During the last five years, enormous progress has been made in understanding the molecular basis of circadian systems, mainly by molecular genetic studies using the mouse and fly. Extensive evidence has revealed that the core clock machinery involves “clock genes” and “clock proteins” functioning as molecular cogs. These participate in transcriptional/translational feedback loops and many homologous clock-components in the fruit fly Drosophila are also expressed in mammalian clock tissues with circadian rhythms. Thus, the mechanisms of the central clock seem to be conserved across animal kingdom. However, some recent studies imply that the present widely accepted molecular models of circadian clocks may not always be supported by the experimental evidence.
The physiological properties of most organisms, from cyanobacteria to human, display a circadian (Latin circa dies, or ‘about a day’) pattern of activity, which is regulated by an endogenous circadian clock. While rhythms controlled by circadian clocks are self-sustained and persist robustly in constant darkness and temperature with a period close to 24 hr, they are entrained by environmental time cues (Zeitgebers) such as light, temperature, and feeding (Soriano, 1981; Rusak et al., 1993). The circadian clock allows organisms to anticipate periodic changes in environmental circumstance and to change their physiological status accordingly. The internal system that maintains circadian rhythms can be formally represented by three different components: namely, an input pathway, the pacemaker itself, and an output pathway (Dunlap, 1999). The input pathway transmits light and/or thermal, non-photic information from external stimuli to the clock (pacemaker). The pacemaker itself, which comprises the mechanism of the endogenous oscillator, is the source of circadian temporal modulation. The output pathway conveys rhythmic information from the central oscillator for eliciting the daily changes in behavioral activity and other physiological processes.
The molecular genetic study of circadian clocks was initiated in the fruit fly Drosophila melanogaster. When held in a light:dark (LD) 12:12 cycle, wild-type Drosophila individuals show a circadian pattern in their locomotor and eclosion behavior. Extensive evidence indicates that the circadian control of both behaviors depends upon several so-called “clock genes” and “clock proteins.” Many of the clock genes identified in Drosophila are also expressed in mammals as well as in other insects. In Drosophila, certain cell clusters in the brain, the lateral neurons (LNs) and dorsal neurons (DNs), express clock genes such as period (per) and timeless (tim) (Liu et al., 1992; Saez and Young, 1996) (Table 1). In particular, the LNs are suggested to be the major site of circadian pacemakers that control daily locomotor activity and pupal eclosion rhythms. In adult flies, the ventral group of LNs (LNvs) have been subdivided into two subgroups of typically five small LNvs (s-LNvs) and four large LNvs (lLNvs) projecting to the dorsal protocerebrum and to the medulla neuropile of the optic lobe, respectively (Stanewsky et al., 1997; Kaneko and Hall, 2000). The l-LNvs also project to the LNvs and the medulla of the contralateral side. Both groups of LNvs express the pigment-dispersing factor (pdf) gene, which is often used to reveal the projection pattern of LNvs (Kaneko and Hall, 2000). The pdf gene encodes the neuropeptide PDF, which is thought to function as a molecule signaling the clock's output to target tissues (see below).
Comparison of properties and functions of Drosophila and mouse clock proteins.
Negative Pacemaker Components
The first identified genetic component of the clock was the period (per) gene discovered by Konopka and Benzer (1971). They systemically screened mutagen-exposed Drosophila for circadian clock mutants showing altered temporal pupal eclosion patterns in constant conditions (constant darkness, DD). Konopka's mutant strains had different circadian eclosion phenotypes showing abnormally long (perL) or short (perS) rhythms and some were arrhythmic (per0) but all mapped to the single genetic locus per. Not only eclosion rhythms were affected by these mutations, but the locomotor activity rhythms were also altered in a similar way. In the mid-1980s, the per gene was cloned and shown to encode a protein, PERIOD (PER), which contains the PAS domain (Fig. 1). PAS is an acronym for the first three proteins found to share this domain: PER, the product SIM of the Drosophila gene single-minded, and a mammalian protein ARNT (aryl hydrocarbon receptor nuclear translocator) (Crews et al., 1988). The PAS domain has since been known to be an important signaling module for protein-protein interaction, and for monitoring a number of environmental changes such as light, oxygen, and redox potential.
