Walking bumblebees see faster

Abstract

The behavioural state of animals has profound effects on neuronal information processing. Locomotion changes the response properties of visual interneurons in the insect brain, but it is still unknown if it also alters the response properties of photoreceptors. Photoreceptor responses become faster at higher temperatures. It has therefore been suggested that thermoregulation in insects could improve temporal resolution in vision, but direct evidence for this idea has so far been missing. Here, we compared electroretinograms from the compound eyes of tethered bumblebees that were either sitting or walking on an air-supported ball. We found that the visual processing speed strongly increased when the bumblebees were walking. By monitoring the eye temperature during recording, we saw that the increase in response speed was in synchrony with a rise in eye temperature. By artificially heating the head, we show that the walking-induced temperature increase of the visual system is sufficient to explain the rise in processing speed. We also show that walking accelerates the visual system to the equivalent of a 14-fold increase in light intensity. We conclude that the walking-induced rise in temperature accelerates the processing of visual information-an ideal strategy to process the increased information flow during locomotion.

Keywords: Bombus terrestris; Gaussian white noise; electroretinograms; photoreceptors; state dependency; temperature.

Figure 1.

Experimental set-up and ERGs of sitting and walking bumblebees stimulated with green light pulses. (a) Schematic illustration of the experimental set-up. The bumblebee was tethered and positioned on top of an air-supported ball. While presenting green light (530 nm) pulses, the ERG was recorded. An infrared lamp and a thermographic camera were used to control and measure the temperature. (b) Upper trace: single pulse of green light (50 ms duration at 1 Hz, pulse intensity: 2 × 10¹⁵ photons cm⁻² s⁻¹). Lower panel: ERG response curves of one example trace of an animal during walking (violet) or sitting (light blue: before walking, dark blue: after walking). The time difference between the onset of the light pulse and the crossing of the half-maximum amplitude gave the latency. (c) Boxplots of the latency at half-maximum of the ERG response during walking (violet) or sitting (light blue: before walking, dark blue: after walking) of 16 bumblebees. The latency at half-maximum was significantly shorter during walking than during sitting (p < 0.001, Wilcoxon test with Bonferroni correction), but no differences were seen when comparing before and after walking (p = 0.1359, Wilcoxon test with Bonferroni correction). Boxplots show median, interquartile range (IQR), whiskers with 1.5 x IQR and outliers greater than 1.5 x IQR.

Figure 2.

Effects of locomotion on the ERG of B. terrestris during stimulation with Gaussian white noise. (a) Section of the Gaussian noise stimulus (black; for whole trace, see electronic supplementary material, figure S1) and the corresponding ERG response (blue). To illustrate the similarity between noise stimulus and ERG response, the noise track was inverted and shifted (right). (b) Cross-correlation between the signal of the photodiode and the ERG to measure the latency of the ERG during stimulation with Gaussian noise. Cross-correlation measures the similarity of two waveforms in the time domain and can assume values between −1 (identical waveforms with inverted signs) and 1 (identical waveforms). The x-value at the minimum of the cross-correlation indicates the latency between the two curves. The minimum during walking (violet) occurs earlier than when the animal is sitting (blue). (c) Boxplots show the responses of 12 animals to the noise stimulus, before walking (light blue; 27.6°C ± 1.7°C), during walking (violet; 29.9°C ± 2.0°C) and after walking (dark blue; 27.5°C ± 2.0°C). The latency (ms) was significantly shorter during walking than during sitting (p < 0.0005, Wilcoxon test with Bonferroni correction), but no differences were seen when comparing before and after walking (p = 0.748, Wilcoxon test with Bonferroni correction). Boxplots show median, IQR, whiskers with 1.5 x IQR and outliers greater than 1.5 x IQR. (d) Time course of latency (upper trace), temperature (middle trace) and forward motion (lower trace). During locomotion (highlighted in violet), the latency decreased and the temperature increased (symbolized with thermographic pictures at the top; schematic drawing shows the position and orientation of the animal). After walking, the values return to their initial level.

Figure 3.

