Social/Olfactory Cues as Zeitgebers
In vertebrates, non-photic, non-thermal inputs are known to be able to synchronize circadian clocks. This includes social interactions and food availability. If food is restricted to a certain time of the day, mammals will change the phase of their circadian behavior to be able to feed, even in the presence of an LD cycle (Damiola et al., 2000). Flies do not appear to be able to do so, suggesting that food availability is not a strong Zeitgeber for flies, at least in the presence of an LD cycle (Oishi et al., 2004).
Flies are, however, sensitive to social cues (Levine et al., 2002b). If a small number of flies with a defined circadian phase are mixed with a larger population having a different phase, they can influence the phase of the larger population. Surprisingly, the contrary is not true, the large population does not seem to be able to phase-shift the smaller. This social phase-shifting is dependent on olfactory cues (Levine et al., 2002b), and motivated the search for pheromones that might be rhythmically produced. Pheromones are produced by oenocytes. These oenocytes have circadian clocks, and the production of at least one key enzyme for pheromone production (DESAT1) is rhythmic (Krupp et al., 2008). The amplitude is quite low (1.5- to 2-fold) at the mRNA and protein levels. However, this is nevertheless probably functionally meaningful, since mutants with DESAT1 levels reduced by ca. 50% have an altered hydrocarbon production profile. Moreover, the concentrations of at least four cuticular hydrocarbons involved in courtship are circadianly regulated. Whether these pheromones actually are responsible for the phase shift observed after social interaction is not known. Interestingly, social interactions have effects on clock gene expression in oenocytes, and on the pattern of expression of desatl and cuticular hydrocarbons (Krupp et al., 2008). There is therefore a complex reciprocal relationship between the oneocyte circadian clock, the production of cuticular hydrocarbons, and social interaction. What could be the purpose of these interactions?
Synchronization of male and female behavior to maximize the chances of finding a mating partner comes to mind. Courtship behavior is influenced by the circadian clock, and the interaction of males and females profoundly alters the pattern of circadian behavior (Sakai and Ishida, 2001a; Fujii et al., 2007). Oenocyte pacemakers have not yet been specifically disrupted, but the data presented above strongly indicate that they play an important role in social/olfactory synchronization of circadian behavior.
Neural Control of Drosophila Circadian Behavior
As discussed in the introduction, circadian clocks regulate various behaviors. The best studied at the neural level is Drosophila’s circadian locomotor rhythms. This section will mostly focus on the neural network that regulates this rhythmic behavior.
Anatomy of the Drosophila Circadian Neural Circuit
Immunohistochemical staining with antibodies directed against PER and other clock proteins, as well as circa-dian GAL4 drivers (per-GAL4, tim-GAL4) or P-element enhancer traps driving reporter genes (gfp, lacZ), were used to identify circadian neurons in Drosophila (see, for example, Helfrich-Forster, 1995; Kaneko et al., 1997; Kaneko and Hall, 2000; Shafer et al., 2006, reviewed in Helfrich-Forster, 2005; Nitabach and Taghert, 2008).
There are about 75—80 circadian neurons per hemisphere in the adult fly brain (all the neuron numbers below are given per brain hemisphere; see Figure 7). They are divided into several groups, named by the anatomical positions of their cell bodies in the brain; these groups therefore do not necessarily represent functional units.
There are three types of ventrolateral neurons (LNvs). The four large LNvs (l-LNvs) and four small LNvs (s-LNvs) are pigment dispersing factor (PDF) positive. As discussed in the next section, PDF is a neuropeptide that plays an important role in communication between circadian neurons. In addition, there is a PDF negative s-LNv usually referred to as the "fifth s-LN"^ There are six dorsolateral neurons (LNds), three lateral posterior neurons (LPNs), and three groups of dorsal neurons (17 DN1s and 2 DN2s, while the number of DN3s is believed to be between 30 and 40). Most of these groups can be further divided in subgroups, based on their size, neurotransmit-ter content, specific GAL4 driver expression or expression of other markers (for example the presence of absence of CRY). This extraordinary diversity suggests a complex division of function among circadian neurons. For a more detailed description of circadian neuron subtypes, see, for example, Kaneko and Hall, (2000), Shafer et al. (2006), Helfrich-Forster et al. (2007a, 2007b), Nitabach and Taghert (2008), Yoshii et al. (2008), Johard et al. (2009), and references therein.
