DBT affects sleep in both circadian and non-circadian neurons

Sleep is a very important behavior observed in almost all animals. Importantly, sleep is subject to both circadian and homeostatic regulation. The circadian rhythm determines the daily alternation of the sleep-wake cycle, while homeostasis mediates the rise and dissipation of sleep pressure during the wake and sleep period. As an important kinase, dbt plays a central role in both circadian rhythms and development. We investigated the sleep patterns of several ethyl methanesulfonate-induced dbt mutants and discuss the possible reasons why different sleep phenotypes were shown in these mutants. In order to reduce DBT in all neurons in which it is expressed, CRISPR-Cas9 was used to produce flies that expressed GAL4 in frame with the dbt gene at its endogenous locus, and knock-down of DBT with this construct produced elevated sleep during the day and reduced sleep at night. Loss of sleep at night is mediated by dbt loss during the sleep/wake cycle in the adult, while the increased sleep during the day is produced by reductions in dbt during development and not by reductions in the adult. Additionally, using targeted RNA interference, we uncovered the contribution of dbt on sleep in different subsets of neurons in which dbt is normally expressed. Reduction of dbt in circadian neurons produced less sleep at night, while lower expression of dbt in noncircadian neurons produced increased sleep during the day. Importantly, independently of the types of neurons where dbt affects sleep, we demonstrate that the PER protein is involved in DBT mediated sleep regulation.


Author summary
Doubletime (dbt) is known as a kinase orthologous to mammalian Casein Kinase I ε (CKIε) and Casein Kinase I δ (CKIδ), which are involved in various biological processes and play an important role in regulation of circadian rhythm. In this study, we first analyzed the role of dbt on sleep in Drosophila, and then mapped its expression pattern and further neuronal mechanisms, in which DBT importantly regulates sleep through PER in both non-clock neurons and clock neurons.

Introduction
Sleep is a conserved behavior that has been observed in a variety of species ranging from insects to mammals [1,2]. During sleep, the organism remains quiescent and has a lower response to outside stimulation. Behavioral and electroencephalogram (EEG) studies have suggested that sleep is controlled by circadian and homeostatic processes [3]. The circadian rhythm process could clearly reflect the changing tendency of sleep and wake activity during the day, and it limits sleep amount in a day [4,5]. As for the homeostatic process, it is produced by the accumulation of sleep pressure in the wake period and the release of this pressure in the following sleep period [1]. Although nearly all animals need sleep, the purpose and function of sleep remain poorly understood. With the development of molecular biology, many aspects of life have been addressed at the gene level, including the mechanisms of sleep. Over recent years, many genes have been shown to be involved in sleep regulation [1,6]. As one of the most important clock genes, DBT is known as a kinase orthologous to mammalian kinases CKIε and CKIδ [7,8] and contributes to many different kinds of functions, including circadian rhythm [7][8][9][10][11][12], planar cell polarity [13,14], programmed cell death [15,16] and growth. Although DBT/CKIε is involved in various biological processes including in the circadian rhythm, its relationship with sleep is still unclear. Intriguingly, a previous report showed that functional consequences of a CKIδ mutation caused familial advanced sleep phase syndrome (FASPS) in mammals [17], suggesting that DBT/CKIε could be involved in sleep regulation. However, it is still not known how DBT/CKIε affects sleep in Drosophila and other organisms. In this study, the sleep phenotype and neuronal mechanism by which DBT affects sleep were analyzed. As a strong effect in dbt mutants was observed, a dbt-GAL4 line was constructed with CRISPR/Cas9 to address the localization of DBT expression, and then we used the GAL4/UAS binary expression approach to drive the expression of dbt RNAi in specific neurons in which DBT is expressed, including the Mushroom Bodies (MB), Fan-Shaped Body (FSB), Ellipsoid Body (EB), Pars Intercerebralis (PI) and clock neurons [18]. Results show that DBT affects sleep in both clock neurons and some parts of the central complex and Mushroom Body, and these effects on sleep are based on PER protein.

Sleep in EMS dbt mutants exhibits altered amounts and phase changes consistent with the circadian period changes
In order to determine if the dbt gene affects sleep we analyzed sleep behavior of two different dbt mutants-dbt S and dbt L , which produce short period and long period circadian rhythm respectively. The dbt S and dbt L are two point-mutations of dbt, with the point mutation resulting in an amino acid change of Pro47Ser in dbt S flies and Met80Ile in dbt L flies [7]. These two mutations were reported to affect stability and circadian rhythmicity of PER by altering PER phosphorylation [8]. In the dbt S mutant, the PER protein has a more rapid phosphorylation [9], and the circadian period of the dbt S mutant in DD is decreased to around 18 hours [8]. By contrast, in the dbt L mutant, the PER protein remains longer than wildtype in DD and the circadian period in DD is increased to around 26.8h [8].
