Central and peripheral circadian oscillator mechanisms in flies and mammals

Circadian clocks are molecular time-keeping mechanisms that reside in a diverse range of cell types in a variety of organisms. The primary role of these cell-autonomous clocks is to maintain their own ∼24 hour molecular rhythm and to drive the rhythmic expression of genes involved in physiology,metabolism and behavior. The key features of a circadian clock are its ability to synchronize (entrain) to environmental time cues, or zeitgebers, such as light-dark or temperature cycles, and to maintain rhythmic function when placed in constant conditions.

In mammals the central `master' clock is located in the suprachiasmatic nucleus (SCN) of the anterior hypothalamus(Klein et al., 1991). The SCN is entrained to light-dark (LD) cycles by a distinct set of photosensitive retinal ganglion cells that project to the SCN through the retinohypothalamic tract (Berson et al., 2002;Hattar et al., 2002;Moore et al., 1995;Provencio et al., 2002). The SCN in turn activates rhythms in behavior (e.g. locomotor activity), by secreting factors (e.g. TGFα and prokineticin) that act locally within the hypothalamus (Cheng et al.,2002; Kramer et al.,2001), and entrains subservient circadian oscillators in peripheral tissues (e.g. liver and kidney) via humoral signals (e.g. glucocorticoids) (Balsalobre et al.,2000a; Balsalobre et al.,2000b; Oishi et al.,1998; Ueyama et al.,1999). Such peripheral oscillators can, however, become uncoupled from the SCN if their specific needs dictate — as occurs in liver, lung and skeletal muscle after entrainment by food(Yamazaki et al., 2000). The SCN maintains robust (>2 week) rhythms when entrained to LD cycles in vivo and cultured in vitro, whereas peripheral oscillators lose rhythmicity after just 4-5 days under the same conditions(Yamazaki et al., 2000).

In flies the central clock is located in a group of 5-6 bilaterally symmetric small ventral lateral neurons (sLN_vs) situated in the lateral brain close to the optic lobes(Helfrich-Forster, 1996). As in the SCN, sLN_vs receive light input from retinal photoreceptors in the compound eyes and extra-retinal photoreceptors within the brain;however, they can also be entrained directly by light that penetrates the cuticle (Helfrich-Forster et al.,2001; Stanewsky et al.,1998). In constant dark (DD) conditions, sLN_vs maintain robust rhythms in gene expression and locomotor activity(Ewer et al., 1992;Frisch et al., 1994;Helfrich-Forster, 1998;Zerr et al., 1990). Peripheral oscillators in flies (e.g. antennal clock cells and Malpighian tubules) can also maintain robust (>7 day) rhythms in cell culture, which suggests that fly peripheral oscillators depend less on the sLN_vs than do their mammalian counterparts on the SCN(Emery et al., 1997;Giebultowicz and Hege, 1997;Plautz et al., 1997). Indeed,precisely how much influence the sLN_vs have over fly peripheral oscillators is not known, given that light penetrating through the cuticle appears to entrain peripheral clocks, thus negating the requirement for a`master' clock to synchronize other oscillators in the fly. Despite this difference, the sLN_vs can be regarded as a central oscillator because, like the SCN, they drive behavioral rhythms in locomotor activity.

The core components involved in central and peripheral clocks of flies and mammals are largely conserved and in both they form interlocked feedback loops in transcription/translation (Fig. 1, Table 1). In flies, most of the information regarding the oscillator mechanism has derived from studies of mRNA and protein levels in whole head extracts. As such, the majority of information concerns oscillators in head peripheral tissues (i.e. photoreceptors and antennae), because the sLN_vs make up only a small percentage (<1%) of the total number of oscillators in the head. In mammals, however, previous work focused primarily on the central SCN oscillator because of the desire to link it to the behavioral rhythm in locomotor activity. Daily fluctuations in mRNA and protein levels can also be readily monitored in the SCN, whereas a similar examination of most clock components in the fly sLN_vs is not possible. Thus, previous comparisons between these species have largely compared fly peripheral oscillators with the mouse central oscillator.

