Sleep facilitates memory by blocking dopamine neuron mediated forgetting

Dopamine Neurons are Regulated by Behavioral State

We previously observed that, after promoting the acquisition of olfactory memories, a small set of DANs that innervate the mushroom body (MB) memory center, intriguingly, display synchronized and ongoing Ca²⁺ based activity after learning that causes the forgetting of early aversive olfactory memories in Drosophila (Berry et al., 2012). While this activity occurs as reoccurring bursts, we noticed that the pattern of activity appeared temporally regulated, occurring in bouts. In order to understand how the DAN based forgetting signal might be regulated, we developed an in vivo imaging assay allowing simultaneous monitoring of a fly's DAN Ca²⁺ activity, via GCaMP^3.0 (Tian et al., 2009) expression using TH-gal4 (Brand and Perrimon, 1993; Friggi-Grelin et al., 2003), and behavior while walking on a ball supported by air (see Experimental Procedures for details; Figure 1A). We focused on two regions of the DAN processes that form synaptic connections to the MBs, referred to as neuropils, one that displays ongoing activity and belongs to the MV1 neuron, and an adjacent control region belonging to the V1 neuron, which is relatively inactive (Figure 1B). Remarkably, a 1 hr simultaneous recording of locomotion and DAN activity revealed that the MV1 neuropil displayed activity resembling the coarse temporal pattern of locomotor behavior (Figure 1C). We used ball rotation data to cluster time points into either behaviorally active or rest states, and found that MV1 neuropil activity was robustly elevated during active states, whereas the V1 neuropil activity remained low in both states, but had a slight decrease during active states (Figure 1D). Furthermore, the MV1 neuropil Ca²⁺ signal was strongly correlated with ball rotation, particularly in the lower frequency domains (freq < 0.002 Hz, or approx. 1 cycle every 8 min or more, timescales consistent with that of locomotor bout structure (Donelson et al., 2012))(Figure 1E and Figure S1A). Finally, we looked at DAN activity during stable transitions into and out of behaviorally active states (> 5 sec stable state before and after transition) by aligning transition segments of recordings across all animals. Interestingly, MV1 DAN activity robustly increased upon transition into active states, while, conversely, dropped during rest states (Figure 1F). V1 activity remained low and was not significantly regulated with behavioral transitions. Together, these data, along with our observations of synchronized activity between MV1 and another DAN, MP1 (Berry et al., 2012), indicate that the ongoing activity from specific sets of DAN involved in forgetting, including MV1, is regulated with the behavioral state of the animal.

Given the strong correlation between DAN activity and locomotor activity, we tested whether DAN activation might promote locomotor activity, that is, whether DAN activity is upstream of locomotor circuits. Two prior studies found no role for these MB innervating DANs in regulating locomotor activity (Liu et al., 2012; Ueno et al., 2012). When we blocked the synaptic output from these DANs, using c150-gal4 (Dubnau et al., 2003) with restricted expression in MV1, MP1, and V1 DANs to drive temperature-sensitive UAS-shi^ts1 function (Kitamoto, 2001), we did not see decreased locomotor activity between temperatures, although the experimental genotype exhibited less activity at high temperature compared to one (gal4 alone), but not both control genotypes (Figure S1B, S1C, and S1E). We also noted from imaging experiments that locomotor activity occasionally occurs while the MV1 neuron is not active (Figure 1C), thus further supporting that locomotor behavior does not require c150-gal4 DAN output. Furthermore, stimulation of these neurons, using UAS-trpA1 (Hamada et al., 2008), did not produce genotype specific and robust increases in locomotor activity (Figure S1D and S1F). But similar to the blocking experiments, high temperature increased the locomotor activity of the two control genotypes (UAS-trpA1 and c150-gal4 alone). These data indicate that c150-gal4 DAN output is neither necessary nor sufficient to acutely drive locomotor activity. We conclude, therefore, that the ongoing signal in MV1 is either downstream of locomotor behavior itself, or is regulated in parallel, by other brain areas that promote arousal and locomotor activity.

