Control of a Central Pattern Generator By an Identified Modulatory Interneurone in Crustacea I. Modulation of the Pyloric Motor Output

A number of rhythmic sequences of behaviour are programmed by central neuronal networks, the endogenous activities of which are determined by the synaptic interactions between neurones in the network and/or by the cellular properties of neurones in the network (see reviews by Fentress, 1976; Grillner, 1977; Delcomyn, 1980). The identification of the neurones in such a network and the determination of their synap tic relationships is, however, insufficient to explain either the flexibility of the behavioural sequences controlled by such networks or the manner in which such sequences are initiated and terminated. In fact, it appears that even those properties determining the ‘endogenous’ activity of pattern generators can be modulated. For example, the bursting activity of oscillatory neurones can be induced, amplified or eliminated (Parnas, Armstrong & Strumwasser, 1974; Barker, Ifshin & Gainer, 1975), and the efficacy of synaptic relationships between neurones can be modified by certain slow synaptic potentials (see review by Hartzell, 1981). Thus, it is of interest to study the mechanisms controlling central pattern generators.

Aspects of the control of pattern generators are examined in the present study, in which we show that the activity of a single identified interneurone can profoundly modify the output of the pyloric network, which is a motor pattern generator in the crustacean stomatogastric nervous system. This neuronal network, which controls the rhythmic movements of the pyloric stomach, is composed of 14 identified neurones, whose synaptic connections are known (Maynard, 1972; Maynard & Selverston, 1975; Selverston, Russell, Miller & King, 1976; Selverston & Miller, 1980). The rhythmic activity of this motor pattern generator is largely due to the fact that some of the neurones are endogenous oscillators and act as pacemakers. In addition, reciprocal inhibitory synapses between the various neurones of the network contribute to the generation of the pyloric rhythm. However, the rhythm and the intensity of the pyloric pattern are also strongly dependent upon input from higher nervous centres (Russell, 1979). This dependence is illustrated by the observations that the oscillatory activity of the pacemaker neurones can be induced or amplified by such inputs (Russell & Hartline, 1981; Moulins & Cournil, 1982) and that such inputs may also control regenerative membrane properties of the non-oscillatory neurones of the network that contribute to the generation of the pyloric pattern (Russell & Hartline, 1978). The source of these inputs has, however, been unknown.

The present study shows that an effective ‘higher centre’ input is a single interneurone which is located in an anterior ganglion and projects directly to the stomatogastric ganglion. We here report that discharge of this neurone profoundly modifies the activity of all the neurones of the pyloric network. The activity of this neurone both increases the frequency of the rhythm and the frequency of spikes within bursts of the neurones (quantitative modifications) and modifies phase relationships between the discharges of the neurones (qualitative modifications). These modifications are established slowly and are of long duration. Thus it appears that a single interneurone can alter a rhythmic motor output. In certain cases, these modifications of the pyloric pattern resemble those which occur when feeding starts in the intact animal. In the accompanying paper (Dickinson & Nagy, 1983) we shall show that these modifications of the pyloric rhythm by a modulatory interneurone are based primarily on the induction and amplification of the regenerative membrane properties which are responsible for the ‘burstiness’ of all the neurones of the pyloric pattern generator.

A preliminary report of this work has been published elsewhere (Nagy, Dickinson & Moulins, 1981).

Both male and female Cape lobsters, Jasus lalandii, (41 animals) and red lobsters, Palinurus vulgaris, (6 animals) were used. Similar results were obtained from the two species; the data illustrated are fromJasus except where otherwise indicated. Animals were maintained in aerated, flowing sea water before use, at a temperature of 16−20 °C.

Experiments were conducted on isolated preparations of the stomatogastric nervous system, pinned on Sylgard-covered Petri dishes and constantly superfused with oxygenated saline (temperature around 20°C). The isolated preparation used here corresponds to the ‘combined preparation’ used by Selverston et al. (1976; Fig. 1A) and consists of the stomatogastric ganglion (STG), the oesophageal ganglion (OG), the two commissural ganglia (CG) and the connecting nerves. The four ganglia were desheathed to allow access to the neuronal cell bodies. Activity of neurones in the stomatogastric ganglion was recorded extracellularly on three pairs of motor nerves: the dorsal and ventral lateral ventricular nerves (dlvn, vlvn) and the medial ventricular nerve (mvn). Intracellular recordings were made from cell bodies of the neurones, and their identities were established by the presence of action potentials in various motor nerves of the stomatogastric ganglion, the time of discharge in the pyloric sequence and synaptic relationships with other neurones of the network (Maynard, 1972; Maynard & Dando, 1974; Maynard & Selverston, 1975).

