The dorsal fan-shaped body is a neurochemically heterogeneous sleep-regulating center in Drosophila

To assess the role of 23E10-GAL4 dFB neurons in sleep regulation, we designed a Split-GAL4 screen [6] focused on the dFB. The Split-GAL4 technology separates the functional GAL4 transcription factor into 2 non-functional fragments, a GAL4 DNA-binding domain (DBD) and an activation domain (AD). Expression of the AD and DBD fragments is controlled by different enhancers and the functional GAL4 transcription factor is reconstituted only in the cells in which both fragments are expressed [6]. We obtained 20 different AD lines based on their associated GAL4 line’s expression in the dFB, as observed using the Janelia FlyLight website [21,30]. Since the goal of our screen was to identify the contribution of 23E10-GAL4 dFB neurons in sleep regulation, we designed a targeted approach by combining these individual AD lines to a 23E10-DBD line [30,31], thus creating 20 new Split-GAL4 lines named FBS for Fan-Shaped Body Splits (see S1 Table for a description of these lines). These newly created lines were screened behaviorally for their ability to modulate sleep and anatomically to identify their expression patterns. To activate the cells, each individual Split-GAL4 FBS line was crossed to: (1) a line expressing both the thermogenetic TrpA1 cation channel [32] and an mCD8GFP construct; and (2) the optogenetic CsChrimson cation channel [33].

For the thermogenetic screen, flies were maintained at 22 °C for 2 days before raising the temperature to 31 °C at the beginning of day 3 for a duration of 24 h. Temperature was then lowered to 22 °C on day 4 for recovery (S1A Fig). For each individual fly, we calculated the percentage of total sleep change between activation day (day 3) and baseline day (day 2) (S1A Fig). As a control for this screen, we used an enhancerless AD construct (created in the vector used to make all the AD lines) combined with 23E10-DBD, since it is the common element in all the Split-GAL4 lines analyzed in this screen. As seen in S1B Fig, acute thermogenetic activation of the neurons contained in 4 of the 20 FBS lines led to significant increases in sleep in female flies (FBS42, FBS45, FBS53, and FBS68), when compared with controls. Individual sleep traces for these 4 lines are shown in S1D–S1H Fig. Sleep profiles of the other 16 FBS lines are shown in S2 Fig. Importantly, analysis of activity counts during awake time reveals that increases in sleep are not caused by a reduction in locomotor activity (S3A Fig), ruling out motor deficits or paralysis. On the contrary, 2 of the sleep-inducing lines even showed an increase in waking activity upon neuronal activation. While increases in total sleep are indicative of increased sleep quantity, this measurement does not provide information about sleep quality or sleep depth. Previous work proposed that increased sleep bout duration is an indication of increased sleep depth [34]. To assess whether sleep quality is modulated when activating FBS lines, we analyzed sleep consolidation during the day and night in these flies. As seen in S3B Fig, daytime sleep bout duration is significantly increased upon thermogenetic activation in 7 out of 20 FBS lines (FBS28, FBS42, FBS45, FBS53, FBS68, FBS72, and FBS84). Interestingly, 3 of those 7 lines did not show an increase in total sleep (FBS28, FBS72, and FBS84). During the nighttime, the effect of high temperature on sleep is obvious as most lines, including the control, display a significant decrease in sleep bout duration when raising the temperature to 31 °C (S3C Fig). These data agree with previous studies documenting that changes in temperature modulate sleep architecture and that high temperature disturbs sleep at night [35–38]. Only 4 of the 20 FBS lines (FBS42, FBS45, FBS53, and FBS68) maintain similar nighttime sleep bout durations before and during thermogenetic activation. Importantly, in 3 of these 4 lines, nighttime sleep bout duration is significantly increased at 31 °C when compared to control flies (S3C Fig).

Since sleep in Drosophila is sexually dimorphic [39–41], we systematically assessed male flies in our experiments. As seen in S4A Fig, thermogenetic activation increases total sleep in males in 7 out of the 20 FBS lines, including the 4 lines that increase total sleep in females (FBS42, FBS45, FBS53, FBS68, FBS72, FBS81, and FBS84). These effects on total sleep are not caused by motor deficits or paralysis as waking activity is not reduced in long sleeping male flies (S4B Fig). Daytime sleep bout duration is significantly increased in 6 out of these 7 lines (S4C Fig). When examining nighttime sleep, thermogenetic activation of FBS45 and FBS68 significantly increases bout duration while FBS42, FBS53, FBS81, and FBS84 show no difference between 22 and 31 °C (S4D Fig). Altogether, our thermogenetic approach identified 8 FBS lines that increase total sleep and/or sleep bout duration when activated (S2 Table).

Because of the strong effect that temperature has on sleep, we sought to confirm our thermogenetic findings using an alternative method to manipulate FBS lines. The logic behind our thinking is that it may be difficult to fully describe and characterize the sleep behaviors of our FBS lines when temperature itself has such a profound effect on sleep. Furthermore, a recent study demonstrated that some of the effects of high temperature on sleep are mediated by GABAergic transmission on dFB neurons [38]. We thus undertook an optogenetic screen using CsChrimson [33]. Our optogenetic experimental setup is described in S1A Fig. Notably, for neurons to be activated when expressing CsChrimson, flies need to be fed all trans–retinal and the neurons must be stimulated with a 627 nm LED light source [33]. Thus, optogenetic approaches provide a much better control on the timing and parameters of activation. Our regular optogenetic activation screen (consisting of a pulse cycle of [5 ms on, 95 ms off] × 20 with a 4 s delay between pulse cycles) gave results that are mostly identical to the thermogenetic approach. That is, all 4 sleep-promoting lines identified using TrpA1 also increase total sleep when activated with CsChrimson in retinal-fed female flies (FBS42, FBS45, FBS53, and FBS68) (S1C Fig). Interestingly, optogenetic activation of neurons contained in FBS70 and FBS72 is sleep-promoting in retinal-fed female flies while it was not using TrpA1 (S1B and S1C Fig). Sleep profiles of all FBS lines subjected to optogenetic activation are shown in S5 Fig. Looking at males, optogenetic activation increases total sleep in all 7 sleep-promoting lines identified using TrpA1 (FBS42, FBS45, FBS53, FBS68, FBS72, FBS81, and FBS84), as well as an 8th line (FBS33) (S6A Fig). Importantly, these optogenetic sleep-promoting effects are not the result of locomotor deficits (S6B Fig for males and S7A Fig for females).

Examination of sleep bout duration in retinal-fed flies indicates that most sleep-promoting FBS lines increase sleep consolidation during the day (S6C Fig for males and S7B Fig for females) and during the night (S6D Fig for males and S7C Fig for females). Interestingly, some FBS lines that did not increase total sleep upon optogenetic activation increased daytime sleep bout duration in females (FBS57, FBS60, FBS81, and FBS84) (S7B Fig) and in males (FBS35, FBS64, and FBS87) (S6C Fig). In addition, optogenetic activation of neurons contained within FBS58 increases nighttime sleep bout duration in females (S7C Fig). These data suggest that these lines may also express in sleep-regulating neurons. Importantly, analysis of sleep parameters in vehicle-fed flies demonstrate that the sleep phenotypes we observed are specific to retinal-fed and LED stimulated flies (S8 Fig for females and S9 Fig for males).

A summary of sleep-modulating effects for all FBS lines in both activation protocols is provided in S2 Table. Taken together, our thermogenetic and optogenetic screens revealed that activating neurons contained within 16 FBS lines increases at least 1 sleep parameter. Among these 16 lines, 4 are consistently increasing total sleep and sleep consolidation in females (FBS42, FBS45, FBS53, and FBS68) and 7 in males (FBS42, FBS45, FBS53, FBS68, FBS72, FBS81, and FBS84), using both activation protocols (S2 Table).

Having identified 16 different Split-GAL4 lines that significantly increase at least 1 sleep parameter when thermogenetically or optogenetically activated, we sought to identify the neurons that are contained within these lines. As seen in S1D–S1H and S2 Figs, 19 out of 20 lines express in dFB neurons, with only FBS25 showing no expression at all in the brain and VNC. Since FBS25 expresses in no neurons at all, it is not surprising that no sleep changes were seen using both activation protocols, as FBS25 should behave like a control. The number of dFB neurons contained within different FBS lines ranges from 4 to 27 per brain on average (S1 Table). Most lines show very little additional expression in the brain and there is no consistency in these non-dFB labeled cells between different lines. However, 18 out of the 19 lines that express in dFB neurons also express in cells in the metathoracic ganglion of the VNC. This is not surprising as our previous study showed that the 23E10-GAL4 driver, on which this Split-GAL4 screen is based, expresses in 4 neurons located in the metathoracic ganglion of the VNC, consisting of 2 TPN1 neurons [42] and the 2 VNC-SP cells [27]. Most FBS lines (17 out of 20) express in neurons that have anatomical features similar to VNC-SP neurons in the VNC and in the brain, including very typical processes that we previously named “bowtie” [27]. These “bowtie” processes are located extremely close to the axonal projections of dFB neurons, making their visualization sometimes difficult depending on the strength of dFB projections’ staining. Importantly, all 16 FBS lines that increase at least 1 sleep parameter when activated express in these VNC-SP-like cells, in addition to dFB neurons. Furthermore, most FBS lines (14 out of 20) also express in cells that appear to be TPN1 neurons [42].

