Altered Patterns of Sleep and Behavioral Adaptability in NPAS2-Deficient Mice

See allHide authors and affiliations

Science  18 Jul 2003:
Vol. 301, Issue 5631, pp. 379-383
DOI: 10.1126/science.1082795


Animal behavior is synchronized to the 24-hour light:dark (LD) cycle by regulatory programs that produce circadian fluctuations in gene expression throughout the body. In mammals, the transcription factor CLOCK controls circadian oscillation in the suprachiasmatic nucleus of the brain; its paralog, neuronal PAS domain protein 2 (NPAS2), performs a similar function in other forebrain sites. To investigate the role of NPAS2 in behavioral manifestations of circadian rhythm, we studied locomotor activity, sleep patterns, and adaptability to both light- and restricted food–driven entrainment in NPAS2-deficient mice. Our results indicate that NPAS2 plays a substantive role in maintaining circadian behaviors in normal LD and feeding conditions and that NPAS2 is critical for adaptability to food restriction.

Many and diverse organisms have dedicated regulatory programs that switch genes on and off as a function of the light:dark (LD) cycle (1, 2). Remarkably, for reasons not well understood, these programs continue to operate at or near a 24-hour cycle in the absence of light (3, 4).

In mammals, a specialized region of the brain termed the suprachiasmatic nucleus (SCN) functions as the master pacemaker of circadian rhythm (5, 6). The electrical and metabolic activities of neurons in the SCN fluctuate rhythmically over the LD cycle (710). In both diurnal and nocturnal animals, the SCN is “on” during the day and “off” at night. A heterodimeric, activating transcription factor composed of the CLOCK and BMAL1 polypeptides turns SCN gene expression on in the day (1113). The CLOCK: BMAL1 transcription factor directly induces the expression of genes encoding negative regulators, including the cryptochrome (Cry) proteins, as well as modulatory proteins exemplified by the products of the Period (Per) genes (1218). The circadian cycle of the SCN is synchronized with the LD cycle by neuronal input from the retina (19, 20).

Organismal rhythms in mammals are now understood to be executed by the interplay between a master SCN clock and molecular clocks expressed in other parts of the body (2124). One of these peripheral clocks is specified by a paralog of CLOCK, designated neuronal PAS domain protein 2 (NPAS2). CLOCK and NPAS2 are similar in amino acid sequence, share BMAL1 as an obligate heterodimeric partner, bind to the same DNA recognition element, are suppressed by the cryptochromes, and commonly depend on favorably reducing ratios of nicotinamide adenine dinucleotide (NAD) cofactors (18, 2527). In contrast, the Clock and Npas2 genes are expressed in regions of the body that mostly do not overlap. The Clock gene is expressed in the SCN (28, 29); Npas2 is not. Instead, Npas2 is expressed in forebrain cortices that process sensory information, as well as regions of the basal ganglia and limbic system involved in the control of anxiety and emotion (30, 31).

To investigate whether CLOCK and NPAS2 are functionally redundant or serve distint roles in circadian behavior, we studied NPAS2-deficient (–/–) mice generated by targeted disruption (30) and backcrossed to the C57BL/6J background over nine generations to create a congenic strain comparable to that used in analysis of CLOCK mice. We used standard assays of wheel running activity to examine circadian periodicity (32). When exposed to a 12-hour:12-hour LD cycle, wild-type mice run at night and rest in the daytime. When released into constant darkness (DD), mice continue to run during the period of subjective darkness and to rest during the period of subjective daylight. Each day, however, mice tend to initiate wheel running slightly earlier, revealing an intrinsic period (τ) a bit shorter than 24 hours. NPAS2 heterozygous (+/–) and NPAS2 (–/–) mice displayed rhythmic locomotor activity largely comparable to that of wild-type (+/+) littermates under normal LD conditions (fig. S1). Rhythmic oscillation of locomotor activity persisted under DD conditions, with wild-type and (+/–) mice both displaying periods of 23.7 hours. NPAS2 (–/–) mice reproducibly displayed a shorter intrinsic period of 23.5 hours that persisted over at least 48 days under conditions of constant darkness. In sharp contrast, CLOCK (–/–) mice studied under similar conditions show a lengthening in period, followed by a rapid deterioration of circadian rhythmicity (28, 33).

