Neuromodulators signal through astrocytes to alter neural circuit activity and behavior

Summary

Astrocytes associate with synapses throughout the brain and express receptors for neurotransmitters that can elevate intracellular calcium (Ca²⁺) ^1-3. Astrocyte Ca²⁺ signaling has been proposed to modulate neural circuit activity ⁴, but pathways regulating these events are poorly defined and in vivo evidence linking changes in astrocyte Ca²⁺ to alterations in neurotransmission or behaviors is limited. Here we show Drosophila astrocytes exhibit activity-regulated Ca²⁺ signaling events in vivo. Tyramine (Tyr) and octopamine (Oct) released from Tdc2⁺ neurons signal directly to astrocytes to stimulate Ca²⁺ increases through the octopamine-tyramine receptor (Oct-TyrR) and the TRP channel Waterwitch (Wtrw), and astrocytes in turn modulate downstream dopaminergic (DA) neurons. Tyr or Oct application to live preparations silenced dopaminergic (DA) neurons and this inhibition required astrocytic Oct-TyrR and Wtrw. Increasing astrocyte Ca²⁺ signaling was sufficient to silence DA neuron activity, which was mediated by astrocyte endocytic function and adenosine receptors. Selective disruption of Oct-TyrR or Wtrw expression in astrocytes blocked astrocyte Ca²⁺ signaling and profoundly altered olfactory-driven chemotaxis behavior and touch-induced startle responses. Our work identifies Oct-TyrR and Wtrw as key components of the astrocyte Ca²⁺ signaling machinery, provides direct evidence that Oct- and Tyr-based neuromodulation can be mediated by astrocytes, and demonstrates that astrocytes are essential for multiple sensory-driven behaviors.

Astrocytes are intimately associated with brain synapses and positioned to broadly regulate synaptic activity. It has been widely proposed that astrocytes modulate neural circuits and behavior ^5-7, although in vivo data demonstrating astrocytes are directly activated by neurotransmission and signal back to neurons to modulate output remains elusive. Astrocytes exhibit dynamic fluctuations in intracellular Ca²⁺ in vitro ^8,9 and in vivo ^10,11 suggesting Ca²⁺ signaling might be a useful measure of astrocyte “activity”. Despite decades of studies on astrocytic Ca²⁺ transients, signaling pathways controlling these transients remain poorly defined, and their in vivo relevance remains controversial. If astrocytes indeed actively participate in information processing in circuits, it is imperative we characterize such mechanisms as they would represent a potentially widespread mechanism for controlling brain function.

We performed an RNAi-based screen in Drosophila where we individually knocked-down ~500 Ca²⁺ signaling-related genes selectively in astrocytes using the astrocyte-specific Gal4 driver line alrm-Gal4 ¹², and assayed larval olfactory-driven chemotaxis behavior (Extended Data Fig. 1a). We found astrocyte knockdown of the TRP channel Water witch (Wtrw)¹³ led to a significant (~50%) decrease in larval chemotaxis toward iso-amyl acetate (IAA)(Fig. 1a). Similar results were found with a second, non-overlapping RNAi construct (Fig. 1a) and was specific to glia as shown by insensitivity to blockade of Gal4/UAS in neurons with elav-Gal80 (Extended Data Fig. 1e). The null allele wtrw^ex was equally defective in chemotaxis, which was rescued by re-expression of wtrw only in astrocytes (Fig. 1a). We also found that blockade of Gal4/UAS driven wtrw^RNAi with tsh-Gal80 in the ventral nerve cord (VNC) resulted in normal chemotaxis responses (Extended Data Fig. 1f), revealing a critical role for VNC astrocytes in this behavior, although this does not exclude an additional role for brain astrocytes. Chemotaxis defects did not result from simple alterations in motility, as wtrw^RNAi animals exhibited normal locomotion (Extended Data Fig. 1g) and light avoidance responses (Extended Data Fig. 1c,h). We conclude that Wtrw is required in astrocytes for normal larval olfactory-driven behavior.

