Introduction

Circadian clocks align physiological and behavioural rhythms with environmental cycles. This enhances fitness by enabling anticipation of daily and seasonal changes (Jabbur et al., 2024; Sharma, 2003). These biological timekeepers integrate multiple cues, like light and temperature, which vary predictably across seasons and latitudes. The duration of light (photoperiod), high temperature (thermoperiod), and the lag of temperature cycles relative to light, all vary seasonally. This is seen across latitudes, albeit with milder seasonal variation in the tropics (Ahrens & Henson, 2018). Since circadian clocks rely on both light and temperature cycles, it is crucial to understand their joint effects on circadian systems (Yoshii et al., 2010). However, most laboratory-based studies have focused on either one of these time cues in isolation.

Drosophila melanogaster has been instrumental in elucidating principles of circadian functioning (Dubowy & Sehgal, 2017). Most of the research has focused on light as a time cue (Helfrich-Förster, 2020) while only a few studies have focused on synchronization of circadian rhythms by temperature cycles, combined light and temperature cycles or natural conditions (Busza et al., 2007; Menegazzi et al., 2013; Yoshii et al., 2009). Existing information regarding the role of fruit fly circadian neuronal and molecular components in regulating activity patterns in response to different environmental cues makes it a useful model for studying the integration of seasonal cues (Allada & Chung, 2010; Hermann-Luibl & Helfrich-Förster, 2015). Its locomotor activity pattern is known to be modulated by both light and temperature (Dubruille & Emery, 2008) and comprises distinct morning and evening activity bouts. Each of these bouts are regulated by morning and evening neuronal subsets (M and E cells) within the Drosophila circadian pacemaker network (Grima et al., 2004; Stoleru et al., 2004). It has been demonstrated that evening activity and E cells track the temperature cycle with high fidelity when temperature rise is phase-lagged with respect to light cycle (Miyasako et al., 2007; Yoshii et al., 2010). This is not the case for morning activity or M cells which remain phase locked to lights-ON. The hypothesis that evening activity exhibits greater temperature sensitivity or temperature tracking remains to be rigorously tested since most prior studies have utilized only one phase relationship between light and temperature cycles. This results in a limited understanding of how dynamic prioritization of time cues, ‘zeitgebers’, by the circadian system can occur (Oda & Friesen, 2011). In contrast, systematic analysis of multiple phase relationships can unravel differential sensitivity of the morning and evening activity and their corresponding circadian neurons. This may underlie the encoding of seasonal cues by the circadian network, enabling optimally timed activity.

In this study, we utilized a range of phase delays of temperature cycles relative to light and systematically assayed activity patterns of flies to gain a nuanced understanding of how the circadian rhythm responds to complex zeitgeber cycles (time cues influencing the state of the circadian system). We also conducted a similar examination of activity patterns of laboratory-selected Drosophila melanogaster “chronotype” populations (artificially selected for divergent timing of fly emergence from pupal case for over 350 generations). This was aimed at assessing how the response of activity patterns to complex zeitgeber cycles evolves upon selection for timing of a circadian behaviour. Moreover, the overlap between the model system’s overt rhythm and underlying neuronal substrates allowed us to make inferences and develop hypotheses regarding the coupling and differential zeitgeber sensitivity between the M and E cells of the circadian neuronal network.

The coupling and connectivity across the circadian network is one of the means by which seasonal information like photoperiod or temperature is encoded into the circadian clock (Li et al., 2024; Schlichting et al., 2016, 2019; Stoleru et al., 2007). However, alternative splicing of the components of the transcription-translation feedback loop (TTFL) forms an intracellular mechanism for encoding seasonal cues. Alternative splicing of circadian genes is widespread across taxa and plays a role in maintaining robust circadian rhythms and their synchronization to time cues (Bartok et al., 2013; Shakhmantsir & Sehgal, 2019). Hence, understanding the temporal expression pattern and the role of alternative splicing of circadian genes under varying environmental conditions in flies can inform not only how a given species synchronizes to zeitgeber cycles but also help generate hypotheses which could be tested across taxa. In D. melanogaster, period, timeless as well as Clock are known to undergo temperature-dependent alternative splicing (Cai et al., 2024; Majercak et al., 1999; Martin Anduaga et al., 2019; Shakhmantsir et al., 2018). TIMELESS plays a key role in integration of both light and temperature via its interaction with the cell autonomous photoreceptor CRYPTOCHROME and temperature-dependent mRNA splicing, respectively (Dubruille & Emery, 2008; Shakhmantsir & Sehgal, 2019). A qualitative examination of expression profiles of its splice variants under temperature cycles with constant darkness versus constant light as well as across different natural cycles suggest that light and temperature cycles may have combined effects on temperature-dependent timeless splicing (Montelli et al., 2015; Shakhmantsir et al., 2018). While timeless splicing under light cycles is well studied, effects of temperature cycle and combined light-temperature cycle remain unclear (Martin Anduaga et al., 2019; Shakhmantsir et al., 2018). To address these gaps, we investigated the expression profile of timeless splice variants and its splicing regulator Psi under more complex environmental cycles.

Splice sites of circadian genes, like period, are known to have natural variations which are correlated with thermal adaptation, indicating that selection can shape circadian gene splicing and temperature responsiveness (Low et al., 2008, 2012). Our laboratory-selected divergent chronotype fly populations are known to exhibit differences in temperature responsiveness of their emergence as well as activity rhythms (Abhilash et al., 2019, 2020; Nikhil et al., 2014; Vaze et al., 2012). Thus, it provides an apt system to test if selection for timing of behaviour/ temporal niche can lead to divergence in timeless splicing and differences in circadian plasticity.

In this study, we show that morning and evening activity bouts of flies exhibit differential temperature sensitivity. Moreover, such differences depend on the phase relationship between light and temperature cycles. We further demonstrate that this differential temperature sensitivity evolves upon selection for timing of emergence. We find that the presence of light cycle along with temperature cycle and the phase relationship between the two can significantly alter temperature-dependent timeless splicing. We also demonstrate for the first time that the transcript of timeless splicing regulator, Psi, is rhythmic under specific environmental conditions. Furthermore, we find that selection lines with divergent emergence timing exhibit differences in the expression profiles of timeless splice variants. Taken together, our studies examining circadian activity rhythm under a range of environmental cycles and linking splicing profiles of a circadian gene to behavioural plasticity, advances our understanding of how circadian systems adapt to complex environmental variability.

