Brain cerebrospinal fluid flow

Abstract

Cerebrospinal fluid flows around and into the brain, driven by intricate mechanisms, with profound implications for human health. According to the glymphatic hypothesis, in physiological conditions, cerebrospinal fluid flows primarily during sleep and serves to remove metabolic wastes like the amyloid-beta and tau proteins whose accumulation is believed to cause Alzheimer's disease. This paper reviews one research team's recent in vivo experiments and theoretical studies to better understand the fluid dynamics of brain cerebrospinal fluid flow. Driving mechanisms are considered, particularly arterial pulsation. Flow correlates closely with artery motion and changes when artery motion is manipulated. Though there are discrepancies between in vivo observations and predictions from simulations and theoretical studies of the mechanism, realistic boundary conditions bring closer agreement. Vessel shapes are considered, and have elongation that minimizes their hydraulic resistance, perhaps through evolutionary optimization. The pathological condition of stroke is considered. Much tissue damage after stroke is caused by swelling, and there is now strong evidence that early swelling is caused not by fluid from blood, as is commonly thought, but by cerebrospinal fluid. Finally, drug delivery is considered, and demonstrations show the glymphatic system could quickly deliver drugs across the blood-brain barrier. The paper closes with a discussion of future opportunities in the fast-changing field of brain fluid dynamics.

FIG. 1.

Inflow of cerebrospinal fluid to the brain, visualized with dye. The brain of a live, anesthetized mouse is seen from above, imaged through the intact skull. Dye is transported much faster than diffusion alone could do, and does not spread uniformly, instead following pathways that nearly match the shapes of arteries (though the dye is not in arteries). Experiments like these give simple evidence that the brain has a fluid transport system distinct from the cardiovascular system — the glymphatic system. Transcranial imaging courtesy of H. Mestre and M. Nedergaard.

FIG. 2.

Methods for imaging CSF flow in the brains of live mice. (a) Transcranial imaging uses an optical microscope with relatively low magnification to image the top surface of the brain, typically with scalp parted but skull intact. (b) Two-photon imaging uses a confocal, two-photon microscope, typically viewing a small region of tissue through a glass cranial window that has been installed in place of a patch of skull, while keeping the dura intact. The square overlaid on (a) indicates the approximate region we typically explore with two-photon imaging.

FIG. 3.

Measuring cerebrospinal fluid flow in the brain of a live mouse. (a) Paths of tracer particles, colored according to their instantaneous speed, closely match fluid motion in the perivascular space. The background image shows nearby blood vessels. (b) The velocity, in many square regions, averaged over time. The measured velocities have typical characteristics of a laminar flow. Two-photon imaging and particle tracking courtesy of S. Holstein-Rønsbo, Y. Gan, and M. Nedergaard.

FIG. 4.

Surface perivascular spaces are open, not filled with porous media. (a) Paths of tracer particles observed in a surface perivascular space via two-photon imaging. The thick blue line and red shapes mark one cross-section. (b) The measured velocity profile, along the cross-section marked in (a), does not match the velocity profile expected for Darcy flow in a porous medium, but closely matches the velocity profile expected for Poiseuille flow in an open space whose cross-section is fit to the measurements. (c) Particle displacements d increase linearly with time t, as expected for Poiseuille flow, not as the square root of time, as expected for Darcy flow. Here L = 40 μm and U = 68 m/s. Adapted from [31].

FIG. 5.

Surface perivascular spaces have nearly optimal shapes. (a) A surface perivascular space can be simply characterized as the region between a circular artery and an elliptical, eccentric outer wall. (b) Flow in vessels with cross-sectional shapes of this sort is fastest when elongation is neither too great nor too little. (c) Using the measured area, artery radius, and eccentricity of three different previously published perivascular spaces, we varied the elongation, calculating the resulting hydraulic resistance. We found a single minimum in each case. The calculated resistances of the observed spaces (triangles) nearly match the minima (circles). Adapted from [34].

