Real time, in vivo measurement of neuronal and peripheral clocks in Drosophila melanogaster

To monitor distinct clock oscillations in real time, in vivo, we designed a genetically encoded reporter that we call LABL. LABL was constructed into an attB cloning vector for Drosophila embryo injection and PhiC31-mediated genome integration (Bischof et al., 2007; Figure 1—figure supplement 1). LABL is comprised of a per promoter fused to mCherry flanked by FRT sequences, which was subsequently fused to Luc2 (pGL4.10) (Figure 1A). Three stop codons placed 3’ of mCherry are designed to block Luciferase expression. The employed per promoter (~6.7 kb) responds to clock regulation through the CLK/CYC transcription activator complex (Bargiello et al., 1984; Darlington et al., 1998). Tissue-specific expression of Flipase (FLP) triggers recombination at the FRT sites, excising mCherry from the genome and leaving Luciferase under per promoter control. To test the functionality of LABL in vitro, we monitored FLP-driven LABL activity in cultured S2 cells (Figure 1B and C). Since S2 cells do not express CLK, which is required to activate the per promoter, we co-transfected LABL plasmid with Clk-CFP, with and without FLP. These cells were subsequently imaged for CFP and mCherry expression and monitored for luminescence. Cells expressing FLP exhibited increased luminescence and a loss of mCherry fluorescence. Cells lacking FLP exhibited no luminescence but increased mCherry fluorescence. Thus, these data demonstrate that the LABL reporter is functional as designed and can be activated by FLP expression.

Having established its functionality in cultured cells, we proceeded to assess LABL in adult flies. The LABL reporter plasmid was used to generate reporter flies which were then monitored in a luminometer using arenas designed to hold 15 flies on top of fly food supplemented with luciferin (Figure 2A). LABL flies carrying the tim-UAS-Gal4 (TUG) pan-circadian tissue driver were crossed to flies carrying UAS-FLP, and the progeny monitored for luminescence activity in constant darkness (Figure 2B). The raw luminescence data exhibited an oscillating rhythm and a gradual decay, as expected (Brandes et al., 1996; Stanewsky et al., 1997).

We next compared pan-circadian tissue luminescence oscillations with behavioural rhythms. The recorded luminescence activity was normalized to the gradual decay in signal and plotted as an average of four experiments (Figure 2C, top panel). Control flies lacking a driver exhibited no discernable oscillation (white line). To characterize the luminescence oscillations, we first quantified the decay in amplitude (the difference in y-value, between peaks and troughs) of signal. Data points were binned into 30-min time intervals and a 48-hr sinusoidal curve was fitted to the data at 24-hr intervals (Figure 2D). The coordinates of the local minima and maxima were recorded, averaged, and plotted over the decay-normalized luminescence signal to reveal the amplitude of oscillation across time (black circles) (Figure 2C). An S-curve fitted to the changing local minima and maxima (black line) revealed points of inflection coinciding with Day 6 of constant darkness (vertical dashed lines). We used this point of inflection, which is the time at which the amplitude falls to 50% between maximum and minimum (A50), as a measure of clock stability because decay of oscillations into arrhythmic transcription may exceed the timeline of this assay.

Genotypically identical flies were measured for locomotor activity and their behavioral rhythms plotted (Figure 2C, bottom panel). We found that the peaks of morning anticipation (yellow circles) decayed rapidly, allowing the evening anticipation peaks (blue circles) to dominate behavioral oscillations in constant dark conditions. Focusing on the change in evening anticipation peaks, we found that a fitted S-curve revealed a point of inflection (A50) at ~day 6, coincident to the A50 observed in luminescence oscillations. We conclude that the decay of amplitude of oscillation of the ubiquitously expressed molecular clock and behavioural rhythms are consistent with each other.

We next characterized the change of period of luminescence oscillation across time. A Morlet wavelet was fitted onto the measured luminescence oscillations (Figure 2E). Period values with highest confidence intervals revealed a steady ~23.5 hr period of oscillation across time; specifically, oscillations occurred with an average period of 23.57 hr over nine days in constant darkness. This oscillation period also mirrored the behavioural period, with luminescence oscillation and behavioural period statistically the same, at ~23.5 hr. Thus, luminescence could be detected in flies in which LABL was activated using UAS-FLP and the pan-circadian Gal4 driver TUG and revealed parallels between transcription oscillations and behavioral rhythms of the fly.

To determine the effectiveness of LABL as a luminescence-based reporter, we compared TUG-activated LABL oscillations to other, ubiquitously expressed luciferase reporters. To this end, we compared our data to data collected from plo (per promoter fused to luciferase) (Brandes et al., 1996) and PER-BG::Luc (per promoter and ~2/3rds of the per gene fused to luciferase) (Stanewsky et al., 1997) flies, two other luciferase-based reporter systems (Figure 3A). Since TUG drives Gal4 expression in all clock cells, we expected TUG-activated LABL luminescence signal to be comparable to plo luminescence signal. Indeed, the oscillation of luminescence signal was comparable in amplitude, period, and phase (Figure 3A). The PER-BG::Luc construct, on the other hand, retains a significant portion of the PER coding region, and since there is a phase delay between per transcription and per mRNA accumulation (So and Rosbash, 1997; Stanewsky et al., 1997), we expected a phase difference between TUG-activated LABL luminescence signal and PER-BG::Luc signal. As anticipated, PER-BG::Luc luminescence signal was phase-delayed compared to both TUG-activated LABL and plo flies (Stanewsky et al., 1997). Thus, LABL is predictably comparable to both plo and PER-BG::Luc flies.