Sleep Reduces Ongoing DAN Activity

Given that the ongoing activity in MV1 was highest during behaviorally active states, we tested the hypothesis that reducing behavioral activity with increased sleep drive would reduce this ongoing activity. The GABA_A agonist, Gaboxadol (or THIP), has been shown to specifically promote deep non-REM sleep in humans (Faulhaber et al., 1997) and rats (Lancel and Faulhaber, 1996), while leaving REM sleep intact; sleep characteristics similar to those occurring during normal homeostatic regulation of sleep. Therefore, we attempted to induce sleep in Drosophila by feeding flies various doses of Gaboxadol (Figure S2A). Shortly after Gaboxadol feeding, long periods of quiescence, (> 5 min) conventionally defined as “sleep” in Drosophila, significantly increased during both day and night (Figure S2B) in a dose dependent manner (Figure S2C). Next, we fed flies Gaboxadol (0.1 mg/ml) for one day and then removed the drug (Figure 2A) to test whether these effects were reversible. Once again, Gaboxadol treatment increased sleep, occurring as bouts with increased duration, but remarkably, total sleep and bout duration actually decreased after drug removal compared to control fed flies (Figure 2B and S2D). These data indicate that less sleep is needed in flies given Gaboxadol the prior day, suggestive of a homeostatic response. Finally, to test the arousability of flies fed Gaboxadol, we delivered a single mechanical stimulus every hour for one day followed by a day of drug treatment (Figure S2E and S2E’). Interestingly, the average evoked activity, post-stimulus, (see Experimental Procedures) was significantly reduced with increasing Gaboxadol dosage (0.1 and 1.0 mg/ml, Figure 2C, Figure S2E’). These data suggest that having Gaboxadol onboard increases arousal thresholds. Altogether, our data indicate that Gaboxadol, similar to effects on mammals, induces bona fide sleep in Drosophila, with hallmark characteristics that include reversible quiescence, homeostatic regulation, and increased arousal thresholds.

In order to observe the effects of Gaboxadol on DAN activity, we perfused varying concentrations of this sleep agent across the brain while performing in vivo imaging of MV1 activity and monitored fly body movement in a recording chamber (see Experimental Procedures, Figure 2D and 2E). Like walking on the ball, ongoing MV1 activity was also regulated with behavioral state in this assay, increasing during bouts of body movement (Figure S2F and S2G). Remarkably, Gaboxadol perfusion rapidly (Figure 2F and Figure S2H and S2H’) and robustly attenuated both fly movement (Figure 2G) and MV1 activity (Figure 2H) at 0.01 and 0.1 mg/ml. Furthermore, we found that the quiescent behavioral state and reduced MV1 activity was fully reversible with wash out (Figure 2I-2L), thus eliminating pharmacological induced damage as a cause of decreased physical and DAN activity.

To rule out non-specific effects of Gaboxadol and extend these results, we increased sleep drive by thermogenetic stimulation of the sleep circuit. Recent studies identified a dorsal fan shaped body (dfsb) circuit in the central brain, specifically represented in the R23E10-gal4 line (Jenett et al., 2012) (Figure 3A and Figure S3A), which acts as the effector arm of the sleep homeostat (Donlea et al., 2011; Donlea et al., 2014). Consistent with these studies, TrpA1 based stimulation of R23E10-gal4 expressing neurons caused a rapid and robust increase in daytime sleep followed by a negative sleep rebound the day after stimulation (Figure 3B and 3C), confirming the dfsb circuit's role in homeostatic sleep regulation. In order to measure DAN activity in vivo while using the gal4-uas system to modulate the sleep circuit, we developed a TH-lexA line (Figure S3B) to express GCaMP^3.0 in the MV1 and V1 DAN neurons, their associated MB neuropil regions, as well as DAN innervation of the anterior inferior medial protocerebrum (PR) (Figure 3D and 3E), a region also exhibiting ongoing activity like MV1 (Berry et al., 2012). While simultaneously measuring movement and DAN activity, we recorded before (“Pre”), during (“Treat”), and after (“Post”) stimulation of dfsb neurons (Figure 3F). As predicted, stimulation of the sleep circuit rapidly decreased fly behavioral activity and was accompanied with a robust decrease in MV1 and PR DAN activity, with no change in the control V1 region (Figure 3G and 3H). Fly behavioral activity was partially restored and ongoing activity in MV1 and PR completely restored to pre-stimulation levels after stimulation of the sleep circuit was ceased. These results, along with those from Gaboxadol administration, indicate that increased sleep drive dramatically reduces the ongoing activity of DANs involved in forgetting.

