Morning and Evening Circadian Pacemakers Independently Drive Premotor Centers via a Specific Dopamine Relay

Summary

Many animals exhibit morning and evening peaks of locomotor behavior. In Drosophila, two corresponding circadian neural oscillators - M (morning) cells and E (evening) cells - each exhibits a corresponding morning or evening neural activity peak. Yet we know little of the neural circuitry by which distinct circadian oscillators produce specific outputs to precisely control behavioral episodes. Here we show that Ring Neurons of the Ellipsoid Body (EB-RNs) display spontaneous morning and evening neural activity peaks in vivo: these peaks coincide with the bouts of locomotor activity and result from independent activation by M and E pacemakers. Further, M and E cells regulate EB-RNs via identified PPM3 dopaminergic neurons, which project to the EB and are normally co-active with EB-RNs. These in vivo findings establish the fundamental elements of a circadian neuronal output pathway: distinct circadian oscillators independently drive a common pre-motor center, through the agency of specific dopaminergic interneurons.

eTOC

Liang et al. describe neural outputs from the Drosophila circadian pacemaker network regulating locomotor rhythms: via a specific dopamine relay, the network forms parallel connections to the central complex, the pre-motor area dictating the balance between rest and activity.

INTRODUCTION

Circadian rhythms provide adaptive value by promoting expression of diverse physiological processes and behaviors at specific times of the day. In mammals, rhythms in hormone release, rest/activity cycles, body temperature, and metabolism are all controlled by the multi-oscillator system of pacemakers in the suprachiasmatic nucleus (SCN) of the anterior hypothalamus. Numerous studies have documented that the SCN uses hormonal and neuronal signaling to provide adaptive phasic information across all times of day (Lehman et al., 1987; Moore and Klein, 1974; Ralph et al., 1990, De la Iglesia et al., 2003; Kalsbeek et al., 2006; VanderLeest et al., 2007). However, the information connecting SCN signaling to neural circuits that translate its outputs is fragmentary. Lacking direct in vivo experimental observations, the definition of circadian output networks remains a significant challenge.

In Drosophila, a prominent circadian output is the daily locomotor activity rhythm, which peaks once around dawn and again around dusk (Figure 1B). The rhythm is controlled by molecular clocks that cycle synchronously within ~ 150 circadian pacemaker neurons (Nitabach & Taghert, 2008). Among these circadian neurons, two separate groups (termed M cells and E cells) control the morning and evening activity peaks respectively (Stoleru et al., 2004; Grima et al., 2004; Yoshii et al., 2004). Previously we reported that different groups of circadian neurons display rhythmic but asynchronous circadian neural activity in vivo: they peak at different yet stereotyped times of day (Liang et al., 2016). These neural activity rhythms depend on their synchronous molecular clocks, but their activity peak times are staggered, by neuropeptide-mediated interactions between circadian neuron groups. This allows the network to create multiple phasic time points (Liang et al., 2017). Consequently, M cells peak in the morning and E cells peak in the evening. The distinct peak times of M cells and E cells could potentially guide output motor circuits to generate independent morning and evening locomotor behavioral peaks. To support this emerging network view of pacemaker regulated behavior, we wished to ask which pre-motor centers transduce circadian timing signals to generate morning or evening phase-specific locomotor activity. As a strategy, we reasoned that spontaneous neural activity patterns corresponding to the daily bimodal activity pattern could help identify the critical premotor elements.

Figure 1. Daily bimodal neural activity patterns of EB ring neurons.

(A) The central complex in the fly brain. EB-RNs, ellipsoid body ring neurons; FSB, fan-shaped body; PB, protocerebral bridge; NO, noduli.

(B) The average locomotor activity histogram and phase distributions of behavioral peaks of wild type R56H10-GAL4/GCaMP6s flies (above, n = 16) and flies in which EB-RNs are silenced by TeTn (below, n = 14). Both genotypes are shown under a 12-hr light: 12-hr dark (LD) cycle (left) and in the first day under constant darkness (DD) (right). Dots indicate SEM.

(C) Average rhythm strength (power) of locomotor activity for 9 days under DD displayed by two genetic controls and by flies in which EB-RNs express TeTn (; asterisk denotes significant differences compared to control (P < 0.0001, Mann-Whitney test).

(D–H) Daily Ca²⁺ activity patterns of the EB ring neuron subgroups: (D) R1 labelled by tim-GAL4, (E) R2 labelled by cry-lexA, (F) R3 labelled by pdfr(F)-GAL4, (G) R4 labelled by R19H08(pdfr)-lexA, and (H) R1–4 labelled by R56H10-GAL4. Left, confocal images of EB ring neurons and diagrams of their concentric arborizations; scale bars, 25 μm. Middle and Right, average Ca²⁺ transients and Ca²⁺ phase distributions for both morning peaks (orange dots and arrow) and evening peaks (blue dots and arrow). Middle - under LD; Right - under DD.

Robie et al. (2017) performed an unbiased screen of sparsely-labeled neuronal groups to determine which when activated could initiate locomotor activity. By this analysis, the strongest candidates were the Ring Neurons of the Ellipsoid Body (EB-RNs) (Figure 1A). In parallel, silencing these same EB-RNs reduced spontaneous locomotor activity (Martín-Peña et al., 2014). EB-RNs are a subset of neurons that constitute the Central Complex - the primary locomotor control center in insects (Strauss and Heisenberg, 1993; Pfeiffer and Homberg, 2014). EB-RNs encode visual landmarks for visuospatial-memory-based orientation and navigation (Neuser et al., 2008; Ofstad et al., 2011; Seelig and Jayaraman, 2013). In the monarch butterfly, EB-RNs are involved in sun-compass navigation (Heinze and Reppert, 2010), which requires timing information from circadian clocks (Froy et al., 2003). Several recent studies have linked activity in subsets of the EB-RN to the regulation of sleep-wake physiology. Liu et al (2016) identified an EB subset whose activity defines an integrator that promotes sleep drive. In addition, two independent groups described synaptic links between a subset of the DN1 group of circadian pacemakers to subgroups of EB-RNs (Lamaze et al., 2018; Guo et al., 2018) which may help regulate sleep/arousal. DN1 cells integrate outputs from M and E cells and regulate both locomotor activity and sleep (Zhang et al., 2010; Liang et al., 2017; Guo et al., 2016). Therefore, we first measured spontaneous activity in EB-RNs in vivo, to see if they represent a point of convergent circadian regulation that could lead to daily bouts of locomotor activity.

