Inertial Encoding Mechanisms and Flight Dynamics of Dipteran Insects

INERTIAL ENCODING MECHANISMS AND FLIGHT DYNAMICS OF DIPTERAN INSECTS

By ALEXANDRA M. YARGER

Submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy

Department of Biology

CASE WESTERN RESERVE UNIVERSITY

May, 2020

CASE WESTERN RESERVE UNIVERSITY

SCHOOL OF GRADUATE STUDIES

We hereby approve the thesis/dissertation of

Alexandra M. Yarger

candidate for the degree of Doctor of Philosophy*.

Committee Chair

Dr. Jessica Fox

Committee Member

Dr. Roy Ritzmann

Committee Member

Dr. Hillel Chiel

Committee Member

Dr. Michael Lewicki

Date of Defense

March 3, 2020

*We also certify that written approval has been obtained

for any proprietary material contained therein.

Dedication

To my father, Jonathon Yarger who told me to do what I want and not what’s reasonable.

Table of Contents

TITLE PAGE...... 1

COMMITTEE APPROVAL SHEET ...... 2

DEDICATION ...... 3

LIST OF FIGURES ...... 6

ACKNOWLEDGEMENTS ...... 7

ABSTRACT ...... 8

CHAPTER 1 ...... 10

Summary ...... 11

Introduction ...... 11 What are halteres? ...... 13 What are the dynamics of the halteres as flies perform natural behaviors? ...... 16 What are the forces acting on the haltere? ...... 17 What are the sensory structures on the halteres that transduce these forces, and how do they function? . 19 How do haltere afferent neurons process force information? ...... 22 Where does haltere information go? ...... 23 How is haltere information integrated with vision to guide wing-steering behavior? ...... 25 How is haltere information integrated with vision for gaze control? ...... 28 Are halteres useful for behaviors other than flight? ...... 29 How do other insects sense forces for flight control? ...... 30 Open questions on haltere function for fly behavior ...... 32

CHAPTER 2 ...... 36

Summary ...... 37

Introduction ...... 37

Methods...... 42 Recording ethological haltere oscillations ...... 42 Preparation for nerve recordings ...... 42 Electrophysiology...... 43 Haltere stimulation ...... 43

Results ...... 44

Spontaneous haltere oscillations are similar in freely-behaving and tethered flies ...... 44 Cells activated by self-generated and motor-driven oscillations phase-lock with similar precision .... 46 Sensilla respond to low amplitudes and frequencies during motor-driven oscillations, but not during self-generated haltere oscillations ...... 49 Eliminating neural feedback does not change the activation threshold...... 50 Amplitude and frequency thresholds change with lateral displacement ...... 50 Sensory neurons can encode haltere lateral displacements using phase information ...... 52

Discussion...... 55

Supplementary materials ...... 59

CHAPTER 3 ...... 61

Summary ...... 62

Methods...... 63 Recording ethological data ...... 63 Tethered oscillations ...... 63 Takeoffs ...... 63 Statistics ...... 64

Results & Discussion ...... 64 Calyptratae flies oscillate their halteres asynchronously during pre-flight behaviors ...... 64 Spontaneous takeoffs in Calyptratae flies are shorter than in non-Calyptratae flies ...... 66 Halteres influence takeoff speed, pitch stability, and leg extension velocity in Calyptratae ...... 67 Flies with fast leg extensions and stout bodies have shorter takeoffs ...... 70 Haltere use and morphology in closely related clades ...... 72

CHAPTER 4 ...... 75

Conclusions ...... 76 Single sensory neurons ...... 76 Behavioral context ...... 78

Future Directions ...... 80 Population code and downstream effects ...... 80 Population recordings of the haltere nerve ...... 81 Contralateral haltere interneurons ...... 82 Neck motor neurons ...... 83 Further flight dynamics and behavior ...... 84 Modeling effects of morphology on takeoff speed and stability ...... 84 Takeoff strategies and force production...... 85 Haltere use in walking behaviors ...... 86 Height and orientation ...... 87 Expanding our catalog of behavioral and phylogenetic diversity ...... 88

REFERENCES ...... 90

List of Figures Chapter 1

Figure 1.1 Photographs of halteres ...... 12 Figure 1.2 Haltere campaniform sensilla...... 15 Figure 1.3 Diagram showing the haltere campaniform sensilla fields ...... 21

Chapter 2

Figure 2.1 Experimental setup ...... 40 Figure 2.2 Self-generated haltere oscillations in tethered flies are similar to those in freely- walking flies ...... 45 Figure 2.3 Amplitude and frequency threshold is lower for motor-driven oscillations ...... 48 Figure 2.4 Amplitude and frequency thresholds depend on the lateral displacement of the haltere ...... 52 Figure 2.5 Lateral displacement results in a phase-shift for some cells, but not others ...... 54 Figure 2.6 A schematic of predicted afferent and downstream activity during straight and rotating flight ...... 58

Chapter 3

Figure 3.1 Calyptratae have distinct behaviors ...... 65 Figure 3.2 Calyptratae flies without halteres lose pitch stability, but not takeoff speed ...... 68 Figure 3.3 Larger sagittal aspect ratios allow flies to take off faster ...... 71 Figure 3.4 Drosophila escapes are fast, but are not affected by haltere removal ...... 73

Chapter 4

Figure 4.1 Haltere sweep angle influences primary afferent firing phase ...... 77 Figure 4.2 Example haltere campaniform sensilla ...... 81 Figure 4.3 Takeoff orientation in two non-Calyptratae flies ...... 88

Acknowledgements

I would like to thank my mentor Jessica Fox who not only provided me with all the tools and support I needed to become an effective researcher, but also showed me that the most important qualities a scientist should have are kindness, enthusiasm, and grit.

I would also like to express my sincere gratitude to my committee, Roy Ritzmann,

Mark Willis, Hillel Chiel, and Michael Lewicki for their thoughtful suggestions and experimental guidance.

I thank Al Pollack, Tom Daniel, Nicholas Kathman, Michael Rauscher, David

Bertsch, Tanvi Deora, and Samuel Fabian for helpful discussions. I thank Sweta Agrawal for creating Figure 1.3 and for haltere SEMs, Rebekka Bamert for flesh fly illustrations,

Michael Tu for custom-written MATLAB software (Centipede), Alexa Smith, Katherine

Jordan, Tasha Johnson, Noah DeFino, and Jesse Fritz for data collection and analysis.

And I thank my family, Jonathon, Elizabeth, Andrew, Benjamin, Megan, Kitri, and Rico for their love and encouragement.

Inertial Encoding Mechanisms and Flight Dynamics of Dipteran Insects

By ALEXANDRA M. YARGER

Abstract

In dipteran insects (true flies), modified hindwings known as halteres detect forces produced by body rotations, and are essential for flight. Halteres are mechanosensory organs with several arrays of sensory cells at their base, and they are one of the characteristic features of flies. Mechanosensory information from the halteres is sent to wing-steering and head movement motor neurons, allowing direct control of body position and gaze. Analyses of the structure and dynamics of halteres indicate that they experience very small aerodynamic forces but significant inertial forces, including

Coriolis forces associated with body rotations. The sensory cells at the base of the haltere detect these forces and allow the fly to correct for perturbations during flight.

The mechanisms by which haltere neurons transform forces resulting from three- dimensional body rotations into patterns of neural spikes are unknown, however. We use intracellular electrodes to record from haltere primary afferent neurons during a range of haltere motions. We find that spike timing activity of individual neurons changes with displacement, and propose a mechanism by which single neurons can encode three- dimensional haltere movements during flight.

However, halteres are not just used for flight. The most recently diverged monophyletic subsection within the Dipteran order, called Calyptratae also use their halters during walking behavior (Hall et al., 2015). We examined the biomechanics of a representative Calyptratae fly and compared it with known wing-haltere mechanics in a non-Calyptratae fly (Deora et al., 2015) that does not use its halters when walking. We also compared the transition behavior (takeoff) that occurs between walking and flying in a variety of Calyptratae and non-Calyptratae fly families. We find that body morphology and haltere use contribute to takeoff speed and stability, but only in the Calyptratae clade.

Chapter 1

Dipteran halteres: perspectives on function and integration for a unique sensory organ

This material was previously published in the journal Integrative and Comparative

Biology: Yarger, A. Y. and Fox, J. L. (2016)

Summary

The halteres of dipteran insects (true flies) are essential mechanosensory organs for flight. These are modified hindwings with several arrays of sensory cells at their base, and they are one of the characteristic features of flies. Mechanosensory information from the halteres is sent with low latency to wing-steering and head movement motoneurons, allowing direct control of body position and gaze. Analyses of the structure and dynamics of halteres indicate that they experience very small aerodynamic forces but significant inertial forces, including Coriolis forces associated with body rotations. The sensory cells at the base of the haltere detect these forces and allow the fly to correct for perturbations during flight, but new evidence suggests that this may not be their only role. This review will examine our current understanding of how these organs move, encode forces, and transmit information about these forces to the nervous system to guide behavior.

Introduction

The four largest orders of holometabolous insects are each defined and named by characteristics of their wings: the Coleoptera (beetles, “sheath wing”), Lepidoptera

(butterflies and moths, “scale wing”), Hymenoptera (ants, wasps, and bees, “membranous wing”), and Diptera (flies, “two wings”; Borror et al., 1992). The last of these, the flies, are thus defined not by what they possess, but by what they lack: a second pair of lift- generating wings. In place of these wings is a pair of small dumbbell-shaped appendages known as halteres (from the Greek word for “dumbbell”; Figure 1.1). The shape of these organs is such that they do not generate significant lift (unlike the hindwings of other insect orders), but rather experience large inertial forces as they beat with high amplitude

11 during flight (Nalbach, 1993). Centuries ago, curious scientists found that if the halteres are removed, flies are unable to fly (Derham, 1714). Derham wrote:

For the keeping the Body steady and upright in Flight, it generally holds true, (if I mistake not,) that all bipennated Insects have Poisses joyn’d to the Body, under the hinder Part of their Wings; but such as have four Wings, or Wings with Elytra, none. If one of the Poisses, or one of the lesser auxiliary Wings be cut off, the Insect will fly as if one Side overbalanced the other, until it falleth on the Ground; so if both be cut off, they will fly aukwardly, and unsteadily, manifesting the Defect of some necessary part. These

Poisses, or Pointells are, for the most Part little Balls, set at the Top of a slender stalk, which they can move every way at pleasure. In some they stand alone, in others, (as in the whole Flesh-Fly Tribe,) they have little Covers or Shields, under which they lie and move. The Use, no doubt of these Poisses and secondary lesser Wings is to poise the

Body, and to obviate all the Vacillations thereof in Flight; serving to the Insect as the long Pole, laden at the Ends with Lead, doth the Ropedancer.

Figure 1.1. A) Photograph of the crane fly Tipula hespera. Black pointers indicate the end-knobs of the halteres. Photo by Armin Hinterwirth. B) Scanning electron micrograph (SEM) of the haltere of a

12 calyptrate fly. Black pointer indicates haltere end-knob, white pointer indicates calypter. Photo by CSIRO

(Science Image 3237).

Early ideas about the function of the halteres included this concept of the halteres as “balancers” and a second concept of halteres as “stimulation organs,” in which haltere oscillations would drive the flight muscles (v. Buddenbrock, 1919), but later data demonstrated that the halteres are organs of control rather than activation (Fraenkel,

1939; Pringle, 1948). Although the complex high-speed flight behaviors of flies have long fascinated scientists, there are still important unanswered questions on how the fly’s exquisite sensory systems operate to guide their aerial acrobatics. This review will focus on the role of haltere sensory input in fly behavior and examine how this essential organ is used in concert with other senses to guide the fly through its environment.

What are halteres?

The anatomy and development of halteres support the hypothesis that these organs evolved from the hindwings of ancestral insects. They are located on the metathorax and have similar hinges to the forewings, and the muscles of the metathorax that move the halteres are serial homologues of the mesothoracic muscles that move the wings (Chan et al., 1998). These muscles are reduced in both size and number from their mesothoracic counterparts, but have the same functions of oscillating the halteres and providing fine steering motions.

Arrays of the mechanosensory structures known as campaniform sensilla are found on both halteres and wings. On the wings of moths, sensilla are found along the wing veins as well as clustered into small fields at the base (Dombrowski UJ, 1991;

Dickerson et al., 2014). On the halteres, these sensilla are found exclusively at the base, and their arrangement has been described in two highly-derived fly species: Drosophila

(Cole and Palka, 1982) and Calliphora (Gnatzy et al., 1987). In these species, the campaniform sensilla are arranged in straight rows and grouped into uniform fields. On both wings and halteres, groups of sensilla at the base are positioned at regions where maximal strain would occur during wing or haltere oscillation and body rotations. Recent data show that the forewings of moths can act as sensors of body rotation (Dickerson et al., 2014), suggesting that the campaniform sensilla on wings and halteres serve similar functions.

Figure 1.2. A) Diagram showing the location at the base of the haltere SEMs were taken B) SEM of halteres of Tipula showing fields of campaniform sensilla (Scale bar: 10 μm). White box indicates location of the following panel. C) SEM of single campaniform field (dBP; scale bar: 10 μm). White box indicates location of the following panel. D) SEM showing individual campaniform sensilla (Scale bar: 1 μm). E)

Diagram showing structure of individual campaniform sensillum(Keil, 1997). Dome-shaped region is composed of flexible cuticle. Downward displacement of this dome is detected by mechanosensory dendritic processes, which relay that information to the haltere nerve. The electrical activity of the campaniform sensilla can be recorded using extracellular electrodes placed in the axon. Adapted from Fox and Daniel, 2008 and Keil, 1997.

In segmented organisms, homeobox (Hox) genes direct the development of serially homologous segments. The Hox gene Ultrabithorax (Ubx) specifies the identity of the third thoracic segment of the insect body (Hersh et al., 2007). Proper hind wing development requires Ubx in a number of taxa, including butterflies, beetles, and flies

(Weatherbee et al., 1998, 1999; Tomoyasu et al., 2005). In Drosophila, Ubx controls the formation of the halteres during metamorphosis: if Ubx is deactivated, the haltere develops into a fully-formed wing (Hersh et al., 2007). This single homeotic gene change results in a radical phenotype change, because Ubx activates a cascade of other target genes (Weatherbee et al., 1999; Galant et al., 2002; Hersh and Carroll, 2005). Genes expressed in wings and repressed in halteres have also been identified, but whether or not they act as direct targets of Ubx regulation is still unknown (Weatherbee et al., 1998;

Crickmore and Mann, 2006; Mohit et al., 2006). Although much remains to be explored beyond the level of Ubx activation, this locus provides clues on how the ancestors of flies' hind wings may have originally evolved into halteres.

What are the dynamics of the halteres as flies perform natural behaviors?

The halteres oscillate when flies are flying, typically at the same frequency and at opposite phase with the wings’ oscillations (Deora et al. 2015; Hall et al. 2015). The tight synchrony of wings and halteres was recently demonstrated to be mechanical, not neural, in origin. When forceps are used to manually move a freshly-killed soldier fly’s (Diptera:

Stratiomyidae) wing up and down, not only does the opposite wing move in synchrony, the halteres also beat in antiphase with both wings. This experiment indicates that the coupling of the wings and halteres is driven by mechanical coupling and not neural input

(Deora et al., 2015). The source of this coupling between the wings and halteres was

16 found to be two small ridges of cuticle known as the subepimeral ridges (Deora et al.,

2015). This mechanical connection ensures that the halteres remain synchronized with the movements of the wings during flight, such that the wings and halteres beat at the same frequency and with a specific phase that does not change between wing strokes. The haltere-wing phase relationship can differ substantially between species, however.

