Primary Culture of Identified Neurones From a Cnidarians

Despite strong evidence, both morphological and physiological, that chemically mediated synaptic transmission occurs in all the cnidarian (coelenterate) classes (see Martin & Spencer, 1983; Satterlie & Spencer, 1988, for reviews), there is insufficient information to identify the transmitter substances. There are, however, several preparations in which it is possible to record intracellularly both pre-and postsynaptically in the Hydrozoa and Scyphozoa [Aglantha neuromuscular junction (Kerfoot et al. 1985); Polyorchis neuromuscular and neuroneuronal synapses (Spencer, 1982; Spencer & Arkett, 1984); Cyanea neuroneuronal synapse (Anderson, 1985)]. The opportunities afforded by these preparations have not been fully exploited. A major reason for this appears to be that the epithelium covering the synaptic sites forms an effective diffusion barrier (King & Spencer, 1979) to substances applied to the bathing medium. Some success at removing these diffusional barriers has been achieved by partial dissociation of the covering epithelium following treatment with collagenase and hyaluronidase (Spencer, 1988). Using the scyphozoan medusa Cyanea (Anderson, 1985, 1987), it has been possible to expose the subumbrellar motor nerve-net by destroying most of the overlying epithelial cells by brief oxidation of the perirhopalial tissue with sodium hypochlorite. This has enabled the study of bidirectional synaptic transmission and a voltage-sensitive sodium current. However, using this isolation technique, it has not been possible to demonstrate postsynaptic effects of applied putative transmitters that mimic natural release (P. A. V. Anderson, personal communication).

We have developed a technique that produces primary cultures of identifiable neurones from Polyorchis penicillatus that will survive for many days and has important experimental advantages over in vivo studies; for example, no diffusion barrier surrounds the neurones. Also, using cultured neurones it is possible to test whether a particular agent is having a direct effect (and not via intervening cells) since a visual check can be made for possible contact with presynaptic neurones. It has been established that all the identified neurones in Polyorchis, and probably in all hydrozoans, belong to electrically coupled networks (Spencer, 1981; Spencer & Arkett, 1984). Therefore any attempts to use such networks, either in vivo or partially isolated, to examine their biophysical properties using voltage-clamp methods will be complicated by neighbouring, coupled cells acting as current sinks and thus preventing the experimenter achieving an effective clamp at the command voltage.

Polyorchis penicillatus was used to develop this culture technique as it is the cnidarian species for which we have most information concerning the distributions, properties and synaptic interactions of identified neuronal networks. Four discrete neuronal networks have been described: (a) the giant swimming motor neurone network (SMN) which innervates the striated muscle of the bell (Spencer, 1981); (b) the burster or B network whose neurones are excitatory presynaptic neurones to the SMNs and the tentacle longitudinal muscle (Spencer & Arkett, 1984); (c) the oscillator or O network whose neurones are photosensitive and are probably presynaptic to both the SMNs and the B neurones (Arkett & Spencer, 1986a,b); (d) the network of neurones that are immunoreactive to antisera raised against peptides containing Arg-Phe-amide (Grimmelikhuijzen & Spencer, 1984). Three neuronal types have now been identified in culture (SMN, B-like and RFamide-containing). These provide the essential cellular elements for examining chemical transmission; namely, a target neurone (SMN) suspected of being responsive to the neuromodulatory effects of Arg-Phe-amide carboxyterminal peptides (Spencer, 1988), a neuronal type that may release Arg-Phe-amides, and a neurone that resembles a type (B) which is known to produce EPSPs in SMNs.

This study describes a technique for culturing hydrozoan neurones and gives preliminary evidence that such neurones show properties of electrical excitability in vitro that are very similar to those seen in vivo. It can be expected that such cultures could be used not only to study the neuropharmacology of chemical transmission but also to give essential, basic data on the membrane currents that are present in each of these neuronal types. Data such as this is lacking for the Hydrozoa and is essential to enable us to reconstruct the early evolution of excitable systems in the Metazoa.

