Identification, Characterisation and In Vitro Reconstruction of an Interneuronal Network of the Snail Helisoma Trivolvis

The identification of interneurones and their connections with follower cells is an important step towards understanding the neuronal basis of behaviour. With the notable exception of the Mauthner neurones of teleost fish (Korn and Faber, 1979), cellular studies of identifiable interneurones in higher animals have usually proved impractical. In the invertebrates, however, individual interneurones that are involved in behaviour patterns such as learning (Kandel, 1976; Byrne et al. 1990; Schacher et al. 1990) and rhythm generation have been identified (Jacklet, 1989; Harris-Warrick and Marder, 1991). The characterization of such interneurones and their follower cells has enabled analysis of entire neuronal circuits that control specific behaviour patterns and/or their individual components (Getting, 1989). Furthermore, very precise studies of neuronal networks in molluscs have recently been facilitated by the ability to partially reconstruct these circuits in vitro (Rayport and Schacher, 1986; Kleinfeld et al. 1990a,b; Syed et al. 1990).

Among the molluscs currently in wide use in neurobiological research are the basommatophoran pulmonates, Helisoma trivolvis, Planorbis corneus and Lymnaea stagnalis and the opisthobranchs, Aplysia californica and Tritonia diomedea. Unlike Lymnaea, Aplysia and Tritonia, however, the central ring ganglia of Helisoma and Planorbis have only recently begun to be mapped (e.g. Sonetti et al. 1982, 1988). In fact, the majority of studies on Helisoma have involved the buccal ganglia and its control of feeding behaviour (e.g. Kater, 1974; Granzow and Kater, 1977; Merickel et al. 1977; Kaneko et al. 1978; Murphy et al. 1985; Jones et al. 1987; Bulloch et al. 1988; Quinlan and Murphy, 1991). Similarly, large identified buccal neurones such as B5 and B19 are being used both in vivo and in vitro to study various aspects of regeneration, synapse formation and neuronal plasticity (Bulloch, 1989). A number of interneurones have been identified in the central ring ganglia of Lymnaea, Planorbis, Aplysia and Tritonia. Knowledge of these interneurones has been fundamental to unravelling the neuronal basis of a number of behaviours (e.g. Kandel, 1976; Getting, 1989; Syed et al. 1990, 1991; Syed and Winlow, 1991a,b). Since the central ganglionic ring of Helisoma is similar to that of Planorbis and is the mirror image of that of Lymnaea, in the present study we searched for interneurones homologous to some of those identified in these species. Using morphological, immunohistochemical, chromatographic, electrophysiological and pharmacological techniques, we have identified a central neuronal network consisting of aminergic and peptidergic interneurones. This interneuronal network acts widely on a variety of follower cells and its identification will greatly facilitate future studies of the nervous system of Helisoma.

Laboratory-bred stocks of an albino strain of Helisoma trivolvis (‘Oregon Red’) of 15–20mm vertical shell height (approximate age 4–6 months) were used unless otherwise specified. These animals were raised as described previously (Jones et al. 1987).

Central ring ganglia were dissected in normal Helisoma saline (in mmol l⁻¹: NaCl 51.3, KCl 1.7, CaCl₂ 4.1, MgCl₂ 1.5 and Hepes, 5, pH7.3) and pinned to the silicone rubber (RTV 616 GE) base of 35mm Petri dishes. Zero-Ca²⁺/high-Mg²⁺ saline contained CaCl₂ 0 mmol l⁻¹, MgCl₂ 1.5mmol l⁻¹, MgSO₄ 13.5mmol l⁻¹, NaCl and KCl being normal. For high-Ca²⁺/high-Mg²⁺ saline, the concentrations of both CaCl₂ and MgCl₂ were raised sixfold, i.e. final concentrations were CaCl₂ 24.6mmol l⁻¹, MgCl₂ 9 mmol l⁻¹.

Conventional electrophysiological techniques were used for recording intracellular neuronal activity using single-barrelled glass microelectrodes filled with a saturated solution of K₂SO₄ or 1 mol l⁻¹ KCl. Amplified signals were recorded on chart paper using a Gould four-channel chart recorder (model 2400). Intracellular Lucifer Yellow stains of the neurones were prepared according to the methods of Syed and Winlow (1989).

Central ring ganglia were removed from snails (10–12mm shell height) and pinned out flat on a small piece of silicone rubber by cutting either the cerebral commissure or one of the cerebropedal/cerebropleural connectives.

