Logo of nihpaAbout Author manuscriptsSubmit a manuscriptNIH Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
J Neurosci. Author manuscript; available in PMC Oct 17, 2013.
Published in final edited form as:
PMCID: PMC3670110

A dynamic deep sleep stage in Drosophila


How might one determine whether simple animals such as flies sleep in stages? Sleep in mammals is a dynamic process involving different stages of sleep intensity, and these are typically associated with measurable changes in brain activity (Blake and Gerard, 1937; Rechtschaffen and Kales, 1968; Webb and Agnew, 1971). Evidence for different sleep stages in invertebrates remains elusive, even though it has been well established that many invertebrate species require sleep (Campbell and Tobler, 1984; Hendricks et al., 2000; Shaw et al., 2000; Sauer et al., 2003). Here we use electrophysiology and arousal-testing paradigms to show that the fruit fly, Drosophila melanogaster, transitions between deeper and lighter sleep within extended bouts of inactivity, with deeper sleep intensities after ~15 and ~30 minutes of inactivity. As in mammals, the timing and intensity of these dynamic sleep processes in flies is homeostatically regulated and modulated by behavioral experience. Two molecules linked to synaptic plasticity regulate the intensity of the first deep sleep stage. Optogenetic upregulation of cyclic adenosine monophosphate (cAMP) during the day increases sleep intensity at night, whereas loss of function of a molecule involved in synaptic pruning, the fragile-X mental retardation protein (FMRP), increases sleep intensity during the day. Our results show that sleep is not homogenous in insects, and suggest that waking behavior and associated synaptic plasticity mechanisms determine the timing and intensity of deep sleep stages in Drosophila.


Most animals endowed with a brain require daily sleep (Campbell and Tobler, 1984; Cirelli and Tononi, 2008), and this state of behavioral quiescence is thought to be crucial for a number of processes related to cognition (Killgore, 2010). Sleep is often first identified in animals by prolonged immobility, but is better characterized by decreased responsiveness to stimuli (Campbell and Tobler, 1984). Other criteria, such as homeostatic regulation, neural correlates in brain activity, and modulation by stimulants such as caffeine are also typically probed in order to measure sleep physiology in animals (van Swinderen, 2006). In humans, sleep is a dynamic process where the brain transitions through different stages of activity such as rapid eye movement (REM) and three stages of non-REM, including slow-wave sleep (SWS). These distinct stages, typically identified in by electroencephalographic (EEG) recordings (Rechtschaffen and Kales, 1968), are associated with different arousal thresholds: in humans lower levels of behavioral responsiveness to stimuli occur during SWS, and comparatively higher responsiveness levels occur during REM sleep (Rechtschaffen et al., 1966). Whereas both REM sleep and SWS have been proposed to play an important role in memory consolidation (Diekelmann and Born, 2010), SWS has been proposed to be specifically involved in maintaining synaptic homeostasis in mammals (Tononi and Cirelli, 2003, 2006).

Recent research on sleep genetics in model organisms, such as the fruit fly Drosophila melanogaster, has provided evidence that sleep processes engage molecular mechanisms that regulate synaptic function (Bushey et al., 2009; Donlea et al., 2009; Gilestro et al., 2009; Bushey et al., 2011), although flies show no evidence of the slow-wave brain activity associated with deep sleep in mammals (Nitz et al., 2002). Although oscillatory brain activity has been found in sleeping crayfish (Ramon et al., 2004), and honeybees and cockroaches appear to display varying levels of sleep intensity (Tobler and Neuner-Jehle, 1992; Sauer et al., 2003; Eban-Rothschild and Bloch, 2008), sleep/wake in Drosophila has been conventionally quantified as a bimodal process. Based on early sleep research in this model, the duration of fly inactivity reflects accrued sleep processes (Andretic and Shaw, 2005), and a fly is considered to be asleep if it has been inactive for five minutes or more (Shaw et al., 2000; Huber et al., 2004). Although the 5-minute inactivity threshold has been invaluable for understanding sleep processes in this model, and has already provided insight on sleep intensity in flies (Huber et al., 2004), it remains unclear whether flies, like mammals, display defined sleep stages which might be associated with specific cellular functions. This question has become increasingly relevant since Drosophila has become a model of choice for investigating molecular processes associated with sleep in health and disease (van Alphen and van Swinderen, 2011). In this study, we investigated whether flies display changing levels of sleep intensity throughout the day and night, and within extended sleep bouts. We then used behavioral and genetic manipulations to measure whether our sleep intensity metrics are affected in predictable ways.

Materials and Methods

Fly stocks and rearing conditions

Flies were raised at 22°C on standard yeast-based Drosophila media, on a 12 hour light-12 hour dark (LD) rhythm. The day before each experiment, adult Drosophila (4–8 days old) were briefly (<5 min) anesthetized using CO2, after which males and females were transferred to plastic vials containing standard fly media. Each vial contained ~40–50 males or females. The next day, individual flies were aspirated into 65mm glass tubes (Trikinetics, Waltham, MA) containing fly media on one end and sealed with a cotton plug on the other end. Canton-S (CS) and white1118 (w1118) are standard laboratory strains. The following lines were ordered from the Bloomington Stock Center: Fmr1Δ50M(stock nr 6930, w1118; Fmr1Δ50M/TM6B; Tb1); elav-Gal4/Cyo; TM2/TM6B. UAS-PACα flies were a gift of Martin Schwärzel (Schroder-Lang et al., 2007) and were crossed to +/elav-Gal4/Cyo; TM2/TMB6.