Eliminating the clock causes behavioural arrhythmicity and loss of transcriptional oscillation. To ensure that the transcription oscillations we observed using LABL were clock-dependent, we monitored TUG-activated LABL luminescence oscillations in a genetic background lacking functional TIM expression (tim¹) (Figure 3B). As expected, both locomotor activity and transcription oscillation of tim¹ flies were arrhythmic (top and middle panel). Importantly, our attempt to fit Morlet wavelets to the luminescence data revealed periods ranging broadly from 48 hr to 72 hr, consistent with circadian arrhythmicity (bottom panel). Elimination of PDF signaling using han⁵³⁰⁴ mutant flies (expressing PDF receptor lacking its transmembrane and cytoplasmic tail) permits rhythmic behavior in the first 1–2 days of constant darkness, but ultimately results in arrhythmic behavior in a majority of flies (Hyun et al., 2005). When we characterized LABL oscillations in a han⁵³⁰⁴ background (Figure 3C), we found that the majority of flies became arrhythmic in their locomotor activity in the second day of constant darkness, as expected. While there was also a rapid decay in transcription oscillation, measured by LABL, closer visual inspection revealed that initial (2–3 days) ~24 hr oscillations decayed into a reproducible 60-hr infradian oscillations with a reliably narrower range of period-fit. Such a decay from an ~24 hr oscillation and coherent (higher confidence wavelet fit) 60-hr infradian oscillation of luminescence was not observable in flies lacking a functional clock (i.e. tim¹ flies). A Kolmogorov-Smirnov test reveals that the period distribution within the 48–72 hr range of the han⁵³⁰⁴ mutant flies was significantly different than that of tim¹ flies (Figure 3B and C, inset).

Because of the unusual observation of infradian rhythms in luminescence, we wanted to determine if we could observe these oscillations using other methods, namely by locomotion assay or immunoblotting. We first measured locomotion rhythms of wild-type (iso31), han⁵³⁰⁴, and tim¹ flies, then conducted a Lomb-Scargle analysis for behavioural period determination (Figure 3—figure supplement 1A). In the 16– 32 hr range, wild-type flies yielded robust ~24 hr rhythms with high power of fit, han⁵³⁰⁴ flies yielded behavioural periods that clustered around 24 hr with a wide distribution and a low power of fit, and tim¹ flies yielded no discernable pattern of behavioural period and a low power of fit. We repeated the analysis for the 48–72 hr range and found two clusters of behavioural periods at approximately 60–62 hr and 50 hr for both wild type and han⁵³⁰⁴ flies. By contrast, tim¹ flies showed no identifiable clustering. All flies had an equivalently low power of fit. Although tempting to suggest the 50 hr and 60 hr clustering patterns we observed in wild-type and han⁵³⁰⁴ flies could be considered infradian rhythms, due to the low power of fit that was comparable to that of tim¹ flies, we do not consider these data reliable representations of infradian oscillations.

We next extracted protein from isolated han⁵³⁰⁴ fly heads, then probed for PER and TIM by immunoblotting to determine if we could observe infradian oscillations (Figure 3—figure supplement 1B). We used protein extracts from heads because protein extracts from whole flies did not yield oscillating protein signal by immunoblot (Figure 3—figure supplement 1C), possibly due to non-oscillating PER protein in ovaries in flies (Hardin, 1994). We observed reliable PER and TIM oscillations from head extracts from both wild-type and the first day of han⁵³⁰⁴ flies in constant darkness. However, the PER and TIM band intensity profiles did not represent reliable oscillations equivalent to a 60 hr period in these immunoblots. We conclude that the subtle amplitude of luminescence oscillation may be too small to observe by either locomotion or immunoblotting assays.

Since the TUG driver can be used to activate LABL in otherwise wild type flies to record oscillating luminescence signal, we wanted to demonstrate that other commonly used circadian drivers can also be used as effectively. We drove FLP expression using Pdf-, DvPdf-, R18H11-, Clk4.1-, Mai179-, and Clk9M-Gal4 (Figure 4A and Figure 4—figure supplement 1). We found that the oscillation with the lowest signal came from flies in which LABL was activated by Pdf-, DvPdf-, and R18H11-Gal4, and this correlates well with the relatively small numbers of neurons that were targeted for LABL activation when compared with TUG-driven LABL (Figure 4A). Clk4.1-, Mai179-, and Clk9M-Gal4 drivers yielded oscillations with higher relative amplitudes than would be expected in proportion to their target neurons (Figure 4—figure supplement 1), suggesting that unintended neurons or tissues may be luminescing (see below). Oscillation periods were maintained at ~24 hr. We conclude that LABL can also be used in conjunction with more tissue-specific drivers.

