INTRODUCTION

Dudai (1990) described that an ideal subject for studying the cellular and molecular bases of learning should, probably, have ten large nerve cells, ten genes, a generation time of 1 week, and the ability to play a cello and recite Shakespeare (note that this sentence originated with W. G. Quinn). Nobody knows such organisms, but gastropod molluscs are such compromises. Gastropod molluscs are excellent experimental preparations that not only neurobiologists but also psychologists are using to help understand the basic mechanisms underlying learning and memory. They have relatively simple central nervous systems (CNSs) with large, identifiable neurons. The neurons are accessible for detailed electrophysiological, biophysical, biochemical, and molecular studies. Such a small brain in invertebrate was recently termed “microbrain” by Mizunami et al. (1999). To date, many scientists have analyzed various neural mechanisms on production of complex, long-lasting behavioral changes of gastropod molluscs including Aplysia, Hermissenda, Limax, and so on (Alkon et al., 1993; Bailey and Kandel, 1993; Frank and Greenberg, 1994; Gelperin et al., 1996; Abel and Kandel, 1998; Alkon et al., 1998; Matzel et al., 1998; Silva et al., 1998; Ito et al., 1999; Kimura et al., 1999). Associative learning in the pond snail, Lymnaea stagnalis (Fig. 1), can be also used as an important system for investigating the neurobiology of learning and memory. It includes classical and operant conditioning, for example, using feeding behavior (Alexander et al., 1982; Audesirk et al., 1982; Kemenes and Benjamin, 1989a, b, 1994; Whelan and McCrohan, 1996; Kemenes et al., 1997; Staras et al., 1998, 1999a, b), respiratory behavior (Lukowiak et al., 1996, 1998; Hermann and Bulloch, 1998; Spencer et al., 1999), withdrawal behavior (Sakakibara et al., 1998), and the isolated brain (Veprintsev and Rosanov, 1967). We here review our recent findings for associative learning, particularly conditioned taste aversion (CTA), in L. stagnalis.

Fig. 1

The pond snail, Lymnaea stagnalis.

CONDITIONING BY CHEMOSENSORY STIMULI

First, we considered that it was necessary to determine whether, in L. stagnalis for the same behavior such as feeding, aversive and appetitive conditioning yield different strengths and periods of either acquisition or retention. To this end, we examined the effects of various chemosensory and physical stimuli on feeding and avoidance behavioral responses (Kojima et al., 1996). Then, using these findings, we constructed classical-conditioning paradigms with aversive and appetitive stimuli (Kojima et al., 1995, 1996). In the aversive conditioning paradigm, an appetitive stimulus (sucrose, conditioned stimulus: CS), which increased the feeding response, was paired with an aversive stimulus (KCl, unconditioned stimulus: UCS), which inhibited feeding behavior. Upon presentation of KCl, the aversive conditioning, which is generally called conditioned taste aversion (CTA), was acquired quickly and persisted for up to a month (Fig. 2A, B). On the other hand, the appetitive conditioning paradigm paired a neutral stimulus (vibration, CS) with an appetitive stimulus (sucrose, UCS). This conditioning took longer to acquire and persisted for a shorter period of time than the CTA (Fig. 2C, D).

Fig. 2

Classical conditioning. The feeding responses evoked by the conditioned stimulus (sucrose or vibration) were counted. See Kojima et al. (1996) for the details. (A, B) Feeding response and time dependence of retention in the aversive (sucrose-KCl) conditioning, i.e. conditioned taste aversion. (C, D) Feeding response and time dependence of retention in the appetitive (vibration-sucrose) conditioning. The areas between the two dashed lines in (B and D) show the means±SEM of feeding responses in the control snails at the immediate conditioning tests. The time in the abscissas of (B and D) is expressed in a logarithmic scale. All data are means±SEM. * indicates p<0.01.

