Drosophila were reared at 25°C on a corn-meal medium under 12 h:12 h light-dark (LD) cycles. UAS-CLKΔ (Tanoue et al., 2004) was a gift from Jadwiga Giebultowicz. UAS-TNT (UAS-TNT-G) (Kaneko et al., 2000) was a gift from Jeff Hall. UAS-per RNAi (perCt-IR) (Martinek and Young, 2000), Pdf-GAL4 (Park et al., 2000), gal118 (Blanchardon et al., 2001), DvPdf-Gal4 (Bahn et al., 2009), and tubulin-GAL80ts (McGuire et al., 2004) were described previously.

The locomotor behavior assay was performed as described in Beuchle et al. (2012) using the Drosophila Activity Monitoring (DAM) System (Trikinetics, Waltham, MA), except that assays were performed at 29∘C in the experiments with adult-restricted conditional GAL4 induction and at 18∘C for developmental GAL4 induction. For adult-restricted GAL4 induction, flies were crossed and raised at 18∘C until 2 days after eclosion. Male flies of appropriate genotypes were then collected and placed in the DAM monitors and assayed for locomotor activity at 29∘C. Flies were first entrained in 12 h:12 h-LD cycles for 4 days and then released in DD for 10–12 days. The light intensity of the incubator was approximately 1000 lux. For experiments with developmental GAL4 expression, flies were raised at 29∘C until 2 days after eclosion, and then the collected flies were assayed at 18∘C. In both sets of experiments, behavioral data were analyzed from the second day in LD. Two to four independent experiments were performed for each genotype. The numbers of flies used in the behavioral assays are indicated in Table 1. The behavioral data were analyzed using FaasX software (Blanchardon et al., 2001). The flies with power over 20 and width over 2.5 h according to the χ² periodogram analysis were defined as rhythmic. The significance threshold was set to 5%. Morning anticipation index (M-index) was calculated as described in Im and Taghert (2010) with minor modifications. The M-index was calculated for individual flies as (sum of activity over 3 h before lights on)/(sum activity over 6 h before lights on) at each day from LD2 to LD4, and the 3 values were averaged to obtain the mean M-index of an individual fly. The mean M-indices were pooled per genotype and presented as boxplots.

Anti-PER and PDF immunostaining of fly brains was performed as described previously (Shafer et al., 2002). The brains were imaged using a Leica SP5 confocal microscope and images were analyzed using Fiji/Image J software (National Institutes of Health). To quantify PER staining intensity, sum slices projections were generated from 2 μm z-section confocal images, and the mean pixel value of each cell and background pixel value was measured. The mean pixel value of each cell in a given subgroup was calculated by subtracting the mean pixel value of the background and plotted as relative intensity normalized to the value of the control group at CT0. PDF levels in the s-LNv dorsal terminals were measured as described in Kozlov et al. (2017). Briefly, the region of interest (ROI) for the axonal termini (from the tip of the s-LNv dorsal termini until where the terminal arbors first branch) was specified manually with the polygon selection tool in Fiji and the intensity sum within each ROI was measured. The representative confocal images were maximum projections generated from the same confocal z-series used for the quantification.

Statistical analysis and data visualization were performed using GraphPad Prism (9.0). A p-value < 0.05 is considered a statistically significant test result. Asterisks indicate p-values, where ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001, and ^****p < 0.0001. ns indicates a non-significant test result. Data were first tested for normality with D’Agostino-Pearson K2 test, and normally distributed data sets were analyzed using parametric tests (ANOVA and unpaired t-test with Welch’s correction) and non-normally distributed data were analyzed with non-parametric tests (the Kruskal-Wallis test and Mann-Whitney U-test).

Welch’s t-test and Mann-Whitney U-test were used for pairwise comparisons of the behavioral rhythmicity between control and test genotype groups, depending on the distribution of each data set. Morning anticipation indices were compared using the Kruskal-Wallis test with Dunn’s multiple comparisons test. Signal intensities of immunofluorescence images were compared using the multiple unpaired t-test with Welch’s correction or ANOVA with Sidak’s multiple comparison’s test.