Although different locomotor activities were produced by these two mutants (see Fig 1M-1O), both of them show a sleep decrease during the nighttime. In addition, during the daytime, a decreased sleep amount was seen in dbt S mutants while dbt L males showed an increase ( Fig  1A-1D  As these two mutants have different circadian phase, we can see a clear phase change in the sleep profile of these two mutants which is consistent with their circadian locomotor activity phase (Fig 1A and 1C; advanced in dbt S and delayed in dbt L ). Since the morning and evening peaks of activity in the dbt S mutant are advanced while slightly delayed in the dbt L mutant ( Fig  1P-1R), the normal sleep totals for 12 hours daytime sleep and 12 hours nighttime sleep may not represent the phase of sleep properly in dbt mutants. Therefore, we divided one day into 4 periods of 6 hours and analyzed the sleep parameters in dbt S and dbt L mutants respectively. The results show that dbt S has a significantly lower sleep amount and sleep bout duration in both late day (ZT7-12) and late night (ZT19-24) (Fig 1E and 1F). Importantly, sleep bout duration is reduced in early day (ZT1-6) and early night (ZT13-18), while overall levels of sleep are unchanged (ZT1-6 & ZT13-18 in Fig 1E and 1F), suggesting that the dbt S mutation decreases sleep quality independently of total sleep amounts. Analysis of activity counts while the flies are awake reveals that the reductions of sleep observed during ZT7-12 and ZT19-24 are not due to hyperactive locomotor activity ( Fig 1G). As for the dbt L mutants, total sleep amount is decreased during ZT1-6, ZT13-18 and ZT19-24, and increased from ZT7-12 ( Fig 1I). Importantly, reduced and elevated levels of sleep are associated with reduced sleep consolidation during ZT1-6, ZT13-18 and ZT19-24, and elevated sleep consolidation during ZT7-12 respectively ( Fig 1J). Importantly, analysis of activity counts while awake indicates that reduction or increases in sleeping levels are not due to hyper-or hypoactivity respectively ( Fig 1K).
There is a possibility that the sleep decrease of dbt S mutants is caused by a circadian phase advance, since in dbt S flies the sleep loss during the night happens in the late night from around ZT19-ZT24 (Fig 1A and 1E), and previous work showed that dbt S flies have a short subjective day but a normal subjective night. This could mean that the night period for dbt S flies is advanced by nearly 6 hours in constant darkness [9], and there may be a corresponding phase advance in LD, leading to the daytime sleep decrease after midday, the early sleep increase seen during late day in the profile, and the early sleep decrease seen after midnight.
As for dbt L mutant, the phase of the sleep cycle is delayed, with persistent sleep happening in the late day and sleep loss happening in the early morning and early evening (Fig 1I). While these sleep assays are conducted in a 12hr:12hr LD cycle, nevertheless the dbt S genotype has an advanced nighttime sleep phase which appears to start at ZT6-7, and dbt L flies have a slightly delayed nighttime phase which starts after ZT12.
https://doi.org/10.1371/journal.pgen.1010035.g001 the locomotor activity phase change in dbt mutants. To better understand the effects of dbt on sleep, we decided to decrease the expression of dbt endogenously.