Closer inspection of actograms of wild-type and NPAS2 (–/–) mice revealed a substantive difference in wheel running during the active, nocturnal period of the circadian cycle. In contrast to wild-type mice, which displayed a prominent break during the middle of the night, NPAS2 (–/–) mice postponed or completely skipped this break (Fig. 1A). To quantify this difference in activity, we generated aggregate activity profiles and calculated half-maximal nocturnal activity (Fig. 1B). Both wild-type and NPAS2 (–/–) mice exhibited high levels of running activity commencing at lights off. However, at 6.6 hours into the dark period [Zeitgeber time (ZT) 18.6], activity in wild-type and NPAS2 (–/–) animals began to diverge. Wheel running activity in wild-type mice began a steep decline, reaching a half-maximal level at 7.8 hours into the dark period (ZT 19.8). During this same period, wheel running activity in NPAS2 (–/–) mice was sustained at high levels. Not until 9.8 hours into the dark period, or a full 2 hours after activity had subsided to half-maximal in wild-type animals, did wheel running activity in NPAS2 (–/–) decline to half-maximal. The difference in nocturnal wheel running behavior between the two genotypes was highly significant [repeated measures analysis of variance (ANOVA) for factor “genotype” F(1,54) = 12.40, P = 0.0009]. During the last hour of darkness, wheel running activity rebounded to a small, secondary peak more prominent in wild-type than in NPAS2 (–/–) mice. Overall, the amount of nocturnal wheel running activity was 19.4% higher in Npas2 homozygotes than wild-type littermates.

Fig. 1.

(A) Representative double-plotted wheel running actograms from a wild-type mouse (left) and an NPAS2 (–/–) mouse (right) (32). In this and all subsequent figures, the alternating light and dark bar indicates the LD cycle. Zeitgeber time (ZT) is plotted on the x axis, and the number of days of recording is plotted on the y axis. (B) For each animal, wheel activity was accumulated in 6-min bins for 8 days and then averaged to yield one 24-hour profile. Data from NPAS2 (–/–) and wild-type mice were then averaged and plotted with respect to ZT. Vertical bars indicate SEM. Longer ticks on the x axis at ZT 19.8 and 21.8 denote the time at which half-maximal nocturnal activity was achieved for wild-type and NPAS2 (–/–) mice, respectively. Half-maximal values were obtained by locating the 6-min bin with the lowest amount of activity, subtracting it from the bin with the highest amount of activity, and dividing by 2. Wheel running–activity profiles were significantly dependent on genotype [F(1,54) = 12.40, P = 0.0009, power 0.950]. A significant genotype × time of night interaction was also present [F(119,6246) = 6.41, P < 0.0001, power 1.00]. From ZT 18.6 through ZT 22.4 (horizontal line), the number of wheel revolutions was significantly higher in NPAS2 (–/–) mice than in wild-type animals (P < 0.05).

This nocturnal break in circadian wheel running activity is detectable in the actograms of many previous studies of C57BL/6J mice (3437). Moreover, recordings of open-field activity in C57BL/6J mice have clearly demonstrated two distinct and equally intense peaks of activity: one coinciding with the transition from lights on to lights off and the other coinciding with the transition from lights off to lights on (38, 39). These data suggest that C57BL/6J mice are behaviorally programmed to take a brief, 2- to 3-hour break in circadian locomotor activity beginning midway through the active period and that this behavior requires NPAS2.

The genotype-specific differences in nocturnal locomotor activity described here do not necessarily translate to differences in sleep. Thus, we next implanted wild-type and NPAS2 (–/–) mice with electroencephalogram (EEG) and electromyogram (EMG) electrodes to monitor their sleep patterns as a function of the LD cycle. Recordings of cortical and myotonic activity were acquired continuously over 48 hours to assess vigilance state. Although implanted mice were free to move about their cages, they were not provided access to running wheels. As shown in Fig. 2A, wild-type mice steadily transited in and out of sleep throughout the daylight period, then became continuously awake for a period of 6 to 7 hours beginning at the onset of darkness. Consistent with previous sleep studies in C57BL/6 mice (40), wild-type mice then experienced a period of sleep lasting several hours followed by wakefulness during the period immediately before light was returned.

Fig. 2.