To conditionally eliminate molecular clockwork only during adulthood in the LNvs, we drove the expression of CLKΔ, a dominant-negative mutant of CLK (Tanoue et al., 2004), using the combination of Pdf-GAL4 and temperature sensitive GAL80 expressed ubiquitously under the tubulin promoter (tub-GAL80^ts). The flies were raised at 18∘C until 2 days after eclosion and adult male flies were maintained at 29∘C during the subsequent experiments (McGuire et al., 2004). To verify if this manipulation effectively blocks molecular clockwork, the brains of flies were immunolabeled with anti-PER and anti-PDF antibodies every 4 h on the third day in constant darkness (DD3) following an entrainment to 12 h:12 h light-dark (LD) cycles for 4 days (Supplementary Figure 1). As expected, PER levels in the s-LNvs were significantly reduced and arrhythmic on DD3. In DN1s, PER levels peaked at CT12 but did not show 24 h rhythmicity. This observation is congruent with the notion that molecular rhythms in the s-LNvs affect rhythmicity of the DN1s (Nitabach et al., 2006; Zhang L.Y. et al., 2010; Beuchle et al., 2012; Jaumouille et al., 2015). PER levels and oscillations in the LNds were not different between control and CLKΔ-expressing flies (Figure 2A). Moreover, PDF signal at the s-LNv dorsal termini was reduced and arrhythmic in CLKΔ-expressing flies (Figures 2B,C). This observation confirms the previous finding that Clk controls PDF levels and accumulation rhythms via regulating vri expression in adult s-LNvs (Gunawardhana and Hardin, 2017).

Having validated that adult-restricted expression of CLKΔ in the LNvs effectively abolishes molecular clockwork, we next assayed locomotor behavior of these flies. Unexpectedly, CLKΔ driven with Pdf-GAL4 and tub-GAL80ts had a modest effect on free-running locomotor rhythms at least up to DD10. Most (98.3%) of the Pdf-GAL4/CLKΔ; tub-GAL80ts/+ flies were rhythmic until DD5, despite with a reduced rhythm power. Their rhythms damped significantly after DD6 compared to control flies (Figures 2D,E and Table 1). To verify the effect of adult-restricted CLKΔ expression in the LNvs, we next turned to the gal1118 driver (Malpel et al., 2004). Gal1118 is expressed in the s- and l-LNvs and weakly expressed in 5 LNds, but the expression in the LNds is detectable only in the flies homozygous for gal1118 (Malpel et al., 2004; Figure 1B). Therefore, we used one copy of gal1118 in combination with tub-GAL80ts to express CLKΔ only during adulthood. These flies remained highly rhythmic during the first 5 days in DD. After DD6, a high proportion of these flies became arrhythmic (Table 1 and Supplementary Figures 2A,B). These results indicate that behavioral rhythms in DD only gradually dampen over several days when clocks in the LNvs are disrupted in adulthood.

To corroborate these findings, we also used UAS-RNAi against PER as an alternative method to temporarily block molecular clockwork. UAS-PER RNAi was driven in the LNvs with Pdf-GAL4 or gal1118 and tub-GAL80ts and its expression was induced by the temperature shift from 18 to 29∘C 2 days after eclosion. For behavioral assays we used ga1118, because many flies carrying Pdf-GAL4, tub-GAL80ts, and UAS-PER RNAi did not survive until the end of behavioral recording for unknown reasons. As with conditional CLKΔ expression, this treatment eliminated PER expression and rhythms in the s-LNvs, verified by anti-PER and anti-PDF immunostaining on DD3 (Supplementary Figure 1 and Figure 3A). Levels and rhythms of PDF accumulation at the dorsal termini of the s-LNvs were also reduced (Figures 3B,C). This observation is congruent with a previous report that CLK represses pdf transcription (Mezan et al., 2016) because PER knockdown should increase CLK/CYC-transcriptional activity (Yu and Hardin, 2006). Free-running locomotor rhythmicity of these flies were not disrupted compared to controls at least up to DD10 (Figures 3D,E and Table 1), consistent with the results of adult-restricted, LNv-targeted CLKΔ expression (Figures 2D,E, Table 1, and Supplementary Figures 2A,B).

