sites of plasticity in the neural circuit mediating tentacle

Sites of Plasticity in the Neural CircuitMediating Tentacle Withdrawal in the SnailHelix aspersa: Implications for BehavioralChange and Learning KineticsSteven A. Prescott1 and Ronald ChaseDepartment of Biology, McGill UniversityMontreal, Quebec, H3A 1B1, Canada

Abstract

The tentacle withdrawal reflex of thesnail Helix aspersa exhibits a complexcombination of habituation andsensitization consistent with thedual-process theory of plasticity.Habituation, sensitization, or a combinationof both were elicited by varying stimulationparameters and lesion condition. Analysis ofresponse plasticity shows that the late phaseof the response is selectively enhanced bysensitization, whereas all phases aredecreased by habituation. Previous datahave shown that tentacle withdrawal ismediated conjointly by parallelmonosynaptic and polysynaptic pathways.The former mediates the early phase,whereas the latter mediates the late phase ofthe response. Plastic loci were identified bystimulating and recording at different pointswithin the neural circuit, in combinationwith selective lesions. Results indicate thatdepression occurs at an upstream locus,before circuit divergence, and is thereforeexpressed in all pathways, whereasfacilitation requires downstream facilitatoryneurons and is selectively expressed inpolysynaptic pathways. Differentialexpression of plasticity between pathwayshelps explain the behavioral manifestationof depression and facilitation. A simplemathematical model is used to show howserial positioning of depression andfacilitation can explain the kinetics of

dual-process learning. These resultsillustrate how the position of cellularplasticity in the network affects behavioralchange and how forms of plasticity caninteract to determine the kinetics of the netchanges.

Introduction

An important goal in studies of learning andmemory is to relate cellular plasticity to behavioralchange. To achieve this, one must consider what ishappening at the neural network level. The impor-tance is twofold. First, different network elementsmay mediate different components of the behav-ioral response, meaning that plasticity in differentnetwork elements will confer learning in differentcomponents of the behavior. Second, relatinglearning kinetics at the cellular level to learningkinetics at the behavioral level may be complicatedby interactions between learning processes at thenetwork level. The tentacle withdrawal reflex ofthe snail Helix aspersa is plastic (Prescott andChase 1996) and is mediated by a relatively simpleneural network (Prescott et al. 1997), which makesit a suitable preparation in which to investigate therelationship between cellular plasticity and behav-ioral change.

Our first goal was to localize plasticity withinthe neural network to help relate cellular plasticityto behavioral change. We have shown previouslythat excitation caused by a brief mechanical stimu-lus to the tip of the tentacle is transformed into aprolonged neuronal discharge in interneurons ofthe tentacle ganglion (Prescott et al. 1997). Pro-longed activity is transmitted through the polysyn-aptic pathway and is thought to be a major deter-minant of muscle response duration based on cor-relative data and direct motoneuron stimulation

1Corresponding author. Present address: Department ofPharmacology and Therapeutics, McGill University, Mon-treal, Quebec, H3G 1Y6, Canada.

LEARNING & MEMORY 6:363–380 © 1999 by Cold Spring Harbor Laboratory Press ISSN1072-0502/99 $5.00

&L E A R N I N G M E M O R Y

363

experiments in Helix (Prescott et al. 1997) andmodeling studies in Aplysia (Lieb and Frost 1997).Neural activity in the monosynaptic pathway con-sists of a phasic burst that is rapidly transmitted tothe muscle, thereby mediating the early phase ofthe muscle response. Similar dichotomous rolesare seen in the parallel monosynaptic and polysyn-aptic pathways that mediate withdrawal reflexes inAplysia (Cleary and Byrne 1993; White et al. 1993;Frost and Kandel 1995; Lieb and Frost 1997). Basedon the different functions of the neural pathways,the current study investigated whether differentialexpression of depression and facilitation in themonosynaptic and polysynaptic pathways can ex-plain why habituation largely determines responseamplitude, whereas sensitization largely deter-mines response duration, as previously describedin Helix (Prescott and Chase 1996).

We also sought to localize sites of plasticity toinvestigate the interactions between learning pro-cesses. In certain systems, habituation and sensiti-zation can develop concurrently and compete todetermine behavioral change and thereby producea pattern of learning (referred to here as dual-pro-cess learning) consistent with the dual-processtheory of plasticity proposed by Groves andThompson (1970). Habituation and sensitizationare often considered opposing processes and,therefore, mutually exclusive. But these forms oflearning do occur together and work toward acommon goal. Sensitization increases the salienceof strong and/or novel stimuli (high informationalvalue), whereas habituation decreases the salienceof repeated stimuli (low informational value) that,as Brown (1998) points out, serves to further en-hance the salience of less frequent, high informa-tion stimuli. The combination of sensitization andhabituation essentially works toward making be-havior effective but not wasteful.

Dual-process learning occurs within diversesystems (for review, see Prescott 1998) includingthe tentacle withdrawal reflex of Helix (Christof-fersen et al. 1981; Zakharov and Balaban 1987;Balaban 1993; Prescott and Chase 1996). The rela-tive positioning of depression and facilitationwithin the neural network is thought to underliethe learning kinetics that characterize dual-processlearning. Thus, the second goal of this paper was toexplain the kinetics of dual-process learning by in-vestigating this phenomenon in a relatively simplesystem.

To achieve the two goals described above, thisstudy made use of the recently improved charac-

terization of the neural circuit mediating tentaclewithdrawal in the snail Helix aspersa (Fig. 1) (Pres-cott et al. 1997). Localization of plastic loci wasachieved by stimulating and recording at differentpoints in the circuit, thereby including or exclud-ing plastic loci in the stimulus-response pathway.Careful analysis of the responses shows the disso-ciation of plasticity in the different phases of theresponse; we relate these changes to differentialplasticity in the constituent pathways of the neuralcircuit. Because habituation and sensitization aredifferentially influenced by stimulus intensity andrepetition, we were able to selectively elicit one orthe other form of plasticity by changing stimulationparameters. Habituation is ideally induced by low-intensity, high-frequency (i.e., repetitive) stimula-tion, whereas sensitization is ideally induced byhigh-intensity, low-frequency stimulation (Thomp-son and Spencer 1966). Moreover, in the tentaclewithdrawal reflex of Helix, habituation and sensi-tization are distinguished by their requirements forinduction: The central nervous system (CNS) isnecessary for the induction of sensitization but is

Figure 1: The neural circuit mediating tentacle with-drawal. Mechanical stimulation is applied to the olfac-tory epithelium. Sensory neurons (S) transmit the neuralinformation through parallel peripheral and centralstimulus-response pathways, the latter passing throughthe cerebral ganglion (CNS). Previous evidence suggeststhat the short latency response is mediated by monosyn-aptic pathways, whereas the later, more prolonged re-sponse is mediated by polysynaptic pathways. Prolon-gation of the neural signal probably occurs during trans-mission through locus 1. Apart from the giant neuronC3, each circle represents a group of cells. Facilitatoryneurons (F) have not been identified but are believed toproject from the cerebral ganglion into the tentacle gan-glion. The numbers indicate potential loci of synapticplasticity as referred to in the text and in Table 1. (I)Interneurons; (M) motoneurons additional to C3. Modi-fied from Prescott et al. (1997).

Prescott and Chase


364

not needed for the induction of habituation,though the peripheral nervous system is sufficientto express both forms of learning (Prescott andChase 1996). Therefore, we were able to elicit ha-bituation, sensitization, and combinations thereofto obtain the data necessary to test a recently pub-lished model of dual-process learning kinetics(Prescott 1998). Some of these data have been pub-lished previously in abstract form (S.A. Prescottand R. Chase, unpubl.).

Materials and Methods

Methodology for the current experiments hasbeen published previously (Prescott and Chase1996; Prescott et al. 1997); it is summarized here,along with variations and additional informationrelevant to the current study. Mature specimens ofthe terrestrial snail Helix aspersa were used for allexperiments. Each snail was anesthetized by injec-tion of ∼3 ml of isotonic MgCl2. All central nervousganglia plus the superior tentacles (rhinophores)were removed and dissected in a Sylgard-coateddish filled with a 1:1 mixture of MgCl2 and normalsnail saline (Prescott and Chase 1996). For CNSlesions, the olfactory nerve and the tentacle retrac-tor nerve were cut leaving only the tentacle forexperimentation. After dissection, the solution wasreplaced with normal saline, and experimentationwas delayed $30 min.

