Foreign Connections are Formed In Vitro by Aplysia Californica Interneurone L10 and Its In Vivo Followers and Non-Followers

A prominent feature of invertebrate nervous systems is the invariance of their neuronal connectivity. Different animals of the same species display stereotyped synaptic connections between identified cells. This anatomical simplicity has prompted the use of invertebrates in the search for functionally defined neuronal circuits. Indeed, much of our current understanding of patterned motor output derives from studies of identified sets of invertebrate neurones that are activated during breathing, locomotion or ingestive behaviour (for a recent review see Getting, 1989). Functional circuits associated with behaviour such as learning and memory are also accessible in the invertebrate nervous system (for a recent review see Byrne, 1987).

The invariance of neuronal connections in invertebrates suggests that the synaptic specificity may be obligatory. The present work addresses this hypothesis by constructing small circuits in vitro with identified neurones from Aplysia californica. This approach follows the work of Ready and Nicholls (1979), who showed that identified neurones from the leech can establish chemical connections in vitro. The techniques of invertebrate cell culture have since been extended to other preparations, notably Aplysia californica (Kaczmarek et al. 1979; Dagan and Levitan, 1981; Schacher and Proshansky, 1983), the pond snail Helisoma trivolvis (Wong et al. 1981), the pond snail Lymnaea stagnalis (Kostenko et al. 1983), the squid Loligo pealei (Parsons and Chow, 1989) and to crustaceans (Cooke et al. 1989). Co-cultures of identified neurones have been used to study the development of electrotonic connections (Hadley et al. 1983; Bodmer et al. 1984), the development and physiology of chemical synapses (Fuchs et al. 1981; Fuchs et al. 1982; Camardo et al. 1983; Haydon, 1988) and mechanisms for short-term and long-term synaptic modification (Rayport and Schacher, 1986; Montarolo et al. 1986; Dale et al. 1988). Recently, co-cultured invertebrate neurones have been used to examine emergent properties of neuronal circuits (Kleinfeld et al. 1990; Parsons et al. 1989a; Syed et al. 1989).

Previous studies have shown that connections not found in vivo can occur in vitro. Nicholls and coworkers used co-cultured identified leech cells and found that many cells formed connections in vitro that largely resembled those in the ganglion, while others formed novel, albeit weak, synapses in culture (Fuchs et al. 1981; Fuchs et al. 1982; Arechiga et al. 1986; Vyklicky and Nicholls, 1988). In cocultures of Aplysia neurones, Schacher et al. (1985) found that giant neurone R2 formed novel connections in vitro, including a connection from R2 to interneurone L10. Novel connections among co-cultured left upper quadrant (LUQ) neurones (Parsons et al. 1989) and other identified neurones (Kleinfeld et al. 1990) have also been observed. In contrast, Schacher and coworkers found that the synaptic specificity of L10 for three groups of cells, i.e. LUQ cells, right upper quadrant (RUQ) cells and R14, was maintained in vitro (Camardo et al. 1983; Schacher,. 1988). The results for Aplysia suggest that L10 may be unusual in its ability to maintain specificity in culture.

Here we examine whether the preservation of synaptic specificity of L10 is common or is exceptional in co-cultures with other Aplysia neurones. The systematic study of this question is facilitated by the results of recent studies of the cellular and synaptic properties of L10 in situ (Frazier et al. 1967; Kandel et al. 1967; Koester and Kandel, 1977; Koester and Alevizos, 1989). In particular: (i) L10 makes connections onto a large number of cells in vivo; many of these are followers and do not make connections back onto L10; (ii) the sign of the postsynaptic response is different for different followers; (iii) many followers are relatively large and thus amenable to isolation and study in culture.

We assayed the connections formed by co-cultured pairs of neurones, consisting of L1 and a follower or non-follower. Two additional co-cultures that did not involve L10 were examined in conjunction with these studies. Our focus was on describing the patterns of synaptic specificity in vitro. Biophysical aspects of the novel connections are not emphasized in the present work.

Preliminary reports of this work have been presented (Kleinfeld et al. 1988; Obaid et al. 1989; Parsons et al. 1989 b).

