Circadian rhythms, periodicities with an approximate 24-h length, are essential physiological functions in almost all organisms on the earth and are generated by endogenous biological clocks. The biological clocks consist of 3 components: input, pacemaker, and output. The endogenous period length is not exactly 24 h and must be entrained by light. In mammals, other entrainment cues such as food intake or temperature, even social cues, can reset the circadian rhythm. The central oscillator in mammals is located in the suprachiasmatic nuclei (SCN) in the anteroventral hypothalamus. Recent studies revealed that the peripheral clocks reside not only in other brain regions but also in peripheral organs and even in cultured cells 1234. The output of the biological clocks includes broad physiological phenomena such as locomotor activity, sleep-wake cycle, hormonal secretion, cardiovascular condition, bowel movement, and even mental states. The disturbance of the biological clocks may cause many dysfunctions from routinely experienced jet-lag or work-shift problems to sleep disturbance and cardiovascular, metabolic or mental diseases 25.

Recent molecular biological studies have revealed that the molecular mechanism of the biological clocks is based on the interlocked loops of transcriptional and translational feedback 678. Among the transcriptional controls in mammals, 3 major clock components for canonical clock genes, i.e., E-box, RORE, and DBPE, have been reported so far 39. The E-box (CACGTG), for example, in the Per1 promoter, is the most well-known regulatory element. It is considered the binding site for the heterocomplex of CLOCK/BMAL1, which is a positive regulatory element. Negative regulatory elements, PERs and CRYs, restrain the transcriptional activity 101112. It is widely known that not only other clock genes are regulated by the E-box but also output regulators or clock-controlled genes are regulated by it as well 1314151617. On the other hand, Bmal1, whose circadian expression exhibits a sort of anti-phase to Per2 expression, is positively regulated by Rorα and negatively regulated by Rev-erbα, through RORE 18192021.

Per2 was literally cloned as a second mammalian period gene 2223, however, gene-knockout analysis revealed a more prominent role for mPER2 in the mammalian clock than PER1 24. Using the in vitro rhythm oscillation monitoring system (IV-ROMS) 325, instead of the canonical E-box (CACGTG), we previously identified an E-box-like sequence (CACGTT) in the Per2 promoter region in vitro 26, which is essential for the cell-autonomous transcriptional oscillation of Per2 as well as for Per2 circadian oscillations in vivo 27. For the cell-autonomous transcriptional oscillation of Per2, not only this E-box-like sequence but also the downstream region is involved in the circadian oscillation of Per2 26. Here, we identified closely spaced E-box or E-box-like elements in the regulatory regions of canonical clock genes including Per2. Our comprehensive approach raised the possibility that not just a single E-box but a direct repeat of E-box-like elements is required for circadian oscillation of core clock and clock-controlled genes.

Our previous study demonstrated that an E-box-like sequence (CACGTT) and its downstream region are essential for transcriptional oscillation of Per2, a crucial component of molecular clocks 26. Comparing with other mammalian genome sequences, within the conserved regions among mammalian (human, rhesus, rat, mouse, and opossum) genomes, we found that an E-box-like sequence followed by another non-canonical E-box with 6-base pairs (bp) between the two (Fig. 1A). In addition to Per2, we found a similar E-box or E-box-like sequence followed by another E-box-like sequence with an intervening space of 6 bp in Per1 and Per3, which are other mammalian period homologues, in their upstream promoter region close to the transcription start site (TSS). The E-box or E-box-like sequences distal and proximal to TSS were dubbed E1 and E2, respectively, and collectively as the "EE-element" (Fig. 1A). These sequences are well conserved among diverse mammals, such as human, rhesus, rat, and mouse, and even the opossum which is evolutionally located at the farthest branch from humans among the mammals. To examine whether these EE-elements are involved in transcriptional oscillation, we performed the IV-ROMS established in our previous study by using the following constructs: The oligonucleotides of 3 (hPer1, hPer2, and hPer3) EE-elements were inserted upstream of the SV40 promoter in a luciferase (luc) plasmid, and the constructed plasmids were used to transfect mouse NIH3T3 cells. All 3 EE-elements gave significant oscillation of luciferase activity but the amplitudes were different among the genes (Fig. 1B). The hPer2 EE-element displayed the most robust oscillation, whereas the hPer1 EE-element gave the lowest amplitude, which correlated with the patterns derived from luc-constructs containing the longer promoter regions 26. The amplitude in circadian oscillation seemed to correlate with the number of mismatches in the E2 sequence. In fact, a mutant hPer3 EE-element, in which the GAG at the end of the E2 sequence was replaced with canonical GTG, exhibited a higher amplitude than the wild-type hPer3 EE-element (see Additional file 1).

