Glymphatic Dysfunction and Brain Waste Accumulation

The glymphatic system—the brain’s lymphatic-equivalent waste-clearance network—operates primarily during slow-wave sleep, flushing metabolic byproducts (tau, amyloid, glutamate, inflammatory mediators) from the interstitial space. This section examines evidence that glymphatic dysfunction in ME/CFS, driven by impaired slow-wave sleep architecture and aquaporin-4 dysregulation, creates a cycle: poor sleep \(\to\) waste accumulation \(\to\) cognitive impairment and neuroinflammation \(\to\) worse sleep. This provides a mechanistic link between unrefreshing sleep and brain fog.

1 Glymphatic System Physiology and Sleep Dependency

The glymphatic system (named by Nedergaard lab, 2013) is a brain-wide fluid exchange network in which cerebrospinal fluid (CSF) flows along perivascular spaces (para-arterial), exchanges with interstitial fluid (ISF) through astrocytic AQP4 water channels, and drains via para-venous spaces and meningeal lymphatics (Xie et al. 2013). Chayama et al. (2026) demonstrated that endogenous neuron-derived proteins drain primarily to dura, skull, and nasal cavity rather than cervical lymph nodes, suggesting that CSF-tracer-based drainage maps may not reflect physiological protein clearance routes (Chayama et al. 2026). This system clears tau, amyloid-beta, and metabolic waste products. Crucially, it operates preferentially during slow-wave sleep, with a 60% increase in interstitial space during SWS compared to waking (Xie et al. 2013). Sleep duration is less important than sleep quality (specifically SWS content): glymphatic clearance is maximally coupled to delta-oscillation (slow-wave) sleep, not total sleep duration (Nemat-Gorgani et al. 2025).

2 Brain Clearance Architecture: Routes, Kinetics, and Brain-Border Immune Interfaces

TipAchievement: Neuronal Protein Tracing Reveals Physiological Brain Clearance Architecture

Chayama et al. (2026) developed a genetic system (Syn-ZsG) to trace neuron-derived protein clearance from the brain to border tissues without the pressure-driven artefacts of conventional CSF tracer injections. By expressing a fluorescent reporter (ZsGreen) selectively in neurons and tracking its distribution across brain border compartments, they revealed that physiological clearance follows a fundamentally different architecture from what CSF-injected tracers suggest (Chayama et al. 2026).

Key findings:

• Distinct routes from CSF tracers. Neuron-derived proteins drain primarily to dura, skull, and nasal cavity — with relatively low levels in superficial and deep cervical lymph nodes. By contrast, CSF-injected tracers (OVA) concentrate in cervical lymph nodes, bypassing the brain itself. This was validated using biorthogonal labeling of endogenous neuronal proteomes (Syn-PheRS), confirming that the ZsGreen distribution reflects native protein handling, not reporter artefact (Chayama et al. 2026).
• Compartmentalized clearance follows a “nearest exit” principle. By restricting ZsGreen expression to specific brain regions (forebrain, striatum, anterior olfactory), the authors demonstrated that clearance is spatially organized: proteins from each region drain preferentially to the nearest border compartment. Dorsal cortical proteins exit through dorsal dura and skull; striatal proteins drain via basal skull and nasal cavity. This implies that different brain regions have different rates of waste clearance and different failure modes in disease.
• Border-specific clearance kinetics. Pulse-chase experiments revealed that skull-associated borders exhibit slow outflow (\(k = -0.008\)), while dura, nasal cavity, and cribriform plate show rapid turnover (\(k = 0.127--0.261\)). Slow skull kinetics may facilitate prolonged antigen retention and immune surveillance.
• Skull-resident B cells surveil brain antigens for immune tolerance. Transcriptomic analysis revealed that skull B cells that sample neuronal proteins upregulate antigen processing and presentation pathways alongside tolerogenic markers (PD-L1, IL10ra, Cd1d1, Zbtb20, Ptpn22) while downregulating pro-inflammatory genes (Tnf, Il1b, type I interferon signaling). This “tolerogenic-presentation” profile suggests that skull marrow serves as a neuroimmune checkpoint — sampling brain-derived antigens and actively suppressing inflammatory responses against them. PD-L1 protein levels were confirmed elevated on ZsGreen+ skull B cells by quantitative immunostaining.
• Disease disrupts clearance through distinct mechanisms. Acute inflammation (LPS) shunted neuronal proteins away from homeostatic border drainage pathways into the bloodstream via vascular leakage. By contrast, amyloid pathology (5XFAD) caused parenchymal retention and border exit obstruction — proteins accumulated in the brain and failed to reach borders, reducing peripheral clearance. These two failure modes — rerouting vs. obstruction — have fundamentally different consequences for neuroimmune surveillance and peripheral antigen exposure (Chayama et al. 2026).

These findings redefine brain clearance as a compartmentalized system of organized pathways and immune niches whose dysfunction may underlie regional vulnerability in neurological disease.

(Certainty: 0.85 for the methodology and foundational mechanisms in mice. Published in Cell, July 2026, rigorous experimental design with multiple validation approaches. No direct human data beyond conservation of meningeal lymphatics between rodents and humans; precise anatomical flow patterns may differ in primates. No ME/CFS models tested.)

Implications for ME/CFS glymphatic research. The nearest exit principle suggests that different ME/CFS cognitive phenotypes may reflect regional failure of specific clearance compartments. A brain fog-predominant patient whose symptoms centre on executive dysfunction (prefrontal cortex) may have impaired dorsal dural/skull clearance, while a patient with memory impairment (hippocampal) may have distinct clearance deficits. This is entirely speculative — no regional glymphatic measurements exist in ME/CFS — but the compartmentalized architecture provides a mechanistic framework for heterogeneity that the current “global glymphatic failure” model cannot capture.

The skull-resident B cell tolerogenic mechanism raises the reverse possibility: in ME/CFS neuroinflammation, failure of skull border immune tolerance could convert a normally protective neuroimmune checkpoint into a site of autoimmune priming. If skull B cells lose their tolerogenic program (PD-L1 downregulation, loss of IL10ra signalling, shift from antigen presentation to pro-inflammatory activation), brain-derived antigens that normally induce tolerance could instead trigger CNS-directed autoimmunity. This is consistent with the autoantibody findings documented in ME/CFS subsets (Chapter Speculative Mechanistic Hypotheses (Section GPCR Autoantibody-Driven Dysfunction)), though the specific skull B cell mechanism has never been studied in ME/CFS.

The demonstration that ICM-injected tracers (the standard methodology in glymphatic research) produce a fundamentally different distribution from endogenously produced neuronal proteins has direct methodological implications for the field: all existing DTI-ALPS studies in fibromyalgia, Long COVID, and ME/CFS measure a proxy signal that may reflect CSF flow patterns rather than parenchymal clearance. This does not invalidate DTI-ALPS — the fibromyalgia and Long COVID studies show meaningful clinical correlations — but it means the signal being measured may not capture the same pathways that actual brain-derived proteins traverse. A normal DTI-ALPS index does not rule out impaired parenchymal clearance, and an abnormal DTI-ALPS index may reflect disturbed CSF flow without necessarily indicating impaired neuronal waste extraction.

3 Multiple Converging Impairments in ME/CFS

In ME/CFS, several factors converge to impair glymphatic function:

Reduced slow-wave sleep. ME/CFS patients show reduced SWS relative to total sleep time, with alpha-delta intrusion pattern disrupting SWS continuity. While total sleep time may be normal or even extended, the quality is compromised.

AQP4 dysregulation. Aquaporin-4 channels, positioned at astrocytic endfeet facing perivascular spaces, are the critical bottleneck for CSF-ISF exchange. Norepinephrine (elevated in ME/CFS due to autonomic dysregulation) inhibits AQP4 function: chronic adrenergic dysregulation in ME/CFS may chronically suppress glymphatic flow (Nemat-Gorgani et al. 2025). Neuroinflammation can additionally depolarize AQP4 (Ding et al. 2025).

Pre-existing neuroinflammation. Activated microglia disrupt perivascular flow dynamics.

4 Norepinephrine Vasomotion as the Glymphatic Pump

Recent mechanistic work has identified the primary physical driver of glymphatic flow: infraslow oscillations in locus coeruleus (LC) norepinephrine (NE) release generate rhythmic vasomotion in cerebral arteries, which in turn drives CSF pulsation through perivascular spaces (Hauglund et al. 2025). Hauglund et al. (2025) demonstrated using LC-specific optogenetics that NE oscillations are both necessary and sufficient for glymphatic clearance during NREM sleep — optogenetic silencing of LC abolished glymphatic transport, while patterned LC stimulation restored it. Zolpidem, a Z-drug commonly prescribed for sleep in ME/CFS, suppressed NE oscillation amplitude by approximately 50% and proportionally reduced glymphatic flow (Hauglund et al. 2025).

In humans, Fultz et al. (2019) provided the first direct evidence of this cascade using simultaneous EEG and fast fMRI: during NREM sleep, large slow waves of neural activity (\(<\) 0.1 Hz) preceded oscillations in cerebral blood volume, which in turn preceded pulsatile CSF inflow into the fourth ventricle (Fultz et al. 2019). The coupling is sequential and directional: neural slow wave → blood volume change → CSF pulse. This establishes that glymphatic clearance depends not merely on sleep stage (SWS vs. REM) but on the integrity of neurovascular coupling within SWS — a distinction with direct ME/CFS relevance. SleepFM provides large-scale validation of this cross-modal coupling principle, showing that the specific coordination cascade described here (neural → vascular → CSF) represents the type of cross-modal physiological synchrony that is the strongest predictor of disease onset (Thapa et al. 2026).

Implications for ME/CFS. ME/CFS patients have documented LC-noradrenergic dysfunction: the NIH deep phenotyping study found reduced DHPG (the primary NE metabolite) in cerebrospinal fluid, indicating impaired central catecholamine turnover (Walitt et al. 2024). If LC-NE oscillatory dynamics are disrupted — whether from reduced LC neuron firing capacity, altered NE synthesis, or dysregulated feedback — the vasomotion-driven CSF pump would operate at reduced amplitude regardless of sleep architecture quality. This creates a double hit: alpha-delta intrusion reduces the time spent in SWS (reducing the window for clearance), while LC-NE dysfunction reduces the efficiency of clearance during whatever SWS occurs.

Arterial pulsatility provides an independent contributor to glymphatic flow. Hablitz and Nedergaard (2021) showed that reduced arterial pulsatility — from cardiovascular deconditioning, low blood pressure, or reduced cardiac output — can decrease glymphatic transport by up to 50% (Hablitz and Nedergaard 2021). ME/CFS patients with orthostatic intolerance (OI) and POTS exhibit precisely this pattern: reduced cerebral perfusion pressure during upright posture and, potentially, reduced nocturnal arterial pulsatility from autonomic dysfunction. The convergence of LC-NE dysfunction, reduced arterial pulsatility, and impaired SWS architecture represents a triple hit on glymphatic clearance.

