SLEEP AND NEURODEGENERATION: EXAMINING POTENTIAL PHYSIOLOGICAL MECHANISMS

Abstract

Purpose of the review:

The purpose of this review is to provide an overview of potential mechanisms mediating the bi-directional relationship between sleep and neurodegenerative diseases such as Alzheimer’s disease. We provide updates on previously proposed mechanisms and identify new mechanisms particularly concerning how sleep disturbances affect memory-related neural circuits.

Recent findings:

In this review, we focus on the multiple mechanisms that potentially mediate the relationship between sleep and Alzheimer’s disease. We present updates for previously hypothesized mechanisms such as sleep-related changes in production/release and clearance of amyloid-β and tau proteins as well as more recently proposed mechanisms relating to tau phosphorylation, the orexin system, astrocytes, and microglia. We also highlight how disruptions in sleep EEG oscillations that underlie memory-related neural circuits, such as slow wave activity, theta bursts, sleep spindles, and gamma ripples, change in Alzheimer’s disease.

Summary:

Disturbed sleep increases Alzheimer’s disease risk via multiple potential mechanisms that suggest multiple targets to test approved and effective treatments of sleep disorders to prevent or delay Alzheimer’s disease.

INTRODUCTION

Alzheimer’s disease (AD) is the most common neurodegenerative disease and is characterized by an asymptomatic “preclinical” period of 15-20 years when amyloid-β (Aβ) aggregates as insoluble extracellular plaque followed by an increase and spread of intraneuronal tau tangles, and subsequently resulting in neuronal loss and the onset of cognitive symptoms.[1] Amyloid plaques may be detected in vivo during this asymptomatic period by imaging and biofluid markers including: 1) ratios of Aβ42/40 and phosphorylated tau (p-tau)/unphosphorylated tau in both cerebrospinal fluid (CSF) and blood;[2–4] and 2) positron emission tomography (PET) neuroimaging using radiotracers for fibrillar Aβ and paired helical filaments of tau.[5, 6]

Although disturbed sleep has been reported in individuals with dementia for decades,[7] the increased use of biomarkers to measure AD pathology (particularly amyloid plaques) in asymptomatic individuals with and without AD pathology has permitted sleep assessments before the onset of cognitive dysfunction and resulted in growing evidence for a bidirectional relationship between sleep and neurodegenerative disorders such as AD (reviewed in [8, 9]). Sleep disturbances have been noted to precede the onset of both cognitive problems and AD pathology.[10, 11] Alternatively, cognitively unimpaired older adults with biomarker evidence for amyloid plaques (i.e., “preclinical” AD) were found to have lower sleep duration,[12] lower sleep efficiency,[13] and lower non-rapid eye movement (NREM) slow wave activity (SWA)[14] compared to cognitively unimpaired older adults without amyloid plaques.

Sleep has potential as both an in vivo marker of brain function and an intervention to prevent or delay the onset of symptomatic neurodegenerative disease. Studies with longitudinal measures of cognitive function, sleep, and brain neurodegeneration pathology are needed to establish causal relationships. Multiple critical factors that both disturb sleep and increase AD risk are outside the scope of this review, however. Common causes of sleep disturbance, such as obstructive sleep apnea, insomnia, and periodic limb movements, are associated with increased AD risk.[15–17] For example, untreated OSA increases both soluble CSF concentrations of Aβ and tau [18] as well as the future risk of cognitive impairment.[17] Socio-structural determinants of health (SDoH) have also been shown to affect both neurodegeneration and sleep. Socioeconomic factors (i.e., environmental, socio-structural, and neighborhood disadvantage) are associated with cognitive impairment and worse AD pathology.[19–26] Further, SDoH, but not genetic ancestry, were recently found to predict dementia prevalence in Latin America [27] as well as AD neuropathology.[28] SDoH were also associated with shorter sleep duration and poorer sleep quality.[29] Although specific sleep disorders and SDoH will not be examined in this review, both of these factors, as well as others, affect the mechanisms discussed below to impact markers of AD pathology and cognitive performance in older adults in addition to the mechanisms discussed here (discussed in [30–34]). Despite this, several potential mechanisms mediating the relationship between sleep and AD have been identified. Understanding these mechanisms may suggest intervention targets. In this review, we will discuss potential mechanisms mediating this bidirectional relationship (Fig 1).

