- Email: [email protected]
A Persistent Cellular Change in a Single Modulatory Neuron Contributes to Associative Long-Term Memory
Current Biology, Vol. 13, 1064–1069, June 17, 2003, 2003 Elsevier Science Ltd. All rights reserved. DOI 10.1016/S 09 60 - 98 22 ( 03 )0 0 38 0- 4
A Persistent Cellular Change in a Single Modulatory Neuron Contributes to Associative Long-Term Memory Nicholas G. Jones,1 Ildiko´ Kemenes, Gyo¨rgy Kemenes, and Paul R. Benjamin* Sussex Centre for Neuroscience School of Biological Sciences University of Sussex Falmer Brighton, BN1 9QG United Kingdom
Summary Most neuronal models of learning assume that changes in synaptic strength are the main mechanism underlying long-term memory (LTM) formation [1]. However, we show here that a persistent depolarization of membrane potential, a type of cellular change that increases neuronal responsiveness, contributes significantly to a long-lasting associative memory trace. The use of a model invertebrate network with identified neurons and known synaptic connectivity had the advantage that the contribution of this cellular change to memory could be evaluated in a neuron with a known function in the learning circuit. Specifically, we used the well-understood motor circuit underlying molluscan feeding [2–4] and showed that a key modulatory neuron involved in the initiation of feeding ingestive movements [5] underwent a long-term depolarization following behavioral associative conditioning [6]. This depolarization led to an enhanced single cell and network responsiveness to a previously neutral tactile conditioned stimulus, and the persistence of both matched the time course of behavioral associative memory. The change in the membrane potential of a key modulatory neuron is both sufficient and necessary to initiate a conditioned response in a reduced preparation and underscores its importance for associative LTM. Results and Discussion Although there are a number of examples of long-term, conditioning-induced changes in the electrical properties of neurons [7–10], it has been difficult to establish a relationship between learning-induced cellular changes and behavioral plasticity. The well-described feeding system of the pond snail, Lymnaea stagnalis, allowed us to investigate how a cellular change in an identified modulatory neuron contributes to associative LTM in a neural network that undergoes well-defined electrical changes after behavioral classical conditioning [6, 11]. Feeding in this snail is a rhythmic motor behavior controlled by an exceptionally well-understood central pat*Correspondence: [email protected] 1 Present address: Sensory Function Group, Centre for Neuroscience, King’s College London, London Bridge, SE1 1UL, United Kingdom.
tern generator (CPG) network [2–5, 12, 13]. In their natural environment, snails carry out rhythmic biting movements and take food from floating pond-weed by using a toothed tongue or radula and are able to form associations between the mechanical and chemical properties of food [14]. In the laboratory, an effective feeding stimulus is sucrose, which we use as the rewarding unconditioned stimulus (US) in our classical conditioning experiments. The focus of the present study is the modulatory CV1a cell, which directly excites the most important feeding CPG interneuron, N1M, (Figure 1) and is thus involved in the initiation and control of feeding movements [5, 15]. Our previous work already has shown that appetitive (reward) classical conditioning leads to electrophysiologically tractable associative LTM in Lymnaea [6, 11]. Here, we show that persistent conditioning-induced depolarization of CV1a makes a key contribution to this associative memory trace. Classical conditioning in intact snails [14] was carried out by pairing touch to the lips (the conditioned stimulus, CS) with sucrose (0.01 M final concentration) in spaced training (15 trials over 3 days, five trials a day, 90 min intertrial intervals). This type of training leads to a longterm memory trace [16] whose electrical correlates were recorded in subsequent electrophysiological experiments. Two measures of behavioral LTM were used [6]. One was the number of conditioned snails showing feeding responses to the touch CS compared with a random control group; the other was the strength of response to the CS (feeding bites per minute) compared with controls. After 15 behavioral training trials, the vast majority of the conditioned animals (22/23, 96%), but only a minority of the random control animals (6/23, 25%), responded to lip touch with an increased feeding rate. The mean feeding response to the lip touch CS in the conditioned group was 5.3 ⫾ 0.6 bites/min, whereas, in the control group, it was ⫺0.9 ⫾ 0.4 bites/min. Both of these parameters of the behavioral response to the CS were significantly greater in the conditioned compared with the control group (p ⬍ 0.0001 [Fisher’s exact test] and p ⬍ 0.01 [t test], respectively). Electrophysiological recording of CV1a in reduced head-brain preparations [6, 11] from trained animals showed an enhanced CV1a response to the CS after conditioning that was part of the electrical correlate of LTM. In preparations made from conditioned animals 1 day after training, touch to the lips induced rhythmic bursts of spike activity, known as fictive feeding, in CV1a and a corecorded feeding motoneuron (Figure 2A, top record). This was different from preparations made from random control animals in which touch induced little or no burst of activity (Figure 2A, bottom record). Statistically, more preparations showed a fictive feeding response to the CS in the conditioned group (6/6, 100%) compared to the control group (2/11, 18%) (Figure 2C, left; Fisher’s exact test, p ⬍ 0.01); the response was also significantly stronger in the conditioned (2.7 ⫾ 0.7 cycles/min) versus the control group (⫺0.9 ⫾ 0.6 cycles/ min) (Figure 2C, right; t test, p ⬍ 0.002). The pattern of
Cellular Mechanism of Long-Term Memory Formation 1065
Figure 1. A Simplified Summary of the Lymnaea Feeding Circuit Synaptic connections are shown between the modulatory Cerebral Ventral 1a (CV1a, [15]) neuron and the feeding central pattern generator (CPG) circuit [27, 28]. CV1a is able to drive the CPG via a monosynaptic excitatory connection with the N1 medial (N1M) CPG neuron [5, 15]. Further types of excitatory and inhibitory synaptic connections occur between the N2 and N3 CPG interneurons (N2d/ v, dorsal or ventral type N2 neuron; N3p/t, phasic or tonic type N3 neuron). The CPG neurons drive rhythmic activity in motoneurons (cell types B1-10) that produce feeding movements via contractions of the muscular feeding apparatus [17]. The bars represent excitatory synaptic connections; the closed circles represent inhibitory synaptic connections.
