The Relationship of the Central Motor Pattern to the Feeding Cycle of Lymnaea Stagnalis

In the two preceding papers (Benjamin & Rose, 1979; Benjamin, Rose, Slade & Lacy, 1979) the electrical activity and distribution of axons of identified neurones in the Lymnaea buccal ganglia were described. The object of this paper is to prove which of the identified cells are motoneurones, and to show how their activity relates to movements of the buccal mass.

From the point of view of the morphology, the Lymnaea buccal mass is an excellent system to study, since accounts of the musculature and feeding movements have already been given by Carriker (1946), Hubendick (1957), and more recently by Goldschmeding & de Vlieger (1975). From the electrophysiological point of view the system also has the advantage that the muscles produce action potentials during contraction. This means that muscle activity can be recorded extracellularly in feeding preparations. In this paper much emphasis will be placed on neurone and muscle recording from the actively feeding preparation, since many of the cell types can only be identified on the basis of their synaptic inputs during feeding cycles. In other molluscs, such as Aplysia, the buccal muscles do not exhibit action potentials (Cohen, Weiss & Kupfermann, 1978), making a similar analysis more difficult. The Aplysia muscle fibres are of larger diameter, however, so that intracellular recording is possible. At this stage we have not been able to record intracellularly from the small muscle fibres of Lymnaea, but have tried to determine the functional role of identified cells in relation to extracellular recording of muscular contraction. These recordings show that in each feeding cycle there are three phases of muscle activity followed by an inactive phase. Motoneurones have been identified for each of the active phases. In addition a functional division occurs in that single input cells are involved in the control of the buccal mass itself, while double input cells may control gut movement and secretion.

All of the recordings in this paper were made on the semi-intact preparation. This consisted of the buccal mass, oesophagus, and attached brain and buccal ganglia. Intracellular recordings from neurones were made as described in the earlier paper (Benjamin & Rose, 1979). Extracellular recordings were made from the buccal muscles using glass suction electrodes in the conventional way. All recordings were made from pairs of muscles or from neurones and muscles on the same side. In order to record from neurones in the buccal ganglia during feeding movements, the buccal ganglia were supported on a small metal table covered with a thin layer of Sylgard. The table was placed underneath the buccal ganglia, care being taken not to damage any of the buccal nerves. This technique usually had the effect of stretching the buccal nerves slightly as the buccal ganglia were pulled away from the buccal mass. The ganglia were held on the table using only one or two micropins, a minimum amount of stretch being applied to the ganglia themselves, since this may uncouple the interneurones which drive the feeding cycle.

In some cases the buccal mass was partially dissected. In order to explore the innervation of the large anterior jugalis muscle the buccal mass was cut in a dorsal longitudinal direction and the two halves of the anterior jugalis were spread out (see Benjamin et al. 1979; fig. 2). To look at the innervation of the buccal sphincter, a lateral longitudinal cut was made in the anterior jugalis near the point where the lateral buccal nerve enters this muscle (Fig. 1), and the anterior jugalis pulled aside with fine forceps to expose the underlying muscles. Similarly the problem of the separation of posterior jugalis and tensor innervation was investigated by making a dorsal longitudinal cut across the top of the posterior jugalis, and folding this muscle back to expose the radula tensor muscles. In some cases the posterior jugalis was isolated with its thin posterior jugalis nerve and attached buccal ganglia.

Cinephotography was carried out on the semi-intact preparation using a Bolex 16 mm reflex camera at 25 frames/s. The feeding movements were investigated both with and without simultaneous extracellular recordings from the buccal muscles. Feeding movements were sometimes induced by dropping crystals of sucrose close to the buccal mass. When simultaneous cinefilm and extracellular recordings were made, the two recordings were synchronized by noting the frame number as the camera was switched on and off. Operation of the camera switch was arranged to cause a steady d.c. deflexion of the extracellular trace, and the times of ‘on’ and ‘off’ were measured from the time base on the extracellular recording. This time interval was then checked with the time interval on the film, knowing the number of frames between ‘on ‘and ‘off’ and the frame speed. If these two measurements agreed satisfactorily, the record was subdivided and the frames corresponding to specific points in the feeding cycle were projected and drawn. This method is accurate in the present case because the feeding cycle itself is relatively slow, lasting 3–10 s, and the method is accurate to within 2–3 frames (or 0·1 s).

A comprehensive description of the structure of the buccal mass in Lymnaea stagnalis has been given by Carriker (1946). Using Carriker’s account as a basis, Goldschmeding & de Vlieger (1975) regrouped the muscles into four concentric muscle systems. From a physiological point of view our main problem has been to distinguish between protractor and retractor muscles, and we have approached this problem by defining the activity of each muscle rather than looking at the functional anatomy of the system. All we need to know initially is how the muscles are arranged externally, and how the underlying muscles lie in relation to these. Fig. 1 should be consulted for the general layout of the muscles of the buccal mass, and Fig. 6 shows the approximate position of the odontophore during different phases of the feeding cycle.

