Dual origins of the intracellular circadian calcium rhythm in the suprachiasmatic nucleus

To analyze the functional links between the circadian rhythms of Ca²⁺, PER2 expression, and spontaneous firing in the SCN, we combined three established methods for measuring each circadian rhythm10,14. The intracellular Ca²⁺ level was monitored by expressing genetically encoded Ca²⁺ sensors, GCaMP6s17, which were transfected into the SCN slice using adeno-associated virus (AAV)10,18. PER2 expression was measured using a bioluminescence reporter, and spontaneous firing was recorded by a multielectrode array dish (MED) with 8 × 8 electrodes of 20 × 20 μm.

The SCN slice of PER2::LUC knock-in newborn mice was explanted on a culture membrane and AAV was transfected (Fig. S1a). Several days later, a flipped slice was placed on MED. Simultaneous recording of the three measures was started on the 10^th culture day, which corresponded to postnatal day 10. Circadian rhythms in GCaMP6s fluorescence (Ca²⁺ rhythms) and PER2::LUC bioluminescence (PER2 rhythms) were monitored by a high-sensitivity CCD camera mounted on an upright microscope, and spontaneous firing was monitored using a MED (Fig. S1b). Circadian Ca²⁺, PER2, and firing rhythms were analyzed in a region of interest (ROI) (20 × 20 μm) on each MED electrode (Fig. 1a).

In the wild-type SCN, robust circadian rhythms of Ca²⁺, PER2, and firing were detected in all ROIs examined in single SCN slices (Fig. 1b). The circadian peak of the Ca²⁺ rhythm was phase-advanced, relative to that of the PER2 rhythm, in the entire area of the SCN (Fig. 1c), showing a phase difference of approximately 6 h on average (Fig. 2b). The circadian peak of the firing rhythm was located between that of the Ca²⁺ and PER2 rhythms, and was phase-advanced to the circadian PER2 rhythm by 2.2 h on average (Figs 1b and 2a). In terms of SD, the phase relation between the circadian Ca²⁺ and firing rhythms was unstable compared with that between the Ca²⁺ and PER2 rhythms [4.3 vs. 1.0, p < 0.001 (Fig. 2a,b)]. The circadian Ca²⁺ and firing rhythms were occasionally (approximately 10% of total electrodes) 180 degrees out of phase (Fig. 2a).

The Cry1,2^−/− SCN exhibited circadian rhythms for all three functions, but they were less robust and more variable compared to those of the wild-type (Fig. 1d–f). The circadian periods of all rhythms were significantly shorter and distributed in a wider range in the Cry1,2^−/− SCN than in the wild-type (Fig. S2). Two peaks appeared to exist in the period distributions, with one located at approximately 23 h, similar to that of the wild-type rhythm, and the other, substantially shorter, at approximately 18 h (Fig. S2). Circadian periods were significantly different between wild-type and Cry1,2^−/− SCN [Ca²⁺ : 24.4 ± 1.0 h vs. 20.0 ± 3.2 h, PER2: 24.9 ± 0.8 h vs. 20.6 ± 3.2 h, Firing: 24.1 ± 1.5 h vs. 20.0 ± 3.1 h (70 ROIs in 4 wild-type slices, 52 ROIs in 4 Cry1,2^−/− slices)]. Therefore, circadian phases of the Cry1,2^−/− SCN were described in terms of degrees [1 degree = 360 degree/period (Fig. 2c)]. The phase relation between the Ca²⁺ and PER2 rhythms was changed in the Cry1,2^−/− SCN [47.2 ± 7.8 degrees, n = 45 (Fig. 2c)], showing a significantly smaller phase difference than in the wild-type (87.8 ± 3.0 degrees, n = 72). Similarly, the phase difference between the Ca²⁺ and firing rhythms was smaller in the Cry1,2^−/− SCN (6.1 ± 11.5 degrees, n = 47) than in the wild-type (37.1 ± 8.0 degrees, n = 71). However, the phase relation between the PER2 and firing rhythms was not changed in the Cry1,2^−/− SCN (41.1 ± 8.3 degrees, n = 47) compared to that in the wild-type (50.7 ± 5.5 degrees, n = 71) (Fig. 2c). When the phase difference between the circadian Ca²⁺ and PER2 rhythms was separately calculated in the ventromedial and dorsolateral regions of the SCN, the difference was significantly larger in the dorsolateral (59.3 ± 12.9 degrees, n = 22) than ventromedial region [13.1 ± 14.9 degrees, n = 25 (Fig. 2c)]. These findings indicate that the circadian organization of the SCN, particularly in the ventromedial region, was significantly changed in the Cry1,2^−/− mice.