Our previous study demonstrated that VNC-SP neurons are cholinergic and that the expression of the Split-GAL4 repressor KZip⁺ [43] under the control of a ChAT-LexA driver effectively blocks the accumulation of reconstituted GAL4 in VNC-SP neurons [27]. Since we observed neurons reminiscent of VNC-SP in all 4 FBS lines that increase sleep when thermogenetically activated in females, we wondered whether VNC-SP neurons are the cells responsible for the sleep increase in these lines. To address this possibility, we expressed TrpA1 and GFP in FBS42, FBS45, FBS53, and FBS68 neurons while simultaneously driving the expression of the KZip⁺ repressor in ChAT-LexA cells. As seen in S10A–S10D Fig, expressing the KZip⁺ repressor in cholinergic cells abolishes GFP expression in the metathoracic ganglion in all FBS>UAS-GFP; ChAT-LexA>LexAop2-KZip⁺ flies. These data demonstrate that VNC-SP neurons are present in the expression pattern of all 4 FBS lines tested. When assessing sleep, expressing the KZip⁺ repressor in ChAT-LexA cells prevents the increase in sleep observed in all FBS>UAS-TrpA1 flies tested (S10E Fig). These data suggest that the neurons responsible for the sleep increase observed when thermogenetically activating FBS42, FBS45, FBS53, and FBS68 cells are the cholinergic VNC-SP neurons. However, since we previously demonstrated that some 23E10-GAL4 dFB neurons are cholinergic [27], it is possible that cholinergic 23E10-GAL4 dFB neurons may also participate in these sleep phenotypes. Based on these data, and our previous work showing that 23E10-GAL4 expresses in VNC-SP cells [27], we believe that all 16 FBS sleep-promoting lines express in VNC-SP neurons, making it impossible for us to assess whether 23E10-GAL4 dFB neurons modulate sleep using these lines.

However, if VNC-SP neurons are present in most FBS lines, why do not we see any sleep changes with one of them (FBS1)? In addition, why are some VNC-SP expressing FBS lines more potent than others at promoting sleep? We hypothesized that these differences may be explained by differences in strength of expression of the reconstituted GAL4 in VNC-SP neurons in FBS lines. To address this, we used a GFP-DD construct which is normally degraded by the proteasome [44]. Degradation is blocked when feeding the flies trimethoprim (TMP) [44] and therefore strength of expression can be assessed. We used 23E10-GAL4 and the VNC-SP Split-GAL4 line previously described [27] as controls to compare expression levels in the VNC-SP neurons of different FBS lines. Flies were maintained on standard food until they reached 5–6 days of age. Next, half of the flies were fed either vehicle control (DMSO) or TMP, to protect GFP-DD from degradation, for 24 h (S11A Fig). Flies were then dissected and GFP levels were measured in the metathoracic ganglion of the VNC. As seen in S11B and S11C Fig, feeding TMP leads to strong increases in GFP signal in VNC-SP neurons when GFP-DD is expressed using 23E10-GAL4, VNC-SP Split, and FBS45. When GFP-DD is expressed in FBS33 cells, we observed a moderate increase in GFP signal intensity. With 2 lines, FBS1 and FBS58, we saw no statistical differences between DMSO-fed and TMP-fed flies. The strength of expression in VNC-SP neurons correlates well with the magnitude of sleep increases seen when activating the different lines (S2 Table). FBS45 is strongly sleep-promoting when thermogenetically and optogenetically activated, while FBS33 only increases total sleep and daytime sleep bout duration in males using optogenetic activation. Finally, FBS58 has a very modest impact on sleep while FBS1 has none. Thus, we conclude that within our FBS lines, the strength of expression in VNC-SP neurons dictates how potently a given line can modulate sleep.

When combining behavioral and anatomical data, we conclude that the numbers of dFB neurons contained within a specific FBS line cannot be predictive of whether the line can strongly promote sleep upon neuronal activation. For example, FBS68 strongly increases all sleep parameters in both sexes and activation protocols while only expressing in 3–6 dFB neurons (S1H Fig and S1 and S2 Tables). Conversely, FBS60 contains 22–27 dFB neurons and only increases daytime sleep bout duration in females using optogenetic activation (S2J Fig and S1 and S2 Tables). The anatomical feature that is predictive of whether an FBS line can modulate sleep or not is the presence/absence of VNC-SP neurons within the pattern of expression. In addition, how strongly a line expresses in VNC-SP neurons dictates how strongly sleep-promoting that line is (S11B and S11C Fig).

In conclusion, our activation screens identified 16 FBS lines, expressing in diverse numbers of dFB neurons, that modulate at least 1 sleep parameter (S2 Table). However, all 16 lines contain the previously identified sleep-promoting VNC-SP neurons. Thus, these lines are not well suited to assess the role of 23E10-GAL4 dFB neurons in sleep.

In our screen, we only identified 2 FBS lines that express in dFB neurons but not in VNC-SP cells (FBS6 and FBS41). No sleep parameters are increased when activating neurons contained within either line. However, both lines only express in a reduced number of dFB neurons (S2 Fig and S1 Table). Thus, these 2 lines are not fully recapitulating the dFB expression pattern of the 23E10-GAL4 driver. Fortunately, we identified an additional FBS line (84C10-AD; 23E10-DBD which we named dFB-Split) that expresses in 18–27 dFB neurons per brain and no other cells in the brain or VNC (Fig 1A, S1 and S2 Movies, and S1 Table). Since we previously reported that 23E10-GAL4 reliably labels 23–30 dFB neurons in our confocal microscopy experiments [27], we conclude that dFB-Split expresses in 78% to 90% of 23E10-GAL4 dFB neurons. Additionally, dFB-Split does not express in the wings, legs, gut, or ovaries (S12 Fig). Though not being a complete replication of all 23E10-GAL4 dFB neurons, we hypothesized that the dFB-Split line is a good tool to assess a role for 23E10-GAL4 dFB neurons in sleep regulation. For clarity, we will refer to the dFB cells contained in the dFB-Split expression pattern as dFB^23E10Ո84C10 neurons.

Fig 1. Activation of dFB^23E10Ո84C10 neurons promotes sleep.

