Mechanisms for the Production of the Motor Output Pattern in Flying Locusts

How does the locust nervous system co-ordinate flight behaviour? Wilson & Weis-Fogh (1962) showed that the wing motion in flight depends on the pattern of alternating bursts of impulses in the depressor and elevator motor neurons. Wilson (1961) established that the thoracic ganglia can produce the basic flight pattern even in the absence of patterned sensory input. So a more precise but limited formulation of the above question is, ‘How does the central nervous system produce the pattern of alternating impulse bursts in the depressor and elevator motor neurons? ‘

One hypothesis (Wilson, 1966) is that the production of the rhythmic bursts depends on the excitatory coupling of synergistic motor neurons (Kendig, 1966; Wilson & Waldron, 1967a). When impulse activity starts in such a group of positively coupled neurons, it tends to spread throughout the group and to reach high frequency. Eventually however the frequency decreases, due to fatigue (or accumulated refractoriness, which has been demonstrated for locust motor axons by Wilson, 19646). Decrease in frequency may also be due to slowly acting collateral inhibition which seems to be present within the synergist group (Wilson & Waldron, 1967 a). Each unit then receives less input from the other units, and is also less responsive. The result may be that all of the units become inactive, and remain inactive until one unit has recovered sufficiently to resume activity in the absence of excitatory feedback from the others. This unit then excites the others, and a new burst of activity begins. If both groups of synergistic motor neurons thus produce rhythmic bursts at the flight frequency, then the only additional factor needed to co-ordinate the whole set of muscles is some mechanism for the alternation of elevator and depressor bursts. This model for the flight system is completed by suggesting that alternation is due to inhibitory interactions between antagonist motor neurons. Such a model has been presented by Wilson (1966), though he did not specify whether these interactions occurred among motor neurons or among interneurons. The latter alternative remains possible, as do a variety of other models involving interneurons. The evidence presented in this paper, however, tends to support Wilson’s model, and it eliminates alternative models which depend solely on motor neuron interactions.

Some of the data relate to another general question : is the flight system anatomically diffuse, as are many other functional systems?

In most experiments adult male Schistocerca gregaria performed tethered flight in the windstream from an open-jet wind tunnel. Flight muscle potentials were recorded using the techniques described in Wilson & Weis-Fogh (1962). These muscle potentials are known to be in one-to-one correspondence with the action potentials of the motor neurons. The activity of a motor neuron, as recorded by the muscle potentials it induces, is referred to as the activity of a ‘unit’ in this paper. In some experiments abdominal muscle potentials during flight were recorded in a similar manner. In order to record activity in the abdominal ventral nerve cord during flight the dorsal half of the abdomen was removed and the peripheral nerves were severed, thus deafferenting the cord and freeing it so that it could be lifted over bipolar silver electrodes.

Starting and stopping of flight were studied in Locusta migratoria and Schistocerca shoshone as well as in Schistocerca gregaria. Flight initiation was elicited by sudden exposure of the head to wind and by removal of substrate contact for the tarsi of the prothoracic legs. Stopping occurred spontaneously or was induced by blocking the wind. Special techniques used in obtaining the results of the first section are described in that section.

A few standard terms have been used in describing the flight pattern. The period is the duration of a wingbeat cycle. The latency is the interval between the first firing in the burst of impulses in one unit, and the first firing in the next burst of another. The phase is the latency divided by the period.

The basic flight pattern consists of alternating bursts of activity in depressor and elevator motor neurons. In this first set of experiments it was shown that the rhythmic bursts in the depressor motor neurons can continue even when there is apparently no activity in the elevator motor neurons. To obtain this result, I held the wings of a tethered locust which was flying in front of a wind tunnel. Even with the wings held still, healthy animals continue to try to fly. That is, the motor neurons are active in the flight pattern. The period increases, presumably because input from the stretch receptors of the wing hinge is reduced (Wilson & Gettrup, 1963). When the wings are held up, the elevator activity decreases (as previously reported in Wilson & Gettrup, 1963 ; Wilson, 1964 a). I was able to place electrodes in about of the elevator muscles at one time, by taking advantage of their approximately linear arrangement to skewer most of them on each side with a single wire electrode. Even those not directly penetrated were near enough to the recording electrode for muscle potentials in them to have reached the electrode by volume conduction. Under these circumstances the periodic depressor activity continues unchanged through whole cycles in which no elevator activity can be seen. Figure 1 shows a record in which there appears to be no elevator activity through as many as four consecutive cycles of output, though a small amount of elevator activity could be obscured by base-line noise. Since transmission from the motor neurons to these muscles is one-to-one, the absence of elevator muscle activity indicates absence of elevator motor axon activity. Portions of the record in which elevator activity is absent end either with the resumption of elevator activity (even though the upward wing position is maintained) or with the termination of all flight activity. Depressor activity may continue for tens of cycles while elevator activity is very low, though probably not entirely absent. Thus the flight system can produce regular, flight-like bursts of activity in the depressor motor neurons while all or most of the elevator motor axons are inactive. Hence the production of rhythmic bursts apparently does not depend on interactions between antagonistic motor neurons.