Details of the dissection, the experimental arrangements and recording techniques have all been previously described (Moulins & Nagy, 1981). Intracellular recordings were made using glass microelectrodes filled with 3M-KC1 (resistances of 20-30 MQ). The same electrodes were used for current injection via a bridge circuit. Platinum wire electrodes were used for extracellular recordings and stimulations of the various nerves. In some experiments, the ganglia were surrounded by petroleum jelly walls so that each ganglion could be bathed in a separate solution. The normal saline used was artificial sea water. To block synaptic activity within a ganglion, the Ca²⁺ ions in the saline were replaced with Mg²⁺, and 12 mm-Co²⁺ was added; this solution is called 0Ca²⁺+Co²⁺ saline throughout the paper. For pharmacological experiments, the drugs to be tested were dissolved in saline (pH 7-45) just before use. These solutions were continuously perfused through the Vaseline chamber (volume approximately 1 ml) and the effects were recorded after 20−30 volume changes (about 30 min). The following drugs were used: atropine sulphate (Sigma), eserine (Sigma), hexamethonium bromide (Sigma), picrotoxin (gift from Fluka A.G.), tetrodotoxin (Sigma) and J-tubocurarine chloride (Sigma).

The basic organization of the pyloric network in Panulirus was first described by Maynard (1972) and Maynard & Selverston (1975). Injasus, this organization is quite similar, except that the pyloric rhythm has a mean frequency of about 1 Hz, instead of about 2 Hz as in Panulirus (see Selverston et al. 1976). The pyloric network consists of 14 neurones (Fig. 1B) whose cell bodies are located in the stomatogastricganglion (Fig. 1A, STG). The rhythmic output of the network (i.e., the pyloric rhythm, Fig. 1C) is driven by endogenous oscillations of the membrane potentials of three electrically coupled pacemaker neurones. One of these, the anterior burster (AB) is an interneurone whose axon enters the commissural ganglia (Fig. 1A, CG); the other two are motoneurones which drive the dilator musculature of the pyloric filter (pyloric dilator neurones, PD). Because they are electrically coupled, these three neurones have very similar and synchronous activity and can be considered as a functional unit. The Constrictor muscles of the pyloric filter are controlled by the lateral pyloric neurone (LP) and eight pyloric neurones (PY); the latter are electrically coupled and hence will be considered as a single functional element. The dilator muscles of the car-diopyloric valve, which controls the entrance to the pyloric filter, are under the control of the ventral dilator motor neurone (VD), whilst the constrictor muscles of this valve are driven by the inferior cardiac motoneurone (IC). The pacemaker neurones rhythmically inhibit the activity of the constrictor motoneurones, resulting in a regular alternation of dilator and constrictor discharges (Fig. 1C). These inhibitory synapses are generally considered to be functionally the most important synapses in determining the expression of the pyloric sequence (Selverston et al. 1976). The other synaptic relationships within the network are shown in Fig. IB.

We have identified an interneurone, the anterior pyloric modulator (APM) (Nagy et al. 1981), whose cell body is located in the oesophageal ganglion (Fig. 1A, OG) and whose discharge activates the pyloric network and leads to major modifications of the pyloric output. Fig. ID shows extracellularly recorded activity of a pyloric motor nerve (the vlvn), illustrating the modifications resulting from a 4 s discharge of APM. The frequency of bursts increases, as do the number and frequency of spikes recorded in each burst.

Activation of APM modifies the activity of the pyloric neurones, which are located in the stomatogastric ganglion (Fig. 2A, STG). Although only one afferent nerve (the stomatogastric, Fig. 2A, stn) enters the stomatogastric ganglion in the isolated preparation, two morphological pathways connect this nerve to the oesophageal ganglion and thence to the cell body of APM: a short pathway, via the oesophageal nerve (on), and a longer pathway, via the inferior oesophageal nerve (ion), the commissural ganglion (CG), and the superior oesophageal nerve (son). Cutting the on does not alter APM’s effects on the pyloric neurones, suggesting that APM exerts its effects via the longer pathway. Furthermore, cutting a single ion or son diminishes, but does not eliminate, the effects of APM on the pyloric neurones, whereas cutting both tons or both sons does eliminate APM’s effects. Thus, it appears that APM influences the pyloric neurones via a bilateral pathway, which passes through both ions and both sons to the stn, and thence to the stomatogastric ganglion.

Three lines of evidence, shown in Fig. 2, indicate that two bilaterally symmetrical axons of APM, both entering the stomatogastric ganglion, are responsible for the effects of APM on the pyloric neurones. There is a one-to-one correspondance of action potentials recorded in the soma of APM and those recorded in the ion and then in the stn (Fig. 2B). In addition, antidromic spikes are recorded in the soma of APM (Fig. 2C) in response to stimulation of the ion (S₁), the ipsilateral ion (S₂) and the sin (S₃). The time required for these antidromic spikes to reach the soma increases as the stimulus is moved from ion to son to stn. Lastly, one can generate spike collisions between antidromic spikes provoked by stimulating one ion (Fig. 2D, S₁) and the ipsilateral son (S₂). The recordings shown here are from a single side, but either side gives identical results, confirming the bilateral symmetry of APM’s axons. To eliminate the possibility that the axons of APM do not actually enter the stomatogastric ganglion, but instead synapse on and activate interneurones in the commissural ganglia (see Fig. 2A), synaptic activity in the commissural ganglia was blocked with a 0Ca²⁺+Co²⁺ (12 mm) saline during the experiments described above.