Figure 1:

Potential Mechanisms Mediating the Bi-directional Relationship Between Sleep and Alzheimer’s Disease (AD). Aging, social and environment factors, and co-morbid medical conditions may lead to sleep disturbances, but are also risk factors for neurodegenerative disease such as AD. Sleep disturbances increase AD risk through multiple potential mechanisms including neural circuit and slow wave sleep (SWS) disruption and orexin dysregulation that increases amyloid-β (Aβ), tau, and phospho-tau (pTau) levels and are associated with astrocytic and microglial dysfunction. These mechanisms promote the formation of AD pathology. Further, sleep disturbances and AD neurodegeneration disrupt neural circuits involved with memory consolidation and other cognitive functions increasing risk of cognitive impairment.

Mechanisms for how sleep disturbances promote Alzheimer pathology

Production/Release and Clearance of Aβ and Tau

Longitudinal sampling of both interstitial fluid (ISF) in mice and CSF in humans demonstrated that soluble Aβ and tau increase during wakefulness and decrease during sleep.[35–38] As predicted by these findings, sleep loss results in increased soluble levels of both Aβ and tau.[38, 39] Changes in sleep/wake activity affect production/release and clearance of Aβ and tau with the end effect of increasing their concentration and potentially promoting AD pathology. Studies in mice found that neuronal activity releases Aβ, tau, and alpha-synuclein into the ISF.[8] Neuronal activity increases during wakefulness and, as expected, sleep loss results in greater release of proteins released with neuronal activity such as Aβ, tau, alpha-synuclein, and neuronal pentraxin 2 (NPTX2).[38–40] Proteins that are not released with neuronal activity, such as neurofilament light chain (NfL) and glial fibrillary acidic protein (GFAP), do not change with prolonged waking.[36]

Sleep loss also decreases clearance of proteins and other metabolites from the brain via several potential mechanisms. Recent evidence strongly supports decreased protein clearance from the brain in sleep-deprived humans. For instance, humans injected intrathecally with gadolinium had increased retention of gadolinium (i.e., decreased clearance) after one night of sleep deprivation.[41] Acute sleep loss also decreases CSF-to-blood clearance of soluble AD biomarkers in healthy participants with paired CSF and blood samples.[42] Sleep-related decreases in clearance may be mediated by glymphatics and changes in bulk fluid flow in the brain,[43–46] changes in expression of the aquaporin-4 channel,[47] diurnal fluctuations of the composition of CSF by the choroid plexus,[48] meningeal lymphatics,[44, 45, 49] autonomic changes during sleep,[50] and the blood-brain barrier.[51] Further, differences in vascular dynamics with large CSF oscillations during NREM sleep and vasodilation during REM sleep likely also play a role in mediating clearance mechanisms.[52]

After AD pathology (e.g., amyloid plaques) begin to form, a feedback loop likely occurs. Sleep disturbances such as lower sleep efficiency[13] occur even in individuals with preclinical AD (i.e., cognitively unimpaired but with amyloid plaques) which lead to sleep loss and increased production/release and decreased clearance of Aβ, tau, and p-tau. Amyloid plaques are also associated with neuronal hyperactivity, independent of changes in sleep/wake activity, resulting in greater neuronal activity and tau accumulation.[53] The relationship between production/release and clearance of Aβ, tau, and other proteins is likely complicated and affected by stage of AD, age, medical comorbidities, and other factors.

Sleep and Tau Phosphorylation

Hyperphosphorylation of tau promotes tau aggregation as intraneuronal tangles that lead to neuronal loss and eventually cognitive symptoms. Further, increased tau phosphorylation is associated with amyloid plaques. Hyperphosphorylation of tau-threonine-217, for example, is a very early sign of amyloid accumulation as insoluble plaque.[54] Recent studies in mice found that sleep-wake activity affected protein phosphorylation in the brain.[55, 56] In humans, sleep deprivation affects tau phosphorylation depending on the specific phosphorylation site.[38] For instance, sleep loss was associated with greater phosphorylation at tau-threonine-217, lower phosphorylation at tau-serine-202, and unchanged phosphorylation at tau-threonine-181. A greater ratio of phosphorylated-tau-threonine-217 to unphosphorylated-tau-threonine-217 (pT217/T217) has been correlated with brain amyloid plaque burden including when plaques are cleared by anti-amyloid monoclonal antibody therapy.[57, 58] Further, these findings support sleep loss as a potential mechanism to increase tau pathology by elevating tau phosphorylation and promoting the formation of neurofibrillary tau tangles and eventually neuronal loss although further research is needed to establish this causality.

