Glymphatic system

The glymphatic system is a proposed pathway by which the brain clears soluble waste despite having no conventional lymphatic vessels in its tissue. In the model, cerebrospinal fluid (CSF) flows into the brain along the spaces that surround arteries, exchanges with the interstitial fluid between cells, and leaves along routes that drain towards lymph nodes in the neck. The name was coined in 2012 for a mechanism said to depend on water channels in astrocytes, a kind of glial cell. That CSF enters brain tissue along blood vessels is well supported in rodents; whether fluid actually flows through the tissue, and whether sleep speeds it up, remain disputed as of October 2026.

§Origins

The 2012 study that named the pathway imaged small fluorescent tracers in the brains of living mice. It reported that CSF enters brain tissue along spaces surrounding penetrating arteries and that interstitial fluid leaves along spaces around veins. Mice lacking the water channel aquaporin-4 (AQP4) showed slower CSF influx and roughly a 70% reduction in interstitial solute clearance, and they cleared injected soluble amyloid-β less well.¹ The authors proposed the name "in light of its dependence on glial water flux, and its subservience of a lymphatic function", making the pathway a candidate route for removing the proteins that accumulate in Alzheimer's disease.

A year later the same group reported that natural sleep and anaesthesia in mice were associated with a 60% expansion of the interstitial space, greater exchange of CSF with interstitial fluid, and faster clearance of β-amyloid.² It became the popular basis for the claim that sleep washes the brain, discussed under Sleep and longevity.

The pathway still needed an exit. In 2015 two groups independently described lymphatic vessels in the dura mater of mice, the outermost membrane around the brain. The vessels carried fluid from the CSF to the deep cervical lymph nodes, and mice engineered to lack them cleared macromolecules from the brain more slowly.³

1. 2012
   Pathway named
   Iliff, Nedergaard and colleagues describe periarterial CSF influx and AQP4-dependent clearance in mice and propose the name glymphatic.
2. 2013
   Sleep linked to clearance
   Xie and colleagues report an expanded interstitial space and faster amyloid-β clearance during sleep and anaesthesia in mice.
3. 2015
   Meningeal lymphatics described
   Two independent mouse studies find lymphatic vessels in the dura that drain CSF to the deep cervical lymph nodes.
4. 2017
   Diffusion challenge
   Verkman's group re-tests the model in mice and rats and finds solute movement consistent with diffusion and independent of AQP4.
5. 2018
   Tracer imaged in human brains
   An Oslo group follows intrathecally injected MRI contrast through the whole human brain over two days.
6. 2019
   CSF waves in human sleep
   Fast fMRI shows pulses of CSF inflow coupled to slow waves during non-REM sleep.
7. 2024
   Opposite sleep result
   Franks and Wisden's group reports that brain clearance in mice is reduced, not increased, in sleep and anaesthesia.
8. 2025
   Published exchange
   Nedergaard's group argues the 2024 study measured redistribution rather than clearance; Franks and Wisden reject every objection.
9. 2026
   Radio-frequency stimulation in mice
   A Beijing group reports that 40.68 MHz ultrashort-wave exposure speeds tracer clearance in mice.

§How it is supposed to work

In the model, CSF in the space around the brain is driven into perivascular spaces, the fluid-filled sleeves around penetrating arteries. Astrocytes wrap these vessels with their endfeet, on which AQP4 is concentrated. That concentration is called polarization, and the hypothesis holds that it eases fluid movement from the perivascular space into the tissue. Convective flow is then supposed to sweep solutes through the interstitium towards spaces around veins, from where they reach the meningeal lymphatics and the neck. Mouse work from the same laboratory reported in 2014 that this exchange declines with age as AQP4 polarization is lost, making the pathway a candidate contributor to brain aging alongside failing Proteostasis collapse and other Hallmarks of aging.

What drives the flow is unsettled. Arterial pulsation was an early proposal. Later mouse studies have pointed to slow vascular oscillations and to synchronized neuronal activity, including slow rhythms imposed with Optogenetics, and that last idea is the premise of current attempts to boost clearance with Non-invasive neuromodulation.

§Measuring it in people

Human evidence rests on three indirect methods. The most direct is intrathecal-contrast MRI. Researchers in Oslo injected the contrast agent gadobutrol into the CSF and scanned repeatedly over two days and again at four weeks; the tracer reached every brain region and cleared more slowly in patients with idiopathic normal pressure hydrocephalus, a condition that causes dementia. In a follow-up, one night of total sleep deprivation in seven people slowed tracer clearance from most brain regions compared with seventeen who slept, and a further night of sleep did not make up the difference.⁴ The method shows where a tracer goes, not by what physical process.

Fast functional MRI is the second. A 2019 study of thirteen sleeping volunteers found that during non-REM sleep neural slow waves were followed by oscillations in blood volume, which were coupled to pulses of CSF inflow at the fourth ventricle.⁵ It measured CSF movement, not waste removal.

The third, DTI-ALPS, is an index computed from routine diffusion MRI along the direction of perivascular spaces in deep white matter. It was proposed in 2017 and is attractive as a glymphatic marker because it needs no injection. A 2026 comparison in 56 people who also had intrathecal-contrast MRI found little tracer in the regions the index samples and only a limited association between the two measures; the authors called ALPS "not well-validated".⁶ That limits any use of the index as one of the Aging biomarkers of brain clearance.