FIG. 6.

Arterial pulsation is a primary driver of flow in perivascular spaces. (a) The instantaneous root-mean-square flow velocity pulses in synchrony with the ECG signal (which indicates cardiac activity), not with respiration. (b) Artery wall velocity closely matches fluctuations in root-mean-square flow velocity. (c) The drug angiotensin II raises the mean arterial pressure (MAP), presumably stiffening artery walls. N = 4 mice. (d) With high blood pressure, time-averaged flow in perivascular spaces (measured using two-photon imaging) is slower, and particle paths (insets) often show upstream motion. (e) Inducing high blood pressure (BP) causes a decrease in mean flow speed and an increase in the backflow fraction; both effects are statistically significant. N = 7 mice. Adapted from [30].

FIG. 7.

Dual-syringe experiments demonstrate that observed flows are not injection artifacts. (a) While injecting a solution of tracer particles with one pump, we withdrew an equal amount of fluid with another. (b) Simultaneous withdrawal and injection eliminates the increase in intracranial pressure observed when injecting without withdrawing. (c) Root-mean-square fluid velocity in the middle cerebral artery is similar in either sort of experiment. Shaded regions indicate standard error of the mean, from 6 single-injection experiments and 6 dual-syringe experiments. (d–e) Neither the mean velocity, nor the fraction of time during which the flow direction reverses, differ significantly from one sort of experiment to the other. Error bars indicate mean ± standard error. Adapted from [50].

FIG. 8.

Realistic end boundary conditions help reconcile results from experiments and simulations. (a) Simulating perivascular spaces with periodic or zero-pressure end boundary conditions predicts the flow in isolation (circuit closed by dashed line); resistance and compliance at the boundaries can approximate the effects of coupling to the rest of the flow pathway (larger circuit). (b) The simulated centerline velocity from Kedarasetti et al. [38] exhibits much larger velocity fluctuations than in vivo measurements, as well as a shifted velocity peak (compare to Fig. 6b). (c) The centerline velocity calculated from the Kedarasetti et al. results, using Eq. 3 to model coupling to the rest of the glymphatic system, exhibits much smaller velocity fluctuations and a velocity peak that occurs later in the cardiac cycle. (d) Flow velocities measured in vivo exhibit fluctuations, peak location, and overall shape that resemble the coupled prediction. Adapted from [51].

FIG. 9.

Swelling soon after stroke is due primarily to cerebrospinal fluid flow. (a) We injected 1 mm spheres to induce stroke in mice by obstructing the middle cerebral artery (MCA). (b) Transcranial imaging shows that cerebrospinal fluid (marked with dye, shown red-orange) rushes into the brain following following a spreading depolarization wave of neural activity (green), primarily in the ipsilateral hemisphere (where the sphere was injected), not in the contralateral hemisphere. (c) We located fronts separating bright from dim regions, for both dye in CSF and spreading depolarization (SD). (d) Tracking fronts showed that CSF inflow lags depolarization. (e) Tracking also showed that CSF accelerated after the depolarization propagated. (f) After depolarization, arterioles constrict, enlarging the surrounding perivascular space, which is then filled with CSF. Here, dextran dye (shown red) was injected intravenously to mark the arteriole, BSA-647 dye (shown green) was injected into the cisterna magna to mark the CSF, and neural activity (shown purple to yellow) was also visible in these genetically modified mice. (g) Particle tracking in surface PVSs shows pulsatile flow before artery obstruction, but smooth flow after, likely because flow during stroke is driven not by artery pulsation but by constriction. Adapted from [54].

FIG. 10.

Osmotic promotion of glymphatic flow might help deliver drugs to the brain. We injected dye into the cisterna magna and also injected either osmotically-neutral NaCl solution or concentrated mannitol solution. Dye entered brain tissue (increasing the mean pixel intensity, MPI) much more quickly in experiments using mannitol, as transcranial imaging shows. Adapted from [25].