RESULTS

Spontaneous daily bimodal activity in EB-RNs in vivo

To test whether EB-RNs regulate circadian locomotor activity, we expressed tetanus toxin light chain (TeTn, Sweeney et al., 1995) to block neurotransmission in the majority of ~60 EB-RNs. As expected, the circadian rhythm of locomotor activity in these flies was impaired under DD, as was the general level of activity (Figure 1B and Table S1). Therefore, to learn about the possible involvement of the EB-RNs in normal rhythmic locomotion (Figure 1C), we then measured in vivo spontaneous activity exhibited by these neurons in otherwise wild-type flies. Using the genetically encoded calcium sensor GCaMP6s (Chen et al, 2013), we performed in vivo Ca²⁺ imaging in living flies for 24 hrs using methods previously described (Liang et al., 2016, 2017). EB-RNs contains several genetically and morphologically distinct subgroups (Renn et al., 1999a). We dissected four EB-RN subgroups using different genetic drivers that use regulatory sequences associated with different circadian clock-related genes: one with sequences from timeless, one from cryptochrome, and two from pdfr (pigment-dispersing factor receptor) (Figure 1D–G and Figure S1A). In both 12-hr light: 12-hr dark (LD) cycles and in constant darkness (DD) conditions, the four different EB-RN subgroups we tested displayed spontaneous, daily Ca²⁺ rhythms (Figure 1D–G and Figure S2A–D). These dynamic signals were not a consequence of rhythmic GCaMP6 sensor expression (Figure S3A–F). The average Ca²⁺ activity profile of each subgroup was bimodal (Hartigans’ dip test, LD: p<0.0001, DD: p<0.05), with a peak around dawn and another around dusk. At these two peaks, with high-frequency light-sheet imaging (1Hz), we found that many EB-RNs showed an increase in the incidence and magnitude of fast spontaneous activity events (Figure S3GH). The two daily peaks corresponded to the times of day when flies showed daily locomotor activity peaks (Figure 1C). The outer subgroup of EB-RNs caused the strongest effects on locomotor activity according to Robie et al. (2017). We tested the same split-GAL4 drivers as reported by Robie et al. (2017) and found that these locomotion-promoting EB-RNs likewise displayed a similar spontaneous daily bimodal activity pattern (Figure S1B; Hartigans’ dip test, p<0.0001). We also confirmed the daily bimodal activity pattern exhibited by different EB-RN subgroups using a separate, circadian-clock-irrelevant driver line to label the majority of EB-RNs (Figure 1H and Figure S2E; Hartigans’ dip test, LD: p<0.001, DD: p<0.01).

To directly test the correlation between EB-RN neural activity and locomotor activity in single flies in our experimental paradigm, we performed in vivo 24-hr Ca²⁺ imaging while simultaneously measuring spontaneous leg movements as a proxy for locomotor activity levels (Figure 2A). EB-RN activity was strongly correlated with such behavioral activity in individual flies, both at a daily time scale (Figure 2F–K) as well as at a shorter (hourly) time scale (Figure 2C–E and Figure S4E–L). Analysis of the shorter timescale indicated that increases in EB-RN activity were coincident with increases in behavioral activity; decreases typically preceded decreases in behavioral activity by a few minutes (Figure 2E). We noted that under DD the imaged flies (Figure 2F) showed higher locomotor activity around the middle of subjective day than did freely-moving flies (Figure 1B - right panel). The increase in activity during the middle of the subjective day was not caused by illumination of the specimen during imaging (Figure S4A–C) and was not observed under LD conditions. Under LD, both the locomotor activity and EB-RN activity clearly showed two daily peaks. Thus, EB-RNs, consistent with their documented roles as pre-motor activity centers, exhibit spontaneous daily neural activity rhythms that precisely correspond to the pattern of circadian locomotor rhythms.

Figure 2. EB ring neuron activity is correlated with locomotor activity.

(A) Illustration of long-term in vivo imaging arrangement combined with infrared measurement of locomotor activity (see Methods).

(B) Map of the major circadian neuron groups and EB ring neurons.

(C1 & C2) Representative recordings of two flies in DD for 24 hr: bars, normalized locomotor activity counts per 10 m; blue traces, Ca²⁺ activity of EB-R2 neurons in the same fly.

(D–E) Average locomotor activity (black) and Ca²⁺ activity of EB-R2 neurons (blue) aligned by (D) increasing phase and (E) decreasing phase of Ca²⁺ activity. The averaged phase lags were calculated by cross-correlation: (D) 0.35 ±7 min; (E) −15±0.5 min.

(F–H) Average locomotor activity (F) and average Ca²⁺ transients of (G) EB-R2 neurons and (H) circadian pacemaker neurons in the same flies (cry-lexA>GCaMP6s; n = 6 flies) during DD1. Dots and shading indicate SEM.

(I–K) Average locomotor activity (I) and average Ca²⁺ transients of (J) EB-R2 neurons and (K) circadian pacemaker neurons in the same flies (cry-lexA>GCaMP6s; n = 5 flies) under LD cycles.

Circadian pacemaker neurons drive EB-RN activity rhythms

The daily neural activity rhythms in EB-RNs could reflect rhythmic sensory inputs, either proprioceptive or visual. For example, recent studies suggest that EB-RNs encode self-motion information (Shiozaki & Kazama, 2017). Therefore, to block ascending proprioceptive sensory inputs, we transected connectives between the brain and ventral nerve cord (between subesophageal and first thoracic neuromeres) immediately before Ca²⁺ imaging. EB-RNs still displayed normal bimodal activity rhythms (Figure 3A). These EB-RN rhythms persisted even when the entire body of the fly was removed immediately before imaging (Figure 3B). Therefore, spontaneous bimodal EB-RN activity rhythms are not a consequence of locomotor behavioral activity. Previous studies also showed that EB-RNs receive large-scale visual inputs (Seelig & Jayaraman, 2013; Omoto et al., 2017; Sun et al., 2017). We therefore removed visual inputs by testing flies in DD (Figure 1) or by testing genetically blind norpA^P24 mutant flies (Figure 3C): in both cases, normal EB-RN activity rhythms persisted. Together, these results indicate that neither visual nor proprioceptive inputs are necessary for spontaneous EB-RN activity rhythms.

Figure 3. EB ring neuron rhythms are driven by clocks, not in response to behavior or sensation.

(A) Daily Ca²⁺ activity patterns of (middle) EB-R2 neurons and (right) circadian neurons under LD immediately after cutting the neural connectives between brain and ventral nerve cord in otherwise intact flies (dotted line indicates level of the cut; n = 10 flies).

(B) Daily Ca²⁺ activity patterns of EB-R2 and circadian neurons under LD immediately after transecting the connection between heads and bodies, then removing the bodies (n = 6 flies).

(C) In blind cry-lexA>GCaMP6s flies (norpA^p24 mutant), (left) average locomotor activity (n = 22 flies) and daily Ca²⁺ activity patterns (middle) of EB-R2 neurons and (right) circadian neurons under LD (n = 6 flies).

(D) In arrhythmic cry-lexA>GCaMP6s flies (per⁰¹ mutants), average locomotor activity (n = 16 flies) and arrhythmic Ca²⁺ activity patterns of EB-R2 and of circadian neurons under LD (n = 7 flies).

(E) In Pdfr-deficient cry-lexA>GCaMP6s flies (pdfr^han5304 mutants), average locomotor activity (n = 8 flies) and Ca²⁺ activity patterns of EB-R2 and of circadian neurons under LD (n = 7 flies).

