The lymphatic network extends throughout the body, filtering cellular waste out of circulating fluids everywhere—except in the brain. Curiously, this energy hog, which falters at the slightest change to its physiology, has no lymph vessels. Instead, scientists have believed for years that the cerebrospinal fluid (CSF) acts as the brain’s lymphatic system. Research published in the August 15 Science Translational Medicine adds substance to this old concept. Using in-vivo imaging to track solutes injected into the brains of mice, “we showed that the CSF goes into the brain just like people thought, but it does so along very specific anatomical structures,” said Jeffrey Iliff, first author on the study led by senior investigator Maiken Nedergaard of the University of Rochester Medical Center in New York. “There’s a specialized anatomy that allows the CSF to move very quickly and very deeply into the brain, exchanging with fluid that’s inside.” The pathway depends on a specific water channel found on astrocytes and seems important for clearing brain amyloid-β in mice, the scientists report. The glymphatic pathway, as the authors named it, since it relies on glia, clears Aβ, suggesting it may help prevent amyloid plaques from accumulating in the brain.

Earlier work in anesthetized cats showed CSF tracers flowing through the brain alongside arteries (Rennels et al., 1990; Rennels et al., 1985), suggesting that the CSF acts in lymph-like fashion. Yet other studies called that idea into question; the CSF appeared to snake through the brain slowly, with unpredictable direction (see Ichimura et al., 1991). In the current study, the Rochester team revisited this issue with state-of-the-art imaging technology.

First, Iliff and colleagues did an experiment the old-fashioned way—injecting fluorescent tracers into the CSF of anesthetized mice, then killing the animals at various time points and fixing their brain tissue to see where the tracers had gone. Fluorescence microscopy of the brain slices confirmed that the tracers were moving along the outsides of blood vessels. When the experiments were repeated in transgenic mice with fluorescently labeled arteries, the results were “pretty unequivocal that [tracer movement] followed arteries only,” Iliff told Alzforum.

To get a clearer picture of where solutes go in real time, the scientists then switched to two-photon imaging, which also enabled them to visualize the brain at greater depths than conventional microscopy. Two-photon microscopy can penetrate down to 400 to 500 microns—not so deep for the human brain but good enough to see most of the way through the cerebral cortex in mice.

Iliff and colleagues injected fluorescent green tracers, ranging in size from 3 to 2,000 kD, into the CSF of anesthetized mice, and watched their movement through cranial windows embedded into the skull. The tracers moved down into the cerebral cortex along arteries, then collected in deep veins where they got drained out of the brain (see video below). “As the fluid courses in along arteries through the tissue and out along veins in a continuous flow, it sweeps along particles that are sitting in between the cells,” Iliff said.

Capturing one frame per minute, two-photon microscopy shows a fluorescent green tracer moving into the brain along an artery (shown in red on left side) and picking up particles (black dots) en route to a vein (shown in blue on right side) for eventual clearance out of the brain. Images were taken at the brain’s surface (top panel), and at 60 and 120 microns below the surface (middle and bottom panels, respectively). [Courtesy of Jeffrey Iliff, University of Rochester Medical Center.]

Those experiments defined the clearance pathway anatomically. But what pushes solutes and fluids through the system? Iliff and colleagues had a hunch that astrocytes played a role. Portions of these glial cells wrap around brain arteries, and those appendages carry a water channel called aquaporin 4 (Aqp4). To test whether the astrocytic water channels were responsible for driving fluid around the arteries and through the brain, the researchers injected mannitol and dextran molecules into the CSF of Aqp4 knockout mice. By two-photon microscopy and slice imaging, the solutes moved through the brain more slowly and got cleared 65 to 80 percent less efficiently than in wild-type mice. (See a brief explanation by Iliff in this three-minute video.)


In this three-minute video, first author Jeffrey Iliff describes how his team used two-photon imaging to delineate a new clearance pathway in the brain. Image courtesy of Jeffrey Iliff, University of Rochester Medical Center

The findings could have ramifications for neurodegenerative diseases that cause abnormal buildup of proteins in the brain. To test if the glymph system clears the brain of Aβ, the team injected fluorescent-tagged Aβ1-40 into the mouse striatum and saw it collecting and aggregating “in those same deep veins that we saw with other tracers,” Iliff said. Aqp4-null mice cleared 125I-labeled Aβ from the brain about 70 percent more slowly than did wild-type animals.