Decreased expression of dbt throughout its multiple expression sites in the brain causes increased daytime sleep and decreased nighttime sleep
To decrease the expression of dbt endogenously, we constructed a dbt-GAL4 line by inserting the T2AGAL4 reading frame fused in frame immediately downstream of the last amino acid in dbt (Fig 2). Then we used the UAS/GAL4 system to express UAS-dbt RNAi with our newly created dbt-GAL4 driver. Not surprisingly, we found that only a few flies with the UAS-dbt RNAi / dbt-GAL4 genotype emerged from around 30-50 pupae, illustrating the key role played by dbt during development. From the emerging progeny, after 3-5 days of sleep evaluation, flies with general down-regulation of dbt exhibited a sleep increase during the daytime and a sleep decrease during the nighttime (Fig 3A and 3B). Interestingly, daytime sleep bout duration is no different while nighttime sleep bout duration is significantly decreased when compared with the 2 controls ( Fig 3C). To exclude the possibility that the sleep increase observed during the daytime in dbt-GAL4/UAS-dbt RNAi flies is caused by lower activity, we calculated the activity counts/awake time. Fig 3D shows that the locomotor activity of dbt-GAL4/UAS-dbt RNAi does not differ significantly from controls during daytime. Importantly, the sleep decrease observed during the night in dbt-GAL4/UAS-dbt RNAi flies is not caused by hyperactivity ( Fig  3D). These results suggest that sleep changes seen in dbt-GAL4/UAS-dbt RNAi flies are not The expression pattern of dbt-GAL4 detected with UAS-mCD8: GFP and co-stained with NC82 in the whole brain. Scale bar = 50um. (F-H) Expression of UAS-mCD8: GFP driven by dbt-GAL4 and stained with anti-GFP (green) and anti-NC82 (red) to localize the expression of dbt. The localization shows that DBT protein is expressed broadly in the mushroom body (upper arrow in F), antennal lobe (lower arrows in F), pars intercerebralis (upper arrow in G), ellipsoid body (lower arrow in G), fan-shaped body (arrow in H). Scale bar = 50um. (I-J) Expression of UAS-mCD8: GFP driven by dbt-GAL4 and stained with anti-GFP (green) and anti-PDF (red) to localize the expression of dbt. The localization shows that DBT protein is also expressed in the lateral neurons (I: l-LNv marked with an arrow; s-LNv are PDF-positive neurons below these) and optic lobes (J). Scale bar = 50um. ns. no significant difference, � p< 0.05, �� p< 0.01 and ��� p< 0.001. Black and grey bars (or lines) respectively represent UAS-/+, -GAL4/+ control flies, while orange bars (or lines) represent the treatment flies. The horizontal bar below each graph presents day (white) and night (black).
https://doi.org/10.1371/journal.pgen.1010035.g003 caused by abnormal locomotor activity, and the overall phase difference in the sleep profile relative to controls suggests a delayed phase for the sleep profile. The delayed phase is most likely due to the long-period circadian phenotype for this genotype (Table 1).
While the DBT protein in the dbt S , dbt L and dbt RNAi /dbt-GAL4 genotypes is expressed in all cells in which dbt is normally expressed, the total extent of dbt expression in the brain of adult flies is still unclear. Previous research showed that dbt transcripts were expressed in both the optic lobes and some central neurons [7,19]. In order to localize the expression of dbt, a UAS-mCD8: GFP reporter was expressed with the dbt-GAL4 driver. Immunostaining with anti GFP, anti NC82 and anti PDF antibodies show that dbt-GAL4 is expressed broadly in the optic lobes ( Fig 3E and 3J), antennal lobe (lower arrow in Fig  3F) and the central brain, including the mushroom bodies (upper arrow in Fig 3F), ellipsoid body (lower arrow in Fig 3G), fan-shaped body (arrow in Fig 3H), and pars intercerebralis (upper arrow in Fig 3G). It is also detected in the large and small LNvs, whose cell bodies express PDF (Fig 3I).

Dbt affects daytime sleep in a manner that is dependent on developmental processes
Dbt is also known as disc overgrown (dco), and besides its function in circadian rhythms, it also involved in some developmental processes, such as the Wnt-TCF signaling pathway and the Hippo signaling pathway [13,20,21], so there is a possibility that the effect of dbt on sleep may depend on its function in developmental processes. To determine if dbt affects sleep through its effects on developmental processes, we constructed a stable line with both tub-GAL80ts and dbt-GAL4, and down-regulated dbt with this line only in the adult or larval stage. In order to inhibit the effect of GAL4 during development, we raised the flies at 18˚C, and analyzed adult sleep at 22˚C for one day and then elevated the temperature to 29˚C to knock down dbt (Fig 4A). Results showed a significant decrease in nighttime sleep after 3 days at 29 degrees accompanied by a lower sleep bout duration ( Fig  4B-4D), but no change during daytime after several days of dbt knock-down (Fig 4A-4C). The reduction of sleep observed during nighttime is not due to hyperactive locomotor activity ( Fig 4E). Consistent results were also obtained in tub-GAL80ts/+; UAS-dbt K/R / dbt-GAL4 flies overexpressing a dominant negative DBT K/R protein [11] and occurred immediately after the elevation to 29˚C (S2 Fig), most likely because overexpression of the dominant negative DBT K/R occurred immediately and did not take as long as the RNAimediated knock-down of DBT. On the other hand, when we raised tub-GAL80ts/UASdicer; UAS-dbt RNAi /dbt-GAL4 flies at 29˚C to down-regulate dbt expression during development, no adult flies emerged (S1 Table). These data indicate that the effect of dbt on daytime sleep is probably due to developmental processes. Importantly, the effect on nighttime sleep is caused by loss of dbt in the adult.