Mechanical stimulation was effected by a 1-secpulse of saline applied to the immobilized olfactoryepithelium (which under natural conditions is lo-cated at the tip of the superior tentacle when thetentacle is extended). The strength of stimulationwas adjusted by changing the pump flow rate.Stimulus strength is referred to as weak (0.23 ml/sec) or strong (0.41 ml/sec); stimulation strongerthan 0.41 ml/sec was not used in the present studybecause inhibition can develop after very strongstimulation (Prescott 1997). The tentacle retractormuscle was attached at its proximal end to a forcetransducer, and its contractile force was measuredisometrically. Neural activity was recorded in theolfactory nerve by use of a suction electrode. Allolfactory nerve recordings were performed with aCNS lesion because it was necessary to sever thenerve and take up the distal stump in the electrode(as opposed to taking up the nerve en passant) toget an extracellular recording of sufficient qualityto clearly identify action potentials. Neural activitywas also recorded intracellularly from the identi-fied motoneuron C3. All data were digitized and

stored on computer (Digidata 1200 A/D converterand Axotape 2.0.2 software, both from Axon In-struments) for later analysis.

TRAINING SCHEDULE

The training schedule is adapted from thatused previously (Prescott and Chase 1996). Thebaseline, or naive, response was determined by thefirst test stimulus (trial 0). Blocks of five trainingstimuli were then applied at 2-min interstimulusintervals; after each block, a single test stimuluswas applied. This was repeated five times for atotal of 31 stimuli. The responses to the six teststimuli (trials 0–5) constitute the principal data inour analysis. All test stimuli (with one exceptiondescribed in Results) were “weak” regardless ofthe stimulus intensity used for training. This allowsfor a direct comparison of responses to test stimulisuch that differences between training conditionsare clearly attributable to training, and not to test-ing. Furthermore, the muscle response to weakstimulation is not saturated at any phase, meaningthat sensitization, if present, would not be pre-vented from causing an observable increase in thatphase.

RESPONSE ANALYSIS

The first 30 sec of each response were re-corded and subsequently analyzed to create a re-sponse profile. For muscle responses, tension wasmeasured at 0.5-sec intervals and plotted againsttime after stimulus onset. For olfactory nerve andC3 responses, action potentials were counted in0.5-sec bins and plotted against time after stimulusonset (using time at midpoint of the bin) to give aspike-frequency histogram. This was repeated fortrials 0–5. Each measurement in each trial was nor-malized to the peak response in trial 0 of the samepreparation (maximal tension or bin measuredover 30 sec) so that each measurement is ex-pressed as a percentage of that peak. Responsesfrom all preparations used in a given training con-dition were averaged. Three-dimensional responseprofiles are used to express the response ampli-tude at different phases of the response and toshow how this changes with repeated stimulation,that is, from trials 0–5.

Changes in the different phases of the re-sponse were quantified in learning curves. Formuscle responses, tension was analyzed at three

PLASTICITY OF THE TENTACLE WITHDRAWAL REFLEX


365

different phases of the response: (1) Rising phasecorresponds to tension at 2 sec and reflects therate of muscle contraction, (2) peak phase corre-sponds to tension at 7 sec and reflects the peakamplitude of muscle contraction, and (3) latephase corresponds to average tension between 15sec and 30 sec. The average response over 15 secwas analyzed because tension tends to fluctuateduring the late phase. Response amplitude duringthe late phase is taken as an index of responseduration because it shows the maintenance ofmuscle tension and it is not confounded bychanges in other phases as can happen with othermeasures of response duration, for example,changes in peak phase will affect duration at halfpeak tension. For nerve and C3 responses, latephase was calculated as above, peak phase corre-sponds to the time indicated in the appropriatefigure legends, and rising phase was not calculated.For each preparation, the response at a particularphase in a given trial is expressed as percent of theresponse at the same phase in trial 0. Responsesfrom preparations within the same training condi-tion were averaged. For nerve and C3 responses,n = 3 preparations; this number is low because ofthe difficulty in maintaining stable recordings overthe duration of training (>1 hr). For muscle re-sponses, n varies for different conditions and isreported in the figure legends. The variation is at-tributable to the fact that muscle recordings weredone simultaneously with nerve or C3 recordingsthat were sometimes lost during training; themuscle record was not discarded when the corre-sponding nerve or C3 record was incomplete.

Data are reported as mean ± S.E. For the pur-poses of statistical testing, data were log trans-formed after normalization. As a response de-creases, further reduction is limited and varianceacross preparations decreases, whereas the oppo-site is true as a response increases; a log transfor-mation redistributes the data more evenly andthereby makes variances more similar. Two-tailedunpaired t-tests were applied on rising phase andlate phase data of each training condition; no testswere applied to peak phase data because changeswere generally intermediate compared with theother two phases. Starting with trial 1, testing wasrepeated until an increasing or decreasing trendreached significance compared with either thebaseline response (trial 0) in the same training con-dition or the matching trial in the control condition(weak, infrequent stimulation; Fig. 2B), or untiltrial 5. A Bonferroni correction was applied based

on the number of times t-tests were repeated fordata of a given phase for a given training condition.Two-way analyses of variance (ANOVAs) were alsoused to analyze the effects of stimulus intensity andfrequency on response plasticity. Statistical analy-sis and curve fitting were performed on either Sig-maStat version 1 or SigmaPlot version 4 (SPSS,Inc.).

Results

CONTROL CONDITION AND SELECTIVEINDUCTION OF SENSITIZATION

A sample naive muscle response to weakstimulation is shown in Figure 2A. Using stimula-tion parameters predicted to not cause any plastic-ity (weak test stimuli at 12-min intervals withoutany intervening training stimuli), five preparationswere tested to verify this prediction and to ensurethat the in vitro preparation was stable over theduration of training. Minimal change in the muscleresponse was observed between trials 0 and 5 (Fig.2B). There was some decrease in the rising phase,but the change was not significant compared withbaseline nor did the kinetics suggest that habitua-tion was responsible for the change. The increasein the late phase was also not significant, but it mayindicate mild sensitization. These data (referred toas “control” later in the text) serve as the standardwith which to compare data from other trainingconditions to demonstrate plasticity in the muscleresponse.

The purpose of the next experiment was toselectively induce sensitization without habitua-tion by applying strong but infrequent stimuli (at12-min intervals). This was the only case in whichstrong stimulation was used for testing; the higherbaseline response may cause an underestimation ofthe percentage increase in the response caused bylearning. The muscle response exhibited the ef-fects of sensitization, but not in all phases (Fig. 2C).The late phase response showed the greatest in-crease, rising to 288 ± 75% by trial 1 (which was asignificant increase compared with control;P < 0.05) and remaining at this level. The peakphase response was also slightly increased. The ris-ing phase response decreased, but the change wasnot significant compared with control, suggestingan absence of learning in this phase.

These data demonstrate two points. First, theyconfirm that strong stimuli can elicit sensitization,whereas weak stimuli do not. Second, they show

Prescott and Chase


366

the selective effects of sensitization on the latephase response and therein suggest that facilitationis restricted to the polysynaptic pathway. As pre-dicted, there is no indication of habituation withweak, infrequent stimulation. Given the absence ofhabituation, it is noteworthy that the effects of sen-sitization are maintained between trials 1 and 5 inthe late phase response (Fig. 2C).

PLASTICITY OF MUSCLE RESPONSEWITH CNS INTACT

The goal of the next experiment was to inducehabituation by increasing the stimulation fre-quency. With weak stimulation at 2-min intervals,all phases of the muscle response showed the ex-

ponential decrease (Fig. 3A) characteristic of ha-bituation (Thompson and Spencer 1966). Reduc-tions in the rising phase and late phase responseswere both significant by trial 1 (P < 0.05).

