Temperature-dependent resetting of the molecular circadian oscillator in Drosophila

Abstract

Circadian clocks responsible for daily time keeping in a wide range of organisms synchronize to daily temperature cycles via pathways that remain poorly understood. To address this problem from the perspective of the molecular oscillator, we monitored temperature-dependent resetting of four of its core components in the fruitfly Drosophila melanogaster: the transcripts and proteins for the clock genes period (per) and timeless (tim). The molecular circadian cycle in adult heads exhibited parallel responses to temperature-mediated resetting at the levels of per transcript, tim transcript and TIM protein. Early phase adjustment specific to per transcript rhythms was explained by clock-independent temperature-driven transcription of per. The cold-induced expression of Drosophila per contrasts with the previously reported heat-induced regulation of mammalian Period 2. An altered and more readily re-entrainable temperature-synchronized circadian oscillator that featured temperature-driven per transcript rhythms and phase-shifted TIM and PER protein rhythms was found for flies of the ‘Tim 4’ genotype, which lacked daily tim transcript oscillations but maintained post-transcriptional temperature entrainment of tim expression. The accelerated molecular and behavioural temperature entrainment observed for Tim 4 flies indicates that clock-controlled tim expression constrains the rate of temperature cycle-mediated circadian resetting.

1. Introduction

Many organisms use circadian clocks to maintain daily rhythms in their physiology and behaviour, and synchronize them to daily environmental cycles in light and temperature. The fruitfly Drosophila melanogaster has emerged as a successful model system that has aided in the discovery of components and pathways underlying daily time keeping in animals. A central role in the Drosophila molecular clock circuits is played by the transcription factor CLOCK/CYCLE (CLK/CYC) [1]. Late in the day CLK/CYC induces the transcription of a number of target genes, including the negative feedback regulators period (per) and timeless (tim). PER/TIM-containing protein complexes accumulate in the cytoplasm with a substantial delay relative to the expression of per and tim mRNA. The casein kinase Iε DOUBLETIME (DBT) contributes to this delay by binding and phosphorylating PER, and targeting it for proteasome-mediated degradation. TIM regulates PER by stabilizing it and controlling its nuclear transfer around midnight. Once in the nucleus PER binds and inhibits CLK/CYC.

Circadian clocks are able to maintain synchrony with their environment by entraining to time cues such as those provided by daily temperature cycles. Intact flies entrain both molecular and behavioural circadian rhythms to environmental temperature cycles [2–7]. Moreover, temperature-dependent rhythms in clock gene–luciferase reporter constructs persist in cultures of dissected tissues [5,6,8], indicating that relevant temperature sensors are distributed throughout the fly. It is conceivable that parts of the circadian cycle itself are directly temperature-sensitive. In fact, splicing of the 3′-terminal intron of the per gene was revealed as a temperature-sensitive determinant of both per expression and daily locomotor activity profiles. Moreover, the 3′-terminal intron of tim was also found to be spliced in a thermally sensitive manner [2]. Meanwhile, in mammals, temperature-dependent expression of the per homologue Per2 was proposed to be governed by Heat Shock Factor 1 (HSF1, [9,10]). Although body temperature is regulated differently in homeotherms (such as mammals) versus poikylotherms (such as Drosophila), it may be used by both types of organism to synchronize peripheral clocks. In mammals, circadian control of a reliable low-amplitude body temperature rhythm is capable of synchronizing clock-controlled gene expression [11]. In Drosophila, body temperature rhythms are determined by the interaction between daily environmental temperature cycles and clock-controlled rhythms in temperature preference [12]. This study provides insights into the way that thermal entrainment acts on the molecular circadian oscillator of peripheral clocks, and examines the involvement of the HSF1/HSP90 pathway and regulation of tim expression in mechanisms of circadian temperature entrainment in Drosophila.

2. Results

(a). Association of the HSF1/HSP90 pathway and regulation of tim expression with daily temperature entrainment of Drosophila locomotor activity

This study primarily addresses the molecular mechanisms of temperature-dependent re-entrainment of circadian rhythms in the adult Drosophila head. Specifically, experiments were performed to determine the ordered temperature-mediated re-entrainment of the transcripts and proteins encoded by the core clock genes per and tim. Analyses were conducted for w¹¹¹⁸ control flies as well as two experimental genotypes, Hsp90^e6D/08445 and y w; tim⁰¹; {tim_P-tim_cDNA} (‘Tim 4′), for which we observed accelerated temperature re-entrainment in behavioural analyses as outlined below.

