ABSTRACT
The Lymnaea buccal ganglion is organized such that the basic feeding rhythm is generated by an interneuronal network which imposes its activity on a set of follower cells. In this paper we extend our earlier observations (Benjamin & Rose, 1979) on the follower cells to show that they receive four consecutive synaptic inputs. The main objective of the paper is to describe the properties of an interneurone called the ‘slow oscillator’ which is capable of initiating feeding cycles. This interneurone will be used in the following paper (Rose & Benjamin, 1981) to drive other members of the interneuronal network in order to determine how it is organized, and to understand the origin and timing of the four synaptic inputs to the follower cells.
INTRODUCTION
Many invertebrate motor systems produce cyclical patterns of burst activity in the control of simple behaviours such as swimming, heart beat and feeding. It has now been found that a number of these systems are organized such that a network of interneurones generates the rhythm itself, this rhythm then being imposed on the motoneurones. This arrangement occurs in such systems as leech swimming (Friesen, Poon & Stent, 1976), Helisoma feeding (Kaneko, Merickel & Kater, 1978), and probably also in the Tritonia swimming system (Getting, 1977). In this paper and the next (Rose & Benjamin, 1981) we will show that this basic arrangement is also present in the Lymnaea feeding system. We will also show that the intemeuronal network has certain novel features when compared with other networks (reviewed by Friesen & Stent, 1978).
The starting point of our present investigations is our earlier analysis of the activity of the buccal motoneurones of Lymnaea (Benjamin & Rose, 1979). We have shown that eight different types of motoneurone could be distinguished, their firing patterns resulting from various combinations of two main phases of synaptic input. The resulting firing patterns were related to movement of the buccal mass during feeding (Rose & Benjamin, 1979). In the first part of this paper we re-examine this mot™ neuronal pattern to emphasize the timing of two minor inputs which were only briefly discussed in the previous work (Benjamin & Rose, 1979). These additional inputs will be shown to be particularly significant when we come to discuss each component of the interneuronal network in the second paper (Rose & Benjamin, 1981).
Our earlier examination of the motoneurones was based on the recording of activity in spontaneously active preparations, and we found considerable variation in the degree of patterning in different animals. It would obviously be very useful to have control over the rhythm generation. We have now found that the largest interneurone in the Lymnaea buccal ganglion will drive the rest of the network in a stereotyped way when steadily depolarized. In this paper we describe the activity of this interneurone, which is called the ‘slow oscillator’, and show how it is connected to some of the motoneurones. In the second paper (Rose & Benjamin, 1981) the slow oscillator will be used to drive the rest of the interneuronal network, and it will be shown that this network is responsible for all of the synaptic inputs to the motoneurones.
METHODS
Methods were essentially the same as described by Benjamin & Rose (1979). The isolated brain and buccal ganglion was pinned out in a perspex chamber with a sylgard base, and bathed in Hepes buffered saline (Benjamin, 1978). Intracellular recordings were made from up to four indentified buccal neurones simultaneously and fed into four high input impedance amplifiers each of which had capacitance compensation and current passing facilities. Potentials were displayed on a storage oscilloscope and permanently recorded on film or on a four-channel pen recorder (Brush 2400). As a test for monosynaptic connexions, microelectrodes were filled with 1 M TEA bromide (Sigma) solution. An interneurone was impaled, and the TEA was allowed to diffuse from the tip of the microelectrode. This technique was effective because of the short distances of the synapses from the interneuronal cell bodies.
RESULTS
Location of cells
In this paper two new types of follower cell will be described in addition to the eight types of motoneurones of our earlier papers (Benjamin & Rose, 1979; Benjamin et al. 1979). Of these two new types, which are called 9 and 10 cells, the 9 cells are located on the ventral surface of the ganglion and are not shown in Fig. 1 which is a dorsal view. The location of a representative 10 cell is shown in relation to cell types 1-8 as described previously (Benjamin & Rose (1979), Fig. 1). Also shown is the position of the slow oscillator (SO) interneurone. This cell, which is about 40 μm in diameter, is located at the point where the 1 and 2 cells are in close contact, thus making it fairly easy to find in different preparations.