The muscles of the buccal mass are organized around the radula and its supporting cartilage, the odontophore. The odontophore consists of two symmetrical spoonshaped structures, united ventrally, and the radula is stretched over its surface, lying in the U-shaped groove between the two cartilages and running posteriorly into the bulbous radula sac (Fig. 6). To understand the muscular arrangement it is best to imagine the odontophore as a large oval-shaped structure seen in side view with its ventral base acting as a fulcrum about which it is rotated during the feeding cycle. Many of the buccal muscles are inserted at this fulcrum point. The unpaired posterior and anterior jugalis muscles are inserted at this point on each side and pass over the top of the buccal mass, the anterior jugalis enclosing the anterior two-thirds and the posterior jugalis the posterior one-third of the buccal mass. It is important to note that while the anterior jugalis is a large thick muscle, the posterior jugalis is only a thin muscle sheet, which besides covering the posterior end of the buccal mass also projects anteriorly beneath the anterior jugalis, suggesting that it is a protractor muscle (Carriker, 1946). Also inserted at the fulcrum are the three paired tensor muscles which wrap around the posterior edges of the odontophore to be inserted on the subradular membrane. Contraction of these muscles stretches the radula over the odontophore during rasping. Of these tensor muscles the largest is the supralateral radula tensor which forms the two large posterior-lateral bulges behind each fulcrum. It is these tensor muscles which are covered by the sheetlike posterior jugalis.

If the anterior jugalis is dissected away, the buccal sphincter can be seen underneath. This muscle encircles the buccal cavity, as does the mandibular approximator which lies just in front of it. There are two important muscles which lie under the buccal sphincter - the dorsal and ventral odontophoral flexors, which insert on the upper and lower (fulcrum point) parts of the cartilage (Goldschmeding & de Vlieger, 1975). We have been unable to record from these muscles, because they form wide thin bands which are difficult to dissect from the thick overlying anterior jugalis, but it is interesting to note that Carriker (1946) assigns to the flexor muscles the function of protraction, and according to his description protraction is brought about by the combined contraction of the posterior jugalis and flexor muscles.

A small group of strand-like muscles are inserted ventrally below the fulcrum, and project anteriorly to insert under the mouth. These include the pre- and postventral protractors and the postventral levators (Fig. 1). Although the postventral levators serve to raise the posterior end of the buccal mass on the forward stroke of the odontophore, it seems that the pre- and postvential protractors are actually retractor muscles in that they pull the ventrally placed fulcrum point forwards during retraction, thus helping in the backward rotation of the odontophore. The buccal retractor muscles themselves are also inserted laterally at each fulcrum point (Fig. 1), and are extrinsic muscles retracting the buccal mass as a whole.

It can be seen therefore that all the larger muscles concerned with the movements of the odontophore are inserted at or near the fulcrum point. Of the remaining muscles the dorso-lateral protractor is the largest of four small muscle pairs involved in turning the buccal mass in the anterior direction, this muscle arising near the oesophagus and running across the sides of the buccal mass to insert in the mouth region (Fig. 1). The posterior end of the radula continues into the bulbous radula sac which is supported dorsally by a group of four small suspensory muscles (Goldschmeding & de Vlieger, 1975; Carriker, 1946). Not included in this description are four muscles which open and close the mouth, and appear to be supplied by the labial nerves (Carriker, 1946), and also such minor muscles as the intra-cartilage tensors and the tensor of the collostylar hood.

From a physiological point of view it is convenient that the most important muscles involved in the feeding cycles can be recorded externally without the need for dissection. However, the buccal sphincter and flexor muscles can only be recorded by dissecting away the anterior jugalis. Although we have attempted to record from these muscles we are not satisfied that we have isolated them from muscle strands of the anterior jugalis, and will not therefore include any information on these muscles in the physiological account. We have also grouped together muscles of similar functions such as the ‘tensor’ (3 tensor muscles), ‘suspensors’ (4 suspensory muscles) although a more complete account would have to differentiate between different muscles in these groups.

In simple terms the feeding cycle consists of protraction and retraction of the radula. In reality it will be shown below that there are two phases of retraction and that there is usually a period of inactivity between the end of retraction of one cycle and the start of protraction of the next. In Lymnaea the feeding cycle therefore consists of four separate phases; protraction (P), retraction (phases R1 and R2), and inactivity (I).

Electromyographic recordings were made with glass suction electrodes placed on the appropriate muscles. Recordings were always made from pairs of muscles on the same side of the buccal mass. The tip diameter of the electrodes was up to 100 μm, so that the activity of large numbers of muscle fibres were being recorded. In relation to the surface area of the more important muscles, this tip diameter is still relatively small. There are certainly variations in the activity of different parts of some muscles (e.g. anterior jugalis, see below) and our electrode size was a compromise between not losing detail of the overall muscle activity and not having too much overlapping activity from a number of motor units. It is nevertheless true to say that we are not absolutely clear at this stage why some muscles have unitary potentials over a large area (e.g. upper anterior jugalis), whereas others produce potentials of a compound shape (e.g. tensors). It appears that the unitary potentials may be produced by single neurones branching over a large area (a single motor unit), and the compound potentials by simultaneous activity in many muscle fibres innervated by different neurones (several motor units). An alternative explanation might be that the unitary potentials result from highly synchronized activity in several neurones supplying the same area.

The overall configuration of the feeding cycle can be distinguished by recording from the anterior and posterior jugalis muscles simultaneously (Fig. 2). A point of confusion immediately arises when we refer to recordings from the posterior jugalis muscle. This is because the posterior jugalis is a thin muscle sheet covering the large tensor muscles, and the suction electrode unavoidably sucks up both the posterior jugalis and some of the underlying tensor muscles. It is difficult to prove which part of the recorded activity is coming from which muscle, although we are now fairly certain of the distinction (see later). For the moment, we will refer to a recording made above the posterior jugalis as a ‘posterior jugalis/tensor’ recording.

