Region-Specific and State-Dependent Astrocyte Ca²⁺ Dynamics during the Sleep-Wake Cycle in Mice

Abstract

Neural activity is diverse, and varies depending on brain regions and sleep/wakefulness states. However, whether astrocyte activity differs between sleep/wakefulness states, and whether there are differences in astrocyte activity among brain regions remain poorly understood. Therefore, in this study, we recorded astrocyte intracellular calcium (Ca²⁺) concentrations of mice during sleep/wakefulness states in the cortex, hippocampus, hypothalamus, cerebellum, and pons using fiber photometry. For this purpose, male transgenic mice expressing the genetically encoded ratiometric Ca²⁺ sensor YCnano50 specifically in their astrocytes were used. We demonstrated that Ca²⁺ levels in astrocytes substantially decrease during rapid eye movement (REM) sleep, and increase after the onset of wakefulness. In contrast, differences in Ca²⁺ levels during non-REM (NREM) sleep were observed among the different brain regions, and no significant decrease was observed in the hypothalamus and pons. Further analyses focusing on the transition between sleep/wakefulness states and correlation analysis with the duration of REM sleep showed that Ca²⁺ dynamics differs among brain regions, suggesting the existence of several clusters, i.e., the first comprising the cortex and hippocampus, the second comprising the hypothalamus and pons, and the third comprising the cerebellum. Our study thus demonstrated that astrocyte Ca²⁺ levels change substantially according to sleep/wakefulness states. These changes were consistent in general unlike neural activity. However, we also clarified that Ca²⁺ dynamics varies depending on the brain region, implying that astrocytes may play various physiological roles in sleep.

SIGNIFICANCE STATEMENT Sleep is an instinctive behavior of many organisms. In the previous five decades, the mechanism of the neural circuits controlling sleep/wakefulness states and the neural activities associated with sleep/wakefulness states in various brain regions have been elucidated. However, whether astrocytes, which are a type of glial cell, change their activity during different sleep/wakefulness states was poorly understood. Here, we demonstrated that dynamic changes in astrocyte Ca²⁺ concentrations occur in the cortex, hippocampus, hypothalamus, cerebellum, and pons of mice during natural sleep. Further analyses demonstrated that Ca²⁺ dynamics slightly differ among different brain regions, implying that the physiological roles of astrocytes in sleep/wakefulness might vary depending on the brain region.

Introduction

Astrocytes, which are the main subtype of glial cells, are essential for central nervous system development and function. Many previous studies have clarified that astrocytes have housekeeping roles in brain function, contributing to ion and neurotransmitter homeostasis, formation, and maintenance of the blood-brain barrier (Sofroniew and Vinters, 2010; Bojarskaite et al., 2020), regulation of blood flow, metabolic support for neurons (Magistretti and Allaman, 2018), neurotransmitter recycling (Sofroniew and Vinters, 2010), and regulation of synaptogenesis and synaptic transmission (Allen, 2014; Allen and Eroglu, 2017). All of these physiological functions of astrocytes are strongly associated with the dynamics of their intracellular calcium (Ca²⁺) concentration. Intrinsic signals, including those involving neurotransmitters, protons, cannabinoids, polyphosphate, and endothelin result in increases in astrocyte Ca²⁺ concentrations (Gourine et al., 2010; Navarrete and Araque, 2010; Filosa et al., 2012; Min and Nevian, 2012; Holmström et al., 2013). In turn, the activated astrocytes release neuroactive substances called gliotransmitters, which activate neurons and vascular smooth muscle (Sasaki et al., 2012; Araque et al., 2014; Beppu et al., 2014; Savtchouk and Volterra, 2018).

In addition, it has become clear that astrocytes are involved in the regulation and physiological functions of mammalian sleep (Frank, 2019). There is evidence that astrocytes regulate sleep pressure through soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE)-dependent adenosine release (Halassa et al., 2009; Florian et al., 2011). Astrocytes also promote the sleep-dependent clearance of brain waste products, such as β-amyloid (Xie et al., 2013). The dysfunction of gap junctions in astrocytes leads to the inability to transfer lactate to wake-promoting orexin neurons in the lateral hypothalamic area, resulting in sleep disorders (Clasadonte et al., 2017), whereas optogenetic stimulation of astrocytes in the hypothalamus increases sleep (Pelluru et al., 2016). Furthermore, astrocytes also regulate cortical state switching (Poskanzer and Yuste, 2016). These results strongly indicate that astrocytes play a pivotal role in sleep as well as other brain functions.

However, astrocytic Ca²⁺ dynamics in natural sleep has been poorly understood, although an imaging study demonstrated that general anesthesia disrupts astrocyte Ca²⁺ signaling in mice (Thrane et al., 2012). Recent studies have recorded astrocyte Ca²⁺ changes during sleep/wakefulness in mice, but to date, data have only been reported for the cortex (Bojarskaite et al., 2020; Ingiosi et al., 2020). Accumulating evidence has shown that astrocytes are heterogeneous with respect to their transcriptomes and functions among various brain regions as well as among various neuronal types (Chai et al., 2017; Morel et al., 2017; Zeisel et al., 2018; Batiuk et al., 2020; Bayraktar et al., 2020; Lozzi et al., 2020). It is well known that neurons in various brain regions, i.e., the hypothalamus, midbrain, and brainstem, control the sleep/wakefulness state, as well as rapid eye movement (REM)/non-REM (NREM) sleep, via a flip-flop circuit. (Sakurai, 2007; Weber and Dan, 2016; Scammell et al., 2017; Liu and Dan, 2019). However, little is known about how astrocyte Ca²⁺ concentrations in brain regions other than the cortex change depending on the sleep/wakefulness state, and whether there are dynamic differences in astrocyte Ca²⁺ levels among the different brain regions. Therefore, in this study, we performed fiber photometry recordings of astrocyte Ca²⁺ dynamics of the cerebellum, cortex, hippocampus, hypothalamus, and pons of mice during the sleep-wake cycle. To optically record astrocyte Ca²⁺ dynamics, we used megalencephalic leukoencephalopathy with subcortical cysts 1 (Mlc1)-tetracygline transactivator (tTA); TetO-YCnano50 bigenic mice (Horikawa et al., 2010; Tanaka et al., 2010, 2012; Kanemaru et al., 2014) in which astrocytes specifically express YCnano50.

In this study, we demonstrated that astrocyte Ca²⁺ levels consistently decrease during REM sleep, and immediately increase with the onset of wakefulness. In contrast, differences in Ca²⁺ levels during NREM sleep were observed in various brain regions. There was a significant decrease in Ca²⁺ levels in the cerebellum, cortex, and hippocampus during NREM sleep, but no significant decrease was observed in the hypothalamus and pons.