The study of circadian neuron projections reveals that most circadian neurons send projections to the dorsal protocerebrum, which is believed to play an important role in locomotor activity control (Kaneko and Hall, 2000; Helfrich-Forster, 2005). One notable exception is the l-LNv group. In addition, projection patterns suggest that circadian neural groups communicate with each other. Contralateral projection of l-LNvs and LNds might help circadian neurons from both hemispheres to communicate with each other (Helfrich-Forster et al., 2007a). Furthermore, dorsal ipsilateral projections from s-LNvs and ventral ipsilateral projections from LNds and DNs probably mutually connect ventrally and dorsally located circadian neurons. An important center of communication appears to be the accessory medulla, where many circadian neuron projections converge, and where visual input might be received (Helfrich-Forster et al., 2007a).
Figure 7 The circadian neurons and their projections in the Drosophila brain. Different groups of circadian neurons are labeled with different colors. The three groups of dorsal neurons (DN1, DN2, and DN3) are all labeled in gray. Projections are of the same color as the cell bodies. For simplicity, only the projections from the neurons of the right hemisphere are shown. Note that the l-LNvs and LNds send contralateral projections.
Thus, there is a complex interconnection between different circadian neurons, which might contribute to the robustness of circadian behavior and to their ability to respond to environmental inputs (see sections 15.4.2 and 15.4.5).
Communication between Circadian Neurons
PIGMENT DISPERSING FACTOR (PDF) and its receptor To understand how circadian behavior is generated, it is necessary to determine how circadian neurons communicate with each other. An important circadian neurotransmitter — and the best characterized — is pigment dispersing factor (PDF). This neuropeptide has a very restricted expression pattern in Drosophila, being expressed only in the four s-LNvs and the four l-LNvs, in four to six abdominal ganglionic neurons (non-circadian), and in three neurons in the tritocerebellum that undergo apoptosis soon after eclosion (Renn et al., 1999). Wild type flies exhibit daily cycling of PDF immunoreactivity within the s-LNvs dorsal projections. These rhythms are self-sustained, and are affected by per mutations (Park et al., 2000). This rhythmic PDF immunoreactivity is interpreted as reflecting rhythmic PDF release.
The Pdf 01 mutation reveals the crucial role played by PDF in the control of circadian rhythms (Renn et al., 1999). Most Pdf01 mutants are arrhythmic in DD, and the rare flies that remain weakly rhythmic have a short-period behavior. In LD, the morning peak of activity is lost (see below) and the evening activity phase is advanced. The behavioral arrhythmicity in DD is correlated with desynchronization between circadian neurons, and a loss of amplitude of molecular rhythms (Peng et al., 2003; Lin et al., 2004). More recent studies demonstrate that PDF has different effects on different circadian neurons; for some of them PDF is necessary for rhythmicity, while for others it accelerates or slows down their pacemaker (Choi et al., 2009; Yoshii et al., 2009c).
The Pdf 01 phenotypes can be phenocopied by eliminating the PDF cells by expressing a pro-apoptotic gene (rpr or hid) with a Pdf-GAL4 driver (Renn et al., 1999). It is also replicated in mutants affecting a class II G protein-coupled receptor (Hyun et al., 2005; Lear et al., 2005a; Mertens et al., 2005), which has been demonstrated to respond specifically to PDF (Mertens et al, 2005). PDF receptor (PDFR) is expressed in many circadian neurons and possibly in non-circadian neurons as well, although direct reliable visualization of PDFR with specific antibodies has proved challenging (Hyun et al., 2005; Lear et al., 2005a, 2009; Mertens et al., 2005; Shafer et al., 2008). The most recent study relied on the use of a large MYC-tagged transgene encompassing the whole PDFR genomic region (Im and Taghert, 2010). It is thus likely to precisely reflect endogenous PDFR expression. PDFR is expressed in specific circadian neurons: the s-LNvs, and a subset of l-LNvs, LNds, DN1s, and DN3s. The presence of PDFR on s-LNvs suggests the existence of an autocrine feedback and communication between large and small LNvs. Interestingly, it appears that a broader range of circadian neurons rapidly responds to bath-applied PDF, based on a PDF-induced signal visualized in dissected brain with a FRET-based cAMP sensor (Shafer et al., 2008). PDFR-expressing neurons might thus communicate with PDFR-negative neurons to relay PDF signals. Recent studies demonstrate that PDFR expression in a subset of DN1s plays an important role for rhythmicity in DD, and for morning activity (Zhang et al., 2010a). In addition, PDF signaling collaborates with CRY to synchronize the so-called E oscillators that control the evening peak of activity (see section 15.4.4) with the LD cycle (Cusumano et al., 2009; Zhang et al., 2009).
