Proc. Nati. Acad. Sci. USAVol. 85, pp. 9356-9359, December 1988Neurobiology

Long-term sensitization in Aplysia increases the number ofpresynaptic contacts onto the identified gill motor neuron L7

(learning/invertebrate/long-term memory/structural correlates/synapse number)

CRAIG H. BAILEY*t AND MARY CHEN**Center for Neurobiology and Behavior, The New York State Psychiatric Institute, 722 West 168th Street, and tDepartments of Anatomy and Cell Biologyand Psychiatry, College of Physicians and Surgeons, Columbia University, New York, NY 10032

Communicated by Eric R. Kandel, August 18, 1988

ABSTRACT We have used the gill and siphon withdrawalreflex of Aplysia to study the morphological basis of thepersistent synaptic plasticity that underlies long-term sensiti-zation. One critical locus for storage of the memory forsensitization is the set of monosynaptic connections betweenidentified siphon sensory neurons and gill and siphon motorneurons. To complement previous morphological studies of thepresynaptic terminals of identified sensory neurons, we exam-ined the effects of long-term sensitization on the structure of anidentified postsynaptic target-the gill motor neuron L7. Wefound an increase in the frequency, size, and vesicle comple-ment of presynaptic contacts onto L7 processes in sensitizedcompared to control animals. Combined, these data indicate astriking increase in the percentage ofthe surface area ofL7 thatis occupied by synaptic contacts after long-term training. Theseresults are consistent with our observations that sensitizationproduces an increase in the synapses that the sensory neuronsmake on their target cells and provide additional support forthe hypothesis that changes in synapse number may representa mechanism underlying long-term memory.

How information is stored in the brain is a question centralto both neurobiology and psychology. Recent studies onseveral higher invertebrates have enhanced our knowledgeabout the synaptic loci and mechanisms that are involved inthe acquisition and retention of various elementary forms oflearning and memory (1, 2).One such model system-the gill and siphon withdrawal

reflex ofAplysia californica-has proven to be a particularlyaccessible behavioral preparation for cellular and molecularstudies of learning and memory (3). This reflex can bemodified by a simple type of nonassociative learning-sensitization-the memory of which can exist in a short-termform lasting minutes to hours (4) and a long-term formpersisting for days to week (5, 6). Previous work (for review,see ref. 7) has shown that the short-term memory forsensitization involves an enhancement in the effectiveness ofthe synapses between siphon sensory neurons and gill andsiphon motor neurons (8-11). More recent electrophysiolog-ical studies examining the amplitude of the monosynapticexcitatory postsynaptic potentials between sensory neuronsand the gill motor neuron L7 have shown that the samesynaptic loci are involved in the storage of the long-termmemory for sensitization (6). This enduring change in thefunctional expression of sensory to follower cell connectionsis accompanied by two classes of morphological change: (i)increases in active zone morphology at identified sensoryneuron synapses (12) and (ii) an increase in the total numberof presynaptic varicosities per sensory neuron (13, 14).

In the present study we have examined further the mor-phological consequences of long-term sensitization by deter-

mining its effect on a postsynaptic target of the sensoryneurons-the identified gill motor neuron L7. We report herethat long-term sensitization is accompanied by an increase inthe frequency, size, and vesicle complement of presynapticactive zones onto the dendritic tree of L7 with a resultantincrease in the percentage of its receptive surface that isoccupied by synaptic contact.

METHODSAnimals were trained for long-term sensitization by theprotocol of Pinsker et al. (5). Animals were individuallyhoused for a minimum of5 days in circulating seawater beforebehavioral training. To assess their responsiveness, wedelivered twojets of seawater to the siphon with a Water Pik.Animals were accepted for the experiment if the meanduration of their first two test responses (interstimulusinterval, 30 sec) was 10 sec or longer. The scores (total of 10stimuli) were ranked and the animals were randomly distrib-uted into a control (untrained) group and a group for long-term sensitization. There were no significant differencesbetween the groups before training. Long-term sensitizationwas produced by giving animals training sessions on fourconsecutive days. Each session consisted of exposure to fourelectrical stimuli (100 mA for 2 sec), each separated by 1.5 hr.All electrical stimuli were delivered to the neck regionthrough bipolar capillary electrodes.