Although the two systems show striking similarities, they also show some equally important differences. Here we outline current understanding of the fly peripheral and mouse central oscillators, paying particular attention to the differences between the two. We then consider more recent findings that suggest that peripheral oscillators of mammals share more in common with those of the fly than the mammalian central oscillator and, hence, that the central oscillators of both organisms may have acquired novel regulatory features that make them specialized versions of a more basic design.

Two interlocked transcriptional/translational feedback loops operate in the Drosophila circadian oscillator: a period/timeless(per/tim) loop and a dClock (dClk) loop(Fig. 1A;Table 1). These feedback loops have recently been reviewed in great detail(Allada et al., 2001;Meyer-Bernstein and Sehgal,2001; Young and Kay,2001); thus we provide only a brief overview. In the per/tim loop, per and tim transcription is activated by dClock-Cycle (dCLK-CYC), a basic-helix-loop-helix (bHLH)/PAS(PER-ARNT-SIM) protein heterodimer that binds canonical E-box sequences(CACGTG) in the per and tim promoters(Allada et al., 1998;Darlington et al., 1998;Hao et al., 1997;Rutila et al., 1998). The Period (PER) and Timeless (TIM) proteins accumulate in the cytoplasm as PER-TIM heterodimers after a substantial phosphorylation-dependent delay and then translocate into the nucleus (Gekakis et al., 1995; Kloss et al.,1998; Martinek et al.,2001; Price et al.,1998; Saez and Young,1996; Vosshall et al.,1994). Nuclear PER-TIM then bind to dCLK-CYC, thereby repressing per and tim transcription(Bae et al., 2000;Lee et al., 1999). Once PER and TIM are degraded, another cycle of per and timactivation can begin. In the dClk loop, dClk transcription is repressed by dCLK-CYC (Glossop et al.,1999), probably indirectly since there are no canonical E-boxes in or around dClk (N.R.J.G. and P.E.H., unpublished). Repression of dClk is relieved when PER-TIM binds to dCLK-CYC, which (indirectly)promotes dClk transcription. The opposing effects of dCLK-CYC and PER-TIM on transcription ensure that per/tim and dClk mRNA transcripts oscillate out of phase (Table 1). In DD conditions the oscillations in per/timtranscription persist for a week or more(Brandes et al., 1996;Stanewsky et al., 1998). dClk mRNA oscillations in DD have not been extensively studied,although dClk mRNA does cycle after 2 days in constant darkness(Bae et al., 1998). cyc mRNA and protein levels are constant throughout the circadian cycle (Bae et al., 2000;Rutila et al., 1998;Wang et al., 2001).

Light entrains the oscillator through a blue-light-responsive pterin/flavin-binding protein, cryptochrome (CRY)(Fig. 1A;Table 1). CRY can bind TIM in response to light, ultimately causing TIM degradation and predictable phase advances or delays in the oscillator(Hunter-Ensor et al., 1996;Lee et al., 1996;Myers et al., 1996;Saunders et al., 1994;Zeng et al., 1996). Transcription of cry is also under control of the oscillator(Table 1). The crymRNA profile reflects that of dClk, albeit delayed by several hours(Egan et al., 1999;Emery et al., 1998;Ishikawa et al., 1999). cry and dClk mRNA levels are similarly perturbed in several clock mutants. In dClk^Jrk and cyc⁰¹ mutants, dClk and crymRNAs are constitutively high, whereas in per⁰¹and tim⁰¹ mutants dClk and cryare constitutively low (Emery et al.,1998). This suggests that cry and dClk are under similar transcriptional regulation. Unlike the robust per/timoscillations, cry mRNA oscillations dampen quickly, tending to trough levels by the third day in DD (Ishikawa et al., 1999).

Although CRY can entrain peripheral oscillators in the fly tissue-autonomously (Emery et al.,2000b), light input to the sLN_vs can also occur through retinal and extra-retinal photoreceptors(Helfrich-Forster et al.,2001). As such, the sLN_vs can still entrain in flies that contain a single amino acid mutation that disrupts the flavin-binding domain of CRY (cry^b). Entrainment in these flies to pulses of light is, however, impaired compared with the wildtype(Stanewsky et al., 1998).

