Regulation of sleep plasticity by a thermo-sensitive circuit in Drosophila

Sleep is a highly conserved and essential behaviour in many species, including the fruit fly Drosophila melanogaster. In the wild, sensory signalling encoding environmental information must be integrated with sleep drive to ensure that sleep is not initiated during detrimental conditions. However, the molecular and circuit mechanisms by which sleep timing is modulated by the environment are unclear. Here we introduce a novel behavioural paradigm to study this issue. We show that in male fruit flies, onset of the daytime siesta is delayed by ambient temperatures above 29 °C. We term this effect Prolonged Morning Wakefulness (PMW). We show that signalling through the TrpA1 thermo-sensor is required for PMW, and that TrpA1 specifically impacts siesta onset, but not night sleep onset, in response to elevated temperatures. We identify two critical TrpA1-expressing circuits and show that both contact DN1p clock neurons, the output of which is also required for PMW. Finally, we identify the circadian blue-light photoreceptor CRYPTOCHROME as a molecular regulator of PMW, and propose a model in which the Drosophila nervous system integrates information encoding temperature, light, and time to dynamically control when sleep is initiated. Our results provide a platform to investigate how environmental inputs co-ordinately regulate sleep plasticity.


Results
Increased ambient temperature alters sleep architecture in Drosophila. To study how sleep architecture dynamically adapts to environmental changes, we measured sleep levels and timing in adult male flies during two consecutive days. On Day 1 of the experimental paradigm, flies were housed at 22 °C, in 12 h light: 12 h dark (LD) conditions. On Day 2, flies were exposed to a range of warm temperatures (27-31 °C) for 24 h, beginning at Zeitgeber Time 0 (ZT0, lights-on). In these and all subsequent experiments, sleep was defined as 5 min of inactivity, as measured by the Drosophila Activity Monitoring (DAM) system, a well-described standard in the field 21 .
Temperature increases resulted in complex changes to the architecture of siesta sleep. Shifting male flies from 22 °C to either 27 °C or 29 °C prolonged siesta sleep towards lights-off (ZT12), yielding a net increase in siesta sleep at 27 °C and 29 °C compared to 22 °C (Fig. 1a,b,f). This may be due to a delay in the initiation of locomotor increases before lights-off (evening anticipation; Fig. S1), an output of the circadian clock previously shown to be temperature-sensitive 22,23 . At 27 °C, siesta onset was slightly advanced relative to 22 °C (Fig. 1a,d), quantified as the change in latency to initiate sleep between cold and hot days (Δ Latency). However, further increases in temperature shifted siesta onset to later time periods. In particular, at ≥ 30 °C we observed a robust delay in siesta onset that was not observed at 29 °C (Fig. 1b-d), contributing to a net reduction in siesta sleep at ≥ 30 °C Scientific RepoRts | 7:40304 | DOI: 10.1038/srep40304 (Fig. 1f). With respect to nighttime sleep, we found that heightened ambient temperatures induced a delay in night sleep onset (Fig. 1a-c,e). In addition, we observed a roughly linear decrease in night sleep levels in response to increasing temperature levels (Fig. 1g). Thus, temperature increases differentially affect the onset of day versus night sleep, with siesta sleep onset both advanced and delayed by temperature increases within a relatively narrow range, and night sleep onset consistently delayed. To our knowledge, the response of siesta onset to elevated temperatures has not previously been characterized, and for simplicity, we refer to the temperature-induced delay in siesta onset as PMW -Prolonged Morning Wakefulness.
We also tested for the presence of PMW in adult female Drosophila. At 22 °C, female Drosophila initiate the siesta later during the day compared to males (Fig. S1), and we did not observe PMW in females when shifted from 22 °C to 30 °C (Fig. S1). Thus, PMW is sexually dimorphic at 30 °C and correlates with a siesta onset occurring earlier in males than in females at 22 °C.
