The neuronal control of flight in dragonflies has been investigated by making intracellular recordings from identified motor neurones, singly and in pairs, in tethered flying and non-flying animals.

  1. All the neuronal circuitry necessary to general flight is contained within the thoracic ganglia.

  2. Large rhythmical fluctuations in membrane potential of flight motor neurones occur during flight. They appear to be generated by intemeurones because there is no evidence that they can be produced by individual motor neurones or by direct interactions between motor neurones. Waveforms in motor neurones that innervate the same muscle are often similar but not identical and a motor neurone may spike several times during a wingbeat. Wave frequency and the phase relation of waves in different motor neurones can change with time.

  3. There are several sources of rhythmical input to motor neurones and each motor neurone receives input from a separate set of interneurones. Non-spiking interneurones have been found.

  4. Pathways of delayed excitation are activated by simultaneous spiking in groups of motor neurones, showing that motor neurones have inputs to interneurones.

  5. Rhythms of two frequencies occur in spiracle motor neurones.

  6. Much proprioceptive information from the wing bases reaches thoracic ganglia and affects motor neurone activity.

It is concluded that, as in the locust, populations of interneurones generate the flight rhythm and drive motor neurones. The diversity of motor neurone activity in dragonflies is presumably related to their well-known aerial manoeuvres.

Neuronal circuitry within the thoracic ganglion of a locust is able to generate a flight rhythm without an input of phasic proprioceptive information (Wilson, 1961 ; Wilson & Wyman, 1965). Direct interactions between motor neurones themselves were initially thought to be responsible for generating flight (e.g. Wilson, 1966), but there is no evidence that such connexions, if they exist, are strong enough to do this. Instead, the involvement of interneurones in generating flight rhythms is indicated Ly intracellular recordings which reveal that flight motor neurones can be depolarized in waves at wingbeat frequencies without spiking (Bentley, 1969, in crickets; Burrows, 1973, 1975a, in locusts). In the ventilation by quiescent locusts such waves of depolarization occur in many motor neurones during each expiratory phase and appear to be communicated to both flight and ventilatory motor neurones by the same interneurones that also drive the motor neurones in the slower ventilatory rhythm (Burrows, 1975a, b). However, it is not obvious how these interneurones could co-ordinate motor neurones during flight because in quiescent locusts they depolarize antagonistic and serially homologous motor neurones in phase with each other. Also the rhythms conveyed by these interneurones originate in the metathoracic ganglion but an isolated mesothoracic ganglion has been shown to be capable of producing a flight rhythm (Wilson, 1961). Further indication of the involvement of interneurones in co-ordinating the spiking of motor neurones during flight is the observation that a spike in one tergosternal wing elevator motor neurone may cause a depolarization in that motor neurone and in its contralateral partner after a delay equal to the wingbeat interval (Burrows, 1973). Burrows also occasionally found similar pathways between the tergosternal motor neurone and a first basalar depressor motor neurone.

This paper extends the conclusion that interneurones rather than direct connexions between motor neurones are responsible for generating flight rhythms in another insect: the dragonfly. In dragonflies an individual motor neurone receives a set of inputs that is different from the inputs to any other motor neurone, which must be a reason for the great aerobatic ability of these insects. The anatomy of the motor neurones from which intracellular recordings are made has been described in the previous paper (Simmons, 1977).

Recordings were made from male and female Hemianax papuensis (Burmeister), caught locally. To prepare for an experiment a dragonfly’s legs were cut off and parts of the thoracic cuticle removed to expose the ventral surfaces of the thoracic ganglia and to permit access to some muscles (see Fig. 5 in the preceding paper; Simmons, 1977). The dragonfly was suspended upside-down in a manner that allowed it to move its wings freely (Fig. 1). A small plastic-covered stainless-steel platform, fixed to a syringe needle through which saline (Eibl, 1974) passed, was manipulated under the meso- and metathoracic ganglion and small pins were stuck into the platform to stabilize ganglia against movement. To prepare a dragonfly in this way took between 1 and 2 h, with less than 40% chance of complete success. This paper is based on recordings from 82 preparations.

