Chronic sleep fragmentation impairs brain interstitial clearance in young wildtype mice

Abstract

Accumulating evidence shows that most chronic neurological diseases have a link with sleep disturbances, and that patients with chronically poor sleep undergo an accelerated cognitive decline. Indeed, a single-night of sleep deprivation may increase metabolic waste levels in cerebrospinal fluid. However, it remains unknown how chronic sleep disturbances in isolation from an underlying neurological disease may affect the glymphatic system. Clearance of brain interstitial waste by the glymphatic system occurs primarily during sleep, driven by multiple oscillators including arterial pulsatility, and vasomotion. Herein, we induced sleep fragmentation in young wildtype mice and assessed the effects on glymphatic activity and cognitive functions. Chronic sleep fragmentation reduced glymphatic function and impaired cognitive functions in healthy mice. A mechanistic analysis showed that the chronic sleep fragmentation suppressed slow vasomotion, without altering cardiac-driven pulsations. Taken together, results of this study document that chronic sleep fragmentation suppresses brain metabolite clearance and impairs cognition, even in the absence of disease.

Graphical Abstract.

Deng et al. found that chronic sleep fragmentation of 30 days significantly suppressed glymphatic influx, which was mirrored by cognitive decline. In vivo two-photon imaging in the awake state revealed that slow vasomotion rather than cardiac driven pulsations was suppressed in chronic sleep fragmented mice. Therefore, vascular dysfunction induced by sleep disturbance might impair brain metabolic waste clearance and increase the risk of developing dementia.

Introduction

Sleep is a highly conserved and physiologically necessary state across all vertebrate species.^1,2 One of the key biological functions of sleep is to promote the clearance of metabolic wastes that accumulate in brain during wakefulness.^2,3 Sleep deprivation impairs cognitive function and can induce irreversible pathologies in the brain.^4,5 Chronic sleep disturbance is often the first sign of an incipient neurodegenerative disease, ⁶ and prolonged partial sleep deprivation accelerates progression in a murine Alzheimer’s disease model.^7,8

During wakefulness, neural activity results in the accumulation of potential neurotoxic metabolic wastes such as amyloid-β (Aβ) and phosphorylated-tau (p-Tau).^9,10 The amyloid-β concentration in cerebrospinal fluid (CSF) declines during sleep and rises during prolonged wakefulness. ⁸ A positron emission tomography (PET) study showed that one-night sleep deprivation increased the Aβ concentration in the hippocampus and thalamus in healthy young humans, ¹¹ whereas obstructive sleep apnea, which is characterized by frequently disrupted sleep architecture, was associated with increased Aβ burden. ¹² We previously reported that chronic sleep fragmentation in healthy young wildtype mice increased intracellular Aβ accumulation. ¹³ We and others have proposed that the biological necessity of sleep lies in its activation of the glymphatic system, a glia-based cerebrospinal fluid – interstitial fluid exchange system present in all vertebrate species studied to date.^3,14 The glymphatic system, which functionally resembles in many aspects the lymphatic system in peripheral tissues, facilitates the export of neurotoxic metabolic wastes that accumulate in brain during wakefulness.^3,14 Numerous studies have shown that the glymphatic function is impaired in patients suffering from neurodegenerative diseases, as well as in experimental mouse models of Alzheimer’s disease and Parkinson’s disease.^{15 –19} These findings provide a framework for deciphering the interconnection between chronic sleep disturbances and neurodegenerative diseases. Growing evidence highlights the importance of arterial pulsations as the driving force for perivascular CSF flow, a key component of glymphatic function. Arterial wall elasticity and pulsatility are known to decrease with aging and in pathologic conditions such as hypertension.^{20 –22} Recent findings suggested that slow vasomotion, along with pulsatility, contributes to driving CSF flow.^23,24 However, it remains unknown how chronic sleep disturbances alter glymphatic function in young healthy animals, and whether vascular dysfunction links sleep disturbances with glymphatic decline.

Herein, we established a protocol for inducing sleep fragmentation (SF) in healthy mice, modified from our previously published protocol in which mice were awoken up to 30 times per hour during their behaviorally inactive light phase.^13,25 We tested the hypothesis that chronic sleep fragmentation impairs glymphatic clearance in brain of healthy rodents, and that slow arterial vasomotion is a key mediator of this causal association. The study has translational implications regarding the importance of optimizing therapies for chronic sleep disturbances.

Materials and methods

Animals

Wildtype male C57BL/6 mice (aged 2–3 months, weight between 25–30 g) were acquired from the institutional experimental animal centers (Shanghai and Hubei, China). Mice were group-housed in a 12:12 light/dark cycle with ad libitum access to food and water. All experiments were approved by the Institutional Animal Care and Use Committees of Fudan University, Shanghai, China and Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Hubei, China. All studies were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. We report all animal data in compliance with the ARRIVE guidelines 2.0, and made efforts to minimize the number of experimental animals.

Sleep fragmentation modeling

We adopted a sleep fragmentation (SF) intervention with modifications from previously published protocols.^5,26,27 In brief, mice in the SF group were housed in their home-cage, which was placed on an orbital rotor (Shiping, Shanghai) programmed to shake at 110 rpm for 10 sec every 2 minutes during the light-on phase (8:00 am–8:00 pm, namely ZT 0–12). The SF intervention lasted for three (SF 3 days) or 30 consecutive days (SF 30 days); the control mice were undisturbed across sleep-wake cycles. We conducted behavioral experiments and measurements of glymphatic function between ZT 5–12 on the day following the cessation of SF interventions. We have established the patency of this protocol in our previous studies.^13,25 Experimenters performed daily checks to insure the availability of food and water, taking care to avoid unnecessary disturbance of the mice. There was no mortality in any of the experimental groups, including mice exposed to chronic sleep fragmentation.

