ABSTRACT
The feeding cycle of Lymnaea is generated by a network of three types of interneurone, N1, N2 and N3. This network is driven by the slow oscillator (SO) interneurone described in the previous paper. Interaction between the different interneurones is dependent on both connectivity and endogenous properties, and utilizes such properties as post-inhibitory rebound and self-feedback within electrically-coupled populations. Each of the four components of the interneuronal network (SO, N1, N2 and N3) is responsible for a different phase of synaptic input to the follower cell population which was previously shown to directly control feeding movements.
INTRODUCTION
In the feeding system of Lymnaea the central rhythm is generated by a network of interneurones which impose their activity on a set of follower cells which are mainly motoneurones. In previous papers (Benjamin & Rose, 1979; Rose & Benjamin, 1981) we showed that the follower cells received four consecutive phases of synaptic input, the first of these inputs arising from a single interneurone called the slow oscillator (SO). Maintained depolarization of the SO initiates bursts in the follower cells which are comparable to bursts observed during spontaneous feeding cycles (Benjamin & Rose, 1979; Rose & Benjamin, 1979). In this paper we use the SO interneurone to drive the rest of the intemeuronal network, consisting of the N1, N2 and N3 groups of cells, each of which is responsible for one of the identified inputs to the follower cell population.
One of the main aims of the paper is to try to determine the inter-connexions between the components of the intemeuronal network. Although we have only been able to provide direct evidence of monosynaptic connexions between some of the cells, many of the other connexions can be deduced from observing spontaneous or driven activity patterns in the interneurones. Our present hypothesis of the functioning of the interneuronal network is given at the end of the paper. Here it will be see that the mechanism of burst production involves the interaction of several endogenous and network properties. Thus, although the N1, N2 and N3 interneurones could be considered as forming a recurrent cyclic inhibitory loop (see Friesen & Stent, 1978), the network does not operate in the manner proposed by Kling & Székely (1968). Instead the recovery from inhibition is rapid in N2 and N3 cells and post-inhibitory rebound processes may be involved. Furthermore, the N1 network seems to be spontaneously rhythmic and homologous with the cyberchron network of the buccal ganglion of Helisoma (Kater, 1974). Although no single principle of operation is involved, the network provides an interesting combination of mechanisms similar to those demonstrated in other systems. The network operation has similarities for instance to the pyloric cycle of the lobster stomatogastric ganglion. Another interesting point is that the rhythm may be manipulated by passing current into a single inter-neurone (Rose & Benjamin, 1981). In particular it will be shown that the rhythm is not simply driven by the SO interneurone but N1, N2 and N3 interneurones also have inhibitory feedback connexions to this interneurone. In previous models of rhythmic networks the idea that the rhythm could be modulated in frequency by an external input has often been proposed (Kling & Székely, 1968; Friesen & Stent, 1978). In the Lymnaea buccal ganglion the SO interneurone is providing an external driving input, but it is also receiving feedback from the rhythm generating network.
METHODS
The methods were exactly as in the previous paper (Rose & Benjamin, 1981).
RESULTS
Location of cells
Fig. 1 shows the location of cell types 1–8 together with the follower cell 10 and the SO interneurone as described in the previous paper (Rose & Benjamin, 1981). In this paper we also describe the activity of three types of interneurones called N1, N2 and N3. These interneurones are all small cells, being 10–25 μm diameter and their positions are variable. The N1 interneurones are located around the edge of the 2 cell between the 2 and 3 cell and between the 2 and 4-group cells. In the N1 interneurone region there is at least one N3 interneurone shown by the filled-in circle in Fig. 1. The N2 interneurones are found around the 3 cell, particularly between the 3 and 4-group cells and are shown as half filled-in circles in Fig. 1. We estimate that there are about ten of the N1 and N2 type interneurones of which six or seven of each type have been drawn in Fig. 1.
Cell types N1, N2 and N3
The main object of this paper is to describe the different types of interneurone in the central pattern generator, and to determine how they are interconnected. There are three types of interneurone, N1, N2 and N3 whose location has been shown in Fig. 1. In the sections which follow we discuss each of these types in turn. A summary diagram of the interactions between these interneurones is given in Fig. 18, which should be referred to if any relationships need to be clarified.
N1 interneurones
An example of spontaneous activity in first network interneurones is shown in Fig. 2, in which two small cells are recorded with a 4 cell. The 4 cell shows the typical pattern of synaptic inputs of a feeding preparation (Benjamin & Rose, 1979). Following each 4 cell burst there is the SO inhibitory input (Rose & Benjamin, 1981). This is followed by a more marked phase inhibition caused by Nr interneurones which overlap with the end of the SO burst (further evidence of this SO and N1 overlap is given in Fig. 6). The second network (N2) input causes an inhibitory notch on the rising phase of the 4 cell as it recovers from N1 inhibition. Simultaneous with first network input on the 4 cell, the two small cells on the middle traces are strongly bursting, and both cells cut-off strongly as the second network input appears on the 4 cell. Therefore, from the point of view of the timing, the two small cells are candidate N1 interneurones. Later in the recording the neurone on the second channel was depolarized with a 1 sec pulse. This had the effect of activating the N1 neurone on the third channel. On termination of the depolarizing pulse a subthreshold plateau of activity remains on the second channel cell which is simultaneous with continued activity in the third channel cell-i.e. there is some reverberation of activity involving the two cells. Also each time the second channel is activated, the 4 cell on the upper channel becomes hyperpolarized, providing evidence that the activated cells are interneurones.