5 Dual-Speed Glymphatic Transport: Fast Advective vs. Slow Diffusive Flow

TipAchievement: Physics-Informed AI Maps Dual-Speed Glymphatic Flow

Toscano et al. (2026) developed MR-AIV (Magnetic Resonance Artificial Intelligence Velocimetry), a physics-informed neural network framework that reconstructs 3D fluid velocity fields from dynamic contrast-enhanced MRI. Applied to the mouse brain, MR-AIV revealed two mechanistically distinct glymphatic transport regimes (Toscano et al. 2026):

• Fast advective flow (\(\sim\) 3 μm/s). Perivascular transport along cerebral arteries, driven by the LC-NE vasomotion pump. This is the high-throughput clearance component that moves waste rapidly through paravascular conduits.
• Slow diffusive transport (\(\sim\) 0.1 μm/s). Interstitial diffusion through brain parenchyma, mediated by AQP4 water channels at astrocytic endfeet. This is the rate-limiting bottleneck for waste that has entered deep tissue.

The 30-fold velocity difference is not a continuous spectrum but a genuine bifurcation between two transport modes: advection (pressure-driven, perivascular) and diffusion (concentration-gradient-driven, interstitial). MR-AIV simultaneously provides tissue permeability estimates and pressure field maps — quantities inaccessible to any other method.

(Certainty: 0.50 for the dual-speed mechanism in mice. Caveats: \(n=5\) mice; not yet applied to disease models or humans; method validated against synthetic ground truth (\(<2%\) reconstruction error) not independent physical flow measurements; no replication. The \(<2%\) error is against computational ground truth — a known optimistic form of PINN validation. Published in Science Advances (2026); unreplicated; no ME/CFS data. Downgraded from 0.75: this is a methodological advance in 5 mice. Certainty calibrated to be consistent with other small-n animal studies in this paper (cf. cert 0.30 for n=4–6 mouse studies).)

Implications for ME/CFS. The dual-speed observation adds quantitative resolution to the glymphatic triple hit: perivascular flow is predominantly NE-vasomotion-driven, while interstitial diffusion is AQP4-dependent — a decomposition the triple-hit already described in functional terms. The velocity numbers (~3 vs. ~0.1 \(\mu\)m/s) are mouse values with unknown human scaling; whether the two transport components can be selectively impaired in disease is entirely untested. The simplest hypothesis — that LC-NE dysfunction impairs both components proportionally, since elevated NE both disrupts vasomotion and depolarizes AQP4 — is the parsimonious null model against which any selective-impairment claim must be tested.

CautionSpeculation: Dual-Speed Impairment May Predict Treatment Response Heterogeneity

Certainty: 0.20. If the advective and diffusive transport components can be selectively impaired (an untested premise), two impairment patterns are logically possible: advective-dominant (reduced perivascular flow with preserved diffusion) and diffusive-dominant (reduced diffusion with preserved perivascular flow). In practice, the most common pattern would be mixed impairment (both reduced), consistent with the triple-hit model which already predicts proportional decline driven by shared upstream factors. The selective-pattern hypothesis is speculative because: (a) LC-NE dysfunction impairs both AQP4 (via elevated adrenergic tone) and vasomotion (via oscillatory failure), making selective impairment mechanistically unlikely unless compensatory mechanisms decouple the two; (b) no ME/CFS data exist on perivascular vs. interstitial flow separation; (c) the 30-fold velocity difference between compartments in healthy mice is a property of spatial anatomy, not independent regulatory systems. This speculation provides a framework for future MR-AIV-based studies but does not assert the existence of distinct subtypes.

Falsifiable prediction: MR-AIV in ME/CFS patients must demonstrate genuinely selective impairment (fast flow \(<50%\) control with slow diffusion \(>70%\) control, or vice versa) in a statistically significant subgroup before the selective-impairment framework is empirically supported. If impairment is proportional across both components (mixed pattern in all patients), the unified-impairment null model is retained and dual-speed decoupling adds no therapeutic stratification value beyond what the triple-hit already predicts.

CautionSpeculation: Dual-Speed Decoupling May Explain the “Tired but Wired” Paradox

Certainty: 0.20. (Reduced from 0.25: the “pump running but drainage clogged” metaphor is inconsistent with the paper’s own mechanism — LC-NE dysfunction impairs vasomotion, so the advective “pump” is already degraded. The corrected framing: both components decline together, but the mismatch between waste production and clearance generates the dissonance.) If ME/CFS involves impaired glymphatic clearance across both transport components (as the triple-hit model predicts), waste accumulates at the same time the brain is actively metabolizing. The resulting physiological dissonance — neuronal activity generating waste faster than even the surviving clearance capacity can remove it — may contribute to the “tired but wired” phenotype alongside the established orexin (Section Constraints on, and Rival Readings of, the Torpor/Sickness-Circuit Model) and adenosine (Section Adenosine Accumulation and Pathological Sleep Pressure) mechanisms. This is a convergent, not competing, mechanism: orexin deficiency and adenosine dysregulation generate the subjective experience of unrefreshing sleep; glymphatic impairment explains why that poor sleep fails to restore cognitive function.

6 Inflammation-Induced Clearance Misdirection and PEM

CautionSpeculation: Post-Exertional Malaise May Involve Inflammation-Induced Routing Disruption of Brain Clearance

Chayama et al. (2026) demonstrated that acute LPS-induced inflammation causes neuronal proteins to bypass homeostatic border drainage routes and leak directly into the bloodstream (Chayama et al. 2026). LPS is a maximal TLR4 agonist producing saturating inflammatory activation; PEM involves more modest cytokine elevations in subsets of ME/CFS patients (Section Inflammatory Cytokine-Induced Somnolence and Fatigue), and systematic review finds cytokine findings are heterogeneous and inconclusive. The step from LPS-induced BBB disruption to PEM-level cytokine effects is a large inferential gap — the two stimuli differ in magnitude, receptor engagement, and tissue distribution — and has not been empirically calibrated.

If future work demonstrates that PEM-level cytokine elevations are sufficient to produce similar routing disruption, the mechanism would predict: (a) plasma levels of CNS-specific proteins (neurofilament light chain, GFAP, tau) may rise during severe PEM episodes; (b) the magnitude of rise should correlate with PEM severity; (c) anti-inflammatory pretreatment should reduce both protein leakage and PEM severity. The routing disruption, if it occurs, would be transient since PEM cytokine spikes resolve within 24–72 hours.

Certainty: 0.25. Directly extrapolated from mouse LPS model (Chayama 2026); no human PEM data on neuronal protein leakage. The existing PEM literature focuses on metabolic danger signals and cytokine amplification; this bloodstream-leakage mechanism is entirely speculative and untested in ME/CFS. CNS protein elevation during exercise has been studied in concussion and stroke populations but not in PEM.

7 Cross-Modal Physiological Decoupling as a Unifying Framework

A recent advance from outside ME/CFS research provides large-scale validation of the decoupling principle that unifies these mechanisms. SleepFM (Thapa et al., 2026) is a multimodal AI foundation model trained on 585,000+ hours of polysomnography data from 65,000 participants, using a novel leave-one-out contrastive learning approach that simultaneously analyses EEG, ECG, EMG, pulse, and respiratory signals (Thapa et al. 2026). Published in Nature Medicine, SleepFM predicts the future onset of 130+ health conditions from a single night’s sleep data.

The model’s central finding is directly relevant to ME/CFS pathophysiology: the most informative signal for disease prediction is not any single modality in isolation, but the degree to which biological systems are out of sync with each other. The researchers observed that body constituents operating in decoupled states — “a brain that looks asleep but a heart that looks awake” — carried the strongest predictive weight for developing multiple diseases (Thapa et al. 2026). Retrospectively, the individual sleep abnormalities documented in ME/CFS research (alpha-delta intrusion, autonomic-sleep dissociation, impaired neurovascular coupling) can be unified under this cross-modal decoupling framework — though this unification is post-hoc and the specific mechanisms linking these abnormalities remain to be tested.

TipAchievement: Cross-Modal Decoupling as a Validated Disease-Predictive Principle

SleepFM demonstrates at scale (\(n=65,000\)) that cross-modal physiological decoupling is a generalisable biomarker of pathology, not a method-specific artefact. The model achieved particularly strong predictions for conditions overlapping with ME/CFS domains: Parkinson’s disease (C-index 0.89), dementia (0.85), hypertensive heart disease (0.84), and myocardial infarction (0.81) (Thapa et al. 2026). The leave-one-out contrastive learning approach — which masks one modality and challenges the model to reconstruct it from the others — is operationally identical to quantifying physiological coupling integrity, a concept central to the autonomic and sleep-architecture research reviewed in this chapter.

(Certainty: 0.80 for SleepFM findings; from Nature Medicine, \(n=65,000\), rigorous methodology, not yet independently replicated. Certainty for ME/CFS-specific application: 0.25 — the model has never been applied to ME/CFS polysomnography data, and its training set likely did not include ME/CFS-labelled patients.)

The decoupling framework maps directly onto each ME/CFS sleep domain described above:

• Alpha-delta sleep (Section Sleep Architecture Failure Hypothesis): Cortical delta oscillations (asleep brain) contaminated by wake-like alpha activity (awake brain) — an intra-modality decoupling within a single EEG channel that SleepFM’s architecture could quantify as reduced self-consistency.

• LC-NE oscillatory dysfunction (Section Neuronal Protein Tracing Reveals Physiological Brain Clearance Architecture): Impaired coupling between neural norepinephrine oscillations (\(\sim\) 0.05 Hz) and cerebral arterial vasomotion — a cross-modal decoupling between EEG and cardiovascular signals of the type SleepFM explicitly detects.

• Autonomic-sleep decoupling: Failure of parasympathetic activation to accompany NREM sleep onset, producing elevated nocturnal heart rate that is discordant with EEG sleep staging — precisely the “brain asleep, heart awake” phenomenon (Thapa et al. 2026).

• Glymphatic cascade decoupling (Section Glymphatic Dysfunction and Brain Waste Accumulation): The sequential neural → vascular → CSF coupling cascade (Fultz et al. 2019) may be disrupted at any step, producing an incoherent oscillatory pattern that cannot drive coordinated fluid clearance (Section Can Coupling Strength Be Estimated from Reduced-Modality Data?).

• Circadian decoupling (Section Melatonin Dysfunction and Circadian Disruption): Loss of normal phase synchrony between melatonin onset, cortisol rhythm, core body temperature, and activity cycles — a temporal decoupling of multiple physiological rhythms (McCarthy 2022).

