Ž .Brain Research 803 1998 9–18

Research report

Selective changes in AMPA receptors in rabbit cerebellum followingclassical conditioning of the eyelid-nictitating membrane response

Stephanie A. Hauge a, Jo Anne Tracy b, Michel Baudry a, Richard F. Thompson a,)

a Neuroscience Program, UniÕersity of Southern California, Los Angeles, CA 90089, USAb Department of Psychology, Program in Neural Science, Indiana UniÕersity, Bloomington, IN 47401, USA

Accepted 12 May 1998

Abstract

Ž .a-Amino-3-hydroxy-5-methylisoxazole-4-propionic acid AMPA receptors are critically involved in several forms of synapticŽ .plasticity proposed to be neural substrates for learning and memory, e.g., long-term potentiation and long-term depression LTD . The

present study was designed to determine changes in cerebellar AMPA receptors following classical conditioning of the eyeblink-nictitat-Ž .ing membrane response NMR in the rabbit. Quantitative autoradiography was used to assess changes in ligand binding properties of

cerebellar AMPA receptors following NMR conditioning elicited by pairing electrical stimulation of the pontine nuclei with an airpuff tow3 x w3 x Ž .the eye. H AMPA and H -6-cyano-7-nitroquinoxaline-2,3-dion CNQX binding were determined following preincubation of

w3 xfrozen–thawed brain tissue sections at 0 or 358C. With 08C preincubation, no significant differences in H AMPA binding to cerebellarAMPA receptors were seen between any of the experimental groups tested. In contrast, preincubation at 358C revealed significant

w3 xdecreases in H AMPA binding to the trained side of the cerebellar cortex resulting from paired presentations of the conditioned and theunconditioned stimuli, while unpaired presentations of the stimuli resulted in no significant effect. With 358C preincubation, there were

w3 x w3 xno significant differences in H CNQX binding between any of the experimental groups and no significant differences in H AMPAbinding in the untrained side of the cerebellum. These results indicate that NMR conditioning is associated with a selective modificationof AMPA-receptor properties in brain structures involved in the storage of the associative memory. Furthermore, they support thehypothesis that cerebellar LTD, resulting from decreased synaptic efficacy at parallel fiber–Purkinje cell synapses mediated by a changein AMPA-receptor properties, is a form of synaptic plasticity that supports this type of learning. q 1998 Published by Elsevier ScienceB.V. All rights reserved.

Keywords: Classical conditioning; Nictitating membrane; a-Amino-3-hydroxy-5-methylisoxazole-4-propionic acid; 6-Cyano-7-nitroquinoxaline-2,3-dion;Quisqualate receptor; Cerebellum; Autoradiography

1. Introduction

Classical conditioning of discrete behavioral responsesis critically dependent upon the cerebellum. For classicalconditioning of the eyelid-nictitating membrane responseŽ .NMR , rabbits learn to associate an initially neutral condi-

Ž .tioned stimulus CS with an aversive unconditioned stim-Ž .ulus US following paired presentations of CS and US.

ŽInitially, rabbits extend the nictitating membrane and.close the external eyelid only in response to the US, thus

Ž .producing an unconditioned response UR , but followingrepeated paired presentations of the two stimuli, rabbits

) Corresponding author. Fax: q1-213-740-5687; E-mail:thompson@neuro.usc.edu

extend the nictitating membrane in response to the CSŽ .alone, producing a conditioned response CR . The mossy

Ž .fiber input from the pontine nuclei PN and other sourcesto the cerebellum conveys essential information concern-ing the CS, and the cerebellar climbing fiber input fromthe inferior olive forms the fundamental pathway for the

w xUS 63,65 . The anatomical circuit for the CR includes: thecerebellum, efferent projections from the anterior interposi-tus nucleus via the superior cerebellar peduncle to themagnocellular red nucleus, and descending rubral projec-tions from the red nucleus to the motor and premotornuclei. A wealth of evidence identifies the cerebellum asthe site of the memory trace for this response. Morespecifically, a localized region of the cerebellum includingthe anterior interpositus nucleus is necessary for bothnormal acquisition and retention of this response

0006-8993r98r$19.00 q 1998 Published by Elsevier Science B.V. All rights reserved.Ž .PII: S0006-8993 98 00525-3

( )S.A. Hauge et al.rBrain Research 803 1998 9–1810

w x5,6,35,36,47,71 . Large cerebellar cortical lesions canmarkedly impair acquisition and retention of the CRw x w x37,41,72 , impair adaptive timing of the CR 41,52 and

w x Žprevent acquisition of conditioned inhibition 41 . Forw x.reviews, see Refs. 30,63,65,73 .

Direct electrical stimulation of the pontine nuclei ormossy fibers in the middle cerebellar peduncle functions asan effective conditioned stimulus. Rabbits trained withpontine or mossy-fiber stimulation paired with a peripheralUS acquire the NMR in a manner that is more rapid and

w xrobust than conditioning with a peripheral-tone CS 60 . Inaddition, some animals initially trained with a pontine-stimulation CS and then switched to a peripheral-tone CSdemonstrate complete, immediate transfer of the CR and ofthe learning-induced neuronal model of the CR in theinterpositus nucleus. Therefore, it is thought that apontine-stimulation CS and a peripheral-tone CS activate a

w xsignificant portion of the same neuronal pathway 61 .Mossy-fiber projections from the PN and other sources

that provide the CS information are glutamatergic andsynapse on the dendritic terminals of granule cells withinthe cerebellar granular layer and on the deep cerebellar