Increased Sleep After Learning Impairs Forgetting

Since ongoing MV1 activity is decreased with increasing sleep drive, we hypothesized that acutely and reversibly increasing sleep drive specifically after learning would reduce DAN mediated forgetting. To test this, we first chose to modulate sleep with Gaboxadol after aversive olfactory conditioning, where populations of flies learn to associate an odor with electric shock. Memory to this association is then tested in a T-maze, giving flies the choice between the trained odor and an unconditioned odor. Since memory from this kind of training decays quickly after training, we sought to increase the rate of Gaboxadol consumption, and thus the rate of sleep onset, by increasing the hunger of flies via starvation prior to feeding. As we observed previously, flies fed Gaboxadol experienced more sleep than controls, and a 16 hr starvation period increased this effect (Figure S4A). Furthermore, we found that flies removed from Gaboxadol food 1 hr after learning (Figure 4A) partially returned to control sleep and activity levels by the 6^th hr (Figure S4B and S4C), and completely by the 8^th hr after learning (Figure S4D and S4E), indicating that these time points were appropriate for testing memory retrieval. This Gaboxadol feeding protocol led to increased sleep (Figure 4B) during the period of memory retention. Remarkably, we found that flies forced to sleep with Gaboxadol treatment after learning exhibited enhanced memory retention at both 6 and 8 hr (Figure 4C and 4D). Similarly, sleep circuit stimulation after conditioning also rapidly and reversibly induced sleep (Figure 4E-4H) and enhanced both 3 and 6 hr memory retention (Figure 4I). Importantly, simultaneous stimulation of the dfsb sleep circuit and c150-gal4 DANs also led to strong sleep induction (Figure 4J and K). Memory retention, in contrast, was markedly decreased (Figure 4L), similar to that observed with stimulation of DANs alone (Berry et al., 2012). Therefore, sleep, after learning, loses it protective qualities when DAN signaling is artificially potentiated. This circuit level epistasis experiment indicates that DAN mediated forgetting is downstream of sleep networks. Together, the data indicate that increased sleep and reduced arousal after learning reduces DAN mediated forgetting of aversive olfactory memories.

Increased Arousal After Learning Increases Forgetting

If the DANs innervating the MB memory center are downstream of arousal circuitry, then they should respond to arousing stimuli. In fact, these neurons have already been shown to respond to many salient stimuli, including odors and electric shock (Mao and Davis, 2009) and temperature changes (Tomchik, 2013). Since mechanical stimuli have been extensively used to arouse flies for sleep deprivation (Shaw et al., 2000), we delivered airpuffs to the fly using a protocol shown to induce arousal in flies (Lebestky et al., 2009), while simultaneously recording fly movement and DAN activity (Figure 5A). We found that the DAN processes in all three areas (MV1, PR, V1) of the mushroom body neuropil exhibited robust responses to each airpuff. However, MV1 responsiveness was maintained across stimuli while the other regions showed attenuated responsiveness across stimuli (Figure 5B). Importantly, both fly movement and ongoing MV1 activity continued at an elevated level just after stimulation (Figure 5C and 5D), indicating a stimulus-induced elevation in arousal and MV1 DAN activity.

