Timing precision in fly flight control: integrating mechanosensory input with muscle physiology

Abstract

Animals rapidly collect and act on incoming information to navigate complex environments, making the precise timing of sensory feedback critical in the context of neural circuit function. Moreover, the timing of sensory input determines the biomechanical properties of muscles that undergo cyclic length changes, as during locomotion. Both of these issues come to a head in the case of flying insects, as these animals execute steering manoeuvres at timescales approaching the upper limits of performance for neuromechanical systems. Among insects, flies stand out as especially adept given their ability to execute manoeuvres that require sub-millisecond control of steering muscles. Although vision is critical, here I review the role of rapid, wingbeat-synchronous mechanosensory feedback from the wings and structures unique to flies, the halteres. The visual system and descending interneurons of the brain employ a spike rate coding scheme to relay commands to the wing steering system. By contrast, mechanosensory feedback operates at faster timescales and in the language of motor neurons, i.e. spike timing, allowing wing and haltere input to dynamically structure the output of the wing steering system. Although the halteres have been long known to provide essential input to the wing steering system as gyroscopic sensors, recent evidence suggests that the feedback from these vestigial hindwings is under active control. Thus, flies may accomplish manoeuvres through a conserved hindwing circuit, regulating the firing phase—and thus, the mechanical power output—of the wing steering muscles.

1. Introduction

Timing plays a critical role in the nervous system, where the ability to rapidly detect and process subtle disturbances in the environment determines whether an animal can attain its next meal or successfully navigate complex, unpredictable terrain. Previous work on a number of animals has made tremendous strides toward uncovering the specialized biomechanical mechanisms and neural circuits used to resolve timing differences with nano- to microsecond resolution [1]. Although much of this work has focused on the detection of timing differences in sensory systems, the importance of sub-millisecond precision in structuring motor output is becoming increasingly appreciated [2]. Indeed, longstanding evidence shows that the timing of neural input determines how a muscle functions in the context of locomotion [3]. These issues converge in the context of insect flight, as these organisms perform behaviours at rapid timescales, suggesting that they operate near the performance limits of neuromechanical mechanisms for control.

Members of Diptera, the true flies, are among nature's most agile flying creatures [4,5]. This exquisite manoeuvrability is due in part to the rapid integration of multimodal sensory input from the visual system and wingbeat-synchronous mechanosensory feedback [6,7]. Furthermore, flies are unique in that they possess specialized mechanosensory organs known as the halteres, which until recently were primarily renowned for being the only true biological ‘gyroscopes,’ but may serve a broader role in flight control [8,9]. In the light of this new evidence, and the wings' established role in maintaining the flight rhythm [6,10], it seems appropriate to re-evaluate our knowledge of how mechanosensory input regulates the muscles that control flight manoeuvres. Here, I outline the role of timing in determining wing steering muscle physiology. I then address the sources of this timing information and detail multiple hypotheses of how flies modulate mechanosensory feedback to execute their aerial manoeuvres. Finally, I place flies’ flight control architecture in an evolutionary context, suggesting a parsimonious scenario that led to the evolution of the haltere. I focus my attention on the blowfly Calliphora vicina in which much of the original physiological work was conducted, and the fruit fly, Drosophila melanogaster, an organism in which modern genetic tools have deepened our insight into Dipteran flight control circuitry. Examining fly flight through this comparative lens allows for an understanding of how flies evolved their remarkable aerial manoeuvrability and provides powerful examples of how the timing of motor output is structured by precise sensory input.

