GABA in the suprachiasmatic nucleus refines circadian output rhythms in mice

We examined the effect of GABA deficiency on the circadian rhythms of spontaneous firing, intracellular Ca²⁺ level, and circadian PER2 rhythm in the fetal SCN cultured on MED. Since the VGAT^−/− mice do not survive after birth³¹, we obtained VGAT^−/− SCN slices at the embryonic day 19 or 20 together with the wild type (WT) and VGAT^+/− littermates from VGAT^+/− females crossed with VGAT^+/− males.

Multiunit activity in the SCN slice was measured continuously for more than 5 circadian cycles and spontaneous discharges with signal to noise ratio (S/N) > 2 were collected from each electrode (Supplementary Fig. 1). The frequency of burst firings (Hz) was evaluated for 10 min records in 100 ms bins (6000 bins) at CT8 when the highest neuronal activity was observed in both VGAT^−/− and WT mice (Supplementary Fig. 2) and the relative abundance of spike expressed as the ratio (%) of bin numbers to the total was plotted against the frequency (Supplementary Fig. 3a). The bin ratio showed a marked decrease around 10 Hz and never exceeded 35 Hz in the WT SCN, whereas more than 3% of the total number of bins exceeded 35 Hz in the VGAT^−/− SCN (Supplementary Fig. 3b). Thus, the firing rate over 35 Hz was regarded as the burst firing in this study. The burst firing abruptly appeared for a few seconds and repeated at 2–3 min intervals (Supplementary Fig. 1). Representative patterns of burst firing in other SCN slices are also illustrated in Supplementary Fig. 2.

The mean firing rate was calculated at every 1 min from all electrodes located at the SCN region (9–19 electrodes/SCN, Supplementary Figs. 1 and 2). The firing rate was typically in the range of 2–18 Hz at the circadian peak in the WT SCN (Fig. 1a, b). By contrast, the circadian firing rhythms in the VGAT^−/− and GAD65^−/−/67^−/− SCN were characterized by two distinct bands of a lower and higher firing rate throughout the circadian phase (Fig. 1a from the VGAT^−/− and GAD65/67^−/− SCN). The band of higher firing rates represented the bins containing the burst firings (higher than 35 Hz) and the band of lower firing rates represented those without burst firings. The two bands were detected from the start of culture and the difference in the mean frequency (Hz) between the two bands was not changed throughout the course of culture (Fig. 1a, Supplementary Fig. 4a), whereas the firing rate itself steadily increased during this period in the VGAT^−/− SCN.

Burst firings in the VGAT^−/− SCN were detected in all electrodes in synchrony (Fig. 2a, b) throughout the entire SCN slice and strongly correlated in time with each other as shown in the heat map of correlation coefficients (Fig. 2c) which was significantly high in the VGAT^−/− SCN (Fig. 2d). Importantly, the occurrence of burst firing in the VGAT^−/− SCN was not dependent on the circadian phase (Fig. 2e, Supplementary Fig. 5). The firing rate and shape of burst varied among the SCN slices but was indistinguishable among the electrodes in the same SCN (Supplementary Fig. 2).

Despite the superposition of burst firing, Chi-square periodogram revealed significant (p < 0.01) circadian firing rhythms in the VGAT^−/− and GAD65^−/−/67^−/− SCN (Fig. 1b). Their circadian periods (VGAT^−/−, 23.63 ± 0.55 h; GAD65^−/−/67^−/−, 23.79 ± 0.44 h) were comparable to those of the WT (23.86 ± 0.42 h) and VGAT^+/− (23.35 ± 0.18 h).

To understand the effects of GABA deficiency on the SCN molecular clock, we examined the circadian PER2 rhythms in SCN slices from VGAT^−/− and GAD65^−/−/67^−/− mice carrying a bioluminescence reporter for the clock gene product PER2 (PER2::LUC)³². Bioluminescence emitted from the PER2 reporter was monitored with an electron-multiplying charge-coupled device (EM-CCD) camera together with spontaneous firing in the SCN.

PER2::LUC rhythms in the VGAT^−/− and GAD65^−/−/67^−/− SCN were as robust as those in WT or VGAT^+/− littermates (Fig. 1d, Supplementary Fig. 6a). Cosine curve fitting was applied to all pixels (4.3 μm × 4.3 μm) to obtain the acrophase (fitted circadian peak phase) (Fig. 1d). Acrophase maps (Fig. 1d, left) and distribution of acrophases by Rayleigh plot (Fig. 1d, right) indicated that the PER2::LUC rhythm of each pixel was synchronized in the entire VGAT^−/− SCN slice to a similar extent to those in the WT SCN. In addition, the distribution of the acrophase in terms of the mean length of vector (r) was not different among genotypes (Fig. 1e). The same results were obtained in the simultaneously recorded PER2::LUC and spontaneous firing of the GAD65^−/−/67^−/− SCN (Fig. 1a–e). The variability of the circadian period in terms of SD was not significantly different among genotypes.

