Central Generation of Bursting in the Feeding System of the Snail, Lymnaea Stagnalis

Invertebrate ganglia have proved useful for the study of integrative mechanisms not only because they contain small numbers of relatively large neurones, but also because many ganglia produce cyclical patterns of burst activity in which component bursts fire in a definite temporal relationship to each other. In motor systems such as those controlling leech swimming (Friesen, Poon & Stent, 1976), Tritonia swimming (Getting, 1977) and Helisoma feeding (Kaneko, Merickel & Kater, 1978), the atual pattern generating mechanism is a property of a network of interneurones, which are often small cells, and these cells impose the rhythm on the motoneurones. This further simplifies analysis by dividing the population into two types with clearly distinguishable properties.

In the leech swimming system the rhythm is generated through cyclic inhibition between four interneurones per segment (Friesen et al. 1976) whereas in Helisoma there is positive feedback within small networks of electrotonically coupled interneurones (Kaneko et al. 1978). In Helisoma the resultant synchronous burst activity in the interneuronal networks produces inhibition in some cells (the protractor motoneurones) and excitation in others (the retractor motoneurones), the overall effect being to produce a protraction-retraction cycle of muscular activity controlling feeding movements.

We have examined the feeding system of Lymnaea, a freshwater pulmonate snail. The novel feature of this system is that there are two phases of synaptic input to the motoneurones, implying the existence of two interneuronal networks firing consecutively. The burst pattern is consequently much more complex than the simple alternation of bursts found in Helisoma, there being at least eight different patterns of bursting in the motoneurones. The situation in which two networks could be interacting may be of some importance when it is considered that many models of rhythm generation in mammals are based on the idea of synaptic interactions between small populations of neurones (e.g. cortico-thalamic system, Andersen & Eccles, 1962; respiratory system, Cohen, 1968) and in Lymnaea such interactions could be studied in identified groups of neurones.

In this paper we give details of the central motor pattern for feeding in Lymnaea. The study involved intracellular recording from identified neurones in the buccal ganglia. In the two following papers we will present anatomical and electrophysiological evidence which shows that many of the cells described in the present paper are motoneurones in the feeding system (Benjamin, Rose, Slade & Lacy, 1979; Rose & Benjamin, 1979). The relationship between the neural pattern and the activity of the buccal musculature leading to food ingestion will be explained in the last paper (Rose & Benjamin, 1979).

Lymnaea stagnalis were obtained from animal suppliers, kept in aerated Brighton tapwater and fed on lettuce for short periods before use. Experiments were carried out on isolated brains in blood, just previously collected from the intact snail, or in Lymnaea bicarbonate buffered saline (Winlow & Benjamin, 1976) or most recently in Hepes buffered saline (Benjamin, 1978). The preparations most often used for electrophysiological recording consisted of the brain plus paired buccal ganglia with the cerebrobuccal connectives intact. Part of the pro-oesophagus and salivary gland ducts remained attached to the buccal ganglia by the dorsobuccal nerves. A semi-intact preparation was used in a few experiments to compare bursting in isolated and feeding preparations and this will be described in the accompanying paper (Rose & Benjamin, 1979). In all experiments the brain was treated for short periods with a proteolytic enzyme (Protease Type V, Sigma, London) which, if used carefully, aided penetration of buccal ganglion neurones without affecting the electrical properties of the neurones or their synaptic connections.

Intracellular recordings were made from the cell bodies of identified buccal neurones, usually two at a time, using pairs of independently manipulated glass microelectrodes filled by the glass-fibre method with the supernatant from a saturated solution of K₂SO₄. In a few experiments three or four neurones were recorded simultaneously. Electrode resistances were in the 20–80 MΩ range. Signals were fed into Bioelectric Instruments Pi amplifiers (Farmingdale, U.S.A.) which allowed negative capacitance compensation and electrode resistance balancing, so that current could be passed through the recording electrode whilst recording resulting changes in membrane potential in a reasonably accurate manner. Potentials were displayed on a storage oscilloscope and permanently recorded on film or pen-recorder.

The paired buccal ganglia are connected to the rest of the central nervous system by long cerebrobuccal connectives, whilst the buccal ganglia themselves are joined by a short buccal commissure (Fig. 1). Nerves from the buccal ganglia innervate muscles forming the buccal mass, the gut and the salivary glands (Carriker, 1946). Details of nervous innervation and arrangement of the muscles of the buccal mass will be given later (Benjamin et al, 1979; Rose & Benjamin, 1979).

Neurones which can be identified as individuals form eight pairs of bilaterally symmetrical cells whose cell body locations are shown in Fig. 1. They were arbitrarily numbered so that each member of a pair had the same number with the prefix letter B to indicate buccal ganglion and then R (right) and L (left) to indicate which ganglion the cell body was in. The cells were thus labelled BR₁, BR₂… BR8 and BL₁, BL₂… BL8. This simple scheme is complicated by the fact that some of these cells occur in groups of two to seven cells having similar properties. Thus there are at least two cells of the 5, 6 and 7 type in each ganglion (only one of each of these cells is shown in Fig. 1). The most prominent grouping occurs with respect to the 4 cells. Each of the large 4 cells (BL₄ and BR₄) are surrounded by a cluster of medium-sized cells having similar electrical properties to each other and to the large 4 cell with which they are associated. We refer to the whole group on each side as the ‘4 group’, this consisting of the 4 cell (BR₄ and BL₄) and surrounding BL₄ cluster cells and BR₄ cluster cells. The 4-group cells are situated on the dorso-posterior surface of each buccal ganglion (Fig. 1). The exact number of BL₄ cluster cells and BR₄ cluster cells is not known but the maximum number so far recorded in each cluster is six.