Further data on the N1 cells is given in Figs. 3 and 4. The cells shown in the middle channel of Fig. 3 and in Fig. 4 oscillated spontaneously with a period of about 20 s (i.e. a relatively inactive preparation). Depolarization of the cell shown in Fig. 3(a) for about 2 s initiated a burst of activity in this cell while simultaneously the two 4 cluster cells became hyperpolarized. It is important to note that the N1 cells do not cause clear 1:1 p.s.p.s. on the 4 cluster cells, for the hyperpolarization takes the form of a smooth summating i.p.s.p. This is presumably because there is a network of these interneurones which are firing synchronously (see Fig. 4). When the N1 interneurone stopped firing, the 4 cluster cell on the lower channel recovered from the inhibition and fired a burst of spikes which caused electrotonic p.s.p.s in the other 4 cluster cell (Benjamin & Rose, 1979). When the cell shown on the middle channel was depolarized for 4 s (Fig. 3b) this resulted in stronger activation. In this case the hyperpolarization of the 4 cluster cells was followed by a second larger-amplitude hyperpolarization which occurred as the middle-channel cell stopped firing. What is presumably happeN1ng here is that the first network interneurones are activating the second network interneurones which cause the second wave of inhibition on the 4 cluster cells (see later). When this second network is activated (Fig. 3b) the N1 cells switchoff more sharply than in Fig. 3 (a). This could be because the N2 interneurones are inhibiting the N1 cells at this point, or it could be a consequence of some intrinsic mechanism of the N1 cells associated with the higher level of activation of these cells.
The results of Fig. 3 amplify those given in Fig. 2, but we need evidence that these are the same cells. This would be provided if we could again demonstrate the reverberation of activity between cells of this type. Later in the experiment shown in Fig. 3, another cell of the same type was penetrated and is shown on the upper channel of Fig. 4. Activation of this cell resulted in a burst of p.s.p.s in the cell of the lower channel (which is the cell on the middle channel in Fig. 3). These p.s.p.s we superimposed on a step change of the membrane potential, which suggests that the two cells are electrotonically coupled. Similarly activation of the lower cell of Fig. 4 always results in a step change in the membrane potential of the upper channel cell. In this case, however, another feature is demonstrated. The first step is simply transmitted to the other cell. At a higher level of depolarization the second step initiates two action potentials which are transmitted as electrotonic p.s.p.s to the upper channel cell. At a still higher level of activation the third step initiates a burst of action potentials in the lower channel cell which are again transmitted as electrotonic p.s.p.s to the upper cell. This time however four p.s.p.s of a larger amplitude appear on the upper channel recording signifying the activation of a third cell. In this case also the membrane potential of both cells does not return rapidly to the resting potential as before, but shows signs of reverberation. Finally at the highest level of depolarizing current (fourth step to lower cell) the whole network goes into sustained activity lasting for about 10 s. The network is thus capable of reverberation as in the previous case (Fig. 2).
Although the results of Fig. 3 show that the 4-group cells become hyperpolarized when the N1 interneurones are activated, it was common to find cells bursting like N1 interneurones but which did not produce strong postsynaptic effects on the 4-group cells when activated. It seems that it is necessary to activate much of the network in order to see clear postsynaptic effects. Also because a number of N1 interneurones are converging on the follower cells, the p.s.p.s tend to be of a smooth compound type occurs spontaneously (Benjamin & Rose, 1979).
If the N1 cells are electrotonically coupled, it might be expected that the steady depolarization of one cell in the network would be transmitted to a number of other cells thus making the system more likely to oscillate. This certainly is the case in some preparations of which Fig. 5 is an example. This is a recording from two N1 interneurones and a 4 cell. The N1 cells were silent both before and after a steady depolarizing current was passed into one of them. Steady depolarization of the upper channel cell initiates a series of bursts at regular intervals (period ~ 14 s). The bursts in the two cells were synchronous and the 4 cell became hyperpolarized at the time of occurrence of each burst. This result is interesting because there is no sign of second network input on the 4 cell record, implying that the first network is capable of oscillating on its own and does not necessarily require the second network to switch it off.
