Extraglandular Innervation of the Salivary Cells of the Leech Haementeria Ghilianii: Neuronal Stimulation Elicits Gland-Cell Action Potentials and Secretion

The salivary cells of the giant leech Haementeria are grouped into two pairs of discrete glands which lie at the base of a long, muscular proboscis (Sawyer et al. 1982). The two large anterior glands each contain about 240 cells ranging in diameter from 400 to 1200 μm; the smaller posterior glands contain cells of about 100 μm in diameter. Each cell has a single process which leaves the gland and enters the proboscis, and release of secretory products appears to occur at the end of each process or ‘ductule’. Some ductules do not travel far from the gland whereas others reach the tip of the proboscis, a distance of more than 5 cm in large leeches.

The proboscis-salivary gland complex is supplied by a paired stomatogastric nerve from the brain, which enters near the proboscis base close to the attachment of the salivary glands. The gland-cell bodies produce overshooting, calcium dependent action potentials in response to depolarizing current applied through an intracellular microelectrode (Marshall & Lent, 1984; Jones et al. 1985; Wuttke & Berry, 1988). We report here that the gland cells are also activated by stimulation of the stomatogastric nerve but that the innervation of each cell is outside the gland, close to the ductule tip; impulses travel along the ductule from the tip into the cell body in the gland.

Specimens of Haementeria ghilianii were obtained from a breeding colony held at Biopharm (UK) Ltd, Swansea, where they were maintained in aquaria at 26°C. The aquarium water contained (mmoll⁻¹): NaCl, 0·63; CaCl₂, 0·07; MgCl₂ 0·07; K₂SO₄, 0·01; Tris maleate, 1 (pH6·0) (Sawyer et al. 1981).

The proboscis and attached salivary glands, together with the brain and first few ganglia of the ventral nerve cord, were removed from the animal after cutting through the oesophagus. The preparation (see Fig. 1) was pinned to stretch the two sheaths which surround the proboscis (the sheaths are attached to the proboscis only at its base). The stomatogastric nerve, which runs between the sheaths from the brain to the base of the proboscis, was freed by cutting the outer sheath. The preparation was then transferred to a Perspex experimental bath (volume 1 ml) where it was pinned to a layer of Sylgard. Careful pinning was essential to avoid the dislodging of microelectrodes in gland cells by movements of the proboscis; pins were placed through the inner sheath and also through the endothelium which invests the salivary glands. Physiological saline flowed through the bath at about 3 ml min⁻¹; its composition (mmol 1⁻¹) was: NaCl, 125; KC1, 4; CaC12,1·8; glucose, 11; Tris maleate, 10 (pH7·4) at room temperature (18–22°C).

The techniques were similar to those described by Wuttke & Berry (1988). Briefly, individual gland cells were impaled with a bevelled, KCl⁻ filled microelectrode (10–20 MΩ) for recording membrane potential. A second electrode was sometimes inserted for passing current. Each electrode was connected to a Digitimer NL102 amplifier, and signals were monitored on a pen recorder (Brush 2200S) and stored on tape (Thorn EMI3000 FM tape recorder). Glass-tipped suction electrodes were used for stimulating the stomatogastric nerve or the tip of the proboscis.

Effects of some putative transmitters were examined by perfusion at known concentration, using a multiway valve (Holder & Sattelle, 1972). The following substances were applied: acetylcholine chloride, 5-hydroxytryptamine creatinine sulphate, dopamine hydrochloride and octopamine hydrochloride (Sigma).

