Experimental animals

Adult flies, Drosophila melanogaster, were used. They were raised on standard cornmeal-glucose-yeast medium at 25°C under a light cycle with 12hr light to 12hr dark (LD12:12). The strains that we used for behavioral assay were Canton-S (wild-type), period⁰¹ (per⁰¹), timeless⁰¹ (tim⁰¹), dClock^Jrk (dClk^Jrk) and cycle⁰¹ (cyc⁰¹).

Recording of locomotor activity

Male flies of about five days old were individually housed in transparent acrylic rectangular tubes (3×3×70 mm). The tube was plugged at one end with agar/glucose medium as food and was sealed with Parafilm, and at the other end with a silicon tube filled with damped absorbent cotton as water source. A moving fly interrupted an infrared beam and the number of interruptions during each 6 min was recorded using a computerized system (Tomioka et al., 1997). The fly tubes and activity sensing system was placed in an incubator (Hitachi, CR32S) in which the light and temperature were controlled. Lighting conditions in the incubator were given by a cool white fluorescent lamp connected to an electronic timer. Light intensity at the animals level was 100 to 300 lx, varying with the proximity to the lamp. Temperature cycles were set by a built-in thermostat driven by an electronic timer. Temperature steps-up and -down were finished within 15 min. Temperature cycles used were composed of an equal duration of thermophase (30°C) and cryophase (25°C) with periods varying from 32 hr (T=32 hr) to 8 hr (T=8 hr) at an interval of 4 hr.

Data analysis

The raw activity data were displayed as conventional double-plotted actograms. For judgment of the activity pattern under different T values, actograms were drawn at the respective Zeitgeber period. The entrainment of activity of individual animals was objectively examined by the chi-square periodogram (Sokolove and Bushell, 1978) with trial periods around a given T. When the calculated value at a trial period equal to T exceeded the 0.05 confidence level, the animal was designated as entrained. Average daily activity patterns were calculated for entrained fly groups to compare the average waveform entrained to temperature cycles of different T values. To determine the phases of peaks of activity rhythms, moving average of 41 data points was performed. The free-running period was calculated by the chi-square periodogram (Sokolove and Bushell, 1978). Analysis of variance (ANOVA) was used to test whether significant differences among the time intervals between either onset of the thermophase or the cryophase and the activity peak were observed at different T values.

Locomotor rhythm of wild-type flies under temperature cycles

We first examined the entrainment of wild-type (Canton-S) flies to temperature cycles under LL and DD. The loco-motor activity rhythm of the flies was recorded under temperature cycles with T=32 hr, 28 hr, 24 hr, 20 hr, 16 hr, 12 hr and 8 hr consisting of an equal duration of thermophase (30°C) and cryophase (25°C). They showed clear entrainment to all temperature cycles under LL without evidence of frequency division or frequency demultiplication (Fig. 1). They showed a large primary peak in addition to small peaks associated with temperature transitions (Fig. 2). The phase of the major peak was apparently dependent on the period of temperature cycles (P<0.001, ANOVA). It occurred in the middle of the thermophase in T=32 hr, gradually delaying toward the cryophase as T was shortened to 20 hr. In T=20 hr or shorter, the peak always occurred in the cryophase (Fig. 2H). The results indicate that locomotor rhythms in the wild-type flies are driven by a circadian clock in temperature cycles even under LL. In DD, however, they showed free-running rhythms even under temperature cycles, except T=24 hr (Fig. 1). Free-running periods were close to 24 hr but substantially varied dependent on T, ranging from 23.8±0.02 hr (T=16 hr, n=8) to 24.9±0.2 hr (T=28 hr, n=8). In T=24 hr, they synchronized to the temperature cycle with a primary peak occurring at the late thermophase (11.3±0.2 hr after the onset of thermophase); the daily activity pattern was similar to that under LL.

Fig. 1

Double plots of representative actograms of wild-type flies recorded under temperature cycles of various periods, consisting of an equal duration of thermophase (30°C) and cryophase (25°C) in LL (right) and DD (left). Each actogram is double-plotted from top to bottom at the respective Zeitgeber period, which is shown, on the left of each row of the panels. White and black bars on the top of the actograms indicate light and dark, respectively. Boxes on the right half of the actograms indicate the thermophase. In DD, the locomotor rhythm was entrained only to temperature cycle of 24 hr (T=24hr) and free-ran in other Ts. In contrast, the rhythm synchronized with temperature cycles of all Ts under LL with the phase of the major peak changing dependent on T. For further explanations see text.

