Relationship between daily rhythms and seasonal timing changes across latitudes in Drosophila littoralis

We investigated the photoperiodic response of D. littoralis strains collected across a latitudinal gradient in Europe (figure 1a) by exposing flies to days with different photoperiods but a constant temperature of 16°C, thus mimicking summer-like (long days and short nights) or winter-like (short days and long nights) light conditions in the laboratory. Photoperiods that the flies experience in nature are shown for four strains stemming from different latitudes in figure 1b.

Photoperiodism in flies was assessed by evaluating the stage of ovarian maturation in females, which is a marker of reproductive diapause [24]. To determine the CNLs at which 50% of the flies of each population enter diapause, we generated photoperiodic response curves for six previously studied strains (Kilpisjärvi, Rovaniemi, Zürich, Ticino, Khobi and Batumi) [12] and three more recently collected strains (Portugal, Petäjävesi and Raattama) (figure 1c; electronic supplementary material, table S1). Correlation analysis revealed a negative correlation between latitude of origin and CNL (figure 1d). This is consistent with previously reported data by Lumme & Oikarinen [6] and Lankinen [12], who found a latitudinal cline in the onset of diapause of D. littoralis. Importantly, the diapause levels of the previously studied strains still follow the same latitudinal cline as previously found [12], despite the long time in captivity, and the three new strains fit well into it, as can be seen from the photoperiodic response curves of all strains (figure 1c; electronic supplementary material, table S1).

Overall, we have established a panel of laboratory inbred D. littoralis strains with a robust photoperiodic phenotype and a clear latitudinal cline in the CNL (figure 1d).

To test whether circadian behaviours of D. littoralis vary also with the latitude of origin, we recorded locomotor activity of five selected D. littoralis strains (Kilpisjärvi, Raattama, Zürich, Portugal and Batumi) in the Drosophila Activity Monitor system (Trikinetics, Princeton, MA, USA) at three different photoperiods (LD8:16, LD12:12 or LD20:4) at 20°C. D. melanogaster, the common laboratory fruit fly, served as a reference as most of the knowledge about the circadian clock in flies derives from this species [25]. Remarkably, a photoperiod of LD12:12 (12 h of light followed by 12 h of darkness, equinox-like) corresponds to winter light conditions at high latitudes, whereas at low latitudes it corresponds to light conditions of spring and autumn days. Furthermore, extreme photoperiods such as LD20:4 are typical summer photoperiods for the northern D. littoralis strains, but these light conditions are not experienced in nature by the southern European strains. By contrast, LD8:16 is a short photoperiod that occurs throughout continental Europe and in the northern part of the D. littoralis range, but that will not impact locomotion of northern populations because under this photoperiod flies are inactive (diapausing).

Individual flies were recorded for 5 days under LD8:16, LD12:12 or LD20:4 and then released into constant darkness (DD) to assess their free-running rhythmicity in the absence of external environmental cues. The average rhythms of the different fly populations are shown in figure 2a, while actograms of individual flies are shown in electronic supplementary material, figure S1. The Lomb–Scargle periodogram was used to quantify the endogenous period length (figure 2b; electronic supplementary material, table S3) and to determine the percentage of rhythmic flies for each strain (figure 2c). As shown previously, D. melanogaster flies are rhythmic when released in DD, regardless of the photoperiod to which they were previously exposed [26]. By contrast, southern D. littoralis strains (Batumi and Portugal) showed a higher rhythmicity in DD when they had previously experienced short, but not long, photoperiods (figure 2a–c). The loss of rhythmicity after exposure to long photoperiods likely reflects the adaptation of the circadian clock of southern flies to only moderate changes in day lengths. D. littoralis strains from the northernmost collection sites (Kilpisjärvi and Raattama), however, rapidly became arrhythmic under DD after all three photoperiods. Behavioural arrhythmicity under constant conditions of northern fly populations is a known trait which has been reported for several species [13–16,19]. Here we show that within a single species this trait is plastic and latitude dependent. The correlation analysis revealed a significant negative correlation between circadian rhythmicity in DD and the latitudes from which the strains originated (figure 2h). The free-running periods of the rhythmic flies were not significantly different in the D. littoralis strains (electronic supplementary material, table S3).