Materials and Methods

Animals

All experimental procedures involving animals were approved by the Animal Care and Use Committee of Tohoku University (approval no. 2019LsA-018) and were conducted in accordance with the National Institute of Health guidelines. All efforts were made to minimize animal suffering and discomfort, and to reduce the number of animals used. Mlc1-tTA; TetO-YCnano50 mice, which were used to monitor the dynamics of intracellular calcium concentration in astrocytes, were produced by crossing Mlc1-tTA mice (Tanaka et al., 2010) with TetO-YCnano50 mice (Kanemaru et al., 2014). The following PCR primer sets were used for mouse genotyping: MlcU-675 (5′-AAATTCAGGAAGCTGTGTGCCTGC-3′) and mtTA24L (5′-CGGAGTTGATCACCTTGGACTTGT-3′) for Mlc1-tTA mice; and tetO-up (5′-AGCAGAGCTCGTTTAGTGAACCGT-3′) and intron-low (5′-AAGGCAGGATGATGACCAGGATGT-3′) for TetO-YCnano50 mice. Mice were housed under a controlled 12 h/12 h light/dark cycle (light on hours 8:30 A.M. to 8:30 P.M.). Mice had ad libitum access to food and water. A total of 14 male mice (nine Mlc1-tTA; TetO-YCnano50 mice and five Mlc1-tTA mice as controls) were used in this study. The following number of mice were used for the recording of each brain area: cerebellum, three mice (five recordings); cortex, three mice (six recordings); hippocampus, one mouse (two recordings); hypothalamus, two mice (four recordings); pons, one mouse (two recordings). For the recording of control Mlc1-tTA mice, three mice were used (eight recordings). For the freely moving recording, two Mlc1-tTA; TetO-YCnano50 mice and two Mlc1-tTA mice as controls were used.

Surgical procedures

Male Mlc1-tTA; TetO-YCnano50 bigenic mice and Mlc1-tTA monogenic mice (≥12 weeks of age) were used. Stereotaxic surgery was performed under anesthesia with pentobarbital (5 mg/kg, i.p. as induction) and with isoflurane (1–2% for maintenance) using a vaporizer for small animals (Bio Research Center) with the mice positioned in a stereotaxic frame (Narishige). Two bone screws were implanted on the skull as electrodes for cortical electroencephalograms (EEGs), and twisted wires (AS633, Cooner Wire) were inserted into the neck muscle as an electrode for electromyograms (EMGs). Another bone screw was implanted in the cerebellum as a ground. All electrodes were connected to a pin socket.

For fiber photometry experiments, a cannula (CF440-10, Thorlabs) with a glass optical fiber (φ 400 μm, 0.39 NA, Thorlabs) was implanted into the cortex (1.2 mm posterior, 3.1 mm lateral from bregma, 0.3 mm depth from the brain surface), hippocampus (1.7 mm posterior, 1.5 mm lateral from bregma, 1.3 mm depth from the brain surface), hypothalamus (1.8 mm posterior, 1.0 mm lateral from bregma, 4.5 mm depth from the brain surface), cerebellum (6.0 mm posterior, 1.0 mm lateral from bregma, and 0.5 mm depth from the brain surface), and pons (5.1 mm posterior, 1.2 mm lateral from bregma, and 3.5 mm depth from the brain surface). All electrodes and optical fiber cannulas were fixed to the skull with dental cement. To fixate the head of mice, a stainless chamber frame (CF-10, Narishige) was also attached to the skull using dental cement. After the surgery, the mice were left to recover for at least 5 d. During the habituation period, mice were placed in a head-fix apparatus (MAG-1, Narishige), by securing them by the stainless chamber frame and placing them into an acrylic tube. This procedure was continued for at least 5 d, during which the duration of head-fixation was gradually extended from ten to 120 min.

In vivo fiber photometry experiments in the head-fixed condition

To detect the dynamics of intracellular Ca²⁺ concentrations in astrocytes, a fiber photometric system (Lucir) was used. A 420 nm violet light-emitting diode (LED; Doric) was used to obtain Ca²⁺-dependent signals. Recording in the head-fixed condition was performed for about 5 h/d. During recording, EEGs and EMGs were recorded continuously, whereas excitation light (20 Hz, 5 ms in width) was intermittently illuminated at random for 4 min each time. The input light was reflected off a dichroic mirror (FF458-Di02, Semrock) coupled to an optical fiber. LED power was 1.12 ± 0.18 mW/mm² at the fiber tip. Light emission of cyan and yellow fluorescence from YCnano50 was collected via an optical fiber cannula, divided by a dichroic mirror (FF509-FDi01, Semrock) into cyan (483/32-nm bandpass filter, Semrock) and yellow (542/27-nm bandpass filter, Semrock), and detected by each photomultiplier (Lucir). Excitation signals were generated by a pulse generator (AWG-50, Elmos) to control the LEDs. Fluorescence data were acquired at a sampling rate of 1 kHz through an analog-to-digital converter [Micro1401-3, Cambridge Electronic Design (CED)]. At the same time as the fluorescence recording, EEG and EMG signals were amplified (DAM50, World Precision Instruments), filtered, and digitized at 1 kHz using an analog-to-digital converter. EEG and EMG signals were high-pass and low-pass filtered at 0.1 and 300 Hz, respectively. Locomotion was induced by pinching the tail with forceps. Data were recorded using Spike2 software (CED).

In vivo sleep/wakefulness recording using freely moving mice

For the analysis of freely moving mice, continuous EEG and EMG recordings were performed through a slip ring (SPM-35-8P-03, HIKARI DENSHI), which was designed so that the movement of the mice was unrestricted. EEG and EMG signals were amplified (AB-610J, Nihon Koden), filtered (EEG, 0.75–20 Hz; EMG, 20–50 Hz), digitized at a sampling rate of 128 Hz, and recorded using SleepSign software version 3 (Kissei Comtec).

Histologic analysis

To confirm the position of the implanted optical fibers, mice were deeply anesthetized with isoflurane and perfused sequentially with 20 ml of chilled saline and 20 ml of chilled 4% paraformaldehyde in phosphate buffer solution (Nacalai Tesque). The brains were removed and immersed in the above fixation solution overnight at 4°C, and then immersed in 30% sucrose in PBS for at least 2 d. The brains were quickly frozen in embedding solution (Sakura Finetek), and cut into coronal sections using a cryostat (CM3050, Leica) at a thickness of 40 and 50 μm. For immunostaining, to confirm the expression of YCnano50 in astrocytes, the brain sections of Mlc1-tTA; TetO-YCnano50 bigenic mice were incubated with mouse anti-S100β antibody (1:1000; S2532, Merck) or mouse anti-GFAP antibody (1:2000; G3893, Merck) overnight at 4°C. Then, the sections were incubated with CF594 donkey anti-mouse IgG (1:1000; 20116-1, Nacalai Tesque) for 1 h at room temperature, mounted onto APS-coated slides, coverslipped with 50% glycerol in PBS, and observed using a fluorescence microscope (BZ-9000, Keyence) or a confocal microscope (LSM800, Zeiss).

Data analysis

Sleep scoring

Polysomnographic recordings were automatically scored offline, with each epoch scored as wakefulness, NREM sleep, or REM sleep by SleepSign (Kissei Comtec), in 4-s epochs, according to standard criteria (Radulovacki et al., 1984; Tobler et al., 1997). All vigilance state classifications assigned by SleepSign were confirmed visually. The same individual, blinded to mouse genotype and experimental condition, scored all EEG/EMG recordings. Spectral analysis of the EEGs was performed by fast Fourier transform, which yielded a power spectral profile with a 1-Hz resolution divided into δ (1−5 Hz), θ (6−10 Hz), α (10−13 Hz), β (13−25 Hz), and γ (30−50 Hz) waves. To quantify EMG amplitude, the root-mean-square (rms) was calculated.

Fiber photometry signal processing

In Figures 1–4, axoGraph was used to calculate yellow fluorescence protein (YFP) to cyan fluorescence protein (CFP; Y/C) ratios. The average value of the YFP intensity and CFP intensity for each light illumination (5 ms) was calculated, and then the Y/C ratio was calculated. For the comparison of Y/C ratios during the sleep/wakefulness states, the Y/C ratio of each episode was normalized with the average value during wakefulness set as 1. For sleep/wakefulness state transition analyses, four consecutive epochs (16 s) of one state followed immediately by eight consecutive epochs (32 s) of a distinct state were used. To assess correlations between Y/C ratio and EEG/EMG power (Figs. 2, 4), spectral densities of EEG signals in every 1-s window were estimated at δ (1–5 Hz), θ (6–10 Hz), α (13–25 Hz), β (13–25 Hz), and γ (30–50 Hz) bands using the multitaper method (http://chronux.org/). EMG power was calculated as the common logarithm of a rms value in every 1-s window. The mean of Y/C ratio in the corresponding 1-s window was also calculated. Pearson's correlation coefficient was computed based on z-scored values in each state.

