DEC1 Modulates the Circadian Phase of Clock Gene Expression

Abstract

DEC1 suppresses CLOCK/BMAL1-enhanced promoter activity, but its role in the circadian system of mammals remains unclear. Here we examined the effect of Dec1 overexpression or deficiency on circadian gene expression triggered with 50% serum. Overexpression of Dec1 delayed the phase of clock genes such as Dec1, Dec2, Per1, and Dbp that contain E boxes in their regulatory regions, whereas it had little effect on the circadian phase of Per2 and Cry1 carrying CACGTT E′ boxes. In contrast, Dec1 deficiency advanced the phase of the E-box-containing clock genes but not that of the E′-box-containing clock genes. Accordingly, DEC1 showed strong binding and transrepression on the E box, but not on the E′ box, in chromatin immunoprecipitation, electrophoretic mobility shift, and luciferase reporter assays. Dec1^−/− mice showed behavioral rhythms with slightly but significantly longer circadian periods under conditions of constant darkness and faster reentrainment to a 6-h phase-advanced shift of a light-dark cycle. Knockdown of Dec2 with small interfering RNA advanced the phase of Dec1 and Dbp expression, and double knockdown of Dec1 and Dec2 had much stronger effects on the expression of the E-box-containing clock genes. These findings suggest that DEC1, along with DEC2, plays a role in the finer regulation and robustness of the molecular clock.

The mammalian molecular clock system consists of various clock genes and their protein products involved in interlocked feedback loops of transcriptional and translational regulation through clock elements such as CACGTG E-box, D-box, and ROR/REV-ERB binding elements (RORE) (18, 24). Among these regulatory sequences, the E box is thought to be the most important element in the molecular oscillatory system, since it is the binding site for the CLOCK/BMAL1 heterodimer, which up-regulates various clock genes, including Dec1, Dec2, Per1, Dbp, and Rev-erbα. In this regulatory system, PER, CRY, and DEC serve as negative factors for transcription from E-box-driven promoters, and the E-box-like element E′ box (CACGTT) was recently shown to be involved in the direct regulation of Per2 and Cry1 genes by CLOCK/BMAL1 (1, 31, 34). RORE, on the other hand, is a clock element to which the transcriptional activators—RORα/RORβ/RORγ—and repressors—REV-ERBα/REV-ERBβ—bind, and ROR and REV-ERB regulate the circadian expression of Bmal1, Clock, Npas2, and Cry1 via RORE (31).

In the mammalian clock system, DEC1 (also known as BHLHB2, STRA13, or SHARP2) and DEC2 (BHLHB3 or SHARP1) serve as transcriptional repressors for CLOCK/BMAL1-enhanced promoter activity, through binding to E boxes or interaction with BMAL1 (12, 14, 18, 26). Among suppressive factors for E boxes, DEC1 and DEC2 can bind directly to E boxes through their basic helix-loop-helix DNA binding domains (18, 26), although it remains unclear whether DEC1 and DEC2 also bind to the E′ boxes. In contrast, PER and CRY interact with the CLOCK/BMAL1 heterodimer but cannot bind directly to E/E′ boxes, since they have no DNA binding domain.

Dec1 expression shows robust circadian rhythms in the suprachiasmatic nucleus and various peripheral tissues (7, 10, 14, 17, 18, 23), and it is further enhanced by light pulse, hypoxia, or growth factors, with the Dec1 level gradually increasing during chondrogenic or osteogenic differentiation (6, 14, 15, 22, 25, 27-29). Dec1 expression can also be elevated in various tissues by refeeding after starvation (17). The remarkable responsiveness of Dec1 to these environmental factors raises the question of whether Dec1 is involved in some type of cross-talk between environmental factors and the clock system, but the precise role of DEC1 in molecular clock systems remains unknown. Expression profiles of clock genes such as Per2, Per3, Clock, Bmal1, Dbp, and Rev-erbα did not show any appreciable changes in Dec1^−/− mouse liver experiments, while 42 clock-controlled genes were affected in the Dec1^−/− liver (11). On the other hand, Clockwork Orange (cwo), a Drosophila homolog of Dec, has been shown to be indispensable for the maintenance of precise circadian rhythms of fly locomotor activity (16, 20, 21).

The object of this study was to determine precisely how DEC1 participates in circadian regulation in the presence of several other suppressive factors for the CLOCK/BMAL1 activity. Assessing the role of each individual clock gene in whole animals has proved difficult so far, since so many repressors are redundantly involved in the feedback inhibition of CLOCK/BMAL1 transactivation. In addition, metabolic status may change the expression profile of some clock genes (17). Considering the complexity of the clock system in vivo, mammalian cells in culture—showing circadian rhythms—may be a useful model for studies of DEC1. Previous studies have shown that brief treatment of various cells with 50% horse serum (or other compounds) triggers rhythmic expression of clock genes (2, 4, 5, 13, 32). In this report, we show that overexpression or deficiency of DEC1 in 50% serum-exposed cells selectively shifted the phase of several clock genes, including Dec2, Per1, Dbp, and Rev-erbα, by binding to the E boxes. And a deficiency of Dec1 and Dec2 markedly decreased the expression of several clock genes. In addition, locomotor activities of mice were affected by disruption of the Dec1 gene. These findings suggest a selective and specific role for DEC1 and DEC2 in the precise adjustment and maintenance of the mammalian molecular clock system.

MATERIALS AND METHODS

Animals.