Each of these decoupling patterns contributes to the central ME/CFS complaint of unrefreshing sleep: restorative sleep requires coordinated synchronisation of brain, autonomic, vascular, respiratory, and circadian systems. When these systems are out of phase, the patient may sleep for adequate duration yet achieve none of the physiological benefits that sleep is supposed to provide.

CautionSpeculation: SleepFM as ME/CFS Decoupling Quantification Tool

SleepFM or an equivalent multimodal sleep foundation model, applied to existing ME/CFS polysomnography datasets, could quantify the degree of cross-modal decoupling in patients versus matched controls. Because the model was trained on 65,000 individuals without ME/CFS-specific labels, its learned representation of “normal” physiological coupling provides a normative reference against which ME/CFS decoupling can be measured. Key predictions: (a) ME/CFS patients will show significantly higher multimodal reconstruction error (indicating greater cross-modal decoupling) than healthy controls. (b) Decoupling magnitude will correlate with symptom severity (fatigue, unrefreshing sleep, cognitive dysfunction). (c) Specific decoupling patterns may distinguish ME/CFS subtypes (sleep-predominant, brain fog-predominant, PEM-predominant — Section Connection to Patient Phenotypes). (d) Post-exertional worsening on Day 2 post-CPET will be reflected in increased decoupling magnitude in the intervening night’s sleep data. (e) Decoupling indices may respond to treatment more sensitively than sleep staging metrics, providing a treatment-response biomarker.

Certainty: 0.25. The model architecture is validated (Nature Medicine, \(n=65,000\)), but SleepFM has not been applied to ME/CFS data; the predictions above are mechanistically grounded but entirely untested. The existence of large-scale PSG datasets at established ME/CFS research centres (Stanford Sleep Medicine Center cohort, Bateman Horne Center) makes this a feasible near-term study. Cross-reference to SleepFM Cross-Modal Decoupling Validation in ME/CFS for a detailed study protocol.

Limitations: SleepFM was trained on clinical PSG populations (patients referred for sleep assessment), not general-population screening, which may introduce referral bias. PSG hardware, electrode montages, and scoring conventions vary between centres, potentially limiting transferability. The model’s original training data and weights are not publicly available as of 2026, and the interpretability of its learned features is limited by the black-box nature of foundation models. No ME/CFS-specific data exist.

NoteOpen Question: Is Cross-Modal Physiological Decoupling a Core Pathophysiological Mechanism or an Epiphenomenon?

SleepFM establishes that decoupling predicts disease, but does the decoupling contribute causally to symptoms, or is it a harmless byproduct of genuine pathophysiological processes (e.g., neuroinflammation, metabolic dysfunction)? This distinction matters clinically: if decoupling is causal, interventions that improve coupling (circadian resynchronisation, autonomic training, slow-wave sleep enhancement) should improve symptoms; if epiphenomenal, such interventions would be cosmetic. The causal question can be tested: does pharmacologically enhancing NE oscillatory amplitude (via trazodone, Section Therapeutic Implications) reduce decoupling indices and simultaneously improve next-day symptoms? Does targeted SWS enhancement normalise cross-modal coupling, and does the coupling improvement mediate the symptomatic benefit? These experiments require a multimodal sleep foundation model or equivalent coupling quantification tool and are not currently feasible without access to SleepFM or a comparable system.

ImportantHypothesis: Glymphatic Failure as Driver of Cognitive Symptoms and Unrefreshing Sleep

The glymphatic system — a brain-wide CSF/ISF exchange network driven by astrocytic aquaporin-4 (AQP4) water channels and perivascular fluid dynamics — clears metabolic waste, tau, and amyloid-\(\beta\) primarily during slow-wave sleep (Xie et al. 2013). In ME/CFS, multiple factors converge to impair this system: (1) reduced SWS content from alpha-delta sleep intrusion, (2) chronic adrenergic dysregulation (elevated norepinephrine) inhibiting AQP4 polarization at astrocytic endfeet, and (3) pre-existing neuroinflammation disrupting perivascular flow dynamics (Nemat-Gorgani et al. 2025). Waste accumulation then activates the NLRP3 inflammasome in microglia, producing IL-1\(\beta\) and IL-18 that further disrupt sleep architecture (Nemat-Gorgani et al. 2025) (Ding et al. 2025). The resulting vicious cycle — impaired glymphatic clearance \(\to\) waste accumulation \(\to\) neuroinflammation \(\to\) worse sleep — may explain why ME/CFS patients report unrefreshing sleep despite adequate total sleep duration (Wostyn and De Deyn 2018) (Nemat-Gorgani et al. 2025). SleepFM provides independent, large-scale validation of the principle that disrupted sleep physiology predicts disease onset (Thapa et al. 2026), though the specific glymphatic component of this prediction has not been isolated. (Certainty: Low-Medium for the ME/CFS-specific application; the glymphatic mechanism itself is High certainty. Toscano et al. 2026 (Toscano et al. 2026) provides quantitative resolution of perivascular vs. interstitial transport regimes — a refinement, not a qualitative shift, since the NE-vasomotion (Hauglund 2025) and AQP4-depolarization models were already mechanistic. Direct glymphatic imaging studies in ME/CFS remain lacking.)

The glymphatic hypothesis additionally has implications for drug safety. Several medications commonly used for sleep in ME/CFS — notably zolpidem and other Z-drugs — suppress the NE oscillations that drive glymphatic flow by approximately 50% (Hauglund et al. 2025). Orexin receptor antagonists, alpha-1 adrenergic blockers (doxazosin, prazosin), and antipsychotics with adrenergic antagonism (low-dose quetiapine) may similarly impair the vasomotion-dependent CSF pump through their effects on noradrenergic signalling (Zhu, Yang, and Hashimoto 2025). SleepFM’s large-scale validation (Thapa et al. 2026, Nature Medicine, n=65,000) that cross-modal physiological decoupling during sleep predicts disease onset across 130+ conditions (Thapa et al. 2026) provides a mechanistic rationale for concern about coupling-disrupting medications; however, the observational design cannot establish whether decoupling is causal or epiphenomenal, and the trade-off between improved subjective sleep (which may itself enhance glymphatic function via SWS consolidation) and potential impairment of NE-driven clearance has not been quantified in any condition. The clinical magnitude of this effect in humans is unknown, but the mechanism raises the possibility that some sleep medications, while improving subjective sleep, may paradoxically impair the waste-clearance function that makes sleep restorative.

CautionWarning: Glymphatic-Impairing Medications in ME/CFS Sleep Management

Several drug classes used for ME/CFS sleep management may suppress the norepinephrine oscillations that drive glymphatic clearance during NREM sleep (Hauglund et al. 2025) (Zhu, Yang, and Hashimoto 2025):

• Z-drugs (zolpidem, zopiclone, eszopiclone): Suppress NE oscillation amplitude by ~50% in animal models (Hauglund et al. 2025). Additionally, Parhizkar et al. (2025) demonstrated that zolpidem increases sleep duration in tauopathy mice but provides zero neuroprotection against tau-mediated neurodegeneration — unlike the DORA lemborexant, which reduced tau phosphorylation and preserved hippocampal volume by 30–40% (Parhizkar et al. 2025). This zolpidem-specific failure has two mechanistic explanations: (i) suppressed NE oscillations (Hauglund 2025) impair glymphatic clearance while zolpidem-induced sedation does not trigger naturalistic sleep physiology; (ii) GABA-A agonism does not address the orexin→PKA→tau phosphorylation pathway that DORAs interrupt.
• Orexin receptor antagonists (suvorexant, lemborexant, daridorexant): LC pathway effects may reduce NE-driven vasomotion (Zhu, Yang, and Hashimoto 2025). However, the DORA class carries a potential countervailing benefit not shared by Z-drugs: orexin receptor blockade reduces PKA-dependent tau phosphorylation (Parhizkar et al. 2025) (Lucey et al. 2023). The net risk–benefit balance between possible NE oscillation suppression and possible tau phosphorylation reduction has not been quantified; both directions remain plausible.
• Alpha-1 adrenergic blockers (doxazosin, prazosin): Directly reduce vasomotion amplitude (Zhu, Yang, and Hashimoto 2025)
• Low-dose quetiapine: Alpha-1 blockade component may impair vasomotion (Zhu, Yang, and Hashimoto 2025)

This evidence is from animal models; clinical magnitude in humans is unknown. This does not constitute a recommendation to discontinue these medications. The trade-off between improved subjective sleep (which may itself enhance glymphatic function via SWS consolidation) and potential impairment of NE-driven clearance has not been quantified. Clinicians should be aware of this emerging mechanistic concern. No ME/CFS-specific data exist.

8 Brain-Border Immune Tolerance Failure as ME/CFS Neuroimmune Mechanism

CautionSpeculation: Skull Border B Cell Failure and CNS-Directed Autoimmunity in ME/CFS

Certainty: 0.20. Chayama et al. (2026) identified a population of skull-resident B cells that sample brain-derived neuronal proteins and mount a tolerogenic response — upregulating PD-L1, IL10ra, Cd1d1, Zbtb20, and Ptpn22 while simultaneously activating antigen processing and presentation machinery (Chayama et al. 2026). This “tolerogenic-presentation” profile operates as a neuroimmune checkpoint: brain antigens that reach the skull border are presented to lymphocytes in a regulatory context that actively suppresses inflammatory responses against CNS proteins.

In ME/CFS, chronic neuroinflammation (microglial activation, Section Microglia Activation and Neuroinflammatory Fatigue; cytokine signaling, Section Inflammatory Cytokine-Induced Somnolence and Fatigue) may disrupt this tolerogenic checkpoint through at least three mechanisms: (1) inflammatory cytokines (IL-1\(\beta\), TNF-\(\alpha\), type I interferons) may downregulate PD-L1 and IL10ra expression on skull B cells, converting them from tolerogenic to immunogenic antigen-presenting cells; (2) the inflammatory rerouting demonstrated by Chayama et al. — LPS shunting brain-derived proteins into the bloodstream via vascular leakage (Chayama et al. 2026) — would deliver brain antigens to systemic lymphoid organs (spleen, peripheral lymph nodes) that lack the skull border’s tolerogenic programming, potentially breaking CNS immune tolerance; (3) skull bone marrow channels are remodeled during neuroinflammation, which may alter B cell maturation within the skull niche and bias developing B cells away from tolerogenic programming.

This mechanism would explain how ME/CFS neuroinflammation could generate CNS-directed autoantibodies (such as GPCR autoantibodies documented in subsets of ME/CFS patients, Chapter Speculative Mechanistic Hypotheses (Section GPCR Autoantibody-Driven Dysfunction)) without requiring systemic autoimmune disease. The skull border checkpoint failure model predicts that: (a) autoantibody-positive ME/CFS patients should show evidence of disrupted skull meningeal immune function (e.g., altered CSF B cell profiles, reduced CSF levels of tolerogenic cytokines IL-10 and TGF-\(\beta\)); (b) anti-inflammatory interventions that reduce neuroinflammation (LDN, microglial modulators) should decrease autoantibody titres if checkpoint failure is driven by inflammatory cytokine signalling; (c) the skull border B cell compartment should show reduced PD-L1 and IL10ra expression in ME/CFS patients vs. controls on autopsy or CSF analysis.

