Calcium and cAMP directly modulate the speed of the Drosophila circadian clock

In order to investigate the role of calcium in regulating circadian clock function, we determined the consequences on the circadian rhythm of emergence of knocking down, specifically in the PG, genes involved in calcium homeostasis. As shown in Fig 1A, reducing the expression of two different calcium channels (cacophony and Ca-alpha1D; Fig 1Aa and 1Ab) and of the IP3 (Inositol 1,4,5, -tris-phosphate) receptor (IP3R; Fig 1Ac) significantly lengthened the periodicity of adult emergence (Fig 1D), without affecting the strength of the rhythmicity (Fig 1E). By contrast, knockdown of SERCA (sarco/endoplasmic reticulum calcium ATPase) in the PG resulted in a significant shortening of the periodicity of eclosion (Fig 1Ba and 1D), also without significantly altering its associated rhythmicity index (Fig 1E). (In this case, SERCA was knocked down acutely at the end of metamorphosis using the GAL80^ts temperature-sensitive genetic system [37] because SERCA RNAi expression in the PG throughout life was lethal.) Such manipulations are expected to cause decreases and increases in calcium levels for cacophony, Ca-alpha1D, and IP3R, and SERCA, respectively, and direct measurement of calcium levels in the PG showed that this was indeed the case (cf. Fig 2C, below). We also probed the possible role of CaMKII (calcium/calmodulin-dependent protein kinase I) and CASK (calmodulin-dependent protein kinase activity) in the PG to determine if target proteins of calcium signaling might have an effect on the eclosion rhythm. As shown Fig 1C and 1D knockdown of CaMKII and of CASK in the PG (Fig 1Ca and Fig 1Cb, respectively) caused a significant lengthening of the periodicity of eclosion compared to that of the control (Fig 1D), without affecting strength of rhythmicity (Fig 1E). These results show that changing the levels of intracellular calcium affects the periodicity of the circadian clock. These changes can be effected by reducing calcium entry (e.g., through CACOPHONY or CA-ALPHA1D) or release from intracellular stores (e.g., via IP3), and could be mediated by CaMKII or CASK. Flies bearing single UAS-RNAi insertions for genes in calcium homeostasis pathway expressed normal circadian rhythmicity of eclosion (S1 Fig).

We measured intracellular calcium levels in the PG in vivo using the genetically-encoded sensor, GCaMP 3.2 (GCaMP) [38]. This sensor has been extensively used and causes no apparent cytotoxicity or behavioral defects [39]. In particular, it caused no defects in the circadian rhythmicity of eclosion (S2A–S2C Fig) or locomotor activity (S2D–S2F Fig) when expressed in the PG using the phm-GAL4 driver [19,40]. We determined the dynamic range and sensitivity of this sensor by incubating PGs with different concentrations of calcium in the presence of 50 μM calcimycin, a calcium ionophore (Fig 2A). As shown in Fig 2B, we found that the readout from the GCaMP sensor increased monotonically between 0 and 300nM calcium.