Brain recordings were performed as described previously (Nitz et al., 2002). Briefly, local field potentials (LFPs) were sampled at 300 Hz as a voltage differential from two glass electrodes inserted into the fly brain, one in each hemisphere. Electrode recording locations were verified by iontophoresis of Texas Red dye and subsequent brain imaging using standard brain dissection techniques and fluorescence microscopy. These were typically deep into either optic lobe, as shown previously (van Swinderen, 2012). To monitor fly activity and rest, the tethered fly (only wild-type females were used) was positioned on a humidified air-suspended ball upon which it could walk or rest at will (Fig. 1A). All flies used for electrophysiology (N=13) survived at least 24 hours. Some flies were tethered in the morning and others in the evening to ensure that day and night effects were not purely a consequence of time on the tether. Filming of the preparation with a webcam (Logitech) at 3 frames/s under infrared light allowed for quantification of fly activity levels. For each frame, a pixelated, logical (black and white) picture was generated which could be subtracted from the preceding frame in order to generate a difference image (Δ pixels) through time. The threshold for immobility was determined offline, by visually comparing the activity trace (as shown in Fig. 1B) with the recorded movie of the fly to determine the average number of pixels that change from frame to frame for a completely inactive fly. This noise is caused by small fluctuations in lighting conditions as well as camera noise. A threshold was chosen that was +2 s.d above the inactivity average for the whole trace (20 or 25 pixels for all flies, see red line in Fig 1E,F). Pixel fluctuations above this threshold are considered to be fly activity, and flies are classified as being awake as soon as their activity is above their individual threshold, even for just one frame. Sleep duration was thus counted from the last instance of any activity (this was the sleep initiation point), plus 5 min or more of inactivity. LFP data were time-stamped to match movement data. Fourier analyses of LFPs were performed using Matlab software. For the summed frequency analysis, 45–55 Hz range was removed to prevent line noise contamination. To quantify LFP dynamics within sleep bouts for day and night time sleep, we selected sleep bouts of 11–20 minutes, because such data presumably included the 5 min sleep “transition” period. Also, these are the longest sleep bouts that occur regularly (3–5 times per fly – Fig. 1D) during both day and night. For every minute of these sleep bouts, a Fourier transform was performed on the LFP data to determine its power spectrum. Low frequencies (<10Hz) were removed to eliminate fluctuations caused by heart rate or respiratory activity. The resulting power spectrum was summed over the 11–40 and 41–80Hz range. The lower frequencies (11–40 Hz) bracket a range where we have previously observed attention-like effects in the LFP(van Swinderen and Greenspan, 2003). We contrast these efffects to 41–80 Hz because these represent an adjoining range of frequencies where we have not observed attention-like effects in the LFP.

Fig. 1
Recording brain activity during fly sleep

Behavioral assays

Behavioural responsiveness to a mechanical stimulus

Flies were individually housed in 65 mm glass tubes (Trikinetics, Waltham, MA), 17 tubes on a tray, two trays per filmed experiment. Arousal thresholds were measured by subjecting flies to a mechanical stimulus using shaft-less vibrating motors (Precision Microdrives, model 312–101). Stimulus intensity was controlled by randomly modulating the voltage used to drive the motors to deliver 5×200 ms pulses (one pulse per second) between 0 and 1.2g (steps of 0.3g), using a custom Matlab program interfacing with the analogue output channels of a USB data acquisition device (1280 LS, Measurement Computing). The unit g is used to express vibration amplitude (1 g equals the gravitational force at the surface of the earth, 980 cm/s2). The Precision Microdrive motors have a linear relationship between input voltage and vibration, where an increase of 0.5V results in an increased vibration of 0.3g, with a maximum of 1.2g at 2.5V. When flies are startled by mechanical stimulation, this triggers a locomotion response (Sawamura et al., 2008). Flies were recorded at 15 frames per second for one minute before (“baseline”) and one minute after the stimulus (“startle”) using a high-resolution camera (Grasshopper, Point Grey Research, Richmond, BC, Canada). Images were stored for offline analysis with custom Matlab software. Stimulus-induced locomotion was calculated by subtracting the average velocity during the minute before stimulus onset from the average velocity during 1 minute after stimulus delivery. Some flies did not respond even at the strongest intensity (1.2g). These were given a 1.5g score and labelled as non-responding (NR). Experiments were conducted between 1pm and 5pm in the afternoon. In a separate set of experiments, a maximal vibration intensity (1.2 g) was delivered every hour over 24 hours, and velocity changes were monitored over 3 days and nights. Circadian effects on arousal were tested by raising flies in light-dark (LD) conditions, and then transferring them to the arousal-testing incubator for three days, with lights turned permanently off (DD condition) or on (LL condition).

Arousal thresholds

Arousal thresholds were tested with sequentially increasing vibration intensities, from 0 to 1.2g, in 0.3g (200ms) increments, every 10s, once an hour over 24 hours. For these experiments, a slightly different protocol was employed, based on visual observation of filmed experiments. Arousal thresholds were calculated by assigning the vibration intensity (g) that evoked a locomotion response (walking at least half the length of the glass tube) in quiescent animals (i.e., flies that had not shown any movement in the preceding minute), and determining the distribution of g values for a strain. Since data were non-parametric, kruskal-wallis analyses were used to test for significant effects among strains or conditions.

Automated arousal testing over 72 hours

For sleep intensity experiments, two trays of 17 flies each were placed in an incubator at 22°C and a 12 hour light:dark cycle (lights on at 8 am). Fly activity over 72hours was recorded using a webcam (Logitech 9000), modified to record infrared light (see http://www.pysolo.net/docs/hacking-a-webcam-to-improve-lighting-conditions/forinstructions). Visible light was filtered using an infrared long pass filter (Edmund optics, stock no NT 43–948). Images were taken with Pysolo (Gilestro and Cirelli, 2009) at one frame per second and stored for offline analysis using custom Matlabsoftware. Every hour, flies were subjected to a stimulus consisting of five 200ms pulses at 1.2g. Responsiveness was determined for each fly at each time point, using the velocity criteria described above. For each stimulus, percentages of responding flies were obtained by determining the fraction of flies that showed a significant increase in velocity, using a two sample Kolgomorov-Smirnov test to determine if the velocities during the minute before and after the startle stimulus are part of the same distribution. Since filming was continuous for these experiments, flies were grouped into different 1 min bins (from 0 to 59 min) depending on how long they had been inactive prior to the test stimulus. Filmed analysis was also performed on some experiments to measure brief awakenings (BA), as in (Huber et al., 2004), since BAs were found in that study to correlate with sleep intensity. Filmed activity traces (velocity) were divided into 1 min bins and converted into binary data (1=active, 0=inactive) based on our activity threshold. By searching for ‘010’ sequences, we were able to calculate the average number of brief awakenings (BA) per hour. Our filmed analyses were consistent with previously published data for decreased brief awakenings in sleep-deprived flies (Huber et al., 2004): BA/hr: male day, ZT1–3, 1.13±0.20; male day, sleep-deprived ZT1–3, 0.65±0.14 (P=0.037); female day, ZT1–3, 0.98±0.15; female day, sleep-deprived ZT1–3, 0.53±0.20 (P=0.045, by t-test comparing means). In addition, there were differences in BA between day and night sleep, consistent with our finding that sleep intensity is different between day and night: BA/hr: male day, 1.01±0.18; male night, 0.40±0.03 (P=0.01); female day, 0.98±0.15; female night, 0.53±0.20 (P=0.001, by t-test comparing means).