Since luminescence oscillations in han⁵³⁰⁴ exhibited differences in amplitude, period and a decay into infradian rhythms, we wanted to determine whether these changes could also be observed using LABL in a different biological context. Aged flies have weaker locomotor behaviour rhythms, lower levels of per transcript, and lower amplitude of luminescence signal from flies expressing luciferase by tim or per promoter, compared to their younger (1–3 day old) counterparts (Koh et al., 2006; Luo et al., 2012; Rakshit et al., 2012; Umezaki et al., 2012; Zhao et al., 2019, Figure 4—figure supplement 2). We therefore reasoned that aged LABL flies in which clocks are no longer robust should display luminescence oscillations with weak or unstable amplitude, and an altered A50, using different drivers. We aged LABL-expressing flies to 39–41 days, measured their luminescence, and compared these values to young flies (Figure 4). In TUG-activated aged LABL flies, we fitted appropriate peaks and troughs to the luminescence signal to determine the A50. We note that in young flies the A50 was at 6.5–7 days, almost identical to our earlier observation (Figure 2C), reflecting the reproducibility of this metric. However, we note that since we do not determine A50 for individual flies, this measure does not readily permit statistical analysis for comparison across genotypes. TUG-activated old LABL flies were too irregular to reliably fit peaks and troughs. When LABL was activated using more specific drivers (i.e. Pdf, DvPdf-, and R18H11-Gal4), the peaks and the troughs of old flies remained low throughout the measurement, and either did not reveal an S-curve of decay or revealed too few peaks and troughs through which to fit a curve. Thus, we conclude that an A50 could not be determined for less stable clock oscillations in aged flies because the amplitude in their endogenous clock oscillations had already decayed to their minimal stable levels.

We also note that the luminescence signal decayed into infradian oscillations in some target neurons in aged flies. On day 3 of constant darkness, Pdf- and R18H11-Gal4 activated LABL flies began to exhibit infradian oscillations. On the other hand, Dvpdf-Gal4 activated LABL flies exhibited circadian oscillations until day 7 of constant conditions, at which point the amplitude of oscillation became too low to draw any further conclusions. Thus, infradian oscillations can develop in older flies that exhibit circadian oscillations when younger, suggesting that infradian oscillations can manifest in flies that are genetically identical but distinct in age.

Dissociated neurons exhibit dampened clock oscillations measured by luminescence (Sabado et al., 2017), suggesting clock coupling may also influence our in vivo luminescence measurements. We therefore wanted to determine whether amplitude decay or ~60 hr infradian luminescence oscillations could emerge from oscillators that are uncoupled. We conducted a simple simulation in which half of the simulated oscillators were phase advanced by different magnitudes and the cumulative oscillation quantified (Figure 5A). Our simulations revealed that the magnitude of a phase advance of half the oscillators changes the amplitude of total observed oscillation but does not change the period. A 12 hr phase advance leads to destructive interference, eliminating the oscillation. We next simulated individual oscillators with the same 24 hr period that were allowed to drift in phase over time. These simulations revealed that permitting the oscillators to randomly assume different phase values dampens the overall oscillation amplitude, with a faster rate of phase drift causing a more rapid dampening effect, but still not altering the period of oscillation (Figure 5B). We noted that such phase drift leads to an S-curve decay of amplitude, as we observed in our biological measurements (Figure 2C). We conclude that allowing for phase-shifted variability of oscillations over time cannot explain the infradian rhythms seen in the data.

Thus, while rhythmic and clock-dependent, for the purposes of this report, we interpret these infradian oscillations to represent arrhythmia from the 24-hr circadian perspective (i.e. they oscillate with a period on the order of days). These data together demonstrated that the LABL signal is clock-dependent, PDF receptor (PDFR)-dependent, and that it reflects circadian locomotor behavior. Importantly, our method of analysis can quantify mutation-caused changes in transcription oscillations in target cells and tissues robustly and consistently.

Although the anatomical locations targeted by circadian Gal4 drivers are well characterized in the brain (Figure 6A), the amplitude of luminescence oscillations in some of the Gal4 drivers used to activate LABL (e.g. Clk4.1-, Mai179-, and Clk9M-Gal4) suggests that these driver lines may express Gal4 in non-neuronal tissue or non-circadian neurons as well. To establish the anatomical regions in which different Gal4 lines activate LABL, we visualized luminescence in whole animals, distinguishing between Gal4 lines that visibly activate LABL in the body from those which do not. Flies in which LABL was activated using Clk4.1-, Mai179-, and Clk9M-Gal4 exhibited luminescence signal from the body, similar to that of TUG (Figure 6B). On the other hand, the Pdf-, DvPdf-, and R18H11-Gal4 drivers showed no luminescence in peripheral tissues (Figure 6B). We therefore monitored luminescence from explanted live fly brains from Pdf-, DvPdf-, and R18H11-Gal4 activated LABL flies (Figure 6C). We found that Pdf-Gal4 activated LABL flies emitted luminescence from a single point located in the ventral lateral region of the brain, where the LNvs are expected to be. DvPdf-Gal4 activated LABL flies emitted luminescence from a wider area in the lateral region of the brain, suggesting that the sources of light are the LNvs and LNds. R18H11-Gal4 activated LABL fly brains emitted luminescence from the dorsal area, in the region DN1s are located. These observations correlated well with expected expression pattern of these Gal4 lines (Figure 6A). To ensure that the luminescence signal recorded from the brain specifically reflected the circadian neuronal cluster in which the Gal4 drivers were expected to activate LABL, we tested for FLP activity using G-TRACE reporter flies co-stained for TIM expression (Figure 6D). G-TRACE consists of the Ubip63 promoter upstream of stop codons flanked by FRT sequences, followed by the GFP gene (Evans et al., 2009). RedStinger (RFP)-positive neurons revealed Gal4-active sites, as shown. GFP-positive neurons that co-stain with TIM antibody revealed that the LNvs (Pdf-Gal4), the LNvs, 5th s-LNv and 3 LNd neurons (DvPdf-Gal4), and DN1s (R18H11-Gal4) had the expected FLP activity. Thus, Pdf-, DvPdf-, and R18H11-Gal4 lines activated LABL in the predicted neuronal clusters (collectively, ‘brain drivers’), indicating that they are suitable for use with LABL to monitor neuron-specific clocks. The remaining drivers are collectively referred to as ‘body drivers’, due to their activity in the fly body.