Transcript profiling of adult Drosophila heads revealed wide-spread temperature-driven as well as temperature-entrained circadian responses [2]. Notably, a set of heat-induced transcripts associated with the HSF1/HSP90 pathway was found to cycle in both wild-type and arrhythmic tim⁰¹ mutant heads. This regulatory signature was of particular interest given studies in mouse liver which proposed that mouse HSF1 acts as a mediator of daily gene expression rhythms in response to body temperature cycles [9,13]. Moreover, in parallel with its regulation of Hsp genes, HSF1 also controlled the master clock gene mPer2, thus providing a mechanism for temperature entrainment of the molecular circadian oscillator [9,14,15]. The possible existence of a similar temperature entrainment pathway in flies was addressed by behavioural phenotyping of genetic mutations affecting the Drosophila HSF1/HSP90 pathway. In particular, Hsp90 mutants were found to affect temperature-mediated resetting of locomotor activity rhythms. When challenged with a large phase advance in their environmental temperature cycle in DD Hsp90^e6D/08445 mutant flies exhibited a significant reduction in the number of ‘transition days’ associated with the process of phase-resetting daily locomotor activity (see the electronic supplementary material, figure S1).

A separate rationale prompted investigation of regulation of the core clock gene tim in circadian temperature entrainment. tim expression is one of the steps in the molecular circadian cycle that are subjected to temperature-dependent regulation. Splicing of the 3′-terminal intron of tim transcript, which is needed to produce full-length TIM protein, occurs preferentially at higher temperatures [2], while tim transcription at colder temperatures becomes more sensitive to light-mediated induction [16]. Both of these control mechanisms, as well as circadian regulation of tim transcription, are lost in Tim 4 flies under DD conditions. This genotype features rescue of the tim⁰¹ mutation by a transgene, which expresses a tim cDNA (lacking the 3′-terminal intron) under control of a 4.3 kb tim promoter fragment [17,18]. Analyses of behavioural temperature entrainment and resetting of Tim 4 in DD revealed that these flies adjusted their locomotor activity rhythms earlier than control flies (see the electronic supplementary material, figure S1).

The above-mentioned acute behavioural resetting phenotype was, in part, associated with weaker ‘free-running’ behavioural rhythms under constant conditions (electronic supplementary material, table S1) [18,19]. Although defects in temperature entrainment of circadian locomotor activity are likely to be attributable to defective signalling in the neural clock circuit, there are reasons to assume that HSP90 and TIM, in principle, could affect temperature entrainment in all clock-bearing cells: (i) HSP90 expression is ubiquitous [20] and TIM is found in all clock-bearing cells [21], (ii) the molecular circuits of clock neurons and peripheral clocks are very similar [1], and (iii) peripheral clocks can autonomously entrain to environmental temperature cycles [5,6,8]. We therefore studied temperature-mediated resetting of molecular circadian rhythms in wild-type and mutant adult heads. As most of the clock gene expression in the head maps to the compound eyes and other tissues outside of the brain [22], our analyses generally represent peripheral clocks.

(b). Temperature-dependent regulation of multiple steps in the molecular circadian cycle

Fly cultures of the three aforementioned genotypes in constant dark (DD) conditions were entrained to a daily temperature gradient (gradually ramping between 16°C and 28°C) and then sampled every 4 h for two additional days under stable temperature entrainment or during 2 days following a 180° phase shift in the daily temperature gradient. Adult head extracts were prepared and analysed for per and tim transcript, and protein time course profiles, using quantitative reverse transcriptase PCR (qRT-PCR) and quantitative Western blotting analyses. Statistical analyses of the effect of genotype on expression levels indicated no effect on tim transcript and only a modest effect on per transcript (see the electronic supplementary material, figure S2). Genotype, however, did appear to strongly affect TIM and PER protein. Most notably, TIM levels were significantly lower in Tim 4 heads. The transgenic rescue construct in this genotype that is lacking some crucial cis-regulatory elements fails to generate rhythmic tim transcript levels [18] (figure 1), and this also affects TIM protein expression.

Figure 1.

Temperature-dependent entrainment and resetting of molecular circadian rhythms in w¹¹¹⁸, Tim 4 and Hsp90^e6D/08445 adult Drosophila heads. Flies were temperature-entrained in DD and then harvested every 4 h over 2 further days either at the original conditions (ZT8, 12, 16, 20, 0, 4, 8, 12, 16, 20, 0, 4; with ZT0 and ZT12 at the temperature maximum and minimum, respectively; solid black line) or directly following a 12 h environmental phase shift (t = 2, 6, 10, 14, 18, 22, 26, 30, 34, 38, 42, 46 h after t0, the time of transfer, which coincided with ZT6; solid grey line). Total protein and RNA from adult head tissue was analysed by quantitative Western blotting and RT-PCR, respectively. Expression ratios were determined relative to constitutively expressed internal controls (Ef1β, HSP70) and normalized to the time course average. The plotted data represent the average (±s.e.m.) of four biological replicates representing stable entrainment (black squares and lines) or resetting (open circles and grey lines). The shaded bars below the graphs represent the pre-shift (top with ZT times in black squares) and post-shift (bottom with t times in open circles) temperature cycles. Note the different patterns of resetting associated with per transcript (early phase advance), tim transcript (later phase advance), and PER and TIM protein (later phase delay), as well as the molecular phenotype of the Tim 4 genotype (loss of tim transcript rhythms and virtually instantaneous resetting of other rhythms). Welch test p-values are indicated whenever gene expression was predicted to significantly vary over the experimental time course. Time points for which post-hoc analyses predicted a significantly increased expression level are annotated (*0.01 <p < 0.05; **0.001 <p < 0.01; ***p < 0.001).