Fly activity

For some experiments, activity was measured using the infrared Drosophila Activity Monitor System (DAMS, Trikinetics, Waltham, MA). As flies move back and forth in the tube, they interrupt an infrared beam at the centre of the tube. These interruptions are counted and stored per minute and analysed offline using Matlab® (2011a, Mathworks, Natick, MA). Fly sleep in these devices was defined as any period of inactivity longer than five minutes, as reported previously (Shaw et al., 2000; Huber et al., 2004). Average daytime sleep bout duration (min ± s.e.m) for our wild-type strain was 22±0.5 (males) and 23±1.0 (females). Average night time sleep bout duration was 32±0.5 (males) and 35±0.5 (females, n=32 flies per sex).

Sleep deprivation

Individual flies in glass tubes were placed in a SNAP device (Shaw et al., 2002) and sleep deprived for 24 hours. One mechanical stimulus was delivered every 20 seconds which caused the flies to be tapped down their respective tubes. Sleep deprivation ran from 8 am till 8 am the next morning, or from 8 pm till 8 pm the next evening, after which flies were transferred to the arousal setup to measure responsiveness during rebound sleep during the flies’ subjective day or night. Sleep deprivation effects were also confirmed by identifying sleep rebound in the Trikinetics setup (see above).


Adult CS flies (3–4 days old) were briefly anesthetized using CO2 and transferred to plastic vials (5mm radius, 65mm length) containing food. The socialized groups consisted of 40–50 flies (males and females were kept separate), The isolated group consists of a single fly in each vial. After four days of exposure to either a social or a socially impoverished environment, flies were transferred to the arousal testing setup.

Image processing

All image processing was done in Matlab (Mathworks, Natick, MA). After each experiment, fly positions were extracted from recorded images using an image subtraction method. For each recording session, a reference image was created by taking the average of all images recorded during that session. This averaging process eliminates all dynamic objects (the flies) and keeps all static objects (the rest of the setup). Fly positions were extracted by subtracting the reference image from each recorded image, transforming the resulting image to a logical black-white image and calculating the centre of mass for each fly. Fly position traces were differentiated to obtain velocity.


PACα activity was induced using two arrays of 32 blue LEDs (458±10 nm, element14.com, part # FNL-U500B22WCSL). During arousal threshold experiments, which were conducted during the day in lit chambers, blue illumination started 10 minutes before the first trial and continued throughout the experiment. During 72hr arousal testing experiments, flies were illuminated with only blue light throughout the light period (8am–8pm), and this was turned off at night.

Curve fitting

To determine when sleep was at its lowest intensity, we fitted 3rd order polynomials to the sleep intensity data, which has the following functional form:


The first turning point of equation (1) corresponds to the moment when deepest sleep is achieved, and can be determined by solving the polynomial’s derivative for zero. The turning point that had the lowest value of x, which corresponds to the local minimum, was selected to determine the time and amplitude of deepest sleep. To visualise sleep onset and intensity, responsiveness data for immobile flies was processed in five steps. 1. Mean responsiveness data was normalised such that the responsiveness of active flies (at 0 minutes of inactivity) was set to a value of 1. 2. Normalised data was smoothed using a 5-point moving average. 3. 3rd order polynomials were fitted to smoothed, normalised means. 4. The turning points in the curve were determined by solving the polynomial’s derivative for zero: dy/dx = 3*ax^2 + 2*bx + c = 0. 5.

Statistical analyses

All data were analyzed in Matlab. Responsiveness to stimuli was analyzed using a two sample Kolgomorov-Smirnov test to determine if the velocities during the minute before and after the startle stimulus are part of the same distribution. Arousal threshold data were non-parametric. Hence, Kruskal-Wallis analyses were used to test for significant effects among strains or conditions. Mean sleep durations were compared by Wilcoxon rank sum test. All data are presented as means ± s.e.m. Comparisons of sleep intensity between strains and/or conditions was done by first fitting sleep intensity data (“fraction response”) by a 3rd order polynomial (see above) and finding the deepest sleep point. A 5-min window was framed around this point, and the original, un-smoothened data within this window were then averaged to determine sleep intensity at this point. These data were compared between strains and conditions to uncover significant differences in sleep intensity (by one-way ANOVA or Wilcoxon rank sum test). Significance was set at P<0.05 for all statistical tests.



To first investigate whether flies might display distinct sleep stages during prolonged bouts of inactivity, we performed brain recordings on individual animals (Fig. 1). In previous studies we have shown that sleep in Drosophila is associated with overall decreased activity in the brain, as measured by local field potentials (LFPs) (Nitz et al., 2002; van Swinderen et al., 2004), although LFP dynamics during sleep were not examined, and sleep was defined according to criteria established for untethered flies (>5 min of inactivity (Shaw et al., 2000; Huber et al., 2004)). To record LFPs from behaving flies, we used a similar preparation as previously published (Nitz et al., 2002), with the difference that the tethered fly could walk (or rest) on an air-suspended ball during recordings (Fig. 1A). Filming of the fly under infrared lighting allowed movement to be quantified (see Materials and Methods) during the day and night (Fig. 1B). Filmed tracking of fly activity levels provided sufficient resolution for detecting when flies stopped moving entirely, and determining what happened in the brain LFP during prolonged inactivity (which is equivalent to immobility at our level of resolution, Fig. 1B). Tethered flies were significantly less active at night, and the night was characterized by longer inactivity bouts than the day (Fig.1C). However, tethered flies did regularly display extended bouts of inactivity (>5 min) in this setup, day or night (Fig. 1D). We are assuming this prolonged in activity is a resting state.

