The synaptic junctions of LE and RF cluster sensory neurones of Aplysia californica are differentially modulated by serotonin

The gill and siphon withdrawal reflex and the tail withdrawal reflex are two excellent systems for investigating the cellular and molecular mechanisms underlying short-term and long-term behavioural modifications (Hawkins et al., 1993; Lechner and Byrne, 1998). The neuronal networks causally related to these reflexes include two main components, a monosynaptic one, the sensory-to-motor neurone synaptic junctions, and a polysynaptic one made up of various excitatory and inhibitory interneuronal synapses (Frost et al., 1988; Trudeau and Castellucci, 1992; White et al., 1993). There are many sites that can be modified when the reflex is facilitated or depressed (Fischer and Carew, 1995; Trudeau and Castellucci, 1993a; Xu et al., 1995). The importance of a specific site and the mechanism involved may depend on the particular protocol being used (Mackey et al., 1987; Byrne and Kandel, 1996). It is readily apparent that a large amount of complexity can be achieved even at the level of a simple withdrawal reflex and in a relatively simple neuronal network. This paper documents another degree of complexity resulting from the fact that the sensory neurones of the RF cluster differ from the sensory neurones in the other clusters.

It is known that the monosynaptic component of withdrawal reflexes in Aplysia californica is a critical site of synaptic modulation occurring during short-term and long-term behavioural changes (Byrne and Kandel, 1996; Castellucci et al., 1978). In these studies, the connections made by the LE sensory neurone cluster of the abdominal ganglion or the VC sensory cluster of the pleural ganglion were investigated. It was repeatedly observed that these synaptic junctions are facilitated by the neuromodulator serotonin (5-hydroxytryptamine, 5-HT) (Ghirardi et al., 1995; Wu et al., 1995). Four sensory neurone clusters have been described in the abdominal ganglion; the LE, rLE, RE and RF clusters (Byrne, 1980; Dubuc and Castellucci, 1991).

The receptive fields of these cells are located on the siphon, the mantle, the branchial cavity and the gill itself (Dubuc and Castellucci, 1991), which is the receptive field of the withdrawal reflex. The properties of these sensory neurones are similar except that the action potentials of the RF neurones are not as sensitive to known modulators of the defensive reflex (Dubuc and Castellucci, 1991). Since the synaptic output of the sensory neurones is to the various motor neurones of the gill and siphon as well as to some motor neurones involved in inking (L14), we wanted to determine whether the RF synaptic junctions were modulated like the other sensory neurones involved in the aversive reflex. Previous studies have shown that pairing sensory neurone tetanization with connective stimulation produces much larger synaptic facilitation than that obtained by tetanization alone (Dubuc, 1990). This was true for all neurones of the LE, rLE and RE clusters, the exceptions being the neurones from the RF cluster. In the latter case, tetanization alone produces a larger effect than that due to pairing with connective stimulation (Dubuc, 1990). Because the connectives are known to contain serotoninergic axons (Longley and Longley, 1986; Hawkins, 1989; Mackey et al., 1989), we thought that this depressing effect could be attributed in part to serotonin. We therefore examined the effects of 5-HT on the RF neurone synapses and compared the results with those for the LE neurone synapses.

Aplysia californica (mass 100–300 g) were purchased from Marine Specimen Unlimited (Pacific Palisades, CA, USA) or Aplysia Resource Facilities (Miami, FL, USA). They were maintained in a large (900 l) tank at 15°C. All experiments were performed at room temperature (22°C) on isolated abdominal ganglia. Before dissection, animals were anaesthetized with an injection of an isotonic MgCl₂ solution (385 mmol l⁻¹) corresponding to approximately one-third of their volume. Dissection of the abdominal ganglion was performed in an extracellular medium consisting of equal volumes of 385 mmol l⁻¹ MgCl₂ and artificial sea water (ASW). The ganglia were pinned to the bottom of a Sylgard-coated chamber (2 ml volume) and the sheath surrounding them was cut using scissors and removed using fine forceps. All preparations were rested under constant superfusion with ASW for at least 2 h before the start of an experiment.

