Circadian and ultradian rhythms of clock gene expression in the suprachiasmatic nucleus of freely moving mice

In the present study, several technical improvements were made to enable bioluminescence recordings of the SCN in freely moving mice (Fig. 1, Supplementary Figure S1). First, cooling the photomultiplier tube (PMT) in the photon counting devise (In vivo Kronos, Atto) to 10 °C reduced dark counts and consequently increased the signal to noise ratio. Background noise was reduced from 834.2 ± 91.0 to 36.2 ± 6.5 counts/min (mean ± SD) and became stable (Supplementary Figure S1). In addition, a convex lens (Plano-Convex lens, #45-081, Edmond) between the optical fiber and PMT optimized the detection area and increased signal strength by ca 40%. Second, the substrate luciferin was delivered to the SCN via an osmotic pump implanted in the body, instead of perfusing it through the lateral ventricle7. Third, to ensure that the mice could move freely, a long (3 m) plastic optical fiber was employed in order to reduce fiber torque. These improvements all together enabled us to monitor bioluminescence in the SCN of freely moving mice up to 3 weeks (Supplementary Figure S1e).

Circadian rhythms in clock gene expression were observed continuously for up to 3 weeks in the SCN of freely moving mice, carrying bioluminescence reporters for the expression of Per1 (Per1-luc), Bmal1 (Bmal1-ELuc) or for the protein product of Per2 (PER2::LUC) (Fig. 1, Fig. 2, Supplementary Figure S1, Figure S2). Histological examination revealed that the tip of an implanted optical fiber was located at the dorsal border of the SCN in most cases and was found within the dorsal part of the SCN in the rest of cases (Supplementary Figure S1c).

The bioluminescence detected by the fiber originated mainly from the dorsal area of the SCN, because only 30% (Bmal1-ELuc) to 50% (Per1-luc, PER2::LUC) of bioluminescence emitted from the ventral area of the SCN was estimated to reach the optical fiber8. The difference between the clock gene reporters is due to the wavelength of lights emitted by the reporters (561 nm vs. 630 nm). The spatial distribution profiles of bioluminescence intensity were obtained using data from cultured coronal slices of the SCN prepared from the same clock gene reporter mice (Supplementary Figure S3). The intensity of bioluminescence was brightest at the medial part of the SCN, whereas the intensity at the border of SCN was less than 10% of the brightest area in the SCN and decreased rapidly with distance from the SCN border.

Significant circadian rhythms were detected in clock gene expressions by Chi square periodogram (p < 0.05) in all mice whose SCN emitted measureable bioluminescence (Fig. 1a). Simultaneously measured behavior activity by a passive infrared sensor also showed significant circadian rhythms (Fig. 1c). Estimates of the circadian period in gene expression rhythm were slightly different among the three strains of mice but not between gene expression rhythms and their respective behavior rhythms (Per1-luc mice, 23.7 ± 0.3 h for bioluminescence and 23.9 ± 0.2 h for behavior activity, n = 8; PER2::LUC mice, 24.3 ± 0.3 h and 24.0 ± 0.3 h, n = 3; Bmal1-ELuc mice, 23.9 ± 0.1 h and 23.9 ± 0.2 h, n = 7, by Chi square periodogram).

The circadian rhythms in clock gene expression measured ex vivo were significantly phase-advanced by approximately 3 hours in all 3 clock gene reporters. Nonetheless, the phase-relationship among the Per1, PER2 and Bmal1 is consistent with results from in situ hybridization9,10 and in ex vivo studies11,12,13.

Unexpectedly, the in vivo circadian rhythms in gene expression were superimposed by episodic bursts of bioluminescence in all three reporters (Fig. 2, Supplementary Figure S2). Chi square periodogram revealed significant ultradian periods in all mice. Most mice showed more than one periodicity of different strength in the ultradian range. The major period was 3.0 ± 0.5 h (n = 8) for Per1-luc, 2.9 ± 0.3 h (n = 3) for PER2::LUC, and 3.2 ± 0.8 h (n = 7) for Bmal1-ELuc, respectively. Wavelet analysis confirmed these ultradian periodicities (Fig. 2, Supplementary Figure S2). Ultradian rhythms were also detected in behavior activity by Chi square periodogram in all mice examined (4.7 ± 0.2 h, n = 8, Per1-luc; 3.7 ± 0.9 h, n = 3, PER2::LUC; 4.6 ± 1.1 h, n = 7, Bmal1-ELuc), and the periods were slightly but significantly longer than those observed in clock genes (p < 0.01, Per1-luc; p < 0.05, PER2::LUC; p < 0.05, Bmal1-ELuc, one-way repeated measure ANOVA).