Connexions between the SO interneurone and the N1 interneurones
In the previous paper (Rose & Benjamin, 1981) it was found that the slow oscillator interneurone could initiate first and second network input to the follower cells. Having identified first network interneurones, it should be possible to find out if there is a direct connexion between the SO and the N1 interneurones. In Fig. 6, steady depolarization of the SO interneurone activates the 4 cell and a cell on the middle channel which appears to be an N1 interneurone. The interesting finding is that the SO produce 1:1 facilitating e.p.s.p.s on the middle channel cell. We have previously shown that the SO may produce facilitating i.p.s.p.s on the 4 cell and facilitating e.p.s.p.s on the 1 cell (Rose & Benjamin, 1981). In this particular case the initial p.s.p. amplitudes on the middle channel cell are very small, so that summation takes place slowly at first. There is therefore a delay before the middle channel cell reaches threshold. When this cell reaches threshold it found that the spikes retain a 1:1 correspondence with SO spikes. Furthermore, there is evidence that this cell is an interneurone, because the simultaneous occurrence of SO spikes and spikes in this cell result in a sharp increase in the amplitude of the i.p.s.p.s on the 4 cell in the ‘SO + N1’ phase, implying that both neurones are converging on the 4 cell. The burst in the middle channel cell is simultaneous with a slowing in the instantaneous firing frequency of the SO, and finally both cells switch off together. The termination of bursts in these cells is simultaneous with the inhibitory notch in the 4 cell which arises from N2 interneurone activity (see later).
The earlier results on the N1 interneurones showed that they typically discharge at high frequency, whereas the cell shown in Fig. 6 is firing action potentials which are Simultaneous with those in the slow oscillator. If our argument about the synchronization of spikes in the SO and the cell shown in Fig. 6 is correct (with the cell being an N1 interneurone), this suggests that in this particular preparation the SO may have been driving this cell alone. Otherwise we would have expected to have seen a compound i.p.s.p. on the 4 cell as a result of discharge of driven N1 interneurones. We therefore tried to record the same connection under conditions in which the N1 network was more excitable (Fig. 7). Fig. 7 shows clearly that the SO causes 1:1 facilitating e.p.s.p.s on two cells which reach threshold and burst with exactly the right timing for them to be N1 interneurones. Simultaneous with the bursts in these two cells, the 8 cell receives first network inhibitory input in the form of a compound i.p.s.p. These two cells again switch off at the same time as the SO interneurone, and at the same time as the 8 cell receives second network inhibitory input.
Although these recordings show a growth in amplitude of successive e.p.s.p.s in N1 cells we cannot state unequivocally that this involves facilitation. It is possible for instance that the effect arises as a result of anomalous rectification in N1 cells. Quantitative analysis of the cycles shown in Fig. 7 (Hindmarsh & Rose, 1980) has shown, however, that the model of facilitation proposed by Mallart & Martin (1967) will accurately predict the time course of summation of these e.p.s.p.s, and the occurrence of this effect at all synapses arising from the SO interneurone provides some additional proof that it may be true facilitation.
There are several other features on the recording of Fig. 7 which are of interest. Thus, following the second network input, the 8 cell starts to fire (Benjamin & Rose, 1979), and during the burst it receives an additional inhibitory input (Rose & Benjamin, 1981). This input also appears on the SO interneurone and on the presumed N1 interneurones, and lasts for a considerable fraction of the feeding cycle (e.g. in 4s cycle). We will show later that this inhibitory input is caused by activity in N3 interneurones. Another point is that when the passage of depolarizing current into the SO is stopped, the presumed N1 interneurones continue to burst and the 8 cell continues to receive first, second and third network inhibitory inputs. The N1 burst in the freely-running oscillation is considerably longer (e.g. 2 s) than those occurring when the SO is driving the feeding cycle (e.g. 1 s), and eventually die out. The SO is therefore able to initiate an oscillation in other interneurones which persists for several cycles once the SO is switched off.
The summation of the facilitatory p.s.p.s on N1 cells is shown in more detail in Fig. 8(a). It was commonly found that the first few impulses in the SO were not followed by a p.s.p. in the N1 interneurone, but after that the p.s.p.s followed in a 1:1 fashion. Also in Fig. 8(a) note that several N1 cells are firing and this results in a compound i.p.s.p. on the 8 cell. In another preparation (Fig. 8(b) we found a cell which was similar to that shown in Fig. 6 in that the spikes were 1:1 with spikes in the SO and the i.p.s.p.s on the 8 cell again followed the two cells in a 1:1 manner and were discrete potentials. This cell did not have any facilitating e.p.s.p.s arising from the SO input.
N2 interneurones
Criteria for identification
When we come to consider the second network interneurones the first question which we have to ask is how do we distinguish them from first network interneurones ? Certainly there are likely to be differences in burst timing and the degree of coupling to other neurones of the same type, but assuming for the moment that a given cell was silent, can we distinguish between the two types simply by firing the cell artificially and looking for postsynaptic effects ? From our previous work (Benjamin & Rose, 1979), we define first network (N1) interneurones as those having the following connexions to motoneurones: ‘i’ to 3, 4, 5, 8, 9 cells/ ‘e’ to 2, 6, 7, 10 cells. By contrast second network (N2) interneurones should have the following connexions: ‘i’ to 4, 5, 7, 8 cells/‘e’ to 2, 3, 9, 10 cells. From this we can conclude that only 3 and 7 cells receive inputs of opposite polarity from the two networks (9 cells may receive some second network inhibitory input (Rose & Benjamin, 1981, Fig. 3) and are difficult to record anyway because of their ventral location). It would seem therefore that we have a very straightforward method of distinguishing N1 from N2 interneurones simply depolarize the cell and observe whether it produces inhibition or excitation of the 3 cell for instance. However, although we initially used this line of reasoning it became clear when third network (N3) interneurones were discovered that this test was not sufficient. Thus N3 interneurones excite the 3 cell and inhibit the 7 cells just as the N2 interneurones do. Therefore the only reliable test to distinguish an interneuronal type is to firstly observe the interneurone in an active patterning state are then depolarize it and look for postsynaptic effects. In the results which follow we will differentiate between those results where the identification is conclusive and those in which several types of interneurone could be involved.