For glyoxylic acid histofluorescence of monoamines, we modified the whole-mount protocol of Audesirk (1985). Briefly, central ring ganglia were immersed in a solution containing 220mmol l⁻¹ glyoxylic acid (Sigma) plus 40mmol l⁻¹ Hepes prepared freshly in normal Helisoma saline and adjusted to pH7.0. After a 30min incubation (at 4 °C) the preparations were blotted to remove excess solution, dried under a hair dryer (cool setting), oriented on a glass coverslip and placed in a 100°C oven for 5min; they were then mounted in a drop of paraffin oil on a slide and viewed under epifluorescence illumination as described by Audesirk (1985).

For whole-mount immunohistochemistry of serotonin- and FMRFamide-containing neurones, pinned preparations were partially desheathed (Syed and Winlow, 1991a,b) and treated with 0.1% protease (Sigma, type XIV) in normal saline for 5–10min. This treatment was followed by thorough rinsing in cold normal saline. At this point, some preparations were selected for Lucifer Yellow ionophoresis to mark specific neurones (see above). The Lucifer Yellow was allowed to diffuse for 2–3h. All preparations were fixed overnight at 4°C in Zamboni’s fixative (2% paraformaldehyde in 0.1 mol l⁻¹ phosphate buffer, pH7.2, plus 15% v/v saturated picric acid). Ganglia were rinsed in 0.187 mol l⁻¹ phosphate-buffered saline (PBS), followed by 4–5 rinses in PBS plus 1% Triton X-100 and then incubated in a blocking solution consisting of 5–10% normal donkey serum, 1% Triton X-100, 0.1% bovine serum albumin and 0.02% sodium azide in PBS.

The preparations were next incubated in primary antisera (diluted 1:200 in blocking solution) for 3–4 days at 4°C. The primary antibodies used were a polyclonal antiserum generated in goat against a serotonin–BSA conjugate (INCstar Corp.) and a polyclonal antiserum generated in rabbit against a synthetic FMRFamide/succinylated thyroglobulin conjugate (generously provided by J. Bishop; see O’Donohue et al. 1984). After incubation in primary antiserum, the preparations were rinsed thoroughly over a 6–8 h period in PBS plus 1% Triton X-100, then immersed in blocking solution for 1–2h before being incubated overnight (at 4°C) in secondary antiserum. The secondary antisera used were a donkey anti-goat IgG/fluorescein isothiocyanate conjugate and a donkey antirabbit IgG/tetramethylrhodamine isothiocyanate conjugate (both from Jackson Immunoresearch). These labelled antisera were used either singly or together, each at a concentration of 1:50 in blocking solution. The preparations were rinsed thoroughly, first in PBS/1% Triton X-100 and finally in PBS alone. They were then mounted on slides in PBS/glycerol solution (85% glycerol, 15% PBS) with 1% w/v n-propyl gallate to reduce fluorophore fading (Giloh and Sedat, 1982). Preparations were viewed under epifluorescence illumination with appropriate filters for fluorescein isothiocyanate, tetramethylrhodamine and Lucifer Yellow. Since the specificity of the FMRFamide antibody for both Helisoma and Lymnaea has been tested rigorously in our previous studies (Bulloch et al. 1988; McKenney et al. 1990) we did not repeat preabsorption controls in this study. Furthermore, Richmond et al. (1991), utilizing high performance liquid chromatography (HPLC) and radioimmunoassay (RIA) techniques, have demonstrated that VD4 contains true FMRFamide along with FLRFamide and GDPFLRFamide.

