Resetting the Circadian Clock by Social Experience in Drosophila melanogaster

See allHide authors and affiliations

Science  06 Dec 2002:
Vol. 298, Issue 5600, pp. 2010-2012
DOI: 10.1126/science.1076008


Circadian clocks are influenced by social interactions in a variety of species, but little is known about the sensory mechanisms underlying these effects. We investigated whether social cues could reset circadian rhythms in Drosophila melanogasterby addressing two questions: Is there a social influence on circadian timing? If so, then how is that influence communicated? The experiments show that in a social context Drosophilatransmit and receive cues that influence circadian time and that these cues are likely olfactory.

Circadian clocks in animals regulate the timing of molecular, physiological, and behavioral rhythms. Environmental features such as photoperiod and temperature cycles reset these biological oscillators, enabling anticipation of dawn, dusk, and season (1–6). Other kinds of cues (“nonphotic”) also influence clock time (7). For example, studies on humans (8), rodents (9), fish (10), and bees (11) have demonstrated social influences on rhythmicity, but underlying sensory mechanisms remain unexplained. It is nonetheless clear that multiple sensory pathways transmit ambient temporal information from the periphery to clock cells in the brain (7).

We investigated social influence on circadian timing in the fruit fly Drosophila melanogaster. We initially hypothesized that the circadian phases [marked by the peak of locomotor activity in DD (constant darkness)] would be more coherent for Drosophilaliving together (group-housed) than those of isolates, because groups of flies might agree about the time of day even without photic cues. Locomotor activity rhythms from group-housed wild-type individuals were compared to those of sibling isolates. After an initial 5 days in 12 hours of light and 12 hours of dark (LD 12:12), isolates and group-housed subjects were maintained for 2 weeks in DD. Isolates were then placed in activity monitors, whereas the group-housed flies were separated and monitored in DD to assess the effects on individual rhythmicity (12).

The effect of this treatment on phase coherence was analyzed with the use of circular statistics (Fig. 1A) (13, 14). The resulting vector angle indicates the mean peak time for each group, and its magnitude indicates phase coherence, with longer tails denoting a tighter distribution of phase estimates around the day (0, no correlation; 1, perfect correlation) (13, 14). The difference in phase coherence was significant (P = 0.02), and there was no effect on phase angle (timing) (P = 0.64), suggesting that the clocks of group-housed individuals in DD are more synchronized than those of isolates (12).

Figure 1

Phase coherence of Drosophilalocomotor activity rhythms is affected by housing and genotype. (A) Group housing increases phase coherence. Phase analysis of locomotor activity rhythms is on the basis of individual records in constant darkness (13, 14). Mean phase estimates for the isolates (asterisks; n = 58) as well as for the group-housed subjects (open circles; n = 87) are plotted in a 24-hour dial. (B) Arrhythmicper0 visitors disperse the phase of wild-type hosts. Open circles represent phase estimates for hosts (n = 34); asterisks indicate wild-type controls (n = 9).

This finding implies that housing arrhythmic mutant individuals (visitors) together with wild-type hosts could destabilize host phase. After 5 days in LD 12:12, we placed a group of wild-type controls (40 per vial) or a group of wild-type hosts plus arrhythmicper0 visitors (32 plus 8, per vial) in DD for 5 days (12) (per0 is a loss-of-functionperiod mutation causing locomotor arrhythmicity). Locomotor activity was then assessed individually in DD for these two groups.

There was a large effect of social experience on the phase of the hosts (Fig. 1B). Wild-type hosts joined by per0 visitors showed dispersed phase coherence (P < 0.02) and mean peak time (P < 0.01) as compared to controls. This further suggests an interaction between circadian clock function and social experience, because coherence and phase (and also strength; fig. S1) of locomotor activity rhythms are influenced by the genotypic characteristics of the biological clocks (or lack thereof) within the social mix (12).

