Alzheimer's disease as a systems-level timing disorder: Circadian disruption of glial immunometabolism, brain clearance, and therapeutic responsiveness

Abstract

Alzheimer's disease (AD) is traditionally conceptualized as a disorder of protein aggregation and neurodegeneration, yet growing evidence indicates that fundamental temporal organization of brain physiology is also disrupted. In the healthy brain, circadian clocks coordinate sleep–wake behavior, glial immunometabolism, astrocytic aquaporin-4 polarity, and glymphatic–lymphatic clearance, aligning immune readiness and proteostasis with daily activity–rest cycles. In AD, this temporal coordination progressively deteriorates, manifesting as sleep fragmentation, instability of rest–activity rhythms, vulnerability of central clock structures, and loss of circadian gating of glial and clearance pathways. These disruptions create phase-inappropriate immune and metabolic states, impair protein clearance, and alter the fate of extracellular vesicles, which may shift from mediators of waste export to facilitators of proteopathic spread. Importantly, circadian failure also constrains therapeutic delivery and biomarker interpretation by modulating blood–brain barrier transport, brain fluid dynamics, and brain-to-blood signal export. We propose that AD can be reframed as a systems-level timing disorder, in which loss of temporal coherence integrates molecular pathology, glial dysfunction, clearance failure, therapeutic inefficacy, and biomarker variability. This framework highlights chrono-pharmacology, chrono-neurotherapeutics, and circadian-informed biomarkers as essential components of precision strategies for AD prevention and treatment.

Highlights

• •Alzheimer's disease involves loss of temporal coordination across circadian timing, sleep-wake organization, and brain physiology.
• •Circadian clocks synchronize glial immunometabolism, AQP4 polarity, and sleep-dependent glymphatic-lymphatic clearance.
• •Circadian disruption drives maladaptive glial states and impairs proteostasis, increasing neuronal vulnerability.
• •The glymphatic-exosome axis governs whether extracellular vesicles mediate clearance or protephathic propagation.
• •Therapeutic efficacy and biomarker readouts in Alzheimer's disease are inherently time-dependent.

1. Introduction

Alzheimer's disease (AD) is classically defined by the accumulation of amyloid-β plaques, tau neurofibrillary tangles, synaptic dysfunction, and progressive neurodegeneration. Decades of research have elucidated molecular mechanisms underlying protein aggregation, neuronal vulnerability, and glial activation, yet disease-modifying therapies have produced modest or inconsistent clinical benefit. This persistent translational gap has prompted a reassessment of prevailing conceptual models, raising the possibility that critical dimensions of disease biology remain under-integrated rather than undiscovered.

One such dimension is time. The healthy brain is not a static biochemical environment but a dynamically organized system in which cellular metabolism, immune surveillance, barrier transport, and waste clearance are synchronized to circadian and sleep–wake rhythms (Hastings et al., 2018; Musiek and Holtzman, 2016). Circadian clocks coordinate neuronal activity, glial function, vascular physiology, and cerebrospinal fluid (CSF) dynamics, aligning energetic demand and proteostasis with predictable cycles of activity and rest. Sleep, in this context, is not merely a behavioral state but a distinct physiological phase characterized by altered brain fluid movement, immune tone, and synaptic homeostasis (Hablitz et al., 2020; Nedergaard and Goldman, 2020). These temporal programs allow the central nervous system to balance information processing during wakefulness with repair, clearance, and recalibration during rest.

Disruption of circadian rhythms and sleep architecture is increasingly recognized as a prominent and early feature of AD (Winer et al., 2019; Musiek et al., 2018). Patients exhibit sleep fragmentation, altered rest–activity patterns, circadian phase shifts, and degeneration of clock-regulating structures, often preceding overt cognitive decline. Importantly, these abnormalities are not confined to advanced disease stages nor fully explained by neurodegeneration alone. Experimental and human data indicate that circadian disruption can interact bidirectionally with amyloid and tau pathology, glial activation, and metabolic stress, suggesting that loss of temporal coherence may contribute actively to disease progression rather than merely reflecting it (Winer et al., 2019; Musiek et al., 2018).

Parallel advances in systems neuroscience have revealed that brain clearance pathways are temporally regulated. Astrocytic aquaporin-4 polarity, glymphatic CSF–interstitial fluid exchange, and downstream lymphatic drainage exhibit state-dependent and circadian modulation (Hablitz et al., 2020; Rasmussen et al., 2018). These pathways are essential for the removal of metabolic waste and protein aggregates, and their efficiency varies across the sleep–wake cycle. Failure of this clearance infrastructure has been implicated in AD pathophysiology, yet it is increasingly clear that clearance dysfunction cannot be fully understood without considering its temporal regulation (Nedergaard and Goldman, 2020; Rasmussen et al., 2018).

At the cellular level, glial immunometabolism is also subject to circadian control. Microglia and astrocytes exhibit time-of-day variation in inflammatory responsiveness, metabolic substrate utilization, and phagocytic activity (Nakanishi et al., 2021; Ding et al., 2024). In a temporally aligned system, such gating constrains immune activation to appropriate phases and promotes resolution and repair. Loss of circadian gating, by contrast, can lock glia into maladaptive metabolic and inflammatory states, amplifying neuronal stress and impairing proteostasis. These observations provide a mechanistic bridge between circadian disruption, chronic neuroinflammation, and progressive neurodegeneration.

Extracellular vesicles (EVs) further complicate this landscape. EVs, including exosomes, participate in intercellular communication and can transport proteins, nucleic acids, and lipids across neural and vascular compartments. In neurodegenerative disease, EVs have been implicated both in the dissemination of pathogenic proteins and in potential clearance or biomarker pathways (Asai et al., 2015; Kalluri and LeBleu, 2020). Emerging evidence suggests that the fate of EVs whether they facilitate export, accumulation, or propagation of pathology is constrained by the integrity and timing of brain clearance systems. This duality underscores the need for a systems-level view in which vesicle biology is embedded within temporally regulated transport networks.