Cisterna magna injection

We performed cisterna magna (CM) cannulations as described previously. ¹⁴ In brief, the day before the measurements of glymphatic influx, mice were implanted with a closed-end PE10 tube into CM under anesthesia with 1.0–2.0% isoflurane inhalation. On the day following the cessation of SF interventions, mice received a CM infusion of 10 μl 1% (weight/volume) fluorescein-conjugated dextran (FITC, 40 kDa, Thermofisher D1845) dissolved in artificial cerebrospinal fluid (aCSF, NaCl 124 mM, KCl 3 mM, NaHCO₃ 26 mM, NaH₂PO₄ 1.24 mM, MgSO₄ 2 mM, CaCl₂ 2 mM, Glucose 10 mM), with an infusion rate of 1 μl/min using a microperfusion pump (BASi, USA). Allowing 20 minutes for circulation, mice were euthanized with 80 mg/kg of body weight pentobarbital sodium for transcardial perfusion with PBS and 4% paraformaldehyde (PFA). Brain samples and the deep cervical lymph nodes (dCLNs) were rapidly removed and post-fixed by immersion overnight in 4% PFA. The tissues were then dehydrated by immersion solutions of increasing sucrose concentration (10, 20, and 30%), each for approximately 24 hours at 4 °C. Coronal frozen brain sections (thickness 100 μm) and lymph node sections (thickness 30 μm) were prepared with a cryostatic microtome (Leica, Germany) at −20 °C. Sections were slide-mounted with glycerin and kept in the dark until imaging.

Preparation of cranial window

We adopted our surgical protocol for preparing a cranial window from previous literature.^20,28 We anesthetized the mice with pentobarbital sodium (50 mg/kg body weight, i.p.), retracted the scalp and removed the fascia on the superior cranial surface using a razor blade. We next glued a metal plate on the skull and exposed the right hemisphere. We drilled a cranial window of 4 mm diameter (centered at bregma −1.5 mm, lateral 2.0 mm) over the somatosensory cortex and kept dura matter intact. The cranial window was covered with a glass coverslip and secured with superglue (Ergo5800, Kisling, Switzerland). Mice were allowed at least post-surgery two weeks’ recovery prior to in-vivo imaging.

In vivo two-photon imaging

We used two-photon microscopy to measure pulsatility of cerebral blood vessels in somatosensory cortex in vivo through the chronic cranial window as previously described.^22,23 Before the in vivo imaging experiments, the mice were habituated to the imaging setup for 20–40 min per day for five consecutive days. Applying the criteria from the prior study, ²³ we excluded vessel diameter changes exceeding 45% of the baseline measurements as likely motion artefacts. On the imaging day, fluorescein isothiocyanate–dextran (FITC–dextran, 70 kDa, 1% w/v, Sigma, USA) dissolved in 150 μl saline was infused into the tail vein 10 minutes prior to the imaging. We used a two-photon microscope (Olympus, Japan) equipped with a 25× water-immersion lens, with excitation wavelength set at 920 nm. For pulsatility measurements, we imaged the transverse sections of the blood vessels at a depth of 80–120 μm from the pial surface in the line scan (x-t) mode at 500–800 Hz for 5000 cycles. On average, 7–20 blood vessels were imaged per mouse. We performed x-t line scans orthogonal to the vessel axis to measure vascular diameters, which we recorded and plotted as a function of time. Vascular pulsatility (in units of μm*ms) was defined as the absolute value of the integral (area under the curve) of the vascular diameter per millisecond epoch. For measuring spontaneous vasomotion, we viewed pial arteries and arteriolar branches by 2x magnification and imaged during 310 s at a speed of 2 μs/pixel, pixel resolution of 0.994 μm (256 × 256), and frame-rate of ∼2.4 Hz (to a total of 750 frames). We identified the arteries by their morphology and direction; arteries are straight and usually have fewer branches than veins. Using FIJI and Matlab custom-written scripts, we drew three to six regions of interest (ROIs) on three to six arteries per field of view (FOV). We then measured the vessel segment size for each frame (defined as the number of pixels per ROI occupied by the vessel) and calculated the mean vessel segment size over all frames. To calculate the percentage of changes in vessel diameters, we divided the difference between vessel segment size for each frame and the mean vessel segment size of all frames by the mean vessel segment size. Application of Fourier transform for the percentage changes of vessel diameters yielded the power spectral density (PSD) of the cyclic vessel diameter oscillations in the frequency range of 0.02–1.2 Hz. When making comparisons at specific frequencies (e.g., 0.1 Hz), we averaged and compared the amplitudes of the PSD for the frequencies of interest from all ROIs per mouse. For two-photon imaging in vivo, we imaged the diameters of the microvessels, and excluded from analysis any vessels with diameter less than 10 μm. The edges of the vessels were identified with the line tool in Fiji software to indicate the turning points of fluorescence intensity. The vessels with diameter (10–60 μm) were included for quantification of pulsatility and slow vasomotion.

EEG/EMG electrodes implantation and recording

We implanted electroencephalography (EEG) and electromyography (EMG) electrodes as the protocols described previously. ²⁹ In brief, mice were anesthetized with pentobarbital sodium (50 mg/kg, i.p.). Two stainless steel screws (1 mm diameter) were implanted on the top of the skull (anteroposterior, +1.0 mm and left-right, −1.5 mm from bregma or lambda) as EEG electrodes. Next, two Teflon-coated silver wires inserted in the neck muscle served for EMG electrodes. The four electrodes were attached to a microconnector, which was fixed to the skull with dental cement. After at least two weeks’ post-surgery recovery, we performed SF interventions and started EEG/EMG continuous recording at ZT 12 of last day of SF modeling. We analyzed the power spectrum of EEG signals between ZT 5–12 on the following day. EEG/EMG signals was amplified, bandpass filtered (EEG, 0.5–30 Hz; EMG, 40–200 Hz), digitized at 128 Hz, and recorded with Vital Recorder software (Kissei Comtec). EEG signals were divided into the traditional power bands: delta (0.5–4 Hz), theta (4–8 Hz), alpha (8–13 Hz) and beta (13–20 Hz). The power at each frequency was analyzed by the fast Fourier transform procedure in SleepSign software (Kissei Comtec, Nagano, Japan).

