Splicing of a core circadian clock gene regulates seasonal adaptations by a winter gating mechanism

Abstract

Organisms adjust their physiology and behavior in response to seasonal changes. The current working model indicates that the circadian clock is involved in this process, but the molecular mechanisms mediating the integration of seasonal cues are still unclear. Notably, the circadian neuropeptide pigment-dispersing factor (PDF), an output of the circadian clock, has been shown to alter its expression and activity in response to seasonal changes to facilitate seasonal adaptations in insects. Here, we show that the alternative splicing of a circadian clock gene, timeless (tim), regulates the seasonal responses through PDF in Drosophila melanogaster. We found that tim-sc, the predominant isoform in winter, is regulated by photoperiod, while the canonical tim-l isoform is not. In addition, we demonstrated that tim-sc is used to maintain physiology and behavior in a “winter lock” state by modulating PDF. Our results support a role of isoform-specific characteristics in providing circadian clock components with the ability to modulate seasonal physiology.

The circadian clock integrates photoperiodic and temperature cues by RNA splicing to promote seasonal plasticity.

INTRODUCTION

Animals, plants, and even bacteria can anticipate and adapt to seasonal changes in environmental conditions (1–4). These adaptations are diverse, involving a wide array of physiological, behavioral, and phenotypical changes such as developmental arrest (i.e., diapause) (5), reproduction (3), hibernation (6), and migration (7–9), among others. Seasonal changes allow organisms to endure long periods of extreme temperature conditions and limited food availability; thus, engaging in these changes needs to be precisely timed. Several phenological cues are used to inform seasonal environmental changes, with photoperiod (i.e., length of day) and temperature being the most prominent ones in temperate climates (10). Little is known about how organisms integrate these cues to coordinate and establish seasonal programs in physiology and behavior, particularly in animals.

In the mid-1930s, Bünning (11, 12) proposed that the circadian clock, the mechanism that allows for maintaining daily rhythms in physiology and behavior, is indispensable for photoperiodism in plants. Since then, studies have shown this to be the case in plants (13) and in several vertebrate (14) and invertebrate (15) animals. In mammals and birds, the suprachiasmatic nucleus, the core circadian structure in the brain, integrates photoperiodic cues and regulates the rhythmic nocturnal release of melatonin from the pineal gland (1, 16). This initiates a hormonal cascade involving the thyroid-stimulating hormone in the pars tuberalis of the pituitary gland and the regulation of tanycytes in the third ventricle of the hypothalamus, which triggers seasonal adaptations (17–24). How circadian outputs, such as melatonin, are regulated in the seasonal context is still unclear, and exploring these questions in seasonal vertebrate models is still challenging given their genetic intractability. Thus, simple insect models have proven useful in this context.

In several insects, including Drosophila melanogaster, temperature has a massive influence on seasonal adaptations in addition to photoperiodic cues (25, 26). Both photoperiod and temperature are integrated by the circadian clock neuronal network (CCNN) to enable decisions on entering, maintaining, and exiting seasonal adaptations (27). These neurons relay seasonal cues to neurosecretory cells in the pars lateralis and pars intercerebralis (PI), which in turn trigger seasonal adaptations by secreting hormone-like peptides (28). The circadian neuropeptide pigment-dispersing factor (PDF) has been reported to play an important role in seasonality across insects (28–35). PDF is a circadian output released by a group of circadian neurons within the CCNN named ventral lateral neurons (LNvs), and it coordinates the CCNN rhythmicity under constant darkness in D. melanogaster (36–39). Under cold conditions, PDF levels drop, allowing the accumulation of the protein EYES ABSENT (EYA) (40). PDF relays both photoperiod and temperature cues to cells in the PI to regulate seasonal adaptations, a feature that seems to be conserved in other insects, including mosquitoes, bugs, and flies (32, 41–45). Since PDF is a circadian output, it is conceivable that the integration of seasonal cues occurs upstream in the circadian clock.

At the molecular level, the circadian clock functions as a transcriptional-translational feedback loop in which positive elements promote the transcription of negative elements that, once translated, block their own transcription in an exquisitely regulated process that takes around 24 hours. In D. melanogaster, the positive elements are clock (clk) and cycle (cyc), and the negative elements are timeless (tim) and period (per) (46–49). Genetic manipulations by mutating or silencing these genes render a diverse array of phenotypes in several insects, highlighting the involvement of clock genes in seasonal adaptations [reviewed in (29)]. Nonetheless, how the circadian clock integrates seasonal cues is still unclear. In this regard, alternative splicing (AS) has arisen as a potential mechanism, given that multiple clock genes exhibit temperature-sensitive AS (50–52).

AS is the process by which different isoforms are generated from the same preprocessed RNA. This mechanism mediates a wide array of regulations, including the generation of protein variants with altered structure and function, or controlling transcripts’ stability and degradation [reviewed in (53)]. AS is key to providing phenotypic plasticity and adaptation to the environment. Hence, a role of AS in seasonal adaptations has been supported by evidence in several models, including plants and animals (54–58). For instance, changes in the AS landscape have been observed in different tissues of hibernating brown bears (59). In seasonal morphs of butterfly Bicyclus anynana, AS affects a subset of genes without changes in gene expression level (60). In the Japanese quail, one particular isoform of Eya3 containing exon 7 is expressed in the pars tuberalis under long days (61), and AS affects the seasonal expression of the blue-light activated cryptochrome in the retina of European robins (62).

Seasonal cues have been shown to modulate AS of core clock genes in several species. In D. melanogaster, AS of the 3′ terminal intron of per is observed under cold temperatures, in the laboratory, and in natural conditions, which contributes to differential accumulation of per to regulate seasonal locomotor adaptations (50, 51, 56). In addition, AS events have been shown to directly affect the structure and function of the core clock proteins. For instance, we recently described that clk undergoes temperature-dependent AS, which generates a CLK isoform with a four–amino acid deletion adjacent to the CLK DNA binding domain, affecting CLK function in regulating circadian output and PDF response to temperature (63). Another clock protein, TIM, undergoes even more pronounced AS-dependent changes in sequence/structure (Fig. 1A) (64). Under cold conditions, two isoforms are generated, tim-cold (tim-c) and tim-short-and-cold (tim-sc), encoding proteins differing in 33 and 507 amino acids, respectively, from the canonical TIM-long (TIM-L) isoform (65–68). Expression of tim-c and tim-sc almost completely replaces tim-l in winter-like cold conditions, making these isoforms particularly interesting in the context of seasonal adaptations. Supporting this notion, tim-null mutants have impaired reproductive arrest response in response to seasonal changes (67). In addition, tim displays naturally occurring, clinally distributed alleles, ls and s, whose protein products exhibit differential light sensitivity (69–71). All these data point to the potential role for tim in photoperiod and temperature cue integration.

Fig. 1. Messenger RNAs encoding TIMELESS isoforms are differentially regulated by photoperiod and temperature.

(A) Schematic of the TIMELESS protein. PER binding domains 1 and 2 (PER BD1 and PER BD2), nuclear localization signal (NLS), and CLD are shown [reviewed in (64)]. Expression of (B) tim-l, (C) tim-c, and (D) tim-sc in summer-like, LP [16-hour (h) light and 8-hour dark cycles; 16:8 LD] at 25°C (red lines) and winter-like, SP (8 hours light and 16 hours dark cycles; 8:16 LD) at 10°C (blue lines) in CS flies, bearing the ls allele. Horizontal bars above the graphs depict light (white) and dark (gray) periods over the 24-hour cycle. Expression of (E) tim-l, (F) tim-c, and (G) tim-sc under the same conditions is shown for the w¹¹¹⁸ flies bearing the s allele. The P values for rhythmicity computed by CircaCompare are shown on the graphs. The P values and statistics for MESOR, phase, and amplitude are presented in the results. (H) Experimental setting for changing photoperiod at 25°C. Expression profiles of (I) tim-l and (J) tim-sc are shown for flies entrained under these conditions. (K) Experimental setting for changing photoperiod at 10°C. Expression profiles of (L) tim-l and (M) tim-sc are shown for flies entrained under these conditions. In (H) to (M), 16:8 LD, 12:12 LD, and 8:16 LD are represented by green, black, and pink lines, respectively. Vertical dotted lines in the graphs represent the phase of peak expression according to CircaCompare. For (B) to (G), N = 3 replicates, with ~100 fly heads per time point. For (I) to (M), N = 3 to 6 replicates, with ~100 fly heads each per time point.