Temporal regulation also extends to therapeutic delivery and biomarker interpretation. Blood–brain barrier transport, brain fluid dynamics, immune tone, and brain-to-blood signal export all vary across the circadian cycle (Zhang et al., 2021; Eckhardt et al., 2025). As a result, identical interventions or sampling strategies may yield different outcomes depending on biological timing. This time dependence offers a plausible explanation for heterogeneity in clinical trials and biomarker studies that is not readily accounted for by molecular pathology alone.

Collectively, these observations motivate a reframing of AD not solely as a disorder of protein aggregation or cell loss, but as a systems-level timing disorder in which circadian disruption destabilizes the coordination between neuronal activity, glial function, clearance capacity, therapeutic responsiveness, and biomarker visibility. In this Review, we synthesize evidence across circadian biology, sleep physiology, glial immunometabolism, brain clearance pathways, extracellular vesicle dynamics, and translational studies to propose an integrated temporal framework for AD. By treating time as an organizing variable rather than a confounder, this perspective aims to clarify mechanistic links between disparate pathological features and to identify chrono-informed strategies for intervention, biomarker development, and clinical trial design.

Terminology note: Throughout this Review, we use “loss of temporal coherence” as the overarching construct to describe progressive breakdown of coordinated circadian timing and sleep–wake organization; we use “circadian disruption” for clock-dependent mechanisms and “sleep fragmentation” for sleep-state–dependent mechanisms.

Box 1: evidence-anchored vs hypothesis-generating components of the timing model.

Within this framework, several components are supported by direct empirical evidence, including circadian regulation of glial immunometabolism, sleep-dependent glymphatic clearance, astrocytic AQP4 polarity, and circadian modulation of blood–brain barrier transport. Other elements such as state-dependent switching of extracellular vesicle fate from clearance toward proteopathic propagation under circadian failure remain integrative and hypothesis-generating. These hypotheses are grounded in established vesicle biology and clearance constraints but require direct experimental testing. We explicitly present these components as testable predictions rather than settled mechanisms.

2. Circadian architecture of the healthy brain

In the healthy brain, “time” is not a background variable; it is an organizing principle. Daily (∼24 h) rhythms emerge from cell-autonomous molecular clocks that are synchronized across a hierarchical network, with the suprachiasmatic nucleus (SCN) acting as the central coordinator of organismal sleep–wake state (Brancaccio et al., 2019). At the core of this system is a transcription–translation negative feedback loop (TTFL): CLOCK:BMAL1 heterodimers bind E-box elements to drive daytime transcription of Per and Cry genes; PER–CRY complexes accumulate, translocate to the nucleus, and repress CLOCK:BMAL1 activity during the circadian night, until PER–CRY degradation permits initiation of the next cycle. This canonical loop is not confined to neurons; TTFLs operate broadly across brain cell types, creating a distributed “timekeeping” substrate that can gate cellular programs relevant to homeostasis (Brancaccio et al., 2019).

A defining feature of healthy circadian organization is that astrocytes are not passive followers of neuronal timekeeping. Within the SCN, astrocytes possess functionally meaningful TTFLs and can encode circadian information in a way that is sufficient to instruct neuronal partners and sustain behavioral rhythmicity. In a striking demonstration of sufficiency, restoring Cry1 selectively in SCN astrocytes (in an otherwise clock-deficient background) re-established molecular oscillations in the SCN and reinstated circadian locomotor behavior, mediated through astrocyte-to-neuron signaling that includes glutamatergic cues (Brancaccio et al., 2019). This establishes a critical baseline concept for the present manuscript: normal sleep–wake biology arises from multicellular clock integration, where astrocytic timing is an active component of system-level synchronization.

Downstream of this timing architecture, circadian state is transduced into compartment-level brain physiology through the gliovascular interface, most notably via aquaporin-4 (AQP4) polarization at astrocytic endfeet. In the healthy brain, perivascular AQP4 localization and the molecular machinery supporting endfoot specialization are dynamically regulated across the day–night cycle. In parallel, glymphatic influx the convective movement of cerebrospinal fluid (CSF) into brain parenchyma along perivascular pathways is not constant: it is enhanced during the rest phase relative to the active phase, aligning waste transport capacity with the sleep-dominant window of tissue repair and proteostatic maintenance (Hablitz et al., 2020). Together, these observations define a baseline “clearance-ready” state in which circadian timing, astrocytic endfoot biology, and perivascular fluid exchange are coordinated rather than independent variables (Hablitz et al., 2020).

Importantly, sleep-dependent clearance is not purely a rodent construct; human sleep is also characterized by state-linked, system-level CSF dynamics. During non–rapid eye movement sleep, slow neural activity and associated hemodynamic changes are coupled to oscillatory CSF movement, supporting the concept that sleep organizes large-scale fluid shifts that could facilitate metabolite redistribution and clearance (Fultz et al., 2019). These human data provide a crucial translational anchor for the baseline model: sleep is a brain-wide physiological state that coordinates electrophysiology, blood dynamics, and CSF flux, thereby plausibly shaping the efficiency of perivascular transport processes.

Taken together, the “healthy reference state” can be summarized as a three-layer architecture: (i) a canonical CLOCK:BMAL1-centered TTFL distributed across brain cell types, (ii) SCN-based multicellular integration in which astrocytic clocks can be behaviorally instructive rather than merely permissive, and (iii) circadian/sleep-linked coupling of astrocytic endfoot specialization (AQP4 polarity) with enhanced rest-phase glymphatic influx and sleep-associated CSF dynamics (Hablitz et al., 2020; Brancaccio et al., 2019; Fultz et al., 2019). This baseline is the necessary foundation for later sections where we argue that AD represents a systems-level synchronization failure where timing signals, glial “readiness,” and clearance pathways become misaligned, converting protective dynamics into vulnerability.

