Acute Physical Stress Elevates Mouse Period1 mRNA Expression in Mouse Peripheral Tissues via a Glucocorticoid-responsive Element*

In mammals, the circadian and stress systems (both centers of which are located in the hypothalamus) are involved in adaptation to predictable and unpredictable environmental stimuli, respectively. Although the interaction and relationship between these two systems are intriguing and have been studied in different ways since the “pre-clock gene” era, the molecular interaction between them remains largely unknown. Here, we show by systematic molecular biological analysis that acute physical stress elevated only Period1 (Per1) mRNA expression in mouse peripheral organs. Although behavioral rhythms in vivo and peripheral molecular clocks are rather stable against acute restraint stress, the results of a series of promoter analyses, including chromatin immunoprecipitation assays, indicate that a glucocorticoid-responsive element in the Per1 promoter is indispensable for induction of this mRNA both in vitro and in vivo. These results suggest that Per1 can be a potential stress marker and that a third pathway of Per1 transcriptional control may exist in addition to the clock-regulated CLOCK-BMAL1/E-box and light-responsive cAMP-responsive element-binding protein/cAMP-responsive element pathways.

uncontrollable stressors have been shown to induce changes in a wide variety of behavioral parameters, including decreases in general locomotor activity and explorative behavior as well as impairment of sexual behavior, feeding, and drinking (27). Social stressors can be uncontrollable. Uncontrollable stresses induce classical stress responses via the autonomic sympathoadrenal axis and the hypothalamic-pituitary-adrenal axis with an acute and strong neuroendocrine activation. The hypothalamic-pituitary-adrenal axis, which mobilizes bodily resources through the release of glucocorticoids, and the autonomic sympathoadrenal axis have been considered to be the main neuroendocrine systems involved in the integrated stress response. Increased hypothalamic-pituitary-adrenal activity and catecholaminergic activity, both interplaying with various other neuroendocrine systems, orchestrate an adequate response to the unexpected challenge.
Various studies on the relationship between stress and circadian rhythms by different kinds of stressors in different species of animals have led to the general idea that the central circadian oscillator appears to be well protected against unpredictable stressful stimuli, but that certain stressors can strongly affect the output of the clock and the expression of the rhythm (1). Moreover, it is well known that serum cortisol levels in humans and rodents show circadian rhythms. However, with fewer studies using mice compared with traditionally chronobiology-preferred animals, such as hamsters and rats, the underlying molecular mechanisms of this classical and intriguing relationship between the circadian and stress systems have remained unclear. Does stress alter the circadian rhythm in mice? Does corticosterone alter circadian rhythms in behavior or gene expression? If stress or corticosterone affects clock gene expression, what is the mechanism? In this study, we systematically examined the temporal expression of mRNAs of clock and clock-related genes in mouse peripheral tissues as well as their behavioral rhythms after an acute restraint stress.

EXPERIMENTAL PROCEDURES
Animals-Male ICR (for molecular study) and BALB/c (for behavioral study) mice purchased 5 weeks postpartum from Japan SLC, Inc. (Hamamatsu, Japan) were housed under 12-h light/12-h dark (LD) cycles over 2 weeks. A dark/dark cycle means complete darkness as a continuation of the dark phase of the last LD cycle, and samplings were made in the third dark/dark cycle or later. All protocols using animals in this study were approved by the Osaka Bioscience Institute Animal Research Committee.
Sample Preparation-ICR mice were restrained for 1 h in a 31-mm, inner diameter, ϫ 85-mm acryl tube and then housed in individual cages or killed immediately, and peripheral tissues and blood were collected. The peripheral tissues collected were immersed overnight at 4°C in RNAlater (Ambion, Inc.), and then total RNA was isolated using an SV total RNA isolation system (Promega Corp.). Serum corticosterone concentration was measured with an enzyme immunoassay kit (Assay Designs, Ann Arbor, MI).
Behavioral Analysis-After entrainment under the LD conditions (lights on from 8:00 to 20:00), BALB/c mice were housed individually in cages equipped with an infrared locomotor recording apparatus (Biotex, Kyoto, Japan) and maintained under dark/dark cycles as described previously (28). To estimate free-running periods, we used a 2 periodogram (29). We estimated periods by using 7 days of data just before and 12 days after the stress-imposed day. To estimate the activity ratio, the average daily activity for 3 days just after the stress-imposed day was divided by that for 10 days just before the stress-imposed day. For resynchronization experiments, LD cycles were advanced 6 h (lights on from 2:00 to 14:00) just after the stress-imposed day.