In this study, we investigated the role of tim AS in the integration of photoperiod and temperature by the circadian clock to modulate seasonality, focusing primarily on winter adaptations. We first describe how photoperiod and temperature shape TIM expression. In particular, we show that tim-sc is highly sensitive to changes in photoperiod as compared with tim-l, resulting in a significant phase advance in tim-sc daily peak expression. TIM-C and TIM-SC are both expressed at the protein level in winter conditions, but TIM-SC is the predominant isoform with regard to expression level. The phase advance observed at the mRNA level is maintained in TIM-SC protein under cold conditions despite the generally lower amplitude in TIM-SC daily protein rhythm. We found that TIM-SC is predominantly nuclear, affecting circadian clock function and outputs. By performing functional analyses on transgenic flies expressing either tim-l or tim-sc, we showed that the molecular rearrangement of the core clock incorporating TIM-SC in winter-like conditions allows for a gating mechanism for overwintering. We observed that flies expressing tim-sc exhibit low PDF levels, similar to what was previously observed in winter-like conditions (40), and this coincides with winter-associated daily locomotor profile and reproductive dormancy. Notably, tim-sc flies did not completely phenocopy flies maintained in winter conditions with regard to regulation within the molecular clock, suggesting that other mechanisms are involved. In summary, we propose a model in which AS-dependent molecular plasticity of clock components is critical to seasonality.

RESULTS

Temperature-dependent tim isoforms are differentially regulated by photoperiod

Timeless mRNA splicing is regulated by temperature (65–68). At 25°C, the dominant isoform is the canonical tim-l, whereas under cold conditions, two isoforms are produced: tim-c and tim-sc (Fig. 1A). It is unknown whether photoperiod, as another major seasonal cue, has any effect on the expression of these isoforms at the mRNA level. To answer this question, we determined the levels of tim-l, tim-c, and tim-sc in heads of wild-type flies entrained under summer-like conditions [long photoperiod (LP), 16:8 light dark (LD) at 25°C] or winter-like conditions [short photoperiod (SP), 8:16 LD at 10°C]. We used both Canton-S (CS) flies and w¹¹¹⁸ flies as they carry either the ls or s allele, respectively (69). This allowed us to have all possible combinations of timeless generated by alternative translation start sites and AS (Fig. 1A). Overall, all timeless isoforms are rhythmically expressed in summer and winter, albeit at different expression levels, amplitudes, and with different peak phases of expression (Fig. 1, B to G; P value for rhythmicity computed by CircaCompare shown on each graph). In CS, tim-l expression is higher in summer as compared to winter [Fig. 1B; midline estimating statistic of rhythms (MESOR) 0.53 versus 0.21, P = 2.4 × 10⁻¹¹], and the phase of its peak expression changes from Zeitgeber time 16.9 (ZT16.9) in summer to ZT11.5 in winter (∆ = 5.4 hours, P = 9.6 × 10⁻⁴). In contrast, the expression of tim-c (Fig. 1C) and tim-sc (Fig. 1D) increases in winter conditions as compared to summer (MESOR tim-c: 0.65 versus 0.50, P = 3.13 × 10⁻¹⁶; MESOR tim-sc: 0.65 versus 0.01, P = 3.8 × 10⁻¹⁸). For tim-c, the peak expression in winter is at ZT11.2, while in summer it is at ZT15.9 (∆ = 4.7 hours, P = 3.8 × 10⁻⁵). For tim-sc, in contrast, the peak phase shows a ∆ value of around 6.3 hours, from ZT16.8 in summer to ZT10.5 in winter, but this difference did not reach significance (P = 0.72). Similar results were observed in w¹¹¹⁸ flies, with the main difference being that the MESOR in tim-l was not significantly different in summer versus winter (MESOR 0.49 versus 0.40, P = 0.10) but with a difference in the amplitude of summer versus winter oscillation of tim-l (0.39 versus 0.13, P = 1.8 × 10⁻³; Fig. 1, E to G). While not significant, possibly because of the lower amplitude of its rhythms, the observed ∆ in peak tim-sc expression was bigger than the one observed in tim-l or tim-c across seasonal conditions in both CS and w¹¹¹⁸, suggesting a differential regulation of tim-sc compared to the other isoforms.

To further investigate whether photoperiod alone has any effect on the peak phase and expression level of tim-sc, newly eclosed w¹¹¹⁸ flies were entrained either in LP (16:8 LD), equinox (12:12 LD), or SP (8:16 LD) at constant 25°C (Fig. 1H). Under these conditions, peak phase of tim-l expression is unchanged as per visual inspection and raw peak values. Algorithmic estimation by CircaCompare displays marginal advances of around 1 hour from LP to equinox, while an additional hour advance is observed from equinox to SP but is not statistically significant (LP: ZT16.9, equinox: ZT15.3, SP: ZT14.0, LP vs Equinox P.adj = 4.5 × 10⁻², Equinox vs SP P.adj = 0.12, and LP vs SP P.adj = 5.2 × 10⁻³; Fig. 1I). The peak expression of tim-sc shows an advancement from LP to equinox, and, in contrast to tim-l, tim-sc shows a sharp advancement from equinox to SP (LP: ZT16.6, equinox: ZT15.1, SP: ZT12.5, LP vs Equinox P.adj = 3.1 × 10⁻², Equinox vs SP P.adj = 2.1 × 10⁻⁵, and LP vs SP P.adj = 2.6 × 10⁻⁷; Fig. 1J). Furthermore, when using the absolute values of expression, we also did not observe significant differences in MESOR (LP: 5.07, equinox: 4.61, SP: 4.35, LP vs Equi P.adj = 0.512, Equi vs SP P.adj = 0.68, and LP vs SP P.adj = 0.512) or amplitude of tim-sc expression (LP: 3.39, equinox: 3.41, SP: 3.54, LP vs Equinox P.adj = 0.98, Equinox vs SP P.adj = 0.39, and LP vs SP P.adj = 0.98).

Temperature alone has been shown to affect the phase of expression of tim and per at 12:12 LD (50). We therefore tested the effect of cold temperature in tim isoform expression by maintaining the flies at either equinox or SP at 10°C (Fig. 1K). Comparing these two conditions, tim-l shows a slight advancement in its peak expression at SP compared to equinox but the difference is not significant (equinox: ZT12.36 and SP: ZT10.63, P = 0.43; Fig. 1L). In contrast, a significant 6-hour advancement is observed in tim-sc in SP compared to equinox at 10°C (equinox: ZT13.85 and SP: ZT7.82, P = 0.0012; Fig. 1M). This suggests that photoperiod can regulate the daily profiles of tim isoforms, particularly the peak phase of expression. While the winter-expressed isoform tim-sc is extremely sensitive to photoperiod, tim-l is largely insensitive to this phenological cue.

To determine whether photoperiodic regulation of tim isoforms translates to changes at the protein level, we detected temperature-dependent TIM isoforms by Western blot under simulated summer (16:8 LD at 25°C) versus winter (8:16 LD at 10°C) conditions. Our existing antibody against TIM (rat polyclonal anti-TIM, R5839, RRID: AB_2782953) was generated using the C-terminal segment of the canonical long isoform, which is missing in TIM-C and TIM-SC isoforms (Fig. 1A and fig. S1, A and B). To overcome this issue, we developed a new antibody that recognizes the shared N terminus of TIM (fig. S1, A and B). We showed that this new antibody (RRID: AB_3713152) recognizes TIM-L at 25°C and TIM-SC at 10°C in fly head lysates (fig. S1, C and D). This was supported by the absence of signal in tim⁰ flies and staining of FLAG-tagged TIM-SC with α-TIM and α-FLAG (fig. S1E). Closer inspection revealed that TIM-SC is observed as two bands at 10°C, possibly due to posttranslational modifications. In addition, two weak bands migrating slightly faster than TIM-L appear primarily at evening time points at 10°C (fig. S1F). The size difference and cold-inducibility are consistent with the notion that these are the TIM-C isoforms. Similar to TIM-SC, posttranslational modifications could have led to the production of multiple TIM-C bands.