Postsynaptic effects
There are a number of small cells located around the 3 cell which when depolarized have the effect of exciting the 3 cell. In the example shown in Fig. 9(a), three of these interneurones were penetrated at the same time as a 3 cell. Depolarization of each in born resulted in a long-lasting (3–4 s) depolarizing wave on the 3 cell. Their effects were additive since firing all three together (Fig. 9(a) results in a more steeply-rising and larger amplitude depolarization of the 3 cell than that caused by firing any one cell individually. In this experiment no evidence was obtained of coupling between these three cells, although similar records have been obtained in which there was weak excitatory coupling between pairs of cells of this type. In the light of remarks made in the previous section they could equally well be N2 or N3 interneurones on the basis of this experiment alone. In some cases (Fig. 9(b) interneurones were found with different membrane properties. The cell fired first in Fig. 9(b) was similar to the cells of Fig. 9(a), but the cell on the second channel had a much lower firing frequency and caused a two-component excitation of the 3 cell, these being an iN1tial rapid depolarization followed by the more usual slow response.
These small cells located around the 3 cell have a number of other postsynaptic effects which suggest that they may be N2 interneurones. In Fig. 10(b) a short burst of 3–4 spikes in one of these interneurones strongly excites the 3 cell as seen previously (Fig. 9), while the 4 cell is firstly inhibited and then recovers quickly to give a post inhibitory rebound waveform (see Benjamin & Rose, 1979, for discussion of P.I 1 and 4 cells). The interesting finding however is that there is a large amplitude hyperpolarization of the 5 cell. Recovery from this hyperpolarization is relatively slow and shows no post-inhibitory rebound. These relationships between 3, 4 and 5 cells are exactly those observed in the isolated ganglion when it is strongly patterN1ng and show that only a few spikes in certain interneurones are capable of producing very pronounced effects on the follower cells. Firing of these cells also causes e.p.s.p.s of up to 5 mV amplitude on the 2 cell (Fig. 10(a)). Benjamin & Rose (1979) have previously noted that the second network excitatory input to the 2 cell is of larger amplitude than the first.
Evidence that the connections from these interneurones to 3 and 4 cells is monosynaptic is given in Fig. 11. Here the cell on the lower-channel causes 1 :1 i.p.s.p.s. on a 4 cell and the usual prolonged depolarization of the 3 cell. When TEA is allowed to diffuse from the electrode into the cell (Fig. 12b), the action potentials of the inter-neurone become prolonged and the i.p.s.p.s. on the 4 cell increase in amplitude. Also the 3 cell depolarization rises more rapidly and the first spike in the interneurone is followed by a more pronounced e.p.s.p. than in Fig. 11 a. The iN1tial large amplitude of the first N2 spike in Fig. 11(b) is an artifact: See Figure legend.
Spontaneous activity
In order to confirm that these cells are second network interneurones we need to show that they fire during the second phase of the cycle. We also want to know something about the inputs to these cells. This type of information can only be obtained by observing their activity during spontaneous feeding cycles (Fig. 12a, 13). In Fig. 12 the preparation was producing patterned activity initially, which gradually deteriorated into non-patterning activity. This is typical of the sequence of events which happens during a recording period of . A presumed N2 interneurone was recorded on channel 3, together with 3, 4 and 5 cell. Evoked activity in this neurone later in the experiment (Fig. 12b) produced inhibition of 4 and 5 cells and excitation in the 3 cell. We conclude that this cell could be an N2 interneurone with relatively weak effects on the follower cells (compare with the interneurone of Fig. 10 b). The interesting finding in regard to this cell is that when the ganglion is producing patterned activity (Fig. 12a), this cell receives an inhibitory input during phase 1 (i.e. synchronous with inhibition on the 3 and 4 cells and the first phase of inhibition on the 5 cell), and an excitation during phase 2. These results suggest that the N2 interneurones may be receiving inhibitory input from N1 interneurones, after which they immediately recover and fire. But the cell of Fig. 12(a) is not firing, and we certainly need evidence that cells of this type are capable of firing during phase 2 of the feeding cycle. This is provided by the recordings of Fig. 13. In Fig. 13(a) the interneurone on the middle channel inhibits the 7 cell and the main 4 cell when polarized (see three bursts on left, note that the inhibition of the 4 cell is confusing because of the presence of electrotonic p.s.p.s). When the cell is spontaneously active (Fig. 13 a fourth burst, Fig. 13b), it receives a burst of i.p.s.p.s simultaneous with excitation of the 7 cell, and then recovers rapidly from this inhibition and fires a strong burst. The i.p.s.p.s on the interneurone appear to be 1:1 with i.p.s.p.s on the 4 cell and e.p.s.p.s on the 7 cell, suggesting that they are driven by a common N1 interneurone(s). Simultaneous with the burst in the interneurone the 7 cell is strongly hyperpolarized. The ‘ei’ sequence of synaptic inputs to the 7 cell is typical of a feeding preparation, and the recording suggests that the interneurone on the middle channel is partly responsible for the ‘i* component on the 7 cell and the second ‘i’ input to the 4 cell (lower channel).