Ž .nuclei DCN , activating both N-methyl-D-aspartateŽ .NMDA and non-NMDA glutamate receptor subtypesw x7–9,14–16,19,24,28,42,43,48,49,51,57,70 . In the molec-ular layer of the cerebellar cortex, parallel fibers, whichare axons of granule cells, provide Purkinje-cell dendrites

w xwith numerous weak excitatory inputs 21,50 . A wealth ofexperimental evidence suggests that L-glutamate is thetransmitter released by parallel fibers in the cerebellar

w xcortex 20,25,26,46,55,74 . Both ionotropic andmetabotropic glutamate receptors mediate parallel fiber–Purkinje cell transmission, and both are thought to becritically involved in the induction and maintenance of

Ž . w xcerebellar long-term depression LTD 1,18,27,34,39 . Asyet, neither the climbing fiber transmitter nor the post-synaptic receptors for the climbing fiber–Purkinje cell orclimbing fiber–DCN synapses have been identified. It isthought, however, that currently identified a-amino-3-hy-

Ž .droxy-5-methylisoxazole-4-propionic acid AMPA recep-wtors do not mediate transmission at these synapses 1,32–

x34,40,53 .In the present study, ligand binding and quantitative

autoradiography were used to investigate potential changesin cerebellar AMPA-receptor properties resulting from thenormal acquisition and retention of the NMR in the rabbit.w3 xH AMPA, an agonist specific for the AMPA family of

w3 xglutamate receptors, and H -6-cyano-7-nitroquinoxaline-Ž .2,3-dion CNQX , an antagonist of this receptor family,

were used to label AMPA receptors in thin frozen–thawedsections from rabbit cerebella preincubated at differenttemperatures. Results indicate that a regional modificationof AMPA-receptor properties is associated with NMRconditioning and that this modification is specific to thesynaptic subpopulation of this receptor type within thetrained side of the cerebellum.

2. Materials and methods

2.1. Animals

Ž .Sixteen male New Zealand white rabbits 2.5–3.0 kgwere individually housed and maintained on food andwater ad libitum. All animals were treated according toNIH guidelines for the welfare of laboratory animals. Theanimals were divided into three groups: those receiving

Ž .paired stimuli Paired, ns5 , pseudorandom unpairedŽ . Ž .stimuli Unpaired, ns6 , and no handling Naive, ns5 .

2.2. Surgery

Animals in the paired and unpaired groups were anes-Ž .thetized initially with Xylazine 8.0 mgrkg and Ketamine

Ž .60.0 mgrkg and were maintained with 1.5–2.0%halothane during surgery. Two insulated, stainless steel

Žbipolar stimulating electrodes approximately 200-mm ex-.posed tips were implanted in the pontine nuclei according

to stereotaxic coordinates and to maximize field potentialsgenerated at the electrode tip by an acoustic stimulus.Stereotaxic coordinates for electrode placement in the right

Žpontine nuclei with lambda positioned 1.5 mm ventral to. Žbregma were approximately depending upon the animal’s.weight : 5.0 mm lateral to lambda, 0.5–1.0 mm rostral to

w x Ž .lambda, and 13.0–15.0 mm ventral 60 see Fig. 1 .Animals were allowed to recover for 7 days followingsurgery prior to commencement of training.

2.3. BehaÕioral training

Animals in both the paired and unpaired groups wereadapted to the training apparatus for 1 h on two consecu-tive days. Paired animals were then trained with standarddelay procedures for classical conditioning of the NMRŽ .intertrial intervals ranging pseudorandomly from 20–40 s

Žwith pontine-stimulation CS 60 mA, 200 Hz, 0.1 ms. Ž .pulses and airpuff US 3 psi corneal airpuff . Condition-

ing was achieved by pairing a 350-ms train of bipolarŽ . Ž .stimulation CS with a coterminating 100-ms airpuff US

directed toward the left cornea. Each training session wascomposed of 12 blocks, with each block consisting of oneCS-alone trial followed by eight paired CS–US trials andone US-alone trial. A CR was defined as a nictitating-membrane movement of 0.5 mm or greater occurringbetween the CS and US onset. Paired animals were trainedto criterion, defined as eight CRs on nine consecutivetrials, and then overtrained for four additional sessions.Unpaired animals received uncorrelated presentations of

Žthe stimulation CS and airpuff US 108 pseudorandomizedpresentations of each stimulus with and average intertrial

.interval of 15 s in a number of sessions equivalent to thatexperienced by the paired subjects and, therefore, matchedthe paired animals exactly for number of training days and

w xexposure to stimuli 60 . Naive animals were not subject totraining, handling, or surgical procedures.

( )S.A. Hauge et al.rBrain Research 803 1998 9–18 11

Ž . Ž .Fig. 1. Summary of CS stimulation sites used in the present study. Paired animals ns5 were given acquisition training asterisks , while unpaired rabbitsŽ . Ž .ns6 were given pseudorandom unpaired training open triangles . LPs lateral pontine nucleus; MPsmedial pontine nucleus; VPsventral pontine

Žnucleus; DLPsdorsolateral pontine nucleus; VLPsventrolateral pontine nucleus. Note that the electrode placements for two animals one paired and one.unpaired were in the middle cerebellar peduncle.

2.4. Tissue preparation

Paired and unpaired animals were sacrificed by decapi-tation under halothane anesthesia 24 h following cessationof training. Naive animals were also sacrificed on the sameday. Following quick removal, the brains were rapidly

Žfrozen in y208C isopentane. Coronal sections 10 mm.thick were cut in a cryostat throughout the rostral–caudal

extent of the cerebellum, thaw-mounted on chrom–alumgelatin coated slides, and kept at y708C until use forligand-binding experiments and histology. In addition, 40-mm sections were taken throughout the extent of thepontine nuclei for verification of electrode placement.