The cycle begins with a burst in the posterior jugalis/tensor which we later show to be the first (or protraction, P) phase of the feeding cycle. Muscle potentials occur at high frequency during this burst and there may be several overlapping units. Retraction in the feeding cycle consists of two phases, beginning with a burst in the lower part of the anterior jugalis (phase Ri; Fig. 2 b), followed by a burst of similar duration in the upper part of the anterior jugalis (phase R2; Fig. 2a). Muscle potentials of the lower part of the anterior jugalis have a compound shape and appear to be made up of the summed activity of a number of motor units. By contrast the burst in the upper part of the anterior jugalis has large unitary potentials occurring at very high frequency at the onset of the burst, suggesting the recording of a single motor unit.

In Fig. 2(c) two suction electrodes were positioned close together near the middle of the anterior jugalis, and it can be seen that at this recording site some of the muscle potentials of the lower part are transmitted to the other electrode which is primarily recording the anterior jugalis upper burst. As the electrode is moved more anteriorly only the second high frequency burst is recorded. This suggests that near the middle of the anterior jugalis there is some overlap of motor units supplying the upper and lower parts of the muscle. It seems unlikely that the upper burst could be coming from the underlying sphincter muscle because of the thickness of the anterior jugalis.

A burst of variable intensity occurs in the posterior jugalis/tensor simultaneous with the lower anterior jugalis burst (Fig. 2b). This second phase of posterior jugalis/tensor firing is usually much larger in amplitude than the first burst (Fig. 2b), although the relative magnitudes of the two bursts depend on the position of the electrode, the second burst tending to be larger the nearer the electrode is to the supramedian radular tensor. Evidence which will be given later suggests that the first burst is due to the contraction of the posterior jugalis muscle, and the second large amplitude burst to the contraction of the large tensor muscles. The variation in amplitude of these two component bursts with the position of the recording electrode is illustrated in Fig. 4(c), where the first burst alone is clearly recorded by the upper channel electrode. This first burst is partially recorded by the other electrode which was placed over the supramedian radular tensor, but in this position the predominant feature is the large amplitude second burst. The explanation therefore appears to be that the tensor burst follows the posterior jugalis burst, and since it terminates before the onset of the burst in the upper part of the anterior jugalis (Fig. 4b), we may conclude that it is synchronous with the first retraction burst (R1) in the lower half of the anterior jugalis.

Each protraction-retraction cycle is separated by a period of virtual inactivity in the jugalis muscles (the I phase). It will be shown later that several identified neurons discharge during this quiescent period, and that some of these project to the oesophagus. It seems likely that this period is associated with peristaltic activity in the oesophagus following the feeding cycle of the buccal mass.

Further detail of the feeding cycle is provided by determining the relationship of activity in other muscles to the protraction-retraction cycle of the jugalis muscles. The postventral protractor muscle discharges at the same time as the tensor burst (Fig. 3 c), and is therefore a retractor muscle, as Kater has also pointed out in Helisoma (Kater, 1974). Futher confirmation of the timing of this burst is given by recording it with the upper anterior jugalis burst (Fig. 3 e), which it precedes. The dorso-lateral protractor and the suspensor muscles of the radula sac fire bursts which are in phase with the protraction burst of the posterior jugalis (Fig. 3 a, d), the suspensor muscles discharging at similar frequency to the posterior jugalis burst, while the dorso-lateral protractor burst is composed of low frequency and large amplitude potentials. Finally, one or two large amplitude potentials occur in the buccal retractor muscle during the second phase of retraction (R2), as is shown by recording this muscle with the anterior jugalis upper burst (Fig. 3b).

The cycle period is defined as the time interval between the onset of one phase of protraction (posterior jugalis burst), and the next. There are considerable variations in the cycle period, some extreme examples being shown in Fig. 4. Typically the cycle is 10 s total duration, but in extreme cases the cycle period may be as short as 5 s (Fig. 4b). The most variable component is the duration of the inactive period (I), which is of the order of 1 s in Fig. 4b), compared with as long as 20 s in other cases (Fig. 2 a). The next most variable component is the protraction phase, although this is not particularly evident in the extracellular recordings shown here. When we come to look at the protraction-retraction cycle in more detail it will be seen that the protraction phase is sometimes very short (Fig. 11 c), and at other times long (Fig. 12 a). Such variations are not seen in the set of paired extracellular recordings (Figs. 2–4) because the short-duration protraction phases are associated with poorly patterning preparations which have been rejected here. The retraction phases (R1 and R2) are relatively constant.

Frequently the cyclical bursting behaviour occurs for periods of a minute or so, punctuated by periods of inactivity or rather random firing typical of the non-feeding preparation. It is possible to initiate feeding cycles by dropping crystals of sucrose close to the buccal mass (Goldschmeding & Jaeger, 1973), and an example of activity initiated in this way is shown in Fig. 4(c). After application of sucrose (arrow) there is intense burst activity in the posterior jugalis and tensor muscles, which slowly declines over 8 or 9 cycles. Such initiated sequences should be compared with the exponentially decaying sequences initiated by food extract in Aplysia depilans (Rose, 1976) and Archidoris (Rose, 1971).

The sequence of muscle potentials is summarized in Fig. 5 for muscles which are accessible to recording without further dissection of the buccal mass. It was found difficult for instance to separate the odontophoral flexor and buccal sphincter muscles from the overlying anterior jugalis muscle and still retain an actively feeding preparation.

