The Specificity of Synapse Formation by identified Leech Neurones in culture

As neurones send out processes towards their targets and form synapses during development or regeneration, they execute a complex sequence of steps. In the initial stages, trophic factors and substratum molecules provide stimuli for neurite extension and cues for guidance. As a last step, the right cells must form the correct connections with each other if neuronal circuits are to generate performance that is functionally useful. In addition, the physiological properties of these synapses must be appropriately tuned. Thus, a particular synapse between particular types of neurones in the CNS has a defined set of properties that depend upon the specialized pre-and postsynaptic structures: the synapse is characteristically excitatory or inhibitory, chemical or electrical or both, rectifying or nonrectifying, strong or weak, facilitating or depressing with trains of impulses, and subject or not subject to modulation.

Numerous questions arise concerning the way in which newly formed synapses acquire this assembly of properties that make up specificity (Vidal-Sanz et al. 1987). These include: (1) how specialized pre-and postsynaptic structures are formed on appropriate regions of the cells, (2) the mechanisms that influence the choice of establishment of a chemical or an electrical synapse and (3) the possible role of the transmitter itself in determining synapse formation. (For example, do all neurones containing serotonin have the ability to connect in the same way with the same type of target cell?) In the present series of experiments we have approached these and related questions that bear on cell recognition and synapse formation between pairs of identified neurones maintained in tissue culture.

For technical reasons it is usually not possible to record physiologically from synapses as they form within the CNS of a vertebrate or an invertebrate. The synapses at which transmission actually occurs are inaccessible and the exact timing and mechanism of synapse formation cannot be measured directly. The sequence of changes in fine structural appearance is hard to define. Moreover, it is usually not possible to record simultaneously with microelectrodes from the two individual cells that constitute the pre-and the postsynaptic elements. Such difficulties arise largely from the geometry of the neurones and the profusion of cells within the CNS.

An idealized preparation would be far simpler and would consist of two identified nerve cells of known function in vitro, with known membrane properties and transmitters. It would be important for synapses to form rapidly, within hours, at well-defined sites close to the soma from which electrical recordings could be made. Since 1979 (Ready and Nicholls, 1979), when identified leech neurones were first isolated from the leech CNS and placed in culture, numerous invertebrates have provided preparations that fulfil these criteria for following synapse formation. For example, neurones isolated from leeches, snails and Aplysia can re-establish chemical or electrical synapses, often with properties resembling those normally formed during development in the CNS of the animal (Fuchs et al. 1982; Liu and Nicholls, 1989; Bodmer et al. 1988; Haydon et al. 1987; Rayport and Schacher, 1986). The experiments described here show that the connections between the isolated leech neurones can form rapidly and reliably, are highly specific and can be used to follow in detail the sequence of steps as chemical and electrical synapses are formed.

The particular cell that is to be isolated can usually be recognized by its shape, size and position in the leech ganglion. Particularly easy to identify without electrical recording are the sensory P, N and T cells, the 5-hydroxytryptamine (5-HT)-containing Retzius cells, the L and the AE motoneurones and the AP cells of unknown function (Nicholls, 1987). Other cells such as the tiny DL or VL serotonergic cells cannot be found simply by visual inspection of the ganglion: they must be recorded from electrically or stained selectively before removal to identify them (Acklin and Nicholls, 1990; Stuart et al. 1974). The cell is pulled out either by lassoing with fine nylon monofilament or by suction into a mouth-held pipette after mild enzyme treatment of the ganglion with collagenase-dispase. Neurones can now be extracted together with a variable length of the axonal arborization or without processes as rounded cell-bodies.

The isolated neurones survive for up to 3 weeks in a suitable medium consisting of Leibowitz-15 with 2 % foetal calf serum and garamycin (Dietzel et al. 1986; Fuchs et al. 1982). A convenient feature is that N, P, T, Retzius and AP cells maintain their distinctive action potential configurations in culture. Hence, in a pair of cells that has made a synapse it is always possible to be certain which cell is which. In spite of such constancy, measurable changes do occur in the distribution over the cell surfaces of ion channels for sodium and potassium with time in culture. For example, immediately after a Retzius cell has been removed, the sodium channels are present in highest concentration in the axons, especially at their broken tips. As outgrowth proceeds, the sodium channels appear in the growth cones and also in the soma (Garcia et al. 1990).

