Electrical activity of the proventriculus of the polychaete worm Syllis spongiphila

The proventriculus of the marine polychaete worm Syllis spongiphila is the section of the alimentary canal, about 1 mm in length, that lies between the proboscis and the intestine. It is composed of a single layer of striated columnar myoepithelial cells that are arranged radially in annular rows around a central lumen (Fig. 1) (Haswell, 1890; Smith, del Castillo & Anderson, 1973). This organ draws sea water and small particulate matter into the alimentary canal by expansion of its lumen. At rest, the lumen is slightly open ; during expansion it is opened by shortening of the muscle cells, and materials are sucked inward via the mouth, pharynx and proboscis. This phase of activity is associated with depolarization of the myoepithelial cells (del Castillo, Anderson & Smith, 1972). The lumen then closes by relaxation, and often lengthening (Anderson et al., in preparation), of the cells, and the materials are pushed in a caudal direction into the intestine. This phase of activity is associated with hyperpolarization of the myoepithelial cells (del Castillo et al. 1972). There seem to be valves at the anterior and posterior ends of the proventriculus that ensure that materials move toward the intestine (Haswell, 1890). In the living proventriculus waves of contractions and relaxations can be seen moving peristaltically in a caudal direction.

In cross-section the lumen of the proventriculus appears as an oval slit with a dorso-ventral axis. Haswell (1890) called the dorsal and ventral grooves of the proventriculus ‘raphes’. The cells that project from the lateral aspects of the lumen may reach 100 μm in length, whereas those that project from the lumen in a dorsal or ventral direction are 40−50 μm in length (Smith et al. 1973). Haswell (1890) observed that the shorter cells adjacent to the dorsal and ventral raphes lack a Z line. Ultrastructural studies (del Castillo et al. 1972 ; Smith et al. 1973) have shown that these cells are composed of a single sarcomere: a central A band, with the thin filaments of the flanking I bands inserted on the ends of the cell.

The other cells of the proventriculus have two sarcomeres ; they are characterized by a single, centrally located Z line (Haswell, 1890; confirmed by Malaquin, 1893 ; Okada, 1928; Schmidt, 1935). Electron micrographs (del Castillo et al. 1972; Smith et al. 1973) show that thin filaments inserted on the Z line extend into the flanking sarcomeres and interdigitate between the thick filaments of each A band. The second sets of thin filaments of the two sarcomeres insert directly on the ends of the cell. The sarcomere length, measured as the H-to-H distance, reaches 40 μm in the lateral cells (Smith et al. 1973).

Because the cells of the proventriculus possess sarcomeres that not only are exceptionally long but also may occur singly, they offer a new level of resolution for physiological studies. In this paper we describe experiments that were performed to study some of the electrophysiological properties of these unusual myoepithelial cells and their synaptic control.

Specimens of Syllis spongiphila were collected in mangrove swamps of northeast Puerto Rico and the Bay of San Juan and shipped to Massachusetts. In the laboratory they were kept individually in small dishes of sea water at 20 °C for periods of up to 1 month. The animals were kept separately because they tended to feed on each other. It was found late in this study, however, that up to 20 animals can be kept in a single, covered 3 in. Petri dish containing a substrate of cheese cloth in sea water about 5 mm in depth.

A worm was placed in a small dish lined with Silgaard containing either natural or artificial sea water (ASW). The artificial sea water used for most experiments, ASW₁ (from Hagiwara & Kidokoro, 1971), had the following ionic composition, expressed in mm/1: Na, 393 ; K, 9; Ca, 9·3 ; Mg, 48·4; Cl, 517·4. The pH was adjusted to 7·4 with 10 mm Tris-HCl. A second artificial solution, ASW₂, was used for the experiments described in the last two parts of the Results section (Influence of Membrane Potential on the Synaptic Potentials and Electrical Coupling). It had the following ionic composition, expressed in mm/1: Na, 457; K, 9·7; Ca, 10·1 ; Mg, 52·5; Cl, 534; HCO₃, 2·5; SO₄, 27.7 (Welsh, Smith & Kammer, 1968). No obvious differences in the activity of the preparations were detected in the two different bathing media.

The body wall around the proventriculus was removed using watchmaker’s forceps. The head and part of the proboscis were drawn into the tip of a suction electrode for indirect stimulation, and the body of the worm posterior to the proventriculus was pinned to the bottom of the chamber with stainless-steel insect pins. Experiments were performed at temperatures ranging from 21 to 23 °C.