Quantitative Real-time Reverse Transcription (RT)-PCR-A Taq-Man low density array (Applied Biosystems), which contained mouse (m) Per1, mPer2, mPer3, mArnt1 (mBmal1), mNpas2, mCry1, mCry2, mBhlhb2 (mDec1), mBhlhb3 (mDec2), mDbp, and mNfil3 (mE4bp4) as clock genes and 18 S rRNA as an internal control, was examined using an ABI PRISM 7900HT sequence detection system (Applied Biosystems). For one port of the TaqMan low density array, 100 ng of cDNA template was mixed with 50 l of 2ϫ TaqMan Universal PCR Master Mix (Applied Biosystems) and filled up to 100 l with distilled water. The reaction was first incubated at 50°C for 2 min at then at 95°C for 10 min, followed by 40 cycles at 95°C for 15 s and 60°C for 1 min.
Each quantitative real-time RT-PCR was performed using the ABI PRISM 7900HT sequence detection system as described previously (19). The PCR primers and probes (5Ј-FAM TM and 3Ј-TAMRA TM ) were designed with Primer Express software (Applied Biosystems), and the sequences of the primers and probes were as follows: mPer1 FW, GAA AGA AAC CTC TGG CTG TTC CT; mPer1 RV, GCT GAC GAC GGA TCT TTC TTG; mPer1 TaqMan probe, CCT CAG GTA TTT GGA GAG CTG CAA CAT TCC; mPer2 FW, CGC CTA GAA TCC CTC CTG AGA; mPer2 RV, CCA CCG GCC TGT AGG ATC T; mPer2 TaqMan probe, AGG CTG TGG ATG AAA GGG CGG TC; mBmal1 FW, GCA GTG CCA CTG ACT ACC AAG A; mBmal1 RV, TCC TGG ACA TTG CAT TGC AT; mBmal1 TaqMan probe, ATC AAG AAT GCA AGG GAG GCC CAC A; 18 S rRNA FW, CGC CGC TAG AGG TGA AAT TC; 18 S rRNA RV, CGA ACC TCC GAC TTT CGT TCT; and 18 S rRNA TaqMan probe, CCG GCG CAA GAC GGA CCA GA. For a 25-l PCR, 50 ng of cDNA template was mixed with the primers and probes to final concentrations of 300 and 2 nM, respectively, and 12.5 l of 2ϫ TaqMan Universal PCR Master Mix. The reaction was first incubated at 50°C for 2 min and then at 95°C for 10 min, followed by 40 cycles at 95°C for 15 s and 60°C for 1 min.
Cosine Fitting Curve-To estimate the phase of cycling genes, we prepared curves of 24-h periodicity with phases (b) from 0 to 23 h at intervals of 1 h, yielding a total of 24 model cosine curves, and calculated the correlation coefficient of the cosine curves and cycling genes. For model cosine curves and cycling genes observed over a series of N conditions, a similarity score can be computed as shown in Equations 1 and 2.
The values of X are set to the mean of observations on X; X then becomes the standard deviation of X, and S(cos, X) is exactly equal to the Pearson correlation coefficient between the values of cosine wave and observation on X. As for the phase of each gene, from 24 model curves, we selected the phase curve best fit with the highest correlation coefficient.
In Silico Search-Computer-aided analysis was performed as described previously (19). Briefly, human, rat, and mouse sequences of Per1 were downloaded from the Celera Data Base System and the NCBI Gene Database. Each Per1 sequence spanning from 9 kb upstream to 4 kb downstream of the transcription start site was obtained. Multiple sequence alignment of these sequences for Per1 was performed using ClustalW Version 1.83 with default parameters. The binding elements were then searched from this alignment using a pattern-finding tool, FUZZNUC (available at bioweb.pasteur.fr/seqanal/interfaces/ fuzznuc.html), with the following consensus sequences allowing for a 1-base mismatch: DBP-binding element, (G/A)T(G/T)A(T/C)GTAA-(T/C); E-box, CACGTG; ROR/REV-ERB element, (A/T)A(A/T)NT-(A/G)GGTCA; and glucocorticoid-responsive element (GRE), AGAA-CANNNTGTTCT. The accession numbers used and sequence numbers analyzed are as follows: human, NT_010718.14 and c6905708 -6891708; rat, rCG34390 and 52960430 -52974430; and mouse, NT_039515.2 and 65661216 -65676215.