Figure 1.

Astrocyte Ca²⁺ dynamics in the cerebellum during sleep/wakefulness states of mice. A, Immunohistochemical analysis demonstrating that YCnano50 is specifically expressed in cerebellar astrocytes in the Mlc1-tTA; TetO-YCnano50 bigenic mouse brain. Left, YFP fluorescence of YCnano50-positive cells (green). Middle, S100β-immunoreactive astrocytes (red). Left, Merged image (yellow). Scale bar: 40 µm. GL, granule cell layer; ML, molecular layer; PL, Purkinje cell layer. B, Schematic drawing showing the fiber photometry system used in this study. Fluorescence emission is applied from the LED light source. Yellow and cyan fluorescence signals were corrected by bandpass filters and enhanced by photomultipliers (PMT). C, The location of the glass optical fiber, which was implanted in the cerebellum of Mlc1-tTA; TetO-YCnano50 bigenic mice (6.0 mm posterior, 1.0 mm lateral from bregma, 0.5 mm depth from the brain surface). Scale bar: 1 mm. D, Example of Y/C ratios (top) and corresponding intensity changes of yellow and cyan fluorescence (bottom) recorded during tail pinch-induced locomotion in Mlc1-tTA; TetO-YCnano50 bigenic mice. The arrows indicate the timing of the tail pinch. E, Box plot summarizing the data from D. The Y/C ratios during locomotion were normalized using the Y/C ratio at 10 s immediately before the locomotion set as 1; *p < 0.05. F, Comparison of GFAP immunoreactivity (red) with (arrowheads) or without (arrows) YCnano50 expression (green) in the hippocampal astrocytes. Scale bar: 15 µm. G, H, Representative traces of EEG, EEG power density spectrum, EMG, cerebellar astrocyte Ca²⁺ signals (Y/C ratio), and Y/C ratio spectrogram in Mlc1-tTA; TetO-YCnano50 bigenic mice (G) and Mlc1-tTA monogenic mice (H). I, Box plot summarizing the data from G, H. Y/C ratios were normalized to the value of each episode, with the average value of awakening set as 1; *p < 0.05. J, Y/C ratios during the transitions between the sleep/wakefulness states. Transitions occurred at time 0. Data are from 4-s intervals characterized by state transitions. The line graph with the colored circles and gray circles are a summary of the data from Mlc1-tTA; TetO-YCnano50 bigenic mice and Mlc1-tTA monogenic mice, respectively; *p < 0.05 versus the fourth epoch immediately before the state transition. K, Bar graph representing the slope of the Y/C ratio of mice at the time of awakening from NREM sleep and REM sleep; *p < 0.05. L–N, Analyses of the correlation between Y/C ratios and episode duration of wakefulness (L), NREM sleep (M), and REM sleep (N). Colored circles and gray circles indicate the summary of data from Mlc1-tTA; TetO-YCnano50 bigenic mice and Mlc1-tTA monogenic mice, respectively. NR, NREM sleep; R, REM sleep; W, wakefulness. Values are shown as means ± SEM.

Figure 4.

Correlation of EEG/EMG and cortical astrocyte Ca²⁺ signals during different sleep/wakefulness states. A–C, Correlation analyses between normalized (z-scored) Y/C ratios from the cortex and normalized (z-scored) EEG power densities in the δ (1–5 Hz), θ (6–10 Hz), α (10–13 Hz), β (13–25 Hz), and γ (30–50 Hz) waves during wakefulness (A), NREM sleep (B), and REM sleep (C). D–F, Correlation analyses between normalized Y/C ratios and normalized rms of EMG during wakefulness (D), NREM sleep (E), and REM sleep (F). The data in this figure were analyzed using 1-s bin sizes. G–I, Bar graphs showing correlation coefficients summarizing the data from A–F. The correlation coefficient of each recording was cross-validated by splitting the data into the first and second halves. Values are shown as means ± SEM.

Figure 2.

Correlation between EEG/EMG and cerebellar astrocytic Ca²⁺ signals during different sleep/wakefulness states. A–C, Correlation analyses between normalized (z-scored) cerebellar astrocytic Y/C ratios, and normalized (z-scored) EEG power densities in the δ (1–5 Hz), θ (6–10 Hz), α (10–13 Hz), β (13–25 Hz), and γ (30–50 Hz) waves during wakefulness (A), NREM sleep (B), and REM sleep (C). D–F, Correlation analyses between normalized Y/C ratios and normalized rms of EMG during wakefulness (D), NREM sleep (E), and REM sleep (F). The data in this figure were analyzed in 1-s bin sizes. G–I, Bar graphs showing correlation coefficients summarizing the data from A–F. The correlation coefficient of each recording was cross-validated by splitting the data into the first and second halves. Values are shown as means ± SEM.

In Figure 5, all data analyses were performed by custom written MATLAB software (MathWorks). To compute normalized Y/C ratios during the time-normalized episodes, Y/C ratios were first normalized by the mean Y/C ratio during wakefulness. To obtain the time-normalized Ca²⁺ dynamics, each episode was segmented into five bins, and the mean Y/C ratio was computed for each bin. Episodes were classified based on vigilance states before and after the episode.

Figure 5.

Dynamics of Ca²⁺ signals during different sleep/wakefulness states and different brain regions. A, The mean profiles of Ca²⁺ signals during time-normalized episodes. In each panel, the duration of each episode was segmented into 5 bins and the mean normalized Y/C ratios were computed in various brain regions. B, Decoding performance of Ca²⁺ signals for different sleep/wakefulness states among various brain regions; *p < 0.05, F_(5,17) = 6.10, one-way ANOVA with the post hoc HSD test. Values are shown as means ± SEM.

Decoding

To decode sleep/wakefulness states based on Ca²⁺ signals, the same approach as described previously was used (Tsunematsu et al., 2020). Briefly, the mean Y/C ratio was computed in each corresponding window (4 s). After training a linear classifier, classification performance was calculated with fourfold cross validation.

Experimental design and statistical analysis

Data are presented as the mean ± SEM unless otherwise stated. Statistical analyses were performed using MATLAB. Multiple group comparisons were performed by one-way ANOVA in samples with a Gaussian distribution, and by the Kruskal–Wallis test in samples with a non-Gaussian distribution, with the post hoc Bonferroni test. In Figure 5B, one-way ANOVA with the post hoc Tukey's HSD test was performed; p values of <0.05 were considered to indicate a statistically significant difference between groups. To calculate effect size, power analyses were performed using G*Power 3.1 (Faul et al., 2007).

