Circadian gating of light-induced arousal in Drosophila sleep

Abstract Circadian rhythms and sleep homeostasis constitute the two-process model for daily sleep regulation. However, evidence for circadian control of sleep-wake cycles has been relatively short since clock-less animals often show sleep behaviors quantitatively comparable to wild-type. Here we examine Drosophila sleep behaviors under different light-dark regimes and demonstrate that circadian clocks gate light-induced arousal. Genetic excitation of tyrosine decarboxylase 2 (TDC2)-expressing neurons suppressed sleep more evidently at night, causing nocturnal activity. The arousal effects were likely mediated in part by glutamate transmission from the octopaminergic neurons and substantially masked by light. Application of T12 cycles (6-h light: 6-h dark) further showed that the light-sensitive effects of TDC2 neurons depended on the time of the day. In particular, light-sensing via visual input pathway led to strong sleep suppression at subjective night, and such an effect disappeared in clock-less mutants. Transgenic mapping revealed that light-induced arousal and free-running behavioral rhythms require distinct groups of circadian pacemaker neurons. These results provide convincing evidence that circadian control of sleep is mediated by the dedicated clock neurons for light-induced arousal.


Introduction
Sleep is a highly conserved behavior among various species (Anafi et al., 2019). While sleep is essential for normal brain function, a sleep-like state is observed even in animals that lack a centralized nervous system (Kanaya et al., 2020;Nath et al., 2017), indicating an ancestral origin. A two-process model has been proposed to explain how animals shape daily sleep (Borbely & Achermann, 1992. Process S sustains sleep homeostasis by gradually increasing 'sleep pressure' during wakefulness and dissipating it during sleep. On the other hand, circadian rhythms constitute Process C that times sleep onset and offset on a 24-h timescale. On top of these two processes, animal sleep could be further modified by external sleep cues or internal states, leading to plasticity in sleep behaviors and sleep-relevant physiology (Beckwith & French, 2019).
Daily oscillations in gene expression are the essence of circadian rhythms at molecular levels and are driven by transcriptional and translational feedback loops (TTFL). The interlocked network of TTFLs is well conserved between flies and mammals, generating the cell-autonomous molecular clocks (Hardin & Panda, 2013;Lim & Allada, 2013;Zheng & Sehgal, 2012). In Drosophila, CLOCK-CYCLE (CLK-CYC) heterodimer activates the transcription of circadian clock genes, including period (per) and timeless (tim). Cytoplasmic PER proteins then associate with TIM to enter the nucleus and inhibit the transcriptional activity of CLK-CYC, thereby suppressing their own transcription. Additional clock genes constitute secondary and tertiary TTFLs to stabilize the core molecular oscillator above.
Although the molecular clock operates in a cell-autonomous manner, the adult fly brain has $150 circadian pacemaker neurons that robustly display circadian expression of clock genes and sustain behavioral rhythms by the network property (Hermann-Luibl & Helfrich-Forster, 2015;Kozlov & Nagoshi, 2019). The pacemaker neurons are divided into distinct groups based on their anatomical, structural and functional heterogeneity. These include large and small lateral ventral neurons (l-LNv and s-LNv, respectively) expressing a circadian neuropeptide called PIGMENT-DISPERSING FACTOR (PDF) (Renn et al., 1999), PDFnegative 5th s-LNv, dorsal LNs (LNd), three different clusters of dorsal neurons (DN1, DN2, and DN3), and lateral posterior neurons (LPNs).
Importantly, s-LNv acts as the morning oscillator to control the anticipatory activity to lights-on in light-dark cycles (LD) and drive free-running locomotor rhythms in constant dark (DD) (Grima et al., 2004;Stoleru et al., 2004), whereas LNd and 5th s-LNv serve as the evening oscillator to anticipate lights-off in LD and sustain behavior rhythms in constant light (LL) (Cusumano et al., 2009;Picot et al., 2007). The circadian function of the two oscillators is consistent with anti-phase rhythms in their intracellular Ca2þ levels (Liang et al., 2016). Nonetheless, it remains elusive how circadian clocks adaptively intervene in sleep regulation upon various sleep cues. In this study, we address the question by demonstrating that molecular clocks in a subset of the Drosophila circadian pacemaker neurons mediate timedependent arousal by light.