Pulses of 0·2–0·5 ms duration were applied via the suction electrode at varying frequencies and intensities to evoke activity from nerve fibres that innervate the proventriculus.

Glass, 3 M-KCl-filled microelectrodes of 10−20 MΩ resistance were used for recording from and injecting current into single muscle cells. To record, the electrodes were connected to the oscilloscope through high input impedence, capacity-neutralized preamplifiers. To inject current, one electrode was connected to a stimulator in series with a 100 MΩ) resistor. The applied current was measured by recording the voltage drop across the input resistance (1 MΩ) of the oscilloscope or across a 10 KΩ resistor to ground. By means of switches, either electrode could be selected to record or to inject current. Recordings were photographed on film with a kymograph camera.

Specific experimental solutions and various modes of stimulation and recording are described in detail below.

The resting potentials of the myoepithelial cells were similar in the two bathing media: 62·5 ± 6·3 mV (S.D., n = 98) in ASW₁ and 59·1 ± 5·5 mV (s.d., n = 91) in ASW₂. These values are about 10 mV greater than those previously reported (del Castillo et al. 1972); further experiments have shown that the low values of resting potentials that were taken into account to arrive at the earlier mean value are not typical.

To determine to what degree the resting potential is dependent upon the distribution of K⁺ ions across the resting membrane, the concentration of potassium in the external bath was varied from 1 to 90 mm, and the resulting changes in the resting potentials were recorded. To eliminate the influence of changing external concentrations of chloride ions (Hodgkin & Horowicz, 1959), these experiments were performed using chloride-free solutions. Either acetate or methane sulphonate anions were substituted for chloride ions, on the assumption that these large anions do not permeate the cell membrane. Each time the bathing solution was changed, the chamber was flushed five times with the new solution. The preparation was allowed to equilibrate from 30−120 s. Three to five cells were then impaled in sequence. Each penetration was maintained for at least 30 s. The electrode was then withdrawn and the resting potential photographed. After each series of resting potentials in a given test solution, the bath was returned to normal ASW and a set of control resting potentials recorded.

Fig. 2 shows the mean values of resting potentials at various external potassium ion concentrations plotted on semilog₁₀ paper. By the method of least squares, it was shown that the points for 9−90 mm-[K]_o fall on a line that has a slope of 48·0 ± 3·3 mV for a tenfold change in the external potassium ion concentration. This is 10 mV less steep than the slope of 58 mV predicted by the Nernst equation for a membrane that behaves solely as a potassium electrode. Thus, it appears that at rest the membrane potential is determined largely by the permeability of the cell membrane to potassium, but that the membrane is also somewhat permeable to other ions. For concentrations of external potassium below 9 mm, the points depart from the line in the direction of decreased membrane potentials ; this is presumably due to the effect of ions other than K+ that determine the resting membrane potential.

A description of the spontaneous activity recorded with microelectrodes from single myoepithelial cells was given by del Castillo et al. (1972). There are two main types of spontaneous events: (a) small hyperpolarizations (most frequently observed) and depolarizations (occasionally observed) usually ⩽ 15 mV in amplitude, occurring sporadically without obvious association with visually detected mechanical activity; and (b) complex, long-lasting potential changes made up of many depolarizing and hyperpolarizing events and usually associated with visible mechanical activity; the depolarizing phases often overshoot the zero reference level. Examples of spontaneous activity recorded with a microelectrode from a single myoepithelial cell are shown in Fig. 3. Potential changes similar to these spontaneous events could be elicited by indirect electrical stimulation.

Because the nerve fibres that innervate the proventriculus are very small and embedded within the raphes, it was not possible to dissect them out and stimulate them individually or in small groups. Instead, stimuli were applied to the head and anterior portion of the proboscis by means of a suction electrode.

