Wakefulness Is Promoted during Day Time by PDFR Signalling to Dopaminergic Neurons in Drosophila melanogaster

Visual Abstract


Introduction
Daily cycles in several environmental factors synchronize endogenous circadian clocks which drive rhythmic sleep/wake patterns in many organisms. Homeostatic mechanisms modulate the amount and depth of sleep, and also allow animals to recover from any sleep deprivation they may have incurred. Together, these processes control the timing and occurrence of sleep and wake states, thereby modulating sleep/wake cycles. Since the discovery that sleep behavior of Drosophila melanogaster is similar to mammalian sleep in several aspects (Hendricks et al., 2000;Shaw et al., 2000), many pathways and neuronal circuits involving sleep homeostat and circadian clocks have been uncovered. Genes such as minisleep (mns) and hyperkinetic (hk) encoding subunits of Shaker potassium channel function in the sleep homeostat (Cirelli et al., 2005;Bushey et al., 2007). More recently, central complex structures such as dorsal fan-shaped body (FB;Donlea et al., 2014) and the ellipsoid body (EB; Liu et al., 2016) have been shown to function as effector and modulator of the sleep homeostat, respectively. Meanwhile, mutations in core circadian clock genes such as Clock (clk) and Cycle (cyc) have been shown to cause impaired timing of sleep as they tend to become nocturnal (Kumar et al., 2012). The circadian neuropeptide pigment dispersing factor (PDF) and its receptor (PDFR) are involved in relaying wake-promoting signals from the circadian pacemaker ventral lateral neurons (LN v s;Parisky et al., 2008;Chung et al., 2009) in response to light input  as well as dopamine (Shang et al., 2011). While it has been suggested that the EB may be the downstream target of this wake-promoting PDF/ PDFR signaling, the evidence in favor of the same is limited (Parisky et al., 2008).
In the recent past, in the quest to uncover output pathways of the circadian clocks that help in timing of sleep/ wake cycles, a few dedicated circuits have been mapped. Most notably, timing of sleep onset at the beginning of night is a function of increased inhibition of wakepromoting large LN v (l-LN v ) by GABA (Liu et al., 2014). On the other hand, sleep is suppressed at the end of night by the action of PDF on PDFR ϩ dorsal neuron 1 (DN 1 ) group of the circadian network which in turn secrete the wakepromoting neuropeptide diuretic hormone 31 (DH31; Kunst et al., 2014). Furthermore, yet another group showed that DN 1 s through glutamate modulate day-time siesta and night-time sleep by inhibiting the morning (small LN v ; s-LN v ) and evening (dorsal lateral neurons; LN d s) activity controlling circadian neurons (Guo et al., 2016). Yet, none of the studies so far have shed light on how circadian neurons may induce wakefulness during the day.
Here, we addressed this question by screening for putative downstream targets of PDFR signaling by altering the levels of pdfr expression in several subsets of neurons -namely, circadian neurons that are known to express pdfr (Hyun et al., 2005;Lear et al., 2005;Mertens et al., 2005;Im and Taghert, 2010) subsets of mushroom body (MB) neurons that are sleep-or wake-promoting (Joiner et al., 2006;Pitman et al., 2006;Cavanaugh et al., 2016), wake-promoting pars intercerebralis (PI;Foltenyi et al., 2007), sleep homeostat EB (Liu et al., 2016), and sleep-promoting FB neurons (Donlea et al., 2011) as well as aminergic neuronal groups, most of which are reported to be wake-promoting (Kume et al., 2005;Crocker et al., 2010). Strikingly, we found that a subset of dopaminergic neurons responds to changes in pdfr expression by changing the levels of day-time sleep, increasing pdfr levels decreases day-time sleep and vice versa. Moreover, we find that PDF ϩ and dopaminergic neurons form synaptic contacts with one another, along with the possibility of the former inhibiting the latter. Thus, our results uncover a dedicated pathway involving signaling from the PDF ϩ neurons perhaps to the PPM3 dopaminergic neurons in the regulation of wakefulness during the day.

Fly strains
Fly strains used along with their source information is listed in Table 1. Briefly, flies were maintained on standard cornmeal medium under LD12:12 at 25°C. All flies used for sleep measurements have been back-crossed to the standard Iso31 or w 1118 (BDSC #5905) background for at least five generations. Pdfr 5304 , Pdfr 3369 , UAS Pdfr RNAi,  UAS dicer, UAS Pdfr, and TH GAL4 have been backcrossed to w 1118 for seven generations.

