Interlocked Feedback Loops Within the Drosophila Circadian Oscillator

See allHide authors and affiliations

Science  22 Oct 1999:
Vol. 286, Issue 5440, pp. 766-768
DOI: 10.1126/science.286.5440.766


Drosophila Clock (dClk) is rhythmically expressed, with peaks in mRNA and protein (dCLK) abundance early in the morning. dClk mRNA cycling is shown here to be regulated by PERIOD-TIMELESS (PER-TIM)–mediated release of dCLK- and CYCLE (CYC)–dependent repression. Lack of both PER-TIM derepression and dCLK-CYC repression results in high levels of dClk mRNA, which implies that a separate dClk activator is present. These results demonstrate that the Drosophila circadian feedback loop is composed of two interlocked negative feedback loops: aper-tim loop, which is activated by dCLK-CYC and repressed by PER-TIM, and a dClk loop, which is repressed by dCLK-CYC and derepressed by PER-TIM.

The circadian oscillators of eukaryotic and certain prokaryotic organisms are controlled through autoregulatory feedback loops in gene expression (1). InDrosophila, five genes have been identified that are necessary for circadian feedback loop function: period(per), timeless (tim),Drosophila Clock (dClk), Cycle(Cyc), and double-time (dbt) (2–9). Three of these genes—per,tim, and dClk—are rhythmically expressed:per and tim mRNA levels peak early in the evening [zeitgeber time (ZT) 13–16, where ZT 0 is lights on and ZT 12 is lights off], and dClk mRNA levels peak late at night to early in the morning (ZT 23 to ZT 4) (2–5,10).

Regulation of per and tim expression has been characterized in some detail. Activation of per andtim transcription is mediated by two basic helix-loop-helix–PAS transcription factors, dCLK and CYC, which form heterodimers that target E-box regulatory elements of the sequence CACGTG in the per and tim promoters (4, 6, 7, 11,12). Although per and tim mRNAs reach peak levels early in the evening (ZT 13–16), PER and TIM levels do not peak until late evening (ZT 18–24) (13,14). This delay results from the initial destabilization of PER by DBT-dependent phosphorylation, followed by the stabilization of PER by dimerization with TIM (8, 9). PER-TIM dimers then move into the nucleus and form a complex with dCLK-CYC activators (15), which results in transcriptional repression (by deactivation) ofper and tim (4).

Comparatively little is known about the regulation of dClkmRNA cycling. The levels of dClk mRNA are low in mutants lacking PER (per 01) or TIM (tim 01) function, which suggests that PER and TIM activate dClk transcription in addition to their roles as transcriptional repressors (5). The mechanism of PER-TIM–dependent activation is not known, but three models have been proposed to account for this activation (5). In the first two models, PER and TIM promote dClk transcription by shuttling transcriptional activators into the nucleus (Fig. 1A) or by coactivating a transcriptional complex (Fig. 1B). In the third model, PER or TIM or both inhibit the activity of a transcriptional repressor complex (Fig. 1C).

Figure 1

Models for how PER and TIM might activate dClk expression. (A) Shuttling transcriptional activators into the nucleus. (B) Coactivating a transcription complex. (C) Inhibiting a transcriptional repressor complex. Gray triangle, transcription complex that promotes dClk expression; black circle, transcription complex that represses dClk expression; striped bar, dClk transcriptional regulatory sequences. The diagram is based on fig. 5 of (5).

To distinguish among these alternative models, we measureddClk mRNA levels in different clock gene mutant combinations. Because dCLK and CYC are both required for perand tim activation, we predicted that mutants lacking functional dCLK (dClk Jrk) or CYC (Cyc 0) would exhibit low levels ofdClk mRNA because the concentrations of the PER and TIM activators (of dClk) would be low. We were surprised to find that the level of dClk mRNA was indistinguishable from the wild-type peak in both mutants (Fig. 2). The levels of dClk mRNA do not vary significantly over the circadian cycle in these mutants (P > 0.05), which is consistent with the lack of a functional circadian oscillator (6, 7).

