A Mechanosensory Pathway to the Drosophila Circadian Clock

See allHide authors and affiliations

Science  31 Jan 2014:
Vol. 343, Issue 6170, pp. 525-528
DOI: 10.1126/science.1245710

Coordinating the Clock

In flies, the mechanosensory chordotonal organs help to coordinate the effects of temperature on circadian cycles. Simoni et al. (p. 525) provide a mechanism by which mechanosensory input is processed to help to synchronize the biological clock in Drosophila melanogaster. The chordotonal organs, which have similarities to the mammalian ear, were also required for sensation of a vibration stimulus and its effects on the endogenous brain clock. The chordotonal organs, present in the joints of the limbs, provide neuronal signals that allow the animal to sense its position or posture—and thus might mediate feedback of a range of behaviors onto the endogenous biological clock.


Circadian clocks attune the physiology of virtually all living organisms to the diurnal cycles of their environments. In metazoan animals, multiple sensory input pathways have been linked to clock synchronization with the environmental cycle (entrainment). Extrinsic entrainment cues include light and temperature. We show that (12-hour:12-hour) cycles of vibration and silence (VS) are sufficient to synchronize the daily locomotor activity of wild-type Drosophila melanogaster. Behavioral synchronization to VS cycles required a functional clock and functional chordotonal organs and was accompanied by phase-shifts of the daily oscillations of PERIOD protein concentrations in brain clock neurons. The feedback from mechanosensory—and particularly, proprioceptive—organs may help an animal to keep its circadian clock in sync with its own, stimulus-induced activities.

The neurocellular network that adjusts an organism’s physiological needs to the diurnal fluctuations of its environment is summarily referred to as the “circadian clock” (1). The tasks associated with the operation of circadian clocks are computationally challenging. In metazoan animals, clock synchronization requires integration of inputs from different sensory modalities, of which light and temperature changes provide major cues.

Chordotonal organs (ChOs) have been linked to temperature entrainment of the circadian clock in adult flies (2). ChOs are internal mechanoreceptors mediating proprioception and the detection of air- and substrate-borne vibrations (3, 4). If signaling from ChOs provides sensory input for the entrainment of the fly’s circadian clock to temperature cycles (2), we reasoned that exposure to a rhythmic mechanical stimulus (Fig. 1, A and B, stimulus details) that excites the fly’s ChOs (fig. S1, response details) might phenocopy temperature entrainment (5) and be sufficient to synchronize the clock and clock-controlled locomotor behavior. To test this, we first entrained adult wild-type flies to 12-hour:12-hour light-dark (LD) cycles (6). Flies were then transferred to constant darkness (DD) and constant temperature (7). One group remained in silence to serve as controls (Fig. 1C, top), whereas a second group was exposed to 12-hour:12-hour vibration:silence (VS) cycles (Fig. 1C, bottom). In the first, 4-day-long VS regime (VS1), vibration onset was delayed by 6 hours from light onset in the preceding LD cycles (L+6h). In a second, 5-day-long vibration regime (VS2), vibration onset was then delayed by another 6 hours (thus now L+12h). At the end of VS2, the flies were released into the final free running (FR) conditions—that is, darkness and silence—in which they were kept for another 5 days (Fig. 1C).

Fig. 1 Synchronization of Drosophila locomotor activity by VS cycles.

(A) Experimental set-up showing a Drosophila activity monitor (DAM) mounted on top of a bass loudspeaker (BLS). Vibrations were monitored by using an accelerometer (ACC). (B) Example of vibratory stimulus sequence played in loop (bottom, acceleration; middle, velocity; top, displacement). (C) (top) Locomotor activity (actogram) of wild-type (CantonS; n = 15 flies) control flies that were exposed to initial 12-hour:12-hour, light-dark (LD) cycles, followed by complete darkness (DD). (bottom) Actogram of experimental flies (CantonS; n = 15 flies) that were exposed to initial 12-hour:12-hour LD cycles, followed by two phase-delayed 12-hour:12-hour VS cycles (VS1-VS2) in DD, followed by DD without VS (free-run, FR). Gray areas, darkness; yellow areas, light; white areas, vibration in darkness. The final 5 days in DD were used to compare the phase of the peak activity during FR between experimental and control groups. Circadian reference time is given relative to the initial LD entrainment (Lights ON = 0). (D) Histograms showing the daily activity averages during three different phases for (bottom) experimental flies (LD, VS2, and FR, n = 75 flies) and (top) control flies (LD, DD, and FR, n = 72 flies). Error bars represent SEM.

