and respiratory rate.

REM sleep, on the other hand, is characterized by the presence of rapid eye movements and a “desynchronized” EEG pattern consisting of low-voltage, high-frequency waves [12], [13]. Most vivid dreams occur in this stage although they also may be present during NREM sleep. A complete loss of postural muscle tone is another hallmark feature of this state, and this REM atonia is considered to be a protective mechanism to prevent dream enactment [13], [14], [15]. However, brief phasic muscle twitches, i.e., isolated or bursts of muscle activity, may occur against the backdrop of “tonic” atonia [13], [14], [15]. Other key physiological changes that occur during REM sleep include autonomic instability with higher and variable heart rate and respiratory rate, penile erection, and cessation of thermoregulation [12], [15].

The timing, amounts, and architecture of these sleep states undergo significant changes across the lifespan, and aging is associated with marked changes in the quality and quantity of both NREM and REM sleep [16], [17], [18], [19]. Similarly, several neurodegenerative and psychiatric disorders are accompanied by significant alterations in sleep behavior [20], [21], [22]. These changes may even precede some of the common neuropsychiatric disorders, especially Alzheimer's disease (AD) and related dementias, Parkinson's disease (PD), and major depressive disorder (MDD) [20], [21], [22]. Irrespective of the cause, insufficient sleep has significant adverse health effects, primarily cardio-metabolic, mood, and cognitive consequences [23], [24], [25], [26]. This review will specifically focus on the cognitive changes associated with inadequate sleep in normal aging and their potential neurological basis.

What are the age-related changes in sleep-wake behavior?

Sleep in infants

The development of sleep occurs in utero, and the cycling of NREM and REM sleep can be detected as early as 28 weeks of gestational age [27]. During the first weeks of life, neonates sleep approximately 16–18 h and exhibit a polyphasic pattern with several sleep episodes throughout a 24-h period [28], [29], [30]. A single sleep episode may consist of one or two sleep cycles of 50–60 min and usually begins with REM sleep (referred to as “active sleep”). REM sleep occupies 50–70 % of total sleep time [28], [29], [30], and it has been theorized that this high level of REM sleep is crucial for normal brain development and for maintaining and establishing new synapses and connections [31]. Rapid eye movements and irregular breathing typical of REM sleep can be observed, but the neonates display a variety of facial and body movements (including smiling, startling, and sucking motions) during REM sleep because the neural circuits controlling REM atonia are not fully developed at this stage [30]. However, by six months of age, these mechanisms ensue and muscle atonia is evident during REM sleep. Sleep cycles also begin with NREM sleep from this point onwards. The duration of sleep gradually decreases across the first year, and 12-month-olds sleep for 12–13 h, in which REM sleep occupies less than 40 % [28], [30], [32], [33]. Moreover, the polyphasic pattern of sleep observed in infants begins to turn into one to two naps. By school-age, sleep is more confined to the nighttime while total sleep time and percentage of REM sleep gradually decrease until the adult stage and beyond [33] (Table 1).

Table 1.

Sleep duration across lifespan in humans.

Age	Number of sleep episodes	Total sleep duration (hours)	REM sleep (% of total sleep duration)
Newborn (0–3 months)	Several	16–18	50–70 % (active sleep)
Infants (3–6 months)	Several	15–17	35–55 % (active sleep)
Infants (6–12 months)	Several	12–15	25–40 %
Toddlers/Preschoolers	2–3	9–10.5	25 %
School-age (1–5 years)	1	8.5–10.5	25 %
Middle childhood (6–12 years)	1	8–10	20–25 %
Teens/Adolescents	1	7.5–8.5	20–25 %
Adults	1	7–8	20–25 %
Older adults	Fragmented (variable)	6.5–7.5	20–25 % (variable due to medical social issues

Sleep in young and older adults

Young, healthy adults typically sleep about 7–8 h per night and go through four to six sleep cycles of 70–120 min duration. NREM sleep is predominant during the first two cycles, while REM sleep gradually increases across the night and is the prominent stage in the last two cycles. REM sleep at this stage of life occupies about 25 % of total sleep time (Table 1). As we age, several drastic changes in sleep behavior occur, including a shift in the timing of sleep, reduced time spent in sleep, and reduced sleep efficiency [17], [18], [19]. It is often difficult to dissociate the sleep changes specifically related to normal aging from those secondary to medical or treatment conditions, but sleep complaints are more common in older adults, and the above sleep changes are observed even in individuals not suffering from any medical conditions. A recent study analyzing sleep data from 69,650 adults from several countries across the globe confirmed the presence of sleep changes during the normal aging process [34]. Several detailed reviews on age-related changes in sleep behavior exist [17], [18], [19], [35], [36], but we herein briefly summarize the most important ones (Table 1) and their putative neurological basis.

A shift in sleep timing is a prominent change that is commonly observed in older adults, and they exhibit phase advance, i.e., go to bed earlier in the night and wake up in the morning earlier than young adults [35], [37], [38], [39]. Their sleep-wake rhythms become less robust, and they also exhibit a reduced ability to adjust to any changes in the phase of these rhythms [35], [37], [38], [39]. Accompanied by such changes in sleep timing, the total time spent in sleep is decreased in older adults. Meta-analysis studies find a linear correlation between the total sleep time and age, with 10–12 min less sleep per decade [16]. For example, total sleep time is about 30 mins less in 60-year-olds compared to 40-year-olds. While some reports indicate a further decline in total sleep time even after age 60 years, others report either no significant change or even a gradual increase that stabilizes at around 70 years of age [16], [40], [41], [42]. Nevertheless, a decrease specifically in SWS, i.e., N3 stage in the elderly is a consistent and most prominent finding across studies [16], [42], [43], [44], [45], [46]. While the percentage of other NREM sleep stages increases, SWS decreases considerably by age 60 and does not appear to further decrease beyond that age. Finally, the percentage of REM sleep remains mostly stable throughout adult life but may decrease after age 60. However, there is a large discrepancy across studies about REM sleep changes in the elderly. While some reports indicate that the percentage REM sleep decreases with age, several other studies find no age-related decline in REM sleep [46], [47], [48], [49], [50], [51], [52]. Thus, the decrease in REM sleep with age, if any, may be very small.