Falsifiable prediction: Skull bone marrow aspirates or CSF B cell profiles from ME/CFS patients should show reduced PD-L1 expression (flow cytometry) and a transcriptional shift away from tolerogenic gene expression modules (Ptpn22, Zbtb20, Cd274 downregulation) compared to age-matched controls. A null result — no difference in skull B cell tolerogenic markers between ME/CFS and controls — would refute the checkpoint failure mechanism.

Limitations: Entirely theoretical. Chayama et al. demonstrated skull B cell tolerogenic function in healthy mice; no human skull B cell data exist for any condition. No ME/CFS skull/CSF B cell profiling has been conducted. The mouse skull B cell phenotype has not been studied under neuroinflammatory conditions. The connection between skull checkpoint failure and clinical autoantibody production is speculative and multi-step. This is a cross-species, cross-condition mechanistic extrapolation with no direct evidence.

9 Post-Exertional Symptom Specificity as Compartment-Specific Overload

CautionSpeculation: Post-Exertional Symptom Heterogeneity Reflects Compartment-Specific Clearance Overload

Certainty: 0.20. Chayama et al. (2026) demonstrated that different brain regions drain to distinct border compartments following the nearest exit principle (Chayama et al. 2026). Physical exertion activates widespread brain regions, but the metabolic waste produced by each region drains to different compartments with different clearance kinetics (skull: slow, \(k = -0.008\); dura/nasal/cribriform: rapid, \(k = 0.127--0.261\)). If a patient has pre-existing impairment in a specific compartment, exertion may overwhelm that compartment’s already-reduced clearance capacity, producing compartment-specific symptom exacerbation.

This predicts that ME/CFS patients with predominantly cognitive PEM (brain fog after exertion, preserved motor function) have primarily impaired dorsal skull/dura clearance (serving prefrontal/executive regions), while patients with predominantly sensorimotor PEM (headaches, dizziness, coordination worsening after exertion) have impaired basal skull/nasal clearance. The classic mixed PEM presentation — affecting cognition, sensory processing, and motor function simultaneously — would reflect multi-compartment impairment, which is likely the most common pattern given the multi-system nature of ME/CFS.

Falsifiable prediction: ME/CFS patients with cognitive-predominant PEM should show greater DTI-ALPS or clearance deficits in dorsal (skull/dura) compartments vs. basal compartments on compartment-resolved imaging. Cognitive exertion tasks should increase CSF neuronal protein levels preferentially in dorsal-cortex-draining compartments, while physical exertion tasks should increase levels in basal/sensorimotor-draining compartments. A null result — no compartment-specific PEM-symptom correlation — would refute the specificity claim.

Limitations: No regional clearance measurements exist in ME/CFS. The nearest exit principle is demonstrated in mice; compartmental boundaries in human brain may be more complex or overlapping. PEM is inherently multi-system and mixed patterns are the norm, making compartment-specific effects difficult to isolate. This is a cross-species mechanistic extrapolation.

10 Skull Channel Remodeling as Clearance Biomarker and Structural Stratifier

CautionSpeculation: Skull Bone Marrow Channel Density as a Structural Biomarker of ME/CFS Clearance Stage

Certainty: 0.20. Chayama et al. (2026) show that skull borders are the slowest clearance compartment (\(k = -0.008\)) and that skull-resident B cells maintain tolerogenic neuroimmune surveillance (Chayama et al. 2026). Skull bone marrow channels — direct vascular connections between the skull marrow and the brain surface — are the anatomical substrate for this surveillance and may remodel in disease. Eme-Scolan et al. (2026) demonstrated that neuroinflammation remodels skull channels to increase immune cell trafficking.

In ME/CFS, two opposing remodeling forces may operate at different disease stages. Early disease (less than 2 years): neuroinflammation-driven channel expansion increases drainage capacity as a compensatory response to impaired clearance. Late disease (more than 10 years): perivascular fibrosis and structural degradation (Section Glymphatic Dysfunction and Brain Waste Accumulation, ~70× recovery timescale) narrow skull channels, reducing both clearance capacity and immune surveillance. High-resolution CT or MRI quantitative channel metrics (channel density, mean diameter, total channel volume) could serve as a structural biomarker distinguishing early (compensatory) from late (decompensated) disease stages, with implications for treatment selection: vasodilatory/anti-inflammatory strategies in dilated-channel early disease, anti-fibrotic/clearance-enhancing strategies in narrowed-channel late disease.

Falsifiable prediction: HRCT-measured skull channel density should be elevated in early ME/CFS (less than 2 years, neuroinflammation-driven) and reduced in late ME/CFS (more than 10 years, fibrosis-driven) vs. age-matched controls. Channel density should correlate inversely with CSF neuronal protein accumulation (r less than \(-\) 0.4) and directly with sleep architecture quality (SWS percentage, r greater than 0.4). A null correlation between channel density and any ME/CFS parameter would refute the structural biomarker hypothesis.

Limitations: No human skull channel remodeling data in ME/CFS. Eme-Scolan et al. is a mouse neuroinflammation study. Skull channel imaging methodology is not clinically validated. The compensatory→decompensated transition is entirely hypothetical. Channel density may vary with age, sex, and vascular risk factors unrelated to ME/CFS.

WarningLimitation: DTI-ALPS Signal May Reflect CSF Flow, Not Protein Clearance from Brain Parenchyma

Chayama et al. (2026) demonstrated that CSF-injected tracers distribute to fundamentally different anatomical compartments than endogenously produced neuronal proteins: 50–80% of ICM-injected tracers are recovered in cervical lymph nodes, while neuron-derived proteins drain primarily to dura, skull, and nasal cavity with minimal cervical lymph node representation (Chayama et al. 2026). This divergence persists even when matching tracer duration (14-day chronic CSF infusion produced the same lymph-node-dominated pattern). The implication: conventional glymphatic measurement techniques — including DTI-ALPS, which measures water diffusion along perivascular spaces — may reflect CSF flow dynamics rather than the actual clearance routes used by brain-derived proteins.

This does not invalidate existing fibromyalgia and Long COVID DTI-ALPS findings (Section Chronic Glymphatic Impairment as a Risk Factor for Accelerated Neurodegeneration in ME/CFS), which show clinically meaningful correlations. But it means: (a) a normal DTI-ALPS index does not exclude impaired parenchymal clearance of neuronal waste — proteins may accumulate in brain tissue while perivascular CSF flow appears normal; (b) the DTI-ALPS signal may be dominated by the fast CSF-to-lymph-node pathway that Chayama et al. show is minimally involved in neuronal protein clearance; (c) novel imaging methods that track endogenous protein movement (rather than CSF tracer flow) are needed for glymphatic assessment in ME/CFS; (d) blood-based measurement of brain-derived proteins (NfL, tau, amyloid-\(\beta\)) may complement DTI-ALPS by directly measuring the output of parenchymal clearance rather than the proxy of perivascular fluid motion.

This limitation applies to all DTI-ALPS-based claims in this chapter. Until a method for tracking neuronal protein clearance in humans (analogous to the Syn-ZsG paradigm) is developed, all glymphatic imaging in ME/CFS should be interpreted with the caveat that the signal being measured may represent only a subset — and possibly not the most functionally relevant subset — of brain waste clearance pathways.

11 CSF:Blood Neuronal Protein Ratio as a Disease Stage and Mechanism Biomarker

NoteOpen Question: Does the CSF:Blood Ratio of Neuronal Proteins Distinguish Inflammatory Rerouting from Obstructive Trapping?

Chayama et al. (2026) demonstrated two mechanistically distinct failure modes for brain clearance — acute inflammation shunts proteins into blood (high blood, low border clearance), while amyloid pathology traps proteins in the parenchyma (low blood, low border clearance) (Chayama et al. 2026). These produce opposite predictions for the CSF:blood ratio of neuron-derived proteins (NfL, tau, amyloid-\(\beta\)): inflammatory rerouting should produce high blood neuronal proteins relative to CSF (ratio less than 1), while obstructive trapping should produce high CSF neuronal proteins relative to blood (ratio greater than 1).

If this framework applies to ME/CFS, simultaneous CSF and blood measurement of neuronal proteins could serve as a disease-stage biomarker with direct therapeutic implications. Early disease (less than 2 years, inflammatory-dominant) would predict response to anti-inflammatory/BBB-stabilising interventions; late disease (more than 10 years, obstructive-dominant) would predict response to clearance-enhancing/SWS-restoring interventions. A flat ratio (\(\sim\) 1) throughout disease would refute the rerouting-vs-trapping distinction and suggest a unitary impairment.

Key research questions: (a) Does the CSF:blood neuronal protein ratio shift over disease course from less-than-1 to greater-than-1 in a longitudinal post-infectious ME/CFS cohort? (b) Does this ratio predict treatment response — anti-inflammatory agents in low-ratio patients, SWS-enhancing agents in high-ratio patients? (c) Can blood-only measurement of multiple neuronal proteins (NfL, p-tau181, GFAP) serve as a clinically accessible proxy for the full CSF:blood ratio, avoiding the need for lumbar puncture? (d) Does the ratio correlate with the specific ME/CFS mechanism (BBB integrity biomarkers for rerouting, structural degradation biomarkers for trapping)?

12 Clearance Architecture as a Cross-Disease Framework

CautionSpeculation: ME/CFS-Long COVID Clearance Architecture Parallel: Post-Infectious Disruption of Compartmentalized Brain Clearance

Certainty: 0.30. Both ME/CFS and Long COVID involve post-infectious onset, sleep architecture disruption, cognitive dysfunction, and (in Long COVID) documented DTI-ALPS abnormalities (Tang 2025: reduced DTI-ALPS in post-COVID sleep disorder, \(n=59\); Chaganti 2025: asymmetrical glymphatic dysfunction with BBB disruption correlation) (Tang et al. 2025) (Chaganti, Talekar, and Brew 2025). The Chayama et al. (2026) compartmentalized clearance framework suggests testable predictions about how the two conditions might differ (Chayama et al. 2026).