Both per mRNA and PER protein levels exhibit rhythmic abundance, reflecting the corresponding behavioral rhythm: the normal circadian molecular rhythm in wild-type flies, the short or long period rhythm in perS or perL mutants, and the lack of any rhythm in per0 flies (Hardin et al., 1990). Characterization of these mutant flies using genetic mosaic and transgenic approaches revealed that both the pupal eclosion and locomotor activity rhythms are dominantly controlled by only a small number of per-expressing LNs in fly brain (Zerr et al., 1990; Liu et al., 1992). The successful identification of per prompted subsequent genetic and biochemical screens to identify additional components of the molecular clock.
The second fly clock gene cloned was timeless (tim) (Sehgal et al., 1994). tim0 flies are arrhythmic, and tim mRNA and levels of its gene product TIMELESS (TIM) protein are almost coincident with those of per and PER, respectively (Hunter-Ensor et al., 1996; Myers et al., 1996). More importantly, tim0 mutants exhibit an elevated level of cytoplasmic PER protein in the fly head, and TIM protein was also isolated as an interaction partner of PER using a yeast two-hybrid assay (Vosshall et al., 1994; Gekakis et al., 1995). In wild-type flies, the spatial expression pattern of TIM is very similar to that of PER, and PER and TIM accumulate in nuclei of clock neurons in a circadian manner (Saez and Young, 1996). Thus, the circadian molecular clock is proposed to be composed of a negative feedback loop mechanism: PER and TIM heterodimerize and translocate to the nucleus, and the resulting nuclear PER:TIM heterodimer inhibits its own transcription (Fig. 2). PAS domains have been shown to be dimerization surfaces, found in proteins that contain a basic helix-loop-helix (bHLH) DNA-binding domain and that often function as heterodimeric partners. This suggested that PER could function as a transcriptional repressor of a bHLH-PAS heterodimeric pair, because PER lacks the bHLH domain (Huang et al., 1993). The TIM sequence is unique and contains neither a known DNA-binding domain nor a PAS domain. Instead, TIM has three ARMADILLO-like domains (ALD) (Kyriacou and Hastings, 2001; Stanewsky, 2002). The ALD has been shown to participate in protein-protein interactions in some cell-adhesion- and cytoskeleton-associated proteins (Hatzfeld, 1999). Two of the TIM ALDs overlap with the region that can bind to PER (Kyriacou and Hastings, 2001; Saez and Young, 1996).