Dec1^−/− mice were generated by inGenious Targeting Laboratory, Inc. (Stony Brook, NY). The 4.7-kb BamHI-BssSI genomic fragment of Dec1, which contains entire coding regions in exons 1 to 5, was replaced with a Neo cassette. The resulting chimeric mice were back-crossed to a C57BL/6J background for three generations. Male Dec1^−/− mice (n = 6) and their wild-type littermates (n = 6) that were 4 to 6 months of age were individually housed under conditions of cycles of 12 h of light followed by 12 h of dark (LD) (lights on 8:00 to 20:00; light intensity in the light phase, 300 lx) for 2 or 3 weeks, after which the mice were released into constant darkness (DD) for 40 days. Locomotor activities were measured using infrared thermal sensors, and spontaneous locomotor activity data obtained were analyzed using Clock Lab software (Actimetrics Inc.). The free-running period of circadian behavioral rhythms was calculated by a chi-square periodogram using activity data from day 25 to 40 in DD. For phase-shifting experiments, the same mice which had been entrained to the LD cycle were subjected to a 6-h phase-advanced shift of an LD cycle. To determine the day when the reentrainment would be completed, phase-angle differences between light-off and activity-onset times of each day after the beginning of the phase shift were calculated for 20 days and compared with mean values of those for 5 days before the phase shift. Static significance was analyzed by a Bonferroni/Dunn post hoc test. All procedures were performed in compliance with the standard principles and guidelines for the care and use of laboratory animals of Hiroshima University Graduate School of Biomedical Sciences.

Cell culture and serum shock.

Human fibroblast-like (stromal/mesenchymal stem) cells isolated from bone marrow were purchased from BioWhittaker (Walkersville, MD). Mouse embryonic fibroblasts were established from day 14.5 embryos of a Dec1^−/− mouse and a wild-type littermate. Serum shocking of human fibroblast-like cells or mouse fibroblasts was performed as described previously (5). After a 2-h incubation with 50% horse serum, the medium was replaced with Dulbecco's modified Eagle's medium-high glucose (Gibco) supplemented with 5% fetal bovine serum. The cells were harvested at the indicated times and subjected to quantitative real-time reverse transcription-PCR (RT-PCR) analysis.

Antibodies.

Rabbit antibodies against mouse DEC1 were produced by immunizing a synthetic peptide fragment of mDEC1:397-411 (Cys-Gln-Ala-Leu-Lys-Gln-Ile-Pro-Pro-Leu-Asn-Leu-Glu-Thr-Lys-Asp). The obtained antibodies were purified by affinity column chromatography. Anti-PER1 antibodies (KI044) were purchased from TransGenic Inc. Anti-CLOCK antibodies (catalog no. sc-6927) and control immunoglobulin G (normal rabbit immunoglobulin G) (catalog no. sc-2027) were from Santa Cruz Biotechnology.

Adenovirus infection.

At 24 h after seeding the cells, infection with adenovirus expressing human Dec1 (ad-Dec1) or adenovirus expressing LacZ (ad-LacZ) was performed as described previously (29). To control expression levels of overexpressed Dec1 in both human fibroblast-like cells and mouse Dec1^−/− fibroblasts, we examined the relationship between the mRNA level of Dec1 and the multiplicities of infection (MOI) of the adenovirus. The mRNA level of overexpressed Dec1 in the human cells was higher than that in the mouse cells at an MOI of 10 PFU/cell, but the levels in both cells were similar at an MOI of 50 and quite high (more than 10-fold higher) compared with the mRNA level in the noninfected human cells (see Fig. S1 in the supplemental material). Using this information, we infected human fibroblast-like cells and mouse Dec1^−/− fibroblasts with ad-Dec1 or ad-LacZ at an MOI of 50. Adenovirus carrying the LacZ gene was generously supplied by Kohei Miyazono (The University of Tokyo) (9). At 6 h after the infection procedure, the culture medium was replaced with fresh medium and cultured for 3 days before being subjected to serum shock. The high level of expression of Dec1 induced by serum shock was confirmed by quantitative RT-PCR analysis and Western blotting using anti-DEC1 antibodies.

siRNA treatment.

Small interfering RNA (siRNA) directed against Dec2 (5′-GACUAUUCCUCUUUGUAUAUG-3′ and 5′-UAUACAAAGAGGAAUAGUCCA-3′) and nonspecific control siRNA (Qiagen) were transfected into wild-type and Dec1^−/− mouse fibroblasts by use of Lipofectamine 2000 (Invitrogen). The transfection efficiencies in these cells were more than 95% as determined with Alexa 488-labeled siRNA. Overall levels of suppression of Dec2 mRNA in Dec1^−/− cells and wild-type cells were 75.8% and 74.7%, respectively, as determined by averaging Dec2 expression levels in those cells across the time points. At 72 h later, the transfected cells were subjected to serum shock for circadian rhythm detection.

RNA extraction and real-time quantitative RT-PCR.

Total RNA was extracted from cultured cells by use of an RNeasy Mini kit (Qiagen) and was subjected to real-time quantitative RT-PCR analysis with an ABI Prism 7900 sequence detection system (Applied Biosystems), using TaqMan probes and primers as follows: 5′-6-carboxyfluorescein (FAM)-TATGGACACAGACAAAGATGACCCTCATGG-6-carboxytetramethylrhodamine (TAMRA)-3′, 5′-GCTCCACTGACTACCAAGAA-3′, and 5′-CTTCCCTTGCATTTTTTATCC-3′ for human Bmal1; 5′-FAM-AGATCAACTGCCTGGACAGCATCCTCAG-TAMRA-3′, 5′-AAGCCTCCAGCTGCTCCTACCA-3′, and 5′-GAGGTTGCAGCTCTCCAGGT-3′ for human Per1; 5′-FAM-TTTGAGGTGCTGATGGTGCGCTTTG-TAMRA-3′, 5′-CTTCCGTGACCTTTCTCAGCA-3′, and 5′-GGTGCGGCTTAGGAACATCAC-3′ for human Rev-erbα; and 5′-FAM-TGACAACAGACGGCAGCATCATCTATGTG-TAMRA-3′, 5′-GGCATTAGATGGCTTCGTCATC-3′, and 5′-AAGGAGAGGTGTGATACTGTCGG-3′ for mouse Npas2. TaqMan probes and primers for human Dec1, Dec2, Per2, and Dbp and mouse Dec1, Dec2, Bmal1, Clock, Dbp, Rev-erbα, Per1, Per2, and Cry1 were described previously (10, 15, 18, 23). TaqMan probes and primers for human Cry1 and 18S RNA were obtained from Applied Biosystems. The values for mRNA levels, relative to internal control 18S RNA levels, represent the means ± standard error of the means (SEM) for three wells. Circadian rhythmicity was analyzed by one-way analysis of variance (ANOVA), and differences between the two groups were analyzed by two-way ANOVA. The peak times of mRNA levels in circadian rhythmicity were analyzed with cosinor fitting procedures, and the average period lengths from the first peak to the third peak were determined.