Using this new reagent, we detected TIM isoforms in head lysates from CS and w¹¹¹⁸ flies entrained in summer (LP at 25°C) or winter (SP at 10°C) conditions (Fig. 2, A and B). As expected, TIM-L is the predominant isoform in summer, both in CS and w¹¹¹⁸. In winter conditions, TIM-L is replaced by TIM-C and TIM-SC. In CS, TIM-L peaks at around ZT20 [Rhythmicity Analysis Incorporating Nonparametric methods (RAIN) P = 3.75 × 10⁻¹⁵; Fig. 2C], while in w¹¹¹⁸, its peak expression occurs at ZT24 (RAIN P = 3.86 × 10⁻¹⁰; Fig. 2F) in simulated summer. The roughly 4-hour delay between mRNA and protein peak for TIM-L is expected, given previously reported posttranslational regulation of TIM [reviewed in (72)]. Similarly, there was an approximately 4- to 5-hour delay between mRNA and protein peak for TIM-C, which is highly expressed in winter. Whereas tim-c mRNA peaks ~ZT11.2, TIM-C protein peaks in the middle of the night at ZT16 in both CS and w¹¹¹⁸ (RAIN CS P = 3.16 × 10⁻⁶, w¹¹¹⁸ P = 1.01 × 10⁻⁶; Fig. 2, D and G, respectively). This suggests that similar posttranslational mechanisms may be at work to regulate TIM-C daily abundance. Last, TIM-SC is not detectable in simulated summer but very abundant in simulated winter. Although TIM-SC was deemed to exhibit daily rhythmicity in both CS and w¹¹¹⁸ (RAIN P = 6.94 × 10⁻⁴ and P = 2.1 × 10⁻²; Fig. 2, E and H, respectively), the amplitudes of these rhythms were much weaker than observed for TIM-L. Nonetheless, we observed a clear ~8-hour phase difference in the expression of TIM-L in summer and TIM-SC in winter in both CS and w¹¹¹ flies. Overall, these data suggest that the expression levels of the different tim isoforms are regulated not only by temperature but also differentially regulated by photoperiod. In particular, the peak mRNA expression of tim-sc exhibits a significant phase advance, which, together with posttranslational regulation, results in an ~8-hour phase advance at the protein level.

Fig. 2. TIMELESS expression differs across different seasonal conditions.

Representative Western blots detecting TIM using an antibody generated from an N-terminal antigen. Fly head lysates from (A) CS flies, bearing the ls allele, or (B) w¹¹¹⁸ flies, bearing the s allele, were used. Flies were entrained under summer-like, LP (16:8 LD) at 25°C conditions or winter-like, SP (8:16 LD) at 10°C conditions, and heads were collected at the indicated time points. Arrows indicated the different isoforms based on band size. Quantification of TIM-L, TIM-C, and TIM-SC normalized expression against HSP70, from CS (C to E, respectively) and w¹¹¹⁸ flies (F to H, respectively). Red lines represent expression at LP 25°C and blue lines represent SP 10°C. Statistical P values for rhythmicity obtained using RAIN are depicted in graphs (C) to (H). The dotted line represents the phase of peak expression of the proteins according to RAIN. For (C) to (H), N = 6 biological replicates, with ~100 fly heads per time point. Horizontal bars above the graphs depict light (white) and dark (gray) periods in summer versus winter conditions.

TIMELESS is predominantly nuclear in winter conditions

We showed that TIM-SC is the main isoform expressed in whole heads under winter-like conditions. However, whether there are seasonal differences at the cellular level within the CCNN is unknown. In particular, TIM-SC is a truncated protein that is missing the well-characterized cytoplasmic localization domain (CLD) in canonical TIM-L, suggesting that the regulation of its subcellular localization could be compromised. Thus, we sought to determine the expression of TIM across the CCNN under different seasonal conditions, with the understanding that TIM-L is dominantly expressed in simulated summer while TIM-SC is preferentially expressed in simulated winter. We first validated our antibody for use in immunofluorescence in intact tissue. As in Western blot, we showed that new TIM antibody specifically detects TIM in whole-brain preparations of wild-type flies entrained at 25°C, while staining is absent in tim⁰¹ flies (fig. S2). We then detected TIM in brains from CS and w¹¹¹⁸ flies entrained under summer conditions (LP 25°C) or winter conditions (SP 10°C). All the major clock clusters were detected, including the dorsal neurons 1 (DN1), the DN3, the dorsal lateral neurons (LNds), and the LNvs (fig. S3). Relatively low levels of TIM were observed at ZT3 in summer or winter conditions (Fig. 3, A and B, respectively). In contrast, higher expression of TIM is observed at ZT15 in both seasonal conditions (Fig. 3, C and D). Simulated winter appeared to alter the subcellular localization of TIM. In summer, TIM is predominantly cytosolic at ZT15 (Fig. 3C, inserts), while in winter, TIM is almost exclusively nuclear (Fig. 3D, inserts).

Fig. 3. TIMELESS is predominantly nuclear in winter-like conditions.

Representative images of TIM expression in brains collected at ZT3 (A and B) or ZT15 (C and D) from flies entrained in summer-like, LP (16:8 LD) at 25°C conditions or winter-like, SP (8:16 LD) at 10°C conditions. The inserts show the main clock neuronal clusters: DN, dorsal neurons; LNvs, ventral lateral neurons; LNds, dorsal lateral neurons. Representative images and quantification of coimmunofluorescence against PDF (magenta) and TIM (green). Samples were collected at ZT18, ZT20, ZT22, and ZT24 from (E and F) CS flies and w¹¹¹⁸ (G and H) entrained in LP 25°C (red boxes) or SP 10°C (blue boxes). Letters represent the significant differences between conditions (different letter: P < 0.05, same letter: P > 0.05). Two-way analysis of variance (ANOVA) with Holm-Šídák’s multiple comparisons test, three independent replicates, with n = 45 to 107 cells from >eight brains per time point, per condition. Scale bars, (A to D) 50 μm and (zoomed inserts) 25 μm. Scale bars, (E to G) 10 μm.

To quantitatively determine the subcellular localization of TIM in summer versus winter conditions, we focused on the LNvs cluster. There are two subtypes of LNvs; large-LNvs (l-LNvs) and small-LNvs (s-LNvs). Both types of LNvs serve as centers for controlling seasonal responses, differentially participating in the modulation of locomotor activity profile during long days and promoting reproductive dormancy in winter conditions, respectively (35, 40, 73). We used immunodetection of the PDF to identify the PDF⁺ LNvs and to delineate the cytosol and nucleus in brains from CS flies entrained in summer or winter conditions (Fig. 3E). PDF has been used as a cytosol marker in published reports as it remains in the cytoplasm across the day (74–76). TIM nuclear localization, while still following a daily pattern, is slightly phase-advanced in l-LNvs compared with s-LNvs (74). Thus, focusing on the l-LNvs would allow us to better capture temporal information than in the compressed s-LNvs nuclear localization schedule. We examined TIM localization in the l-LNvs at ZT18, ZT20, ZT22, and ZT24. Overall, TIM is more nuclear in winter than in summer [Fig. 3F; F(1,680) = 935.3, P < 0.0001]. Under summer conditions, TIM is predominantly cytosolic at ZT18 and progressively becomes more nuclear by ZT24 (0.305 versus 0.367, respectively; P < 0.0001). In winter conditions, TIM is already more nuclear at ZT18 as compared to ZT24 in summer (0.459 versus 0.367, P < 0.0001). Even with this, TIM nuclear localization increases at later times in winter condition (ZT18: 0.459 versus ZT24: 0.514, P < 0.0001). To explore whether TIM-SC nuclear localization is sensitive to tim N-terminal alleles, we conducted the same experiments in w¹¹¹⁸ flies expressing only s-tim. Similar results were found in brains from these flies (Fig. 3G). TIM nuclear localization increases with time in both summer and winter [Fig. 3H; F(3,459) = 38.21, P < 0.0001], and TIM predominantly nuclear localization is observed in winter as compared to summer [F(1, 459) = 229.0, P < 0.0001]. These data suggest that not only is TIM isoform expression changing across seasons but so is its subcellular localization.

TIM-SC expression regulates locomotor activity profile

We showed that the expression of TIM isoforms as well as their subcellular localization change across seasons in response to temperature and photoperiod. However, whether these dynamics are important for seasonal adaptations is unknown. Thus, we sought to determine the role of the main TIM isoforms in summer and winter, TIM-L and TIM-SC, respectively, in seasonal adaptations. To do this, we generated two transgenic fly lines expressing either TIM-L or TIM-SC cDNA under ~4.7 kb of the tim promoter (Fig. 4A). As expected, these flies, when used to rescue tim⁰ mutants, express only TIM-L or TIM-SC at 12:12 LD at 25°C (Fig. 4B). To describe the effect of the selective expression of these isoforms on circadian locomotor activity, flies were kept in 12:12 LD at 25°C for 4 days and then released into constant darkness (DD) for 7 days to assay free-running period. As expected, yw control flies show a daily morning and evening peak aligning with the light transitions in LD, which is sustained when flies are placed in DD (Fig. 4, C and G). Changes in activity at the light transitions are also observed in tim⁰¹ in LD, but no anticipation of these transitions is observed, and rhythms quickly disappear in DD, suggesting a startle response to light instead of rhythms in LD (Fig. 4, D and H).