Second, we demonstrated a sensory preconditioning in L. stagnalis (Kojima et al., 1998). An appetitive sucrose solution (CS1) and weak vibration (CS2) were first associated, and then the CS2 and an aversive KCl solution (UCS) were done. To build the conditioning, two different training procedures, massed and spaced, were examined. It is well known in psychology that spaced training, interposing a rest interval between the multiple training sessions, produces stronger and longer-lasting memory than massed training, which has the same number of training sessions with no rest interval (Hintzman, 1974). After the both training, the sensory preconditioning was built: significantly fewer feeding response to the CS1 became elicited; slower latency to the first bite to the CS1 was induced (Fig. 3). However, no significant differences on the memory retention between these training procedures were found in the sensory preconditioning, possible because this conditioning included a neutral conditioning (CS1-CS2) (see Kojima et al., 1996 for the details).

Fig. 3

Sensory preconditioning. The feeding responses and the latencies to the first bites evoked by the first conditioned stimulus (sucrose) were measured. See Kojima et al. (1998) for the details. (A–C) Massed training. (D–F) Spaced training. S, V and K are sucrose (CS1), vibration (CS2), and KCl (UCS) stimulus training, respectively. The areas between the two dashed lines in (C and F) show the means±SEM of feeding responses in the control snails at the immediate conditioning tests. The time in the abscissas of (C and F) is expressed in a logarithmic scale. All data are means±SEM. * indicates p<0.05.

Third, an operant conditioning that L. stagnalis suppressed its naturally occurring behavior of escape from a water tank was examined by using a negative reinforcement (i.e. an aversive KCl stimulus) prepared outside the tank (Fig. 4A, Kobayashi et al., 1998). During the training period, the number of escapes from a tank was strongly suppressed. One of behavioral factors for this suppression was confirmed as the elongation of latency to the first escape after training. The effects on the memory retention were examined in the massed and spaced training procedures. The memory retention by the massed training was observed within 20 min after training (Fig. 4B–D). By the spaced training, the learning acquisition was found to be stronger, which was observed as the slower latency to the first escape, than by the massed training, but the longer-lasting memory retention, which had been expected first, was not formed (Fig. 4E1–I). These results suggest that once L. stagnalis recognize the external environment is safe after training, they may extinguish their memory of the past situation quickly, resulting in no or very little difference in the memory retention by two different training procedures in this operant conditioning. Compared with the behavior in classical conditioning (Kojima et al., 1995, 1996, 1998) and its neural mechanism described in the next sections (Kojima et al., 1997, Nakamura et al., 1999a, b, c), these findings may help to address not only the neural basis of operant conditioning but also the relation between the classical and operant conditioning.

Fig. 4

Operant conditioning. See Kobayashi et al. (1998) for the details. (A) Schematic presentation of training apparatus for the operant conditioning. (B–D) Massed training. (E–I) Spaced training. The voluntary activity shown in (I) was measured from the conditioning snails in the pre-test at Day 1 and in the post-test at Day 4. All data are means±SEM. * and ** indicate p<0.05 and p<0.005, respectively.

CELLULAR MECHANISM OF CONDITIONED TASTE AVERSION

Based on the behavioral experiments for the CTA in L. stagnalis (Kojima et al., 1995, 1996), we proposed a neuromodulatory model (Fig. 5A). When the CS (sucrose) is followed by the UCS (KCl) in the training session, the association of the CS and UCS potentiates an inhibitory pathway, resulting in suppression of the feeding response to the CS. Taking account of the underlying neural circuits so far reported (Benjamin and Elliott, 1989; Ferguson and Benjamin, 1991; Syed and Winlow, 1991; Elliott and Kemenes, 1992, McCrohan and Kyriakides, 1992; Inoue et al., 1996a, b; Yeoman et al., 1996), our model further proposed that sensory neuron(s) sensitive to the appetitive CS excite the feeding central pattern generator (CPG) in the CS pathway to induce the feeding response, and that sensory neuron(s) for the aversive UCS excite interneurons in the UCS pathway, resulting in the withdrawal response (Fig. 5B). The cerebral giant cells (CGCs) exert weak excitatory monosynaptic and strong inhibitory polysynaptic influences upon the neuron 1 medial (N1M) cells in the CPG, and the repetitive firing of the CGCs results in inhibitory influence on the N1M cells (Yeoman et al., 1996). Thus, it was presumed that the inhibitory influence of the CGCs upon the N1M cells may be potentiated in the conditioned snails to suppress the feeding response. To test the validity of this model, we examined the differences in synaptic inputs to the N1M cells by stimulating the CGCs in the conditioned and control snails (Kojima et al., 1997), and further studied whether or not the association of the chemosensory inputs (sucrose and KCl) occurs in the CGCs (Nakamura et al., 1999a, b).