ChIP assays.

Chromatin immunoprecipitation (ChIP) assays of Dec1, Dec2, Per1, Dbp, and Rev-erbα E boxes and the Per2 E′ box were performed using a ChIP assay kit (Upstate) according to the manufacturer's instructions. DNA samples recovered from human fibroblast-like cells were subjected to PCR using gene-specific primers as follows: 5′-AGCTAGCAAGGGGATATTCC-3′ and 5′-TTGGAGCTACGTGTTCTACC-3′ were used for Dec1, 5′-GCCTTTGGGAAGATGAATAGC-3′ and 5′-AAGTGTGAAGCAGTTGGTCC-3′ for Dec2, 5′-AAAGAGGCGCAGCGTATCTC-3′ and 5′-TTGCATAATGCCGGGCACTG-3′ for Per1, 5′-TTTGGGCACATTGAGGTCTC-3′ and 5′-CTGTACCTTTAAGCACCAGC-3′ for Dbp, 5′-CTCGTTACATAATGAGCTCC-3′ and 5′-CAGGAATGGCTCCATGTTAC-3′ for Rev-erbα, and 5′-ATGTAAAAAGAGCGACGGGC-3′ and 5′-AGCAGCCCAAGGAACTTCC-3′ for Per2. Obtained PCR products were subjected to agarose gel electrophoresis. The experiments were repeated two to four times, and similar results were obtained each time.

Plasmid constructions.

Oligonucleotides of three tandem sequences containing human Per2 CACGTT E′ boxes (underlined) with 6 bp of flanking sequences (CGCGGTCACGTTTTCCAC) were subcloned into pGL3-TK (30). Construction of pGL3-TK containing three E boxes in the Dec1 and Per1 promoter and expression vectors for Clock, Bmal1, and Dec1 has been described previously (14, 18). Cry1 expression vector was generously supplied by Masaaki Ikeda, Saitama Medical School.

Transient transfection and luciferase reporter assay.

NIH 3T3 cells were seeded at 3 × 10⁴ cells per 16-mm-diameter well 24 h before transfection. Luciferase reporter plasmid (2 ng per well) and mouse Clock expression and Bmal1 expression vectors (50 ng each per well) were cotransfected with or without Dec1 or Cry1 expression vectors (5 or 10 ng per well) by use of Lipofectamine 2000. As an internal standard, 0.05 ng of pRL-SV40 (Promega) was cotransfected. The total concentration of DNA was adjusted to 112.05 ng per well with an empty vector (pcDNA3.1/Zeo; Promega). After incubation for 4 h, the medium was replaced with fresh medium. The cells were incubated for a further 48 h and subjected to a luciferase reporter assay using a dual-luciferase reporter assay system (Promega). Luciferase activities were normalized by internal control activities. The values represent the means ± SEM of the results obained for four wells. Statistical significance was analyzed by a Student t test.

Electrophoretic mobility shift assays (EMSA).

Human DEC1 protein was synthesized using TNT-coupled reticulocyte lysate systems (Promega). The double-stranded oligonucleotides of the DEC1 E box (5′-agCGTTGTCCAACACGTGAGACTCATctca-3′) and the Per2 E′ box (5′-CGCGCGCGGTCACGTTTTCCACTATGctca-3′) were labeled using [³²P]dCTP (GE Healthcare) and a DNA polymerase I Klenow fragment (New England Biolabs). Synthesized human DEC1 protein was incubated with approximately 40 fmol (20,000 cpm) of ³²P-labeled probe for 10 min at room temperature in 15 μl of 10 mM Tris-HCl (pH 8.0)-0.5 mM dithiothreitol-10% glycerol-1 μg of poly(dI-dC)-50 mM NaCl-5 mM MgCl₂, and then the mixtures were subjected to 5% polyacrylamide gel electrophoresis in 12 mM Tris-HCl-125 mM glycine-1 mM EDTA electrophoresis buffer. In competition experiments, a 25-, 50-, or 100-fold molar excess of unlabeled Dec1 E box or Per2 E′ box was added.

RESULTS

Rhythmic expression of Dec1 and its profile of binding to CACGTG E box in living cells.

Since brief treatment of various cells with 50% horse serum triggers rhythmic expression of clock genes in culture (5), we used this technique to determine whether an expression profile of Dec1 mRNA would have circadian rhythmicity in human fibroblast-like cells (bone marrow stromal cells). In these cells, the Dec1 mRNA levels were higher than those in human skin fibroblasts (data not shown). The expression pattern of Dec1 after 50% serum treatment for 2 h showed a robust circadian rhythm, with peaks at approximately 20 h and 44 h after the start of serum stimulation (Fig. 1A). The circadian mRNA profile was similar to that of Per1, with the peak observed at the same phase.

FIG. 1.

