Neuro-Muscular Mechanisms of Abdominal Pumping in the Locust

Much interest centres at present on the neural mechanisms which control centrally patterned rhythmical activity in arthropods. Yet in only one example has identification of an underlying oscillator within the CNS been claimed - that driving the scaphognathite beat in crabs (Mendelson, 1971). Now that it is possible to record synaptic and spike activity from cell bodies within ganglia (Hoyle, 1970) and to know something of the morphology of neurones within the CNS through the injection of cobalt (Pitman, Tweedle & Cohen, 1972), it should soon be possible to discover more about the motor neurones and interneurones responsible for some forms of rhythmic activity in insects.

Locust ventilation continues to invite experiments because of the regularity and persistence of the output under trauma. Before central analysis can be attempted, however, more must be known about the peripheral motor output and the functions of abdominal muscles; this paper attempts to supply some of the needed information.

In locusts and dragonfly larvae pumping movements comprise alternate contractions of expiratory and inspiratory muscles (Mill & Pickard, 1972), whereas in cockroaches and many other insects expiratory muscles alone are present, inspiration being achieved by cuticular elasticity (Farley, Case & Roeder, 1967). Recordings from abdominal connectives in crickets (Huber, 1960), locusts (Miller, 1966) and cockroaches (Farley et al. 1967) have demonstrated the presence of bursts of impulses phase-locked with expiration. Evidence is discussed here which supports the hypothesis of earlier workers that the bursts occur in interneurones which co-ordinate the motor output and thereby achieve a well synchronized ventilatory stroke in all abdominal segments.

Schistocerca gregaria Forskål was obtained from the Centre for Overseas Pest Research in London and cultured under normal conditions.

Electromyograms were made from the muscles of an intact, restrained locust by inserting insulated copper wires through the cuticle into appropriate muscles. For recording from lateral nerve trunks the abdomen was cut down the mid-dorsal line, pinned out flat on a block, tissue-side uppermost, and perfused with Ringer’s solution (Usherwood, 1968). Under these conditions rhythmical contractions in ventilatory muscles continued for several hours. Suction electrodes or stainless-steel hooked wires surrounded with a paraffin and petroleum jelly mixture were used for recording from nerve trunks which were usually cut distally to exclude sensory input. Records were also taken from abdominal connectives either intact or after they had been de-sheathed and split into bundles. Intracellular records were taken from muscle fibres with micropipettes filled with 0·6 M-K₂SO₄ and having a tip resistance of 5–20 MΩ. Some experiments were carried out in a chamber continuously ventilated with 5 % CO₂ in air.

Muscles are numbered according to the system of Snodgrass (1935) and Albrecht (1953). There are several descriptions of the innervation of the pre-genital segments of Acrididae (references in Seabrook, 1968), but they are not based on the use of electrophysiological techniques to track the course of axons. However, Tyrer (1971) and Hinkle & Camhi (1972) have used this method to describe some features of the innervation of the dorsal longitudinal muscles in the abdomen.

In the present study nerves have been traced using methylene blue and electrophysiological methods. Most attention has been paid to the fourth abdominal segment, but this is similar to the other pre-genital segments. The metathoracic ganglion fuses with the first three abdominal ganglia during development. This ganglionic mass supplies nerves to the metathorax and to the first three abdominal segments. The thoracic ganglia are labelled GI, GII and GIII: the separate abdominal ganglia are labelled G4 to G8.

Each pre-genital abdominal ganglion gives rise to two pairs of lateral nerve trunks, an anterior dorsal trunk (labelled Ni) and a posterior ventral trunk (labelled N2). In addition a median nerve leaves posteriorly from the dorsal side of each ganglion and passes to the next posterior ganglion, giving rise to a pair of transverse nerves on the way (Fig. 1). Cobalt chloride, introduced into the cut end of Ni and caused to migrate up the axons and into the ganglion by electrophoresis (Pitman et al. 1972; Iles & Mulloney, 1971) has indicated the positions of some cell bodies in the ganglion whose axons run in N1, and has also shown that six of these axons have cell bodies in the next anterior ganglion (Fig. 2D, E). Dorsal longitudinal muscles in the prothorax (Shepheard, 1970),’mesothorax and metathorax (Neville, 1963) receive much of their innervation from the next anterior ganglion, and it may be that the abdominal dorsal longitudinals, innervated by Ni, also receive part of their innervation from the next anterior ganglion. The positions of some cell bodies in G 4, whose axons run in N2 and in the median nerve, are shown in Fig. 2C. The median nerve has been shown electrophysiologically to contain at least four motor axons each of which divides and runs into the two transverse nerves ; four cell bodies appear ventrally in the ganglion when cobalt is introduced into the median nerve (Fig. 2B). In addition neurosecretory axons from the next posterior ganglion are believed to join the median nerve at the point where it divides into the transverse nerves (cf. Chalaye, 1966; Smalley, 1970), and their cell bodies have a dorsal location (Fig. 2A). Probably only a proportion of the neurones with wider axons in peripheral trunks have been stained so far.

