Neuropeptides PDF and DH31 hierarchically regulate free-running rhythmicity in Drosophila circadian locomotor activity

To re-evaluate the function of DH31 in regulating the circadian rhythms of locomotor activity, we focused on a double mutant of Dh31^#51 (a loss-of-function mutant)¹⁶ and Pdf¹ (a null mutant)⁷ and examined the phenotypes for rhythmicity, free-running period, morning anticipation and evening activity peaks.

We determined that both w¹¹¹⁸ (WT) and Dh31^#51 mutant flies maintained a robust free-running rhythmicity (WT: 92% rhythmic, power = 1371.7, Dh31^#51: 93% rhythmic, power = 678.2) (Fig. 1A,F,G and Table 1). These data are consistent with previous reports using Dh31^#51 mutants¹⁵ and another Dh31 mutant¹⁴. In contrast, the Pdf¹ mutants exhibited a weak free-running rhythmicity (40% rhythmic, power = 243.0) (Fig. 1A,H and Table 1), which is also consistent with previous reports^9,10. However, we determined that the free-running rhythmicity of Dh31^#51;Pdf¹ double mutants was strongly disrupted: 92% of the flies exhibited an arrhythmic phenotype, and only 8% showed weak amplitudes (power = 226.7) (Fig. 1A,I and Table 1). These data indicate that the Dh31^#51;Pdf¹ double-mutant phenotype is more severely arrhythmic than either single mutant, suggesting that DH31 is involved in modulating free-running rhythmicity.

We also determined that the free-running period of the Dh31^#51;Pdf¹ double mutants was 23.1 h (Fig. 1I and Table 1), which was slightly longer than that of the Pdf¹ mutants (22.5 h; Fig. 1H and Table 1) and shorter than that of the Dh31^#51 mutants (24.4 h; Fig. 1G and Table 1). These data indicate that the Dh31 mutation did not enhance the shorter period caused by the Pdf mutation.

In terms of morning anticipation, compared to WT flies, both the Dh31^#51;Pdf¹ double mutants and Dh31^#51 mutants exhibited abnormal morning anticipation (Fig. S1 and Table S1). However, the anticipation indexes of the Dh31^#51;Pdf¹ double mutants and Dh31^#51 mutants were not significantly different (Table S2). Furthermore, both the Dh31^#51;Pdf¹ double mutants and Pdf¹ mutants similarly exhibited approximately one-hour advanced peaks in evening activity, which occurred at ZT 10.5–11 (Figs 1D,E and S1 and Tables S3, S4). Thus, the Dh31 mutation did not enhance the abnormal patterns of morning and evening anticipation caused by the Pdf mutation.

In summary, we determined that the lack of DH31 strongly enhanced the arrhythmic phenotype induced by the Pdf mutation but did not affect the free-running period, morning anticipation or timing of the evening activity peak in Dh31^#51;Pdf¹ double mutants. Therefore, we examined the functions of DH31 in regulating free-running rhythmicity, with a particular focused on the relationship between DH31 and PDF.

Given that Dh31^#51;Pdf¹ double-mutant flies exhibited disrupted free-running rhythmicity and that Dh31^#51 mutants still showed normal free-running rhythmicity (Fig. 1), it is likely that an abnormal Dh31^#51 phenotype might be invisible in the presence of normal PDF signaling. To examine this possibility, we asked whether DH31 expression in clock neurons could overcome the changes in rhythmicity identified in Dh31^#51;Pdf¹ double mutants.

Because we and others showed that DH31 is expressed in a subset of circadian clock cells (DN1ps)^14,15, we first expressed DH31 in DN1ps using R18H11-Gal4 (a DN1p driver) (Fig. S4B). However, 73% of the rescued flies exhibited arrhythmicity (Figs S3G, S4B, and Table 1: R18H11-Gal4 > UAS-Dh31, Dh31^#51;Pdf¹), which was similar to the results for the Gal4 and UAS control flies (Fig. S3D,F and Table 1: R18H11-Gal4/+, Dh31^#51;Pdf¹, +/UAS-Dh31, Dh31^#51;Pdf¹). These data suggest that DH31 expression in R18H11-Gal4-expressing neurons is not sufficient to restore free-running rhythmicity.

