Glymphatic influx and clearance are accelerated by neurovascular coupling

Abstract

Functional hyperemia, also known as neurovascular coupling, is a phenomenon that occurs when neural activity increases local cerebral blood flow. Because all biological activity produces metabolic waste, we here sought to investigate the relationship between functional hyperemia and waste clearance via the glymphatic system. The analysis showed that whisker stimulation increased both glymphatic influx and clearance in the mouse somatosensory cortex with a 1.6-fold increase in periarterial cerebrospinal fluid (CSF) influx velocity in the activated hemisphere. Particle tracking velocimetry revealed a direct coupling between arterial dilation/constriction and periarterial CSF flow velocity. Optogenetic manipulation of vascular smooth muscle cells enhanced glymphatic influx in the absence of neural activation. We propose that impedance pumping allows arterial pulsatility to drive CSF in the same direction as blood flow, and we present a simulation that supports this idea. Thus, functional hyperemia boosts not only the supply of metabolites but also the removal of metabolic waste.

Extended Data Fig. 1 |. Whisker stimulation causes Ca²⁺ and hemodynamic responses in contralateral barrel cortex.

a. Adult wildtype mice were head-plated under ketamine/xylazine anesthesia and exposed to unilateral whisker stimulation (5Hz 10 ms, 10 psi, 30 s per stimulation). Macroscopic imaging was utilized to map neuronal activation (Ca²⁺ signaling) in GCaMP mice. b. The Ca²⁺ response in the stimulated and unstimulated hemisphere across whisker stimulation in Thy1-GCaMP6S mice (n=8 mice). c. Representative images of the Ca²⁺-response to three different whisker stimulation protocols (1, 5 or 10 Hz with 10 ms pulse length). The image is averaged across five stimulations. d. Ca²⁺ traces from barrel field cortex in response to 30 sec whisker stimulation protocols in Thy1-GCaMP6S mice (frequencies: 1, 5 or 10 Hz and pulse lengths: 5, 10 or 15 ms) (n=7 for 5 and 10ms, n=5 for 15ms). e. Area under the curve (AUC) of tested pulse lengths (5, 10 or 15 ms, 5 Hz) (n=7 mice). One-way ANOVA w Tukey’s correction: P=0.0097, 5 vs. 10ms; P=0.16, 5 vs. 15ms; P=0.99, 10 vs. 15ms. f. AUC of traces from Fig. 1C comparing the effect of the three stimulation protocols on Ca²⁺-signaling (left) or cerebral blood flow (right) (n=7). One-way ANOVA w Dunnett’s correction: Ca²⁺: P=0.005, 1 vs. 5Hz; P=0.17, 1 vs. 10Hz; P=0.006, 5 vs. 10Hz. CBF: P=0.04, 1 vs. 5Hz; P=0.04, 5 vs. 10Hz. g. The total hemoglobin (HbT) change in barrel cortex in response to unilateral whisker stimulation (5 Hz 10 ms, 10 psi). Regions of interest (ROI) are depicted above barrel and motor cortex in both hemispheres. h. The hemodynamic response during whisker stimulation in barrel cortex (pink and orange) and in control regions (green and blue). The gray dashed line represents the HbT response in the stimulated barrel cortex (orange), when elevating the air pressure from 10 to 20 psi. Right: AUC of the HbT in barrel field cortex, when applying air pressure of 10 or 20 psi (n=6 mice). Two-tailed paired t-test: P=0.013. i. Hemodynamic response in the same mice before application of any whisker stimulation. Data are represented as mean ±SEM. *=P<0.05, **=P<0.01.

