The glymphatic system and its role in cerebral homeostasis

Abstract

The brain’s high bioenergetic state is paralleled by high metabolic waste production. Authentic lymphatic vasculature is lacking in brain parenchyma. Cerebrospinal fluid (CSF) flow has long been thought to facilitate central nervous system detoxification in place of lymphatics, but the exact processes involved in toxic waste clearance from the brain remain incompletely understood. Over the past 8 yr, novel data in animals and humans have begun to shed new light on these processes in the form of the “glymphatic system,” a brain-wide perivascular transit passageway dedicated to CSF transport and interstitial fluid exchange that facilitates metabolic waste drainage from the brain. Here we will discuss glymphatic system anatomy and methods to visualize and quantify glymphatic system (GS) transport in the brain and also discuss physiological drivers of its function in normal brain and in neurodegeneration.

INTRODUCTION

The brain’s high-energy demand is paralleled by high metabolic waste production. In most body organs, the lymphatic vasculature is responsible for metabolic waste drainage and fluid homeostasis. While the meninges covering the brain and spinal cord are equipped with lymphatics (3, 4, 69), the brain parenchyma itself is devoid of lymphatic vessels. The tight blood-brain barrier (BBB) restricts solute and large fluid shifts and alternate waste elimination systems are operational in brain tissue. Cerebrospinal fluid (CSF) produced in the choroid plexuses of the cerebral ventricles, in addition to other roles, is thought to play an important role for detoxification of brain tissue in place of lymphatics (19, 20, 58, 89). The “glymphatic system” concept was brought to prominence in 2012 (47) shedding new light on CSF transport and brain waste drainage processes. The glymphatic system (GS) is described as a perivascular transit passageway for CSF and interstitial fluid (ISF) exchange that facilitates metabolic waste drainage from the brain parenchyma in a manner dependent on aquaporin 4 (AQP4) water channels on glial cell (47). Several excellent reviews of the GS are available, and we refer readers to these for more details (1, 9, 46, 77, 99, 103). Here we will focus on 1) GS anatomy, 2) methods to visualize and quantify GS transport in the brain, and 3) discuss physiological drivers of GS function in normal brain and in the setting of neurodegeneration.

COMPOSITION OF THE GLYMPHATIC SYSTEM

The GS is located beyond the BBB and comprises the entire perivascular space (PVS) within the brain parenchyma (47). The PVS is constructed as a coaxial system where the inner cylinder is the BBB-tight vessel (e.g., artery, arteriole, capillary, venule, or vein) and the outer cylinder is made of astrocytic end-feet processes that envelop the entire cerebral vasculature. The outer perimeter of the PVS is not “tight” due to gaps (20–30 nm) between the astrocytic end-feet processes (74). The cortical penetrating arterioles are surrounded in part by a layer of pia mater, and at this level the PVS is a fluid-filled space referred to as the Virchow-Robbin space, from where it eventually merges into the basal lamina at the level of the capillary. The basal lamina is located in between the vessel wall and the astrocytic end feet. Under normal conditions, the capillary cell types do not make direct contact with the PVS and are always separated by the basal lamina (82, 89). In humans, the PVS can be detected in the brain parenchyma by magnetic resonance imaging (MRI) as tube-like structures that run perpendicular to the brain’s surface in directions that are spatially correlated with perforating vessels thought to be primarily arterial (51, 106). Cortical PVS can be observed in young, healthy brain (Fig. 1) but is more common in the aging brain, and abnormally dilated PVS is associated with cerebral small vessel disease (cSVD) and other neurological disease states (26, 106, 107). In the human and rodent brain, the PVS communicates with the subarachnoid space as evidenced by multiple studies showing that tracer uptake is visible in PVS following in vivo administration of tracers into CSF (vide infra).

Fig. 1.

T2-weighted MRIs acquired on a 1.5T MRI instrument (GE Signa HDx 1.5using a T2 fast-spin-echo pulse sequence (4-mm thick slices) from the brain of a healthy 35-yr-old female. Perivascular spaces are clearly visible (yellow arrows) in a typical pattern. Data courtesy of Joanna Wardlaw.

