1. Neuroscience
Download icon

The roles of vision and antennal mechanoreception in hawkmoth flight control

  1. Ajinkya Dahake
  2. Anna L Stöckl  Is a corresponding author
  3. James J Foster
  4. Sanjay P Sane
  5. Almut Kelber
  1. Lund University, Sweden
  2. Tata Institute of Fundamental Research, India
Research Article
  • Cited 3
  • Views 1,472
  • Annotations
Cite this article as: eLife 2018;7:e37606 doi: 10.7554/eLife.37606

Abstract

Flying animals need continual sensory feedback about their body position and orientation for flight control. The visual system provides essential but slow feedback. In contrast, mechanosensory channels can provide feedback at much shorter timescales. How the contributions from these two senses are integrated remains an open question in most insect groups. In Diptera, fast mechanosensory feedback is provided by organs called halteres and is crucial for the control of rapid flight manoeuvres, while vision controls manoeuvres in lower temporal frequency bands. Here, we have investigated the visual-mechanosensory integration in the hawkmoth Macroglossum stellatarum. They represent a large group of insects that use Johnston’s organs in their antennae to provide mechanosensory feedback on perturbations in body position. Our experiments show that antennal mechanosensory feedback specifically mediates fast flight manoeuvres, but not slow ones. Moreover, we did not observe compensatory interactions between antennal and visual feedback.

https://doi.org/10.7554/eLife.37606.001

Introduction

The impressive aerobatic manoeuvres of insects provide an insightful model for the neural control of flight (Frye and Dickinson, 2001; Fuller et al., 2014). Insect flight requires continual sensory feedback, both on the position of the body relative to the environment, as well as on perturbations to body position. Visual feedback provides key information about flight parameters including ground speed, distance to obstacles and targets, and aerial displacements (for a review see Srinivasan et al., 1999). However, visual estimation of self-motion (Fuller et al., 2014; Hung et al., 2013) is limited by its temporal resolution and substantial latency to flight muscle activation (Sherman and Dickinson, 2004; Suver et al., 2016). This may often be too slow to control very fast aerial manoeuvres, which require rapid sensory feedback before perturbations become uncontrollably large and thus energetically costly to the animals (Bender and Dickinson, 2006).

Avoiding the temporal limitations set by the visual system, insects use mechanosensors to sense their own motion, as these can transduce perturbations on much faster time scales (Yarger and Fox, 2016). The halteres of Dipteran insects are a classic example of gyroscopic function in active flight (Fraenkel and Pringle, 1938; Nalbach, 1994; Pringle, 1948). Halteres are club-shaped mechanosensory structures that were evolutionarily derived from the hind-wings. They vibrate at the wing beat frequency and can sense rotations in any axis, and provide crucial sensory input to stabilise flight after perturbations (Ristroph et al., 2010). Halteres, however, are a special feature of only Dipteran (and Strepsipteran) insects (Pix et al., 1993). How do flying insects from other orders, which also require fast feedback for stable flight, control flight manoeuvres without halteres? This question is especially interesting in insects active in dim light, as the visual systems of many insects trade off temporal acuity for sensitivity (Warrant, 1999; Warrant, 2017), thus rendering visually based flight control even less reliable and increasing the need for mechanosensory feedback control.

Sphingids are a group of flying insects that are able to fly over a wide range of light intensities, due to their superposition compound eyes and additional neural adaptations (O'Carroll et al., 1996; O'Carroll et al., 1997; Stöckl et al., 2017a; Theobald et al., 2010). The effect of light intensity on their visual flight control has been quantified recently (Sponberg et al., 2015; Stöckl et al., 2017a). Moreover, the crepuscular hawkmoth Manduca sexta has been shown to use information provided by antennal mechanosensors, which may function similar to Dipteran halteres (Sane et al., 2007). The mechanosensory Johnston's organs, present at the pedicel-flagellar joint of the antennae, are stimulated by deflections of the antennal flagellum, and are sensitive to a wide range of frequencies (Dieudonné et al., 2014), which far exceed the temporal response range of the visual system (Stöckl et al., 2017a; Theobald et al., 2010). After ablation of their flagella, the Johnston’s organs of M. sexta no longer receive relevant information, causing flight instability in these moths, whereas re-attachment of the flagella statistically significantly improves their flight performance (Sane et al., 2007). Impaired flight performance following flagellar ablation was also observed in other Lepidopteran species, such as the tortoise-shell butterfly Aglais urticae (Gewecke and Niehaus, 1981; Niehaus, 1981) and the diurnal swallowtail moth Urania fulgens (Sane et al., 2010). Although the above studies underscored the importance of antennal mechanosensors for natural flight, the severe behavioural impairment caused by flagellar ablation meant that the precise contributions of antennal mechanosensors to flight control remained an open question, as did their integration with the visual sense.

To address these questions, we chose an insect model which retains both the motivation and ability to fly after flagella ablation: the diurnal hawkmoth Macroglossum stellatarum. M. stellatarum feed from flowers while hovering in front of them, and previous studies have underscored the importance of visual feedback on their flower tracking behaviour (Farina et al., 1995; Farina et al., 1994; Kern, 1998; Stöckl et al., 2017a). Using this hawkmoth, we were able to test the role of antennal mechanosensors for the control of stationary hovering flight, as well as for flight manoeuvres at controlled temporal frequencies, focussing on the integration of visual and mechanosensory information (Figure 1A). Here, we show that in M. stellatarum, antennal mechanosensors play a key role in the control of hovering flight, specifically in the control of fast flight manoeuvres (rapid turns). Furthermore, we show that visual and antennal mechanosensory feedback operate in different frequency bands, with no sign of compensatory interaction.

Results

Antennectomised hawkmoths performed less stable hovering at a stationary flower

To quantify the role of antennal mechanosensory feedback in free flight, we trained hawkmoths (Macroglossum stellatarum) in a flight cage to approach and hover in front of an artificial flower with sugar solution provided in a nectary at its centre (see Materials and methods). Individual moths were tested in three antennal conditions: with intact antennae (control, blue, Figure 1B and Figure 1—figure supplement 1) with ablated flagella (ablated, red) and with re-attached flagella (reattach, green). The number of animals taking off and feeding from the flower decreased significantly with flagella ablation, but returned to control levels following reattachment (Table 1, Supplementary file 1). Yet, depending on light levels, a substantial proportion of individuals with ablated flagella still approached and fed from the flower (60% in bright and 36% in dim light), making it possible to study the combined roles of vision and antennal mechanosensory feedback on flight control in more detail.

Figure 1 with 1 supplement see all
Flight control in hawkmoths requires vision and mechanosensation.

(A) Flight control in insects requires sensory feedback on perturbations of body position. The visual system supplies such feedback, but with comparably long response latencies. In addition, insects use mechanosensory systems to control their position in the air, which provide rapid feedback and thus are crucial for fast flight manoeuvres. Here, we investigated the role of antennal mechanosensation and vision on flight control in the hummingbird hawkmoth Macroglossum stellatarum. (B) In order to quantify the effects of antennal mechanosensation in free flight, we subjected each hawkmoth to three treatments: intact antennae (control, blue), ablated flagella (ablated, red) and re-attached flagella (reattach, green). To quantify the role of vision, we tested these three antennal treatments in two different light intensities (bright: 3000 lux, corresponding to partially overcast daylight and dim: 30 lux, corresponding to sunset intensities). (C) All conditions were tested in free hovering flight at artificial flowers, which were either stationary (hovering) or moved at different temporal frequencies (manœuvre). (D) We used a stimulus composed of a sum-of-sines to sample distinct frequencies with similar velocities (amplitude adjusted accordingly), as well as a stimulus ramping up in frequency, while retaining similar amplitude (chirp).

https://doi.org/10.7554/eLife.37606.002
Table 1
Proportion of hawkmoths performing specific behaviours across antennal conditions and light intensities.

Proportion of trials in which animals performed the following behaviours: no flight, flight (but no tracking of the flower), tracking. This dataset is based on the animals participating in the moving flower experiments. Of the total number of animals, 27 control, 22 flagella ablated, and 14 re-attached moths were tested in both light intensities. Some were tested multiple times to collect the necessary tracking data, and thus have contributed multiple trials to this dataset. Statistical comparisons were performed using multinomial regression including individual identity as a random factor, to model the rates of one of the three behaviours as a function of antennal condition and light intensity (without interaction terms). Statistical significance is indicated by: *p < 0.05, **p < 0.01, ***p < 0.001. For statistical details, see Supplementary file 1.

https://doi.org/10.7554/eLife.37606.004
ControlAblatedReattach
brightNo flight0.030.110.05
Flight0.150.290.1
Tracking0.820.60 ***0.85
Total384220
 dim ***No flight0.030.340.08
 Flight0.090.300.08
 Tracking0.880.36 ***0.84
Total356625

When approaching the flower, moths with ablated flagella had distinctly longer and more tortuous flight trajectories than moths in the control and flagella re-attached conditions (Figure 2—figure supplement 1), pointing toward an impairment of flight control due to flagella ablation. To further quantify the effect of flagella ablation and re-attachement on flight performance, we initially focused on the hovering flight of the hawkmoth in front of a stationary flower, where its body position could be closely monitored (Figure 1C) and the target position was clearly defined by the position of the flower on which it fed. For these stationary flower experiment, we analysed the hovering flight of six animals, which performed in all three antennal conditions and in both light intensities (see Materials and Methods).

When hovering in front of the stationary flower, we noticed that ablated moths had a greater positional variation in relation to their target position than moths in the other two antennal conditions, as evident from their thorax position over time (blue line, Figure 2A). We quantified the amplitude of these thoracic movements across a range of frequencies from 0.5 to 50 Hz (Note that these frequencies are not related to flower movement, since the flower in this experiment remained stationary. The frequency analysis refers to the movement of the thorax of the animals). In all three antennal conditions, the amplitude of this thoracic ‘jitter’ decreased with increasing frequencies (i.e. the animals performed smaller movements at higher frequencies, Figure 2B; Figure 2—figure supplement 2A). At 3000 lux, the thoracic jitter of ablated moths was statistically significantly larger than that of the other two antennal conditions between 0.7 and 5 Hz, and between 8 and 11 Hz (Figure 2B, Supplementary file 2), whereas the difference between control and re-attached condition was not statistically significant. Thus, re-attaching the flagellum restored flight performance close to the control state.

Figure 2 with 2 supplements see all
Flagella ablated hawkmoths showed greater thorax and abdomen movements during hovering flight at a stationary flower.

(A) When hawkmoths hovered in front of a stationary flower at 3000 lux, it was notable that flagella ablated moths jittered around their target position with larger amplitudes than moths of the other two antennal conditions, as quantified by the position of their thorax. The nectary is centered at 0 mm in this graph. (B) The thorax of moths with ablated flagella jittered with significantly higher amplitudes than the other two antennal conditions at frequencies between 1 and 5 Hz. There was no significant difference between control and re-attached moths. (C) The position of the abdomen in the three antennal treatments showed a similar trend to the thorax: the flagella ablated moths exhibited significantly larger abdomen jitter in the frequency range between 0.5 and 10 Hz than the other two treatments. (B, C) Lines show average, and shaded areas ± SEM. Statistical significance is indicated below the plots as: black p < 0.001, dark grey: p < 0.01, grey p < 0.05, white p > 0.05. Post-hoc tests were performed as part of a general linear model including antennal treatment and frequency (binned to the logarithmic scale) as factors, see Supplementary file 2 and Supplementary file 3.

https://doi.org/10.7554/eLife.37606.005

Because many insects, including hawkmoths, use abdominal movements for aerial stabilization during flight (Camhi, 1970; Dyhr et al., 2013; Hinterwirth et al., 2012), we also quantified the movement of the abdomen over the same frequency range. Across all three antennal conditions, the abdominal and thoracic movements revealed similar trends; in flagella ablated moths, their magnitude was statistically significantly larger than in control moths or flagella re-attached moths over the entire range of frequencies tested (Figure 2C, Supplementary file 3). Like thoracic jitter, abdominal jitter of moths with re-attached flagella was not statistically significantly different from control moths (except for one frequency: 1.66 Hz).

Since hovering is a dynamically unstable flight mode (Liang and Sun, 2013; Wu and Sun, 2012), hovering animals need continual sensory feedback to maintain a fixed position (Cowan et al., 2014). Both the visual system and antennal mechanosensory systems could provide sensory feedback to correct for deviations from the target position. Because flagella ablated moths showed larger positional jitter, especially at higher frequencies, we conclude that antennal mechanosensory feedback is required for the control of hovering flight. Without antennal input, the feedback about deviations from the target position, likely supplied by the visual system and therefore slower, causes moths to drift further from their target position before a corrective manoeuvre can be initiated. This in turn results in greater thoracic and abdominal movements.

