Sleep disorders and Alzheimer ’ s disease pathophysiology: The role of the Glymphatic System. A scoping review

Background: Alzheimer ’ s disease (AD) is highly intertwined with sleep disturbances throughout its whole natural history. Sleep consists of a major compound of the functionality of the glymphatic system, as the synchronized slow-wave activity during NREM facilitates cerebrospinal and interstitial long-distance mixing. Objective: The present study undertakes a scoping review of research on the involvement of the glymphatic system in AD-related sleep disturbances. Design: we searched Medline, Embase, PsychInfo and HEAL-link databases, without limitations on date and language, along with reference lists of relevant reviews and all included studies. We included in vivo, in vitro and post-mortem studies examining glymphatic implications of sleep disturbances in human populations with AD spectrum pathology. A thematic synthesis of evidence based on the extracted content was


Introduction
The lymphatic system serves the dual role of regulating peripheral fluid homeostasis, while facilitating immune surveillance (Moore and Bertram, 2018).The lymphatic system is organized in a network of channels, the density of which is proportional to the vascular network that accompanies it.This relates to the metabolic needs of the associated tissues (Liao and Padera, 2013).However, the lymphatic system of the Central Nervous System (CNS) differs significantly, by lacking conventional lymphatic vessels.In fact, the glymphatic system, as discovered by Iliff and colleagues (Iliff et al., 2012), is the lymphatic system of the CNS that resides in the perivascular space (PVS) of cerebral vasculature and is formed by the astrocytic endfeet, that are full of polarized aquaporin-4 receptors (AQ4).Within PVS, cerebrospinal fluid (CSF) circulates, being mixed continuously with interstitial fluid (ISF), along its way, through the function of AQ4.The lack of lymphatic vessels, along with the fact that glymphatic drainage is astrocyte-dependent, allow glia to exert their fundamental immunological properties in a compartment with no tight boundaries.Put simply, the territory of the CNS glymphatic seems to be the PVS, in which CSF is mixed with ISF the perivascular space.The perivascular space is divided into periarteriolar, pericapillary and perivenular spaces, which CSF flows within to permeate the brain parenchyma (Wardlaw et al., 2020).Perivascular space encompasses the whole CNS, constituting the conduit of the lymphatic system, where it is possible for glia to surveil all the areas.
In a recent review, we proposed a novel conceptual framework that underpins the functionality of the glymphatic system (Astara et al., 2023).We postulated that the function of the glymphatic system is regulated and mediated by a) arterial pulsation, whose pulsatile waveform reflects CSF flow in the periarterial space, b) respiration and posture, as they modify intrathoracic pressure which affect cerebral and meningeal venous drainage and perivenous CSF efflux and c) cerebral autoregulation, as it establishes arterial-venous equilibrium, which in turn adjusts CSF flow.
Nevertheless, the most fundamental element that regulates glymphatic functionality is considered to be sleep, and particularly slow wave activity during non-rapid eye movement (NREM) sleep.NREM introduces a state-dependency aspect of the glymphatic system, as CSF oscillatory motion is aligned to, not only cerebral vasculature, but also to neuronal slow waves, allowing long-distance circulation and clearance (Astara et al., 2023).
The benefits of sleep have undoubtedly been well-documented.For instance, it is intertwined with cognitive performance and emotional processing (Walker, 2009).Likewise, its implications in diseases, such as neurodegenerative disorders, have gained significant attention.Sleep is disrupted in neurodegenerative diseases, like Alzheimer's disease (AD).AD is the most common type of dementia in older individuals aged ≥ 65 years, characterized by the accumulation of amyloid plaques and hyperphosphorylated tau (Tahami Monfared et al., 2022).AD patients are even more likely to suffer from sleep disorders, and especially, sleep-disordered breathing (SDB) (Guarnieri et al., 2012), which, if untreated, are strongly associated with accelerated cognitive decline and conversion to a major neurocognitive disorder, even at a younger age (Lee et al., 2019;Leng et al., 2017).AD has been linked to altered neural activity oscillations, both during wakefulness and sleep (Hassainia et al., 1997).The architecture of sleep appears to be disrupted in patients with dementia, with stages 3, 4 and rapid eye movement (REM) sleep covering a lower-than-usual portion of the overall sleep duration.
Moreover, AD is characterized by global slowing of the posterior alpha rhythm as a reflection of degenerating dominant oscillatory networks active in the resting, awake brain (Prinz et al., 1982).The REM latency increases and the quality of sleep (measured as waking time in bed and number of night-time wakes) decreases.All these changes seem to become more intense in later stages of the disease.Furthermore, evidence suggest that AD-related epileptiform discharges occur more frequently during night sleep, especially during non-REM stages (Horváth et al., 2017).Better sleep consolidation has been proposed to attenuate the negative effects of the presence of the APOE ε4 allele towards developing AD, the annual rate of cognitive decline as well as the accumulation of neuropathological findings, such as neurofibrillary tangle density (Lim et al., 2013).More recent, prospective studies, confirm the previously mentioned associations and provide compelling support on the notion of the disrupted sleep patterns being involved in the defected clearance of biomolecules implicated in the pathology of AD (Lucey et al., 2019;Targa et al., 2021).
Sleep disturbances in AD seem to be a common element of the varied clinical, imaging and laboratory manifestations.In fact, several genotypes have been identified (Bellenguez et al., 2022), while recent studies based on data-driven and unbiased computational phenotyping of AD, are suggestive of the existence of AD phenotypes (Kim et al., 2020).Despite decades of research, the underlying mechanisms driving AD pathophysiology remain incompletely understood.Sleep disorders related with the glymphatic system have recently received increased attention in the natural history of Alzheimer's disease.While many previous studies have explored the relationship between sleep disturbances and AD, a comprehensive examination of the sleep-related glymphatic implications in the natural history of AD is lacking.Recent evidence has postulated that excessive protein aggregates may be the result of an imbalance between production and clearance mechanisms (Tarasoff-Conway et al., 2015).Since Aβ is known to be removed, in part, by the flow of ISF, the glymphatic system could be a crucial determinant of the removal rate of Aβ even in early-onset forms of AD as major pathophysiologic element in AD Thus, there is an increasing interest in investigating its clearance mechanisms as potential targets for AD treatment.
Several emerging studies support the fundamental role of sleep, as the optimal spatiotemporal environment for glymphatic clearance.Spatial environment refers to context, the paravascular and interstitial space, in which CSF is mixed with ISF, and which is enlarged by 60% during sleep, allowing more efficient clearance of metabolic waste (Xie et al., 2013).The temporal environment, in this context, stems from circadian optimization of the CSF influx into the interstitial space, as indicated by studies in murine models, which show the presence of a temporal dimension of AQ4 polarization in perivascular astrocytic endfeet (Hablitz et al., 2020).Hence, it becomes evident that glymphatic regulation is coupled with sleep physiology.In other words, glymphatic impedance and sleep disturbances become intertwined and may create a noxious milieu resulting in neurodegeneration.This is likely not specific to AD, but generally implicated in neurodegeneration, chronic or even acute, such as traumatic brain injury (Hayden, 2023;Opel et al., 2019).Relevant to AD, sleep disturbances may hinder glymphatic clearance, perpetuating, if not activating the disease process.
Primary research on the glymphatic system is thriving, and there is an abundance of multidimensional information on these complex interactions in various populations (preclinical and clinical) with AD pathology.However, there are important fundamental questions that remain unanswered, which would provide a framework for understanding and expanding this research.For instance, are there established glymphatic implications of sleep disturbances and sleep disorders?If so, how are these related to the onset and progression of AD? What are the therapeutic implications, and what translational approaches have been applied so far?
While previous systematic reviews have explored the potential connections between sleep disorders and the glymphatic system in healthy populations and Alzheimer's disease (Chong et al., 2022;Sangalli and Boggero, 2023), significant methodological limitations and inconsistencies exist and fundamental questions remain unanswered.It is particularly challenging to systematically evaluate and synthesize the primarily hypothesis-driven, diverse, and heterogeneous translational research applications in the field of glymphatic research.This challenge is likely exacerbated and becomes methodologically problematic when employing a systematic review methodology, which, by definition, seeks to address specific questions of a relatively narrow scope grounded in a well-defined concept.
We propose that this is mainly due to the lack of a conceptual framework for thematic synthesis and critical appraisal of the existing diverse evidence.Based on our recently published evidence-based theoretical framework for the functionality of the glymphatic system (Astara et al., 2023), we propose a scoping review, examining conceptually broader research questions that may give a new direction in terms of collecting and appraising the diverse existing evidence, but also allow a better understanding of key concepts, gaps and inconsistencies in the literature and the nature of their origin (for example methodological and conceptual problems), and thereby offer new directions for future research.
Therefore, we opted for a scoping rather than systematic review, to provide an overview of the existing literature, which is extremely difficult to methodologically collect, without proper parameterization.This scoping review aims to fill this knowledge gap by systematically examining the available evidence from various sources.By adopting a systematic review methodology, we aim to recognize the milestones in this research, identify research gaps, and synthesize the findings to offer a comprehensive overview of the sleep-related glymphatic implications in AD pathophysiology.
We aimed to identify and synthesize literature examining glymphatic implications in relation to sleep disturbances in the spectrum of AD manifestations.The glymphatic system was not established until recently (Iliff et al., 2012), and its involvement in neurophysiology remains elusive.In a relevant review (Astara et al., 2023) we distinguished the functional elements underlying the physiology of the glymphatic system.Here, taking a step further to the pathophysiological elements, we aimed to delve into sleep, as the fundamental functional element of the glymphatic that highlights its state-dependency.
We therefore focused on the following broad research questions: • Are there implications of glymphatic system dysfunction in ADrelated sleep disturbances?• What may be the processes though which glymphatic system dysfunction, as a result of sleep disturbances, may contribute to the onset and progression of AD spectrum pathology and the disease itself?• What are the therapeutic implications, and what translational approaches have been applied so far?