Rapid transport of tracers from CSF into the parenchymal perivascular network under carefully controlled physiological conditions was documented in early work by Rennels and coworkers (89) in cat brain by administering horseradish peroxidase into CSF. Two decades later, CSF and solute transport along the PVS of pial arteries and cortical penetrating arterioles of live mice was visualized in real time using two-photon microscopy by administering fluorescently tagged dyes into the cisterna magna (47). These pioneering in vivo studies revealed that small molecular weight (MW) solutes moved rapidly (5–10 min) into the periarterial space (but not perivenous space) and from there into the ISF space (47). Furthermore, waste solutes including soluble Aβ_1–40 injected into brain parenchyma were shown to migrate from the ISF into the PVS of the large central veins inferring that perivenous conduits served as exit pathways connecting to lymphatic networks outside the brain. Importantly, it was also documented that the astrocytic AQP4 water channels were important for rapid periarterial influx of CSF and solutes as well as for drainage of soluble Aβ (47). The importance of the AQP4 water channels for rapid CSF-ISF exchange has been contested (1, 76, 100), and alternate physiological factors (e.g., ISF volume changes and vascular pulsatility) may be more important for time-efficient GS transport and waste drainage (vide infra). It is also not known exactly how the AQP4 water channels regulate GS function. Recently, a novel study using multiple echo time arterial-spin-labeling MRI demonstrated slower than normal water exchange times in the brain (suggesting slow water transport across the PVS into parenchyma) in transgenic mice models lacking AQP4 water channels compared with mice with normal AQP4 water channels (80). Figure 2 is an illustration of the principal anatomical components of the GS and highlights that in normal brain solutes in the ISF drain toward the perivenous space.

Fig. 2.

Illustration of the glymphatic system of the brain. In principle, the glymphatic system comprise a periarterial influx pathway and a perivenous pathway for cerebrospinal fluid (CSF) transit that are coupled to the interstitial fluid space via the aquaporin 4 (APP4) water channels. The AQP4 water channels are positioned on the glial end feet that make up the outer perimeter of the perivascular space; the inner perimeter is the vascular basement membrane. CSF flows into the periarterial space and mixes with ISF whereby waste solutes (black particles) are propelled toward the perivenous conduits for ultimate drainage out of the brain.

WHOLE BRAIN GS FUNCTION

To visualize and quantify GS function in the whole rodent brain, we administered paramagnetic gadolinium-tagged contrast molecules into CSF of the rat via the cisterna magna in combination with dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) (45). The paramagnetic contrast agents shorten the T1 relaxation time thereby eliciting signal changes on the T1-weighted MRIs enabling tracking of solute transport in CSF and brain parenchyma (45). Using this approach, we demonstrated that small MW paramagnetic contrast molecules moved rapidly in the subarachnoid space, along pial arteries and more slowly transited into brain parenchyma in a specific anatomical pattern (45). We noted that brain regions with the most rapid CSF and solute transport included the brainstem, hypothalamus, olfactory bulb, frontal cortex, cerebellum, and the ventral hippocampus (45). We have since refined the MRI-based GS transport technique and determined than only a relatively small (19–20%) fraction of the contrast agent administered into the CSF enters into the rat brain parenchyma over 2.5 h (66). In addition, we showed that MR contrast influx and clearance from brain parenchyma is dependent on body position (67) and the anesthetic used (8).

The GS transport pattern in the rat brain using DCE-MRI is very similar to that observed using the same method in nonhuman primate brain (37) and humans (28, 29, 92). In humans, intrathecal lumbar administration of MR contrast (Gadobutrol, MW: 605 Da) has been performed to diagnose dural tears in otherwise normal subjects (91). In these clinical DCE-MRI studies, MR contrast enhancement in brain parenchyma was observed in a pattern similar to the rodent brain with largest uptake in areas adjacent to large arteries including the anterior, middle, and posterior cerebral arteries (91). In the human brain, regions with most significant uptake after administration of contrast into the lumbar intrathecal space (6–9 h) included the brainstem, cerebellum, frontal cortex, and limbic regions (hippocampus, amygdala, accumbens, and entorhinal cortex) (91).