Sharp microelectrodes were pulled from omega-dot borosilicate glass (WPI, Sarasota, FL, USA) and filled with 2 mol l⁻¹ potassium acetate. Their resistances were between 10 and 20 MΩ. Experiments were performed in current-clamp mode, and the voltage signals were recorded through Axoclamp 2B amplifiers (Axon Instruments). Computer acquisition and analysis of the data were performed using PClamp. A modified version of CLAMPEX (Dr M. V. Storozhuk) was also used to measure evoked excitatory postsynaptic potential (EPSP) amplitudes. Data are presented as means ± S.E.M. Student’s t-test was used for statistical comparisons; all probabilities reported are two-tailed.

Artificial sea water (ASW) was composed of (in mmol l⁻¹): NaCl, 460; KCl, 10; CaCl₂, 11; MgCl₂, 55; and Hepes buffer, 10 (pH 7.6). The composition of the 2:1 ASW, which contained twice the normal concentration of Mg²⁺ and 1.25 times the normal concentration of CaCl₂ was (in mmol l⁻¹): NaCl, 368; KCl, 8; CaCl₂, 13.8; MgCl₂, 110; and Hepes buffer, 10 (pH 7.6) (Trudeau and Castellucci, 1992). Experiments with 5-HT (and corresponding controls) summarized in this report were performed in the modified ASW (2:1 ASW) to minimize possible indirect effects of 5-HT on the connections under study. Similar results, however, were observed in ASW containing normal concentrations of Mg²⁺ and Ca²⁺.

In pilot experiments, we rarely observed short (<5–7 ms) and constant-latency EPSPs in ink motor neurones (L14 neurones; Carew and Kandel, 1977a) when the presynaptic neurone was an LE cell; longer-latency (10–15 ms) EPSPs were usually observed. The latencies of these EPSPs were highly variable from trial to trial and they were abolished in bathing solution containing a high concentration of divalent cations (2:1 ASW; Trudeau and Castellucci, 1992), suggesting that these synaptic inputs are indirect. It was important to compare the action of 5-HT on RF and LE synapses in the same preparation to minimize the variation between animals, and we therefore used synaptic junctions with good convergence (more than 50 % of the sampled neurones had a connection), the RF–L14 and LE–LFS neuronal pairs.

In our experiments, the presynaptic neurones were stimulated intracellularly every 30 s and, after the fifth trial, a small volume of 5-HT solution was added to the chamber to obtain a final 5-HT concentration of 10 μmol l⁻¹. The results of the first series of experiments are summarized in Fig. 1. We found that the LE–LFS connections were facilitated in the presence of 5-HT, but there was no facilitation at the RF–L14 connections. On average, the amplitude of the EPSPs of LE neurones for trials 6–10 was increased by 64±24 % (P<0.05, N=6) compared with the control trials (mean of trials 4 and 5). In contrast, the amplitude of the EPSPs of the RF sensory neurones was decreased by 47.8±4.3 % (P<0.01, N=6).

To distinguish between an inhibitory effect or a simple depression of the EPSPs of the RF neurones, we performed a series of parallel control experiments in which no 5-HT was added. To test the generality of the observation, we also examined another RF follower neurone, the gill motor neurone L7. The results of this series of experiments are summarized in Fig. 2A. We found that the synaptic depression was normal in the control experiments (N=10) and that in presence of 5-HT there was no facilitation but an inhibition in addition to the synaptic depression. On average, the amplitude of the RF–L14 EPSPs (trials 6–10) was 52.6±8.5 % (N=5) compared with the control or reference trials (mean of trials 4 and 5). The mean amplitude of the RF–L7 EPSPs was 67.9±9 % (N=5). The corresponding control values obtained in the absence of 5-HT were 86.4±2.1 % for the RF–L14 connections and 91±5.4 % for the RF–L7 connections. Thus, both RF–L14 and RF–L7 connections were inhibited by 5-HT, although the effect tended to be less pronounced in L7. Compared with their respective control groups, the EPSP amplitudes in the two experimental groups were decreased by 39.1 % in L14 (P<0.001) and by 25.4 % in L7 (P<0.01). It is unlikely that the effect of 5-HT is due to shunting of the postsynaptic cell. As illustrated in Fig. 2B, no change in input membrane resistance was observed in neurone L14 (N=4).