Circadian Ca²⁺ and PER2 rhythms in the wild-type SCN showed regional specificity in a coronal slice. For regional comparison of the circadian Ca²⁺ and PER2 rhythms, we used pixel-level analysis10,19,20. Note that each pixel has 2.3-μm resolution, which is smaller than the size of cell body. We found that the circadian peak phases of Ca²⁺ and PER2 rhythms were relatively phase-delayed in the narrow area from the dorsal to the medial border of the third ventricle and in the ventral area (Fig. 1c). The regional specificity of the circadian phases was not prominent in the Cry1,2^−/− SCN (Fig. 1f). Pixel-level analysis also revealed that a large area of the dorsolateral SCN in the Cry1,2^−/− mice was lacking circadian Ca²⁺ rhythms but did display circadian PER2 rhythms (Fig. 3a,b top). Whereas no difference was observed in the area with significant circadian PER2 rhythms (Fig. 3b bottom). The deviation in acrophase, an index of network synchronization, was significantly increased in the circadian Ca²⁺ rhythms of the Cry1,2^−/− SCN (Fig. 3c top), whereas there was no difference in the circadian PER2 rhythms (Fig. 3c bottom). On one hand, the amplitude was significantly reduced in the Cry1,2^−/− SCN for both circadian rhythms (Fig. 3d).

Intracellular Ca²⁺ in the SCN has been demonstrated to oscillate in a circadian fashion9,10,11,21, and to show a stable phase relation with circadian firing9 and PER2 rhythm10,11. It has been reported that Ca²⁺ levels are mediated by intracellular stores9,22; blocking Ca²⁺ influx abolishes the rhythmic expression of clock genes in the SCN23. In these previous reports, the circadian peak of intracellular Ca²⁺ levels phase advanced to the circadian firing peak, indicating that the intracellular Ca²⁺ level was not a direct consequence of neuronal activity9. The present study confirmed these findings in wild-type mice by demonstrating a 2.2 h phase advance of the circadian Ca²⁺ rhythm to the firing rhythm. Importantly, the phase relation of the two rhythms was changed in the Cry1,2^−/− SCN and the phase difference was abolished (Fig. 2). Thus, the circadian Ca²⁺ rhythm and the firing rhythm in the SCN are dissociable. On the other hand, the phase relation between circadian Ca²⁺ and PER2 rhythms was also changed in the Cry1,2^−/− SCN. Interestingly, the phase relation was kept unchanged between the circadian PER2 and firing rhythms. These findings indicate that the circadian rhythm of intracellular Ca²⁺ in the wild-type SCN is regulated by at least two circadian oscillations of different origins, one generated by the core loop, represented by the circadian PER2 rhythm, and the other is the rhythmic input from the SCN neural network.

Ca²⁺ in the input pathways is critical for resetting the core loop and, thus, the coherence of the cellular circadian rhythms in the SCN network. Ca²⁺ modifies clock gene expression via the cAMP-responding element (CRE)11,13 and/or CaMKII activation, which activates E-box dependent gene expression through the phosphorylation of CLOCK and the dimerization with BMAL124. In addition, Ca²⁺ is critical for the output pathways as it triggers release of transmitter and modulates action potentials via Ca²⁺ activated channels, including BK channels25. Blocking neuronal firing by TTX reduces the amplitude of circadian Ca²⁺ rhythms by only approximately 30%10. Therefore, the remaining circadian rhythm could be under the control of an endogenous oscillation(s). Thus, two factors of different origins may contribute to the circadian Ca²⁺ rhythm.

Bmal1 was reported to be critical for circadian Ca²⁺ rhythms. Suppression of Bmal1 expression results in amplitude reduction and period elongation of Ca²⁺ rhythms in single SCN cells26. Bmal1 expression is regulated by an auto-feedback loop2 which is interlocked with the core loop2,3, and is dissociable from the core loop represented by circadian PER2 rhythms27. The change in the phase relationship between circadian PER2 and Ca²⁺ rhythms could be interpreted as the result of the recoupling of two intracellular oscillations in the absence of circadian input from the SCN neural network. Alternatively, the two factors may contribute to the shape formation of circadian Ca²⁺ rhythms. The circadian Ca²⁺ rhythm consists of an endogenous component and an exogenous masking component from the SCN neural network. CRY double deficiency and TTX may block the latter component and deform the circadian Ca²⁺ rhythm, which apparently changes the phase relationship between them (Fig. 4).