2.5. Ligand binding and autoradiography

2.5.1. AMPA bindingTissue sections were first equilibrated to room tempera-

ture and were then preincubated at either 358C or 08C inŽ .100-mM Tris–acetate buffer pH 7.4 containing 100 mM

EGTA for 45 min. Note that the pH of the Tris–acetatebuffer was adjusted to 7.4 at both preincubation tempera-tures. Sections were then incubated at 0–48C for 45 min in

Ž .50-mM Tris–acetate buffer pH 7.4 , 100 mM EGTA,containing 50 mM potassium thiocyanate and 100 nMw3 x Ž .H AMPA Spec. Act. 60 Cirmmol, NEN-Dupont .Non-specific binding was defined as the amount of

( )S.A. Hauge et al.rBrain Research 803 1998 9–1812

w3 xH AMPA binding measured in the presence of 1 mMquisqualate. Following the incubation, the sections werewashed at 0–48C in the incubation buffer twice for 10 s,once for 5 s in 50% incubation buffer, and then dippedthree times in distilled water. Sections were then rapidlydried under a stream of warm air and used for autoradiog-

w xraphy 69 .

2.5.2. CNQX bindingTissue sections were first equilibrated to room tempera-

ture and were then preincubated at either 358C or 08C inŽ .100-mM Tris–acetate buffer pH 7.4 containing 100 mM

EGTA for 45 min. Sections were then incubated at 0–48CŽ .for 45 min in 100 mM Tris–acetate buffer pH 7.4 , 100

mM EGTA, containing 100 mM glycine and 50 nMw3 x Ž .H CNQX Spec. Act. 18.3 Cirmmol, NEN-Dupont .Non-specific binding was defined as the amount ofw3 xH CNQX binding measured in the presence of 1 mMquisqualate. Following the incubation, the sections werewashed at 0–48C in the incubation buffer twice for 10 s,once for 5 s in 50% incubation buffer, and then dippedthree times in distilled water. Sections were then rapidlydried under a stream of warm air and used for autoradiog-

w xraphy 69 .

2.5.3. Autoradiography and statistical analysisŽ .Autoradiographic film Hyperfilm, Amersham, IL was

pressed against the tissue sections and against radioactivityŽ .standards ARC, St. Louis, MO for 2 weeks. Films were

developed at room temperature for 3–5 min in KodakGBX developer and fixer. Films were then analyzed with

Žthe aid of an image analysis system BRAIN software.running on the DUMAS system from Drexel University to

w3 x w3 xdetermine the amount of H AMPA or H CNQX bindingpresent within different regions of each tissue section aswell as the amount of binding to the entire section. Opticaldensities were converted into pmolrmg protein using thestandards present on the same film.

Binding ratios were calculated by dividing the amountof binding to a specific region by the amount of binding tothe whole section. Differences in binding ratios among all

Ž .three experimental groups naive, paired, and unpairedwere determined for each region by a One-way ANOVA

Žand LSD post-hoc analysis P values -0.05 were consid-. Ž .ered significant Statistica software, Statsoft, Tulsa, OK .

3. Results

All the animals in the paired group achieved criterion,defined as eight out of nine consecutive CRs, with an

Žaverage number of trials to criterion of 187.1 standard.error 34.5 . Pontine-electrode placements for animals in

the paired and unpaired groups are represented in Fig. 1.Upon reaching criterion, paired animals were overtrainedfor four sessions. Animals in the unpaired group exhibitedless than 10% CR performance.

Table 1w3 x w3 x ŽEffect of NMR conditioning on H AMPA and H CNQX binding 08C

. Žpreincubation to coronal sections of rabbit cerebellum from paired,.unpaired, and naive animals

Ligand Region Group

Paired Unpaired Naive3w xH AMPA left HVIrg 1.03"0.04 1.07"0.05 1.08"0.05

left HVIrm 1.50"0.14 1.59"0.06 1.70"0.10left lateral g 1.01"0.07 1.05"0.02 1.01"0.05left lateral m 1.57"0.19 1.70"0.05 1.80"0.12left medial g 0.96"0.06 1.07"0.05 1.15"0.05left medial m 1.54"0.12 1.73"0.09 1.87"0.09

3w xH CNQX left HVIrg 1.03"0.10 0.84"0.06 0.96"0.03left HVIrm 1.49"0.07 1.30"0.14 1.46"0.08left lateral g 1.01"0.09 0.82"0.07 0.91"0.01left lateral m 1.58"0.05 1.59"0.12 1.50"0.00left medial g 1.02"0.06 0.98"0.09 0.92"0.05left medial m 1.54"0.05 1.60"0.13 1.47"0.07

Values represent the amount of binding measured in a specific regiondivided by the amount of binding to the entire section, as described in

Ž . Ž .Section 2, and are means of five paired and naive and six unpairedanimals per group "S.E.M.HVIrgsHVI granular layer.HVIrmsHVI molecular layer.Lateral gsgranular layer lateral to HVI.Lateral msmolecular layer lateral to HVI.Medial gsgranular layer medial to HVI.Medial msmolecular layer medial to HVI.

w3 xNo significant differences in H AMPA binding wereŽobserved among any of the groups paired, unpaired, and

.naive in any of the regions measured following tissueŽ .preincubation at 08C Fig. 2A,C,E and Table 1 . With 358C

preincubation, however, significant differences inw3 x Ž .H AMPA binding were found in the trained left side ofthe cerebellum. More specifically, analysis of binding lev-