Does PMW simply represent an acute avoidance response to rapidly increased ambient temperature? To rule out this possibility, we shifted male flies from 22 °C to 30 °C at ZT12 rather than ZT0 and measured sleep the following morning, after 12 h at elevated temperature (Fig. 2a). Indeed, under these conditions we still observed robust PMW, and the magnitude of PMW was equivalent to that caused by a shift from 22 °C to 30 °C at ZT0 (Fig. 2b,c). Thus, PMW is not solely a reaction to a rapid environmental change, but is a behavioral response linked to high temperatures during the morning.
PMW is GLASS-and CRYPTOCHROME-dependent. Since PMW occurs shortly after lights-on, we tested whether PMW could be modified by mutations that impact light-sensing pathways, the circadian clock, or both. We examined PMW in three photoreceptor mutants where signaling through the compound eye is abolished. norpA P41 is a loss of function allele in the phospholipase C-β -encoding gene norpA, a critical component in the canonical light transduction pathway 24,25 ; GMR-hid flies express the pro-apoptotic gene hid in all photoreceptor cells 26 ; and gl 60j mutants are developmentally blind due to loss of GLASS, a transcription factor required for photoreceptor cell development 27 . Surprisingly, PMW was still observed in norpA P41 and GMR-hid males, yet was suppressed in gl 60j homozygotes (Fig. 2d). GLASS is also required for the development of a subset of clock cells termed DN1p neurons 26,28 , providing a possible explanation for this discrepancy (see below). Interestingly, we also found that PMW was suppressed by loss of CRYPTOCHROME (CRY), a circadian blue-light photoreceptor 29 (Fig. 2d; see Discussion). Finally, to test for a direct role of the clock in gating PMW, we generated a new timeless (tim) null allele by replacing the tim coding sequence with a mini-white + reporter gene using homologous recombination (tim ko ; see Materials and Methods). TIM is an essential component of the negative arm of the circadian transcription-translation feedback loop 30 . We confirmed that clock-driven morning and evening anticipation are lost in tim ko homozygote males (Fig. S2), and that tim ko homozygotes are arrhythmic in constant-dark conditions (data not shown). PMW was reduced in tim ko homozygotes, but not fully suppressed (Fig. S2). These results suggest that light-and temperature information collectively drive PMW, with the light-sensing pathway involving the CRY photoreceptor. Furthermore, the circadian clock may play a modulatory role in this process.
PMW requires the TrpA1 thermo-receptor. What molecular pathways signal sleep-relevant temperature information to regulate siesta onset? The Drosophila genome encodes several thermo-sensory proteins 31 . Of these, TrpA1, a cation-conducting channel active at 30 °C, is required for temperature-induced changes in the phase of morning and evening anticipation 23 . We therefore tested whether TrpA1 also impacted PMW. Indeed, a loss of function mutation in TrpA1 (TrpA1 1 ) suppressed PMW, while PMW was still robustly observed in a paired genetic control ( Fig. 3a-c). These results were confirmed using a previously validated TrpA1-RNAi 32 transgene to knock down TrpA1 expression throughout the Drosophila nervous system using the pan-neuronal driver elav-GAL4. In the TrpA1 knockdown background, PMW was suppressed (Fig. S3), similarly to TrpA1 1 homozygotes ( Fig. 3a-c). In contrast, null or strongly hypomorphic mutations affecting the Gr28b 33 and Pyrexia 34,35 thermo-receptors did not suppress PMW (Fig. S3). From the above data, we conclude that TrpA1 is the critical thermo-sensor that mediates PMW.
As shown previously 23 , loss of TrpA1 also suppressed temperature-induced changes in morning and evening anticipation (Fig. 3a,b), but had no effect on temperature-induced delay of sleep onset during the night (Fig. 3d). Furthermore, sleep loss at 30 °C was still strongly observed in the early-middle of the night in TrpA1 1 mutants (Fig. 3b). Thus, temperature-dependent modulation of day sleep, but not night sleep, appears strongly dependent on TrpA1.