R18H11-Gal4 is expressed in only ~ four to six DN1ps¹⁴, and we also found that DH31 is expressed in anterior DN1s (DN1as) (Supplemental Fig. S5). As such, DH31 expression from non-R18H11-Gal4-expressing DN1s might also play a role in regulating free-running rhythmicity. We therefore expressed DH31 using tim-Gal4 (a driver for all clock cells) in the Dh31^#51;Pdf¹ double mutants and assessed whether DH31 expression alone could prevent the severe arrhythmicity. We determined that 51% of the rescued flies showed restored free-running rhythmicity (Figs 2B and S3P, and Table 1: tim-Gal4 > UAS-Dh31, Dh31^#51;Pdf¹), while the Gal4 and UAS control flies from the double-mutant background still exhibited severe arrhythmic phenotypes (Figs 2B and S3B,F, and Table 1: tim-Gal4/+, Dh31^#51; Pdf¹, +/UAS-Dh31, Dh31^#51; Pdf¹). The data indicated that DH31 expression in tim-Gal4-expressing neurons in Dh31^#51;Pdf¹ mutants is sufficient to recover a similar level of rhythmicity to that of Pdf¹ mutants (Fig. 2A,B) and suggested that DH31 in tim-Gal4-expressing neurons contributes to free-running rhythmicity.

Given that Pdf¹ single mutants exhibited an arrhythmic phenotype, PDF should be able to regulate a free-running rhythm without DH31. To confirm this hypothesis, PDF was expressed in LNvs using Pdf-Gal4 (a LNv driver) in Dh31^#51;Pdf¹ double-mutant flies. These flies strongly recovered their free-running rhythm, with a rhythmicity level that was similar to that of Dh31^#51 and WT flies (Figs 2A,C and S3O, and Table 1: Pdf-Gal4 > UAS-Pdf, Dh31^#51;Pdf¹), suggesting that PDF secretion from LNvs is sufficient to restore rhythmicity in these double-mutant flies. Taken together, we concluded that PDF and DH31 regulate free-running rhythmicity in a hierarchical fashion, in which PDF functions in a primary role, and DH31 functions in a secondary role.

Importantly, the disruption of morning anticipation and the advanced peak in evening activity in Dh31^#51;Pdf¹ double mutants were also recovered in Pdf-Gal4 > UAS-Pdf, Dh31^#51;Pdf¹ flies (S2I,L,M and Tables S1 and S2). These data also highlight that the overexpression of PDF from LNvs is sufficient to restore normal phenotypes in the double mutants and to prevent the abnormal morning anticipation phenotype in Dh31 mutants.

To determine how PDF and DH31 regulate free-running rhythmicity at the cellular level, we first verified that both PDF and DH31 act on clock cells to regulate free-running rhythmicity. Membrane-tethered peptides have both linker and anchor peptides that couple with the cell membrane, which results in cell-autonomous binding and the constant activation of the receptors on specific cells^17,18. By using tethered-PDF (t-PDF), we determined that t-PDF expression in all clock cells using tim-Gal4 restored rhythmicity to 71% in the double mutants, showing a rhythmicity close to that of the Dh31^#51 flies (Figs 3A,C and S3B,H,I, and Table 1: tim-Gal4 > UAS-t-Pdf, Dh31^#51;Pdf¹). Similarly, t-DH31 expression in all clock cells using tim-Gal4 also restored rhythmicity to 37%, which was similar to the rhythmicity of the Pdf¹ flies (Figs 3A,B and S3A–C, and Table 1: tim-Gal4 > UAS-t-Dh31, Dh31^#51;Pdf¹). Therefore, we confirmed that both PDF and DH31 act on clock cells to regulate free-running rhythmicity.