Long COVID’s documented endothelial dysfunction and BBB disruption (McAlpine 2026) would be expected to produce the inflammatory rerouting pattern — brain antigens shunted into blood through leaky vasculature. ME/CFS, particularly with longer disease duration and the documented structural degradation of the glymphatic apparatus (Section Glymphatic Dysfunction and Brain Waste Accumulation, ~70× recovery timescale), would be expected to shift toward the obstructive trapping pattern — proteins retained in parenchyma with reduced border clearance. This predicts that Long COVID should show higher blood:CSF neuronal protein ratios (inflammatory rerouting), while chronic ME/CFS should show higher CSF:blood ratios (obstructive trapping). The Chaganti 2025 finding of asymmetrical glymphatic dysfunction in Long COVID is consistent with the nearest-exit principle: unilateral BBB disruption should produce unilateral clearance impairment, because each hemisphere drains to its ipsilateral border compartments.

If these predictions are confirmed, the therapeutic implications diverge: Long COVID may benefit more from endothelial-stabilising interventions (tight junction stabilisers, anti-inflammatory agents), while chronic ME/CFS may require clearance-enhancing interventions (SWS restoration, position optimisation, glymphatic pressure gradient support) to overcome the obstructive bottleneck. A null cross-condition comparison — both conditions showing identical blood:CSF ratios — would suggest that post-infectious clearance disruption is a unitary rather than compartment-specific phenomenon.

Falsifiable prediction: Long COVID patients should show higher blood NfL relative to CSF (blood:CSF ratio greater than 1, inflammatory rerouting), while ME/CFS patients with more than 5 years duration should show higher CSF NfL relative to blood (CSF:blood ratio greater than 1, obstructive trapping). Unilateral BBB disruption on DCE-MRI in Long COVID should correlate with ipsilateral DTI-ALPS asymmetry.

Limitations: No direct comparison study exists. CSF measurements require lumbar puncture, limiting sample sizes. Comorbid Long COVID + ME/CFS patients confound the prediction. The degree of compartmentalization in humans may differ from the mouse nearest-exit model. Tang and Chaganti are single studies with moderate sample sizes and partially overlapping populations; replication in independent cohorts is pending.

The cross-modal decoupling framework (Section Post-Exertional Malaise May Involve Inflammation-Induced Routing Disruption of Brain Clearance) suggests therapeutic strategies organised by the specific coupling deficit rather than by drug class. Because decoupling may contribute causally to unrefreshing sleep rather than being merely an epiphenomenon, interventions that restore physiological coordination during sleep may produce symptomatic benefit beyond what sleep-stage metrics alone predict — though this causal hypothesis remains untested and the correlation between decoupling and disease in SleepFM (Thapa et al. 2026) does not establish that coupling restoration improves outcomes.

CautionSpeculation: Pharmacological Coupling Restoration Strategies

Several existing drug classes may improve cross-modal coupling during sleep through distinct mechanisms, though none has been formally tested in ME/CFS for this indication:

• Autonomic coupling: Low-dose propranolol (10–20 mg at bedtime) may reduce nocturnal sympathetic tone, improving EEG–ECG coherence by dampening excessive cardiac responsiveness to cortical arousals. Clonidine (0.05–0.1 mg at bedtime) may stabilise LC-NE oscillatory amplitude without suppressing it, distinct from Z-drugs which suppress oscillation amplitude by \(\sim\) 50% (Hauglund et al. 2025).

• Thalamocortical coupling: Pregabalin (25–75 mg at bedtime) modulates CaV\(\alpha\) 2\(\delta\) calcium channel subunits and may reduce alpha intrusion into delta sleep, though gabapentinoid sleep architecture effects in ME/CFS are documented as adverse and the coupling hypothesis does not override established safety warnings.

• Metabolic coupling: Adenosine homeostasis may be modifiable through caffeine ablation protocols (4–8 weeks complete elimination) combined with S-adenosylmethionine (SAMe, 200–400 mg BID) to support methylation-dependent adenosine clearance pathways. This strategy remains entirely speculative; no controlled trial of adenosine modulation for sleep coupling exists.

• Orexin-mediated coupling: Timing of orexin antagonists (suvorexant, lemborexant) to the individual’s DLMO window may improve state-transition coupling (wake → sleep) without impairing LC-NE oscillatory dynamics. Administering at habitual bedtime rather than 2–3 hours before DLMO may miss the physiological coupling window for the sleep-onset transition.

Certainty: 0.20–0.30 across agents. Each intervention has mechanistic plausibility derived from the decoupling framework but zero direct evidence in ME/CFS for coupling-specific endpoints. Human dosing data for propranolol (autonomic), clonidine (noradrenergic), pregabalin (thalamic), and SAMe (methylation) exist for other indications but not for ME/CFS sleep coupling. The coupling endpoints themselves (EEG–ECG coherence, phase-amplitude coupling at \(<\) 0.1 Hz) are research measurements not validated as clinical outcomes.

Contraindications and interactions: Propranolol may worsen orthostatic intolerance in POTS patients (hypotension exacerbation). Clonidine may cause morning sedation in sensitive patients; start at 0.025 mg and titrate over weeks. Pregabalin carries dependency risk and withdrawal syndrome; reserve for patients with documented alpha-delta phenotype who have failed non-pharmacological approaches. SAMe may trigger mania in bipolar spectrum patients. All medications above must be screened against common ME/CFS co-prescriptions: fludrocortisone (additive hypotension risk with propranolol), midodrine (opposing mechanisms with clonidine), and LDN (unknown interaction). For bedbound patients, see Chapter Urgent Action Plan for Severe Cases. No human ME/CFS dosing data exist for any of these agents used for coupling-specific endpoints.

Falsifiable prediction: Patients selected by decoupling phenotype (high EEG–ECG coherence deficit → propranolol; high alpha-delta intrusion → pregabalin) will show greater symptomatic improvement than unselected patients receiving the same agent. Coupling-matched treatment should outperform coupling-mismatched treatment by \(>\) 30% on sleep refreshment scores.

CautionSpeculation: Non-Pharmacological Coupling Enhancement Strategies

Physiological interventions may improve coupling integrity with lower risk profiles than pharmacological approaches:

• Slow-paced breathing (6 breaths/min, 15 minutes before bedtime): Enhances cardiorespiratory coupling and increases vagal tone via the baroreflex arc. Practising immediately before sleep onset may prime the cardiorespiratory coupling mechanism to persist into early NREM sleep (Zuraikat, Benson, and Hall 2021). This is low-risk, zero-cost, and caregiver-implementable for moderate patients; bedbound/severe patients may lack the respiratory muscle endurance for sustained paced breathing.

• Core temperature modulation: A warm bath or shower (40°C, 15–20 minutes) 90 minutes before bedtime triggers peripheral vasodilation and a subsequent core temperature drop that facilitates the wake-to-sleep autonomic transition. This may improve brain–autonomic coupling at the sleep-onset transition, where decoupling is often maximal. Not suitable for patients with POTS (heat exacerbates vasodilation-mediated hypotension) or MCAS (heat may trigger degranulation).

• Transcutaneous vagal nerve stimulation (tVNS): Low-intensity stimulation (tragus or cymba conchae, 25 Hz, below sensory threshold) during the first NREM cycle may enhance parasympathetic tone and stabilise autonomic oscillations without disrupting sleep architecture. Experimental; no ME/CFS tVNS-sleep data exist.

• Lateral sleeping position: Left lateral decubitus position enhances glymphatic transport geometry versus supine or prone, and may optimise CSF flow even when neurovascular coupling is impaired (Lee et al. 2015). Already recommended for severe patients in Chapter Urgent Action Plan for Severe Cases. Zero cost.

Certainty: 0.15–0.25. The mechanisms are physiologically grounded (respiratory-cardiac coupling, thermoregulatory sleep facilitation) but no ME/CFS-specific coupling outcome data exist. These interventions are low-risk and may be trialled without prescription or specialist equipment, making them accessible even to patients without sleep laboratory access.

Falsifiable prediction: A 4-week protocol combining slow-paced breathing (pre-bedtime) + temperature modulation (90 min pre-bedtime) will improve EEG–ECG coherence during the first NREM cycle by \(>\) 20% and increase subjective sleep refreshment scores by \(>\) 30% versus sleep hygiene alone.

CautionSpeculation: Home-Based Decoupling Monitoring for Severe Patients

Consumer-grade wearables (chest-strap ECG monitors, consumer EEG headbands, pulse oximeters) can approximate some SleepFM decoupling metrics at home. A simplified coupling index derived from ECG (R-R interval variability at \(<\) 0.1 Hz during estimated NREM sleep) and respiratory rate variability may serve as a proxy for full PSG-based decoupling scores. For severe ME/CFS patients (bedbound, unable to attend sleep laboratories), caregiver-administered nocturnal recordings using a single-lead ECG patch and respiratory band could enable coupling-trend monitoring to guide treatment timing (e.g., detecting worsening coupling → adjusting medication schedule, increasing rest). The Bayesian coupling inference model (Section Home-Based Decoupling Monitoring for Severe Patients) may enable estimation of full decoupling scores from reduced-modality data.

Certainty: 0.20. The concept of reduced-modality coupling inference is plausible given SleepFM’s leave-one-out architecture, but has not been validated for home use, consumer-grade hardware, or severe ME/CFS populations. Signal quality and artefact rejection from movement are significant practical barriers in bedbound patients with positional restrictions.

Falsifiable prediction: A home ECG-based coupling index will correlate with laboratory PSG-based decoupling scores with \(r > 0.70\) and will detect PEM-predictive worsening trends with \(>\) 72 hours lead time.

13 Mathematical Models of Physiological Decoupling

The cross-modal decoupling framework can be formalised mathematically, enabling quantitative predictions and model-guided experimental design.

CautionSpeculation: Multi-Oscillator Phase-Lock Model of Sleep Physiology

The interactions among brain (delta/alpha EEG oscillators), autonomic (R-R interval oscillations), vascular (vasomotion oscillations), and respiratory (breathing rate oscillations) systems during sleep can be modelled as coupled oscillators in the Kuramoto framework, where each system \(i\) has a natural frequency \(\omega_i\), and coupling strength \(K_{i,j}\) determines how strongly system \(i\) entrains to system \(j\). In health, coupling strengths are high enough to maintain phase synchrony across systems at the infraslow (\(<\) 0.1 Hz) range critical for glymphatic pumping. In ME/CFS, one or more coupling constants are pathologically reduced, causing partial or complete desynchronisation.

The model predicts: (a) system-level coupling loss propagates nonlinearly — a 20% reduction in neural–vascular coupling \(K_\text{nv}\) may reduce overall glymphatic clearance by >50% because the sequential coupling cascade (neural → vascular → CSF) amplifies upstream deficits; (b) partial restoration of any single coupling pathway (e.g., autonomic–neural via propranolol) will improve overall synchrony by less than the deficit implies, because phase coherence requires sufficient coupling strength across all connected pairs; (c) the most efficient coupling-restoration target depends on the specific decoupling phenotype: autonomic-decoupled patients require autonomic-targeted interventions, thalamocortical-decoupled patients require calcium-modulating interventions.