Gain and linear coherence between the stimulus and the ERG for experiments with walking bumblebees. (a) To assess which frequency range of the ERG response was specifically affected, we calculated gain (i) and linear coherence (ii) and averaged over 11 animals (for individual responses see electronic supplementary material, figures S3 and S4). The gain was fitted with a second-order low-pass filter (dashed lines). During walking, a shift to higher frequencies was observed (grey shaded area indicates right-shift of fit). Dips at 150 Hz result from a noise artefact that could not be eliminated. (b) Parameters of second-order low-pass fit to the gain. The amplitude β did not differ between any of the conditions ((i); p > 0.05, Wilcoxon test with Bonferroni correction). The time constant τ became significantly smaller during walking ((ii); p = 0.001, Wilcoxon test with Bonferroni correction). There was a small, but significant difference in τ between the two sitting conditions ((ii); p = 0.014, Wilcoxon test with Bonferroni correction). Walking had no effect on the damping factor ζ ((iii); p > 0.1, Wilcoxon test with Bonferroni correction), but there was a small, but significant difference between the two sitting conditions ((iii); p = 0.004, Wilcoxon test with Bonferroni correction). Boxplots show median, IQR, whiskers with 1.5 x IQR and outliers greater than 1.5 x IQR.

Figure 4.

Effects of temperature on ERGs of photoreceptor cells in bumblebees stimulated with noise stimulus. (a) Cross-correlation between the signal of the photodiode and the ERG to measure the latency of the ERG during stimulation with Gaussian noise. The minimum of the cross-correlation occurred earlier, when the bumblebee was heated up while it was sitting (orange) compared with when the animal was just sitting (shades of blue). (b) Boxplots show the responses of 22 animals to the noise stimulus, before heating (light blue; median temperature: 26.9°C ± 1.2°C), while heated (orange; 31.9°C ± 2.0°C) and after return to initial temperature (dark blue; 27.0°C ± 1.8°C). The latency was significantly shorter when heated than at the initial temperature (p = 3 × 10⁻⁵, Wilcoxon test with Bonferroni correction), but no differences were seen when comparing the initial temperatures before and after heating experiments (p = 0.0534, Wilcoxon test with Bonferroni correction). Boxplots show median, IQR, whiskers with 1.5 x IQR and outliers greater than 1.5 x IQR.

Figure 5.

Gain and linear coherence between the stimulus and the ERG for experiments in which the bumblebees were heated by an infrared lamp. (a) Averaged gain and coherence showed a clear shift to higher frequencies in heated animals (n = 22; for more details see electronic supplementary material, figures S5 and S6). Dashed lines are fits of second-order low-pass filter functions. In the heated condition, a shift to higher frequencies was observed (grey shaded area indicates right shift of fit). Dips at 150 Hz result from a noise artefact that could not be eliminated. (b) Parameters of second-order low-pass filter fits. The amplitude β of the fit function increased significantly during heating compared with the condition before heating (p = 0.0012, Wilcoxon test with Bonferroni correction), whereas there were no significant differences in the amplitude between the sitting conditions and between the heating condition and the after heating condition (p > 0.02, Wilcoxon test with Bonferroni correction). The time constant τ was significantly smaller during heating compared with both sitting conditions (p < 0.0001, Wilcoxon test with Bonferroni correction), but showed no significant difference between the sitting conditions (p = 0.592, Wilcoxon test with Bonferroni correction). The damping factor ζ was significantly increased during heating compared with both sitting conditions (p ≤ 0.001, Wilcoxon test with Bonferroni correction), while it did not differ between the sitting conditions (p = 0.287, Wilcoxon test with Bonferroni correction). Boxplots show median, IQR, whiskers with 1.5 x IQR and outliers greater than 1.5 x IQR.

Figure 6.

Comparison of temperature and intensity effects. (a) Latency for walking (violet; 60 data points from 12 animals) and heated (orange; 110 data points from 22 animals) animals plotted against the eye temperature. With increasing temperature, the latency of the ERG decreases. Linear model predictions for the final model (fixed effects of temperature and treatment, individual effects on latency and temperature response) are shown as lines, with bootstrapped 95% confidence intervals represented by shaded areas (for more details, see electronic supplementary material, figures S8 and S9). (b) The latency of ERG responses from 17 animals was measured after adaptation to four different intensities (1.74 × 10¹⁴, 7.27 × 10¹⁴, 1.31 × 10¹⁵ and 2.41 × 10¹⁵ photons cm⁻² s⁻¹). As the light intensity increases, the latency decreases. Boxplots show median, IQR, whiskers with 1.5 x IQR and outliers greater than 1.5 x IQR.