Two populations of TrpA1-expressing neurons are necessary for PMW. We next sought to identify subpopulations of TrpA1-expressing neurons that transduce thermo-sensory information to drive PMW. Recent work has shown that a group of TrpA1-expressing neurons defined by the TrpA1[SH]-GAL4 driver modulates the phase of morning anticipation in response to temperature changes 23 . To test if these neurons also play a role in PMW, we used TrpA1-RNAi to knockdown TrpA1 in TrpA1[SH]-neurons. Indeed, PMW was reduced when TrpA1 expression was inhibited in TrpA1[SH]-neurons ( Fig. 4a,b,d). To test whether additional TrpA1-expressing neurons also influenced PMW we screened several driver lines shown, or predicted, to label TrpA1-positive neurons. From this mini-screen, we found that expression of TrpA1-RNAi using pickpocket-GAL4 (ppk-GAL4), also robustly suppressed PMW (Fig. 4a,c,d). We further found that acute inhibition of TrpA1[SH]-and ppk-neuron output using temperature-sensitive dominant-negative shibire (UAS-shi ts ), which blocks endocytosis of synaptic vesicles at 30 °C but not 22 °C 36 , also suppressed PMW (Fig. 4e). However, neither TrpA1 knockdown nor inhibition of synaptic transmission in TrpA1[SH]-and ppk-neurons suppressed temperature-induced delays in nighttime sleep onset (Fig. S4). Thus, TrpA1 expression in, and neurotransmitter release from, TrpA1[SH]-and ppk-neurons are required for PMW, and these circuits primarily impact daytime, as opposed to nighttime, sleep onset.
Scientific RepoRts | 7:40304 | DOI: 10.1038/srep40304 What are the neuro-anatomical correlates of PMW suppression through blocking TrpA1-signaling? ppk-GAL4 is widely used for labeling TrpA1-expressing sensory class IV multi-dendritic (mdIV) neurons in the larval and adult body wall 37,38 , but is also expressed in additional neurons in the adult legs, wings and antennae (data not shown). Interestingly, comparison of the projection patterns of TrpA1[SH]-and ppk-neurons in the adult brain suggested a potential commonality: both drivers encompass neurons that project to the dorsal posterior protocerebrum (DPP: Fig. 4f, arrows). As part of our mini-screen we also tested eight promoter fragments of the ppk-promoter fused to GAL4 (see Materials and Methods). Of these, the ppk[200871]-GAL4 driver labeled mdIV neurons on the adult body wall and exhibited a similar projection pattern to ppk-GAL4 in the suboesophageal ganglion (SOG) region of the brain (

Dorsal-projecting TrpA1[SH]-and ppk-neurons are distinct cell-types. Since both the TrpA1[SH]-
and ppk-positive populations include neurons that send projections to the DPP and regulate PMW, we wondered whether the TrpA1[SH]-and ppk-GAL4 drivers label a common set of sensory neurons. TrpA1[SH]-GAL4 labels internal thermo-sensory AC neurons, whose axons project to the DPP from cell bodies located close to the antennal lobes 39 . In support of the above premise, we stochastically observed AC cell bodies when examining fluorescently labeled TrpA1[SH]-and ppk-neurons (Fig. S5). Thus, we used an intersectional strategy to provide more definitive evidence for common, or distinct, circuits labeled by TrpA1[SH]-and ppk-GAL4. We drove expression of the GAL4-inhibitory protein GAL80 under control of the ppk-promoter (ppk-GAL80 40 ). In the presence of ppk-GAL80, we observed a strongly penetrant loss of ppk-GAL4 expression using CD4::TdTomato (CD4::TdTom) as a fluorescent reporter (Fig. 5a). We confirmed suppression of ppk-GAL4 by ppk-GAL80 at the behavioral level by driving UAS-shi ts in ppk-GAL4/ppk-GAL80 males. In this background we still observed robust PMW (Fig. 5b), in contrast to the effect of driving UAS-shi ts with ppk-GAL4 in the absence of ppk-GAL80 (Fig. 4e). Thus, ppk-GAL80 robustly suppresses ppk-GAL4 activity. Is the same true for TrpA1[SH]-GAL4? Unlike ppk-GAL4, TrpA1[SH]-GAL4-driven CD4::TdTom fluorescence was clearly observed in the presence of ppk-GAL80 (Fig. 5a), and expression of UAS-shi ts in a TrpA1[SH]-GAL4/ppk-GAL80 background still inhibited PMW at 30 °C (Fig. 5b). These data demonstrate that the critical TrpA1[SH]-and ppk-positive neurons required for PMW are distinct populations.
