CGRP neurons mediate sleep-specific circadian output in Drosophila

Summary

Background

Imbalances in amount and timing of sleep are harmful to physical and mental health. Therefore, the study of the underlying mechanisms is of great biological importance. Proper timing and amount of sleep is regulated by both the circadian clock and homeostatic sleep drive. However, very little is known about the cellular and molecular mechanisms by which the circadian clock regulates sleep. In this study we describe a novel role for DIURETIC HORMONE 31 (DH31), the fly homologue of the vertebrate neuropeptide CALCITONIN GENE RELATED PEPTIDE (CGRP), as a circadian wake-promoting signal that awakens the fly in anticipation of dawn.

Results

Analysis of loss-of-function and gain-of-function Drosophila mutants demonstrates that DH31 suppresses sleep late at night. DH31 is expressed by a subset of dorsal circadian clock neurons that also express the receptor for the circadian neuropeptide PIGMENT DISPERSING FACTOR (PDF). PDF secreted by the ventral pacemaker subset of circadian clock neurons acts on PDF receptors in the DH31-expressing dorsal clock neurons to increase DH31 secretion before dawn. Activation of PDFR in DH31 positive DN1 specifically affects sleep and has no effect on circadian rhythms, thus constituting a dedicated locus for circadian regulation of sleep.

Conclusions

We identified a novel signaling molecule (DH31) as part of a neuropeptide relay mechanism for circadian control of sleep. Our results indicate that outputs of the clock controlling sleep and locomotor rhythms are mediated via distinct neuronal/cellular channels.

Introduction

Sleep is an essential physiological and behavioral process, conserved widely across diverse animal clades [for review, see 1]. However, the function of sleep remains largely elusive. Sleep loss is detrimental to memory performance and general health, and excessive sleep deprivation can lead to death [2–4]. Furthermore, it is not only important that we sleep but also when we sleep, as shift-work, jet-lag, and genetic disorders of circadian timing are associated with various physical and mental disorders [for review, see 5, 6]. The amount and timing of sleep are controlled by both the circadian control system and homeostatic sleep drive [7–10]. In vertebrates circadian rhythms are controlled by the suprachiasmatic nucleus (SCN) of the hypothalamus [for review, see 11]. Despite the recent explosion of knowledge concerning how the SCN keeps circadian time, the cellular and molecular mechanisms by which timekeeping information is propagated to the sleep system remain very poorly understood. So far, three circadian SCN output signals regulating sleep have been described (prokinecticin 2, cardiotrophin-like cytokine, and TGF-α) but their mechanisms of action are unknown [12–15].

The fruit fly, Drosophila melanogaster, has proven to be an outstanding system for investigating the basic cellular and molecular mechanisms that regulate sleep and circadian rhythms [for review see, 16, 17–19]. Genes regulating these behaviors and the basic neural architecture of sleep and circadian rhythms are evolutionarily conserved [for review see, 17, 20]. Drosophila are diurnal animals that mostly sleep at night, but also during a midday “siesta” [21, 22]. This daily rhythm of sleep is dependent on a functional circadian clock [22]. The circadian control network of the fly comprises ~150 neurons localized in six anatomically distinct cell groups divided into lateral (lLNv, sLNv, and LNd) and dorsal (DN1, DN2, and DN3) clusters [for review, see 18, 23]. Despite the essential role of the circadian clock in regulating sleep, only the PIGMENT DISPERSING FACTOR (PDF)-expressing LNvs, the key pacemaker neurons of the circadian network [24], have been implicated in the control of sleep [25–28]. Furthermore, the cellular and molecular mechanisms by which PDF or other unknown signals propagate out of the circadian network to influence sleep remain a mystery.