Positive Pacemaker Components
A behavioral screen in the mouse produced a semi-dominant long-period mutant called Clock (mClk) (Vitaterna et al., 1994). Homozygotes generate either long-period loco-motor activity rhythms or are arrhythmic. In 1997, positional cloning and sequencing analysis revealed that mClk gene product, mCLOCK (mCLK), is a bHLH-PAS-containing transcription factor (King et al., 1997). Subsequent molecular genetic studies clarified that mCLK interacts directly with another bHLH-PAS family protein, BMAL1 (brain and muscle ARNT like protein 1, MOP3) (Gekakis et al., 1998; Takahata et al., 1998). Disruption of BMAL1 gene causes an immediate loss of locomotor activity rhythms in mice. Furthermore, the mCLK:BMAL1 heterodimer was shown to bind to the specific DNA element E-box having a nucleotide sequence CACGTG found in the 5′ flanking region of the mouse period homolog (mPer1), and this heterodimer enhances transcription of mPer1 (Takahata et al., 1998). Homology cloning approaches identified Clock (Clk) and cycle (cyc), as Drosophila counterparts to the mouse's clock genes mClk and BMAL1, respectively (Allada et al., 1998; Darlington et al., 1998). These two genes encode proteins containing a bHLH-PAS domain, and the gene products, CLOCK (CLK) and CYCLE (CYC) can dimerize and bind directly to the E-box enhancers present in the per and tim promoters, as well as stimulate transcription of both the per and tim genes, respectively. Moreover, it was found that if PER and TIM proteins are both produced in cultured Drosophila cells and allowed to undergo nuclear localization, transcriptional activation through the E-box by CLK:CYC dimer is remarkably suppressed (Darlington et al., 1998). Thus, a transcription/translation autoregulatory feedback loop has been defined as a driving force for the clock's molecular mechanism. The mammalian circadian clock seems to be quite similar to this fly clock. The bHLH-PAS dimeric pair CLK:CYC resides in the nucleus where it occupies the Ebox elements in the per and tim genes, positively regulating their transcription. PER and TIM protein levels continue to rise throughout the day to their peak levels in the early evening, a few hr after the peak level of per and tim mRNAs. The two proteins heterodimerize and translocate into the nucleus where they inhibit the transcriptional activity of the CLK:CYC dimer, thus repressing their own transcription. As both PER and TIM proteins are degraded before dawn, this process is relieved, lifting repression of the CLK:CYC dimer, and thereby starting another cycle of PER and TIM accumulation.
It has been demonstrated that PER/TIM's transcriptional inhibition of their own genes is due to their direct interaction with the CLK:CYC complex. This physical interaction prevents E-box-binding of CLK:CYC but does not disrupt the association between CLK and CYC (Lee et al., 1999). This might be brought about by an allosterically induced conformational change in the CLK:CYC heterodimer or by a masking of the DNA-binding region in the CLK:CYC heterodimer, leading to a reduction in DNA-binding activity. PER and TIM could be in the CLK:CYC complex together, although the stoichiometry and their modes of interaction have not yet been elucidated.
Other Components for Posttranscriptional Regulation
Molecular genetic studies in Drosophila identify some additional circadian components. These circadian regulatory molecules act to refine the transcriptional/translational feedback loops. Light changes the levels of these additional clock components, resetting the clock to different time of day. In Drosophila, the entrainment of the circadian clock relies on the degradation of TIM in response to light (Zeng et al., 1996). CRYPTOCHROME (CRY) is the major mediator for the rapid light-induced degradation of TIM (Ceriani et al., 1999), which has been demonstrated to be mediated by a ubiquitin-proteasome pathway (Naidoo et al., 1999).
The double-time (dbt) gene encodes a Drosophila homolog of the mammalian casein kinase Iε (Kloss et al., 1998). Phosphorylated forms of PER normally appear after PER protein levels reach a peak and begin to decline. In null mutant flies of dbt gene (dbtP), even though per mRNA levels are normal, hypophosphorylated PER accumulates at higher than normal levels throughout the day, indicating that the effects on the protein are post-transcriptional (Price et al., 1998). In fact, it has been demonstrated that DBT protein physically associates with both PER and the PER:TIM heterodimer, and phosphorylates PER (Kloss et al., 2001). In late subjective day when monomeric PER is synthesized in the cytoplasm, DBT binds to PER and promotes its phosphorylation, leading to the degradation of PER and subsequent accumulation of TIM. High concentrations of TIM promote a formation of the stable complex of PER:TIM:DBT, that can enter the nucleus during early subjective night. Over an 8–10 hr period, the nuclear PER:TIM:DBT complex is converted to the PER:DBT complex, and concomitantly progressive repression of per and tim transcription results in a decreased accumulation of the PER:TIM:DBT complex. Nuclear DBT may progressively phosphorylate PER, leading to its nuclear (Kloss et al., 2001).