QUANTIFICATION OF GS FUNCTION

As described above, GS transport can be observed in the entire brain using the DCE-MRI approach (8, 45, 66, 67). However, supplementary analysis is required to quantify transport and to extract differences in GS transport flow across brain regions and across experimental groups. Current techniques for quantifying GS transport include assessment of time-signal-curves of brain parenchymal solute uptake or clearance (8, 28, 45), kinetic analysis (67), or k-means cluster analysis (45, 50). These analytical strategies have provided valuable information but are limited because solute transport across brain regions is heterogeneous causing generalized kinetic models to fail.

VISUALIZATION OF CSF TRANSPORT INTO BRAIN USING OPTIMAL MASS TRANSPORT

We were the first to model GS transport based on DCE-MRI images using the traditional optimal mass transport (OMT) formulation (86, 87). The theory of OMT seeks the most feasible way to redistribute mass from one given distribution to another while minimizing the associated cost of transportation (85). In the first approach, we made several assumptions including the notion that glymphatic CSF-solute transport was governed principally by advection (86) as originally proposed (47). These initial results revealed aberrant CSF “streaming” patterns of contrast solutes into brain parenchyma (86). Transport by pure advection has been a subject of controversy surrounding GS transport and several studies have suggested diffusion dominant solute transport in neuropil (82, 100). We further improved the OMT-based computational analysis with the ultimate goal of visualizing how PVS pathology might alter GS transport and waste drainage. We introduced a novel visualization framework, “GlymphVIS” (30) using a more physiologically relevant model inspired by the work of Benamou and Brenier (7). Specifically, in the GlymphVIS model we added a diffusion term in the standard continuity equation to better model both advection and diffusion thereby more accurately modeling the CSF-solute transport in the brain parenchyma [for more detail, see Elkin et al. (30)]. Figure 3 shows the effect of increasing the diffusion term in the optimal transport algorithm. With minimal or no diffusion term, the OMT presentation of CSF parenchymal streamlines is not aligning with physiological evidence of MR contrast uptake in live rodent brain (Fig. 3A, arrows); however, with more diffusion weighting (Fig. 3B), the aberrant parenchymal CSF pathways have disappeared and the uptake pattern better match what is observed on the MRI data strongly suggesting that parenchymal GS transport is governed by both advection and diffusion. We have further validated these data using phantoms (61). Figure 4A shows the conventional visualization of glymphatic transport in whole rat brain based on DCE-MRIs and “% signal increase from baseline” 1.5 h after administration of MR contrast into the CSF via cisterna magna. The color-coded map shows the spatial distribution of CSF tagged with MR contrast demonstrating that CSF and the contrast solute have penetrated into the cerebellum, midbrain, olfactory bulb, and along the PVS of the middle cerebral artery (MCA) as highlighted in Fig. 4C. Figure 4B shows the same data set processed by the GlymphVis algorithm with advection and diffusion terms deriving CSF streamlines created by proximity and similar curvature using the QuickBundles algorithm (35). These streamlines show brain parenchymal CSF flow patterns at a fixed point in time. Please note that the CSF streamlines along the MCA (Fig. 4D) are a matching contrast uptake in the original data (Fig. 4C). We are currently extending the GlymphVIS analysis to include visualization of CSF pathlines. These pathlines represent the time-varying CSF trajectories and can be used to determine particle attributes including solute speed and flux in one comprehensive figure. Moreover, we are exploring and comparing the advantages (and disadvantages) of both Eulerian and Lagrangian coordinates in visualizing the flow (61).

Fig. 3.

