Clearance from the mouse brain by convection of interstitial fluid towards the ventricular system

In this study we investigated clearance from the brain and the role of peri- and paravascular
pathways herein. The perivascular route is considered to be located along basement
membranes within the vessel wall 10]. On the other hand, the paravascular space is located in between the brain parenchyma
(astrocyte endfeet) and the vessel, and is filled with CSF 9]. The contribution of diffusion and bulk flow in the removal of metabolites and waste
from brain extracellular fluids is a longstanding debate that recently has regained
interest. Based on earlier work from Cserr 14]–16], Rennels 4], Weller 10], and others, Abbott 6] concluded that ISF is not a static fluid, although a later review by Sykova and Nicholso
13] concludes that bulk flow is likely to be restricted to perivascular spaces and the
main mechanism for exchange in the brain extracellular space is diffusion. More recent
work of the Nedergaard group 9], 17], 18] suggests that paravascular pathways provide an influx route for CSF along arteries,
which mixes with ISF and leaves the brain via venous outflow pathways. However, as
pointed out by Hladky and Barrand 19], this view is at variance with many earlier studies and awaits further confirmation.

Bulk flow from the striatum towards the ventricular system

The current study presents several arguments for the presence of bulk flow of interstitial
fluid from the striatum towards the ventricles. The first evidence is the transport
of the tracers after injection into the striatum, where the tracers distribute from
the injection towards the nearest lateral ventricle. Both F-500 and TR-3 moved along
extracellular spaces, albeit that TR-3 was also taken up by parenchymal cells, in
accordance with in vitro data 20]. Such transport could theoretically be caused by diffusion, convection, or both.
Discrimination between these processes is possible based on the distances traveled
and the directionality of such transport. F-500 would be expected to diffuse a far
smaller distance than TR-3 which, in our hands, was not the case. In addition, F-500
clearly moved towards the ventricles, and accumulated strongly at the ependymal layer.
The accumulation is suggestive of a sieving process, where ISF is transported to the
ventricle and part of the tracer stays behind. For the small TR-3, transport distances
for diffusion would still be in the order of magnitude that we found 13], but also here the tracer preferentially moved towards the ventricles. This dye accumulated
to a lesser extent, reflecting the possibility that it passes the ependymal layer
more easily than F-500. Taken together, this suggests a significant contribution of
tracer drainage by ISF bulk flow from the striatum into the ventricular system. In
the lateral ventricle, ISF mixes with CSF once it passes the ependymal cells at the
ventricle edge. Then, tracers follow CSF bulk flow along the ventricular system and
get further distributed over the subarachnoid space.

The choroid plexus as a potential exit route from the CSF

Floating in the ventricles, the choroid plexus is considered as the main source of
CSF. In this study, we found that TR-3 penetrates into the choroid plexus. Distances
are short, so penetration could be the result of diffusion from the ventricular CSF.
Alternatively, it has been shown that active transport can be bi-directional over
the epithelial cells of the choroid plexus 21]. The epithelial cells of the choroid plexus constitute the blood-CSF barrier (BCSFB)
that limits the movement of solutes from blood to the CSF and vice versa. However,
like the blood–brain barrier, the BCSFB contains transport systems for influx and/or
efflux of nutrients, metabolic products and ions, and might also be involved in detoxification
processes 21], 22]. This also opens the possibility that the choroid plexus may function as an exit
route for waste products from the CSF. Indeed, ex vivo data show the removal of amyloid
? from the choroid plexus 23] and in vivo, there is a correlation between choroid plexus dysfunction and amyloid
? removal 24]. Nevertheless, the current data did not allow us to discriminate between diffusion
and active transport of this particular tracer over the BCSFB.

