Inhibition of The Respiratory Pattern-Generating Neurons by an Identified Whole-Body Withdrawal Interneuron of Lymnaea Stagnalis

Rhythmic behaviors such as locomotion, respiration and feeding are generally thought to be mediated by discrete groups of synaptically interconnected neurons termed central pattern generators (CPGs) (Delcomyn, 1980; Kristan, 1980; Selverston, 1980; Pearson, 1985, 1993; Getting, 1988, 1989; Calabrese et al. 1989; Harris-Warrick and Johnson, 1989; Jacklet, 1989). It therefore seems logical to anticipate that the neuronal basis of behavioral choices may also hinge upon interactions between various CPG neurons. For instance, the neural network underlying a higher-priority behavior, such as an escape response, may also play an active role in terminating other ongoing motor programs.

The neuronal networks mediating higher-priority behaviors have been shown to suppress incompatible motor functions either at the behavioral or at the motor neuronal level (Davis et al. 1974a,b; Kyriakides and McCrohan, 1988; Teyke et al. 1990; Arshavsky et al. 1994a,b; Norekian and Satterlie, 1996). Whether the rhythm-generating neurons (CPG neurons) of the incompatible behaviors are also directly affected by the neurons of higher-priority behaviors has not yet been determined. Since communications between higher-order CPG neurons mediating different behavioral repertoires are necessary for behaviorally meaningful movements to occur, a detailed knowledge of their interactive capabilities is, therefore, critical for understanding the cellular basis of seemingly distinct motor functions.

The freshwater mollusc Lymnaea stagnalis is a useful model system for studying the cellular and molecular mechanisms underlying regeneration and synapse formation (Bulloch and Syed, 1992), the cellular basis of feeding (Benjamin, 1983), locomotion (Haydon and Winlow, 1986; Winlow and Haydon, 1986; Syed et al. 1988), respiration (Janse et al. 1985; Syed and Winlow, 1991b; Syed et al. 1991; Moroz and Winlow, 1992), whole-body withdrawal (Sakhorov and S.-Rózsa, 1989; Ferguson and Benjamin, 1991a,b; Syed and Winlow, 1991a; Winlow et al. 1992), egg-laying behavior (Ter Maat et al. 1989, 1992) and peptidergic signaling mechanisms (Boer et al. 1987). Of particular interest to us are two antagonistic motor functions: respiration and whole-body withdrawal behavior. In Lymnaea stagnalis, the behavioral and neuronal correlates of these behaviors are well characterized, although the exact cellular mechanisms that coordinate the interactions between these incompatible motor programs are not well understood.

Upon encountering a noxious stimulus, L. stagnalis terminates ongoing behaviors, such as feeding, respiration and locomotion, and withdraws its head–foot complex into the shell, an action pattern often referred to as whole-body withdrawal (Ferguson, 1984; Ferguson and Benjamin, 1991a,b; Syed and Winlow, 1991a). This is the only escape response available to the animal and, as it is a factor determining survival, it takes priority over other behaviors such as locomotion and respiration (Winlow et al. 1992). A network of identified, electrically coupled neurons, the LPeD11 and RPeD11 neurons, mediates whole-body withdrawal behavior in L. stagnalis (Syed and Winlow, 1991a).

To accomplish respiratory movements, i.e. opening and closing of its respiratory orifice, the pneumostome, L. stagnalis must extend its head–foot complex some distance out of the shell, a behavior that is incompatible with the whole-body withdrawal described above. Both behavioral and neuronal correlates of the respiratory behavior in L. stagnalis are well characterized (Syed et al. 1990, 1991; Syed and Winlow, 1991b). Various environmental stimuli, such as a shadow reflex, a light on–off response or a mechanical touch to the head–foot complex during spontaneously occurring respiratory movements, induce pneumostome closure and an immediate termination of the respiratory behavior (Syed, 1988; Lukowiak et al. 1994).

In the present study, we sought to determine the neuronal loci at which the behavioral selection between the withdrawal and respiration programs takes place. We provide direct evidence that an identified, higher-order, whole-body withdrawal interneuron, RPeD11, terminates the respiratory motor programs at the rhythm generator (CPG) level.

Laboratory-reared stocks of the freshwater snail Lymnaea stagnalis (L.), with an approximate shell length of 25–30 mm, were maintained at 18 °C in well-aerated artificial pond water and fed on lettuce. In order to enhance fictive aerial respiratory activity in the isolated central nervous system (CNS) preparations, animals were kept overnight in non-aerated pond water.