We next expanded our question to ask whether phase among hosts could be influenced by visitors from another “time zone” (12). Two LD 12:12 cycles with the start of the light phase (lights-on) occurring 6 hours apart were established, with “early” and “late” control individuals housed in vials in the respective incubators for 5 days. On the fifth day, visitors (8 per vial) from one of the incubators were mixed with hosts (32 per vial) from the other incubator 9 hours after lights-on in the early incubator (3 hours after lights-on in the late incubator). The controls and mixed groups were placed immediately in DD for another 5 days, and activity was individually monitored in DD for 5 days thereafter.

Here (Figs. 2 and3), the analysis was extended to include stability and timing of individual peaks for each animal. Points correspond to estimates of mean phase and its variability for an individual subject. The arrows shown in Figs. 2 and 3 summarize each group of points; length describes the mean estimate of the dispersion of the peak phase, and direction indicates mean timing of occurrence of the daily peak in locomotor activity. A significant overall difference between the vectors can result from differences in phase angle, phase coherence, or a combination of the two; thus, this analysis is more general and more conservative than the method used in Fig. 1, because it is not linked to a hypothesis about effects on phase coherence or mean phase time (12).

Figure 2

The interaction between flies from different “time zones” alters circadian phase. (A) Open circles show late controls (n = 16); asterisks mark early controls (n = 16). (B) Early control individuals (asterisks; n = 16) were compared with early hosts (open circles; n = 47). (C) Late control individuals (open circles; n = 16) were compared with late hosts (asterisks; n = 42).

Figure 3

Effects of perS visitors on perL hosts depend on the host:visitor ratio. Controls (asterisks; n = 17 in all panels) andperL hosts (open circles) (A) at a 4:1 ratio (n = 27), (B) at a 2:1 ratio (n = 19), and (C) at a 1:1 ratio (n = 17) were compared.

Figure 2 depicts comparisons between early and late control individuals, early controls and early hosts, and late controls and late hosts. After 5 days in DD followed by behavioral monitoring, early and late control groups maintained a mean difference of 4.2 hours (a reduced difference from 6 to 4.2 hours presumably stems from variability among the individual free-running circadian clocks during DD). There were significant differences between the early and late controls (Fig. 2A; P < 0.01), a weaker, nonsignificant effect of the late visitors on the early hosts ( Fig. 2B;P = 0.12), and a significant effect of the early visitors on the late hosts (Fig. 2C; P < 0.01). The effect is directional: Early visitors affected phase among late hosts, whereas late visitors did not affect early hosts as strongly, implying that social influence on locomotor rhythms depends on when (subjective time) the stimulus is provided. Time-varying responses underlie clock resetting in general (5, 6), suggesting that phase response curves could be developed for these social interactions (7, 9).

Would a similar effect on phase occur between mutant strains? A fast-clock (perS ) and a slow-clock (perL ) period mutant display, respectively, advanced and delayed evening activity peaks in LD 12:12 (15). Experimental design was the same as in the previous experiment, except only one incubator and one LD cycle were used. For the studies with per0 (Fig. 1B) and early and late visitors (Fig. 2), we maintained a ratio of 32 hosts to 8 visitors or 4:1. Here, we examined the effects of 2:1 (27 hosts:13 visitors) and of 1:1 ratios, maintaining a constant number of 40 flies per vial in DD, to assess whether the size of a social subgroup matters. Analysis for hosts versus control (perL alone) was the same as inFig. 2.

Consistent with our results from early visitors and late hosts (of similar genotype), perS visitors had a significant effect on perL hosts at ratios of 4:1 (Fig. 3A;P = 0.01) and 2:1 (Fig. 3B;P = 0.04) but had a weaker (nonsignificant) effect at 1:1 (Fig. 3C; P = 0.1). In addition, and consistent with the previous early-versus-late experiment, there was no effect of perL visitors on theirperS hosts at any ratio. These experiments indicate that the composition of the social group plays a role in the communication of timing signals.

