Period-independent novel circadian oscillators revealed by timed exercise and palatable meals

We first used a training protocol that combined restricted feeding of both chow and a palatable meal (peanut butter) to establish palatable meal anticipatory activity (PAA) in wild-type mice in the light-dark cycle (see Supplementary Fig. S1; the protocol was slightly modified from Keith et al.22). After training, mice were provided with ad libitum chow and given peanut butter (for 1 h) at either ZT4 or ZT10 (Supplementary Fig. S2). Consistent with Keith et al.22, the robustness of PAA was phase-dependent; all mice fed peanut butter at ZT4 (6 of 6) exhibited PAA, but only 4 out of 7 mice displayed PAA when fed at ZT10. We next tested whether PAA persisted in constant conditions by removing the daily palatable meal. After termination of daily palatable meal access, mice in both the ZT4 (5 of 6) and ZT10 (3 of 7) groups had sustained PAA when provided only with ad libitum chow (Supplementary Figs S2 and S3). These data suggest that there is an endogenous, self-sustained oscillator entrained by palatable food. There was no correlation between the amount of peanut butter consumed and robustness of anticipatory activity (Supplementary Fig. S4).

It is possible that the PAA we observed resulted from entrainment of the FEO by the scheduled restricted feeding of chow during the training portion of the protocol. To address this potential caveat, we asked whether mice would entrain to a daily palatable meal without the training step (i.e. with no chow restricted feeding). For this, mice were maintained with ad libitum chow and peanut butter was provided for 1 h at ZT4 or ZT10. For ZT4, we found that 4 out of 6 mice expressed PAA starting 1–2 h before peanut butter was placed in the cage (Fig. 1 and Supplementary Fig. S5). PAA persisted for 2 days after termination of daily peanut butter feeding (Supplementary Fig. S6). In contrast, none of the mice (0 of 6) fed peanut butter at ZT10 developed PAA. There was no correlation between the amount of peanut butter consumed and robustness of anticipatory activity (Supplementary Fig. S7). These data demonstrate that mice developed PAA to scheduled peanut butter during ad libitum chow and PAA persisted when the palatable meal was removed. Moreover, training the mice with coincident restricted chow and scheduled peanut butter was not necessary for the development of palatable meal anticipatory activity.

In our initial experiment, we could not rule out the possibility that PAA was driven by the SCN. Thus, we next tested whether timed palatable meal could reveal PICO in arrhythmic Period mutant mice, in which the molecular clocks in the SCN and peripheral organs are disabled. In constant conditions (ad libitum chow and constant darkness) Per1/2/3 triple mutant mice did not have circadian rhythms of wheel-running activity, demonstrating that the SCN was disabled in these mice (Fig. 2, Supplementary Fig. S8; they displayed ultradian behavior rhythms typical of circadian mutant mice). We gave periodic 1-h peanut butter access to Per1/2/3 triple mutant mice in constant darkness with ad libitum chow (Fig. 2, Supplementary Fig. S8). When peanut butter was given at a 21-h interval (T21), half of the mice (4 of 8) showed consolidated PAA with a 21-h period (Fig. 2, Supplementary Fig. S8; mice #39, 26, 68 and 02). Another mouse developed a ~17-h period PAA rhythm (Supplementary Fig. S8; #27). The remaining 3 mice did not develop consolidated activity prior to peanut butter feeding (Fig. 2 and Supplementary Fig. S8; #29, 30 and 35); however, the ultradian rhythmicity in these mice was affected by the treatment. There was no correlation between the development of consolidated activity and peanut butter consumption (Supplementary Fig. S9).