Following 5 min of inactivity, fly brain LFP activity becomes quite flat (Fig. 1E), as shown previously (Nitz et al., 2002). We are assuming this is sleep. In contrast, the waking LFP is more active, even when the animal is not displaying any detectable movement for a few seconds (Fig. 1F). Analyses of LFPs confirmed that brain activity (11–80Hz) in wild-type flies is significantly attenuated during sleep (Fig. 2A). However, day and night-time sleep are not equivalent: the sleeping fly brain is significantly more active during the day than during the night (Fig.2A). Wake LFP, in contrast, is not different between the day and the night (Fig. 2A). We next examined LFP activity through time during extended (11–20 min) sleep bouts (Fig. 2B). We found that LFP power (11–80 Hz) changes considerably during a sleep bout (Fig. 2C,D), although there was variability among flies in the timing of LFP changes in power. A closer examination of different LFP frequency domains (low versus high) revealed significant variability in the lower frequencies (11–40 Hz), depending on time since sleep initiation (Minutes asleep, Fig. 2E), but no changes in higher frequencies during extended sleep bouts (Fig. 2F).

Fig. 2
Changes in brain activity during fly sleep

Arousal thresholds

Transiently floored 11–40Hz activity during sleep (Fig. 2E) suggested a deeper sleep stage, because average LFP power is already lower during sleep than wake (Fig. 2A). To determine whether these dynamic sleep processes are behaviorally relevant, we investigated whether flies displayed different levels of behavioral responsiveness during sleep. To efficiently test and measure behavioral responsiveness in flies, we designed an apparatus to deliver brief (200ms) mechanical stimuli of varying (randomized) intensities (0–1.2g, see Materials and Methods) to groups of flies housed individually in small glass tubes (Fig.3). Flies were filmed, and their locomotor responses to the stimuli were automatically tracked (Fig. 3A). Startle responses increase with stimulus intensity in both males and females (Fig. 3B,C). By probing hourly for responses to robust vibration intensities (1.2 g) we were able to query how average behavioral responsiveness changed throughout several consecutive days and nights (Fig. 3D). As expected, responsiveness to the vibrations followed a robust day/night pattern, with decreased responsiveness at night and increased responsiveness during the day when flies are exposed to 12hr light-dark (LD) cycles (Fig. 3E–G). Flies kept in constant darkness (DD) also displayed increased responsiveness during their subjective day (Fig. 3H), but flies kept in constant light (LL) displayed an intermediate level of behavioral responsiveness that was not different between day and night (Fig. 3I). Our startle paradigm thus accurately probes fly arousal under the LD conditions that are traditionally used to study sleep in Drosophila.

Fig. 3
Stimulus-induced locomotion

We therefore next examined arousal thresholds specifically in quiescent flies. By probing only non-moving animals with increasing vibration intensities (Fig.4A), we determined the minimal vibration intensity needed to evoke movement (arousal threshold, see Materials and Methods) in different individuals. By probing hourly over several days and nights, we then determined how arousal thresholds change over 24 hours (Fig. 4B). As expected, arousal thresholds are significantly lower during the day than at night, in males and females (P<0.001, 1 d.f., H=401.01, Kruskal-Wallis test for medians, N = 60, only male data is shown). Indeed, the arousal threshold profiles we found (Fig.4B) closely resemble more conventional graphs plotting hourly sleep duration in Drosophila based on a 5 min inactivity threshold (Hendricks et al., 2000; Shaw et al., 2000; Huber et al., 2004; Andretic and Shaw, 2005) (and see Fig. 8E in this study), validating our approach as a reliable indicator of arousal state in flies (see also Materials and Methods, where we describe “brief awakenings” per hour, which was found to be a reliable indicator of sleep intensity in flies, as shown in an earlier study (Huber et al., 2004)).

Different levels of sleep intensity in quiescent flies
Fig. 8
Effect of socialization on fly sleep intensity

Sleep intensity

Sleep duration and sleep intensity are not necessarily correlated – as suggested by our brain recording experiments. To investigate sleep intensity in our assay, we applied our arousal-testing paradigm to examine responsiveness levels during extended bouts of inactivity (as determined by video tracking), during the day or night. We measured behavioral responsiveness to a robust vibration stimulus (1.2g) every hour for three days, and then determined whether the duration of prior immobility (only inactive flies were analyzed) predicted responsiveness probability. We compiled all ranges of inactivity from 1 to 59 minutes, for day versus night (see Materials and Methods). We confirmed that night-time sleep was on average deeper than daytime sleep, but also found that responsiveness probability varied depending on the time elapsed (Fig.4C,D). We confirmed that wild-type flies are significantly less responsive after 5 min of immobility (Shaw et al., 2000; Huber et al., 2004), in males and females, day or night (Fig.4E,F), but also that flies already become significantly less responsive than moving flies after only 1 min of inactivity (P<0.01 for males and females, day or night). Responsiveness decreases gradually in wild-type flies (i.e., sleep intensity increases), until a deeper sleep stage is achieved after ~10 min of immobility, for day or night sleep, in males and females (Fig.4C–F). The timing of this deeper sleep stage is consistent with an early observation of a distinct postural change occurring at 14 min in a similar preparation (Hendricks et al., 2000). Interestingly, after this first period of deepest sleep intensity, responsiveness levels then appear to change again: sleep becomes lighter at around 20 min, and a second deep sleep stage is evident around 30 min in both sexes, day or night, ending in comparatively lighter sleep as the quiescent hour proceeds (Fig.4C–F). Indeed, analysis of extended sleep bouts during the day revealed that male flies that have been quiescent for 45–60 min are as highly responsive as flies that have been quiescent for less than 5 min (Fig.4E). Sleep intensity in flies thus depends on how long they have been asleep, but this relationship is nonlinear with some evidence of what appears to be cycling behavior in arousability. Responsiveness differed significantly across the 12 5 min bins (one-way ANOVA, males, day: F(11,46) = 15.2, P<0.001; males, night: F(11,46) = 23.16, P<0.001; females, day: F(11,46) = 23.59, P<0.001; females, night: F(11,46) = 16.49, P<0.001).