Electrodes consisting of a pair of 50 µm. silver wires, insulated to their tips, were placed in the neck to stimulate the connectives, and in various muscles to record from them and stimulate the axon terminals of motor neurones antidromically. Platinum hooks or suction electrodes were used to record from and stimulate nerve trunks.

When wings were manipulated their angular movement was registered by using a potentiometer, the spindle of which was attached by a crank to the wing. Sometimes wing movements produced by the animal were recorded by using a capacitance device (Stange & Hardeland, 1970). Abdominal movements were monitored by a semi-conductive strain gauge. A jet of air, directed at the head and monitored by a thermistor, increased tendency to ‘fly’.

Experiments were performed at 22 °C. Local aeshnid dragonflies are seen flying in numbers when the temperature exceeds 18 °C.

Glass microelectrodes, drawn with glass fibres inside, filled with 2 M potassium acetate and with resistances of 40−80 MΩ were used to record intracellular potentials from motor neurones. The tips of these electrodes were strong enough to penetrate the ganglionic sheath but flexible enough to follow some movements of the ganglia. Penetration of the sheath was often aided when tension was applied to it by raising the platform supporting the ganglia. The potential recorded through an electrode fell by 30−50 mV as it passed through the sheath, and entry into a neurone was signified by the appearance of synaptic potentials or spikes rather than any further gross potential changes. The criteria used to identify motor neurones were the same as those described by Hoyle & Burrows (1973), except that it was not feasible to monitor the muscle tension.

Anatomical nomenclature is the same as used in the previous paper (Simmons, 1977).

Patterns of motor neurone spikes during flight and their central origin

During tethered flight bursts of spikes at about 20/s occur in motor neurones that innervate elevator and depressor muscles, bursts to the elevator muscles alternating with bursts to the depressors. Observations with a stroboscope show that wingbeat frequency in tethered dragonflies varies between 10 and 40 beats/s, and that all four wings can show considerable independence in their movements.

All the flight motor neurones studied produce fast twitches in the muscles they innervate. Myograms are often difficult to interpret because of the number of motor neurones, each of which may spike several times in a wingbeat, and because electrodes often move when the muscles contract vigorously. Myograms in Fig. 2 show some features of recording during flight.

Recording en passant from nerve trunks gives a clearer picture of the pattern of individual motor neurone spikes (Fig. 3), although spikes to one muscle alone cannot usually be recorded in this way as nerves often serve several muscles and are arranged in a way that makes it hard to record from a branch that innervates only one muscle (previous paper, Simmons, 1977). The occurrence of the largest spikes can be correlated visually with the contractions of the major flight muscles.

To test whether the central nervous system contains all the neuronal machinery necessary for producing flight, the feedback loops that convey information about wing movements were eliminated by cutting all peripheral nerves of the thoracic ganglia (Fig. 3b, c). Patterns of spikes resembling those in tethered flight persist after the operation, although usually at greatly reduced frequency. The magnitude of the effect that proprioceptive information has on wingbeat frequency is difficult to assess because even in intact animals the frequency is variable.

When the intervening connectives are severed, patterns of spikes resembling those in flight continue in both meso- and metathoracic ganglia. Flight could seldom be elicited by electrical stimuli to anterior and posterior connectives when both head and abdomen were removed, unlike the situation in locusts (Wilson & Wyman, 1965), probably because it is difficult, with gross stimulation, to mimic commands for flight in dragonflies.

The patterning of motor neurone spikes is variable when recorded from nerves at the start of flight. Sometimes elevator and sometimes depressor motor neurones spike first, and over the first few cycles the durations and the intervals between spike bursts become more regular. This is apparently not due to mutilation of the dragonfly for it has also been found in intact aeshnids (Pond, 1973).