Results

Temperature sensitivity of evening activity is uncovered by out-of-sync environmental cycles in Drosophila melanogaster

To test the hypothesis that morning and evening activity bouts differ in temperature sensitivity and to understand zeitgeber prioritization under complex environmental cycles, we conducted experiments on four wild-type Drosophila melanogaster populations (control_1-4). First, flies were entrained to synchronized light (light: dark, LD) and temperature cycles (thermophase: cryophase, henceforth, TC) with lights-ON coinciding with temperature rise (LDTC in-sync). Subsequently, they were exposed to out-of-sync light and temperature cycles (LDTC out-of-sync). Six such experiments, ranging from 2 hr to 12 hr out-of-sync phase relationship between LD and TC, were imposed (Figure 1A). Comparison of activity profiles between LDTC in-sync and the 6 hr out-of-sync condition (4^th row of the schematic in Fig 1A) shows a pronounced phase shift in evening activity but not morning activity (Figure 1B). Similar activity profiles for all the phase relationships between LD and TC are shown in Figure S1A1-6.

Figure 1.

The evening activity phase was delayed with the delay in TC relative to LD, whereas morning activity remained phase locked to lights-ON. We quantified these changes by comparing mean phase shifts and sum of squared differences (SSD) between synchronized and out-of-sync LDTC conditions (for details see Materials and Methods). Both metrics revealed that the degree of activity waveform change differed between morning and evening bouts and also depended on the phase difference between light and temperature (Figure 1C&D). Values for change in the mean phase of activity during the morning window are positive for 8, 10, and 12 hours of delay between the temperature and light cycles (Figure S1A4-6). It is likely that this is due to high nocturnal activity and startle activity during the morning window around the high to low temperature transition (marked by grey arrows). The negative phase shift in evening activity (Δφ_CoM) confirmed its tracking of delayed temperature cycles (Figure 1D). Despite generally following temperature, evening activity remained aligned with light during extremely large phase delays (10–12 hours), possibly due to the highly unnatural phase relationships. These results confirm that morning and evening locomotor activities differ in temperature sensitivity, with this differential effect is contingent on the phase relationship between light and temperature zeitgebers.

Having gained a nuanced understanding of the circadian behavioural responses to complex zeitgeber cycles, we next sought to characterise the effect of environmental regimes with cyclic light and temperature on circadian gene splicing, a well-known mechanism for integrating temperature information into the molecular circadian clock. To this end, we profiled the expression of total timeless and all its known splice variants in one of the wild-type fly populations under three environmental conditions: (1) temperature cycles only (TC 12:12) in constant darkness, (2) synchronized light and temperature cycles (LDTC in-sync, lights-ON aligned with thermophase onset), and (3) phase-shifted light and temperature cycles (LDTC out-of-sync, thermophase onset delayed by 6 hours relative to lights-ON). We also temporally profiled Psi, a timeless splicing regulator implicated in activity phasing under temperature cycles (Foley et al., 2019).

Light: dark cycles modulate temperature-driven timeless splicing

We analysed the temporal expression profile of total timeless using primers targeting exonic regions common to all variants. Timeless expression was rhythmic under TC 12:12 and LDTC in-sync but not under LDTC out-of-sync (Figure 2A1-3). Comparing TC 12:12 and LDTC in-sync revealed that the presence of a light cycle along with a temperature cycle significantly altered the timeless expression profile (Table S1A), consolidating expression into a sharp peak at ZT15 with increased relative amplitude (Figure 2A4, S2A). Expression profiles also differed significantly between LDTC in-sync and out-of-sync regimes, underlining the role of zeitgeber phase relationship in shaping expression and rhythmicity (Figure 2A5, Table S1B).

Figure 2.

We next assessed temperature-induced timeless splice variants. The cold-induced variant tim-cold was rhythmic under all three regimes (Figure 2B1-3), with expression significantly altered by light presence and showing higher amplitude under LDTC in-sync compared to TC 12:12 (Figure 2B4, S2B, Table S1C). Tim-cold’s peak phase shifted under LDTC out-of-sync relative to in-sync, tracking cryophase onset, with significant differences in its expression profiles between these regimes (Figure 2B5, Table S1D). The other cold-induced variant, tim-short and cold, was rhythmic under TC 12:12 and LDTC in-sync but not under LDTC out-of-sync (Figure 2C1-3). It peaked sharply at ZT15 under TC 12:12 but under LDTC in-sync, it peaked at ZT19 with a gradual rise from ZT11. Thus, upon comparing its expression profiles under TC and LDTC in-sync, a significant regime * time point interaction and a difference in amplitude were detected (Figure 2C4, S2C, Table S1E). Its pattern was differed between LDTC in-sync and out-of-sync, suggesting that its expression profile and rhythmicity depended on light-temperature phase relationships (Figure 2C5, Table S1F). The warm temperature-induced variant, tim-medium, was rhythmic across all three regimes (Figure 2D1-3). Comparison between TC 12:12 and LDTC in-sync revealed that tim-medium expression was significantly altered upon the presence of an LD cycle along with TC (Figure 2D4, Table S1G). It showed a more gradual rise and fall under LDTC in-sync compared to the steep rise-and-fall pattern under TC 12:12 (Figure 2D4). Between LDTC in-sync and out-of-sync, the phase relationship significantly modulated its expression profile, with peak levels occurring at ZT19 under out-of-sync conditions (Figure 2D5, Table S1H).

Rhythmic expression of timeless splicing regulator, Psi

PSI, a splicing regulator involved in timeless splicing, displayed rhythmic mRNA expression under TC 12:12 and LDTC in-sync but lacked statistical rhythmicity under LDTC out-of-sync (Figure 2E1-3). To our knowledge, this is the first demonstration that Psi exhibits environmental regime dependent rhythmicity, further linking timeless splicing regulation to environmental cues. While visual differences in its expression profiles across the three environmental regimes may be apparent (Figure 2E4&5), we did not detect any statistically significant differences (Table S1I&J).

