Cannabinoid Signaling Recruits Astrocytes to Modulate Presynaptic Function in the Suprachiasmatic Nucleus

Visual Abstract


Introduction
Circadian rhythms are 24-h cycles in behavioral and physiologic processes such as sleep/wake cycles, cognitive function, and hormone release. In mammals, the master circadian oscillator is located in the suprachiasmatic nucleus (SCN) where neurons expressing a molecular clock generate a 24-h timing signal. The network of SCN neurons and astrocytes, which also express molecular clocks, are required for a precise and stable circadian signal (Welsh et al., 2010;Jackson, 2011;Mohawk and Takahashi, 2011). A single astrocyte can interact with hundreds of synapses, enabling these cells to regulate network activity within brain regions (Pirttimaki and Parri, 2013). Astrocytes regulate clock gene expression and neuronal synchrony in culture systems, but how astrocytes regulate neuronal function within the SCN remains largely unknown (Beaulé et al., 2009;Marpegan et al., 2009Marpegan et al., , 2011Barca-Mayo et al., 2017;Brancaccio et al., 2017;Tso et al., 2017;Svobodova et al., 2018). Astrocytes regulate neuronal function in multiple ways, from buffering extracellular ion and neurotransmitter concentrations (Bellot-Saez et al., 2017), to regulating blood oxygen and metabolism (Sonnay et al., 2018;Takata et al., 2018), and actively responding to external cues through intracellular signaling molecules such as Ca 2ϩ and inducing the release of neuromodulators such as ATP, glutamate, or adenosine (Fiacco et al., 2009;Losi et al., 2014).
Endocannabinoids, endogenously generated lipophilic molecules, act as retrograde signals from neurons to regulate presynaptic neurotransmitter release via activation of G-protein-coupled cannabinoid-1 receptors (CB1Rs; Ohno-Shosaku et al., 2001;Araque et al., 2017). The production and metabolism of endocannabinoids have diurnal patterns, indicating they may be under circadian clock control (Valenti et al., 2004;Liedhegner et al., 2014;Koch et al., 2015). Cannabinoid receptor activation blocks lightinduced phase shifts of circadian behavior (Sanford et al., 2008;Acuna-Goycolea et al., 2010). In addition, cannabinoid signaling increases neuronal firing within the SCN by decreasing presynaptic GABA release (Acuna-Goycolea et al., 2010). Given this evidence of interactions between the circadian and cannabinoid systems, surprisingly little is known about how cannabinoids alter SCN function and circadian clock timing. Endocannabinoids can alter neuronal function by activating astrocyte signaling pathways and the release of gliotransmitters (Navarrete and Araque, 2008). Here, we hypothesize that cannabinoid signaling activates an intracellular Ca 2ϩ signaling pathway in astrocytes and releases neuromodulators to alter SCN neuronal function and ultimately change circadian clock timing.

Ethical approval
All animal care, handling, and housing were approved in advance by the Institutional Animal Care and Use Committee at Oregon Health & Science University.

Animals and housing
Both male and female mice (three to six months of age) on a C57/BL6 background were used. Mice were group housed on a 12/12 h light/dark (LD) cycle with food and water ad libitum. GFAP-Cre mice (B6.Cg-Tg(Gfapcre)73.12Mvs/J; The Jackson Laboratory, RRID:IMSR_ JAX:012886), where Cre recombinase expression is driven by the promotor for the astrocytic marker glial fibrillary acidic protein (GFAP), were used to target expression to SCN astrocytes. Tail samples were sent to an external facility for genotyping (Transnetyx, Inc).

Stereotaxic surgery
Male and female GFAP-Cre mice were anesthetized with isoflurane and fixed in a stereotaxic frame. Next, two 90-nl boluses, 30 s apart of either AAV9.CAG.Flex. GCaMP6m.WPRE.SV40 (Penn Vector Core), AAV5-DIO-hM3Dq-mCherry (Addgene), or an equal mixture were bilaterally injected into the SCN to coordinates: x, -0.4, y, Ϯ0.2, and z, -5.8 from bregma. SCN (150 m) slices were prepared on days 10 -21 after injection as described above. Successful injections were confirmed by observing expression of the green reporter GCaMP6 or mCherry expression from the Gq DREADD in the SCN.

