Initiation and Modulation of Action Potentials in Salivary Gland Cells of Haementeria Ghilianii by Putative Transmitters and Cyclic Nucleotides

Leech salivary cells are electrically excitable, producing calcium-dependent action potentials, accompanied by secretion, in response to nerve stimulation (Lent and Dickinson, 1984; Marshall and Lent, 1988; Wuttke et al. 1989). Very little is known, however, about the nature of gland-cell innervation. In the jawed leeches Hirudo medicinalis and Macrobdella decora, Marshall and Lent (1988) found that acetylcholine, y-aminobutyric acid, dopamine and noradrenaline were without effect on secretion (though they did not examine for electrophysiological actions), whereas 5-hydroxytryptamine (5-HT) initiated secretion and excited the gland cells into impulse activity. Similar effects were produced by stimulation of the serotonergic Retzius cells, indicating that these neurones contribute to the innervation of salivary cells. The location of the synaptic input in jawed leeches is unknown. Each salivary cell extends a single process (ductule) anteriorly to the base of one of the three jaws; such ductules may reach 2 cm in length. Innervation possibly occurs close to the cell body, in view of the appearance there of subthreshold depolarizing responses to Retzius cell stimulation (Marshall and Lent, 1988).

In contrast, salivary gland cells of the proboscis-bearing leeches, Haementeria ghilianii and Oosthuizobdella garoui, are innervated near the ductule ending and action potentials are conducted back to the cell body; no signs of synaptic potentials are seen in the soma (Wuttke et al. 1989; Cooper and Berry, 1990). In these leeches the salivary cell bodies are collected into a distinct salivary gland whereas in H. medicinalis they are distributed throughout the musculature of the head. Although electrophysiological experiments failed to demonstrate any innervation of the salivary glands of H. ghilianii, the gland-cell bodies show responses to a number of putative transmitters (Wuttke et al. 1989), and axon profiles have been observed in the glands (Walz et al. 1988). Furthermore, Glover (1987) detected 5-HT in whole salivary glands of H. ghilianii, and Marshall and Lent (1985) found nerve fibres with a positive reaction to anti-serotonin antibody running along the length of the proboscis sheath and entering the salivary glands. The present paper analyses the pharmacology of the salivary gland cells of H. ghilianii and discusses its possible physiological significance.

Salivary glands were dissected from specimens of Haementeria ghilianii (de Filippi) obtained from our breeding colony and maintained in aquaria at 26°C. The glands were secured to a layer of Sylgard at the base of a Perspex experimental bath (volume 0.25 ml) by placing pins through the overlying endothelium. Physiological saline or other solutions (see below) flowed through the bath at a rate of about 10 bath vols min^-1. Solutions could be changed rapidly without affecting recording conditions (Holder and Sattelle, 1972).

Observations on secretion were made with isolated preparations in which the proboscis/salivary gland complex was left intact. Unstimulated preparations produced little or no secretion, and any stimulation of secretory activity could be clearly seen at the tip of the proboscis.

Individual gland cells were impaled under visual control with two bevelled, KC1-filled microelectrodes (10-20MΩ); one electrode was used for recording mem-brane potential and the other for passing either constant-current pulses to measure input resistance or direct current to alter the membrane potential. The overlying sheath was not removed or softened by enzymes. The largest cells (600-1000 μm in diameter) in the anterior salivary glands were generally chosen for study but it was not easy to maintain stable recordings with two electrodes, probably because of the greatly invaginated nature of the cell membrane (Walz et al. 1988); electrodes tended to come out gradually from the cell. This problem was largely overcome by mounting each electrode on a high-speed stepper (Digitimer SCAT-02), which facilitated repositioning or repenetration without loss of membrane potential. The electrodes were each connected to a Digitimer NL102 amplifier, and signals were monitored on a storage oscilloscope (Tektronix 5111) and pen recorder (Brush 2200S) and stored on tape (Thorn EMI 3000 FM tape recorder).

Membrane currents were measured with a two-electrode voltage-clamp amplifier (Axoclamp 2A, Axon Instruments). Cells were generally held at, or close to, their resting membrane potential, and the effects of substances on the currents evoked by depolarizing or hyperpolarizing voltage steps were examined. Linear leakage currents could be subtracted electronically before displaying and storing the data.

