Proc. Nati. Acad. Sci. USA Vol. 85, pp. 2373-2377, April 1988 Neurobiology Long-term memory in Aplysia modulates the total number of varicosities of single identified sensory neurons (learning/habituation/sensitization/structural correlates/synapse) CRAIG H. BAILEY*t AND MARY CHEN* *Center for Neurobiology and Behavior, The New York State Psychiatric Institute, 722 West 168th Street, New York, NY 10032; and tDepartments of Anatomy and Cell Biology and Psychiatry, College of Physicians and Surgeons, Columbia University, New York, NY 10032 Communicated by Eric R. Kandel, January 11, 1988 ABSTRACT The morphological consequences of long- identified sensory neuron synapses that occur following term habituation and sensitization of the gill withdrawal reflex long-term habituation and sensitization. in Aplysia californica were explored by examining the total In the present study we have used horseradish peroxidase number of presynaptic varicosities of single identified sensory (HRP) to label the presynaptic terminals (varicosities) of neurons (a critical site of plasticity for the biochemical and sensory neurons. We have then used serial 20-num "slab- biophysical changes that underlie both types of learning) in thick" sections to count the total number of varicosities per control and behaviorally trained animals. Sensory neurons sensory neuron in control and behaviorally modified ani- from habituated animals had 35% fewer synaptic varicosities mals. We have found that sensory neurons from long-term than did sensory neurons from control animals. In contrast, habituated animals have fewer varicosities than do sensory sensory neurons from sensitized animals had twice as many neurons from control animals and that sensory neurons from varicosities per sensory neuron compared to controls, as well long-term sensitized animals have more varicosities. as enlarged neuropil arbors. These changes suggest that modulation of synapse number may play a role in the main- METHODS tenance of long-term memory. Sensory neurons from three groups of animals were exam- An issue central to the study of learning and memory is the ined in this study: control (untrained) animals (n = 16) and functional relationship between synaptic structure and the animals trained for long-term habituation (n = 10) or long- prolonged changes in synaptic effectiveness that accompany term sensitization (n = 10). long-term behavioral modification. It has become clear that Animals were trained for long-term habituation by the synaptic connections in the mature nervous system of both protocol of Carew et al. (19) and for long-term sensitization vertebrates and invertebrates exhibit a substantial degree of by the protocol of Pinsker et al. (20). In each case the structural plasticity with normal operation (1) and with behavioral memory for the task is retained for several learning (2-4). Thus, changes in environmental complexity, weeks. Animals were individually housed for a minimum of social structure, or training in the intact animal (5-11), as 5 days in circulating seawater before behavioral training. To well as electrical stimulation in vitro (12, 13) and in vivo assess their responsiveness, we delivered two jets of seawa- (14-16), have been shown to alter both the number of ter to the siphon with a WaterPik. Animals were accepted for synapses and their structure. The role these structural alter- the experiment if the mean duration of their first two test ations may play in learning and memory has been difficult to responses (interstimulus interval, 30 sec) was 10 sec or assess in complex systems, where the contribution of indi- longer. The scores (total of 10 stimuli) were ranked, and the vidual synapses to the learning process is largely unknown. animals were randomly distributed into a control (untrained) The neural parsimony offered by several higher invertebrate group and groups for long-term habituation and long-term preparations has successfully bridged this gap and has sensitization. There were no significant differences among proven useful for the cellular analysis of behavioral prob- the groups before training. Long-term habituation was pro- lems. We have used one such model system-the gill and duced by giving animals 10 sessions of habituation training siphon withdrawal reflex of Aplysia californica-to study per day for 10 days. Long-term sensitization was produced two simple forms of nonassociative learning (habituation and by giving animals training sessions on four consecutive days. sensitization) both of which can exist in a short-term form Each session consisted of exposure to four electrical stimuli lasting up to 1 hr (17, 18) and a long-term form lasting >3 (100 mA for 2 sec), each separated by 1.5 hr. All electrical weeks (19, 20). Several aspects of the biophysical and stimuli were delivered to the neck region through bipolar biochemical mechanisms that underlie short-term habitua- capillary electrodes. tion and sensitization are understood and involve changes in Retention of habituation and sensitization was tested 1 day synaptic effectiveness produced