Figure 1. Larval chemotaxis and startle-induced responses require the astrocyte-expressed TRP channel Water witch.

a, Chemotaxis assay (n=12). b, Gentle touch assay (n=30). c, Pseudocolored maximum intensity projections of 15min movies, averaged traces of 16 individual astrocytes and quantifications of the frequency of somatic Ca²⁺ transients (n=10, 160 cells total). Scale bar, 50μm. d, Responses of astrocytes to neurotransmitters/neuromodulators in the presence of tetrodotoxin (n=6, 96 cells total). *p<0.05, **p<0.01, n.s., not significant, Error bars, s.e.m. Wilcoxon and Mann-Whitney tests followed by Bonferroni-Holm post hoc test (a). One-way ANOVA followed by Tukey's post hoc test (b,c,d).

Ca²⁺ transients in mammalian astrocytes in awake behaving mice are induced during periods of elevated arousal ^14-16. We therefore used a gentle anterior touch assay¹⁷ to examine the role of astrocyte Ca²⁺ signaling in larval startle-induced behaviors. Crawling larvae touched anteriorly with a hair responded by pausing and continuing forward (Type I response), or by moving backward and executing an escape response (Type II response) (Extended Data Fig. 1b). Control animals exhibited roughly equal frequencies of Type I and Type II responses, however, wtrw^RNAi in astrocytes resulted in a dramatic alteration in behavior: ~80% of larvae exhibited Type I responses, a phenomenon mimicked by wtrw^ex mutants (Fig. 1b) and the phenotype is independent of neurons (Extended Data Fig. 1i). These data indicate astrocyte expressed Wtrw also modulates startle-induced behavioral changes in Drosophila larvae.

To explore in vivo Ca²⁺ signaling we developed a semi-dissected preparation to image the larval central nervous system (CNS). GCaMP6s was used to image astrocyte cytosolic Ca²⁺ changes (UAS-GCaMP6s), mCherry (UAS-mCherry) was used as a reference for astrocyte position, and dorsal astrocyte cell bodies were imaged in the VNC. We found astrocyte cell bodies exhibited coordinated, population-wide slow oscillations (termed somatic Ca²⁺ transients) (Extended Data Fig. 1j-I; Supplementary video 1). Somatic Ca²⁺ transients occurred approximately every 2 minutes and exhibited an average ~11% change in dF/F (Extended Data Fig. 1m). Interestingly, blocking neuronal activity with tetrodotoxin (TTX) suppressed transients by approximately 60%, as did application of the broad Ca²⁺ channel blocker lanthanum chloride (LaCl₃)(Extended Data Fig. 1n). Similar astrocyte Ca²⁺ transients were observed when we imaged intact immobilized larvae (Extended Data Fig. 1o), indicating our dissected preparation preserves in vivo patterns of astrocyte activity.

TRP channels regulate Ca²⁺ levels in astrocytes ¹⁸, we therefore reasoned Wtrw might drive Ca²⁺ signaling in Drosophila astrocytes. Control larvae exhibited 8-9 rhythmic oscillations in somatic Ca²⁺ transients over 15 minutes. In contrast, astrocyte-specific wtrw^RNAi led to a roughly 50% decrease in somatic astrocyte Ca²⁺ transients, which was also observed in the wtrw^ex mutant (Fig. 1c). Bath application of acetylcholine, glutamate, and γ-aminobutyric acid in the presence of TTX did not elicit a change in Ca²⁺ levels in astrocytes. In contrast, application of tyramine (Tyr) or octopamine (Oct), the invertebrate analogues of norepinephrine, which has been shown to induce Ca²⁺ transients in mammalian astrocytes ^14,15,19,20, potently elevated somatic Ca²⁺ levels in astrocytes (Fig. 1d), indicating astrocyte somatic Ca²⁺ signaling is regulated by these neuromodulators.