Flagella ablation reduced flower tracking performance at high flower movement frequencies

After observing impaired flight stability during flower approach and stationary hovering with flagella ablation, we went on to examine its effects on flight manoeuvres at specific temporal frequencies. To this aim, we moved the artificial flower along a controlled trajectory while the hawkmoths were feeding from the nectary, thus eliciting flight manoeuvres of controlled frequencies and amplitudes while the moths were tracking the flower (Figure 1C). To probe the moths’ manoeuvrability at different amplitudes and speeds of flower movement, we used two movement patterns: one pattern was generated from a sum of sine-waves ranging from 0.5 to 8.9 Hz. They decreased in amplitude with increasing frequency to retain a constant velocity (‘sum-of-sines’, Figure 3A), which allowed the hawkmoths to track the entire stimulus successfully. The second movement pattern had a constant amplitude, while its frequency increased over time from 0 to 7.3 Hz over time (‘chirp’, Figure 1D and 3C), thus resulting in increasing flower velocity. This stimulus was designed to test the limits of the hawkmoths’ manoeuvrability, as the increasing velocity made it more challenging for them to track the artificial flower. We analysed the flight performance of 12 moths, which tracked both stimuli in all antennal conditions and two light intensities.

Figure 3 with 4 supplements see all
Flagella ablated hawkmoths showed reduced tracking performance of flowers moving at high frequencies.

(A, C) Trajectories of hawkmoths tracking moving flowers with the sum-of-sines (A) and the chirp (C) stimulus. Trajectories of the different antennal conditions are stacked for comparability. When tracking a moving robotic flower at 3000 lux, hawkmoths with ablated flagella often overshot the movements of the flower, specifically at higher frequencies. With increasing frequencies, moths also increasingly lagged behind the phase of flower movements more strongly. While the amplitude in the sum-of-sines stimulus was adjusted such that moths of all conditions could track the entirety of the stimulus, the chirp stimulus forced moths with too large overshoots and phase-lags to loose contact with the flower, and abort tracking (see red tracks in C). (B) Together, overshooting and phase-lags resulted in an increased tracking error of flagella ablated moths with the sum-of-sines stimulus at frequencies between 2 and 6 Hz, compared to both the control and re-attached condition. Linear mixed-effects models were used to compare the tracking error of the different antennal treatments with respect to frequency. Colours indicate significance (black p < 0.001, dark grey: p < 0.01, grey p < 0.05, white p > 0.05, Supplementary file 4). The red indicator on the x-axis gives the median frequency at which flagella ablated moths aborted tracking the chirp-stimulus (D). Curves show the mean and 95% confidence intervals of the mean, calculated in the complex plane. (D) For the chirp stimulus, we compared the movement frequency of the flower, at which the moths aborted tracking across antennal treatments, showing that flagella ablated moths lost contact with the flower at significantly lower frequencies than the control and re-attached condition. A Friedman test was used to compare between the treatments (***p<0.001, **: p<0.01, *p<0.05, Supplementary file 5).

https://doi.org/10.7554/eLife.37606.009

When tracking the sum-of-sines stimulus, tracking performance of control hawkmoths was consistent with previous investigations of intact individuals of this species (Farina et al., 1995; Farina et al., 1994; Stöckl et al., 2017a). To quantify the accuracy of the tracking performance, we used a metric that evaluates their accuracy in tracking both the amplitude (Figure 3—figure supplement 1D–F) and the phase (Figure 3—figure supplement 1G–I) of the flower movement, termed tracking error (Roth et al., 2014; Sponberg et al., 2015). In all antennal conditions, the control moths tracked the sum-of-sines stimulus accurately at low flower frequencies (Figure 3A): at 3000 lux, their tracking errors were close to 0 for flower movements up to 1 Hz, indicating nearly perfect tracking (Figure 3B). With increasing frequency, tracking errors increased, but there was no statistically significant difference in tracking error between antennal conditions for frequencies below 2 Hz (Figure 3B, Supplementary file 4). At higher flower frequencies, flagella ablated moths overshot the flower movements, resulting in a greater lag between the position of the moth and the flower (Figure 3—figure supplements 1,2) and thus larger tracking errors (Figure 3B): in the range of 2 to 5 Hz, tracking errors were statistically significantly higher for the flagella ablated moths than for both the control moths and moths with re-attached flagella (Supplementary file 4). In this flower frequency range, hawkmoths with re-attached flagella also had statistically significantly higher tracking errors than control moths. Thus, the reduction of antennal mechanosensory feedback impaired flight control specifically at the higher temporal frequencies of flower movement, which compelled the moths to perform faster turns. The ability of flagella ablated moths to track at frequencies below 2 Hz suggests that vision (and possibly other sensory modalities) provide feedback that is sufficiently fast to enable control of slower manoeuvres.

To ensure that the differences in flight performance between the three antennal conditions was independent of the specific type of flower movement, we presented the same hawkmoths with a ‘chirp’ stimulus in which the amplitude of flower movement was held constant, while the temporal frequency continuously increased from 0 to 7.3 Hz (Figure 1D), and with it the velocity of the flower. Unlike the sum-of-sines stimulus, which moths in all antennal conditions were able to track in its entirety, this stimulus was designed to test the limits of the hawkmoths' manoeuvrability, as the increasing velocity made it increasingly difficult for the moths to track the flower. At low flower frequencies, hawkmoths of all antennal conditions tracked this stimulus with high fidelity (Figure 3C). However, as the flower frequency increased, flagella ablated moths tended to overshoot the position of the flower at the end of each sideways movement, when the flower movement changed direction. The accumulated phase lag and overshoot were eventually large enough to cause the moths to lose contact with the nectary and abort flower tracking (Figure 3C). Only 1 out of 12 ablated moths succeeded in following the flower movement during the entire stimulus at 3000 lux. In contrast, all control and 10 out of 12 re-attached moths tracked the flower until the maximum frequency. As a measure of effective flight performance, we quantified the flower frequency at which hawkmoths in the different antennal conditions aborted flower tracking (Figure 3D). At 3000 lx, there was no statistically significant difference between the control and re-attached flagella moths (Supplementary file 5). Flagella ablated hawkmoths aborted flower tracking at a median frequency of only 4.4 Hz, statistically significantly lower than the other two conditions (Figure 3D, Supplementary file 5). While re-attached hawkmoths did not show a statistically significant difference in tracking abortion frequency from control moths, looking at the flight tracks tracking the chirp stimulus in more detail (Figure 3—figure supplement 3) revealed some differences in tracking performance: the power they shared with the stimulus at frequencies above 4 Hz was higher than that of ablated moths, but lower than that of control moths (Figure 3—figure supplement 3E), and similarly, their phase delay did not increase as quickly with frequency as that of ablated moths, but quicker than for control moths (Figure 3—figure supplement 3E).

Despite the differences in the movement patterns of the two stimuli and their demands on flower tracking, we observed similar trends in hawkmoth flight performance. Moths in all antennal conditions tracked flower movements well at low frequencies, whereas flagella ablated hawkmoths were statistically significantly impaired at higher frequencies compared to the other two antennal conditions. The average flower frequency at which ablated moths failed to track the chirp stimulus was consistent with the frequency range for which the tracking error with the sum-of-sines-stimulus was greatest, despite the difference in flower velocity between stimuli. These data show that antennal feedback is crucial for fast turns - or directional changes - which are associated with changes in body posture.

Slow visual feedback impaired flower tracking performance of all antennal conditions

In the experiments described so far, hawkmoths of all antennal conditions did not differ in their flight performance when performing slower movements (at lower frequencies of positional jitter when hovering at a stationary flower, and at flower frequencies when tracking a moving flower). At these movement frequencies, feedback from other sensory modalities likely mitigated the problems in flight control caused by antennal ablation. In particular, visual feedback is known to provide information about changes in insect body (head) position in flight (for a review see Srinivasan et al., 1999), albeit with longer latencies and a lower frequency range than mechanosensory feedback (Sane et al., 2007). Because the latency of visual feedback depends on the ambient light intensity (Stöckl et al., 2017a), we next tested how the reliability of visual feedback affected the hawkmoth’s flight performance by decreasing the ambient light intensity in combination with the antennal manipulations. We tested the same group of hawkmoths at an illumination of 30 lux, close to the light intensity limit at which these diurnal hawkmoths are still able to reliably approach and feed from the artificial flowers (Stöckl et al., 2017a). If flagella ablated hawkmoths relied mainly on visual feedback for flight control when they lack antennal mechanosensory feedback, their flight performance should be poorer under low light, as compared to bright light. However, we did not expect any difference in the performance of hawkmoths in the control and re-attached conditions, because these moths receive fast feedback from their antennal mechanosensors.

To quantify the effect of light intensity on flight performance during stationary hovering, we calculated the difference in the amplitude of thoracic and abdominal movements for individual moths of all three antennal conditions (Figure 4C,D). We found no statistically significant effect of antennal condition on the average difference in thorax (Supplementary file 7) or abdomen (Supplementary file 13) jitter between dim and bright light, suggesting that differences in the temporal acuity of visual inputs did not additionally affect the impact of flagella ablation on flight performance.

Light intensity had the same effect on all antennal treatments.

To test the effect of visual feedback and its possible interaction with antennal mechanosensory feedback on flower tracking, we performed all experiments both in bright (3000 lux) and dim (30 lux) light intensities. Hawkmoths showed reduced tracking performance of artificial flowers moving at higher frequencies in dim light, due to the slowing of their visual system (Figure 2—figure supplement 2, Figure 3—figure supplements 1,4). Here, we compare tracking performance between bright and dim light across antennal treatments. (A) We quantified the difference in frequency between light intensities at which moths reached a tracking error of 1 with the sum-of-sines stimulus. There was no significant difference (Supplementary file 8) between antennal conditions, suggesting that vision reduced tracking performance in dim light irrespective of the presence or absence of mechano-sensory feedback. (B) Similarly, there was no significant difference between the tracking performance in dim and bright light for the chirp stimulus (quantified as the difference of tracking abortion frequency at the two light conditions) (Supplementary file 6). (C–D) We determined the difference between the log-transformed magnitude spectra for thorax (C) and abdomen (D) jitter in bright and dim light. No significant effect of antennal condition was found using Friedman comparisons of the average difference in thorax or abdomen movements (Supplementary files 7 and 13). Lines show average, and shaded areas ± SEM.

https://doi.org/10.7554/eLife.37606.016

We then went on to quantify the effect of light intensity on flight performance in the moving flower experiments. While we did observe an increase in tracking error of flagella ablated hawkmoths with the sum-of-sines stimulus in dim light (Figure 3—figure supplement 1B), particularly at lower flower frequencies, we observed the same effect in the control and flagella re-attached conditions (Figure 3—figure supplement 1A,C). We quantified the effect by comparing the difference in flower frequency at which each individual reached a tracking error of 1 in dim and bright light across antennal conditions (Figure 4A). There was no significant difference between antennal conditions (Supplementary file 8). Thus, light intensity affected flower tracking in general (as has been shown previously, (Sponberg et al., 2015; Stöckl et al., 2017a), but did not interact with antennal condition. Experiments using the chirp stimulus further confirmed this finding (compare Figure 3D and Figure 3—figure supplement 4). To compare the performance of moths in dim and bright light, we measured the difference in flower frequency at which moths aborted tracking with the chirp stimulus in dim and bright light across antennal conditions (Figure 4B). Also for this stimulus, we did not find statistically significant differences between antennal conditions (Supplementary file 6), indicating that visual feedback did not compensate for the loss of mechanosensory feedback in flagella ablated moths. Instead, the slower visual processing affected flight control similarly in all antennal conditions.

Discussion

Although visual and mechanosensory feedback is known to play a prominent role in the control of insect flight, it is not clear how these inputs are integrated by the insect brain to generate behaviour. In Dipteran flies, which use halteres as gyroscopic sensors, vision and mechanosensation operate in frequency ranges that are complementary (Mureli and Fox, 2015; Yarger and Fox, 2016). A natural question arising from these studies is: how do insects that lack halteres process mechanosensory and visual feedback? To address this question, we here investigated how visual inputs from compound eyes and mechanosensory inputs from antennal Johnston’s organs control flight in combination. For both stationary hovering and flight manoeuvres during flower tracking in Macroglossum stellatarum, our data show that antennal mechanosensory input is crucial for control of fast flight manoeuvres, while visual input controls the slower ones - similar to observation in flies.