Identified interneurones were isolated from their respective ganglia as described below. These freshly dissected single neurones were placed into 10 µl of 0.2mol l⁻¹ pentafluoropropionic acid (PFPA) and frozen until analysis (within 3h). Immediately before analysis, the samples were individually passed through two freeze–thaw cycles and the contents of the tube were diluted to 100 µl into an autosampler microvial. The chromatographic system consisted of a Waters HPLC pump (model 6000A), an ESA COULOCHEM guard cell at +0.75V, a Waters WISP 710B autosampler, a Beckman Ultrasphere 5µm octadecyl ion-pairing column (4.6mmX45mm) thermostatted at 50°C, a graphite/cyanoacrylate electrode and a BAS amperometric detector set for +0.65V. The detector signal was digitized and recorded using the Waters/Dynamic Solutions Maxima data acquisition system. The autosampler introduced 90 µl of sample and the separation was performed at 50°C on a Beckman Ultrasphere 5µm C-18 column (4.6mmX45mm) using a mobile phase consisting of 30mmol l⁻¹ trisodium citrate, 10mmol l⁻¹ citric acid, 1 mmol l⁻¹ ethylenediaminetetraacetic acid (EDTA), 100mmol l⁻¹ sodium perchlorate, 10mmol l⁻¹ sodium dodecylsulphate (SDS) at a flow rate of 2mlmin⁻¹. Samples were individually processed. Standardization was carried out using freshly prepared external standards in 0.2mol l⁻¹ PFPA. The detection limit was 0.1pmol of dopamine or serotonin. In pilot experiments, the level of monamines detected was variable; the data reported here were from snails maintained at 4°C overnight and dissected in cold saline, a procedure which eliminated this variability.

The current study of Helisoma neurones used the procedure for culture of Lymnaea neurones described in detail by Ridgway et al. (1991). This procedure was adapted for Helisoma by (1) preparation of medium at pH7.4 and (2) use of collagenase/dispase and trypsin at 0.67mgml⁻¹ and 1.33mgml⁻¹, respectively. Plated neurones exhibited neurite outgrowth within 12–24h; intracellular recordings from these sprouted neurones were made after 48h as described earlier. Neurotransmitters such as dopamine and FMRFamide were pressure-ejected through fire-polished glass electrodes using a Picospritzer II (General Valve, Co.).

This paper identifies a network of three Helisoma interneurones that are interconnected and that synapse upon a large number of follower cells. The locations of these neurones and their histochemical and morphological properties are described and we then demonstrate the network interactions between these interneurones. Finally, we describe the reconstruction of this network in cell culture and demonstrate that these neurones make specific connections in vitro similar to those observed in vivo.

With the exception of the paired giant serotonergic cells of the cerebral ganglia (Granzow and Kater, 1979), little is known regarding neurones of the central ring ganglia of Helisoma (see Introduction). In order to facilitate studies of Helisoma, including the present examination of interneurones, we have prepared a cell map of the cerebral, pedal, pleural, parietal and visceral ganglia (Fig. 1). This map shows (1) individually identifiable giant cells (e.g. LPeD1, RPeD1, previously designated as P1 and P5, respectively, see Haydon et al. 1984; McCobb et al. 1988) and (2) the major cell clusters (e.g. VA, VB, etc.) and/or ganglionic segments. Details of the morphological, histological and immunohistochemical properties of most of these cells are beyond the scope of the present report and will appear elsewhere (see, for example, Culver et al. 1992). However, in the following sections reference is made to a number of groups of follower cells (e.g. Tables 2 and 3).

The neurone, which we now designate as LPeD1, has the largest soma located on the dorsal surface of the left pedal ganglion, medial to the statocyst (Fig. 1). In adult Helisoma (vertical shell diameter 15–20mm), the soma of LPeD1 has a diameter of 100–150 µm. Intracellular Lucifer Yellow staining of this neurone showed it to have a single main axon that passed through the left pleural and parietal ganglia where it branched and projected to the periphery via the left internal and external parietal nerves. A fine branch was also observed to enter the visceral ganglion from where it subsequently projected to the periphery via the anal nerve (Fig. 2A). Fine axon collaterals of neurone LPeD1 were observed in the left pedal, pleural and parietal ganglia as well as the visceral ganglion (Fig. 2A).

Owing to its apparent homology with the Lymnaea giant dopamine cell (RPeD1; Cottrell, 1977; Cottrell et al. 1979; McCaman et al. 1979), the transmitter of LPeD1 was suspected to be dopamine. Preparations were therefore processed with an induced histofluorescence technique for monoamines, i.e. glyoxylic acid. Isolated central ring ganglia processed in this manner revealed many neurones with the characteristic blue and yellow-green fluorescence associated with dopamine and serotonin respectively (e.g. Audesirk, 1985). The neurone LPeD1 was amongst those with prominent blue fluorescence (Fig. 2C), whereas the giant neurone of the right pedal ganglion (RPeD1) exhibited yellowgreen (serotonin) fluorescence (not shown). For further analysis of its transmitter phenotype, neurone LPeD1 was isolated from the ganglion and processed for dopamine and serotonin content by HPLC. Individually isolated somata of neurone LPeD1 were found to contain a high level of dopamine (Table 1; Fig. 3), but serotonin was not detectable.