The syaggy (sgg) gene encodes another kinase SGG that modulates a further circadian clock component. SGG is a homolog of glycogen synthase kinase-3β, and promotes TIM phosphorylation, which regulates the timing of nuclear entry of the PER:TIM complex (Martinek et al., 2001). Recent studies reveal that Drosophila casein kinase 2β (CK2β) also plays an important role in the molecular clock (Akten et al., 2003). In the Andante fly, which is a mutant of the casein kinase 2β gene, PER and TIM accumulate to abnormally high levels and the nuclear translocation of PER and TIM is significantly delayed. This suggests that CK2β promotes nuclear entry of clock proteins by direct phosphorylation of clock proteins or by activation of a secondary kinase such as DBT and SGG.
Drosophila Slimb protein is a member of the F-box/WD40-repeat protein family of the SCF (Skp1/Cullin/F-box protein) ubiquitin E3 ligase complex that targets phosphorylated proteins for degradation. Recent research reveals that Slimb is an essential component of the circadian clock (Grima et al., 2002; Ko et al., 2002). Flies mutant for slimb are behaviorally arrhythmic, and can be rescued by targeted expression of wild-type Slimb in their clock cells. In head extracts from slimb mutants, highly phosphorylated PER and TIM are constitutively present throughout the day. Slimb interacts preferentially with DBT-mediated phosphorylated PER and targets it for rapid degradation by 26S protea-some. Slimb-dependent periodic degradation of PER and/or TIM is thought to be crucial for the maintenance of normal circadian rhythmicity.
Interlocked Feedback Loops
Several lines of evidence have revealed that PER and TIM can stimulate transcription of the Clk gene and that CLK:CYC heterodimers repress Clk transcription, forming interlocked feedback loops (Bae, et al., 1998; Glossop et al., 1999). These positive and negative aspects of clock proteins are important for the high amplitude modulation of circadian oscillation and for fine-tuning of the circadian clock. Recent studies reveal that VRILLE (VRI), the product of the vrille gene encoding a basic leucine zipper (bZIP) transcription factor, bridges between the interlocked feedback loops of the Drosophila circadian clock (Cyran et al., 2003; Glossop et al., 2003). VRI was first described as a transcription factor essential for signaling in the decapentaplegic pathway during Drosophila embryogenesis (George and Terracol, 1997). Differential-display screenings for clock-controlled transcripts identified vri as an important circadian component (Blau and Young, 1999). Analysis of the promoter sequence revealed that vri contains an E-box, and that expression of vri is indeed driven by CLK:CYC. The vri gene expresses in the LNs, and its mRNA cycles in phase with those of per and tim. Overexpression of vri in clock cells by means of the GAL4-UAS system not only results in a longer period of or arrhythmia in locomotor activity, but also affects severely the expression of PER and TIM (Blau and Young, 1999). Glossop et al. (2003) and Cyran et al. (2003) independently reported that VRI directly regulates Clk transcription by specific binding to Clk promotor elements and functions as a negative component of the Clk feedback loop. In this context, the positive regulation of Clk by PER/TIM occurs indirectly, by preventing CLK:CYC function from activating vri. Such an interconnected feedback mechanism is also found in the mammalian circadian clock. In mice, the orphan nuclear receptor REV-ERBα plays an analogous role to VRI, repressing BMAL1 expression by binding its response element within the BMAL1 promotor.