In mammals, two interacting transcriptional/translational feedback loops also operate (Fig. 1B;Table 1): an mPer/mCryloop and a Bmal1 loop (reviewed byAllada et al., 2001;Meyer-Bernstein and Sehgal,2001; Young and Kay,2001). The mammalian homolog of dCLK, mCLK, forms heterodimers with BMAL1 (the mammalian CYC homolog) and drives rhythmic transcription from E-boxes in three mammalian Period genes (mPer1, mPer2 and mPer3) (Gekakis et al.,1998; Hogenesch et al.,1998; Jin et al.,1999). Of these genes, mPer1 and mPer2 are further involved in the oscillator; mPer3 appears to be involved in output rhythms (Bae et al.,2001). mCLK-BMAL1 is also thought to activate transcription of the two cry homologs in mammals, mCry1 and mCry2. Activation of mCry1 presumably occurs through a canonical E-box situated ∼300 bp upstream of the mCry1 promoter(Kume et al., 1999). No such E-box has been identified at the mCry2 locus, but mCry2 is believed to be regulated similarly to mCry1 and the mPersgiven that they exhibit similar mRNA cycling phases(Table 1) and that mPer and mCry transcript levels are low in mClkmutant mice (Shearman et al.,2000).

As in Drosophila, mPER1 and mPER2 proteins undergo a substantial phosphorylation-dependent delay in accumulation with respect to their mRNAs(Akashi et al., 2002;Lowrey et al., 2000;Takano et al., 2000). mPER1 and mPER2 peak in concert with the mCRYs(Table 1), form mPER-mCRY complexes, and translocate to the nucleus(Griffin et al., 1999;Kume et al., 1999;Vielhaber et al., 2001;Yagita et al., 2002). mCRYs(perhaps in conjunction with mPERs) then act as potent negative regulators by directly interacting with the mCLK-BMAL1 heterodimer(Griffin et al., 1999;Kume et al., 1999;Lee et al., 2001). mPER2 has also been shown to positively regulate Bmal1 transcription, similarly to dPER positive regulation of dClk. However, mPER2-dependent activation of Bmal1 is thought to occur indirectly via coactivation or nuclear shuttling of an activator rather than inhibition of a repressor as in the case of dClk (Shearman et al., 2000). This mechanistic difference explains why low levels of Bmal1 transcript are seen in the SCN of mClk mutant mice:activators and repressors of Bmal1 transcription are both dependent on mCLK-BMAL1.

Light entrains the SCN by inducing transcription of mPer1 and mPer2 (Albrecht et al.,1997; Albrecht et al.,2001; Shearman et al.,1997; Shigeyoshi et al.,1997; Zylka et al.,1998b). Analysis of mCRYs in the SCN initially suggested that they do not play a role in entrainment analogous to that of CRY in Drosophila: the mPer genes are still activated in response to light in mCry double-knockout mice(Okamura et al., 1999). However, the mCrys were recently shown to be involved in entraining the SCN to light (Selby et al.,2000; Thompson et al.,2001), but this role is carried out in the retina, where opsin-based photoreceptors also function to entrain the SCN(Berson et al., 2002;Freedman et al., 1999;Hattar et al., 2002;Provencio et al., 2002).

The circadian oscillator mechanism in peripheral tissues appears to be different from that in the SCN. In peripheral tissues, Bmal1regulation is similar to that of dClk in the fly, and different from Bmal1 regulation in the SCN, since Bmal1 mRNA is constitutively high in mClk mutant mice(Oishi et al., 2000). This result implies that mCRY (and perhaps the mPERs) stimulate Bmal1expression by relieving mCLK-BMAL1-dependent repression of Bmal1. This mCLK-BMAL1-dependent repression can occur directly, through interactions with either the Bmal1 promoter or the Bmal1 activator protein, or indirectly by activation of a Bmal1 repressor that interacts with the Bmal1 promoter or the Bmal1 activator itself (Fig. 1B). Intriguingly,mouse Bmal1 knockouts have low levels of the mutant Bmal1mRNA in liver (Bunger et al.,2000). It is difficult to reconcile this result with the high levels of wild-type Bmal1 mRNA in the liver of mClk mutant mice, although loss of positive regulatory elements, alteration of mutant Bmal1 mRNA stability, or transcriptional interference from the selectable marker could lead to low levels of mutant Bmal1 mRNA in the Bmal1 knockout. Even though genetic elements involved in mCLK-BMAL1-dependent repression are clearly present at the Bmal1locus, a different mechanism operates in the SCN, because Bmal1transcripts are tonically low in mClk mutant mice. Thus, positive regulation of Bmal1 by mPER2 in the SCN must occur through a separate, or redundant, mechanism.