In the description that follows we have found it convenient to abbreviate the above terminology when referring to neurones with similar electrical properties. For instance the term ‘4 cell’ refers to either BL₄ or BR₄, ‘4 cluster cell’ refers to any cell from the BR₄ cluster or BL₄ cluster, ‘7 cell’ refers to either BL₇ or BR₇, and so on, for all the bilaterally symmetrical pairs of neurones. The same convention will be used in The accompanying papers (Benjamin et al. 1979 ; Rose & Benjamin, 1979).

The largest neurones shown in Fig. 1 can be recognized with certainty before electrode penetration because of their relatively constant cell body locations on the ganglion surface. Thus the 1, 2, 3 and 4 cells can be identified on the basis of their size and location and also because the 2 and 3 cells are often paler orange than the i and 4 cells. Although the 5, 6, 7 and 8 cells are usually found in the positions indicated in Fig. 1, their exact locations are not constant from preparation to preparation, and identification of cell type has to be based on recognition of the synaptic inputs following impalement.

The synaptic inputs and spike activity recorded from buccal neurones in Lymnaea did not result from electrical stimulation of nerves, and did not occur as a result of sensory activity evoked in the periphery as the result of muscular contraction. The activity we described here occurred in isolated preparations of buccal ganglia and brain, but the same patterns of activity occurred in semi-isolated preparations (Rose & Benjamin, 1979) where feeding behaviour could be observed. The effectiveness of synaptic inputs in evoking spike activity varied considerably both between preparations and in the same preparation at different times. Obvious phase locking of burst activity in buccal neurones was seen when the inputs were large and regular in occurrence. From work on semi-intact preparations we know that phase related burst activity of this type leads to feeding (Rose & Benjamin, 1979). We have termed such isolated preparations where burst activity is so dramatically timed, ‘feeding preparations’. In ‘non-feeding’ preparations, synaptic inputs of a similar type were present but potentials were small in amplitude, irregular, and of low frequency.

All the cells we have identified in the buccal ganglia of Lymnaea show burst activity which can be related to the occurrence of synaptic inputs, although the endogenous properties of the neurones are also of great importance. We have found it convenient to divide the identified buccal neurones into two types depending on whether the cells receive one or two phases of synaptic input per cycle of bursting. If a cell receives a single synaptic input per cycle then it always occurs in the first phase of the double input cells.

The single and double synaptic inputs produce a complex pattern of spike activity in the eight types of identified neurone. An overall impression of the temporal relations of buccal cell bursting can be obtained by reference to Fig. 14.

It is convenient to discuss 4-group cells first because these cells are easily recorded and the synaptic inputs of other buccal neurones can readily be compared with them.

The 4-group cells receive a simultaneous inhibitory input and when this input ceases the cells burst synchronously (Figs. 2 a, 3 a, c, e), although the largest 4-group cells (BL₄ and BR₄) may start bursting after the 4 cluster cells (Fig. 3 a). This synchronous bursting appears to involve two processes - rebound from inhibition and electrical connections within the group. We will discuss these two points in turn.

The inhibitory post-synaptic potential (i.p.s.p.) which precedes bursts in 4-group cells is a compound summating potential which has a similar time of onset and form in any pair of 4-group cells whether recorded in the same or opposite buccal ganglia. This is shown for cells in opposite ganglia (4 cells, Fig. 2a, 4 cluster cells, figure 3e) and for cells in the same ganglia (Fig. 3 c). Trains of unitary i.p.s.p.s often precede the compound summating i.p.s.p. in 4-group cells (Fig. 2b, 4 cells; Fig. 3 b, 4 cluster cells). The function of this type of inhibitory input is unknown, and will not be referred to again. In feeding preparations where 4-group cells are regularly bursting, the i.p.s.p.s preceding the burst may cause little or no change in membrane potential (Fig. 3 e) or may even be reversed-to yield a depolarizing potential (Fig. 3 c, bottom trace). This apparently occurs because the large hyperpolarization following bursts in some cells drives the membrane potential beyond the reversal potential of the i.p.s.p. occurring before the next burst. It was interesting that the i.p.s.p. of the 4 cluster cell of figure 8 c, d was hyperpolarizing in periods when the preparation showed little bursting activity (Fig. 8 c) but reversed to a depolarizing potential when the cell was bursting in a precisely phasic manner with another bursting neurone (BL₃) (Fig. 8c). The fact that the i.p.s.p.s of 4-group cells can be reversed suggests that they are mediated by chemical synaptic transmitters.

We think that the bursts of spikes in 4-group cells result from post-inhibitory rebound following the i.p.s.p.s, although we cannot disprove that some component of the depolarisation during the burst is due to the activity of an excitatory interneurone. If the i.p.s.p. was small or absent, no bursting occurs. The i.p.s.p.s do not inhibit on-going spike activity because the previous burst terminates before the onset of the next i.p.s.p. (Fig. 1a). In strongly bursting 4-group cells, a rapid depolarization follows the i.p.s.p.s, and this reaches a peak and slowly declines (Figs, 2a, 3c, d). The burst is often terminated by a rapid post-burst hyperpolarization (Figs, 2a, 3c), the degree of hyperpolarization varying considerably even in cells from the same preparation (compare top and bottom traces of Fig. 3e). The ability of i.p.s.p.s to initiate bursts also varies in different cells, so that a burst will follow an i.p.s.p. in one cell but not inevitably in another (compare top and bottom traces in Fig. 3b, d), 4 cells being less likely to fire than 4 cluster cells (Fig. 3d). 4 cells were also different from 4 cluster cells in that they tended to fire with greater delay after the end of an i.p.s.p. (Fig. 3a) and so the burst started later and finished later than 4 cluster cells. In general the 4 cells, the largest 4-group neurones, were the least excitable of the 4-group neurones and this is interesting because it fits with results from vertebrate motoneurones whose readiness to fire is also a function of size (Henneman, Somjen & Carpenter, 1965).