Extended Data Fig. 2 |. CSF tracer influx is symmetric across the hemispheres in control non-stimulated mice.

a. Fluorescence signal (mean pixel intensity (MPI)) of the CSF tracer (70 kDa dextran) in each hemisphere of the control mice (n=7). b. Average fluorescent signal change across 90 seconds in left and right hemisphere of control mice receiving no whisker stimulation (n=7). Normalized according to tracer intensity at stimulation start (t=30 seconds). c. Front-tracking of the CSF tracer spread during macroscopic imaging demonstrates higher velocity in the stimulated hemisphere (n=8). Two mice from Figure 1E were not eligible for front-tracking analysis. Two-tailed paired t-test: P=0.033. d. Left: Group overlay of the fluorescent signal displaying the parenchymal influx of the CSF tracer in ex vivo brain sections spanning the barrel cortex area (anterior/posterior (A/P): 0.3 to −1.7) in control mice receiving no whisker stimulations. Each hemisphere was analyzed (regions of interest are outlined by white dotted lines). Right: Percentage distribution of tracer (3 kDa dextran) in the two hemispheres (49.7 ±1.1% vs. 50.3 ±1.1%). The total tracer signal in each brain was set to 100% (n=6). Two-tailed paired t-test: P=0.77. Data are represented as mean ±SEM. *=P<0.05.

Extended Data Fig. 3 |. Puncture of the perivascular space membrane increases CSF flow velocity.

a-c. Absolute values of the average change in pial artery diameter (A), downstream CSF flow velocity (B) and cross-stream CSF flow velocity (C) during functional hyperemia (n=7). d. The downstream velocity increases dramatically in mice with punctured perivascular space membrane (n=5). Mean CSF flow velocity: 70.5 ±3.8 μm/s.

Extended Data Fig. 4 |. CSF tracer moves from arterial to venous perivascular spaces over time.

a. Adult NG2-dsRed mice were intracisternally injected (70 kDa dextran) under ketamine/xylazine anesthesia. At 30 or 120 minutes after infusion start, mice were intracardially perfused with lectin (WGA-647) followed by 4% paraformaldehyde (PFA). b. Representative section from 120 min group (−2.4 mm bregma), inset = lateral cortex. Analysis pipeline from left to right: vessels were scored as positive for perivascular CSF tracer (yellow circles). Vessels were excluded if there was no tracer deeper than 200 μm from the brain surface (arrow heads). Using NG2-dsRed, vessels were scored as arteries (red) or veins (blue) if they had banded or diffuse dsRed signal, respectively. Finally, the number of arteries and veins with tracer was counted for each brain region. c-e. Number of vessels (C) (scored as arteries/veins as well as uncategorized), arteries (D), or veins (E) with CSF tracer at each brain subregion (left), and pooled into cortical, deep, and total (right) (3 brain sections/mouse; −1.8, −2.4, and −3.2 mm bregma) (n=3). f. Representative images of tracer distribution in arteries from lateral cortex at 30 minutes (top) and 120 minutes (bottom). WGA-647 labeling is high in arteries (arrows), but low in veins. NG2-dsRed signal is banded in arteries (arrows) and diffuse in veins. CSF tracer (70 kDa dextran) concentration in arterial perivascular spaces is high at 30 minutes (arrow, top panel), and in part phagocytosed by perivascular macrophages around arteries and veins at 120 minutes (bottom panel). g. Model diagram depicting high CSF tracer at arterial perivascular spaces at 30 minutes, and high CSF tracer at venous perivascular spaces at 120 minutes post infusion. Data are represented as mean ±SEM.

Extended Data Fig. 5 |. Tracer inflow and parenchymal spread is larger in anesthetized compared to awake mice.

a. Overlay of Figure 1E and 4D: CSF tracer signal (mean pixel intensity (MPI)) across 30 minutes in the stimulated and unstimulated hemisphere of both anesthetized (n=10) and awake mice (n=8) in response to whisker stimulations. Gray bars: 30 sec whisker stimulation. b. Parenchymal spread of CSF tracer in the cortices (ROI outlined by white dotted lines) in anesthetized (n=7) and awake mice (n=8) measured as the %Area coverage (12.6 ±1.8% vs. 4.6 ±1.5%). Two-way ANOVA with Sidak correction, P=0.0029. Data are represented as mean ±SEM. **=P<0.01.