(A) Representative confocal stack of a female 84C10AD; 23E10-DBD (dFB-Split)>UAS-mCD8GFP brain (left panel), VNC (middle panel), and the location where VNC-SP “bowtie” processes are seen in many FBS lines, but not in dFB-Split (right panel). We observed 22.52 ± 0.59 (n = 23) dFB neurons in dFB-Split>UAS-mCD8GFP brains. Green, anti-GFP; magenta, anti-nc82 (neuropile marker). (B) Sleep profile in minutes of sleep per hour for day 2 (LED OFF, blue line) and day 3 (LED ON, red line) for retinal-fed empty control (Empty-AD; 23E10-DBD) and dFB-Split female flies expressing CsChrimson subjected to a 1 Hz optogenetic activation protocol. (C) Box plots of total sleep change in % ((total sleep on day 3-total sleep on day 2/total sleep on day 2) × 100) for female control (Empty-AD; 23E10-DBD) and dFB-Split flies expressing CsChrimson under 1 Hz optogenetic activation (cycles of 5 ms LED ON, 995 ms LED OFF). Two-way ANOVA followed by Sidak’s multiple comparisons revealed no differences between control and dFB-Split. n.s. = not significant. n = 20-29 flies per genotype and condition. (D) Sleep profile in minutes of sleep per hour for day 2 (LED OFF, blue line) and day 3 (LED ON, red line) for retinal-fed empty control (Empty-AD; 23E10-DBD) and dFB-Split female flies expressing CsChrimson subjected to our regular optogenetic activation protocol (5 ms LED ON, 95 ms LED OFF, with a 4 s delay between pulses). (E) Box plots of total sleep change in % ((total sleep on day 3-total sleep on day 2/total sleep on day 2) × 100) for female control (Empty-AD; 23E10-DBD) and dFB-Split flies expressing CsChrimson under regular optogenetic activation. Two-way ANOVA followed by Sidak’s multiple comparisons revealed no differences between control and dFB-Split. n.s. = not significant. n = 59-82 flies per genotype and condition. (F) Sleep profile in minutes of sleep per hour for day 2 (LED OFF, blue line) and day 3 (LED ON, red line) for retinal-fed empty control (Empty-AD; 23E10-DBD) and dFB-Split female flies expressing CsChrimson subjected to a 10 Hz optogenetic activation protocol (cycles of 5 ms LED ON, 95 ms LED OFF). (G) Box plots of total sleep change in % for female control (Empty-AD; 23E10-DBD) and dFB-Split flies expressing CsChrimson under 10 Hz optogenetic activation. Two-way ANOVA followed by Sidak’s multiple comparisons revealed that sleep is significantly increased in dFB-Split>CsChrimson females. ***P n = 22-24 flies per genotype and condition. (H) Sleep profile in minutes of sleep per hour for day 2 (LED OFF, blue line) and day 3 (LED ON, red line) for retinal-fed empty control (Empty-AD; 23E10-DBD) and dFB-Split female flies expressing CsChrimson subjected to a 20 Hz optogenetic activation protocol (cycles of 5 ms LED ON, 45 ms LED OFF). (I) Box plots of total sleep change in % for female control (Empty-AD; 23E10-DBD) and dFB-Split flies expressing CsChrimson under 20 Hz optogenetic activation. Two-way ANOVA followed by Sidak’s multiple comparisons revealed that sleep is significantly increased in dFB-Split>CsChrimson females. **P n = 26-38 flies per genotype and condition. (J) Sleep profile in minutes of sleep per hour for day 2 (LED OFF, blue line) and day 3 (LED ON, red line) for retinal-fed empty control (Empty-AD; 23E10-DBD) and dFB-Split female flies expressing CsChrimson subjected to a 50 Hz optogenetic activation protocol (cycles of 5 ms LED ON, 15 ms LED OFF). (K) Box plots of total sleep change in % for female control (Empty-AD; 23E10-DBD) and dFB-Split flies expressing CsChrimson under 50 Hz optogenetic activation. Two-way ANOVA followed by Sidak’s multiple comparisons revealed that sleep is significantly increased in dFB-Split>CsChrimson females. *P P n = 25-38 flies per genotype and condition. (L) Box plots of total sleep change in % for female control (Empty-AD; 23E10-DBD) and dFB-Split flies expressing CsChrimson under constant optogenetic activation. Two-way ANOVA followed by Sidak’s multiple comparisons revealed that sleep is significantly increased in dFB-Split>CsChrimson females. n.s. = not significant, ****P n = 17-27 flies per genotype and condition. (M) Box plots of total sleep change in % for female control (Empty-AD; 23E10-DBD) and dFB-Split flies expressing CsChrimson under 20 Hz optogenetic activation obtained with the DAM5H multibeam system. Two-way ANOVA followed by Sidak’s multiple comparisons revealed that sleep is significantly increased in dFB-Split>CsChrimson females. n.s. = not significant, **P n = 12-15 flies per genotype and condition. (N) Box plots of total sleep change in % for female control (Empty-AD; 23E10-DBD) and dFB-Split flies expressing CsChrimson under 50 Hz optogenetic activation obtained with the DAM5H multibeam system. Two-way ANOVA followed by Sidak’s multiple comparisons revealed that sleep is significantly increased in dFB-Split>CsChrimson females. *P P n = 35-54 flies per genotype and condition. (O) Setup for combined multibeam and video analysis. (P) Multibeam (left) and video analysis (right) of retinal-fed control and dFB-Split>UAS-CsChrimson female flies. Recordings were performed between ZT3-5 for 10 min of baseline (Bs) followed by 10 min of 20 Hz optogenetic activation (Act). For both multibeam and video analysis, data was analyzed in 1 min bin. Flies could perform the following 3 behaviors, walking, micromovements (in place movements like feeding or grooming) or rest (no movements at all), and percentage of time spent in each behavior over 10 min is shown. If a fly walks during a minute bin, the output for that minute is walking, independently of other behaviors that could be performed during the same minute. If a fly performs a micromovement during a minute bin, but shows no walking, the output for that minute is micromovements. If a fly shows no walking or micromovements for a minute bin, the output for that minute is resting. For video analysis, behaviors were manually scored. Two-way repeated measures ANOVA followed by Sidak’s multiple comparisons test found that LED activated retinal-fed dFB-Split>UAS-CsChrimson flies spend less time walking and more time resting than on baseline. No changes in micromovements were observed. No changes were seen in control flies. Similar results were obtained with multibeam and video analysis. n.s. = not significant, ** P P n = 11-12 flies for each genotype. (Q) Video analysis of retinal-fed control and dFB-Split>UAS-CsChrimson female flies. Recording was performed at ZT1-2 for 30 min on consecutive days during baseline (OFF) and activation (ON). Behaviors were manually scored, and amount of time spent in each behavior over 30 min is shown. Two-way ANOVA followed by Sidak’s multiple comparisons test found that retinal-fed dFB-Split>UAS-CsChrimson flies with LED ON sleep significantly more than on baseline day (OFF), **** P P P P n = 8 flies for each genotype. (R) Arousal threshold in vehicle-fed and retinal-fed dFB-Split>UAS-CsChrimson female flies. Percentage of flies awakened by a stimulus of increasing strength (20%, 40%, 60%, 80%, or 100% of maximum strength) with and without 627 nm LEDs stimulation. Two-way ANOVA followed by Sidak’s multiple comparisons indicates that activation of dFB^23E10Ո84C10 neurons reduce the responsiveness to the 20% and 40% stimulus strength when compared with non-activated flies. No difference in responsiveness is seen at the strongest stimulus (60%, 80%, and 100%). Two-way ANOVA followed by Sidak’s multiple comparisons indicates that in vehicle-fed flies no difference in responsiveness is seen between LED stimulated and non-stimulated flies. ****P P n = 16-33 flies per genotype and condition. The raw data underlying parts C, E, G, I, K, L, M, N, P, Q, and R can be found in S1 Data. AD, activation domain; DBD, DNA-binding domain; dFB, dorsal fan-shaped body; VNC, ventral nerve cord.

https://doi.org/10.1371/journal.pbio.3003014.g001

Having identified a genetic driver that is dFB-specific and includes the majority of the 23E10-GAL4 dFB neurons, we sought to investigate the contribution of these cells to sleep. First, we employed a 1 Hz optogenetic activation protocol and found no effects on sleep in female flies (Fig 1B and 1C). Next, we used our regular optogenetic protocol (consisting of a pulse cycle of [5 ms on, 95 ms off] × 20 with a 4 s delay between pulse cycles) and again obtained no sleep increases when activating dFB^23E10Ո84C10 neurons in females (Fig 1D and 1E). Note that this activation protocol is sufficient to increase sleep using multiple other drivers (S1C Fig) and the VNC-SP Split-GAL4 [27], indicating that it is effective at activating some neurons. Why then does this protocol not cause an effect in dFB^23E10Ո84C10 neurons? Previous studies have demonstrated that the dFB is under strong dopaminergic inhibition [15,17,20,45] and that dopamine is a key factor regulating whether dFB neurons are active or silent [20]. In addition, it was shown that sleep deprivation increases the activity of dFB neurons, switching them to the active state [16]. We hypothesize that in our experiments, sleep pressure is low, as flies are allowed to sleep normally before activation. Thus, the inhibitory activity of dopaminergic inputs to the dFB should be relatively strong and it is possible that our regular activation protocol may not be sufficient to overcome this inhibition. To test this possibility, we decided to increase the intensity of the optogenetic activation protocol. As seen in Fig 1F and 1G, a 10 Hz activation of dFB^23E10Ո84C10 neurons increases total sleep in females. In addition, it significantly increases daytime sleep bout duration (S13B Fig). These effects are not due to a locomotor deficit as waking activity is unaffected by 10 Hz optogenetic activation in females (S13A Fig). Importantly, this increase in daytime sleep consolidation is not seen in vehicle-fed flies (S13E). We found similar sleep-promoting effects when looking at dFB-Split>CsChrimson females and males subjected to a 20 Hz optogenetic activation (Fig 1H and 1I for females and S14G Fig for males). In addition, a 20 Hz optogenetic activation increases daytime and nighttime sleep bout duration in female and male flies (S14B and S14C Fig for females and S14I and S14J Fig for males). Again, these effects are not due to a locomotor deficit as waking activity is unaffected by a 20 Hz activation (S14A Fig for females and S14H Fig for males).

We next used a 50 Hz activation protocol. As seen in Fig 1J and 1K, total sleep is increased in dFB-Split>CsChrimson females when activated with a 50 Hz protocol. Furthermore, daytime and nighttime sleep bout durations are increased in females using this activation protocol (S15B and S15C Fig). These effects are not the result of abnormal locomotor activity (S15A Fig) and are not seen in vehicle-fed flies (S15E and S15F Fig). Similar behavioral data was obtained when activating dFB^23E10Ո84C10 neurons with a 50 Hz optogenetic protocol in male flies (S15G–S15M Fig). Since our data suggested a direct relationship between the intensity of the optogenetic activation protocol and the capacity of dFB^23E10Ո84C10 neurons to increase sleep, we decided to turn the 627 nm LEDs on constantly. As seen in Fig 1L, constant LED activation of dFB^23E10Ո84C10 neurons significantly increases sleep in females, indicating that once the activation protocol is sufficient (above 10 Hz), dFB^23E10Ո84C10 neurons can reliably promote sleep.