Significant tim transcript oscillations under stable entrainment were detected for w¹¹¹⁸ and Hsp90^e6D/08445, with peak tim mRNA levels approximately 4 and 8 h after the temperature maximum, respectively (figure 1; electronic supplementary material, figure S3 and table S2). Upon exposure to an anti-phase temperature cycle, the tim transcript rhythms in both of these genotypes continued to oscillate in their original phase for approximately 20 h before completing most of the imposed phase shift by phase advances within the next approximately 12 h (figure 1; electronic supplementary material, figure S3 and table S3).

per transcript on the other hand exhibited significant rhythms in Hsp90^e6D/08445 but not in w¹¹¹⁸ heads (figure 1; electronic supplementary material, figure S3 and table S2). In Hsp90^e6D/08445 heads, the temperature-entrained peak for per transcript roughly coincided with that of tim transcript. Interestingly, significant rhythmic regulation of per transcript was observed in all genotypes during temperature-dependent resetting. Under these conditions, w¹¹¹⁸ per transcript expression patterns resembled those observed for Hsp90^e6D/08445. Moreover, resetting of per transcript rhythms by phase advances occurred within hours of the environmental shift and well in advance of resetting of the three other molecular rhythms examined here (tim transcript, TIM and PER protein).

TIM and PER protein expression exhibited significant daily rhythms for stably temperature-entrained w¹¹¹⁸ and Hsp90^e6D/08445 flies (figure 1; electronic supplementary material, figure S3 and table S2). Relative to the environmental cycle, maximal PER and TIM protein expression occurred shortly before or, in the case of PER in Hsp90^e6D/08445, at the temperature minimum. The phase of peak TIM protein levels in w¹¹¹⁸ and Hsp90^e6D/08445 heads showed a delay of less than 4 h relative to the peak of tim mRNA, which is shorter than that reported for wild-type under light-entrained conditions [23], while the delay between per mRNA and PER protein levels in Hsp90^e6D/08445 heads of slightly more than 4 h is more in line with previous observations for wild-type under LD conditions [4,23] (figure 1; electronic supplementary material, table S2). Despite the dampened and abnormal daily pattern of per transcript in w¹¹¹⁸ head extracts, PER protein maintains roughly the same phase relationship relative to tim transcript and TIM protein in temperature-entrained w¹¹¹⁸ and Hsp90^e6D/08445 flies. During temperature-mediated re-entrainment of w¹¹¹⁸ and Hsp90^e6D/08445 flies significant changes were observed for PER and TIM protein (figure 1), although the effect for PER in w¹¹¹⁸ was observed only for ANOVA (p < 0.006) and not for the Welch test (p = 0.11). Temperature-dependent resetting of TIM and PER protein rhythms became notable in w¹¹¹⁸ and Hsp90^e6D/08445 heads after approximately 20 h of exposure to the anti-phase daily temperature gradient, and accomplished most of the imposed phase shift by phase-delaying over the next half day. Thus, while PER and TIM protein reset during the same time interval as tim transcript, they do so by phase shifting in the opposite direction (figure 1; electronic supplementary material, figure S3 and table S3). PER protein does appear to track TIM protein reasonably well over the time courses representing both stably entrained and resetting conditions, consistent with the well-established requirement for TIM to stabilize PER [24].

(c). Temperature-mediated molecular re-entrainment is faster in Tim 4 peripheral clocks

In Tim 4 heads, significant oscillations were detected for per but not tim transcript under conditions for stable entrainment (figure 1; electronic supplementary material, figure S3). The absence of tim transcript oscillations was expected based on previous characterizations of this genotype [18]. Maximal per transcript was detected at the temperature minimum, which is delayed relative to peak per transcript for Hsp90^e6D/08445. Remarkably, re-entrainment of per transcript rhythms for Tim 4 flies was essentially complete within half a day, while re-entrainment still appeared to be incomplete after the 2-day time course for Hsp90^e6D/08445 flies.