To extract further behavioural markers known to be under the control of the circadian clock, we analysed the average activity profiles of the different D. littoralis strains under the LD conditions that preceded their release into DD. We found that southern strains showed a bimodal activity pattern with a clear morning peak (M_peak) and evening peak (E_peak) and a reduced activity (increased sleep) in the middle of the light phase (siesta or light-phase sleep). The northern strains (Raattama and Kilpisjärvi) instead showed a higher variability in activity between individuals, resulting in noisier activity patterns (figure 2d). The strain from Zürich showed an intermediate activity profile between the southern and northern strains (figure 2d) as was already seen for its rhythmicity in DD (figure 2a–c). Notably, under LD8:16, the activity pattern of the southern D. littoralis strains (Batumi and Portugal) was more similar to the activity of D. melanogaster than to that of the northern D. littoralis strains (Raattama and Kilpisjärvi). Under LD12:12, all strains showed some degree of bimodal activity, with the northern D. littoralis strains showing higher activity levels with less pronounced siesta (figure 2d,f). At LD20:4, the siesta (light-phase sleep) became less pronounced in all flies, but it remained less deep in the northern strains (figure 2d,f; electronic supplementary material, figure S2), revealing a clear latitudinal cline (figure 2k). In addition, major differences between D. melanogaster and the different D. littoralis strains occurred in the timing of M_peak and E_peak under LD20:4. D. melanogaster and the southern D. littoralis strains from Batumi and Portugal showed only little morning activity, which, if present, started well after light-on (L_ON). Evening activity, on the other hand, started early and decreased already before light-off (L_OFF). The northern strains (Raattama and Kilpisjärvi) showed instead a more pronounced M activity shortly after L_ON and an E activity that remained high until the end of the day (figure 2d). The quantification of the M_peak and E_peak under the longest photoperiod (LD20:4) showed a late M_peak and early E_peak in the southern strain (Portugal), an intermediate phenotype in the central Europe strain (Zürich) but an early M_peak and late E_peak in the northern ones (Raattama and Kilpisjärvi; figure 2e). For the southern Batumi strain, we could not determine the E_peak because it was masked by the strong L_OFF peak (see below). The timing of M_peak and E_peak of the northern strains was more plastic than that of the southern strains meaning that they could better follow L_ON and L_OFF under very long photoperiods. The values of the estimated M and E peaks were significantly different between northern and southern D. littoralis strains tested and showed a clear latitudinal cline (figure 2i,j).

Other features of the daily activity patterns of Drosophila such as the pronounced sudden increases of locomotor activity directly after L_ON and L_OFF transitions are known to be independent of the circadian clock [27,28]. In D. melanogaster these L_ON and L_OFF peaks, also referred to as startle responses, were present under every photoperiod tested (figure 2d). Interestingly, the startle response at L_ON is completely absent in all D. littoralis strains, while the startle response at L_OFF is reduced in most strains except for the southern Batumi strain (figure 2d). Absence of the startle responses has already been reported in other fly species of the virilis group collected at high latitudes, such as Drosophila lummei and D. ezoana [11,19], suggesting that this is due to a decrease in light sensitivity with increasing latitude.

In summary, we found a negative correlation of several clock-dependent phenotypes with the latitude of origin, such as the average percentage of rhythmicity in DD (figure 2h), phase of the M_peak (figure 2i) and the E_peak (figure 2j), and sleep hours during the light phase (figure 2k). All these measured values decreased with increasing latitude. The present results illustrate that the adaptation of D. littoralis to different latitudes is not only reflected in seasonal responses such as reproductive diapause (figure 1), but also in the circadian clock that controls daily behaviour (figure 2).

