Connections Between Thoraco-Coxal Proprioceptive Afferents and Motor Neurons in the Locust

Animals need information about the relative positions of their appendages to control posture and coordinate movement. This information is provided by proprioceptors, receptors specialised for detecting and signalling muscle activity or joint position and movement. In both vertebrates and invertebrates, proprioceptive input is important during walking in the production of temporally ordered and precise movements (e.g. Grillner and Zangger, 1979; Pearson, 1985; Sillar et al., 1987), in regulating the transition from one phase of movement to another (e.g. Grillner and Rossignol, 1978; Bässler, 1993) and in the reinforcement of ongoing motor activity during stance (Pearson, 1995a). In quiescent animals, resistance or stretch reflexes predominate, in which muscular activity is recruited by proprioceptive afferents to oppose an imposed movement. During locomotion, these reflexes are substantially altered, often reversing into assistance reflexes in which muscular activity is increased in response to an imposed movement in the same direction (for a review, see Pearson, 1995a).

Arthropod limbs possess a variety of different types of proprioceptor, which may be internal, e.g. chordotonal organs and muscle receptor organs, or external, e.g. hair plates (for reviews, see Mill, 1976; Field and Matheson, 1998). Many of these receptors mediate reflex activity in motor neurons (e.g. Bush, 1962; Burrows, 1987; El Manira et al., 1991). The motor effects of afferent neurons innervating the femoral chordotonal organ (FeCO), for example, which spans the joint between the femur and tibia in the legs of insects, include both resistance and assistance reflexes (Field and Burrows, 1982; Bässler et al., 1986). These afferents connect either directly to motor neurons innervating the muscles that move the tibia or to several populations of local and intersegmental interneurons (for reviews, see Burrows, 1987, 1996).

While the femoro-tibial joint in insects is a hinge joint, with two main muscles controlling the angle between the femur and tibia, the thoraco-coxal joint is a universal joint, allowing movement in three planes. In walking locusts, the metathoracic coxa is moved by six muscle groups (Duch and Pflüger, 1995), so that any movement of the coxa involves a particular combination of muscle actions. Several muscles that move the hind wing also insert on the dorsal coxa, and thoraco-coxal receptors may be sensitive to the contraction of these and other wing muscles (Stevenson, 1997). Coordination of this joint is likely to be of crucial importance to the animal, because its movement profoundly affects leg position, but relatively little is known about the physiology of its proprioceptors. In the locust hind leg, the thoraco-coxal joint has six internal and four external proprioceptors monitoring its movement and position (Bräunig et al., 1981; Pflüger et al., 1981). The internal proprioceptors consist of three chordotonal organs and a strand receptor associated with the articulating membrane of the joint, a group of two multipolar cells associated with the anterior joint chordotonal organ and a myochordotonal organ (Bräunig et al., 1981). In the mesothoracic thoraco-coxal joint, which has an arrangement of proprioceptors somewhat different from that of the metathorax, activity in the chordotonal organ afferents excites motor neurons innervating ipsilateral thoraco-coxal and coxo-trochanteral muscles (Hustert, 1982), with both resistance and assistance reflexes occurring in response to imposed movement. These reflexes, however, vary considerably in intensity, suggesting that direct connections between afferents and motor neurons are either relatively weak or absent (Hustert, 1982). Nothing is known about the connections made by metathoracic thoraco-coxal proprioceptor afferents. This study addresses the paucity of information regarding the control of this joint by investigating connections within the metathoracic ganglion between the afferents of two internal thoraco-coxal proprioceptors, the myochordotonal organ (MCO) and posterior joint chordotonal organ (pjCO), and motor neurons innervating thoraco-coxal and wing muscles. It is shown that some afferent neurons of the MCO make direct chemical synapses with motor neurons innervating thoraco-coxal muscles M121, M126 and M125, while motor neurons innervating other thoraco-coxal and wing muscles receive indirect inputs from afferents of the MCO and pjCO. Preliminary results have been published elsewhere (Wildman, 1998).