In addition to the effect on the central clock's components, VRI affects the output system through a peptide named PDF (Blau and Young, 1999). PDF is an insect homolog of pigment-dispersing hormone, PDH, which was isolated through its ability to disperse pigment granules in the visual system of crustaceans, where it was first discovered (Rao and Riehm, 1993). The pdf gene of D. melanogaster encodes 102-amino acids and this precursor protein (prepro-PDF) consists of a signal peptide and a PDF-associated peptide (PAP) followed by the mature 18-amino acid PDF peptide with α-amidation signal (Park and Hall, 1998). The primary structure of Drosophila PDF is highly conserved among other members of the PDH/PDF family. In Drosophila, the spatial expression pattern of PDF is restricted largely to the LNs (Kaneko and Hall, 2000), and in insects, a number of pieces of evidence have revealed that PDF is a possible key output signal from the circadian clock. For example, microinjecting PDF into the cockroach optic lobe shifts the phase of its circadian rhythmicity (Petri and Stengl, 1997). PDF injection into the housefly's optic lobe enlarges the axons of monopolar cells in the first neuropile, or lamina, which normally change their caliber dynamically with a circadian rhythm (Pyza and Meinertzhagen, 1996). This suggests that daytime release of PDF might regulate the diurnal rhythmic size changes of these neurons in vivo. Furthermore, genetic studies on the pdf gene of Drosophila strongly suggest that PDF functions as a circadian transmitter.
From analysis of the null mutation for pdf (pdf01), Renn et al. found that, although flies lacking PDF are well entraining under LD conditions, the evening peak in locomotor activity occurs about 1 hr earlier than that in wild-type flies (Renn et al., 1999). The mutants did not anticipate the light-on signal as well either. After 2–3 days in constant dark conditions, many pdf01 mutants became arrhythmic. In pdf01flies, LNvs are normal in number and morphology, and the mutant phenotypes can be rescued with exogenously introducing wild-type pdf. Although there is no circadian rhythm in the abundance of pdf mRNA in wild-type flies, mutations in the Clk or cyc genes reduced pdf expression and PDFimmunolabeling in the l-LNvs (Blau and Young, 1999). Disruption of vri oscillations also results in reduced PDFimmunoreactivity, but the abundance of pdf mRNA is not affected (Blau and Young, 1999). Notably, in wild-type flies, PDF peptide levels were found to be constant in the somata of LNvs, but to cycle at their axon terminals, indicating the rhythmic PDF release (Park et al., 2000). In null mutants of per and tim genes, pdf mRNA levels are normal, although the circadian accumulation of PDF is eliminated. These results suggest that periodic release of PDF is important for circadian behavioral rhythmicity.
Ectopic expression of PDF in the vicinity of the ordinary targets of LNvs generates behavioral arrhythmia, but in more remote locations it does not (Helfrich-Förster et al., 2000). This indicates that when PDF is released it acts locally as a signal rather than systemically. This was further supported in a recent study by Williams et al. (2001) on the neurofibromatosis-1 (Nf1) gene. That study revealed that Nf1 null mutations generate abnormalities in circadian locomotor activity and enhancement of activity in mitogen-activated kinase (MAPK). In wild-type flies, phospho-MAPK levels are higher during the night, although there is no significant difference in the pdf mutants. Immunocytochemical labeling has revealed a circadian oscillation of activated MAPK in the vicinity of nerve terminals of LNvs, suggesting a functional coupling of PDF to Ras/MAPK signaling (Williams et al., 2001). These studies strongly suggest that the PDF peptide functions as neurotransmitter and exert its activity via a post- and/or pre-synaptic membrane receptor.
Although evidence supporting the possibility that PDF is a pacemaker peptide has also been accumulating, the entity with which PDF interacts, namely the PDF receptor, has yet to be characterized. Interestingly, Chuman et al. (2002) demonstrated that prepro-PDF in the cricket Gryllus bimaculatus contains a putative nuclear localization signal (NLS) instead of an obvious signal sequence for entry into a secre-tory pathway. Moreover, immuno-cytochemical analyses have revealed that Gryllus PDF can localize to the nucleus in a certain group of neurons (Chuman et al., 2002). Growing evidence has revealed that neuropeptides and neurohormones such as angiotensin II (Ang II), dynorphin B, and fibroblast growth factor (FGF) can all function in the nucleus as intracrine molecules (Bouche et al., 1987; Ventura et al., 1998; Cook et al., 2001). In addition, microinjection of PDF to cockroach produces time-dependent phase shifts of loco-motor activity rhythm, indicating that a transient change in PDF level could cause a stable change in molecular components of circadian clock as described above (Petri and Stengle, 1997). Considering these results and the fact that under some circumstances PDF affects components of the master clock, the site of action for PDF may be in the nucleus as well as at a membrane surface receptor.