Rhythmic expression of Dec1 and Per1 mRNA and binding of DEC1 and PER1 protein to CACGTG E-box in human fibroblast-like cells. (A) Rhythmic expression of clock genes (including Dec1 and Per1) was induced by treatment with 50% horse serum for 2 h. Total RNA of the induced cells was extracted every 4 h and subjected to quantitative RT-PCR analysis of Dec1 and Per1 mRNA expression. (B) To observe the time course of DEC1, PER1, and CLOCK protein binding to E boxes in the Dec1 and Per1 promoters, the same cells were subjected to ChIP assays using anti-DEC1, anti-PER1, and anti-CLOCK antibodies. The extracted DNAs were amplified with primers specific for human Dec1 or Per1 and subjected to agarose gel electorphoresis. Locations of E boxes and PCR primers in the Dec1 and Per1 genes are shown. (C) ChIP assays of DEC1 binding to the E boxes of various clock genes in human fibroblast-like cells infected with Dec1-overexpressing or LacZ-overexpressing cells. After the serum shock, cells were harvested at the indicated times and subjected to ChIP assays using anti-DEC1 antibodies. The extracted DNAs were amplified with primers specific for human Dec1, Dec2, Per1, Dbp, and Rev-erbα. Locations of the E box and PCR primers in the Dec2, Dbp, and Rev-erbα genes are also shown. (D) Transcriptional induction of Dec1 mRNA was examined at up to 12 h after the serum shock. (E) By use of Western blot analysis, expression levels of DEC1 protein in cells infected with ad-Dec1 or ad-LacZ were determined before and after the serum shock. (F) ChIP assays of CLOCK binding to the E box and E′ box in Dec1-overexpressing cells and control LacZ-expressing cells. The samples precipitated with anti-CLOCK antibodies were amplified with primers specific for human Dec1 and Per2. Locations of the E′ box and PCR primers in the Per2 gene are also shown. IgG, immunoglobulin G.

Although DEC1 is a transcriptional repressor that binds to the E boxes of some clock genes in EMSA (12, 14, 18), the status of its binding to the E boxes in living cells had not yet been demonstrated. Therefore, we examined the profile of DEC1 binding to the CACGTG E boxes in Dec1 and Per1 promoters by use of ChIP assays with anti-DEC1 antibodies. The results revealed that DEC1 bound to the E boxes in the fibroblast-like cells and that the DEC1 binding profile exhibited circadian rhythmicity. The binding levels of DEC1 protein were maximal when the mRNA levels of Dec1 and Per1 began to decrease (at 24 and 48 h after serum shock) (Fig. 1B). In contrast, binding of PER1 protein to the same E boxes was maximal when the expression levels of Dec1 and Per1 began to increase (at 36 h after serum shock). These findings demonstrated that DEC1 and PER1 bind to the E boxes in almost opposite phases of the circadian rhythms. On the other hand, profiles of binding of CLOCK to the E boxes in the Dec1 and Per1 promoters did not show significant circadian rhythmicity. The absence of circadian rhythmicity in CLOCK binding to E boxes in human fibroblasts corresponds to the previous observation in mice liver by Lee et al. (19).

Binding of DEC1 to E boxes in the various clock genes in Dec1-overexpressing and control cells.

DEC1 binding to E boxes of various clock genes in human cells infected with ad-Dec1 or control LacZ was also examined. The DEC1 binding to E boxes of Dec1, Dec2, Per1, Dbp, and Rev-erbα was observed by ChIP analysis using anti-DEC1 antibodies and specific primers for each gene. The DEC1 binding to the E boxes in the control LacZ-expressing cells markedly increased within 2 h after serum shock and started to decrease at 4 h (Fig. 1C). The level of DEC1 protein that bound to the E boxes declined to a minimal level at 12 h, while the DEC1 protein binding level in Dec1-overexpressing cells maintained high levels for at least 12 h, irrespective of the presence or absence of serum stimulation. The profile of DEC1 binding to the E boxes in LacZ-expressing cells correlated with the changes in the expression level of Dec1 mRNA after serum stimulation (Fig. 1D), and the enhanced binding of DEC1 protein to the E boxes in Dec1-overexpressing cells coincided with constitutively increased Dec1 protein levels in the cells (Fig. 1E). To investigate the effect of overexpressed DEC1 on the binding of CLOCK/BMAL1 heterodimer to the E box, we performed ChIP assays using anti-CLOCK antibodies. DEC1 overexpression suppressed—but did not abolish—the binding of CLOCK to the E box in the Dec1 promoter (Fig. 1F). However, it had little effect on CLOCK binding to the Per2 E′ box.

Effect of Dec1 overexpression on circadian rhythms of clock gene expression.

Human fibroblast-like cells overexpressing Dec1 (ad-Dec1) or control LacZ (ad-LacZ) were exposed to 50% serum for 2 h, after which total RNA was extracted every 4 h. Expression profiles of clock genes—including Dec1, Dec2, Per1, Dbp, Rev-erbα, Bmal1, Per2, and Cry1—showed robust circadian rhythmicity in both Dec1-overexpressing and LacZ-expressing cells (one-way ANOVA; P < 0.01) (Fig. 2). However, Dec1 overexpression delayed the phases of circadian profiles of endogenous Dec1 and Dec2 mRNA for 6.3 and 5.6 h, respectively, relative to those seen with control cells, with first peak around 24 h after serum shock (Fig. 2 and Table 1), and the period lengths of Dec1 and Dec2 in Dec1-overexpressing cells (25.8 and 25.6 h) were about 2 h longer than those seen with control cells (23.9 and 23.8 h). The overexpression also delayed the peak phases of Per1, Dbp, and Rev-erbα—which carry E boxes in their regulatory regions—for 2.3, 3.3, and 2.3 h, respectively, at the first peak, and the period lengths of Per1, Dbp, and Rev-erbα were 25.4, 25.5, and 25.8 h, respectively, for the Dec1-overexpressing cells and 23.8, 23.7, and 24.0 h, respectively, for the control cells. Furthermore, overexpression of Dec1 reduced the expression levels of Dec1, Dec2, and Dbp, probably because of the continuous interaction of DEC1 with E boxes. In contrast, overexpression of Dec1 had little effect on the initial phases of circadian expression of Per2, Cry1, and Bmal1, which do not contain conserved E boxes in their regulatory regions. However, the period lengths of Per2 and Cry1 expression were also lengthened by Dec1 overexpression. The lengths were 25.5 and 24.0 h, respectively, for the Dec1-overexpressing cells and 23.9 and 23.0 h, respectively, for the control cells, resulting in an apparent phase delay at the third or fourth peak. In addition, Dec1 overexpression increased the expression level of Bmal1 mRNA.