DN1p clock neurons are required for PMW and contact thermo-sensory neurons. Loss of the circadian photoreceptor CRY suppresses PMW (Fig. 2d). Therefore, we wondered whether subsets of clock   43,44 . Interestingly, the output of DN1p neurons during the morning has also been shown to be temperature-dependent 44 , and the excitability of DN1p neurons peaks around dawn 45 , the time period in which PMW occurs (Fig. 1c). Furthermore, the development of DN1p neurons, but not s-LN v s, is GLASS-dependent 26,28 , and as shown above, loss of GLASS suppresses PMW (Fig. 2d). Therefore, we tested for a direct role for DN1p neurons by acutely inhibiting DN1p synaptic output at 30 °C, accomplished by driving UAS-shi ts with the driver clk4.1M-GAL4 (4.1 M), which labels both CRY-positive and -negative DN1p neurons in the adult brain 44 . When shifted to 30 °C, inhibiting DN1p output suppressed PMW, whereas PMW was intact in control lines (Fig. 6a-c). In contrast, inhibiting DN1p output did not alter the delay in night sleep onset at 30 °C (Fig. S6). Similar expression of UAS-shi ts in the CRY-positive s-LN v s and LN d s using mai179-GAL4 43 did not suppress PMW (Fig. S6). Since blocking classical neurotransmitter release does not inhibit PDF exocytosis 46 , we also tested for PMW in pdf null males (pdf 01 ) 42 . In this background, PMW was present, albeit slightly reduced (Fig. S6). These results suggest that DN1p clock neurons are wake-promoting in the early morning at elevated temperatures and undertake a privileged role within the circadian CRY-positive network in regulating PMW.
DN1p cell bodies are located in the DPP, potentially in close proximity to projections from thermo-sensory ppk-and TrpA1[SH]-neurons. To confirm this, we used orthogonal binary systems to drive distinct fluorophores in both ppk-and DN1p neurons. Indeed, we observed a clear overlap between ppk-and DN1p-projections (Fig. 6d). To study potential connectivity between thermo-sensory and DN1p neurons, we used GRASP 47 to test for physical interactions between ppk-and DN1p neurons, and TrpA1[SH]-and DN1p neurons, using PDF immuno-reactivity of s-LN v axons as a marker for the location of DN1p projections 48 . In both cases, expression of complementary split-GFP fragments in either ppk-and DN1p-neurons (Fig. 6e), or TrpA1[SH]-and DN1p-neurons (Fig. 6f) resulted in GRASP fluorescence. In contrast, we only observed minimal GRASP fluorescence between ppk-and PDF-neurons (Fig. S6). Collectively, these results suggest that DN1p-neurons receive dual input from two distinct populations of thermo-sensory neurons to drive temperature-induced increases in morning arousal.