Thus the main factor responsible for synchronizing the bursting activity of 4-group cells is common i.p.s.p. input mediated by chemical transmitters. The other factor is the electrotonic coupling that occurs between 4-group cells. Application of long square pulses of depolarizing or hyperpolarizing current to one 4-group cell results in an attentuated but similar response in another randomly selected neurone in the same (Fig. 3f) or opposite ganglion (Fig. 2c), with no obvious delay in the onset of the post-synaptic response compared with that in the current-injected cell. Also spike activity in one 4-group cell frequently resulted in small depolarising potentials in another cell (Fig. 2e). Many such potentials could be seen during the burst of a weakly bursting neurone (Fig. 2b) or if 4-group cells were hyperpolarized to prevent bursting (Fig. 2d). It is noticeable that few electrotonic e.p.s.p.s. could be seen in non-feeding preparations when 4 cells, in particular, may not be bursting (bottom trace of Fig. 3 d). It seems likely, at least during the burst, that electrotonic junctions may play a role in synchronizing the activity of 4-group cells in that a component of the burst depolarization comes from the spike activity in other 4-group cells. Whether the junctions are important in other parts of the burst cycle is more problematical, particularly as many of the junctions between 4-group cells are of very low efficacy.

Several factors affect the measurement of the efficacy of 4-group electrotonic junctions. A true value of the transfer of potential from one cell to the other at the synapse is impossible to obtain because it is likely that the site of the junctions is some distance from the recording site in the cell bodies (see Benjamin et al 1979). Also values of even an apparent coupling coefficient are difficult to obtain because of the input resistance fluctuations due to synaptic input and spike activity. The highest values of coupling ratio were obtained in the absence of i.p.s.p. input and thus spike activity. Under these conditions a value of d.c. coupling ratio (ratio of steady potential in the post-synaptic cell to that in the presynaptic cell) of up to 0·15 was obtained for ipsilateral cells. This value was obtained for the pair of cells shown in Fig. 3f. It can be seen that the value of the coupling ratio for these cells is the same for both depolarizing and hyperpolarizing current pulses of the same amplitude, which indicates a nonrectifying electrotonic synapse. Similar results were obtained for pairs of 4-group cells in opposite ganglia (Fig. 2c), except that in this case the coupling ratio was lower, as might be expected in consideration of the greater conduction distances for subthreshold potentials, between these cells.

As well as differences in coupling ratio between 4-group cells in different locations, values for particular cell couplings varied among preparations. Thus the d.c. value of coupling ratio for BL₄ and BR₄ varied in value from 0·03 to 0·01 in different preparations, and this reflected the situation in the whole 4-group network. When junctional efficacy of BL₄ and BR₄ was high, electrotonic junctions could be demonstrated between any randomly selected pair of ipsilateral or contralateral 4-group cells (although the coupling ratio for small cells recorded in opposite ganglia could be very difficult to distinguish against the noise level). When coupling ratios were low between BL₄ and BR₄ it was difficult to find electrotonic junctions between all cells located in different ganglia. The low efficacy of junctions in some preparations may result from an uncoupling due to physical stretch of the ganglia when they are pinned out.

Arguments about the role of electrotonic junctions in burst synchrony apply to some other buccal neurones which also have common synaptic inputs and are electrotonically coupled.

In summary, it seems that the chemically mediated i.p.s.p.s are the chief factor responsible for initiating bursts in 4-group cells. Electrotonic junctions may be important in reinforcing the synchronising effect of common i.p.s.p. inputs especially during the burst. Lastly, the endogenous properties of 4-group cells make them liable to rebound from inhibitory inputs (unlike other buccal cells, see section below for 5 cells) and may be important in burst termination, as there is no obvious synaptic input that serves this function.

It will be seen in the following sections that the synaptic inputs of other buccal neurones occur in strict temporal relationship to the i.p.s.p.s of 4-group cells, and that these inputs are the main determinant of burst patterning in the buccal ganglia.

These large neurones receive simultaneous excitatory input (Fig. 4b) which results in common patterns of firing (Fig. 4a). Like 4-group cells, BL₁ and BR₁ are electrotonically connected (Fig. 4d) and this may be an additional factor in synchronizing 1-cell firing. The excitatory inputs of 1 cells occur at the same time as the i.p.s.p.s of 4-group cells (Fig. 5 a) and so bursts in 1 cells are followed by bursts in 4-group cells in feeding preparations (Fig. 5 c).

The periodic excitation of 1 cells results from summating compound excitatory post-synaptic potentials (e.p.s.p.s) with similar time of onset and form in pairs of recorded 1 cells (Fig. 4b). Details of membrane potential fluctuations within the compound e.p.s.p.s are very similar (Fig. 4c), which suggests that BL₁ and BR₁ must be driven by the same interneurone or by more than one tightly coupled cell. A second type of excitatory input occurs on 1 cells and consists of low level continuous e.p.s.p.s which can be observed interspersed between the larger compound e.p.s.p.s (Figs. 4b, 5 b). Recent work in this laboratory (C. McCrohan, unpublished results) shows that these low amplitude e.p.s.p.s result from the activity of an interneurone in the cerebral ganglion of Lymnaea. No spike activity results from this second type of input.