Prior to dissection, the animals were anesthetized in a 10 % solution of Listerine (ethanol 21.9 %, menthol 0.042 %) in normal saline. Animals failed to respond to a gentle touch to their head–foot complex within 10 min of Listerine treatment and were, therefore, considered to be anesthetized. These animals recovered completely within 20 min when placed back in normal pond water. The anesthetized animals were deshelled and the central ring ganglia were removed as described elsewhere (Syed and Winlow, 1991a,b). The isolated CNS was pinned to the bottom of the dissection dish and was maintained in a normal Hepes (50 mmol l⁻¹) -buffered Lymnaea saline (51.3 mmol l⁻¹ NaCl, 1.7 mmol l⁻¹ KCl, 4.1 mmol l⁻¹ CaCl2, 1.5 mmol l⁻¹ MgCl2, see Ridgway et al. 1991). The outer connective tissue sheath was removed using fine forceps, whereas the inner connective sheath was softened by applying a crystal of protease (Sigma type XIV) directly to the central ganglia for 15 s, followed by immediate wash-out with cold (4 °C) normal saline. Semi-intact animals were prepared according to previously published methods (Syed et al. 1991). Briefly, the pneumostome, lung and mantle cavity with intact innervation from the CNS were isolated and pinned to the bottom of the dish as described above.

Conventional intracellular recording techniques were used (Syed and Winlow, 1991a,b). Specifically, glass microelectrodes (1.5 mm internal diameter, WPI) were pulled on a vertical electrode puller (Kopf, 700C) and filled with a saturated solution of K2SO4. The tip resistance of these electrodes typically ranged from 25 to 35 MΩ. The intracellular signals were amplified via preamplifiers (Getting, model 5; Dagan, 8700), displayed on a storage oscilloscope (Tektronix, R5103N) and recorded on a Gould four-channel chart recorder (Brush 2400). To prepare high-Ca²⁺/high-Mg²⁺ salines, the concentrations of these divalent cations in the normal saline were raised ([Ca²⁺] to 24 mmol l⁻¹ and [Mg²⁺] to 12 mmol l⁻¹).

Established cell culture techniques for L. stagnalis neurons were used as described previously (Syed et al. 1990; Ridgway et al. 1991). Briefly, the animals were dissected under sterile conditions and, following a series of antibiotic washes and subsequent enzymatic treatments, the central ring ganglia were pinned to the bottom of the dissection dish. The connective tissue sheath was carefully removed with fine forceps and individual somata were removed using a fire-polished glass pipette coated with Sigma-Cote (Sigma). Neurons were isolated by applying gentle pressure through a Gilmont syringe and were plated onto culture dishes (Falcon, 3001) coated with poly-L-lysine containing brain-conditioned media (see Ridgway et al. 1991, for details). Following 24 h of sprouting and subsequent physical overlap between the neurites, both the pre-and postsynaptic neurons were viewed under an inverted microscope (Zeiss Axiovert 135, with Nomarski attachments). The sprouted neurons were impaled using the electrophysiological techniques described above and photographed with a 35 mm camera mounted on the inverted microscope.

In the first series of experiments, we demonstrate that the whole-body withdrawal interneuron RPeD11 terminates the spontaneously occurring respiratory rhythm in the isolated CNS preparation. Evidence is also presented to suggest that this inhibition may be direct as it persists in high-Ca²⁺/high-Mg²⁺ saline and also in the in vitro cell culture. A diagrammatic representation of the central ring ganglia of L. stagnalis depicting the locations of the neurons used in this study is presented in Fig. 1A. The identified interneuron RPeD11, involved in whole-body withdrawal, is located on the dorsal surface of the right pedal ganglion. A homolog of RPeD11 is located in the left pedal ganglion and is termed LPeD11 (Syed and Winlow, 1991a). Detailed morphological profiles of these cells have been published previously (Syed and Winlow, 1991a). These two neurons are electrically coupled to each other and also to a large number of body wall (cerebral A, CeA) and foot musculature (pedal G cluster, PeG) motor neurons (Fig. 1B). Together, these cells constitute the neural network that mediates whole-body withdrawal behavior in L. stagnalis (Syed and Winlow, 1991a).