Methods
A scoping review was conducted and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) (Table S1) (Tricco et al., 2018).The stages include: 1) the identification of information sources relevant to the research question and the construction of search strategy, 2) a selection of studies eligible for the review, 3) data synthesis.

Data sources and charting, search strategy and protocol registration
We conducted a systematic literature search in several databases, including Medline (via Ovid), Embase, PsychInfo and HEAL-link (via Ovid, a Greek national database which operates as a consortium composed of 43 institutions and has access to over 20 content providers, https://www.heal-link.gr/en/home-2/).The initial search was implemented in December 2022 and there were no limits on date, language, and publication status.We also searched reference lists of previously published reviews and we manually reviewed the references of all included studies.
At least two independent reviewers jointly developed a data-charting form to determine which variables to extract.The two reviewers inspected all abstracts and then the full texts identified from the searches.The process was facilitated by the use of Rayyan (htt ps://www.rayyan.ai/)(Ouzzani et al., 2016).Conflicts were resolved by discussions between the reviewers, and in case of further disagreement, the decision for inclusion was made by discussion between the senior authors (AL, GV, MS).At least two reviewers independently decided whether the studies met the inclusion criteria.A form was developed and used blindly by at least 2 authors to confirm relevance and extract study characteristics such as first author name and publication year, population type, intervention type or characteristics of sleep disorders suggestive of glymphatic implications and the results of the analysis of their intercorrelation extrapolated to AD.
The following keywords were used: Alzheimer (and variations), neurodegeneration (and variations), protein aggregation, glymphatic (and variations), paravascular (and variations), Virchow-Robin Space (and variations), aquaporin-4 (and variations), arterial pulsation (and variations), convective bulk flow, cerebral autoregulation (and variations), brain homeostasis (and variations), sleep (and variations), circadian (and variations), fluid oscillations, systems biology.The search was limited to in vitro, in vivo and/or post-mortem human population and animal studies were excluded.
A follow-up search was conducted in December 2023 to identify any additional reviews published after the initial search.One study was found eligible and was included.A search for original studies and review articles in Google with no date restrictions was also conducted at this time; only the first 100 hits (as sorted by relevance by Google) were screened.No additional relevant original studies were identified.

Eligibility criteria
A two-stage screening process was used to assess the relevance of studies identified in the search.Studies were eligible for inclusion if they broadly described the involvement of the glymphatic system in sleep pathophysiology of either cognitively normal tested for AD biomarkers or AD patients.For sleep disturbances, all methods, either subjective (e. g., questionnaires) or objective (like polysomnography or actigraphy) were eligible.Regarding inclusion criteria for either cognitive normal or AD participants, all studies incorporating established biomarkers for AD (b-amyloid and tau proteins) in either CSF or plasma, were included.Additionally, populations diagnosed with any combination of clinical criteria and structural and functional imaging, like Magnetic Resonance Imaging (MRI) and Positron Emission Tomography (PET), AD biomarkers and cognitive assessment (Mini-Mental State Examination, Montreal Cognitive Assesment, etc) were eligible.Lastly, studies pooling samples retrospectively from previous cohort studies or databases from specialized departments, were included.Exclusion criteria were: i) all types of review studies and meta-analyses, ii) Animal studies, both in vitro and in vivo iii) Dementias other than AD, iv) research protocols focusing merely either on AD or sleep disturbances, without examining K. Astara et al. any parameters related directly or indirectly to the glymphatic system.

Data synthesis
A thematic synthesis of evidence based on the extracted content was applied and is presented below in a narrative way (Fig. 1) (Page et al., 2021).From a total of 3035 recordings, 70 papers were eligible for inclusion in the following thematic areas depicted in Fig. 2.

Protein aggregation and toxicity: indirect evidence of sleep disturbances' glymphatic implications and their links to CSF AD biomarkers in preclinical populations
Despite our extensive search, the existing literature on sleep disorders in AD was scarce and very heterogeneous in its areas of focus.We therefore found it necessary to propose a new, evidence-based conceptual framework for the thematic synthesis of this evidence.The proposed framework aligns with our scoping review's overarching objectives, with an organization that does not compromise the goal of comprehensively mapping the existing literature.
A recent metanalysis highlighted the involvement of amyloid toxicity in sleep disturbances in AD (Harenbrock et al., 2023).Aβ toxicity seems to impair the restorative mechanisms of sleep, by disrupting the functionality of specific regions and their interconnections.This disruption may even have clinical implications.It could be postulated that in AD, amyloid burden accumulates in specific sleep-active brain regions in a dose-dependent manner as a result of disruption of synaptic transmission (Wang et al., 2009, p. 200), which may extend to intergroup neural oscillations in networks (Palop and Mucke, 2010) and their functional connectivity (Bero et al., 2012).Notably, tau deposition, a major mechanism in AD pathogenesis, has also been associated with propagation by sleep deficiency, as well (Lv et al., 2022).Given the suggestive evidence of significant involvement of AD biomarkers in sleep disturbances, which may impact AD progression, we propose a data synthesis that indicates a roadmap connecting sleep disturbances to AD within the context of the glymphatic system.We therefore firstly examine protein aggregation in both acute and chronic sleep deprivation in cognitively normal participants, in order to identify possible glymphatic participation in the course of the disorder.First, we grouped together three studies involving healthy individuals, who were acutely sleep deprived and assessed for Aβ toxicity after one night of sleep deprivation (Lucey et al., 2018;Ooms et al., 2014;Shokri-Kojori et al., 2018).We then expanded our interest beyond amyloid and discussed four studies that referred to several markers of proteinopathies in chronic sleep disturbances (Fjell et al., 2018;Holth et al., 2019;Kim et al., 2022;Sprecher et al., 2017), as well as nine studies that explored the reflection of proteinopathy on sleep architecture (Brown et al., 2016;Ju et al., 2017Ju et al., , 2013;;Mander et al., 2022Mander et al., , 2015Mander et al., , 2013;;Olsson et al., 2018;Spira et al., 2013;Sprecher et al., 2015) in order to integrate their glymphatic interrelationship.Lastly, we focused on possible glymphatic implications in the conversion of preclinical to clinical AD.We found and examined eight studies introducing glymphatic involvement in protein aggregation and disease progression (Banerjee et al., 2017;Fultz et al., 2019;Gertje et al., 2021;Jeong et al., 2022;Mattia et al., 2003;Preston et al., 2003;Rajendran et al., 2006;Silverberg et al., 2001).