Loss of expression of dbt in clock neurons causes a sleep decrease during the nighttime, while loss of expression in other neurons leads to elevated daytime sleep
DBT is a kinase that plays many different roles in various biological processes [7][8][9][10][11][12][13][14][15][16]. To limit its function in specific tissues, and to examine the circadian effect of dbt on the sleep phenotype, we used the UAS/GAL4 system to express UAS-dbt RNAi with two clock neuron driverstim-GAL4 and per-GAL4. A UAS-dicer was separately included in the tim-GAL4 or per-GAL4 and UAS-dbt RNAi to enhance the effect of UAS-dbt RNAi . Down-regulation of dbt expression with a tim-GAL4 driver causes a sleep decrease during the nighttime in LD (Fig 5A and 5B) but also causes a reduced rhythmicity (22.6%) and a longer period (26.2hrs) in constant darkness (DD) ( Table 1). However, with the per-GAL4 driver, sleep was increased during the daytime with no significant difference during the night time ( Fig 5E and 5F), and these changes were reflected with a change of sleep bout duration (a decrease with tim-GAL4 at night and an increase with per-GAL4 during the day) (Fig 5C and  5G). Note that even though down-regulation of dbt expression with per-GAL4 or tim-GAL4 alters sleep, the locomotor activity of these flies during awake time has no statistically significant difference from controls, meaning that the altered sleep of these flies is not due to the difference between their activities (Fig 5D and 5H).
It is possible that these different results are caused by the different expression patterns of the two GAL4 drivers. To verify expression patterns of the two drivers in the fly brain, we crossed each of them with a UAS-mCD8: GFP reporter, and their expression patterns in the brain were detected after dissecting and immunostaining with anti GFP. After microscopy, we find that tim-GAL4 is clearly expressed mostly in the circadian neurons ( Fig 5I-5L) including the l-LNv and s-LNv (Fig 5J), DNs (Fig 5K), optic lobe (lower arrow in Fig 5I) and LNd (arrow in Fig 5L), while per-GAL4 is expressed widely in the brain (Fig 5M-5R), including in two lateral neurons (LNvs) (arrow in Fig 5N) and dorsal neurons (DNs) (arrow in Fig 5O), ellipsoid body (EB) (upper arrow in Fig 5P and 5Q), antennal lobe (AL) (lower arrow in Fig 5P), medio-lateral antennal lobe tract (lower arrow in Fig 5R) and lateral horn (upper arrow in Fig 5R).
In order to exclude the possibility that dbt may affects sleep through glial cells, in which tim-GAL4 is expressed [22], we used the glial-specific repo-GAL4 driver to reduce the expression of dbt in glial cells and tested the sleep phenotype. Results showed that the sleep amount of UAS-dicer; UAS-dbt RNAi /repo-GAL4 did not have a significant difference with the control flies (S3I and S3J Fig). Thus, according to the previous reports and result in this paper, we find that the sleep difference of UAS-dicer; UAS-dbt RNAi /tim-GAL4 was because of the lower expression of dbt in clock neurons rather than glial cells or any other neurons. Given the different effects that down-regulation of dbt with per-GAL4 and tim-GAL4 have on sleep, dbt affects sleep through both clock neurons and non-clock neurons.
Overexpression of dbt K/R with the tim-GAL4 driver leads to a sleep decrease during the nighttime DbtK38R (dbt K/R ) is another site-specific mutation of dbt; the mutant protein can bind PER in vivo but lacks the protein kinase activity of DBT WT and thereby inhibits the activity of DBT WT by competitive binding to PER as a dominant negative. When dbt K/R is expressed in clock neurons of the fly, a very long period or arrhythmicity is produced [11]. In this study, we find that down-regulation of dbt with the tim-GAL4 driver and RNAi can cause a sleep decrease during the nighttime. In order to verify this conclusion, we used a tim-GAL4 driver to drive the expression of UAS-dbt K/R to determine if the same sleep phenotype is produced. As we expected, one group (group 1) of tim-GAL4/+; UAS-dbt K/R /+ flies show a sleep decrease during the nighttime, while its sleep amount during the daytime is not changed (Fig 6A and  6B), and the sleep bout duration of these flies is reduced during the nighttime (Fig 6C), a phenotype consistent with UAS-dicer/tim-GAL4; UAS-dbt RNAi /+ flies. Additionally, the activity during awake time of tim-GAL4/+; UAS-dbt K/R /+ flies is not changed, which means the sleep reduction of these flies during the nighttime is not due to hyperactivity (Fig 6D).