FIG. 2.

The phase delays of circadian expression of E-box-containing clock genes caused by Dec1 overexpression after serum shock. Human fibroblast-like cells were infected with ad-Dec1 (red lines) or ad-LacZ (black lines). Three days later, the cells were incubated with 50% horse serum for 2 h, and total RNA was extracted from these cells at the indicated times after the start of serum shock and subjected to quantitative RT-PCR analysis. Expression profiles of all examined clock genes showed significant circadian rhythms (one-way ANOVA; P < 0.01). Immediate-early induction of Dec1 and Per1 expression is indicated by asterisks. Numbers in parentheses represent differences between the two groups (two-way ANOVA; P < 0.01, P < 0.05, or P > 0.05).

TABLE 1.

Effect of overexpression (ad-Dec1) and deficiency (Dec1^−/−) of Dec1 and knockdown (siRNA) of Dec2 on circadian expression of clock genes at the first peak after serum shock^a

Gene expressed	% Phase difference change (h) and up- or down-regulation result for indicated treatment and cell category					Presence or absence of conserved E box (CACGTG)
	ad-Dec1		Dec1 KO (Dec1+/+ [mouse])	Dec2 siRNA
	Dec1+/+ (human)	Dec1−/− (mouse)		Dec1+/+ (mouse)	Dec1−/− (mouse)
Dec1	+6.3, ↓↓	ND	ND	−3.9, ↑	ND	Present
Dec2	+5.6, ↓	+3.6, ↓	−2.8, ↑	ND	ND	Present
Per1	+2.3	+2.9	−2.0	n.s.	−3.3, ↓	Present
Dbp	+3.3, ↓	+1.9	−2.5	−1.0	−1.3, ↓↓	Present
Rev-erbα	+2.3	+1.8	−2.6	−1.5	−2.0, ↓	Present
Per2	n.s.	n.s.	n.s.	n.s.	−1.1	Absent
Cry1	n.s.	n.s.	n.s.	n.s.	n.s.	Absent

^aHuman fibroblast-like cells and mouse wild-type (Dec1^+/+) or Dec1^−/− fibroblasts were infected with ad-Dec1 or transfected with Dec2 siRNA. Differences in clock gene expression profiles for the two groups in the period from 8 h to 36 h after serum shock were analyzed by two-way ANOVA. Values represent the phase difference (in hours) between the gene expression rhythms of the experimental and control groups at the first peak. Positive and negative values indicate advanced and delayed phase differences from the circadian peak of each control gene, respectively. KO, knockout; n.s., no significant difference; ↓, lesser down-regulation (20 to 40% decrease of area under the curve) of average expression level; ↓↓, greater down-regulation (>40%) of average expression level; ↑, up-regulation (20 to 40%) of average expression level; ND, not determined.

The phase advance of E-box-containing clock gene expression rhythms in Dec1^−/− cells.

If the phase delay of the clock gene expression caused by Dec1 overexpression is indeed related to the physiological role of DEC1, DEC1 deprivation should have the opposite effect on the phase of clock gene expression. To test this hypothesis, we isolated fibroblasts from embryos of Dec1^−/− and wild-type mice. Dec1 expression in the Dec1^−/− fibroblasts was abolished, whereas it showed a robust circadian rhythmicity in wild-type fibroblasts after serum shock (Fig. 3A). The expression profiles of Dec2, Per1, and Dbp were significantly different in the Dec1^−/− fibroblasts, with the first peaks detected 2.8, 2.0, and 2.5 h, respectively, earlier than those of wild-type (Dec1^+/+) cells. Interestingly, long-term observation revealed that the phase advance in the circadian expression rhythms of Dec2, Per1, and Dbp was gradually reduced each circadian cycle, and finally no differences were detected at the fourth cycle. Accordingly, the period lengths of Dec2, Per1, and Dbp circadian expression in the Dec1^−/− cells were longer than in the wild-type cells. The lengths of Dec2, Per1, and Dbp were 24.7, 25.2, and 25.0 h, respectively, in the Dec1^−/− cells and 23.6, 24.4, and 24.0 h, respectively, in the wild-type cells. In contrast, Dec1 deficiency had little effect on the phases and periods of Per2 and Cry1 circadian expression.

FIG. 3.

Effect of Dec1 deficiency on E-box-containing clock genes and effects of Dec1 overexpression in Dec1^−/− cells. (A) Dec1^−/− (blue lines) and wild-type (Dec1^+/+; black lines) fibroblasts obtained from a Dec1^−/− mouse and a wild-type littermate were treated with 50% horse serum. The mRNA expression levels (means ± SEM; n = 3) of Dec1, Dec2, Per1, Dbp, Per2, and Cry1 were determined every 4 h by quantitative RT-PCR analysis. Immediate-early induction of Dec1 and Per1 expression is indicated by asterisks. (B) Dec1^−/− fibroblasts were infected with ad-Dec1 (red lines) or ad-LacZ (blue lines) and subjected to serum shock for circadian expression of clock genes. The mRNA levels were determined. Numbers in parentheses represent differences between the two groups (two-way ANOVA; P < 0.01, P < 0.05, or P > 0.05).