In addition to sleep amounts, EEG spectral characteristics undergo significant changes with age. For example, slow-wave activity in the EEG (SWA; 0.75–4.5 Hz delta oscillations), which is considered to be an indicator of SWS depth or intensity, progressively decreases with age. The amplitude, slope, and density of these slow-wave oscillations decrease and a drastic reduction in absolute SWA power (up to 80 %) is observed in the elderly, especially during the first half of the night [17], [42], [46], [53], [54]. While some studies also find a reduction in average frequency of slow-waves (by 0.1 Hz), it is not a consistent observation. As SWA also reflects a need for sleep, it is likely that older adults simply “need” less sleep [55], [56]. EEG spindle activity, another key feature of NREM EEG (N2 stage), also declines with age [42], [47], [57], [58]. Similar to SWA, the frequency of spindle activity shows very little variation with age, but the amplitude, duration, and density of sleep spindles are reduced significantly [47], [57], [58]. Finally, even during REM sleep, a significant reduction in SWA (especially, 1.25–4 Hz) along with an increase in higher frequency (10–20 Hz) activity is observed in the elderly [53], [59]. As discussed in the later part of this review, both SWA and spindle activity have important implications in cortical development and memory processing.

Another major problem experienced by older adults is a deterioration of sleep continuum or maintenance [16], [42], [59], [60], [61], [62]. Although it is commonly assumed that older adults have difficulty falling asleep, many studies do not find an increase in sleep latency, but they consistently report a decline in sleep maintenance [16], [42], [59], [60], [61], [62]. Older adults wake up more often during the night; both brief arousals (3–14 s duration) and longer waking bouts are increased, leading to fragmented nocturnal sleep [43], [45], [59], [63], [64]. Consistent with this, sleep efficiency (ratio of total sleep time to time in bed expressed in percent) decreases with age, even after 60 years of age [16], [45].

Older individuals may often display increased sleepiness and hypersomnolence during the daytime, but these changes are likely related to other conditions such as disease and treatment, changes in daytime behavioral habits, or environmental factors [65], [66], [67]. Epidemiological and experimental sleep restriction studies in humans have found that sleepiness and hypersomnia during the daytime are not related to normal aging [65], [38], [68]. Indeed, healthy older adults were more alert and were more effectively able to maintain wakefulness and cognitive performance after sleep deprivation [38], [69], [70], [71]. It can be argued that older adults could accumulate as much, or more, sleep pressure as younger adults after sleep deprivation, but they are unable to exhibit sleep rebound due to a reduced ability to fall asleep. However, Klerman and colleagues found that while healthy older adults wake up more frequently than younger adults, they also fall back asleep at the same rate as younger adults [63]. These data, taken together with reduced SWA during the nighttime, suggest that healthy older adults may “need” less sleep than young adults.

Collectively, normal aging is associated with a shift in sleep timing, reduced SWS, and impaired sleep maintenance in humans. Most of these sleep changes appear to be stable after 60 years of age among older adults with excellent health, but negative medical, social, and lifestyle conditions may exacerbate these changes.

Sleep in aged animals

Several animal models, such as flies, rodents, and non-human primates, have been shown to exhibit age-related sleep changes, but we will limit our discussion to rats and mice used in a large proportion of these studies. The average lifespan of laboratory mice and rats are 24 and 36 months, respectively, with 18- to 24-month-olds considered ‘aged’ and 10- to 14-month-olds considered ‘middle-aged’ [72], [73], [74], [75], [76]. Sleep data from these age groups were compared to 2- to 4-month-olds (young group) in most studies investigating age-related changes.

Young (adult) rats and mice sleep about 50–60 % of the 24-h period (40–55 % NREM sleep, 5–10 % REM sleep) when maintained on a 12 h:12 h light: dark cycle. As nocturnal rodents, they spend a majority (60–75 %) of the light period in sleep and a similar proportion of the dark period in wakefulness. Unlike humans, sleep in these rodents is polyphasic in nature, and they go through 150–300 sleep-wake cycles in a 24-h period. NREM sleep bouts may transition to either wakefulness or REM sleep, but almost all REM bouts terminate in wakefulness. Both the amounts and architecture of sleep-wake states are reported to be altered with age, but the findings are largely inconsistent, presumably due to differences in strain and genetic background of animals used in these studies. For example, most studies report no significant changes in the total time spent in NREM or REM sleep with age in both species whereas a few studies find a decrease or increase in these states either during the entire 24-h period or particularly during active or rest periods in older animals [77], [78], [79], [80], [81], [82], [83], [84], [85], [86], [87], [88], [89]. Nevertheless, a decrease in the amount of wakefulness and a corresponding increase in NREM sleep during the active period was noted in both rats and mice, which is similar to higher daytime sleepiness observed in older humans. Consistent with this higher sleep during the active period, diurnal variation of NREM and REM sleep is reduced and the daily sleep-wake rhythms are flattened in older animals [77], [79], [83], [90], [81], although this effect is not consistent across strains [80], [91], [92]. A most prominent change that is reliably observed both in aged rats and mice is the fragmented pattern of sleep, characterized by shorter sleep-wake bouts and more frequent transitions between states [77], [78], [82], [86], [81], [92], [93]. While older animals wake up more frequently from sleep, they also fall asleep at the same rate, suggesting that the ability to maintain sleep and wakefulness, rather than to initiate sleep, declines with age.

In addition to spontaneous sleep, the homeostatic sleep control is also dysregulated in aged animals. After 12–48 h sleep deprivation, older rats display smaller and delayed sleep rebound as well as less-consolidated sleep bouts and reduced SWA delta power during the recovery sleep [80], [94]. In contrast to rats, older mice exhibit higher SWA delta power during the recovery sleep after sleep deprivation and it decays more slowly, suggesting that older mice may dissipate sleep pressure slower than young mice [87], [95]. Moreover, SWA during spontaneous NREM sleep is also found to be higher in aged mice [87], [95]. These data suggest that aging may be associated with a persistently higher homeostatic sleep “need” in mice.

Collectively, animal aging is associated with sleep-wake fragmentation, reduced circadian variation of sleep-wake rhythms, deficits in sleep homeostasis, and an altered EEG spectral profile [80], [90], [81], [92], [96], [97], [98], [99], [79], [85], [100], [101]. As these changes are similar to those observed in humans, animal models of aging have been greatly helpful to understand the neural basis of sleep changes in the elderly.

