Closing the Circadian Loop: CLOCK-Induced Transcription of Its Own Inhibitors per and tim

See allHide authors and affiliations

Science  05 Jun 1998:
Vol. 280, Issue 5369, pp. 1599-1603
DOI: 10.1126/science.280.5369.1599


The circadian oscillator generates a rhythmic output with a period of about 24 hours. Despite extensive studies in several model systems, the biochemical mode of action has not yet been demonstrated for any of its components. Here, the Drosophila CLOCK protein was shown to induce transcription of the circadian rhythm genes periodand timeless. dCLOCK functioned as a heterodimer with aDrosophila homolog of BMAL1. These proteins acted through an E-box sequence in the period promoter. Thetimeless promoter contains an 18–base pair element encompassing an E-box, which was sufficient to confer dCLOCK responsiveness to a reporter gene. PERIOD and TIMELESS proteins blocked dCLOCK's ability to transactivate their promoters via the E-box. Thus, dCLOCK drives expression of period and timeless,which in turn inhibit dCLOCK's activity and close the circadian loop.

In animals, plants, or prokaryotes, activities such as locomotion or photosynthesis do not occur with equal probability throughout the 24-hour day but are organized by an endogenous circadian oscillator. The oscillator allows the organism to anticipate daily environmental fluctuations rather than merely respond to them. InDrosophila, two essential oscillator components,period (per) and timeless(tim), have mRNA transcript levels that cycle with a circadian rhythm (1). Mouse homologs of per are also regulated in a circadian fashion (1). Thus, the core mechanism of the circadian oscillator is likely to be conserved between Drosophila and mammals.

In Drosophila, the per and tim mRNA oscillations are controlled in large part by transcriptional regulation (2) and some posttranscriptional processes (3). Point mutations in the coding region can change the length of the cycle or abolish it (1), indicating that PER and TIM proteins control their own oscillations. Overexpression of a pertransgene down-regulates the endogenous gene (4). PER and TIM proteins are nuclear at specific times of day, in such a way that nuclear entry coincides with a downturn in mRNA levels. As the nuclear protein levels fall, the mRNA levels rise (5). Thus, the manner in which PER and TIM regulate their transcription is thought to be through rhythmic repression in a negative feedback loop.

However, it is not known whether PER and TIM interact directly with transcription factors or influence transcription indirectly. Although neither protein has a recognizable DNA binding domain, PER shares with some transcription factors a region of homology termed the PAS domain (6), which functions as a protein-protein interaction interface (1) and often mediates heterodimerization between distinct bHLH-PAS transcription factors (7). TIM, which lacks the PAS domain, interacts with PER through the PER PAS domain (8). A mouse gene,Clock, encodes a bHLH-PAS protein whose wild-type function is necessary for the maintenance of circadian rhythms (9). An E-box sequence within the per promoter is required for high-level expression and robust cycling of transcription in transgenic flies (10). Because E-boxes are well-defined targets of bHLH-PAS protein heterodimers, we reasoned that Clock exists in flies and activates transcription of per andtim. In this model, PER and TIM would then abrogate the activity of CLOCK in a rhythmic fashion. This would provide the biochemical steps necessary to close the circadian loop.

We first determined whether a Drosophila homolog of the mouse Clock (mClock) gene existed and tested whether the protein product could activate per ortim transcription. A low-stringency screen of a cDNA library from adult heads with an mClock probe encoding the bHLH and PAS domains yielded overlapping clones with a high degree of homology to mClock (Fig. 1A) (11). Library screening did not reveal any clones with significant similarity to other bHLH-PAS proteins. At high stringency, a single band was seen on a Southern (DNA) blot of fly genomic DNA when a fragment encompassing dclock PAS B was used as the probe, which suggests that dclock is a single-copy gene (12). Sequence analysis revealed differential splicing. In variant A, the reading frame is intact and encodes a full-length protein of 115.7 kD. In variant B, the coding region goes out of frame after the bHLH domain (13). dCLOCK is highly homologous to both mCLOCK and NPAS2 (MOP4) (14, 15). dCLOCK and mCLOCK share 60% identity in the bHLH domain, 43% in PAS A, and 70% in PAS B (Fig. 1A). Both dCLOCK (1023 amino acids) and mCLOCK (855 amino acids) are rich in glutamine and have long polyglutamine stretches. We mapped the chromosomal position of dclock to 66A (16).

