nature neuroscience • volume 4 no 3 • march 2001 289

articles

Distinct cellular and molecular mechanisms underlie differenttemporal phases of hippocampal LTP: induction requires elevat-ed postsynaptic calcium1, early maintenance is mediated by theactivation of protein kinases2, and later phases depend on genetranscription3,4 and protein synthesis5–7. Consistent with the pre-sumption that synaptic plasticity provides the neural basis forlong-term information storage in the brain, several forms of long-term memory involving both hippocampal and cortical struc-tures also depend on protein synthesis8–10.

Following induction of LTP in the dentate gyrus, a cascade ofgenes is activated at different time points11,12. The transient acti-vation of immediate early genes is believed to be a critical stepin these successive waves of gene expression. Zif268 (also knownas Krox-24, EGR-1 or NGFI-A) is one such immediate early gene,encoding a zinc finger transcription factor13–15. In the dentategyrus, Zif268 mRNA is transiently upregulated between ten min-utes and two hours following LTP-inducing stimulation16–18, atime course that precludes a role for Zif268 protein in the earlieststages of LTP, but that is consistent with an involvement in thetransition from the early phase of LTP to the protein synthesis-dependent late phase19. Expression is also upregulated in the insu-lar cortex following LTP-inducing stimulation of afferents fromthe basolateral amygdala20, suggesting a role for Zif268 in activ-ity-dependent neuronal function in other brain regions.

The association between Zif268 activation and LTP has beenknown for over a decade. More recent studies have provided evi-dence that Zif268 mRNA is upregulated in the hippocampus inthe behaving rat following exposure to a novel stimulus21, andin the inferior temporal gyrus of the monkey during associativelearning22. However, it remains uncertain whether the activationof Zif268 is necessary for the generation of late LTP, or for theformation of long-term memories. To address these questions,

we examined short- and long-term synaptic plasticity in the den-tate gyrus of mice with a targeted inactivation of the Zif268gene23, using both anesthetized and freely moving animals. Wehave also examined a variety of hippocampus-dependent andindependent tasks that make demands on short- and/or long-term memory. Our results demonstrate that activation of Zif268is required for the expression of long-term but not short-termsynaptic and behavioral plasticity.

RESULTSHippocampal anatomy in Zif268–/– miceWe used several histochemical and immunohistochemical mark-ers to examine hippocampal anatomy in control and mutantmice. Nissl staining (data not shown) and NeuN immunoreac-tivity (which labels a neuronal-specific DNA-binding protein24;Fig. 1a) were indistinguishable in wild-type (Zif268+/+) andhomozygous mutant (Zif268–/–) mice. Parvalbumin labels a cal-cium-binding protein in a subpopulation of GABAergic interneu-rons in the hippocampus25, again showing similar staining inboth Zif268+/+ and Zif268–/– mice (Fig. 1b). Synaptophysinimmunoreactivity in synaptic layers also appeared normal inZif268–/– mice (Fig. 1c). Thus, basic neuronal architecture wasnot affected by the Zif268 mutation.

Short-term plasticity and LTP in anesthetized miceWe measured LTP and short-term plasticity of both the field exci-tory postsynaptic potential (fEPSP) and population spike in ure-thane-anesthetized mice. Paired-pulse stimulation (inter-pulseinterval, 10–100 ms) at intensities sub-threshold for a popula-tion spike resulted in facilitation of the fEPSP, maximal at inter-pulse intervals of 10–20 ms, in both Zif268+/+ and Zif268–/– mice(Fig. 2a). Paired-pulse facilitation of the fEPSP is attributable to

A requirement for the immediateearly gene Zif268 in the expressionof late LTP and long-term memories

M. W. Jones1, M. L. Errington1, P. J. French1, A. Fine1, T. V. P. Bliss1, S. Garel2, P. Charnay2, B. Bozon3, S. Laroche3 and S. Davis3

1 Division of Neurophysiology, National Institute for Medical Research, Mill Hill, London NW7 1AA, UK

2 Biologie Moléculaire du Développement, INSERM U-368, Ecole Normale Supérieure, 75230 Paris, France

3 Laboratoire de Neurobiologie de l’Apprentissage, de la Mémoire et de la Communication, CNRS UMR 8620, Université Paris-Sud, 91405 Orsay, France

Correspondence should be addressed to S.D. (sabrina.davis@ibaic.u-psud.fr)

The induction of long-term potentiation (LTP) in the dentate gyrus of the hippocampus is associatedwith a rapid and robust transcription of the immediate early gene Zif268. We used a mutant mousewith a targeted disruption of Zif268 to ask whether this gene, which encodes a zinc finger transcrip-tion factor, is required for the maintenance of late LTP and for the expression of long-term memory.We show that whereas mutant mice exhibit early LTP in the dentate gyrus, late LTP is absent whenmeasured 24 and 48 hours after tetanus in the freely moving animal. In both spatial and non-spatiallearning tasks, short-term memory remained intact, whereas performance was impaired in testsrequiring long-term memory. Thus, Zif268 is essential for the transition from short- to long-termsynaptic plasticity and for the expression of long-term memories.

©20

01 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://n

euro

sci.n

atu

re.c

om

© 2001 Nature Publishing Group http://neurosci.nature.com

290 nature neuroscience • volume 4 no 3 • march 2001

presynaptic mechanisms26. Paired-pulse stimulation at intensi-ties just above threshold for evoking a population spike resultedin complete suppression of the population spike at short inter-stimulus intervals (10–25 ms) and spike facilitation at longerintervals in both Zif268+/+ and Zif268–/– mice (Fig. 2b). Spikefacilitation peaked at inter-pulse intervals of 60–80 ms in all micetested. This profile of spike depression at short inter-stimulusintervals followed by facilitation at longer intervals reflects recur-rent inhibition and disinhibition and is therefore, in part, a post-synaptic phenomenon. Thus, short-term presynaptic plasticityand network excitability in the dentate gyrus seem to be normalin Zif268 mutant mice.