Sleep assays
Three-to 6-d-old mated females were individually housed in glass tubes (65 mm in length, 3 mm in diameter) with sucrose medium (5% sucrose and 2% agar) on one end and cotton plug on the other and activity was recorded in DAM2 monitors (Drosophila activity monitoring system, Trikinetics). The DAM system works on the principle that whenever a fly crosses the middle of the tube, it breaks an infra-red beam which gets detected by infrared sensors and recorded as activity counts. Flies were housed in light and temperature-controlled environments with 12 h of light (ϳ300 -500 lux) and 12 h of darkness (LD12:12) at 25°C in incubators (MIR-273, Sanyo; DR-36VLC8 Percival Scientific Inc.) for a period of 3 d. Activity was binned at 1 min and sleep parameters such as daytime and night-time sleep duration, bout length and number and activity per waking minute were estimated using PySolo (Gilestro and Cirelli, 2009), while sleep profiles and sleep latency were obtained from a custom-made Microsoft Excel spreadsheet template.

Statistical analysis
Change in day-time sleep is calculated as difference between day-time sleep of experimental flies and GAL4 or UAS control flies as a percentage of day-time sleep of GAL4 or UAS control flies. For comparison of sleep parameters, one-way ANOVA with genotype as fixed factor followed by post hoc Tukey's HSD test was conducted. For comparison of GFP fluorescence intensity, two-way ANOVA with genotype and time point as fixed factors followed by post hoc Tukey's HSD test was conducted. Significance level for all tests was set at p Ͻ 0.05.

PDFR signaling promotes wakefulness specifically during the day
To establish a phenotype on the basis of which, our screen to uncover downstream targets of PDFR could be designed, we examined two previously established loss-of-function mutants of the pdfr gene, pdfr 5304 and pdfr 3369 . A previous study had reported that both daytime and night-time sleep of these mutants is significantly higher than that of background control flies (Chung et al., 2009). However, we found that both mutants after backcrossing to the widely used Iso31 (w 1118 ) background for seven to eight generations exhibited significantly higher sleep only during the day-time under a standard LD12:12 cycle at 25°C (Fig. 1A,C; w 1118 vs pdfr 5304 , Student's two-tailed t test, t (0.05,2,46) ϭ -2.93, p Ͻ 0.05; w 1118 vs pdfr 3369 , Student's two-tailed t test, t (0.05,2,38) ϭ -6.33, p Ͻ 0.00001). Moreover, day-time sleep of pdfr 5304 and pdfr 3369 mutants was also different from one another (one-way ANOVA, F (2,59) ϭ 21.52, p Ͻ 0.00001), which was a peculiar observation seen in independent experiments. Functional analysis of the different domains that are deleted in these mutants, the former carries a deletion which excludes all the transmembrane domains and C terminus while the latter has a smaller deletion with loss of only one transmembrane domain and C terminus (Hyun et al., 2005) may explain why day-time sleep levels are different in the two mutants. Nonetheless, both mutants sleep much higher as compared to their background controls. Total sleep is significantly higher than the controls only in one of the mutants ( Fig. 1B; w 1118 vs pdfr 5304 , t (0.05,2,46) ϭ -1.92, p ϭ 0.06; w 1118 vs pdfr 3369 , t (0.05,2,38) ϭ -3.68, p Ͻ 0.005). However, night-time sleep of both the pdfr mutants was not different from that of the controls ( Fig. 1D; w 1118 vs pdfr 5304 , t (0.05,2,46) ϭ -0.09, p ϭ 0.93; w 1118 vs pdfr 3369 , t (0.05,2,38) ϭ 0.13, p ϭ 0.9). These differences in sleep were not due to differences in activity levels (Extended Data Fig. 1-1A,B). Although the activity per waking minute is significantly lower for one of the mutants (Extended Data Fig. 1-1B; w 1118 vs pdfr 5304 , t (0.05,2,46) ϭ 2.58, p Ͻ 0.05; w 1118 vs pdfr 3369 , t (0.05,2,38) ϭ 1.91, p ϭ 0.06) as compared to the control, this result was not seen in replicate experiments using the same genotypes (data not shown). Furthermore, the increase in daytime sleep seen in the pdfr mutants is also seen during subjective day time, when these flies are transferred to constant darkness (DD) at 25°C (Extended Data Fig.  1-1C,D; w 1118 vs pdfr 5304 , t (0.05,2,41) ϭ -5.51, p Ͻ 0.00001; w 1118 vs pdfr 3369 , t (0.05,2,35) ϭ -5.24, p Ͻ 0.00001). Moreover, we confirmed that back-crossing has not resulted in loss of the pdfr mutation by the observation that the behavioural phenotype of advanced evening activity peak (Hyun et al., 2005;Lear et al., 2005;Mertens et al., 2005) is reproduced under LD12:12 (Extended Data Fig. 1-1A). Taken together, these data suggest that absence of func- tional PDFR results in increased sleep duration specifically during the day.