Certainty: 0.25. (0.20→0.25: SleepFM (Thapa et al. 2026) demonstrates cross-modal decoupling during sleep predicts disease onset across 130+ conditions, supporting the coupling-constant framework at the general-principle level; coupling constants not empirically characterised, no specific ME/CFS mechanism validated) The Kuramoto framework is well-established for biological oscillator systems (circadian biology, neural synchronisation), but has not been applied to multi-system sleep physiology in ME/CFS. Coupling constants for healthy sleep systems are not empirically characterised. The nonlinear clearance prediction is consistent with the Hauglund et al. demonstration that 50% NE oscillation suppression reduces glymphatic clearance proportionally (Hauglund et al. 2025).

Falsifiable prediction: A Kuramoto model with coupling constants fit to healthy control PSG data will reproduce ME/CFS decoupling patterns when coupling constants are reduced by 30–50%. The model will predict that a 20% increase in any single coupling constant improves overall phase coherence by \(<\) 12%, but 20% increases across two coupled pathways produce \(>\) 35% improvement.

NoteOpen Question: Can Coupling Strength Be Estimated from Reduced-Modality Data?

SleepFM’s leave-one-out contrastive learning architecture demonstrates that one modality can be reconstructed from the others. Bayesian inference, with SleepFM-derived prior distributions over physiological coupling patterns, may enable estimation of full five-modality decoupling scores from two or three consumer-accessible signals (ECG + respiratory band + pulse oximeter). This would make decoupling quantification feasible outside sleep laboratories, transforming a research tool into a deployable clinical metric. The key uncertainty is whether the information loss from reduced modalities degrades the coupling signal below clinical utility thresholds. Cross-reference to proposed study in Section SleepFM Cross-Modal Decoupling Validation in ME/CFS.

14 Long-Term Consequences: Neurodegeneration Risk from Chronic Glymphatic Impairment

The glymphatic system clears not only acute metabolic waste but also the protein aggregates implicated in neurodegenerative disease: tau, amyloid-\(\beta\), and alpha-synuclein. If glymphatic clearance is chronically impaired in ME/CFS — as the mechanisms in Sections Glymphatic Dysfunction and Brain Waste Accumulation and Neuronal Protein Tracing Reveals Physiological Brain Clearance Architecture predict — the question of long-term neurodegeneration risk arises.

The evidence linking sleep disruption to neurodegenerative protein accumulation is quantitative. Holth et al. (2019) demonstrated in mice that brain interstitial fluid (ISF) tau increases approximately 90% during normal wakefulness compared to sleep, and that chronic sleep deprivation accelerates tau seeding and spreading in a tauopathy model (Holth et al. 2019). In humans, a single night of sleep deprivation increased cerebrospinal fluid tau by more than 50% (Holth et al. 2019). Ju et al. (2017) showed that targeted disruption of slow-wave activity (SWA) — using acoustic stimulation to degrade SWS without reducing total sleep time — elevated CSF amyloid-\(\beta\) 40 and amyloid-\(\beta\) 42 levels, with SWA disruption magnitude correlating with amyloid increase (\(r=0.61\)) (Ju et al. 2017). Actigraphy-measured sleep quality also correlated with CSF tau levels (Ju et al. 2017). Crucially, Ju et al. disrupted SWA quality rather than sleep duration — precisely the pattern seen in ME/CFS alpha-delta sleep.

The closest tested analogue is Long COVID. Chaganti et al. (2025) measured glymphatic function using DTI-ALPS (diffusion tensor imaging along perivascular spaces) in Long COVID patients with brain fog and found significant reduction in the left-hemisphere DTI-ALPS index compared to controls, inversely correlated with blood-brain barrier permeability (Chaganti, Talekar, and Brew 2025). Tang et al. (2025) independently found reduced DTI-ALPS in post-COVID sleep disorder patients (\(n=59\)), with strong correlation between DTI-ALPS and sleep quality (\(r=-0.64\)) and partial reversibility over time (Tang et al. 2025). Both studies provide the methodological blueprint — and the mechanistic precedent — for equivalent measurements in ME/CFS.

CautionSpeculation: Chronic Glymphatic Impairment as a Risk Factor for Accelerated Neurodegeneration in ME/CFS

Certainty: 0.35. (0.30→0.35: Chayama et al. 2026 (Chayama et al. 2026) demonstrates that brain-derived protein clearance is compartmentalized and that disease disrupts clearance through distinct mechanisms — inflammation drives vascular leakage into blood, while proteinopathy causes parenchymal retention — providing mechanistic granularity beyond the prior “global glymphatic impairment” model. 0.25→0.30 previously: SleepFM (Thapa et al. 2026) demonstrates cross-modal decoupling during sleep predicts disease onset across 130+ conditions including neurodegenerative diseases (C-index 0.85 for dementia, 0.89 for Parkinson’s), supporting glymphatic coupling disruption as a mechanistic pathway at the general-principle level; no ME/CFS-specific neurodegeneration data) ME/CFS patients with chronic impairment of slow-wave sleep quality (alpha-delta intrusion), LC-NE oscillatory dysfunction (Section Neuronal Protein Tracing Reveals Physiological Brain Clearance Architecture), and reduced arterial pulsatility (OI/POTS) may accumulate tau and amyloid-\(\beta\) at rates exceeding age-matched controls. If sustained over years to decades, this triple-hit on glymphatic clearance could elevate long-term risk of Alzheimer’s disease or related tauopathies.

Chayama et al. (2026) provide direct experimental evidence that inflammation and proteinopathy disrupt brain clearance through fundamentally distinct mechanisms (Chayama et al. 2026). In the 5XFAD amyloidosis model, brain-derived proteins were trapped within the parenchyma and failed to reach border exit routes, reducing peripheral clearance to blood by ~45% at 7 months relative to 2 months of age. By contrast, acute LPS-induced inflammation shunted proteins directly into the bloodstream through a compromised BBB, bypassing homeostatic border drainage entirely. Both mechanisms could operate in ME/CFS: chronic neuroinflammation (Section Microglia Activation and Neuroinflammatory Fatigue, Section Inflammatory Cytokine-Induced Somnolence and Fatigue) may drive the inflammatory rerouting pattern, shunting brain antigens into the peripheral circulation where they encounter systemic immune cells outside the tolerogenic environment of the skull border; while the progressive protein accumulation predicted by impaired clearance could eventually produce the obstructive pattern, further reducing border drainage. The specific combination — inflammatory rerouting early in disease vs. obstructive trapping in chronic disease — is untested and mechanistically speculative, but the Chayama et al. framework provides testable predictions for distinguishing them by measuring the ratio of brain-derived proteins in blood (high with rerouting, low with obstruction) vs. CSF (high with obstruction, normal with rerouting).

Supporting evidence:

• One night of human sleep deprivation → \(>\) 50% CSF tau increase (Holth et al. 2019)
• Targeted SWA disruption (not sleep duration reduction) → elevated CSF amyloid-\(\beta\) (\(r=0.61\)) (Ju et al. 2017)
• Chronic mouse sleep deprivation → tau pathology spreading (Holth et al. 2019)
• Long COVID shows reduced DTI-ALPS glymphatic index, persistent at 12 months (Chaganti, Talekar, and Brew 2025) (caveat: DTI-ALPS may reflect CSF flow rather than parenchymal protein clearance, per @lim-ch15-dti-alps-caveat; the correlations are clinically meaningful regardless of which clearance pathway they track)
• Post-COVID sleep disorder correlates with DTI-ALPS reduction (\(r=-0.64\)), partially reversible (Tang et al. 2025) (same DTI-ALPS caveat applies)
• Inflammatory LPS rerouting and amyloid obstructive trapping demonstrated as mechanistically distinct by Chayama 2026 (Chayama et al. 2026)
• An inflammation-driven route to CNS pathology (independent of clearance failure) is supported cross-disorder: meta-analyses found Alzheimer’s inflammation dissociable from psychiatric comorbidity and mechanistically upstream of neurodegeneration, with IL-17A and G-CSF shared between Alzheimer’s and depression/PTSD (Kuring et al. 2026) (Kuring et al. 2023) (Section bidirectional mood inflammation template). This adds a second, clearance-independent mechanism by which chronic inflammation in ME/CFS (if sustained) could contribute to CNS degeneration — complementary to, not mutually exclusive with, glymphatic impairment. An exposure-defined longitudinal version of the same clearance-independent route comes from a large electronic-health-record cohort: acute brain-parenchymal inflammation (encephalitis) predicts a two- to five-fold increase in long-term dementia risk, strongest for non-infectious/post-infectious inflammatory (autoimmune) etiologies (Aditi et al. 2026) (Section encephalitis dementia precedent).
• DTI-ALPS signal may reflect CSF flow rather than parenchymal clearance of brain-derived proteins (Chayama et al. 2026) — normal DTI-ALPS does not exclude impaired neuronal waste extraction

Testable predictions:

• 1. ME/CFS patients with \(>\) 5 years disease duration show elevated plasma p-tau181 and/or NfL vs. age-matched controls
• 2. ME/CFS patients show reduced DTI-ALPS glymphatic index, correlating with brain fog severity and disease duration; however, DTI-ALPS normalcy does not rule out impaired parenchymal clearance
• 3. Interventions improving SWS quality in ME/CFS (sodium oxybate, trazodone, circadian resynchronisation) reduce CSF tau relative to baseline
• 4. Registry-linkage studies show elevated age-adjusted dementia incidence in ME/CFS cohorts vs. general population
• 5. Blood:CSF ratio of neuron-derived proteins (NfL, tau) distinguishes inflammatory rerouting (high blood, low CSF) from obstructive trapping (low blood, high CSF)

Important limitation: This is extrapolation from sleep deprivation studies, Long COVID imaging, and mouse genetic tracing to a disease in which neither glymphatic function nor neurodegeneration biomarkers have been directly measured. The Chayama et al. findings are in mice; conservation of the nearest exit principle and the distinct failure modes to humans, particularly in ME/CFS, is entirely unstudied. The speculation is mechanistically grounded but all specific ME/CFS predictions remain untested.

The glymphatic hypothesis provides a mechanistic rationale for a clinically observed but poorly explained phenomenon: ME/CFS patients often sleep for adequate or extended durations yet report profoundly unrefreshing sleep and persistent cognitive impairment (Wostyn and De Deyn 2018). Since glymphatic clearance is maximally coupled to delta-oscillation (slow-wave) sleep — not total sleep duration — the alpha-delta intrusion pattern documented in ME/CFS polysomnography (non-delta electroencephalographic activity during NREM sleep) would suppress glymphatic flow regardless of how long the patient sleeps (Nemat-Gorgani et al. 2025) (Xie et al. 2013). Treatment implication: improving sleep quality (SWS content) is likely more therapeutically relevant than extending sleep duration (Nemat-Gorgani et al. 2025).

