Astrocytes Control Circadian Timekeeping in the Suprachiasmatic Nucleus via Glutamatergic Signaling

Abstract

The suprachiasmatic nucleus (SCN) of the hypothalamus orchestrates daily rhythms of physiology and behavior in mammals. Its circadian (∼24 hr) oscillations of gene expression and electrical activity are generated intrinsically and can persist indefinitely in temporal isolation. This robust and resilient timekeeping is generally regarded as a product of the intrinsic connectivity of its neurons. Here we show that neurons constitute only one "half" of the SCN clock, the one metabolically active during circadian daytime. In contrast, SCN astrocytes are active during circadian nighttime, when they suppress the activity of SCN neurons by regulating extracellular glutamate levels. This glutamatergic gliotransmission is sensed by neurons of the dorsal SCN via specific pre-synaptic NMDA receptor assemblies containing NR2C subunits. Remarkably, somatic genetic re-programming of intracellular clocks in SCN astrocytes was capable of remodeling circadian behavioral rhythms in adult mice. Thus, SCN circuit-level timekeeping arises from interdependent and mutually supportive astrocytic-neuronal signaling.

Keywords: GABA; NMDAR2C; SCN; astrocytic-neuronal interactions; calcium oscillations; circadian; circadian behavior; circuit synchronization; extracellular glutamate; membrane potential oscillations.

Figure 1

Circadian Landscape of Interlocked, Anti-phasic Oscillations in SCN Neurons and Astrocytes (A–E) Photomicrographs and representative traces of SCN slices transduced with AAVs encoding Syn-RCaMP1h reporting neuronal [Ca²⁺]_i paired with (A) Syn-ArcLightD reporting membrane voltage; (B) gfaABC1D-GCaMP3 reporting astrocytic [Ca²⁺]_i; (C) gfaABC1D-LCK::GCaMP6 detecting [Ca²⁺]_i in astrocytic microdomains; and (D and E) iGluSnFR driven by Syn or GFAP, respectively, to detect [Glu]_e localized on neuronal or astrocytic cell membranes. ArclightD y axis inverted for presentation purposes. SCN area is outlined (white dashed lines) for presentation purposes. (F) Circadian landscape of reporters in (A)–(E), plotted relative to phase of RCaMP1h (n depicted on graph; each dot is a biological replicate). PER2::LUC is used to register circadian time (CT12). Photomicrographs are false colors LUT ΔF/F. (G) Fluorescent traces and ratios showing GFAP-iGluSnFR fluorescence upon MSO treatment, compared with co-expressed GFAP-hM3DGq::mCherry, used as an internal control. n = 3 for both reporters. (H) Representative traces of circadian oscillations of PER2::LUC in SCN slices treated with MSO or vehicle, in serum-free conditions. (I and J) Bar graphs showing amplitude ratio (I) and relative amplitude error (RAE) ratio (J) in SCN slices treated with MSO or vehicle. (K) Representative traces of PER2::LUC oscillations of SCN slices treated with astrocytically (blue line) or neuronally (red line) restricted flex-taCasp3-TEV to specifically ablate those populations. (L) Mean ± SEM iGluSnFR traces from astrocytically ablated or neuronally ablated SCN slices. Scale bars, 100 μM. Measurement windows in gray. (M) Bar graphs showing variations of iGluSnFR fluorescent intensity ratios in astrocytically or neuronally ablated SCN slices. All bar graphs are mean ± SEM; n experimental replicates depicted on bars. Statistical tests are as follows: (I and J) one-way ANOVA, Bonferroni corrected; (M) two-tailed paired t test. ^∗p < 0.05; ^∗∗∗∗p < 0.0001. See also Figures S1–S4 and Movie S1.

Figure 2

Glutamatergic Regulation of Circadian Synchrony of SCN Neurons (A) Representative PMT PER2::LUC traces and bar graphs of amplitude and RAE ratio of SCN slices treated with a cocktail of inhibitors of glial glutamate transporters (EAAT1 and EAAT2) (20 μM UCPH-101 + 10 μM WAY-213613), or with DL-TBOA (200 μM), which also inhibits the neuronal EAAT3. (B) Representative multi-channel imaging traces of Syn-RCaMP1h, PER2::LUC, and Syn-iGluSnFR cellular oscillations plus mean traces (black lines) showing effects of DL-TBOA treatment on these reporters. (C and D) Representative Rayleigh plots (C) and bar graphs of mean vector length (D) showing Syn-RCaMP1h, PER2::LUC, and iGluSnFR oscillators before DL-TBOA treatment, in the presence of the drug and after washout (N_osc/SCN > 100, n = 3). (E) Period scatter of RCaMP1h, PER2::LUC, and iGluSnFR cellular reports in SCN treated with DL-TBOA. (F) Circadian waveforms of [Glu]_e and [Ca²⁺]_i, and corresponding estimates of skewness for SCN treated with DL-TBOA. Bar graphs are mean ± SEM; n experimental replicates depicted on bars. Statistical tests are as follows: two-way repeated-measures ANOVA, Bonferroni corrected. ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001; ^∗∗∗∗p < 0.0001. See also Figure S5.