Screen for downstream targets of PDFR signaling
Previous studies that have characterized the expression pattern of pdfr using different antibodies against PDFR and/or promoter-mediated expression of cellular tags such as myc have revealed pdfr expression in a subset of circadian clock neurons, PI, EB, and ϳ50 as yet non-characterized non-clock cells (Hyun et al., 2005;Lear et al., 2005;Mertens et al., 2005;Parisky et al., 2008;Im and Taghert, 2010). Therefore, on the basis of the predicted expression pattern of pdfr and potential sites in the vicinity of PDF projections, as well as those that function in sleep/wake regulation, we altered expression of pdfr in a total of 26 GAL4 lines including distinct subsets of circadian clock neurons, MB, PI, central complex, and some neurotransmitter/peptide systems. Our interest was to identify driver lines whose targets responded with both an increase in day-time sleep on downregulation and a decrease in day-time sleep on overexpression (OEX) of pdfr. Moreover, to rule out non-specific effects on daytime sleep of either the GAL4 or UAS parental line, we required the experimental flies to be significantly different as compared to both parental controls to be considered as a hit.
Quite surprisingly, downregulation and/or OEX in subsets of circadian clock neurons, which had previously been reported to modulate activity/rest rhythms in LD as well as in DD (Im and Taghert, 2010) did not show an effect on day-time sleep (Fig. 2). While downregulation of pdfr in ϳ12-14 DN1 p s using Clk 4.1M GAL4 (Zhang et al., 2010) resulted in a significant increase in day-time sleep as compared to both parental controls ( Fig. 2A; Extended Data Fig. 2-1), OEX of pdfr in the same subset of neurons did not result in a corresponding decrease in day-time sleep ( Fig. 2B; Extended Data Fig. 2-2). Moreover, downregulation of pdfr in almost all PDFR ϩ clock neurons using Pdfr (B) GAL4 (Im and Taghert, 2010) resulted in an increase in day-time sleep but this was significantly different only from the UAS parental control ( Fig. 2A; Extended Data Fig. 2-1). OEX of pdfr using the same driver however resulted in significant decrease in day-time sleep only as compared to the GAL4 control ( Fig. 2B; Extended Data Fig. 2-2). Moreover, when pdfr was downregulated and/or overexpressed in a different combination of essentially the same cluster of circadian clock neurons (Cry GAL4-39; Klarsfeld et al., 2004), consistent effects on day-time sleep were not observed (Fig. 2). These results together lead to the interpretation that circadian clock neurons may not be major downstream targets of PDFR signaling that regulates day-time wakefulness.
In a recent study, it was found that PDF ϩ neurons communicate with DN1s, which then communicate with DH44 ϩ PI neurons that brings about rhythmic locomotor activity under DD conditions (Cavanaugh et al., 2014). Given our finding that DN1s are most likely not the downstream targets of PDF ϩ neurons for sleep regulation, we next asked whether the PI neurons were direct recipients of PDF signals for modulation of day-time sleep. Barring a few non-specific parental effects on sleep, none of the 5 PI-specific GAL4 drivers we screened showed any significant effects on day-time sleep when pdfr was downregulated and/or overexpressed (Fig. 2). Thus, although PI neurons appear to be well-placed to receive PDF signals, our finding suggests that they are not required for sleep regulation by the PDF ϩ neurons. In the light of our results and previous findings that PI neurons modulate sleep and wake levels (Foltenyi et al., 2007;Crocker et al., 2010), it appears that the PDF signaling and PI neurons are in different pathways of sleep and wake regulation.
