Is there a cerebral lymphatic system?

The brain is unique among virtually all somatic organs in its lack of a conventional lymphatic vasculature^1–3. In the periphery, the lymphatic circulation facilitates the clearance of extracellular proteins and excess fluid from the interstitium, a role critical to tissue homeostasis and function^{4, 5}. Yet within the brain, despite its complex architecture and high metabolic activity, and neural cells’ sensitivity to changes in the extracellular environment, no specialized organ-wide anatomic structure has yet been identified that facilitates the efficient ‘lymphatic’ clearance of extracellular solutes and fluid from the brain parenchyma.

Current understanding of brain insterstitial solute clearance

For small molecules and hydrophobic compounds, efflux across the blood brain barrier (BBB) is relatively unrestricted. Molecules that are substrates for specific BBB transporters are also readily cleared from the brain^{6, 7}. Other compounds must be cleared from the brain interstitium to the cerebrospinal fluid (CSF) compartment, where they are ultimately eliminated to the blood stream via arachnoid granulations or to peripheral lymphatics along cranial nerves^{1, 8, 9}. However, the distances between much of the brain tissue and the CSF compartments are too great for efficient clearance by simple diffusion, particularly for large molecules (such as peptides and proteins) with low diffusion coefficients⁶. Rather, the clearance of these interstitial solutes from the brain is attributed to bulk flow, by which convective currents of interstitial fluid (ISF) ‘sweep’ solutes along at a high rate that is largely independent of molecular size^{1, 2, 6, 7}.

In a controversial series of studies, Grady and colleagues^{10, 11} suggested that brain ISF may exchange with CSF along ‘paravascular’ routes surrounding cerebral blood vessels. As these findings appeared to be subsequently refuted by Cserr and colleagues^{12, 13}, such ‘retrograde’ movement of CSF into the brain parenchyma is now thought to be of comparatively minor physiological importance¹. However, if a substantial amount of CSF moves through the brain interstitium, and if this flux occurs along defined anatomical pathways, this would fundamentally alter our understanding of how CSF facilitates the clearance of interstitial solutes and metabolic wastes from the brain.

The glymphatic pathway: a paravascular pathway for interstitial solute clearance

In a recent study¹⁴ we define for the first time a brain-wide anatomical pathway that facilitates the exchange of CSF and ISF and the clearance of interstitial solutes from the brain. This pathway consists of 3 elements: a para-arterial CSF influx route, a para-venous ISF clearance route, and a trans-parenchymal pathway that is dependent upon astroglial water transport via the astrocytic aquaporin-4 (AQP4) water channel (represented in schematic form in Figure 1A).

Figure 1. Schematic of glymphatic pathway function in normal and diseased brain.

Conceptual framework for the failure of glymphatic interstitial solute clearance after diffuse injury. (A) In the healthy brain, CSF from the subarachnoid space rapidly enters the brain along paravascular channels surrounding penetrating arteries (green arrow) and exchanges with brain ISF. ISF and solutes are cleared to paravascular spaces surrounding large caliber draining veins (orange arrows). Convective bulk fluid flux between the paravascular CSF influx and ISF efflux pathways is facilitated by astroglial water transport through AQP4 expressed exclusively along perivascular astrocytic endfeet. This convective bulk flow facilitates the clearance of interstitial solutes from the brain. (B) Reactive astrogliosis that occurs after diffuse injury such as microinfarction or mild traumatic brain injury causes in the mis-localization of AQP4 from the perivascular endfeet to the rest of the astrocytic soma. This results in the loss of efficient interstitial bulk flow, and the failure of glymphatic interstitial solute clearance and may contribute to the deposition of extracellular and intracellular protein aggregates (such as amyloid β or tau) after diffuse injury.

Using in vivo 2-photon and ex vivo confocal imaging of small molecular weight fluorescent CSF tracers, we found that a large proportion (>40%¹⁴) of subarachnoid CSF rapidly enters the brain parenchyma along paravascular spaces surrounding penetrating arteries throughout the brain. CSF tracer entered the brain initially through the Virchow-Robin space, then followed arterial vascular smooth muscle basement membrane to reach the basal lamina of the brain capillary bed. At all levels of this paravascular route, CSF tracer entered into the interstitial space, reflecting the exchange of CSF and ISF¹⁴. Para-arterial CSF influx extended throughout the brain, and appeared to occur along virtually all penetrating arteries. ISF clearance pathways, in contrast, were restricted to a specific group of large caliber draining veins. Fluorescent tracer injected directly into the interstitium of the cortex, striatum or thalamus was cleared medially to the internal cerebral veins and great vein of Galen and ventro-laterally to the caudal rhinal vein¹⁴.