We can therefore conclude that APM is a ‘T’-shaped neurone, sending one axon down each ion; these axons traverse the commissural ganglia to enter the sons, and then the single stn. The two axons, which are about 4cm in length, thus enter the stomatogastric ganglion, where they exert their effects. APM is the only one of the 12 neurones in the oesophageal ganglion which has this morphology and which can modify the pyloric activity. Consequently this neurone is relatively easy to identify using electrophysiological tests.

The shape of the APM action potentials in Fig. 2C and D shows a certain complexity: the spikes are overshooting (reaching +15 to +20 mV) and have a shoulder on the rising phase. If the axon of APM is stimulated repetitively, the shoulder becomes increasingly pronounced (Fig. 3B, C, D). After a number of stimuli, the overshooting spike is replaced by a much smaller spike (about 40mV), the amplitude of which corresponds to that of the shoulder on the larger action potential (Fig. 3D, E).By analogy to certain neurones in locusts (Heitler & Goodman, 1978; Hoyle & Dagan, 1978), we suspect that the larger spikes (Fig. 3B) are action potentials that fully invade the soma, whilst the smaller spikes are axonal spikes, which are revealed as the excitability of the soma progressively decreases during repetitive stimulation. This type of spiking pattern suggests the presence of an inexcitable zone at the junction of the two APM axons (Fig. 3A, inset; see also Heitler & Goodman, 1978). When antidromic stimulation produces an action potential which does not invade the soma, that action potential likewise fails to invade the contralateral ion (Fig. 3E); however, action potentials which do invade the soma also invade the contralateral ion (Fig. 3B, C; circle, bottom trace). Thus if the soma of APM is sufficiently excitable and a somatic spike is produced (spontaneously or with current injection), an action potential in each APM axon is propagated towards the stomatogastric ganglion. In other words, each action potential recorded in the soma of APM is accompanied by two action potentials which reach the stomatogastric ganglion via the ions, sons and stn.

Activity in APM causes activation of all the neurones of the pyloric network (Fig. 4) This activation is accompanied by major modifications of the pyloric pattern although the extent of these modifications is dependent in part upon the previous level of activity in the pyloric network.

It has been shown in both intact lobsters (Rezer & Moulins, 1980) and in isolated nervous systems (Nagy, 1981) that the pyloric rhythm can take one of two general forms: a slow, irregular rhythm with a cycle period greater than 2 s and in which some the pyloric neurones may not be active (Fig. 4A, before APM firing; Fig. 4B1) and a fast, regular rhythm with a period of 1-2 s, in which all the pyloric neurones are active (Fig. 4C, before APM firing).

If APM is induced to fire when the pyloric rhythm is slow and irregular, its overall effect takes one of two forms, both of which are accompanied by extensive increases in spike frequency of all the pyloric neurones. Either the overall frequency of pyloric bursting is markedly increased (acceleration of the pyloric rhythm, Fig. 4A) or it is accelerated only slightly and temporarily (Fig. 4B). In the latter case, the pyloric neurones, especially the constrictor neurones (LP and PY), tend to fire in plateaus that are considerably longer than normal (see Fig. 10D, PY). If, on the other hand, the pyloric rhythm is already rapid and regular, the major effects of an APM discharge are increased amplitude of oscillation and increased spike frequency in all pyloric neurones, but little or no increase in pyloric cycle frequency (Fig. 4C).

As seen in Fig. 4A, activity in APM can provoke a strong acceleration of a spontaneously slow pyloric rhythm. This effect is also shown in Fig. 5B, in which the pyloric frequency is represented by the frequency of membrane potential oscillations in the pacemaker neurone PD. The increase in its oscillation frequency after a 6s APM discharge is illustrated by plotting oscillation frequency as a function of time (Fig. 5D; open circles). Maximum frequency is reached 4−5 s after the start of APM spiking; the frequency of the pyloric rhythm then slowly decreases and is still greater than the control 45 s after the end of the APM discharge. It can thus be seen that the duration of this effect of APM is eight-to ten-fold longer than the duration of the discharge that provoked it. Fig. 5C shows that, in contrast to the strong acceleration seen in Fig. 5B, when the pyloric rhythm is rapid (Frequency 1−1·2 Hz), APM firing does not affect the pyloric frequency. This insensitivity of the frequency of the rapid pyloric rhythm is depicted graphically in Fig. 5D (filled squares). APM still affects the pyloric neurones, however, as seen in the other modifications it provokes (e.g. changes in amplitude of oscillations and in the efficacy of synaptic inputs, which will be discussed shortly; Fig. 5C, LP). Note that the maximum frequency reached by the slow rhythm (Fig. 5D, open circles) under the influence of APM is very close to that of the rapid rhythm (Fig. 5D, filled squares; approximately 1 Hz). Although the two cases represented in Fig. 5 are near the extremes of spontaneous pyloric frequency, the maximum frequency reached is around 1 Hz in all cases in which APM causes an acceleration of the pyloric rhythm.