We previously found that in the northern D. littoralis strain Kilpisjärvi the circadian clock protein PERIOD (PER) was expressed in all circadian clock neurons known in D. melanogaster (figure 3a), with the exception of the dorsal clock neurons 2 (DN₂) and the large ventrolateral neurons (l-LN_v) [18]. Here, we extended this analysis to a southern D. littoralis strain (Batumi) and additionally performed immunostaining against the clock components Par Domain Protein 1 (PDP1) and PDF. In the DN₂ of both D. littoralis strains, we still could not detect clear staining, but PDP1 and PDF staining were present in the l-LN_v of both strains (electronic supplementary material, figure S3), suggesting that all clock neurons are present in D. littoralis, but that some of them express only small amounts of clock proteins and are therefore difficult to detect.

We performed time series of immunostaining in D. littoralis from Kilpisjärvi and Batumi, with an antibody against PDP1 under three photoperiods (LD8:16, LD12:12 and LD20:4) to quantify circadian protein oscillation in the clock neurons. PDP1 staining was quantified in three groups of lateral clock neurons (the l-LN_v, s-LN_v and the dorsolateral neurons, LN_d) as these groups are responsible for timing M and E activity peaks in D. melanogaster [29], and PDP1 has been shown to oscillate strongly in them [30]. Furthermore, we quantified PDP1 staining in one group of dorsal neurons, the DN₁, which is of similar importance in the control of M and E peaks of D. melanogaster [31], most likely by promoting the siesta between them [32]. Therefore, a subgroup of the DN1_p has been named ‘siesta neurons’. We revealed interesting differences in PDP1 oscillations between the two strains and D. melanogaster.

In D. melanogaster, PDP1 oscillated in all clock neurons examined and under all photoperiods, albeit with reduced amplitude under photoperiods different from LD12:12, which is consistent with previous observations (figure 3b, table 1; electronic supplementary material, tables S4 and S5) [30]. Under very long photoperiods (LD20:4) the PDP1 oscillations became bimodal and, most strikingly, very high in the DN₁ (the siesta neurons) and LN_d (the E neurons) (figure 3b, table 1; electronic supplementary material, tables S4 and S5). This matches the activity patterns under LD20:4, where the siesta was long and E activity most pronounced (figure 2d).

In the southern D. littoralis strain (Batumi), PDP1 cycling was found to be rhythmic in most clock neurons at LD8:16 and LD12:12, albeit the oscillations were not as synchronous as in D. melanogaster (figure 3c, table 1; electronic supplementary material, tables S4 and S5). However, under the very long photoperiod (LD20:4), the LN_d and l-LN_v showed a bimodal pattern with high amplitude, while PDP1 levels became again high in the DN₁ and lost rhythmicity in the s-LN_v (figure 3c; table 1; electronic supplementary material, table S4). Again, this staining pattern coincides with the locomotor activity patterns, which resembled those of D. melanogaster, but were generally noisier, particularly under the very long photoperiod (figure 2d).

In the northern D. littoralis strain (Kilpisjärvi), we could see a clear staining of neither PDP1 nor PER in the s-LN_v (electronic supplementary material, figure S3g). Moreover, under LD8:16, we detected no significant daily cycling of PDP1 in any of the clock neurons (table 1), and if there were oscillations, these were largely out of phase with each other (figure 3d; electronic supplementary material, table S4). Under LD12:12, the LN_d and l-LN_v oscillated in a daily manner, while the oscillations in the DN₁ were not significant (table 1). Under the very long photoperiod, the oscillations of the l-LN_v and LN_d became bimodal and of high amplitude (figure 3d). Indeed, JTK-cycle analysis revealed significant 12 h rhythms but no significant rhythms in the 24 h range (table 1; electronic supplementary material, tables S4 and S5). By contrast to the southern D. littoralis strain, PDP1 levels in the DN₁ were much lower and did not increase with increasing photoperiod in the northern strain (figure 3d, table 1; electronic supplementary material, table S5). Again, this pattern of PDP1 cycling fits to the behavioural data, where the flies showed no clear organization in morning and evening activity under short days but exhibited morning and late evening activity with little siesta in between under longer days (figure 2d).