Aβ toxicity after one night of sleep deprivation in cognitive normal populations
Three studies investigated acute Aβ toxicity after one night of sleep deprivation, either partial or total, in healthy individuals (Lucey et al., 2018;Ooms et al., 2014;Shokri-Kojori et al., 2018) (Table 1).Lucey et al. showed a 30% increase in overnight Aβ production rate without affecting Aβ38/Aβ40 and Aβ42/Aβ40 ratios, suggesting a glymphatic effect irrespective of β-secretase activity (Lucey et al., 2018).In another study, this increase in amyloid accumulation was found to be region-dependent, namely in the right hippocampus and thalamus, as a consequence of the connection between Aβ production and neuronal activity (Shokri-Kojori et al., 2018).A randomized clinical trial in middle-aged men focused on the diurnal effect of Aβ production during both unrestricted and restricted sleep (Ooms et al., 2014).In normal sleep duration, a 6% reduction was observed, with the lowest Aβ42 values occurring at 10 AM.Restricted sleep diminished this effect on Aβ42 values.It should be noted that these effects were not replicated for Aβ40, P-tau, and T-tau, but a trendline was observed (Ooms et al., 2014).

Other protein-aggregates in sleep perturbation in cognitive normal populations
In order to identify alterations in the production/clearance rate in other proteinopathies, research should extend its focus to relatively chronic disturbances of sleep.So far, only four studies have addressed poor sleep and CSF-biomarkers, all in healthy populations (Fjell et al., 2018;Holth et al., 2019;Kim et al., 2022;Sprecher et al., 2017) (Table 2).
After one night of total sleep deprivation, CSF tau levels were significantly increased by over 50%, along with synuclein (Holth et al., 2019).An extended cohort study with a 5-year follow-up interval combined a sample of cognitively healthy and older adults with mild cognitive impairment (MCI) and supported these results, by applying CSF biomarker ratios (Kim et al., 2022).The p-Tau/Aβ ratio was associated with score severity in Informant-reported sleep disturbance questionnaire, along with accelerated changes in brain morphometry  and cognition.APOE4 also exhibited a significant interaction with the severity scale and led to faster brain atrophy (Kim et al., 2022).
Delving further into CSF biomarkers, a large-scale cross-sectional study by Sprecher et al. examined a variety of CSF biomarkers along with self-reported sleep disturbances among healthy older adults with a family history of sporadic AD.While single-pathology markers like tau and amyloid were associated with sleep adequacy, when expressed in ratios, they were also associated with daytime sleepiness.MCP-1/Aβ42, and YKL-40/Aβ42 ratios, which are indicators of neuroinflammation and astroglial activation instigated by amyloid plaques, were also correlated with greater somnolence and sleep adequacy respectively.It is worth pointing that no significant interactions with confounding factors like APOE e4 status, age and comorbidities affected variance in CSF markers.In addition, neither sleep-disordered breathing (SDB) nor sleep duration were associated with CSF biomarkers.This may seem like a contrasting result, given the variety of evidence that supports their interrelationship.Partly, these results were replicated in Aβ positive but cognitively healthy older adults, with CSF levels of total-and P-tau, as well as YKL-40-predicting poorer sleep in a 3-year follow-up, independently of brain atrophy (Fjell et al., 2018).This outcome could be Table 2 Characteristics of studies that assessed the production/clearance rate in other proteinopathies.Abbreviations: AD= Alzheimer's disease, PSG= polysomnography, IRSD= inventory-informant-reported sleep disturbance, MCP-1 = monocyte chemoattractant protein, NfL= neurofilament light chain, GFAP= astrocytic glial fibrillary acidic protein, YKL-40 = Chitinase-3-like protein 1 (alternatively: CHI3L1).

Table 3
Indications of perturbed sleep architecture due to AD biomarkers.Abbreviations: CSF= Cerebrospinal fluid, PSG= polysomnography, MHPG= 3-methoxy-4hydroxyphenylglycol, HVA= homovanillic acid, 5-HIAA= 5-Hydroxyindoleacetic acid, MTL= medial temporal lobe, NfL= neurofilament light chain, NSE= neuronspecific enolase, FABP= fatty-acid-binding protein GFAP= astrocytic glial fibrillary acidic protein, YKL-40 = Chitinase-3-like protein 1 (alternatively: CHI3L1), S-100B= S100 calcium-binding protein B, sTRM2 = soluble triggering receptor expressed on myeloid cells 2, mPFC= medial prefrontal cortex, fMRI= functional MRI.attributed to the relatively mild severity of SDB and the fact that it was estimated with self-reports and not the gold standard of PSG.Notably, from the results above, APOE phenotype failed to be associated with CSF biomarkers of AD.APOE e4 has been documented as a fundamental genetic risk factor for AD, suggesting an influence in both amyloid and tau levels (Yamazaki et al., 2019), but studies in preclinical AD have offered mixed results, (Herukka et al., 2007;Hong et al., 2022;Mofrad et al., 2020), perhaps due to varying sex-dependent severity (Liu et al., 2021).Further longitudinal studies with follow up data and serial amyloid and tau PET imaging data would clarify impacts of APOE e4 on AD-related biomarkers.

The effects of protein aggregation on sleep architecture in cognitive normal populations
EEG alterations reflect neuronal activity, which in turn is locally associated with Aβ production.Multiple studies have addressed this connection, supporting the hypothesis of glymphatic impedance in clearing metabolic waste (Brown et al., 2016;Ju et al., 2017Ju et al., , 2013;;Mander et al., 2022Mander et al., , 2015Mander et al., , 2013;;Olsson et al., 2018;Spira et al., 2013;Sprecher et al., 2015).During sleep, especially deep NREM sleep, metabolic waste is optimally cleared, leading to increased amyloid concentration and burden on brain areas associated with NREM regulation (Kang et al., 2009).Congruent with these findings, it could be suggested that deep NREM fragmentation might serve as an early pathophysiological element that develops -if not triggers -AD.
Three studies incorporated self-reported sleep quality and amyloid burden estimation, through CSF analysis (Brown et al., 2016) and [ 11 C] PiB PET (Spira et al., 2013;Sprecher et al., 2015) (Table 3).Brown et al. associated sleep latency with higher levels of brain Aβ, irrespective to APOE ε4 status, in healthy individuals above 60 years old (Brown et al., 2016).Spira et al. exhibited, correlation of shorter sleep duration, as well as worse perceived sleep quality, with greater amyloid burden.These associations remained significant after excluding participants with either MCI or dementia (Spira et al., 2013).Concomitantly, Spercher et al. associated diminished sleep quality with increased amyloid burden, but they differ in the association between sleep duration and β-amyloid (Sprecher et al., 2015).This could be attributed to the younger age of the sample included, as abnormalities may become evident in PET later in life (Palmqvist et al., 2016).Nevertheless, they incorporated more detailed sleep domains, which demonstrated an association of amyloid burden with somnolence.In addition, there was no association with symptoms consistent with SDB, probably again due to mild severity and not applying objective estimation (PSG).
Regarding actigraphy studies, one applied actigraphy alone (Ju et al., 2017) and another a combination of actigraphy and PSG (Ju et al., 2013).Both studies seem to support the results of the previously mentioned studies on self-reported quality of sleep, with Ju YES et al. extending the relationship of amyloid depositionbad sleep quality-frequent napping, in the preclinical stage of AD (Ju et al., 2017).Ju YS et al. aimed to elucidate these data by highlighting specifically synaptic degeneration as a contributor to increased amyloid burden and NREM disruption, rather than glia activation or general protein clearance mechanisms (Ju et al., 2013).Although tau did not show a correlation with slow-wave activity disruption like β-amyloid, it was associated with actigraphic parameters of sleep quality.Overall, these results may suggest that β-amyloid undermines quality of NREM earlier in the natural history of AD, while tau has more prolonged effects on sleep quality over an extended time interval.
Four studies aimed to unveil alterations on NREM sleep (Mander et al., 2022(Mander et al., , 2015(Mander et al., , 2013;;Olsson et al., 2018).Three studies displayed the effect of amyloid burden, specifically on mPFC, a brain area highly active during deep NREM, as it generates slow-wave activity.Mander et al., 2013 demonstrated an age-related atrophy of mPFC, which consequently predicted the canonical age-related decline in NREM slow-wave activity (Mander et al., 2013).This effect was partially implicated in the long-term memory impairment.Two years later, the same team identified increased amyloid burden as a contributor for mPFC atrophy and disruption of slow-wave activity, which exacerbated worse performance in memory retention tests (Mander et al., 2015).A few years later, Mander et al., 2022 offered further insight into the immunological continuum of neurodegenerative progress within the context of sleep disturbances (Mander et al., 2022).CSF biomarkers of astrocytic and microglial activation (GFAP, YKL-40, S100B, and log sTREM2) were consistently associated with markers, such as Aβ proteins, tau proteins and markers of neuronal integrity like α-synuclein, neurogranin and NfL.In addition, deficits in sleep spindle expression, particularly at faster frequencies over fronto-central EEG derivations, were also associated with these biomarkers.These findings suggest that even before β-amyloid pathology becomes evident, glial activation can facilitate early neurodegeneration by facilitating tau phosphorylation and synaptic degeneration-associated sleep spindle.Clinically, this association is related to memory deficits in cognitively unimpaired older adults at risk for AD (Mander et al., 2022).
However, the study by Olsson and colleagues showed that partial sleep deprivation for 8 nights affected only orexin levels in young healthy individuals, while CSF Аβ remained unchanged.PSG analysis exhibited a significant reduction in all sleep stages, but NREM (Olsson et al., 2018).This inconsistency could be attributed to the methodological design of the study, which focused on partial and not total sleep deprivation.Therefore, since slow-wave sleep (SWS) was not affected by partial sleep deprivation, it could be suggested that clearance of metabolic waste remained effective.Afterall, orexin seems to be responsible for calibrating REM sleep, without interfering with NREM (Feng et al., 2020;Kantor et al., 2009).