Brachyceran (short-antennaed) flies, like soldier flies, oscillate their halteres almost exactly opposite their wings (180 degrees out of phase). However, more ancient nematoceran (long-antennaed) flies, which include crane flies, midges, and mosquitoes, exhibit wing-haltere phases that remain consistent between wingbeats, but vary substantially between species. These differences in wing-haltere coordination suggest that the activity of the haltere and wing sensors could vary according to a group’s evolutionary relationships (Hall et al., 2015). This in turn affects the mechanisms used to produce adaptive motor responses, which determine the flight behavior of the fly.

In addition to moving the halteres in flight, some calyptrate flies also oscillate their halteres while walking, without oscillating their wings (Fraenkel, 1939; Miller,

1977; Sandeman and Markl, 1980). They are able to oscillate them at their typical wingbeat frequency, but without engaging the large indirect flight muscles (Hall et al.,

2015), suggesting that the small muscles intrinsic to the haltere (Chan et al., 1998) are responsible for the oscillation. The left and right halteres also show variable phase relationships while walking, in stark contrast to their strict coupling during flight, demonstrating that the thoracic mechanisms that maintain haltere synchrony in flight are not activated while walking (Hall et al., 2015).

What are the forces acting on the haltere?

As the haltere oscillates during walking or flight, several forces act upon it. These forces are largely inertial, with significantly smaller aerodynamic forces (Nalbach, 1993).

Body rotations in space produce forces acting primarily on the end-knobs of the halteres, as a result of their large and fast-moving mass (Nalbach, 1993). These end-knob forces cause torsion or bending of the stalk, which activates fields of mechanosensory structures located at the base of the haltere (Pringle, 1948; Eberle et al., 2015). Analysis of these forces demonstrated that halteres are not wind sensors, nor do they detect air currents resulting from the beating of the wings. Halteres are positioned close to the body, in the cleft between the abdomen and thorax, where air currents caused by body rotations have a negligible effect on haltere movements (Nalbach, 1993). Furthermore, using high-speed video analysis of tethered flying blowflies, Nalbach (1993) calculated the relative contributions of each force during a body rotation and found that the angular velocity dependent force (Coriolis force) had the largest contribution for all three directions of rotation (yaw, pitch, and roll). The halteres only experience Coriolis force while rotating, but always experience a large inertial force when flapping.

Thompson et al. (2009) predicted the end-knob trajectory of the haltere during body rotations of various speeds. When a fly is flying in a straight path, each haltere vibrates in a plane. However, when the fly is rotating, the Coriolis force deflects the halteres laterally in addition to their up- and-down movement, resulting in a figure-of- eight trajectory for the end-knob (Thompson et al., 2009). At different velocities, the phases of the maximum out-of-plane displacement with respect to the haltere oscillation are different. An array of sensilla with precise spike timing and high sensitivity for particular phases of the haltere’s oscillatory movement would be able to detect this

18 change in maximum displacement phase using a population code, and thus encode the fly’s rotational velocity (Fox et al., 2010). Directional selectivity of the campaniform sensilla, as implied by the ovoid shape of some of the sensillar caps (Zill et al., 1981;

Thurm, 1984; Gnatzy et al., 1987), could further refine the encoding by increasing sensitivity to forces occurring in specific directions.

Differences in haltere morphology can influence the forces acting on the haltere.

More ancient families, such as Tipulidae (crane flies), possess halteres with long stalks

(Figure 1.1A). The halteres of more derived flies are generally shorter and in some families, such as Calliphoridae (blow flies), specialized flaps of wing tissue called calypters cover the haltere (Figure 1.1B). Pringle (1948) hypothesized that the function of calypters was to prevent wind turbulence from affecting haltere movements, allowing more precise detection of body position, but this idea has not been formally tested. The stalk of the haltere is also not always straight, and in more derived families tends to curve to reflect the shape of the body. This minimizes the amount of air space between the end- knob and the sides of the abdomen and thorax, which, like the calypters, could decrease wind resistance. In all halteres, the bulb is asymmetric, with the larger portion of the bulb posterior. The degree of this asymmetry can vary across species but generally increases with evolutionary derivation, from a low asymmetry in crane flies (Fox and Daniel, 2008) to extreme asymmetry in hoverflies (Ball and Morris, 2015). This asymmetry may lead to increased torsion on the haltere during body rotations, providing a potentially informative force to the campaniform sensilla (Eberle et al., 2015).

What are the sensory structures on the halteres that transduce these forces, and how do they function?

The halteres possess several distinct fields of sensory cells at their base (Figure

1.2). These fields, which vary in number and shape depending on species (Gnatzy et al.,

1987; Fox and Daniel, 2008), are composed of campaniform sensilla. These are dome- shaped protrusions located on the surface of the exoskeleton (cuticle), and are found on all insects in various locations. Attached inside the dome of each campaniform sensillum is a modified cilium extending from the sensory process of a neuron (Figure 1.2E). When the cuticle bends and the surface of the dome is deflected, the neurons generate graded receptor potentials mediated by mechanically-activated ionotropic TRPN channels. Upon reaching threshold, these receptor potentials are then encoded as action potentials

(Walker et al., 2000; Kernan, 2007). These action potentials are sent to the central nervous system to be interpreted (Smith, 1969; Thurm, 1984; Keil, 1997; Chapman et al.,

2013). In blowflies (Calliphora), there are more than 400 campaniform sensilla per haltere, which make up the majority of the campaniform sensilla found on the fly’s body

(Gnatzy et al., 1987). The sheer number of campaniform sensilla on the haltere suggests that flies may detect forces with high resolution, but the upstream effects of this fine- scale sensitivity are not currently known.

The haltere campaniform sensilla are clustered into groups, the number of which can vary by species. Almost all flies have all the same groups of fields (Figure 1.3), though there are some families that do differ. These groups generally comprise cells of a similar orientation, suggesting directional selectivity for particular forces. Anatomical studies show that different fields project their afferents to different specific locations in the thoracic ganglion. In particular, afferents providing input to the first basalar muscle of the wing-steering complex originate from a single field of campaniform sensilla (Chan

20 and Dickinson, 1996). The position-specific projections from each field suggest that as different forces impact different sets of campaniform sensilla, the information will be processed in a spatiotopic way by the central nervous system (Barth et al., 1984; Thurm,

1984; Dickinson, 1990; Chan and Dickinson, 1996). A spatiotopic arrangement of force sensitivity could allow higher neural centers to efficiently interpret force information from the halteres, but this has not been demonstrated as a mechanism of haltere information processing.

Figure 1.3. Diagram showing the haltere campaniform sensilla fields on both the dorsal and ventral sides of the haltere: dorsal Hick's papillae (dF1), basal plate (dF2), dorsal scapal plate (dF3), unbestimmte Papille

(dSi), ventral Hick's papillae (vF1), and ventral scapal plate (dF2)(Gnatzy et al., 1987; Chan and Dickinson,

1996).

The other major mechanosensory structures involved in sensing haltere movements are chordotonal organs (Pflugstaedt, 1912). Chordotonal organs, unlike campaniform sensilla, exist beneath the cuticle and typically respond to stretch or vibration as opposed to distortion or bending (Chapman et al., 2013). Their sensory nerve endings attach between two internal points and when those points move apart, stretching

21 the nerve cell, the difference in length is detected and transformed into action potentials.

There are far fewer chordotonal organs at the base of the haltere than campaniform sensilla (two, versus 400 campaniform sensilla), so it is predicted that they are less important for detecting and transmitting rotational information from haltere movements

(Pringle, 1948). However, their function remains elusive as their axonal projections are localized within the relatively massive bundle of axons from the campaniform sensilla, making them difficult to visualize or access electrophysiologically (Chan and Dickinson,

1996).

How do haltere afferent neurons process force information?

The axons of the campaniform sensilla neurons project out of the base of the haltere as a nerve bundle and enter the thoracic ganglion. Extracellular recordings from the haltere nerve showed that when halteres are oscillated, bursts of action potentials occur in synchrony with the oscillation frequency (Pringle, 1948). When flies are then rotated, these bursts break down into asynchronous spikes. Instead of bursts occurring at particular phases of haltere oscillations, spike times are spread throughout the duration of the haltere movement. Pringle suggested that this reflects recruitment of different groups of sensilla as the fly rotates.

The oscillation-synchronous action potentials in the haltere nerve can be fired at frequencies well beyond the fly’s natural wingbeat (and, by extension, haltere beat) frequency: crane flies with a natural wingbeat frequency of 40 Hz show haltere- synchronous firing in the primary afferent neurons up to 150 Hz when the haltere is mechanically oscillated (Fox and Daniel, 2008). These spikes are highly phase-locked to

22 the haltere’s movement, showing some of the most precise spike-timing measured in any neuron (Fox and Daniel, 2008).

This phase-locking activity suggests that small changes in the haltere’s movement might result in large changes in the spike timing activity of the population of haltere afferent neurons. Analysis of the encoding properties of haltere afferent neurons suggests that a population code, in which each haltere afferent fires a spike at a particular phase of movement, may be able to rapidly and accurately encode the out-of-plane haltere movements associated with body rotations (Thompson et al., 2009; Fox et al., 2010).

Where does haltere information go?

Axons within the haltere nerve project to the neuropil in a region-specific manner, in which dye fills of the neurons in specific fields of campaniform sensilla reveal projections to different locations in the central nervous system. However, these projections are numerous and somewhat diffuse for each field, suggesting that haltere information is sent to many post-synaptic targets within the central nervous system.

Projections to the metathorax, posterior and mid-mesothorax, and prothorax are most prominent, and the anterior-most termination is located in the posterior subesophageal ganglion (Chan and Dickinson, 1996). The majority of specific synaptic connections between haltere afferents and their targets remain unknown, but one particular synapse between the haltere nerve and a wing steering motor neuron (mnb1, the neuron of the first basalar muscle), has been described (Fayyazuddin and Dickinson, 1996). The haltere afferents activating the synapse at mnb1 originate from the campaniform field dF2

(Fayyazuddin and Dickinson, 1996). This single known connection suggests that other

23 campaniform fields may also project to wing-steering muscles, but this is certainly not the only possibility. Each sensilla field projects to particular regions within the thoracic ganglion along an ipsilateral tract within the dorsal tectulum, but their post-synaptic targets are currently unknown (Chan and Dickinson, 1996).

The synapse between the haltere afferents and mnb1 consists of two components: one fast and electrical, and the other slow and chemical. The electrical synapse is the source of the phase-locked firing pattern of mnb1 with the halteres during flight. With each haltere beat, information from haltere campaniform sensilla is sent almost instantaneously to the wing steering muscles (Fayyazuddin and Dickinson, 1996). This fast electrical synapse allows changes in body position detected by the halteres to be sent directly to the wing steering muscles with a very short delay (<5 ms).

In addition to the electrotonic input from the haltere, mnb1 receives mechanosensory input from the wings. To determine how haltere and wing afferents influence the firing of mnb1, Fayyazuddin and Dickinson electrically stimulated the wing and the haltere nerves, varying the phase of stimulation. They found that mnb1 was entrained by the wing afferent stimulation over a wider range of phase differences than by haltere afferent stimulation. The activity at the haltere afferent-mnb1 synapse indicates that the phase of firing in mnb1 is constrained by the wing afferents during straight flight, but that phase-shifted spikes from the haltere can change mnb1’s spike timing and thus, adjust the tension on the wing-steering basalar muscle to turn the fly (Fayyazuddin and

Dickinson, 1999). Rotational perturbations cause the haltere to be laterally deflected

(Thompson et al., 2009) and the dF2 campaniform field is oriented such that it could detect this deflection for certain types of rotation (Pringle, 1948; Gnatzy et al., 1987;

Chan and Dickinson, 1996). The haltere afferents activating the synapse at mnb1 originate from dF2 (Fayyazuddin and Dickinson, 1996), and thus dF2 may be essential for detecting Coriolis forces caused by rotations in a particular direction, while other directions of rotation may activate other campaniform fields. These careful electrophysiological experiments thus provided a mechanism by which flies can use haltere sensory information to sense rotations and rapidly adjust their motor output.

How is haltere information integrated with vision to guide wing-steering behavior?

Because flies lacking halteres are unable to maintain stable free flight (Derham,

1714), finer analyses of the behavioral effects of haltere input have been limited. By tethering flies and observing their attempted body rotations through torque sensors or wingbeat amplitude measurements, however, we can determine how the loss of haltere input influences behavior and draw inferences about how the fly’s brain uses haltere information. Flies adjust the tension in direct wing-steering muscles to control the amplitude and attack angles of their wings (Heide, 1983; Dickinson and Tu, 1997), and thus haltere inputs to motoneurons controlling these muscles have been of particular interest.

An experiment in which flies were rotated in various directions while fixating a closed-loop visual target provided evidence that halteres detect body rotations in each of the major rotational axes. Intact flies were able to adjust their wingbeat amplitudes to maintain a stable flight path (correcting for imposed body rotations) when rotated in any direction, but when one haltere was ablated, the flies’ ability to correct their flight path was unilaterally diminished for motions toward the ablated side (Dickinson, 1999). In

25 flies with both halteres removed, the flight path correction was nearly eliminated. The fly’s wing-steering adjustments were correlated with the body’s rotational velocity, and not its position or acceleration, indicating that halteres detect the velocity of the fly’s body rotations and act as gyroscopes to stabilize the fly against imposed rotations.

How do the halteres contribute to behavior when the flies are allowed to turn themselves, rather than experiencing imposed turns? During free flight, flies exhibit periods of smooth linear motion interspersed with rapid turns called saccades. Bender and

Dickinson (2006) found that when flies were fixed to a magnetic tether, such that they were able to perform free turns in the yaw direction, the amplitudes of their body saccades (rapid turns) did not depend on visual feedback. Instead, saccade amplitude was only affected when the halteres were weighted or ablated, increasing or decreasing the

Coriolis forces acting on the haltere, respectively. When the mass was increased, saccade amplitude decreased, and when mass was decreased, saccade amplitude increased

(Bender and Dickinson, 2006). These results indicate that when flies perform voluntary, high-speed rotations, feedback from the halteres is essential to properly execute the behavior. Furthermore, body saccades are under sensory feedback control, and are not ballistic maneuvers: despite their rapidity, they can be modulated throughout their execution.

If haltere feedback controls the amplitude and duration of saccades, which are often stimulated by visual input, how might visual information integrate with haltere information to control flight behavior? It is clear that flies employ both vision and halteres to maintain stable flight. The halteres and eyes are tuned to complementary speeds of rotations for correcting the body position of the fly. When flies steer their

26 wings in response to a moving visual stimulus, peak responses occur at lower speeds, whereas peak responses to physical rotations of the fly’s body occur at higher speeds

(Sherman and Dickinson, 2003). The compensatory head roll response shows a similar pattern of range fractionation (Hengstenberg, 1988; Schwyn et al., 2011). The faster haltere system provides rapid feedforward information, whereas the slower visual system permits more accurate feedback (Sherman and Dickinson, 2003; Schwyn et al., 2011).

Because the halteres and eyes detect different ranges of angular velocities, the integration of these two separately tuned senses increases the fly’s ability to correct its body position over a wide range of angular velocities.