In addition to the change in quantity of elevator activity, when the wings are held up, there is a change in timing. As the elevator activity decreases, it occurs relatively later in the wingstroke cycle. In one animal this phase shift sometimes became so extreme that the recorded elevator potentials were synchronous with the depressor potentials (Fig. 2). This synchrony persisted for as many as forty wingbeat cycles, before the elevators gradually resumed their normal earlier phase.

When the wings are held down, rather than up, the elevator activity increases and the depressor activity decreases (as previously reported in Wilson & Gettrup, 1963; Wilson, 1964a). As the elevator activity increases it tends to fill the entire interval between depressor bursts. In this respect the system is asymmetric, for the depressors rarely produce such prolonged bursts when the wings are held up.

With steady electrical stimulation of a flight motor axon, the intervals between successive firings increase because of the accumulation of refractoriness (Wilson, 1964b). During normal bursts, the usual observation is just the opposite; successive intervals decrease (Fig. 3). The depressor bursts show this acceleration, even when there is little or no elevator activity, as in the experiments described in the previous section. This indicates that the acceleration of depressor motor neuron activity is not due to decreasing inhibition from elevator motor neurons during the course of the depressor burst. Nor is the acceleration due to sensory input, since depressor bursts show this acceleration even in the absence of patterned sensory input (Wilson, 1967).

In normal flight, when the wingbeat period increases, both the latency from depressor to elevator activity and the latency from elevator to depressor activity increase proportionately (fig. 14 in Wilson & Weis-Fogh, 1962; fig. 1 in Waldron, 1967). The graphs of Fig. 4 show that as a consequence the phase of the elevator in the depressor cycle remains approximately constant as period increases. This relationship holds equally well for comparisons of muscles innervated from the same ganglion or from different ganglia. One important exception to this phase-constancy occurs when the amount of activity in the elevators changes relative to that in the depressors ; an example of this was given in the first section. Furthermore, the phase of an individual unit advances when the number of impulses per burst in that unit increases (Waldron, 1967).

Flight almost always begins with a burst of activity in many elevator muscles. Often thia elevator burst terminates just as the depressor activity begins (Fig. 5). Production of a more or less normal flight pattern frequently begins immediately with the initial depressor activity.

Sometimes, before the flight pattern begins, there is more prolonged elevator activity with scattered depressor impulses (Fig. 6). If several depressor units are active simultaneously there is a concurrent pause in the elevator activity. When the earliest depressor activity fails to initiate flight, the addition of new depressor units or improved synchronization of the depressors does generally lead to the production of the flight pattern, i.e. to the production of bursts of activity by the elevators, to regular repetition of depressor bursts, and to alternation of antagonist bursts.

The most common deviation from the normal flight pattern during early cycles is the production of abnormally long elevator bursts which terminate only during, rather than before, the next depressor bursts. The occurrence of these long elevator bursts does not seem to be correlated with the frequency of firing within the burst, the timing of the beginning of the burst, or any other feature of the output pattern. This is only one of the circumstances in which the elevators show more tendency to produce prolonged bursts than the depressors do. This difference has already been described for the experiments in which the wings are held stationary. Before the production of the flight pattern begins, depressors are typically active in brief bursts, whereas activity in elevator units is more prolonged. At the end of flight long bursts of activity are common in elevators and rare in depressors. Another difference is that, before the production of flight pattern begins, impulses in different synergist units tend to be synchronous for depressors, but not for elevators. The causes of these differences are unknown, but they may be simple quantitative differences in the properties of elevator and depressor motor neurons.