The stomatogastric nerve was filled with cobalt to trace its pathways in the proboscis and in the brain. In isolated preparations of (a) supra- and sub- oesophageal ganglia plus the first few ganglia of the nerve cord, and (b) the proboscis and attached salivary glands, the cut end of the stomatogastric nerve was placed into a Vaseline trough filled with distilled water. The rest of the preparation was covered with saline containing 18 mmol 1⁻¹ MgCl₂ to relax the nerve, which otherwise tended to contract and pull out of the trough (there are muscle fibres in the sheaths of nerve trunks and of ganglia). The length of nerve was kept as short as possible to reduce diffusion distance. After 5 min the distilled water was replaced by aqueous 5 % cobalt chloride containing bovine serum albumin (0·013 g 100 ml⁻¹ solution; Strausfeld & Obermayer, 1976) or aqueous 6% hexammine cobalt (III) chloride (Sigma). The trough was closed with Vaseline and more saline added to the rest of the preparation. Preparations were kept at 8°C for 12–24h (brain) or up to 44h (proboscis).

The cobalt was subsequently precipitated for 15–30 min in 1 % sodium sulphide (w/v) in physiological saline (pH 7·4) and the preparation then washed in distilled water and fixed in Carnoy’s fixative. After silver intensification (Bacon & Altman, 1977) tissues were dehydrated in ethanol, cleared in methyl salicylate and embedded in Ralmount.

Secretory products are released at the outer surface of the proboscis tip and in the proboscis lumen. These secretions were separately sucked into a fine syringe needle and their proteins were resolved by SDS polyacrylamide gel electrophoresis under reducing conditions (Laemmli, 1970), using a Pharmacia Phast System and gradient gels (10–15%). The following relative molecular weight markers (Pharmacia) were co-electrophoresed: phosphorylase b, 94000; bovine serum albumin, 67 000; ovalbumin, 43000; carbonic anhydrase, 30000; soybean trypsin inhibitor, 20100; α-lactalbumin, 14400. Gels were subsequently stained with silver (Heukeshoven & Dernick, 1985).

In addition, individual gland-cell bodies were ruptured to extrude their cytoplasm which was collected in a syringe needle and subsequently analysed as above.

The preparation consisted of the proboscis and attached salivary glands, the brain and first few body ganglia, and the paired stomatogastric nerve which runs between two sheaths and enters the proboscis near its base (Fig. 1). The inner sheath encloses the proboscis and is attached at one end to its base and at the other (in the intact animal) to the mouth. The proboscis lies freely within this sheath and is able to extend several centimetres beyond its opening.

For further details of the anatomy see Sawyer et al. (1982) and Sawyer (1986).

When the stomatogastric nerve was back-filled with cobalt chloride its two main trunks were found to enter the brain and branch into the nerves to the mouth; processes also ran anteriorly into the sheath but there were no stained cell bodies within the brain. With hexammine cobalt chloride, however, extremely fine processes could be seen coming from the main trunks, and cell bodies in the brain and first segmental ganglion were stained. Some of the cells, which apparently form bilaterally symmetrical pairs, seem large enough to be identifiable and recorded with microelectrodes (Fig. 2). No processes were seen to pass back to other ganglia of the ventral nerve cord.

The two stomatogastric nerve trunks enter the proboscis near its base and each runs forward towards the tip. They could be traced for several millimetres but disappeared well before reaching the end of the proboscis. Processes could also be traced posteriorly to the level of the salivary glands but it was not possible to determine whether they entered the glands. Silver intensification was necessary to follow them this far and the procedure disrupted the glands. At the entry point of the stomatogastric nerve into the proboscis there is an intricate network of stained fibres and a cluster of stained cell bodies which lie in the inner sheath (Fig. 3). These cell bodies are not visible in living, unstained preparations but cells with the same characteristic pattern of distribution are clearly revealed by Methylene Blue. They do not take up Neutral Red, which tends to stain monoaminergic neurones (Lent, 1981).

The stomatogastric nerve thus contains axons of both central and peripheral neurones.