Physiological saline had the following composition (mmol 1 ^-1): NaCl, 125; KC1, 4; CaCl2, 1-8; glucose, 11; Tris maleate, 10 (pH7.4) at room temperature (18-22°C). Ionic mechanisms were investigated by ion substitution or addition of channel blocking drugs. Na⁺ was replaced by N-methyl-D-glucamine, and variations in [K⁺] were compensated by changes in [Na⁺], Tetraethylammonium and 4-aminopyridine (with an equimolar reduction in [Na⁺]) were used to block K⁺ currents, and Co²⁺ was used to block Ca²⁺ influx.

Putative transmitters, second messengers, agonists and antagonists were applied by perfusion at known concentrations. The following substances were used: acetylcholine chloride, 5-hydroxytryptamine creatinine sulphate, y-aminobutyric acid (GABA), glutamic acid (monosodium salt), dopamine hydrochloride, DL-octopamine hydrochloride, cyclic 3’,5’-adenosine monophosphate and its dibu-tyryl and 8-bromo derivatives (all sodium salts), cyclic 3’,5’-and cyclic 2’,3’-guanosine monophosphate (sodium salts), theophylline, methysergide maleate, carbamylcholine chloride (carbachol), pilocarpine hydrochloride, tetraethyl-ammonium chloride (TEA⁺), tetramethylammonium hydroxide, hexamethonium chloride, nicotine, strychnine, d-tubocurarine chloride and atropine. Apart from methysergide (Sandoz), all chemicals and drugs were obtained from Sigma Chemical Co.

Of the various substances tested, acetylcholine (ACh) produced the largest effect on membrane potential. Perfusion of the salivary gland with 10⁻⁴moll^-1ACh produced a depolarization of 25.9±6.2mV (N = 30, mean±s.D.) accompanied by a reduction in input resistance and the generation of 1-3 action potentials (Fig. 1). This response remained unchanged with repeated applications of ACh, provided the membrane potential was allowed to return to control levels. There was desensitization with prolonged ACh perfusion, the response declining to 24±11% of its maximal value within 2 min (N =9). A concentration of 10⁻⁶moll^-1 ACh produced a subthreshold depolarization of 6.5±2.1mV (N =6). The minimum concentration of ACh required to produce a detectable depolarization was ICT⁷ mol 1 ^-1. Acetylcholine had no effect on the amplitude and shape of action potentials.

The magnitude of the ACh response as a function of membrane potential is shown in Fig. 2 and plotted in Fig. 3. The reversal potential of the response was difficult to determine directly in most cells. Although there is no electrical coupling, which presents problems in some other salivary glands (e.g. Bahls, 1987), the large size of the cells and the extent of delayed rectification made it necessary to pass large amounts of current, which often could not be carried by the microelectrodes. In those cases where measurements were possible, the reversal potential was about —25 mV (Fig. 3).

The response to ACh was dependent on external Na⁺. Replacement of Na⁺ by N-methyl-D-glucamine produced a complete block of response in five cells, and a reduction to 18 % in a further cell.

Several ACh agonists were tested for their ability to mimic the response to ACh in salivary cells (Fig. 4). Each substance was tested at a concentration of 10⁻³moir¹ on 10 cells. Carbachol, tetramethylammonium and nicotine produced similar-sized depolarizations to those caused by 10⁻⁴mol 1 ^-1 ACh, though the time course was prolonged for nicotine and carbachol. On these cells, ACh produced a depolarization of 31.2±6.3mV, and the responses to carbachol, tetramethylammonium and nicotine were, respectively, 33.3±5.3, 33.1±4.5 and 29±5.4mV. Choline and pilocarpine (Fig. 4) produced little or no response (5.8±4.3 and 1.5±1.1 mV, respectively).

The actions of various cholinergic antagonists were examined by comparing a control response to 5x10⁻⁵moll^-1 ACh with a test response after perfusing the gland for about 5 min with an antagonist at a concentration of 5 x ICC⁴ mol 1 ^-1. The most powerful antagonists tested were d-tubocurarine (N =5; Fig. 5C) and strychnine (N=2; not illustrated), which completely blocked the response to ACh. Recovery was slow, requiring at least 25 min for d-tubocurarine and 40 min for strychnine. Tetraethylammonium (Fig. 5A) reduced the response to 24±5 % of control (N=4), hexamethonium (Fig. 5B) to 54±22% (N =4) and atropine (not illustrated) to 35 ±18% (A=6).