mPer1 Promoter Constructs and Transient Reporter Assay-Bacterial artificial chromosome clone RP23-26L6 was purchased from Invitrogen (BACPAC Resource Center at Children's Hospital Oakland Research Institute). The 6.301-kb region upstream of the translation initiation codon of mPer1 was fused to the luciferase gene in the pGL3-Basic vector (P1F). Deletion constructs P1K and P1S were prepared by removing the region starting at 3.806 and 2.369 kb, respectively, upstream from the translation initiation site of mPer1. The mutant constructs (P1FdM, P1FpM, and P1Fd/pM) for the distal GRE (dGRE) and the proximal GRE (pGRE) were prepared by changing AGA ACA CGA TGT TCC to AGA CAG CGA TGT TCC and GGA ACA TCC TGT TCT to GGA CAG TCC TGT TCT, respectively.
NIH3T3 cells were seeded in 24-well plates at a density of 4 ϫ 10 4 cells/well 1 day prior to transfection. Transfection was carried out using PolyFect (Qiagen Inc.). 3 h after transfection, dexamethasone (final concentration of 0.1 M; Sigma) was added to the medium. 27 h after transfection, the cells were washed with phosphate-buffered saline and harvested in 100 l of passive lysis buffer (Promega). Luciferase activities were assayed with the Dual-Luciferase reporter assay system (Promega Corp.) using a Turner Designs TD-20/20 luminometer. A Renilla luciferase plasmid was used for cotransfection to normalize each transfection assay. In this study, the cells were cotransfected with 200 ng of the pGL3 firefly luciferase reporter plasmid and 20 ng of the phRL-TK Renilla luciferase internal control plasmid.
Chromatin Immunoprecipitation (ChIP) Assays-ChIP assays were carried out on NIH3T3 cells and mouse liver. We performed ChIP assays on mouse liver as described (available at genomecenter.ucdavis.edu/farnham/protocols/tissues.html) with slight modifications. After sonication to obtain DNA fragments 100 -600 bp in length, cellular debris was removed by centrifugation, and the lysates were diluted 1:4 in ChIP dilution buffer (50 mM Tris-HCl (pH 8.0), 167 mM NaCl, 1.1% Triton X-100, and 0.11% sodium deoxycholate supplemented with 1 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, and 10 g/ml leupeptin). Nonspecific background was removed by incubating samples with a salmon sperm DNA/protein G-agarose slurry for 4 h at 4°C with rotation. The samples were centrifuged, and 0.1 volume of the recovered supernatants was stored as an input sample, whereas the rest was incubated overnight with 5 g of the indicated antibodies at 4°C with rotation. Antibodies against the glucocorticoid receptor (GR) (catalog no. sc-1002) and normal rabbit IgG (catalog no. sc-2027) were purchased from Santa Cruz Biotechnology, Inc. The immunocomplexes were collected with 10 l of protein G-agarose slurry for 4 h at 4°C with rotation. The beads were sequentially washed with the following buffers: 50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% SDS, and 0.1% sodium deoxycholate; 50 mM Tris-HCl (pH 8.0), 500 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% SDS, and 0.1% sodium deoxycholate; and 10 mM Tris-HCl (pH 8.0), 0.25 M LiCl, 1 mM EDTA, 0.5% Nonidet P-40, and 0.5% sodium deoxycholate.