fictive feeding in CV1a and motoneuron B1 induced by touch after conditioning (Figure 2A) was typical of the electrical activity known to underlie behavioral feeding in the intact system [17], and we use this activity to monitor the systems level expression of the LTM trace in the Lymnaea feeding network [6, 11]. The occurrence of burst patterns in the motoneurons indicated that the CPG network had been activated, because motoneurons only fire rhythmically as a result of synaptic input provided by the CPG [18]. These experiments showed that CV1a always fired during conditioned fictive feeding and was active along with other components of the feeding network that it controls. Thus, the rhythmic activation of CV1a by the CS following behavioral conditioning is part of the global electrophysiological readout of the associative LTM trace. To examine whether changes in CV1a’s electrical properties could be contributing to LTM, we looked in more detail for cellular changes in CV1a cells following training. The most striking change was a persistent depolarization of the CV1a membrane potential. This depolarization increased the probability of CV1a firing in response to the CS. In the example shown in Figure 2A, top record, the CV1a cell has a resting potential of ⫺57 mV in a conditioned snail, and it fires in response to the tactile lip stimulus (the CS). In contrast, in the example of a control snail (Figure 2A, bottom record), the CV1a resting potential is ⫺68 mV, and it shows no response to lip touch. Statistically, the mean CV1a resting potential in the conditioned group (n ⫽ 6) was found to be significantly more depolarized, ⫺56 ⫾ 2 mV, compared with that in the random control group (n ⫽ 11), ⫺67 ⫾ 2 mV (t test, p ⬍ 0.03). Moreover, the level of fictive feeding
response (cycles/min) recorded in the CV1a cells (Figure 2C) was significantly correlated (Pearson’s R, 0.6; p ⬍ 0.01) to CV1a resting potential. The effect of conditioning was therefore to depolarize the resting potential of the CV1a cells by an average of 11 mV and to increase CV1a’s responsiveness and that of the entire feeding CPG to the CS. This depolarization was seen in the CV1a cells in all 6 preparations in this first experiment and in all 26 preparations from conditioned snails in 2 subsequent experiments (combined mean resting potential, ⫺54 ⫾ 1 mV [n ⫽ 32]). This was significantly different from the combined mean control CV1a resting potential in the same three experiments, ⫺65 ⫾ 2 mV (n ⫽ 19, t test, p ⬍ 0.001), so the depolarization of CV1a following conditioning was highly consistent. In further experiments, other electrical changes in CV1a that are known to affect neuronal responsiveness, such as a change in spike threshold, membrane input resistance, and spike frequency in response to current steps of different amplitude, also were compared in preparations from conditioned (n ⫽ 7) and control snails (n ⫽ 8); but, unlike the membrane potential, no significant differences were found in any of these other electrical parameters. Neither were there any differences between the two groups in the frequency of fictive feeding cycles driven by CV1a in response to a steady depolarization of around 25 mV; this finding indicates no difference between the ability of “conditioned” and “control” CV1as to activate the feeding CPG. It seems therefore, that the main lasting effect of conditioning on CV1a was to shift its resting potential nearer to its firing threshold. The change in membrane potential induced by conditioning appears to be restricted to the CV1a cells. We tested if cellular changes occurred in other types of neurons in the feeding system, such as the motoneurons corecorded with CV1a or another important modulatory neuron (the Cerebral Giant Cell, [4]), but found no evidence for electrical changes in these cells. Next, we showed that depolarizing the CV1a cell to artificially mimic the changes brought about by conditioning was sufficient to produce an LTM-like electrophysiological trace in the feeding network. By depolarizing CV1a in preparations from naı¨ve snails to ⫺56 mV by using a second electrode for current injection, we could induce an increase in the fictive feeding response to touch (Figure 2B, top traces) similar to that recorded following conditioning (Figure 2A, top traces). At the recorded (nonmanipulated) CV1a membrane potential of ⫺67 mV (Figure 2B, bottom traces), touch produced no fictive feeding response, and this was similar to the lack of response seen in the random control group of the earlier experiment (Figure 2A, bottom traces). With the membrane potential of a single CV1a set at ⫺56 mV, the majority of naı¨ve preparations (12/14, 86%) showed a fictive feeding response to touch (Figure 2D, left), significantly more (Fisher’s exact test, p ⬍ 0.05) than at ⫺67 mV (6/14, 43%). The response to touch at ⫺56 mV was 1.6 ⫾ 0.5 cycles/min, significantly greater (t test, p ⬍ 0.05) than at ⫺67 mV (0.1 ⫾ 0.3 cycles/min, Figure 2D, right). The ability to mimic the effects of conditioning by CV1a membrane potential manipulation is an important part of the evidence showing that changes in the membrane potential in this specific cell type are contributing to the LTM trace.
Current Biology 1066
Figure 2. Electrophysiological Correlate of the LTM Trace Behavioral classical conditioning produces a CS-induced increase in the systems level electrical activity (fictive feeding response). The conditioning-induced increase in the fictive feeding response was mimicked in naı¨ve preparations by the depolarization of CV1a and was significantly reduced in conditioned preparations by the hyperpolarization of CV1a. (A) Top two traces: the lip touch CS evokes rhythmic fictive feeding activity in CV1a and motoneuron B1 in a preparation from a conditioned animal. Bottom two traces: the lip touch CS evokes no rhythmic fictive feeding activity in a preparation from a control animal. (B) Top two traces: the CS evokes rhythmic fictive feeding activity in CV1a and motoneuron B1 in a preparation from a naı¨ve animal when CV1a is artificially depolarized to ⫺56 mV by using a second current injection electrode. Bottom two traces: the CS evokes no rhythmic fictive feeding activity in the same preparation as in (A) with CV1a at its recorded (i.e., nonmanipulated) membrane potential (MP). (C) Significantly more conditioned than control preparations respond to the CS (*, p ⬍ 0.002, Fisher’s exact test, only the percentage of responding preparations is shown for each group here and in [D] and [F]). The response to the CS, which was calculated as the post minus prestimulus difference in fictive feeding activity [6, 11], is significantly stronger in the conditioned versus the control group (*, p ⬍ 0.01, t test). (D) Significantly more naı¨ve preparations respond to the CS (*, p ⬍ 0.05, Fisher’s exact test), and the response is significantly stronger (*, p ⬍ 0.05, t test) when a CV1a is depolarized compared to recorded membrane potential. (E) The touch CS evokes only weak fictive feeding activity in both CV1as and motoneuron B1 in a preparation from a conditioned animal when both CV1as are hyperpolarized to control membrane potential levels. The dashed lines above the CV1a traces indicate the level of membrane potential before artificial hyperpolarization. (F) Significantly fewer conditioned preparations in which CV1a depolarization was reversed by artificial hyperpolarization respond to the lip touch CS (*, p ⬍ 0.05, Fisher’s exact test), and the response is also significantly weaker in this group than in the group of conditioned preparations with CV1as at recorded MP (*, p ⬍ 0.04, t test).