As shown above, repetitive stimulation (i.e.,high-frequency) tends to cause habituation,whereas strong stimulation tends to cause sensiti-zation. High-frequency, strong stimulation shouldsimultaneously cause habituation and sensitization.The effects of this sort of stimulation are shown inFigure 3B. The rising phase habituated but at aslower rate than with repetitive weak stimulation;the decrease was significant by trial 2 (P < 0.05).Habituation of the peak phase was also reducedsuch that there was virtually no change in the re-sponse compared with baseline. Plasticity in the

Figure 2: Plasticity of muscle response atlong interstimulus intervals. Trial stimuliare applied at 12-min interstimulus inter-vals; no intervening training stimulationis given. The box in this and subsequentfigures shows a circuit diagram with sitesof stimulation and recording indicated.(A) Example of muscle response to weakmechanical stimulation in naive prepara-tion. The duration of stimulation is markedby a thick bar at bottom, left of the trace.The three different phases of the responseto be measured are also marked: (m) risingphase, response at 2 sec; (n) peak phase,response at 7 sec; (l) late phase, responseover the last 15 sec. (B1) Response pro-files, weak stimulation; n = 5 prepara-tions. (B2) Learning curves, weak stimula-tion. Error bars in this figure and all otherfigures represent S.E. The late phase re-sponse increases slightly, whereas the ris-ing phase decreases slightly; none of thesechanges is significant compared withbaseline (trial 0) (unpaired t-tests withBonferroni correction for five repeatedtests for each phase). These data are con-sistent with an absence of plasticity; thistraining condition (weak, infrequentstimulation) is therefore referred to as thecontrol condition. (C1) Response profiles,strong stimulation; n = 3 preparations.Small graph shows a 90° clockwise ro-tated view of the large graph. (C2) Learn-

ing curves, strong stimulation. The late phase response is significantly increased by trial 1 compared with the control datain B2 [asterisk (*) P < 0.05; unpaired t-test]. The rising phase response did not change significantly between trials 1 and5 compared with control data [(ns) not significant; unpaired t-tests with Bonferroni correction for five tests]. The peak phaseshowed intermediate changes. These data indicate that habituation does not result from low-frequency stimulation but thatsustained sensitization can be induced by strong, infrequent stimulation.



367

late phase was the most profoundly altered withthe response transiently rising above baseline fortrials 1–3. The response reached 182.3 ± 40.7% intrial 1, which, though not significantly larger thantrial 1 of the control condition, was significantlylarger than trial 1 in Figure 3A (P < 0.05). It is veryimportant to notice the decrease in the late phaseresponse between trials 1 and 5 (Fig. 3B). The non-monotonic changes in late phase response ampli-tude are consistent with dual-process learning andsuggest that habituation and sensitization occur si-multaneously. This is in striking contrast to thelearning kinetics for the late phase response in Fig-ure 2C wherein only sensitization is evident.

The rising phase of the muscle response is me-diated monosynaptically, whereas the late phase ismediated polysynaptically (see Introduction). Thefindings indicate that, with weak stimulation, ha-

bituation affects all phases of the muscle response,suggesting that depression is expressed in bothmonosynaptic and polysynaptic pathways. Withstrong stimulation, only the late phase shows clearevidence of sensitization, suggesting that facilita-tion occurs selectively in the polysynaptic path-way. Reduction in the rate of habituation of therising phase response with strong stimulation com-pared with weak stimulation does not indicate sen-sitization because habituation occurs more slowlywith higher stimulus intensities without any con-tribution from sensitization (Thompson and Spen-cer 1966; see below); notably, the learning curvefor this phase still displays a monotonic exponen-tial decrease. The peak phase is presumably medi-ated jointly through both monosynaptic and poly-synaptic pathways, which would explain its inter-mediate plasticity.

Figure 3: Plasticity of muscle responsewith CNS intact. (m) Rising phase, re-sponse at 2 sec; (n) peak phase, re-sponse at 7 sec; (l) late phase, re-sponse over the last 15 sec. (A1) Re-sponse profiles, weak stimulation; n = 4preparations. (A2) Learning curves,weak stimulation. All phases of themuscle response show an exponentialdecrease, though at slightly differentrates, as is standard for habituation. Re-ductions in the rising phase and latephase are both significant by trial 1[asterisk (*) P < 0.05; unpaired t-tests].(B1) Response profiles, strong stimula-tion; n = 5 preparations. Small graphshows a 90° clockwise rotated view ofthe large graph. (B2) Learning curves,strong stimulation. Different phases ofthe muscle response exhibit very differ-ent changes. The rising phase still ex-hibits habituation though at a reducedrate compared with A2; the reduction issignificant by trial 2 [asterisk (*),P < 0.05; unpaired t-tests with Bonfer-roni correction for two tests]. In con-trast, the late phase transiently sensi-tizes; the increase is not significantcompared with control but is significantcompared with the equivalent point onA [ asterisk (*), P < 0.05; unpairedt-test]. Changes in the peak phaseare again intermediate compared withchanges in the other two phases. Thenonmonotonic changes (as seen in the late phase response) are a defining feature of dual-process learning. These dataindicate that habituation affects all phases of the response, whereas sensitization predominantly affects the late phase.

Prescott and Chase


368

STATISTICAL ANALYSIS OF EFFECTSOF STIMULUS PARAMETERS

Two-way ANOVAs were used to further inves-tigate the effects of stimulus intensity and fre-quency on plasticity of the rising phase and latephase of the muscle response. Comparing the latephase responses in trial 1 for different stimulationparameters (data from Figs. 2 and 3), a two-wayANOVA showed that intensity had a significant ef-fect (Fintensity = 10.00, P < 0.01, df = 1). Frequencyhad an effect that fell just short of significanceafter allowing for the effects of intensity(Ffrequency = 4.28, P = 0.06, df = 1). However, atwo-way ANOVA on the late phase response intrial 5 showed that frequency had a significant ef-fect on plasticity after allowing for differences inintensity (Ffrequency = 10.06, P < 0.01, df = 1),whereas intensity no longer had a significant effect(Fintensity = 2.67, P > 0.05, df = 1). Repeating theANOVA on trials 2–4 showed that intensity had asignificant effect in trial 2 (P < 0.05), but the effectwas not significant in later trials. In contrast, theeffects of frequency only became significant at trial3 (P < 0.05) and increased in later trials. The factthat stimulus frequency and intensity both affectplasticity of the late phase response is further evi-dence that both habituation and sensitization com-pete to determine net changes in this phase of theresponse and, moreover, that the balance betweenthe two learning processes shifts in favor of habitu-ation as training progresses.

Analysis of the rising phase response showed adifferent pattern of effects. Two-way ANOVAs ontrials 1–5 showed that stimulus intensity did nothave a significant effect on plasticity in any of thetrials. Stimulus frequency, on the other hand, had asignificant effect in all trials (at least P < 0.05).These results are consistent with the argument thatthe rising phase of the response is affected by ha-bituation but not by sensitization.

PLASTICITY OF MUSCLE RESPONSEAFTER CNS LESION

Previous studies in Helix have shown that therate, amplitude, and duration of muscle contrac-tion are reduced after a CNS lesion (Prescott andChase 1996; Prescott et al. 1997), but the responseis still quite robust and can exhibit the effects oflearning. Unlike habituation that does not requirethe CNS, sensitization requires the CNS for induc-tion but not for expression (Prescott and Chase

1996). Therefore, preparations with CNS lesionswere trained with either weak or strong stimuli forcomparison to the plasticity in preparations withthe CNS intact. The prediction was that habitua-tion would be intact, whereas sensitization wouldbe absent in those preparations with a CNS lesion.

As when the CNS is intact, repeated weakstimulation after a CNS lesion resulted in habitua-tion of all phases of the muscle response (Fig. 4A).Reductions in the rising phase and in the peakphase were both significant by trial 1 (P < 0.05 andP < 0.01, respectively). Repeated strong stimula-tion with a CNS lesion, on the other hand, did notelicit the same plasticity as with an intact CNS.Instead, all phases of the response habituated withstandard kinetics (Fig. 4B); reduction in the risingphase was significant by trial 1 (P < 0.05), and re-duction in the late phase was significant by trial 4(P < 0.01). Furthermore, kinetics of plasticity inthe late phase were qualitatively changed in thatthere was no transient increase in the responseabove baseline as seen when the CNS was intact(Fig. 3B). These changes are attributable to a loss offacilitation caused by the CNS lesion, which is con-sistent with the necessity of the CNS for sensitiza-tion’s induction (see above). Habituation of the ris-ing phase is virtually unchanged from that seen inFigure 3B, supporting the previous argument thatfacilitation does not occur in the monosynapticpathway.