A second effect of APM activity is an increase in the spike frequency within the bursts of all of the pyloric neurones. This is shown graphically in Fig. 6, which corresponds to the recording of Fig. 4C, in which the pyloric frequency was relatively high before APM was induced to fire. Two major differences between the effects of APM on the constrictor motoneurones (LP, PY) and on the pacemaker neurones (PD) can be seen. Firstly, the extent of the increase in spike frequency is considerably greater for the constrictors (80−90% increase) than for the pacemakers (15−17% increase). Secondly, the duration of the increase is greater for the constrictor than for the pacemaker neurones. Spike frequency of the constrictors is still greater than control 45 s after the APM discharge (it returns to pre-stimulation level in about Bmin); here again the effects last eight to ten times the duration of the APM firing. In the pacemakers, on the other hand, the increase in spike frequency lasts only 4 −5 s after the APM discharge. This stronger activation of the constrictors is also seen when the pyloric rhythm is spontaneously slow (Fig. 10D, see also Dickinson & Nagy, 1983).

The third major effect of an APM discharge is a modification of the amplitudes of oscillations of the pyloric neurones (Fig. 7). An APM discharge (6s, about 30 Hz) causes an increase in overall oscillation amplitude in both the pacemaker PD and the constrictor LP (25% and 35% increases, respectively) (Fig. 7C,_1,2). In the constrictor PY, however, amplitude first decreases (to a maximum of 30%), then increases (8%) (Fig. 7C,₃). These effects are more easily understood if one examines separately the temporal evolution of the peaks and troughs of the oscillations (Fig. 7B). For all three neuronal types, the peak reaches a potential that is 3 −5 mV less negative (more depolarized) than normal (Fig. 7B, P). The increase in spike frequency in all three neurones corresponds to this change in the peaks of oscillation. In contrast, the troughs of the oscillations reach a potential that is 5 −6 mV more negative (more hyperpolarized) in the pacemaker PD and the constrictor LP, but is 12 mV less negative (more depolarized) in the constrictor PY (Fig. 7B, T). In PY the peaks remain depolarized after the troughs have returned to the control level, thus explaining the double effect on the overall amplitude. It is noteworthy that the temporal evolution of the amplitude effects of APM’s discharge is analogous to that of both spike frequency and the frequency of the pyloric rhythm (long latency, long duration). Because the bursting activity of the pyloric neurones is due both to synaptic connectivity within the network and to endogenous properties of the neurones (see Selverston et al. 1976; Russell & Hartline, 1978), it may be asked which of these two factors is modified by APM’s discharge and which results in the observed changes in the amplitude of oscillation of the pyloric neurones. The parallel depolarizations of the peaks of oscillation in the three neuronal types may result from modifications APM provokes in the properties of all three neurones (see Dickinson & Nagy, 1983). In contrast, as will be shown below, the evolution of the troughs of the oscillations is at least partly due to changes in the efficacy of synapses within the pyloric network after an APM discharge.

In addition to the quantitative modifications of the pyloric rhythm discussed above, APM provokes qualitative modifications that involve changes in the phase relationships amongst the various neurones. These changes are caused by modifications in the efficacy of synapses within the network. Amongst those affected are the inhibitory synapses between the pacemaker neurones (PD) and the constrictor neurones (LP and PY), synapses which are thought to be functionally the most important in determining the phase relationships between the firing of the neurones in the pyloric network (Selverston et al. 1976; Selverston & Miller, 1980).

We first considered the effects of APM on the inhibition of LP by PD (Fig. 8A). The apparent efficacy of this synapse is reflected in the absolute value to which the membrane potential of LP falls due to the inhibition by PD (Fig. 8B, filled triangles). This value is much less negative after an APM discharge than is the control (compare Bz with B₁) indicating that the efficacy of the synapse has decreased.The development of this potential before, during and after APM firing is shown in Fig. 8C (LP). The decreased efficacy of the inhibition of LP by PD is established slowly (maximum decrease 4−6 s after the start of APM firing) and is of long duration (lasting for at least 30s after a 6s APM discharge). Fig. 8C also shows simultaneous changes in action potential frequency in PD (curve marked PD, same curve as Fig. 6A). The observed decrease in synaptic efficacy is unexpected for two reasons: firstly, it occurs when the spike frequency in the presynaptic neurone PD increases; secondly, it occurs when the membrane potential of the postsynaptic neurone LP is less negative (and hence further from the reversal potential for the ipsps), which should increase the apparent hyperpolarization of LP.