Tdc2⁺ neurons are the only known source of Tyr and Oct in the larval VNC ²¹. To explore their relationship with astrocytes we expressed the red-shifted Ca²⁺ indicator R-GECO1 in Tdc2⁺ neurons (using tdc2-LexA/LexAop-R-GECO1) and GCaMP6s in astrocytes (using alrm-Gal4/UAS-GCaMP6s) and examined in vivo activity. We observed a striking positive correlation between Tdc2⁺ neuron activity and somatic astrocyte Ca²⁺: when Tdc2⁺ neurons were active, astrocyte somatic Ca²⁺ levels increased, and when Tdc2⁺ neurons were silent, astrocyte somatic Ca²⁺ levels decreased (Fig. 2a; Supplementary video 2). A similar correlation was observed in intact larvae (Extended Data Fig. 1p). Moreover, the amplitude and duration of somatic astrocyte Ca²⁺ rise was tightly correlated with Ca²⁺ spikes in Tdc2⁺ neurons (Fig. 2a). When we chronically silenced Tdc2⁺ neurons by expressing the K⁺ leak channel Kir2.1²², rhythmic oscillations in astrocyte Ca²⁺ were eliminated (Fig. 2b; Extended Data Fig. 2a); and acute optogenetic blockade of Tdc2⁺ neuron activity using halorhodopsin (tdc2-Gal4>UAS-NpHR) led to a decrease in astrocyte somatic Ca²⁺ (Fig. 2c; Extended Data Fig. 2b). Interestingly, astrocyte Ca²⁺ signaling was also blocked by the α1 adrenergic receptor antagonist terazosin (Fig. 2d; Extended Data Fig. 2c), which inhibits arousal-evoked Ca²⁺ responses in cortical astrocytes in mammals ^14,15, further supporting the notion that astrocyte modulation is highly conserved between species. Consistent with astrocyte Ca²⁺ signaling being regulated by Oct and Tyr, we found somatic astrocyte Ca²⁺ transients were nearly absent (~80% reduced) in tdc2^RO54 mutants, which lack both Tyr and Oct, although they persisted in tβh^nM18 mutants, which lack Oct but retain Tyr signaling (Fig. 2e,f). Finally, Tdc2⁺ neurons were activated when olfactory neurons were optogenetically stimulated in intact larvae (using or83b-Gal4), suggesting sensory cues flow to Tdc2⁺ neurons, which are upstream of astrocyte Ca²⁺ signaling (Extended Data Fig. 2d).

Figure 2. Tdc2⁺ neurons fire rhythmically to drive astrocyte somatic Ca²⁺ transients through the octopamine-tyramine receptor (Oct-TyrR).

a, Correlation analyses of neuronal activity and somatic Ca²⁺ transients in astrocytes. Averaged traces of 16 individual astrocytes (GCaMP6s) and 4 pairs of Tdc2⁺ neurites (R-GECO1) from the same sample. Vertical orange bars highlight concomitant activity in astrocytes and neurons (* marked events are shown in inset). Amplitude and duration of individual Ca²⁺ transients in astrocytes correlate highly with activity of Tdc2⁺ neurons. b, Frequency of Ca²⁺ transients in astrocytes (n=10,160 cells total). c,d, Relative changes in numbers of Ca²⁺ transients after acute blockade of octopamine/tyramine signaling by halorhodopsin (c) or terazosin (d) (n=6, 96 cells total). Black dots, when either 561nm light or terazosin was delivered. e, Representative traces of astrocytic Ca²⁺ transients in mutants defective in tyramine/octopamine signaling. f,g, Frequency of Ca²⁺ transients in tdc2^RO54, tβh^nM18 (f, n=6, 96 cells total) and Oct-TyrR^hono mutants (g, n listed for each genotype, 16n cells total). h, Chemotaxis assay (n=12). i, Gentle touch assay (n=30). *p<0.05, **p<0.01, n.s., not significant. Error bar, s.e.m. Paired t-test (c,d), one-way ANOVA followed by Tukey's post hoc test (b,f,g,h,i).