Following tetanic stimulation, both genotypes showed significant and similar potentiation of the population spike50–60 minutes after tetanus (2.8 ± 0.5 mV increase in Zif268+/+

and 2.8 ± 0.7 mV in Zif268–/– mice, p < 0.05 versus baseline;Fig. 2c). LTP of the fEPSP slope was also similar in the twogenotypes (23 ± 5.3% in Zif268+/+ and 21 ± 6.1% in Zif268–/–

mice, p < 0.05 with respect to baseline; Fig. 2d).

Long-lasting synaptic plasticity in freely moving miceWe investigated LTP over several days in freely moving mice. Teststimulus intensities required to evoke responses with comparablefEPSP slopes and population spike amplitudes were not differ-ent between genotypes (Fig. 3a). Following two days of baselinerecording, tetanic stimulation of the perforant path induced sig-nificant potentiation of the population spike in all genotypes (p < 0.05 versus baseline for spike amplitude 50–60 minutes aftertetanus in Zif268+/+ and Zif268–/– mice, p < 0.01 in Zif268+/–

mice). The decrease in the latency of the population spike aftertetanic stimulation made it impractical to compare the initialslope of the fEPSP before and after LTP induction (Fig. 3a). Thus,as in anesthetized mice, potentiation in mutants was similar tothat in wild types in the first hour following induction. Wild-type mice maintained significant spike potentiation 24 and 48 hours following the tetanic stimulation (p < 0.05 versus con-trol). In contrast, neither Zif268+/– nor Zif268–/– mice supportedsignificant potentiation at these time points (p > 0.06 versus con-trol values, p < 0.05 versus Zif268+/+ mice; Fig. 3b). Thus, theZif268 gene is necessary for the expression of the later phases ofLTP in the dentate gyrus.

Lack of expression of Zif268 in the hippocampusWe examined levels of Zif268 mRNA in Zif268+/+ and Zif268–/–

mice by in situ hybridization. The results confirmed a completeabsence of Zif268 mRNA in the Zif268–/– mice, whereas in the

articles

Fig. 1. Gross hippocampal anatomy is normal in Zif268–/– mice. (a)NeuN immunoreactivity, (b) parvalbumin staining and (c) synaptophysinimmunoreactivity all appeared similar in Zif268–/– (right) and Zif268+/+

(left) mice.

Fig. 2. Short-term synaptic plasticity and LTP in the dentate gyrus ofanesthetized Zif268 mutant mice. (a) At low stimulus intensities,paired-pulse facilitation of EPSP amplitude shows the same dependenceon inter-stimulus interval (ISI) in Zif268–/– (�, n = 5) and Zif268+/+ mice(�, n = 4). Sample responses show facilitation at 15 ms ISI in a Zif268–/–

mouse. (b) At stimulus intensities just above threshold for evoking apopulation spike, paired-pulse depression and facilitation of the popula-tion spike also show the same dependence on ISI in Zif268–/– andZif268+/+ mice. Sample responses show depression (10 ms ISI) andfacilitation (50 ms ISI) from a Zif268–/– mouse. (c, d) LTP of the popula-tion spike (c) and the field EPSP (d) in Zif268–/– and Zif268+/+ mice (�,n = 5, and �, n = 4, respectively). Mean changes in the amplitude of thepopulation spike (c), and percentage changes in the slope of the fieldEPSP (d), plotted with respect to the mean values in the 10 min pre-ceding tetanic stimulation (arrows). Test stimuli were given to the per-forant path at a frequency of 1 per 30 s. A similar level of potentiationof the population spike and EPSP was seen in Zif268+/+ and Zif268–/–

mice for the first hour after tetanus.

a b

c

d

a

b

c

©20

01 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://n

euro

sci.n

atu

re.c

om

© 2001 Nature Publishing Group http://neurosci.nature.com

Zif268+/+ mice, Zif268 expression was normal and LTPinducible (Fig. 4). The construct used to inactivateZif268 in the mutant mice23 involved the insertion ofa lacZ cassette downstream of the promoter; in situhybridization demonstrated normal constitutive andLTP-regulated expression of the lacZ gene in theZif268–/– mice (Fig. 4), suggesting that the activity ofthe Zif268 promoter was not affected by the mutation.This was confirmed by in situ hybridization with a com-mon probe derived from the 5´ untranslated region ofZif268, and located upstream of the site of insertion oflacZ. As expected, the common probe detected equiva-lent levels of LTP-inducible mRNAs in Zif268+/+ andZif268–/– mice (Fig. 4).

Learning and memory deficits in Zif268 mutant miceWe next examined whether the deficits in synaptic plasticity inZif268 mutant mice, particularly evident in the late phases of LTP,were paralleled by corresponding deficits in learning and memory.

We first evaluated short-term, working memory by testing spon-taneous alternation in a T-maze, a spatial behavior that dependson the integrity of the hippocampus27. All three genotypes (n = 6per genotype) showed normal alternation between the arms visit-ed on the first and second runs of a trial when a 30-second delaywas imposed between runs (Zif268+/+, 75 ± 6.8%; Zif268+/–, 76 ± 3.0%; Zif268–/–, 69 ± 5.1% alternation; p < 0.05 versuschance in each case). Mice were also tested with a ten-minutedelay between runs; again, there was no significant difference inthe levels of alternation among genotypes (Zif268+/+, 70 ± 4.4%;Zif268+/–, 71 ± 7.7%; Zif268–/–, 60 ± 6.1%; n = 12 per group).All mice made a similar mean number of entries into the armsduring habituation, indicating a similar level of explorato-ry behavior and locomotion. This type of spatial workingmemory can therefore occur in the absence of Zif268

nature neuroscience • volume 4 no 3 • march 2001 291

expression. To assess the role of Zif268 in longer-lasting formsof spatial and associative memory, we trained mice in four otherbehavioral tasks.

