Coordination of Locomotor and Cardiorespiratory Networks of Lymnaea Stagnalis by A Pair of Identified Interneurones

Many of the complex types of behaviour exhibited by animals are a result of coordination between two or more motor centres of the central nervous system. Numerous examples are available where such coordination has been demonstrated in both vertebrates (Viala, 1986; Feldman and Ellenberger, 1988; Cohen, 1987) and invertebrates (Hughes and Wiersma, 1960; Ikeda and Wiersma, 1964; MacMillan et al. 1983). However, few studies have uncovered the circuitry underlying this coordination, and even fewer have identified specific coordinating neurones (see Bush and Clarac, 1985). Many researchers are currently interested in determining the degree and importance of neural integration between various aspects of behaviours. Some of the questions that need to be answered in this area concern the possible existence of multifunctional neurones (Delcomyn, 1987; Ritzmann et al. 1980; Atwood and Wiersma, 1967; Bernard et al. 1989; Hooper and Moulins, 1989). For example, do those neurones that coordinate various neural elements of a given behaviour also play an integral role in coordinating seemingly unrelated behaviours, or are coordinating neurones specified for each behaviour? To answer such questions, many researchers have turned to examining the behaviours exhibited by invertebrates, whose ‘simple’ nervous systems are often very amenable to neurophysiological studies.

In recent years, the freshwater pulmonate snail Lymnaea stagnalis (L.) has become a popular subject for studies of such simple behaviour patterns as feeding (Egelhaaf and Benjamin, 1983; Elliott and Benjamin, 1985; McCrohan, 1984; Kyriakides and McCrohan, 1988), reproduction (see Boer et al. 1987), whole-body withdrawal (Ferguson, 1984; Benjamin et al. 1985), locomotion (Syed, 1988; Syed et al. 1988) and respiration (van der Wilt et al. 1988; Syed et al. 1990). Although in most of these studies the pertinent effector organs and central neurones involved in the specific behaviour have been identified, very little is known regarding the coordination (e.g. spatiotemporal relationships) between these various behaviours, an exception being the work of Kyriakides and McCrohan (1988). The functional integration of the behaviourial repertoire of an animal is a determining factor in survival, particularly in the face of environmental stress. For example, if Lymnaea is subjected to a noxious stimulus during normal activity, behaviour such as respiration or locomotion is terminated immediately and the animal withdraws its entire head-foot complex into the shell. This fixed action pattern is called whole-body withdrawal (Ferguson, 1984). To satisfy its respiratory needs, however, a pulmonate snail must expose its pneumostome (respiratory orifice) to the air. In Lymnaea, pneumostome opening and closing movements can only be achieved if the body is extended out of the shell, i.e. a behaviour that is in opposition to withdrawal. During respiration, locomotor behaviour is also suppressed. Furthermore, analysis of locomotion in Lymnaea has revealed that well-coordinated movements of the left and right body wall and foot musculature occur in the absence of whole-body withdrawal or respiratory activity (Syed, 1988). Thus, a significant amount of coordination must exist among these three seemingly distinct behaviours. Here we present evidence that a pair of bilateral, electrically coupled interneurones coordinate the activities of locomotor and whole-body withdrawal motoneurones in Lymnaea. Furthermore, these two interneurones, identified as left and right pedal dorsal 11 (L/RPeDll), make chemical connections with heart motoneurones (Benjamin et al. 1988) and with neurones that participate in respiratory behaviour (Syed and Winlow, 1988a,b; N. I. Syed, D. Harrison and W. Winlow, in preparation).