Given that pdfr is expressed in the EB, and the suggestion that they could be the output neurons of PDF effects on sleep and wake levels (Parisky et al., 2008), we downregulated and overexpressed pdfr using GAL4 drivers that distinctly label the EB. We found that downregulation of pdfr using c119 GAL4 led to an increase in day-time sleep Another GAL4 driver targeting the EB did not show these effects on day-time sleep on downregulation and OEX of pdfr (c232 GAL4; Fig. 2). Thus, these results suggest that EB may not be downstream of PDFR signaling in sleep and wake modulation. Im and Taghert (2010) reported that in addition to circadian clock neurons, PI, and EB, there are ϳ50 cells in the brain that are PDFR ϩ . We hypothesized that these 50 cells could potentially be any one of the MB and/or FB cells, neurons both of which are implicated in sleep regulation (Joiner et al., 2006;Pitman et al., 2006;Donlea et al., 2011;Donlea et al., 2014) and which may lie in the vicinity of projections of the PDF ϩ s-LN v neurons. Not so surprisingly, none of the GAL4 lines labeling either MB or FB showed significant and opposite effects on day-time sleep on downregulation and OEX of pdfr (Fig. 2). Inter-estingly, however, 4 GAL4 drivers showed strong significant effects on day-time sleep on either downregulation only or OEX only of pdfr. Out of these, when pdfr was downregulated using the 30y GAL4 which labels the ␣/␤ lobes strongly and the rest of the MB weakly (Aso et al., 2009), day-time sleep was significantly higher as compared to both GAL4 and UAS controls ( Fig  OEX of pdfr using UAS pdfr crossed with GAL4 lines expressed in subsets of circadian clock neurons, PI, MB, central complex, and major neurotransmitter systems. Bars represent percentage increase (positive values) or decrease (negative values) in day-time sleep of experimental flies with respect to that of either GAL4 (gray) and UAS (black) parental controls. Asterisks above the bars indicate level of significance when a one-way ANOVA with genotype as factor followed by post hoc Tukey's test was done on raw day-time sleep levels. Several lines when used to downregulate pdfr show a significant increase in day-time sleep, but do not show a corresponding decrease in day-time sleep when pdfr is overexpressed (Clk 4.1M GAL4, 30y GAL4, c5 GAL4, c119 GAL4), whereas a few lines show a significant decrease in day-time sleep when pdfr is overexpressed, but no corresponding increase in day-time sleep is seen when pdfr is downregulated (121y GAL4, 104y GAL4). However, in two lines (Ddc GAL4, TH GAL4), when pdfr is downregulated, there occurs a significant increase in day-time sleep, and when pdfr is overexpressed, there occurs a significant decrease in day-time sleep. All other details are as in Figure 1. For all genotypes, n Ͼ 24 flies. Results representative from at least two independent experiments. See also Extended Data Figures 2-1, 2-2. more restricted driver (c5 GAL4) for FB thereby revealing non-specific effects of the OEX using the broad driver.
We next focused on a few neurotransmitter/peptide clusters that have previously been known to regulate sleep and wake such as dopamine, serotonin, octopamine and neuropeoptide F (NPF; Kume et al., 2005;Yuan et al., 2006;Crocker and Sehgal, 2008;He et al., 2013). To our surprise, when we downregulated pdfr in serotonergic and dopaminergic neurons using Ddc GAL4, as well as dopaminergic neurons alone using TH GAL4, we found that day-time sleep was significantly higher than the parental controls ( Fig. 2A; Extended Data Fig. 2-1). Moreover, when we overexpressed pdfr using the same GAL4 drivers, we found that day-time sleep was significantly lesser than the parental controls ( Fig. 2B; Extended Data Fig. 2-2). However, there was no significant effect of either downregulating or overexpressing pdfr in either NPF ϩ or octopaminergic neurons on day-time sleep. Taken together, these results suggest that dopaminergic neurons are the most likely candidate for being the downstream targets of PDFR signaling to modulate day-time sleep and wake levels.

PDFR signaling to dopaminergic neurons promotes day-time wakefulness
We Night sleep latency: DR, F (2,89) ϭ 1.1, p ϭ 0.34; OEX, F (2,88) ϭ 2.81, p ϭ 0.07). Thus, these results lead us to hypothesize that decreasing PDFR signaling to dopaminergic neurons increases day-time sleep, while increasing PDFR signaling to dopaminergic neurons suppresses day-time sleep and makes it fragmented, in addition to delaying sleep onset, suggesting that PDFR signaling to dopaminergic neurons is necessary for initiating and maintaining day-time wakefulness.

PDFR ؉ PPM3 neurons modulate day-time wakefulness
Dopaminergic neurons labeled on the basis of reactivity to antibody against Tyrosine Hydroxylase (anti-TH), which is the rate-limiting enzyme for dopamine synthesis, are divided into several subsets based on their anatomic location (Mao and Davis, 2009). There are two subsets present in the anterior brain (protocerebral anterior medial and lateral; PAM and PAL, respectively) and five subsets in the posterior brain (protocerebral posterior medial and lateral; PPM and PPL, respectively; PPM1-3 and PPL1-2). Of these, two previous studies have implicated a pair of bilaterally located PPL1 neurons (Liu et al., 2012) and a unilateral PPM3 neuron (Ueno et al., 2012) in promoting wakefulness through the inhibition of sleep-promoting FB. We asked whether the PDFR signaling is acting on either or both of these subsets to promote wakefulness specifically during the day. We used the previously created and characterized GAL4 drivers (TH-A   To identify the dopaminergic neurons that receive signals from PDF, we used the previously described Pdfr myc line (Im and Taghert, 2010) where myc is fused to the Pdfr gene, such that labeling MYC labels most PDFR ϩ neurons including clock neurons and ϳ50 as yet uncharacterized non-clock neurons. We co-stained adult brains of pdfr myc flies with antibodies against TH and MYC and examined any overlap that may exist between TH ϩ and PDFR ϩ neurons. We found that one to two PDFR ϩ neurons always lie in the vicinity of PPL1 and PPM3 subset of dopaminergic neurons (Extended Data Fig. 4-5A; n ϭ 22 hemispheres). On careful examination, we found that in three out of 11 brains visualized, there was one PPM3 neuron in each hemisphere that was both TH ϩ and PDFR ϩ (Extended Data Fig. 4-5B, left). The low number of brains showing TH ϩ and PDFR ϩ PPM3 neurons could be because of high background and low affinity of the anti-MYC antibody. However, in none of the brains could we detect any overlap between PPL1 TH ϩ and PDFR ϩ neurons, although they were quite close to each other (Extended Data Fig. 4-5B, right). Thus, we can only conclude that perhaps one PPM3 neuron per hemisphere may express the PDFR.