Here we identify the neuropeptide DIURETIC HORMONE 31 (DH31) as a circadian clock output factor that controls fly sleep. DH31 and its receptor (DH31-R1) are homologous to vertebrate CALCITONIN GENE RELATED PEPTIDE (CGRP) and its receptor (CLR) [29–32], which have been shown to increase locomotion in zebrafish [33], but whose role in sleep has not been investigated. By analysis of DH31 loss-of-function and gain-of-function mutations we demonstrate that DH31 is a wake-promoting signal that acts late at night to arouse flies in anticipation of dawn. DH31 secretion by a specific subset of DN1 circadian clock neurons mediates its wake-promoting effect, and functional imaging of a genetically encoded fluorescent voltage indicator reveals that the DH31-expressing DN1s are most electrically active before dawn. Furthermore, DH31 secretion by DN1 clock neurons is directly regulated by PDF signals from sLNvs. This PDF-to-DH31 peptide relay is specific to sleep and is not involved in circadian timekeeping, thus constituting a dedicated locus for circadian regulation of sleep. The receptors for PDF and CGRP are both class B1 GPCRs, which signal through G-α,s, adenylate cyclase, and intracellular cAMP. Our results reveal a novel class B1 neuropeptide relay mechanism for circadian control of sleep, and provide the foundation for further analysis of the cellular and molecular mechanisms by which DH31/CGRP neuropeptides regulate sleep.

Results

DH31 loss-of-function flies sleep more, and more deeply, at night

CGRP neuropeptides affect stress responses and anxiety in vertebrates, which strongly influence sleep [5, 34, 35]. To test the hypothesis that DH31 is involved in regulating sleep, we characterized flies with a P-element transposon insertion in the third intron of the DH31 locus (Figure 1A). This transposon insertion (DH31^KG09001) reduces DH31 mRNA levels by ~20-fold (Figure 1B) and almost completely abolishes anti-DH31 immunoreactivity (Figure 1C), establishing DH31^KG09001 as a severe loss-of-function allele.

Figure 1. DH31 loss-of-function mutation increases sleep.

(A) Structure of the DH31 gene, indicating the P-element insertion KG09001 in the third intron. (B) qPCR indicates that DH31 mRNA levels are reduced by 20-fold in DH31^KG09001 flies compared to wild-type. Levels of mRNA were normalized to expression of Rpl32. (C) Maximum-projection confocal images of the anterior and posterior regions of whole-mount adult brains of DH31^KG09001 and wild-type CantonS flies (CS). Brains were stained with antibodies to synaptic protein BRUCHPILOT to visualize neuropils (NC82, red) and DH31 (white). DH31 immunoreactivity is undetectable in DH31^KG09001 flies while CS flies have broad expression in the brain. Scale bar = 100µm. (D) Sleep profiles of DH31^KG09001 homozygous (red, n = 27), DH31^KG09001/DH31-Df hemizygous (blue, n = 39), and w¹¹¹⁸ control flies (black, n = 55). Lights-on (dawn) is at ZT0 (Zeitgeber Time 0) and lights-off (dusk) is at ZT12. Reduction of DH31 expression increases sleep at night. (E) Quantification of sleep for DH31^KG09001 (red), DH31^KG09001/DH31-Df (blue), and w¹¹¹⁸ control flies (black). (F-G) Quantification of sleep consolidation. DH31 loss-of-function flies have fewer sleep bouts of longer duration at night. (H) DH31 loss-of-function flies have unimpaired locomotion while awake as reflected in beam crossings per waking minute (activity index). (I-L) DH31 loss-of-function flies are less sensitive to the wake-promoting effect of carbamazepine (CBZ). (I-J) Sleep profiles of DH31^KG09001 (red, n = 31) and w¹¹¹⁸ (black, n = 32) on normal food (I) and food containing 0.4 mg/ml CBZ (J). (K) DH31^KG09001 flies are less sensitive than w¹¹¹⁸ (black, n = 32) to suppression of sleep induced by feeding CBZ. (L) Nighttime sleep of DH31^KG09001 flies (red) and w¹¹¹⁸ flies (black) on CBZ food normalized to control food. Error bars indicate the SEM. Statistical comparisons are by one-way ANOVA with Dunnet’s post-hoc paired-comparison test, except for sleep bout durations, which are compared using the Wilcoxon rank-sum paired-comparison test. n.s. = not significant, * = p < 0.05, ** = p < 0.01, *** = p < 0.001. See also Figure S1.

We employ the well-established infrared beam-crossing locomotor assay to measure sleep [36]. Since genetic background substantially influences fly sleep [37], we extensively backcrossed the DH31^KG09001 allele into a common genetic background (w¹¹¹⁸). We also generated hemizygous flies with DH31^KG09001 opposite a chromosomal deficiency spanning the DH31 locus (Df(2L)Exel(7038), hereafter referred to as DH31-Df). We quantified two key features of sleep, duration and consolidation, with consolidation reflected in the number and duration of sleep bouts. A low number of sleep bouts of longer duration reflects high consolidation, and vice versa.