These coordinated circadian and sleep-linked programs establish a physiological reference state in which glial timing, astrocytic endfoot specialization, and glymphatic–lymphatic clearance are temporally aligned (Fig. 1).

Fig. 1.

| Circadian coordination of glial state and brain clearance in the healthy brain. Circadian timing organizes multicellular brain physiology to align immune–metabolic state, fluid dynamics, and clearance capacity under healthy conditions. In the healthy brain, central circadian signals coordinated by the suprachiasmatic nucleus (SCN) synchronize astrocytic and microglial states across the sleep–wake cycle. At the molecular level, CLOCK–BMAL1–driven transcriptional programs generate rhythmic PER/CRY feedback that gates downstream cellular functions. During the rest/sleep phase, polarized aquaporin-4 (AQP4) expression at astrocytic endfeet facilitates cerebrospinal fluid–interstitial fluid (CSF–ISF) exchange within perivascular spaces, promoting glymphatic influx and downstream lymphatic efflux. This temporally gated clearance window supports efficient solute transport, including regulated extracellular vesicle (EV) export toward lymphatic and systemic circulation. Together, circadian clock integration, glial state regulation, and sleep-dependent clearance define a coordinated reference architecture that maintains proteostasis and brain-to-periphery communication under physiological conditions.

3. Circadian breakdown in Alzheimer's disease

A consistent clinical signal across the AD continuum is that the circadian system does not simply weaken with age but becomes progressively disorganized. In contrast to the advanced sleep–wake phase disorders commonly observed in healthy older adults, a characteristic phenotype in AD is irregular sleep–wake rhythm disorder, defined by a lack of a clear 24-h sleep–wake pattern with prolonged wakefulness at night and irregular sleep bouts during the day (Leng et al., 2019).

Sleep fragmentation and rest–activity rhythm instability across the AD continuum. Across studies that include preclinical AD (amyloid pathology without cognitive symptoms), mild cognitive impairment, and clinically manifest AD, behavioral markers of circadian disruption repeatedly include rest–activity rhythm fragmentation and altered sleep timing (Leng et al., 2019). Notably, across diverse cohorts, a recurring pattern is high fragmentation with only slight reduction (or no change) in rhythm amplitude, suggesting that the more consistent phenotype may be loss of consolidation and stability rather than a uniform collapse of circadian strength.

Objective sleep phenotyping using actigraphy offers a practical way to quantify this instability. In cohorts of cognitively unimpaired older adults, actigraphy studies summarized and discussed in recent studies associate higher sleep fragmentation, longer sleep latency, and greater wake after sleep onset (WASO) with increased AD risk and accelerated cognitive decline, and link specific actigraphy metrics (e.g., shorter total sleep time, lower sleep efficiency) with higher tau-related biomarker signals in cognitively unimpaired individuals (Stankeviciute et al., 2025). Importantly for interpretation, these reports emphasize that WASO and fragmentation indices capture complementary aspects of disrupted sleep continuity (duration versus frequency of awakenings), strengthening confidence that “sleep fragmentation” is not a single noisy variable but a measurable phenotype with separable components (Stankeviciute et al., 2025).

Loss of temporal coherence, not a single stereotyped phase shift. An evidence-based difficulty in this literature is that phase shifts are mixed across studies, and variability across individuals can be substantial (Leng et al., 2019). This is not a weakness of the field so much as an important biological message: AD-related circadian disruption appears better described as a loss of temporal coherence (i.e., fragmented rest–activity organization with scattered behavior across 24 h) rather than one uniform “phase advance” or “phase delay” that would apply across patients and stages (Leng et al., 2019).

In fact, a recent Lancet Neurology review explicitly notes that behavioral circadian markers (sleep timing, daytime sleepiness, rest–activity rhythms) are far more commonly measured than biological markers (melatonin, cortisol, body temperature), underscoring why behavioral fragmentation must be interpreted as a systems-level phenotype that can reflect both sleep and circadian mechanisms (Leng et al., 2019).

Biological substrates: vulnerability of the master clock and clock-linked homeostatic programs. A biologically grounded substrate for rhythm collapse is the suprachiasmatic nucleus (SCN), the master circadian pacemaker. In postmortem human hypothalamus across progressive AD neuropathologic change, spatial in-situ proteomic profiling showed that SCN neurons at Braak VI had markedly higher phosphorylated tau than Braak 0, with neurofibrillary tangles reported in the SCN (Son et al., 2024). Crucially, this study also reports increases in glial proteins already at Braak I in the SCN (with different patterns in nearby control nuclei), supporting the interpretation that SCN vulnerability can emerge early and may contribute to circadian disturbances observed even before onset of cognitive disorder.

Beyond the SCN itself, circadian disruption in neurodegeneration increasingly intersects with cell-type-specific clock-linked homeostatic programs. A Neuron study demonstrated that BMAL1 (ARNTL), a core circadian transcription factor was shown to regulate astrocyte activation programs and to influence tau and α-synuclein pathology through an axis involving the autophagy chaperone BAG3, with BAG3 reported as expressed in disease-associated astrocytes in human AD samples (Sheehan et al., 2023). This is important for the logic of this review: it supports an evidence-based view that “circadian disruption” in AD is not only behavioral timing disturbance, but also includes disruptions in molecular programs that gate glial state and proteostatic resilience.

Summary. Taken together, the selected human and mechanistic evidence supports three cautious but robust conclusions. First, AD is associated with clinically meaningful disruption of circadian organization, including irregular sleep–wake patterning and high rest–activity fragmentation. Second, the dominant clinical signature is better framed as loss of consolidation and temporal coherence than as a single stereotyped phase shift. Third, circadian breakdown has plausible biological substrates spanning early SCN glial dysregulation and later tau pathology in the master clock, and clock-linked astrocyte proteostasis programs that intersect proteopathic vulnerability.