The most negative values of LP’s membrane potential (troughs of LP oscillations) are normally determined by the inhibition LP receives from PD (Fig. 8Bj, filled triangles). We have seen that the LP potential resulting from the inhibition by PD is less negative after an APM discharge; however, after such a discharge, the absolute value of the troughs of LP’s oscillations is more negative (Fig. 7Bz, T). This apparent contradiction can be explained by the observation that the hyperpolarization of LP results from two separate inhibitions (Fig. 8A): that from PD (Fig. 8B1, filled triangle) and that from the PY neurones (Fig. 8B1, open triangle). After an APM discharge, the efficacy of the inhibitory synapse of PD onto LP decreases, but that of the inhibitory synapse of PY onto LP increases, and the value of the LP membrane potential determined by the inhibition from PY becomes more negative (Fig. 8B,2, open triangle). Fig. 8D (LP) shows the evolution of this potential as a function of time after an APM discharge, as well as the parallel evolution of spike frequency in PY (Fig. 8D, PY; same curve as Fig. 6C). Consequently, when APM fires, it is the inhibition from PY rather than that from PD that determines the extreme values of the LP troughs. Thus, a discharge of APM causes a reversal of the relative importances of the inhibitions produced in LP by the pacemaker neurones PD and by the constrictor neurones PY. Functionally, this allows LP to maintain or increase its oscillatory activity, yet to modify substantially the phase relationships between its discharges and the discharges of the pacemaker neurones (see below).

Another synaptic relationship which is functionally important in determining the pyloric rhythm is the chemically mediated synaptic inhibition exerted by the pacemaker neurones (PD) upon the constrictor neurones PY. Again the apparent efficacy of the synapse can be characterized by the membrane potential of PY at the maximum of the inhibition. This corresponds to the troughs of the PY oscillations, which were shown in Fig. 7B,₃ (T) to become less negative after an APM discharge. At the same time, the spike frequency of the presynaptic neurone (PD) increases (see Fig. 8C, PD). Thus, the efficacy of the inhibition of PY by PD decreases in a manner analogous to that of the inhibition of LP by PD (compare Fig. 7B ₃, T and 8C, LP).

Overall, an APM discharge is accompanied by a marked decrease in the apparent efficacy of those synapses which are functionally of greatest importance in the pyloric network, the inhibitory synapses of the pacemakers on the constrictor motoneurones. This results in changes in phase relationships between the discharges of the various neurones (Fig. 8E, F). The phases of discharge of the constrictor neurones (LP and PY) in the pacemaker cycle are significantly advanced when APM fires (see legend of Fig. 8E, F for explanation of phase relationships). This modification of the phase relationships induced by APM discharge can lead to the suppression (Fig. 8F, arrow) of the characteristic silent period which usually occurs between the firing of the antagonistic neurones PD and LP (Fig. 8E, t; see also extracellular record in Fig. ID, before APM firing).

In addition, the APM discharge is accompanied by an increase in the efficacy of another inhibitory synapse, that of one constrictor (PY) on another (LP). This modification can be explained by the strong activation (and resultant increased spike frequency) of the presynaptic element (PY). In contrast, the decreased synaptic efficacies appear paradoxical because they occur when spike frequency (and amplitude of oscillation) in the presynaptic neurone (PD) increases. We will show in the following paper (Dickinson & Nagy, 1983) that this can be explained by modifications of the membrane properties of the postsynaptic neurones.

Activity in APM inverts the relative efficacy of two synapses of the same type (inhibitory chemical) that a single neurone (LP) receives from two others (PD and PY). Firing in APM can likewise invert the relative importance of two types of synapse (electrical and inhibitory chemical) which exist simultaneously between two neurones.

Within the set of dilator neurones such a double synaptic relation exists between the pacemaker neurone (PD) and the ventral dilator neurone (VD) (Maynard, 1972; Maynard & Selverston, 1975; Fig. 9A). They are linked by an electrical synapse, and the pacemaker neurones inhibit neurone VD by a chemical synapse. The electrical synapse between PD and VD tends to synchronize the cyclically recurring depolarizations of the two neurones (Fig. 9C). However, at the peak of firing, chemical inhibition from PD causes VD to stop firing, thus maintaining a difference in phase between these two neurones. The exact phase relationship of PD and VD discharge at any time is determined by the dynamic balance between the effects of the electrical synapse and those of the inhibitory chemical synapse. When APM fires, the efficacy of the inhibitory synapse from PD onto VD decreases to almost zero and the bursts inPD and VD become synchronous (Fig. 9D). Thus, activity in APM results in a shift of the balance between the two synapse types in favour of the electrical synapse. This effect continues for some time after the end of the APM discharge (Fig. 9B), and then progressively recovers. During the recovery, intermediate cases in which VD fires double bursts can be seen (Fig. 9E). In these cases, at the peak of its discharge PD can briefly inhibit VD (via the chemical synapse), but after that the electrical synapse (which depolarizes VD) predominates again, and the burst of VD resumes. The decreased efficacy of the chemical synapse cannot be accounted for by changes in firing of the presynaptic element, for both spike frequency and amplitude of oscillation of PD increase rather than decrease during and after an APM discharge.

