GABA_B receptors play an essential role in maintaining sleep during the second half of the night in Drosophila melanogaster

The fruit fly Drosophila melanogaster has become a well-accepted model for sleep research (reviewed by Cirelli, 2009; Minot et al., 2011). As in mammals, it has been shown that the sleep-like state of Drosophila is associated with reduced sensory responsiveness and reduced brain activity (Nitz et al., 2002; van Swinderen et al., 2004), and is subject to both circadian and homeostatic regulation (Hendricks et al., 2000; Shaw et al., 2000). Similarly to in humans, monaminergic neurons (specifically dopaminergic and octopaminergic neurons) enhance arousal in fruit flies (Andretic et al., 2005; Kume et al., 2005; Lebestky et al., 2009; Crocker et al., 2010), whereas GABAergic neurons promote sleep (Agosto et al., 2008). As in humans, GABA advances sleep onset (reduces sleep latency) and prolongs total sleep (increases sleep maintenance) (Agosto et al., 2008). Brain regions possibly implicated in the regulation of sleep in D. melanogaster are the pars intercerebralis (Foltenyi et al., 2007; Crocker et al., 2010), the mushroom bodies (Joiner et al., 2006; Pitman et al., 2006; Yuan et al., 2006) and a subgroup of the pigment dispersing factor (PDF)-positive neurons called the l-LN_v neurons (Parisky et al., 2008; Sheeba et al., 2008a; Chung et al., 2009; Lebestky et al., 2009; Shang et al., 2011). The l-LN_v belong to the circadian clock neurons, indicating that in flies, as in mammals, the sleep circuit is intimately linked to the circadian clock and that the mechanisms employed to govern sleep in the brain are evolutionarily ancient.

The l-LN_v are conspicuous clock neurons with wide arborisations in the optic lobe, fibres in the accessory medulla – the insect clock centre – and connections between the brain hemispheres (Helfrich-Förster et al., 2007a). Thus, the l-LN_v neurons are anatomically well suited to modulate the activity of many neurons. In addition, their arborisations overlap with those of monaminergic neurons (Hamasaka and Nässel, 2006). Several studies show that they indeed receive dopaminergic, octopaminergic and GABAergic input and that they control the flies' arousal and sleep (Agosto et al., 2008; Parisky et al., 2008; Kula-Eversole et al., 2010; Shang et al., 2011). Furthermore, the l-LN_v are directly light sensitive and promote arousal and activity in response to light, especially in the morning (Shang et al., 2008; Sheeba et al., 2008a; Sheeba et al., 2008b; Fogle et al., 2011).

A part of the sleep-promoting effect of GABA on the l-LN_v has been shown to be mediated via the fast ionotropic GABA_A receptor Rdl (Resistance to dieldrin) (Agosto et al., 2008). Rdl Cl⁻ channels are expressed in the l-LN_v (Agosto et al., 2008) and, similar to mammalian GABA_A receptors, they mediate fast inhibitory neurotransmission (Lee et al., 2003). As expected, GABA application reduced the action potential firing rate in the l-LN_v, whereas application of picrotoxin, a GABA_A receptor antagonist, increased it (McCarthy et al., 2011). Furthermore, an Rdl receptor mutant with prolonged channel opening and consequently increased channel current significantly decreased sleep latency of the flies after lights-off, whereas the downregulation of the Rdl receptor via RNAi increased it (Agosto et al., 2008).

Nevertheless, the manipulation of the Rdl receptor had no effect on sleep maintenance. Because the latter is significantly reduced after silencing the GABAergic neurons (Parisky et al., 2008), other GABA receptors must be responsible for maintaining sleep. Suitable candidates are slow metabotropic GABA_B receptors that are often co-localised with ionotropic GABA_A receptors (Enell et al., 2007). In Drosophila, like in mammals, the metabotropic GABA_B receptors are G-protein-coupled seven-transmembrane proteins composed of two subunits, GABA_B-R1 and GABA_B-R2 (Kaupmann et al., 1998; Mezler et al., 2001). The GABA_B-R1 is the ligand binding unit and GABA_B-R2 is required for translocation to the cell membrane and for stronger coupling to the G-protein (Kaupmann et al., 1998; Galvez et al., 2001). In this study we show that the l-LN_v do express metabotropic GABA_B-R2 receptors and that these receptors are relevant for sleep maintenance but not for sleep latency. Thus, metabotropic and ionotropic GABA receptors are cooperating in sleep regulation.