Neural basis of sleep changes in aging

Research on animal models during the past 100 years revealed enormous information on the neural mechanisms controlling sleep-wake behavior. They indicate that sleep-wake states and the associated physiological changes are controlled by several neurochemically distinct neuronal cell groups and their complex interactions. Age-related changes in these cell groups, their projections, neurotransmitter signaling, and the molecular machinery may underlie sleep alterations with normal aging. Various medical and treatment conditions, including neuropsychiatric illness and sleep disorders, or lifestyle factors may independently contribute to these sleep abnormalities. Herein, we will primarily focus on the age-related changes in sleep-wake circuits.

NREM sleep-promoting cell groups

It is well established that the preoptic area (POA) in the rostral hypothalamus is a key sleep-promoting region in mammals [102], [103], [104], [105], [106]. Two distinct cell groups within the POA, namely, the ventrolateral preoptic nucleus (VLPO) and the median preoptic nucleus (MnPO), contain a high density of neurons that are maximally active during NREM sleep [106], [107], [108], [109], [110], [111], [112], [113]. Neurotoxic lesions of VLPO, MnPO, or large lesions encompassing both structures induce profound insomnia in laboratory rodents, and the magnitude of sleep loss is proportional to the lesion size, indicating that sleep-promoting neurons may be distributed throughout the POA [114], [115], [116], [117], [118]. Sleep-promoting POA neurons are mostly GABAergic, but subsets express a variety of neuropeptides, including galanin, corticotrophin-releasing hormone (CRH), tachykinin, and dynorphin [113], [119], [120], [121], [122]. Selective activation of these cell populations increases wake-NREM sleep transitions and NREM sleep amounts, whereas their inhibition reduces spontaneous NREM sleep and state transitions in mice [119], [123], [124]. In addition, focal deletion of POA galaninergic neurons in mice abolishes the ability to increase sleep amounts after sleep deprivation, indicating that these neurons are necessary for sleep homeostasis [124]. The human equivalent of the sleep-promoting POA is known as the intermediate nucleus of the hypothalamus (INH) or the sexually dimorphic nucleus [125]. The number of galanin-positive neurons in this region correlates with the time spent in consolidated sleep bouts (estimated based on the number and duration of activity bouts), suggesting that INH galaninergic neurons may play a role in sleep similar to rodent VLPO galanin neurons [126]. The mean volume of INH and the total number of neurons in this area are reduced in older people [126], [127], [128]. Importantly, the number of galanin neurons within the INH may be specifically reduced in older individuals [126], [129], suggesting that the loss of these neurons underlies sleep fragmentation and decreased SWS in the elderly, potentially contributing to normal cognitive decline. The number of INH galanin neurons is further reduced in older subjects with Alzheimer's disease (AD), implying that more severe sleep fragmentation and sleep loss might contribute to or worsen the cognitive impairment in AD [126].

In addition to the POA, other forebrain structures such as the zona incerta as well as hindbrain structures including the parafacial zone, ventrolateral periaqueductal gray, dorsal raphe, the rostromedial tegmental region, and other brainstem regions may contain sleep-active neurons and possess sleep-promoting capabilities [130], [131], [132], [133], [134], [135], [136]. Thus, the sleep-promoting network appears to be more widespread than previously thought and may be distributed throughout the brain, but any age-related changes in these regions that could potentially contribute to sleep changes are currently unknown.

Wake-promoting cell groups

The sleep-promoting neurons in the preoptic and extra-preoptic regions may directly or indirectly inhibit a brain-wide network of wake-promoting neurons (Fig. 1A). This network includes the noradrenergic neurons in the locus coeruleus (LC), dopaminergic neurons in the ventral periaqueductal gray (vPAG) and the ventral tegmental area (VTA), histaminergic neurons in the tuberomammillary nucleus, orexinergic and neurotensinergic neurons in the lateral hypothalamus, and glutamatergic neurons in the supramammillary, parabrachial (PB), and the pedunculopontine tegmental (PPT) regions [107], [137], [138], [139], [140]. Selective activation of each of these cell groups induces sustained arousal, whereas their inhibition or ablation decreases spontaneous wake amounts and the ability to remain awake for a longer duration [139], [140], [141], [142]. Each of these wake-promoting cell groups may produce cortical arousal via multiple pathways – 1) by activating other wake-promoting cell groups, 2) by inhibiting sleep-promoting neurons, 3) by directly activating the cortex, and 4) by indirectly activating the cortex through their projections to the basal forebrain and/or thalamus (Fig. 1A) [102], [107], [137]. The mutual inhibition between wake- and sleep-promoting cell groups (especially the VLPO) is the basis of the “flip-flop” model of sleep-wake control proposed by Saper and colleagues [107]. In this model, wake-promoting neurons inhibit the sleep-promoting VLPO and thereby disinhibit their own firing, leading to consolidated wakefulness.

Fig. 1.