As mentioned above, habituation is reducedwhen stimulus intensity is increased (Thompsonand Spencer 1966). This is an intrinsic property ofhabituation and does not rely on the incrementaleffects of sensitization. The reduction of habitua-tion in the absence of sensitization is clearly seenby comparing the plasticity elicited by weak (Fig.4A) and strong (Fig. 4B) stimulation after a CNSlesion. The argument for reduced habituationrather than occult sensitization is supported by thefact that the decrease is monotonic, rather thanshowing any transient increase suggestive of sensi-tization.

PLASTICITY OF OLFACTORY NERVE RESPONSEAFTER CNS LESION

Having investigated the learning kinetics oftentacle withdrawal, the next step was to explorethe underlying network plasticity. We began bylooking at plasticity in the earliest (most upstream)part of the circuit and then moved progressivelydownstream. The numbered loci in Figure 1 indi-



369

cate possible sites of plasticity and generally followthe chronology of our investigation.

To investigate plasticity in the upstream cir-cuit, the CNS was lesioned and the distal stump ofthe olfactory nerve was taken up in a suction elec-trode to record afferent neural activity. Figure 5Ashows a sample recording from the olfactorynerve. Response profiles are shown only for strongstimulation (Fig. 5B1); data from weak stimulationare very similar. Learning curves for both stimulusstrengths are shown in Figure 5B2. The peakphase, occurring very early in the response (1.25sec; see Fig. 5A), was not significantly differentfrom the baseline response at any trial, for bothweak and strong stimulation. Given the short la-tency to the peak phase, the majority of this re-sponse constitutes activity in the monosynapticpathway because activity in the polysynaptic path-way is delayed by an upstream synapse and, more-

over, never seems to reach such a high firing fre-quency. The absence of plasticity in the peakphase is therefore consistent with a direct projec-tion of sensory neurons to the CNS (see Fig. 1),meaning that there is no plastic synapse upstreamof the recording site nor that decreased neuronalexcitability plays any role in depression. The shortduration of the nonplastic phase (see Fig. 5B1) re-flects the phasic nature of activity in the monosyn-aptic pathway.

The prolonged neural activity transmittedthrough the polysynaptic pathway (i.e., latephase), on the other hand, showed a progressivedecrease with training. This occurred for bothstrong and weak stimulation, as expected with aCNS lesion (because there is no sensitization). Thereduction compared baseline was significant bytrial 1 for both stimulus strengths (P < 0.05). Themost likely explanation for this reduction is depres-

Figure 4: Plasticity of muscle responseafter CNS lesion. (m) Rising phase, re-sponse at 2 sec; (n) peak phase, re-sponse at 7 sec; (l) late phase, responseover last 15 sec. (A1) Response profiles,weak stimulation; n = 5 preparations.(A2) Learning curves, weak stimulation.As with the CNS intact, weak stimulationresults in habituation of all phases of theresponse. Reductions in the rising phaseand late phase are both significant bytrial 1 [asterisk (*), P < 0.05; asterisks(**), P < 0.01; unpaired t-tests]. (B1) Re-sponse profiles, strong stimulation; n = 3preparations. (B2) Learning curves,strong stimulation. Reduction in the ris-ing phase is still significant at trial 1 [as-terisk (*) P < 0.05; unpaired t-test),whereas the reduction in the peak phasereaches significance at trial 4 [asterisk(*), P < 0.05; unpaired t-tests with Bon-ferroni correction for four tests]. Thesecurves are markedly different from theequivalent curves in Fig. 3B2, that is,there is no transient increase abovebaseline. This confirms that the CNS isnecessary for the induction of sensitiza-tion, which also reflects the heterosyn-aptic nature of facilitation. Comparisonof A and B shows that even in the ab-sence of sensitization, the rate and de-gree of habituation are less under condi-tions of strong stimulation comparedwith weak stimulation, consistent withthe standard kinetics of habituation.

Prescott and Chase


370

sion at the synapses between sensory neurons andinterneurons (locus 1 on Fig. 1) because prolongedfiring occurs in the interneurons.

PLASTICITY OF C3 RESPONSE WITH CNS INTACT

C3 is a giant motoneuron that contributes sub-stantially to the mediation of tentacle withdrawal(Prescott et al. 1997). Its large size allows for rela-tively good intracellular recording, which also pro-vides the benefit of selective, intracellular stimula-tion. The convergent input onto C3 from both sen-sory neurons and interneurons (Fig. 1) is reflectedin its response, which consists of a high-frequencyburst followed by more prolonged, tonic activity(Fig. 6A).

Unlike the olfactory nerve response, the C3response showed a decrease in its peak phase overthe course of training for both weak and strongstimulation (Fig. 6B,C). The responses were con-sistently lower than baseline for both stimulusstrengths, but the reduction did not reach signifi-cance in either case. Given the very high firingfrequency in the peak phase, decreased firing maynot accurately reflect decreased synaptic input toC3 because of response saturation, which couldexplain the attenuated decrement in this phase(Fig. 6B,C). As expected, the decrease was slower

in the case of strong stimulation. The C3 responsedecrement is probably mediated through depres-sion at the synapse between sensory neurons andC3 (locus 2).

With weak stimulation, the late phase C3 re-sponse also decreased (Fig. 6B); the reduction wassignificant by trial 1 (P < 0.05). This could be dueto depression at locus 1 and/or at locus 3. Usingthe simplifying assumption of linear additivity tosubtract the effects of depression at locus 1 (seefigure legend for details), Figure 7A demonstratesthat the reduction in the late phase olfactory nerveresponse is sufficient to account for reduction inthe late phase C3 response. This suggests that de-pression occurs upstream at locus 1 rather thanfarther downstream at locus 3.

Changes in the late phase response were quali-tatively different depending on whether trainingwas with weak or strong stimulation (Fig. 6, cf. Band C). After training with strong stimulation, thelate phase C3 response showed a transient increase(Fig. 6C) reminiscent of plasticity in the late phasemuscle response with the CNS intact (Fig. 3B). Theresponse at trial 1 was not, however, significantlyincreased compared with baseline or with Figure6B. However, a one-way ANOVA on the late phaseresponse using data from trials 1–5 showed thatstimulus intensity did have a significant effect

Figure 5: Plasticity of olfactory nerveresponse after CNS lesion. (A) Exampleof olfactory nerve response to weak me-chanical stimulation in a naive prepara-tion. The tops of the action potentialswere truncated during recording. Theduration of stimulation is marked by athick bar at bottom, left of the trace. Thetwo different phases of the response tobe measured are also marked: (n) peakphase, response at 1.25 sec; (l) latephase, response over the last 15 sec.(B1) Response profiles, strong stimula-tion; n = 3 preparations. Response pro-files for weak stimulation (not shown,n = 3 preparations) are very similar tothose for strong stimulation. (B2) Learn-ing curves, weak and strong stimulation.Consistent with the findings in Fig. 4, fa-cilitation does not occur in the absenceof the CNS. The peak phase shows no

significant change from baseline for either weak or strong stimulation [(ns) not significant; unpaired t-tests with Bonferronicorrection for five tests]. In contrast, prolonged neural activity of the late phase is significantly reduced compared withbaseline by trial 1 for both stimulus strengths [asterisk (*), P < 0.05; unpaired t-tests]. This pattern of depression isconsistent with plasticity at locus 1.



371

(Fintensity = 7.58, P < 0.05, df = 1) that suggeststhat sensitization affects the late phase. Despite thevariability between preparations, plasticity in thelate phase response seen after strong stimulationappears quite different from that seen after weakstimulation.

The above data provide evidence for facilita-tion in at least one component of the central poly-synaptic pathway. Past experiments in which theCNS was lesioned after induction of sensitizationsuggest, in fact, that the majority of facilitation isprobably expressed in the peripheral pathways(Prescott and Chase 1996). Facilitation in periph-eral pathways cannot be directly recorded for tech-nical reasons, but data reported below confirm thatsuch facilitation does occur and appears to be

much more robust than in the central pathway.Common expression of facilitation between cen-tral and peripheral pathways could be explainedby facilitation at locus 1 and/or facilitation at bothloci 3 and 5. Facilitation at the neuromuscular junc-tion (loci 6 and 7) is also possible.