During the first (or protraction, P) phase of the cycle the most important feature is the burst in the posterior jugalis. The suspensor muscles of the radula sac, and the dorsolateral protractors are also active during this phase. It is also probable that the dorsal odontophoral flexor (not recorded) acts in conjunction with the posterior jugalis in everting the radula to the mouth (Carriker, 1946).

Retraction is divided into two phases; R1 and R2. During R1 the tensor muscles, anterior jugalis lower, and postventral protractor muscles are active. During R2 the upper anterior jugalis and buccal retractor fire. Between the end of R2 and the beginning of the next cycle is the inactive phase (I) in which no buccal muscles participate. Variations in the cycle period are associated with differences in the durations of the inactive and protraction periods. These variations will be shown later to be associated with variations in the durations of the underlying postsynaptic potentials of the motoneurones.

The feeding cycle in most molluscs involves two main types of movement, namely rotation of the buccal mass relative to the body (using extrinsic muscles), and movement of the radula relative to the buccal mass (using intrinsic muscles). In laboratory- reared animals, whose body wall is semi-transparent, the rotation of the buccal mass within the body cavity can be clearly seen. However, we have only filmed movements of the isolated buccal mass, in which the extrinsic muscles which suspend the buccal mass and the muscles round the mouth have been cut. This description is therefore primarily concerned with radula movements within the buccal mass. A further constraint is that because the muscles round the mouth have been cut, the radula does not emerge properly from the mouth, and details of the rasping movements cannot be distinguished. Nevertheless, the filmed sequences of movements are useful because they define the relationship between muscle activity and the different phases of protraction and retraction.

A complete feeding cycle is shown in Fig. 6. In the resting state the postero-ventral edges of the two halves of the odontophore bulge out ventrally, and the posterior jugalis can be clearly distinguished (shaded region). The approximate orientation of the odontophore within the buccal mass is indicated by a bar line, and the oval-shaped odontophore cartilage has been drawn schematically on each diagram to help explain the external shape changes. During the slow protraction phase (frames 1–4) the buccal mass changes to an ovoid shape as the radula is pushed towards the mouth, and the area occupied by the posterior jugalis is progressively diminished. In frames 3–4 the leading edge of the radula can be seen pressing against the dorsal wall of the buccal mass anteriorly at the end of the protraction phase/start of the rasp movement. It appears that the two halves of the odontophore become fairly closely approximated at this stage, explaining the ovoid shape of the buccal mass. It will be shown later on that the diminishing area of the posterior jugalis is associated with posterior jugalis muscle activity. It is also worth noting that during this protraction phase there is evidence of a forward thrusting of the whole buccal mass which corresponds to the forward rotation of the buccal mass seen in the intact animal.

Retraction (Fig. 6, frames 5–9) is a continuous process, in which the radula is moved backwards and downwards, and its leading edge is rotated towards the oesophagus. As the radula is moved backwards and rotated, the radula sac appears ventro-posteriorly (frame 6). The radula sac enlarges and swings forwards as retraction continues. This results in the radula assuming a vertical orientation (frame 8) compared with the near-horizontal position at the start of retraction (frame 5). The radula is then rotated backwards beyond the vertical position, and its leading edge can be seen pushing up against the roof of the buccal mass at the point where it joins the oesophagus (frame 9). Finally the radula returns to the resting position (frame 10).

Although the sequence shown in Fig. 6 illustrates the main features of the cycle, it does not enable us to distinguish between the two phases of retraction (R1 and R2). By filming the dorsal surface of the buccal mass it is possible to watch the backward movement of the radula by the indentation that its leading edge makes in the dorsal wall of the buccal mass (Fig. 7). A distinct V-shaped line can be seen as the radula moves backwards. This backward movement occurs as two separate phases, which correspond to phases R1 and R2 of the muscle recordings (see below). The first phase involves the rotation of the radula to the vertical position (Fig. 7, frames 3-5). This slow phase R1 is followed by R2, a powerful and very rapid continuation of the backward rotation of the radula (frames 6-8). It has been necessary to space the frames closer together to detect this rapid movement. This second phase of retraction (R2) is accompanied by contraction of the upper part of the anterior jugalis which forces the radula towards the oesophagus (frame 8), completing the cycle.

As a final confirmation of our interpretation, six cycles were filmed and the activity of pairs of muscles recorded simultaneously. A typical example is shown in Fig. 8 with recordings from the posterior jugalis/tensor and anterior jugalis upper. As expected, the initial posterior jugalis burst discharges during protraction (Fig. 8, frames 1, 2). The activity which follows is associated with the tensor burst (phase Ri, frames 3, 4), and involves the beginning of retraction with the emergence of the radula sac. The muscle potentials from the tensor muscle are much smaller than usual in this recording. The radula is in a near-vertical position at the end of the R1 phase. However, the most dramatic event is the powerful contraction of the anterior jugalis, whose electrical activity corresponds with a marked narrowing of the anterior half of the iuccal mass. This is the rapid movement referred to previously, and its rapidity is ‘reflected in the high frequency of the anterior jugalis upper burst (phase R2). It is this rapid movement which forces the vertically positioned radula backwards, delivering food to the oesophagus. Recordings of this type therefore prove that the two phases of retraction (R1 and R2) correspond to a slow R1 followed by a fast R2 backward rotation of the radula.