Next, we reasoned that increasing the arousal after learning would accelerate DAN mediated forgetting. To test this, we developed a population arousal device (Figure 6A) that allowed us to deliver mechanical stimuli (2 sec stimulus every 1 min over 80 min) to flies in population vials after aversive olfactory learning (see Experimental Procedures for details). We found that mechanical stimuli delivered for the first 80 min after learning significantly aroused populations of flies, leading to an overall increase in activity between each stimulus (Figure S5), with activity levels dropping back to control levels after treatment (Figure 6B and 6C). Importantly, mechanical stimulation after learning caused a robust decrease in 3 hr memory for wild-type Canton-S flies (Figure 6D). However, acquisition and immediate memory retrieval were not disrupted by prior mechanical stimulation (Figure 6E), indicating that the stimuli must be delivered after learning to observe its disruptive effects. Remarkably, we found that blocking neuronal output of c150-gal4 DANs specifically during mechanical stimulation blocked the forgetting induced by this treatment (Figure 6F). Therefore, our data indicate that increasing arousal after learning accelerates DAN mediated forgetting.

Fly Stocks

Fly stocks were cultured on standard food at room temperature. The following lines were used for experiments, crosses, and to generate stocks: Canton-S, w^CS10 (or +), TH-gal4 (3^rd chromosome; Friggi-Grelin et al., 2003), R23E10-gal4 (attP2, 3^rd chromosome; Jenett et al., 2012; Donlea et al., 2014), c150-gal4 (3^rd chromosome; Dubnau et al., 2003), MBgal80 (2^nd chromosome; Krashes et al., 2007), TH-lexA (2^nd chromosome; this study, see below), UAS-shi^ts1 (PJFRC100-20XUAS-TTS-shibire(ts1)-p10 in attP2, 3^rd chromosome; Pfeiffer et al., 2012), UAS-trpA1 (2^nd chromosome; Hamada et al., 2008), UAS-mCD8::GFP (X chromosome; Lee and Luo, 1999), UAS-GCaMP^3.0 (2^nd chromosome; Tian et al., 2009), UAS-RFP (2^nd chromosome; Pramatarova et al., 2003), lexAop-GCaMP^3.0 (pJFRC27-13XLexAop2-IVS-GCamp3-p10 in attP2, 3^rd chromosome; Pfeiffer et al., 2012), lexAop-mtdTomato (pJFRC48-13XLexAop2-IVS-myrtdTomato in su(Hw)attP1, 3^nd chromosome; gift from Dr. G. Rubin), lexAop-myr::GFP (pJFRC59-13XLexAop2-IVS-myr::GFP-p10 in attP2, 3^rd chromosome; Pfeiffer et al., 2012). For all Gaboxadol sleep/activity and memory experiments (Figure 2A-2C, Figure 4A-4D, Figure S2A-S2E’, and Figure S4) and mechanical stimuli experiments (Figure 6B-6E and Figure S5), Canton-S flies were used. Crosses were raised at 25°C with 70% relative humidity with a 12 hr light-dark cycle, with the exception of DAM monitoring (see below) experiments involving TrpA1 stimulation (Figure S1B-S1F and Figure 3B and 3C) that were maintained at a lower 23°C to minimize baseline TrpA1 stimulation. All experiments were conducted starting at lights on. For more details on genetic crosses, see Extended Experimental Procedures.

TH-lexA Construction

To begin construction, we obtained nlsLexA:GADfl plasmid (#26232, Addgene; Pfeiffer et al., 2010) and pCaSpeR-TH (containing GAL4 encoding sequences) from Dr. S. Birman (Friggi-Grelin et al., 2003). We removed the GAL4 encoding sequence from pCaSpeR-TH via digestion with BamHI restriction enzyme and a FseI restriction enzyme site polylinker was placed at the BamHI site to facilitate the cloning of the nlsLexA:GAD^fl open reading frame. nlsLexA:GAD^fl was PCR amplified using FseI tailed PCR primers and the product was inserted into the newly made FseI site in the GAL4-less pCaSpeR2-TH vector upstream of the TH gene open reading frame. This produced the pCasDTH-nlsLexA:GAD^fl construct used for injections (Rainbow Transgenic Flies, Inc.) to create the TH-lexA transgenic flies used in this study.

