Unpaired Median Neurones in a Lepidopteran Larva (Antheraea Pernyi): I. Anatomy And Physiology

Bilaterally symmetrical nerve cells with bifurcating axons have been described in a number of insect species, mostly after backfilling dye into nerves that innervate muscle blocks (locusts and cockroaches, see Hoyle, 1978; crickets, Davis & Alanis, 1979; Hemiptera, Davis, 1977; Coleoptera, Christensen & Carlson, 1981, 1982; Lepidoptera, Taylor & Truman, 1974; Casaday & Camhi, 1976; Rheuben & Kammer, 1980; Kondoh & Obara, 1982). The association of these cells with nerve trunks that innervate muscles (Hoyle, Dagan, Moberly & Colquhoun, 1974; Davis, 1977; Davis & Alanis, 1979; Kondoh & Obara, 1982; and the present study) suggests that this class of neurones may be involved in the control of the musculature in these phylogenetically diverse groups of insects. In the locust and cockroach many bilaterally symmetrical neurones have been shown to have overshooting action potentials in their somata (Crossman, Kerkut, Pitman & Walker, 1971). In contrast, most insect neurones appear to lack soma spikes (Orthoptera, Hoyle & Burrows, 1973; Dictyoptera, Fourtner & Pearson, 1977; Odonata, Simmons, 1977; Lepidoptera, Rind, 1983).

The present study investigates the anatomy and physiology of two unpaired median neurones (MCI and MC2) in the abdominal segments of A. pernyi, and includes a detailed description of the muscular anatomy of the segment as no previous accounts in the literature were adequate. Results indicated that the two median cells investigated in the present study were uniquely identifiable neurones. Several physiological and anatomical similarities between MCI and MC2 and the dorsal unpaired median (DUM) neurones of locusts and cockroaches were noted.

Caterpillars of the Chinese oak silkmoth Antheraea pernyi were raised from hatching to their third instar in batches of 50-100 in constant light conditions at 18-28°C, depending on the rate of growth required. The animals were fed on beech, hawthorn, oak or sweet chestnut leaves during the summer, and on evergreen oak leaves during the winter. Nutritional deficiencies of evergreen oak were overcome by spraying freshly cut leaves with the water-soluble components of the diet of David & Gardiner (1966). To reduce the spread of infectious diseases, the third-instar animals were transferred to individual polystyrene containers in which a small piece of moistened ‘oasis’ was placed. Animals were fed daily until spinning started at the end of the fifth instar. Five days after spinning, pupae were removed from their cocoons to monitor their development, and were kept at 10-28°C, depending on when adults were required. Following eclosion, moths were kept in the dark and observed until they had mated. Eggs were surface-sterilized in 5% sodium hypochlorite solution in dilute detergent, rinsed in distilled water then allowed to develop at 18-28°C until hatching. Using these methods a continuous culture of Antheraea pernyi was maintained through-out the year.