Fig. 2

A–G: Averaged and smoothed daily activity patterns of wild-type flies in LL that were calculated from frequency folded data at a period indicated by T. Shaded area indicates the thermophase. n indicates the number of flies entrained. Note that a primary peak occurred in the middle of the thermophase (arrow) in addition to small peaks at temperature transitions in T=32hr, gradually delayed to the middle to late cryophase as the temperature cycle was shortened. The time difference from the onset of thermophase and the primary peak changed dependent on T. H: Mean phase (±SEM) of major peaks in temperature cycles under LL. The phase of the primary peak occurring at the middle of thermophase in T=32hr gradually delayed toward the late cryophase as T was shortened.

Locomotor rhythm of clock mutants under temperature cycles in constant darkness

Thermoperiodic entrainability of arrhythmic mutant flies carrying per⁰¹, tim⁰¹, dClk^Jrk or cyc⁰¹ was examined by recording their locomotor activity in the temperature cycles under DD. Fig. 3 shows representative actograms. In all mutant flies, locomotor activity clearly synchronized with temperature cycles of various T values without evidence of frequency division or frequency demultiplication. dClk^Jrk and cyc⁰¹ flies showed a definite tendency to be more active in the thermophase than the cryophase. Activity increased immediately after the start of thermophase in all Ts (Figs. 3 and 4). The phase of activity peak always occurred at about 2.5 hr after the onset of thermophase (Fig. 5B), and there was no systematic dependency of the phase on T except T=28 hr in cyc⁰¹ (P>0.05, ANOVA).

Fig. 3

Double plots of representative actograms of per⁰¹, tim⁰¹, dClk^Jrk and cyc⁰¹ flies recorded under temperature cycles of various periods, consisting of an equal duration of thermophase (30°C) and cryophase (25°C) in DD. Actograms were plotted from top to bottom at the respective Zeitgeber period that was shown on the left of each row of the panels. Black bars on the top of the actograms indicate constant darkness. Boxes on the right half of the actograms indicate the thermophase. In all flies, the locomotor rhythm synchronized with temperature cycles in all Ts. Note that the phase of the activity changed dependent on the period of temperature cycles in per⁰¹ and tim⁰¹ flies, while in dClk^Jrk and cyc⁰¹ flies the active phase consistently occurred in the thermophase. For further explanations see text.

Fig. 4

Averaged and smoothed daily activity patterns of per⁰¹, tim⁰¹, dClk^Jrk and cyc⁰¹ flies in DD that were calculated from frequency folded data. Shaded areas indicate the thermophase. n indicates the number of flies used. Note that the activity phase was always in the thermophase in dClk^Jrk and cyc⁰¹ flies while, in per⁰¹ and tim⁰¹ flies, a major peak that occurred in the middle of the thermophase (arrow) in addition to small peaks occurring at temperature transitions in T=32 hr, gradually delayed to the middle to late cryophase as the temperature cycle shortened. The time difference from the onset of thermophase and the major activity peak was 6∼8 hr and a discernible trough (arrow head) occurred about 4 hr after the onset of cryophase in per⁰¹ and tim⁰¹ flies.

per⁰¹ and tim⁰¹ flies also entrained to the temperature cycles (Figs. 3 and 4); they were also more active during the thermophase than during the cryophase except for T=8 hr under which activity was concentrated in the cryophase. Average activity curves revealed a major peak that occurred 6∼8 hr after the onset of thermophase in addition to two rather small peaks associated with temperature transitions. The former gradually became to occur later relative to the temperature cycle as T was shortened, and eventually in the cryophase in T=8 hr. In real time, it occurred slightly but significantly earlier as the T values became smaller in per⁰¹ flies (Fig. 5A)(P<0.01, ANOVA), while, in tim⁰¹ flies, the peak occurred slightly later at T=32 hr, 24 hr and 8 hr (Fig. 5A)(P<0.01, ANOVA). Another point to be noted is that the per⁰¹ and tim⁰¹ flies showed a trough at about 4∼6 hr after the onset of cryophase in all Ts except T=8 hr (Figs. 3 and 4), where the trough occurred in the thermophase. The trough was more robust in per⁰¹ than in tim⁰¹ flies.