Rather than representing a simple amplification of amyloid or tau burden, AD is increasingly characterized by loss of temporal coherence across glial states and clearance pathways (Fig. 2).

Fig. 2.

| Alzheimer's disease as a systems-level disorder of circadian timing and clearance. Alzheimer's disease reflects a breakdown of circadian coordination that disrupts glial state switching and renders brain clearance transport-limited. In the healthy state (left), intact circadian coordination maintains synchronized clock phase across cellular compartments, enabling rhythmic astrocytic and microglial switching, polarized aquaporin-4 (AQP4) localization at astrocytic endfeet, consolidated sleep, and efficient glymphatic–lymphatic flow. In Alzheimer's disease (right), circadian desynchronization produces phase dispersion, leading to phase-inappropriate immune activation, microglial metabolic misalignment, and loss of AQP4 polarity. These alterations impair sleep-dependent cerebrospinal fluid–interstitial fluid exchange and downstream lymphatic efflux, resulting in extracellular vesicle (EV) retention and misrouting that favors pathological propagation over clearance. Collectively, this loss of temporal coherence generates a progressive clearance deficit in which biomarker visibility and therapeutic responsiveness become constrained by transport capacity rather than target engagement alone.

4. Circadian Clock–Dependent vs sleep-state–dependent mechanisms

Evidence supports a distinction between mechanisms primarily driven by intrinsic circadian clock function and those dependent on sleep state. Circadian clock–dependent processes include BMAL1-linked glial transcriptional programs, rhythmic blood–brain barrier transport and efflux capacity, and selective vulnerability of the suprachiasmatic nucleus and downstream clock networks. In contrast, sleep-state–dependent mechanisms include slow-wave–associated cerebrospinal fluid dynamics, glymphatic–lymphatic clearance efficiency, and the cumulative effects of sleep fragmentation on metabolic and inflammatory tone.

In humans, these dimensions are partially inseparable, and current actigraphy-based measures reflect integrated system behavior rather than pure circadian phase (Musiek and Holtzman, 2016; Leng et al., 2019; Xie et al., 2013).

Circadian gating of glial immunometabolism. A core (and often underappreciated) premise is that glial immune–metabolic state is not a constant “set point,” but a time-structured program. In isolated mouse microglia, transcripts linked to inflammation, nutrient utilization, and antioxidation show daily rhythmicity, aligning metabolic demand with predictable changes in activity/rest across the light–dark cycle (Wang et al., 2020). In that framework, BMAL1 functions as a gatekeeper coupling clock phase to microglial immune tone and metabolic gene programs: when Bmal1 is disrupted, microglia shift toward lower pro-inflammatory gene expression with higher anti-inflammatory/antioxidant programs, alongside altered metabolic-gene expression and measurable functional consequences (including increased phagocytic capacity in BV-2 knockdown experiments) (Wang et al., 2020). Collectively, these findings indicate that the brain's innate immune system has an intrinsic timing logic; when timing breaks, immune-metabolic responses can become mismatched to brain state.

This timing logic becomes especially consequential in AD, where microglia repeatedly face a persistent, proteotoxic stimulus. In primary microglia, acute Aβ exposure induces a canonical metabolic reprogramming from oxidative phosphorylation toward glycolysis, accompanied by inflammatory activation, and this response depends on the AKT–mTOR–HIF-1α signaling axis, with glycolytic flux functionally linked to inflammatory output (Baik et al., 2019). Critically, the same study shows that chronic amyloid-β exposure does not simply “maintain” the acute glycolytic program. Instead, microglia transition into an innate immune–tolerant/dysfunctional state marked by reduced inflammatory cytokine output and reduced phagocytosis, with broad metabolic defects involving both glycolysis and OXPHOS (i.e., not a clean “glycolysis stays on” story) (Baik et al., 2019). That pattern strongly supports a “metabolic jetlag” concept in AD: repeated or chronic danger signaling may push microglia through maladaptive state transitions until the system becomes phase-inappropriate and functionally exhausted not merely inflamed.

Astrocytes add a second, complementary layer of circadian control: clock disruption can be sufficient to induce reactive programs in a cell-autonomous manner. Astrocyte-specific or astrocyte-enriched Bmal1 loss induces astrocyte activation and inflammatory gene expression, mechanistically linked (in this model system) to altered glutathione-S-transferase (GST)/glutathionylation-related signaling rather than simple depletion of total glutathione. Functionally, loss of astrocytic Bmal1 worsens neuronal survival in co-culture contexts, consistent with the idea that clock integrity is part of the astrocyte's “support contract” with neurons (Lananna et al., 2018). Taken together, these findings support a model in which microglial and astrocytic clocks regulate distinct but convergent pathways immune–metabolic switching in microglia and redox/glutathione-linked activation states in astrocytes that collectively shape inflammatory setpoints and tissue resilience across the day.

Summary: when circadian architecture is intact, glia can cycle between brain-state-appropriate modes (vigilance/defense vs restoration/repair). When clocks degrade (as in AD and aging contexts), glia may become locked into wrong-phase programs either persistently primed, or paradoxically tolerant/exhausted and both outcomes can impair proteostasis and neuronal support. This is the mechanistic bridge from “timing disorder” to “immunometabolic failure”: the novelty is not that metabolism matters, but that time-of-day is a hidden variable that can flip the sign and efficiency of glial responses, making the same stimulus produce different outcomes depending on circadian phase and chronicity.

Sleep–glymphatic–lymphatic coupling as a dynamic proteostasis program. A healthy adult brain does not clear metabolites and proteopathic proteins at a constant rate; instead, it time-shares clearance capacity across behavioral state (sleep vs wake) and circadian phase, thereby matching waste removal to daily cycles of neuronal activity and protein release. In this framework, sleep is not simply “rest” as it is an active systems state in which CSF–ISF exchange is upshifted, and the downstream lymphatic outflow pathways become the obligatory exit route for brain-derived solutes. This coupling matters because the CNS lacks classical parenchymal lymphatic vessels, so large-scale protein export depends on (i) convective paravascular influx and interstitial exchange (glymphatic transport) and (ii) meningeal/cervical lymphatic drainage acting in series (Rasmussen et al., 2018).