Highlighting the route of amyloid, does not clearly highlight the glymphatic history of AD
Glymphatic impedance drives, to some extent, the deleterious effects of protein aggregation on sleep even in preclinical populations.Then, we aimed to identify the possible involvement of the glymphatic system in the conversion from cognitively normal to clinical AD.The imbalance in the production and clearance of Aβ is a crucial triggering point for the onset of AD.There are implications in which Aβ causes toxicity, either in the interstitial space, leading to neuronal apoptosis with the formation of senile plaques, or intravascularly, inducing cerebral amyloidosis (Reiss et al., 2018).Based on the existing evidence (Table 4), it could be hypothesized that the existence of Aβ in these two different compartments of the CNS might reflect the glymphatic system's effort to clear these toxic aggregates.However, this finding failed to be extended to the principal imaging marker of glymphatic functionality, namely enlarged perivascular space (EPVS).
Three studies were included to offer some clarity.Rajendran et al. utilized human post-mortem tissues from 3 AD participants and investigated the possible involvement of exosomes (Rajendran et al., 2006).Their results have shown accumulation and enrichment of exosomal proteins around amyloid plaques, indicative of exosome-associated Aβ burden, before entering the perivascular space.In murine models, only synaptic transmission was related to secreted Aβ in the ISF (Cirrito et al., 2005).Perhaps, the exact route of Aβ efflux may contain inherent information that mechanistically predicts the ability of optimal clearance and, hence, the evolution of the AD.With the extraction of protein aggregates, CSF production rate has been found reduced in a small sample of AD patients, suggestive of glymphatic impedance (Silverberg et al., 2001).Suboptimal clearance results in entrapment of Aβ along the whole glymphatic route.Preston et al. identified that Aβ in the brain parenchyma were continuous with Aβ in vascular walls but not vice versa (Preston et al., 2003).Therefore, Aβ elimination impairment does not affect interstitial space only, but also the perivascular route, in which it enters and further on drains along the walls of cortical arteries to leptomeningeal arteries.The co-existence of Aβ in both perivascular and in interstitial spaces is accompanied by coupled electrophysiological and hemodynamical oscillations, as indicated in a small study with 14 K. Astara et al.AD patients (Mattia et al., 2003).An increase in theta, along with a decrease in alpha frequency bands with a topographic distribution over the central and posterior EEG scalp regions, were associated with bilateral reduction in the temporoparietal and sensorimotor regional cerebral blood volume (rCBV).The total level of hypoperfusion was associated with the EEG findings (Mattia et al., 2003).
Notably, Fultz and colleagues replicated these results in healthy individuals, by offering insights in glymphatic kinetics (Fultz et al., 2019).They observed that CSF oscillation mirrored NREM rhythm, whereas during wakefulness it was comparatively attenuated and reflecting respiratory motion.Furthermore, with blood oxygen level-dependent (BOLD) signal after functional magnetic resonance imaging (fMRI) indicated hemodynamic anticorrelation with CSF motion.Thus, with EEG oscillations preceding, neural, CSF and BOLD dynamics were found coupled during sleep (Fultz et al., 2019).
Regarding human studies in AD, the published data seem to validate the above findings, but not associated with proteinopathies.Two studies examined EPVS in basal ganglia (BG), centrum semiovale (CSO) and hippocampi (HP) implementing PET imaging and two CSF analysis.BG-EPVS was correlated with specific deficits in cognitive functions (language and frontal/executive function), regardless of amyloid burden (Jeong et al., 2022), and it predicted AD likelihood (Banerjee et al., 2017).Hippocampal EPVS correlated with general cognition decline and AD diagnosis, independently of amyloid burden, in both PET and CSF analysis (Gertje et al., 2021;Jeong et al., 2022).CSO-EPVS was associated with the rate of cognitive decline and AD diagnosis and advancement, as well (Banerjee et al., 2017;Jeong et al., 2022).These results are in contrast with one recent pilot study in which middle-aged participants with preclinical AD were divided into two groups based on their amyloid status (positive or negative) and had their EPVS associated with CSF analysis (Vilor-Tejedor et al., 2021).Increased EPVS in CSO was associated with tau burden in Aβ + participants, while EPVS in both CSO and BG maintained their correlation with Аβ40, independent of Aβ status.These findings indicate a glymphatic involvement in early AD (Vilor-Tejedor et al., 2021).Similarly, in another study including patients with neurodegenerative diseases (with a minor proportion being AD) both CSO-EPVS and CSF-AQ4 levels correlated with CSF-tTau.Notably, BG-EPVS did not exhibit significant association, implying that mainly CSO-EPVS is implicated in neurodegeneration.Evidence demonstrating association of BG-EPVS with neurodegeneration may be driven by confounding factors like aging and cardiovascular burden (Bown et al., 2022;Evans et al., 2023).
Based on the above, it could be postulated that glymphatic impedance may perpetuate protein aggregation in a non-specific manner, as it extends from preclinical to clinical neurodegeneration.Furthermore, glymphatic dysfunction seems to be a consequence of non-specific sleep disturbances related to SWS disruption.Nevertheless, further studies incorporating imaging markers of the glymphatic system and CSF analysis could clarify the glymphatic trajectory of protein aggregates in AD-specific populations.

Potential glymphatic markers in SDB
A study with OSAS patients suggested the potential causal association of the glymphatic drainage and the cognitive decline (Wang et al., 2023).It was displayed that PVS was increased in bilateral frontal cortex and the basal ganglia, as well as bilateral lateral ventricles and the fourth ventricle compared to controls, while all consistently indicated that PVS enlargement occurred especially in patients with severe OSAS or hypoxemia.Cognitive assessment tests were negatively correlated, to a varying degree, with the aforementioned areas, while AHI and ODI were both positively correlated with bilateral frontal cortex (Wang et al., 2023).Notably, CPAP therapy for 1 month did not show significant differences (Wang et al., 2023).Furthermore, two studies incorporated diffusion tensor imaging (DTI) with the perivascular space (DTI-ALPS) method in OSAS patients (Lee et al., 2022;Roy et al., 2022).The ALPS index is an indicator of glymphatic system activity, by estimating water diffusion.When close to 1, it represents minimal water diffusion along the PVS.In both studies, DTI-ALPS in patients with OSAS was significantly lower.Lee and colleagues further demonstrated that DTI-ALPS index was negatively correlated with the AHI and ODI in NREM sleep (Lee et al., 2022).Since perivascular diffusivity was found reduced in OSAS, it was investigated in detail, deconstructing it into its axonal bundles (Roy et al., 2022).Periventricular projection and association diffusivity analysis unveiled a correlation of SaO2 with fiber projections from one hemisphere across the other and AHI with association fiber areas from deep, i.e. white matter, to superficial, i.e. neocortex (Roy et al., 2022).These findings may be attributed to the hypoxic episodes induced in apneas, as murine stroke-models have shown diminished water diffusivity in all axonal bundles in the ischemic areas (Shereen et al., 2011).