Oregon R was used as a wild-type strain for GABA_B-R2, PDF and GAD1 immunohistochemistry. For visualizing GABA_B-R2 receptors we also used a GABA_B-R2-GAL4 line (Root et al., 2008) (kindly provided by Jing Wang, University of California, San Diego) to express green fluorescent protein (GFP) with the binary UAS-GAL4 system (using UAS-s65tGFP, stock 1522, Bloomington Stock Center, Bloomington, IN, USA). The specificity of the GABA_B-R2-GAL4 line for GABA_B-R2-expressing neurons has been demonstrated previously for the adult olfactory sensory neurons (OSNs) (Root et al., 2008) and seems to be correct also for most the other labelled neurons, judging from a wide overlap between immunostaining with a GABA_B-R2 antiserum and GABA_B-R2-GAL4-driven GFP (Hamasaka et al., 2005). In order to downregulate the GABA_B-R2 receptor specifically in the PDF neurons (s-LN_v and l-LN_v), we used Pdf-GAL4 (Park et al., 2000) to either express UAS-GABA_B-R2-RNAi (Root et al., 2008) (provided by Jing Wang) alone, or to simultaneously express UAS-GABA_B-R2-RNAi and UAS-Dicer2 (no. 60012, Vienna Drosophila RNAi Center, Wien, Austria). In the first experiment, the Pdf-GAL4 driver and UAS-GABA_B-R2-RNAi effector lines crossed to white¹¹¹⁸ were taken as controls, and in the second experiment, the Pdf-GAL4 driver and UAS-GABA_B-R2-RNAi effector lines crossed to UAS-Dicer2 were taken as controls. In a second set of experiments, we drove an independent UAS-Trip-GABA_B-R2-RNAi line (no. 27699; Bloomington Stock Center) together with dicer2 under control of Pdf-GAL4 in order to downregulate GABA_B-R2. All flies were raised on Drosophila food (0.8% agar, 2.2% sugar-beet syrup, 8.0% malt extract, 1.8% yeast, 1.0% soy flour, 8.0% corn flour and 0.3% hydroxybenzoic acid) at 25°C under a 12 h:12 h light:dark (LD) cycle and transferred to 20°C at an age of ~3 days.

For PDF and GFP co-labelling, whole-mount brains were dissected in PBS-TX (0.01 mol l⁻¹ phosphate-buffered saline with 0.5% Triton X-100, pH 7.2) before fixing them in ice-cold 4% paraformaldehyde (in 0.1 mol l⁻¹ phosphate buffer, pH 7.2) for 1 h. For GABA_B-R2 immunolabelling, entire fly heads were fixed for 3 h prior to brain dissection. The fixed and dissected brains were washed in PBS-TX, blocked for 2 h in 5% normal goat serum (NGS, in PBS-TX) and then incubated with primary antibodies at 4°C overnight or for 48 h. Primary antibodies were used in the following concentrations in PBS-TX with 5% NGS: anti-PDF 1:2500 and anti-GABA_B-R2 1:10,000 (for description of primary antibodies, see below). After washing several times in PBS-TX, the brains were incubated in secondary fluorochrome-conjugated antibodies in PBS-TX with 5% NGS either overnight at 4°C or for 2 h at room temperature. We used Alexa Fluor antibodies (Molecular Probes, Carlsbad, CA, USA) of 488 nm (goat anti-mouse) or 546 nm (goat anti-rabbit) and Cy2- or Cy3-tagged IgGs (goat anti-mouse or anti-rabbit; Jackson ImmunoResearch, West Grove, PA, USA) at a dilution of 1:1000. After incubation with the secondary antibodies, brains were washed in PB-TX and subsequently mounted in Vectashield medium (Vector Laboratories, Burlingame, CA, USA) on glass slides, all in the same orientation.