Ž Ž .els in the granular F 1,9 s 5.125606, P - 0.05,. Ž Ž .ANOVA and molecular F 1,9 s21.12522, P-0.0015,.ANOVA layers of left cerebellar cortical lobule HVI

w3 xshowed significant decreases in H AMPA binding inŽpaired animals as compared to unpaired 24% and 25%

.decreases in granular and molecular layers respectivelyŽ Ž .and in the molecular layer F 1,8 s10.53899, P-0.015,

. ŽANOVA as compared to naive animals 26% decrease;.Fig. 2B,D,F and Fig. 3A and Table 1 . Note that measure-

ments of binding in the molecular layer included bindingw3 xin Purkinje-cell bodies. H AMPA binding in cortical

regions adjacent to HVI in the left cerebellum, however,remained unchanged. No significant differences were ob-served in the left side of the cerebellum between animals

Žin the unpaired and naive groups Fig. 2B,D and Fig. 3A.and Table 1 . In addition, no significant differences in

w3 x Ž .H AMPA binding were seen on the untrained right sideof the cerebellum between any of the three groups in any

Ž .of the regions measured data not shown . Finally, now3 xsignificant differences in H CNQX binding were seen

between paired, unpaired, and naive animals in any regions

( )S.A. Hauge et al.rBrain Research 803 1998 9–18 13

w3 xFig. 2. Effects of classical conditioning on H AMPA binding to cerebellum determined under different preincubation conditions. Tissue sections wereŽ . Ž . w3 xpreincubated at either 08C A, C, and E or 358C B, D, and F , and H AMPA binding was performed as described in Section 2. At the end of the

Ž . Ž .incubation, sections were dried and exposed to autoradiographic film. Representative images from naive A and B , pseudo-conditioned C and D , andŽ .paired E and F rabbits are shown.

( )S.A. Hauge et al.rBrain Research 803 1998 9–1814

Fig. 3. Effects of classical conditioning on ligand binding properties of AMPA receptors. Autoradiograms similar to those shown in Fig. 2 werequantitatively analyzed with the aid of an image analysis system. Optical densities were converted to pmolrmg protein, and the results were expressed as

Ž . w3 xbinding ratios as described in Section 2. Results are mean"S.E.M. per group. A H AMPA binding measured in left cerebellar lobule HVI followingŽ . w3 x Ž .preincubation at 358C. B H CNQX binding measured in left cerebellar lobule HVI following preincubation at 358C. C Schematic representation of

Ž .areas quantified. ‘)’ indicates significance P-0.05 as compared to values in naive animals. HVIscortical lobule HVI; Gsgranular layer;Msmolecular layer; HVIrGsHVI granular layer; HVIrMsHVI molecular layer; Lateral Gsgranular layer lateral to HVI; Lateral Msmolecularlayer lateral to HVI; Medial Gsgranular layer medial to HVI; Medial Msmolecular layer medial to HVI.

( )S.A. Hauge et al.rBrain Research 803 1998 9–18 15

Ž .of the trained left side following preincubation at either 0Ž .or 358C Fig. 3B and Table 1 .

4. Discussion

The present results indicate that classical conditioningof the NMR is associated with a regionally selective

w3 xdecrease in H AMPA binding to synaptic subpopulationsw3 xof cerebellar AMPA receptors. Decreases in H AMPAw3 xbinding, without corresponding changes in H CNQX

binding, to these receptor subpopulations suggest that someproperties of existing receptors are modified due to NMR

w3 xconditioning. Significant differences in H AMPA bindingwere found between paired and unpaired animals andbetween paired and naive animals; differences were not

Žfound, however, between unpaired and naive animals see.Fig. 3A and C and Table 1 . They were also specifically

localized to the molecular layer, including the Purkinje cellbodies, and the granular layer of cortical lobule HVI on the

Ž .trained left side of the cerebellum. In particular, measure-w3 xments of H AMPA binding to cortical regions adjacent to

area HVI of the left cerebellum revealed no significantŽdifferences between the three groups see Fig. 2B,D,F and

.Fig. 3A and C . These results suggest, therefore, thatconditioning-induced changes in AMPA-receptor proper-ties were not the result of stress or of the stimulationparadigm but rather were caused by an associative processoccurring within left cerebellar cortical lobule HVI.

Different tissue preincubation temperatures have beenused in previous studies to analyze the properties of differ-

Žent subpopulations of AMPA receptors synaptic vs. non-. w3 x w3 xsynaptic labelled by H AMPA and H CNQX. Initial

studies revealed that preincubation of frozen–thawed ratbrain sections at 358C resulted in a significant decrease inw3 xH AMPA binding as compared to binding to sectionspreincubated at 08C. This decrease was attributed to a lossof high-affinity binding prevalent in the cell-body layersw x58,69 . More recent studies have shown that these high-af-

w3 xfinity H AMPA binding sites represent primarily non-synaptic, immature receptors located intracellularly in theendoplasmic reticulum or early Golgi apparatus. Con-

w3 xversely, H AMPA binding sites that are labelled follow-ing tissue preincubation at 358C represent primarily low

w xaffinity synaptic receptors 59 . Therefore, the decrease inw3 xH AMPA binding in left cerebellar cortical lobule HVIresulting from NMR conditioning reflects changes in low-affinity synaptic AMPA receptors.