At the mechanistic core of this brain-wide clearance circuit sits the astrocyte endfoot: polarized AQP4 expression at perivascular endfeet provides a permissive interface for CSF–ISF exchange and thus for bulk solute movement along glymphatic pathways (Hablitz et al., 2020; Mestre et al., 2018). Across multi-lab re-analyses, genetic loss of AQP4 or disruption of its perivascular localization reduces CSF tracer transport, and variables such as anesthesia and age can substantially modulate observed transport highlighting that “clearance capacity” is a state-dependent physiologic phenotype, not a fixed anatomical constant.

A key healthy-brain principle, therefore, is circadian gating of clearance efficiency: glymphatic influx is higher during the rest phase, and AQP4 polarization itself shows day–night patterning, consistent with a clock-aligned architecture that schedules fluid movement and solute export (Hablitz et al., 2020). The conceptual implication for this review is important: when we talk about “brain clearance,” we are really talking about a rhythmic fluid-transport program that requires synchrony between (a) astrocytic water handling, (b) vascular/perivascular dynamics, and (c) sleep timing and depth.

Downstream, the brain's “waste stream” must eventually leave the cranial compartment. Meningeal lymphatic vessels provide a direct anatomical route linking CNS solute clearance to the peripheral immune and metabolic recycling systems. In particular, lymphatic vessels at the skull base contribute to CSF drainage and solute efflux to deep cervical lymph nodes, reinforcing that glymphatic transport is not an end in itself, it is an upstream feeder into lymphatic outflow (Ahn et al., 2019). This creates a serial bottleneck logic: even if parenchymal CSF–ISF exchange is intact, impaired lymphatic egress can still throttle overall clearance efficiency (and conversely, robust lymphatic outflow cannot compensate for diminished paravascular exchange).

Critically, multiple human-relevant observations align with this sleep-dependent clearance model. Reviews synthesizing rodent and emerging human evidence emphasize that glymphatic activity is primarily increased during sleep, with experimental and imaging work supporting that sleep loss and disrupted sleep architecture can alter brain protein homeostasis (Nedergaard and Goldman, 2020; Rasmussen et al., 2018). In humans and mouse models, the sleep–wake cycle modulates extracellular tau dynamics, supporting the broader claim that sleep timing and quality are upstream regulators of proteostatic load (Holth et al., 2019). Put bluntly: in a healthy brain, sleep-aligned clearance is how the CNS keeps protein synthesis and protein export in balance and any chronic misalignment (circadian or sleep fragmentation) is expected to create a slow, cumulative “clearance debt” even before overt neurodegeneration emerges (Nedergaard and Goldman, 2020).

Taken together, these findings support a model in which glymphatic clearance is not a constitutive property but a temporally gated systems function, whose failure in AD is better understood as loss of synchronization than loss of capacity.

The Glymphatic–Exosome axis: transport-limited clearance and vesicle-mediated propagation. In a healthy brain, extracellular vesicles (EVs) especially exosomes should be viewed not only as intercellular messengers, but also as mobile “containers” whose trafficking is constrained by brain-scale fluid dynamics. Exosomes are endosome-derived EVs (∼100 nm on average) with complex cargo (nucleic acids, proteins, lipids, metabolites) that can reflect cell state and enable multi-component biomarker readouts in biofluids. Yet, critically, their in vivo functions can be difficult to infer from bolus dosing, because supraphysiologic exosome exposure in animals can induce strong phenotypes, underscoring the need to interpret trafficking/biology within physiological transport constraints rather than “payload-only” logic. This immediately raises a systems question for AD: even if a vesicle is informative or therapeutic, what routes and states actually allow it to move, exit, and be cleared? (Kalluri and LeBleu, 2020).

A key transport axis is the glymphatic–meningeal lymphatic continuum, which can move brain-derived solutes toward cervical lymphatics and ultimately the blood. In vivo, suppression of glymphatic function using clinically relevant manipulations (including sleep deprivation) reduces convective glymphatic efflux and sharply reduces tracer clearance to cervical lymphatics; strikingly, these same manipulations can suppress or eliminate the expected rise of blood biomarkers released from injured brain, demonstrating that “brain-to-blood” signals can be transport-limited, not merely “production-limited.” In other words, peripheral EV/protein biomarkers (and, by extension, vesicle-associated species) should be interpreted through the lens of state-dependent clearance capacity, where sleep-linked glymphatic flow can act as a gate on what appears in blood (Plog et al., 2015).

Within AD-relevant proteinopathy, these transport constraints intersect with a second axis: microglia-driven vesicular propagation of tau. In a rapid in vivo tau propagation paradigm, microglial depletion dramatically suppressed tau propagation and reduced downstream physiological dysfunction; mechanistically, the work supports a model in which microglia phagocytose tau-bearing neuronal material/synapses and then secrete tau in exosomes, which can efficiently transmit tau to neurons. Importantly, systemic inhibition of exosome synthesis (nSMase2 inhibition with GW4869) suppressed tau propagation in a circuit-dependent manner (e.g., suppressing DG propagation while sparing some EC accumulation), implying that early/local tau accumulation and network spread may have different dependencies on exosome biogenesis and/or drug distribution. This supports a “two-phase” conceptualization: initiation may be less exosome-dependent in some regions, whereas propagation along connected circuits can be strongly exosome-mediated (Asai et al., 2015).