Intracellular potentials in motor neurones during flight

The most striking feature of recordings from motor neurones of flying tethered dragonflies is the appearance of large rhythmical fluctuations, or waves, in membrane potential (Figs. 4, 5). Often the waveforms in each member of a pair of motor neurones differ markedly from each other in a particular recording. Spike frequency is always highest at the peaks of waves. If the waves were movement artifacts it would be expected that groups of spikes would often appear at parts of waves other than the peaks, and that waves recorded simultaneously from two different motor neurones would be similar because both electrode tips would be moved at the same time. Many discrete synaptic potentials occur during waves. Waves can occur without spikes (Fig. 5b) but usually each motor neurone spikes several times in a wing beat, and six spikes is the maximum number recorded at the peak of a wave (Fig. 5c-dvm1, third wave).

Simultaneous recordings from motor neurones innervating the same muscle, or from bilaterally homologous motor neurones, are often similar (Fig. 4). Often recordings are more complex, showing features such as waves which change in frequency, or waves in different motor neurones altering their phase relation (Figs, 5d, e). Waves can occur at different frequencies in different motor neurones on the same side of the ganglion at the same time (Fig. 5 c) and waves in a motor neurone may be out of phase with bursts of spikes in other motor neurones that innervate the same muscle (Fig. 5c). The highest frequency of waves that was recorded was 50 Hz (Fig. 5 a) in one preparation, and the lowest was 10 Hz (Fig. 5g) in another preparation.

Inputs to motor neurones in non-flying dragonflies

Synaptic potentials without apparent pattern occur continually in motor neurones of quiescent dragonflies (Figs. 6, 7). Occasionally spikes also occur, often associated with the appearance of small waves and probably concerned with movements such as wing twisting which occur in resting dragonflies. A small proportion of synaptic inputs are common to different motor neurones but, in contrast to the situation in the locust (Burrows, 1975 a), most are not, even when the motor neurones innervate the same muscle (Figs. 6a, b, 7). Regular variation in the amplitude of waves was occasionally seen in flight motor neurones immediately following flight (Fig. 6d). Such beating was most often found in highly dissected restrained dragonflies.

Passage of 10 −20 nA depolarizing current into a motor neurone cell body often elicits spikes in that motor neurone. In some motor neurones, especially those of the depressor muscle pm1, rapidly switching off a hyperpolarizing current is a more effective method of producing spikes. Wave-like potentials were not caused by injection of intracellular current. A spike in a motor neurone of a non-flying dragonfly, hether produced spontaneously or by passage of current, cannot be correlated with potentials in other motor neurones in so far as can be judged by visual inspection of recordings (Fig. 6 a, 7). Depolarizing or hyperpolarizing a motor neurone is without effect on other motor neurones that innervate the same muscle (Fig. 7).

Inhibitory potentials are sometimes clearly seen in flight motor neurones (Figs. 8 a, b). Two sets of flight motor neurones with their axons in nerve 3 often receive inhibitory inputs when spikes occur in their antagonists.

There are a number of sources of excitatory inputs to motor neurones, including air currents directed to the antennae (Figs. 8c, d) or leading edges of the wings, abrupt movements in the visual field, and mechanical disturbance to the abdomen.

No relation was found between abdominal ventilatory movements and potentials in flight motor neurones. In the motor neurones which innervate the thoracic spiracular closer muscles prolonged depolarization with spikes is observed during abdominal inspiratory movements, and conversely I.P.S.P.’S at about 50 Hz, which inhibit spiking, are observed during expiratory movements (Fig. 9). I.P.S.P.’S also occur irregularly during expiration. A similar pattern has been reported from locust spiracular motor neurones (Burrows, 1974; 1975b).