Results

Fiber photometry recording of Ca²⁺ signals in cerebellar astrocytes

To elucidate the dynamics of intracellular Ca²⁺ concentration in astrocytes during sleep/wakefulness states, we used Mlc1-tTA; TetO-YCnano50 bigenic mice. Mlc1 is an astrocyte-specific protein with unknown function, which is highly expressed in perivascular astrocyte end-feet and astrocyte-astrocyte contacts (Boor et al., 2005; Teijido et al., 2007). We first focused on cerebellar astrocytes, and investigated the Ca²⁺ changes. It has been reported that the activity of Bergmann glial cells, a specific type of radial astrocyte in the cerebellum, is inhibited in the anesthetized state compared with in the awake state (Hoogland et al., 2009; Nimmerjahn et al., 2009). Therefore, we hypothesized that cerebellar astrocytes may demonstrate state-dependent Ca²⁺ dynamics throughout the sleep-wake cycle. To this end, we took advantage of the strong fluorescence intensity of cerebellar astrocytes in Mlc1-tTA; TetO-YCnano50 bigenic mouse line (Kanemaru et al., 2014).

To confirm astrocyte-specific expression of YCnano50 in the cerebellum, we performed immunostaining of astrocytes using S100β, an astrocyte-specific marker. The merged images show that YCnano50 was exclusively observed in cerebellar astrocytes in Mlc1-tTA; TetO-YCnano50 bigenic mice, and ectopic expression was not observed (Fig. 1A). Astrocytes in the cortex and hippocampus sparsely expressed YCnano50 as previously reported (Kanemaru et al., 2014). In this study, fiber photometry was used to record changes in astrocyte Ca²⁺ concentrations from head-fixed mice, as Y/C ratios (Fig. 1B). We first confirmed whether Y/C ratios calculated from the YFP fluorescence and CFP fluorescence precisely reflect the changes in astrocyte Ca²⁺ concentrations under our experimental conditions. For this purpose, we analyzed changes in the Y/C ratio during tail pinch-induced locomotion in mice, as it has previously been reported that astrocyte Ca²⁺ concentrations increase with locomotion and the startle response in mice (Nimmerjahn et al., 2009; Srinivasan et al., 2015; Bojarskaite et al., 2020). To deliver excitation light and collect fluorescence signals from the cerebellum, a glass optical fiber was implanted into the cerebellum of mice (Fig. 1C). Immediately after the tail pinch-induced locomotion, YFP fluorescence and CFP fluorescence showed changes in opposite directions, resulting in an increase in the Y/C ratio (Fig. 1D). The average Y/C ratio of the 10 s immediately before the tail pinch-induced locomotion was set to 1. The average Y/C ratio of the 10 s immediately after the tail pinch-induced locomotion was normalized, and a significant increase in Y/C ratio on locomotion was observed (n = 4 from three recording sessions and one animal; paired t test, *p < 0.05).

To assess the effect of YCnano50 expression in astrocytes, we examined whether there are any signs of reactive astrocyte. We compared with the hippocampal astrocytes which have sparce expression of YCnano50. There was no morphologic change and no signs of upregulation of glial fibrillary acidic protein (GFAP) immunoreactivity between YCnano50-expressing astrocytes and intact astrocytes (Fig. 1F), indicating no toxic effects. To further confirm the effect of YCnano50 expressed in bigenic mouse astrocytes and the effect of the head-fixed condition on sleep architecture, we compared the duration of each sleep/wakefulness episode and time spent in each sleep/wakefulness state during the light period (9 A.M. to 3 P.M.) between the mice. To determine the sleep/wakefulness state of mice, EEG and EMG electrodes were implanted into their skulls and neck muscles. There were no significant differences between Mlc1-tTA; TetO-YCnano50 bigenic mice (six recording sessions and two animals) and Mlc1-tTA monogenic mice in the freely moving condition [six recording sessions and two animals; episode duration of wakefulness: Kruskal–Wallis: F_(3,28) = 180.9, p = 0.10, no significant difference (NS); episode duration of NREM: Kruskal–Wallis: F_(3,28) = 495.6, p < 0.05, followed by multiple comparisons by the Bonferroni test: p = 1, NS; episode duration of REM: Kruskal–Wallis: F_(3,21) = 92.2, p = 0.16, NS; time spent in wakefulness: Kruskal–Wallis: F_(3,28) = 11.1, p = 0.94, NS; time spent in NREM: Kruskal–Wallis: F_(3,28) = 23.1, p = 0.85, NS; time spent in REM: Kruskal–Wallis: F_(3,28) = 112.6, p = 0.27, NS; Table 1]. On the other hand, the episode duration of NREM sleep was significantly reduced in mice in the head-fixed condition (without hypothalamus, 16 recording sessions and six animals) compared with mice in the freely moving condition (six recording sessions and two animals; episode duration of NREM: Kruskal–Wallis: F_(3,28) = 495.6, p < 0.05, followed by multiple comparisons by the Bonferroni test: p < 0.05; Table 1). However, no significant differences were observed between the other conditions.

Table 1.

Sleep architecture in mice in the head-fixed condition and freely moving condition

	Wakefulness	NREM	REM
Episode duration (s)
Free-moving condition of Mlc1-tTA mice	158.7 ± 25.5	187.3 ± 14.9	87.8 ± 14.5
Free-moving condition of Mlc1-tTA; TetO-YCnano50 mice	197.0 ± 24.6	205.8 ± 18.7	62.3 ± 6.0
Head-fixed condition without recording from hypothalamus of Mlc1-tTA; TetO-YCnano50 mice	134.6 ± 32.7	105.7 ± 18.4*	90.9 ± 7.5
Head-fixed condition with recording from hypothalamus of Mlc1-tTA; TetO-YCnano50 mice	102.1 ± 36.9	106.3 ± 13.8	86.8 ± 26.4
Time spent in each state (%)
Freely moving condition of Mlc1-tTA mice	42.6 ± 4.9	52.2 ± 3.4	5.2 ± 1.1
Freely moving condition of Mlc1-tTA; TetO-YCnano50 mice	46.6 ± 3.1	49.5 ± 2.7	3.9 ± 0.9
Head-fixed condition without recording from hypothalamus	48.4 ± 6.2	43.4 ± 5.1	8.2 ± 1.5
Head-fixed condition with recording from hypothalamus	41.9 ± 7.4	51.6 ± 5.2	6.5 ± 3.4

Data are shown as means ± SD.

*p < 0.05, head-fixed condition without recording from hypothalamus versus freely moving condition of Mlc1-tTA mice and freely moving condition of Mlc1-tTA; TetO-YCnano50 mice.

State-dependent Ca²⁺ dynamics in cerebellar astrocytes

We next analyzed the dynamics of intracellular Ca²⁺ concentration in astrocytes during sleep/wakefulness states. Ca²⁺ signal dynamics were recorded during the light period (9 A.M. to 3 P.M.). Excitation light (20 Hz, 5 ms in width) was intermittently illuminated at random for 4 min. Y/C ratios gradually decreased during sleep, showing the lowest value during REM sleep, and instantaneously increased with awakening in Mlc1-tTA; TetO-YCnano50 mice (Fig. 1G). A decrease in Y/C ratio represents a decrease in Ca²⁺ concentration, as we have previously reported (Natsubori et al., 2017; Tsutsui-Kimura et al., 2017; Yoshida et al., 2020). In contrast, Y/C ratios did not change with sleep/wakefulness state in control Mlc1-tTA mice (Fig. 1H). The spectrogram of Y/C ratio in Mlc1-tTA; TetO-YCnano50 mice exhibited state-dependent changes at below ∼0.25 Hz, whereas small fluctuations at above ∼0.25 Hz appeared even in Mlc1-tTA mice. These results indicate that signals below ∼0.25 Hz reflect state-dependent Ca²⁺ signals in astrocyte. Normalized Y/C ratios in wakefulness, NREM sleep, and REM sleep in cerebellar astrocytes of Mlc1-tTA; TetO-YCnano50 mice were 1.000 ± 0.001 (n = 66 episodes from 5 recording sessions and 3 animals), 0.988 ± 0.002 (n = 63 episodes from 5 recording sessions and three animals), and 0.944 ± 0.008 (n = 17 episodes from three recording sessions and 2 animals), respectively (Kruskal–Wallis: F_(2,143) = 56.68, p < 0.05, followed by multiple comparison by the Bonferroni test: p < 0.05, effect size f = 1.07; Fig. 1I). In contrast, normalized Y/C ratios during wakefulness, NREM sleep, and REM sleep in cerebellar astrocytes of Mlc1-tTA mice were 1.000 ± 0.001 (n = 60 episodes from 5 recording sessions and 3 animals), 1.000 ± 0.001 (n = 70 episodes from 5 recording sessions and three animals), and 0.997 ± 0.001 (n = 33 episodes from 5 recording sessions and three animals), respectively (Kruskal–Wallis: F_(2,160) = 4.04, p = 0.13, NS, effect size f = 0.17; Fig. 1I). These results indicate that Ca²⁺ concentrations of cerebellar astrocytes change substantially with sleep/wakefulness state in mice.