First, we tested spatial learning and memory in an open-fieldwater maze, using a massed training protocol where the acquisi-tion phase was conducted within a two-hour period. All genotypestook the same time to find the escape platform in the first trainingtrial (Fig. 5a). As the training progressed, there was a reduction inescape latencies for all three genotypes. However, both Zif268+/–

and Zif268–/– mice took significantly longer than Zif268+/+ mice,suggesting they had a learning deficit (F2,24 = 8.14; p < 0.01). A cor-responding memory deficit was evident in a probe trial done 48 hours after training, during which Zif268+/+ mice showed a sig-nificant preference for the target quadrant (F3,32 = 6.01, p < 0.01),whereas Zif268+/– and Zif268–/– (F < 1 for both genotypes) miceshowed no spatial bias for the training quadrant (Fig. 5b). Swimspeeds were similar for all genotypes, and mice did not exhibitfloating behavior. The deficits observed in the acquisition and

articles

Fig. 3. LTP in the dentate gyrus of awake Zif268 mutantmice. (a) Stimulus intensity required to evoke field potentialswith comparable EPSP slopes and population spike ampli-tudes was independent of genotype. Sample responses areshown for mice of each genotype before (dotted line) and 10 min after (solid line) tetanic stimulation. Values for themean amplitudes (± s.e.m.) of the population spike are given1 h and 48 h after the tetanus (*p < 0.05, **p < 0.01 com-pared to pre-tetanus amplitudes). (b) Time course of LTP inthe awake animal. Baseline responses were recorded for 2 days (20 min per day). A tetanus was delivered (arrow) onthe second day, and responses were monitored for 1 h, andagain for 20 min per day for the next 2 days. Zif268+/+ (�, n= 7), Zif268+/– (�, n = 11) and Zif268–/– (�, n = 6) mice allshowed significant potentiation of the population spike 1 hafter tetanic stimulation, but only Zif268+/+ mice maintainedthis potentiation over the following 48 h.

Fig. 4. Expression of Zif268 in the dentate gyrus. In situ hybridiza-tion shows that mRNAs for Zif268 (in Zif268+/+ mice) and lacZ(in Zif268–/– mice) are expressed in the dentate gyrus 1 h follow-ing LTP-inducing stimulation of the perforant path. (Comparehybridization signal in the left (tetanized) and right (untetanized)dentate gyrus in each case.) A hybridization probe for a region ofthe Zif268 gene transcribed in both Zif268–/– and Zif268+/+ mice(‘common’) revealed similar expression in mice of both geno-types (right).

a

b

©20

01 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://n

euro

sci.n

atu

re.c

om

© 2001 Nature Publishing Group http://neurosci.nature.com

292 nature neuroscience • volume 4 no 3 • march 2001

retention (and/or recall) of this task could not therefore beexplained by aberrant motor behavior.

Complex tasks can often be more efficiently learned withextended and distributed training28. Using a protocol in whichtraining was distributed over ten days, all genotypes showed sim-ilar acquisition curves (Fig. 5c), and all groups spent more timein the training quadrant in a probe trial given eight days later (allp-values < 0.01; Fig. 5d). Thus, with extended training, Zif268mutant mice show normal acquisition and long-term recall.

As spatial learning takes place over many trials, we examinedperformance in three non-spatial tasks in which learning is

achieved after a single trial (conditioned taste aversion and socialtransmission of food preference) or after a brief training period(novel object recognition). In conditioned taste aversion, water-deprived mice learn to associate a novel taste (15% sucrose solu-tion) with the malaise induced by injection of lithium chloride29.When subsequently offered the choice between water and sucrosesolutions, animals tend to avoid the sucrose. This task is not hip-pocampus-dependent, but requires structures including the baso-lateral amygdala and insular cortex. Zif268 mRNA is upregulatedfollowing induction of LTP in the pathway connecting these twostructures20. We found that 24 hours after LiCl injection, miceshowed significant aversion to the sucrose (p < 0.05 versus NaCl-injected Zif268+/+ mice; Fig. 6). However, neither Zif268+/– norZif268–/– mice exhibited significant aversion at 24 hours (p > 0.1).Analysis of variance confirmed a significant difference in aver-sion index between genotypes (F2,24 = 9.58; p < 0.01 for threeLiCl-injected groups). The same effects were also observed 48 hours after LiCl injections (data not shown). The three groupsthat were injected with NaCl showed no preference for sucroseor water, either 24 or 48 hours after injection.

These data show that Zif268 is necessary for generating along-term memory of an association formed in a single trial.Because of the malaise induced by injections of LiCl, we wereunable to test the mice at short intervals, so we could not deter-mine whether short-term memory of the association was affect-ed in Zif268 mutants. Therefore, in two other experiments, weused a within-animal protocol to test for both short- and long-term memory. In the first task, we used social transmission offood preference to test one-trial learning both immediately and24 hours after learning. This is an olfactory discrimination taskin which rodents show preference for a novel food that hasrecently been smelled on the breath of another (demonstrator)animal30. Food-deprived demonstrator mice were given eithercocoa- or coriander-scented food to eat over a period of twohours and then allowed to interact with observer mice for 20 minutes. Food-deprived observer mice were then given free

articles

Fig. 5. Spatial navigation in the watermaze. (a) During massed training, allmice took the same amount of timeto escape the water on the first trial(Tr1). During acquisition, all micelearned to locate the hidden platform,although the Zif268+/– (�, n = 9) andZif268–/– (�, n = 9) were slower tolearn the task than the Zif268+/+ mice(�, n = 9). (b) In a probe trial given 48 h later, Zif268+/+ mice showed aspatial bias for the training quadrant(*p < 0.01), but this was not the casewith Zif268+/– or Zif268–/– mice (F < 1for both genotypes), which distrib-uted their time equally among the 4quadrants. (c) If extended and distrib-uted training was given, all mice(Zif268+/+, n = 11; Zif268+/–, n = 13;Zif268–/–, n = 10) learned the task atsimilar rates and (d) showed reten-tion of the learning in a probe trialgiven 8 days later (*p < 0.01).