Finally, the beads were washed twice with 10 mM Tris-HCl (pH 8.0) and 1 mM EDTA. The immunocomplexes were then eluted by the addition of 200 l of ChIP direct elution buffer (10 mM Tris-HCl (pH 8.0), 300 mM NaCl, 5 mM EDTA, and 0.5% SDS) and incubated for 4 h at 65°C. The remaining RNAs and proteins were digested by the sequential addition of RNase A (final concentration of 20 g/ml) for 30 min at 37°C and proteinase K (final concentration of 50 g/ml) for 1 h at 55°C. The DNA was recovered by phenol/chloroform/isoamyl alcohol (25:24:1) extraction and ethanol precipitation. For NIH3T3 cells, we slightly modified the above method. Approximately 1 ϫ 10 6 cells were collected in a 1.5-ml tube and cross-linked by the addition of formaldehyde for 10 min. Following the addition of glycogen to stop cross-linking, the cells were collected by centrifugation and washed once with phosphate-buffered saline. After centrifugation, the cell pellets were resuspended in 250 l of SDS lysis buffer (50 mM Tris-HCl (pH 8.0), 10 mM EDTA, and 1% SDS supplemented with 1 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, and 10 g/ml leupeptin) and incubated on ice for 20 min. After sonication, cellular debris was removed by centrifugation, and the lysates were diluted 1:8 in ChIP dilution buffer. The following steps were same as described above for the protocol used for liver samples. The primer sequences for ChIP assays were as follows:

Effect of Restraint Stress on Behavioral Rhythms-We first examined whether acute physical stress influences circadian behavioral rhythms.
Mice were subjected to restraint stress during circadian time (CT) 1-2 and CT4 -5, when mPer1 was strongly induced as described below (see Fig. 3), and during CT14 -15, when mPer1 was least induced (see Fig. 3). The mice showed a clear free-running rhythm of locomotor activity in the dark/dark cycle after 2-week entrainment in the LD cycle (Fig. 1A). The values of the free-running period of activity of the control mice ranged from 22.7 to 24. 1 h (Fig. 1B). The mean free-running period averaged 23.50 Ϯ 0.08 h (n ϭ 20). Some of the stress-imposed mice seemed to show a longer period; however, statistical analyses revealed no significant difference regardless of time when exposed to the acute stress. Daily activity was then compared as the activity ratio, defined as daily activity after treatment divided by that before treatment. No significant difference was found between stress-imposed and control mice regardless of the time treated (Fig. 1C). Mice were subjected to a 6-h advance in the LD cycle after restraint stress, and the number of days to resynchronization of the activity rhythm was counted (Fig. 1D). No significant difference was observed between stress-imposed and control mice. These in vivo behavioral data suggest that mouse circadian clocks behave with variability, but are largely unaffected by a 1-h restraint stress.
Effect of Restraint Stress on Temporal Expression of Clock Genes in Peripheral Tissues-To explore the effect of restraint stress on the peripheral clocks, we next examined the temporal expression of clock genes in mouse peripheral tissues after a restraint stress. Mice were housed for 2 weeks under the LD conditions and then exposed to a 1-h restraint stress at Zeitgeber time (ZT) 1-2 or ZT14 -15. Approximately 12 h after the restraint stress, we started to collect samples of peripheral tissues every 3 h and examined the temporal expression of mPer1, mPer2, and mBmal1 mRNAs by quantitative real-time RT-PCR. Data on temporal expression of mPer1, mPer2, and mBmal1 mRNAs, which were obtained from each sample for stress ending at ZT2 and ZT15, together with controls, were plotted for three different tissues, i.e. liver, heart, and kidney (Fig. 2, A-C, respectively). As seen in behavioral rhythms in vivo, interindividual variability was observed, especially in the stress-imposed groups. We applied a "cosine fitting curve method" to objectively simulate the circadian curve and to estimate the peak time of cycling clock genes. Cosine curves demonstrating the highest scores in the Pearson correlation coefficient were accepted as the fittest cosine curves of cycling clock genes. The peak times and Pearson correlation coefficients in the liver obtained by this cosine fitting curve method were ZT13 and 0.82 ( Fig. 2A, panel a) and ZT0 and 0.85 (panel i), respectively. The peak times of these three genes according to this simulation were likely to be the same and were not affected by the 1-h restraint stress ending at either ZT2 or ZT15. We also obtained similar results for the heart and kidney (Fig. 2, B and C, respectively). The peak times and correlation coefficients were ZT11 and 0.71 (Fig. 2B, panel a) (ZT2 and ZT15) in the heart and mPer2 (ZT2) in the kidney were delayed compared with those in the controls; however, the Pearson correlation coefficient was relatively small. Although phase shifts of even 1 h or so would indeed be undetectable, these results argue that, against a 1-h restraint stress, peripheral molecular clocks are rather stable.