15 Progressive Degradation of the Glymphatic Apparatus: Why Sleep Restoration Alone Is Insufficient

Certainty: 0.20. An under-recognized consequence of chronic glymphatic impairment is that the clearance machinery itself degrades over time. Years of impaired flow do not merely allow waste to accumulate — they progressively damage the anatomical structures that perform clearance, creating a recovery lag that extends far beyond the timescale at which sleep architecture can be pharmacologically restored.

Three sites of structural degradation. (1) AQP4 depolarization at astrocytic endfeet. Aquaporin-4 channels are normally concentrated in the perivascular endfoot membranes where they create the osmotic gradient driving CSF-ISF exchange. Chronic adrenergic dysregulation — elevated nocturnal norepinephrine sustained over years — progressively delocalizes AQP4 from these membranes, redistributing it across the astrocytic cell body where it cannot contribute to perivascular fluid transport (Nemat-Gorgani et al. 2025). This is not an acute inhibitory effect reversible over hours (like NE suppression of AQP4 conductance); it is a protein trafficking defect that requires endfoot repolarization, which in turn requires sustained normalization of perivascular NE tone — a process measured in weeks to months, not nights. (2) Reactive astrogliosis and endfoot retraction. Chronic neuroinflammation drives astrocyte reactivity, which includes morphological changes to the endfoot processes that wrap around cerebral vessels. Reactive astrocytes retract endfeet, widening the perivascular gap and reducing the contact surface area through which AQP4-mediated transport occurs (Ding et al. 2025). This structural remodeling — analogous to podocyte foot process effacement in the kidney glomerulus — persists after the inflammatory stimulus resolves, because astrocytic cytoskeletal reorganization involves intermediate filament accumulation (GFAP upregulation) that takes months to reverse. (3) Perivascular fibrosis and basement membrane thickening. The paper already documents basement membrane thickening as a microvascular feature of ME/CFS (Section Engineered Exosome-Mediated HSP70 mRNA Delivery as Proof-of-Principle for CNS mRNA Therapy). Chronic waste product accumulation in the perivascular space — including misfolded proteins, inflammatory mediators, and extracellular matrix fragments — triggers perivascular fibrosis analogous to the interstitial fibrosis seen in chronic kidney disease. A thicker, less compliant perivascular basement membrane resists the pressure gradients that drive glymphatic flow, regardless of whether NE oscillations and SWS architecture are normalized.

The recovery lag. These three structural changes impose a temporal decoupling between sleep quality and clearance efficiency. Pharmacological restoration of SWS architecture (via trazodone, DORA) can improve the signal (NE oscillations, delta power, SWS duration) within days — the relevant pharmacology is receptor-level (orexin blockade, 5-HT2A antagonism, H1 blockade), and the relevant timescale is the drug’s half-life plus the time to establish a new steady-state sleep microarchitecture, approximately 3–5 nights. But the machinery that converts that signal into clearance (AQP4 polarity, endfoot contact area, perivascular compliance) requires structural remodeling at timescales governed by protein trafficking, cytoskeletal reorganization, and extracellular matrix turnover:

Component	Acute restoration	Structural recovery	Ratio	Basis
SWS architecture (delta power, spindle density)	3–7 days	—	1×	H1/5-HT2A blockade + orexin normalization; EEG response measurable night 1
AQP4 endfoot repolarization	7–14 days (partial)	2–4 months (full)	~20×	Protein trafficking: half of mislocalized AQP4 redistributes to endfeet on the timescale of astrocytic membrane recycling (~2 weeks); full repolarization requires multiple turnover cycles plus sustained low perivascular NE
Astrocytic endfoot re-extension	2–4 weeks (initial GFAP reduction)	3–6 months (endfoot contact area restoration)	~40×	Cytoskeletal reorganization: intermediate filament (GFAP) half-life is weeks in reactive astrocytes; disassembly + process re-extension + stable contact formation is a sequential multi-week process each, totaling months
Perivascular fibrosis / BM regression	Not meaningful	6–18 months	~100×	Extracellular matrix turnover: collagen IV half-life in basement membranes is 120–300 days in the CNS; MMP-mediated degradation + inhibited new collagen deposition + scar remodeling is the slowest biological process in the recovery cascade

The weighted average across components — factoring that SWS architecture normalizes in ~5 days while full structural recovery spans ~12 months — produces a composite ratio of approximately 70× (range 20–100× depending on the dominant site of impairment). This is not a precise measurement; it is an order-of-magnitude estimate from the biological timescales of the relevant structural processes. The practical implication: a patient who achieves objectively normal sleep architecture after years of non-restorative sleep may still have impaired glymphatic function — the pump is running at full speed but the pipes are narrowed, the seals are leaky, and the osmotic gradient is blunted. Sleep architecture normalization and clearance efficiency normalization are separated by two orders of magnitude in time.

What closes the gap. Four interventions target different nodes in this structural recovery: (1) sustained NE normalization (DORA or low-dose propranolol/clonidine at bedtime, over months) allows AQP4 repolarization — this is the slowest step because it requires persistent perivascular NE tone reduction, not acute receptor blockade; (2) anti-inflammatory strategies (LDN, microglial modulation) reduce reactive astrogliosis and permit endfoot re-extension; (3) glymphatic pressure gradient support (slow-paced breathing at 6 breaths/min, lateral sleeping position, elevated head position) mechanically optimizes flow through still-compromised pathways; (4) pacing (reducing daily metabolic waste production at source) reduces the clearance burden that the damaged apparatus must handle, buying time for structural recovery. No single intervention accelerates all three sites simultaneously. The recovery timescale is not measured in nights. It is measured in months of consistent multi-modal intervention.

Falsifiable prediction. ME/CFS patients with >5 years of non-restorative sleep who achieve normalized PSG sleep architecture (delta power, SWS percentage, spindle density) after 4 weeks of trazodone + DORA should still show reduced DTI-ALPS glymphatic index compared to age-matched controls, while patients treated for >6 months should show progressive DTI-ALPS improvement approaching (but not necessarily reaching) control levels. A null result — DTI-ALPS normalization within 4 weeks of sleep architecture restoration — would refute the structural degradation claim and support a purely functional (reversible) model of glymphatic impairment.

Limitations. Entirely theoretical. No study has measured AQP4 polarization, astrocytic endfoot morphology, or perivascular compliance in ME/CFS patients before and after sleep restoration. The structural remodeling timescales are inferred from analogous processes (kidney podocyte recovery, reactive astrogliosis resolution in TBI models) with unknown applicability to chronic glymphatic impairment. DTI-ALPS is an indirect measure of glymphatic function and its sensitivity to the proposed structural changes is unvalidated. The claim that sleep architecture normalization does not equate to clearance normalization is a testable prediction, not an established finding.

Progressive improvement under sustained restorative sleep. The structural degradation model is not a claim that damage is permanent. It is a claim that the recovery timescale is ~70× the restoration timescale. If consistent restorative sleep is maintained for months — via nightly DORA + trazodone + glymphatic position optimization — the same structural processes that degrade over years would begin to reverse at the rates in the table above. The trajectory is nonlinear: at week 2, AQP4 has partially repolarized but endfeet are still retracted and perivascular fibrosis is unchanged, yielding a DTI-ALPS improvement of perhaps 10–15% of the deficit. At month 3, AQP4 is fully polarized and endfeet have begun re-extending, clearing the middle of the deficit. At month 12, perivascular BM has remodeled sufficiently that flow resistance approaches control levels. The patient may feel no different for the first 6–8 weeks despite objectively improving glymphatic function, because accumulated waste must be cleared before net benefit accrues (the brain must first pay down the waste debt before the balance sheet shows positive clearance). Then, at roughly month 3 — when AQP4 is fully repolarized, endfeet are partially re-extended, and accumulated waste has been substantially cleared — cumulative clearance gains cross into symptomatic territory: less brain fog, reduced PEM trigger sensitivity, more consistent cognitive stamina. This threshold timing is consistent with patient reports that sleep interventions “finally started working” after 8–12 weeks of sustained use. The model predicts that treatment discontinuation at week 2–4 produces no lasting benefit, while 3–6 months of continuous normalized sleep produces progressive and partially retained improvement, and 12+ months approaches the ceiling of structural recovery.

{{/* Phase 10a retroactive synthesis: brain-clearance architecture capstone (condenses ~19 clearance environments in ch15 + ch08 bridges) */}}

TipSynthesis: Brain Clearance Architecture as an Integrative Framework for ME/CFS Neurological Symptoms

The environments in this chapter assemble the Chayama 2026 brain-clearance findings, the Toscano 2026 dual-speed glymphatic model, and the orexin–vasomotion literature into a single architectural framework for how waste-clearance failure could produce ME/CFS neurological symptoms — while keeping the large gap between the animal mechanisms and any human ME/CFS measurement in full view. The organising insight is that physiological brain clearance follows a nearest-exit principle to dura, skull, and nasal routes rather than the cervical-lymphatic route that CSF-tracer studies implied (Neuronal Protein Tracing Reveals Physiological Brain Clearance Architecture), and that it operates in two regimes — fast perivascular advection driven by noradrenergic vasomotion and slow interstitial diffusion (Physics-Informed AI Maps Dual-Speed Glymphatic Flow). If the rodent clearance anatomy translates to humans and if clearance is impaired in ME/CFS — both untested — this architecture would make several otherwise-disconnected ME/CFS observations coherent. First, it distinguishes two failure modes with opposite biomarker signatures: inflammatory rerouting (proteins shunted to blood; predicted high plasma, low CSF) versus obstructive trapping (parenchymal retention; predicted low plasma, high CSF), a distinction proposed both as a disease-stage marker (Does the CSF:Blood Ratio of Neuronal Proteins Distinguish Inflammatory Rerouting from Obstructive Trapping?, DTI-ALPS Signal May Reflect CSF Flow, Not Protein Clearance from Brain Parenchyma) and as a possible axis separating Long COVID from chronic ME/CFS (ME/CFS-Long COVID Clearance Architecture Parallel: Post-Infectious Disruption of Compartmentalized Brain Clearance). Second, it supplies a mechanistic route from inflammation to symptom via orexin: an inflammation→PGE2→orexin-suppression→reduced NE-vasomotion→impaired-clearance loop (The Orexin–Vasomotion–Glymphatic Triad: A Unified Sleep Failure Model, PGE₂–EP3 Self-Sustaining Feedback Loop: Orexin Suppression as a CNS Disease Maintenance Mechanism), which — episodically amplified by exertion — yields a PEM “ratchet” (Glymphatic Failure as a Mechanism for Post-Exertional Cognitive Crash) and a self-reinforcing sleep–clearance vicious cycle explaining unrefreshing sleep (Glymphatic Failure as Driver of Cognitive Symptoms and Unrefreshing Sleep). Third, the nearest-exit / compartment structure offers a candidate explanation for why PEM symptoms are regionally heterogeneous (cognitive vs sensorimotor), via compartment-specific clearance overload (Post-Exertional Symptom Heterogeneity Reflects Compartment-Specific Clearance Overload), and raises the possibility that chronic neuroinflammation breaks the skull-border B-cell tolerogenic checkpoint (Skull Border B Cell Failure and CNS-Directed Autoimmunity in ME/CFS) — with skull-channel remodelling as a speculative structural biomarker (Skull Bone Marrow Channel Density as a Structural Biomarker of ME/CFS Clearance Stage). Long-term, chronic clearance failure is proposed as a neurodegeneration risk factor (Chronic Glymphatic Impairment as a Risk Factor for Accelerated Neurodegeneration in ME/CFS).