The classical circadian clock mutations, period short (per^S) and period long (per^L) shorten and lengthen rhythmic behaviour, respectively (Konopka and Benzer, 1971). We wanted to further validate LABL using these clock mutants. The per^S mutation produces a residue substitution that blocks phosphorylation of PER by DBT (Chiu et al., 2008; Kivimäe et al., 2008). This mutation in per leads to an early destabilization of the protein before dawn and higher amplitude of luciferase oscillation under the control of the per promoter, causing advanced evening anticipation and short behavioural rhythms (Edery et al., 1994; Li and Rosbash, 2013; Marrus et al., 1996; Top et al., 2018). When assessed using LABL, per^S mutant flies indeed exhibited the expected short-period oscillations of luminescence (Figure 7—figure supplement 1A–1D). Each period of luminescence oscillation correlated well with period of behavioural rhythms. The amplitude of luminescence oscillation measured in per^S flies in which LABL was activated using TUG driver was greater than their wild-type counterparts (Figure 2 and Figure 7—figure supplement 1A), as expected (Top et al., 2018). This effect was not present in pdf-Gal4 activated and R18H11-Gal4 activated LABL flies (Figure 4A and Figure 7—figure supplement 1B and D). Amplitude stability was also differently affected across different neuronal clusters, with the per^S mutation advancing the A50 by a few days in Pdf-Gal4 activated LABL flies and in DvPdf-Gal4 activated LABL flies, while R18H11-Gal4 activated LABL flies maintained a more stable amplitude over time (Figure 4A and Figure 7—figure supplement 1B–1D).

The per^L mutation is a residue substitution in the PER PAS domain that interferes with TIM binding (Gekakis et al., 1995). As suggested by its name, this mutation causes a long behavioural period under free-running conditions (Konopka and Benzer, 1971). This mutation also affects temperature compensation ability in the circadian clock and leads to weaker clock oscillations (Huang et al., 1995; Konopka et al., 1989; Rutila et al., 1996). In LABL flies with a per^L genetic background, we found that all oscillations were noisier, with lower oscillatory amplitude and longer period, consistent with these earlier reports (Figure 7—figure supplement 1E–1H). The apparent noisiness of the oscillation prevented us from determining an A50, which suggests that this parameter is limited to more stable oscillations. We also found that in DvPdf-Gal4 activated LABL flies, signal from luminescence oscillations became too weak on day six of constant darkness to reliably measure, suggesting that the collective signal from these neurons does not detectably oscillate. Our data confirm the features of clock oscillations in per^S and per^L genetic backgrounds, but also uncover differences in distinct neuronal clusters, further validating LABL as a reliable tool to measure distinct circadian clocks.

Elimination of the clock component tim causes arrhythmic locomotor activity and clock oscillations as measured by LABL (Figure 3B). Elimination of PDF signaling leads to arrhythmic locomotor activity in the majority of flies and a transition from ~24 hr to narrow-range infradian oscillations (Figure 3C). This loss of PDF signaling also causes loss of clock synchrony between clock neurons within the fly brain (Lin et al., 2004; Yoshii et al., 2009). We therefore sought to directly measure clock activity in distinct circadian neurons through activation of LABL in mutants lacking PDF signaling (han⁵³⁰⁴). Pdf- and DvPdf-Gal4 retained transcription oscillation in the absence of PDF signaling, with R18H11-Gal4 retaining oscillations until the end of Day 6, while Clk4.1-, Mai179-, Clk9M-Gal4, and TUG began to deteriorate after one cycle of oscillation (Figure 7A), in parallel with observed development of arrhythmic behaviour in locomotor output (Figure 3C). Thus, signal from LABL flies activated by body drivers appears to deteriorate relatively rapidly. In all cases, loss of PDF signaling in the han⁵³⁰⁴ mutants revealed higher amplitude of luminescence oscillation, to varying degrees, relative to wild type, consistent with earlier observations of high amplitude PER-LUC oscillations in subsets of clock neurons in a pdf¹ genetic background (Versteven et al., 2020; Yoshii et al., 2009).