The nature of the changes in these receptors is notclearly defined at this point. The lack of change inw3 xH CNQX binding under the same experimental condi-tions indicates that the effects are not simply due to a

Ž .decrease in receptor number see Fig. 3B . Significantw3 xdecreases in H AMPA binding could be attributed to a

change in the affinity of AMPA for the receptor resultingfrom a subtle conformational change affecting the recep-

tor’s ligand binding domain. They could also arise from amodification of the receptor affecting the coupling of thebinding of the ligand with the receptor’s channel domain.Agonist binding measured at equilibrium represents thecomplex endpoint of the ligand binding to the receptor’srecognition domain followed by transition of the receptorto a desensitized state. An antagonist, however, might label

w xa different state of the receptor 10 .It has previously been shown that astrocytes have

w xneurotransmitter receptors and transporters 62 . Morespecifically, Bergmann glia in the molecular layer of therat cerebellum exhibit dense staining when immunola-belled with antibodies against GluR1 and GluR4 but not

w xGluR2r3 subunits of the AMPA receptor 54 . Therefore,modulation of astrocytic AMPA receptors in left cerebellarcortical lobule HVI as a result of NMR conditioning mustalso be considered. Recently, it has been shown thatgene-knockout mice lacking glial fibrillary acidic proteinŽ .GFAP demonstrate a deficiency in LTD at Purkinjecell–parallel fiber synapses as well as significant impair-

w xments in eyeblink conditioning 56 , suggesting a role ofBergmann glia in this form of learning.

Morphological changes in Purkinje cells in the cerebel-lar cortex have been reported to result from ‘enriched’

w xmotor experience and motor learning tasks 2,17,31 . Theseinclude increases in synaptic structures on Purkinje andother neurons as well as other changes. At this point therelationship between these reported morphological changesand the currently reported neurochemical changes is notclear. However, both sets of data demonstrate learning-in-duced neuronal–synaptic plasticity in the cerebellar cortex.

w3 xSignificant increases in H AMPA binding in the hip-pocampus have been reported following induction of

w xlong-term potentiation in rats 68 and as a result of bothdelay and trace classical conditioning of the NMR in

w xrabbits 66,67 . Furthermore, it has been proposed thatw3 xthese observed increases in H AMPA binding were corre-

lated with increases in AMPA-receptor function within thew xhippocampus 44 . It is therefore tempting to suggest that

w3 xthe significant decreases in H AMPA binding observedin the present study are associated with a decreased func-tional response of AMPA receptors in select regions of thecerebellum. This would be consistent with the hypothesisthat cerebellar LTD, which results in a decrease in synapticefficacy at parallel fiber–Purkinje cell synapses mediatedby a change in AMPA-receptor properties, may be a formof synaptic plasticity that supports NMR conditioning inthe cerebellar cortex. It is not likely that decreasedw3 xH AMPA binding could be linked to modulation ofclimbing fiber–Purkinje cell synapses because, as previ-ously mentioned, neither glutamate nor currently identifiedAMPA receptors are thought to mediate synaptic transmis-sion at these synapses. It has been postulated that cerebel-lar LTD is caused in part by a change in the phosphoryla-tion of AMPA receptors on Purkinje cell dendrites. Thiswould result in AMPA-receptor desensitization followed

( )S.A. Hauge et al.rBrain Research 803 1998 9–1816

by a transition to a stable desensitized state establishingw xLTD 11,23 . In addition, activation of endogenous PLA2

is thought to play a role in the formation of cerebellar longw xterm depression 38 and has been shown to cause a

reduction of AMPA-receptor affinity under certain experi-w xmental conditions 3,45 . Possibly then, the modification of

ligand binding properties we observed is associated withthis postulated desensitized state of AMPA receptors. Inaddition, the observed differences in AMPA receptors inthe granular layer resulting from NMR conditioning sug-gest that there is a second site of synaptic modification thatoccurs with this form of learning at the mossy fiber–gran-ule cell synapse.

Cerebellar LTD has been implicated in several learningparadigms. It has been argued that adaptation of thevestibulo-ocular reflex occurs in the flocculus of the cere-

w xbellum as a consequence, in part, of LTD 22 . LTD hasbeen implicated in classical conditioning of the rabbitNMR as decreases in some Purkinje cell simple spikefiring rates in the CS period, particularly in lobule HVI,occur immediately prior to the appearance of CRsw x12,13,64 . As mentioned previously, results of studiesusing gene knockout mice, such as GFAP knockout micew x w x56 and PKCg knockout mice 4 , have indicated that theability to induce cerebellar LTD is associated with normalacquisition of the eyelid-nictitating membrane responsew x29,30 . In conclusion, modifications of synaptic cerebellarAMPA receptors in the molecular and granular layers ofthe trained side of the cerebellum resulted from the normalacquisition and retention of the NMR, thus supporting thehypothesis that cerebellar LTD is a form of synapticplasticity that underlies this form of learning.

Acknowledgements

We thank L. Chen and S. Bao of the University ofSouthern California and Dr. Cara Wellman of IndianaUniversity for comments on this manuscript. We alsothank Carol Dersarkissian for aid in surgery and behavioraltraining. This work was supported by National ScienceFoundation grant IBN-9215069, Office of Naval Researchgrant N00014-95-1152, National Institute of Mental Healthgrant 5P01-MH52194, and a grant from Sankyo to RichardF. Thompson.