The frequency and size of episodic bursts were examined to characterize episodic bursts of clock gene reporters. Episodic bursts occurred either sporadically or sequentially. In the latter case, bursts overlapped, leading a stepwise increase in bioluminescence that lasted up to several hours. Since most sporadic episodes persisted for approximately one hour, episodic bursts of less than 2 h duration were used for further analyses. A total of 691 episodic bursts were detected during the first 5 circadian cycles under DD for all mice (Per1-luc, n = 342; PER2::LUC, n = 106; Bmal1-ELuc, n = 243). The mean number of episodic bursts per mouse was 42.8 ± 10.8/5 cycles for Per1-luc, 35.3 ± 20.7/5 cycles for PER2::LUC and 34.7 ± 8.0/5 cycles for Bmal1-ELuc. The mean duration of episodic burst was 70.5 ± 7.5 min for Per1-luc, 85.0 ± 5.0 min for PER2::LUC and 75.5 ± 9.5 min for Bmal1-ELuc, respectively. The frequency and the size of episodic burst depended on the circadian phase of respective reporter genes (Fig. 3). The size of episode was expressed as an area under the curve (AUC). AUC was calculated by summing bioluminescence above the baseline. Both the number of episode and AUC of Per1-luc were largest in the late subjective day (CT6-CT12), those of PER2::LUC were in the early subjective night (CT12-CT18), and of Bmal1-ELuc in the early subjective day (CT0-CT6), respectively (p < 0.05, one-way repeated measure ANOVA), where CT12 was defined as the onset of behavior activity in terms of circadian time (CT) which was a time unit divided by a circadian period. The mean interval from the initial trough of a burst to the episodic peak was 25 min, irrespective of reporter gene (Fig. 4).

An episodic burst of bioluminescence was often accompanied by a bout of behavior activity (activity bout). To know whether there is causality between episodic bursts of bioluminescence and activity bouts or not, we examined their temporal as well as stoichiometric relationships. Activity bouts were detected mainly in the subjective night (CT12-CT24) (one-way repeated measure ANOVA, p < 0.01), irrespective of reporter gene (Fig. 3). Therefore, the peak times of bioluminescence bursts and activity bouts were antiphasic for Bmal1-ELuc and substantially out of phase for Per1-luc. The desynchronized phase relationship between episodic bursts and activity bouts was also revealed by wavelet analysis (Fig. 2, Supplementary Figure S2). The intensity of ultradian rhythms fluctuated in a circadian fashion in both measures, but the peak intensity was out of phase in Per1-luc and Bmal1-ELuc. Thus, there was no temporal correlation in the frequency or in the AUC between episodic bursts of bioluminescence and activity bouts at least in Per1-luc and Bmal1-ELuc.

The mean sizes of activity bout and accompanying bioluminescence or of episodic burst and accompanying behavior activity were compared to examine stoichiometric relations between the two measures (Fig. 4). In the activity bout reference, the mean activity bout peaked after 10 min at 17.1 to 20.2 counts/5 min. Accompanying bioluminescence started to increase almost simultaneously with an activity bout and peaked in 25 min (Per1, Bmal1) or in 35 min (PER2) on average. However, the peak bioluminescence level was variable (52.3–112.0 counts/5 min). By contrast, in the bioluminescence burst reference, the mean episodic burst peaked in 25 min in all gene reporters and the mean peak level was 138–233 counts/5 min, significantly larger than those obtained with the activity bout reference by 2–3 times, irrespective of the clock genes (p < 0.01, one-way repeated measure ANOVA). On the other hand, the mean peak levels of accompanying behavior activity (3–6 counts/5 min) was approximately one thirds of the peak levels obtained with the activity bout reference, and significantly lower than those, irrespective of the clock genes (p < 0.01, one-way repeated measure ANOVA). The behavior activity did not increase significantly in parallel with the episodic burst of PER2::LUC. Correlation analysis failed to reveal any significant relationship with respect to AUC and the peak level between activity bouts and accompanying bioluminescence or between episodic bursts of bioluminescence and accompanying behavior activity in all clock gene reporters. Thus, there was no evidence for a stoichiometric correlation between the episodic burst and the activity bout.

