Small-amplitude head oscillations result from a multimodal head stabilization reflex in hawkmoths

Abstract

In flying insects, head stabilization is an essential reflex that helps to reduce motion blur during fast aerial manoeuvres. This reflex is multimodal and requires the integration of visual and antennal mechanosensory feedback in hawkmoths, each operating as a negative-feedback-control loop. As in any negative-feedback system, the head stabilization system possesses inherent oscillatory dynamics that depend on the rate at which the sensorimotor components of the reflex operate. Consistent with this expectation, we observed small-amplitude oscillations in the head motion (or head wobble) of the oleander hawkmoth, Daphnis nerii, which are accentuated when sensory feedback is aberrant. Here, we show that these oscillations emerge from the inherent dynamics of the multimodal reflex underlying gaze stabilization, and that the amplitude of head wobble is a function of both the visual feedback and antennal mechanosensory feedback from the Johnston's organs. Our data support the hypothesis that head wobble results from a multimodal, dynamically stabilized reflex loop that mediates head positioning.

1. Introduction

Reflexive behaviours in animals rely on feedback from various sensory modalities to achieve control and stability in diverse contexts [1]. From a systems perspective, such feedback-control loops possess inherent dynamical properties that depend on the gains and latencies of their sensorimotor components [2,3]. When these parameters are affected by disease or injury, the dynamics of feedback-control also change thereby altering the rhythmic output of these behaviours [4]. To probe into the constitution of these reflex systems, it is thus useful to adopt a strategy in which the latencies of the sensory feedback are altered either directly, by inducing delays in transduction pathway (by chemical or genetic means, for example), or indirectly, by creating conditions that cause delayed sensory transduction, for example lowering light levels to increase visual transduction latencies [5].

In flying insects, several reflexes constituted as negative-feedback loops help to maintain the body posture during flight [1]. These reflexes, which are distributed along the body, include the positioning of their antennae [6], head stabilization [7,8] and abdominal flexion [9]. Each of these behaviours is a necessary feature of insect flight and must be mutually coordinated. In this context, flight-related reflexes in insects are particularly interesting as they are very rapid, and hence push the limits of nervous control. In particular, the head stabilization reflex is crucial for flight owing to its intimate association with vision [8]. Because insects cannot move their eyes relative to their heads, gaze stabilization is mostly achieved through compensatory head movements that help reduce motion blur of wide-field images during fast aerial manoeuvres [8,10,11].

Head stabilization is also interesting because it involves multimodal sensory feedback. In Diptera, it is mediated by visual feedback from eyes and mechanosensory feedback from halteres (modified hindwings in flies) [8,10] in addition to neck prosternal organs [12]. Recent work has shown that in non-Dipteran insects such as hawkmoths which lack halteres, head stabilization is mediated by visual and antennal mechanosensory feedback derived from the Johnston's organ (JO), which is highly sensitive and spans the pedicel–flagellar joint, thus monitoring the relative movements in this joint [13]. Together, these data show that head stabilization results from a multimodal feedback circuit, and is essential for flight control [13–16]

It is well-known that underdamped feedback-controlled dynamical systems exhibit oscillatory outputs [3,17]. Alternatively, oscillatory outputs may result from central pattern generator (CPG)-like neural circuits in the absence of sensory feedback [18]. Are such oscillatory outputs evident in feedback-controlled behaviors such as the head stabilization reflex? Moreover, if present, do these oscillations emerge from the feedback-driven dynamics, in which case their characteristics should be altered by changes in the feedback from individual sensory modalities? To test these hypotheses, we conducted a series of experiments in tethered oleander hawkmoths, Daphnis nerii, in which we manipulated the visual and antennal mechanosensory feedback, thereby indirectly altering the magnitude of the sensory feedback. Here, we show that hawkmoths display small-amplitude head oscillations (henceforth called head wobble) that depend on both visual and antennal mechanosensory feedback, consistent with our hypotheses. Our data are consistent with the hypothesis that such oscillatory dynamics emerge naturally from a dynamically stable reflex that determines the head positioning behaviour. When visual feedback is altered by changing light levels or antennal mechanosensory feedback by reducing the flagellar load on the JO, the dynamics of the head movements are altered. Thus, head wobble in hawkmoths provides insights into the dynamics of multimodal reflexes.

2. Material and methods

(a) . Moth breeding

All experiments were performed on adult oleander hawkmoths, Daphnis nerii, from our laboratory culture (for husbandry details, see [13]). The choice of D. nerii as a study system was motivated by their nocturnal lifestyle, as they can fly stably under very low light levels (operating light levels approx. 0.05–250 lux; P. Kalyanasundaram; S.P. Sane 2013, unpublished observations).