Recent studies have proposed that activation of all 23E10-GAL4 expressing neurons (including dFB^23E10Ո84C10 cells) promotes micromovements, such as grooming, rather than sleep [29,46]. It is important to note that these conclusions were made using either a thermogenetic approach that may not be sufficient (29° instead of 31° for TrpA1 activation) or a very brief (5 min) 1 Hz optogenetic activation protocol. As mentioned above, we showed that a 24 h long 1 Hz optogenetic activation of dFB^23E10Ո84C10 neurons was insufficient to increase sleep (Fig 1B and 1C). To further validate that sufficient optogenetic activation of dFB^23E10Ո84C10 neurons increases sleep, and not any other behavior (feeding, grooming, or in-place micromovements) that could be falsely registered as sleep by the single beam DAM2 system, we employed the more sensitive multibeam activity monitors (DAM5H, Trikinetics). These monitors contain 15 independent infrared beams, separated by 3 mm across the tube length, resulting in flies being continuously monitored by at least 1 beam during the experiment. First, we sought to validate the sensitivity of the multibeam system to detect different behaviors. We loaded 19 wild-type Canton-S flies in 65-mm long glass tubes and monitored behavior for 10 min using the multibeam system and video recording concurrently (S16A Fig). Video analysis revealed that flies always perform one of 7 different behaviors: rest (total immobility), walking (movement), feeding (micromovement), grooming (micromovement), posture change (a small dip of the abdomen, micromovement), proboscis extension (micromovement), and single leg movement (micromovement) (S16B Fig, Movies 1 and 2 available on https://osf.io/64zh9/?view_only=d058f62c75254556a22f11bd28979287). Our analysis of the multibeam data obtained from the 19 flies revealed that when considering all 7 behaviors, the multibeam sensitivity is 88.9% (S16C Fig). When looking at individual behaviors, sensitivity is above 90% for rest, walking, feeding, and grooming (S16C Fig). However, the multibeam performs poorly for detection of posture change, proboscis extension, and single leg movement (S16C Fig). Notably, these 3 behaviors are extremely subtle with only a very minor movement of one body part, and do not resemble the jerky multiple legs movements that have been described for 1 Hz activation of all 23E10-GAL4 neurons [29]. Additionally, the accuracy of the multibeam system is above 90% for walking and micromovements and 78% for rest (S16C Fig).

Interestingly, we observed that posture change, proboscis extension, and single leg movement events are always surrounded by periods of rest (S16D Fig top panels and S16E Fig, bottom panels, Movies 1 and 2 available on https://osf.io/64zh9/?view_only=d058f62c75254556a22f11bd28979287). Previous studies have suggested that such events are associated with sleep [29,47], suggesting that a failure of the multibeam system to detect them may not lead to a dramatic overestimation of sleep, especially considering the low frequency of such behaviors (S16B Fig). If we consider that posture change, proboscis extension, and single leg movement are part of sleep, the accuracy of the multibeam system for correctly detecting rest jumps to 95.9% (S16C Fig).

As seen in Fig 1M and 1N, 20 and 50 Hz optogenetic activations significantly increase sleep in dFB-Split>CsChrimson females as assessed using the DAM5H multibeam system, ruling out the possibility that we misregistered micromovements as sleep. However, to further support our findings, we performed video analysis. First, we coupled video and multibeam analysis of control and dFB-Split>CsChrimson females during 10 min of baseline recording and the subsequent 10 min of optogenetic activation at 20 Hz (Fig 1O). We found that upon optogenetic activation of dFB^23E10Ո84C10 neurons, walking is significantly reduced, and rest (total immobility with no micromovements) is increased (Fig 1P and Movies 3–16 available on https://osf.io/64zh9/?view_only=d058f62c75254556a22f11bd28979287). Importantly, we observed no change in micromovements upon optogenetic activation (Fig 1P and Movies 3–16 available on https://osf.io/64zh9/?view_only=d058f62c75254556a22f11bd28979287). Both multibeam and video analysis produced similar results, reinforcing the validity of the multibeam system to properly register sleep, movements, and micromovements (Fig 1P). Second, we performed video analysis of multiple behaviors before and during constant optogenetic activation of dFB^23E10Ո84C10 neurons and we found that times spent feeding, grooming, and walking are significantly reduced in dFB-Split>CsChrimson flies during activation, while time spent sleeping is significantly increased (Fig 1Q). Thus, based on our multibeam and video analysis, we conclude that either a 20 Hz or a constant optogenetic activation of dFB^23E10Ո84C10 neurons increases sleep, rather than micromovements.

Finally, we investigated whether activation of dFB^23E10Ո84C10 neurons modulate arousal threshold by applying mechanical stimulations of increasing strength to dFB-Split>CsChrimson flies. As seen in Fig 1R, when dFB^23E10Ո84C10 neurons are activated (retinal-fed, LED ON constant), a significantly lower percentage of flies are awakened by the 20% and 40% stimulus strength, compared with non-activated flies (retinal-fed, LED OFF). Importantly, at the higher stimuli strength (60%, 80%, and 100%), there are no differences between activated and non-activated dFB-Split>CsChrimson flies, indicating that these flies can respond to a stimulus of sufficient strength (i.e., these flies are not paralyzed). These data suggest that constant optogenetic activation of dFB^23E10Ո84C10 neurons increases arousal threshold, an indicator of sleep depth.

To further support that activation of dFB^23E10Ո84C10 neurons increases sleep and not micromovements, we sought to assess whether the state induced by activation of dFB^23E10Ո84C10 neurons can support one of sleep’s proposed functions. The relationship between sleep and memory is well documented [13,48,49]. In particular, sleep plays a role in the consolidation of STM into LTM [50]. Employing genetic and pharmacological activation, previous studies have demonstrated that inducing sleep after a courtship memory training protocol sufficient to create STM, but not LTM, could convert that STM into LTM in Drosophila [14,18,19]. Sleep depriving flies while activating sleep-promoting neurons abrogated the formation of LTM, demonstrating that sleep was needed for this effect [14]. More recent studies have shown that post-learning neuronal reactivation of dopaminergic neurons that are involved in memory acquisition is needed for LTM consolidation [18]. Importantly, this reactivation necessitates sleep and the activity of vFB neurons during a narrow time window after learning [18,51]. Based on their data, the authors proposed that vFB and dFB neurons promote sleep in response to different types of experiences and that these neurons underlie different functions of sleep [18]. We wondered whether the sleep induced by activation of dFB^23E10Ո84C10 neurons is capable of converting STM to LTM. To test this, we used a 1 h courtship training protocol that does not create LTM and activated dFB^23E10Ո84C10 neurons optogenetically for 23 h following the end of training. LTM was tested 24 h after the onset of training (Fig 2A). As seen in Fig 2B, sleep is significantly increased when activating dFB^23E10Ո84C10 neurons post-training. As expected, control flies and dFB-Split>CsChrimson flies trained for 1 h, without increasing sleep post-training, show no LTM (Fig 2C). However, activating dFB^23E10Ո84C10 neurons for 23 h after a 1 h courtship training session led to LTM, as indicated by a significantly higher suppression index (SI) (Fig 2C). Individual values of courtship indices for untrained and trained males are presented in S17A Fig. These data demonstrate that activation of dFB^23E10Ո84C10 neurons following training can convert STM to LTM. An alternative hypothesis is that activation of dFB^23E10Ո84C10 neurons promotes LTM consolidation independently of sleep. To rule out this possibility, we sleep-deprived dFB-Split>CsChrimson flies following the 1 h training session while they were optogenetically activated. This manipulation resulted in sleep loss (Fig 2B) and no LTM (Fig 2C). Thus, we conclude that the sleep that is induced by activation of dFB^23E10Ո84C10 neurons can convert STM to LTM, supporting that the behavior observed is indeed sleep.

Fig 2. The sleep induced by dFB^23E10Ո84C10 neurons activation can consolidate LTM.

(A) Schematic of protocol for LTM consolidation. A 1 h training period that only generates STM is followed by optogenetic activation of dFB^23E10Ո84C10 neurons for 23 h post-training (with or without sleep deprivation). Flies were tested 24 h after the onset of training. (B) Post-training sleep change for different genotypes and condition. The ~23 h time period post-training was matched to the equivalent time period on the baseline day for comparison. Unpaired parametric t test for Empty-Split groups. Kruskal–Wallis test of multiple comparisons for dFB-Split groups using the vehicle condition as the control. ****P P n = 46-54 flies per genotype and condition. (C) Courtship LTM shown as the SI (SI = 100 * (1 – (CI Trained/ CI Untrained))) of trained fly groups tested 24 h after the onset of training. Wilcoxon signed-rank test using H_0: SI = 0. Sample size (untrained:trained) from left to right on the graph, n = 54 (27:27), 51 (27:24), 49 (27:22), 51 (26:25), 46 (21:25), and 53 (27:26), respectively. Courtship indices P S1 Data. dFB, dorsal fan-shaped body; LTM, long-term memory; SI, suppression index; STM, short-term memory.

https://doi.org/10.1371/journal.pbio.3003014.g002

After demonstrating that activation of dFB^23E10Ո84C10 neurons increases sleep duration and sleep depth, and that this sleep consolidates LTM, we performed experiments to silence the activity of these neurons chronically by expressing the hyperpolarizing inward rectifying potassium channel Kir2.1 [52]. First, we assessed whether the constitutive expression of Kir2.1 in dFB^23E10Ո84C10 neurons leads to anatomical defects. As seen in S18A and S18B Fig, we observed no gross morphological defects in dFB^23E10Ո84C10 neuron numbers or processes when expressing Kir2.1, compared with controls.