Flagellar re-attachment improves flight performance

We have shown that flight control in the diurnal hawkmoth M. stellatarum requires feedback from antennal mechanosensors. As also observed in previous experiments by Sane et al. (2007), re-attaching the flagellum restored flight performance by reloading the Johnston’s organs, both for stationary hovering and flower-tracking behaviors. This is consistent with the growing body of evidence (Dieudonné et al., 2014; Gewecke and Niehaus, 1981; Niehaus, 1981; Sane et al., 2010) that Lepidoptera use antennal Johnston's organs for flight control. One possible way how antennal mechanosensors might impact flight control is by providing feedback for head stabilisation. While Hymenopteran insects (Polistes humilis) seem to purely rely on visual information to stabilise their head during roll manoeuvres (Viollet and Zeil, 2013), preliminary data from hawkmoths shows that they might require antennal feedback for head stabilisation (Sane et al., 2018).

As demonstrated previously (Sane et al., 2007), while flagellum re-attachment improved flight performance statistically significantly compared to the flagella ablated condition, it did not restore it to the level of intact animals in all experimental conditions. Feedback provided with re-attached flagella restored flight performance when hovering at stationary flowers to levels that were statistically not significantly different from the control group (Figure 2). Similarly, moths with re-attached flagella tracked the chirp stimulus at the moving flower for similar lengths as control animals (Figure 3C,D) – although impairments in flower tracking at higher flower movement frequencies compared to the control condition were visible upon a more detailed analysis of the flight tracks (Figure 3—figure supplement 3). There were also statistically significant differences in the flight performance of re-attached and control moths with the sum-of-sines stimulus (Figure 3A,B). These differences between experimental conditions suggest that flagella re-attachement did not entirely restore flight performance back to the levels of intact animals. One reason for this might be that some properties of the re-attached flagella differed from those of intact animals. Re-attached flagella were not connected to the haemolymph system of the hawkmoth and thus dried out, which reduced their weight by more than 50% (see Materials and methods). Since the flagella are thought to provide a mass that inertial forces act on (Sane et al., 2007), changes in weight may considerably alter the sensory input to the Johnston’s organs. Changes in the flexibility of the flagella due to moisture loss may also contribute to this effect. Moreover, there are also mechanosensors along the length of the flagellum, which may be important for flight control. When the antennal nerve is severed, these mechanosensory units remain inactive even after flagellar reattachment, which may add to the observed deterioration in their ability to control flight. The roles of these mechanosensors and of the weight and flexibility of the flagella need to be further explored in future experiments.

Antennal mechanosensation and vision operate in different frequency bands

Our experiments quantified the frequency range in which antennal mechanosensory feedback is required for the control of flight in M. stellatarum moths using a moving flower which the animals tracked to initiate flight manoeuvres at different temporal frequencies. We demonstrated that flagella ablated hawkmoths can track flowers moving at frequencies below 2 Hz with the same fidelity as hawkmoths with intact antennae (Figure 2 and 3). This suggests that control of slower manoeuvres is not as dependent on antennal mechanosensory feedback, as is the control of faster manoeuvres. On the other hand, flagella ablated hawkmoths performed statistically significantly worse than moths with intact and re-attached flagella at flower movements above 2 Hz, where more rapid turns are required to follow the lateral trajectory of the moving flower. Our findings are mirrored in the study of Dipteran flight control: slower rotations of fruit flies are tuned stronger to visual feedback, whereas faster rotations require feedback from haltere mechanosensors (Sherman and Dickinson, 2003).

It is noteworthy that the response delays in the flower tracking of M. stellatarum we observed were very short compared to visuomotor feedback loops measured in other hawkmoths (Sponberg et al., 2015) and other flying insects (Reiser et al., 2012; Viollet and Zeil, 2013): with the sum-of-sines stimulus at 3000 lux, hawkmoths in the control condition showed a phase lag of approximately 90° at the highest temporal frequency of 8.9 Hz (Figure 3—figure supplements 1,2), suggesting a response delay of less than 30 ms. Similarly, the phase delay with the chirp stimulus at 6.5 Hz in the control condition at 3000 lux was close to 80° (Figure 3—figure supplement 3), indicating a response delay of 35 ms. This would indeed be one of the fastest visuomotor transformations described in insects. Considering that visually tracking the flower likely involves computing a directional motion component, and the fastest latencies of wide-field motion neurons in blowflies are approximately 25 ms (Warzecha and Egelhaaf, 2000), it is unlikely that such fast flower tracking responses are purely elicited by vision in M. stellatarum. Recent work by Roth et al. (2016) in Manduca sexta has demonstrated that mechanoreceptors on the proboscis can play a role in monitoring flower position, in addition to visual input. Mechanosensors have much faster transduction than most visual receptors. We cannot exclude a mechansensory component to flower tracking originating from the proboscis in M. stellatarum, but the phase lags observed in M. sexta for purely mechanosensory tracking, even though shorter than those for purely visual tracking at higher temporal frequencies, are still distinctly larger than the ones observed for flower tracking in our experiments (Roth et al., 2016). Another possible explanation for the short tracking delays at high frequencies is a direct mechanical coupling between the head of the hawkmoth and the flower via the proboscis. The flowertracking responses of the hawkmoths might include this mechanical coupling, which could explain their extraordinarily fast responses at high flower movement frequencies. The change in responses across antennal conditions and light intensities shows that there is, nevertheless, a strong sensorimotor component of the behaviour, and since the potential mechanical coupling, as well as putative mechanosensory input from the proboscis, were present in all antennal conditions as well as light intensities, they did not affect the observed results with respect to visual and antennal mechanosensory feedback. Moreover, we did observe changes in flight performance upon antennal ablation and re-attachement both during flower approach (Figure 2—figure supplement 1) and hovering at stationary flowers (Figure 2, Figure 4C,D), where mechanosensory inputs and mechanical coupling did not play a role.

Thus, we conclude that mechanosensory feedback from the antennae is essential for the control of fast flight manoeuvres, which require corrective movements to occur in timescales that may not be sufficient for the transduction of visual feedback. This again is analogous to the finding that the control of fast saccadic rotations in Dipterans mainly requires mechanosensory feedback from the halteres, while vision plays a relatively marginal role (Bender and Dickinson, 2006; Sherman and Dickinson, 2003).

Vision does not compensate for the loss of antennal mechanosensation in hawkmoth flight control

Both vision and mechanosensation contribute to insect flight control, and the mechanistic underpinnings of this multimodal integration are subject of many ongoing investigations. In Dipteran flies, vision and haltere mechanosensation operate in complementary frequency ranges, and while both inputs are required for stable flight under most circumstances (Yarger and Fox, 2016), they do not seem to compensate for each other (Mureli and Fox, 2015). Antennal movements also depend on feedback from multiple sensory modalities. For example, in honeybees, airflow on the antennae and optic flow influence antennal positioning in tethered as well as free flight (Roy Khurana and Sane, 2016). In the Oleander hawkmoth Daphnis nerii, visual feedback modulates antennal positioning in a similar way (Krishnan and Sane, 2014).

Here, we tested how vision and antennal mechanosensation in combination influence flight control during flower tracking. Using a bright and a low light intensity, we manipulated the temporal resolution of visual responses (Stöckl et al., 2017a; Stöckl et al., 2016). In dim light, the low speed and reduced reliability of the visual input to flight control causes larger tracking errors when flowers move at high frequencies for all antennal conditions (Figure 4). This effect is explained by the fact that visual input is essential for moths to identify and track the flower movement relative to their own position – antennal mechanosensors cannot provide the required information (Figure 1A). Because visual processing is slower in dim light, moths face greater difficulties in resolving fast flower movements, which causes failure in tracking (Sponberg et al., 2015; Stöckl et al., 2017a).

We did not observe a specific effect of light intensity on flight control in the flagella ablated moths. This suggests that, even at higher resolution under brightly lit conditions, visual feedback is unable to mitigate the instability caused by the loss of antennal mechanosensory feedback. Two main hypotheses could explain this finding: first, the contributions of vision and mechanosensation contribute to the motor outputs via separate parallel pathways, whose functions do not overlap. This is unlikely, as recent recordings of descending neurons in Oleander hawkmoth show that they respond to both visual and mechanosensory stimulation (Mohan et al., 2017). Alternatively, vision and mechanosensation share descending pathways but operate in different frequency ranges, and the visual input is too slow to compensate for the lack of antennal mechanosensory feedback. The latter hypothesis is consistent with physiological studies showing that mechanosensors in Johnston’s organ respond to antennal displacements at frequencies of up to 100 Hz in the hawkmoth M. sexta (Sane et al., 2007), whereas the wide-field motion-sensitive neurons of the same species cease to respond at temporal frequencies above 20 Hz (Stöckl et al., 2017a), at which most mechanosensors of the Johnston’s organ only show a weak response. Eventually, an assessment of the physiological responses of descending neurons that activate the flight muscles is required to reveal the mechanisms of integration of visual and mechanosensory information in control of flight in hawkmoths.

Conclusion

Antennal mechanosensation represents one strategy for flying insects to obtain rapid sensory feedback about changes in self-motion, which is crucial for flight control. We showed here that in the diurnal hawkmoth M. stellatarum, mechanosesory feedback from antennae is required for the control of fast flight manoeuvres and rapid deviations from their hovering position, whereas their visual system drives the control of slower manoeuvres. These findings detail a striking similarity to the interaction between mechanosensory halteres and vision in the Dipteran flight control model, and for the first time dissect the combined role of visual and antennal mechanosensory feeback for flight control in hawkmoths, which may be representative for many other non-Dipteran insects.

Materials and methods

Animals

Wild adult Macroglossum stellatarum L. (Sphingidae), were caught in Sorède, France. Eggs were collected and the caterpillars raised on their native host plant Gallium sp. The eclosed adults were allowed to fly and feed from artificial flowers similar to the experimental flowers, in flight cages (70 cm length, 60 cm width, 50 cm height) in a 14:10 hr light:dark cycle for at least one day before experiments.

All animals were tested with intact antennae first (control), then with ablated flagella (ablated), and finally with re-attached flagella (reattach) as described below (Figure 1—figure supplement 1). Only data from animals that could be tested under all three antennal conditions was included in the final data analysis.

Surgery: flagella ablation and re-attachment

Request a detailed protocol

For flagella ablation, moths were held, by their thorax under a dissection microscope and their flagella were clipped with a pair of surgical scissors, while retaining 5–10 annuli (Figure 1—figure supplement 1B). This ensured that Johnston’s organs, located at the base of the antennae, were left intact but unloaded. Ablated flagella were preserved in a plastic petri dish with wet tissue to prevent them from drying and losing shape until they were re-attached to the same individual. Moths were left to recover from the surgery and tested on the following day.

To re-attach the flagella, moths were immobilized by cooling at 3° C for 8 min, followed by 2 min at −20° C. Flagella were quickly attached to the flagellar stump with a small amount of superglue (Loctite Super Glue Gel, Henkel, Figure 1—figure supplement 1C). After ensuring that the flagella were properly attached, moths were placed inside a plastic box (10 cm x 10 cm x 8 cm) on a wet tissue paper for 10 min to keep them quiescent and ensure proper reattachment. In case an animal broke the re-attached flagella, a spare one of similar size was used to repeat the re-attachment procedure. Moths were then allowed to recover for a day, before being used in experiments.

We noticed that the flagella lost moisture once re-attached. To quantify the reduction in weight due to moisture loss, we weighed a set of flagella directly after surgery and a few days later when they had dried. Dry flagella had statistically significantly lower weights than freshly ablated flagella (moist: 1.2 ± 0.2 mg, dry: 0.4 ± 0.3 mg; median and inter-quartile range, Wilcoxon rank sum test, z-value = −5.915, p<0.001). We could not determine the weight of the glue used for reattachment, but it is unlikely to exceed the difference between dried and moist flagella, considering the tiny amount of glue used.

To obtain a general idea of the weight ratios of the antenna, head and body of individual hawkmoths, we measured these quantities in six freshly sacrificed animals (three male, three female). The resulting average head:antenna ratio was 3.38:1 ± 0.68 standard deviation, with a weight of 10.5 ± 1.4 mg for the head and 3.1 ± 0.31 mg for the two antennae. The animals weighed an average of 252 ± 66 mg.

