Abstract
Mammalian sleep is regulated by a homeostatic process that increases sleep drive and intensity as a function of prior wake time. Sleep homeostasis has traditionally been thought to be a product of neurons, but recent findings demonstrate that this process is also modulated by glial astrocytes. The precise role of astrocytes in the accumulation and discharge of sleep drive is unknown. We investigated this question by selectively activating basal forebrain (BF) astrocytes using designer receptors exclusively activated by designer drugs (DREADDs) in male and female mice. DREADD activation of the Gq-protein-coupled pathway in BF astrocytes produced long and continuous periods of wakefulness that paradoxically did not cause the expected homeostatic response to sleep loss (e.g., increases in sleep time or intensity). Further investigations showed that this was not because of indirect effects of the ligand that activated DREADDs. These findings suggest that the need for sleep is not only driven by wakefulness per se, but also by specific neuronal-glial circuits that are differentially activated in wakefulness.
SIGNIFICANCE STATEMENT Sleep drive is controlled by a homeostatic process that increases sleep duration and intensity based on prior time spent awake. Non-neuronal brain cells (e.g., glial astrocytes) influence this homeostatic process, but their precise role is unclear. We used a genetic technique to activate astrocytes in the basal forebrain (BF) of mice, a brain region important for sleep and wake expression and sleep homeostasis. Astroglial activation induced prolonged wakefulness without the expected homeostatic increase in sleep drive (i.e., sleep duration and intensity). These findings indicate that our need to sleep is also driven by non-neuronal cells, and not only by time spent awake.
Introduction
An evolutionarily conserved feature of sleep is that it is regulated by two major processes (Borbély et al., 2016). A biological clock provides timing (circadian) signals that organize sleep and wakefulness across the 24-h day (Fisher et al., 2013). A homeostatic mechanism increases sleep drive and (in some species) intensity as a function of prior wakefulness (Allada et al., 2017). Great progress has been made characterizing the circadian regulation of sleep and wakefulness. For example, the anatomic location of a principal biological clock and the key molecular mechanisms of timekeeping are known (Rosenwasser and Turek, 2022). In contrast, far less is known about the cellular basis of sleep homeostasis. No master mammalian homeostat has been discovered and comparatively little is known about the molecular cascades in brain cells necessary for sleep homeostasis (Allada et al., 2017). Nevertheless, until recently, the traditional view has been that sleep homeostasis is a product of neurons (Frank, 2013).
We showed that sleep homeostasis is also dependent on glial astrocytes (Halassa et al., 2009; Ingiosi et al., 2020; Ingiosi and Frank, 2022). Astrocytes perform many critical functions in the brain that make them well-positioned to mediate sleep homeostasis (Ingiosi and Frank, 2023). For example, astrocytes express receptors for wake-promoting neuromodulators like noradrenaline (Ding et al., 2013; Martin-Fernandez et al., 2017), and astrocytes release sleep-promoting substances like ATP/adenosine which alters neuronal activity in canonical sleep-wake nuclei (Halassa et al., 2009; Kim et al., 2014; Choi et al., 2022). In addition, conditional inhibition of gliotransmission in astrocytes inhibits sleep drive, as measured by sleep-wake behavior and electroencephalograph (EEG) activity (Halassa et al., 2009). Under these conditions, transgenic mice show less compensatory responses to sleep loss, which suggests that they can stay awake with less accumulating sleep drive. Conditional deletion of astrocytic membrane-bound receptors to circulating neuromodulators (Ingiosi and Frank, 2022) or intracellular calcium regulating proteins produces similar effects (Ingiosi et al., 2020). This suggests that astrocytes respond to extracellular signals (ligands) originating from neurons (e.g., classic neurotransmitters or neuromodulators).
To explore this further, we investigated a key pathway that links membrane-bound receptors to secondary intracellular cascades in astrocytes. More specifically, mammalian astrocytes express in vivo several membrane-bound receptors that couple to G-proteins. Therefore, experimentally manipulating these pathways engages native mechanisms present in vivo. This can be accomplished by using selective astroglial expression of designer receptors exclusively activated by designer drugs (DREADDs) coupled to the Gq-protein pathway. In astrocytes, Gq activates the phospholipase C pathway which increases intracellular calcium through inositol 1,4,5-trisphosphate-mediated release from internal stores (Chai et al., 2017; Durkee et al., 2019; Shen et al., 2021; Vaidyanathan et al., 2021). Muscarinic acetylcholine receptors (Durkee et al., 2019), group I metabotropic glutamate receptors (Spampinato et al., 2018), and histamine H1 receptors (Kárpáti et al., 2019) are expressed by astrocytes and trigger Gq-protein-coupled cascades. Research on the impact of astroglial Gq-DREADD activation on sleep expression is limited, but recent studies show the effects are region-specific. Gq-DREADD activation of cortical astrocytes increases nonrapid eye movement sleep (NREMS) duration (Vaidyanathan et al., 2021) whereas astroglial activation in the lateral hypothalamus increases wakefulness (Cai et al., 2022). However, the impact of astroglial DREADD activation on sleep homeostasis has yet to be explored.
We examined the effects of chemogenetically activating the Gq pathway in basal forebrain (BF) astrocytes. This is because the BF is comprised of several classes of neurons that influence sleep and wake time as well as sleep homeostasis (Brown et al., 2012; Yang et al., 2017). Furthermore, the role of BF astrocytes in sleep homeostasis has not been investigated. We find that activating Gq-DREADDs in BF astrocytes leads to hours of sustained wakefulness without the normal compensatory changes in sleep time or intensity.
Materials and Methods
Animals
B6;FVB-Tg(Aldh1l1-cre)JD1884Htz/J (Aldh1l1-Cre; 023748) and B6N;129-Tg(CAG-CHRM3*,-mCitrine)1Ute/J (hM3Dq; 026220) mice were obtained from The Jackson Laboratory (Bar Harbor, ME, USA). Heterozygous Aldh1l1-Cre+/− male mice were bred with heterozygous hM3Dqfl/− female mice to produce Aldh1l1-Cre−/−; hM3Dqfl/− control (Ctrl) mice and Aldh1l1-Cre+/−; hM3Dqfl/− (Gq+/−) experimental littermates. The non-DREADD-expressing Ctrl mice served as both surgical and pharmacological controls for any potential indirect effects of our procedures. Mice were housed in standard cages at an ambient temperature of 24 ± 1°C on a 12/12 h light/dark cycle with food and water ad libitum. All experimental procedures were approved by the Institutional Animal Care and Use Committee of Washington State University and conducted in accordance with National Research Council guidelines and regulations for experiments in live animals.