2. Motor control of flight manoeuvres

Unlike other extant flying taxa such as birds and bats, all the musculature for power and control in flying insects resides in the thorax [11–16]. The majority of the thoracic volume is occupied by the power muscles, which are responsible for the back-and-forth motion of the wings and provide all the mechanical power for flight. There are two groups of power muscles: the dorsoventral muscles, which contract during the upstroke, and their antagonists, the dorsolongitudinal muscles, which contract during the downstroke [11,12]. The power muscles do not attach to the wing directly; instead, they insert onto the thorax, with their reciprocal contraction resulting in its deformation, which is transmitted to the wings via the complex wing hinge. As flying insects radiated, they became much smaller, and consequently, the required wingbeat frequency needed to support their weight increased. For example, whereas the large damselfly Megaloprepus coerulatus flaps its wings at only 5.5 Hz, the tiny midge Forcipomyia has a reported wingbeat frequency of 1046 Hz [11]. The typical so-called synchronous muscle which contracts in a one-to-one fashion with neural excitation would not be able to cycle calcium fast enough each wing stroke to produce the necessary forces to generate lift. In flies and many other orders, contraction of the power muscles is decoupled from their neural excitation; a single muscle action potential leads to multiple contraction cycles. As a result, they are termed asynchronous flight muscle. Asynchronous muscle has evolved independently at least seven times in flying insects and reflects one solution regarding the relative importance of power versus precision in muscle function.

The other side of this trade-off are the small synchronous steering muscles that regulate flight manoeuvres. These muscles insert directly on the wing hinge onto invaginated cuticular elements known as sclerites. The motion of the sclerites changes the conformation of the wing hinge, leading to subtle changes in wing motion and large changes in aerodynamic force production [4,12,13]. The wing hinge is made of four sclerites, known as the first to fourth axillaries. For historical reasons, the fourth axillary is known as the Hg. Additionally, the basalar sclerite, which is not part of the wing hinge, connects to the hinge via a ligamentous connection and influences wing motion. Each sclerite except for the second axillary is associated with a subset of steering muscles that insert onto it.

The flight steering system of Drosophila is composed of only 12 muscles divided among the four sclerites, compared with 13 in Calliphora (figure 1a), and each steering muscle is innervated by a single motor neuron. Previous electrophysiological recordings in Calliphora along with recent calcium imaging experiments in Drosophila revealed that the wing steering system can be further divided into two broad classes: the tonic and phasic muscles ([17–20]; figure 1b). Tonic muscles regulate stabilization reflexes and are active throughout flight, firing once per wingstroke at a precise time, or phase, in the stroke cycle (figure 1c). The only mechanism available to flies for controlling tonic muscle activity is regulating the phase of activation [6,18–22]; flies are largely unable to regulate the spike rate of tonic muscle firing, with some exceptions [22]. Phasic muscles control active manoeuvres and are typically quiescent, firing in bursts where each action potential is phase-locked to the stroke cycle [18–20,23]. The classification of the wing steering muscles into tonic and phasic can be further subdivided by anatomical location (figure 1b). At least one muscle of each physiological class inserts onto each sclerite, and each sclerite is hypothesized to control a specific mode of wing motion [17].

Figure 1.

Anatomy and function of the dipteran wing steering system. (a) Steering manoeuvres in Drosophila and other flies are controlled by a set of steering muscles that attach to sclerites, which control the conformation of the wing hinge. Redrawn from Lindsay et al. [17]. (b) The muscles associated with each sclerite can be divided into two functional classes: muscles that are tonically active (yellow) or phasically recruited (red) during flight. (c) The tonic muscles allow fine-scale control of manoeuvres, whereas phasic muscles are engaged for large, rapid manoeuvres. Both types of steering muscles receive descending visual commands and are activated at precise phases of the wingstroke. Arrows do not indicate the size of manoeuvres controlled by each muscle type. (Online version in colour.)

