Calcium-Dependent Action Potentials in Giant Salivary Gland Cells of the Leech Haementeria Ghilianii: Calcium Removal Induces Dependence on Sodium

The salivary gland cells of the giant Amazon leech Haementeria ghilianii are extremely large (up to 1·2 mm in diameter) and do not show electrical coupling or dye-coupling (Sawyer et al. 1982; Marshall & Lent, 1984; Jones et al. 1985). They produce overshooting action potentials that are dependent on calcium, with no measurable contribution by sodium (Marshall & Lent, 1984). For example, spikes are abolished by removal of external calcium but not by removal of sodium or addition of tetrodotoxin; strontium or barium can replace calcium but the spikes are abolished by cobalt, manganese and the calcium antagonist methoxyverapa-mil.

The action potential seems a likely stimulus for secretion and it is thus important to understand fully its various properties. We therefore decided to extend the studies of Marshall & Lent (1984) on the mechanism of generation of resting and action potentials. It was found that removal of external calcium does not block action potential generation but induces prolonged action potentials which are mediated by sodium. Under normal conditions the spike seems to be mediated solely by calcium. This sodium permeability in the absence of external calcium, which is a feature of certain other calcium channels (e.g. Hess & Tsien, 1984), is shown to be an important consideration in studies of stimulus-secretion coupling in the salivary gland of Haementeria.

Specimens of Haementeria ghilianii were obtained from a breeding colony [Biopharm (U.K.) Ltd, Swansea] and maintained in aquaria at 26°C. The aquarium water had the following composition (mmolI⁻¹): NaCl, 0·63; CaCl₂, 0·07; MgCl₂, 0·07; K₂SO₄, 0·01; Tris-maleate, 2 (pH 6·0) (Sawyer et al. 1981).

Experiments were performed on the anterior salivary glands which were dissected from the animal together with the posterior glands and a small piece of attached proboscis (see Sawyer et al. 1982 for details of the anatomy). The preparation was pinned to the Sylgard base of a Perspex experimental bath (volume 0·25 ml) and immersed in a continuous flow of physiological saline (see below). Solutions were fed into the bath by gravity and removed by aspiration, with a flow rate of about 10 bath volumes min⁻¹. Solutions could be changed rapidly without affecting recording conditions (Holder & Sattelle, 1972). A Ag-AgCl reference electrode was placed in a solution of 3 mol 1⁻¹ KC1 which was connected to the bath via a KCl-agar bridge (2% agar in 3 mol 1⁻¹ KC1) placed near the outflow.

Conventional intracellular recording and stimulating techniques were used. Individual cells were impaled with a glass microelectrode containing 3 mol 1 ¹ KC.I which was connected to a Digitimer NL102 amplifier. For current injection, a second KCl-filled electrode, connected to a similar amplifier, was inserted into the cell and the current was passed and monitored by the amplifier.

Initially it was sometimes difficult to insert a second electrode in spite of the large size of the cells. This problem was also noted by Marshall & Lent (1984) and may result from the extensive infoldings of the cell membrane. A microelectrode beveller was therefore constructed according to the method described by Kripke & Ogden (1974). A 0·3 μm lapping film (Scotch 3M) was used instead of 0·05μm micropolish, and an acoustic monitoring system was used for controlling the bevelling process (Kaila & Voipio, 1985). Electrode resistances of 60-80 MΩ, measured in saline, fell to 10-20 MΩ after bevelling. This process made microelec-trode penetration easier and improved the stability of recordings.

Signals were monitored on a digital storage oscilloscope (Famell DTS 12) and a pen recorder (Brush 2200S). Most of the records were stored on tape (Thorn EMI3000 FM tape recorder).

The various solutions used are shown in Table 1.

