Cerebrospinal fluid turnover as a driver of brain clearance

Cerebrospinal fluid (CSF) has historically been considered to function as a sink for brain‐derived waste disposal. Recent work suggested that CSF interacts even more intensely with brain tissue than previously recognized, through perivascular spaces that penetrate the brain. Cardiac pulsations, vasomotion, and respiration have been suggested to drive CSF flow in these perivascular spaces, thereby enhancing waste clearance. However, the intrinsic role of CSF production in relation to its distribution volume (turnover) is not an explicit component of recent concepts on brain clearance. Here, we review the work on CSF turnover and volume, focusing on preclinical evidence. Herein, we highlight the use of MRI in establishing CSF‐related parameters. We describe the impact of sleep, effect of anesthesia, aging, and hypertension on CSF turnover, and how this relates to brain clearance. Evaluation of the available evidence suggests that CSF turnover is a major determinant in brain clearance. In addition, we propose that several putative drivers of brain clearance, but also conditions associated with impaired clearance, such as aging, may actually relate to altered CSF turnover.

conditions such as brain injury, 9 small infarcts, 10 hypertension, 11 and models of cerebral amyloid angiopathy, 12 were found to have a detrimental effect on these parameters.
A common denominator of the proposed driving forces of brain clearance is that they involve oscillations in blood vessel diameter, intracranial pressure, or both, which generate an oscillatory motion of CSF.These may be present in the ventricular system, subarachnoid space (SAS), or extend along the perivascular spaces.Oscillations differ in time scale and magnitude, related to the frequencies of the heartbeat, respiration, and vasomotion.Although oscillations of CSF are likely to enhance solute dispersion between the CSF and interstitial fluid (ISF) through mixing, the question of whether oscillations of fluid generate net flow remains debated. 13,14In this respect, the synthesis of CSF by the choroid plexus is a more obvious source of CSF net flow.In the current landscape of brain clearance research, the role of CSF formation as a potential driver of brain clearance appears to be underappreciated, although some modeling studies clearly indicate its importance. 15The aim of this review is therefore to investigate if CSF turnover, which is defined as CSF production divided by its distribution volume, could affect brain waste clearance, and whether proposed driving forces and conditions of impaired clearance could in fact be attributable to their impact on CSF turnover.