Cellular Mechanism of Long-Term Memory Formation 1067
Previous work already showed that the effect of tactile lip conditioning did not generalize to other parts of the body [14]. It was also important to show that the effect of CV1a depolarization in mimicking the electrical correlate of LTM was similarly restricted to the original site of the CS application, the lips; otherwise, the effects of changing CV1a membrane potential would be nonspecific, unlike the behavioral change. We did the same type of artificial depolarization of the CV1a as in the previous experiment but targeted an alternative touch site, the tentacle [11, 14] (n ⫽ 11 preparations). As before, the response to touch was compared at either ⫺56 or ⫺67 mV in the same preparation. With tentacle touch, there was no significant difference (t test, p ⫽ 0.6) between fictive feeding responses at ⫺56 mV (0.1 ⫾ 0.4 cycles/min) and ⫺67 mV (⫺0.2 ⫾ 0.4 cycles/min), indicating that the effects of depolarization in naı¨ve animals were confined to the lip touch site. This was confirmed when the data on the effects of lip touch were compared with those on tentacle touch. The mean fictive feeding response difference to lip touch at ⫺56 mV was significantly greater, 1.5 ⫾ 0.5 cycles/min, than the 0.1 ⫾ 0.4 cycles/min recorded for the tentacles at the same CV1a membrane potential (t test, p ⬍ 0.05). The necessity of CV1a depolarization for the electrophysiological correlate of the LTM trace was assessed by experiments in which artificial hyperpolarization of the CV1a cells was carried out in preparations made from conditioned snails. If depolarization of CV1a was important for the memory trace, its reversal by hyperpolarization would significantly reduce the fictive feeding response to the CS. For this experiment, it was necessary that both CV1as (the left and right homologs) were hyperpolarized (Figure 2E) to remove the full effects of conditioning-induced CV1a depolarization from the feeding system. In these double-electrode experiments (Figure 2F), when CV1a depolarization was reversed, both the number of preparations responding to the CS (7/11, 63%) and the conditioned fictive feeding response (1.7 ⫾ 0.4 cycles/min) were significantly reduced (Fisher’s exact test, p ⬍ 0.05; t test, p ⬍ 0.04) compared with conditioned preparations with CV1a cells at recorded membrane potential (14/14 [100%] and 2.6 ⫾ 0.4 cycles/ min). This type of experiment provided evidence that conditioning-induced depolarization of CV1a was necessary for the expression of memory trace. To further strengthen the link between a cellular electrophysiological change and behavioral LTM, we showed that the persistence of conditioning-induced CV1a depolarization matched the duration of both the behavioral and electrophysiological memory trace. Previous behavioral analysis showed that the memory produced by 15 trials is stable between 1 and 4 days but declines to control levels by 21 days posttraining [16]. For a direct comparison with the behavior, we tested CV1a membrane potential and fictive feeding responses to the CS in conditioned and control preparations on each of the first 4 days and at 21 days after conditioning. The data from the first 4 days were combined because ANOVAs showed no differences in these parameters between either the conditioned or the control groups (n ⫽ 3–5 animals per group) tested on each of the 4 days following conditioning. In this 4-day period, the
Figure 3. Time Course of the Cellular Change in CV1a The conditioning-induced cellular change in the feeding modulatory interneuron CV1a shows a similar temporal pattern to the systems level electrical correlate of LTM (fictive feeding). (A) CV1as from the conditioned group show a more depolarized membrane potential in the first 4 days following conditioning compared with CV1a cells from the random control group (*, p ⬍ 0.0001, t test), but, by 21 days, there is no statistical difference in CV1a membrane potential between the two groups. (B) Fictive feeding, the electrophysiological correlate of the behavioral LTM trace, shows a similar temporal pattern with elevated fictive feeding responses to the lip CS in conditioned snails compared with control snails in the first 4 days after conditioning (*, p ⬍ 0.0001, t test), but it shows no difference at 21 days. These results indicate that both the cellular and systems level correlate of the LTM memory trace persist for at least 4 days after conditioning but are lost after 21 days. These data are consistent with previously published behavioral data [16] showing a similar time course for LTM retention and loss.
CV1a membrane potential (Figure 3A) was significantly more depolarized (t test, p ⬍ 0.0001) and the fictive feeding response to the CS (Figure 3B) was significantly stronger (t test, p ⬍ 0.0001) in the conditioned (n ⫽ 14) compared with the control group (n ⫽ 12) (⫺55 ⫾ 1 mV versus ⫺67 ⫾ 2 mV and 2.6 ⫾ 0.4 cycles/min versus ⫺0.9 ⫾ 0.6 cycles/min). In contrast, on the 21st day after training, the conditioned (n ⫽ 6) and control preparations (n ⫽ 6) were very similar in their CV1a resting potential (conditioned, ⫺61 ⫾ 2 mV; control, ⫺65 ⫾ 2 mV) and response to touch (conditioned, 0.2 ⫾ 0.4 cycles/ min; control, 0.2 ⫾ 0.3 cycles/min). A comparison was also made between conditioned preparations from the earlier and later time periods. CV1as from conditioned snails were significantly less depolarized (⫺61 ⫾ 2 mV versus ⫺56 ⫾ 1 mV; t test, p ⬍ 0.02), and the CS-induced fictive feeding responses were significantly weaker (0.2 ⫾ 0.4 cycles per minute versus 2.6 ⫾ 0.4 cycles per minute; t test, p ⬍ 0.001) at 21 days than in the 1- to 4-day time period. These experiments show that retention of the memory trace and its eventual loss follow the same temporal pattern as the electrical changes in the CV1a cells. This would be expected if the depolarization of the CV1as was an essential part of the electrical mechanism underlying behavioral LTM. We have demonstrated that behavioral classical conditioning induced long-lasting neuronal plasticity in a single modulatory neuron type controlling feeding in the snail Lymnaea. We found a persistent conditioning-
Current Biology 1068
induced depolarization in neuron CV1a that brought the neuron closer to the threshold for firing. This depolarization makes CV1a more likely to fire in response to the conditioned stimulus and thus drives the feeding CPG network by the direct synaptic pathway to the N1M CPG interneuron (Figure 1). The conditioning-induced depolarization functionally increases the excitability of CV1a and allows the conditioned stimulus to more effectively drive feeding rhythms. General electrical parameters of the cell (e.g., input resistance) remain unchanged. This may indicate that the intrinsic properties of CV1a are not involved in the depolarization, although there are a number of important examples in which long-lasting depolarization occurs with no underlying net input resistance change [19–21]. The persistent change in CV1a membrane potential may have a presynaptic origin, i.e., it is generated by a persistently enhanced extrinsic input, although this seems unlikely, as synaptic input would have been expected to cause some change in input resistance. Whatever its origin, the change in CV1a membrane potential should be regarded as a key element of the conditioned response, because our experiments showed that its depolarization was both sufficient and necessary for an increased response of the whole feeding system to the CS. But how can a cellular change in a modulatory neuron like CV1a lead to a defined conditioned response to a specific CS? We believe that selectivity is ensured by CV1a being embedded in a small network with a highly specialized function that responds only to a narrow range of food-related sensory inputs. These sensory inputs arise mainly from the lips, which normally come into contact with food during rhythmic feeding [22], unlike other structures, such as the tentacles. Indeed, our previous work showed no generalization of the classically conditioned response from the lips to the tentacles [14], and, here, we provided further experimental evidence for response specificity by showing that artificial depolarization of naı¨ve CV1as selectively increases the response to lip touch. Long-term, conditioning-induced changes in the electrical properties of identified neurons occur in other molluscan models of associative learning. These include changes in membrane potential [7, 23] and input resistance [9, 24]. In vitro-evoked LTP in mammalian cortical and hippocampal slices also had suggested that lasting membrane depolarization might be a mechanism for associative neuronal plasticity [25, 26]. However, our work provides the first example of a persistent conditioning-induced cellular change, a depolarization, in an identified modulatory cell type in a CPG-driven motor system in which the cellular changes can be directly linked to behavior. The evidence showing that the depolarization is both sufficient and necessary for eliciting the electrophysiological correlate of the memory trace provides a particularly strong case for the role of cellular changes in LTM. Acknowledgments Support was provided by the Biotechnology and Biological Sciences Research Council (P.R.B., I.K. and N.G.J.) and the Medical Research Council (G.K. and N.G.J.).