(b) . Treatments

We anaesthetized the moths by keeping them at −20°C for approximately 8 min. The moths were dorsally tethered by attaching a 3 mm neodymium magnet with cyanoacrylate glue on the back of their thorax; this magnet was attached to one of opposite polarity on the tether pole. The tether pole was oscillated using a stepper motor to which it was attached (further details in [13]).

(c) . Filming procedure

The moths were filmed from a frontal view with a high-speed camera (V611, Vision research, Ametek, USA) at a frame rate of 1200 frames per second for the imposed roll experiments and at 600 frames per second for the static tether experiments. We used small white markers on black paint on the moth head for contrast and easy digitization. A ring of infrared LEDs provided additional illumination during filming.

(d) . Experimental conditions

We recorded the head wobble behaviour of moths under two conditions: imposed roll and static tether.

(e) . Imposed roll stimulus

For the imposed roll stimulus, we rotated the tether with a peak-to-peak amplitude of 60° (±30°) at a frequency of 2 Hz (figure 1b), substantially lower than the 10 Hz typical frequency typical of head wobble (see Results), for 8 cycles. We replaced the beginning and end of each stimulus period with a linear ramp to avoid sudden movement of the tether. In the experiments involving imposed roll stimulus, the ambient light levels were maintained at approximately 250 lux (twilight). There were five experimental groups: control, flagella-clipped, flagella-reattached, Johnston's organ (JO) glued (in which the pedicel–flagellar joint was glued) and its corresponding sham procedure in which we applied glue to a few annuli above the pedicel–flagellar joint. These treatments are detailed below.

Figure 1.

(a) A qualitative model illustrating the feedback loops in head stabilization. (b) A schematic illustrating the experimental set-up, head wobble and the angle used for analysis. The white-and-red concentric circle on the moth's head signifies that the axis of rotation is in the direction pointing out of the page. (c(i), d(i)) Representative raw plots showing time-series of θ_thorax (black line) and θ_{head−thorax} (purple line) in the presence of imposed roll in the control (c(i)) and flagella-clipped (d(i)) moths. Note the distinct small-amplitude oscillations superimposed upon the large-amplitude compensatory head movements in d(i) and d(ii). (c(ii), d(ii)). Representative Fourier transforms of the time-series of θ_thorax and θ_{head–thorax} in a control (blue trace, c(ii)) and flagella-clipped (orange trace, d(ii)) moths. (e,f) Boxplots comparing wobble amplitude (e(i), f(i)) and frequency (e(ii), f(ii)) between control (blue box), flagella-clipped (orange box) and flagella-reattached (red box) moths, and between sham (left box, f(i,ii)) and Johnston's organ glued moths (right box; f(i,ii)) in the imposed roll experiments. In all boxplots, the data points are displayed on the boxplots. For comparison in figure 1e(i); *signifies p < 0.05 using Kruskal–Wallis test followed by Nemenyi test. For comparison in figure 1f(i), **p = 0.0037 using Wilcoxon rank sum test.

(f) . Control group

Moths were anaesthetized and tethered but underwent no further manipulations. Thus, their antennal mechanosensory feedback was intact.

(g) . Flagella-clipped moths

Flagella were excised around the 3rd/4th annulus, drastically reducing the mechanical load on JO and disrupting the antennal mechanosensory feedback.

(h) . Flagella-reattached moths

Flagella were cut and reattached to their stumps using cyanoacrylate glue. This treatment ensured that only the mechanosensory feedback from JO was restored, whereas feedback from the rest of the antenna remained disrupted owing to the severed nerve.

(i) . Johnston's organ glued group/sham treatment

Moths were kept on a metal plate placed atop an ice-bath throughout the treatment procedure (approx. 30 min) to render them relatively immobile for the duration of the treatment. In both groups, the head and base of the antennae were descaled to expose the pedicel–flagellar joint, following which cyanoacrylate glue was applied in the pedicel–flagellar joint in the Johnston's organ (JO) glued group, and approximately at the junction between 2nd and 3rd annuli in the sham group to ensure that the gluing procedure itself did not affect head stabilization. By gluing the pedicel–flagellar joint, feedback from JO, which spans that joint, was severely reduced, and hence similar to moths with clipped flagella. By contrast, the sham treatment controlled for the effects of the glue application. All the groups were allowed to recover for 30 min post-treatment before starting the experiment.