The PER2::LUC rhythms of four genotypes were also recorded by photo multiplier tube (PMT). The circadian period, the standardized amplitude in the first circadian cycle and the damping ratio during the first six cycles were not significantly different among them (Supplementary Fig. 6), which were consistent with those of circadian firing rhythm measured on MED (Fig. 1). These results indicate that GABA is neither necessary for the generation nor the synchronization of the circadian PER2::LUC rhythm in SCN cells.

VGAT or GAD65/67 deficiency induced burst firings of high frequency in the SCN slice, which could be due to a lack of tonic release of GABA in the SCN. In order to test this possibility, we applied GABA in the culture medium for the VGAT^−/− SCN slice to see whether the noisy circadian rhythm with superimposed burst firings was rescued or not. As expected, burst firings in the VGAT^−/− SCN were abolished by GABA application, the effective dose of which was as little as 3 mM (Fig. 3a, c, Supplementary Figs. 7 and 8a, b). Lower doses failed to reduce the burst firing (Supplementary Fig. 7). Importantly, even during GABA application, nearly a half of the electrodes examined (44%, 36/82, n = 5 slices) showed significant circadian rhythms in neuronal activity without burst firings, but the rest of electrodes did not exhibit any neuronal activity (56%, 46/82, n = 5 slices). Pixel level analysis of PER2::LUC images revealed that GABA application did not affect the circadian rhythms or distribution of the acrophases (Fig. 3e, g, i). These results support the conclusion that burst firings are ascribed to a lack of GABA signaling.

To test whether or not suppression of GABA action mimics the burst firings, a cocktail of GABA_A and GABA_B receptor antagonists, bicuculline and sacrofen, was applied to the WT SCN slice. Upon application of GABA antagonists, the WT SCN showed burst firings similar to those detected in the VGAT^−/− SCN (Fig. 3b, d, Supplementary Fig. 8c, d). The firing rate showed two bands which persisted for 4 days. The circadian firing rhythms were still obvious even though superimposed by burst firings. The circadian PER2::LUC rhythm was not affected at all by GABA antagonists (Fig. 3f). The standardized circadian amplitude (from trough to peak divided by peak value) on the cycle immediately before the drug application and that on the second cycle after the application were not significantly different. PER2::LUC images revealed that GABA antagonists did not change the circadian phase distribution on pixel level (Fig. 4h, j). These results support the conclusion that burst firings are due to a lack of GABA signaling.

Genetic deficiency of GABA in the SCN modulated circadian rhythms in neural firing, without affecting the circadian molecular oscillation (Fig. 1, Supplementary Fig. 5). Intracellular Ca²⁺ showed circadian rhythms^33,34 and played an important role in the transmission of input and output signals to and from the cellular circadian oscillation in the SCN^33–36. To investigate the effects of VGAT deficiency on circadian Ca²⁺ rhythms, hSyn-GCaMP6s, a fluorescent intracellular Ca²⁺ sensor was introduced into the cultured SCN slice carrying a PER2::LUC reporter using adeno-associated virus (AAV). The slice was set on an MED probe for simultaneous measurement of fluorescence (Ca²⁺), bioluminescence (PER2::LUC), and spontaneous firing.

The WT SCN slice showed robust circadian PER2::LUC, intracellular Ca²⁺, and spontaneous firing rhythms (Fig. 4a, c, Supplementary Movie 1) as previously reported^33,37. Interestingly, intracellular Ca²⁺ in the VGAT^−/− SCN occasionally showed a spike-like increase which was superimposed on the circadian Ca²⁺ rhythm, while PER2::LUC exhibited robust circadian rhythms similar to those of the WT SCN (Fig. 4d, Supplementary Movie 2). The circadian firing rhythms showed two distinct bands. Time-lapse Ca²⁺ imaging at 3 s intervals for 30 min revealed the calcium spikes in the VGAT^−/− but not in the WT SCN (Fig. 4e, f, Supplementary Movie 3). In order to examine the temporal relationship between the burst firing and the calcium spike, Ca²⁺ imaging was performed with a higher time resolution at 100 ms intervals for 1 min (Fig. 4g, h, Supplementary Movie 4). The calcium spikes occurred synchronously throughout the entire SCN slice for several seconds (Fig. 4d, f, h). The membrane potential also showed burst firings throughout the slice in synchrony with calcium spikes (Fig. 4h). These findings indicate that lack of GABA signaling induces a spike-like increase in intracellular Ca²⁺ throughout the entire SCN slice in synchrony with burst firings.