Follower cells and their inputs
In the Lymnaea buccal ganglion the central pattern is generated by a network of interneurones, which impose their activity on a set of follower cells. The inter neuronal network itself has four subcomponents, and for simplicity these may considered to be connected in a chain (Fig. 2d). For the complete interneuronal circuit the reader is referred to Fig. 18 of the following paper (Rose & Benjamin. 1981). The feeding cycle is generated by consecutive firing of members of this chain. The object of this paper is to show that feeding cycles may be initiated by manipulation of the membrane potential in the first element of the chain, the ‘slow oscillator’ (SO) interneurone.
Before attempting to describe the activity of the interneurones it is important to revise our earlier description of the follower cells. Previously (Benjamin & Rose, 1979) it was shown that there were 8 different types of follower cell, each receiving one or two consecutive phases of synaptic input during spontaneous feeding cycles. Two minor inputs were also briefly described. Further recording has shown the existence of two more cell types, 9 and 10 cells, and has helped to clarify the timing of the two minor inputs. These findings are incorporated into our earlier summary diagram (Benjamin & Rose, 1979, Fig. 14) in Fig. 2. Since the pattern generator produces the inputs to the follower cells, the activity of the complete system can now be understood by aligning Fig. 2 for the follower cells with Fig. 18 in the following paper (Rose & Benjamin, 1981), which shows the intemeuronal activity together with a few representative examples of follower cells taken from Fig. 2. The exact role of each of the intemeuronal inputs to the patterning of the follower cells will now be discussed with reference to Fig. 2.
In Fig. 2 it is apparent that there are four phases of synaptic input (SO, Ni, N2 and N3) occurring sequentially. In our earlier work we considered only two main biases of synaptic input (Benjamin & Rose, 1979). It will be shown in the following paper (Rose & Benjamin, 1981) that these two main phases of synaptic input arise from activity in the N1 and N2 interneurones. The N1–N2 activity is preceded by activity in the SO cell, which initiated the cycle, and followed by N3 which completes the cycle. The SO and N3 inputs to the follower cells are minor in the sense that they do not greatly affect the patterning of the follower cells. One further technical point is that although the sequence of activity is SO → NI → N2 → N3, there is some overlap of activity between Ni and SO (Rose & Benjamin, 1981) so that the N1 phase has to be labelled ‘Ni + SO’ to indicate this overlap.
The two cell types shown at the top of Fig. 2 (1 cell, 6 cell) are both depolarized by N1 input; the 6 cells being protractor motoneurones and the 1 cell supplying the salivary glands (Benjamin et al. 1979; Rose & Benjamin, 1979). Evidence will be given later (Fig. 7) that the SO interneurone has an input on the 1 cell, and this has been included on the diagram. The 9, 4-group and 8 cells have been arranged together since they form a group (see below). The 9 cells receive inhibitory input from the N1 interneurones and recover rapidly, whereas the larger 4-group cells have an inhibitory input from N1 and to a lesser extent from N2 also. The 8 cells receive inhibitory input from Ni cells and have a strong inhibitory input from N2 cells. Evidence will be given later (Fig. 7) of a SO input to the 4-group cells, and this has been included together with a possible SO input to 9 cells (Fig. 3). The input from the N3 interneurones to 4 and 8 cells consists of a succession of discrete hyperpolarizations which allow the N2 input and cause interruptions of the 4-group and 8 cell bursts. Clear examples of the N3 input are provided in Figs. 7, 17 and 18 of the following pap® (Rose & Benjamin, 1981).
The 3 cells receive inhibition during the N1 phase, and excitation during N2, which causes bursting. Small amplitude e.p.s.p.s. follow the 3 cell burst, these originating from activity in N3 interneurones (Rose & Benjamin, 1981), Figs. 16 and 17). The 5 cells and the 7 cells both receive inhibition during the Na phase, but the 5 cells are inhibited and the 7 cells excited during the Ni phase. Both 5 and 7 cells receive an input from the N3 interneurones which results in oscillations in the membrane potential as shown in Fig. 16 and 17 (for the 7 cell) of the following paper (Rose & Benjamin, 1981). Both the 2 cells and the new 10 cells receive excitation during both N1 and N2 phases.