During VS1, wild-type flies showed an initial activity peak after vibration onset, which decreased throughout the remaining vibration part (Fig. 1C and fig. S2). During VS2, flies again showed increased activity immediately after vibration onset, which declined rapidly (Fig. 1, C and D, and fig. S2). In contrast to VS1, flies now also exhibited increased activity several hours before vibration onset, which is reminiscent of anticipatory behavior in LD cycles (Fig. 1D and fig. S2). To quantify the behavioral activity occurring before vibration onset (the anticipatory activity component), we determined the ratio of the activity in the 4-hour time window before the V-phase and the total activity in the S-phase [compare with (8)]. The resulting entrainment index (EI) revealed that anticipatory activity was significantly increased (Fig. 2B). To further probe whether the activity patterns during the VS cycles resulted from a clock-controlled synchronization of behavioral activity, as indicated by the EI calculation, we conducted a phase analysis of the activity peaks in the final FR conditions between flies exposed to VS cycles and controls (fig. S3) (7, 9). The FR activity peaks were in phase with those of the last VS cycle, demonstrating that the circadian clock driving these rhythms had indeed been stably synchronized (Figs. 1, C and D, and 2C). In the control group, activity peaks free-ran from the synchronized phase set during the initial LD cycle and hence occurred significantly earlier (mean difference, 4.9 hours; P < 0.001, Watson-Williams-Stevens test) (Figs. 1, C and D, and 2C; fig. S3; and table S1) than those of the experimental group.

Fig. 2 Requirement of a functional clock and chordotonal organs for synchronization to VS cycles.

(A) Average locomotor actograms (top, control flies; bottom, experimental flies) for per01, per01 13.2, tilB1, and nocteP mutant flies. Quantification is provided in (B) and table S1. Stimulus sequence and actogram shadings are provided in Fig. 1. (B) Entrainment Index: quantification of behavioral activity anticipating the V-onset as a measure for entrainment. Red dotted line indicates random distribution of activity in the 4-hour window (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, one-way analysis of variance) (7). (C) Average locomotor activities for final 5 days in FR conditions of experimental flies exposed to VS cycles (vibrated, red) and control flies not exposed to VS cycles (FR, gray). Solid lines represent sinusoidal fits to the average locomotor data (circles, control flies; triangles, experimental flies). To facilitate appreciation of phase differences, both data and corresponding fit values have been normalized to the maximum of the fit function. The arrhythmic activities of per01 flies have been normalized to their respective mean value. Raw behavioral data of all genotypes in (C) are provided in (A), Fig. 1C, and figs. S4 and S7. CantonS, n = 75 flies (vibrated), n = 72 flies (FR); CantonS without antennae: n = 28 flies (vibrated), n = 30 flies (FR); per01: n = 32 flies (vibrated), n = 32 flies (FR); per01 13.2: n = 21 flies (vibrated), n = 22 flies (FR); tilB1: n = 24 flies (vibrated), n = 25 flies (FR); nocteP: n = 16 flies (vibrated), n = 16 flies (FR). Error bars represent SEM.

Not all flies synchronized their activity to the vibration cycles (fig. S4). We therefore assessed each fly’s synchronization by inspecting individual actograms without any knowledge about the experimental treatments of the particular fly under investigation. This “observer-blind” analysis revealed that across eight independent experiments, ~53% of all flies (n = 312) synchronized to vibration cycles (table S1). The reasons for this incomplete synchronization are unclear. The vibration stimulus used across our experiments was chosen for its experimental reliability [easy quantifiability and reproducibility (7)] and was not optimized for behavioral efficacy. Two-frequency vibration as used in this study may only excite a certain fraction of the animals’ ChOs, and other, spectrally more complex stimuli might prove behaviorally more efficient. Indeed, ablating the animal’s antennae, which changes the stimulus perceived by the flies, was already sufficient to increase synchronization rates to ~74% (Fig. 2, B and C; fig. S5; and table S1).