Various attempts were made to determine whether the action potentials which are produced in gland-cell bodies by applied depolarizing current (see Fig. 11) are conducted along the ductules. It would be extremely difficult to penetrate ductules with a microelectrode once they have entered the proboscis because they disappear amongst the musculature. Also, the proboscis produces continuous movements which may dislodge an electrode even from a cell body in a well-secured gland. Ductules are also difficult to trace within the gland itself because they form its core, surrounded by the cell bodies. Some ductules come close to the surface at various points, and these could be impaled and shown to produce action potentials. In no case, however, was it possible to trace a ductule for more than 3 mm from its cell body.

Extracellular recordings, using a suction electrode attached to the tip of the proboscis, failed to reveal any potential changes associated with spikes elicited in a cell body. If the tip of the proboscis was stimulated by current pulses, however, a subthreshold depolarizing response occurred in certain cells after each stimulus (Fig. 4A). These responses followed stimuli on a one-to-one basis up to frequencies of 15 min⁻¹ (Fig. 4B) although in some cells there were intermittent failures. Occasionally they disappeared altogether but could be reinstated by a brief period of high-frequency stimulation (Fig. 4B). The responses appear to represent invasion of the cell body by action potentials in the ductule, propagated from the tip. Evidence is presented below that the cell bodies are not innervated, and that these brief depolarizing responses are not synaptic potentials. In the following account they will be referred to as ‘ductule spikes’, i.e. intrasomatically recorded potentials which are produced by electrotonic spread of current associated with an action potential in the ductule. Their relatively small size presumably results from the large difference between membrane area of the ductule and of its cell body (ductules have a diameter of about 30 μm in the gland whereas cell bodies range up to 1200 μm and their membranes are greatly invaginated; Walz et al. 1988).

The typical configuration of a ductule spike recorded in the cell body is shown in

Fig, 4A. Amplitudes ranged from 2 to 30mV, occasionally reaching spike threshold for the soma. In some cells the amplitude remained constant during repetitive activation whereas in others there was wide variation. There was usually a small undershoot (0·5–2 mV). The latency of the response was 1–2s. It was difficult to estimate conduction velocity because ductules are coiled in the proboscis, but values of 5 cm s⁻¹ were obtained using the length of a fully stretched proboscis. Similar responses, with shorter latency, could be obtained after removing much of the proboscis and stimulating the remaining cut end. It was not possible to elicit responses to tip stimulation in all cells investigated, however, presumably because their ductule terminated nearer the base of the proboscis. All stimuli produced contractions of the proboscis.

Three lines of evidence indicate that the responses to stimulation of the proboscis tip are not synaptic potentials, (i) They persisted in saline containing 18 mmol 1⁻¹ Mg²⁺ to block synaptic transmission (Fig. 5). This high Mg²⁺ concentration prevented contraction of the proboscis produced by stimulation of the stomatogastric nerve, (ii) The responses were not potentiated by hyperpolarizing the gland cell with applied current; in some cells hyperpolarization to −110 mV produced little or no change in amplitude, and in others the response became much reduced (Fig. 6). The possibility was considered that they were synaptic potentials produced by a decrease in conductance, but there was no inversion even at −120 mV. The decrease in amplitude with hyperpolarization presumably resulted from a block of the incoming ductule spike further from the cell body, (iii) The responses could be blocked by collision with intrasomatically evoked impulses travelling in the opposite direction (Fig. 7). This indicates that they were produced by ductule spikes conducted from the tip (Parnas et al. 1969). An alternative explanation for these results is that the responses are EPSPs produced by a nonrectifying electrotonic synapse, but this is extremely unlikely: no gland cells are known to receive such input, and it would probably be ineffective in such large cells.