9 and 10 cells
We give here examples of the firing patterns of 9 and 10 cells. This is simply for reference since these cells have not been discussed previously.
The 9 cells occur almost entirely on the ventral surface of the buccal ganglia. The ventral surface contains many cells of similar size (30–40 μm) so that it is more difficult to identify individual cells visually than it is on the dorsal surface. Two such cells are shown on the upper and lower channels of Fig. 3 (a). The cell on the upper channel is most typical of a 9 cell, showing very rapid recovery and high frequency burst discharge following the (N1 + SO) period of inhibition. The timing of the 9 cell discharge has been determined by recording it with a 4 cell on the dorsal surface of the opposite buccal ganglion. This was achieved by twisting the buccal ganglion so that the left dorsal and right ventral surfaces were uppermost. The 4 cell received similar inputs to the 9 cell except that recovery was slowed in BL4 by the presence of an N2 input. The slow recovery of the lower cell following the N1 phase may also be due to this cell receiving N2 input.
Although we have seen that there are differences in the rates of recovery of the ventral surface cells (Fig. 3) those with a very rapid recovery from inhibition were more common and we will refer to ventral surface cells having typically this rapid recovery as ‘9 cells’. In fact 9, 4-group and 8 cells (Benjamin & Rose, 1979) may be considered together as a group. All of these cells receive first network (N1) inhibitory input, but there are variations in the extent of second network (N2) input. Thus 9 cells receive mainly Ni input, 4 group cells have some N2 input which slows down their rising phase before the burst, and 8 cells represent the other extreme to 9 cells, with a very strong N2 input in addition to the Ni input. In support of this idea of a 9, 4-group, 8 cell group, we found that 9 cells were electrotonically connected to each other and to 4 cells, although we have not investigated this coupling quantitatively. The 9 cells produce no intemeuronal effects on stimulation. A further point is that the 9 cells receive unitary i.p.s.p.s preceding the Ni inhibitory input. This presumably arises from activity in the SO interneurone, although we have been unable to prove this because of the location of the SO and 9 cells on opposite surfaces of the ganglion.
The 10 cells are located as a group in the position shown in Fig. 1. In common with many of the other follower cells they receive two consecutive inputs, these both being excitatory, so that the 10 cells are similar to the 2 cells (Benjamin & Rose, 1979). In Fig. 3(6), a 10 cell is recorded with a 7 and a 3 cell. The first network input (N1) Produces a slowly rising depolarization of the 10 cell, which occurs at the same time as excitation of the 7 cell and inhibition of the 3 cell. The second network input (N2) strongly depolarizes the 3 and 10 cells and inhibits the 7 cell.
The slow oscillator interneurone
During our search for possible interneurones, a cell was found which could initiate complete feeding cycles when steadily depolarized. As will be shown below, this cell is responsible for the inhibitory input to the 4-group cells which precedes the first phase (N1) input. In the following paper (Rose & Benjamin, 1981) it will be shown that this cell excites other interneurones either directly or indirectly, and that there is inhibitory feedback from these cells back onto this interneurone. As a consequence of the feedback, steady depolarization causes the membrane potential of this cell to oscillate slowly. We have decided to call the cell the ‘slow oscillator’ (SO) interneurone because of the slow membrane potential oscillation and also because it has a characteristic low instantaneous firing frequency which distinguishes it from other buccal interneurones.