Restraint Stress Induces mRNA Expression of mPer1, but Not That of Other Clock Genes in Mouse Peripheral Tissues-To ascertain whether acute physical stress influences circadian clocks at the molecular level, we subjected mice to restraint stress for 1 h every 4 h in 1 day and then examined transcripts in their peripheral tissues sampled after stress treatment by quantitative real-time RT-PCR. Fig. 3A shows the mRNA expression levels of canonical clock genes (mPer1, mPer2, mPer3, mBmal1, mNpas2, mCry1, mCry2, mDec1, mDec2, mDbp, and mE4bp4) in mouse liver after 1 h of restraint stress. The circadian patterns of controls were consistent with data reported previously (19,30). Mice were subjected to restraint stress starting at ZT23, ZT3, ZT7, ZT11, ZT15, or ZT19 so that mice were free again at ZT0, ZT4, ZT8, ZT12, ZT16, or ZT20, respectively. At all time points tested except ZT16, the amount of mPer1 mRNA was significantly elevated (Fig. 3A  and supplemental Fig. S1). The -fold change in the mPer1 increase seemed to be larger in light than in darkness. However, a 1-h restraint stress failed to affect the expression of the other clock genes (mPer2, mPer3, mBmal1, mNpas2, mCry1, mCry2, mDec1, mDec2, mDbp, and mE4bp4) with two exceptions (ZT8 for mPer3 and ZT20 for mDec1) (Fig. 3A). We also confirmed that the 1-h restraint stress increased only mPer1 mRNA expression in the other peripheral tissues examined: heart, lung, and stomach (Fig. 3, B-D, respectively; and supplemental Fig. S1). Among these three tissues, again only mPer1 expression was significantly increased by this kind of stress, with certain exceptions at some points. As it is well known that acute stress elevates serum corticosterone levels (31), we investigated whether the serum corticosterone level was increased immediately after the 1-h restraint stress. The level FIGURE 1. Effects of a 1-h restraint stress on behavioral parameters in mice. A, the activity records of BALB/c mice are shown. The light regime consisted of 2 weeks in an LD cycle, followed 24 days in a dark/dark cycle. The acute restrain stress subjected is indicated by boldface dots. B, free-running periods were compared between pre-stress (PrS) and post-stress (PoS) periods. C, the daily activities of stress-imposed mice were compared with those of control mice. The activity ratio indicates the daily activity after stress divided by that before stress. D, the numbers of days until resynchronization of the activity rhythm were counted. Mice were exposed to stress at ZT2 or ZT15 on the day just before a 6-h advance of the LD cycle. Dots refer to sample values, and short bars refer to the average of values. A-C, n ϭ 13 (CT2 stress), n ϭ 12 (CT15 stress), and n ϭ 5 (others); D, n ϭ 4 (ZT2 and ZT15) and n ϭ 3 (control (ctrl)). showed a 3-40-fold increase at the various time points tested compared with the relative controls (supplemental Fig. S2), in which the serum corticosterone level fluctuated in a circadian manner as described (32).

FIGURE 2. Effects of a 1-h restraint stress at ZT1 (ZT2) or ZT14 (ZT15) on the temporal expression of clock genes in mouse liver (A), heart (B), and kidney (C). Patterns for
GREs in the Per1 Genome Sequence-Although we observed no significant difference in the behavior of in vivo rhythms or peripheral molecular clocks between the stress-imposed and control mice, the finding that the restraint stress immediately induced only mPer1 mRNA led us to hypothesize that a functional element that is conserved in the Per1 genome sequence but not in other clock gene sequences among species is responsible for this rapid induction of Per1. To test this hypothesis, we analyzed the promoter of mPer1. Our previous "dry" approach revealed that there are four E-box elements and two DBPbinding elements in Per1 genome sequences, which are conserved among human, rat, and mouse (19). However, these two elements are also present in the genome sequences of other clock genes. For example, E-box elements are present in Dbp and Per2 promoter regions, and DBP-binding elements are found in Per3 promoter regions (19). Further in silico analysis revealed a candidate cis-element, i.e. a GRE-like element (AGA ACA NNN TGT TCC) (Fig. 4A, upper panel), that is conserved among the three mammalian genomes, but is not seen in the other genome sequences of the canonical clock genes described here. The reason why we call it a GRE-like element is that the consensus sequence of the GRE is AGA ACA NNN TGT TCT, i.e. the last T was changed to C in the Per1 genome sequence. In addition, we found another GRE-like element within the first intron, although this element is not conserved among these three species (Fig. 4A, lower panel). Glu-cocorticoid is known to be an indicator of acute stress (supplemental Fig. S2) (31), and its secretion over a 24-h period shows a circadian pattern (supplemental Fig. S2) (32) and is thought to be controlled by a circadian clock within the SCN (33).
