Formation of Specific Connections in Culture by Identified Leech Neurones Containing Serotonin, Acetylcholine and Peptide Transmitters

Identified neurones of invertebrates, dissected one by one out of the central nervous system and maintained in culture, have proved valuable for approaching problems concerned with synapse formation and specificity (Fuchs, Henderson & Nicholls, 1982; Dagan & Levitan, 1981; Hadley, Kater & Cohan, 1983; Schacher, Rayport & Ambron, 1985). For example, the giant Retzius cells of the leech which are known to contain and secrete serotonin (5-HT) continue to do so in culture (Henderson, 1983). Moreover, they make electrical synapses with each other and make chemical synapses on the P sensory neurone (Fuchs et al. 1982; Dietzel, Drapeau & Nicholls, 1986). In these examples the pattern of connections made by each type of cell is highly specific. Thus, the same P cells which do not become electrically coupled to Retzius cells in culture do become coupled to a different type of cell, the L motoneurone (Fuchs, Nicholls & Ready, 1981). Specificity of synaptic connections in culture, however, is not identical to that in the ganglion: L and Retzius cells become electrically coupled in culture, a connection that does not occur in the ganglion during development (Fuchs et al. 1982).

In the present series of experiments we have analysed a further set of electrical and chemical connections established in culture, with particular emphasis on certain neurones that have been shown to contain a peptide transmitter. These are the heart excitor motoneurone (HE) and the anterior pagoda (AP), a large cell of unknown function. Both cells are cholinergic (Wallace, 1981) and both contain a peptide transmitter, resembling in its immunoreactivity the cardioactive tetrapeptide known as FMRFamide (phenylalanyl-methionyl-arginyl-phenylalanyl-amide; Kuhlman, Li & Calabrese, 1985a, b ). In addition Kuhlman et al. (1985b) have provided evidence that the FMRFamide-like peptide is used by leech neurones as a transmitter modulating the heart rhythm.

It will be shown that specific electrical connections are established by peptide-containing cells with regularity after 2—6 days in culture; chemical synaptic inter-actions can also be demonstrated, but only occasionally. Accordingly we have used the presence or absence of electrical synapses between neurones as a convenient and reliable assay for specificity. The principal problems we have approached are: (1) What is the pattern of electrical connections established by the HE and AP cells, containing the same two transmitters? (2) Do they become connected to one another, to the Retzius cell containing 5-HT and to the P cell, containing an unknown transmitter? (3) Do certain cells, such as the P cell, preferentially make rectifying electrical junctions? (4) Is peptide located in vesicles at characteristic release sites? (5) What chemical synaptic connections develop with peptide-containing cells as presynaptic or postsynaptic elements?

Techniques for removing individual neurones from ganglia, maintaining them in culture and recording electrically were as described elsewhere (Dietzel et al. 1986). In brief, cells such as the HE, AP, Retzius and P were identified by their shapes, sizes and positions and in addition by their distinctive electrical properties within the ganglion and in culture. Cells were removed one by one from desheathed ganglia that had been treated with collagenase-dispase (2 mg ml^-1, Boehringer Mannheim Corporation). After isolation, cells were washed in Leibowitz 15 medium containing 2% foetal calf serum with gentamycin (0·1mgml^-1, Schering Corporation) and glucose (30 mmol 1⁻¹). Pairs of cells were plated in microwells coated with polylysine or concanavalin A (ConA). Cells on ConA sprouted more extensively than those on polylysine (Chiquet & Acklin, 1986). Electrical recordings were made with micro-electrodes filled with potassium acetate.

Synthetic peptides were obtained from Bachem AG (Bubendorf, Switzerland). 4 mg FMRFamide [Phe-Met-Arg-Phe-NH₂(2-trifluoroacetate)] was dissolved together with 4 mg bovine serum albumin (BSA; Fluka, Buchs, Switzerland) in 0·5 ml distilled water. To this solution, 100 mg l-ethyl-3(3-dimethylaminopropyl)-carbodiimide-HCl (Sigma) dissolved in 0·25 ml water was added (Jennes & Stumpf, 1983). After incubation for 60 min at room temperature on a shaker, the mixture was dialysed against distilled water in the cold, and subsequently lyophilized.

