Abstract
'New' memories are initially labile and sensitive to disruption before being consolidated into stable long-term memories1, 2, 3, 4, 5. Much evidence indicates that this consolidation involves the synthesis of new proteins in neurons6, 7, 8, 9. The lateral and basal nuclei of the amygdala (LBA) are believed to be a site of memory storage in fear learning10. Infusion of the protein synthesis inhibitor anisomycin into the LBA shortly after training prevents consolidation of fear memories11. Here we show that consolidated fear memories, when reactivated during retrieval, return to a labile state in which infusion of anisomycin shortly after memory reactivation produces amnesia on later tests, regardless of whether reactivation was performed 1 or 14 days after conditioning. The same treatment with anisomycin, in the absence of memory reactivation, left memory intact. Consistent with a time-limited role for protein synthesis production in consolidation, delay of the infusion until six hours after memory reactivation produced no amnesia. Our data show that consolidated fear memories, when reactivated, return to a labile state that requires de novo protein synthesis for reconsolidation. These findings are not predicted by traditional theories of memory consolidation.
The idea that new memories go through an initial labile period before being consolidated into stable long-term memories is an entrenched part of psychological and neurobiological models of memory12. For example, there is considerable evidence that the formation of a long-term memory can be disrupted by certain treatments, such as systemic drug injections or electroconvulsive shock, given shortly after training, but that the same treatments given several hours or days later have no effect. One of the most commonly used drug manipulations involves the administration of drugs that block the translation of RNA into protein. Studies of this type indicate that memory consolidation involves protein synthesis5, 6, 7, 8, 9.
It has also been reported that electroconvulsive shock or systemic drug administration given after memory reactivation (retrieval) can cause an amnesia for the original learning13, 14, 15, 16, which indicates that consolidated memories might become labile when retrieved, and might even require reconsolidation. Here we examine whether reconsolidation involving protein synthesis is required for retrieved memories to persist. We use a behavioural paradigm, auditory fear conditioning, for which the neural circuit underlying memory formation is well characterized17, 18, 19. This allows us to manipulate memory at its presumed locus of storage, in contrast to past studies in which drugs were administered systemically. Specifically, we target infusions of anisomycin, an inhibitor of protein synthesis, to the LBA, a region implicated in fear learning by lesion, pharmacological and physiological findings17, 18, 19. Previously, we showed that infusions of anisomycin given after training block long-term but not short-term memory of auditory fear conditioning11. Here we examine the effects of similar manipulations administered after retrieval.
Rats were given a single pairing of a tone (conditioned stimulus, CS) and foot-shock (unconditioned stimulus, US). On test days, immobility (freezing) was used as an index of fear learning20. Twenty-four hours later, the rats received a single CS presentation (test 1) immediately followed by bilateral infusions of anisomycin or vehicle (artificial cerebrospinal fluid; ACSF) into the LBA. Freezing in test 1 was specific to the CS and comparable across groups. An analysis of variance (ANOVA) that compared freezing during the pre-CS or CS periods across groups indicated that there was no interaction between these two variables (F(2,18) = 1.6), nor an effect of group (F(2,18) = 1.4). However, there was a significant effect of period (F(1,18) = 160, P < 0.01). Twenty-four hours after test 1, the rats were presented with three CSs (test 2). Anisomycin produced a dose-dependent decrease in freezing in response to the CS in test 2 (Figs 1, 2a–c). An ANOVA revealed a main effect of group ( F(2,18) = 12, P < 0.01). A Newman–Keuls post hoc test revealed that the low-dose anisomycin and ACSF groups were similar to each other (P > 0.05), but both were significantly different from the high-dose group (P values < 0.01). Extinction was observed over the three CS presentations (main effect of trial ( F(2,36) = 7, P < 0.01), but there was no interaction between trials and group (F < 1 ). This effect of anisomycin requires that the memory be actively retrieved, as omission of the CS before anisomycin infusion in test 1 led to normal conditioned fear responses in test 2 (Fig. 2d, e ; no main effect of group (F < 1). The latter finding rules out the possibility that the deficit we observed is due to a disruption of late waves of protein synthesis21 that may be necessary for the consolidation of the original learning, or to nonspecific effects such as damage to the amygdala by the drug. It is unlikely that the high dose of anisomycin damaged the amygdala after CS transmission, as there was no histological evidence of amygdala damage. Furthermore, rats relearn fear conditioning normally when the same dose of anisomycin is infused into the amygdala after initial learning. The fact that anisomycin infusions after reactivation of the fear memory produced amnesia for the original learning indicates that reactivation of a consolidation fear memory may place it in a labile state, one that has to be reconsolidated by protein synthesis to remain usable to the organism in future situations.