PDF ؉ and TH ؉ neurons form synaptic contacts in sLN v axons
Based on our results with altering pdfr levels that show that dopaminergic neurons are downstream targets of PDFR signaling, we next examined the nature of communication between PDF ϩ and dopaminergic neurons. Because PDF is a neuropeptide, it can have long-range non-synaptic effects on downstream neurons expressing the PDFR (Nässel and Winther, 2010). We conducted a GFP reconstitution across synaptic partners (GRASP) experiment which relies on two independent binary systems allowing the expression of two membrane-bound GFP fragments in different sets of neurons, such that GFP is reconstituted and fluoresces only when the fragments are present at synaptic distances (Gordon and Scott, 2009). A similar experiment done previously had shown the presence of synapses between PDF ϩ and dopaminergic neurons in the LN v dendrites (Shang et al., 2011). However, here, we asked whether synapses between PDF ϩ and dopaminergic neurons occur specifically in the region of LN v axons since we wished to examine postsynaptic targets of PDF. We costained adult brains of flies in which dopaminergic neurons expressed GFP1-10 fragment and PDF ϩ neurons expressed GFP11 fragment with anti-GFP antibody that specifically labels reconstituted GFP and anti-PDF to visualize the LN v projections. We found that reconstituted GFP signal was specifically detected in the ascending part of the dorsal projection of sLN v (Fig. 5A) which is an axonal process. However, when either fragment was individually driven in the PDF ϩ neurons or dopaminer-gic neurons alone, no GFP signal was detected (Fig.  5B,C) showing that the antibody does not recognize individual fragments of GFP. This shows that PDF ϩ and dopaminergic neurons form synaptic connections especially in the axonal projections of s-LN v , thus bolstering our finding that dopaminergic neurons are downstream of PDFR signaling. Furthermore, in brains with dopaminergic neurons labeled with promoter driven GFP (TH GAL4 Ͼ UAS GFP), as well as with anti-TH, we find a dopaminergic projection intersecting the ascending dorsal projection of s-LN v (Fig.  5D, asterisk).

Auxiliary role of sLN v in mediating wake-promoting effects of l-LN v
While previous studies suggest a negligible role for the s-LN v in the sleep/wake circuit Chung et al., 2009), s-LN v have been proposed to have a secondary role in promoting wake-mediating effects of l-LN v (Parisky et al., 2008;Potdar and Sheeba, 2012). To explore their role further, we made use of previously reported toxic version of Huntingtin protein expression (Htt Q128, referred to as Q128, non-toxic form referred to as Q0) to selectively render s-LN v dysfunctional (Sheeba et al., 2010) while simultaneously changing the electrical properties of the remaining l-LN v by expressing the bacterial sodium channel NachBac (NB; Nitabach et al., 2006). We compared sleep patterns of the following genotypes of flies: those in which s-LN v were dysfunctional but l-LN v were normally firing (s -L ϩ Q128 and NCQ128), some neurons from both LN v subsets were ablated (s Ϯ L Ϯ rpr, apoptotic gene reaper; Potdar and Sheeba, 2012), both LN v subsets hyperexcited (s H L H NBQ0) and s-LN v were dysfunctional but l-LN v were hyperexcited (s -L H NBQ128) with their respective controls in which both LN v were normally firing and functional (s ϩ L ϩ Q0 for s -L ϩ Q128, s ϩ L ϩ GAL4 for s Ϯ L Ϯ rpr, s ϩ L ϩ NCQ0 for s H L H NBQ0 ands -L ϩ NCQ128, NC refers to dORKNC1, which is a non-conducting potassium channel; Nitabach et al., 2002). Additionally, s -L ϩ NCQ128 served as control for s -L H NBQ128. Because l-LN v wake-promoting effects are primarily mediated by light , we examined sleep levels of these flies in LD12:12 cycles with different day-time light intensities. We observe altered levels of day-time sleep only when the LN v are ablated, but never when they are hyperexcited either completely or partially (Extended Data Fig. 5-1A,B; two-way ANOVA, 10 lux: F (7,478) ϭ 4.5, p Ͻ 0.0005; 300 lux: F (7,468) ϭ 8.27, p Ͻ 0.00001; 2000 lux: F (7,476) ϭ 13.91, p Ͻ 0.00001). Moreover, the increased levels of day-time sleep in s Ϯ L Ϯ flies is seen only when the light intensity is low (ϳ10 lux), but with increasing light intensity, day-time sleep levels are comparable to s ϩ L ϩ flies, suggesting that remaining LN v that have not been ablated can modulate day-time sleep effectively especially in the presence of saturating light intensities (Extended Data Fig. 5-1A,B, left panels). Fur-continued ϭ 32 flies) and UAS Pdfr/ϩ (n ϭ 32 flies) controls (right). Day-time sleep is increased with downregulation, whereas it is decreased with OEX of pdfr in a subset of dopaminergic neurons driven by TH-D' GAL4. All other details are as in Figure 1. Results representative from at least two independent experiments. See also Extended Data Figures 4-1, 4-2, 4-3, 4-4, 4-5.