“We think this is how things are cleared if nothing else can clear them, like a pathway of last resort,” Iliff told Alzforum. This glymphatic pathway influenced the movement of 759-dalton and 3-kD tracers, but did not affect larger molecules around 2,000 kD. Size is not all that matters. The glymph cleared Aβ (~5 kD) faster than a similar-sized dextran molecule, suggesting that other factors besides size influence clearance rate. Iliff said the most likely candidates for clearance along this path would be hydrophilic molecules that lack specific transporters and cannot leave the brain by rapid diffusion.

Other molecules may be cleared by other mechanisms. “The relative contributions of the glymphatic system, arachnoid villi transport, and active blood-brain barrier transport should be determined, though will likely need to be done for individual molecules,” noted John Cirrito of Washington University School of Medicine in St. Louis, Missouri (see full comment below).

Clearing Aβ is a major therapeutic strategy, and many groups are testing immunotherapy approaches, which are based on large immunoglobulins. It is hard to predict how the glymph system might affect clearance by these therapies. “Antibodies that activate microglia and cause phagocytosis of plaque would not need bulk flow clearance. However, antibodies that target soluble Aβ, like Lilly’s solanezumab, may use this pathway more,” Cirrito told Alzforum via e-mail. “That said, the glymphatic system discriminates by size, and an antibody bound to Aβ would be 40-fold larger than Aβ alone, so its clearance through the pathway may be hindered.”

Other scientists found the study interesting because it sheds light on a new cellular player in the brain’s perivascular clearance pathways. “It’s not only vascular smooth muscle cells in small penetrating cerebral arteries that regulate these systems, but astrocytes, too,” said Berislav Zlokovic of the University of Southern California, Los Angeles.

To examine the role of astrocytes in clearing brain Aβ, Iliff and colleagues are crossing AD transgenic lines with mice deficient in Aqp4. They also plan to mechanically disrupt the CSF clearance pathway in 3xTg AD and APPswe/PSEN1dE9 mice to see if that speeds amyloid deposition, or enhances CSF flux and see if that slows plaque buildup.—Esther Landhuis




  1. This study importantly adds to our knowledge that passive clearance pathways from the brain interstitial and cerebrospinal fluid could potentially play an important role in clearing toxic molecules from the brain. This is in addition to previously demonstrated vascular clearance across the blood-brain barrier, which typically requires the presence of specific transporters at the luminal side of blood vessels to eliminate toxins away from brain. Vascular clearance is more rapid than the passive clearance pathway. In addition to previously demonstrated roles of vascular smooth muscle cells in small penetrating cerebral arteries, the study sheds a new light on possible role of astrocytes in the regulation of the passive clearance route, which further cements the concept that cells of the neurovascular unit may critically determine the levels of different potential neurotoxins in brain.

    View all comments by Berislav Zlokovic
  2. Iliff and colleagues use a clever approach to describe a brand-new bulk flow clearance pathway for interstitial fluid (ISF) and cerebrospinal fluid (CSF) water as well as molecules; fluids drain through paravenous spaces within parenchyma to be eliminated. Interestingly, this is specific to paravenous spaces, not other blood vessels, and is particularly reliant on spaces regulated by aquaporin 4. Aquaporin 4-null mice have reduced perivascular space, which presumably would restrict water flow, and could be responsible for the reduced clearance of molecules. If this system is active in other locales, then it is possible that other flow-regulating molecules besides Aqp4 could be active there instead.

    This “glymphatic system” affects small molecules more so than larger molecules, likely due to physical restrictions within the pathway. Determining what physical obstructions are responsible for this differential size preference will be interesting. But more than size alone determines the rate of clearance, since Aβ is cleared faster than a similarly sized dextran molecule. The relative contribution of the glymphatic system, arachnoid villi transport, and active blood-brain barrier transport should also be determined, though will likely need to be done for individual molecules.

    We generally think about bulk flow clearance pathways as an amorphous entity that clears CSF molecules through arachnoid villi, or that clears ISF into the CSF. In contrast, the glymphatic system is a new bulk flow pathway that can be linked to specific molecules and regulation. This system also takes into account physical space, which is something we don’t usually think about in terms of clearance pathways. It will be interesting to see how other molecules are handled by this clearance pathway, as well as what changes due to disease could alter it.

    View all comments by John Cirrito
  3. Nice work, that now adds “moving” data to the description by Roy Weller in the 1980s, who found and denoted the “perivascular drainage channels” (PVDC). These PVDCs were long forgotten by the time the first “AD genes” were discovered. The puzzle pieces now start to fall into place giving a clearer picture of Aβ production and removal from the brain—first flush and then transport. Different groups have shown how to get rid of Aβ via the blood-brain barrier using ABC transporters, till now it was quite uncertain how Aβ makes it to the vessel walls. Also, the choroid plexus gets into the focus since it is tightly connected to the PVDCs via the CSF spaces.