Connexions from N2 interneurones to the SO interneurone
We have previously observed that the slow oscillator (SO) interneurone is switched off during phase 2 of the feeding cycle (Rose & Benjamin, 1981). Since the N2 interneurones fire during this phase, it is important to discover whether they can inhibit the SO. Findings on this point are given in Fig. 14, with more detail of the p.s.p.s in Fig. 15. In the resting condition of non-feeding preparations the SO is silent (Fig. 14), and under these conditions a burst of impulses in the N2 interneurones produce a weak excitation of this cell (Fig. 14, first burst in N2). This is shown in greater detail in Fig. 15(a), in which it can be seen that each spike in the N2 interneurone is followed by 1:1 i.p.s.p.s in the 4 cell and 1 :i e.p.s.p.s. in the SO cell. The 3 cell of Fig. 15(a) is again strongly excited by the interneurone burst.
When the SO is depolarized so that it fires, the e.p.s.p.s produced by N2 interneurone stimulation become reversed to hyperpolarizing potentials. This is shown detail in Fig. 15(b) (before arrow). In Fig. 14, from another preparation, we see that when the SO is firing, a burst in the N2 interneurone results in excitation of the 3 cell and a pronounced but delayed inhibition of the SO. It could be argued that these inhibitory inputs to the SO might be arising spontaneously, but in a control experiment in the same preparation as shown in Fig. 14 in which the SO was depolarized without firing the N2 interneurone, it was found that the SO was only receiving intermittent inhibition, and was not bursting in the idealized cyclic fashion (Rose & Benjamin, 1981). Second network (N2) interneurones therefore appear to inhibit the SO interneurones.
N3, interneurones
Rose & Benjamin (1981) reported the presence of another input to the follower cells besides that arising from the SO, N1 and N2 interneurones. The 4 cell recording on the upper channel of Fig. 16(a) in fact shows all four inputs particularly clearly, Immediately following each 4 cell burst the SO input comes onto the cell for about it, this input consisting of a succession of discrete i.p.s.p.s occurring at low frequency.
This is followed by a compound i.p.s.p. which further hyperpolarizes the cell for a period of about . This compound i.p.s.p. is caused by activity in N1 and SO interneurones and is simultaneous with a smooth hyperpolarization of the 3 cell and a depolarization of the 7 cell. Immediately following the N1 input the 4 cell starts to depolarize, probably as a result of a post-inhibitory rebound mechanism (Benjamin & Rose, 1979). The post-inhibitory rebound may be reinforced by a synaptic input from the cerebral giant cell (McCrohan & Benjamin, 1980b). This recovery is interrupted by a third inhibitory input arising from N2 interneurones, and simultaneously the 7 cell is strongly hyperpolarized. Normally the 3 cell would be strongly excited at this point, but for some reason this is not occurring in Fig. 16(a). When the 4 cell finally discharges, the burst is interrupted at about 100 ms intervals by inhibitory potentials. Thus the 4 cell does not burst continuously as we previously thought (Benjamin & Rose, 1979) but the bursting is broken up into about 8 short bursts of 50–200 ms duration each. Simultaneous with the inhibitory potentials on the 4 cell, inhibitory potentials also occur on the 7 cell and excitatory potentials occur on the 3 cell. Examination of the cell shown in channel 3 of Fig. 16(a) shows that it receives two successive inhibitory inputs when the N1 and N2 interneurones are presumed to be firing, and then recovers and fires a burst. This burst consists of about 8 sub-groups of variable amplitude spikes. If we look at one of the cycles recorded at a higher speed (Fig.16) it becomes clear that the spikes on the cell of channel 3 are exactly simultaneous with the inhibitory potentials on the 4 and 7 ells and with the excitatory potentials on the 3 cell. Furthermore, if the cell on channel 3 is depolarized (Fig. 16b) it causes inhibition of the 4 cell and excitation of the 3 cell. We conclude that this is a separate interneuronal type. It is not clear at the present time whether there is more than one interneurone of this type, although it is possible that the variable amplitude spikes are the result of electrotonic conduction of spikes from other members of the network, as occurs in the buccal ganglion of Planorbis (Berry, 1972). It is also interesting that the e.p.s.p.s which this interneurone type causes on the 3 cell (Fig. 16a) appear to to be the same as those produced on the 3 cell by firing the cerebral giant cell (McCrohan & Benjamin, 1980b, Fig. 8). This suggests that the cerebral giant cell is capable of exciting the N3 interneurone(s). It can now be seen why it is difficult to separate N2 and N3 interneurones simply by depolarization of the cell, since both N2 and N3 interneurones inhibit the 4 cell and cause a long-duration excitation of the 3 cell. The only clear way of distinguishing the two types is on the basis of spontaneous activity, for the N3 interneurones receive two phases of synaptic input whereas the N2 interneurones receive only one. Also, the N3 interneurones fire immediately after the N2 interneurones. The significance of these timing relationships will be discussed the next section.