In quiescent locusts pumping takes place in the dorso-ventral plane. Dorso-ventral expiratory muscles are innervated mainly by N 2 of the same segment (see Table 1), while the inspiratory muscles, which expand the sterna ventrally, are innervated by the median nerve of the preceding segment. Thus the antagonistic muscles of one segment are innervated by two ganglia. Records, to be described below, from a ganglion therefore comprise expiratory activity to one segment and inspiratory activity to the next posterior segment. The dorso-ventral inspiratory muscles, which are at the anterior end of the segment, may have evolved from intersegmental muscles arising in the preceding segment. Segments 1 and 2 lack inspiratory muscles, and the corresponding axons in the median nerves supply spiracle muscles. Thus the GIII complex gives rise to four median nerves, the anterior two supplying two axons each to the opener and closer muscles of spiracles 3 and 4, while the third and fourth median nerves supply two large axons to each of the inspiratory muscles of segments 3 and 4, and two small axons to the closer muscles of spiracles 5 and 6. The opener muscles of these spiracles are supplied by paired lateral axons running in N1. More posterior spiracles receive a similar pattern of innervation. Those spiracles which normally serve an expiratory function (nos. 5 to 10) may therefore be capable of some unilateral independence as is known to be the case in Blaberus (Miller, 1973). An outline of the median nerve supply to spiracles and ventilating muscles is given in Fig. 3.

Auxiliary forms of pumping appear when the insect is stressed, and these include longitudinal telescoping movements by the abdomen (Miller, 1960). Longitudinal expiratory muscles are innervated by N1, and inspiratory muscles by N2 (Table 1).

Intracellular records from muscle fibres in segment 4, together with extracellular records from peripheral nerves, have shown that normally two, but sometimes three, excitatory axons contribute to ventilatory activity in each fibre. Two ‘slow’ axons usually produce responses of 0·5–5 mV and 2–10 mV respectively in each fibre and a third ‘fast’ axon produces responses 10–45 mV size but the last has not been seen in muscles 186, 192, 194, 195 or 196 (Fig. 4) (Table 1). A complete inventory of the motor axons supplying pumping muscles has not been compiled, nor is the total number of axons supplying each muscle known in most cases. In muscles 182, 183, 187 and 190 small hyperpolarizations have at times been seen, probably coming from activity in inhibitory axons, but they do not make regular contributions to the pumping cycle and are presumably important in other forms of activity.

Normal pumping is accompanied by discrete alternating motor bursts in N 2 and the median nerve. Bursts are also formed in some N1 axons synchronous with those in N 2 even when no apparent longitudinal movement occurs. They probably provide tension in longitudinal muscles and thus prevent extension of the abdomen when the sterna are lifted during expiration. Well-formed bursts occur in N1 during hyperventilation. Most analyses have been carried out on the compound bursts of N 2 and the median nerve, or on intracellular records from selected muscles. The phasing of activity in different muscles during expiration has not been examined, but this and other aspects are the subjects of a current study by Mr R. Hustert at Cologne.

Bursts in N2 and the median nerve, recorded at about the same distance from G 4, normally show no overlap. In one preparation the mean interval between expiration and inspiration was 12 msec, and between inspiration and expiration, 85 msec (Figs. 5 and 6). Conduction delays probably do not materially alter these values at the periphery. Overlaps between antagonistic motor neurone bursts do occasionally occur, however, for example, when ventilation is slow and the firing frequencies of motor units are lower than normal. These observations suggest that reciprocal inhibitory coupling, if it occurs between the antagonists, does so mainly at the level of interneurones and not at that of motor neurones.

In intact locusts the duration of inspiratory activity in each cycle is usually less variable than that of expiratory activity, the latter showing a good correlation with cycle time (Fig. 7). When ventilation is slow, expiratory neurones may start to fire ionically as soon as inspiration ceases and they then produce a burst at higher frequency with the expiratory movement some time later. Alternatively the expiratory Stroke may be divided into two active phases separated by a plateau of tonic firing corresponding to a period of maintained compression (Miller, 1965). Other patterns may occur, particularly when the ventilation frequency is low (Hustert, pers. com.). Muscle records show that expiratory bursts are normally initiated by slow units which accelerate and then decelerate towards the end of the burst (Fig. 8A). Fast units, if they participate, usually fire later in the burst as has already been noticed in this (Hinkle & Camhi, 1972) and other arthropod systems (Davis, 1971). The two inspiratory axons each fire at a similar and steady frequency throughout inspiration, sometimes showing a slight acceleration (Fig. 8B).