Altogether our observations thus far demonstrate that the circadian system in Drosophila melanogaster exhibits a differential temperature sensitivity between morning and evening locomotor activity, with evening activity more closely tracking temperature cycle shifts while morning activity remains predominantly light-locked. At the molecular level, we find that the expression and rhythmicity of temperature-induced timeless splice variants and their regulator Psi are fundamentally shaped not only by ambient temperature, as previously shown, but also by light and temperature cycles as well as by the phase relationship between these environmental cues.

To examine how evolution of different circadian phases may influence the integration of light and temperature cues, we extended our behavioural and molecular analyses to laboratory-selected Drosophila populations that have been divergent phasing of their emergence rhythms. Given that it has been previously established that the circadian clocks of these populations have indeed been diverged (Abhilash et al., 2020; Kumar et al., 2007; Nikhil et al., 2016), assaying their locomotor activity patterns under a range of environmental conditions along with timeless splicing dynamics in these selection lines allowed us to examine how adaptive changes in the circadian system impact temperature sensitivity of the evening locomotor activity and the molecular mechanisms underlying environmental signal integration. This comparison provides critical insights into the evolutionary plasticity of circadian behaviour and its molecular regulation under complex environmental cycles.

Evolution of temperature sensitivity: Late-selected flies exhibit enhanced evening activity phase shifts

Late-selected flies show the greatest delay in evening activity, followed by control flies, with early-selected flies showing the least delay (Figure 3A), while morning activity remains consistent across groups. This is especially clear in the phasing of activity drop after lights-OFF under the out-of-sync condition (indicated by grey arrows in Figure 3A). This trend among the selection lines can be observed across different phase lags between TC and LD (Figure S1A-C). We quantified timing (mean phase change) and activity pattern (SSD) changes during morning and evening periods between synchronized and each of the out-of-sync cycles. Morning activity responses were similar across the selection lines (Figures 3B, 3C), but evening activity differed distinctly (Figures 3D, 3E).

Figure 3.

Mixed model ANOVA indicated significant effects of selection and TC-LD phase relationship on the change in evening activity between LDTC in-sync and out-of-sync environmental cycles (Table S2C&D), with a significant interaction indicating differential responsiveness among lines (Table S2D). Morning activity also showed interaction effects, likely due to differences in startle responses, especially when temperature changed before the morning activity bout (marked by grey arrow in Figure S1A4-6, Table S2A&B). It is clear from the mean phase change that evening activity of late flies tracks delayed temperature cycles to a greater extent than that of early or control flies (Figure 3E).

These findings highlight that artificial selection for timing of emergence results in distinct behavioural adjustments to environmental cues, particularly in the evening activity phase. This underscores the value of these selection lines as powerful models to investigate the correlated evolution underlying flexibility and adaptation of circadian behaviour to changing environmental conditions upon temporal niche segregation. A previous study described global differences in activity rhythms among the selection lines under TC 12:12 (Abhilash et al., 2020). Our results along with such previously established differences in temperature sensitivity of activity rhythms, make these selection lines an apt system to study correlated evolution of temperature-induced circadian gene splicing.

Impact of long-term selection on circadian regulation of timeless splicing

We assayed total timeless, its splice variants, and the splicing regulator Psi in one block, i.e., early₄, late₄, and control₄ under three regimes: TC 12:12, LDTC in-sync, and LDTC out-of-sync to test whether selection drove correlated evolution in timeless splicing. Two comparisons were made: (1) differences among lines within each regime per transcript, and (2) changes in their expression profiles across regimes.

Total timeless was rhythmic under TC 12:12 in all lines, with early₄ showing significantly higher expression at ZT07 and ZT11 and a broader, flatter pattern in control₄ and late₄ lines (Figure 4A1). This resulted in higher relative amplitude and advanced mean phase in early₄ (Figure 4A4). Under LDTC in-sync, timeless remained rhythmic in all three lines. Early₄ had elevated expression at ZT11, late₄ had significantly lower expression at ZT15, indicating line-specific transcriptional adjustments (Figure 4A2). Under LDTC out-of-sync, rhythmicity was lost in all lines, likely due to environmental desynchrony and inherently high biological variability in our populations (Figure 4A3).

Figure 4.

Comparing TC 12:12 and LDTC in-sync showed that combined light and temperature cycles delay and consolidate timeless expression (Figures B1-3). Early₄ maintained high amplitude across both, while control₄ and late₄ showed amplitude expansion under LDTC, reflecting clock plasticity (Figure 4B4). Comparing LDTC in-sync and out-of-sync revealed amplitude reduction and rhythmicity loss across lines, particularly drastic in late₄ flies (Figure 4C1-3). All statistical comparisons are listed in Table S3.

Tim-cold showed robust rhythmicity across all regimes in early₄, control₄, and late₄ (Figure 5A1-3). Under TC 12:12, early₄ exhibited reduced amplitude, while late₄ maintained higher tim-cold levels for longer during the cryophase. These differences in their profiles are reflected in mean phase trends and significantly higher expression in late₄ flies during cryophase (Figure 5A4). This correlates with previously established differences in activity phasing and temperature sensitivity among the selection lines (Abhilash et al., 2020). Under LDTC out-of-sync, peak expression was uniformly delayed to ZT19 across lines, matching delayed cryophase onset (Figure 5B1-3). Notably, at ZT11, tim-cold levels increased in control₄ and especially late₄ under out-of-sync compared to in-sync LDTC conditions, while early remained stable, revealing time-of-day dependent differential regulatory responses to environmental desynchrony (Figure 5B4).

Figure 5.

The cold-induced tim-sc variant was rhythmic under TC 12:12 and LDTC in-sync but not under LDTC out-of-sync for all three lines (Figure 5C1-3). Under TC 12:12, early₄ showed a sharp post-peak decline at ZT19, reflecting selection line-dependent transcript modulation (Figure 5C1). Loss of tim-sc rhythmicity may also be contributed by increased biological replicate variability and a missing time point. All statistical comparisons for cold-induced timeless transcripts are in Table S4, with additional profile comparisons in Figure S3.