To investigate whether both E1 and E2 were necessary for generating the cell-autonomous oscillation, we again performed IV-ROMS with the use of the mutant hPer2 EE-element constructs. Not only the double E-box-like mutations (E1mE2m) but also the single E-box-like mutation (E1m or E2m) resulted in drastically reduced circadian transcription or led to an almost complete loss of transactivation (Fig. 2). A longer Per2 promoter (~2 kb) with double E-element mutants showed that the amplitude was heavily reduced, with circadian transcription hardly being observed (see Additional file 2). These results are similar to those obtained with mutations of 2 ROREs in the Bmal1 promoter 21 and suggest that these 2 E-box-like elements function in Per2 transactivation.

To see whether endogenous clock transcription factors, BMAL1 and CLOCK, could bind to the EE-element, we investigated the binding of endogenous BMAL1 and CLOCK to hPer2 double-stranded DNA fragment immobilized on streptoavidin beads (Fig. 3). For this experiment, we raised specific antibodies against BMAL1 (see Additional file 3). Mouse liver lysates were prepared at 4-h intervals and immunoblotted with antibodies against BMAL1 or CLOCK (indicated as PC in Fig. 3A). When the lysates were incubated with the bead-bound Per2 oligonucleotides and electrophoresis performed, followed by immunoblotting, bands corresponding to BMAL1 and CLOCK were detected (Fig. 3A). The shifted bands seen correspond to phosphorylated BMAL1; and its phosphorylation level peaked around at CT (circadian time) 12, which is consistent with our previous data obtained with the Per2 promoter 26. In contrast, the oligonucleotides with the same length but having scrambled sequences bound neither BMAL1 nor CLOCK proteins. The amounts of BMAL1 and CLOCK pulled down with mutant E1m or with the double-mutant E1mE2m were drastically reduced or almost undetectable compared with those pulled-down by the wild-type EE-element (Fig. 3B). This result suggests that the endogenous BMAL1 and CLOCK proteins bound to the E1 element but not to E2. On the other hand, when we used the E2m oligonucleotide, precipitated BMAL1 and CLOCK proteins were substantially detected. These results suggest that the binding of CLOCK/BMAL1 to E1 and that of unknown factors to E2 are necessary for cell-autonomous circadian transcription.

We next examined the length of the space between E1 and E2, because the spaces between two E-box-like elements were 6 bp in all of 3 Pers EE-elements (Fig. 1A). To see whether the space length was crucial for cell-autonomous circadian transcription, we constructed different spaces (4~8 bp) between E1 and E2 of the hPer2 EE-element and performed IV-ROMS (Fig. 4). sp7, in which there were 7 bp spacing nucleotides between the 2 E-elements, yielded a circadian rhythmicity of Per2 transcription but with a slightly diminished amplitude compared with the wild-type oligonucleotide. The amplitude of sp5 and sp8 (5-bp and 8-bp space, respectively) or sp4 (4-bp space) displayed almost no or no oscillation of Per2, respectively. These results demonstrate that the spacing between the 2 E-elements was critical for rhythm generation and that the 6-bp (or 7-bp) space was necessary for cell-autonomous transcriptional oscillation. Thus the EE-element for circadian rhythm may be defined as follows: 1) E1 and E2 allow 1 and 2 base mismatches, respectively, and 2) the space between the 2 E-elements is 6 or 7 bp (Table 1).

Previous studies 282930 in which clock or clock-controlled genes were extensively examined at the transcriptional level by DNA microarrays revealed that a total of 1551 genes exhibited circadian oscillation. So we examined how many genes contained the EE-element that met our criteria described above in their promoter regions within 1 kb upstream from the TSS. As a consequence of in silico analysis, 30 EE-elements in 29 genes were found among 1510 genes, which sequences are available from the VISTA database (Table 2). The genes that appeared in different papers and organs were clock and clock-controlled genes, such as Nr1d1 (also known as Rev-erbα), Dbp, Avp and Tef, besides Per2. This result strongly suggests that not a single E-box but a direct repeat of E-box-like elements is essential for the generation of cell-autonomous circadian oscillation. To confirm this, by using IV-ROMS we investigated whether the EE-element in Dbp was also required for its rhythmicity. The hDbp EE-element showed circadian oscillation as well as the hPer2 one (see Additional file 4). The mutants of E1 and/or E2 in the EE-element as well as the space mutants, sp4, sp5, and sp8 displayed no rhythmicity (see Additional file 4), which is consistent with the results of Per2 described above.