The ionic mechanism of ductule spikes was determined by substitution of extracellular sodium and by use of cobalt to block calcium channels. Action potentials in the cell body are calcium-dependent: they are unaffected by removal of external sodium and are reversibly blocked by 5mmoll⁻¹ cobalt (Marshall & Lent, 1984; Jones et al. 1985; Wuttke & Berry, 1988). Ductule spikes recorded in the cell body were not blocked by sodium removal (Fig. 8; sodium was replaced by N-methyl-D-glucamine). There was, however, an increase in the number of failures, whose cause is unknown; it did not appear to be related to the transient hyperpolarization produced by Na⁺ removal because failures continued to occur during the return to the original membrane potential, though the time course of the change in membrane potential could be different in the ductule. Addition of 5 mmol I⁻¹ cobalt usually produced intermittent failures of response rather than a complete block. Responses were reversibly abolished by 10 mmol 1⁻¹ cobalt but showed a slow recovery when cobalt was washed out (approx. 20 min) whereas recovery of somatic action potentials from 5 mmol 1⁻¹ cobalt takes only a few minutes (Wuttke & Berry, 1988). These differences in cobalt concentration and recovery time may be due to more restricted access of solutions to the ductules which are surrounded by muscle fibres in the proboscis. The results indicate that the ductule spike is Ca²⁺-dependent but they are not as clear-cut as those for the soma spike.

Gland cells were never observed to produce spontaneous action potentials or show signs of spontaneous synaptic input, even in preparations where the brain and first few body ganglia were left attached. If the stomatogastric nerve was stimulated electrically, single stimuli produced a contraction of the proboscis (and of the stomatogastric nerve itself) but no effect on salivary cell bodies. With repetitive stimulation of less than about 10 s⁻¹, the gland cells remained inactive while the proboscis became strongly contracted and subsequently produced vigorous peristaltic waves which started near the tip (the orifice at the tip usually remained open). Even without stimulation the proboscis was active; if unrestricted it waved around, became coiled and then uncoiled, and periodically contracted very strongly, but it rarely showed peristalsis. Above about 10 s⁻¹, repetitive stimulation of the stomatogastric nerve for 10–20 s produced in some gland cells responses which resembled ductule spikes elicited by stimulation of the proboscis tip (Fig. 9). Evidence is presented below that they are indeed ductule spikes and not synaptic potentials. The responses sometimes stopped after a few seconds until the nerve was stimulated again, and sometimes they continued for several minutes, occurring at frequencies up to 45 min⁻¹ (Fig. 10). There was a delay of at least 15 s between the beginning of stimulation and the first response, and delays of 1 min or more were recorded. Similar responses occurred if the nerve was cut and the distal end stimulated. Ductule spikes sometimes reached spike threshold for the cell body (especially the first in a series) but the responses were generally subthreshold. Considerable variation was found between responses to the same stimulation, even in a single cell; this was not correlated with any obvious movements of the proboscis. Ductule spikes produced by nerve stimulation were abolished by 18 mmol 1⁻¹ Mg²⁺ after about 20 min but they could still be elicited by stimulation of the proboscis tip after 1 h.

For a number of reasons, which are explained in the Discussion, it seemed likely that the salivary cells were innervated near the ductule tip rather than in the gland itself. In a cell whose ductule terminated at the end of the proboscis, therefore, the response to nerve stimulation should have been abolished by cutting the proboscis tip. It was indeed found that ductule spikes produced by nerve stimulation disappeared after removing about a 1 mm length from the proboscis tip, whereas stimulation of the cut end of the proboscis continued to produce ductule spikes on a one-for-one basis.

Confirmation that the responses to nerve stimulation were ductule spikes rather than synaptic potentials was more difficult than for the responses to stimulation of the proboscis tip. For example, the absence of a fixed relationship between stimulus and response meant that collision experiments could not be reliably performed. Hyperpolarization of the soma to −100 mV, however, produced little change, or a reduction in amplitude of responses to stimulation of the nerve, providing further evidence that they are not synaptic potentials.

The effects of acetylcholine (ACh), 5-hydroxytryptamine (5-HT), dopamine and octopamine were investigated by perfusion (3 ml min⁻¹). It was necessary to distinguish between effects on the ductule tip in the proboscis and effects on the cell body in the gland. Recordings from salivary-cell bodies were therefore made in a bath where the proboscis tip alone could be exposed to transmitter by means of a Vaseline partition which separated it from the glands (volume of the compartment containing the tip was 0·5 ml; the presence of a ductule at the tip was confirmed by the presence of a response to tip stimulation).