(j) . Static tether experiments

In the static tether case, the tether was kept in a fixed position and the groups were filmed in three light conditions (approx. 200 lux: clear-sky twilight; approx. 20 lux: overcast twilight; <0.01 lux: clear night sky/quarter moon). We used a lux meter (centre 337; range: 0.01–40 000 lux) to measure the ambient light intensity around the moth. The order of the light conditions was randomized across moths. In these experiments, there were two experimental groups: control and flagella-clipped, prepared as described above. In both the experiments the moths were flapping during the recording.

(k) . Data analysis

Using a custom C++ program in an OpenCV library, we digitized and computed the following angles in MATLAB (also see [13]; figure 1b):

• θ_thorax : angle between thorax and frame vertical;

• θ_head : angle between head and frame vertical;

• θ_{head−thorax} : angle between head and thorax.

For imposed roll at 2 Hz, we measured the magnitude of the highest peak between 5 and 15 Hz from the Fourier transforms of θ_{head−thorax} in all the moths (figure 1c–f; also see figure S1A–F). In static tether experiments, the stimulus frequency is zero, and hence we computed the magnitude of the highest peak between 2 and 15 Hz from the Fourier transform (figure 2; also see figure S1G,H).

Figure 2.

(a–f) Static tether experiments in control and flagella-clipped moths at approximately 200 lux (a,b), approximately 20 lux (c,d) and dark (e,f). Representative Fourier transforms of the time-series of θ_thorax and θ_{head−thorax} are shown in the left panel for approximately 200 lux (a), 20 lux (c) and less than 0.01 lux (e). In these plots, control moths are shown by the blue trace and flagella-clipped moths by the orange trace. The boxplots compare wobble amplitude between control (blue box) and flagella-clipped moths (orange box) at approximately 200 lux (b, white), approximately 20 lux (d, grey), dark (f, black). The outlier in (f) corresponds to a moth that showed more than normal head movement. (g) Boxplot comparing wobble frequency between control and flagella-clipped conditions at the three light levels. For (b), ***

p = 0.0006 by Wilcoxon rank sum test. For (d) *p = 0.0393 by Wilcoxon rank sum test.

3. Results

(a) . Head wobble amplitude increases when antennal mechanosensory feedback is reduced

In moths that were subjected to an externally imposed roll stimulus, we observed a distinct head wobble in both control and flagella-clipped moths superimposed upon large-amplitude compensatory head movements (figure 1c,d). The frequency of this head wobble was distinct from the frequencies of the imposed roll (2 Hz) and wingbeat (approx. 30 Hz; figure 1c(ii), see also figure 2a,c,e). In control moths (example in figure 1c; also see electronic supplementary material, figure S2A), the amplitude of the wobble was 1.3 ± 0.2° (blue box, n = 11; median ± standard error of mean; figure 1e(i)) and frequency was 10.6 ± 0.5 Hz (blue box, figure 1e(ii)), whereas in flagella-clipped moths (example in figure 1d; electronic supplementary material, figure S2B), the wobble amplitude was 3.8 ± 0.5° (orange box, figure 1e(i), n = 10) and frequency was 11.4 ± 0.4 Hz (orange box, figure 1e(ii)). In moths whose flagella were reattached, thereby restoring mechanical load on the JO, the wobble amplitude was 1.4 ± 0.4° (red box, n = 8; example in electronic supplementary material, figure 1A,B; figure 1e(i)) and frequency was 10.8 ± 0.7 Hz (red box, figure 1e(ii); electronic supplementary material, figure S2C), similar to the control case. The head wobble amplitude was significantly greater in flagella-clipped moths than in either control or flagella-reattached moths (p = 0.0002, Kruskal–Wallis test, followed by Nemenyi test; figure 1e(i)). However, the wobble frequencies in the control, flagella-clipped and flagella-reattached moths were not significantly different from each other (figure 1e(ii)).

The above experiments showed that antennal mechanosensory feedback is required to minimize head wobble. The key mechanosensory organ in the antenna is the JOs, which spans the pedicel–flagellar joint. To determine its role, we glued the pedicel–flagellar joint to reduce JO feedback (JO glued group) and compared these data with a sham group in which glue was applied a few annuli above the pedicel–flagellar joint. As in the flagella-clipped group, we observed a prominent wobble at 10.4 ± 0.4 Hz in the JO glued group (n = 8, figure 1f(i,ii); electronic supplementary material, figure S2D,E). The wobble amplitude was significantly enhanced in JO glued animals (3.2 ± 0.7°, n = 8) as compared with the corresponding sham treatment (1.2 ± 0.2°, n = 12) (p = 0.0037, Wilcoxon rank sum test; figure 1f(i)). The wobble frequencies in both cases were not significantly different (figure 1f(ii)). These data are consistent with the hypothesis that head wobble in hawkmoths depends on mechanosensory feedback from JO (figure 1a; [13]).