References

w x1 A. Ajima, T. Hensch, R.T. Kado, M. Ito, Differential blockingaction of Joro spider toxin analog on parallel fiber and climbing

Ž .fiber synapses in cerebellar Purkinje cells, Neurosci. Res. 12 1991281–286.

w x2 B.J. Anderson, A.A. Alcantara, W.T. Greenough, Motor-skill learn-ing: changes in synaptic organization of the rat cerebellar cortex,

Ž .Neurobiol. Learn. Mem. 66 1996 221–229.

w x3 M. Baudry, G. Massicotte, S. Hauge, Opposite effects of phospholi-w3 xpase A2 on H AMPA binding in adult and neonatal membranes,

Ž .Dev. Brain Res. 61 1991 265–267.w x4 C. Chen, K. Masanobu, A. Abeliovich, L. Chen, S. Bao, J.J. Kim,

K. Hashimoto, R.F. Thompson, S. Tonegawa, Impaired motor coor-dination correlates with persistent multiple climbing fiber innerva-

Ž .tion in PKCg mutant mice, Cell 83 1995 1233–1242.w x5 L. Chen, S. Bao, J.M. Lockard, J.K. Kim, R.F. Thompson, Impaired

classical eyeblink conditioning in cerebellar-lesioned and PurkinjeŽ . Ž .cell degeneration pcd mutant mice, J. Neurosci. 16 1996 2829–

2838.w x6 G.A. Clark, D.A. McCormick, D.G. Lavond, R.F. Thompson, Ef-

fects of lesions of cerebellar nuclei on conditioned behavioral andŽ .hippocampal neuronal response, Brain Res. 291 1984 125–136.

w x7 M.J. Courtney, J.J. Lambert, D.J. Nicholls, The interactions betweenplasma membrane depolarization and glutamate receptor activationin the regulation of cytoplasmic free calcium in cultured cerebellar

Ž .granule cells, J. Neurosci. 10 1991 3873–3879.w x8 S.G. Cull-Candy, J.R. Howe, D.C. Ogden, Noise and single channels

activated by excitatory amino acids in cerebellar granule cells inŽ .explant culture, J. Physiol. London 400 1988 189–222.

w x9 E.D. D’Angelo, P. Rossi, J. Garthwaite, Dual-component NMDAŽ .receptor currents at a single central synapse, Nature 346 1990

467–470.w x10 B. Edmonds, A.J. Gibb, D. Colquhoun, Mechanisms of activation of

glutamate receptors and the time-course of excitatory synaptic cur-Ž .rents, Annu. Rev. Physiol. 57 1995 495–519.

w x11 J.C. Fiala, S. Grossberg, D. Bullock, Metabotropic glutamate recep-tor activation in cerebellar Purkinje cells as substrate for adaptivetiming of the classically conditioned eye-blink response, J. Neurosci.

Ž .16 1996 3760–3774.w x12 M.R. Foy, R.F. Thompson, Single unit analysis of Purkinje cell

discharge in classically conditioned and untrained rabbits, Soc.Ž .Neurosci. Abstr. 12 1986 518.

w x13 M.R. Foy, D.J. Krupa, J. Tracy, R.F. Thompson, Analysis of singleunit recordings from cerebellar cortex of classically conditioned

Ž .rabbits, Soc. Neurosci. Abstr. 18 1992 1215.w x14 M.E. Freeman, J.D. Lane, J.E. Smith, Turnover rates of amino acid

neurotransmitters in regions of rat cerebellum, J. Neurochem. 40Ž .1983 1441–1447.

w x15 J. Garthwaite, A.R. Brodbelt, Synaptic activation of N-methyl-D-aspartate and non-N-methyl-D-aspartate receptors in the mossy fibrepathway in adult and immature rat cerebellar slices, Neuroscience 29Ž .1989 401–412.

w x16 J. Garthwaite, A.R. Brodbelt, Glutamate as the principal mossy fibretransmitter in rat cerebellum: pharmacological evidence, Eur. J.

Ž .Neurosci. 2 1990 177–180.w x17 W.T. Greenough, G.S. Withers, B.J. Anderson, Experience-depen-

dent synaptogenesis as a plausible memory mechanism, in: I.Ž .Gormezano, E.A. Wasserman Eds. , Learning and Memory: The

Biological Substrates, Lawrence Erlbaum Associates, Hillsdale, NJ,1992, pp. 209–229.

w x18 T. Hirano, Synaptic transmission between rat inferior olivary neu-rons and cerebellar Purkinje cells in culture, J. Neurophysiol. 63Ž .1990 181–189.

w x19 I. Holopainen, M.O.K. Enkvist, K.E.O. Ackerman, Glutamate recep-tor agonists increase intracellular calcium independently of voltagegated Ca channels in rat cerebellar granule cells, Neurosci. Lett. 98Ž .1989 57–60.

w x20 D.B. Hudson, T. Valcana, G. Bean, P.S. Timiras, Glutamic acid: astrong candidate as the neurotransmitter of the cerebellar granule

Ž .cell, Neurochem. Res. 1 1976 73–81.w x21 M. Ito, The Cerebellum and Neural Control, Raven Press, New

York, 1984.w x22 M. Ito, Long-term depression as a memory process in the cerebel-

Ž .lum, Neurosci. Res. 3 1986 531–539.w x23 M. Ito, Cerebellar mechanisms of long-term depression, in: M.

( )S.A. Hauge et al.rBrain Research 803 1998 9–18 17

Ž .Baudry, R.F. Thompson, J.L. Davis Eds. , Synaptic Plasticity:Molecular, Cellular, and Functional Aspects, MIT Press, Cambridge,1993, pp. 117–128.

w x24 T. Kaneko, K. Itoh, R. Shigemoto, N. Mizuno, Glutaminase-likeimmunoreactivity in the lower brainstem and the cerebellum of the

Ž .adult rat, Neuroscience 32 1989 79–98.w x25 M. Kano, M. Kato, Mode of induction of long-term depression at

parallel fibre–Purkinje cell synapses in rabbit cerebellar cortex,Ž .Neurosci. Res. 5 1987 544–556.

w x26 M. Kano, M. Kato, Quisqualate receptors are specifically involvedŽ .in cerebellar synaptic plasticity, Nature 325 1987 276–279.

w x27 M. Kano, M. Kato, H.S. Chang, The glutamate receptor subtypemediating parallel fibre–Purkinje cell transmission in rabbit cerebel-

Ž .lar cortex, Neurosci. Res. 5 1988 325–337.w x28 K. Keinanen, W. Wisden, B. Sommer, P. Werner, A. Herb, T.A.