Sleep behavior analysis
A single-male fly was loaded into a 65 Â 5 mm (length x diameter) glass tube containing 5% sucrose and 2% agar (behavior food). Locomotor behaviors in individual flies were quantified by the number of the infrared beam crosses per minute and recorded by the Drosophila Activity Monitor (DAM) system (Trikinetics). For sleep analysis, flies were entrained under 12-h light: 12-h dark (LD) cycles for 5 days at 25 C. For sleep analysis in LL or DD, flies were entrained under the standard LD cycle for 3 days at 25 C and subsequently transferred to LL or DD cycles. For T12-cycle (6-h light: 6-h dark) sleep analysis, flies were entrained under the standard LD cycle for 2 days at 25 C and then transferred to T12 cycles. For TrpA1 experiments, flies were entrained under LD cycles for 3 days at 21 C prior to the temperature shift to 29 C . They were further incubated in LD cycles for 3 days at 29 C. For oral administration of MK801, (þ)-MK801 maleate (Tocris) was dissolved in the behavior food at a final concentration of 0.05 mg/mL. Flies were loaded onto the drug-containing food on the first day of LD cycles, and their locomotor behaviors were similarly recorded as above. Raw behavior data were collected by the DAM system and analyzed using an Excel macro as described previously (Pfeiffenberger et al., 2010). The asleep bout was defined if a given fly showed no detectable activity longer than 5 min.

Whole-mount brain imaging
Whole brains were dissected from transgenic flies and incubated in PBS containing 3.7% formaldehyde for 28 min at room temperature. The fixed brains were washed 20 min Â 3 times with PBS containing 0.3% of Triton X-100 and then mounted in a VECTASHIELD mounting medium (Vector Laboratories). Fluorescence images were obtained using an LSM780 laser-scanning microscope (Zeiss).

Statistical analysis
Kruskal-Wallis ANOVA, one-way ANOVA, Welch's ANOVA, unpaired t-test, and Mann-Whitney test were performed using GraphPad Prism 9. Aligned rank transformation (ART) one-way and two-way ANOVAs were performed using R (version 4.2.0) as described previously .

TDC2-expressing neurons suppress nighttime sleep likely via co-transmission of octopamine and glutamate
Octopamine is one of the monoamine neurotransmitters in insects and is closely related to norepinephrine in mammals (Roeder, 2005). It is implicated in a broad range of insect physiology including aggression, motor behaviors, alcohol tolerance, metabolism, and arousal (Roeder, 2020;Selcho & Pauls, 2019). Octopamine biosynthesis requires tyrosine decarboxylase 2 (TDC2), which removes the carboxyl group from tyrosine, producing the octopamine precursor tyramine. Genetic deficiency in the octopamine biosynthesis pathway leads to long daytime sleep in Drosophila (Crocker & Sehgal, 2008). On the other hand, transgenic excitation of the TDC2-expressing octopaminergic neurons suppresses nighttime sleep, defining octopamine as a wake-promoting neurotransmitter (Crocker & Sehgal, 2008;Kayser et al., 2015). Intriguingly, most Drosophila octopaminergic neurons are also glutamatergic, and their co-transmission has been shown to play differential roles in aggression and courtship behaviors (Sherer et al., 2020). We thus asked if the dual neurotransmission would contribute to octopamine-relevant sleep regulation.
Indeed, RNA interference (RNAi)-mediated depletion of vesicular glutamate transporter (VGLUT), tyramine b hydroxylase (an enzyme converting tyramine to octopamine), or TDC2 in TDC2 neurons comparably suppressed short nighttime sleep in Tdc2>TrpA1 flies (Figure 1(D) and supplemental Figure S2). By contrast, TDC2 neuron-specific expression of the choline acetyltransferase (ChAT) RNAi transgene negligibly affected the sleep phenotypes in Tdc2>TrpA1 flies (supplemental Figure S2). It is thus likely that non-cholinergic expression of the ChAT-Gal80 transgene may block Tdc2>TrpA1 expression (Kayser et al., 2015). Although previous studies have validated the effectiveness of our VGlut RNAi transgene (Aguilar et al., 2017;Takagi et al., 2017), we cannot exclude the possibility of offtarget effects in RNAi experiments. Nonetheless, heterozygosity of the glial glutamate transporter genderblind (gb) (Augustin et al., 2007) rescued short nighttime sleep in Tdc2>TrpA1 flies (Figure 1(E)). We reason that the loss of gb function leads to low glutamate levels at the synapse (Augustin et al., 2007), limiting Tdc2>TrpA1 effects on sleep. Finally, oral administration of the ionotropic glutamate receptor N-methyl-D-aspartate receptor (NMDAR) antagonist rescued short nighttime sleep in Tdc2>TrpA1 flies while negligibly affecting control sleep (Figure 1(F)). Together, these results support that TDC2 neurons implicate glutamate transmission in their sleep suppression (Zimmerman et al., 2017).