Experimental setup

Request a detailed protocol

We used a robotic flower assay as our experimental setup. This assay was first pioneered by Farina et al. (1994) and Sponberg et al. (2015), also used in Stöckl et al. (2017a). A flight cage (of the same size as the holding cage) was lined with soft muslin cloth and covered with black cloth on the outside, on three sides, while the front and top were sealed with Perspex windows to allow filming. An artificial flower (48 mm in diameter, on a 140 mm stalk) at the centre of the flight cage, with a nectary (opening of 8.3 mm diameter) filled with 10% sucrose solution, could be moved sideways (in arcs around the central pole). The position of the flower was controlled by a stepper motor (0.9 degree/step resolution, 1/16 microstepping, Phidgets, Inc.). The motor was interfaced using the Phidget21 MATLAB library (https://www.phidgets.com/docs21/Language_-_MATLAB) with custom written code shared by Simon Sponberg (Sponberg et al., 2015). In short, we set the position of the motor using the 'CPhidgetStepper_setTargetPosition' command of the 'phidget21' library according to the trajectory of the stimuli (Source Data 1) The cage was illuminated from above with an adjustable white LED panel and diffuser (CN-126 LED video light, Neewer, dimensions: 7.9 × 15.8 cm, 126 individual LEDs, colour temperature: 5400K). The light intensity was set to 3000 lux for the bright light condition and 30 lux for the dim light condition (measured with a Hagner ScreenMaster, B. Hagner AB, Solna, Sweden, at the position of the artificial flower). In addition, two 850 nm IR LED lights (LEDLB-16-IR-F, Larson Electronics) provided illumination for the infrared-sensitive high-speed video cameras (MotionBLITZ EoSens mini, Mikrotron) used to film the flower and moths. Videos were recorded at 100 fps, allowing us to record sequences of up to 28 s, which were required for our analysis of flower tracking. One camera was placed on top of the cage to film the flower and moth from above during all tests. For experiments with the stationary flower, a second camera providing a rear view was placed on a tripod outside the experimental cage, at approximately 30 cm distance from the artificial flower.

Behavioral experiments

Request a detailed protocol

Eclosed moths were taken from their holding cage and placed in small individually marked cardboard boxes, in which they would be held between trials. For the duration of the experiment, moths were only given access to sucrose from the artificial flower during trials in the experimental cage. A single hawkmoth at a time was introduced into the experimental cage.

We performed two sets of experiments: in the first one, we filmed the moth’s approach to and hovering at a stationary flower with both the top and the rear camera. In the second one, we filmed the moth tracking a moving flower using only the top camera. In this set of experiments, we started moving the artificial flower once the moth began to feed from it. We used two different types of movements, the ‘sum-of-sines’ stimulus and the ‘chirp’ stimulus, in the same flight bout. The first 16 s of the sequence thus comprised of the pseudo-random sum-of-sine stimulus composed of the following 14 frequencies, which were prime multiples of each other to avoid harmonic overlap: 0.5, 0.7, 1.1, 1.3, 1.7, 1.9, 2.3, 2.9, 3.7, 4.3, 5.3, 6.1, 7.9, 8.9 Hz. High frequencies had lower amplitudes and vice-versa, to assure equal velocities at all frequencies (Figure 1D and Source Data 1). The sum-of-sines stimulus was followed by a brief stationary phase of 0.5 s, and then the 11 s lasting chirp stimulus with fixed movement amplitude of 10.4 mm and frequencies increasing over time from 0 up to 7.3 Hz (Figure 1D and Source Data 1). To avoid startling the animals, we did not initiate the movement of the flower abruptly at full amplitude, but rather slowly ramped up (and ramped down) the amplitude of the stimuli over half a second before and after the ‘sum-of-sines’ and the ‘chirp’ stimulus. We excluded these portions of the stimulus from our analysis: we extracted 10 s of the flight path with the ‘sum-of-sines’ stimulus for analysis, 9.5 s from the ‘chirp’ stimulus, always starting at the same position of the stimulus for all animals.

The protocols were similar for both sets of experiments. Each individual was tested six times: in three antennal conditions (intact control, ablated and with reattached flagella), and in two light intensities (3000 and 30 lux). Because M. stellatarum were less motivated to fly in dim light, we first tested the moths in dim light, when they were hungriest and had the highest motivation to forage, and in bright light (3000 lux) later the same day. If a moth did not track both the sum-of-sines and the chirp stimulus (or the stationary flower for at least 6 s), we repeated the test the next day, until a full set of data was collected and the experiment moved on to the next condition. This experimental strategy gave flagella ablated (and re-attached) moths a chance to adapt to their altered mechanosensory feedback, and practice flying and tracking the flower on several days before succeeding. Indeed, our observations suggest that hawkmoths learned to adjust their flight to the lack or change of mechanosensory feedback, as the initial flight attempts of many flagella ablated (and to a lesser degree re-attached) moths showed more severe impairments than consecutive attempts.

Datasets

Request a detailed protocol

Our final datasets include only individuals that tracked the flower in all three antennal conditions in both light intensities. We used six individuals for the experiment with the stationary flower, performed one trial of 6 s at the stationary flower in each condition, and 12 different individuals with the moving flower, which performed one trial (comprised of the sum-of-sines and chirp stimulus) in each condition. Thus, the data analysis of the stationary and moving flower experiments (Figures 24) has a balanced design, with paired measures for all three antennal conditions and the two light intensities. Moreover, we characterized the general behaviour of all hawkmoths that were part of the moving flower experiment, including those that did not complete all antennal conditions and light intensities, and thus were not included in any further analysis. Thus, the general behaviour scores (Table 1) does not comprise a balanced design, and contains repeated measures (which were accounted for in the statistical analysis, see Data Analysis below).

Data analysis

Request a detailed protocol

The positions of the flower and the hawkmoth were digitised from the videos using the DLTdv5 software for MATLAB (Hedrick, 2008Dyhr et al., 2013). In experiments with stationary flowers, both the approach and the stationary hovering were digitised, whereas in experiments with moving flower, only sequences during which the proboscis of a moth was in contact with the nectary were rated as ‘tracking’ and digitized (as in Sponberg et al., 2015; Stöckl et al., 2017a). In the top view, a point on the flower, and a reliably identifiable point on the pronotum of the moth were used for reference. From the rear view videos, we used the centre of the nectary, the centre of the pronotum and the centre tip of the abdomen (Figure 3).

General behaviour

Request a detailed protocol

We characterised the general behaviour of all hawkmoths in the moving flower experiments, including those that did not complete all antennal conditions and light intensities, and thus were not included in any further analysis. We classified their behaviour into three different categories (Table 1): non-flying (animals which would not take off after 5 min in the experimental cage), flying (animals which flew but would not feed from the flower) and tracking (animals feeding from and tracking the flower, at least partially). We used multinomial logistic regression (package mlogit v0.2–4: (Croissant, 2013) to model the rates of one of the three behaviours (non-flying, flying, tracking, Table 1) as a function of antennal condition and light intensity, including the identity of individual moths as a random factor (Supplementary file 1).

Stationary flower experiments

Request a detailed protocol

To compare the stability of hovering flight between the different antennal conditions and light intensities, we analysed the position of the thorax and abdomen for a 6 s interval of hovering at the flower nectary during feeding (given perfect hovering, the thorax should retain a stable position, because the flower was immobile). We quantified the amplitude of thorax and abdomen movements across different movement frequencies by Fourier transforming their position over time (Figure 2B,C). To assess the effect of antennal condition across frequencies, we applied a linear mixed-effects model (Bates et al., 2015) with antennal condition, frequency and their interaction as fixed effects and individual identity as a random effect on the log-transformed magnitudes of body movement. We confirmed that the full model did explain the variance better than reduced versions of the model (likelihood ratio test) before performing post-hoc comparisons using the ‘lmerTest’ package in R (Kuznetsova et al., 2017).

To compare these measures across light intensities, we calculated the difference between the log-transformed magnitude spectra of thorax and abdomen position in bright and dim light for each antennal condition (Figure 4C,D). We then compared these using general linear models of the same form as above.

Sum-of-sines movement

Request a detailed protocol

We used system identification analysis (Cowan et al., 2014) to characterise hawkmoth flower tracking performance. This analysis is possible because the sum-of-sines stimulus fulfils the requirement of linearity, that is it generates the same flower tracking performance at different amplitudes and phase relationships (see Supplement, Stöckl et al., 2017a). Hawkmoth flower tracking can be described by two components: gain and phase (Farina et al., 1994; Sponberg et al., 2015; Stöckl et al., 2017a). Gain relates the amplitude of flower movement to hawkmoth movement (one for perfect tracking), while the phase describes the lead or lag of the hawkmoth with respect to the flower movement (0 for perfect tracking). We used a metric called tracking error ε (Roth et al., 2014; Sponberg et al., 2015), which incorporates effects of both gain and phase to quantify tracking performance of hawkmoths (Figure 3B). It is calculated as the complex distance between the moth’s response H(s) and the ideal tracking conditions (gain = 1, phase lag = 0), where s is the Laplace frequency variable:

(1) ϵ(s)=H(s)(1+0i)

A tracking error of 0 means perfect tracking (comprising a gain of 1 and a phase lag of 0), while the tracking error is one if the hawkmoth and flower movement are uncorrelated (e.g. when either the hawkmoth remains stationary and the flower moves, or vice versa). We calculated average tracking errors and their confidence intervals within antennal conditions by averaging data in the complex plane, to avoid artefacts resulting from separating gain and phase components when transforming them and averaging in the non-complex plane (see Stöckl et al., 2017a for discussion).

Since our tracking error metric is a complex value and was only transformed into the non-complex plane after averaging across individuals, it is not straightforward to find appropriate statistical tests to compare tracking error (as well as gain and phase) across antennal conditions and light intensities. Linear mixed effects models would be well suited, but complex data might not fulfil all of the assumptions these models are based on. Lacking an alternative, we had to rely on these tests, as did previous studies with the same approach (Roth et al., 2016; Sponberg et al., 2015) to compare the effect of antennal conditions across frequencies as fixed effects, including individual identity as a random effect. We are confident that overall trends identified as statistically significant by these models are indicative of biologically relevant effects, but advice caution when interpreting differences in significance at individual flower movement frequencies isolated from the overall trend.

To compare tracking performance across light intensities, we calculated the difference in the flower frequency at which tracking error reached one for both dim and bright light intensity within antennal conditions (similar to Stöckl et al., 2017a), and compared these across antennal conditions (Figure 4A). For statistical comparisons, we used the Friedman test, which is a non-parametric test that accounts for repeated measures.

Chirp movement

Request a detailed protocol

The chirp stimulus does not fulfil the linearity criterion, because it does not generate the same flower tracking performance at different amplitudes and phase relationships, but rather contains a saturation non-linearity which makes it increasingly harder for moths to track the flower with increasing flower frequency. Thus, the system identification analysis we used for the sum-of-sines stimulus could not be applied (Roth et al., 2014).

We therefore determined the flower frequency, at which each individual lost proboscis contact with the flower (i.e. failed at tracking the flower) as a measure of flower tracking performance across frequencies. This measure gave an absolute cut-off frequency at which moths could no longer track the oscillating flower. Because this data was non-parametric and included repeated measures, we used a Friedman test to compare the paired data. To compare the tracking performance across light intensities, we calculated the difference in flower frequency between dim and bright light at which each individual in each antennal condition stopped tracking the flower. These differences between light conditions were then compared across antennal conditions using a Friedman test to retain information about the paired data (Figure 4B).

In order to resolve the finer differences in flower tracking performance at the chirp stimulus between the three antennal conditions, we also performed a frequency analysis, and calculated the amplitude spectrum of the responses, as well as the cross power spectrum density and phase relationship between moth and flower tracks (Figure 3—figure supplement 3).

References

  1. 1
  2. 2
  3. 3
    Sensory control of abdomen posture in flying locusts’
    1. JM Camhi
    (1970)
    The Journal of Experimental Biology 52:533–537.
  4. 4
  5. 5
    Mlogit: Multinomial Logit Model
    1. Y Croissant
    (2013)
    R Package Version.
  6. 6
  7. 7
  8. 8
    The regulation of distance to dummy flowers during hovering flight in the hawk moth macroglossum stellatarum
    1. WM Farina
    2. D Varjú
    3. Y Zhou
    (1994)
    Journal of Comparative Physiology A 174:239–247.
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
    Visual and mechanosensory integration by descending interneurons in hawkmoths
    1. U Mohan
    2. M Maitri
    3. SP Sane
    (2017)
    Society of Integrative and Comparative Biology Abstract. Accessed January 7, 2017.
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
    The Gyroscopic Mechanism of the Halteres of Diptera
    1. JWS Pringle
    (1948)
    Philosophical Transactions of the Royal Society B: Biological Sciences 233:347–384.
    https://doi.org/10.1098/rstb.1948.0007
  29. 29
    ATask-Level Model for Optomotor Yaw Regulation in Drosophila Melanogaster: A Frequency-Domain System Identification Approachin ‘Decision and Control
    1. MB Reiser
    2. E Roth
    3. NJ Cowan
    4. MH Dickinson
    (2012)
    IEEE 51st Annual Conference On’, IEEE. pp. 3721–3726.
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51

Decision letter

  1. Ronald L Calabrese
    Reviewing Editor; Emory University, United States
  2. Eve Marder
    Senior Editor; Brandeis University, United States

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "The roles of vision and antennal mechanoreception in hawkmoth flight control" for consideration by eLife. Your article has been reviewed by three peer reviewers, including Ronald L Calabrese as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Eve Marder as the Senior Editor. The following individual involved in the review of your submission has agreed to reveal their identity: Noah Cowan (Reviewer #2).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

In this manuscript, the authors present an analysis of the interaction of antennal mechanosensory and visual input in the stabilization of hovering flight in a diurnal moth. The analysis compares three antennal conditions (intact control, antennae removed, antennae reattached) and two light conditions mid-range and low, each with two stimuli stationary flower feeding, and oscillating flower feeding consisting of two movements: chirp and sum of sines. They find that for both stimuli in the mid-range lighting, moths show degraded tracking at higher frequencies with removal of antennae and recover somewhat when antennae are reattached, compared to control (intact antennae). In low lighting, all three conditions behave similarly poorly, even at low frequencies. These results are consistent with the emerging conclusion that visual input is essential for moths to identify and track the flower movement relative to their own position – antennal mechanoreceptors cannot provide that required information but are necessary to support fast flight maneuvers. Vision and mechanoreception thus act in different frequency domains and do not compensate. These studies complement those showing similar interactions between haltere mechanoreception and vision in dipterans (most notably Drosophila) by extending them to the greater number of insect orders that lack halteres and thus make the findings of wide interest.