Behavioral testing

For all the behavioral experiments, mice were habituated in the behavioral room for an hour before testing.

Open field test

The open field test (OFT) was conducted as described previously. ¹³ Each mouse was placed in the center of the OFT box made with white opaque Plexiglas (25 × 25 × 28 cm), and allowed for free exploration for 5 minutes. The trajectories were recorded by video monitoring and analyzed with the AnyMaze software (Stoelting Co., USA).

Novel object recognition

We conducted the novel object recognition (NOR) test with modifications from a published procedure. ³⁰ In brief, for familiarization, on the first day we placed the mice in the open field containing two identical red cylindrical objects (A1 and A2) and allowed for 10-minute free exploration. The following day, we replaced one of the objects with a green cube of similar dimensions (novel object), with the other red cylinder (familiar object) in place, and then placed the mouse for 5-minute free exploration. We manually scored time spent in contacting the novel and familiar objects. We regarded either staring at or sniffing the object as “active exploration”, and disregarded episodes of sitting or sleeping next to the object or climbing on it, or any explorations lasting less than 2 seconds.

Immunofluorescence staining

Brain samples were harvested after transcardial perfusion fixation with 4% PFA and post-fixation followed by dehydration with a sucrose gradient as described above. We prepared coronal brain cryosections of 20 μm thickness. After permeabilization with 0.5% Triton X-100 (Sigma-Aldrich, USA), sections were incubated with 5% bovine serum albumin dissolved in PBS for 1 hour at room temperature. The incubations with primary antibodies (rabbit anti-AQP4, 1:200(Protein tech, China) and rat anti-CD31, 1:200(BD Pharmingen, China)) were conducted at 4 °C for 16 h. Secondary antibody (Alexa594-conjugated donkey anti-rabbit IgG, 1:500; Abcam, USA, Alexa488-conjugated donkey anti-rabbit IgG, 1:1000, or and Alexa555-conjugated Goat anti-rat IgG, 1:500; Abcam, USA) incubations were conducted for 1 hour at room temperature. Slices were then sealed using 50% glycerin in PBS. Images were acquired at 10× and 20× magnification with a fluorescence microscope (Olympus, Japan). Representative images were acquired at 20× or 40× magnification with laser confocal microscope (NIKON, Japan).

Meninges dissection and immunostaining

Meninges were dissected following previously published protocols.^31,32 In brief, mice were euthanized with pentobarbital sodium (80 mg/kg body weight, i.p.) and transcardially perfused with 0.1 M of PBS for 5 min. We carefully removed the top of the skull and immersed it, along with the meninges attached on the inner cranial surface, in 4% PFA for 24 hours at 4 °C. The fixed meninges were then dissected from the skullcap and after gently rinsed with PBS, incubated with PBS containing 5% bovine serum albumin and 0.3% Triton-X-100 for 1 hour at room temperature, followed by incubation with appropriate dilutions of primary antibody (Anti-LYVE1, Rabbit IgG, 1:200; Abcam, USA) overnight at 4 °C. The whole-mount of meninges from each mouse were rinsed three times in PBS, and then incubated with the secondary antibody (Alexa488-conjugated donkey anti-rabbit IgG, 1:1000; Abcam, USA) for 1.5 hours at room temperature. Finally, the meninges were sealed between two coverslips with ProLong™ Gold Antifade Mountant (Thermofisher, USA) The LYVE1 (+) regions were selected with the threshold tool in the Fiji software and normalized to the total area of the sample, as described previously. ³²

Quantitative analysis of fluorescent CSF-tracer distribution

For each brain sample, eight brain slices from different coronal levels were selected for imaging with a fluorescence microscope (Olympus BX5, Japan) following a consistent imaging protocol. Equivalent coronal positions were used for all biological replicates. Tracer influx was quantified by a blinded investigator using Fiji/Image J software (NIH, USA). The target brain regions, cerebral cortex, subcortex (including caudate putamen, lateral septal nucleus, thalamus and hypothalamus) and hippocampus were manually outlined in each slice; the mean fluorescence intensity within these ROIs and the positive area fractions under a constant threshold was averaged among the eight brain slices. Data from the sleep fragmentation group were normalized to the averaged readouts of the control group.

Evaluation of AQP4 depolarization

Based on the images acquired from AQP4 immunofluorescence staining, the AQP4 polarization pattern around vasculatures was quantified as described previously. ²¹ In brief, a rectangular ROI was placed perpendicular to the vessel axis, with the mid-line matching the center of vessels measuring 10–20 μm in diameter. We plotted fluorescence intensity of AQP4 along the long axis of the ROI. To calculate the AQP4 polarization index, we calculated the ratio between the peak intensity of the vascular endfeet and the average intensity in parenchyma. Alternatively, we plotted the AQP4 fluorescence intensity as a function with the distance away from the edges of the vasculature. The slope of the declining intensity served to evaluate the dispersion of AQP4 protein distribution around the vessels. We randomly selected approximately 50 cortical blood vessels per mouse for quantification. The measurements were conducted with Fiji software (NIH, USA).