We then measured changes in calcium levels in the PG during the course of the day (and subjective day) at the beginning of metamorphosis (white pre-pupal stage, WPP). We chose this stage because the PG is intact this time and the circadian clock is fully functional [41], unlike in animals prior to emergence, when PG cells are undergoing apoptosis [42], which, in our hands, rendered such measurements variable and inconsistent. Overall, GCaMP readings were between 0.6 and 0.8, which corresponds to 50-300nM (see Fig 2B), and is consistent with published values. For instance, work done using Manduca sexta PG using Fura-2-loaded PG cells, reported a basal calcium concentration in the range of 50-200nM and maximal levels of 300-500nM following stimulation with the PTTH neuropeptide [43]. Importantly, genetic manipulations expected to increase and decrease the levels of calcium (expression in the PG of RNAi to cacophony, Ca-alpha1D, and IP3R, and RNAi to SERCA, respectively) did indeed cause the corresponding changes in GCaMP fluorescence (Fig 2C). This indicates that the GCaMP sensor can detect physiologically-relevant changes in intracellular calcium levels in the PG. We then measured the GCaMP signal at different times of day in the PG of wildtype animals of the same developmental age (WPP stage) under a 12h light: 12h dark regime (LD) (Fig 2D) and under conditions of constant darkness and temperature (DD)(Fig 2E). Frequency analyses of these data using JTK_cycle [44] and Lomb-Scargle [45] failed to reveal any significant circadian rhythmicity. Nevertheless, a more restricted ANOVA analysis revealed that calcium levels reached a minimum at 12h after lights on (beginning of the night; typically referred to as Zeitgeber (ZT) time 12, where ZT0 is lights-on) under LD conditions (Fig 2D), which was delayed about 3h under DD conditions (circadian time (CT) 15, early subjective night; where CT0 is the start of the subjective day)(Fig 2E). The absence of circadian rhythmicity in these data is surprising. However, calcium oscillations in the PG are known to be influenced by the brain [41]. Thus, it is possible that the timecourse we recorded in the PG reflects the sum of two circadian inputs, one driven by the brain, and the other produced by the PG itself. Our results contrast with those obtained previously, by Morioka and colleagues [41], which showed that baseline intracellular calcium concentrations in the PG reached a single minimum at the beginning of the day under LD condition. However, their readings were done every 6 hours and skipped the timepoint at which we observed an even lower (and minimal) level. In arrhythmic per⁰ mutants, we observed a single minimum under LD conditions at ZT9 (Fig 2F); we observed no significant changes under DD conditions (Fig 2G), as would be expected for an arrhythmic genotype.

In parallel experiments we evaluated the role of cAMP in the control of the circadian clock that controls the timing of adult fly emergence. We found that knockdown of the cAMP phosphodiesterase (encoded by the dunce gene) in the PG shortened the periodicity of eclosion (Fig 3Aa and 3B), whereas knockdown in the PG of the calcium-dependent adenylate cyclase (encoded by the rutabaga gene), lengthened the periodicity of eclosion (Fig 3Ab and 3B), without affecting the strength of the rhythms (Fig 3C). Such manipulations are expected to cause increases and decreases in cAMP levels, respectively, and direct measurement of cAMP levels in the PG showed that this was indeed the case (cf. Fig 4C).

Furthermore, RNAi inhibition of others elements of the cAMP pathway in the PG also affected the periodicity of eclosion. Indeed, knockdown of Epac (Exchange protein directly activated by cAMP) and Rap1 (Ras-related protein 1 GTPase) in the PG caused a significant lengthening of the periodicity of eclosion (Fig 3Ac, 3Ad and 3B). Flies bearing single UAS-RNAi insertions for genes in cAMP pathway expressed normal circadian rhythmicity of eclosion (S3 Fig)

We next monitored cytoplasmic cAMP levels in the PG, using the genetically-encoded Epac1 camps cAMP sensor at the beginning of metamorphosis. This sensor has previously been used to measure cAMP levels in Drosophila neurons [31,33], and driving its expression in the PG did not affect the rhythmicity of eclosion (S4A–S4C Fig) or of locomotor activity (S4D–S4F Fig). Stimulating the PG with 100 μM forskolin (FSK), an activator of most forms of adenylyl cyclase [46], caused the Epac1 signal to fall (as expected for a FRET-based sensor, for which an increase in cAMP would cause an decrease in signal) to a steady state level within 10 min (Fig 4A). In order to determine the sensitivity of this sensor, we measured the signal produced by increasing concentrations of 8-Br-cAMP, a membrane permeable analog of cAMP (Fig 4B). The FRET signal in the PG was highest for basal conditions, and decreased monotonically until 8-Br-cAMP concentrations of around 700nM. Importantly, genetic manipulations expected to increase and decrease the levels of cAMP (expression in the PG of RNAi of dunce and rutabaga, respectively) did indeed cause the corresponding changes in Epac1 FRET signal, indicating that the Epac1 sensor can detect physiologically-relevant changes in intracellular cAMP levels in the PG (Fig 4C).