Juvenile Aplysia californica (1–10 g; stage 12), raised in mariculture, were either a gift of the Howard Hughes Medical Institute (Woods Hole, MA) or purchased from Sea Life Supply (Sand City, CA). Identified neurones in the abdominal ganglion were located by their relative position and size and by the morphology of their axonal branching patterns (Frazier et al. 1967; Kandel et al. 1967). Neurones were isolated and maintained in cell culture as described earlier (Dagan and Levitan, 1981; Schacher and Proshansky, 1983; Kleinfeld et al. 1990). Pairs of neurones were plated on the same day with their axons crossed approximately 200 μm below the soma. This procedure ensured that the neurones remained in close proximity prior to their attachment to the culture dish. The morphology and relative position of the cells was sketched at the time of each plating. Extensive outgrowth was present 2–3 days after plating, as shown for co-cultured L10 and R15 in Fig. 1. Co-cultures in which either cell lost its axonal segment were rejected for further study; such cells may be incapable of forming chemical connections (Schacher et al. 1985; Kleinfeld et al. 1990).

A potential systematic error in our experiments arose from incorrect identification of neurones. This error was judged to be negligible for particularly distinctive cells such as L7 and R15. For other neurones this error was estimated to be no more than 10 %.

Standard current-clamp methods were used. Neurones were penetrated with single-barrel microelectrodes filled with 4 mol l⁻ potassium acetate at their tips and backfilled with 3 mol l⁻ KC1 (R_E=15 –20 M Ω). Electrotonic coupling between neurones was measured in response to small, hyperpolarizing currents as described previously (Bennett, 1977). Postsynaptic potentials were measured in response to action potentials elicited in the presynaptic cell by the injection of brief, suprathreshold current pulses. For some measurements the postsynaptic cell was penetrated with separate current-passing and voltage-sensing electrodes. A synapse was judged to be present if we observed a postsynaptic response with a magnitude of at least 5 mV after 1–10 spikes in the presynaptic neurone; significant responses less than 5 mV in amplitude following multiple spikes in the presynaptic cell were never observed. A synapse was judged to be absent if a significant postsynaptic response was not observed after firing the presynaptic cell at the high rate of 10–20 spikes s⁻ for 5 s; this criterion was modified for L10/R15 co-cultures (see Results).

All recordings were made in an artificial sea water (ASW) that consisted of 460mmoll⁻ NaCl, 10mmoll⁻ KC1, 10mmoll⁻ Hepes, 10 mmol l ⁻¹ glucose, 55 mmol l⁻ MgCl₂ and 10mmoll⁻ CaCl₂, pH7.6, at room temperature. Cocultures were examined 3–5 days after plating.

We examined co-cultures of L10 with six of its followers in the ganglion (L7, L11, L12, L13, R15 and R16) and two non-followers (a second L10 and L14A). In situ, L10 forms purely inhibitory inputs to L11, L12 and L13, purely excitatory inputs to R15 and R16, and a dual-action input to L7, with inhibition followed by excitation (Kandel et al. 1967) (Fig. 2). None of these followers or non-followers is reported to form connections back to L10.

The firing properties of the neurones, either alone or in co-culture, are discussed first. We then present results on the novel reciprocal connections and unidirectional connections that are established in vitro. A large number of co-cultures were prepared so that the probability of forming a given type of synapse could be established (Table 1).

The firing properties of many neurones in the intact ganglion have been characterized. Interneurone L10 produces an aperiodic bursting output in vivo (Frazier et al. 1967; Koester et al. 1974). Follower R15 produces a stereotypic (‘parabolic’) bursting output consisting of trains of action potentials separated by periods of hyperpolarization (Strumwasser, 1965; Alving, 1968). In contrast, followers L7, L11, L12, L13 and R16 and the non-follower L14A produce nonbursting output (Frazier et al. 1967). We first examined the firing properties of these cells in culture.

Interneurone L10 was studied in two different culture environments, i.e. isolated L10 neurones (N=10) or co-cultures of L10 with one of its followers (N=77) (Fig. 1). Intracellular recordings showed that about half the LlOs fired with a rhythmic pattern of output activity, regardless of the culture environment. A typical pattern consisted of well-defined bursts of action potentials, lasting for 25–50 s, separated by longer periods of silence (Fig. 3A,B). The rhythmic, bursting pattern continued unabated for up to 10 h. On rare occasions (N=3) we observed a rhythmic output pattern with a relatively short period between bursts (Fig. 3C). The rhythmic bursting patterns observed in vitro are distinctly periodic compared with those observed in vivo.