In summary, it appears that APM can modify the phase relationships between the discharges of the pacemaker neurones and VD by shifting the dynamic balance which exists between the effects of an electrical and an inhibitory chemical synapse.

Firing of the APM neurone engenders major and lasting modifications of the activity of the pyloric neurones. The expression of these modifications is in part dependent upon the previous activity level of the pyloric network, as discussed earlier, but is also dependent upon the intensity of the activity in APM (Fig. 10). Increasing spike frequency in APM does not result in a simple increase in activity of the pyloric neurones, but rather modifies, to differing extents, the various characteristics of the discharges of the pyloric neurones (amplitude of oscillation, spike frequency, frequency of the rhythm, etc.). This more or less alters the pyloric rhythm. As shown in Fig. 10B, firing of APM at 2-5 Hz causes increased spike frequency and amplitude of oscillation in the PY neurone (compare to Fig. 10A), but has little effect on the overall frequency of the pyloric rhythm (see the extracellular recording of pacemaker neurone PD in the dlvn, Fig. 10B). When APM fires at 6 Hz, however, the rhythm is substantially accelerated (Fig. 10C). Finally, after a 20 Hz discharge of APM (Fig. 10D), the pyloric rhythm first accelerates, then stops and the PY neurone fires in long plateaus. Thus, in the example shown, activity in APM results in a slow but intense rhythm, a rapid and intense rhythm and an intense non-rhythmic activity of the pyloric neurones, respectively, as firing frequency in APM is increased. These effects have been observed repeatably throughout the same experiment, and from one preparation to another.

In this example, APM spike frequencies as low as 2·5 Hz produce distinct modifications of the pyloric pattern (compare Fig. 10B and A). It is interesting to note that APM will, at least in isolated preparations, fire spontaneously at frequencies equal to or greater than 2·5 Hz. One such example is shown in Fig. 10E, in which APM fires in irregular bursts. We have frequently recorded such bursting patterns of activity in both Jasus and Palinurus. However this rhythm is variable from one animal to another and thus far we have not been able to correlate it with any other rhythm or to identify any sources of afferent input which might trigger this activity in APM.

We have shown that APM modifies the expression of the motor pattern produced by the whole pyloric network. These effects are established slowly (seconds) and persist longer (five to ten times) than the duration of the APM discharge provoking them. In addition, as is shown in the following paper (Dickinson & Nagy, 1983), APM modifies voltage-dependent properties of its postsynaptic neurones. This suggests that APM does not exert classical synaptic effects, having short actions (milliseconds), but is instead modulatory in nature (see Discussion). It was therefore of particular interest to determine the kind of receptors (and the neural transmitter) involved. The pharmacological tests presented below suggest that APM acts via muscarinic cholinergic receptors.

The effects of APM on the pyloric neurones are blocked by atropine, a cholinergic antagonist which is known to block muscarinic receptors in vertebrates. When the stomatogastric ganglion is bathed in 5 X 10⁻⁴M-atropine (in saline), the effects of APM are reversibly blocked (Fig. 11B) (in 2·5 X 10⁻⁴M-atropine, the effects of APM are greatly decreased). In contrast, solutions of d-tubocurarine and hexaméthonium (cholinergic antagonists which block nicotinic receptors) do not block the effects of APM, even at concentrations as high as 10⁻³M (Fig. 11C, D). Note that atropine, although it completely blocks APM’s effects, does not modify synaptic activity within the pyloric network, as shown by the fact that PD still inhibits the constrictor neurones (Fig. 1 IB, PY). The effects of atropine are completely reversible, as seen in Fig. 11E in which the effects of APM, blocked by 10⁻³ M-atropine (Fig. 1lE,₁), returned after 1 h of perfusion with normal saline (Fig. E2).

That the effects of APM on pyloric neurones involve cholinergic muscarinic receptors could also be shown using muscarinic agonists. Perfusion of the deafferented stomatogastric ganglion with muscarinic agonists (pilocarpine, oxotremorine) has been shown to induce rhythmic activity in previously silent pyloric pacemaker neurones (Marder & Paupardin-Tritsch, 1978; Anderson, 1980). This activation resembles to some extent the activation induced by APM. However, other afferent fibres to the stomatogastric ganglion, known to activate the pacemaker neurones, are also thought to be cholinergic (Russell & Hartline, 1981; Sigvardt & Mulloney, 1982) thus rendering the actions of muscarinic agonists difficult to interpret.