Essential revisions:

The expert reviews are provided, which will require rewriting and some new analyses. Important points include:

1) All three reviewers agreed in consultation that a control diagram is needed to help the reviewers contextualize the findings and point the way forward for further mechanistic analyses.

2) In discussion among the reviewers, there was some concern about the phase analysis and the reviewers discussed whether cross correlation would be a better strategy. However they concluded that cross correlation could be tricky with a chirp or sum of sines. Indeed, cross-correlation should be done at a given frequency. If these data are available then a cross correlation might be in order.

3) The expert reviews provided should all be addressed; they are detailed but consistent and complementary.

Reviewer #1

Concerns

1) I found it confusing that the essential idea or hypothesis that visual input is essential for moths to identify and track movement relative to their own position regardless of frequency is not made clear up front. The presentation could be made a lot clearer, if the contrast between mid-range and low light was presented first for the intact condition. Perhaps this problem will be ameliorated by the inclusion of a control diagram.

Reviewer #2:

Concerns

The most significant issue is the lack of a clear interpretation of the results. The results of this paper are quite interesting; the removal of a "postural" self-motion (similar to proprioception or vestibular) feedback system, the antennae, affects tracking of the exogenous motion of a flower. Why is this so interesting? Because of the subtle but evidently important interaction between these distinct feedback loops (diagram in Author response image 1):

One could explain (at least) qualitatively this interaction, as depicted in a graphical feedback control model such as the one above. It seems interesting that ablating the antenna disrupts the inner loop of the control system, making the outer loop (vision and mechanoreception from proboscis) not as effective at tracking. How does the modulation of these inner-loop dynamics (based on ablation) hinder outer-loop control? This is ripe for interesting computational modeling – such modeling itself could be saved for future work, but the description of this problem, which now is only vaguely hinted at, would elevate the paper substantially. The diagram above is just a rough cut and needs to be fleshed out, but I believe it to be a reasonable stab at the topology of the feedback system in question and if the authors agree, I encourage them to adapt and include something similar. Some possible issues with my above diagram that will require greater thought by the authors:

– the summing junction after the antennal and visuo-mechanosensory blocks is a simplifying assumption,

– before the plant perhaps there should be some sort of CNS integration

– Probably it is a +/- but the second junction may be better as rectangle labeled

"multisensory integration" where the inner-workings are left as future work.

In addition to that we have a number of detailed comments.

Detailed comments:

1) The authors use the word "significantly" even for data that, while statistically significantly different, are not that different from a controls engineering point of view. This is most notable in the discussion of Figure 3B, where the tracking errors of the re-attached and ablated moths were "significantly different" but compared to controls, they were quite similar.

2) Sometimes ablated/reattached moths perform more like controls, and sometimes they perform more like antennectomized moths. Can this be fleshed out a bit? For example as in point #1, the reattached and antennectomized moths performed similarly, but in the chirp task, control and re-attached were more similar.

3) It is unclear that how many trials are performed for each individual animal in each different case. It seems "one set of data", but it should be described in the paper. If you did not do more than one trial per animal (e.g. multiple passes of sums-of-sines), why not? If you did, how many did you do, how did you perform averaging, etc.? As long as that can be clarified, the results are compelling and seem to support the overall claims of the paper.

4) The paper has no citations to the Cohen lab papers that include models of haltere-based flight control in flies over the last 8 years that include roll, pitch, and yaw perturbations. There are crucial experimental differences; you perturb target motion (tracking task) and they provide a mechanical perturbation (i.e. a disturbance), so the experimental topology is different but perhaps their data could give you some insights into the inner-loop control structure?

5) The paper assumes that is the proboscis mechanoreception has "little or no" feedback contribution. Recent and crucial work by Roth et al. that the authors site quite clearly indicates otherwise for a related species of moth. This manuscript does use data in dim and bright light in the phase before proboscis contact with the nectary for a stationary flower (Figure 2—figure supplement 1), in which case this assumption seems valid, but mechanoreception is known to play a large role in tracking a moving flower (Roth et al.). That said, I'm not sure why this rather dubious assumption is needed.

6) Tracking error decreases significantly after reaching to its peak (Figure 3B, Figure 3—figure supplement 1). It is not intuitively clear why or how tracking error would decrease at higher frequencies. Is there any explanation for that? I have some concern that phase lags may have "wrapped" but that the analysis performed didn't identify that wrapping. Could that be possible? See especially Figure 3—figure supplement 1, last row.

7) As shown in Figure 1D, the power for chirp stimulus is much larger than the power for sum-of- sines. In subsection “Chirp movement”, it is mentioned that "chirp stimulus does not fulfill the linearity criterion".

Was this measured quantitatively? If so, how? Maybe if chirp had similar amplitude as the sum-of-sines then it would be linear. Is there a reason for chirp stimulus to have such large amplitude (and consequently high power)?

8) The frequency response could be calculated using the data from the chirp stimulus (by using cross spectral density and power spectral density). Also, in Figure 3C, using a second x-axis to show the frequency would be helpful.

9) The result for chirp stimulus tracking for control and reattached are very similar (Figure 3C, 3D). What is the explanation for this high similarity?

10) In the Results section paragraph four, the sentence is vague: "abdominal jitter of moths with re-attached flagella differed significantly from control moths at only one frequency (1.66 Hz)". What's the explanation for that? Why only one frequency? Is this meant to be the same frequency that is listed as 1.7 Hz in subsection “Behavioural experiments”?

11) In paragraph two of subsection “Flagella ablation reduces flower tracking performance at high frequencies in hawkmoths”, based on Figure 3B, the tracking error doesn't look significantly higher. Does it mean statistically significant?

12) Figure 4C, 4D: The plots in three colors have overlapped each other so much that they are unclear. Other methods can be used to show the change of noise in different frequencies.

13) In Figure 4C, it is unclear why the dimension of amplitude is shown as Hz.

Reviewer #3:

Concerns

– A parallel is made by authors between halteres role in Dipteran and flagella in hawkmoth. However, it is surprising that the work of Itai Cohen group was not cited in the Introduction (paragraph two). Cohen was probably the first to measure disturbance rejection in free flying fruitfly around the three axes of rotation (roll, pitch and yaw).

It is worth noting that halteres have an essential role in the stabilization of the fly, which is not the case in hawkmoth without flagella: the tracking accuracy is degraded but the flight is still stable in hawkmoth. Only the jitter seems to be higher without flagella but the flight remains stable in hover: the tracking seems to be slower indeed. I do not see any instability according to the definition proposed by control engineers. For example in figure 2, hovering flight with ablated flagella is not instable but less accurate. I suggest authors mention more a more accurate control than an instable control.

– Introduction paragraph four: it is mentioned that antennal mechanosensors play a key role but what kind of role? Authors should add a block diagram to clarify their model and to show clearly the closed-loop control the hawkmoth position:

– the tracking error between the insect and target position will be shown

– the inner loop based on the antenna block could be shown with respect to the control of the head in body orientation.

As the antennal ablation does not introduce instability but less accurate visual tracking, authors could discuss the fact that antennal mechanoreceptors could act as the prosternal organs in fly to allow the animal to measure the orientation of its head with respect to the insect body (head in boy orientation). As suggested by Viollet and Zeil, 2013, prosternal organs may be involved in a mechanoreceptive feedback on head position relative to the thorax. Antennal mechanoreceptors in hawkmoth could play a similar role: they could just improve the control (accuracy) of the head orientation (gaze).

– Subsection “Sum-of-sines movement”: about the tracking error, please clarify why you used this particular metric defined by equation 1.

– What kind of algorithms and software were used to estimate the gain and phase? I recommend having an open access to the programs (code).

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "The roles of vision and antennal mechanoreception in hawkmoth flight control" for further consideration at eLife. Your revised article has been favorably evaluated by Eve Marder (Senior Editor), a Reviewing Editor, and two reviewers.

The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:

The authors have done an excellent job of addressing the reviewer concerns except in two points.

Essential Revisions:

– Adding the detailed block diagram figure:

The expert reviewers are in strong agreement in consultation that the present control diagram is inaccurate. They recommend that the attached block diagram topology with inner loop be adopted. See the detailed reviews for their reasoning. If the authors insist that a 'simpler' diagram is needed, then the attached block diagram with NO inner loop may be used. Again see the detailed reviews for their reasoning.

– Phase analysis:

There are serious concerns about the phase analysis that were not adequately addressed in revision. See the detailed reviews for how these concerns can be addressed.

Reviewer #2:

Block diagram.

The authors introduce a different block diagram than what was suggested originally, they mentioned that they don't want to limit their conclusion to flower tracking so they introduce a more general block diagram. They want to mention the whole flight control in their block diagram, "not just tracking of a moving flower". The authors' new block diagram (Figure 1A) is unfortunately topologically incorrect. A "postural perturbation" is not defined in this paper anywhere that I can find but let us assume that when the diagram is "reduced" to the present paper, it is a moving flower (since that is all the paper addresses in the way of perturbations, although I wouldn't ever refer flower motion as a "postural perturbation" because it is a sensory perturbation that may lead to postural sway but only indirectly through the visuomotor system; not incorrect but not very clear either).

In any case, the moth's motion is subtracted from the flower motion before going into the visual system but self-motion is not subtracted from flower motion before going to the antennae (unless you are somehow modeling the antenna as a wind sensor which is unlikely). Since I am revealing myself in this review I would be happy to discuss this point further but as drawn this diagram does not make sense. The self-motion feedback to the antenna is grounded and is not subtracted from the sensory feedback. A perfectly acceptable approach would be to remove the subtraction bubble altogether, and draw it in the way that I have suggested.

(You would need to re-do the graphics inside the Sensory Input block to not give the sense that self-motion feedback only goes to the antennae.) The left arrow could be something like "Exogenous perturbation" and the return arrow could be "Self-motion feedback" or, maybe "Exafferent perturbation" for the left incoming error and "Reafferent feedback" for the return path.

Phase lag

Second, the problem with your currently unwrapped phase is that it is -π/2 at 10Hz, which means that the time lag between the input stimulus and the moth motion is 25ms, which seems extremely fast. I've never seen it in any visuomotor control paper in any species (from external motion to animal motion). If you look for example at Roth et al., 2016 he shows roughly 3π/4 to 2π (almost 360o) of phase lag, which corresponds to something like 75ms to 100ms which is a lot more sensible. This is very much in line with the feedback delay estimated by Sponberg, 2016 in their "Luminance dependence" paper. It is not possible it is only 25ms of total time lag (delay + low-pass mechanical phase lag). I don't see a problem with the approach but it is a completely unbelievable result. There are many possible sources of this.

One problem I've had is when I have data streams from different sources that get temporally offset or that temporally drift. This can introduce leads / lags. Another more likely possibility is that the roll off is so fast that your attempt to unwrap just misses it. See the incredible roll off in Eatai Roth's recent PNAS paper. The fastest latency I've seen estimated is Dyhr et al., 2013 (hawkmoth abdomen) which was 41ms but keep in mind that was just to the abdomen not the entire flight behavior, and it is quite possible that the high-pass "lead" filter helped mask some of that delay. Even still at 10Hz, the phase lag was π (180o) from stimulus to abdominal movement. The flight mechanics would surely introduce more phase lag.

Based on the error analysis in the complex plane (which I never really doubted even if I got a bit confused at one point – The explanation re: reduced tracking error at high frequencies is reasonable and I should have realized it before.) I don't think this will affect the main conclusions of the paper but it really does need to be addressed.