Having established the pattern of muscle activity, the next problem is to relate activity in identified neurones to muscle recordings. This subject will be introduced by giving evidence for 1:1 nerve and muscle activity to the anterior jugalis and posterior jugalis/tensor muscles in the non-feeding preparation. The finding of a 1:1 relationship between an identified neurone and muscle does not necessarily prove that the neurone is a motoneurone. It is possible, for instance, that the neurone is electro- tonically coupled to another cell which is the motoneurone for that muscle. While we have observed electrotonic coupling as a common organizational feature in the buccal ganglion we have only rarely observed such 1:1 following of action potentials between electrotonically coupled cells. Usually the coupling is not sufficiently strong for such following to occur. Even if such coupling did occur, it could at least be stated that some elements of a given set of synchronously firing neurones were motoneurones. The main problem is that of relating the activity of the identified groups of neurones (Benjamin & Rose, 1979) to identified muscles. In this respect it is important to demonstrate the relationship between nerve and muscle activity in the actively feeding system. The non-feeding preparation can give information on visually identified cells such as the 4-group cells, but it is necessary to have an actively feeding preparation to be able to identify other cell types on the basis of their synaptic inputs (see Benjamin & Rose, 1979). In this analysis we demonstrate this relationship for buccal cell types 3-8 as identified in the previous paper (Benjamin & Rose, 1979), cell types 1 and 2 not being included, since the anatomical work has shown that their axons go exclusively to the salivary glands and oesophagus respectively and not to the buccal musculature (Benjamin et al. 1979). Where possible some comments will be made on the ability of the muscle to follow at different frequencies, but generally emphasis has been placed on determining the overall nerve-muscle relationship rather than discussing specific details. In all recordings (Figs. 9–13) muscle activity was recorded on the same side as the neurone.

In the non-feeding preparation there is usually a low level of nerve and muscle activity, and sometimes many of the neurones are silent. Consequently it is easier to show 1:1 firing since the muscle potentials are not obscured by background spike activity. The limitation is that it is only really possible to identify the 3 cell and 4- group cells with certainty, since these cells are usually visually identifiable, and often have a characteristic inhibitory input followed by rebound excitation. Even in nonfeeding preparations the inhibitory inputs to the 4-group cells are often of sufficient amplitude to give rise to a short post-inhibitory burst.

The anterior jugalis muscle is innervated by 4-group snd 8 cells. Of these cells the most striking results were obtained from the main 4 cell, which was observed many times to have a 1:1 relationship of nerve to muscle action potentials (Fig. 9c). A number of 4 cluster cells also innervate the same muscle (Fig. 9 a). By exploring the surface of the anterior jugalis with a suction electrode it was shown that there was no consistency in the areas innervated by specific cells from animal to animal. In particular the main 4 cell sometimes covered a wide area, while in other preparations it was fairly restricted. A representative example showing the spread of the innervation from three 4 cluster cells is shown in Fig. 9(a). The only generalization which we can make is that the main 4 cell usually supplies a much larger area of this muscle than any of the 4 cluster cells. During feeding the anterior jugalis undergoes quite complex movements, and it is interesting that 4-group cell bursts have different times of onset (Benjamin & Rose, 1979). This will presumably lead to differential contractions of different parts of the muscle. However, a given pattern of spatial innervation cannot ba distinguished presumably because there are variations in the depths at which the fibres run, the recordings being made from the muscle surface.

The posterior jugalis muscle is a thin sheet overlying the large tensor muscles. A suction electrode placed over the posterior jugalis invariably sucks up some of the tensor muscles as well. It is easy to demonstrate that the main 4 cell and some 4 cluster cells produce 1:1 muscle potentials on this posterior jugalis/tensor combination (Fig. 10a, c). We have observed small cells in the region of the 5 cell (Fig. 10b, d) and a large number of small cells in the region of the buccal commissure which also innervate this area. The main question which arises is whether the 4-group cells innervate both muscles. Anatomically the posterior jugalis is thought to be innervated only by the thin posterior jugalis nerve (Carriker, 1946), and our back-injections of this nerve show that only one or two small cells in the buccal ganglion project down this nerve (Benjamin et al. 1979). By contrast the tensor muscles are innervated by a branch of the ventrobuccal nerve according to Carriker (1946), and it has been shown previously that a number of 4-group cells project down this nerve (Benjamin et al. 1979). It seems likely therefore that the 1:1 firing of 4-group cells on to the posterior jugalis/ tensor combination is actually due to the innervation of the tensor muscles by the 4-group cells. Evidence that this is in fact the case may be obtained by cutting the posterior jugalis dorsally and folding it back to expose the tensor muscles. The main 4 cell can then be shown to cause 1:1 following of muscle potentials recorded either |rom the tensor muscles directly or from the tensor muscles recorded through the posterior jugalis. No 1:1 following is observed by recording directly from the posterior jugalis. It could be argued that the ventral buccal nerve might also innervate the posterior jugalis, and that fine branches to this muscle had been cut in the dissection procedure. It is clearly very difficult technically to be absolutely certain that there is no innervation of the posterior jugalis from 4-group cells. Later we will show that 6 cells also innervate the posterior jugalis/tensor. In this case it seems likely that the 6 cells constitute the small cells which project down the posterior jugalis nerve to the posterior jugalis muscle. One way to demonstrate this in the non-feeding preparation would be to isolate the posterior jugalis and attached posterior jugalis nerve and buccal ganglion, and to record from the 6 cells to show their 1:1 innervation of this muscle. We have not concentrated on this experiment because of the difficulty of finding and identifying 6 cells, but it might be technically feasible.