Production and characterization of antiserum (code B7873/3) to GABA_B-R2 was described previously (Hamasaka et al., 2005). In western blots of brain tissue from Drosophila the antiserum detected a band of appropriate size (Hamasaka et al., 2005), and in GABA_B-R2 immunohistochemistry on Drosophila brains the antiserum labelled neurons that were in close proximity to neural elements immunopositive for a vesicular GABA transporter, GAD and a GABA_A receptor (RDL), strongly indicating that the GABA_B-R2 antiserum is reliably detecting a GABA_B receptor (Enell et al., 2007). The anti-GABA_B-R2 specificity was further established by diminished GABA_B-R2 immunolabelling after GABA_B-R2 RNAi in specific neurons (OSNs) in the antenna/antennal lobe (Root et al., 2008). Furthermore, physiological/pharmacological evidence (using Ca²⁺ imaging) for GABA_B-R2 expression has been produced for the larval s-LN_v and the adult OSN (Hamasaka et al., 2005; Root et al., 2008). Recently, physiological evidence for metabotropic GABA signalling was also provided for the adult s-LN_v using cAMP imaging (Lelito and Shafer, 2012).

The mouse anti-PDF was provided by Justin Blau (obtained from the Developmental Studies Hybridoma Bank).

We performed laser scanning confocal microscopy (Leica TCS SPE, Leica, Wetzlar, Germany) to analyse immunofluorescent brains. To avoid bleeding through, we used sequential scanning (laser lines 488 and 532 nm). Confocal stacks of 1.5 μm thickness were acquired. For quantifying intensity of GABA_B-R2 immunostaining, the settings of the confocal microscope (laser power, gain and contrast) were kept the same for all preparations. The staining intensity of GABA_B-R2 labelling was then quantified on single confocal images by two different methods. To compare labelling intensity between GABA_B-R2 immunostaining and GABA_B-R2 GFP reporter staining, we coded the images and judged by eye whether the intensity in each single PDF-positive neuron was strong, weak or completely absent. Subsequently, we calculated the percentage of l-LN_v and s-LN_v with strong, weak and no staining (Fig. 1J,K). To judge the degree of GABA_B-R2 downregulation in the l-LN_v and s-LN_v by RNAi, entire confocal stacks were imported into ImageJ (Fiji distribution; http://fiji.sc/wiki/index.php/Fiji or http://rsb.info.nih.gov/ij/). Individual somata of l-LN_v and s-LN_v were selected using the magic wand selection tool on single confocal stacks. The average staining intensity of the selected area was calculated (grey value 0: no staining; grey value 255: maximal staining), and the same was done for the neighbouring area outside the neurons to obtain the staining intensity of the background. The background staining was subtracted from the value gained for the soma. An average staining intensity was calculated for all l-LN_v and s-LN_v of one brain hemisphere, and finally the values of all 11–13 brain hemispheres were averaged. After testing for normal distribution, a one-way ANOVA was used to test for significant staining differences between control flies and flies in which GABA_B-R2 was downregulated.

The locomotor activity of single male control flies and flies with downregulated GABA_B-R2 was recorded using the TriKinetics DAM2 System (TriKinetics, Waltham, MA, USA). Monitors were put in light-tight boxes (recording units) with white light LED illumination of 47.6 μW cm⁻² (Lumitronix LED-Rechnik GmbH, Jungingen, Germany) during the light phase. These recording units were placed in a climate chamber with a constant temperature of 20±0.5°C and 60±1% relative humidity. Flies were exposed to 12 h:12 h LD cycles for 11 days and subsequently kept in constant darkness (DD) for another 10 days. Only male flies with an age of 3 to 6 days were taken for the experiments. Flies that died during the experiment were excluded from the calculations.

