All procedures were in strict accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and were approved by Tufts University and Institutional Animal Care and Use Committees. All animals were housed on a 12 h light/dark cycle and were given standard chow and water ad libitum.

The VIPP mouse line was created by fusing the type I Ins(1,4,5)P₃ 5′-phosphatase (IPP) construct to the Venus construct and co-expressing the Lac-Z reporter (De Smedt et al., 1997). The original IPP construct in pcDNA3 (De Smedt et al., 1997) was released by BamHI/XbaI digestion and cloned into pUC19 cloning vector. The coding region of Venus (Evanko and Haydon, 2005) was PCR amplified, digested with BamHI/NcoI, and ligated in-frame into the corresponding sites of IPP in pUC19. For experiments involving cultured astrocytes, the VIPP fragment was cloned into the expression vector pcDNA3.1. To make the tetO-transgenic vector, the VIPP fragment was excised from pcDNA3.1 using BamHI (blunted)/XbaI, and the ligated into StyI (blunted)/SpeI digested pTg1-tetO vector. The pTg1-tetO-plasmid has been described elsewhere (Pascual et al., 2005). TetO-VIPP and tetO-LacZ were digested to remove bacterial sequences and co-injected into C57BL/6J fertilized zygotes. Founders were identified by PCR, and subsequently crossed with GFAP-tTA animals (Pascual et al., 2005). All breedings were performed with 40 mg/kg doxycycline (Dox, Bioserv, Frenchtown, NJ, USA) in the food, and offspring were maintained on Dox until weaning to suppress transgene expression during embryonic life and post-natal development. VIPP or littermate control C57BL/6J mice were used.

Horizontal hippocampal slices (300–330 μm) from 6 to 10-week-old mice were prepared as described previously (Pascual et al., 2005). Briefly, the brain was rapidly removed and chilled with cold (4°C) artificial cerebrospinal fluid (ACSF) of the following composition (in mM): 124 NaCl, 3.1 KCl, 2 MgCl₂, 1 CaCl₂, 26 NaHCO₃, 1 NaH₂PO₄, 10 glucose, 1 Na-pyruvate, and 0.6 ascorbic acid (pH 7.4 adjusted with 95% O₂, 5% CO₂) or 85 NaCl, 2.5 KCl, 1.25 NaH₂PO₄, 0.5 CaCl₂, 4 MgCl₂, 25 NaHCO₃, 75 sucrose, 25 glucose, and 0.5 ascorbic acid (pH 7.4 with 95% O₂, 5% CO₂). The brain was then cut with a Leica VT 1000S vibratome and CA3 to CA1 connections were severed. Slices were incubated at 30–32°C for 1 h recovery and were recorded at 32–34°C.

A multicell bolus loading was performed as previously described (Li et al., 2003). Rhod-2/AM, was dissolved in 20% pluronic acid in DMSO to yield a concentration of 20 mM. For cell loading, this solution was diluted 1/100 with an injection solution containing (in mM) 150 NaCl, 2.5 KCl, 10 HEPES, pH 7.4. An injection pipette (1–2 MΩ) was filled with this solution and inserted into stratum radiatum of CA1 region, and ejected using positive pressure (1 min, 70 kPa; Picospritzer NPI, Germany). One hour later images were acquired using a Q-imaging cooled CCD camera and ImagePro software.

Wild type and VIPP brain slices were isolated and responses to stimulation were quantified as delta fluorescence, expressed as ΔF/F = (F₁-F₀)/F₀, where F₀ and F₁ are the value of the fluorescence in the ROI at rest and the given time point, respectively. Slices were stimulated with either 100 μM ATP application or electrical stimulation of the Schaffer collateral fibers. Stimulation pulses were elicited using a 125 μm concentric Pt-Ir electrode. Either 1 s of 100 Hz or five theta burst trains were used to stimulate slices. Theta burst stimulation was 100 Hz for 40 ms with 200 ms intervals. Recordings of astrocyte responses to stimulation were made in area CA1.

Cortical astrocyte cultures were prepared from 1 to 2-day-old mice. Cells were grown in modified Eagle’s medium containing 10% fetal bovine serum, 2 mM L-glutamine, 40 mM D-glucose, 14 mM sodium bicarbonate, 1% sodium pyruvate, and 1% penicillin/streptomycin in 95% air/5% humidified CO₂. Confluent cultures were shaken overnight followed by a media change and a subsequent round of shaking. After the final shaking, adherent cells were dissociated using 0.25% trypsin and 1 mM EDTA, and replated on glass coverslips. Cells were used after 2–4 days in culture.