The most striking characteristic of the clock neuronal network in most of the virilis group strains studied so far is that the s-LN_v were not marked by antibodies against PDF [13,15–17,19]. Furthermore, most of these studies showed additional non-clock PDF-positive neurons in the dorsal brain of these species, which were recently shown to be CRZ-positive LNCs [18]. Here, we investigated whether these PDF staining patterns were conserved in the five D. littoralis strains stemming from different latitudes. We found that the s-LN_v terminals of the five strains were marked by anti-PDF to different degrees (figure 4a). While the more southern strains Batumi, Portugal and Zürich showed prominent staining in the s-LN_v terminals, this staining was very faint in the northern strain Raattama and completely absent in Kilpisjärvi as found previously. Another notable difference between the strains was the intensity of the PDF immunostaining in the arborizations of the CRZ neurons. While the Kilpisjärvi strain showed high PDF immunoreactivity in both somata and arborizations of the CRZ neurons, the PDF staining intensity in the arborizations of the CRZ neurons appeared to decrease with latitude (figure 4a).

In D. melanogaster, PDF is essential for rhythmic behaviour under DD conditions [33], and it oscillates in a daily and circadian manner in the terminals of the PDF-positive s-LN_v with a maximum in the morning [34]. Therefore, we tested whether PDF also oscillates in the s-LN_v of the behaviourally rhythmic southern D. littoralis strain from Batumi. Under LD12:12, we found a significant cycling of PDF staining intensity in the s-LN_v terminals with a peak at L_ON (figure 4b; electronic supplementary material, table S8). The PDF oscillation occurs in phase with the nuclear PDP1 oscillation that has already been shown in figure 3c and is again shown as a reference in figure 4b. This result indicates a rhythmic release of PDF from the s-LN_v terminals in the southern strain during LD cycles as was shown in D. melanogaster [34].

To assess whether the cycling of PDF continues under DD conditions, we immunostained the brains of the flies of the behaviourally rhythmic D. littoralis strain Batumi and the behaviourally arrhythmic strain Raattama with antibodies against PDF and PDP1 every 3 h on the second day of DD after being entrained to LD12:12 (figure 4c). We found that the molecular oscillations appear to continue in Batumi, although only PDF cycling turned out to be significant by JTK_CYCLE analysis (electronic supplementary material, tables S6–S8). In the northern strain Raattama, PDF did not significantly oscillate in the s-LN_v terminals (figure 4c; electronic supplementary material, tables S6–S8), which fits well with its behavioural arrhythmicity under DD (figure 2; electronic supplementary material, figure S1).

Remarkably, the amplitude of the PDF oscillation in the s-LN_v terminals follows the same general trend (figure 4d) as the behavioural phenotypes described above (figure 2), suggesting a possible causal effect.

Previous studies in D. melanogaster have shown that PDF is involved not only in the circadian control of behavioural rhythms, but most likely also in the seasonal control of reproductive dormancy [21,22]. The s-LN_v terminals of reproductive flies are significantly longer and more prominently marked by anti-PDF than those of dormant flies [21]. In high-latitude D. littoralis flies, these terminals are PDF-negative, but instead the neurosecretory CRZ neurons are stronger stained by the PDF antibody in reproductive flies (in LD20:4, 23°C) than in diapausing flies (in LD8:16, 10°C) (figure 5b) [18]. By contrast, CRZ levels are low in the same neurons in reproductive flies and high in diapausing flies (figure 5b) [18]. This suggests a putative antagonistic role of PDF and CRZ in the neurosecretory CRZ neurons in seasonal adaptation of northern D. littoralis flies.