| CSF PRODUCTION BY THE CHOROID PLEXUS AND OTHER POTENTIAL SOURCES
CSF is present in the ventricles, and in the SAS surrounding both the brain and spinal cord.It has several roles, with its main function being homeostasis of the central nervous system.It also provides protection from mechanical impact, reduction of the brain's effective weight, and transportation of, for example, hormones and nutrients. 16,17CSF is composed primarily of water and salts, mixed with proteins, hormones, and neurotransmitters.Several studies have determined the choroid plexus as the most prominent site of CSF production. 17A recent review by MacAulay et al. provides an excellent overview of the molecular mechanisms of CSF formation. 17The choroid plexus is found in all four ventricles and consists of a specialized vascular bed, stromal tissue and a monolayer of epithelial cells, lined with microvilli to maximize the contact area with the inner space of the ventricles. 16The choroid plexus acts as a blood-CSF barrier (BCSFB), due to the presence of tight junctions located between the choroid plexus epithelial cells. 18These proteins establish a strong connection between cells, effectively impeding the passage of a wide range of substances from the blood into the CSF. 18These proteins form a tight bond between cells that prevents movement of a large variety of substances from blood to CSF.A human MRI study revealed increasing choroid plexus volume with increasing age, which is consistent with earlier histological findings in rats. 19,20With arterial spin labeling (ASL) MRI a reduction in blood perfusion of the choroid plexus with advancing age was found, suggesting the choroid plexus volume increases to compensate for the loss in blood flow. 19ter homeostasis is of great importance in the body, especially in the brain, where increasing water levels can lead to brain swelling and increased intracranial pressure.Therefore, maintaining adequate water levels in brain cells, CSF, and ISF, is of great importance.An essential component in the movement of water molecules between these compartments is the presence of aquaporins (AQP). 21AQP are transmembrane proteins that allow the movement of water and a few other molecules, including glycerol, urea, and monocarboxylate. 21The most abundant aquaporin type in the central nervous system is AQP4, which is expressed on astrocytic endfeet surrounding blood vessels and at the glia limitans surrounding the brain and spinal cord. 22AQP are a crucial element of the glymphatic theory, facilitating fluid and solute movement from the perivascular spaces to the brain interstitium. 3Pharmacological blocking of AQP4 channels in mice leads to decreased tau clearance, supporting their role in brain clearance. 23In addition, a reduction was seen in perivascular AQP4 polarization in older mice, concomitant with an observed 40% reduction in the clearance of injected amyloid-β in these mice. 8These and many other findings suggest an important link between impaired AQP4 functioning and reduced brain clearance.However, the exact role of AQP is heavily debated in this context. 246][27] A possible other source proposed here is the secretion or leakage from blood vessels into the ISF, or directly into the CSF.This source of CSF formation could originate from blood vessels throughout the brain.Recent support for this production route comes from research using ASL, where CSF-ASL signal was not only found in the choroid plexus, but also in the SAS throughout the cortex. 28Nonetheless, the extent to which the labeled water efflux is offset by the influx of water from the CSF into the bloodstream remains uncertain with the utilization of this method.Consequently, the limited ability to ascertain the net contribution of this source to CSF formation restricts the extent of conclusions that can be drawn through this approach.
The production of CSF in mice is determined to be around 50-150 nL/min, depending on experimental conditions, age, sex, and type of anesthesia, 29 while in humans it is estimated to be around 0.3-0.4mL/min. 30The formation of CSF is not a trivial parameter to measure, as it is easily affected by the experimental approach.Measurements of CSF production can be classified as either direct (measured in the ventricles) or indirect (measured in the cisterna magna). 31Examples of the methodologies used for measuring CSF production include the indirect perfusion method, the Welch method, and the air-replacement method. 31The indirect perfusion method, also called the ventriculo-cisternal perfusion method, entails the infusion of tracer into the lateral ventricle and subsequent collection from the cisterna magna, allowing a calculation of CSF production based on the tracer dilution.The Welch method compares the hematocrit between aortic blood samples and venous blood samples from the choroid plexus, with the resulting volume loss serving as a measure of CSF production.The infusion of air in combination with roentgenographic imaging of the ventricles is used in the air-replacement method.In this method, CSF production can be determined through fluid and time measurements in the ventricles. 31Limitations in these methodologies can be found in interference with changes in intracranial pressure, the invasive nature of the procedure (including anesthetics), and the assumptions regarding homogenous mixing of tracers.MRI overcomes some of these limitations, and several MRI applications have been used to estimate CSF production, such as phase-contrast MRI (PC-MRI) and ASL-MRI. 31PC-MRI combines phase contrast and cardiac cine technologies to quantify flow, and has been widely used to study CSF dynamics. 32,33Previous CSF flow measurements using PC-MRI measured the velocity in the cerebral aqueduct. 33,34However, CSF flow measurements using this method are shown to be confounded by the respiratory cycle. 35ASL-MRI can visualize cerebral perfusion by magnetically labeling arterial blood without contrast injection. 28,36Using this method, labeled water crossing the BCSFB can be measured, providing information regarding the water exchange between these compartments in the brain.