Can halteres influence flight behavior when the fly’s body is not rotating? In quiescent preparations of Calliphora, the intrinsic muscles of the halteres, but not the wing-steering muscles, are stimulated by moving wide-field visual stimuli (Chan et al.,

1998). This curious result suggests that visually guided, self-generated turns could be driven by an active movement of the halteres, triggering a corrective reflex to steer the fly’s wings. This manipulation of the fly’s control loop would take advantage of the fly’s haltere-driven reflexes to escape from a reflex loop that would otherwise trap them in straight flight (Chan et al., 1998).

This interesting hypothesis would imply that flies might rely on their halteres to perform visually-stimulated turns. When halteres are removed from tethered flying flies, however, they are still able to steer their wings toward moving figures, but they struggle to stabilize moving wide-field visual stimuli (Mureli and Fox, 2015). Halteres are thus not necessary for visually guided wing-steering movements, and visual information can reach wing-steering muscles through some haltere-independent connection. However,

27 halteres do contribute to wing-steering maneuvers in a context-dependent manner, even when the behavior is separated from body rotations.

How is haltere information integrated with vision for gaze control?

Primary afferent neurons from the haltere synapse on neck motoneurons (NMNs) that control the position of the fly’s head (Strausfeld and Seyan, 1985; Milde et al., 1987;

Strausfeld et al., 1987). This input allows the head to counter-rotate against the body’s rotation, keeping the head stable even as the body rotates behind it. Because the fly’s eyes cannot move independently of its head, stabilization of the head via visual and haltere input keeps the visual scene fixed on the retina, and prevents self-generated optic flow from entering the visual information stream (Hengstenberg, 1991; Lappe et al.,

1999).

Hengstenberg tested flies’ ability to stabilize their heads when rotated over a large range of angular velocities and found that in the roll direction of rotation, flies were able to maintain their head position in response to angular motion only at speeds above 50°/s.

He found that flies were most adept at detecting angular rotations at ~1500°/s

(Hengstenberg, 1988). Nalbach & Hengstenberg (1986) showed velocity dependencies for yaw and pitch rotations as well, but specific ranges similar to those published for roll were not defined. When haltere inputs were removed entirely, the fly's ability to perceive roll movements at high angular velocities (above 1000 degrees per second) disappeared.

These results indicate that the fastest stabilization responses are supported entirely by the relatively faster mechanosensory pathway from the halteres.

Flies are still able to perform compensatory head movements without the use of their halteres at slower angular velocities (Hengstenberg, 1991; Schwyn et al., 2011).

However, differences between this visually-guided response and that of a purely mechanosensory compensation do exist. Head stabilization outputs due to visual inputs alone are slower and longer-lasting than haltere-driven reflexes (Hengstenberg, 1991).

Thus, the visual and mechanosensory systems work together to stabilize the visual field of the animal: first, by quickly responding to fast changes, and second, by maintaining that response until it is fully corrected (Hengstenberg, 1991).

A possible mechanism for the integration of visual and haltere inputs to head steering has recently been described (Huston and Krapp, 2009). A sub-population of

NMNs exhibit what are described as “gating-like” properties of multi-sensory integration.

When these NMNs were recorded extracellularly and the fly was given visual stimuli alone, these neurons did not fire action potentials. However, when the halteres were concurrently activated by oscillating them with a small motor while the visual stimulus was presented, the NMNs fired action potentials. Intracellular recordings showed that

NMNs receive subthreshold inputs from both visual motion and haltere movements. The subthreshold inputs from the halteres were larger than the visual motion inputs, and were also phase locked with the haltere oscillations. These results indicate that the “gating- like” behavior of the NMNs is a result of the convergence of these two subthreshold inputs.

Are halteres useful for behaviors other than flight?

The necessity of halteres in flight has been well documented, but little is known about their use in other behaviors. Flies in the clade Calyptratae (a large group including some of the most speciose families like Calliphoridae and Muscidae) have been documented to oscillate their wings while walking in addition to during flight (Fraenkel,

1939; Miller, 1977; Sandeman and Markl, 1980; Hall et al., 2015). The oscillation of the haltere is similar in amplitude and frequency during walking and flight for these flies, and the halteres always oscillate when walking or when flying. Similar haltere oscillations have been observed during walking in Micropezidae (Hall et al., 2015), which are outside the Calyptratae but which use walking as a primary means of locomotion. Indeed, some species of Micropezidae are wingless (McAlpine, 1990), indicating that their haltere movements are used exclusively for walking. All other families of Diptera do not generally oscillate their halteres while walking, but always do so while flying.

Are these movements important to the fly’s behavior, or are they simply an epiphenomenon of walking? When a species of fly that oscillates its halteres during walking (Sarcophaga bullata) is challenged to walk on a vertical surface, removal of the halteres impairs its ability to climb, as well as its ability to resist falling when perturbed.

Haltere removal had no effect on the climbing or falling behavior of Drosophila, which does not move its halteres during walking. This indicates that haltere inputs are behaviorally relevant to those species which oscillate them while walking, and that they aid in walking behavior when the task is challenging (Hall et al., 2015).

How do other insects sense forces for flight control?

Our current understanding of haltere function suggests that any oscillating mass may act as an inertial sensor during flight, opening the possibility that non-dipteran insects may detect body rotations using other body parts. A haltere-like function for an oscillating appendage has been demonstrated in two other orders, the Strepsiptera and the

Lepidoptera. In the Strepsiptera, which are more closely related to beetles than to flies, the front pair of wings is modified into a haltere-like organ (Wigglesworth, 1946;

McKenna and Farrell, 2010; Wiegmann et al., 2011a). Though Strepsipterans are not only rare, but tiny, difficult to collect, and short-lived, Pix et al. (1993) were able to demonstrate through behavior and morphology that the specialized forewings of male strepsipterans perform the same function as the dipteran halteres. Rotational movements of the body combined with the oscillating forewings produce forces that can be detected by fields of campaniform sensilla located at the base of the forewings. Pix et al. showed that the forewing sensors transmit body rotation information to the head and abdomen to produce compensatory movements (Pix et al., 1993).

In the Lepidoptera, the antennae have been demonstrated to play a haltere-like role in flight control. Antennae of the hawkmoth Manduca sexta exhibit small amplitude oscillations of their antenna at constant angles during flight (Sane et al., 2007). By separating the long part of the antenna (the flagella) from the mechanosensory cells at the antennal base (Böhm's bristles and the Johnston’s organ; Johnston, 1855; Böhm, 1911), then reattaching them and assessing flight performance, Sane et al. (2007) determined that the mechanosensory role of the antenna is essential for flight stabilization: when the flagella are removed, the moths were no longer able to maintain stable flight. After reattachment of the flagella, flight performance was restored. The passive oscillations of

31 the antennae, combined with their mass, allow them to experience Coriolis forces during body rotations. Thus, the antennae of moths may play a haltere-like role in their flight behavior.

We noted above that the halteres are serial homologs of the wings, which themselves contain fields of campaniform sensilla. These sensilla permit the wings to act as both actuators of movement, and as sensors of that movement (Dickinson, 1990;

Dickerson et al., 2014). Recent work indicates that the forewings of Manduca can act as pitch rotation sensors, driving a corrective abdominal movement when stimulated

(Dickerson et al., 2014). These data suggest that the wings of insects acted as sensors of body rotation before their specialization into halteres. Further questions remain about the trade-offs between a lift-generating role and a sensing role for the halteres: whereas moths perform both sensing and actuation with both pairs of wings, flies have separated the lift and sensing functions such that halteres provide no lift and act as sensors only.

What were the evolutionary forces driving this trade-off? This remains an open question, but one that is applicable to one of the largest and most successful groups of animals on the planet.

Open questions on haltere function for fly behavior

Through careful anatomical, electrophysiological, and behavioral studies, significant progress has been made towards understanding the function of these curious but essential sensory organs. Several questions remain, and current research on these topics will lead us towards a more global understanding of sensory information processing for fly flight.

1) Do the broad encoding properties of haltere primary afferent neurons suggest a broad role in behavior?

The electrophysiological responses of haltere primary afferent neurons to motion stimuli show that they provide rapid, precise input to the central nervous system over a broad range of stimulus frequencies (Pringle, 1948; Fox and Daniel, 2008; Fox et al.,

2010). These results demonstrate that haltere sensory neurons are able to play their canonical role as body rotation sensors, but also that they are able to transmit force information well beyond the narrow range of frequencies in which the Coriolis and oscillation-related inertial forces might occur. Rather, recordings from haltere neurons suggest that any in-plane vibration of the haltere above a certain amplitude threshold should result in spike activity in the primary afferent neurons. There are two important implications of this finding: first, the halteres may play a role in sensing other forces beyond those associated with body rotations, and second, the movements of the haltere during natural behavior are the primary determinant of the neural response, with neural tuning playing only a very minor role in structuring the information sent to the nervous system.

2) How is sensory information from the haltere decoded into behavioral commands?

Because there is little filtering of information at the neural level, spikes resulting from different types of haltere movements will reach the central nervous system. What are the resulting responses in higher-order neurons, and how are these interpreted into behavioral commands at the motor level? Presently, the integration of haltere input has been studied in only two groups of peripheral motoneurons: the wing-steering

33 motoneurons (Fayyazuddin and Dickinson, 1996, 1999) and the head-steering NMNs

(Huston and Krapp, 2009). However, the haltere afferents project to many locations in the thoracic ganglion, as observed by dye-filling anatomical descriptions (Chan and

Dickinson, 1996), and their post-synaptic targets are unknown. Similarly, the responses of these post-synaptic targets to haltere input, as well as their responses to inputs from other sensory systems, motor feedback, or state-dependent modulation, are unknown.

Haltere input has been demonstrated to provide behaviorally relevant information during multiple behaviors (Dickinson, 1999; Hall et al., 2015), but it is not known how the input is decoded during these behaviors.

3) How is haltere information represented in the brain?

Our current understanding of haltere inputs is limited to synapses at the periphery, in the wing and neck motoneurons. How is haltere information represented in neurons in the brain? Some haltere information remains peripheral only, and is used exclusively for the fast reflexes necessary for high-speed flight (Land and Collett, 1974; Wagner, 1986).

However, Kathman and Fox (2019) showed that cells in the central brain also respond to haltere inputs and that the timing and rates of neural spiking and can be modulated by sensory input from the visual system. It is also possible that the haltere information represented in the brain is integrated with other sensory information, to modulate slower or more complex responses to stimuli or experiences (Kathman and Fox, 2019). Current research suggests that in the central complex, a brain region known to integrate feedback from multiple sensory systems as well as motor information, the firing of specific cells indicates the fly’s position and orientation in space (Seelig and Jayaraman, 2015). Might halteres have a place in this encoding scheme? This question is somewhat difficult to

34 answer in Drosophila, because haltere movement occurs only during flight, confounding haltere input with similar-frequency input from the campaniform sensilla on the flapping wing. However, in those flies showing a diversity of haltere movements and behaviors, haltere input to higher-level centers may provide a means of sensory flexibility and control, allowing the fly to transition from one behavior to another at high speed and precision.

Chapter 2

Single mechanosensory neurons encode lateral displacements using precise spike timing and thresholds

This material was previously published in the journal Proceedings of the Royal Society

B: Yarger, A. Y. and Fox, J. L. (2018)

Summary

During locomotion, animals rely on multiple sensory modalities to maintain stability. External cues may guide behavior, but they must be interpreted in the context of the animal’s own body movements. Mechanosensory cues that can resolve dynamic internal and environmental conditions, like those from vertebrate vestibular systems or other proprioceptors, are essential for guided movement. How do afferent proprioceptor neurons transform movement into a neural code?

In flies, modified hindwings known as halteres detect forces produced by body rotations, and are essential for flight. The mechanisms by which haltere neurons transform forces resulting from three-dimensional body rotations into patterns of neural spikes are unknown, however. We use intracellular electrodes to record from haltere primary afferent neurons during a range of haltere motions. We find that spike timing activity of individual neurons changes with displacement, and propose a mechanism by which single neurons can encode three-dimensional haltere movements during flight.

Introduction

Tapping on the window pane, a fly attempts to cross the invisible glass barrier.

Each time she collides, her body tumbles, but she never falls completely. How are flies so adept at righting themselves when knocked off balance by an unseen obstacle? One adaptation they possess is a set of specialized sensory organs called halteres (Figure

2.1A-C). Flies cannot fly without halteres, as a major function of halteres is to detect body rotations that occur during flight (Derham, 1714; Pringle, 1948; Fayyazuddin and

Dickinson, 1999; Sherman and Dickinson, 2003). The halteres’ primary function is to

37 provide sensory input to wing-steering (Fayyazuddin and Dickinson, 1996, 1999) and neck motoneurons (Strausfeld and Seyan, 1985; Milde et al., 1987; Huston and Krapp,

2009) so that the fly can maintain a stable flight path as well as control its gaze.

During a body rotation, several forces act on the halteres. The two largest forces are the inertial force from wingbeat-synchronous haltere oscillations (Figure 2.1A), and the Coriolis force (Nalbach, 1993). The Coriolis force causes torsion at the haltere base that can alter its trajectory, particularly its lateral displacement (Nalbach, 1993; Pix et al.,

1993; Thompson et al., 2009; Eberle et al., 2015). The forces resulting from changes in the haltere’s position are detected by fields of mechanosensory campaniform sensilla located at the base (Figure 2.1C-E; Smith, 1969; Agrawal et al., 2017). Neurons within the sensilla provide precise, phase-locked responses when the haltere is oscillated

(Pringle, 1948; Fox and Daniel, 2008).

How are haltere deflections encoded by haltere sensory neurons? The spiking activity of primary afferent neurons is the only representation of the haltere’s movement available to the nervous system, and thus, information contained in these spikes is essential for flight. Our current knowledge of information processing in single afferents is limited to fully restrained animals experiencing artificial haltere movements (Pringle,

1948; Fox and Daniel, 2008; Fox et al., 2010). Measuring spiking activity in flying flies is challenging, because the haltere nerve runs directly past two sets of large flight muscles contracting at high frequency. Still, data from tethered experiments led to two non-exclusive hypotheses that may explain mechanisms by which haltere neurons encode changes in haltere stroke plane.

First, a change in stroke plane could recruit specific afferents. This mechanism was first proposed in 1948 by JWS Pringle in his comprehensive study of haltere mechanics and physiology (Pringle, 1948). Under this hypothesis, different types of body rotations would activate different populations of campaniform sensilla. A downstream neuron could distinguish turns of different directions and magnitudes by integrating the activity of sensilla groups. Different neurons could have different thresholds of activation that change with stroke plane. A neuron that is activated when the haltere oscillates above a particular amplitude and frequency in one plane could be activated at a lower or higher amplitude and frequency in another. Thus, the threshold of activation could encode information about the haltere’s lateral displacement.