Interneurones

Current pulses applied via an electrode which has passed through the motor neurone cell body layer into the neuropile often produce spikes in the flight motor neurones. Both spiking and non-spiking interneurones were occasionally penetrated, but seldom for more than a few seconds. In one preparation, when nerves to flight muscles were cut, greatly reducing the hazard of movement artifacts, stable recordings from an unidentified non-spiking interneurone revealed depolarizing potentials associated with bursts of spikes in some elevator motor neurones. Passage of hypolarizing current into this cell through the electrode produced spikes in some depressor motor neurones (Fig. 10).

Pathways of delayed excitation between motor neurones

An antidromic spike in a motor neurone is sometimes followed, after a fixed delay, by an excitatory potential in the same neurone (Figs. 11, 12b). This effect was observed in motor neurones controlling the depressor muscles pm1 and dvm3 and the elevator muscle dvm1 In contrast, when spikes were elicited by intracellular injection of current no delayed depolarization occurred. Such delayed excitation of a motor neurone is unlikely to involve proprioceptive feedback as it was observed in preparations where wings were restrained, as well as in preparations where wings were free to move, and the wings did not always move appreciably in the less restrained preparations. Conclusions from these results are discussed later and summarized in Fig. 14.

The delay between an antidromic spike and the following potential is usually 40 − 50 ms, which is the most common interval between wingbeats of a tethered flying dragonfly, but a delay of 20 ms was once found for a pm2 motor neurone (Fig. 11 (a), i – iii).

A pathway of delayed excitation was twice demonstrated to exist between a dvm3 motor neurone and a pmx motor neurone (Fig. 12(b), i – iii). On both occasions an antidromic spike in the dvm3 motor neurone was followed by a variable wave of depolarizing potential in the dvm3 motor neurone and, after 45 ms, by a depolarizing potential in the ipsilateral pm1 motor neurone. In contrast, a train of orthodromic spikes in the dvm3 motor neurone, elicited by releasing it from applied hyperpolarizing current, had no such effect, probably because, as discussed later, such treatment elicits spikes in only one motor neurone (Fig. 12 a). Antidromic spikes in dvm1 motor neurones may be followed by waves of depolarizing potentials in their contralateral partners and in various unidentified neurones. Delayed excitation between elevator and depressor or between depressor and elevator motor neurones has not been found, although appropriate tests have often been made. No interaction of any kind has been found during experiments where two motor neurones that innervate the same muscle were penetrated simultaneously. This would allow (a) orthodromic stimulation of either a single motor neurone or two motor neurones simultaneously, and (b) monitoring of whether shocks to axon terminals in muscles elicited antidromic spikes in only one motor neurone or in two.

Proprioceptive inputs to thoracic ganglia

Absence of one wing appears not to impair greatly the ability of a dragonfly to fly, and I have captured dragonflies which had been flying with one wing atrophied. Some indication of the ability of a dragonfly to adjust flight movements was gained when a marked female dragonfly from which one hind wing had been removed was recaught while copulating and laying eggs; activities performed in flight.

Nerves 1C and 1D contain many sensory axons from the wing, but in contrast to the locust, none is individually recognizable by the size of its spikes. Some receptors respond to wing elevation, others to depression; some are phasic, others tonic (Figs. 13a, b).

Forced wing movements produce synaptic potentials in motor neurones: elevation of any wing excites depressor motor neurones and depression excites elevator motor neurones (Figs. 13 c, d), but the effects of numerous receptors of joints, hairs and thoracic chordontonal organs cannot be distinguished. The anatomical arrangement of nerves, muscles and endocuticle precludes simultaneous recording of sensory spikes and motor neurone P.S.P.’S in a search for monosynaptic connexions, and attempts to record from sensory neurones with wires implanted in veins at the wing base failed because the muscular activity swamped the sensory activity.

Any neuronal mechanism proposed for the generation and control of dragonfly flight must not exclude the variability in patterns of muscular contraction underlying the great aerial manoeuvrability of dragonflies. Mechanisms which regulate the timing of contraction and the power output of different muscles make use of variation in the number of motor neurones that spike and the number of spikes in individual motor neurones during a single wing beat, and also of alteration in phase relations of depolarizing waves in different motor neurones (see Figs. 4, 5).