To judge the role of PDF in seasonal adaptation in a southern D. littoralis strain, we quantified the immunostaining intensity of PDF and CRZ in the CRZ neurons of the strain Batumi under diapause-inducing conditions (LD8:16, 10°C) and reproductive conditions (LD20:4, 23°C). These neurons can be subdivided by morphology into two groups, six smaller CRZ_1–6 neurons and one bigger neurosecretory CRZ_CN neuron per hemisphere (figure 5a, white arrow) [18,35], hence we measured separately the staining intensities of the two clusters. We compared directly the newly acquired data in Batumi to the previously published ones in Kilpisjärvi [18]. As previously found in the Kilpisjärvi strain, the CRZ immunostaining intensity in both CRZ clusters was also higher in the Batumi strain under diapausing conditions (figure 5b). However, the PDF intensity in the CRZ neurons of the Batumi strain was not significantly different between the two conditions in CRZ_1–6 neurons and slightly increased in the CRZ_CN under diapausing condition (figure 5b). Next, we measured the size of the CRZ neuron somata and found that these were significantly larger under long summer days in both strains (figure 5c) suggesting that more peptides can be stored in the CRZ neuron somata when the flies are reproductive.

The present results indicate that CRZ plays a similar role in inducing diapause in both strains, while PDF in the CRZ neurons is playing a different role in the southern D. littoralis strain compared with the northern one. To test whether PDF from the s-LN_v terminals showed a seasonal variation in the southern D. littoralis strain, as it does in D. melanogaster, we quantified the PDF staining intensity in the PDF-positive s-LN_v terminals in Batumi under simulated winter and summer days. We found a significantly higher PDF staining in the s-LN_v terminals of reproductively active flies than in diapausing flies (figure 5d). This suggests that southern D. littoralis flies control diapause in a similar way to how D. melanogaster flies control dormancy.

Flies originating in northern latitudes face very rapid changes in photoperiod that can best be tracked by weak plastic clocks. Such plastic clocks can better follow not only very long Zeitgeber cycles (see above), but also very long photoperiods [11,19,46]. Indeed, as latitude increased, we found a stronger coupling of M_peak and E_peak to the beginning and end of the photoperiod. We hypothesize that the circadian clock of the common ancestor of the virilis group was like that of the southern strains of D. littoralis and that, being weaker than the clock of D. melanogaster, it allowed this group to invade more northern latitudes. Their clocks, once they reached northern latitudes, may have weakened further, becoming fast-dampening clocks. In parallel, the tracking of dawn and dusk by the weak M and E oscillators improved further.

The next question is whether the different phases of M_peak and E_peak affect the CNL for diapause induction in southern and northern D. littoralis strains. Assuming that the time interval (phase relationship) between E_peak and M_peak codes for night length as predicted by the internal coincidence model [47,48], this time interval is very short for northern flies under LD20:4 and longer for southern flies, as the latter cannot couple their E_peak and M_peak to dusk and dawn under extreme photoperiods (see figure 2e). It is not difficult to imagine that, in the northern strains, the short night signal also sets the CNL to a short night, while the longer night signal of the southern flies leads to a longer CNL. Alternatively, the CNL could also be influenced by the time interval between the beginning of the night and the M_peak, which is also very short for the northern flies under LD 20:4 and significantly longer for the southern flies. This measurement of night length would correspond to the quantitative version of the external coincidence model, which generally fits flies better than the internal coincidence model [2,11,36,41]. In this model, the assessment of night length is based on the accumulation of a ‘diapause’ substance during darkness. As soon as this substance reaches a certain reference threshold, diapause is initiated. In the northern strains, the critical threshold might be reached earlier because diapause-inhibiting factors such as PDF are reduced (see below).

At the molecular level, we found differences in the daily oscillations of several clock components in southern and northern D. littoralis strains that are consistent with the observed behaviour at different photoperiods and constant darkness. In the southern species, the daily PDP1 oscillations were robust and highly synchronous between the assessed clock neurons under short days while they were barely significant and quite out of phase in the northern D. littoralis strain. However, during long days, significant 12 h rhythms occurred in a few clock neurons in the northern D. littoralis strain, which might drive the observed bimodal activity pattern with M_peak and E_peaks. Furthermore, PDP1, PER and PDF were either completely absent or not oscillating in the s-LN_v of the northern strain. This could explain the flies’ behavioural arrhythmicity under DD conditions since these neurons are important for robust activity rhythms under DD conditions in D. melanogaster [33,49,50].