For the establishment of synapses, pairs of identified neurones are placed close together in various configurations (see Fig. 1 and below). Neurones will also form synapses when placed at a distance from one another after process growth has occurred. Such synapses are, however, less favourable for physiological and morphological analysis since the sites of transmission cannot be recognized by simple inspection within the profusion of neuronal processes.

The presence or absence of a synaptic connection, as well as the mode of transmission, depends on which two cells have been apposed in culture. The results obtained so far with leech neurones are shown in Table 1. The following are a few examples, (a) Purely electrical synapses without rectification form between pairs of L cells, (b) Rectifying electrical synapses develop when an L cell is paired with a P sensory cell. (In this case depolarization of the P cell spreads well to the L cell, but hyperpolarization does not. Conversely, hyperpolarization but not depolarization spreads from L to P.) No signs of chemically mediated interactions occur at these electrical synapses, (c) By contrast, purely chemical transmission with 5-HT as the transmitter develops when Retzius cells are paired with P cell. no sign of electrical transmission can be detected, (d) When a Retzius cell is paires with another Retzius cell, mixed chemical and electrical synapses develop (without rectification), (e) Touch sensory (T) cells have not been seen to establish connections in culture. Numerous other connections are shown in Table 1 (see Fuchs et al.1982; Aréchiga et al. 1986; Liu and Nicholls, 1989).

What these connections demonstrate is orderliness and regularity. The same types of cells always form the same types of connections. One exception to this rule is provided by the N cells. These two cells, medial and lateral in their locations in the ganglion, are sensory for nociception and have similar receptive fields. They differ in their membrane properties and connections, and a specific antibody has been found that binds to the medial but not the lateral N cell. These two N cells also make distinctive connections in culture. Like the P cell, the lateral N cell, when paired with Retzius, L, AE or AP cells, always makes a rectifying electrical connection (such that depolarization spreads better than hyperpolarization from the lateral N cell to its partner and vice versa). The medial N cell, however, behaves differently. It can make a rectifying or a nonrectifying synapse on the same partner. In some cultures the medial N cell forms a nonrectifying junction with the Retzius cell while in others the junction shows rectification (Vycklicky and Nicholls, 1988). The factors that influence this choice by the medial N cell are not known.

Some but not all of the connections made in culture resemble those seen in the ganglion: Retzius to Retzius, Retzius to P, P to L and L to L connections seen in the dish are virtually identical to those occurring in the animal. At the same time, certain connections fail to reappear between these cells. Although Retzius cells form chemical connections on P cells in the ganglion and in culture, the normal strong chemical connection from P to Retzius cells is not reproduced in culture, nor is the strong chemical connection from P to L. We cannot at present explain these failures. A reasonable assumption would be that certain factors conducive to synapse formation are present in embryos but lacking in the dish.

In addition, novel types of connections not seen in the ganglion can form in culture. The strong electrical coupling between Retzius and L cells and AE cells provide examples (Aréchiga et al. 1986). Again, this result, though not explained, is hardly surprising. Presumably cell-cell recognition is only one of many factors that determine the patterns of connections that develop during embryogenesis. One can imagine that neurones that have the potential for forming synapses never do so because they never come into contact within the nervous system. In culture the same two neurones can be directly apposed and contact one another.

Together these results emphasize the importance of cell-cell recognition as the final step in synapse formation. By mechanisms that are not known, a neurone, such as the Retzius or the P cell, can decide what type of synapse it will form, and whether it will be the pre-or the postsynaptic element, depending solely on the partner it encounters, without reference to the fluid environment or the substratum.

Do all cells containing the same transmitter make similar connections? This question has been answered by isolating DL and VL cells (Acklin and Nicholls, 1990). These two small neurones, like Retzius cells, contain and secrete 5-HT as their transmitter. Yet, as shown in Table 1, they make quite different connections from those of Retzius cells. They form purely electrical synapses on the P cell with no sign of chemical transmission (Fig. 2). Hence the type of transmitter that a neurone liberates does not on its own determine how its target cell is selected.