LNvs are the main clock cells that regulate locomotor activity rhythms^2,3,5, and bath application of PDF or DH31 can activate LNvs¹⁹. Therefore, we assessed whether DH31 and PDF act on LNvs to regulate rhythmicity. However, neither t-DH31 nor t-PDF expression in LNvs using Pdf-Gal4 was able to restore rhythmicity compared to the phenotypes of the Gal4 control flies (Figs S3K,L,M and S4C,D and Table 1: Pdf-Gal4 > UAS-t-Dh31 and Dh31^#51;Pdf¹, Pdf-Gal4 > UAS-t-Pdf, Dh31^#51;Pdf¹). Therefore, our data suggest that both PDF and DH31 are less likely to act on LNvs to modulate rhythmicity.

We subsequently focused on DN1s because small LNvs (sLNvs) project to DN1s^20–22, PDF acts on DN1ps to regulate locomotor activity^22,23, and Pdfr expression in DN1ps restores the dampened free-running rhythm in Pdfr mutant flies²³. We determined that flies with t-PDF expression in DN1ps using R18H11-Gal4 showed 83% rhythmicity, which was close to the rhythmicity of the Dh31^#51 flies (Figs 3A,C and S3J, and Table 1: R18H11-Gal4 > UAS-t-Pdf, Dh31^#51;Pdf¹). Therefore, we confirmed that t-PDF expression in DN1ps rescues the severe disruption of free-running rhythm identified in Dh31^#51;Pdf¹ double mutants.

To determine whether DH31 also acts on the same group of clock cells, we expressed t-DH31 in DN1ps and assessed the effect on free-running rhythmicity. When t-DH31 was expressed in the DN1ps of Dh31^#51;Pdf¹ double-mutant flies using the R18H11-Gal4 driver, the free-running rhythmicity was restored to the same level as that of the Pdf¹ flies (38%, Figs 3A,B and S3E, and Table 1: R18H11-Gal4 > UAS-t-Dh31, Dh31^#51;Pdf¹). The rhythmicity of these flies was significantly different from that of the Gal4 or UAS control flies from the double-mutant background (Fig. 3B and Table 1). These results suggest that DH31 also acts on DN1ps to modulate free-running rhythmicity. Thus, our data suggest that both PDF and DH31 can act on DN1ps to regulate free-running rhythmicity.

To further examine the role of DH31 in the arrhythmic phenotype, we focused on the molecular oscillations in each clock cell. Given that the Pdf mutation causes abnormal molecular oscillations in clock cells^8,9 and that the Dh31 mutation enhanced the abnormal free-running rhythms caused by the Pdf mutation (Fig. 1A), we suspected that molecular oscillations in the clock cells of Dh31^#51;Pdf¹ double mutants would be severely dampened compared to those of Pdf¹ mutants.

To test this possibility, we examined the molecular oscillations in each group of clock cells (LNv, LNd, and DN1) by measuring the expression of Vrille (VRI), which is a component of the second clock feedback loop in the core molecular clock system²⁴. In most cases, severe arrhythmicity in the Dh31^#51;Pdf¹ double mutants was identified by three days after the shift from LD to DD; thus, we compared the expression levels of VRI in each mutant maintained at DD3 (Fig. S3Q). We determined that the intensities of VRI expression in LNvs and LNds in all mutant flies (Dh31^#51, Pdf¹ and Dh31^#51;Pdf¹) were less than those in WT flies at ZT 13 and 19 (Figs 4A,B and S6 and Tables S5, S6). Notably, the peak of VRI expression in LNds in Dh31^#51 was delayed compared to that in WT and the other mutants (Figs 4B and S6), suggesting that DH31 is involved in regulating molecular rhythms in LNds. Furthermore, the Pdf¹ and Dh31^#51;Pdf¹ mutants exhibited a severe disruption in the molecular rhythm in their DN1s (Figs 4C and S6 and Tables S5, S6). However, there was no significant difference between the Pdf¹ and Dh31^#51;Pdf¹ mutants in LNvs, LNds or DN1s (Figs 4 and S6 and Tables S5, S6). Therefore, the data did not show a significant difference in VRI expression between Pdf¹ and Dh31^#51;Pdf¹ double mutants, suggesting that the Dh31 mutation does not enhance the abnormal molecular oscillations caused by the Pdf mutation. Thus, the role of Dh31 in regulating free-running rhythmicity may differ from that of Pdf.