Retention of sensitization was tested 1 day after comple-tion of the last training session. The behavioral performanceofanimals-control and long-term sensitized-was estimatedby comparing their behavioral scores 24 hr after the comple-tion of training with their pretraining scores. The ratios aftertraining/pretraining were as follows: long-term sensitizedanimals were significantly different from control [5.3 ± 1.1(mean ± SEM, n = 4) vs. 0.91 ± 0.12 (mean ± SEM, n = 4);t = 4.01, P < 0.011.Within 48 hr of the end of training, animals were anesthe-

tized by intracoelomic injection of isotonic MgCl2. Theabdominal ganglion connected to the gill and abdominal aortawas removed from each animal and transferred to a solutionof supplemented artificial seawater containing a high Mg2+content (200 mM) and a low Ca2' content (1 mM) to blocksynaptic transmission during pinning and desheathing. Afterdesheathing, the ganglion was bathed in seawater with anormal Mg2' (55 mM) and Ca2' (10 mM) content for aminimum of 30 min before we attempted to identify L7. L7was identified on the basis of its size, location, electrophys-iological characteristics, the occurrence of stereotypic spon-taneous synaptic input, and, most importantly, the unique gillmovements (15) and contractions ofthe wall ofthe abdominalaorta (16) it produces when caused to fire by the intracellularinjection of current. Once L7 was identified, its soma wasslowly pressure-injected with horseradish peroxidase (typeVI, purchased from Sigma) at a concentration of4% in 0.5 Mpotassium chloride. After a 20-hr incubation period to allow

9356

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

2, 2

021

Proc. Natl. Acad. Sci. USA 85 (1988) 9357

the horseradish peroxidase to fill the neuropil arbor of L7 aswell as its axons in the branchial, genital-pericardial, andsiphon nerves, ganglia were fixed, histochemically pro-cessed, and embedded in Epon (17). Each L7 was seriallysectioned using 20-Itm slab-thick sections.To analyze the fine structure of the L7 processes and

quantitate the frequency of presynaptic contacts onto itsdendritic arbor, 20 slab-thick sections from similar regions ofthe left hemiganglion neuropil in control (n = 4) and exper-imental (n = 4) animals were re-embedded and thin-sectioned. Complete sets of 100-200 serial thin sections weretaken from each re-embedded thick section. All horseradishperoxidase-labeled L7 profiles on each of 10 equally spacedthin sections from each re-embedded slab-thick section werephotographed and analyzed through a blind procedure. Toexamine in detail the fine structure and extent of synapticinput onto the postsynaptic surface of each neuron, weanalyzed a total of 1267 L7 profiles taken from eight animals.

Quantitation of L7 processes and presynaptic contactsonto them was made using a Bioquant II digitizing tablet(R&M Biometrics, Nashville, TN) interfaced with an AppleIle microcomputer running the Bioquant II morphometrysoftware. The surface areas of relevant structures werecomputed by multiplying the thickness of the section (esti-mated to average 0.1 aum as judged by interference color) bythe linear values obtained by tracing the perimeter (L7profiles) or length (presynaptic active zones) on prints en-larged to a final magnification of 60,000. The number ofvesicles associated with each active zone was determined bycounting the total number of vesicle profiles in each sectionthat fell within 30 nm of the presynaptic active zone mem-brane (12, 18).