Surprisingly, total sleep is significantly increased when expressing Kir2.1 in dFB^23E10Ո84C10 neurons in female flies (Fig 3A). This effect on sleep is accompanied by an increase of daytime sleep bout duration (Fig 3C) while consolidation at night is unchanged (Fig 3D). These enhancements on sleep are not due to locomotor deficits (Fig 3B). We obtained identical behavioral results when expressing Kir2.1 in dFB^23E10Ո84C10 neurons of male flies (S18C–S18F Fig). These results are somewhat surprising as both chronic silencing and optogenetic activation of dFB^23E10Ո84C10 neurons result in sleep increases. We thus decided to investigate further. First, we repeated these behavioral experiments using the multibeam system. As seen in Fig 3E, chronic hyperpolarization of dFB^23E10Ո84C10 neurons increases total sleep as assessed with DAM5H monitors. Thus, chronic hyperpolarization of dFB^23E10Ո84C10 neurons increases sleep and sleep consolidation without disrupting the gross morphological properties of these cells. To further investigate the effects of silencing dFB^23E10Ո84C10 neurons, we employed an acute silencing approach by expressing Shi^ts1 [53]. Acute silencing (24 h, Fig 3F) of dFB^23E10Ո84C10 neurons has no effect on sleep in females (Fig 3G) or males (Fig 3H). Altogether, our chronic silencing data further demonstrate that dFB^23E10Ո84C10 neurons regulate sleep. However, it is puzzling that both activation and chronic silencing lead to sleep increases, and that acute silencing has no effect at all.

Fig 3. Chronic hyperpolarization of dFB^23E10Ո84C10 neurons promotes sleep and impairs sleep homeostasis.

(A) Box plots of total sleep (in minutes) for control and dFB-Split>Kir2.1 female flies. A one-way ANOVA followed by Tukey’s multiple comparisons revealed that dFB-Split>Kir2.1 female flies sleep significantly more than controls. *P P P n = 40-85 flies per genotype. (B) Box plots of locomotor activity counts per minute awake for flies presented in A. A Kruskal–Wallis ANOVA followed by Dunn’s multiple comparisons revealed no differences between controls and dFB-Split>Kir2.1 female flies. n.s. = not significant. n = 40-85 flies per genotype. (C) Box plots of daytime sleep bout duration (in minutes) for flies presented in A. A Kruskal–Wallis ANOVA followed by Dunn’s multiple comparisons revealed that daytime sleep bout duration is increased in dFB-Split>Kir2.1 female flies. ****P P n = 40-85 flies per genotype. (D) Box plots of nighttime sleep bout duration (in minutes) for flies presented in A. A Kruskal–Wallis ANOVA followed by Dunn’s multiple comparisons revealed no differences between controls and dFB-Split>Kir2.1 female flies. n.s. = not significant, **P n = 40-85 flies per genotype. (E) Box plots of total sleep (in minutes) for control and dFB-Split>Kir2.1 female flies measured with the DAM5H multibeam system. A two-tailed Mann–Whitney U test revealed that dFB-Split>Kir2.1 female flies sleep significantly more than controls. ***P n = 49-61 flies per genotype. (F) Diagram of the experimental assay for acute silencing. Sleep was measured at 22 °C for 2 days to establish baseline sleep profile. Flies were then shifted to 31 °C for 24 h at the start of day 3 to silence the activity of the targeted cells by activating the Shi^ts1 actuator, and then returned to 22 °C on day 4. White bars (L) represent the 12 h of light and black bars (D) represent the 12 h of dark that are oscillating daily. (G) Box plots of total sleep change in % for control (Empty-AD; 23E10-DBD>UAS-Shi^ts1), and dFB-Split>UAS-Shi^ts1 female flies upon thermogenetic silencing. A two-tailed unpaired t test revealed no differences between controls and dFB-Split>UAS-Shi^ts1 female flies. n.s. = not significant. n = 50 flies per genotype. (H) Box plots of total sleep change in % for control (Empty-AD; 23E10-DBD>UAS-Shi^ts1), and dFB-Split>UAS-Shi^ts1 male flies upon thermogenetic silencing. A two-tailed unpaired t test revealed no differences between controls and dFB-Split>UAS-Shi^ts1 male flies. n.s. = not significant. n = 60-62 flies per genotype. (I) Cumulative sleep lost then gained for Empty-AD; 23E10-DBD>UAS-Kir2.1 and dFB-Split>UAS-Kir2.1 female flies during 12 h of mechanical sleep deprivation (D+SD) and 48 h of sleep recovery. (J) Box plots of total sleep recovered in % during 48 h of sleep recovery following 12 h of sleep deprivation at night for Empty-AD; 23E10-DBD>UAS-Kir2.1 and dFB-Split>UAS-Kir2.1 female flies. Two-tailed Mann–Whitney U tests revealed that sleep rebound is significantly decreased at all time points between controls and dFB-Split>UAS-Kir2.1 female flies. However, Wilcoxon signed-rank tests (indicated by ^# on graph) using H_0: % sleep recovered = 0 revealed that dFB-Split>UAS-Kir2.1 female flies have a sleep rebound significantly greater than 0 at time points 4 h, 6 h, 12 h, and 24 h. ***P P ####P ###P #P n = 25-54 flies per genotype. The raw data underlying parts A, B, C, D, E, G, H, I, and J can be found in S1 Data. AD, activation domain; DBD, DNA-binding domain; dFB, dorsal fan-shaped body.

https://doi.org/10.1371/journal.pbio.3003014.g003

Multiple studies have suggested a role for dFB neurons in regulating sleep homeostasis. In particular, sleep deprivation increases the excitability of dFB neurons [16] and augmented sleep pressure switches dFB neurons from an electrically silent to an electrically active state [20]. Furthermore, reducing the excitability of dFB neurons by reducing levels of the Rho-GTPase-activating protein encoded by the crossveinless-c (cv-c) gene leads to a defect in sleep homeostasis [16]. We previously demonstrated that hyperpolarizing all 23E10-GAL4 neurons blocks sleep homeostasis; however, this effect was not due to VNC-SP neurons [27]. Thus, we concluded that there must be 23E10-GAL4 expressing neurons that are not VNC-SP cells that are involved in sleep homeostasis. We hypothesized that dFB^23E10Ո84C10 neurons could be the cells responsible for sleep homeostasis. To test this possibility, we expressed Kir2.1 in dFB^23E10Ո84C10 neurons, subjected flies to 12 h of mechanical sleep deprivation (SD) during the night and monitored recovery sleep for the subsequent 48 h. We observed that control flies show a gradual recovery of lost sleep over the 48 h period (Fig 3I). Hyperpolarizing dFB^23E10Ո84C10 neurons led to an interesting pattern as dFB-Split>Kir2.1 flies show some recovery during the first 4–6 h following SD, but then did not show additional gains during the remaining recovery period (Fig 3I). To further demonstrate our finding, we quantified sleep recovery after 4 h, 6 h, 12 h, 24 h, and 48 h. As seen in Fig 3J, sleep recovery is significantly reduced between controls and dFB-Split>Kir2.1 flies at all time points. However, in dFB-Split>Kir2.1 flies, sleep recovery at 4 h, 6 h, 12 h, and 24 h is significantly different from 0, indicating that these flies show a weaker, but not nonexistent, homeostatic rebound (Fig 3J). These data suggest that hyperpolarizing dFB^23E10Ո84C10 neurons do not fully block sleep homeostasis, especially in the immediate period following sleep deprivation. Interestingly, another work showed that hyperpolarizing all 23E10-GAL4 neurons does not impair early (6 h) sleep homeostasis [28]. In conclusion, our data suggests that there may be multiple groups of neurons involved in sleep homeostasis in the fly brain, with dFB^23E10Ո84C10 neurons being an important, but not the sole, player.