Calcium imaging
Ca 2ϩ measurements were obtained by recording images at an excitation wavelength of 380 nm supplied via a Lambda 10-3 filter wheel (Sutter Instrument Company) passing through a UG11 optical filter to restrict harmonic wavelengths above 400 nm, reflected via a 400-nm DCLP dichroic mirror, and through a Leica 40ϫ/0.80 UV water immersion objective (Leica Biosystems). Emitted light passed through a 510 Ϯ 40 nm filter (Chroma Technology) and was recorded with a cooled charge-coupled device camera (CCD-1300-Y/HS, Princeton Scientific) with acquisition time and binning adjusted to minimize photobleaching and maximize recording speed via the digital imaging software Metafluor (Molecular Devices).
To record the long-term changes in [Ca 2ϩ ]i due to bath application of WIN, images were sampled every 10 s. This was chosen based on previous work investigating Ca 2ϩ signaling in cortical astrocytes, which demonstrated that most agonist-induced events lasted longer than 10 s (Poskanzer and Yuste, 2016). Regions of interest (ROIs) were selected based on visual identification of independent intensity differences from the rest of the slice. To evaluate changes of ROI intensity after drug treatment, the baseline was defined as the 300 s before test agent application. ⌬F/F ϭ (ROI intensity-baseline intensity)/ baseline intensity. To eliminate experimenter bias in defining a positive or negative response, drug-induced change in Ca 2ϩ was defined as a ⌬F/F Ͼ2 SD away from baseline (Poskanzer and Yuste, 2016).
In separate experiments measuring faster, spontaneous Ca 2ϩ events, and for paired electrophysiological recordings and GCaMP imaging of spontaneous Ca 2ϩ events, images were acquired every 2 s, based on previous work in the SCN investigating spontaneous Ca 2ϩ signals (Brancaccio et al., 2017), and to enable better resolution for washout experiments without photobleaching the epifluorescence. For these experiments, ⌬F/F ϭ ROI intensity/ average slice intensity. Average slice intensity was used to eliminate large baseline shifts in intensity on drug treat-ment that may have been due to decreasing CB1R activity with AM251, or by photobleaching produced by the higher acquisition rate. An event was defined as a time point that was Ͼ2 SD away from the average intensity of the previous 30 s.

Bioluminescence assays
Long-term organotypic slice cultures of the SCN were prepared from PER2::LUC mice. These slices allow for long-term recording of molecular clock rhythms that follow stable patterns of expression throughout the entire recording session, indicating that the SCN network remains largely intact (Yoo et al., 2004). Slices were treated within the first 4 d in the early subjective day, circadian time (CT)1-CT6 (where CT 12 is defined as peak bioluminescence), for 1 h with 3 M WIN (Sigma-Aldrich), 0.09% DMSO, 0.2 M DPCPX, a combination treatment of WIN Research Article: New Research and DPCPX, 100 M adenosine (Sigma-Aldrich), or 0.38% sterile water. Data were acquired and analyzed with Lumicycle Analysis software (Actimetrics, Inc). Baseline drift in Lumicycle recordings was corrected by subtracting a third order or less polynomial followed by fitting with a damped sine wave. Only recordings with Ͼ75% goodness-of-fit to a dampened sine wave were used for analysis. Phase shifts were calculated by comparing two predictions for the time of the first peak post-treatment: one prediction calculated from at least three cycles before treatment and a second prediction derived from at least three cycles after treatment. The difference between these two predictions indicated the size of the phase shift (Besing et al., 2012;Hablitz et al., 2014).