In spite of the lower TIM protein levels in Tim 4 heads (see the electronic supplementary material, figure S2), significant daily changes were detected under conditions of temperature-dependent resetting (figure 1) and stable temperature entrainment (electronic supplementary material, figure S3). As for per transcript, the stably entrained peak phase of TIM protein in Tim 4 heads showed a delay relative to the phase of the TIM maximum for Hsp90^e6D/08445 (or w¹¹¹⁸). In addition, temperature-dependent resetting of TIM protein was very fast (figure 1; electronic supplementary material, figure S3 and table S3) and, though not detected as significant rhythms, the PER protein profiles (figure 1; electronic supplementary material, figure S3) appear consistent with accelerated resetting of the molecular clock in Tim 4 flies upon exposure to an anti-phase temperature gradient.

Thus, the Tim 4 genotype causes not only a loss of tim transcript rhythms (figure 1; electronic supplementary material, figure S3) [18] and a reduction in TIM protein levels (electronic supplementary material, figure S2), but also altered phase relationships of per transcript and TIM protein relative to environmental temperature cycles, and very quick temperature-dependent molecular resetting of the circadian clock in the adult fly head. The apparently almost instant molecular circadian resetting in Tim 4 heads could reflect either temperature-driven regulation in the absence of self-sustaining peripheral clock function or a residual peripheral clock with an altered sensitivity to temperature-dependent resetting. While the locomotor activity rhythms of Tim 4 flies suggest that they have a weakened but active circadian clock in their circadian pacemaker neurons (electronic supplementary material, table S1), this assay does not reflect clock function elsewhere in the fly. To directly monitor peripheral molecular rhythms following either stable temperature entrainment or partial temperature-mediated resetting, circadian gene expression was examined for the tim-luc transgene, which expresses the firefly luciferase enzyme under control of the tim promoter [25]. Whole fly in vivo tim-luc-generated luminescence rhythms were observed under constant conditions for temperature-entrained y w {tim-luc}; tim⁰¹; {tim_P-tim_cDNA} (‘tim-luc Tim 4′) experimental and y w {tim-luc} (‘tim-luc’) control flies following subsequent exposure for 0, 1 or 2 days to an anti-phase environmental temperature gradient. Although there were some significant genotype- and condition-dependent differences for the rhythmicity, rhythmic power and period length, the majority of flies for all genotype/condition combinations showed detectable circadian rhythmicity (electronic supplementary material, table S4). Thus, it was possible to analyse temperature-dependent resetting of the phase of free-running molecular luciferase activity (figure 2). The tim-luc Tim 4 flies exhibited obvious phase shifts in response to treatment with one or two anti-phase temperature cycles, with a stronger effect for the latter. By contrast, the clocks in tim-luc flies were much more resistant to temperature-mediated re-entrainment of whole fly luciferase activity rhythms. Therefore, the Tim 4 genotype maintains significant central and peripheral clock function but is associated with accelerated temperature-dependent resetting of both locomotor activity and peripheral molecular rhythms.

Figure 2.

Temperature-dependent resetting of peripheral circadian luciferase activity in tim-luc and tim-luc Tim 4 flies. Female flies of the indicated genotypes (see the electronic supplementary material, table S4 for numbers) were placed on luciferin sugar agar media and entrained to a daily 16–28°C temperature gradient in DD and then exposed to the same conditions in the opposite phase for 0, 1 or 2 days prior to in vivo luciferase assays conducted in DD at constant ambient temperature in an automated TopCount luminescence counter. Luminescence counts were averaged for each genotype/treatment combination. Resulting profiles were detrended, normalized and analysed for rhythmicity using BRASS software. Along with the experimental data (thin lines and squares), fitted rhythmic components are indicated by the smooth thick lines. The latter are used to illustrate the effect of temperature re-entrainment on the resulting circadian phase in the bottom panels. Note that resetting of circadian molecular phase is much more pronounced for the tim-luc Tim 4 flies following exposure to either 1 or 2 anti-phase temperature cycles. (Online version in colour.)

(d). per transcript levels are regulated by both temperature-driven transcription and splicing

A potential explanation for the dampened daily per transcript profile in temperature-entrained w¹¹¹⁸ fly heads is that the high-amplitude clock-regulated oscillations observed in per transcript [26] are counteracted by a temperature-driven mechanism favouring a different phase relationship with the environmental temperature cycle. To address this issue, the temperature-regulated daily per and tim transcript profiles in adult heads for the w¹¹¹⁸, Hsp90^e6D/08445 and Tim 4 genotypes were compared with those observed for arrhythmic y w; tim⁰¹ heads (figure 3a). Indeed, in the absence of a functional clock, highly significant temperature-driven regulation of per mRNA was observed with peak and trough phases that coincided with the temperature minimum and maximum, respectively, as well as a 4.9-fold peak/trough ratio. By contrast, total tim mRNA levels showed only modest temperature-driven changes (1.7-fold peak/trough) in the arrhythmic tim⁰¹ genotype. Flies from the Tim 4 genotype, which lacked circadian control of tim mRNA expression (figure 1; electronic supplementary material, figure S3) [18] and had reduced behavioural rhythmicity (electronic supplementary material, figure S1 and table S1), strongly resembled y w; tim⁰¹flies for their per and tim expression profiles in the presence of a daily temperature gradient. However, the presence of a wild-type tim gene and a stronger circadian clock in w¹¹¹⁸ and Hsp90^e6D/08445 fly heads resulted in very different daily mRNA profiles for per and tim.