Among the genes identified by in silico analysis (Table 2), arginine vasopressin (Avp) is a neuropeptide physiologically synthesized in the hypothalamus in a circadian manner but pathologically expressed by small-cell lung cancer. In the latter case, upstream stimulatory factor (USF), another bHLH transcription factor, activates the vasopressin promoter via the adjacent non-canonical E-boxes in tandem with a 6-nucleotides space between them and the second E-box is important for full enhancer function of the first E-box 31. Furthermore, hypoxia-inducible factor 1 (HIF-1), a bHLH-PAS transcription factor, has recently been reported to regulate some hypoxia-inducible genes such as vascular endothelial growth factor (VEGF) through an HIF-1 binding site and its downstream HIF-1 ancillary sequence, both of which are non-canonical E-boxes 32. The space between these motifs in this case is 8 nucleotides and is crucial for activity of promoter. Not only mammalian genes but also fly clock genes contain a direct repeat of E-box-like elements, as discovered in this study. The Drosophila timeless promoter includes an E-box-like sequence dubbed TER1 and is considered as an EE-element together with its 7-bp downstream canonical E-box. This TER1 plays a major role in tim transcription 33. These reports in addition to this study demonstrate that not a single E-box but an EE-element is crucial for transcriptional activation by certain bHLH transcription factors.

Our results suggest that 2 dimeric complexes of transcription factors bind to the 2 E-box-like elements with a short distance between them, such as 6 nucleotides. To support this possibility, we simulated this binding of bHLH transcription factors to the EE-element on the DNA by using the crystal structure of the bHLH domain of the Myc-Max heterodimer bound to DNA (PDB ID: 1NKP) 34. Since 1 turn of double-helix DNA is approximately 34 Å and requires 10.4 nucleotides, the distance from the beginning of E1 to the end of space, which consists of 12 nucleotides, spans 39.2 Å (34/10.4 × 12, Fig. 5A). Two binding faces form a 55.8-degree angle (360/10.4 × 12–360, Fig. 5B). The maximum dimension of the bHLH structure, which lies along the DNA, is about 32 Å (Fig. 5C). This dimension is within the 39 Å calculated above and allows the binding. When the space is 5 nucleotides, for example, the distance between 2 complexes shortens approximately 3 Å horizontally compared with the case of 6 nucleotides; however, the angle between them becomes 20.6°. This precludes the possibility that the bHLH structure binds to the EE-element, because 2 whole structures including PAS domains may collide with each other. This simulation supports our contention that 6 nucleotides between 2 E-box-like elements, not 5 nucleotides, allow the binding of bHLH transcription factors to the EE-element on the DNA.

As a candidate for unknown factors that bind to E2, another type of transcription factor can be a possible candidate as well as the bHLH family transcription factor including HIF1 35 or the bZip family transcription factor E4BP4 (also known as NFIL3) 36. Our preliminary EMSA suggests that some proteins may bind to E2 (see Additional file 5). Alternatively, different forms of the CLOCK and BMAL1 complex can be involved in this circadian transcription. The circadian transcriptional complex includes at least CLOCK, NPAS2, BMAL1, PERs and CRYs, probably in different combinations, and these proteins can be posttranslationally modified 37. While E1 element is sufficient for binding of CLOCK and BMAL1, it is possible that binding to E2 can occur only after CLOCK and/or BMAL1 originally binds to E1. Thus E2 can be necessary for the binding of different forms of the complex or perhaps the binding to E2 is important for complex activation. Further studies, including identification of unknown factors or the circadian transcription complex and their crystal structures, will be required to elucidate the precise mechanism of transcriptional activation via an EE-element.

Here we identified a direct repeat of E-box-like elements consisting of a canonical E-box or non-canonical E-box in tandem with an E-box-like element, having a short interval of 6 base pairs between them. Although other regions outside the EE-element seem to be active in robust circadian oscillation with high amplitude 38, the EE-element identified in this study, not a single E-box, is the minimal required element for the generation of cell-autonomous transcriptional oscillation of clock and clock-controlled genes.

YN and MY carried out experiments and drafted the manuscript. AT discussed the results. MY, HS, and TY carried out the in silico study. AY participated in the coordination and provided financial support. TN carried out the structural design. TT participated in the experimental design and coordination and wrote the paper. All authors read and approved the final manuscript.