Fig. 1C shows a summary diagram depicting the synaptic organization of the respiratory CPG neurons (modified from Syed and Winlow, 1991b). Brain preparations, isolated from animals that had been maintained overnight under anoxic conditions, exhibited periodic, spontaneous respiratory episodes (Fig. 2A). In this preparation, simultaneous intracellular recordings were made from right pedal dorsal 1 (RPeD1) and visceral dorsal 4 (VD4). As the input 3 interneuron (IP3I) is situated on the opposite side to the VD4 and RPeD1 neurons (Syed et al. 1990), indirect evidence for IP3I activity was obtained by recording from a pneumostome opener muscle motor neuron (visceral J, VJ). IP3I activity, which underlies expiration (pneumostome opening), was recorded as a series of characteristic bursts in a VJ cell and RPeD1 (Benjamin and Winlow, 1981). Bursting activity in VD4 represents inspiration (pneumostome closing) (Syed and Winlow, 1991b). In previously quiescent preparations, RPeD1 stimulation induced respiratory activity in the CPG network (Syed et al. 1990). One such respiratory cycle is shown in Fig. 2B. RPeD1 inhibits VD4 and initiates IP3I activity by a dual inhibitory–excitatory synapse (Syed et al. 1990). IP3I in turn excited a VJ cell as well as RPeD1 itself. A combined inhibitory effect of both RPeD1 and IP3I activity caused excitation of VD4 (Fig. 2B; see also Fig. 1C and Syed and Winlow, 1991b) via mechanisms other than post-inhibitory rebound excitation (Syed et al. 1992).

In semi-intact preparations, we have previously demonstrated that a gentle mechanical touch to the pneumostome area terminated respiratory activity (Syed and Winlow, 1991a). In the present study, we considered the possibility that this inhibition of the respiratory CPG neurons was mediated by RPeD11.

To determine the effects of RPeD11 on spontaneously occurring respiratory CPG activity, simultaneous intracellular recordings were made from RPeD11, the respiratory CPG neurons RPeD1 and VD4 and a motor neuron VJ cell. Indirect evidence for the activation of IP3I was again obtained from characteristic excitatory discharges in a VJ cell and RPeD1 (Benjamin and Winlow, 1981). The isolated CNS preparations generated regular, rhythmical respiratory discharges (Fig. 3A), similar to those seen in semi-intact preparations (Syed et al. 1991). Brief electrical stimulation of RPeD11 (at a frequency of 3–5 Hz for a duration of 5–10 s) during spontaneously occurring respiratory activity terminated the respiratory rhythm in all of the CPG neurons recorded from various different preparations (N=20) (Fig. 3A). Furthermore, the next episode of respiratory activity in both RPeD1 and the VJ cell was weaker (i.e. fewer spikes in RPeD1, shorter duration of the effect of IP3I bursts on the VJ cell) and uncoordinated, and it failed to activate VD4. Normal rhythmic respiratory activity resumed in the second episode following RPeD11 stimulation (Fig. 3A). A similar but repeated stimulation of RPeD11 (two stimuli at a frequency of 3–5 Hz, with an interval of 15–20 s) during spontaneously occurring respiratory activity terminated the respiratory rhythm for 30–60 min (N=10) (Fig. 3B). These data demonstrate that the electrical stimulation of RPeD11 in an isolated brain preparation is sufficient to inhibit the spontaneously occurring respiratory rhythm.

To determine whether RPeD11 inhibited respiratory CPG neurons directly or via polysynaptic pathways, simultaneous intracellular recordings were made from RPeD11, RPeD1 and VD4 in isolated brain preparations. In the presence of normal saline, activity in RPeD11 inhibited the activities of both RPeD1 and VD4 (Fig. 4A). Unfortunately, in neither of these preparations was it possible to resolve 1:1 inhibitory postsynaptic potentials (IPSPs) between RPeD11 and RPeD1/VD4. These inhibitory connections, however, persisted in the presence of high-Ca²⁺/high-Mg²⁺ saline (N=9) (Fig. 4B), suggesting the possibility that they may be monosynaptic in nature. Owing to anatomical constraints (dorsal versus ventral), similar data were not obtained for IP3I.

To determine the effects of RPeD11 stimulation on the pneumostome closer muscle (PM), on the columellar muscle (CM) and on the activity of RPeD1, a semi-intact preparation (lung–mantle, pneumostome and CNS, Syed et al. 1991) was used. RPeD11 stimulation in these semi-intact preparations induced contractions of both PM and CM, while inhibiting the spontaneous activity in RPeD1 (N=10) (Fig. 5A). These data suggest that activity in RPeD11 is sufficient to inhibit respiratory behavior both at the CPG and at the affector organ level. A noxious stimulus delivered to the pneumostome area (touching with a pair of blunt forceps), however, activated RPeD11 for several seconds, generating either single spikes or bursts of action potentials (Fig. 5B). This excitation of RPeD11, in turn, was correlated with a prolonged inhibition of RPeD1. Taken together, these data demonstrate that, regardless of its stimulation parameters (direct or indirect, short versus prolonged bursts of spikes, etc.), RPeD11 activity is sufficient to inhibit the respiratory CPG neuron in the semi-intact animals.