Astrocytes were transfected using Fugene (Roche, Indianapolis, IN, USA) and 24–72 h post-transfection, astrocytes cotransfected with IPP and DsRed were loaded with Fluo-4 by incubation in Fluo-4/AM, and Ca²⁺ imaging was performed on the stage of an inverted microscope. Astrocytes transfected with VIPP were loaded with Rhod-2 by incubation in Rhod-2/AM and Ca²⁺ imaging was performed in a similar fashion. Ins(1,4,5)P₃ 5′-phosphatase activity at 1 μM substrate level was performed on VIPP transfected COS-7 cells as described previously (Terunuma et al., 2008).

Mice aged 8–12 weeks were anesthetized with isoflurane and placed into a stereotaxic frame. Electroencephalogram (EEG) head implants (Pinnacle Technology, Inc.) were placed as previously described (Clasadonte et al., 2014). After surgery, mice were subcutaneously injected with buprenorphine (0.08 mg/kg) and lactated Ringer’s solution, and fed moistened rodent food. After 5 days of postoperative recovery, lightweight recording cables were connected to the head implants and mice were placed in Circular Plexiglas cages (Pinnacle Technology, Inc.) and acclimated for a week.

Electroencephalography/electromyography activity was continuously monitored for 48 h. EEG/EMG activity starting at Zeitgeber time (ZT) 0 was measured for 24 h (baseline), followed by 6 h of sleep deprivation enforced by gentle handling. EEG/EMG activity was monitored during the 6 h of sleep deprivation and 18 h of recovery sleep. During data acquisition, EEG signals were high pass filtered at 0.5 Hz and low pass filtered at 40 Hz. EMG signals were high pass filtered at 0.5 Hz and low pass filtered at 100 Hz. The amplifier system (Pinnacle Technology, Inc.), sampled at 250 Hz with a PAL 8400 data acquisition system (Pinnacle Technology, Inc.).

Sleep stages were scored visually based on 4 s epochs by a trained experimenter using SleepSign for Animal software (Kissei Comtec). Wakefulness consisted of low-amplitude, high-frequency EEG, and high EMG activity; REM sleep consisted of low-amplitude EEG activity with prominent 5–8 Hz frequency and low EMG activity; NREM sleep consisted of high-amplitude, low-frequency EEG with little EMG modulation. Time spent in each state was expressed as a percentage of the total recording time in 1 h time bins. EEG power spectra of consecutive 4 s epochs [fast Fourier transform (FFT) routine; Hanning window] were calculated. The EEG power during NREM sleep from 0.5 to 1.5 Hz was defined as low frequency slow wave activity (lf-SWA) and was used as a quantitative measure of sleep pressure and homeostatic sleep drive (Franken et al., 2001; Cirelli et al., 2005). The EEG power of lf-SWA during NREM sleep was used to assess hour-by-hour sleep power. Hour by hour lf-SWA was normalized to the last 4 h of the baseline day. The EEG power from 5.0 to 8.0 Hz was defined as theta activity and was used as a measure of theta power. To compare theta power across genotypes, raw data were normalized to 0.5 to 1.5 Hz frequency range. Normalization in this range was chosen because it best represented the raw data and there were no differences in power between wild type and VIPP mice in this frequency range.

Eight to ten week old animals were cardiac perfused with phosphate buffered saline (PBS) and paraformaldehyde (PFA). The brain was extracted, post-fixed in PFA for 1 h at 4°C, placed overnight in 30% sucrose and frozen at -80°C. Coronal sections (40 μm) were cut, washed with PBS, and permeabilized with 0.1% Triton X-100, 5% horse or goat serum in PBS. Unless otherwise indicated the antibodies used were from Millipore, Temecula, CA, USA. We used mouse NeuN antibody, 1:1000, rabbit NG2 antibody, 1:1000, rabbit GFAP antibody (Sigma, St. Louis, MO, USA), 1:1000, and rabbit Iba1 antibody (Wako Chemicals, Richmond, VA, USA), 1:500. Secondary antibodies conjugated to either Alexa 633 or 568 were used. Sections were mounted and visualized on either Fluoview 1000 confocal microscope (Olympus, Center Valley, PA, USA) or A1 confocal microscope (Nikon, Tokyo, Japan). Venus fluorescence was detected at 515 nm.

For all experiments except sleep studies, comparisons between two groups were conducted with Student’s t-test. For sleep studies, comparisons between two groups were made using a Mann–Whitney U test. Groups with multiple data points were compared using two-way repeated measures ANOVA, followed by a Tukey’s post hoc multiple-comparisons test. Statistical significance was defined as p < 0.05. Data are presented as mean ± SEM unless otherwise stated.