Taken together, these results support the idea that rhythmic release of PDF synchronizes most Drosophila circadian neurons, and thus maintains circadian behavior in DD, and properly phases circadian behavior in LD. It should be noted that a recent study has challenged the notion that rhythmic PDF release is needed for circadian rhythms in DD. Indeed, in flies expressing a modified PDF in the LNvs, no rhythms in PDF signals were found in the dorsal termini, yet flies were normally rhythmic (Kula et al., 2006).
Interestingly, the expression of a PDF peptide that is tethered to plasma membranes in dorsally located cir-cadian neurons can improve rhythmicity in Pdf 01 flies, although the rhythms thus obtained are not normal (Choi et al., 2009). Most rescued flies exhibit a complex behavior, with simultaneous complex rhythms of 22-hour and 26-hour periods. Thus, hyperactivation of the PDFR with highly expressed membrane-tethered PDF results in internal de-synchronization, with some circadian neurons running fast and some slow, probably based on the way their pacemakers respond to PDF signaling (phase advances or delays). Interestingly, when the LNvs are hyperexcited with the expression of NachBac driven by Pdf-GAL4, a similar complex behavior arises (Sheeba et al., 2008a). If this genetic manipulation is done in Pdf 01 flies, rhythmicity is surprisingly rescued, although with a short period. This strongly suggests the existence of additional LNv neurotransmitters that can, under specific circumstances, compensate for the lack of PDF (see the next section).
Other neurotransmitters of circadian neurons It has recently been found that PDF-positive s-LNvs do indeed express an additional neuropeptide: short neuropeptide F (sNPF) (Johard et al, 2009). sNPF is also found in two LNds. In addition, three other neuropeptides are expressed in some circadian neurons: ion transport peptide (ITP) in the fifth s-LNv and in one LNd, neuropeptide F (NPF) in the ITP-positive LNd and two other LNds, and IPNamide in two DN1s (named DN1as for their anterior position) (Shafer et al., 2006; Johard et al., 2009). The function of these neuropeptides in the control of circadian behavior is not yet known. However, NPF might be responsible for the difference between male and female circadian behaviors. In LD, females are much more active during the middle of the day than males (Helfrich-Forster, 2000). Interestingly, NPF is only expressed in the LNds in males, and not in females (Lee et al., 2006).
Classical neurotransmitters have also been found in specific circadian neurons. A Cha-GAL4 driver shows that the fifth s-LNv and the two sNPF-positive LNds are cholinergic.
Two DN1as, some DN1ps, and several DN3s are found to express the vesicular glutamate transporter (DvGluT), which indicates that they are glutamatergic neurons (Hamasaka et al., 2007). Interestingly, the s-LNvs express the metabotropic glutamate receptor DmGluRA. Furthermore, in DmGluRA mutants or in flies with RNAi-medi-ated DmGluRA knockdown, morning activity — which is controlled by the s-LNvs (see section 15.4.4) — is increased. In addition, DmGluRA mutant larvae show increased photophobic behavior, which is positively controlled by the s-LNvs (Mazzoni et al., 2005; Hamasaka et al., 2007). These data indicate that the LNvs receive inhibitory gluta-mate signals from glutamatergic DNs through DmGluRA (Hamasaka et al., 2007). This is confirmed by the observation that glutamate application leads to decreased intra-cellular calcium in larval s-LNvs, which is associated with decreased neuronal activity.