RESULTSCell L7 is a primary motor neuron for the gill and mantle shelfin Aplysia and is involved in the defensive reflex of thesemantle organs. It receives both direct and indirect (by way ofinterneurons) input from the mechanoreceptor sensory neu-rons of the siphon skin and these connections are critical locifor the changes in synaptic effectiveness that underlie be-havioral sensitization (3).The three-dimensional morphology of L7 has been exam-

ined by Winlow and Kandel (19). Using cobalt chloride as amarker substance, this light microscopic study revealed thedistribution and extent of the major processes of L7 withinthe abdominal ganglion. Since the finest ramifications of thecell were most likely not visualized, this study primarilyprovided a low-resolution view of the architecture of L7.To address this issue in more detail, we have begun a

combined light and electron microscopic examination of themorphology of L7 using horseradish peroxidase and serial20-,um slab-thick sections of Epon-embedded material. Thisapproach has permitted the visualization of the finest pro-cesses of the neuron and thereby facilitated quantitativeanalysis of the synaptic organization of the dendritic arbor ofL7.Long-Term Sensitization Increases the Number ofPresynap-

tic Contacts onto the Dendritic Tree of L7. In the present studywe have examined the effects of long-term sensitization onthe fine structure of the dendritic field ofL7 and on the extentof the synaptic input it receives. Using random thin sectionstaken through the horseradish peroxidase-labeled secondaryand tertiary branches ofthe proximal dendritic tree of L7, wequantitated changes in the receptive surface of L7 in controland behaviorally modified animals.We began by examining the incidence ofpresynaptic active

zones onto L7 processes in both behavioral groups. Quanti-tative ultrastructural analysis of 1267 L7 profiles revealed anincrease in the frequency of presynaptic contacts onto L7

Table 1. Number of L7 profiles, number of presynaptic profileswith active zones onto L7 profiles, and ratio of L7 profilesreceiving presynaptic contact in sensitized and control animals

No. Ratio of L7presynaptic profiles with

L7 profiles, active zones presynaptic contact,Animal n onto L7 %

Sensitized1 190 89 472 275 130 473 148 54 374 75 38 51

Control1 78 13 172 124 22 183 133 29 224 244 55 23

processes in sensitized compared to control animals [0.45 ±0.03 (mean ± SEM, n = 4) vs. 0.197 ± 0.02 (n = 4), t = 7.43,P < 0.001; Table 1]. The number of synaptic contacts persquare micrometer of L7 surface area increases (1.72 ± 0.06vs. 0.61 ± 0.05, t = 13.36, P < 0.001) as does the incidenceof multiple synaptic contacts onto the same postsynapticprofile (0.06 ± 0.008 vs. 0.01 ± 0.004, t = 4.8, P < 0.01; Fig.1). Both the size (0.49 ± 0.01 Am vs. 0.4 ± 0.02 Am, t = 3.1,P < 0.05) and vesicle complement (4.1 ± 0.1 vs. 2.8 ± 0.2,t = 5.9, P < 0.01) of the active zones at these presynapticcontacts onto L7 dendritic processes are larger in sensitizedanimals than in controls (Table 2). Combined, these dataindicate a striking increase in the percentage of the surfacearea of L7 that is occupied by synaptic contacts afterlong-term training (0.084 ± 0.002 vs. 0.025 ± 0.003, t = 16.86,P < 0.001; Fig. 2).*

In addition to quantitating the effects of long-term sensi-tization on the nature and extent ofpresynaptic contacts ontoL7, we have also begun to characterize the morphology ofthepostsynaptic components at these synapses. We have foundan increase in the frequency of small (<0.5 nm), spine-likepostsynaptic L7 processes at these synapses in sensitizedcompared to control animals (0.72 ± 0.06 vs. 0.40 ± 0.09, t- 3.04, P < 0.05; Fig. 1C).

DISCUSSIONPrevious studies of long-term sensitization of the gill andsiphon withdrawal reflex in Aplysia have indicated that acommon locus-the synapses between identified sensoryneurons and follower cells-undergoes both functional andstructural alterations. Electrophysiological examination hasshown a 2-fold increase in the amplitude of monosynapticexcitatory postsynaptic potentials between siphon mechano-receptors and the gill motor neuron L7 in sensitized com-pared to control animals (6). Similarly, quantitative morpho-logical analysis of the identified presynaptic varicosities ofsensory neurons has shown an increase in the number, size,and vesicle complement of their active zones (12) as well asa near doubling ofthe total number of varicosities per sensoryneuron (13, 14). The nature and extent of both the functional