Another issue to consider for localizing facili-tation is the relationship between facilitation andprolongation and their respective sites of expres-sion. Prolongation is the phenomenon wherein theexcitation caused by a brief stimulus is transformedinto a prolonged neuronal discharge. Previous ex-periments have shown that the transformation oc-curs in the tentacle ganglion, most likely at locus 1,and that the degree of prolongation is proportionalto the signal intensity (Prescott et al. 1997). On the

Figure 6: Plasticity of C3 response.(A) Example of C3 response to weak me-chanical stimulation in a naive prepara-tion. The duration of stimulation ismarked by a thick bar at bottom, left ofthe trace. The two different phases of theresponse to be measured are alsomarked: (n) peak phase, response at1.75 sec; (l) late phase, response overthe last 15 sec. This example shows arelatively robust late phase response.(B1) Response profiles, weak stimula-tion; n = 3 preparations. (B2) Learningcurves, weak stimulation. The peakphase response decreases between trials0 and 5, but the reduction does notreach significance compared with base-line [(ns) not significant; unpaired t-testswith Bonferroni correction for five tests];however, response saturation may at-tenuate the observable decrease in spikenumber (see Results). Depression at lo-cus 2 is the most probable cause for thedecrease in the peak phase C3 response.Reduction in the late phase comparedwith baseline is significant by trial 1 [as-terisk (*), P < 0.05; unpaired t-test]. De-pression at locus 1 is the most probablecause for this decrease (see Results).(C1) Response profiles, strong stimula-tion; n = 3 preparations. (C2) Learningcurves, strong stimulation. Reduction ofthe peak phase is less than with weakstimulation [the response is not significantly altered from baseline; (ns) not significant; unpaired t-tests with Bonferronicorrection for five tests], but the kinetics are otherwise quite similar with the response showing a progressive decrement.In contrast, the late phase of the C3 response shows a transient increase very similar to that seen in the late phase muscleresponse (Fig. 3D). The increase at trial 1 is not significant compared with baseline or to data in B2 [(ns) not significant;unpaired t-tests]; however, results of a one-way ANOVA described in the text suggest that sensitization does affect the latephase response. These data are consistent with selective expression of facilitation in the polysynaptic pathway.

Prescott and Chase


372

basis of the influence of signal intensity, facilitationcausing increased transmission at locus 1 would beexpected to enhance prolongation, whereas facili-tation downstream would likely fail to have anysuch effect. It seems highly plausible, therefore,that facilitation conferring sensitization in the late

phase response occurs at locus 1, but this does notexclude plasticity farther downstream in either thecentral or peripheral polysynaptic pathways.

PLASTICITY AT OTHER SITES IN THE CIRCUIT

Having investigated plasticity in the centralpathway, the next goal was to understand plastic-ity at analogous sites in the peripheral pathway.Although direct measurement of plasticity in thispathway is complicated by technical difficulties,data already presented allow one to infer that re-duction of the neural signal in the peripheral path-way (reflected by decreased muscle response afterCNS lesion; see Fig. 4) is mostly attributable todepression at an early locus. For habituation of theearly phase, depression probably occurs at the pe-ripheral synapse between sensory neurons and mo-toneurons (locus 4). Using the same method ofsubtraction as in Figure 7A, Figure 7B shows thatthe majority of reduction in the late phase muscleresponse is most likely attributable to depression atlocus 1. In both cases, the localization of depres-sion is in keeping with previous data (see above)indicating that the output synapses of sensory neu-rons are prone to depression. There may also beplasticity farther downstream, such as depressionat the neuromuscular junction (locus 7), but anysuch contribution is probably minor.

We examined the question of neuromuscularplasticity mediated through the central pathway bydirectly exciting C3 by injection of depolarizingcurrent (∼1.0 nA over 8 sec) to cause spikingequivalent in intensity to that caused by weak me-chanical stimulation of the olfactory epithelium.This intracellular stimulation was repeated usingthe same training schedule as in other experi-ments; total spike number and peak muscle re-sponse were measured. Between trials 0 and 5, theC3 response decreased by only 3.3%, whereas themuscle response decreased by only 4.0% (n = 3preparations). Repeated stimulation thereforecaused neither reduction in C3’s excitability nordepression of C3’s output (i.e., at locus 6).

The possibility of post-tetanic potentiation atC3’s neuromuscular junction (locus 6) was alsotested. Various intensities, durations, and combina-tions of stimulation were tried, but in no instancewas there evidence of increased synaptic transmis-sion to the muscle. Results also suggests that C3 isnot responsible for inducing heterosynaptic facili-tation. The possibility of heterosynaptic facilitationat locus 6 under other training conditions is not

Figure 7: Localization of depression. (A) Depression inthe central pathway, weak stimulation. Original data forlate phase responses have been fit with single exponen-tial curves. Reduction of the late phase olfactory nerveresponse indicates depression at locus 1 (late phase datafrom Fig. 5B2, m), whereas reduction of the late phaseC3 response indicates the summed depression at loci 1and 3 (late phase data from Fig. 6B2, .). Depression atlocus 3 (d) was calculated by subtracting depression atlocus 1 from depression at loci 1 + 3, using the simpli-fying assumption of linear additivity. Inferred plasticityat locus 3. The results indicate that depression at locus 3contributes little to total depression. (B) Depression inthe peripheral pathway, weak stimulation. The samestrategy is used as in A to subtract depression at locus 1(late phase data from Fig. 5B2, m) from depression atloci 1 + 5 + 7 (late phase data from Fig. 4A2, l) to showthat depression at locus 5 and/or 7 makes only a minorcontribution compared with depression at locus 1. (s)Inferred plasticity at loci 5 and 7. Together with thefindings of Figs. 4 and 5, these data indicate that decre-ment of sensory neuron output is the main cause of de-creased transmission through the neural circuit.



373

ruled out by these experiments, but it seems un-likely given other results. All available data suggestthat locus 6 is nonplastic. Separate stimulation ex-periments suggest that post-tetanic potentiation isalso absent from other neuromuscular junctions(see below).

INDUCTION OF SENSITIZATIONBY ELECTRICAL STIMULATION

Given the necessity of the CNS for the induc-tion of sensitization, we hypothesized that facili-tatory neurons project from the CNS to the periph-ery where they effect an increase in synaptic trans-mission selectively in the polysynaptic pathway(see Fig. 1). Given the upstream location of depres-sion relative to facilitatory neurons, activation offacilitatory neurons would tend to wane as depres-sion develops, thereby causing a reduction in fa-cilitation. This has important implications for dual-process learning (see Discussion).

Because facilitatory neurons have not yet beenidentified in this system, the scenario describedabove cannot be directly demonstrated. As an al-ternative, the olfactory nerve was lesioned andelectrically stimulated at its distal end, first, toshow that facilitation can be expressed in the pe-riphery as suggested earlier in the text, second, toshow that this stimulation can elicit sensitization

(therein supporting the existence of centrifugal fa-cilitatory neurons), and third, to investigate the ki-netics of sensitization elicited in such a manner.The stimulus used for each training trial consistedof 50 pulses delivered at 10 Hz, each pulse being20 msec in duration and 500 mV in intensity; weakmechanical stimulation was used for the test trials.

Direct stimulation of the olfactory nerve elic-ited sensitization (Fig. 8). The rising phase re-sponse, although increased, was not significantlychanged in any trial compared with control data.On the other hand, the increase in the late phaseresponse was significant by trial 1 (P < 0.01). Be-cause the CNS had been lesioned, these results sug-gest that the axons of central facilitatory neuronswere directly excited, causing facilitation in theperiphery. The effects of sensitization were similarto those observed in response to mechanical stimu-lation (Figs. 2 and 3); specifically, the late phaseshowed a large increase, whereas the other twophases showed only small increases. One could ar-gue that this sensitization was possibly caused bypost-tetanic potentiation rather than by heterosyn-aptic facilitation, but such an argument is not sup-ported by the differential effects of sensitizationdepending on response phase.

An important point to notice in Figure 8 is thekinetics of the sensitization. Although there is aslight decrease after the largest increase (at trial 1)

Figure 8: Plasticity of muscle responsewith olfactory nerve stimulation. Electricalstimulation was applied to the distal endof the cut olfactory nerve for training;weak mechanical stimulation was used fortesting. (m) Rising phase, response at 2sec; (n) peak phase, response at 7 sec; (l)late phase, response over the last 15 sec.(A) Response profiles; n = 3 preparations.The small graph shows a 90° clockwiserotated view of the large graph. (B) Learn-ing curves. The rising phase increasedabove baseline, but the response is notsignificantly larger than control data atany trial [(ns) not significant; unpaired t-tests with Bonferroni correction for fivetests]. The late phase response showed thegreatest increase and is significantly largerthan control data by trial 1 [asterisks (**),P < 0.01; unpaired t-test). This pattern ofeffects in which the late phase is greatly increased whereas the other phases are less affected is similar to the effectsobserved with training by mechanical stimulation (Figs. 2 and 3). These data therefore support the argument that facili-tatory neurons project from the cerebral ganglia (as shown in Fig. 1) and effect facilitation peripherally in the polysynapticpathways.