After de-afferentation of the abdomen, or isolation of the thoracic and abdominal nerve cord, the frequency of ventilatory bursting declines but the pattern of firing remains essentially unchanged. In intact locusts sensory discharges in phase with ventilation have been recorded from N1 and N 2 similar to those seen in cockroaches by Farley & Case (1968); their contribution is unimportant for the basic patterning of the motor output, but they may play a role in controlling some aspects of the output both tonically and phasically (cf. Davis, 1969; Pearson, 1972).

Brain removal also results in a considerable reduction of pumping frequency, whereas the subsequent removal of other parts of the CNS anterior to GIII has less effect (Fig. 9). Low-threshold CO₂ receptors are believed to lie in or near the brain (Miller, 1960) and their activity may maintain the ventilatory frequency. Continued ventilatory bursting can be recorded from a completely isolated GIH-G8 preparation, or from any single ganglion removed from such a preparation. However, while an isolated GIII continues to produce short inspiratory bursts and longer more variable expiratory firing, as in the intact locust, a single abdominal ganglion, or the G4–G8 chain, produces cycles of activity at a much reduced frequency (sometimes only if treated with CO₂ or submerged in Ringer) with short expiratory bursts and long intervening periods of inspiratory firing (Fig. 10).

The bursting frequency in an intact locust or in the isolated cord can be increased by continuous electrical stimulation of the anterior thoracic connectives. In intact locusts it can be raised to 3 Hz by this means. The response may be produced through the action of command interneurones running between the head and the metathoracic ganglion (Miller, 1966).

During slow ventilation the abdominal segments may sometimes be seen to contract in sequence, either with anterior or with posterior segments leading, but as ventilation accelerates contractions become apparently synchronous in all segments. A reduction of intersegmental delays at higher cycling frequencies has been reported in the gill beat of Limulus (Fourtner, Drewes & Pax, 1971) and it also characterizes locust pumping.

Simultaneous recordings have been made from homologous nerve trunks of up to four segments in a ventilating locust in order to determine the order in which the abdominal ganglia produce motor bursts. Most attention has been paid to records from N2 and the median nerve of GIII, G4, G5 and G6 (Fig. 11). Table 2 shows the results of a comparison of the timing of expiratory bursts from GIII and G4, and histograms of the delays between the different ganglia are shown in Fig. 12.

Although there is much variation in the order in which the ganglia produce expiratory or inspiratory bursts, anterior ganglia tend to produce motor bursts earlier than posterior ones and GIII is usually in the lead. Variations may occur from cycle to cycle within one preparation, as well as between preparations. During fast ventilation delays of about 10–15 msec occur between adjacent ganglia and a total delay of 50–75 msec exists between GIII and G8. However, because of the large amount of variation, not much significance can be attached to these figures. When ventilation is slow, or when parts of the anterior cord have been removed, intersegmental delays may be in the order of hundreds of milliseconds.

Extracellular records from abdominal connectives show bursts of spikes phase-locked to, but slightly in advance of, the expiratory motor bursts (Fig. 13). They can be recorded in the connective at any point between GIII and G8, but the spikes are larger nearer the anterior end. Connective bursts continue after de-afferentation and can be recorded in an isolated GIII–G8 chain. They also appear in the posterior connectives of an isolated GIII. Single spikes have been tracked between GIII and G8 propagating at 1–1·35 m/sec (at 21 °C) with no detectable reduction in velocity as they pass through intervening ganglia. Left and right connective units seem to act independently. They cannot be recorded in the absence of GIII. It seems therefore that they occur in a single interneurone in each connective which extends throughout the abdominal nerve cord without intervening synapses and is activated in GIII.

Sometimes the latter part of a connective burst can be seen to be formed by more than one unit. Some of the additional activity may be in the axons of the six motor cells whose somata are in the ganglion anterior to that from which their axons emerge, as already described. Additional bursts of spikes propagated anteriorly and phase-locked with expiration can sometimes be seen in intact locusts and they are probably in interneurones with afferent connexions.