Inputs from Non-Circadian Neurons
The s-LNvs are probably integrating multiple inputs from various neuronal cell types. Indeed, the larval s-LNvs have been found to respond to application of GABA and serotonin (5HT), with decreased intracellular calcium (Hamasaka et al., 2005; Hamasaka and Nassel, 2006). GABA and 5HT staining showed that some GABAergic and serotoninergic processes terminate near the dendrites of larval and adult LNvs. Interestingly, 5HT receptors are expressed in adult LNvs, and serotonin appears to modulate CRY-dependent circadian photoresponses (Yuan et al., 2005). In addition, recent studies show that GABA inhibition of LNvs induces sleep (Parisky et al., 2008). Indeed, the large LNvs promote arousal (Shang et al., 2008; Sheeba et al., 2008b). Histaminergic projections from the Hofbauer-Buchner eyelet are found near the LNvs, and thus presumably participate in light entrain-ment (Hamasaka and Nassel, 2006; Veleri et al., 2007).
Neural Control of Morning and Evening Activity
As already discussed, what is relevant for flies in the wild is to be able to cope with day/night cycles. Thus, they have to be able to integrate temperature and light inputs, and generate properly phased circadian rhythms. Under a standard LD cycle (12 hours of light and 12 hours of darkness at 25°C), flies show a bimodal activity — with a morning peak (M peak) before light on and an evening peak (E peak) before lights off (Figure 1). There are, in addition, non-circadian acute, transient responses to the light transitions (positive masking). Similar M and E peaks are observed under temperature cycles.
Two complementary approaches based on the GAL4/ UAS system (Brand et al., 1994) were used to identify the neural substrates generating the M and E peaks (Blanchar-don et al., 2001; Grima et al, 2004; Stoleru et al, 2004). The first was the ablation of specific circadian neural groups with targeted expression of a pro-apoptotic gene (hid). The second was to rescue PER expression in a restricted number of circadian neurons of per0 flies. To achieve a good level of cellular resolution, different GAL4 drivers were used. GAL80 repressor transgenes were also introduced as tools to restrict GAL4-mediated expression (Lee and Luo, 1999; Stoleru et al., 2004).
These studies concluded that the PDF-positive s-LNvs control the M peak and are thus also referred to as M oscillators, while the fifth s-LNv and a subset of LNds (with possibly a few DN1s) control the E peak and are thus called E oscillators. Indeed, ablation of PDF cells — as mentioned in section 15.4.2.1 — results in the loss of the M peak, while per0 rescue with Pdf-GAL4 restores specifically the M peak (Renn et al., 1999; Grima et al., 2004; Stoleru et al. , 2004). The M peak is not rescued with a driver targeting the l-LNvs, but can be rescued when the s-LNvs are targeted (Grima et al., 2004; Cusumano et al., 2009). Ablation of all cry13-GAL4-positive circadian neurons (LNvs, LNds, and a few DN1s) results in the loss of both M and E peaks, but if Pdf-GAL80 is added to protect the LNvs from HID expression, the morning peak is preserved (Stoleru et al., 2004). Finally, rescue in a subset of LNds and in the fifth s-LNv is sufficient to restore evening activity in per0 flies (Grima et al., 2004; Rieger et al., 2009). Surprisingly, an M peak is observed in flies that have no PER in both large and small PDF-positive LNvs, but PER in all other circadian neurons. This seems at odds with the fact that the PDF-positive LNvs are required for M activity. The proposed explanation is that circadian neurons driving the M and E peaks are coupled oscillators (Stoleru et al., 2004). This is supported by the observation that PDF-negative cells send projections toward the s-LNvs (see above). The PDF-negative circadian neurons would thus be able to activate PDF-positive neurons to generate the M peak even in the absence of a circadian clock in
PDF-positive circadian neurons. It should be noted that although PDF-positive s-LNvs are sufficient for morning activity, they might not be required. Indeed, it is possible to produce flies without functional s-LNvs by expressing a poly-Q containing fragment of the Huntingtin protein in PDF-positive neurons (Sheeba et al., 2010). This either kills the s-LNvs or completely blocks PDF expression, but leaves the l-LNvs intact. Morning anticipation is preserved in these flies, but flies are arrhythmic in DD, as expected. Taken together with the absence of morning anticipation in flies without PDF cells, this suggests that the l-LNvs can substitute for the s-LNvs. There might therefore be a certain degree of plasticity in the circadian system (see also below).