Figure 1

Analysis of dCLOCK protein sequence and mRNA expression. (A) Alignment ofDrosophila and mouse CLOCK proteins, showing percent identity (similarity) between bHLH, PAS A, and PAS B regions. The extent of the PAS A and PAS B domains shown are as defined by the PAS-PYP module (36). The glutamine-rich (Q-rich) region contains polyglutamine (poly Q) repeats indicated in black. (B) Northern blot analysis of polyadenylated (poly A) mRNA from whole flies (lane 1), total RNA extracted from S2 cells (lane 2), and total head RNA (lane 3), hybridized with a riboprobe covering nucleotides 530 to 761 (37). Ten micrograms of RNA was loaded in each well. Because of the small difference in size, the known splice variants could not be distinguished in this assay. (C) Time course of dclock mRNA levels. RNase protection analysis was done on head RNA from flies harvested throughout a light-dark cycle (38). The white and black bars beneath the figure represent the periods of light and dark, respectively, with lights on at ZT0 and lights off at ZT12. Lane A is a positive control for splice variant A, which gives rise to the upper doublet encoding the full-length protein. Lane B is a control for variant B. The probe hybridizing to this variant is clipped at a two-base mismatch and gives rise to the lower doublet encoding truncated proteins. The relevant bands are marked with an arrowhead. The band in the lower panel is a protection of RP49 RNA for reference. The zeitgeber time is indicated above the lanes. RNA from appendages (App.), bodies (Body), and S2 cells is also shown. (D) Variant A (squares) and B (triangles)dclock RNA levels from Fig. 1C were normalized to RP49 values and plotted with zeitgeber time on the x axis. Light and dark periods are reversed relative to Fig. 1C to highlight the bimodality. The assays were performed three times with indistinguishable results on independently obtained samples. A representative data set is shown.

A dclock probe hybridized to a single band at about 5 kb on a Northern (RNA) blot of whole fly RNA (Fig. 1B). dclock was detectable in RNA isolated from head, body, and appendage fractions, indicating that dclock was expressed widely (Fig. 1C). This correlates well with the observation that independent oscillators are present throughout the fly (17). To examine the temporal expression profile of dclock, we analyzed RNA isolated from fly heads harvested throughout 24 hours of a light-dark cycle. Unlikeper and tim, dclock mRNA oscillated in a bimodal fashion (Fig. 1, C and D). It peaked at zeitgeber time 5 (ZT5) and at ZT23, whereas per and tim peak at ZT16 (1). Preliminary experiments indicated that the drop at ZT1 occurred even when the animals were kept in darkness (12). Whether this profile is reflected at the protein level is not known. To explore the possible significance of splice variant B, we measured its level of expression. Splice variant B was expressed weakly at all times of day, and it cycled in phase with the full-length form (Fig. 1, C and D).

To investigate the role of dCLOCK in the transcriptional regulation ofper and tim, we used a transient transfection assay in Drosophila Schneider (S2) cells. Untransfected S2 cells did not express detectable levels of dclock mRNA (Fig.1, B and C), PER, or TIM (8). We cotransfected a construct expressing dclock along with a construct driving luciferase expression from either the per or timpromoters (Fig. 2A). dCLOCK caused an approximately 20- and 200-fold induction of the period andtimeless promoters, respectively.