  4. Obviously, aquaporin 4 is important to this method of clearance, but it may not be the only aquaporin that is important to amyloid clearance (or proper amyloid function).

    For example, Alan Basbaum's group at the University of California, San Francisco, has investigated olfactory ensheathing glia (OEG). They found that these cells express the water channel aquaporin 1 (AQP1) and propose "that AQP1 expression represents an important distinguishing characteristic of OEG."

    There may be additional aquaporins which are important to other specialized functions.


    . Olfactory ensheathing glia express aquaporin 1. J Comp Neurol. 2010 Nov 1;518(21):4329-41. PubMed.

  5. This study illustrates that considerable progress can be made in Alzheimer's research if we obtain better insight into the normal physiology of Aβ.

    Researchers and clinicians now use abnormal concentrations of Aβ in CSF to diagnose AD, and even use changes in Aβ concentrations as outcome parameters in pharmaceutical studies. Yet we know surprisingly little of how Aβ is transported to CSF following its "birth" in neurons. As a consequence, we are unable to accurately predict the effect of "failure" of one or more systems (e.g., clearance) on CSF Aβ levels. We have recently attempted to provide an overview of these pathways, and of the factors that may influence it, from the viewpoint of trying to understand why CSF Aβ is reduced in AD (Spies et al., 2012).

    A better understanding of the mechanisms involved may help us understand Alzheimer's, and is essential to interpret the effects of anti-amyloid therapies on CSF Aβ.


    . Reviewing reasons for the decreased CSF Abeta42 concentration in Alzheimer. Front Biosci. 2012;17:2024-34. PubMed.

  6. We are intrigued by this important technological advance in helping to understand the fate of solutes within the CNS. We previously showed that patients in the first AN1792 Aβ immunotherapy trial (Elan Pharmaceuticals) developed increased severity of cerebral amyloid angiopathy (CAA) as plaques were cleared from the brain (Boche et al. 2008)—a change that was also observed in animal models. We interpreted this finding as reflecting Aβ, solubilized from plaques by antibody, tracking to the arteries along the previously defined perivascular drainage pathway (Weller et al., 2009), which involves the basement membranes of the artery wall rather than the perivascular/paravascular space. This previously defined system seems, at least at this stage, to better explain the changes in CAA induced by immunotherapy, as in CAA, the Aβ is deposited within rather than around the artery wall. However, there may clearly be multiple pathways for exit of solutes from the brain that are important in this context. The understanding of these drainage systems is particularly topical at the moment, as overload of these systems in an aged brain seems likely, at least in our view (Boche et al., 2010), to be the cause of the amyloid-related imaging abnormalities (ARIA), which are hampering current immunotherapy trials for Alzheimer’s disease (Sperling et al., 2012).


    . Consequence of Abeta immunization on the vasculature of human Alzheimer's disease brain. Brain. 2008 Dec;131(Pt 12):3299-310. PubMed.

    . Lymphatic drainage of the brain and the pathophysiology of neurological disease. Acta Neuropathol. 2009 Jan;117(1):1-14. PubMed.

    . Neuropathology after active Abeta42 immunotherapy: implications for Alzheimer's disease pathogenesis. Acta Neuropathol. 2010 Sep;120(3):369-84. PubMed.

    . Amyloid-related imaging abnormalities in patients with Alzheimer's disease treated with bapineuzumab: a retrospective analysis. Lancet Neurol. 2012 Mar;11(3):241-9. PubMed.

  7. Iliff et al. have presented a very interesting study. They observed that tracers injected into the CSF via the cisterna magna of mice extended into the brain along paravascular pathways around arteries but not around veins. The lower-molecular-weight tracers extended into the brain parenchyma and into paravenous compartments, and then into the CSF. A similar distribution of tracer was observed following intraparenchymal injections into the cerebral cortex and deep grey matter of the mouse brain. The authors suggest that homoeostasis of interstitial fluid is maintained by the flow of CSF through the brain and, from their experiments using aquaporin 4 knockout mice, they conclude that astrocytes may be involved in this pathway. This work confirms and extends the work of Rennels (Rennels et al., 1985) in a rather elegant way.