In contrast to the large ACh-evoked responses, 5-hydroxytryptamine (5-HT) produced only a small subthreshold depolarization. Its ability to initiate secretion, however, prompted a detailed study, including voltage-clamp experiments that revealed effects on voltage-gated inward and outward currents. This paper analyses the variety of effects of 5-HT and the possible involvement of cyclic nucleotides, rather than the pharmacology of 5-HT receptors.

Perfusion of the salivary glands with 5-HT produced a depolarizing response that failed to reach threshold even with concentrations as high as 10⁻⁴ moll⁻¹ (responses ranged from 4 to 10 mV, N=10). There was an associated increase in membrane resistance (18±12%, N=9), which was larger if the membrane potential was reset to its original value by applied current (Fig. 6). Action potentials, elicited by injection of depolarizing current, were increased in amplitude and duration by 5-HT (Fig. 6). Observable effects on these variables occurred at concentrations of about 10⁻⁷moll^-1, and there was no desensitization over periods of at least 15 min (longest period tested). The spike undershoot tended to be reduced by 10⁻⁴moll^-1 5-HT, but was little affected by lower doses.

Action potentials could be elicited by anode break at the termination of a hyperpolarizing current pulse, and these were elicited more easily in the presence of 5-HT (Fig. 6). 5-HT also enabled action potentials to be produced more readily by depolarizing pulses, and it reduced the adaptation that occurs during maintained depolarization. In the absence of 5-HT it was impossible to produce more than three or four action potentials with a depolarizing pulse, and usually only one was produced, whereas in the presence of 5-HT a train of 35 impulses or more could be elicited.

The effects of 5-HT were examined further by voltage-clamp studies. Since 5-HT does not initiate action potentials, but modifies their properties, emphasis was placed on the analysis of voltage-gated conductances. It proved difficult, however, to pass sufficient current through the microelectrodes to clamp cells adequately at potentials beyond threshold. Penetration with low-resistance electrodes was impossible without loss of membrane potential and excitability (this extreme sensitivity may be a general feature of leech salivary cells; for example, Marshall and Lent (1988) could not usually hold H. medicinalis cells (up to 200 pm in diameter) for more than 3 min with a single microelectrode). We therefore generally tested subthreshold depolarizations for effects of 5-HT that might underlie the increased size and duration of action potentials and the increased excitability. Fig. 7 shows that depolarizing pulses induced an inward current consisting of an initial fast transient followed by a longer-lasting component. In the presence of 10⁻⁵moll^-1 5-HT the current was greatly increased in amplitude recovery to initial levels followed washing with normal saline. The threshold concentration for this effect was about 10⁻⁷ mol 1 ⁻¹. These and the following experiments were all performed on at least four cells and typical results are shown.

Since the cells produce purely Ca²⁺-dependent action potentials (Wuttke and Berry, 1988), it seemed likely that the inward current was carried by Ca²⁺, and attempts were made to distinguish between a potentiating effect of 5-HT on this Ifurrent and a blocking action on K⁺-mediated outward currents. No outward current is apparent in Fig. 7 but, with larger depolarizing pulses or in cells with lower membrane potentials, the inward current was followed by a long-lasting outward current (see Fig. 8). Addition of 10 mmol 1 ^-1 CoCl₂ blocked the voltagegated inward current, indicating that it is indeed mediated by Ca²⁺. CoCl₂ also reduced the outward current, suggesting the presence of a Ca²⁺-activated K⁺ conductance. The effect of 5-HT was now to reduce or abolish the outward current (Fig. 8). Similar effects were produced by 5 mmol 1 ^-1 CoCl₂, except that the transient component of the inward current was not fully blocked by Co²⁺ and it was still potentiated by 5-HT. This may indicate that there are two types of Ca²⁺ channel, differing in rate of inactivation and sensitivity to Co²⁺, but this was not investigated further.

Although the depression of outward current by 5-HT would enhance the net inward current, this did not exclude the possibility of an additional, potentiating effect on the voltage-gated Ca²⁺ current. To test this possibility we attempted to block K⁺ currents by addition of 50 mmol 1 ^-1 TEA⁺ and 10 mmol 1 ^-14-aminopyridine. Outward currents were greatly reduced, and the inward current now showed little response to 5-HT (Fig. 9). These results suggest that 5-HT enhances inward current by reducing outward K⁺ currents.