Raw data actograms were created using ActogramJ for ImageJ (Schmid et al., 2011). The average activity levels of all individual flies were calculated for the entire LD and DD period. To do so, the beam crosses per minute from days 2–11 (LD) and days 12–21 (DD) were first averaged for each single fly and then the activity values of all flies of one genotype were averaged. This calculation was possible during DD, because the periods of the individual flies were close to 24 h. During the LD period, average activity levels were additionally determined during the first and second halves of the day and the night, respectively, to be compared with the sleep values during the same time periods (see below). To see whether GABA_B-R2 downregulation affects the speed of the circadian clock, we determined the free-running periods of all individual flies during days 12 to 21 in DD by periodogram analysis (Sokolove and Bushell, 1978) combined with a chi-square test with a 5% significance level (Schmid et al., 2011).

The sleep analysis was only performed during the LD cycles (from day 2 to day 11). Calculations of total sleep were performed for each hour of the day using a macro written in Microsoft Excel 2007. Average sleep bout durations were calculated for 6 h intervals using the very same macro. In previous studies sleep was defined as period of inactivity longer than 5 min (Hendricks et al., 2000; Shaw et al., 2000; Andretic and Shaw, 2005; Ho and Sehgal, 2005). At first we used the same criterion, but then we switched to a more stringent one (a period of inactivity longer than 10 min as definition of sleep) in order to make very sure that periods of inactivity that resemble rest rather than sleep are not included. An average sleeping curve was calculated for each genotype out of the hourly sleep duration determined for each individual fly. Furthermore, average total sleep and average sleep bout duration were calculated in 6-h intervals for each genotype for the first and second halves of the day and night, respectively. Sleep latency after lights-off was determined as the time in minutes after lights off until the first sleep bout of at least 10 min occurred. To do so, we imported the TriKinetics file into Excel and searched manually for the first occurrence of 10 zeros in a row after lights-off for each individual fly.

Previous studies have shown that there are large numbers of GABA-producing neurons in the adult brain of Drosophila (Enell et al., 2007; Okada et al., 2009). The processes of some of these arborise in the region of the accessory medulla that contains presumed dendrites of the PDF neurons (Hamasaka et al., 2005). Thus GABAergic neurons may converge on PDF neurons.

To identify whether PDF neurons express metabotropic GABA_B receptors, we used two different markers: we expressed GFP with a GABA_B-R2-specific GAL4 driver line (Root et al., 2008) and we used an antiserum raised against the GABA_B-R2 protein (Hamasaka et al., 2005). In both cases we co-stained with anti-PDF to judge whether the two signals overlap (Fig. 1A–I). We found that the PDF-positive l-LN_v neurons are reliably marked by both methods. GABA_B-R2-GAL4-driven GFP was present in 98.8% of the l-LN_v labelled by anti-PDF (Fig. 1J) and in 100% of the l-LN_v that were labelled with the GABA_B-R2 antiserum (Fig. 1K). Whereas the GABA_B-R2-antibody labelling was similarly strong in all four l-LN_v neurons (Fig. 1G), GFP was strongly expressed in only two to three of the four l-LN_v neurons and weak in the remaining one to two cells (Fig. 1A,D). The PDF-positive s-LN_v were not reliably stained by GABA_B-R2-GAL4-driven GFP (Fig. 1A): only 39.7% of the s-LN_v showed a GFP signal at all and just 2.7% of them revealed a strong signal (Fig. 1J). Nevertheless, the GABA_B-R2 antibody revealed the s-LN_v reliably (Fig. 1G,K): 100% of the s-LN_v showed a prominent punctuate staining. We conclude that the GABA_B-R2 receptor is expressed on all PDF neurons (the l-LN_v and the s-LN_v), but that GABA_B-R2-GAL4 drives only weakly in the s-LN_v.