The dGRE in the mPer1 Genome Is Functional- Fig. 4B presents a schematic diagram of the mPer1 gene, including its promoter region covering 6.3 kb upstream of the translation initiation codon. Within this region, there are five E-boxes and one CRE, all of which were found experimentally to be functional elements for mPer1 transcription (25,34). In this study, we found, by in silico analysis as described above, two GRE-like elements, one located adjacent to the CRE in the 5Ј-promoter region (dGRE) and the other in its first intron (pGRE). To determine whether these two GRE-like elements are functional or not, we conducted a luciferase assay using transfected NIH3T3 cells.
P1F, which includes a 6301-bp sequence upstream of the translation initiation site of mPer1 flanked by the luciferase gene in the pGL3-Basic vector, showed a 5-10-fold increase in luciferase activity after treatment with the GR agonist dexamethasone compared with that after EtOH (solvent) treatment (Fig. 4, C and D), which is consistent with a previous report (35). To learn whether the GRE(s) is involved in the increase in luciferase activity in response to dexamethasone, we constructed a series of deletion  mutants and point mutants of the mPer1 promoter region. The luciferase assay using deletion mutants demonstrated that a dexamethasone-responsive element(s) resided between positions Ϫ3806 and Ϫ2369 (Fig. 4C). This result suggests that the dGRE is crucial for mPer1 induction by dexamethasone. Moreover, the results obtained for the point mutants clearly indicated that of the two GREs, only the dGRE was functional (Fig. 4D). Taken together, these results strongly suggest that the dGRE in the mPer1 promoter is necessary and sufficient for glucocorticoid signaling to elevate the amount of mPer1 mRNA.
The GR Controls Per1 Expression in Vivo-To assess the association of the GR with the native mPer1 promoter in NIH3T3 cells in the presence or absence of dexamethasone treatment, we performed ChIP assays. Immunoprecipitated samples were subjected to semiquantitative RT-PCR (Fig.  5B) and quantitative real-time RT-PCR (Fig. 5C) using primers designed to specifically amplify the dGRE and pGRE sites in the mPer1 promoter (Fig.  5A). As a negative control, we amplified the mPer2 exon 1 region as described (36). As demonstrated in Fig. 5, we observed that the GR bound to both the dGRE and pGRE in the mPer1 promoter and its first intron, respectively, upon dexamethasone treatment, but did not bind to them upon ethanol treatment. In contrast, ChIP assays using negative control primers displayed no binding to the GR in the presence or absence of dexamethasone treatment. To confirm whether the GR binds to GREs in the mPer1 promoter in vivo, we again performed ChIP assays using immunoprecipitated liver samples derived from mice before and after stress treatment. As shown in Fig. 6, by comparison with a control (normal rabbit IgG), stress strongly induced GR binding to the GRE in the mPer1 promoter. GR binding to the dGRE was significantly higher than that to the pGRE. Again no binding was observed using negative control primers. These data represent compelling evidence that the GR binds to the GRE in the mPer1 promoter in vivo and that the association of the GR with the dGRE in the mPer1 promoter is pronounced in response to pathophysiological stimuli such as stress.
Restraint Stress Effects Correlate with Corticosterone Effects-To connect the evidence that restraint stress effects are mediated by the GRE identified above, we manipulated plasma levels of corticosterone by injections of corticosterone into mice in the absence of stress and quantified the mPer1 mRNA expression in the liver. The expression of mPer1 transcripts increased in a corticosterone injection dose-dependent manner (Fig. 7A), suggesting that the restraint stress effects link to the above GRE results in vivo. Moreover, we systematically examined the expression levels of canonical clock genes in mouse liver 2 h after corticosterone injections. Of 11 clock gene mRNAs, only mPer1 mRNA exhibited a significant increase (Fig. 7B), consistent with the result of acute stress shown in Fig. 3. These results indicate that restraint stress effects correlate with corticosterone effects by way of the GRE and that restraint stress activates Per1 expression via corticosterone in vivo.