Since RPeD11 is electrically coupled to its contralateral homolog LPeD11 (Syed and Winlow, 1991a), we tested the possibility that inhibition of the respiratory neurons by RPeD11 may have been mediated indirectly via LPeD11. Since saline containing high levels of divalent cations does not perturb electrical connections (Syed and Winlow, 1991a), we used intracellular current injection to manipulate the membrane potentials of either RPeD11 or LPeD11 selectively. Simultaneous intracellular activity was recorded from RPeD11, LPeD11 and RPeD1 in an isolated brain preparation. The presence of electrical coupling between RPeD11 and LPeD11 was demonstrated with the passage of both hyperpolarizing and depolarizing current pulses between the cells (Fig. 6). Stimulation of LPeD11 with depolarizing current injection produced electrotonic potentials in RPeD11 maintained at rest (−65 mV). It did not, however, inhibit RPeD1 activity (Fig. 6A). RPeD11 was then depolarized near its firing threshold (−55 mV) and LPeD11 was stimulated once again. LPeD11 now successfully generated spikes in RPeD11 which, in turn, inhibited RPeD1 activity (Fig. 6B). These experiments were highly reproducible, and the data (N=7) suggest that RPeD1 is inhibited directly by RPeD11 and not via LPeD11. To determine whether LPeD11 contributes to RPeD11-induced inhibition of RPeD1, the current was injected into RPeD11 in order to activate a previously depolarized LPeD11 (held near its firing threshold). Under these experimental conditions, regardless of the spiking activity in LPeD11, activity in RPeD11 produced a similar inhibition in RPeD1 (Fig. 7A,B). These data serve to demonstrate further that RPeD1 inhibition is mediated by RPeD11 and that LPeD11 does not contribute significantly to this inhibition.

Although the synaptic connections between RPeD11 and the respiratory CPG neurons persisted in the presence of high levels of divalent cations, it was impossible to discern 1:1 IPSPs in the target cells. Moreover, since RPeD11 has extensive electrotonic couplings with a large number of other motor neurons (Syed and Winlow, 1991a), it was difficult to obtain unequivocal evidence that the connections were indeed direct. Furthermore, owing to anatomical constraints, we had yet to confirm the nature of the synaptic connections between RPeD11 and the IP3I in isolated brain preparations in vivo. Therefore, in order to determine the nature of these connections, we utilized in vitro cell culture, where specific synapses between L. stagnalis neurons can be reconstructed reliably.

Individual cell somata of the respiratory CPG neurons RPeD1, VD4 and IP3I were isolated from their respective ganglia and cultured in vitro with RPeD11 under conditions that support neurite outgrowth and synapse formation (Syed et al. 1990; Ridgway et al. 1991). When cultured with an individual respiratory interneuron, RPeD11 always formed inhibitory synapses. Specifically, when co-cultured with RPeD1 (N=13, Fig. 8A), VD4 (N=7, Fig. 9A) or IP3I (N=3, Fig. 10A), RPeD11 formed inhibitory synapses with these cells (Figs 8B, 9B, 10B) that were similar to those observed in vivo (Figs 3, 4). In these in vitro preparations, in the absence of other synaptic and non-synaptic influences, unitary 1:1 IPSPs were often discernible, particularly between RPeD11 and VD4 and between RPeD11 and IP3I (Figs 9B, 10B). These data demonstrate that the inhibitory connections between RPeD11 and the respiratory CPG neurons can be reconstructed in vitro, providing us with an opportunity to investigate in detail the pharmacological basis of synaptic interactions between these cells.

In this study, we have provided evidence that a higher-order, whole-body withdrawal interneuron RPeD11 directly inhibits respiratory CPG neurons. At the neuronal level, our data are consistent with the behavioral observation that stimuli which induce whole-body withdrawal terminate respiratory behavior (Winlow et al. 1992).