The dorsal surface of fifth-instar Antheraea pernyi larvae was smeared with petroleum jelly, sharpened tungsten hooks were inserted into the head capsule and anal skin flap and the animal was chilled at 0°C until just immobilized (normally 8-12min). It was then suspended in a dissection chamber, a dorsal incision was made from the tail to the head capsule and the resulting flaps of skin were fixed onto swivelling pins at the side of the chamber. The gut and silk glands were ligatured and removed, care being taken not to disrupt the tracheal supply to the ganglion under study. Saline modified from Weevers (1966) (in mmol 1^-1: NaCl, 20; KC1, 30; MgCl₂, 20; CaCl₂, 3; sucrose, 160; NaH₂PO₄, 10-9; Na₂HPO₄, 4-1; pH 6-6) was then superfused through the body cavity at a rate of 1-2 ml min^-1. Abdominal ganglion 3, 4, 5 or 6 was then lifted onto a vinyl-covered stainless-steel platform and pinned out with small insect pins approximately 2 mm long. Up to 20 such pins were inserted into connective tissue and large tracheae to immobilize the ganglion for recording. The connective tissue sheath covering the ganglion was focally softened by applying a 1 % protease solution (Sigma Type VI) in a flat-bottomed pipette for 8-12 min. Microelectrodes pulled on a Brown & Flaming pipette puller (Sutter Instrument Co.), filled with 3 mol 1^-1 KC1 or 2 mol 1^-1 potassium acetate solution, with resistances of 20-50 MΩ were used to make intracellular recordings. Penetration of median cells was aided by the use of a piezoelectric prodder (Weevers, 1980) driven by square wave pulses from a stimulator. Signals from the microelectrode were passed through a home-built electrometer with internal bridge balance circuitry, displayed on an oscilloscope (Tektronix 5111 A) and simultaneously recorded on a Racal Store 4FM tape recorder. Intracellular recordings of muscle activity were made in a similar way except that electrodes were of lower resistance (1-10 MΩ) and were mounted on a flexible silver wire (Woodbury & Brady, 1956).

Extracellular recordings of neural activity were made en passant from nerve trunks with a chloride-coated silver wire hook electrode which could be retracted into a paraffin-filled glass sleeve. Recordings were made between the hook and an indifferent electrode placed in the body cavity via an Isleworth A103 pre-amplifier. Nerve trunks were stimulated by a fine pair of chloride-coated silver wires insulated except at their tips. Recordings were played into a Digitimer DS900 transient recorder and plotted on a Gould x,y plotter.

For intracellular dye filling, microelectrodes were filled with 100 mmol 1^-1 hexammine cobaltic chloride, 3% Lucifer Yellow (Stewart, 1981) or 4% horse-radish peroxidase (Boehringer Mannheim, grade 1) in 0·5 mol 1^-1 potassium acetate solution. Dye was injected into physiologically identified cells by passing 1-10nA constant-current pulses, 500ms long at 1Hz for 5-30min. Cobalt-filled cells were then treated with dilute ammonium sulphide solution (two drops in 10 ml of phosphate buffer, pH 7·0) for up to 5 min, rinsed in fresh phosphate buffer and fixed in 5 % formaldehyde in buffer for 30 min. After washing overnight in running tap water, ganglia were rinsed twice in distilled water and then intensified using the method of Bacon & Altman (1977). Ganglia were then dehydrated in ethanol, cleared in methyl benzoate and xylene, and mounted in neutral Canada balsam.

Cells filled with horseradish peroxidase were processed according to the method of Roberts & Clarke (1982). Lucifer Yellow fills were fixed in 5 % formaldehyde in 100 mmol 1^-1 phosphate buffer pH 7·0, dehydrated in ethanol and mounted in Fluoromount (Edwin Gurr Ltd).

The general neuroanatomy of fifth-instar caterpillars was revealed by intra vitam injection of 1-2 ml of 0·17 % Methylene Blue in saline, reduced with Rongalite (see Pantin, 1946, for exact details). After 15 min the animal was dissected in the normal way from the dorsal midline, either under phosphate buffer (pH 7·0) or under saturated ammonium molybdate solution. Six preparations were studied in detail in this way.

Of the many neurones with cell bodies in the midline of the abdominal ganglia of larval Antheraea pernyi, only two were recorded that had axons projecting bilaterally and symmetrically in both nerves 1 of the ganglion. These neurones were named median cells 1 and 2 (MCI and MC2), and were found to be uniquely identifiable. The anatomy of the segment in fifth-instar larvae is described first to provide a frame of reference for the description of the morphology of the two median cells.

Methylene Blue staining of nerve trunks in the abdomen revealed fine branches that were only seen with great difficulty in living material. Fig. 1 shows data collected from six preparations using this technique and subsequently confirmed in many in vivo preparations. The diagrams show the structure of nerves at four stages of dissection starting after the removal of the gut, silk glands, gonads and fat body. The four stages progress from the innermost layer of longitudinal muscles outwards to the oblique muscles directly below the cuticle. The nomenclature of nerve branches was based on the system of Pipa & Cook (1959) as it was felt that the system of Beckel (1958) was too inflexible to allow for the variability of branching seen in different preparations.