Figure 1: Schematic representation of the amygdala at four different rostral–caudal planes.
![Figure 1 : Schematic representation of the amygdala at four different rostral|[ndash]|caudal
planes. Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, or to obtain a text description, please contact npg@nature.com](/nature/journal/v406/n6797/images/406722aa.0.jpg)
The numbers represent the posterior coordinate from bregma. Injector placements in the LBA are represented by the filled symbols; black filled squares represent ASCF group placements, grey filled triangles represent the low-dose anisomycin, and black filled circles represent high-dose group. L, lateral nucleus; B, basal nucleus; C, central nucleus. The placements for subsequent experiments all demonstrate similar distributions as in this experiment and therefore are not shown.
High resolution image and legend (121K)Figure 2: A test of whether consolidated fear memories can become labile when reactivated.
a, The behavioural procedure used for experiment 1A. b, Freezing
to the CS on test 1 was comparable across groups and was specific to the CS.
c, Intra-LBA anisomycin infusions after reactivation of a consolidated
fear memory produce amnesia for the original learning, as measured on test
2. d, e, Rats demonstrated normal memory if the CS was omitted
before anisomycin. d, The behavioural procedure used for experiment
1B. Rats were placed in the test chamber and received infusions of anisomycin.
e, Percent freezing on test 2. Figure legend is applicable to both
c and e. Vertical open-headed arrows represent infusions. All data
points represent group means 
s.e.m.
Protein synthesis inhibitors typically impair the consolidation of new
memories when they are administered during a specific time window (which varies
from minutes to hours) after learning8. Administration of such
drugs after this time window does not affect memory. We next asked whether
a time window also exists for reconsolidation, by delaying anisomycin infusions
for 6 h after retrieval. Freezing in test 1 was specific to the CS
and comparable across groups. ANOVAs revealed no period 
group interaction
(F < 1) and no effect of group (F
= 1) but a significant effect of period (F(1,13) =
124, P < 0.01). In contrast to anisomycin infusion
immediately after retrieval, infusion 6 h after retrieval had no effect
(Fig. 3). ANOVAs revealed that in test 2 there was no
main effect of group (F(1,13) = 4), no trial
group interaction (F(2,26) = 2.7) and no main effect
of trial (F(2,26) = 3). Note that the nonsignificant
impairment produced by delaying anisomycin is much smaller than that seen
when anisomycin is given immediately after CS reactivation (compare with Fig. 2c). Thus, both consolidation and reconsolidation have
time windows within which protein synthesis is required if a memory is to
persist.
Figure 3: Intact memory if anisomycin infusions are delayed by 6 h.
![Figure 3 : Intact memory if anisomycin infusions are delayed by 6|[thinsp]|h. Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, or to obtain a text description, please contact npg@nature.com](/nature/journal/v406/n6797/images/406722ac.0.jpg)
a, The behavioural procedure used for experiment 2. Vertical open-headed
arrows represent infusions. b, Freezing on test 1 was specific
to the CS and comparable across groups. c, Percent freezing during
test 2. The groups are not significantly different. All data points represent
group means s.e.m.
In all of the previous experiments, the time between training and CS presentation
for test 1 was about 24 h. We next investigated whether reconsolidation
has a temporal gradient. For example, older memories may be more thoroughly
consolidated and thus may be resistant to becoming unstable when retrieved.
To test this, we waited for 14 days between conditioning and test 1 (Fig. 4). Postponing test 1 by 14 days produced an incubation
effect22 such that freezing in test 1 was higher than in the
previous experiment. Freezing during test 1 was specific to the CS and was
comparable across groups. There was no period group interaction (
F(1,9) = 1.1) and no effect of group (F =
1), but a significant effect of period (F(1,9) = 440
, P < 0.01). However, the performance of the
groups in test 2 diverged—the group receiving intra-LBA infusions of
anisomycin showed significantly less freezing than the controls. ANOVAs revealed
a main effect of group (F(1,9) = 20.6,
P < 0.01) and trials (F(2,18) = 7,
P < 0.01), but no interaction between these variables (
F < 1). Thus, blockade of protein synthesis in the LBA after
memory reactivation caused amnesia of the original learning, even though the
learning took place 14 days before the reactivation and drug treatment. Even
well consolidated memories are labile and subject to disruption when reactivated.
Figure 4: Fourteen days after training, anisomycin infusions after reactivation of the memory still produce amnesia.
a, The behavioural procedure used for experiment 3. Vertical open-headed
arrows represent infusions. b, Freezing during test 1 was specific
to the CS and was comparable across groups. c, Percent freezing on
test 2. All data points represent group means s.e.m.