Solutions 1, 2, 3 (except V-methyl-D-glucamine; see solution 4) and 5 were adjusted to pH 7·4 by adding about 10 mmol 1⁻¹ NaOH, bringing the final Na⁺ concentration to about 125 mmol 1⁻¹. EGTA produced a pH change which was corrected by addition of small amounts of Tris. Solution 4 was adjusted to pH 7·4 with lmoll⁻¹HCl. The level of free Ca²⁺ in the Ca²⁺ concentration-buffered salines was 10⁻⁴moll⁻¹ for those solutions containing NTA, and 2xl0⁻⁵moll⁻¹ for the solution containing HEDTA (see Tsien & Rink, 1980; Fukushima & Hagiwara, 1985).

Ouabain (Strophanthin G, 2xl0⁻⁵moll⁻¹) was employed as a specific blocker of the Na⁺/K⁺ exchange pump. Cobalt (II) ions (5 mmol 1⁻¹ CoCl₂) were used to block Ca²⁺ channels. All chemicals were obtained from Sigma Chemical Co. except CoCl₂ (BDH Chemicals, Ltd) and NaCH₃SO₄ (Hopkin & Williams).

Experiments were performed at room temperature (20–23°C). For one set of experiments the temperature was rapidly lowered by about 10°C by switching the incoming flow of saline first through an appropriate length of tubing immersed in ice. A copper-constantan thermocouple was placed next to the preparation in the bath to monitor temperature.

Resting membrane potential varied quite widely from cell to cell, ranging from −40 mV to −80 mV. The lower values did not usually appear to be associated with cell damage and could be sustained for several hours, indicating a genuine wide Idifference in membrane potential between cells. It should be noted that the cells do not form a homogeneous population but vary widely in size (400-1200 μm in diameter) and histochemistry (Sawyer et al. 1982). In a sample of 100 cells from 40 glands the average membrane potential was -56·1 mV ± 9·7 mV (S.D.). In normal saline there was no spontaneous spike activity or evidence of synaptic input such as the miniature potentials found in salivary cells of the cockroach (House, 1980; Ginsborg & House, 1980) and snail (Kater et al. 1978a).

Cell input resistance was between 2 and 14 MΩ. This value tended to be smaller in the larger cells but also varied with membrane potential (Fig. 1). Resistance increased with hyperpolarization up to about -80 mV and then began to decrease. Cells with high resting membrane potential thus tended to show mainly inward (anomalous) rectification. This non-linear relationship between current and voltage meant that conductance changes produced by any particular procedure had to be checked for indirect effects of changes in membrane potential. Similar current-voltage relationships were found in each of 12 cells that were studied systematically. Results using current pulses were the same as those using a continuous current which was progressively increased under manual control (see Marmor, 1975). With larger hyperpolarizing current pulses the potential reached a peak and then rapidly declined to a steady level, indicating a time-dependent inward rectification which increased with strength of pulse (Fig. 1Bii; see Stuenkel, 1985; Bostock & Grafe, 1985). Similar delayed rectification was apparent for subthreshold depolarizing pulses. When a depolarizing or hyper-polarizing pulse was terminated there was a brief (‘rebound’) hyperpolarization or depolarization, respectively. The transient depolarization increased with the size of the hyperpolarizing pulse until threshold was reached and an action potential was elicited (Fig. 1).

No calculations of specific membrane resistance were made because the membrane is deeply and extensively infolded (Walz et al. 1988) so that membrane surface area cannot easily be estimated from cell diameter. In addition, there is the presence of the ductule to consider; this is a single, long process of the cell which leaves the gland and enters the proboscis (Sawyer et al. 1982).

Simultaneous recordings from pairs of cells confirmed the results of previous studies that there is no intercellular coupling (Marshall & Lent, 1984; Jones et al. 1985). This property is unusual because the cells of almost all exocrine and endocrine glands of vertebrates and invertebrates are coupled to their neighbours (Petersen, 1976, 1980; see Kater, 1977, for an exception). It seemed possible (if unlikely) that the presence of coupling was being masked by effects of pH: low intracellular pH is known to uncouple cells (Bennett et al. 1978). Intracellular pH was therefore increased by addition of 20 mmol 1⁻¹ trimethylamine to the saline (see Thornas, 1984), although no pH measurements were made. [NaCl] was reduced by 20 mmol 1⁻¹ in compensation, and external pH was maintained at its normal level. However, applied current pulses still failed to pass from cell to cell.