To investigate sleep intensity stages by awakening animals with a stimulus, as we have done, presents a fundamental conundrum: individual animals can only be awoken once during a sleep bout, after which they are obviously no longer asleep. So how can we investigate sleep dynamics in flies behaviorally? Our population assay suggests that, on average, flies are more deeply asleep after 10 and 30 min of immobility, compared to other times – especially females (Fig. 4C,D). Although our electrophysiology supports the view that sleep intensity levels may be dynamic throughout a sleep bout (Figs. 1&2), it is nevertheless difficult to demonstrate behaviorally that this is happening in individual animals. To get around this problem, we questioned whether different groups of flies may be contributing disproportionally to different parts of the sleep intensity curves that we uncovered (Fig. 4C,D). In that experiment, we probed hourly for responsiveness, and for all animals that were immobile at the time of probing, we determined how long they had been immobile, thereby assigning them to an inactivity bin between 1 min and 59 min. We found that responsiveness levels were normally distributed among all inactive flies tested (Lilliefors test for normally, P<0.05, Fig. 5A,B). Plotting the contributions of each fly to different inactivity bins also revealed these to be evenly distributed across individuals (Fig. 5C,D). That is, any fly that had provided responsiveness data for 20–40 min inactivity bins, for example, also provided responsiveness data for 0–20 min inactivity bins (i.e., these were not different groups of flies). However, it is clear that all flies contributed less to 20–40 min inactivity bins, compared to 0–20 min and 40–60 min (Fig. 5E,F). One explanation for this distribution of the data could be that sleep intensity levels are indeed cycling within an extended sleep bout: there may be fewer 20–40 min events because flies that have slept through the end of their first sleep cycle will most likely initiate the next sleep cycle, and therefore not wake up during the deeper stages (i.e., centered around 30 min) of their second sleep cycle. Most flies, however, would naturally awaken after 20 min, which would be consistent with our averaged data showing that sleeping flies are comparatively more responsive then (Fig. 4CD), and with the average sleep bout length being ~20–30 min in our wild-type strain (see Materials and Methods). We therefore focused on the first 20 min of sleep, which includes the first deep sleep stage, for the remainder of this study.

Fig. 5
Individual fly sleep data distribution

To better identify the first deep sleep stage, we fit a 3rd order polynomial to smoothened responsiveness data (see Materials and Methods), with the first local minimum in the curve indicating the time of deepest sleep (Fig.6A,B, vertical dashed lines). We identified the first deep sleep stage occurring between 12 and 16 min inactivity in wild-type flies, irrespective of time of day or sex, which suggests strongly regulated processes controlling sleep intensity in Drosophila. When we tested for responsiveness every 30 min instead of every hour in a replicate set of experiments, the first deep sleep stage seemed earlier and deeper, especially in males (Fig.6C,D), suggesting that a 30 min testing regime may be increasing sleep drive (i.e., producing some sleep deprivation). Similarly, our brain-recording paradigm yielded a comparatively earlier onset of lower 10–40Hz activity (Fig. 2E), suggesting that the tethered recording setup may also be causing some sleep deprivation. We therefore focused on an hourly sleep probe in subsequent experiments.

Fig. 6
The first deep-sleep stage

Behavioral effects on sleep intensity

We next examined the effect of sleep drive on sleep intensity dynamics. In humans, SWS duration and intensity increase proportionally to prior wake time, and decrease as sleep is restored throughout the night (Borbely and Achermann, 1999) – evidence that sleep is a homeostatic process. Some of our data already suggests a homeostatic process: quiescent male flies are less responsive in the first hours of the night, compared to equally quiescent flies in the last hours of the night (Fig.4B). As in humans, extended wakefulness in Drosophila males and females results in increased sleep drive (pressure to sleep)(Hendricks et al., 2000; Shaw et al., 2000). To more directly determine whether increased wakefulness altered sleep intensity profiles, we sleep-deprived flies for 24 hours (see Materials and Methods) and measured sleep intensity for the following day or night. We found that 24 hours of sleep deprivation precipitated the onset of deep sleep in males and females (Fig. 7A,B, vertical dashed lines). To quantify sleep intensity at this time we averaged the (non-smoothened) data for a 5 min window around the point of deepest sleep (Fig. 7C), termed the deep sleep response probability (RP). Deep sleep RP was significantly lower in sleep-deprived flies, but only for nighttime sleep (Fig. 7D). Importantly, sleep intensity returns to comparatively lighter levels after a full day of recovery, for both day and night in males (Fig. 7E, one-way ANOVA F(1,135)=5.42, and F(1,135)=11.78, respectively). Female recovery from sleep deprivation was less clear, where only daytime responsiveness appeared to recover on the second day (Fig. 7F, one-way ANOVA F(1,135)=11.14).

Fig. 7
Effect of sleep deprivation on the first deep sleep stage

Other behavioral manipulations predicted to increase sleep intensity are socialization and learning. The synaptic homeostasis hypothesis for sleep (Tononi and Cirelli, 2006) describes an interaction between wake experience and sleep drive: information encoded during wake leads to an overall increase in synaptic strength in the brain, which would then be proportionally normalized (i.e., downscaled) during the deeper stages of sleep. Since socialization has been proposed as one way to increase sleep drive in Drosophila (Ganguly-Fitzgerald et al., 2006), we tested the effect of this simple manipulation on sleep intensity in our arousal paradigm. Socialized male flies (males kept with other males for 4days, see Materials and Methods) slept more deeply subsequently during the day, compared to similarly handled male flies kept in isolation (Fig.8A,B). In contrast, sleep intensity in female flies was not affected by intra-sexual socialization (Fig.8C,D), even though sleep duration was altered (Fig. 8E,F), as measured by more traditional (i.e., infra-red beam crossing) sleep metrics. This result highlights an important point: longer sleep duration does not necessarily mean flies are sleeping more deeply. The dimorphismin sleep intensity between males and females may be due to different behavioral interactions among males or females, with more complex male social interactions, such as aggression and courtship, producing a greater sleep drive during the day. These social interactions may, for example, be engaging learning and memory pathways that affect sleep intensity in males.