Specimens of Polyorchis penicillatus were collected from Bamfield Inlet, on the West coast of Vancouver Island, and kept in running sea water (10–12°C) at Bamfield Marine Station or in cooled, recirculated artificial sea water (8–13°C) in Edmonton, for about 1 week.

The procedure used for dissociation is summarized in Fig. 1. All steps were carried out at room temperature (20–25°C), unless stated otherwise. Jellyfish with a bell diameter of 1·5–2 cm were used. Vela were removed, in an anaesthetic solution of 1:1 isotonic MgCl₂ and sea water, by cutting as close to the ring canal as possible. Strips of velum containing the nerve-rings were cut out, divided into 1–2 mm segments, and placed for 10 min in a silicon-coated test tube (Siliclad, Clay-Adams; used at 0·01% v/v in distilled water) containing artificial sea water (ASW) of the following ionic concentrations (in mmol l⁻¹): NaCl, 376; MgCl₂, 41·4; Na₂SO₄, 26; KC1, 8·5; CaCl₂, 10; Hepes, 10; and gentamycin sulphate, 50 mg l⁻¹ (Sigma G-3632). pH was adjusted to 7·5 using Imoll⁻¹ NaOH. The tissue was rinsed again for 5 min in ASW, and then exposed to ASW free of divalent cations (CaCl₂ and MgCl₂ were replaced with 76·5 mmol l⁻¹ choline chloride and 1 mmol l⁻¹ EGTA) for 10–15 min. A brief rinse in ASW was followed by digestion in collagenase (Type I or Type XI, Sigma) at 100 units ml⁻¹ (Schmid & Alder, 1984) in ASW for 3–5 h at 20–25 °C. The time required for satisfactory digestion varied from animal to animal. If cells tended to slough off easily during dissection, the digestion was stopped after 3h. Tissue from freshly captured animals tended to be less fragile and digestion was prolonged.

The enzyme solution was then replaced with 8·0 ml of ASW, some of which was pipetted off to leave a volume of 0·5–2·0 ml. The tissue was immediately triturated through the fire-polished tip (inner diameter 0·5–0·8mm) of a silicon-coated (see above, for test tube) Pasteur pipette. Approximately 10 gentle strokes against the end of the tube were usually sufficient to cause dissociation, leaving only a few pieces of mesogloea. Excessive trituration tended to reduce the yield of large neurones. In later experiments, a 10 min rinse in ASW free of divalent cations immediately followed digestion, and preceded trituration in normal ASW, to reduce aggregation (Fig. 1).

The cell suspension obtained from one animal was divided among 2–8 culture dishes coated with a layer of mesogloeal substratum (see below). Generally, a large drop (0·2–0·3 ml) of suspension was deposited in the middle 1–1·5 cm² of the dish to make the cells more accessible for electrophysiology. After cultures had settled for at least 1 h at 10–15 °C, about 2 ml of medium (unsupplemented ASW with gentamycin) was added. Cultures were held at 10–15°C and the medium replaced daily. Healthy cultures could be kept in this way for 4–10 days and, on occasion, for up to 2 weeks. In more recent experiments, the culture medium was changed only once after plating; this seemed to allow longer survival.

Several potential substrata were examined (tissue-culture plastic, slide glass, poly-lysine, gelatin). Of these, only mesogloea gave reliable attachment of cells and reasonable longevity. To prepare a mesogloeal substratum, swimming bells were dissected into halves or quarters and the peduncle removed. Gently stirring the bell sections from one animal in 50 ml of distilled water for 1 h removed all cells from the mesogloea. The mesogloeal lamella on the subumbrellar side was then peeled off and discarded. The remaining exumbrellar jelly was rinsed briefly (30 s) several times until the supernatant was no longer cloudy. Salts were diluted out by soaking the mesogloea in distilled water for at least 3 h, with water changes every hour. Cleaned mesogloea was placed with the exumbrellar surface upwards, on plastic Petri dishes (Falcon 3001 or 1008) or on glass slides and was dried flat at 25–30°C for 1–2 days in a dust-free environment.