AD biomarkers in patients with OSAS
There exists a significant variability in AD biomarkers among OSAS patients, and the results from several studies are inconsistent.
To start with, one study included middle-aged, cognitively healthy individuals and explored amyloid burden related to breathing disorders.SDB was associated with greater amyloid burden in the left precuneus, posterior cingulate, calcarine, and cuneus regions (André et al., 2020).Similar results regarding regional amyloid burden have been replicated (Yun et al., 2017).In particular, OSAS severity was also correlated with increased brain amyloid burden, namely decrease in CSF Aβ42 and an increase in cortical PiB-PET uptake, in cognitively normal elderly, longitudinally (Sharma et al., 2018).Hypoxia was identified as a principal contributor in amyloid burden, probably in concordance with increased metabolism observed in these areas (André et al., 2020).A recent study expanded the association of SDB with regional atrophy in Аβ positive and cognitively unimpaired individuals (André et al., 2023).Apnea-hyponea index (AHI) and ODI severity were correlated with lower volume in specific regions of the medial temporal lobe, in which AD clinical manifestation onsets.Notably, despite the majority of participants having objectively medium or severe SDB, subjective evaluation of symptomatology was not proportionate (André et al., 2023).
AD-biomarkers have also been examined in the peripheral blood of OSAS patients, with the evidence being rather robust.Higher plasma levels of Aβ40, Aβ42, t-tau, p-tau were found in OSAS groups (Bhuniya et al., 2022;Bu et al., 2015;Kong et al., 2021).From all the SDB indices examined, the ones regarding oxygen saturation (oxygen desaturation index (ODI) and mean SaO2) were found consistently significant with all amyloid and tau proteins (Bhuniya et al., 2022;Bu et al., 2015;Kong et al., 2021).From a sleep-architecture perspective, amyloid burden was negatively associated with N3 and REM stages, and positively with NREM-AHI (Bhuniya et al., 2022;Jackson et al., 2020).
However, evidence from CSF analysis is inconclusive.In some comparative studies, CSF concentrations of Aβ40 have been found either positively (Przybylska-Kuć et al., 2019) or negatively (Ju et al., 2016) associated with SDB indices.Concomitantly, CSF Aβ42 exhibited either no differences (Ju et al., 2016;Przybylska-Kuć et al., 2019), decreased levels (Fernandes et al., 2022) or increased levels in OSAS populations (Liguori et al., 2017).CSF t-Tau and p-Tau, which are typically increased in AD, were found decreased in OSAS in one study only, along with other neuronally-derived proteins (Ju et al., 2016).
An interesting study by Osorio et al. not only replicated the aforementioned results but also stratified them based on APOE allele phenotypes (Osorio et al., 2014).Aβ42 showed a significant negative correlation with the APOE Е2 + group, a non-significant positive correlation with the APOE ε3 + group, while no differences were observed in the APOE ε4 + group.Conversely, p-tau was only positively associated with the APOE Е3 + group (Osorio et al., 2014).Actigraphically calculated AHI was not correlated with any of the CSF AD biomarkers.This could be attributed to either underestimation of AHI compared to the gold-standard PSG-calculation, or to the relatively mild severity of SDB in the population examined.
AD risk in SDB patients has been associated with specific indexes, both AHI and nocturnal oxygen saturation, in REM and NREM sleep (Tsai et al., 2022).In a large-scale, longitudinal study, the presence of OSAS in the preclinical stages of AD was associated with lower levels of CSF concentrations of T-tau and P-tau, as well as brain florbetapir PET amyloid load and higher CSF Aβ42 levels.Obstructive sleep apnea status influenced the annual change of CSF Aβ42, T-tau, and P-tau, but not florbetapir PET amyloid load (Bubu et al., 2019).Conversely, OSAS severity, in terms of AHI and ODI, was associated with increased levels of exosomal Aβ42, t-tau and P-T181-tau (Sun et al., 2022).
Three studies implemented PET scan to estimate amyloid burden in OSAS patients (Elias et al., 2018;Fernandes et al., 2022;Jackson et al., 2020).Aβ was found increased in OSAS, severity (Jackson et al., 2020), BMI and APOE ε4 dependent (Elias et al., 2018).Fernades et al. suggested that the transition to the clinical manifestation of AD, accompanied by reduced levels of CSF Aβ42 and dysregulated glucose  metabolism in the pathologically implicated brain areas, is significantly associated with the presence of OSAS (Fernandes et al., 2022).Likewise, a post-mortem study replicated this SDB-driven hypothesis of AD, by correlating sleep apnea severity with Aβ burden in the hippocampus, the region incriminated for the clinically evident AD (Owen et al., 2021).Interestingly, NFT and amyloid burden, although prevalent in the majority of brainstem samples (a region thought to be where the preclinical stages of AD start) failed to show significant differences with sleep apnea severity (Owen et al., 2021).
To sum up, the results show an overall tendency of increased amyloid and tau burden in OSAS patients, deconstructing the natural history of AD.Notably, hypoxic parameters were strongly evident in the association of OSAS and AD.Experimental studies in murine models support the fact that hypoxia may lead to both overproduction of Aβ and its reduced degradation in the brain.Hypoxia influences amyloid by increasing beta-secretase for Beta-Secretase 1 (BACE-1) gene expression in a HIF-1a-dependent way (Shiota et al., 2013).Similarly, genetic knockout of BACE-1 gene resulted in reduction of β secretase activity and, thus, of amyloid production (Roberds et al., 2001).Additionally, hypoxia leads to neuronal cell death mediated by tau hyperphosphorylation (Fang et al., 2019).
However, an inconsistency in findings was observed primarily in CSF, and to a lesser extent in serum levels of AD biomarkers.One could hypothesize that this issue could be attributed to a glymphatic-driven etiology.Respiratory effort increases around apneic episodes, leading to elevated intrathoracic and intracranial pressure.The outcome is increased venous pressure and impedance in the outflow compartment of the glymphatic system.As a consequence of this, amyloid burden residing in the ISF cannot be effectively cleared through the CSF, which may explain the lower concentrations peripherally (CSF and plasma).This hypothesis offers new insights into the natural history of the disease, complementing, or even challenging the strictly pathological classification, and is worthy of further research as its core elements remain to be elucidated.

Circadian dysregulation
Sleep is driven by a homeostatic function, namely sleep pressure (mediated by adenosine accumulation) and the circadian rhythms (Deboer, 2018).Sleep-wake cycle shows circadian features, and is mainly driven by the activity of the hypothalamic suprachiasmatic nucleus (SCN) (Vansteensel et al., 2008).The SCN controls circadian secretion of melatonin, the hormone responsible for sleep onset, entrained to the light/dark cycle.Allusions to the regulation of the glymphatic system by circadian rhythms have primarily been made by Hablitz and colleagues, who indicated a time-dependent polarization of AQ4 in the perivascular astrocytic endfeet (Hablitz et al., 2020).