Fig. 5

Cellular mechanism of conditioned taste aversion. See Kojima et al. (1997) for the details. (A, B) Neuromodulatory model. A white and a black arrow or circle indicate an excitatory and an inhibitory influence or synapse, respectively. A dashed line part in (A) shows a variability of transmission strength. A solid and a dashed line in (B) indicate a monosynaptic and a polysynaptic projection, respectively. CV1: cerebral ventral 1 cell; CGC cerebral giant cell; N1M: neuron 1 medial cell; SN: sensory neuron; IN: interneuron; MN: motoneuron. (C–E) Enhancement of IPSP to the feeding central pattern generator. Typical IPSPs (C) and their summarized data (D) in the N1M cells, which were induced by depolarizing current injection into the CGCs during a bar, were recorded in the conditioned (filled circles) and control (open circles) snails. In (E), we defined the conditioned snails, which bit less than twice for 90 sec by the conditioned stimulus (sucrose), as the good performers, and compared the IPSPs at 12 sec after the onset of CGC depolarization. Backward indicates backward-conditioning control. The data are means±SEM. * and ** indicate p<0.05 and p<0.01, respectively.

Inhibitory postsynaptic potential (IPSP) which was evoked in the N1M cell by activation of the CGC was larger and lasted longer in the conditioned snail than that in the control snail (Fig. 5C–E). The electrical properties of the cell bodies of CGCs and the responses of the CGCs to the chemosensory inputs were not changed during the CTA. These results, together with the previous report indicating the existence of excitatory projection from the N1M cells to the buccal motoneurons 1 (B1 cells) involved in feeding behavior (Kemenes and Elliott, 1994), suggested that enhanced IPSP in the N1M cells may underlie the suppression of feeding responses in the CTA of L. stagnalis.

The neural circuits from the chemosensory neurons receiving the CS and UCS have not yet been demonstrated. Generally, as compared to the motoneuronal system, the chemosensory neuronal system has been poorly understood as regards its role in feeding behavior of gastropod molluscs (Kemenes et al., 1986). Even in L. stagnalis, only six types of endings of primary neurons are known (Zylstra, 1972). These neurons have been morphologically classified as sensory neurons in the epithelia of lips, tentacles, and dorsal surface of head (Zylstra, 1972; Zaitseva and Bocharova, 1981). Neither the secondary chemosensory neurons nor the interneurons directly following the primary ones have been identified. Therefore, to understand the cellular mechanism of the CTA, the neural pathways which transmit taste signals to the feeding CPG in L. stagnalis should be elucidated clearly.

We thus targeted to clarify pathways for the feeding responses between the chemosensory neurons and the regulatory neurons, such as the CGCs. To achieve this goal, the lip and tentacle nerves of L. stagnalis, were characterized using histological techniques (Nakamura et al., 1999a). Anatomical drawings showed the detailed distributions of the superior lip, median lip, and tentacle nerves in the lip and tentacle; particularly it was found that the mouth is mainly innervated by the superior lip nerve. The neurons in the CNS by backfilling of the superior lip nerve and/or the median lip nerve with fluorescence dyes made some clusters, whereas those stained from the tentacle nerve made other clusters. These stained neurons were not part in the CPG or its regulatory neurons for feeding. These results, therefore, suggested that the superior lip nerve may be employed as a principal factor in the chemosensory transduction from the mouth, and that no direct projections from the CPG or its regulatory neurons for feeding to the lip and tentacle nerves.