Figure 5.
Anatomic connections between TH ϩ and PDF ϩ neurons. A, Reconstituted GFP (GRASP) signal was detected in brains of thermore, day or night-time sleep levels are not altered when s-LN v are dysfunctional and l-LN v are normally firing (s -L ϩ ). Interestingly, the finding that flies with hyperexcited LN v show unchanged day-time sleep levels even in low light intensity LD12:12 cycles suggest that lightresponsive l-LN v can be saturated in terms of their firing capacity with as low light intensity as 10 lux. However, night-time sleep levels are always significantly lower than controls when both LN v are hyperexcited (s H L H ; Extended Data Fig. 5-1A,B, right panels). Interestingly, night-time sleep levels of flies with dysfunctional s-LN v but hyperexcited l-LN v (s -L H ) is significantly reduced as compared to controls, but always higher than the s H L H flies (Extended Data Fig. 5-1A,B, right panels). In fact, under LD12:12 with low light intensity days (10 lux), we find that night-time sleep levels of s -L H flies are comparable with their s ϩ L ϩ controls (Extended Data Fig. 5-1A,B, right-top). These results validate that l-LN v modulate wakefulness and further show that functional s-LN v are required to mediate these effects.

PDFR signaling inhibits PPM3 neuronal activity specifically during the day
Previous studies have shown that binding of PDF to PDFR results in a strong increase of cyclic AMP (cAMP) levels (Mertens et al., 2005; and moderate increase of intracellular calcium (Ca 2ϩ ) levels when expressed in HEK293 cells (Mertens et al., 2005). To test whether cAMP is the second messenger involved in mediating wakefulness through PDFR signaling, we overexpressed using TH-F3 GAL4 either the catalytic or regulatory subunit of cAMP-dependent protein kinase A (PKA; PKACA and PKAR) which increases or reduces PKA activity, respectively. We found no significant changes in day-time sleep as a result of increasing or decreasing PKA activity (Extended Data Fig. 6-1A,B,D; one-way ANOVA, F (2,86) ϭ 0.44, p ϭ 0.64), although daytime sleep of TH-F3 GAL4 Ͼ UAS PKAR flies was significantly lower as compared to only the UAS PKAR control flies (Extended Data Fig. 6-1C; one-way ANOVA, F (2,77) ϭ 4.41, p Ͻ 0.05). This shows that cAMP may not be the secondary messenger responding to PDFR signaling in the TH-F3 ϩ neurons, as changing PKA activity levels has negligible effects on day-time sleep.