Are the same cells responsible for the morning and evening peaks under a TC cycle? To a large extend, yes (Busza et al., 2007). However, when the cells driving the M and E peaks in LD are ablated, a low-amplitude circadian evening peak can still be detected. Moreover, rescue in per0 flies restricted to these cells does not entirely restore normal behavior under a TC cycle. Hence, additional circa-dian neurons respond specifically to temperature cycles, and can influence circadian behavior in adult flies (Busza et al., 2007). The identity of these temperature-sensitive circadian neurons is not yet known, because of lack of proper circadian drivers. However, the DN2s and LPNs are likely candidates. Indeed, when flies are exposed simultaneously to an LD and a TC cycle with an abnormal phase relationship, most circadian neurons follow the LD cycles, but the DN2s and the LPNs choose the TC cycle (Miyasako et al., 2007). Moreover, the DN2s and LPNs express low CRY levels, or possibly none (Klarsfeld et al., 2004; Yoshii et al., 2008). This could explain their preference for TC cycles. Finally, in larvae, in which the circadian system is considerably simpler, strong evidence indicates that the DN2s (and/or maybe the fifth s-LNv) can synchronize the LNvs with TC cycles (Picot et al., 2009).
Interestingly, while adult flies can quickly synchronize their circadian behavior with LD cycles, they are slow to entrain to TC cycles (Busza et al., 2007; Currie et al., 2009). This slow pace is determined by the PDF-positive circadian neurons, since in their absence flies are very sensitive to temperature input. Also, weakening the circadian molecular pacemaker in PDF-positive LNvs accelerates the rate of TC entrainment (weaker oscillators are more sensitive to inputs; Pittendrigh et al., 1991). The conclusion is that the robust pacemaker in the s-LNvs slows down entrainment, while PDF-negative circadian neurons are very sensitive to temperature cycles. The balance between TC-resistant and -sensitive neurons could be crucial for flies to be responsive to temperature cycles, without excessively responding to a Zeitgeber that is not as reliable as the LD cycle (Busza et al., 2007).
Plasticity in the Circadian Neural Network
The different sensitivity of circadian neurons to specific Zeitgebers is a first neural mechanism that might help flies to adapt their behavior to their ever-changing environment. Another is the modulation of the output of specific circadian neurons by environmental input. A striking example is the photic modulation of the respective role of PDF-positive and PDF-negative circadian neurons in the control of circadian locomotor rhythms (reviewed in Dubruille and Emery, 2008; Nitabach and Taghert, 2008; Sheeba et al., 2008c). Under constant darkness, as described above, the PDF-positive s-LNvs drive cir-cadian locomotor behavior; they are necessary for these rhythms, and determine their pace (Renn et al., 1999; Stoleru et al., 2005). In constant light (LL), as we have seen, flies are usually arrhythmic, but if the CRY input pathway is compromised, flies are rhythmic (Emery et al., 2000a). Interestingly, these LL rhythms are not driven by the PDF-positive circadian neurons, but by the E oscillator (three to four LNds + fifth sLNv) or a subset of DN1s (Murad et al, 2007; Picot et al, 2007; Stoleru et al., 2007). Moreover, PDF is not necessary for LL rhythms (Murad et al., 2007; Picot et al., 2007). ‘The presence or absence of light thus modulates the role played by a subset of PDF-negative and PDF-positive circadian neurons. Evidence indicates that this modulation is dependent on visual photoreceptors, and regulates the output of the relevant circadian neurons (Picot et al., 2007). Indeed, cryb flies with the E oscillators being the only cells with a functional circadian clock are rhythmic in LL but not in DD, even though the circadian pacemaker of the E oscillator keeps running under DD conditions as well. The reverse is also true: cryb flies with a functional clock restricted to the PDF-positive circadian neurons (M oscillator) are rhythmic only in DD but not in LL, even though under the latter conditions the molecular pacemaker free-runs. This light-dependent contribution of circadian neurons is probably important for dealing with days of different photoperiod. Indeed, under long day conditions, the PDF-negative circadian neurons (the E oscillators and/ or the DN1s, presumably) play a more important role in determining the phase of circadian behavior, while under short photoperiod the PDF-positive LNvs (M oscillators) dominate (Stoleru et al., 2007). Fitting with this idea, the PDF-positive LNvs (the large ones, probably) appear to be the dominant contributors to the advance zone of the PRC (see section 15.3), while PDF-negative circadian neurons dominate in the delay zone (Stoleru et al., 2007; Shang et al., 2008; Tang et al., 2010) (for further discussion, see Stoleru et al., 2007; Dubruille and Emery, 2008).