Orexin and AD Biomarkers and Pathology

Substantial evidence supports a role for the orexin system in the development of AD pathology. Orexin-A and orexin-B (also called hypocretin-1 and hypocretin-2) are two wake-promoting neuropeptides encoded by a common precursor 131 amino acid polypeptide, prepro-orexin (reviewed in [59]). Consistent with a neurotransmitter that regulates sleep-wake activity, CSF orexin-A levels fluctuate with a diurnal pattern.[60] A recent study in mice found that orexin neurons become hyperexcitable with age, lowering the threshold for sleep-wake transitions and leading to greater sleep fragmentation,[61] potentially altering production/release and clearance mechanisms discussed above. Knocking out the orexin gene in amyloid precursor protein (APP) transgenic mice that develop amyloid deposition also led to a marked decrease in amyloid pathology in the brain.[62]

Dual orexin receptor antagonists (DORAs) block orexin peptides at orexin receptor 1 (OXR1) and orexin receptor 2 (OXR2) promoting sleep. Treatment of APP transgenic mice with a DORA decreased soluble Aβ concentrations and amyloid plaques.[35] A recent study showed that a DORA (suvorexant) acutely decreased both soluble Aβ levels and tau phosphorylation in cognitively unimpaired amyloid-negative individuals age 45-65.[63] Interestingly, suvorexant did not increase measures of sleep suggesting that the effect of orexin on AD biomarkers may be mediated by sleep- and non-sleep mechanisms. Orexinergic blockade at OXR1 and OXR2 increases sleep, but these G protein-coupled receptors affect multiple downstream pathways, such as β-arrestin-2, p38 mitogen-activated protein kinase (MAPK), and the extracellular signal regulated kinase (ERK).[64–66] MAPK and ERK phosphorylate tau at multiple sites and β-arrestin-2 has been implicated Aβ generation. Future studies are urgently needed to compare the effect of DORAs on AD biomarkers compared to other sleep-inducing drugs (e.g., zolpidem) and to identify potential new targets for intervention by determining if non-sleep mechanisms mediated by orexins contribute to AD risk.

Sleep and Glial Cells

Emerging research increasingly implicates glial cells, such as astrocytes and microglia, in the relationship between sleep and AD. First, both astrocytes and microglia regulate sleep via calcium-dependent processes with astrocytes specifically implicated in slow wave sleep.[67, 68] Second, sleep loss in mice activates microglia and astrocytes. [69, 70] For astrocytes, activation leads to phagocytosis of presynaptic elements of large synapses and may represent the brain’s response to prolonged waking.[70] The causal effect of increased synaptic pruning by astrocytic phagocytosis on cognition needs further study, but may be one factor that contributes to adverse effects of prolonged waking on cognitive performance. Further, a recent study in APP/PS1 mice that develop amyloid deposition found that optogenetically activating astrocytes at the endogenous frequency of slow waves restored slow wave activity, reduced amyloid deposition, and improved memory performance.[71]

Acute and chronic sleep loss also increase inflammation (reviewed in [72]). Microglia drive tauopathy in an ApoE-dependent manner[73] as well as by increasing T cell infiltration.[74] Microglia also react to amyloid plaques leading to an immune response such as increased cytokines, complement proteins, interleukins, and chemokines (reviewed in [75, 76]) suggesting that sleep and the microglial response may interact. Indeed, sleep deprivation has been found to exacerbate microglia reactivity in transgenic mice that develop amyloid plaques even before plaques develop.[77] These findings support that sleep loss results in an increased inflammatory response that may promote neuronal injury. Further, optogenetically restoring sleep reduces amyloid pathology by increasing Aβ clearance by microglia.[78] These findings suggest that astrocytes promote disrupted sleep and AD pathology, while disrupted sleep activates microglia even in the absence of AD pathology.