Optogenetic manipulation of sleep intensity

To investigate a possible link between learning and sleep intensity in flies (see (Seugnet et al., 2008)), we acutely manipulated a molecular pathway that has been associated with synaptic plasticity and memory formation. Learning and memory processes are closely tied to cyclic adensosine monophosphate (cAMP) signaling pathways in neurons, where increased cAMP activity results in increased synaptic efficacy (Huang et al., 1994) as well as in transcription of genes involved in up-regulating synaptic strength (Dash et al., 1990). Fly memory mutants defective in cAMP regulation, such as dunce1 (which has chronically increased cAMP levels) sleep less (Hendricks et al., 2001) and do not show experience-dependent changes in sleep duration (Ganguly-Fitzgerald et al., 2006). We therefore sought to specifically increase cAMP activity only during the day, when most learning behavior presumably occurs, to see whether acute molecular control of plasticity mechanisms in awake flies altered sleep intensity. To best achieve this, we used an optogenetic approach: a transgenic adenylyl cyclase that is activated by blue light, PACα (Schroder-Lang et al., 2007), under control of the UAS/Gal4 expression system (Brand and Perrimon, 1993). Under blue light, PACα converts ATP into cAMP in the Gal4-specified neurons (Fig. 9A), initiating a signaling cascade that modulates synaptic function (Huang et al., 1994; Schroder-Lang et al., 2007). We found that acute, pan-neuronal activation of PACα (in elav-Gal4/UAS-PACα flies) significantly increased arousal thresholds (decreased responsiveness), to levels similar to dunce1mutants (Fig. 9B, p<0.01, Kruskal Wallis test comparing medians, 3 d.f., H=12.63). When adenylyl cyclase activation was continued for the flies’ entire day (but turned off at night), elav-Gal4/UAS-PACα flies displayed chronically decreased responsiveness to mechanical stimuli, during the day and night, compared to controls also exposed to blue light (Fig. 9C). However, sleep intensity was only significantly increased at night, after blue light was turned off, compared to both genetic controls (Fig.9D–F). This suggests that cAMP-related signaling that accrues during the day has long-lasting effects on sleep intensity at night.

Fig. 9
Adenylyl cyclase activity increases night-time sleep intensity

Synaptic homeostasis and sleep intensity

The synaptic homeostasis hypothesis for sleep proposes that synaptic potentiation accrued during the day is proportionally downscaled during sleep (Tononi and Cirelli, 2003, 2006) (Fig. 10A). A protein closely associated with synaptic remodeling and sleep function is FMRP, produced by the fragile-X mental retardation gene (Fmr1), which regulates synaptic pruning (Tessier and Broadie, 2008) and synaptic plasticity (Mercaldo et al., 2009). Sleep duration in flies is tightly regulated by dFmr1, the Drosophila homologue of the gene (Wan et al., 2000). In Drosophila, loss of dFmr1 results in overgrown dendrites (Pan et al., 2004; Bushey et al., 2011), while dFmr1 over-expression shows opposite effects: reduced dendritic branching and loss of synaptic differentiation (Pan et al., 2004). Increased sleep duration in dFmr1 loss of function mutants is therefore thought to result from less efficient synaptic downscaling during sleep, although an association with deeper sleep, predicted in humans (Tononi and Cirelli, 2006), has never been shown in flies. Our results thus far suggest that deeper sleep may be associated with homeostatic functions in the fly brain (Fig. 10B). What would be the effect on sleep intensity if a component of the synaptic downscaling machinery were defective? Using our arousal paradigm to test dFmr1 mutants, we found that a loss of function allele of dFmr1 (d50) displayed deeper night-like sleep during the day (Fig.10C,D). Female dFmr1mutants showed a similar significant effect (Fig.10E,F). Thus, rather than further deepening night-time sleep (as in the case of sleep-deprived flies, Fig. 7D), loss of function of dFmr1 mainly deepens daytime sleep.

Fig. 10
Loss ofdFmr1 gene function increases daytime sleep intensity


The discovery over a decade ago that Drosophila melanogaster sleeps (Hendricks et al., 2000; Shaw et al., 2000) has revolutionized approaches to studying sleep functions in animal models. This is because the powerful tools associated with Drosophila genetic analysis could be applied to understanding sleep functions, which had been traditionally studied in higher animals less amenable to molecular genetic analysis. Increasingly, Drosophila sleep phenotypes are now being used as a model to study the molecular underpinnings of cognitive disorders, such as schizophrenia and mental retardation (van Alphen and van Swinderen, 2011), with the idea that a fundamental connection is likely to exist between many brain disorders and defects in sleep function. One of the proposed functions of sleep that may provide this connection is synaptic homeostasis (Tononi and Cirelli, 2006). According to this hypothesis (which remains actively debated (Frank, 2012; Tononi and Cirelli, 2012)), deep sleep in mammals serves to proportionally downscale synaptic strengths across the brain, thereby decreasing energy and space requirements while preserving the relative synaptic weights that are a manifestation of learning and plasticity in the brain. Indeed, some cellular and molecular evidence for synaptic downscaling during sleep has come from recent research in Drosophila (Donlea et al., 2009; Gilestro et al., 2009; Bushey et al., 2011). However, one setback with using Drosophila to study this mammalian-centric hypothesis, is that deep sleep specifically has been proposed as the stage during which synaptic downscaling occurs, and delta (0.5–4 Hz) “slow” waves as the mechanism that may be enabling downscaling (Tononi and Cirelli, 2006). Yet, despite earlier work suggesting simple approaches to identify sleep intensity in flies (Huber et al., 2004; Andretic and Shaw, 2005), most Drosophila sleep studies often equate total sleep duration with sleep intensity: the longer flies are inactive (as determined by infra-red beam crossing devices), the more sleep functions are presumably being accomplished. Our study shows that sleep duration does not necessarily equate with sleep intensity: flies can sleep more lightly or more deeply at different times of the day or times since sleep onset. In addition, there is no strong evidence so far of slow-wave activity during sleep in flies or other insects (van Swinderen, 2006). So, the use of the powerful Drosophila model to investigate sleep functions originally proposed for mammals may have met with some resistance because there was no evidence of different sleep stages in flies, and no evidence of distinct electrical signatures in the fly brain associated with sleep intensity – two key sleep criteria in mammals. Our study identifies a deep sleep stage in flies.

A synthesis of our electrophysiological, behavioral, and genetic manipulations demonstrate that sleep in flies, as in mammals and birds, is a dynamic, heterogeneous state – which suggests that different sleep stages are a fundamental characteristic of sleep in any animal. Sleep in Drosophila transitions through stereotypical epochs of increasing and decreasing intensity, and our results show that behavioral and genetic manipulations can alter the timing and intensity of the first deep sleep stage (characterized in Fig. 6). Importantly, different strains may display different baseline sleep intensity profiles (e.g., baseline sleep appears lighter in Pacα/+ controls, Fig. 9DE), so a 5 min threshold for defining sleep in Drosophila may not be appropriate for all strains. Then, daytime sleep is lighter on average than nighttime sleep, so accumulating sleep duration metrics by combining day and night sleep (e.g., “total sleep”) may not always be valid. Finally, sleep intensity can become lighter as flies remain inactive, so total length of a sleep bout does not necessarily reveal what kind of sleep is occurring at any one time. This may especially be the case for mutant strains, where sleep duration may not reflect normal sleep functions. We find, for example, that dFmr1 mutants sleep more deeply during the day, while nighttime sleep intensity remains unchanged. This suggests that loss of FMRP-related synaptic downscaling transfers nighttime sleep functions to the day, to perhaps offset the less efficient downscaling occurring during the night in the mutant. On the other hand, cAMPup-regulation during the day deepens nighttime sleep, perhaps because more downscaling is required following this artificial up-regulation of synaptic activity. Finally, our socialization protocol appears to only increase daytime sleep intensity, and only in males.