Tim-medium was rhythmic across all three lines and regimes (Figure 6A1-3). Upon comparing its expression profile under LDTC in-sync to TC 12:12 for the selection lines (Figure 6B1-3), we found that control₄ and late₄ shift in a similar manner with there being an overall delay in their profiles but significant change (Table S5D). However, early₄ cross 50% of their peak expression levels by ZT03 and ZT07 under TC but not under LDTC in-sync regime while the peak and decline in early₄ tim-medium levels occur at the same time between the two regimes (Figure 6B2), resulting in its profile being altered significantly between these two regimes (Table S5D).

Figure 6.

Psi was rhythmic in control₄ under TC 12:12, as described before, but not in early₄ or late₄ (Figure 6C1). It was, however, rhythmic for early₄, control₄ as well as late₄ flies under LDTC in-sync regimes (Figure 6C2) thus, demonstrating distinct splicing regulator dynamics linked to selection between TC versus LDTC in-sync. Psi was rhythmic only in early₄ under LDTC out-of-sync regime and not in control or late (Figure 6C3), where higher biological replicate variation may have contributed to the loss of detectable rhythmicity in addition to environmental asynchrony. All statistical comparisons conducted for tim-medium and Psi transcripts have been listed in Table S5 and all plots for profile comparisons not discussed here are a part of Figure S4.

Overall, these results detail rhythmic expression patterns and selection-dependent modulation of timeless splice variants and the splicing factor Psi across different environmental cycles. The observed variation among biological replicates and occasional arrhythmicity under LDTC out-of-sync regime in our populations could be stemming from their large standing genetic variation but appears to be compounded by the transcript being assayed as well as environmental asynchrony. This underscores both the biological complexity and sensitivity of circadian molecular regulation to environmental and genetic factors, highlighting important considerations for future studies of clock plasticity and adaptation. Altogether, our results provide strong evidence for correlated evolution of splicing dynamics of a circadian gene upon selection for timing of emergence rhythm.

Discussion

Using an experimental protocol comprising incremental phase delays of temperature cycles relative to light, we reveal that evening activity exhibits greater temperature sensitivity compared to morning activity. Through our long-term laboratory selection approach, we find that this property of evening activity can undergo correlated evolution upon selection for divergent timing of emergence. We also demonstrate that the expression pattern of temperature-induced timeless splice variants is modulated by light:dark cycles. Furthermore, their expression profiles diverge upon selection for timing of emergence. We also find correlations between altered timeless splicing and diverged temperature sensitivity of activity rhythms among the selection lines.

Differential temperature sensitivity of morning and evening activity

The widespread distribution of crepuscular activity across diverse lineages suggests it is an evolutionary adaptation balancing predation avoidance, light availability, and temperature regulation (Bennie et al., 2014; Carroll & Harvey-Carroll, 2023; Kirmse, 2024; Kirse et al., 2025; Rojas et al., 1999; Wong & Didham, 2024). The unparalleled advantages of the fly model system make it valuable to study how complex environmental cycles regulate crepuscular activity at molecular, neuronal and behavioural levels.

Previous studies demonstrated that in D. melanogaster, the molecular clock phase in evening neurons and CRY-negative neurons tracks the temperature cycle (Miyasako et al., 2007; Yoshii et al., 2009, 2010). However, such studies often used a single-phase displacement which can limit insights into the full range of circadian responses (Oda & Friesen, 2011). While Harper et al., 2016, assayed activity under a range of different phase relationships between light and temperature cycles, the temperature rise was advanced with respect to lights-ON leading to startle responses and complicating interpretation. In nature temperature rise always lags dawn, making it crucial to study the response of circadian system to phase lagged temperature cycles compared to lights-ON. Moreover, the study by Harper et al., 2016 was aimed at assessing the role of cryptochrome in the phasing of evening activity. Thus, the hypothesis of the two activity bouts having differential temperature sensitivity was still under-examined.

Our experimental design, incorporating incremental phase delays of the temperature cycle relative to light, enabled rigorous testing of this hypothesis under naturally plausible phase relationships between the two zeitgebers. We found that morning and evening activity bouts exhibit differential temperature sensitivity with evening activity phase tracking the delay in temperature cycle to a greater extent (Figure 1 This difference between the activity bouts was greatest under a 6-hour phase delay between the two zeitgebers, with evening activity re-aligning with the light cycle under extreme or unnatural phase lags of 10 or 12 hours (Figure 1C&D). This phase dependence underscores the importance of using incremental phase differences instead of single-phase displacement in circadian studies to avoid misleading or incomplete interpretations regarding zeitgeber integration. These behavioural dynamics under a range of phase delays between light and temperature prompt further exploration of how specific neuronal oscillators within the fly circadian network and their interactions may be mediating this phase-dependent response.

In Drosophila melanogaster, morning activity is regulated by PDF^+ve LNvs, while PDF^-ve small LNv and LNds govern evening activity (Allada & Chung, 2010). Integrating this cellular framework with our data and prior modelling of emergence rhythms under similar LDTC conditions (Oda & Friesen, 2011), we propose a model describing coupling and temperature sensitivity of these oscillators.