Exogenic amyloid-β1-40 clearance evaluation

We measured the clearance of exogenous human Amyloid-β1-40 in awake mice following a published protocol with minor modifications. ³³ Mice were pre-implanted with cannula in the prefrontal cortex (AP 1 mm, ML 2 mm) at a depth of 2 mm below the pial surface. At 24 hours post implantation 1 µl of a 10 nmol/L solution of human amyloid β1-40 in aCSF was infused at a rate of 0.1 µl/min. After allowing diffusion of the human Amyloid β1-40 for 20 min, the mice were euthanized with pentobarbital sodium (80 mg/kg body weight, i.p.) and the brains were collected. The infused cerebral hemisphere was homogenized with an automatized homogenizer, and the total amount of residual Amyloid β1-40 was measured with a human Aβ ELISA kit (CSB-E10684h, Cusabio, China).

Statistical analysis

All statistical analysis was performed on GraphPad Prism 9 (GraphPad Software). Data distribution normality was examined by the Shapiro-Wilk test. For comparisons between two groups, two-tailed unpaired Student t tests were performed when datasets followed a parametric distribution, and Mann–Whitney tests were performed if the normality assumption was violated. To compare the power density of EEG, we used two-way ANOVA test. P < 0.05 was considered statistically significant. All data are represented as boxplots with min/max whiskers or mean ± standard deviation (SD).

Results

Sleep fragmentation induces cognition decline and anxiety-like behaviors

To model sleep fragmentation (SF), the mice were exposed to a repeated cycle of cage shaking (10 sec on, 110 sec off) during the light-on phase continuously for either three days (SF 3 days) or 30 days (SF 30 days). We compared the SF mice after their last exposure with their condition-matched control group (Figure 1a). We have previously found that chronic sleep fragmentation induces significant cognitive deficits to the Morris water maze (MWM) and novel object recognition (NOR) tests.^13,34 Herein, NOR test consistently revealed that SF 3 days and SF 30 days interventions both resulted in a significant reduction in the time fraction spent in exploring the novel object in comparison with respective controls (Figure 1b–d). The open field test showed that both sleep fragmentation groups showed comparable total distance traveled with their control groups (Figure 1e–g). Sleep fragmentation groups showed reduced entering of the central zone, suggesting anxiety-like behavior (Figure 1f). In summary, SF 3 days and SF 30 days treatment both induced cognitive impairment in young adult wildtype mice.

Figure 1.

Sleep fragmentation induces cognition decline and anxiety-like behaviors. (a) The experimental design of the study and a schematic illustration of the sleep fragmentation (SF) model. (b) Schematic illustration of the novel object recognition test. (c–d) The exploration time spent with the familiar and novel objects (left) and the fraction of time spent on the novel object (right) on the test day are compared between control versus SF 3 days (c), and between control versus SF 30 days (d). (e) Schematic illustration of the open field test. (f–g) The distance travelled (left), times of entering central zone (middle), and the percentage of distance travelled in central zone (right) are compared between control versus SF 3 days (f), and between control versus SF 30 days (g). Comparisons of the times of entering the central zone and the percentage of distance in central zone used Mann Whitney test, while the other comparisons used unpaired t-test. Data are presented as boxplots with all datapoints. ****P<0.0001, ***P < 0.001, **P < 0.01, *P < 0.05.

Three days of sleep fragmentation does not alter glymphatic function

During the light-on phase between ZT 5–12, while the mice were primarily inactive, glymphatic influx was higher as compared with the light-off, active phase. ² To assess the effects of SF 3 days on glymphatic function, we implanted a cannula in the cisterna magna on the last day of SF. Twenty-four hours later, while awake, the mice were infused with fluorescence tracer into the cisterna magna (Figure 2a). There were no statistical differences in tracer signal intensity and area fractions in cortex, subcortex, hippocampus, or the whole brain regions between the SF 3 days group and control group (Figures 2b–d). Neither did the mean intensity of fluorescence and area covered by tracer in neck lymph node exhibit significant differences between the two groups (Figures 2e–g). An evaluation of clearance of exogenous human Amyloid-β1-40 infused into the prefrontal cortex revealed a non-significant trend towards increased clearance in the SF 3 days group as compared to normal sleep controls (Figure 2h–i). In summary, brain glymphatic function was unaffected by acute sleep fragmentation lasting three days.

Figure 2.

Acute three days SF does not alter glymphatic function. (a) Schematic figure indicating the cisterna magna (CM) injection of fluorescence tracer (FITC-Dextran, 40 kDa), quantification of mean intensity and area fraction of the glymphatic tracer influx in the brain slices (Lower). (b) Representative images of fluorescence tracer influx in rostral-to-caudal brain slices (left) and tracer influx in the cortex (right boxes). Scale bar, 1000 μm. (c–d) Brain regional fluorescence intensity (c) and fluorescence tracer distributing area Continued.fraction (d) are compared between control vs SF 3 days mice. Data were normalized to the average of the control group. Brain regions include whole brain (whole), cortex (Ctx), subcortex (SubCtx), and hippocampus (Hippo). Comparison of the normalized area fraction of subcortex in panel D used Mann Whitney test, all the other comparisons used unpaired t test. (e) Schematic and representative images of tracer influx into deep cervical lymph nodes (dCLNs), scale bar, 200 μm. (f–g) Fluorescence intensity (f) and fluorescence tracer distributing area fraction (g) are compared between control vs SF 3 days mice. Data were normalized to the average of the control group. Unpaired t test was used. (h) The schematic illustration of measuring the clearance efficacy of exogenic human Amyloid-β1-40 in awake mice. (i) The concentration of human Amyloid-β1-40 (unit: ng/ml) measured in the brain tissue after 20 min circulation time are compared between control versus SF 3 days. N = 5 for each group. Unpaired t test was used. (j) Representative images of meninges that were counterstained with DAPI (blue) and LYVE1 (green) of control and SF 3 days mice. Scale bar, 1 mm and (k) The area fraction of LYVE1 (+) regions normalized with the area size of the meningeal territory. N = 4 for each group. Mann Whitney test was used. Data are presented as boxplots with all datapoints. ns, not significant.

