Recurrent Circuitry for Balancing Sleep Need and Sleep

Abstract

Sleep-promoting neurons in the dorsal fan-shaped body (dFB) of Drosophila are integral to sleep homeostasis, but how these cells impose sleep on the organism is unknown. We report that dFB neurons communicate via inhibitory transmitters, including allatostatin-A (AstA), with interneurons connecting the superior arch with the ellipsoid body of the central complex. These "helicon cells" express the galanin receptor homolog AstA-R1, respond to visual input, gate locomotion, and are inhibited by AstA, suggesting that dFB neurons promote rest by suppressing visually guided movement. Sleep changes caused by enhanced or diminished allatostatinergic transmission from dFB neurons and by inhibition or optogenetic stimulation of helicon cells support this notion. Helicon cells provide excitation to R2 neurons of the ellipsoid body, whose activity-dependent plasticity signals rising sleep pressure to the dFB. By virtue of this autoregulatory loop, dFB-mediated inhibition interrupts processes that incur a sleep debt, allowing restorative sleep to rebalance the books. VIDEO ABSTRACT.

Keywords: Drosophila melanogaster; arousal; central complex; ellipsoid body; fan-shaped body; relaxation oscillator; sleep homeostasis; sleep pressure.

Figure 1

dFB Neurons Regulate Sleep via AstA (A) R23E10-GAL4-driven expression of the presynaptic marker ^GFPDSyd-1 (orange) and the dendritic marker DenMark (magenta) in dFB neurons. Axon terminals are concentrated in one layer of the dFB; dendrites extend into the dorsal protocerebrum. (B) AstA immunostaining (red) overlaps with the axonal projections of sleep-promoting dFB neurons expressing R23E10-GAL4-driven CD8::GFP (blue). (C) Homozygous AstA^MB10261 mutants (red) sleep less than heterozygous controls (gray) during the course of a 24-hr day (left, mean ± SEM, n = 91–113 flies per group). White and black bars denote periods of light and darkness, respectively. Two-way repeated-measures ANOVA of the hour-by-hour sleep time course detects a significant genotype × time interaction (p < 0.0001). Total sleep is reduced by ∼25% in AstA^MB10261 mutants compared with heterozygous controls (p < 0.0001, Mann-Whitney test). Circles symbolize individual flies; horizontal lines indicate group means. (D) Homozygous AstA^MB10261 mutants (red) exhibit shorter sleep bouts than heterozygous controls (gray) during the day (left, mean ± SEM; p < 0.0001, Mann-Whitney test) and night (right, mean ± SEM; p < 0.0001, Mann-Whitney test). (E) After overnight sleep deprivation for 12 hr, homozygous AstA^MB10261 mutants (red, left, mean ± SEM, n = 121–124 flies per group) and flies expressing AstA^RNAi under the control of R23E10-GAL4 (red, right, mean ± SEM, n = 51–55 flies per group) show a reduced sleep rebound relative to heterozygous AstA^MB10261 mutants or parental controls, respectively (left: p < 0.0001, Mann-Whitney test; right: p = 0.0189, Kruskal-Wallis ANOVA). The asterisk on the right denotes a significant difference from both parental controls in pairwise post hoc comparisons. (F) Expression of AstA^RNAi under the control of R23E10-GAL4 (red) reduces sleep compared with parental controls (light gray, R23E10-GAL4/+; dark gray, UAS-AstA^RNAi/+) (mean ± SEM, n = 31–34 flies per group). White and black bars denote periods of light and darkness, respectively. Two-way repeated-measures ANOVA of the hour-by-hour sleep time course detects a significant genotype × time interaction (left, p < 0.0001); one-way ANOVA detects a significant genotype effect on total sleep time (right, p < 0.0001). Circles symbolize individual flies; horizontal lines indicate group means. The asterisk denotes a significant difference from both parental controls in pairwise post hoc comparisons. (G) Overexpression of AstA under the control of R23E10-GAL4 (red) increases sleep compared with parental controls (light gray, R23E10-GAL4/+; dark gray, UAS-AstA/+) (mean ± SEM, n = 14–16 flies per group). White and black bars denote periods of light and darkness, respectively. Two-way repeated-measures ANOVA of the hour-by-hour sleep time course detects a significant genotype × time interaction (left, p < 0.0001); one-way ANOVA detects a significant genotype effect on total sleep time (right, p < 0.0001). Circles symbolize individual flies; horizontal lines indicate group means. The asterisk denotes a significant difference from both parental controls in pairwise post hoc comparisons. See also Figure S1.