Circadian clocks and light collaboratively control sleepregulatory effects of TDC2 neurons Nighttime sleep-specific effects of TDC2 neurons led us to hypothesize that their sleep regulation may involve circadian clock-or light-dependent mechanisms. To address this question, we first examined if the wake-promoting effects of TDC2 neurons would disappear in the absence of functional circadian clocks. TDC2 neuron-specific expression of the voltage-gated sodium channel NaChBac (Tdc2>NaChBac) has been shown to induce short nighttime sleep (Crocker & Sehgal, 2008; Figure 2(A)). We note, however, that daily sleep profiles of Tdc2>NaChBac flies often vary among independent experiments, and the phenotypic instability may involve developmental effects of the constitutive excitation of TDC2 neurons. Nonetheless, we found that Tdc2>NaChBac potently suppressed nighttime sleep even in arrhythmic clock mutants harboring loss-of-function alleles in the circadian clock genes per (per 01 ) (Figure 2(B)) and PDF receptor (Pdfr 5304 ) (Figure 2(C)). In fact, Tdc2> NaChBac rather increased daytime sleep in the clock-less genetic backgrounds compared to heterozygous transgenic controls. These results suggest that TDC2 neurons do not necessarily require circadian clocks for arousal, but their light-dependent effects are somehow exaggerated in those clock mutants. By contrast, a hypomorphic mutation in the leaky sodium channel narrow abdomen (na) partially rescued short nighttime sleep in Tdc2>NaChBac flies (Figure 2(D)), possibly implicating na-dependent clock neuron activity in their sleep regulation (Flourakis et al., 2015;Lear et al., 2005).
To assess the light-dependent effects of TDC2 neurons on sleep, we compared sleep behaviors in either DD or LL cycles after LD entrainment. Tdc2>NaChBac suppressed sleep in both conditions, but there were apparent differences in the sleep phenotypes. Tdc2>NaChBac suppressed the subjective daytime sleep in DD (Figure 3(A)), suggesting that light may mask Tdc2 effects on daytime sleep in LD. Consistent with this idea, Tdc2>NaChBac suppressed sleep more robustly in DD than in LL ( Figure 3). Moreover, the VGlut-Gal80 transgene suppressed the wake-promoting effects of TDC2 neurons in DD but not in LL (Figure 3). These lines of evidence support a model that light specifically silences the wake-promoting effects of glutamatergic TDC2 neurons.
Endogenous circadian rhythms free-run in DD, whereas LL abolishes the clock. Accordingly, we employed T12 cycles (6-h light: 6-h dark) to better distinguish between light and circadian effects on TDC2 neuron-dependent sleep ( Figure  4(A)). For a simple description of our observations below, we divided the two T12 cycles into four 6-h phases (i.e. L1, D1, L2, and D2). Under these conditions, control flies showed high activity at the beginning of each phase, likely reflecting their startling behaviors in response to light transitions (Figure 4(B)). The very striking effects of light on sleep were observed under T12 cycles since light substantially suppressed wild-type sleep during the L2 but not the L1 phase. On the other hand, Tdc2>NaChBac flies displayed distinct sleep phenotypes. TDC2 neurons suppressed D1 sleep in the subjective day more robustly than D2 sleep in the subjective night (Figure 4(B)). We speculate that the desynchrony between light and circadian time (i.e. L2 phase) may transiently titrate the output pathway of Tdc2>NaChBac neurons, only partially suppressing D2 sleep. Alternatively, but not exclusively, 12-h sleep deprivation over the D1-L2 phase may cause a rebound D2 sleep that is resistant to the Tdc2>NaChBac manipulation. Excitation of more broad neurons in the fly brain (121Y > NaChBac) led to nocturnal behaviors in the standard LD cycles comparable to those in Tdc2>NaChBac flies (Figure 4(C) and supplemental Figure  S3). However, light masked the wake-promoting effects of 121Y > NaChBac under T12 cycles regardless of the time of the day, possibly indicating the loss of circadian clock function (see below). These results further support that circadian clocks and light coordinately control the wake-promoting effects of TDC2 neurons.