Except in overdamped or critically damped cases, dynamically stable processes naturally generate oscillatory outputs (manifest in this case as a head wobble) owing to the gains and delays of the feedback loop associated with each sensory modality [17] This hypothesis predicts that the reduction of the feedback from the sensory modalities should increase the wobble amplitude, as the system drifts further from its set-point. Consistent with this prediction, wobble amplitude increased when the antennal mechanosensory input was reduced (figure 1e(i),f(i)).

(b) . Visual and antennal mechanosensory feedback influence head wobble frequency and amplitude respectively

In the experiments described above, we observed the small-amplitude head wobble superimposed with large-amplitude compensatory head movements elicited by an externally imposed roll stimulus which induces optic flow on the retina of the moth. Does head wobble also occur in the absence of the compensatory head movements? Moreover, does head wobble depend on visual feedback in addition to antennal mechanosensory feedback? To address these questions, we conducted a series of experiments on moths attached to a static tether. We indirectly altered the latency of visual feedback [5] by conducting the experiments in three ambient light conditions that varied by three orders of magnitude (approx. 200 lux, approx. 20 lux, <0.01 lux) (figure 2a–f). For each of these light conditions, we repeated the antennal manipulations.

At an ambient light intensity of approximately 200 lux, the wobble amplitude in control moths (blue box, 0.7 ± 0.1, n = 7) was lower than in flagella-clipped moths (orange box, 2.4 ± 0.2°, n = 7; p = 0.0006, Wilcoxon rank sum test) (figure 2a,b; electronic supplementary material, figure S3A,B). Thus, antennal mechanosensory feedback helps minimize head wobble amplitude at approximately 200 lux. We observed a similar result at  light intensity of approximately 20 lux (figure 2c,d; electronic supplementary material, figure S3C,D), in which the wobble amplitude in flagella-clipped moths (orange box, 2.2 ± 0.4°, n = 8) exceeded the value in control moths (blue box, 1 ± 0.1°, n = 6; p = 0.0393, Wilcoxon rank sum test). At 20 lux, the variability of head wobble amplitude values was greater for flagella-clipped moths, suggesting that they had greater difficulty in minimizing head wobble, perhaps owing to the increased latency in visual feedback combined with reduction of antennal mechanosensory feedback.

Under dark conditions (<0.01 lux), the wobble amplitude in flagella-clipped moths (0.8 ± 0.3°, n = 8) was similar to control moths (0.7 ± 1.4°, n = 9) (p = 0.9125, Wilcoxon rank sum test) (figure 2e,f; electronic supplementary material, figure S3E,F). In the flagella-clipped moths, wobble amplitude in dark conditions was significantly lower than that at approximately 20 lux (p < 0.05, Friedman test, followed by Bonferroni correction, electronic supplementary material, figure S1G; figure 2c–f), but the wobble amplitudes at approximately 200 lux and approximately 20 lux were similar (approx. 200 lux: 2.4 ± 0.2°, n = 7; approx. 20 lux: 2.2 ± 0.4°, n = 7). In flagella-clipped moths, we observed low wobble amplitudes in the dark, probably because in this case they operate under severely reduced feedback.

Is wobble frequency a function of these multimodal inputs? At all the light levels, the frequency of head wobble was similar between control (approx. 200 lux: 8.7 ± 0.5 Hz; approx. 20 lux: 7.2 ± 1 Hz; dark: 2.7 ± 1) and flagella-clipped (approx. 200 lux: 9.7 ± 0.5 Hz; approx. 20 lux: 7 ± 0.4 Hz; dark: 4.8 ± 1.4 Hz) moths (figure 2g). We hypothesized that wobble frequency will decrease as light intensity decreases, owing to increased latency of the visual feedback. To test this, we compared the flagella-clipped moths for which trials of all light levels were present because they likely depend on visual feedback. Consistent with the hypothesis, wobble frequency slightly reduced when light levels went from approximately 200 lux (9.7 ± 0.4 Hz, n = 7) to approximately 20 lux (7 ± 0.4 Hz, n = 7), but the reduction was greater when the illumination was <0.01 lux (6.3 ± 1.5 Hz; electronic supplementary material, figure S1H, p < 0.05, Friedman test followed by Bonferroni correction). Thus, the experiments on static tether also highlight the importance of both visual and antennal mechanosensory feedback for minimizing head wobble.