by modulation of the Ca2+ after the completion of the respective training sessions. The current at a common locus-the presynaptic terminals of behavioral performance of animals-control, long-term ha- identified mechanoreceptor sensory neurons (21-24). Less- bituated and long-term sensitized-was estimated by com- well characterized are the morphological mechanisms that paring their behavioral scores 24 hr after the completion of underlie habituation and sensitization. In particular, it is not training with their pretraining scores. These ratios (after/ known whether structural alterations at sensory neuron training/pretraining) were as follows: long-term habituated synapses contribute to the transition of the short-term form animals were significantly different from their controls [0.11 to one of longer duration. To address these issues, we ± 0.04 (mean ± SEM, n = 10) vs. 1.18 ± 0.12 (mean + examined the nature and extent of morphological changes at SEM, n = 8); t= 9.679 sec; P < 0.0001, two-tailed test] and long-term sensitized animals were significantly different The publication costs of this article were defrayed in part by page charge from their controls [9.08 ± 1.3 (mean ± SEM, n = 10) vs. payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact. Abbreviation: HRP, horseradish peroxidase. Downloaded by guest on September 27, 2021 2373 2374 Neurobiology: Bailey and Chen Proc. Natl. Acad Sci. USA 85 (1988) 0.95 + 0.07 (mean ± SEM, n = 8); t = 5.64; P < 0.0001, term habituation. To maximize the chances of finding la- two-tailed test]. beled sensory neuron processes in thin sections, multiple Within 48 hr of the end of training, animals were anesthe- cells in each animal were injected with HRP, but it was not tized by intracoelomic injection of isotonic MgCl2 (25). The feasible to examine if long-term training had any effect on abdominal ganglion was removed from each animal and the total number of presynaptic varicosities per sensory transferred to a solution of supplemented artificial seawater neuron. containing a high concentration of Mg2+ (200 mM) and a low In the present study, we have addressed this question concentration of Ca2+ (1 mM) to block synaptic transmis- directly by quantitatively analyzing the total axonal arbor of sion during pinning and desheathing. After desheathing, the single HRP-filled sensory neurons. The total number of ganglion was bathed in seawater with a normal concentration varicosities per sensory neuron are summarized in Fig. 1 and of Mg2+ (55 mM) and Ca2+ (10 mM) for a minimum of 30 show a trend parallel to our initial observations on active- min. To label presynaptic varicosities, a single sensory zone morphology. However, the present result is even more neuron from each animal was chosen at random, identified dramatic, because it involves modulation of each neuron's (26), and intrasomatically pressure-injected with HRP (type entire synaptic field. Sensory neurons from control animals VI, Sigma) at a concentration of 20 mg/ml in distilled water. had on average 1291 varicosities; this was reduced 35% to Following a 2-hr incubation period to allow HRP to fill the 836 varicosities per sensory neuron in long-term habituated neuropil arbor of the sensory neuron as well as its axon in animals. In contrast, sensory neurons from long-term sensi- the siphon nerve, ganglia were fixed, histochemically pro- tized animals had approximately twice as many varicosities cessed, and embedded in Epon (27). Each sensory neuron (2697 varicosities) compared to their controls (1320 varicos- was completely reconstructed by serial 20-gm slab-thick ities). sections, and the total number of HRP-labeled varicosities The alterations in synapse number that accompany long- for each cell was counted through an experimenter-blind term memory in Aplysia suggest a number of parallels with procedure using high-resolution light microscopy (27). A the sequence of structural events that has been described total of 36 sensory neurons were analyzed in this fashion. throughout the normal development ofthe nervous system in The axonal branching pattern of individual mechanorecep- other animals. For example, the reduction in varicosity tor sensory neurons has been described, and the types of number during long-term habituation may be similar to the varicosities that could be found along their neuropil arbor phenomenon of synapse elimination. In contrast, the dra- have been characterized (27). Although the branching pat- tern is quite complex, relatively few types of varicosities are 3000 r found. These include (0) small relatively symmetrical bead- like varicosities along fine branches; (ii) more eccentrically placed varicosities; (iii) expansions that occur at some but not all branch points; and (iv) terminal varicosities that range 2500 k- from spherical to club-shaped. The first two categories account for =90% of the total number of varicosities. All of c these are easily and reliably identified in 20-,um-thick sec- tions and can be accurately counted by using long-working- z 0 2000 [- distance high-resolution objectives.