The named nerve trunks were all consistently identifiable in different animals; finer branching was extremely variable but that represented in Fig. 1 is typical. The pattern of branching was basically similar to that described in the earlier studies of Libby (1959) and Beckel (1958) (in Hyalophora cecropia) with the following exceptions. Beckel’s nerve All which was shown to run from Nlb3 near the spiracle and anastamose with N2d in the next anterior segment was not seen in Antheraea pernyi (see Fig. 1C). Libby (1959) also failed to record this branch. Second, the long branch of N2b (see Fig. 1C) was shown by Beckel as being distally attached to a branch of Nlc; this connection was never seen. Four anastamoses were seen; N1 joined mN both near the ganglion and near the spiracle. The small branch mNa (see Fig. 1A) which anastamosed with N2c of the next anterior ganglion was that shown by Taghert & Truman (1982), in Manduca sexta, to contain the axons of three identified bursicon-containing cells. The fourth anastamosis was again with mN where Nlb3 joined it just ventral to the spiracle.

The targets of each of the named branches are summarized in Table 1. Nerves Nlbl, Nlb2, Nlcl, Nlc2, Nlfl, Nlf2, Nlf3 and N2b were never seen to have any connections with muscle fibres and appeared to innervate either the cuticle or identified sense organs. Nerve trunks N1a1, Nla2, Nla3, Nlb3, Nid, Nle, Nig, Nlh and Nlj could only be traced to bundles of muscle fibres; no branches could be traced to run to the cuticle or to identified sense organs.

Motor units were characterized in the following way. By placing a fine stimulating electrode on a muscle fibre it was possible to stimulate a single motoneurone and all of its collaterals, and by using a movable intracellular recording electrode the extent of the motor unit could be defined. An extracellular recording electrode was placed on the proximal root of the nerve and used to check that only a single antidromic spike was elicited by stimulation.

The results obtained using these methods are shown in Fig. 2 (for the sake of simplicity the gut, silk glands, fat body, nervous system and tracheae have been omitted). The nomenclature used is that of Lyonet (1762) and Forbes (1914), except for muscles π and ψ, which were not identified by these authors. The shading of the muscles reflects the source of their innervation, the 19 stippled motor units were supplied by motoneurones with somata in the next anterior segmental ganglion via Nl. Four motor units (crosshatched) were supplied by motoneurones with cell bodies in the same segment’s ganglion and axons in Nl. Ten muscles (striped) were supplied by motoneurones with axons in N2. The spiracle opener (spo) is a non-muscular ligament (Beckel, 1958) and the alary muscle, heart and spiracle closer were innervated by the median nerve (mN).

Of all the muscles tested, only one (γ) was shown to be innervated by more than one axon. All fibres in this muscle showed two sizes of EJP in spontaneous activity and in response to graded stimulation of N2. All other motor units had only one size of spontaneous or evoked EJP. No evidence for inhibitory input was seen in any muscles in the abdomen of Antheraea pernyi larvae.

Two neurones, MCI and MC2, were recorded that had axons in both right and left nerves 1 of abdominal ganglia and were distinguishable from all other neurones by this criterion and by their characteristic intracellular physiology. It was of interest, therefore, to examine their central morphology to determine whether these two cells were uniquely identifiable, to see how they differed from one another and to compare their anatomy with that of DUM (dorsal unpaired median) cells in other insects. Intracellular dye fills of the two median cells were obtained on many occasions; typical examples are shown in Fig. 3. The most striking feature of the cell was the symmetry of its axons, dendrite distribution and branching at least to the level of secondary branching. The posteriorly sited soma gave rise to a single neurite that ran forward without branching quite deep within the neuropile. At a level just behind the roots of nerve 1 the neurite rose sharply, and bifurcated into two axons which ran over the dorsal surface of the neuropile and then out of right and left nerves 1 of the ganglion. Using the method of McKenzie & Vogt (1976) the vertical distribution of dendrites was mapped for four regions of the neuropile (see Fig. 4) from the fine focus control on a microscope. This indicated that dendritic branching was largely restricted to the dorsal, dorsomedial and dorsolateral regions of the neuropile.