Both DH31^KG09001 and DH31^KG09001/DH31-Df female flies sleep more at night than w¹¹¹⁸ controls (Figures 1D and E). Hemizygous flies are more severely affected, consistent with them being completely null for one allele of DH31. The nighttime sleep of DH31 loss-of-function flies is also more consolidated than that of w¹¹¹⁸ control flies, manifested in fewer sleep bouts of longer duration (Figures 1F and G). We also combined DH31^KG09001 with nearby deficiencies from the same genetic background as DH31-Df, but which do not span the DH31 locus. These flies possessing one wild-type DH31 allele sleep less than the hemizygous flies (Figures S1A–C). These results establish that increases in sleep in DH31^KG09001 flies are caused by DH31 loss-of-function. None of these effects on sleep can be attributed to general deficits in locomotor activity, as DH31 loss-of-function flies engage in as much locomotor activity while awake as w¹¹¹⁸ controls flies (Figures 1H).

The increase in sleep of DH31 loss-of-function flies is most prominent shortly before dawn. A statistically significant increase in sleep in DH31 loss-of-function flies is only observed in the second half of the night (Figures S1D and E). The increased late-night sleep of DH31 loss-of-function flies suggests that DH31 is a negative regulator of sleep maintenance that awakens flies in anticipation of dawn.

We also tested whether DH31 loss-of-function flies would be resistant to pharmacological suppression of sleep. Carbamazepine (CBZ) enhances desensitization of Drosophila GABA_A receptors and thereby suppresses sleep [38, 39]. As expected, CBZ administered by feeding significantly suppresses sleep duration (Figures 1I and J). However, the sleep-suppressing effect of CBZ is strongly attenuated in DH31^KG09001 homozygous flies (Figures 1K and L), indicating that DH31 signaling is important for CBZ-induced suppression of sleep. This suggests that under normal conditions GABA-mediated promotion of sleep is mediated in part through suppression of DH31 secretion.

Overexpression of DH31 suppresses sleep at night

Since DH31 loss-of-function flies sleep more at night (Figure 1), we tested whether increasing DH31 signaling causes flies to sleep less (Figure 2). Overexpression of DH31 using the UAS/GAL4 system [40] with a pan-neuronal nsyb driver suppresses nighttime sleep (Figures 2A and B). Overexpression of DH31 significantly suppresses sleep specifically during the second half of the night, and this decrease in sleep is accompanied by an increase in sleep fragmentation (Figures S2A–C). These effects cannot be attributed to a non-specific increase in locomotor activity, as locomotor activity while awake is no more than in controls (Figure S2D). DH31 acts primarily on the DH31 receptor (DH31-R1) but has also been shown to modestly activate the receptor for the circadian neuropeptide PDF (PDFR) [41, 42], which also promotes wake. To rule out the possibility that the wake-promoting effect of DH31 is mediated through activation of PDFR we pan-neuronally overexpressed DH31 in the han⁵³⁰⁴ PDFR null-mutant [43], which has the same effect on sleep as in PDFR wild-type flies (Figures S2E–I). These effects of DH31 overexpression further support the conclusion that DH31 down regulates sleep maintenance late at night and awakens the fly before dawn.

Figure 2. DH31 overexpression suppresses sleep.