5-HT also influenced an inward current induced by hyperpolarization. Fig. 10 shows that after a hyperpolarizing jump, in addition to the instantaneous current change (linear leakage current was not subtracted), a large voltage-dependent inward current developed slowly and reached a peak within 10-20 s (depending on voltage). Its amplitude was approximately doubled by 10⁻⁵moll^-1 5-HT (Fig. 10; the threshold concentration for an effect was 10⁻⁷ mol 1 ^-1). This inward rectifier is not specific to K⁺, but carries approximately equal K⁺ and Na⁺ currents (Wuttke and Berry, 1991).

A detailed analysis was not made of the pharmacological profile of 5-HT responses, but the 5-HT antagonist methysergide was used to provide information on the specificity of the actions of 5-HT. Methysergide (10⁻⁴moll^-1) blocked the depolarization, the increase in membrane resistance and the increased excitability produced by 10⁻⁵moll^-1 5-HT. The potentiating effect of 10⁻⁵moll^-1 5-HT on the inward current, elicited by depolarizing or hyperpolarizing steps, was reduced by 10⁻⁴ mol 1 ^-1 methysergide. All the effects of the blocking drug were reversed by washing with normal saline.

Dopamine and octopamine produced subthreshold depolarizations, and in some cells enhanced excitability, confirming the results of an earlier study (Wuttke et al. 1989). Two additional substances were tested in the present study: glutamate (IO^-4moll^-1) produced a small depolarization (4.4±l.lmV) and reduction in membrane resistance (to 62±10%, 7V=7), and GABA (10⁻⁴moll^-1) had no effect at all (not shown). Since the electrophysiological effects of these substances were small, and none of them stimulated salivary secretion, they were not analysed further.

Electrophysiological effects of cyclic 3’,5’-adenosine monophosphate (cyclic AMP) and cyclic 3’,5’-guanosine monophosphate (cyclic GMP) were examined to determine whether these substances might mediate the actions of any neuroglandular transmitter. Particular attention was paid to comparing the effects of cyclic nucleotides with those of 5-HT and ACh, which produced the major responses.

Perfusion of the salivary glands with 1 mmol 1 ^-1 cyclic GMP produced a small depolarization (3.6±3.2mV, N =9) and reduction in membrane resistance (to 54±19%, N =5), and a progressive decline in spike amplitude until the cells became very difficult to excite (Fig. 11). These effects of cyclic GMP were specific to the 3’,5’ form; 2’,3’-cyclic GMP had little or no effect. The results indicate that the actions of 5-HT and ACh are not mediated via this cyclic nucleotide.

There was little effect of cyclic GMP on spike duration; an initial slight shortening was followed by a small lengthening of the action potential as the amplitude fell. There was, however, a dramatic shortening of the duration of action potentials that had been artificially prolonged either by addition of 50 mmol 1 ^-1 TEA⁺ (which produced action potentials of about 6 s duration) or by removal of external Ca²⁺ (which results in Na⁺-dependent action potentials lasting tens of seconds; Wuttke and Berry, 1988). For example, the duration of Na⁺-dependent action potentials was reduced by more than 90% within 1 min (Fig. 12). The speed of the response, compared with the gradual reduction in spike amplitude in normal saline, may indicate an extracellular effect. Subsequently, there was a decline in spike height, but this was associated with a progressive depolarization (Fig. 12). These experiments were not pursued further, but they clearly show complex modulatory effects of cyclic GMP on gland-cell action potentials.

Addition of 1 mmol 1 ^-1 cyclic AMP or the membrane-permeable dibutyryl cyclic AMP generally produced little effect. The non-hydrolysable 8-bromo derivative of cyclic AMP (0.5-1 mmol 1 ^-1), however, mimicked all the effects of 5-HT on membrane potential, membrane resistance, action potential duration and adaptation (Fig. 13) and on membrane current evoked both by depolarizing voltage steps (Fig. 14) and by hyperpolarizing steps (not shown). The effects were slowly reversible, requiring at least 30 min for full recovery. The possibility that cyclic,AMP could mediate the effects of 5-HT was further tested by application of the phosphodiesterase inhibitor theophylline (2 mmol 1 ⁻¹). Unfortunately, no clearcut results were obtained; any effect of theophylline was certainly not large.