To determine if EB-RN activity rhythms are driven by molecular clocks, we measured Ca²⁺ activity in circadian-defective per⁰¹ (null) mutant flies (Konopka & Benzer, 1971), which fail to display circadian clock-dependent anticipatory behavior. Although per⁰¹ flies still had two peaks of startle responses (to the lights-on and lights-off stimuli under LD cycles), daily Ca²⁺ activity patterns in EB-RNs were arrhythmic (Figure 3D). Therefore, EB-RN activity rhythms specifically correlate with - and entirely depend on - circadian clock signals that regulate daily behavioral peaks. Notably, EB-RNs do not exhibit measurable expression of the core clock gene period, which is highly expressed and cycles in circadian pacemaker neurons (Figure S1C) (Kaneko and Hall, 2000). Furthermore, manipulations that alter the pace of circadian clocks in a subset of circadian neurons reportedly shifted the phase of locomotor activity phases as previously reported (Stoleru et al., 2005; Yao and Shafer, 2014), while the same manipulation within EB-RNs did not affect locomotor behavior (Figure S1D–F). Thus, we conclude that daily EB-RN activity rhythms are downstream of circadian timing information provided by circadian pacemaker neurons. To test whether circadian neurons regulate EB-RNs, we impaired a crucial signal within the pacemaker network, the neuropeptide PDF (Renn et al., 1999b). In PDF receptor mutant (pdfr^han5304) flies (Han et al., 2005), the EB-RNs activity pattern under LD transformed to a daily unimodal one (Hartigans’ dip test, p=0.23): the morning activity peak was lost, and the evening peak was advanced (Figure 3E; Watson-Williams test, p=0.00012). This neural activity pattern mirrors the changes in locomotor activity pattern typically displayed by pdfr^han5304 flies. Meanwhile, EB-RNs responded to thermogenetic and pharmacogenetic activation of PDF-releasing neurons (Figure S5A–B). Thus, EB-RN activity rhythms could be driven (directly or indirectly) by PDF-expressing circadian pacemaker neurons.

Distinct circadian neurons dictate the separate phases of EB-RN activity

In contrast to EB-RNs, all circadian pacemaker neurons showed single daily peaks of activity. This difference suggested that the daily two-peak activity pattern of EB-RNs could be generated by a combination of different circadian neuronal outputs. We considered the simplest model, that M cells could drive a morning activity peak in EB-RNs while E cells could independently drive EB-RN evening activity. To begin to test this possibility, we first determined that EB-RNs responded to the selective activation of M cells (four s-LNv) by ATP application to brains expressing ATP-gated cation channel P2X2 (Lima and Miesenböck, 2005) in M cells (Figure 4A). A similar design to selectively activate E cells, the 5^th s-LNv and three PDFR-positive LNd (Im and Taghert, 2011), produced correspondent EB-RN responses of comparable amplitude (Figure 4BC and Figure S5D–F). These results support the proposition that both M cells and E cells have functional connections with EB-RNs.

Figure 4. Daily activity phases of EB-RNs are dictated by M and E cells.

(A) Illustration and averaged response traces of M cells (s-LNv) and EB-R2 neurons to ATP application in flies with P2X2 expressed in M cells (n = 6 flies).

(B) Illustration and averaged response traces of E cells (three LNd and the 5^th s-LNv neurons) and EB-R2 neurons to ATP application in flies with P2X2 expressed in E cells (n = 5 flies).

(C) Maximum Ca²⁺ changes of EB-R2 after ATP application in experimental flies from panels A and B and from controls (+/UAS-P2X2;cry-lexA>GCaMP6s; n = 3 flies, also see Figure S2D–F).

(D) Average locomotor activity in DD1 of controls (+>SGG;cry-lexA>GCaMP6s;, n = 16 flies).

(E) Left - average locomotor activity in DD1 of flies expressing SGG in PDF neurons (using PDF-GAL4 (M-cell)>SGG, n = 16 flies). Right - phase comparisons of morning and evening activity between +>SGG and M-cell>SGG. Note that only the morning activity phase was significantly different (advanced, * P < 0.05, Watson-Williams test).

(F) Left - daily Ca²⁺ activity patterns during DD1 of EB-R2 neurons in +>SGG;cry-lexA>GCaMP6s flies (solid line, n = 6) and in M-cell>SGG;cry-lexA>GCaMP6s flies (dash line, n = 6). Note that the morning peak of EB-R2 was significantly advanced in M-cell>SGG (orange arrow). Right - phase comparisons of EB-R2 Ca²⁺ activity peaks between +>SGG (black points and arrow) and M-cell>SGG (gray points and arrow). The morning and evening activity peaks are displayed separately.

(G) Daily Ca²⁺ activity patterns and phase comparisons of circadian neurons in the same flies from (F): +>SGG flies (solid line, n = 6) and in M-cell>SGG flies (dash line, n = 6). Only peak phases of M cells (s-LNv) were significantly advanced in M-cell>SGG (* P < 0.05, Watson-Williams test).

(H) Left - average locomotor activity in DD1 of flies expressing SGG in E cells (using split-GAL4 line GMR_MB122B; E-cell>SGG, n = 16 flies). Right - phase comparisons of morning and evening locomotor activity between +>SGG and E-cell>SGG. Only evening activity phase was significantly different (advanced, * P < 0.05, Watson-Williams test).

(I) Daily Ca²⁺ activity patterns and phase comparisons of EB-R2 neurons between +>SGG flies (solid line, n = 6) and in E-cell>SGG flies (dash line, n = 8). The evening peak of EB-R2 were significantly advanced in E-cell>SGG (blue arrow).

(J) Daily Ca²⁺ activity patterns and phase comparisons of circadian neurons between +>SGG flies (solid line, n = 6) and in E-cell>SGG flies (dash line, n = 8). Only peak phases of E cells (LNd) were significantly advanced in E-cell>SGG (* P < 0.05, Watson-Williams test).

As a more stringent test, we then asked whether selectively accelerating M or E oscillators would selectively influence the phase of either the morning and/or evening peak of EB-RN Ca²⁺ activity. Overexpressing Shaggy (SSG) using pdf-GAL4 (M-cell>SGG) to accelerate the molecular clocks selectively in morning oscillators advanced the morning peak of locomotor activity (cf. Stoleru et al., 2005) and the M-cell activity peak (Figure 4DE). In these flies, we found that only the morning peak of EB-RN Ca²⁺ activity was phase-advanced, while their evening Ca²⁺ peak phase was unaffected (Figure 4FG). This result suggests that the morning peak of EB-RNs activity is dictated predominantly by the phase of M cell activity. In parallel, we then asked whether the evening peak of EB-RNs is dictated by the phase of E cell activity. Overexpressing SGG in E cells by using a split-GAL4 line GMR_MB122B (Liang et al., 2017) selectively advanced the evening behavioral peak (Figure 4IJ) and the E-cell activity peak (Figure 4H–J). In these flies, we found that only the evening peak of EB-RN Ca²⁺ activity was phase-advanced, and by comparable amplitude to what we previously observed with M cell acceleration and M peak advance. Taken together, these results reveal essential circuit links to demonstrate that M cells and E cells can independently dictate the two distinct phases of EB-RN pre-motor activity.