Originally, the left and right body wall motoneurones of Lymnaea, identified as the left and right cerebral A cluster neurones (L/RCeA cluster), were found to be electrically coupled with their contralateral as well as their ipsilateral homologues (Haydon, 1982; Ferguson, 1984). However, the basis for the coupling between contralateral clusters was puzzling since CeA neurones have only ipsilateral projections. Here we present data demonstrating that L/RPeDll coordinate the activities of L/RCeA cluster neurones via electrotonic coupling. The electrical coupling between LCeA cluster and RCeA cluster neurones was lost after selective ablation of L/RPeDll. However, the L/RCeA cluster neurones remained coupled with their ipsilateral (i.e. neighbouring) cells. We also provide evidence that although L/RPeDll participate in whole-body withdrawal behaviour, they do not induce this behaviour when stimulated electrically. Furthermore, we demonstrate that the interneurone visceral dorsal 4 (VD4, Janse et al. 1985) (=VWI of Benjamin, 1984), which controls the inspiratory phase of the respiratory behaviour (Syed and Winlow, 1988b; Syed et al. 1990), has inhibitory effects on foot and body wall motoneurones as well as on L/RPeDll. These studies therefore suggest that L/RPeDll coordinate the activities of locomotor motoneurones which innervate the left and right sides of the body, and that they are also multifunctional interneurones, acting on neuronal networks that control the cardiorespiratory system of Lymnaea.

Specimens of Lymnaea stagnalis (L.) were usually obtained from animal suppliers and occasionally collected from the Leeds-Liverpool canal. These snails were maintained at 10–16 °C in aerated pond water obtained locally and fed on lettuce, supplemented with tropical fish food. All experiments were performed on 1–4 g snails bathed in standard snail saline (Benjamin and Winlow, 1981). Isolated brains of Lymnaea were prepared for electrophysiological and morphological studies and maintained in snail saline buffered to pH7.8–7.9 as previously described (Benjamin and Winlow, 1981). Several modified salines were also used. For zero-Ca²⁺/high-Mg²⁺ saline, calcium in the normal saline was replaced by magnesium, the latter being raised from 2 to 6 mmol l⁻¹. For high-Ca²⁺/high-Mg²⁺ saline, concentrations of both divalent cations were raised sixfold: Ca²⁺ to 24 mmol l⁻¹, Mg²⁺ to 12 mmol l⁻¹. Individual neurones were impaled with glass microelectrodes (10-20 MQ) and recordings of neuronal activity were obtained using conventional techniques. Electrophysiological signals were amplified, displayed and recorded by conventional means (Benjamin and Winlow, 1981).

Lucifer Yellow (CH) stains of the neurones were prepared according to the methods of Syed and Winlow (1989). Briefly, microelectrodes were filled with 10% Lucifer Yellow dissolved in double distilled and deionized water. These dye-filled electrodes had a tip resistance of 20–80MΩ. Prior to the impalement and withdrawal of electrodes from the neurones a constant holding current of +2 nA was applied to prevent the leakage of the dye from the tip of the electrode. Upon successful penetration of the cells, this current was switched off. The Lucifer Yellow was then injected into the soma by applying a constant −2 nA current for 10–20 min. These preparations were left overnight to allow the spread of dye and were fixed in buffered formalin (formaldehyde, 4% in 0.1 moll⁻¹ sodium phosphate buffer to pH7.4) for 3h. The fixed brains were dehydrated using a series of ascending concentrations of ethanol, then defatted and cleared by incubation in dimethyl sulphoxide and methyl salicylate. Cleared tissues were mounted in FluoroSave (Calbiochem) on depression slides and observed using a Leitz Dialux 20EB incident fluorescence microscope. Successfully injected neurones were photographed using 400 ASA Kodak Ektachrome slide film and drawn using a camera lucida.

The effects of sensory inputs on the interneurones were tested using a newly developed semi-intact preparation (Syed, 1988; N. I. Syed, D. Harrison and W. Winlow, in preparation), as were the effects of the interneurones on whole-body withdrawal behaviour. Briefly, animals were anaesthetized by bubbling 2% halothane into the pond water as described by Girdlestone (1986). Anaesthetized animals were transferred to a chamber designed for these experiments (Syed, 1988). Fine nickel wires were attached to the animals at various points using tissue glue (cyanoacrylate adhesive). All these wires were extended and attached to the wall of the chamber by Plasticine, allowing the animal to be suspended in the saline in a manner similar to the Tritonia preparation described by Willows et al. (1973). Once firmly suspended, a small mid-dorsal body incision was made and the body wall was gently retracted with blunt hooks attached to the nickel wires. The brain was lifted by inserting a wax-covered spatula held on a micromanipulator (Syed, 1988). One end of a cotton thread was attached to the tension transducer and the other to the foot or body wall musculature of the animal. Upon completion of surgery, the animals were allowed to recover from anaesthesia. After several washes in normal saline, intracellular recordings were made from central neurones and muscle tension was recorded using tension transducers.