PDF is not only involved in the circadian clock of several insects [51–53] and in behavioural plasticity [54], but it has an additional role in photoperiodism. Blow flies (Protophormia terraenovae) cannot discriminate between long and short days when their PDF-positive clock neurons are ablated [55]. Culex pipiens mosquitoes need PDF to remain reproductive under long day conditions [56], while linden bugs (Pyrrhocoris apterus) [51] and brown-winged green bugs (Plautia stali) [57] need PDF to enter diapause under short days. A role of PDF in photoperiodism was most recently also suggested for the pea aphid Acyrtosiphon pisum [58,59]. In D. melanogaster, PDF released from the s-LN_v acts on the insulin-producing MNCs and inhibits diapause [21,22]. Thus, PDF is needed to keep the flies in the reproductive state similar to mosquitoes. The same seems to be true in the southern D. littoralis strain Batumi, which shows enhanced PDF staining in the s-LN_v terminals under long photoperiods as previously shown in D. melanogaster [21]. Thus, we propose that the diapause-inhibiting effect of PDF is conserved in D. littoralis, although the way in which PDF acts differs in the southern and northern strains. The northern strain from Kilpisjärvi lacks PDF in the s-LN_v terminals and consequently also the diapause-inhibiting effect via the MNCs. This might facilitate the induction of diapause in the northern flies already at longer photoperiods and result in shorter CNLs (see above). On the other hand, under reproductive conditions, the northern strains strongly express PDF in the CRZ neurons and in their fibres running to the corpora cardiaca. There, PDF may act on PDF receptors on adipokinetic hormone-expressing cells and increase metabolic activity [60]. PDF may even be released via the corpora cardiaca into the circulation and could target peripheral tissues that express the PDF receptor [61,62]. In northern D. littoralis, Pdf transcript and PDF peptide levels increase in the CRZ-positive neurons under summer conditions [18]. This could mean that PDF stemming from the CRZ-positive neurons may have overtaken the role of clock neuron-derived PDF in promoting the flies’ reproductivity in summer, although this way seems clearly less effective than the way via the insulin-producing MNCs. CRZ on the other hand, is a good candidate for inducing diapause when days shorten in autumn, and this appears to be similarly true in southern and northern strains as we show here. In summary, northern and southern D. littoralis appear to differ in the way by which PDF keeps them in the reproductive state, but not in the way CRZ promotes diapause. Nonetheless, PDF and CRZ are clearly not the only neurohormones involved in diapause regulation. More studies are needed to unravel the influence of the endogenous clocks on the neurohormonal system in D. littoralis and other insects.

Our study extends the results of previous studies that found a latitudinal cline in eclosion rhythms of D. littoralis to the adult locomotor activity. We found that several behavioural, neuronal and molecular features of the adult circadian clock show latitudinal clines that are consistent with the latitudinal cline of the photoperiodic diapause, suggesting the potential involvement of the circadian timekeeping system in seasonal adaptation. In particular, differences in the expression pattern and oscillation of the neuropeptide PDF in the fly brain could be responsible for the observed differences in seasonal diapause timing between D. littoralis populations living at high and low latitudes. This is an important step in revealing the role of the circadian clock in the seasonal response of insects.

The different D. littoralis strains were collected over different years throughout Europe. The strains collected in Kilpisjärvi (69° N) and Rovaniemi (66° N) (Finland), Zürich (47° N) and Ticino (46° N) (Switzerland), Batumi (41° N) and Khobi (42°N ) (Georgia) were first collected and described in the study by Lankinen [12]. The Portugal strain was collected in Portugal (41° N) in the year 2000. The strains collected in Raattama (68° N) and Petäjävesi (62° N) (Finland) were kindly provided by Maaria Kankare and were collected in 2020 and 2021, respectively (figure 1a). These strains were reared at 23°C under light–dark cycles (LD) 20:4 with 60–70% RH. The D. melanogaster stock used was the Canton-S wild-type line [63] and was reared under LD12:12 at 25°C with 60% RH. All flies were fed on standard cornmeal/agar medium with yeast.