Figure 3.

Temperature-driven regulation of per transcript levels and splicing. Temperature-dependent average daily transcript profiles (±s.e.m.; n ≥4) were examined in arrhythmic mutant flies for per and tim as well as Gal4 and per::luc transgenes expressed from a 4.2 kb per promoter. Significant variation of gene expression with the daily 16–28°C temperature gradient is reflected by the annotated Welch test p-values. Time points with significantly increased expression levels according to post-hoc analyses are indicated (*0.01 < p < 0.05; **0.001 < p < 0.01). (a) Daily per and tim transcript profiles in y w; tim⁰¹ (with annotated statistics) and three other genotypes (annotated in the electronic supplementary material, figure S3). Note that, unlike the transcript profiles for Hsp90^e6D/08445 and ‘wild-type’ (w¹¹¹⁸), the ones for Tim 4 closely match those of arrhythmic y w; tim⁰¹ flies; also, the clock-independent temperature-driven regulation of daily transcript levels that is detected in the tim⁰¹ background for per is much more prominent than that for tim. (b) Daily profiles in tim⁰¹ mutant heads of native B’ (3′-UTR spliced) and A (3′-UTR unspliced) per transcripts as well as their ratio. Note that the transgenic per::luc transcript lacks the 3′ end of the per gene and is therefore not detected here. (c) Daily transcript profiles for transgenic per reporter genes. The y per⁰¹ w; {per_P(4.2kb)-Gal4} flies represented by the data in the left-hand panel exhibited daily cycling in the native per⁰¹ message, but not in the transgenic Gal4 transcript expressed under control of a 4.2 kb per promoter fragment. By contrast, analysis of the y w {per::luc}; tim⁰¹ flies (right-hand panel) revealed significant mRNA oscillations for a per::luc transgene encompassing not only the 4.2 kb upstream promoter, but also 5.6 kb of the downstream per gene. Moreover, additional reactions using primers targeting the first intron of per mRNA rather than the luciferase gene revealed significant daily rhythms in per pre-mRNA. In this case, pre-mRNA from both the native per gene and the per::luc transgene was detected.

The 3′-terminal intron of per is known to be alternatively spliced in a temperature-dependent manner [27]. Increased splicing of the per 3′-UTR at colder temperatures is associated with increased transcript and protein levels, and contributes to temperature-dependent modulation of the daily pattern of locomotor activity [27], and was therefore hypothesized to contribute to the observed temperature-driven regulation of per transcript. Hence, the daily expression profiles for both the B’ (3′-UTR spliced) and A (3′-UTR unspliced) native per transcript isoforms were determined by qRT-PCR for two arrhythmic mutant fly lines (y w; tim⁰¹ and y w BG-luc; tim⁰¹, see below) exposed to a temperature gradient (figure 3b). Consistent with previous studies, a trend in the ratio of spliced to unspliced per was found that indicated increased levels of splicing at colder temperatures. This effect was statistically significant for the y w BG-luc; tim⁰¹ line as well as for the combined data for both lines. Nevertheless, in both lines, not only 3′UTR-spliced B’-type but also 3′UTR-unspliced A-type native per transcripts displayed highly significant temperature-driven regulation with maximal expression at the daily temperature minimum. The reproducible increase in the levels of A-type transcript at the coldest time of day, in particular, could not be explained by temperature-dependent splicing at the 3′UTR and pointed to an alternative temperature-driven regulation of per transcript levels. Moreover, BG-luc, a per::luc transgene that lacks the 3′UTR as well as the last three coding exons of per, still showed a robust temperature-driven transcript profile (figure 3c; see below). Therefore, the contribution of transcriptional regulation to temperature-driven per mRNA expression was investigated.