Mechanisms for how sleep disturbances promote cognitive dysfunction

Abnormalities in Sleep EEG and Neurodegenerative Disease

Recent advancements in the study of sleep electroencephalography (EEG) have shed light on its perturbations in neurodegenerative diseases. Notably, Alzheimer’s disease is associated with alterations in frequency components and amplitudes within sleep EEG, including: 1) loss of oscillatory events termed K-complexes,[79, 80] 2) a reduction of low-frequency SWA,[14, 81–83] and 3) diminution of sleep spindles[14, 81–85]. Individuals with alpha-synucleinopathies undergo similar EEG changes, and these individuals further demonstrate lower cognitive abilities in relation to lower SWA[86] and reduced spindle activity.[87] SWA is proposed to provide neuroprotective properties in preventing and slowing neurodegenerative disease,[88, 89] including a role of SWA in the protecting cognitive reserve among individuals with AD pathology.[90] Advanced signal processing methods have further identified changes in the coupling of discrete oscillatory events—particularly slow oscillations, theta bursts, and sleep spindles—within NREM sleep EEG in AD, correlating with molecular pathology and cognitive impairments.[81, 91, 92] REM sleep EEG abnormalities, such as alterations in frequency composition, are also noted in both individuals with AD[92] and in individuals with the prodromal synucleinopathy REM sleep behavior disorder.[93, 94] Further, the potential application of sleep EEG as a digital biomarker for AD is gaining traction, indicating a key role in sleep’s oscillations in predicting, detecting, and monitoring neurodegenerative disease.[91, 95, 96] These advancements provide a lens through which elements of sleep EEG can be correlated with the crucial roles of oscillatory events in facilitating brain communication and coordination of cellular activities. Collectively, they deepen our understanding of how observable oscillatory activity supports cognitive processes in both healthy brain functioning and in the context of neurodegenerative diseases (Table 1).

Table 1:

Oscillatory Events Potentially Linked to Symptomatic Neurodegenerative Disease

Oscillatory Event Type	Defining Characteristics	Role in Cognitive Functions	Link to Neurodegenerative Disease
Slow Wave Activity (SWA)	Orchestrates cortical activations through upstates and downstates, synchronizing large brain regions.[97, 120, 136]	Essential for sleep-dependent memory processing and synaptic regulation.[117, 131, 137-139]	Degradation of SWA linked to early Alzheimer’s disease pathology through cortical amyloid and tau accumulation and possible LC dysfunction.[14, 81–83, 99–101]
Theta Bursts	Oscillatory events preceding slow waves, observed in cortical and subcortical regions.[108–110]	Proposed parallel function in sleep to wake-state bursts of theta activity for hippocampal-dependent memory retrieval.[111]	Loss of power and timing imprecision in MCI and early AD, indicating potential disruptions in hippocampal neural circuits.[91]
Sleep Spindles	Discrete oscillatory events synchronized in cortical and subcortical structures during NREM sleep;[108–110, 114, 115] occur in isolation and coupled to slow wave events.[123, 124]	Integral to memory processing and consolidation, linking hippocampal and limbic structures with association cortexes during memory replay.[116, 117, 131, 140]	Aging linked to lower-frequency spindle coupling; abnormalities in spindle phase alignment and frequency consistency associated with memory impairment and AD pathology.[81, 91, 92, 128]
Gamma Ripples	Bursts of high-frequency gamma activity associated with slow waves and spindles; occur as sharp wave ripples (SWR) coupled to slow waves and gamma ripples coupled to spindles.[108, 109, 129]	Facilitate the binding of cortical regions and synchronization of neuronal activity during memory reactivation and consolidation.[129, 131, 140]	Abnormalities in sharp wave ripples associated with AD pathology;[132, 133] APOE4 expression impacts hippocampal microstructure and ripple production.[134, 135]

Slow Wave Activity

SWA plays a crucial role in orchestrating cortical activation, alternating between “upstates” and “downstates” states and synchronizing large regions of brain cortex.[97] This synchronous activity is critical in sleep-dependent memory processing and synaptic regulation (reviewed in [98]). The locus coeruleus (LC) also exhibits synchronous activation with upstates and downstates states of SWA, suggesting a link between the noradrenergic system and cortical activations during SWA upstates.[99] Given the early involvement of the LC in AD,[100, 101] this is a potential neuroanatomical mechanism through which AD pathology might impact SWA. Other studies have demonstrated anatomical specificity for cortical accumulation of amyloid and tau in association with disrupted SWA.[14, 81, 83] Together, these studies provide connections between neuroanatomical structures and SWA physiology that inform our mechanistic understanding of the relationships between sleep and AD.