We next focused on Ca²⁺ dynamics during the transition between sleep/wakefulness states (Fig. 1J). Y/C ratios gradually decreased after the onset of NREM sleep following wakefulness. Sixteen seconds after the start of NREM sleep, Y/C ratios significantly decreased compared with when mice were awake (n = 26 episodes from three recording sessions and two animals; one-way ANOVA: F_(11,300) = 6.79, p < 0.05, followed by multiple comparisons by the Bonferroni test: p < 0.05 vs the fourth epoch immediately before state transition). A slow decrease in Y/C ratios was also observed in the transition from NREM sleep to REM sleep, but there was no significant difference (n = 10 episodes from two recording sessions and two animals; one-way ANOVA: F_(11,108) = 2.35, p < 0.05, followed by multiple comparison by the Bonferroni test: p ≥ 0.05, NS). During the transition from NREM and REM sleep to wakefulness, the Y/C ratio increased (from NREM to wakefulness: n = 19 episodes from four recording sessions and three animals; from REM to wakefulness: n = 14 episodes from three recording sessions and two animals; from NREM to wakefulness, one-way ANOVA: F_(11,216) = 5.47, p < 0.05, followed by multiple comparisons by the Bonferroni test: p < 0.05 vs the fourth epoch immediately before state transition; from REM to wake, one-way ANOVA: F_(11,156) = 12.87, p < 0.05, followed by multiple comparisons by the Bonferroni test: p < 0.05 vs the fourth epoch immediately before state transition). However, the slopes were completely different between from NREM to wakefulness and from REM to wakefulness. The slope was calculated by dividing the difference in Y/C ratios immediately before and after the transition by 4 s. The slope from REM to wakefulness (n = 14 episodes from three recording sessions and two animals) was significantly larger than the slope from NREM to wakefulness (n = 19 episodes from four recording sessions and 3 animals; unpaired t test, p < 0.05; Fig. 1K). These results indicate that the signaling pathway that increases intracellular Ca²⁺ concentrations might differ between from REM sleep and from NREM sleep, although they do not appear to precede the 4-s bin.

Next, correlation analysis was performed to investigate the association between the episode duration of wakefulness, NREM, and REM, and changes in Ca²⁺ concentrations (Fig. 1L–N). There was no significant correlation between wakefulness and NREM sleep in either Mlc1-tTA; TetO-YCnano50 mice (wakefulness: n = 41 episodes from three recording sessions and two animals; NREM: n = 24 episodes from three recording sessions and two animals; wakefulness: r = 0.14, p = 0.38, NS; NREM: r = −0.23, p = 0.28, NS) or in Mlc1-tTA mice (wakefulness: n = 46 episodes from five recording sessions and three animals; NREM: n = 20 episodes from four recording sessions and three animals; wakefulness: r = 0.13, p = 0.40, NS; NREM: r = −0.18, p = 0.45, NS; Fig. 1L,M). In contrast, there was a significant negative correlation between the episode duration of REM and Y/C ratio in Mlc1-tTA; TetO-YCnano50 mice (n = 10 episodes from two recording sessions and two animals; r = −0.79, p < 0.05; Fig. 1N), which is in good agreement with the gradual decrease in Ca²⁺ level during REM sleep. On the other hand, no significant correlation was observed in Mlc1-tTA mice (n = 12 episodes from two recording sessions and two animals; r = 0.09, p = 0.79, NS; Fig. 1N). These results indicate that Ca²⁺ concentration in cerebellar astrocytes decreases as REM sleep episode duration increases.

Correlation between EEG/EMG and Ca²⁺ signals in cerebellar astrocytes during sleep/wakefulness

Our results up to this point demonstrated that cerebellar astrocyte Ca²⁺ concentrations change dynamically with sleep/wakefulness state. Therefore, we next investigated whether Ca²⁺ fluctuations in cerebellar astrocytes also correlate with electrophysiological features during each sleep/wakefulness state. For this purpose, we investigated the association between Y/C ratios and cortical EEGs and EMGs during wakefulness, NREM, and REM sleep (50 recording sessions and three animals). Cortical EEGs were analyzed by dividing them into δ (1–5 Hz), θ (6–10 Hz), α (10–13 Hz), β (13–25 Hz), and γ (30–50 Hz) wave components. Regarding EMGs, their magnitudes were evaluated by calculating the rms. Then, their correlation with Y/C ratios was analyzed. During wakefulness, although Y/C ratios showed a significant negative correlation with α waves and EMG rms (Fig. 2A,D), the effect was weak (Fig. 2G), suggesting little cofluctuation of Ca²⁺ and electrophysiological signals during wakefulness. A negative correlation between EMG rms and Y/C ratio imply that Y/C ratio does not reflect simple motion-related signals. During NREM sleep, Y/C ratios were positively correlated with δ waves and EMG rms, and negatively correlated with θ, α, and β waves (Fig. 2B,E,H). During REM sleep, however, Y/C ratios demonstrated a positive correlation with δ, α, and β waves and EMG rms, and a negative correlation with θ waves (Fig. 2C,F,I). Although this correlation analysis demonstrated significant differences, no highly positive nor highly negative correlation was identified. A minor but statistically significant correlation between cerebellar astrocyte Ca²⁺ concentrations and electrophysiological features, particularly during sleep, was identified.