The ionic basis of the resting potential was investigated by varying the concentration of external ions. In agreement with Marshall & Lent (1984), at high external K⁺ concentrations the values for resting potential conformed closely to the slope predicted by the Nernst equation for a potassium equilibrium potential. At low K⁺ concentrations, however, the membrane potential did not simply deviate from predicted values but showed a depolarization of 1-5 mV in Immoir¹ K⁺ and 3-10mV in Ommoir¹ K⁺ (Fig. 2A). This could be attributed to a block of an electrogenic sodium pump (see below).

Na⁺ removal produced a hyperpolarization ranging from 5 to 35 mV (mean 15 ±6·5 mV, N = 18). The maximum level was reached in about 3 min and was associated with an increase in membrane resistance (Fig. 2B); recovery occurred quickly on reintroduction of Na⁺. After a few minutes in Na⁺-free solution, however, the membrane potential usually recovered slowly, and replacement of Na⁺ produced a small transient depolarization. Marshall & Lent (1984) reported no change in membrane potential or input resistance on Na⁺ removal but theyl used a different substitute (arginine). They did not investigate the effects of Ca²⁺ or Cl⁻ on membrane potential, and these ions were therefore studied here in some detail.

Removal of Ca²⁺ and addition of 1 mmol 1⁻¹ EGTA produced a depolarization of 15 mV (±5 mV, N =16) and a reduction in membrane resistance to 54% (±23%, N = 9) within about lmin. Excitability was increased and spontaneous action potentials were usually produced, making the measurements of potential and resistance difficult. Fig. 2C shows an example where spontaneous firing did not occur, and resistance measurements were made with pulses that were too small in amplitude to elicit rebound spikes. The effects of prolonged Ca²⁺ removal on resting membrane potential are described in the next section.

Effects of low-chloride (3-6 mmol 1⁻¹) or chloride-free solutions were examined using three different chloride substitutes. Sulphate-based saline produced a reversible hyperpolarization of 15 mV (±5·5 mV, N = 20) which was associated with an increase in membrane resistance (Fig. 3), except in one cell with a high resting membrane potential (−73 mV) where a decrease in resistance occurred. In every case, however, the resistance change was dependent on membrane potential rather than Cl⁻ removal per se (Fig. 3B). Recovery on replacement of Cl⁻ was faster than the response to its removal, and in some cases there was a small, transient depolarization. Similar results were obtained with a modified sulphate-based saline in which Ca²⁺ and K⁺ concentrations were increased by 8·2 mmol 1⁻¹ and 0·5 mmol 1⁻¹, respectively, to allow for their reduced activities produced by sulphate (Hodgkin & Horowicz, 1959; Mullins & Noda, 1963).

With methylsulphate as a Cl⁻ substitute there was a rapid hyperpolarization of 4-8mV (±2·9mV, N = 6). Gluconate also produced a small hyperpolarization (4·2 ±1·4 mV, N = 19) although in 12 cells this was transient and was followed by a depolarization of 8·6 ± 4·2 mV (Fig. 3). Membrane resistance increased during the hyperpolarizing phase and decreased during depolarization, but again the effects were attributable solely to changes in membrane potential: similar changes were obtained in normal saline by shifting the membrane potential by the same amount with applied current, and when the membrane potential in Cl∼-free saline was returned to its initial level there was no measurable change in resistance.

In no case were the responses indicative of a high Cl⁻ permeability, where one would expect an immediate depolarization on removal of chloride (because of the change in Cl⁻ equilibrium potential) followed by a hyperpolarization on return to normal Cl⁻ levels (Hodgkin & Horowicz, 1959; Barber, 1987). The large hyperpolarizations (up to 24 mV) which were found only in sulphate-based saline suggest either that this ion is not totally impermeant or that it has effects of its own. However, the fact that all three substitutes produced a hyperpolarization indicates that this is a true response to Cl⁻ removal rather than an artefact produced by the replacement ion. No systematic studies were made of its mechanism, but an increase in K⁺ permeability appeared unlikely because pulses of 10mmoll⁻¹K⁺ produced similar depolarizations in normal and sulphate-based saline.