Decay of oscillation amplitude is an indicator of clock robustness. We therefore quantified the A50 value for the different neurons and tissues in which LABL was activated. Stability of luminescence oscillation is represented as a bar (see top of luminescence recordings in Figure 7A), with the dark region representing the early minima/maxima within the S-curves, the light region representing the later minima/maxima within the S-curve, and the gradient representing the difference between points of inflection measured. A comparison of A50 values of different oscillators revealed that it can range from 7 days, in the case of Mai179-Gal4 activated wild-type LABL flies, to 2 days in Clk9M-Gal4 activated wild-type LABL flies (Figure 7B and Figure 4—figure supplement 1). Although we have not determined the extent of influence of the genetic background the GAL4 drivers have on the A50, in all cases elimination of functional PDFR shifted the A50 earlier in constant dark conditions, or eliminated it altogether. han⁵³⁰⁴ flies with Pdf- and DvPdf-Gal4 activated LABL maintained their oscillations despite an advanced A50 (Figure 7A and B), in agreement with earlier immunofluorescence data that demonstrate LNvs and LNds continuing to oscillate in the absence of PDF signaling (Lin et al., 2004). han⁵³⁰⁴ flies with Pdf-Gal4 activated LABL exhibited a high amplitude relative to their wild-type counterparts for ~5 days. This is similar to an observed increase in tim promoter activity in s-LNvs in flies lacking PDF signaling (Mezan et al., 2016). On the other hand, DvPdf- and R18H11-Gal4 activated han⁵³⁰⁴ LABL flies exhibited amplitude peaks similar to their wild-type counterparts after the first cycle of oscillation. Interestingly, with loss of functional PDFR, R18H11-Gal4 activated LABL flies had a relatively unchanged (~half day difference) A50 but relied on PDF signaling to maintain rhythmicity (becoming arrhythmic after Day 6). The large A50 advance observed in DvPDF-Gal4 activated han⁵³⁰⁴ LABL flies is possibly due to phase dispersal in the s-LNvs and phase advance in the LNds caused by a loss of PDF signaling (Lin et al., 2004). Overall, our data suggest that body-driver activated LABL flies are dependent on functional PDFR to maintain rhythmic oscillations. Of the brain drivers, R18H11-Gal4 activated LABL flies were dependent on functional PDFR to maintain rhythmic oscillations in the long term (beyond 6 days), but Pdf- and DvPdf-Gal4 activated LABL flies continued to oscillate in flies lacking PDF signaling, as also observed in pdf¹ flies (Yoshii et al., 2009). From these data, we confirm that PDF signaling is critical for rhythmic locomotor activity and some neuronal clocks, but is not necessary to sustain the ‘master clock’ neurons (LNvs) (Lin et al., 2004; Yoshii et al., 2009).

Flies lacking PDF (pdf¹) exhibit shorter behavioural period as well as advanced PER nuclear translocation in LNds (Lin et al., 2004; Renn et al., 1999). We therefore quantified changes in period of transcription oscillations in the different clocks to determine if shortened behavioural rhythms were caused by shortened oscillatory periods in distinct clocks. Examination of the averaged peaks and troughs of transcription oscillation revealed a consistent phase advance in PDF- and R18H11-Gal4 activated LABL flies lacking functional PDFR (Figure 7A). Surprisingly, elimination of functional PDFR caused longer observed clock oscillation periods in different regions of the brain, except those where R18H11-Gal4 was used to activate LABL flies, which maintained wild-type-like oscillatory periods (Figure 7C). This is inconsistent with earlier reports using 8.0-luc which shows a shorter oscillatory period (Versteven et al., 2020) and the short behavioural period of pdf¹ and pdfr mutants (Hyun et al., 2005; Renn et al., 1999). However, this reporter lacks the period promoter (but encodes the entire PER protein) and is expressed only in some circadian neurons (Veleri et al., 2003), offering a possible explanation for this discrepancy. Our data are also inconsistent with previously published locomotor data that show a shortened period of pdf¹ flies, when arrhythmic animals are excluded (Renn et al., 1999). Therefore, the observed shorter behavioural rhythms may be caused by (1) other clocks that oscillate with a shorter period, (2) a phase change of some of the clocks that normally converge to regulate behavioural rhythms, uncoupled by the loss of PDF (e.g., Figure 5), or (3) elimination of PDFR not entirely reflecting the effect of a pdf¹ fly. Indeed, flies lacking a functional PDFR (han⁵³⁰⁴) or PDF (pdf¹) exhibit a phase delay or advance, respectively, of tim or per transcription oscillation in oenocytes, and PDFR is responsive to other neuropeptides (e.g. DH31), which may explain the differences that we have observed (Krupp et al., 2013; Mertens et al., 2005).

Periods of luminescence oscillation of different tissues in han⁵³⁰⁴ flies were lengthened to different degrees. Luminescence from Mai179-Gal4 activated LABL in han⁵³⁰⁴ flies exhibited a~30 hr period before becoming arrhythmic on day 3, as compared to a~24 hr period in wild-type flies (Figure 7C). A time course revealed that all oscillators that lose rhythmicity due to loss of PDF signaling appeared to form narrow range infradian rhythms (Figure 7D). Of these, the body drivers (Clk4.1-, Mai179-, Clk9M-Gal4) appeared to show a more disorganized period decay of period in constant darkness (Figure 7D). Importantly, we conclude from these data that LABL can demonstrate that PDF signaling has distinct effects on different neurons and tissues in clock stability, period, and amplitude (Figure 7E).

