Circadian Transcription Contributes to Core Period Determination in Drosophila

Circadian rhythms are widespread in nature and help to maintain internal temporal order as well as anticipate daily environmental changes [1]. They use self-sustained biochemical oscillators that generate oscillations at the molecular, physiological, and behavioral levels [2,3].

Results over the past 15 years have highlighted the importance of transcription to circadian biology [4]. In eukaryotic systems, a large fraction of mRNAs, perhaps 10% or more, undergoes circadian transcription (e.g., [5,6]). Circadian transcriptional oscillations contribute to myriad physiological and behavioral outputs in diverse tissues of eukaryotic organisms (e.g., [5,7–10]). Recent data from humans, mice, and flies indicate that numerous syndromes and even pathologies result from a disruption of these daily oscillations [11–16].

A conserved heterodimeric transcription factor, constituted by the proteins Clock and BMAL1 (CLK–BMAL) in mammals and Clock and Cycle (CLK–CYC) in flies, sits at the top of the system that generates circadian transcriptional oscillations [17–22]. These complexes direct the transcription of direct target genes, some of which encode repressors of the activity that leads to their transcription. These repressor proteins, chiefly Timeless and Period in flies or Cryptochrome and Period in mammals, accumulate over the course of many hours and ultimately result in the repression of CLK–CYC or CLK–BMAL activity, respectively [23–28]. A complete cycle takes approximately 24 h and is entrained or reset to exactly 24 h by the daily light–dark (LD) cycle.

These transcriptional cycles constitute the core circadian transcriptional feedback loop of flies and mammals. There are also subsidiary loops involving additional repressors and activators, but genetic evidence indicates that they are less important to circadian timekeeping [29–31]. The circadian transcriptional feedback loop was originally proposed in flies and based on the circadian oscillation of per transcription as well as the role of PER in the parallel timing of behavioral and transcriptional oscillations [23,25]. Subsequent evidence made a direct role of PER in transcriptional repression more likely [18,24,32–34].

There is also an important contribution of post-transcriptional and post-translational regulation to circadian timekeeping in both the fly and the mammalian systems. In Drosophila, genetic evidence indicates that major alterations in circadian period result from mutations of key kinase genes, and there is similar evidence in mammals. For example, the key Drosophila clock gene doubletime (dbt) encodes CKIɛ, and its mammalian relative is also a clock gene [35–37]. This importance of phosphorylation to circadian timekeeping even derives from studies of humans with advanced sleep phase syndrome [11–13,38]. Manipulation of phosphatase activities within Drosophila clock cells also affects circadian period [39,40].

Major targets of these post-translational modifications appear to be the transcriptional repressors PER and TIM. Their modification status as well as the rates with which these modifications take place have a major influence on their degradation rate [35,38,41–50]. Modification of PER may additionally influence its transcriptional repressor activity or the timing of this activity [34,38,51–53]. It is also likely that the repression of CLK–CYC activity occurs at least in part via CLK phosphorylation, which may be mediated by a PER–DBT complex and/or a PER–TIM–DBT complex [42–44,54].

The importance of post-translational modification to period determination has been strengthened by recent results from cyanobacteria [55]. The three key clock proteins—KaiA, KaiB, and KaiC—are transcription factors. However, recombinant versions of these proteins undergo circadian oscillations of association and modification state in vitro (KaiC has autokinase and autophosphatase activity) in the absence of transcription and without nucleic acids [56,57]. These results make it very likely that the core circadian system in cyanobacteria is predominantly if not exclusively post-translational and suggest that circadian transcriptional regulation is a downstream output feature, unnecessary for core circadian timekeeping. This raises the possibility that a similar situation occurs in flies and mammals: the core circadian system may be primarily post-translational (e.g., based on the temporal modification of PER and TIM). Consistent with this notion, Yang and Sehgal have shown that circadian locomotor activity rhythms can occur with per- and tim-expressing transgenes missing their natural promoters [58]. This work extended previous indications that behavioral rhythms require PER activity but do not require circadian transcription of the per gene [59].

To pursue the contribution of transcription to core circadian timekeeping in Drosophila, we have analyzed the in vivo effects of a CYC–viral protein 16 (VP16) fusion gene. VP16 is a potent transcriptional activator derived from Herpes virus [60] and imparts to the CLK–CYC-VP16 complex enhanced transcriptional activity relative to the normal CLK–CYC heterodimeric complex. This is based on activity in S2 cells as well as flies expressing CYC-VP16. These flies also have increased levels of CLK–CYC direct target gene mRNAs, including those from per and tim. Moreover, the CYC-VP16-expressing flies have short periods, implicating circadian transcription in period determination. Taken together with more detailed molecular analyses of these flies as well as behavioral assays of strains missing the normal per promoter, we suggest that CLK–CYC-mediated transcription of the per gene is important for period determination.