§The dispute

Three parts of the model are contested. The first is whether fluid flows through brain tissue at all. A 2017 re-test in mice and rats found that the spread of fluorescent dextrans depended on their size as diffusion predicts, did not change just after cardiac arrest, and was not impaired by deleting the AQP4 gene. A 2022 critical review concluded that solutes move both in and out along periarterial routes, that there is no evidence the main outflow is perivenous, and that any flow through the interstitium is too slow compared with diffusion to matter.⁷ The second is AQP4 dependence, where knockout studies from different laboratories disagree.

The third is sleep. A 2024 study from Nicholas Franks and William Wisden injected dye directly into brain tissue of male mice and reported that its movement was independent of sleep, waking and anaesthesia, while clearance was markedly reduced during sleep and anaesthesia.⁸ In 2025, Nedergaard's group argued that the study measured movement of dye within the brain rather than its removal, that the dye crosses the blood–brain barrier, and that the anaesthetic was given once and mostly wore off. Franks and Wisden replied that all of the objections were "misunderstandings, misleading, or simply incorrect", and countered that most studies reporting faster clearance during sleep actually measure tracer entering the brain from the CSF, which they called an unproven stand-in for removal.

Flushed, or diffused?

Proponents point to periarterial tracer influx, which even critical reviewers accept, the AQP4-knockout results, the sleep effect in mice, and human MRI in which an intrathecal tracer reaches the whole brain. Critics reply that solute spread in rodent tissue fits diffusion, that AQP4 dependence has not replicated across laboratories, that a study measuring removal from tissue found the opposite sleep effect, and that influx is not clearance. The two camps use different tracers, injection sites, anaesthetics and definitions, so most experiments cited by one side do not test the claim the other side makes.

§Alzheimer's disease, aging and brain injury

The disease case rests almost entirely on mice. A 2024 study in 5XFAD mice, a strain engineered to accumulate amyloid, found that one-hour sessions of 40 Hz flickering light and sound increased CSF influx and interstitial efflux in cortex, together with more AQP4 polarization and wider meningeal lymphatic vessels. Inhibiting glymphatic clearance abolished the stimulation's reduction of amyloid.⁹

In October 2026 a Beijing group reported a similar result with 40.68 MHz ultrashort-wave radio-frequency energy, a physiotherapy modality, applied at a non-thermal setting through electrodes held 2 cm from the skull.¹⁰ In healthy young male mice under anaesthesia for the tracer measurements, a single 10-minute session increased tracer arrival in cervical lymph nodes at 60 minutes; by 120 minutes the groups no longer differed. In awake mice, exposure raised delta-band power in hippocampal and EEG recordings. The tracer effect was blunted in a chemogenetically silenced hippocampus and abolished when AQP4 was blocked with a drug. After closed-skull brain injury, treated mice had less brain water at one hour and three days. In six-month-old female 5XFAD mice, seven daily sessions improved tracer clearance, lowered amyloid-β immunofluorescence (7–8 mice per group) and improved two memory tasks. The authors note that the whole experiment was run as one cohort without independent replication.

What the ultrashort-wave study did not show

Compressed into a headline, the study becomes a way to flush Alzheimer's plaque from the brain in minutes. It does not show that. The 10-minute effect was faster tracer drainage in healthy mice. The amyloid reduction followed a week of daily sessions in a mouse model and was a fall in staining intensity, not removal; tau was not measured. AQP4 was redistributed towards blood vessels and more abundant, and channel opening was not measured. No person was exposed.

The human side is thinner. Slower tracer clearance in normal pressure hydrocephalus and after one sleepless night are associations in small clinical groups. No human study has shown that improving glymphatic function slows Alzheimer's disease, and the idea rests on the plausibility of the mouse work rather than on a trial in people.

§Outlook

The next tests are methodological. If the field agrees on what counts as clearance, meaning removal from the brain rather than entry into it, the sleep dispute becomes testable, and the claims made for stimulation devices become checkable. In humans, intrathecal MRI requires injecting contrast into the CSF and DTI-ALPS is poorly validated, so a non-invasive measure with a known relationship to tracer removal would change what can be asked in people.

Interest in clearance also bears on Targeted drug delivery, because a drug given into the CSF has to reach brain tissue by the same routes, and on arguments over Engineered sleep reduction, which assume that sleep's functions can be replaced. It also fits the geroscience case that brain aging and dementia share causes, and the Compression of morbidity and Healthspan arguments, though none yet has human glymphatic data behind it.

The strongest objection to the whole framework remains the 2017 one. If solutes in brain tissue move mainly by diffusion, then the interventions now being designed to speed glymphatic flow in mice are optimizing a process that may be smaller than the model assumes, and their benefits in mouse models would need another explanation.

§See also

• Alzheimer's disease
• Sleep and longevity
• Non-invasive neuromodulation
• Proteostasis collapse
• Aging biomarkers
• Engineered sleep reduction
• Hallmarks of aging

§References