To assess the functional importance of the connectivity between PDF ϩ and dopaminergic neurons, we next examined intracellular Ca 2ϩ levels in dopaminergic neurons at two time points, one during day [zeitgeber time (ZT)4; 4 h after lights-ON] and another during night (ZT14) in the presence and absence of functional PDFR (Pdfr 5304 mutant background). To quantify Ca 2ϩ levels, we used the recently developed CaLexA method which relies on calcium-dependent-nuclear transport of VP16:LexA to drive GFP downstream of LexAop responder element (Masuyama et al., 2012). When the CaLexA transgenes are expressed using the broad dopaminergic driver TH GAL4, we find that one to two neurons of some dopaminergic subsets notably within the PAL, PPM2, and PPM3 clusters express GFP (Table 2). However, differential expression of GFP depending on time point and genotype is observed only in approximately one to two PPM3 neurons per hemisphere (Table 3). We find that in the presence of functional PDFR, at ZT4 when the levels of PDF are also high (Park et al., 2000), the amount of GFP ϩ signal seen in continued flies expressing LexAop CD4::spGFP11 under Pdf LexA control and UAS CD4::spGFP1-10 under TH GAL4 control (n ϭ 22 hemispheres). GRASP signal colocalized with ascending portion of s-LN v dorsal projections labeled with antibody against PDF. Highlighted region is magnified in the right-most panel. Control flies lacking expression of (B) spGFP1-10 (n ϭ 16 hemispheres) and (C) spGFP11 (n ϭ 16 hemispheres) do not show GRASP signal. Results representative from two independent experiments. Arrowheads indicate non-specific staining. Expression of GFP using TH GAL4 and colabeling PDF (D, left) and using antibodies against TH and PDF (D, right) in wild-type flies reveal dopaminergic projection in the vicinity of ascending portion of s-LN v dorsal projection as indicated by asterisks. Scale bar: 20 m. See also Extended Data Figure 5-1. the PPM3 neurons is quite low (Fig. 6A,B; two-way ANOVA, F (1,70) ϭ 10.85, p Ͻ 0.005). However, at ZT14 when the levels of PDF are low (Park et al., 2000), amount of GFP ϩ signal seen in PPM3 neurons is significantly higher than that at ZT4 (Fig. 6A,B). In the absence of functional PDFR, at both ZT4 and ZT14, the GFP ϩ signal is high and not different from each other (Fig. 6A,B). Importantly, at ZT4, the GFP ϩ signal is significantly higher in the absence of functional PDFR than in its presence (Fig. 6). This shows that PDF acting on PDFR in the PPM3 dopaminergic neurons decreases Ca 2ϩ levels specifically during the day time.  Figure 1. See also Extended Data Figures 6-1, 6-2.
Given that dopaminergic neurons are wake-promoting, inhibiting them should inhibit wakefulness. Yet, increasing PDFR in PPM3 neurons which, in accordance with our calcium level quantification should cause increased inhibition, results in decreased day-time sleep. Similarly, absence of pdfr leads to reduced inhibition (as seen from increased Ca 2ϩ levels) of PPM3 neurons, yet behaviourally the flies sleep more during the day time. This indicates that PDF/PDFR signaling is acting on those PPM3 neurons that are, in effect, sleep-promoting. Thus, while a majority of dopaminergic neurons may be wakepromoting there may still be 1-2 PPM3 neurons which are PDFR ϩ and which effectively promote sleep. to examine this heterogeneity, we expressed the Drosophila temperature-sensitive cation channel dTRPA1 (Hamada et al., 2008) in different subsets of dopaminergic neurons using TH-D1, TH-D' and TH-F3 GAL4 drivers and examined the sleep levels of flies at a low ineffective temperature of 21°C as well as at a high activating temperature of 29°C. As reported in an earlier study (Liu et al., 2012), we find that flies sleep lesser both during the day and night when dopaminergic neurons driven by TH-D1 and TH-D' GAL4 are hyperexcited (Extended Data Fig. 6-2; one-way ANOVA, TH-D1: F (2,76) ϭ 23.16, p Ͻ 0.00001; TH-D': F (2,78) ϭ 28.88, p Ͻ 0.00001). However, when neurons expressed by the TH-F3 GAL4 are hyperexcited, flies tend to sleep as much as their GAL4 and UAS parental controls do, especially during the day time (Extended Data Fig.  6-2; one-way ANOVA, F (2,85) ϭ 2.29, p ϭ 0.1). This can happen only if these neurons do not actually have any effect on sleep, or if they are a heterogeneous group of wake-promoting and sleep-promoting neurons, such that hyperexciting both leads to a cancellation of effects caused by both groups. Given that TH-F3 GAL4-driven dopaminergic neurons have effects on sleep when pdfr levels are altered, our results point toward the possibility of one to two PDFR ϩ PPM3 neurons that are also sleeppromoting.

Discussion
Dopamine is primarily involved in promoting wakefulness (Andretic et al., 2005;Kume et al., 2005) and is known to act on l-LN v (Shang et al., 2011) as well as inhibit sleep-promoting dFB (Liu et al., 2012;Ueno et al., 2012) to carry out its wake-promoting function. Here, we uncover that certain dopamine neurons are in fact sleeppromoting and through the inhibitory action of PDFR signaling, wakefulness gets promoted specifically during the day. additional experiments that use optogenetic techniques can shed more light on whether these dopaminergic neurons promote sleep directly, or indirectly by preventing wakefulness either through a gating mechanism or by a permissive role. Interestingly, a previous study has found that dopamine acts on l-LN v to promote wakefulness (Shang et al., 2011) and we find that PDFR signaling acts on dopamine neurons, suggesting a feedforward pathway for wake promotion, where dopamine acting on l-LN v promotes the inhibition of sleeppromoting dopaminergic neurons by PDFR signaling. The identity of dopamine neurons acting on l-LN v and those responding to PDFR signaling may differ which can be uncovered with additional experiments.