Figure 2.1. Experimental setup. A) Schematic showing the location of the haltere and the natural oscillation trajectory that occurs during flight and other behaviors. B) Schematic of dissection. Wings and legs were removed before tethering. An iron filing was glued to the tip of the haltere beside the magnet

40 attached to the motor. The dorsal side of the thorax was opened to expose the haltere nerve for intracellular recording. C) SEM of haltere. White square shows location of part D. D) SEM of campaniform sensilla located at the base of the haltere. E: Schematic of campaniform sensilla. Arrows represent direction of deflection to activate sensilla. E) Hypothesis for sensory neuron phase shifts during body rotations (adapted from Pringle, 1948). Top trace (blue): haltere elevation during a single oscillation. Bottom five traces: activity of four different sensilla (represented by different colors) during four different conditions. Top to bottom: large right turn, small right turn, stable flight, small left turn, large left turn. Haltere images on the right (tan) show approximate haltere lateral displacement during each of these five conditions

In the second hypothesis, individual afferent neurons could change the timing of their action potentials as the haltere’s position changed. Afferent neurons are tightly phase-locked to haltere oscillation (Pringle, 1948; Fox and Daniel, 2008; Fox et al.,

2010), and in experiments in which the haltere was oscillated in a single plane, different neurons fire at unique times with respect to the oscillation. In an encoding scheme based on spike timing, information about body rotation could be obtained from the spike time of a single campaniform sensillum. This hypothesis was also introduced in Pringle’s work, in which he postulated that some cells will shift their firing phase corresponding to the magnitude of the turn (Figure 2.1F). At the time, Pringle was unable to record from single sensory cells to confirm this: an attempt to simulate a yaw turn during an extracellular recording resulted in a “confused jumble of spikes” that were difficult to interpret.

Here, we use sharp intracellular electrode recordings to test both of these hypotheses and determine if lateral haltere displacements (like those a fly might experience during a body rotation) result in lower thresholds of activation or in phase shifts in haltere sensory neurons. The spiking activity we recorded provides evidence

41 supporting both hypotheses and provides a potential mechanism for encoding body rotations through haltere lateral displacement.

Methods

Recording ethological haltere oscillations

A high-speed camera captured the halteres of freely-walking flies at 2000 frames per second (Fastec Imaging, San Diego, CA, USA). Wings were removed to prevent flight and to ensure haltere visibility. Positions of the haltere tip and base were digitized using DLTdataviewer (Hedrick, 2008). In some trials, two synchronized cameras were used to observe both halteres simultaneously.

Preparation for nerve recordings

Experiments were performed on adult flesh flies (Sarcophaga bullata; Carolina

Biological, Burlington, NC, USA) taken from our lab colony. Animals were given food and water ad libitum, and kept on a 12h light/dark cycle at 27°C. Flies were anesthetized at 0°C for 2-3 minutes. Wings, legs and calypters were removed before tethering the flies dorsal side up to a wooden tether. The exoskeleton and wing muscles above the haltere nerve were removed to expose the nerve for intracellular recording (Figure 2.1B inset).

Insect saline (in mM: 150 NaCl, 3.1 KCl, 5.4 CaCl2, 1.0 MgCl2, 2.0 NaOH, 5.0 N-tris

[hydroxymethyl] methyl-2-amino ethanesulfonic acid (TES); Blagburn and Sattelle,

1987) was added to the thoracic cavity to cover the neural tissue. Prior to each experiment, we verified that the fly was able to spontaneously oscillate both halteres at natural oscillation frequency (~200Hz).

Electrophysiology

We recorded intracellularly from single axons in the haltere nerve using sharp quartz electrodes (Figure 2.1B inset). Glass capillaries with outer diameter of 1.0mm and inner diameter of 0.70mm (Sutter Instruments, QF100-70-10) were pulled to a resistance of ~15 MΩ using a laser puller (Sutter Instruments P-2000) and filled with 1 M KCl. The indifferent electrode was a hooked silver wire, placed in the hemolymph, that doubled as a support platform for the haltere nerve. The signal was sampled at 16kHz, filtered and amplified through an AM Systems amplifier (Model 1600, Carlsborg, WA, USA). Data were recorded onto a computer using Spike2 (CED, Cambridge, UK) and a Micro 1401

AD board (CED). We used custom-written MATLAB software (The MathWorks, Natick,

MA, USA) for spike phase and frequency analysis. In a minority of cells, there were multiple spikes per oscillation. This only occurred during motor-driven oscillation experiments at low frequencies. In these cases, we used the first spike in the burst to measure the phase of activation.

Haltere stimulation

During intracellular recordings, flies were allowed to oscillate their halteres naturally (self-generated oscillations), and the halteres were also oscillated by a servo motor (motor-driven oscillations). To oscillate the haltere with a servo motor, we attached an iron filing (100-300um diameter) to the haltere bulb with glue (Figure 2.1B;

Loctite Super Glue-Ultragel Control, Düsseldorf, Germany). A servo motor (Aurora

Scientific, Ontario, Canada) with a small rare-earth magnet glued to the lever arm was then used to move the haltere at various amplitudes and frequencies. Motor commands

43 were controlled by custom MATLAB software and sent to the motor (USB-6229 M

Series, National Instruments, Austin, TX, USA). The motor-driven amplitudes and frequencies tested were similar to those observed in self-generated haltere motions.

In our initial experiments, we allowed the haltere to rest at a natural angle relative to the body, and oscillated the haltere dorso-ventrally at this angle (Movie S1: center position). This angle varied between animals. To observe how changing the haltere’s lateral displacement influenced its spiking activity, the motor was placed either anterior or posterior to the natural resting position (center) of the haltere, drawing the haltere anterior or posterior with respect to the body (Movie S1). Haltere movements were captured by high speed video at 2000 fps (Fastec Imaging) and digitized as above. The lateral displacement angle was measured using photographs of the fly’s dorsal aspect taken through a cell phone camera mounted on a dissecting microscope (Mi Platform,

Scientific Device Laboratory, IL, USA). Lateral displacement angles were similar across animals.

Results

Spontaneous haltere oscillations are similar in freely-behaving and tethered flies

Flesh flies, like other members of the Calyptratae, oscillate their halteres during both flight and walking behaviors (Figure 2.1A; Sandeman and Markl, 1980; Hall et al.,

2015). When walking, the oscillations occur in short bursts, similar to those observed in tethered flies (Figure 2.2A). We compared oscillation behavior in freely-walking and animals that were tethered and prepared for electrophysiology. The tethered flies lacked wings, legs, and indirect flight muscles, but were still observed to spontaneously oscillate

44 their halteres (Hall et al., 2015). Wings were removed before filming in both walking and tethered flies, so we observed the wing stumps to ensure that the indirect flight muscles were not being activated in an attempt to fly.

Figure 2.2. Self-generated haltere oscillations in tethered flies are similar to those in freely-walking flies.

A) Top trace: haltere elevation during a self-generated oscillation in a tethered animal. Lower trace: haltere elevation during a self-generated oscillation in a freely-behaving animal. B-C) Each point represents a

45 single animal’s average amplitude (B) or frequency (C) oscillation and the bars represent the median value for all animals (N=7 freely-walking and 5 tethered animals) in each condition. D) Each point represents the vector strength of a single cell during self-generated (gray) and motor-driven (black) oscillations. Solid horizontal bars represent the median vector strength for all cells recorded during either self-generated or motor-driven oscillations. E) Each arrow represents a single cell recorded during self-generated (gray) and motor-driven (black) oscillations. The length of the arrow shows the vector strength of that cell and the direction shows the cell’s phase of activation.

The amplitude of oscillation was similar for freely-walking (median amplitude

147.0 ±33.3°) and tethered flies (median amplitude 144.8±32.5°; n = 23 bouts in 7 freely- walking and 5 tethered animals, Wilcoxon rank-sum test, p = 1; Figure 2.2B). The frequency of oscillation was also similar for freely-walking (median frequency

120.7±67.52Hz) and tethered flies (median frequency 173.6±67.12Hz; n = 24 bouts in 7 freely-walking and 5 tethered animals, Wilcoxon rank-sum test, p = 0.64; Figure 2.2C).

These frequencies were also within ranges previously reported for tethered flying

Sarcophaga(Hall et al., 2015). These results indicated that the self-generated haltere oscillations observed in tethered flies are similar to haltere movements that occur in behaving flies.

Cells activated by self-generated and motor-driven oscillations phase-lock with similar precision

When flies oscillate their halteres during flight and other behaviors, the motion is driven by haltere muscles located within the thorax, beneath the base of the haltere (Chan et al., 1998; Hall et al., 2015). Though these muscles are greatly reduced in size compared with the wing muscles, they are capable of oscillating the halteres independently from the wings (Chan et al., 1998; Hall et al., 2015). When we oscillated the halteres in our

46 experiments, the motion was not driven by the wing or haltere muscles, but instead through the connection of the iron filing attached to the haltere bulb and the magnet attached to the motor. Both types of actuation (self-generated & motor-driven) produced similar motion, as captured by high-speed video (Movie S2; Figure 2.3A-B), but it is possible that these two sources of haltere motion (motor vs. muscles) do not produce the same types of responses in the primary afferents. We compared the activity of cells during spontaneous self-generated oscillations to that of our motor-driven experiments and found that responses to both self-generated oscillations and motor-driven oscillations were phase-locked (Figure 2.2D). Phase-locking activity was quantified using vector strength (Batschelet, 1981). The calculated vector strengths were very high (between 0.81 and 1) for all cells recorded. The median vector strength during self-generated oscillations was 0.97±0.05 and the median vector strength during motor-driven oscillations was 0.96±0.04. These groups were not significantly different (Wilcoxon rank sum test p=0.99; Figure 2.2D). We also found that cells responded with a wide distribution of preferred phases in both self-generated and motor-driven experiments

(Figure 2.2E).

Figure 2.3. Amplitude and frequency threshold is lower for motor-driven oscillations. A) Top trace: haltere elevation during a self-generated oscillation. Lower trace: activity of a single mechanosensory cell in the haltere nerve. The cell spikes in response to the haltere oscillations only when oscillations have high amplitude and frequency. B) Top trace: elevation of a haltere during a motor-driven oscillation. Lower trace: activity of a single mechanosensory cell in the haltere nerve. The cell spikes in response to the haltere

48 oscillations only when oscillations have a high amplitude. C-D) Average amplitude (C) and frequency (D) threshold at which each cell responded during self-generated and motor-driven haltere oscillations. Each point represents a single cell’s average response threshold and the bars represent the median value for all cells recorded in each condition. E-F) The average amplitude (E) and frequency (F) threshold at which a cell responded during self-generated oscillation before and after cutting the nerve. Each point represents a single cell’s average response threshold (n=2 trials per cell); horizontal lines represent the median value for all cells recorded in each condition.

Sensilla respond to low amplitudes and frequencies during motor-driven oscillations, but not during self-generated haltere oscillations

Primary afferent cells were activated at much lower frequencies and amplitudes during motor-driven haltere oscillations than during self-generated oscillations (Figure

2.3A). The amplitude and frequency of the oscillations were gradually increased (a

“sweep” in frequency or amplitude) to determine the thresholds at which cells respond to motor-driven oscillations. Self-generated oscillations naturally increase and decrease in amplitude and frequency over time (Figure 2.3A), so we were able to compare the thresholds of activation for the motor-driven and self-generated haltere oscillations.

Across all cells, self-generated oscillations resulted in a higher amplitude and frequency threshold of activation than motor-driven oscillations.

The median amplitude threshold for self-generated oscillations was 140±25.7°, which was higher than the median motor-driven oscillation amplitude threshold of

26±23.8° (n = 15 self-generated bouts/cells and n = 7 motor-driven bouts/cells in 7 animals, Wilcoxon rank-sum test, p << 0.001; Figure 2.3C). The frequency threshold for self-generated oscillations was 93+-56.9Hz and for the motor-driven oscillations, it was

14±9.1Hz (n = 14 self-generated bouts/cells and n =11 motor-driven bouts/cells in 7 animals, Wilcoxon rank-sum test p << 0.001; Figure 2.3D).

Eliminating neural feedback does not change the activation threshold

The high-speed video recordings of self-generated and motor-driven oscillations showed virtually identical trajectories (Figure 2.3A-B; Movie S2). Why are the responses to motor-driven oscillations elicited at much lower frequencies and amplitudes?

Behavioral state changes have been shown to change the sensitivity of peripheral sensory neurons in flies (Chiappe et al., 2010; Maimon et al., 2010; Suver et al., 2012). We hypothesized that feedback from the central nervous system could raise the firing thresholds of haltere neurons during active oscillations. We removed any direct modulatory inputs to the primary afferents by cutting the haltere nerve and recording from the primary afferent axons above the cut. The amplitude and frequency threshold of activation remained the same in intact and nerve-cut preparations.

The amplitude thresholds for flies with intact (median amplitude threshold

142±10.3°) and cut haltere nerves (median amplitude threshold 140±18.6°) were similar

(n = 6 bouts recorded in 6 cells for each condition in 2 animals, Wilcoxon rank-sum test, p = 0.76; Figure 2.3E). The frequency thresholds for flies with intact (median frequency threshold 182±14.1Hz) and cut haltere nerves (median frequency threshold 174±14.1Hz) were also similar (n = 6 bouts recorded in 6 cells for each condition in 2 animals

Wilcoxon rank-sum test, p = 1; Figure 2.3F).

Amplitude and frequency thresholds change with lateral displacement

To test whether the differences in frequency and amplitude thresholds we saw in the self-generated versus motor-driven oscillations were due to a slight difference in haltere trajectory, we compared activation thresholds within individual cells at different lateral displacements (Movie S1). Deflections were similar across all flies, but due to differences in body size and tethering position, there were slight variations in the deflections for each experiment. More anterior angles of oscillation resulted in lower frequency thresholds across all cells (Wilcoxon rank sum test: p=0.007, N=7 cells, center median=72±16.4Hz, anterior median=22±15.5Hz; Figure 2.4A-B). When the haltere was pushed posterior from the center, there were no responses in any cells even at the highest frequency tested (150Hz; Figure 2.4A-B).

Figure 2.4. Amplitude and frequency thresholds depend on the lateral displacement of the haltere. A) Top trace: haltere position over time during a gradual sweep (increase then decrease, 0-150-0Hz) in frequency.

Lower three traces: activity of a single sensory neuron at three lateral displacements. B) Each circle represents a single cell’s frequency threshold for a particular lateral displacement direction (posterior, center, or anterior). Black bars represent the median value for all cells recorded in each condition. No action potentials were elicited in any cells at any tested frequency in the posterior position during our frequency sweep experiments. N = 7 cells, one trial per cell. C) Top trace: haltere position over time during an amplitude sweep. Lower three traces: activity of a single sensory neuron at three different lateral displacements. In this example, we used an amplitude sweep from 0-60 degrees. D) Each circle represents the lowest amplitude that elicited a spike at a particular lateral displacement direction (posterior, center, or anterior). Black bars represent the median value for all trials recorded in a single cell for each condition (N

= 10 trials).

To measure the amplitude threshold for different lateral displacements, we increased the amplitude in discrete steps and in sweeps. Within individual cells, the preferred lateral displacement position (at which the cell was most likely to respond to low amplitude oscillations) was consistent across trials. Most of the cells (7 out of 8) were most likely to respond to low amplitude oscillations in the anterior position and least likely in the posterior position (Kruskal-Wallis test: p << 0.001, n=10 trials per cell, posterior median=32.3±1.73°, center median=30.9±1.51°, anterior median=27.3±3.5°;

Figure 2.4C-D). One cell was most likely to respond to low amplitude oscillations in the center position, with equal likelihood of response in the anterior and posterior positions.

These results are consistent with Pringle’s first hypothesis that changes in the stroke plane can recruit specific afferents to encode changes in haltere trajectory.

Sensory neurons can encode haltere lateral displacements using phase information

Pringle’s second hypothesis stated that individual afferent neurons could change their spike timing as the haltere’s position changed. We oscillated the haltere at frequencies between 5 and 150Hz at different angles relative to the fly’s body. All cells remained phase-locked to the haltere oscillations regardless of position. We observed three different types of encoding using spike phase.