In mice with suprachiasmatic nucleus (SCN; mammalian master clock neurons) lesions, organs become asynchronous (Sinturel et al., 2021). In flies, the PDF-expressing ventral lateral neurons are considered to be the master clock neurons (Grima et al., 2004; Stoleru et al., 2004). We reasoned that eliminating PDF-signaling would serve to reduce the master clock influence on peripheral organs in flies. This may explain, in part, the difference we observe in LABL activated by brain clocks compared to body clocks (Figure 7). Although PDFR protein has not been shown to be expressed in the peripheral organs, pdfr mRNA has (Li et al., 2022), suggesting that clocks in these tissues may partially depend on PDF-signaling. To characterize tissue-specific peripheral clock drivers, we monitored the stability of tissue-specific clock function in constant dark conditions in both wild-type controls and han⁵³⁰⁴ mutants. Specifically, we examined LABL oscillations using the neuronal driver elav-Gal4 (Saitoe et al., 2001; Schuster et al., 1996; Sink et al., 2001), intestinal drivers esg- and NP3084-Gal4 (Bonnay et al., 2013; Croker et al., 2007; Lee et al., 2016; Strand and Micchelli, 2011), fat body drivers C564- and LSP2-Gal4 (C564 also expresses in hemocytes and some male reproductive tissues) (Hrdlicka et al., 2002; Paredes et al., 2011; Takeuchi et al., 2015; Zaidman-Rémy et al., 2006), and muscle driver mef2-Gal4 (also expresses in some neurons [LNvs]) (Blanchard et al., 2010; Cripps et al., 1998; Gajewski et al., 1997, Figure 8A). These drivers activated LABL to reveal wild-type clock oscillations with ~24 hr periods (Figure 8—figure supplement 1).

If sufficient amounts of PDFR protein for signaling efficacy are not expressed in these organs despite measured pdfr transcript (Li et al., 2022), oscillations may be expected to remain undisrupted in the han⁵³⁰⁴ genetic background. However, our observation of high-amplitude luminescence oscillations in flies lacking PDF signaling is consistent with earlier high-amplitude observations (Versteven et al., 2020; Yoshii et al., 2009) and our own observations (Figure 7A) that PDF signaling is relevant in these tissues (Figure 8B). Furthermore, the noisiness of these oscillations in flies lacking PDF-signaling, similar to hepatocyte oscillations in SCN-lesioned mice (Sinturel et al., 2021), suggests that PDF signaling from the master clocks plays some type of stabilizing role for these peripheral clocks.

The measured peripheral clocks appear to respond differently to a loss of PDF signaling. We found that the fat body drivers, C564-, and LSP2-Gal4, revealed phase-advanced luminescence oscillations in han⁵³⁰⁴ flies, but these oscillations were otherwise similar to wild-type. NP3084-Gal4 driven LABL lost rhythmic oscillations within 5 days, whether or not PDF-signaling was present. Interestingly, mef2- and esg-Gal4 activated han⁵³⁰⁴ LABL flies exhibited a loss of measurable oscillation between days 5–6 and 4–6, respectively, followed by a re-established rhythm and also exhibited earlier A50 values compared to their wild-type counterparts (Figure 8B and C). We also note that oscillations in the mef2- and esg-Gal4 driven LABL flies, when rhythms are reestablished, reappear in reverse phase. This echoes an observation made in the cry^– LNds, which normally oscillate in reverse phase with respect to the cry⁺ LNds, but oscillate in synchrony with them in a pdf¹ background (Yoshii et al., 2009). Together, our data suggest that PDF signaling has varying influence on different peripheral clocks.

Quantification of clock periods in peripheral tissues revealed distinct periods in wild-type flies (Figure 8D). Importantly, LABL activated by fat body-specific drivers (C564- and LSP2-Gal4), but also for intestine driver NP3084, exhibited oscillation periods greater than 48 hr, suggesting that these clocks may not be as robust (e.g. compared to elav⁺ clocks) in constant darkness. Elimination of PDF signaling appeared to destabilize the period of oscillation of these and all other measured peripheral clocks to varying degrees (Figure 8D). To uncover the extent of period stability in the absence of PDF signaling, we monitored changes in period of peripheral clocks over time (Figure 8E). Intestine-specific Gal4 drivers (esg- and NP3084-Gal4) revealed transcriptional oscillations that stabilized in infradian oscillations in flies lacking PDF signaling, similar to that observed using body clock drivers (Figure 7D). Of note, NP3084-Gal4 activated LABL flies also exhibited infradian oscillations in a wild type PDFR background. On the other hand, muscle- and fat body-specific drivers (mef2-, C564-, and LSP2-Gal4) appeared to maintain a rhythmic ~24 hr transcription oscillation, though these were unstable, as demonstrated by intermittent deviations (Figure 8E) and observed visually (Figure 8B). Interestingly, periods of clock oscillation in the fat body were shorter in the absence of PDF signaling as compared to the wild type (Figure 8B and D), contrasting the longer periods observed in the same mutant background in Pdf- and DvPdf-Gal4 activated LABL flies (Figure 7C), but similar to shorter behavioural periods measured by locomotion (Hyun et al., 2005). Thus, the LABL reporter system is able to distinguish between loss of rhythmic oscillation and destabilized oscillation, as well as changes in the period of transcription oscillation, of circadian clocks in peripheral tissue.