Since Drosophila senses light via multiple pathways (Helfrich-Forster, 2020; Mazzotta et al., 2020), we asked if the genetic deficit in any specific light-sensing pathway could modify L2 sleep suppression. Light did not suppress L2 sleep in eyeless mutant gl 60j (Moses et al., 1989) and genetically blind NorpA 36 flies (Bloomquist et al., 1988; Figure  5(B)). However, transgenic ablation of the photoreceptor neurons in the eye (GMR > hid) negligibly affected L2 sleep suppression ( Figure 5(B)). We noted that early night sleep in GMR > hid flies was reduced in the standard LD cycle compared to control. Consequently, the 24-h sleep profiles of GMR > hid flies were very similar between LD and T12 cycles, likely indicating their insensitivity to light-induced arousal in the L2 phase. CRYPTOCHROME (CRY) is a cellautonomous, blue-light sensor protein that is expressed in a subset of circadian pacemaker neurons and implicated in the light entrainment of circadian gene expression and behaviors (Emery et al., 2000). However, cry deletion flies displayed significant L2 sleep suppression ( Figure 5(B)), indicating a redundant role of the deep brain light sensor in the lightinduced arousal. Together, these results support that the visual light-sensing suppresses sleep in a circadian clockdependent manner, allowing the misalignment of circadian time and light to arouse diurnal flies only at night.

Non-PDF clock neurons mediate circadian control of light-induced arousal
Individual groups of circadian pacemaker neurons in the adult fly brain have unique neuroanatomical positions for light sensing and play differential roles in shaping circadian behaviors (Helfrich-Forster, 2020;Mazzotta et al., 2020). We thus asked which part of the clock neuron network would be responsible for timely regulating light-induced arousal.
expressing clock cells (tim-Gal4 > PER) fully rescued per mutant sleep and suppressed L2 sleep in T12 cycles ( Figure  6(A)). Selective blockade of the transgenic PER expression in tim-expressing PDF neurons (tim-Gal4, Pdf-Gal80 > PER) did not interfere with the rescue of per mutant sleep ( Figure  6(B)). Moreover, PER expression only in PDF-expressing clock neurons (Pdf-Gal4 > PER) negligibly affected per mutant sleep, showing T12-entrained sleep profiles with no L2 sleep suppression (Figure 6(C)). These results suggest that molecular clocks in PDF neurons are neither necessary nor sufficient for the circadian gating of light-induced arousal. PER rescue by other Gal4 drivers failed to suppress L2 sleep in per mutant flies (supplemental Figure S4). We reason that those Gal4 drivers did not induce sufficient levels of transgenic PER expression in a key subset of non-PDF clock neurons for L2 sleep suppression. Alternatively, circadian control of the light-induced arousal may require molecular clocks in the majority of non-PDF clock neurons. Data represent mean ± SEM (n ¼ 15-74 for clock mutants; n ¼ 16-54 for light-sensing mutants). n.s.: not significant; Ã p < 0.05, ÃÃ p < 0.01, ÃÃÃ p < 0.001, ÃÃÃÃ p < 0.0001 to control (w 1118 ) as determined by ART one-way ANOVA with Wilcoxon rank-sum test.

Discussion
Light is a strong environmental cue for animal behaviors, including circadian rhythms and sleep. Diurnal flies display more consolidated sleep at night, whereas genetic excitation of wake-promoting TDC2 neurons suppresses sleep more potently in the absence of light (Crocker & Sehgal, 2008;Kayser et al., 2015). Our data suggest that the light-sensitive arousal from the octopaminergic TDC2 neurons may depend on their glutamate transmission, consistent with a role of the dual transmission in Drosophila aggression behavior (Sherer et al., 2020). By employing T12 cycles, we further demonstrate that circadian clocks gate the lightinduced arousal, suppressing sleep only when animals face the misalignment of external light with their internal circadian time. We reason that the circadian control of light-induced arousal likely acts as a neural mechanism for adaptive advantage to diurnal animals. The brain loci of individual clock neurons are anatomically distinct, making differences in their light sensitivity and neural function. Photoreceptor neurons in the compound eye send signals toward the medulla, while the extraretinal photoreceptor Hofbauer-Buchner (HB) eyelets project into the accessory medulla (AME) (Helfrich-Forster, 2020). The AME harbors dendrites from select circadian pacemaker neurons, differentially transmitting light information from the compound eyes or the HB eyelets to individual groups of clock cells (e.g. l-LNv, s-LNv, 5th s-LNv, ITP-LNd, anterior DN1, and DN3) (Li et al., 2018). For instance, the compound eyes send excitatory signals to l-LNv via cholinergic Figure 6. Non-PDF clock neurons mediate circadian control of light-induced arousal. (A-C) LD (day 2) and T12 (day 5) sleep in per 01 mutants expressing the per transgene (UAS-per16) in select clock neurons (tim-Gal4, all tim-expressing cells; tim-Gal4þPdf-Gal80, tim-positive and Pdf-negative cells; Pdf-Gal4, all Pdf-expressing neurons). The circadian pacemaker neurons corresponding to each combination of Gal4 and Gal80 transgenes were depicted on the left. Data represent mean ± SEM (n ¼ 25-46 for tim-Gal4; n ¼ 19-39 for tim-Gal4þPdf-Gal80; n ¼ 16-32 for Pdf-Gal4). n.s.: not significant; ÃÃ p < 0.01, ÃÃÃÃ p < 0.0001 as determined by ART two-way ANOVA with Wilcoxon rank-sum test.
