Role of aquaporin-4 polarization in extracellular solute clearance

A growing body of evidence suggests that excess interstitial fluid (ISF), solutes and waste products are cleared from the brain parenchyma via perivascular spaces through the so-called glymphatic system [1, 2]. The original groundwork of brain waste clearance pathways was performed by Cserr and colleagues in 1970–1980s who observed that waste from the brain was cleared in a size-independent manner along perivascular spaces [3–5]. The glymphatic model extends these observations and describes that flow of extracellular fluid in the brain involves three components: (1) cerebrospinal fluid (CSF) influx into the brain along the perivascular spaces (PVS) of penetrating arterioles, (2) fluid flow through the parenchyma, and (3) efflux of interstitial fluid and waste products from the brain in the PVS along veins. The initial glymphatic study and several follow up studies have shown that extracellular solute clearance is facilitated by the astrocytic water channels aquaporin-4 (AQP4), that are highly enriched in astrocytic endfeet [2, 6]. Later studies have demonstrated that glymphatic flow is driven by arterial blood vessel dynamics [7–9], is more effective in sleep [10–13] and under certain types of anesthesia [14], and is less effective in advanced age [15].

It is unclear through which mechanism the passive water channel AQP4 affects glymphatic flow. A role for AQP4 in brain fluid flow dynamics was already first suggested with its discovery in astrocytes in the late 1990s [16], and the direct role for AQP4 in clearance of injected fluid was first suggested in 2004 by Papadopolous et al. [17]. In the context of the glymphatic system, it has been suggested that the pronounced polarization of AQP4 to the perivascular endfeet in particular is important for both tracer influx [6] and efflux [18], but this has only been evaluated with non-quantitative fluorescence microscopy techniques. Also, the overall importance of AQP4 in brain waste clearance has been questioned by others in the field [19]. Hence, there is a need for quantitative assessment of the role of AQP4 polarization on solute clearance from the brain.

Here we aimed to quantitatively evaluate the role of AQP4 and AQP4 polarization to the astrocytic endfeet on glymphatic efflux of extracellular solutes. We measured clearance of 0.18 kDa [³H]mannitol and 500 kDa [³H]dextran infused through an acutely implanted cannula into the striatum of mice devoid of AQP4 (Aqp4^–/– mice) and mice that lack the AQP4 anchoring protein α-syntrophin, which results in 90% loss of the perivascular pool of AQP4 (Snta1^–/– mice) [20–22]. Although acute intrastriatal cannula implantation has been shown to impair glymphatic flow [23], we aimed to use similar techniques to what was used in the original glymphatic system reports [2, 10] to directly compare how solute clearance was affected by overall AQP4 removal and AQP4 polarization to the endfeet. We found that removal of AQP4 attenuated clearance of both 0.18 kDa [³H]mannitol and 500 kDa [³H]dextran, similar to the original reports [2]. Surprisingly, clearance of large and small solutes differed in Snta1^–/– mice. We found that the clearance of large 500 kDa [³H]dextran, but not the small 0.18 kDa [³H]mannitol was reduced in Snta1^–/– mice. Lastly, we evaluated how age within normal adult range affected the clearance of large and small molecules. We found that the clearance of 500 kDa [³H]dextran increased with age in adult mice, while clearance of 0.18 kDa [³H]mannitol remained unaffected. Our quantitative measures confirm that AQP4 is important for clearance of extracellular solutes and add novel insights that the polarization of AQP4 to the astrocytic endfeet and age within the adult range are more important for the clearance of large than small solutes.

Loss of AQP4 polarization is a hallmark of a wide range of brain pathologies, such as epilepsy [33], stroke [34], idiopathic normal pressure hydrocephalus [35] and Alzheimer’s disease [36]. Therefore, it is crucial to understand whether loss of endfoot AQP4 affects the clearance of extracellular solutes. Using a quantitative method of radioactive tracer infusion to the striatum of anesthetized mice, we here show that the subcellular polarization of AQP4 to the astrocytic endfeet may have a greater impact on clearance of large versus small molecules. Moreover, in accordance with the literature [2–4], we found a similar clearance for 0.18 kDa [³H]mannitol with a hydrodynamic radius of 0.4 nm and the ~2800 times larger 500 kDa [³H]dextran with a hydrodynamic radius of 16 nm. Lastly, we show that clearance of large molecules increased with age within the adult age range, while clearance of small molecules was not affected.

Astrocytic endfeet almost completely ensheath the blood vessels of the brain and likely represent a barrier between the PVS and the parenchyma hindering free flow of fluids and solutes [37]. Iliff et al. [2] suggested that the highly concentrated expression of AQP4 in the endfeet was important for the glymphatic system, and recent semi-quantitative studies of the distribution of injected fluorescent tracers in Snta1^–/– mice have suggested both an attenuated influx [6] and efflux [18] of tracers in these mice. Partly in agreement with these reports, using quantitative assessment of clearance of radioactive tracers, we demonstrate that perivascular pool of AQP4 is more important for glymphatic clearance of large than small extracellular solutes. We found no difference in the clearance of [³H]mannitol between Snta1^–/– and WT mice, while for 500 kDa [³H]dextran we found a reduced clearance in the Snta1^–/– mice 2 and 3 h after injection. This suggests that the highly polarized expression of AQP4 cannot account for all the effects of AQP4 for brain solute clearance. We did not observe a significant increase in clearance between 2 and 3 h timepoints. The fact that tracer clearance is faster initially, is in line with existing literature [2, 10, 23], and could be explained by a model where tracer clearance is concentration dependent, suggesting a role for diffusion. However, similar clearance rates between vastly different molecular sizes of tracers observed by us and shown in a range of previous studies [2, 3], suggest that convection plays a role. Likely our observations are best explained by a combination of parenchymal diffusion, which could be one of the mechanisms contributing to faster initial clearance, and perivascular convection, accounting for size-independence of overall clearance [27]. Lastly, we cannot rule out potential issues with variability or technical concerns such as infusion triggered local pressure increases or clumping of the 500 kDa dextran molecules which could clog the drainage pathways.