To test whether behavioral entrainment to VS cycles requires a functional clock, we performed the above experiment in the background of the per01 mutation—a loss-of-function allele of the central clock gene period (10). On average, per01 mutant flies displayed higher activity during the silent phase and lower activity during the vibration phase (fig. S6), but individual actograms revealed a “noisy” activity pattern during the entire vibration part compared with constitutively high or low activity in silence (fig. S4). The per01 flies did not show anticipatory behavior during VS cycles (fig. S6), suggesting that this anticipation requires a functional clock. We tried to rescue the VS entrainment in per01 flies by introducing a period construct (per01 13.2) that restores behavioral and molecular rhythms (8, 11, 12). EI calculation and observer-blind actogram classification showed that per01 13.2 flies synchronized to the VS cycles (Fig. 2, A and B; figs. S4 and S6; and table S1) with the phase difference between the VS-exposed and control flies differing significantly (P < 0.01) (fig. S3). Although on average per01 13.2 flies show a 24-hour period (table S1), the silent control flies do show a lengthened period during the part of the experiment corresponding to VS2 (Fig. 2A, DD days 8 to 12). They thereby acquire a later phase as compared with that of the VS-exposed flies before entry into the final 5 experimental days (used to calculate the phase differences between silence and VS-exposed flies); as a result, the phase relation between the activity curves of experimental and control flies is reversed as compared with that of CantonS flies (Fig. 2C and fig. S3).

To explore the molecular requirements of vibration-dependent entrainment, we tested flies with mutations in tilB (13) and nocte (14), two genes important for both temperature synchronization of the fly’s circadian clock and structural integrity of ChOs (2, 13). Flies carrying either the loss-of-function allele tilB1 or the hypomorphic allele nocteP failed to synchronize to vibration cycles and instead free-ran throughout the experiment (Fig. 2, figs. S5 and S7, and table S1).

If VS cycles synchronize clock-controlled behavior, they should also synchronize the oscillations of clock gene products in the neurons driving this behavior. We therefore compared the FR bioluminescence oscillations in brain clock neurons of two groups of flies initially synchronized to the same LD cycle and expressing a PER-LUCIFERASE (PER-LUC) fusion protein in subsets of their clock neurons (8, 15): experimental flies that had been exposed to 12-hour:12-hour VS cycles and control flies that had not been exposed to VS cycles (Fig. 3). Both experimental and control flies displayed circadian oscillations of PER-LUC protein levels in constant conditions (τ = 22.7 ± 0.2 hours) (Fig. 3). In flies exposed to VS cycles, however, the phase of the molecular oscillation was shifted compared with that of flies kept in silence. The observed molecular phase shift was sensitive to the phase of the VS entrainment regime: If the VS cycles were delayed [time shift Δt[vib] = +6 hours] relative to the initial LD entrainment, the molecular oscillations appeared to be shifted to the left, with the peaks occurring ~5 hours before those of the FR control group (Fig. 3, left); if the VS cycles were advanced (Δt[vib] = –6 hours), however, the peaks appeared shifted to the right, occurring ~3.5 hours after those of the controls (Fig. 3, right). Thus, mechanical stimulation can change the phase of the Drosophila molecular clock and function as Zeitgeber.

Fig. 3 Phase shifts of PERIOD protein oscillations in central clock neurons caused by rhythmic mechanical stimulation.