We then measured the FRET signal in the PG at different times of day in wildtype animals of the same developmental age (WPP stage). As was the case for the calcium measurements, we did not detect any significant rhythmicity in the circadian range using the JTK_cycle [44] and Lomb-Scargle [45] spectral analyses. However, ANOVA analysis did reveal that under LD conditions cAMP levels were highest (lowest FRET signal) at ZT18, and lowest (highest FRET signal) around lights-off (ZT12)(Fig 4D). Under DD conditions highest cAMP levels were delayed by 3h (to CT0), whereas the lowest cAMP levels occurred at the middle of the subjective day (CT6)(Fig 4E). Finally, in arrhythmic per⁰ mutants, under LD conditions cAMP levels were highest at ZT18 and lowest around ZT21 (Fig 4F; similarly to rhythmic controls under LD conditions; Fig 4D); we observed no significant changes under DD conditions (Fig 4G), as would be expected for an arrhythmic genotype.

The effects of manipulating calcium and cAMP levels in the PG on the periodicity of emergence suggest that these messengers could directly alter the clock TTFLs, and a variety of indirect evidence [30,47] suggests that they may act on the TTFL via CREB. As shown in Fig 5Aa and 5B, knockdown of CREB in the PG caused the expression of an arrhythmic pattern of emergence. Furthermore, using cre-luciferase as a reporter for clock output [29] showed that reducing CREB expression in the PG rendered this clock arrhythmic (Fig 5D). An arrhythmic pattern of emergence was also obtained when CREB was knocked down in all clock cells (Fig 5Ab and 5B), and when it was knocked down in all clock cells except for the PG clock (Fig 5Ac and 5B), compared to controls (Fig 5B and 5C) indicating that CREB plays a key role in all clock cells. Flies bearing transgenes for CREB RNAi for alone, and for the phm-GAL4 and tim-GAL4 drivers alone expressed normal circadian rhythmicity of eclosion (S3 Fig and S1 Fig, respectively).

Our results show that manipulations of calcium and cAMP levels in the PG affect the rhythmicity of eclosion by directly affecting PG clock function, and that these actions may be mediated by CREB. In order to determine if a similar situation obtains for the central clock, we investigated the consequences on the circadian rhythmicity of eclosion of manipulating calcium and cAMP levels in central brain pacemaker neurons. Consistent with the results obtained for the PG, manipulations expected to decrease calcium (knockdown of calcium channels, cacophony and Ca-alpha1D) and cAMP levels (knockdown of rutabaga) in critical Small Ventral Lateral pacemaker neurons (sLNv) using the pdf-GAL4 driver, lengthened of the periodicity of eclosion (Fig 6Aa–6Ac and 6C). By contrast, knockdown of dunce, which is expected to increase the levels of cAMP in these pacemaker neurons, caused a shortening of the periodicity of eclosion (Fig 6Ad and 6C). Finally, similar manipulations using the timeless-GAL4 (tim-GAL4) driver, which is expressed in both central and peripheral pacemaker cells, produced similar results (Fig 6B and 6D). Flies bearing the pdf-gGAL4 driver alone expressed normal circadian rhythmicity of eclosion (S1 Fig). These results are comparable to the ones obtained following similar manipulations in the PG, and suggests that calcium and cAMP signaling plays a similar role in the central and the peripheral clock with respect to the circadian control of adult emergence.