(A) Sleep profiles of flies overexpressing DH31 in all neurons (nsyb>DH31, blue, n = 30) compared to empty>DH31 control (black, n = 28). (B) Flies overexpressing DH31 pan-neuronally (blue) sleep significantly less at night than controls (black). (C) Sleep profiles of flies overexpressing DH31 using R20A02. (D) Overexpression of DH31 using the DH31-derived R20A02 driver (red, n = 26) significantly reduces nighttime sleep compared to empty control, other DH31-derived drivers (R21E02, R21C09, and R20D02, black, n = 25, 26, and 26) have no effect. (E) Conditional activation of R20A02 neurons using the heat-sensitive cation channel dTRPA1. Flies were maintained for two days at 21.5°C to measure baseline sleep, shifted to 29.5°C for two days, and then returned to 21.5°C for two more days. Activation of R20A02 neurons (red, n = 32) decreases sleep compared to control (black, n = 31). (F) R20A02>dTRPA1 flies sleep significantly less on both days of heat induction than empty>dTRPA1 control flies. (G) Maximum-projection confocal images of the anterior and posterior regions of whole-mount adult fly brains expressing myrGFP driven by R20A02. Brains were stained with antibodies against NC82 (magenta) and GFP (green). GFP is expressed in the mushroom body (MB), antennal lobe (AnL), ellipsoid body (EB), suboesophageal ganglion (SOG), medulla (Med) and lobula (Lob) of the optic lobes (OL), as well as in somata of the dorsal neuron 1 group of circadian clock neurons (DN1) and their axonal projections in the posterior superior medial protocerebrum (pSMP). Scale bars = 50 µm. Error bars indicate the SEM. All statistical comparisons are made by one-way ANOVA with Dunnet’s post-hoc test. n.s. = not significant, * = p < 0.05, ** = p < 0.01, *** = p < 0.001. See also Figure S2–S4.

To identify cellular sources of DH31 capable of regulating sleep, we overexpressed DH31 using GAL4 drivers from a recently established collection generated at Janelia Farm Research Campus designed for targeting restricted cell populations in the CNS [44]. Each of the lines in this collection expresses GAL4 under the control of sequences derived from non-coding enhancer regions of neurally expressed genes [45]. Of four lines tested derived from enhancer fragments of the DH31 locus, only one (R20A02) significantly suppresses sleep when used to drive DH31 expression (Figures 2C and D, Table S1). This suppression of sleep is greatest in the second half of the night, accompanied by decreased sleep consolidation (Figures S3A–C) and increased waking locomotor activity (Figure S3D). To determine whether the neurons targeted by R20A02 intrinsically suppress sleep in the absence of exogenous DH31 expression, we expressed the heat-sensitive cation channel dTRPA1 [46] using R20A02 and elevated the ambient temperature to increase neuronal activity. dTRPA1-mediated activation of R20A02 neurons significantly suppresses both daytime and nighttime sleep (Figures 2E and F). Daytime sleep during the two days after cessation of dTRPA1 activation is significantly increased, reflecting homeostatic sleep rebound (Figures S3E). These results indicate that R20A02 neurons are intrinsically wake-promoting.

To determine the identity of R20A02 neurons, we drove expression of membrane-bound myristolated-GFP (myrGFP). R20A02 is active in the suboesophageal ganglion (SOG), antennal lobe (AnL), mushroom bodies (MB), medulla and lobula of the optic lobes (OL), ellipsoid body (EB) of the central complex, and DN1 circadian clock neurons that project to the posterior superior medial protocerebrum (pSMP) (Figures 2G, Figures S4A–D). Comparing the expression of R20A02 to the three DH31-derived lines that did not influence sleep when used to drive DH31 expression, the only two brain regions unique to R20A02 were EB and the DN1 clock neurons (Figures 2, Figures S4E–J).

DH31-expressing DN1 clock neurons regulate sleep

Endogenous DH31 is expressed in both DN1s and the EB (Figures S5A and B). To further refine the identities of DH31-expressing neurons that regulate sleep, we returned to the Janelia GAL4 driver collection. Exogenous expression of DH31 using the R43D05 driver line significantly suppresses nighttime sleep. This driver is active in circadian clock neurons and its ability to suppress sleep is strongly reduced by the introduction of the cryGAL80 suppressor transgene (Figures S5C–E, Table S1), which specifically inhibits GAL4 activity in circadian clock neurons [47]. This indicates that DH31 secretion by circadian clock neurons is capable of regulating sleep.

The R18H11 Janelia driver line is active in a restricted subset of DN1 clock neurons (4–6) that endogenously express DH31. It exhibits very limited expression in other brain regions, namely the SOG and in a diffuse set of cells in the lateral brain region, but no expression can be detected in the EB (Figures 3A and B). Restoring expression of DH31 in DH31^KG09001 homozygous mutant flies using the R18H11 driver rescues waking during the three hours before dawn (Figure 3C). Immunostaining for DH31 in these flies reveals detectable expression solely in DN1 clock neurons (Figure 3D). Restoring DH31 expression using the R20A02 driver also rescues waking before dawn (Figure 3C) and restores expression of DH31 in DN1s as well as in the EB (Figure 3E). This cell-specific rescue in DH31 loss-of-function flies supports a role for endogenous DH31 secretion by DN1s in waking the fly before dawn.