Surgical procedures
EEG and EMG electrode and cannulae implantation
Adult male and female mice [Ctrl (n = 12; n = 6 females) and Gq+/− (n = 10; n = 5 females); 10–14 weeks old)] were anesthetized with isoflurane and stereotaxically implanted with two chronic guide cannulae (C315GS-5/SPC; P1 Technologies) in basal forebrain (from bregma AP: 0.0 mm, ML: ±1.62 mm, DV: −5.0 mm; Zant et al., 2016; McKenna et al., 2020). Two additional adult male mice (n = 1 Ctrl; n = 1 Gq+/−; 13 weeks old) were implanted with 1 chronic guide cannula in the right lateral ventricle (from bregma AP: −0.5 mm, ML: +1.25 mm, DV: −2.0 mm; Ingiosi et al., 2015) to assess global impacts of astroglial DREADD activation. All mice were also implanted with four electroencephalographic (EEG) screw electrodes (AMS120/3; Antrin Miniature Specialties) contralaterally over frontal (2) and visual (2) cortices, and two electromyographic (EMG) wire electrodes were implanted in the nuchal muscles as previously described (Ingiosi et al., 2020; Ingiosi and Frank, 2022). EEG electrodes and guide cannulae were fixed to the skull with dental acrylic. Patency of the cannulae was maintained with indwelling dummy cannulae (C315DCS-5/SPC; P1 Technologies). Mice were allowed to recover from surgery for at least 7 d before habituation to the recording environment.
Telemeter and cannulae implantation
We assessed changes in core temperature and gross motor activity to ensure that DREADD activation did not produce abnormal changes in physiology or behavior that might indirectly impact sleep-wake expression and regulation. A separate group of adult male and female Gq+/− mice (n = 5; n = 3 female; 9–14 weeks old) were anesthetized with isoflurane and implanted with a telemetry device (G2 E-mitter; STARR Life Sciences Corp.) in the peritoneal cavity as previously described (Ingiosi et al., 2020; Ingiosi and Frank, 2022). A suture was used to secure the telemeter to the abdominal musculature, and wound clips were used to close the skin. Mice were then implanted with 2 chronic guide cannulae in the BF as described above and capped with indwelling dummy cannulae. Two anchor screws (Antrin Miniature Specialties) were placed contralaterally over frontal and visual cortices. Guide cannulae and anchor screws were fixed to the skull with dental acrylic. Body weight, hydration, and fecal output were monitored daily for 8 d after surgery at which point wound clips were removed.
Experimental design: sleep and polysomnographic analyses
CNO experiments
After recovery from cannulae and EEG/EMG electrode implantation, mice were individually housed in polycarbonate recording cages and connected to a lightweight, flexible recording cable (Ingiosi et al., 2019, 2020; Ingiosi and Frank, 2022). Mice habituated to the recording cable for at least 3 d before data collection. Once habituated, baseline (BL) EEG and EMG data were collected for 24 h starting at light onset while mice were left undisturbed. The next day, either clozapine N-oxide dihydrochloride (CNO; 0.36 mm, i.e., 0.003 mg/kg for a 25-g mouse; HB6149: Hello Bio Inc.) or vehicle (saline) was injected intracranially (Stachniak et al., 2014; Barbano et al., 2020) into the BF via guide cannulae (250 nl per cannulae at ∼50 nl/min; 500 nl for an intracerebroventricular cannula) in freely behaving mice using an internal cannula (C315IS-5/SPC; P1 Technologies) attached to a Hamilton syringe by silicone tubing (2415500; Dow Corning Corporation). Current evidence suggests intracranial delivery of CNO circumvents the off-target effects observed after systemic injections (Mahler and Aston-Jones, 2018). CNO and vehicle injections occurred during Zeitgeber time (ZT)0–ZT1 within 45 min of light onset [18.56 ± 13.01 min (mean ± SD) from light onset] using a counterbalanced schedule separating injections by at least 48 h. Mice were left undisturbed after each injection, and EEG and EMG data were collected for 24 h. This approach allowed us to make within-subject vehicle versus CNO comparisons as well as Ctrl versus Gq+/− between-subject comparisons.
Sleep deprivation and J60 control experiments
We performed two additional experiments in the same Ctrl and Gq+/− mice that received vehicle and CNO. We first measured the homeostatic response to 6 h of sleep deprivation (SD) to determine whether the mice had an intact homeostatic response to SD. This would also provide comparison data to changes following DREADD activation in the Gq+/− mice. At least 48 h after the CNO/vehicle injections, Ctrl and Gq+/− mice underwent 6-h SD via gentle handling starting at light onset as previously described (Halassa et al., 2009; Ingiosi et al., 2020; Ingiosi and Frank, 2022). SD via gentle handling involves arousing mice (e.g., tactile, auditory stimuli) when their EEG/EMG and/or behavior (e.g., posture, quiescence) is predictive or indicative of sleep. Mice were then left to recover undisturbed for 18 h. Post-SD data were compared (and/or normalized where applicable) to time-matched values from the BL day.
Second, we determined whether a different DREADD ligand [JHU37160 dihydrochloride (J60); HB6261; Hello Bio Inc.] reproduced the sleep-wake effects of CNO in Gq+/− mice. This was prudent because CNO has been shown to have off-target effects on sleep-wake expression in non-DREADD-expressing mice (Varin et al., 2018; Traut et al., 2023). J60 is a potent ligand with high DREADD selectivity and occupancy (Bonaventura et al., 2019; Peeters et al., 2020; Costa et al., 2021). In addition, no off-target effects of J60 have been reported thus far for a variety of measures (Bonaventura et al., 2019; Zhang et al., 2020; Costa et al., 2021; Fleury Curado et al., 2021; Desloovere et al., 2022), including sleep time in C57Bl/6J mice (Fleury Curado et al., 2021). Following the SD experiment, mice were allowed at least 48 h of additional recovery and then were injected intracranially with 0.35 mm J60 (i.e., 0.003 mg/kg for a 25-g mouse) during ZT0 – 1 [250 nl per cannulae at ∼50 nl/min; 17.75 ± 13.0 min (mean ± SD) from light onset], and EEG and EMG data were recorded for 24 h. These data were compared with results obtained from vehicle treatments used in the CNO experiments.
Polysomnography and EEG data analyses
EEG and EMG data were collected using a GRASS 7 polygraph system (Natus Medical Incorporated) via a lightweight recording cable. The signals were amplified, digitized, and processed at 128 Hz using VitalRecorder acquisition software (v3.0.0.0; SleepSign for Animal, Kissei Comtec Co, LTD). EEG and EMG data were high-pass and low-pass filtered at 0.3 and 100 Hz and 10 and 100 Hz, respectively (Ingiosi et al., 2020; Ingiosi and Frank, 2022). Wakefulness, NREMS, and rapid eye movement sleep (REMS) were determined from EEG and EMG data by visual inspection of the EEG waveform, EMG activity, and fast Fourier transform (FFT) using SleepSign for Animal (v3.3.8.1803; Kissei Comtec Co, Ltd.). Vigilance states were scored in 4-s epochs by an investigator blinded to experimental conditions (Ingiosi et al., 2020; Ingiosi and Frank, 2022). These data were then used to calculate time-in-state, bout duration, bout frequency, and latency to state. For vehicle, CNO, and J60 injection days, only postinjection data were included in calculations. Time-in-state was expressed as a percentage of total recording time (TRT) in 2-h bins. Minimum bout lengths were defined as ≥7 consecutive epochs (≥28 s) for wakefulness and NREMS and ≥4 consecutive epochs (≥16 s) for REMS (Ingiosi et al., 2020; Ingiosi and Frank, 2022). Frequency and duration of vigilance state bouts were shown as differences from a control condition by subtracting time-matched vehicle data from CNO or J60 data or time-matched BL data from SD data (Ingiosi and Opp, 2016; Ingiosi et al., 2019; Ingiosi and Frank, 2022). These differences were then expressed in 6-h bins. Latency to NREMS and REMS was calculated in two ways. First, we determined elapsed time postinjection or post-SD before a bout of average length which was calculated from the 24-h vehicle data (or 24-h BL data for SD). We also calculated the latency to the first 4-s epoch of NREMS and REMS, an analysis which makes no assumption about minimum bout duration.