| CSF FLOW THROUGH THE VENTRICULAR SYSTEM AND ACROSS THE BRAIN SURFACE FOLLOWING PERIVASCULAR SPACES
CSF flow patterns can be visualized using dynamic contrast-enhanced (DCE) MRI.In a preclinical study, this technique was used in combination with intracisternal paramagnetic contrast administration in CSF to trace CSF flow throughout the brain. 37After 10 min, contrast was detected at the cisterna magna and fourth ventricle. 37,38Subsequently, contrast moved both rostral along the cerebral arteries, as well as caudal to the spinal cord. 38DCE-MRI signal was observed to distribute to the olfactory bulb, rhinal fissure, and cerebellum after 20 min, where it continued along the middle cerebral artery after 30 min. 37,38Similar findings regarding the CSF route were reported in other research. 39,40Contrast was found to be cleared from the brain, including the hippocampus, after 6 h postintracisternal injection. 40human DCE-MRI study using intrathecal administration of gadobutrol imaged at various time points, showed a similar CSF flow pattern as demonstrated in preclinical data. 41Contrast was seen to distribute from the foramen magnum via perivascular spaces along the anterior, middle, and posterior cerebral arteries to the front of the brain. 41This distribution was uniformly throughout the SAS surrounding the brain, in an antegrade manner along the cerebral arteries.While the spatial distribution pattern appeared similar in humans compared with rodents, contrast in human brain parenchyma remained detectable after 24 h, suggesting that contrast clearance was slower in humans compared with rodents. 41sed on ex vivo fluorescence imaging, it was suggested that the CSF compartment on the surface exhibits continuity with the perivascular spaces surrounding the penetrating arteries, potentially serving as an influx route for CSF transport. 37,42Nevertheless, it is important to consider that postmortem artifacts could potentially impede the interpretation of tracer distribution patterns. 43,44In vivo experiments have demonstrated that both smaller molecules (< 1 kDa) and larger molecules (200 kDa) are transported through the perivascular spaces at similar rates, indicating the prevalence of bulk flow in their transport. 37While microspheres display oscillatory flow on the brain surface, they do not penetrate the perivascular spaces along the penetrating vessels. 43,45Consequently, it remains unclear whether bulk flow extends from the brain surface into the brain tissue, or if larger particles such as microspheres and cells are hindered by semipermeable membranes within this region.
The influence of anesthesia on CSF flow should not be disregarded in these preclinical studies.Previous investigations have demonstrated that the choice of anesthetic can significantly impact the dynamics of CSF signal enhancement in DCE-MRI. 46Specifically, isoflurane anesthesia has been associated with slower CSF signal enhancement compared with the use of ketamine/xylazine anesthesia. 46Under ketamine/xylazine anesthesia, the contrast agent was observed to travel via the ventral part of the brain along the major arteries of the circle of Willis to the nasal turbinates, with some penetration into the brain parenchyma and further exit through the pharyngeal lymphatic vessels. 46Conversely, under isoflurane anesthesia, a lesser amount of signal was detected in the olfactory bulb, while a substantial portion of the contrast agent was diverted towards the cervical area. 46The role of anesthetics was systematically examined in a study that evaluated the effect of various types of anesthetics on CSF tracer influx. 47The results indicated that tracer influx into the brain parenchyma was highest in the group under ketamine/xylazine anesthesia, followed by isoflurane/dexmedetomidine, while isoflurane or avertin alone resulted in the lowest tracer influx. 47Similarly, Ma et al. 44 found a reduction in brain penetration of tracer by isoflurane, compared with a mixture of ketamine/medetomidine.One possible explanation for these findings is that certain anesthetics, particularly isoflurane, facilitate the rapid exit of tracers via the CSF to the lymphatic system, or limit brain penetration because of inhibition of vasomotor function.

| CSF EXIT ROUTES
The efflux routes of CSF have recently been reviewed by Proulx 48 and include several sites of drainage from the cranium and spinal cord.Historically, textbooks have ascribed an important role for drainage to arachnoid villi projecting into the venous sinus, but Proulx argued that support for a role of these structures as sites of CSF outflow is actually relatively weak.Instead, evidence for perineural outflow routes is much stronger.This includes perineural sheaths surrounding olfactory nerves that cross the cribriform plate to allow CSF drainage into nasal lymphatics.This pathway and other perineural exit routes at the base of the brain subsequently connect to cervical lymph nodes.Support for drainage along spinal cord nerves to lymphatics is also present. 49Finally, dural lymphatics have been discovered to be an important factor in CSF drainage. 50,51There is evidence that dural lymphatics connect to skull bone marrow, which may play a role in immunological responses, 52 but whether these contribute substantially to CSF outflow remains to be established.