Figure 3. Restoration of DH31 expression in DN1s of DH31^KG09001 flies rescues late-night awakening.

(A) Maximum-projection confocal images of the posterior (top panel) and anterior regions (bottom panel) of whole-mount adult fly brains expressing myrGFP driven by R18H11-GAL4. Brains were stained with antibodies against NC82 (magenta) and GFP (green). GFP is expressed in the SOG and in DN1 circadian clock neurons that send projections to the pSMP. Scale bars = 50 µm. (B) Co-labeling of endogenous DH31 (magenta), GFP-expression driven by R18H11 (green), and the core endogenous circadian protein PERIOD (PER, blue) reveals colocalization in five DN1s (arrows). Scale bar = 10 µm. (C) DH31 rescue flies (DH31^KG09001; R18H11>DH31, red, n = 91 and DH31^KG09001; R20A02>DH31, blue, n = 90) sleep significantly less in the three hours before dawn than control flies (DH31^KG09001; empty>DH31, black, n = 93). (D and E) Maximum-projection confocal images of anterior (left) and posterior (right) regions of whole-mount adult brains expressing DH31 using R18H11 (D) or R20A02 (E) in DH31^KG09001 mutant flies. Scale bar = 50 µm. Error bars indicate the SEM. All statistical comparisons are by one-way ANOVA with Tukey-HSD post-hoc test. n.s. = not significant, * = p < 0.05, ** = p < 0.01, *** = p < 0.001. See also Figure S5.

To test whether DH31-expressing DN1s regulate sleep in the absence of exogenous DH31 expression we activated them using dTRPA1. Temperature-induced activation of R18H11 neurons suppresses sleep (Figures 4A and C), and this suppression is strongest in the hours before and after dawn (arrows in Figure 4A). To confirm that this suppression of sleep is mediated by the DH31-expressing DN1s, we inhibited dTRPA1 expression in clock neurons using cryGAL80, thereby significantly reducing suppression of sleep (Figures 4B and C). These results demonstrate that DH31-expressing DN1 circadian clock neurons are intrinsically wake-promoting.

Figure 4. Activation of DH31-expressing DN1 clock neurons suppresses sleep.

(A) Conditional activation of R18H11 neurons (R18H11>dTRPA1, red, n = 26) decreases sleep compared to empty>dTRPA1 control (black, n = 29). Suppression of sleep is strongest in the hours before and after lights-on (arrows). (E) Suppressing dTRPA1 expression in circadian clock neurons using cryGAL80 (R18H11+cryGAL80>dTRPA1, blue, n = 24) significantly reduces this effect compared to control flies (empty+cryGAL80>dTRPA1, black, n = 31). (F) Quantification of sleep for empty>dTRPA1 with or without cryGAL80 (black) and R18H11>dTRPA1 with (blue) or without cryGAL80 (red). cryGAL80 strongly suppresses the wake-promoting effect of dTRPA1 activation in R18H11 neurons. Error bars indicate the SEM. All statistical comparisons are by one-way ANOVA with Tukey-HSD post-hoc test. n.s. = not significant, * = p < 0.05, ** = p < 0.01, *** = p < 0.001.

Cell-autonomous activation of DH31 receptors suppresses sleep

To identify cellular targets of DH31 that promote waking activity, we cell-autonomously activated DH31-R1 by transgenic expression of membrane-tethered DH31 [tDH31, 48, 49]. tDH31 is a genetically encoded DH31-R1 agonist comprising the DH31 peptide covalently linked to the cell surface via a C-terminal GPI (glycosylphosphatidylinositol) anchor [49]. GPI anchoring limits the action of tDH31 to the cell in which it is expressed, where it cell-autonomously activates native DH31-R1 without affecting DH31-R1 on neighboring cells [49]. Importantly, tDH31 does not activate other class B1 neuropeptide receptors, including PDFR [48, 49].