(A) Schematic representation of the environmental conditions before flies were transferred to the bioluminescence counter. For the delay and advance experiment, flies were transferred at ZT20 and ZT8, respectively, relative to the previous vibration onset. Colors and shadings referring to light and stimulus conditions are the same as in Fig. 1. (B) Normalized bioluminescence activity of FR 8.0-luc flies—which express a PER-LUC fusion construct in a subset of the clock neurons (7, 8, 15)—after exposure to VS cycles (red) and control flies not exposed to VS (gray). VS cycles that were delayed (Δt[vib] = +6 hours; experimental flies, n = 15; controls, n = 11) relative to the initial LD entrainment lead to an advance of the molecular oscillations (left), whereas VS cycles that were advanced (Δt[vib] = –6 hours; experimental flies, n = 10; controls, n = 8) relative to the initial LD entrainment lead to a delay (right) compared with PER-LUC expression of control flies not exposed to VS stimuli. Bioluminescence data was de-trended before fitting of a sinusoidal model (solid lines) (fig. S8) (7). Both data and corresponding fit values have been normalized to the maximum of the fit function to highlight the phase difference. Two additional delay experiments (Δt[vib] = +6 hours; experimental flies, n = 17; controls, n = 25) and one advance experiment (Δt[vib] = –6 hours; experimental flies, n = 9; controls, n = 10) were performed, and molecular phase shifts in the same direction as shown in (B) were observed. Error bars represent SEM.

The phenotypes of nocteP and tilB1 mutants, together with the crucial role of chordotonal organs for both mechanical and temperature-dependent circadian entrainment, suggest that both modes of entrainment share a common molecular, cellular, and potentially mechanistic basis. When directly comparing the activities of wild-type flies synchronized to the two different Zeitgebers, the activity peaks in relation to the entraining stimuli acquired similar phases (fig. S2). At the end of two phase-delayed temperature cycles (TCs), wild-type flies exhibited their main activity peak at the late cold/early warm phase, whereas after comparable VS cycle shifts, the main activity peak occurred during the late silent or early vibration phase (Fig. 1, C and D, and fig. S2).

Our results reveal a mechanosensory input pathway to the fly’s circadian clock that requires signaling from ChOs. Although external, diurnally fluctuating mechanical stimuli can also act as extrinsic Zeitgebers, a circadian pattern of mechanoreceptor activation will inevitably result from every locomotor activity that is patterned in a circadian way. ChOs act as proprioceptors and are located at almost every joint of the insect body (3). The summary output of an animal’s ChOs would thus be a faithful monitor of its overall (locomotor) activity.

Circadian clocks respond to light and numerous other factors (16, 17), including, for example, social interactions (18, 19), drug administration (20), temperature (21), or feeding protocols (22), forcing the question of how these are being integrated to compute a central clock time. The finding that—in flies—proprioceptor activation, which accompanies all forms of locomotor behavior, can reset the clock offers a potential solution to this problem: All environmental stimuli that lead to changes in the animal’s behavior would also affect the clock through the concomitant changes in proprioceptor activation. A proprioceptive clock entrainment would automatically weight environmental stimuli according to their respective ability to generate locomotion, which is a good indicator of the stimuli’s evolutionary importance. An animal’s activity can indeed affect its circadian clock (23, 24), and some nonphotic influences, such as certain drugs, only affect the clock if the animal is free to move (25). Proprioceptive feedback to the clock offers a mechanism for such activity-dependent clock entrainment.

Supplementary Materials

Materials and Methods

Figs. S1 to S10

Table S1

References (2631)

  • * These authors contributed equally to this work.

  • Present address: Department of Life Science, Division of Cell and Molecular Biology, Imperial College London, South Kensington, London SW7 2AZ, UK.

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: This work was supported by grants from the Biotechnology and Biological Sciences Research Council (BB/H001204/1 to R.S. and BB/G004455/1 to J.T.A.), the Human Frontier Science Program (to J.T.A.), and the European Union FP6 Integrated Project “EUCLOCK” (to R.S.). M.P.T. received funding from the Engineering and Physical Sciences Research Council (EP/F500351/1). Additional data, including raw data, are presented in the supplementary materials. The authors thank CoMPLEX students W. Ashworth and M. Ransley for their help, W. Potter for technical support, and P. Dayan from the Gatsby Computational Neuroscience Unit at University College London for valuable discussions.
View Abstract

Navigate This Article