Two features of an individual burst were seen with every L10. (i) The amplitude of the action potentials within a burst increased with time until it reached a plateau value 5-10 s after the onset of the burst (Fig. 3B,C). (ii) The burst terminated with a subthreshold depolarization (Fig. 3B,C).

L10 neurones that did not burst could be induced to fire tonically by the constant injection of small depolarizing currents. A current of about 0.1–0.5 nA was needed to drive a cell just above its threshold for firing, at which it fired at a rate of 2–3 spikes s⁻¹. The firing properties of cultured LlOs were uncorrelated with the input resistance of the cell (R_In=35±15MΩ S.D.) and the basis for the heterogeneity among L10 cells from different animals is unclear.

Follower R15 neurones, cultured in isolation, produced an output pattern consisting of bursts of action potentials with a pronounced hyperpolarization between bursts (Fig. 4A). These features closely approximate to the behaviour reported for R15 in vivo (Strumwasser, 1965; Meech, 1979). Similar results were found with other cultured R15 cells (N=6). A typical burst pattern of R15 cocultured with L10 is shown in Fig. 4B. The bursts occurred periodically, but the durations of the bursts were shorter than those reported for R15 in vivo. Bursting by R15 in vitro has also been reported by Kramer et al. (1988).

None of the other neurones we examined (L7, L11, L12, L13, L14A and R16) exhibited rhythmic bursting in vitro, in agreement with their behaviour in vivo. Each of these cells could be induced to fire tonically by the constant injection of pmall, depolarizing currents. The behaviour of L14A was notable in that its rate of firing tended to accelerate, for a period of about 30 s, following the onset of activity. A similar increase in firing rate occurs in vivo (Carew and Kandel, 1977).

Co-cultures of L10 paired with followers L7, LU, L12 or L13 formed reciprocal inhibitory connections. The connections from L10 to each of L11, L12 and L13 were reminiscent of those in vivo (Kandel et al. V)6T). The connection from L10 to L7 differed; a dual-action synapse, with inhibition followed by excitation, occurs in vivo. Novel connections were also observed: each of these followers formed a connection to L10. These results are summarized in Fig. 2 and Table 1.

Neurone L13 formed a particularly strong novel connection to L10. Typically, inhibitory postsynaptic potentials (IPSPs) with a magnitude of 10–20 mV followed a single action potential in L13 (Fig. 5A-C). This IPSP did not facilitate but it did fatigue to about half its maximum value after repeated stimulation of L13. A connection from L13 to L10 occurred in all but one co-culture (12 out of 13 cocultures). In addition, the type of connection found in vivo from L10 to L13 was present in about half the cultures (Fig. 5D). In vitro, the IPSP in L13 was observable following a brief train of action potentials in L10 but not following a single action potential, in contrast to the connection from L13 to L10.

Neurone L7 also formed strong, novel synaptic connections to L10 (12 out of 24 co-cultures). We observed IPSPs with a magnitude of 2–5 mV following a single action potential in L7 (Fig. 6A-C). Further, we typically observed spontaneous, brief hyperpolarizing events about 50 s after the initial IPSP (Fig. 6C).

A solely inhibitory input from L10 to L7 was also present in about half the cultures (Fig. 6D). There was a significant probability, at the level of two standard deviations, that co-cultured L7 and L10 cells that expressed one connection would express the reciprocal connection. It was interesting that there were no instances in which an excitatory response or a dual-action response was observed in L7. This is in contrast to the response observed in vivo (Wachtel and Kandel, 1971) and previously reported to occur in vitro (Schacher, 1988).