A single indirect stimulus usually evoked a hyperpolarizing response (Fig. 4,a). The amplitudes of the single responses were variable and could be controlled only approximately by adjustment of the intensity of the stimulus. In most preparations depolarizing responses were elicited only with repeated stimuli (see below); often, however, single stimuli applied to the anterior end of the animal gave rise to biphasic responses that can be interpreted as composed of an initial depolarizing response followed by a hyperpolarizing one (Fig. 4,b, c). It was often but not always possible to evoke single depolarizing responses by apparently fatiguing or inactivating the nerve fibres that produce hyperpolarizing synaptic events by prior application of a train of high frequency stimuli lasting several seconds. The effect of this stimulation on the hyperpolarizing responses, when successful, was essentially irreversible. The preparations would continue to give depolarizing responses for several hours, but hyperpolarizing responses would seldom reappear. Indeed, it was found in preparations that had not been vigorously stimulated but that were 4-6 h old that the hyperpolarizations but not the depolarizations became difficult to elicit. A depolarization from such a preparation is shown in Fig. 4(d). It is possible that the nerve fibres that elicit hyperpolarizations are more susceptible to damage and deterioration than those that elicit depolarizations; or the mechanism(s) for transmitter renewal after depletion may operate very slowly in the terminals that elicit hyperpolarizations.

A comparison of the hyperpolarizing and depolarizing events in Fig. 4 shows that the time course of the depolarizing synaptic potentials is faster than that of the hyperpolarizing ones. In 32 records from 5 preparations that permitted measurement of simple depolarizations, the rise time to peak ranged from 20 to 35 ms. The rise time to peak of simple hyperpolarizations, measured in 103 records from 6 preparations, ranged from 25 to 75 ms. (Simple depolarizations occurred in two of the six experiments exhibiting simple hyperpolarizations.) The mean values of the rise times varied significantly from experiment to experiment (P ⩽ 0·05), so it was not possible to compare statistically the difference between the pooled samples of the two sets of data. Nevertheless, it can be seen in the histogram of Fig. 5 that the majority of the depolarizations rise to peak amplitude faster than the majority of the hyperpolarizations. The decay times of the simple depolarizations were also faster than those of the simple hyperpolarizations. The time to decay to half amplitude ranged from 20 to 55 ms for the depolarizations (n = 29) and 62−135 ms for the hyperpolarizations (n = 87).

Fig. 6 shows the responses recorded from a single cell at different frequencies of stimulation. Frequencies up to 4 Hz gave rise to mainly hyperpolarizing potentials, characterized by a fast initial rise and slower decay. At frequencies from 5 to 8 Hz the amplitudes of the hyperpolarizations decreased slightly and small depolarizations occurred. At frequencies of 10 Hz and greater the depolarizations added to form serrated waves of varying degrees of complexity. The hyperpolarizations decreased further in amplitude with increased frequency of stimulation. Trains of 20-25 evoked activity that resembled spontaneous complex potential changes, with both hyperpolarizing and depolarizing components.

The simplest hypothesis to explain these findings is that the proventriculus is innervated by two populations of nerve fibres, an excitatory population that leads to depolarization of the myoepithelial cells and an inhibitory one that leads to hyperpolarization. Because the type of potentials produced depended mostly upon the frequency and not the intensity of stimulation, one may also suggest that the excitatory nerve terminals show a large degree of facilitation, compared to the inhibitory nerve terminals; this facilitation would be required to release sufficient transmitter to cause depolarizing potentials. Thus at low frequencies of stimulation, only inhibitory terminals release sufficient transmitter to elicit a response. With increasing frequencies of stimulation the excitatory terminals release increasing amounts of transmitter. Atwood (1973), in reviewing the differences between phasic and tonic motor axons in Crustacea, pointed out that phasic axons are often characterized by a high initial output of transmitter, poor facilitation and rapid fatigue. The inhibitory population of neurones in the Syllis preparation may be similar to the phasic axons of Crustacea. Indeed, the hyperpolarizing responses do not exhibit facilitation, they seem to decrease in amplitude with increasing frequencies of stimulation and they appear to fatigue and disappear altogether following trains of stimuli lasting several seconds. The tonic axons of Crustacea, on the other hand, are characterized by slow initial output of transmitter, good facilitation and resistance to fatigue (Atwood, 1973). These characteristics seem to resemble those of excitatory nerve fibres in the Syllis preparation which appear to have a low initial output of transmitter and which elicit depolarizing responses that clearly do not decrease in amplitude with increasing frequencies of stimulation.

An alternative hypothesis to explain the dependence on frequency of stimulation would be that one or more intemeurones is interposed in the excitatory pathway. If the interposed synapse(s) showed a high degree of facilitation and required higher rates of stimulation to activate the excitatory motoneurone, then low-frequency stimulation would elicit only hyperpolarization in the myoepithelial cells, while high-frequency stimuli would evoke depolarization. On the basis of the available evidence it is not possible to decide between these two possibilities.