Extended Data Fig. 6 |. Optogenetic stimulation increases total tracer influx.

a. Representative image of optogenetic stimulation above the middle cerebral artery in Sm22-Cre:Ai32-ChR2 mice. b. The individual traces from the optogenetic mice shown in Figure 5C (n=6). Blue bars: laser stimulation (30 seconds each; 10 Hz 50 ms). c. Average change in tracer signal (mean pixel intensity (MPI)) during and between stimulations in control animals (Sm22-Cre^−/−:Ai32-ChR2) (n=4). Two-way ANOVA with Sidak correction, P=0.83, stim; P=0.99, unstim. Data are represented as mean ±SEM.

Figure 1.. Functional hyperemia increases periarterial CSF influx.

a, Adult wild-type mice were exposed to unilateral whisker stimulation (30 s each), and macroscopic imaging through an intact skull was used to map neuronal activation (Ca²⁺), hemodynamic signals (IOS) or CSF tracer transport under KX anesthesia. b, Whisker stimulation (5 Hz, 10 ms) increased neuronal activity (top) and cerebral blood flow (bottom) only in the contralateral hemisphere. c, Left – Ca²⁺ traces from barrel field cortex in response to different whisker stimulation protocols (n = 7, five stimulations/mouse). Right – relative changes in cerebral blood flow (CBF) in the MCA area (n = 7, five stimulations/mouse). d, Representative image of tracer influx around the MCA. Dotted circles, ROIs. e, Fluorescence signal (MPI) of the tracer influx (70 kDa dextran; n = 10 mice). Gray bars show 30 s whisker stimulation. f, Area under the curve (AUC) across 30 min tracer influx for the stimulated (stim and unstim hemisphere, n = 10 mice; Fig. 1e) and control group (left and right hemisphere, n = 7 mice; Extended Data Fig. 2a). Two-way ANOVA w Tukey’s correction: P = 0.0705, stim vs. unstim; P = 0.0022, stim vs. left; P = 0.0006, stim vs. right. g, Change in tracer signal across all 20 stimulations per mouse. Normalized according to tracer intensity at stimulation start (t = 30 s). h, Analysis of tracer efflux from MCA branches into surrounding tissue at 15 and 30 min after whisker stimulation start. Linear ROIs are located outside the perivascular space. i, The intensity profiles of CSF tracer in the tissue surrounding the MCA. Right, AUC of the intensity profiles (n = 10). Two-way ANOVA w Sidak correction: P = 0.0133, t = 15; P < 0.0001, t = 30. j, Parenchymal CSF tracer distribution was analyzed in vibratome sections containing somatosensory cortex. k, Left – group overlay of the parenchymal influx of CSF tracer ex vivo. Dashed lines, analyzed regions. Right – percentage distribution of tracer (MPI; 3 kDa dextran) in each hemisphere (52.8 ±1.1% vs. 47.2 ±1.1%; n = 7 mice, eight brain sections/mouse). Two-tailed paired t-test: P = 0.049. l, Linear fit to the first 20 min inflow from Fig. 1e, where tracer influx dominates relative to clearance. Data are represented as mean ± s.e.m. *P < 0.05, **P < 0.01, ****P < 0.0001. See also Extended Data Fig. 1 and 2 and Supplementary Video 1.