Under the experimental conditions, the time from the beginning of elevator activity to the beginning of flight-patterned activity could be as brief as 20 msec, or as long as i min. In the latter cases the activity of the various elevator and depressor muscles tended to increase and decrease together. The only significant difference in the starting behaviour of the three species studied was that, under the experimental conditions, Schistocerca gregaria and Locusta migratoria generally started flight quite abruptly, with a brief, simple elevator burst, whereas Schistocerca shoshone often showed longer and more varied activity preceding flight. In contrast, there was no difference of this kind, or of any other observable kind, which correlated with differences in the initial forewing position. (Forewings were sometimes folded between flights and sometimes were left mechanically locked in the open position.) The elevator activity at the beginning of flight is associated with the unfolding of the wings, but the programming for this appears not to depend on reflex feedback, but rather to be built into the central nervous system.

At the end of flight the wingbeat cycle may slow down, most units become less active, and some may become inactive. The last burst of elevator activity may be prolonged as in Fig. 5b. There is rarely disruption of the flight pattern of the kind sometimes seen at the beginning of flight. The order in which the activity in different muscles ends is approximately opposite to the order in which different muscles become active at the beginning of flight. Generally the tergo-sternal elevator muscles become active first, then the other elevators, and lastly the depressors, though there is considerable variation within and between animals. Mesothoracic elevator muscles become active slightly earlier than do metathoracic. There is no evidence for lesser excitability of mesothoracic units. This suggests that threshold differences are not the reason that the cycle of activity during normal flight occurs about 6-10 msec, later in most motor neurons of the mesothoracic ganglion than it does in the metathoracic motor neurons (Wilson & Weis-Fogh, 1962).

The 6−10 msec, timing difference can be observed already in the first cycle of flight, and, for depressor units, even before the over-all flight patterning begins. Even though metathoracic depressors are active earlier than mesothoracic depressors already in the first cycle, the initial burst of elevator activity does not terminate earlier for metathoracic units (Fig. 5). Thus there is a general reciprocal relationship between depressor and elevator activity, but no special relationship for antagonists from the same ganglion as compared to those of different ganglia. In fact, aside from the timing differences already described, there were no differences between units of the meta-and mesothoracic ganglia, and there was no evidence for greater co-ordination of activity in motor neurons of the same ganglion as opposed to motor neurons from different ganglia.

Figure 7 shows a typical record of activity in a ventral abdominal muscle during flight. The muscle becomes active at the same phase in each flight cycle. In recordings from the deafferented ventral nerve cord of the abdomen there is a general increase in activity at the beginning of flight. A few of the recorded units show precise phasing relative to the flight cycle. These units may be active in short bursts in each cycle or once per cycle as in Fig. 76 or only in occasional cycles. Since the cord has been deafferented, this cannot be local reflex activity, but rather must represent information about the flight cycle carried in the abdominal central nervous system. The phase of the activity in the ventral muscles suggests that the activity serves to hold the abdomen steady as the thorax moves up and down during the flight cycle.

Let us return to the original question: ‘How is the C.N.S. organized so that it produces the pattern of alternating bursts of activity in depressor and elevator motor neurons?’ The only well-established interactions in the locust thoracic C.N.S. are between synergistic motor neurons. Synaptic excitatory interactions are shown by the intracellular recordings of Kendig (1966 and unpublished) and by a correlational analysis of non-flight activity in the motor neurons (Wilson & Waldron, 1967 a). As described earlier, a group of neurons which is positively coupled in this way can produce rhythmic bursts of activity.

In fact, if the flight pattern is produced by interactions of motor neurons only, it seems that the only reasonable hypothesis for the production of the rhythmic bursts is that this burst production depends on positive coupling of synergistic motor neurons. The observation that rhythmic bursts continue in the depressor motor neurons even when there appears to be no activity in the elevator motor neurons indicates that the mechanism for the production of the flight bursts cannot depend on interactions between antagonistic motor neurons. This conclusion is supported by the argument of Wilson (1964b), that the pause between bursts of activity in antagonistic motor neurons indicates that production of bursts does not depend on the depressors turning off the elevators and vice versa. On the other hand, the individual motor neurons do not seem to have an inherent tendency to be active in the flight pattern (Wilson, 1961 ; Wilson & Wyman, 1965). Thus, if the flight pattern can be generated by the motor neurons, the rhythmic bursts must be due to the positive interactions among synergists. This positive-feedback system could produce the observed acceleration of the activity within the rhythmic bursts. Bursts produced by model networks of two or more positively coupled, electronic analogue, neuron models (Lewis, 1967) show just such an acceleration, since the input to each unit increases during the bursts as activity in the other units increases (Wilson & Waldron, 1967,b). Though the known positive coupling of synergistic motor neurons and accumulation of refractoriness in motor neurons should result in a tendency to produce rhythmic bursts, as described, it may be that interneurons play a supplementary or even a primary role in the production of the rhythmic bursts. It is only if interneurons do play a role that there are plausible alternative explanations for the observations presented here. The question of interneuron participation in pattern generation has now become a central problem in the analysis of the flight pattern generator (Wilson & Waldron, 1967 b).