The following lines were used in the study: tim-UAS-Gal4 (A3) (Blau and Young, 1999), Pdf-Gal4 (Park et al., 2000), DvPdf-Gal4 (Guo et al., 2014), R18H11-Gal4 (Guo et al., 2016), Clk4.1-Gal4 (Zhang et al., 2010a; Zhang et al., 2010b), per^S and per^L (Konopka and Benzer, 1971), elav-Gal4 (gift from Michael W. Young); Pdfr⁵³⁰⁴ (han⁵³⁰⁴) (Hyun et al., 2005) (gift from Paul Taghert); Mai179-Gal4 (Picot et al., 2007) (gift from C. Helfrich-Foerster); Clk9M-Gal4 (Kaneko et al., 2012) (gift from O. T. Shafer); 3xUAS-FLP2::pest (Nern et al., 2011) (Janelia Research Campus); PDF receptor mutant han⁵³⁰⁴ (Hyun et al., 2005; Mertens et al., 2005) (gift from P. H. Taghert); esg-Gal4 (gift from Ben Ohlstein); NP3084-Gal4 (gift from David Walker). G-TRACE flies (32251) Evans et al., 2009; mef2-Gal4 (27390); C564-Gal4 (6982); LSP2-Gal4 (6357) were acquired from BDSC. Genotypes of all flies used in experiments are summarized in Supplementary file 1. Flies were reared on standard cornmeal/agar/yeast/molasses medium at room temperature (22 °C) or 18 °C in ambient laboratory light. Flies were entrained in a light-dark cycle (12 hr:12 h) at 25 °C.

LABL was constructed into an attB cloning vector for Drosophila embryo injection and PhiC31-mediated genome integration (Bischof et al., 2007; Figure 1—figure supplement 1). The minimal per promoter (Bargiello et al., 1984) (ranging from restriction sites SphI to XbaI) was first cloned into the pattB vector. The 5’ untranslated region, ranging from XbaI site to the per ATG start codon was rebuilt using standard PCR: the ATG was conserved and an FRT sequence followed by a NotI restriction site was added. mCherry was amplified using standard PCR, its start codon eliminated, and a 5’ NotI restriction site, three 3’ stop codons and an EcoRI restriction site added. Luc2 (Promega) was amplified using standard PCR its start codon eliminated, a 5’ EcoRI restriction site followed by FRT sequence added and then 3’ XhoI and KpnI restriction sites added. The plasmid was then assembled and injected into embryos (BestGene Inc).

Luminescence of flies were measured using the LumiCycle 32 Color (Actimetrics). Custom 35 mm plates (designed by Actimetrics) were used to adapt the LumiCycle 32 for Drosophila use. D-luciferin potassium salt (Cayman Chemicals or Gold Biotechnology) was mixed with standard fly food to a final concentration of 15 mM. Volume of food in the plate was sufficient to limit fly movement along the z-axis of the plates. 15 flies were placed in each plate and covered with a cover slip. Each of the four replicates were recorded in a different position in the luminometer to ensure equal representation of data from each of the four photomultiplier tubes. Luminescence from each plate was recorded in 4-min intervals on a standard Windows-operated PC using software by Actimetrics.

Luminescence of transfected S2 cells were measured using a liquid scintillation counter LS6000IC (Beckman) in single-photon collection mode. Cells were transfected in six-well plates at 80% confluency using Effectene transfection reagent (Qiagen) following the manufacturer’s protocol. Cells were transfected with 200 ng total DNA with the indicated plasmids distributed in equivalent amounts. Plasmids used were pAc-Clk-CFP (actin promoter driven clock fused to CFP) (Top et al., 2018), pAc-FLP (actin driven Flipase; gift from Nicholas Stavropoulos), LABL plasmid (above). Cells were fed 24 hr after transfection and lysed another 24 hr later in Cell Culture Lysis Reagent (Promega). Extracts were mixed with Luciferase Assay Reagent (Promega) in a 1:5 ratio and monitored for luminescence.

Actimetrics analysis software was used to normalize the exponential decay of luminescence signal over days using a polynomial curve fit, with no smoothing. Data were exported into.csv files for analysis. Custom python code was used to organize luminescence data into 30 min bins and quantify peaks, troughs and decay of oscillating luminescence signal (LABLv9.py; https://www.top-lab.org/downloads↗ or https://github.com/deniztop/LABL↗) (Top, 2022). A sinusoidal curve was fitted to each day,+/-half a day, with the x- and y- coordinates for the local maxima and minima identified and quantified. Overlapping values for each peak and trough were averaged for x- and y- coordinates before being plotted,+/-standard deviation. Graphpad Prism 9 software was used to fit an S-curve to these points, and the calculated value for IC50 was used to identify the point of inflection to identify the reported A50. Custom python code was also used to quantify periods of oscillations using a Morlet wavelet fit (waveletsv4.py; https://www.top-lab.org/downloads↗ or https://github.com/deniztop/LABL↗) (Top, 2022), as previously described by others (Leise and Harrington, 2011). The bottom 25% of best fits were omitted from analysis. Computer-generated oscillation data were used to confirm the parameters for analysis (data not shown): bandwidth = 3; central frequency = 1; data points in an hour = 15 (i.e. 15x4 min in an hour); zero-hour mark = 10 (10:00 am, subjective dawn); period of shortest wave match = 16; period of longest wave match = 72; increment of wavelet fit = 1; percent threshold for discarding data as a fraction = 0.25. Highest confidence interval values were used to plot period across time. The time (x-value) when an oscillation was considered arrhythmic was defined as the last 25 data points before a wavelet fit exhibited a jump from ~24 hr to ~60 hr, and averaged.