The EEG was fast Fourier transformed to produce power spectra between 0 and 60 Hz in 0.5-Hz bins (Ingiosi and Frank, 2022). All spectral data were normalized to undisturbed BL EEG spectra. Each spectral bin was expressed as a percentage of the total power in BL wakefulness, NREMS, and REMS averaged across the three vigilance states for the entire 24-h BL period (Ingiosi et al., 2020; Ingiosi and Frank, 2022). We defined δ as 0.5–4 Hz, low δ as 0.5–1.5 Hz, high δ as 2–4 Hz, θ as 5–9 Hz, α as 10–15 Hz, β as 15–30 Hz, and γ as 30–60 Hz. As not all mice were equally awake or asleep in any given time bin, we used the following rules for calculating mean changes in EEG spectra as described previously (Franken et al., 2001). A mouse had to spend ≥5 min in wakefulness or NREMS or ≥1 min in REMS (per time bin) for that EEG data to be included in the mean state analyses. In addition, for statistical comparisons, at least five mice per condition had to contribute to EEG spectra measurements per time bin. EEG epochs with visually detected artifacts were excluded from spectral analyses. Similar rules were applied to hourly changes in NREMS δ power, defined as mean FFT power in δ (0.5–4 Hz), low δ (0.5–1.5 Hz), or high δ (2–4 Hz; Halassa et al., 2009; Hubbard et al., 2020; Ingiosi et al., 2020; Ingiosi and Frank, 2022). FFT power within the δ band, low δ band, or high δ band for each time bin postinjection or post-SD recovery were normalized to the average NREMS δ, low δ, or high δ band value, respectively, from the last 4 h of the undisturbed BL light period (h9–12) and expressed as a percentage shown in 2-h bins (adapted from Franken et al., 1991).
Experimental design: measurement of core temperature and activity
We measured changes in core temperature and gross activity after CNO and J60, because if present, they might indirectly alter sleep. After postoperative recovery from telemeter and cannulae implantation, mice were individually housed in standard mouse cages and given at least 5 d to habituate to the recording environment. As described above, mice were injected with either CNO or vehicle during ZT0–ZT1 using a counterbalanced design separating the injections by 72 h. Three days later, all mice were injected with J60 during ZT0–ZT1 as described above. After each injection, mice were left undisturbed, and core body temperature and cage activity were monitored continuously for 24 h. Core body temperature (°C) and gross cage activity (counts) data were captured from the abdominal telemetry device, transmitted to an energizer/receiver (ER-4000; STARR Life Sciences Corp.), and recorded with VitalView software (v5.1; STARR Life Sciences Corp.). Data were collected every minute for 24 h postinjection and averaged across 2-h bins to determine 24-h diurnal/nocturnal patterns in response to vehicle, CNO, and J60.
Immunohistochemistry
We used immunohistochemistry to (1) confirm astroglial-specific expression of the DREADDs and (2) verify ligand activation of DREADDs in astrocytes via astroglial cFos expression. Brains were obtained from the same mice used for sleep phenotyping at least 4 d after J60 injections. Ctrl (n = 3, females = 2) and Gq+/− (n = 3, females = 2) mice were injected with 0.36 mm CNO bilaterally in the BF (250 nl per cannulae at ∼50 nl/min) during ZT0–ZT1, as described, and left undisturbed for ∼90 min. Mice were then transcardially perfused with 1× PBS followed by 10% buffered formalin. Brains were extracted and immediately postfixed in 10% buffered formalin overnight at 4°C. To protect against freezing artifacts, brains were transferred to 4°C 30% sucrose in 1× PBS for 24–48 h and subsequently frozen and stored at −80°C until processing.
Frozen brains were sectioned coronally at 30 µm on a Thermo-Fisher CryoStar NX50 cryostat in a series of six and stored free-floating in cryoprotectant. For immunohistochemical staining, free-floating sections were first washed 3 × 10 min in 1× PBS and then incubated for 30 min in a blocking solution of 2% normal goat serum (S-1000-20; Vector Laboratories) and 0.1% Triton X-100 (T8787-50ML; Sigma-Aldrich). Tissue was then incubated for 35–40 h at 4°C on a gentle rocker with primary antibodies diluted in blocking solution. Primary antibodies were used to amplify native mCitrine signal (1:1000, polyclonal guinea pig anti-GFP; 132-004; Synaptic Systems; detects GFP variants including yellow fluorescent proteins like mCitrine; Griesbeck et al., 2001) and to identify astrocytes (1:1000, monoclonal rabbit anti-S100β; ab52642; Abcam), cFos expression (1:1000, monoclonal mouse anti-cFos; 309-cFOS; PhosphoSolutions), and neurons [1:1000, monoclonal mouse anti-NeuN; RBFOX/NeuN (1B7); Novus Biologicals]. Sections were then incubated for 1 h at room temperature with the following Alexa Fluor fluorophore-conjugated secondary antibodies diluted 1:1000 in blocking solution to generate fluorescence contrast of the primary antibodies for confocal detection: goat anti-guinea pig 488 (for GFP; A11073; Thermo Fisher Scientific), donkey anti-rabbit 594 (for S100β; ab150132; Abcam), goat anti-mouse (IgG1) 647 (for cFos; A21240; Thermo Fisher Scientific), and goat anti-mouse (IgG1) 594 (for NeuN; A21125; Thermo Fisher Scientific). After contrasting, sections were washed three times in 1× PBS, mounted onto generic 50 × 70 mm glass slides using 24 × 50 mm #1.5 thickness Gold Seal 3422 coverslips (50-189-9137; Fisher Scientific) and DAPI Fluoromount-G (17984-24; Electron Microscopy Sciences).