Recent in vivo studies in mouse liver have shown that mPER1 and mPER2 are rate limiting for the nuclear translocation of mCRYs and, in the SCN, mPERs require mCRYs to achieve and maintain nuclear translocation(Shearman et al., 2000;Vielhaber et al., 2001;Yagita et al., 2002). The low levels of Bmal1 mRNA in mCry-deficient mice and the dependence of mPERs and mCRYs on each other for nuclear translocation suggest,therefore, that nuclear mPER and/or mCRY is necessary for the positive regulation of Bmal1. Whether it is nuclear mPER, mCRY or both, the positive regulatory effect of these factors on Bmal1 transcription(via relief of mCLK-BMAL1-dependent repression) ensures that the mPers and Bmal1 cycle in anti-phase with one another(Table 1).

An obvious difference between the fly and mammalian oscillators is that mCRYs appear to have taken over the role of TIM in facilitating mPER nuclear localization and, indeed, it is now thought that a mouse TIM homolog (mTIM)does not operate in central or peripheral mammalian oscillators(Field et al., 2000;Zylka et al., 1998a). So, if mCRYs have replaced mTIM in the mouse circadian oscillator, how do the functions of mCRYs and dCRY differ?

A role for Drosophila CRY as an inhibitor of dCLK-CYC (analogous to that of mCRY in inhibiting mCLK-BMAL1) has not been investigated. However,in LD cycles, CRY starts to accumulate in the dark at a time when per/tim transcription (by dCLK-CYC) first starts being repressed(Table 1). Given the role of mCRYs in the mammalian SCN, re-examination of previous results in the fly also suggests that mCRY and dCRY are more functionally conserved than was previously believed. For example, in Drosophila cell culture, CRY nuclear localization is dependent on both PER and TIM (CRY and TIM alone are insufficient) and appears to be induced by light(Ceriani et al., 1999). However, CRY-TIM interactions are also evident in the dark(Ceriani et al., 1999;Rosato et al., 2001). Further,flies containing the cry^b mutation have constitutive intermediate levels of per and tim transcripts and PER and TIM protein in whole heads (Stanewsky et al., 1998). In the photoreceptors of these mutants TIM protein,and presumably PER and PER-TIM, remains predominantly cytoplasmic(Stanewsky et al., 1998). These phenotypes (i.e. cytoplasmic TIM and intermediate levels of perand tim mRNAs) precisely mimic those of per⁰¹flies (Hunter-Ensor et al.,1996), which suggests that CRY and PER cooperatively translocate to the nucleus in fly peripheral clocks. One caveat is that nuclear TIM, and presumably PER, is present in the renal Malpighian tubules of cry^b mutants(Ivanchenko et al., 2001). It will be interesting to see whether, as is the case for mPER in mCry-deficient mice, such staining is actually peri-nuclear(Shearman et al., 2000).

In Drosophila peripheral oscillators CRY might therefore facilitate PER-TIM nuclear translocation, or vice versa, just as mPERs and mCRYs require each other for nuclear entry in mammals. Such a conserved role is supported by the recent finding that dCRY can engage in protein-protein interactions with dPER independently of TIM(Rosato et al., 2001). Thus,in addition to mediating PER nuclear localization in flies, TIM might also facilitate movement of CRY into the nucleus. In mammals, mPERs and mCRYs have bypassed this requirement for TIM, thus rendering TIM redundant. Hence,although the roles of TIM in flies and mice have clearly diverged, the functions of the PERs and CRYs might still have been largely retained in fly and mouse peripheral tissues and the SCN, but not the sLN_vs, where CRY is not critical for oscillator function(Emery et al., 2000a;Emery et al., 2000b;Stanewsky et al., 1998).