To manipulate the transcriptional activation potential of the CLK–CYC heterodimer, we generated a fusion protein between the CYC protein and the strong and well-characterized viral transcriptional activator VP16 (Figure 1A) [60]. Current indications are that all activator activity of the CLK–CYC heterodimer normally comes from the polyglutamine region of CLK (Figure 1A) [18], so we considered that VP16 might increase the activity of a CLK–CYC-VP16 heterodimer. As an initial assay, DNA encoding the fusion protein was transfected into S2 cells along with a standard timeless promoter-luciferase (tim-luc) reporter gene [18,61], which responds well to CLK–CYC activity.

Transfection of the fusion protein gene has little or no activity (Figure 1B). This is expected and reflects the absence of its partner CLK from S2 cells [18]. In contrast, transfection of a CLK gene alone partners with endogenous CYC and potently increases reporter gene activity (Figure 1B), identically to what has been reported previously [18]. Cotransfection of CLK with CYC-VP16 increases activity a further 5-fold (Figure 1B), which presumably reflects the transcriptional activation potential of VP16. Importantly, cotransfection of CLK with CYC or with another VP16 fusion protein (GAL4-VP16) has no effect over transfection with CLK alone (Figure S1A and unpublished data). An assay of endogenous tim mRNA expression by real-time PCR and TIM protein by western blotting gives rise to similar results: CYC-VP16 alone has no activity, whereas CLK plus CYC-VP16 cotransfection has considerably more activity than CLK alone (Figure 1C and Figure S1B). Moreover, coexpression of CYC-VP16 rescues activity of the truncated CLK^Jrk protein in this tissue culture assay system (Figure S1C); CLK^Jrk is missing most of its activation domain [17].

CLK-driven transcription is inhibited by double-stranded RNAs (dsRNAs) against the 5′ and 3′ untranslated regions (UTRs) of the endogenous cyc mRNA present in S2 cells (Figure 1B and 1C). In contrast, activity due to cotransfection of CLK and CYC-VP16 is insensitive to incubation with the same dsRNAs (Figure 1B and 1C). This is because the CYC-VP16 expression plasmid does not carry the cyc UTRs. The result indicates that most CLK activity is derived from the CLK–CYC-VP16 heterodimer. Neither CLK–CYC nor CLK–CYC-VP16 has activity on a tim-luc reporter with mutant E-boxes [61], indicating that the CLK–CYC-VP16 fusion has DNA-binding properties similar to wild-type CLK–CYC (Figure 1B). Because there is no detectable endogenous per expression in S2 cells, even after clk expression [62], the higher target gene mRNA levels are likely the consequence of a stronger transcriptional activation independent of any possible weaker PER-mediated repression on CYC-VP16.

Cotransfection with per cDNA inhibits CLK–CYC-VP16 activity, similar to what is observed for CLK–CYC activity (Figure 1D) [18,34] Given the entirely different nature of the VP16 activator compared to the polyglutamine region of CLK and the 5-fold increase in activity, this suggests that per repression involves a similar inhibition of CLK–CYC and CLK–CYC-VP16, probably an inhibition of DNA binding [54]. The similar properties of the two heterodimers are despite the much more potent activity of the former.

To generate flies with cyc-vp16 expression in circadian cells, we created uas-cyc-vp16 transgenic flies and crossed them to tim-gal4 driver lines. We then assayed circadian locomotor behavior in these tim-cyc-vp16 flies (Figure 2A and 2B, top). They were robustly rhythmic with ∼22-h periods, approximately 2 h shorter than those of wild-type flies. Figure 2C summarizes comparable period shortening by uas-cyc-vp16 combined with a highly spatially restricted circadian driver (pdf-gal4) and with two broader expression drivers (actin-gal4 and the pan-neuronal elav-gal4). This indicates that the ∼22-h period is not an idiosyncrasy of the tim-gal4 driver. Moreover, the short period was not simply caused by cyc overexpression. This is because elav-cyc flies (uas-cyc rather than uas-cyc-vp16 in combination with the same elav-gal4 driver) have a wild-type–like period (Figure 2C, bottom). We thus attribute the period-shortening effect to increased transcriptional activity from the CLK–CYC-VP16 heterodimer within circadian cells.