Adult desert locusts (Schistocerca gregaria Forskål) of either sex were taken from our crowded colony. Individuals were mounted ventral side uppermost in Plasticine, with their legs restrained. The meso- and metathoracic ganglia and left anterior coxal rotator muscle were exposed by removing a section of exoskeleton (basisternum). The connectives anterior to the mesothoracic ganglion and posterior to the metathoracic ganglion were cut, as were all the metathoracic nerves except for nerve 2C and sometimes nerves 4 and 3A on the left-hand side, depending on the muscles being investigated. All mesothoracic nerves except for nerves 5 were cut. The body cavity was continuously perfused with locust saline (Usherwood and Grundfest, 1965). In experiments in which synaptic transmission was investigated, calcium (4 mmol l⁻¹) was replaced with the same concentration of magnesium.

The apodeme of the anterior coxal rotator muscle (M121; all muscles are named and numbered after Snodgrass, 1929) was cut at its point of insertion onto the ventral coxal rim and held in a pair of fine forceps mounted on a vibrator (Ling Dynamic, V101). Sinusoidal length changes were applied to the muscle under open-loop conditions to stimulate MCO afferents mechanically. Bipolar silver wire hook electrodes or en passant suction electrodes made from polyethylene tubing were used to record spikes from peripheral nerves and to stimulate the nerves electrically. The conduction velocities of some afferents were measured with two recording electrodes separated as far as possible from each other. Electromyograms were recorded from muscles in the thorax using pairs of fine steel pins pushed through the cuticle. Intracellular recordings were made from motor neuron somata using conventional techniques (for details, see Kuenzi and Burrows, 1995). Motor neurons were identified by the presence of antidromic spikes evoked by electrical stimulation of either peripheral nerve 3C (for M121 and the common inhibitor, CI1) or the muscles themselves. Data were digitised at a minimum sampling frequency of 5 kHz using a CED1401 interface and Spike2 software (Cambridge Electronic Design).

The metathoracic MCO is associated with the coxal rotator muscle (M121), which consists of two groups of muscle fibres, anterior and posterior, both of which originate on the metathoracic sternum and share a common insertion onto the anterior ventral rim of the coxa. The MCO inserts anteriorly on the ventral surface of the anterior part (M121a) of the muscle (Fig. 1A) and on the sternal cuticle, close to the origin of the posterior part of the muscle (M121p). The axons of the sensory cells (approximately 30) project to the metathoracic ganglion in nerve 2C (N2C). Some send projections into the first three abdominal neuromeres, which are fused with the metathoracic neuromere to form the metathoracic ganglion. Projections are also sent to the mesothoracic ganglion (Bräunig et al., 1981). N2C also contains the axons of five afferent neurons innervating the pjCO. These axons leave N2C distally, close to the base of the MCO, and form nerve 2C2 (N2C2). The pjCO is situated near the ventral posterior edge of the coxal rim and inserts on the joint membrane and posterior corner of the triangulum (part of the ventral exoskeleton). The axons of the pjCO afferent neurons project to the metathoracic ganglion only (Bräunig et al., 1981). N2C contains no motor neuron axons, nor is it known to contain axons of other sensory neurons.

Under open-loop conditions, sinusoidal length changes applied directly to M121 evoked afferent spikes in N2C (Fig. 1B), but not in N2C2. Most of these spikes occurred during the stretch phase of the stimulus, which corresponds to leg retraction, although in some preparations release-sensitive afferents could be distinguished, signalling a decrease in length of M121 (protraction). Electrically evoked contraction of thoraco-coxal and wing muscles evoked afferent spikes in N2C and N2C2 (Fig. 1C), even with the apodeme of M121 cut free from its insertion onto the coxal rim and only the origin on the sternum remaining intact. Spontaneous contractions of both anterior and posterior parts of M121 under closed-loop conditions also evoked afferent spikes in N2C (Fig. 1D).

Electromyogram recordings were made from four of the six groups of thoraco-coxal muscles (Table 1). In five preparations, the insertion of M121 on the metathoracic coxal rim was left intact and all sensory nerves except for N2C were cut. The vibrator was used to move the coxa, resulting in a rotation about the long axis of the coxa of approximately 20 °. This resulted in an increase in length of M121, slight decreases in the lengths of M123/4 and M125, and a slight increase in the length of M126, after which an unassisted return to the resting position was allowed. These experiments were performed under closed-loop conditions. An increase in spike frequency, including the recruitment of motor units not previously active, occurred in all these muscles during the imposed phase of the stimulus (Fig. 2). Muscle activity declined during the return phase, although the level of activity in M126 remained relatively high for several seconds after the movement had ceased. Reflex activation of motor neurons adapted rapidly, with reflex activity often being lost after three consecutive movements separated by intervals of less than 2–3 s.