Selective photoinactivation of neurones was carried out using previously developed methods (Miller and Selverston, 1979; Bulloch and Kater, 1982; Bulloch et al. 1984; Elliott and Kleindienst, 1990). Briefly, the cells were first filled with 10% Lucifer Yellow and then, using a portable HBO 100 W d.c. mercury arc lamp and a light guide, exposed to a beam of high-intensity blue light. We found that these Lucifer-Yellow-filled neurones are killed within seconds if they are injected with depolarizing current during the period of blue light exposure. The soma and its main processes began to disintegrate, resting membrane potential was lost and cells became inexcitable (see Results).

The paired interneurones L/RPeDll are located on the dorsal surface of the left and right pedal ganglia, respectively. These neurones are whitish-orange in colour, have soma diameters of 30–50 μm, and are situated at the pedal end of the pedal-pleural connectives, anterior to the statocyst (Fig. 1). They are the largest cells located within the pedal G clusters (Slade et al. 1981). Ionophoretic injections of Lucifer Yellow into neurones L/RPeDll reveal that they have similar morphologies and that together their processes encircle the lower ganglionic ring (i.e. pedal, pleural, parietal and visceral ganglia) and enter the cerebral ganglia (Fig. 2). Neurone LPeDll has extensive neuropilar arborizations in all the above ganglia and has two main axons, one running ipsilaterally and the other contralaterally (Fig. 2A). Simultaneous injection of Lucifer Yellow into both left and right PeDll (Fig. 2B) revealed that, although these neurones do not have main axons in the superior, median or inferior pedal nerves (see Slade et al. 1981), a small axonal branch of each cell does project towards (but does not enter) the superior pedal nerve.

Motoneurones to the foot lie within the pedal G clusters (L/RPeG cluster) (McCrohan and Winlow, 1985; Winlow and Haydon, 1986; Haydon and Winlow, 1986). Cerebral A cluster (CeA) neurones innervate the body wall musculature (Haydon, 1982) and are termed whole-body withdrawal motoneurones (Ferguson, 1984; Benjamin et al. 1985). The CeA cluster neurones are a group of 20–25 orange-coloured cells (30–40 μm soma diameter) located in each cerebral ganglion (Fig. 1). They have axons in the (ipsilateral) superior and inferior cervical nerves of the pedal ganglia and extensive branches in ipsilateral cerebral, pleural and pedal ganglia (Fig. 3A.B).

Functional electrical connections are present between neurones L/RPeDll and PeG clusters (Fig. 4A,B). Most ipsilateral CeA cluster neurones are known to be electrically coupled to each other (Haydon, 1982; Ferguson, 1984; Benjamin et al. 1985). Furthermore, L/RCeA cluster neurones are also known to be electrically coupled to their contralateral homologues (Ferguson, 1984). We found that interneurones L/RPeDll were also electrically coupled to L/RCeA cluster peurones (Fig. 4C). In the present study, to rule out the possibility of chemical synaptic transmission, all experiments where the presence of electrical coupling is demonstrated were performed in zero-Ca²⁺/high-Mg²⁺ salines.