Putting these pieces together yields the section's central systems-level insight: vesicles can be both “vehicles of spread” and “vehicles of clearance,” depending on whether the brain's clearance infrastructure is functioning and whether vesicle biogenesis is acting as a protective export pathway or a propagation route. The exosome literature explicitly frames this duality in neurodegeneration: tau and Aβ are found in exosomes across contexts, and exosome biogenesis can plausibly be pathogenic (propagating toxic assemblies) or protective (exporting/limiting toxic oligomers, or enabling removal) depending on disease, cell type, and in vivo context. Therefore, any AD narrative that centers EVs must incorporate (i) circuit biology of propagation (microglia–exosome tau transfer), and (ii) state-dependent bulk transport/clearance (glymphatic-to-lymphatic-to-blood) as co-equal determinants of what vesicles do and what vesicles signal (Asai et al., 2015; Kalluri and LeBleu, 2020).

5. Chrono-pharmacology and chrono-delivery in Alzheimer's disease

A central implication of viewing AD as a disorder of temporal regulation is that therapeutic efficacy cannot be evaluated independently of when an intervention engages the brain. Blood–brain barrier (BBB) transport, CSF dynamics, and neuroimmune tone are not temporally uniform but fluctuate across the 24-h cycle, such that molecular target engagement at one time of day may correspond to biological disengagement at another. In this framework, “dose” and “delivery” become inseparable from circadian phase, transforming time from a nuisance variable into a determinant of pharmacological success (Musiek and Holtzman, 2016).

One of the clearest demonstrations of temporal gating occurs at the BBB itself. Experimental studies show that endothelial efflux capacity exhibits circadian rhythmicity, with transporter activity peaking during the active phase and reaching a nadir during the rest phase; critically, this rhythmicity collapses in circadian clock–deficient conditions (Zhang et al., 2021).

Mechanistically, the endothelial molecular clock regulates intracellular magnesium oscillations via TRPM7, which in turn modulate the activity of efflux transporters such as P-glycoprotein (Zou et al., 2019).

These findings establish that brain exposure to systemically delivered agents is inherently time-dependent. From a review perspective, the implication is straightforward but often overlooked: for drugs that are transporter substrates or rely on transcytotic mechanisms, non-timed dosing effectively averages across permissive and non-permissive BBB states, diluting apparent efficacy.

Temporal structure also governs clearance physiology. Human multimodal imaging demonstrates that non-rapid eye movement sleep is accompanied by large-amplitude, low-frequency oscillations that couple neural activity, hemodynamics, and pulsatile CSF inflow at approximately 0.05 Hz (Fultz et al., 2019). This defines sleep as a distinct fluid-dynamic state rather than a passive reduction in activity, with direct consequences for solute redistribution and clearance.

When sleep is fragmented or slow-wave activity is diminished both common features across the AD continuum the physiological window that supports efficient clearance may be truncated or lost. From a trial-design standpoint, this raises the possibility that clearance-dependent therapies are systematically disadvantaged in cohorts with unmeasured or uncontrolled sleep disruption.

Immune–metabolic interventions further illustrate why timing cannot be treated as secondary. In ageing models, restoring myeloid bioenergetic balance by targeting prostaglandin E₂–EP2 signaling reverses cognitive decline, normalizes mitochondrial respiration, and improves hippocampal-dependent memory (Minhas et al., 2021). Although these interventions are often described as “peripheral,” their cognitive impact depends on endothelial and immune-to-brain coupling pathways that themselves exhibit circadian modulation.

This reinforces a broader principle: even therapies not explicitly designed to target circadian biology may succeed or fail depending on when they intersect with time-varying immune and vascular states.

Taken together, these observations provide a coherent explanation for heterogeneity in AD therapeutic trials. Circadian and sleep disruption can emerge early in disease and may actively contribute to pathogenesis rather than merely reflecting neurodegeneration.

Patients therefore differ not only in molecular pathology but also in the integrity of temporal control over BBB transport, CSF dynamics, and immune metabolism. Such differences define plausible “chrono-phenotypes,” in which the same intervention may be effective only within circadian-permissive windows that are preserved in some individuals but eroded in others.

The overarching conclusion of this section is not that timing replaces target selection, but that time is a hidden dimension of pharmacology in the brain. Many Alzheimer's trials may have failed not solely because of incorrect targets or insufficient doses, but because interventions were delivered without regard to the biological time at which the brain is most receptive to delivery, engagement, and clearance.

6. Chrono-neurotherapeutics: synchronizing intervention with brain state

A precision chrono-neurotherapeutics framework for AD begins with the recognition that the brain is not a steady-state compartment. CSF production, interstitial fluid (ISF) exchange, BBB transport, and perivascular mixing are all regulated across the 24-h cycle by circadian timing mechanisms that integrate molecular clocks, sleep–wake state, and vascular physiology. As a result, solute distribution and clearance are dynamically modulated rather than continuous, rendering therapeutic delivery intrinsically time-dependent rather than purely dose-dependent (Vizcarra et al., 2024).

Importantly, AD is not simply a condition in which chronotherapy could be layered onto an otherwise intact system. Experimental and clinical evidence indicates that sleep–wake architecture and sleep-dependent homeostatic circuits are disrupted early in disease progression. In the AppˆNL-G-F knock-in mouse model, hippocampal hyperexcitability emerges alongside alterations in sleep structure, and sleep-active melanin-concentrating hormone (MCH) neurons in the lateral hypothalamus projecting to hippocampal CA1 are recruited as a compensatory mechanism that progressively deteriorates with age (Calafate et al., 2023). Mechanistically, MCH signaling dampens synaptic transmission and modulates firing-rate homeostasis in CA1 pyramidal neurons, and exogenous MCH is sufficient to reverse excessive excitatory drive in this model, directly linking sleep-circuit dysfunction to network instability.