The presence of functional clocks in developing LNvs is required for driving normal locomotor rhythms in adulthood (Goda et al., 2011; Beuchle et al., 2012). Consistent with this notion, conditional expression of CLKΔ only during development until eclosion in the LNvs resulted in an immediate behavioral arrhythmia in DD (Figures 4A,B). This finding verifies that conditional expression of CLKΔ is an effective tool to disrupt molecular clocks and highlights the differential requirements for LNvs’ clocks during development and adulthood in the functioning of the pacemaker circuit. Previous studies have shown that inactivating per in pacemaker neurons during development does not affect behavioral rhythms of adult flies but depleting CYC during development abolishes locomotor rhythms in adults (Ewer et al., 1990; Goda et al., 2011). Our results are in agreement with these observations and support that CLK/CYC activity has non-clock roles during development.

How could locomotor rhythms persist several days under constant conditions while the M-cells are molecularly arrhythmic? As proposed by a number of other studies (Guo et al., 2014; Yao and Shafer, 2014; Bulthuis et al., 2019; Delventhal et al., 2019; Schlichting et al., 2019), we hypothesized that other pacemaker neurons compensate arrhythmic LNvs to drive free-running rhythms. In particular, recent studies have described that CRISPR-mediated ablation of molecular clocks in the LNvs does not cause strong behavioral phenotypes in DD, whereas clock knockout in the M- and all or part of LN-EO cells using Mai179-GAL4 or DvPdf-GAL4 significantly reduces behavioral rhythmicity (although the data of the DvPdf-GAL4-mediated clock knockout are not displayed) (Delventhal et al., 2019; Schlichting et al., 2019). Therefore, LN-EO cells are the likely candidates of the surrogate main pacemaker. To test this idea, we next expressed CLKΔ with the DvPdf-GAL4 driver during adulthood. The majority of these flies remained rhythmic until DD5, thereafter became arrhythmic (Figures 5A,B and Table 1). However, the average rhythm power of DvPdf-GAL4/CLKΔ; tub-GAL80ts/+ flies were approximately the same as that of two control groups until DD10. Since DvPdf-GAL4 is expressed in the M-cells and two out of four LN-EO cells (Figure 1B), these observations suggest the possibility that clocks in PDF-negative pacemaker neurons, including two LNds that do not express DvPdf-GAL4 (i.e., CRY-positive, sNPF-positive LNds), compensate the loss of clocks in the M-cells and maintain rhythmic locomotor output for several days.

Our results thus far are in line with previous works and suggest that non-M-cells can output behavioral rhythms independently of the M-cells, or they input signals to the M-cells, which then output behavioral rhythms without the need of molecular clocks in the M-cells. The possibility that non-M-cells output behavioral rhythms via M-cells is backed by the evidence that presence of the s-LNvs is essential for locomotor rhythms (Helfrich-Forster, 1998; Stoleru et al., 2004). To test this hypothesis further, we sought to block neuronal output of the M-cells with or without disrupting molecular clocks in the M-cells. We first blocked output of the LNvs during adulthood by conditionally expressing tetanus toxin light chain (TNT) (Sweeney et al., 1995) using the combination of gal1118 and tub-GAL80^ts. Gal1118 was used instead of Pdf-GAL4 due to the ease in establishing the desired genotypes in co-expression experiments. TNT is a protease that cleaves n-synaptobrevin, syntaxin or SNAP-25, thereby inhibits synaptic transmission (Sweeney et al., 1995) and neuropeptide release (Ding et al., 2019). These flies exhibited locomotor rhythms with a significantly reduced power already during the first 5 days in DD. The power of the rhythmicity was further reduced after DD6 (Figures 6A,B). Approximately 25% of these flies were arrhythmic before DD5, and thereafter approximately 80% of them became arrhythmic (Figure 6C and Table 1). This finding is largely congruent with the results presented in Kaneko et al. (2000), where the same UAS-TNT (UAS-TNT-G) line used in the present study was constitutively driven with Pdf-GAL4. In contrast, one study reported that constitutive expression of TNT with Pdf-GAL4 had no effect on both LD and DD behavior, using another UAS-TNT insert, UAS-TNT-E (Umezaki et al., 2011). This discrepancy is likely attributed to the difference in expression levels, as UAS-TNT-G renders a higher level of expression than UAS-TNT-E (Kaneko et al., 1997).