The electrotonic junction between 1 cells was demonstrated by passing.square pulses of depolarizing and hyperpolarizing current into one cell of the pair and recording a similar but attenuated response in the other cell (Fig. 4d). The electrotonic junction connecting 1 cells appears to be non-rectifying (Fig. 4.d) and of low efficacy (coupling ratio of BL₁ and BR₁ junction is 0-05 in Fig. 4d), although one has to take into account that the junction between these cells is liable to be some distance from the recording sites in the cell bodies (see the anatomical studies of Benjamin et al. 1979). It is obvious from the recording of Fig. 4b that a spike in BL₁ does not result in a corresponding spike in BR₁ even at the peak of the compound e.p.s.p input. Overall, it seems that the electrotonic junction connecting the 1 cells cannot be the most important factor determining spike synchrony compared with the occurrence of common e.p.s.p.s.

Endogenous factors appear to be important in determining the response of 1 cell to e.p.s.p. input. Even in feeding preparations, 1 cells rarely fire more than a few spikes in a burst (Figs. 5b, 8b, 13c) and this seems to be due to the adaptive properties of 1 cells. The e.p.s.p.s of the 1 cell in Fig. 5b show a double peak with the 1 cell firing a spike at the top of the rising phase of each of these peaks but not during maintained apparently suprathreshold depolarisation. In general, the 1 cells show a tendency to fire during or at the peak of the rising phase of the e.p.s.p., but their spike generating mechanism adapts to subsequent maintained depolarizations.

The temporal relationship of 1 cell and 4-group cells firing was examined by paired recordings. Hyperpolarizing both cell types to prevent firing shows that the i.p.s.p.s of 4-group cells occur at the same time as the e.p.s.p.s of 1 cells (Fig. 5 a). Details of post-synaptic potential form in both cell types are related, so that the double depolarizing peaks of the BRi of Fig. 5b are matched by similar hyperpolarizing peaks in the BR₄. This results in spikes in 1 cells which are followed by rebound spikes in 4-group cells (Fig. 5b). Long-term recordings of 1 cells and 4-group cells in feeding preparations show the bursts of 1 cells are regularly followed by bursts in 4-group cells (Fig. 5 c). There is usually a gap between the end of a 4-group burst and the beginning of the next burst of 1 cell activity (Fig. 5 b, c). No connexions have been found between 1 cells and 4-group cells, and it can be seen from Fig. 5 a that inhibition of 1 cells does not prevent the occurrence of large regular i.p.s.p.s in 4-group cells.

6 cells are small and difficult to find. They rarely show bursting activity in isolated buccal ganglion preparations, and it was necessary to use semi-intact feeding preparations to record large burst-generating synaptic inputs. In consequence, we have concentrated our efforts on recording 6 cells with 4-group cells to see how they fire in relation to other buccal neurones. We have not looked for synaptic connexions between 6 cells themselves.

Fig. 6 shows a 6 cell recorded with a 4 cell in an isolated (Fig. 6b), and a semiintact preparation (Fig. 6a). It can be seen that the i.p.s.p.s of 4 cells occur at the same time as summating compound e.p.s.p.s in 6 cells. Only in the semi-intact preparations are the e.p.s.p.s large enough to generate bursts of activity. In this case bursts of spikes in the 6 cell were followed by bursts in the 4 cell (Fig. 6a). Thus the 6 cells fire at the same time as 1 cells although the number of spikes in a 6-cell burst is much higher than that of a 1 cell.

The description so far has revealed a system in Lymnaea which is similar to that found by Kater (1974) in the buccal ganglion of Helisoma. Thus the 4-group cells of Lymnaea have similar inputs and firing patterns to the protractor motoneurones, the 6 cells to the retractor motoneurones, and the 1 cells to the salivary gland motoneurones of Helisoma. Nevertheless, a fundamental disagreement arises at this point because we assign the opposite functions to the 4-group cells and 6 cells ; that is, we define 4-group cells as retractor motoneurones and 6 cells as protractor motoneurones. The third paper in this series (Rose & Benjamin, 1979) will provide evidence for this statement. On the other hand, the 1 cells in Lymnaea have axonal projections to the salivary gland ducts (Benjamin et al. 1979), which suggests that they are homologous to Kater’s salivary-gland motoneurones. A summary of the single input system is shown in the top three traces of Fig. 14, and this can be compared with Kater’s Fig. 23 (Kater, 1974).

None of the cell types we describe in the rest of the results as double input cells have been described in Helisoma.

3 cells are large neurones which were easy to record in the isolated buccal ganglia. Like other pairs of bilaterally symmetrical neurones in the buccal ganglia of Lymnaea they receive common synaptic inputs (Fig. 7 a, this can be compared wib) and are electrotonically coupled (Fig. 7c). As a result, 3 cells fire synchronously (Fig. 7b). 3 cells receive common inhibitory inputs (Fig. 7a), which are followed by powerful depolarizing waves upon which action potentials are superimposed (Fig. 7b). We will argue that the depolarizing wave is due to interneuronal activity, but cannot rule out the possibility of rebound excitation as part of the mechanism for burst generation. The inhibitory input to 3 cells occur at the same time as the inhibitory input to 4 group cells (Fig. 8 c) and the excitation on 1 cells (Fig. 8a). Thus bursts of spikes in 1 cells (and thus 6 cells) are followed by those of 3 cells (Fig. 8b) which themselves are synchronous with those of 4-group neurones (Fig. 8d).