This observation carries a critical implication for chrono-neurotherapeutics: the physiological window most favorable for restoration of sleep-associated downscaling of neuronal activity and fluid-mediated clearance may itself be fragmented or weakened in AD. Consequently, optimal intervention timing cannot be inferred from clock time alone, but must be anchored to actual sleep physiology, as quantified by objective sleep metrics. In this context, chronotherapy is not simply “night-time dosing,” but state-matched intervention, aligned with residual homeostatic capacity rather than assumed circadian phase (Calafate et al., 2023).

A second translational dimension situates chrono-neurotherapeutics within a broader prevention-to-intervention continuum. The 2024 Lancet Standing Commission emphasizes that a substantial proportion of dementia risk is attributable to modifiable factors and that multi-component prevention strategies including sleep health and vascular-metabolic risk control remain relevant across the lifespan (Livingston et al., 2024).

From a systems perspective, these interventions act upstream of therapeutic delivery by preserving the integrity of circadian and sleep-dependent brain physiology upon which clearance pathways, barrier function, and neuronal resilience depend. Thus, chrono-neurotherapeutics should be understood not as a narrow scheduling optimization, but as an integrated strategy in which risk-factor modification sustains the biological rhythms that determine therapeutic receptivity (Livingston et al., 2024).

Taken together, these data justify a pragmatic but conceptually novel conclusion: effective Alzheimer's interventions must be synchronized with brain state, not simply administered according to external time. As disease progression progressively distorts sleep circuitry and circadian organization, patient-specific phenotyping of sleep integrity and brain fluid dynamics may become a prerequisite for precision timing. Chrono-neurotherapeutics therefore reframes treatment not as a static molecular correction, but as a dynamic alignment between intervention and the brain's remaining capacity for homeostatic regulation.

7. Chrono-biomarkers at the periphery: aligning circadian signatures, PET pathology, and neuron-derived EV readouts

A practical way to make the circadian–glymphatic–exosome concept clinically actionable is to treat time as a biomarker dimension i.e., to ask whether readily scalable peripheral signals can report (i) where a person sits on the circadian fragmentation/timing spectrum and (ii) whether that circadian state is already coupled to AD molecular pathology. Recent human actigraphy-to-imaging work directly supports this bridge: in older adults with objective early cognitive impairment, earlier circadian timing (acrotime) was associated with higher Aβ and tau PET, and greater within-day rhythm fragmentation (intradaily variability, IV) related to higher tau in Braak III/IV regions; critically, tau in Braak I–IV mediated the association between acrotime and verbal memory one of the cleanest human demonstrations that circadian timing can map onto cognition through AD pathology rather than only through “sleep symptoms” (Eckhardt et al., 2025).

This is where neuron-derived extracellular vesicles (NDEVs/nEVs) from blood become a high-novelty enabling layer: they are increasingly positioned as a “molecular window into the brain” because neuronal EVs can be detected in blood and captured via immunoaffinity approaches (e.g., L1CAM-based strategies used in multiple AD studies), allowing pathway-level biomarkers (tau species; insulin-signaling nodes; synaptic/neuronal stress proteins) to be tracked longitudinally (Delgado-Peraza et al., 2023). In a late-middle-aged at-risk cohort, individuals with subtle cognitive decline (milder than MCI/dementia) showed higher nEV total tau, p181-tau, and p231-tau than stable counterparts, with age-dependent divergence; additionally, longitudinal change in insulin signaling biomarkers differed between decliners and stable participants, and multi-marker panels were able to classify decline status with meaningful performance in training and test sets (Eren et al., 2020). Together with the actigraphy–PET mediation findings, this suggests a testable “chrono-biomarker” logic: circadian timing/fragmentation could be used to stratify when and in whom EV-accessible molecular pathology is most detectable, and EV panels could be used to determine whether circadian disruption is merely symptomatic or already coupled to tau/insulin-resistance–linked neurodegenerative cascades (Eckhardt et al., 2025; Eren et al., 2020).

Finally, the circadian–clearance axis provides a mechanistic reason why timing should matter for what reaches blood: human sleep is accompanied by coordinated slow-wave neural activity, hemodynamic oscillations, and large CSF flow dynamics, consistent with sleep-linked clearance physiology (Fultz et al., 2019). Complementary experimental evidence shows that glymphatic activity can determine whether brain-derived protein biomarkers rise in the periphery sleep deprivation and other manipulations suppressed biomarker appearance in blood in a murine injury paradigm, supporting the principle that brain-to-blood signal transport is glymphatic-state dependent (Plog et al., 2015). In a review framing, the central insight is that chrono-disruption can simultaneously (i) exacerbate pathology, for example by coupling tau and Aβ dynamics to circadian timing, and (ii) distort biomarker visibility through altered clearance and transport, creating a scenario in which the same individual may appear biologically different depending on when and how sampling is performed. This highlights a critical gap in current AD biomarker strategies, which are largely time-agnostic, despite emerging human evidence indicating that temporal context is an integral component of both disease biology and its measurement.

Together, circadian disruption of glial immunometabolism and brain clearance pathways may create time-dependent windows of vulnerability and therapeutic responsiveness in Alzheimer's disease (Fig. 3).

Box 2: key testable predictions of the systems-level timing model.

• •Time-of-day–dependent variability in peripheral neuron-derived extracellular vesicle (nEV) tau and related biomarkers, reflecting circadian gating of brain-to-blood clearance rather than changes in pathological production alone.
• •Improved peripheral biomarker detectability following restoration of sleep consolidation, occurring prior to measurable reductions in amyloid or tau burden, consistent with enhanced clearance and export rather than immediate disease modification.
• •Enhanced central nervous system exposure and efficacy of therapeutic agents when dosing is aligned to circadian-permissive windows, without the need for dose escalation, due to rhythmic blood–brain barrier transport and fluid dynamics.
• •Stronger coupling between circadian fragmentation metrics (e.g., actigraphy-derived intradaily variability) and regional tau pathology, particularly in early Braak-stage regions, supporting temporal coherence as a modifier of vulnerability.
• •Divergent extracellular vesicle fate under intact versus impaired clearance states, with preserved circadian–sleep coupling favoring vesicle-mediated export and biomarker signaling, and circadian failure favoring vesicle retention and proteopathic propagation.