Homogenized mesogloea was also used as a substratum. In this case cleaned mesogloea from one animal was minced coarsely, placed in 50 ml of distilled water, and blended at top speed for 30s (VirTis 23 blender). 2ml of the resulting suspension was plated in each culture dish and left to settle at room temperature. After 1–3 h, excess suspension was discarded, and the dishes were dried overnight at 25–30°C. This method produced about 20 coated dishes with mesogloea from one animal. Homogenized substrata gave optically superior images to whole mesogloea; however, cultures survived a day or two longer on whole mesogloea. Cultureware containing homogenized or whole mesogloea substratum was kept at room temperature in dust-free conditions until used (within a week), or at 0–4°C if kept longer.

The velum was dissected from the bell margin as described above, and pinned onto a Sylgard-coated Petri dish with cactus spines (from Opuntia sp.). Microelectrodes were filled either with 5% Lucifer Yellow CH (Sigma L-0259) in distilled water and backfilled with Imoll⁻¹ LiCl or with 200 mmol l⁻¹ 5(6)-carboxyfluorescein (Sigma C-7153) in 10 mmol l⁻¹ Hepes buffer at pH 7–8 (Sigma H-3375) and backfilled with Imoll⁻¹ KC1. Electrode resistances were 30-100MQ. A neurone was penetrated with a dye-filled electrode for 10–25 min. Hyperpolarizing current pulses of 2–8 nA lasting 10–30 s were alternated with brief checks of the resting potential and the presence of plateau spikes (characteristic of the SMNs; see Fig. 5) on depolarization. To limit the region through which the dye spread, two incisions were made through the nerve-ring, 2–3 mm apart. Several (up to seven) fills were made in this segment of the SMN network, until all cells (some 200 neurones) were fluorescent.

The labelled section of nerve-ring was dissected out and dissociated as described above, to give 0·2–0·3ml of suspension. This was plated on dry mesogloea on a glass slide, and allowed to settle and attach for 3–5 h in a moist chamber at 10–15°C. The slide was then briefly rinsed with ASW to remove unattached cells. For observations, a coverslip was placed on the suspension, with supporting coverslips forming a cavity. Labelled cells were photographed both under ultraviolet illumination and with Nomarski-DIC optics (see Fig. 4). In most cases, these slide cultures were kept coverslipped for one to several days in moist, cool conditions. Carboxyfluorescein remained visible for at least 3 days, but Lucifer Yellow lost much of its brightness within 1 day.

All treatments were carried out at 4°C. Two-day-old cultures were fixed for 1 day with 4% paraformaldehyde in ASW (pH7·5), rinsed in ASW free of divalent cations for 1 h (to prevent phosphate salt precipitation), and fixed for 5 more days with 4% paraformaldehyde in phosphate-buffered saline (PBS) (440 mmol l⁻¹ NaCl, 35 mmol l⁻¹ sodium phosphate, pH7·0). After a 15 min wash in PBS, cultures were incubated for 4h in 0·2 mol l⁻¹ glycine in distilled water, and for 1 h in PBS-Triton (0·25% Triton-X). Rabbit antiserum 146II (1/200 in PBS-Tri-ton + 0·25% human serum albumin; Grimmelikhujizen, 1985) against RFamide was applied overnight (18-24 h). Cultures were then washed for 1 h in PBS-Triton. FITC-labelled goat-anti-rabbit IgG (1/200) was applied for 18–24 h. After a 10 min wash in PBS, cultures were mounted in glycerol and photographed under ultraviolet epi-illumination.

One-to five-day-old cultures were used for intracellular recording. Dissociated neurones were observed through a Nikon Diaphot inverted microscope and, in some cases, photographed with phase-contrast and/or Nomarski optics. Electrophysiology was carried out at room temperature, using 30–150 MΩ electrodes filled with 2 mol l⁻¹ potassium acetate. Grounding was through a chloridized silver wire across an agar-ASW bridge (5% in ASW) or with a Ag-AgCl pellet. The membrane potential was monitored across a bridge circuit (Getting model 5), amplified and stored on FM tape. For current-passing, the bridge was balanced in the bath. Penetrations were achieved by ringing the capacity-compensation circuit, or by very gently tapping the manipulator. Recordings could be maintained for several minutes to more than an hour.