Glymphatic transport visualized by optimal mass transport (OMT) analysis based on dynamic contrast enhanced MRIs obtained from a live rat after MR contrast administration into the cerebrospinal fluid (CSF). The OMT-based analysis derives “CSF transport pathlines,” which are shown as a color-coded map overlaid on the corresponding volume rendered anatomical MRI. We are showing the effect of increasing the diffusion term in the optimal transport algorithm. Specifically, with a minimal or absent diffusion term in the OMT analysis, the pattern of CSF parenchymal streamlines does not align well with physiological evidence of MR contrast uptake in live rodent brain (A, arrows on nonexisting CSF pathlines). However, with more diffusion “weighting” (B), the aberrant parenchymal CSF pathways have disappeared and the uptake pattern better matched what is observed on the MRI data, strongly suggesting that parenchymal glymphatic system transport is governed by both advection and diffusion. Scale bars = 3 mm.

Fig. 4.

A: the conventional visualization of glymphatic transport in whole rat brain based on dynamic contrast-enhanced MRIs expressed as “% signal increase from baseline” 1.5 h after administration of MR contrast into the cerebrospinal fluid (CSF) via cisterna magna. The color-coded map shows the spatial distribution of CSF tagged with MR contrast demonstrating that CSF and the contrast solute have penetrated into the cerebellum, midbrain, and olfactory bulb and along the perivascular space of the middle cerebral artery as highlighted in C. B: the same data set processed by the GlymphVis algorithm with advection and diffusion terms deriving CSF streamlines. These streamlines show brain parenchymal CSF flow patterns at a fixed point in time. Please note that the CSF streamlines including transport along the middle cerebral artery (D) are well matched to the contrast uptake in the original data (compare withA and C). Scale bars = 2 mm.

THE GS OPERATES MORE EFFICIENTLY IN THE SLEEPING BRAIN WHEN COMPARED WITH WAKEFULNESS

All the initial experiments on the GS were carried out on mice anesthetized with ketamine/xylazine (47). Nedergaard’s team proceeded to test the GS system’s functionality in different arousal states and discovered that the GS function differed between sleep and wakefulness (109). Specifically, they discovered that the GS influx of solutes into brain parenchyma was increased ∼80% in sleep compared with wakefulness inferring that the GS system is largely nonfunctioning in wakefulness (109). Furthermore, drainage of Aβ from brain parenchyma was ∼40% more efficient during sleep or anesthesia with ketamine/xylazine (KX) when compared with wakefulness (109). A dramatic increase in the ISF volume fraction (>40–60%) between sleep and wakefulness controlled by norepinephrine (NE) was discovered and attributed to the enhanced GS function. Specifically, it was suggested that solute transport in the ISF was less restrictive in sleep when compared with wakefulness (109). Collectively, these experiments also implied that GS function and waste clearance was inefficient in wakefulness regardless of the presence of AQP4 water channels. Of note, the observed increase in GS transport with KX anesthesia compared with wakefulness was attributed to an associated increase in slow wave delta power and decreased central norepinephrine (NE) tone (109). From these data, one can infer that the enhanced GS transport observed with KX anesthesia was mediated by xylazine and not by ketamine, which is known to increase central NE (57, 63). Xylazine is an alpha-2 receptor agonist and blocks central NE release (75) similar to the hypnotic dexmedetomidine (13, 54) used clinically for sedation and as an adjuvant for general anesthesia. In support of this statement, it was also demonstrated that anesthesia with dexmedetomidine and low-dose isoflurane increased GS transport twofold when compared with isoflurane alone (8).

GS FUNCTION AND PHYSIOLOGICAL DRIVERS

Rennels and coworkers (89) showed that unilateral carotid artery ligation impeded perivascular influx of CSF and tracer molecules leading them to conclude that normal arterial pulsatility was a major driver. Iliff and colleagues validated these data and further showed that an increase in pulsatility with dobutamine enhanced GS transport (48). More recent studies conducted using large 1.0-μm microspheres and direct visualization of the large PVS around pial surface arteries showed that CSF transport was indeed pulsatile and bulk flow driven at the surface of the brain (78). In addition, high pulsatility as observed in acute hypertension was shown to be ineffective in driving solute transport in the PVS (78). Paradoxically, while it appears that the rate of CSF-ISF exchange and Aβ clearance is highest during nonrapid eye movement (NREM) sleep (109), this physiological state is actually associated with periods of significantly reduced blood pressure, cerebral blood flow, and cerebral pulsatility (59, 60, 72, 96). However, the question of whether or not lower magnitude pulsatility might be overall more efficient for GS transport during sleep needs further investigation.

