A subclass of evening cells promotes the switch from arousal to sleep at dusk

SUMMARY Animals exhibit rhythmic patterns of behavior that are shaped by an internal circadian clock and the external environment. While light intensity varies across the day, there are particularly robust differences at twilight (dawn/dusk). These periods are also associated with major changes in behavioral states, such as the transition from arousal to sleep. However, the neural mechanisms by which time and environmental conditions promote these behavioral transitions are poorly defined. Here, we show that the E1 subclass of Drosophila evening clock neurons promotes the transition from arousal to sleep at dusk. We first demonstrate that the cell-autonomous clocks of E2 neurons alone are required to drive and adjust the phase of evening anticipation, the canonical behavior associated with “evening” clock neurons. We next show that conditionally silencing E1 neurons causes a significant delay in sleep onset after dusk. However, rather than simply promoting sleep, activating E1 neurons produces time- and light- dependent effects on behavior. Activation of E1 neurons has no effect early in the day, but then triggers arousal before dusk and induces sleep after dusk. Strikingly, these phenotypes critically depend on the presence of light during the day. Despite their influence on behavior around dusk, in vivo voltage imaging of E1 neurons reveals that their spiking rate does not vary between dawn and dusk. Moreover, E1-specific clock ablation has no effect on arousal or sleep. Thus, we suggest that, rather than specifying “evening” time, E1 neurons act, in concert with other rhythmic neurons, to promote behavioral transitions at dusk.


INTRODUCTION
Daily rhythms of behavior depend on both timing information from internal circadian clocks and predictable changes in the external environment. In particular, the rising and setting of the sun produce dramatic changes in environmental light at dawn and dusk, respectively. Given the ethological importance of these transitions between daytime and nighttime, discrete changes in behavioral state often occur around these times, such as the switch from wakefulness to sleep.
However, while much is known about how environmental cues, such as light, entrain the circadian clock, it remains unclear whether and how the clock interacts with these environmental changes at twilight to produce specific patterns of behavior.
Circadian rhythms are centrally controlled by groups of pacemaker neurons in the brain that rhythmically express core clock genes. 1 In Drosophila, the central clock network is comprised of approximately 75 pacemaker neurons in each brain hemisphere. 2 The clock network promotes robust rhythms of locomotor activity and sleep that demonstrate the ability of a fly to both predict and react to changing environmental conditions. As crepuscular animals, flies exhibit two peaks of activity, one occurring at dawn and one occurring at dusk. These two peaks of activity are thought to be separately controlled by different groups of lateral neurons (LNs) in the clock network: "morning cells" (the PDF-positive small ventral LNs/s-LNvs) and "evening cells" (E cells; the PDF-negative 5 th s-LNv and dorsal LNs/LNds). [3][4][5][6][7] Drosophila exhibit an increase in sleep during midday ("siesta"), but, as dusk approaches, flies slowly increase their activity in anticipation of the light-to-dark transition. After dusk, they react to nightfall by rapidly transitioning from peak activity into sleep. [8][9][10] A subset of dorsal clock neurons (DN1ps) has been shown to contribute to evening anticipatory behavior at the end of the "siesta". 7 However, most work on evening anticipation has focused on the E cells, as it has previously been shown that the evening cells, while the E1 neurons are atypical clock neurons that play a newly defined role in promoting the behavioral transition from wakefulness to sleep at dusk.