Previous studies have suggested that the transmitter of the LE and VC connections is an excitatory amino acid. It has also been shown that these synaptic connections were blocked by antagonists of non-NMDA receptors such as CNQX or DNQX (Trudeau and Castellucci, 1993b; Dale and Kandel, 1993). The unusual effect of 5-HT at the RF connections led us to examine whether these synaptic junctions were differently affected by antagonists such as CNQX. We found that the RF–L14 synaptic connections were not unusual in this regard: they were effectively blocked by CNQX (100 μmol l⁻¹). One example of such an experiment is illustrated in Fig. 3A. In these experiments, CNQX was added after the fifth trial (N=3) or after the third trial (N=1). The time course of the effect is shown in Fig. 3B (control, N=4; experimental, N=3). A summary of the results of the four experiments is shown in Fig. 3C; when normalized to the two trials preceding the application of CNQX, the values for the experimental group were 20.5±5.5 % of the corresponding values for the control group. The difference between the two groups was statistically significant (P<0.001). These results are similar to those obtained for the LE connections by Trudeau and Castellucci (1993).

Previous studies have shown that inking and gill withdrawal behaviour in A. californica are rather different from each other in several aspects. Inking is a high-threshold behaviour with a steep input/output relationship approaching an all-or-none response; gill withdrawal is a low-threshold behaviour with a graded input/output relationship, and it is graded as a function of the stimulus intensity (Carew and Kandel, 1977a; Shapiro et al., 1979). These differences are at least partly due to the high threshold and electrical coupling of L14 neurones (Carew and Kandel, 1977a–c). The absence of inking sensitization to strong repeated neck stimuli (Carew and Kandel, 1977a) may indicate other differences between the inking and withdrawal circuits. In this regard, the RF neurones may contribute to these differences. In the present study, we observed a differential effect of 5-HT on the EPSPs evoked by the LE and the RF sensory neurones: while LE–LFS connections were facilitated by 5-HT, the RF–L14 connections were not (Figs 1, 2). Although our results may seem to be unusual for the sensory neurones of the gill and the siphon reflex, they are not unique. It has been shown previously (Rosen et al., 1989), that 5-HT inhibits some sensory neurone connections to neurones in the cerebral ganglion. Our experiments extend this observation for the RF cluster of the abdominal ganglion. We think that modulation of RF connections by 5-HT is likely to occur in simple learning protocols such as sensitization or classical conditioning. This view is supported first by the fact that many serotoninergic processes are found in the neuropile of the abdominal ganglion and, second, that serotoninergic neurones send their processes to the abdominal ganglion and that they are activated by tail shocks (Glanzman et al., 1989; Hawkins, 1989; Longley and Longley, 1986; Mackey et al., 1989) or connective stimulation.

Mechanism of the inhibitory effect of 5-HT on RF connections It is unlikely that shunting contributes to the inhibitory effect of 5-HT on RF connections. For example, the 5-HT effect is not accompanied by any change in input resistance in cell L14 (Fig. 2B). However, we cannot exclude the possibility that local shunting occurs in the neuropile region. Since the mechanoreceptor properties and the receptive field types of the RF neurones do not differ from those of the other cluster cells (Dubuc and Castellucci, 1991), we wanted to determine whether any differences other than the modulation by 5-HT existed. One obvious test was to verify whether the RF connections were blocked by the same AMPA/kainate receptor antagonist, CNQX, that is effective at the LE or VC connections (Trudeau and Castellucci, 1993; Dale and Kandel, 1993). We found that the RF connections are blocked by CNQX (Fig. 3), suggesting that, in this respect, RF–L14 connections are similar to the connections made by the sensory neurones from the other clusters. A similar conclusion (Gapon and Kupfermann, 1996) was drawn for the sensory connections in the cerebral ganglion, which are also inhibited by 5-HT (Rosen et al., 1989). It has been shown that, in the presence of 5-HT, there is an increase in cyclic AMP levels in the sensory neurones (Bernier et al., 1982). Thus, there is a possibility that a different type of 5-HT receptor is present on the RF neurones which could be associated with a decrease in cyclic AMP level; such a receptor has recently been cloned from the central nervous system of A. californica (Angers et al., 1998). It is not yet known whether it is present on the RF neurones.

We wish to thank France Cartier for typing the manuscript. This work was funded in part by grants MT-12092 and MT-14142 from the Medical Research Council of Canada and from NIMH grant MH50525 to V.F.C.