Posteriorly propagated bursts in each connective usually start at a low frequency, and then accelerate to a steady firing of 50–100 spikes/sec, at about which time the expiratory motor burst starts (Fig. 14). The durations of connective and motor bursts are correlated even though the motor burst often starts ca. 100 msec later (Fig. 15); but there is no relation between the detailed structure of connective and motor bursts. The delay between the onset of connective and expiratory motor bursts sometime shows a correlation with total ventilatory cycle time, but this is not always so (Fig. 16). The start of the connective burst overlaps the end of the preceding inspiratory burst, sometimes by more than 100 msec, but there is usually a pause between the end of the connective burst and the start of the next inspiratory burst. The connective unit fires tonically throughout temporary pauses in ventilation until the next stroke is initiated.

Section of both GII–III connectives leads to slower and sometimes less well co-ordinated ventilation. The spike frequency in each connective unit may be lower and the ‘burst’, phase-locked with expiration, may now comprise a period of uniform tonic firing. Most expiratory bursts are still represented in every ganglion, but occasionally they appear in GIII alone; they are always accompanied by connective bursts and the failures therefore seem to occur more posteriorly within abdominal ganglia (Fig. 17). Even when there is no G 4 expiratory burst, G 4 inspiratory firing may show a temporary drop in frequency which coincides with the connective burst. This suggests that the connective unit weakly inhibits inspiratory firing. At other times bursts may appear in abdominal ganglia with no accompanying GIII burst, but there is then no connective burst.

With one connective cut between GIII and G4 normal coordination usually continues posteriorly if pumping is strong, with a symmetrical output in the two N2 of G4. However, abdominal ganglia may again occasionally fail to match Gill expiratory bursts. At other times they have been seen to form bursts at twice the frequency of GIII (Fig. 18). Frequency doubling of this sort occurs when the GIII bursts are long and G 4 fires only at the beginning and end of them, the two short expiratory bursts thus formed being separated by a period of inspiratory firing in the G 4 median nerve. Activity in the G 4 median nerve is therefore not incompatible with connective-unit firing. Thus while the inspiratory motor neurones apparently receive weak inhibition from the connective unit, they are strongly inhibited during expiratory activity within their own ganglion.

Electrical stimulation of abdominal connectives in an intact or isolated GIII–G8 preparation produces immediate excitatory responses in N2 expiratory motor nerves in several segments, and cessation of activity in inspiratory nerves. Stimulation may initiate a ventilatory cycle, and it is possible to pace ventilation in this way (cf. Mill, 1970). Such stimulation may affect a co-ordinating system which enables more or less synchronized activity to appear in several segments, as will be discussed below.

Electrical stimulation of the lateral nerves is without effect on the median nerves. However, stimulation of the median nerves, at least in isolated ganglia, re-sets the phase of spikes in the train and produces a temporary inhibition of excitatory activity in N2. The effect is probably brought about by antidromic stimulation of the motor nerves since sensory axons are believed to be absent from the median nerves. The result suggests that the median nerves form collateral inhibitory synapses on the N2 motor axons, or an antecedent interneurones.

Each locust abdominal ganglion when isolated is capable of producing burst cycles in ventilatory nerves. We can therefore consider four possible ways in which such activity may be co-ordinated in the intact insect to produce a well synchronized ventilatory output.

1.Fast-acting intersegmental proprioceptive loops may co-ordinate the output, a ventilatory stroke in one segment rapidly initiating one in an adjacent segment.

However, although proprioceptors are active during pumping in intact locusts, isolated cords maintain a co-ordinated rhythm and sensory input cannot therefore be the only co-ordinating mechanism.

2.Each segmental oscillator may trigger activity in the next ganglion as it becomes active. In the crayfish and lobster swimmeret systems a cycle in one segment is believed to cause the next to fire through the activity of co-ordinating interneurones whose bursts are phase-locked to the motor bursts, but start slightly after them (Stein, 1971). While such a system may co-ordinate locust ventilation when GIII is absent, it does not account for the very short intersegmental delays nor explain the connective bursts which are propagated all the way down the cord from GIII in advance of motor bursts.

3.Each segmental oscillator may be triggered by activity in interneurones which run the length of the abdominal cord. Once triggered the pattern of the burst cycle may be determined by each segmental ganglion. A separate set of command interneurones running throughout the abdominal cord may serve to establish a common level of excitability in all ganglia, as in the swimmeret system (Stein, 1971). Such a system does not, however, explain the correlation between the durations of connective and expiratory motor bursts and again fails to account for the form of the connective bursts.