The follower cells of LPeD1 were identified by making simultaneous intracellular recordings from LPeD1 and a large number of identified neurones and neuronal clusters in the central ring ganglia. These recordings revealed the presence of excitatory, inhibitory and biphasic postsynaptic potentials between LPeD1 and follower neurones (Figs 4–6; Table 2). To obtain evidence for the chemical and monosynaptic nature of these connections, preparations were superfused with zero-Ca²⁺/high-Mg²⁺ and high-Ca²⁺/high-Mg²⁺ salines respectively. In all cases tested, zero-Ca²⁺/high-Mg²⁺ saline treatment blocked excitatory (Fig. 4), inhibitory (Fig. 5) and biphasic connections (Fig. 6) from LPeD1 to its follower cells (N=18). As a test for monosynaptic connections, preparations were then superfused with high-Ca²⁺/high-Mg²⁺ saline, which raises the threshold for action potentials and should reduce the likelihood of polysynaptic transmission (Berry and Pentreath, 1976). In no case tested (N=21) was the connection from LPeD1 blocked by this treatment, nor was there any obvious increase in latency such as might occur if intervening neurones were involved.

In addition to the follower cells described above, LPeD1 was also found to synapse on both RPeD1 and VD4 interneurones. The postsynaptic effect of LPeD1 on RPeD1 was inhibitory (Fig. 7A) whereas its electrical stimulation excited VD4 (Fig. 7D). These two connections were also blocked in zero-Ca²⁺/high-Mg²⁺ saline, but were not affected by high-Ca²⁺/high-Mg²⁺ saline (Fig. 7).

This is the largest cell present on the dorsal surface of the right pedal ganglion (Figs 1, 2B), with a soma diameter in the range of 80–100 µm in 15–20mm animals. Intracellular injection of Lucifer Yellow revealed this cell to have a main axon passing through the right pleural, right parietal and visceral ganglia. One branch of this axon projected into the right parietal nerve, whereas the other projected to the periphery via the anal nerve. Fine axonal processes were also observed in the right pedal, right pleural, right parietal and visceral ganglia (Fig. 2B).

Since this cell is also a probable homologue of the Lymnaea LPeD1 neurone (Kyriakides et al. 1989), whole mounts of central ring ganglia of Helisoma were processed for serotonin immunoreactivity using a polyclonal antiserum. This approach revealed many serotonin-immunoreactive somata and fibres throughout the central ring ganglia (data not shown). Amongst many serotonin-immmunoreactive somata in the pedal ganglia, readily identifiable by its size and location was RPeD1 (Fig. 2D). When isolated and processed individually for monoamine content by HPLC, the neurone RPeD1was found to contain a high level of serotonin (Table 1, Fig. 3), but no detectable dopamine.

Despite repeated and rigorous attempts, we failed to locate postsynaptic follower cells of RPeD1 within the pleural, parietal or visceral ganglia. In this context, we made simultaneous recordings from RPeD1 and neurones located in the visceral ganglion, neurone VD4 and the A–M cluster neurones (20–40 cells tested in each preparation, N=7), in the right pleural ganglion, A, B, C cluster neurones (20–25 cells tested in each preparation, N=6) and also in the right parietal ganglion, A–F cluster neurones (20–30 cells tested in each preparation, N=13). In the left pedal ganglion, however, neurone LPeD1 was found to be inhibited by RPeD1 (Fig. 8), although this connection was not found in all ganglia (observed in 7 out of 10 preparations). In all those seven preparations where this connection was demonstrated, it was found to be chemical and probably monosynaptic in nature based upon the criteria described above (Fig. 8).

A small, opalescent white cell was identified on the dorsal surface of the visceral ganglion near the right parietal ganglion. On the basis of its probable homology with neurone VD4 in Lymnaea (see Discussion), we have designated this cell as VD4 (Fig. 1). Owing to its location and size (50–70 µm in 15–20mm animals), this cell is not as easily discernible as either LPeD1 or RPeD1. When stained with Lucifer Yellow, we found this cell to be a true interneurone, with its axonal projections confined within the central ring ganglia as well as entering the right cerebrobuccal connective (Fig. 9). Neurone VD4 had two main axons which encircled the lower ganglionic ring and axon collaterals in all central ring ganglia. When processed for monoamine content by HPLC, neither dopamine nor serotonin was detected (Table 1, Fig. 3).