Chronic sleep fragmentation suppresses glymphatic function

On the day after the SF 30 days intervention, glymphatic function was evaluated as above. Macroscopic imaging of whole brains showed that the CSF tracer fluorescence was lower in SF 30 days mice versus normal sleep controls (Figure 3a). An analysis of CSF tracer influx in the coronal brain sections with standard methods,^14,35 revealed that the mean tracer intensity and influx area in brain slices extending from rostral to caudal was significantly lower in SF 30 days mice as compared with normal sleep controls (Figure 3b–d). This reduction was significant in subregions of cortex and subcortex, but not in hippocampus (Figure 3c–d). The perivascular fluorescence tracer distribution around the cortical penetrating vessels was suppressed by SF 30 days (Figure 3b, zoom-in view). CSF tracers are partly exported from cisterna magna to the deep cervical lymph nodes (dCLNs).^32,36,37 Ex vivo analysis of the lymph nodes showed that tracer efflux to the dCLNs was suppressed in SF 30 days mice compared with controls (Figure 3e–g). In addition, the clearance of exogenous infused Amyloid β1-40 in prefrontal cortex was significantly decreased in SF 30 days versus normal sleep control mice (Figure 3h–i). In summary, SF for 30 days significantly suppressed glymphatic influx into the brain, CSF efflux via the dCLNs and Aβ1-40 clearance from prefrontal cortex, suggesting an overall impairment of CSF transport.

Figure 3.

Chronic sleep fragmentation suppresses glymphatic function. (a) Schematic figure indicating the cisterna magna (CM) injection of fluorescence tracer (FITC-Dextran, 40 kDa) and quantification of mean intensity and area fraction of the glymphatic tracer influx in the brain slices (Upper). Representative images of the whole brain tracer distribution under the macroscopic imaging for SF 30 days and normal control mice (Lower). (b) Representative images of fluorescence tracer influx in rostral-to-caudal brain slices(left) and tracer influx in the cortex (right boxes). Scale bar, 1000 μm. (c–d) Brain regional fluorescence intensity (c) and fluorescence tracer distributing area fraction (d) are compared between control vs SF 30 days mice. Data were normalized to the average of the control group. Brain regions include whole brain (whole), cortex (Ctx), subcortex (SubCtx), and hippocampus (Hippo). Unpaired t test or Mann Whitney test was used. (e) Schematic and representative images of tracer influx into deep cervical lymph nodes (dCLNs), scale bar, 200 μm. (f–g) Fluorescence intensity (f) and fluorescence tracer distributing area fraction (g) are compared between control vs SF 30 days mice. Data were normalized to the average of the control group. Unpaired t test was used. (h) The schematic illustration of measuring the clearance efficacy of exogenic human Amyloid-β1-40 in awake mice. (i) The concentration of human Amyloid-β1-40 (unit: ng/ml) measured in the brain tissue after 20 min circulation time are compared between control versus SF 30 days. N = 5 for each group. Unpaired t test was used. (j) Representative images of meninges that were counterstained with DAPI (blue) and LYVE1 (green) of control and SF 30 days mice. Scale bar, 1 mm. (k) The area fraction of LYVE1 (+) regions normalized with the area size of the meningeal territory. N = 5 for each group. Unpaired t-test was used. Data are presented as boxplots with all datapoints. **P < 0.01, *P < 0.05, ns indicating not significant.

Sleep fragmentation did not alter meningeal lymph vessels

We labelled meningeal lymph vessels with antibodies against the lymphatic marker, LYVE1. The area fraction of the LYVE1 positive region in SF 3 days mice was comparable with controls (Figure 2j–k), while the area fraction in SF 30 days mice showed a trend towards reduction as compared with controls (Figure 3j–k). Therefore, SF did not significantly alter area of meningeal lymph vessels.

Chronic sleep fragmentation suppressed slow vasomotion but not cardiac pulsatility

Vascular pulsatility driven by the cardiac cycle is an important driver of periarterial CSF inflow,^{20 –22} which is both reduced in animal models of normal aging and hypertension.^20,21 Our present assessment of vascular cardiac-driven pulsatility in awake mice through a chronic cranial window did not reveal significant changes in SF 30 days mice as compared with controls (Figure 4a, c). Intrinsic slow vasomotion and functional hyperemia are additional drivers of brain fluid clearance.^{23,38 –41} The frequency of slow vasomotion in absence of external stimuli falls within a broad range of frequencies centered around 0.02–0.1 Hz in the non-anesthetized human, mouse, and rat brains.^24,37 We evaluated oscillations of the arterial wall by Fourier transform analysis and plotted the range of 0.02–1.2 Hz (Figure 4d). The power spectrum density of frequency bands (0.05–0.15 Hz, low) was significantly lower in mice exposed to SF 30 days compared with controls (Figure 4f). Comparing the amplitude of the power spectrum density at 0.05 and 0.1 Hz showed a significant reduction in amplitude at 0.1 Hz in SF 30 days mice versus control mice (Figure 4h, j). The data thus suggest that sleep fragmentation for 30 days drastically suppressed spontaneous vasomotion, but not cardiac cycle-related pulsatility, in cortical arterial vessels in awake mice.

Figure 4.