Finally, we investigated the effects of manipulating calcium and cAMP levels in clock neurons on the circadian rhythm of locomotor activity of adult flies. As shown in Fig 7A and 7C and also summarized in Table 1, knockdown in sLNv neurons (using the pdf-GAL4 driver) of dunce shortened (Fig 7Ab, Fig 7C), whereas knockdown of rutabaga and IP3R lengthened (Fig 7Aa and 7Ac, Fig 7C), the periodicity of the locomotor activity rhythm. Previous work that manipulated calcium levels in sLNv by expressing buffer protein parvalbumin (PV) showed that PV overexpression eventually caused a lengthening of the free-running period [27]. Yet, this result was only visible in flies bearing several (>2) copies of the UAS-PV transgene and also, somewhat paradoxically, only after several weeks in DD. The relatively modest lengthening obtained when expressing IP3R RNAi in sLNv neurons, and the lack of detectable change following knockdown of cacophony and Ca-alpha1D (Fig 7Ca) may be due to low effectiveness of these RNAi lines in causing significant changes in calcium levels, and/or may require a longer monitoring period under DD conditions. Finally, and in line with previous work [27] knockdown of these genes in all clock cells (Fig 7B and 7D; Table 1) severely disrupted the free-running rhythms of locomotor activity. Similarly, knockdown of CREB in sLNv neurons (Fig 7Ad, Fig 7C) or in all clock neurons (Fig 7Bd, Fig 7D) caused the expression of an arrhythmic pattern of locomotor activity. Overall, these results suggest that calcium and cAMP levels affect clock function itself.

Mammalian and insect circadian clocks are produced by intracellular transcriptional/translational feedback loops (TTFL). In the case of neuronal circadian pacemakers, the activity of the cells that make up the pacemaker is then coordinated through transmitter and neuropeptide action. In the mammalian SCN, these messengers cause changes in the intracellular levels of calcium or cAMP and are known to directly affect the TTFL via CREB-mediated changes of transcription [24,48]. In Drosophila, by contrast, the relationship between changes in calcium and cAMP and the functioning of TTFLs is less clear. For example, different neurons within the central brain pacemaker exhibit distinctly different phases of calcium oscillation while expressing essentially identically-timed rhythms of clock gene expression [49]. This suggests that the pattern of neuronal activity of different fly pacemaker neurons may be an emergent property that depends on their interconnection. Likewise, use of a CRE-luc transgene to probe CREB expression in the Drosophila brain [50] showed that most neuronal clusters, including non-clock neurons, expressed a circadian rhythmicity of bioluminescence, suggesting that CREB could be part of the clock’s output mechanism.

Here we explored directly the relationship between calcium and cAMP levels and clock function using the simpler, non-neuronal, peripheral clock housed within the PG endocrine gland. We found small but significant changes in calcium and cAMP levels during the course of the day, which persisted under constant conditions, suggesting that these second messengers could be part of the output of the clock. These changes did not show a circadian rhythmicity. This contrasts with previously reported measurements of cAMP levels in the adult fly brain, which showed a single maximum [28] and may reflect the fact that the PG clock is a slave of the brain clock [19,35]. Importantly, however, we found that manipulations that increased or decreased the levels of calcium and cAMP in the PG caused, respectively, a shortening and lengthening of the periodicity of eclosion, suggesting that they may also affect clock TTFL function. In the case of calcium, these manipulations were effective when they targeted calcium channels as well as elements involved in calcium storage, revealing that both mechanisms could mediate changes in clock periodicity. Furthermore, PG-specific knockdown of CaMKII and CASK also affected clock period, suggesting that the effects of changes in calcium levels on the clock could be effected by phosphorylation. These results suggest that the effects on locomotor activity rhythmicity caused by reducing calcium levels in central neurons [27] could, at least in part, also be due to actions on the intracellular TTFLs. For the case of cAMP, our results similarly suggest that the shortened circadian periodicity of adult locomotor activity rhythmicity reported for dunce mutants [28], may be due to direct actions on the brain clock TTFL.