These experiments were performed in Oulu, while all the other experiments presented in this study were performed in Würzburg. The flies were maintained at 16°C, LL and fed on malt medium [64]. Different groups of freshly eclosed female flies were entrained to different photoperiods at 16°C. Their ovaries were dissected after 3 weeks of entrainment; following Lankinen et al. [41] the flies were scored as reproductive in the presence of fully developed eggs and diapausing when little to no yolk was present. Photoperiodic response curves (PPRCs) of the different strains were calculated using the function drm (dose response models) from the R package drc (dose response curve, v4.2.3 [41,65]). The CNL is found at the point of inflection of the PPRC (ED50 = 50% of the flies are diapausing) (see figure 1c).

The locomotor activity of male flies aged between 3 and 7 days post-eclosion was recorded with the Drosophila Activity Monitor (DAM) system (Trikinetics, Princeton, MA, USA) at 20°C. We decided on 20°C because this is an average value between the summer temperatures in the far north and the far south and should be comfortable for all strains. Single flies were placed in glass tubes with a diameter of 0.5 cm for D. melanogaster and 0.7 cm for D. littoralis. The tube was filled on one end with food consisting of 4% saccharose and 2% agar in water, while the other end was blocked by a foam plug. The activity was recorded for approximately 10 days under different LD cycles: LD8:16, LD12:12 or LD20:4 at 20°C. Afterwards, the flies were exposed to constant darkness (DD) to assess their rhythmicity under constant conditions for at least 9 days. Of the 10 days recorded in LD, only the last 5 were used for the calculation of average activity profiles and activity peaks, while circadian rhythmicity, period length and power in DD were calculated for the first 9 days of constant darkness.

To quantify the staining intensity of the neuropeptides PDF and CRZ in the CRZ neurons of reproductive and diapausing flies, 60 virgin females of the Batumi strain were collected and exposed to diapause-inducing conditions (LD8:16, 10°C) or reproductive conditions (LD20:4, 23°C). After 3 weeks, the flies were collected at ZT1 (Zeitgeber time 1, which corresponds to 1 h after lights-on) and fixed as described below. The flies’ ovaries were dissected in a phosphate-buffered saline solution (PBS) and their reproductive state was determined as described above. Afterwards, the brains were dissected and further stained as described below.

Immunohistochemistry was performed on male flies between 3 and 7 days post-eclosion following the protocol from [66]. We immunostained the flies’ brains to define the anatomical position and staining intensity of the circadian clock proteins PER and PDP1 and the circadian clock neuropeptide PDF in different strains of D. littoralis. Flies were fixed for 3.5 h in 4% paraformaldehyde dissolved in PBS with 0.5% Triton X100 (PBST) added. The flies were then dissected after three washes of 10 min with PBS. Subsequently the brains were blocked in a 5% normal goat serum (NGS) solution in PBST 0.5% (this was substituted by normal donkey serum, NDS, when the primary antibody solution contained anti-PER-goat) overnight at 4°C. The brains were then incubated in the primary antibody mix (see table 2 for information on the antibodies) which consisted of the primary antibodies with 5% NGS or NDS and 0.02% NaN₃ in PBST 0.5% for 2 days at 4°C and 1 day at room temperature. After six PBST washes of 10 min, the brains were incubated overnight at 4°C in the secondary antibody mix (see table 2) consisting of the secondary antibodies with 5% NGS (or NDS) in PBST 0.5%. Lastly, the brains were embedded with Vectashield (Vector Laboratories, Burlingame, CA, USA) on microscope slides with approximately 150 µm spacer. The images were acquired with a Leica SPE confocal microscope (Leica Microsystems, Wetzlar, Germany), equipped with a photomultiplier tube and solid-state lasers (488, 532 and 635 nm) for excitation. Images were scanned with a 20-fold magnification glycerol immersion objective (HC PL APO; Leica Microsystems). The confocal stacks had a 2 μm z-step size and either 1024 × 1024 pixels for the quantifications (figures 4 and 5) or 2048 × 2048 (electronic supplementary material, figure S3) with a pixel size of 537 × 537 nm² and a voxel size of 0.537 × 0.537 × 2 μm³. The images were further processed using Fiji (v 1.53 t [69]). Brightness and contrast were eventually changed linearly.