Figure 2

A Mutation in AstA-R1 Reduces Sleep (A) Homozygous AstA-R1^MB07922 mutants (blue) sleep less than heterozygous controls (gray) during the course of a 24-hr day (left, mean ± SEM, n = 46–48 flies per group). White and black bars denote periods of light and darkness, respectively. Two-way repeated-measures ANOVA of the hour-by-hour sleep time course detects a significant genotype × time interaction (p < 0.0001). Total sleep is reduced by ∼25% in AstA-R1^MB07922 mutants compared with heterozygous controls (p < 0.0001, Mann-Whitney test). Circles symbolize individual flies; horizontal lines indicate group means. (B) Homozygous AstA-R1^MB07922 mutants (blue) exhibit shorter sleep bouts than heterozygous controls (gray) during the day (left, mean ± SEM; p = 0.0012, Mann-Whitney test) and night (right, mean ± SEM; p < 0.0001, Mann-Whitney test). (C) After overnight sleep deprivation for 12 hr, homozygous AstA-R1^MB07922 mutants (blue) show a reduced sleep rebound relative to heterozygous controls (gray) (mean ± SEM, n = 107–112 flies per group; p < 0.0001, Mann-Whitney test).

Figure 3

AstA-R1 Functions in Helicon Cells to Regulate Sleep (A) The AstA-R1 enhancer element in R22H05-GAL4 drives transgene expression in a small number of neurons in the brain (blue), which include four cells whose neurites contact AstA-immunopositive puncta in the dFB (red). (B) Central complex neurons labeled by R22H05-GAL4 (blue) connect the superior arch to the ellipsoid body via fibers that pass through the dFB. The connecting fibers closely adjoin sleep-promoting dFB neurons marked by R23E10-LexA (red). (C) R22H05-GAL4-driven expression of the presynaptic marker ^GFPDSyd-1 (orange) and the dendritic marker DenMark (magenta) in helicon cells. Axon terminals are concentrated in the bulb and in rings of the ellipsoid body; dendrites extend into the superior arch. (D) Expression of AstA-R1^RNAi under the control of R22H05-GAL4 (blue) reduces sleep compared with parental controls (light gray, R22H05-GAL4/+; dark gray, UAS-AstA-R1^RNAi/+) (mean ± SEM, n = 63–64 flies per group). White and black bars denote periods of light and darkness, respectively. Two-way repeated-measures ANOVA of the hour-by-hour sleep time course detects a significant genotype × time interaction (left, p < 0.0001); Kruskal-Wallis ANOVA detects a significant genotype effect on total sleep time (right, p < 0.0001). Circles symbolize individual flies; horizontal lines indicate group means. The asterisk denotes a significant difference from both parental controls in pairwise post hoc comparisons. (E) Expression of AstA-R1^RNAi under the control of R24B11-GAL4 (blue) reduces sleep compared with parental controls (light gray, R24B11-GAL4/+; dark gray, UAS-AstA-R1^RNAi/+) (mean ± SEM, n = 30–32 flies per group). White and black bars denote periods of light and darkness, respectively. Two-way repeated-measures ANOVA of the hour-by-hour sleep time course detects a significant genotype × time interaction (left, p < 0.0001); Kruskal-Wallis ANOVA detects a significant genotype effect on total sleep time (right, p < 0.0001). Circles symbolize individual flies; horizontal lines indicate group means. The asterisk denotes a significant difference from both parental controls in pairwise post hoc comparisons. (F) After overnight sleep deprivation for 12 hr, flies expressing AstA-R1^RNAi under the control of R22H05-GAL4 or R24B11-GAL4 show a reduced sleep rebound relative to parental controls (mean ± SEM, n = 59–64 flies per group; p < 0.0001, Kruskal-Wallis ANOVA). Asterisks denote significant differences from both parental controls in pairwise post hoc comparisons. See also Figures S2 and S3.