Mice carrying a Per1-luc11 (hetero or homozygous), PER2::LUC12 (homozygous) or Bmal1-ELuc24 (heterozygous) reporter gene on a C57BL/6j background were used. PER2::LUC mice were originally produced in the Northwestern University. The animals had been backcrossed with C57BL/6j background mice (supplied from CREA Japan, Inc.) more than 10 generations when arrived at our laboratory. Per1-luc and Bmal1-ELuc mice were produced in YS New Technology Institute (Tochigi, Japan) and Institute of Advanced Industrial Science and Technology (Tsukuba, Japan), respectively. They were generated from fertilized eggs of C57BL/6j background; though created on a C57BL/6j background, we back-crossed these mice with C57BL/6j background mice as well for more than 7 generations. The mice were reared in our animal quarters where environmental conditions were controlled (light-dark cycle (LD); lights-on 6:00–18:00; light intensity approximately 100 lx at the cage bottom; humidity 60 ± 10%). Food and water were available ad libitum. Sterilized wooden chips were placed on the floor of an animal cage. Experiments were conducted in compliance with the rules and regulations established by the Animal Care and Use Committee of Hokkaido University under the ethical permission of the Animal Research Committee of Hokkaido University (Approval No. 08-0279).

Male and female mice were used in the present study at 2–6 months old (3.9 ± 1.4 months old, mean and SD). Mice were individually housed in polycarbonate cages (115 mm wide, 215 mm long, 300 mm high), and placed in a light-tight and air-conditioned box (40 × 50 × 50 cm). Spontaneous movements were measured by a passive infrared sensor which detects a change in thermal radiation from an animal due to behavior activity25. The amount of behavior activity was automatically recorded every min by computer (The Chronobiology Kit; Stanford Software System).

Surgical operation was performed under isoflurane anesthesia. A handmade guide cannula (inside diameter 1.12 mm, outer diameter 1.48 mm) was stereotaxically inserted into the brain (0.2 mm posterior and 0.2 mm lateral to the bregma, and 3.0 mm from the surface of the skull) and fixed by a dental resin. After a recovery period of more than 4 days, a polymethyl methacrylate optical fiber was inserted (fiber diameter, 0.5 mm; surface cladding, 0.25 mm thick) into the guide cannula aimed at the SCN (5.8 mm from the surface of the skull) and fixed with dental resin (Supplementary Figure S1b, c). More than 4 days later, an osmotic pump (flow speed, 0.5 μl/hr; pump volume, 200 μl; 2002, Alzet) containing D-luciferin K (100 mM) in saline was implanted in the peritoneal cavity for recording of Per1-luc and Bmal1-ELuc. For PER2::LUC recording, an osmotic pump (flow speed, 0.11 μl/hr; pump volume, 100 μl; 1004, Alzet) was implanted subcutaneously at the shoulders to deliver D-luciferin Na (50 mM) into the lateral ventricle. D-luciferin Na was dissolved in artificial cerebrospinal fluid (ACSF). A catheter connected to an osmotic pump was passed under the skin to the top of the head and inserted in to the lateral ventricle through a guide tube implanted beforehand at 0.6 mm posterior to the bregma, 1.4 mm to the lateral from the midline and 2.2 mm in depth.

Three to five days after the last surgery, bioluminescence recording began in constant darkness (DD). At the end of bioluminescence measurement, mice were transcardially perfused with physiological saline solution followed by 4% paraformaldehyde in 0.1 M phosphate buffer (PB) under ether anesthesia. Brains were cryoprotected in 20% sucrose in 0.1 M PB. Serial coronal sections (30 μm) were made by Cryostat (Leica) and stained with cresyl violet to identify the location of the optical fiber tip.

In vivo measurement of bioluminescence was performed under DD to exclude the penetration of environmental lights. The day of release into DD was defined as Day 0. Bioluminescence emitted from the SCN was counted every minute via an optical fiber which was connected to a photon counting device (In vivo Kronos, Atto) equipped with a PMT (Hamamatsu Photonics). When bioluminescence was not detected in the first several days of measurement, the recording was stopped.

In the ex vivo experiment, mice reared in LD were euthanized by cervical dislocation and decapitated. The brain was rapidly removed and 300 μm coronal slices including the SCN was made by a microslicer (D.S.K: DTK-1000; Dosaka EM) in cooled Hanks’ Balanced salt solution (SIGMA). A trimmed slice of bilateral SCN was placed on a culture membrane (Milicell-CM, Millipore Corporations) in a 35-mm Petri dish. 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, the composition of which was described previously26. Bioluminescence from the SCN slices was measured for 1 min at 10-min intervals or 1min intervals with a luminometer (Lumicycle, Actimetrics or Kronos, Atto). The day of slice preparation was defined as day 0.