Statistical analysis
All statistical analysis was performed with PASW Statistics 18 (IBM Analytics). For comparisons of means in samples with normal distributions and homogeneous variances (as indicated by a Levene's test), an independent samples t test or ANOVA was used for comparisons between two means or two or more means, respectively, followed by Fisher's Bonferroni adjusted post hoc test when necessary. In cases of a non-normal distribution (as indicated by a Shapiro-Wilk test) or unequal variances (Levene's test), a nonparametric Mann-Whitney U test or Kruskal-Wallis test was used for comparisons between two means or two or more means, respectively, followed by a median test for post hoc analyses. Significance was ascribed at p Ͻ 0.05. A repeated measures analysis was used for change over time within a cell or a slice, when appropriate. In the case of a non-normal distribution, a Friedman test followed by post hoc Wilcoxon signed-rank tests with the alpha level Bonferroni adjusted for multiple comparisons was used.
After demonstrating that WIN decreases the frequency of mGPSCs, we next sought to determine whether astrocytes played a role in cannabinoid signaling. Astrocytic metabolic function was inhibited with FC (1 M), an inhibitor of the Krebs cycle preferentially taken up by astrocytes (Navarrete and Araque, 2008). FC application did not significantly change mGPSC frequency or amplitude compared to 0.01% DMSO controls (Fig. 1F,G). However, FC completely occluded the WIN reduction of the mGPSC frequency (FCϩWIN: -2.0 Ϯ 8.58%), and had no effect on the mGPSC amplitude, indicating that proper astrocytic functioning was necessary for the presynaptic effects of WIN (Fig. 1F,G; Tables 1-Tables 3).

The CB1/2R agonist WIN induces an increase in intracellular astrocytic Ca 2؉
To understand how cannabinoids influence astrocyte signaling, we targeted astrocytes by using a mouse line where Cre recombinase expression is driven by the astrocytic marker GFAP. To validate the specificity of the Cre recombinase expression to astrocytes in the SCN, we bred GFAP-Cre mice with the chloride sensor (Clsensor) mouse line, a chloride reporter that has robust expression and a strong fluorescent signal (Waseem et al., 2010;Klett and Allen, 2017). We then compared the expression of the Clsensor to that of neuronal and glial cells markers in the SCN. The GFAP staining and fluorescent Clsensor signal The treatment/control ratio was calculated for each recorded neuron and then averaged for each treatment group.
co-localized completely ( Fig. 2A-F). Clsensor expression was not detected in cells positive for NFHC ( Fig. 2G-L) or the Neural/Glial antigen proteoglycan (data not shown), a marker for polydenrocyte glia (Nishiyama et al., 2009). Note that NFHC was used instead of more commonly used neuronal markers, such as neuronal nuclear antigen and neuron specific enolase, which are barely detectable in SCN neurons (Morin et al., 2011;Moldavan et al., 2015). These data demonstrate that the GFAP-Cre mice are an appropriate model to target astrocytes in the SCN. Endocannabinoids recruit astrocytes to mediate synaptic transmission by initiating intracellular Ca 2ϩ signaling cascades (Navarrete and Araque, 2010;Bindocci et al., 2017). Here, we tested the hypothesis that activation of CB1/2Rs activates an intracellular Ca 2ϩ signaling pathway in SCN astrocytes (Fig. 3). An adeno-associated virus containing GCaMP6, an intensity based Ca 2ϩ reporter that is flanked by loxP sites (Chen et al., 2013), was injected into the SCN of GFAP-Creϩ animals to enable monitoring of Ca 2ϩ signaling in SCN astrocytes. Astrocyte regions were defined as soma or non-soma by shape; somas were identified as more circular with thin processes branching from the center. This distinction was made because astrocytes differentially, spatiotemporally, regulate Ca 2ϩ influxes throughout their somas and processes Tong et al., 2013;Bindocci et al., 2017). Increases or decreases of intracellular Ca 2ϩ were defined as events if the amplitude was Ͼ2 SD from baseline, with variable responses showing both a significant increase and a significant decrease (Irwin and Allen, 2013). WIN (3 M) application increased [Ca 2ϩ ]i in 52.5% of the somas. The non-soma regions showed similar responses with increased [Ca 2ϩ ]i in 55.2% (Fig. 3B,F). Controls were treated with 0.01% DMSO, the concentration     (Fig. 3D,H) (Fig. 3H). The magnitude of these responses were smaller than responses without thapsigargin, although only the non-soma group, where the sample size was larger, was statistically significant (non-soma: 0.122 Ϯ 0.011 ⌬F/F, soma: 0.123 Ϯ 0.012 ⌬F/F; Table 3). These data demonstrate that activation of CB1 receptors by WIN activates an intracellular Ca 2ϩ signaling pathway in SCN astrocytes.