To determine whether Tdc2⁺ neurons signal directly to astrocytes we knocked down the two Tyr receptors TyrR and TyrRII, and Oct-TyrR, a dual specificity receptor capable of binding both Oct and Tyr ^23,24, in astrocytes and assayed Ca²⁺ signaling. Depletion of TyrR or TyrRII had no effect (Data not shown), however depletion of Oct-TyrR strongly suppressed astrocyte somatic Ca²⁺ transients. We observed strong inhibition of astrocyte somatic Ca²⁺ transients in the Oct-TyrR^hono homozygous mutant, and in Oct-TyrR^hono/+ heterozygous mutants, which were previously shown to have dominant effects ²⁵ (Fig. 2g). Loss of one copy of Oct-TyrR (Oct-TyrR^hono/+) also led to deficits in chemotaxis behavior similar to those observed with astrocyte-specific knockdown of wtrw; these behavioral changes were enhanced in the Oct-TyrR^hono homozygous mutant, or by astrocyte-specific knockdown of Oct-TyrR with two copies of an Oct-TyrR RNAi construct. Co-expression of Oct-TyrR^RNAi and wtrw^RNAi, did not enhance chemotaxis defects, suggesting they act in a common genetic pathway (Fig. 2h). Finally, astrocyte-specific depletion of Oct-TyrR led to a strong inhibition of Type II escape responses in the gentle anterior touch assay (Fig. 2i). We conclude Wtrw and Oct-TyrR are required in astrocytes to trigger somatic Ca²⁺ transients in response to Oct and Tyr, and astrocyte Wtrw and Oct-TyrR are essential for normal olfactory-driven chemotaxis and startle-induced escape responses.

Astrocytes support neuronal function in many ways and it is plausible that alterations in astrocyte Ca²⁺ signaling might modify upstream Tdc2⁺ neuron firing to influence behaviors. We therefore examined the frequency and amplitude of firing events in Tdc2⁺ neurons in wtrw^ex null mutants and found no alterations compared to controls (Extended Data Fig. 2e). That Tdc2⁺ neurons fire normally in wtrw mutants indicates that blocking astrocyte Ca²⁺ signaling does not lead to a global disruption of neuronal activity in the larval CNS.

Biogenic amines often regulate dopaminergic (DA) neurons ²⁶. Triple labeling revealed that astrocytes, Tdc2⁺ neurons and DA neurons are in close proximity in the larval CNS (Extended Data Fig. 2f). When we imaged DA neuron activity (using th-Gal4/UAS-GCaMP6s) we found wtrw^ex mutants exhibited a significant increase (~50-70%) in the frequency of DA Ca²⁺ transients in multiple segmental clusters of TH⁺ cells. A similar phenotype was observed in Oct-TyrR^hono/+ mutants (Fig. 3a), although there were no significant changes in the average amplitude of DA neuron responses (Extended Data Fig. 2g). Thus Wtrw and Oct-TyrR negatively regulate DA neuron firing. Consistent with this observation, we found bath application of Oct or Tyr, but not vehicle (DMSO) to larval CNS suppressed DA neuron activity within seconds (Fig. 3b). Strikingly, depletion of the Oct-TyrR and Wtrw specifically from astrocytes using RNAi potently suppressed Tyr-induced Ca²⁺ elevation in astrocytes and silencing of DA neurons (Fig. 3c; Extended Data Fig. 2k). These data indicate that Tyr signals through astrocyte Oct-TyrR and Wtrw to inhibit downstream DA neuron activity. To explore how DA neuron modulation regulates larval chemotaxis behavior we performed a series of acute manipulations. We first expressed the temperature sensitive channel TrpA1 in DA neurons and found after 24-hour ectopic activation of DA neurons larvae exhibited profound chemotaxis defects similar to those observed when Wtrw or Oct-TyrR were depleted from astrocytes (Extended Data Fig. 2h). In addition, feeding larvae a dopamine D1-like receptor antagonist rescued deficits associated with wtrw^RNAi in astrocytes in chemotaxis assays (Extended Data Fig. 2i). These data support the notion that the increased DA neuron activity observed after depletion of astrocyte Oct-TyrR and Wtrw is responsible for defects in chemotaxis behavior.