The importance of the spontaneous oscillations in arterial tone and diameter that occur in multiple vascular beds including in the brain for GS function is unknown. Termed “vasomotion” these rhythmic but very low-frequency (ranging from ∼3 to 25/min depending on vessel size) variations in arterial/arteriolar smooth muscle tone can produce fluctuations in vessel diameter comparable to those resulting from cardiac contraction and are influenced by a range of factors including general anesthesia and sedation (18, 43, 49, 64). While not generally considered to play a major role in bulk CSF flow, given the proposed role of the GS and clearance of Aβ, there has been particular interest in the potential relationship between impaired vasomotion and cerebral amyloid angiopathy (25). Notably, a recent study documented that spontaneous vasomotion correlated with perivascular clearance of solutes in live, awake mice and this driving force was impaired in mice with cerebral amyloid angiopathy (104). As with cardiac pulsatility, vasomotion is decreased during non-REM sleep (110). Implications of these findings in terms of GS function remain unclear, but when considered in conjunction with decreased cardiogenic arterial pulsatility during sleep, the data suggest that two physical factors thought to be the main drivers of GS function may actually be diminished when CSF-ISF exchange is highest.

LYMPHATIC VESSELS IN THE MENINGES

Lymphatic vessels were documented in the cerebral dura mater covering the human brain several decades ago (12). Using newer, state-of-the art imaging techniques and molecular markers of the lymphatic endothelial cells, the meningeal lymphatics were rediscovered in 2015 and thoroughly described in mice both structurally and functionally (4, 69). Meningeal lymphatics were observed using Prox1-GFP and Vegfr^3+/LacZ reporter mice and immunofluorescence in the dura mater surrounding the brain in a particular pattern (4, 69). Specifically, the majority of lymphatic vessels were observed to run toward the base of the skull along the transverse sinus, the sigmoid sinus, and the retroglenoid and rostral rhinal vein (4, 21, 69, 71). In areas of the skull foramina, lymphatic vessels could be observed to exit along meningeal portions of internal carotid artery and along cranial nerves (3). In other rodent studies, it was shown that meningeal lymphatics develop and mature after birth and their growth and development is dependent on vascular endothelial growth factor C (VEGF-C) (3). The functionality of the meningeal lymphatics was demonstrated by injecting inert tracers into the brain parenchyma of the Prox1-GFP mice and drainage of the tracer could be located on the meningeal lymphatics and at the level of the deep cervical lymph nodes (4, 69). Furthermore, a transgenic mouse with loss of dural lymphatics had reduced macromolecule drainage from the brain, but paradoxically, no increase in intracranial pressure suggesting alternative pathways for fluid and solute drainage (4). A recent post mortem study in humans confirmed the presence of lymphatic vessels in the dura; however, Aβ deposition in the wall of dural lymphatic vessels was absent (36) suggesting that these drainage pathways (at least in humans) might not be implicated in severe AD pathology.

Intriguingly, lymphatic vessels along the dural sinuses and along meningeal artery can be visualized in human brain after intravenous administration of a MR contrast agent (2). The visualization of lymphatic vessels is based on the fact that after intravenous administration of Gadobutrol, the MR contrast molecule leaks out of the dural blood vessels and travels through the ISF space into adjacent lymphatics (2). By implementation other special MR pulse sequences, the MR contrast signal from blood can be eliminated and after subtraction the meningeal lymphatics can be revealed. Using this novel in vivo approach to study meningeal lymphatics in humans will allow further investigations into the functionality and role of meningeal lymphatics in normal brain and in neurodegenerative disease states.