Neurons utilize endocannabinoid signaling to alter astrocyte function
We demonstrated that WIN-induced CB1R activation increased astrocytic [Ca 2ϩ ]i (Fig. 3). However, agonistinduced activation of CB1R does not address endogenous, spontaneous Ca 2ϩ signaling in astrocytes. Spontaneous Ca 2ϩ events change both local and network wide intracellular signaling within and across astrocytes, disruption of which has been linked to neuropathology (Shigetomi et al., 2016). We determined the role of endogenous cannabinoid signaling in regulating these spontaneous Ca 2ϩ events within GCaMP6 expressing astrocytes by applying AM251 (Fig. 4). Total event numbers were summed for the 3 min prior, 3 min during, and 3 min after AM251 treatment. AM251, reduced the frequency of [Ca 2ϩ ]i events in 42.9% of somas. In non-soma regions AM251 reduced the number of spontaneous Ca 2ϩ events in 44.1% of regions (Fig. 4D). We found no difference in event amplitude between soma and non-soma regions before, during, or after AM251 treatment ( Fig. 4E; Table 4).
Both soma and non-soma regions responded similarly in the regions in which AM251 decreased the spontaneous Ca 2ϩ event frequency, with the number of events decreasing during treatment and a significant recovery during 3 min of washout ( Fig. 4C; Table 4). Similarly, soma and non-soma regions with an increased frequency of Ca 2ϩ events saw a return to baseline after treatment  Endocannabinoid signaling acts as a retrograde signal from a postsynaptic neuron to presynaptic axon terminals, and our model suggests that in the SCN astrocytes are necessary for this process, responding to a postsynaptic neuronal release of endocannabinoids with a Ca 2ϩ signaling event. To test this hypothesis, neurons were recorded in whole cell patch clamp mode and voltageclamped at -60 mV, then depolarized (10 pulses from Ϫ80 to ϩ20 mV, 100 ms duration, 5 Hz) while Ca 2ϩ events were recorded from GCaMP6 expressing astrocytes (Fig.  5). The number of Ca 2ϩ events 30 s before and after neuronal depolarization were compared before, during and after AM251 application to determine whether depolarization increased Ca 2ϩ events in a CB1R-dependent manner. Before cannabinoid receptor blockade, the depolarization protocol significantly increased the number of spontaneous Ca 2ϩ events (pre-depol: 1.47 Ϯ 0.08 events, post-depol: 2.10 Ϯ 0.07 statistics in Table 4). AM251 treatment prevented the depolarization-induced increase in Ca 2ϩ event frequency (pre-depol: 1.18 Ϯ 0.07 events, post-depol: 1.33 Ϯ 0.07 events), an effect that recovered after 3 min of washout (pre-depol: 1.04 Ϯ 0.06 events, post-depol: 1.72 Ϯ 0.06 events; Fig. 5C,D). There were no differences in event amplitude distributions among any groups (Fig. 5E). These experiments demonstrated that depolarization of a postsynaptic neuron activated Ca 2ϩ signals in astrocytes that were dependent on endocannabinoid signaling.

Adenosine signaling is necessary for WIN response
Cannabinoid signaling, both via endocannabinoids and the synthetic cannabinoid WIN, activated a Ca 2ϩ response in astrocytes, and WIN can decrease release of GABA in SCN neurons in an astrocyte-dependent manner. Next, we wanted to identify the mechanism by which astrocytes decrease presynaptic GABA release. In other areas of the brain, astrocytes release glutamate that alters presynaptic function by activating metabotropic glutamate receptors (Liu et al., 2004;Perea and Araque, 2007;Rusakov, 2015). We demonstrated that this is not the case in the SCN by using ACPTII, a competitive metabotropic receptor inhibitor. ACPTII alone did not significantly alter mGPSC frequency or amplitude. However, on addition of WIN the mGPSC frequency significantly decreased (ACPTIIϩWIN: -19.2 Ϯ 4.5%, statistics in Table 5) without significantly altering amplitude, indicating metabotropic glutamate receptors are not required for cannabinoidinduced astrocyte-mediated presynaptic changes ( Fig.  6A; Table 5).