A preparation known as the work-loop technique has demonstrated the importance of the activation phase for determining muscle function [3]. By contrast to experiments where muscle function is examined under isometric (constant length) or isotonic (constant force) conditions, the work-loop method is most suitable for muscles that undergo cyclic changes in length, as during locomotion. Imposing cyclic length changes while stimulating the muscle at a specific phase of oscillation and measuring force output allows calculation of the muscle's work per cycle, or mechanical power output. This technique has been used in animals across a range of taxa to demonstrate that depending on the locomotor context, muscle can act as an actuator, brake or strut [3,21,24–29]. Such biomechanical analyses on the tonic first basalar muscle (wB1) demonstrated that this muscle exhibits negative power output at wingbeat frequency; it effectively operates as a variable stiffness spring to absorb energy during flight [21]. The mechanical power output of wB1 has important consequences for wing motion and the production of flight forces. Phase advances of wB1 are associated with increased wingbeat amplitude, and thus greater aerodynamic forces [18,19,30]. Experiments directly stimulating the phasic second basalar wing steering muscle, wB2, at different points in the stroke cycle showed that wB2's effect on wingbeat amplitude is highly sensitive to the activation phase [23]. Thus, through the action of a limited set of muscle and motor neuron pairs, flies can finely control wing motion and aerodynamic forces at rapid timescales.

3. How do flies regulate wing steering muscle activity?

Flies’ reliance on the timing of motor signals to control wing motion necessitates the question of the source of this precise phase control. Unlike the power muscles, the wing steering muscles are not supplied by any central pattern generator [12,31,32]. Descending commands from the brain are one potential source for controlling the steering muscles. Indeed, anatomical work across dipterans shows that a number of multimodal descending interneurons project from the brain to the wing neuropil of the ventral nerve cord (VNC; [33–35]). However, there is not yet any electrophysiological evidence that the brain sends commands to the wing steering muscles in a phase-locked manner [36]. Instead, descending commands are sent via slow, graded potentials that are consistent with a spike rate coding mechanism, not timing.

Previous researchers hypothesized that flies use wingbeat-synchronous mechanosensory feedback to regulate the steering muscle activation phase ([6,32]; figure 2). At the wing base and along its blade, flies are equipped with dozens of mechanosensory structures known as campaniform sensilla. These dome-shaped sensors possess a single neuron that is directly underneath a cap embedded in spongy tissue [37]. Cuticular strain due to bending or twisting results in the cap rising or falling, pulling on the neuron's dendrite and eliciting an action potential that is phase-locked to the stimulus cycle ([38]; figure 2a). Campaniform sensilla are found across insects, particularly in places that experience significant bending or twisting, such as the wings or legs [37,39–44]. Experiments systematically ablating each wing showed that flies only need mechanosensory feedback from one wing to entrain the firing phase of wB1 [6,32]. Thus, the previous working hypothesis of how flies regulate wing steering muscle motor neuron phase was that slow descending visual input converges with rapid mechanosensory feedback at the level of the motor neuron (figure 2b; [6]). Depending on the direction of visual motion, this multimodal feedback will shift the firing phase of the motor neuron, changing the biomechanical properties of the muscle (figure 2b). These sub-millisecond timing shifts will very slightly change the fly's stroke kinematics, but have large consequences on the resulting aerodynamic forces and moments.

Figure 2.

Flies integrate visual and mechanosensory input to regulate wing steering muscle phase or activation. (a) Hypothesis of how flies structure the firing of wing steering muscle motor neurons, redrawn from Heide [6]. Relatively slow, graded visual commands (green) are combined with wingbeat-synchronous mechanosensory input (purple) from embedded sensors from the wings and halteres, either adjusting firing phase (Δφ, black and yellow) or gating (red) of tonic or phasic wing steering muscle motor neurons, respectively. (b) The effects of changing the activation phase of the tonic wing steering muscle wB1. This wing steering muscle has a preferred phase of activation, φ₀, near the downstroke–upstroke transition. Descending commands shift the firing phase to an earlier or later portion of the stroke cycle, φ_0+vis, which changes the mechanical power output of the muscle. These biomechanical changes either increase or decrease muscle stiffness each wingstroke, resulting in subtle kinematic changes and large modulations in aerodynamic force production. (Online version in colour.)