The evening oscillator group is heterogeneous (Shafer et al., 2022), containing light-sensitive neurons expressing the cell-autonomous photoreceptor CRYOTOCHROME (CRY) as well as receiving light input from morning cells and temperature-sensitive neurons lacking CRY (Beckwith & Ceriani, 2015). LNds also receive input from temperature-sensitive dorsal neurons (Li et al., 2024). Our findings support that evening and morning activity differ in temperature responsiveness, with the evening oscillator showing greater phase-dependent sensitivity. Previous work shows that fly emergence circadian rhythm too tracks the shifts in temperature albeit within a specific range before realigning to light, mirroring our observed evening activity patterns (Pittendrigh and Bruce, 1959). Another study describes the numerical simulations of limit-cycle oscillator models, which capture the main features of the emergence patterns observed under such a range of phase relationships between light and temperature cycles (Oda & Friesen, 2011). They show that such dynamics can arise from mutually coupled oscillators without asymmetry in the coupling between them or in the sensitivity to their respective dominant zeitgebers. Considering the evening oscillator’s heterogeneity and these insights, we propose a model in which distinct neuronal subpopulations integrate light and temperature inputs to regulate evening activity timing in fruit flies (Figure 7). This allows evening activity to track temperature within a defined phase range but revert to light cues under extreme phase differences, enabling seasonal adjustment through modulation of oscillator phase relationships. Beyond modelling the observed activity data, certain aspects of this model can be experimentally tested by using cell-specific knockout or knockdown of synchronizing factors such as the neuropeptide pigment dispersing factor (PDF) or of photoreceptor cryptochrome, and assessing their effects on activity under LD-TC conditions.

Figure 7.

timeless splicing as a node for integration of seasonal cues

In addition to neuronal circuit mechanisms, molecular pathways such as temperature-dependent alternative splicing of the timeless gene offer a complementary route for integrating seasonal cues into the circadian system. It is well established that timeless undergoes temperature-dependent splicing: cooler temperatures upregulate tim-cold (tim-c) and tim-short and cold (tim-sc), while warmer temperatures increase tim-medium (tim-M) (Martin Anduaga et al., 2019; Shakhmantsir et al., 2018). These prior studies have primarily examined timeless splicing under LD cycles with constant cool or warm temperatures but without precise quantitative comparisons across environmental conditions with cyclic temperature and cyclic light.

Our study systematically profiled total timeless and its splice variants under three distinct environmental conditions (i) TC – temperature cycle under constant darkness, (ii) LDTC in-sync – synchronous light and temperature cycle (iii) LDTC out-of-sync – temperature rise was delayed by 6 hours with respect to lights-ON. We found that timeless expression was rhythmic under TC and LDTC in-sync (Figure 2A1&2). However, the amplitude and consolidation were higher under LDTC in-sync, reflecting synergistic effects of synchronous environmental cues (Figure 2A4, Figure S2A). In contrast, LDTC out-of-sync regime delayed and dampened timeless expression highlighting the critical effect of phase alignment between zeitgeber cycles on timeless rhythmicity (Figure 2A5).

The cold induced isoform, tim-cold, showed robust rhythmicity across all regimes (Figure 2B1-3). While tim-cold expression profile was significantly altered by the presence of light cycle and its phase relationship with temperature cycle (Figure 2B4&B5, Table S1C&D), its peak expression was always phase locked to cryophase onset. However, the other cold-induced transcript, tim-sc showed more complex dynamics, it peaked with tim-cold only under TC but had a gradual rise and a delayed peak under LDTC in-sync (Figure 2C1&2). This suggests that its peak expression is not universally locked to temperature drop. Its expression pattern under TC and LDTC in-sync suggest that tim-sc may act as a seasonal detector, integrating both light and temperature information (Figure 2C4, Table S1E). On the other hand, tim-cold may relay immediate temperature drop information to the circadian clock. Our hypothesis is supported by the presence of tim-sc in temperate D. simulans and D. melanogaster and its absence in tropical D. yakuba, while tim-cold is expressed in all three species (Martin Anduaga et al., 2019). Additionally, tim-cold likely binds AGO1-miRNA complexes, whereas TIM-SC protein regulates PERIOD stability, highlighting functional differences between these cold-induced isoforms (Martin Anduaga et al., 2019).

The warmth-induced transcript tim-medium is under strong post-transcriptional regulation by the RNAi machinery, and the protein of this transcript is not functional (Martin Anduaga et al., 2019; Shakhmantsir et al., 2018). Therefore, it is thought to contribute to delays in TIM protein profile under warmer temperatures by effectively reducing the levels of the canonical tim-long variant available for translation (Shakhmantsir et al., 2018). In our study, the rhythmic expression of tim-medium across all environmental conditions highlights its robust regulation (Figure 2D1-3). Furthermore, the significant alterations in its expression profile caused by the presence of a light cycle and the phase relationship between light and temperature underscore the complex interplay of these cues in likely modulating TIM protein levels via the expression dynamics of tim-medium (Figure 2D4&5, Table S1G&H).

Critical to timeless alternative splicing are the splicing factors PSI and PRP4, which likely regulate timeless splicing in opposite directions: prp4 knockdown upregulates tim-medium while Psi knockdown downregulates it but increases the cold-induced isoforms under LD cycles with constant temperature (Foley et al., 2019; Shakhmantsir et al., 2018). Due to Psi’s role in activity phasing under TC we conducted time point profiling of this splicing regulator under three environmental conditions (Foley et al., 2019). We show for the first time that Psi expression is rhythmic under certain environmental conditions (Figure 2E1&2).

Previous work suggests a positive correlation between Psi and tim-medium levels and a negative correlation between Psi and levels of tim-cold and tim-sc (Foley et al., 2019). We observe that this correlation is most clearly upheld under temperature cycles wherein tim-medium and Psi peak in-phase while the cold-induced transcripts start rising when Psi falls (Figure 8A1&2). Under LDTC in-sync conditions this relationship is weaker, even though Psi is rhythmic, potentially due to light-modulated factors influencing timeless splicing (Figure 8B1&2). This is in line with Psi’s established role in regulating activity phasing under TC 12:12 but not LD 12:12 (Foley et al., 2019). This is also supported by the fact that while tim-medium and tim-cold are rhythmic under LD 12:12 (Shakhmantsir et al., 2018), Psi is not (datasets from Kuintzle et al., 2017; Rodriguez et al., 2013). Assuming this to be true across genetic backgrounds and sexes and combined with our similar findings for these three transcripts under LDTC out-of-sync (Figure 2B3, D3&E3), we speculate that Psi rhythmicity may not be essential for tim-medium or tim-cold rhythmicity especially under conditions wherein light cycles are present along with temperature cycles (Figure 8C1&2). Our study opens avenues for future experiments manipulating PSI expression under varying light-temperature regimes to dissect its precise regulatory interactions. We hypothesize that flies with Psi knocked down in the clock neurons should exhibit a less pronounced shift of the evening activity under the range LDTC out-of-sync conditions for which activity is assayed in our study. On the other hand, its overexpression should cause larger delays in response to delayed temperature cycles due to the increased levels of tim-medium translating into delay in TIM protein accumulation.