Reviewer #3:

There are just three points I would like to address again:

– the new block diagram is not accurate enough. I suggest separating the vision block from the antennal mechanosensors. A visual error can result from a difference between the moth's head orientation and the flower's position. These three signals (head's orientation, flower position and visual error) must be indicated on the diagram. This visual error can then be sent into the vision block, the output of which can be sent to the central integration block. As the mechanosensors seem to play a major role in the stabilisation of the moth, I suggest inserting this block in an inner loop with the motor system block. I agree that this point needs further experiments to determine precisely the function of the mechanosensors.

– Would it be possible that antennae act as a lead compensator (derivator) as the oscillations (jitter) are reduced? This point could be addressed in the discussion.

– Discussion about the role of the antenna to stabilise the head on the basis of the work of Viollet and Zeil (JEB paper) is not included.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "The roles of vision and antennal mechanoreception in hawkmoth flight control" for further consideration at eLife. Your revised article has been favorably evaluated by Eve Marder (Senior Editor), a Reviewing Editor, and one reviewer.

The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:

This is an unusual case where the reviewer's rationale for a rather minor required revision requires a rather long argument. Basically this reviewer calls for a caveat to be added to the Discussion in a short paragraph (or a few sentences to an existing paragraph). This caveat will not change the impact of the work, but will help the reader understand the rather remarkable tracking ability of this moth. Revision can be very swift and will not require re-review.

Reviewer #2:

The authors have done a remarkable job addressing my comments. I particularly appreciate their effort on technical issues such as providing extra data, revised PDFs, etc. Scientifically, I am convinced by the arguments in the revised manuscript; the updated block diagram and the other more minor issues I raised have also been addressed.

The one remaining issue about which I found myself concerned was the issue of phase lag. 90o at 8.9Hz is really quite extraordinary in the animal kingdom and having done system ID on moths, fish, humans, cockroaches, fruit flies, and even non-moving system ID on the jamming avoidance response in electric fish (where there is no "inertial low pass filter"), I've never seen such a short delay on a sensorimotor feedback loop except maybe from some work on haltere feedback. But not only have I been able to recapitulate the results based on their uploaded MATLAB data files, but also was able to reproduce the results from the raw image data, performing my own image tracking. In fact, in that video the phase lag at 8.9Hz is a mere 66o(with excellent SNR), corresponding to a mere 20ms visuo-movement response (total phase lag, including delay and mechanical phase lag). I just simply don't believe that is possible.

The authors claim that the animal is smaller and has a higher wingbeat frequency and therefore 'in-cycle' control would mean low phase lag, but I do not believe the synapses are any faster in this moth than in any other insect – and not even a fruit fly with hundreds of wingbeats per second can respond that fast to visual perturbations (fastest shown about 30ms I believe).

However, I do think I have a possible explanation, which is that there is a direct, mechanical coupling between the flower and the head of the moth via the proboscis. It is the only thing that I can see from the videos that could explain this extraordinary response. (Note that I measured to the tip of the head, not the thorax, when measuring the 20ms lag at 8.9Hz). That said, I think it is clear from the video that there is still a strong sensorimotor component to the behavior and the comparison being made in the paper includes this mechanical coupling for both antenna intact and antenna-ablated conditions. So I do not think that the possible mechanical coupling undermines the results in anyway, since the coupling was present in both conditions.

However, I would appreciate if the authors could acknowledge that these phase lags are unusually short (no known examples in the literature), and that some mechanical coupling may be playing a role. This can be a discussion point and needn't be a major point. It should also say that any such mechanical coupling would be present in both groups (intact vs ablated antenna), and doesn't impact the main findings of the paper.

As an extra step I looked at the bode plots from flower to thorax and flower to abdomen. The thought is that "yanking" the proboscis to the left and right might rotate the body quickly (and therefore show a low phase lag to the rostral end of the animal) but may not move the thorax directly. The moth seems to be rotating around the thorax. I did see much greater phase lags – on the order of 50ms – when I looked at the flower-to-thorax (near the rear of the thorax) transfer function.

https://doi.org/10.7554/eLife.37606.033

Author response

Essential revisions:

The expert reviews are provided, which will require rewriting and some new analyses. Important points include:

1) All three reviewers agreed in consultation that a control diagram is needed to help the reviewers contextualize the findings and point the way forward for further mechanistic analyses.

We agree with the reviewers that a control diagram is beneficial to visualize the findings, and to support future control theoretical analyses, and have therefore added it to Figure 1.

2) In discussion among the reviewers, there was some concern about the phase analysis and the reviewers discussed whether cross correlation would be a better strategy. However they concluded that cross correlation could be tricky with a chirp or sum of sines. Indeed, cross-correlation should be done at a given frequency. If these data are available then a cross correlation might be in order.

We are a little bit puzzled at the concern with our phase analysis we applied to the sum-of-sines data, as this has been used in similar contexts in a range of previous studies with very similar (if not identical) settings (Sponberg, et al., 2015, Roth, et al., 2016, Stöckl et al., 2017), and has a strong theoretical foundation (Roth, Sponberg, and Cowan, 2014). We have demonstrated how our data fulfils the criteria required for this type of analysis. We have responded to the specific concerns related to the analysis in detailed comments below. Should these not alleviate the concerns, then we would be very grateful for a discussion with the reviewers on how to best improve the analysis of the data.

3) The expert reviews provided should all be addressed; they are detailed but consistent and complementary. All minor comments should be addressed.

Below, we respond to all suggestions point by point, and highlight how we have implemented them in the revised manuscript.

Reviewer #1

Concerns

1) I found it confusing that the essential idea or hypothesis that visual input is essential for moths to identify and track movement relative to their own position regardless of frequency is not made clear up front. The presentation could be made a lot clearer, if the contrast between mid-range and low light was presented first for the intact condition. Perhaps this problem will be ameliorated by the inclusion of a control diagram.

As the reviewer suggests, we hope that the inclusion of a control diagram in Figure 1 helps to dispel any remaining confusions on this point, and also supports our discussion of the importance of vision for flight control in the introduction.

Reviewer #2:

Concerns

The most significant issue is the lack of a clear interpretation of the results. The results of this paper are quite interesting, the removal of a "postural" self-motion (similar to proprioception or vestibular) feedback system, the antennae, affects tracking of the exogenous motion of a flower. Why is this so interesting? Because of the subtle but evidently important interaction between these distinct feedback loops. (Diagram in Author response image 1):

Author response image 1

One could explain (at least) qualitatively this interaction, as depicted in a graphical feedback control model such as the one above. It seems interesting that ablating the antenna disrupts the inner loop of the control system, making the outer loop (vision and mechanoreception from proboscis) not as effective at tracking. How does the modulation of these inner-loop dynamics (based on ablation) hinder outer-loop control? This is ripe for interesting computational modeling – such modeling itself could be saved for future work, but the description of this problem, which now is only vaguely hinted at, would elevate the paper substantially. The diagram above is just a rough cut and needs to be fleshed out, but I believe it to be a reasonable stab at the topology of the feedback system in question and if the authors agree, I encourage them to adapt and include something similar. Some possible issues with my above diagram that will require greater thought by the authors:

– the summing junction after the antennal and visuo-mechanosensory blocks is asimplifying assumption,

– before the plant perhaps there should be some sort of CNS integration

– Probably it is a +/- but the second junction may be better as rectangle labeled

"multisensory integration" where the inner-workings are left as future work.

While we agree with the reviewer’s interpretation, we would like to highlight that it is only covering a part of our findings. We have also shown an impairment in flight control during the approach flight to the flower, as well as during hovering in front of a stationary flower, in addition to the impairments observed while the flower is moving and forcing the moths to perform flight manoeuvres to track it. We therefore conclude that our results show that the postural self-motion feedback is required for all aspects of hawkmoth flight, not just tracking of a moving flower. Thus, tracking of a moving flower is just one specific aspect of the effect we are describing, namely that hawkmoths require feedback from antennal mechanosensors for stable flight.

The other main finding of our study is that postural control through vision and antennal mechanosensation seem to operate in different frequency bands (as they do in flies). Thus, we suggest that both vision and antennal mechanosensation contribute to this postural control loop, and that they are parallel systems, which don’t seem to be able to compensate for each other.

Given our interpretation of the results, the fact that ablating the flagella makes tracking a moving flower less accurate is almost a necessary conclusion – because even if the moths can still perfectly determine the movement of the flower using vision and potential mechanosensory feedback from the proboscis, they cannot follow the movement due to their impaired postural control in the air. This leads to a worse flower tracking performance – not because the sensory information about flower movement is impaired, but because moths with ablated flagella are poorer at stabilizing flight than intact moths.

We have highlighted the roles of vision and antennal mechanosensation for postural control in a control model (Figure 1), which will help the interpretation our results. However, it is not clear how flower tracking and its sensory control interacts with the postural control circuit, and our study was not designed to investigate this question (but to lay the groundwork for the role of antennal mechanosensors in postural control in hawkmoths). Hence, we propose a control diagram that is more general, rather than a specific one that integrates flower tracking and postural control. We feel that a more meaningful model can be proposed once we have more information on the mechanisms of integration between the two control circuits, in a manner that is experimentally testable, which is not currently the case. We therefore agree with the reviewer that this would be a very interesting question for future studies.

In addition to that we have a number of detailed comments.

Detailed comments:

1) The authors use the word "significantly" even for data that, while statistically significantly different, are not that different from a controls engineering point of view. This is most notable in the discussion of Figure 3B, where the tracking errors of the re-attached and ablated moths were "significantly different" but compared to controls, they were quite similar.

We used the word “significant” in the sense of statistically significant, as this, in our eyes, is an objective way to describe the data. We are aware that this might not always relate to significant differences in control engineering points of view, or in biological terms, but we deemed it the least subjective way to determine differences. We therefore have now rephrased all our uses of the word “significant” to clarify its meaning.

2) Sometimes ablated/reattached moths perform more like controls, and sometimes they perform more like antennectomized moths. Can this be fleshed out a bit? For example as in point #1, the reattached and antennectomized moths performed similarly, but in the chirp task, control and re-attached were more similar.

We agree with the reviewer that this point has to be discussed at more length, and therefore expanded our discussion of this point in the manuscript (Discussion – Flagellar re-attachment improves flight performance). In short: we suggest that flagella re-attachment improves, but does not entirely restore flight performance (postural control) to control levels (as has also been observed in a previous study on hawkmoths, Sane et al., 2007). We discuss reasons for this in the manuscript. One likely reason why we observe slightly different trends between conditions in the different experimental paradigms is that they challenge the postural control system of the moths to different degrees. All experimental paradigms though show the same basic result: flagella re-attachment improves flight performance, but does not entirely restore it. This change in flight performance occurs mainly at higher frequencies of movement.

3) It is unclear that how many trials are performed for each individual animal in each different case. It seems "one set of data", but it should be described in the paper. If you did not do more than one trial per animal (e.g. multiple passes of sums-of-sines), why not? If you did, how many did you do, how did you perform averaging, etc.? As long as that can be clarified, the results are compelling and seem to support the overall claims of the paper.

We have now compiled the information about the number of animals and number of trials used in a section of the Materials and methods labelled “Datasets” and hope this makes it clearer. To summarize: the results shown in Figures 2-4 and related supplementary material were obtained from a paired experimental design: 6 animals were used in the stationary flower experiment, which performed one trial at the flower in each antennal condition and each light intensity. 12 different animals were used in the moving flower experiment, which similarly performed one trial (comprising the chirp and sum-of-sines stimulus) in each antennal condition and light intensity. Only the behavioural scores in Table 1 do not represent a balanced design. They were obtained from all animals that participated in the moving flower experiments, including those that only performed in a single or a few conditions – which has been considered in the statistical analysis (for details see Materials and methods, Datasets).

4) The paper has no citations to the Cohen lab papers that include models of haltere-based flight control in flies over the last 8 years that include roll, pitch, and yaw perturbations. There are crucial experimental differences; you perturb target motion (tracking task) and they provide a mechanical perturbation (i.e. a disturbance), so the experimental topology is different but perhaps their data could give you some insights into the inner-loop control structure?

We appreciate the reviewer’s suggestion and have added a reference to the work of the Cohen group to this the section about flight stabilisation in the Introduction.

5) The paper assumes that is the proboscis mechanoreception has "little or no" feedback contribution. Recent and crucial work by Roth et al. that the authors site quite clearly indicates otherwise for a related species of moth. This manuscript does use data in dim and bright light in the phase before proboscis contact with the nectary for a stationary flower (Figure 2—figure supplement 1), in which case this assumption seems valid, but mechanoreception is known to play a large role in tracking a moving flower (Roth et al). That said, I'm not sure why this rather dubious assumption is needed.