During 5 days of sleep monitoring, we also found another group of experimental flies (group 2: tim-GAL4/+; UAS-dbt K/R /+) showing a sleep change every single night (a cycle of high sleep one night and low sleep the next) (Fig 6I and 6J), a pattern which is slightly different from the tim-GAL4/+; UAS-dbt K/R /+ group 1 flies with a stable sleep decrease during the nighttime (Fig 6E and 6F). The tim-GAL4/+; UAS-dbt K/R /+ were from the same genotype and experiments, but they produced two different sleep phenotypes categorized into two groups, in which some independent repeats only exhibited one type of sleep phenotype (group 1), other independent repeats only exhibited another sleep phenotype (group 2), and some independent repeats simultaneously exhibited two kinds of phenotypes (i.e., group 1 and group 2). From the sleep profiles of single flies of tim-GAL4/+; UAS-dbt K/R /+, we found that the rate of the tim-GAL4/+; UAS-dbt K/R /+ (group 2) sleep change could range from around 20% to 50% from one night to the next.
As a previous study showed that overexpression of dbt K/R produced either longer period flies or arrhythmic flies, we wondered if its effects on sleep are based on circadian differences. Therefore, we analyzed the locomotor activity of the 2 groups of tim-GAL4/+; UAS-dbt K/R / + flies in LD condition, and their circadian rhythmicity and periods in DD condition. The results show that the activity of group 1 tim-GAL4/+; UAS-dbt K/R /+ flies has a stable higher activity during the nighttime (Fig 6G and 6H), while the activity of group 2 still changes from one night to the next, with a night of high activity followed by a night of low activity (Fig 6K  and 6L). Therefore, the long-period oscillations of sleep are produced by long-period changes in activity levels. As for the detection of circadian rhythm and period of these flies, the results show that both groups of tim-GAL4/+; UAS-dbt K/R /+ have lower circadian rhythmicity and longer periods ( Table 2), but group 2 tim-GAL4/+; UAS-dbt K/R /+ has an even longer circadian period and is more arrhythmic than group 1, potentially leading to a lack of entrainment of sleep to a 24 hr LD. Finally, the expression of a UAS-dbt WT with either the tim-GAL4 or per- GAL4 driver has no effect on the sleep profile (S5 Fig)-a result consistent with its weak effects on circadian rhythms [11].
Then we assessed if this phenotype requires PER to be manifested in tim-GAL4/+; UAS-dbt K/ R /+. We expressed dbt K/R with tim-GAL4 in per 0 background and found that the sleep decrease during the nighttime of all days disappeared, while the sleep amounts were no different from the per 0 /Y; UAS-dbt K/R /+ and per 0 ; timGAL4/+ controls (Fig 6M and 6N). In addition, the variations of phenotype group 2 on nighttime sleep and activity in the per 0 /Y; UAS-dbt K/R /tim-GAL4 flies also disappeared (Fig 6O and 6P). These data indicate that the sleep phenotypes of dbt K/R expression, with a decrease during the nighttime, is dependent on PER protein.

Dbt affects daytime sleep through the central complex and mushroom body neurons in a manner that is also based on Period protein
Previous reports showed that central complex and mushroom body neurons are critical in sleep regulation [2,23,24]. To determine if dbt affects sleep through other sites of expression in Average sleep amount during daytime and nighttime with high expression of dbt K/R from tim-GAL4 driver. n (UAS-dbt K/R /+) = 42, n (tim-GAL4/+) = 52, n (tim-GAL4/+; UAS-dbt K/R /+ group1) = 51. (C) Average sleep bout duration during daytime and nighttime with high expression of dbt K/R from tim-GAL4 driver. n (UAS-dbt K/R /+) = 42, n (tim-GAL4/+) = 52, n (tim-GAL4/+; UAS-dbt K/R /+ group1) = 51. (D) Activity counts/awake time with high expression of dbt K/R from tim-GAL4 driver. n (UAS-dbt K/R /+) = 42, n (tim-GAL4/+) = 52, n (tim-GAL4/+; UAS-dbt K/R /+ group1) = 51. (E) 5 days sleep profiles of flies with high expression of dbt K/R in clock neurons (group1). n (UAS-dbt K/R /+) = 42, n (tim-GAL4/+) = 52, n (tim-GAL4/+; UAS-dbt K/R /+ group1) = 51.  (Fig 7A-7H), fan-shaped body (Fig 7I-7L) or ellipsoid body ( Fig  7M-7P) causes a sleep increase during daytime but does not affect circadian period, in contrast with the effect obtained with tim-GAL4 (Table 1), the data for the other drivers, including 44409, C232, 23E10-GAL4 and dilp2-GAL4, that did not cause strong effects are shown in S3 Fig. In addition, according to the analysis of activity counts/awake time, we find that the increased daytime sleep phenotype, is not caused by abnormal activity in these flies (Fig 7D,  7H, 7L and 7P). The daytime sleep in individual dbt-GAL4/UAS-dbt RNAi flies with both a long circadian rhythm and arrhythmia have also been analyzed respectively, and the results show that the daytime sleep is increased in both of these flies, meaning the sleep effect of dbt during daytime is independent of the circadian rhythm phenotype (S6 Fig).