Chronic sleep fragmentation alters the brain vascular function. (a) Schematic diagram of in vivo two-photon imaging of vascular pulsatility of cortical vessels with line scan mode. Linear ROIs were applied in the traverse sections of the blood vessels at depth of 80–120 μm from pial surface and scanned with line scan (x-t) mode. Vascular pulsatility (in units μm*ms) was defined as the Continued.absolute value of the integral of the vascular diameter per millisecond. (b–c) Vascular pulsatility are compared between control versus SF 3 days (b, N = 5 for each group), and between control versus SF 30 days (c, N = 4 versus 5). Unpaired t test was used. (d) Schematic diagram of evaluating vascular vasomotion with two-photon imaging. Representative images acquired with image scan mode (xy-t) are displayed in somatosensory cortex in awake mice (left). Rectangle ROIs (dash line) were applied on the selected arterial segments to calculate the area size (middle). The normalized area size of ROI was plotted against time course indicating the temporal dynamics of the vessel constriction and dilation in the selected arterial segment (right). (e–f) The time course data of the normalized area size of ROIs were processed with fast Fourier transform (FFT) analysis. The power spectrum density is plotted against the arterial spontaneous oscillation frequency (range, 0.02–1.2 Hz) for control versus SF 3 days (e), and for control versus SF 30 days (f). The inserts showed the enlarged view of the frequency range 0.02–0.2 Hz (dash box). The frequency range was divided into four bands, very low (0.02–0.05 Hz), low (0.05–0.15 Hz), high (0.15–0.60 Hz), and very high (0.60–1.2 Hz). (g–h) Histograms summarized the power density of very high, high, low and very low frequency bands between control and SF 3 days (g), and between control and SF 30 days (h). Multiple t-tests with Holm-Šídák method were used. (i–j) The amplitude at 0.1 Hz and 0.05 Hz in the power spectrum density are compared for both groups. In total, n = 30 vessels from five SF 3 days mice and n=28 vessels from five control mice were included (i); in total n = 23 arteries from five SF 30 days mice and n = 20 arteries from four control mice were included (j). Unpaired t tests were performed for the comparisons at F = 0.05 Hz, and Mann Whitney test for the comparisons at F = 0.1 Hz. Data are presented as mean ± SD (gray or green region) in (e–f) and boxplots with all datapoints in b–c and g–j. *P < 0.05, **P < 0.01, ns indicating not significant.

Acute sleep fragmentation did not alter slow vasomotion and cardiac pulsatility

Vascular pulsatility and slow vasomotion did not differ between SF 3 days mice and normal sleep controls (Figure 4b). Slow vasomotion did not differ between SF 3 days and controls across all frequency bands (Figure 4e, g). The amplitudes of the frequency spectrum at 0.05 or 0.1 Hz were similar between SF 3 days and controls groups (Figure 4i). The data suggests that sleep fragmentation of three days does not alter vascular function as regards to cardiac pulsatility or slow vasomotion in awake mice.

Vascular polarization of AQP4 is only mildly reduced by chronic sleep fragmentation

Aquaporin (AQP4) water channels are primarily localized in perivascular astrocytic processes plastered around the cerebral vasculatures. ²¹ Using immunostaining, we measured the declining slope of AQP4 fluorescence signal as a function of the distance from the vessel wall. We found a slight but significant reduction in the declining slope in SF 30 days mice versus controls, suggesting that AQP4 protein expression was less polarized around the vascular wall after 30 days of SF (Figure 5a–c). We saw a non-significant trend towards reduced ratio of fluorescence intensity between the perivascular area and neighboring parenchymal regions in SF 30 days mice (Figure 5c). Total AQP4 protein level in brain homogenates was not altered by SF 30 days (Figure 5d; Supplementary Figure 1). We purified cerebral vessels to measure the perivascular AQP4 expression (Figure 5e; Supplementary Figure 1), ⁴² finding only a trend towards a reduction in vascular AQP4 in SF 30 days mice (Figure 5e). Thus, the modest changes in AQP4 vascular polarization suggest that reactive changes in astrocytes are not a major factor in the suppression of glymphatic transport observed in SF 30 days mice. This conclusion is supported by the previous finding that GFAP expression is not upregulated by 30 days of SF. ²⁵

Figure 5.

Chronic sleep fragmentation slightly alters the perivascular AQP4 distribution and enhances theta prevalence. (a) Representative images of AQP4 and CD31 immunostaining in cortex of control and SF 30 days mice. The right schematic displays the change of intensity of AQP4 fluorescence along with the line extending perpendicular from the vascular wall to the parenchyma. Scale bar, 100 µm. (b) The decay slope of AQP4 fluorescence intensity change within 10 µm distance was calculated as the formula: slope = Δ F/10, as the insert in (b). N = 5 mice per group, in total 50 vessels per animal. AU, arbitrary unit. (c) Comparison of the decay slope of AQP4 fluorescence intensity (left) and AQP4 polarization index (right) between control and SF 30 days mice. Unpaired t-tests were used. (d–e) Quantification of the amount of AQP4 protein of the whole brain or perivascular AQP4 protein in SF 30 days and control mice by western blot. Relative quantification of AQP4 with β-actin in whole brain tissue (d) and in the perivascular tissue (e) is compared between control and SF 30 days mice. N = 5 mice per group, unpaired t tests were used. (f) EEG power density are compared between control and SF 30 days mice during ZT 5–12. Black line indicates the frequency band range showing significant difference. (g) A diagram of EEG power bands division is displayed (upper panel). Prevalence of delta, theta, alpha and beta power bands are compared between control and SF 30 days mice (lower panel, N = 5 versus 6). Unpaired t tests and two-way ANOVA test were used. Data are presented as mean ± SD (gray or green region) in b, f and boxplots with all datapoints in c, d, e, g. **P < 0.01, *P < 0.05, ns indicating not significant.