Figure 2.. Perivascular CSF flow kinetics is closely controlled by arterial diameter.

a, Adult wild-type mice were head-plated, and a cranial window inserted above the MCA under KX anesthesia. Microspheres (1 μm) were injected into the cisterna magna before exposure to unilateral whisker stimulation. Two-photon imaging was used to map CSF flow velocity in the perivascular space of the MCA. b, The path of fluorescent microspheres in the perivascular space (top left) was tracked (top right) and used to calculate the mean flow speed (bottom). The superimposed trajectories of the tracked microspheres outline the perivascular space. c, Average downstream CSF flow velocity in the period before whisker stimulation (n = 7 mice). d, Cross-correlation of the downstream CSF flow velocity and artery diameter (left). Cross-correlation of the downstream CSF flow velocity and vessel wall velocity (right) demonstrating little phase difference. e, Artery diameter changes were measured and averaged across five linear ROIs spanning the artery lumen. The artery boundaries were automatically registered based on fluorescence intensity. f, Percentage artery diameter change (red) and vessel wall velocity (brown) during functional hyperemia (n = 7). g, Left – maximum arterial diameter change from baseline. Right – time from stimulation start to peak diameter change (Diam) and peak change in wall velocity (Wall v.; n = 7 mice). h, The time-averaged velocity direction field (green arrows) shows net transport in the downstream direction. i, Percentage change of downstream velocity (V_d) during functional hyperemia (n = 7). j, Left – maximum downstream velocity change from baseline. Right – time from stimulation start to peak velocity (n = 7 mice). k, The time-averaged velocity direction field (green arrows) shows the net transport in the cross-stream direction. l, Percentage change of cross-stream velocity (V_c) during functional hyperemia (n = 7). m, Average cross-stream velocity during baseline (t = 0-30s) and stimulation period (t = 30-60s ). Two-tailed paired t-test, P = 0.26 (n = 7 mice). Data are represented as mean ± s.e.m. See also Extended Data Fig. 3 and Supplementary Video 2.

Figure 3.. Perivascular spaces – transfer of CSF tracers and dynamic changes.

a, Adult wild-type mice were head-plated, and a cranial window was inserted above the middle cerebral artery under KX anesthesia. CSF tracer (3 kDa dextran) was injected into the cisterna magna before exposure to unilateral whisker stimulation. Two-photon imaging was used to map width changes of artery and perivascular space. The width changes were averaged across five linear ROIs. b, Artery radius and PVS width during baseline (n = 6 mice). c, Kymograph of artery and perivascular space during a 90 s protocol with 30 s stimulation. d, Change of arterial radius and perivascular space width during functional hyperemia (n = 6). e, Movement of the outer perivascular space wall. Right – baseline width and first peak (n = 7). Two-tailed paired t-test: P = 0.0015. f, Adult NG2-dsRed mice were intracisternally injected under KX anesthesia. At 30 or 120 min after injection start, mice were intracardially perfused with lectin (WGA-647) followed by fixation with 4% PFA. g, Cortical CSF tracer distribution at 30 min. Tracer is present along arteries (top, arrow) and absent from veins (bottom, arrow). h, Cortical CSF tracer distribution at 120 min. Tracer is less often present along arteries (top, arrow) and more often along veins (bottom, arrow) than at 30 min. i, Left – proportion of arteries with CSF tracer at 30 min (green) and 120 min (blue). Right – relative change in total number of arteries labeled with CSF tracer at 120 min compared to 30 min. j, Left – proportion of veins with CSF tracer at 30 min (green) and 120 min (blue). Right – relative change in number of veins with CSF tracer at 120 min compared to 30 min. For i and j, 30 min: n = 3 mice, 349 arteries, 52 veins; 120 min: n = 3 mice, 384 arteries, 241 veins. Two-tailed paired t-test: P = 0.0099. Data are represented as mean ± s.e.m. **P < 0.01. See also Extended Data Fig. 4

Figure 4.. Functional hyperemia does not increase glymphatic influx during wakefulness.