The existence of a deep sleep stage in Drosophila raises the question of why flies also display a substantial amount of lighter sleep, especially during the day. Indeed, the non-equivalence of daytime and night-time sleep was observed in most of our experiments and manipulations, supporting the idea that daytime and night-time sleep achieves distinct functions in flies (Ishimoto et al., 2012), which we propose can also be understood as lighter and deeper sleep stages. If the function of deep sleep in flies and other animals is associated with synaptic downscaling, why is there also a need for extensive periods of light sleep? Are the behavioral defects in dFmr1mutants due to defective deep sleep processes or due to lost light sleep processes? Our finding that sleep intensity is regulated by two molecules (FMRP and cAMP) involved in plasticity and synaptic remodeling allows us to speculate that a distinct suite of molecules will be expressed to achieve specific sleep functions in correlation with the timing and depth of the deep sleep stage (Fig. 10B). By considering how sleep intensity and timing changes following behavioral manipulations or in mutant strains, future studies should uncover precise, functional roles for sleep processes in the Drosophila model. Although flies do not appear to display the “delta” slow waves (Nitz et al., 2002) which have been proposed as a mechanism for synaptic downscaling during mammalian deep sleep (Tononi and Cirelli, 2006), it is possible that reduced oscillatory activity within behaviorally-relevant frequency ranges, such as 11–40 Hz, may accomplish similar synaptic downscaling functions in simpler animals with smaller brains. Although we have not combined electrophysiology with arousal threshold experiments in this study, the first deep sleep stage in Drosophila appears to be matched by transiently floored 11–40Hz LFP activity in the brain. Since increased activity in a similar LFP frequency range has been associated with selective attention and choice behavior in awake flies (van Swinderen and Greenspan, 2003; Tang and Juusola, 2010), it is possible that transiently decreased activity in this same frequency range reflects a homeostatic downscaling response during deep sleep in insects. Future studies should establish the connection, if any, between global changes in electrical activity in the fly brain and expression of genes involved in regulating synaptic function.


We thank Martin Schwaerzel for generously providing the UAS-PACα strain, Paul Shaw for discussions on the manuscript, Angelique Paulk for brain dissections and artwork, and Yan-Qiong Zhou for help on the fly-ball design. This work was supported by an Australian Research Council Discovery project grant (DP1093968) and Future Fellowship (FT100100725) to BvS.


Bart van Alphen: Designed the research, designed the startle paradigm, conducted experiments, analyzed the data, and wrote the paper.

Melvyn Yap: Designed and conducted experiments, analyzed data.

Leonie Kirzsenblat: Designed and conducted experiments, analyzed data.

Benjamin Kottler: Designed the startle paradigm.

Bruno van Swinderen: Designed the research, analyzed the data, wrote the paper.