The initial discovery of the SO cell is shown in Fig. 4(a). When the SO cell was depolarized in a stepwise manner, the firing frequency of the 5 cell decreased. After about 2 s a large amplitude hyperpolarizing wave appeared in the 5 cell, and at the same time the SO stopped firing. An inhibitory wave also appeared on the 4 cell recording every time a burst was initiated in the SO, this waveform beginning slightly after the onset of the SO burst, and terminating at the time of appearance of the large amplitude hyperpolarizing wave on the 5 cell. The simplest explanation of these findings would appear to be that the SO is initiating activity in the Ni followed by the N2 interneuronal network, the former causing inhibition of the 4 cell and slowing of the 5 cell, and the latter causing the large amplitude hyperpolarization of the 5 cell. The initiation of this sequence of events occurs each time the SO is depolarized (Fig. 4a). If instead the SO is steadily depolarized (Fig. 4b), periodic inhibition of this cell is observed. This inhibition occurs at the same time as the large amplitude hyperpolarization of the 5 cell as observed previously (Fig. 4a).
Functional role of the SO interneurone
So far preliminary evidence has been given which shows that when the SO is depolarized by applied current, it causes p.s.p.s in the 4 cell and initiates a burst pattern in the 5 cell similar to that seen in normal feeding (Benjamin & Rose, 1979; Rose & Benjamin, 1979). In this section we give evidence that the SO makes a mono-synaptic connexion with the 4 cell, and also examine the role of the SO when it is spontaneously active.
We have tested for a monosynaptic connexion between the SO and the 4 cell by injecting TEA into the SO neurone. TEA should increase the duration of the SO action potentials, which in turn will potentiate the amplitude of the i.p.s.p.s on the 4 cell if a monosynaptic connexion is involved. If instead, there is an interposed interneurone between the SO and the 4 cell, this potentiation should not occur (discussed by Berry & Pentreath, 1976). Fig. 5(a) is a recording from a spontaneously active SO cell and a 4 cell. The bursts in the SO are very similar to those seen previously when the SO was artifically depolarized (Fig. 4), except that the onset of bursting in the SO is slightly delayed and there are fewer action potentials than in previous recordings, i.e. the SO is presumably less powerfully activated in this particular case. Also in this example the 4 cell is in a strongly-bursting state and it is just possible to distinguish i.p.s.p.s on the 4 cell which are apparently 1:1 with SO spikes. TEA was then allowed to diffuse from the electrode tip, and the SO spikes lengthened (Fig. 56). As the action potentials increased in duration in the SO cell, the i.p.s.p.s on the 4 cell increased markedly in amplitude and were 1:1 with SO spikes suggesting that this is a monosynaptic connexion. The period of the oscillation changed little on TEA injection, although the periodic inhibition of the SO was less distinct.
We have seen that the SO is capable of spontaneous activity. Yet in many preparations we have observed inputs to the follower cells which cause feeding cycles when the SO was silent. Is the SO therefore just playing a modulatory role, or is it essential for feeding cycles to take place ? One way of answering this question is to prevent the firing of the SO when it is spontaneously active and observe the effects on follower cells. In Fig. 6, the spontaneously-active SO was hyperpolarized for 26 s while observing activity in 4 and 10 cells. When the SO is switched off, the 4 cell burst cycle period slows from to 6 s, while the first phase of excitation to the 10 cell is markedly lengthened. It is interesting that during the period of hyperpolarization the SO shows a slight hyperpolarization at the time of the first network (N1) input and a stronger hyperpolarizing wave at the time of occurrence of the second network input. Also note that the 4 cell continues to switch off rapidly when the SO is silent, so that we may conclude that the i.p.s.p. input from the active SO to the 4 cell (Figs. 4, 5) is not itself switching the 4 cell burst off.
The fact that the switching off of the SO only slightly affects the cycle period of the follower cells suggests that it is only playing a modulatory role in the feeding cycle.
Postsynaptic effects of the SO interneurone
Evidence has been given to show that the SO interneurone makes a monosynaptic connexion with the 4-group cells. Occasionally the i.p.s.p.s occurring on the 4-group cells are of larger amplitude than usual, and can be clearly seen to be in 1:1 relation to SO spikes. In the example given in Fig. 7(a), the 4 cell i.p.s.p.s gradually increase in amplitude during each SO burst. The i.p.s.p.s show most marked amplitude increase during the period of N1 input (labelled ‘SO + N1’).A similar phenomenon has also been observed in the 1 cell following SO spikes (Fig. 76). In this case the e.p.s.p.s following SO spikes again grow in amplitude, and this amplitude increase is also more marked during the N1 phase (again labelled ‘SO + N1’).