Figure 2

Transcriptional activation ofper and tim by dCLOCK. (A) Specific activation of period and timeless promoters by dCLOCK. Drosophila S2 cells were transfected with pAct expression plasmids as a source of dCLOCK, dARNT, or dSIM. These were cotransfected with per-luc, tim-luc, or CME-lacZ reporter plasmids (35). (B) The E-box was necessary for dCLOCK-mediated activation of period promoter fragments. S2 cells were transfected with either dCLOCK or pAct as indicated. Reporter plasmids contain a portion of the per promoter fused to the hsp70basal promoter driving lacZ (10). The 69- and 154-bp fragments correspond to –563 to –494, and –603 to –449 of the per promoter, respectively. 154 (Δ1) refers to a deletion from –529 to –519, and 154 (Δ2) to –539 to –529, both of which remove a portion of the E-box (10). (C) The E-box was sufficient for activation of per andtim promoters by dCLOCK. Reporters contained four copies of an 18-bp element from either the per or timpromoters fused to hsp70 driving luciferase (39). Identity between the per andtim E-boxes and the surrounding sequence is indicated in bold. In mutant E-box targets, the central CG of the E-box was replaced with GC (underlined). Plasmids expressing dCLOCK or dCLOCK(ΔQ) were cotransfected alone or in combination. dCLOCK(ΔQ) contains amino acids 1 through 792 of dCLOCK, omitting two of the three glutamine repeats.

To determine whether the effect of dCLOCK was specific for theper and tim promoters, we asked whether dCLOCK could induce transcription from the CNS midline element (CME). This element is a target for the bHLH-PAS partners dSIM and dARNT (TANGO) (18). dCLOCK was unable to induce transcription from this promoter element in S2 cells (Fig. 2A). Conversely, dSIM and dARNT were unable to induce the per or tim promoters, although they did induce expression from the CME. Thus, dCLOCK was not a promiscuous E-box activator but was rather a specific activator of the per and tim promoters. The sequence of splice variant B encoded two proteins that could act as dominant negative inhibitors (13). However, we did not see an effect of this splice variant on the ability of dCLOCK to induce expression (12). Although exogenous dARNT is necessary for high-level dSIM-mediated transactivation of the CME in S2 cells (18), dARNT did not cause a further increase in the transcriptional activity of dCLOCK (see below).

We also tested a 154–base pair (bp) fragment and a 69-bp subfragment of the per promoter, both of which confer robust expression and circadian cycling of mRNA in transgenic flies (10). Both fragments were sufficient for the dCLOCK-mediated induction (Fig. 2B). Furthermore, dCLOCK did not activate target constructs containing the 154-bp fragment with deletions in the E box (CACGTG), which abolish high-level mRNA cycling in transgenic animals (10) (Fig. 2B). Thus, dCLOCK's activity on this element of the period promoter closely correlates with the pattern observed in transgenic flies.

Sequence analysis of the tim promoter revealed a 10-bp consensus with the per E-box region (Fig. 2C). To test whether per and tim E-boxes supported dCLOCK activation, we tested multimerized E-boxes fused to a luciferase expression cassette. Both per and tim E-box targets were activated by dCLOCK (Fig. 2C). Constructs containing mutated E-boxes did not support transcriptional activation by dCLOCK. A truncated dCLOCK protein lacking two of the three polyglutamine repeats [dCLOCK(ΔQ)] only weakly activatedper and tim E-boxes, which is consistent with the idea that this region corresponds to the transactivation domain. Furthermore, this protein antagonized full-length dCLOCK activation ofperiod and timeless E-boxes. The activity of this truncation correlates well with the finding that the mouse circadian mutant Clock, in which part of the glutamine-rich region is deleted (9), behaves genetically as an antimorph (19).

bHLH-PAS transcription factors most often function as heterodimers (18, 20). We initially predicted that the partner for dCLOCK was dARNT, a common partner for several bHLH-PAS proteins (18, 20, 21). When dsim is transfected into S2 cells, little activation of its target element is seen without cotransfection of dARNT (18). However, exogenous dARNT had no effect on the level of induction by dCLOCK (Fig.2A), which suggests that dCLOCK acts either as a homodimer or with another partner endogenous to S2 cells. In an accompanying paper (22), evidence is provided that the bHLH-PAS transcription factor BMAL1 (MOP3) (15, 23) is the partner for mCLOCK. To test whether this is true for Drosophila, we obtained a Drosophila homolog of BMAL1 through the expressed sequence tag (EST) database. Like dclock, per, and tim (1), dbmal1 is expressed in the fly head as determined by Northern blot (12). However, when cotransfected with dCLOCK, this clone had no effect on the induction of the per promoter (12). Northern blot analysis showed dbmal1 expression in naı̈ve Schneider cells (12), in contrast to the absence of detectable signal when the same blot was probed with dclock (Fig. 1B). High levels of endogenous dBMAL1 may obscure any contribution of exogenous dBMAL1 (24). We employed both a two-hybrid and a one-hybrid yeast assay (25) as a heterologous system to test whether dCLOCK and dBMAL1 interact and bind the per E-box. dCLOCK and dBMAL1 specifically interacted with each other in yeast cells (Fig. 3A) and bind to a region encompassing the per E-box but not to an E-box mutant (Fig.3B). Together with the results reported in mammals (22,26), this strongly suggests that dCLOCK and dBMAL1 form a complex that activates period and timelesstranscription in vivo. We localized dbmal1 to chromosomal position 76C (27).