It was surprising that L10, a cholinergic neurone (Giller and Schwartz, 1971), did not elicit a dual-action response in L7 in vitro. In particular, the dual-action response characterized in vivo can be mimicked by the application of acetylcholine to L7 (Frazier et al. 1967). One possible explanation for this loss is that L7 may lose the capacity for dual-action responses to cholinergic stimulation in vitro. To test his we analyzed co-cultures of L7 with R2 (N=3); R2, like L10, is cholinergic (Giller and Schwartz, 1971), although it may contain additional transmitters (Schacher et al. 1985). We found that single action potentials in R2 produced an excitatory postsynaptic potential (EPSP) in L7, while rapid firing unmasked an IPSP. The dual-action response, with inhibition followed by excitation, is clearly seen following a burst of action potentials in R2 (Fig. 7). The in vitro response in L7 following activation of R2 was similar to that observed in vivo following excitation of L10 (Wachtel and Kandel, 1971). The synapse from R2 to L7 is novel and demonstrates that L7 is capable of forming a dual-action synapse in vitro.

The connection from L10 to followers L11 and L12 observed in vivo (Kandel et al. 1967) was also present in vitro (Fig. 8B,D). However, L11 and L12 formed novel inhibitory connections to L10 (Fig. 8A,C). Unlike the situation with L13 or L7, a single action potential in L11 elicited a relatively weak (ΔV<lmV) response in L10. However, a strong summed response was observed following a brief train of action potentials in LU (Fig. 8A). Connections of similar strength were observed for the connection from L12 to L10 (Fig. 8C). These novel connections were found in about three-quarters of the co-cultures studied (13 out of 19 for L11/L10 and 7 out of 9 for L12/L10).

In the light of the partial preservation of the synaptic function observed in vivo for the L10 to L7 connection, it was of interest to examine the biophysical properties of L10 connections to followers L11, L12 and L13 (Figs 5D, 8B,D). We considered these properties in the light of the novel input from L10 to L7 (Fig. 6D). The postsynaptic response was examined in selected cells using a two-electrode current-clamp. The results for L11 co-cultured with L10 are shown in Fig. 9A. With L11 poised at −45 mV, just below its threshold potential, action potentials in L10 elicited both fast IPSPs and a slow IPSP. The fast IPSP inverted at a potential of about −60mV and showed some evidence of facilitation (Fig. 9A). It could be reversibly blocked by the addition of 100 μmoll⁻ tubo-curarine to the ASW (data not shown). This response was similar to the chloride-mediated IPSP characterized in situ (Kehoe, 1972a). The slow IPSP inverted at a potential of about −80 mV (Fig. 9A). It was not affected by tubocurarine but was blocked by the addition of 500 μmoll⁻ tetraethylammonium to the ASW (data not shown). This response was similar to the potassium-mediated IPSP that occurs in vivo (Kehoe, 1972b). These and related data for co-cultures with L12 (N=4) show that L10 was capable of forming inhibitory connections in culture that functioned similarly to those in the ganglion. Furthermore, the strength of these synapses was maintained with repeated bursting output from L10. This was demonstrated in a co-culture of L10 with Lll, where L11 had been activated by the injection of a small, constant current (Fig. 9B). The bursts in L10 cause the functional inhibition of L11 over repeated periods of about 10³s.

Neurones R15 and R16 receive excitatory input in vivo (Kandel et al. 1967). In vitro, L10 also made connections onto these cells, but the input was inhibitory rather than excitatory. Reciprocal connections of these followers back onto L10 were not observed in vitro. In addition, L14A, a non-follower of L10, formed a novel inhibitory connection to L10.

The inhibition of R15 by L10 was initially deduced from the effects of L10 on the pattern of rhythmic bursting in R15. Activation of L10 resulted in a marked increase in the period between bursts in R15 (Fig. 10A). Further, action potentials in L10 could terminate a burst in R15 prematurely. This functional assay implies that L10 inhibits R15, although we did not observe IPSPs in R15 during the period between bursts (Fig. 10A). However, IPSPs were observed in R15 in response tq stimulation of L10 when R15 was depolarized to a potential of at least -50 mV (Fig. 10B). For these measurements we waited until R15 stopped firing at each level of depolarization. There were no instances in which an excitatory response in R15 was observed.

Inhibitory connections were also formed from L10 to R16 (6 out of 7 instances) (Fig. 11A). IPSPs with magnitudes of 1–2mV were observed following single action potentials in L10. The recovery time of the IPSP was relatively fast (about 1 s), like that observed with R15. However, the IPSP in R16 was clearly discernible at its resting potential, −55 mV, in contrast to the IPSP in R15 (cf. Figs 10B, 11A). There were no instances in which an excitatory response in R16 was observed. Further, we never observed a connection from R16 to L10.

