Drosophila Free-Running Rhythms Require Intercellular Communication

To test whether the LN_vs can support free-running circadian locomotor activity rhythms independently of other functional clock cells, we restricted pacemaker activity to these few PDF-expressing cells. CYCLE (CYC) is a bHLH–PAS protein (Rutila et al. 1998) and forms a heterodimeric transcription factor with CLOCK (CLK), another bHLH–PAS protein (Allada et al. 1998). CYC is an essential component of the Drosophila circadian oscillator transcriptional feedback loop (Glossop et al. 1999). The cyc¹ nonsense mutation completely eliminates molecular oscillations, and the direct target genes period (per) and timeless (tim) mRNAs are essentially undetectable (Rutila et al. 1998). Behavioral rhythms are also absent in the cyc¹ homozygous mutant strain (Rutila et al. 1998). We rescued cyc¹ specifically in the LN_vs, by using a well-characterized pdf–GAL4 driver (Renn et al. 1999) in combination with a UAS–CYC transgene to express ectopically wild-type CYC. Since CYC is apparently not a rate-limiting component of active dCLK–CYC complexes (Bae et al. 2000) and does not undergo molecular oscillations itself (Rutila et al. 1998), we expected that CYC overexpression would not cause circadian oscillator dysfunction. Indeed, the presence of the two transgenes did not affect locomotor activity rhythms in a wild-type background (Figure 1C, right panel).

The rescued mutant flies (pdf–GAL4;UAS–CYC,cyc¹/cyc¹) were examined by two independent criteria. First, molecular oscillations were assayed by in situ hybridization with a tim probe (Figure 1A and 1B). tim RNA levels undergo robust cycling in wild-type flies, with a trough at ZT3 and a peak at ZT15 (Sehgal et al. 1994). This is also true within all individual clock neurons (Zhao et al. 2003). tim mRNA cycled in the LN_vs (Figure 1A and B), indicating successful rescue of the molecular oscillator within these cells. The fact that other clock neurons were still tim mRNA-negative (Figure 1A and B) suggests that CYC and the rest of the molecular machinery can function cell autonomously, at least in the LN_vs under these light–dark (LD) conditions. The observed oscillations are also not passively driven by light, since they persisted in DD, at least in the s-LN_vs (Figure S1). Second, locomotor activity rhythms were examined by standard behavioral criteria. The transgenic flies were completely arrhythmic in DD. They were also arrhythmic under LD conditions, as the flies failed to anticipate the discontinuous transitions from light to dark or from dark to light (see Figure 1C, left panel; Rutila et al. 1998). In summary, the behavioral phenotypes were indistinguishable from those of the parental cyc¹ mutant strain.

The insufficiency of LN_v molecular rhythmicity indicates that one or more additional groups of rhythmic clock neurons are required for behavioral rhythmicity. We considered that robust molecular cycling under extended constant darkness conditions might be a good criterion for identifying these cell groups, because prior biochemical studies showed that some head and brain locations undergo damping of molecular oscillations under free-running conditions (Hardin 1994; Stanewsky et al. 1997). This conclusion has been extended by more recent immunohistochemical observations (Yang and Sehgal 2001; Shafer et al. 2002). The criterion of maintaining persistent and robust molecular rhythms in DD therefore suggests that only a limited set of brain locations are likely to be free-running pacemaker candidates. In order to identify these neurons, we assayed fly brains by tim in situ hybridization after 8 days in DD. To our surprise, we found that all tim-expressing brain cell groups (including both large ventral lateral neurons [l-LN_vs] and small ventral lateral neurons [s-LN_vs], doral lateral neurons [LN_ds], and all three groups of dorsal neurons [DNs]) still cycle robustly at this time (Figure 2). Previous studies have reported that the l-LN_vs fail to maintain oscillations at the beginning of DD (Yang and Sehgal 2001; Shafer et al. 2002). We have reproduced these observations, but noticed that the l-LN_vs “adapt” to constant conditions by becoming rhythmic once again after about 2 days in DD (data not shown). These results clearly distinguish the brain from the eyes and other peripheral tissues, which rapidly lose coherent molecular oscillations under free-running conditions (Hardin 1994; Plautz et al. 1997; Stanewsky et al. 1997; Giebultowicz et al. 2000). Although this approach failed to identify the additional neuronal groups necessary for behavioral rhythms, it suggests that many of these brain neuronal groups might act together in a network to support robust rhythms.

This association between robust molecular oscillations in all brain clock cells and behavioral rhythms in DD also made us consider the role of the neuropeptide PDF. The Pdf¹ mutant strain is unique among identified Drosophila circadian mutants, as it has little effect under LD conditions, but loses behavioral rhythmicity gradually and specifically in DD (Renn et al. 1999). This phenotype might reflect a disassociation between behavioral rhythmicity and the underlying molecular oscillations, as predicted from the role of PDF as a circadian output signal; it is proposed to connect the molecular oscillation in the LN_vs to locomotor activity (Renn et al. 1999).

We considered a completely different interpretation, namely, that PDF contributes to the functional integration of several brain clock neuronal groups, which is necessary to sustain molecular as well as behavioral rhythmicity under constant conditions. This fits well with previous studies of PDF in other organisms (Rao and Riehm 1993; Petri and Stengl 1997). In contrast to the canonical output model, this possibility suggests that the Pdf¹ mutant might manifest unusual molecular oscillations within clock neurons, especially under DD conditions. To address this issue experimentally, we examined Pdf¹ mutant flies by tim in situ hybridization.