Recent computational studies and behavioural experiments support this form of multimodal integration [45–47]. Simple models can recapitulate the behaviour of the wB1 or wB2 motor neurons just by layering visual feedback on top of wingbeat-synchronous mechanosensory feedback [46]. Moreover, flies can perform visually mediated steering manoeuvres when the wings are the sole source of wingbeat-synchronous feedback following experimental ablation [45,47]. Yet, these responses are greatly diminished compared with intact flies, underscoring the importance of the full suite of mechanosensory feedback available to flies in flight. Moreover, wing motion—and the resulting mechanosensory feedback from wing campaniforms—are the direct result of the wing steering muscles. Wing campaniforms can only provide information about the current status of the wing, not the animal's future intentions. In addition, wing mechanosensory feedback entrains both the ipsilateral and contralateral wB1 [6], making manoeuvres that require asymmetric changes in wingstroke amplitude difficult. Thus, the wings and visual system may provide input to set the firing phase of the wing steering muscles. But the question of how flies adjust activation phase independently of the wingstroke to control manoeuvres remains.

4. The haltere, flies' metronome/gyroscope

Although the wings provide crucial mechanosensory input on a stroke-by-stroke basis, flies also rely on feedback from the halteres to structure the output of the wing steering system. These tiny, club-shaped structures beat during flight and serve no direct aerodynamic function. Nevertheless, halteres retain several features that are consistent with their evolutionary history as hindwings. For example, whereas the wings contain dozens of campaniform sensilla, these sensory structures have hypertrophied on the haltere such that Drosophila have 140 haltere campaniforms compared with the wing's 48 while Calliphora have 340 haltere campaniforms compared with 150 on the wing ([8,9,40,41,48–52], figure 3a). Halteres vary broadly regarding their relative size and shape; however, across species, haltere campaniforms are arranged in highly stereotyped groups [8,9,50]. Further, haltere and wing campaniforms are serially homologous [53]. Indeed, mutations of the Drosophila Hox gene Ultrabithorax transform the halteres into a second pair of wings [39]. Additionally, haltere motion is controlled by a power muscle—the first identified asynchronous muscle [54]—and a small set of steering muscles (figure 3b) [9,54–58].

Figure 3.

Halteres help flies maintain aerial stability. (a) Confocal micrograph of the dorsal surface of the right haltere of Drosophila. Each haltere possesses hundreds of strain-sensitive campaniform sensilla that are found on the base and stalk of the haltere's dorsal and ventral aspects. dF2: dorsal field 2; dF3: dorsal field 3. Scale bar: 50 µm. (b) In addition, the haltere has a power muscle, hDVM, that is responsible for driving its oscillations during flight and seven tiny steering muscles: the basalares (hB1, hB2) and the axillaries (hI1, hI2, hIII1–3), all of which receive visual input. Reproduced from Dickerson et al. [9]. (c) During flight, the haltere beats antiphase to the wings with an amplitude of nearly 180° (red). During rotational perturbations (black), the haltere experiences the Coriolis force, changing its tip path (blue) and altering the pattern of strain at the organ's base and stalk. These new strain patterns recruit additional campaniform sensilla each stroke. Field dF2 may be particularly sensitive to these out-of-plane motions, and may thus confer the haltere its gyroscopic capabilities. Haltere tip path changes due to Coriolis forces are exaggerated for clarity. (d) Haltere-mediated detection of the Coriolis force triggers equilibrium reflexes of the wings and neck, rapidly stabilizing flight posture and gaze. (Online version in colour.)

The proliferation of the haltere's sensory apparatus is such that this tiny organ can entrain the timing of the ipsilateral wB1 [6]. Based on the number of campaniform sensilla the haltere possesses, von Buddenbrock argued that halteres served as ‘stimulation organs’ for the flight circuit, providing the necessary input for the normal production of flight muscle forces [59]. More recent electrophysiological experiments have demonstrated that wing and haltere feedback structures wB1 firing on a wingstroke-to-wingstroke basis [10]. Thus, one major role for the haltere is to act as a metronome for the wing steering system by providing rhythmic mechanosensory feedback during normal flight.