Another important output gene is lark. The lark gene was identified in a mutagenesis screen for genes affecting the circadian eclosion pattern of Drosophila (Newby and Jackson, 1993). The circadian function of lark appears to be specific for eclosion, since lark mutants have no significant effect on locomotor rhythms. Molecular analysis has shown that the LARK protein is an RNA-binding protein containing RNA recognition motifs and a retroviral-type zinc finger, suggesting that LARK plays a role in regulating protein translation (Newby and Jackson, 1996; McNeil et al., 1999). LARK abundance changes in a circadian manner in Drosophila neurons expressing crustacean cardioactive peptide (CCAP), which is a neuropeptide required to initiate eclosion in several different insects (Zhang et al., 2000). Since lark mRNA does not cycle and LARK protein rhythmicity is abolished in the per0 mutant, the rhythm in LARK is likely to be regulated by a circadian post-transcriptional mechanism (McNeil et al., 1998). The target of the LARK protein is not known yet, but ccap mRNA might be a potential target, in regard to its clock function.
The temporal expression pattern of clock genes and proteins and their interactions have been examined mainly based on the experiments that were carried out by using extracts from whole fly heads or expression systems with cultured cells. It should therefore be noted that the results obtained or the conclusions deduced from such studies may not necessarily reflect the processes in vivo in the clock neurons.
Shafer et al. (2002) have constructed a detailed time course of PER and TIM expression and their subcellular localization patterns in the LNvs by immunocytochemical analysis using a panel of antisera against PER and TIM. Surprisingly, they reveal that PER and TIM have different profiles of nuclear accumulation in the LNvs. PER accumulates in the nucleus during the early night, whereas TIM is restricted to the cytoplasm of the LNvs through much of that time. Moreover, l-LNvs and s-LNvs have a similar timing of PER nuclear localization, but differed by about 3 hr in the nuclear accumulation of TIM. These features are not predicted by the present model of the circadian molecular clock in Drosophila and are contradictory to the concept that PER and TIM function as an obligate heterodimer in their nuclear transport as well as in the suppression of CLK/CYC-dependent transcription. The result that PER appears in nuclei at least 3 hr prior to TIM suggests that most nuclear PER is not complexed with TIM at that time, but it does not preclude a catalytic role for TIM in PER nuclear entry. This could explain the apparent absence of nuclear TIM during the middle of the night, and that PER and TIM translocate into the nucleus as their PER/TIM heterodimer, and that subsequently TIM is exported to the cytoplasm. Such clock protein shuttling is known in mammalian PER proteins (Vielhaber et al., 2001; Yagita et al., 2002). Thus, this possibility needs to be tested in Drosophila.
Nakamura and coworkers investigating the circadian rhythm of spontaneous discharges among individual neurons in the SCN of Clk mutant mice, using slice cultures and dispersed cell cultures (Nakamura et al., 2002), have revealed that circadian rhythms exist in 77% of slice cultures and 15% of dispersed cell cultures. These results are also inconsistent with current hypotheses about molecular clocks. According to the model for the mammalian molecular clock, if any of the core clock components are lost, then circadian oscillations will also be abolished. Thus, even widely accepted molecular models of circadian clocks are not always able to explain all experimental evidence. Thus, it is clearly important for the elucidation of the molecular mechanisms of the clock to investigate the expression patterns of clock genes and also the localization and the interactions of clock proteins, and to do this within a single clock cell and/or clock tissue in situ.
We express our sincere gratitude to Prof. Ian A. Meinertzhagen (Dalhausie University, Halifax, Canada) for his kind reviewing the manuscript and correction of the English.