E2 neuron clocks drive evening anticipation
Most prior work investigating the function of evening cells has focused on manipulating the core molecular clock within these cells while monitoring changes in rhythmic locomotor behavior.
However, these studies have mostly used genetic driver lines that label multiple subclasses of evening cells. [3][4][5][6]11,[19][20][21] We hypothesized that evening anticipatory behavior is attributable to the E2 neurons, because the broader drivers previously used typically included these cells (E1 + E2 neurons in Mai179-GAL4 22 and MB122B-split GAL4 20 or E2 + E3 neurons in dvpdf-GAL4, 23 Fig.   1A). To test whether E2 neurons are responsible for the previously observed effects on anticipation, we compared core clock manipulations in a combined E1+E2 driver, R78G02-GAL4 24 (Fig. 1B) against an E2-specific driver, R54D11-GAL4 17,25 (Fig. 1C). Using R78G02-GAL4, we ablated core clock rhythms in both E1 and E2 neurons through tissue-specific knockout of the core clock gene timeless (tim) using a previously-validated guide RNAs and assessed daily locomotor behavior in 12:12 light:dark (LD) cycles. 26 As expected, this manipulation significantly reduced evening anticipatory behavior, highlighting the necessity of the molecular clock in evening cells for normal evening anticipatory behavior (Figs. 1D and 1E). In addition, changing the period length of molecular clocks in the evening cells has been shown to alter the phase of activity in the evening. [5][6][7]11,27 Indeed, we also found that shortening the period length of the core clocks in both E1 and E2 neurons using the hyperactive kinase DBT S advanced the phase of . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.28.555147 doi: bioRxiv preprint evening activity (Figs. 1F and 1G), 28 further emphasizing the importance of the molecular clock in the evening cells for proper evening behavior.
Interestingly, knocking out tim in E2 neurons alone using R54D11-GAL4 similarly reduced evening anticipation (Figs. 1H and 1I), phenocopying the effect observed with tim knockout in both E1 and E2 neurons. In addition, overexpression of DBT S using R54D11-GAL4 significantly advanced the phase of evening activity (Figs. 1J and 1K), again reproducing the phenotype seen with DBT S overexpression in both E1 and E2 neurons. Thus, manipulating the molecular clocks of E2 neurons alone is sufficient to produce the effects on evening activity previously described when manipulating multiple subclasses of evening cells.
One model for clock neuron function postulates that the cell-autonomous clock within each clock neuron drives rhythmic electrical activity to drive cyclical behavior. To test whether E2 neurons regulate evening behavior through their electrical activity, we conditionally silenced E2 neurons using the inwardly rectifying K + channel Kir2.1 in combination with the auxin-inducible GAL4 expression system (AGES). 29,30 Unexpectedly, auxin-induced silencing of E2 neurons with Kir2.1 did not impair evening anticipation, compared to controls ( Fig. S1A and S1B). Next, we activated E2 neurons using the heat-sensitive cation channel dTrpA1. 31 Similarly, thermogenetic activation of E2 neurons had no significant effect on evening anticipation, although it did lead to a significant increase in nighttime activity (Figs. S1C-S1E). Thus, these data suggest that E2 neuron control of evening anticipation is clock-dependent, but not activity-dependent.
Interestingly, ablating clock rhythms in E1 neurons using tim CRISPR knockout had no effect on evening anticipation, suggesting that, unlike E2 molecular clocks, E1 molecular clocks are not necessary for evening anticipation (Figs. S2A and S2B). In addition, shortening the period length in E1 neurons using UAS-DBT S had only a modest effect with R15C11-GAL4 and inconsistent effects with Trissin-GAL4 (Figs. S2C and S2D), suggesting that the clocks in E1 neurons play a less important role than E2 neurons in setting the phase of evening activity.
We next used AGES-induced expression of Kir2.1 to assess if conditionally silencing E1 neurons affects circadian locomotor activity. Strikingly, silencing E1 neurons in adult flies causes a significant delay in the rapid offset of activity after dusk at zeitgeber time (ZT) 12 (Fig. 2B).
Flies exhibit long, consolidated bouts of sleep at night after dusk, 8-10 but evening cells have not been previously shown to influence sleep timing. To address the possibility that E1 neurons affect the timing of sleep at night, we additionally measured the timing of the first sleep bout after dusk and found that E1 conditional silencing significantly delays sleep onset after dusk (Fig. 2C).
Because fly clock neurons have previously been described as either "wake promoting" or "sleep promoting" (reviewed in Shafer and Keene, 2021), 34 we suspected that E1 neurons might represent a novel "sleep promoting" cluster of clock neurons. To test whether E1 neurons induce sleep when activated, we used dTrpA1 to conditionally activate E1 neurons throughout the day. Surprisingly, activating E1 neurons produced time-and light-dependent effects on sleep. Instead of solely promoting sleep, E1 neuron activation induces almost no behavioral change in the morning, but a significant decrease in sleep in the afternoon (Figs. 2D and 2E, S2E). Perhaps most strikingly, . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.28.555147 doi: bioRxiv preprint the arousal-promoting effects of activating E1 neurons abruptly ceased at dusk, even though the E1 neurons were still being activated (Fig. S2F). Taken together, these findings suggest that E1 neurons play a multifaceted role in patterning behavior around dusk.
To further support that the observed effects on sleep and arousal were due to manipulating E1 neurons and not off-target neurons present in Trissin-GAL4 or R15C11-GAL4, we tested a split-GAL4 combination that only labels E1 neurons in the central brain, R18H11-AD; R78G02-DBD Fig. S2G). Using E1 split-GAL4 to silence E1 neurons with Kir2.1 significantly delayed sleep onset (Figs. S2H and S2I). Moreover, thermogenetic activation of E1 neurons using E1 split-GAL4 recapitulated the time-and light-dependent effects on sleep seen with Trissin-GAL4 and R15C11-GAL4 (Figs. S2J and S2K). These results using E1 split-GAL4 confirm that the dramatic effects on behavior around dusk result from manipulating only the four E1 LNds.