Summary of interactions
The evidence given here and in the preceding paper (Rose & Benjamin, 1981) suggest that the four types of interneurones (SO, N1, N2 and N3) are connected as shown in Fig. 18(a), and discharge is illustrated in Fig. 18(b). Before we evaluate the evidence for the particular network illustrated, let us simply consider the firing pattern which such a network could generate (Fig. 18b). The cycle begins with a burst in the SO interneurone which excites the first network (N1) cells via a facilitating synapse. When the depolarization of several N1 interneurones reaches the threshold or spike initiation some form of positive feedback operates within the N1 network to Produce a distinct burst. During this burst the N2 and N3 interneurones are inhibited by the N1 interneurones. Then, by some unknown mechanism (see below) the N1 bursts suddenly switch off. This has the effect of releasing the N2 interneurones from inhibition, and they fire by rebound excitation probably reinforced by positive coupling within the N2 network. The N2 burst inhibits the SO interneurone and the N3 interneurones. The N2 interneurones fire at high frequency initially and gradually adapt until the N2 burst terminates rapidly releasing the N3 interneurone(s) from inhibition. The N3 interneurone(s) now discharge by rebound excitation and cause the inhibitory waveforms on the SO and N1 interneurones. Finally the N3 interneurones switch off and the SO begins to fire again to initiate another cycle. During the cycle the follower cells receive inputs from the various interneurones. Two representative examples are shown in Fig. 18(b) (3 and 4 group cells). Thus the 4 cell receives inhibitory inputs from all four components of the interneuronal network which has been discussed in detail with reference to Fig. 16, while the 3 cell receives inhibition from N1 interneurones and excitation from N2 and N3 interneurones.
Now we must consider the evidence for this scheme, and whether it has any weaknesses. Firstly, it should be emphasized that we have only shown the miN1mum number of connexions needed to explain our results. There are 12 possible inter-connexions between the four components of the network, and each of these connexions could be either inhibitory or excitatory, or could even be electrical, so there are 36 possible types of connexion between our four components. Any given network will probably have a maximum number of 12 single interconnexions, so that in our circuit (Fig. 18a) five are missing. Of the connexions illustrated we have only clearly demonstrated two. These are the excitatory connexion from the SO to N1 interneurones (Figs. 6–9) and the inhibitory connexion from the N2 interneurones to the SO (Figs. 14 and 15). The inhibitory connexions from N1 to N2 and N3 are inferred from the firing patterns (Figs. 12, 13, 16, 17) as are the inhibitory connexions from N2 to N3 and that from N3 to N1 and SO (Figs. 16, 2, 7, 8). The difficulty of proving some of these connexions lies partly in the fact that in many cases we are dealing with groups of interneurones, so that depolarization of one cell activates the whole group and unitary p.s.p.s cannot be distinguished in the cells to which this group is connected. Nevertheless further recording from the appropriate pairs of cells is technically possible and our scheme provides the basis for further work. In relation to the five connexions not shown in Fig. 18(a), we have obtained partial evidence that the SO can excite N2 and N3 interneurones although this is at a low level even if it exists. Our recordings provide no evidence of a connexion from N3 to N2. This leaves the connexion from Nr to SO and from N2 to N1 to be considered. We have noted several times that the instantaneous frequency of the SO decreases as the N1 interneurones start to burst, which suggests that the N1 interneurones might be inhibiting the SO. Similarly the N1 interneurones switch off rapidly when the N2 interneurones fire suggesting that the N2 interneurones might be inhibiting the N1 cells. In fact this raises the difficult question of what is happening at the critical point at which the SO and N1 cells switch off and the N2 cells switch on. There are several possibilities to consider. Firstly it is possible that unidentified interneurones are involved. Thus there might be an interneurone which switched N1 cells on and off or which switched N2 interneurones on and off. Secondly it is possible that the N1 interneurones burst spontaneously either individually or as a group. Evidence for spontaneous oscillation has been given in Fig. 5. If the N1 interneurones did oscillate spontaneously, then they would be the interneurones which drive the feeding rhythm, with the N2 oscillation being entrained and the SO providing modulation of the N1 rhythm. It should be noted that in this case it is still possible that the N2 interneurones could have inhibitory feedback to the N1 interneurones. This would simply reinforce the spontaneous termination of N1 burst.