Immunostaining and Microscopy

Whole brains were isolated and processed in a manner similar to the Janelia Farm's Fly Light Project protocol (Jenett et al., 2012). See Extended Experimental Procedures for details.

Fly on Ball

In concept, fly on ball experiments were setup similar to a previous study (Seelig et al., 2010). To avoid disturbing both physical activity and DAN activity, flies were not anesthetized during any imaging procedure. For more details see Figure S6 and Extended Experimental Procedures.

Gaboxadol and In Vivo imaging

Female flies, 3-6 day old, were positioned and attached inside a plastic recording chamber (different than fly on ball setup above) using myristic acid and dissected as previously described (Berry et al., 2012). Fresh physiological saline (described above) was used in these experiments. For video recordings, the fly's body was illuminated with a halogen light source (Schott, #A20500). Preliminary experiments using Fluorescein dye (#F6377, Sigma-Aldrich) showed that perfusion of 2 mg/ml dye at 10 ml/min through 0.0812 inch inner diameter tubing (Cole Palmer, #95609-42), flowing from the entry point of the chamber, created full dye concentration in the brain after 10 sec. This gave us a rough estimate of how long it would take to get the full concentration of Gaboxadol into the brain. Therefore, during Gaboxadol perfusion experiments (Figure 2D-2H, and Figure S2H and S2H’) we perfused at 10 ml/min for 10 sec prior to reducing speed to 1 ml/min and recording movement and DAN activity for 20 min. For Gaboxadol wash out experiments (Figure 2I-2L), we perfused (1 ml/min) fresh saline for 10 min while measuring movement and DAN activity (“Pre”), then perfused Gaboxadol (0.01 mg/ml at 1 ml/min) for 5 min (significant effects on movement and DAN activity were seen as early as ~5 min, Figure S2H and S2H’), then recorded 10 min of movement and DAN activity (“Treat”), then washed out Gaboxadol with fresh saline at high perfusion rate (10 ml/min) for 35 min, and finally reduced flow rate to 1 ml/min for final 10 min recording (“Post”). See “Data Analysis: In Vivo Imaging” in Extended Experimental Procedures for details on subsequent analysis of video and confocal recordings.

Sleep Circuit Stimulation and In Vivo Imaging

For sleep circuit stimulation during in vivo imaging (Figure 3D-3H), fly age, preparation, dissection, saline, perfusion conditions, and lighting were the same as for “Gaboxadol and In Vivo Imaging” above except for the addition of in-line heater/cooler (Harvard Apparatus, #64-0353) just upstream of the perfusion chamber. This in-line heater/cooler was cooled with a liquid cooling system (Warner Instruments, #641922) and the temperature was regulated with a temperature controller (Harvard Apparatus, #64-0352) using feedback from a thermistor positioned upstream and near the fly head capsule. Using a perfusion rate of 3 ml/min, a temperature of 23°C was maintained near the brain of the fly during a 10 min recording of movement and DAN activity (“Pre”), then the temperature was either ramped up to 34°C (to ensure that the brain located some distance from the probe reaches at least 32°C) and maintained for a 60 min recording (“Treat”) before being brought back to 23°C, or for a control group of flies, it was maintained at 23°C during “Treat” window. Finally a 10 min recording was made at 23°C (“Post”). See “Data Analysis: In Vivo Imaging” in Extended Experimental Procedures for details on subsequent analysis of video and confocal recordings.

Locomotor Activity and Sleep Monitoring

For experiments measuring locomotor activity and sleep in flies (except in vivo imaging experiments), we used the Drosophila Activity Monitoring (DAM) system (Trikinetics, Inc.). DAM monitoring was conducted on 3-5 day old female flies. See Extended Experimental Procedures for more details.