In addition to their role as metronomes for the flight circuit, halteres are commonly known as the only true biological gyroscopes. Derham first noted that flies crash catastrophically once their halteres are removed and named them ‘haltere’ for their role in maintaining a fly's balance through the air [60]. Although this simple experiment elegantly demonstrated that halteres are necessary for flight, how these unique organs helped flies stay aloft remained elusive for centuries. In an extensive analysis that remains at the forefront of our understanding of haltere function after more than 70 years, Pringle analysed the dynamics of a moving haltere and the anatomy of its campaniforms [8]. He postulated that during rotations, the haltere experiences an inertial force that is orthogonal to its original plane of oscillation, also known as the Coriolis force:

$$F_{\mathrm{C}}=2m\omega \times v,$$ 4.1

where m is the mass of the haltere end knob, ω is the angular velocity of the body and v is the haltere tangential (tip) velocity. Coriolis forces would change the haltere tip path from a straight line to a figure-eight or oval, depending on the direction of rotation (figure 3c). Furthermore, Coriolis forces only occur in a rotating frame of reference, and later analyses showed that only Coriolis forces acting on the haltere can inform flies of the direction and magnitude in which they are rotating [61,62]. Thus, the haltere could help flies monitor their self-motion, similar to the vertebrate vestibular system [63–66].

A great deal of behavioural and electrophysiological work supports a gyroscopic function for the haltere. For example, flies that are subjected to mechanical rotational stimuli exhibit compensatory reflexes of the head and wings, which are abolished upon haltere removal [67–70]. Furthermore, the campaniform sensilla at the haltere's base, known as dorsal field 2 (dF2; figure 3a), provide direct, excitatory input to the ipsilateral wB1 motor neuron via a mixed electrotonic and chemical synapse [52,71,72]. Feedback from dF2 has the ability to induce a phase advance in wB1's firing pattern at specific stimulus intervals relative to wing mechanosensory feedback [10]. The sensilla of this field are arranged in a way that suggests they are most sensitive to the shear strains that occur when the haltere tip path is deflected as a result of Coriolis forces [8,73], and along with a similarly arranged chordotonal organ, may endow the haltere with its gyroscopic capacity. Also, haltere feedback is nonlinearly integrated with visual input to control the firing of neck muscle motor neurons, pointing to a role in gaze stabilization for extracting visual motion cues [74,75].

One expectation that follows from equation (4.1) is that gyroscopic sensing should exhibit the characteristics of a high-pass filter, with haltere-mediated reflexes being more sensitive to faster rotations. Behavioural experiments indicate that this is indeed the case, whereas visually mediated reflexes are more sensitive to slower rotations [67,70,76]. More recently, a number of papers have used a control-theoretic framework to model the haltere's role in flight control [77–82]. Flies subjected to an applied torque from a magnetic field with a metal pin that is dorsally tethered recover from this perturbation within five wingbeats [77]. Fitting the body dynamics to a linear control model suggests that in the context of rotational perturbations, haltere feedback acts as a proportional-integral controller on the body's angular velocity [77,79–81]. Such a control architecture may allow flies to correct for the current angular velocity error signal due to Coriolis forces via the proportional term while also adjusting for estimates of the past error via the integral term. Based on experiments using a dynamically scaled model of a flapping fly, a separate hypothesis instead suggests that the haltere provides active feedback during flight—particularly about the yaw axis—which can be modelled using a proportional controller alone [78,82]. This may allow the visual system to operate with higher sensitivity, or gain, without making the fly unstable. At this time, further experiments are needed to distinguish between these models. Regardless, flies combine haltere gyroscopic sensing with visual feedback to broaden their sensitivity to external perturbations during flight, making them remarkably stable.