In order to determine the transmitter phenotype of neurone VD4, whole mounts of Helisoma central ring ganglia were processed for FMRFamide immunoreactivity using a polyclonal antiserum. A large number of neurones and their fibres, including neurone VD4, were found to be immunoreactive to this antibody. Although FMRFamide immunoreactive somata of VD4 were easily discernible, we confirmed these findings by a double staining procedure. In these preparations, neurone VD4 was first visually identified and stained with Lucifer Yellow dye and then preparations were processed for FMRFamide immunoreactivity (Fig. 9B,C). Neurone VD4, but not LPeD1 and RPeD1, was found to be immunoreactive to the FMRFamide antibody. In a recent study, utilizing a variety of techniques such as HPLC and RIA, Richmond et al. (1991) have also shown Helisoma VD4 to contain true FMRFamide as well as two other members of the FMRFamide family.

The follower cells of VD4 were found in nearly all ganglia of the central ganglionic ring. In most instances, these connections were inhibitory in nature. For example, all of the putative neurosecretory cells examined in the visceral and parietal ganglia (identified on the basis of their white coloration) were found to be inhibited by VD4 (Fig. 10A,B). Furthermore, the effects of VD4 on various pedal ganglia neurones were also inhibitory (e.g. Fig. 10C). In addition to these inhibitory connections, a few excitatory connections were also observed in the pedal, parietal and visceral ganglia (e.g. Fig. 10D) (Table 3). Furthermore, electrical stimulation of VD4 inhibited the activities of both LPeD1 and RPeD1 (Fig. 11A). To test whether the connections between VD4 and LPeD1 and RPeD1 were chemical, we first superfused our preparations with zero-Ca²⁺/high-Mg²⁺ saline. In this saline the connections between VD4 and its follower cells, e.g. LPeD1 and RPeD1, were significantly reduced in amplitude (Fig. 11B), suggesting that they are chemical in nature. To test the monosynaptic nature of these connections, preparations were superfused with high-Ca²⁺/high-Mg²⁺ saline. All the connections between VD4 and its follower cells, including LPeD1 and RPeD1, were found to persist under these experimental manipulations (Table 3; Fig. 11C).

Since our immunohistochemical data suggested that neurone VD4 contained an FMRFamide-like peptide, we tested the possibility that FMRFamide and some of its analogues were active. Simultaneous pressure-ejection of FMRFamide (range 10⁻⁵–10⁻⁷ mol l⁻¹) on follower cells LPeD1 and RPeD1 mimicked the effects of electrical stimulation of VD4 (Fig. 11D). We also tested analogues of FMRFamide such as FLRFamide, SDPFLRFamide and GDPFLRFamide. When bath-applied to LPeD1 or RPeD1, all of these peptides were found to have inhibitory effects similar to either VD4 stimulation or exogenous application of FMRFamide (data not shown).

Since it is often difficult in vivo to demonstrate unequivocally that connections between neurones are truly monosynaptic and that the effects of various exogenously applied substances are direct, we turned to in vitro cell culture. To test the monosynaptic and specific nature of connections between LPeD1, RPeD1 and VD4, we isolated these neurones from their respective ganglia. When these neurones were plated on poly-L-lysine-coated dishes containing conditioned medium they exhibited neurite outgrowth and also synapse formation (Fig. 12).

We first cultured only one pair of interneurones per dish and tested the ability of the sprouted interneurones to re-establish specific synaptic connections. When simultaneous intracellular recordings were made, these neurones were often found to have established appropriate synaptic connections. For example, a mutually inhibitory chemical connection developed between LPeD1 and RPeD1 in 76% of preparations (N=17) (Fig. 12B,C). In the other 24% of preparations, however, we detected the presence of electrical coupling between these neurones, no chemical connection being observed (Fig. 12D).

When LPeD1 and VD4 were cultured as a pair, LPeD1 developed an appropriate excitatory connection with VD4 (N=7) (Fig. 12E). Similarly, when VD4 was co-cultured with either LPeD1, RPeD1 or both, it re-established its specific inhibitory connections with these cells (N=11) (Fig. 12F). In these same preparations, however, no connections were observed between RPeD1 and VD4, a result which parallels the situation in vivo. Although the connections made in vitro from VD4 to neurones LPeD1 and RPeD1 were of the same type as those observed in vivo, it was difficult to resolve the occurrence of unitary inhibitory postsynaptic potentials (IPSPs). In contrast, unitary PSPs were often visible in connections made by LPeD1 and RPeD1 onto their follower cells (see Fig. 12B, C and E).