We agree with the reviewer that it the question whether the hawkmoths use mechanosensory feedback from their proboscis to track flowers does not affect our results, and we therefore removed this discussion from our manuscript.

6) Tracking error decreases significantly after reaching to its peak (Figure 3B, Figure 3—figure supplement 1). It is not intuitively clear why or how tracking error would decrease at higher frequencies. Is there any explanation for that? I have some concern that phase lags may have "wrapped" but that the analysis performed didn't identify that wrapping. Could that be possible? See especially Figure 3—figure supplement 1, last row.

Phases are not unwrapped in the analysis and calculation of the tracking error, as these all took place in the complex plane (for detailed discussion of phase wrapping, see Sponberg, 2015, supplementary material. We used exactly the same analysis methods). For the visual presentation in Figure 3—figure supplement 1 we did unwrap the phases – but again, this did not affect any of the tracking error calculations.

Tracking error decreases after the peak, because the animals are barely tracking these frequencies. Since tracking error describes the distance of the moth’s gain and phase from ideal tracking (gain=1, phase=0) in polar coordinates, if the moth has 0 gain, it has by definition a tracking error of 1. This is very nicely visualised in Figure S4 in Sponberg et al., 2015, which I have reproduced for clarification here.

7) As shown in Figure 1D, the power for chirp stimulus is much larger than the power for sum-of- sines. In subsection “Chirp movement”, it is mentioned that "chirp stimulus does not fulfill the linearity criterion".

Was this measured quantitatively? If so, how? Maybe if chirp had similar amplitude as the sum-of-sines then it would be linear. Is there a reason to have such large amplitude (and consequently high power) for chirp stimulus?

The linearity required for this type of analysis as laid out by Roth et al., 2014 and Sponberg et al., 2015, demands that the tracking performance (the specific shape of the gain and phase responses) does not depend on the amplitude or phase relationship of the stimulus (within reasonable limits). We therefore determined the linearity of the sum-of-sines stimulus for this hawkmoth species in previous experiments, using a smaller amplitude and a new set of randomised phases for the different flower movement frequencies (see Stöckl et al., 2017, Figure S1).

The chirp stimulus was made to test the limits of the moth’s tracking performance, and forces some (especially in the ablated condition) to abort tracking, as they are not able to follow the movement of the flower at the higher speeds any more. In a sense, it was built to expose non-linearities, in the failure of the moth’s tracking. Very likely, though we did not test this, moths would be able to track higher frequency flower movements at smaller amplitudes, and fewer at larger amplitudes – violating the scaling criterion of linearity. Indeed, it is possible that at smaller stimulus amplitudes, we might find a regime where the chirp stimulus produces responses fulfilling the linearity criterion. However, our stimulus does not, and therefore we chose to use a different type of analysis.

8) The frequency response could be calculated using the data from the chirp stimulus (by using cross spectral density and power spectral density). Also, in Figure 3C, using a second x-axis to show the frequency would be helpful.

A frequency axis in Figure 3C is a very good idea. We added it to the figure.

We also thank the reviewer for this suggestion of using a spectral analysis to look into the tracking of the chirp stimulus in more detail – as we could not use the system identification approach. We have added Figure 3—figure supplement 3 in the supplement, containing the amplitude spectra of the flight tracks of the different antennal conditions, as well as the cross power spectrum density and cross spectrum phase. We added the phase analysis, because the moths in the ablated condition have a rather high jitter in their flight tracks (as also shown in Figure 2). The phase analysis highlights at which frequencies the moth actually had a consistent phase relationship with the flower movement, and shows that moths with ablated flagella did not have it at the higher flower frequencies – in contrast to re-attached and control moths.

In essence, the spectral analysis shows a similar trend as the analysis of the sum-of-sines stimulus, and thus adds detail to our analysis we did not obtain before with just looking at the tracking abortion frequency: the re-attached moths range in between the ablated and the control ones in terms of flower tracking at the higher frequencies. Interestingly, this seemingly small difference in performance between moths with re-attached vs. ablated flagella was enough for most (though not all, see Figure 3D and Figure 3—figure supplement 2) of the re-attached moths to track the entire stimulus to the end like control moths, while all except one of the moths with ablated flagella had to abort flower tracking well before the end of the stimulus. We added the respective discussion to the manuscript as well (subsection “Flagella ablation reduces flower tracking performance at high flower movement frequencies” and “Sum-of-sines movement”).

9) The result for chirp stimulus tracking for control and reattached are very similar (Figure 3C, 3D). What is the explanation for this high similarity?

See our responses to point 2.

10) In the Results section paragraph four, the sentence is vague: "abdominal jitter of moths with re-attached flagella differed significantly from control moths at only one frequency (1.66 Hz)". What's the explanation for that? Why only one frequency? Is this meant to be the same frequency that is listed as 1.7 Hz in subsection “Behavioural experiments”?

We don’t have a mechanistic explanation for why only this particular frequency was statistically significantly different. Since it does not follow an overall trend (all other frequencies were not significantly different), and the frequency sampling was arbitrary (in the sense that we don’t know which frequencies are relevant for hawkmoth flight or if some are more than others), we would not put too much emphasis on this statistical result. However, it is part of our findings and as such we do report it.

The 1.7 Hz is the frequency at which the flower moves, while the 1.66 Hz is the frequency at which the abdomen jitters in the power spectrum. As has been pointed out by reviewer 1, the two different types of frequency analysis are confusing, and we have edited the manuscript to make clearer when we talk about flower frequencies and when we talk about the frequency analysis of the moth’s thorax or abdominal movement in stationary flower trials (see responses reviewer 1, point 1 for details).

11) In paragraph two of subsection “Flagella ablation reduces flower tracking performance at high frequencies in hawkmoths”, based on Figure 3B, the tracking error doesn't look significantly higher? Does it mean statistically significant?

Yes. We added this to all uses of the word “significant” in the manuscript, to avoid confusion (see responses to point 1 for details).

12) Figure 4C, 4D: The plots in three colors have overlapped each other so much that they are unclear. Other methods can be used to show the change of noise in different frequencies.

The fact that the three colours overlap is highlighting our results: there is no significant difference between the three conditions. We agree with the reviewer though that it is hard to identify individual traces, and therefore we have reduced the line width to make identification easier (see Figure 4).

13) In Figure 4C, it is unclear why the dimension of amplitude is shown as Hz.

This is a mistake on our side; the amplitude should be in “mm” and has been corrected.

Reviewer #3:

Concerns

– A parallel is made by authors between halteres role in Dipteran and flagella in hawkmoth. However, it is surprising that the work of Itai Cohen group was not cited in the Introduction (paragraph two). Cohen was probably the first to measure disturbance rejection in free flying fruitfly around the three axes of rotation (roll, pitch and yaw).

We appreciate the reviewer’s suggestion and have added a reference to the work of the Cohen group to this section of the Introduction.

It is worth noting that halteres have an essential role in the stabilization of the fly, which is not the case in hawkmoth without flagella: the tracking accuracy is degraded but the flight is still stable in hawkmoth. Only the jitter seems to be higher without flagella but the flight remains stable in hover: the tracking seems to be slower indeed. I do not see any instability according to the definition proposed by control engineers. For example in figure 2, hovering flight with ablated flagella is not instable but less accurate. I suggest authors mention a more accurate control than an instable control.

We do not entirely agree with this interpretation. The fact that the jitter in hovering is greater in moths without flagella than in intact moths (as the reviewer points out) is a strong indication that the flight is less stable without antennal mechanosensory feedback – since the moth cannot keep its hovering position reliably and has to use larger corrective movements (quantified as “jitter”). Moreover, when approaching the flower, moths without flagella also fly on much more tortuous paths (Figure 2—figure supplement 1.), giving more indications that flight stability in all flight modes is affected by flagella ablation, not just during flower tracking. We would also argue that the lower accuracy of flower tracking in flagella-ablated moths is due to a lack of control of body position (which also causes the elevated jitter, and the more tortuous flight paths) rather than an impairment in tracking, because all senses required for flower tracking itself are not impaired (vision and potentially proboscis mechanosensation).

We would therefore argue that flight in hawkmoths (in all modes: hovering on one spot, tracking the flower while hovering, forward flight etc.) are less stable (less controlled) without flagella – rather than that tracking is less accurate, as this is only a consequence of the less controlled / more unstable body position. We hope this becomes even clearer with the addition of the control diagram, where we highlight the contributions of the different senses to the different aspects of flight and flower tracking.

– Introduction paragraph four: it is mentioned that antennal mechanosensors play a key role but what kind of role? Authors should add a block diagram to clarify their model and to show clearly the closed-loop control the hawkmoth position:

– the tracking error between the insect and target position will be shown

– the inner loop based on the antenna block could be shown with respect to the control of the head in body orientation.

We agree with the reviewer that a control diagram is of great benefit to clarify our control hypothesis, and added it to Figure 1. See more detailed discussion on this in the responses to reviewer 2.

As the antennal ablation does not introduce instability but less accurate visual tracking, authors could discuss the fact that antennal mechanoreceptors could act as the prosternal organs in fly to allow the animal to measure the orientation of its head with respect to the insect body (head in boy orientation). As suggested by Viollet and Zeil, 2013, prosternal organs may be involved in a mechanoreceptive feedback on head position relative to the thorax. Antennal mechanoreceptors in hawkmoth could play a similar role: they could just improve the control (accuracy) of the head orientation (gaze).

As explained above, we do not entirely agree with the reviewer’s interpretation of our results that flagella ablation leads to less accurate visual tracking – and suggest that antennal mechanoreception is important for flight control, rather than visual tracking. It is possible, though, that the antenna contribute to head stabilisation – the experiments performed in our study are not suited to shed light on this question. Preliminary data are indeed suggesting that antennae might be involved in head stabilisation, in addition to visual mechanisms. We included the reference in the Discussion.

– Subsection “Sum-of-sines movement”: about the tracking error, please clarify why you used this particular metric defined by equation 1.

Because it is a well-established metric (Roth, Sponberg, and Cowan, 2014; Sponberg et al., 2015; Roth et al., 2016; Stöckl et al., 2017, to describe flight performance in these flower tracking experiments, and allows for a readout of an intuitive metric that combines both the effect on gain and phase.

– What kind of algorithms and software were used to estimate the gain and phase? I recommend having an open access to the programs (code).

We used exactly the same method as specified in the extensive supplementary material in Sponberg et al., 2015, theoretical foundation of which was laid out in Roth, Sponberg and Cowan, 2014. We made this clearer in our Materials and methods section now.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Essential Revisions

– Adding the detailed block diagram figure

The expert reviewers are in strong agreement in consultation that the present control diagram is inaccurate. They recommend that the attached block diagram topology with inner loop be adopted. See the detailed reviews for their reasoning. If the authors insist that a 'simpler' diagram is needed, then the attached block diagram with NO inner loop may be used. Again see the detailed reviews for their reasoning.

We did not intend for our schematic to be a control diagram for flower tracking, which would indeed be inaccurate, as reviewers 2 and 3 point out. We thank reviewer 2 for pointing out that our term “positional perturbations” can be misleading, and therefore agree with the reviewers that it should not remain in its current form. We thank the reviewers for understanding our unease with the detailed control diagram and their proposal of adding a simplified version. We have integrated this control diagram in Figure 1 in the revised version of the manuscript.

– Phase analysis.

There are serious concerns about the phase analysis that were not adequately addressed in revision. See the detailed reviews for how these concerns can be addressed.

We have responded to reviewer 2’s concerns about the phase analysis with specific reference to their comments (see below).

Reviewer #2:

Block diagram:

The authors introduce a different block diagram than what was suggested originally, they mentioned that they don't want to limit their conclusion to flower tracking so they introduce a more general block diagram. […] The left arrow could be something like "Exogenous perturbation" and the return arrow could be "Self-motion feedback" or, maybe "Exafferent perturbation" for the left incoming error and "Reafferent feedback" for the return path.

See our responses to the editor’s point 1.

Phase lag:

Second, the problem with your currently unwrapped phase is that it is -π/2 at 10Hz, which means that the time lag between the input stimulus and the moth motion is 25ms, which seems extremely fast. I've never seen it in any visuomotor control paper in any species (from external motion to animal motion). If you look for example at Roth et al., 2016 he shows roughly 3π/4 to 2π (almost 360o) of phase lag, which corresponds to something like 75ms to 100ms which is a lot more sensible. This is very much in line with the feedback delay estimated by Sponberg, 2016 in their "Luminance dependence" paper. It is not possible it is only 25ms of total time lag (delay + low-pass mechanical phase lag). I don't see a problem with the approach but it is a completely unbelievable result. There are many possible sources of this. One problem I've had is when I have data streams from different sources that get temporally offset or that temporally drift. This can introduce leads / lags. Another more likely possibility is that the roll off is so fast that your attempt to unwrap just misses it. See the incredible roll off in Eatai Roth's recent PNAS paper.