To determine the importance of astrocytic IP₃/Ca²⁺ signals, we used molecular genetics to selectively impair IP₃/Ca²⁺ signaling in astrocytes. IP₃ is metabolized through type I Ins(1,4,5)P₃ 5′-phosphatase (IPP) activity (De Smedt et al., 1997). We transiently transfected cultured astrocytes to express IPP and demonstrate that it attenuates agonist-induced Ca²⁺ signaling in these cells (Supplementary Figure S1A; 83 ± 2% reduction of the Ca²⁺ signal, n = 10 coverslips). To generate a fluorescent variant of IPP for use in transgenic animals, we tagged the N-terminal of IPP with the YFP variant Venus. Transient transfection and over expression of Venus-IPP fusion protein (VIPP) attenuated Ca²⁺ signaling in cultured astrocytes (Supplementary Figure S1B; 67 ± 3% reduction of the Ca²⁺ signal, n = 6 coverslips). When the transgenic vector was transfected into COS-7 cells, anti-GFP blotting of protein lysates revealed a band corresponding to the predicted size of the Venus-IPP fusion protein (Supplementary Figure S1C), and biochemical analysis showed a 40-fold increase in IPP activity (Supplementary Figure S1D).

We generated astrocyte-specific transgenic animals using the Tet-Off system (Figure 1A) (Pascual et al., 2005; Halassa et al., 2009) to allow conditional expression of VIPP in astrocytes (VIPP mice). TetO.VIPP and tetO.β-gal were co-injected into C57BL/6J fertilized zygotes to generate tetO.VIPP founder lines, which were screened for transgene expression by being crossed with GFAP.tTA mice (Figure 1A) (Pascual et al., 2005; Halassa et al., 2009). Bigenic offspring of one line showed abundant Venus fluorescence when examined by confocal microscopy (Figures 1B–H). Double-labeling experiments demonstrated that VIPP was selectively expressed in astrocytes. Venus fluorescence was observed in 39% of hippocampal area CA1 GFAP positive astrocytes (n = 527 out of 1343 from six animals; Figures 1C,D,I), but was never detected in neurons (NeuN; Figures 1B,H), NG2-glia (NG2; Figure 1E), oligodendrocytes (olig1; Figure 1F), or microglia (Iba1; Figure 1G). In addition to hippocampal VIPP expression, we also observed transgene expression in the brainstem (Supplementary Figure S3A), and minimal expression in the cortex. Double labeling with anti-NeuN never showed transgene expression in neurons in any brain region.

We studied stimulus-induced astrocytic Ca²⁺ signaling in the hippocampus, a brain region with prominent theta activity. We found astrocytic Ca²⁺ was attenuated in slices derived from VIPP mice. The time-integral of the Ca²⁺ response was attenuated from a wild-type value of 2083 ± 360 ΔF/F₀^∗s to 156 ± 61 ΔF/F₀^∗s in astrocytes visually confirmed to express VIPP (p < 0.0016, t-test) in response to exogenous application of ATP (100 μM). Astrocytes with undetectable VIPP exhibited Ca²⁺ signals that were not significantly different from wild-type values (Figures 1J,K). Similarly, synaptic activity-dependent recruitment of astrocytic Ca²⁺ signals was attenuated in VIPP expressing astrocytes (Figures 1L–O). Ca²⁺ signals in VIPP expressing astrocytes were attenuated from 3327 ± 940 ΔF/F₀^∗s to 253 ± 70 ΔF/F₀^∗s (p < 0.025, t-test; Figures 1L,M) following five trains of theta burst stimulation. The ability of VIPP to attenuate astrocytic Ca²⁺ signals was not stimulus-specific since the response to 100 Hz tetanic trains was also significantly reduced in VIPP expressing astrocytes (p < 0.006, t-test; Figures 1N,O).

To examine the importance of astrocytic IP₃/Ca²⁺ signaling in sleep regulation, we asked whether VIPP expression altered the duration of each vigilance state. The percent time spent in REM sleep was increased in VIPP expressing mice, compared to wild type mice (Figure 2; VIPP, n = 12; WT, n = 8). This was most apparent during the light phase (subjective nighttime), and corresponded to an increase in number of REM bouts (represented in Figures 3A,B and quantified in Figure 3C, VIPP = 63.75 ± 1.6, wild type = 56.5 ± 2.0; p < 0.01). However, REM bout duration was unchanged in VIPP mice (Figure 3D). Mice spent significantly more time in REM sleep during the second half of the light phase and the second half of the dark phase (Figure 2C, ZT6-12; VIPP = 8.6 ± 0.4%, wild type = 6.8 ± 0.3%; p < 0.01; ZT18-24; VIPP = 4.9 ± 0.4%, wild type = 3.9 ± 0.4%; p < 0.05). During the second half of the light phase, when VIPP mice spent a far greater percentage of time in REM sleep, compared to wild type mice, they spent significantly less time awake (Figure 2A, ZT6-12; VIPP = 34.5 ± 1.3%, wild type = 39.8 ± 2.0%; p < 0.05). Throughout the light/dark cycle, VIPP and wild type mice spent similar time in NREM (Figure 2B). These data suggest that astrocyte IP₃/Ca²⁺ signaling specifically modulates the REM phase of sleep, particularly in the later portion of the inactive period.