Figure 3. Astrocytes mediate tyramine-induced inhibition of dopaminergic neurons through Oct-TyrR.

a, Enhanced activity of dopaminergic neurons in wtrw^ex mutant, Oct-TyrR^hono /+ heterozygotes (n=10, 80 neurites). s, subesophageal segments. t, thoracic segments. b, Tyramine and octopamine (2.5mM) inhibit the activity of dopaminergic neurons (n=6, 48 neurites). c, Astrocyte-specific RNAi for Oct-TyrR or wtrw attenuates inhibition of tyramine on dopaminergic neurons (n=6, 48 neurites). Black dots, when tyramine was perfused. *p<0.05, **p<0.01, n.s., not significant. Error bar, s.e.m. Paired t-test (b, same treatment, c, same genotype), one-way ANOVA followed by Tukey's post hoc test (a,b,c).

How do astrocytes regulate DA neuron firing? Astrocytic release of ATP and signaling through purinergic receptors is one mechanism by which astrocytes appear to modulate neuronal activity ^27,28. The Drosophila genome contains a single seven transmembrane purinergic receptor, AdoR, that is most similar to mammalian adenosine receptors ²⁹. Pretreatment of larval CNS preparations with the AdoR antagonist SCH-442416 blocked Tyr-induced DA neuron silencing without altering overall DA neuron activity before Tyr treatment. Similar results were observed in AdoR null mutants (Fig. 4a; Extended Data Fig. 2j). Purinergic signaling often results from vesicular release of ATP, which can be hydrolyzed to adenosine by ectoenzymes. We manipulated exocytosis selectively in astrocytes using a dominant negative version of the dynamin Shibire (Shi^DN) and found this suppress the ability of Tyr to silence DA neurons. This blockade was not enhanced further by application of AdoR antagonists and had no effect on baseline DA neuron activity before Tyr treatment (Fig. 4b; Extended Data Fig. 2l). Finally, to determine whether astrocyte Ca²⁺ signaling through TRP channels is sufficient to suppress DA neuron firing we expressed the wasabi (AITC)-sensitive Drosophila TrpA1 channel in astrocytes and assayed the effects of AITC application on astrocyte Ca²⁺ signaling and DA neuron firing. Exposure of control animals to AITC did not alter astrocyte Ca²⁺ levels, however application of AITC to preparations expressing TrpA1 in astrocytes led to an increase in GCaMP6s signals of astrocytes of ~20% (Extended Data Fig. 2m), similar to that elicited by Tyr application (Fig. 1d). Moreover, AITC activation of astrocyte Ca²⁺ signaling resulted in a strong suppression of DA neuron activity. This effect was reversible by application of AdoR antagonists and did not occur when AITC was applied to control (i.e. non-TrpA1 expressing) animals (Fig. 4c; Extended Data Figure 2n). Thus TrpA1 channel mediated increases in astrocyte Ca²⁺ signaling are sufficient to silence DA neurons in vivo. DA neuron silencing is likely be mediated by ATP release from astrocytes and subsequent activation of AdoR on DA neurons.

Figure 4. Tyramine mediated inhibition of dopaminergic neurons depends on adenosine receptor and glial endocytic function.

a, AdoR is required for tyramine mediated inhibition of dopaminergic neurons (n=6, 48 neurites). b, Blockade of endocytic function by dominant negative shibire (shi^DN) attenuates inhibition of tyramine on dopaminergic neurons (n=6, 48 neurites). Black dots, when tyramine was perfused. c, Ca²⁺ influx to astrocytes through TrpA1 is sufficient to inhibit dopaminergic neurons (n=6, 48 neurites). Black dots, when AITC was perfused. d, Model. Olfactory and likely mechanosensory information flow towards Tdc2⁺ neurons that release Oct and Tyr to activate the Oct-TyrR on astrocytes and in turn astrocyte Ca²⁺ entry through the TRP channel Wtrw. Increase in astrocyte Ca²⁺ is sufficient to silence DA neurons through a mechanism requiring the adenosine receptor AdoR, potentially through astrocyte ATP release and its breakdown to adenosine. Inhibition of DA neuron activity by astrocytes is essential for normal chemotaxis and startle-induced reversal behaviors. *p<0.05, **p<0.01, n.s., not significant. Error bar, s.e.m. Paired t test (a,b,c, same genotype), Wilcoxon and Mann-Whitney tests followed by Bonferroni-Holm post hoc test (a,b), one-way ANOVA followed by Tukey's post hoc test (c).