In conclusion, the membrane potential of gland cells appears to be largely generated by K⁺, with a small contribution from an electrogenic Na⁺ pump and a somewhat variable Na⁺ permeability.

Gland cells produced action potentials in response to depolarizing current pulses or when hyperpolarizing pulses were terminated (anode-break excitation). The overshoot was up to +28 mV, peak-to-peak amplitude between 70 and 110 mV, and spike duration 100-400 ms. Repolarization occurred in two phases and there was a large after-hyperpolarization (see Fig. 4). Impulses produced by a hyperpolarizing or depolarizing pulse were of similar amplitude and configuration, but repetitive activation tended to result in progressively smaller action potentials. Adaptation was rapid, and rarely more than three or four spikes were produced by maintained depolarization.

Marshall & Lent (1984) concluded that the action potential is dependent on Ca²⁺. We have repeated most of their experiments and can confirm their conclusion. For example, removal of external Na⁺ was without effect on spike amplitude, the overshoot was increased by raised Ca²⁺ concentration, Co²⁺ blocked the action potential, and Ba²⁺ and Sr^24- could substitute for Ca²⁺.

Fig. 4 compares an action potential elicited in normal saline with one recorded after 5 min in saline containing 7V-methyl-D-glucamine in place of Na⁺ (when the liyperpolarizing response to Na⁺ removal had reached a steady level). Typically there was no effect of Na⁺ removal on the rate of rise or amplitude of the action potential (Fig. 4), although sometimes the amplitude increased slightly. Increasing the concentration of external Na⁺ tended to reduce spike amplitude. This indicates a lack of any direct involvement of Na⁺ in the generation of the action potential. In four out of nine cells, however, Na⁺ removal reduced spike duration by shortening the plateau phase (Fig. 4). Measurements were made for at least 30 min in Na⁺-free saline to ensure removal of extracellular Na⁺ from the gland, although evidence is presented below that this occurs very rapidly.

Ca²⁺-free saline was found not to abolish the action potential but to depolarize the cell and increase its excitability, resulting in action potentials of reduced amplitude (peaking at about OmV) and greatly increased duration. Examples of typical responses to Ca²⁺ removal are shown in Fig. 5. In general, spikes progressively lengthened, lasting up to several minutes (Fig. 5A), but in some cases they became much longer (15 min or more) or showed no repolarization until Ca²⁺ was replaced or a large inward current was injected (Fig. 5B,C). The depolarization induced by Ca²⁺-free solution was associated with a decrease in membrane resistance which became especially pronounced during the peak of the action potential. The depolarizing phase of the spike showed a gradual reduction associated with a progressive increase in membrane resistance until the spike was suddenly terminated (see first few spikes of Fig. 5C). It was usually possible to predict which action potentials would be greatly prolonged because, after an initial small repolarization (as if about to terminate the spike), the membrane depolar-ized again and input resistance remained low or decreased further (see longest response in Fig. 5C). These long-duration depolarizations with a maintained plateau will be referred to as plateau potentials (see Yang & Lent, 1983).

The prolonged action potentials produced by Ca²⁺ removal were reduced in amplitude by about 10 mV when external Na⁺ concentration was halved to 63 mmol 1⁻¹ (Fig. 6); they were reduced by a further 10 mV in 31 mmol T¹ Na⁺ (not shown) and were rapidly and reversibly abolished by complete removal of Na⁺ (Fig. 7). Li⁺ could substitute for Na⁺, producing action potentials of similar amplitude and duration (Fig. 7).

Action potentials in Ca²⁺-free saline containing either Na⁺ or Li⁺ were blocked by addition of 5 mmol 1⁻¹ CoCl₂ (Fig. 8). These results suggest that the action potentials are produced by Na⁺ or Li⁺ flowing through Ca²⁺ channels.