  • Andretic R, Shaw PJ. Essentials of sleep recordings in Drosophila: moving beyond sleep time. Methods Enzymol. 2005;393:759–772. [PubMed]
  • Blake H, Gerard R. Brain potentials during sleep. Am J Physiol. 1937;119:692–703.
  • Borbely AA, Achermann P. Sleep homeostasis and models of sleep regulation. J Biol Rhythms. 1999;14:557–568. [PubMed]
  • Brand AH, Perrimon N. Targeted gene expression as a means of altering cell fates and generating dominant phenotypes. Development. 1993;118:401–415. [PubMed]
  • Bushey D, Tononi G, Cirelli C. The Drosophila fragile X mental retardation gene regulates sleep need. J Neurosci. 2009;29:1948–1961. [PMC free article] [PubMed]
  • Bushey D, Tononi G, Cirelli C. Sleep and synaptic homeostasis: structural evidence in Drosophila. Science. 2011;332:1576–1581. [PMC free article] [PubMed]
  • Campbell SS, Tobler I. Animal sleep: a review of sleep duration across phylogeny. Neurosci Biobehav Rev. 1984;8:269–300. [PubMed]
  • Cirelli C, Tononi G. Is sleep essential? PLoS Biol. 2008;6:e216. [PMC free article] [PubMed]
  • Dash PK, Hochner B, Kandel ER. Injection of the cAMP-responsive element into the nucleus of Aplysia sensory neurons blocks long-term facilitation. Nature. 1990;345:718–721. [PubMed]
  • Diekelmann S, Born J. The memory function of sleep. Nat Rev Neurosci. 2010;11:114–126. [PubMed]
  • Donlea JM, Ramanan N, Shaw PJ. Use-dependent plasticity in clock neurons regulates sleep need in Drosophila. Science. 2009;324:105–108. [PMC free article] [PubMed]
  • Eban-Rothschild AD, Bloch G. Differences in the sleep architecture of forager and young honeybees (Apis mellifera) J Exp Biol. 2008;211:2408–2416. [PubMed]
  • Frank MG. Erasing synapses in sleep: is it time to be SHY? Neural Plast. 2012;2012:264378. [PMC free article] [PubMed]
  • Ganguly-Fitzgerald I, Donlea J, Shaw PJ. Waking experience affects sleep need in Drosophila. Science. 2006;313:1775–1781. [PubMed]
  • Gilestro GF, Cirelli C. pySolo: a complete suite for sleep analysis in Drosophila. Bioinformatics. 2009;25:1466–1467. [PMC free article] [PubMed]
  • Gilestro GF, Tononi G, Cirelli C. Widespread changes in synaptic markers as a function of sleep and wakefulness in Drosophila. Science. 2009;324:109–112. [PMC free article] [PubMed]
  • Hendricks JC, Finn SM, Panckeri KA, Chavkin J, Williams JA, Sehgal A, Pack AI. Rest in Drosophila is a sleep-like state. Neuron. 2000;25:129–138. [PubMed]
  • Hendricks JC, Williams JA, Panckeri K, Kirk D, Tello M, Yin JC, Sehgal A. A non-circadian role for cAMP signaling and CREB activity in Drosophila rest homeostasis. Nat Neurosci. 2001;4:1108–1115. [PubMed]
  • Huang YY, Li XC, Kandel ER. cAMP contributes to mossy fiber LTP by initiating both a covalently mediated early phase and macromolecular synthesis-dependent late phase. Cell. 1994;79:69–79. [PubMed]
  • Huber R, Hill SL, Holladay C, Biesiadecki M, Tononi G, Cirelli C. Sleep homeostasis in Drosophila melanogaster. Sleep. 2004;27:628–639. [PubMed]
  • Ishimoto H, Lark A, Kitamoto T. Factors that Differentially Affect Daytime and Nighttime Sleep in Drosophila melanogaster. Front Neurol. 2012;3:24. [PMC free article] [PubMed]
  • Killgore WD. Effects of sleep deprivation on cognition. Prog Brain Res. 2010;185:105–129. [PubMed]
  • Mercaldo V, Descalzi G, Zhuo M. Fragile X mental retardation protein in learning-related synaptic plasticity. Mol Cells. 2009;28:501–507. [PubMed]
  • Nitz DA, van Swinderen B, Tononi G, Greenspan RJ. Electrophysiological correlates of rest and activity in Drosophila melanogaster. Curr Biol. 2002;12:1934–1940. [PubMed]
  • Pan L, Zhang YQ, Woodruff E, Broadie K. The Drosophila fragile X gene negatively regulates neuronal elaboration and synaptic differentiation. Curr Biol. 2004;14:1863–1870. [PubMed]
  • Ramon F, Hernandez-Falcon J, Nguyen B, Bullock TH. Slow wave sleep in crayfish. Proc Natl Acad Sci U S A. 2004;101:11857–11861. [PMC free article] [PubMed]
  • Rechtschaffen A, Kales A. A manual of standardized terminology, techniques and scoring system of sleep stages in human subjects. Los Angeles: Brain Information Service/Brain Research Institute; 1968. [PubMed]
  • Rechtschaffen A, Hauri P, Zeitlin M. Auditory awakening thresholds in REM and NREM sleep stages. Percept Mot Skills. 1966;22:927–942. [PubMed]
  • Sauer S, Kinkelin M, Herrmann E, Kaiser W. The dynamics of sleep-like behaviour in honey bees. J Comp Physiol A Neuroethol Sens Neural Behav Physiol. 2003;189:599–607. [PubMed]
  • Sawamura N, Ando T, Maruyama Y, Fujimuro M, Mochizuki H, Honjo K, Shimoda M, Toda H, Sawamura-Yamamoto T, Makuch LA, Hayashi A, Ishizuka K, Cascella NG, Kamiya A, Ishida N, Tomoda T, Hai T, Furukubo-Tokunaga K, Sawa A. Nuclear DISC1 regulates CRE-mediated gene transcription and sleep homeostasis in the fruit fly. Mol Psychiatry. 2008;13:1138–1148. 1069. [PMC free article] [PubMed]
  • Schroder-Lang S, Schwarzel M, Seifert R, Strunker T, Kateriya S, Looser J, Watanabe M, Kaupp UB, Hegemann P, Nagel G. Fast manipulation of cellular cAMP level by light in vivo. Nat Methods. 2007;4:39–42. [PubMed]
  • Seugnet L, Suzuki Y, Vine L, Gottschalk L, Shaw PJ. D1 receptor activation in the mushroom bodies rescues sleep-loss-induced learning impairments in Drosophila. Curr Biol. 2008;18:1110–1117. [PMC free article] [PubMed]
  • Shaw PJ, Cirelli C, Greenspan RJ, Tononi G. Correlates of sleep and waking in Drosophila melanogaster. Science. 2000;287:1834–1837. [PubMed]
  • Shaw PJ, Tononi G, Greenspan RJ, Robinson DF. Stress response genes protect against lethal effects of sleep deprivation in Drosophila. Nature. 2002;417:287–291. [PubMed]
  • Tang S, Juusola M. Intrinsic activity in the fly brain gates visual information during behavioral choices. PLoS One. 2010;5:e14455. [PMC free article] [PubMed]
  • Tessier CR, Broadie K. Drosophila fragile X mental retardation protein developmentally regulates activity-dependent axon pruning. Development. 2008;135:1547–1557. [PMC free article] [PubMed]
  • Tobler II, Neuner-Jehle M. 24-h variation of vigilance in the cockroach Blaberus giganteus. J Sleep Res. 1992;1:231–239. [PubMed]
  • Tononi G, Cirelli C. Sleep and synaptic homeostasis: a hypothesis. Brain Research Bulletin. 2003;62:143–150. [PubMed]
  • Tononi G, Cirelli C. Sleep function and synaptic homeostasis. Sleep Med Rev. 2006;10:49–62. [PubMed]
  • Tononi G, Cirelli C. Time to be SHY? Some comments on sleep and synaptic homeostasis. Neural Plast. 2012;2012:415250. [PMC free article] [PubMed]
  • van Alphen B, van Swinderen B. Drosophila strategies to study psychiatric disorders. Brain Res Bull 2011
  • van Swinderen B. Competing visual flicker reveals attention-like rivalry in the fly brain. Front Integr Neurosci. 2012;6:96. [PMC free article] [PubMed]
  • van Swinderen B, Greenspan RJ. Salience modulates 20–30 Hz brain activity in Drosophila. Nat Neurosci. 2003;6:579–586. [PubMed]
  • van Swinderen B, Nitz DA, Greenspan RJ. Uncoupling of brain activity from movement defines arousal States in Drosophila. Curr Biol. 2004;14:81–87. [PubMed]
  • van Swinderen . Sleep in invertebrates. In: Kaas JH, editor. Evolution of Nervous Systems. Oxford: Academic Press; 2006.
  • Wan L, Dockendorff TC, Jongens TA, Dreyfuss G. Characterization of dFMR1, a Drosophila melanogaster homolog of the fragile X mental retardation protein. Mol Cell Biol. 2000;20:8536–8547. [PMC free article] [PubMed]
  • Webb WB, Agnew HW., Jr Stage 4 sleep: influence of time course variables. Science. 1971;174:1354–1356. [PubMed]
PubReader format: click here to try


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...