We subsequently focused on the functions of PDF and DH31 receptors in locomotor activity rhythms. In Dh31 receptor (Dh31r) loss-of-function mutant flies (Dh31r^f05546/Df(2 R)BSC273, referred to here as Dh31r¹/^Df), Dh31r mRNA levels in the head were 38% of those observed in w¹¹¹⁸ flies²⁵. We recently found that these mutant flies showed abnormal TPRs; however, they exhibited normal locomotor activity rhythms (Fig. 5A,D,H and Table 1: Dh31r^1/Df)²⁵. Given the phenotype of Dh31^#51;Pdf¹ double mutants in locomotor activity rhythms, we expected the Dh31r mutation to also enhance the Pdfr mutant phenotype. To this end, we created Pdfr⁵³⁰⁴;Dh31r^1/Df double mutants. First, we confirmed that Pdfr⁵³⁰⁴ mutants exhibited a weak rhythmicity (51% rhythmic and power = 323.6), shorter period (23.3 h), loss of morning anticipation and advanced phases in the evening activity peak (Fig. 5A,B,F and Tables 1 and S1), which were very similar to the phenotype of Pdf¹ mutants. These findings suggest that Pdfr⁵³⁰⁴ mutants represent a phenocopy of Pdf¹ mutants^12,26,27. However, the Pdfr⁵³⁰⁴;Dh31r^1/Df double-mutant phenotype still exhibited 60% rhythmicity (Fig. 5A,I and Table 1), which was similar to that of the Pdfr⁵³⁰⁴ single mutants (51% rhythmic, Table 1). The morning anticipation and phase-advanced evening activity peak phenotypes were also not influenced by the double mutation of Pdfr and Dh31r (Fig. 5E and Tables S1 and S2). Thus, these data indicate that the Dh31r mutation did not enhance the arrhythmic phenotype caused by the Pdfr mutation.

Furthermore, as a control, we generated Pdfr⁵³⁰⁴;Dh31^#51 double mutant flies and tested their locomotor activity rhythms. Pdfr⁵³⁰⁴;Dh31^#51 double mutant flies showed only 23% rhythmicity in free running (Fig. 5A,G and Table 1), which was significantly lower than that of Pdfr⁵³⁰⁴ or Pdfr⁵³⁰⁴;Dh31r^1/Df double-mutant flies (51% or 60%, respectively) (Fig. 5A,F,I and Table 1). The results suggested that both PDF and DH31 signals regulate robust locomotor activity rhythms but that DH31R is unlikely to regulate locomotor activity rhythms.

We demonstrated a novel function of DH31 in regulating Drosophila locomotor activity rhythms. We showed that Dh31^#51 mutants maintained a robust free-running rhythm (Fig. 1 and Table 1), whereas Dh31^#51;Pdf¹ double-mutant flies exhibited a severe disruption of their free-running rhythm compared to Pdf¹ mutants (Fig. 1 and Table 1). These findings suggest that Dh31^#51 mutants maintain a robust free-running rhythm because the primary factor, PDF, can sustain a strong rhythm (Figs 1 and 6). We showed that ~40% of Pdf¹ single-mutant flies exhibited a preserved rhythmic state, which is because DH31 can partially support free-running rhythmicity. Thus, the severe disruptions of free-running rhythm in Pdf¹ and Dh31^#51 double-mutant flies is likely caused by the loss of both pathways.

PDF is secreted from the main circadian neurons, LNvs, and acts on other clock cells through PDFR to synchronize and maintain robust molecular rhythms^11,19. We showed that PDF expression from LNvs in Dh31^#51;Pdf¹ mutants restored rhythmicity (Fig. 2C), in contrast to t-PDF expression in LNvs (Fig. S4D), indicating that an autoreceptor of PDF signals in LNvs is not sufficient to maintain rhythmicity. Instead, we showed that t-PDF expression in DN1ps restored rhythmicity, which suggests that PDF signaling in DN1ps is sufficient to maintain robust free-running rhythmicity (Fig. 3C). Recently, the responsiveness to PDF was shown to be strongly altered for 24 h via RalA GTPase in sLNvs²⁸. Therefore, we expect that the continuous activation of PDFR by t-PDF generates rhythmic downstream signaling in PDFR-expressing neurons.