For averages of spike parameters, one action potential per cell penetrated was measured from tape recordings. The amplitude of an action potential was measured from the beginning of the steeply rising phase to the peak. The duration was measured from the same starting point to the point where the repolarizing phase reached the resting membrane potential present 50ms before the spike. The amplitude of the afterhyperpolarization was measured from the latter point to the maximum level of the hyperpolarization. The duration of this hyperpolarization was measured from the end of the spike to where the membrane potential reached the resting level again. Rising and falling slopes were calculated from tape-recorded action potentials; the most linear part of a trace was used. The rising slopes of swimming motor neurone spikes were calculated for 1–2 ms; plateau and rapidly repolarizing slopes were calculated for 8–30 ms and 3–10 ms, respectively. Rising and falling slopes of short-duration spikes were calculated for 1–2·-4 ms. All values are expressed as mean ± standard deviation.

Isolated cells and clusters (50–400 μm in diameter) of cells settled within an hour of plating, attached to the mesogloea, and began extending processes. Epithelial cells were abundant during the first 2 days in culture and often formed monolayers of varying sizes (50μm to >2mm in diameter). After 1 day, processes bearing growth cones were seen radiating from the clusters. Over several days, many of the cells forming the monolayers and clusters floated away, leaving isolated, processbearing cells. Several neuronal types could be recognized morphologically, electrophysiologically or immunocytochemically. These included the swimming motor neurones (SMNs) and small neurones (Fig. 2).

1–5% of neurones counted in 2-to 4-day-old cultures were distinguishable by their large size (soma 30–50 μm long, with processes up to 200 μm) and clear cytoplasm. The nucleus was usually surrounded by membranous structures (Figs 3A, 4B), which appeared as concentric sheaths or as vacuoles and often extended into the processes, where they appeared as longitudinal bands or folds. Similar bands were sometimes also seen scattered along the neurites. These structures may represent the reflexive membrane infoldings observed in an ultrastructural study of the swimming motor neurones (Spencer, 1979).

Initially these large cells were fusiform or multipolar with short, wide processes, bearing lamellipodial growth cones (Fig. 3A). As cultures aged, the processes became elongated and often tapered, and growth cones became smaller. Fig. 3B shows an SMN with a long nontapered process. Neurites could be straight, apparently with no other attachment to the substratum besides their growth cones, or flat and spread at additional contact points, becoming quite irregular.

Fluorescent labelling of the dye-coupled SMN network after physiological identification, followed by dissociation, was carried out in seven preparations (Fig. 4). Results confirmed that the large cells described above were SMNs. The fluorescent cells in culture were usually large (30–50μm long), presented a clear cytoplasm and membranous inclusions around the nucleus (Fig. 4B). Since cultures were photographed only 1 or 2 days after plating, neurites were not always present and, if present, were usually short. Some small fluorescent cells (soma 10–15 μni long) were also observed, suggesting the cultured SMNs have a range of sizes.

The mean resting potential of isolated SMNs was −53 ± 20mV (N = 16), and ranged from -33 to −95 mV for different cells. Spencer (1981) measured an average in vivo resting potential for SMNs of −57 mV, and Anderson & Mackie (1977) measured a mean of −60 ± 5 mV. Approximately half the cells penetrated were excitable to depolarizing current pulses of 0·5–4·0 nA or produced spikes after being released from prolonged hyperpolarization with 1·0–4·0 nA of current (‘rebound’ or ‘anode-break stimulated’ spikes). Of these excitable neurones, 13% were spontaneously active.