Salivary secretion in H. ghilianii is elicited by stimulation of the stomatogastric nerve, but attempts to discover the neuroglandular transmitter are complicated by the fact that innervation appears to occur outside the salivary gland near the ductule terminal, which is inaccessible to microelectrodes (Wuttke et al. 1989). Axon profiles and 5-HT-containing nerve fibres have been described in the gland (Marshall and Lent, 1985; Walz et al. 1988), but there is no electrophysiological or anatomical evidence of synaptic connections there. Effects on the cell body, studied here, are therefore almost certainly extrasynaptic. In the nervous system, extrasynaptic receptors generally have no known function but may show similarities to subsynaptic receptors, and thus offer clues to the nature of any actions at the synapse (e.g. in the leech nervous system; Sargent et al. 1977). We have devoted most attention to 5-HT because it was the only putative transmitter tested that initiated secretion and it seems a likely candidate for the neuroglandular transmitter in H. ghilianii (Wuttke et al. 1989). Acetylcholine was also studied in some detail because it was the only putative transmitter to excite the gland cells into impulse activity.

Application of acetylcholine (ACh) to the salivary gland elicits a depolarization in gland-cell bodies that is abolished by removal of external Na⁺. The reversal potential for the ACh response was found to be approximately —25 mV, which is considerably more negative than the Na⁺ equilibrium potential (approx. +45 mV; Wuttke and Berry, 1990), and suggests that K⁺ permeability is also increased, though this was not tested.

The ACh receptors on the salivary cells of H. ghilianii show similarities to those described in the salivary cells of molluscs (Barber, 1982,1985; Bahls, 1987). Apart from annelids, only gastropod molluscs have been shown to have electrically excitable salivary gland cells, though their cellular organisation is quite different; molluscan salivary glands have an acinar structure and cell-cell coupling, typical features of all salivary glands that have been examined except for those of leeches (and probably oligochaetes; Marshall and Lent, 1988). In the mollusc Helisoma trivolvis ACh produces a Na⁺-(and probably K⁺-) dependent depolarization that is mimicked by the cholinergic agonists carbachol and tetramethylammonium, with a weak response to pilocarpine (Bahls, 1987). Choline is, however, a potent agonist, and nicotine produces only a weak response, in contrast to their effects in H. ghilianii, though choline activates salivary cells in the proboscis leech Oosthuizobdella garoui (Cooper and Berry, 1990). The effects of cholinergic antagonists are also similar in H. ghilianii and H. trivolvis:. atropine, TEA⁺, hexaméthonium, strychnine and d-tubocurarine all reduce ACh depolarizations, with the last two being the most potent and the effect of strychnine being difficult to reverse. The lack of selectivity in their sensitivity to cholinergic antagonists makes classification of the receptors as muscarinic or nicotinic difficult.

One of the main differences between the ACh response in H. ghilianii and H. trivolvis is that the molluscan salivary cells show only minimal desensitization to prolonged application of ACh and there is an after-hyperpolarization generated by the Na⁺ pump (Bahls, 1987). No after-hyperpolarization is seen in H. ghilianii, in spite of the presence of a strongly electrogenic Na⁺ pump (Wuttke and Berry, 1988), suggesting a relatively small influx of Na⁺ (perhaps as a result of desensitization).

Apart from H. ghilianii and O. garoui, no detailed pharmacological studies appear to have been made of leech salivary cells. Although the salivary glands in these two species are similar in structure and physiology, some differences are apparent. For example, ACh (10⁻⁴moll^-1) does not affect spike amplitude in H. ghilianii whereas concentrations as low as 10⁻⁷ mol 1 ^-1 produce a large increase in spike height in O. garoui (Cooper and Berry, 1990). More marked differences are seen for 5-HT (see below).

It is possible that ACh receptors similar to those found in the cell body may occur near the ductule ending, in the region of the synaptic input. For example, application of ACh to the proboscis tip produces the same response as application to the soma, i.e. 1-3 action potentials (conducted along the ductule to the cell body; Wuttke et al. 1989). The failure of bath-applied ACh to elicit secretion could be due to rapid desensitization, as seen in the soma, and does not necessarily preclude ACh from consideration as a neuroglandular transmitter. Also, the rapid adaptation shown by the cells restricts their activation by bath-applied ACh. Brief applications from a micropipette might lessen the effects of desensitization and adaptation, and more closely resemble the effect of any cholinergic input. It would not be easy, however, to use this method at the ductule terminal because these cannot yet be precisely located for any one cell body.