Figure 8.

Our detailed characterisation and comparison of timeless splicing across environmental regimes places timeless splicing as a pivotal node in seasonal cue integration since it is influenced not only by ambient temperature but also by light cycle and its phase relationships with temperature cycle. Our study underscores a complex interplay of environmental signals in regulating activity patterns as well as circadian gene splicing.

Another key aspect of seasonal variation is photoperiod; since period splicing is sensitive to photoperiodic conditions, timeless splicing may be as well (Majercak et al., 2004). Previous studies have characterised naturally occurring variation in splice site strength in the circadian gene period along with clinal studies of its SNPs and its correlation with seasonal adaptation highlights the evolutionary importance of circadian gene splicing (Low et al., 2008, 2012). Such clinal studies for splice site strength within timeless in D. melanogaster could provide insights into their role in adaptation to varying environmental cycles. While clinal studies are indispensable for studying trait evolution, including that of circadian rhythms, they are largely correlational and ill-equipped to isolate the influence of specific environmental factors (Abhilash & Sharma, 2016). On the other hand, laboratory selection experiments, while lacking ecological complexity, address these limitations by focusing on a specific selection pressure. These provide advantages such as no assumptions regarding ancestry and ancestral state of the trait of interest, replicate populations, stronger causal inference regarding the effect of selection as well as controlled population genetics (Abhilash & Sharma, 2016). Therefore, through our approach of using laboratory populations selected for divergent timing of fly emergence, we could investigate how circadian system adaptions influence evening activity temperature sensitivity and timeless splicing and gain insights into evolution of seasonal cue integration under a temporal niche segregation-like scenario.

Correlated evolution of temperature sensitivity of evening activity and circadian gene splicing upon selection for timing of emergence

Late-selected flies diverged in the temperature sensitivity of their evening activity, with their evening activity phase tracking the shift in temperature cycle to a greater extent than that of early and control flies (Figure 3). We hypothesise that under out-of-sync LDTC conditions, PERIOD and TIMELESS expression in CRY-negative neurons will shift most in late populations, mirroring evolved plasticity of evening activity in response to temperature cues. The divergence observed in late flies taken together with our findings for control (wild-type populations) indicate that differential temperature sensitivity of morning and evening activity bouts can undergo correlated evolution upon selection for timing of circadian emergence. Such behavioural dynamics in our selection lines make it an apt system to explore how our circadian organisation model proposed for control (Figure 7) may have been altered upon selection. We propose a model for the circadian organisation in late which can explain the behavioural pattern observed (Figure 9). We propose testable simplistic parsimonious models considering our own observations, known connectivity across the circadian network, and the findings from a previous study (Oda & Friesen, 2011).

Figure 9.

Building on the behavioural divergence observed, we investigated whether timeless splicing plays a role in this adaptation by examining molecular correlates that might underlie differences in temperature sensitivity among the selection lines. Under TC 12:12, timeless expression is advanced in early₄ and delayed in late₄ relative to control flies (Figure 4A1&A4), mirroring their activity rhythm phases (Abhilash et al., 2020). Mathematical models propose that a larger amplitude oscillator is predicted to have reduced resetting capacity, whereas smaller amplitude oscillators reset more readily (Barrett & Takahashi, 1997; Winfree, 2001). Overexpression of clock components reduces oscillator amplitude and increases resetting capacity (Busza et al., 2007), making the clock more responsive to environmental perturbations (Pittendrigh et al., 1991). Early₄ having the highest amplitude of timeless expression (Figure 4A4) aligns with them taking the longest to re-entrain activity rhythms after jetlagged temperature cycles, while late flies re-entrain fastest (Abhilash et al., 2020). Since late₄ and control₄ flies do not differ in timeless amplitude, other factors likely contribute to the rapid resetting under temperature cycle jetlag shown by late flies compared to control, perhaps reflecting previously described divergence in their circadian organisation (Figure 7 vs. Figure 9).

The most striking differences in timeless splicing among the selection lines were observed for tim-cold and the warm temperature-induced tim-medium transcripts, both rhythmic across all three lines and environmental conditions (Figure 5A1-3, Figure 6A1-3). Tim-cold transcript levels were significantly elevated in late₄ flies during cryophase under TC 12:12 (figure 5A4), consistent with their heightened temperature responsiveness (Abhilash et al., 2020). Comparing TC and LDTC in-sync regimes suggests that LD modulates tim-cold expression differentially across selection lines (Figure 5B1-3). The phase relationship between light and temperature cycles (LDTC out-of-sync vs. in-sync) also impacts tim-cold, with late₄ flies showing the largest change at ZT11 (Figure 5B4). However, linking these molecular shifts to behaviour mandates mutant or transgenic studies targeting tim-cold. While tim-medium expression profile was significantly altered when LD was present along with TC in early₄, this was not the case for control₄ and late₄ flies (Table S5D, Figure 6B1-3). Thus, not only is tim-medium expression profile altered in response to LD in addition to TC in wildtype flies (control₄) as mentioned before, but this response diverged upon selection for timing of emergence rhythm. Under LDTC out-of-sync, tim-medium accumulation was delayed, peaking at ZT19 in early₄ and control₄ lines (Figure S4A1&2), whereas late₄ flies maintained high expression from ZT11 to ZT19, broadening the profile (Figure S4A3). Advanced tim-medium rise is known to delay TIM protein accumulation and thus would delay activity (Shakhmantsir et al., 2018). Accordingly, tim-medium expression under LDTC out-of-sync aligns with the delayed evening activity observed in late₄ flies (Figure 3A). Taken together, these findings strongly indicate that timeless splicing dynamics evolve under selection for divergent timing of fly emergence. However, our molecular conclusions are currently limited to one population per selection line, unlike our behavioural data drawn from four populations for each selection line.