Received: March 31, 2003 Revised: April 28, 2003 Accepted: April 29, 2003 Published: June 17, 2003 References 1. Kandel, E.R. (2001). The molecular biology of memory storage: a dialogue between genes and synapses. Science 294, 1030– 1038. 2. Staras, K., Kemenes, G., and Benjamin, P.R. (1998). Patterngenerating role for motoneurons in a rhythmically active neuronal network. J. Neurosci. 18, 3669–3688. 3. Staras, K., Kemenes, I., Benjamin, P.R., and Kemenes, G. (2003). Loss of self-inhibition is a cellular mechanism for episodic rhythmic behavior. Curr. Biol. 13, 116–124. 4. Straub, V.A., and Benjamin, P.R. (2001). Extrinsic modulation and motor pattern generation in a feeding network: a cellular study. J. Neurosci. 21, 1767–1778. 5. Kemenes, G., Staras, K., and Benjamin, P.R. (2001). Multiple types of control by identified interneurons in a sensory-activated rhythmic motor pattern. J. Neurosci. 21, 2903–2911. 6. Staras, K., Kemenes, G., and Benjamin, P.R. (1999). Cellular traces of behavioral classical conditioning can be recorded at several specific sites in a simple nervous system. J. Neurosci. 19, 347–357. 7. Crow, T.J., and Alkon, D.L. (1980). Associative behavioral modification in Hermissenda: cellular correlates. Science 209, 412–414. 8. Moyer, J.R., Jr., Thompson, L.T., and Disterhoft, J.F. (1996). Trace eyeblink conditioning increases CA1 excitability in a transient and learning-specific manner. J. Neurosci. 16, 5536–5546. 9. Brembs, B., Lorenzetti, F.D., Reyes, F.D., Baxter, D.A., and Byrne, J.H. (2002). Operant reward learning in Aplysia: neuronal correlates and mechanisms. Science 296, 1706–1709. 10. Rosenkranz, J.A., and Grace, A.A. (2002). Dopamine-mediated modulation of odour-evoked amygdala potentials during pavlovian conditioning. Nature 417, 282–287. 11. Jones, N.G., Kemenes, G., and Benjamin, P.R. (2001). Selective expression of electrical correlates of differential appetitive classical conditioning in a feeding network. J. Neurophysiol. 85, 89–97. 12. Straub, V.A., Staras, K., Kemenes, G., and Benjamin, P.R. (2002). Endogenous and network properties of Lymnaea feeding central pattern generator interneurons. J. Neurophysiol. 88, 1569– 1583. 13. Jones, R. (2002). Snails with rhythm. Nat. Rev. Neurosci. 3, 918. 14. Kemenes, G., and Benjamin, P.R. (1989). Appetitive learning in snails shows characteristics of conditioning in vertebrates. Brain Res. 489, 163–166. 15. McCrohan, C.R., and Kyriakides, M.A. (1989). Cerebral interneurones controlling feeding motor output in the snail Lymnaea stagnalis. J. Exp. Biol. 147, 361–374. 16. Kemenes, G., and Benjamin, P.R. (1994). Training in a novel environment improves the appetitive learning performance of the snail, Lymnaea stagnalis. Behav. Neural Biol. 61, 139–149. 17. Rose, R.M., and Benjamin, P.R. (1979). The relationship of the central motor pattern of the feeding cycle of Lymnaea stagnalis. J. Exp. Biol. 80, 137–163. 18. Rose, R.M., and Benjamin, P.R. (1981). Interneuronal control of feeding in Lymnaea stagnalis. II. The interneuronal mechanisms generating feeding cycles. J. Exp. Biol. 92, 203–228. 19. Swandulla, D., and Lux, H.D. (1984). Changes in ionic conductances induced by cAMP in Helix neurons. Brain Res. 305, 115–122. 20. Kemenes, G., S.-Ro´zsa, K., and Carpenter, D.O. (1993). CyclicAMP mediated excitatory responses to leucine enkephalin in Aplysia neurons. J. Exp. Biol. 181, 321–328. 21. Ross, S.T., and Soltesz, I. (2001). Long-term plasticity in interneurons of the dentate gyrus. Proc. Natl. Acad. Sci. USA 98, 8874–8879. 22. Staras, K., Kemenes, G., and Benjamin, P.R. (1999). Electrophysiological and behavioral analysis of lip touch as a compo-
Cellular Mechanism of Long-Term Memory Formation 1069
23.
24.
25. 26.
27.
28.
nent of the food stimulus in the snail Lymnaea. J. Neurophysiol. 81, 1261–1273. Gainutdinov, K.L., Chekmarev, L.J., and Gainutdinova, T.H. (1998). Excitability increase in withdrawal interneurons after conditioning in snail. Neuroreport 9, 517–520. Antonov, I., Antonova, I., Kandel, E.R., and Hawkins, R.D. (2001). The contribution of activity-dependent synaptic plasticity to classical conditioning in Aplysia. J. Neurosci. 21, 6413–6422. Ito, K.I., and Hicks, T.P. (1994). Masked long-term potentiation in kitten visual cortex in vitro. Exp. Brain Res. 100, 175–180. Bortolotto, Z.A., and Collingridge, G.L. (1995). On the mechanism of long-term potentiation induced by (1S,3R)-1-aminocyclopentane-1,3-dicarboxylic acid (ACPD) in rat hippocampal slices. Neuropharmacology 34, 1003–1014. Elliott, C.J.H., and Benjamin, P.R. (1985). Interactions of patterngenerating interneurons controlling feeding in Lymnaea stagnalis. J. Neurophysiol. 54, 1396–1411. Brierley, M.J., Yeoman, M.S., and Benjamin, P.R. (1997). Glutamatergic N2v cells are central pattern generator interneurons of the Lymnaea feeding system: a new model for rhythm generation. J. Neurophysiol. 78, 3396–3407.