Fig. 6. Conditioned taste aversion. Aversion indices (volume sucroseconsumed/total volume consumed; mean ± s.e.m.) for Zif268+/+ (n = 8),Zif268+/– (n = 10) and Zif268–/– (n = 10) mice, following exposure toLiCl or to NaCl. When sucrose was followed by an intraperitonealinjection of lithium chloride, Zif268+/+ mice avoided sucrose and prefer-entially drank water 24 h after conditioning (*p < 0.05). In contrast, nei-ther Zif268–/– nor Zif268+/– mice showed a significant aversion tosucrose at 24 h. Control mice in which drinking the novel sucrose solu-tion was followed by a sodium chloride injection (n = 4 for each geno-type) drank similar volumes of sucrose and water.

a b

c d

©20

01 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://n

euro

sci.n

atu

re.c

om

© 2001 Nature Publishing Group http://neurosci.nature.com

nature neuroscience • volume 4 no 3 • march 2001 293

choice between cocoa and coriander either 30 secondsafter the interaction, or 24 hours later. Thirty secondsafter the interaction, all genotypes showed a significantpreference for the food they had smelled on the breath of thedemonstrators (all p-values < 0.05; Fig. 7a). Twenty-four hourslater, however, whereas Zif268+/+ mice still showed a preferencefor the demonstrated food, Zif268+/– and Zif268–/– mice con-sumed similar amounts of both food types (Fig. 7b). Analysisof the consumption of each scented food (F1,53 = 1.05; p > 0.05)showed no difference among genotypes, and the amount of foodconsumed by demonstrators (F2,52 = 1.95; p > 0.05) was equiv-alent, suggesting that there was no bias toward a particular food,and that the demonstrators for each genotype consumed enoughfood to transmit the smell. These data show that Zif268 is impor-tant for the formation of long-term memory for an olfactoryevent, whereas short-term memory for the same event is notdependent on Zif268.

Rodents tend to explore a novel object in preference to afamiliar object. Novel object recognition is hippocampus depen-dent31, and we used this as a second task in which to assess bothshort- and long-term memory in the same groups of animals.Mice were allowed to explore 2 objects for 20 minutes. Followinga 10-minute or 24-hour delay, one of the familiar objects wasreplaced with a novel object, and the time spent exploring thenovel and familiar objects was measured. At the ten-minutedelay, mice of all genotypes spent significantly longer exploringthe novel object than the familiar object (all p-values < 0.05; Fig. 7c), indicating similar levels of motivation and short-termmemory. In contrast, with a 24-hour interval between explor-ing the familiar and novel objects, Zif268+/+ and Zif268+/– micestill preferentially explored the novel object (p < 0.01) whereasZif268–/– mice did not (p > 0.05; Fig. 7d). Thus, as with socialtransmission of food preference, Zif268+/+ mice were able toretain information acquired 24 hours earlier, whereas Zif268–/–

mice could not. The performance of Zif268+/– mice was againintermediate, although, in contrast to their performance in socialtransmission of food preference, they exhibited significant reten-tion at 24 hours.

DISCUSSIONIn this paper, we demonstrate an absence of late phase LTP inthe dentate gyrus of freely moving mice with a targeted inacti-vation of the immediate early gene Zif268. The observation thatLacZ reporter mRNA is upregulated after tetanic stimulation sug-gests that signaling events upstream of Zif268 transcription arenot affected in mice with the mutant gene. We conclude that theabsence of late LTP is due to a failure in the synthesis of down-stream effector proteins encoded by genes for which Zif268 is anobligatory transcription factor.

As a consequence of pituitary and ovarian defects, homozy-gous Zif268 mutant mice have a reduced body size and are ster-ile, although heterozygous mice are phenotypically normal interms of size and fertility23. However, LTP decayed at a similarrate in both heterozygous and homozygous mice, which alsoshowed similar deficits in two of the learning tasks that placeddemands on long-term memory, suggesting that endocrine dys-function does not contribute to the observed phenotype. Histo-logical examination using cellular, neuronal and presynapticmarkers confirmed similar cell densities and hippocampal archi-tecture in wild-type and mutant mice, showing that disruption ofZif268 had no gross effects on hippocampal circuitry.

Basal synaptic transmission, neuronal excitability and short-term plasticity were normal in mutant mice. All genotypesshowed an equivalent magnitude of LTP for the first hour fol-lowing its induction. However, whereas LTP was maintained forat least 48 hours after induction in wild-type mice, LTP in bothheterozygous and homozygous mutant mice decayed to base-line within 24 hours. The level of Zif268 mRNA in heterozygousanimals is approximately half that seen in wild-type mice (datanot shown), suggesting that this is insufficient to achieve thelevels of Zif268 activation required for the successful expressionof late LTP.