Amplitudes of spontaneous and rebound spikes in isolated SMNs (50 ± 11 mV, N = 18; Fig. 5A) were lower than those recorded from neurones in vivo (Fig. 5B) using the method of Spencer (1981). For example, Anderson & Mackie (1977) and Satterlie & Spencer (1983) measured spike amplitudes of 80-100 mV for neurones in vivo. Action potential time courses in vitro were characteristic of action potentials in intact SMNs (Fig. 5A,B; Anderson & Mackie, 1977; Spencer, 1981; Satterlie & Spencer, 1983). In vitro spikes depolarized at a mean maximum rate of 20 ± 13 Vs⁻¹ (N = 15), overshot on occasion and then slowly repolarized along a plateau at a mean rate of –1·1 ±0·3 Vs⁻¹(N = 11). The final repolarizing phase was more rapid, at –4·4 ±3·3Vs⁻¹ (N=15). Mean spike duration, from the beginning of the steeply rising phase to the time when the repolarizing phase reached resting potential, was 37 ± 11ms (N = 18) and ranged from 25 to 50ms.

Satterlie & Spencer (1983) found a range of 8–50 ms for neurones in vivo. The spikes were followed by afterhyperpolarizations with a mean amplitude of −13 ± 6 mV and a mean duration of 260 ± 140 ms (N = 17).

The action potential peak and the shoulder at the end of the plateau were seen to vary in shape from sharp to rounded, particularly when depolarizing stimuli were used to elicit the spike. Action potential duration varied during a recording, and appeared to be related to the degree of hyperpolarization when the cell was excited. This was most evident in the rebound action potential trains elicited upon anode-break stimulation (Fig. 5C): spikes generated during the depolarizing return to resting potential were invariably of short duration (10–20 ms) and lacked plateaux, whereas spikes occurring later (after 20 ms) in the train showed more distinct plateaux. We did not detect any change in spike shape related to the age of cultures.

Most of the cells with neurone-like morphology were small (somata <20μm long). Cells with somatai lengths of 10–20μm could be separated into two fairly distinct morphological classes: neurones with spindle-shaped, indistinct somata (Fig. 6A), and bipolar or multipolar neurones with more distinct, convex somata and narrow, sometimes beaded, processes (Fig. 7). Cells from these two classes could be identified by electrophysiology and immunocytochemistry against RFamide, respectively. Very small cells, with somata 5–10μm long were also observed but will not be described further, as no attempt was made to record from them and they did not stain immunocytochemically.

One population of small neurones present in cultures was characterized by narrow, indistinct, spindle-shaped somata (10–20μm in length) and processes which could be tapered and straight, or flat and irregular (Fig. 6A). As for the SMNs, the cytoplasm was clear, but the nucleus was less obvious than in the SMNs. Few, if any, membranous inclusions were seen around the nucleus. However, these cells often showed single, phase-yellow spots near the bases of their processes.

These neurones had a mean resting potential of −55 ± 9 mV, ranging from −45 to −80 mV (N = 6). Injection of small (approx. 0·1 nA) depolarizing current pulses or release from strong hyperpolarization elicited spikes (Fig. 6B) which were of short duration (5·4 ± 1·4 ms; N = 7), had a mean amplitude of 25 ± 7 mV and were followed by a large afterhyperpolarization of −18 ± 7 mV which lasted 52 ± 27 ms. Both the depolarizing and the repolarizing slopes of the spikes were of similar magnitude, measuring 12 ±2 Vs⁻¹ and −11 ±2 Vs⁻¹, respectively (N = 6).

Action potentials recorded from these short-spike neurones (SSNs) closely fesembled those that can be recorded from neurones of an identified outer nerveing network, the B system (Fig. 6C), but spike amplitudes were smaller in vitro than in vivo (75–80 mV, from Spencer & Arkett, 1984). Resting potentials were lower in B neurones (mean −40 mV, from Spencer & Arkett, 1984) than in isolated SSNs. Occasionally, recordings from small cells without processes gave action potentials with the characteristics described above. Fig. 6D shows recording of a spontaneous train of spikes from one such neurone. It can be seen how similar the firing pattern is to that observed in B system neurones (Fig. 6E).