Intracellular dye filling indicated that there were no significant differences between the central morphology of MCI and MC2; in fact, they could only be distinguished by their peripheral branching pattern. This is shown by the dye fills drawn in Fig. 3. Apart from characteristic differences in peripheral axon branching, the only difference between MCI and MC2 was that MCI tended to have the more dorsal cell body. However, this was not a reliable distinguishing characteristic. In 78 % of preparations where measurements were recorded (32 out of 41) MCI had its cell body within 50 μm of the dorsal ganglion surface (measured with the fine advance of the micromanipulator). MC2 had a dorsally located soma in only 38 % of preparations (10 out of 36). There was no significant tendency for one cell body to be more posteriorly located than the other.

Peripheral axons of the median cells were traced with three methods. First, antidromic impulses were evoked by stimulation of peripheral nerve trunks and were recorded in the cell soma; this indicated the presence of an axon branch at the point of stimulation. The paucity of synaptic input from nerve 1 made this method reliable and quick. Orthodromic tracing used the presence of extracellular potentials recorded in peripheral nerve trunks and time-locked to intracellularly recorded soma spikes. This method was slow, it often required signal averaging and could damage peripheral nerves. The third method was to inject dye into the soma of the median cell and let it diffuse into the axons. This method was rather capricious; however, it did give unequivocal confirmation of the basic branching patterns.

MCI was distinguished by the presence of an axon branch in Nlb3 whereas the other median cell (MC2) always had an axon branch in N1 beyond the spiracle. No neurone with branches in both of these nerves was ever seen. The branch of MCI in Nlb3 could usually be traced no further than the anastamosis with the median nerve in Antheraea pernyi. In a few preparations, axons of MCI could be traced a short distance down the nerve rami running from Nlb3 to muscles l,q,θ and the spiracle closer muscle, but in no case was extensive branching found in any of these muscles. Both MCI and MC2 sometimes had axons in the branch of Nla which innervated the ventral longitudinal and oblique muscles. In 53 % of preparations where detailed tracing was carried out (9 out of 17) MCI had axon branches traced into Nla; 79 % of MC2 cells had axons in Nla (11 out of 14 cells). The extent of this branching was highly variable but could never be traced beyond the major nerve trunks innervating blocks of muscle.

The distribution of axons was very variable. In summary, occasionally branches of either MCI or MC2 were found in every branch of Nl that innervated blocks of muscle but they were never found in any nerve trunks that were not seen to innervate muscle blocks (see Table 1). A significant detail of this tracing was that the latency of antidromic impulses often increased dramatically within a short distance along a nerve trunk innervating a block of muscle, presumably as the axon narrowed near its termination. Conduction velocity varied from 1ms^-1 in major nerve trunks (e.g. Nl) to 0·02ms^-1 in the ventral muscle block. The extent of median cell branching was very limited in comparison with that of motoneurones; within muscle blocks, branches were only detected in nerves supplying large bundles of fibres, never in the fine branches that ramified over single muscle fibres. Typical patterns of branching of MCI and MC2 from a single preparation are shown schematically in Fig. 5.