Consistent with this interpretation is the period of tim-cyc-vp16 in combination with the classic per^s allele; these flies have ∼17-h periods, 2 h shorter than the canonical per^s 19–20 h phenotype (Figure 2B, bottom). The additive nature of tim-cyc-vp16 and per^s suggests that they shorten period in independent ways, the former by increasing transcription of CLK–CYC direct target genes and the latter by causing more rapid PER turnover [50].

To further study the period-shortening effect of tim-cyc-vp16, we characterized the molecular clock of these flies. To this end, we added a tim-luc or a per-luc reporter gene to the tim-cyc-vp16 strain (generating tim-luc-cyc-vp16 flies or per-luc-cyc-vp16 flies).

The expression of luciferase is robustly rhythmic in tim-luc-cyc-vp16 flies and isolated wings. The patterns are similar to those of wild-type tim-luc flies, but luciferase levels were about 2–3 times higher (Figure 3A for isolated wings and Figure S2A for intact flies). This is a comparable activity difference to what was observed above between CLK–CYC and CLK–CYC-VP16 in S2 cells (Figure 1B and 1C). Robust cycling and an even greater activity difference are observed with the per-luc reporter gene (Figure 3B).

Normalization to the first peak of the oscillations in Figure 3B and 3A allowed a useful comparison between the controls and the per-luc-cyc-vp16 and the tim-luc-cyc-vp16 profiles (Figure 3C and Figure S2B). The normalized pairs are very similar, but the CYC-VP16 curves are phase-advanced as they decrease more rapidly and then increase more rapidly during the next cycle (Figure 3C and Figure S2B). The peaks remain coincident, almost certainly reflecting entrainment to the superimposed 24-h LD cycle. Careful observation of the tim-luc reporter in constant darkness (DD) conditions reveals shorter circadian period in CYC-VP16 flies, in parallel with the behavior (Figure S2B). The damping oscillations of the wing transcriptional reporters in DD (always true in our hands) precluded a precise period determination.

To compare these reporter effects with those on bona-fide circadian mRNAs, microarray assays were performed on tim-cyc-vp16 head RNA from Zeitgeber time 15 (ZT15) and ZT3 (the timepoints when the CLK target genes have the peak and trough mRNA amounts in wild-type flies) and compared to the same timepoints from wild-type flies (Figure 4A and 4B). CLK–CYC direct target gene (tim, per, vrille (vri), and par domaine protein 1 (pdp1)) mRNA peak levels increase 2–3-fold, and an increase is also observed in trough levels (Figure 4A). The increase in trough levels suggests that they normally result from residual CLK–CYC activity that resists repression and/or that there is a minority of CLK–CYC-expressing cells that lack a robust circadian repression system. The microarray results are qualitatively similar although quantitatively less striking than the reporter gene assays shown above (Figure 3A and 3B). This may reflect the longer half-lives of the CLK–CYC direct target gene mRNAs relative to the luciferase reporter mRNAs or another level of post-transcriptional regulation. It is also possible that the reporter genes have a larger transcriptional response to CYC-VP16 than the CLK–CYC direct target genes.

In contrast to these direct target genes, maximal values for cycling mRNAs that peak at the opposite time of day are not increased in tim-cyc-vp16 flies (Figure 4B). Trough levels are decreased, however, suggesting that this might reflect an increase in the level of a transcriptional repressor protein, itself the product of a CLK–CYC direct target gene (e.g., VRI [29,30]).

We also tested whether the increase in CLK-mediated transcription was predominantly due to impaired per repression. To this end, we measured the effect of the CYC-VP16 protein in a per null mutant (per¹) background [63]. The tim and vri mRNA levels are increased in per¹ flies, comparable to the increase in the S2 cell (also without PER) experiments (Figure 1B). This indicates that transcription is increased independent of any more subtle effects on per repression.

Although we attribute the shorter period of the cyc-vp16 flies to a direct enhancement of transcription, it is still possible that the VP16 activation domain has a subtle effect on some other aspect of repression, which then only indirectly enhances transcription. Therefore we decided to assay the periods of transgenic flies carrying increasing numbers of copies of the Clk genomic region. Introduction of additional copies of the Clk transgene shortens circadian period and increases CLK–CYC-mediated transcription similar to the effects of the cyc-vp16 transgene (Figure 4A and 4B).

Homozygous ClkAR flies have significantly diminished levels of functional CLK and very low amplitude transcriptional oscillations of core clock genes [64]. As a consequence, these mutant flies do not have circadian activity patterns in DD or even in standard LD conditions. In addition they do not show the typical burst of activity at the beginning of the light cycle present in wild-type flies (lights-on startle response). Because CYC-VP16 increases CLK-driven transcription, we tested it for rescue of circadian activity in the ClkAR mutant background. Although introduction of CYC-VP16 into the ClkAR background failed to rescue circadian locomotor activity rhythms in DD conditions, most of the abnormal features of LD behavior conditions were restored: this included the presence of behavioral cycles (higher diurnal than night activity) as well as the lights-on startle response (Figure 6).