A – Simplified view of the neural network regulating wake and NREM sleep in mammals. Several cell groups in the hypothalamus and brainstem are involved in the generation and maintenance of sleep and wakefulness. Wake-promoting cell groups activate the cortex through their projections to the thalamus and/or basal forebrain. In addition, they inhibit sleep-promoting cell groups and disinhibit themselves in the process. The concurrent activation of cortical regions and inhibition of sleep-promoting cell groups help to produce consolidated wakefulness. Conversely, the VLPO and other sleep-promoting cell groups may actively inhibit this wake-promoting network to induce sleep. B – Simplified view of neural network regulating REM sleep. The SLD contains REM sleep-generating neurons and these neurons are under the inhibitory control of “REM-off” neurons in the vlPAG/LPT. The interactions between these populations determine the timing and amounts of spontaneous REM sleep, but several other cell groups (primarily the MCHergic, orexinergic, monoaminergic and cholinergic neurons) modulate REM sleep by targeting “REM-on” and “REM-off” neurons. Red and green arrows indicate, respectively, inhibitory and activating influences on target regions. POA – Preoptic area; ZI – Zona incerta; RMTg – Rostromedial tegmental nucleus; vlPAG – Ventrolateral periaquaductal gray; DR – Dorsal raphe; PZ – Parafacial zone; LH – lateral hypothalamus, TMN – Tuberomammillary nucleus; SUM – Supramammillary nucleus; VTA – Ventral tegmental area; vPAG – Ventral periaqueductal gray; PPT – Pedunculopontine tegmental nucleus; PB – parabrachial nucleus; LC – Locus coeruleus; LPT – lateral pontine tegmentum; SLD – Sublaterodorsal nucleus; MCH – Melanin-concentrating hormone. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Each of the wake-promoting cell groups mentioned above is affected to varying degrees in normal aging; the changes in orexin neurons are especially well understood. Several deficits in the orexin signaling system, including marked reductions in orexin neuron numbers, orexin peptide levels, and orexin receptor expression, are observed in aged animals [143], [144], [145], [146], [147], [148]. Additionally, aged animals exhibit a significant degradation of orexinergic projections and reduced responsiveness of target neurons to orexin, indicating a diminished orexin activity on target sites [149], [150]. The orexin neurons that still remain may also undergo changes on a cellular or molecular level. For example, orexin neurons display an altered pattern of diurnal activity and an increased endoplasmic reticulum stress during aging [89]. Orexin system dysfunction in humans and animals destabilizes both sleep and awake states, causing more frequent sleep-wake transitions [151], [152]. Thus, sleep fragmentation and less robust sleep-wake rhythms in the elderly may be at least in part due to orexin system dysfunction. Finally, the remaining orexin neurons also appear to become hyper-excitable, which may increase nighttime arousals as well as sleep fragmentation in the elderly [148].

In contrast to orexin neurons, the TMN or LC wake-promoting neurons are preserved across normal aging [153], [154], [155]. However, decreased activity of LC neurons, especially during the active period, decreased levels of dopamine beta-hydroxylase (necessary for norepinephrine synthesis) in the LC, and decreased noradrenaline reuptake in the LC terminals are observed in aged animals, suggesting a reduced LC functionality in the aging brain [156], [157], [158]. Similarly, aged animals exhibit elevated levels of histamine metabolites in the cerebrospinal fluid and a decreased binding of histamine receptors in the cortex, suggesting diminished TMN action on target neurons [155], [159]. It is likely that other wake-promoting systems also deteriorate during aging and contribute to sleep alterations, but the experimental evidence is currently lacking.

REM sleep-regulating cell groups

Early transection studies established that the pons is both necessary and sufficient for generating REM sleep [15]. A restricted region within the pons known as the sublaterodorsal nucleus (SLD) contains glutamatergic neurons that are specifically active during REM sleep [160], [161], [162]. Non-specific neurotoxic lesions involving all SLD neurons or genetic lesions targeting specifically glutamatergic neurons robustly reduce REM sleep amounts in laboratory rodents [163], [164], [165]. Muscle atonia during REM sleep is also diminished or abolished in these animals, and they exhibit exaggerated phasic muscle twitches and a variety of involuntary motor movements (ranging from simple body jerks to complex behaviors such as walking and running) while in REM sleep [163], [164], [165]. Similar dream-enactment behavior is also observed in humans, and this disorder is termed REM sleep behavior disorder (RBD) [166], [167]. RBD predominantly affects older adults; the first signs generally appear after 50 years of age [166], [167]. SLD glutamatergic neurons inhibit spinal motor neurons by activating GABA/gycinergic neurons in the ventromedial medulla (VMM) as well as spinal interneurons to induce REM atonia [15], [168], [169], [170]. Neurotoxic or genetic lesions not only in the SLD, but also along this pathway i.e., VMM or spinal cord produced varying degrees of REM sleep without atonia accompanied by dream-enactment behavior [164], [171], [172], [173].

SLD neurons are inhibited by a “REM-off” region comprising the vlPAG and the adjacent lateral pontine tegmentum (LPT) in the midbrain (Fig. 1B) [161], [174]. GABAergic neurons in the vlPAG/LPT exhibit their lowest activity during REM sleep, and this activity reduction begins before the onset of REM sleep [174]. Selective activation of these neurons strongly suppresses REM sleep while their inhibition increases REM sleep levels [174], [175]. Thus, vlPAG/LPT GABA neurons may play a “permissive role” in REM sleep regulation. The interactions between these neurons and SLD glutamatergic neurons may determine the timing and quantity of spontaneous REM sleep, but other cell groups also modulate REM sleep by targeting these two regions (Fig. 1B) [168]. For example, melanin-concentrating hormone (MCH)-expressing neurons in the lateral hypothalamus promote REM sleep primarily by inhibiting the vlPAG/LPT [176]. In contrast, orexin neurons located in the same region can activate vlPAG/LPT neurons and prevent REM sleep [177].

As described previously, REM sleep appears to be minimally altered in normal aging, but major neurodegenerative and psychiatric disorders, common in older people, are associated with significant changes in REM sleep. Patients with PD display reduced REM sleep in addition to night-time insomnia and daytime sleepiness, and a large proportion of these patients is afflicted with RBD [178], [179]. Based on above data, it is conceivable that age-related neurodegenerative changes in the subcoeruleus, the human equivalent of SLD, or in the VMM underlie RBD in PD. Consistently, lesions in the pons are observed in humans with RBD [180], [181], [182]. In addition to PD, significant REM sleep dysregulation is observed in major depression, which is also common among the elderly. Individuals with depression exhibit a short REM latency (time to first REM sleep from sleep onset) and significantly higher REM sleep levels [183], [184]. However, it is unclear if the increase in REM sleep is a physiological response to regain emotional balance or a comorbid condition. Nevertheless, selective REM sleep deprivation alleviates depressive symptoms, and several antidepressants reduce or eliminate REM sleep and improve mood in both animals and humans [183], [185]. Animal studies point out that lesions in the ventromedial prefrontal cortex induce depressive behavior and the associated REM sleep changes simultaneously [186], [187]. Considering the functional abnormalities in the prefrontal cortex observed in aged individuals [60], [188], it is highly likely that medial prefrontal functional deficits underlie REM sleep changes in older subjects with depression.