A second difference between oscillators in the mouse SCN and fly periphery concerns the transcriptional activator that is rhythmically expressed: dClk in flies (mClk is constitutively expressed) and Bmal1 in mammals (cyc is constitutively expressed). Currently, the activators of dClk and cyc and of mClk and Bmal1 are not known; thus any explanation for this`switch' remains speculative. Interestingly, however, dClktranscription is repressed by dCLK-CYC(Glossop et al., 1999),although this repression is probably indirect owing to the absence of consensus E-boxes in or around dClk (N.R.J.G. and P.E.H.,unpublished). Recently, Yu et al., have shown in vitro that mCLK-BMAL1 heterodimers can, either directly or indirectly, repress Bmal1transcription (Yu et al.,2002). Further, a functional analog of mCLK, NPAS2, that can take the place of mCLK in some mammalian peripheral oscillators (e.g. forebrain)can also repress Bmal1 transcription in cell culture when co-transfected with BMAL1 (Reick et al.,2001). Importantly, these in vitro results accurately reflect Bmal1 regulation in vivo in peripheral tissues(Oishi et al., 2000). These findings suggest that negative feedback loops from the positive elements of the fly and mouse oscillators (i.e. dCLK-CYC repression of dClk and mCLK/NPAS2-BMAL1 repression of Bmal1) operate through similar, but as yet unknown, mechanisms.

So why is dClk rhythmically expressed in flies and Bmal1rather than mClk rhythmically expressed in mammals? dClk and cyc are both bHLH-PAS-protein-encoding genes that presumably derived from a single common ancestor and were once under the same regulatory control. Thus, after activation, both genes would have been subject to dCLK-CYC-dependent repression. Since the cyclic accumulation of the dCLK-CYC dimer would require only one of these proteins to cycle, regulation of the other could be more flexible. Hence, elements concerned with cycrepression were lost, and the fly oscillator appears to have chosen cyclic regulation of dClk over cyc. Conversely, in mammals, Bmal1 was chosen over mClk, and so mClk regulation became more flexible. In support of this theory is the fact that in all mammalian clock tissues examined thus far, Bmal1 is always present and cycles at the mRNA level (Table 1). mClk and Npas2, by contrast, show variability in their spatial expression and cyclic nature. For instance, mClk expression in the SCN is constitutive, whereas Npas2expression cycles in the vasculature (seeTable 1 and below).

Do dClk and Npas2 share any similarities in their regulation? Recent in vivo work on the mouse vasculature has shown that Npas2 mRNA transcripts cycle. This suggests that, unlike its constitutively expressed analog mClk, Npas2 regulation is more like that of dClk. Bmal1 transcripts also cycle in the vasculature,peaking at the same time as those of Npas2(Table 1). This observation is particularly interesting because it shows that the oscillatory phases of both positive elements in the mouse vasculature are more similar to those of the fly than to those of the mouse SCN.

A third difference between fly peripheral and mouse central circadian oscillators is the anti-phase cycling of the core transcripts: mPer1and mPer2 expression peaks at CT 04-10 hours whereas Drosophila per peaks at CT 14-15 hours, and Bmal1 peaks at CT 12-21 hours whereas dClk peaks at CT 23-05 hours (seeTable 1)^†Note that, in a variety of mammalian peripheral tissues, Per1 and Per2 transcripts reach peak levels in the early night (ZT/CT 09-17 hours) (Table 1); moreover,luciferase reporter gene assays in rats show that mPer1 oscillations in DD in cultured peripheral tissues peak in the mid to late night — in anti-phase with the SCN oscillations(Yamazaki et al., 2000). Likewise, the patterns of dClk and Bmal1 cycling are also similar in the fly and mouse periphery: dClk mRNA peaks at ZT/CT 23-04 hours, whereas Bmal1 transcripts peak at ZT/CT 21-05 hours(Npas2 mRNA in the vasculature also peaks at ZT/CT 01-05 hours). The phases of the Per and Bmal1 oscillations in mammalian peripheral tissues therefore closely reflect those in the fly and are dramatically different from those in the SCN, which peak much earlier. The entrainment of fly and mammalian peripheral oscillators is necessarily different since, in flies, light acts directly through CRY(Emery et al., 2000a;Emery et al., 2000b;Stanewsky et al., 1998), but in mammals it acts indirectly via the SCN. This indirect action, mediated by humoral factors, imposes a delay that is likely to account for the difference in the phases of mammalian central and peripheral oscillators.