We also showed that molecular oscillations in DN1s were strongly dampened in Pdf¹ mutants compared with WT flies (Fig. 4C). These data are consistent with previous studies in which the molecular oscillations of PER in Pdf¹ mutants held under DD conditions were dampened in DN1s¹⁰ and the genetic manipulation of the circadian clocks in PDF-positive cells altered the molecular rhythms in DN1ps^23,29. Furthermore, Pdfr expression in DN1ps has been reported to prevent the arrhythmic phenotype in Pdfr⁵³⁰⁴ mutants²³. These findings support the idea that PDF is secreted from LNvs and acts on DN1ps to regulate free-running rhythmicity.

Furthermore, we showed that t-DH31 expression in DN1ps rescued the Pdf¹ and Dh31^#51 double-mutant phenotypes, which suggests that DH31 acts on DN1ps to regulate rhythmicity (Fig. 3B). Although it has been suggested that DH31 release might increase at dawn¹⁴ and that DH31-mRNA expression levels oscillate for 24 h¹³, how t-DH31 expression causes rhythmic behavioral output remains unclear. Because DH31 can modestly activate PDFR in vitro¹², we cannot exclude the possibility that t-DH31 overexpression might simply activate PDFR in DN1ps instead of the intrinsic PDF signals, thereby restoring locomotor activity rhythms in the flies. However, the rhythmicity of Dh31^#51;Pdf¹ mutants overexpressing t-DH31 in tim-Gal4-expressing neurons or R18H11-Gal4-expressing DN1ps only reached levels similar to that of the Pdf¹ single-mutant flies (Fig. 3A,B). Therefore, DH31 likely acts on DN1ps separately from the PDF pathway.

Although we and others have shown that DH31 is expressed in a subset of DN1ps^14,15, DH31 expression using R18H11-Gal4 did not rescue the Pdf¹ and Dh31^#51 double-mutant phenotypes (Fig. S4B), suggesting that DH31 expression in R18H11-Gal4-expressing neurons is insufficient to maintain rhythmicity. Instead, we showed that DH31 is expressed in DN1as (Fig. S5) and that DH31 expression in tim-Gal4-expressing neurons rescued the phenotype (Fig. 2B), which suggests that DH31 expression in clock neurons maintains rhythmicity. That said, given that DH31 is expressed in nonclock neurons^14,15 and that tim-Gal4 is expressed in nonclock cells³⁰, we cannot exclude the possibility that DH31 expression in nonclock neurons might play a role in rescuing the severe phenotype of Dh31^#51;Pdf¹ mutants. Alternatively, although DH31 expression in LNvs was not detectable via anti-DH31 antibody staining¹⁵, a recent RNA-seq analysis detected Dh31 gene expression in both LNvs and DN1s¹³. Therefore, DH31 expression from LNvs may potentially act on DN1s to support locomotor activity rhythms.

In summary, we propose that PDF and DH31 regulate free-running rhythms in a hierarchical fashion in DN1ps (Fig. 6). As t-DH31 or t-PDF expression in DN1ps resulted in a similar level of rhythmicity as that observed in flies expressing t-DH31 or t-PDF, respectively, in tim-Gal4-expressing neurons (Fig. 3), DN1ps are at least one of the important clock cells that regulate free-running rhythmicity.

Given that Dh31^#51;Pdf¹ mutants exhibited severe arrhythmicity in free-running rhythm, we speculated that the severe arrhythmic phenotype might be a result of abnormal molecular oscillations. However, the molecular oscillations of Dh31^#51;Pdf¹ mutants were similar to those of Pdf¹ mutants (Fig. 4). Therefore, the molecular mechanisms by which DH31 regulates free-running rhythms still remain unclear. Importantly, the peak of VRI expression in LNds in Dh31^#5 was at ZT 19, which was delayed compared with those of WT flies and the other mutants (Fig. 4B). The data suggested that DH31 is involved in the regulation of molecular oscillations in LNds. Because LNds are the evening pacemaker^2,3, the delayed VRI oscillations in LNds might be associated with the longer period of free-running rhythm in Dh31^#51 (24.4 h, Table 1).