Of the four transmitter candidates, only ACh elicited spikes. At 10⁻⁶moll⁻¹ there was a small subthreshold depolarization but at 10⁻⁴moll⁻¹ a large, rapid depolarization resulted in 1–3 action potentials and a large drop in membrane resistance. The effect was transient, however, and membrane potential and resistance returned to normal after about 5 min exposure (Fig. 11).

Depolarizations were also produced by 5-HT, dopamine and octopamine, but these did not reach threshold even at concentrations of 10⁻⁴moll⁻¹. A marked effect on excitability was produced by 5-HT (and in some cells by dopamine).

Fig. 12 shows the typical depolarization and increase in membrane resistance produced by 10⁻⁴moll⁻¹ 5-HT. Suprathreshold depolarizing current now evoked prolonged spiking whereas before and after 5-HT application the response to the same current was a single spike (gland cells rarely produce more than three or four spikes when depolarized by applied current). The influence of 5-HT on excitability is not simply to depolarize the cell towards threshold and thereby increase the effect of the current; in the absence of 5-HT it is impossible to elicit prolonged firing by increasing the current or by applying prior hyperpolarization to remove any inactivation of voltage-dependent channels.

5-HT and dopamine affected the shape of the action potential, tending to produce an increased overshoot, reduced undershoot and increase in duration.

5-HT, dopamine and octopamine at concentrations of 10⁻⁶–10⁻⁴moll⁻¹ produced no effect on the cell body when applied to the proboscis tip. ACh at similar concentrations produced one or two ductule spikes, resembling the effect of application to the cell body but without the transient depolarization (Fig. 13); this effect was not, however, observed in all gland cells tested.

Some salivary cells secrete into the lumen of the proboscis and others on to its outer surface (Sawyer et al. 1982). Little or no secretory product was seen at the surface in unstimulated preparations bathed in physiological saline, but secretion in response to stimulation could be clearly observed. Secretions stuck for a while to the proboscis; they were not readily soluble and dispersed slowly. At high magnification they could be seen to emerge from discrete points which appeared to be single ductule endings. The ductules secrete from their tip directly on to the proboscis surface; they do not discharge first into a common duct (Walz et al. 1988; this contrasts with jawed leeches such as Hirudo;Marshall & Lent, 1988).

Secretion could be elicited by repetitive stimulation of the stomatogastric nerve, and by exposure of the preparation to 10⁻⁴moll⁻¹ 5-HT, but no effect of dopamine, octopamine or ACh was observed. The most powerful stimulant of secretion was calcium-free saline which was applied to the glands alone, with the proboscis in a separate compartment containing the normal concentration of calcium. Removal of calcium elicits spontaneous action potentials in gland-cell bodies, produced by sodium flowing through calcium channels (Wuttke & Berry, 1988). It seems likely that these action potentials are conducted along the ductules and initiate the secretion observed at the proboscis tip.

The protein composition of salivary gland extracts has been documented (Budzynski et al. 1981a,b) but little is known of the composition of secreted material. Our interest focused on the secretions at the proboscis tip because they can clearly be seen forming from individual ductules, and thus offer the possibility of studying secretory mechanisms using single cells. Secretion was initiated by repetitive stimulation of the stomatogastric nerve, and material from the proboscis kip and lumen was sucked separately into a fine syringe needle for subsequent. protein analysis. Also, cytoplasm was removed from a selection of gland-cell bodies for analysis.