| CSF VOLUME
CSF volume is the sum of all CSF present in the ventricles, SAS, sulci, and perivascular spaces.Accurate quantification of whole brain CSF volume is very challenging.Segmentation of high-resolution T1-weighted MRI images is traditionally used to estimate the fraction of gray matter, white matter, and CSF. 53,54This technique suffices to quantify CSF in the larger ventricles and SAS; however, it suffers heavily from partial volume effects and therefore is not capable of accurately quantifying fluid volume with subvoxel resolution in the smaller perivascular spaces and dural lymphatics.6][57] These techniques suffer much less from partial volume effects because CSF volume can be quantified as a volume percentage, where signal in the large ventricles is scaled to containing 100% CSF.However, to the best of our knowledge, to date, this approach has not been applied in preclinical studies.

| DOES CSF TURNOVER AFFECT BRAIN CLEARANCE?
Brain clearance refers to the removal of waste products from the brain tissue itself (not CSF), which is mostly studied via injection of tracer substances into the brain interstitium.In addition, many studies have investigated brain clearance after injection of tracers into the CSF compartment, usually the cisterna magna or the ventricles.When studying brain clearance after CSF infusion, tracers first need to enter the brain tissue across the pia mater or via the perivascular spaces along penetrating arteries.Subsequently, tracers may leave the brain either along basement membranes of arteries 58 or along perivascular spaces of veins, 3 directly cross the pia mater and re-enter the CSF, follow white matter tracts, 59 or directly enter the CSF of the ventricles.
Mechanistically, CSF flow may directly sweep away tracers and waste from the brain by bulk flow when these enter the CSF, but also support a concentration gradient for diffusion of interstitial solutes toward the CSF.Brain clearance, therefore, depends on CSF flow, but not in a oneto-one relationship 60 (Figure 1).Thus, CSF flow apparently may bypass the brain tissue under certain conditions.An example of this is when isoflurane is used in brain clearance studies.Isoflurane increases CSF production, 29 yet it negatively impacts tracer penetration into the brain tissue and its subsequent removal in comparison with ketamine-xylazine anesthesia. 61The combined effect of reduced tissue penetration and increased CSF production appears to enhance tracer outflow to the blood, bypassing the brain tissue. 29,44It is conceivable that perivascular spaces surrounding penetrating arteries may not consistently provide complete accessibility.This could be attributed to factors such as vasodilation, potentially leading to a reduction in the size of these perivascular spaces.Yet, there is a great overlap between conditions of altered CSF production and brain clearance.These include both physiological processes and experimental interventions.A clear example of the latter is provided by studies using acetazolamide, a drug that inhibits carbonic anhydrase.A study by Barbuskaite et al. used acetazolamide to reduce intracranial pressure and showed that it reduced intracranial pressure via its action on CSF production. 62In the same study, inhibition of CSF production also reduced the dispersion of dye over the brain surface via the CSF.These observations correlate well with the results obtained by Lundgaard et al., who   showed that acetazolamide reduced brain clearance of both inulin and lactate, although in that study the effect on CSF production was not determined. 63Further evidence for the relationship between CSF production, tracer distribution, and brain clearance is discussed below.