Pan-neuronal expression of tDH31 suppresses sleep at night (Figures 5A and B), consistent with the suppression of nighttime sleep induced by overexpression of soluble DH31 (Figure 2). This establishes that DH31 suppresses sleep by acting on DH31-R1 in neurons, and not in glia or other non-neuronal cells. More restricted neuronal expression of tDH31 with a variety of GAL4 drivers has no effect on sleep (Figure 5E, Table S2), perhaps because DH31 promotes wake by acting on DH31-R1-expressing neurons not targeted by any of these drivers. Alternatively, DH31 may promote wake by acting on a widely distributed network of DH31-R1-expressing neurons. Flies expressing tDH31 in all neurons not only sleep less, their sleep is also less consolidated, reflected in increased number of sleep bouts of shorter duration (Figure 5C and D). These effects cannot be attributed to general locomotor hyperactivity, as locomotor activity while awake is no more than that of control flies (Figure 5E). The suppression of sleep by tDH31 is most prominent during the second half of the night (Figure S6), consistent with the timing of the sleep effects in DH31 loss-of-function and gain-of function mutant flies (Figures 1–3). tDH31-induced sleep suppression is stronger than that induced by DH31 overexpression (Figures 2 and 3), consistent with the dependence of DH31 secretion on neural activity as opposed to the constitutive activation of DH31-R1 by tDH31.

Figure 5. Constitutive activation of DH31 receptor suppresses sleep.

(A) Sleep profiles of flies expressing tDH31 pan-neuronally (nsyb>tDH31, blue, n = 30) compared to empty>tDH31 (black, n = 31) and nsyb>tSCRPDF controls (red, n = 31). (B) Constitutive activation of DH31 receptor (DH31-R1) significantly reduces nighttime sleep. (C-D) Constitutive activation of DH31-R1 increases the number of sleep bouts (C) while reducing their duration (D), reflecting reduced sleep consolidation. (E) The sleep-suppressing effect of pan-neuronal tDH31-expression is not due to general locomotor hyperactivity. (F) Overexpression of tDH31 using DH31-R1-derived GAL4 lines, lines active in the central complex (CX), circadian clock neurons, pars intercerebralis (PI), and GAL4 lines active in neurons expressing classical neurotransmitters and biogenic amines (broad). Values plotted are differences in nighttime sleep compared to empty>tDH31 or the respective GAL4 driver crossed to UAS>tSCRPDF. Error bars indicate the SEM. Statistical comparisons are by oneway ANOVA with Dunnet’s post-hoc paired-comparison test, except for sleep bout durations, which are compared using the Wilcoxon rank-sum paired-comparison test. Statistical analysis of nighttime sleep in (F) was by Student’s t-test adjusted for multiple comparisons using the Benjamini-Hochberg method. n.s. = not significant, * = p < 0.05, ** = p < 0.01, *** = p < 0.001. See also Figure S6.

PDFR activation in DH31-expressing DN1s suppresses sleep late at night

The circadian neuropeptide PDF secreted by sLNv pacemaker neurons suppresses sleep [25, 27, 28], but the cellular targets upon which PDF acts to regulate sleep remain unknown. However, a number of studies implicate the DN1s as a direct target of PDF signals from sLNvs that are important for regulating circadian rhythms [48, 50–58]. Furthermore, PDF secretion by sLNvs is highest around dawn [59]. We thus hypothesized that DH31 secretion by DN1s is directly regulated by PDF activation of PDFR late at night, constituting a pacemaker-controlled sleep-regulating output of the circadian clock. Consistent with this hypothesis we detected PDFR expression in the DH31-expressing DN1s (Figure S7A).