Cell groups regulating sleep-wake rhythms

In addition to the aforementioned homeostatic control mechanisms, sleep-wake behavior is strongly modulated by the circadian timing system. The circadian rhythms in sleep-wakefulness are generated by the suprachiasmatic nucleus (SCN), which is considered to be the master circadian pacemaker [189], [190]. A molecular machinery involving circadian clock genes and a transcriptional-translational feedback loop in SCN neurons is crucial for generating circadian rhythms in physiology and behavior [189], [190]. SCN neurons do not send direct projections to any of the sleep- and wake-promoting cell groups, but multisynaptic pathways via the subparavantricular zone and dorsomedial hypothalamus are proposed to orchestrate sleep-wake rhythms [102], [191]. In addition to controlling the timing of sleep-wake states, the SCN is also implicated in active wake promotion. SCN lesions increase total sleep amounts in non-human primates, but not in other species [192], [193], [194]. Nevertheless, SCN lesions in rodents abolish sleep-wake rhythms by increasing sleep during the active periods and decreasing it during their rest periods [193], [194]. Thus, age-related changes in the SCN and the circadian timing system may underlie the shift in sleep timing and the daytime sleepiness in the elderly. While the total number of SCN neurons is not reduced, a subset expressing vasoactive intestinal peptide decreases with age, and their numbers correlate with the amplitude of motor activity rhythms [195], [196], [197]. In addition, numerous changes at the cellular and molecular level have been reported. A large proportion of SCN neurons becomes silent or ceases to fire rhythmically in aged mice [198]. Moreover, the phase synchrony between SCN neurons is grossly disturbed, resulting in a reduced amplitude of multi-unit firing [199]. Finally, the production of key neuropeptides in SCN neurons involved in circadian behavior including vasopressin, vasoactive intestinal peptide, gastrin-releasing peptide, and neurotensin decrease with age [200]. All these changes potentially alter SCN output signaling and alter sleep-wake and other rhythms in older animals and humans.

Collectively, aging is associated with a loss or reduction in functionality of wake- and sleep-promoting cell groups and the circadian timing system, which may underlie changes in sleep timing as well as reduced sleep amounts and quality in the elderly.

What is the contribution of sleep to memory processes in healthy aging and disease?

NREM sleep and memory

The benefits of sleep for memory processes have long been recognized. In one of the first experiments to assess this relationship, Jenkins & Dallenbach (1924) had their participants learn word lists and noted that a night of sleep significantly increased subsequent recall [201]. Since then, it has become clear that memories are “encoded” (information received and placed in initial storage) and “retrieved” (recalled) during waking time, but “consolidated” (stabilized in cortical networks for long-term storage) mainly during sleep. Here we will focus on the role of sleep in the consolidation of memories. It is important to note the distinction between “synaptic consolidation” (or cellular consolidation), describing events leading to changes in select synapses (e.g., through long term potentiation (LTP)), as opposed to “systems consolidation,” which describes the process of transforming a liable network connection to a long-term, integrated, and stable memory [202]. In this way, synaptic consolidation functions as a subunit of systems consolidation since many specific synapses need to be modified to consolidate a memory trace [203].

Through this process, new memories (new information) are stored in both hippocampal and neocortical networks, but retrieval is initially critically dependent on the hippocampus. Memory consolidation describes a process during which repeated replay of memory sequences during sleep (and quiet wake) strengthens and integrates memories in cortical areas for long-term storage and decreases the dependency on the hippocampus [204]. One example of memory replay during sleep is the firing sequences of so-called “place cells” in the hippocampus, which are active during maze exploration in a specific pattern, and the pattern is replayed during subsequent (but not preceding) NREM sleep [205], [206], [207], [208]. Later experiments revealed that memory replay occurs primarily during NREM sleep (but also during quiet wake) and may be compressed in time [209]. Reactivation of specific neuronal sequences during sleep can also occur in the cortex and subcortical regions such as the ventral striatum and VTA [210]. Simultaneous multi-unit recordings indicate that hippocampal reactivation is coordinated with activity in the neocortex, striatum, amygdala, or VTA, usually with a time delay of 40–50 ms, suggesting a spread from the hippocampus to other sites [207]. This coordination may differ with the nature of the experience in that more rewarding or more aversive stimuli produce stronger couplings between the hippocampus and, for example, the amygdala [211].

The replay of memories and the direction of information flow are also modulated by the differential release of specific neurotransmitters during sleep-wake states. Acetylcholine (Ach) promotes wakefulness and decreases the output of CA1 neurons. Ach levels are high during waking and low during NREM sleep, which may facilitate information flow “towards” the hippocampus during wake but “from” the hippocampus during NREM sleep [212], [213]. Interestingly, even learning tasks that have been labeled as “non-hippocampus dependent,” such as the novel object recognition test (NOR), still benefit from sleep [214]. Besides Ach, dopamine signaling (from the VTA) may play an important role in memory consolidation in that it can tag certain synapses for later strengthening when activated during e.g., a rewarding stimulus [215]. In one experiment, McNarma and colleagues (2014) artificially activated dopamine signaling from the VTA to the hippocampus during the encoding phase of a new spatial memory and showed that the dopamine-enhanced sequence was reactivated during subsequent sleep, leading to improved memory performance upon recall [216].

In addition to rodent research probing the interaction of sleep and memory consolidation, there is a growing body of research in human participants indicating that memories are consolidated during sleep. When participants are asked to learn word lists to be recalled after a delay, the best performance is observed if the participants slept during that delay [217]. Moreover, it appears that sleep not only helps to consolidate memories but also to analyze the information and to obtain the gist – perhaps to reduce the amount of data needing to be stored. In one experiment by Wagner and colleagues, participants learned stimulus–response sequences and gradually performed better over time [218]. However, performance in some participants sharply improved as they discovered a hidden rule underlying all of the sequences. Twice as many participants discovered this rule if the practice trials were followed by 8 h of nighttime sleep rather than 8 h of wake, suggesting that sleep facilitated the discovery of the hidden rule. Thus, sleep may represent an opportunity for the brain to engage in a sort-of “unsupervised learning” where the new information is compared to existing information [219].

But how does the brain transfer or integrate memories during sleep? One clue might be in the specific brain oscillations, especially during NREM sleep. Researchers have identified several oscillations that are specifically important for memory consolidation, involving reactivation of select memories and modulation of specific synapses spanning several brain regions. As described below, these components may be particularly crucial for fine-tuning the timing of the processes leading to memory consolidation.