Furthermore, the lip and tentacle nerves of L. stagnalis, were characterized using electrophysiological techniques (Nakamura et al., 1999b). When the activity of those nerves was induced in lip-tentacle preparations, aversive taste (KCl) signals were transmitted through all the lip and tentacle nerves, but appetitive (sucrose) signals could be recorded only through the superior lip nerve. In the CNS immersed in high Mg²⁺-high Ca²⁺ saline, electrical stimuli applied to any of the nerves failed to induce action potentials in the CGCs, implying that the signals are polysynaptically transmitted to the CGCs (Fig. 5B). Intracellular recordings revealed that the CGCs in semi-intact preparations received both appetitive and aversive taste signals not only through the superior lip nerve but also through the median lip nerve (Fig. 6). In addition, an osphradium was ruled out as a candidate for appetitive reception. The present results, together with our preceding data arrived at by the histochemical analyses (Nakamura et al., 1999a), indicated that the appetitive taste transduction responsible for generating the feeding responses is performed through the superior lip nerve with some contribution of the median lip nerve. The data showing that the CGCs can receive various taste signals suggested that they may play a crucial role in feeding behavior as demonstrated in the study of CTA (Fig. 5B).

Fig. 6

Intracellular recordings of the CGCs in response to appetitive and aversive chemical stimuli applied to the lips in semi-intact preparations. See Nakamura et al. (1999b) for the details. (A) Schematic presentation of the semi-intact preparation. (B, C) Responses of the CGCs to sucrose and KCl which are used for the acquisition of conditioned taste aversion.

On the other hand, serotonin (5-HT) has been emphasized of neurotransmitters in the CNS of L. stagnalis, because 5-HT is known to play an important role to activate the feeding motor pattern (Yeoman et al., 1994, 1996; Kemenes et al., 1997). The CGCs are known to be serotonergic (Boer et al., 1984; Tuersley and McCrohan, 1988; Yeoman et al., 1996). Although these previous studies in immunohistochemistry let us gain deep insights into the neurobiology of L. stagnalis, all the results were obtained by two-dimensional (2-D) observation according to conventional methods, leading some defects that a 3-D reconstruction of immunoreactive neurons in the CNS was ambiguous and that discrimination of individual small neurons from cell clusters was difficult. Therefore, we examined 3-D arrangement of 5-HT-like immunoreactive neurons in the CNS of L. stagnalis, by a combination of immunohistochemistry and confocal laser scanning microscopy (Hatakeyama and Ito, 1999a). In addition to the confirmation of previously identified serotonergic neurons, some important new findings were obtained (see Hatakeyama and Ito, 1999a for the details). These results could produce the exactly 3-D maps for 5-HT-like immunoreactive neurons in the CNS, and will help the further analyses of neural networks employing 5-HT.

Recently, γ-aminobutyric acid (GABA) has been noted. Injection of GABA into the haemocoel of intact L. stagnalis resulted in the modulation of feeding activity (Romanova et al., 1996). Such injection also immediately evoked the rhythmic movements of radula, including opening the mouth and three-phase cyclic radula biting: protraction, rasping, and swallowing. Therefore, we examined the distribution of GABA-like immunoreactive neurons in the CNS of L. stagnalis using a 3-D analysis by immunohistochemistry combined with confocal laser scanning microscopy (Hatakeyama and Ito, 1999b). Our results showed that GABA is ubiquitously located in the CNS, particularly that the paired GABA-like immunoreactive neurons in the buccal ganglia resemble a pair of buccal motoneurons 19 (B19 cells) of Helisoma trivolvis (Kater, 1974; Richmond et al., 1991). To our knowledge, this is the first description of such neurons in L. stagnalis. Functional studies demonstrated that application of GABA to the B19 cells in H. trivolvis stimulated the rhythmic patterned activity, which resembled fictive feeding (Richmond et al., 1994). Taken together, these results emphasized the need for further analysis of the functional role of GABA-like immunoreactive neurons in L. stagnalis (corresponding to the B19 cells of H. trivolvis) and their influence of the activity of feeding motor pattern.