In Pdf¹ flies, all clock neurons had robust tim RNA oscillations in LD, and the cycling phase and amplitude were comparable to those of wide-type flies (Figure 3A). The mutant flies were then released into DD and assayed at various times thereafter. In the first day of DD, cycling was similar to that observed in LD (Figure 3B). By the fourth day of DD, however, the cycling amplitude was much reduced in all clock neurons (Figure 3C and 3D). This was most evident from the unusually high signal in the CT2 sample; in wild-type flies, no tim signal was detected in any clock neuron at this timepoint (Figure 3C, left panels). There was also a reduced signal strength at the peak time, CT14 (Figure 3C, fourth panel from the left). The result parallels the damping of behavioral rhythms in the Pdf¹ mutant strain (Renn et al. 1999).

Despite the gradual fading of locomotor activity rhythms in DD, a significant fraction of Pdf¹ mutant flies is still weakly rhythmic after 4 d of DD (Renn et al. 1999). By tracking their locomotor activity phases, we observed that most of them had accumulated an approximately 4-hour phase advance relative to wild-type flies by the fourth day in DD. This is consistent with the measured ca. 23-hour periods of these weakly rhythmic flies (1-hour phase advanced per day for 4 days) as well as their advanced evening activity peak in LD (Renn et al. 1999). Quantitation of the tim in situ hybridization signal showed that there was a comparable one-point (4 h) advance in the peak of tim RNA and also confirmed the reduced cycling amplitude (Figure 3D). In order to eliminate the possibility that the observed damping is caused by the asynchrony of the Pdf¹ fly population, locomotor activities were tracked in real time. Individual flies were then removed from the monitors to assay tim RNA levels. Identical damped molecular oscillations were also observed in this case (data not shown). Taken together, the results indicate an excellent quantitative correspondence in phase and amplitude between the tim RNA rhythms and the behavioral rhythms in all clock neurons of the Pdf¹ strain.

To extend these observations, we also assayed cryptochrome (cry) mRNA oscillations by in situ hybridization. cry is expressed in a similar clock neuron pattern to tim, but it has a peak expression at ZT2 and a trough at ZT14 (Emery et al. 1998; Zhao et al. 2003). This phase is opposite to that of tim and other CLK–CYC direct target genes and reflects the fact that cry is only indirectly regulated by this heterodimeric transcription factor; CLK–CYC directly regulates the transcription factors PDP1 and VRILLE, which then regulate cry (Cyran et al. 2003; Glossop et al. 2003). Despite these differences between tim and cry, a similar result was obtained for cry in the Pdf¹ strain in the fourth day of DD (Figure 4), i.e., a reduced cycling amplitude compared to the fourth day of DD in a wild-type strain. This is suggested by the in situ pictures and is strongly indicated by the quantitation (Figure 4). The correspondence between the tim and cry mRNA patterns indicates that the entire circadian transcriptional program damps in the mutant strain in DD, which underlies the behavioral damping.

It is noteworthy that the mRNA oscillations damp uniformly in the Pdf¹ mutant strain, including the PDF-expressing LN_vs (see Figures 3 and 4). Since PDF is a neuropeptide (Rao and Riehm 1993), it is unlikely to exert a direct intracellular effect on the LN_v transcriptional machinery. A more conservative interpretation is that PDF maintains intercellular communication between individual LN_v neurons (Petri and Stengl 1997) and/or between the LN_vs and other cells; the communication is essential for self-sustained molecular rhythms within the LN_vs. Although this “feedback” could be quite indirect, the l-LN_vs project to the contralateral LN_vs through the posterior optic tract. Moreover, the s-LN_vs project dorsally to the superior protocerebrum, the location of the DNs. (Helfrich-Förster 1995). These anatomic features suggest that PDF might bind directly to clock neurons.

To test this hypothesis, in vitro biotinylated PDF peptide was incubated with fixed adult brains under near physiological conditions. The bound peptide was then detected in situ with a streptavidin-conjugated enzymatic amplification reaction. The vast majority of the signal localized with numerous cells at the periphery of medulla (Figure 5A). This is exactly where the l-LN_vs send large arborizations as their centrifugal projections (Helfrich-Förster 1995). Importantly, signal was also detected coincident with the LN_vs (Figure 5B) and likely DN3 clock neurons (Figure 5C) within the superior protocerebrum region, i.e., the bound peptide colocalized with GFP when the brains were from a strain with GFP-labeled clock neurons. Staining intensity was temporally constant; i.e., there was no systematic variation in signal intensity with circadian time. Although we obtained identical results with two differently biotinylated PDF peptides and there was no staining with two other biotinylated control peptides, we had difficulty to compete specifically the signal with nonbiotinylated PDF (see Materials and Methods). Moreover, PDF peptide staining of clock neurons was not reliably detected in every brain, in contrast to optic lobe staining. Nonetheless, we never detected peptide staining of other neurons in the vicinity of the LN_vs; i.e., signal in this region of the brain was always coincident with the GFP-labeled LN_vs. The peptide staining therefore suggests that PDF acts on the LN_vs in an autocrine or paracrine fashion as well as on other clock neurons, but the results do not exclude additional, more indirect modes of action.