There are now known to be several examples of normally specific Ca²⁺ channels which become non-selectively permeable to monovalent cations in the absence of Ca²⁺ (e.g. Hess & Tsien, 1984). In such circumstances it has often been found that addition of trace amounts of Ca²⁺ which are too low in concentration to carry significant inward current may block the movement of other ions through the channel (Aimers et al. 1984). Some indication of this phenomenon was observed in the gland cells when Ca²⁺-free saline (which, with 1 mmol 1⁻¹ EGTA, probably contained less than 10⁻⁸mol 1⁻¹ free Ca²⁺; Miller & Mörchen, 1978) was replaced by a solution containing buffered Ca²⁺ at a concentration between 2xl0⁻⁵ and 10⁻⁴moll⁻¹. Spike duration quickly became reduced and small changes were seen in spike amplitude: in some cases there was an increase but in others a consistent and reversible decrease (Fig. 9). At the concentration of Ca²⁺ which produced these effects, Ca²⁺ did not appear to contribute significantly to the action potential |because regenerative responses were abolished by Na⁺ removal (not shown). The precise nature of this reduction in spike amplitude and duration by trace amounts of Ca²⁺ will need to be clarified by measurements of membrane current, but the results are consistent with a block by Ca²⁺ of Na⁺ or Li⁺ passing through Ca²⁺ channels.

Ca²⁺ removal produced Na⁺ spikes rather than abolition of spike activity. Fig. 10A shows that even after about 1 h in Ca²⁺-free solution containing 5 mmol 1⁻¹ EGTA the peak depolarization of the spike showed practically no change. After Ih there was a steady decline in spike amplitude, probably caused by the associated depolarization rather than removal of Ca²⁺per se. The eventual depolarization and cessation of spiking were not unexpected since prolonged Ca²⁺ depletion is likely to produce deleterious effects, and EGTA may have toxic effects of its own at the concentration used (Miller & Mórchen, 1978).

Fig. 10B shows a recording from a different cell where peak-to-peak spike amplitude increased dramatically after immersion for 35 min in Ca²⁺-free solution. However, for the full 35-min period there was no change in peak depolarization indicating no tendency for Ca²⁺ removal to abolish action potentials.

The sudden hyperpolarization seen in Fig. 10B is an extreme example, but there was a general tendency for the membrane potential in Ca²⁺-free saline to increase at some stage beyond the original level found in normal saline: this was in spite of the initial depolarizing response to Ca²⁺ removal. In Fig. 10A, for example, the membrane potential following the second plateau potential is held negative to the initial level for about 5min. Hyperpolarizations beyond the original membrane potential were also observed after a period of exposure to Ca²⁺-free solution when normal saline was reintroduced. This is evident in the three recordings in Fig. 5 and for the first (short) period of Ca²⁺ removal in Fig. 10A. In view of the extreme duration of Na⁺-dependent action potentials in Ca²⁺-free saline, and the associated high membrane conductance, it seemed likely that intracellular [Na⁺] would rise markedly and that the hyperpolarizations may reflect increased activity of an electrogenic Na⁺ pump. In support of this view it was found that treatments that are known to block the pump, such as removal of external K⁺ or addition of ouabain, produced a large decrease in potential if applied during the hyperpolariz-ing phase following Ca²⁺-free spike activity (Fig. 11). The hyperpolarization and pump activity increased with time spent in Ca²⁺-free solution (Fig. 11A) and they were especially large if the pump was blocked during the period of firing by removal of K⁺ (presumably because this increased intracellular Na⁺). Increased activity of the pump was evident for more than 1 h after a 10-min period of Ca²⁺ (and K⁺) depletion, and progressively declined as the membrane potential gradually recovered (Fig. 11B).

When Li⁺ replaced Na⁺ in these experiments, spike activity was not followed by a hyperpolarization or increase in Na⁺ pump activity. Since Li⁺ is not a substrate for the Na⁺ pump (Thornas, 1969) this provides further evidence that the hyperpolarization which follows Na⁺ spikes in Ca²⁺-free saline is indeed gener-ated by a Na⁺ pump.