In the next paper (Rose & Benjamin, 1981) we will show a clear example of apparently facilitating e.p.s.p.s occurring in the N1 interneurones following SO spikes. We will also show, that when the N1 interneurones reach threshold and fire, N1 interneurone spikes are sometimes synchronized with SO spikes. The present observations on 1 and 4 cells then become explicable in terms of SO and N1 activity. The explanation would be that the SO fires first and causes facilitation on Ni, 4 and 1 cells. Towards the end of the SO burst, SO and Ni overlaps. In most cases several Ni cells fire towards the end of the SO burst, giving smooth p.s.p.s on follower cells during the ‘SO + N1’ phase. In the examples given in Fig. 7 it is probable that those. N1 interneurones which are firing are synchronized with SO spikes. Since SO and N1 cells converge on 1 and 4 cells their postsynaptic effects will summate in these cells. The recordings of Fig. 7 may therefore demonstrate a rather complicated sequence of events in which there is firstly a facilitation of p.s.p.s during the ‘SO’ phase, which is inforced by synchronized inputs from SO and N1 cells during the ‘SO + Ni’ phase.
Effects of injecting current into the SO interneurone
We have made some preliminary observations on the effects of passing maintained and transient currents into the SO interneurone.
Steady currents
If a steady depolarizing current is passed into the SO interneurone and the current strength is progressively increased, the SO fires at increasing frequency and the period of the oscillation shortens. In the example given in Fig. 8 the period of the oscillation shortens more rapidly at lower current strengths (Fig. 8a–b) than at higher current strengths (Fig 8b–c). More detailed measurements of the inhibitory waveform which occurs periodically in the SO cell shows that the slope of the hyperpolarizing phase of the i.p.s.p. waveform becomes steeper with increasing stimulus strength. It is interesting that as the rate of onset of inhibition in the SO increases so does the rate of depolarization of the 3 cell (Fig. 8). At this stage we have not quantified the period-input current relationship of the SO, but this would be an important change to measure in more detail as a test of any model of the overall system.
Transient response
In a few preparations we observed that each time a short (1s) depolarizing pulse was passed into the SO interneurone, the pulse was followed by a prolonged (40–50) activation of the SO neurone in which activity oscillated periodically (Fig. 9). As the SO cell oscillates it is interesting that the amplitude of the waveform recorded in the 10 cell gradually increases (Fig. 9, upper channel). Increasing activation of the 10 cell is associated with increasing inhibition of the SO. Thus the 10 cell is initially below spike threshold, but starts spiking at the end of the recording as the SO neurone stops firing. At the end of SO firing, the SO receives successive waves of input from the second network N2 (see Rose & Benjamin, 1981) which also fire the 10 cell. This type of response therefore shows that there may be an imbalance in the activation of different intemeuronal networks. Similar growth and decay charges of different bursts have previously been observed in induced sequences in the buccal ganglion of Aplysia depilans (Rose, 1976) and in Lymnaea by the addition of sucrose to semi-intact preparations (Rose & Benjamin, 1979).
DISCUSSION
Is the SO cell a command interneurone ?
There are theoretically two basic mechanisms by which cyclic behaviour patterns might be initiated by cells controlling a rhythmic network. Firstly there could be a cell or group of cells which served to trigger the cyclical behaviour which then outlasts the triggering stimulus. Secondly the cyclical behaviour could arise through maintained activity in a cell or group of cells which would thus provide a continuous input to the rest of the system. As an illustration of this difference it is interesting to consider that in the Tritonia swimming system it was originally thought that a group of cells (the TGNs) triggered escape swimming (Willows & Hoyle, 1969), but later work (Getting, 1975, 1976, 1977) has suggested that the TGNs are unimportant in this pect and that swimming cycles result from maintained activity in a group of nitemeurones (C2 cells). The SO interneurone in Lymnaea is of this second type, for maintained depolarization initiates feeding cycles. However, within this category a further qualification can be added, depending on whether maintained activity in the interneurone is either necessary or sufficient or both necessary and sufficient for the behaviour to occur.