Dusk switches E1 behavioral output from arousal to sleep
Our 24-hr activation protocol suggested that the light-to-dark transition at dusk plays a key role in determining the behavioral output of the E1 neurons. However, to ensure the striking, lightdependent effect observed with this manipulation was not an artifact of 12-hr dTrpA1 activation before dusk, we activated E1 neurons for 3 hrs before and after dusk. Thermogenetic activation of E1 neurons using this regimen decreased sleep during the light phase and also advanced sleep onset time after lights-off ( Figs. 3A and 3B, S3A). Notably, this latter phenotype is opposite to that seen with silencing E1 neurons, suggesting that E1 neural activity is responsible for timing sleep onset after dusk.
After observing the sleep-advancing effects of E1 activation, we hypothesized that E1 neurons might simply promote arousal when activated in light conditions and sleep when activated . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. in dark conditions. Alternatively, E1 neuron output might instead be sensitive to the transition from light to dark, meaning the sleep-promoting effects of E1 activation in the dark would not occur without a preceding light phase. To distinguish between these hypotheses, we activated E1 neurons for 24 hrs in constant darkness (DD), expecting to see an increase in sleep if our initial hypothesis was correct. Strikingly, E1 neuron activation in DD had no effect on locomotor behavior compared to controls (Figs. 3C and 3D, S3B), suggesting that E1 neuron output is sensitive to the transition from light to dark, rather than simply promoting arousal during the daytime and sleep during the nighttime. Overall, light appears to play two roles in determining the behavioral output of E1 neurons. First, our DD activation data suggest that light is a permissive signal that allows E1 neurons to influence behavior. Second, the data from our 6-hr activation around dusk suggest that the light-to-dark transition switches E1 behavioral output from promoting arousal to inducing sleep.

Light input pathways act in parallel to gate E1-mediated behavior
Because light has a strong gating effect on E1 behavioral output in DD, we hypothesized that a specific light-sensing pathway might control that gating effect. Many clock neurons, including E1 neurons, can sense light in a cell-autonomous manner through the deep brain photoreceptor CRY. 35,36 To test whether light sensation through CRY was gating E1 neuron behavioral output, we performed 24-hr dTrpA1 activation of E1 neurons in a cry null mutant background (cry 02 ). 37 If CRY were necessary for E1 output behavior, one would expect no change in behavior with E1 activation. However, similar to activation of E1 neurons in a wild-type background, activating E1 neurons in the cry 02 background in LD produced an afternoon reduction in sleep (Figs. S3C and   S3D).
. CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.28.555147 doi: bioRxiv preprint Outside of CRY, light can directly and indirectly interact with clock neurons via imageforming and non-image-forming visual pathways (reviewed in Yoshii et al., 2016). 38 Thus, we tested whether light sensation through the visual organs might be required for gating E1 behavioral output. To do this, we performed 24-hr activation of E1 neurons in LD in null mutants for the phospholipase C (PLC) required for visual transduction, norpA (norpA P24 ). 39 Again, E1 neuron activation in the norpA P24 background resulted in decreased sleep in the afternoon (Figs. S3E and S3F), suggesting that light sensation through visual pathways is also not, on its own, necessary for proper E1 behavioral output. Although the norpA P24 allele should cause blindness in the compound eyes, ocelli, and H-B eyelets, some reports have suggested that photoreception in the Rhodopsin 6 (Rh6) expressing H-B eyelets is PLC-independent. 40,41 Because the HB-eyelets form monosynaptic connections with clock network neurons, 18,42,43 we further verified that the HBeyelets were not necessary for gating E1 output behavior by thermogenetically activating E1 neurons for 24 hrs in an rh6 null mutant background (rh6 1 ). 44 The rh6 locus ends within 50 bases of the trissin locus, in which the Trissin-2A-GAL4 is inserted, so we were only able to perform the confirming that the HB-eyelets are not necessary for gating E1 activation behavior. Because neither photoreception through CRY nor the visual system appear to be individually necessary for gating E1 output behavior, we suspect that these systems act in parallel to sense environmental light conditions and allow E1 activation to alter arousal and sleep behavior. Indeed, these systems appear to function in a redundant manner for other circadian behaviors, such as photoentrainment, photoperiod adjustment, and evening peak behavior in LD. 38,45