The activity of a single identified interneurone, the anterior pyloric modulator (APM), in the oesophageal ganglion of the rock lobsters Jasus lalandii andPalinurus vulgaris, can modify the output of the pyloric pattern generator. Thus a single neurone can alter a rhythmic motor behaviour. These extensive effects appear to be characteristic of a neuromodulatory process.

It seems that APM directly activates each of the neurones of the pyloric network. Its two axons project directly into the stomatogastric ganglion. Although these axons pass through the two commissural ganglia and may make synaptic connections there, such connections are not necessary for the activation of the pyloric neurones. This is indicated by the fact that APM activates the pyloric neurones even when synaptic activity in the commissural ganglia is blocked. Moreover, all of the 30 neurones in the stomatogastric ganglion have been studied (Selverston et al. 1976) and it is known that there are no local interneurones presynaptic to the neurones of the pyloric network. Thus the effects of APM on pyloric activity cannot involve a local interneurone in the stomatogastric ganglion, and a direct action of APM on the pyloric neurones is a possibility that must be considered. This is also suggested by the observation that APM’s effects involve cholinergic muscarinic receptors and that the existence of such receptors has been proposed at least for the pacemaker neurones PD (Marder & Paupardin-Tritsch, 1978). In addition, some recent experiments on Palinurus vul-pris (unpublished) suggest the presence of monosynaptic connections between APM and the PY neurones. Excitatory postsynaptic potentials (epsps) were correlated one to-one with spikes in APM. These epsps, in contrast to APM’s modulatory effects, are sensitive to and can be reversibly blocked by d-tubocurarine (a nicotinic cholinergic blocker), indicating that APM is cholinergic. The simplest explanation for these results is that APM acts directly upon nicotinic receptors on the pyloric neurones to produce epsps, and that it simultaneously acts upon muscarinic receptors on the same neurones to produce the modulatory effects we have described in this paper.

Although we have not been able to exclude the possibility that APM acts presynap-tically in the stomatogastric ganglion upon another fibre afferent to this ganglion (and to the pyloric network), the available evidence suggests a direct action of APM’s axons on the pyloric neurones. We may now ask whether APM has synapses upon each of the pyloric neurones or whether it directly affects only certain of these neurones. Most of the synaptic connections within the pyloric network are blocked by either curare or picrotoxin (Marder & Paupardin-Tritsch, 1978; Bidaut, 1980). However, APM is able to influence all the pyloric neurones in the presence of either drug (see Dickinson & Nagy, 1983 for APM’s effects in the presence of picrotoxin). Thus, APM must be directly and separately activating each of the pyloric neurones.

An APM discharge influences all the neurones of the pyloric pattern generator, thus modifying the output of the network and changing the entire pyloric motor pattern. Because the muscular activity controlled by the pyloric neurones is well known (Hartline & Maynard, 1975), it is possible to envisage a number of behavioural implications of this modulation by APM. The most common effect of an APM discharge is an increased activity in the pyloric neurones (seen as an increase in spike frequency within bursts), which would be likely to increase the strength of muscular contractions. This activation is particularly strong in the constrictor motoneurones, which control the muscles performing the power stroke of the pyloric movements.

More far-reaching functional consequences of an APM discharge could result from the qualitative alteration of the pyloric pattern that is produced by such a discharge. The motor pattern generated by the pyloric network is determined both by the oscillatory activity of the pacemaker neurones (Fig. 12A, PD-AB) and by the reciprocal inhibitory synapses within the network, the most important being those of the pacemakers onto the constrictors (Selverston et al. 1976). We have shown firstly that APM decreases the functional importance of these synapses relative to others in the network (Fig. 12C), and secondly that it can alter the oscillatory activity of the pacemaker neurones, either by transforming the pyloric pattern from a rhythmic to a non-rhythmic one, or by increasing the frequency of the pyloric rhythm. These effects of APM can influence three aspects of the pyloric behavioural sequence. The first of these is the phase relationship between the contractions of the different pyloric muscles. The decreased efficacy of the synaptic input from pacemaker to constrictor neurones after firing of APM leads to phase shifts of the discharges of pyloric neurones relative to each other. Therefore APM activity must lead to the same phase shifts between the contractions of the pyloric muscles, particularly the antagonistic dilator and constrictor muscles.

The second modification which can be induced by APM is a temporary suppression of rhythmic behaviour and its replacement by strong firing in plateaus of the pyloric neurones, especially of the constrictor neurones (see Fig. 10D). This represents a profound alteration of the pyloric motor pattern. In the intact animal, rhythmic activity in the pyloric stomach would cease, and the strong and long-lasting contractions of the constrictor muscles could then push food out of the pyloric stomach.

The third type of modification which can be provoked by APM is an acceleration of a previously slow pyloric rhythm to a maximum frequency of about 1 Hz, an acceleration which is accompanied by recruitment of previously silent neurones. This corresponds to a shift from a slow, irregular rhythm to a fast, regular pyloric rhythm, which in the intact animal has been observed to occur abruptly when feeding starts, the fast rhythm continuing for several hours after the end of a meal (Rezer & Moulins, 1980). The long duration of the effects of APM suggests that it might be involved in this transition between the two kinds of pyloric rhythm during feeding.