To what extent do synapses in culture resemble those in the ganglion? First, in their physiological properties the newly formed synapses display normal characteristics. Thus, the nonrectifying synapses established between Retzius, L or AP cells are indistinguishable from those in situ. Similarly, the rectifying P to L or N to L junctions show normal voltage sensitivity. Chemically mediated transmission is more flexible, with marked facilitation and a slow modulation that depends on the holding potential in the presynaptic cell. Quantal release of 5-HT has been demonstrated by measuring miniature potentials and by statistical analysis of fluctuations in the amplitudes of evoked synaptic potentials (Henderson et al. 1983). These experiments, together with voltage-clamp studies, show that changes in release rather than postsynaptic mechanisms account for facilitation ani modulation (Stewart et al. 1989). As soon as the synapses form in culture they as endowed with these dynamic characteristics (see below).

The electron microscopy of synapses in culture reveals fine structural features similar to those in situ. A problem for making such comparisons is that the chemical synapses most intensively studied in vitro, the serotonergic Retzius to P and Retzius to Retzius connections, have not been identified by electron microscopy in the leech central nervous system. Nevertheless, the principal characteristics shown in Fig. 3 are typical for 5-HT synapses in general. Dense-cored vesicles form a ring around clusters of clear vesicles that are apposed to active zones in the presynaptic Retzius cell (Kuffler et al. 1987). At these sites the pre-and postsynaptic membranes are thickened; the extracellular space is widened and contains densely staining material. In the postsynaptic P cell cisternae are evident. The chemical synapses formed by Retzius cells on one another exhibit similar morphological characteristics, Many but not all of these features appear within hours as the synapses form in culture (see below).

The dense-cored and clear vesicles turn over after prolonged stimulation. Markers such as horseradish peroxidase, ferritin and colloidal gold in the external fluid are taken up and transferred from coated vesicles, cisternae and vesicles (Kuffler et al. 1987). As expected, only the presynaptic Retzius cell takes up label after stimulation of both cells with a high external K⁺ concentration (50 mmol l^-1). Colloidal gold particles have been observed in both dense-cored and clear vesicles.

We do not yet know which type of vesicle corresponds to the quantum of release. Gap junctions have not yet been sought or identified at electrical synapses formed by leech neurones in culture.

The experiments illustrated in Fig. 4 and Table 2 were made to determine whether the properties and speed of synapse formation are influenced by the particular sites on the surfaces of the two cells that come into contact. Neuronal cell bodies within the ganglion of the leech are virtually free of synapses (see, however, French and Muller, 1986) and all contacts are made at a distance in the neuropile. Are cell bodies, initial segments and axons equivalent in their ability to form pre-and postsynaptic structures, as well as chemical or electrical synapses?

Neurones were apposed in various configurations: (1) with their somata touching (this was the site of contact in all earlier experiments); (2) with initial segment touching initial segment; (3) with initial segment touching soma and (4.with axon touching axon. The differences were dramatic. For convenience the results are summarized in Table 2 for the mixed chemical and electrical synapses that develop between pairs of Retzius cells in culture (Liu and Nicholls, 1989).

With soma to soma contact between Retzius cells (A), electrical transmission appeared first, after about 3 days, becoming stronger over the next 6 days. Chemical transmission invariably developed later at about 3–8 days, but was often masked by the strong electrical coupling (alas! no method has yet been found for uncoupling leech electrical junctions). With initial segment to initial segment apposition (B), the sequence and reliability were quite different. Within 6–8 h bidirectional, chemically mediated transmission appeared. Often chemical transmission in one direction was stronger than in the other. These chemical synapses became stronger hour by hour and showed clear facilitation and depression. Their fine structure examined at 12–24 h resembled that observed in mature synapses in culture. The site could be easily identified at the tips of the apposed initial segment. Only after a prolonged delay of 24 h was there evidence of weak electrical transmission (which is easier to detect, by physiological recording, than chemical transmission). Unlike soma to soma contact, the sequence, timing and reliability of synapse formation were precise. That the tip of the initial segment represented a preferential site for chemical synapse formation was demonstrated by apposing cells with the tip of one Retzius cell initial segment touching the soma of another (C). Again, chemical synaptic transmission appeared in 6–8 h, became stronger and was followed after about 24h by electrical transmission. But, with this configuration the chemical transmission was in only one direction - from initial segment to soma, never in the reverse direction. Y. Liu (in preparation) has further studied synapse formation by the same cells with their axon tips touching (D). For those experiments the Retzius cell was pulled out with a length of axon intact. The broken tips of the axons were apposed in culture, a technically difficult manoeuvre. With the Retzius neurones in this configuration electrical transmission appeared before chemical transmission.