Figure 4

dFB Neurons Inhibit Helicon Cells and Their Visual Responses (A) Morphology of a single biocytin-filled helicon cell. (B) Membrane potential of the helicon cell shown in (A) during UP and DOWN states. (C) Responses of helicon cells to visual stimulation (blue bars, 1.5 s illumination at 450–490 nm). Top: membrane potential. Bottom: spike rasters of 8 helicon cells during five visual stimulation trials. Left: helicon cell responses before the activation of P2X2-expressing dFB neurons with ATP. Right: responses of the same 8 helicon cells after the activation of P2X2-expressing dFB neurons with ATP. (D) Visually evoked changes in spike frequency (left) and membrane potential baseline (right) before and after the activation of P2X2-expressing dFB neurons with ATP. Kruskal-Wallis ANOVA detects a significant difference in visually evoked spike frequency changes between groups (p = 0.0003). Asterisks indicate significant differences in planned pairwise post hoc comparisons (black brackets); gray brackets denote pairwise comparisons without significant differences. Paired t test fails to detect a significant difference in visually evoked changes in membrane potential baseline (p = 0.2045). (E) Frequency (slice angle) and probability (slice radius) of locomotor bouts as a function of helicon cell activity (n = 13 cells). Ninety-four bouts were visually evoked and 73 were self-initiated; of the self-initiated bouts, 58 occurred during UP states and 15 during DOWN states. The probability of a visual stimulus to elicit a locomotor bout was 0.7234; the probability of self-initiated movement was 0.0707/s during UP and 0.0179/s during DOWN states. Note that probabilities are plotted on a logarithmic scale. (F) Membrane potentials of helicon cells following the application of AstA (red) or a peptide with a scrambled AstA sequence (gray) or following the application of AstA to helicon cells expressing AstA-R1^RNAi under the control of R22H05-GAL4 (blue). Traces are averages of 20 peptide applications. (G) Hyperpolarizations evoked by AstA or by the activation of P2X2-expressing dFB neurons with ATP. Left: recordings from helicon cells. Right: recordings from neuroendocrine cells in the pars intercerebralis (PI). Circles symbolize average responses of individual cells to 20 peptide applications (n = 7–8 cells per condition); horizontal lines indicate group means. Kruskal-Wallis ANOVA detects a significant difference between groups (p < 0.0001); asterisks indicate significant differences from control conditions in pairwise post hoc comparisons.

Figure 5

Helicon Cells Gate Locomotion (A) Temperature-inducible expression of Kir2.1 under the control of R24B11-GAL4 increases sleep (n = 32–62 flies per group). Circles symbolize individual flies; horizontal lines indicate group means. Two-way repeated-measures ANOVA detects a significant genotype × temperature interaction (p < 0.0001); the asterisk indicates a significant difference from both parental controls in pairwise post hoc comparisons. (B) Temperature-inducible expression of Kir2.1 under the control of R24B11-GAL4 reduces the percentage of flies awakened by a visual stimulus (n = 22–23 flies per group). Circles symbolize individual flies; horizontal lines indicate group means. Two-way ANOVA detects a significant genotype × temperature interaction (p = 0.0024); the asterisk indicates a significant difference from both parental controls in pairwise post hoc comparisons. (C) Closed-loop optogenetic control of helicon cell activity. The walking speed of a fly (blue) is continuously monitored, and photostimulation (orange) is triggered after > 3 min of inactivity. Each stimulation block consists of three optical pulses that are repeated every 30 s until the next movement occurs. (D) Inactivity-triggered photostimulation decreases sleep in retinal-fed flies expressing CsChrimson under R24B11-GAL4 control (blue, n = 14 flies) relative to vehicle-treated controls (gray, n = 12 flies) (p = 0.0066, t test). Circles symbolize individual flies; horizontal lines indicate group means. (E) Retinal-fed flies expressing CsChrimson under R24B11-GAL4 control (orange, n = 14 flies) receive fewer optical stimuli than vehicle-treated controls (gray, n = 12 flies) (p = 0.0154, t test). Circles symbolize individual flies; horizontal lines indicate group means. (F and G) Locomotor activity (blue, top) and exposure to photostimulation (orange, bottom) of 12 vehicle-treated (F) and 14 retinal-fed flies (G) expressing CsChrimson under R24B11-GAL4 control. Individuals in each group are sorted in descending order of locomotor activity during photostimulation. Matching rows in the activity and photostimulation plots report simultaneously logged data from the same individual. Colored squares represent 15-min time bins. Within each bin, the percentages of time spent moving or exposed to photostimulation are color-coded according to the look-up tables on the right. The 24-hr experimental period is preceded by a night of baseline sleep. Stimulation light pulses notwithstanding, the animals were raised and kept in constant darkness.