The easiest cells to record from in the feeding preparation are the large 4-group and 3 cells. A common feature in proteased preparations is that the burst recorded over the supra-median radular tensor (plus posterior jugalis) often has a large amplitude spike-like deflexion at the onset, followed by a high-frequency burst of muscle potentials. The 4-group and 3 cells discharge at the same time as this burst, but an examination of the 3 cell recorded at high speed (Fig. 11) reveals that there is no 1:1 relationship between the high-frequency 3-cell burst and the complex muscle potentials of the tensor muscle. Recordings from a 4 cell (Fig. 11 b) and a 4 cluster cell (Fig. nc) together with the posterior jugalis/tensor, not only show that these cells, fire at the same time as the high-frequency tensor bursts, but also show that thq inhibitory wave occurs at the same time as the first burst in the posterior jugalis. It is nevertheless difficult to base the alignment of the neural sequence with the muscle sequence on information of this kind, because i.p.s.p.s in 4-group cells often occur with little change in membrane potential (Benjamin & Rose, 1979).

Much more information is given by recording muscles with the double input 5, of and 8 cells, since there are more clearly defined transition points in a cycle.

The recording of Fig. 12(a) from the 5 cell and the posterior jugalis/tensor is particularly illustrative because the first and second bursts of the posterior jugalis/ tensor are of large amplitude. The first point is that the 5-cell burst occurs during the inactive (I) phase. The first-phase inhibition on the 5 cell is simultaneous with the first (protraction) burst in the posterior jugalis/tensor, and the second phase of inhibition occurs during the first phase of retraction (R1) simultaneous with the second or ‘tensor’ burst of the posterior jugalis/tensor. The 5 cell recovers from inhibition during phase R2 and fires once more during the I phase. The 5-cell burst itself is not therefore synchronized with any activity in the buccal musculature. This is in agreement with the anatomical findings (Benjamin et al. which showed that the 5 cell sends a single axon to the dorsobuccal nerve. It seems likely that the 5 cell, like the 2 cell, is concerned with oesophageal movements following the transfer of food to the oesophagus at the end of the feeding cycle.

The results for the 5 cell are compatible with a similar recording from the 8 cell (Fig. 12b). Here the first (protraction burst) is of small amplitude because of the position of the electrode. This first burst (phase P) occurs at the same time as the first phase of inhibition on the 8 cell. The second inhibitory wave-form is synchronized with the ‘tensor’ burst (phase Ri) as in the 5 cell. Although the p.s.p.s are very similar in the 5 and 8 cells, the 8 cell shows rebound excitation immediately following phase R1, and fires during phase R2. This suggests that the 8 cell is involved in the contraction of the upper part of the anterior jugalis. The 8 cell also has an axon in the lateral buccal nerve, which supplies this muscle (Benjamin et al. 1979). Confirmation of this prediction is given by the recording shown in Fig. 12(d) from the 8 cell and the upper anterior jugalis in a feeding preparation. Although it can be clearly seen that the 8 cell causes large amplitude muscle potentials in this muscle following a period of inhibition, the recording is confusing because the first phase of inhibition on the 8 cell is very short in duration (Fig. 12 d, arrow). We confirmed that there was a very short first phase of inhibition by recording the 8 cell with a 4 cluster cell, which has only the first inhibitory input. The recording of Fig. 12 (d) therefore appears as a succession of second inhibitory phases, followed by rebound excitation. The 1:1 relationship of 8-cell activity to muscle potentials in the upper anterior jugalis is shown in more detail in Fig. 9 (b).

The 7-cell activity also ties in with that of the 4, 5 and 8 cells (Fig. 12 c). The typical acceleration of the 7-cell burst occurs during the protraction phase, the 7 cell receiving excitation at the same time as the 5 and 8 cells receive the first phase of inhibition. Although it is possible that the 7 cell could generate part of the first (protraction) burst in the posterior jugalis, this seems unlikely, since there was no 1:1 correspondence of nerve and muscle potentials in Fig. 12(c) even when this cell was firing at low frequency. The excitation on the 7 cell is followed by a second phase of inhibition during phase R1 of the muscle cycle, as found for 5 and 8 cells. While recording from a 7 cell the whole surface of the buccal mass has been explored with a suction electrode, including the dissected buccal sphincter muscle. Since no innervation to any muscle was found we suggest that the 7 cell may supply the oesophagus. Goldschmeding, Bruins & Everts (1977) show a cell ‘β’ next to the 3 cell which projects down the dorsobuccal nerve only, and from its size and position it is possible that this is the 7 cell.

Having established the timing of the neural and muscle sequences in terms of double input cells, we can re-examine the single input cells. Since the 6 cells receive excitation simultaneous with the first phase of inhibition of 3, 4, 5 and 8 cells, and are not synaptically driven at any other part of the cycle, it seemed likely that that the 6 cells could cause the first phase protraction burst in the posterior jugalis. That this is the case is proved by the recordings of Fig. 13(a), (b). In Fig. 13(a) the 6 cell fires at low frequency in a 1:1 correspondence to small muscle potentials in the posterior jugalis/tensor. When this preparation is showing patterned feeding cycles (Fig. 13 b) the 6 cell receives strong excitation during the protraction phase of the cycle. Unfortunately, because the 6 cell discharges at such a high frequency, individual muscle action potentials fuse together to form a poorly distinguishable burst of small amplitude, and it is no longer possible to distinguish the 1:1 correspondence. It is clear, however, that the 6-cell burst and the first burst in the posterior jugalis correspond. The 1:1 firing and the fact that the cell fires in exactly the phase which we would predict for the 6 cell is convincing evidence that 6 cells are involved in the protraction phase.