SWA can be further categorized into slow oscillations (<1 Hz) and higher frequency delta waves (1-4 Hz),[102] and this distinction has been beneficial in elucidating specific EEG features associated with AD, such as the selective loss of slow oscillations concurrent with amyloid deposition.[82] Recent studies have also proposed the existence of various subtypes of slow oscillations, classified by their shapes, associations with spindle subtypes, and predominant neuroanatomical locations.[103–105] Furthermore, analyses of slow wave slopes, which represent “transition frequencies” between the downstates and upstates, have identified compositional changes in SWA related to aging, cognitive changes, and AD pathology.[92, 106, 107] Analyses incorporating these transition frequencies of SWA have also revealed variations in the spindle and theta burst coupling properties of high and low transition frequency SWA in relation to levels of molecular AD biomarkers and symptoms of mild cognitive impairment (MCI), as assessed by the Clinical Dementia Rating Scale (CDR).[91]

Theta Bursts

Theta bursts, recently identified as oscillatory events preceding slow wave occurrences, have been observed in cortical and subcortical regions.[108–110] Their precise function remains under investigation, but a leading hypothesis suggests a parallel function in sleep to the wake-state bursts of theta activity that occur during hippocampal-dependent memory retrieval.[111] In this context, the generation of theta rhythms might originate from intrinsic hippocampal theta oscillators involved in pattern completion activities during memory recall.[112] The noted loss of power and timing imprecision of coupled theta bursts in the context of MCI and early AD indicates potential disruptions in these underlying hippocampal neural circuits. Damage to medial temporal lobe structures, including the hippocampi, may underlie diminished theta burst activity in sleep EEG, providing a direct connection to the compromised hippocampal pattern completion and separation activities that produce amnestic deficits observed in MCI and early AD.[113]

Sleep Spindles

Sleep spindles, discrete oscillatory events traversing cortical and subcortical structures during NREM sleep, play a key role in linking hippocampal and limbic structures with association cortexes[108–110, 114, 115] and are integral to memory processing and consolidation.[98, 116–119] The oscillatory circuit that generates spindle waveforms has been localized to the reticular nucleus of the thalamus,[120] which typically produces peak EEG power around 13-14 Hz, although spindles span a broad frequency range.[121] Higher frequency spindles are more prevalent posteriorly, whereas lower frequency spindles are commonly recorded anteriorly.[104, 122] Spindles occur both as isolated events and in coordinated temporal coupling with slow waves.[123, 124] These coupled spindles are implicated in hippocampal-dependent memory consolidation through their temporal organization of gamma ripples, which are nested within spindle cycles.[98, 108, 109] Intracranial recording studies have revealed distinctions in spindle-to-gamma ripple coupling corresponding to neuroanatomical sources, with posterior hippocampal spindles demonstrating this coupling pattern and anterior hippocampal spindles more often lacking ripple coupling.[109] These findings suggest that spindle subtypes, varying by anatomical location, may have distinct roles in memory processing and synaptic regulation. Additionally, the coupling of spindles with slow waves differs by spindle frequency, with higher frequencies appearing earlier in the upstates of slow wave events.[103, 104]