Our study further reveals distinct rhythmicity of Psi, a timeless splicing regulator, among early₄, control₄, and late₄ flies, highlighting selection’s impact on this splicing regulator. Psi rhythmicity is disrupted in early₄ and late₄ flies under temperature cycles but maintained in control₄ (Figure 6C1). Under synchronized LDTC, Psi is rhythmic across all lines (Figure 6C2). Under desynchronized LDTC, rhythmicity is largely confined to early₄ flies, with high variance suggesting complex gene-environment interactions shaping its expression (Figure 6C3). Given PSI’s role in neural splicing and developmental processes (Labourier et al., 2002), it is plausible that Psi has been a direct target of selection for emergence timing.

This raises a compelling question: What functional roles do timeless splice variants serve in emergence rhythms, which are regulated by the same central clock as activity rhythms (Qiu & Hardin, 1996)? Emergence rhythms of flies deficient in specific splice variants would illuminate how seasonal signals and environmental temperatures influence holometabolic development via the circadian clock. Our selection lines too, provide an ideal system to investigate this question, as they exhibit divergent temperature sensitivity in emergence rhythms as well (Abhilash et al., 2019; Nikhil et al., 2014). Since introgressing transgenic constructs into these backgrounds may be challenging, RNA sequencing of brains from freshly emerged flies—particularly under TC or LDTC out-of-sync regimes—could reveal variant-specific expression relevant to differential temperature sensitivity of emergence rhythms. Multiple studies conducted across generations on these laboratory-selected fly populations, including the current one, consistently demonstrate that selection for emergence timing using light as the sole cue leads to divergence in temperature sensitivity of both emergence and activity rhythms.(Abhilash et al., 2019, 2020; Nikhil et al., 2014; Vaze et al., 2012). These studies also highlight that laboratory selection has uncovered novel and unexpected genetic correlations, such as between emergence timing and temperature responsiveness of circadian rhythms. Through the current study we reinforce the notion that selection for behavioural phasing fundamentally reshapes the neuronal network and molecular circadian machinery underlying these rhythms.

In summary, our findings demonstrate that the temperature sensitivity of evening activity depends on the degree of misalignment between the light and temperature cycles; most likely due to the heterogeneous nature of evening oscillator neurons. This highlights how circadian systems can dynamically prioritize different environmental signals, a principle relevant to understanding how organisms cope with shifting environmental features or seasonal changes. Our detailed profiling of timeless splice variants under different light and temperature regimes revealed that in addition to ambient temperature, timeless splicing is modulated by the presence of light cycles along with temperature cycles and the phase relationship between the two. This shows that alternative splicing of core clock genes is a key mechanism for integrating multiple time cues, a process which may likely be conserved across taxa. Using the laboratory-selection lines, we reveal that selection for emergence timing leads to correlated evolution of temperature sensitivity of evening activity and timeless splicing. Our findings show how differential temperature response of the activity components can provide a means to fine tune behaviour under complex environmental cycles. Additionally, we highlight how timeless splicing is dynamically regulated by these cycles and how selection shapes both behaviour and circadian gene splicing. Altogether these findings emphasize the circadian clock’s complexity and adaptability.

Materials and Methods

Fly strains

Wildtype fly populations and laboratory-selected divergent chronotype fly populations

Four large and outbred populations were used to derive four of each early, control and late selection lines. Therefore, the set of selection lines comprises 12 populations – early_1-4, control_1-4 and late_1-4. The early, control and late selection lines with the same subscript indicate that they have been derived from the same ancestral population (referred to as ‘blocks’). In contrast, different subscripts indicate independent genetic substructures. The early and late selection lines have been subjected to selection for phase of entrainment of adult fly emergence for more than 400 generations now while the control populations remain unselected, comprising of flies emerging throughout the day and represent wildtype flies (detailed maintenance protocol is described in Kumar et al., 2007). It is important to note that the wildtype control or the selected early and late are reared in large numbers (∼1200 flies in roughly 1:1 sex ratio for each population) and therefore, have high standing genetic variation.

One generation of common rearing wherein no selection pressure is applied and flies emerging throughout the day are collected to form the parental generation from which egg collection is conducted for the experimental flies. This minimises the maternal and non-genetic inheritance effect on the measured trait (Bonduriansky & Day, 2009).

All the blocks of early, control and late have been used to test for the differential responses of the activity rhythm under environmental regimes with light and temperature cycles. Below is the table listing the number of generations of selection these populations had undergone at the time when each of these activity assays were performed.

Locomotor activity recording

3-5 days old virgin male flies from each of the 12 populations (4 blocks of each of the three selection lines post common rearing, early_1-4, control_1-4 and late_1-4) were housed in locomotor tubes with corn food medium and placed in Drosophila Activity Monitors (DAM system by Trikinetics) to measure their activity counts. Assays were carried out in Sanyo incubators (MIR-154, Sanyo, Tokyo, Japan). Light intensities were adjusted to 70 lux using 298 neutral-density Lee filter and a LI-COR light meter. The maintenance of these populations is carried out under 70 lux, and light intensities lower than 100 lux minimise startles. The temperature cycle used had a warm/thermophase temperature of 28 °C and a cool/cryophase temperature of 19 °C. For each of the six experiments in which TC lagged LD, the activity was first recorded under conditions wherein LD and TC were in-sync and were transferred to one of the LDTC out-of-sync regimes after 4 days. This allowed for calculations to compare the waveform under the LDTC out-of-sync regime to that of LDTC in-sync regime from the same generation/set of flies for each experiment.

Figure 10.

Activity profile analyses

The activity profiles for each individual under the in-sync and out-of-sync regimes were calculated by averaging the daily profiles under each of the regimes. When a set of flies were transferred from the in-sync regime to an out-of-sync regime, the first couple of cycles were eliminated from the calculation for average profiles since all flies attained stable phases after day 2 under the new LDTC regime. The activity counts in average profile for each individual under both in-sync and out-of-sync regime were normalised by the total average activity from that individual to obtain activity level per 15 minutes. These two profiles for each individual were then used to assess changes in the activity waveform between the two regimes.