4.Each set of segmental expiratory motor neurones may be driven by an interneurone which determines the initiation, duration and possibly the intensity of their activity. Such a unit, in relaying patterned information, would qualify as a co-ordinating interneurone and be comparable to those postulated in other ventilatory systems (Farley et al. 1967) and in swimmerets (Stein, 1971). It might by-pass or suppress the activity of segmental oscillators as do the interneurones which supply the stomatogastric ganglion of crayfish and which when stimulated at more than 2 Hz suppress local oscillations and drive motor neurones directly (Dando & Selverston, 1972).

With a conduction velocity of 1–1·35 m/sec the postulated co-ordinating interneurone in locusts would be expected to excite G4 7–10 msec after Gill, and to excite G8 25–35 msec after GIII. The measured mean delays are somewhat greater, but the large variability probably means that factors other than conduction delays account for much of the difference. Fluctuating thresholds, or states of excitability, varying independently within ganglia may be more important in determining the firing order than the timing of instructions delivered by the co-ordinating interneurone. The long latency (ca. 100 msec) between the arrival of the first connective spike and the start of the expiratory motor burst suggests that temporal summation, perhaps also affected by local conditions, is important. Locust ventilation apparently provides an example where burst cycles executed by the motor neurones of one segment are organized by neurones in another. Similarly the activity of anterior spiracles, and head pumping, is thought to be organized from the metathoracic ganglion (Miller, 1967), while abdominal motor neurones may respond to a flight oscillator several segments away (Waldron, 1967; Hinkle & Camhi, 1972).

Bentley (1969 a) recorded from the neuropile of the cricket mesothoracic ganglion and found motor units which fired bursts in phase with ventilation. During the interburst they were hyperpolarized by IPSPs. He detected other units firing during the interburst which may have supplied the IPSPs. His results suggest that ventilatory and other motor neurones receive much inhibitory as well as excitatory input.

A model to explain locust ventilation on the lines of alternative 4 above was put forward by Miller (1966). This must now be modified to account for more recent results. The model (Fig. 19) comprises a burst-forming pacemaker situated in the metathoracic ganglion. The ‘discharge’ or relaxation phase of the pacemaker, which is of relatively constant duration, determines the onset and duration of inspiration. It does so by inhibiting the firing of two autoactive units – the co-ordinating interneurones which run the length of the abdominal cord. The duration of the ‘charge’ phase of the pacemaker, which corresponds to expiration and intervening pauses, is controlled extrinsically by command fibres which respond to CO₂, oxygen and other factors elsewhere in the CNS. (An oscillator with similar properties was postulated by Bentley (1969b) to account for the chirp intervals in stridulating crickets.) The co-ordinating interneurones fire throughout this period; the patterning of their activity may be determined by intervening cells or by the direct action of the pacemaker on them. Bursts in the co-ordinating interneurones always precede expiratory motor bursts from GIII and abdominal ganglia. Action by the interneurones on Other neurones within Gill may be looked upon as similar to that in more posterior ganglia. In each ganglion their input is summed by a further interneurone which then distributes excitatory activity to the expiratory motor neurones through synapses with appropriate thresholds. Activity in the co-ordinating interneurones also weakly inhibits the inspiratory motor neurones. These neurones fire at a more or less steady frequency which is usually independent of the cycle time, but they may show a slight acceleration during the burst. The phase of their spikes in a train can be re-set by antidromic stimulation. These features suggest that the motor neurones are autoactive and that they weakly excite each other. They may also inhibit the expiratory motor neurones. No direct evidence for excitatory coupling between synergists has been obtained but there is some evidence from antidromic stimulation for inhibitory coupling between antagonists in one direction, although it cannot yet be said whether this is direct or via an interneurone.

The intraganglionic interneuronal distributor may itself be capable of initiating burst cycles when the co-ordinating interneurone is inactive. Its bursts are short and they determine the expiratory stroke. They are separated by longer periods of endogenous firing by the inspiratory motor neurones. Bursts are simultaneously transmitted through long collaterals to neighbouring ganglia where they initiate further expiratory bursts. Such a mechanism accounts for co-ordinated activity in the absence of GIII. However, in the intact system the endogenous activity of segmental oscillators is normally suppressed by the co-ordinating interneurone.

Many features of this model are hypothetical and unsupported by evidence, but it bears some resemblance to the models put forward by Pearson (1972) to account for cockroach walking and by Bentley (1969b) to explain cricket stridulation. It is hoped that intracellular recordings which are now being made from some of the units involved will provide more direct evidence.

We are indebted to our colleagues Dr Caroline Pond and Dr Malcolm Burrows for assistance and valuable discussions, and to the Science Research Council for financial support.