Sleep patterns of wild-type and NPAS2 (–/–) mice. (A) Sample hypnograms derived from stage-scored data from two wild-type and two NPAS2 (–/–) mice over 24 hours of baseline recording in a 12-hour:12-hour LD cycle. Each point represents a stage score assigned for a 15-s epoch, and every other 15-s epoch is plotted (32). (B) Two-day average (±SEM) of percent time spent in NREM and REM sleep in wild-type and NPAS2 (–/–) mice. During the light periods, no differences between genotypes were noted. During the dark periods, NPAS2 (–/–) mice spent significantly less time in NREM sleep than wild-type mice [F(1,37) = 10.14, P = 0.0029, power 0.890]. A significant effect of hour of the night [F(11,407) = 21.68, P < 0.0001, power 1.0] and significant genotype × hour of the night interaction [F(11,407) = 3.12, P = 0.005, power 0.993] were also detected. A similar but nonsignificant trend was observed for REM sleep.

NPAS2 (–/–) mice, like their wild-type littermates, displayed normal patterns of sleep throughout the light period. However, during the active, nocturnal period, they remained awake nearly continuously for the first 8 to 9 hours of darkness (Fig. 2A). The mutants did show brief periods of sleep very late in the 12-hour dark period, consistent with a late decline in wheel running activity (Fig. 1). We quantitatively compared the sleep/wake patterns of 20 wild-type mice and 19 NPAS2 (–/–) mice over 48 hours of recording (Fig. 2B). Non–rapid eye movement sleep (NREMS) in wild-type mice peaked 8 hours into the dark period (ZT 20), whereas the peak in NREMS in Npas2 homozygotes did not occur until 10 hours into the dark period. Relative to wild-type littermates, NPAS2 (–/–) mice showed a 27% reduction in NREMS and a 23% reduction in REM sleep at night, which roughly balanced their 19.3% increase in wheel running activity.

Because Npas2 is expressed in the somatosensory cortex and limbic system nuclei (30), we hypothesized that the altered patterns of sleep and locomotor activity in NPAS2 (–/–) mice might be due to a lack of modulatory input from somatosensory and limbic systems to the master clock in the SCN. If so, the circadian rhythmicity in NPAS2 (–/–) mice may be more reliant on the Zeitgeber of light vs. dark than are the rhythms of their wild-type littermates. To test this hypothesis, we compared the wheel running activity of wild-type and NPAS2 (–/–) mice in response to a 4-hour phase advance in the LD cycle. Remarkably, on the first day of the 4-hour advance, NPAS2 (–/–) mice began running an average of 3.0 (±0.15) hours earlier than normal, whereas wheel running by the wild-type littermates was advanced 2.2 (±0.21) hours (Fig. 3). Moreover, the rate of adaptation to the phase advance was significantly different between genotypes. Complete entrainment as assessed by locomotor activity was observed for animals of both genotypes 5 to 6 days after the change in the timing of the LD cycle. The enhanced adaptability to a phase advance observed in the NPAS2 (–/–) mice did not appear to be due to a compensatory increase in Clock gene expression, as levels of Clock gene expression in the forebrain do not differ between the two genotypes (fig. S2).

Fig. 3.

Effect of a 4-hour phase advance on activity onset in wild-type and NPAS2 (–/–) mice. (A) Actograms from a wild-type and an NPAS2 (–/–) mouse showing baseline activity and response to a 4-hour phase advance. Arrow indicates beginning of phase advance. (B) Summary of the mean daily advance in activity onset over 6 days. The onset of wheel running was dependent on genotype [F(1,54) = 4.41, P = 0.04, power 0.531] with a highly significant difference on the first day. The significant genotype × day of advance interaction [F(5,270) = 2.36, P = 0.04, power 0.751] indicates that the rate of adjustment was faster in NPAS2 (–/–) mice than in wild-type mice.

The light-dominated circadian activity cycle of rodents can be disrupted by a paradigm of restricted feeding (RF). Under a normal LD cycle and ad libitum feeding conditions, mice and rats feed at night. If, however, food is restricted to the light period, they will feed in the day as is needed for survival. As a part of this adaptive response, rodents develop anticipatory locomotor activity wherein they run each day immediately before food is made available. This behavior is reproducible and strictly adherent to the circadian cycle, reflective of an inherent program of behavioral adaptation (41). However, RF does not depend on the SCN (42), and RF does not change the rhythm of the molecular clock in the SCN (43, 44).

We first investigated whether the effects of RF on anticipatory locomotor activity were linked to changes in the molecular rhythm of the forebrain, where Npas2 is expressed. Wild-type C57BL/6J mice were exposed to two RF paradigms. In the first, mice were left on a normal LD cycle with food restricted to a 4-hour window initiated 3 hours into the light period (ZT 3); in the second, mice were switched to constant darkness at the onset of RF. As shown in Fig. 4A, mice left on the standard LD cycle developed a peak of locomotor activity immediately preceding food presentation in addition to maintaining a shortened nocturnal bout of activity. In contrast, mice shifted to constant darkness adapted a single period of wheel running activity immediately preceding food presentation.