Fig. 4. Transgenic rescue lines expressing specific timeless isoforms recapitulate seasonal changes in locomotor activity.

(A) Flies expressing only tim-l or tim-sc were generated by insertion of a transgenic construct bearing the tim-l or tim-sc coding DNA sequence under a ~4.7-kb of the timeless promoter in tim⁰¹ null-mutant flies. (B) Representative Western blot detecting TIM using an antibody generated from an N-terminal antigen in samples from tim-l or tim-sc rescue flies. Locomotor activity was monitored for 4 days in 12:12 LD at 25°C and then for 7 days in constant darkness (DD). Representative double plot actograms and average activity profiles in LD of yw control flies (C and G), tim⁰¹ null-mutant flies (D and H), tim-l rescue flies (E and I), and tim-sc rescue flies (F and J). Data for three independent experiments, with n = 92 to 95 flies per condition. Gray boxes on the graphs represent the dark period, while the light periods are represented as white boxes. The arrows in graphs (G to J) represent the peak of evening activity. (K) Quantification of the phase of the evening activity peak in yw (black box), tim-l (orange box), and tim-sc (purple box). Letters represent the significant (different letter: P < 0.05, same letter: P > 0.05) differences between conditions. Kruskal-Wallis test, followed by Dunn’s test. (L) Periodogram for yw (black line), tim null (pink line), tim-l (orange line), and tim-sc (purple line) extracted from the data collected in DD. Vertical dotted lines represent the free-running period for each genotype. The solid line represents the expected significance threshold (P < 0.05) across periods. Data for three independent experiments, with n = 81 to 87 flies per condition. (A) was created in BioRender. S.H. (2026) https://BioRender.com/x7ypk6k.

One of the main seasonal adaptations observed in Drosophila is the plasticity of the evening locomotor activity (40, 50, 77, 78). During cold days, flies advance their evening activity to potentially match the short days of winter and promote the chances of receiving midday heat. While on hot summer days, flies delay their evening activity to avoid the high heat. Introduction of tim-l rescues both daily activity in LD and sustained rhythmic locomotor activity in DD (Fig. 4, E and I, and Table 1). On the other hand, the introduction of tim-sc only recovers the evening peak of activity, while weakly preserving the rhythmicity of this peak on DD (Fig. 4, F and J, and Table 1). Quantification of the phase of the evening peak of activity shows a difference across genotypes [Fig. 4K; H(2) = 122.2, P < 0.0001]. In particular, there is a 2.5-hour advancement of the peak in tim-sc rescue flies as compared to yw control flies (8.753 versus 11.22, P < 0.0001). This advancement is of similar magnitude to the one observed in Pdf⁰¹ flies (fig. S4) (40, 79). A marginal, but significant, 1.28-hour advancement was observed in tim-l rescue flies as compared to the yw control (10.00 versus 11.22, P < 0.0001) while differing in 1.25 hours with the peak of tim-sc flies (10.00 versus 8.753, P < 0.0001).

Table 1. Daily activity profiles of Drosophila fly lines expressing specific timeless isoforms.

Genotype	Rhythmicity (Qp.act/Qp.sig)	% Rhythmic	Period (hours)	n
yw	1.10 ± 0.014	70.1	23.41 ± 0.07	87
yw; tim01	0.907 ± 0.005	1.2	-	84
yw; tim01; tim-l	1.057 ± 0.016	56.8	23.67 ± 0.25	81
yw; tim01; tim-sc	1.002 ± 0.011	33.3	24.91 ± 0.39	84

Last, the introduction of tim-l rescued circadian rhythmicity to the level of yw control flies (Fig. 4, C to F, and Table 1), with a period of around 24 hours (Fig. 4L, black and orange lines, and Table 1). Rhythmicity is partially rescued by the introduction of tim-sc, albeit with a longer free-running period (Fig. 4L, purple line, and Table 1). All in all, tim-sc rescue flies do not behave as tim⁰¹ or tim-l flies, suggesting that TIM-SC have an active function within the clock that differs from the canonical TIM-L isoform.

Tim-sc generates a winter lock to promote successful overwintering

We previously demonstrated that PDF levels in the s-LNvs dorsal terminals are a crucial regulator that prompts flies to overwinter (40). In summer, PDF levels are high, maintaining summer locomotor architecture and promoting reproduction. In winter, a reduction in PDF levels reshapes daily locomotor activity profiles by modulating the evening peak and triggers reproductive dormancy. Consistent with the idea that PDF expression can be differentially regulated by tim AS, tim-sc flies show an advancement in their evening peak of activity similar to Pdf-null mutants in 12:12 LD cycles at 25°C (Fig. 4J and fig. S4). To determine whether PDF expression is indeed altered in flies expressing tim-l versus tim-sc, we assayed the levels of PDF in the terminals of the s-LNvs, which are important for seasonal adaptations. As expected, PDF levels are overall reduced in tim-sc flies as compared to tim-l flies [Fig. 5, A and B; F(1,103) = 30.12, P < 0.0001]. In addition, PDF levels are rhythmic in tim-l flies while arrhythmic in tim-sc flies (Fig. 5B; RAIN P = 1.1 × 10⁻⁵ and P = 0.5, respectively).

Fig. 5. Timeless splicing acts as a seasonal lock to modulate PDF.

(A) Representative immunofluorescence of PDF in the terminals of the small s-LNvs in tim-l and tim-sc rescue lines. Flies were entrained in 12:12 LD cycles at 25°C for 3 days and collected at ZT12 and ZT24. (B) Quantification of PDF intensity on the s-LNvs in tim-l (orange line) and tim-sc (purple line) flies. Two-way ANOVA with Šídák’s multiple comparisons test, *P < 0.05, **P < 0.005, ***P < 0.0005, from two independent experiments with a total n = 8 to 10 brains per condition. The P value for rhythmicity for tim-l and tim-sc obtained using RAIN is shown in orange and purple letters, respectively. (C) Representative immunofluorescences of PDF in the s-LNvs terminals from flies entrained in summer-like, LP (16:8 LD) at 25°C, or winter-like, SP (8:16 LD) at 10°C, conditions. Samples were collected at ZT3 and ZT15. (D) Quantification of experiments in (C). Two-way ANOVA with Holm-Šídák’s multiple comparisons test. Letters represent the significant differences between conditions (different letter: P < 0.05, the same letter: P > 0.05). (E) Representative images of ovaries from tim-l and tim-sc rescue flies entrained in summer-like conditions (LP 25°C) or winter-like conditions (SP 10°C). (F) Dormancy, as the percentage of flies in a population in ovaries with <1 egg or immature egg, in summer (left) or winter (right) in tim-l (orange bars) and tim-sc (purple bars) flies. Unpaired t test, *P < 0.05, N = 3 replicates with ~20 flies each. ns, not significant.

Considering that PDF levels are low in tim-sc flies and high in tim-l flies, we hypothesized that tim isoform exchange from TIM-L in summer to TIM-SC in winter underlies the changes in PDF observed across seasons. To test this, we took tim-l and tim-sc flies and entrained them under summer (LP 25°C) or winter (SP 10°C) conditions and probed against PDF at ZT3 and ZT15. If tim isoform exchange is necessary for PDF-mediated seasonal adaptation, tim-l and tim-sc flies would be unresponsive to environmental conditions, and PDF would stay high or low, respectively, regardless. However, this is not what we observed. In summer, tim-l flies display high levels of PDF that are reduced when placing the flies in winter conditions [Fig. 5, C and D, orange boxes; F(3,126) = 28.90, P < 0.0001]. On the other hand, tim-sc flies have low levels of PDF both in summer and winter (Fig. 5, C and D, purple boxes). PDF levels do not appear to cycle in winter in either tim-l or tim-sc flies as no difference is observed between ZT3 and ZT15 (tim-l: 2.699 versus 2.769, P = 0.9991; tim-sc: 2.367 versus 2.402, P = 0.9991). This suggests that tim isoform exchange is not required for the reduction of PDF from summer to winter; rather, the presence of TIM-SC locks PDF levels in a winter state, while locking the isoforms also affects rhythmicity.