A second class of small neurones, with somatai diameters of 10-20 pm, could be characterized by their immunoreactivity using anti-RFamide serum (Fig. 7). Such cells had 1–3 narrow (approx. 1 μm in diameter), often long processes which could extend for more than 100 μm and usually showed many small (2–3 μm) varicosities. Other neurones having a very similar morphology, such as long narrow processes and small somata, did not stain for the RFamide peptide. No electrophysiological recordings were obtained from neurones with these morphologies.

Various combinations of SMNs and small neurones could be observed in cultures one to several days after plating. Neuronal complexes of anastomosing SMNs, overlapping small neurones, or of both neuronal size classes could be observed. The presence of chemical synapses in these ‘networks’ has not yet been examined electrophysiologically. However, simultaneous intracellular recording from two closely apposed SMNs revealed the presence of electrical coupling.

Clusters of cells which settled initially (see beginning of Results section), could be followed for several days. They were often seen to give rise to neuronal complexes once the overlying epithelial cells had sloughed off, suggesting the complexes were ‘preformed’ in the settling clusters. In other cases, neurones which settled far apart (3–5 soma lengths) were seen to extend processes which overlapped with those of distant neighbours and, in some cases, neurites were seen approaching nearby cells, appearing to have turned at some point in their extension.

We have described a technique for isolating different identifiable neuronal types from a cnidarian in order to study individual features of each type, such as their morphology, passive and active electrical properties, and sensitivity to putative neurotransmitters or neuromodulators. Among the different neurones isolated from the jellyfish Poly orchis penicillatus, the swimming motor neurones have proved to be the most easily identified and the most accessible for electrophysiological recording. Their distinctive morphology is similar in vitro and in vivo: membrane inclusions and clear cytoplasm were seen in an ultrastructural study by Spencer (1979), and their large size and wide processes were also visible in Lucifer Yellow-filled preparations (Spencer & Satterlie, 1980; Spencer, 1981). Fluorescent labelling before dissociation (Fig. 4) has shown that these morphological features can be used to identify SMNs reliably. Swimming motor neurones were electro-physiologically identified by their characteristic action potentials with plateaux showing similar time courses in culture (Fig. 5A) and in the intact network (Fig. 5B; see also, Anderson & Mackie, 1977; Spencer, 1981). That this spike shape occurs in isolated, cultured neurones shows that the plateau phase does not result from the temporal summation of several short-duration action potentials from adjacent neurones in the network but is due to summed membrane currents which are intrinsic to each neurone (see discussion by Spencer, 1981). In culture, these action potentials show a variability in duration (25–50 ms; Fig. 5C) which is similar, though not identical, to that observed in the semi-intact preparation (8–50ms; Anderson & Mackie, 1977; Spencer, 1981; Satterlie & Spencer, 1983). The variability in vivo may result from changes in activation states of channel proteins, with spike duration varying with the value of the resting potential when a spike is initiated (Anderson, 1979; Spencer, 1981). Such an explanation is also suggested by the occurrence, in vitro, of short-duration SMN spikes immediately following sudden release from hyperpolarization (Fig. 5C, spike no. 1). It is possible that a rapid repolarization current, presumably carried by potassium ions and having characteristics similar to I_A (Hille, 1984), is released from inactivation during the hyperpolarizing prepulse. Dissociated SMNs prepared by the present method will be used for a more detailed description of this phenomenon using voltage-clamp techniques.

The small neurones which were identified will offer us the possibility of studying synaptic interactions in culture, as they represent networks which are thought to be presynaptic to the SMNs in vivo. The cultured short-spike neurones (SSNs) generate rapid action potentials (Fig. 6B) that can appear in spontaneous bursts (Fig. 6D) and which resemble recordings from the B system neurones (Fig. 6C,E, respectively). These are the only neurones in semi-intact preparations known to produce such short-duration spikes (Spencer & Arkett, 1984). We believe this is strong evidence that SSNs are B system neurones which are presynaptic to the SMNs. This can be confirmed, in the future, by fluorescent labelling and dissociation of the B network, followed by electrophysiological recording from isolated, labelled B neurones.