Recently, the intracellular calcium rhythms in each clock cell were reported to be nonsynchronous and associated with morning and evening peaks in locomotor activity⁴. DH31 signaling may possibly contribute to the downstream output that controls molecular rhythms in pacemaker processes, such as intracellular calcium rhythms. Given that PDF from sLNvs regulates strong molecular rhythms in DN1ps and generates robust free-running rhythms under constant conditions^10,23,29, DH31 may help maintain vigorous output signals downstream of the molecular clocks in DN1ps.

We recently showed that both Dh31r^1/Df mutants and flies undergoing Dh31r knockdown in their neurons showed normal rhythmicity in the locomotor activity rhythm²⁵. In contrast to Dh31^#51;Pdf¹ double mutants, Pdfr⁵³⁰⁴;Dh31r^1/Df double mutants did not enhance the arrhythmicity observed in Pdfr single mutants (Fig. 5A,F,I and Table 1), which suggests that DH31R does not complement PDFR function; thus, DH31R does not function as a receptor for DH31 in this context. Given that Dh31r^1/Df flies showed a strong abnormality in the TPR phenotype²⁵, it is more likely that DH31R does not play an important role in locomotor activity rhythms. However, Dh31r^1/Df mutants are not null²⁵, and we cannot exclude the possibility that a small amount of residual DH31R might drive robust locomotor activity rhythms with the PDF pathway.

Which receptors might function with DH31 to regulate free-running rhythmicity? Given that DH31 can activate PDFR in vitro¹², bath applications of DH31 can activate LNvs via PDFR¹⁹ and DH31 can function as a ligand of PDFR in TPR at the onset of night¹⁵, PDFR may function as a receptors for both DH31 and PDF in the regulation of free-running rhythmicity. However, because the arrhythmicity of Pdfr⁵³⁰⁴ mutants was not as severe as that of Dh31^#51;Pdf¹ mutants (Fig. 5), PDFR does not appear to act as a receptor for DH31 in this context (Fig. 6).

Both DH31R and PDFR are class II G-protein coupled receptors (GPCRs), which also include Hector and Diuretic hormone 44 receptors 1 and 2 (DH44R1 and DH44R2, respectively)^31,32. Interestingly, the DH44R1 and DH44R2 ligand DH44 has been implicated in circadian output circuits^21,33. Therefore, although there is no evidence from in vitro or in vivo experiments, these receptors might nevertheless function as receptors for DH31 to regulate free-running rhythmicity.

The orchestration of neuropeptides is critical for regulating circadian clock functions in species that range from flies to mammals. In mammals, several neuropeptides, including vasoactive intestinal polypeptide (VIP), arginine vasopressin (AVP) and neuromedin S (NMS), are expressed in the SCN, which is the center for circadian clock control^34,35. The hierarchy of neuropeptide signaling contributes to circadian function in the SCN³⁶. Several recent studies in Drosophila have identified the neuropeptides, including ion transport peptide (ITP)³⁷, neuropeptide F (NPF)³⁸, allatostatin A³⁹, short neuropeptide F⁴⁰, leucokinin³³ and DH44²¹, that regulate locomotor activity and sleep. However, given that DH31 complements the function of PDF in regulating free-running rhythmicity in the same clock cells, DH31 not only serves as one of the neuropeptides that regulates circadian rhythms but also might selectively influence PDF function in the regulation of free-running rhythms. Thus, our findings shed new light on the next steps required to improve our understanding of the core neuropeptide regulatory mechanisms involved in the circadian rhythm.