The localization result of per-GAL4 shows that this per driver is widely expressed in the fly brain, including in clock neurons, EB, AL and the lateral horn (Fig 5M-5R). The previous reports showed that this driver is also expressed in parts of the central complex neurons in addition to clock neurons [22,25], and dbt has a similar expression pattern as this driver (Fig 3E and  3F). In addition, as DBT has a close relationship with PER in the circadian program, which is important in sleep regulation, there is a possibility that dbt affects sleep in both circadian and non-circadian neurons that are both dependent on PER. Therefore, to investigate if dbt effects on sleep through the mushroom body and central complex neurons are dependent on PER protein, we decreased the expression of dbt with 201Y (Fig 7Q and 7R), C309 (Fig 7S and 7T), C205 (Fig 7U and 7V) and C819 (Fig 7W and 7X) in a per 0 mutant background. Without PER, the sleep increase during the daytime could not be detected with decreased expression of dbt in the mushroom body neurons and central complex neurons. Thus, the effect of dbt on sleep is dependent on PER. Although we cannot determine whether PER and DBT are expressed in the same neurons to mediate the daytime increase, at least DBT regulates sleep through PER.
Therefore, decreased expression of dbt in the mushroom body, fan-shaped body and ellipsoid body causes an increased sleep amount during daytime, while decreased expression in circadian neurons causes a sleep decrease during nighttime, thereby confirming the possibility that dbt affects sleep through circadian neurons (decreased sleep at night), mushroom body and central complex neurons (increased sleep during the day, as does per-GAL4 expression).
Furthermore, we knocked down the expression of dbt only in adult flies driven by the drivers of the mushroom body, fan-shaped body and ellipsoid body, in conjunction with the tub-GAL80ts repressor. The results showed that down-regulation of dbt only in the adult stage did not impact sleep during the daytime, indicating that effects of dbt on sleep in these neurons is dependent on development (S7 Fig). As we have mentioned above, decreased expression of dbt with dbt-GAL4 causes a sleep increase during the daytime and a sleep decrease during the nighttime. Decreased expression of dbt in clock neurons causes a sleep decrease during the nighttime while the daytime sleep does not change. Finally, decreased expression of dbt in the mushroom body or central complex neurons only causes a sleep increase during the daytime while the nighttime sleep does not change. Therefore, reduced expression with dbt-GAL4 generates both the effects produced by knock-down in circadian and noncircadian neurons, presumably because expression is reduced in both types of neurons.

Discussion
In circadian rhythms, DBT is an important kinase that can degrade PER during the day, with PER stabilization in a binary complex with TIM at night and eventual movement of the PER/ TIM complex into the nucleus around midnight [8,26]. Due to its multiple functions in the circadian clock, loss of function of dbt may cause different circadian mutations such as short period, long period and arrhythmic flies. There are several reports showing that the human sleep disorder Familial Advanced Sleep Phase Syndrome (FASPS) could be explained as mutations of phosphorylation sites in PER or mutations of CKIε or δ (human orthologues of dbt) [17,27], thereby indicating a function of dbt on clock-dependent sleep regulation. We monitored sleep in two different dbt mutants. Both mutants show a sleep decrease during the nighttime, but the difference is that in the dbt S mutant, the sleep decrease happens in the late evening while in the dbt L mutant, it happens in the early morning and the whole night. The timing of these phenotypes suggests that the phase change of locomotor activity may be the main reason why these different dbt mutants affect sleep differently. As different sleep phenotypes were seen in the two mutants, the sleep phenotype produced by loss of expression for wild type dbt was assessed. There is a report showing that dbt is expressed widely in the optic lobes, clock neurons including in the lateral neurons, and has the same pattern as per transcription in the adult head [7]. In order to investigate the localization of dbt and to down-regulate the expression of dbt endogenously, a dbt-GAL4 line was constructed. Then we used a UAS-mCD8: GFP reporter expressed with dbt-GAL4 to observe the expression sites of dbt. The localization of dbt-GAL4 expression pattern in our study shows that dbt is not only expressed in clock neurons but is also broadly expressed in the mushroom body and central complex neurons. To specifically assess the function of all these neurons in the regulation of sleep by dbt, a UAS-dbt RNAi line was expressed with the dbt-GAL4 driver to decrease the expression of dbt in all of these areas. This knock-down produced two dbt phenotypes-elevated daytime sleep and reduced nighttime sleep.