Although anisomycin blocks reconsolidation, it is possible that the impairment is due to nonspecific effects that render the amygdala temporarily dysfunctional for reasons other than protein synthesis inhibition. To provide compelling evidence that any manipulation acts specifically on the molecular mechanisms mediating consolidation, as opposed to producing nonspecific effects, at a minimum it is necessary to show that memory is intact shortly after training but impaired later23, 24, 25, 26. For example, rats freeze normally 4 h but are impaired 24 h after receiving intra-LBA anisomycin infusions immediately after training11. Using the same logic and applying it to reconsolidation, if the effects of anisomycin infused into the LBA shortly after memory reactivation are specific to reconsolidation, then freezing should be normal at 4 h, but impaired at 24 h, after CS presentation. We refer to these two time points are post-reactivation short-term memory (PR-STM) and post-reactivation long-term memory (PR-LTM), respectively.
The two groups exhibited comparable freezing during test 1 (
Fig. 5). An ANOVA revealed a significant effect of period (
F(1,10) = 147, P < 0.01), no effect
of group (F < 1) and no interaction between these
two variables (F < 1). Rats treated with anisomycin
immediately after test 1 exhibited intact PR-STM but deficient PR-LTM. ANOVAs
on the scores of the PR-STM test showed that there was no main effect of group
(F < 1), no interaction between trial group
(F < 1) and no effect of trial (F
< 1). Conversely, a similar analysis on the PR-LTM scores did reveal
a significant effect of group (F(1,14) = 14,
P < 0.01). Furthermore, there was no interaction between group
and trial (F < 1) but there was a main effect of
trial (F(2,28) = 5, P < 0.05
). The fact that animals can accurately perceive, evaluate and respond
to the CS 4 h after anisomycin infusion shows that the amygdala is
functionally intact at the time of the PR-STM test, and thus that anisomycin
did not affect reconsolidation by producing nonspecific effects. This pattern
of findings, intact PR-STM and impaired PR-LTM, localizes the effects of anisomycin
on fear behaviour to the molecular processes mediating reconsolidation.
Figure 5: Amnesia following anisomycin is not due to nonspecific effects.
a, The behavioural procedure used for experiment 4. Vertical open-headed
arrows represent infusions. Anisomycin infusions impaired post-reactivation
long-term memory (PR-LTM), but had no effect on post-reactivation short-term
memory (PR-STM). b, All rats demonstrated comparable freezing scores
on test 1. c, Scores on the PR-STM test 4 h after reactivation
and anisomycin. d, Scores on the PR-LTM test 24 h after reactivation
and anisomycin. The key is applicable to both c and d. All data
points represent group means s.e.m.
The conventional view of memory consolidation predicts that blockade of protein synthesis should block new learning in test 1 of these experiments, which is an extinction test. The only new learning that occurs is about the failure of the CS to predict the US. Blockade of protein synthesis should therefore block extinction, and thus enhance memory, according to the conventional view. In contrast, however, blockade of protein synthesis had the opposite effect—it eliminated the memory rather than making it stronger.
These results provide evidence that fear memories, once retrieved, must undergo protein-synthesis-dependent reconsolidation in the LBA or nearby areas to remain accessible at later times. The full implications of this finding for fear and other memories are not understood at present. It is possible that not all memories require reconsolidation. There may be a range of parameters within which reactivation of a memory converts it into a labile state, possibly involving the extent of experience with the particular learning situation, the kind of learning system engaged and the motivational state of the subject at the time of learning and retrieval.
Reconsolidation may reflect the dynamic nature of the process by which new information is added to existing stores. It has long been believed that memory retrieval is an active or constructive process by which old information is integrated with the current knowledge base of the organism27. Reconsolidation may be part of the neural mechanism through which constructed memories are stored for later constructions.
Current models of learning propose that the production of new proteins is necessary for structural encoding of recent experiences in long-term memory6, 7, 8, 9. In addition, it now appears that new proteins are also required to maintain memories that have been reactivated. It seems unlikely that retrieval reverses the structural changes induced by original learning. Rather, some property of retrieval may destabilize the structural changes such that they now have to be reconsolidated with the aid of new proteins. The fact that animals demonstrate intact freezing 4 h after anisomycin indicates that the structural changes may remain functional for at least 4 h. Particularly important for future work will be the clarification of the physiological basis of memory lability during retrieval and the requirement for reconsolidation. It is possible that some modification of the synaptic tagging hypothesis28, 29, which proposes that active synapses are given molecular markers that help stabilize synapses by capturing proteins made in the cell nucleus, might account for lability and reconsolidation, although this remains to be seen.