E1 neuron firing does not vary between dawn and dusk
Long-term Ca 2+ imaging in clock network neurons has repeatedly shown that clusters of clock network neurons have daily rhythms of intracellular Ca 2+ . [11][12][13] Additionally, electrophysiological analyses of multiple clusters of clock network neurons reveal daily changes in neural activity that correlate with the time of day when those cells influence behavior. [46][47][48] Whether the electrical activity of LNd neurons varies across circadian time has not been previously addressed. However, these cells, as a group, do exhibit significantly higher levels of intracellular Ca 2+ at or just before dusk. [11][12][13] Because our silencing and activation experiments suggest that E1 LNds influence behavior around dusk, we hypothesized that their neural activity in vivo would be significantly higher at dusk compared to dawn. To test whether E1 neurons are more active at dusk (ZT12- 13) or dawn (ZT0-1), we measured spiking frequency in E1 LNds in vivo using the voltage indicator Voltron2 (Fig. 4A). 49 E1 neurons have symmetrical contra-and ipsilateral branching in the superior medial protocerebrum (SMP), 16,17 allowing us to record changes in membrane potential and calculate spiking events from all four E1 LNds simultaneously in a single region of interest.
Surprisingly, we did not detect a difference in LNd spiking rate between dawn and dusk (Figs. 4B and 4C), suggesting that changes in E1 activity are not necessary for their temporally restricted behavioral influence around dusk.

E1 neurons rely on a non-cell-autonomous clock to properly time behavioral output
Our E1 silencing and 24-hr activation experiments suggest that E1 behavioral output is temporally restricted to the evening, but our in vivo measurements of E1 activity suggest that their firing rate does not change between dawn and dusk. We next ensured that E1 behavioral output was being temporally gated by the circadian clock by performing thermogenetic activation of E1 neurons for . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made Clock neurons are characterized by high-amplitude, rhythmic expression of core clock genes. 2 Cell-autonomous clocks are thought to largely govern rhythmic output from these neurons with input from other clock neurons or environmental stimuli. 1,2,38,50,51 Thus, we next asked whether per is specifically required in E1 neurons for timing their behavioral output. We performed tissue-specific CRISPR to knockout per 26 in E1 neurons and then thermogenetically activated these cells for 24 hrs. Strikingly, this manipulation did not impair the timing of E1 output, showing no effect on behavior in the morning and a significant decrease in sleep in the afternoon (Figs. 5C and 5D). Using 3 UAS transgenes might dilute GAL4-UAS activity. Thus, to ensure that per tissue-specific CRISPR was effective in the presence of a 3 rd UAS-transgene, we performed PER immunostaining and found PER was largely eliminated in E1 neurons, even with the inclusion of an additional UAS-mCD8::GFP transgene (Fig. 5E). These data suggest that E1 neuron output is temporally restricted by a non-cell-autonomous clock that limits the behavioral influence of E1 neurons to the afternoon and evening.
It is possible that E1 behavioral output is timed by both the cell-autonomous clock and a non-cell-autonomous clock, the latter of which could compensate for the loss of the former. To test whether cell-autonomous clocks in E1 neurons were sufficient for timing E1 output behavior, we repeated the 24-hr E1 thermogenetic activation in the per 01 background while simultaneously rescuing per expression in E1 neurons using UAS-per. 16. 52 We found that restoring per expression in E1 neurons was not sufficient to properly pattern E1 output, as thermogenetic activation of E1 neurons in flies with per expression only in E1 clock neurons resulted in a significant reduction in . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.28.555147 doi: bioRxiv preprint sleep from ZT0-12, similar to that seen in per 01 null animals ( Figs. S4A and S4B). These findings suggest that E1 neurons significantly, if not wholly, rely on a non-cell-autonomous clock to properly time their behavioral output.