GLYMPHATIC TRANSPORT IN AGING, NEUROTRAUMA, AND NEURODEGENERATION

Brain parenchymal influx of CSF and Aβ drainage from ISF is significantly reduced in old mice when compared with young and middle-aged mice (62). The decline in GS transport function in aging mice is multifactorial and ascribed to loss of perivascular AQP4 polarization and neuroinflammation (62). GS function has also been shown to be decreased in a mouse model of AD (81), in traumatic brain injury (TBI) (44, 83, 88), and in stroke (34). Glymphatic clearance of tau in the “hit and run” TBI mouse model was shown to be reduced acutely after the insult and associated with later onset altered global AQP4 expression and loss of perivascular AQP4 polarization secondary to inflammation (44, 88). Specifically, in the TBI mouse model the temporal trajectories of intracranial pressure changes and tissue edema (peaking 3 days after TBI) were different from those of AQP4 expression changes (peaked at 7 days) post-TBI suggesting that the water channels were not directly related to edema information acutely after TBI (88). To summarize, in conditions of aging, TBI, and stroke, the perivascular CSF passage through brain tissue is deficient and GS transport and waste drainage is therefore less efficient. However, the underlying pathophysiology of impaired CSF influx in these various pathologies are different. For example, in stroke and TBI, CSF influx is nearly absent in the ischemic/lesioned hemisphere when compared with the contralateral side, secondary to tissue trauma (34), causing loss of vascular pulsatility and tissue edema with obliteration of the periarterial conduits in the parenchyma. In aging, the periarterial CSF influx and CSF-ISF exchange is compromised primarily secondary to perivascular inflammation and loss of AQP4 perivascular polarization (62).

Glymphatic transport has also been studied in animal models of cerebral small vessel disease (cSVD) (5). cSVD is frequently observed in the elderly human brain, and a common cSVD subtype is associated with thickening of the cerebral arterioles, so-called “arteriolosclerosis.” Arteriolosclerosis can progress to fibrinoid necrosis, microhemorrhage, or microinfarction, and capillaries are also affected and sometimes venules (32, 55, 97, 107). The pathogenesis of sporadic arteriolosclerosis cSVD is largely unknown but thought to result from hypertension, vasospasm or “failure of the endothelial barrier function” and ultimately impaired oxygen delivery to the tissues (70, 107). MRI-based diagnosis of cSVD include the presence of small subcortical infarcts, white matter hyperintensities, enlarged PVS, lacunes, microbleeds, and cerebral atrophy (22, 107). Thickening of the arterial wall and dilated PVS are thought to impair oxygen delivery to the tissue similar to what is documented in multiple sclerosis where tissue hypoxia is widespread (24, 73) although the precise mechanism is unknown. Rodent models of spontaneous hypertension have been used to investigate the effect of chronic hypertension on cSVD pathology in the brain. While some reports document increased GS bulk-flow-driven transport in the spontaneously hypertensive rat due to changes in vessel stiffness and arterial pulse wave velocity (5), others report that overall CSF-ISF exchange is reduced (78, 79).

HYPOXIA AND CSF TRANSPORT: IMPLICATIONS FOR HIGH-ALTITUDE SICKNESS

To the best of our knowledge, no animal experiments have investigated the effect of high-altitude hypoxia on GS transport. The potential mechanisms involved in potential GS changes in high-altitude sickness is discussed below and are based on current evidence of CSF transport in conditions where hypoxia is thought to be implicated (e.g., cSVD, stroke, and traumatic brain injury) and inspired by excellent recent reviews by Lawley et al. (65), and Hackett and Roach (40). Furthermore, given the common involvement of deep white matter in high-altitude cerebral edema (HACE) and cSVD we will highlight how perivascular transport of CSF might be affected in conditions of acute mountain sickness (AMS) and HACE. Symptoms of AMS include headache, fatigue, nausea, and vomiting and sleep disturbance; and the headache component is thought to involve pain transmission via the trigemino-cervical complex like in migraine headache (14). The much more severe condition of HACE is rare but can occur with rapid ascents to altitudes of >4,000 m and afflicted subjects have ataxic gait and altered mentation (40). The prime “insult” instigating altitude sickness is obviously related to hypoxia; however, it is currently not possible to predict who will be susceptible to developing AMS or HACE (40, 94). A hypothesis proposed states that individuals susceptible to high-altitude sickness are those with less intracranial and intraspinal “compliance” or a lower CSF-to-brain parenchymal tissue volume ratio (94). This hypothesis has been indirectly supported by studies demonstrating that older subjects have a lower incidence of AMS compared with younger subjects at moderate altitude (93) and evidence of higher CSF-to-brain tissue volume in elderly when compared with young adults (39). An in-depth discussion of the CNS compliance hypothesis was recently presented by Lawley et al. (65) and readers are referred to this excellent review for details of the proposed CSF pathophysiology in high-altitude illness. Here, we briefly discuss pathophysiology of AMS and HACE from the point of view of GS transport and brain waste drainage. Assuming that the primary outcome in high-altitude illness is rapid onset hypoxia, cerebral overperfusion, increased sympathetic activity and “brain swelling” secondary to vasodilation, and BBB compromise (in the setting of HACE) (40), several key points regarding how GS transport might be affected can be inferred as described in the following sections.