Cultured mammalian fibroblasts also contain a circadian clock that is set by various pharmacological agents (e.g. serum shock, dexamethasone and endothelin 1) thought to mimic endogenous humoral factors(Balsalobre et al., 2000a;Balsalobre et al., 1998;Yagita et al., 2001). The oscillator in these cells has properties similar to those of that in other mammalian tissues: Per mRNA cycles in antiphase to Bmal1mRNA; PER1 and PER2 accumulation is delayed compared with their peak mRNA levels; clock gene expression can be entrained by SCN cells; and clock function is dependent on Cry genes(Allen et al., 2001;Balsalobre et al., 1998;Yagita et al., 2001). The oscillator mechanism in cultured fibroblasts is presumed to be like that in other peripheral tissues, but regulation of the Bmal1 loop has not been tested to determine whether this is the case. Moreover, the phase of clock gene expression is expected to be similar to that in other peripheral tissues, but it is difficult to assess the phase of expression relative to an LD cycle since the clock is set by pharmacological agents.

Interestingly, peripheral oscillators in other vertebrates have phases closer to the SCN than to peripheral tissues of mammals. In the Xenopus eye, xPer1, xCry1, xCry2a and xCry2b mRNAs peak at ∼CT 18-02 hours (Zhu and Green, 2001; Zhuang et al.,2000) and Bmal1 peaks at CT 11-15 hours (C. Green,personal communication). A similar situation exists in the chicken pineal gland: cCry2 and cPer2 mRNAs peak at CT 22-02 hours and cBal1 and cBmal2 mRNAs peak at CT 10-14 hours(Bailey et al., 2002;Fukada and Okano, 2002;Yamamoto et al., 2001). Zebrafish peripheral oscillators (e.g. those in the liver, eye, pineal gland,kidney and an embryonic cell line) are unusual because they are directly light entrainable, as is the case with Drosophila oscillators, but phase shifts are mediated by the light-induced expression of mPer2, as in the mammalian SCN (Pando et al.,2001; Whitmore et al.,2000). The phase of zebrafish peripheral oscillators is similar to those of Xenopus and chickens: zfPer1, zfPer2 and zfPer3 peak at ZT 00-06 hours and zfClk, zfBmal1 and zfBmal2 peak at ZT 12-18 hours(Cermakian et al., 2000;Pando et al., 2001;Whitmore et al., 1998). Thus,in non-mammalian vertebrates, the phase of rhythmically expressed clock genes in peripheral tissues is a few hours earlier than the mammalian SCN and almost 12 hours earlier than mammalian peripheral tissues. In each case, the rhythmic repressors are essentially antiphase to the rhythmic activators, owing to the interlocked feedback loop mechanism.

Previous comparisons between the roles of mCRY and dCRY suggested that their functions had diverged significantly. mCRY appeared to be a core clock component that has no photoreceptive properties, whereas dCRY seemed to be a cell-autonomous photoreceptor that conveys light information to the oscillator without being intimately embedded in the core workings.

Recent work has shown that, in the fly, CRY may serve both roles. In the central oscillator it serves as the predominant photoreceptor but is dispensable within the core clockwork: the cry^b mutant has wild-type free-running locomotor activity in constant light conditions that render wild-type flies arrhythmic (Emery et al., 2000a). In peripheral oscillators, however, CRY may well serve as an integral clock component. Two key experiments have shown this. The first relies on the fact that the circadian clock in flies can be entrained by temperature cycles, thus, allowing the photoreceptive role of CRY to be uncoupled from its oscillator function. When wild-type flies are entrained in DD to temperature cycles and then transferred to constant temperature they show clock-driven behavioral rhythms in locomotor activity (which are LN dependent) and in electroantennogram (EAG) responses to odorant (an antennal output that is thought to be driven by oscillator cells in the antennae)(Krishnan et al., 2001;Wheeler et al., 1993). When the cry^b mutant is treated similarly it retains a functional sLN_v oscillator but loses EAG rhythms, which shows that the peripheral antennal clock is non-functional(Krishnan et al., 2001). Importantly, loss of the EAG output rhythm in temperature-entrained cry^b flies results from a non-functional oscillator,because rhythmic per and tim mRNA transcription in antennae is severely crippled compared with that in temperature-entrained wild-type flies (Krishnan et al., 2001). Although it remains possible that CRY is also a thermoreceptor, the most parsimonious interpretation of this result is that the non-functional antennal clock results from a dependence of the oscillator on CRY. The second experiment focused on the renal Malpighian tubules (Mts) and the oscillating nature of the clock proteins PER and TIM in this tissue. Ivanchenko et al. showed that, in cry^b flies, the oscillations of PER and TIM in Mts are disrupted in LD conditions and abolished in DD, whereas PER/TIM oscillations in the sLN_vs are essentially wild-type in LD and DD(Ivanchenko et al., 2001). Taken together, these data indicate that CRY is an essential component of the oscillator in peripheral, but not central, clocks.