In some cells, the phase of activation varied across lateral displacement angles: changing the lateral position of the haltere before oscillating it resulted in a different firing phase for the leading spike in each burst (Figure 2.5A-B). In other cells, changing the lateral displacement of the haltere had no effect on the spike phase (Figure 2.5C-D).

Finally, some cells instead of the phase shifting continuously in one direction with increasing lateral displacement, these cells’ phase shifted in the same direction regardless of whether the haltere was pushed anterior or posterior from center (Figure 2.5E-F).

Figure 2.5. Lateral displacement results in a phase-shift for some cells, but not others. A, C, E) Top trace: elevation of the haltere. Lower traces: activity of an example cell that was oscillated in different planes

(lateral displacements) relative to the body. The angle of each haltere displacement was measured relative to the fly’s longitudinal axis and “center” was defined as the angle at which the haltere rested naturally. B,

D, F) Each point represents the phase of activation for a single haltere oscillation in the cell shown at left.

Black bars represent median phase of activation for all oscillations at a given lateral displacement for that cell. Three of N=12 cells are shown here: 2 cells shifted phase in one direction (as in A & B), 6 cells did not shift phase (as in C & D), 2 cells shifted phase in two directions (as in D & E). The remaining 2 cells shifted phase, but only two lateral displacements were recorded and thus could not be classified.

Overall, half of the cells shifted phase and the other half did not (n = 12 cells). We categorized two of the phase-shifting cells as “one-way” because their phase of activation shifted continuously in one direction as we moved the haltere from anterior to posterior

(e.g. Figure 2.5A-B). Two other cells were classified as “two-way” because as the haltere was moved from anterior to center, the phase shifted in one direction, but as the haltere was moved from center to posterior, the phase shifted in the opposite direction (e.g.

Figure 2.5 E-F). In two cells, we were only able to collect data for two lateral displacement positions. There was a phase-shift observed in both cells, but without a third position to compare, it is unclear whether they are “one-” or “two-way” shifting cells. Six cells did not shift their phase in response to lateral displacement (e.g. Figure 2.5

C-D). All twelve cells also exhibited the amplitude and frequency threshold shifts described in Figure 2.4.

Discussion

Seventy years ago, Pringle’s extracellular recordings led to two hypotheses for the mechanism of encoding body rotations in the firing patterns of haltere primary afferent neurons (Pringle, 1948). Here, we provide support for both hypotheses and show that not only can haltere stroke plane be encoded using the firing phase of individual campaniform sensilla, but also that specific afferents are recruited as the haltere’s position changes.

Our experiments showing that the threshold of activation is dependent on the stroke plane (Figure 2.4) support Pringle’s hypothesis that neurons can be recruited when the haltere lateral displacement changes. We show that the adequate stimulus to activate a single neuron in one plane of oscillation may not be adequate in a different plane.

The haltere campaniform sensilla are organized into five distinct fields such that their directional preferences of activation vary (Fraenkel and Pringle, 1938; Pringle,

1948). In particular, the dF1, dF2, and vF1 fields are composed of rows of sensilla oriented parallel to the haltere stalk (presumably to detect the large flapping motion), whereas the vF2 and dF3 fields contain rows of sensilla arranged perpendicular to the stalk, and are maximally sensitive to lateral displacements (Fraenkel and Pringle, 1938;

Gnatzy et al., 1987; Fayyazuddin and Dickinson, 1996; Agrawal et al., 2017). Though the small size of the primary afferent axons prevents the simultaneous measurements of their physiological properties and specific locations, the encoding properties reported here are consistent with the hypothesis that afferent neurons from some, but not all, fields of sensilla will be activated preferentially when the haltere is moved out of its natural stroke plane.

The second hypothesis, that the timing of action potentials could encode haltere motion, was supported by data showing that firing phase shifts with lateral displacement.

Because the phase will shift with different haltere motions, a downstream neuron would be able to use the spike timing of primary sensilla to distinguish different types of haltere motion. The location of a particular sensillum could determine whether its phase will shift in response to a change in haltere trajectory. We used the same types of stimuli for all cells recorded, so if the axons from which we recorded emanated from different

56 sensilla fields, it is not surprising that different cells responded in different ways to the same stimuli.

Both mechanisms of encoding provide information about the haltere’s movement, and both change spike timing in the primary afferents. Similar seemingly-redundant neural codes occur in other sensory systems. Leeches encode touch using relative spike latencies, number of spikes, and instantaneous firing rate (Thomson and Kristan, 2006).

The lateral line system in fish uses both phase and rate coding (Bleckmann, 2008), as does the somatosensory system of primates (Harvey et al., 2013). In visually-stimulated interneurons in locusts, different spike train characteristics (firing rate, peak firing time, and spike count) control specific actions of the downstream motoneurons (Fotowat et al.,

2011). In haltere primary afferents, the two encoding mechanisms have the same effect on a downstream neuron: a depolarizing spike would arrive at a different time when the haltere is laterally deflected. Recordings of downstream neurons will be required to investigate how the two mechanisms described here influence their spiking activity. In

Drosophila, second-order neurons in the antennal lobe receiving the same vibration stimulus express different populations of ion channels, allowing them to encode, in parallel, both the overall energy and the phase/frequency of the vibration (Azevedo and

Wilson, 2017). Similarly, downstream haltere-sensing neurons may respond in different ways to the encoding mechanisms we have described. Regardless, a downstream neuron will likely need to integrate information from multiple afferents to gain a fuller representation of the body’s rotation (Figure 2.6).

Figure 2.6. A schematic of predicted afferent and downstream activity during straight and rotating flight.

Ai) When the fly is not rotating, primary afferents fire action potentials at various phases with respect to the haltere oscillation cycle (top blue trace), or do not fire at all. Five schematic neurons are shown. Aii) When the fly rotates, some afferents are activated or shift their phase of firing. Bi) In straight flight, a downstream interneuron would receive depolarizing EPSPs from the active afferents at various phases of the oscillation, but the membrane is not depolarized above threshold. Bii) During a rotation, EPSPs from multiple afferents would occur simultaneously and could sum to reach threshold. Ci) Traces of action potentials from three different afferents in the same fly when the haltere is oscillated in an anterior position. Each occurs at a unique phase. Cii) When the haltere is moved to a posterior position, the three cells fire at more similar phases. One cell’s action potential occurs sooner and one later, as compared to the anterior position, and the third cell (black trace) does not shift phase.

We predict that populations of sensilla converge onto interneurons (like contralateral haltere interneurons [cHINs]; Strausfeld and Seyan, 1985) or other neurons that are activated in response to different types of haltere motion. For example, there may be a population of primary afferents that are activated at different phases of the haltere

58 oscillation during stable flight (Fox et al., 2010), then shift phases during a yaw rotation so that they then fire simultaneously (Figure. 2.6A). These neurons could converge onto a single interneuron or motoneuron, depolarizing it above its threshold only in response to that specific yaw motion. Similarly, neurons that are silent during straight flight and activated when the haltere is moved laterally could also depolarize a downstream neuron, making it more likely to reach threshold and fire (Figure. 6B). In this way, a downstream neuron could act as a coincidence detector to sense body rotations from the input of a population of sensilla. The postsynaptic integration of input from multiple afferents will depend on the specific properties of the target neurons: recordings from known targets show EPSPs that range from ~1-2 ms (in the wing-steering motoneuron mnb1

(Fayyazuddin and Dickinson, 1996) to ~20 ms (in neck motoneurons; Huston and Krapp,

2009a) in width. These different integration times would infer different sensitivities to the specific timing of input from the afferents. Further recordings from downstream neurons are necessary to fully test this hypothesis, but our data suggest that this is a possible mechanism for encoding the haltere’s position and thus distinguishing different types of body rotations (Figure. 2.6C). Using both of Pringle’s hypothesized encoding mechanisms, a single sensillum becomes an integral part of a population code that can dynamically represent haltere motion.

Supplementary materials

Movie S1: Lateral displacement stimulus. The fly’s head is to the right and the magnet attached to the servo motor oscillates the iron filing glued to the haltere. The recording electrode and ground wire are also visible.

Movie S2: Self-generated and motor-driven haltere oscillations. The fly’s head is to the left and the haltere motions are visible under both self-generated and motor-driven conditions.

Chapter 3

Haltere use and body morphology predict takeoff dynamics in Dipteran insects

Summary

The order Diptera (true flies) are named for their two wings because their hindwings have evolved into specialized mechanosensory organs called halteres. Flies use halteres to detect body rotations and maintain stability during flight and other behaviors (Derham, 1714; Fraenkel and Pringle, 1938; Pringle, 1948; Hall et al., 2015;

Yarger and Fox, 2018). The most recently diverged monophyletic subsection within the dipteran order, the Calyptratae (Wiegmann et al., 2011a), possess distinctive behavioral traits. Calyptratae flies are able to move their halteres independently from their wings and oscillate their halteres during walking behavior (Hall et al., 2015). Here, we document

Calyptratae haltere movements during non-flight behaviors and demonstrate that this subsection of flies uses their halteres to stabilize their bodies during takeoff, whereas non-Calyptratae flies do not. Flies of the Calyptrate are able to take off more rapidly than non-Calyptratae flies, and they gain this speed in part from their stout body morphology and leg extension velocity. This speed comes at the expense of stability, which can be mitigated by haltere use. A closely-related non-Calyptratae species (D. melanogaster) also has a stout body and rapid takeoff, but does not use its halteres to stabilize itself.

Body morphology is thus a good predictor for takeoff duration, but phylogeny better predicts haltere use and contribution to takeoff duration and stability. Haltere use allows for greater speed and stability during fast escapes, but only in the Calyptratae clade.

Methods

Recording ethological data

A high-speed camera (Fastec Imaging, San Diego, CA, USA) captured tethered and freely-behaving flies at 1000-3000 frames per second. Positions of the wings, halteres, head, abdomen, and center of mass were digitized using DLTDATAVIEWER

(Hedrick, 2008). In many trials, two synchronized cameras were used to observe both sides of the animal simultaneously. In trials where wings or halteres were removed or cut, flies were anaesthetized at 0℃ for 2–3 min before surgery. Experiments were performed on adult female black soldier flies (Hermetia illucens; Symton, College Station, TX), and blue bottle flies (Calliphora vomitoria; Josh’s Frogs, Owosso, MI, USA) taken from laboratory colonies. Animals were given food and water ad libitum, and kept on a 12 L:

12 D cycle at 25℃. All additional flies used were collected in Ohio, USA and identified to family level with a dichotomous key (Borror et al., 1992).

Tethered oscillations

Wing-cutting experiments were similar to Deora et al. (Deora et al., 2015). Flies were anesthetized as described above, then tethered dorsal side up to a wooden tether. To shorten the wings, we used a pair of scissors (Fine Science Tools Inc., Foster City, CA) under a dissecting microscope and imaged the wings before and after each cut to measure their relative lengths. Tethered flies were allowed to initiate haltere and wing oscillations spontaneously. Occasionally, oscillations were elicited with a gentle air puff.

Takeoffs

Flies were anesthetized and placed inside plastic tubes (8mm diameter opening) with an electrical breadboard at the top. After waking up, the flies climbed to the top of the platform, from which they could take off. The platform’s surface was metal and connected to a power supply and a resistor and could be triggered to heat up, which would elicit an escape takeoff. The same platform was used at room temperature to observe spontaneous takeoffs, in which no external stimuli were provided. Due to their small size, fruit fly takeoffs were filmed using a smaller platform (a plastic pipette tip) and escapes were elicited by gently touching the abdomen with a paint brush bristle.

Statistics

Haltere synchrony was quantified using vector strength (Batschelet, 1981). All data were non-normally distributed. Comparisons between groups were made using the

Wilcoxon rank sum test with Bonferroni correction. Correlations between two variables were made using Spearman’s rank correlation.

Results & Discussion

Calyptratae flies oscillate their halteres asynchronously during pre-flight behaviors

Although the Calyptratae subsection exhibits extensive genetic and ecological diversity (Stireman et al., 2006; Wiegmann et al., 2011a; Kutty et al., 2019), they display similar locomotor behaviors. All Calyptratae flies (for example, the blow fly Calliphora vomitoria, Figure 3.1A) oscillate their halteres when they are walking, whereas non-

Calyptratae flies (for example, the soldier fly Hermetia illucens, Figure 3.1A; but with the exception of some species of Micropezidae; Figure 3.1B (Hall et al., 2015)) only move their halteres when the wings are also flapping. One notable difference between

64 the haltere movements during walking and takeoff is that the left and right halteres are only synchronized when the wings begin to move. Although blow flies move their halteres differently than soldier flies during walking, the mechanisms that synchronize the halteres in flight— physical linkages including the subepimeral ridge and the scutellum (Deora et al., 2015)— appear to still be in place. Thus, the observed differences in haltere movements during walking are not likely a result of biomechanics of the wings, halteres, and thorax.

Figure 3.1. Locomotor behaviors across fly families. A) Wing and haltere amplitude for a soldier fly (upper traces) and a blow fly (lower traces) during walking (green shaded region) and takeoff (blue shaded region). Blow flies move their halteres during walking and soldier flies do not. Both flies have synchronized haltere oscillations during takeoff. B) Diptera phylogeny of relevant families adapted from Wiegmann et al. (Wiegmann et al., 2011a). Calyptratae flies shown in blue. Non-Calyptratae flies shown in red and black. C) Example wingbeat (teal) and haltere (blue) oscillation frequency at different cut wing lengths for an individual soldier fly (upper) and blow fly (lower). Left and right halteres are only synchronized when the wings are intact and flapping. D) Haltere-haltere vector strength during haltere oscillations in flesh flies where the wings are intact and flapping (left), cut and flapping (center), and intact and stationary (right). E) Time from downstroke initiation (left) and number of wingbeats (right) before

65 feet lose contact with ground during spontaneous takeoffs in representative fly families. Each data point represents an individual species within its color-coded family (1-7 individual animals per species; 1-3 takeoffs per animal).

Deora et al. (Deora et al., 2015) showed that in soldier flies, the wings and halteres are synchronized to each other through mechanical linkages in the thorax. This synchronization breaks down when the wings flap at artificially high frequencies, as they do when the wings are shortened (Deora et al., 2015, 2017). Synchronization of wings and halteres breaks down at high wingbeat frequencies in a Calyptratae fly (the blow fly) as well (Figure 3.1C). Like the soldier fly, the blow fly’s wings must be intact and flapping for the wings and halteres to be linked (Figure 3.1D).

Spontaneous takeoffs in Calyptratae flies are shorter than in non-Calyptratae flies

Although the same mechanical coupling exists for both Calyptratae and non-

Calyptratae representative species (blow flies and soldier flies), which allows for synchronization of the halteres during takeoff, takeoff execution differs between the clades. We observed flies of several species during both escape and spontaneous takeoffs.

We consider spontaneous takeoffs to be those not elicited by any obvious external sensory stimulus. We measured the time between takeoff initiation (the first frame when the wings began moving downward) and takeoff completion (the frame when all feet were no longer in contact with the platform). We found that both the takeoff duration and number of wingbeats in the takeoff differed significantly between the Calyptratae and non-Calyptratae (rank sum p=0.012, p=0.011 respectively). Flies in the Calyptratae were much faster, with takeoffs lasting an average of 6.75+/-1.77 ms and 1.57+/-0.42

66 wingbeats, compared to the non-Calyptratae with takeoffs lasting 49.1+/-32.0 ms and

12.0+/-13.3 wingbeats (Figure 3.1E).