4. Discussion

Oscillatory dynamics in animal behaviour may arise in diverse scenarios. One, rhythmic actions may result from endogenous oscillatory outputs generated by neural circuits such as central pattern generators [18]. In such cases, sensory inputs are not essential for generating rhythmic patterns, which can exist even in deafferented preparations [19]. However, sensory inputs serve to modify the frequency and amplitude of the rhythmic output. Two, rhythmic output may result from the mechanics of the behavioural apparatus in the animal. For instance, the flapping frequency of certain insects emerges from mechanical resonance of the musculoskeletal system of their thorax [20]. Three, oscillations may naturally arise from the negative-feedback processes involved in stabilizing motor output [1,17]. Such oscillations have been studied in diverse systems, including hover-feeding from flowers in hawkmoths [16], dynamic whole-body oscillations in weakly electric fish [21,22], postural sway [23], postural ankle oscillations during constant heel elevation [24], slow eye oscillations [25], constant hand and finger positioning [26,27] in humans. In these examples, the dynamical output changes when sensory or motor components undergo an alteration either induced experimentally [16,27] or due to disease or injury [26]. In other cases, such oscillations may serve specific functions. For instance, whole-body oscillations in weakly electric fish are hypothesized as an instance of active sensing which enables the fish to control their position by altering the gain of their reafferent feedback to maintain constant sensory slip [28]. Alternatively, like microsaccades or eye tremors, the oscillations may serve perceptual functions, including visual scanning and performance of high-acuity tasks [29], or to prevent perceptual fading. Although we cannot reject these putative functional scenarios, our study suggests that, in moths, head wobble emerges from the dynamics of a multisensory head stabilization reflex (figure 1a) and is minimized in moths with intact sensory feedback (control, figure 1c(i,ii)).

In Diptera, mechanosensory feedback for head stabilization is derived from halteres as well as the prosternal organ [10,30,31]. Although head oscillations are reported in Diptera, the role of haltere and visual inputs is unclear. For instance, in tethered blowflies, head fluctuations of ±10° occur in the absence of an imposed roll stimulus in a homogeneous visual background [7]. Small-amplitude head oscillations were observed in Drosophila at a frequency of approximately 15 Hz in the presence of static visual features, but were absent when the flies were presented with a uniform background [32]. In hawkmoths, halteres are absent and the prosternal organ is not yet described. However, previous studies show antennal mechanosensory feedback from the JO is crucial for flight [14–16] as well as head stabilization [13]. This study renders further support to the role of antennal mechanosensory feedback in maintaining head position. When antennal mechanosensory feedback from JO was altered, hawkmoths faced difficulties in maintaining their aerial position, as manifest from whole-body oscillations in a flower-tracking assay in the diurnal hawkmoth Macroglossum stellatarium [16]. The amplitude of jitter during hover-feeding in flagella-clipped moths was greater than in the flagella-intact and flagella-reattached conditions, similar to the wobble amplitude in Daphnis nerii. In these experiments, the wobble or jitter emerges as the nervous system tries to maintain a stable state using sensory feedback. Because the head stabilization was adversely affected (i.e. head wobble was enhanced), both vision and flight should be adversely affected, consistent with the above study [16].

A growing number of studies show that the brain integrates sensory cues from multiple sensory modalities to generate complex behaviours [10,30,33]. To generate a stable behavioural output these systems are organized as negative-feedback loops the dynamical properties of which depend on the gain and latencies of the cross-modal sensory input and motor components. Our study suggests that such behaviours may be characterized by small-amplitude oscillations that depend on the rates of their sensory feedback.

Data accessibility

All data are accessible at the following link: https://data.mendeley.com/datasets/xndr5g45db/draft?a=1bb29b13-57fb-4ca6-9dd7-8db1cd903209. Electronic supplementary material, figures S1–S3 accompany this paper [34].

Authors' contributions

P.C.: conceptualization, formal analysis, investigation, methodology, visualization, writing—original draft, writing—review and editing; U.M.: formal analysis, methodology, writing—review and editing; S.P.S.: conceptualization, formal analysis, funding acquisition, project administration, resources, supervision, visualization, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

The authors declare no competing interests.

Funding

Our study was funded by the Air Force Office of Scientific Research (AFOSR), no. FA2386-11-1-4057 and no. FA9550-16-1-0155, and the National Centre for Biological Sciences (Tata Institute of Fundamental Research) to S.P.S. We acknowledge support of the Department of Atomic Energy, Government of India, under project no. 12-R&D-TFR-5.04-0800, and NCBS core computational facilities, supported under project no. 12-R&D-TFR-5.04-0900.