Previous analyses have demonstrated that a 4.2 kb fragment of the per upstream promoter is sufficient to confer circadian regulation to transgenic reporters [28]. To examine whether this segment also harboured sequences responsible for temperature-driven regulation of per mRNA, qRT-PCR analyses of head extracts from arrhythmic y per⁰¹ w; {per_P(4.2kb)-Gal4} flies were conducted for both the transgenic Gal4 and the native per⁰¹ message (figure 3c). While the per⁰¹ mRNA, as expected, presented significant daily regulation in the presence of a temperature cycle (p = 0.01), this was not the case for the Gal4 reporter transcript. Thus, the 4.2 kb upstream per promoter was unable to confer temperature-driven regulation of mRNA expression in this context. Next, we tested whether other sequences from the native per gene played a role in the observed thermal regulation of per transcript levels. For this purpose, expression profiles were determined from heads of y w BG-luc; tim⁰¹ flies exposed to a daily temperature cycle for the BG-luc transgenic fusion transcript encompassing part of per as well as the fire fly luciferase gene (figure 3c), as well as native per transcripts (figure 3b). In contrast to the per_P(4.2kb)-Gal4 transgene [29], the BG-luc transgene fusion used in this experiment [30] did exhibit highly significant temperature-driven regulation (p < 0.001) that was comparable to that observed for native per. As the BG-luc transgene does not just include the aforementioned 4.2 kb per promoter fragment, but also 5627 bp of the per sequence downstream of the transcription start site [30], it appears that cis-regulatory sequences involved in temperature-mediated regulation of per mRNA expression reside in this additional segment of the per gene. Finally, we asked whether temperature-driven expression could be observed for nascent per transcript. To test this, qRT-PCR analyses were performed on the above-described y w BG-luc; tim⁰¹ head extracts using an amplicon in the first intron of per. Indeed, a significant temperature-driven oscillation (p < 0.02) was detected in this experiment (figure 3c). Moreover, the expression profile for per pre-mRNA strongly resembled that for the BG-luc transgenic mRNA, as well as native per mRNA (cf. figure 3b,c). Thus, our results indicate temperature-driven transcriptional regulation of per by intragenic cis-regulatory elements.

3. Discussion

(a). Environmental temperature cycles separately impact at least three molecular parameters of the circadian cycle

As described above, daily rhythms in tim transcript, per transcript, and TIM or PER protein levels responded differently to temperature-mediated resetting. In the heads of wild-type adults challenged with a 180° phase shift in their environmental temperature cycle, per transcript reacted quickly with an increase in rhythmic amplitude and an apparent advance in phase, whereas about 20 h later both a phase advance in tim transcript and a phase delay in PER and TIM protein were initiated. tim and per mRNA and protein rhythms are known to be functionally linked in the circadian cycle in a number of ways: (i) of course, PER and TIM protein are produced from per and tim transcript; (ii) TIM protein is known to stabilize PER protein by preventing DOUBLETIME-mediated phosphorylation and subsequent turnover of PER protein [31]; and (iii) PER and TIM feedback negatively on the expression of per and tim transcripts via PER-mediated inhibition of the CLK/CYC transcription factor [32]. As per transcript is the first to show temperature-mediated resetting, its response cannot be a secondary consequence of a phase shift in tim transcript or PER or TIM protein. Conversely, the delayed resetting responses of tim transcript or TIM and PER protein do not simply follow from the earlier changes in per transcript. Specifically, whereas per transcript shows a strong and early phase advance as well as an increase in amplitude, this is not tracked by the resetting profile for PER protein, which shows later phase delays and no obvious increase in amplitude. Instead, the PER protein profile resembles that of TIM protein during temperature-mediated re-entrainment. Indeed, this would be predicted based on the requirement for TIM to stabilize PER [24]. The delayed temperature-dependent phase advancing of tim transcript also does not obviously result from the changes in either per transcript or PER or TIM protein. Any effect of per transcript levels or TIM protein levels on tim transcript levels would be expected to be mediated via PER protein levels. However, PER protein levels initially reset with an extended trough phase, which would be expected to be associated with a delay in negative feedback on CLK/CYC activity, and therefore an associated increase in tim transcript levels rather than the decrease that is observed. Finally, TIM protein rhythms do not simply track tim transcript levels during temperature-mediated re-entrainment as it is difficult to explain how the phase-delaying pattern of TIM protein would result from the phase advancing of tim transcript or why TIM protein levels would peak prior to tim transcript levels during the second day. Taken together, exposure of wild-type flies to a 180° phase shift in their environmental temperature cycle appears to have separate effects on per transcript, tim transcript and TIM protein rhythms in the adult head (see the electronic supplementary material, figure S5). The diversity of temperature-mediated resetting responses among the four central molecular circadian rhythms analysed in this study suggests that the environmental temperature cycle interacts in a complex manner with the circadian clock mechanisms and impacts many points in the molecular circadian cycle.