An inverted Leica Microsystems DMi8 laser scanning microscope was used for image capture. Briefly, to identify sections containing basal forebrain structures, slides containing multiple mounted sections were first tile imaged at 5× with polarized light using the “Navigator” function in Leica Application Suite X software. Stereotaxic coordinates of our sections were then identified based on visual comparison against an atlas (Franklin and Paxinos, 2008). Using the Navigator function, BF structures were then digitally circumscribed to define a region of interest and optically sliced to capture volumes for quantification (15 1-µm thick sections at 512 × 512 pixels/frame using 20× HC PL APO 0.75 NA CS2 objective, 1.5× digital zoom). We primarily focused on the medial septum (MS) and the vertical limb of the diagonal band of Broca (VDB; +0.74 – +1.1 from bregma), regions that did not show cannulae-induced autofluorescent artifacts or gliosis. To quantify nuclear cFos expression in BF astrocytes we required (1) that the fluorescence signals from DAPI and cFos completely colocalize in the x-y-plane in a max-projection and when viewing the volume, be completely colocalized along the z-axis by optical cross-section and (2) colocalization of DAPI+/cFos+ nuclei with S100β tertiary labeling. Counts were obtained from 10 whole mount brain sections from three Ctrl mice [three to four brain sections/mouse, 26 regions of interest (ROIs)] and 10 brain sections from three Gq+/− mice (three to four brain sections/mouse, 27 ROIs). Each ROI was 0.13 mm2. Density of cFos+S100β+ double labeling was calculated as a percentage by dividing the total number of double-labeled cFos+S100β+ cells by the total number of S100β+ cells and multiplying by 100 for each brain section.
Statistical analyses
Plots were generated using SigmaPlot (v11.0, Systat Software) and R (v4.1.1). SPSS for Windows 25 (IBM Corporation) was used for statistical analyses. Data are shown as means ± SEM. Normality of the data was determined via Shapiro–Wilk or Kolmogorov–Smirnov tests. Normally distributed data were assessed with parametric tests: paired t test, unpaired t tests, one sample t test, general linear model for repeated measures (RM; repeated measures ANOVA). In cases when data were not normally distributed or there were missing cells, we used nonparametric tests: Wilcoxon signed-rank tests, Mann–Whitney U tests, and Kruskal–Wallis tests. Comparison of cFos+S100β+ density was made with a Mann–Whitney U test. RM using time (hours) as the repeated measure and either genotype (Ctrl vs Gq+/−) as the between-subjects factor or treatment (vehicle vs CNO or J60; BL vs SD) as the within-subjects factor was used for comparisons of several sleep and physiological measurements. These included time-in-state, bout frequency, bout duration, BL versus SD NREMS EEG δ power, core body temperature, and cage activity. For postinjection data, time-in-state (as % TRT), bout frequency, and bout duration RM comparisons were made over all time intervals for the full 24-h period. BL versus SD time-in-state and NREMS δ power was assessed over the initial 6 h recovery period in the light phase (h7–12). RM was also used for comparisons of EEG power spectra using frequency (Hz) as the repeated measure and treatment (vehicle vs CNO or J60) as a within-subjects factor over the full 0- to 30-Hz range, or, as indicated, over the δ (0.5–4 Hz) and θ (5–9 Hz) frequency ranges. RM comparisons were tested for sphericity, and a Greenhouse–Geisser correction was applied when needed. Post hoc pairwise comparisons with Sidak corrections for multiple comparisons were performed when there were significant interaction effects or main effects of genotype or treatment. Comparisons of vehicle versus CNO or J60 on NREMS δ power, low NREMS δ power, and high NREMS δ power were made using Kruskal–Wallis tests. Sidak corrections for multiple comparisons were applied to post hoc pairwise comparisons. One sample t tests were used to determine whether changes in bout frequency or bout duration differed from 0. Paired t tests or Wilcoxon signed-rank tests were used as appropriate for comparisons of γ band power as well as sleep latency between treatments (vehicle vs CNO or J60). Unpaired t tests or Mann–Whitney U tests were used as appropriate for comparisons of sleep latency between genotypes (Ctrl vs Gq+/−). When possible, sex was entered as a between-subjects factor separately for Ctrl and Gq+/− mice for the measures described above. An α level <0.05 was used to indicate significance. To facilitate readability, most statistical results are provided as tables (Table 1–Table 9).
Statistical output from repeated measures ANOVA comparisons of time in state, number of bouts per hour, and bout duration
Results
CNO activation of astroglial Gq-DREADDs in BF promotes sustained wakefulness
We first determined the impact of activating BF astrocytes on sleep-wake architecture and EEG activity. To do this, we expressed Gq-DREADDs selectively in astrocytes by crossing Aldh1l1-Cre+/− mice with hM3Dqfl/− mice to produce Aldh1l1-Cre+/−; hM3Dqfl/− (Gq+/−) mutant offspring and Aldh1l1-Cre−/−; hM3Dqfl/− control (Ctrl) littermates (Fig. 1A–C). We then activated astroglial Gq DREADDs by delivering the DREADD ligand clozapine N-oxide dihydrochloride (CNO) directly to the BF. We adopted procedures from previously published work (Adamsky et al., 2018) to verify CNO increased activity in BF astrocytes (Fig. 1D–F). As shown in Figure 1F, CNO in Gq+/− mice significantly increased cFos expression in BF astrocytes compared with CNO in Ctrl mice (Mann–Whitney U, U = 5.00, p < 0.001).
Gq-DREADDs are expressed in astrocytes of Aldh1l1-Cre+/− hM3Dqfl/− (Gq+/−) mutant mice and CNO induces astroglial cFos expression in basal forebrain (BF) of Gq+/− mice. A, The surrogate mCitrine reporter for Gq-DREADD expression is not detected in BF of control (Ctrl) mice that do not express Gq-DREADDs when amplified by an anti-GFP antibody nor does it co-localize with S100β-labeled astrocytes. Inset shows nonspecific fluorescence observed in a no primary antibody (anti-GFP) control section incubated with secondary antibody only. Scale bar is 50 µm. B, C, The Gq-DREADD mCitrine reporter amplified by anti-GFP colocalizes with the B, astroglial marker S100β but not with the C, neuronal marker NeuN in BF of Gq+/− mice. Inset shows mCitrine (anti-GFP) and S100β colocalization at a higher magnification. Scale bar for B is 50 µm (inset is 25 µm). Scale bar for C is 25 µm. D, CNO-induced cFos expression in S100β-labeled astrocytes in Gq+/− BF. Examples of cFos+S100β+ double-labeled cells are marked with yellow arrowheads. Scale bar is 25 µm. E, Representative brain section showing the medial septum (MS) and vertical diagonal band of Broca (VDB) circumscribed by the dashed yellow line (bregma +1.1). F, CNO increases cFos expression in BF astrocytes (labeled with S100β) in Gq+/− mice compared with CNO-treated Ctrl (Mann–Whitney U, *p < 0.05). Values are from three Ctrl mice (26 ROIs from 10 brain sections; 3–4 brain sections/mouse) and from three Gq+/− mice (27 ROIs from 10 brain sections; 3–4 brain sections/mouse). Brain sections were collected ∼90-min post-CNO injection.