Rabbits were immunized subcutaneously with 250 μg FMRFamide-BSA conjugate dissolved in 0·75 ml water and emulsified with 0·75 ml complete Freund’s adjuvant (Gibco). Boosts with the same amount of peptide conjugate in incomplete adjuvant were made 1, 2 and 3 months later. Rabbits were bled 1 week after every boost. Antiserum titres were determined by an enzyme-linked immuno-adsorbent assay (ELISA; Trivers, Harris, Rougeot & Dray, 1983) in which peptide (10μgml^-1 in phosphate-buffered saline, PBS) was bound to microtitre wells (Falcon dishes, 3911), which were then incubated with serial dilutions of antisera (in 10% horse serum in PBS). Dilutions giving half-maximal response against FMRFamide in this assay were 1:200 to 1:400 after the first boost and 1:1600 to 1:3200 after the third boost. No reaction with 1:100 diluted antisera was found against substance P, somatostatin and met-enkephalin (for further controls, see below).

Leech ganglia were stained with anti-FMRFamide antisera as described by Kuhlman et al. (1985a,b). Briefly, isolated ganglia were fixed overnight in the cold with 4 % paraformaldehyde in 0·12 mol I^-1 sodium phosphate buffer (pH 7·4). After opening the connective tissue capsules with forceps, the ganglia were incubated with 0·4% Triton X-100 (Merck, Darmstadt) in 0·1 moll^-1 sodium phosphate buffer (pH 7-4) for 2h, followed by incubation in the same solution containing 10 % horse serum. The tissue was stained overnight in the cold with anti-FMRFamide anti-serum diluted 1:200 to 1:800 in serum/Triton/phosphate buffer and, after several washes, with rhodamine-labelled goat anti-rabbit IgG (Cappel; diluted 1:50) for 2h at room temperature. Ganglia were mounted in glycerol and viewed with epi-fluorescence optics.

In each segmental ganglion, a characteristic set of about 20 neurones was labelled by the anti-FMRFamide antisera. A commercial antiserum against FMRFamide (Cambridge Research Biochemicals, UK) yielded a similar staining pattern. Pre-incubation of the diluted (1:200) antisera with synthetic FMRFamide or Tyr-FMRFamide (100μgml^-1; overnight in the cold) completely abolished their ability to stain neurones in leech ganglia. In contrast, pre-incubation with substance P, somatostatin, met-enkephalin, met-enkephalin-Arg-Phe (which contains the free acid form of the FMRF sequence at the C-terminus) and gastrin tetrapeptide (Trp-Met-Asp-Phe-NH₂) did not neutralize the anti-FMRFamide antisera. 100 μg ml^-1 porcine neuropeptide Y (whose C-terminus is -Arg-Tyr-NH₂) decreased, but did not abolish the specific staining of leech neurones by our anti-FMRFamide antisera. This suggests that our antisera are specific for the amidated C-terminus of FMRFamide and thus comparable to antisera produced in other laboratories (Kuhlman et al. 1985a,b, Dockray et al. 1983).

Isolated neurones and heart muscle fibres were fixed and embedded in micro-wells. Each solution was fully exchanged around the cells by gentle pipetting. Neurones were fixed for 5 min in 0·6% glutaraldehyde and 0·4% paraformaldehyde in 0·08moll^-1 cacodylate buffer, pH7-4, 342mosmoll^-1, post-fixed in 1 % osmium tetroxide in 0·1moll^-1 cacodylate buffer for 5 min, and serially dehydrated in ethanol. Embedding was in Epon which was poured into the microwells, substituted three times and allowed to infiltrate the cells for at least 4 h before polymerizing in an oven at 75°C for 4h. From microwells coated with polylysine or concanavalin A it was possible to separate the cured Epon containing the cultured cells by inserting the edge of a razor blade between the Epon and the well. The Epon plug was mounted on a dowel with the cells at the flat surface on top. In this manner the first sections were cut through the interface between the cells and the bottom of the microwell. Sections were counterstained with 50 % ethanolic uranyl acetate for 6 min and lead citrate for 2·5 min.

Labelling of FMRFamide-like peptide in the HE and AP cells required modification of the normal fixation protocol (see above). Standard fixation methods which maintain good morphology of the cultured neurones prevented labelling by the FMRFamide antiserum. Successful labelling was achieved only after reducing the glutaraldehyde concentration from 0·6% to 0·4% and by increasing the paraformal-dehyde concentration from 0·4% to 2·5%. It was also necessary to reduce the concentration of the postfixation osmium tetroxide from 1 % to 0·05 %, resulting in reduced contrast of the tissue as well as poorer preservation.