If the winter lock hypothesis is correct, placing tim-sc flies even under summer conditions will promote overwintering. To test this, we subjected tim-l and tim-sc flies to summer or winter conditions and examined ovary development as a readout of reproductive dormancy. Consistent with our hypothesis, tim-sc flies have higher levels of dormancy in summer as compared to tim-l flies [Fig. 5, E and F, left; 19% versus 6.7%; t(4) = 3.7, P = 0.0208]. As expected, both tim-l and tim-sc enter dormancy in winter-like conditions [Fig. 5, E and F, right; 96.33% versus 98.00%, t(4) = 0.61, P = 0.574].

Last, to test whether splicing plasticity is also required to alter locomotor activity profile in a similar fashion, we place tim-l and tim-sc flies in LP 25°C and SP 10°C (fig. S5). As in previous reports, yw control flies can adjust their evening activity to follow the lights-off transition under long, summer days. Notably, neither tim-l nor tim-sc flies are able to do so (fig. S5, A and B). In contrast, under winter-like conditions, both tim-l and tim-sc flies can match the light schedule as activity is concentrated at midday (fig. S5C). Also, tim-sc flies show significantly higher activity in winter as compared to tim-l or yw flies (fig. S5D). In summary, our data show that TIM-SC generates a winter lock, gating PDF in cold, short days to promote a winter program.

TIM-SC reshapes the molecular clockwork through PER activity

PDF is indirectly regulated by CLK (80, 81). Given the importance of TIM in the molecular clock and its function in regulating CLK activity, the molecular remodeling of the core clock resulting from the expression of TIM-SC in winter conditions may result in the observed reduction of PDF, which underlies the winter lock mechanism. To determine the effect of having TIM-SC instead of TIM-L in the molecular clockwork, we assayed the expression and function of TIM heterodimeric partner PERIOD in tim-l and tim-sc flies. We first assayed PER levels by immunoblotting in whole head extracts from tim-l and tim-sc flies (Fig. 6A). As expected, PER levels cycle across the day in tim-l flies, with a peak abundance at ZT16 (Fig. 6B; RAIN P = 5.32 × 10⁻⁴). PER abundance is significantly reduced in tim-sc flies as compared to tim-l flies [Fig. 6B; F(1,60) = 47.44, P < 0.0001]. In addition, PER daily expression is not rhythmic in tim-sc flies (RAIN P = 0.91).

Fig. 6. TIM-SC expression affects key core clock features.

(A) Representative Western blot detecting PERIOD in tim-l and tim-sc rescue lines entrained in 12:12 LD at 25°C. (B) Quantification of normalized PER expression in tim-l (orange line) and tim-sc (purple line) fly lines. HSP70 was used for normalization. The vertical dotted line represents the phase of the peak expression, and the P value represents the rhythmicity statistic according to RAIN. Two-way ANOVA with Šídák’s multiple comparisons test, N = 6 replicates, with ~100 flies per time point, per genotype. *P < 0.05, **P < 0.005, ***P < 0.001, ****P < 0.0001. (C) Representative coimmunofluorescence against PDF (magenta) and PER (green) in tim-l and tim-sc fly lines. Scale bar, 2 μm. (D) Samples were collected at ZT18, ZT20, and ZT22 in flies entrained in 12:12 LD cycles at 25°C. Two-way ANOVA with Holm-Šídák’s multiple comparisons test, N = 3 replicates, n = 23 to 50 cells from >eight brains per time point, per condition. Letters represent the significant differences between conditions (different letter: P < 0.05, the same letter: P > 0.05). (E) CLOCK occupancy on per promoter in heads of tim-l (orange bar) and tim-sc (purple bar) flies. Two-way ANOVA with Holm-Šídák’s multiple comparisons test, N = 4 replicates with ~500 flies per time point, per genotype. Letters represent the significant differences between conditions (different letter: P < 0.05, the same letter: P > 0.05). (F) per expression on heads from tim-l (orange dots) and tim-sc (purple dots) rescue lines. Two-way ANOVA with Holm-Šídák’s multiple comparisons test, N = 4 replicates with ~100 flies per time point, per genotype. *P < 0.05. The dotted line represents the phase of the peak expression, and the P value represents the rhythmicity statistic according to RAIN.

A possible hypothesis explaining the reduced PER levels is its altered nuclear localization, resulting in accelerated phosphorylation-dependent degradation. To evaluate subcellular localization of PER, we detected PER by immunofluorescence using a specific antibody (Fig. 6C and fig. S6). As in Fig. 3, we focused on the LNvs by costaining against PDF. While PER is more nuclear in tim-l flies, its distribution in tim-sc flies is sparse across the cytosol and the nucleus (Fig. 6C). Quantification of the nuclear fraction of total PER indeed shows a reduction in nuclear localization of the protein in tim-sc flies as compared with tim-l flies [Fig. 6D; F(1,234) = 45.41, P < 0.0001].

Considering that PER localization fails to reach high nuclear levels in tim-sc flies, we hypothesized that its downstream function of inhibiting CLOCK-dependent DNA binding and transcription may be affected. To address this, we conducted chromatin immunoprecipitation (ChIP), pulling down CLK in tim-l and tim-sc flies, followed by quantitative polymerase chain reaction (qPCR) against the perCRS promoter region (Fig. 6E) to detect CLK-DNA binding. Consistent with our hypothesis, CLK binding to the per promoter is significantly elevated in tim-sc flies compared to tim-l flies [F(1,27) = 25.78, P < 0.0001; MESOR P = 5.56 × 10⁻⁵]. Notably, CLK is binding rhythmically to the per promoter in both tim-l and tim-sc flies, regardless of the arrhythmic PER levels in the latter (P = 0.02 and P = 0.01, respectively). This is in agreement with the rhythmic expression of tim-c and tim-sc in simulated winter conditions (Fig. 1)

Last, considering that CLK binding to per promoter is higher in tim-sc–expressing flies, we hypothesized that this would result in increased per mRNA expression. However, this is not what we observed. Expression of per is higher only at ZT8 in tim-sc flies (0.667 versus 0.280, P = 0.0164), and it is significantly reduced at ZT16, the peak expression of per in tim-l flies (Fig. 6F; 0.801 versus 0.423, P = 0.0193).

Our data suggest that TIM-SC expression affects PER abundance and localization, affecting its role in removing CLK from the DNA. Despite CLK-DNA binding remaining rhythmic in tim-sc flies, the expression of per mRNA (CLK target) is not rhythmic. This suggests that a partial remodeling of the molecular clockwork may promote seasonal adaptations.

DISCUSSION

The circadian clock has been functionally linked to seasonal adaptations; however, the exact role of the clock in seasonal timing or to what extent the clock is required to establish, maintain, and exit these adaptations is still unclear. In this study, we demonstrate that the splicing of a core clock component, tim, is a crucial factor in maintaining winter adaptations in Drosophila. Our data suggest that the presence of TIM-SC in winter, gated by cold temperatures, alters the function of the molecular clock. TIM-SC is primarily present in the nucleus across the day, and this preferential localization promotes PER down-regulation, which elevates CLK binding to the DNA. This remodeling affects PDF levels, preventing its increase in summer conditions, keeping the animals in reproductive arrest and a constant winter locomotor profile in a process we termed “winter lock” (Fig. 7).

Fig. 7. Model depicting the role of timeless isoform exchange in seasonal adaptations.

During summer, in LPs and warm days, the predominant TIM-L isoform achieves its function of chaperoning PER into the nucleus to inhibit interaction of CLK-long to the DNA. CLK-long is the predominant isoform of CLOCK in warm temperatures (63). This promotes the normal daily regulation of PDF, permitting reproduction and normal daily locomotor activity. During winter, the predominant TIM-SC isoform, missing the CLD, is predominantly localized in the nucleus. Under these conditions, PER levels are down-regulated, and the predominant CLK isoform in the cold, CLK-cold, has higher binding to the DNA. The presence of this “cold molecular arrangement” is sufficient to lock the PDF levels in winter-like states, keeping flies in reproductive arrest and locomotion patterns associated with winter. Created in BioRender. S.H. (2026) https://BioRender.com/o1jldo4.

In natural conditions, coinciding environmental signals permit organisms to accurately change their seasonal behavior and physiology. AS offers an important strategy to functionally diversify the genome to respond to these changes. Here, we show that the previously described isoform of tim, tim-sc, is not only sensitive to temperature but also to photoperiod. Differential regulation of clock gene splicing by different seasonal cues has been observed in other species. In the model plant Arabidopsis thaliana, the splicing of core clock genes TOC1 and ELF3 is suppressed by short days, while cold temperature suppresses CCA1 and ELF3 AS, and induces TOC1 splicing (52, 82, 83). The AS of tim is photosensitive as opposed to thermosensitive in a northern Drosophila species, Drosophila montana (55). These highlight the conservation of splicing of clock genes as a general process by which the clock integrates seasonal environmental cues. Whether the splicing of tim or other clock components in D. melanogaster is affected by other seasonal cues similar to D. montana remains to be explored.