Fig. 4.

Behavioral and molecular adaptation of wild-type mice to two paradigms of restricted feeding (32). (A) Representative actograms from C57BL/6J males subjected to 4 hours restricted feeding (RF) during LD (left) or during DD (right, start of DD indicated by arrow). Periods during which food was available are shaded. (B) Double-plotted Northern blots showing changes in circadian gene expression in the mouse forebrain in response to RF. C57BL/6J mice received food ad libitum (left) or were subjected to RF under either LD (middle) or DD (right) conditions.

In what way is the molecular clock of the forebrain affected by RF? Under normal LD and ad libitum feeding conditions, and consistent with previous reports (25, 44, 45), mRNA expressed from the Per2, Per3, and Cry1 genes peaked in abundance around ZT 16, with BMAL1 gene expression reaching its highest level at ZT 0 (Fig. 4B, fig. S3). Mice subjected to RF but left under a normal 12-hour:12-hour LD cycle revealed bimodal patterns of Per2 and Cry1 gene expression with one peak of mRNA abundance occurring at the onset of darkness (ZT 12). For Per2, the newly acquired mRNA peak was concordant with daytime access to food (ZT 4), whereas for Cry1, the newly generated peak was concordant with the onset of food anticipatory running activity (ZT 0). The peak of Per3 mRNA abundance was advanced by 4 hours such that it coincided with the onset of nocturnal running (ZT 12). Animals that were shifted to constant darkness concomitant with the onset of RF developed single peaks of Per2, Per3, Cry1, and Bmal1 gene expression fully 12 hours antiphase to their normal pattern.

Knowing that NPAS2 is expressed in the forebrain (30, 31) and is necessary for rhythmic oscillation of Per and Cry gene expression (25), we studied the adaptation of NPAS2 (–/–) mice to RF. Remarkably, a large proportion of NPAS2 (–/–) mice became sick (inactive, piloerected, and hypothermic) within 4 to 8 days of the RF challenge. Of 11 mutant mice tested, two died and seven became severely ill. In contrast, only 5 of 13 wild-type littermates became ill and none died. Moreover, the surviving NPAS2 (–/–) mice were judged to be sick for a significantly longer time (3.0 ± 0.7 days) than their wild-type littermates [0.79 ± 0.5 days; t(22) = 2.2, P = 0.007].

It was possible to protect NPAS2 (–/–) mice from the lethal effect of RF by placing mouse diet pellets directly on the cage floor of the sick animals and increasing the period of food availability from 4 hours to 6 hours (32). Using a new group of 18 wild-type and 15 NPAS2 (–/–) mice, we ascertained that both food consumption and body weight of NPAS2 (–/–) mice were substantially reduced relative to wild-type littermates 1 week after exposure to the RF paradigm. As shown in Fig. 5A, food intake for mice of both genotypes dropped precipitously on the first day of RF, and then rose steadily for 2 days. From day 5 through day 13, a time during which NPAS2 (–/–) mice exhibited outward signs of sickness, food intake was significantly lower in the mutant mice. Initially, mice of both genotypes lost roughly 25% of total body weight (Fig. 5B). From day 8 until the end of the study, body weight in the NPAS2 (–/–) mice was significantly lower than that of wild-type littermates. By both of these criteria, food consumption and body weight, NPAS2 (–/–) mice appeared to be compromised in adaptation to RF.

Fig. 5.

Effect of RF on food intake, body weight, and food anticipatory activity (FAA) in wild-type and NPAS2 (–/–) mice (32). (A) Mean food intake (± SEM) during RF was significantly different between genotypes [F(1,30) = 9.84, P = 0.004, power = 0.88]. *P < 0.01 by t test. (B) Mean body weight (± SEM) during RF was dependent on genotype [F(1,30) = 5.019, P = 0.032, power = 0.576]. *P < 0.02 by t test. (C) Mean daily amount (±SEM) of FAA during 15 days of RF. Acquisition of FAA was significantly slower in NPAS2 (–/–) than in wild-type littermates [F(1,29) = 27.6, P < 0.0001, power = 1.00]. *P < 0.04 by t test. (D) Actograms of wheel running in wild-type (left) and NPAS2 (–/–) (right) mice over 14 days of baseline followed by 15 days of RF in DD. Light-gray shading indicates time during which food was available; arrow indicates the beginning of DD.