Ultrathin sections were mounted on parloidion-coated nickel slot grids. The sections had to be further treated according to the method of Trapp et al. (1981) to expose the antigenic sites and to achieve any antiserum staining. The sections were treated as follows: grids were (1) placed on a drop of sodium ethoxide (saturated solution diluted 1:3 in 30 % ethanol) for 3 min to remove some of the embedding Epon, (2) rinsed in Tris-HCl (0·5 moll^-1, pH 7·6), (3) placed on a drop of hydrogen peroxide, 0·2% for 3 min, (4) rinsed in Tris-HCl, (5) incubated on a drop of FMRFamide antiserum with PBS (diluted 1:100) for 18 h at 4°C, (6) rinsed in Tris-HCl, (7) incubated on a drop of protein-A-coated colloidal gold, 8nm in diameter (Roth, Bendayan & Orci, 1978) (optical density at 525 nm, 0·073), in PBS for 4h at room temperature, and (8) rinsed in water and counterstained normally.

In confirmation of results obtained by Kuhlman et al. (1985a, b), antiserum prepared against purified FMRFamide labelled a small group of neurones in segmental ganglia. Most intensely fluorescent were the HE cells. The annulus erector (AE) motoneurones were also labelled, one on each side, as were about 15 small, unidentified cells in each ganglion (Fig. 1B). Axons of these neurones could be seen in connectives and roots. The identification of neurones was made (a) by visual inspection and (b) in several experiments by labelling the cell first with the fluorescent dye Lucifer Yellow injected from the microelectrode and then with antiserum (Fig. 1A). In certain cells, notably the AP cells (of unknown function), staining with antiserum was weak or absent in freshly dissected ganglia (see Kuhlman et al. 1985a). However, when roots and connectives were cut in the animal, or when ganglia were maintained in culture for 3 days the FMRFamide-like reactivity increased, as assessed by staining. Tests made on cells at various times after isolation showed that the ability to react with antiserum persisted for 8 days or longer.

The populations of dense-core and clear vesicles in peptide-containing HE and AP cells were different to those in Retzius cells. Dense-core vesicles consisted of a single population with concentric dark centres in both AP cells (Fig. 3A) and HE cells. The mean diameter of these vesicles was 91 nm (range 61-131 nm). By contrast, the Retzius cells (Henderson, Kuffler, Nicholls & Zhang, 1983) contain two classes of dense-core vesicles; large eccentric, diameter 111 nm, and small concentric, diameter 71 nm. An additional distinction between the dense-core vesicles of these cell types was the darkness of the core which is more intensely stained in the Retzius cell than in the AP or HE cells.

In the Retzius cells an additional population of clear, round vesicles with a mean diameter of 38nm has been described (Kuffler, Nicholls & Drapeau, 1986). At presumed transmitter release sites in the Retzius cell a cluster of these clear vesicles is apposed to presynaptic densities, surrounded by a cap of large and small dense-core vesicles (Kuffler et al. 1986). Like Retzius cells, the HE and AP cells also contained clear, round vesicles (Fig. 3A) with a mean diameter of 42nm (range 35-61 nm). The HE cells were found to have a second population of clear vesicles with a mean diameter of 26 nm (range 22-28 nm). These smaller vesicles were consistently found in clusters at the cell membrane. Both the dense-core and large clear vesicles of the HE and AP cells were found distributed throughout the cell bodies and in clusters near cell membranes of the cell body and processes (Fig. 3A).

Antiserum to FMRFamide was used to localize the putative transmitter in the HE and AP cells. After fixing and embedding the cells for electron microscopy, the thin sections were exposed to the FMRFamide antiserum and the location of bound antibodies was marked by tagging them with colloidal gold coated with protein A. The antibodies to FMRFamide were found in the dense-core vesicles in both the HE (Fig. 3E) and AP cells. Antibody against 5-HT, which labels dense-core vesicles in Retzius cells, did not label the vesicles in the HE or AP cells. Owing to difficulties in maintaining the antigenic sites recognized by the anti-FMRFamide antibodies when preparing the cells for electron microscopy, modifications of the fixation procedures were necessary; as a result, the preservation of morphology is not optimal. Never-theless, more than 70% of the dense-core vesicles in anti-FMRFamide antiserum-labelled HE cells were labelled by at least two colloidal gold particles: many vesicles were found with 20 associated gold particles. Anti-FMRFamide antiserum did not label the clear vesicle populations. This result, on its own, does not rule out the presence of peptide that might have been lost during fixation.