As described above, L/RCeA cluster neurones have only ipsilateral projections, yet they were found to be electrically coupled to their contralateral homologues. Interneurones L/RPeDll, in contrast, not only have ipsilateral and contralateral projections but are also electrically coupled to both L/RCeA cluster neurones. To test whether the electrical coupling between L/RCeA cluster neurones was via these interneurones, we selectively ablated either LPeDll or RPeDll by intracellular injection of 10% Lucifer Yellow followed by exposure of the preparations to blue light for 5–10 min (Bulloch et al. 1984; Elliott and Kleindienst, 1990). The effectiveness of this photoablation was checked in two ways: (1) visually by observation of the immediate disintegration and fragmentation of the cell body, and (2) electrophysiologically, by the loss of electrical activity and resting membrane potential (Fig. 5). Normally, the loss of resting membrane potential occurred in 5–10 min, but this could be achieved immediately by depolarising the injected cell (Fig. 5). When either left or right PeDll was selectively killed by this procedure, the electrical coupling between L/RCeA cluster neurones persisted (Fig. 6A). However, when neurones L/RPeDll were both photo-inactivated, the electrical coupling between L/RCeA cluster neurones was lost (Fig. 6B). Cerebral A cluster neurones within the ipsilateral ganglion, however, remained coupled even after the removal of both interneurones from the circuit (Fig. 6C). These data suggest that the activities of the contralateral cerebral A cluster neurones are coordinated via neurones L/RPeDll.

In addition to their electrical connections, interneurones L/RPeDll also make chemical synapses with a wide variety of neurones. The giant cells visceral ventral 1 and 2 (Wl, W2), (Benjamin and Winlow, 1981) were found to have axonal branches in various visceral and parietal nerves. In addition, they were found to project to the periphery via the pedal nerves, which innervate foot and body wall musculature (Fig. 7A). Some of the visceral F group neurones also had axon projections in various pedal nerves (Fig. 7B) and received inputs in common with locomotor motoneurones (N. I. Syed, unpublished data). The function of Wl, W2 and visceral F group neurones is unknown, but electrical stimulation or spontaneous action potentials in either left or right PeDll produced 1:1 inhibitory postsynaptic potentials (IPSPs) in Wl and W2 (not shown here) and excitatory postsynaptic potentials (EPSPs) in VF cells (Fig. 8A,B). These connections between L/RPeDll and VVl and VV2 and VF group neurones were significantly blocked when bathing solution Ca²⁺ was replaced with Mg²⁺ (Fig. 8A,B), providing evidence for the chemical nature of these connections. Evidence that these connections might be monosynaptic is also provided in Fig. 8, where these connections are shown to persist in high-Ca²⁺/high-Mg²⁺ salines.

Intracellular stimulation of L/RPeDll also produced EPSPs of constant latency in all visceral E (VE) group neurones examined, including two cells that are electrically coupled to each other and that have been described as heart motoneurones (Benjamin et al. 1988) (Fig. 8C). The connections between VE group neurones and L/RPeDll were also significantly reduced by replacing the Ca²⁺ in normal saline with other divalent cations, such as Mg²⁺ (Fig. 8C) or Co²⁺. Furthermore, the connections between PeDll and VE group neurones persist in high-Ca²⁺/high-Mg²⁺ saline (Fig. 8C), suggesting the possibility of a monosynaptic pathway.

It has recently been shown that a visceral J cell (VJ cell) is a pneumostome opener muscle motoneurone, while a visceral K cell (VK cell) is a pneumostome closer muscle motoneurone (N. I. Syed, D. Harrison and W. Winlow, in preparation). Electrical stimulation of RPeDll produced IPSPs in the VJ cell and EPSPs in the VK cell (Fig. 9A,B). Similarly, stimulation of neurones L/RPeDll produced EPSPs in VG group neurones (Fig. 9C), a group that receives inhibitory inputs during pneumostome opening (Syed, 1988), but whose exact function is not yet known. These connections were also blocked in zero-Ca²⁺/high-Mg²⁺ saline but not in high-Ca²⁺/high-Mg²⁺ saline (Fig. 9A,B and C).

Several right parietal A group neurones (RPA group) (Benjamin and Winlow, 1981) have been found to be motoneurones to the mantle cavity musculature (N. I. Syed, D. Harrison and W. Winlow, in preparation). Neurones L/RPeDll were found to have electrical coupling with two of these RPA group neurones (Fig. 10A). Another pair of electrically coupled neurones present in the visceral and parietal ganglia are visceral dorsal 1 (VD1) and right parietal dorsal 2 (RPD2) (Boer et al. 1979; Benjamin and Winlow, 1981; Benjamin and Pilkington, 1986). Electrical stimulation of RPeDll had excitatory effects on both VD1 and RPD2 (Fig, 10A). The RPA group neurone, which was found to be electrically coupled to RPeDll, had no effect on VD1 and RPD2 (Fig. 10B). All the connections found between neurones L/RPeDll and the respiratory neurones were reversibly blocked when the Ca²⁺ in the normal saline was replaced with Mg²⁺ (Fig. 11) or Co²⁺. These divalent cation manipulations did not affect the electrical coupling between RPeDll and RPA group neurones (Fig. 11).