a, Adult wild-type mice were head-plated and chronically implanted with a cannula in the cisterna magna. Mice were adapted to awake head fixation before injection of CSF tracer (3 kDa dextran) and exposure to unilateral whisker stimulation during awake two-photon imaging. b, Representative images of tracer influx in awake mice during unilateral whisker stimulation (5 Hz with 10 ms pulses, 20 psi for 30 s with 60 s intervals; dashed circles–ROIs above the middle cerebral artery (MCA)). c, Tracer distribution in the whole brain after ended experiment. Images are overlaid for the entire group. Top, ventral side; bottom, lateral side receiving whisker stimulation (n = 6). d, Tracer signal (mean pixel intensity) in the stimulated and unstimulated hemisphere in response to whisker stimulations. Gray bars, 30 s whisker stimulation. Inset – zoom of tracer intensity between 20 min and 30 min. e, Comparison of tracer influx (area under curve) between awake (n = 8; Fig. 4d) and anesthetized mice (n = 10; Fig. 1e). Two-way ANOVA w Tukey’s correction: P < 0.0001. f, Left – ex vivo analysis of the parenchymal tracer distribution of brain tissue sections spanning the barrel cortex region. Tracer distribution overlaid for the whole group (n = 8). Dashed lines, analyzed regions. Right – the percentage distribution of tracer into each hemisphere (53.4 ±2.8% vs. 46.6 ±2.8%; n = 8, eight brain sections/mouse). Two-tailed paired t-test: P = 0.27. g, Mice were injected with a yellow fluorescence protein (YFP)-based chloride sensor expressed in the astrocytic cytosol 2 weeks before insertion of a chronic cranial window. Mice were injected with an i.v. tracer (70 kDa dextran) and exposed to unilateral whisker stimulation during awake two-photon imaging. Representative images of a pial and penetrating segment of the MCA and the corresponding ROIs (yellow) used for diameter analysis. Diameter change of pial (white background, n = 4) and penetrating (pink background, n = 3) segments of the MCA in awake mice during whisker stimulation. Two-tailed paired t-test: P = 0.034, pial; P = 0.054, penetrating. h, Distance between the penetrating artery and surrounding astrocytes during whisker stimulation in wakefulness (n = 4). Two-tailed paired t-test: P = 0.0161. Data are represented as mean ± s.e.m. *P < 0.05, ****P < 0.0001 See also Extended Data Fig. 5a.

Figure 5.. Arterial diameter changes and not neuronal activation per se increases glymphatic influx.

a, We developed an optogenetic mouse line (Sm22-Cre:Ai32-ChR2), where photoactivation of ChR2 in smooth muscle cells leads to arterial constriction. Adult mice were head-plated under KX anesthesia and intracisternally injected with a CSF tracer (2,000 kDa, fluorescein isothiocyanate (FITC)) before exposure to unilateral optogenetic stimulation (473 nm, 10 Hz, 50 ms; 30 s every minute) during macroscopic imaging. Control mice have no expression of channelrhodopsin-2 (Sm22-Cre^−/−:Ai32-ChR2). Right – light stimulation increases intracellular Ca²⁺ in the smooth muscle cells causing depolarization and vasoconstriction. Bottom – representative kymograph of arterial constriction across two optogenetic stimulations (black bars, 30 s optogenetic stimulation). b, Representative images of CSF tracer influx during 30 min circulation. Optogenetic stimulations were given on one hemisphere (stim) every minute for 30 s (10 Hz, 50 ms). c, Average change in fluorescent tracer signal (MPI) across 30 min of glymphatic influx with optogenetic stimulations (blue bars) (n = 6). Mean trace. d, Average signal change during and between stimulations (n = 6 mice). Two-way ANOVA w Sidak correction, P = 0.005 (laser on) and P = 0.009 (laser off). e, Top – overlay of dorsal and ventral whole-brain images from optogenetic mice. Bottom – percentage distribution of tracer in the two hemispheres (55 ±2% vs. 45 ±2%; n = 7 mice). Two-tailed paired t-test, P = 0.046. f, Linear fit to the first 20 min of inflow from Fig. 5c, where tracer influx dominates relative to clearance. The linear fits are used to estimate CSF flow rates in the perivascular space based on the tracer signal changes. Data are represented as mean ± s.e.m. *P < 0.05, **P < 0.01. See also Extended Data Fig. 6 and Supplementary Video 3.