Discussion
How plasticity of distinct sleep periods is regulated at the molecular and circuit levels is unclear. Here we show that the TrpA1 thermo-sensor imparts temperature-sensitivity to siesta sleep in Drosophila, but modulates night sleep to a more subtle degree. Furthermore, we define a novel circuit linking thermo-sensory cells to clock neurons that, in turn, delay sleep onset in response to elevated temperatures. Modulation of Drosophila sleep by temperature has recently been examined, but only up to an ambient temperature of 29 °C 19,20 . At this temperature, siesta sleep was shown to increase relative to 25 °C in both male and female flies 20 , and our results are consistent with this finding (Fig. 1f and Fig. S1). However, we find that in male flies, this effect is specific to the 27-29 °C range. At higher temperatures that would nonetheless be common during summer months (≥ 30 °C), siesta sleep is reduced. In particular, we observed a clear delay in siesta onset at ≥ 30 °C that we term PMW (Fig. 1c,d). The timing and magnitude of siesta sleep is sexually dimorphic 7 , with male flies initiating sleep earlier in the morning at 22 °C when females are still active. While the causes of this sex-specific sleep pattern are still being elucidated, it is noticeable that females do not show PMW at 30 °C. We suggest that the relative hyperactivity of females in the morning masks the effect of temperature on arousal, and later in the afternoon, circadian and/or homeostatic mechanisms act to initiate sleep, whether at mild or higher ambient temperatures. These results imply that arousal during the early morning is particularly sensitive to temperature increases. The circuits we have identified suggest an explanatory basis for this effect.
We found that TrpA1 acts in two distinct thermo-sensory subpopulations defined by the TrpA1[SH]-and ppk-GAL4 drivers to drive PMW, and that both DPP-projecting TrpA1[SH]-and ppk-neurons make physical contact with DN1p neurons that promote arousal in the early morning at 30 °C (Figs 4 and 6). When ectopically expressed, enhanced synaptic transmission induced by TrpA1 can be detected at 26 °C and is further increased at 29 °C 39 . We hypothesize that DN1p neurons receive weak excitatory drive from DPP-projecting TrpA1[SH]-and ppk-neurons, perhaps due to low TrpA1 expression or intrinsic excitability in each cell-type. In this model, excitatory drive scales with temperature 39 , and simultaneous input from TrpA1[SH]-and ppk-neurons, in combination with strong TrpA1-dependent activation of both circuits, is required to cause robust DN1p firing. This, in turn, prolongs arousal during morning periods.
Our model, combined with prior literature, suggests a mechanism for the relatively specific effect of TrpA1 signalling and DN1p activation on the onset of siesta, rather than night sleep. Under LD conditions, DN1p neurons promote morning anticipation, i.e increased locomotion before lights-on, and this output is reduced at low temperatures 44,49 . Recent work has also shown that thermo-genetic activation of CRY-positive DN1p neurons with a distinct driver (R18H11-GAL4 50 ) induces a PMW-like phenotype (Fig. 4 of ref. 51), further supporting a role for DN1p neurons in this process. Importantly, the intrinsic excitability of DN1p neurons is under circadian control, peaking between ZT20-ZT4 and reaching a minimum between ZT8-16 due to clock-dependent oscillations in the resting membrane potential 45 (RMP). PDF signalling from s-LN v s further enhances DN1p excitability in the late night/early morning 52,53 . Thus, DN1p neurons are 'primed' to receive excitatory input from thermo-sensory neurons during the early morning.
Consistent with clock-and PDF-dependent increases in DN1p excitability during the morning, we observed that loss of both the clock protein TIM, and PDF, reduce the magnitude of PMW (Figs S2 and S6). We further demonstrate a role for the blue-light photoreceptor CRY as an essential molecular regulator of PMW (Fig. 2d). CRY is expressed in s-LN v , LN d and DN1p clock neurons 41,54 , and cry transcription is clock-controlled 55 . Inhibiting neurotransmitter or neuropeptide release from the s-LN v s and LN d s does not phenocopy the effect of loss of CRY on PMW ( Fig. 2 and Fig. S6). In contrast, inhibiting DN1p output fully suppresses PMW (Fig. 6), as observed in cry null males (Fig. 2). Thus, the most parsimonious hypothesis is that CRY is acting in DN1p neurons.