The effect of CYC-VP16 on the transcriptional profiles of the reporters and CLK–CYC direct target mRNAs suggested that the period-shortening effect might be simply due to a CYC-VP16-mediated change in the timing or level of per transcription. To test this possibility, we assayed the period of tim-cyc-vp16 flies in the context of uas-per (i.e., a period gene that can be driven constitutively by GAL4 but not by CLK–CYC or by CLK–CYC-VP16). Importantly, Sehgal and co-workers [58] have shown previously that uas-per can rescue the arrhythmic per¹ genotype (per¹; elav-gal4; uas-per), and we verified this finding (Figure 7A). Importantly, the elav-gal4 driver in combination with uas-cyc-vp16 (and a wild-type per gene) also manifests the ∼2-h period shortening as shown above (Figure 2C). However, these two transgenes in combination with the uas-per and per¹ only shorten circadian period by 20 min (Figure 7A–7C, and Figure S3B). This indicates that an increase in the levels and/or timing of per transcription is a major contributor to CLK–CYC-VP16 period shortening. We also note the broad distribution of individual fly periods from genotypes containing the uas-per; per¹ combination compared to the much tighter distribution in genotypes containing a proper per promoter (Figure 7C and Figure S3C); this is an additional indication that per transcription contributes to period determination (see Discussion section).

This role of per transcription is consistent with previous reports showing a relationship between per gene dose and behavioral period: more per genes cause shorter periods [65–67]. To determine if other ways of increasing per transcription also give rise to period shortening, we compared behavioral period between genotypes with one or two doses of uas-per (Figure 7A and Figure S3C). Rather than shortening period, however, the extra copy of uas-per slightly lengthens it. This is consistent with previous reports showing that overexpression of a uas-per transgene does not shorten period [44,58,68]. Taken together with other data shown above, we conclude that the short period of tim-cyc-vp16 requires not just increased levels of per mRNA but proper timing of the per transcriptional increase.

To address the role of transcription in core circadian timekeeping in the Drosophila system, we have analyzed the effects of a cyc-vp16 fusion gene in S2 cells as well as in flies. VP16 is a potent and well-studied transcriptional activator, which imparts to the CLK–CYC-VP16 heterodimer enhanced activity relative to that of the normal CLK–CYC complex. This increased activity is manifested with reporter genes, and transgenic flies also have increased levels of CLK–CYC direct target gene mRNAs, including those from per and tim. Importantly, the cyc-vp16-expressing flies have a short period, implicating circadian transcription in period determination. As this short period and proper period control more generally require a per promoter, we suggest that CLK–CYC-VP16 drives increased per transcription, which leads to more rapid accumulation of PER and a consequent advanced phase of per repression. This is also consistent with reporter gene profiles in cyc-vp16-expressing flies. The results indicate that circadian transcription contributes to core period determination in Drosophila.

This conclusion fits with several other pieces of data from the Drosophila system. First, recent studies have identified the transcriptional repressor–encoding gene clockwork orange (cwo) as a clock gene [62,69,70]. The protein product synergizes with PER and aids the repression of CLK–CYC direct target genes. Importantly, mutations in cwo or changes in cwo expression cause substantial period changes. Second, an increase in per gene dose leads to flies with short periods. There is a decrease of approximately 0.5 h for each additional gene copy up to about four copies, which have a ∼22-h period (e.g., [65]). Third, a hemizygous deletion that includes clock lengthens circadian period by about 0.5 h [17]. Although this deletion removes more DNA than just clk (including the adjacent clock gene pdp1), our results indicate that additional copies of the clk locus indeed shorten the circadian period of otherwise wild-type flies (Figure 5). All of these observations are qualitatively similar to the increase in transcription and period shortening caused by expression of cyc-vp16 in flies.

Because of the molecular analyses (Figure 3 and Figure S2), we suspect that it is the timing of per transcription rather than a simple increase in per mRNA levels that causes the period shortening by expression of cyc-vp16. As the reporter genes contain proper per and tim promoters, their profiles indicate that per and tim transcription decreases more steeply and then increases more steeply in the cyc-vp16 flies (Figure 3C and Figure S2B). The steeper decrease presumably reflects a faster accumulation of active PER repressor, and the steeper increase reflects the enhanced potency of CLK–CYC-VP16. In addition, we note that an increase in per dose with a uas-per transgene slightly increases rather than decreases period (Figure 7A and Figure S3C) [44,58]. This genetic requirement for the per promoter also emphasizes the contribution of proper transcriptional regulation to period determination.