Verdoon, B. Sakmann, P.H. Seeburg, A family of AMPA-selectiveŽ .glutamate receptors, Science 249 1990 556–560.

w x29 J.J. Kim, L. Chen, S. Bao, W. Sun, R.F. Thompson, Geneticdissections of the cerebellar circuitry involved in classical eyeblinkconditioning, in: S. Nakanishi, A.J. Silva, S. Aizawa, M. KatsukiŽ .Eds. , Gene Targeting and New Developments in Neurobiology,Japan Scientific Societies Press, Tokyo, 1996, pp. 3–15.

w x30 J.J. Kim, R.F. Thompson, Cerebellar circuits and synaptic mecha-Ž .nisms involved in classical eyeblink conditioning, TINS 20 1997

177–181.w x31 J.A. Kleim, R.A. Swain, C.M. Czerlanis, J.L. Kelly, M.A. Pipitone,

W.T. Greenough, Learning-dependent dendritic hypertrophy of cere-bellar stellate cells: plasticity of local circuit neurons, Neurobiol.

Ž .Learn. Mem. 67 1997 29–33.w x32 T. Knopfel, E. Audinat, B.H. Gahwiler, Climbing fibre responses in

olivo-cerebellar slice cultures: I. Microelectrode recordings fromŽ .Purkinje cells, Eur. J. Neurosci. 2 1990 726–732.

w x33 T. Knopfel, I. Vranesic, C. Staub, B.H. Gahweiler, Climbing fibreresponses in olivo-cerebellar slice cultures: II. Dynamics of cytosolic

Ž .calcium in Purkinje cells, Eur. J. Neurosci. 3 1990 343–348.w x34 A. Konnerth, I. Llano, C.M. Armstrong, Synaptic currents in cere-

Ž .bellar Purkinje cells, Proc. Natl. Acad. Sci. U.S.A. 87 19902662–2665.

w x35 D.J. Krupa, J.K. Thompson, R.F. Thompson, Localization of aŽ .memory trace in the mammalian brain, Science 260 1993 989–991.

w x36 D.J. Krupa, R.F. Thompson, Reversible inactivation of the cerebellarinterpositus nucleus completely prevents acquisition of the classi-

Ž .cally conditioned eye-blink response, Learning Memory 3 1997545–556.

w x37 D.G. Lavond, J.E. Steinmetz, M.H. Yokaitis, R.F. Thompson, Reac-quisition of classical conditioning after removal of cerebellar cortex,

Ž .Exp. Brain Res. 67 1987 569–593.w x38 D.J. Linden, Phospholipase A2 controls the induction of short-term

versus long-term depression in the cerebellar Purkinje neurons inŽ .culture, Neuron 15 1995 1393–1401.

w x39 D.J. Linden, J.A. Conner, Long-term synaptic depression, Annu.Ž .Rev. Neurosci. 18 1995 319–357.

w x40 I. Llano, A. Marty, C.M. Armstrong, A. Konnerth, Synaptic- andagonist-induced excitatory currents of Purkinje cells in rat cerebellar

Ž .slices, J. Physiol. 434 1991 183–213.w x41 C.G. Logan, Cerebellar cortical involvement in excitatory and in-

hibitory classical conditioning, Stanford University, 1991, unpub-lished doctoral dissertation.

w x42 R.L. Makowiec, J.J. Cha, J.B. Penney, A.B. Young, Cerebellarexcitatory amino acid binding sites in normal, granuloprival, and

Ž .Purkinje cell deficient mice, Neuroscience 42 1991 671–681.w x43 W.F. Maragos, J.B. Penney, A.B. Young, Anatomic correlation of

w3 xNMDA and H -TCP-labelled receptors in rat brain, J. Neurosci. 8Ž .1988 493–501.

w x44 S. Maren, G. Tocco, S. Standley, M. Baudry, R.F. Thomspon,Postsynaptic factors in the expression of long-term potentiation

Ž .LTP : increased glutamate receptor binding following LTP induc-Ž .tion in vivo, Proc. Natl. Acad. Sci. U.S.A. 90 1993 9654–9658.

w x45 G. Massicotte, J. Bernard, M. Baudry, Postnatal changes in AMPAreceptor regulation by phospholipase A2 treatment of synaptic mem-branes: temporally differential effects of agonist and antagonist

Ž .binding, Dev. Brain Res. 66 1992 203–208.w x46 W.J. McBride, M.H. Aprison, K. Kusano, Contents of several amino

acids in the cerebellum, brain stem, and cerebrum of the ‘staggerer’,‘weaver’, and ‘nervous’ neurologically mutant mice, J. Neurochem.

Ž .26 1976 867–871.w x47 D.A. McCormick, R.F. Thompson, Cerebellum: Essential involve-

ment in the classically conditioned eyelid response, Science 223Ž .1984 296–299.

w x48 D.T. Monaghan, C.W. Cotman, Distribution of N-methyl-D-aspar-w3 xtate-sensitive L- H glutamate binding sites in rat brain, J. Neurosci.