Three observations indicate that there are several sources of rhythmical input to motor neurones: (a) the variability in frequency of waves in motor neurones (Fig. 5); (b) the changes in phase relations between waves in different motor neurones (Figs. 5 d, e); and (c) beating in the amplitudes of waves sometimes recorded following flights (Fig. 6d). Few P.S.P.’S that are common to a pair of motor neurones in a quiescent dragonfly have been found (Figs. 6a, b, 7), indicating that each motor neurone receives a set of inputs which is distinct from those of any other motor neurone. It is not surprising that direct connexions between motor neurones were not found as these would reduce the independence of action of different motor neurones. Many inputs to motor neurones are likely to come from non-spiking interneurones, which drive (potential changes in motor neurones of cockroaches (Pearson & Fourtner, 1975) and locusts (Burrows & Siegler, 1976), and have been found in central ganglia of cicadas (Simmons, in preparation). Recordings from motor neurones alone cannot identify the non-spiking interneurones that drive them because probably few non-spiking interneurones produce discrete P.S.P.’S. Although recordings from non-spiking interneurones in dragonflies have been made, none have been found which drive motor neurones during flight (the motor neurone spikes in Fig. 10 are not patterned as they are during flight).

The demonstration of pathways of delayed excitation involving motor neurones is important in showing that motor neurones can have inputs to other neurones. To demonstrate a pathway of delayed excitation involving a dragonfly flight motor neurone, an antidromic spike is essential. These pathways were never found following spikes elicited in single motor neurones by current injected into them. The reason is not known but the following hypothesis could explain the difference. Antidromic spikes were produced by stimulating axon terminals in muscles and since each muscle is innervated by several motor neurones with axons of similar diameter (previous paper, Simmons, 1977) the stimulus would produce simultaneous spikes in a group of motor neurones. The ‘delay’ box in Fig. 14, which shows elements in a pathway of delayed excitation, would require spikes in more than one motor neurone for activation. Delayed excitation was not always demonstrated, and can have a different latency at different times (Fig. 11) so that inputs from central neurones to the ‘delay’ box are also required. Variation in motor-neurone spike patterning may involve alternative pathways of delayed excitation in that functional groups of motor neurones, some involving motor neurones of different flight muscles, may be involved in different flight manoeuvres. Because these pathways are often not demonstrable in quiescent dragonflies the best way to reveal pathways of delayed excitation would be by antidromic stimulation of motor neurones out of phase with their expected time of spiking during flight. These experiments would be difficult, however, as flights by tethered, dissected dragonflies are short and unpredictable.

Many interneurones are clearly involved in generating and controlling dragonfly flight. The synaptic potentials that are seen in flight motor neurones of a quiescent dragonfly (Figs. 6, 7) are presumably inputs from interneurones that drive the motor neurones during flight. If so, it is clear that flight rhythms are generated and coupled to motor neurones by non-spiking and spiking interneurones. The interneurones responsible for generating flight rhythms are possibly continually active, as continually active intemeuronal oscillators have been found by recording from locust motor neurones (Burrows, 1975 a, b) and from cicada tymbal motor neurones (Simmons, in preparation). The start of flight may involve alteration in the coupling between rhythmically active interneurones and between interneurones and motor neurones rather than the sudden initiation of the flight rhythm within intemeuronal oscillators. In the execution of complex flight manoeuvres there is presumably a subtle and appropriate balance between central oscillators driving flight motor neurones, proprioceptor excitation from the wings and elsewhere, interaction from interneurones controlling other flight neurones, and descending excitation and inhibition from the head.