A final detail may be added in regard to the main 4 cell. This cell fires in 1:1 correspondence with the lower anterior jugalis at low frequencies (Fig. 13 c). During feeding cycles (Fig. 13 d) the inhibition on the 4 cell corresponds with the protraction phase. In this particular preparation the 4 cell receives a second phase of synaptic inhibition (Fig. 13 d, arrow) like the 8 cells. This leads to delayed firing of the 4 cell relative to the 4 cluster cells, and has been remarked on previously (Benjamin & Rose, 1979). The delay is such that the 4 cell may contribute partially to contraction of the anterior jugalis during phase R2, in conjunction with the 8 cell. Comparison of the similar muscle recordings in Fig. 13(A) and (d) from two different preparations shows that the 4 cell receives inhibition as the 6 cell receives excitation during the protraction phase of the cycle, entirely in accord with previous findings (Benjamin & Rose, 1979).

It is now possible to describe how the burst activity of cell types 1-8 are correlated with the activity of the muscles described earlier. The neurones which have been proved to have a 1:1 relationship to buccal muscles are shown in Fig. 14, while the alignment of the remaining neurones with the feeding cycle is given in Fig. 15. Only a few representative muscles are shown in Fig. 14, and only the phases of the cycle itself in Fig. 15.

The cycle begins with a burst in the 6 cells (Fig. 14), which causes a burst in the posterior jugalis (first burst in posterior jugalis/tensor), and protraction of the radula. Undoubtedly there are more neurones involved in protraction since the suspensor, flexor and dorso-lateral protractor muscles also contribute to this phase. Although we have found other candidate neurones which burst in this phase their identification as motoneurones is complicated by the fact that intemeurones also discharge during this period, and we have not been able to demonstrate the 1:1 relationship as done for the 6 cells.

The retraction movement is divided into two phases (R1 and R2). In phase R1 the 4-group cells fire, causing the contraction of the lower part of the anterior jugalis and the tensor muscles. These muscles cause the rasp movement, the anterior jugalis contracting differentially to produce a backward movement, and the tensor muscles stretching the radula over the odontophore. The postventral protractor muscles also contract, moving the fulcrum towards the mouth and thus helping to swing the odontophore to a vertical position.

Phase R2 is primarily caused by the burst in the 8 cell, which causes contraction of the upper part of the anterior jugalis. This forces the odontophore beyond the vertical position so that food can be released into the oesophagus. The 4 cell may also contribute to this movement. The buccal retractor also pulls the whole buccal mass backwards during R2. Finally the odontophore returns to the resting position, and stays there until the start of the next protraction phase.

In Fig. 15 the remaining cell types are aligned with the feeding cycle. The physiological and anatomical (Benjamin et al. 1979) evidence suggests that none of these cells are involved with the control of buccal muscles. The 1 cells are probably involved with salivary gland activity as in Helisoma (Kater, 1974), and the 3 cells may also supply glandular cells in the walls of the buccal mass and oesophagus (Benjamin et al. 1979). Anatomically, the 2, 3, 5 and probably also the 7 cells, supply the gut (Benjamin et al. 1979). The alignment of neural activity to the feeding cycle shown in Fig. 15 is based on recordings of 5 cell (Fig. 12a), 3 cell (Fig. n a), and 7 cell (Fig. 12c) activity, and the remaining 1 and 2 cells are shown in their usual temporal positions relative to 3, 5 and 7 cells (Benjamin & Rose, 1979). Bursting in the 5 cell is clearly restricted to the |nactive period (I), and this activity may be combined with short bursts in the 2 cell (Benjamin & Rose, 1979) to produce oesophageal contractions.

The object of this paper was to relate the pattern of neural activity of identified neurones in the buccal ganglion to the sequence of muscular contractions during feeding. It would be very difficult to give a complete account of this relationship because the buccal mass has 46 separate muscles (Carriker, 1946). However, most of these muscles are arranged as symmetrical pairs, many occur in groups having similar functions (e.g. 4 suspensor muscles, 3 paired tensor muscles), and this number also includes small muscles which play an insignificant part in the feeding cycle. A further simplification arises when it is realized that most of these muscles discharge in separate phases of a 4-phase feeding cycle (Fig. 5). The main result of this paper is that activity in 6, 4 group and 8 cells are components of the protraction, R1 and R2 phases respectively, and activity in the 5 cell occurs in the inactive phase (I). By obtaining examples of burst activity in these cells during feeding cycles, it has been possible to align the neural cycle with the muscle activity cycle (Figs. 14, 15). It is also interesting that the distinction between single and double input cells discussed in the first paper (Benjamin & Rose, 1979) seems to have a functional significance. Thus the single input 6 and 4 groups cells are involved in controlling buccal mass movements, whereas the double input 2, 3, 5 and 7 cells send axons to the gut (Benjamin et al. 1979). The exception to this scheme is that the 8 cells are double input cells which supply the upper part of the anterior jugalis causing contraction during phase R2. However, the contraction of the upper part of the anterior jugalis is essentially the initiating phase of peristalsis in the oesophagus, and in this sense the 8 cell is also a double input cell whose activity relates to gut function.

If the single and double input cells serve different functions, can they be active as independent groups ? The hypothesis put forward in the first paper was that the two phases of input might result from activity in two oscillatory networks of intemeurones, with the first network being capable of entraining the second. In isolated ganglia we have often observed double input cells which go into long periods of firing in which only the second input is present, suggesting that the proposed second oscillator is capable of independent activity. It is possible that a series of feeding cycles could be followed by activity in the gut alone, and that this could result from independent bursting of the second oscillator which drives the 2, 3, 5, 7 and 8 cells only.