Figure 2

Peak levels of dClk mRNA are present throughout the circadian cycle indClk Jrk and Cyc 0 mutants. (A) Wild-type, dClk Jrk, andCyc 0 flies maintained in LD12:12 for a minimum of 3 days were collected and subjected to ribonuclease (RNase) protection assays as described (2). Total head RNA (12 μg) was hybridized to a dClk RNA probe that protects a 290-nucleotide (nt) fragment (5) and an RP49 probe that protects a 59-nt fragment (2). After RNase digestion, reactions were run on a 6% denaturing polyacrylamide gel. Numbers at the top of each lane correspond to the ZT of sample collection. M indicates the 123–base pair (bp) marker lane. RP49 was included as a control for the total amount of RNA in each reaction. (B) dClk and RP49 protection products were quantified with a PhosphorImager and the dClk/RP49 ratios were plotted. mRNA levels are shown for wild-type (•),dClk Jrk (▴), and Cyc 0(▪) flies. The dClk/RP49 values were normalized to the peak dClk level in the wild type at ZT 1, which was set to 1.0. Error bars represent standard deviations based on five independent experiments. The light and dark bars represent times when lights were on or off, respectively. Analysis of these results by analysis of variance showed no significant differences indClk mRNA over the day (P > 0.05) indClk Jrk and Cyc 0 mutants.

The high level of dClk mRNA in the absence of dCLK-dependent PER accumulation indicates that PER-dependent dClkactivation does not occur by nuclear localization of an activator or by coactivation (Fig. 1, A and B). However, the possibility remains that low levels of per and tim transcripts indClk Jrk or Cyc 0 mutants (6, 7) lead to some active PER-TIM dimer formation and subsequent activation of dClk transcription. To eliminate this possibility, we measured dClk mRNA levels in per 01;dClk Jrk andper 01;Cyc 0 double mutants. In both cases, the levels of dClk mRNA observed under light-dark (LD) or constant dark (DD) conditions were close to the peak level in wild-type flies (Fig. 3), indicating that PER-TIM activatesdClk transcription through derepression (Fig. 1C).

Figure 3

Peak levels of dClk mRNA persist in per 01;dClk Jrkand per 01; Cyc 0 double mutants. (A) Wild-type, per 01,per 01;dClk Jrk, andper 01;Cyc 0 flies were maintained, collected, and assayed as in Fig. 1. Numbers at the top of each lane correspond to the time of sample collection. M indicates the 123-bp marker lane. RP49 was included as a control for the total amount of RNA in each reaction. (B)dClk and RP49 protection products were quantified with a PhosphorImager and the dClk/RP49 ratios were plotted. The dClk/RP49 values at ZT 1 (open bars) or ZT 13 (solid bars) were normalized to the peak dClk level in the wild type at ZT 1, which was set to 1.0. Error bars represent standard deviations based on three independent experiments. (C) Wild-type, per 01,dClk Jrk,per 01;dClk Jrk,Cyc 0, andper 01;Cyc 0 flies were entrained in LD12:12 for 3 days and collected at circadian times 1 (CT 1) and 13 (CT 13) during the first day of DD. RNase protection assays were performed as in Fig. 1. Numbers and symbols are the same as in (A). (D) dClk and RP49 protection products from (C) were quantified as described in (B).dClk/RP49 values are shown for CT 1 (gray bars) and CT 13 (solid bars). Normalization was performed as described in (B). This experiment was repeated with similar results.

The dClk repressor that is removed as a result of PER-TIM accumulation appears to be either dCLK-CYC itself or a repressor that is activated by dCLK-CYC. When comparing the levels of dClkbetween per 01 flies andper 01;dClk Jrk orper 01;Cyc 0 double mutants, the presence of active dCLK and CYC results in the repression of dClk transcript accumulation. Inper 01 mutants, dClk mRNA is at low but detectable levels (5) (Fig. 3). This suggests that in the absence of PER-TIM derepression, dClktranscription reaches a steady state in which activation and dCLK-CYC–dependent repression equilibrate to produce low levels ofdClk mRNA transcripts and, hence, of dCLK protein. Inper 01 and tim 01 mutants,per and tim transcription is constitutive andper and tim transcripts are relatively low in abundance (2, 16). This result can be explained by the partial activation of per andtim by low levels of dCLK-CYC dimers in the absence of PER-TIM repression.