The change in the phase relationship among three cellular circadian rhythms in the Cry1,2^−/− SCN could be explained by regional specificity of the circadian network in the SCN. Circadian Ca²⁺ rhythms on the cellular level were ubiquitous in the wild-type SCN with a phase advance in the dorsal region of the SCN. Previously, TTX treatment in cultured SCN slices was shown to dissociate the circadian Ca²⁺ rhythms in the dorsal and ventral regions of the SCN10. In this case, the circadian Ca²⁺ rhythms in the ventral region were phase-delayed, showing a larger phase difference between the two regions. If the phase delay shifts by TTX were specific to the circadian Ca²⁺ rhythms, the phase relation with other circadian rhythms would also be changed.

On the other hand, the Cry1,2^−/− SCN lacks the circadian Ca²⁺ rhythm in the dorsolateral region, despite the expression of circadian PER2 rhythms (Figs 1f and 3a), although the amplitude was much smaller than wild-type (Fig. 3d). Since this region corresponds to the area of AVP-containing neurons, the Ca²⁺ rhythms may be critical for proper functioning of transmitter release. Recently, dysfunction of AVP signaling was demonstrated to be responsible for the loss of coherent circadian rhythms in the Cry1,2^−/− SCN15,28. An abolishment of circadian rhythms by AVP neuron-specific Bmal1 knockout in this region is known to deteriorate the behavior rhythms by splitting the activity onset and offset7. These findings indicate that AVP neurons in the dorsal regions of the Cry1,2^−/− SCN lose their mutual couplings for coherent circadian rhythms, as well as their communication to the ventral region of the SCN. We recently reported that AVP content is extremely reduced in Cry1,2^−/− SCN both in neonates and adults, and found that AVP and VIP signaling is differentially involved in the network of cellular oscillations depending on the developmental stages15. A lack of circadian signals from the dorsal to ventral region may change the phase relation between circadian PER2 and Ca²⁺ rhythms in these neurons.

In conclusion, the Cry1,2^−/− SCN shows an altered phase relation of circadian Ca²⁺ and PER2 rhythms in the ventral region, where the neural communication from AVP-containing neurons in the dorsal regions was substantially attenuated. The altered phase relation in the Cry1,2^−/− SCN was well explained in terms of dual signals which contribute the phase and/or the shape of intracellular Ca²⁺ rhythms and have different origins: a signal from the core molecular feedback loop and a signal from the neural network in which the cell is involved.

All experimental protocols and methods performed were approved by the animal experiment ethics committee at Hokkaido University (approval no. 15–0153), and were in accordance with Hokkaido Univesity guidelines for the care and use of laboratory animals.

C57BL/6 J background mice were used for control experiments. Cry1,2^−/− mice, originally obtained from Tohoku University, were backcrossed with C57BL/6 J mice for more than 13 generations in our animal facility, and were bred with PER2::LUC mice carrying a PER2 fusion luciferase reporter. Mice were born and bred in our animal quarters under controlled environmental conditions (temperature: 22 ± 2 °C, humidity: 60 ± 5%, 12 h light/12 h dark, with lights on from 0600 to 1800 h). Light intensity was around 100 lx at the cage surface. The mice were fed commercial chow and tap water ad libitum.

Decapitation was performed at the middle of the light phase. The brains of neonate mice (1-day old, both male and female) were rapidly removed and dipped in ice-cold balanced salt solution comprising (in mM) 87 NaCl, 2.5 KCl, 7 MgCl₂, 0.5 CaCl₂, 1.25 NaH₂PO₄, 25 NaHCO₃, 25 glucose, 10 HEPES, and 75 sucrose. A 200 μm coronal brain slice containing the mid rostrocaudal region of the SCN was carefully prepared using a vibratome (VT 1200; Leica). We confirmed previously that the dorsolateral and ventromedial regions mostly contained AVP and VIP neurons, respectively10. The bilateral SCNs were dissected from the slice using a surgical knife and explanted onto a culture membrane (Millicell CM; pore size, 0.4 μm; Millipore) in a 35-mm Petri dish containing 1.0 mL of DMEM (Invitrogen) and 5% FBS (Sigma-Aldrich). Before the recordings, membranes with cultured SCN slices were cut out, flipped over, and transferred to glass-base dishes (35 mm, No.1-S; AGC Techno Glass) or MED dishes (MED-P210A) that were collagen coated (Cellmatrix type 1-c; Nitta Gelatin) and supplemented with 180–250 μL DMEM and 5% supplement. MED dishes were sealed with O₂-permeable filters (membrane kit, High Sens; YSI) using silicone grease compounds (SH111; Dow Corning Toray).