Surprisingly, the palatable meal-induced rhythm continued for up to 7 cycles after the termination of peanut butter feeding. The free-running periods of these rhythms were ~21-h (Supplementary Table S1). We then fed the mice peanut butter for 1 h each day on a 24-h cycle (T24). None of the Per1/2/3 triple mutant mice developed PAA to the 24-h palatable meal schedule (Fig. 2, Supplementary Figs S8 and S9). In addition, the T24 peanut butter feeding did not alter the ultradian activity rhythm. Thus, the PAA rhythm has two fundamental properties of a circadian oscillator: it has a range of entrainment (e.g. entrains to T21 but not T24) and it persists in constant conditions. Together, these data suggest that the PAA rhythm is generated by a circadian oscillator. We named this novel circadian oscillator the palatable meal-inducible circadian oscillator (PICO). Moreover, because PICO persists in the absence of functional PERIOD 1, 2, and 3, it is a non-canonical circadian clock.

The methamphetamine sensitive circadian oscillator (MASCO or methamphetamine-induced oscillator, MAO) is a non-SCN circadian oscillator. The rhythmic output of MASCO can only be observed when low dose methamphetamine is chronically administered24,25,26,27. We measured the MASCO period from the same mutant mice used in the timed palatable meal experiments. After the palatable meal-induced behavior rhythms (PICO outputs) were extinguished (evidenced by the presence of ultradian, but no circadian rhythms), mice were administered low-dose methamphetamine (0.005%) in their drinking water. Consistent with our previous study, all Per1/2/3 mutant mice exhibited consolidated activity28. Mice initially had a ~18-h period of wheel-running activity rhythm that persisted for several days and then switched to a stable 22-h period (Fig. 3, Supplementary Fig. S10 and Table S1).

Mice develop anticipatory activity to scheduled palatable meals, and low doses of methamphetamine reveal rhythmic behavior. Both of these stimuli are potent rewarding stimuli that activate the dopaminergic system18,20,29,30. Likewise, voluntary running wheel activity elevates dopamine in the rodent brain31,32. Therefore, we tested if timed access to a running wheel would also elicit a circadian rhythm in Per1/2/3 mutant mice. Six of the Per1/2/3 triple mutant mice used in the palatable meal experiments were kept in constant darkness and housed with a locked running wheel. The wheel was unlocked for 1 h each cycle on a 21-h interval (T21). When the wheel was unlocked, 5 of the 6 mice ran on the wheels (Fig. 4, Supplementary Figs S11 and S12). Unlike the PAA during scheduled peanut butter access, mice did not anticipate the recurrent unlocking of the wheel. Instead, general activity, measured by passive infrared sensors, increased shortly after the wheel was unlocked and remained elevated for 1–2 h after the wheel was relocked. The consolidated wheel-induced activity developed in 3 of the 6 mice (Fig. 4 and Supplementary Fig. S11; mice #68, 29 and 30). The 3 mice that exhibited rhythmic behavior had the most wheel-running revolutions during the 1 h of wheel access (Supplementary Fig. S12).

When the mice were released into constant conditions (the wheel remained locked), the wheel-induced rhythm free-ran for many days with a ~21-h period (). Interestingly, in one of the cages the wheel-locking mechanism failed during constant conditions and the wheel was unlocked for a day. The consolidated 21-h activity rhythm of the mouse in this cage quickly changed to an ultradian rhythm (; #29). Supplementary Table S1 Supplementary Fig. S11

We then gave the same Per1/2/3 mutant mice 1 h of wheel access on a 24-h interval (T24). Two (out of 6) mice developed consolidated activity. When the wheel was permanently locked, their rhythms free-ran with periods of 21-h and 23-h, respectively. Throughout the course of our experiments, only one mouse (#68) exhibited consolidated activity in all three treatments (peanut butter, methamphetamine, running wheel). With the exception of this mouse, the mice that displayed PAA (#39, 26, 27, 02) did not display wheel-induced consolidated rhythms. Similarly, the mice that showed wheel-induced rhythms (#29, 30) did not display PAA.