The fundamental plan of the molluscan nervous system has remained surprisingly stable, allowing both groups and individual homologous neurones to be identified between ganglia in different classes (Dorsett, 1986; Croll, 1987). This is exemplified by the present study of Helisoma where we took advantage of previously identified interneurones in Lymnaea and Planorbis and, despite different ganglionic arrangements, were able to identify three homologous interneurones, namely left pedal dorsal 1, right pedal dorsal 1 and visceral dorsal 4.

In the present study we have proposed a standardized system to map the Helisoma central ring ganglia. This system is based on that proposed by Benjamin and Winlow (1981) for Lymnaea which has been successfully used and extended by others. The central ring ganglia of Helisoma are the mirror images of those of Lymnaea, so the Lymnaea nomenclature system can readily be adapted for this species (Culver et al. 1992; Richmond et al. 1991). Since a large number of laboratories use Helisoma, we believe a nomenclature system is required and, in addition, a common system will be a tremendous asset for comparative studies with Lymnaea.

Berry and Cottrell (1973, 1975) described a multi-action neurone (the giant dopamine cell, GDC) in the sinistral basommatophoran pulmonate Planorbis corneus. The cell body of this neurone is located in the left pedal ganglion and was known to contain dopamine (Powell and Cottrell, 1974). In a closely related species, Lymnaea stagnalis (L.), a cell similar to the GDC of Planorbis was identified in the right pedal ganglion (McCaman et al. 1979; Cottrell et al. 1979) and was designated as RPeD1 (Benjamin and Winlow, 1981). Since Lymnaea is a dextral pulmonate and its central ganglionic ring is a mirror image of that of Planorbis, the giant dopamine-containing cells of the two animals are thought to be homologous. These dopaminergic neurones from both Planorbis (Berry and Cottrell, 1973, 1975, 1979; Cottrell, 1977) and Lymnaea (Winlow et al. 1981) are monosynaptically connected to a large number of follower cells in the visceral and right parietal ganglia. Although the exact function of the GDC in Planorbis is unclear, in Lymnaea the giant dopaminergic neurone RPeD1 is known to be involved in respiratory behaviour (Syed et al. 1990; Syed and Winlow, 1991a). Earlier, a Helisoma neurone was designated as P1 (pedal 1), based primarily on its location and was thought to be the homologue of both the Planorbis GDC and Lymnaea RPeD1 (McCobb et al. 1988). It was also reported that the ‘left pedal giant cell’ of Helisoma contained 0.80pmol per cell of dopamine (McCaman et al. 1979), an estimate remarkably close to that of this study (Table 1). The morphology and synaptic contacts of this neurone, however, were previously unknown. We have recently begun to map the nervous system of Helisoma and have developed a nomenclature system which is compatible with the system in common use for the Lymnaea central nervous system. Based upon this naming system for identifiable neurones located on the dorsal surface of the left pedal ganglia, we have designated this cell as Helisoma LPeD1. We confirmed using the glyoxylic acid histofluorescence technique and HPLC that this cell contains dopamine and has a similar morphology to that of both the GDC and RPeD1 of Planorbis and Lymnaea, respectively. Electrophysiologically, LPeD1 of Helisoma was also found to make synaptic contacts with a number of cells located in both the visceral and left parietal ganglia which appear to be chemical and monosynaptic. Given the high content of dopamine in the Helisoma LPeD1 and the pharmacological characterisation of the synapses made by its homologues in Planorbis and Lymnaea, dopamine is clearly a strong candidate as the transmitter at LPeD1 synapses in Helisoma. The range of pharmacological studies required to establish the identity of the transmitter and receptor subtypes at each of the LPeD1–follower cell connections, however, was beyond the scope of the present study.

It has recently been shown that, in culture, Helisoma LPeD1 re-established specific synaptic connections with appropriate follower cells, but not with non-follower cells (Syed et al. 1992). Furthermore, in vitro, the giant dopamine cells of Helisoma and Lymnaea make appropriate synapses upon each other’s follower cells (Syed et al. 1992).