As we have remarked in the introduction, Lymnaea differs from Aplysia in that the muscles appear to generate action potentials. In certain buccal muscles in Aplysia contractions are produced by summation and facilitation of excitatory junction potentials (Orkand & Orkand, 1975; Cohen et al. 1978) and the muscles contract in a graded fashion proportional to the size of these junctional potentials. By contrast the buccal muscles in Helisoma produce action potentials, this being confirmed by intracellular recording from the posterior jugalis muscle (Kater, Heyer & Hegmann, 1971). In relation to the innervation pattern we have found that several motoneurones can innervate the same muscle (e.g. anterior jugalis), and that certain motoneurones can innervate more than one muscle (e.g. 4 cell to anterior jugalis and tensor muscles (see also Goldschmeding, 1977)). In Helisoma different motoneurones also innervate different parts of the anterior jugalis (Kater, 1974), while in Aplysia the lower extrinsic protractor (Orkand & Orkand, 1975) and accessory radula closer (Cohen et al. 1978) also receive more than one excitatory axon.

Another important result is that activity in the semi-intact preparation is very similar to that recorded from the isolated ganglion, implying that feeding cycles are generated centrally. But how is this activity initiated and maintained, and is it modified by proprioreceptive feedback ? It is certainly possible to initiate sequences in the semiintact buccal mass using sucrose (Fig. 4c), but at this stage we do not know whether this provides a phasic or slowly decaying stimulus to the pattern generator, or whether the chemosensory stimulus acts by way of higher order interneurones. The fact that isolated buccal ganglia can produce periods of intense cyclical feeding suggests that feeding can also be initiated internally. In Helisoma, Granzow & Kater (1977) have shown that feeding cycles can be initiated by passing current into certain interneurones in the cerebral ganglia. Gillette & Davis (1977) have also demonstrated feedback from the buccal ganglia to cerebral neurones in Pleurobranchea. The relationship between the buccal and cerebral ganglia in Lymnaea is at present being investigated in this laboratory (McCrohan, unpublished observations). In relation to proprioreceptive input, Kater & Rowell (1973) have shown that stretching of mechanoreceptors in the wall of the buccal mass of Helisoma evokes e.p.s.p.s in retractor motoneurones and i.p.s.p.s in protractor motoneurones, and suggest that the mechanoreceptors cut off protractor activity before the inhibitory input to the protractors (Kater & Rowell, 1973; fig. 13). Kater (1974; fig. 23) seems to have modified this idea, in that he shows the mechanoreceptor input as being simultaneous with the inhibitory input to protractors from the intemeurones. By comparison all that we can say with certainty is that in the isolated buccal ganglion of Lymnaea the 4-group cells terminate sharply before the inhibitory input from intemeurones, suggesting that mechanoreceptors are not important in limiting the duration of 4-group bursts at high cycle frequencies.

Two previous studies are directly relevant to the present work. Goldschmeding & de Vlieger (1975) have discussed the feeding cycle of Lymnaea stagnalis based on anatomy and visual observation, while Kater (1974) has based his description of the closely related pulmonate Helisoma trivolis on electromyographic and neurone recording. The major difference between our work and that of Goldschmeding & de Vlieger is that we put much greater emphasis on the rasping and retraction phases (R1 and R2) as active phases of the cycle, whereas the latter authors regard retraction as a passive relaxation of all muscles, with 19 of the 28 muscle types involved in protraction. This difference arises primarily because Goldschmeding & de Vlieger claim that the anterior jugalis contracts during protraction and that the tensor muscles also stretch the radula over the cartilage during this phase. By contrast we have shown that the anterior jugalis contracts strongly during the two retraction phases (R1 and R2) and that the tensor muscles contract during Ri. In other respects the two descriptions are similar - for instance, during protraction we both agree that the dorso-lateral protractor, suspensory muscles, and posterior jugalis are active, and during the return of the radula to the resting position the posterior odontophoral protractor and buccal retractor contract. Goldschmeding & de Vlieger’s interpretation was partly based on the finding that the main 4 cell supplied both the anterior and posterior jugalis, which should therefore contract together. We claim that the posterior jugalis recording is largely tensor muscle activity, and that it is the tensor and anterior jugalis which contract together in retraction.

On comparison of our results with Kater’s findings in Helisoma, there is some similarity as regards the activity patterns (Kater, 1974; fig. 8). We both find that the posterior jugalis and dorso-lateral protractor are protractor muscles, the tensor is involved in rasping (our phase Ri) and the buccal retractor and anterior jugalis in retraction. Kater does not find two phases of anterior jugalis activity, and has the posterior odontophoral protractor muscle firing in the later phase of retraction (our R2). What is very puzzling is that Kater’s protractor motoneurone group have a marked hyperpolarization preceding the burst of action potentials, the retractor motoneurone bursts being in antiphase to these protractor motoneurones. This is exactly the opposite of the system in Lymnaea, where it is the retractor (4-group) motoneurones which have the hyperpolarization, and the protractor (6 cells) are in antiphase. Our description also differs in that we add the activity of double input cells (2, 3, 5, 7) and show that 2, 5 and 7 cells are active during the inter-feeding phase. It appears that the 1 cell is similar in controlling the salivary glands.

We thank the M.R.C. for financial support, and Mike Land for help with the cinephotography.