A third possibility is that the SO interneurone might drive the N1 cells, which in turn feed back inhibitorily to the SO (i.e. reciprocal inhibition). In view of the fact that N1 cells oscillate spontaneously with the SO switched off (Fig. 7, Fig. 6 (Rose & Benjamin, 1981)) this seems unlikely. Therefore, unless we assume the existence of other interneurones, we are forced to the conclusion that the most likely explanation for the generation of the rhythm is that it originates from activity in the N1 interneurones. This will be reconsidered in the Discussion.
DISCUSSION
The Network
In Fig. 18(a) we presented a network which represented the most likely way in which the buccal interneurones are interconnected. There are several features of this network which need further explanation. Firstly, the network diagram does not distinguish between single cells and groups of cells. Although we have chosen to represent each component as a circle to avoid excessive complication, it is clear that the SO is a single cell in each buccal ganglion, whereas the N1 and N2 interneurones appe.ar to be groups of cells. The N3 interneurone(s) may also be a group of cells, although we have not yet recorded from two of them simultaneously. Also we have not represented the positive feedback which exists between N1 interneurones and probably also between N2 interneurones. This again was a deliberate omission since we do not at present know the substructures of these networks. It is possible for instance that the N1 group also contains inhibitory interneurones which feedback on other members of the group producing recurrent inhibition within the N1 network. Another simplification which we have introduced is the representation of connexions between SO, N1, N2 and N3 by single lines. This would be accurate for connexions between single cells, but we are dealing with interconnexions between subpopulations, of neurones and it is clear that any given cell may connect with a number of cells in the next network, and in some cases a cell in one network may not be connected to a cell in another network even though there is an arrow connecting the two networks in our diagram. We have also assumed that most interneurones are multiaction. For instance the second network (N2) interneurones excite the 3 cell and inhibit the 4 cell, N3 and SO. Alternatively it is possible that the N1, N2 and N3 cell types could each be orgaN1zed as two subsets of interneurones, one subset being inhibitory and the other excitatory to follower cells. The cyberchron network of Helisoma appears to be orgaN1zed like this (Kater, 1974). At a more general level it should also be noted that Fig. 18(a) is only one of a number of alternative ways of representing the circuit. We have chosen this form of representation because it gives the impression of propagation down a chain on neurones with recurrent feedback operating. In fact we could have equally represented the circuit as a closed loop involving N1, N2 and N3 cells (i.e. recurrent cyclic inhibition) with the SO modulating activity in the loop. Such a circuit could not be so easily related to the activity pattern however, and it would be somewhat of a distortion to consider N1-N2-N3-N1 as forming a loop because the N3 to N1 inhibitory connexion is functionally not very effective, whereas N1, N2 and N3 are clearly members of a chain.
The output pattern
From our results it is clear that the SO interneurone is capable of driving the feeding rhythm by way of N1 interneurones. Yet in the preceding paper (Rose & Benjamin, 1981) we showed that the SO interneurone was not necessary for the generation of feeding cycles, and we have therefore been forced to consider either that other interneurones are involved or that the buccal network is in some way spontaneously rhythmic. Each of these points needs to be considered in more detail.
Other interneurones
As we have already remarked, either the N1 or N2 networks could be driven directly by other unidentified interneurones. These unientified interneurones could be endogenously active and be the true originators of the feeding rhythm. Alternatively the buccal interneurones could be involved in feedback connections to unidentified interneurones possibly located in the cerebral ganglion. McCrohan & Benjamin (1980 a) have shown that the cerebral giant cells in Lymnaea burst during phase 1 of the feeding cycle and have postsynaptic effects on buccal motoneurones (McCrohan & Benjamin, 1980b). One of the effects which they demonstrated was that excitation of the cerebral giant cell activates slow and fast e.p.s.p. responses in the 3 cell. From the present work we know that slow and fast e.p.s.p.s in the 3 cell originate from activity in N2 and N3 interneurones, so that the possibility of an input from the cerebral giant cell to the buccal interneuronal network is realistic. McCrohan (personal communication) has recently discovered another interneurone in the cerebral ganglion which is similar to the slow oscillator interneurone in its modulation of feeding cycles. It is therefore possible that buccal-cerebral loops may participate in the termination of the N1 burst. Reciprocal connexions involving interneurones in the cerebral and buccal ganglia of Pleurobranchia have been demonstrated by Davis and colleagues (Davis, 1977).