Figure 6.. Laser stimulation in optogenetic mice increases CSF downstream velocity.

a, Left – we developed an optogenetic mouse line (Sm22-Cre:Ai32-ChR2), where photoactivation of ChR2 in smooth muscle cells leads to arterial constriction. Adult mice were head-plated, and a cranial window was inserted above the middle cerebral artery under KX anesthesia. Microspheres (1 μm) were injected into the cisterna magna before exposure to unilateral optogenetic stimulation (473 nm, 10 Hz, 50 ms; 30 s with 60 s interval) during macroscopic imaging. Middle – microspheres are traveling in the perivascular space. The laser fiber was aimed at the middle cerebral artery segment. Right – particle tracking flowmetry was applied to calculate mean flow speed of microspheres in the perivascular space. b, The average downstream CSF flow velocity between t = 0-30 s (n = 5). c, Average CSF flow velocity of microspheres in response to optogenetic stimulation (n = 5 mice). d, Average particle number within the arterial segment in response to optogenetic stimulation (n = 5 mice). e, Representative images of 1 μm microspheres moving around pial arterial branches at branching points (white arrowheads, n = 5). The image is a 30-s maximum projection showing the microsphere trajectories before any whisker stimulations. The blue arrow indicates the direction of the CSF flow, which is parallel to the blood flow. f, As the laser stimulation ceases, the arterial dilation pushes particles from the pial arterial to venous perivascular spaces (yellow arrowhead, n = 3). Data are represented as mean ± s.e.m.

Figure 7.. Functional hyperemia increases glymphatic clearance.

a, Adult wild-type mice were head-plated under KX anesthesia. CSF tracer (3 kDa dextran) was injected into the cisterna magna and circulated for 30 min before exposure to unilateral whisker stimulation during macroscopic imaging. b, Left – representative image of CSF tracer in the perivascular spaces along the middle cerebral artery. ROIs are outlined by dotted circles. Right – change in tracer signal (MPI) in the middle cerebral artery area across 30 min of unilateral whisker stimulation (n = 6). c, The change in tracer signal from t = 0 to t = 30 min. Dark pixels denote less tracer signal. The representative image was averaged across 20 frames (4 s) to improve resolution. d, Quantification of the total change in mean pixel intensity for each hemisphere across the 30 min (n = 6). Two-tailed paired t-test: P = 0.037. Data are represented as mean ± s.e.m. *P < 0.05.

Figure 8.. Impedance pumping models reproduce characteristics of flows driven by functional hyperemia and optogenetic stimulation.

a, The first model is based on arterial dilation, which we experimentally induced by unilateral whisker stimulation. b, In the first model, an arteriolar wall (red, shown in cross-section) actively dilates and relaxes in a small region (shaded red), with wall motion spreading in both directions along the artery, causing the flow of CSF (arrows) in the surrounding perivascular spaces. c, Local wall motion across one dilatory cycle (enlargement of red shaded area in b). d, The volume flow rate through one cross-section of the perivascular space (dashed line in b) varies over each dilation cycle. Flow is decreased during local dilation, with a slight lag, and increases during local relaxation. e, Over many cycles, the flow carries passive tracers to the right along the perivascular space (enlargement of region marked in b). f, The net flux (cumulative volume of fluid moving rightward through the cross-section shown in b) oscillates and increases steadily over time. Gray bars, whisker stimulations. g, The second model is based on local vasoconstriction, which experimentally is obtained by optogenetic stimulation of mice expressing ChR2 in smooth muscle cells. h,i, In our second model, the artery constricts instead of dilating, driving a flow with different spatial structures and different volume flow rate. j-l, In the constriction model, as in the dilation model, tracer moves to the right and net flux increases over time. The direction of the tracer (to the right, not left) is determined by the location of the active arterial diameter change, not by whether it is dilation or constriction. Blue bars, laser stimulations. See also Extended Data Fig. 7 and Supplementary Video 4 and 5.