In the next experiment, we looked at whether the phase advance observed in Dec1^−/− cells could be restored by Dec1 overexpression. Overexpression of human Dec1 in mouse Dec1^−/− cells phase-delayed the expression rhythms of Dec2, Per1, and Dbp for 3.6, 2.9, and 1.9 h, respectively, at the first peak and significantly reduced the expression level of Dec2 (Fig. 3B and Table 1). In contrast, the phases of Per2 and Cry1 expression rhythms were not altered. These findings proved that DEC1 is indeed involved in control of the circadian phase and amplitude of the E-box-containing clock genes, with weaker effects on those of E′-box-containing clock genes.

Effects of Dec2 knockdown in Dec1^−/− cells on circadian expression of clock genes.

To determine whether DEC2 has similar effects on the expression of clock genes, siRNA was used to knock down Dec2 expression. Treatment with siRNA against Dec2 resulted in a phase advance of circadian expression rhythm of Dec1 and Dbp for 3.9 and 1.0 h, respectively, at the first peak and slightly increased the expression level of Dec1 (Fig. 4A and Table 1), although it did not change the expression of Per1 and the second peak time of Dbp. In addition, transfection of Dec2 siRNA into Dec1^−/− cells markedly decreased the amplitude of Dbp and Per1 circadian expression and further advanced their phases by 1.3 and 3.3 h, respectively (Fig. 4B and Table 1), whereas it had only weak effects on the phases and expression levels of Per2 and Cry1 (Table 1). The disruption of both Dec1 and Dec2 expression had a much greater effect on circadian expression of Dbp and Per1 than that of either Dec1 or Dec2 alone, and knockdown of both Dec1 and Dec2 decreased the expression levels of Rev-erbα, Bmal1, Clock, and Npas2 (Fig. 4C and Table 1).

FIG. 4.

Effect of Dec2 knockdown on circadian expression of clock genes. (A) Dec2 (green lines) or nonspecific control (black lines) siRNA was transfected into wild-type (Dec1^+/+) fibroblasts. At 72 h later, the cells were treated with 50% horse serum. The mRNA expression levels (means ± SEM; n = 3) of Dec2, Dec1, Per1, and Dbp were determined by quantitative RT-PCR analysis. Immediate-early induction of Dec1 and Per1 expression is indicated by asterisks. (B) Expression levels (means ± SEM; n = 3) of Per1 and Dbp in Dec1^−/− fibroblasts transfected with Dec2 (green lines) or control (blue lines) siRNA were determined. (C) Expression levels (means ± SEM; n = 3) of Bmal1, Clock, and Npas2 in Dec1^−/− fibroblasts transfected with Dec2 siRNA (green lines) were compared with those in wild-type (Dec1^+/+) cells transfected with control siRNA (black lines). Numbers in parentheses represent differences between the two groups (two-way ANOVA; P < 0.01, P < 0.05, or P > 0.05).

Stronger direct action of DEC1 on E box than on E′ box.

Although the circadian expression of the Per2 gene is regulated by CLOCK/BMAL1, the gene does not contain the CACGTG E box in its regulatory region. The elements responsible for the regulation are CACGTT E′ boxes. Since the effect of Dec1 overexpression or deficiency on the circadian expression of Per2 was much less than that on CACGTG E-box-carrying clock genes, we compared the effects of DEC1 on transcriptional activities from E-box- and E′-box-carrying promoters. In luciferase reporter assays, DEC1 abolished the CLOCK/BMAL1-enhanced transcriptional activities of Dec1 and Per1 E boxes, whereas it had far less effect on the promoter activity of the Per2 E′-box-containing construct (Fig. 5A). In contrast, CRY1 abolished the CLOCK/BMAL1-enhanced transcriptional activity of both the Per2 E′ box and the Dec1 and Per1 E boxes. Accordingly, in ChIP analysis the Per2 E′-box region did not coprecipitate with anti-DEC1 antibodies in either the presence or the absence of the stimulation by 50% serum, but the E-box regions in Dec1 and Per1 promoters did coprecipitate with anti-DEC1 antibodies (Fig. 5B). On the other hand, both the Per2 E′ box and the E boxes of Dec1 and Per1 coprecipitated with anti-PER1 antibodies 2 h after serum stimulation. In EMSA, DEC1 bound to both the E box and the E′ box, but the intensity of the shifted band of the E box associated with DEC1 protein was much greater than that of the E′ box (Fig. 5C). In addition, unlabeled Dec1 E-box oligonucleotides competed with a ³²P-labeled Dec1 E box or Per2 E′ box more effectively than unlabeled Per2 E′-box oligonucleotides did. These findings indicate that the binding affinity of DEC1 for the Dec1 E box is higher than that for the Per2 E′ box. The stronger binding of DEC1 to E boxes probably selectively delayed the circadian phases of clock genes containing E boxes, whereas PER1 indirectly bound to both the E box and the E′ box via protein-protein interaction with CLOCK/BMAL1 after posttranslational modification.

FIG. 5.

Comparison of clock gene regulation profiles of E-box-containing clock genes such as Dec1 and Per1 genes to that of the E′-box-containing Per2 gene. P values were calculated by using Student t tests (**, P < 0.01). (A) Luciferase reporter assays of reporter genes containing three tandem copies of the Dec1 or Per1 E box or the Per2 E′ box were performed as described in Materials and Methods. The reporter constructs were cotransfected with Clock and Bmal1 expression vectors (50 ng each) together with various amounts of Dec1 or Cry1 expression vectors as indicated. Luciferase activities (means ± SEM; n = 4) were measured after a 48-h incubation. (B) ChIP assays of human fibroblast-like cells by use of anti-DEC1 and anti-PER1 antibodies were performed before, and 2 h after, serum shock. (C) Binding of DEC1 to the Dec1 E box (CACGTG) and the Per2 E′ box (CACGTT) was examined by EMSA. The ³²P-labeled Dec1 E-box (lanes 1 to 7) or Per2 E′-box (lanes 8 to 14) probe was incubated with human DEC1 protein synthesized by an in vitro transcription/translation system. Competition experiments were performed using a 25-, 50-, or 100-fold molar excess of unlabeled Dec1 E box (lanes 2 to 4 and 9 to 11) or Per2 E′ box (lanes 5 to 7 and 12 to 14). Arrows indicate the shifted bands of radiolabeled Dec1 E box or Per2 E′ box bound to DEC1 protein. IgG, immunoglobulin G.