Altogether, the working model of the haltere until recently was that it serves as a mostly passive sensor. Campaniform sensilla on the stalk fire single, phase-locked action potentials during normal beating [83–85], providing wingbeat-synchronous feedback to the wB1 muscle. This ensures that wB1 fires at the proper phase of the stroke cycle. Then, during sudden instabilities due to wind gusts, differential wing loading or wing damage, the haltere experiences Coriolis forces resulting from the fly's body rotation. The Coriolis forces alter the haltere's tip path trajectory, changing the pattern of strain at its base. This new strain pattern recruits additional campaniform sensilla, which fire at different phases of the stimulus cycle [85–87]. Through the campaniforms' connection to the wing steering system, this increased activity elicits a phase advance in wB1. As a result, wingbeat amplitude increases, resulting in increased aerodynamic force production, allowing the fly to correct for instabilities in its flight path (figure 3d).

5. The control-loop hypothesis: converting graded visual input into a phase code

The reflexes mediated by the halteres are extremely rapid. For example, stimulation of haltere campaniforms results in depolarizations of the wB1 motor neuron in approximately 4 ms [71,72]. The existence of such fast reflex loops connecting sensing to motor output via the haltere poses a serious problem to understanding how flies execute voluntary flight manoeuvres [88]. Recent work on a number of insect flight behaviours suggests a role for efference copy in suppressing visually mediated reflexes during voluntary manoeuvres [89–91]. In the case of silencing the haltere's reflexes during active turns, alternative strategies that take into account its evolutionary history as a hindwing exist. Previous electrophysiological studies on Calliphora demonstrated that the haltere steering muscles receive descending visual input [58]. This result suggests that flies may actively tune the strength of haltere mechanosensory feedback to achieve flight turns, rather than use the haltere for passively detecting gyroscopic forces. In what is known as the control-loop hypothesis, descending visual commands activate the haltere muscles, resulting in changes to the structure's motion or mechanics. The resulting changes alter the strain experienced by haltere campaniform sensilla, recruiting additional mechanosensors to fire each stroke cycle at different phases of haltere oscillation. Then, owing to the haltere's direct connection to the wB1 muscle, wingbeat amplitude, and thus aerodynamic forces increase, allowing the fly to change its heading. This way, the visual system could indirectly initiate voluntary changes in the flight path through its connection to the haltere steering muscles.

New evidence using the neurogenetic tools uniquely available in Drosophila has confirmed the control-loop hypothesis as a plausible mechanism for regulating the timing and activation of the wing steering system via three lines of evidence [9]. First, the haltere steering muscles are tuned to rotations about the body's cardinal axes, matching an identified descending interneuron (DNHS1) that projects to the haltere neuropil of the VNC [36]. This result suggests that the visual system controls the haltere muscles, implying the modulation of mechanosensory input via visual input [9]. Second, the haltere primary afferent projections in the brain are also responsive to global visual rotations during tethered flight, indicating that haltere motion is controlled in a similar manner to the wings. Finally, activation of the haltere steering muscle motor neurons either alters the activation phase of wB1 or recruits wB2. Taken together, these data are the first direct observations that flies can use changes in haltere steering muscle activity to co-opt mechanosensory reflexes, thereby enabling the execution of voluntary manoeuvres.

Although the control-loop hypothesis is a mechanism unique to flies, it is based on the haltere's evolutionary history. Indeed, past work in locusts demonstrates that mechanosensory feedback from the hindwing is strongly coupled to the motor neurons of the forewing control muscles [42,92,93]. In addition, ablating hindwing mechanosensory feedback in the hawkmoth Manduca sexta alters forewing kinematics and resultant lift force production [94]. Taken together, the picture that emerges is one where the regulation of wing steering muscle activation phase was the principal selective pressure that led to the evolution of the haltere (figure 4; [59,60]). In this scenario, flies' evolutionary ancestor was a large, four-winged flyer that used a central pattern generator to generate the flight rhythm, layering wingbeat-synchronous mechanosensory feedback and descending visual commands for steering commands (figure 4a). Owing to the tight neuromechanical coupling between the fore- and hindwings, this insect could maintain flight stability but had poor manoeuvrability. As flying insects radiated and decreased in body size, the evolution of asynchronous flight muscle marked a transition from a neural oscillator to a biomechanical oscillator for powering flight. While this mechanism enabled small fliers to achieve the necessary high wingbeat frequencies to stay aloft, the steering muscles still required precise timing input for executing manoeuvres. Thus, wingbeat-synchronous mechanosensory feedback became an essential modality for controlling wing motion (figure 4b). These flies may have been more manoeuvrable compared with their ancestor because of their reduced body size. However, the activation phase of the steering muscles in these animals would still depend on sensory feedback due to bending of the fore and hindwings.