Other immediate early genes have also been implicated in LTPand learning. For example, mRNA for Homer, a protein that

articles

Fig. 7. Discrimination learning at short and long delays.Olfactory discrimination in social transmission of food prefer-ence (a, b) and object recognition (c, d) is normal at shorttime intervals but impaired at long intervals in Zif268 mutantmice. (a) Preference index exhibited by the observer mice forthe scented food eaten by the demonstrator mouse (demon-strated food consumed/total food consumed) at 30-s delay (a)and 24-h delay (b) after interaction between demonstrator andobserver mice. All mice (n = 10 for each genotype) learned thediscrimination when there was minimal delay (*p < 0.05),showing a significant preference above chance level for thefood eaten by the demonstrator mice. When a 24-h delay wasimposed, only Zif268+/+ mice (n = 10; *p < 0.05) retained thememory for the odor. Neither Zif268+/– (n = 7) nor Zif268–/–

(n = 9) mice showed a significant preference for the flavor towhich they had been exposed during the interactive period. (c, d) Percent time spent exploring the novel object (timespent exploring the novel object/total time × 100) at a 10-minand 24-h delay in the object recognition task. (c) All mice (n = 8for each genotype) spent significantly more time exploring thenovel object at the 10 min delay (*p < 0.05). (d) After a 24-hdelay, Zif268+/+ and Zif268+/– mice spent more time exploringthe novel object (*p < 0.05), whereas Zif268–/– mice no longershowed a preference for the novel object.

a b

dc

©20

01 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://n

euro

sci.n

atu

re.c

om

© 2001 Nature Publishing Group http://neurosci.nature.com

294 nature neuroscience • volume 4 no 3 • march 2001

binds metabotropic glutamate receptors, is upregulated followingLTP induction32,33. Arc (also known as Arg 3.1), another mRNAspecies upregulated after LTP34,35, is necessary for both late LTPand spatial learning36, as shown by antisense techniques. Togeth-er with our results, these data suggest that several immediate earlygenes act together to establish late LTP.

To examine the involvement of Zif268 in short- and long-term memory, we used a variety of behavioral tasks makinguse of single or repeated training, different types of reinforce-ment and the processing of spatial or non-spatial information.Some of these tasks are hippocampus dependent, and some arenot. Our results provide evidence that at least some forms ofshort-term memory are intact in the Zif268 mutant mice,because they demonstrated normal levels of spontaneous alter-nation, an innate behavior that relies upon spatial workingmemory. In common with early LTP, spontaneous alternationrequires activation of the NMDA receptor and phosphoryla-tion of protein kinase C37, events that are upstream of genetranscription, and that seem to be sufficient to mediate short-term memory. Consistent with the notion that loss of func-tional Zif268 does not affect short-term memory processes,Zif268 mutant mice were also able to perform olfactory dis-crimination in social transmission of food preference, and visu-al discrimination in an object recognition task, when minimaldelay was imposed.

In contrast to the lack of effect on short-term memory, long-term memory in Zif268 mutant mice was severely impaired. Inthree forms of learning, conditioned taste aversion, olfactorydiscrimination and novel object recognition, homozygous miceexhibited no significant recall when tested 24 hours later. Thesefindings support the notion that Zif268 is critical in the expres-sion of the long-term memory trace. Spatial navigation is a morecomplex type of learning, requiring several trials to reach crite-rion. During massed training, both heterozygous and homozy-gous mice showed impaired acquisition and a severe deficit inlong-term spatial memory assessed in probe trials 24 hours later,suggesting that Zif268 is required either for memory consolida-tion during and following the process of learning or for retrieval.We found that spatial memory deficits in Zif268 mutant micecould be completely rescued by extended and distributed train-ing. Similar deficits in the consolidation of learning and thepotential for rescue by extended training have been reported inmice with targeted disruption of the gene encoding the cAMPresponse element binding protein CREB38. In addition, geneticstudies in Drosophila have suggested that the ability to form long-term memories following disruption of CREB signaling is influ-enced by the temporal structure of the training schedule39. Theupstream regulatory elements of the Zif268 gene include sixserum response elements (SRE) and two cAMP response ele-ments (CRE)40,41, and LTP-dependent transcriptional regula-tion of Zif268 is controlled by the mitogen-activated proteinkinase (MAPK) pathway42. Both CREB and MAPK signalinghave been implicated in synaptic plasticity43,44 and certain typesof learning45,46. Our data therefore strengthen the evidence thatZif268 is an essential participant in the signaling cascade requiredfor synaptic and behavioral plasticity, with the proviso that inthe absence of Zif268, other unidentified signaling pathways canbe recruited under certain conditions (such as distributed train-ing) to allow information to enter an accessible long-term mem-ory store.

In summary, activation of Zif268 is essential for stabilizingsynaptic plasticity in the hippocampus and for the expression ofhippocampal and non-hippocampal forms of long-term mem-

ory. Our results also establish an isomorphism between hip-pocampal LTP and hippocampus-dependent learning withrespect to the effects of the Zif268 mutation: short-term plastic-ity and short-term memory are unaffected, whereas long-termplasticity and long-term memory are impaired.

METHODSAnimals. Mice were generated using 129/SV ES cells injected intoC57BL/6J blastocysts and backcrossed onto a C57BL/6J background23.The mutation involved insertion of a lacZ-neo cassette between pro-moter and coding sequence, and addition of a frameshift mutation atthe level of an Ndel restriction site corresponding to the beginning of theDNA-binding domain. Age-matched (2–8 month old) Zif268+/+,Zif268+/– and Zif268–/– littermates were tested in several behavioraltasks, with the sequence of tasks randomized. A subset of mice was usedfor electrophysiology. All procedures carried out at CNRS were con-ducted in accordance with recommendations of the EU (86/609/EEC)and the French National Committee (87/848) and those at NIMR inaccordance with the UK Animals (Scientific Procedures) Act (1986).

Spontaneous alternation. Mice were habituated to the T-maze for 10–20 min/day for two days. During testing (4–8 trials/day for threedays), they were placed in the start box for 30 s and then given a forcedchoice run (randomly assigned). On entering the open arm, they wereheld there for 30 s, then returned to the start box, held for 30 s or 10 minand released, with both arms accessible. The percent number of trials onwhich mice alternated was subjected to angular transformation and dif-ferences analyzed with Student’s t-test.