Long-duration compound i.p.s.p.s precede 3-cell bursts. We were impressed by the similarity of latency and shape of i.p.s.p.s in pairs of recorded 3 cells and 4-group cells (Fig. 8,c), and at first thought that the rapid depolarization following inhibition of 3 cells was due to rebound from inhibition as with 4-group cells. However, two pieces of evidence support the idea that a second input is mainly responsible for the depolai izing wave in 3 cells. Firstly, a depolarizing wave sometimes occurs in 3 cells without a preceding i.p.s.p. (Fig. 11b). Secondly, a double depolarizing wave occurs in some 3 cells without an i.p.s.p. preceding the second burst (Fig. 7b). Neither of these results can be explained on the basis of simple post-inhibitory rebound following the 3-cell i.p.s.p., and the fact that a depolarizing potential can occur independently of inhibition suggests that it is mediated by another sort of interneurone whose firing is timed to follow the interneurone(s) producing the i.p.s.p.s. The exact nature of this second input will be made clearer when we have discussed 5, 7 and 8 cells.

There is considerable variation in the amplitude and duration of the peak depolarization of 3 cells, but compared with 4-group cells the depolarization always has a slow rate of decline (Fig. 8d). Superimposed on the slow decline of 3 cells are small peaks of depolarization which are synchronous on BL₃ and BR₃ (Fig. 7 a, b).

Common synaptic inputs on 3 cells are presumed to be the main factor producing synchronized bursts in 3 cells (Fig. 7b). The parallel changes in membrane potential of recorded cells are very obvious when 3 cells in the same preparation are compared (Fig. 7 a). Low-level non-rectifying electrotonic junctions connect 3 cells (Fig. 7 c) and perhaps play a part in spike synchronization. Hyperpolarization of one member of a pair of 3 cells to prevent firing reveals small depolarizations in this cell from the spike activity of its partner which is still firing (Fig. 7d). This suggests that the electrotonic junction can provide a small component to the depolarizing wave underlying burst activity.

The evidence for a second input to 3 cells is indirect, but other buccal cells where the two phases of synaptic input were inhibitory provide convincing evidence for a second input, as will be seen for the 5 cells.

5 cells are small, but it has been possible to record many cells of this type in feeding preparations. Thus pairs of 5 cells in the same or opposite buccal ganglia have been recorded (unlike other small buccal neurones like the 6 cells) and it has been possible to record them with all the other buccal ganglion cell types we describe in this paper.

The most characteristic feature of 5 cells is that they receive two phases of inhibitory input per burst cycle (Fig. 9c). If a 5 cell is recorded with a 3 cell (Fig. 9a) or 4-group cell (Fig. 9b) it can be seen that the first phase of inhibition is synchronous with the inhibitory input preceding the bursts of 3 and 4-group cells, and thus occurs at the same time as the e.p.s.p.s of the 1 and 6 cells. However, instead of the depolarizing wave which follows the i.p.s.p.s of 3 cells and 4-group cells, a second and stronger wave of inhibition occurs (Fig. 9 a, b) which we will call the second phase of inhibition. Following release from this second phase of inhibition, the 5 cell begins to fire usually without rebound excitation and continues to fire at a steady rate until the onset of the first phase of inhibition of the next cycle (Fig. 9a). Summarizing the firing pattern of 5 cells in comparison with other buccal neurones, we find that the 5-cell burst begins soon after the end of the 3- and 4-group cells’ bursts, and terminates at the start of the bursts in 1 and 6 cells (for summary diagram, see Fig. 14).

There are certain secondary features of 5-cell discharge patterns which are of We have recorded pairs of 5 cells in left and right buccal ganglia, and as with other bilaterally symmetrical neurones they receive identical synaptic inputs leading to synchronous bursting (Fig. 9c). However, unlike other bilaterally symmetrical neurones in the buccal ganglia, no electrotonic coupling could be demonstrated between BL₅ and BR₅. The lack of synaptic connexions between 5 cells in opposite ganglia is supported by our neuroanatomy (Benjamin et al. 1979), which showed no contra-laterally projecting 5-cell axons.

7 cells are small and variable in location. Nevertheless we have been able to record many cells of this type and characterize their firing pattern in relation to other buccal neurones.

The 7 cells may be closely compared with the 5 cells in that they also receive two consecutive phases of synaptic input which occur at the same time as those of 5 cells (Fig. 10,a). The 7 cells differ from the 5 cells in that the first phase input is excitatory, but are similar in that the second phase is a large-amplitude inhibitory input (Fig. 10a). This timing is confirmed by recording a 7 cell with a 4-group cell when the excitation of the 7 cell corresponds with the inhibition of the 4-group cell (Fig. 10b, c). Like the 5 cells, the 7 cells usually recover slowly from the second-phase inhibitory input, an exponential decrease in membrane potential usually giving rise to steady firing (Fig. 10b). However, the discharge rate accelerates sharply during the first (excitatory) phase of synaptic input of the next cycle (Fig. 10b). Occasionally some acceleration of spike activity is seen immediately after release from the second phase of inhibition, suggesting rebound excitation (Fig. 10c, following last inhibitory input in 7 cells) but this is not as pronounced as in 4-group cells (or 8 cells, see next section). The overall relationship of 7 cells to single input cells can be considered by reference to Fig. 10c. Two 7 cells from the right buccal ganglion were recorded with a 4-group cell and a 1 cell. It can be seen that the first-phase excitatory input on the 7 cells occur at the same time as the inhibition on the 4-group cell and the excitation on the 1 cell. The second-phase inhibitory input of the 7 cell occurs after the single input of the 1 cell and 4-group cell. It is interesting to note that the 4-group cell of Fig. 10c also receives a slight second phase of inhibition. The significance of this second input will be considered in the next section dealing with 8 cells. interest. There are large variations in the duration of the first phase of inhibition of 5 cells. When the first phase is longer than about 2 s, discharge of action potentials of about half normal amplitude are observed usually towards the end of the i.p.s.p. when membrane potential is falling (Fig. 9c, last double input at right of trace). This characteristic feature of the 5 cells is presumably due to invasion of the soma by blocked axonal spikes. Despite variations in the duration of both the first and second phases of 5-cell inhibition, the amplitude relationship of the two phases is constant, the second phase always being much larger in amplitude compared with the first (Fig. 9a, b, c) and much larger than the i.p.s.p.s seen in 4-group cells and 3 cells from the same preparation (Fig. 9 a, b).