While temperature affects the overall expression of tim-sc, we demonstrated that the timing of its peak expression is regulated by photoperiod (Fig. 1). We have previously identified a similar modulation of PDF and EYA expression at the mRNA and protein levels by photoperiod and temperature (40). It is possible, then, that this is a common feature of seasonal cue integration in D. melanogaster. The photoperiodic sensitivity we observed in tim-sc was not observed in the tim-l canonical isoform. Thus, the coincidence of photoperiod and temperature may be important for regulating seasonality in D. melanogaster. Studies exploring how the coincidence of environmental cues differentially modulates AS are warranted. A recent study showed that seasonal morphs in the butterfly B. anynana display differential splicing that could potentially integrate with differential expression in modulating seasonal plasticity, further supporting this idea (60). Nonetheless, the functional relevance of the photoperiodic regulation of tim-sc remains unknown. It may be that the photoperiodic regulation of this isoform is secondary and not required or that it participates at another level we have yet to consider. Future experiments are required to explore this. Moreover, although we detected daily rhythms in all RNA variants across seasons, the physiological significance of these rhythms, especially the ones with low amplitudes, remain unclear. Thus, it is important to be cautious while interpreting these results.

At the protein level, we observed a phase advancement of TIM-SC expression. This is likely a combined effect of temperature and photoperiod, and posttranslational modifications, accounting for the delay of around 4 hours in the peak mRNA and peak protein levels. We found that the expression of TIM-SC is slightly rhythmic under winter-like conditions. We previously reported that TIM-SC is not rhythmic at 10°C in equinox (67). It is possible that the coincidence of low temperatures and SPs enhances the rhythmicity of TIM-SC. Our mRNA work supports this notion as tim-sc is particularly sensitive to photoperiod.

TIM-L and TIM-C levels are overall low during daytime, while TIM-SC expression, albeit rhythmic, remains high even during the daytime (Fig. 2). This is observed regardless of the N-terminal variations given by the ls-tim or s-tim alleles since it is observed in CS and w¹¹¹⁸ flies regardless. Thus, the sensitivity of TIM-SC to light may be reduced as compared with the canonical TIM-L. TIM sensitivity to light is mediated by its binding to the photosensitive protein CRYPTOCHROME (CRY) (84–87). Since the C-terminal CRY-binding domain is missing in the TIM-SC isoform, CRY-TIM interaction and subsequent TIM degradation may be affected. Further experiments would be required to test this directly.

One notable difference we observed was the subcellular localization of TIM in summer and in winter. Under summer conditions, when TIM-L is the predominant isoform, TIM expression is mostly cytosolic during early night and becomes nuclear at later times. In contrast, in winter conditions, when TIM-SC is the more highly expressed isoform, TIM is observed predominantly in the nucleus throughout the night. TIM-SC is missing a CLD, which has been shown to promote TIM retention and accumulation in the cytosol before its interaction with PER and subsequent nuclear translocation (88). Thus, it is possible that once translated, TIM-SC is immediately translocated into the nucleus, preventing its cytosolic accumulation. It is also possible that TIM-C expression constitutes part of the detected TIM signal in winter conditions as both TIM-SC and TIM-C are recognized by our antibody. TIM-C lacks nuclear export signal, key motif for exportin 1–dependent nuclear export (64), suggesting higher likelihood of nuclear localization. Experiments are warranted to dissect the localization of TIM-SC and TIM-C, respectively, but would require isoform-specific antibodies.

PER abundance and nuclear accumulation rely on TIM function (88, 89). If TIM-SC is actively translocated to the nucleus once it is produced, we would expect to see PER shuttling at the same time. However, this is not what we observed. Flies expressing only tim-sc display lower levels of PER, and it is mostly cytosolic. A possibility is that TIM-SC translocates to the nucleus without PER due to a reduced binding affinity as compared with TIM-L. Consistent with this, TIM-SC is missing half of the second PER binding domain, potentially compromising this interaction. A previous study showed that TIM-SC interacts with PER, so other mechanisms could also be at play (66). Delayed nuclear accumulation of PER and TIM is key for timekeeping since both PER and TIM accumulate for several hours before translocation (89, 90). This is tightly regulated by kinases, including casein kinase 2 (CK2), CK1α, DOUBLETIME (91–94), and SHAGGY (95–97). If TIM-SC and PER indeed form a complex, a differential interaction with any of these kinases could contribute to PER’s reduced translocation and/or expression. In addition, those kinases that translocate with PER and are involved in removing CLK from the DNA, including CK1α, would also be reduced. Further experiments are required to test these hypotheses. Moreover, PER nuclear localization may be either phase-delayed or advanced in tim-sc flies. Given that our sampling schedule is limited to three time points, it is possible that we are not capturing this phenotype. Nonetheless, the reduction in nuclear PER is consistent with an overall increase in CLK binding to the DNA, suggesting a reduction in PER-dependent removal of CLK from the DNA. Although we observed more CLK binding to the DNA, we did not observe an increase in per mRNA as an output. This suggests that on-DNA CLK is still repressed (98). A possibility is that PER-independent modulation of CLK by CK2 is altered. CK2 extensively phosphorylates TIM and likely interacts and translocates with it, thus modulating CLK activity (94, 97). Further experiments are required to clarify this. Overall, these support the notion that TIM-SC, which is preferentially expressed in winter, remodels the molecular clockwork, producing a winter clock that differs in function and output from the summer clock.

Studies in several species, including D. melanogaster (99–101), the bean bug Riptortus pedestris, the silk moth Bombyx mori (102, 103), the cabbage beetle Colaphellus bowringi (104), and the house mosquito C pipiens (44, 105), have functionally shown that the expression of core clock genes is required for seasonal adaptations (27, 29). Yet, the extent and direction of effects observed by knocking down or mutating clock genes have been variable. It is possible that the ability of the clock to influence seasonality does not rely on individual genes/isoforms but on the inherent plasticity of the clock to seasonal cues through splicing. Here, we tested this hypothesis by locking the tim isoform to the canonical tim-l or the winter-predominant tim-sc. Flies expressing tim-sc have a significant advancement in the evening peak of activity as compared to tim-l–expressing flies (Fig. 4). This locomotor architecture is similar to the one observed in flies kept in cold conditions (40, 106). The tim-sc flies do not behave as null-mutants since they have anticipation and, they partially rescue circadian rhythmicity consistent with a previous study (66). This highlights the idea of a remodeling of the clockwork in winter as opposed to a loss of function. We previously showed that the advancement of the evening peak relates to the levels of PDF (40). In winter, low PDF levels render advancement in the evening peak, similar to Pdf-null mutants (79). Consistently, flies expressing tim-sc show low levels of PDF as compared with tim-l–expressing flies (Fig. 5). This again suggests that tim-sc does not result in a tim null system since PDF is constitutively high in tim⁰¹ mutants (107).

Maintaining or exiting overwintering is key for survival (5). Here, we showed that locking the plasticity of the clock resulted in impaired seasonal adaptations. While tim-l flies are still able to enter into dormancy, regulated by a reduction in PDF, tim-sc flies are not able to exit their winter state since no increase in PDF or ovary development was observed when placing the flies in summer conditions. tim-sc transcript is not required for the reduction of PDF, but it is required to maintain low PDF levels in winter conditions (Fig. 5). It is therefore possible that both TIM-L and TIM-SC are important in playing roles at different levels of the dormancy process. Supporting this, flies missing both TIM-L or TIM-SC are unable to enter reproductive dormancy, as shown in tim-null mutant flies (67), unlike tim-l– or tim-sc–expressing flies described here. tim-sc flies do not completely phenocopy winter adaptations seen in wild-type flies reared in winter conditions (40). In that sense, other mechanisms might be in place to trigger this PDF reduction.

Even under winter conditions, PDF levels in the terminals of the s-LNvs are rhythmic in wild-type flies (40). At first sight, this is not what we observed in tim-l or tim-sc flies. It is possible that only two time points (ZT3 and ZT15) used in this study are unable to capture rhythmicity under these conditions. Another explanation is that rhythmicity is determined by isoform exchange (i.e., the presence of both isoforms is required) or by the TIM-C isoform, which is also expressed under winter conditions. Here, we mainly focused on the role of TIM-L/TIM-SC exchange in seasonal physiology. Nonetheless, TIM-C is also expressed under winter conditions, albeit at lower levels than TIM-SC. TIM-C conserves many of the domains of the canonical TIM-L isoform, and it is possible that the low amounts of TIM-C are sufficient to complement TIM-SC in the rhythmicity of PDF in winter. Low TIM-C levels may help maintain a functional circadian clock, albeit dampened, thereby explaining the partial rhythmicity in the cold, as recently shown (63). Having a dampened circadian clock has also been linked to the adaptation of changing photoperiods in Drosophila ezoana and also proposed in aphids (34, 108). Future experiments are required to test these hypotheses and to dissect the functional role of TIM-C.