Rhythmic N1 and N2 activity
In our previous work (Benjamin & Rose, 1979; Rose & Benjamin, 1979) we noted that the Lymnaea buccal ganglion appeared to be more complicated than the closely-related Helisoma described by Kater and co-workers (Kater, 1974; Kaneko, Merickel & Kater, 1978). From the present work we now know that the two phases of synaptic input described previously in Lymnaea (Benjamin & Rose, 1979) arise from activity in N1 and N2 networks. However, if we assume that the N1 interneurones are spontaneously active, are driving the N2 interneurones, and are modulated by the SO interneurone, then the conclusion which may be reached is that the fundamental generator of the feeding rhythm in Lymnaea is the spontaneous oscillation of N1 interneurones. In this respect the Lymnaea and Helisoma systems seem to be basically similar. Thus in Helisoma the rhythm is generated by a network of electrically-coupled ‘cyberchron’ interneurones. Activity spreads through these interneurones by a positive feedback mechanism and is terminated by an unknown process (Merickel, Kater & Eyman, 1978). This primary oscillator of the Helisoma system seems to be analogous to our N1 interneuronal network, within which there is also electrical coupling which leads to bursting through positive feedback. In this sense our work raises the same questions as in Helisoma-namely, how does an electrically-coupled network generate a bursting rhythm ? At this stage there does not seem to be any difficulty in understanding in general terms how the bursts switch on, in that models of electrically-coupled networks will initiate bursts through a positive feedback mechanism (Merickel, et al., 1978). The main problem is how do the bursts terminate ? This is in fact the problem which we referred to earlier in saying that the N1 bursts terminate by an unknown process. One possibility that has been put forward is that synchronization of action potentials reduces junctional shunting and thereby increases the time constants in the population. This has the effect of prolonging synchronous afterpotentials (Getting & Willows, 1978). Such synchronization of action potentials has not been observed among N1 interneurones. Perhaps a more likely explanation would be that within the N1 population there are cells which cause inhibition of the rest of the network. According to this possibility excitation would spread among N1 interneurones with the inhibitory cells being activated towards the end of the burst period. Another possible explanation is that one or more of the N1 interneurones produces bursts endogenously which drives the rest of the network. Other explanations have been discussed by Kaneko, et al., (1978) and reviewed by Berry & Pentreath (1977). In earlier work (Benjamin & Rose, 1979) we gave evidence that the cells responsible for the second phase of synaptic input might be spontaneously active. We now know that this input originates from activity in N2 interneurones. Our observations, particularly on fatigued preparations support the view that the N2 interneurones are capable of spontaneous oscillation on their own, without being driven by N1 interneurones. In many preparations we observed that early in the recording period, the N1 and N2 networks were active and locked together with N2 following N1. The N1 network frequently dropped out after leaving the N2 network which was still rhythmically active. This suggests that both networks are capable of spontaneous oscillation but that normally the N1 network entrains the N2 network.
Relationship to other systems
We have already commented on the similarities between the rhythmic N1 interneurones in Lymnaea and the cyberchron neurones in Helisoma. A further feature of the Lymnaea system is that the N1 interneurones drive N2 and N3 interneurones. The bursts in the N2 and N3 interneurones could result from a mechanism related to postinhibitory rebound (PIR) found in other systems (Perkel & Mulloney, 1974), although this simple interpretation is complicated by the inherent rhythmicity of the N2 interneurones. The network could therefore be summarized as a system with two successive stages of rebound driven by an oscillator (N1 interneurones). Considered, in this way an interesting comparison can be drawn between the generation of the feeding rhythm in Lymnaea and the pyloric rhythm of the lobster stomatogastric ganglion (Hartline & Maynard, 1975; Maynard & Selverston, 1975; Selverston et al., 1976). In the lobster pyloric rhythm, cell type PD is endogenously active and inhibits PY and LP cells just as the rhythmic N1 network in Lymnaea inhibits N2 and N3. In the lobster, termination of the PD burst as followed by recovery of LP and PY neurones. The LP recovery is more rapid than PY, so that onset of the LP burst reestablishes inhibition of the PY set and also the PD set. Similarly in Lymnaea the recovery of the N2 cells is more rapid than N3 so that the onset of the N2 burst reestablishes inhibition of the N3 cells and probably contributes also to inhibition of N1. In the lobster, adaptation of the LP cells eventually leads to release of the PY cells, which then inhibit LP and release PD from inhibitory input. In Lymnaea there is a similar adaptation of N2 cells which eventually release N3 cells, although the N3 cells do not appear to inhibit N2 cells once they are bursting, and differ from the lobster pyloric rhythm slightly in the N3 cells inhibit N1 cells. The principles of organization of these two rhythms therefore have some similarities, although the buccal system is more complicated because of certain additional endogenous mechanisms which result in rapid recovery of the N2 cells, and because of the additional SO interneurone which drives and receives feedback from the rhythm generator.
This comparison raises the question of whether the N1-N2-N3-N1 loop should be considered as an example of recurrent cyclic inhibition as discussed in general terms by Friesen & Stent (1978). It is clear, however, that the loop does not operate in the manner proposed by Kling & Székeley (1968). This is mainly because in their model the crucial factor is that recovery from inhibition is slow for each component of the network, whereas in Lymnaea recovery is rapid and the N1 and N2 subnetworks are also rhythmically active. This reinforces the general conclusion that in the understanding of the output patterns of rhythmic networks, endogenous properties must be considered of equal importance to the connectivity itself. From a theoretical point of view however the buccal network on Lymnaea introduces the idea that the rhythm generator may feedback to the source of external driving input, in this case the slow oscillator interneurone.
ACKNOWLEDGEMENTS
R. M. R. was supported by an M.R.C. grant to P. R. B.