The increased transcriptional potency of CLK–CYC-VP16 is unlikely to be a consequence of impaired PER-mediated repression, due in turn to some structurally anomalous feature of the artificial fusion protein. This is because the stronger activation of CLK direct targets by the CLK–CYC-VP16 dimer is apparent even in the absence of PER (Figures 1C and 4C). Shorter periods due to more potent transcription is also the conclusion of Figure 5, which shows that increasing clk gene dose (an independent and “more natural” way to increase CLK-mediated transcription) leads to molecular and behavioral changes that resemble those observed in tim-cyc-vp16 flies. Finally, cyc-vp16 expression rescues several aspects of the ClkAR phenotype (Figure 6). This suggests that these features are due to low direct target mRNA levels, which are increased by the more potent CLK–CYC-VP16 complex. The failure to rescue the behavioral arrhythmicity of homozygous ClkAR flies may reflect a requirement for minimal CLK levels, which would not be expected to increase by the addition of CYC-VP16.

The robust behavioral and molecular rhythms of cyc-vp16 flies (Figure 2 and 3) more generally indicate that CLK–CYC-VP16 circadian function, including the mechanism(s) that temporally activate or repress transcription of this hyperactive complex, must be similar to those that regulate the activity of the wild-type CLK–CYC complex. This is also because the increased transcription as well as RNA levels in tim-cyc-vp16 flies suggests that most CLK–CYC direct target gene transcription is carried out by CYC-VP16 rather than endogenous CYC. Because the VP16 activation domain almost certainly functions differently from the CLK polyglutamine region, this indicates that the recruitment of specific activator and/or repressor proteins is unlikely to play a prominent, mechanistic role in the circadian regulation of transcription. A more likely mechanism involves the cyclical inhibition of CLK–CYC DNA binding. Importantly, this notion is consistent with recent chromatin immunoprecipitation results from the mammalian as well as the fly system [54,71]. Nonetheless, we suggest that per transcription as well as DNA binding of the CLK–CYC dimer to per E-boxes is the actual timekeeper of the circadian cycle during the mid-late day, when they are both increasing. This predicts that the additional activation power of VP16 indirectly shortens the DNA binding time of the CLK–CYC-VP16 dimer by accelerating the rate of PER accumulation and function. This hypothesis also fits well with the behavioral and molecular defects observed in cwo mutant flies [62,69,70].

The emphasis on the per promoter is seemingly contradicted by the rhythmicity of flies missing not only this promoter but also the tim promoter [58]. In our hands as well, per¹; elav-gal4; uas-per flies are largely rhythmic despite weak rhythms, and their average period is near-normal. However, the period distribution of individual flies is unusually broad (Figure 7C and Figure S3C), indicating a contribution of the per promoter to the proper control of period within individual flies—even without CYC-VP16. Moreover, luciferase recordings from these transgenic flies show poor or no transcriptional oscillations (unpublished data). These observations suggest that individual neurons from this per¹; elav-gal4; uas-per strain might be impaired even more than indicated by the behavioral rhythms of this strain (i.e., circadian brain circuitry might help to compensate for poor core circadian function within individual cells). This is analogous to the superior circadian performance of behavioral rhythmicity and the suprachiasmatic nucleus (SCN) from mutant mouse strains compared to that of individual tissue culture cells (mouse embryonic fibroblasts) derived from the same strains [72].

The role of circadian transcription described in this study complements the well-documented role of PER, TIM, and CLK post-translational regulation in period determination [34,35,43,44,48,49,54,73–75]. Given the parallel role of mammalian CLK and BMAL1 to CLK and CYC, it would be surprising were there not a similar contribution of circadian transcription to mammals. This suggests that there is a division of labor in animals between transcriptional and post-translational regulation of circadian timekeeping, which may even be temporally segregated. In contrast and as mentioned above, recent indications are that post-translational regulation is the pre-eminent mechanism in cyanobacteria. It is also the case that individual bacterial cells keep excellent circadian time, essentially indistinguishable from the culture [76]. This contrasts with individual eukaryotic cells, for example, separated SCN cells, which show substantially more variation in period than the intact SCN or organism [72,77]. All of these considerations suggest that the intracellular timekeeping mechanism of animals is different from that of cyanobacteria. We suggest that this important difference between systems reflects their separate origins, a view that is supported by the lack of sequence conservation between cyanobacterial and animal clock proteins.