Prescott and Chase


374

in the late phase response, the increase is main-tained well above baseline. In this regard, the ki-netics of plasticity in the late phase response aremore similar to those observed in Figure 2C than tothose in Figure 3B; furthermore, there is no evi-dence of decreases in the other phases of the re-sponse. These results suggest that repeated stimu-lation of the olfactory nerve, although causing sen-sitization through direct excitation of thefacilitatory neurons, did not elicit habituation be-cause the site of stimulation was downstream ofthe locus of depression.

Discussion

Careful analysis of the tentacle withdrawal re-flex in Helix shows that different phases of theresponse can change independently through learn-ing and are differentially influenced by habituationand sensitization. Previous data from this systemand from Aplysia (see Introduction) suggest thatthe early phase of withdrawal responses is medi-ated by monosynaptic neural pathways. Data fromFigure 6 show that the monosynaptic pathway dis-plays depression, which is consistent with habitu-ation of the rising phase muscle response seen inFigures 3 and 4. The polysynaptic pathway, whichis believed to be important for the later part of theresponse by virtue of its prolonged activity, alsoexpresses depression (Figs. 5 and 6), the effects ofwhich are reflected in the late phase muscle re-sponse (Figs. 3 and 4). But whereas depression iscommon to both pathways, facilitation is selec-tively expressed in the polysynaptic pathway assuggested by the dissociation of plasticity in theearly and late phase responses in both C3 (Fig. 6)and the muscle (Figs. 2, 3, and 8). These results areconsistent with our previous report of the differ-ential effects of habituation and sensitization onresponse amplitude and duration (Prescott andChase 1996). Recent experiments in Tritonia(Brown et al. 1996) and in Aplysia (Hawkins et al.1998) have also demonstrated dissociative changesin different components of behavior, leading theauthors to postulate that plasticity at different lociin the network underlies changes in different com-ponents of the behavior.

Table 1 summarizes the localization of plastic-ity within the neural network mediating tentaclewithdrawal in Helix. The vast majority of plasticityoccurs at the output synapses of sensory neurons(loci 1, 2, and 4), but, as noted above, net plasticityis not the same between monosynaptic and poly-

synaptic pathways, with facilitation selectively ex-pressed in the latter (most likely at locus 1). Thebalance between the opposing forms of plasticityis significantly influenced by the intensity and fre-quency of stimulation. By taking advantage of thissensitivity to stimulation parameters, as well as thenecessity of the CNS for sensitization, data werecollected to investigate how depression and facili-tation interact to determine net plasticity and thekinetics of dual-process learning, as will be ex-plained below.

HABITUATION AND DEPRESSION

It is widely held that synaptic depression iscausally related to behavioral habituation (seeChristoffersen 1997) though other mechanismscausing reduced neural transmission have been de-scribed, such as decreased sensory neuron excit-ability (Walters et al. 1983). In Aplysia, homosyn-aptic depression is very robust at sensory neuronoutput synapses (Castellucci et al. 1970). Plasticityat an analogous, upstream position seems to occurin many circuits (for reviews, see Menzel andBicker 1987; Prescott 1998). Data indicate that thetentacle withdrawal reflex of Helix similarly ex-presses depression at the sensory neuron outputsynapses (loci 1, 2, and 4).

There are many consequences of depressionoccurring so early in the circuit. Because depres-sion precedes divergence of monosynaptic andpolysynaptic pathways, depression affects trans-mission through both pathways, as seen in Aplysia(Hawkins et al. 1981). The ubiquitous effects ofdepression on the rate (rising phase), peak ampli-

Table 1: Summary of plastic loci

Synapse

Numberon

Fig. 1 Depression Facilitation

S-I 1 yes yesS-C3 2 yes noI-C3 3 no noS-M 4 yes noI-M 5 no (?)C3-muscle 6 no noM-muscle 7 no (?)

(C3) Identified motoneuron; (F) facilitatory neurons;(I) interneurons; (M) motoneurons additional to C3;(S) sensory neurons.



375

tude (peak phase), and duration (late phase) of theresponse seen in this study support the same con-clusion. Another consequence of depression soearly in the circuit is that input to downstreamfacilitatory neurons will wane as depression devel-ops, which has important consequences for thekinetics of learning (see below).

SENSITIZATION AND FACILITATION

Just as depression is thought to cause habitu-ation, facilitation is thought to underlie sensitiza-tion. An increase in synaptic transmission can beeffected by many different mechanisms (Fisher etal. 1997). By using the word facilitation, we wishonly to indicate that transmission through the plas-tic locus is increased and that the mechanism re-sponsible for this change is heterosynaptic, that is,produced by influences extrinsic to the plastic lo-cus. In the case of tentacle withdrawal, the needfor heterosynaptic modulation is illustrated by thenecessity of the CNS for facilitation (cf. Figs. 3 and4), especially given that facilitation can be ex-pressed in the periphery (Fig. 8; Prescott andChase 1996).

In Aplysia, the mechanism for presynaptic fa-cilitation has been well described (Castellucci andKandel 1976; for review, see Byrne and Kandel1996). A similar mechanism probably occurs in He-lix. Facilitation also tends to occur at an upstreamposition but only at a subset of the synapses thatdisplay depression. In other words, depressiontends to be cellwide, but facilitation is branch spe-cific depending on where modulatory transmitteris received and where it induces facilitation (Clarkand Kandel 1984; Martin et al. 1997). The branchspecificity of facilitation is important for explainingthe differential effects of sensitization on differentcomponents of the behavior as observed in Aplysia(e.g., Stopfer and Carew 1996) and in the currentstudy (Figs. 2, 3, 6, and 8). The most plausiblescenario for facilitation in the tentacle withdrawalreflex is that facilitatory neurons projecting fromthe CNS (Fig. 1) cause facilitation peripherally, inthe polysynaptic pathways.

Although the effects of facilitation predomi-nate in the polysynaptic pathway underlying ten-tacle withdrawal in Helix, precise localization ofthe facilitated loci is uncertain (Table 1). A largecomponent probably occurs at locus 1, first, be-cause of facilitation’s common expression in cen-tral and peripheral polysynaptic pathways and, sec-ond, because of the position of locus 1 relative to

the site of prolongation. However, both lines ofargument are circumstantial, and facilitation’smore precise localization warrants further investi-gation. In Aplysia for instance, facilitation at sen-sory neuron synapses is well documented, butcomparatively recent findings have emphasizedplasticity at other sites and by other mechanismssuch as a reduction of inhibition (Frost et al. 1988;Fischer and Carew 1993; Trudeau and Castellucci1993a,b; Cohen et al. 1997). Although inhibitioncan occur in the circuit mediating tentacle with-drawal, modulation of inhibition does not contrib-ute to plasticity with the stimulus strengths used inthis study (Prescott 1997).

DUAL-PROCESS THEORY OF LEARNING

Interactions between learning processes areimportant in determining how plasticity developsand how it is expressed. At the synaptic level, theeffects of depression influence whether spikebroadening or vesicle mobilization is the predomi-nant cause of facilitation (Klein 1995; Byrne andKandel 1996). At the network level, Hawkins et al.(1998) have suggested recently that interactionsbetween habituation and inhibition may underliesome of the differences between dishabituationand sensitization previously described in Aplysia(e.g., Marcus et al. 1988). Dual-process learning,characterized by transient sensitization followedby habituation (Groves and Thompson 1970), isalso thought to result from interactions betweenlearning processes at the network level (Prescott1998). The plasticity exhibited by the tentaclewithdrawal reflex is consistent with the dual-pro-cess theory of plasticity.

The results of this study support our hypoth-esis that the relative positioning of depression andfacilitation affects how these forms of plasticity in-teract to determine the kinetics of dual-processlearning. As stimulation is repeated, depression oc-curs at sensory neuron output synapses. Withweak stimulation, depression is robust and occursin the absence of facilitation. With stronger stimu-lation, depression still occurs but is weaker andslower to develop, consistent with the standardfeatures of habituation (Thompson and Spencer1966), and, moreover, the changes in synaptictransmission are confounded by the introductionof facilitation. Like depression, facilitation is ex-pressed at an upstream locus, but it may not beubiquitous in all neural pathways given its branchspecificity (see above).