Spatial navigation. The water maze (1.5 m diameter) contained an escapeplatform (10 cm) in a fixed position. After a 1-day habituation period(2 blocks of 4 trials with the hidden platform in the center of the maze)massed trials were given the following day (5 blocks of 5 trials; 120 s inter-trial interval; 60 s maximum swim time; 15 min inter-block interval). A60 s probe trial was run 48 h post training. In distributed training, micewere given 1 block of 4 pre-training trials, and, during acquisition, 2 blocks (5-h interval) of 4 trials (60-s interval; maximum swim time of90 s) a day for 10 days. A 90-s probe trial was given 8 days later. ANOVAwas conducted on escape latencies (acquisition) and time in each quad-rant (probe trial).

Conditioned taste aversion. Water-deprived mice were trained for 3 daysto drink ad libitum from 2 identical water bottles in the home cage (30 min/day). Bottles were weighed to measure consumption. On theconditioning day, mice received 15% sucrose from 2 identical bottles for30 min. One hour later, mice were either injected with 0.9% saline, or0.3 M LiCl (10% body weight i.p.). At 24 and 48 h after conditioning,mice could choose between water and sucrose for 30 min. The aversionindex (volume of sucrose consumed × 100/total volume consumed) wassubjected to angular transformation and analyzed with Student’s t-test(aversion index) and ANOVA (group difference).

Social transfer of food preference. Mice were habituated to test cagesand placed on a food-restricted diet for several days. On test days,demonstrators were given scented food (coriander, 0.3%, or bitter cocoa,2.0% by weight) for 2 h. Observers were then exposed to demonstra-tors for 20 min, and placed in test cages either 30 s or 24 h later, wherethey had free access to both foods for 20 min. Both male and femalemice were used in this task, and as performance was similar betweensexes, animals were grouped for analysis. Food consumed was calculat-ed by weighing, before and after testing. The preference ratio (demon-strated food eaten × 100/total food eaten) was analyzed using Wilcoxonranked scores test.

Object recognition. Mice were habituated to test boxes (30 × 20 × 10 cm)for 20 min per day for 3–4 days. Two objects were placed in the box infixed positions, and mice were allowed two 10-min exploration sessions,with a 10-min interval in home cages. We quantified the time mice spentwith both forelimbs orientated toward and within 4 cm of either object.

articles©

2001

Nat

ure

Pu

blis

hin

g G

rou

p

htt

p:/

/neu

rosc

i.nat

ure

.co

m© 2001 Nature Publishing Group http://neurosci.nature.com

nature neuroscience • volume 4 no 3 • march 2001 295

During the test phase (10 min or 24 h later), one object was replaced bya novel object, and time spent exploring the objects was recorded. Explo-ration indices (time exploring object × 100/total exploration time) werecompared using Student’s paired t-test.

Surgery and electrophysiology. Acute experiments were done under ure-thane anesthesia. A concentric bipolar stimulating electrode was positionedin the perforant path, and a glass micropipette recording electrode in thehilus of the ipsilateral dentate gyrus. Pairs of pulses (inter-pulse intervals,10–100 ms) were used to study paired-pulse facilitation at two intensities,one to evoke a pure EPSP and the other a population spike of approximately1 mV. Three responses were collected at each and averaged. Test responseswere evoked by monophasic stimuli (1/30 s; 100–300 µA, 50–60 µs). Pulsewidth was doubled during tetani, which consisted of 6 series of 6 trains of6 stimuli at 400 Hz, 100 ms between trains, 20 s between series.

Long-term plasticity was studied in the dentate gyrus of freely mov-ing mice as described47. Following surgery, animals were allowed to recov-er (at least seven days) before recording, and were habituated to therecording chamber. Test responses elicited by monophasic pulses (80 µs,26–250 µA, 1/30 s) were recorded for 20-min periods on consecutivedays at an intensity that evoked a 1–3 mV population spike. Followingtwo days of stable baseline, a tetanus was delivered to the perforant path(same parameters as above, except pulse duration 100 µs). Responseswere measured for 60 min after tetanus and again, for 20 min at 24 h and48 h after tetanus. Data are expressed as change in amplitude (mV) rel-ative to the baseline measured over the 10 min before tetanic stimula-tion. Student’s t-test was used to compare mean levels of potentiation.

In situ hybridization. One hour after tetanic stimulation, brains wereremoved, frozen on dry ice and stored at –70°C. Sections (14 µm) weremounted on poly-L-lysine coated glass slides, and in situ hybridization wasdone as described17. After fixation, sections were hybridized overnight at42°C in 100 ml buffer containing 50% formamide, 4× SSC (150 mM sodi-um chloride/50 mM sodium citrate), 10% dextran sulphate, 5× Denhardt’s,200 mg/ml acid-alkali cleaved salmon testis DNA, 100 mg/ml long-chainpolyadenylic acid, 25 mM sodium phosphate (pH 7.0), 1 mM sodiumpyrophosphate and 100,000 c.p.m. radiolabeled probe (∼ 1 ng/ml) underparafilm coverslips. Sections were washed in 1× SSC at 55°C (30 min), 0.1× SSC at room temperature (5 min), dehydrated in 70% and 95% ethanol,and exposed to autoradiographic film. [35S]ATP end-labeled probes (NEN)were generated using terminal deoxynucleotidyl transferase (Promega,France) according to manufacturer’s instructions. A 50-fold excess of unla-beled oligonucleotide was used as negative control. Oligonucleotides had aunique sequence (Oswel, UK). Probe sequences were as follows: Zif268,CCGTGGCTCAGCAGCATCATCTCCTCCAGTTTGGGGTAGTTGTCC,complementary to nucleotides spanning amino acids 2–16 (ref. 13). LacZ,TTGGTGTAGATGGGCGCATCGTAACCGTGCATCTGCCAGTTTGAG,complementary to nucleotides 261–305. The probe against the 5´ UTR ofZif268, common to both wild-type and mutant mice, was GGGTTACAT-GCGGGGTGCAGGGGCACACTGCGGGGAGT, complementary tonucleotides 90–128 upstream of the AUG start codon.