Physiological evidence for chemical synaptic transmission by HE and AP cells acting as presynaptic neurones was obtained in only a few pairs of cells (see below). This made it difficult to search for release sites by electron microscopy. Neverthe-less, in the few pairs of cells where the presynaptic HE cell had been placed in close apposition to a heart muscle fibre (Fig. 3C) and chemical synaptic transmission was observed physiologically, vesicle clusters of the type shown in Fig. 3D were observed. Dense-core vesicles in these clusters reacted with FMRFamide antiserum. As shown below (Fig. 9) impulses in the HE cell evoked synaptic potentials and contractions in the muscle fibre. The clustering of vesicles close to the cell membrane is reminiscent of the appearance seen in a P cell that made chemical synapses with an AP cell (Fig. 3B) and in boutons en passant of Retzius cells. Also clusters of the smaller population of clear vesicles (26 nm in diameter) were seen in the regions of the HE cell close to the heart muscle fibres without surrounding dense-core vesicles.

Both AP cells and HE cells in culture maintained their membrane properties, giving small characteristic action potentials. These were often larger than those seen in situ (Fuchs et al. 1981). HE and AP cells made electrical connections with a number of target cells. Such electrical connections developed reliably in 2-6 days between healthy pairs of cells cultured on polylysine or concanavalin A. The strength of coupling varied but the pattern of connections did not. Thus, HE cells became connected to other HE cells, to AP cells and to Retzius cells by non-rectifying junctions; AP cells showed a similar pattern of connections (Fig. 4). In each case, depolarizing or hyperpolarizing current injected into one cell (upper traces, low gain) spread to the other cell (lower traces, high gain). Routine exploration by electron microscopy has, so far, not revealed characteristic gap junctions between HE, AP and Retzius cells in culture.

AP cells also became electrically coupled to P cells. However, as shown in Fig. 5, these junctions rectified: depolarization spread more readily from P cells to AP cells, hyperpolarization more readily from AP cells to P cells. Rectifying junctions with similar properties developed between P and HE cells (not shown). So far no non-rectifying junctions made by P cells have been observed in leech ganglia or in culture.

As observed before (Fuchset al. 1981, 1982; Dietzeleia/. 1986), Retzius and P cells never became electrically coupled.

These results provide examples showing that cells in culture can make specific connections with one another and that certain targets are excluded. Under the conditions of culture, two cells that may never be in contact in the ganglion can be brought into close apposition, allowing novel, ‘abnormal’ connections to appear. Thus, AP and HE and Retzius cells are not demonstrably coupled in the ganglion.

Chemically mediated synaptic interactions developed between pairs of cells plated close together in culture. The HE and AP cells (containing peptide as well as acetylcholine, ACh) could be presynaptic to Retzius and P cells, producing IPSPs. Conversely, P cell impulses could evoke chemically mediated EPSPs in AP cells (see below). Detailed study of these synaptic interactions was precluded by features such as the low frequency with which they could be observed. The strong electrical coupling between cells often made it impossible to tell whether a chemical component was also present. Moreover, in those pairs of cells in which chemically mediated synaptic potentials could be unequivocally observed, they were often small and fatigued after a few trials (four or five). In total, we have only observed 12 unambiguous chemical interactions out of hundreds of trials. The possible reasons for these failures are discussed below.

Examples of chemical synaptic interactions are shown in Figs 6, 7 and 8 for pairs of cells in which there was no electrical coupling. The hyperpolarizing synaptic potentials in Retzius or P cells evoked by HE or AP stimulation arose after a delay and had a prolonged time course. From the sign of the potential, the delay and the evanescent quality, as well as the absence of electrical coupling, these potentials were clearly chemical in origin. A particularly obvious postsynaptic potential is shown in Fig. 9, in which the presynaptic cell was an HE neurone that had sprouted and made contact with an isolated heart muscle fibre in culture. This is the same pair of cells shown in the electron micrograph of Fig. 3D in which a close apposition of the HE cell and a heart muscle fibre can be seen.