DEVELOPMENTAL STUDY OF CONDITIONED TASTE AVERSION

The relationships between development and learning were strongly emphasized in gastropod molluscs, particularly in Aplysia californica by Carew and his colleagues (Marcus et al., 1994; Nolen and Carew, 1994; Marcus and Carew, 1998), because neurobiologists have long speculated that growing processes involved in the development of CNS may persist in the adult where they could subserve learning and memory. As a first step to study the relationships between the development and learning in the CNS of L. stagnalis, we examined developmental changes in the acquisition and retention of CTA (Yamanaka et al., 1999). We found that L. stagnalis developed ability of a CTA as a long-term memory through three critical stages (see Fig. 7A for the outline of developing snails). Embryos in veliconcha (stage 25) started to respond to appetitive sucrose at the first critical stage. This response was in good agreement with morphological observations (Meshcheryakov, 1990) that embryos at this developmental stage seemed to be physically ready to eat. However, they could not associate this appetitive stimulus (CS) with an aversive stimulus of KCl (UCS). At the second critical stage, embryos just before hatching (stage 29) acquired the CTA, but the conditioned response did not persist. Through this stage, they may acquire learning ability to safely seek out food in an external environment. At the third critical stage, immature snails with a 10 mm shell could use a long-term memory to maintain the conditioned response. This memory persisted for at least a month, showing that now they are able to maintain a long-term memory so that they can safely eat a variety of food when they cover wide territory to search for a mate. The findings indicated that the development of learning ability in snails, which secures acquisition of better survival ability, is coincident with the major changes in their life cycle.

Fig. 7

Developmental changes in conditioned taste aversion. See Yamanaka et al. (1999) and Sadamoto et al. (1999) for the details. (A) Outline of developing snails kept at 20°C. The shell length is expressed in a logarithmic scale. ED and PND indicate embryonic days and postnatal days, respectively. St.: embryonic stage according to Meshcheryakov's criteria (1990). (B–E) Increase in numbers of cells and that in volumes of the buccal (B, C) and cerebral (D, E) ganglia. The shell lengths of 0.4 mm, 0.7 mm, 1.2 mm, 2 mm, 10 mm, and 20 mm correspond to St. 24, St. 25, St. 29 embryos, juvenile snails, immature snails, and adult snails, respectively.

Next, to provide the anatomical substrate upon which the CTA is superposed during the development, we examined the number of cells and the volumes of the buccal and cerebral ganglia in L. stagnalis at the critical developmental stages described above (Sadamoto et al., 1999). The buccal and cerebral ganglia include the majority of neurons involved in the CTA. We found that the numbers of cells in these ganglia are almost saturated in the immature snails, but the volumes of these ganglia still increase from the immature snails to the adults (Fig. 7B–E). These results suggested that most of the cells indispensable to the CTA emerge at the immature stage, but that individual cells in the ganglia continue to enlarge even in adulthood. Furthermore, a pair of key neurons (CGCs) for the CTA were found to mature at the immature stage. This study provided the anatomical substrate upon the long-term CTA, by which snails can eat safe food in a wide territory. Furthermore, we studied the synthesis and storage of 5-HT in the CGCs, using L. stagnalis at the above critical developmental stages. There was a positive correlation between the acquisition of CTA and the first appearance of 5-HT immunoreactivity in the CGCs at the embryonic stage 29 (Yamanaka et al., unpublished data). These results, therefore, indicated that the development of a pair of key neurons (CGCs) for the CTA stimulates the developmental changes in the learning ability.