In mice, a similar problem of dissecting the role of mCRYs as photoreceptors was encountered. Because mCRYs are a core component of the oscillator, mCRY-double-knockout mice exhibit arrhythmic behavior (i.e. the central oscillator is non-functional) (van der Horst et al., 1999;Vitaterna et al., 1999). mCRYs were not believed to have photoreceptive properties, because in the SCNs of these mice, light pulses administered in vivo can induce transcription of core clock genes: mPer1 and mPer2(Okamura et al., 1999). The interpretation was that photic input to the oscillator can occur in the absence of mCRYs. Freedman et al. and Selby et al., have shown that mice lacking either classical opsin-based photoreceptors [retinal degeneration (rd/rd) mutants missing rods and most cones] or mCry1 and mCry2 can in fact entrain to LD cycles(Freedman et al., 1999;Selby et al., 2000). However,mice lacking both photoreceptor types have severely impaired ability to entrain, which indicates that opsin and mCRY have redundant roles in the transduction of light information (Selby et al., 2000). Further support for mCRYs involvement in light transduction comes from experiments eliminating opsin-based photoreception by ocular retinaldehyde deprivation, which leaves photic signaling to the SCN intact (Thompson et al.,2001). The discovery of another mammalian photoreceptor,melanopsin, in retinal ganglion cells that project to the SCN also suggests that, in common with the sLN_vs in Drosophila, the SCN has redundant mechanisms for entrainment(Hattar et al., 2002;Provencio et al., 2002).

These studies suggest that dCRY and mCRYs function similarly in flies and mammals. In both, they serve as core clock components and/or as photoreceptors depending on the tissue. A similar situation exists in zebrafish, but the six zfCry genes in this species have become specialized, functioning as either core clock components or as photoreceptors(Cermakian et al., 2002;Kobayashi et al., 2000). Since photoentrainment of zebrafish peripheral tissues is dependent on blue-light photoreceptors, and therefore probably cryptochromes(Cermakian et al., 2002), it will be interesting to determine how CRY photic signaling has evolved from a TIM-degradation-based light entrainment mechanism in Drosophila to a zfPer2-induction-based mechanism in zebrafish.

As studies of circadian oscillators are rapidly progressing three important features are becoming clear. (1) Central and peripheral oscillator mechanisms are different in flies. The sLN_vs do not require CRY for oscillator function, whereas peripheral oscillators do. A further example of central versus peripheral differences is found in the moth, in which PER nuclear cycling is absent in the brain even though it cycles in peripheral tissues much in the same way as it does in Drosophila(Sauman and Reppert, 1996). Thus, even between species as closely related as flies and moths, the central oscillator has acquired species-specific differences. (2) Central and peripheral oscillator mechanisms in mammals are different. The Bmal1negative-feedback loop in the SCN is regulated differently than that in peripheral tissues. (3) Peripheral oscillators in flies and mammals show more similarities within their molecular workings than they do with their respective central oscillators. This is most evident when one considers the requirement for CRY and the analogous regulation of the dClk/Bmal1negative-feedback loops, which suggest that the function of core clock components is highly conserved. Thus, in order to advance our understanding of the `general oscillator', comparisons between species should focus on peripheral oscillators. From this, it should then be easier to identify the species-specific modifications of the central oscillator that have arisen.

We thank P. E. Hardin lab members and Greg Cahill for helpful discussions,and Wangjie Yu and Hitoshi Okamura for comments on the manuscript. This work was supported by grants from the National Institutes of Health (DC04857 and MH61423) to P.E.H.