If we compare the discharge of 7 cells with that of other buccal neurones we can see that they begin to fire at the end of 4-group cell bursts (Fig. 10b), fire steadily until the start of the 1 cell excitatory input (Fig. 10c), when they accelerate, and then cease to fire as the 4-group cells switch on (Fig. 10a). A summary of 7-cell activity in relation to other buccal neurones, which includes 3 cells and 6 cells, is shown in Fig. 14.

Simultaneous recording of two 7 cells shows that they receive identical synaptic inputs (Fig. 10c), and this must be responsible for the similar patterns of spike activity seen in these cells. No synaptic connexions have been found between 7 cells. Injecting a powerful depolarizing current pulse into one 7 cell produces no change in the membrane potential or spike activity of another 7 cell recorded at the same time (Fig. 10c, arrowed).

8 cells are rather difficult to find and, although we have been able to record their activity in relation to other buccal neurones, pairs of 8 cells have not been recorded.

Like the 7 cells, the 8 cells receive two consecutive phases of synaptic input in each cycle of bursting. The first phase is inhibitory and occurs at the same time and is of similar duration to the excitatory input of the 7 cells (Fig. 11 a). A second phase of 8-cell inhibition, which is usually larger than the first (Fig. 11b), accompanies the second inhibitory input on 7 cells (Fig. 11 a). Recovery from inhibition is rapid in 8 cells, and leads to a fast burst of spikes in contrast to 7 cells, which begin to fire slowly after the end of the i.p.s.p. (Fig. 11a). Post-inhibitory rebound bursts of spikes in 8 cells are followed by low-frequency spike activity (Fig. 11 a) or inactivity (see Rose & Benjamin, Fig. 12 a, 1979) until the next cycle of synaptic input. In non-feeding preparations, 8 cells continue to fire at high frequency until the next period of synaptic inhibition (Fig. 11 b). The timing of 8-cell bursts was confirmed by recording an 8 cell with a 3 cell. In this case, the first phase of inhibition in the 8 cell occurs at the same time as the inhibition preceding the 3 cell burst (first cycle of synaptic input from the left). It is interesting that the first phase of synaptic input may be absent from 8 cells and 3 cells in non-feeding preparations (Fig. 11 b, second cycle of synaptic input from the left).

We can now summarize the activity of 8 cells in relation to other buccal neurones. The second phase of inhibitory inputs to 8 cells are followed by a burst of spikes which precede the start of slow-frequency firing of 7 cells (Fig. 11a). The second phase of inhibition in 8 cells is accompanied by excitation in 3 cells (Fig. 11 b) so that a 3-cell burst is followed by an 8-cell burst. The overall sequence of buccal cell bursts, which includes those of 8 cells, is shown in Fig. 14. Bursts in 1 cells, 6 cells and 7 cells start the cycle, and are followed by activity in 3 cells and 4-group cells, which are followed kn turn by 8-cell bursts. 5 and 7 cells fire slowly until the beginning of the next cycle of oynaptic input. The significance of this sequence will be explained in Rose & Benjamin (1979). It is sufficient to state here that the consecutive bursts of action potentials in 6 cells, 4-group cells, 8 cells and 5 cells provide the central motor program for the four-phase feeding cycle seen in the semi-intact preparation.

Rebound excitation following inhibitory synaptic input seems to be an important mechanism in 8 cells, and they share this presumed endogenous response mechanism with the 4-group cells. We have defined 4-group cells as single input neurones with inhibitory input leading to bursts of spikes by post-inhibitory rebound, and 8 cells as double-input neurones with a similar burst-generating mechanism. However, in some preparations the largest 4-group cells (BL₄ and BR₄) and several other of the largest 4 cluster cells receive a component of second-phase inhibition. The 4 cell of Fig. 10c has a second phase inhibitory input which hyperpolarizes the neurone just when the depolarizing component of the first phase of inhibition would have led to a burst. The second i.p.s.p. on the 4-group cell thus delays the onset of the burst just like the second phase of inhibition of an 8 cell. This overlapping of electrophysiological properties of 4-group and 8 cells has functional significance in that the 4-group cells and 8 cells innervate different, but adjacent parts of the anterior jugalis muscle in the feeding system (Rose & Benjamin, 1979), and also project along the same peripheral nerves (Benjamin et al. 1979). The evidence for this statement and its functional significance will be explained in Rose & Benjamin (1979).

2 cells are the second largest pair of neurones in the buccal ganglia. They are easy to identify but their synaptic inputs are complex and produce patterns of spikes which were often difficult to relate to activity in other buccal neurones. This is mainly because they receive a powerful inhibitory input which is not shared by other identified cells in the buccal ganglia.