How might calcium and cAMP ultimately affect the Drosophila clock TTFL? Circumstantial evidence suggests that these second messengers may funnel their actions through the calcium/cAMP-dependent transcription factor, CREB, as occurs in the mammalian SCN [22–24, 26]. Indeed, the consequences of acute [30] and sustained [47] changes in pacemaker excitability affect TTFLs and are associated with changes in CREB expression. Here, we found that targeted knockdown of CREB in the PG caused a loss of circadian rhythmicity of eclosion, revealing that CREB plays a key role in the PG clock TTFL function; a similar result was obtained following CREB knockdown in the central clock, indicating that CREB may also be critical for neuronal clock TTFLs. Interestingly, overexpression of CREB-binding protein (CBP) in timeless-expressing cells causes an arrhythmic pattern of eclosion and locomotor activity as well as abnormal expression of CLOCK/CYCLE-induced clock genes [51], which suggest that CREB and CBP proteins may play a critical role in the control of the period transcriptional activity, consistent with the presence of CREB binding sites [29] in this gene. Overall, the picture that is emerging from a variety of studies is that calcium and cAMP are actors in the output of the clock, and also play an important role in translating, via CREB, the actions of stimuli external to the clock into changes in TTFL speed and/or phase. Nevertheless, we also found that knockdown of Epac in the PG also caused a significant lengthening of periodicity, which suggests a role for the Epac pathway in regulating circadian clock periodicity, as occurs in the SCN [23]. Although Epac can lead to the activation of CREB, it can also regulate MAP kinase signaling [52], and MAP kinase p38 knockdown in Drosophila clock neurons induces a lengthening of the periodicity of locomotor activity by altering the degree of phosphorylation of PER protein, which delays its translocation to the nucleus [53]. This suggests an additional mechanism through which cAMP could alter TTFL cycling.

The circadian rhythm of Drosophila adult emergence depends on functional clocks in both the brain and the PG [19,35] and recent evidence shows that the brain transmits time information to the PG via the PTTH neuropeptide acting on its receptor, TORSO [19]. Yet, it is unclear how circadian rhythmicity is produced by this system, since PTTH expression does not show circadian rhythmicity in Drosophila [19,54,55] nor is there any evidence for daily oscillations in molting hormone titers during fly metamorphosis [56]. One possibility is that circadian rhythmicity of emergence is accomplished by fine-tuning the levels of ecdysone (the precursor of the active steroid, 20-hydroxyecdysone) production and/or secretion via calcium and cAMP. Although PTTH acts via the tyrosine kinase coupled receptor TORSO, calcium and cAMP have long been associated with PTTH transduction [57], and more recently, calcium has also been shown to play a critical role in ecdysone secretion from the PG [58]. Thus, changes in calcium and cAMP levels due to clock activity could impinge upon the PTTH transduction pathway, ultimately causing changes in the rate of fall of ecdysone titers, thereby delaying or accelerating the rate of metamorphosis and causing emergence to be confined to a specific window of time. Such a mechanism of gating would not require circadian oscillations in 20-hydroxyecdysone titers, but only subtle changes to the rate of fall of 20-hydroxyecdysone titers, and result in the relatively wide eclosion gate observed in this insect species. This mechanism implies that pathways external to the clock and PTTH transduction could also affect the pace of metamorphosis through cross-talk with the PTTH transduction pathway. The detection of complex patterns of calcium oscillations in the PG [58] suggests that much remains to be understood about the functioning of this critical endocrine organ.

This research was approved by the Bioethics and Biosecurity committees of the Universidad de Valparaíso, Chile.