APM’s action on the pyloric activity is typical of a neuromodulatory action (see Kandel, 1976; Kandel, Krasne, Strumwasser & Truman, 1979; Dismukes, 1979; Kupfermann, 1979; Daly, Hoffer & Dismukes, 1980; review by Hartzell, 1981). The characteristics are (1) membrane properties of a relatively large number of neurones are modified; we have shown that APM influences all the neurones of the pyloric network: (2) effects are established slowly and have a long duration (tens of seconds to hours); we have seen that the modifications provoked by APM are first visible above 1 s after the start of APM firing, that they peak 3−5 s later (although APM’s spikes reach the stomatogastric ganglion in about 100 ms), and that they last eight to ten times longer than the discharge which provoked them; (3) effects are dependent on the voltage of the postsynaptic elements; all of the modifications of the pyloric pattern provoked by APM can be explained by voltage-dependent modifications of membrane properties of the pyloric neurones (Dickinson & Nagy, 1983); (4) the action modifies the effects of other synaptic inputs; we have seen that APM’s discharge drastically alters the relative efficacy of all the synaptic relations within the pyloric network. Therefore, the influence exerted by APM on the pyloric motor pattern fulfills the main criteria for a neuromodulatory process.

We have seen that APM can profoundly change the expression of the pyloric activity in several ways, and that some of these changes resemble the modifications of the pyloric motor pattern which correspond in intact animals to the initiation of a behaviour adapted for feeding. Therefore, one might suggest that APM could contribute to the triggering of this pyloric pattern associated with feeding. APM might thus be an example of the theoretical concept of a modulatory element integrated into a command system (see Kupfermann & Weiss, 1978). A possible role for APM in a command system is suggested by the observation that it modifies the membrane properties of pyloric neurones and thus changes their sensitivity to synaptic inputs (Dickinson & Nagy, 1983). It is therefore possible that APM conditions the pyloric neurones to respond to input from other elements in a command system.

A number of studies have suggested that it is the receptors involved rather than the nature of the transmitter that determine a modulatory synaptic effect (see review by Dismukes, 1979). Thus, a transmitter used as a briefly acting neuromediator at a given synapse may also be involved in neuromodulation at another or even at the same synapse.

In the stomatogastric ganglion of lobsters, a number of substances may be involved in chemical synaptic transmission, including acetylcholine, GABA and glutamate, which are putative transmitters at synapses within the pyloric network (Marder & Paupardin-Tritsch, 1978; Bidaut, 1980). Others, such as dopamine (Kushner & Maynard, 1977; Raper, 1979; Anderson & Barker, 1981) and octopamine (Barker, Kushner & Hooper, 1979), are thought to mediate transmission from higher order nervous centres. Our experiments suggest that the modulation of pyloric activity provoked by APM is mediated by acetylcholine acting on muscarinic receptors. It is interesting to note that all of the known slow potentials mediated by acetylcholine utilize muscarinic receptors (e.g. parasympathetic ganglion cells in the mudpuppy, Hartzell, Kuffler, Stickgold & Yoshikami, 1977; sympathetic neurones in the toad, Schulman & Weight, 1976; Weight, Schulman, Smith & Busis, 1979 and in the rabbit, Dun, Kaibara & Karczmar, 1978).

In addition, studies of several non-neuronal cell types have shown that the activation of muscarinic receptors can cause modifications of regenerative membrane properties; thus, the plateau phase of cardiac action potentials in vertebrates is decreased (Giles & Noble, 1976; Ten Eick, Nawrath, McDonald &Trautwein, 1976 curves, 1976), slow waves in intestinal muscles of mammals are increased (Bolton, 1971), and the plateau phase of activity in pancreatic endocrine cells is increased (Gagerman, Idahl, Meissner & Taljedal, 1978). In the pyloric network, activation of muscarinic receptors on the PD pacemaker neurones with muscarinic agonists (pilocarpine, oxotremorine) induces bursting activity in these neurones (Marder & Paupardin-Tritsch, 1978; Anderson, 1980). Moreover, electrical stimulation of the inferior ventricular nerve induces a long-term enhancement of bursting in the pacemaker neurones, and this effect might be mediated by muscarinic receptors (Russell & Hartline, 1981). But, before the present study, nothing was known about modifications of the activity of the non-pacemaker neurones. The long-term modulation by an APM discharge, which most probably involves muscarinic receptors, will be shown in the next paper (Dickinson & Nagy, 1983) to be due essentially to induction and amplification of membrane properties underlying the burstiness in all of the pyloric neurones.

We are grateful to Professor M. Moulins for helpful criticisms of the manuscript. This work was supported by the DGRST (Grants 80P6049 and 81E0545) and by a CNRS/NSF exchange fellowship.