Dopaminergic neurons regulate EB-RNs

None of the ~150 circadian pacemaker neurons in Drosophila project directly to the EB (Helfrich-Förster, 2005). We therefore asked through which interneurons M cells and E cells might regulate daily neural activity in EB-RNs. A set of two dopaminergic (DA) neurons (named PPM3-EB) appeared as prominent candidates: they innervate the EB; further, they can initiate locomotor activity and promote ethanol-induced locomotor activity (Kong et al., 2010). First we established that PPM3-EB neurons spontaneously displayed a daily bimodal neural activity pattern in vivo (Hartigans’ dip test, p<0.0001), similar to that of the EB-RNs and consistent with their putative involvement in the daily profile of locomotor activity (Figure 5A). To study the precise relationship between activity in EB-RNs and that in PPM3-EBs, we employed dual-color Ca²⁺ imaging in single fly brains (Figure 5B). This method separated Ca²⁺ activity signals from these two anatomically-overlapping neuron groups, by simultaneously recording a green signal in PPM3-EB (GCaMP6s), and a red signal in EB-RNs (jGECO1a, Dana et al., 2016). We found that the spontaneous Ca²⁺ activity patterns of EB-RNs were highly correlated with those of PPM3-EB, but poorly correlated with those of the l-LNv circadian neurons, which were also labelled by jGECO1a (Figure 5C–F and Figure S6C). This result suggests that PPM3-EB and EB-RNs are closely connected: they receive common inputs, and/or one receive synapses from the other. We next turned to anatomical analysis using the GRASP method (GFP reconstitution across synaptic partners - Feinberg et al., 2008). We expressed two complementary GFP fragments in PPM3-EB and EB-RNs respectively to demonstrate membrane appositions (potential synaptic connections) that normally exist between these two groups of neurons (Figure S6A–B). Furthermore, we asked whether the morning and evening activity peaks in PPM3-EB are independently dictated by M cells and E cells respectively, as we showed for the bimodal activity patterns in the EB-RNs. Again, we overexpressed SSG using pdf-GAL4 (PDF>SGG) to selectively advance the M-cell activity peak (as in Figure 4DE). In these flies, we found that only the morning peak of PPM3-EB was phase-advanced, while the evening peak phase was unaffected (Figure 5G–I). This result suggested that the daily activity rhythm of PPM3-EB are driven by the same set of circadian neuron outputs as EB-RNs. Both PPM3-EB and EB-RNs responded to the bath-application of PDF and to the pharmacogenetic activation of PDF neurons (Figure S6DE). In response to the activation of PDF neurons, PPM3-EB responded more quickly than EB-RNs, which is consistent with PPM3-EBs occupying a physiological role ‘upstream’ of EB-RNs. Together, these results support a model in which circadian pacemaker neurons indirectly activate as many as ~60 pairs of EB-RNs by first activating two pairs of dopaminergic neurons, the PPM3-EB.

Figure 5. Daily bimodal pattern of PPM3-EB.

(A) The position of the PPM3-EB neurons central complex in the fly brain (left). Daily bimodal Ca²⁺ activity patterns of PPM3-EB under LD (middle) and DD (right) (n = 6 flies each).

(B) Illustration of dual-color Ca²⁺ imaging: GCaMP6s in PPM3-EB and jRGECO1a in EB-R2 and in circadian neurons.

(C) Example traces of Ca²⁺ activity in EB-R2, PPM3-EB, and l-LNv neurons (sampling rate, 1Hz).

(D) Average Ca²⁺ activity traces from (C) aligned by (left) PPM3-EB peak and (right) l-LNv peak.

(E) As in (D), average Ca²⁺ activity traces from all flies (n = 8 flies).

(F) Correlation of Ca²⁺ activity (Pearson’s r) between EB and PPM3 is considerably stronger than that between EB and l-LNv (p=0.0009, paired t-test after Fisher’s Z-transform).

(G–I) Daily Ca²⁺ activity patterns of PPM3 neurons during DD1 (G) in GAL-only control flies (n = 6 flies), (H) in UAS-only control flies (n = 6 flies), and (I) in M-cell>SGG flies (n = 5 flies).

(J) Phase comparisons of PPM3 neuron activity peaks for both morning (orange) and evening (blue) phases between controls and M-cell>SGG flies. The morning PPM3 peak was significantly advanced in M-cell>SGG, while the evening PPM3 peak was unaffected (Watson-Williams test).

We tested this model by blocking neurotransmission in PPM3-EB neurons, thereby asking if their specific output is necessary for proper locomotor rhythmicity. Using intersectional genetics (GMR92G05-GAL4 and TH-Flp), we restricted the expression of tetanus toxin (TeTn, Sweeney et al., 1995) to the two pairs of PPM3-EB. The locomotor activity of these flies was largely arrhythmic under DD (Figure 6AB). This behavioral deficit was comparable with, and even more severe than that caused by blocking neurotransmission in the majority of EB-RNs (Figure 1B and Table S1). Importantly, while the molecular clocks and Ca²⁺ rhythms of circadian pacemaker neurons in these flies were intact (Figure S7 and Figure 6E), the daily bimodal neural activity pattern of EB-RNs was severely impaired (Figure 6CD). Likewise, knocking down DA receptors DopR2 or D2R in the majority of EB-RNs also impaired rhythmicity in locomotor activity under DD (Figure 6F and Table S1). Significantly, these same DA receptor knockdowns also suppressed the daily bimodal neural activity pattern of EB-RNs (Figure 6G–K). Hence we propose that dopaminergic input from PPM3-EB neurons forms a critical relay to regulate EB-generated locomotor activity, according to a multi-phasic circadian schedule.

Figure 6. PPM3-EB and EB-RNs constitute downstream elements of a circadian output motor circuit.

(A) Group-averaged actograms of control flies (left) and one expressing tetanus toxin (TeTn) in PPM3-EB neurons to block neurotransmission (right).

(B) Average rhythm strength (power) of genotypes in (A) for 9 days under DD; asterisk denotes significant differences compared to control (P < 0.0001, Mann-Whitney test).

(C) Daily Ca²⁺ activity patterns of EB-R2 neurons during DD1 in control flies (solid line, n = 6) and ones with TeTn expressed in PPM3-EB neurons (PPM3>TeTn, dash line, n = 6).

(D) The amplitude of daily morning and evening Ca²⁺ peak in EB-R2 neurons in control and PPM3>TeTn flies. The amplitude difference in the evening peak was significantly different (*P < 0.05, Mann-Whitney test).

(E) Daily Ca²⁺ activity patterns of circadian neurons during DD1 in the same flies from (C): control flies (solid lines, n = 6) and PPM3>TeTn flies (dash line, n = 6).

(F) Average rhythm strength (power) of genotypes for 9 days under DD in which DA receptors are knocked down in EB-RNs using R56H10-gal4; asterisk denotes significant differences compared to control (P < 0.05, Kruskal-Wallis test followed by post hoc Dunn’s tests). See more behavioral controls and statistics in Table S1.

(G–I) Daily Ca²⁺ activity patterns of EB-R neurons under DD1 in (G) WT (n = 8 flies), (H) flies with (H) D2R-knockdown and (I) DopR2-knockdown in EB-R neurons (n = 5 and 5 flies).

(J–K) The amplitude of daily (J) morning and (K) evening Ca²⁺ peak in EB-R neurons was reduced in DA-receptor-knockdown flies (*P < 0.05, Mann-Whitney test).