To identify the ionic species that produce depolarizing and hyperpolarizing potential changes, the preparation was equilibrated with artificial sea water of different ionic compositions. The head end of the animal was stimulated with constant intensity pulses of 0·5 ms duration at a frequency of 20 Hz, and potential changes were recorded intracellularly from single myoepithelial cells.

• (a) Chloride-free solutions. Replacement of all of the chloride ions by acetate or methane sulphonate ions resulted in a complete block of the hyperpolarizing but not the depolarizing component of the complex potential changes. Fig 7 illustrates that the effect is reversible, and that the hyperpolarization is restored upon return to normal ASW. These results suggest that chloride ions contribute to the hyperpolarization by moving inward across the subsynaptic membrane.

• (b) Sodium-free solutions. When Na⁺ was completely replaced by Tris or choline, no responses at all were recorded from the myoepithelial cells in response to stimulation. This lack of response is probably due to block of conduction in the nerve fibres.

• (c) Calcium-free solutions. Removal of Ca²⁺ from the external solution led to the almost complete abolition of both hyperpolarizing and depolarizing responses (Fig. 8). It is possible that the absence of Ca²⁺ ions leads to a block of transmitter release from all presynaptic terminals or to a block of nerve conduction or both.

• (d) Influence of manganese ions. Mn²⁺ ions are known to compete with but not to replace calcium ions in various preparations (Hagiwara & Nakajima, 1966; Hagiwara & Takahashi, 1967). Fig. 9 shows the results of indirect stimulation in a solution of ASW containing, in addition to the normal concentration of calcium ions, 4 mm-Mn^a+ ions. The depolarizing component of the response was almost completely suppressed, but the hyperpolarizing component seemed unaffected by the presence of manganese ions.

If the presynaptic terminals of both the inhibitory and excitatory nerve terminals required the same amount of calcium for transmitter release, and if the presynaptic binding sites for calcium were identically susceptible to competition by manganese, then one could interpret these results to mean that the main action of manganese ions is at the postsynaptic membrane. Since the hyperpolarizing response remains in the presence of manganese ions, one could assume that presynaptic release of transmitter is not affected and that manganese probably blocks the inward movement of calcium ions across the postsynaptic cell membrane.

However, in view of the difference observed between the responses of the two populations of nerve terminals to varying frequencies of stimulation, it is not unlikely that the calcium requirements of the two populations are different. If, for example, the excitatory nerve terminals must facilitate to elicit a significant response, they may require more calcium ions than the inhibitory terminals to cause the increased transmitter release. Therefore, an alternative interpretation of these results may be that the Mn²⁺ ions in these experiments blocked specifically the release of transmitter from the excitatory presynaptic terminals. On the basis of this experiment, it is not possible to decide between these two interpretations.

To change the membrane potential of the myoepithelial cells, a microelectrode was placed in one cell and used to inject current pulses of approximately 0·8 s duration and varying intensities ; a second microelectrode was placed in an adjacent cell for recording. Because the cells of the proventriculus are coupled electrically (see the next section), it was not necessary to place the two microelectrodes in the same cell. While the membrane potentials of the myoepithelial cells were displaced by the injected current pulses, single, constant intensity pulses of 0·2–0·5 ms duration were applied with the suction electrode to the head of the animal. To determine the reversal potential of the hyperpolarizing synaptic potentials, it was important that the response to the indirect stimulus was a simple hyperpolarization at the resting membrane potential. When it was complicated by a depolarizing component (as in Fig. 4,b, c), increasing the membrane potential resulted in an increase in the amplitude of the depolarization and a consequent masking of the concomitant decrease in amplitude and reversal of the hyperpolarization. Such a result is shown in Fig. 10.

Fig. 11 illustrates the reversal of a simple hyperpolarizing synaptic response. The reversal potential in this test was –105 mV. The mean reversal potential for eight measurements taken from two preparations was –104 ± 3 mV.

To examine electrical coupling between the cells of the proventriculus, two microelectrodes were used, one to pass rectangular current pulses of 350-800 ms duration and varying intensities, and the other to record the resulting changes in potential. For each pair of impalements, the directionality of coupling was studied by reversing the roles of the two electrodes, so that current pulses were applied through the original recording electrode and potential changes were measured with the original stimulating electrode. Pairs of impalements were made at different interelectrode distances along the rostral-caudal, dorsal-ventral and left-right (i.e. across the raphe) axes. From these sets of recordings, current-voltage curves were plotted, as shown in Figs. 12 and 13.