| DOES SLEEP ENHANCE BRAIN CLEARANCE VIA AN INCREASE IN CSF TURNOVER?
The choroid plexus is under the control of circadian rhythms, with the highest CSF production in humans during the night. 64,65The time course of CSF production negatively correlates with CSF levels of amyloid β in humans, suggesting that the enhanced production of CSF contributes to the clearance of amyloid β.Sleep and circadian rhythmicity are closely related, but they are not identical.While the production of CSF in humans peaks at night, subjects in that study were awake at each time point, stressing the importance of circadian rhythms on CSF production. 65Recent work by Eide et al. addressed the role of sleep itself in waste clearance. 66These authors showed that poor sleep is associated with higher brain tissue levels of intrathecal infused tracer, indicated by DCE-MRI over time.This suggests that the removal pathways from the brain are negatively affected by sleep deprivation.In another study, already after one night of sleep deprivation, the burden of amyloid β was increased in certain brain areas. 67Unfortunately, direct measurements of the impact of sleep itself on CSF production in humans are currently lacking.In rodents, there is even a greater paucity of data on circadian CSF production and disentanglement from the impact of sleep.This is probably due to the fact that it is not trivial to measure CSF production in rodents both awake and during sleep.In addition, it should be acknowledged that the level of waste products in CSF is not only determined by CSF turnover, but also the production rate of waste may depend on the state of arousal.
Other factors may also contribute to the beneficial effects of sleep on brain clearance.Xie et al. suggested an expansion of the brain extracellular space during sleep in mice, which may facilitate clearance of the interstitium. 7In vivo imaging of naturally sleeping mice demonstrated varying blood vessel sizes that were correlated with changes in perivascular space size depending on the sleep state. 68Body posture also affected CSF efflux patterns, with most efficient brain clearance in the lateral sleeping position. 69A more recent study has identified the appearance of synchronized vasomotion and CSF oscillations during nonrapid eye movement sleep as an important contributor to brain clearance. 70However, while sleep was found to greatly enhance CSF oscillations, this does not imply net production of CSF.Taken together, several indices of CSF circulation depend on sleep, 71 but data on net production of CSF during sleep are lacking.
F I G U R E 1 Hypothesized contribution of CSF turnover to brain clearance.(A) Exchange of solutes between the CSF and brain interstitium depends on CSF turnover as it sweeps waste from the brain and maintains a concentration gradient for diffusion.(B) This process is facilitated by perivascular spaces along penetrating vessels.In specific circumstances, the involvement of perivascular spaces surrounding the penetrating vessels in the clearance process may be hindered.(C) One example of this hindrance could be the persistent vasodilation and subsequent closure of these spaces.CSF, cerebrospinal fluid.

| AGE REDUCES CSF TURNOVER, DOES THIS IMPACT BRAIN CLEARANCE STUDIES?
In mice, aging was found to be associated with a gradual decrease in CSF production. 29A similar finding was reported in rats, 72 while humans are also known to show a decrease in CSF production with age. 73Additionally, in older mice, there is a noticeable decline in CSF drainage toward the lymphatic system. 74,75Furthermore, there are noticeable changes in the characteristics and integrity of the meningeal lymphatic vessels, indicating a potential reduction in their functionality as a result of aging. 76Because aging at the same time is associated with brain atrophy and ventricular enlargement, CSF volume is expected to increase.This implies that CSF turnover decreases substantially with aging.The removal of waste prod- parenchyma.It is important to note that the model assumes diffusion in the brain interstitium and does not incorporate perivascular spaces.The simulations show that various models of CSF efflux, via the cribriform plate, arachnoid granulations, or meningeal lymphatics, all speed up clearance from years (based on diffusion only) to days. 14When we apply this model and assume that CSF production is reduced by 50% related to aging, both initial dispersion and subsequent removal are delayed in the CSF compartment and brain parenchyma. 73In addition, at later stages, the concentration of tracer is increased in the parenchyma because of the persistent influx from the CSF (Figure 2).When we model an increase in CSF volume (by a 1 mm expansion of the SAS) to mimic brain atrophy, the peak concentration in the CSF is lowered, while both dispersion and removal in the CSF and parenchyma are delayed.A combination of lowered CSF production and enlarged CSF volume further delays both the initial dispersion and removal from both compartments.The most important message from these findings is that in experimental studies where tracers are infused to study clearance or glymphatic function, the data are heavily dependent on CSF production and volume.A snapshot of tracer distribution taken at early time points may be interpreted as originating from impaired glymphatic function, while it could actually also be attributed to changes in CSF turnover. 8