We have shown that reactivation changes the status of a consolidated fear memory to a labile one that must be reconsolidated using de novo protein synthesis to persist. Like consolidation itself, reconsolidation has a temporal window during which blockade of protein synthesis is effective, and beyond which it is not. Furthermore the reconsolidation process, like the consolidation process, has a short-term phase that is not dependent on protein synthesis. A definition of consolidation based on 'new' memories is insufficiently broad to describe these data. We propose, in keeping with the original suggestion by Misanin et al.13, that as a first approximation 'active' rather than 'new' memories be viewed as labile, subject to disruption, and requiring protein-synthesis-dependent consolidation processes.
Methods
Subjects
Subjects were adult male Sprague–Dawley rats from Hilltop Labs. Rats were housed individually in plastic Nalgene cages and maintained on a 12/12 h light/dark cycle. Food and water were provided ad libitum.
Surgery and histology
Under Nembutal anaesthesia (45 mg kg
-1), rats were implanted bilaterally with 22-gauge stainless steel
cannulas into the lateral amygdata. Coordinates were 3.0 mm posterior
to bregma, 5.3 mm lateral to the midline and 8.0 mm ventral
to the skull surface. Rats were given at least 5 days to recover before experimental
procedures. All procedures were in accordance with the NIH Guide, and were
approved by the NYU Animal Care and Use Committee. At the end of the experiment,
using standard histological methods, animals were perfused and their brains
sectioned at 50 
m thickness. The sections were stained using
cresyl violet and examined by light microscopy for cannula penetration into
the LBA.
Intra-LBA infusions
Drugs were infused slowly using
an infusion pump into the LBA at 0.25 l min-1
. Following drug infusion, cannulas were left in place for an additional
minute to allow diffusion of the drug away from the cannula tip. Anisomycin
(Sigma) was dissolved in equimolar HCl, diluted with ACSF and adjusted to
pH 7.4 with NaOH. Although the lateral nucleus was the main target, the 0.5-l
infusions also probably affected the adjacent basal nucleus. We therefore
refer to the affected area as the lateral and basal amygdala (LBA).
Apparatus
Conditioning and tone testing were conducted in different chambers. For conditioning, rats were placed in a Plexiglas rodent conditioning chamber (chamber A) with a metal grid floor (Model E10-10, Coulbourn Instruments) that was enclosed within a sound attenuating chamber (Model E10-20). The chamber was dimly illuminated by a single house light. For tone testing, rats were placed in a different Plexiglas chamber (ENV-001, MedAssociates), which has been shown to minimize generalization from the conditioning environment11. The tone-testing chamber (chamber B) was brightly lit with three house lights and contained a flat black Formica floor that had been washed with peppermint soap. A micro-video camera was mounted at the top of the chamber so that rats could be videotaped during testing.
General behavioural procedures
Rats were placed in chamber A and after a 5-min acclimatizing period, given a single conditioning trial consisting of a 30-s presentation of a 5-kHz, 75-dB tone CS that ended at the same time as a 2.0-mA, 1-s food shock US. Rats were then returned to their home cages. The next day, 24 h later, rats were placed in chamber B and given a single 30-s CS presentation (test 1) to reactivate the memory. Twenty-four hours after test 1, rats were returned to chamber B and given three CS presentations (test 2).
Experiment 1A
Rats were infused with either 62.5 g
per 0.5 l per side (n = 8) or 6.2 g
per 0.5 l per side (n = 7) anisomycin
or ACSF (n = 6) immediately after CS termination during
test 1. The highest dose of anisomycin was chosen based on a previous study
showing >90% suppression of protein synthesis in cortex using this concentration30. Previous data have shown that post-training intra-LBA infusions
of the high but not low dose blocked consolidation of fear conditioning11.
Experiment 1B
During test 1 no CS was presented while the animals explored chamber B, but rats still received an infusion of vehicle (n = 6) or the high dose of anisomycin ( n = 7) at the end of the exposure to chamber B.
Experiment 2
High-dose anisomycin (n = 8) or vehicle (n = 7) infusions were performed 6 h after test 1. Animals were transported to the infusion room, received the infusion and were then returned to their home cage.
Experiment 3
Fourteen days were inserted between conditioning and test 1. After CS reactivation, rats received either high-dose anisomycin (n = 6) or vehicle (n = 5) infusion.
Experiment 4
After test 1 animals received either vehicle (n = 8) or high-dose anisomycin ( n = 8) infusions into the LBA. An extra test was inserted 4 h after test 1 (post-reactivation short-term memory, PR-STM) during which animals received three CS presentations. Test 2 was performed as described above, 24 h after reactivation (post-reactivation long-term memory, PR-LTM). These time points were chosen based on the findings that freezing is intact 4 h, but impaired 24 h, after conditioning11.