From the current-voltage curves shown in Fig. 12 it is clear that there is extensive electrical coupling along all three directions. Further, since there was no difference when the stimulating and recording electrodes were interchanged (see filled and open symbols in Fig. 12), it is clear that there is no rectification in any of these directions.

It is interesting that the depolarizations, especially those of cells near the point at which current was injected, often tended to be greater in amplitude than the hyperpolarizations produced by current pulses of the same magnitude but opposite polarity. This is shown by the upward deflexion of the current-voltage curve in the depolarizing quadrant of Fig. 12. This phenomenon may be similar to the anomalous rectification of the K+ channels observed by Katz (1949) and others (e.g. Adrian & Freygang, 1962) in frog twitch muscle; however, we have no evidence at this time to prove that the rectification observed in the Syllis preparation is due to the properties of the K⁺ channels of the membrane. The upward deflexion of the current-voltage curve suggests a possible regenerative capability of the membrane, but depolarizations up to the maximum we have been able to produce (39 mV more positive than the resting potential) have not evoked a regenerative response.

Further information on the properties of the electrical coupling was obtained from experiments such as the one shown in Fig. 13, in which electrodes were placed at different distances apart in the rostral-caudal direction. When the slopes of the current-voltage curves (obtained from the hyperpolarizing quadrant) were plotted against distance between electrodes on semi-log₁₀ paper, a straight line resulted (Fig. 14). These data indicate that the electrotonic potentials decrease exponentially with distance from the site of current injection.

Length constants were determined from five experiments of this type in which data had been obtained from three or more pairs of impalements at different distances. The mean value of the space constant, λ, was 0·52 mm (range 0·40–0·66 mm). Assuming a diameter of about 25 μm for a single cell (Smith et al. 1973), the mean value of λ indicates that an event will decay to 1/e of its value over a distance of about 20 cells from the point of current injection.

The presence of extensive electrotonic coupling between the myoepithelial cells of the proventriculus raises the question of whether the coupling might be responsible for coordination of contraction during peristaltic swallowing. This was examined by comparing two experiments. In both experiments two microelectrodes were used to record simultaneously the responses of myoepithelial cells at two different positions in the rostral-caudal direction. In the first experiment (Fig. 15) a third microelectrode was used to inject current pulses into a cell rostral to the two recording electrodes. In the second experiment (Fig. 16) a suction electrode was used to evoke responses indirectly via nerve fibres that innervate the proventriculus. In Fig. 15 it is clear that the caudal response is much slower and smaller than the rostral response, as would be expected from the previously described characteristics of the coupling. In addition, since the cells do not show clear regenerative properties, it is unlikely that the coupling among them could be a major factor in the transmission of activity along the proventriculus. In Fig. 16, where responses were indirectly evoked from two pairs of cells 0·37 and 0·83 mm apart, there is no indication of a delay between the initiation of the rostral and caudal postsynaptic events. These results suggest that nerve fibres innervate the proventriculus at many points along its length, and that significant potential changes will occur in cells due to direct synaptic innervation before the potential changes due to electrotonic coupling have reached equivalent levels. To detect a delay at the electrode separations and sweep speeds used in these experiments, the conduction velocities of the nerve fibres would have to be on the order of 0·4 ms or slower. The actual conduction velocities are not known; electron micrographs of the nerve fibres do not reveal obvious myelination (Smith et al. 1973), and the largest axons observed in cross-section are only 2−3 μm in diameter (although it must be emphasised that a systematic study of the neural elements has not been made). On the basis of Rushton’s (1951) theoretical predictions, unmyelinated fibres of 1−3 μm in diameter would conduct at 2−4 m/s. Assuming a velocity of 2 m/s and no intemeurones in the conduction pathway, then the time required to conduct an event over the o·8 mm distance between the two recording electrodes would be only 0·4 ms. In fact, neural conduction would traverse the entire length of a proventriculus 1 mm in length in 0·5 ms. Since peristalsis may require some hundreds of milliseconds for a wave of contraction to travel along the proventriculus, the delay between activation of rostral and caudal cells in peristalsis is not likely to be due to the time required for neuronal conduction. But neither is the sequential activation of cells likely to be due to electrotonic coupling (as argued above). A reasonable hypothesis for the present is that a pattern-generating neuronal network located rostral to the proventriculus coordinates peristalsis and that cells of the proventriculus are activated synaptically either directly or via electrotonic coupling with their close neighbours.