In order to rule out developmental effects, we added a tub-GAL80ts to inhibit GAL4 mediated expression of dbt RNAi during development and then down-regulated the expression of dbt for 3-5 days after the adults emerged from the pupae. The sleep assay results show that the effect of dbt on daytime sleep is likely to be based on developmental processes, while the sleep reduction during the nighttime is due to the loss of expression of dbt in the adult stage. Thus, we propose that the dbt kinase may affect sleep through different pathways that depend on different groups of neurons.
As it is known that dbt is expressed in clock neurons, the tim-GAL4 reported to be mainly expressed in clock neurons and in some glial cells was used to down-regulate dbt [22]; the flies showed a sleep reduction during the nighttime. As use of a glial GAL4 drivers to down regulate dbt did not cause any significant changes in sleep, the effect of knocking down dbt with tim-GAL4 on sleep is likely due to clock neurons rather than glial cells and central complex neurons. Moreover, the decreased sleep amount during the nighttime is consistent with the sleep phenotype of the dbt mutants. As the RNAi-system has the possibility of off-target effects on other mRNAs, the UAS-dbt K/R line was used here as another dbt loss-of-function methodology. K38 and the surrounding region are conserved in Drosophila dbt and vertebrate CKIε/δ, and the mutation can cause a longer circadian period or arrhythmicity in a dominant negative manner [11]. A previous study shows that with altered kinase activity a mutant of CKIε/δ may cause sleep advance or delay in humans [17,28], which is the same phenotype produced in Drosophila. For example, sleep is advanced in the short period dbt S mutant and delayed in the longer period dbt L flies. Although various sleep phenotypes may happen in dbt mutants, none of them show the unique sleep phenotype that dbt K/R shows. The sleep analysis of UAS-dbt K/R / tim-GAL4 indicates that the dbt K/R mutant can cause either sleep loss during the nighttime or a sleep decrease every other night, and this sleep profile is dependent on PER protein. It is also possible that this alternating sleep profile is caused by a longer period (~30 h) in the UASdbt K/R /tim-GAL4 line-a period that persists even in LD because it cannot entrain to the 24 hr LD cycle. While two different sleep phenotypes were shown with this genotype, the sleep amount during the nighttime is still decreased in both groups in UAS-dbt K/R /tim-GAL4 flies, and this result is consistent with the result for knock-down with UAS-dbt RNAi and the tim-GAL4 driver. These results show that down-regulation of dbt levels or overexpression of a dominant negative dbt mutant in clock neurons can both cause a sleep decrease, which is dependent on PER because the effect is eliminated in a per 0 mutant.
With down-regulation of dbt in clock neurons and dbt-GAL4 neurons, the flies exhibit a sleep decrease during the nighttime. When the per-GAL4 driver (another clock neuron driver) was used to reduce the expression of dbt transcription, the flies showed a sleep increase during the daytime only. The expression patterns of the tim-GAL4 and per-GAL4 drivers have been described [22] as different. The per-GAL4 is not only expressed in clock neurons but also in the central brain. That might be the main reason why the UAS-dbt RNAi /per-GAL4 flies have a higher daytime sleep. However, the expression pattern of the per-GAL4 driver in clock neurons is not as extensive as the tim-GAL4 and dbt-GAL4 drivers. That is, the per-GAL4 used in this study is expressed only in 2-3 cells of the l-LNv ( Fig 5N) and DN1s (Fig 5O), while tim-Gal4 and dbt-Gal4 are broadly expressed in clock cells such as the LNvs (Figs 3I, 3J, and 5I-5L). As per-GAL4 is also expressed in clock neurons, it should have the potential to reduce sleep during the nighttime, but this has not been detected. It is possible that down-regulation of dbt with the per-GAL4 driver is not strong enough to cause a circadian effect, so that sleep reduction during the nighttime may not be observed. In support of this, the per-GAL4 driver did not lengthen circadian period as did tim-GAL4 and dbt-GAL4 during DD (Table 1). These results mean that the circadian clock may not be the only source of sleep change in dbt mutants.