A Persistent Cellular Change in a Single Modulatory Neuron Contributes to Associative Long-Term Memory Nicholas G. Jones,1 Ildiko´ Kemenes, Gyo¨rgy Kemenes, and Paul R. Benjamin* Sussex Centre for Neuroscience School of Biological Sciences University of Sussex Falmer Brighton, BN1 9QG United Kingdom
Summary Most neuronal models of learning assume that changes in synaptic strength are the main mechanism underlying long-term memory (LTM) formation [1]. However, we show here that a persistent depolarization of membrane potential, a type of cellular change that increases neuronal responsiveness, contributes significantly to a long-lasting associative memory trace. The use of a model invertebrate network with identified neurons and known synaptic connectivity had the advantage that the contribution of this cellular change to memory could be evaluated in a neuron with a known function in the learning circuit. Specifically, we used the well-understood motor circuit underlying molluscan feeding [2–4] and showed that a key modulatory neuron involved in the initiation of feeding ingestive movements [5] underwent a long-term depolarization following behavioral associative conditioning [6]. This depolarization led to an enhanced single cell and network responsiveness to a previously neutral tactile conditioned stimulus, and the persistence of both matched the time course of behavioral associative memory. The change in the membrane potential of a key modulatory neuron is both sufficient and necessary to initiate a conditioned response in a reduced preparation and underscores its importance for associative LTM. Results and Discussion Although there are a number of examples of long-term, conditioning-induced changes in the electrical properties of neurons [7–10], it has been difficult to establish a relationship between learning-induced cellular changes and behavioral plasticity. The well-described feeding system of the pond snail, Lymnaea stagnalis, allowed us to investigate how a cellular change in an identified modulatory neuron contributes to associative LTM in a neural network that undergoes well-defined electrical changes after behavioral classical conditioning [6, 11]. Feeding in this snail is a rhythmic motor behavior controlled by an exceptionally well-understood central pat*Correspondence: [email protected] 1 Present address: Sensory Function Group, Centre for Neuroscience, King’s College London, London Bridge, SE1 1UL, United Kingdom.
tern generator (CPG) network [2–5, 12, 13]. In their natural environment, snails carry out rhythmic biting movements and take food from floating pond-weed by using a toothed tongue or radula and are able to form associations between the mechanical and chemical properties of food [14]. In the laboratory, an effective feeding stimulus is sucrose, which we use as the rewarding unconditioned stimulus (US) in our classical conditioning experiments. The focus of the present study is the modulatory CV1a cell, which directly excites the most important feeding CPG interneuron, N1M, (Figure 1) and is thus involved in the initiation and control of feeding movements [5, 15]. Our previous work already has shown that appetitive (reward) classical conditioning leads to electrophysiologically tractable associative LTM in Lymnaea [6, 11]. Here, we show that persistent conditioning-induced depolarization of CV1a makes a key contribution to this associative memory trace. Classical conditioning in intact snails [14] was carried out by pairing touch to the lips (the conditioned stimulus, CS) with sucrose (0.01 M final concentration) in spaced training (15 trials over 3 days, five trials a day, 90 min intertrial intervals). This type of training leads to a longterm memory trace [16] whose electrical correlates were recorded in subsequent electrophysiological experiments. Two measures of behavioral LTM were used [6]. One was the number of conditioned snails showing feeding responses to the touch CS compared with a random control group; the other was the strength of response to the CS (feeding bites per minute) compared with controls. After 15 behavioral training trials, the vast majority of the conditioned animals (22/23, 96%), but only a minority of the random control animals (6/23, 25%), responded to lip touch with an increased feeding rate. The mean feeding response to the lip touch CS in the conditioned group was 5.3 ⫾ 0.6 bites/min, whereas, in the control group, it was ⫺0.9 ⫾ 0.4 bites/min. Both of these parameters of the behavioral response to the CS were significantly greater in the conditioned compared with the control group (p ⬍ 0.0001 [Fisher’s exact test] and p ⬍ 0.01 [t test], respectively). Electrophysiological recording of CV1a in reduced head-brain preparations [6, 11] from trained animals showed an enhanced CV1a response to the CS after conditioning that was part of the electrical correlate of LTM. In preparations made from conditioned animals 1 day after training, touch to the lips induced rhythmic bursts of spike activity, known as fictive feeding, in CV1a and a corecorded feeding motoneuron (Figure 2A, top record). This was different from preparations made from random control animals in which touch induced little or no burst of activity (Figure 2A, bottom record). Statistically, more preparations showed a fictive feeding response to the CS in the conditioned group (6/6, 100%) compared to the control group (2/11, 18%) (Figure 2C, left; Fisher’s exact test, p ⬍ 0.01); the response was also significantly stronger in the conditioned (2.7 ⫾ 0.7 cycles/min) versus the control group (⫺0.9 ⫾ 0.6 cycles/ min) (Figure 2C, right; t test, p ⬍ 0.002). The pattern of
Cellular Mechanism of Long-Term Memory Formation 1065
Figure 1. A Simplified Summary of the Lymnaea Feeding Circuit Synaptic connections are shown between the modulatory Cerebral Ventral 1a (CV1a, [15]) neuron and the feeding central pattern generator (CPG) circuit [27, 28]. CV1a is able to drive the CPG via a monosynaptic excitatory connection with the N1 medial (N1M) CPG neuron [5, 15]. Further types of excitatory and inhibitory synaptic connections occur between the N2 and N3 CPG interneurons (N2d/ v, dorsal or ventral type N2 neuron; N3p/t, phasic or tonic type N3 neuron). The CPG neurons drive rhythmic activity in motoneurons (cell types B1-10) that produce feeding movements via contractions of the muscular feeding apparatus [17]. The bars represent excitatory synaptic connections; the closed circles represent inhibitory synaptic connections.