Astrocytes modulate plasticity and cortical oscillations as well as sleep homeostasis (Fellin et al., 2009; Halassa et al., 2009; Lee et al., 2014). Using genetically modified mice that have attenuated gliotransmission, the dnSNARE mouse, it was shown that the gliotransmitter adenosine (derived from ATP) is necessary for the accumulation of sleep pressure (Halassa et al., 2009). Given that Ca²⁺ signaling can contribute to gliotransmission (Porter and McCarthy, 1997; Araque et al., 1998; Buzsàki, 2002; Hua et al., 2004; Kreft et al., 2004; Chen et al., 2005; Pryazhnikov and Khiroug, 2008; Paukert et al., 2014), we asked whether we could reproduce the dnSNARE phenotype of impaired sleep homeostasis by disrupting astrocytic IP₃/Ca²⁺ signaling. We compared baseline SWA and SWA following sleep deprivation in VIPP and wild type littermate mice. VIPP mice neither exhibited alterations in SWA (0.5–4 Hz) nor low frequency-SWA (lf-SWA, 0.5–1.5 Hz) either in baseline conditions or following sleep deprivation compared to wild type littermates (Supplementary Figure S2) suggesting that IP₃/Ca²⁺ signaling does not modulate NREM sleep homeostasis.

The magnitude of the homeostatic increase in SWA delta power that occurs with prolonged wake is positively correlated with the duration of subsequent NREM sleep (McCarley, 2007). REM sleep is also homeostatically regulated; REM-specific sleep deprivation results in REM specific rebound sleep (Kitka et al., 2009). However, the regulation of REM sleep homeostasis is poorly understood. VIPP mice show enhanced REM rebound sleep after 6 h of sleep deprivation, compared to wild type mice (Figure 4). The greatest difference is seen in the later phase of the light cycle, during ZT10-11 (Figure 4C; p < 0.05, Tukey’s post hoc test). These data indicate that astrocyte IP₃/Ca²⁺ signaling modulates the homeostatic increase in REM sleep after SD.

We next examined power spectra associated with each vigilance state. We found enhanced EEG theta power in VIPP mice, compared to wild type mice (Figure 5). This effect was present during wakefulness, NREM and REM sleep. However, the most robust effect occurred during REM sleep, where theta is the dominant frequency in the power spectrum; although, theta is also a prominent frequency in wakefulness during active stimulus processing (Colgin, 2013). When a power analysis was performed on the entire 24 h baseline day, VIPP mice exhibited enhanced theta power (Figure 5A; WAKE: p < 0.05 at 7.1–8.1 Hz; NREM: p < 0.05 at 3.6–4.1, 5.8–6.3 Hz; REM: p < 0.05 at 5.9–8.1 Hz). When overall theta power (5–8 Hz) was averaged and compared across genotypes, VIPP mice expressed significantly increased power during REM sleep (p < 0.05; wild type: 3.6 ± 0.5; VIPP: 5.1 ± 0.4) but not during other vigilance states.

To assess a time-of-day effect on theta power, the baseline day was broken down into the light and dark phase and power spectra were analyzed for each vigilance state within these time bins. In both the light and dark phase, average theta power was significantly increased during REM sleep (Figures 5B,C). During the light phase, VIPP mice exhibited enhanced power within the theta frequency range across both NREM and REM sleep (Figure 5B; WAKE: NS; NREM: p < 0.05 at 3.9–4.1, 5.3, 5.8–6.3 Hz; REM: p < 0.05 at 5.4–8.1 Hz). When theta power was averaged across the entire 5–8 Hz frequency range, an increase in average theta power was observed in VIPP mice during REM sleep only (inset: wild type: 3.5 ± 0.5; VIPP: 5.0 ± 0.4; p < 0.05). During the dark phase, VIPP mice exhibited enhanced theta power across all vigilance states (Figure 5C; WAKE: p < 0.05 at 7.1–8.1 Hz; NREM: p < 0.05 at 3.9 Hz; REM: p < 0.05 at 6.1–8.1 Hz), but when averaged across frequencies, VIPP mice had significantly greater theta power during REM sleep only (wild type: 3.7 ± 0.6; VIPP: 5.8 ± 0.6; p < 0.05). These data suggest that astrocyte IP₃/Ca²⁺ signaling modulates theta power, especially during REM sleep.