On the basis of these observations, we propose that interlocked negative feedback loops mediate circadian oscillator function inDrosophila (Fig. 4). Late at night, PER-TIM dimers in the nucleus bind to and sequester dCLK-CYC dimers. This interaction effectively inhibits dCLK-CYC function, which leads to the repression of per and timtranscription and the derepression of dClk transcription. As PER-TIM levels fall early in the morning (ZT 0–3), dCLK-CYC dimers are released and repress dClk expression, thereby decreasingdClk mRNA levels so that they are low by the end of the day (ZT 12) (5). Concomitant with the drop indClk mRNA levels (through dCLK-CYC–dependent repression) is the accumulation of per and tim mRNA (through E-box–dependent dCLK-CYC activation) (2, 3). AsdClk mRNA falls to low levels early in the evening (ZT 15), the levels of dCLK-CYC also fall (17), leading to a decrease in per and tim transcription and an increase indClk mRNA accumulation. A new cycle then begins as high levels of PER and TIM enter the nucleus and dCLK starts to accumulate late at night (17, 18).

Figure 4

Model for gene regulation within theDrosophila circadian oscillator. During the late evening (right), PER-TIM dimers (closed and open squares, respectively) enter the nucleus and bind dCLK-CYC dimers (closed and open circles, respectively), thereby repressing per-timactivation. Concurrently, the binding of PER-TIM dimers to dCLK-CYC releases dCLK-CYC–dependent repression of dClk, thus enabling dClk transcription via a separate activator or activator complex (triangle). By midday (left), high levels of dCLK-CYC (in the absence of PER-TIM) serve to activateper-tim transcription and repress dClktranscription (either directly or through intermediate factors). As the circadian cycle progresses, PER-TIM dimers accumulate and enter the nucleus during the late evening to start the next cycle. Dashes, maximal repression; plus signs, maximal activation; wavy lines, mRNA.

These observations also fit well with the regulation ofDrosophila cryptochrome (cry), whose mRNA cycles in phase with that of dClk (17). LikedClk, cry mRNA transcripts are constitutively low in per 01 mutants and constitutively high indClk Jrk or Cyc 0 single mutants and inper 01;dClk Jrk orper 01;Cyc 0 double mutants (19). These striking similarities between dClkand cry mRNA phases (in the wild type) and dClkand cry mRNA levels in circadian mutants suggest that thecry locus may be regulated by the same PER-TIM release of dCLK-CYC repression mechanism as dClk.

These results reveal the existence of a dClk feedback loop and its regulatory interactions with the well-characterizedper-tim feedback loop. One clear prediction from these experiments is that there is a separate activator of dClkexpression. Such an activator is indicated by the high levels ofdClk mRNA in the absence of PER and of either dCLK or CYC. This observation is somewhat surprising because the presence of this activator is independent of factors that control the expression of other clock genes (that is, PER, dCLK, and CYC).

Data supporting the existence of interlocked per-tim anddClk feedback loops were obtained from whole heads, raising the possibility that dClk expression in small subsets of “clock-specific” cells such as the locomotor activity pacemaker cells (that is, lateral neurons) (20–22) could be masked bydClk expression in other tissues. However, the autonomy and synchrony of per expression in diverse tissues in the head and body suggest that the circadian feedback loop mechanism is the same in all tissues (23) and argue against fundamental tissue-specific differences in the feedback loop mechanism.

An important aspect of circadian biology is how the clock regulates clock-controlled genes (CCGs). In mammals, it has been shown in vitro that CLOCK and BMAL-1 (the mammalian ortholog of CYC) activate vasopressin gene transcription and that all three mouse PERs and TIM repress this activation, resulting in peak vasopressin mRNA transcripts by midmorning (ZT 6) (24). Although this mode of regulation may be more general for CCGs whose mRNA transcripts peak in phase with per (or mPer), it does not explain how CCGs that cycle in antiphase are regulated. The results presented here provide a possible mechanism by which the clock regulates CCGs whose mRNAs cycle in antiphase to those of per. The similarities between dClk and cry mRNA profiles in the wild type and in several single and double circadian mutants suggest that PER-TIM release of dCLK-CYC repression may serve a more general role in regulating CCG mRNAs that cycle in antiphase to per mRNA.

  • * These authors contributed equally to this work.

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


View Abstract

Navigate This Article