PKA-DEPENDENT REGULATION OF SYNAPTIC ENHANCEMENT

As described repeatedly, the CGCs play a crucial role in the regulation of CTA in L. stagnalis. However, the mechanisms of signal transduction from the CGCs to the follower buccal interneurons and motoneurons are not clear. To our knowledge, only the work by McCrohan and Gillette (1988) suggested that cyclic AMP (cAMP) is a candidate of the second messenger in the CGCs by demonstrating the appearance of Na⁺ current by the cAMP injection. We thus examined whether cAMP-dependent protein kinase (PKA) contributes to enhancement of a monosynaptic connection between the presynaptic CGCs and the postsynaptic B1 cells (Nakamura et al., 1999c). Injection of cAMP into the CGCs or inhibition of phosphodiesterase by isobutylmethylxanthine in the CGCs increased the amplitude of excitatory postsynaptic potential (EPSP) in the B1 cells, whereas no changes were detected in the electrical properties of the CGCs (Fig. 8A–C). The synaptic enhancement in the B1 cells was completely blocked by inhibition of PKA in the CGCs but did not require a de novo protein synthesis due to a PKA phosphorylation (Fig. 8D, E). The increase in the EPSP amplitude of B1 cells was associated with the increase in the amount of 5-HT release from the CGCs (Fig. 8F, G). These results thus provided the physiological evidence of the direct regulation of a synaptic enhancement by PKA in the CNS of L. stagnalis, indicating the completely different mechanism from that in the well-studied siphon- and gill-withdrawal reflex in A. californica (Abel and Kandel, 1998).

Fig. 8

PKA-dependent regulation of synaptic enhancement between the CGCs and the follower B1 motoneurons. See Nakamura et al. (1999c) for the details. (A–C) Synaptic enhancement by injection of cAMP into the CGCs. Single EPSPs recorded in the B1 cells 3 hr after the onset of injections are shown in (A), and compound EPSPs are shown in (B). The changes in the amplitudes of compound EPSPs recorded in the B1 cells 3 hr after the onset of injections are summarized in (C). (D, E) Blockade of the increase in EPSP amplitude in the B1 cells by PKA-inhibitor (Rp-cAMPS) application on the cerebral ganglia including the CGCs. The changes in the EPSP amplitudes recorded in the B1 cells 3 hr after the onset of injections are summarized in (E). (F, G) Dose-response of the EPSP amplitude in the B1 cells to 5-HT₃ antagonist. The EPSP amplitudes shown in (F) were recorded in the B1 cells 3 hr after the onset of injections of cAMP (filled circles) and 5′AMP (open circles), when the 5-HT₃ antagonist was applied on the whole CNS. The concentrations of 5-HT₃ antagonist for the half maximum blockade are summarized in (G). All data are means±SEM. * and ** indicate p<0.01 and p<0.001, respectively.

INVOLVEMENT OF NITRIC OXIDE

The function of nitric oxide (NO), particularly a specific role in feeding behavior, has been started to be examined (Elofsson et al., 1993; Moroz et al., 1993, 1994a, b, 1995, 1999; Elphick et al., 1994, 1995; Martinez, 1995; Korneev et al., 1998; Serfözö et al., 1998). To examine whether NO-generative neurons are included in the central circuitry for generation of feeding pattern in L. stagnalis, two staining techniques for NADPH diaphorase and 5-HT were applied for its CNS (Sadamoto et al., 1998). The former technique is known to show localization of NO synthase (NOS); the latter is well employed as a marker for the feeding circuitry because 5-HT is a main transmitter in it (Kemenes, 1997). In the buccal ganglion, B2 cells were found to be a pair of putative NO-generative neurons (Fig. 9A). These motoneurons are not involved directly in the coordination of feeding pattern (Fig. 9B) but are activated simultaneously with feeding to control the oesophageal and gut tissues for digestion. Taking account of the diffusion effects of NO, the NO released from the B2 cells, when the feeding is started, is considered to sufficiently modulate the feeding circuitry. In the cerebral ganglion, the superior lip, median lip, and tentacle nerves included both putative NO-generative fibers and serotonergic fibers. These fibers were not identical, but the NO released in the nerves may activate the serotonergic fibers, resulting in the influence upon the initiation of feeding. Therefore, our findings clearly showed that NO is not involved in transmission within the central circuitry for feeding, but suggested that NO can crucially affect feeding behavior, such as initiation and modulation of the feeding pattern.