We next expressed both CLKΔ and TNT in the LNvs using gal1118 during adulthood. Strikingly, these flies became immediately arrhythmic in DD (Figures 6A–C and Table 1). In other words, loss of neural transmission and loss of molecular clockwork in adult LNvs cumulatively cause the rapid decline of rhythmic behavioral output. This finding further suggests that molecular clocks and neural output function of the M-cells are independent components that add up to enable robust rhythmic behavioral output in DD. Even when clocks are disrupted, the M-cells can receive inputs from other pacemakers and transmit output signals to the output circuit. When TNT is expressed in the M-cells, two scenarios are possible: a surrogate master pacemaker bypasses the M-cells to control locomotor output; or the M-cells still produce output signals via TNT-insensitive transmitters/peptides or by electrical coupling via gap junctions (Schneider and Stengl, 2006; Ramakrishnan and Sheeba, 2020).

It has been shown that morning anticipatory activity requires the presence of the M-cells and PDF neuropeptide (Renn et al., 1999; Grima et al., 2004; Shafer and Taghert, 2009) but can be observed in the absence of per rhythms within the M-cells (Stoleru et al., 2004). We examined whether morning anticipation is impaired when molecular clocks are disrupted during adulthood in the LNvs. Adult-restricted PER RNAi in the LNvs, which abolishes PER rhythms and reduces PDF levels and rhythms in the s-LNv axonal termini (Figures 3A–C), did not impair morning anticipation (Figures 7A,B, compare perRNAi/+; gal1118/tubG80 and gal1118/tubG80). These results are congruent with the report that per expression in the M-cells is not required for morning anticipation (Stoleru et al., 2004). However, the morning anticipation index (M-index) (Sheeba et al., 2010) was significantly reduced when CLKΔ was conditionally driven in adulthood with Pdf-GAL4 or gal1118 (Figures 7A,B, see PdfG4/ClkΔ; tubG80/+, ClkΔ/+; gal1118/tubG80, PdfG4/+; tubG80/+, gal1118/tubG80, and ClkΔ/+; tubG80/+). Similarly, adult-restricted CLKΔ expression using Dvpdf-GAL4 reduced morning anticipation compared with controls (Figures 7A,B, compare DvPdfG4/ClkΔ; tubG80/+, DvPdfG4/+; tubG80/+, and ClkΔ/+; tubG80/+). CLKΔ expression in adult LNvs reduces the levels and diurnal rhythms of PDF accumulation in the s-LNv dorsal projections (Figures 2B,C) as is the case with adult-restricted PER RNAi. These results validate that adult-restricted expression of CLKΔ using GAL4/GAL80ts is immediately in effect and further suggest that loss of PER rhythms and reduction in axonal PDF levels in the s-LNvs during adulthood do not necessarily impair morning anticipation. These data instead suggest the involvement of other factors that are regulated by CLK in generating morning anticipation.

In this regard, it is noteworthy that morning anticipation was reduced in the flies expressing TNT in adult LNvs compared with controls and even further reduced when CLKΔ and TNT were co-expressed (Figures 7A,B, compare TNT/+; gal1118/tubG80, CLKΔ/TNT; gal1118/tubG80, TNT/+; tubG80/+, and gal1118/tubG80). As TNT is known to block both neurotransmitter and neuropeptide release (Sweeney et al., 1995; Ding et al., 2019), the results suggest the possibility that, in addition to the PDF, neurotransmitter, such as glycine (Frenkel et al., 2017) or the short Neuropeptide F (sNPF) (Johard et al., 2009), may be involved in the normal morning anticipation.

We also noticed that evening activity peak was not apparently advanced in all the genotypes tested (Figure 7A). Free-running rhythms were not shortened either (Table 1). This was surprising because loss of PDF advances evening peak in a 12 h:12 h LD cycle at the temperature near 25°C (Renn et al., 1999). The lack of effect on evening activity even in the flies that have reduced morning anticipation is probably because their PDF levels are only partially reduced (Figures 2B,C, 3B,C). Additionally, the fact that all the behavioral experiments were performed at 29°C, the temperature that suppresses daytime activity (Majercak et al., 1999; Parisky et al., 2016), likely masked the effects on the evening activity peak.