2 cells receive both excitatory (Fig. 12b) and inhibitory synaptic inputs (Fig. 12a, c) which can result in similar patterns of spike activity in pairs of cells from the same preparation (Fig. 12a) but often does not, so that a 2 cell from one ganglion could fire whilst its partner from the other ganglion was silent (Fig. 12 c). No synaptic connexions were recorded between 2 cells and this may be one reason for the lack of spike synchrony.

We have classed 2 cells as double input cells because their excitatory input lasts throughout the two phases of synaptic input in a double-input cell, although there may not be two clear peaks of depolarization within the e.p.s.p.s waveform. If a 2 cell is hyperpolarized to prevent firing and recorded with a 4-group cell (Fig. 13 a) then the depolarization of the 2 cell can be seen to start at the same time as the inhibition of the 4-group cell, rises to a peak by the end of the 4-group cell i.p.s.p., and then gradually decays until the end of the 4-group cell burst. This shows that the excitatory input to 2 cells covers the same time period as the double inputs of cells such as the 5 cells (Fig. 9b). In some preparations the excitatory input has a more obvious double form (Fig. 12b) with the first phase of excitation always lower in amplitude than the second. The double excitatory input to 2 cells produces long bursts of spikes and their timing can be confirmed by recording 2 cells with other buccal neurones. For instance, the burst of 2 cell spikes in Fig. 13c starts at about the same time as the 1-cell burst but continues long after the end of the 1-cell activity. Long bursts of spikes in the 2 cell of Fig. 13 (b) occur at the same time as the inhibition followed by excitation (and bursting) in the 3 cell. However, other shorter bursts occur between the long bursts of spikes in the 2-cell record of Fig. 13 (b), and these are associated with the occurrence of membrane hyperpolarizations not present on the 3 cell (or any other identified cell type in the buccal ganglion).

Following the long oursts of spikes due to double-excitatory inputs on the 2 cells of Fig. 13(b) shorter bursts of spikes are superimposed on the depolarizing peaks of membrane oscillations. Several short bursts of spikes may occur before the next long burst (Fig. 13b). We think that the short bursts of 2 cell spikes are produced following release from inhibitory synaptic inputs which are responsible for the hyperpolarizing components of the membrane potential oscillations. Our arguments for periodic hyperpolarizations due to synaptic activity are not conclusive, and part of the hyperpolarizations following long 2-cell bursts could be due to an endogenous mechanism (see below). However, it is difficult to explain the identical hyperpolarizations recorded in BL₂ and BR₂ as being due to an endogenous oscillatory property of the 2-cell membrane because they occur in the absence of spike activity in one member of a pair (Fig. 12 c) and in the absence of any synaptic connexions between 2 cells. The absence of synaptic connexions between 2 cells has been shown in many experiments. For instance, it can be seen that post-synaptic potentials in the BL₂ of Fig. 12(c) did not occur as a result of spikes in BR₂. The lack of connexion between 2 cells will be confirmed by anatomical evidence in Benjamin et al. (1979). Some of the larger hyperpolarizations which follow long bursts of spikes in 2 cells (Fig. 13 b) could be due to an endogenous mechanism because molluscan neurones are known to hyperpolarize following bursts of spike activity (Junge & Stephens, 1973).

In summarizing factors responsible for spike activity in 2 cells we can say that two types of synaptic input mediate bursting. Double excitatory inputs generate long bursts of action potentials which start at the beginning of i-cell bursts (Fig. 13 c) and continue until the end of 3-cell bursts (Fig. 136). In between these long bursts, several short-duration bursts of spikes are superimposed on depolarizing peaks of membrane potential oscillations due to inhibitory synaptic input and the occurrence of these bursts following inhibition is not related to spike activity in other buccal neurones (e.g. Fig. 13b). Strongly synchronized spike activity in 2 cells is less common than in other bilaterally symmetrical neurones in the buccal ganglia and we relate this to the absence of 2-cell connexions and to the low amplitude of synaptic inputs to these cells in many preparations. The spike activity of 2 cells often seems diffuse compared with other buccal neurones and this impression is confirmed when the spike pattern of 2 cells is compared with the overall feeding pattern in the summary diagram of Fig. 14.

We have summarized the synaptic inputs of Lymnaea buccal neurones in Fig. 14 and their electrotonic connexions in Fig. 15. We show two cycles of synaptic input and evoked spikes (Fig. 14) with the cycle beginning at the start of the 1- and 6-cell bursts and ending at the end of the 5-cell bursts. This corresponds with the feeding cycle of the buccal mass, which will be described in the accompanying paper of Rose & Benjamin (1979).

The rather complex pattern of spike activity in the buccal feeding system is produced by synaptic inputs coming from outside the population of recorded neurones, but interacting with their endogenous properties and reinforced by electrotonic connexions between certain neurones of the same type (Fig. 15). A simple model of how the system might work is for one population of interneurones to be responsible for the first phase of input and another to be responsible for the second phase. In phase terminology, the two populations of interneurones would need to oscillate with the same period but the phase of one set of interneurones would need to be delayed with respect to the other to produce two phases of synaptic input per cycle of feeding. Sometimes variations in the firing patterns of the two underlying oscillators occur spontaneously; for instance, in the 3 cell and 8 cell of Fig. 11 the first input is always followed by the second but the second input may occur on its own. Thus it appears that one of our supposed oscillators has become spontaneously uncoupled from the other.