The daily bimodal pattern of PPM3-EB suggested these dopaminergic neurons may release DA twice a day, once in the morning and again in the evening, to modulate the neural activity of EB-RNs. To test this hypothesis, we first measured the responses of EB-RNs to pharmacogenetic activation of PPM3-EB (Figure 7AB) and to bath-application of dopamine (Figure 7C). Using GRAB_DA, a genetically encoded GPCR-activation-based-DA sensor (Sun et al., 2018), we then asked when and how often during the 24 hr day EB-RNs in vivo receive direct DA modulation. The GRABDA sensor in the EB-R3 subgroup responded to bath application of DA with increased fluorescence (Figure 7D). We then recorded this signal in vivo for 24 hours and observed a spontaneous, daily bimodal pattern of activation (Figure 7E), similar in pattern and phasing to the Ca²⁺ activity pattern of upstream dopaminergic PPM3-EB neurons and to the Ca²⁺ activity pattern of EB-RNs themselves. Collectively, these results suggested that the daily bimodal dopaminergic modulation from PPM-EB to EB-RNs relays two distinct circadian phase points from the pacemaker neurons system. This dopaminergic modulation helps to generate daily bimodal patterns in the neural activity of EB-RNs and thus daily bimodal patterns in locomotor activity (Figure 7F).

Figure 7. EB-RNs receive daily bimodal dopamine inputs from PPM3-EB neurons.

(A) Average traces of EB-R2 neurons and circadian pacemaker neurons labelled by cry-LexA, responding to ATP (left) in flies with P2X2 expressing PPM3-EB DA neurons (n = 5 flies) and (right) UAS-P2X2 only controls (n = 5 flies). Shaded area indicates duration of drug application. Error bars denote SEM.

(B) Maximum Ca²⁺ signal changes in individual cells within 4 min after ATP bath application.

(C) Left - average traces of EB-RNs responding to the bath-application of dopamine (10⁻⁴ M) at ZT6 (n = 4 flies). Right - maximum Ca²⁺ signal changes in individual cells within 3 min after DA bath application.

(D) Left - average traces of change from dopamine sensor GRAB_DA2m in EB-R3 neurons responding to steps of DA bath application (10⁻⁶ M to 10⁻⁵ M to 10⁻⁴ M) at ZT6–8 (n = 4 flies). Right - maximum Ca²⁺ signal changes in individual cells within 2 min after different concentration of DA.

(E) The daily spontaneous bimodal pattern of signal from dopamine sensor GRAB_DA2m in EB-R3 during DD1 (n = 5 flies).

(F) Model of the circadian output pathway for locomotor activity rhythms: circadian pacemaker M cells and E cells independently activate EB-RN pre-motor circuits around dawn and dusk respectively, through a relay by PPM3-EB dopaminergic neurons.

DISCUSSION

Locomotor activity in Drosophila follows a daily bimodal rhythm that peaks around dawn and again around dusk. By measuring spontaneous neural activity in vivo across the 24 hr day, we found that morning and evening circadian oscillators independently activate the premotor Ring Neurons of the Ellipsoid Body through the agency of PPM3-EB dopaminergic neurons. These findings provide the most detailed insights available in any model system by which pre-motor pathways are organized in response to phasic circadian pacemaker information. In addition, they indicate an unexpectedly obligate role for dopamine in the neural control of daily rhythmic locomotor activity. We based our conclusions on four lines of evidence: (1) Both PPM3-EB and EB-RNs display daily spontaneous bimodal neural activity patterns that precisely correlate with locomotor activity patterns peaking around dawn and dusk (Figure 1 & 5A). (2) Locomotor activity closely followed changes in EB-RN activity (Figure 2) while EB-RN activity was itself highly correlated with PPM3-EB activity (Figure 5B–F). (3) Different phases of EB-RN circadian-rhythmic neural activity relied on independent inputs from circadian pacemakers, M cells and E cells, but did not rely at all on visual inputs or on the execution of locomotor behavior (Figure 3 and 4). (4) Both EB-RN activity rhythms and normal locomotor activity rhythms required PPM3-EB inputs (Figure 6A–E); normal locomotor activity rhythms also required DA receptors on EB-RNs to receive inputs from PPM3-EB DA neurons (Figure 6F–K). These data together support a model that features outputs from M cells and E cells sequentially and independently generating the two daily peaks of activity PPM3-EB DA neurons. These non-circadian PPM3-EB DA neurons in turn relay the phasic information to activate as many as ~60 pairs of EB-RNs, thereby generating the bimodal daily locomotor activity rhythm (Figure 7E).

Output circuits downstream of circadian pacemaker neurons

Our findings constitute important steps in relating the activities of distinct circadian pacemaker neurons to downstream neural circuits. Selcho et al. (2017) recently described circadian pacemaker control of a peripheral clock in Drosophila to control steroid hormone secretion whose titers gate subsequent adult emergence (eclosion). In that output pathway, s-LNv activate the peptidergic PTTH neurons, which in turn activate the peripheral Prothoracic Gland. With respect to locomotor behavior, we found that M (s-LNv) cells and E (LNd) oscillators independently control the morning and evening neural activity phases in EB-RNs (Figure 4) and in PPM3-EB (Figure 5G–I). Two recent studies linked a different subset of circadian pacemakers (DN1s) to subgroups of EB-RNs, via subsets of neurons in the Anterior Optic Tubercle (Lamaze et al., 2018; Guo et al., 2018). By manipulating activity in this pathway, both groups found effects on the balance between sleep and wake states. Thus, increasing lines of research indicate circadian- and sleep-regulating circuits impart timing information to govern behavior through the classic pre-motor centers of the Central Complex.

Based on previous limited screens, two other groups of identified peptidergic neurons were implicated as components of output circuits for locomotor activity rhythms in Drosophila. The two groups included ones that express the diuretic hormone 44 (DH44), an orthologue of mammalian CRF (Cavanaugh et al., 2014), and ones that express leucokinin (LK) (Cavey et al., 2016), whose receptor is related to the neurokinin receptors. DH44 neurons receive synaptic inputs from DN1 pacemaker neurons, and both DH44- and LK- neurons are required for proper locomotor activity rhythms under constant darkness (DD) conditions. However, the connectivity by which these two groups of neuroendocrine neurons promote locomotor activity, and phase-restrict it to morning or evening times, is uncertain. The daily two-peak pattern of locomotor activity is different from the daily activity pattern of either LK neurons (which are more active in the evening - Cavey et al., 2016), or that of DH44 neurons (which are more active in mid-day -Bai et al., 2018). Additionally, recent studies showed that LK neurons also mediate hunger signals to promote locomotor activity (Yurgel et al., 2018; Zandawala et al., 2018). These observations suggest that during daily locomotor activity peaks, flies might move to seek food, seek a mate, and/or respond to other internal drives, which might be coupled with circadian timing. Therefore, several parallel pathways may converge within the EB-RN pre-motor circuit to generate and shape daily behavioral patterns.