It has been assumed that neuromodulators alter neural circuit activity and behaviors by directly acting on neurons; however, astrocytes (and other glia) express a diverse array of neuromodulatory receptors suggesting they may contribute to the global effects of these neurotransmitters. In this study we show neuromodulatory signaling in at least some cases flows through astrocytes. We demonstrate Drosophila astrocytes exhibit robust Ca²⁺ signaling events remarkably similar to those observed in awake behaving mice ^14-16. Similar to the global activation of astrocyte Ca²⁺ signaling by norepinephrine in mammals, release of tyramine (and likely octopamine) in vivo triggers synchronous activation of astrocytes. We identify Wtrw and Oct-TyrR as key components of the astrocyte Ca²⁺ signaling axis and their loss of function phenotypes provide direct evidence that astrocyte Ca²⁺ signaling modulates animal behaviors. Furthermore, we show that during suppression of DA neuron activity, Oct/Tyr signal directly to astrocytes, and astrocyte Ca²⁺ transients execute neuromodulatory events downstream (Fig. 4d). Octopamine and tyramine have been linked to arousal and aggression in insects and are considered to be functionally equivalent to norepinephrine in mammals. The profound modulation of astrocyte Ca²⁺ levels by norepinephrine in mammals ^14-16 suggests that the Tyr/Oct/norepinephrine-mediated activation of astrocyte Ca²⁺ signaling, and in turn astrocyte-based neuromodulation, are ancient features of the metazoan nervous system. A reassessment of the cellular basis of neuromodulation is therefore warranted, with further reflection on direct roles for astrocytes in transducing neuromodulatory signals in neural circuits.

Material and Methods

Fly stocks and husbandry

Flies were cultured in standard cornmeal food at 25°C in 12h/12h light/dark cycles. Fly stocks include: w¹¹¹⁸, alrm-Gal4 ²⁶, alrm-LexA::GAD, UAS-wtrw-gfp, LexAop-R-GECO1 (Bloomington, 52224), tdc2-Gal4 (Bloomington, 9313), 20XUAS-IVS-GCaMP6s (Bloomington, 42746), 13XLexAop2-IVS-GCaMP6s-p10 (Bloomington, 44274), UAS-Kir2.1-EGFP (Bloomington, 6595), UAS-Oct-TyrR^RNAi (Bloomington, 28332), th-Gal4 (Bloomington, 8848), UAS-wtrw^VDRC (VDRC, 107423), Oct-TyrR^hono (Kyoto, 109038), UAS-shi^DN (Bloomington, 5822), UAS-TrpA1 (Bloomington, 26263), UAS-NpHR (Kyoto, 117008), UAS-CsChrimson (Bloomington, 55135), AdoR^KG03964ex (Bloomington, 30868), tdc2-LexA::p65 ²⁷, UAS-wtrw^{2.23 28}, wtrw^{ex 29}, tβh^{nM18 30}, tdc2^{RO54 31}, th-LexA::p65 ³², tsh-Gal80 ³³.

Behavioral tests

Chemotaxis and phototaxis

These two assays were performed using protocols described previously with minor modifications ³⁴. Briefly, pools of ~100 3^rd instar larvae (108-120h after egg lay) were allowed to freely move for 5 minutes on petri dishes with according settings for chemotaxis assay or phototaxis assay, then the numbers of larvae in odorant/vehicle circles, or in light/dark quadrants were scored respectively. Response indices were calculated as: RI_chemotaxis=(N_odor-N_vehicle)/(N_odor+N_vehicle), RI_phototaxis=(N_dark-N_light)/(N_dark+N_light). Isoamyl acetate (IAA) was diluted to 1:5000 in mineral oil, and pure mineral oil was used as the control. Light intensity in phototaxis assays was 2400 lux white light (5400K). Thermogenetic manipulation of dopaminergic neurons was accomplished by transferring 3^rd instar larvae from 18°C to 25°C to allow activation of TrpA1 for 24h before testing. For chemotaxis assays related to D1-like or D2-like dopamine receptor inhibitors 3^rd instar larvae were fed on cornmeal food mixed with inhibitors for 24h before testing.