The reflex activation of M121, the muscle with which the MCO is associated, was investigated in more detail than that of other muscles. The motor nerve (N3C1) supplying M121 contains the axons of two excitatory neurons (Table 1), innervating the anterior and posterior parts of the muscle separately (Duch and Pflüger, 1995; Hoffmann and Pflüger, 1990), and an axon of common inhibitor 1 (CI1). Action potentials recorded extracellularly from this nerve had clearly different amplitudes, with the largest being those of the phasic motor neuron innervating the posterior part of M121 and the smallest being those of CI1. In most preparations, the tonic motor neuron, which innervates the anterior component of the muscle (M121a), was the only neuron spiking while the preparation was at rest, although sensory stimuli, e.g. touching the mesothoracic tarsi, could evoke short bursts of spikes in all three neurons. Stretch applied directly to M121 under open-loop conditions evoked an increase in the number of spikes in the tonically active motor neuron, but not in either the phasic motor neuron or CI1. This resistance reflex varied in strength both between preparations and over time. Intracellular recordings from the two excitatory motor neurons of M121 showed that both neurons are depolarised when the muscle is stretched (Fig. 3). The amplitude of these synaptic potentials was often large enough to increase the frequency of spikes in the motor neuron innervating M121a, but adapted over the course of the first few stimuli of a series. A similar decrease in the frequency of afferent spikes occurred in N2C.

The connections between N2C afferent neurons and thoraco-coxal motor neurons were established by recording intracellularly from motor neuron somata while stimulating N2C electrically or evoking spikes in sensory afferents by stretching M121. These experiments showed that some MCO afferents make direct, monosynaptic chemical synaptic connections with excitatory motor neurons innervating M121, M126 and M125.

In most preparations, one or two MCO afferent neurons had spikes that were of sufficient amplitude in the recording to be discriminated reliably from the spikes of other neurons. These afferents spiked in response to an increase in M121 length, although some also spiked at a lower frequency during the release phase of the stimulus. Although these afferents had similar response properties, it is not known whether they represented the same neurons in different preparations. When these spikes were used to trigger a signal averager, it was found that they caused depolarising postsynaptic potentials in 33 % of the recordings from the motor neuron innervating M121p and in 50 % of those from the motor neuron innervating M121a (Fig. 4B). In only two of 34 preparations could single postsynaptic potentials, correlated on a 1:1 basis with afferent spikes, be seen without signal averaging (Fig. 4A). Postsynaptic potentials evoked by MCO afferents were also seen in 20 % of the recordings from M126 motor neurons. The mean conduction velocity of N2C afferents was 1.08 m s⁻¹ (range 0.88–1.56 m s⁻¹, N=63) and sensory spikes thus took a mean of 1.85 ms (range 1.03–2.73 ms) to reach the metathoracic ganglion, over a mean distance of 2.0 mm (range 1.6–2.4 mm, N=32). Postsynaptic potentials detected by signal averaging had latencies of 2.9–4.9 ms (mean 3.85 ms, N=15) with respect to the peripherally recorded spike, giving a mean central latency of 2.0 ms (range <1–3.9 ms, N=15). Signal averaging triggered from single afferents recorded from N2C2 provided no evidence of direct connections between pjCO afferents and any of the thoraco-coxal motor neurons.