What sensory mechanisms underlie social effects on circadian function? Visual or thermal cues are unlikely; social interactions and locomotor activity measurements took place in darkness and at constant temperature. Studies on courtship in Drosophila demonstrated that communication between flies may occur over a short distance by means of endogenous volatile chemical signals (16,17). We asked whether chemosensory signaling could synchronize the phase of circadian activity in wild-type flies. Humidified air was pumped through either a vial containing food plus 10 to 15 wild-type flies (“fly air”) or a similar vial without flies (“neutral air”). Both vials were simultaneously maintained in a 24-hour LD cycle at 25°C. Outflow from each vial reached individuals in an otherwise completely isolated activity monitor such that half received fly air and half received neutral air. Individuals receiving fly air were synchronized, whereas those receiving neutral air were more dispersed (P = 0.05; Fig. 4A). This indicates that chemical signals generated by wild-type Drosophila can synchronize individuals maintained in constant darkness and further suggests a cue that is rhythmically produced and short-lived.

Figure 4

Chemosensory signaling mediates social influence on circadian timing. (A) The peak phases of locomotor activity rhythms generated by isolated wild-type individuals receiving neutral air (asterisks; n = 13) are significantly more dispersed than those generated by individuals receiving fly air (open circles; n = 8). (B) Disruptive effects ofper0 visitors are not evident on the peak phase of olfactory mutants. The parasbl-2 controls (open circles; n = 26) are indistinguishable from sibling hosts (asterisks; n = 15). (C) Effects of per0 visitors are not evident on the peak phase of per-7.2 transgenic hosts (open circles;n = 32); transgenic controls indicated by asterisks (n = 25).

We used the allelic olfactory mutants parasbl-1 (18) and parasbl-2 (19) to ask whether the sense of smell might be involved. The para locus encodes a voltage-gated sodium channel, and the sbl alleles produce generalized deficits in olfactory responses (19, 20), including to odors emanating from other flies (21, 22). Individuals from these strains were capable of detecting sucrose and light on the basis of assays of gustation and phototaxis (12, 23). Hypothetically, if smell detects the timing signal, thenper0 visitors would not disrupt phase in the mutants as they disrupt phase in the wild type (Fig. 1B). There was no effect of arrhythmic visitors on the phase of locomotor rhythms inparasbl-2 hosts (P = 0.4; Fig. 4B),parasbl-1 (P = 0.3), contrary to wild-type visitors in a control experiment (P = 0.02).

We cannot rule out parallel involvement of auditory or tactile cues, but because olfactory mutant responses are not disrupted byper0 , olfaction is likely required for the social effect observed in the wild type. However, there is another possibility: Although clock cells in the central brain regulate locomotor activity rhythms, autonomous circadian clocks reside in a variety of Drosophila tissues, including several associated with sensory structures (24). We considered whether temporal regulation of sensory input (as opposed to the input per se) might be required for effective social communication. Accordingly, we employed a transgenic strain, per-7.2 (12), in which per+ expression is restricted to certain clock neurons within the central brain (25). Behavioral rhythmicity is normal in this strain (25), but temporal regulation of responses to odors by the antennal nerve is eliminated (26). Figure 4C shows no disruptive influence ofper0 visitors on the phase of per-7.2hosts (P = 0.5; the wild-type positive control as above,Fig. 4B). This suggests that recognition of a social cue depends on the temporal control of sensory input by peripheral clocks.

These findings show that circadian clocks may be reset by social communication in Drosophila; that this communication may reflect genotype, experience, and composition of the group; and that the mechanism underlying these effects is likely chemosensory. In addition, this mode of social communication appears to rely on the distributed property of the circadian system, whereby temporal gating of peripheral sensory input informs the central clock–controlled regulation of behavior. Finally, neural mutations and gene manipulations in Drosophila can now be used to dissect social rhythm–regulating interactions.

Supporting Online Material

Materials and Methods

Fig. S1

References and Notes

  • * To whom correspondence should be addressed. E-mail: hall{at}


Stay Connected to Science

Navigate This Article