Fig. 9

Histochemistry to reveal the relation between NO-generative neurons and feeding circuitry in the buccal ganglia. See Sadamoto et al. (1998) for the details. (A) NADPH-diaphorase staining in the right buccal ganglion. An arrow points the B2 motoneuron. (B) 5-HT immunohistochemistry showing the feeding circuitry. bc: buccal commissure; dbn: dorsobuccal nerve; cbc: cerebrobuccal connective.

To demonstrate that NO is generated in the CNS accompanied with feeding behavior, we measured the increase in NO concentration at the buccal ganglia in semi-intact preparations of L. stagnalis using an NO specific electrode, when the lips of these preparations were stimulated by sucrose (Kobayashi et al., unpublished data). The NO concentration at the buccal ganglia was significantly increased by an application of sucrose to the lips. The rhythmic NO response well corresponded to the rhythmic bursting (fictive feeding response) of the B2 cells. These data provided the first direct evidence that NO is actually generated in the CNS of L. stagnalis and involved in a specific behavior such as feeding.

IMAGING BY NEW MICROSCOPES

New microscopes have been recently developed to image morphology and function of multiple sites in tissues or small cellular elements that are inaccessible by conventional light and electron microscopy and electrophysiological techniques. In the neurobiology of learning and memory, one of the most striking findings is that long-term memory involves morphological changes in neurons (Bailey and Kandel, 1993). To observe the fine structures of neurons isolated from the CNS of L. stagnalis, we thus attempted to detect fine 3-D structures of living neuronal terminals with at least 40-nm vertical information, and succeeded in observing the real-time dynamics of their terminals, such as attraction and repulsion with each other, using an atomic force microscope (AFM) for a liquid environment (Nagayama et al., 1996, 1997; see Shao et al. for the detail of AFM). The time-dependent fine dynamics of the neurons obtained here is beyond the resolution power of an optical microscope, and actually it has never been accomplished with an electron microscope that requires of fixing and staining.

More recently, we began to image the changes in membrane potential by using an optical recording technique (see Kojima et al., 1999 for the detail of our optical recording system) as well as the those in intracellular Ca²⁺ concentration (Alkon et al., 1992; Collin et al., 1992; Ito et al., 1994). The electrical responses in the cerebral ganglion and the median lip nerve of L. stagnalis were measured, when a short current pulse was delivered to the median lip nerve, which transmits chemosensory signals to the cerebral ganglion (Kojima et al., unpublished data). We detected a composite depolarizing response in the cerebral ganglion, which consisted of a sharp depolarizing response corresponding to a compound action potential and a following slow depolarizing response. After pharmacological analyses, the slow depolarizing response was found to originate from glial cells. Our results thus provided the first evidence for neuron-independent signals in glial cells of gastropod molluscs and implied the contribution of the glial signaling to the chemosensory processing in the CNS of L. stagnalis.

CONCLUDING REMARKS

The studies of CTA in L. stagnalis provide us with insight into the richness of cellular correlates of associative learning behavior in the relatively simple CNS, or microbrain. The next critical step in our understanding of the neural mechanisms of associative learning in L. stagnalis is to proceed studies at molecular level. In A. californica, it has been possible to identify some of the genes and gene products that contribute importantly to induction of the long-lasting neural and behavioral changes underlying the long-term memory (Abel and Kandel, 1998). Unfortunately, the molecular analyses involved in the associative learning of L. stagnalis are future work. We just started the molecular cloning for genes concerned with the CTA (Sadamoto et al., unpublished data). We really hope that many molecular biologists take part in studying the molecular mechanisms of associative learning in L. stagnalis and that gene-knockout snails will be produced in near future.