All the samples quantified were stained simultaneously under the same conditions; images were then acquired with the same settings. The quantification was carried out in Fiji on raw images. Since different types of signals were quantified, we used different methods for each quantification. To quantify the nuclear PDP1 staining intensity, a 3 × 3 px² region of interest (ROI) was generated with the Fiji Rectangle selection tool on the brightest focal plane of the nucleus. To quantify the PDF staining intensity in the s-LN_v terminals, the ROIs were set on the 10 pixels with the highest intensity. To quantify the PDF and CRZ staining intensity in the CRZ neurons in the diapausing conditions, the ROIs were generated around the whole cell body. For all the different quantifications, the mean grey value of the ROI was then subtracted by the background intensity ROI value which was set near the quantified region and with a comparable ROI area. Finally, for quantifying the PDF staining intensity in the PDF terminal of Batumi flies kept under long and short days, we used the method described in Hermann-Luibl et al. [70]. First, 8 to 10 confocal stacks containing the s-LN_v terminals were overlayed as a maximal projection, using the function Stacks › Z Project… and Max Intensity as projection type in Fiji. Second, the PDF staining in the CRZ neurons had to be removed. For this purpose, the brains had been fluorescently immunostained by anti-CRZ and anti-PDF, and the CRZ channel was subtracted twice from the PDF channel in ImageJ (Process › Math › Subtract…). This reliably eliminated the PDF staining in the CRZ arborizations, but in rare cases some staining in the cell bodies remained, which was removed manually. In some cases, some PDF staining in the posterior optic commissure or the optic lobes was visible, which was also removed manually. In the next step, the background was subtracted, so that everything in the image remained black, except the PDF terminals. Subsequently, the pixel intensity of the entire image was measured with the tool Analyze > Measure in Fiji.

Except for the actograms (electronic supplementary material, figure S1), which were generated with ActogramJ (a Fiji plugin) [71], all other graphs were generated with R (v4.2.3) via Rstudio (v2023.09.1+494), using the ggplot2 package [72]. The locomotor activity and sleep graphs were generated using the Rethomics framework [73]. The sleep amount was calculated with the function sleep_dam_annotation which scores as asleep individuals which are immobile for a minimum of 5 min. The average locomotor activity was calculated starting from the averaged locomotor activity of single flies over the last 5 days of LD; this was subsequently averaged among all flies and smoothed with a moving average filter over 11 of the 1 min values. The latitude map plot was generated with the packages rnaturalearth [74] and sf [75]. Further modifications to the figures were made using Inkscape (v1.3).

The photoperiod lengths recordings (figure 1b) were collected from: Open meteo > historical weather over 4 years, from 2019 to 2023 (https://open-meteo.com/en/docs/historical-weather-api↗).

The statistical analysis was performed in R. In general, all data were tested for normal distribution using Shapiro’s normality test and for homogeneous variance using the Levene test. Based on the results of these tests, different approaches were taken, as described in the image captions. To compare the sleep during the light phase, the morning and evening peaks between strains, a Kruskal–Wallis test was performed followed by pairwise comparison using the Wilcoxon rank sum test. The Lomb–Scargle periodograms were generated using the rethomics framework [73]. The cycle analyses were performed by JTK_CYCLE [76] using the R package DiscoRhythm (v1.14.0) [77]. The single ROI intensities were averaged per brain and the analysis was run on the averaged brain values.

The evening peak quantification was carried out as described in Vaze et al. [11]. Briefly, the activity of single flies was averaged through 5 LD days, and the data were duplicated to obtain a double plot. The activity was then smoothed by local regression with a 0.02 span and finding the maximum values of the data with a rolling filter over 11 data points. The peaks were then found at the intersection points between the curve generated and the original activity data.

Raw data and data analysis are available in the electronic supplementary material [78].