Gel electrophoresis showed the presence of two main protein bands in material from the proboscis tip, whereas secretions in the lumen contained at least 10 bands (Fig. 14). The tip proteins were not found in the lumen, and their relative molecular masses (19 500 and 28 000) did not correspond to that of the anticoagulant, hementin (120 000), which is present in gland extracts (Malinconico et al. 1984). Interestingly, the cytoplasm of certain gland-cell bodies showed these same two bands (Fig. 14) whereas other cells had many bands, some of which corresponded to those found in lumen secretions but none to those in tip secretions (Fig. 14). More than 50 cell bodies in three leeches were examined to determine whether it was possible to identify visually the cells containing the two proteins corresponding to those found in tip secretions. Middle-sized cells (approx. 600 μm in 15cm long leeches), which stained poorly with Methylene Blue, were consistently found to contain these two proteins, whereas other neighbouring cells did not (Methylene Blue was used because it was found to stain gland cells differentially, and this might have reflected their different secretory content). These cells were scattered throughout the gland and did not appear distributed in any particular pattern. The findings with Methylene Blue are important because there is currently no other means of distinguishing between cell types in a living gland.

The salivary gland cells of Haementeria resemble nerve cells in that depolarization of the cell body produces action potentials which are conducted along a process (which may be several centimetres long) to its ending where release of secretory product occurs. It is tempting, therefore, to regard the action potential as the trigger for secretion, perhaps providing an influx of calcium for exocytosis. Action potentials appear to elicit secretion in other exocrine glands such as the salivary gland and pedal gland of molluscs (Kater, 1977; Kater et al. 1978; Barber, 1983; Goldring et al. 1983) and in certain vertebrate endocrine glands (Petersen, 1980; Ozawa & Sand, 1986). Our main evidence in favour of this view is that calcium-free saline, which produces spontaneous repetitive firing in gland-cell bodies (Wuttke & Berry, 1988), also elicits copious secretion at the proboscis tip. In these experiments the glands alone were exposed to calcium-free solution by placing them in a separate compartment from the bulk of the proboscis which was bathed in normal saline.

A further link between the action potential and secretion is indicated by the fact that stimulation of the stomatogastric nerve elicits action potentials in gland cells and also causes secretion. However, secretion may also be elicited independently of action potentials, e.g. by 5-HT (see below).

Neuronally evoked gland-cell impulses recorded in the soma appear to be initiated outside the gland near the ductule tip and are then conducted to the cell body; they resemble the ductule spikes produced by direct stimulation of the proboscis tip. For cells with a ductule terminating at the end of the proboscis, removal of the proboscis tip abolishes the response to nerve stimulation but not to direct stimulation of the proboscis. Our reason for examining the effect of cutting the proboscis tip was the observation by Walz et al. (1988) of synapses onto ductules in electron microscope sections of the proboscis tip. These authors found no evidence of synapses in the salivary gland itself (although occasional axon profiles were seen) and we have found no electrophysiological evidence of innervation of the gland. If innervation does occur only near the ductule tip then it seems unnecessary to produce propagated ductule spikes. Synapses may, however, be up to 100 μm from the ductule tip (Walz et al. 1988), perhaps too far to influence release directly. It is not known whether the conduction of action potentials to the cell body has any functional significance.

Application of the natural transmitter to the ductule tip would be expected to mimic the effect of nerve stimulation in producing ductule spikes and secretion. Synapses are found within 100 μm of the ductule ending at the surface of the proboscis (Walz et al. 1988) so that putative transmitters probably do not have to penetrate deeply into the proboscis to elicit a response. Nevertheless, saliva contains enzymes which presumably enable the penetration of mammalian skin, yet the proboscis is not obviously affected by these secretions and may well be less permeable than other leech tissues. This is indicated by the time taken for 18 mmol 1⁻¹ Mg²⁺ to block proboscis contractions produced by stimulating the stomatogastric nerve (20min). Nevertheless, Glover (1987) found that, although he could detect serotonin in the proboscis and salivary glands of Haementeria, serotonin-immunoreactive processes were found only in the proboscis, and he suggested that the salivary glands may have a barrier to the antisera. In the present experiments two of the putative transmitters tested (serotonin and acetylcholine) produced rapid responses when applied to the proboscis tip or to the gland-cell bodies. These results suggest that there may be no special barrier in the glands or proboscis, but simply that certain parts of the tissue are less accessible than others, especially those deep in the proboscis.