A novel inhibitory connection to L10 was formed by L14A (Fig. 11B). The IPSPs had a magnitude of 1-2 mV following single action potentials in L10. There was only one instance of a connection from L10 to L14A (7V=12). A single occurrence is an unreliable indicator (see Materials and methods) and we conclude that L10 does not innervate L14A.

The results above show that L10 is capable of forming atypical inhibitory connections with a variety of other neurones from the abdominal ganglion. We therefore investigated whether one L10 could innervate other L10 cells in vitro. Only a single instance, out of 38 possible instances, was observed. This small occurrence is an unreliable indicator (see Materials and methods). We thus conclude that L10 does not innervate other L10 cells in vitro.

A well-defined specificity of connections was found in co-cultures of Aplysia californica interneurone L10 with in situ followers and non-followers (Fig. 2 and Table 1). This specificity partially coincided with that observed in vivo (Kandel et al. 1967). In vivo connections whose sign was preserved in vitro were made by L10 onto neurones Lll, L12 and L13 (Figs 5D, 8B,D). Connections in vitro that did not preserve the sign of corresponding synapses in vivo were made by L10 onto R15, R16 and L7 (Figs 6D, 10A, 11A). Connections that have never been observed to occur in vivo, in terms of either direction or sign, were formed in vitro between L10 and followers L7, Lll, L12, L13 and R16 and non-follower L14A (Figs 5D, 6D, 8B,D, 11A,B). All these novel connections were inhibitory and some were strong (Fig. 5A,C).

The synaptic response observed in two of our co-cultures differs from that reported by Schacher (1988). In those studies the connections between L10 and followers L7 and R15 preserved their in vivo specificity. The basis for the differences with these particular synapses is unclear. However, novel inhibitory connections in addition to those reported here have been observed in vitro. A novel connection from R2 to L10 has been reported by Schacher et al. (1985) and connections involving L10 and identified cerebral neurones are discussed by Chiel et al. (1986).

It is striking that all the novel connections to L10 in vitro were inhibitory (Table 1). Similarly, the known connections to L10 in vivo from other interneurones are also exclusively inhibitory (Kandel et al. 1967; Koester and Alevizos, 1989). This suggests that the specificity of these connections may be regulated by the postsynaptic cell, L10. Postsynaptic regulation of the synaptic morphology of Aplysia sensory neurones for motor neurone L7 has been demonstrated by Glanzman et al. (1989). A precedent for postsynaptic regulation of synapse formation has also been described in mammalian nervous systems (Sur et al. 1988). The morphology of presynaptic, but foreign, inputs is adjusted and resembles the morphology of the native input (Metin and Frost, 1989).

The followers that formed novel connections to L10 in vitro do not innervate L10 in vivo. One limiting factor that could account for the absence of these novel connections in vivo is physical proximity. However, many of these followers receive input from L10 in the ganglion (Frazier et al. 1967) and their processes are in intimate contact. Our results suggest that it is unlikely that physical separation prevents the formation of such connections in vivo. It should be noted that, in vivo, L10 is innervated by other interneurones (Kandel et al. 1967; Segal and Koester, 1980). These presynaptic neurones may effectively compete for L10 territory and thus prevent the followers from forming reciprocal connections onto L10. A role for synaptic competition is suggested by the reduced frequency of novel connections for specific followers when other followers are present in vitro (D. Kleinfeld, T. D. Parsons and F. Raccuia-Behling, unpublished results). Thus the use of co-cultures of L10 with single followers may optimize the conditions for the formation of novel connections. A precedent for this interpretation comes from the observed elimination of foreign inputs onto salamander muscle fibres by native inputs (Dennis and Yip, 1978).

We thank B. Friedman for her active interest in this work, H. J. Chiel, S. Schacher and F. Strumwasser for useful discussions, and T. Capo and the Howard Hughes Medical Institute for kindly supplying us with Aplysia. This work was supported by AT&T Bell Laboratories, Public Health Services grant NS 16824 to BMS and ALO and a Veterinary Medical Scientist Training Program Fellowship to TDP.