Various studies have shown that the central complex neurons and mushroom body are crucial to the regulation of sleep [2]. For instance, the activation or silencing of the mushroom body may promote or reduce sleep of flies, showing that the mushroom body plays an important role in sleep regulation [24,29]. Additionally, the fan-shaped body and ellipsoid body have been reported to contribute to sleep/wake promotion and homeostasis regulation [23,30]. In order to localize the site of mediation by dbt of the developmental effect on daytime sleep, we down-regulated dbt in several central complex and mushroom body neurons with different GAL4 drivers, and the results indicated that down-regulation of dbt in these neurons resulted in an increased daytime sleep, which is consistent with the effect caused by the per-GAL4 driver. In conclusion, dbt may affect sleep through both circadian neurons and noncircadian neurons. As we have mentioned above, the sleep decrease during the nighttime in UAS-dbt K/ R /tim-GAL4 flies is dependent on PER, and the localization of dbt and per drivers shown in this paper indicates that both of the genes may be expressed in the central complex neurons and some part of the mushroom body neurons. Previous reports also showed that per-GAL4 is expressed broadly in the central nervous system [22,25], and PER is expressed in more neurons after knocking down DBT [8]. Therefore, we determined if dbt effects on daytime sleep are also dependent on PER. To investigate this hypothesis, we decreased the expression of DBT with various GAL4 lines in a per 0 background, with the disappearance of the daytime sleep increase in the treatment flies. This result shows that dbt effects on daytime sleep are also dependent on PER. The requirement of PER for both the DBT-dependent increase in daytime sleep and decrease in nighttime sleep indicates that these effects are likely central to the sleep mechanism rather than indirect effects of DBT on sleep. Which developmental pathway contributes to dbt's effects on sleep regulation, and how do the circadian and noncircadian cells produce different effects on sleep? Additional work will address these issues.

Generation of dbt-GAL4 line with a Crispr-Cas 9 approach
The dbt-GAL4 line was constructed with the Crispr-Cas 9 approach by inserting the T2AGAL4 reading frame fused in frame immediately downstream of the last amino acid in dbt. The donor, which has dbt regions extending from 3R:31055021 to 31058029 with a T2A-GAL4 sequence and a removable marker LoxP-RFP-LoxP downstream of the last amino acid of GAL4, was carried in the pBluescript-SK (+) vector.
The sgRNAs were designed with CRISPR Optimal Target Finder (http://tools.flycrispr. molbio.wisc.edu/targetFinder/) [41]. With the goal of enhanced cleavage of the target site, they were designed as follows: gRNA1 (genomic target sequence + gRNA scaffold sequence): GCAAATAATATTTTATCGTTT+GTTTTAGAGCTAGAAATAGCAAGTTAAAATAAG GCTAGTCCGTTATCAACTTGAAAAAGTGGCACCGAGTCGGTGC gRNA2 (genomic target sequence + gRNA scaffold sequence): GTATCGTTTAGGTTGCGACGC+GTTTTAGAGCTAGAAATAGCAAGTTAAAATAA GGCTAGTCCGTTATCAACTTGAAAAAGTGGCACCGAGTCGGTGC; gRNA3 (genomic target sequence + gRNA scaffold sequence): GTTGAATGTATCCAAGCGGC+GTTTTAGAGCTAGAAATAGCAAGTTAAAATAAG GCTAGTCCGTTATCAACTTGAAAAAGTGGCACCGAGTCGGTGC Cas9 mRNA was produced following the protocol of Basset et al. [42]. The construction of donor, RNA mix injection and screening of transgenic flies were performed by Fungene Biotech (http://www.fungene.tech/). The G0 flies were crossed with w 1118 ; TM3, Ser/TM6, Tb line, and the positive flies were identified by the RFP marker which is expressed in the eyes and eventually removed by crossing with the Cre; D/Sb (line 851 from Bloomington Drosophila Stock Center).

Behavioral assays
3-5 days old flies were placed in 65mm×5mm glass tubes with 0.54% agar, 1.1% yeast, 5.6% corn powder, 6.8% molasses and 0.1% Tegosept fly food. Sleep assays were performed at 25˚C,~60% humidity, 12h light/12h dark (LD) conditions, and for circadian assays, the flies were entrained in LD for 3 days and detected in constant darkness (DD) for 5-7 days. Locomotor activity was detected using the DAM system (Drosophila activity monitoring system, Trikinetics, Waltham, MA), and periods of inactivity lasting 5-min or more were quantified as sleep as previously described. Flies in experiments with tub-GAL80ts were raised and analyzed at either 18˚C, 22˚C or 29˚C as specified in Results. The data of sleep analysis is following the method of Driscoll et al [43] with Matlab and pysolo [44]. Circadian activity assays and analysis were conducted by faasX software.

Statistical analysis
All the sleep profiles and the significance analyses are produced by GraphPad Prism 8.0. All the parameters of statistical analysis are shown in S2-S12 Tables.

Resource ability
Materials availability. Fly lines and antibodies not available commercially can be obtained from one of the three labs involved in this study (Zhao, Price or Dissel).