(b). Hsp90^e6d/08445 and Tim 4 mutants both exhibit accelerated temperature-dependent resetting of behavioural rhythms, but only the latter show a similar effect on peripheral molecular rhythms

We examined the relationship between temperature-mediated re-entrainment of daily locomotor activity rhythms and daily gene expression rhythms in the adult head for two mutant genotypes (Hsp90^e6D/08445 and Tim 4) that showed faster thermal re-entrainment of daily locomotor rhythms (electronic supplementary material, figure S1). At the molecular level, per and tim transcript and protein rhythms in Hsp90^e6D/08445 heads before and during temperature-mediated resetting resembled those observed for w¹¹¹⁸ control flies. It therefore appears that the behavioural phenotype during temperature-dependent resetting of Hsp90^e6D/08445 flies could be attributable to clock neuron-specific changes in temperature entrainment or molecular oscillator function. By contrast, Tim 4 flies showed faster temperature-mediated resetting of both behavioural and peripheral molecular rhythms. Hence, it is likely that the altered state of the molecular circadian oscillator resulting from the manipulation of tim expression in this strain is responsible for the observed accelerated temperature re-entrainment of locomotor activity rhythms. In Tim 4 mutants, however, tim transcript rhythms, which normally exert a strong circadian drive in the oscillator, are lost and residual phase-shifted TIM oscillations are derived exclusively from post-transcriptional mechanisms that remain to be identified. Given the absence of introns in the transgenic Tim 4 construct, these mechanisms may act (post-)translationally rather than at the level of splicing. The phase-shifted and weakened TIM protein rhythms in Tim 4 flies may account for increased responsiveness of the oscillator to environmental temperature. Given TIM's role in controlling PER stability, rhythmic accumulation of TIM protein is thought to play a critical role in PER-dependent negative feedback [33]. Indeed, PER protein rhythms in Tim 4 heads are also less robust and show a trend that resembles the TIM profile. Only the stably entrained per transcript rhythm is clearly more pronounced in Tim 4 heads than in w¹¹¹⁸ or Hsp90^e6D/08445 flies, but this reflects temperature-driven rather than clock-dependent regulation. Whereas per transcript responds rapidly to temperature-mediated resetting in all genotypes examined and tim transcript appears temperature insensitive in Tim 4 flies, the TIM protein profile correlates well with the rate of temperature resetting across the examined genotypes. Thus, of the three separable molecular temperature responses identified in this study, the temperature-mediated regulation of TIM protein is of most direct relevance to the early temperature resetting phenotype of Tim 4 flies.

(c). A new mechanism for temperature-driven regulation of per transcript

In this study, per transcript was found to respond early and in a clock-independent manner to a phase-shifted environmental temperature cycle. Our study complements previous analyses of cold-induced per expression responses to single temperature steps or square wave temperature cycles in wild-type as well as arrhythmic mutant flies [2,34]. Our results indicated that temperature-dependent splicing at the 3′UTR of per, which contributes to upregulation of per transcript levels at colder temperatures [27], only partially explained the observed temperature-driven regulation of per. Moreover, temperature-driven regulation of per persisted in the absence of the native 3′UTR and was detected at the pre-mRNA level. Taken together, our analyses point to cold-induced transcription of per that is mediated by cis-regulatory elements residing in a 5627 bp region downstream of the transcription start site. Our findings match predictions from previous behavioural studies [4,35] suggesting that circadian temperature entrainment is at least partially independent of thermo-sensitive splicing of the per 3′UTR.

(d). Comparison of circadian temperature entrainment in Drosophila and mammals

Given the negative feedback exerted by HSP90 on HSF1 activity [36,37], one possible explanation of the accelerated temperature-dependent behavioural resetting observed for the Hsp90^e6D/08445 mutant genotype was that it would be associated with increased HSF activity, which by analogy with mammalian HSF1 could modulate clock gene expression in a temperature-dependent manner. Another apparent analogy with mammalian temperature entrainment is the temperature-driven regulation of per transcript levels. However, our analyses point to several differences between the molecular temperature entrainment mechanisms of Drosophila and mammals. First, unlike mouse Per2, Drosophila per levels are upregulated in response to decreased rather than increased temperatures. Thus, unlike its mammalian counterpart, Drosophila per is not co-regulated with HSF1 target genes. Second, no evidence was found to support a role for the HSF1/HSP90 in the temperature-controlled expression of Drosophila per as temperature-driven regulation of per persists in tim⁰¹; Hsp90^e6D/08445 mutants (see the electronic supplementary material, figure S4). Thus, while circadian temperature entrainment in both Drosophila and mammals involves transcriptional regulation of per genes, the underlying mechanisms are different.