Chronic sleep fragmentation enhances theta power at the time of glymphatic suppression

To assess the neural activity in the time window for measuring glymphatic function, we recorded EEG and EMG signals after the cessation of the SF 30 days treatment. Previous studies have shown that anesthesia increases glymphatic influx as a function of EEG delta prevalence, but with a weak trend towards a negative correlation for theta prevalence. ³⁵ Present results for the EEG power spectrum density recorded at ZT 5–12 showed that theta prevalence was significantly higher in SF 30 days mice than in normal sleep controls. Delta prevalence and other power bands did not differ significantly (Figure 5f–g).

Sleep fragmentation does not increase blood brain barrier permeability

We next test if the blood brain barrier (BBB) is compromised in sleep fragmentation mice by using two alternative approaches to measure influx of Evans Blue across the BBB. Lipopolysaccharide (LPS)-treated animals were used as a positive control (Supplementary Figure 2). The analysis showed that neither SF 3 days nor SF 30 days treatments altered Evans Blue uptake relative to control mice, indicating a lack of effect of acute and chronic sleep deprivation on BBB permeability.

Discussion

In this study, chronic sleep fragmentation in mice lasting 30 days led to cognitive impairments and anxiety-like behaviors, consistent with prior reports^13,43 (Figure 1). Focusing on glymphatic transport, we found that chronic sleep fragmentation significantly decreased CSF influx into the brain, as well as CSF efflux to the cervical lymph nodes, pointing to a global suppression of brain fluid flow (Figure 3). To explore the underlying mechanisms of this phenomenon, we examined several factors known to drive glymphatic fluid transport. In our two-photon microscopy evaluation, SF 3 days or SF 30 days treatments were without effects on cardiac-driven arterial pulsatility. In contrast, we saw a striking dampening of slow vasomotion in SF 30 days group relative to their control group, but not in the SF 3 days group. (Figure 4). To our knowledge, this is the first study documenting that chronic sleep fragmentation potently suppresses glymphatic fluid transport in the absence of an underlying neurological disorder. This observation has important clinical implications, potentially identifying the missing link between and its association with increased risk of dementia.^{44 –47}

Sleep fragmentation generates consistent outcomes in human and rodent models

Sleep fragmentation differs from sleep restriction (shorter sleep duration) and sleep deprivation (extended wakefulness) with respect to the frequent awakenings from sleep. ⁴⁸ Sleep fragmentation has been modeled experimentally in humans^49,50 and rodent models.^5,26 Due to ethical and safety considerations, human sleep loss studies rarely extend for more than three days.^49,50 In an MRI study, one night of sleep deprivation suppressed glymphatic clearance in human brain. ⁵¹ Across the various approaches to study sleep fragmentation in humans, most results show that sleep fragmentation evokes increased daytime sleepiness, decrements in attention, slowed reaction time, less cognitive flexibility, decreased working memory, and impaired mood in healthy human subjects.^{49,50,52 –55} Periodic or randomized rotating chambers such as the model adopted in this study have been used to study the behavioral and pathological outcomes of sleep fragmentation in rodents.^5,26 Similar to human findings we found that the SF 3 days intervention resulted in impairments of memory and induced anxiety-like behaviors (Figure 1). Other have shown that mice with chronic sleep fragmentation developed cognitive impairments, reduced catecholaminergic and orexinergic neuronal projections, glial activation, and elevated hypercapnic arousal threshold,^5,25 albeit with total sleep duration and circadian distribution of sleep comparable with control values, possibly due to adaptive homeostatic processes.^5,26 Two nights of sleep fragmentation impaired insulin sensitivity in healthy human subjects, ⁵⁶ while in wildtype mice, nine days of sleep fragmentation in combination of high fat diet impaired glucose tolerance. ⁵⁷ Several weeks of sleep fragmentation in mice altered their metabolic pattern by enhancing food intake behaviors, abdominal adipose tissue accumulation, and insulin resistance.^58,59 In our previous PET imaging study, we found increased cerebral uptake of the glucose analogue ¹⁸F-fluorodeoxyglucose (FDG) in SF 30 days mice compared with controls. ²⁵ We also saw hippocampal transcriptomic changes in SF 30 days mice closely resembled findings in other models of chronic stress. ²⁵ Thus, based on the existing literatures, acute sleep fragmentation provokes comparable behavioral outcomes in healthy human subjects and wildtype mice.

Sleep fragmentation affects the key mediators of glymphatic fluid transport

The cardiac cycle and slow vasodynamics are the two major drivers of glymphatic transport. These drivers together initiate low-pressure gradients that propel peri-arterial CSF inflow to the brain parenchyma. The driving forces of glymphatic transport have been visualized by MRI imaging of human brain and imaging of live rodent brain.^{20,39,40,60‒62} One of the most striking present observations is that sleep fragmentation of 30 days suppressed slow vasomotion in mouse brain (Figure 4). In mice, slow vasomotion has much lower frequency (∼0.1 Hz) than cardiac pulsatility (∼5–10 Hz), but slow vasomotion evokes much larger amplitude changes in vascular diameter (10–15%) compared to cardiac pulsation (2–4%).^20,23,37 In human and rodent brains, slow vasomotion falls within a broad range of frequencies centered around 0.1 Hz. ²⁴ An elegant human study combining ventricular flow imaging with recordings of EEG and fMRI BOLD signals revealed that CSF pulsation (∼0.05 Hz) anticorrelated with cerebral blood volume (CBV) dynamic changes and CSF flow in the 4th ventricle during NREM sleep. ³⁹ Autonomic regulation of vascular tone is another factor potentially contributing to slow vasomotion. ⁴¹ Furthermore, inspiration can drive CSF flow in the subarachnoid space, the ventricles ⁶³ and the aqueduct of Sylvius. ⁶³ Respiratory pulsations (∼0.3 Hz in mice) drive a low-pressure venous drainage system via cyclically changing the perivenous space volume and extracranial pressure.^40,64 In the present study, there was a trend towards attenuation of the vasodynamics in the frequency range 0.15–0.60 Hz in the SF 30 days treatment group (Figure 4). It remains to be further determined whether respiration contributes to the suppression of glymphatic function by sleep fragmentation. Similar to the hemodynamic response driven by neural activity, autonomic arousal during NREM sleep generates a transient decrease in CBV, which also promotes CSF pulsation, albeit on a slower time scale. ⁴¹ It remains to be established whether the CBV/CSF phenomena are overlapping, but it seems likely that CSF flow could be driven by the multiple mechanisms that change vascular volume. ⁶⁵ We recently reported that non-significant trend towards an increased blood pressure after SF 30 days compared with controls. ²⁹ Another group using the same model found a minor but significant blood pressure increase after 8 weeks of sleep fragmentation. ⁶⁶ They also reported a transient increase in heart rate confined to the initial weeks of sleep fragmentation. Therefore, we suspect that blood pressure and heart rate were likely unaffected at the end of SF 30 days interventions.