Figure 3

Interaction of dCLOCK with dBMAL1 and binding of the heterodimer to an E-box in theper promoter. (A) Yeast two-hybrid assay showing specific interaction of dCLOCK with dBMAL1. Shown are yeast patches expressing the indicated LEXA bait (rows) with the indicated VP16 hybrid (columns). p65 is full-length rat synaptotagmin (negative control), expressed as a LEXA fusion. Blue precipitate indicates cumulative β-Gal activity, resulting from activation of the lacZ reporter gene by protein-protein interaction. Each triplicate represents three independent transformants (40). dCLOCK consists of amino acids 1 through 496 expressed as a LEXA fusion. VP16-dBMAL1 consists of amino acids 1 through 413 expressed as VP16 fusion. (B) Yeast one-hybrid assay showing binding of the dCLOCK-dBMAL1 heterodimer to the per E-box site. Shown are yeast patches expressing the indicated pairs of proteins (rows) and transformed with the indicated reporter construct (columns). DNA binding results in activation of the lacZ reporter gene. The LEXA domain has no effect on DNA binding in this one-hybrid assay. insert, reporter strain lackingper upstream sequences; 21-bp, mut, reporter strain with E-box element scrambled; 21-bp, wt, same 21-bp fragment with wild-type E-box sequence; 69-bp, wt, larger region from the perpromoter containing an intact E-box (41).

To test the prediction that PER and TIM proteins prevent dCLOCK-induced expression from their E-boxes, we cotransfected per andtim cDNAs expressed from the Drosophila actin 5Cpromoter (Fig. 4). We observed a 4.5-fold reduction in dCLOCK-mediated reporter induction. Both perand tim were required to elicit this effect, probably because of their mutual dependence for nuclear localization (8). PER and TIM exhibited no negative effect on expression from the E-box multimer when dCLOCK was excluded or when the E-box was mutated. Thus, PER and TIM negated dCLOCK's transcriptional activity rather than repressing dCLOCK independent transcription. These data are in accord with a model in which either PER or TIM binds to either dCLOCK or dBMAL1, sequestering it in a nonfunctional complex.

Figure 4

PER and TIM inhibition of dCLOCKmediated transactivation. S2 cells were transfected with combinations of pAct expression plasmids (dCLOCK, PER, and TIM) and wild-type or mutantper E-box reporters as indicated (42).

When our results are placed in the context of the current understanding of the Drosophila circadian oscillator, dCLOCK closes the feedback loop. A dCLOCKdBMAL1 complex drives expression ofper and tim by binding an E-box that is present in their promoters. With time, PER and TIM heterodimers accumulate, translocate to the nucleus (1), and act as dominant negative inhibitors of dCLOCK-dBMAL1. As mRNA and protein levels fall, the inhibition is relieved, which allows dCLOCK-dBMAL1 to initiate a new round of synthesis. The position that dCLOCK holds in driving the expression of circadian oscillator components is reminiscent of the role played by the white collar genes inNeurospora (28). However, the ability of WC-1 and WC-2 to directly activate the frequency promoter has not yet been established.

It is tempting to speculate that the Drosophilafour-component transcriptional feedback loop described here is sufficient to generate a rudimentary circadian rhythm. This oscillation would be amplified by other, unknown proteins that regulate RNA stability (3), protein stability (29), and phosphorylation (30) of the oscillator components.

  • * Co–first authors.

  • To whom correspondence should be addressed. E-mail: stevek{at}


View Abstract

Navigate This Article