Calcium channels are known to be very sensitive to temperature (Byerly et al. 1984; Narahashi et al. 1987). However, little work seems to have been done on the effects of temperature on Ca²⁺ spikes and those produced by non-specific current when Ca²⁺ is removed. We were particularly interested to see whether low temperature would reduce the variability of spike duration in Ca²⁺-free saline by producing predominantly plateau potentials.

The effects of rapid cooling by 10°C are shown in Fig. 12. In normal saline there vas a depolarization of 13-5 mV (±5 mV, N = 19) and an increase in membrane esistance (despite the depolarization) of 180% (±27%, N= 16). Action poten-tials showed an increase in overshoot of 3-8 mV and a decrease in undershoot of 4-9 mV; depolarizing and repolarizing phases were slowed and spike duration was increased 3-to 4-fold. In Ca²⁺-free saline there was also a depolarization (6-9 mV) and increase in resistance on cooling, though spontaneous impulse activity usually made measurements difficult. Peak depolarization of the spike was now reduced in amplitude, however, by 10-15 mV and spike duration was unaffected or reduced (Fig. 12B). This reduction in spike amplitude was much too fast to be caused by a slowing of the sodium pump. The results suggest that the Ca²⁺ channel may be differentially affected by temperature depending on the presence or absence of^ Ca²⁺.

The purpose of the present experiments is to provide an electrophysiological basis for studies on excitation-secretion coupling in the salivary gland of Haementeria, a preparation whose special advantages for this type of work have been well documented (Marshall & Lent, 1984; Jones et al. 1985; Sawyer, 1986). Our results confirm the conclusion of Marshall & Lent (1984) that the salivary cell action potential is dependent on Ca²⁺ [most of the criteria suggested by Hagiwara & Byerly (1981) to identify Ca²⁺-dependent action potentials have been satisfied]. Action potentials were found to persist for an hour or more in zero-Ca²⁺ solution containing 5 mmol 1⁻¹ EGTA although one of the tests of Ca²⁺ dependence proposed by Hagiwara & Byerly (1981) is a block of the action potential by removal of external Ca²⁺. There are several examples, however, where this does not occur because the action potential changes to a dependence on Na⁺ when Ca²⁺ is removed (Prosser et al. 1977; Miller & Mörchen, 1978; Minota & Koketsu, 1983; Yang & Lent, 1983; Yoshida, 1983; Jmari et al. 1987). The Na⁺ current passes through Ca²⁺ channels which become permeable to monovalent cations in the absence of Ca²⁺ (Hess & Tsien, 1984; Fukushima & Hagiwara, 1985; Lansman et al. 1986; Matsuda, 1986; McCleskey et al. 1986; Tsien et al. 1987). The criterion of Hagiwara & Byerly (1981) that removal of Ca²⁺ should block a Ca²⁺-dependent action potential is, of course, satisfied if the test is made in the absence of Na⁺. It is noteworthy that, in the examples given above, the action potential produced by Ca²⁺ removal tends to be smaller, much longer lasting and more variable in duration than normal, but maintains a fairly rapid repolarizing phase (as in Haementeria).

Although the ability to pass non-specific current may be a general property of Ca²⁺ channels (Aimers & McCleskey, 1984), there are several examples of Cam-dependent action potentials that do become blocked by Ca²⁺ removal (Patlak, 1976; Weisblat et al. 1976; Fukuda et al. 1977; Mizunami et al. 1987) or become greatly attenuated (Hadley et al. 1980; Goldring et al. 1983). Thus there remained the possibility in the present experiments that failure of the action potential to disappear in Ca²⁺-free solution was due to incomplete washout of Ca²⁺ from the gland. This is most unlikely, however, because a depolarization and increase in membrane conductance occurred within a minute or two of Ca²⁺ removal, and the peak depolarization of the action potential rapidly declined to a level which remained steady for about 1 h, even during large changes in spike undershoot. Most importantly, Na⁺ removal barely affected the normal spike but produced a rapid, reversible block of impulse activity in Ca²⁺-free saline. Finally, activation of the Na⁺ pump by impulse activity, and the rapid effect on spike configuration of changing from Ca²⁺-free to low-Ca²⁺ solution are also explained in terms of effective exchange of Ca²⁺ in the gland. In Ca²⁺-free solution the salivary cell action potentials are maintained by Na⁺ or Li⁺ which appear to pass through Ca²⁺ channels because the spikes are blocked by 5 mmol 1⁻¹ Co²⁺.