Activation of a Ca 2؉ response in astrocytes changes neuronal function in the SCN
After demonstrating that cannabinoids increase [Ca 2ϩ ]i in SCN astrocytes and release adenosine, we next asked whether activation of astrocytic Ca 2ϩ signaling pathways could cause an adenosine-dependent decrease in mG-PSC frequency independent of cannabinoid signaling. Designer receptors exclusively activated by designer drugs (DREADDs) are mutated muscarinic acetylcholine receptors that respond only to CNO, a metabolite of acetylcholine not traditionally found in mice, allowing for cell type specific expression and activation of G-protein signaling (Rogan and Roth, 2011). We expressed Gq coupled DREADDs together with GCaMP6 in GFAP-Creϩ astrocytes (Fig. 8). Application of the DREADD agonist CNO increased [Ca 2ϩ ]i in 55.6% of soma regions (13% no response, 16.7% decrease, and 14.8% variable), and produced a similar response in non-soma regions (56.2% increase, 9.4% no response, 20.8% decrease, and 13.6% variable, statistics in Table 5; Fig. 8C). The magnitude of the [Ca 2ϩ ]i increases was not significantly different between soma and non-soma regions (soma: 0.402 Ϯ 0.0828 ⌬F/F, non-soma: 0.269 Ϯ 0.018 ⌬F/F; Fig. 8D). After depleting Ca 2ϩ stores with thapsigargin (1 M), the magnitude of the CNO-induced Ca 2ϩ responses in the soma and non-soma regions was reduced (soma: 0.126 Ϯ 0.007 ⌬F/F, non-soma: 0.169 Ϯ 0.038 ⌬F/F). Similar to the WIN ϩ thapsigargin experiment, this effect was significant in the non-soma groups but not the soma groups due to smaller sample size of cell body regions after Ca 2ϩ depletion with thapsigargin (Table 5). Application of CNO in the absence of Gq DREADD expression had no effect on the mGPSC amplitude (CNO without DREADD: Ϫ4.8 Ϯ 1.4%) or frequency (CNO without DREADD: Ϫ10.2 Ϯ 2.1%) compared to DMSO controls (Table 6). Because CNO did not alter the mGPSC frequency or amplitude when applied without the Gq DREADD, and DMSO does not significantly increase GCaMP6 signaling (Fig. 3) we did not investigate the effects of CNO (10 M) on GCaMP6 signaling in the absence of DREADD expression.
We hypothesized that activating a Ca 2ϩ response in astrocytes would decrease presynaptic GABA release in the SCN because the Gq DREADD activation increased [Ca 2ϩ ]i similar to that induced by WIN. Indeed, mGPSC frequency was significantly reduced in CNO treated groups compared to controls (CNO: -18 Ϯ 4%) while the amplitude showed no difference (Fig. 9A-C; Tables 4-Tables 6). The reduction of mGPSC frequency by WIN was eliminated by blocking A1Rs with DPCPX (DPCPXϩCNO: -6.1 Ϯ 6.5%; Table 6) while the mGPSC amplitudes were not   (Fig. 9D-F; Tables 1, 6), indicating that like cannabinoid signaling, activation of a Ca 2ϩ response in astrocytes initiates a presynaptic inhibition of GABA release dependent on adenosine receptor signaling.