The astroglial AQP4 water channel is expressed in a highly polarized manner in perivascular astrocytic endfeet that immediately bound these paravascular CSF influx and ISF clearance pathways (Figure 1A, 2A)^{15, 16}. We proposed that these perivascular water channels may facilitate the convective bulk flow of fluid from the para-arterial CSF influx pathway through the interstitium, and along the para-venous clearance route. To test this, we evaluated paravascular CSF influx in global Aqp4 knockout mice by both in vivo 2-photon and ex vivo fluorescence imaging. Compared to wild type controls, CSF influx into and through the parenchyma of Aqp4-null mice was dramatically reduced¹⁴. Similarly, when we evaluated the rate of interstitial solute clearance from the brain using a radio-tracer clearance assay, we found that interstitial solute clearance was reduced by ~70% in Aqp4-null mice. As detailed in our recent study¹⁴, these findings demonstrate that AQP4-dependent bulk flow couples CSF influx along the para-arterial pathway to ISF clearance along the para-venous route, forming an organ-wide system that facilitates the clearance of interstitial solutes from the brain parenchyma. Based upon this glial dependence and the functional and structural homology to the peripheral lymphatic system, we have termed this glio-vascular pathway the ‘glymphatic’ system (Figure 1A).

Figure 2. Changes in AQP4 localization after diffuse injury.

(A) Immunofluorescent double-labeling demonstrates that in the healthy young mouse brain, AQP4 expression is highly localized to perivascular astrocytic endfeet surrounding to entirety of the cerebral microvasculature. (B) 7 days after mild traumatic brain injury, widespread reactive astrogliosis (GFAP-immunoreactivity) is observed throughout the ipsilateral cortex. In regions of reactive astrogliosis, AQP4 localization is severely perturbed, exhibiting a loss of polarization to the endfoot process and increased somal labeling. Similar expression patters are observed after diffuse microinfarction (manuscript under review).

Soluble amyloid β is (Aβ) present in the ISF of the healthy young brain and the failure of Aβ clearance is thought to underlie the deposition of Aβ plaques associated with Alzheimer’s disease progression^10,11. We next evaluated whether soluble Aβ constitutes one of the solutes cleared from the brain interstitium along the glymphatic pathway. When fluorescently-labeled Aβ was injected into the cortex or striatum, it accumulated around the same paravascular pathways observed with other fluorescent tracers¹⁴. We also measured the clearance of radio-labeled Aβ injected directly into the striatum of wild type and Aqp4-null mice. In Aqp4-null mice, radio-labeled Aβ clearance was reduced by ~65% compared to wild type animals, suggesting that AQP4-dependent bulk flow along the glymphatic pathway constitutes a key mechanism of clearance of soluble Aβ from the brain interstitium¹⁴.

The effect of diffuse gliotic injury on glymphatic pathway function

Reactive astrogliosis is a cellular response to injury common to many mechanistically distinct forms of brain injury, including ischemic and traumatic brain injury, and characterized by changes in astrocyte morphology and molecular expression patterns^17–19. While more severe ischemic and traumatic brain injury is accompanied by glial scar formation, low-intensity injury frequently results in diffuse and long-lasting reactive astrogliosis. This is reflected in two recent studies from our group. In a mouse model of diffuse microinfarction exhibiting only low-level aggregate ischemic burden, widespread reactive gliosis was evident throughout the cortex and striatum for up to a month post-injury (manuscript under review). Similarly, in a mild traumatic brain injury model, widespread cortical and subcortical reactive gliosis was evident for at least one month post-injury in the absence of frank tissue destruction (manuscript under review).

Changes in AQP4 expression are often observed in conjunction with reactive astrogliosis. After ischemic or traumatic brain injury^{20, 21}, AQP4 expression is typically elevated. As these studies employ moderate to severe ischemic and traumatic brain injury, much of this may be attributable to altered AQP4 expression within the glial scar. In our own studies of microinfarction (manuscript under review) and mild traumatic brain injury, changes in AQP4 expression within regions of diffuse reactive gliosis are more complex. General AQP4 expression is elevated in gliotic regions 7 days after diffuse microinfarction, but normalizes by 14 days post-injury. The distribution of AQP4 expression, however, remains perturbed for at least 1 month post-injury. Rather than the highly polarized perivascular localization observed in normal brain, AQP4 in reactive astrocytes exhibits a marked reduction in polarity, with lower perivascular APQ4 immunoreactivity and higher somal AQP4 immunoreactivity. Similar patterns of AQP4 dysregulation are also observed in reactive astrocytes following mild traumatic brain injury (Figure 2B).

In light of the critical role that perivascular AQP4 plays in the glymphatic clearance of interstitial solutes, including soluble Aβ¹⁴, changes in AQP4 localization following diffuse injury may have critical implications for the pathogenesis of conditions such as vascular dementia and traumatic brain injury. We propose that mis-localization of AQP4 from the perivascular endfeet to the astrocytic soma prevents the efficient directional flux of water into and out of the paravascular spaces that contribute to interstitial solute clearance (Figure 1B). This may cause the widespread failure of waste clearance from the diffusely gliotic brain tissue, resulting in the accumulation of neurotoxic metabolites such as Aβ, in addition to the extracellular and intracellular cytotoxic protein aggregates that are the hallmark of neurodegenerative diseases such as Alzheimer’s disease and chronic traumatic encephalopathy. In this way, reactive gliosis, through its detrimental effects on interstitial waste clearance, may be a key driver of pathology under conditions of diffuse ischemic or traumatic brain injury, and may represent a key target for therapeutic intervention.