Circadian dysregulation of the glymphatic in healthy population
We therefore explored the evidence referring to the glymphatic regulation in the context of circadian rhythmicity, starting with healthy individuals, without dementia.We identified four eligible studies (Table 7).
In the brain parenchyma, CSF production follows a circadian pattern as shown by Nilsson et al.In a study by Nilsson et al., CSF production was measured for 24 h in 6 healthy volunteers using MRI.The observed circadian variations indicated that there is a minimum CSF production around 6 pm and a maximum around 2 am (Nilsson et al., 1992).Moreover, Ly and colleagues uncoupled the two sleep driving forces and examined cortical excitability (Ly et al., 2016).They included healthy participants who underwent prolonged (29 h) wakefulness under constant conditions, cortical excitability, measured as electric neuronal activity after TMS, followed clear circadian dynamics for the whole duration of the test (Ly et al., 2016).The group also observed similar circadian patterns in the brain wave state (EEG measurements of theta power) and the behavior of the participants (vigilance task).Interestingly, the observed variations in cortical excitability were significantly more pronounced in the participants with the largest cortisol secretion amplitude, a feature which is thought to be an indicator of how strong the circadian wake promoting signal will be.
Recent studies were able to unravel the potential molecular basis of the sleep disturbance in AD.Genetic alterations in the astrocytic end-feet expressing AQ4 protein, emerged as a potential key player.The hypothesis describes that these mutations in the AQ4 gene impair the brain's CSF flow rate and its ability to successfully clear Aβ.In a genetic data analysis performed by Rainey-Smith et al., the group identified specific AQ4 single nucleotide polymorphisms (SNPs) that were associated with reduced sleep quality and an increased Aβ burden (Rainey--Smith et al., 2018).Interestingly, in individuals homozygous for some of the identified SNPs, the examined sleep parameters were significantly worse and the levels of Aβ burden measured with PiB-PET were close to mild AD diagnosis.In addition, in a more recent study carried out by Larsen et al., a whole AQ4-haplotype (low AQ4-expressing HtMi variant), which demonstrates decreased AQ4 expression, was associated with increased slow wave energy (SWE) during NREM that were necessary to initiate proper glymphatic flow.The SNPs included in this haplotype have been previously associated with AD and cognitive decline, showcasing a direct link between reduced AQP4, disturbed  sleep intensity and AD risk (Ulv Larsen et al., 2020).

Circadian dysregulation of the glymphatic in AD population
Neurons secreting vasoactive intestinal polypeptide (VIP), reside within SCN, consisting of 10% of the population of SCN neurons.VIP neurons are integral to maintaining circadian rhythmicity by receiving light from the retina and maintain a 24-hour activity (Ee and Ry, 2001).In AD, either SCN volume reduction or interruptions in signaling, dysregulate circadian rhythms and sleep-wake cycle (Homolak et al., 2018).We approached glymphatic perturbation by including 3 thematic areas: volumetric degeneration of SCN (3 studies identified), circadian dysregulation by actigraphic analysis and altered gene expression (7 studies identified) (Table 8).

Volumetric loss of SCN in AD
Several studies have indicated a substantial loss of neurons along the circadian route, both age-related and AD-pathology-related (La Morgia et al., 2016;Wang et al., 2015;Zhou et al., 1995).SCN changes varies between the studies, which show either decreased population of neurons (Zhou et al., 1995) or no significant differences compared to non-demented controls (Wang et al., 2015).This could be due to the underestimation of the extent of the SCN degeneration, as the VIP neurons involved occupy a small volume relative to the total area.Therefore, as can be deduced from the actigraphic data, even the underestimated loss of VIP neurons leads to circadian dysregulation, like phase delay in AD patients (Wang et al., 2015).Interestingly, the number of VIP neurons in the SCN was positively associated with the normalized amplitude of the 24-hour activity rhythm (Wang et al., 2015).Similarly, another in vivo and post-mortem study investigating other photoreceptors, the melanopsin retinal ganglion cells (mRGCs), showed both age-related findings, like optic neuropathy, and AD-related ones, with abnormal mRGC morphology with consistent amyloid deposition.In the subgroup that underwent actigraphy, despite the small sample size, sleep efficiency was reported significantly reduced (La Morgia et al., 2016).All findings support the plausible role of impaired reception and encoding of light stimuli contributing to the circadian rhythm dysfunction in AD.

Circadian alterations as identified by actigraphic analyses and genetic expression
SCN degeneration is correlated with circadian rhythm impairment which disturbs the sleep/wake cycle in AD.In a study where 11 mild and 9 moderate AD patients were compared to 12 healthy individuals on their propensity to daytime sleepiness when alertness should be at its peak, results indicated a significantly higher propensity in AD patients compared to the healthy group that was even correlated to the severity of the disease (moderate AD patients demonstrated significantly higher levels of daytime sleepiness than mild AD patients).Moreover, excessive daytime sleepiness was correlated with increased cognitive dysfunction, further highlighting the importance of circadian dysregulation in the progression of AD (Bonanni et al., 2005).
Circadian dysregulation is evident even in the preclinical stages of AD.A large cross-sectional study including cognitive normal elderly unveiled and association of amyloid burden with rest-activity rhythm fragmentation, through reported increased intradaily variability (Musiek et al., 2018).Similarly, in a rather small sample of AD, actigraphic analysis showed positive correlations between the severity of rest-activity rhythm fragmentation and the reduced 24 h amplitude, findings possible mediated by a diminished amount of total melatonin secretion (Mishima et al., 1999).
In AD, two post-mortem and one in vivo studies investigated changes in circadian clock gene expression (Cermakian et al., 2011;Cronin et al., 2017;Wu et al., 2006).Wu et al. found altered clock gene expression (regarding hPer1, hbeta1-ADR, hCry1 and hBmal1 mRNA) in both preclinical and clinical AD (Wu et al., 2006).Cronin et al. incorporated not only histopathological tissues from AD brain samples, but also in vitro experimentation with AD patient fibroblasts and identified possible epigenetic alterations regarding circadian dysregulation in AD (Cronin et al., 2017).Cermakian and colleagues expanded their research interest beyond the pineal gland and found variable clock gene expression in the cingulate cortex and the bed nucleus of the stria terminalis, in both control and AD samples.Although AD samples exhibited similar 24 h rhythmicity, inter-gene and inter-region correlations were attenuated (Cermakian et al., 2011).
Pillai and colleagues examined the immunological effects of circadian dysregulation in mild AD patients, by analyzing a multiplex panel of CSF and serum biomarkers of immunity, inflammation and neurodegeneration, and associating them with certain indexes of circadian rhythm (Pillai et al., 2021).Their unique results offer some insight in the complex intercorrelated immunological elements between AD and sleep disturbances.However, more extended studies are required to replicate these results.

Interventions
After exploring the available evidence of glymphatic implications of sleep disturbances in AD pathophysiology, we sought for potential interventions that could reinforce the management of AD.We found 13 studies that examined optimization of breathing during sleep (like CPAP) and re-regulation of circadian rhythms (like supplements) and demonstrated explicit efficacy as adjuvant therapeutic options (Table 9).

Acute effects of CPAP
The only currently published study directly addressing the glymphatic system aimed at identifying alterations of the perivascular space (PVS) in OSAS patients after one month of CPAP therapy (Wang et al., 2023).The study reported that the enlargement of PVS in the bilateral frontal cortex, the basal ganglia, bilateral lateral ventricles and the fourth ventricle was associated with OSAS severity and the existence of hypoxemia.However, CPAP therapy did not significantly change PVS volume (Wang et al., 2023).Perhaps, the relative short duration of therapy, along with the participants' unknown amyloid status (which in the above chapters was found correlated with PVS enlargement and highly prevalent in OSAS patients) may have underestimated the true potential of the results.
In groups of patients with both OSAS and AD, both studies that examined acute effects of CPAP therapy exhibited potential benefits.Rosenzweig et al. implemented CPAP therapy, along with phycoeducation (guiding participants for a better mental well-being through psychotherapeutic strategies) and lifestyle modifications for one month, and aimed at associating such EEG-architecture alterations with neurocognitive ones (Rosenzweig et al., 2016).Their results suggest active neuroplastic reorganizing changes within the long-range cortico-cortical pathways, and within cortico-striatopallidal-thalamocortical loop volleys.Although this study does not directly involve AD patients, it offers a pathophysiological connection with AD, as the disruption of thalamocortical oscillators has been implicated in the manifestation of AD (Jagirdar and Chin, 2019;Katsuki et al., 2022).Cooke and colleagues showed that not only after one night of treatment, but also 3 weeks of therapy, sleep architecture changed by enhancing Stage 2 sleep and deep sleep after one night and by decreasing WASO, Stage 1, arousals, and increasing Stage 3 after 3 weeks.A trend of increase in sleep onset was also exhibited (Cooke et al., 2009a).Similar results have been replicated in patients with OSAS but without dementia, with reduced portion of Stage 1 sleep and more percent REM sleep after one night of treatment (Loredo et al., 2006), as well as increased amounts of SWS (Fietze et al., 1997;Parrino et al., 2005).Ancoli-Israel et al. demonstrated in the first and only so far randomized double-blind placebo-controlled trial, that the group of AD-OSAS patients after 3 or 6 weeks of treatment, had their cognitive capabilities of episodic verbal learning and memory and some aspects of executive functioning such as cognitive flexibility and mental processing speed improved (Ancoli-Israel et al., 2008).This evidence may suggest an OSAS-related onset of AD and not vice versa.Glymphatic dysregulation is in line with such hypotheses, as the suboptimal clearance of metabolic waste due to sleep disturbances, leads to accumulation of protein aggregates in the interstitial space, which further worsens CSF-ISF mixing.
In a study cited elsewhere, regarding the correlation of amyloid and tau burden in OSAS, a small subgroup of patients was included which reported the use of CPAP for unknown time interval (Elias et al., 2018).The reported use of CPAP had no effect on cognitive assessment or amyloid findings (Elias et al., 2018).