To understand the role of GABA in the SCN in the output signal to behavior, mice with an SCN-specific deletion of VGAT were generated by injecting AAV9-hSyn-GFP-Cre into the SCN of VGAT^flox/flox mice (SCN-VGAT depleted mice). As control, AAV9-hSyn-hrGFP was injected into the SCN of VGAT^flox/flox mice. Three weeks later, mice were transferred from a light–dark (LD) cycle to constant darkness (DD). Circadian behavioral rhythms were persisted in SCN-VGAT depleted mice but the activity was disrupted under DD (Fig. 5a–e). The amount of behavioral activity in the dark period of the LD cycle was reduced in the SCN-VGAT depleted mice (Fig. 5d) and the subjective day–night difference in the activity was also reduced under DD (Fig. 5f). The cycle-to-cycle variability increased and the Qp value in the Chi-square periodogram decreased in the SCN-VGAT depleted mice (Fig. 5c). However, the free-running period was not different between the SCN-VGAT depleted and control mice. Post-hoc immunostaining revealed that VGAT expression was significantly attenuated in the SCN of SCN-VGAT depleted mice compared to the control SCN by ca. 31% (Fig. 5f, g). These results suggest that a lack of GABA function in the SCN deteriorates circadian behavioral rhythms.

VGAT^−/−70 and GAD65^−/−/67^−/−71 mice on the C57BL/6J background and VGAT^{flox/flox70,72} mice on the 129/SvEv (back-crossed with C57BL/6J mice for at least two generations) were used. They were bred with mPer2^Luc knock-in mice carrying a PER2::LUC fusion reporter³². Mice were bred and reared in the animal quarter at Hokkaido University Graduate School of Medicine, where environmental conditions were controlled (lights-on 6:00–18:00 h; light intensity approximately 100 lx at the bottom of cage: humidity 60 ± 10%). Both male and female mice were used in the present study. Experiments were conducted in compliance with the rules and regulations established by the Animal Care and Use Committee under the ethical permission of the Animal Research Committee of Hokkaido University (Approval No. 08-0279), Nagoya University (Approval No. 18257), and Gunma University (Approval Nos. 07-109, 09-047, and 14-006).

Since VGAT^−/− and GAD65^−/−/67^−/− die postnatally, heterozygote mice (VGAT^+/− or GAD65^+/−/67^+/−) were mated to obtain homozygote fetuses (VGAT^−/− or GAD65^−/−/67^−/−). The fetal SCN were taken under isoflurane anesthesia at the embryonic day 19 or 20 and the fetus genotype was examined later. Coronal brain slices were cut 300 μm with a tissue chopper (Mcllwain) and the SCN was dissected at the mid rostro-caudal region. A paired SCN was cultured on a Millicell-CM culture insert (Millipore Corporation) or multi-electrode array dish (MED, Alpha MED Scientific) pre-coated with collagen (Cellmatrix type 1-C; Nitta Gelatin), as described previously⁵⁵. Briefly, the slice was cultured in air at 36.5 °C with 1.2 ml Dulbecco’s modified Eagle’s medium (Invitrogen) with 0.1 mM D-luciferin K and 5% supplement solution. After 5–22 days, the cultured SCN slices were subjected to the continuous measurement of bioluminescence and spontaneous firing.

Bioluminescence from the whole SCN slice was measured with a luminometer equipped with a PMT (Lumicycle, Actimetrics) as previously described⁵⁵. The intensity of bioluminescence measured for 1 min at 10 min intervals was expressed in relative light units (RLU; counts/min).

Bioluminescence in time-series images was measured using an electron-multiplying (EM) charge-coupled device (CCD) (ImagEM; Hamamatsu Photonics) or CCD (ORCA-ІІ; Hamamatsu Photonics) camera attached to the bottom port of ECRIPSE E1000 (Nikon) and the top port of ECRIPSE TE2000-U or ECLIPSE 80i (Nikon). The images were captured every 60 min with an exposure time of 59 min. The intensity of bioluminescence was expressed in relative light units (RLU; counts/h) and analyzed at pixel level of 2.3 μm × 2.3 μm for ImagEM and 4.3 μm × 4.3 μm for ORCA-ІІ.

Extracellular action potential was continuously recorded for more than 5 days in the SCN slice cultured on a MED probe with 64 planar electrodes (20 × 20 µm; MED-P210A) arranged in an 8 × 8 grid embedded in the center of a 0.7 mm × 0.7 mm area (Alpha MED Scientific). Spike discharges with signal-noise ratio >2.0 were collected by Spike Detector software (Alpha MED Scientific) as previously described⁵⁵. The number of spikes was calculated in 1 min or 100 ms bins for each electrode covered by the SCN slice. The SCN slice on a MED probe was placed in a mini-incubator installed on the stage of an upright microscope (ECRIPSE E1000, Nikon) or inverted microscope (ECRIPSE TE2000-U, Nikon) for simultaneous recording of PER2::LUC bioluminescence and spontaneous firing.