Effect of Dec1 deficiency on circadian behavioral rhythms of mice.

To clarify the role of DEC1 in vivo, we compared the circadian phenotype of Dec1^−/− mice with that of wild-type mice by monitoring their spontaneous locomotor activities under DD as well as LD conditions. Dec1^−/− mice showed robust activity rhythms in LD, with phases of activity onset and end not significantly different from those of wild-type mice (data not shown), but the average circadian period in DD was 24.10 ± 0.05 h (mean ± SEM; n = 6), which was significantly longer than that of wild-type mice (23.95 ± 0.03 h; n = 6 [P < 0.05; Student t test]) (Fig. 6A and B). We also subjected Dec1^−/− and wild-type mice to a simulated jet lag experiment by phase-shifting an LD cycle. Following a 6-h phase-advancing shift of the LD cycle, behavioral rhythms of Dec1^−/− mice showed gradual phase advance and were reentrained on day 7 after the phase shift, while it took 10 days for the wild-type mice to reach complete reentrainment after the 6-h phase advance of the LD cycle (Fig. 6C and D; also see Table S1 in the supplemental material). The activity onset times of Dec1^−/− mice from day 3 to day 9 after the phase shift were significantly earlier than those of wild-type mice (Fig. 6D). Thus, the reentrainment of behavioral rhythms in Dec1^−/− mice was significantly faster than that in wild-type mice.

FIG. 6.

Circadian phenotypes of Dec1^−/− mice. (A) Representative double-plotted locomotor activity of a wild-type mouse (Dec1^+/+) and a Dec1^−/− mouse. Mice were maintained in LD for the first 12 days of the record and then released into DD. A gray background in the right half of the depiction of activity data indicates the dark periods. Black and gray horizontal bars at the top of each activity record indicate the dark and light phases in LD. (B) Distribution of the free-running periods of wild-type (+/+) and Dec1^−/− mice. Closed circles indicate individual periods, and open boxes indicate the average period (± SEM) for a genotype. P values were calculated by using Student t tests (*, P < 0.05). (C) Effects of a 6-h phase advance on activity onset. Representative double-plotted activity data for a wild-type mouse (Dec1^+/+) and a Dec1^−/− mouse before and after the phase-advance shift of an LD cycle by 6 h. Open arrows indicate the beginning of the phase advance. Black and white horizontal bars at the top of each panel depicting activity data indicate dark and light phases of an LD cycle, respectively, before and after the phase shift. (D) Phase shifts of activity onset times before and after a 6-h phase advance of an LD cycle. The onset times from day 1 to day 22 were significantly different between Dec1^−/− and wild-type mice (two-way ANOVA; P < 0.01); the onset phases were significantly earlier from day 3 to day 9 for Dec1^−/− mice compared with wild-type control results (*, P < 0.05; **, < 0.01 [Bonferroni/Dunn post hoc test]). Open and closed arrowheads indicate the days when reentrainment was completed (day 7 for Dec1^−/− mice; day 10 for wild-type mice). The activity onset time under LD entrainment was slightly earlier (by 0.2 ± 0.1 h) than light-off time for Dec1^−/− mice, a result which was not significantly different from that seen with wild-type mice (0.2 ± 0.2 h).

DISCUSSION

In the present study, we demonstrated that Dec1 overexpression in both human and mouse cells caused a phase delay in circadian rhythms of the expression of CACGTG E-box-containing genes such as Dec1, Dec2, Per1, Dbp, and Rev-erbα (Fig. 2 and 3B and Table 1) in the first cycle after serum shock but that Dec1 deficiency advanced the circadian phase of E-box-containing genes (Fig. 3A and Table 1). In addition to these phenomena, Dec1 overexpression decreased the expression level or amplitude of Dec1, Dec2, and Dbp, whereas Dec1 deficiency increased the expression of Dec2 mRNA. Thus, overexpression and deficiency of Dec1 had virtually opposite effects on the phase and expression levels of clock genes. On the other hand, Dec1 overexpression had less effect on the Per2 and Cry1 circadian expression driven by CACGTT E′ boxes (Fig. 2 and 3B and Table 1) than that on the E-box-containing genes. Supporting this observation, luciferase reporter assays showed a lower inhibitory effect of DEC1 on the Per2 promoter, and ChIP and EMSA analyses confirmed the weaker binding of DEC1 to E′ box of Per2 (Fig. 5). DEC1 binding to E′ box in living cells may be too weak to form a stable complex in ChIP analysis or to cause the phase shift of Per2 or Cry1 expression in the initial peak. Furthermore, Dec1 overexpression suppressed CLOCK binding to the E box but did not affect CLOCK binding to the E′ box (Fig. 1F). Thus, circadian rhythms of E-box-containing and E′-box-containing genes were dissociated within the same cells by overexpression or suppression of Dec1, although the dissociated phases in Dec1^−/− cells may resynchronize after three or more circadian cycles.