Individual flies were analyzed for locomotor activity for three days in a light-dark cycle, followed by seven days in constant darkness using the Drosophila Activity Monitor System 5 (Trikinetics). Circadian behaviour was determined in the analysis of time in constant darkness. Periodicity of rhythmic locomotor activity was determined using ImageJ plug-in ActogramJ (Lomb-Scargle) (Schmid et al., 2011).

To measure loss of mCherry signal from LABL plasmid in S2 cells, cells were grown in Lab-Tek II chamber slides (Nunc, Rochester, NY). Cells were imaged using a DeltaVision system (Applied Precision, Issaquah, WA) equipped with an inverted Olympus IX70 microscope (60 X oil objective, 1.42 N.A.), a CFP/YFP/mCherry filter set and dichroic mirror (Chroma, Foothill Ranch, CA), a CCD camera (Photometrics, Tucson, AZ), and an XYZ piezoelectric stage for locating and revisiting multiple cells.

Fly brains were collected at ZT23, fixed, mounted, and imaged using Leica confocal microscopy as previously described (Top et al., 2018; Top et al., 2016). Briefly, fly heads were fixed in PBS with 4% paraformaldehyde and 0.5% Triton X-100. Brains were dissected and washed in PBS with 0.5% Triton X-100. Brains were then probed with a 1:1000 dilution of rat anti-TIM (Myers et al., 1996), a 1:1000 dilution of chicken anti-GFP (Aves Labs), and a 1:1000 dilution of rabbit anti-mCherry (Rockland). Washed brains were re-probed using a 1:500 dilution of goat anti-chicken conjugated to Alexa-488, a 1:200 dilution of donkey anti-rat conjugated to Alexa-594, and a 1:200 dilution of donkey anti-rabbit conjugated to Alexa-647 secondary antibodies (Jackson Immunological). Brains were mounted using Fluoromount G (Beckman Coulter, Brea, CA) and imaged using a ZEISS LSM 710 confocal microscopy at ×40 magnification.

Fly heads were collected from iso31 and han⁵³⁰⁴ flies at 6-hr intervals, specifically CT03, 09, 15, and 21. Protein was extracted from fly heads as described earlier (Top et al., 2016). Protein samples were quantified using a Lowry Assay (Bio-Rad Cat. 5000112) and transferred to PVDF membrane using standard methods. Blots were probed with antibodies against PER (1:500) (gift from Michael W. Young), TIM (1:3000) (Myers et al., 1996), and tubulin (1:20,000) (Sigma, Cat. T6199). HRP-conjugated secondary antibodies against rat (Jackson, Cat. 712-035-153) and mouse (Jackson, Cat. 715-036-151) were used at 1:10,000 dilution.

Flies were kept at 25 °C under light-dark cycles in vials containing 1% agar and 5% sucrose food supplemented with 15 mM luciferin for at least 3 days prior to the experiment. Images were taken using a Luminoview LV200 (Olympus) bioluminescence imaging system, equipped with an EM-CCD camera (0.688 MHz EM-CCD CAM-ImageEM X2, Hamamatsu, Japan) at a time point corresponding to the peak phase of expression (between ZT 18 and ZT 20).

Adult flies were anesthetized with ether and dry mounted on a CellView glass bottom imaging dish (35x10 mm; Greiner Bio One). Once the position of the fly was determined by bright field microscopy, bioluminescence was measured for 5 min with an EM gain of 400, using a UPLSAPO 20 x Apochromat, followed by 600ms exposure to fluorescence light. After this, another bright field image was captured and compared to the initial image, in order to ensure that the fly had not moved.

Preparation of the brains was similar to that described previously (Schubert et al., 2020; Versteven et al., 2020). Briefly, brains were dissected under red light in ice cold Ca²⁺ free Ringer’s solution and placed on a CellView glass bottom imaging dish (35x10 mm; Greiner Bio One) treated with Heptane glue, where the culture medium was added. The medium contains 20% heat-inactivated fetal bovine serum, 1% Penicillin-Streptomycin and 0.75 µM Luciferin in Schneider’s medium (Sigma-Aldrich). Bioluminescence images were acquired with a UPLSAPO 40 x by exposing the brains for 5 min with an EM gain of 400. For each brain, a Z-stack of 13 slices was taken, each slice containing a bioluminescence, a fluorescence (200ms), and a bright field image. All images were processed in FIJI (Schindelin et al., 2012) and combined with bright field images in GIMP (https://www.gimp.org/↗).