We found that activating BF astroglial Gq-DREADDs in Gq+/− mice induced long periods of continuous wakefulness (≥6 h) and an overall increase in wake time for the entire light phase (Figs. 2, 3A; Tables 1, 2) when CNO was injected during ZT0–ZT1. This was accompanied by increased latencies to sleep (Fig. 4; Table 3) as well as reduced sleep time, bout frequency, and bout duration for NREMS and REMS for the entire 12-h light period (Fig. 3; Tables 1, 2). During the dark period, Gq+/− sleep was more fragmented after CNO compared with vehicle as reflected by shorter, more frequent bouts of wakefulness and NREMS (Fig. 3; Tables 1, 2). Dark period sleep time after CNO did not differ from vehicle values except for a transient increase in NREM sleep time in the first 2 h of the dark phase (h13–14; Fig. 3A).
Statistical output from one-sample t tests (compared with 0) for the change (from vehicle) in the number of bouts per hour and bout duration after CNO or J60
Statistical output for comparisons of latencies to NREMS and REMS after vehicle, CNO, or J60
CNO activation of Gq-DREADDs in basal forebrain astrocytes promotes wakefulness. Representative spectrograms, hypnograms, and EEG and EMG traces show responses to vehicle or CNO injections to the BF from an A, Ctrl mouse and a B, Gq+/− mouse. Open and closed bars below spectrograms represent the light and dark periods, respectively. C, Gray boxes show 30-min subsets of Gq+/− post-CNO injection data from B for low power, “fragmented” sleep (left) and “recovered” sleep (right). D, Red boxes show 40-s subsets of Gq+/− post-CNO injection data extracted from the gray boxes in C. Injection timepoints are highlighted by red rectangles in the hypnogram. Freq, frequency; Inj, injection; W, wakefulness; N, NREMS; R, REMS.
Gq+/− mice show prolonged, consolidated wakefulness, and reduced sleep after CNO. A, Time spent in wakefulness (left), NREMS (middle), and REMS (right) after vehicle or CNO delivery to BF during ZT0–ZT1 shown as a percentage of total recording time in 2-h bins for Ctrl (top) and Gq+/− (middle) littermates. Bottom row compares Ctrl and Gq+/− responses to CNO (repeated measures ANOVA). Sidak corrections for multiple comparisons were applied to post hoc pairwise comparisons (*p < 0.05). B, Change in bout frequency (top) and bout duration (bottom) shown as CNO–vehicle differences in 6-h bins for wakefulness (left), NREMS (middle), and REMS (right). *, different from 0 (one-sample t test). #, different from Ctrl (repeated measures ANOVA with post hoc Sidak-corrected pairwise comparisons). Open and closed bars on the x-axis represent the light and dark periods, respectively. Values are means ± SE from n = 12 Ctrl and n = 10 Gq+/− mice. Dots in B are data from individual mice. p < 0.05. See also Tables 1 and 2.
CNO increases latency to sleep in Gq+/− mice. Latency to NREMS and REMS after vehicle or CNO injections in BF defined as A, time to the first bout of average duration (based on 24-h vehicle data) or B, time to the first 4-s epoch postinjection. Note the log10 scale on the y-axis (vehicle vs CNO: paired t test or Wilcoxon signed-rank; Ctrl vs Gq+/−: unpaired t test or Mann–Whitney U). Values are means ± SE from n = 12 Ctrl and n = 10 Gq+/− mice. Dots are data from individual mice. *p < 0.05. See also Table 3.
We then verified that CNO had no effects on sleep architecture in Ctrl mice. This is a necessary control when using CNO as it has been shown to induce off-target effects that can indirectly impact sleep (Huber et al., 2000b; MacLaren et al., 2016; Manvich et al., 2018; Jendryka et al., 2019; Traut et al., 2023). The same dose delivered to Ctrl littermates had no impact on sleep architecture compared with vehicle (Figs. 2, 3; Tables 1, 2). In addition, between-subjects comparisons of Ctrl versus Gq+/− CNO responses recapitulated the within-subject vehicle versus CNO comparisons in Gq+/− mice (Figs. 3, 4). We also found CNO injected intracerebroventricularly to globally activate astroglial Gq-DREADDs did not reproduce the effects of intra-BF CNO injections (data not shown) indicating diffusion of CNO outside the BF is an unlikely explanation for these results.
CNO-induced wakefulness is not associated with increased sleep drive
CNO-induced wakefulness in Gq+/− mice did not result in homeostatic and compensatory increases in sleep time, continuity, or intensity at sleep onset as would be expected after sleep deprivation (SD) by gentle handling (Franken et al., 2001). Once Gq+/− mice fell asleep after the CNO injection, they exhibited reduced NREMS δ power in the δ band (0.5–4 Hz) as well as the low δ band (0.5–1.5 Hz), which is more sensitive to manipulations in astrocytes (Halassa et al., 2009; Ingiosi et al., 2020; Ingiosi and Frank, 2022; Fig. 5A; Table 4). In addition, CNO did not affect the high NREMS δ power band (2–4 Hz) in Gq+/− mice (Fig. 5A; Table 4); the latter is thought to be a marker of wake inertia when elevated within NREMS (Hubbard et al., 2020). CNO also had no effect on NREMS δ power in Ctrl mice (Fig. 5A). NREMS EEG spectral power in the low δ, α, and β frequency bands also decreased for at least 10 h post-CNO delivery in Gq+/− mice (Fig. 5B; Table 5). We then examined EEG changes in CNO-induced wakefulness to determine whether other indices of increased sleep drive were also absent [e.g., waking EEG θ activity (Vyazovskiy and Tobler, 2005; Zhang et al., 2014), leakage of slow waves into wakefulness (Huber et al., 2000a)]. We found that waking θ activity did not increase, nor did δ “leak” into waking or REMS spectra (Fig. 6; Table 5).
Statistical output from Kruskal–Wallis comparisons of vehicle versus CNO or J60 for NREMS δ power, low NREMS δ power, and high NREMS δ power
Statistical output from repeated measures ANOVA comparisons of vehicle versus CNO for Gq+/− EEG spectra
CNO does not increase sleep propensity in Gq+/− mice. A, Normalized NREMS δ power (0.5–4.0 Hz; top), low NREMS δ power (0.5–1.5 Hz; middle), and high NREMS δ power (2.0–4.0 Hz; bottom) shown in 2-h bins postinjection for Ctrl (left) and Gq+/− (right) mice (Kruskal–Wallis with post hoc Sidak-corrected pairwise comparisons). Open and closed bars on the x-axis represent the light and dark periods, respectively. Values are means ± SE from n = 12 Ctrl and n = 10 Gq+/− mice. *p < 0.05. See also Table 4. B, Normalized NREMS EEG spectral power for Gq+/− mice shown in 2-h bins starting from the time at which (h9–10) ≥5 mice spent ≥5 min in NREMS within the time bin post-CNO injection. Light and dark blue squares above the x-axis denote frequency bins with significant vehicle versus CNO differences (light blue: repeated measures ANOVA, 0.5–4 Hz; dark blue: repeated measures ANOVA, 0–30 Hz). Values are means (line) ± SE (shading). Sidak corrections for multiple comparisons were applied to post hoc pairwise comparisons. p < 0.05. See also Table 5.