Whole-body withdrawal and escape responses have been shown to take priority over other behaviors in molluscs, including L. stagnalis (Syed and Winlow, 1991a; Winlow et al. 1992), Aplysia californica (Walters and Erickson, 1986), Planorbis corneus(Arshavsky et al. 1994a,b), Helix lucorum (Balaban, 1983; Maximova and Balaban, 1984), Tritonia diomedia (Getting, 1977) and Clione limacina (Huang and Satterlie, 1990; Norekian and Satterlie, 1996). The sole exceptions appears to be Pleurobranchaea and the holoplanktonic pteropod mollusc Clione limacina, where feeding behavior dominates other behaviors, such as head withdrawal, righting and mating (Davis et al. 1974a; Norekian and Satterlie, 1996). In most animals, the dominance of escape responses over other behaviors is thought to stem from the whole-body withdrawal neurons dominating other neural networks that mediate incompatible behaviors. In most of these species, inhibitory synaptic connections between withdrawal-associated neurons and those involved in the incompatible behaviors have been observed. Since the inhibitory effects of withdrawal neurons were primarily observed either at the behavioral or at the motor neuronal level (see Introduction), the extent to which the withdrawal circuit directly influenced neurons at the level of higher-order neurons remained undetermined.

We have previously demonstrated that the respiratory motor neurons were monosynaptically influenced by RPeD11 (Syed and Winlow, 1991a). In the present study, we provide evidence for the inhibition of L. stagnalis respiratory CPG interneurons (RPeD1, VD4, IP3I) by a whole-body withdrawal neuron RPeD11. Our data are, therefore, consistent with the hypothesis that neurons mediating higher-priority motor functions inhibit incompatible behaviors not only at the motor neuronal but also at the higher-order CPG neuron level. The behavioral significance of such an extensive suppression of motor command, at all levels of its motor output, would be to ensure an effective and efficient termination of an incompatible behavior.

As described above, during spontaneously occurring respiratory movements in a freely behaving animal, a gentle mechanical touch to the pneumostome area causes its closure and the termination of the respiratory behavior (Lukowiak et al. 1994). When the animal is not ventilating, however, a stronger stimulus applied to the head–foot complex of both L. stagnalis (Sakhorov and S.-Rózsa, 1989) and Planorbis corneus (Arshavsky et al. 1994b) causes a forceful opening of the pneumostome. The behavioral significance of these pneumostome openings may be to reduce the body volume, by removing air from the lung and hemolymph from the mantle cavity (Sakhorov and S.-Rózsa, 1989; Arshavsky et al. 1994b). In Helix lucorum, however, noxious stimuli that induce whole-body withdrawal always cause pneumostome closure (Balaban, 1983).

In both Helix lucorum and Planorbis corneus, the neurons that control withdrawal behavior appear to control the pneumostome as well. For instance, a command neuron, LPa3, underlies whole-body withdrawal behavior in Helix lucorum. In addition to activating withdrawal reflexes, electrical stimulation of LPa3 in a semi-intact preparation induces pneumostome closing movements (Balaban, 1983; Vehovszky et al. 1989). In Planorbis corneus, the electrical stimulation of the withdrawal neuron DRN1 induces pneumostome opening movements (Arshavsky et al. 1994b). These data, therefore, suggest that, in addition to inducing withdrawal reflexes, the neurons controlling withdrawal behavior may also exert motor control over the pneumostome musculature.

This generic involvement of the pneumostome in both respiratory and withdrawal behaviors led some researchers to suggest that these two motor functions were controlled by the same network (Sakhorov and S.-Rózsa, 1989). For instance, Sakhorov and S.-Rózsa (1989) raised the possibility that the IP3I activity (pneumostome openings as observed during respiration) in L. stagnalis may be a component of the whole-animal withdrawal behavior. Our data, however, are not consistent with their hypothesis. We have previously demonstrated, in a semi-intact preparation, that a gentle mechanical touch to the pneumostome (inducing whole-body withdrawal) terminates IP3I activity in the middle of its discharge (Syed and Winlow, 1991a), suggesting that it is not a component of the withdrawal reflex. In the present study, we have demonstrated inhibitory synaptic connections between the withdrawal neuron RPeD11 and the respiratory CPG neurons (RPeD1, VD4 and IP3I). Moreover, we have also demonstrated in a semi-intact preparation that a noxious stimulus delivered to the pneumostome area inhibits the respiratory interneuron RPeD1 (Fig. 6). Taken together, both the behavioral analysis and the network configuration reinforce our view that pneumostome opening and the underlying motor discharges (IP3I activity), as observed during active respiratory movements, are not a component of the withdrawal reflex.