Figure 3

NMDAR Assemblies Containing NR2C Mediate Effects of Extracellular Glutamate on Circadian Rhythmicity in SCN Slices (A and B) Representative PMT traces from SCN slices expressing PER2::LUC pre-treated with MK-801 (12.5 μM) and DNQX (12.5 μM), alone or in combination, before DL-TBOA (A) or vehicle addition (B). (C) Bar graphs showing effects of pre-treatment with MK-801, alone or in combination with DNQX in slices treated with DL-TBOA or vehicle, respectively. (D) Representative PMT traces of PER2::LUC SCN slices treated with the NR2C antagonist DQP-1105 (50 μM) or vehicle. (E and F) Bar graphs showing (E) amplitude ratio and (F) period difference in SCN slices treated with DQP-1105, when compared to slices treated with vehicle and NR2A or NR2B antagonists (1 μM TCN-201 and 1 μM Ro-25-6981, respectively). All bar graphs mean ± SEM; n experimental replicates depicted on bars. Statistical test is as follows: (A–F) two-way repeated-measures ANOVA, Bonferroni corrected; n experimental replicates depicted on bars. ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗∗p < 0.0001. See also Figure S6.

Figure 4

NR2C Inhibition Alters Spatiotemporal Wave of Clock Gene Expression by Selectively Impairing Synchronization of Dorsal NR2C⁺ SCN Neurons (A) Confocal micrographs of SCN slices from Grin2C-iCre mice transduced with AAV-CAG-flex-tdTomato and counterstained with AVP antiserum to highlight SCN cytoarchitecture. (B) Representative frame from time-lapse recording of PER2::LUC SCN slices from Grin2C-iCre mice expressing flex-tdTomato to label NR2C⁺ neurons. Boxed areas show dorsal and ventral SCN, as defined by tdTomato signal. (C and D) Raster plots (C) and bar graphs (D) of amplitude ratio of PER2::LUC oscillations in ventral and dorsal regions of SCN slices treated with DQP-1105. (E and F) Representative Rayleigh plots (E) and bar graphs (F) of mean vector length of ventral and dorsal PER2::LUC oscillators, as defined by Grin2C-tdTomato expression, before DQP-1105 treatment and in the presence of the drug. (G and H) (G) Photomicrographs and Poincaré plots of spatiotemporal waves of PER2::LUC quantified by CoL before (black trajectory) and during (white trajectory; red on Poincaré plot) DQP-1105 treatment, within a single SCN and across several replicates (H). SCN outline is for presentation purposes. All bar graphs are mean ± SEM; n experimental replicates depicted on bars. Statistical tests are as follows: (D and F) two-way repeated-measures ANOVA, Bonferroni corrected; (G) paired two-tailed t test, n = 3. ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗∗p < 0.0001. Scale bars, 50 μm. See also Figure S7 and Movie S2.

Figure 5

NR2C Inhibition Depolarizes SCN Neurons and Impairs Coherence of Circadian Oscillations of Membrane Potential, [Ca²⁺]_i, and Clock Gene Expression (A) Representative electrophysiological traces showing membrane potential recorded from dorsal SCN neurons during the circadian day (CT3.5–CT9.5) or night (CT15.5–CT21.5). Black bars indicate serial vehicle and DQP-1105 treatments. Traces rescaled to highlight the extent of DQP-1105-induced depolarization. (B) Scatterplots of change in membrane potential during focal DQP-1105 application relative to DMSO (mean ± SEM; CT3.5–CT9.5, n = 17; CT15.5–CT21.5, n = 18; each time point sampled from four SCNs; unpaired two-tailed t test). Projected circadian oscillations of neuronal [Ca²⁺]_i and [Glu]_e are shown for presentation purposes. (C) Scatterplots of change in spontaneous firing rates (SFRs), during either DQP-1105 or DMSO treatment, normalized to baseline SFR, recorded immediately before treatment (two-way repeated-measures ANOVA, with Bonferroni correction; n = 17). (D) Representative aggregate traces showing changes of simultaneously recorded neuronal [Ca²⁺]_i, membrane potential, and PER2::LUC, induced by DQP-1105 and after washout. (E and F) Bar graphs showing severe reduction in the number (E) and quality (high RAE) (F) of oscillations of neuronal [Ca²⁺]_i, membrane potential, and PER2::LUC at single-cell level (mean ± SEM, N_osc/SCN > 60, n = 3; two-way repeated-measures ANOVA, Bonferroni corrected; ^∗∗∗p < 0.001).