Functional hyperemia, also known as neurovascular coupling, induced by whisker stimulation increased glymphatic influx in rats, whereas optogenetic stimulation of vascular smooth muscle cells increased glymphatic influx in the absence of neural activation. ³⁸ We recently reported the SF 30 days treatment significantly suppressed the functional hyperemia upon whisker stimulation as compared with control animals. ²⁹ These findings highlight that vascular function plays important role in glymphatic function regulation. The absence of significant vascular functional changes in SF 3 days mice seems consistent with the lacking effects on glymphatic system (Figure 2). We thus draw the conclusion that slow vasomotion, but not cardiac-driven arterial pulsatility, is suppressed by chronic sleep fragmentation, which would imply that that repetitive sleep interruptions perturb cerebrovascular dynamics.

The persistent stress due to chronic fragmentation of sleep might potentially generate structural vascular malformations. Classically, stress responses involve increased norepinephrine release and dysregulation of hypothalamic-pituitary-adrenal (HPA) axis. Hypothalamic release of corticotropin-releasing factor elicits pituitary release of adrenocorticotropic hormone and the consequential release of cortisol and catecholamines from the adrenal gland.^67,68 A chronic overactivation of the HPA axis could lead to increased arterial stiffness and atherosclerosis.^67,69 Indeed, epidemiological studies have previously documented a strong association between chronic stress and cardiovascular diseases.^70,71 This is supported by our recent finding that SF 30 days treatment increased the density of the cortical microvasculature, but did not trigger any arteriole remodeling evident to immunolabeling against α-SMA (α-smooth muscle actin) and laminin. ²⁹ Extracellular norepinephrine levels drop abruptly upon the onset of REM sleep. ⁷² Indeed, central norepinephrine signaling is a key element in the regulation of glymphatic function. For example, pharmacological blockade of adrenergic receptors significantly enhances glymphatic influx. ³ Interestingly, our previous data showed that chronic sleep fragmentation significantly reduced the total duration of REM sleep, but not the duration of NREM sleep and wakefulness. ²⁹ Based on these findings, we propose that norepinephrine dysregulation due to chronic sleep fragmentation may contribute to glymphatic impairments.

Neural activity has also been shown to modulate glymphatic fluid transport. ³⁵ Thus, cortical EEG slow delta prevalence correlated strongly with glymphatic influx, while beta prevalence correlated inversely with glymphatic influx in anesthetized mice, with a trend towards a positive correlation for theta prevalence. ³⁵ In this study based on an analysis on non-anesthetized mice, the EEG power spectrum was recorded between ZT 5–12, or within the time window for the glymphatic influx studies done in parallel (Figure 5). The implications of a trend towards an increased theta wave prevalence for glymphatic fluid transport are unknown. However, theta burst stimulation have been shown to facilitate glymphatic clearance efficiency after sleep deprivation. ⁷³

Polarized AQP4 expression is a fundamental requirement for intracerebral fluid transport, since CSF tracer influx and clearance of tracers injected into the brain are suppressed in AQP4 knockout mice.^14,74,75 Other studies have shown that reduced vascular polarization of AQP4 in astrocytic endfeet is associated with declining CSF inflow.^21,76,77 A recent study reported that AQP4 polarization is under dynamic circadian regulation, which peaks at the zenith of glymphatic function during the behaviorally inactive phase. ² However, we saw only minor changes in AQP4 polarization in SF 30 days mice versus control (Figure 5), rendering it unlikely that AQP4 is involved in SF-induced suppression of glymphatic flow.

Clinical relevance

The present observations indicate that chronic sleep fragmentation induces suppression of slow vasomotion, which in turn impairs glymphatic fluid transport in young, healthy wildtype mice. This observation builds upon the clinical evidence linking sleep disturbances to an increased risk of developing cognitive decline and proteinopathies. This is this first study to document that sleep fragmentation, in the absence of an underlying neurological or medical disorder, impairs glymphatic function. Our exclusive use of male mice limits the generalizability of present results. The results substantiate the established linkage between chronic insomnia and vascular pathologies, including vascular dementia.^78,79 Clinical translation of present results could entail the development of sleep medications that do not dampen slow vasomotion, thereby maintaining glymphatic flow and metabolite clearance. An attractive non-pharmacological treatment option for sleep disturbance is voluntary physical exercise, which improves vascular reactivity^80,81 and glymphatic function in animal studies, ⁸² and reduces sleep disturbances in the clinic. ⁸³

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Supplementary material

Supplemental material for this article is available online.