It is clear from the experiments described that potential changes can be produced in the muscle cells of the proventriculus both by synaptic activity and by electrotonic propagation from neighbouring cells. The experiments using indirect stimulation strongly suggest that the proventriculus is innervated by two populations of nerve fibres, one inhibitory and one excitatory. However, because these fibres are small and inaccessible, it is not possible at present to test whether there are interneurones involved in the pattern of innervation or whether there are differences in fatigability and properties of facilitation of the two populations of nerve endings.

Ultrastructural studies of the Syllis proventriculus (Smith et al. 1973) have not revealed the presence of ‘typical’ neuromuscular synapses (i.e. regions of close apposition between the nerve terminal membrane and the sarcolemma). Indeed the only vesicle-filled nerve terminals seen in electron micrographs are separated from the muscle cells by a space of 0·3–1·0 μm which is filled with loosely grouped collagen fibres. Whether or not these are the nerve terminals that directly control the activity of the myoepithelial cells remains uncertain, since it is always possible that conventional junctions exist but have not been detected so far. It must be emphasised that the large distance between the observed nerve terminals and the myoepithelial cells is not sufficient to disqualify them as the presynaptic components of the nerve-muscle junctions in this preparation. Indeed, autonomic postganglionic neurones that innervate heart and smooth muscle possess numerous varicosities along their terminal branches from which transmitter substances are thought to be released; some varicosities may be only a few hundred ångströms distant from the fibres while others may be much farther away (e.g. Grillo, 1966; Devine & Simpson, 1967). It is interesting that the terminals observed in electron micrographs of the proventriculus appear to be of two classes, one which possesses large, spherical vesicles and another which possesses small, flattened vesicles (Smith et al. 1973). Studies in a crustacean muscle (Atwood, Lang & Morin, 1972) have shown that vesicles of different morphological appearance mediate different functions, (i.e. excitation and inhibition). It is possible that the same occurs in Syllis.

There is considerable electrical coupling between the cells of the proventriculus. Ultrastructural studies have shown that the apposed membranes of these cells are separated by an extracellular gap of about 20 Å and that the membranes are interdigitated in localized regions; no specialized intercellular connexions have been observed (Smith et al. 1973). It is likely that the narrow gap between cells presents a large resistance to radial ionic flow toward the outside surface of the proventriculus and that this resistance is further increased by the tortuous path resulting from the interdigitation of the membranes. Thus if the relative resistance encountered by current flow is greater along the gap than that encountered across it, the net result will be for current to flow from cell to cell instead of along the gaps between cells.

There is also coupling across the raphes. In electron micrographs these grooves appear as spaces devoid of cells about 35 μm wide (Smith et al. 1973). In this case it can be hypothesized that the resistance to current flow within the raphe is low whereas that to current flow across the extracellular coverings of the inner and outer surfaces of the proventriculus is relatively high, so that net current flow occurs across the raphe.

During synaptically elicited hyperpolarization it appears that the membrane increases its permeability to Cl⁻ ions. The possible contribution of K⁺ ions to the hyperpolarizations can be largely excluded by the fact that no hyperpolarizations have been observed during stimulation in chloride-free solutions. It seems unlikely that the identity of the ion or ions that are involved in the excitatory synaptic potentials can be determined by means of indirect stimulation, since manipulation of Na⁺ or Ca²⁺ ions, the most likely candidates, interferes with either nerve conduction or transmitter release or both. Therefore, it appears that it will be necessary to first identify the depolarizing transmitter substance and then apply it iontophoretically while changing the ionic concentrations in the bath.

Since the depolarizations often overshoot the zero reference level and the currentvoltage curves are skewed upward in the depolarizing quadrant, it seems possible that the membranes of the myoepithelial cells are capable of regenerative activity, but so far we have not succeeded in evoking active responses by applied depolarizations. It is also possible that the overshoot may be due to synaptic activity alone. Further experiments are required to distinguish between these two possibilities.

We wish to thank Mr F. McKenzie for supplying us with animals. We are grateful to Dr Charles Edwards, State University of New York at Albany, for the loan of equipment and many helpful suggestions. Finally we wish to thank Dr Richard F. Olivo for critically reading portions of the manuscript. This work was supported in part by National Institutes of Health Grant i Roi NS12196-01 PHY to M.A. and USPHS NIH grant nos NS-07464 and K6N-14938 to J.d.C.