Previous studies have proposed that dFB neurons are GABAergic [54,55]. However, these data were obtained using a driver that is not dFB-specific [27]. In addition, some dFB neurons transcribe both the vesicular glutamate transporter (VGlut) and the vesicular acetylcholine transporter (VAChT) [56]. Furthermore, we showed in our previous work [27] that some 23E10-GAL4 dFB neurons are cholinergic. Taken all together, these data indicate that the neurochemical identity of dFB neurons is uncertain. We thus investigated whether dFB neurons are GABAergic, glutamatergic, or cholinergic by expressing GFP in dFB^23E10Ո84C10 neurons and staining them with antibodies to GABA, choline acetyltransferase (ChAT, the enzyme necessary to produce acetylcholine), and the VGlut. As seen in Fig 4A, we observed no GABA staining in dFB^23E10Ո84C10 neurons, but instead found that a minority of these cells express ChAT only or VGlut only, while more than 50% of them express both neurotransmitters (Fig 4B and 4C). Note that while this manuscript was in revision, a new preprint demonstrated that dFB neurons express VGlut, VAChT, and ChAT [57], in agreement with our findings and previously published data [56]. Our immunohistochemical analysis thus uncovered that dFB^23E10Ո84C10 neurons can be divided in 3 subgroups: neurons that express ChAT and VGlut (ChAT⁺, VGlut⁺ cells), neurons that express only VGlut (ChAT^–, VGlut⁺ cells), and neurons that express only ChAT (ChAT⁺, VGlut^– cells) (Fig 4C, right).

Fig 4. Neurochemical identity of dFB^23E10Ո84C10 neurons.

(A) Representative confocal stack of a female dFB-Split>UAS-mCD8GFP brain stained with GFP and GABA antibodies and focusing on dFB^23E10Ո84C10 cell bodies. Green, anti-GFP; magenta, anti-GABA. (B) Representative confocal stack of a female dFB-Split>UAS-mCD8GFP brain stained with GFP, VGlut and ChAT antibodies and focusing on dFB^23E10Ո84C10 cell bodies. White arrow indicates a VGlut only positive dFB^23E10Ո84C10 neurons. Yellow arrows show ChAT only positive dFB^23E10Ո84C10 neurons. Yellow asterisks indicate dFB^23E10Ո84C10 neurons positive for VGlut and ChAT. Green, anti-GFP; gray, anti-VGlut, magenta, anti-ChAT. (C) Quantification of VGlut only positive, ChAT only positive and VGlut and ChAT positive dFB^23E10Ո84C10 neurons. We observed 22.76 ± 0.58 (n = 13) GFP positive dFB^23E10Ո84C10 neurons per brain, 3.31 ± 0.71 cells were ChAT only positive, 4.69 ± 0.73 cells were VGlut only positive, 13.77 ± 0.93 cells were ChAT and VGlut positive, and 0.85 ± 0.25 cells were negative for both ChAT and VGlut. (D) Representative confocal stack of a female 84C10-AD; Gad1-DBD>UAS-mCD8GFP. Green, anti-GFP; magenta, anti-nc82 (neuropile marker). (E) Representative confocal stack of a female 84C10-AD; ChAT-DBD>UAS-mCD8GFP. Green, anti-GFP; magenta, anti-nc82 (neuropile marker). (F) Representative confocal stack of a female VGlut-AD; 84C10-DBD>UAS-mCD8GFP. Green, anti-GFP; magenta, anti-nc82 (neuropile marker). (G) Quantification of the numbers of dFB neurons contained within the 2 Split-GAL4 lines presented in E and F. We observed 16.00 ± 0.41 (n = 4) dFB cells for 84C10-AD; ChAT-DBD and 22.20 ± 0.49 (n = 5) dFB neurons for VGlut-AD; 84C10-DBD. No other expression was seen in the brain and VNC for these lines. (H) Sleep profile in minutes of sleep per hour for day 2 (LED OFF, blue line) and day 3 (LED ON, red line) for retinal-fed Empty-Split control females expressing CsChrimson subjected to a 50 Hz optogenetic activation protocol (cycles of 5 ms LED ON, 15 ms LED OFF). (I) Sleep profile in minutes of sleep per hour for day 2 (LED OFF, blue line) and day 3 (LED ON, red line) for retinal-fed VGlut-AD; 84C10-DBD>CsChrimson female flies subjected to a 50 Hz optogenetic activation protocol (cycles of 5 ms LED ON, 15 ms LED OFF). (J) Sleep profile in minutes of sleep per hour for day 2 (LED OFF, blue line) and day 3 (LED ON, red line) for retinal-fed 84C10-AD; ChAT-DBD>CsChrimson female flies subjected to a 50 Hz optogenetic activation protocol (cycles of 5 ms LED ON, 15 ms LED OFF). (K) Box plots of total sleep change in % ((total sleep on activation day-total sleep on baseline day/total sleep on baseline day) × 100) for control (Empty-Split>CsChrimson), VGlut-AD; 84C10-DBD>CsChrimson and 84C10-AD; ChAT-DBD>CsChrimson female flies under a 50 Hz optogenetic activation protocol. Two-way ANOVA followed by Sidak’s multiple comparisons revealed that activating 84C10-AD; ChAT-DBD neurons significantly increases sleep. n.s. = not significant, ****P n = 32-47 flies per genotype and condition. (L) Box plots of daytime sleep change in % ((daytime sleep on activation day-daytime sleep on baseline day/daytime sleep on baseline day) × 100) for control (Empty-Split>CsChrimson), VGlut-AD; 84C10-DBD>CsChrimson and 84C10-AD; ChAT-DBD>CsChrimson female flies under a 50 Hz optogenetic activation protocol. Two-way ANOVA followed by Tukey’s multiple comparisons revealed that activating 84C10-AD; ChAT-DBD and VGlut-AD; 84C10-DBD neurons significantly increases daytime sleep and that activating 84C10-AD; ChAT-DBD neurons increase sleep more than activation of VGlut-AD; 84C10-DBD cells. n.s. = not significant, ***P P n = 32-47 flies per genotype and condition. (M) Sleep profile in minutes of sleep per hour for Empty-Split>Shi^ts1 females maintained at 22° (baseline) and 31° (activation). (N) Sleep profile in minutes of sleep per hour for VGlut-AD; 84C10-DBD>Shi^ts1 females maintained at 22° (baseline) and 31° (activation). (O) Sleep profile in minutes of sleep per hour for 84C10-AD; ChAT-DBD>Shi^ts1 females maintained at 22° (baseline) and 31° (activation). (P) Box plots of total sleep change in % ((total sleep on activation day-total sleep on baseline day/total sleep on baseline day) × 100) for control (Empty-Split>Shi^ts1), VGlut-AD; 84C10-DBD>Shi^ts1 and 84C10-AD; ChAT-DBD>Shi^ts1 female flies. A Kruskal–Wallis ANOVA followed by Dunn’s multiple comparisons revealed that acutely silencing VGlut-AD; 84C10-DBD and 84C10-AD; ChAT-DBD neurons significantly decreases sleep. ****P n = 26-60 flies per genotype. (Q) Box plots of locomotor activity counts per minute awake for flies presented in P. Two-way repeated measures ANOVA followed by Sidak’s multiple comparisons revealed that acutely silencing VGlut-AD; 84C10-DBD and 84C10-AD; ChAT-DBD neurons does not lead to hyperactivity. n.s. = not significant, *P P n = 26-60 flies per genotype. (R) Box plots of total sleep recovered in % during 24 h of recovery following 12 h of sleep deprivation at night for Empty-Split>Shi^ts1, VGlut-AD; 84C10-DBD>Shi^ts1 and 84C10-AD; ChAT-DBD>Shi^ts1 female flies. Following sleep deprivation, flies were either maintained at 22° (control) or at 31° (to silence neurons). Kruskal–Wallis test followed by Dunn’s multiple comparisons revealed that acutely silencing 84C10-AD; ChAT-DBD neurons blocks sleep homeostasis. No differences were seen between lines at 22°. n.s. = not significant, **P P n = 28-57 flies per genotype and condition. The raw data underlying parts C, G, K, L, P, Q, and R can be found in S1 Data. AD, activation domain; ChAT, choline acetyltransferase; DBD, DNA-binding domain; dFB, dorsal fan-shaped body; VGlut, vesicular glutamate transporter; VNC, ventral nerve cord.