Halteres influence takeoff speed, pitch stability, and leg extension velocity in

Calyptratae

We chose representative species from Calyptratae and non-Calyptratae and analyzed their flight dynamics during escape takeoffs. In insects, escape takeoffs tend to be faster and less stable than spontaneous takeoffs (Trimarchi and Schneiderman, 1995a,

1995b; Burrows and Wang, 2007; Card and Dickinson, 2008a). Thus, differences between the Calyptratae and non-Calyptratae groups will likely be more obvious during the escape takeoffs than in spontaneous takeoffs. In both species, we gently removed the halteres and examined the behavior of haltere-ablated flies to determine the role of haltere sensory input in structuring takeoff behavior.

Consistent with multi-family comparisons above (Figure 3.1E), blow fly

(Calyptratae) spontaneous takeoffs were shorter than soldier fly (non-Calyptratae) spontaneous takeoffs (Figure 3.2A; rank sum p<0.0001). Escape takeoffs were also significantly shorter in blow flies than in soldier flies (rank sum p=0.006). Removing the halteres significantly increased blow fly takeoff duration (rank sum p=0.0007), slowing them to the same duration as soldier fly takeoffs (Figure 3.2A; rank sum p=0.845).

Removing soldier fly halteres had no effect on their takeoff duration (rank sum, p=0.729).

Haltere removal thus lengthens escape takeoff duration in a representative Calyptratae fly, but not in the non-Calyptratae soldier fly.

Figure 3.2. Calyptratae flies without halteres lose speed, pitch stability, and leg extension velocity during escape takeoffs. A) Time from wing downstroke initiation to takeoff in blow flies (blue) and soldier flies (red) for intact spontaneous (left), intact escape (middle), and haltereless escape (right) takeoffs (n=12-15 takeoffs per condition, Bonferroni correction threshold=0.007). B) Schematic demonstrating digitized pitch change (yellow) and translation (cyan) over time during an escape takeoff in a haltereless solder fly (left) and a haltereless blow fly (right). C) Mean pitch change divided by mean translation in blow flies (blue) and soldier flies (red) during the first three wingbeats of intact spontaneous (left), intact escape (middle), and haltereless escape (right) takeoffs (n=9-11 takeoffs per condition, Bonferroni correction threshold=0.007). D) Raw traces (gray), mean (dark red or dark blue), and median (red or blue) of pitch change divided by mean translation in blow flies (blue) and soldier flies (red) during the first three wingbeats of intact spontaneous (left), intact escape (middle), and haltereless escape (right) takeoffs. Some raw traces extend beyond the axis limits E) Raw example traces of middle leg femur-tibia angle (black) and thorax elevation (purple) in blowflies (left) and soldier flies (right). Takeoffs indicated by shaded region. Black dotted line indicates takeoff initiation and gray dotted line indicates liftoff. Top traces show halteres intact escape takeoffs. Bottom traces show halteres removed escape takeoffs. F) Raw traces of middle leg femur-tibia in blowflies (left) and soldier flies (right). Top traces show halteres intact escape takeoffs. Bottom traces show halteres removed escape takeoffs. Example traces represented in E are shown in black. Each trace represents the leg angle only during the takeoff G) Middle leg extension velocity in blow flies

(blue) and soldier flies (red) during takeoff (start of thorax elevation to liftoff) for intact escape (left), and haltereless escape (right) takeoffs (n=7-8 takeoffs per condition, Bonferroni correction threshold=0.0125).

We next examined the haltere’s contribution to stability during takeoff. We measured change in pitch and translation (distance traveled at the center of mass) over the course of the first three wingbeats of each takeoff (Figure 3.2B), because blow fly and soldier fly spontaneous takeoffs were completed within 2-3 wingbeats (Figure 3.1E). We define a metric of pitch stability as pitch change (º) divided by translation (mm). We found that blow flies are less stable in pitch than soldier flies during escape takeoffs with or without halteres removed, but they are equally stable during spontaneous takeoffs

(Figure 3.2C, D; rank sum p<0.0001, p = 0.0001, p=0.315 respectively). Removing the halteres of soldier flies did not change their pitch stability, which was equal in both escape and spontaneous takeoffs. In blow flies, however, escape takeoffs were less stable than spontaneous takeoffs (p=0.0001), and removing the halteres made their escape takeoffs even less stable (p=0.0001; Figure 3.2D).

Most insects take off by producing force with jumping legs, flapping wings, or a combination of both, but jumping strategies tend to produce faster escapes than flapping

(Govind and Dandy, 1972; Pond, 1972; Weis-Fogh, 1973; Sunada et al., 1993; Trimarchi and Schneiderman, 1995a, 1995b; Burrows and Wang, 2007; Card and Dickinson, 2008a;

Fontaine et al., 2009; Chen et al., 2013; Muijres et al., 2017). To quantify the contribution of jumping legs in blow fly and soldier fly takeoffs, we measured the middle leg femur- tibia angle from takeoff initiation (initial elevation of the thorax) to liftoff (time when feet lost contact with ground; Figure 3.2E-G). We found that blow flies have faster middle leg extension velocities during escape takeoffs than soldier flies, which could contribute to

69 their shorter takeoff durations (Figure 3.2E-G). We also found that haltere removal decreases blow fly middle leg extension velocity, but has no effect on soldier flies

(Figure 3.2E-G). This further demonstrates the necessity of haltere input in producing fast escapes, only in the Calyptratae blow flies.

Flies with fast leg extensions and stout bodies have shorter takeoffs

Of the Calyptratae flies we collected, all were observed to use both jumping legs and flapping wings to take off. We predicted that flies that require fewer wingbeats to take off will also rely more on generating lift from jumping legs. Calyptratae flies in general required fewer wingbeats to take off than non-Calyptratae (Figure 3.1E), indicating that Calyptratae may rely more on jumping legs than flapping wings to produce lift than the non-Calyptratae. We measured the middle leg extension velocity during spontaneous takeoffs in each of the representative families and found that takeoff duration was significantly correlated with leg extension velocity (Figure 3.3A).

Drosophilidae were by far the fastest in both categories, however they are some obvious difference in overall body shape between the stout-bodied Drosophila melanogaster and for example, the elongate soldier fly. To investigate the contribution of these different body morphologies, we measured the sagittal aspect ratio (SAR) of the flies by taking the ratio of the dorsoventral and anteroposterior axes for each fly collected, and in doing so, quantified a difference in body shape (Figure 3.3B).

Figure 3.3. Increased leg extension velocity and larger sagittal aspect ratios allow flies to take off faster. A) Correlation between leg extension velocity and time from wing downstroke initiation to takeoff for intact spontaneous takeoffs in all representative families (burgundy) and in all families excluding Drosophilidae (black). B) Schematic demonstrating sagittal aspect ratio (SAR; dorsoventral axis length divided by anteroposterior axis length) of representative families. C) SAR of all representative fly families. Each data point represents an individual species within its color-coded family (n=1-7 individuals per species). D) Correlation between SAR and time from wing downstroke initiation to takeoff for intact spontaneous takeoffs in all representative families. E) Correlation between SAR and number of wingbeats to takeoff for intact spontaneous takeoffs in all representative families. F) SAR plotted against species’ wingbeat frequency for intact spontaneous takeoffs in all representative families.

Calyptratae flies had significantly higher SARs than the non-Calyptratae flies

(Figure 3.3C; rank sum p=0.0016). SAR was significantly correlated with takeoff speed

71 and number of wingbeats to takeoff (Figure 3.3D-E; Spearman’s rank correlation p=0.001, p=0.0002 respectively). SAR was not correlated with wingbeat frequency, suggesting that a shorter body is not an indicator of faster flight dynamics (Figure 3.3F;

Spearman’s rank correlation p=0.751). The non-Calyptratae superfamily most closely related to the Calyptratae subsection is Ephydroidea, which contains the family

Drosophilidae (Wiegmann et al., 2011a). The common fruit fly Drosophila melanogaster has SAR and takeoff speed similar to the Calyptratae flies. Like other non-Calyptratae flies, fruit flies do not use their halteres when walking, but unlike non-Calyptratae they share a similar body shape to the Calyptratae clade (high SAR). This indicates that the increased speed of takeoff exhibited by the Calyptratae flies may also be linked to their body morphology, rather than haltere use only.

Haltere use and morphology in closely related clades

Increasing surface area increases drag, but it also decreases stability (Alexander,

2004). Soldier flies are more stable than blow flies and removing their halteres does not affect their stability, suggesting that soldier flies do not require haltere information during takeoffs.

Intact blow flies extend their legs faster than soldier flies (Figure 3.2G) and

Calyptratae in general have faster leg extensions for spontaneous takeoffs than non-

Calyptrate, but the common fruit fly is the exception (Figure 3.3A). Fruit flies are similar in body shape and leg extension velocity to the Calyptratae, but they do not use their halteres in the same ways. We predicted that fruit flies, unlike the blow flies will still be able to perform fast takeoffs with halteres removed, because they have not been shown to rely on haltere input for non-flying behaviors.

Fruit fly takeoffs were conservatively defined as the first two wingbeats (based on spontaneous takeoff results; Figure 3.1E). We found that like blow flies, fruit flies have short takeoffs (Figures 3.1E, 3.4A) and escape takeoffs in intact fruit flies were not significantly different from blow fly takeoffs in duration (rank sum p=0.334). Fruit fly escape takeoffs, like blow fly escape takeoffs, are less stable than their spontaneous takeoffs (Card and Dickinson, 2008a). Unlike blow flies, however, fruit fly escape takeoff duration and pitch stability were not affected by haltere removal (Figure 3.4).

Because removing the halteres did not affect escape takeoff duration or pitch stability, fruit flies, like soldier flies, do not require haltere information to perform escape takeoffs.

Figure 3.4. Fruit fly escapes are fast, but are not affected by haltere removal. A) Time from wing downstroke initiation to takeoff in fruit flies for intact escape (left), and haltereless escape (right) takeoffs.

B) Raw traces (gray), mean (dark burgundy) and median (burgundy) of pitch change divided by mean translation in fruit flies during the first two wingbeats of intact escape (left), and halteres removed escape (right) takeoffs. C) Mean pitch change divided by mean translation normalized to body length in blow flies (blue), fruit flies (burgundy), and soldier flies (red) during intact escape (left), and halteres removed escape (right).

Fruit flies have been shown to generate most or all (depending on escape type) of their lift through jumping legs (Trimarchi and Schneiderman, 1995b, 1995a; Card and

Dickinson, 2008a; Fontaine et al., 2009), whereas other non-Calyptrate flies like mosquitos and hover flies generate lift primarily from the wings (Chen et al., 2013;

Muijres et al., 2017). Taken together, our analysis of takeoff dynamics across the fly phylogeny suggests that body morphology (stouter flies) and takeoff strategy (jumping over flapping) result in shorter takeoff durations. But for flies in the Calyptratae, taking advantage of the sensory halteres during the transition to flight behavior provides a measure of control during these high-speed maneuvers.

Chapter 4

Conclusions and future directions

Conclusions

In chapter 1, I outlined our current understanding of how halteres move, how their sensors encode forces, and how they transmit information about these forces to the nervous system to guide behavior. I described the sensory structures responsible for detecting strain on the haltere, how the forces acting on these sensors change during body rotations, where in the nervous system this information is transmitted, the types of information that are integrated with haltere information, and what motor outputs result from changes in haltere position information. I also described some open questions that, if answered, might lead us toward a more global understanding of sensory information processing for fly flight. Do halteres play a role in sensing other forces beyond those associated with body rotations? How do higher-order neurons interpret primary afferent activity that results in behavioral commands at the motor level? Do responses change with inputs from other sensory systems, motor feedback, or state-dependent modulation?

And how and why is haltere information represented in the brain?

Single sensory neurons

In chapter 2, I show how the movement of a fly’s body in flight can be encoded by single mechanosensory neurons acting among a population, demonstrating a mechanism for translating movement into predictable patterns of action potentials. We propose a mechanism by which single neurons can encode three-dimensional movement of a mechanosensory organ. Using sharp quartz glass electrodes, I recorded intracellularly from single axons in the haltere nerve. I found that neural activity in the primary afferent cells of the haltere is highly phase-locked to the oscillations of the

76 halteres. However, the specific phase of the neuron’s activity, and the oscillation frequency and amplitude threshold above which sensory neurons will respond, are dependent upon the haltere’s plane of oscillation. In my experiments, I changed the sweep angle of the haltere relative to the fly’s body (its lateral displacement; Figure. 4.1).

As a result of the change in sweep angle, the firing phase and threshold of individual sensory neurons also changed, providing a mechanism for a single cell to encode the haltere’s position. Body rotations that occur during flight also cause lateral displacements of the haltere, suggesting that body rotations may be encoded by the activation threshold and phasing of these neurons (Yarger and Fox, 2018).

Figure 4.1. Haltere sweep angle influences primary afferent firing phase

The encoding mechanisms described here were proposed to be implemented during free-flight body rotations, the sensors on the halteres are not necessarily limited to flight behavior. As demonstrated in chapter 3, the halteres are active during walking in some flies and takeoff in all. Therefore, these findings may provide inspiration for sensors and sensory processing algorithms that can be implemented in flying machines not only during free-flight, but also during transitions from ground to air and back again.

Harnessing mechanisms used by flies to maintain stability in flight and other behaviors could mean spectacular advances in the efficiency and stability of drone technology.

Behavioral context

In our exploration of sensory processing and identification of the mechanisms that encode body position, we must also take into consideration the behavioral contexts in which these systems function. In order to understand the principles of information processing in the nervous system, we need to know what types of information are available to the nervous system in natural contexts and whether that information is behaviorally relevant. Chapter 3 begins to catalogue the types of haltere behaviors used by flies and some of the differences that exist between relatively closely related animals.

Calyptratae flies, the most recently diverged subsection of the dipteran order, were observed to possess distinctive behavioral traits. These flies are able to move their halteres independently from their wings and also oscillate their halteres during walking behavior in addition to flight (Hall et al., 2015). We confirmed that the biomechanics of the thorax responsible for synchronization of the halteres and wings is consistent for

Calyptratae and non-Calyptratae representatives (Deora et al., 2015). We hypothesize that the Calyptratae flies have taken advantage of a sensory system that was already in place and are now able to extract inertial information from the halteres even when the halteres are not synchronized (during walking behavior).

We examined takeoff behaviors in both Calyptratae and non-Calyptratae fly families. Calyptratae flies tended to have stouter bodies (higher ratio of thorax height to body axis length, or sagittal aspect ratio; SAR) and shorter takeoff durations. However,

78 one non-Calyptratae fly species that we measured (Drosophila melanogaster) had a similar SAR to the Calyptratae and also had similarly short takeoff durations. Drosophila also happens to be a member of the superfamily (Drosophilidae) most closely related to the Calyptratae subsection (Wiegmann et al., 2011b). Drosophilidae flies do not oscillate their halteres independently from their wings and do not use their halteres during walking behavior. This gave us the opportunity to examine whether body morphology or haltere usage determined their takeoff dynamics. Drosophila proved to be more similar to soldier flies than to blow flies when the halteres were removed: haltere removal did not affect takeoff duration, even though the intact Drosophila had a short-duration takeoff like the blow flies. These results show that body morphology is a good predictor of takeoff duration, but that phylogeny is a better predictor for haltere contribution to takeoff duration & stability. In other words, haltere use allows for greater speed and stability during fast escapes, but only in the Calyptratae clade.