Gaboxadol Effects on Sleep and Arousability

For differential Gaboxadol dosage experiments looking at 24 hr sleep (Figure S2A-S2C), just prior to lights on, the flies were tapped into new glass tubes containing food with either no (0 mg/ml) or a range of concentrations (0.01, 0.1, 1.0 mg/ml diluted in water) of Gaboxadol and the glass tubes were placed into DAM monitors. In a separate multiday experiment (Figure 2A-2C), flies were put into control food tubes for a day (“Pre”), then just before lights on the next day, were quickly transferred to tubes with 0 (“Ctrl”) or .1 mg/ml Gaboxadol for a treatment day (“Treat”), and finally, were transferred back to control food for a “Post” day. In order to measure the effects of Gaboxadol on the “arousability” of flies, we designed a protocol similar to previous methods (van Alphen et al., 2013; Donlea et al., 2014). A strong 5 sec vibrational stimulus is given to the DAM monitors (via Velcro attachment to the population arousal device, see below) once every hour for two days (Figure S2E). The first day all flies are kept on control food (“Pre”), but just prior to lights on the second day, flies were transferred to 0, 0.01, 0.1, or 1.0 mg/ml Gaboxadol food tubes for the treatment day (“Treat”). We aligned all the stimulus points and observed that most of the change in activity upon stimulation occurred primarily within 15 min after stimulus (Figure S2E’). Therefore, to quantitate the response of flies to stimuli, we calculated an “Evoked activity” as the difference between a fly's activity 15 min before and after stimulus (Figure S3E). Finally, we calculated the mean evoked activity (Figure 2C) by averaging evoked activity at all the day time stimulus points for all the flies in a given group.

To evaluate the effects of starvation on Gaboxadol efficacy (Figure S4A), 50-60 flies were first placed in standard large vials with normal stock food and then transferred into starvation vials with 1.0% Agar (water but no nutrients) for 0, 2, 5, or 16 hr prior to aspirating individual flies into tubes with food containing either Gaboxadol (0.1 mg/ml), or control food (“Ctrl”). Sleep was then monitored for 3 hr with the DAM system. For Gaboxadol feeding after olfactory learning (Figure 4A and 4B, and Figure S4C-S4G), 50-60 flies were first starved for 16 hr (as above) and then subjected to standard aversive olfactory conditioning (see below). Individual flies were then aspirated into DAM tubes containing either control food (“Ctrl) or 0.1 mg/ml Gaboxadol (“Gab”) for either 1 hr (Figure 4A and 4B) or 3 hr (Figure S4B). Flies were then tapped into new DAM tubes with either control food (“Ctrl→Ctrl” or “Gab→Ctrl”) or into fresh 0.1 mg/ml Gaboxadol food (“Gab→Gab”) for the remainder of the recording.

Sleep Circuit Stimulation and Sleep Monitoring

Experiments measuring sleep and locomotor activity after sleep circuit stimulation (Figure 3B and 3C), were conducted the same as with Gaboxadol and sleep monitoring above, with the following exceptions. After one day of DAM monitoring at 23°C (“Pre”), monitors were either moved to 32°C, or kept at 23°C, for a second day (“Stim”), and finally moved back to, or kept at, 23°C for an additional day (“Post”). For experiments measuring stimulation induced sleep after olfactory learning (Figure 4E-4H, 4J, and 4K), 50-60 flies were subjected to standard aversive olfactory conditioning (see below) and then individual flies were aspirated into 23°C or 32°C DAM tubes/monitors and sleep was monitored. At 15 min prior to either 3 or 6 hr post learning, 32°C monitors were then moved to 23°C (23°C monitors were also moved but kept at 23°C to keep stimuli induced by transferring monitors similar between temperature groups).