DISCUSSION
Since the first experiments almost two decades ago described the ability of evening cells to control evening anticipation in Drosophila, 3,4 multiple lines of evidence have suggested that the evening lateral neurons are made up of distinct subclasses: E1, E2, and E3. [14][15][16][17]53,54 However, to date there has not been a systematic attempt to define the behavioral roles of these subclasses in controlling circadian behavior. Here, our results show that between the CRY/PDFR-expressing E1 and E2 neurons, the cell-autonomous clocks of E2 neurons are primarily responsible for proper evening anticipation, the defining behavior of evening cells, and potently impact the phase of evening activity.
In this study, we also describe a behavioral role for the E1 LNds for the first time. E1 neuron activity appears to be important for flies to properly transition to sleep after dusk, and their behavioral output is strikingly reliant on the light-to-dark transition at dusk. However, we also found that their firing rate is not different between dawn and dusk. Additionally, their cellautonomous clocks are dispensable for properly timing E1 behavioral output, suggesting they depend on timing information from other clock neurons to perform their functions at dusk (Fig.   5F).
Cell-autonomous clocks, but not neural activity, in E2 neurons are required for evening anticipation . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.28.555147 doi: bioRxiv preprint Although we showed that E2 neuron clocks are involved in evening anticipation, their neural activity appeared to be dispensable for this process, as neither silencing nor activation significantly affected evening anticipation. One reason E2 neural activity might be insulated from impacting anticipation is that E2 neurons receive direct synaptic input from the visual system and are strongly depolarized by light. Thus, if E2 neurons signal the evening phase of the day, then classical neurotransmission would be a poor method of communicating that information, as it would be confounded by firing induced by visual input. Instead, we hypothesize that the mechanism by which E2 neurons signal impending dusk is clock-dependent, but not activity-dependent. For example, a potential mechanism could be signaling via the constitutive secretory pathway, as the contents of these vesicles could be under circadian-dependent transcriptional/translational control, while their release could be Ca 2+ independent. 55

E1 neurons control sleep timing in a light-dependent manner
To our knowledge, the E1 LNds represent the first example of a well-defined clock neuron subcluster that promotes both arousal and sleep, in this case in a light-dependent manner. Where does light act to regulate E1-mediated behavior? E1 neurons express CRY and are weakly depolarized by light through E2 neurons. 18 However, we showed that CRY expression was not necessary for E1 activation-induced arousal. Moreover, thermogenetic activation of E1 neurons should more strongly depolarize these cells compared to light stimulation. Thus, we hypothesize that the gating effects of light on E1 neuron output occur downstream of E1 neurons.
Interestingly, we also showed that the visual system is not necessary for gating E1 output behavior. Because neither CRY nor visual input are required for E1-induced arousal, we suggest that parallel light pathways are involved in gating E1 output behavior. Multiple circuits could . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.28.555147 doi: bioRxiv preprint potentially receive external light information through parallel pathways. The ellipsoid body (EB), a neuropil closely associated with navigation and locomotion, contains CRY-expressing ring neurons and receives extensive synaptic input from the visual system. 11,56 Additionally, tuberculobulbar (TuBu) neurons are part of the input pathway to the EB from the visual system and are also postsynaptic to CRY-positive clock neurons. [57][58][59] It is possible that such parallel pathways exist to detect changes in a broad spectrum of light, as dusk changes both the brightness and composition of daylight.

The clocks of E1 neurons are not required for their function at dusk
Clock neurons have traditionally been defined by their strong, rhythmic expression of core clock genes. It is generally assumed that both cell-autonomous expression of these genes and changes in spiking activity are necessary for a clock neuron subtype to function at a particular time of day. 2,12,13,50 While E1 neurons impact behavior at dusk, we find that E1 LNds do not vary their firing rate between dawn and dusk. Although we initially predicted significant differences in E1 firing between these two timepoints, our behavioral experiments are consistent with these physiological results, suggesting that the cell-autonomous clocks of E1 LNds are of little importance for their behavioral output at dusk. Instead, we suspect that the rhythmicity of E1 neuron signaling is generated by a postsynaptic cluster of clock neurons (such as the dorsal neurons). For example, coupled networks of neuropeptidergic signaling are common in central clocks, including the mammalian SCN and Drosophila clock network. 14,60 Thus, the dusk-specific actions of E1 neurons could be encoded by postsynaptic, clock-dependent synthesis of a neuropeptide receptor, such as TrissinR or sNPFR. 61,62 . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.28.555147 doi: bioRxiv preprint