To test whether activation of PDFR in DH31-expressing DN1s regulates their wake-promoting activity, we expressed tPDF in these cells using R18H11. Constitutive activation of PDFR in DH31-expressing DN1s suppresses sleep late at night (Figures 6A and D, Figures S7B). This is consistent with the effect of dTRPA1-mediated neuronal activation (Figure 4) and suggests that PDFR activation electrically excites DH31-expressing DN1s. It was recently shown that bath-applied PDF electrically excites some DN1s, although that study targeted a larger subset of DN1s whose neurochemical identities are unknown [58]. The sleep-suppressing effect of tPDF expression is blocked by cryGAL80 (Figure 6B and D, Figures S7C), establishing that it is due to PDFR activation in the DH31-expressing DN1s. To determine whether the suppression of sleep induced by activation of PDFR in DN1s is mediated by regulation of DH31 secretion we expressed tPDF in DH31^KG09001 homozygous mutant flies. The suppression of sleep over the twelve-hour night by PDFR activation is abolished in DH31^KG09001 mutant flies (Figures 6C and D) although a small, statistically significant decrease in sleep remains during the second half of the night (Figures S7D). This residual wake-promoting effect of PDFR activation could be mediated by the low level of DH31 still expressed in DH31^KG09001 mutant flies, and is consistent with the more severe sleep phenotype of hemizygous DH31^KG09001/DH31-Df flies than homozygous DH31^KG09001 flies. These results establish that DH31 is a PDF-regulated sleep-controlling circadian clock output.

Figure 6. Activation of PDFR in DH31-expressing DN1s suppresses sleep by regulating DH31 secretion late at night.

(A–C) Sleep profiles of flies expressing tPDF in DH31-expressing DN1s (R18H11>tPDF, blue) compared to empty>tPDF (black) and R18H11>tSCRPDF (red) controls. (A) Activation of PDFR in DH31-expressing DN1s of DH31 wild-type flies reduces sleep at night. (B) Inhibiting GAL4 activity in circadian clock neurons by cryGAL80 suppresses the wake-promoting effect of PDFR activation. (C) The wake-promoting effect of PDFR activation is severely reduced in DH31^KG09001 mutant flies. (D) Quantification of sleep during the night. Error bars indicate the SEM. All statistical comparisons are by one-way ANOVA with Dunnet’s post-hoc test. n.s. = not significant, * = p < 0.05, ** = p < 0.01, *** = p < 0.001. (E-F) Representative optical recordings of membrane potential in cell bodies (green) and processes (orange and purple) of DH31-expressing DN1s in adult brains explanted at ZT22 (A) and ZT10 (B). Scale bars = 10 µm. (G) Standard deviation of the optical signal was computed for each soma region (unfilled circles), with mean ± SEM (filled circles). The standard deviation at ZT22 (n=15) is significantly greater than at ZT10 (n=12) (unpaired t-test, p < 0.05). (H) Power spectrum was computed for each soma region using fast Fourier transform with 0.2 Hz bin width. Powers at each frequency were averaged (±SEM) across soma. ZT22 power is significantly greater than ZT10 power (two-way ANOVA with repeated measures, p < 0.001). See also Figure S7.

Electrical activity of DH31-expressing DN1s is high late at night

If DH31 secretion by DN1s awakens the fly in anticipation of dawn, then electrical activity of DH31-expressing DN1s is predicted to be high late at night. To directly measure electrical activity, we expressed the genetically encoded fluorescent voltage indicator ArcLight [59, 60] using the R18H11 driver. This approach has already been successfully used to demonstrate that PDF is released in a circadian fashion by sLNv terminals in the dorsal protocerebrum [59]. Optical recording of membrane potential in whole-brain explants demonstrates that the DH31-expressing DN1s are more active late at night (ZT22) than late in the day (ZT10) (Figures 6E–H). This further supports the hypothesis that the circadian neuropeptide PDF awakens the fly through PDFR dependent activation of DH31 secretion by DN1s late at night.

DH31 is a sleep-specific circadian clock output

Since DH31 is expressed in circadian clock neurons and is secreted by these neurons with a daily rhythm, we asked whether DH31 regulates not only sleep but also circadian timekeeping. DH31^KG09001 homozygous and DH31^KG09001/DH31-Df hemizygous loss-of-function flies exhibit identical strength and period of free-running circadian locomotor rhythms as w¹¹¹⁸ control flies (Figures 7A–C, upper panels). Furthermore, constitutive activation of DH31-R1 by pan-neuronal expression of tDH31 has no effect on rhythm strength, although it does induce a small decrease in free-running period (Figures 7A–C, middle panels). Since PDF signaling to DN1s is critical for morning anticipatory locomotor activity and robust free-running locomotor activity rhythms [49, 55, 61–64], we constitutively activated PDFR in the DH31 expressing DN1s using tPDF and found no effects on free-running circadian rhythms (Figures 7A–C, lower panels). These results indicate that DH31 only very weakly influences circadian rhythms, and only in the context of non-physiological constitutive activation of DH31-R1 throughout the entire brain, thus establishing DH31 as a PDF-regulated sleep-specific output of the circadian clock that awakens the fly in anticipation of dawn.