Slow oscillation

Slow oscillations (SOs) are the result of a cortex-wide shift of neural activity patterns from the individual, independent activity to a coordinated rhythm alternating between an “up-state” and a “down-state” (Fig. 2A). Each state lasts about 100–500 ms, resulting in a < 1 Hz rhythm in which most cortical neurons in layers 2–3 and 5 take part [220]. These SOs often travel from anterior to posterior cortical regions but also reach subcortical structures, including the hippocampus [221]. Artificially enhancing the SO, e.g., by auditory clicks, significantly augments memory consolidation [222]. Also, in a paw reaching task in rats, when the task involved either one of the front paws, the amplitude of the SO was increased within the contralateral motor cortex in subsequent sleep, suggesting a use-dependent increase of the SO in specific brain regions [223].

Fig. 2.

Brain oscillations for memory consolidation during sleep. A – In the adult brain, thalamic spindles occur during the rising phase of the cortical slow oscillation (SO) and hippocampal sharp-wave ripples nestle in the excitable throughs of the spindles, facilitating memory processes. In contrast, in the aging brain (B), NREM sleep is reduced and fragmented, cortical SOs are less ample, and thalamic spindles are less frequent and also tend to occur out of phase with SOs. In addition, the frequency of hippocampal ripples co-occurring with spindles is reduced. These age-related alterations appear to negatively impact cognitive performance in the elderly.

Researchers often refer to faster frequencies (1–4 Hz) within SWA as delta waves, which may have separate mechanisms of origin and serve different purposes from SO [224], [225]. While the precise relationship between SO and delta waves is still not fully understood, it is clear that the amplitude of delta waves is augmented most with prolonged time spent in waking at all ages [226]. The decline of delta waves between early adolescence and young adults by as much as 50 % is thought to stem from a reduction in synaptic pruning and an associated decline in cerebral metabolic rate [227].

Sleep spindles

Sleep spindles are generated by neurons in the reticular thalamic nucleus (RTN) and spread to cortical as well as subcortical regions [228], [229], including the hippocampus [230]. Spindle density correlates with post-sleep retrieval performance [231]. They are often formed during the beginning of the SO up-state [232], and their coupling to the SO is enhanced in areas associated with prior learning [233], [234]. A study by Latchoumane and colleagues showed that when spindles are artificially triggered by optogenetic stimulation of RTN neurons (at 8 Hz), they only enhance memory when they are tightly coupled with the SO up-state [235]. One of the mechanisms by which SO-spindle events may enhance memory consolidation during sleep is by priming cortical pyramidal neurons to receive inputs through a reduction of inhibitory tone to pyramidal cell dendrites [236], making them more receptive to incoming long-range signals at that moment.

Sharp-wave ripples

While thalamic spindles may prime cortical neurons to receive input at specific times during the SO cycle, hippocampal ripples (80–300 Hz) may deliver the message. Ripples nest in the excitable troughs of spindles and, therefore, also in the up-states of the SO (Fig. 2A). This constellation is thought to facilitate information transfer from hippocampal to cortical networks [232], [237]. Generally, cortical SO trigger thalamic spindles, which in turn trigger hippocampal ripples, but ripples and spindles can also elicit SOs [233], [235]. This facilitates information flow between the cortex and hippocampus in both directions so that specific memories can be marked for replay and long-term storage [238]. Selective suppression of ripples impairs spatial memory [239], but all three rhythms are important and depend on each other for memory consolidation.

Taken together, the SO provides the necessary conditions for memory consolidation by reducing the external information reaching the cortex during the down-state and triggering thalamic spindles during the up-state. Thalamic spindles then prime cortical pyramidal neurons to receive afferent inputs from the hippocampus during ripples which are nested in the excitable throughs of spindles (Fig. 2A). Using the existing thalamocortical network, spindles then direct which cortical areas receive those inputs, as evidenced by the high degree of task-dependent localization of spindles [240]. In this way, the thalamus plays a similar role for memory consolidation during sleep than for directing selective attention during wakefulness [241], [242]. Importantly, post-learning sleep enhances the coupling of SO-spindle-ripple events, and the degree of coupling predicts subsequent memory performance [235], [243]. As described above, an overall reduction in SWS and sleep fragmentation are observed in older individuals. Moreover, the characteristics of both SO and spindles in SWS are also altered: Per unit of SWS time older brains tend to produce fewer SOs and with a reduced amplitude [244]. Moreover, the density, amplitude, and duration of sleep spindles are reduced, while the spindle frequency is unaffected or slightly increases [47], [245] (Fig. 2B). Most importantly though, the coupling of the SO-spindle-ripple events is reduced, leading to impaired memory consolidation [244], [246]. Specifically, while sleep spindles in the adult brain tend to arrive precisely timed to the beginning of the SO up-state, in elderly brains, sleep spindles often occur “too early” and more scattered, missing the best time window for information transfer between memory systems [60]. Not surprisingly, aging also affects hippocampal sharp-wave ripples, reducing their density and the oscillatory frequency within each ripple [247].

In addition to these alterations to timing and coherence of sleep oscillations, the overall directionality and possible origins of sleep oscillations changes in elderly brains. In multi-electrode EEG recordings in humans, it was found that sleep spindles tend to originate over frontal areas in young adults, while this focal origin seems much more dispersed in the elderly [60], [245]. One of the root causes for this topographic dispersion and the changes in sleep oscillations in elderly subjects may be a reduction in brain volume in regions generating these phenomena. Muehlroth and colleagues found a significant correlation between gray matter loss in the mPFC (presumed to be the origin of the SO), thalamus (origin of sleep spindles), and hippocampus (origin of ripples) as well as a decline in SO-spindle coupling and memory consolidation during sleep [246]. Consistent with this idea, patients with AD or with mild cognitive impairment also show reduced sleep spindle activity [248], which might contribute to memory decline in these populations.

To ameliorate the cognitive effects of decreasing sleep quality and quantity in the elderly, several distinct options are available – 1) treating comorbid factors such as chronic pain or sleep apnea, 2) implementing non-pharmacological means including improving sleep hygiene, relaxation therapy, or cognitive behavioral therapy (CBT), and 3) pharmacological treatment. Considering the first two options, studies have shown that treating sleep apnea with continuous airway pressure or using CBT in insomnia patients can improve sleep and with it restore some cognitive and memory functions [249], [250]. However, in many cases insufficient sleep remains an unsolved problem and those patients may require pharmacological interventions.