Weeks & Kristan (1978) have described an interneurone (cell 204) which is located in each segmental ganglion of the leech, and which elicits swimming activity from the isolated nerve cord when steadily depolarized. In this case it can be proved that activity in this interneurone is sufficient for the initiation and maintenance of swimming, but is not necessary in the sense that if two of these cells are hyperpolarized swimming could still be elicited by nerve stimulation. This is in contrast to Tritonia, where hyperpolarization of a C2 interneurone is effective in stopping cyclical bursts in response to peripheral stimulation. On the basis of these considerations it would be misleading to classify the SO interneurone as a command cell since feeding will occur even when this cell is hyperpolarized (Fig.6). In this respect it is similar to cell 204 of the leech.
Other cells which initiate feeding
There is evidence of cells having similar properties to the SO in other buccal ganglia. Gillette, Gillette & Davis (1980) have described a symmetrical pair of interneurones, the ventral white cells, in the buccal ganglia of Pleurobranchaea which have the ability to drive the feeding rhythm. These cells may produce either short bursts in phase with the feeding cycle or prolonged depolarizations which recur at long intervals. During the prolonged bursts the action potentials progressively broaden, the increase in spike duration being correlated with an acceleration of feeding cycle frequency. Although involved in the initiation of feeding cycles, the ventral white cells therefore have quite different electrophysiological properties to the SO interneurones of Lymnaea. Bulloch & Dorsett (1979) have also described a pair of neurones, the white cells, which may be involved in the initiation of patterned activity in the buccal ganglia of Tritonia hombergii.
Other interneurones which have synaptic connexions with buccal motoneurones in Lymnaea are the cerebral giant cells (C.G.C.s) (McCrohan & Benjamin, 1980b). Tonic firing of the C.G.C.s increases the intensity of bursting in a number of motoneurones, but the C.G.C.s are themselves incapable of initiating feeding cycles or of affecting the frequency of ongoing cycles. In spontaneously active preparations the C.G.C.s usually fire tonically, although when they do fire phasically they receive excitation in the Ni phase and inhibition in the Nz phase (McCrohan & Benjamin, 1980a). Recent work has also suggested that the C.G.C.s have direct effects on the buccal intemeuronal network including the SO interneurone (McCrohan, personal communication).
The follower cell inputs
In our earlier work (Benjamin & Rose, 1979) we showed that the follower cells were driven by two main phases of synaptic input. In the following paper (Rose & Benjamin, 1980) these will be shown to arise from activity in N1 and N2 networks. This scheme has been extended in this paper to include two minor inputs, SO and N3, so that cycle of activity is divided up into 4 separate phases of synaptic input (SO, N1, N2, N3, Fig. 2). In relation to the follower cells it is interesting to consider whether the SO and N3 inputs have any functional significance. It has been shown in Fig. 6 that SO activity is not essential for the termination of 4-group cell bursts, although it may be acting to reinforce those mechanisms which normally terminate bursts in 4-group cells. The input which we have called ‘N3’ is more interesting. Although it is difficult to see what function this input performs in inhibiting cells which are already hyperpolarized (5 cells, 7 cells) it is clear that the N3 inhibitory input fractionates 4-group and 8 cell bursts into a series of short duration bursts. Since 4-group and 8 cells are retractor motoneurones, this implies that retraction takes place as a ratchet-like movement, perhaps enabling the radula teeth to cut into the food more effectively or release it more effectively into the oesophagus (Rose & Benjamin, 1979). It is difficult to provide a functional explanation of the excitatory N3 input to the 3 cell, since this occurs after the 3 cell burst, and all the e.p.s.p.s are subthreshold.
ACKNOWLEDGEMENTS
We would like to thank Miss C. R. McCrohan for reading the manuscript R. M. R. was supported by an M.R.C. grant to P. R. B.