Aliquots of the adeno-associated virus (AAV, serotype 2/1) (1 μL) harboring GCaMP6s under the control of the human synapsin-1 promoter (produced by the University of Pennsylvania Gene Therapy Program Vector Core: https://www.med.upenn.edu/gtp/vectorcore/↗) were inoculated onto the surface of the SCN cultures 1 d after the preparation of the SCN slices. Infected slices were further cultured for at least 8 d before imaging. The titer of the GCaMP6s vector was 1.89 × 10¹³ genome copies/mL. We validated the specificity and transduction efficiency of AAV encoding hSyn-driven calcium indicator10. The calcium probes were expressed in SCN neurons, but not in glial cells, in the entire regions including the dorsolateral and ventromedial SCN, and the transduction efficiency was nearly 100%.

The SCN slice was cultured in 100% air on a MED dish with 8 × 8 electrodes of 20 × 20 μm. Before the recording, the MED dish was placed in a mini-incubator installed on the stage of the microscope (ECRIPSE 80i, Nikon). GCaMP6s fluorescence, PER2::LUC bioluminescence, and spontaneous firing were simultaneously recorded while culturing with the culture media supplemented with 0.5 mM D-luciferin K. Fluorescence and bioluminescence were recorded with a cooled CCD camera at −80 °C (512 × 512 pixels, 2.3 × 2.3 μm in pixel size) (ImagEM, Hamamatsu Photonics) every 60 min with an exposure time of 2–3 s for fluorescence and 3540 s (59 min) for bioluminescence, respectively, for 3–6 days. As exposure time was triple digits different, the spectrum overlap was negligible. GCaMP6s was excited at cyan color (475/28 nm) with an LED light source (Light Engine; Lumencor Inc) and the fluorescence was visualized with a 495-nm dichroic mirror and 520/35-nm emission filters (Semlock). Fluorescence and bioluminescence signals were analyzed in a region of interest (ROI) on each MED electrode (20 × 20 μm). The signal intensity was expressed as the average intensity of pixels involved in a ROI. Multi-unit spontaneous firings were recorded using an MED 64 system (Alpha MED Scientific). Spike discharges with signal–noise ratio >2.0 were collected by Spike Detector software (Alpha MED Scientific) as previously described14. The number of spikes per min was calculated for each electrode. Peak phases of the rhythms were estimated by the midpoint of rising and falling limbs of detrended circadian rhythms that intersected the x-axis.

Statistical analyses were performed using Prism GraphPad (GraphPad Software). The group mean is presented as the mean ± sem. The 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. Ventromedial and dorsolateral regions were selected based on the rhythmic area in Fig. 3.

For quantification of the rhythms in the SCN network, we used a custom-made program for creation of acrophase maps, as described previously7,10,19,20. Briefly, fluorescence and bioluminescence images were smoothed with the median filter (one pixel), and converted to eight-bit intensity. The time series of the images in each pixel, {Yj (ti); ti = 1,2, …, N (h)}, was fitted to a cosine curve, with yj(t) = yj (t; Mj, Aj, Cj, Tj) = Mj + Aj * cosine (2pi (t − Cj)/Tj), using a least-squares regression method, where j, yj (t) is the signal intensity at time t (h), Mj is the mesor, Aj is the amplitude, Cj is the acrophase, and Tj is the period of the images. The goodness of fit was statistically evaluated by Percent Rhythm accounted for by the fitted cosine wave (Pearson product–moment correlation analysis) at a significance level of p < 0.001. The percentage of the rhythmic areas in the SCN was estimated as the ratio of statistically rhythmic pixels (by Percent Rhythm, p < 0.01) to total pixels blighter than background level (>mean + 5 SD).