This specific contribution of different circadian neurons to the PRC has an interesting implication. CRY photoresponses are not entirely cell-autonomous, but are dependent on circuit properties. Again, the idea is that specific circadian neurons promote either phase advances or phase delays (Shang et al., 2008; Tang et al., 2010). CRY being necessary for these responses, CRY presumably functions in these circadian neurons to reset their local clock through TIM degradation. However, with the PRC, it is the phase of the s-LNvs that ultimately has to be reset, because these cells are the ones driving behavior in DD — the condition in which phase is measured after the light pulse. So is CRY also degrading TIM in the s-LNvs to contribute to behavior resetting? Surprisingly, at least in the delay zone of the PRC, recent evidence indicates that TIM degradation might be neither necessary nor sufficient for a delay to occur (Tang et al., 2010). This suggests that the s-LNvs reset their clock only through network signaling. Interestingly, the amplitude of the phase delay can be correlated with the number of DN1s that show TIM degradation (for example, by using low light intensity pulses) (Tang et al., 2010). The number of l-LNvs that are present in genetically manipulated flies also determines the amplitude of phase advance (Shang et al., 2008). This indicates that the s-LNvs integrate neuronal signals coming from other circadian neurons. The origin of these signals would encode directionality of the phase shift, while the number of neurons signaling the shift would encode the magnitude of the phase changes. Interneuronal variations in CRY levels (Benito et al., 2008; Yoshii et al., 2008) could explain how light intensity determines phase amplitude: it would take greater light intensity to trigger TIM degradation and reset local clocks in circadian neurons with low CRY levels. Similar principles probably apply to temperature input as well. Indeed, as discussed above, the PDF-positive LNvs are poorly sensitive to temperature input, but PDF-negative circadian neurons are much more sensitive and can entrain them to a temperature cycle (Busza et al., 2007; Miyasako et al., 2007; Picot et al., 2009; Tang et al., 2010). We can thus hypothesize that the s-LNvs are poorly sensitive to inputs so that they can serve as a center of integration from input-sensing neurons that detect different information about the environment (temperature increases and decreases, time of sunset or sunlight, light intensity, etc.). In apparent contradiction to this hypothesis is the fact that CRY is expressed in the s-LNvs (Emery et al., 2000b; Klarsfeld et al., 2004; Benito et al., 2008; Yoshii et al., 2008), and is required in these cells for the phase-shifting effect of light pulses (Tang et al., 2010). Maybe CRY functions very differently in the s-LNvs — for example, by participating in signaling cascades that reset the circadian pacemaker in response to synaptic inputs from the l-LNvs or from the DN1s.
We started this section with an example of output modulation by light. Another example is the effects light and temperature have on output from a subset of DN1s targeted by a GAL4 driver that contains a fragment of the Clk promoter. Using this driver in various rescue experiments (per0, pdfrhan, na1; narrow abdomen (na) is an important channel in circadian neurons; Lear et al., 2005b), it was found that these cells play an important role in light-induced positive masking at light-on (startle response), for morning anticipation under standard LD conditions, and for DD rhythms as a target of PDF signaling (Zhang et al., 2010a). Interestingly, their contribution to LD behavior is affected by the light intensity, as well as by temperature (Zhang et al., 2010b). In per0 rescue experiments, they were shown to generate predominantly an M peak or an E peak, or to promote both peaks, depending on light and temperature conditions. Light intensity did not affect the amplitude or phase of molecular rhythms in the DN1s, which indicates that it is the DN1 output pathway that is under photic and thermal control. Under the same environmental conditions, output from the E oscillators was unaffected, which means that the integration of photic and thermal inputs is specific to DN1 output. However, under other specific light conditions (moonlight used during the night instead of complete darkness) the E oscillators can also generate morning activity, while under a standard LD cycle they do not (Rieger et al., 2009).
All these results thus indicate that specific circa-dian neurons can function as both M and E oscillators, depending on environmental conditions. It is not yet possible, however, to exclude that within the E oscillators and the clk-GAL4-positive DN1s there are different cell types, some specific to morning activity and some specific to evening activity. Indeed, these cell groups are heterogeneous (Johard et al., 2009; Zhang et al., 2010b). It seems more likely, though, that the role of these circadian neurons is plastic, and that this plasticity helps Drosophila to cope with daily changes in its environment.