Dopaminergic neurons under circadian regulation

Previous studies in flies and mice have shown that DA modulates circadian pacemaker circuits (Chang et al., 2006; Grippo et al., 2017; Hirsh et al., 2010; Klose et al., 2016; Langraf et al, 2016; Shang et al., 2011, 2013). Our findings here show that circadian pacemaker neurons also regulate DA neuron activity. DA neurons responded to circadian neuron outputs (Figure 4AB) and showed spontaneous circadian neural activity rhythms that were correlated with behavior (Figure 5). These findings correspond to earlier studies in mammals showing that circadian rhythms in DA neuron activity, and in striatal DA content, are dependent on master circadian pacemaker neurons in the suprachiasmatic nuclei (SCN) (Smith et al., 1992; Sleipness et al., 2006; Luo et al., 2008; Fifel et al., 2018). Deficits of DA neurons in patients and in animal models of Parkinson’s disease caused dysregulation of circadian locomotor activity patterns and of sleep (Videnovic and Golombek, 2017). Consistent with our model, a DA-deficient mouse model displays dampened and fragmented locomotor activity rhythms, yet possesses normal SCN molecular clocks (Taylor et al., 2009; Kudo et al., 2011). It remains to be determined whether DA in mammals, as in Drosophila, represents the critical agent by which circadian outputs activate pre-motor centers to adaptively schedule locomotor activity.

The effects of DA to organize proper circadian control of locomotor behavior may be related to its well documented effects in Drosophila to promote arousal, especially forms of arousal associated with changes in sleep and circadian rhythm states (Andretic et al., 2005; Birman, 2005; Kume et al., 2005; Lima and Miesenbock, 2005; Lebestky et al., 2009; Liu et al., 2012). A recent study suggests that PDF signals from circadian neurons promote wakefulness by suppressing daytime activity in the PPM3 DA neurons (Potdar and Sheeba 2018). However, we favor an alternative model which is based on the results described above, including both manipulations of PPM3 physiology as well as measurements of normal PPM3 24-hr activity patterns in vivo. We propose PPM3-DA neurons promote wakefulness and locomotor activity in the morning following excitation by the M oscillators, and perhaps directly by PDF.

How can circadian timing signals modulate activity in central complex to promote rhythmic locomotor activity?

Our results suggest that a major focus of circadian timing signals to regulate locomotor activity resides in the Central Complex (CX), the decision-making circuit that dictates the balance between locomotion and rest. Within the CX, EB neurons transform sensory inputs into goal-directed motor outputs (Sun et al., 2017; Shiozaki and Kazama, 2017). The final motor output is subject to many signals reflecting the internal state: for example, hunger signals transmitted through the leucokinin-expressing neurons promote locomotor activity (Yurgel et al., 2018; Zandawala et al., 2018). Here, we propose that the circadian system promotes locomotor activity in the dawn and dusk episodes by increasing the probability of the EB-RNs to favor activity over rest. A similar action on EB-RNs appears to underlie sleep promotion by dorsal fan-shaped body (dFSB) neurons (Donlea et al., 2017). dFSB neurons effectively suppress sensory-triggered movements by inhibiting EB-RNs via helicon cells and thereby instigate less activity and more rest. Thus, sleep and circadian signaling antagonistically converge on the EB-RN system to influence the level of motor output.

In addition to motor outputs, parts of EB circuit also signal the sleep drive: Liu et al. (2016) showed that a subgroup of EB-RNs, R2 (called R5 by Omoto et al., 2017) registers sleep debt and thereby constitutes an integral part of the sleep homeostat mechanism. How can this be reconciled with our finding that the EB-RNs (including R2s) exhibit neural activity in concert with locomotor behavior? We propose that, because the level of locomotor activity is directly encoded by EB-RN activity, a subgroup of them (R2s) incorporates the amount of locomotor activity along with duration of wakefulness to help generate sleep drive. Therefore, although they receive common circadian pacemaker and DA inputs, and although they exhibit common activation periods at dawn and at dusk, different subgroups of EB-RNs likely have specialized downstream functions in behavioral control.

Modulation of locomotion in vertebrates

The circuit features described by this work link pacemaker centers to premotor centers via specific DA interneurons: could there be parallels across phylogeny? We note that a region between the vertebrate midbrain and hindbrain - called the Mesencephalic Locomotor Region (MLR) - is implicated in the initiation and control of locomotion in numerous species, including lamprey, cat, rat, and monkey (reviewed by Rycko and Duboc, 2013). Stimulation of the MLR elicits coordinated locomotion with varying speed and gait, depending on stimulus location (Caggiano et al., 2018). The MLR is heterogenous and includes the cuneiform and sub-cuneiform nuclei and the pedunculopontine nucleus. A traditional view held that DA-modulated MLR activity indirectly, via its projections to the Basal Ganglia (Surmeier et al., 2007; Rycko and Duboc, 2013). In lampreys however, the majority of DA neurons extend descending projections to the MLR that can modulate locomotor output (Ryczko et al., 2013). More recently, Ryczko et al (2016) described direct DA inputs to MLR in the rat, from neurons that project to the striatum but also descend to the pedunculopontine nucleus. These authors propose that these direct DA inputs amplify subsequent MLR descending locomotor command. Also, Sharma et al., (2018) have described direct descending projections from A13 DA neurons of the mouse medial Zona Incerta to MLR nuclei. These authors speculate that A13 DA actions in the MLR may exhibit especially long time-courses, because these neurons lack expression of the Dopamine Transporter. Together these observations support the hypothesis of a phylogenetically-conserved DA to premotor pathway in vertebrates that amplifies signals to initiate and/or shape episodic locomotion. Thus, it may be worth investigating whether there exist functional connections between SCN circadian pacemaker outputs and these descending DA cell groups.

STAR Methods

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for reagents may be directed to and will be fulfilled by the Lead Contact, Paul H. Taghert (taghertp@wustl.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Fly stocks.

Flies were reared on standard cornmeal/agar food at room temperature. Before imaging experiments, flies were entrained under 12 h light: 12 h dark (LD) cycles at 25°C for at least 3 days. Male flies, 3–10 days old, were selected for imaging. The following fly lines were previously described: tim(UAS)-GAL4 (Blau & Young 1999), pdfr(F)-GAL4 and pdfr(B)-GAL4 (Im & Taghert 2011), GMR56H10-GAL4 (Sun et al., 2017), GMR69F08-GAL4 (Liu et al., 2016); split-GAL4 lines: GMR_MB122B and GMR_SS00681 (Liang et al., 2017), GMR_SS002769 (Robie et al., 2017); cry-LexA (Liang et al., 2017), pdf-LexA (Shang et al., 2008), TH-LexA (Berry et al 2015); TH-Flp (Xie et al., 2018); UAS-SGG (Martinek et al., 2001), UAS-P2X2 and LexAop-P2X2 (Yao et al., 2002), LexAop-jGECO1a (Dana et al., 2016), UAS-GCaMP6s and LexAop-GCaMP6s (Chen et al., 2013), UAS-GRAB_DA4.4 (Sun et al., 2018), UAS-DopR1-miRNA and UAS-DopR2-miRNA (Liu et al., 2017), UAS-D2R-miRNA and UAS-DopEcR-miRNA (Xie et al., 2018); per⁰¹ (Konopka & Benzer 1971), norpA^P24 (Ostroy & Pak 1974), and pdfr^han5403 (Hyun et al., 2005). UAS-(FRT.stop)-TeTn (BL67690), GMR19C08-LexA (BL52543), GMR56H10-GAL4 (BL61644), and GMR92G05-GAL4 (BL48416) were obtained from Bloomington Stock Center. The cry-LexA line was a gift from Dr. F Rouyer (CNRS Gyf, Paris). All transgenic lines were either generated in the w1118 (iso31) background or backcrossed more than 4 generations into this background. For mutant flies, per⁰¹, norpA^P24, and pdfr^han5403, the phenotypes of behavioral rhythms and circadian pacemaker neural activity rhythms were measured to help interpret EB-RN neural activity rhythms.