It is worth reiterating that studies on the salivary cells of Haementeria are free from the problems of electrical coupling which may seriously complicate the interpretation of results in other glands. For example, Hadley et al. (1980) could not perform reliable quantitative measurements of action potential characteristics in salivary cells of Planorbis because spike amplitude and configuration varied with the number of neighbouring cells which were firing (functional coupling tends to become reduced when cells fire synchronously; see Getting, 1974). In many circumstances it is difficult to determine whether recorded electrical activity comes from the impaled cell or spreads from coupled neighbours; this problem is particularly serious if different cell types are present, when even qualitative changes may become difficult to interpret. Ion substitution experiments may be complicated by effects on coupling resistance in addition to membrane resistance (removal of external Ca²⁺ or Cl⁻, for example, may produce uncoupling, with a resultant increase in input resistance; Asada & Bennett, 1971). None of these problems occurs in Haementeria, and the absence of innervation in the gland (W. A. Wuttke, R. T. Sawyer & M. S. Berry, in preparation) eliminates problems of indirect actions on presynaptic elements (see Ascher et al. 1976).

Electrical excitability is not a feature of mammalian exocrine glands (Petersen, 1980) but is found in certain endocrine gland cells such as the pancreas (Matthews & Sakamoto, 1975), adenohypophysis (Kidokoro, 1975) and adrenal gland (Brandt et al. 1976). Among invertebrates, the salivary glands of insects are inexcitable (Ginsborg & House, 1980; House, 1980) whereas those of molluscs produce action potentials (Kater et al. 1978b; Goldring et al. 1983; Barber, 1983). Molluscan pedal gland cells (which secrete mucus) also produce impulses (Kater, 1977). Electrically excitable secretory cells invariably seem to have a Ca²⁺ component to their spikes, although there may also be a large or small contribution by Na⁺. In Haementeria, only a Ca²⁺ component is evident.

It is generally accepted that action potentials in gland cells act as a stimulus for secretion by providing an influx of Ca²⁺ which is necessary for exocytosis (Hagiwara & Byerly, 1981). The situation is analogous to Ca²⁺ entry into the presynaptic element of a nerve cell (the similarity is particularly marked in Haementeria where the impulse travels along a ductule towards the release site in the proboscis). We were surprised, therefore, to find in our early experiments that Ca²⁺ removal increased the amount of secretory product around the tip of the proboscis (W. A. Wuttke & M. S. Berry, unpublished results), but this accords with our present findings that Ca²⁺ removal actually excites the gland cells. There remains the problem, however, of the mechanism of secretion in the absence of external Ca²⁺. One possibility is that Na⁺ influx results in the release of Ca²⁺ from intracellular stores and this triggers release of secretory products (Lowe et al. 1976). The most likely explanation, however, is that Ca²⁺ is not easily washed from the compact, muscular proboscis. Stimulation of the nerve to the proboscis elicited contractions after more than 30 min in Ca²⁺-free saline containing 5 mmol 1⁻¹ EGTA, indicating the presence of sufficient Ca²⁺ for release of neurotransmitter. Long periods have also been found necessary to remove Ca²⁺ from ganglia in Aplysia (Kehoe,1969). It must be stressed, however, that although our results may be in agreement with a Ca²⁺-dependent, impulse-evoked release of secretion, they do not provide any positive evidence. Experiments are under way on the mechanism of secretion in Haementeria but little is currently known.

We thank Dr R. T. Sawyer for many helpful discussions. This work was supported by a grant from the Science and Engineering Research Council to MSB and R. T. Sawyer (no. GR/D/54774).