The SCN slice was cultured on a Millicell-CM culture insert. Adeno-associated virus (AAV; serotype rh10) was harbored with GCaMP6s, a genetically encoded calcium sensor, under the control of human synapsin-1 promoter (University of Pennsylvania Gene Therapy Program Vector Core). An aliquot of AVV was inoculated onto the surface of the cultured SCN slice 3–5 days after the preparation. The next day of AAV infection, the SCN slice was transferred onto the MED probe. After culturing for 7–14 days on the MED, simultaneous measurements of bioluminescence, fluorescence, and neuronal activity were started. The MED was placed in a mini-incubator installed on the stage of an upright microscope (ECRIPSE-80i, Nikon) equipped with an EM-CCD camera (ImagEM, Hamamatsu Photonics). Fluorescence and bioluminescence were recorded with an EM-CCD camera at −80 °C every 60 min with an exposure time of 2–3 s and 59 min, respectively. For the measurement of fluorescence, an LED-based light source (Retra Light Engine; Lumencor) was used. Fluorescent calcium sensor (GCaMP6s) was excited at cyan color (475/28 nm) with LED light source, and visualized with 495 nm dichroic mirror and 520/35 nm emission filters (Semlock).

Mice of 8–10 months old were anesthetized with isoflurane (induction dose ~2%, maintenance dose ~1%) and placed in a stereotaxic instrument (Kopf). Next, 120 nl of AAV9-hSyn-GFP-Cre (3.3 × 10¹² GC/ml) or AAV9-hSyn-hrGFP (1.5 × 10¹³ GC/ml) was injected into the SCN of VGAT^flox/flox mice bilaterally (±0.0 mm posterior, ±0.2 mm lateral, −5.7 mm ventral from Bregma) with a glass micropipette and an air pressure injection system (BJ-110, BEX).

Spontaneous movements were measured by a passive infrared sensor which detects changes in animal thermal radiation due to movement⁷³. The amount of movement was recorded every minute with computer software (Clocklab, Actimetrics).

After recording locomotor activity, mice were anesthetized with pentobarbital (50 mg/kg, i.p.) and perfused with saline followed by a 10% formalin solution (Wako). Brains were extracted and post-fixed in the same solution for 24 h at 4 °C, followed by a 30% sucrose solution at 4 °C for at least 2 days. 40 μm coronal sections of the SCN were made using a cryostat (Leica) and stored at 4 °C in PBS. Immunostaining was performed as previously described⁵⁵. The primary antibody was rabbit anti-VGAT (1:1000 dilution, SIGMA) and the secondary was CF 594-conjugated Donkey anti-rabbit antibody (1:1000 dilution, Biotium). Brain sections were counterstained with DAPI and mounted to examine with a fluorescence microscope (BZ-9000, Keyence). Fluorescent images were captured with the same intensity and expose time. The intensity of VGAT was analyzed with the ImageJ software. Every third slice (in total, 4 slices) was collected from each brain for analysis. The border of the SCN was determined with the aid of DAPI-stained images. The average intensity of VGAT staining was calculated for the four slices and expressed as intensity/pixel.

The properties of circadian rhythm in bioluminescence signals were analyzed at pixel level, using a cosine curve fitting method as described previously^74,75. Rayleigh plot was also used for analyses with Oriana4 (Kovach Computing Services). The circadian bioluminescence rhythm recorded with a PMT was evaluated by a Chi-square periodogram (ClockLab, Actimetrics) using records of 7 consecutive days with a significance level of p < 0.01. Circadian amplitude was calculated by subtracting the circadian trough value from the peak in a cycle and was further standardized by dividing this amplitude by the circadian peak value to eliminate possible slice biases as described previously⁵⁵. A correlation coefficient of spontaneous firing was calculated by Matlab (Mathworks). Circadian period of behavioral rhythm was determined by a Chi-square periodogram using records of 28 consecutive days under DD. The onset of behavioral rhythms was determined with the Clocklab software. When the GFP signals were not observed in the SCN with post-hoc immunochemistry, all data of the corresponding animal (one of nine VGAT^flox/flox mice) were excluded from the analyses. Cre-GFP was observed in the nucleus while hrGFP (control) was expressed in the cell body and axon; GFP signals in control mice were often detected outside the SCN as well.

Student’s t-test was used when two independent group means were compared, and Welch’s t-test was used when the variances of two group means were different. Paired t-test was used when two dependent group means were compared. A one-way ANOVA with a post-hoc Tukey–Kramer test was used to analyze a single time series data. A two-way ANOVA with post-hoc t-test was adopted when two independent time series data were compared (Statview or Statcel 3).

Further information on research design is available in thelinked to this article. Nature Research Reporting Summary

The data that support the findings of this study are available from the corresponding author on reasonable request.