State-dependent astrocyte Ca²⁺ dynamics among various brain regions

We next assessed whether state-dependent changes in astrocyte Ca²⁺ concentrations are observed not only in the cerebellum but also in other brain regions, and whether the dynamics differ depending on the brain region. Fiber photometry recordings were performed using a glass optical fiber from the cortex and hippocampus, which have been reported to show diversity in neural activity corresponding to the sleep/wakefulness state (Vyazovskiy et al., 2009; Grosmark et al., 2012; Watson et al., 2016; Niethard et al., 2017), and from the hypothalamus and pons, which play a crucial role in the regulation of sleep/wakefulness (Sakurai, 2007; Tsunematsu et al., 2014; Hayashi et al., 2015; Weber et al., 2015; Weber and Dan, 2016; Scammell et al., 2017), using Mlc1-tTA; TetO-YCnano50 mice. To deliver excitation light and collect fluorescence signals, a glass optical fiber was implanted into the brains of mice (Fig. 3A). Hypothalamic recording of fiber-implanted in head-fixed Mlc1-tTA; TetO-YCnano50 mice (four recording sessions and two animals) demonstrated no differences in sleep/wakefulness architecture compared with recordings from other brain regions in head-fixed Mlc1-tTA; TetO-YCnano50 mice (16 recording sessions and six animals), although the implanted optical fiber damages the brain and causes similar to glial scarring around the optical fiber (episode duration of wakefulness: Kruskal–Wallis: F_(3,28) = 180.9, p = 0.10, NS; episode duration of NREM: Kruskal–Wallis: F_(3,28) = 495.6, p < 0.05, followed by multiple comparisons by the Bonferroni test: p = 1, NS; episode duration of REM: Kruskal–Wallis: F_(3,21) = 92.2, p = 0.16, NS; time spent in wakefulness: Kruskal–Wallis: F_(3,28) = 11.1, p = 0.94, NS; time spent in NREM: Kruskal–Wallis: F_(3,28) = 23.1, p = 0.85, NS; time spent in REM: Kruskal–Wallis: F_(3,28) = 112.6, p = 0.27, NS; Table 1). Interestingly, astrocyte Ca²⁺ levels dynamically changed depending on the sleep/wakefulness state throughout the brain regions that were monitored, and significantly decreased in all areas during REM sleep [n = 69 wakefulness (W), 68 NREM sleep (NR), and 25 REM sleep (R) episodes from five recording sessions and three animals in the cortex; n = 23 (W), 24 (NR), and 8 (R) episodes from two recording sessions and one animal in the hippocampus; n = 60 (W), 69 (NR), and 20 (R) episodes from four recording sessions and two animals in the hypothalamus; n = 40 (W), 43 (NR), and 34 (R) episodes from two recording sessions and one animal in the pons; Kruskal–Wallis: F_(2,159) = 29.3, p < 0.05, effect size f = 0.47 in the cortex; Kruskal–Wallis: F_(2,52) = 39.1, p < 0.05, effect size f = 1.34 in the hippocampus; Kruskal–Wallis: F_(2,146) = 30.0, p < 0.05, effect size f = 0.80 in the hypothalamus; Kruskal–Wallis: F_(2,114) = 55.8, p < 0.05, effect size f = 1.13 in the pons, followed by multiple comparisons by the Bonferroni test: p < 0.05; Fig. 3B]. However, significant decreases in Y/C ratios during NREM sleep were observed in the cortex and hippocampus compared with that during wakefulness. In contrast, in the hypothalamus and pons, no significant differences were observed between the Y/C ratios during wakefulness and NREM sleep, but Y/C ratios were significantly reduced during REM sleep compared with during NREM sleep. These results suggest that astrocyte Ca²⁺ levels are at a minimum during REM sleep, whereas they are high during wakefulness. This trend is consistent with that observed in the cerebellum. On the other hand, Ca²⁺ dynamics during NREM sleep vary depending on the brain region.

Figure 3.

Astrocyte Ca²⁺ dynamics during different sleep/wakefulness states in various brain regions. A, Images indicating the location of the glass optical fiber, which was implanted into the cortex, hippocampus, hypothalamus, and pons. Arrows indicates the tip of the optical fiber. Scale bar: 500 µm (cortex and hippocampus) and 1 mm (hypothalamus and pons). B, Box plot summarizing the data of the normalized Y/C ratios obtained from the cortex, hippocampus, hypothalamus, and pons; *p < 0.05. C, Y/C ratios for the transition of sleep/wakefulness states in multiple brain regions. Black filled symbols indicate p < 0.05 versus the fourth epoch immediately before state transition in each brain region. D, Correlation analyses between episode durations of REM sleep and Y/C ratios in the cortex, hippocampus, hypothalamus, and pons. NR, NREM sleep; R, REM sleep; W, wakefulness. Values are represented as means ± SEM.

Further analyses during the sleep/wakefulness state transition clarified the variety of Ca²⁺ dynamics among the different brain regions (Fig. 3C). Focusing on the transition from wakefulness to NREM sleep, no significant changes were seen in the cortex, hypothalamus, and pons (n = 44 episodes from six recording sessions and three animals in the cortex; n = 38 episodes from two recording sessions and one animal in the hypothalamus; n = 17 episodes from one recording session and one animal in the pons; Kruskal–Wallis: F_(11,516) = 11.7, p = 0.39, NS in the cortex; Kruskal–Wallis: F_(11,444) = 3.2, p = 0.99, NS in the hypothalamus; Kruskal–Wallis: F_(11,192) = 7.3, p = 0.77, NS in the pons). However, Y/C ratios gradually decreased in the hippocampus, and the difference became statistically significant after 24 s from state transition (n = 12 episodes from two recording sessions and one animal in the hippocampus; Kruskal–Wallis: F_(11,132) = 49.2, p < 0.05, followed by multiple comparisons by the Bonferroni test: p < 0.05 vs the fourth epoch immediately before state transition in the hippocampus). A slow decrease in Y/C ratios in the hypothalamus and pons was observed from NREM to REM sleep (n = 6 episodes from two recording sessions and one animal in the hypothalamus; n = 8 episodes from two recording sessions and one animal in the pons; Kruskal–Wallis: F_(11,60) = 39.9, p < 0.05 in the hypothalamus; Kruskal–Wallis: F_(11,84) = 57.8, p < 0.05 in the pons, followed by multiple comparisons by the Bonferroni test: p < 0.05 vs the fourth epoch immediately before state transition), whereas it was almost constant in the cortex but was increased in the hippocampus (n = 15 episodes from fix recording sessions and three animals in the cortex; n = 5 episodes from two recording sessions and one animal in the hippocampus; Kruskal–Wallis: F_(11,168) = 0.98, p = 0.99, NS in the cortex; Kruskal–Wallis: F_(11,48) = 23.4, p < 0.05, followed by multiple comparisons by the Bonferroni test: NS vs the fourth epoch immediately before state transition in the hippocampus). At the time of transition from NREM sleep to wakefulness, Y/C ratios increased in the hippocampus (n = 12 episodes from 2 recording sessions and 1 animal in the hippocampus; Kruskal–Wallis: F_(11,132) = 89.3, p < 0.05, followed by multiple comparisons by the Bonferroni test: p < 0.05 vs the fourth epoch immediately before state transition in the hippocampus), but there was no significant difference in the cortex, hypothalamus, and pons (n = 34 episodes from seven recording sessions and three animals in the cortex; n = 25 episodes from two recording sessions and one animal in the hypothalamus; n = 7 episodes from one recording session and one animal in the pons; Kruskal–Wallis: F_(11,396) = 31.8, p < 0.05, followed by multiple comparisons by the Bonferroni test: NS vs the fourth epoch immediately before state transition in the cortex; Kruskal–Wallis: F_(11,288) = 6.0, p = 0.87, NS in the hypothalamus; Kruskal–Wallis: F_(11,72) = 5.4, p = 0.91, NS in the pons). There were consistent increases immediately after the state change from REM sleep to wakefulness in all brain regions that were monitored (n = 20 episodes from six recording sessions and three animals in the cortex; n = 6 episodes from two recording sessions and one animal in the hippocampus; n = 12 episodes from two recording sessions and one animal in the hypothalamus; n = 12 episodes from two recording sessions and one animal in the pons; Kruskal–Wallis: F_(11,228) = 109.5, p < 0.05 in the cortex; Kruskal–Wallis: F_(11,60) = 55.5, p < 0.05 in the hippocampus; Kruskal–Wallis: F_(11,132) = 46.8, p < 0.05 in the hypothalamus; Kruskal–Wallis: F_(11,132) = 61.2, p < 0.05 in the pons, followed by multiple comparisons by the Bonferroni test: p < 0.05 vs the fourth epoch immediately before state transition).