Fig. 5

Mean phase (±SEM) of major peaks in temperature cycles under DD (closed symbols) or LL (open symbols). Shaded areas indicate the thermophase. Ordinate indicates periods of temperature cycles and abscissa the time after the onset of thermophase (A, B and D) or cryophase (C). In per⁰¹ and tim⁰¹ flies, primary peak occurs about 6∼8 hr after the onset of thermophase in DD (A) but about 2∼3 h after the onset of cryophase in LL (C). The major peaks of dClk^Jrk and cyc⁰¹ occurred consistently about 2∼3 hr after the onset of thermophase both in DD and LL (B and D). For further explanations see text.

To clarify whether the rhythmic patterns of the mutant flies are direct response to temperature cycles, we transferred the flies from temperature cycles to constant temperature and vice versa. Upon transfer from temperature cycles to constant 25°C, all 4 mutant flies became arrhythmic after showing a trough associated with a transition from the thermophase to the cryophase (Fig.6). The activity level at constant temperature was nearly intermediate between the peak and trough in the temperature cycle. When transferred from constant 25°C to temperature cycles of 25°C 12hr : 30°C 12 hr (T=24 hr), dClk^Jrk and cyc⁰¹ exhibited a large activity peak in the first thermophase that persisted thereafter (Fig. 7), whereas tim⁰¹ and per⁰¹ became to show clear entrainment at 2nd or 3rd cycle, respectively. Interestingly, in per⁰¹ a trough occurred around 4∼6 hr after the onset of cryophase on the first cycle, becoming more prominent day by day to reach at the least level on the third cycle. A similar but less obvious tendency was also observed in tim⁰¹ flies.

Fig. 6

Averaged longitudinal activity of per⁰¹, tim⁰¹, dClk^Jrk and cyc⁰¹ flies recorded consecutively in temperature cycle of 30°C 12 hr: 25°C 12 hr and then in constant 25°C under DD. Shaded areas indicate the thermophase. n indicates number of flies used. On transfer to constant 25°C, activity rhythms disappeared after showing a trough associated with a transition from 30°C to 25°C in all flies. For further explanations see text.

Fig. 7

Averaged longitudinal activity of per⁰¹, tim⁰¹, dClk^Jrk and cyc⁰¹ flies recorded consecutively in constant 25°C and then temperature cycle of 30°C 12 hr: 25°C 12 hr under DD. Shaded areas indicate the thermophase. On transfer to temperature cycle, a robust peak immediately appeared in the first thermophase in dClk^Jrk and cyc⁰¹ flies, while at least two or more cycles were necessary for clear entrainment in per⁰¹ and tim⁰¹ flies. Arrows and arrow heads indicate a primary peak and a trough, respectively. For further explanations see text.

Locomotor rhythm of clock mutants under temperature cycles in constant light

The locomotor activity of arrhythmic mutant flies was recorded in temperature cycles under LL. Fig. 8 shows representative actograms. Again, all mutant flies clearly showed activity rhythms synchronized with temperature cycles without showing evidence of frequency division or frequency demultiplication. In dClk^Jrk and cyc⁰¹ flies, the activity patterns were quite similar to those under DD (Fig. 9): the peak occurred about 2∼3 hr after the onset of thermophase, and there was no dependency of peak phase on T for both dClk^Jrk and cyc⁰¹ flies (Fig. 5D) (P>0.05, ANOVA). However, per⁰¹ and tim⁰¹ flies reversed from thermoactive in DD to cryoactive in LL (Figs. 5C and 8), exhibiting a large peak about 2∼3 hr after the onset of cryophase. The peak phase was significantly dependent on T in tim⁰¹ flies (Fig. 5C) (P<0.01, ANOVA). The activity in the thermophase was remarkably reduced especially in per⁰¹ flies. Interestingly, tim⁰¹ flies showed a small peak about 6 hr after the onset of thermophase, which seemed identical to the one occurred under DD.

Fig. 8

Double plots of representative actograms of per⁰¹, tim⁰¹, dClk^Jrk and cyc⁰¹ flies recorded under temperature cycles of various periods, consisting of an equal duration of thermophase (30°C) and cryophase (25°C) in LL. Actograms were plotted from top to bottom at the respective Zeitgeber period that was shown on the left of each row of the panels. White bars on the top of the actograms indicate constant light. Boxes on the right half of the actograms indicate the thermophase. In all flies, the locomotor rhythm synchronized with temperature cycles in all Ts. Note that the active phase is antiphase to that in DD (see Fig. 3) in per⁰¹ and tim⁰¹ flies. For further explanations see text.