METHOD DETAILS

Nomenclature.

The nomenclature of ellipsoid body (EB) subgroups in this study follows Renn et al., (1999a), which was revised by Omoto et al., (2017) reflecting the introduction of more specific driver lines. The EB subgroup labelled by cry-lexA and GMR69F08-GAL4 (also see Liu et al., 2016) was called R2, they were re-named R5 by Omoto et al. (2017). The EB subgroup labelled by GMR19C08(pdfr)-lexA was called R4, they were re-named R2 by Omoto et al. (2017).

In vivo fly preparations.

The surgical procedure for Drosophila in vivo calcium imaging followed methods described in Liang et al., (2016, 2017). Following CO₂ anesthetization, flies were mounted by inserting the neck into a narrow cut in an aluminum foil base. Thus, the foil permitted immersion of the head by saline during preparatory surgery and in vivo imaging, while the body remained in an air-filled enclosure. To access circadian pacemaker neurons on one side of the head, a single antenna, a portion of the dorso-anterior head capsule, and a small part of one compound eye were removed from the side ipsilateral to imaging. To access EB-RNs, both antennae and a portion of the dorso-anterior head capsule were removed, while the compound eyes remained intact. The entire surgery was typically ~15 min in duration. For experiments that entailed transection of connectives, or removal of the entire body, the surgery was conducted with fine forceps prior to brain-exposing surgery. The wounds were then closed by application of a bio-compatible silicone adhesive (Kwik-Sil, WPI, USA).

In vivo calcium imaging.

Imaging was conducted with custom Objective Coupled Planar Illumination (OCPI) microscopes (Holekamp et al., 2008), as described in Liang et al., (2016, 2017). Briefly, OCPI uses a cylindrical lens to generate a ~5μm thick light sheet, which was coupled to the focal plane of the objective. For 24-hr imaging, the objective coupled light sheet was scanned across the fly brain through the cranial window every 10 min to capture stacks of images. Each stack contained 20 to 40 separate images with a step size of 5 to 10 microns. For each image, exposure time was not more than 0.1 s. During 24-hr imaging, fresh hemolymph-like saline (HL3; 5 mM KCl, 1.5 mM CaCl2, 70 mM NaCl, 20 mM MgCl2, 10 mM NaHCO3, 5 mM trehalose, 115 mM sucrose, and 5 mM HEPES; pH 7.1) was perfused continuously (0.1–0.2 mL/min). Light-dark cycle stimulation during in vivo calcium imaging was delivered using a white Rebel LED (Luxeon) controlled by an Arduino UNO board (Smart Projects, Italy) as described in Liang et al., (2017). High-frequency imaging was conducted with a speed-optimized OCPI microscope (Greer and Holy, 2018). Image stacks were captured every 10 s (Figure S5A and S6E), every 2 s (Figure 4A and 7A–C, Figure S5B–F), or every 1 s (Figure 5C–F, Figure S3G and S6D). For each image, exposure time was not more than 0.04 s. For pharmacological tests, each fly was treated once. After 1 or 5-min baseline recordings, 1mL of 0.1 mM PDF solution, 1 mM dopamine solution, or 10 mM ATP solution (pH adjusted to 7) was manually added to a 9 mL static HL3 bath over a ~2 s period. PDF was purchased from Neo-MPS (San Diego, CA, USA) at a purity of 86%.

Cameleon-based calcium imaging.

For Figure S2C, FRET imaging was performed as described in Liang et al., (2016). At certain ZTs, living brains were dissected and imaged with an Epifluorescent Olympus BX61 microscope. Exposure times were 20 ms for YFP - FRET and 500 ms for CFP donor. For each ZT point, the ratio values of YFP over CFP were collected from at least 45 cells that were found in at least 5 brains.

In vivo dopamine imaging.

Dopamine imaging was conducted following the same procedure as in vivo calcium imaging using the OCPI microscope. The dopamine sensor GRAB_DA2m is derived from DA1m (Sun et al. 2018), but contains additional point mutations in the inserted cpEGFP. These mutations significantly increase dF/F ~ 3 fold comparing with DA1m.

Locomotor monitoring during imaging.

During 24-hr in vivo calcium imaging, Drosophila locomotor activity was measured by an infrared detector (LTE-301)/emitter (940nm, LTE-302) circuit. The infrared emitter was aimed toward the body of the fly and the detector received the infrared light transmitted through the fly (shown in Figure 2A). Both the body and leg movements can cause changes in transmitted light intensity. The analog signal from the infrared detector was transmitted through an Arduino UNO board with 100Hz sampling rate. The infrared emitter was shut off for 10 seconds every 10 min, allowing the microscope to acquire complete volume brain scans. The daily fly locomotor activity pattern was then calculated by counting the activity events within each 10-min bin. The activity events were identified by time-points when the infrared detector signal was out of the range for standard deviation by 3-fold. Then the normalized event count trace was aligned with the EB-R2 neuron calcium signal of the same fly (Figure 2c). The Pearson’s correlation coefficient between these two signals was calculated. To test their correlation at an hourly time scale, these two signals then were averaged by a method similar to spike-triggered averaging. 4-hr windows (1 hr before and 3 hr after the trigger point) of calcium signals were aligned by the local maximum (increasing phase) or local minimum (decreasing phase) of calcium signal derivatives. The locomotor signals occurring in these 4-hr windows were then averaged. Analysis was performed using R 3.3.3.

Locomotor activity.

To examine the circadian rhythms of locomotor activity, individual flies was monitored using Trikinetics Drosophila Activity Monitor (DAM) system for 6 days under light-dark (LD) cycles and then for 9 days under constant darkness (DD) condition. χ² periodogram with a 95% confidence cutoff and SNR analysis were used to measure circadian rhythmicity and periodicity (Levine et al., 2002). Arrhythmicity were defined by a power value (χ² power at best period) less than 10, width lower than 1, a period less than 18 hrs or more than 30 hrs. To find the phases of morning and evening peaks, each 24-hr day was split into two halves. For LD, it was split at ZT6. For DD1, it was split at the time of the manually selected midday “siesta”. Then the morning peak and evening peak were then determined by the maximum activity in each half.

Immunocytochemistry.

Immunostaining for PER and beta-Gal followed previous descriptions (Liang et al., 2016). Briefly, fly brains were dissected in ice-cold, calcium-free saline and fixed for 15m in 4% paraformaldehyde containing 7% picric acid (v/v) in PBS. Primary antibodies included rabbit anti-PER (1:5000; kindly provided by Dr. M. Rosbash, Brandeis Univ.; Stanewsky et al., 1997) and mouse anti-beta-galactosidase (1:1000; Promega, Madison, WI, Cat. #Z3781, Lot #149211). Secondary antisera were Cy3-conjugated (1:1000; Jackson Immunoresearch, West Grove, PA). Images were acquired on the Olympus FV1200 confocal microscope. PER protein immunostaining intensity was measured in ImageJ-based Fiji (Schindelin et al., 2012).

Data reporting.

No statistical methods were used to predetermine sample sizes. The selection of flies from vials for imaging and behavioral tests were randomized. The investigators were not blinded to fly genotypes.