We also performed correlation analysis to investigate the association between episode duration of REM and changes in Ca²⁺ concentration among the brain regions (Fig. 3D). Changes in cortical and hippocampal Y/C ratios were not found to correlate with episode duration of REM sleep (n = 22 episodes from six recording sessions and three animals in the cortex; n = 6 episodes from two recording sessions and one animal in the hippocampus; r = 0.28, p = 0.21, NS in the cortex; r = 0.61, p = 0.20, NS in the hippocampus). In the hypothalamus and pons, however, episode duration of REM sleep and changes in Y/C ratios showed a significant negative correlation (n = 11 episodes from two recording sessions and one animal in the hypothalamus; n = 5 episodes from two recording sessions and one animal in the pons; r = −0.67, p < 0.05 in the hypothalamus; r = −0.97, p < 0.05 in the pons). These results suggest that the dynamics of astrocyte Ca²⁺ concentration vary depending on the brain region. A longer episode duration of REM sleep does not induce a further decrease in Ca²⁺ level in various regions of the brain, such as in the cortex and hippocampus.

Next, we analyzed the correlation between electrophysiological features from cortical EEGs and EMGs among the sleep/wakefulness states, and astrocyte Ca²⁺ concentrations, using the same methods as in Figure 2. As we compared cortical EEGs in this experiment, we focused on analyzing its correlation with the Ca²⁺ dynamics of cortical astrocytes (6 recording sessions and three animals). During wakefulness, Y/C ratios showed a weak negative correlation with α and β waves, and a weak positive correlation with γ waves (Fig. 4A,D,G). It was reported that an increase in the power value of δ and α waves and a decrease in the power value of γ waves indicate a decrease in the arousal level during wakefulness (McGinley et al., 2015). Cortical astrocyte Ca²⁺ levels appear to be low when mice are in quiet wakefulness. During NREM sleep, cortical Y/C ratios were positively correlated with δ waves, and were negatively correlated with θ, α, and β waves, consistent with those in the cerebellum (Fig. 4B,E,H). During REM sleep, however, Y/C ratios demonstrated a positive correlation with δ, and a negative correlation with θ and α waves (Fig. 4C,F,I). Although the Y scale has become too large to see the trend because of outliers, EMG power and Y/C ratio were positively correlated during both NREM and REM sleep (Fig. 4E,F). In this correlation analysis, minor but significant differences were observed also in the cerebellum. Compared with the results of the cerebellum, in the cortex, there was a tendency toward a correlation between astrocyte Ca²⁺ concentrations and electrophysiological features, not only during sleep but also during wakefulness.

Region-specific Ca²⁺ dynamics in astrocytes

We further quantified the regional differences in Ca²⁺ signals, including in the cerebellum. First, we quantified Ca²⁺ dynamics during each sleep/wakefulness episode, by categorizing the episodes into five states depending on the sleep/wakefulness state before and after each episode (Fig. 5A). Each episode was divided into five time-segments, and the average Ca²⁺ signal was calculated for each brain region (n = 6 recording sessions and three animals in the cortex; n = 2 recording sessions and one animal in the hippocampus; n = 4 recording sessions and two animals in the hypothalamus; n = 2 recording sessions and one animal in the pons; n = 5 recording sessions and three animals in the cerebellum).

Based on the results of Figures 3, 5A, there are likely to be three clusters in astrocyte Ca²⁺ signaling quality (Table 2). The first cluster is the cortex and hippocampus. Both cortical and hippocampal astrocytes showed a significant reduction in Ca²⁺ concentration during NREM sleep. No significant change was observed during the state transition from NREM sleep to REM sleep. In addition, no correlation was observed between astrocyte Ca²⁺ concentration and the duration of REM sleep episodes. The second cluster is the hypothalamus and pons. Astrocytes of both regions maintained their Ca²⁺ concentrations during NREM sleep, which decreased during REM. The third cluster is the cerebellum. In addition, Ca²⁺ signals during the state transition and during REM sleep episodes were completely identical between the hypothalamus and pons. Cerebellar astrocytes include characteristics of the other two clusters, i.e., their Ca²⁺ changes associated with sleep/wakefulness states tend to resemble those of the cortex and hippocampus. However, as in the hypothalamus and brainstem, Ca²⁺ changes negatively correlated with the durations of REM episodes.

Table 2.

Summary of region-specific and state-dependent astrocyte Ca²⁺ dynamics in mice

		Cortex	Hippocampus	Hypothalamus	Pons	Cerebellum
Normalized Y/C ratio	NR (vs W)	↓	↓	→	→	↓
	R (vs W)	↓	↓	↓	↓	↓
	R (vs NR)	↓	→	↓	↓	↓
State transition	W to NR	→	↓	→	→	↓
	NR to R	→	→	↓	↓	→
	NR to W	→	↑	→	→	↑
	R to W	↑	↑	↑	↑	↑
In episode	R (from NR, to W)	→	→	↓	↓	↓
	NR (from W, to R)	↓	↓	→	→	↓
Episode duration	R	→	→	↓	↓	↓

NR, NREM sleep; R, REM sleep; W, wakefulness.

Next, we analyzed the extent to which astrocyte Ca²⁺ signals can predict the ongoing sleep/wakefulness state, and whether there are any differences in decoding performance among the brain regions. For this purpose, average Ca²⁺ signals were computed in each epoch, and a linear classifier was trained with fourfold cross validation (Tsunematsu et al., 2020). The decoding performance based on Y/C ratio of hippocampus and cerebellum in Mlc1-tTA; TetO-YCnano50 mice was significantly higher than that in Mlc1-tTA mice (p < 0.05, one-way ANOVA; Fig. 5B). This result indicates that hippocampal and cerebellar astrocyte activity demonstrate sleep/wakefulness state-dependency. In addition, we also clarified the differences in decoding performance among the different brain regions. Decoding performance of the cerebellum was significantly higher than that of the cortex (p < 0.05, one-way ANOVA; Fig. 5B). Thus, the physiological role of astrocytes in sleep/wakefulness might vary depending on the brain region.

Discussion

In the present study, we analyzed astrocyte Ca²⁺ dynamics during different sleep/wakefulness states among various brain regions. We demonstrated that astrocyte Ca²⁺ concentrations decreased during sleep, reached a minimum during REM sleep, and increased during wakefulness in all brain regions that were recorded. Further analyses indicated that there are at least three astrocyte clusters, comprising astrocytes from different brain regions. Although the association between changes in astrocyte Ca²⁺ concentrations and sleep/wakefulness were consistent in general among the brain regions, we found that detailed Ca²⁺ dynamics varies depending on the brain region.

Technical differences compared with other similar studies

Several recent studies measuring astrocyte Ca²⁺ concentrations during sleep/wakefulness states have been reported (Bojarskaite et al., 2020; Ingiosi et al., 2020). The results of these studies are in good agreement with our results showing that astrocyte Ca²⁺ concentrations/signals decrease during sleep and increase during wakefulness. However, there are several differences. We showed a decrease in Ca²⁺ concentration in the cortex during NREM sleep, which is consistent with previous studies (Bojarskaite et al., 2020; Ingiosi et al., 2020). However, our present study demonstrated that no increase in Ca²⁺ level was observed during the transition from wakefulness to NREM sleep in the cortex, which is inconsistent with a recent study (Ingiosi et al., 2020). Moreover, both papers reported that the Ca²⁺ concentration once increased after transition from sleep to wakefulness then slightly decreased within 15 s. Our study, however, did not show a Ca²⁺ overshoot in any of the recorded brain regions, although we showed 32 s of data posttransition. These differences may be because of differences experimental conditions.