The apparent resting potential of the two median cells compared with the baseline potential when the microelectrode was outside the ganglion was -19·2 ±4·2 mV (N = 10). It is likely, however, that this does not represent the true difference in potential between the inside of the cell and the surrounding extracellular fluid. Whilst advancing the microelectrode through a ganglion, a complex series of potentials was often recorded before a median cell was penetrated, even when the cell was clearly visible as the most dorsal neural soma. Typically, a negative potential of up to -70 mV (probably as a perineurial glial cell was penetrated) was the first event recorded; this was typically followed by an apparent positive potential averaging +33·1 ±4·0 mV (N =10), which was pre-sumed to represent a transperineurial potential, as has been reported in the cockroach (Treherne & Schofield, 1981). Thus a better estimate of the real resting potential of the median cells is given by the difference between the apparent resting potential and the transperineurial potential, i.e. -52·3 ± 8·2 mV (N = 10).

One of the distinguishing characteristics of MCI and MC2 was the presence of long-duration overshooting action potentials recorded in the neural soma. Impulse duration averaged 21·9 ± 3·8 ms (N= 14) for MCI and 20·4 ± 7·2 ms (N = 12) for MC2 measured at half peak amplitude, and had a distinctive profile with a pronounced inflexion on the falling phase (see Fig. 6). The peak amplitude varied between preparations but usually lay between 70 and 90 mV above resting potential (average peak amplitude for MCI 83·6 ±13·8 mV, N=14; for MC2 74·8 ± 11·5 mV, N = 12).

Action potentials in the soma of the median cells were followed by a large afterhyperpolarization 10-20mV below resting potential (MCI, -15·0 ± 5·2 mV, N = 14; MC2, -14·3 ±4·6 mV, N = 12). The durations of the initial hyperpolariz-ations were (average for MCI) 350 ± 89 ms (N = 14) and (for MC2) 289 ± 85 ms (N = 12), measured at half peak amplitude. Following the initial hyperpolarization was an even longer duration afterpotential (LAH) (see Fig. 6C,D) that lasted up to 20 s after each impulse and was up to 8 mV in peak amplitude below resting potential [average for MCI —4·8±l·7mV (N =14), for MC2 — 4·l±l·7mV (N = 12)]. This potential was visible in about 80% of median cells recorded and had a strong depressive effect on the spontaneous firing rate. In some median cells there was a measurable increase in cell conductance during the course of LAH. It was apparent that the resting potential of the median cells was usually poised very close to the threshold for spike initiation, since injection of even small amounts of depolarizing current evoked impulses. However, during the course of LAH, the current threshold for production of impulses increased (see Fig. 7). This indicates that the effect of even small hyperpolarizations of the median cell membrane on the spontaneous firing rate of these neurones could be quite profound. The appearance of the long afterhyperpolarization depended on the presence of a preceding soma spike; in Fig. 8, antidromic action potentials elicited soma spikes that were followed by an LAH; this was unaffected by a short hyperpolarizing current pulse injected through the recording electrode into the median cell. However, when the hyperpolarizing current pulse was just sufficiently strong to block the initiation of a soma spike, no long afterhyperpolarization was evoked.

Spontaneous synaptic activity was seen in all recordings from median cells where the ventral nerve cord was intact. These spontaneous synaptic potentials were all depolarizing and varied in size from 1 to 5 mV in different preparations (average size 2·9 ±1·0mV for 21 preparations). Their frequency also varied considerably, from 5-20 distinguishable events per second were seen in immobile preparations; however, this increased during spontaneous muscular activity.

Paired recordings from MCI and MC2 in the same ganglion revealed that the two cells showed a similar temporal pattern of firing, although the spikes did not correlate one-for-one and the basal rate of firing often differed between the two neurones (see Fig. 9A). At higher gain it was clear that MCI and MC2 shared much common input; in fact, synaptic events correlated exactly one-for-one in the two cells in all recordings (Fig. 9B). No evidence was found for direct connections between the two median cells in the same ganglion; impulses initiated by intracellular current pulses in one cell evoked no time-locked synaptic potentials in the other cell even when the putative postsynaptic cell was systematically depolarized and hyperpolarized. In view of the slowness of the peripheral effects of these neurones (Brookes, 1988), similarly slow modulatory influences were looked for; in spite of this no evidence for any sort of interaction was seen on membrane potential, input impedance or firing rate.