These experiments provide further support for strong homology between the giant dopamine neurones of these different species.

A giant neurone present on the dorsal surface of the left pedal ganglion of Lymnaea and right pedal ganglion of Planorbis is known to contain serotonin (Audesirk, 1985; Cottrell et al. 1979; Kyriakides et al. 1989). In Lymnaea, this serotonergic neurone has been designated as LPeD1, but it has not been found to make synaptic connections with any other neurone (Kyriakides et al. 1989). Despite some morphological and electrophysiological information, the function of Lymnaea LPeD1 is unknown. A Helisoma homologue of Lymnaea LPeD1 was apparently used in some cell culture experiments (Haydon et al. 1984) where it was referred to as P5. We found this cell to have similar morphology to that of Lymnaea LPeD1 (Kyriakides et al. 1989) and, like that of Lymnaea, this cell also contains serotonin as assayed by glyoxylic acid histo-fluorescence, immunohistochemistry and HPLC (Table 1). In accordance with our new nomenclature system for Helisoma, we now designate this cell as Helisoma RPeD1. Although we found this cell to have reciprocal inhibitory connections with LPeD1 in some preparations, the functional significance of RPeD1 is at present unknown. The mutual inhibitory connections between Lymnaea LPeD1 and RPeD1 have never been reported in normal preparations. However, following axotomy and regeneration, Benjamin and Allison (1985) found a novel excitatory synapse between RPeD1 and LPeD1.

Helisoma RPeD1, in common with the giant dopamine cell of both Lymnaea and Helisoma, has axons that project to the periphery. Since these neurones have been found to innervate a number of identified neurones, both LPeD1 and RPeD1 are designated as interneurones in accord with common usage.

A true interneurone located in the visceral ganglion of Lymnaea, with axonal branches confined within the central ring ganglia, was originally identified as the visceral white interneurone (V.W.I.) (Benjamin, 1984). This cell is peptidergic and synapses on many cells in all the central ring ganglia (Benjamin, 1984). A neurone morphologically and electrophysiologically similar to V.W.I. was later described by Janse et al. (1985), who named it visceral dorsal 4 (VD4). Extensive electrophysiological and morphological investigations have revealed that V.W.I. and VD4 are synonymous (Syed, 1988). A homologue of the Lymnaea VD4 has not yet been identified in Planorbis but, in the present study, we appear to have found such a cell in Helisoma. In order to be consistent with previously existing maps of Lymnaea neurones (Winlow and Benjamin, 1976), we have adopted the nomenclature proposed by Janse et al. (1985) (Syed, 1988; Syed and Winlow, 1989, 1991a,b; Syed et al. 1990). Interneurone VD4 of both Lymnaea and Helisoma are immunoreactive for FMRFamide and have very similar morphologies. Our recent studies have indicated that neurone VD4 of Lymnaea does indeed use one of the FMRFamide family of peptides as a transmitter, but also uses a second, as yet unidentified, transmitter (McKenney et al. 1990). From the results of the present study of

Helisoma, an FMRFamide peptide (which could be either a tetrapeptide and/or a heptapeptide, see Richmond et al. 1991) is a strong candidate as the transmitter used at VD4 to RPeD1 and LPeD1 connections (Fig. 12D); if a second transmitter exists in Helisoma VD4, it is unlikely to be dopamine or serotonin (Table 1). However, the apparent absence of unitary EPSPs from VD4 to both LPeD1 and RPeD1 and other follower cells, as well as the long-lasting action of VD4 at some connections, may be due to FMRFamide acting as a local modulator rather than via a classical synapse. The interneurones in both species have a number of follower cells in almost all the central ring ganglia. However, the intrinsic properties of these homologous neurones differ slightly. For example, the Lymnaea VD4 typically fires bursts of spikes (Syed, 1988; Syed et al. 1990) and less commonly single spikes, whereas the Helisoma VD4 always fired single spikes (data not shown).