Alternation of bursts could be due to inhibitory interactions between antagonist motor neurons, even though these interactions are not necessary for the production of the rhythmic bursts. During flight initiation elevator activity decreases during depressor bursts. This implies that the interaction between antagonistic motor neurons or possible antecedent interneurons is indeed inhibitory, at least in one direction. In fact, this inhibition may play an important role in the initiation of flight ; as the elevator activity changes from continuous pre-flight activity to the rhythmic bursts of flight, the necessary interruptions in the elevator activity usually start when new depressor activity starts. Inhibitory interactions between antagonists are also suggested by the observation that the relative phase of antagonist bursts remains approximately constant as period varies. Wilson (1966) has shown that such constant phase relationships are typically produced by networks of mutually inhibiting neuron models, whereas the activity in mutually exciting neuron models more often shows constant latency relationships. This is because the firing of one neuron tends to occur during increases in the net excitation from the other ; with excitatory interactions this increase occurs during the brief time of the rather rapid rise of the excitatory synaptic potential, whereas with inhibitory interactions this increase in net excitation occurs during the longer time of the slow decrease in the inhibitory synaptic potentials. Thus in the latter case there can be variability in latency. This variability in latency is unlikely with excitatory interactions unless the rising phase of the excitatory potentials is lengthened due to temporal summation or facilitation or both.

The mechanism for the production of periodic bursts in the lobster cardiac ganglion is the same as has been postulated here, being dependent on positive coupling among the group of synergist neurons and on accumulation of fatigue in them (Hagiwara, 1961). Baumgarten & Nakayama (1964) have evidence that production of the respiratory rhythm in the cat depends on the positive coupling of synergistic neurons, the accumulation of fatigue in active units, and inhibitory coupling between antagonists. In both these cases the activity in some units accelerates during the early part of each burst. This acceleration is similar to that observed in flight motor neurons of the locust and has been observed also in the flight motor neurons of satumiid moths (Kammer, 1967a), and the respiratory motor neurons in the abdomen of the dragonfly larva (Mill & Hughes, 1966). Evidence that the production of rhythmic bursts does not depend on the normal interactions between antagonistic units has also been found in related systems. Units that are antagonists during flight can produce synchronous bursts at the flight frequency in the singing grasshopper (N. Elsner, personal communication) and in shivering Lepidoptera (Kammer, 19676). Bursting in the ‘A’ cells of the mud crab’s stomatogastric ganglia continues even in the absence of activity in the reciprocating ‘B’ cells (Maynard & Burke, 1966).

In many cases it has been observed that functional systems are anatomically diffuse. A variety of evidence indicates that this generalization applies to the flight system of the locust. Even though an isolated thoracic ganglion can produce the basic flight pattern, there is little difference in the co-ordination within ganglia as compared to between ganglia in the intact animal. This applies to the activity at the beginning of flight, to the roughly constant phase relationships for antagonists, and to the correlations of the number of firings per burst in different units during flight (Waldron, 1967). Two flight muscles are innervated by axons from the prothoracic ganglion, and yet the temporal pattern of activity in these muscles does not show special characteristics relative to those innervated from the pterothoracic ganglia. Furthermore, there is well-coordinated activity in the muscles and in the C.N.S. of the abdomen. Dugard (1967) reports well-phased activity in the leg muscles. The simplest interpretation of these observations is that there is a single functionally integrated flight system which extends through many ganglia.

In conclusion, the data implies that, if the C.N.S. flight-pattern generator consists of motor neurons only, the production of rhythmic bursts depends on the positive coupling of synergistic motor neurons, while the alternation of antagonist bursts depends on inhibitory coupling between depressor and elevator motor neurons. Though the positive coupling of the synergistic motor neurons very probably does play some role in the production of the rhythmic bursts, interneurons may also play a role in this and in producing alternation of antagonist bursts. Though these interactions can occur within a single thoracic ganglion, the flight system is not anatomically well-localized, but rather extends through most of the ventral nervous system.

Prof. Donald Wilson has been a most stimulating and helpful teacher. The work was done during the tenure of a Predoctoral National Science Foundation Fellowship. Financial support came in part from NIH grant number NB 03927 and Air Force grant number AFOSR1246 to D. M. Wilson.