Paired recordings of median cells in different ganglia also showed identical common synaptic input, with events in one cell being followed one-for-one after a short delay by events in the more posterior cell (Fig. 10B). The delay between synaptic events recorded in different ganglia indicated that the presynaptic neurones conducted impulses from the anterior to the posterior of the animal at a conduction velocity that varied from 0·2 to 1·0ms^-1 in different preparations.

The source of the tonic synaptic input was investigated by selective lesions of the nervous system in a preparation where MCI was simultaneously recorded in ganglia 4 and 6 (Fig. 10A,B). The two cells received identical synaptic input when the two connectives were intact. However, severing the left connective between the ganglia led to a disruption of the one-to-one correspondence of synaptic events (Fig. 10C), not all EPSPs in the anterior cell were followed by an EPSP in the more posterior neurone. Cutting the left connective anterior to ganglion 4 led to a resumption of one-to-one correspondence (Fig. 10D). Cutting the right connective between the ganglia led to a virtual silencing of spontaneous synaptic activity in the median cell in ganglion 6 (Fig. 10E). In fact, cutting both connectives at any point posterior to the suboesophageal ganglion along the ventral nerve cord always abolished spontaneous synaptic activity in all median cells posterior to the lesion. Lesion of the circumoesophageal connectives did not significantly affect the appearance of spontaneous EPSPs. These results taken together indicate that the excitatory drive for the median cells arose predominantly in the suboesophageal ganglion, that the axons conveying it descended the length of the nerve cord, that it was not completely synchronized in the two sides and that it did not cross from one side to the other within the abdominal ganglia. Electrical stimulation of the cut ends of connectives evoked synaptic activity similar to the spontaneous activity seen in preparations with an intact nerve cord. Interestingly, stimulation of the peripheral nerves 1 and 2 also evoked EPSPs in most preparations, indicating that there were inputs other than those descending from the suboesophageal ganglion.

The muscles of several families of lepidopteran larvae have been described elsewhere (Cossidae, Lyonet, 1762; Notodontidae, Lubbock, 1859; Lasiocampi-dae, Noctuidae and Sphingidae, Forbes, 1914; Saturniidae, Beckel, 1958; Libby, 1959; Galleria mellonella, Randall, 1968; Manduca sexta, Rheuben & Kammer, 1980). However, the description given here is based on both anatomical and physiological evidence and thus describes for the first time the muscular anatomy in terms of motor units. The most striking feature of the segmental anatomy was the extreme simplicity of the innervation and musculature. The association of MCI and MC2 axons with nerve trunks that innervated bundles of muscle fibres rather than with nerve trunks running to identified sense organs or to the cuticle was very striking.

In none of the Methylene Blue or in vivo preparations were peripheral nerve cell bodies seen on segmental nerves 1 and 2, although such neurones have been described in the stick insect and in the blowfly larva (Finlayson & Osborne, 1968), and in the median nerve of three families of lepidopteran larvae (Hinks, 1975; Griffiths & Finlayson, 1982). In the noctuid, arctiid and sphingid caterpillars examined by these authors, peripheral neural somata were confined to the distal end of the median nerve near the alary muscles, and were never seen on branches supplying somatic muscles. Since axons of the median cells were never traced to the distal median nerve it is unlikely that they synapse onto or interact directly with any peripheral neurosecretory cells.

The consistent central anatomy of the median cells revealed by intracellular dye filling in the present study indicates that these cells are uniquely identifiable. They probably correspond to the bilaterally symmetrical neurones described by Taylor & Truman (1974) in abdominal ganglion 4 of larval Manduca sexta. MCI and MC2 have also been recorded and dye-filled in abdominal segments 3, 4, 5, and 6 in A. pernyi.