Too little is known about the interneurones involved to make detailed comparisons between the neuronal generation of flight in dragonflies and in locusts, or between flight and other rhythmical behaviour. In neither locust nor dragonfly flight systems have direct connexions between flight motor neurones been demonstrated conclusively, ut pathways of delayed excitation between motor neurones have been found in both insects, although in locusts only two muscles, each innervated by a single motor neurone have been investigated (Burrows, 1973). In quiescent locusts the flight motor neurones received depolarizing inputs at the flight frequency and are depolarized in phase with ventilation (Burrows, 1975 a), whereas in quiescent dragonflies no patterning in synaptic inputs to flight motor neurones is observed. This difference may reflect a greater variety in the rhythms and output of the neuronal machinery for flight in dragonflies as compared with locusts, or a failure to emphasize that most of the inputs to locust flight motor neurones come from non-spiking interneurones.

I thank Professor G. A. Horridge for initiating this work and for guidance in performing it. Dr E. E. Ball provided helpful advice during the work. Dr S. Shaw and Dr D. Dvorak criticized the manuscript during various stages of preparation.

Bentley
,
D. R.
(
1969
).
Intracellular activity in cricket neurons during the generation of behaviour patterns
.
J. Insect Physiol
.
15
,
677
699
.
Burrows
,
M.
(
1973
).
The role of delayed excitation in the coordination of some metathoracic flight motoneurons of a locust
.
J. comp. Physiol
.
83
,
135
164
.
Burrows
,
M.
(
1974
).
Modes of activation of motoneurons controlling ventilatory movements of the locust abdomen
.
Phil. Trant. R. Soc. Lond. B
369
,
29
48
.
Burrows
,
M.
(
1975a
),
Co-ordinating interneurones of the locust which convey two patterns of motor commands: their connexions with flight motoneurones
.
J. exp. Biol
.
63
,
713
733
.
Burrows
,
M.
(
1975b
).
Co-ordinating interneurones of the locust which convey two patterns of motor commands: their connexions with ventilatory motoneurones
.
J. exp. Biol
.
63
,
735
753
.
Burrows
,
M.
&
Siegler
,
M. V. S.
(
1976
).
Transmission without spikes between locust interneurones and motoneurones
.
Nature, Lond
,
262
,
322
224
.
Eibl
,
E.
(
1974
).
Verlauf des Hômerven und Projektion der im ‘Hömerven’ verlaufenden Fasern im Prothorakalganglion von Grillen mit Hilfe der CoS Methode
.
Hausarbeit zur Ersten Staatsprüfung für das Lehramt an Gymnasien
,
Cologne University
.
Hoyle
,
G.
&
Burrows
,
M.
(
1973
).
Neural mechanisms underlying behavior in the locust, Schistocerca gregaria. I. Physiology of identified motoneurons in the metathoracic ganglion
.
J. Neurobiol
.
4
,
3
41
.
Pearson
,
K. G.
&
Fourtner
,
C. R.
(
1975
).
Nonspiking interneurons in walking system of the cockroach
.
J. Neurophyriol
.
38
,
33
52
.
Pond
,
C. M.
(
1973
).
Initiation of flight and preflight behaviour of anisopterous dragonflies, Aeschna spp
.
J. Insect Physiol
.
19
,
2225
2229
.
Simmons
,
P. J.
(
1977
).
The neuronal control of dragonfly flight. I. Anatomy
.
J. exp. Biol
,
71
,
123
140
.
Stange
,
G.
&
Hardeland
,
R.
(
1970
).
Eine Methode zur Registrierung der Laufaktivität von kleinen Insekten
.
Oecologica, Berl
.
5
,
400
405
.
Wilson
,
D. M.
(
1961
).
The central nervous control of flight in a locust
.
J. exp. Biol
.
38
,
471
490
.
Wilson
,
D. M.
(
1966
).
Central nervous mechanisms for the generation of rhythmic behaviour in Arthropods
.
Symp. Soc. exp. Biol
.
20
,
199
228
.
Wilson
,
D. M.
&
Wyman
,
R. J.
(
1965
).
Motor output patterns during random and rhythmic stimulation of locust thoracic ganglia
.
Biophys. J
.
5
,
121
43
.