In Lymnaea, during respiratory movements the locomotor and whole-body withdrawal motor activity must be inhibited (N. I. Syed, D. Harrison and W. Winlow, in preparation). Identified interneurone VD4 (Janse et al. 1985) (VWI of Benjamin, 1984) is implicated in respiratory behaviour (Syed and Winlow, 1988a,b; Syed et al. 1990; N. I. Syed and W. Winlow, in preparation). Stimulation of VD4 inhibited the electrically coupled interneurones L/RPeDll and also the motoneurones coupled to them, i.e. L/RPeG and L/RCeA cluster neurones (Fig. 12). The effects of VD4 on most follower cells are slow and difficult to resolve as unitary 1:1 responses. Both the electrical and chemical connections of L/RPeDll with their follower cells described in the present study are summarized in Fig. 13.

Previously, L/RCeA cluster neurones of Lymnaea were described as wholebody withdrawal motoneurones (Benjamin et al. 1985). In the present study, however, these CeA cluster neurones were found to be electrically coupled to interneurones L/RPeDll. We therefore investigated the role of these interneurones in whole-body withdrawal behaviour. To test the relationship between the withdrawal behaviour and the activity in L/RPeDll, photic or mechanical stimuli, which induced withdrawal behaviour, were applied to semi-intact preparations, either by illumination of the head-foot complex or by pressure application with a blunt glass rod held on a micromanipulator. The results obtained from these experiments showed that although L/RPeDll received excitatory inputs during induced withdrawal behaviour, strong electrical stimulation of these cells did not cause whole-body withdrawal (Fig. 14), as was suggested for the electrically coupled L/RCeA cluster neurones by Benjamin et al. (1985).

Most rhythmic behaviour, such as locomotion and respiration, requires coordination between various parts of the body so that movements or postures achieved are effective and useful. Probably the most effective and prompt way to coordinate various neural elements is via electrotonic coupling. The propagation of action potentials in these electrically coupled systems is faster than chemical transmission and, therefore, allows the animals to respond rapidly during a stereotyped behaviour (Marder, 1984). In the invertebrates, a number of preparations have been observed where extensive electrotonic coupling among neurones comprising a rhythm generator forms a positive feedback loop. This feedback, in turn, causes a synchronised discharge of the interconnected neurones (Farmer, 1970; Getting, 1974; Kaneko et al. 1978; Friesen, 1985; Nusbaum et al. 1987; Syed et al. 1988; Koester, 1989).

Such synchrony becomes more important if the survival of the animal is at stake. When confronted with a noxious stimulus, Lymnaea retracts its entire head-foot complex into the shell; a behaviour that can be considered an escape response This is achieved by the abrupt cessation of locomotion and respiratory activity, followed by a simultaneous contraction of body wall and foot muscles. Therefore, from a hierarchical point of view, the whole-body withdrawal behaviour takes priority over other behaviours, such as locomotion or respiration. Since the effector organs involved in these different behaviours are the same, an interaction and coordination between neural elements controlling these muscles must exist.

In Lymnaea we have identified a pair of electrically coupled interneurones that are also electrically coupled to foot and body wall motoneurones, thus forming an integrated network that may serve to coordinate locomotory and respiratory motor activities. In addition, these interneurones modulate cardiorespiratory activity via chemical connections with appropriate motoneurones. We believe that the electrical coupling between homologous contralateral groups of motoneurones is maintained via L/RPeDll because the selective ablation of both interneurones results in the loss of connectivity between the motoneurones. Such a loss of connectivity is not without physiological significance, for it would allow the two sides of the body to function independently. Dysynchrony between left and right sides of the body is evident in turning and twisting movements of Lymnaea, although no direct electrophysiological studies have been carried out to show decoupling of motor centres. In an electrically coupled network, such as that described in the present study, the decoupling between L/RCeA cluster neurones could be efficiently achieved through modulating the activity of interneurones L/RPeDll. A mechanism (e.g. inhibitory synaptic input) that would selectively inactivate these two interneurones would allow the contralateral CeA clusters to act independently.