Prescott and Chase


376

Unlike depression, which is by a homosynap-tic process, facilitation is heterosynaptic and relieson facilitatory neurons for its induction. These fa-cilitatory neurons are excited via sensory neuronswhose output is prone to depression; therefore,the induction of sensitization will wane as depres-sion develops upstream and causes reduced activa-tion of facilitatory neurons. We refer to this reduc-tion in sensitization’s induction as the habituationof sensitization. Facilitation is also expressed up-stream, but its pathway specificity is such that fa-cilitation does not enhance the induction of facili-tation, that is, there is no positive feedback loop(see Prescott 1998). Therefore, the relative posi-tioning of plasticity in the network may explain thelongstanding observation that sensitization tendsto habituate (Lehner 1941; Thompson and Spencer1966; Pinsker et al. 1970).

Our hypothesis is, therefore, that depression isexpressed upstream of the site at which facilitationis induced and that the consequent interaction be-tween the learning processes leads to the kineticsof dual-process learning. Hill and Jin (1998) havedescribed a very similar pattern of plasticity in thecricket cercal system. When depression and facili-tation occur together in that system, learning ki-netics characteristic of dual-process learning areobserved. Furthermore, all of the sensory neuronsynapses display depression thought to be medi-ated by a presynaptic mechanism. Facilitation alsooccurs but only at a subset of synapses. The hy-pothesis is that facilitation is mediated by a retro-grade signal from the postsynaptic cell (Davis andMurphy 1993). The important observation is thatthe same relative positioning of depression and fa-cilitation as observed in the present paper givesrise to dual-process learning kinetics in the cricketcercal system.

Having not yet identified the facilitatory neu-rons in the tentacle withdrawal reflex, we have notbeen able to establish with certainty whether re-duced input to facilitatory neurons is responsiblefor the habituation of sensitization in Helix. Re-duced input to the facilitatory neurons would oc-cur if innervation of those neurons were predomi-nantly through the monosynaptic pathway; suc-cessful induction of sensitization by short-duration,high-intensity stimulation of the olfactory nerve(Fig. 8) is consistent with this type of innervation.Reduced input to central facilitatory neurons isalso consistent with the observation that facilita-tion has a greater effect on peripheral pathwaysthan on central ones (see Results). Hypothetically,

depression of facilitatory neuron output could alsocause the habituation of sensitization, but the ob-servation that sensitization does not wane whenthe olfactory nerve is repeatedly stimulated (Fig. 8)argues against such a mechanism and further sug-gests that decremental changes are occurring up-stream of the putative facilitatory neurons. The ca-pacity to cause dishabituation by stimulation else-where than the tentacle (Prescott and Chase 1996),indicates that the capacity for incremental changeis intact and, therefore, that depression predomi-nantly affects the induction rather than the expres-sion of facilitation. Dishabituation is, of course, animportant criterion for habituation (Thompson andSpencer 1966), and it suggests that depression offacilitatory neuron output should not occur, butone must be cautious in equating the mechanismsof dishabituation and sensitization (as earlier pa-pers did, e.g. Groves and Thompson 1970) givenmore recent findings (Marcus et al. 1988).

In a recent review of the dual-process theoryof plasticity (Prescott 1998), a set of differentialequations was presented with the intention of de-scribing the kinetics of dual-process learning basedon a logical interpretation of the interactions be-tween learning processes. As shown in that paper,data previously available to test the model wereless than ideal. Some of the experiments presentedhere were specifically designed to elicit habitua-tion and sensitization separately and thereby allowdetermination of the kinetics of the individuallearning processes by fitting the appropriatecurves. The effects of pure habituation and puresensitization can be described by simple differen-tial equations:

Pure habituation:

dEH/dt = −h~EH − Emin!. (1)

Pure sensitization:

dES/dt = s~Emax − ES!. (2)

E represents synaptic efficacy, and subscripts iden-tify the learning process causing E to change. Forhabituation, EH decreases at rate h to minimumasymptote Emin. For sensitization, ES increases atrate s to maximum asymptote Emax (see Fig. 9,legend). Equations 1 and 2 can be combined todescribe the habituation of sensitization:

Habituating sensitization:

dEHS/dt = s ? EH~t!/100@~Emax − 100!EH~t!/100+ 100 − EHS#. (3)



377

Equation 3 is written such that the rate and extentof sensitization decrease as habituation develops.Equation 3 is slightly modified from its originalform (Prescott 1998), but only to account for thedifferent scale in which E is expressed. The effectsof habituating sensitization and habituation areadded to determine the net plasticity given thatdepression and facilitation are most likely ex-pressed in parallel in this system.

The numerical values for parameters in equa-tions 1 and 2 were determined by fitting the curvesfor pure habituation and pure sensitization (Fig. 9).These values were then used in equation 3 to cal-culate the predicted plasticity associated with dual-process learning. Comparison of the predictedplasticity with data from experiments in which de-

pression and facilitation developed concurrently(Fig. 3B) shows that the simple mathematicalmodel is reasonably successful in describing thekinetics of dual-process learning (r2 = 0.84; Fig. 9);this value of r is significant (P < 0.05).

The apparent success of the model in dealingwith the current data is not sufficient to validatethe model, but it does demonstrate the model’sutility for future research in this and other systems.For instance, if the neurons responsible for sensi-tization of the tentacle withdrawal reflex wereidentified, it would be desirable to determine thekinetics of depression in their input (or possiblyoutput) for use in equation 3, that is, to calculatethe kinetics of the habituation of sensitization. Ifthe prediction is correct that habituation of sensi-tization is faster and more robust than responsehabituation (Prescott 1998), the systematic over-shoot of the predicted dual-process plasticity inFigure 9 would be eliminated. It would also bebeneficial to test the model over a range of stimu-lus frequencies and intensities. This would evalu-ate how well the model predicts the kinetics ofdual-process learning with quantitatively differentcombinations of depression and facilitation.

Findings from the current study illustrate theimportance of identifying the location of plasticchanges within a neural network. First, differentialexpression of plasticity in the various elements ofthe network can give rise to dissociative changes inthe various components of behavior, such as therate, peak amplitude, and duration of the response.Second, plasticity at a particular locus in a networkcan affect plasticity occurring elsewhere in the net-work giving rise to such phenomena as dual-pro-cess learning. Clearly, cellular plasticity must beconsidered at the network level if one wishes tofully explain the modification of behavior by learn-ing.

AcknowledgmentsThis research was financially supported by a

postgraduate scholarship to S.A.P. and an operating grant toR.C., both from the Natural Sciences and EngineeringResearch Council of Canada. We thank V. Castellucci, P.Drapeau, G. Pollack, and S. Ratte for their helpful advice.

The publication costs of this article were defrayed inpart by payment of page charges. This article must thereforebe hereby marked “advertisement” in accordance with 18USC section 1734 solely to indicate this fact.

ReferencesBalaban, P. 1993. Behavioral neurobiology of learning interrestrial snails. Prog. Neurobiol. 41: 1–19.

Figure 9: Comparison of empirical data with math-ematical model for dual-process learning. Pure sensiti-zation (j, Es) is represented here by plasticity in the latephase muscle response shown in Fig. 2C2; the data arefit (r2 = 0.99) by an exponential curve described by thedifferential equation dES / dt = 4.0 (287.5 − ES). Pure ha-bituation (m, EH) is represented by plasticity in the latephase muscle response shown in Fig. 4B2; the data arefit (r2 = 0.96) by the equation dEH / dt = −0.43(EH − 19.5). The net plasticity resulting from concurrenthabituation and habituating sensitization (s, EHS + EH

− 100), that is, dual-process learning, is represented byplasticity in the late phase muscle response shown inFig. 3B2; in this case, the curve is not a least squares fitof the data but, instead, represents the predicted plastic-ity using the numerical values from the above two curvefits in the equation for habituating sensitization,dEHS / dt = 4.0 ? EH(t) / 100 [(287.5 − 100)EH(t) / 100 +100 − EHS]. This equation was numerically integratedusing the fourth-order Runge–Kutta method to calculateEHS as a function of time, and the final net plasticity wasdetermined as EHS(t) + EH(t) − 100. The resultant curvedescribes the empirical data quite well (r2 = 0.84, whichis significant, P < 0.05) suggesting that this model mayaccurately describe the nonlinear interactions betweendepression and facilitation that produce the kineticscharacteristic of dual-process learning.