RFamide-containing neurones were identified immunocytochemically in 2-day-old cultures, and may play a role in modulating SMN excitability (Spencer, 1988). Their origin in culture has recently been questioned by Alder & Schmid (1987) who have shown that free-floating myoepithelial cells isolated from the medusa Podocoryne carnea (Hydrozoa, Anthomedusae) can transdifferentiate and divide, producing FMRFamide-containing cells 4 days, at the earliest, after an intense and prolonged collagenase treatment (1400 units ml⁻¹,25–28°C for 3 h) which removes the mesogloea. This could also occur in Poly orchis (Hydrozoa, Anthomedusae). However, since the presence of a mesogloeal substratum in our experiments probably prevents this transdifferentiation (Schmid, 1978), and since the cultures we stained with anti-RFamide were too young to have transdifferentiated (2 days old; see Alder & Schmid, 1987), it is unlikely that many, if any, of the dissociated RFamide neurones arose by transdifferentiation.

RFamide-containing neurones and short-spike neurones may constitute two distinct cell populations, as suggested by their morphological differences. Unfortunately, these differences may be a result of the fixation procedure since cells with wide, tapered processes (similar to SSNs) have been seen to become cells with narrow, beaded processes (similar to RFamide neurones) when the medium was agitated or replaced with hypotonic artificial sea water. The relationship between SSNs and RFamide neurones must be determined by recording from isolated neurones to identify SSNs, followed by staining with antiserum against RFamide.

Spike amplitudes observed for SMNs (Fig. 5A,B) were lower than those observed in vivo (Anderson & Mackie, 1977; Spencer, 1981; Satterlie & Spencer, 1983), and a similar reduction was seen for SSNs (Fig. 6B,C) relative to their in vivo counterparts (Spencer & Arkett, 1984). This may have resulted from the intense and prolonged treatment with collagenase, as most preparations of this enzyme contain trace activities of trypsin (1·5 units per 1000units collagenase I, from Sigma), that may be sufficient to damage the ionic channels which carry the action potential currents. However, since the shapes of spikes in vivo and in vitro are so similar this seems unlikely. A more attractive possibility is that the ensheathing epithelial cells of in vivo preparations could either regulate the extracellular ionic environment or act as high-resistance barriers preventing shunting of any leakage currents due to damage by the microelectrode. It is not clear what effect isolation of neurones, which are normally members of an electrically coupled network, should have on spike amplitude. On the one hand, one might expect some degree of summation of action potentials in an electrically coupled population but, on the other, decreased input resistance due to coupling should reduce the amplitude of action potentials. Finally, the decreased excitability of SMNs and the reduced amplitude of action potentials in vitro may be due to the ionic composition of the bathing medium or to the presence of gentamycin. Anderson & Schwab (1984) found that motor neurones of Cyanea were difficult to excite when bathed in ASW but that when a bathing medium containing less Mg²⁺ and more K⁺ was used the neurones became far more excitable. It is possible that the ionic composition of the ASW used in our experiments is different from the extracellular fluid in vivo. Voltage-clamp recordings of peak currents from neurones dissociated with different enzymatic treatment, or from neurones in the presence or absence of apposed epithelial cells, could help clarify the relative contribution of these factors in altering action potential amplitude.

We plan to use this culture system to describe the electrophysiological properties of each neuronal type in vitro using voltage-and current-clamp and to give qualitative and quantitative descriptions of the effects of putative neurotransmitters or neuromodulators on neuronal function in vitro. We expect that this information on target cells in culture, combined with experimental data on presynaptic release in cultured neuronal complexes and in semi-intact preparations, will allow us to determine the nature of chemical transmitters in this primitive nervous system.

We thank Dr C. J. P. Grimmelikhuijzen for providing antiserum 146II; Ms Nancy McFadden for help with RFamide immunocytochemistry in culture; and Dr Larry Guilbert for a brief but fruitful discussion on neuronal labelling. This research was supported by NSERC operating grant no. A0419 to ANS, and an AHFMR Studentship to JP.