It is unclear how AQP4 contributes to glymphatic clearance. It is tempting to think that the channels facilitate transcellular fluid flow across the endfoot sleeve based on the dense clustering of these channels at the endfoot membrane facing the vessel, but it is unclear how such transcellular fluid flow could occur and what would be the driving forces acting upon this system. The resistance to flow has been modeled to be considerably lower between endfeet than through AQP4 [27]. If fluid flow in between adjacent endfeet is the main exit route for fluid from the PVS to the parenchyma and vice versa, one could imagine that these slots are subject to gating mechanisms that are affected by the presence of AQP4. Yet, to the best of our knowledge, no study has been able to address whether endfoot coverage of vessels in vivo can vary or is subjected to regulation. It has also been suggested that the increased size of the extracellular space in Aqp4^–/– mice could play a role [19]. However, in this case one would rather expect an increased clearance in the Aqp4^–/– mice, given the proposed role of increased extracellular space during sleep in the glymphatic system [10]. Similarly, in Snta1^–/– mice, a higher baseline extracellular space has been shown, albeit to a lesser extent than in Aqp4^–/– mice [38]. However, solute influx from the CSF to the interstitial space seems to be AQP4 independent as shown by Smith et al., via application of fluorescent dextrans to the brain surface under constant pressure [39]. The concentration of AQP4 in endfeet has been reported to be higher around capillaries and veins than around arterioles [2]. This observation might suggest that AQP4 is more important for processes not linked with hydrostatic forces associated with perivascular bulk flow around arterioles. One avenue worth exploring in future studies may be to what extent the capillaries contribute to ISF production in the brain, as already suggested in the pioneering work of Cserr et al. [40]. Even though equally as elusive as the role of AQP4 in the hydrostatic framework of the glymphatic system, potentially AQP4 could play a role in facilitating BBB production of ISF. Finally, a large and growing body of literature suggests that endfeet possess the machinery to modulate vessel tone, even though to what extent endfeet partake in vascular control remains controversial [41–43]. In Haidey et al. they propose that stretch-sensitive TRPV4 channels in the astrocytic endfeet play a role in modulating vasomotion [43]. Potentially AQP4 may play a role in the same signaling pathways, enabling the endfeet to react to vascular dynamics, which again may influence the driving forces of the glymphatic system. Future studies should address the mechanistic underpinnings of how AQP4 affects clearance of solutes, for example evaluate whether AQP4 could regulate the morphology of endfeet, the inter-endfoot gaps or vascular dynamics; and selectively manipulate AQP4 expression and polarization levels in arterioles, venules and capillaries and evaluate how this contributes to clearance of small and large molecules.

The role of AQP4 in interstitial waste clearance as originally proposed by Iliff et al., and reported in numerous follow-up studies was challenged by Smith et al. who did not detect any differences between Aqp4^–/– and WT mice [19]. Our findings are in line with the original Iliff et al. study, as we find that extracellular solute removal is reduced in mice lacking AQP4. However, compared with Iliff et al., we find a somewhat smaller effect of AQP4 on clearance. The reason for these discrepancies with the original study is not entirely clear, but could be due to differences in volume of tracer injected (0.5 µl in our study vs. 1 µl in Iliff et al.), different injection methods, or different protocols for radioactivity counting.

Intraparenchymal injection pressures could explain different conclusions about glymphatic flow in different studies [6]. Arguably, injection pressures could easily exceed any potential hydrostatic driving force created by arterial pulsations, confounding assessment of direction of flow in the parenchyma, and potentially also exit pathways. Hence, we did go to great lengths in order to minimize any pressure artifacts by using a very low speed of infusion and a minimal injected volume (0.5 µl over 30 min). Moreover, glymphatic flow is reduced after acute cannulation [23]. While we cannot exclude that tissue responses to acute cannulation, such as BBB disruption, inflammation, swelling, global CNS responses such as intracranial pressure or recovery of lost CSF volume, or steric effects might differ between genotypes and could contribute to different clearance results, we aimed to use similar methods as in the original glymphatic system report that described the effect of AQP4 [2] for direct comparison of the effect of AQP4 polarization to the endfeet. Future studies should confirm our findings using less invasive methods such as the real-time in vivo clearance assay for quantification of glymphatic efflux in which tracers are infused via a chronic implanted cannula and clearance is measured in periphery avoiding post mortem artifacts [23].

One surprising finding from our dataset was the significant positive correlation between age and clearance. Previous studies have demonstrated that clearance is negatively affected by age [15]. However, these studies are not directly comparable to ours as they investigated young (2–3 months) middle aged (10–12 months) and old mice (18 months), whereas we analyzed the dependency on age in mice within our experimental groups (ranging from 60 to 150 days of age). From our data, it is apparent that one should take the age of mice carefully into consideration, even within groups of adult mice, as this may influence the outcome. One potential mechanism through which clearance could depend on age is expression of AQP4. Even though we did not observe any overall differences in the levels of AQP4 between the groups of 60, 100 and 150 days old mice, it could be that polarization of AQP4 to the endfeet or that selective AQP4 expression around arterioles, venules and capillaries varies in this age range.

To conclude, using a quantitative method we confirm findings from previous semi-quantitative fluorescence studies that both the presence and subcellular distribution of AQP4 is important for brain solute clearance, and add that large and small solutes might be cleared by partly different mechanisms, where removal of large molecules partly depends on AQP4 polarization to the endfeet and increases with age within adult range.