Prescott and Chase


378

Brown, G.D. 1998. Nonassociative learning processesaffecting swim probability in the seaslug Tritonia diomedea:Habituation, sensitization and inhibition. Behav. Brain Res.95: 151–165.

Brown, G.D., W.N. Frost, and P.A. Getting. 1996.Habituation and iterative enhancement of multiplecomponents of the Tritonia swim response. Behav. Neurosci.110: 486–491.

Byrne, J.H. and E.R. Kandel. 1996. Presynaptic facilitationrevisited: State and time dependence. J. Neurosci.16: 425–435.

Castellucci, V. and E.R. Kandel. 1976. Presynaptic facilitationas a mechanism for behavioral sensitization in Aplysia.Science 194: 1176–1178.

Castellucci, V., H. Pinsker, I. Kupfermann, and E.R. Kandel.1970. Neuronal mechanisms of habituation anddishabituation of the gill-withdrawal reflex in Aplysia.Science 167: 1745–1748.

Christoffersen, G.R.J. 1997. Habituation: Events in the historyof its characterization and linkage to synaptic depression. Anew proposed kinetic criterion for its identification. Prog.Neurobiol. 53: 45–66.

Christoffersen, G.R.J., K. Frederiksen, J. Johansen, B.I.Kristensen, and L. Simensen. 1981. Behavioral modificationof the optic tentacle of Helix pomatia; effect of puromycin,activity of S-100. Comp. Biochem. Physiol. 68: 611–624.

Clark, G.A. and E.R. Kandel. 1984. Branch-specificheterosynaptic facilitation in Aplysia siphon sensory cells.Proc. Natl. Acad. Sci. 81: 2577–2581.

Cleary, L.J. and J.H. Byrne. 1993. Identification andcharacterization of a multifunction neuron contributing todefensive arousal in Aplysia. J. Neurophysiol.70: 1767–1776.

Cohen, T.E., S.W. Kaplan, E.R. Kandel, and R.D. Hawkins.1997. A simplified preparation for relating cellular events tobehavior: Mechanisms contributing to habituation,dishabituation, and sensitization of the Aplysiagill-withdrawal reflex. J. Neurosci. 17: 2886–2899.

Davis, G.W. and R.K. Murphy. 1993. A role for postsynapticneurons in determining presynaptic release properties in thecricket CNS: Evidence for retrograde control of facilitation. J.Neurosci. 13: 3827–3838.

Fischer, T.M. and T.J. Carew. 1993. Activity-dependentpotentiation of recurrent inhibition: A mechanism fordynamic gain control in the siphon withdrawal reflex ofAplysia. J. Neurosci. 13: 1302–1314.

Fisher, S.A., T.M. Fischer, and T.J. Carew. 1997. Multipleoverlapping processes underlying short-term synapticenhancement. Trends Neurosci. 20: 170–177.

Frost, W.N. and E.R. Kandel. 1995. Structure of the network

mediating siphon-elicited siphon withdrawal in Aplysia. J.Neurophysiol. 73: 2413–2427.

Frost, W.N., G.A. Clark, and E.R. Kandel. 1988. Parallelprocessing of short-term memory for sensitization in Aplysia.J. Neurobiol. 19: 297–334.

Groves, P.M. and R.F. Thompson. 1970. Habituation: Adual-process theory. Psychol. Rev. 77: 419–450.

Hawkins, R.D., V.F. Castellucci, and E.R. Kandel. 1981.Interneurons involved in mediation and modulation ofgill-withdrawal reflex in Aplysia. I. Identification andcharacterization. J. Neurophysiol. 45: 304–314.

Hawkins, R.D., T.E. Cohen, W. Greene, and E.R. Kandel.1998. Relationships between dishabituation, sensitization,and inhibition of the gill- and siphon-withdrawal reflex inAplysia californica—effects of response measure, test time,and training stimulus. Behav. Neurosci. 112: 24–38.

Hill, A.A.V. and P. Jin. 1998. Regulation of synapticdepression rates in the cricket cercal sensory system. J.Neurophysiol. 79: 1277–1285.

Klein, M. 1995. Modulation of ion currents and regulation oftransmitter release in short-term synaptic plasticity: The riseand fall of the action potential. Invert. Neurosci. 1: 15–24.

Lehner, G.F.J. 1941. A study of the extinction ofunconditioned reflexes. J. Exp. Pyschol. 29: 435–456.

Lieb, J.R., Jr. and W.N. Frost. 1997. Realistic simulation ofthe Aplysia siphon-withdrawal reflex circuit: Roles of circuitelements in producing motor output. J. Neurophysiol.77: 1249–1268.

Marcus, E.A., T.G. Nolen, C.H. Rankin, and T.J. Carew.1988. Behavioral dissociation of dishabituation, sensitization,and inhibition in Aplysia. Science 241: 210–213.

Martin, K.C., A. Casadio, H. Zhu, E. Yaping, J.C. Rose, M.Chen, C.H. Bailey, and E.R. Kandel. 1997. Synapse-specific,long-term facilitation of Aplysia sensory to motor synapses: Afunction for local protein synthesis in memory storage. Cell91: 927–938.

Menzel, R. and G. Bicker. 1987. Plasticity in neuronalcircuits and assemblies of invertebrates. In The neural andmolecular bases of learning (ed. J.-P. Changeux and M.Konishi), pp. 433– 472. John Wiley and Sons, Toronto,Canada.

Pinsker, H., I. Kupfermann, V. Castellucci, and E.R. Kandel.1970. Habituation and dishabituation of the gill-withdrawalreflex in Aplysia. Science 167: 1740–1742.

Prescott, S.A. 1997. “Interactions of habituation andsensitization at the network level illustrated by the tentaclewithdrawal reflex of a snail.” MSc thesis, McGill University,Montreal, Canada.

———. 1998. Interactions between depression and



379

facilitation within neural networks: Updating the dual-processtheory of plasticity. Learn. & Mem. 5: 446–466.

Prescott, S.A. and R. Chase. 1996. Two types of plasticity inthe tentacle withdrawal reflex of Helix aspersa aredissociated by tissue location and response measure. J.Comp. Physiol. A 179: 407–414.

Prescott, S.A., N. Gill, and R. Chase. 1997. The neural circuitmediating tentacle withdrawal in Helix aspersa, with specificreference to the competence of the motoneuron C3. J.Neurophysiol. 78: 2951–2965.

Stopfer, M. and T.J. Carew. 1996. Heterosynaptic facilitationof tail sensory neuron synaptic transmission duringhabituation in tail-induced tail and siphon withdrawalreflexes of Aplysia. J. Neurosci. 16: 4933–4948.

Thompson, R.F. and W.A. Spencer. 1966. Habituation: Amodel phenomenon for the study of neuronal substrates ofbehavior. Psychol. Rev. 73: 16–43.

Trudeau, L.-E. and V.F. Castellucci. 1993a. Functionaluncoupling of inhibitory interneurons plays an important rolein short-term sensitization of Aplysia gill and siphonwithdrawal reflex. J. Neurosci. 13: 2126–2135.

———. 1993b. Sensitization of the gill and siphonwithdrawal reflex of Aplysia: Multiple sites of change in theneuronal network. J. Neurophysiol. 70: 1210–1220.

Walters, E.T., J.H. Byrne, T.J. Carew, and E.R. Kandel. 1983.Mechanoafferent neurons innervating tail of Aplysia. I.Response properties and synaptic connections. J.Neurophysiol. 50: 1522–1542.

White, J.A., I. Ziv, L.J. Cleary, D.A. Baxter, and J.H. Byrne.1993. The role of interneurons in controlling thetail-withdrawal reflex in Aplysia: A network model. J.Neurophysiol. 70: 1777–1786.

Zakharov, I.S. and P.M. Balaban. 1987. Neural mechanismsof age-dependent changes in avoidance behaviour of thesnail Helix lucorum. Neurosci. 23: 721–729.

Received July 30, 1998; accepted in revised form April 16,1999

Prescott and Chase


380

sites of plasticity in the neural circuit mediating tentacle

Documents