CNO reduces EEG spectral power in Gq+/− mice. Normalized EEG spectral power for wakefulness (top) and REMS (bottom) from Gq+/− mice shown in response to vehicle and CNO. Wakefulness plots start immediately after injection. REMS plots start from the first time bin at which ≥5 mice showed REMS postinjection. Light and dark blue squares above the x-axis denote frequency bins with significant vehicle versus CNO differences (light blue: repeated measures ANOVA, 5–9 Hz; dark blue: repeated measures ANOVA, 0–30 Hz). Values are means (lines) ± SE (shading) based on data from n = 10 Gq+/− mice. Sidak corrections for multiple comparisons were applied to post hoc pairwise comparisons. p < 0.05. See also Table 5.
Next, as an additional control, we determined whether Gq+/− mice had the ability to produce a compensatory response to SD by other means (i.e., gentle handling). We found that sleep homeostasis as measured by NREMS EEG activity, sleep time, and sleep continuity was intact in Gq+/− mice. Six hours of SD using gentle handling produced the expected increase in NREMS δ power, NREMS time, and sleep continuity (defined as fewer, but longer, NREMS bouts) during recovery compared with undisturbed baseline (BL) conditions (Fig. 7; Tables 6, 7).
Statistical output from repeated measures ANOVA comparisons of baseline (BL) versus 6-h sleep deprivation (SD) via gentle handling for NREMS δ power, low NREMS δ power, and time in state for Gq+/− mice
Statistical output from one-sample t tests (compared with 0) for the change [from baseline (BL)] in the number of bouts per hour and bout duration after 6 h of sleep deprivation (SD) via gentle handling for Gq+/− mice
Sleep deprivation via gentle handling increases sleep propensity in Gq+/− mice. Normalized NREMS δ power for the A, full δ (0.5–4 Hz) and B, low δ (0.5–1.5 Hz) bands shown in 2-h bins for undisturbed baseline (BL) sleep and recovery sleep after 6-h sleep deprivation (SD) via gentle handling for Gq+/− mice. *, different from BL (repeated measures ANOVA with post hoc Sidak-corrected pairwise comparisons). C, Time spent in wakefulness (left), NREMS (middle), and REMS (right) under baseline conditions and after 6-h SD shown as a percentage of total recording time in 2-h bins for Gq+/− mice. *, different from BL (repeated measures ANOVA with post hoc Sidak-corrected pairwise comparisons). Cross-hatched bars on the x-axis for A–C denote the 6-h SD period. Open bars on the x-axis denote the light period recovery phase. Change in D, bout frequency and E, bout duration shown as SD–BL differences for the first 6 h of recovery sleep (i.e., light period) post-SD. *, different from 0 (one-sample t test). Latency to the F, first bout of average duration and G, first epoch for NREMS and REMS after 6-h SD. Values are means ± SE from n = 10 Gq+/− mice. Dots in D–G are data from individual mice. p < 0.05. See also Tables 6 and 7.
γ EEG activation in CNO-induced wakefulness
As discussed above, we examined several EEG bands during CNO-induced waking to determine whether sleep drive was accumulating during this time. To further assess whether CNO-induced waking was qualitatively different from normal waking, we measured changes in the γ band (30–60 Hz) post-CNO (h1–8) in Gq+/− mice. γ oscillations in wakefulness are associated with attention (Fries et al., 2001; Gregoriou et al., 2009; Rouhinen et al., 2013; Vinck et al., 2013), conscious perception (Melloni et al., 2007), and memory (Pesaran et al., 2002; Colgin et al., 2009; Carr et al., 2012). CNO did not affect the waking γ band in Gq+/− mice compared with vehicle [% total power: 0.031 ± 0.0033 (vehicle) versus 0.031 ± 0.0028 (CNO); paired t test: t(9) = 0.02; p = 0.981].
DREADD activation: effects on body temperature and motor activity in Gq+/− mice
We assessed CNO induced changes in core body temperature and cage activity in Gq+/− mice, as large changes in these parameters might indirectly impact sleep and EEG activity (Harding et al., 2020). CNO reduced core body temperature for ∼16 h postinjection and reduced cage activity during the dark period (Fig. 8A; Table 8). Consequently, we tested whether a different DREADD ligand, JHU37160 dihydrochloride (J60), reproduced the sleep architecture effects of CNO without changing body temperature in Gq+/− mice. J60 is a potent ligand with high DREADD selectivity and occupancy (Bonaventura et al., 2019; Peeters et al., 2020; Costa et al., 2021). J60 had similar effects on sleep architecture and EEG activity as CNO without affecting core body temperature or motor activity in Gq+/− mice (Fig. 8B; Table 8). Like CNO, J60 increased wakefulness for ∼8 h post-J60 injection in Gq+/− mice compared with vehicle (Figs. 9, 10A; Table 1) with corresponding reductions in NREM and REM sleep time, bout frequency, and bout duration (Fig. 10A,B; Tables 1, 2). J60 also increased latencies to NREMS and REMS in Gq+/− mice (Fig. 11; Table 3). Gq+/− sleep expression in the dark period after J60 did not differ from vehicle (Fig. 10; Tables 1, 2).
Statistical output from repeated measures ANOVA comparisons of vehicle versus CNO or J60 for core body temperature and cage activity for Gq+/− mice
Statistical output from repeated measures ANOVA comparisons of vehicle versus J60 for Gq+/− EEG spectra
CNO, but not J60, reduces core body temperature and cage activity in Gq+/− mice. A, Core body temperature (top) and cage activity (bottom) for Gq+/− mice shown in 2-h bins in response to vehicle and CNO injected during ZT0–ZT1. B, Core body temperature (top) and cage activity (bottom) for Gq+/− mice shown in 2-h bins in response to vehicle and J60 injected during ZT0–ZT1. Data are shown as means ± SE (repeated measures ANOVA with post hoc Sidak-corrected pairwise comparisons). Open and closed bars on the x-axis represent the light and dark periods, respectively. n = 5 Gq+/− mice. *p < 0.05. See also Table 8.
J60 activation of Gq-DREADDs in basal forebrain astrocytes promotes wakefulness. Representative spectrograms, hypnograms, and EEG and EMG traces show responses to vehicle or J60 injections to the BF from the same A, Ctrl and B, Gq+/− mice shown in Figure 2. Open and closed bars below spectrograms represent the light and dark periods, respectively. C, Gray boxes show 30-min subsets of Gq+/− post-J60 injection data from B for lower power, “fragmented” sleep (left) and “recovered” sleep (right). D, Red boxes show 40-s subsets of Gq+/− post-J60 injection data extracted from the gray boxes in C. Injection timepoints are highlighted by red rectangles in the hypnogram. Freq, frequency; Inj, injection; W, wakefulness; N, NREMS; R, REMS.