Single electrical stimuli (10 V, 50 μs) applied to N2C reliably evoked compound postsynaptic potentials in excitatory motor neurons innervating thoraco-coxal muscles M121a,p, M123/4, M125 and M126. Postsynaptic potentials were also evoked in the common inhibitor 1 neuron (CI1). The postsynaptic potentials consisted of an initial, short-latency depolarising component, followed by a second component of longer latency (Fig. 5). The mean latency of the first components of stimulus-evoked postsynaptic potentials in motor neurons innervating M121, M125 and M126, measured between the time of stimulation and the onset of the postsynaptic potential, was 4.0 ms (range 2.8–4.7 ms, N=51). The central latencies of the postsynaptic potentials were therefore 2 ms or less, consistent with monosynaptic connections. The latencies of stimulus-evoked postsynaptic potentials in motor neurons innervating M123/4 and in CI1 were longer (mean 6.3 ms, range 4.8–9.9 ms, N=16), suggesting indirect connections. The amplitude of both postsynaptic potential components showed adaptation, being greatest in response to the first stimulus of a series (Fig. 5A). Both components could reach spiking threshold, but this was more common in tonic than in phasic motor neurons. Although the first component was simple in shape, the second showed greater variability in waveform, often consisting of several peaks of depolarisation. A decrease in stimulus strength usually resulted in a decrease in the amplitude of the second component before the first was noticeably affected (Fig. 5B). The membrane potential often took over 500 ms to return to its pre-stimulus level, although this varied widely between motor neurons and preparations. The effects of electrical stimuli on M126 were particularly marked, with single stimuli evoking long (sometimes more than 1 s in duration) trains of spikes in at least two of the three excitatory motor neurons innervating this muscle. Repetitive stimulation resulted in a decrease in both the number of spikes evoked by each stimulus and the duration of the depolarisation of the motor neuron (Fig. 5C).

In 52 % of the recordings from tonically active motor neurons innervating M121, M123/4 and M126, the second component of the stimulus-evoked postsynaptic potential was hyperpolarising or contained a hyperpolarising component. In one preparation, the hyperpolarising component had a duration of approximately 400 ms in the motor neuron innervating M121a (Fig. 5D, bottom trace). In the other cases, the initial depolarising postsynaptic potential was followed by a short (approximately 40 ms) hyperpolarisation, which in turn was followed by a complex depolarising postsynaptic potential with a very variable amplitude and waveform (Fig. 5D, top trace).

Although direct connections between pjCO afferents and motor neurons were not found, electrical stimulation of N2C2, which contains the five afferent neurons of the pjCO, evoked compound depolarising postsynaptic potentials in the motor neurons innervating M121a, M121p, M125, M126 and M123/4. This indicates that polysynaptic pathways exist between afferents and these motor neurons. A small-amplitude effect was seen in CI1 in 30 % of preparations. The postsynaptic potentials had smaller amplitudes, slower rise times and longer central latencies (mean 15 ms, range 11–26 ms, N=27) than those evoked by stimulating N2C of the same preparations (Fig. 6A). Many postsynaptic potentials evoked by N2C2 stimulation were subthreshold, but in some preparations bursts of spikes were caused by single stimuli (Fig. 6B).

The identification of motor neurons innervating muscles other than M121 made it necessary to leave the innervation of the muscles intact. As a result, muscular activity evoked by electrical stimulation of N2C or N2C2 may in turn have evoked further afferent activity in N2C (e.g. Fig. 6B). This may have contributed to the longer-latency components of postsynaptic potentials recorded in these motor neurons. However, long-duration postsynaptic potentials with complex and variable waveforms were still present in motor neurons innervating M121 in response to N2C stimulation when all other nerves were cut (Fig. 5A,B,D). This indicates that much of the input to motor neurons evoked by electrical stimulation of N2C is due to the afferent barrage resulting from this stimulation, rather than being the result of secondary sensory input resulting from peripheral effects.

Stimulus-evoked compound postsynaptic potentials in motor neurons innervating M121 and M123/4 decreased gradually to 30 % or less of their original amplitudes after 60 min in low-Ca²⁺/high-Mg²⁺ saline (Fig. 7). This is consistent with the presence of chemical synapses in the pathways between N2C afferents and the motor neurons. A sudden drop in postsynaptic potential amplitude, expected where spiking interneurons are interposed between afferent and motor neurons, was not seen, although the longer-latency depolarising components and the hyperpolarising component of the postsynaptic potentials in motor neurons innervating M121 must result from interneuronal input. Upon return to normal saline, the postsynaptic potentials increased gradually in amplitude, reaching 50–60 % of their original amplitude after a further 60 min.