To investigate whether GABA_B receptors on the PDF neurons might be relevant for locomotor activity rhythms and the general activity level, we downregulated the receptors by expressing two different constructs of UAS-RNAi for GABA_B-R2 (with or without dicer2) in the PDF neurons (using Pdf-GAL4). First, we measured the intensity of immunolabelling in the PDF neurons in comparison to control flies (that carried only the RNAi construct). We observed a significant reduction in staining intensity only for one of the RNAi constructs that had previously been employed successfully (Root et al., 2008), and only if it was combined with dicer2. Furthermore, the knock-down was only significant in the l-LN_v but not in the s-LN_v neurons (Fig. 2A). The latter observation is in agreement with our findings that GABA_B-R2 antibody staining revealed stronger signals in the s-LN_v than in the l-LN_v and that Pdf-GAL4 usually drives less strongly in the s-LN_v as compared with the l-LN_v (Renn et al., 1999).

After we had made sure that the GABA_B-R2 downregulation is preferentially working in the l-LN_v, we monitored locomotor activity of the flies for 11 days in 12 h:12 h LD followed by 10 days of DD, both at 20°C. We found that the flies with significantly downregulated GABA_B-R2 in the l-LN_v were significantly more active than the control flies, and this was true during both LD and DD (Fig. 2B,C). No significant differences in the activity level were present between controls and RNAi flies in the two lines, in which no GABA_B-R2 knock-down was detectable by antibody staining (data not shown). The activity pattern of the flies was in principle similar for all strains, showing bimodal patterns with activity bouts in the morning and evening under LD conditions and more unimodal patterns under DD conditions. GABA_B-R2 downregulation had no effect on the activity rhythms under DD conditions; the great majority of flies (86–100%) were rhythmic, free-ran with a period of approximately 24 h (Fig. 2D), which was of wild-type-like power (Fig. 2E).

Next, we analysed the effects of GABA_B-R2 downregulation on sleep. Again, we did not see any significant differences in sleep between controls and RNAi flies of the two lines, in which the GABA_B-R2 knock-down appeared inefficient (data not shown). But, we found that the flies with significantly downregulated GABA_B-R2 receptors in the l-LN_v showed a largely reduced amount of night-time sleep during LD (Fig. 3). This difference was mainly due to a reduced amount of sleep in the second half of the night (Fig. 3A). During the 6-h interval in the late night, the experimental flies showed 30 min less sleep in total as compared with the controls (Fig. 3C). Simultaneously, they decreased their average sleep bout duration significantly (to 108 min in the 6-h interval, in contrast to 135 and 160 min, respectively, of the control strains) (Fig. 3D). Obviously, the flies with downregulated GABA_B-R2 receptors were clearly unable to maintain sleep in the second half of the night. They woke up early and displayed a high amount of locomotor activity. This is also clearly visible in the actograms of the flies (Fig. 2C).

No effects on sleep were visible during the day or the first half of the night. Downregulation of GABA_B-R2 receptors in the l-LN_v did also not increase sleep latency. After lights-off, the experimental flies fell asleep as fast as flies of Control 2 (UAS-dicer2;Pdf-GAL4/+) (Fig. 3B). Just the flies of Control 1 (UAS-dicer2;UAS-GABA_B-R2-RNAi) fell asleep faster.

As stated above, the expression of the GABA_B-R2 RNAi in the PDF neurons also influenced the overall activity level of the flies, but in contrast to sleep, activity level was increased throughout the entire 24-h day (all four 6-h intervals; Fig. 3E). Thus the regulation of the overall activity level cannot be by the same mechanism as sleep regulation.

Here we show that metabotropic GABA_B-R2 receptors are expressed on the PDF-positive clock neurons (LN_v neurons), and that their downregulation in the l-LN_v by RNAi results in: (1) a higher activity level throughout the day and night and (2) reduced sleep maintenance in the second half of the night. Neither sleep onset nor circadian rhythm parameters were affected by the downregulation. We conclude that GABA signalling via metabotropic receptors on the l-LN_v is essential for sustaining sleep throughout the night and for keeping activity at moderate levels throughout the 24-h day (preventing flies from hyperactivity). A major caveat of RNAi is off-target effects, particularly when Gal4 drivers are expressed in large numbers of non-target neurons. Though we have downregulated GABA_B-R2 in only eight neurons per brain hemisphere and were careful to correlate the behavioural effects of our knockdown experiments with observation and measures of GABA_B-R2 immunostaining in the s-LN_v and l-LN_v, it is still possible that some effects were due to off-target knockdown of other membrane proteins. Nevertheless, given the fact that no such effects have been reported in the previous paper that used the same GABA_B-R2 RNAi line (Root et al., 2008), we think it is unlikely that the behavioural effects described here were due to off-target knockdown of other genes.