The results of our comparative experiments demonstrate the importance of examining sensory processing in multiple behavioral contexts. By examining the halteres in different flies and during different behaviors, we expanded our concept of how halteres are used and how they can encode information. Prior to our discovery of asynchronous haltere motion, it was assumed that the left and right halteres need to be synchronized in order to be effective sensors. We now know that some flies (particularly Calyptratae) may be able to detect haltere information during behaviors other than free flight. We demonstrated that in Calyptratae flies, the halteres do contribute to takeoff dynamics and that the halteres are likely important for walking behavior as well (Daltorio and Fox,

2018).

Future Directions

Population code and downstream effects

Campaniform sensilla are ubiquitous in insects (Zill & Seyfarth, 1996; Zill,

Schmitz, & Büschges, 2004). Campaniform sensilla afferents are all activated by strain, and variation in size, orientation, and body location endow these sensors with specificity of the motions they encode. Although we have seen in crane flies (Fox and Daniel,

2008b), blow flies (Fayyazuddin and Dickinson, 1996, 1999; Huston and Krapp, 2009), and now flesh flies (Yarger and Fox, 2016) that the haltere campaniform sensilla primary afferents are precisely timed with haltere oscillations, there is a wide variety of sensor arrangements and morphologies across flies (Agrawal et al., 2017). Agrawal et al. showed that the Schizophora flies (Figure 3.1B) tend to have small halteres with highly stereotyped organization (Figure 4.2A,D) when compared with the Nematoceran clades

(Figure 4.2C).

Figure 4.2. Example haltere campaniform sensilla. A) Schizophora, family Helomyzidae. B) Orthorrapha, family Therevidae C) Nematocerean, family Culicidae. D) Schizophora, subsection Calyptratae, family

Tachinidae. Scale bars denote 10μm

Regardless of the type or location of a campaniform sensilla, all are activated by strain and all halteres oscillate in synchrony with the flapping wings. Though we see similarities in individual cells’ responses to the haltere oscillations, once the information of the entire population is combined, even small differences could result in vastly different encoding schemes. Small differences in the development and life-history of individuals of the same species might also result in translating the same type of body rotation into different patterns of cellular activation. It is unknown what the variability of sensor arrangements and morphologies is within species, but the variability we see between species indicates that past the primary afferents, multiple population encoding strategies are possible.

Population recordings of the haltere nerve

One relatively straightforward way of determining whether the population code proposed in Chapter 2 exists would be to record from the haltere nerve extracellularly using tetrode, tungsten, or similar techniques that allow for recording many individual neurons at once. There are hundreds of campaniform sensilla on each haltere, so the possible combinations of populations (chapter 2, recruitment hypothesis) as well as their relative timing (chapter 2, phase hypothesis) is not trivial. By repeating the lateral displacement experimental paradigm while recording from a larger population of neurons, we could demonstrate the recruitment and phase shifts from our proposed model for how haltere deflections and body rotations are encoded by large populations of

81 afferents in the haltere nerve. Similar experiments were attempted by Pringle (1942); however, because there is such a large population of these cells, sorting spikes from individual neurons was not possible at the time. Since we can now better isolate individual cells and because extracellular recordings are far more stable than intracellular, we could not only test our model for the population code, but also use Pringle’s technique of physically rotating the entire fly. This would allow us to determine the activity of the primary afferents for particular types of body rotations and further investigate how different properties of rotation (speed, direction, duration) influence the population code.

In the experiments described in chapter 2, we used large lateral displacement as a representation for a body rotation because the displacements that occur under natural conditions are too small to replicate with our experimental setup. It was therefore necessary to exaggerate those displacements to be far greater than what occurs for an actual body rotation during free-flight (Nalbach, 1993; Thompson et al., 2009). Even if perfect replications of the motions proposed in the models were possible, they are still only models and the natural haltere trajectories could still be different under free-flight conditions. The closer we can get to a natural body rotation, the more accurate our measurements for how the nervous system encodes rotations will be. We are far more likely to be able to replicate natural conditions while simultaneously recording extracellularly rather than intracellularly because intracellular recordings require a far more stable setup.

Contralateral haltere interneurons

The haltere primary afferent neurons synapse on interneurons in the thorax.

Interneurons integrate and process information from sensory neurons and direct that

82 information to motor neurons, which in turn command the muscles that produce behavior.

The interneurons most likely to be viable candidates for recording are located in the thoracic ganglion, where they can receive information not only from the halteres, but also from other sensory structures throughout the body. There are 8-10 contralateral haltere interneurons (cHINs) that receive inputs directly from the haltere nerves (Strausfeld and

Seyan, 1985). As interneurons may receive input from multiple campaniform sensilla, I expect that the individual interneurons would respond to a wider range of haltere motions than the individual primary afferents. There are also only 8-10 cHINs (Strausfeld and

Seyan, 1985), so each is likely able to detect multiple types of body rotations. However, it’s possible that some cHINs are not important for encoding haltere displacements at all.

So while you might expect to see cHINs that respond to one or both halteres as they are displaced, other cHINs could be sensitive to all haltere displacements, but also be highly selective for haltere amplitude or frequency. Determining the sensitivity of these downstream neurons will not only provide us with a mechanism for how the population encodes haltere position, but also tell us what aspects of haltere position are important to the animal and are thus being transmitted to the downstream neurons.

Neck motor neurons

Haltere information may ascend to the brain and be integrated with other sensory information, like vision, to modulate slower or more complex responses to stimuli or experiences (Kathman and Fox, 2019), but for reflexive behaviors, the subsequent information transformation occurs at the motor neurons. To determine how information from the haltere neurons and cHINs are integrated, one location to investigate would be the neck motor neurons (NMNs). Here we could record the activity of the NMNs during

83 haltere stimulation to determine what types of motor responses derive from specific sensory inputs detected by the halteres. Huston and Krapp (2009) have recorded from

NMNs during haltere stimulation and found that some of these neurons need both haltere and visual inputs in order to fire action potentials. In Huston and Krapp’s study, the halteres were oscillated in a single plane, at one amplitude and at only two frequencies.

We would investigate how varying the haltere oscillation amplitude, frequency, and lateral displacement influences the activity of neck motor neurons in order to read the

“code” that creates head movements. If we are able to catalog the neck muscle responses that result from different haltere motions, we could use the resulting head movements to predict the types of rotations the haltere motions represent as they are translated into head position corrections. Recording the NMN activity would allow us to determine how haltere position information is encoded by the motor neurons to produce the appropriate head adjustment in response to a body rotation.

Further flight dynamics and behavior

Modeling effects of morphology on takeoff speed and stability

When all other variables can be controlled, to what extent does morphology govern takeoff speed and stability? Simulating takeoffs through computer models allows us to explore the physics of the takeoff without as many of the confounding factors that come along with living organisms (life history, state dependence, etc). Modeling standardized fly shapes without sensory or behavioral specializations will allow us to investigate the effects of morphology on takeoff dynamics independent from other variables. An example of something we would like to control for is takeoff strategy. As I

84 described in chapter 3, most insect species use some combination of jumping legs and flapping wings to become airborne (Govind and Dandy, 1972; Pond, 1972; Weis-Fogh,

1973; Sunada et al., 1993; Trimarchi and Schneiderman, 1995b, 1995a; Burrows and

Wang, 2007; Fontaine et al., 2009; Chen et al., 2013; Muijres et al., 2017), but within individuals, jumping has been shown to be a faster takeoff strategy (Burrows and Wang,

2007; Card and Dickinson, 2008a). If we can standardize the upward force by having it originate from the center of mass rather than legs or wings, it would allow us to determine to what extent the body shape actually contributes to takeoff stability and duration. Even though SAR is correlated with takeoff duration, the amount of force produced by the legs compared with the wings is also likely also correlated with takeoff duration. The effect of body morphology may even be negligible compared to the effect of takeoff strategy. We expect that both will contribute, but if takeoff strategy contributes to increased speed far more than SAR can, the implication would be that changing takeoff strategy is a more effective way of increasing speed than changing body morphology.

Takeoff strategies and force production

Flying animals need to produce upward force in order to take off, but where this force comes from depends on the animal and on the type of takeoff it is performing. One of the most distinct behaviors of Drosophila, which has been previously described (Card and Dickinson, 2008b, 2008a), is the movement of the legs before the wings during takeoff (Card and Dickinson, 2008b, 2008a). In all other fly species we observed, the wings moved first. Most insects take off by producing force with both jumping legs and flapping wings (Govind and Dandy, 1972; Pond, 1972; Weis-Fogh, 1973; Sunada et al.,

1993; Trimarchi and Schneiderman, 1995b, 1995a; Burrows and Wang, 2007; Fontaine et al., 2009; Chen et al., 2013; Muijres et al., 2017), but it is unclear to what extent different fly species use their legs versus their wings to generate lift. Leg and wing forces during takeoff have been quantified in hover flies (Chen et al., 2013) and mosquitos (Muijres et al., 2017) and both use their wings more than their legs to take off. This is in contrast to

Drosophila, which use their legs more during escape takeoff (Trimarchi and

Schneiderman, 1995b, 1995a; Card and Dickinson, 2008a; Fontaine et al., 2009). Though these three species demonstrate the variability of takeoff strategies in flies, this is insufficient for generalized rules or models that describe the wing and leg force contributions for different types of flies or different types of takeoffs.

Haltere use in walking behaviors

Various walking behaviors have been measured to determine which aspects of walking require haltere information. For each of these experiments, flies with intact halteres executed a walking behavior, then the halteres were gently removed using forceps, and the behavior was repeated. We collected high speed video of each condition

(with and without halteres intact) and analyzed each animal’s performance. Some of the tasks included: walking up stairs, being shaken by a cell phone motor, walking upside- down, walking upward/downward/sideways on a vertical surface, transitioning from vertical to upside down, being shaken on a belly dancer, and standing on the inside edge of a vertical petri dish that was being rotated by a servo motor. In each of these experiments we saw no difference in performance with halteres intact versus removed, so it is still unclear why some flies move their halteres during simple walking tasks while others do not. However, based on the results of Chapter 3, it is likely that the true

86 behavioral relevance of asynchronous walking haltere oscillations can only be observed in flies behaving under extreme conditions. One extreme condition where the halteres have been shown to be involved is being dropped (Daltorio and Fox, 2018). Daltorio and

Fox (2018) demonstrated that the halteres allow flesh flies to detect rapid changes in acceleration. The results of chapter 3 show that the halteres are also involved in takeoffs, another extreme condition. It is not surprising that we can only observe these differences under extreme conditions because selective pressures typically need to be extreme in order to result in observable differences in behavior (Grant and Grant, 2002; Donihue et al., 2018). It is therefore important that we explore not only the simple behaviors, but also the extremes where selective pressures are strong enough to require specializations.

Height and orientation

Every insect has adapted specialized behaviors suitable for the environment they inhabit. If you compare the body orientation of a marsh fly to that of a soldier fly during a takeoff, though these two animals have very similar body morphologies, they do not have similar takeoff strategies. The marsh fly we observed stood up at an angle with its prothoracic legs in the air before taking off, whereas the soldier fly was closer to horizontal and had all six legs in contact with the ground (Figure 4.3).

Figure 4.3. Takeoff orientation in two non-Calyptratae flies.

The marsh fly’s behavior might seem unusual without knowing anything about its ecology, but this particular fly was collected from a cattail adjacent to a pond. Marsh flies tend to rest on reeds and so they typically do takeoff from a more vertical body orientation (and often upside-down!) under natural conditions (Berg and Knutson, 1978).

Takeoff dynamics might also be affected by the orientation of the animal or by the surface from which it takes off. Another area we could explore is how takeoff performance changes when these variables are altered and whether the results could be predicted not only by the body morphology, but also by the natural conditions and behaviors of the animals being tested.

Expanding our catalog of behavioral and phylogenetic diversity

A selection of representative families was used to produce the results of Chapter

3, but the diversity of the dipteran order is vast. Further investigation into the differences not only between the Calyptratae and non-Calyptratae groups but also within each of these groups is still necessary. We have uncovered some similarities that exist within each clade, but in doing this we may have oversimplified the complexity and diversity that exists for takeoff behavior. There are always exceptions to the rules we try to apply to nature and by discovering the ways in which animals have diverged from these imagined laws we can better understand where the limitations actually exist and where multiple strategies can effectively serve. There are already some known anomalies in terms of walking and takeoff behavior including the Micropezidae flies (stilt-legged flies), which are the only non-Calyptratae family to include some species shown to use their halteres during walking (Hall et al., 2015). There are also Phoridae flies (scuttle

88 flies) that are noted for their distinctive escape behavior, in which they run away rather than take off (Varney and Noor, 2010) and there are various flightless flies that still possess halteres (Jacobs, 1900; Miller, 1977; De Jong and Cilibertti, 2014). Further, we of course cannot discount other variables such as sexual dimorphism and life history. But, to determine the rules and limitations of sensory processing in locomotor stability it is necessary that we first produce a catalog of the natural behaviors. This way we can determine which aspects of behavior (takeoff time, pitch stability, etc.) are most relevant and ubiquitous to all dipteran insects.

References

Agrawal S, Grimaldi D, Fox JL (2017) Haltere morphology and campaniform sensilla

arrangement across Diptera. Arthropod Struct Dev 46:215–229.

Alexander D, David E (2004) Nature’s flyers : birds, insects, and the biomechanics of

flight. Johns Hopkins University Press.

Azevedo AW, Wilson RI (2017) Active Mechanisms of Vibration Encoding and

Frequency Filtering in Central Mechanosensory Neurons. Neuron 96:446-460.e9.

Ball S, Morris R (2015) Britain’s Hoverflies: A Field Guide. Princeton University Press.

Barth FG, Ficker E, Federle H-U (1984) Model studies on the mechanical significance of

grouping in compound spider slit sensilla (Chelicerata, Araneida). Zoomorphology

104:204–215.

Batschelet E (1981) Circular statistics in biology. Academic Press.

Bender JA, Dickinson MH (2006) A comparison of visual and haltere-mediated feedback

in the control of body saccades in Drosophila melanogaster. J Exp Biol 209:4597–

4606.

Berg CO, Knutson L (1978) Biology and Systematics of the Sciomyzidae. Annu Rev

Entomol 23:239–258.

Blagburn JM, Sattelle DB (1987) Nicotinic acetylcholine receptors on a cholinergic nerve

terminal in the cockroach, Periplaneta americana. J Comp Physiol A 161:215–225.

Bleckmann H (2008) Peripheral and central processing of lateral line information. J

Comp Physiol A 194:145–158.

Böhm L (1911) Die antennalen Sinnesorgane der Lepidopteren. Arb Zool Inst Wien:219–

246.

Borror DJ, Triplehorn CA, Johnson NF (1992) An introduction to the study of insects.

Fort Worth ; Montréal: Saunders College.

Burrows M (2007) Kinematics of jumping in leafhopper insects (Hemiptera,

Auchenorrhyncha, Cicadellidae). J Exp Biol 210:3579–3589.

Card G, Dickinson M (2008a) Performance trade-offs in the flight initiation of

Drosophila. J Exp Biol 211:341–353.

Card G, Dickinson MH (2008b) Visually mediated motor planning in the escape response

of Drosophila. Curr Biol 18:1300–1307.

Chan WP, Prete F, Dickinson MH (1998) Visual input to the efferent control system of a

fly’s “gyroscope”. Science 280:289–292.

Chan WPP, Dickinson MH (1996) Position-specific central projections of

mechanosensory neurons on the haltere of the blow fly, Calliphora vicina. J Comp

Neurol 369:405–418.