Assessing phase lability and changes in activity waveform under LDTC out-of-sync conditions compared to LDTC in-sync conditions

The activity profiles were divided into a 12-hour morning window (6 hours before and after lights-ON) and an evening window (6 hours before and after lights-OFF). The following calculations were conducted for the activity within each of these time windows without filtering the data or elimination of startle peaks to conduct unbiased assessment of the response of the morning and evening activity components.

Sum of squared differences (SSD)

The activity level in a particular 15-minute bin under the LDTC out-of-sync regime was subtracted from the activity level in the same 15-minute bin under the LDTC in-sync regime of the same individual. This difference within a particular bin was then squared. These squared values were then summed up over the 12 hour window to give an SSD value for an individual. SSD values were calculated for the morning and the evening windows. The larger the value, the greater the change in the activity waveform within that time window. The SSD values from all the individuals from a population were then averaged to yield the mean SSD for that population. Thus, resulting in 12 SSD values (early_1-4, control_1-4 and late_1-4) for morning as well as the evening window.

Change in the mean phase of activity (ΔφCoM)

The mean phase of activity was calculated within the morning and evening time windows from the normalised average activity profiles for the in-sync regime for an individual (Batschelet, 1981). The same calculations were carried out for the out-of-sync regime as well. Change in the mean phase of activity during the morning window was calculated by subtracting the value of the mean phase under out-of-sync regime from that of in-sync regime for a given individual. The same calculation was conducted for the evening window as well. The change in mean phase of activity within each time window from all the individuals from a population were then averaged to yield the average change in phase for that population. Thus, resulting in 12 Δφ_CoM values (early_1-4, control_1-4 and late_1-4) for morning as well as the evening window.

Statistical analysis

For both SSD and change in mean phase within each of the time windows for all six experiments, the twelve population values were used to conduct mixed model ANOVA with selection (early, control and late) and environmental regime as the fixed factors and blocks (1-4) as the random factor to assess differential responses of the two activity components (morning and evening) among the selection lines.

Time point sample collection for assaying timeless and Psi transcripts

3-5 days old virgin male flies from block 4 of early, control, and late selection lines were collected and housed in glass vials with corn-medium food. These vials were then placed in an incubator with a set environmental regime (MIR-154, Sanyo, Tokyo, Japan). The environmental regimes used in this study, along with the number of generations of selection before the common rearing for experimental flies, are tabulated below.

Figure 11.

RNA extraction and quantitative real-time polymerase chain reaction

The flies were decapitated and at least 40-45 heads were used per sample to extract RNA. RNA was extracted using TRIzol/TRI reagent (Sigma-Alrich) as per the manufacturer’s instructions from the fly heads. 1 µg of this RNA was treated with DNase to remove any genomic DNA contamination. This served as the template for first-strand complementary DNA (cDNA) synthesis using the Maxima first-strand cDNA synthesis kit. Approximately 10-20 ng of cDNA was used in each qPCR reaction. The qPCR reactions were conducted in a Bio-rad CFX96 RT-PCR detection system. The qPCR mixture was subjected to 95 °C for 3 min, followed by 40 cycles of 95 °C for 10 s, 55 °C / 46 °C for 10 s, and 72 °C for 30 s, followed by a melting curve analysis to confirm that only one PCR product/peak is detected.

Analyses of transcript expression profiles

The cycle threshold values obtained for each of the experimental transcripts were normalised by the expression of a housekeeping gene (rp49/RpL32) to obtain the relative expression of the experimental transcript for a biological replicate, for that time point, using the comparative Ct method (as per example 5 in Schmittgen & Livak, 2008). Each reaction was set up in duplicates and an average value of the two was used for the calculation of relative expression. Average expression and error across the three biological replicates have been plotted. The expression profiles have also been plotted with the expression level at each point scaled by the peak expression value and expressed as a percentage of the peak expression. This scaling was conducted for each biological replicate separately, thus facilitating both the visualisation of variation in expression levels across time points among biological replicates and comparison of profiles across regimes/experiments. These peak normalised expression profiles were used to quantify the mean phase of expression as well as the amplitude of expression (maximum – minimum expression level). RAIN analysis was used to detect rhythmicity in each dataset across a cycle (Thaben & Westermark, 2014). Two-way ANOVA tests with selection and time point as the factors were used to detect differences among the selection lines under a particular environmental regime. Amplitude and peak normalised expression patterns were used to test for the effect of environmental regime on the expression profile of a transcript in the wild-type control population followed by testing for differential effect of the environmental regime on the expression profiles among the selection lines (early, control and late).

GraphPad Prism v.8 was used for conducting all the one-way and two-way ANOVA tests. Mixed model ANOVA with three factors as well as repeated measures ANOVA were implemented in Statistica v.7.

Note: ‘expression pattern’, where used in this study refers only to the temporal expression pattern of the transcript across the time points under a given environmental regime.

All data was plotted using GraphPad Prism v.8

Supplementary information

Figure S1.

Figure S2.

Figure S3.

Figure S4

Table S1.

Table S2.

Table S3.

Table S4.

Table S5.

Table S6.

Data availability

All datasets will be deposited on Dryad.

Acknowledgements

We thank Abhilash Lakshman for experimental design in the early stages of the project and Shephali Dansana for carrying out initial pilot experiments which led to this project. We also thank research intern Prakruthi for carrying out RNA isolation for some of the samples. We would like to acknowledge financial support through intramural funding from the Jawaharlal Nehru Centre for Advanced Scientific Research (JNCASR). We also thank Rajanna, Muniraju and Samuel for their technical assistance.

Additional information

Author Contributions

Pragya Niraj Sharma., conceptualization, methodology, investigation, formal analysis, and writing-original draft and Sheeba Vasu., conceptualization, methodology, supervision, writing-review and editing, and funding acquisition.

Funding

Jawaharlal Nehru Centre for Advanced Scientific Research (JNCASR) (PhD fellowship)

• Pragya Niraj Sharma

Jawaharlal Nehru Centre for Advanced Scientific Research (JNCASR) (Intramural consumable grant)

• Vasu Sheeba