Hypoxic Vasodilation and Increased Cerebral Blood Flow

Adaptive mechanisms to optimize oxygen delivery during high-altitude hypoxia involve an extraordinary network of direct and reflex pathways that ultimately affect ventilation and hemodynamics. Several clinical studies using MRI and arterial spin labeling pulse sequences have documented significant (∼5–20%) increases in CBF (68, 105) in younger subjects with acute exposure to high altitude as well as reduced cerebral vascular reactivity (CVR) (105). Hypoxic arterial vasodilation by MR angiography was confirmed in the human brain at high altitude (68). Furthermore, enlargement of the cerebral venous sinuses by MR venography was also documented in human subjects exposed to a hypoxic challenge (108). Although no study as of yet has investigated solute CSF and parenchymal transport under conditions of high-altitude hypoxia, we have documented impaired GS transport in the setting of isoflurane-induced enlargement of the venous sinuses (8). It is likely, therefore, that high-altitude induced global vasodilation will negatively impact CSF influx and therefore GS transport.

Heart Rate and Respiration

Clinical research studies in healthy subjects have shown that during ascent to high-altitude heart rate and ventilation increase (31, 90, 101). The increased heart rate is caused by the associated hypoxemia (e.g., $P{a}_O{}_{{}_2}$ lower than 50 mmHg has been documented at 12–15,000 ft (31) and is chemoreflex instigated via chemosensitive cells located in the carotid bodies and the aortic body (42). Similarly, the increased minute ventilation (primarily increased tidal volume) at high altitude is also primarily mediated via low arterial O₂ and stimulation of the chemoreceptors (42). A moderate increase in heart rate would in principle increase CSF influx and facilitate enhanced CSF-ISF exchange (48). However, the more influential physiological driver of CSF dynamics at high altitude is likely to be an increase in respiratory tidal volume. Thus human studies have shown that voluntary deep inspiratory breathing is a major driver CSF fluid flow through the cerebral ventricles and basal cisterns (27). The mechanism underlying deep inspiratory breathing on accelerating CSF flow dynamics is related to a more negative thoracic pressure during inspiration, which will directly impact hydrostatic pressure gradients for flow along the perivenous conduits into meningeal lymphatics (27). Furthermore, a recent study showed that during normal human sleep, slow oscillating neural activity precedes coupled waves of blood and CSF flow in the brain (33). Thus, based on these data, one might hypothesize that at high altitude, the beneficial effects of increased respiratory tidal volume on CSF fluid flow would serve to counteract the negative effects of nocturnal hypoxemia and restless sleep (31) on overall waste drainage. More studies are needed to explore the potential beneficial effect of maximizing deep inspiratory breathing at high altitude for prevention of AMS.