If the neurons involved in mediating the respiratory and withdrawal behaviors are different, how are these two behaviors coordinated? Owing to the lack of information regarding the identity of the respiratory neurons in both Helix lucorum and Planorbis corneus, the synaptic relationships between withdrawal and respiratory neurons have not been worked out. In L. stagnalis, however, we have now provided direct evidence that the withdrawal neuron RPeD11 not only terminates the respiratory rhythm at the higher CPG level but also induces pneumostome closure. We believe that RPeD11 may achieve this task not only by inhibiting the respiratory CPG neurons directly but also by inducing the contractions of both withdrawal (columellar) and pneumostome opener muscles.

To deduce the direct versus indirect nature of synaptic connections between RPeD11 and the respiratory CPG neurons, we recorded intracellularly in a solution containing a high concentration of divalent cations. Under these experimental conditions, the synaptic connections between RPeD11 and the respiratory CPG neurons persisted, suggesting that they may be monosynaptic. Moreover, stimulation of other electrically connected neurons, such as LPeD11, did not mimic the effect of RPeD11 stimulation on the target cells, again suggesting that RPeD11 may affect its targets directly. Similar methods have previously been used to demonstrate monosynaptic connections in L. stagnalis (Syed and Winlow, 1991a). However, owing to extensive electrotonic coupling between RPeD11 and other connected cells (L/RCeA, L/RPeG), providing unequivocal evidence for direct versus indirect polysynaptic pathways was still difficult. The presence of electrically coupled neurons in any given circuit often makes it difficult to determine the monosynaptic nature of connections between the neurons in question. Various biochemical and cell deletion techniques, such as treatment with solutions containing a high level of divalent cations (Berry and Pentreath, 1976), enzyme injections (Marder, 1984) and photoinactivation of the neurons (Miller and Selverston, 1979; Elliott and Kleindienst, 1990; Syed and Winlow, 1991a), have elucidated the roles of individual electrically coupled neurons in various circuits. Owing to the extensive electrical coupling between the many motor neurons underlying L. stagnalis withdrawal behavior, these approaches are much less feasible in this case.

To demonstrate that the connections between RPeD11 and the respiratory CPG neurons were likely to be monosynaptic, we constructed these synapses in vitro. Adult L. stagnalis neurons, like those of many other invertebrate species, regrow their processes in culture and reform specific synapses similar to those observed in vivo (Bulloch and Syed, 1992). In a number of these invertebrate species, inappropriate synapses may also form in culture (Chiquet and Nicholls, 1987; Kleinfeld et al. 1990). In L. stagnalis, however, we found RPeD11 to establish inhibitory synapses with its target cells in vitro. Although the synapses between RPeD11 and the respiratory neurons described in this study were of the appropriate type (inhibitory), the compound IPSPs were of larger amplitude and longer duration than those recorded in the in vivo preparations. The augmentation of synaptic responses in vitro may have been due in part to the absence of the synaptic and non-synaptic interactions with a multitude of other cells that often make it difficult to resolve such subtle responses in vivo.

In summary, we have provided evidence that the neuronal networks underlying the respiratory and whole-body withdrawal behaviors are organized in a hierarchical manner, with the withdrawal network taking priority over the respiratory circuit. RPeD11, which plays a key role in coordinating the whole-animal withdrawal behavior, also inhibits respiratory CPG neurons directly and, therefore, serves as a main determinant of the behavioral hierarchy between whole-body withdrawal and respiration. These data provide a unique example in which an incompatible behavior, i.e. respiration, is terminated by the whole-body withdrawal neuron RPeD11 at all levels of CPG organization. Specifically, RPeD11 terminated respiratory rhythm generation at the rhythm generator (CPG) level (Fig. 3A,B), motor command was terminated at the motor neuronal level (Syed and Winlow, 1991b) and the expression of behavior was further curtailed at the muscular level (Fig. 5A). In conclusion, our data suggest that, in order for it to be effective, a neuronal circuit situated hierarchically higher than the others must terminate incompatible motor commands simultaneously at all levels of motor output organization.

We acknowledge the excellent technical assistance provided by Mr M. Wali Zaidi and Mrs. A. Rao. The authors also wish to thank Dr G. Spencer and Mr S. Schultz for their critical comments during the preparation of this manuscript. This work was supported by an MRC operating grant to N.I.S. T.I. was supported by a postdoctoral fellowship from the Ministry of Education, Science and Culture of Japan. N.I.S. is an AHFMR Scholar, a Parker B. Francis fellow and an Alfred P. Sloan fellow.