MCI and MC2 share a number of features with DUM cells that have been studied extensively in locusts and cockroaches. First, they share the characteristic bilateral symmetry of their branching (Crossman et al. 1971). Second, they have action potentials of relatively long duration that invade their cell bodies. Third, they have axons that can be traced into nerves supplying blocks of muscle, a feature of many of the unpaired neurones stained by backfilling from peripheral nerve trunks in cockroach and locust ganglia (Hoyle, 1978), in a cricket (Davis & Alanis, 1979), in a hemipteran (Davis, 1977), in fireflies (Christensen & Carlson, 1981, 1982) and in silkmoth larvae and adults (Taylor & Truman, 1974; Casaday & Camhi, 1976; Rheuben & Kammer, 1980; Kondoh & Obara, 1982). Lastly, MCI and MC2 have been shown to have modulatory effects on a number of somatic muscles (Brookes, 1988) similar to those described for DUMETi in the locust (Evans & O’Shea, 1977; O’Shea & Evans, 1979). However, there are also a number of differences between MCI and MC2 in Antheraea pernyi larvae and DUM cells in locusts and cockroaches. Both cells in this study had extensive peripheral branching. Several DUM cells with exclusively central branching have been described (Crossman et al. 1971), but no such cells were recorded in A. pernyi larvae. The long afterhyperpolarization seen after impulses in MCI and MC2 has not been reported in DUM cells in other insects. The one-for-one correspondence of synaptic events in MCI and MC2 was not reported for DUM cells in the locust metathoracic ganglion, although common synaptic inputs were seen (Hoyle & Dagan, 1978).

MCI and MC2 both had overshooting soma action potentials which distinguished them from the majority of insect neurones. It appears that the biophysical basis of the difference between spiking and non-spiking neural somata in insects is due to the presence of a stronger repolarizing outward potassium current in non-spiking somata which prevents the initiation of overshooting action potentials (Goodman & Heitler, 1979; Miyazaki, 1980). Goodman & Heitler (1979) speculated that DUMETi had a reduced outward potassium current as a mechanism for prolonging the axon spike, thus potentiating transmitter release. Such a mechanism might be expected to cause increased soma excitability. They support their reasoning with the observation that the soma spike of DUMETi is unlikely to play an important physiological role as it often failed when recordings were made at 32°C, the temperature at which the locusts were raised.

Although the suggestions of Goodman & Heitler (1979) are attractive for explaining the presence of a soma spike in DUMETi, it has been shown that the soma spike of MCI and MC2 correlates with the appearance of a long afterhyperpolarizing potential (LAH) that strongly limits their excitability. The effect of this potential was to maintain a low basal rate of discharge which could be overridden by increased descending synaptic drive during periods of excitability. As such, it would appear that the soma spikes of MCI and MC2 and the associated LAHs probably play a significant role in determining the firing of the cells. Another possible role for the soma spikes in these cells could be to coordinate impulse initiation in the two axons. During prolonged periods of recording, soma spikes always evoked impulses in axons that could be detected peripherally. Consideration of the symmetry of the cells suggests that there may be two spike initiation sites, as has been reported for DUMETi in the locust (Heitler & Goodman, 1978); in spite of this, axon spikes were extremely rarely seen during intracellular recording. Thus, bilateral coordination of firing in the axons may be another role for the soma spike. This is of obvious adaptive significance considering the widespread peripheral effects of the median cells (see Brookes, 1988).

The suboesophageal ganglion has classically been associated with the initiation and maintenance of movement in insects (see, for example, Roeder, 1963; Kien, 1983). In the light of the involvement of MCI and MC2 with the modulatory control of somatic muscles (see Brookes, 1988) it is perhaps not surprising that much of their synaptic input should arise in this ganglion. The sharing of synaptic input between median cells in different ganglia is likely to have the effect of ensuring that the entire somatic musculature supplied by these cells receives a similar pattern of modulatory input.