The framework’s honest status is that its mechanistic core rests on rodent work (Chayama Syn-ZsG tracing; Toscano n=5 mice, unreplicated), while every ME/CFS application is speculative — the constructive symptom-linking speculations sit at certainty 0.15–0.35 precisely because glymphatic function has never been directly imaged in ME/CFS. DTI-ALPS, the most-cited human proxy, may reflect CSF flow rather than the parenchymal protein clearance Chayama shows drains to different compartments (DTI-ALPS Signal May Reflect CSF Flow, Not Protein Clearance from Brain Parenchyma), so even the existing indirect human data may not measure what the model needs. No study has measured AQP4 polarization, dual-speed transport, or CSF:blood neuronal-protein ratios in ME/CFS patients. The central open question is therefore not whether brain clearance is biologically important — the rodent and neurodegeneration literature strongly suggest that — but whether clearance failure is a causal driver of ME/CFS symptoms or a downstream consequence of the sleep disruption and neuroinflammation documented elsewhere in this chapter; the single most decisive experiment is direct glymphatic imaging (MR-AIV or dynamic contrast MRI) in an ME/CFS cohort with paired CSF:blood neuronal-protein measurement.

Consequence: If waste-clearance failure genuinely drives ME/CFS brain symptoms, it would reframe unrefreshing sleep, brain fog, and post-exertional cognitive crashes as a plumbing problem — the brain generating waste faster than it can be flushed — which would make sleep-quality (not sleep-duration) interventions and clearance-preserving drug choices central to treatment, and would predict a specific blood-versus-spinal-fluid protein pattern that could stage the disease; but because the direct evidence is entirely from mice and no clearance measurement yet exists in ME/CFS patients, this is a well-organised research programme awaiting its first decisive human imaging study, not a basis for current clinical decisions.

References

Aditi, Thomas Blackwell, Xiang Fang, Surendra Sharma, Matthew Mendoza, Alan Landay, Georgiy Golovko, and Parimal Samir. 2026. “Long-Term Risk of Dementia Following Encephalitis: A Large-Scale Retrospective Cohort Study of Electronic Health Records.” Journal of Neurology 273 (9): 511. https://doi.org/10.1007/s00415-026-14017-3.

Chaganti, Jitender R., T. K. Talekar, and Bruce J. Brew. 2025. “Asymmetrical Glymphatic Dysfunction in Patients with Long Covid Associated Neurocognitive Impairment — Correlation with BBB Disruption.” BMC Neurology 25 (1): 112. https://doi.org/10.1186/s12883-025-04133-4.

Chayama, Yuki, Narsimha R. Rao, Dheeraj Perla, Ian Bowers, Seung Eon Cho, Malabika Das, Shabnam Etemadi, et al. 2026. “Physiological Brain Clearance Architecture Revealed by Neuronal Protein Tracing.” Cell 189: 1–19. https://doi.org/10.1016/j.cell.2026.04.048.

Ding, Feng, Kai Nie, Jian Wang, Yong Zhang, and Peng Lei. 2025. “Glymphatic System Impairment in Alzheimer’s Disease and Potential Therapeutic Strategies.” Ageing Research Reviews 93: 101800. https://doi.org/10.1016/j.arr.2024.101800.

Fultz, Nina E., Giorgio Bonmassar, Kawin Setsompop, Robert A. Stickgold, Bruce R. Rosen, Jonathan R. Polimeni, and Laura D. Lewis. 2019. “Coupled Electrophysiological, Hemodynamic, and Cerebrospinal Fluid Oscillations in Human Sleep.” Science 366 (6465): 628–31. https://doi.org/10.1126/science.aax5440.

Hablitz, Lauren M., and Maiken Nedergaard. 2021. “The Glymphatic System: A Novel Component of Fundamental Neurobiology.” Journal of Neuroscience 41 (37): 7698–7711. https://doi.org/10.1523/JNEUROSCI.0619-21.2021.

Hauglund, Natalie L., Mie Andersen, Klaudia Tokarska, Tessa Radovanovic, Celia Kjaerby, Frederikke L. Sørensen, Zuzanna Bojarowska, et al. 2025. “Norepinephrine-Mediated Slow Vasomotion Drives Glymphatic Clearance During Sleep.” Cell 188 (3): 606–622.e17. https://doi.org/10.1016/j.cell.2024.11.027.

Holth, Jerrah K., Sarah K. Fritschi, Chanung Wang, Nigel P. Pedersen, John R. Cirrito, Thomas E. Mahan, Mary Beth Finn, et al. 2019. “The Sleep-Wake Cycle Regulates Brain Interstitial Fluid Tau in Mice and CSF Tau in Humans.” Science 363 (6429): 880–84. https://doi.org/10.1126/science.aav2546.

Ju, Yo-El S., Sharon J. Ooms, Courtney Sutphen, Shannon L. Macauley, Margaret A. Zangrilli, Gina Jerome, Anne M. Fagan, et al. 2017. “Slow Wave Sleep Disruption Increases Cerebrospinal Fluid Amyloid-\(\beta\) Levels.” Brain 140 (8): 2104–11. https://doi.org/10.1093/brain/awx148.

Kuring, J. K., J. L. Mathias, L. Ward, and G. Tachas. 2023. “Inflammatory Markers in Persons with Clinically-Significant Depression, Anxiety or PTSD: A Systematic Review and Meta-Analysis.” Journal of Psychiatric Research 168: 279–92. https://doi.org/10.1016/j.jpsychires.2023.10.009.

———. 2026. “Inflammatory Markers Associated with Dementia: A Systematic Review and Meta-Analysis.” Journal of Psychiatric Research 197: 26–40. https://doi.org/10.1016/j.jpsychires.2026.02.044.

Lee, Hedok, Lulu Xie, Mei Yu, Hongyi Kang, Tian Feng, Rashid Deane, Jean Logan, Maiken Nedergaard, and Helene Benveniste. 2015. “The Effect of Body Posture on Brain Glymphatic Transport.” Journal of Neuroscience 35 (31): 11034–44. https://doi.org/10.1523/JNEUROSCI.1625-15.2015.

Lucey, Brendan P., Haiyan Liu, Cristina D. Toedebusch, David Freund, Tiara Redrick, Samir L. Chahin, Kwasi G. Mawuenyega, et al. 2023. “Suvorexant Acutely Decreases Tau Phosphorylation and a\(\beta\) in the Human CNS.” Annals of Neurology 94 (1): 27–40. https://doi.org/10.1002/ana.26641.

McCarthy, Michael J. 2022. “Circadian Rhythm Disruption in Myalgic Encephalomyelitis/Chronic Fatigue Syndrome: Implications for the Post-Acute Sequelae of COVID-19.” Brain, Behavior, & Immunity – Health 20: 100412. https://doi.org/10.1016/j.bbih.2022.100412.

Nemat-Gorgani, Neda, Benjamin A Plog, Feng Ding, Kai Nie, Jian Wang, Yong Zhang, and Peng Lei. 2025. “Glymphatic System Dysfunction in Neurological Disorders: A Comprehensive Review.” Nature Reviews Neurology 21: 123–38. https://doi.org/10.1038/s41582-024-00912-3.

Parhizkar, Samira, Xin Bao, Wei Chen, Nicholas Rensing, Yusi Chen, Maia Kipnis, Sophie Song, et al. 2025. “Lemborexant Ameliorates Tau-Mediated Sleep Loss and Neurodegeneration in Males in a Mouse Model of Tauopathy.” Nature Neuroscience 28 (7): 1460–72. https://doi.org/10.1038/s41593-025-01966-7.

Tang, Y. L., H. B. Chen, P. Liu, Y. H. Liao, and A. Xie. 2025. “Glymphatic Function Alterations in Sleep Disorder Patients Post-COVID-19: A Longitudinal DTI-ALPS Study.” Nature and Science of Sleep 17: 1377–90. https://doi.org/10.2147/NSS.S522745.

Thapa, R., M. R. Kjaer, E. Mignot, J. Zou, et al. 2026. “A Multimodal Sleep Foundation Model for Disease Prediction.” Nature Medicine. https://doi.org/10.1038/s41591-025-04133-4.

Toscano, Juan Diego, Yisen Guo, Zhibo Wang, Mohammad Vaezi, Yuki Mori, George Em Karniadakis, Kimberly A. S. Boster, and Douglas H. Kelley. 2026. “MR-AIV Reveals in Vivo Brain-Wide Fluid Flow with Physics-Informed AI.” Science Advances. https://doi.org/10.1126/sciadv.aeb0404.

Walitt, Brian, Komudi Singh, Samuel R LaMunion, et al. 2024. “Deep Phenotyping of Post-Infectious Myalgic Encephalomyelitis/Chronic Fatigue Syndrome.” Nature Communications 15 (1): 907. https://doi.org/10.1038/s41467-024-45107-3.

Wostyn, Peter, and Peter Paul De Deyn. 2018. “The Putative Glymphatic Signature of Chronic Fatigue Syndrome: A New View on the Disease Pathogenesis and Therapy.” Medical Hypotheses 118: 142–45. https://doi.org/10.1016/j.mehy.2018.07.007.

Xie, Lulu, Hongyi Kang, Qiwu Xu, Michael J Chen, Yonghong Liao, Mark Thiyagarajan, John O’Donnell, et al. 2013. “Sleep Drives Metabolite Clearance from the Adult Brain.” Science 342 (6156): 373–77. https://doi.org/10.1126/science.1241224.

Zhu, Ting-Ting, Jian-Jun Yang, and Kenji Hashimoto. 2025. “Noradrenergic Modulation of Glymphatic Clearance: Implications for Neuropsychiatric Disorders and Mortality.” Molecular Psychiatry 30 (9): 4432–34. https://doi.org/10.1038/s41380-025-03051-8.

Zuraikat, F M, S Benson, and M H Hall. 2021. “Later Sleep Timing and Greater Sleep Variability Are Associated with Adverse Health Outcomes.” Applied Physiology, Nutrition, and Metabolism 46 (12): 1467–76. https://doi.org/10.1139/apnm-2020-0032.