Electrical stimulation of N2C also caused postsynaptic potentials in metathoracic motor neurons innervating wing muscles M113 (first tergosternal=wing elevator) and M127 (first basalar=wing depressor). The postsynaptic potential in M113 was usually more complex in waveform and longer in duration than that in M127, with several depolarising peaks and a duration of 250 ms or more (Fig. 8). In most preparations, the postsynaptic potential in M113 had an initial component of very small amplitude (Fig. 8A), although in two instances this component was larger and coincided with the postsynaptic potential in M127 of the same animal (Fig. 8B). The latencies of the postsynaptic potentials in motor neurons innervating M113 and M127 were greater than those in the thoraco-coxal motor neurons (mean 9.0 ms, range 8.1–12.3 ms, N=22). Spikes were rarely evoked, and only by the first stimulus of a series. Stimulation of N2C2 caused depolarising postsynaptic potentials in motor neurons innervating M113, but these were more variable in latency, amplitude and waveform than those evoked by N2C stimulation (Fig. 8B). Only 30 % of motor neurons innervating M127 responded to N2C2 stimulation, and these responses were variable in sign and waveform, small in amplitude and adapted rapidly. Signal averaging was unable to detect postsynaptic potentials caused by single N2C or N2C2 afferents in either neuron. This, together with the relatively long latency of stimulus-evoked postsynaptic potentials, suggests that the pathways between MCO and pjCO afferents and wing muscle motor neurons are polysynaptic.

The presence in the motor neurons innervating M121 and M126 of depolarising postsynaptic potentials with short, fixed latencies which follow spikes of MCO afferents suggests that monosynaptic connections exist between some MCO afferents and these motor neurons (Fig. 9). Afferent activity in N2C can evoke spikes in thoraco-coxal motor neurons, suggesting that the postsynaptic potentials are excitatory. The central latencies of these postsynaptic potentials and of compound postsynaptic potentials evoked by electrical stimulation of N2C were similar in motor neurons innervating M121, M125 and M126 to those in other chemical synapses in the locust, for example between wing stretch receptor afferents and wing motor neurons (Burrows, 1975) and between afferents of the femoral chordotonal organ and tibial motor neurons (Burrows, 1987). Only in relatively few preparations, however, could individual postsynaptic potentials in motor neurons be correlated with afferent spikes. These postsynaptic potentials were also relatively small in amplitude compared with those evoked in tibial motor neurons by afferents of the FeCO, which reach 2.5 mV in amplitude (Burrows, 1987). Connections between MCO afferents and motor neurons may thus be relatively weak, consistent with the predictions of Hustert (1982). Alternatively, postsynaptic potentials may have been masked by depolarisation resulting from damage caused by the microelectrode, although the injection of hyperpolarising current into these cells failed to reveal postsynaptic potentials.

Another explanation could be that only a limited number of the total population of MCO afferents may synapse with a particular motor neuron, and it is possible that in some preparations stretch applied to M121 failed to activate these particular units. Of the approximately 30 afferent neurons known to innervate this proprioceptor (Bräunig et al., 1981), only two or three with large-amplitude stretch-evoked spikes could be discriminated reliably in any one preparation, and it is not known whether these always represented the same afferents. Particular afferents have been shown to connect to only some of the total number of motor neurons innervating a particular muscle in both the locust femoral chordotonal organ (FeCO) (Burrows, 1987) and crayfish coxo-basal chordotonal organ (CBCO) (Le Ray et al., 1997), with any one motor neuron receiving inputs from a particular group of afferents, but not from all.

Monosynaptic connections between proprioceptive afferents and motor neurons are a feature of neural circuits involved in motor control in both vertebrates and invertebrates. In vertebrates, 1a afferent neurons from muscle spindles synapse directly onto agonist motor neurons and inhibit antagonist motor neurons via inhibitory interneurons (for a review, see Matthews, 1972). Similar connections exist in arthropods: in crayfish, monosynaptic connections have been demonstrated between afferents of the CBCO and motor neurons of the coxo-basal joint muscles (El Manira et al., 1991). In locusts, some afferents of the FeCO, which spans the joint between the femur and tibia, make direct chemical synapses with flexor tibiae motor neurons, whereas others connect with the slow extensor tibiae motor neuron (Burrows, 1987). Similarly, afferents of a chordotonal organ that monitors the position and movement of the tarsus relative to the tibia make direct connections with tarsal motor neurons (Laurent, 1987). Receptors that monitor the metathoracic coxo-trochanteral joint, which include two strand receptors and a muscle receptor organ, make direct connections with trochanteral motor neurons (Skorupski and Hustert, 1991) and some mesothoracic coxal hair plate afferents connect with thoraco-coxal motor neurons (Kuenzi and Burrows, 1995). The direct connections described here are thus consistent with this general principle of neural organisation.