Long-term effects of CPAP
In a small group of patients with severe SAS and mild-to-moderate AD underwent CPAP therapy for 6 months and exhibited a deceleration in further cognitive decline, in terms of the median annual cognitive decline (Troussière et al., 2014).Concomitantly, OSAS patients who received CPAP for one year, had their neuronal-derived exosome levels of Aβ42, T-tau, and P-T181-tau significantly lower than age-and genger-matched controls; biomarkers that had been associated with cognitive decline (Sun et al., 2022).Similarly, patients with OSAS who received CPAP or other therapy for averagely 3.9 years, exhibited a significantly reduced AD development, while the presence of OSAS was significantly associated with a higher incidence of AD (Tsai et al., 2020).The results are consistent with those of previous studies and confirm the association between OSAS and AD.Regarding cognitive status, sustained use of CPAP therapy (13.3 months on average), Cooke et al. report an increased likelihood of improvement in executive functioning, but no association with a significant effect on the rate of global cognitive decline.It should be noted that in this study, participants consisted of a small group who were not randomly assigned (continued use versus discontinuation of CPAP) (Cooke et al., 2009b).Alongside the aforementioned benefits, CPAP+ group's depressive symptoms, daytime sleepiness and subjective sleep quality either showed less deterioration or improved from the end of the RCT to follow-up compared to the CPAP− group.Depressive symptoms are important to be considered, as Ayalon et al. showed better compliance of CPAP in AD-OSAS patients who performed better in the Cornell Scale for Depression (Ayalon et al., 2006).Despite the small sample which contained mild-to-moderate AD and moderate OSAS, excluding any possible correlation with the severity of the diseases, to our knowledge, this is the only study that examined adherence of CPAP therapy in such group.
However, in a larger-scale longitudinal cohort by Osorio et al., although the presence of SDB was correlated with onset of cognitive decline in both MCI-to-AD and AD groups, CPAP therapy did not exert significant improvement (Osorio et al., 2015).In addition, this study displayed a stronger association in SDB+ MCI-to-AD group, than in SDB+ AD-specific group.When both groups were adjusted for potential confounders, significance became a trend in both groups (Osorio et al., 2015).This discrepancy with the above results may be attributed to the greater sample size, which offered the opportunity to unveil possible correlations within AD-severity strata.All the previous studies contained rather small sample groups, with most of them being non-blinded.It should be noted, though, that in this large cohort study, the investigation concerned the umbrella-term of SDB and not particularly its most prevalent and severe syndrome, as above.This signifies that severity-related strata did not concern breathing disorders, implanting a potential intrinsic bias.

Alternative or adjunctive therapies
When CPAP fails, is unavailable, or patients are not adherent, alternative treatment methods are available.Only one study examined an alternative therapy, namely uvulopalatopharyngoplasty (UPPP), and the incidence of dementia in OSAS patients.UPPP is the best alternative and most commonly performed surgical procedure for OSAS, but is less efficient than CPAP (Cho et al., 2018).The study by Cho et al. contained all types of dementia and showed that untreated OSAS increased the likelihood of dementia.UPPP helped to lower all dementia incidence.However, the strongest correlation was found with vascular dementia, perhaps due to the hypoxic preconditioning that allows quicker adaptation in ischemic events (as witnessed in vascular dementia), as the authors suggest.It should be emphasized that the candidates for surgical treatment of OSAS more likely are younger and with less comorbidities, making their profile more suitable to respond to treatment and to generally slow down the onset of neurodegeneration.
Since circadian dysregulation plays an independent and interdependent role in AD pathophysiology, supplemental melatonin agents were a considered potentially beneficial, as well as an agent with direct effects on glymphatic clearance.Wade et al. tested prolonged-release melatonin supplements of 2 mg dosage in AD patients and reported satisfactory efficiency, especially in the subgroup complaining primarily for insomnia (Wade et al., 2014).To our knowledge, no studies regarding the effect of therapeutic interventions in circadian rhythms on glymphatic phenotypes exist.
To sum up, despite most studies not being powered to detect interarm differences in cognitive changes, the evidence supports the aforementioned outlined trendline of OSAS-related onset or progression of AD.CPAP has been shown to optimize not only breathing and oxygenation, but also macrostructure of sleep.Therefore, sleep disturbances are ameliorated, which in turn restore glymphatic functionality.Interventions should focus on circadian dysregulation, as well.Either light-therapy or melatonin substitutes could slow down the progression of the disease, in the same way as the widespread drugs currently on the market.