Acquisition of anticipatory locomotor activity in response to RF was also delayed in NPAS2 mutants (Fig. 5, C and D). Between day 3 and day 11 of RF, food anticipatory activity was significantly reduced in NPAS2 (–/–) mice. By day 11 of the study, however, the amount of food anticipatory activity was indistinguishable between the two genotypes. Similar results were seen when RF was conducted in LD (fig. S4). Moreover, food anticipatory activity established in LD conditions persisted in mice of both genotypes when food was presented ad libitum (fig. S5). Limiting food availability to 4 hours during the dark for mice kept on a 12-hour:12-hour LD cycle did not produce genotype-dependent differences in food intake or body weight (fig. S6). Thus, the impaired ability of NPAS2 (–/–) mice to acquire food anticipatory activity did not appear to be secondary to a metabolic deficiency, but was due to dysfunction in the CNS mechanisms responsible for developing a food-entrainable oscillator in a timely manner.

We hypothesize that in mice, and possibly other mammals, CLOCK and NPAS2 function as nonredundant regulators of circadian behavior. CLOCK, by virtue of its role in the SCN, is light-regulated and participates in specifying light-driven aspects of circadian behavior. NPAS2, by virtue of its role in the forebrain, is regulated by sensory stimuli and may participate in the execution of sensory-driven behavior. When food is provided ad libitum, mice forage and feed at night. In so doing, animals are exposed to a diverse array of stimulatory sensory input typical of wakefulness. We hypothesize that wakefulness, coupled with its abundant access to sensory stimuli, represents the underlying entraining input for NPAS2. As such, under normal conditions, the two entrainment conduits, light (CLOCK) and sensory stimulation (NPAS2), are applied in synchrony. When food is not available at night, mice must adapt their behavior to survive. Under such circumstances, animals forage and feed in the day, and the two entrainment conduits are applied asynchronously.

In the absence of NPAS2, we hypothesize that circadian behavior is dictated primarily by the rhythmic output of the SCN. Because this output is phase-locked to the LD cycle, NPAS2 (–/–) animals are virtual “slaves” to light. This interpretation is consistent with the observations of Pitts and colleagues who have found that CLOCK (–/–) mice show an enhanced adaptability to RF (46).

In response to RF, the endogenous rhythm of the SCN remains unchanged, yet those of peripheral tissues adapt (43, 47, 48). As demonstrated herein, the NPAS2-driven rhythm in the forebrain also adapts in response to RF. We speculate that the entraining influence that adapts NPAS2 to a new, food-driven rhythm will be traced to neuronal activity (49). Because NPAS2 is expressed in somatosensory, visual, and auditory cortices and in the olfactory tubercules and bulb (30, 31), it is possible that new, daytime electrical input transmitted through these brain nuclei represents the entraining signal causing adaptation of the molecular clock operative in the forebrain. Because this system, unlike the SCN, can adapt in the absence of light/dark cues, it may also facilitate behavioral adaptation. If so, this may explain why NPAS2 (–/–) mice adapt poorly to RF.

The concept of a CLOCK/NPAS2 balance may also explain the unanticipated effects NPAS2 appears to exert on locomotor activity and sleep. The phenomenon of bimodal activity has been studied in an inbred strain of mice, designated CS mice. CS mice exhibit spontaneous splitting of circadian rhythm under constant darkness (35, 50). The SCN tissue of split-rhythm mice displays unimodal, light-driven peaks of Per1, Per2, and Per3 gene expression (35). In contrast, in the cerebral cortex, Per gene expression undergoes bimodal fluctuation with peaks corresponding temporally with split-activity rhythms as deduced by wheel running. Npas2 expression corresponds with the regions showing bimodal peaks of Per gene expression in split-rhythm mice (50). As demonstrated here, the bimodal activity and sleep patterns apparent in wild-type mice are lacking or blunted in NPAS2 (–/–) mice. It may thus be the case that cerebrocortical regions of the mammalian brain help specify a bimodal pattern of behavioral activity.

Remnants of this crepuscular (dawn/dusk) activity, so prominent in species ranging from mosquitoes to deer and fox, may extend to humans, as exemplified by our propensity to tire midday. Could it be that cultures embracing an afternoon siesta are optimally tuned to our endogenous circadian rhythm?

Supporting Online Material

Materials and Methods

Figs. S1 to S6


References and Notes

View Abstract

Navigate This Article