Figure 4.

Proposed scenario leading to the evolution of halteres. (a) Flies' evolutionary ancestor was a four-winged flyer that used a central pattern generator along with sensory feedback. (b) The radiation of flying insects led to a decrease in body size along with necessary increase in wingbeat frequency, leading to the evolution of asynchronous flight muscle. In many small fliers, the timing of the control muscles may depend on mechanosensory feedback from the hindwing. (c) In flies, the control-loop pathway allows the regulation of wing steering muscle timing or activation separate from the detection of aerodynamic forces via the haltere. Visual commands (1) to the haltere muscles (2) leads to the recruitment of additional campaniform sensilla at each stroke (3); changing the firing phase of tonic wing steering muscles activates phasic muscles (4). The haltere's gyroscopic sensing may be an elaboration of its timing function. Reproduced from Dickerson et al. [9]. (d) Flies' evolutionary ancestor had a relatively large body size and low wingbeat frequency, making it passively stable in flight but with poor manoeuvrability. The evolution of Diptera—and the haltere's dual role as an adjustable timing mechanism and gyroscope—allow these insects to both reject instabilities and rapidly initiate changes in their flight trajectories. (Online version in colour.)

Relying on both wings for producing aerodynamic forces and controlling the timing of the wing steering system would require complex information processing to disambiguate wing bending due to normal flapping from perturbations or voluntary steering commands. Although recent modelling efforts suggest that such processing could be accomplished by the wings [95], the evolution of the haltere and the control-loop pathway represents a clean division of labour between the production of aerodynamic forces and regulating the precision of motor output. The wings produce flight forces and their embedded campaniforms set the rhythm of the wing steering system. Meanwhile, the halteres convert visual commands into a phase code for the wing steering muscles via the haltere motor system, dynamically controlling the steering muscle activation phase or recruitment on a stroke-by-stroke basis, independently of the wings (figure 4c). The detection of rotational perturbations via the Coriolis force is, therefore, a likely elaboration of this conserved circuit. Thus, although commonly described as flies’ ‘gyroscope,’ a more accurate description of the haltere is that it is a multifunctional sensory organ; the haltere enables flies to maintain stability in the face of external perturbations without sacrificing manoeuvrability at rapid timescales (figure 4d). Furthermore, the haltere's direct connection to the neck motor system and its crucial role in gaze stabilization suggest that the visual system regulates the level of haltere feedback and vice versa.

6. Future directions

The potential for understanding how precise mechanosensory feedback structures the timing of motor output is greater than ever before, especially with the advent of new genetic tools for labelling and manipulating specific cell types in Drosophila. Moreover, all flying insects possess four wings or wing-like structures, making the control-loop hypothesis a useful starting point for studying the role of timing across insects, although the role of local VNC circuits or descending commands cannot be excluded. Thus, as genome editing approaches become more tractable in traditional model systems of insect flight that are not genetic models, we will gain comparative insight into this problem at a mechanistic level. The use of these modern methods will highlight the tremendous importance of timing in neural systems as well as the selective pressures that led to the diversification of nature's most successful organisms.

Data accessibility

This article has no additional data.

Competing interests

I declare I have no competing interests.

Funding

This work was supported by startup funds from UNC Chapel Hill and NSF award IOS-2006284.