The first difference is the Ca²⁺ sensor that was used. We used the fluorescence resonance energy transfer-based, ultrasensitive genetically encoded Ca²⁺ indicator YCnano50 (Horikawa et al., 2010). On the other hand, other studies have used GCaMP6f (Chen et al., 2013). YCnano50 has high Ca²⁺ affinity (Kd = 50 nm) compared with GCaMP6f (Kd = 375 nm). Because of the ability of YCnano50 to detect subtle basal changes in the Ca²⁺ concentrations, our study was able to clarify the distinct differences in astrocyte Ca²⁺ concentrations between NREM and REM sleep.

The second difference is the method of expression of the Ca²⁺ sensor. Previous studies expressed the sensor using an adeno-associated virus, whereas we used Mlc1-tTA; TetO-YCnano50 bigenic mice to express the sensor in an astrocyte-specific manner. Because we used genetically modified mice rather than virus infection, the expression level and pattern of YCnano50 were constant among mice, and hence consistent data could be obtained. Although the sensors were substantially expressed in the cerebellum, pons, and hypothalamus, the sparce expression was observed in the cortex, and hippocampus in the genetically modified mice. It cannot exclude the possibility that the expression level of YCnano50 have made a difference from previous studies.

The third difference is the optical imaging approach that was used. As we recorded Ca²⁺ levels in astrocytes using the fiber photometry system, it was possible to analyze deeper regions of the brain, such as the hypothalamus and pons. However, our method only enables the measurement of the sum of changes in Ca²⁺ concentrations of cells. Our previous fiber photometry experiments using the same fiber optics and similar light intensities demonstrated that the detection limit is at a depth of ∼700 µm (Natsubori et al., 2017). Assuming a tissue refractive index of 1.5, we detected signals with a diameter of ∼800 µm at the 700 µm from tips of fiber optics. It indicates that the sum of Ca²⁺ signals of multiple nuclei and subregions were recorded although signals from specific brain region were detected at least. The Ca²⁺ dynamics at the nuclei, single-cell and subcellular level remains unknown. In recording from deep region, especially from the hypothalamus, the implanted optical fiber damaged the tissue and caused similar to glial scaring. Although there was no effect on the sleep/wakefulness state, the conclusion should be handled with care. Further detailed analyses are needed.

Nevertheless, taking advantage of our method, we succeeded in comparing astrocyte Ca²⁺ dynamics among various brain regions.

Correlation between electrophysiological features and astrocyte Ca²⁺ concentrations

Here, we investigated the correlation of cerebellar and cortical astrocyte Ca²⁺ concentrations with cortical oscillations. During NREM sleep, we found that consistent negative correlations between Y/C ratio and α/β power across recordings in the cerebellum and the cortex. Although β oscillations have long been implicated in long-range communication between brain regions (Fries, 2015), the function of such middle frequency oscillations during NREM sleep remains unclear. During REM sleep, we found negative correlations between Y/C ratio and θ power. Because θ power is a prominent biomarkers of REM sleep (Brown et al., 2012), astrocytic Ca²⁺ signals in the cortex and the cerebellum may reflect the depth and/or quality of REM sleep.

Possible mechanisms regulating astrocyte Ca²⁺ concentrations

We observed similar Ca²⁺ concentration changes in different brain regions. This was an interesting result that was inconsistent with neural activity. For instance, cortical neurons activate during wakefulness and REM sleep (Vyazovskiy et al., 2009; Watson et al., 2016; Niethard et al., 2017), hippocampal neurons fire less during REM sleep (Grosmark et al., 2012; Miyawaki and Diba, 2016), and an increase in the firing rate in the brainstem is observed during REM sleep (Hobson et al., 1975; Weber et al., 2015; Tsunematsu et al., 2020). Thus, it has been reported that neural activity patterns vary depending on the brain region.

It has also been reported that Ca²⁺ concentrations are affected by G-protein-coupled receptors (GPCRs) expressed in astrocytes via neurotransmitter release that accompanies neural activity. In general, the activation of astrocyte GPCRs increases astrocyte Ca²⁺ levels (Cornell-Bell et al., 1990; Takata et al., 2011; Jacob et al., 2014; Corkrum et al., 2020), although in some instances a decrease in Ca²⁺ level has been reported (Jennings et al., 2017). However, we observed global Ca²⁺ concentration changes, and therefore it is possible that a neurotransmitter that shows varying release patterns throughout the brain during different sleep/wakefulness states controls astrocyte Ca²⁺ concentration. Noradrenaline has been reported to increase astrocyte Ca²⁺ levels (Bekar et al., 2008; Paukert et al., 2014; Oe et al., 2020). Noradrenergic neurons located in the locus coeruleus project to the entire brain. Changes in firing rates of noradrenergic neurons showed a similar pattern to the Ca²⁺ dynamics of astrocytes (Takahashi et al., 2010; Tsujino et al., 2013). Taken together, noradrenaline released from noradrenergic neurons during wakefulness might increase astrocyte Ca²⁺ concentrations throughout the brain. In addition, considering that astrocyte Ca²⁺ levels gradually decrease during REM sleep, noradrenaline might also act as a volume transmitter, because microdialysis studies have reported that the concentration of noradrenaline in the brain decreases during sleep (Park, 2002; Bellesi et al., 2016). It has been reported that not only noradrenaline but also glutamate, acetylcholine, and GABA increase astrocyte Ca²⁺ levels (Cornell-Bell et al., 1990; Kang et al., 1998; Araque et al., 2002; Sun et al., 2013; Perea et al., 2016). Thus, the effects of these neurotransmitters as well as neuropeptides which regulate sleep/wakefulness state should be elucidated.

Possible heterogeneity of astrocytes among various brain regions

The results of this study surprisingly implicate that astrocytes in different brain regions may have different functions in sleep/wakefulness. Based on our results, we classified astrocytes into the following three clusters: cluster 1, astrocytes in the cortex and hippocampus; cluster 2, astrocytes in the hypothalamus and pons; and cluster 3, astrocytes in the cerebellum.

Astrocytes have recently been clarified to be a heterogeneous population. Transcriptional analyses have demonstrated that astrocyte gene expression patterns differ among and within brain regions and can be classified into several types (Chai et al., 2017; Zeisel et al., 2018; Batiuk et al., 2020; Bayraktar et al., 2020). The gene expression patterns indicated that cortical and hippocampal astrocytes have similar transcriptional profiles (Morel et al., 2017; Lozzi et al., 2020). In contrast, Bergmann glial cells are a type of cerebellar astrocyte with unique morphologic and transcriptional characteristics, although there are other glial cells, i.e., velate astrocytes, in the cerebellum. Bergmann glial cells express Ca²⁺-permeable AMPA receptors composed of the GluA1 and GluA4 subunits (Saab et al., 2012), implying that they have different intracellular Ca²⁺ dynamics. Our results may explain part of the differences in the functions of astrocytes depending on the brain region, as well as the differences in their cellular transcriptomes. In our study, the number of mice in which the hippocampus and pons was analyzed was limited, although the effect size was 1.34 and 1.13, respectively. Furthermore, the effect size calculated from the data of the cortex was 0.47, indicating a medium effect. Thus, further analyses are required before making any definite conclusions.

Decoding performance for ongoing sleep/wakefulness states was significantly higher in the hippocampus and cerebellum of Mlc1-tTA; TetO-YCnano50 mice than Mlc1-tTA mice, suggesting that astrocyte Ca²⁺ dynamics might not only show state-dependent fluctuation, but also contribute to the control of the sleep/wakefulness state itself. Further research should be performed with the caveat that astrocyte functions in sleep/wakefulness states might vary among different brain regions.