Other clock components also undergo temperature-dependent AS in D. melanogaster, including per and clk, suggesting a widespread remodel of the molecular clockwork that could also be in play. In D. melanogaster, AS of the 3′ terminal intron of per is observed under cold temperatures, which contributes to a differential accumulation of per affecting seasonal locomotor adaptations (50, 51, 56). We recently showed that clk undergoes temperature-dependent splicing, producing clk-cold (63). The product, CLK-COLD, is missing a key residue, serine-13, that is required for phosphorylation-dependent inhibition of CLK. Mutations of serine-13 to aspartic acid resulted in flies that did not show a reduction in PDF under cold conditions. Thus, it is likely that clk splicing regulates overall PDF levels, while tim splicing gates the transition from low to high levels from winter to summer. It is still unknown how this is achieved. It is possible that other CLK outputs, such as HR38 and SR (80), could be modulating PDF at the transcriptional level, and VRILLE at the posttranscriptional level (81). Also, whether clk splicing isoforms are also affected by photoperiod, and/or how photoperiod and temperature collectively modulate clk splicing in the context of seasonal adaptations, remains to be explored. In addition, the exact interaction among these cold-induced clock components and how they modulate PDF levels in response to temperature changes warrants future research.

In summary, we showed that the AS of a core clock gene, tim, is key to locking flies in a winter program. Since maintaining overwintering features is key for survival, the winter lock provided by TIM-SC is perhaps a key adaptation, hence explaining the regulation of tim splicing in several insect species (55, 66). Whether similar mechanisms in which splicing of clock components lock the system either in summer or winter are also observed in vertebrates remains to be explored. The core molecular architecture is conserved across species, and splicing of clock components is a common feature for environmental integration in other species (109–111). Thus, the possibility of splicing or its misregulation mediating seasonal physiology on health and disease could be explored (112–114). Our results also support a role of photoperiod in regulating gene expression in Drosophila. Future studies exploring the extent of this regulation could provide insights into the regulation of seasonal physiology and behavior by splicing. Last, this study could shed light on the molecular adaptations to seasons of other insects, including pests and vectors. These features can be exploited to generate insecticides that hijack the splicing of seasonal components to promote mortality by impairing winter adaptations.

MATERIALS AND METHODS

Animals

Fly stocks were reared on a standard corn-yeast diet and maintained at 12:12 LD cycles at 25°C. w¹¹¹⁸ flies were obtained from the Bloomington Drosophila Stock Center (BDSC, stock #3605), and CS flies were a gift from N. Fuenzalida-Uribe and A. Guezzi (University of Puerto Rico, from BDSC #64349). Timeless rescue flies were generated as indicated below.

Generation of timeless rescue lines expressing individual tim isoforms

To generate tim-l–expressing flies, pattB-{tim(WT)-3XFLAG-6XHIS} corresponding to the full-length ls-tim allele was used as a backbone (94). PCR mutagenesis was conducted using NEB Q5 Directed Mutagenesis Kit (New England Biolabs) to mutate the second translation start site (s-tim start site; ATG to GGG) to Gly using the following primers: l-tim_mut_F 5′ GAAATCGGTTGGGGACTGGTTAC 3′ and l-tim_mut_R 5′ ACTTTATCAAAGTTCTGATTATTC 3′. The PCR product was used for KLD reaction (KLD enzyme mix, New England Biolabs), and the digested product was used for transformation in DH5α Escherichia coli. Transformants were selected, and plasmid DNA was sent for sequencing (Genewiz, Azenta) to confirm the desired mutation. Final pattB{tim-l-3XFLAG-6XHIS} plasmid was sent for injection to BestGene (Chino Hills, CA). To generate the tim-sc–expressing flies, pattB{tim-l-3XFLAG-6XHIS} was used as a backbone, and the extra C-terminal segment between the end site for tim-sc and the end site for tim-l sequence was deleted by PCR mutagenesis as described before, using the following primers: TIM-Q5delCterm_R (TIM-SC) 5′ ACCAGGAGCATACCGTTTGG 3′ and 3x FLAG N-term F 5′ GACTACAAAGACCATGACGGT 3′. Both tim-l and tim-sc transformants were crossed to a tim-null mutant background using tim⁰¹ flies (115) to obtain the rescue flies used in all experiments.

Total RNA extraction and quantitative RT-PCR

Total RNA extraction and quantitative reverse transcription PCR (RT-PCR) were performed as previously described (40, 63). Flies were collected on dry ice, and the heads were separated using prechilled sieves. TRI reagent (Sigma-Aldrich, St. Louis, MO) was used to extract total RNA from heads. RNA quantification was carried out on a NanoDrop Spectrophotometer (Thermo Fisher Scientific, Waltham, MA). For reverse transcription, we used the Superscript IV cDNA Synthesis kit (Invitrogen, Waltham, MA) following the manufacturer’s instructions, with 1 μg of RNA as starting material. qPCR was performed on a Bio-Rad CFX96 Real Time System (Bio-Rad, Hercules, CA) using SsoAdvanced Universal SYBR green super mix (Bio-Rad, Hercules, CA) following the following program: initial 95°C for 30 s, followed by 40 cycles of 95°C for 5 s, and 60°C for 30 s. Primers used are as follows: for tim-l: 5′ CCCTTATACCCGAGGTGGAT 3′ and 5′ TGATCGAGTTGCAGTGCTTC 3′, for tim-c: 5′ GCATCTGTGTACGAAAAGGA 3′ and 5′ ATGTAACCTATGTGCGACTC 3′, for tim-sc: 5′ AACACAACCAGGAGCATAC 3′ and 5′ ATGGTCCACAAATGTTAAAA 3′, for per: 5′ GACCGAATCCCTGCTCAA 3′ and 5′ GTGTCATTGGCGGACTTC 3′, and for cbp20: 5′ GTCTGATTCGTGTGGACTGG 3′ and 5′ CAACAGTTTGCCATAACCCC 3′, whose expression was used for normalization. Data were analyzed using the ΔΔCt method.

TIM antibody generation

TIM antibody (RRID: AB_3713152) was generated by Antibodies Inc. (Davis, CA). To generate an antibody against the shared N terminus of all TIM isoforms, a peptide comprising the region from 292 to 590 amino acids was generated (fig. S1, A and B), purified, and injected into rabbits. Serum was processed by affinity purification with TIM peptide, and the purified TIM antibody was validated by immunodetection of TIM in flies, Drosophila S2 cells (fig. S1), and in brains for immunofluorescence (fig. S2).

Protein extraction and Western blot

Protein extraction was carried out as previously described (40, 63). Briefly, the flies were collected on dry ice and heads separated from the bodies using chilled sieves. Whole heads were then used for total protein extraction. Around 150 μl of extraction buffer [0.1% glycerol, 20 mM Hepes (pH 7.5), 50 mM KCl, 2 mM EDTA, 1% Triton X-100, 0.4% NP 40, 1 mM dithiothreitol (DTT), aprotinin (10 μg/ml), leupeptin (5 μg/ml), pepstatin A (1 μg/ml), and 0.5 mM phenylmethylsulfonyl fluoride] was added for 50 μl of heads for each sample, ground using a motorized pestle, and then centrifuged. The supernatant containing the proteins was collected, quantified, aliquoted in SDS loading buffer, and stored at −20°C until electrophoresis. Electrophoresis was carried out in 8% polyacrylamide SDS gels for TIM and 10% for HSP70, and then the proteins were transferred to nitrocellulose membranes (0.45 μm, Bio-Rad, Hercules, CA) on a Trans-Blot semi-dry transfer system (Bio-Rad, Hercules, CA). Membranes were incubated in blocking solution (5% Blotting Grade Blocker Non-Fat Dry Milk, Bio-Rad, Hercules, CA) for 1 hour before antibody incubation overnight at room temperature. Membranes were washed in 0.05% TBST (0.05% Tween 20 in 1× tris-buffered saline) and incubated with secondary antibody for 2 hours before final TBST washes and imaging. For imaging, membranes were incubated with Clarity ECL reagent (Bio-Rad, Hercules, CA) and imaged using a ChemiDoc MP Imaging system (Bio-Rad, Hercules, CA). The dilution of the antibodies is as follows: rabbit α-TIM antibody (RRID: AB_3713152) at 1:3000, mouse α-HSP70 (SAB4200714, Sigma-Aldrich) at 1:7000, α-PER (GP5620; RRID: AB_2747405) at 1:3000, α–rabbit–immunoglobulin G (IgG) horseradish peroxidase (HRP)–conjugated (Cytiva, Marlborough, MA) at 1:2000, and α–mouse-IgG HRP–conjugated (Cytiva) at 1:7000.