https://doi.org/10.1371/journal.pbio.3003014.g004

To confirm that 23E10-GAL4 dFB neurons are not GABAergic, we expressed GFP using 23E10-GAL4 and stained brains with an antibody to GABA. As seen in S19A Fig, we observed no GABA staining in 23E10-GAL4 dFB neurons. In addition, both a Gad1-AD; 23E10-DBD and 84C10-AD; Gad1-DBD Split-GAL4 lines show no expression in dFB neurons (Figs 4D and S19B). Thus, we conclude that dFB^23E10Ո84C10 neurons are not GABAergic, contrary to previous reports. To strengthen our findings about the glutamatergic and cholinergic nature of dFB^23E10Ո84C10 neurons, we employed a Split-GAL4 strategy using the 84C10 component of dFB-Split paired with neurotransmitter-specific Split elements. An 84C10-AD; ChAT-DBD line expresses in about 16 dFB neurons (ChAT⁺, VGlut⁺ and ChAT⁺, VGlut^– cells; named dFB^ChATՈ84C10) per brain and no other cells in the brain and VNC (Fig 4E and 4G and S3 and S4 Movies). A VGlut-AD; 84C10-DBD line expresses in 22 dFB cells (ChAT⁺, VGlut⁺ and ChAT^–, VGlut⁺ cells; named dFB^VGlutՈ84C10) per brain and shows no other expression in the brain and VNC (Fig 4F and 4G and S5 and S6 Movies). Note that while this manuscript was in preparation, a preprint reported that dFB neurons are glutamatergic and that optogenetic activation of these cells weakly, but significantly, increases sleep [58]. However, the Split-GAL4 line used in this study, VGlut-AD; 23E10-DBD may not be dFB-specific, making the interpretation of the behavioral data difficult. We have created 2 different VGlut-AD; 23E10-DBD lines using the only 2 VGlut-AD lines known to us. As seen in S19C and S19D Fig and S7 and S8 Movies, both lines express in dFB neurons, but they also express in 8 neurons that are not dFB cells. Note that both independent lines show similar expression in these 8 non-dFB neurons. Thus, our data suggest that VGlut-AD; 23E10-DBD is not a dFB-specific tool and cannot be used to manipulate only dFB neurons. Taken together, our immunostaining data and Split-GAL4 approach indicate that dFB^23E10Ո84C10 neurons are not homogenous when it comes to neurochemical identity and can be divided in 3 categories. Considering that acetylcholine is the main excitatory neurotransmitter in the fly system and that glutamate is inhibitory in dFB neurons [58], in addition to the high level of recurrent connectivity in the dFB [58,59], it is reasonable to assume that these different categories of dFB^23E10Ո84C10 neurons may differentially affect sleep.

To investigate this possibility, we employed optogenetic activation of dFB^VGlutՈ84C10 and dFB^ChATՈ84C10 neurons. While these tools are more homogeneous than the dFB-Split line (which contains the 3 populations of dFB neurons), they are not perfect, as both express in 2 of the 3 dFB neurons populations. When we employed our regular optogenetic activation protocol, we saw no effects, perhaps unsurprisingly based on the data obtained when activating all dFB^23E10Ո84C10 neurons (S20A Fig). We then increased the frequency of the optogenetic stimulation to 20 Hz and found that activating dFB^ChATՈ84C10 neurons significantly increased sleep while no effects were observed with dFB^VGlutՈ84C10 cells (S20B Fig). A 50 Hz optogenetic activation of dFB^ChATՈ84C10 neurons strongly increases total sleep while it does not with dFB^VGlutՈ84C10 neurons (Fig 4H–K). However, when only examining daytime sleep, we found that a 50 Hz activation of dFB^VGlutՈ84C10 neurons significantly increases sleep, although significantly less than activation of the cholinergic neurons (Fig 4L). Sleep is not increased during the night when activating dFB^VGlutՈ84C10 neurons (S20C Fig). Both daytime and nighttime sleep are increased when activating dFB^ChATՈ84C10 neurons (Figs 4L and S20C). These increases in sleep are not caused by a general deficit in locomotor activity (S20F Fig). These data suggest that within the heterogeneous dFB population, cholinergic neurons play a major role in the sleep-promoting capacity of this region. We obtained similar effects using the multibeam system, ruling out that we have falsely classified periods of micromovements as sleep (S21A–S21D Fig). Further inspection of sleep architecture revealed that while activation of dFB^VGlutՈ84C10 neurons increases daytime sleep, it does so by increasing the number of sleep episodes that flies initiate during the day rather than sleep consolidation (S20D and S20E Fig). During the night, sleep bout numbers are increased, but sleep consolidation is reduced (S20G and S20H Fig). Thus, it appears that activation of dFB^VGlutՈ84C10 neurons increases sleep initiation but does not promote sleep consolidation. This is in marked contrast to the activation of dFB^ChATՈ84C10 neurons, which increases sleep consolidation during day and night (S20E and S20H Fig). None of these effects are observed in vehicle-fed controls (S20I–S20M Fig). These data suggest that dFB^VGlutՈ84C10 and dFB^ChATՈ84C10 neurons are likely modulating different aspects of sleep and position dFB^ChATՈ84C10 neurons as the strongest modulators of sleep and sleep consolidation within the dFB.

We then acutely silenced the activity of dFB^VGlutՈ84C10 and dFB^ChATՈ84C10 neurons by expressing Shi^ts1. As seen in Fig 4M–P, silencing either set of dFB neurons significantly reduces sleep, particularly at night. These effects are not due to hyperactivity (Fig 4Q). Altogether, these data confirm that dFB^VGlutՈ84C10 and dFB^ChATՈ84C10 neurons regulate sleep. Finally, we investigated homeostatic sleep in flies in which dFB^VGlutՈ84C10 and dFB^ChATՈ84C10 neurons were silenced during the recovery period. As seen in Fig 4R, there were no differences in sleep homeostasis between controls, VGlut-AD; 84C10-DBD>Shi^ts1, and 84C10-AD; ChAT-DBD>Shi^ts1 flies when maintained at 22° post-sleep deprivation. However, raising the temperature to 29° to activate Shi^ts1 led to an abrogation of sleep rebound in flies with silenced dFB^ChATՈ84C10 neurons while not affecting sleep rebound in flies with silenced dFB^VGlutՈ84C10 neurons (Fig 4R).

To investigate whether cholinergic and glutamatergic transmission plays a role in the sleep modulatory capacity of the dFB, we expressed RNAi constructs against the VGlut and the VAChT in dFB^VGlutՈ84C10 and dFB^ChATՈ84C10 neurons. As seen in Fig 5A and 5B, expressing 2 RNAi lines against VAChT and 2 RNAi lines against VGlut in dFB^VGlutՈ84C10 neurons significantly reduces sleep in female flies. These effects are not caused by hyperactivity (Fig 5C). For dFB^ChATՈ84C10 neurons, we found that expressing one VAChT RNAi line (27,684) or one VGlut RNAi line (40,845) significantly increases sleep (Fig 5A and 5B), and that these effects are not due to a locomotor defect (Fig 5C). The second VGlut line (40,927) increases sleep during the daytime but it also leads to hypoactivity (Fig 5A–C), making conclusions about its use difficult.

Fig 5. Cholinergic and glutamatergic transmission in dFB neurons.

(A) Sleep profile in minutes of sleep per hour for Empty-Split control, VGlut-AD; 84C10-DBD and 84C10-AD; ChAT-DBD females expressing RNAi against VAChT (80,435 and 27,684) and VGlut (40,927 and 40,845). (B) Box plots of total sleep (in minutes) for flies presented in A. One-way ANOVA followed by Tukey’s multiple comparisons for each individual RNAi line. n.s. = not significant, **P P n = 20-68 flies per genotype. (C) Box plots of locomotor activity counts per minute awake for flies presented in A. Kruskal–Wallis ANOVA followed by Dunn’s multiple comparisons for each individual RNAi line. n.s. = not significant, *P P n = 20-68 flies per genotype. (D) Box plots of total sleep change in % for female control (Empty-Split), VGlut-AD; 84C10-DBD, and 84C10-AD; ChAT-DBD flies expressing CsChrimson and RNAi against GFP (control), VAChT (line 27,684) or VGlut (line 40,845) under 50 Hz activation. Two-way ANOVA followed by Sidak’s multiple comparisons for each individual RNAi line. n.s. = not significant, *P P P n = 15-43 flies per genotype and condition. The raw data underlying parts B, C, and D can be found in S1 Data. AD, activation domain; DBD, DNA-binding domain; dFB, dorsal fan-shaped body; VAChT, vesicular acetylcholine transporter.

https://doi.org/10.1371/journal.pbio.3003014.g005

These data not only suggest that both cholinergic and glutamatergic signaling act in dFB neurons to regulate sleep, but also highlight differences between dFB^VGlutՈ84C10 and dFB^ChATՈ84C10 neurons. Interestingly, a recent preprint reported that reducing VGlut levels in all 23E10-GAL4 expressing neurons reduces sleep [58]. However, whether these phenotypes can be attributed to dFB neurons is uncertain.

To investigate whether cholinergic and glutamatergic transmission is necessary for the sleep-promoting capacity of dFB^VGlutՈ84C10 and dFB^ChATՈ84C10 neurons when optogenetically activated, we expressed RNAi constructs against GFP (control), VAChT, and VGlut, while also expressing CsChrimson. As seen in Fig 5D, we found that reducing VAChT or VGlut levels in dFB^ChATՈ84C10 neurons does not block the sleep increase observed when using a 50 Hz optogenetic activation in female flies. There are multiple possible explanations for these findings. First, a subset of dFB^ChATՈ84C10 neurons also express glutamate (ChAT⁺, VGlut⁺ cells), so it is possible that these neurons use either glutamatergic or cholinergic transmission to downstream targets. This would explain why reducing levels of only one of these 2 single vesicular transporters cannot block the sleep increase obtained with optogenetic activation. Secondly, it is possible that there is enough vesicular transporter remaining in our RNAi experiments, enabling signaling to downstream targets. Alternatively, dFB^ChATՈ84C10 neurons may use another neuromodulator to promote sleep. Given that the central complex, which contains the dFB, is one of the most peptidergic areas of the fly brain [60,61], we think that it is extremely likely that dFB neurons are peptidergic. In fact, a recent preprint has identified many peptides expressed in the dFB [57]. For dFB^VGlutՈ84C10 neurons, we observed no increase in total sleep when expressing GFP RNAi and CsChrimson, in agreement with our findings expressing only CsChrimson (Fig 4K). However, when expressing both VAChT or VGlut RNAi in addition to CsChrimson, we found that a 50 Hz optogenetic activation of dFB^VGlutՈ84C10 neurons significantly increases sleep (Fig 5D). These data suggest that VGlut and VAChT may reduce the sleep-promoting capacity of dFB^VGlutՈ84C10 neurons and that these cells may also use another neuromodulator to increase sleep when sufficiently activated.