Fig. 3.

| The glymphatic–extracellular vesicle axis as a determinant of chrono-therapeutic efficacy and biomarker readout. Extracellular vesicle (EV) fate, therapeutic exposure, and biomarker visibility are gated by circadian phase and brain clearance state. (A) Under conditions of intact clearance, EVs preferentially undergo brain-to-periphery export, supporting peripheral biomarker detection. When clearance is impaired, EVs are retained within the brain and may instead facilitate pathological propagation. (B) Circadian regulation of blood–brain barrier transport and efflux capacity creates phase-dependent differences in drug exposure, such that identical doses administered at different circadian phases can yield divergent brain concentrations and therapeutic effects. (C) Biomarker readouts are similarly time-dependent: integration of central imaging modalities (e.g., PET tau imaging) with peripheral measures such as neuron-derived EVs, when aligned to circadian phase, can distinguish biological signal from sampling noise. Together, these principles motivate chrono-informed intervention and measurement strategies to improve therapeutic efficacy and interpretability in Alzheimer's disease.

8. Conclusions and future directions

AD has long been framed as a disorder of protein aggregation, synaptic failure, and neuronal loss. The synthesis presented in this Review supports a complementary and integrative perspective: AD is also a disorder of loss of temporal coherence, in which circadian timing, sleep–wake structure, glial immunometabolic readiness, and brain clearance capacity progressively lose coordination. Importantly, this loss of temporal coherence provides a unifying explanation for several features of AD that are otherwise difficult to reconcile, including early sleep disruption, heterogeneous glial responses, impaired proteostasis, variable therapeutic efficacy, and inconsistent biomarker readouts.

Across molecular, cellular, and systems levels, circadian clocks emerge as active regulators rather than passive background oscillators. In the healthy brain, clock-driven programs align glial immune tone, astrocytic water handling, and sleep-dependent glymphatic–lymphatic clearance with predictable daily cycles. In AD, disruption of these programs converts adaptive state transitions into maladaptive, phase-inappropriate responses. Microglia may cycle through acute inflammatory activation into tolerant or metabolically rigid states, astrocytes may lose clock-linked redox and support functions, and clearance pathways may become uncoupled from sleep. These changes do not simply reduce efficiency; they fundamentally alter the logic by which the brain responds to proteotoxic stress.

This temporal framework also reframes the EV literature. Rather than treating EVs as intrinsically pathogenic or protective, their role becomes conditional on brain state and clearance timing. Under intact circadian and sleep-dependent transport, vesicles may participate in waste export and biomarker signaling; under clearance failure, the same vesicular pathways may facilitate proteopathic spread. Recognizing this conditionality helps reconcile apparently conflicting findings and emphasizes the need to interpret EV biology within a systems-level temporally regulated transport context.

From a translational standpoint, the implications are substantial. Therapeutic engagement with the brain is inherently time-dependent, shaped by circadian variation in BBB transport, brain fluid dynamics, immune tone, and clearance capacity. Consequently, treatment failure cannot always be attributed to incorrect targets or insufficient dosing; in some cases, interventions may have been delivered outside biologically permissive windows. This insight motivates a shift toward chrono-pharmacology and chrono-neurotherapeutics, in which timing is treated as a controllable variable rather than a source of noise.

Species differences and translational considerations warrant explicit acknowledgment. Rodent circadian phase, sleep architecture, and glymphatic dynamics differ quantitatively from humans, particularly with respect to polyphasic sleep structure, rest–activity distribution, and relative circadian amplitude. These differences constrain direct temporal extrapolation of clock phase and sleep-state timing from animal models. However, they do not negate conserved principles of clock-gated immune regulation, astrocytic endfoot biology, and sleep-associated fluid dynamics that operate across species. Increasingly, human multimodal imaging, actigraphy–PET studies, and peripheral biomarker analyses provide translational anchors that align core elements of the timing framework with human AD biology.

Future research directions should prioritize three complementary areas. First, temporal phenotyping should be incorporated into both observational studies and clinical trials, using objective measures of sleep, rest–activity rhythms, and circadian stability to define biologically meaningful subgroups. Second, mechanistic dissection of clock-linked glial programs across disease stages will be essential to distinguish reversible temporal misalignment from irreversible degeneration. Third, time-aware biomarker strategies, including longitudinal sampling and integration of neuron-derived extracellular vesicles with imaging and behavioral measures, may improve sensitivity and interpretability by aligning measurement with underlying biology.

In summary, viewing AD as a systems-level timing disorder does not replace established molecular models; rather, it contextualizes them within a dynamic temporal architecture that shapes vulnerability, progression, and therapeutic responsiveness. Treating time as an organizing variable opens new avenues for precision intervention, trial design, and biomarker development, and offers a principled framework for integrating diverse strands of AD biology into a coherent whole.

CRediT authorship contribution statement

Duc-Hiep Bach: Writing – review & editing, Writing – original draft, Supervision, Conceptualization. Thanh Liem Nguyen: Writing – review & editing, Supervision.

Ethics approval and consent to participate

Not applicable. This review synthesizes previously published studies and did not involve new studies with human participants or animals.

Consent for publication

Not applicable.

Availability of data and materials

No new datasets were generated or analyzed for this manuscript. All data discussed are from previously published sources cited in the article.

Disclosure

D.-H.B., utilized ChatGPT and Gemini to assist in the structural outlining of the manuscript and to generate initial summaries of selected literature. D.-H.B., manually selected all reference materials, verified the accuracy of AI-generated summaries against the original texts, and revised the final manuscript to ensure intellectual integrity. The author(s) take full responsibility for the final content.

Funding

No funding was received to support this study.

Declaration of competing interest

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