    As Iliff et al. emphasized, maintenance of the external environment for neurons and other cells is an important factor in maintaining normal function in the CNS. The absence of traditional lymphatics in the brain has led to the suggestions that solutes in the interstitial fluid are either cleared directly into the blood, into the CSF, or by perivascular lymphatic drainage along basement membranes in the walls of capillaries and arteries of the cerebral circulation into regional lymph nodes in the neck. Evidence that vascular basement membranes are a route for drainage of solutes is derived from intraparenchymal injections of tracers (Carare et al., 2008; Hawkes et al., 2011) and from the observation that amyloid-β accumulates in the basement membrane pathways in capillary and artery walls as cerebral amyloid angiopathy (CAA) (Herzig et al., 2006; Weller et al., 2009). The amyloid-β that accumulates in capillary and artery walls in transgenic APP mice is derived from neurons (Herzig et al., 2006) rather than from extraneural sources. This suggests that amyloid in CAA is deposited during perivascular drainage from the brain along basement membranes within the walls of capillaries and arteries. Although Iliff et al. are skeptical about the existence of basement membrane drainage pathways within capillary and artery walls, the interstitial fluid drainage system described by these authors does not explain the distribution of amyloid-β in the walls of cortical and leptomeningeal arteries in CAA.

    As new technologies, including two-photon imaging, are introduced further to the study of fluid and solute clearance from the CNS, the discrepancies identified above will hopefully be resolved. Whatever the final solution proves to be, facilitating elimination of amyloid-β from the aging brain will remain an important therapeutic target for Alzheimer’s disease.


    . Solutes, but not cells, drain from the brain parenchyma along basement membranes of capillaries and arteries: significance for cerebral amyloid angiopathy and neuroimmunology. Neuropathol Appl Neurobiol. 2008 Apr;34(2):131-44. Epub 2008 Jan 16 PubMed.

    . Mechanism of cerebral beta-amyloid angiopathy: murine and cellular models. Brain Pathol. 2006 Jan;16(1):40-54. PubMed.

    . A paravascular pathway facilitates CSF flow through the brain parenchyma and the clearance of interstitial solutes, including amyloid β. Sci Transl Med. 2012 Aug 15;4(147):147ra111. PubMed.

    . Evidence for a 'paravascular' fluid circulation in the mammalian central nervous system, provided by the rapid distribution of tracer protein throughout the brain from the subarachnoid space. Brain Res. 1985 Feb 4;326(1):47-63. PubMed.

    . Cerebral amyloid angiopathy in the aetiology and immunotherapy of Alzheimer disease. Alzheimers Res Ther. 2009;1(2):6. PubMed.

    View all comments by Roy Weller
  8. If this "glymphatic" flow indeed drains exclusively to veins, it seems unlikely to account for a large amount of Aβ clearance (and even less likely to account for pathogenic deficiencies in that clearance), because cerebral amyloid angiopathy (CAA) is found predominantly at arteries and arterioles in both humans and animal models of AD. Could this be additional evidence that active transport via receptors of the low-density lipoprotein receptor (LDLR) family is important for Aβ clearance? Yes, but there is at least one other possibility: that CAA represents Aβ arriving in the brain from the vasculature. It's quite clear that some fractional amount of plasma-borne Aβ does cross into the brain, and the occasional occurrence generates great intrigue (e.g., Sutcliffe et al., 2011). Could it be that the Aβ entering via this route is being swept passively in an artery-to-vein flow until sequestered by the tunica media and/or adventitia?


    . Peripheral reduction of β-amyloid is sufficient to reduce brain β-amyloid: implications for Alzheimer's disease. J Neurosci Res. 2011 Jun;89(6):808-14. PubMed.

Make a Comment

To make a comment you must login or register.


Paper Citations

  1. . Rapid solute transport throughout the brain via paravascular fluid pathways. Adv Neurol. 1990;52:431-9. PubMed.
  2. . Evidence for a 'paravascular' fluid circulation in the mammalian central nervous system, provided by the rapid distribution of tracer protein throughout the brain from the subarachnoid space. Brain Res. 1985 Feb 4;326(1):47-63. PubMed.
  3. . Distribution of extracellular tracers in perivascular spaces of the rat brain. Brain Res. 1991 Apr 5;545(1-2):103-13. PubMed.

Other Citations

  1. 3xTg AD

External Citations

  1. APPswe/PSEN1dE9

Further Reading

No Available Further Reading

Primary Papers

  1. . A paravascular pathway facilitates CSF flow through the brain parenchyma and the clearance of interstitial solutes, including amyloid β. Sci Transl Med. 2012 Aug 15;4(147):147ra111. PubMed.