| IMPAIRED CSF FLOW IN HYPERTENSIVE MODELS
Hypertension is a well-known risk factor for the development of cerebrovascular disease, consequently increasing the risk of several diseases, including Alzheimer's disease and stroke.Hypertension can induce vascular changes that impair vascular elasticity, thereby contributing to the development of vascular pathologies.The stiffening of cerebral blood vessels may potentially exert adverse effects on the exchange of waste products at the BCSFB, possibly through the impairment of oscillations by the vessel wall. 77human study using phase-contrast cine MRI compared various CSF flow parameters between nontreated patients with essential hypertension and control participants without symptoms of atherosclerosis.The authors found a decrease in average flow, volume, and velocity of CSF in patients with hypertension. 78In addition, increased systolic blood pressure was associated with lower CSF flow dynamics. 78E-MRI revealed ventricular reflux of CSF in spontaneous hypertensive rats compared with normotensive rats, suggesting altered CSF flow dynamics and thereby possibly impaired brain clearance. 11An increased backflow of CSF was also reported in mice using particle tracking. 43Earlier in vivo experiments from our group showed no evidence of differences in the blood-brain barrier and BCSFB permeability for small fluorescein molecules in hypertensive rats. 79However, other investigators found that hypertensive rats showed a 36% decrease in the delivery of labeled arterial water into the ventricular CSF, suggesting that there might be a defective BCSFB function for water molecules. 36Moreover, MRI revealed differences in several anatomical structures in animals with hypertension compared with normotensive animals, namely, larger lateral, third, and fourth ventricles, a smaller corpus callosum, and a smaller intracranial volume. 80Whole brain apparent diffusion coefficient (ADC) values were also lower in hypertensive rats, all adding up to the notion that the fluid balance is altered with hypertension. 80

| CONCLUSION
In the field of brain clearance research, the role of CSF turnover seems underappreciated as an important contributor to waste removal.CSF turnover sweeps away waste coming from the brain, but also maintains a concentration gradient from ISF to CSF for diffusive transport.The glymphatic concept for brain clearance does not explicitly incorporate CSF turnover.Yet, reduced CSF turnover, resulting from increased CSF volume, decreased CSF production, or a combination of both, is expected to slow down endogenous waste clearance.In addition, reduced CSF turnover, for instance with aging, may also obscure experimental results, because it affects tracer concentration and distribution kinetics.We think that in the context of human studies, MRI would be the appropriate noninvasive technique to measure both CSF volume and production.When it comes to animal studies, MRI has the greatest potential for simultaneously measuring these parameters within the same animal under physiological conditions.However, it is important to acknowledge that limitations pertaining to spatial resolution exist, necessitating the utilization of sophisticated methods to accurately determine CSF volume by accounting for partial volume effects.Additionally, further advancements are required to address confounding factors such as respiration when measuring CSF production at the level of the aqueduct. 35In particular, accurate data on total CSF volume in rodents are currently lacking, hindering estimations of CSF turnover under physiological and pathological conditions in preclinical studies.
ucts from the brain is obviously decreased under these conditions, but what are the consequences of reduced CSF turnover for experiments that address brain clearance and glymphatic function?Hornkjøl et al. recently published a model that allows studying the relationship between tracer dispersion in the CSF and in the human brain interstitium. 15In the model, a certain volume of tracer is infused into the CSF at the level of the foramen magnum.The concentration of CSF contrast is then modeled in a spatial and temporal manner in both the CSF compartment and brain F I G U R E 2 Modeling of tracer dispersion and clearance based on Hornkjøl et al. 15 (A) Theoretical profile of tracer kinetics in the CSF and brain parenchyma of a human under normal conditions (blue lines), after reduced CSF production (yellow), after enlargement of CSF volume (green), and a combination of reduced CSF production and volume enlargement (purple).(B) Illustration along the longitudinal fissure of the brain model showing the different compartments of the brain model with the ventricular system (brown), the cerebral aqueduct (red), the white matter (light blue), gray matter (mid blue), and SAS (dark blue).Examples of tracer distribution, (C) 5, (D) 24, and (E) 72 h after injection from the case with a large SAS and normal production.CSF, cerebrospinal fluid; prod., production; SAS, subarachnoid space.