QUANTIFICATION AND STATISTICAL ANALYSIS

Imaging data analysis.

Calcium imaging data analysis was as described previously (Liang et al., 2016, 2017). Images were acquired by a custom software, Imagine (Holekamp et al, 2008) and processed in Julia 0.6 including non-rigid registration, alignment and maximal projection along z-axis. Then ImageJ-based Fiji was used for rigid registration and to manually select regions of interest (ROIs) over individual cells or groups of cells. Average intensities of ROIs were measured through the time course and divided by average of the whole image to subtract background noise. For spontaneous calcium transients, each time trace was then calculated as ΔF/F=(F-F_min)/F_mean. The minimal intensity over 24-hr recording was regarded as the baseline intensity in order to average the ΔF/F across different trials. In these different trials, to avoid the influence of the single-photon imaging preparation and the ambient recording environment, 24-hr recording sessions were randomly begun at different Zeitgeber Times. Consequently, the calcium levels at the first timepoints (F0) could be either at the peak, the trough, or in the middle of daily calcium rhythms. In order to compare between different trials, F_min was used as the baseline. The ΔF then was divided by F_mean rather than F_min, because F_min (the baseline fluorescence of GCaMP6s) may be zero (below black level of the camera) during 24-hr recordings. For 24-hr time traces, traces of certain cell type ROIs were first aligned, based on Zeitgeber Time and averaged across different flies. Hartigans’ dip test and Silverman’s test were used to testify whether the averaged 24-h time traces are unimodal or bimodal (Hartigan & Hartigan, 1985; Silverman, 1981). The phase relationship between traces was estimated by cross-correlation analysis. The 24-hr-clock circular plot of phases reflected both mean peak time and phase relationships of the same cell-group traces from different flies. For neurons with daily bimodal patterns (EB-RNs and PPM3-EB DA neurons), each trace was split into two parts: ZT18-ZT6 (morning) and ZT6-ZT18 (evening) to estimate the morning and evening peak phases respectively. For dual-color imaging traces, all signals were filtered (high-pass, 1/30 Hz). To ‘spike’-triggered average simultaneous traces of three cell types (Figure 5DE), the peaks of selected cell-type signal were identified by the local maximum of that signal after a low-pass filter (0.2Hz). Unfiltered signals of three cell types were then aligned by these peaks to calculate the averaged traces for individual cell types. For pharmacological calcium responses, each time trace was normalized by the initial intensity (F/F0). The maximum change was calculated by the maximum difference of normalized intensities between baseline and after drug application. The latency (onset time constant) was calculated by the duration from drug application to the time when the trace reached 63.2% of maximum change. The parametric or nonparametric tests were selected based on F variance tests. All statistics tests are two-sided. All the sample size information (n values and what the n represents), as well as specific statistical methods, are listed in corresponding figure legends. Trace analysis and statistics were performed using R 3.3.3 and Prism 7 (GraphPad, San Diego CA).

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Chemicals, Peptides, and Recombinant Proteins
Adenosine 5′-triphosphate	Sigma	A5394
Dopamine hydrochloride	Sigma	H8502
Drosophila pigment dispersing factor (PDF)	Neo-MPS	NA
Experimental Models: Organisms/Strains
Drosophila: tim(UAS)-GAL4	(Blau and Young, 1999)	NA
Drosophila: pdfr(B)-GAL4	Taghert Lab, (Im and Taghert, 2010)	NA
Drosophila: pdfr(F)-GAL4	Taghert Lab, (Im and Taghert, 2010)	NA
Drosophila: pdf-LexA	(Shang et al., 2008)	NA
Drosophila: LexAop-nSyb-spGFPl-10, UAS-CD4-spGFPll	Bloomington (Frank et al., 2015)	RRID:BDSC_64315
Drosophila: 20XUAS-IVS-GCaMP6s(attP40)	Bloomington (Chen et al., 2013)	RRID: BDSC_42746
Drosophila: 13XLexAop2-IVS-GCaMP6s-plO(su(Hw)attPl)	Bloomington (Chen et al., 2013)	RRID: BDSC_44274
Drosophila: LexAop-jGECOla	Bloomington (Dana et al., 2016)	RRID: BDSC_63794
Drosophila: UAS-(FRT.stop)-TeTn	Bloomington	RRID: BDSC_67690
Drosophila: UAS-SGG	Bloomington (Martinek et al., 2001)	RRID: BDSC_5435
Drosophila: UAS-GRABDA4,4	Yulong Li (Sun et al., 2018)	NA
Drosophila: UAS-P2X2	Orie Shafer (Yao et al., 2002)	NA
Drosophila: UAS-cameleon2.1	Bloomington (Miyawaki et al. 1999)	RRID: BDSC_6901
Drosophila: LexAop-P2X2	Orie Shafer (Yao et al., 2002)	NA
Drosophila: cry-LexA::GAD	F. Rouyer (CNRS Gyf, Paris) (Liang et al., 2017)	NA
Drosophila: TH-LexA	(Berry et al 2015)	NA
Drosophila: GMR_SS00681	Gifts from Drs. Dionne, Nern and Rubin (Janelia Research Center, VA) (Liang et al.,2017)	NA
Drosophila: GMR_MB122B	Gifts from Drs. Dionne, Nern and Rubin (Janelia Research Center, VA) (Liang et al.,2017)	NA
Drosophila: GMR_SS002769	Janelia Research Center (Robie et al., 2017)	NA
Drosophila: GMR56H10-GAL4	Bloomington (Sun et al., 2017)	RRID:BDSC_61644
Drosophila: GMR96F08-GAL4	Bloomington (Liu et al., 2016)	RRID: BDSC_39499
Drosophila: GMR92G05-GAL4	Bloomington (Xie et al., 2018)	RRID: BDSC_48416
Drosophila: GMR19C08-LexA	Bloomington	RRID: BDSC_53543
Drosophila: norpAP24	Bloomington (Ostroy & Pak 1974)	RRID: BDSC_9048
Drosophila: per01	(Konopka and Benzer, 1971)	NA
Drosophila: pdfrhan5403	(Hyun et al., 2005)	NA
Drosophila: UAS-DopRl-miRNA	Wu Lab (Liu et al., 2017)	NA
Drosophila: UAS-DopR2-miRNA	Wu Lab (Liu et al., 2017)	NA
Drosophila: UAS-D2R-miRNA	Wu Lab (Xie et al., 2018)	NA
Drosophila: UAS-DopEcR-miRNA	Wu Lab (Xie et al., 2018)	NA
Drosophila: TH-Flp	Wu Lab (Xie et al., 2018)	NA
Software and Algorithms
R	http://www.R-project.org/	Version: 3.3.3
Julia	https://julialang.org/	Version: 0.6
Prism 7	GraphPad	https://www.graphpad.com/
Fiji	(Schindelin et al., 2012)	https://fiji.sc/
Imagine	(Holekamp et al., 2008)	http://holylab.wustl.edu/

Highlights.

• The Drosophila circadian pacemaker network promotes rhythmic daily activity

• EB ring neurons (EB-RNs) exhibit spontaneous morning and evening activity peaks

• The two EB-RN peaks are independently driven by M and E circadian pacemakers

• Both M and E pacemakers regulate EB-RN activity via specific dopamine interneurons