As an electrically coupled network, these Lymnaea neurones have features in common with the neural network that underlies the escape behaviour of crayfish and hermit crabs. In crayfish, giant fibres cause synchronous excitation of the abdominal flexor motoneurones via electrotonic coupling (See Wine and Krasne, 1982). Similarly, in the hermit crab, each giant fibre drives the ipsilateral segmental giant neurones through electrical coupling (Heitler and Fraser, 1987). In Lymnaea, another pair of electrically coupled interneurones, the cerebral giant cells (CGCs) has been shown to coordinate the buccal motor output underlying rhythmic feeding behaviour. However, the connections between CGCs and buccal interneurones and motoneurones are known to be chemical in nature (see McCrohan and Winlow, 1985). Similar CGC and buccal neurone circuitry has been described for other gastropods, such as Planorbis corneus (Berry and Pentreath, 1976), Aplysia californica (Weiss et al. 1978) and Philine aperta (Barber, 1983). In Helisoma trivolvis, which is closely related to Planorbis and Lymnaea, the CGCs are not electrically coupled to each other (Granzow and Kater, 1977), but other methods of coordination between these CGCs (e.g. via other interneurones) may exist. Recently, Kyriakides and McCrohan (1988) have suggested that, in Lymnaea, coordination of the buccal feeding rhythm with foot and body wall movements may occur through inputs originating from buccal ganglia interneurones. Coordination of various behaviours may, therefore, involve a number of interneurones distributed among the different central ganglia to form a complex higher-order network. Modulation of the synaptic connections within such a network would allow certain behaviours to have priority over others or allow switching between behavioural states (Hooper and Moulins, 1989; Harris-Warrick and Johnson, 1989; DiCaprio, 1990; see also Selverston, 1989).

Interneurones L/RPeDll are not only electrically coupled to the withdrawal motoneurones, but also make chemical connections with cardiorespiratory motoneurones. The VJ and VK cells of the visceral ganglion have been shown to be the motoneurones to the pneumostome opener muscle and pneumostome closer muscle, respectively (N. I. Syed, D. Harrison and W. Winlow, in preparation). These motoneurones and others receiving common synaptic inputs (e.g. VG cells) are driven by respiratory interneurones and fire alternating bursts of action potentials during spontaneously occurring respiratory behaviour (N. I. Syed, D. Harrison and W. Winlow, in preparation). The electrical stimulation of L/RPeDll caused the inhibition of VJ cells while exciting VK and VG cells. These findings suggest that when L/RPeDll are spontaneously active, such as during locomotion or withdrawal behaviour, they should have an inhibitory effect on respiratory motor output, as is found during normal behaviour. In addition to these chemical connections, interneurones L/RPeDll make electrical connections with mantle cavity muscle motoneurones (RPA group neurones), which are also involved in the respiratory behaviour (N. I. Syed, D. Harrison and W. Winlow, in preparation).

In locusts, a pair of interneurones that makes simultaneous synaptic contact with flight and respiratory motoneurones has previously been described by Burrows, (1975a,b, 1982). These interneurones not only make extensive chemical connections with 30 flight motoneurones but also synapse upon 20 ventilatory motoneurones (Burrows, 1975a,b, 1982). These locust interneurones have reciprocal effects on antagonistic ventilatory motoneurones, depolarizing those that spike during expiration and hyperpolarizing those that spike during inspiration (Burrows, 1975b). In addition to these similarities with the locust intemeuronal network, L/RPeDll of Lymnaea excite heart motoneurones (Benjamin et al. 1988). This excitation could serve to increase the cardiac output, particularly when there is an increased demand for blood supply during locomotion. Another pair of electrically coupled neurones of Lymnaea, VD1 and RPD2, receives inputs during respiratory behaviour (Syed and Winlow, 1988a,b; N. I. Syed and W. Winlow, in preparation). Both VD1 and RPD2 are sensitive to changes in external (Janse et al. 1985; van der Wilt et al. 1988) and are follower cells of L/RPeDll. The actions of L/RPeDll on VD1 and RPD2 are exclusive and cannot be induced via stimulation of other electrically coupled cells (e.g. RPA group neurones). The interneurones L/RPeDll also make chemical connections with several other neurones (e.g W1 and VF group neurones), which receive inputs in common with locomotor and respiratory motoneurones (Syed, 1988), but whose exact functions remain to be determined. All these results support a multifunctional role for interneurones L/RPeD 11 in integrating cardiorespiratory output with other motor behaviours.