Although there is no direct evidence, 5-HT is a good candidate for effecting neuroglandular transmission in H. ghilianii (Wuttke et al. 1989). Better evidence is available for the jawed leech, H. medicinalis, where application of 5-HT or stimulation of the serotonergic Retzius cells in the first segmental ganglion evokes depolarization and action potentials in the salivary cells (Marshall and Lent, 1988).

In H. ghilianii, cobalt back-filling failed to demonstrate any processes of the Retzius cells in the stomatogastric nerve, which is the only source of innervation of the gland cells from the central nervous system (Wuttke et al. 1989). Stimulation of the Retzius cells did not evoke impulse activity in the gland cells (W. A. Wuttke and M. S. Berry, unpublished observation), though this was perhaps not unexpected since 5-HT itself does not activate the cells. This contrasts with the salivary cells of O. garoui, where 5-HT does not depolarize the cell body but does initiate action potentials, apparently at, or near, the ductule terminal where innervation is believed to occur (Cooper and Berry, 1990).

H. ghilianii is unusual in that 5-HT elicits salivary secretion but does not activate the salivary cells. This substance does, however, produce a multitude of effects on the cells: a depolarization and increase in input resistance, an increase in the amplitude and duration of action potentials, an increase in excitability, a reduction of adaptation during maintained depolarization, an increase in net inward current evoked by subthreshold depolarization, and an increase in an inwardly rectifying current. The relatively low concentration of 5-HT required to produce these effects (approx. 10∼⁷moll^-1) and their reduction or blockade by the 5-HT antagonist methysergide indicate that they are mediated via specific 5-HT receptors, though no attempt was made to perform a pharmacological classification.

The depolarizing response to 5-HT is probably caused by a reduction in K⁺ conductance, because it is associated with an increase in input resistance. 5-HT reduces the outward K⁺ current induced by subthreshold depolarizations, thus potentiating the net inward current. These effects on K⁺ conductance probably underlie the changes in excitability of the gland cells and the changed configuration of the action potentials and their rate of adaptation. The K⁺ current was greatly reduced by Co²⁺, suggesting the presence of a Ca²⁺-activated K⁺ conductance. A direct effect of Co²⁺ on the K⁺ current is possible (Pollock et al. 1985), but substitution of Ca²⁺ with Ba²⁺ or Sr²⁺, or removal of external Ca²⁺ without substitution by divalent cations, produced an increase in spike duration (Wuttke and Berry, 1988), again indicating a reduction in K⁺ conductance.

In other systems, 5-HT may mediate its synaptic effects, including those on voltage-dependent channels, via cyclic nucleotides (Gerschenfeld et al. 1986). In H. ghilianii the slow onset of the responses to 5-HT compared with those to ACh (cf. Figs 2 and 6) suggested the presence of a second messenger. Cyclic GMP was ruled out because of its opposing effects on input resistance, excitability and amplitude of action potentials. Cyclic AMP and its membrane-permeable dibutyryl derivative failed to produce any marked effect, but 8-bromo-cyclic AMP mimicked all the effects of 5-HT. We cannot be certain that the effects of 8-bromocyclic AMP represent the physiological actions of cyclic AMP, though this phosphodiesterase-resistant nucleotide mimics the effects of cyclic AMP in many other cells (e.g. Levitan and Levitan, 1988). In the bag cells of Aplysia calif arnica, where there is good evidence that 5-HT exerts its actions through cyclic AMP, both cyclic AMP and dibutyryl cyclic AMP failed to elicit a response, whereas 8-substituted derivatives produced responses similar to those of 5-HT (Kaczmarek et al. 1978). In H. ghilianii, experiments with theophylline did not provide supportive evidence, and more data are obviously required.

In H. medicinalis salivary cells, Marshall and Lent (1988) consider that the Ca²⁺-dependent action potentials trigger secretion, with Ca²⁺ acting as the coupling agent. In H. ghilianii also, Ca²⁺-dependent action potentials elicit secretion, but they are evidently not the only stimulus: 5-HT is a non-excitatory secretagogue, presumably acting directly on the release site at the ductule terminal (Wuttke et al. 1989). If 5-HT acts via Ca²⁺ influx, it is difficult to see how sufficient Ca²⁺ could enter during the small 5-HT-evoked depolarization (unless this is bigger at the terminal than that seen in the cell body). The present results suggest that cyclic AMP may be the coupling agent for 5-HT.