fictive feeding in CV1a and motoneuron B1 induced by touch after conditioning (Figure 2A) was typical of the electrical activity known to underlie behavioral feeding in the intact system [17], and we use this activity to monitor the systems level expression of the LTM trace in the Lymnaea feeding network [6, 11]. The occurrence of burst patterns in the motoneurons indicated that the CPG network had been activated, because motoneurons only fire rhythmically as a result of synaptic input provided by the CPG [18]. These experiments showed that CV1a always fired during conditioned fictive feeding and was active along with other components of the feeding network that it controls. Thus, the rhythmic activation of CV1a by the CS following behavioral conditioning is part of the global electrophysiological readout of the associative LTM trace. To examine whether changes in CV1a’s electrical properties could be contributing to LTM, we looked in more detail for cellular changes in CV1a cells following training. The most striking change was a persistent depolarization of the CV1a membrane potential. This depolarization increased the probability of CV1a firing in response to the CS. In the example shown in Figure 2A, top record, the CV1a cell has a resting potential of ⫺57 mV in a conditioned snail, and it fires in response to the tactile lip stimulus (the CS). In contrast, in the example of a control snail (Figure 2A, bottom record), the CV1a resting potential is ⫺68 mV, and it shows no response to lip touch. Statistically, the mean CV1a resting potential in the conditioned group (n ⫽ 6) was found to be significantly more depolarized, ⫺56 ⫾ 2 mV, compared with that in the random control group (n ⫽ 11), ⫺67 ⫾ 2 mV (t test, p ⬍ 0.03). Moreover, the level of fictive feeding
response (cycles/min) recorded in the CV1a cells (Figure 2C) was significantly correlated (Pearson’s R, 0.6; p ⬍ 0.01) to CV1a resting potential. The effect of conditioning was therefore to depolarize the resting potential of the CV1a cells by an average of 11 mV and to increase CV1a’s responsiveness and that of the entire feeding CPG to the CS. This depolarization was seen in the CV1a cells in all 6 preparations in this first experiment and in all 26 preparations from conditioned snails in 2 subsequent experiments (combined mean resting potential, ⫺54 ⫾ 1 mV [n ⫽ 32]). This was significantly different from the combined mean control CV1a resting potential in the same three experiments, ⫺65 ⫾ 2 mV (n ⫽ 19, t test, p ⬍ 0.001), so the depolarization of CV1a following conditioning was highly consistent. In further experiments, other electrical changes in CV1a that are known to affect neuronal responsiveness, such as a change in spike threshold, membrane input resistance, and spike frequency in response to current steps of different amplitude, also were compared in preparations from conditioned (n ⫽ 7) and control snails (n ⫽ 8); but, unlike the membrane potential, no significant differences were found in any of these other electrical parameters. Neither were there any differences between the two groups in the frequency of fictive feeding cycles driven by CV1a in response to a steady depolarization of around 25 mV; this finding indicates no difference between the ability of “conditioned” and “control” CV1as to activate the feeding CPG. It seems therefore, that the main lasting effect of conditioning on CV1a was to shift its resting potential nearer to its firing threshold. The change in membrane potential induced by conditioning appears to be restricted to the CV1a cells. We tested if cellular changes occurred in other types of neurons in the feeding system, such as the motoneurons corecorded with CV1a or another important modulatory neuron (the Cerebral Giant Cell, [4]), but found no evidence for electrical changes in these cells. Next, we showed that depolarizing the CV1a cell to artificially mimic the changes brought about by conditioning was sufficient to produce an LTM-like electrophysiological trace in the feeding network. By depolarizing CV1a in preparations from naı¨ve snails to ⫺56 mV by using a second electrode for current injection, we could induce an increase in the fictive feeding response to touch (Figure 2B, top traces) similar to that recorded following conditioning (Figure 2A, top traces). At the recorded (nonmanipulated) CV1a membrane potential of ⫺67 mV (Figure 2B, bottom traces), touch produced no fictive feeding response, and this was similar to the lack of response seen in the random control group of the earlier experiment (Figure 2A, bottom traces). With the membrane potential of a single CV1a set at ⫺56 mV, the majority of naı¨ve preparations (12/14, 86%) showed a fictive feeding response to touch (Figure 2D, left), significantly more (Fisher’s exact test, p ⬍ 0.05) than at ⫺67 mV (6/14, 43%). The response to touch at ⫺56 mV was 1.6 ⫾ 0.5 cycles/min, significantly greater (t test, p ⬍ 0.05) than at ⫺67 mV (0.1 ⫾ 0.3 cycles/min, Figure 2D, right). The ability to mimic the effects of conditioning by CV1a membrane potential manipulation is an important part of the evidence showing that changes in the membrane potential in this specific cell type are contributing to the LTM trace.
Current Biology 1066
Figure 2. Electrophysiological Correlate of the LTM Trace Behavioral classical conditioning produces a CS-induced increase in the systems level electrical activity (fictive feeding response). The conditioning-induced increase in the fictive feeding response was mimicked in naı¨ve preparations by the depolarization of CV1a and was significantly reduced in conditioned preparations by the hyperpolarization of CV1a. (A) Top two traces: the lip touch CS evokes rhythmic fictive feeding activity in CV1a and motoneuron B1 in a preparation from a conditioned animal. Bottom two traces: the lip touch CS evokes no rhythmic fictive feeding activity in a preparation from a control animal. (B) Top two traces: the CS evokes rhythmic fictive feeding activity in CV1a and motoneuron B1 in a preparation from a naı¨ve animal when CV1a is artificially depolarized to ⫺56 mV by using a second current injection electrode. Bottom two traces: the CS evokes no rhythmic fictive feeding activity in the same preparation as in (A) with CV1a at its recorded (i.e., nonmanipulated) membrane potential (MP). (C) Significantly more conditioned than control preparations respond to the CS (*, p ⬍ 0.002, Fisher’s exact test, only the percentage of responding preparations is shown for each group here and in [D] and [F]). The response to the CS, which was calculated as the post minus prestimulus difference in fictive feeding activity [6, 11], is significantly stronger in the conditioned versus the control group (*, p ⬍ 0.01, t test). (D) Significantly more naı¨ve preparations respond to the CS (*, p ⬍ 0.05, Fisher’s exact test), and the response is significantly stronger (*, p ⬍ 0.05, t test) when a CV1a is depolarized compared to recorded membrane potential. (E) The touch CS evokes only weak fictive feeding activity in both CV1as and motoneuron B1 in a preparation from a conditioned animal when both CV1as are hyperpolarized to control membrane potential levels. The dashed lines above the CV1a traces indicate the level of membrane potential before artificial hyperpolarization. (F) Significantly fewer conditioned preparations in which CV1a depolarization was reversed by artificial hyperpolarization respond to the lip touch CS (*, p ⬍ 0.05, Fisher’s exact test), and the response is also significantly weaker in this group than in the group of conditioned preparations with CV1as at recorded MP (*, p ⬍ 0.04, t test).