The direct binding affinity of DEC1/DEC2, but not PER/CRY, for E-box sequences is the most intriguing difference between DEC and PER/CRY as negative components of the molecular clock. We found here that DEC1 action is selective for the clock genes containing E boxes, while PER/CRY can suppress the expression of both E-box- and E′-box-containing clock genes. Dec1 overexpression thus resulted in phase shifts of all of E-box-containing genes examined in this study, whereas forced expression of Per2 at a high level almost abolished the circadian rhythms of endogenous Per2 and Dbp expression in NIH 3T3 cells and, at a relatively low level, reduced the amplitude of the gene expression but had little effect on the circadian phase (33). Among a number of suppressive factors—DEC1, DEC2, PER1, PER2, PER3, CRY1, and CRY2—some may have redundant functions, and some distinct ones, in the molecular clock system. These findings indicate that DEC1/DEC2 and PER/CRY have distinct roles derived from discrete selectivity characteristics with respect to the E boxes and E′ boxes.

Dec1 overexpression markedly enhanced DEC1 binding to the E boxes of various clock genes (Fig. 1C), whereas these clock genes in Dec1-overexpressed cells still displayed rhythmic expression. In agreement with these observations, Dec1 overexpression suppressed—but did not abolish—CLOCK binding to the E box (Fig. 1F). A CLOCK/BMAL1 heterodimer and DEC1 may compete for the E box, depending upon the equilibrium between these transcription factors. Even in the absence of appropriate rhythms of DEC1 action, the PER2 and CRY1 loops (Fig. 7) seem virtually intact, a state which could be sufficient to maintain rhythmicity.

FIG. 7.

A model of DEC and PER functions. DEC1 directly binds to the CACGTG E boxes of clock genes and causes a phase delay in their circadian rhythms, whereas, because of very weak binding to the CACGTT E′ box, it does not change the circadian phase of Per2 expression. The effects of DEC on Dec1 and Dec2 expression (DEC loop) are much stronger than the effects on Per1 and Dbp expression, while the suppressive effect of PER on Per1 and Per2 expression (PER/CRY loop) seems to be stronger than that on Dec1 and Dec2 expression.

DEC1 may modulate the circadian phases of some clock genes without influence on amplitude of the rhythmicity. Dec1 expression is known to be induced by various exogenous or endogenous stimuli—including light pulses, hypoxia, growth factors, and feeding. And since different tissues respond to different environmental signals for their tissue-specific functions, DEC1 may constitute a distinct input pathway through E-box-containing clock genes such as Dec1, Dec2, Per1, and Dbp for modulation of circadian clocks in tissue-specific manners.

In the present study, free-running periods of behavioral rhythms of Dec1^−/− mice in DD were slightly but significantly longer than those of wild-type mice (Fig. 6A and B). In addition, Dec1^−/− mice reentrained to a 6-h phase advance of an LD cycle faster than wild-type mice did (Fig. 6C and D). Accordingly, knockdown of Clockwork Orange, a Drosophila homolog of Dec, showed circadian behavioral rhythms with 2- to 3-h-longer periodicities than the wild type (16, 20, 21); Npas2^−/− or Rorα^+/− mice also showed faster reentrainment following a 4-h advance of an LD cycle (3, 8), possibly because of weakened clock controls. Both Dec and Clockwork Orange are therefore necessary for precise control of circadian rhythms, although the effect of Dec1 deficiency on the behavioral rhythms of mice seems to be less robust than that of Clockwork Orange for Drosophila species (16, 20, 21). However, this difference may be derived from overlapping functions of Dec1 and Dec2 in maintaining and entraining circadian rhythms. Drosophila species seem to have only one Clockwork Orange.

The variation of the circadian phases caused by Dec1 overexpression among E-box-containing clock genes was unexpected. Dec1 overexpression had its greatest effect on the phases of Dec1 and Dec2, a moderate effect on those of Per1 and Dbp, and the least effect on that of Rev-erbα (Fig. 2 and Table 1). Thus, DEC1 appears to be more effective in regulating its own feedback loop (DEC loop) than in regulating the interlocked loop with PER and CRY (Fig. 7). Similarly, PER and CRY may be more effective in their own feedback loop (PER/CRY loop) than in the interlocked loops. Per2 overexpression had a greater suppressive effect on endogenous Per2 expression than on Dbp expression (33). The varying effectiveness of DEC1 with respect to Dec1/Dec2, Per1/Dbp, and Rev-erbα expression in living cells may be accounted for by the presence of different numbers of E boxes at different positions, different binding affinities of DEC1 for the E boxes with different surrounding sequences, and/or different numbers of other clock elements.

Gene expression patterns of Dec1 and Per1 after serum shock were remarkably similar (Fig. 2), but we found that the patterns of timing of DEC1 and PER1 binding to E boxes were different (Fig. 1B). DEC1 may rapidly translocate into the nucleus and bind to E boxes after its transcriptional induction, whereas there is a long time lag before the occurrence of PER1 binding to E boxes via protein-protein interaction with a CLOCK/BMAL1 heterodimer after elevated transcription of Per1.

Unexpectedly, Bmal1, Clock, and Npas2 expression was down-regulated by the knockdown of both Dec1 and Dec2 (Fig. 4C and Table 1). The mechanism underlying the decreases in Bmal1, Clock, and Npas2 levels in the Dec1 and Dec2 doubly deficient cells is not yet known, but Ror mRNA levels were also reduced in the cells (data not shown). Since RORs are transcriptional activators for Bmal1, Clock, and Npas2, reduction of Ror expression is assumed to induce down-regulation of Bmal1, Clock, and Npas2.

In conclusion, our findings indicate a physiological role for DEC1 in the circadian rhythms of clock gene expression in living cells. DEC1 altered the circadian phase of various clock genes through its direct interaction with E boxes. The degree of the phase delay induced by Dec1 overexpression was most prominently displayed in results associated with the circadian expression of Dec2, and forced expression of Dec1 had little effect on the expression of E′-box-containing genes, suggesting that the DEC loop is partially separated from the PER/CRY loop. We also found that DEC2 had similar effects on E-box-containing clock genes and that the double knockdown of Dec1 and Dec2 markedly decreased the amplitude of the circadian expression of several clock genes, emphasizing the essential role of the DEC loop in the molecular clock system.