Sawyer et al. (1982) noted a similarity between the salivary glands of H. ghilianii and neural ganglia. Such comparisons were shown by Marshall and Lent (1984) to be misleading because the cells are clearly not neuronal and their processes are not part of an integrative neuropile, but are totally independent of one another. Nevertheless, the individual salivary cells do bear many resemblances to neurones; for example, they extend a long process (10 cm or more) that conducts action potentials to the terminal where release of secretory products occurs, apparently as a result of Ca²⁺ influx (Wuttke et al. 1989). The variety of transmitter-activated or modulated conductances demonstrated here and the electrophysiological effects of cyclic nucleotides increase the similarity. The various effects of 5-HT on the salivary cells have all been found in neurones: e.g. the depolarization resulting from a decrease in resting K⁺ conductance (Siegelbaum et al. 1982), the decrease in subthreshold voltage-gated K⁺ current, with a net increase in inward Ca²⁺ current (Paupardin-Tritsch et al. 1981), the increase in amplitude and duration of action potentials (Gerschenfeld et al. 1986), the reduction in spike frequency adaptation (Andrade and Nicoll, 1987), and the potentiation of inwardly rectifying current (Benson and Levitan, 1983). Such effects allow transmitters not only to excite or inhibit but also to influence the pattern of action potential discharge, to control the way in which neurones integrate incoming synaptic input and to modulate transmitter release. Subtle effects are unlikely to be important in the salivary cells of H. ghilianii, which appear to remain inactive until feeding occurs and then simply pour out copious secretions. Nevertheless, several features suggest the possibility of a modulatory role for 5-HT, which does not excite the gland cells. It would be well placed to exert such effects if it is released from the synapses onto the gland-cell terminals observed under the electron microscope (Walz et al. 1988). This innervation is reminiscent of input onto presynaptic terminals, for example where 5-HT has facilitatory effects on transmitter release via cyclic AMP (Kandel and Schwartz, 1982). Similarly, in the gland cells, 5-HT might increase release of secretory products by its effects on the amplitude and duration of the action potentials that invade the terminal. Certainly the action potential is a stimulus for secretion (Wuttke et al. 1989), though it is not known whether it is the natural stimulus.

Although the effects of 5-HT on the cell body are almost certainly direct, the possibility exists that 5-HT elicits secretion indirectly via an effect on the nervous system. Repetitive stimulation of the stomatogastric nerve (leading from the brain to the proboscis) elicits gland-cell action potentials and secretion, although it is possible that the effects are mediated by a nerve plexus in the proboscis (see Wuttke et al. 1989). Depending on the source of the innervation, 5-HT may thus stimulate plexus neurones or, perhaps, initiate or modulate the release of transmitter from the presynaptic terminals at the ductule ending. It would be difficult to test for such indirect effects because most procedures designed to alter synaptic transmission would also be expected to change secretion. If 5-HT does act by stimulation of the nervous system, however, it does not mimic the effect of electrical stimulation, which elicits gland-cell action potentials.

An obvious feature of the 5-HT effects is that they all tend to increase cell excitability. This includes the effect on the inward rectifier in the few cells with a high membrane potential where the rectifier is activated, but could include many more cells if their resting membrane potential in vivo is a little higher than that recorded in isolated glands. One possibility is that 5-HT may raise the normally low level of excitability and allow a second, excitatory transmitter to produce a maintained discharge during feeding. This is purely speculative, but the possibility that two transmitters are used is suggested by the presence of small clear vesicles and large dense-cored vesicles in synaptic terminals onto ductules at the proboscis tip (Walz et al. 1988).

The situation is thus complex. 5-HT does not excite the gland cells and appears, from its electrophysiological effects, to be suited to a modulatory role. Nevertheless, it stimulates secretion, so perhaps a dual role should be considered. In this respect it is noteworthy that, even at high concentrations, 5-HT elicits relatively little secretion compared with amounts that can be elicited in other ways, for example by activating the gland cells with Ca²⁺-free saline (Wuttke et al. 1989) or by procedures that increase intracellular pH (W. A. Wuttke and M. S. Berry, unpublished observation). Certainly 5-HT alone is not a potent secretagogue, leaving scope for additional, modulatory actions.

This work was supported by a grant from the Science and Engineering Research Council (no. GR/F/17087). We are grateful to the SmithKline (1982) Foundation for provision of a pulse generator.