In contrast to e.g. treatments for sleep apnea or CBT for insomnia, pharmacological interventions have been shown to alter EEG characteristics in addition to improving sleep amounts, latency, and fragmentation [251]. For example, benzodiazepine (BZ) drugs, historically prescribed for insomnia, particularly in the elderly, decrease NREM sleep EEG frequencies < 10 Hz and increase faster frequencies [252], possibly contributing to the well described next-day side effects such as cognitive impairments and amnesia [253], [254]. BZs enhance GABA activity by binding to GABA_A receptors, thus globally reducing brain activity and inducing sedation. Newer drugs, non-benzodiazepine benzodiazepine receptor agonists (also called Z-drugs), such as Zolpidem or Eszopiclone, are structurally distinct from BZs and target specifically the alpha1 subunit on GABA_A receptors [255]. But, Z-drugs still share some of the side effects of BZs in that they impair the < 10 Hz component of NREM sleep EEG [252] and are thought to interfere with memory consolidation processes, especially in the elderly [256]. However, Z-drugs do enhance NREM EEG phenomena above 10 Hz, including spindles and, at specific concentrations under controlled conditions, have been shown to improve certain aspects of memory performance [257], [258]. Another pharmacological option to increase sleep, and thereby memory performance in the elderly might be to boost melatonin levels. Endogenous melatonin, produced in the pineal gland, is thought to promote sleep via MT1 and MT2 receptors on neurons in the SCN, thus modulating the circadian timing of sleep. While melatonin itself has a short half-life time of only 20–40 min, prolonged release formulations, such as Circadin, are entering the market and yield promising results especially for elderly patients [259], [260]. Compared to BZs (Temazepam) and Z-drugs (Zolpidem), Circadin does not alter EEG spectra during NREM sleep [261] and, therefore, might be a way forward to help elderly individuals to regain sleep and improve memory performance [262].

REM sleep and (emotional) memories

In contrast to NREM sleep, the available data on the effects of REM sleep on memory processes are more contradictory. In fact, some researchers argue that REM sleep may not have any function in memory consolidation whatsoever [263], [264], [265]. They point to the fact that several classes of antidepressant drugs (SSRIs, MOAIs, TCAs) drastically reduce or even abolish REM sleep, yet no detrimental effects, or sometimes even positive effects, on cognitive and memory functioning are observed [266], [267]. Moreover, one study reports that an individual who suffered a brainstem lesion at age 20 recovered and was found to be entirely devoid of REM sleep when re-examined at age 33. In the interim, this person had completed law school, become a successful attorney, and led a normal life [268], indicating that even complete loss of REM sleep for several years may have minimal impact on memory processes.

However, a vast amount of research in humans and animals does support the notion that REM sleep has a role to play in the modulation of memories during sleep, although the methodological tools to best investigate the specific role of REM sleep are under debate. Thus, for example, early studies report an increase in REM sleep duration after learning a new task [269], [270], suggesting that REM sleep may help to consolidate the newly acquired information. However, both NREM and REM sleep amounts are typically increased after learning [271], [272]. To separate the contributions of REM sleep from those of NREM sleep, researchers have used different experimental designs, such as the “split-night paradigm” for human subjects [273] or the “inverted flowerpot technique” for rodents. However, the split night paradigm suffers from circadian influences and is not useful to clearly delineate between the memory functions of NREM and REM sleep as both sleep phases are present (albeit to different degrees) during the first and second half of sleep. On the other hand, the inverted flowerpot method is stressful to the animals, which itself could affect subsequent performance in memory tasks [274], [275].

So if at all, what is the function of REM sleep for memory consolidation processes? Since REM sleep is the time when most of our vivid dreaming occurs, and the EEG patterns resemble those of waking activity (albeit with a prominent hippocampal theta rhythm, particularly in rodents), it was theorized that REM sleep serves as a break between NREM sleep sessions [276]. During this break, the specific processes that serve to consolidate memories during NREM sleep would be interrupted to allow the brain a recovery time without actually waking up. This line of thought is supported by findings suggesting that the firing frequency of hippocampal pyramidal cells increases during individual NREM episodes but decreases sharply during each subsequent REM sleep [277].

Recent research on sleep and memory processing has focused on the formation or pruning of synaptic contacts in the hippocampus and cortex during REM and NREM sleep [278], [279]. One of the proposed theories is that NREM sleep may mainly serve to increase synaptic strength underlying new memories while REM sleep may help to modify (increase or decrease) synaptic strength to improve the signal-to-noise ratio [280], [281] so that during retrieval in wakefulness, the desired memory traces stand out from similar alternatives. Consistently, several studies found that pruning of synapses and spines preferentially happens during REM sleep [282], [283].

One complicating factor in determining the contributions of certain sleep stages for memory processes is the differential nature of memories and the associated memory tests. Testing participants' performance of a newly learned finger-tapping sequence after a certain sleep intervention might yield different results than when participants are tested on how well they remembered a new poem. Overall, it appears that NREM sleep is important for modulating declarative (implicit) memories (such as recalling the names of the last 10 US presidents) and REM sleep for modulating non-declarative (explicit) memories (e.g., riding a bike) [284]. Another way to classify types of memories is whether they are emotionally tagged, i.e., are eliciting positive, negative, or neutral emotions. Mounting evidence suggests that REM sleep may particularly serve to modulate emotional memories [285], [286], such as conditioned fear responses [287]. However, whether REM sleep potentiates or de-potentiates these memories is unclear [288]. If, as several reports indicate, REM sleep serves to enhance emotional memories during sleep [289], [290], then it might be reasonable in certain circumstances to advocate for REM sleep deprivation as a therapeutic tool after traumatic events to prevent these memories from being strengthened [291]. However, if REM sleep serves to de-potentiate emotional memories and thus helps blunt the impact of traumatic events on the organism [292], then REM sleep would aid the restoration of emotional balance and help the organism to be ready for next day adventures. Consistent with this view, fMRI findings indicate that selective REM sleep (but not NREM sleep) deprivation decreases the connectivity between the medial prefrontal cortex (involved in logical thinking) and the amygdala (involved in emotional processing). In one study, REM sleep-deprived individuals rated negative emotional stimuli more strongly negative and had stronger amygdala activation responses to negative stimuli than rested fellow participants [293]. So, not only might REM sleep modulate how we deal with previously encountered emotional situations, but it also affects how we approach future emotionally charged situations [294]. Similarly, a recent study by Scullin and colleagues (2019) finds that REM sleep modulates future-directed “prospective memory,” i.e., the ability to “remember to remember” that a task needed to be done at a certain time. The authors show that prospective memory performance declines with age, but the decline is particularly steep in people with a marked reduction in REM sleep. Overall, the available research on the contributions of sleep to memory processes is underlining sleep as a key factor for memory consolidation. What is concerning though is that sleep amounts and quality decrease with age and in certain diseases, inducing memory decline over time. Restoring sleep should therefore be a key priority to facilitate a healthy ageing process.