tAs one control for possible nonspecific effects of the trainingprotocol, we examined the incidence, size, and vesicle complementof presynaptic active zones onto unidentified postsynaptic profilesin regions of the neuropil contiguous to the dendritic arbor of L7.Approximately 1500 ,um2 of tissue was sampled and no significantdifferences were found with respect to the number per squaremicrometer (0.068 ± 0.003 vs. 0.074 ± 0.003), length (0.28 ± 0.007Aum vs. 0.34 ± 0.06 Aum) or vesicle complement (1.8 ± 0.14 vs. 2.4+ 0.5) of presynaptic active zones onto unlabeled postsynapticprofiles between sensitized and control animals.

Neurobiology: Bailey and Chen

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

2, 2

021

9358 Neurobiology: Bailey and Chen

Vi sv Qj. X

FIG. 1. Increased synaptic contact onto L7 processes afterlong-term sensitization training. The incidence of multiple presynap-tic contacts (*) onto the same L7 profile is increased in sensitizedcompared to control animals. This convergence can occur onintermediate-size processes (A) or on smaller profiles (B and C).Long-term sensitization also increases the frequency of narrow,spine-line postsynaptic L7 processes at these synapses (C). Allprofiles belonging to L7 are easily identified by a blanket of electron-dense horseradish peroxidase reaction product. (Bar = 0.5 .tm.)

Table 2. Length of active zones and number of associatedvesicle profiles in presynaptic terminals contacting L7processes in sensitized and control animals

No. vesicleActive zones Active zone profiles associated

Animal analyzed, n length, ,um with each active zoneSensitized

1 89 0.486 ± 0.02 4.0 ± 0.32 130 0.523 ± 0.02 4.3 ± 0.53 54 0.466 0.03 4.3 ± 0.44 38 0.473 ± 0.04 3.9 ± 0.3

Control1 13 0.352 ± 0.03 2.5 ± 0.42 22 0.367 ± 0.04 2.3 ± 0.53 29 0.439 ± 0.02 3.1 ± 0.34 55 0.450 ± 0.03 3.1 ± 0.2Results represent mean ± SEM.

and structural changes at identified sensory neuron synapsesare consistent with the known behavioral efficiency oflong-term sensitization (5, 6).

In the present study we have examined the effects oflong-term sensitization on the structure of an identifiedpostsynaptic target ofthe sensory neurons-the identified gillmotor neuron L7. Our initial quantitative survey was focusedprimarily on the extent of synaptic contacts onto the recep-tive surface of L7. The results indicate a dramatic increase inthe number and percentage of occupancy of synapses ontothe dendritic tree of L7. In addition, there appears to be anincreased convergence ofpresynaptic contacts onto the samepostsynaptic profile after long-term training.The increase in presynaptic input during long-term sensi-

tization is accompanied by an increased percentage of syn-apses occurring on small (<0.5 ,um) L7 processes. This couldbe due to a shift in the position of synapses onto the L7neuropil arbor or perhaps to an increased outgrowth in thenumber of small processes from L7. Preliminary qualitativelight microscopic observations (data not presented) supportthe second possibility, with an apparent increase in focaloutpocketing of finger-like spines from L7 in sensitizedcompared to control animals. Further analysis will be nec-essary to determine whether such an increase is a consistentpostsynaptic correlate of long-term sensitization in Aplysiaand whether it contributes to an increase in postsynapticsurface membrane area, as has been reported in vertebrates(refs. 20, 21; for review, see ref. 22).That the increases in morphological features ofpresynaptic

contacts onto L7 are likely to result, at least in part, from thebehavioral training protocol is supported by our analysis ofactive zones at presynaptic terminals that synapse ontounlabeled postsynaptic profiles. If, for instance, long-termtraining was producing a general, nonspecific increase in thenumber or size of presynaptic contacts, such an increaseshould be evident in a quantitative analysis of neuropilcontiguous to the territory of L7. By contrast, our datasuggest that the alterations in presynaptic features are dif-ferentially expressed at a known postsynaptic locus for thechanges in efficacy that underlie sensitization-the gill motorneuron L7. This finding provides additional support for thehypothesis that the morphological alterations we have ob-served share some specificity to the learning experience.Although the neurons of origin for the unlabeled presyn-

aptic terminals contacting L7 are not known in the presentstudy, it is attractive to think that some percentage of thesecontacts belong to sensory neurons. This would be consistentwith the results of biophysical studies indicating a trendtoward a higher percentage of sensory cell connections ontoL7 after long-term training (6). To address directly thestructural correlates of this will require simultaneous exam-