Cellular Mechanism of Long-Term Memory Formation 1067
Previous work already showed that the effect of tactile lip conditioning did not generalize to other parts of the body [14]. It was also important to show that the effect of CV1a depolarization in mimicking the electrical correlate of LTM was similarly restricted to the original site of the CS application, the lips; otherwise, the effects of changing CV1a membrane potential would be nonspecific, unlike the behavioral change. We did the same type of artificial depolarization of the CV1a as in the previous experiment but targeted an alternative touch site, the tentacle [11, 14] (n ⫽ 11 preparations). As before, the response to touch was compared at either ⫺56 or ⫺67 mV in the same preparation. With tentacle touch, there was no significant difference (t test, p ⫽ 0.6) between fictive feeding responses at ⫺56 mV (0.1 ⫾ 0.4 cycles/min) and ⫺67 mV (⫺0.2 ⫾ 0.4 cycles/min), indicating that the effects of depolarization in naı¨ve animals were confined to the lip touch site. This was confirmed when the data on the effects of lip touch were compared with those on tentacle touch. The mean fictive feeding response difference to lip touch at ⫺56 mV was significantly greater, 1.5 ⫾ 0.5 cycles/min, than the 0.1 ⫾ 0.4 cycles/min recorded for the tentacles at the same CV1a membrane potential (t test, p ⬍ 0.05). The necessity of CV1a depolarization for the electrophysiological correlate of the LTM trace was assessed by experiments in which artificial hyperpolarization of the CV1a cells was carried out in preparations made from conditioned snails. If depolarization of CV1a was important for the memory trace, its reversal by hyperpolarization would significantly reduce the fictive feeding response to the CS. For this experiment, it was necessary that both CV1as (the left and right homologs) were hyperpolarized (Figure 2E) to remove the full effects of conditioning-induced CV1a depolarization from the feeding system. In these double-electrode experiments (Figure 2F), when CV1a depolarization was reversed, both the number of preparations responding to the CS (7/11, 63%) and the conditioned fictive feeding response (1.7 ⫾ 0.4 cycles/min) were significantly reduced (Fisher’s exact test, p ⬍ 0.05; t test, p ⬍ 0.04) compared with conditioned preparations with CV1a cells at recorded membrane potential (14/14 [100%] and 2.6 ⫾ 0.4 cycles/ min). This type of experiment provided evidence that conditioning-induced depolarization of CV1a was necessary for the expression of memory trace. To further strengthen the link between a cellular electrophysiological change and behavioral LTM, we showed that the persistence of conditioning-induced CV1a depolarization matched the duration of both the behavioral and electrophysiological memory trace. Previous behavioral analysis showed that the memory produced by 15 trials is stable between 1 and 4 days but declines to control levels by 21 days posttraining [16]. For a direct comparison with the behavior, we tested CV1a membrane potential and fictive feeding responses to the CS in conditioned and control preparations on each of the first 4 days and at 21 days after conditioning. The data from the first 4 days were combined because ANOVAs showed no differences in these parameters between either the conditioned or the control groups (n ⫽ 3–5 animals per group) tested on each of the 4 days following conditioning. In this 4-day period, the
Figure 3. Time Course of the Cellular Change in CV1a The conditioning-induced cellular change in the feeding modulatory interneuron CV1a shows a similar temporal pattern to the systems level electrical correlate of LTM (fictive feeding). (A) CV1as from the conditioned group show a more depolarized membrane potential in the first 4 days following conditioning compared with CV1a cells from the random control group (*, p ⬍ 0.0001, t test), but, by 21 days, there is no statistical difference in CV1a membrane potential between the two groups. (B) Fictive feeding, the electrophysiological correlate of the behavioral LTM trace, shows a similar temporal pattern with elevated fictive feeding responses to the lip CS in conditioned snails compared with control snails in the first 4 days after conditioning (*, p ⬍ 0.0001, t test), but it shows no difference at 21 days. These results indicate that both the cellular and systems level correlate of the LTM memory trace persist for at least 4 days after conditioning but are lost after 21 days. These data are consistent with previously published behavioral data [16] showing a similar time course for LTM retention and loss.
CV1a membrane potential (Figure 3A) was significantly more depolarized (t test, p ⬍ 0.0001) and the fictive feeding response to the CS (Figure 3B) was significantly stronger (t test, p ⬍ 0.0001) in the conditioned (n ⫽ 14) compared with the control group (n ⫽ 12) (⫺55 ⫾ 1 mV versus ⫺67 ⫾ 2 mV and 2.6 ⫾ 0.4 cycles/min versus ⫺0.9 ⫾ 0.6 cycles/min). In contrast, on the 21st day after training, the conditioned (n ⫽ 6) and control preparations (n ⫽ 6) were very similar in their CV1a resting potential (conditioned, ⫺61 ⫾ 2 mV; control, ⫺65 ⫾ 2 mV) and response to touch (conditioned, 0.2 ⫾ 0.4 cycles/ min; control, 0.2 ⫾ 0.3 cycles/min). A comparison was also made between conditioned preparations from the earlier and later time periods. CV1as from conditioned snails were significantly less depolarized (⫺61 ⫾ 2 mV versus ⫺56 ⫾ 1 mV; t test, p ⬍ 0.02), and the CS-induced fictive feeding responses were significantly weaker (0.2 ⫾ 0.4 cycles per minute versus 2.6 ⫾ 0.4 cycles per minute; t test, p ⬍ 0.001) at 21 days than in the 1- to 4-day time period. These experiments show that retention of the memory trace and its eventual loss follow the same temporal pattern as the electrical changes in the CV1a cells. This would be expected if the depolarization of the CV1as was an essential part of the electrical mechanism underlying behavioral LTM. We have demonstrated that behavioral classical conditioning induced long-lasting neuronal plasticity in a single modulatory neuron type controlling feeding in the snail Lymnaea. We found a persistent conditioning-
Current Biology 1068
induced depolarization in neuron CV1a that brought the neuron closer to the threshold for firing. This depolarization makes CV1a more likely to fire in response to the conditioned stimulus and thus drives the feeding CPG network by the direct synaptic pathway to the N1M CPG interneuron (Figure 1). The conditioning-induced depolarization functionally increases the excitability of CV1a and allows the conditioned stimulus to more effectively drive feeding rhythms. General electrical parameters of the cell (e.g., input resistance) remain unchanged. This may indicate that the intrinsic properties of CV1a are not involved in the depolarization, although there are a number of important examples in which long-lasting depolarization occurs with no underlying net input resistance change [19–21]. The persistent change in CV1a membrane potential may have a presynaptic origin, i.e., it is generated by a persistently enhanced extrinsic input, although this seems unlikely, as synaptic input would have been expected to cause some change in input resistance. Whatever its origin, the change in CV1a membrane potential should be regarded as a key element of the conditioned response, because our experiments showed that its depolarization was both sufficient and necessary for an increased response of the whole feeding system to the CS. But how can a cellular change in a modulatory neuron like CV1a lead to a defined conditioned response to a specific CS? We believe that selectivity is ensured by CV1a being embedded in a small network with a highly specialized function that responds only to a narrow range of food-related sensory inputs. These sensory inputs arise mainly from the lips, which normally come into contact with food during rhythmic feeding [22], unlike other structures, such as the tentacles. Indeed, our previous work showed no generalization of the classically conditioned response from the lips to the tentacles [14], and, here, we provided further experimental evidence for response specificity by showing that artificial depolarization of naı¨ve CV1as selectively increases the response to lip touch. Long-term, conditioning-induced changes in the electrical properties of identified neurons occur in other molluscan models of associative learning. These include changes in membrane potential [7, 23] and input resistance [9, 24]. In vitro-evoked LTP in mammalian cortical and hippocampal slices also had suggested that lasting membrane depolarization might be a mechanism for associative neuronal plasticity [25, 26]. However, our work provides the first example of a persistent conditioning-induced cellular change, a depolarization, in an identified modulatory cell type in a CPG-driven motor system in which the cellular changes can be directly linked to behavior. The evidence showing that the depolarization is both sufficient and necessary for eliciting the electrophysiological correlate of the memory trace provides a particularly strong case for the role of cellular changes in LTM. Acknowledgments Support was provided by the Biotechnology and Biological Sciences Research Council (P.R.B., I.K. and N.G.J.) and the Medical Research Council (G.K. and N.G.J.).