transmission, whereas the HB eyelets send inhibitory histaminergic and excitatory cholinergic inputs to l-LNv and s-LNv, respectively (Schlichting et al., 2016).
The PDF-expressing s-LNv plays a crucial role in the behavioral response to lights-on (i.e. morning activity) in LD and free-running locomotor rhythms in DD (Grima et al., 2004;Stoleru et al., 2004). On the other hand, light directly activates PDF-expressing l-LNv via the cell-autonomous photoreceptor CRY and promotes arousal at night (Fogle et al., 2011;Parisky et al., 2008;Shang et al., 2008;Sheeba et al., 2008). l-LNv is also post-synaptic to dopaminergic and octopaminergic neurons, mediating their arousal effects in a light-and circadian clock-dependent manner (Shang et al., 2011). Nonetheless, we found that circadian control of light-induced arousal (i.e. L2 sleep suppression in T12 cycles) was evident even in cry deletion mutants. PER expression in PDF neurons was also dispensable for L2 sleep suppression in T12 cycles, indicating that neither light-activated l-LNv nor free-running s-LNv clocks are implicated in this process.
In fact, evidence for light sensitivity or light-dependent function of non-PDF clock neurons is abundant. As a part of the circadian pacemaker neuron network, LNd and DN1 receive PDF signals from s-LNv. However, subsets of non-PDF neurons (i.e. 5th s-LNv, LNd, anterior DN1, and DN3) show light-induced responses independent of PDF neurons (Li et al., 2018). The posterior DN1 (DN1p) expresses the leaky sodium channel NA important for the behavioral response to light (Lear et al., 2005;Zhang, Chung, et al., 2010). Light actually suppresses the behavioral output from DN1p (Chatterjee et al., 2018;, and the DH31-expressing DN1 suppresses nighttime sleep with no overt effects on circadian rhythms (Kunst et al., 2015). DN1p has also been mapped for light-dependent temperature preference (Head et al., 2015). Nonetheless, the lack of DN1p response to light might explain our observation that transgenic PER rescue in per mutant DN1p was insufficient for circadian control of light-induced arousal.
Heterogenous DN1s make contacts with various postsynaptic targets. These include (1) the evening oscillator (i.e. LNd and 5th s-LNv) via the metabotropic glutamate transmission, exhibiting inhibitory effects on the evening locomotor activity (Guo et al., 2016); (2) sleep-promoting tubercular-bulbar (TuBu) neurons that receive visual input and relay both sleep-and wake-promoting signals from DN1 via excitatory and inhibitory transmissions, respectively, to the ellipsoid body ring (EB-R) neurons (Guo et al., 2018;Lamaze et al., 2018); and (3) pars intercerebralis (PI) neurons for wake promotion (Cavanaugh et al., 2014;Guo et al., 2018). We speculate that molecular clocks in the network of non-PDF neurons relay the timing information for integrating light and clocks into sleep regulation, and DN1 might act as a key locus for the intervention.
Wild-type and clock-less flies display well-distinguishable sleep profiles in T12 cycles, providing a sleep environment where circadian control of sleep can be robustly detected. Considering the dedicated roles of individual clock neurons in specific aspects of circadian behaviors and sleep-wake cycles, future studies should further illustrate the hierarchy and plasticity of the non-PDF clock neuron network for adaptive sleep regulation to environmental disturbance.