DISCUSSION
In this study, the restraint stress failed to influence the circadian mechanisms involving behavioral rhythms in vivo (Fig. 1), perhaps because of interindividual differences in stress sensitivity and in the functioning of the circadian system (37). Before mammalian clock genes FIGURE 5. In vivo binding of the GR to both the pGRE and dGRE in the mPer1 promoter analyzed by ChIP in NIH3T3 cells. A, schematic diagram of the mouse Per1 promoter and primers used for ChIP assay. Ex1 and Ex2, exons 1 and 2, respectively. B, representative results of semiquantitative RT-PCR. Cross-linked nuclear extracts were isolated from NIH3T3 cells before (0hr) and after 1 h of dexamethasone (Dex) or EtOH treatment and subjected to immunoprecipitation (IP) with anti-GR and anti-normal rabbit IgG antibodies. Anti-normal rabbit IgG antibody and mPer2 primers were used as controls for immunoprecipitation and PCR, respectively. C, quantification of ChIP by quantitative real-time RT-PCR. Quantitative real-time RT-PCR was performed on the same samples as described for B. All data presented are the means Ϯ S.E. of three independent samples. were identified, Horseman and Ehret (38) reported that glucocorticoid injection, achieving a concentration 10 times greater than that of dexamethasone by the protocol of Balsalobre et al. (35) as described below, is a circadian Zeitgeber for the body temperature rhythms in rats. However, that report also indicated much individual variation in phase shifts to dexamethasone injection. On one hand, it is reasonable from a teleological view that stress does not alter circadian behavior in mice. If stress can easily alter behavioral rhythms in vivo, no stable rhythms would be found, negating the primary reason for having fundamental circadian rhythms.
We also demonstrated that the restraint stress had little effect with respect to the resetting of circadian time at the molecular level in peripheral tissues such as liver, heart, and kidney (Fig. 2). To analyze this matter objectively, we chose the cosine fitting curve method to elucidate the peak time and calculated the Pearson correlation coefficient to confirm reliability. Although 11 of 14 stress-imposed samples indicated the same time peaks compared with respective controls, three points (Fig. 2,  B, panels b, and c; and C, panel b) showed altered time peaks. The values of the Pearson correlation coefficient were, however, under 0.75, which means less reliability with respect to the time peak. This demonstrates that circadian mechanisms at the molecular level in peripheral tissues have robustness against an acute 1-h restraint stress. Combined with behavioral results, we conclude that the circadian rhythm is rather stable against acute stress.
As referenced above, Balsalobre et al. (35) showed that dexamethasone, a synthetic form of corticosterone, can reset the circadian time at the molecular level in peripheral tissues. However, we could not demonstrate this resetting mechanism. As to why our restraint stress could not alter the circadian clock, we consider that the dosage of dexamethasone used (35) was much larger compared with the level of corticosterone under physiological and stress-imposed conditions. The total serum corticosterone was approximately 1.0 g/mouse under the conditions of our protocol, whereas the total dexamethasone was approximately 50 g/mouse according to the protocol of Balsalobre et al. (35). Furthermore, the titer and biological half-life of dexamethasone are approximately 30-fold higher and 4-fold longer, respectively, than those of corticosterone (39). In other words, it is difficult to mimic the serum corticosterone concentration under physiological conditions using dexamethasone. Furthermore, one possibility for unresponsiveness is that the peripheral circadian clocks carry some feedback or compensatory systems against an elevation of the serum corticosterone concentration by acute likely events such as the 1-h restraint stress. In contrast, it has been postulated that a rapid accumulation of Per1 mRNA in the SCN causes phase shifting by light stimuli (40). However, some reports have demonstrated that the rapid induction of Per1 mRNA is not necessary to reset the circadian time against certain stimuli, e.g. non-photic stimuli for mice (41) and glucose treatment for cultured cells (42). Thus, we consider, as another possibility, that it is neither necessary nor sufficient for a phase shift to rapidly accumulate Per1 mRNA and that the rapid induction of mPer1 mRNA by this type of stress might be involved in other cellular and/or physiological mechanisms rather than the resetting of circadian time. It would be interesting, for example, to explore downstream genes by transcriptome analysis and to examine whether or not this mPer1 induction influences other cellular and/or physiological mechanisms.