4. Material and methods

(a). Fly strains

All stocks were maintained on standard yeast cornmeal agar food. Hsp90 mutant genotypes w¹¹¹⁸, w*;Hsp90^e6D/TM6B-Tb¹ and Hsp90⁰⁸⁴⁴⁵,ry⁵⁰⁶/TM3-ry^RK-Sb¹-Ser¹ were obtained from the Bloomington Stock Center. y w; tim⁰¹, Tim 4, y w tim-luc, y w BG-luc and y per⁰¹ w; per_P-Gal4 flies have all been described previously [17,25,30,38,39].

(b). Locomotor activity analyses

Adult flies were assayed for locomotor behaviour as described previously [4,40]. The transition day counts used in the electronic supplementary material, figure S1 were interpolated from individual actograms. The circadian behavioural period lengths in the electronic supplementary material, table S1 were determined by χ²-periodogram analysis (for 7-day intervals of 25°C DD free-running conditions). Individual flies lacking significant (p < 0.01) periodicity in the circadian range (15–35 h) were considered arrhythmic. Flies with circadian periodicity were designated as weakly rhythmic if their relative rhythmic power (determined by the ratio of the peak amplitude divided by the significance threshold) was 1.5 or smaller and rhythmic if their relative rhythmic power exceeded 1.5.

(c). Gene expression analyses

qRT-PCR was carried out as described previously [40] (for primer sequences, see the electronic supplementary material, table S5). Previously published protocols [2,4] were adjusted to accommodate use of a LI-COR Odyssey infrared imaging system for quantitative analysis. In particular, Immobilon-FL PVDF membrane Odyssey-blocking buffer and IRDye-conjugated secondary antibodies (used at 1/10 000 dilution) were substituted for nitrocellulose membrane, 5% non-fat dry milk 1× TBST blocking buffer and horseradish peroxidase-linked secondary antibodies, respectively. Polyclonal rabbit anti-PER and rat anti-TIM and monoclonal mouse anti-HSP70 (Sigma H 5147) were used as primary antibodies at 1/10 000 dilution. An internal control sample was included on each blot to allow for comparison of expression levels across blots.

(d). In vivo luciferase assays

Bioluminescence monitoring of files was carried out following published protocols [30]. Briefly, 100 μl of 5% sucrose 1% agar 0.07% Tegosept medium containing 15 mM luciferin (GOLDBIO) was added to every other well of a white 96-well microtitre plate (Optiplate, Perkin Elmer). Individual flies were added to each well and covered with a clear plastic dome. Plates were analysed in a TopCount NXT luminescence counter. Luminescence counts were analysed for period and relative amplitude error (RAE) by an iterative, coupled fast Fourier transform–nonlinear least squares (FFT-NLLS) [41] multicomponent cosine analysis using the BRASS software package. RAE corresponds to the ratio of the 95% confidence interval for the amplitude by the amplitude estimate. Individual fly profiles with RAE values of less than 0.7 were considered rhythmic, greater than or equal to 0.7 were considered weakly rhythmic, and those for which the program returned no data (RAE > 1) were considered arrhythmic.

(e). Statistical analyses

The Freeman–Halton extension of Fisher's exact test was used for pairwise comparisons for the effect of genotype on the distribution of rhythmic, weakly rhythmic and arrhythmic behavioural and molecular profiles in the electronic supplementary material, tables S1 and S2. The SPSS software package (IBM) was used to conduct analysis of variance (ANOVA) and Welch's test statistics combined with Games-Howell and Tamhane's T2 post-hoc analyses for gene expression data in figures 1 and 3, electronic supplementary material, figures S2 and S3, the number of transition days in the electronic supplementary material, figure S1C, the behavioural period length and relative rhythmic power in the electronic supplementary material, table S1, and the molecular period length and RAE in the electronic supplementary material, table S4. The assumption of homogeneity of variances was examined using Levene's test. Because this assumption was frequently violated, the reported p-values come from Welch's test (which is robust in this condition) rather than ANOVA. Both types of post-hoc analyses are also reliable in this context. The quoted p-value ranges are from Tamhane's T2, which gave slightly more conservative estimates. The time course transcript and protein expression data in figures 1 and 3, and electronic supplementary material, figure S3, were analysed separately for each genotype to determine the effect of time on gene expression. In the analyses for figure 1, each time point representing the entrained condition was treated separately, whereas in the analyses for the electronic supplementary material, figure S3, equivalent ZT time points collected at the same phase on consecutive days were pooled. The effect of genotype on the number of transition days for the data in the electronic supplementary material, figure S1BC was determined separately for each phase shift protocol–gender combination. The effect of genotype on average expression levels was determined for the electronic supplementary material, figure S2 with data pooled across all time points.

Data accessibility

Data used to create the figures and supplementary figures and tables are available via the Dryad repository: http://doi.org/10.5061/dryad.7747n.

Funding statement

Funding by the National Institutes of Health (R01 GM78339 to H.W.) and the University of Southampton contributed to this work.