A neural circuit for sensing and reacting to dusk
One model for signaling time in clock neurons, both in the Drosophila clock network and the mammalian SCN, is that subsets of clock neurons exhibit daily rhythms of neural activity with different phase relationships throughout the day. 12,63,64 While this mechanism is undoubtedly important for timekeeping in central clocks, even some neurons in the SCN do not exhibit rhythms in clock gene expression or electrical activity. [65][66][67][68] We propose that E1 LNds illustrate an additional mechanism for timekeeping in the clock network in which some clock neurons signal the occurrence of an important zeitgeber landmark, like dusk. In other words, E1 neurons do not strictly function as "evening" neurons, but instead coordinate behaviors at dusk, a light-dependent time window.
We do not predict that E1 neurons sense and react to dusk on their own. Instead, we suggest that E1 neurons act as clock-accessory neurons that themselves are not rhythmic but still contribute to a larger circuit that together maintains daily, rhythmic behavior. In this case, E1 neurons prime the "dusk circuit" for transitioning from wakefulness to sleep, while other components of the dusk circuit signal the time and light conditions necessary to execute that behavior. Given the presence of functional parallels between the Drosophila clock network and the SCN, it is tempting to speculate that a similar circuit motif for sensing and reacting to twilight could also be present in mammals.

ACKNOWLEDGEMENTS
We thank R. Allada and O. T. Shafer for kindly sharing Drosophila reagents. We also thank the Bloomington Stock Center for fly stocks (supported by NIH grant P40OD018537). We thank the Lavis lab for providing the JaneliaFluor dye for Voltron2 recording. We thank members of . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023.

DECLARATION OF INTERESTS
The authors declare no competing interests.
. CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.28.555147 doi: bioRxiv preprint    . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made   (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.28.555147 doi: bioRxiv preprint . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.28.555147 doi: bioRxiv preprint  . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.28.555147 doi: bioRxiv preprint  rh6 1 (gray, n=45), and UAS-dTrpA1/+; rh6 1 rh6 1 (gray, n=49) flies. Mean ± 95% confidence interval is shown. One-way ANOVA with Bonferroni post hoc test.
. CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.28.555147 doi: bioRxiv preprint   . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.28.555147 doi: bioRxiv preprint . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made  . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.28.555147 doi: bioRxiv preprint

Transgenic fly strains
All GAL4, split-GAL4, and tissue-specific CRISPR lines were obtained from the Bloomington Drosophila stock center. Other flies were obtained as described in the key resources table.

Immunostaining
Flies were collected and anaesthetized on ice between ZT22-2 for optimal PER visualization. and were all diluted at a 1:1000 dilution. After secondary antibody incubation, brains were washed in PBS-T four times over at least 2 hrs in PBS-T at room temperature. Brains were then mounted on glass microscope slides using VectaShield Plus (Vector Labs) and size 1.5 coverslips.
. CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.28.555147 doi: bioRxiv preprint

Confocal microscopy
Confocal images were acquired using an LSM710 microscope and Zen Black software (Zeiss).
Pinhole aperture and slice thickness were adjusted to software recommendations for the given fluorophores and imaging parameters. Images shown are max intensity projections of multiple z slices. Central brain images to show driver specificity were taken using a 25x objective under oil immersion ("Plan-Neofluar" 25x/0.8). Zoomed in images of clock neuron GFP/PER expression were taken using a 63x objective under oil immersion.

Behavioral measurements
All lines used in behavioral experiments were outcrossed at least four times to the iso 31 background (BDSC 5905). Activity and sleep behavior were measured using the Drosophila Activity Monitoring system (Trikinetics) and the 5-minute of inactivity threshold for identifying sleep. 10 Anticipation slope, sleep onset, and sleep amount were calculated using the Rethomics behavioral analysis package in R. 70 Evening anticipation phase was calculated using PHASE. 71 For all behavioral experiments, 2-4 day old male flies were loaded into sucrose locomotor tubes (5% sucrose, 2% agar), with the day of loading and following day discarded to allow for recovery from carbon dioxide anesthetization. For UAS-dTrpA1 experiments, flies were raised at 23°C. Baseline behavioral data were collected at 22°C for these flies before being raised to 28°C for 24 hrs activation or 29°C for 6 hr activation. For DD UAS-dTrpA1 experiments, one day of baseline behavior was recorded in LD (not shown) before releasing the animals into DD for one baseline day, then one day at 28°C. For all other experiments, flies were raised at 25°C and behavioral data were recorded at 25°C.
. CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.28.555147 doi: bioRxiv preprint For conditional expression experiments using AGES, 30 flies were raised on standard cornmeal molasses agar. Adult (2-3 day old) male flies were placed in NAA-containing (2 mM; PhytoTech Labs, N610) sucrose locomotor tubes for behavioral analysis.