Gq+/− mice show prolonged, consolidated wakefulness and reduced sleep after J60. A, Time spent in wakefulness (left), NREMS (middle), and REMS (right) after vehicle or J60 delivery to BF during ZT0–ZT1 shown as a percentage of total recording time in 2-h bins for Ctrl (top) and Gq+/− (middle) littermates. Bottom row compares Ctrl and Gq+/− responses to J60 (repeated measures ANOVA). Sidak corrections for multiple comparisons were applied to post hoc pairwise comparisons (*p < 0.05). B, Change in bout frequency (top) and bout duration (bottom) shown as J60–vehicle differences in 6-h bins for wakefulness (left), NREMS (middle), and REMS (right). *, different from 0 (one-sample t test). #, different from Ctrl (repeated measures ANOVA with post hoc Sidak-corrected pairwise comparisons). Open and closed bars on the x-axis represent the light and dark periods, respectively. Values are means ± SE from n = 12 Ctrl and n = 10 Gq+/− mice. Dots in B are data from individual mice. p < 0.05. See also Tables 1 and 2.
J60 increases latency to sleep in Gq+/− mice. Latency to NREMS and REMS after vehicle or J60 injections in BF defined as A, time to the first bout of average duration (based on 24-h vehicle data) or B, time to the first 4-s epoch postinjection. Note the log10 scale on the y-axis (vehicle vs J60: paired t test or Wilcoxon signed-rank; Ctrl vs Gq+/−: unpaired t test or Mann–Whitney U). Values are means ± SE from n = 12 Ctrl and n = 10 Gq+/− mice. Dots are data from individual mice. *p < 0.05. See also Table 3.
Like CNO, J60-induced wakefulness did not increase sleep drive in Gq+/− mice as measured by changes in sleep time, sleep continuity, and EEG activity. As was true for CNO-induced wakefulness, J60-induced wakefulness did not lead to increases in NREMS δ power, low NREMS δ power, or high NREMS δ power (Fig. 12A,B; Tables 4, 9) in Gq+/− mice compared with vehicle. Similarly, J60-induced waking EEG did not exhibit elevated θ after J60, nor was there elevated δ in wake or REMS spectra (Fig. 13; Table 9). The waking γ band was also unaffected during the initial wakefulness period (h1–4) after J60 in Gq+/− mice [% total power: 0.032 ± 0.0034 (vehicle) vs 0.026 ± 0.0023 (J60); paired t test: t(9) = 1.40; p = 0.195].
J60 does not increase sleep propensity in Gq+/− mice. A, Normalized NREMS δ power (0.5–4.0 Hz; top), low NREMS δ power (0.5–1.5 Hz; middle), and high NREMS δ power (2.0–4.0 Hz; bottom) shown in 2-h bins postinjection for Ctrl (left) and Gq+/− (right) mice (Kruskal–Wallis with post hoc Sidak-corrected pairwise comparisons). Open and closed bars on the x-axis represent the light and dark periods, respectively. Values are means ± SE from n = 12 Ctrl and n = 10 Gq+/− mice. *p < 0.05. See also Table 4. B, Normalized NREMS EEG spectral power from Gq+/− mice shown in 2-h bins starting from the time at which (h3–4) ≥5 mice spent ≥5 min in NREMS within the time bin post-J60 injection. Light and dark blue squares above the x-axis denote frequency bins with significant vehicle versus J60 differences (light blue: repeated measures ANOVA, 0.5–4 Hz; dark blue: repeated measures ANOVA, 0–30 Hz). Values are means (lines) ± SE (shading). Sidak corrections for multiple comparisons were applied to post hoc pairwise comparisons. p < 0.05. See also Table 9.
J60 reduces EEG spectral power in Gq+/− mice. Normalized EEG spectral power for wakefulness (top) and REMS (bottom) from Gq+/− mice shown in response to vehicle and J60. Statistics are not shown for REMS h17–18 as only four mice had REMS after J60 during this time bin. Wakefulness plots start immediately after injection. REMS plots start from the first time bin at which ≥5 mice showed REMS postinjection. Dark blue squares above the x-axis denote frequency bins with significant vehicle versus J60 differences (repeated measures ANOVA, 0–30 Hz). Values are means (lines) ± SE (shading) based on data from n = 10 Gq+/− mice. Sidak corrections for multiple comparisons were applied to post hoc pairwise comparisons. p < 0.05. See also Table 9.
We also found that sleep expression of Ctrl mice was mostly unaffected by J60 compared with vehicle (Figs. 9–12; Table 1–Table 4). Between-subjects comparisons of Ctrl versus Gq+/− J60 responses recapitulated Gq+/− vehicle versus J60 within-subject comparisons as well (Figs. 10, 11).
Sex differences
There were minor sex differences following different manipulations in either Ctrl or Gq+/− mice, but these did not change the overall effects induced by DREADDs (data not shown).
Discussion
We investigated the role of BF astrocytes in sleep expression and regulation. We find DREADD activation of the Gq-pathway in BF astrocytes produces long periods of continuous waking that paradoxically do not trigger compensatory changes in sleep. Previous studies showed as little as 70 min of sustained wakefulness increases NREMS EEG δ power (a canonical index of sleep drive) in adult mice across diverse mouse strains (Franken et al., 2001). In the present study, activating BF astroglial Gq-pathways produced ≥6 h of sustained wakefulness with no compensatory changes in sleep architecture or NREMS EEG activity. This finding is unlikely to be explained by indirect effects of the DREADD ligand CNO as these results were reproduced using a second DREADD ligand (J60) and not found in non-DREADD-expressing Ctrl mice treated with DREADD ligands. We discuss our main results below in more detail.
BF astrocytes induce waking without increasing sleep drive
Gq-DREADD activation of BF astrocytes produced hours of waking without the expected signs of increased sleep drive. In rodents, comparable amounts of sleep loss via gentle-handling, forced locomotion, or novel object introduction reliably induces compensatory increases in NREMS δ power, sleep continuity (fewer but longer NREMS bouts), and to a lesser extent, sleep time (Franken et al., 2001; Bellesi et al., 2013; Dispersyn et al., 2013). Increased sleep drive also reportedly increases waking EEG θ and δ during SD (Huber et al., 2000a; Vyazovskiy and Tobler, 2005; Zhang et al., 2014). In contrast, despite producing waking amounts (≥6 h) that lead to saturating sleep homeostatic responses in mice (Franken et al., 2001), DREADD-induced sleep loss produced no compensatory changes in any of these metrics.
There are few examples of such complete dissociations between time awake and sleep drive in mammals (Greene and Frank, 2010). For example, there are a handful of gene mutations that influence sleep homeostasis, but in most cases sleep homeostasis is blunted, not eliminated as we show here (Shaw and Franken, 2003; Bjorness et al., 2009; Greene and Frank, 2010; Mang and Franken, 2015; Ingiosi et al., 2020; Wimmer et al., 2021; Ingiosi and Frank, 2022). Even in infancy, when SD does not increase NREMS δ power, a need for sleep is present as sleep drive rapidly increases with SD, and during recovery, neonates show compensatory changes in sleep time or continuity (Frank et al., 1998; Dumoulin Bridi et al., 2015). Our findings are also unlikely explained by the quality of wakefulness following BF astroglial activation. The effects of different types of waking experiences versus time awake are relatively modest and not always observed (Greene and Frank, 2010; Guillaumin et al., 2018; Milinski et al., 2021). Lastly, studies reporting similar prolonged waking after lateral hypothalamic astrocyte activation (Cai et al., 2022) or neuronal activation (in mammillary bodies; Pedersen et al., 2017) did not examine in detail compensatory changes in subsequent sleep and EEG activity. Therefore, the latter studies are inconclusive with respect to this question.