In certain jawed leeches such as Hirudo, 5-HT plays an important role in feeding behaviour (Lent & Dickinson, 1984) and elicits salivary cell impulses and secretion (Marshall & Lent, 1988). 5-HT also produced secretion in Haementeria, but without ductule spikes. Its effect on the cell body was a subthreshold depolarization and increase in excitability. Stimulation of the stomatogastric nerve did elicit ductule spikes, though a rather high frequency of stimulation was necessary to produce such spikes, and they did not occur in every cell. Dopamine was similar to 5-HT in increasing the excitability of gland cells (though less extensively) but it did not initiate secretion. Octopamine was without effect on secretion and had little effect on electrophysiology. ACh produced action potentials when applied to the cell body or ductule tip but these showed rapid adaptation and there was little or no effect on secretion. The results thus suggest only 5-HT as a transmitter candidate. The fact that it does not elicit ductule spikes indicates that the action potential is not the sole (or necessarily the natural) stimulus for secretion. Similarly in the pars intermedia of the vertebrate pituitary gland, secretion is produced by action potentials but can also occur in their absence, apparently as a result of spontaneous subthreshold fluctuations in membrane potential (Tomiko et al. 1984). In the carnivorous mollusc Philine aperta action potentials and excitatory synaptic potentials may independently cause release of secretion (Barber, 1983). A hormonal role for 5-HT is possible in Haementeria, though this was discounted in Hirudo by Lent & Dickinson (1984). 5-HT may possibly act as a cotransmitter because the synaptic terminals onto ductules in the proboscis tip contain small clear vesicles and large dense-core vesicles (Walz et al. 1988). It will be necessary to determine whether 5-HT actually occurs in the proboscis tip and whether 5-HT antagonists block the secretory response to nerve stimulation.

Additional evidence for a role of 5-HT in Haementeria comes from the finding by Marshall & Lent (1985) of nerve fibres with a positive reaction to anti-serotonin antibody which run along the length of the proboscis sheath and enter the salivary glands. Although it was not possible to resolve contacts with individual salivary cells, neuronal 5-HT is clearly present in their vicinity, and Marshall & Lent regard it as probably one of the natural stimuli for secretion. Glover (1987) also detected 5-HT in whole salivary glands. Our results show that there may be 5-HT receptors on the gland cell bodies but these are unlikely to have any direct role in secretion because 5-HT does not elicit action potentials. Even if the neuronal 5-HT acts to increase gland cell excitability, it is difficult to see how this could influence secretion at the distant ductule tip. There is, however, the intriguing possibility that the cell bodies are activated by a different transmitter and that 5-HT acts to ensure that repetitive firing can occur; in the absence of 5-HT the gland cells rarely produce more than three or four action potentials in response to maintained depolarization. Nerve fibres in the gland may be separate from the stomatogastric nerve and not activated by its stimulation (salivary gland cells in the mollusc, Philine, are innervated by independent central and peripheral neurones; Barber, 1983). The long delay between stimulation of the stomatogastric nerve and production of ductule spikes may also be indicative of indirect innervation of the ductule [though long-latency effects of transmitters are known; e.g. Gerschenfeld & Paupardin-Tritsch (1988) have described a 5-HT-induced K⁺ current in snail neurones with a latency of 30–60 s and a time course of 3–5 min following a 500 ms ionophoretic application of 5-HT], It is noteworthy that stimulation of gland-cell bodies by calcium removal elicited far more secretion than even prolonged high-frequency stimulation of the stomatogastric nerve. Any neuronal stimulation of the gland-cell bodies could thus supplement the neuronal stimulation of the ductule tip to produce maximal release of secretory products.