Histology. Mice were perfused with 4% paraformaldehyde in PBS andbrains post-fixed overnight. Free-floating sections (40 µm) were washed4 times in PBS (0.1 M, 10 min per wash). Cresyl violet was used as a gen-eral cell stain, anti-NeuN (1/6000, Chemicon) to mark viable cells, anti-synaptophysin (1/100, Boehringer Mannheim, France) to mark presynapticboutons, and anti-parvalbumin (1/1000, Sigma). Non-specific epitopeswere blocked by incubation in 10% normal goat serum and 0.1% TritonX-100 in PBS for 1 h. Sections were incubated with the primary antibod-ies for 48 h at room temperature and then washed three times in PBS. Sec-ondary antibody was applied for 1 h at room temperature, and the sectionswashed in PBS (×3). Immunostaining was visualized using an ABC elitesystem (Vector Labs, France) and a VIP revelation kit (Biosystems, France).

ACKNOWLEDGEMENTSThis work was supported in part by CNRS PICS programme (N°756). We thank

P. Veyrac and M. Guegan for doing the immunohistochemistry, and S. Hiard for

rearing and genotyping the mice.

RECEIVED 23 OCTOBER 2000; ACCEPTED 16 JANUARY 2001

1. Lynch, G., Larson, J., Kelso, S., Barrionuevo, G. & Schottler, F. Intracellularinjections of EGTA block induction of hippocampal long-termpotentiation. Nature 305, 719–721 (1983).

2. Soderling, T. R. & Derkach, V. A. Postsynaptic protein phosphorylationand LTP. Trends Neurosci. 23, 75–80 (2000).

3. Nguyen, P. V., Abel, T. & Kandel, E. R. Requirement of a critical period oftranscription for induction of late-phase LTP. Science 256, 1104–1107(1994).

4. Frey, U., Frey, S., Schollmeier, F. & Krug, M. Influence of actinomycin D,an RNA synthesis inhibitor, on rat hippocampal neurons in vivo and invitro. J. Physiol. (Lond.) 490, 703–711 (1996).

5. Bliss, T. V. P. & Collingridge, G. L. A synaptic model of memory: long-termpotentiation in the hippocampus. Nature 361, 31–39 (1993).

6. Frey, U., Krug, M., Reymann, K. G. & Matthies, H. Anisomycin, aninhibitor of protein-synthesis, blocks late phases of LTP phenomena in thehippocampal CA1 region in vitro. Brain Res. 452, 57–65 (1988).

7. Otani, S. & Abraham, W. C. Inhibition of protein synthesis in the dentategyrus, but not the entorhinal cortex, blocks the maintenance of long-termpotentiation in rats. Neurosci. Lett. 106, 175–180 (1989).

8. Meiri, N. & Rosenblum, K. Lateral ventricle injection of the proteinsynthesis inhibitor anisomycin impairs long-term memory in a spatialmemory task. Brain Res. 789, 48–55 (1998).

9. Squire, L. R. & Barondes, S. H. Memory impairment during prolongedtraining in mice given inhibitors of cerebral protein synthesis. Brain Res.56, 215–225 (1973).

10. Rosenblum, K., Meiri, N. & Dudai, Y. Taste memory: the role of proteinsynthesis in gustatory cortex. Behav. Neural Biol. 59, 49–56 (1993).

11. Fazeli, M. S., Cobet, J., Dunn, M. J., Dolphin, A. C. & Bliss, T. V. P. Changesin protein synthesis accompanying long-term potentiation in the dentategyrus in vivo. J. Neurosci. 13, 1346–1353 (1993).

12. Davis, S. & Laroche, S. A molecular biological approach to synapticplasticity and learning. C. R. Acad. Sci. III 321, 97–107 (1998).

13. Milbrandt, J. A nerve growth factor-induced gene encodes a possibletranscriptional regulatory factor. Science 85, 7857–7861 (1987).

14. Lemaire, P., Revelant, O., Bravo, R. & Charnay, P. Two genes encodingpotential transcription factors with identical DNA-binding domains areactivated by growth factors in cultured cells. Proc. Natl. Acad. Sci. USA 85,4691–4695 (1988).

15. O’Donovan, K. J., Tourtellotte, W. G., Milbrandt, J. & Baraban, J. M. TheEGR family of transcription-regulatory factors: progress at the interface ofmolecular and systems neuroscience. Trends Neurosci. 22, 167–173 (1999).

16. Cole, A. J., Saffen, D. W., Baraban, J. M. & Worley, P. F. Rapid increase ofan immediate early gene messenger RNA in hippocampal neurons bysynaptic NMDA receptor activation. Nature 340, 474–476 (1989).

17. Wisden, W. et al. Differential expression of immediate early genes in thehippocampus and spinal cord. Neuron 4, 603–604 (1990).

18. Richardson, C. L. et al. Correlation between the induction of an immediateearly gene, zif268, and long-term potentiation in the dentate gyrus. BrainRes. 580, 147–154 (1992).

19. Abraham, W. C., Dragunow, M. & Tate, W. P. The role of immediate earlygenes in the stabilization of long-term potentiation. Mol. Neurobiol. 5,297–314 (1991).

20. Jones, M. W., French, P., Bliss, T. V. P. & Rosenblum, K. Molecularmechanisms of long-term potentiation in the insular cortex in vivo. J. Neurosci. 19, RC36, 1–8 (1999).

21. Hall, J., Thomas, K. L. & Everitt, B. J. Rapid and selective induction ofBDNF expression in the hippocampus during contextual learning. Nat.Neurosci. 3, 533–535 (2000).