Chapman RF, Simpson SJ, Douglas AE (2013) The insects: structure and function.

Cambridge University Press.

Chen MW, Zhang YL, Sun M (2013) Wing and body motion and aerodynamic and leg

forces during take-off in droneflies. J R Soc Interface 10:20130808.

Chiappe ME, Seelig JD, Reiser MB, Jayaraman V (2010) Walking modulates speed

sensitivity in Drosophila motion vision. Curr Biol 20:1470–1475.

Cole ES, Palka J (1982) The pattern of campaniform sensilla on the wing and haltere of

Drosophila melanogaster and several of its homeotic mutants. J Embryol Exp

Morphol 71:41–61.

Crickmore MA, Mann RS (2006) Hox control of organ size by regulation of morphogen

production and mobility. Science 313:63–68.

Daltorio KA, Fox JL (2018) Haltere removal alters responses to gravity in standing flies.

J Exp Biol 221.

De Jong H, Cilibertti P (2014) How cold-adapted flightless flies dispersed over the

northern hemisphere: phylogeny and biogeography of the snow fly genus Chionea

Dalman (Diptera: Limoniidae). Syst Entomol 39:563–589.

Deora T, Gundiah N, Sane SP (2017) Mechanics of the thorax in flies. J Exp Biol

220:1382–1395.

Deora T, Singh AK, Sane SP (2015) Biomechanical basis of wing and haltere

coordination in flies. Proc Natl Acad Sci U S A 112:1481–1486.

Derham W (1714) Physico-theology: Or, A demonstration of the being and attributes of

God, from His works of creation. Being the substance of XVI sermons preached in

St. Mary Le Bow-church, London, at the honble Mr. Boyle’s lectures, in the years

1711 and 1712. London: W. Innys.

Dickerson BH, Aldworth ZN, Daniel TL (2014) Control of moth flight posture is

mediated by wing mechanosensory feedback. J Exp Biol 217:2301–2308.

Dickinson MH (1990) Linear and nonlinear encoding properties of an identified

mechanoreceptor on the fly wing measured with mechanical noise stimuli. J Exp

Biol 151:219–244.

Dickinson MH, Tu MS (1997) The function of Dipteran flight muscle. Comp Biochem

Physiol Part A Physiol 116:223–238.

Dickinson MH (1999) Haltere-mediated equilibrium reflexes of the fruit fly, Drosophila

melanogaster. Philos Trans R Soc Lond B Biol Sci 354:903–916.

Dombrowski UJ (1991) Untersuchungen zur funktionellen Organisation des Flugsystems

von Manduca sexta (L.).

Donihue CM, Herrel A, Fabre A-C, Kamath A, Geneva AJ, Schoener TW, Kolbe JJ,

Losos JB (2018) Hurricane-induced selection on the morphology of an island lizard.

Nature 560:88–91.

Eberle AL, Dickerson BH, Reinhall PG, Daniel TL (2015) A new twist on gyroscopic

sensing: body rotations lead to torsion in flapping, flexing insect wings. J R Soc

Interface 12:20141088.

Fayyazuddin A, Dickinson MH (1996) Haltere afferents provide direct, electrotonic input

to a steering motor neuron in the blowfly, Calliphora. J Neurosci 16:5225–5232.

Fayyazuddin A, Dickinson MH (1999) Convergent mechanosensory input structures the

firing phase of a steering motor neuron in the blowfly, Calliphora. J Neurophysiol

82:1916–1926.

Fontaine EI, Zabala F, Dickinson MH, Burdick JW (2009) Wing and body motion during

flight initiation in Drosophila revealed by automated visual tracking. J Exp Biol

212:1307–1323.

Fotowat H, Harrison RR, Gabbiani F (2011) Multiplexing of motor information in the

discharge of a collision detecting neuron during escape behaviors. Neuron 69:147–

158.

Fox JL, Daniel TL (2008) A neural basis for gyroscopic force measurement in the

halteres of Holorusia. J Comp Physiol A Neuroethol Sensory, Neural, Behav Physiol

194:887–897.

Fox JL, Fairhall AL, Daniel TL (2010) Encoding properties of haltere neurons enable

motion feature detection in a biological gyroscope. Proc Natl Acad Sci U S A

107:3840–3845.

Fraenkel G (1939) The function of the halteres of flies (Diptera). Proc Zool Soc London

109:69–78.

Fraenkel G, Pringle JWS (1938) Biological Sciences: Halteres of Flies as Gyroscopic

Organs of Equilibrium. Nature 141:919–920.

Galant R, Walsh CM, Carroll SB (2002) Hox repression of a target gene: extradenticle-

independent, additive action through multiple monomer binding sites. Development

129:3115–3126.

Gnatzy W, Grünert U, Bender M (1987) Campaniform sensilla of Calliphora vicina

(Insecta, Diptera). I. Topography. Zoomorphology 106:312–319.

Govind CK, Dandy JWT (1972) Non-fibrillar muscles and the start and cessation of flight

in the milkweed bug,Oncopeltus. J Comp Physiol 77:398–417.

Grant PR, Grant BR (2002) Unpredictable evolution in a 30-year study of Darwin’s

finches. Science 296:707–711.

Hall JM, McLoughlin DP, Kathman ND, Yarger AM, Mureli S, Fox JL (2015) Kinematic

diversity suggests expanded roles for fly halteres. Biol Lett 11.

Harvey MA, Saal HP, Dammann JF, Bensmaia SJ (2013) Multiplexing Stimulus

Information through Rate and Temporal Codes in Primate Somatosensory Cortex

Pack CC, ed. PLoS Biol 11:e1001558.

Hedrick TL (2008) Software techniques for two- and three-dimensional kinematic

measurements of biological and biomimetic systems. Bioinspir Biomim 3:034001.

Heide G (1983) Neural mechanisms of flight control in Diptera. In: BIONA-report 2

(Nachtigall W, ed), pp 35–52. Stuttgart: G. Fischer.

Hengstenberg R (1988) Mechanosensory control of compensatory head roll during flight

in the blowflyCalliphora erythrocephala Meig. J Comp Physiol A 163:151–165.

Hengstenberg R (1991) Gaze control in the blowfly Calliphora: a multisensory, two-stage

integration process. Semin Neurosci 3:19–29.

Hersh BM, Carroll SB (2005) Direct regulation of knot gene expression by Ultrabithorax

and the evolution of cis-regulatory elements in Drosophila. Development 132:1567–

1577.

Hersh BM, Nelson CE, Stoll SJ, Norton JE, Albert TJ, Carroll SB (2007) The UBX-

regulated network in the haltere imaginal disc of D. melanogaster. Dev Biol

302:717–727.

Huston SJ, Krapp HG (2009) Nonlinear integration of visual and haltere inputs in fly

neck motor neurons. J Neurosci 29:13097–13105.

Jacobs J (1900) Diptères. Ann la Société Entomol Belgique 44:106–107.

Johnston C (1855) Original communications: Auditory apparatus of the Culex mosquito.

Q J Microsc Sci:97–102.

Kathman ND, Fox JL (2019) Representation of Haltere Oscillations and Integration with

Visual Inputs in the Fly Central Complex. J Neurosci 39:4100–4112.

Keil TA (1997) Functional morphology of insect mechanoreceptors. Microsc Res Tech

39:506–531.

Kernan MJ (2007) Mechanotransduction and auditory transduction in Drosophila.

Pflügers Arch Eur J Physiol 454:703–720.

Kutty SN, Meusemann K, Bayless KM, Marinho MAT, Pont AC, Zhou X, Misof B,

Wiegmann BM, Yeates D, Cerretti P, Meier R, Pape T (2019) Phylogenomic

analysis of Calyptratae: resolving the phylogenetic relationships within a major

radiation of Diptera. Cladistics 35:605–622.

Land MF, Collett TS (1974) Chasing behaviour of houseflies (Fannia canicularis). J

Comp Physiol 89:331–357.

Lappe M, Bremmer F, van den Berg AV (1999) Perception of self-motion from visual

flow. Trends Cogn Sci 3:329–336.

Maimon G, Straw AD, Dickinson MH (2010) Active flight increases the gain of visual

motion processing in Drosophila. Nat Neurosci 13:393–399.

McAlpine DK (1990) A new apterous micropezid fly (Diptera: Schizophora) from

Western Australia. Syst Entomol 15:81–86.

McKenna DD, Farrell BD (2010) 9-genes reinforce the phylogeny of holometabola and

yield alternate views on the phylogenetic placement of Strepsiptera. PLoS One

5:e11887.

Milde JJ, Seyan HS, Strausfeld NJ (1987) The neck motor system of the fly Calliphora

erythrocephala II. Sensory organization. J Comp Physiol A 160:225–238.

Miller PL (1977) Haltere activity in a flightless hippoboscid fly, Crataerina pallida. J

Insect Physiol 23:855–860.

Mohit P, Makhijani K, Madhavi MB, Bharathi V, Lal A, Sirdesai G, Reddy VR, Ramesh

P, Kannan R, Dhawan J, Shashidhara LS (2006) Modulation of AP and DV

signaling pathways by the homeotic gene Ultrabithorax during haltere development

in Drosophila. Dev Biol 291:356–367.

Muijres FT, Chang SW, van Veen WG, Spitzen J, Biemans BT, Koehl MAR, Dudley R

(2017) Escaping blood-fed malaria mosquitoes minimize tactile detection without

compromising on take-off speed. J Exp Biol 220:3751–3762.

Mureli S, Fox JL (2015) Haltere mechanosensory influence on tethered flight behavior in

Drosophila. J Exp Biol 218:2528–2537.

Nalbach G (1993) The halteres of the blowfly Calliphora. J Comp Physiol A 173:293–

300.

Pflugstaedt H (1912) Die halteren der dipteren. Z Wiss Zool 100:1–59.

Pix W, Nalbach G, Zeil J (1993) Strepsipteran forewings are haltere-like organs of

equilibrium. Naturwissenschaften 80:371–374.

Pond CM (1972) The initiation of flight in unrestrained locusts, Schistocerca gregaria. J

Comp Physiol 80:163–178.

Pringle JWS (1948) The gyroscopic mechanism of the halteres of Diptera. Philos Trans R

Soc Lond B Biol Sci 233:347–384.

Sandeman DC, Markl H (1980) Head movements in flies (Calliphora) produced by

deflexion of the halteres. J Exp Biol 85:43–60.

Sane SP, Dieudonné A, Willis MA, Daniel TL (2007) Antennal mechanosensors mediate

flight control in moths. Science 315:863–866.

Schwyn DA, Hernandez Heras FJ, Bolliger G, Parsons MM, Krapp HG, Tanaka RJ

(2011) Interplay between feedback and feedforward control in fly gaze stabilization.

In: World Congress, pp 9674–9679.

Seelig JD, Jayaraman V (2015) Neural dynamics for landmark orientation and angular

path integration. Nature 521:186–191.

Sherman A, Dickinson MH (2003) A comparison of visual and haltere-mediated

equilibrium reflexes in the fruit fly Drosophila melanogaster. J Exp Biol 206:295–

302.

Smith DS (1969) The fine structure of haltere sensilla in the blowfly Calliphora

erythrocephala (Meig.), with scanning electron microscopic observations on the

haltere surface. Tissue Cell 1:443–484.

Stireman JO, O’Hara JE, Wood DM (2006) Tachinidae: Evolution, Behavior, and

Ecology. Annu Rev Entomol 51:525–555.

Strausfeld NJ, Seyan HS (1985) Convergence of visual, haltere, and prosternai inputs at

neck motor neurons of Calliphora erythrocephala. Cell Tissue Res 240:601–615.

Strausfeld NJ, Seyan HS, Milde JJ (1987) The neck motor system of the fly Calliphora

erythrocephala. J Comp Physiol A 160:205–224.

Sunada S, Kawachi K, Watanabe I, Azuma A (1993) Performance of a butterfly in take-

off flight. J Exp Biol 183.

Suver MP, Mamiya A, Dickinson MH (2012) Octopamine neurons mediate flight-

induced modulation of visual processing in Drosophila. Curr Biol 22:2294–2302.

Thompson RA, Wehling MF, Evers JH, Dixon WE (2009) Body rate decoupling using

haltere mid-stroke measurements for inertial flight stabilization in Diptera. J Comp

Physiol A Neuroethol Sens Neural Behav Physiol 195:99–112.

Thomson EE, Kristan WB (2006) Behavioral/Systems/Cognitive Encoding and Decoding

Touch Location in the Leech CNS.

Thurm U (1984) Beiträge der ultrastruckturforschung zur aufklärung sensorischer

mechanismen. Verhandlungen der Dtsch Zool Gesellschaft.

Tomoyasu Y, Wheeler SR, Denell RE (2005) Ultrabithorax is required for membranous

wing identity in the beetle Tribolium castaneum. Nature 433:643–647.

Trimarchi JR, Schneiderman AM (1995a) Flight initiations in Drosophila melanogaster

are mediated by several distinct motor patterns. J Comp Physiol A 176:355–364.

Trimarchi JR, Schneiderman AM (1995b) Initiation of flight in the unrestrained fly,

Drosophila melanogaster. J Zool 235:211–222. v. Buddenbrock W (1919) Die vermutliche losung der halterenfrage. Pflug Arch

175:125–164.

Varney RL, Noor MAF (2010) The scuttle fly. Curr Biol 20:R466–R467.

Wagner H (1986) Flight performance and visual control of flight of the free-flying

housefly (Musca Domestica L.) III. Interactions between angular movement induced

by wide- and smallfield stimuli. Philos Trans R Soc B Biol Sci 312:581–595.

Walker RG, Willingham AT, Zuker CS (2000) A Drosophila mechanosensory

transduction channel. Science (80- ) 287:2229–2234.

Weatherbee SD, Halder G, Kim J, Hudson A, Carroll S (1998) Ultrabithorax regulates

genes at several levels of the wing-patterning hierarchy to shape the development of

the Drosophila haltere. Genes Dev 12:1474–1482.

Weatherbee SD, Nijhout HF, Grunert LW, Halder G, Galant R, Selegue J, Carroll S

100

(1999) Ultrabithorax function in butterfly wings and the evolution of insect wing

patterns. Curr Biol 9:109–115.

Weis-Fogh T (1973) Quick Estimates of Flight Fitness in Hovering Animals, Including

Novel Mechanisms for Lift Production. J Exp Biol 59.

Wiegmann BM et al. (2011a) Episodic radiations in the fly tree of life. Proc Natl Acad

Sci U S A 108:5690–5695.

Wigglesworth VB (1946) Organs of equilibrium in flying insects. Nature 157:655.

Yarger AM, Fox JL (2016) Dipteran Halteres: Perspectives on Function and Integration

for a Unique Sensory Organ. Integr Comp Biol 56:865–876.

Yarger AM, Fox JL (2018) Single mechanosensory neurons encode lateral displacements

using precise spike timing and thresholds. Proceedings Biol Sci 285:20181759.

Zill S, Schmitz J, Büschges A (2004) Load sensing and control of posture and

locomotion. Arthropod Struct Dev 33:273–286.

Zill SN, Moran D, Varela FG (1981) The Exoskeleton and Insect Proprioception: II.

Reflex Effects of Tibial Campaniform Sensilla in the American Cockroach,

Periplaneta Americana. J Exp Biol 94.

Zill SN, Seyfarth E-A (1996) Exoskeletal Sensors for Walking. Sci Am 275:86–90.

101