Secretions are released mainly into the lumen of the oesophagus and proboscis, and also onto the outer surface of the proboscis tip. The tip secretions were found to contain two major proteins which are probably involved with aiding penetration of the mammalian host’s skin, but the mechanism by which the soft proboscis is able to achieve this is unknown. Gel electrophoresis of extruded cytoplasm showed the same two protein bands in certain gland cell bodies. The cells could be identified in living glands on the basis of size and staining properties but extensive studies have revealed no distinguishing electrophysiological features between cells (Marshall & Lent, 1984; Wuttke & Berry, 1988). Histochemically, the gland contains three serous and two mucous secretions (Sawyer et al. 1982). One of the serous secretions (SI), which is probably a proteolytic polypeptide, occurs exclusively in cells whose ductules terminate at the proboscis tip (Sawyer et al. 1982). These cells may thus correspond with those found in the present study to release at the tip.

The anticoagulant hementin appears from histochemical studies to be released into the lumen of both the oesophagus and proboscis, where it prevents coagulation of ingested blood (Sawyer et al. 1982). In support of this, the present experiments showed a protein band of corresponding molecular weight in lumen secretions but not in those from the proboscis tip.

Secretory vesicles pack the cell bodies (Walz et al. 1988) so it is likely that secretory proteins contribute to the observed electrophoresis bands from extruded cytoplasm. In lumen secretions, bands were found that did not correspond to those in any cell, and these may represent secretions from the posterior glands which were not examined here. Some of the protein bands found in the cytoplasm did not correspond to any secretory product and probably represent nonsecretory proteins or secretory protein released in small amounts (they are perhaps associated with the nucleus which is extremely large in gland cells; Sawyer et al. 1982). The ability to collect cytoplasm for analysis should allow study of the metabolism of secretory products in individual cells.

The mechanism of release of secretions is unknown. Marshall & Lent (1985) conclude that since no acinar or duct structure is present it is likely that the various secretions are formed entirely by exocytosis without an accompanying volume of solvent, resulting in small volumes of concentrated product. Electron micrographs of the proboscis tip (Walz et al. 1988) add support to this view: the ductules are densely packed with secretory vesicles and the only other organelles are restricted to a narrow layer near the plasmalemma. Surprisingly, however, copious secretions can be seen to pour out of certain small, discrete areas of the proboscis tip, on occasions reminiscent of toothpaste being squeezed from a tube. It is difficult to see how this could occur by exocytosis, given the restricted access of vesicles to the ductule tip, which is only about 2 μm in diameter. One gains the impression that the tip of the ductule must have broken away (e.g. apocrine secretion). The number of sites at which this occurs is far smaller than the total number of ductule endings known to occur in the proboscis tip, so that more than one mechanism of secretion may be operating (or only a proportion of the cells are secreting). It would be interesting to look for evidence of exocytosis in electron micrographs of the proboscis tip from an actively secreting preparation. Evidence of apocrine-type secretion could be sought by testing for the presence of vesicle membranes in the secretory product.

The special advantages of this preparation for the study of excitation-secretion coupling are well documented (Marshall & Lent, 1984; Jones et al. 1985; Wuttke & Berry, 1988). The present experiments demonstrate four further important features. First, the secretions can be collected for assay, and different protein constituents resolved by gel electrophoresis. Second, the cytoplasm of individual cell bodies can be collected for analysis of secretory proteins. Third, the cells which secrete at the proboscis tip can be identified. Finally and most importantly, the secretions at the proboscis tip come from discrete points on the proboscis surface and appear to emanate from single cells. Since the salivary cells are not electrically coupled (Sawyer et al. 1982) this offers a unique opportunity to monitor secretion in a single, functionally isolated cell in situ which can be penetrated with microelectrodes.

We thank Gabriele Glaser for preparation of the gel. This work was supported by a grant from the Science and Engineering Research Council (no. GR/D/54774).