Glymphatic drainage during sleep

Consolidating memories, extracting gist from new information, and resetting our emotional compass are all important tasks for a sleeping brain. At the same time, the brain may also be engaged in a more fundamental but vitally important task during sleep – removal of metabolites from waking neuronal activity – to prevent the deterioration of brain functioning. Seminal work in the research group of Maiken Nedergaard [295] describes how cerebrospinal fluid (CSF) enters the brain parenchyma by traveling in the peri-arterial space, surrounding incoming arterial blood vessels (Fig. 3A). From here, CSF is transported via aquaporin-4 channels (AQP4) into the endfeet of astroglia cells lining the peri-arterial space. The CSF becomes ISF when released from the astroglia into the interstitial space, where it can take up waste products for removal before being channeled into the peri-venous space via AQP4 channels on astroglia with their endfeet lining venous capillaries. One factor regulating the glymphatic flow is sleep [295]. Specifically, during NREM sleep (or anesthesia producing high SWA), but not during REM sleep or wakefulness, the extracellular space around neurons and glia expands, allowing increased glymphatic drainage and removal of waste products, including soluble misfolded proteins such as amyloid-beta (Aβ) [295], [296], whose accumulation may be causal to AD pathology and cognitive decline. Interestingly, levels of Aβ measured in the CSF fluctuate with the circadian rhythm in that Aβ levels are high during waking time and low during sleep [297]. Current research suggests a bidirectional interaction between poor sleep and rising levels of Aβ where poor sleep leads to more Aβ accumulation, which in turn worsens sleep [298]. In APP transgenic mice (a popular mouse model of AD), sleep states are more fragmented [299], and total amounts of both NREM sleep and REM sleep decrease with advancing AD pathology, but normal sleep patterns can be rescued with an early intervention clearing soluble Aβ and limiting the accumulation of insoluble Aβ plaques [300]. Although poor sleep does not co-occur with progressing pathology in all AD mouse models, chronic sleep-fragmentation or specific disruption of SWS has been shown to increase Aβ levels and neuroinflammation in 3xTg-AD mice [301].

Fig. 3.

Brain glymphatic system in the infant, adult, and aged brain. A – The infant/toddler brain contains fewer astrocytes compared to adults, but they express higher amounts of AQP4 channels in their endfeet lining the perivascular space of arterial vessels. This facilitates CSF flow into the brain parenchyma where it takes up brain waste products before leaving the interstitial space through astrocyte endfeet lining the venous perivascular space. Although there is a high level of waste production in infant brains (due to a high metabolic rate), the balance between waste production and waste clearance is optimal because sleep amounts in infants are high. B – When compared to infants, adult brains contain higher levels of astrocytes, but they express lower amounts of AQP4 channels. However, lower metabolic rates in adults and consequent reduction in waste production allows the glymphatic system to clear brain waste within only 7–9 h of daily sleep. C – In aged brains, AQP4 channels in astrocytes are increasingly mislocated (away from the endfeet lining the perivascular space), reducing fluid flow and removal of waste products. Coupled with reduced sleep amounts and quality, metabolic waste removal becomes increasingly tenuous, which leads to accumulation of insoluble Aβ into plaques and cognitive decline.

Several factors may play a role in the effectiveness of the glymphatic flow and waste product clearance. In infants and children, the level of brain metabolism is very high due to processes pertaining to brain development and maturation, which uses as much as 60 % of the body’s metabolic rate [302]. Moreover, while neurogenesis is largely complete by birth, the number of astrocytes is low and only increases (by several fold) over the first years of life [303]. To manage a high level of metabolic waste products with few astrocytes, infants’ brains sport higher numbers of AQP4 channels in astrocyte endfeet [304] and reserve large parts of the day for sleep / clearance (Fig. 3A). In the adult brain, metabolism levels are relatively low since developmental processes are complete, leading to fewer waste products needing to be removed. While astrocytes in adult brains show lower levels of AQP4 channels in their endfeet, the number of astrocytes present in the adult brain are higher, allowing for optimal clearance within normal levels (7–9 h) of sleep (Fig. 3B). In the aged human and mouse brain, numbers of astrocytes remain high [305], [306] and metabolic waste production is similarly low compared to the adult stage. However, with advancing age, there is a tendency for AQP4 channels to be mislocated away from glia endfeet lining the peri-arterial and peri-venuous space [307] (Fig. 3C). Combined with a decline in sleep duration and sleep quality in aged brains, the mislocation of AQP4 channels means that glymphatic clearance is less efficient, allowing metabolic waste products such as Aβ to accumulate and form insoluble plaques that can become harmful to neurons and lead to cognitive decline.

Conclusion

Sleep, an essential component of our daily lives, undergoes changes as we age: from infants to adolescents, adults and the elderly. The neuronal machinery regulating sleep-wake states generally adapts to the current needs of the organism to facilitate functions such as memory consolidation or glymphatic clearance. With increasing age, we are more likely to experience poor sleep, which could be due to aging processes per se or disease and medication conditions. Poor sleep on the other hand affects many physiological functions, potentially leading to severe health consequences, including mental illness, cardiovascular or neurodegenerative diseases. In addition, poor sleep can accelerate the progress of an existing disease or worsen its symptoms. Thus, improving sleep should alleviate symptoms or even prevent or delay disease onset. It is therefore important to maintain healthy sleep throughout all stages of aging. Future research should focus on devising newer strategies, both pharmacological and non-pharmacological, to improve sleep quality in the elderly.

Funding

This work was supported by United States National Institutes of Health grant R01NS119223 (RV), Harvard Brain Research Initiative seed grant (RV) and startup funds from Auburn University, United States (DK).

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.