Polysynaptic pathways also exist between N2C afferents and motor neurons. The presence of hyperpolarising components and of depolarising components with longer latencies and durations in electrically evoked compound postsynaptic potentials indicates the existence of interneurons in the pathways between afferents and motor neurons (Fig. 9). Preliminary results have indicated that some afferents of the MCO make direct connections both with local spiking interneurons with somata in the ventral midline region of the metathoracic ganglion and with intersegmental interneurons. Afferents of the FeCO make connections with several different populations of spiking local interneurons, nonspiking local interneurons and intersegmental interneurons (Burrows, 1996; Field and Matheson, 1998), all of which may synapse with motor neurons.

Stevenson (1997) suggested that chordotonal organs associated with the coxa may be important in signalling the contraction times of individual wing muscles. He also showed that trains of electrical stimuli delivered to branches C and D of nerve 2 could occasionally evoke delayed excitatory responses in motor neurons innervating wing muscles M119 and M129. The present study has shown that even single stimuli delivered to N2C can result in excitation of two different wing muscles (M113 and M127), and that MCO and pjCO afferents make polysynaptic connections with motor neurons innervating these muscles (Fig. 9). It would clearly be interesting to identify some of the interneurons involved in flight motor behaviour with which N2C afferents connect. The role played by thoraco-coxal receptors during flight could be investigated by comparing the steering ability of intact locusts with that of locusts in which these receptors have been denervated, using experiments similar to those carried out by Möhl (1993).

A myochordotonal organ was first described in the meropodite of crustacean walking appendages (Barth, 1934), where it signals the end point of the flexion stroke (Fourtner and Evoy, 1973). In locusts, a myochordotonal organ is present in the prothorax, associated with the lateral longitudinal neck muscles (Shepheard, 1973), but nothing is known about its physiology. Some of the metathoracic MCO afferents clearly respond to movements imposed on M121 or on the coxa as a whole (Figs 1B, 2), but recordings have yet to be made from this proprioceptor during natural movements of the leg, such as walking.

Mechanical stimulation of the mesothoracic chordotonal organ system, which monitors coxal position in this body segment and which is not present in the metathorax, results in both assistance and resistance reflexes in thoraco-coxal motor neurons (Hustert, 1982). In the present study, movement applied to the metathoracic coxa caused an increase in activity in all of the thoraco-coxal muscles investigated (Fig. 2), including both resisting and stabilising components which are consistent with the connections reported here. The variability and rapid adaptation of this response suggests that the reflex effects of N2C afferent input are under central control, a feature common to many reflex pathways (Pearson, 1995b). Together with the relative weakness of the monosynaptic connections between afferents and motor neurons, this suggests that the anatomical complexity of the thoraco-coxal joint necessitates relatively complex control circuitry. A high degree of flexibility in the reflex effects of proprioceptive feedback is likely to be needed for this joint to be able to perform a large repertoire of movements under many different behavioural and mechanical conditions. This contrasts with the more predictable and stronger reflexes that assist in the control of the anatomically simpler femoro-tibial joint. It is clear that understanding the role of proprioceptive feedback in the control of movement and posture requires knowledge not only of the functional organisation of these reflex circuits but also of the behaviour- and context-dependent modulation of this organisation. Studying thoraco-coxal proprioceptors may thus provide new information about the complexity of networks controlling movement and posture.

This work was supported by a Newton Trust grant to Malcolm Burrows. I thank M. Burrows, T. Friedel, M. Gebhardt, T. Matheson, O. Morris and two anonymous referees for their constructive comments on the manuscript, and S. Ott for his help with Fig. 1A.