We demonstrated that 1-h restraint stress caused a rapid increase in the level of mPer1 mRNA, as a stress-responsive marker, but not in that of other clock genes (mPer2, mPer3, mBmal1, mNpas2, mCry1, mCry2, mDec1, mDec2, mDbp, and mE4bp4) in mouse peripheral tissues (Fig. 3  and supplemental Fig. S1). A previous work reported that forced swim- FIGURE 7. Corticosterone-dependent induction of mPer1 mRNA in mouse liver. A, liver total RNAs were isolated from corticosterone-or phosphate-buffered saline (PBS)injected mice and subjected to quantitative real-time RT-PCR using mPer1 primers. The relative levels of mPer1 mRNA were normalized to the corresponding 18 S rRNA levels. B, the induction of other clock and clock-controlled genes by corticosterone was examined using liver samples treated with 10 mg/kg corticosterone. The relative levels of each mRNA were normalized to the corresponding 18 S rRNA levels. Each mRNA amount treated with phosphate-buffered saline was set to 1. All data presented are the means Ϯ S.E. of three independent samples. *, p Ͻ 0.05; **, p Ͻ 0.01 compared with the relative control (Student's t test). ming, another form of stress, elevates mPer1 mRNA expression in the paraventricular nuclei of the hypothalamus, but fails to do so in the mouse SCN and liver (43). This discrepancy between the latter finding (no increase in the liver) and our present result may be partly because our protocol of restraint stress provides a stronger stress than the protocol of forced swimming. 1) The time during which the animals were subjected to forced swimming (15 min) was shorter than that of the restraint stress (60 min); and 2) the serum corticosterone concentration, which is one of the indices of a stress response, after the restraint stress (supplemental Fig. S2) was 2-3-fold higher than that found after forced swimming (44). Hence, we can hypothesize that the elevation of the mPer1 mRNA in response to stressors is dependent on the concentration of serum corticosterone. We indeed showed that the elevation of the mPer1 mRNA was dependent on the concentration of corticosterone injected in vivo (Fig. 7A). Combined with serum corticosterone concentration results following restraint stress (supplemental Fig. S2), these data indicate the existence of a relaying pathway (stress/corticosterone/Per1) and suggest the existence of GREs in the Per1 promoter region as a link between corticosterone and Per1.
For the first time, we have presented experimental evidence showing that a "functional" GRE exists in the Per1 promoter region and that this GRE is necessary and sufficient for glucocorticoid signaling to cause a rapid increase in the level of Per1 mRNA (Fig. 4). Moreover, ChIP assays revealed that this GRE is indispensable for this mPer1 mRNA induction both in vitro and in vivo (Figs. 5 and 6). Our findings, combined with previous reports, imply that a third pathway for control of Per1 transcription exists, i.e. the glucocorticoid signal in response to stress or other equivalent environmental cues, in addition to the above CREB/ CRE pathway by photic stimuli as well as by transactivation via E-box by CLOCK-BMAL1. Fig. 8 summarizes the schematic models of molecular mechanisms of Per1 transcription. First, the expression of Per1 transcripts exhibits robust circadian rhythms both in the SCN and in peripheral tissues (13,14,19). The most well known mechanism of circadian oscillation as a positive limb is that a heterocomplex of clock transcription factors, CLOCK-BMAL1, binds to the E-box to activate Per1 transcription (Fig.  8A) (20). Second, photic stimulation at the subjective night induces mPer1 and mPer2 mRNAs (but not mPer3 mRNA) in the SCN (15,16), and the induction of mPer1 mRNA is required for light-or glutamateinduced phase shifting (45). The mPer1 and mPer2 promoters (but not the mPer3 promoter) contain CREs that bind CREB, and CREB acts as a pivotal end point of signaling pathways for the regulation of mPer genes by light (Fig. 8B) (25). Third, as shown in this study, in response to stress or other equivalent environmental cues, the glucocorticoid signal induces transient induction of Per1 without correlation with clock regulation. The GR bound to glucocorticoid binds to the GRE in the Per1 promoter and activates Per1 expression (Fig. 8C). Our findings provide not only compelling evidence that the GR binds to the GRE in the Per1 promoter in vivo, but also a new insight into a novel mechanism of Per1 transcription in non-clock functions. Future comprehensive study including non-clock physiology will be needed as a next step in order for us to decipher the physiological significance of this third pathway in Per1 transcription.