Our results are in line with a former study describing the location of GABA_B receptors in D. melanogaster (Hamasaka et al., 2005). The ionotropic GABA_A receptor Rdl has also been identified on the l-LN_v neurons and has been shown to regulate sleep, but its downregulation delayed only sleep onset and did not perturb sleep maintenance (Parisky et al., 2008). In contrast, silencing GABAergic signalling influenced sleep onset and sleep maintenance, indicating that GABA works through the fast Rdl receptor, and also implying a longer-lasting signalling pathway. GABA_B receptors are perfect candidates in mediating slow but longer-lasting effects of GABA. Often, GABA_A and GABA_B receptors cooperate in mediating such fast and slow effects. For example, in the olfactory system, GABA_A receptors mediate the primary modulatory responses to odours whereas GABA_B receptors are responsible for long-lasting effects (Wilson and Laurent, 2005).

In D. melanogaster, GABA_B receptors consist of the two subunits GABA_B-R1 and GABA_B-R2, and only the two units together can efficiently activate the metabotropic GABA signalling cascade (Galvez et al., 2001; Mezler et al., 2001). In our experiments, we downregulated only GABA_B-R2, but this manipulation should also have decreased the amount of functional GABA_B-R1/GABA_B-R2 heterodimers and, therefore, reduced GABA_B signalling in general. Taking into account that sleep maintenance in the second half of the night was already significantly impaired by an ~46% reduction in detectable GABA_B-R2 immunostaining intensity in the l-LN_v clock neurons, it can be assumed that GABA_B receptors account for an even larger portion of the sleep maintenance than detected in our experiments. Thus GABA_B receptors play a crucial role in mediating GABAergic signals to the l-LN_v neurons, which are needed to sustain sleep throughout the night. This is mainly due to the maintenance of extended sleep bout durations in the second half of the night. When signalling by the GABA_B receptor is reduced, sleep bouts during this interval are significantly shortened, leading to less total sleep.

Most importantly, we confirm the l-LN_v as important components in regulation of sleep and arousal (Agosto et al., 2008; Parisky et al., 2008; Kula-Eversole et al., 2010; Shang et al., 2011). In contrast, the s-LN_v seem to be not involved in sleep–arousal regulation but are rather important for maintaining circadian rhythmicity under DD (reviewed by Helfrich-Förster et al., 2007b). One caveat in clearly distinguishing the function of s-LN_v and l-LN_v is the fact that both cell clusters express the neuropeptide PDF and, as a consequence, Pdf-GAL4 drives expression in both subsets of clock neurons. Though we did not see a significant GABA_B-R2 knock-down in the s-LN_v, we cannot completely exclude that GABA_B-R2 was slightly downregulated in these clock neurons and that this knock-down contributed to the observed alterations in sleep. To restrict the knock-down to the s-LN_v we used the R6-GAL4 line that is expressed in the s-LN_v but not in the l-LN_v (Helfrich-Förster et al., 2007a). We observed neither a reduction in GABA_B-R2 staining intensity in the s-LN_v nor any effects on sleep in the second half of the night (F.G. and C.H.-F., unpublished observations). The lack of any visible GABA_B-R2 downregulation in the s-LN_v with R6-GAL4 is in agreement with observations of Shafer and Taghert (Shafer and Taghert, 2009), who could completely downregulate PDF in the s-LN_v using Pdf-GAL4 but not using R6-GAL4. Thus, R6-GAL4 is a weaker driver than Pdf-GAL4 and is obviously not able to influence GABA_B-R2 in the s-LN_v. Nevertheless, in our experiments the R6-Gal4-driven GABA_B-R2 RNAi led to flies that had slightly higher diurnal activity levels and less diurnal rest than the control flies (F.G. and C.H.-F., unpublished observations; supplementary material Fig. S1). This suggests that GABA_B-R2 was downregulated somewhere else. When checking the R6-GAL4 expression more carefully we found that it was not restricted to the brain, but was also present in many cells of the thoracic and especially the abdominal ganglia (F.G. and C.H.-F., unpublished observations). Given the broad expression of GABA_B-R2, a putative knock-down in the ventral nervous system is likely to affect locomotor activity.