In addition to being a useful tool for determining the intrinsic properties of identified neurones (e.g. endogenous or conditional bursters) and the monosynaptic nature of their connections, in vitro preparations have also proved valuable in the study of synaptic specificity. The mechanisms that determine synaptic specificity, however, are as yet largely obscure. One approach in this area has been to examine synapse formation between identified pre- and postsynaptic neurones cultured in vitro. Several invertebrate preparations, such as the leech (see Chiquet and Nicholls, 1987; Nicholls et al. 1990) and the molluscs Aplysia (Schacher and Proshansky, 1983; Schacher et al. 1985; Schacher, 1988; Kleinfeld et al. 1990a,b), Helisoma (Haydon, 1988; Cohan et al. 1990) and Lymnaea (Syed et al. 1990), have helped our understanding of specific synapse formation in vitro. In almost all of these preparations, neurones have been shown to maintain their intrinsic properties. The literature is divided, however, regarding the specificity of synapses in culture. For instance, although leech identified neurones make specific connections with appropriate targets similar to those seen in vivo, some novel connections are also formed (Fuchs et al. 1981). Novel connections have also been observed between Aplysia neurones (Camardo et al. 1983; Schacher et al. 1985; Kleinfeld et al. 1990a,b). While the formation of novel connections is interesting and worthy of further study, it does cause difficulties when one is examining the specificity of synapses. In our hands, the synaptic connections formed between specific Lymnaea neurones in vitro were always appropriate; no inappropriate connections were observed (Syed et al. 1990). It is possible that the novel connections observed by others in vitro may have existed during early development and were later retracted. Therefore, the presence of novel connections may be transient when examined over a long time course, as has been demonstrated for in vivo novel connections of Helisoma (Cohan et al. 1987). These differences might also be attributed to the differences in tissue culture procedures. For instance, when leech or Aplysia neurones are cultured on concanavalin A, the chemical connections usually observed on poly-L-lysine coated dishes are no longer discernible; instead, electrical connections predominate (Lin and Levitan, 1987). Interestingly, substratum conditions are also known to affect the distribution of calcium channels in cultured leech neurones (Ross et al. 1988). However, the occurrence of electrical rather than chemical synaptic connections between LPeD1 and RPeD1 in 24% of in vitro preparations is clearly not attributable to different substratum conditions. Although it is possible that the novel electrical synapses observed between LPeD1 and RPeD1 in 24% of in vitro preparations might be transient connections that precede chemical synapses, we did not test this possibility rigorously. In these preparations we obtained electrophysiological data 48 h after plating and did not observe inappropriate or novel electrotonic synapses except between LPeD1 and RPeD1. Our data at present, therefore, cannot explain the reasons for the occurrence of electrotonic coupling between LPeD1 and RPeD1. It is, however, possible that these neurones may have different genetic programmes for synapse formation, having a higher degree of plasticity in this regard than the other neurones.

The identification of individual neurones and characterization of their synaptic connections is important if we are to understand fully the cellular basis of specific behaviour patterns. In Lymnaea, RPeD1, VD4 and the input 3 interneurone (IP3I; as yet unidentified in Helisoma) are known to be central pattern-generating interneurones that initiate and control respiratory behaviour (Syed et al. 1990; Syed and Winlow, 1991a,b). In addition to the behavioural significance of these Lymnaea neurones, their connectivity patterns are well characterized both in vivo (Syed and Winlow, 1991a) and in vitro (Syed et al. 1990). In contrast, the function of the serotonergic cell (LPeD1 in Lymnaea, RPeD1 in Helisoma) is unknown. Despite apparent locational and morphological similarities between Lymnaea neurones RPeD1 and VD4, with Helisoma neurones LPeD1 and VD4, respectively, the behavioural function(s) of these Helisoma interneurones has yet to be determined.

The interneurones identified in this study and their known follower cells are shown schematically in Fig. 13. The physiological significance of the various neurones and groups of follower neurones in Helisoma has yet to be determined. However, in Lymnaea the function of many apparently homologous neurones is known. For example, most visceral and parietal neurones are known to be involved in cardiorespiratory control (Buckett et al. 1990; Syed and Winlow, 1991a,b; Syed et al. 1991). Most pedal neurones, in contrast, control locomotion (Syed and Winlow, 1989). From these studies it seems reasonable to speculate that LPeD1 and VD4 of Helisoma may be part of the circuit that controls cardiorespiratory behaviour. Additionally, neurone VD4 is a multi-action interneurone that may also be involved in controlling the activity of pedal locomotor neurones. The present detailed morphological and electrophysiological characterisation of these interneurones should facilitate future studies of the neural basis of behaviour in Helisoma. Furthermore, our new in vitro model will provide opportunities to study synapse formation and plasticity in a manner unapproachable in the intact ganglia.