Brain Swelling

Increased blood volume and brain volume increases are documented in high-altitude illness (38, 56). In HACE, vasogenic edema (evaluated by MRI and T2 relaxation) has been documented in deep white matter (corpus callosum) (41). However, whether or not cytotoxic edema occurs in AMS or HACE is contentious. In AMS, one study documented very small increases in T2 values in the splenium of the corpus callosum with exposure to hypoxia (56). The same study also reported that the apparent diffusion coefficient (ADC) increased during the hypoxic episode in most brain regions but in AMS subjects minor decreases in the ADC were documented, which suggest the presence of cytotoxic edema (56). Regardless, “brain swelling” in high-altitude illness is associated with displacement of CSF (94, 108). These data strongly suggest that CSF transport from the subarachnoid space into periarterial conduits in the brain parenchyma is compromised in high-altitude illness and CSF-ISF exchange and waste clearance will consequently decline. Furthermore, the formation of edema will further compromise cerebral perfusion eventually causing ischemia thereby instigating a vicious cycle toward aggravating the insult. It is tempting to speculate that the diversion of CSF away from brain parenchyma in the case of brain swelling in high-altitude sickness might be advantageous. CSF can certainly exit from the cranium without having to pass through the brain parenchyma (23, 53), and these alternate pathways would facilitate maintaining lower ICP. Intriguingly, cisternotomy and diversion of CSF (referred to clinically as “CSF-shift edema”) in severe cases of TBI have been shown to decrease brain swelling, mortality, and morbidity in afflicted subjects (15, 16).

Sleep Disturbances

Interrupted sleep and sleep disturbances have been documented in humans at high altitude (52, 84, 95). A recent metanalysis highlighting polysomnographic sleep studies revealed a reduction in nonrapid eye movement (NREM) sleep and reduction in slow wave sleep at high altitude (10). Because GS transport is most efficient during slow wave sleep (109), it could be inferred that brain waste drainage is impaired in subjects with restless sleep at high altitude and thus potentially contribute to the pathogenesis of acute mountain sickness (AMS). In support, a recent positron emission tomography study using a radioactive Aβ ligand showed increased uptake of Aβ in the brain of healthy human subjects after one night of sleep deprivation (98). Furthermore, a clinical research study conducted at high altitude, demonstrated that hypoxemia, unstable nocturnal ventilation (central apnea), and restless sleep were early symptoms in subjects who later developed AMS (31). There is currently a gap in knowledge regarding the effect of central or obstructive sleep apnea (OSA) on GS transport. However, increased perivascular space visibility on brain MRI images, a marker of cerebral small vessel disease (11, 106), is associated with OSA (17, 102) suggesting perivascular space dysfunction and indirectly inferring glymphatic transport impairment (106). Clearly, more studies on the impact of obstructive and central sleep apneas on CSF transport, GS transport, and waste drainage are needed to the further understanding of the pathogenesis of AMS.

In conclusion, the current concept of how the glymphatic system operates in the central nervous system (CNS) under normal conditions and in states of neurodegeneration was reviewed here. We also revealed that there is limited information on how states of hypoxia affect GS solute transport and waste drainage in the live brain. Furthermore, the reader must be aware that most of the discussion and data pertaining to hypoxia and pathophysiology of AMS from the point of view of GS transport are based on experiments conducted in rodents. Currently, a major barrier to understanding GS transport is the lack of noninvasive imaging technologies for accurate tracking solute transport and waste drainage in the human CNS. Future research efforts should focus on developing sensitive and specific biomarkers for tracking aberrant CSF fluid flow dynamics and endogenous waste drainage in real time.

GRANTS

This work was supported by National Institutes of Health Grants R01-AG-048769, RF1-AG-053991, and R01-AG-057705 and the Leducq Foundation (16/CVD/05).

DISCLOSURES

P. M. Heerdt has consultant agreements with Baxter International and Baudax Bio. No conflicts of interest, financial or otherwise, are declared by the other authors.

AUTHOR CONTRIBUTIONS

H.B. conceived and designed research; H.B. and R.E. interpreted results of experiments; H.B., R.E. and J.W. prepared figures; H.B., P.M.H., J.W. and A.T. drafted manuscript; R.E., P.M.H., S.K., Y.X., H.L., J.W. and A.T. edited and revised manuscript; H.B., R.E., P.M.H., S.K., Y.X., H.L., J.W. and A.T. approved final version of manuscript.