Both the moth’s and the flower’s movement were reconstructed from the same video frames, so it is very unlikely that the observed phase delays were influenced by an offset. We plotted the phase responses without unwrapping the phases, and thus can confirm that unwrapping did not influence the shape of the phase diagram (see below left, the control group at 3000 lux). We further confirmed this by using the same frequency analysis we used for the chirp stimuli in Figure 3–figure supplement 3 (using Matlab’s ‘cspd’ function, which does not unwrap the phases). The example below is also for the control group, 3000 lux, but we have confirmed this for all groups and light conditions. Note that in the analysis to the left, we only extracted gain and phases at the frequencies present in the stimulus – hence the slightly different shapes of the curves.

To avoid similar concerns about the phase unwrapping for the article’s future readers, we have now included the spectral analysis for the sum-of-sines stimulus in the supplement.

Author response image 2

The fastest latency I've seen estimated is Dyhr et al., 2013 (hawkmoth abdomen) which was 41ms but keep in mind that was just to the abdomen not the entire flight behavior, and it is quite possible that the high-pass "lead" filter helped mask some of that delay. Even still at 10Hz, the phase lag was π (180o) from stimulus to abdominal movement. The flight mechanics would surely introduce more phase lag.

We would like to present a few considerations that might render the short phase delays less incredible, compared to those observed previously. An important point to bear in mind is that the previous studies referenced were conducted on Manduca sexta, (Roth et al., 2016 and Dyhr et al., 2013), which is a nocturnal hawkmoth, with a much higher body weight and larger size than M. stellatarum. Typically, the motor responses of smaller insects are faster than the motor responses of larger insects. Please consider the following:

1) The wing beat frequency of M. stellatarum is 80 Hz, which is considerably faster than that of Manduca sexta which flaps at 20-25 Hz (see Stoeckl et al., 2017). Thus, the flight mechanics provide much less of a low-pass filter. If flight could be corrected within one wing beat, it would take 40-50 ms in M. sexta but only 12 ms in M. stellatarum. Moreover, the neural mechanisms controlling flight should be considerably faster if they operate at wing beat frequency.

2) In addition, typically visual responses in diurnal insects are faster than in nocturnal insects. As may be expected, the visual system of the diurnal M. stellatarum is distinctly faster than that of the crepuscular/nocturnal M. sexta (see Stoeckl et al., 2017).

For these reasons, a response time of 25 ms (roughly two wing beats) may not be as surprising as the reviewer suggests.

Importantly, however, interpreting a specific phase delay as a direct readout for the system’s response delay has several caveats of which we should be aware. Indeed, decrease in phase values might be caused by factors other than the “pure” tracking response of moths.

For instance, a random relation between the moth and the stimulus would lead to an average phase delay of 0 – albeit with a very high variance. This is visible in the analysis of flower tracking with the chirp stimulus in Figure 3—figure supplement 3. Here, one would not argue either that the decrease in phase delay starting around 4Hz is representative of a very fast visual delay – but rather of the moths no longer tracking the stimulus consistently and instead generating random phase delays, which eventually average to 0. A similar effect may occur for the sum-of-sines stimulus. Note especially in the above plot, on the right, how the phase approaches 0 for the “gap” in the stimulus around 7Hz.

Moreover, the moths are not “perfect” in their flight patterns – they always have a certain jitter in their position (not just jitter caused by the wingbeat frequency, but also at frequencies well below it). This is evident in the amplitude spectrum of the thorax’s position in Figure 2. This positional jitter may affect the gain and phase analysis of flower tracking to some degree, as it increases the gain at the jitter frequencies, and corrupts the phase relationship, since the positional jitter is not correlated with flower movement.

A superposition of the phase delays resulting from tracking (which decrease consistently as a function of frequency) with those resulting from positional jitter (which should cluster on average around 0) can result in intermediate phase responses. This may be less of a concern at lower flower movement frequencies, where the amplitude of flower movement far exceeds the positional jitter amplitudes, but it exerts greater influence at higher flower movement frequencies, at which the amplitude of flower movements is only around 0.4 mm.

It is noticeable that the positional jitter is very distinct in the ablated and re-attached groups between 6 and 11 Hz for stationary hovering – and that in these groups the flower tracking gain rises again, starting at 6 Hz, while the phase delay decreases. In the control group, for which the jitter at these frequencies is less pronounced, there is no rise in gain and associated decrease in phase delay, but a levelling in both parameters, which might be caused by the positional jitter. As a note: this levelling in gain and phase responses is something that has been observed in all hawkmoths tested with this paradigm so far (see Sponberg et al. 2015; Roth et al., 2016; Stoeckl et al., 2017).

In our analysis, it is not possible to separate potential positional jitter from “true” flower tracking movements, because we cannot decide which part of the movement is “intentional” and which is not. Hence, it is important to be cautious when interpreting the exact magnitude of the gain and phase delay at the small stimulus amplitudes. In future experiments, we aim to avoid such small flower movement amplitudes, which are confounded with self-generated positional jitter in the animals.

As Reviewer 2 points out, these considerations are important but do not affect the conclusions of our study, especially since the “unusual” parts of the phase response occur at frequencies greater than 5 Hz, at which the tracking gain is very low, and the tracking error in all three species approaches 0. Our analysis of differences in tracking error is focused on lower frequencies, at which the interesting differences between species occur.

Based on the error analysis in the complex plane (which I never really doubted even if I got a bit confused at one point – The explanation re: reduced tracking error at high frequencies is reasonable and I should have realized it before.) I don't think this will affect the main conclusions of the paper but it really does need to be addressed.

Reviewer #3:

There are just three points I would like to address again:

– the new block diagram is not enough accurate. I suggest separating the vision block from the antennal mechanosensors. A visual error can result from a difference between the moth's head orientation and the flower's position. These three signals (head's orientation, flower position and visual error) must be indicated on the diagram. This visual error can then be sent into the vision block, the output of which can be sent to the central integration block. As the mechanosensors seem to play a major role in the stabilisation of the moth, I suggest inserting this block in an inner loop with the motor system block. I agree that this point needs further experiments to determine precisely the function of the mechanosensors.

See our responses to the editor’s point 1.

– Would it be possible that antennae act as a lead compensator (derivator) as the oscillations (jitter) are reduced? This point could be addressed in the discussion.

It is possible that ultimately, the function of the control circuits that the mechanosensory input feeds into, could be described as a lead compensator in control theory. We have not investigated the neural control of the described behaviour and did not build and test a control theory model of the behaviour of nervous system.

– Discussion about the role of the antenna to stabilise the head on the basis of the work of Viollet and Zeil (JEB paper) is not included.

In the previous version of the manuscript, we included a short discussion of the possibility that antennal feedback is required for head stabilisation, and also referred to preliminary data giving evidence for this. We now added the Viollet and Zeil paper to this Discussion (subsection “Flagellar re-attachment improves flight performance”).

[Editors' note: further revisions were requested prior to acceptance, as described below.]

This is an unusual case where the reviewer's rationale for a rather minor required revision requires a rather long argument. Basically this reviewer calls for a caveat to be added to the Discussion in a short paragraph (or a few sentences to an existing paragraph). This caveat will not change the impact of the work, but will help the reader understand the rather remarkable tracking ability of this moth. Revision can be very swift and will not require re-review.

Reviewer #2:

The authors have done a remarkable job addressing my comments. I particularly appreciate their effort on technical issues such as providing extra data, revised PDFs, etc. [...] The moth seems to be rotating around the thorax. I did see much greater phase lags – on the order of 50ms – when I looked at the flower-to-thorax (near the rear of the thorax) transfer function.

We thank reviewer 2 for their insightful analysis and comments on the moth flower tracking responses. As suggested by the editor, we have added a paragraph of Discussion to the manuscript, were we highlight that the short delays in flower tracking are exceptional when it comes to sensorimotor (and visuomotor) feedback, and that a possible explanation for this might be mechanical coupling between the flower and the hawkmoth via the proboscis. We highlighted the respective paragraph (paragraph two “Antennal mechanosensation and vision operate in different frequency bands”) in colour for easy visibility.

https://doi.org/10.7554/eLife.37606.034

Article and author information

Author details

  1. Ajinkya Dahake

    1. Vision Group, Lund University, Lund, Sweden
    2. National Centre for Biological Sciences, Tata Institute of Fundamental Research, Bangalore, India
    Contribution
    Data curation, Formal analysis, Investigation, Writing—review and editing
    Contributed equally with
    Anna L Stöckl
    Competing interests
    No competing interests declared
  2. Anna L Stöckl

    Vision Group, Lund University, Lund, Sweden
    Present address
    Chair of Behavioral Physiology and Sociobiology, Würzburg University, Würzburg, Germany
    Contribution
    Data curation, Formal analysis, Visualization, Methodology, Writing—original draft, Writing—review and editing
    Contributed equally with
    Ajinkya Dahake
    For correspondence
    anna.stoeckl@uni-wuerzburg.de
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-0833-9995
  3. James J Foster

    Vision Group, Lund University, Lund, Sweden
    Contribution
    Formal analysis, Writing—review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4444-2375
  4. Sanjay P Sane

    National Centre for Biological Sciences, Tata Institute of Fundamental Research, Bangalore, India
    Contribution
    Conceptualization, Supervision, Funding acquisition, Writing—review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-8274-1181
  5. Almut Kelber

    Vision Group, Lund University, Lund, Sweden
    Contribution
    Conceptualization, Resources, Formal analysis, Supervision, Funding acquisition, Methodology, Project administration, Writing—review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-3937-2808

Funding

Vetenskapsrådet (VR621‐2012‐2212)

  • Almut Kelber

Knut och Alice Wallenbergs Stiftelse

  • Almut Kelber

Carl Tryggers Stiftelse för Vetenskaplig Forskning (15:108)

  • James J Foster

Erasmus+ (Erasmus Mundus Scholarship)

  • Ajinkya Dahake

Air Force Office of Scientific Research (FA2386‐11‐1‐ 4057)

  • Sanjay P Sane

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Merry and Leigh Foster for help with capturing the parental moths in France, Michael Pfaff and Joaquin Goyret for performing pilot experiments, Simon Sponberg for installing our robotic flower set-up and for critical comments on the manuscript, Marie Dacke for allowing us to use two high speed cameras, David O'Carroll for inspiring discussions, Karin Nordström for valuable comments on the manuscript, and Eric Warrant for financial support of AS.

Senior Editor

  1. Eve Marder, Brandeis University, United States

Reviewing Editor

  1. Ronald L Calabrese, Emory University, United States

Publication history

  1. Received: April 16, 2018
  2. Accepted: December 8, 2018
  3. Accepted Manuscript published: December 10, 2018 (version 1)
  4. Version of Record published: December 21, 2018 (version 2)

Copyright

© 2018, Dahake et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,472
    Page views
  • 181
    Downloads
  • 3
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Neuroscience
    Matthew J Davidson et al.
    Research Article Updated

    Research on the neural basis of conscious perception has almost exclusively shown that becoming aware of a stimulus leads to increased neural responses. By designing a novel form of perceptual filling-in (PFI) overlaid with a dynamic texture display, we frequency-tagged multiple disappearing targets as well as their surroundings. We show that in a PFI paradigm, the disappearance of a stimulus and subjective invisibility is associated with increases in neural activity, as measured with steady-state visually evoked potentials (SSVEPs), in electroencephalography (EEG). We also find that this increase correlates with alpha-band activity, a well-established neural measure of attention. These findings cast doubt on the direct relationship previously reported between the strength of neural activity and conscious perception, at least when measured with current tools, such as the SSVEP. Instead, we conclude that SSVEP strength more closely measures changes in attention.

    1. Neuroscience
    Margaret M Cunniff et al.
    Research Article Updated

    Many genes have been linked to autism. However, it remains unclear what long-term changes in neural circuitry result from disruptions in these genes, and how these circuit changes might contribute to abnormal behaviors. To address these questions, we studied behavior and physiology in mice heterozygous for Pogz, a high confidence autism gene. Pogz+/- mice exhibit reduced anxiety-related avoidance in the elevated plus maze (EPM). Theta-frequency communication between the ventral hippocampus (vHPC) and medial prefrontal cortex (mPFC) is known to be necessary for normal avoidance in the EPM. We found deficient theta-frequency synchronization between the vHPC and mPFC in vivo. When we examined vHPC–mPFC communication at higher resolution, vHPC input onto prefrontal GABAergic interneurons was specifically disrupted, whereas input onto pyramidal neurons remained intact. These findings illustrate how the loss of a high confidence autism gene can impair long-range communication by causing inhibitory circuit dysfunction within pathways important for specific behaviors.