Fig. 9

Averaged and smoothed daily activity patterns of per⁰¹, tim⁰¹, dClk^Jrk and cyc⁰¹ flies in LL that were calculated from frequency folded data. Shaded areas indicate the thermophase. n indicates the number of flies used. The major peaks consistently occurred in the cryophase in per⁰¹ and tim⁰¹ flies, and in the thermophase in dClk^Jrk and cyc⁰¹ flies. A small peak (arrow) appeared in the middle to late thermophase in T=20 hr or longer in tim⁰¹ flies. For further explanations see text.

Entrainment to temperature cycle in wild-type flies

The results of the present study, not only confirms but also extends the previous reports that wild-type flies of Drosophila melanogaster can be entrained to temperature cycles in DD (Wheeler et al., 1993; Tomioka et al., 1998). However, the entrainment was achieved only to the cycle of T=24 hr, and the flies showed free-running rhythm with a period close to 24 hr in Ts shorter or longer than 24 hr. The result indicates that temperature cycle is a much weaker entraining agent than light because light can entrain the rhythm in a wide range of Ts (Helfrich-Förster, 2001). Probably, temperature pulses cause only small phase-shifts in free-running flies. This explanation is also supported by the fact that, in DD, per^S and per^L flies also entrained to the temperature cycles with periods close to their natural free-running periods (Tomioka et al., 1998).

Under LL, in contrast to under DD, wild-type flies never free-ran but showed activity rhythms clearly synchronized with the given temperature cycles with wide range of Ts, without evidence of frequency division or frequency demultiplication. It has been shown that LL stops the circadian oscillation not only behavioral but also molecular level (Price et al., 1995). The rhythm observed in the present study has characteristics full-filling the empirical rule for the one driven by the circadian oscillation, however: the phase of a primary activity peak changed dependent on the period of temperature cycles. The fact supports our previous hypothesis that temperature cycle can drive the circadian oscillation even in LL where the clock is stop in constant temperature (Tomioka et al., 1998) and suggests that temperature cycle entrains the rhythm through the pathway different from that for photic entrainment. The locomotor rhythm in Drosophila is believed to be driven by a circadian oscillation generated by an auto-regulatory feedback loop including rhythmic expression of per and tim at a period of about 24 hr (Reppert and Weaver, 2000). Their product proteins, PER and TIM, form heterodimer, enter the nucleus and inhibit their own transcription through inactivation of their transcription factors, dCLK and CYC, products of dClk and cyc (Williams and Sehgal, 2001). Western blot analysis using anti-PER antibody with proteins extracted from fly heads revealed an oscillation in PER abundance at least in T=24 hr and 32 hr under LL (Ibuki, M., et al., unpublished data), suggesting that the temperature induced rhythm in LL involves the rhythmic expression of per probably through the autoregulatory feedback loop. An important question that should be addressed in a future study is how temperature drives the oscillation once stopped by constant light. It is also to be answered whether the same molecular oscillation that works in LD or DD (cf. Williams and Segal, 2001) is working in temperature cycle with Ts considerably shorter or longer than 24 hr.

Thermoperiodically induced rhythms in mutants lacking the per-feedback loop

The present study revealed that all mutant flies carrying mutation in clock genes, which induces arrhythmicity in constant conditions, exhibited rhythms in thermoperiodic conditions. However, the rhythms were quite different from those of wild-type flies in respect to that they never free-run in DD in Ts longer or shorter than 24 hr and that their peak phase was rather stable over a wide range of Ts, suggesting that the underlying mechanism differs from that for wild-type flies.

The rhythmic pattern was remarkably different between mutant groups lacking so-called negative and positive components. In dClk^Jrk and cyc⁰¹ flies lacking either of the positive components, dCLK or CYC (Allada et al., 1998; Rutila et al., 1998), the thermperiodically induced rhythms seemed to be direct response to temperature cycles. This statement is based on the following facts. Their rhythmic patterns were always quite similar through temperature cycles with various Ts both under DD and LL, with peaks always appearing about 2 hr after the onset of thermophase, and with the active phase always being confined in the thermophase.