Figure 6

NR2C Antagonism Decouples Pre-synaptic and Cytosolic [Ca²⁺]_i in SCN Neurons (A and B) (A) Circadian profiles (mean ± SEM, n = 5) of SCN slices co-transduced with SyF::GCaMP3 and Syn-RCaMP1h AAVs to detect pre-synaptic or cytosolic [Ca²⁺]_i, respectively. Gray area shows higher nighttime fluorescence of pre-synaptic [Ca²⁺]_i in comparison to cytosolic [Ca²⁺]_i and circadian variation of pre-synaptic/cytosolic neuronal [Ca²⁺]_i ratios (B). (C) Representative traces of SyF::GCaMP3 and Syn-RCaMP1h fluorescence in SCN slices before and during DQP-1105 treatment. (D) Mean ± SEM of baseline SyF::GCaMP3 and Syn-RCaMP1h fluorescence in the presence of DQP-1105. (E) Mean ± SEM ratios of cytosolic/pre-synaptic baseline fluorescence before DQP-1105 and in the presence of the drug (paired two-tailed t test; ^∗p < 0.05, n = 5). (F) Cartoon depicting the proposed model for the astrocytic-neuronal intercellular axis in the dorsal SCN.

Figure 7

SCN Astrocytes Control Spatiotemporal Wave of Circadian Gene Expression in Juvenile Slices (A) Schematic of strategy, with floxed Ck1ε^Tau/Tau PER2::LUC SCN transduced with AAVs encoding Cre recombinase driven by Syn or GFAP promoters to assess the effects of neuronal and astrocytic Ck1ε^Tau/Tau knockout on circadian rhythms. (B) Representative PMT traces of PER2::LUC SCN from neuronal- or astrocyte-restricted Ck1ε^Tau/Tau knockout (mean ± SEM, p < 0.0001, n = 4). (C) Bar graphs showing period of PER2::LUC, neuronal, and astrocytic [Ca²⁺]_i circadian oscillations co-detected in neuronal- or astrocyte-restricted Ck1ε^Tau/Tau knockout, respectively. (D) Poincaré plots revealing CoL dorsalization in astrocytic-restricted Ck1ε^Tau/Tau knockout, within a single SCN (left) and across different SCNs (right, n = 3; ANOVA). SCN outlined for presentation purposes. Bar graphs are mean ± SEM; n experimental replicates as depicted on bars. Statistical test is as follows: two-way repeated-measures ANOVA, Bonferroni corrected; ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001; ^∗∗∗∗p < 0.0001. dpt, days post-transduction. See also Movie S3.

Figure 8

SCN Astrocytes Control Circadian Patterns of Locomotor Activity in Adult Mice (A) Representative double-plotted actograms of wheel-running behavior in Ck1ε^Tau/Tau mice in DD (gray area), following stereotaxic SCN targeting by AAVs expressing Syn- or GFAP-restricted Cre; absent in EGFP-expressing controls. Note period lengthening following surgery in both Syn- and GFAP-Cre mice. Insets show effective targeting of SCN corresponding to plotted actograms and distinct morphologies of Syn/GFAP-EGFP targeted SCN cells. (B) Bar graphs showing period in DD before and after stereotaxic surgery in Syn- and GFAP-Cre targeted mice and respective EGFP controls. Windows of period detection are color-coded as reported on the actograms. Bar graphs are mean ± SEM; n experimental replicates depicted on bars. (C) Confocal micrographs showing counterstaining of GFAP-mCherry::Cre with the astrocytic markers GFAP and Aldh1L1. (D) Inset: merge showing co-localization of mCherry⁺ with GFAP and Aldh1L1 (arrows) in SCN targeted area as highlighted in (C). Fluorescent signals also presented as single channels. (E) Co-localization rates of GFAP-mCherry::Cre⁺ cells with the astrocytic markers GFAP and Aldh1L1 (N_{GFAP-mCherry::Cre}⁺ = 328; n = 3). Statistical test is as follows: two-way repeated-measures ANOVA, Bonferroni corrected; ^∗∗∗∗p < 0.0001. Scale bars, 50 μm. See also Figure S8.