Flies were cultured at 25 °C with 50% humidity under a 12 h light:12 h dark cycle. Flies were kept on a standard yeast and molasses diet (per 1 L: 50 g yeast, 15 g sucrose, 28 g corn syrup, 33.3 ml molasses, 9 g agar). Fly stocks used in this study are listed in S3 Table.

Sleep was assessed as previously described [11,19,27], and 4- to 10-day-old virgin females or males were used as described in figure legends. Briefly, flies were placed into individual 65 mm tubes and all activity was continuously measured through the Trikinetics Drosophila Activity Monitoring System (DAM2, www.Trikinetics.com, Waltham, Massachusetts, United States of America). Locomotor activity was measured in 1-min bins and sleep was defined as periods of quiescence lasting at least 5 min. For multibeam monitoring, we used the DAM5H monitors (www.Trikinetics.com, Waltham, Massachusetts, USA).

For Fig 1P, flies expressing CsChrimson were loaded into individual 65 mm tubes with food supplemented with 400 μm all-trans retinal and monitored by the DAM5H system while simultaneously being video recorded. The multibeam system was set to record moves (fly movement from one beam to another), counts (fly movement within a beam), and position in the tube. Flies were monitored and recorded for 10 min of baseline activity with LEDs OFF followed immediately by 10 min of activation with LEDs ON at 20 Hz. Multibeam data and video analysis was performed in the same manner as for multibeam validation with Canton-S flies.

For Fig 1Q, video recording was performed as previously described [27]. Flies fed food supplemented with 400 μm all-trans retinal (Sigma, #R2500) or vehicle were loaded into individual 65 mm tubes and placed on custom-built platforms. One webcam for 2 platforms was used to record the flies for 30 min starting at ZT1 (1 h after lights turn on in the morning). Fly behavior was manually scored and categorized as walking, grooming, feeding, resting (inactivity 5 min). Videos were recorded on consecutive days for baseline (LED OFF) and activation (LED ON). Three separate experiments were performed and the pooled data is shown.

Sleep deprivation was performed as previously described [19,27,66,67]. Briefly, flies were placed into individual 65 mm tubes and the sleep-nullifying apparatus (SNAP) was used to sleep deprive flies for 12 h during the dark phase. For Fig 3J, sleep homeostasis was calculated for each individual fly as the ratio of the minutes of sleep gained above baseline after 4 h, 6 h, 12 h, 24 h, and 48 h of recovery sleep divided by the total min of sleep lost during 12 h of sleep deprivation. For Fig 4R, flies were maintained at 22° during baseline and sleep deprivation. Following 12 h of sleep deprivation, a group of flies was maintained at 22° during recovery while another one was transferred to 29° to activate Shi^ts1. Sleep homeostasis was calculated for each individual fly as the ratio of the minutes of sleep gained above baseline during 24 h of recovery sleep divided by the total min of sleep lost during 12 h of sleep deprivation.

Arousal threshold was measured on flies that were optogenetically activated. Control flies and flies expressing CsChrimson were loaded into individual 65 mm glass tubes containing food supplemented with 400 μm all-trans retinal (Sigma, #R2500) or vehicle control. After 3 days, the tubes were placed onto custom 3D printed trays, 16 tubes per tray. Video of 2 trays was recorded using a Raspberry Pi NoIR camera. Arousal stimulus was provided by three coreless vibration motors (7 × 25 mm) capable of 8,000–16,000 rmp vibration. Vibration pattern and intensity were controlled by a DRV2605 haptic controller (Adafruit) paired with a Raspberry Pi 3 Model B+. A custom program was written in python to control the video and motor vibration. The programmed sequence starts with turning video recording on. Ten minutes after recording starts the stimulus pattern begins. The DRV2605 vibration effect used is transition click (effects 62–58) starting at 20% intensity and increasing in 20% increments to 100% intensity. Vibration occurred for 200 ms followed by 800 ms no vibration repeating 4 times followed by a 15 s break before the next increasing stimulus starts. Video recording continued for 10 min after final vibration stimulus. This program repeated once per hour from ZT3 to ZT9. Videos were manually scored. Flies inactive during the 5 min directly before the stimulus began were scored as sleeping. Sleeping flies were then assessed for movement during the increasing intensity stimulus. Arousal was reported based on the percentage of flies moving at each vibration intensity. Arousal threshold was scored on consecutive days for baseline (LED OFF) and activation (LED ON). The code for the arousal threshold can be found at https://github.com/mcellinj/Dissel-lab-arousal-apparatus.

Flies were cold anesthetized, brains and VNC were dissected in ice cold Schneider’s Drosophila Medium (Gibco, 21720024), and blocked in 5% normal goat serum. Following blocking, flies were incubated overnight at 4 °C in primary antibody, washed in PBST, and incubated overnight at 4 °C in secondary antibody. Primary antibodies used were chicken anti-GFP (1:1,000; Aves Labs, Inc, #GFP-1020), mouse anti-bruchpilot (1:50; Developmental Studies Hybridoma Bank (DSHB), nc82-s), mouse anti-ChAT (1:100; DSHB, Chat4b1-c), rabbit anti-vGlut (1:500; DiAntonio lab [68], Washington University), and rabbit anti-GABA (1:1,000; Sigma-Aldrich, #A2052). Secondary antibodies used were goat anti-chicken AlexaFluor 488 (1:800; Invitrogen, #A32931), goat anti-mouse AlexaFluor 555 (1:400; Invitrogen #A32727), goat anti-rabbit AlexaFluor 555 (1:400; Invitrogen, #A32732), goat anti-mouse AlexaFluor 633 (1:400; Invitrogen, $A21052), and goat anti-rabbit AlexaFluor 633 (1:400; Invitrogen, #A21071). Brains and VNCs were mounted on polylysine-treated slides in Vectashield H-1000 mounting medium. Imaging was performed on a Zeiss 510 meta confocal microscope using a Plan-Apochromat 20× or Plan-Neofluar 40× objective. Z-series images were acquired with a 1 μm slice size using the same settings (laser power, gain, offset) to allow for comparison across genotypes. Images were processed and analyzed using ImageJ.

Naïve males were collected upon eclosion and isolated into glass sleep tubes; 48+ hours prior to training, males were transferred into new glass sleep tubes containing either vehicle or all-trans retinal dissolved in food and loaded into monitors. The 24 h period prior to training served as the baseline sleep measurement. Naïve males were transferred into 13 mm × 100 mm glass culture tubes with food at the bottom, with respective vehicle or all-trans retinal additions, and paired for 1 h with a mated trainer female (trained group) or alone (untrained group). Males were then transferred back into sleep tubes on the respective diet and sleep was monitored for the 23 h post-training period. Flies subjected to LED activation were placed under 627 nm LED light at 20 Hz (5 ms ON and 45 ms OFF continuous cycles) for the post-training period. The sleep deprived group experienced LED activation and sleep deprivation over the 23 h post-training period using the SNAP [19,27,66,67], and 24 h after the onset of training, male flies were placed into courtship chambers with a mated tester female and video recorded for the 10 min testing period. The courtship index (CI) was calculated using CI = (Time Spent on Courtship Behaviors/ Total Time of Test) * 100. The SI was then calculated using SI = 100 * (1 – (CI Trained/ CI Untrained)). Trainer and tester female mating status was visually confirmed after don juan-GFP pairing prior to training.

Courtship videos were recorded for 10 min at 20 fps through OBS Studio software. Videos were analyzed using FlyTracker software [69,70] and outputs were saved as JAABA files. The visualization tool within FlyTracker was used to confirm the assignment of male and female flies and any switches in labeling were corrected using this feature. The videos and corresponding output files were scored using JAABA, a MATLAB-based machine learning program [71]. Positive and negative frame examples were used to train 3 different scoring classifiers: OriChase (includes orientation and chase), SingWing (singing and wing extension), and MountCop (includes mounting and copulation). The accuracy of these classifiers was verified by built-in ground truth data and comparison to hand-scored tests. JAABAPlot was used to output the CI for each male fly and the resulting SI was calculated from these values.