How might CRY influence PMW, and since cry transcription is under clock control, why is PMW not fully suppressed in tim mutants? While CRY undertakes several roles in the Drosophila nervous system 29,[56][57][58][59] , recent work has shown that CRY additionally mediates acute light-dependent increases in clock cell excitability via interaction with the potassium β -subunit Hyperkinetic 60,61 . CRY stability is light-dependent, and thus CRY protein levels increase during the night 55 . We hypothesize that in the early morning, strongly-expressed CRY confers light-dependent excitation to DN1p neurons, enhancing the effect of excitatory drive from TrpA1-expressing neurons. Loss of the negative feedback loop of the circadian clock results in constant low-level transcription of cry 55 . However, CRY protein may still accumulate during the night and promote PMW in the early morning. This may explain why loss of TIM reduces, but does not fully suppress, PMW (Fig. S2). Further experiments are required to test the predictions outlined above, and to identify the critical TrpA1-expressing cells that contact DN1p neurons.
In summary, we propose that DN1p neurons integrate both TrpA1-dependent temperature-and CRY-dependent light-information with clock-driven changes in intrinsic excitability to time sleep onset during the early day. In the wild, signalling from a wide array of sensory modalities must be computed in parallel to match sleep onset with environmental conditions. Our work provides a framework to unravel how multi-sensory processing in the Drosophila nervous system facilitates dynamic control of sleep timing.
Behavioral assays. Individual 2-4 day old males were loaded into glass tubes (Trikinetics) containing 2% agar and 4% sucrose. For all experiments shown in this manuscript, Trikinetics monitors were housed in temperature-and light-controlled incubators (LMS, UK). Light intensity was measured to be between 700-1000 Lux using an environmental monitor (Trikinetics). Locomotor activity was recorded in 1 min bins using the Drosophila Activity Monitoring (DAM) system (Trikinetics). During temperature-shift experiments, flies were left for 22 °C prior to recording for 24 h. Activity counts were subsequently measured for 24 h at 22 °C, then for 24 h at elevated temperature (27 °C, 29-31 °C). The time taken for ambient temperature to increase from 22 °C to 30 °C was approximately 25 min. Sleep was defined as a period of inactivity of at least 5 min 67 . A modified version of a previously described Microsoft Excel script 68 was used to measure all sleep parameters detailed in this article. Siesta sleep onset is defined as the latency of the first sleep bout and siesta offset as the end of the last sleep episode. We note that, using incubators from LMS (UK) in the lab of J.E.C.J, consistent and significant delays in siesta onset were observed in response to increasing ambient temperature from 22 °C to 30 °C in control lines. A delayed onset using Percival DR36VLC8 incubators (IA, USA) was also observed in the lab of K.K, although the magnitude of the effect was marginally smaller. Thus, at 30 °C, differences in incubator model may contribute to slightly altered effect sizes. When shifting flies from 22 °C to 31 °C, substantial and highly significant delays in siesta onset of iso31 flies were observed using both incubator models. Sleep graphs were generated using GraphPad Prism 6.
Immunohistochemistry and confocal microscopy. Adult male Drosophila brains were immuno-stained as described previously 10 . Briefly, brains were fixed in 4% paraformaldehyde for 20 min at RT, and blocked in 5% goat serum for 1 h at RT. Primary antibodies used were as follows: Statistical analysis. Since many of the datasets derived from sleep experiments exhibited a non-normal distribution, the following statistical tests were used. Firstly, for binary analysis of whether temperature increases did or did not alter a given sleep parameter, Wilcoxon signed rank test was used, with experimental medians compared to a theoretical median of zero. When simultaneously comparing multiple genotypes, Kruskal-Wallis tests were used, followed by Dunn's post-hoc tests. All statistical analysis was performed using GraphPad Prism 6.