All flies were raised in 12 h light/dark cycles at 25 °C; Zeitgeber Time (ZT) 0 was lights-on, ZT12 was lights-off. w¹¹¹⁸ (RRID:BDSC_3605) flies were used as WT flies. UAS-Dh31 was a kind gift of Dr. Paul Taghert. Membrane-tethered DH31 (UAS-t-DH31-ML:B4) and membrane-tethered PDF (UAS-t-PDF-ML:M2a) were used¹⁷. yw; Pdf¹(RRID:BDSC_26654) flies, yw;Dh31^#51 (FBal0304655) flies¹⁶ and Pdfr⁵³⁰⁴ (RRID:BDSC_33068) flies were backcrossed with w¹¹¹⁸ flies. tim-Gal4 (RRID:BDSC_7126) (expressed in all clock neurons), Pdf-Gal4 (RRID:BDSC_6900) (expressed in LNvs), and R18H11-Gal4 (RRID:BDSC_48832) (expressed in ~4–6 DN1s)^14,20 were obtained from the Bloomington Drosophila Stock Center (stock #7126, # 6900 and #48832, respectively). All Gal4 driver and UAS reporter flies from the Dh31^#51;Pdf¹double-mutant background were generated via chromosome recombination with w;Dh31^#51;Pdf¹ double-mutant flies. Dh31r¹ is a P-element insertion mutant (PBac{WH}Dh31-R^f05546) and was obtained from the Exelixis Collection at the Harvard Medical School. Dh31r^Df is a deletion mutant (Df(2R)BSC273) (RRID:BDSC_23169) and was obtained from the Bloomington Drosophila Stock Center (stock # 23169).

Locomotor activity assays and data analysis were performed as previously described^15,41,42. Flies were reared under 12 h light/dark (LD) cycles at 25 °C. Male flies (1 to 5 days old) were used in the locomotor activity experiments. A Drosophila Activity Monitoring (DAM) system (http://www.trikinetics.com/↗) was placed in an incubator (MIR-154, Sanyo Scientific, Japan). Lights in the incubator (15-W cool white fluorescent lamps (FL15D, TOSHIBA, Japan)) were connected to an electric timer; the light intensity was approximately 800 lux. Locomotor activity was monitored in 12 h LD cycles for five days and in a constant dark condition for more than ten days at 25 °C. The data were analyzed using Actogram J software⁴³. Free-running periods and power values were calculated using a chi-square periodogram^42,44, and flies with a power value < 100 were defined as arrhythmic. Only data from rhythmic flies were averaged to generate a double-plotted actogram. Morning anticipation index (AI) values were calculated as previously described^22,45,46. Briefly, AI = (total activity 3 h before lights-on)/(total activity 6 h before lights-on). The AIs of all flies over days 2–5 of the LD cycles were averaged in each genotype. The AIs for different genotypes were compared using Tukey’s multiple-comparison test. The time of evening peaks in all flies over days 2–5 of the LD cycles were averaged in each genotype. The averaged time of evening peaks for different genotypes were compared using Tukey’s multiple-comparison test.

The signals from VRI antibody staining are relatively strong and specific⁴⁷. Given that VRI is not degraded by light, we think that VRI staining is easier to handle compared with TIM staining. Therefore, we used VRI to examine the molecular oscillations in each group of clock cells (LNv, LNd, and DN1) in Fig. 4. w¹¹¹⁸, Dh31^#51, Pdf¹, and Dh31^#51;Pdf¹ flies were raised under 12 h LD cycles at 25 °C. Adult male flies were subsequently entrained to LD cycles for 3 to 4 days and then shifted to constant dark conditions for 3 days (DD3). The flies were fixed at CT 1, 7, 11 or 19 with 4% paraformaldehyde in PBST (PBS plus 0.3% Triton X-100) for 2 h at room temperature, following brain dissection. Immunostaining was performed as previously described¹⁵. Briefly, 5% normal goat serum in PBST was used for blocking and antibody incubations with guinea pig anti-VRI antibody (1:200)⁴⁷ and donkey anti-guinea pig Alexa Fluor® 647 (RRID:AB_10895029) (1:200, Jackson ImmunoResearch). Mounted brains were scanned using a Zeiss LSM5 Pascal confocal microscope. Images were digitally projected as Z-stacks for immunohistochemical analysis. ImageJ software was used to quantify the intensity of the immunostaining signal in each single cell (LNv, LNd and DN1). After background subtraction, the total intensity of each type of clock cell in a brain hemisphere was determined, and the average intensity of 10 brain hemispheres was calculated in Excel.

The authors declare no competing interests.