Figure 7. The PDF-to-DH31 sleep-suppressing relay does not affect circadian rhythms.

(A) Double-plotted average actograms spanning 10 days in constant darkness (DD). Top: DH31 loss-of-function mutants. Middle: Constitutive DH31-R1 activation throughout the brain by pan-neuronal expression of tDH31. Bottom: Constitutive activation of PDFR in DH31-expressing DN1s. (B) Average free-running period. Error bars indicate the SEM. All statistical comparisons are by one-way ANOVA with Dunnet’s post-hoc test. n.s. = not significant, ** = p < 0.01. (C) Percentage of rhythmic (R, blue) and arrhythmic flies (AR, black). Statistical comparisons are by χ² test.

Discussion

Vertebrate CGRP has been implicated in controlling anxiety [65, 66] and stress response [67, 68]. While not previously addressed experimentally, the intimate relationship between stress, anxiety, and sleep [5] suggest that CGRP might regulate sleep. Potentially relevant to this possibility is the recent observation that acute activation of CGRP signaling in zebrafish larvae increases spontaneous locomotor activity and decreases quiescence [33]. Our analysis of gain-of-function and loss-of-function mutant flies establishes that DH31, the Drosophila homologue of CGRP, is a negative regulator of sleep maintenance that awakens the animal in anticipation of dawn. This finding motivates investigation of a potential role for CGRP in the regulation of vertebrate sleep.

Using powerful tools for cell-specific neuronal manipulation we identified a highly restricted subset of DN1 circadian clock neurons that secrete DH31 late at night to awaken the fly in anticipation of dawn. Neuropeptides secreted from the circadian pacemaker in the mammalian SCN, such as prokinecticin 2 and cardiotrophin-like cytokine, are important clock outputs [69], but their cellular targets and molecular mechanisms remain unknown. In flies, PDF-expressing sLNv pacemaker neurons are known to be upstream of DN1s [55, 57, 58], sLNvs are most active around dawn [59, 70], and PDF secretion from LNvs suppresses sleep at night [25–27]. However, how PDF signals propagate out of the circadian clock network to regulate sleep remains unknown.

Here we show that PDF secreted by the sLNv pacemaker neurons activates PDFR in the DH31-expressing DN1s to increase neuronal activity and DH31 secretion late at night, thereby awakening the fly in anticipation of dawn. This is consistent with an earlier report demonstrating that flies lacking PDF or PDFR sleep more late at night [28]. However, unlike DH31, PDF also promotes wake during the day [25–27]. This suggests that PDF regulation of daytime sleep is mediated by neurons other than the DH31-expressing DN1s. PDFR is expressed by circadian clock neurons in addition to DN1s, and PDF signaling to these other clock neurons could be responsible for the wake-promoting effect of PDF during the day.

While PDF plays a key role in circadian timekeeping [24, 47, 49, 53, 55, 61, 62, 64, 71–74], neither constitutive activation of PDFR in the DH31-expressing DN1s nor manipulation of their secretion of DH31 affect free-running circadian rhythms. This establishes the PDF-to-DH31 neuropeptide relay as a novel sleep-specific output of the circadian pacemaker network, independent from and parallel to the outputs that drive circadian rhythms themselves. Since the basic cellular and molecular organization of vertebrate and insect circadian networks is conserved, this motivates the search for similar mechanisms in the SCN. Future studies are required to identify the neuronal targets of sleep-regulating DH31 signals and the cellular and molecular mechanisms by which DH31-R1 activation in these targets induces wake.

Experimental Procedures

Sleep was measured using the well-established infrared beam crossing assay [36]. Circadian rhythm measurement, immunohistochemistry, qPCR, and imaging were performed as described previously[48, 59, 75]. Detailed descriptions of the methods are available in the Supplemental Experimental Procedures.