Lack of evidence for bulk flow from the subarachnoid space and paravascular spaces
into the brain parenchyma

CSF and ISF may not only communicate at the ependymal layer of the ventricles, but
also at the pial level 16]. In our experiments, after injection into the cisterna magna, the mixture of fluorescent
tracers spread in the subarachnoid space. We observed that TR-3 entered the parenchyma
close to the SAS and arteries, while F-500 was confined to the SAS and paravascular
spaces. The TR-3 signal decreased steeply in deeper cortex layers, and was detected
up to about 0.2 mm deep. Based on the diffusion coefficient for TR-3 in brain tissue
25], and an average diffusion time of 15 min in our experiments, one would estimate that
this tracer may diffuse over an average distance of ~0.31 mm. This is more than the
average distance that we observed: ~0.064 mm. Therefore, the penetration of tracer
appears to be less than can be anticipated on the basis of diffusion. In fact, the
small penetration depth suggests that diffusion is possibly hindered by bulk flow
in the opposite direction. Thus, we suggest that in the current settings, there was
limited penetration of small solutes from the SAS into the brain by diffusion (Fig. 4), and no bulk flow from the SAS and paravascular spaces into the brain parenchyma.
This is further substantiated by the distribution of the high molecular tracer along
the paravascular spaces. Thus, F-500 showed a steep decrease in intensity along arteries
entering the brain, while a constant or even an increased concentration of dye would
be expected due to a sieving action with paravascular inflow of CSF. A more likely
explanation for the distribution of tracers over the SAS and along paravascular spaces
could therefore be the strong mixing movement in the subarachnoid space due to arterial
pulsations 26].

Bulk flow of ISF: a continental divide

The current study suggests the presence of bulk flow of ISF from the striatum toward
the ventricular system. We found no evidence for a contribution of recirculation or
re-entrance from the sub-arachnoid or para-vascular spaces to this ISF flow. At this
point, our interpretation differs from the glymphatic concept 9]. In particular, we did not observe outflow along parenchymal veins towards the subarachnoid
space. Leptomeningeal veins may have shown tracers simply because the tracers are
present in the SAS. Rather, the picture that emerges from our data is that ISF flow
is generated by the transport of fluid from the capillaries to the interstitium. In
other organs, such fluid transport is believed to be a very local phenomenon. According
to the Starling balance of hydrostatic and oncotic pressures, fluid leaves higher
pressure capillaries, close to the arterioles, and is mostly reabsorbed by venous
capillaries. The remainder is removed by the lymphatic system. In the brain, fluid
transport from capillaries to interstitium depends on active transport of ions and
solutes rather than Starling forces, and reabsorption is thought to be absent 19]. Such fluid therefore needs to find its way to the CSF compartment. This would generate
a cumulating flow, small in regions far from the ventricles (like rivulets high up
in the mountains), and much larger when approaching the ventricles. Occasionally (Fig. 6f) we observed that tracer present close to the cortical surface finds its way to
the brain surface along blood vessels. These findings resemble the observations by
Weller 10]. We therefore speculate that there is a ‘continental divide’, a point of highest
interstitial pressure where generated ISF flows to either the ventricles or the cortex.
We suggest that in the mouse this watershed is situated very close to the cortical
surface. Indeed, Arbel-Ornath 27] found that when tracers are injected in the cortex, they quickly drain out of the
brain along arteries. In this analogy, peri- and paravascular spaces could be seen
as fjords, allowing fluid transport from the mountains to the sea. It should be mentioned
that the concept of fluid generation from capillaries is controversial 28]–31], and we see the above view as a working hypothesis. Many issues remain to be addressed,
including the nature and regulation of water transport over the BBB, the amount of
generated ISF/CSF volume as compared to production by the choroid plexus, the interstitial
pressure profile, relevance for removal of toxic products, and relevance for other
species and conditions, notably for humans.

Limitations

Injection of bulk fluid may be a confounding factor when studying interstitial fluid
transport. The contribution of the infusion rate to tracer distribution is difficult
to establish. We used an infusion rate far less than many theoretical 32] and experimental studies 10], but the contribution of the pumping rate still may not be completely ignored. In
pilot studies we varied infusion rates both for the striatum and cisterna magna infusions,
and found no clear differences in tracer distribution. Even when the infusion rate
was reduced to zero, allowing only diffusion from the tip of the needle, fluorescein
(0.38 kD) was found to subsequently drain from the striatum towards the lateral ventricle.
Alternatively, increased perfusion rates and/or volumes applied to cisterna magna
infusion did not appreciably alter the distribution of the tracers (data not shown).

Finally, the use of anesthetics in the present study might have an effect on the fluid
movement. Indeed Xie L et al. showed that during anesthesia the drainage of extracellular
fluid in mice is increased compared to the awake condition, but is similar to the
results obtained during physiological sleep 17].