Recent investigations of the whole-body withdrawal system of Lymnaea provided evidence for motoneuronal function, and L/RCeA cluster neurones were described as the largest group of withdrawal motoneurones (Ferguson, 1984; Benjamin et al. 1985). According to Haydon (1982) and Winlow and Haydon (1986), the L/RCeA cluster neurones are left and right body wall motoneurones. Using our newly developed semi-intact preparation, we demonstrated that L/RCeA cluster neurones and cells coupled to them are spontaneously active during locomotion. These electrically coupled neurones fire synchronous discharges of action potentials during terrestrial locomotion (Syed, 1988). In experiments described here, we have demonstrated that the induction of wholebody withdrawal behaviour via mechanical stimulation of the head-foot complex excites previously quiescent L/RPeDll (Fig. 13) and other motoneurones coupled to them (not shown here). Nevertheless, strong electrical stimulation of L/RPeDll neurones did not induce whole-body withdrawal. We believe that, since the earlier experiments were carried out on severely restrained or extensively dissected animals, it is likely that the presence of motor activity in appropriate muscles and motor nerves was taken as an indication that the animal was engaged in withdrawal behaviour. From experiments described here it is apparent that, although L/RPeDll do participate in whole-body withdrawal behaviour, their prime function appears to be the coordination of foot and body wall musculature, as observed during locomotion (Syed, 1988). Since whole-body withdrawal in Lymnaea represents a form of escape behaviour, it is not surprising that some elements of the neural circuitry resemble those involved in escape responses of other animals. As mentioned before, electrotonic coupling plays an important role in the escape tail-flip of various crustaceans (Wine and Krasne, 1982; Heitler and fraser, 1987). Intemeuronal networks are involved in most escape responses, including escape swimming in leeches (Stent et al. 1978; Stent and Kristan, 1981) and molluscs, such as Tritonia diomedia (Getting, 1988, 1989) and Clione limacina (Arshavsky et al. 1985; Satterlie, 1989). In Clione it is interesting that the circuitry underlying both slow and fast swimming involves considerable electrotonic coupling between identified interneurones and motoneurones. Chemical connections are also important, especially in switching between the two swimming patterns (Satterlie, 1989). The role played by interneurones L/RPeDll in Lymnaea whole-body withdrawal is not yet clear, but they may help to filter sensory inputs so that the appropriate motor programme (withdrawal, locomotion, etc.) can be triggered.

In conclusion, a pair of electrically coupled interneurones has been shown to be electrically coupled to both foot and body wall motoneurones, providing a pathway by which contralateral motoneurones can be coordinated and modulated. These interneurones are also involved in modulating the cardiorespiratory system through chemical connections with heart motoneurones and respiratory interneurones. Furthermore, although interneurones L/RPeDll participate in wholebody withdrawal, their prime function appears to be the coordination of locomotor outputs. Left and right PeDll thus provide an example of interneurones specialized to serve both coordinating and multifunctional roles.

We thank Dr R. L. Ridgway and Dr A. G. M. Bulloch for their helpful and critical comments during preparation of this manuscript. Thanks are also due to Dr R. L. Ridgway for photographic assistance and to Mr D. Harrison for technical assistance. We also thank Caroline Collins for typing the manuscript. This work was supported by an SERC grant to W.W. N.I.S. was a University of Leeds Scholar and held an ORS award.