What then may explain the unusual finding of wakefulness without sleep drive? The absence of a normal homeostatic response to sleep loss is not explained by indirect effects of CNO or an underlying defect in Gq+/− mice. For example, CNO in Ctrl mice did not alter sleep-wake architecture (Figs. 3, 5) indicating CNO did not induce off-target effects on our sleep measures. While CNO decreased core temperature in Gq+/− mice, this fell within physiological ranges that occur across the sleep-wake cycle (Kozak et al., 1995; Oka et al., 2003; Jhaveri et al., 2007). Moreover, this change in core temperature is an unlikely explanation for our results. This is because a second DREADD ligand (J60), that did not change core temperature or activity (Fig. 8), reproduced the main effects of CNO on sleep time, architecture, and homeostasis (Figs. 10, 12).
There is also no evidence that simply expressing DREADDs in BF astrocytes impaired sleep homeostasis. More specifically, sleep homeostasis was intact in Gq+/− mice as they responded with expected homeostatic responses to 6 h of SD by gentle-handling (in the absence of CNO/J60). Like other mouse strains (Franken et al., 1999, 2001), 6-h SD by gentle-handling induced compensatory changes in NREMS δ power, sleep time, and sleep continuity during recovery (Fig. 7). These findings demonstrate Gq+/− mice have the capacity to respond to extended waking (i.e., the sleep homeostat is intact) but fail to do so when waking is instead produced by BF astrocyte activation.
An alternative explanation is activating BF astrocytes changes the activity of surrounding BF neurons that play different roles in generating waking and, separately, the homeostatic sleep response to waking. The BF is comprised of a heterogenous collection of neurons that may play different roles in sleep/wake, EEG activity, and homeostatic responses to sleep loss (Xu et al., 2015; Yang et al., 2017). For example, BF cholinergic neurons (ChAT+) are specifically linked to mammalian sleep homeostasis. Ablating ChAT+ neurons reduces NREMS compensatory responses to SD and waking EEG changes indicative of sleep drive (Kalinchuk et al., 2008, 2015; but see Peng et al., 2020). Selective DREADD activation of BF GABAergic neurons produces long periods of wakefulness similar to what we report, but once sleep commences, the homeostatic response is observed (NREMS δ power increases; Anaclet et al., 2015). Although speculative, our findings suggest Gq-mediated activation of BF astrocytes leads to a complex activation of these circuits, such that waking is triggered (possibly via GABAergic activity) while cholinergic activity is inhibited (explaining the lack of a homeostatic sleep response). This explanation is also in keeping with the fact BF astrocytes do not have long-range connections to forebrain or hindbrain canonical sleep/wake centers, meaning their effects are likely mediated by surrounding neurons that have such projections. However, it is also possible astroglial DREADD activation propagates via a syncytium which could conceivably contribute to distal effects as has been suggested for intracortical responses (Vaidyanathan et al., 2021).
Mechanisms downstream of astroglial Gq activation
Astroglial Gq activation triggers several downstream events that influence activity of surrounding neurons in ways that might explain our results. These include gliotransmission, neurotransmitter uptake, and metabolic neuronal support. Of these, gliotransmission enjoys the most empirical support based on studies in vitro, in situ, and in vivo. For example, studies in vitro, in situ, and in vivo show astroglial Gq activation stimulates gliotransmission of ATP (Gordon et al., 2009; Martin-Fernandez et al., 2017; Iwai et al., 2021; Kofuji and Araque, 2021). ATP has diverse effects on neurons depending on their complement of adenosine or purinergic receptors. The ATP metabolite adenosine via A1 receptors can inhibit neurons or excite neurons via disinhibition (via GABAergic interneurons; Arrigoni et al., 2006; Hawryluk et al., 2012; Yang et al., 2013). Directly activating BF purinergic receptors can also profoundly alter sleep-wake architecture (Yang et al., 2018). Therefore, local release of astroglial ATP might be expected to have complex effects on BF neurons. In contrast, while astroglial Gq activation can influence neurotransmitter and ion uptake (Wang et al., 2012; Devaraju et al., 2013) and changes in astroglial-neuronal metabolism (Loaiza et al., 2003), there is less support for these mechanisms in vivo.
Future directions
Our results raise several questions beyond the scope of this single study to answer. It will be important to determine changes in surrounding brain cells following BF astroglial Gq activation (and inhibition). Does this manipulation lead to a specific, complex activation and inhibition of BF neurons as we propose? If so, what mediates this diversity of responses? Could this be related to the impact of different neuromodulators (and their intracellular signals) on astrocytes? For example, global knockout of astroglial β2-adrenergic (Gs-coupled) receptors reduces homeostatic response to sleep loss but does not eliminate it (Ingiosi and Frank, 2022). Activating different astroglial G-protein-coupled pathways in cortex also has different effects on cortical sleep measures (Vaidyanathan et al., 2021). Therefore, different astroglial intracellular signaling pathways may have distinct roles in sleep expression and regulation.
If gliotransmission is the most plausible mechanism to explain how astroglial changes result in neuronal changes (which is to be determined), what are the roles of other putative gliotransmitters shown to influence sleep-wake architecture and regulation? For example, astroglial Gq-DREADD activation induces release of the excitatory neurotransmitter glutamate (Scofield et al., 2015; Durkee et al., 2019), and chemogenetic activation of BF neuronal subtypes produces different components (e.g., cholinergic-induced suppression of EEG spectra; Anaclet et al., 2015) of the astroglial-induced phenotype described here. Alternatively, BF astrocytes may be a heterogeneous population whose activation induces different downstream impacts on BF neurons.
Lastly, what are the implications of producing wakefulness without cost? There are at least two worth discussing. The first is this narrows the search for the need for sleep and by extension, sleep function (Benington, 2000). The second is it holds the promise of creating waking brains that need less sleep.
[While in review, Peng et al. (2023) reported BF astroglial Gq-DREADD activation using a different promoter (GFAP vs our promoter, Aldh1l1) reproduced some of our results (e.g., reduced REMS time and NREMS δ power).]
Footnotes
This work was funded by National Institutes of Health Grants K99 NS119293 (to A.M.I.), R01 MH099544 (to M.G.F.), and R01 NS114780 (to M.G.F.). We thank Dr. Christine Muheim, Dr. Kristan Singletary, and Sabrina Koh for their technical assistance.
The authors declare no competing financial interests.
- Correspondence should be addressed to Marcos G. Frank at marcos.frank{at}wsu.edu