Discussion
The link between sleep disturbances and Alzheimer's disease (AD) development has received a lot of research in recent years.Extended research has been conducted to investigate the potential function of the glymphatic system in explaining this link.We present a data-driven summary of the evidence from 70 studies across various thematic areas, including protein aggregation and toxicity, glymphatic implications of sleep disturbances, circadian dysregulation in AD, and potential glymphatically extracted interventions.While the majority of the evidence is indirect, our findings synthesis provides useful insights and directions for future research.
The perivascular space is the crucial compartment which conjoins the modalities of the glymphatic system dysfunction, senile-dependent inflammatory processes, and sleep dysregulation.Enlarged perivascular space is related to amyloid burden and adjacent glial activation, while hindering CSF-ISF mixing.Concomitantly, amyloid reverberations have their toll on sleep and, in particular NREM sleep; a physiologically temporal window during which CNS has the opportunity to optimally clear its metabolic waste of wakefulness.
Such effects were identified not only in AD populations, but also in cognitively healthy individuals, with disturbed sleep patterns, acutely or chronically and, perhaps, age-independently.As NREM disturbances emerge long before the clinical manifestation of AD, with reported symptoms suggestive of minor cognitive impairment, they warrant further exploration of the causative intercorrelation between Aβ-related sleep disturbances and cognitive decline.In the future, the investigation of possible Aβ-related sleep EEG biomarkers, could expand not only current knowledge on neurodegenerative diseases, but also the interventions to minimize -if not halttheir progression.
Immune dysregulation is a recognized feature in AD, albeit quite hard to put into biomarkers.Notably, CSF changes in specific proteins (such as Aβ or even synuclein) following sleep deprivation suggest that the alterations are more likely due to perturbed production/release of certain proteins rather than changes in global CSF-ISF clearance.It should be noted that for the diagnosis of neurodegenerative diseases, like AD, the absolute concentrations of CSF biomarkers per se are not utilized, but rather CSF biomarker ratios.This results in capturing multiple pathologies and their temporal interrelationship with greater predictive power, which is a better indicator of disease staging (Racine et al., 2016).
Afterall, there are several hypotheses for the etiology of AD pathogenesis (1) amyloid, (2) tau, and (3) infectious origin hypotheses.The amyloid hypothesis is the most prevalent but does not explain all aspects of AD (De Strooper and Karran, 2016).The tau hypothesis posits that tau hyperphosphorylation and accumulation into intraneuronal NFTs is the causative agent of AD, answering some but not all questions left unanswered by the amyloid hypothesis, as well as the full progression of AD pathology on its own (Maccioni et al., 2010).The final prominent hypothesis for AD pathogenesis is the infectious origin model.This hypothesis proposes that the initiating step in AD development is a pathogenic insult, either viral or bacterial, that triggers the formation of amyloid plaques and NFTs resulting in the onset of AD (Vavougios et al., 2020).The infectious origin hypothesis is relatively new and much of the data supporting it remains circumstantial.It should be noted that since AD encompasses a variety of manifestations and probably different phenotypes, these hypotheses may not be mutually excluded.
Additionally, each biomarker exhibits different kinetics.For instance, half-life of tau is greater than amyloid-β, signifying that its accumulation will become evident in chronic situations (Yamada et al., 2014).Aβ40 and Aβ42 may each be produced at different intracellular locations and by different mechanisms, since it has been reported that Aβ42 aggregates faster than Aβ40 (Hartmann et al., 1997;Wilson et al., 2002).AD biomarkers were principally assessed via CSF analysis, while several studies incorporated PET scans, in which biological relevance has been validated.EEG-biomarkers during wakefulness have, also, been investigated, but data is scarce.However, during sleep, the results were more robust, but their validation is pending, due to confounding factors, like age and undervalued SDB-comorbidity.Consequently, as the implications of sleep disturbances in glymphatic-related AD pathogenesis are well-documented, alternative proposed interventions, like CPAP and melatonin supplements, have demonstrated explicit efficacy in cognitive performance, but the results indicate no effect on amyloid plaques.This deduction may seem contradictory at first, but recent evidence has raised the concern that the relationship of neurodegeneration-amyloid plaque formation-cognitive impairment is neither linear nor mutually inclusive (Hagberg et al., 2020;Nelson et al., 2009).Hence, in terms of interventions, our research provides possible therapy routes for glymphatically derived proteins.Modifying sleep patterns, improving sleep quality, and improving glymphatic function may hold promise in slowing the progression of Alzheimer's disease.However, it is vital to highlight that the efficacy and long-term impacts of these therapies require comprehensive examination.
We further examined the alterations of perivascular space, associated with AD.The enlargement of perivascular space (EPVS) in specific brain regions has been implicated in several diseases.For instance, EPVS in centrum semiovale (CSO-EPVS) are associated with increased lobar microbleeds or cerebral amyloid angiopathy, while EPVS in basal ganglia (BG-EPVS) are related to deeper brain regions (BG, thalamus, internal capsule, and external capsule) or infratentorial regions (brainstem and cerebellum) with or without concomitant lobar MB, higher WMH volumes, hypertension, and ageing (Charidimou et al., 2013;Doubal et al., 2010).In cognitively unimpaired individuals, EPVS in both BG and CSO were associated with age, tau pathology and biomarkers of synaptic dysfunction in Aβ positive participants (Vilor-Tejedor et al., 2021;Yamasaki et al., 2022;Zeng et al., 2022).Zeng et al. focused on glia activation as an additional contributor to EPVS, in order to expand current knowledge.Their results replicated the correlation between CSO-EPVS and tau pathology and identified microglial activation as a link between the two (Zeng et al., 2022).BG-EPVS was consistently associated with age, gender and intracranial volume (Zeng et al., 2022).
Consequently, current data bring BG-EPVS into sharper focus as an early, but non-specific element, principally dependent on age and cardiovascular risk factors, while CSO-EPVS seems more specific to AD pathology, occurring in advanced stages, as it is associated with inflammatory processes that propagate tau pathology.These results were replicated in studies with CSF analysis of participants from the AD continuum, but not in the ones incorporating PET imaging, as presented in our review.Although both CSF measures and PET imaging have been found equally accurate in identifying early AD (Palmqvist et al., 2015), this inconsistency of the results suggests a pathophysiologic and glymphatic-driven difference between the two methods.
Our review of the published data suggests that glymphatic involvement in AD pathogenesis is not specific.In fact, several studies have indicated the same results of glymphatic impedance, outside of AD spectrum (Brown et al., 2018;Paradise et al., 2021).Taking into account the gaps in determining not only AD but all neurodegenerative diseases, it could be deduced that a redefinition of neurodegenerative diseases and a reapproach as spectrums and not distinguished pathological entities is required.An alternative concept introducing "CNS interstitial fluidopathy" has been suggested by Taoka and Naganawa (2021), which supports the glymphatic kinetics as crucial in their pathology.Beyond neurodegeneration, they proposed other diseases, like traumatic brain injury, subarachnoid hemorrhage and ischemic stroke, and idiopathic normal pressure hydrocephalus, as cases in which the glymphatic system stars.
While this work provides important new insights into the interplay between sleep disturbances, glymphatic function, and AD etiology, significant gaps and limitations remain.The vast majority of research presents observational and preclinical data, making it difficult to establish causal correlations.In addition, there is a lack of research designed to explore the direct implications of SDB in AD pathogenesis.Thus, firm conclusions cannot be reached, and glymphatic implications are mostly speculative.
We propose that the mixed findings in AD biomarkers may be attributed to differences in sources of measured concentrations (peripherally through plasma and CSF vs inside the CNS compartment).In SDB patients, both blood and respiratory pressures fluctuate significantly, especially around apneic episodes (Hudgel and Harasick, 1990;Xu et al., 2018).Elevation of BP disrupts CSF influx within the ISF, whereas increased respiratory effort leads to elevated intrathoracic and intracranial pressure, signifying increased venous pressure and impedance in the outflow compartment of the glymphatic system.By blocking the efflux of the glymphatic system, AD biomarkers may not be cleared and, therefore, cannot be detected peripherally.This hypothesis may explain the discrepancies discussed above and is worthy of future research.
Chronic hypercapnia and hypoxia are key features of the pathophysiology of SDB syndromes, (Brzecka, 2007), while NREM and REM sleep are disturbed around apneic episodes (Jones et al., 2014).Although we identified inconsistencies in terms of sleep parameters (like AHI, ODI and nocturnal saturation), we found that suboptimal nocturnal oxygenation emerged as a pivotal factor in glymphatic dysfunction.The pathophysiological hallmarks that link neurodegeneration in SDB and AD are hypoxia, glial activation and glymphatic impedance.Intermittent hypoxia inflicts neuronal loss in prone brain areas that are metabolically demanding, like the ones associated with cognitive functions.Impairment of functional connectivity in regions involved in memory consolidation during sleep has been attributed to nocturnal oxygen desaturation (Naismith et al., 2020).Hypoxic environment in such areas has been associated with increased amyloid (Shiota et al., 2013) and tau phosphorylation (Zhang et al., 2014), signifying glial activation and inflammatory processes, as well.In turn, the overactivation of neurotoxic profile under the weight of the neuroprotective one, propagates AD pathology and neurodegeneration in other CNS regions, contributing to the advancement of the disease (Ulland et al., 2021).The glymphatic system is thought to be capable of counteracting these detrimental effects, by clearing metabolic waste and reducing inflammation.However, as discussed above, the perivascular space is adversely affected, leading to entrapment of protein aggregates and suboptimal CSF-ISF exchange.
The main limitation of this work, as underlined above, is the hypothesis-driven synthesis and presentation of direct and indirect associations with the glymphatic system.Data is scarce, and the physiology and pathophysiology of glymphatic function in humans remains under-researched.However, our work provides a framework for new research directions.Larger-scaled, longitudinal studies with follow-up, including AD-specific populations, could offer some insight and disentangle the glymphatic proportion of AD pathogenesis.
New insights into the potential role of the glymphatic system on the natural history of AD may complement or even challenge the strictly pathological classification and expand the phenotypes of AD pathology.Through longitudinal and interventional studies, future research should aim to elucidate the underlying mechanisms linking sleep disorders, glymphatic dysfunction, and phenotypes of AD pathology.Furthermore, research into the influence of various sleep disorders on glymphatic function and the efficiency of tailored therapies is needed.
In conclusion, evidence from the 70 studies included in this manuscript supports the notion of an interrelationship between sleep disorders and AD pathogenesis, mediated by the glymphatic system.Sleep deprivation appears to disrupt glymphatic clearance mechanisms, which may contribute to protein aggregation and neurotoxicity.Understanding these intricate interconnections is critical for developing novel therapeutic strategies aimed at preserving glymphatic function and slowing the progression of Alzheimer's disease.More studies are needed to determine causality, validate therapies, and eventually apply these findings into clinical practice.

Fig. 2 .
Fig. 2. Bubble Plot of Thematic Areas distinguished for the Scoping Review.It should be noted that some citations are repeated in several thematic areas, due to relevance.

Table 1
Characteristics of studies that assessed Aβ toxicity after one night of sleep deprivation.Abbreviations: PSG= polysomnography, PET= positron emission tomography.
(continued on next page)