Drosophila strains were raised on standard cornmeal media and, unless noted, were maintained at room temperature (20–22°C) on a 12/12 hours light/dark cycle. The following GAL4 drivers were used: phm-GAL4 [40], tim-GAL4 and pdf-GAL4 (generously provided by Paul Taghert, Washington University, USA). The phm-GAL80 strain has previously been described [19]. The UAS-GCaMP 3.2 calcium sensor was kindly provided by Julie Simpson (HHMI, Janelia Park, USA) and the UAS-Epac1 sensor was obtained from Paul Taghert [33]. UAS-RNAi lines were obtained from the National Institute of Genetics, Japan (NIG), the Vienna Drosophila RNAi Center, Vienna, Austria (VDRC) and the Bloomington Drosophila Stock Center at Indiana University, Bloomington, USA (BL). Typically, a number of RNAi lines were tested and the most effective one was then used for the majority of the experiments. The lines used and their source is indicated below; the line chosen for the results reported here are indicated in bold: for IP₃R (Inositol 1,4,5, -tris-phosphate receptor; CG1063): BL#25937, VDRC#6486; Dmca1A (cacophony or Calcium-channel protein α₁subunit A; CG1522): BL#27244, VDRC#104168; Dmca1D (Calcium-channel protein α₁subunit D; CG4894): BL#25830; SERCA (sarco-endoplasmic reticulum calcium ATPase or Calcium ATPase 60A; CG3725): VDRC#4474, VDRC#107446, BL#25928; dCREB-A (cyclic-AMP response element binding protein A; CG7450): BL#31900, VDRC#110650, BL#27648; CASK (calmodulin-dependent protein kinase activity; CG6703): BL#27556, BL#35309, BL#32857; dCaMKII (Calcium/calmodulin-dependent protein kinase II; CG18069): BL#29401; dunce (cAMP phosphodiesterase or PDE4; CG32498): BL#27250, rutabaga (adenylyl cyclase or AC: CG9533): BL#27035, NIG#9533R-1, VDRC#5569, VDRC#101759; Rap1 (Ras related protein1 GTPase; CG1956): BL#29434, VDRC#110757; D-EPAC (Exchange protein directly activated by cAMP; CG34392): BL#29317, VDRC#50372, VDRC#50373.

Crosses (30–60 females + 15 males) were reared at 20°C under LD 12:12 (lights-on at noon). Resulting pupae were placed in Trikinetics eclosion monitors (Trikinetics, Inc., Waltham, MA, USA). The following day the lights were permanently turned off before lights-on and eclosion monitored for 7–8 days. For experiments using tub-GAL80^ts, flies were raised at 20°C and entrained under LD 12:12 cycles. The resulting pupae were placed in the eclosion monitors and the temperature raised to 28.5°C and kept at this temperature for the remainder of the experiment.

Crosses (10 females + 5 males) were reared at 20°C under LD 12:12 light-dark cycle (lights-on at noon). Male progeny were collected under CO₂ anesthesia on the day of eclosion, aged and entrained for 5–6 days, then placed in Trikinetics activity monitors. Their activity was then monitored under LD 12:12 conditions for 7–10 days followed by 7–10 days under DD conditions.

Periodicity of eclosion and locomotor activity records was analyzed using the Maximum Entropy Spectral Analysis (MESA) and Autocorrelation functions using the Matlab-based software from the “Fly Toolbox” software package [59]. The power of the eclosion and locomotor activity rhythms was determined using MESA analysis and the strength of rhythmicity was quantitated using the rhythmicity index (RI) derived from the Autocorrelation analysis, and categorized as rhythmic (RI≥0.3), weakly rhythmic (RI between 0.1–0.3) or arrhythmic (RI≤0.1 and obvious aperiodic records) [60].

Imaging was carried out as described in Selcho et al. (2017) [19]. Briefly, brains were dissected in ice-cold Schneider’s medium (Sigma-Aldrich) containing 1% antibiotic solution (10,000Uml^-1 penicillin and 10 mg ml^-1 streptomycin; Sigma-Aldrich), placed on poly-lysine coated FluoroDish plates (WPI, FL, USA). They were then covered with Schneider’s medium containing 1% antibiotic solution, supplemented with 10% fetal bovine serum and 10 mgml^-1 insulin (Sigma-Aldrich), and containing 1mM luciferin (Potassium Salt, Gold Biotechnology Inc. MO, USA). Preparations were viewed using an LV200 microscope (Olympus, Japan) under 20X magnification and imaged for 72–96 h with an Evolve 512 camera (Photometrics, Tucson, AZ, USA) using 20min exposures and 200 X gain. Records were detrended using FIJI [63].

Statistical analyses were carried out using Prism 6.0 (Graphpad Software Inc, CA). t-tests and one-way ANOVA, followed by Tukey’s post hoc multiple comparison analyses were used for normally distributed data; Wilcoxon rank sum tests were used for non-normally distributed data.

All relevant data are within the paper and its Supporting Information files.