Gentle touch assay

Adapted from ³⁵, early 3^rd instar larvae (84-96h after egg lay) were struck in the thoracic segments with a hair while moving straight. No response, a stop, head retraction and turn were grouped into type I response, initiation of at least one single full body retraction or multiple full body retractions were categorized as type II reversal responses.

Larval motility

3^rd instar larvae (108-120h after egg lay) were videotaped and the maximal velocity was calculated by dividing the travel distance measured when larvae move straight without pause by time.

Molecular biology and transgenic flies

pUAST-wtrw::gfp was made by cloning full length wtrw tagged with gfp at C terminal into pUAST. Transgenic flies were obtained by standard germline injection (BestGene Inc, Chino Hills, CA).

Chemicals, buffers and antibodies

For live preparation studies all the chemicals were diluted in live imaging buffer (110mM NaCl, 5.4mM KCl, 1.2mM CaCl₂, 0.8mM MgCl₂, 10mM D-glucose, 10 mM HEPES, pH adjusted to 7.2), including 1μM tetrodotoxin, 0.1mM LaCl₃, 2.5mM octopamine, 2.5mM glutamate, 2.5mM acetylcholine, 2.5mM GABA, 150μM AITC, 10μM SCH-442416, 1mM terazosin, 0.5mM tyramine was used unless otherwise stated. 0.2mg/ml SCH-23390 (dopamine D1 like receptor antagonist), 0.2mg/ml Eticlopride (dopamine D2 like receptor antagonist), 10mM all trans retinal. Rabbit anti-GAT was used at 1:5000.

Live imaging and data analysis

1^st instar larvae were immersed in halocarbon oil 27 and immobilized between coverslips and slides to image the activity of neurons and astrocytes through the cuticle in intact larvae. 3^rd instar wandering stage larvae were used to dissect the CNS. Briefly, intact CNS were transferred to a silicon pad after dissection and incubated with 100μl live imaging buffer. Samples were allowed to sit at room temperature for 15 minutes before imaging. To examine the effects of lanthanum chloride (LaCl₃) or tetrodotoxin (TTX) on calcium transients, samples were pre-incubated in buffer containing chemicals for 15 minutes before imaging. Movies were acquired by using Volocity software on a spinning disc microscope equipped with a 20X (dissected CNS) or 40X (intact larvae) water immersion objective at 2 frames/second. Changes of GCaMP intensity of ROIs were calculated and expressed as (F_t1...n-F)/F (first 10 frames were averaged as F), which then were plotted over time. Plotted traces were semi-automatically analyzed by the multiple peak fitting function of Igor Pro (WaveMetrics Inc., OR) to obtain the numbers and amplitudes of peaks.

Bath application of drugs

Samples were prepared as described above, incubated with imaging buffer (with/without 1μM TTX). At ~240s (astrocytes) or ~300s (neurons) after imaging, 100μl live imaging buffer containing 2X final concentrations of chemicals (with/without 1μM TTX) were directly added to the medium, and imaging continued. GCaMP6s intensity of ROIs were plotted and analyzed as mentioned above. For terazosin experiments we first examined preparations for 6 min (control, pre-exposure window), terazosin was perfused for 3 min, and we then imaged for an additional 6min (exposed window). Data points of 6-9min that represented the perfusion window were omitted in graph.

Optogenetic activation/inactivation of neurons

For CsChrimson activation of olfactory neurons larvae were grown on cornmeal food daily topped with 100 μl all trans retinal for 3 days. Early 3^rd instar larvae were immobilized between coverslips and slides in halocarbon oil 27, activity of Tdc2⁺ neurons reflected by GCaMP6s were continuously monitored except when red light (640nm, 3.41mW/ mm²) was delivered through an epifluorescence light source. Halorhodopsin mediated inhibition of Tdc2 neurons was accomplished by growing larvae on cornmeal food topped daily with 100 μl all trans retinal for 4-5 days. Larval brains were dissected from wandering 3^rd instar larvae and prepared as described above for imaging somatic calcium transients in astrocytes. GCamP6s signals were recorded for 6 minutes, then halorhodopsin was activated (within ~30 seconds) by alternating delivering 561nm (0.26mW/mm², 3s, for photoinhibition) and blue light 488nm (500msec, for imaging GCaMP6s labeled astrocytes) for another 6min. Data points from two consecutive movies were then assembled together to show changes of somatic calcium transients.