The appeal of these experiments is that they set the stage for examining cellular and molecular events in synapse formation. At what other synapses can one follow the sequence of steps in such detail and so directly? Experiments are now in progress to examine the cytoskeleton during synapse formation at predetermined sites.

The molecular composition of the substratum upon which a leech neurone is plated determines the extent and pattern of its outgrowth, as well as the distribution of voltage-sensitive calcium channels in the newly grown processes. Of particular interest is a laminin-like molecule that can be extracted from the CNS (Chiquet et al. 1988; Masuda-Nakagawa et al. 1988). With this high molecular weight protein as a substratum, certain neurones such as the Retzius cell, start to grow within 30min and send out long slender processes with few branches. The processes contain large numbers of Ca²⁺ channels: stimulation leads to Ca²⁺ entry that can be measured optically using the Ca²⁺-sensitive dye Arsenazo III (Ross et al. 1987, 1988). The location of the laminin molecule can be studied in leech ganglia and connectives using a specific monoclonal antibody. In the adult nervous system, the laminin is restricted to the basal lamina surrounding ganglia and connectives, and is not in direct contact with axons or cell bodies. In developing leech embryos, however, or after injury, the laminin is distributed differently. Laminin is observed at sites in the connectives that link ganglia along which axons will grow during development or regenerate after injury (L. Masuda-Nakagawa, unpublished observations).

A second substratum upon which leech neurones grow profusely is the plant lectin concanavalin A. Retzius cells plated on concanavalin A rapidly send out processes after about 30min. But, unlike those on laminin, the Retzius cell processes become curved, thick and show marked fasciculation (see Fig. 5). Such processes contain relatively few calcium channels compared to those grown on laminin. Thus, different substrata can have markedly different effects on outgrowth rates, patterns and Ca²⁺ channel distribution. This result raises the possibility that certain other substratum molecules might be located at key sites within the CNS or even on particular regions of neuronal surfaces. In principle, such molecules could facilitate or inhibit synapse formation by mechanisms analogous to those used by agrin at the skeletal neuromuscular junction. As yet the molecules that influence synapse formation by leech neurones in culture have not been identified.

The formation of a synapse represents the end-point of a series of decisions made by two neurones. The principal focus of the present experiments was the demonstration of cell-cell recognition in the absence of external cues, using the formation of (or failure to form) chemical or electrical synapses as indicators. With two identified cells close together in culture one can monitor with microelectrodes the mechanism of transmission from outset to maturity (something that cannot usually be achieved within the CNS). That chemically mediated transmission, with its vastly more complex structural requirements, could precede the formation of electrical junctions was a surprise, as was the finding of regional preferences over different regions of the cell surface. We are, however, still far from understanding molecular mechanisms and cytoskeletal rearrangements that cause outgrowth, pattern formation, cessation of growth, the formation of specialized pre- and postsynaptic endings or the selection of the chemical or electrical mode of transmission. How does the nature of the partner determine these events? It seems that simple adherence could not on its own provide an explanation.

For nervous systems consisting of millions of cells rather than hundreds or thousands it seems that the same principles might apply but for large populations rather than for single cells. The simple leech, with only 14 sensory neurones it each ganglion to tell it where a body segment is being stroked, pressed or bitten, has presumably to specify each single cell to a far greater extent to obtain information about the outside world and to regulate its behaviour.

We wish to thank our colleagues at the Biocenter in the Department of Pharmacology, particularly Drs W. B. Adams, L. Masuda-Nakagawa and R. R. Stewart. We also wish to thank Mr P. Bättig and Ms J. Wittker for their excellent technical assistance and secretarial help. Dr R. Guggenheim and Mr D. Mathys kindly provided assistance at the Electron Microscopy Laboratory of the Geology Institute. This work was supported by grants from the Swiss National Fonds to JGN (3.556-0.86) and to DPK (3.516-0.86).