Wild-type male C57BL/6J mice were obtained from the E. K. Wakeland Mouse Breeding Core (UT Southwestern, Dallas, TX USA) or from our breeding colony at UT Southwestern. Per1/2/3 triple mutant mice were generated by intercrossing Per1^+/−/Per2^−/−/Per3^−/− mice (C57BL/6J N15) for 4 to 6 generations in our breeding colony28. All mice were bred and maintained in 12L:12D illuminated by fluorescent bulbs (Ecolux, 32W). Weaned mice were group-housed in cages without running wheels with ad libitum chow and water (20–23 °C and 18–68% relative humidity). Male wild-type C57BL/6J (7–18 weeks old) and male and female Per1/2/3 triple mutant mice (10–55 weeks old) were used for experiments. All experiments were conducted in accordance with the guidelines of the National Institutes of Health Guide for the Care and Use of Laboratory Animals. All procedures were approved by the Institutional Animal Care and Use Committee at UT Southwestern Medical Center (Protocol#: 2013-0035).

Circadian behavior recordings were conducted in light-tight ventilated boxes (22–23 °C, 19–54% relative humidity). Light (7 μW/cm²/s, 55 lux inside the cage) was generated by green LEDs controlled by Chamber Controller (Phenome Technologies, Inc Chicago, IL USA). Mice were singly housed in plastic cages (length × width x height: 29.5 × 11.5 × 12.0 cm) with running wheels (diameter 11 cm). Wheel revolutions were continuously recorded every minute by the ClockLab system (Actimetrics, Wilmette, IL USA). A passive infrared sensor (product ID 189, Adafruit, NYC, NY USA) located 10 cm above the cage lid was used to monitor general activity in the timed wheel access experiment. Cages and water bottles were changed once every 3 weeks. An infrared viewer (FIND-R-SCOPE Infrared Viewer; FJW Optical Systems, Inc. Palatine, IL USA) was used to perform maintenance and feeding in the dark without exposing mice to visible light.

Restricted feeding of chow was done manually by placing and removing chow (Teklad Global 18% Protein Rodent Diet 2918; Harlan, Madison WI USA) on the bottom of the cage. Peanut butter (Jif® Creamy Peanut Butter, 50% fat, 25% carbs and 22% protein; The J.M. Smucker Company, Orrville, OH USA) was added and removed manually in a 35 mm plastic petri-dish lid on the bottom of cage. The protocols are detailed inand the time of each treatment is indicated on each actogram. Supplementary Fig. S1

During methamphetamine treatment, a bottle containing regular water was replaced with a bottle containing 0.005% methamphetamine (Sigma-Aldrich, Inc. St. Louis, MO USA). The time when methamphetamine water was available is indicated on each actogram.

A large push-pull solenoid (product ID 413, Adafruit, NYC, NY USA) was placed on the cage lid and blocked the rotation of the running wheel. The wheel was unlocked or locked via the Pick and Hold Solenoid Driver Module (PH-ET-01, Optimal Engineering Systems, Inc, Van Nuys, CA, USA) controlled by ClockLab. The pickup voltage (24 V) was set for 1.2 sec and followed by a hold voltage (3 V). The schedule for wheel-unlocking is indicted on each actogram.

Each actogram was generated using 6-min bins and the percentile option in the ClockLab analysis software. Twenty-four hour group average activity profiles were generated from 6-min bin activity files using ClockLab. For each activity profile, either 3 days of ad libitum chow, all days during 1-h peanut butter access, or the first 3 days of ad libitum chow after peanut butter treatment were averaged for individual mice. Then group average profiles were generated by averaging individual activity profiles. PAA was defined as the total number of wheel revolutions during the 2 h prior to peanut butter access. The individual daily average of anticipatory activity during peanut butter feeding was used for PAA analysis. Mean peanut butter consumption was compared between groups by the Wilcoxon-Mann-Whitney unpaired two-tailed test. Spearman’s correlation test was used for correlation analysis. For both statistical tests, alpha level was set to 0.05. The free-running period was calculated by fitting a regression line to the onset of activity detected by ClockLab (default criteria settings). The days used for period analysis (at least 5 cycles) are indicated on each individual actogram in the supplemental figures.