Figure 6

Helicon Cells Excite R2 Neurons of the Ellipsoid Body (A and B) Anterior (A) and dorsal (B) views of helicon cells labeled by R78A01-GAL4 (blue) and R2 neurons marked by R48H04-LexA (yellow). (C) Membrane potentials of R2 neurons during optogenetic stimulation of helicon cell activity (orange bars, 2 s illumination at 630 nm). Top: membrane potential traces (left) and changes in membrane potential baseline (right) in the presence and absence of retinal or in the combined presence of retinal and 1 μM tetrodotoxin (TTX). The lack of retinal or the presence of TTX blocks the optogenetically induced depolarization (n = 6–13 cells per group; p = 0.0002, Kruskal-Wallis ANOVA). Asterisks indicate significant differences from control conditions in pairwise post hoc comparisons. (D) EPSP rasters (left) and EPSP frequency modulation (right) of five R2 neurons during five optogenetic stimulation trials of helicon cell activity in retinal-fed flies. One-way repeated-measures ANOVA detects a significant effect of photostimulation (p < 0.0001). (E) EPSP rasters (left) and EPSP frequency modulation (right) of five R2 neurons during five optogenetic stimulation trials of helicon cell activity in vehicle-treated flies. One-way repeated-measures ANOVA fails to detect a significant effect of photostimulation (p = 0.8376). (F) Spike raster (left) and membrane potential (right) of an R2 neuron during optogenetic stimulation of helicon cell activity (orange bars; 50 ms illumination at 630 nm). Each of 9 consecutive light pulses elicits an action potential (AP). See also Figures S3 and S4.

Figure 7

Intense Helicon Cell Activation Induces Rebound Sleep (A) Locomotor activity (blue) of 12 vehicle-treated (top) and 12 retinal-fed flies (bottom) expressing CsChrimson under R24B11-GAL4 control during 24 hr of photostimulation at 20 Hz and a 24-hr recovery period immediately afterward. Individuals in each group are sorted in descending order of inactivity during the recovery period. Colored squares represent 15-min time bins. Within each bin, the percentage of time spent moving is color-coded according to the look-up table on the right. Stimulation light pulses notwithstanding, the animals were raised and kept in constant darkness. (B) Photostimulation generates rebound sleep in retinal-fed flies expressing CsChrimson under R24B11-GAL4 control (blue, n = 12 flies) but not in vehicle-treated controls (gray, n = 12 flies) (p < 0.0001, one-way ANOVA). Circles symbolize individual flies; horizontal lines indicate group means. Asterisks indicate significant differences in planned pairwise post hoc comparisons (black brackets); gray brackets denote pairwise comparisons without significant differences.

Figure 8

The Sleep Homeostat as a Relaxation Oscillator Helicon cells respond to visual input and play a permissive role in locomotion, either by virtue of their excitatory synapses with R2 neurons or through other pathways. R2 neuron activity generates sleep pressure that is communicated to dFB neurons via currently unidentified synaptic connections or non-synaptic mechanisms. The activation of dFB neurons during sleep inhibits helicon cells and, thus, impedes the flow of visual signals to R2 neurons; this raises the visual sensory threshold, blocks locomotion, and reverses the build-up of sleep pressure due to R2 neuron activity, which is driven, in part, by excitation from helicon cells. Because dFB neurons switch between electrical activity and silence, the sleep homeostat functions as a relaxation oscillator akin to the electrical circuit on the right. Here, a capacitor (C) is charged through a resistor (R) and discharged through a neon bulb (Ne) when the voltage across the capacitor exceeds the ignition threshold of the bulb. Common to the biological and electrical circuits is the conversion of a continuous process (changes in sleep pressure or voltage) into binary state changes (an organism that is asleep or awake; a bulb that is lit or dark).