Both cannabinoid and adenosine signaling change circadian clock timing
Cannabinoid signaling alters both astrocytic and neuronal function in the SCN. Cannabinoids also block lightinduced phase shifts, but had no effect on clock timing at night (Sanford et al., 2008;Acuna-Goycolea et al., 2010). Experiments were performed to determine whether CB1R activation can phase shift the molecular clock when administered during the day using the PERIOD2::LU-CIFERASE mouse model (Yoo et al., 2004). We applied WIN (3 m) for 1 h to PER2::LUC SCN cultures during the early day (CT1-CT6). WIN significantly phase advanced PER2::LUC rhythms compared to DMSO (0.01%) vehicletreated controls (WIN: 2.5 Ϯ 1.3 h, vehicle: 0.2 Ϯ 0.1 h), indicating that cannabinoids can indeed influence circadian timing ( Fig. 10; Table 6). We tested the dependency of this phase shift on A1R signaling. We find the A1R antagonist DPCPX does not induce a phase shift when applied for 1 h during the day (DPCPX: 0.3 Ϯ 0.2 h), and co-treatment of WIN and DPCPX does not produce the phase advance seen with WIN alone (DPCPXϩWIN: 0.1 Ϯ 0.2 h; Fig. 10B; Table 6). There was no change in period length between pre-and post-treatment under any treatment paradigm (pre-treatment period: 25.9 Ϯ 0.1 h, post-treatment period: 25.8 Ϯ 0.2 h; Table 6). These data indicate that cannabinoid signaling in the clock center of the brain can alter circadian timing in an A1R-dependent manner. Since the WIN-induced decrease in mGPSC frequency and phase advance of PER2::LUC rhythms are dependent on A1R activation, we hypothesized that adenosine would have similar phase shifting effects as WIN. Indeed, application of adenosine (100 M) during the early day for 1 h significantly phase advanced the PER2:: LUC rhythms compared to vehicle-treated controls (mean Ϯ SEM, adenosine: 3.7 Ϯ 1.7 h, vehicle: 0.5 Ϯ 0.3 h (Fig.  10C,D; Table 6). Similar to the WIN experiments, there was no significant effect of adenosine treatment on period length (Table 6), although there was a significant shortening of period after treatment (pre-treatment period: 27.1 Ϯ 0.4 h, post-treatment period: 25.5 Ϯ 0.3 h) which may have been due to a different vehicle (water instead of DMSO).