In vivo voltage imaging
Heterozygous male flies expressing R18H11-AD; R78G02-DBD>UAS-Voltron2 (0-5 day old) were moved onto standard cornmeal molasses agar containing 0.5 mM all-trans retinal (Millipore-Sigma, R2500) 2-3 day before experiments. On the day of the recording, flies were taken out of their home incubator 2.5-3 hrs before the recording was to take place. Flies were fixed into a 3Dprinted chamber with a chemically etched stainless steel shim bottom using UV-curable adhesive (Loctite AA 3972) on the head and thorax. The legs and proboscis were immobilized with beeswax. The imaging area was exposed by removing the cuticle, muscles, fat bodies, and trachea covering the dorsal part of the brain. The ocelli were kept intact and fixed behind the head against Trehalose dihydrate, 10; glucose, 10; sucrose, 2; NaHCO3, 26; NaH2PO4, 1; CaCl2, 1.5; MgCl2, 4. Following dissection, the chamber was filled with fresh dye-containing saline (1 µM JF552-HaloTag ligand), and the chamber then placed into a dark, humidified box for 1hr. The dye was washed out with three 5 min perfusions with saline at 1-1.5 mL/min. Each perfusion was followed by 10-15 mins of passive washing without perfusion. Widefield fluorescence imaging was performed using a fixed stage microscope (Olympus, BX51WI), a 60x water immersion objective (LUMPlanFl/IR, Olympus), and EMCCD camera (iXon Ultra, ANDOR). Excitation with a 530 nM LED (Luxeon) was controlled using a Mightex light source (BLS-1000-2) and BIOLED I/O . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.28.555147 doi: bioRxiv preprint controller (Mightex, BLS-019). Images were acquired at 849 Hz with 8x8 binning using Nikon imaging software (NIS-Elements).
Voltage imaging data were analyzed in MATLAB. Videos were motion-corrected using NoRMCorre, 72 and ROIs were automatically selected using k-means clustering. After mean pixel intensity from the ROI was extracted for each frame, the trace was corrected for photobleaching by fitting a polynomial curve to the trace and subtracting the curve from the original trace. The trace was then filtered using a second order Butterworth filter with a 0.5 normalized cutoff frequency. Because the E1 neurons exhibited slow, oscillating subthreshold changes as well as spikes, the moving baseline fluorescence intensity was approximated using a 1-8 Hz bandpass filter. Spikes were automatically identified by finding local minima that were below two standard deviations from the estimated baseline fluorescence. These automatically detected spikes were manually inspected and corrected as needed. Spiking frequency was calculated by dividing the total number of identified spikes by the length of the recording.

Quantification and statistical analysis
Statistical analyses were performed using Prism (GraphPad). For behavior experiments, pairwise comparisons were made between each experimental group (GAL4>UAS) and their respective genetic controls (GAL4/+ or UAS/+). Differences between normally distributed data were calculated using one-way ANOVA with Bonferroni post-hoc test to correct for multiple comparisons. Differences in non-normally distributed data were calculated using Kruskal-Wallis with Dunn's post-hoc test to correct for multiple comparisons. Sleep onset time, which was normally distributed in some experiments, but not in others due to the floor effect created by dusk, . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.28.555147 doi: bioRxiv preprint were all treated as though they were not normally distributed. For comparing two groups to each other, such as the spiking frequency in Voltron2 recordings, a two-tailed Student's t test was used.

Data availability
Code used for analyzing behavioral and voltage imaging data available at the following DOI: https://zenodo.org/badge/latestdoi/683218355. . CC-BY 4.0 International license available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint this version posted August 29, 2023. ; https://doi.org/10.1101/2023.08.28.555147 doi: bioRxiv preprint