Ž .5 1986 2909–2919.w x49 S. Nakanishsi, Molecular diversity of glutamate receptors and impli-

Ž .cations for brain function, Science 258 1992 597–603.w x50 T. Nieuwenhuys, J. Voogd, C. van Huijzen, The Human Central

Nervous System: A Synopsis and Atlas, 3rd edn., Springer, Berlin,1988.

w x51 J.M.M. Olson, J.T. Greenamyre, J.B. Penney, A.B. Young, Autora-diographic localization of cerebellar excitatory amino acid binding

Ž .sites in the mouse, Neuroscience 22 1987 913–923.w x52 S.P. Parrett, B.P. Ruiz, M.D. Mauk, Cerebellar cortex lesions disrupt

Ž .the timing of conditioned eyelid responses, J. Neurosci. 13 19931708–1718.

w x53 D.J. Perkel, S. Hestrin, P. Sah, R.A. Nicoll, Excitatory synapticŽ .currents in Purkinje cells, Proc. R. Soc. London B 241 1991

116–121.w x54 R.S. Petralia, R.J. Wenthold, Light and electron immunocytochemi-

cal localization of AMPA-selective glutamate receptors in the ratŽ .brain, J. Comp. Neurol. 318 1992 329–354.

w x55 M.E. Sandoval, C.W. Cotman, Evaluation of glutamate as a neuro-Ž .transmitter of cerebellar parallel fibers, Neuroscience 3 1978 199–

206.w x56 K. Shibuki, H. Gomi, L. Chen, S. Bao, J.J. Kim, H. Wakatsuki, T.

Fujisaki, K. Fujimoto, A. Katoh, T. Ikeda, C. Chen, R.F. Thompson,S. Itohara, Deficient cerebellar long-term depression, impaired eye-blink conditioning, and normal motor coordination in GFAP mutant

Ž .mice, Neuron 16 1996 587–599.w x57 R.A. Silver, S.F. Traynelis, S.G. Cull-Candy, Rapid-time-course

miniature and evoked excitatory currents at cerebellar synapses inŽ .situ, Nature London 355 1992 163–166.

w x58 S. Standley, N. Irvin, M. Baudry, Differential subcellular localiza-tion of two populations of glutamaterAMPA receptors in rat telen-

Ž .cephalon, Neurochem. Int. 25 1994 287–293.w x59 S. Standley, G. Tocco, N. Wagle, M. Baudry, Identification and

characterization of a new species of GluR2 subunit of AMPAŽ .receptors in rat forebrain, Soc. Neurosci. Abstr. 27 1997 369.

w x60 J.E. Steinmetz, D.J. Rosen, P.F. Chapman, D.J. Lavond, R.F.Thompson, Classical conditioning of the rabbit eyelid response witha mossy-fiber stimulation CS: I. Pontine nuclei and middle cerebel-

Ž .lar peduncle stimulation, Behav. Neurosci.. 100 1986 878–887.w x61 J.E. Steinmetz, Neuronal activity in the rabbit interpositus nucleus

during classical NM-conditioning with a pontine-nucleus-stimulationŽ .CS, Psychol. Sci. 1 1990 378–382.

w x62 V.I. Teichberg, Glial glutamate receptors: likely actors in brainŽ .signalling, FASEB J. 5 1991 3086–3091.

w x63 R.F. Thompson, The neurobiology of learning and memory, ScienceŽ .233 1986 941–947.

w x64 R.F. Thompson, Neural mechanisms of classical conditioning inŽ .mammals, Philos. Trans. R. Soc. London B 329 1990 161–170.

w x65 R.F. Thompson, D.J. Krupa, Organization of memory traces in theŽ .mammalian brain, Annu. Rev. Neurosci. 17 1994 519–549.

w x66 G. Tocco, K.K. Devgan, S.A. Hauge, C. Weiss, M. Baudry, R.F.

( )S.A. Hauge et al.rBrain Research 803 1998 9–1818

Thompson, Classical conditioning selectively increases AMPA re-Ž .ceptor binding in rabbit hippocampus, Brain Res. 559 1991 331–

336.w x67 G. Tocco, A.J. Annala, M. Baudry, R.F. Thompson, Learning of a

hippocampal-dependent conditioning task changes the binding prop-erties of AMPA receptors in rabbit hippocampus, Behav. Neural

Ž .Biol. 58 1992 222–231.w x68 G. Tocco, S. Maren, T.J. Shors, M. Baudry, R.F. Thompson,

w3 xLong-term potentiation is associated with increased H AMPAŽ .binding in rat hippocampus, Brain Res. 573 1992 228–234.

w x69 G. Tocco, G. Massicotte, S. Standley, R.F. Thompson, M. Baudry,Effect of temperature and calcium on the binding properties of theAMPA receptor in frozen rat brain sections, Eur. J. Neurosci. 4Ž .1992 1093–1103.

w x70 J.A. Vignon, A. Privat, I. Chaudieu, A. Thierry, J.M. Kamenka, R.

w3 x Žw3 x .Chicheportiche, H Thienyl-phencyclidine H TCP binds two dif-ferent sites in rat brain: localization by autoradiographic and bio-

Ž .chemical techniques, Brain Res. 378 1986 133–141.w x71 C.H. Yeo, M.J. Hardiman, M. Glickstein, Classical conditioning of

the nictitating membrane response of the rabbit: I. Lesions of theŽ .cerebellar nuclei, Exp. Brain Res. 60 1985 87–98.

w x72 C.H. Yeo, M.J. Hardiman, M. Glickstein, Classical conditioning ofthe nictitating membrane response of the rabbit: II. Lesions of the

Ž .cerebellar cortex, Exp. Brain Res. 60 1985 99–113.w x73 C.H. Yeo, Cerebellum and classical conditioning of motor response,

Ž .Ann. New York Acad. Sci. 627 1991 292–304.w x74 A.B. Young, M.L. Oster-Granit, R.M. Herdon, S.H. Snyder, Glu-

tamic acid: selective depletion by viral induced granule cell loss inŽ .hamster cerebellum, Brain Res. 73 1974 1–13.