Discussion
In the SCN, postsynaptic neurons recruit astrocytes via endocannabinoid signaling to modulate presynaptic GABA release. Activation of CB1Rs with the cannabinoid agonist WIN decreases the frequency but not the amplitude of mGPSCs in the SCN, consistent with a reduction in GABA release. This effect is dependent on astrocyte function, and A1R activation. WIN also induces an astrocytic Ca 2ϩ signaling cascade. Mimicking this increased Ca 2ϩ signal in astrocytes using DREADD technology   causes a decrease in mGPSC frequency that is also dependent on A1R activation. Finally, blockade of endogenous CB1R activation with AM251 decreases both spontaneous Ca 2ϩ events, and the number of Ca 2ϩ events induced by depolarization of a postsynaptic neuron indicating neuronal-derived endocannabinoid signaling modulates astrocytic Ca 2ϩ signaling in the SCN. We propose a model whereby postsynaptic neuronal activity generates endocannabinoid release, activating cannabinoid receptors on astrocytes and activating an intracellular Ca 2ϩ signaling pathway, causing the release of adenosine and activation of A1Rs on the presynaptic neuron to decrease GABA release. Adenosine itself decreases mGPSC frequency, supporting this hypothesis. In addition to this novel model of astrocyte recruitment to modulate GABAergic signaling in the SCN, daytime application of either WIN or adenosine phase advanced PER2::LUC rhythms, indicating a conserved mechanism for modulation of circadian timing.
There are a multitude of cannabinoids, from plant derived such as cannabidiol or (-)-trans-⌬ 9 -tetrahydrocannabinol found in marijuana, to endocannabinoids produced by the body such as 2-arachidonoylglycerol and anandamide, to pharmacological agents designed to activate specific cannabinoid receptors (Stella, 2010;Le Boisselier et al., 2017). The affinity of these compounds for cannabinoid receptors varies greatly (McPartland et al., 2007) and cannabinoid receptors can have multiple binding sites (Lauckner et al., 2005;McPartland et al., 2007;Khajehali et al., 2015;Hua et al., 2017). To control for these confounds, we not only activated CBRs via the potent agonist WIN, but blocked CB1Rs, specifically, with AM251 to investigate the role of endocannabinoid signaling instead of relying on observations made with exogenous cannabinoids. Conventionally, CB1Rs are identified as Gi-coupled G-protein-coupled receptors, but WIN binds CB1 in such a way that it couples to Gq G-proteins and promotes release of Ca 2ϩ from internal stores (Lauck-  , 2005). Our data supports these findings such that the effects of WIN in the SCN are dependent on CB1R activation, Gq signaling increases astrocytic intracellular Ca 2ϩ to a similar extent as WIN, and blockade of CB1R decreases Ca 2ϩ events. Further work must be done to identify which endocannabinoid, specifically, is responsible for these Ca 2ϩ events, but given the complexity of CB1R binding and function this may be difficult to discern. Traditionally, cannabinoids act as a retrograde signal whereby after a depolarizing event a postsynaptic cell synthesizes and releases endocannabinoids that activate cannabinoid receptors on the presynaptic cell to regulate neuronal excitability . Our model supports previous work challenging this model, implicating astrocytes as necessary intermediates to fine tune retrograde responses (Navarrete and Araque, 2008Araque, , 2010Min et al., 2012;Di et al., 2013;Viader et al., 2015). In the SCN, there has been growing evidence that astrocytes actively modulate neuronal function and circadian rhythmicity (Barca-Mayo et al., 2017;Brancaccio et al., 2017;Tso et al., 2017), broadening the role of astrocytes from mediating retrograde responses from neurons to actively integrating time-of-day information to the SCN network. A recent study demonstrated that astrocyte-specific deletion of BMAL1, an essential circadian gene, causes bimodal activity of animals in constant darkness. Blockade of GABA(A) over several days was able to rescue this be-havioral phenotype, indicating that astrocytes regulate GABAergic signaling (Barca-Mayo et al., 2017). The current study supports these observations, demonstrating that astrocytes are necessary for modulation of GABAergic tone. Utilizing the circadian tau mutant, where a selective deletion of a mutant form of casein kinase in astrocytes causes a circadian period mismatch in neurons and astrocytes of 22 and 24 h, respectively, astrocytes have been shown to influence circadian free-running period, or day length (Tso et al., 2017). Future work will investigate the ability of astrocytes to synchronize SCN neurons, perhaps via cannabinoid signaling.
We chose to study cannabinoid signaling in the SCN during the day for several reasons. First, previous work demonstrated that endocannabinoids may act as a nonphotic cue (Acuna-Goycolea et al., 2010), which are more effective during the day. Next, the endogenous CB1 agonist 2-arachidonoyl-glycerol shows a diurnal rhythm, with a peak during the day (Valenti et al., 2004;Liedhegner et al., 2014). Finally, endocannabinoid signaling is different from many transmitter signaling systems in that the endocannabinoids are synthesized on demand in response to increases in neuronal activity (Navarrete and Araque, 2008), and SCN neurons are most active during the day (Green and Gillette, 1982). Astrocytes in the SCN have increased global intracellular calcium at night, with relatively little calcium signaling during the day (Brancaccio  model suggests this increased calcium at night reflects active extracellular glutamate buffering by the astrocytes via NMDAR dependent mechanisms (Brancaccio et al., 2017(Brancaccio et al., , 2019. Although providing an explanation to the potential roles of astrocytes in setting periodicity and neuronal activity at night, relatively little is known about astrocytic function during the day. We suggest that decreased [Ca 2ϩ ] i during the day fits our model, enabling astrocytes to fine-tune neuronal activity locally at synapses compared to broadly buffering glutamate across the SCN. This is supported both by AM251 and FC having no direct effect on neuronal activity and by AM251 decreasing spontaneous, local astrocytic [Ca 2ϩ ] i , supporting our hypothesis that endogenous cannabinoid signaling in the SCN involves astrocytes. One endogenous, physiologic condition in the SCN that could induce cannabinoid signaling may be a light stimulus that excites neurons broadly over the SCN. Future work may test the hypothesis that cannabinoid signaling in astrocytes is a mechanism to fine-tune neuronal responses to light.