22. Miyashita, Y., Kameyama, M., Hasegawa, I. & Fukushima, T.Consolidation of visual associative long-term memory in the temporalcortex of primates. Neurobiol. Learn. Mem. 70, 197–211 (1998).

23. Topilko, P. et al. Multiple pituitary and ovarian defects in Krox-24(NGFI-A, Egr-1)-targeted mic. Mol. Endocrinol. 12, 107–122 (1998).

24. Wolf, H. K. et al. NeuN: a useful neuronal marker for diagnostichistopathology. J. Histochem. Cytochem. 44, 1167–1171 (1996).

25. Kosaka, T., Katsumaru, H., Hama, K., Wu, J. Y. & Heizmann, C. W.GABAergic neurons containing the Ca2+-binding protein parvalbumin inthe rat hippocampus and dentate gyrus. Brain Res. 419, 119–130 (1987).

26. Katz, B. & Miledi, R. The role of calcium in neuromuscular facilitation. J. Physiol. (Lond.) 195, 481–492 (1968).

27. Gerlai, R. A new continuous alternation task in T-maze detectshippocampal dysfunction in mice: a strain comparison and lesion study.Behav. Brain Res. 95, 91–101 (1998).

28. Rasmussen, M., Barnes, C. A. & McNaughton, B. L. A systematic test ofcognitive mapping, working memory and temporal discontiguity theoriesof hippocampal function. Psychobiology 17, 335–348 (1989).

29. Garcia, J., Kimmeldorf, D. J. & Koelling, R. A. Conditioned aversion tosaccharin resulting from exposure to gamma radiation. Science 122,157–158 (1955).

30. Strupp, B. J. & Levitsky, D. A. Social transmission of food preference inadult hooded rats (Rattus norvegicus). J. Comp. Physiol. 98, 257–266

articles©

2001

Nat

ure

Pu

blis

hin

g G

rou

p

htt

p:/

/neu

rosc

i.nat

ure

.co

m© 2001 Nature Publishing Group http://neurosci.nature.com

296 nature neuroscience • volume 4 no 3 • march 2001

(1984).31. Rampon, C. et al. Enrichment induces structural changes and recovery

from nonspatial memory deficits in CA1 NMDAR1-knockout mice. Nat.Neurosci. 3, 238–244 (2000).

32. Brakeman, P. R. et al. Homer: A protein that selectively binds metabotropicglutamate receptors. Nature 386, 284–288 (1997).

33. Kato, A. et al. Novel members of the Vesl/Homer family of PDZ proteinsthat bind metabotropic glutamate receptors. J. Biol. Chem. 273,23969–23975 (1998).

34. Link, W. et al. Somatodendritic expression of an immediate early gene isregulated by synaptic activity. Proc. Natl. Acad. Sci. USA 92, 5734–5738(1995).

35. Lyford, G. L. et al. Arc, a growth factor and activity-regulated gene,encodes a novel cytoskeleton-associated protein that is enriched inneuronal dendrites. Neuron 14, 433–445 (1995).

36. Guzowski, J. F. et al. Inhibition of activity-dependent Arc proteinexpression in the rat hippocampus impairs the maintenance of long-termpotentiation and the consolidation of long-term memory. J. Neurosci. 20,3993–4001 (2000).

37. Walker, D. L. & Gold, P. E. Intrahippocampal administration of both the D-and L- isomers of AP5 disrupts spontaneous alternation behavior andevoked potentials. Behav. Neural Biol. 62, 151–162 (1994).

38. Kogan, J. H. et al. Spaced training induces normal long-term memory inCREB mutant mice. Curr. Biol. 7, 1–11 (1996).

39. Yin, J., Del Vecchio, M., Zhou, H. & Tully, T. CREB as a memorymodulator: induced expression of a dCREB2 activator isoform enhances

long-term memory in Drosophila. Cell 81, 107–115 (1995).40. Changelian, P. S., Feng, P., King, T. C. & Milbrandt, J. Structure of the

NGFI-A gene and detection of upstream sequences responsible for itstranscriptional induction by nerve growth factor. Proc. Natl. Acad. Sci.USA 86, 377–381 (1989).

41. Sakamoto, K. M. et al. 5´ upstream sequence and genomic structure of thehuman primary response gene, EGR-1/TIS8. Oncogene 6, 867–871 (1991).

42. Davis, S., Vanhoutte, P., Pagès, C., Caboche, J. & Laroche, S. TheMAPK/ERK cascade targets both Elk-1 and cAMP response element-binding protein to control long-term potentiation-dependent geneexpression in the dentate gyrus in vivo. J. Neurosci. 20, 4563–4572 (2000).

43. Impey, S., Obrietan, K. & Storm, D. R. Making new connections: role ofERK/MAP kinase signaling in neuronal plasticity. Neuron 23, 11–14(1999).

44. Bailey, C. H., Bartsch, D. & Kandel, E. R. Toward a molecular definition oflong-term memory storage. Proc. Natl. Acad. Sci. USA 93, 13445–13452(1996).

45. Silva, A. J., Kogan, J. H., Frankland, P. W. & Kida, S. CREB and memory.Annu. Rev. Neurosci. 21, 127–148 (1998).

46. Blum, S., Moore, A. N., Adams, F. & Dash, P. K. A mitogen-activatedprotein kinase cascade in the CA1/CA2 subfield of the dorsal hippocampusis essential for long-term spatial memory. J. Neurosci. 19, 3535–3544(1999).

47. Davis, S., Bliss, T. V. P., Dutrieux, G., Laroche, S. & Errington, M. L.Induction and duration of long-term potentiation in the hippocampus ofthe freely moving mouse. J. Neurosci. Methods 75, 75–80 (1997).

articles©

2001

Nat

ure

Pu

blis

hin

g G

rou

p

htt

p:/

/neu

rosc

i.nat

ure

.co

m© 2001 Nature Publishing Group http://neurosci.nature.com