Statistical analysis

No statistical methods were used to predetermine sample size. All n numbers represent biological replicates. The experiments were not randomized or blinded. The Shapiro–Wilk test (normal distribution if P>0.05) was used to determine the normality of data. Statistical comparisons were performed using one-way ANOVA followed by Tukey's post hoc test or paired Student's t-test (two-tailed). Non-normally distributed data were compared using Wilcoxon and Mann-Whitney tests followed by Bonferroni-Holm post hoc test. P values less than 0.05 were considered significant. All the data in bar graphs are expressed as mean ± s.e.m.

Extended Data

Extended Data Figure 1. Synchronous somatic Ca²⁺ transients in Drosophila astrocytes.

a-c, Brief diagrams of behavioral tests. d, tsh-Gal80 suppression of alrm-Gal4 activity. Nc82, neuropil (NP). Astrocytes (Astro), alrm>myr::mtdTomato. Vnc, ventral nerve cord. Scale bar, 50μm. e,f, Chemotaxis assay (n=12). g, Locomotion assay (n listed). h, Light avoidance assay (n=12). i, Gentle touch assay (n=24). j, GCaMP6s and mCherry expression in astrocytes. Scale bar, 50μm. k, Representative pseudocolored images of 4 continuous Ca²⁺ transients. Scale bar, 50μm. l, Traces of normalized GCaMP6s intensity over mCherry of 10 individual astrocytes in 15min live imaging window. m, Averaged traces of individual somatic Ca²⁺ transients. n, somatic Ca²⁺ transients in astrocytes with treatments of TTX and LaCl₃ (n=10, 160 cells). o, GCaMP6s expression in astrocytes and traces of 10 individual astrocytes from an intact larva. Scale bar, 20μm. Grey bars (l,o) represent population rise/fall in GCaMP6s signals. p, GCaMP6s labeled astrocytes and R-GECO1 labeled Tdc2⁺ neurites and their averaged traces in an intact larva. Scale bar, 20μm. *p<0.05, **p<0.01, n.s., not significant. Error bar, s.e.m. Wilcoxon and Mann-Whitney tests followed by Bonferroni-Holm post hoc test (e,n), one-way ANOVA followed by Tukey's post hoc test (f,g,h,i).

Extended Data Figure 2. Somatic Ca²⁺ transients of astrocytes inhibit the activity of dopaminergic neurons.

a-c, Representative traces of astrocyte Ca²⁺ transients with blockade of tyramine/octopamine signaling. d, Stimulation of olfactory neurons activates Tdc2⁺ neurons (n=3-4). Scale bar, 25μm. e, Activity of Tdc2⁺ neurons are not altered in wtrw mutants (n=8, 48 neurites). f, Astrocytes, Tdc2⁺ neurons and dopaminergic neurons in larval CNS. Dorsal (arrows point to astrocyte somas), medial (neurites intermingled with ramified processes of astrocytes for monitoring activity are labeled) and ventral (cell bodies of Tdc2⁺ and dopaminergic neurons) images from the boxed region are shown right. s, subesophageal. t, thoracic. Scale bar, 50μm. g, Amplitude of Ca²⁺ spikes of dopaminergic neurons (n=10, 80 neurites). h,i, Chemotaxis assay (n listed). j, Number of Ca²⁺ spikes of dopaminergic neurons (n=6, 48 neurites). k, Responses of astrocytes to tyramine (0.5mM) in the presence of TTX (n=6, 96 cells total). l, Number of Ca²⁺ spikes of dopaminergic neurons (n=6, 48 neurites). m, AITC induces Ca²⁺ influx to astrocytes expressing TrpA1 (n=5, 80 cells). Scale bar, 50μm. n, Number of Ca²⁺ spikes of dopaminergic neurons (n=6, 48 neurites). *p<0.05, **p<0.01, n.s., not significant. Error bar, s.e.m. One-way ANOVA followed by Tukey's post hoc test.