In contrast, per⁰¹ and tim⁰¹ flies lacking either of the negative components, PER or TIM, changed from thermo-active to cryoactive in shorter Ts in DD. The peak phase was also found to change slightly but statistically significantly dependent on the period of temperature cycles. These facts suggest that their activity rhythms are not a direct response to temperature cycles but driven by an endogenous timing mechanism. The fact that the primary peak consistently occurred around 6 hr after the transition from the cryophase to the thermophase in all Ts suggests that the timing mechanism seems to be reset by the transition and set the timing of the peak phase. The involvement of circadian oscillation in per⁰¹ and tim⁰¹ flies is further suggested by the results of transfer of flies from constant temperature to temperature cycles. The activity peak as well as trough never occurred immediately after the transfer but gradually became prominent after several transient cycles. The transients may be explained as the period during which an underlying weak oscillator synchronizes with the given cycle. The oscillator seems to preserve the character of highly damped oscillator since the rhythm almost immediately disappeared when transferred from temperature cycles to constant temperature.

The involvement of per-less circadian oscillation in loco-motor rhythms has also been demonstrated for the per⁰¹ flies entrained photoperiodically (Helfrich and Engelmann, 1987; Helfrich-Förster, 2001). The oscillation demonstrated here occurred when exposed to thermoperiods. It is an interesting question whether these oscillations are based on the same mechanism. Since the thermoperiodically-induced oscillatory components were observed in the flies lacking either PER or TIM but not in flies lacking the positive components, it seems likely that dClk and cyc are somehow involved in the thermoperiodically-induced oscillation. A similar thermoperiodic entrainment of the circadian oscillation was recently reported for the frq⁹ mutant of the bread mold Neurospora crassa (Merrow et al., 1999). Although frq⁹ mutants are arrhythmic under constant conditions (Aronson et al., 1994), they exhibit clear oscillations in their conidiation under temperature cycles with aspects of typical circadian entrainment. The frq-independent oscillation is explained to be driven by a temperature entrainable oscillator called FLO that involves metabolic components (Iwasaki and Dunlap, 2000; Morgan et al., 2001). The relationship between the Neurospora frq-independent oscillation and the Drosophila-thermoperiodically induced oscillation is an interesting issue to be addressed in future studies.

The thermoperiodically induced oscillation was also reported for cockroaches and crickets (Rence and Loher, 1975; Page, 1985). Their circadian clocks driving the loco-motor rhythm reside in the optic lobe. Even after the optic lobes are bilaterally removed, the locomotor activity can be entrained to the temperature cycles with some circadian characteristics. In cockroaches Leucophaea maderae, similarly to D. melanogaster, the rhythms were driven by cycles with wide range of Ts (12∼48 hr) but with predictable T-dependent changes in active phase (Page, 1985). It seems thus a rather general scheme that the circadian system is composed of a temperature entrainable secondary oscillator in addition to the core oscillator involving the autoregulatory feedback loop. Probably the temperature entrainable oscillator is driven by the light entrainable circadian clock in the light of the master-slave organization proposed by Pittendrigh et al. (1958) long time ago.

Light-dependent reversal of the active phase in per⁰¹ and tim⁰¹ flies

The results of the present study demonstrated that the active phase of per⁰¹ and tim⁰¹ flies showed a reversal from the thermophase under DD to the cryophase under LL in a wide range of Ts. This confirms our previous reports for the light-dependent reversal of active phase in per⁰¹ flies in 24 hr temperature cycles (Tomioka et al., 1998). There are multiple photoreceptors known for entrainment of circadian rhythms in Drosophila: they are the compound eyes, ocelli, Hofbauer-Buchner's (H-B) eyelet and a deep brain blue-light receptor, cryptochrome (Yasuyama and Meinertzhagen, 1999; Helfrich-Förster et al., 2001). These photoreceptors cooperate to synchronize the circadian locomotor rhythms. Since the photic information for the light-dependent reversal of active phase is abolished in eya;per⁰¹ and so;per⁰¹ double mutant flies, which lack compound eyes and both compound eyes and ocelli, respectively, but retain the H-B eyelet and cryptochrome, the photic information necessary for this reversal is most likely received by the compound eyes as suggested previously (Tomioka et al., 1998). Interestingly, dClk^Jrk and cyc⁰¹ flies did not show the reversal. Although it is to be answered how dClk and cyc mediate the light-dependent reversal of temperature preference in per⁰¹ and tim⁰¹ flies, the photic pathway seems to be disrupted in dClk^Jrk and cyc⁰¹ flies.