Our results on the l-LN_v certainly do not exclude a role of GABA in the circadian clock controlling activity rhythms under DD conditions (here represented by the s-LN_v). In mammals, GABA is the most abundant neurotransmitter in the circadian clock centre in the brain – the suprachiasmatic nucleus (van den Pol and Tsujimoto, 1985). GABA interacts with GABA_A and GABA_B receptors, producing primarily but not exclusively inhibitory responses through membrane hyperpolarisation (Wagner et al., 1997; Choi et al., 2008). GABA signalling is important for maintaining behavioural circadian rhythmicity, it affects the amplitude of molecular oscillations and might contribute to synchronisation of clock cells within the suprachiasmatic nucleus (Liu and Reppert, 2000; Albus et al., 2005; Aton et al., 2006; Ehlen et al., 2006). The same seems to be true for fruit flies. The s-LN_v neurons of adults alter cAMP levels upon GABA application on isolated brains in vitro (Lelito and Shafer, 2012). Hyperexcitation of GABAergic neurons disrupts the molecular rhythms in the s-LN_v and renders the flies arrhythmic (Dahdal et al., 2010). Thus, GABA signalling affects the circadian clock in the s-LN_v. We found that flies with downregulated GABA_B-R2 receptors had slightly longer free-running periods than the control flies, but this turned out to be only significant in comparison with Control 2 and not to Control 1 (Fig. 2D). Dahdal et al. (Dahdal et al., 2010) found similar small effects on period after downregulating GABA_B-R2 receptors, but a significant period lengthening after downregulating GABA_B-R3 receptors. This indicates that GABA signals via GABA_B-R3 receptors to the s-LN_v and was confirmed in vitro in the larval Drosophila brain by Ca²⁺ imaging (Dahdal et al., 2010). Nevertheless, the study of Dahdal et al. (Dahdal et al., 2010) does not rule out that GABA signals via GABA_B-R3 plus GABA_B-R2 receptors on the adult s-LN_v. We found a rather strong expression of GABA_B-R2 receptors in these clock neurons, and were not able to downregulate it significantly by RNAi, although we used dicer2 as amplification. Dahdal et al. (Dahdal et al., 2010) did not use dicer2, and they also did not measure the effectiveness of the downregulation of GABA_B-R2 by RNAi immunocytochemically directly in the s-LN_v. Thus, the exact GABA_B receptors that mediate GABA responses in the adult s-LN_v need still to be determined.

In summary, we conclude that the l-LN_v subgroup of the PDF-positive clock neurons is a principal target of sleep-promoting and activity-repressing GABAergic neurons and sits at the heart of the sleep circuit in D. melanogaster. Thus, the sleep circuitry of flies is clearly more circumscribed and simpler than that of mammals. Mammals have many targets of sleep-promoting GABAergic neurons, and the circadian clock seems to have a mainly modulatory and less direct influence on sleep (Mistlberger, 2005). The fly sleep circuitry may therefore have condensed the mammalian arousal and sleep stimulating systems (e.g. monaminergic, cholinergic, peptidergic and GABAergic systems) into a simpler and more compact region, which seems to largely coincide with the eight PDF-positive l-LN_v cells of the circadian circuit.

We thank Justin Blau, Jae Park and Jim Wang for supplying fly strains and antibodies.