Proc. Natl. Acad. Sci. USA 85 (1988)

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

2, 2

021

Proc. Natl. Acad. Sci. USA 85 (1988) 9359

10

LI)Z0>J V

)OD

00D0

8

4

2

p<O.OOI

CONTROL(N=4)

ENSITIZI(N=4)

FIG. 2. Percentage of the surface area of L7 that is occupied bypresynaptic contacts in control and long-term sensitized animals.Long-term training increases both the frequency and size of presyn-aptic contacts onto L7 with a resultant increase in the percentage ofreceptive surface of L7 that is occupied by synaptic contact.

ination of unequivocally identified pre- and postsynapticcomponents of the sensory to L7 synapse. Once the mor-

phological equivalents for this monosynaptic connectionhave been determined to naive animals, a more-refinedquantitative comparison between morphological, biophysi-cal, and biochemical changes during long-term sensitizationwill be possible.The results of the present study are consistent with our

earlier observation that long-term sensitization produces an

increase in the synapses that the sensory neurons make on

their target cells (12-14). Detection of these increases afterindependent examination of both pre- and postsynaptichalves of the sensory to L7 connection strengthens theevidence that the morphological alterations occur at the samesynaptic locus as do the functional changes that have beendescribed after long-term sensitization (6). These data alsosuggest that the increase in synaptic effectiveness thatunderlies long-term sensitization may be, at least in part,presynaptic in nature; this notion would be consistent withthe results of a recent quantal analysis of long-term facilita-tion at sensory to L7 synapses grown in culture (23).The increase in the number of synapses that we have

observed during long-term sensitization in Aplysia is similarto results from studies in the mammalian nervous system thatindicate increases in synapse number after various forms ofenvironmental manipulation and learning tasks (refs. 20, 21,24-28, for review, see refs. 22 and 29). This convergence ofresults across species suggests that synapse formation maybe a highly conserved feature accompanying long-term be-havioral modification (30). Little is known concerning themolecular events that underlie these anatomical changes.Two mechanisms that have been proposed (for discussion,see ref. 31) are the modification of preexisting proteins (32,33) or differential gene expression and the synthesis of newproteins (1, 31, 34, 35). Some support for the secondpossibility has come from recent studies in Aplysia wherecomputer-assisted quantitative two-dimensional gel analysishas identified specific proteins in which [35S]methioninelabeling is altered after long-term sensitization training (36).Whatever the underlying biochemical mechanisms, the mor-

phological correspondence between the invertebrate and

vertebrate studies indicates that learning may resemble aform of neuronal growth across a broad segment ofthe animalkingdom and suggests that the stability required for a long-term memory may be provided by alterations in the numberof synaptic connections.

We thank Dr. E. R. Kandel for critical review of the manuscript,K. Hilten and L. Katz for the preparation of figures, and H. Ayersfor typing the manuscript. This work was supported by GrantMH37134 from the National Institute of Mental Health, Scope B ofGrant GM23540 from the National Institute of General MedicalSciences, and the McKnight Endowment Fund for Neuroscience.