Studies of several invertebrate motor systems (reviewed by Kennedy & Davis, 1977) have provided us with examples in which pattern generation occurs in network(s) of interneurones, which then impose their activity on follower motoneurones. Our own unpublished observations on the interneurones of the buccal ganglion of Lymnaea also suggest that interneurones generate the motor pattern in this preparation. However, we were impressed by the complexity of motoneurones firing patterns of Lymnaea compared with these other systems. In Helisoma, for example, positive feedback mediated by electrotonic coupling of interneurones generates synchronous burst activity in the premotor networks which in turn produces inhibition and excitation in the follower cells (Kaneko et al. 1978). The burst pattern which results is a simple alternation of activity between two types of motoneurone which is related to protraction and retraction of the radula. By comparison, the Lymnaea buccal ganglion generates a pattern which is of a different order of complexity, involving two phases of synaptic input and at least eight types of follower motoneurones. Since there are two phases of synaptic input in Lymnaea instead of the one in Helisoma, the simplest explanation would seem to be that there are at least two interneuronal networks each capable of generating one phase of synaptic input of the type found in Helisoma, these networks being coupled in such a way that they fire consecutively. Then if both networks synapse on certain motoneurones we obtain the following combinations of synaptic inputs: ie (inhibition followed by excitation) = 3 cells; ii = 5 cells, 8 cells; ei = 7 cells; ee = 2 cells. Of course at this stage we cannot state unequivocally that there are two interneuronal networks, but only that there are two main phases of synaptic input in which these combinations occur. Other neurones in Lymnaea were classed as single-input neurones because they received just the first interneurone network’s input as follows: e = 1 cells, 6 cells; 1 = 4-group cells. The overall effects on spike activity of these single or double inputs is summarized in Fig. 14.

From a functional point of view there are two further features which are important. Firstly the imposed synaptic input operates on cells with different endogenous properties, such as post-inhibitory rebound, spike adaptation, threshold, and possibly post-burst hyperpolarization. Post-inhibitory rebound (PIR) occurs frequently in molluscan ganglia (Kandel, Frazier & Wachtel, 1969; Siegler, Mpitsos & Davis, 1974; Kater, 1974), and has also been found in hippocampal neurones of the cat (Kandel & Spencer, 1961), and modelled by Perkel & Mulloney (1974). In the Lymnaea buccal ganglion it is an important mechanism for ‘i’- and ‘ii’-type cells. Thus the 4-group cells and 8 cells rebound strongly from inhibitory synaptic input whereas the 5 cells show little or no PIR (Fig. 14). This feature is important from the point of view of the functioning of the buccal musculature (Rose & Benjamin, 1979). Post-inhibitory rebound from the first input plus excitation from the second input may be the underlying cause of the high-frequency firing of the 3 cells. Several factors may determine the observed properties of the 4-group cells. Thus differences in the times of burst onset between 4-group cells (Fig. 3 a) may reflect differences in spike threshold and in the degree of PIR within the population. The sharp cut-off of strongly bursting 4-group cells and 3 cells before the next synaptic input has occurred, could involve endogenous properties such as post-burst hyperpolarization (Junge & Stephens, 1973). For 1 cells, spike adaptation is important, since these cells only fire during rapid changes in potential. None of the 1-cell spike activity is autoactive and this appears to be different from the situation in the slug Limax, where cells with equivalent function (salivary gland motoneurones) show endogenous bursting (Prior & Gelperin, 1977).

The second factor which modifies the effects of synaptic input is the degree of electrotonic coupling between the buccal neurones themselves. In Lymnaea there is no evidence of synaptic connexions between neurones of different types (i.e. types 1-8). However, we do have to take into account the effect of low-level electrotonic connexion within groups of the same type, and between bilaterally symmetrical pairs of cells or groups of cells (summarized in Fig. 15). Functionally such coupling can produce synchronization of activity (Wilson, 1966). Recent work on electrotonically coupled networks in molluscs has demonstrated reverberatory activity which builds up and suddenly terminates (Getting & Willows, 1974). The 4-group cells of Lymnaea form such an electrotonically coupled network but no such build-up of burst activity has been observed and it is more likely that there is a deceleration of spike activity following the onset of the burst, which may be either the result of adaptation (Wilson, 1966), or an effect accompanying PIR (Perkel & Mulloney, 1974). The termination of 4-group cell’s bursts may be a result of the overall activity falling below a critical point when the positive feedback fails to operate, or may involve endogenous properties such as post-burst hyperpolarization.

The buccal mass of Lymnaea is a symmetrical structure (Carriker, 1946) and it is therefore important to provide synchronization of motoneurone activity of cells of the same type but on the opposite sides. This is achieved by synchronous synaptic input to, and electrotonic coupling of, cells of the same type but on the opposite sides. Such electrotonic coupling has been observed between bilaterally symmetrical cells in the buccal ganglia of other molluscs (e.g. Levitan, Tauc & Segundo, 1970; Berry, 1972), and in Helisoma there is evidence of separate populations of electrotonically coupled interneurones in left and right buccal ganglion, both these populations being themselves electrotonically coupled (Kater, 1974). In Lymnaea the combination of strong synchronous synaptic input plus electrotonic coupling results in good spike synchrony of cells from opposite buccal ganglia (i cells, 3 cells and 4-group cells) whereas weak inputs and no coupling results in relative lack of synchrony (2 cells). Strong synaptic inputs alone can also counteract lack of electrotonic coupling to produce close synchrony (e.g. 5 cells).

R. M. R. was supported by an M.R.C. grant to P. R. B.