Received: March 31, 2003 Revised: April 28, 2003 Accepted: April 29, 2003 Published: June 17, 2003 References 1. Kandel, E.R. (2001). The molecular biology of memory storage: a dialogue between genes and synapses. Science 294, 1030– 1038. 2. Staras, K., Kemenes, G., and Benjamin, P.R. (1998). Patterngenerating role for motoneurons in a rhythmically active neuronal network. J. Neurosci. 18, 3669–3688. 3. Staras, K., Kemenes, I., Benjamin, P.R., and Kemenes, G. (2003). Loss of self-inhibition is a cellular mechanism for episodic rhythmic behavior. Curr. Biol. 13, 116–124. 4. Straub, V.A., and Benjamin, P.R. (2001). Extrinsic modulation and motor pattern generation in a feeding network: a cellular study. J. Neurosci. 21, 1767–1778. 5. Kemenes, G., Staras, K., and Benjamin, P.R. (2001). Multiple types of control by identified interneurons in a sensory-activated rhythmic motor pattern. J. Neurosci. 21, 2903–2911. 6. Staras, K., Kemenes, G., and Benjamin, P.R. (1999). Cellular traces of behavioral classical conditioning can be recorded at several specific sites in a simple nervous system. J. Neurosci. 19, 347–357. 7. Crow, T.J., and Alkon, D.L. (1980). Associative behavioral modification in Hermissenda: cellular correlates. Science 209, 412–414. 8. Moyer, J.R., Jr., Thompson, L.T., and Disterhoft, J.F. (1996). Trace eyeblink conditioning increases CA1 excitability in a transient and learning-specific manner. J. Neurosci. 16, 5536–5546. 9. Brembs, B., Lorenzetti, F.D., Reyes, F.D., Baxter, D.A., and Byrne, J.H. (2002). Operant reward learning in Aplysia: neuronal correlates and mechanisms. Science 296, 1706–1709. 10. Rosenkranz, J.A., and Grace, A.A. (2002). Dopamine-mediated modulation of odour-evoked amygdala potentials during pavlovian conditioning. Nature 417, 282–287. 11. Jones, N.G., Kemenes, G., and Benjamin, P.R. (2001). Selective expression of electrical correlates of differential appetitive classical conditioning in a feeding network. J. Neurophysiol. 85, 89–97. 12. Straub, V.A., Staras, K., Kemenes, G., and Benjamin, P.R. (2002). Endogenous and network properties of Lymnaea feeding central pattern generator interneurons. J. Neurophysiol. 88, 1569– 1583. 13. Jones, R. (2002). Snails with rhythm. Nat. Rev. Neurosci. 3, 918. 14. Kemenes, G., and Benjamin, P.R. (1989). Appetitive learning in snails shows characteristics of conditioning in vertebrates. Brain Res. 489, 163–166. 15. McCrohan, C.R., and Kyriakides, M.A. (1989). Cerebral interneurones controlling feeding motor output in the snail Lymnaea stagnalis. J. Exp. Biol. 147, 361–374. 16. Kemenes, G., and Benjamin, P.R. (1994). Training in a novel environment improves the appetitive learning performance of the snail, Lymnaea stagnalis. Behav. Neural Biol. 61, 139–149. 17. Rose, R.M., and Benjamin, P.R. (1979). The relationship of the central motor pattern of the feeding cycle of Lymnaea stagnalis. J. Exp. Biol. 80, 137–163. 18. Rose, R.M., and Benjamin, P.R. (1981). Interneuronal control of feeding in Lymnaea stagnalis. II. The interneuronal mechanisms generating feeding cycles. J. Exp. Biol. 92, 203–228. 19. Swandulla, D., and Lux, H.D. (1984). Changes in ionic conductances induced by cAMP in Helix neurons. Brain Res. 305, 115–122. 20. Kemenes, G., S.-Ro´zsa, K., and Carpenter, D.O. (1993). CyclicAMP mediated excitatory responses to leucine enkephalin in Aplysia neurons. J. Exp. Biol. 181, 321–328. 21. Ross, S.T., and Soltesz, I. (2001). Long-term plasticity in interneurons of the dentate gyrus. Proc. Natl. Acad. Sci. USA 98, 8874–8879. 22. Staras, K., Kemenes, G., and Benjamin, P.R. (1999). Electrophysiological and behavioral analysis of lip touch as a compo-
Cellular Mechanism of Long-Term Memory Formation 1069
23.
24.
25. 26.
27.
28.
nent of the food stimulus in the snail Lymnaea. J. Neurophysiol. 81, 1261–1273. Gainutdinov, K.L., Chekmarev, L.J., and Gainutdinova, T.H. (1998). Excitability increase in withdrawal interneurons after conditioning in snail. Neuroreport 9, 517–520. Antonov, I., Antonova, I., Kandel, E.R., and Hawkins, R.D. (2001). The contribution of activity-dependent synaptic plasticity to classical conditioning in Aplysia. J. Neurosci. 21, 6413–6422. Ito, K.I., and Hicks, T.P. (1994). Masked long-term potentiation in kitten visual cortex in vitro. Exp. Brain Res. 100, 175–180. Bortolotto, Z.A., and Collingridge, G.L. (1995). On the mechanism of long-term potentiation induced by (1S,3R)-1-aminocyclopentane-1,3-dicarboxylic acid (ACPD) in rat hippocampal slices. Neuropharmacology 34, 1003–1014. Elliott, C.J.H., and Benjamin, P.R. (1985). Interactions of patterngenerating interneurons controlling feeding in Lymnaea stagnalis. J. Neurophysiol. 54, 1396–1411. Brierley, M.J., Yeoman, M.S., and Benjamin, P.R. (1997). Glutamatergic N2v cells are central pattern generator interneurons of the Lymnaea feeding system: a new model for rhythm generation. J. Neurophysiol. 78, 3396–3407.