1. Kandel, E. R. & Schwartz, J. H. (1982) Science 218, 433-442.2. Alkon, D. L. (1984) Science 226, 1037-1045.3. Kandel, E. R. (1976) Cellular Basis ofBehavior (Freeman, San

Francisco).4. Pinsker, H., Kupfermann, I., Castellucci, V. F. & Kandel,

E. R. (1970) Science 167, 1740-1742.5. Pinsker, H. M., Hening, W. A., Carew, T. J. & Kandel, E. R.

(1973) Science 182, 1039-1042.6. Frost, W. N., Castellucci, V. F., Hawkins, R. D. & Kandel,

E. R. (1985) Proc. Natl. Acad. Sci. USA 82, 8266-8269.7. Kandel, E. R., Klein, M., Castellucci, V. F., Schacher, S. &

Goelet, P. (1986) Neurosci. Res. 3, 498-520.8. Castellucci, V. F. & Kandel, E. R. (1974) Proc. Natl. Acad.

Sci. USA 71, 5004-5008.9. Klein, M. & Kandel, E. R. (1978) Proc. Nati. Acad. Sci. USA

75, 3512-3516.10. Castellucci, V. F., Kandel, E. R., Schwartz, J. H., Wilson,

F. D., Nairn, A. C. & Greengard, P. (1980) Proc. Nati. Acad.Sci. USA 77, 7492-74%.

11. Klein, M. & Kandel, E. R. (1980) Proc. Natl. Acad. Sci. USA77, 6912-6916.

12. Bailey, C. H. & Chen, M. (1983) Science 220, 91-93.13. Bailey, C. H. & Chen, M. (1988) Proc. Natl. Acad. Sci. USA

85, 2373-2377.14. Bailey, C. H. & Chen, M. (1989) J. Neurosci., in press.15. Kupfermann, I. & Kandel, E. R. (1969) Science 164, 847-850.16. Alevizos, A. & Koester, J. (1986) Soc. Neurosci. Abstr. 12,499.17. Bailey, C. H., Thompson, E. B., Castellucci, V. F. & Kandel,

E. R. (1979) J. Neurocytol. 8, 415-444.18. Bailey, C. H. & Chen, M. (1988) J. Neurosci. 8, 2452-2459.19. Winlow, W. & Kandel, E. R. (1976) Brain Res. 112, 221-249.20. Greenough, W. T. & Volkmar, F. R. (1973) Exp. Neurol. 40,

491-504.21. Chang, F.-L. F. & Greenough, W. T. (1982) Brain Res. 232,

283-292.22. Greenough, W. T. (1984) Trends Neurosci. 7, 229-233.23. Dale, N., Schacher, S. & Kandel, E. R. (1988) Science 239,

282-285.24. Globus, A., Rosenzweig, M. R., Bennett, E. L. & Diamond,

M. C. (1973) J. Comp. Physiol. 82, 175-181.25. Garey, L. J. & Pettigrew, J. D. (1974) Brain Res. 66, 165-172.26. Greenough, W. T., Juraska, J. M. & Volkmar, F. R. (1979)

Behav. Neural Biol. 26, 287-297.27. Turner, A. M. & Greenough, W. T. (1985) Brain Res. 329, 195-

203.28. Camel, J. E., Withers, G. S. & Greenough, W. T. (1986)

Behav. Neurosci. 100, 810-813.29. Greenough, W. T. & Chang, F.-L. F. (1985) in Synaptic Plas-

ticity, ed. Cotman, C. W. (Guilford, New York), pp. 335-372.30. Greenough, W. T. & Bailey, C. H. (1988) Trends Neurosci. 11,

142-147.31. Goelet, P., Castellucci, V. F., Schacher, S. & Kandel, E. R.

(1986) Nature (London) 322, 419-422.32. Crick, F. (1984) Nature (London) 312, 101.33. Saitoh, T. & Schwartz, J. H. (1985) J. Cell Biol. 100, 835-842.34. Agranoff, B. W. (1970) in Protein Metabolism of the Nervous

System, ed. Lajtha, A. (Plenum, New York), pp. 533-543.35. Davis, H. P. & Squire, L. R. (1984) Psychol. Bull. 96, 518-559.36. Castellucci, V. F., Kennedy, T. E., Kandel, E. R. & Goelet, P.

(1988) Neuron 1, 321-328.

Neurobiology: Bailey and Chen

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

2, 2

021