Aging is linked with a shift towards more low-frequency spindle coupling, a phenomenon yet to be fully explored in the context of pathological aging or neurodegenerative diseases, although with some data indicating this shift correlates with reduced memory performance and brain atrophy.[125] It is worth noting that this frequency shift occurs in the context of more generalized age-dependent shifts in the amplitudes and frequencies of EEG components,[126, 127] although possible connections between frequency shifts of coupled spindles with these more widespread changes have not been delineated. The phase-dependent coupling of slow waves with spindles also changes with age, and abnormalities in the phase alignment of spindle coupling are associated with memory impairment and AD pathology.[81, 92, 128] Advanced signal processing techniques employing time-frequency representations have revealed more specific spindle neural circuit physiology, mapping loss of spindle-generating neural circuit precision in both the temporal coupling and frequency components of spindles to levels of molecular AD pathology.[91]

Gamma Ripples

Gamma ripples, characterized by bursts of high-frequency gamma activity in association with slow wave events and spindles, are proposed to facilitate the binding of cortical regions and synchronization of neuronal activity during memory reactivation.[129] A specific gamma activity pattern, known as a sharp wave ripple (SWR), occurs in association with successful recall and cortical reinstatement of memories during the wake state.[98, 130] This system is mirrored in the coupling of SWRs to slow waves during sleep and is implicated in similar cortical reinstatement and memory reactivation processes during sleep-dependent memory processing.[98] The gamma bursts that occur cyclically within spindle oscillations may also have a distinct role in mediating spike-timing-dependent plasticity in the process of memory consolidation.[131] Currently, technical limitations of surface EEG restrict our ability to measure these events in individuals with neurodegenerative diseases. Nonetheless, insights from mouse models of AD demonstrate that abnormalities in SWRs occur in association with AD pathology and may drive synaptic dysregulation and memory deficits.[132, 133] Additionally, animal experiments involving the human APOE4 gene suggest its expression has disruptive impacts on hippocampal microstructure and function, leading to inefficient SWR functions that may drive cognitive impairments.[134, 135] These data indicate there are potentially harmful neurodevelopmental effects of APOE4 expression on hippocampal-mediated gamma ripple functions, although additional studies are needed to confirm whether these effects occur in human neurodevelopment.

CONCLUSION

Disturbed sleep affects multiple systems in humans critical to maintaining brain health and preventing neurodegenerative diseases such as Alzheimer’s disease. In this review, we focused on mechanisms that potentially mediate the relationship between sleep and neurodegenerative processes such as AD. These mechanisms suggest targets to test approved and effective treatments of sleep disorders to prevent or delay AD. Multiple studies support the critical role of sleep in the production/release and clearance of proteins like Aβ and tau. However, there is increasing evidence implicating other mechanisms in the relationship between sleep and AD. Hyperphosphorylated tau is critical to the formation of intraneuronal tau tangles that lead to neuron loss and are closely associated with cognitive symptoms. Studies in mice and humans have recently implicated sleep-wake activity with protein phosphorylation in the central nervous system. Sleep loss also affects astrocytes and microglia. One of the ways astrocytes respond to this sleep loss is to increase phagocytosis of synaptic elements, likely in response to prolonged wake-associated synaptic activities, and potentially providing a mechanism for prolonged wakefulness to injury synapses. Sleep loss also activates microglia, promoting a neuroinflammatory response that could further lead to neural injury. Finally, orexin activity increases AD pathology. Fortunately, there are now DORAs approved for the treatment of insomnia that block orexin at its receptors and also decrease soluble AD biomarkers Aβ and p-tau.

We also extensively discussed how sleep affects memory-related neural circuits. Individuals with symptomatic AD have impaired memory and other cognitive functions. Increasing evidence supports that sleep disturbances also impair these circuits, potentially impairing cognitive performance earlier in AD pathogenesis. Sleep monitoring by EEG, even with a single-channel EEG device worn in the home, is able to measure the oscillatory events produced by these neural circuits. Monitoring sleep’s oscillatory events may be a marker to assess risk of symptomatic AD or to track the response to disease modifying therapies such as anti-amyloid monoclonal antibodies (e.g., lecanemab and donanemab).

Further research is needed, however, to determine how factors associated with both disrupted sleep and neurodegeneration, such as aging, sex, and SDoH, affect the mechanisms discussed in this review. Given the dual effects of many of these factors on both sleep and AD (e.g., age, sex, SDoH), studies will need diverse and representative samples of participants to examine the mechanisms discussed in this review, how these factors may interact, mediate, and/or modify them over time, and how these factors account for group differences in sleep disturbances and AD risk.