Olfactory Conditioning and Memory Retention

For all memory experiments in this study, standard aversive olfactory training and testing was used similar to what has been previously described (Beck et al., 2000). For details on conditions and procedures see Extended Experimental Procedures. For experiments examining the effect of Gaboxadol on memory retention (Figure 4C and 4D), flies were first either starved (“S”, see starvation conditions above) or not (“NS”), trained with olfactory conditioning (“A”), and then put on 2% agarose: 5% sucrose food with either 0.1 mg/ml Gaboxadol (“G”) or not (“C”) for 1 hr before being transferred to new vials with control 2% agarose: 5% sucrose food. Memory retention was measured at either 6 or 8 hr (“R”). For experiments looking at the effect of sleep circuit stimulation on memory retention (Figure 4I), flies were first trained with olfactory conditioning (“A”) at 23°C and then they were either tapped into new 23°C or 32°C glass vials with 2% agarose: 5% sucrose food during memory retention. Finally, vials were either returned to 23°C, or kept at 23°C, 15 min prior to testing (“R”) memory retention at 3 or 6 hr. For experiments measuring memory retention after mechanical stimulation (Figure 6D), flies were first trained with olfactory conditioning (“A”), then tapped into food-less vials (flies can get stuck in food during mechanical stimulation) and were either mechanically stimulated (“Stim”, see “Population arousal assay” below), or not (“Ctrl”), for 80 min after learning prior to measuring memory retention at 3 hr (“R”). Experiments measuring memory retention after blocking c150-gal4 DANs during mechanical stimulation (Figure 6F) followed the same procedure, except flies were trained (“A”) at 23°C, mechanically stimulated (“Stim”) or not (“Ctrl”) at either 23 °C (Figure 6F, left) or 32 °C (Figure 6F, right) for 80 min after learning, and then tapped into 23 °C vials with normal stock food until 3 hr memory retention was measured (“R”).

Airpuff Stimulation of DANs

For experiments where airpuffs were used to stimulate flies (Figure 5), a protocol similar to one previously shown to induce arousal in Drosophila (Lebestky et al., 2009) was used. Briefly, we used a programmable relay (10C1D-D-V2 ZEN unit, OMRON) to deliver six 200 msec airpuffs (2 L/min), delivered at 5 sec intervals via a pipette tubing placed ~2.5 mm from the ventral-frontal surface of the thorax, abdomen, and legs of flies hanging in the perfusion chamber. Both movement and DAN activity were measured before, during, and after airpuff stimulation. See Extended Experimental Procedures for details on processing and calculation of DAN activity and movement. Peak DAN responses were calculated by finding the maximum activity during each 5 sec interval following the beginning of the airpuff protocol. A 15 sec time window before and after the six airpuffs, “Pre” and “Post” respectively, was defined to look for airpuff-induced changes in the mean fly movement and DAN activity.

Population Arousal Assay

For experiments where we mechanically stimulated flies in population vials (Figure 6), a custom-made arousal device (Figure 6A) was used to mechanically stimulate populations of flies in vials standardly used for olfactory conditioning in the lab (see Extended Experimental Procedures for details. For measuring population activity (Figure 6B and 6C), 4 pairs of vials of 50 Canton-S flies (25 male / 25 female) were trained with aversive olfactory conditioning (same as above except, for simplicity, flies were trained only to the one odor, 0.05% 3-octanol), and placed onto the mechanical stimulus platform which was either still (“Ctrl”) or shaking for 80 min before stopping (“Stim”). A video camera (Firefly MV, Point Grey) was used to record the population activity of these 4 pairs of vials containing live flies plus one pair of vials of dead flies (see “Data Analysis: Population Activity” section in Extended Experimental Procedures) during and after stimulation.

Data analysis

For details regarding analysis of both in vivo imaging and population activity data see Extended Experimental Procedures.

Statistical Methods

Statistics were performed using Prism 5 (Graphpad). All tests were two tailed and confidence levels were set at α = 0.05. A detailed explanation of the tests and comparisons made for each experiment is presented in Table S1 in the order they appear in the text. Unless otherwise stated, non-parametric tests were used for in vivo imaging data, video camera data, and DAM system monitoring, while parametric tests were used for olfactory memory comparisons as PI values are normally distributed (Tully et al., 1994).