The role of s-LN v in modulating sleep and wake has been explored in some detail in the recent years. s-LN v have also been shown to promote sleep via short NPF (sNPF) as well as myoinhibitory peptide (MiP) by inhibiting the wake-promoting l-LN v (Shang et al., 2013;Oh et al., 2014). Here, we show that PDF ϩ s-LN v make synaptic contacts with dopaminergic neurons (Fig. 5) and that PDFR signaling inhibits the downstream dopaminergic neurons (Fig. 6) to promote wakefulness during the day. Moreover, we have shown a secondary role for s-LN v in modulating wake-promoting effects of l-LN v . Yet, how this wake-promoting signal which originates in the l-LN v gets relayed to the s-LN v is not understood. Furthermore, from our screen it is clear that this function is not mediated via PDFR signaling among the LN v , as downregulating and overexpressing pdfr in s-LN v (Clk 9M GAL4 and Pdf GAL4) do not result in any sleep defects. Thus, l-LN v to s-LN v wake-promoting signal is independent of PDF while s-LN v to dopamine wake-promoting signal requires PDFR signaling.
PDFR being a class B1 GPCR utilizes cAMP as its second messenger Kunst et al., 2015), although there is evidence for Ca 2ϩ also acting as the second messenger (Mertens et al., 2005). For most of the functions of PDF including stabilizing core clock proteins such as TIMELESS and PERIOD in different target neurons such as DN 1 s and s-LN v , cAMP is the major secondary messenger (Li et al., 2014;Seluzicki et al., 2014). Moreover, it is thought that different actions of PDF of slowing and speeding up of morning and evening clock neurons is also mediated by different components of cAMP signaling mechanism (Duvall and Taghert, 2012;Duvall and Taghert, 2013). However, here we show that for the function of regulating wake levels during the day time, PDFR signaling changes levels of intracellular Ca 2ϩ in dopamine neurons with negligible role for cAMP signaling, suggesting a mechanism by which a neuropeptide that has diverse effects on its downstream targets can modulate different functions independently. We therefore identify a unique subset of downstream targets for PDFR signaling among the dopamine neurons that promote wakefulness depending on time of day.
Interestingly, in our screen we note that there are several driver lines using which there are significant changes in day-time sleep but with only one type of manipulation of pdfr levels (Clk 4.1M, 30y, 104y, 121y GAL4). This may be due to ineffective downregulation of pdfr achieved through the Pdfr RNAi line with these particular drivers. Given that PDF is a neuropeptide which can have longrange non-synaptic effects (Nässel and Winther, 2010), even misexpressing it (104y and 121y GAL4) in different substrates has resulted in altered day-time sleep levels. Because DH31 can also respond to PDFR (Kunst et al., 2015), it is possible that these effects could be mediated by DH31 binding to misexpressed PDFR. However, we find that this may not be the case as downregulating DH31-receptor in these regions does not cause changes in sleep levels (data not shown). Thus, we can conclude that in regions previously not known to express pdfr, misexpression of pdfr can cause sleep level deficits suggesting that PDF can act in regions which are not direct targets yet may lie in the vicinity of LN v projections.
The role of PDF/PDFR signaling is well-known in synchronizing the free-running molecular rhythms in neurons across the circadian network (Peng et al., 2003;Lin et al., 2004;Wu et al., 2008;Yoshii et al., 2009). PDFR signaling in the "evening" neurons (LN d and 5 th s-LN v ) is important for appropriate phasing of the evening bout of activity in light/dark cycles (Yao and Shafer, 2014;Guo et al., 2016). While the role of PDF as a wake-signal has been known, here we demonstrate that a subset of dopaminergic neurons is downstream of the PDF/PDFR signaling. While the PDFR expression is not conclusive, we show that perhaps one PPM3 neuron per hemisphere may express the PDFR. Additional experiments that more directly test the functional connectivity between dopaminergic neurons and PDF ϩ neurons, as well as responsiveness of dopaminergic neurons to PDF may result in a clearer picture. Downregulating pdfr in these neurons results in increase of day-time sleep, which is a phenocopy of the sleep behavior of loss-of-function pdfr whole-body mutants. On the other hand, overexpressing pdfr in these neurons leads to decrease of day-time sleep specifically. We further show that PDF and dopaminergic neurons make synaptic contacts with each other at the site of the axonal projection of s-LN v . Moreover, the effect of PDFR signaling on the PPM3 neurons appears to be inhibitory, suggesting that the PDFR ϩ PPM3 neurons promote sleep. Taken together, we conclude that wake-promoting LN v make synaptic connections with sleep-promoting dopaminergic neurons and promote wakefulness specifically during the day time through inhibitory PDFR signaling.