The strongest chemical connection we observed was from P to AP cells and took the form of excitatory potentials (Fig. 8). Even in the presence of electrical coupling such potentials were obvious. Hence, the P cell which is never presynaptic to Retzius cells (in literally hundreds of trials, see Fuchs et al. 1982; Dietzel et al. 1986) is able to give rise to chemically mediated excitatory synaptic potentials in a different neurone, the AP.

The results obtained in these experiments extend and confirm earlier observations made with cultured leech neurones. In particular it is clear that the formation of connections by pairs of cells in culture is not random but specific (Table 1). Certain types of cells become connected by electrical synapses while others do not. For example, the P cell never makes electrical connections with the Retzius cell, but does become electrically coupled to L, to AP and to HE cells by rectifying junctions (see also Fuchs et al. 1982). Other cells not included in Table 1 also show specificity: N cells form chemical synapses with L cells, with a weak, rectifying electrical component; L cells form non-rectifying junctions with each other (see Fuchs et al. 1981). In some instances the chemical and electrical connections generated in culture resemble those seen in ganglia. That novel, unusual connections can also develop in culture seems hardly surprising for the following reason: in culture two cells are placed in close apposition and need only little outgrowth to produce connections. Cell—cell recognition is presumably the principal factor involved in the decision as to whether or not a synapse should be formed. By contrast, connections in the ganglion are formed as the result of a complex sequence of steps occurring during development. Neurones send out processes which reach their appropriate destinations and there stop growing and form connections with appropriate targets. Not only must growth be in the correct direction but the target cell must at that time be able to receive an input. In addition, factors such as competition and cell death may play a role. Cell-cell recognition, while important, is only the last step. Whatever the molecular mechanisms involved in the ultimate selection, the opportunity for it to be expressed has to occur if a synapse is to be formed during development. There may also be a change of specificity as a result of a change in environment during transfer from ganglion to culture.

Chemically mediated synaptic interactions resembling those seen in the ganglion were also observed in culture. In culture, P cells make excitatory connections with AP cells and HE motoneurones chemically excite cardiac muscle fibres. At the same time, chemical synapses that develop strongly in the ganglion may fail to appear in culture. Thus, P cells have never been observed to make chemical synapses with Retzius cells in culture, a connection that is strong within the ganglion. A ‘failure’ to observe physiological signs of a chemical synaptic interaction does not preclude the possibility of morphological specializations with release of insufficient numbers of quanta for electrophysiological detection. Fine structural studies have, in general, confirmed our physiological results, in that anatomical correlates can be found for pairs of cells where chemical transmission can be definitely established (Kuffler et al. 1986).

The weak and evanescent nature of the chemical synaptic potentials observed in these studies has unfortunately prevented critical analyses of the underlying mechanisms. HE and AP cells contain both ACh and FMRFamide-like peptide (Kuhlman et al. 1985a, b). At present, we do not know which of the two transmitters is involved in producing effects observed in culture. To test effects of peptides and antagonists to ACh and to measure conductance changes in the postsynaptic membrane, more stable recordings of synaptic potentials are required. It is tempting to speculate that the potentials that arose after long delays and persisted for long periods involved second messengers and were mediated by peptides, but the possibility that they arose from actions of ACh cannot be ruled out. In addition, it is not certain from antisera studies whether the peptide in dense-core vesicles is in fact FMRFamide, and, indeed, the peptide within leech neurones has not yet been sequenced. Studies at other invertebrate synapses have shown that antisera generated against FMRFamide can bind to other small peptides with identical or similar C-terminal structure (Dockray et al. 1983).

A serious question that remains concerns why the chemical interactions we observed were so weak and so infrequent. One probable reason is that the culture conditions are not optimal and that particular factors may be missing. For example, whether or not sprouting by leech neurones occurs in culture depends critically upon the nature of the substrate : neurones grow rapidly and extensively on the plant lectin concanavalin A and on connective tissue capsule (Chiquet & Acklin, 1986). The same neurones grow only slowly on polylysine. Other molecules, as yet unidentified in the culture fluid or the substrate, may be important for synapse formation.

We are grateful to Dr W. Adams and Ms Robin Davis for reading this paper and for their expert advice, to Ms Hedi Niederer for her unfailing technical help, to Mr P. Baettig for his skilled help with the photography and to Ms J. Wittker for excellent secretarial work. Supported by grant 3.525-1.83 from the Swiss Nationalfond. Dr H. Arechiga was supported by a grant from the Hoffmann-La-Roche Stiftung.