To examine the spatial distribution of bioluminescence intensity in the SCN and peri-SCN regions, image records of the cultured SCN were obtained with a CCD camera and analyzed on a pixel level. Coronal SCN tissue slices were prepared from the mice carrying the same reporter genes used in the in vivo experiment. The analyses of SCN images were done at the circadian peak phase.

Bioluminescence and behavior values were integrated in 5 min bins and the time series record of both measures were analyzed from Day 1 (1^st circadian cycle) to Day 5 (5^th circadian cycle) under DD. The onset of the 1^st circadian cycle was assigned to 24 h in local time (CT18) of Day 0. Significance of periodicity in the ultradian as well as circadian range was evaluated by Chi square periodogram (p < 0.05). Chi square periodogram for ultradian period was conducted after subtracting the circadian fluctuation from the original data. The circadian fluctuation was calculated by a 4 h moving average method. In most mice examined, significant peaks were detected at two to three ultradian ranges which seem more or less common to them. We calculated the mean value at the period ranges with the largest Qp value and regarded it as the major ultradian period. Wavelet analysis was applied to evaluate ultradian rhythms and periodic changes in the intensity of ultradian rhythms in a range from 0 to 12 hours27, using MatLab software. The bandwidth parameter was set to 3 and the center frequency to 1. Evaluation of rhythmicity and determination of periods were done by Chi square periodogram. Circadian periods of behavior rhythm were calculated by Chi square periodogram and from the slope of a regression line fitted to activity onsets.

Peak phases of circadian clock gene rhythms were calculated on Day 1 in DD and in culture by a geometric method described previously28. The activity onset was determined by ClockLab software (Actimetrics) for the calculation of circadian time (CT), where the activity onset was defined as CT12. When the results of in vivo and ex vivo experiments were compared, the local time was used as a common time reference.

Episodic bursts of bioluminescence were analyzed for their periodicity, dependency of occurrence (frequency) and of the size (amplitude and duration) on the circadian rhythm, and stoichiometric as well as temporal relation to activity bouts that often accompanied episodic bursts. An episodic burst of bioluminescence was defined as a bioluminescence increase whose amplitude was larger than 2 times of the mean level in the 5 day recording for at least 30 min. An episodic increase of behavior activity (activity bout) was defined as an abrupt increase in behavior activity from 0 counts, lasting at least 30 min. The magnitude of activity bout should be more than twice as large as the mean 30 min activity in the whole series of recording. The duration of burst or bout was defined as a length from the initial trough (bioluminescence) or the bin of 0 counts (behavior) to the end trough where the value lowered below the initial trough in a burst or became 0 counts in a bout. The peak latency of the episode was defined as the interval from the initial trough to the peak of the episodic burst or the activity bout. The size of the episodic burst was expressed by the area under the curve (AUC) which was calculated by summing bioluminescence above the baseline connecting the initial and end trough.

Stoichiometric as well as temporal analyses were carried out to examine potential causality between the episodic burst of bioluminescence and the activity bout. The mean number of occurrence as well as AUC of episodic bursts and activity bouts were assigned to 4 circadian phases (CT0-6, CT6-12, CT12-18, CT18-24) using the circadian peak as the phase reference. The analysis was performed for each clock gene reporter. Stoichiometric analysis was carried out for pairs of episodic bursts and activity bouts. In these analyses, only sporadic episodes or activity bouts which subsided within 60 min were used to exclude complexity due to overlapping episodes or bouts. Since an episodic burst was not necessarily accompanied by an activity bout, two different methods were introduced for analysis. First, bioluminescence data were averaged in reference to an activity bout (activity bout reference). Second, activity data were averaged in reference to an episodic burst (bioluminescence burst reference). Bioluminescence in individuals was calculated by subtracting the baseline values. The individual mean values were calculated first and then the mean of individual means were obtained for each reporter. The total number of activity bouts examined was 140 for Per1-luc, 51 for PER2::LUC and 113 for Bmal1-ELuc, and the total number of episodic bursts was 175 for Per1-luc, 35 for PER2::LUC and 145 for Bmal1-ELuc. Correlation analysis was performed with respect to AUC and the peak level for the pairs of activity bout and accompanying bioluminescence and for the pairs of episodic burst of bioluminescence and accompanying behavior activity.

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