ChIP-qPCR

ChIP-qPCR was conducted as described in Cai et al. (63). Fly heads were collected on dry ice and then ground using a chilled mortar and pestle. Nuclear extraction buffer [10 mM tris-HCl (pH 8.0), 0.1 mM EGTA (pH 8.0), 10 mM NaCl, 0.5 mM EDTA (pH 8.0), 1 mM DTT, 0.5% Tergitol NP-10, 0.5 mM spermidine, and 0.15 mM spermine] was used to homogenize the tissue with a glass dounce homogenizer. After filtration using a 70-μm cell strainer and centrifugation, the nuclear fraction was obtained by sucrose gradient. This fraction was used for cross-linking with 0.3% formaldehyde for 10 min, followed by cross-linking quenching with glycine, and then sonicated to fragment the DNA. Samples were centrifuged at 10,000 rpm for 10 min, and the supernatant was processed for pull-down. Guinea pig α-CLK antibody (GP6139, RRID: AB_2827523) (94) bound to Dynabeads (Thermo Fisher Scientific, Waltham, MA) was used for CLK pull-down. Samples were incubated with the beads for 2 hours at 4°C in IP buffer [50 mM tris-HCl (pH 7.5), 2 mM EDTA (pH 8.0), 1% Triton X-100, 0.1% sodium deoxycholate (DOC), 150 mM NaCl, and 0.5 mM EGTA (pH 8.0)], and then beads were pelleted and washed in washing buffer [50 mM tris-HCl, 1 mM EDTA (pH 8.0,) 1% Triton X-100, 0.1% DOC, Acetylated Bovine Serum Albumin (AcBSA;10 μg/ml, Promega, Madison, WI), 100 mM KCl in 1× phosphate-buffered saline (PBS,) 150 mM NaCl, 5 mM EGTA (pH 8.0), and 0.1% SDS]. Beads were then eluted with elution buffer [50 mM tris-HCl (pH 8.0), 10 mM EDTA (pH 8.0), 1% SDS, 1 mM DTT, 50 mM NaCl, proteinase K (4 U/ml, NEB, Ipswich, MA), and ribonuclease A (50 μg/ml, Thermo Fisher Scientific, Pleasanton, CA] and de-crosslinked overnight at 65°C. QIAquick PCR purification kit (Qiagen, Germantown, MD) was used to purify the DNA, which was quantified and used for qPCR. For perCRS, we used 5′ TGCCAGTGCCAGTGCGAGTTCG 3′ and 5′ TGCCTGGTGGGCGGCTGG 3′. For background deduction, the average of amplification of two regions was used: 2R intergenic (CP023338) and X intergenic (FBgn0003638). The primers for 2R were 5′ TCAGCCGGCATCATTAGCAGCCG 3′ and 5′ TCGTGTGCGGGAATCTCTGCCG 3′, and for X 5′ ACTGCGTATTCAGGATACATGCC 3′ and 5′ TGTCCACTTTAATTGATTGCGTGG 3′.

Drosophila activity monitoring

Daily locomotor activity was monitored using the Drosophila Activity Monitoring System (DAMS, TriKinetics, Waltham, MA) as described previously (40, 116). Data were analyzed using ShinyR DAMS (117), and actograms were generated with ActogramJ (118) using the average activity of the whole population. Activity profiles were obtained by averaging the daily activity of flies across the entire 4 days of LD. The evening peak phase was estimated as the highest point of the evening activity peak, as described in previous publications (40), using 15-min binned data. Circadian activity was calculated from data collected over 7 days in DD. Only flies that survived until the last day of the experiment were included in the circadian analysis. The 15-min binned data were used to calculate rhythmicity in DD using a chi-square periodogram. The calculated power (Qp.act) for all flies was used to generate the periodogram reported in Fig. 4. Significance thresholds are indicated as Qp.sig.

Reproductive dormancy

Reproductive dormancy assays were conducted as previously described (35, 40). Newly eclosed female flies were collected between ZT4 and ZT10 and placed in incubators (Tritech Research, Los Angeles, CA) in either simulated summer (16:8 LD at 25°C) or simulated winter (8:16 LD at 10°C) for a week. After this period, flies were collected in 70% ethanol, and ovaries were dissected in 1× PBS. Images of individual ovaries were acquired using an EVOS microscope (Thermo Fisher Scientific), and the percentage of dormancy in a population was quantified as the number of ovary pairs with no eggs or immature eggs over the total. Three trials were conducted, each consisting of ~20 flies.

Whole-brain immunofluorescence

Immunofluorescence was conducted as previously described (40). Flies were collected at the indicated time points/conditions in 4% paraformaldehyde (PFA) in 0.2% PBST (0.2% Triton 100 in 1× PBS) and incubated at room temperature for 40 min with agitation. After that, the brains were dissected in 0.2% PBST and kept on ice until all conditions were dissected. Dissected brains were refixed with 4% PFA for 20 min and then washed four times with 0.2% PBST for 10 min each. Brains were then incubated in 5% normal goat serum in 0.2% PBST for 30 min, followed by overnight incubation with the primary antibody. The brains were then washed four times in 0.2% PBST at room temperature for 10 min each. Brains were incubated for 2 hours with the secondary antibodies diluted to their individual working concentration in 5% normal goat serum. Following four 10-min washes with 0.2% PBST, the brains were incubated in 50% glycerol at 4°C for 30 min. Samples were mounted using ProLong Diamond Antifade Mountant (Thermo Fisher Scientific, Waltham, MA) and left in the dark to cure for at least 24 hours before imaging. The dilution of the primary antibodies was as follows: rabbit α-TIM (RRID: AB_3713152) 1:300, mouse α-PDF 1:200 (Developmental Studies Hybridoma Bank), and rabbit α-PER (Rb s3968-1; RRID: AB_2747406) 1:500. The dilution of the secondary antibodies is as follows: goat α-rabbit IgG (H + L) Alexa Fluor 488 (Invitrogen, Waltham, MA) 1:500 and goat α-mouse IgG Alexa Fluor 647 conjugate (Cell Signaling Technology, Danvers, MA) 1:1000. PDF C7 was deposited in the DSHB by J. Blau (DSHB Hybridoma Product PDF C7).

Images were captured with a Leica SP8 confocal microscope equipped with excitation diodes at 638 nm and OPSL exciting at 488 nm. For the experiments determining PDF levels in the s-LNvs dorsal terminals, images were taken every 1 μm using a 40× oil objective and digital zoom. For images of the whole brain, a 20× objective was used instead. All images were analyzed using Fiji, and PDF levels were assessed as the intensity in the s-LNvs dorsal terminal normalized by the background signal in the same dorsal region. To quantify the nuclear and total amount of PER and TIM, images of the LNvs cell bodies were taken every 1 μm, and a cross-sectional plane for each cell body was selected for quantification. Laser power, detector gain, and digital and optical zoom were kept constant across samples to ensure accurate quantification.

Quantification and statistical analysis

For XY graphs, data were represented as means ± SEM. For all other representations, whiskers and boxes were used, and the result of each biological replicate is represented in the graphs as a single dot. For circadian statistics, we used CircaCompare (119) or RAIN (120) as indicated for each figure. CircaCompare was used if both conditions in a graph were rhythmic; otherwise, we used RAIN. CircaCompare cannot be used to compare two rhythms unless both are rhythmic. CircaCompare uses the nonlinear mixed-effects models to estimate how well a set of data fits the formula measure ~κ + α * cos(τ − φ), where κ represents the MESOR, a rhythm-adjusted mean, α represents amplitude, τ is the period, refers to the time in hours, and φ is the phase. If the data fit a sinusoidal curve of this form, even if it has a low amplitude, the results will be a significant P value. The specific test used for each dataset is defined in the figure legends, and the statistics for each test are reported in the results section, along with the corresponding P values. All datasets were tested for normality using a Shapiro-Wilk normality test, and the results were used to inform the best test to use.

Supplementary Materials

This PDF file includes:

Figs. S1 to S6