Memory is a dynamic process that allows for adaptation to the demands of a continuously changing environment. The ability to update old memories in accordance with new experiences is crucial for maintaining their relevance over time. Particularly, it has been shown that after retrieval (or reactivation), memories may undergo a cycle of labilization and restabilization commonly known as reconsolidation (Nader et al., 2000). The labile state induced by this process can thus allow changes in memory strength and/or content (De Oliveira Alvares et al., 2013). This has been most extensively studied in aversive conditioning paradigms in rodents and humans and is of potential relevance to the management of psychiatric disorders involving dysfunctional memories (Monfils and Holmes, 2018).

Repeated exposure to a conditioned stimulus (CS) in the absence of an aversive unconditioned stimulus (US) also leads to a progressive reduction in fear responses, commonly known as extinction. However, extinction is thought not to erase the original memory; instead, it induces new learning that transiently inhibits fear expression (Bouton, 2002). Therefore, the fear memory typically reemerges with the passage of time (spontaneous recovery), exposure to the US (reinstatement), or when the CS is presented independently of the extinction context (renewal) (Rescorla and Heth, 1975; Archbold et al., 2010; Bouton et al., 2012). Thus, behavioral strategies that can weaken traumatic memories and reduce memory recovery can be relevant for improving the effectiveness of extinction.

Reconsolidation has been described in several experimental paradigms and species, from invertebrates to humans, suggesting that it might be a fundamental property of memories (Nader and Einarsson, 2010). Beyond its biological role in memory updating, it opens a window of opportunity as a potential therapeutic strategy to modify pathological memories. Several studies in the last decades have attempted to pharmacologically or behaviorally disrupt the reconsolidation of traumatic memories (Beckers and Kindt, 2017), as at least in some situations, this strategy can be less susceptible to spontaneous or induced recovery than traditional extinction (Duvarci and Nader, 2004; Monfils et al., 2009). However, most treatments that interfere with memory reconsolidation are toxic and cannot be readily administered to humans. Thus, in spite of almost two decades of research on memory reconsolidation, the evidence for the efficacy of reconsolidation-blocking treatments in clinical settings is still limited (Monfils and Holmes, 2018).

Research on reconsolidation-extinction boundaries suggests that the transition from one process to the other depends on the degree of mismatch between the original memory and the reactivation experience. Many studies have suggested that some degree of mismatch or prediction error is necessary for reconsolidation to occur (Lee, 2009; Fernández et al., 2016); however, if prediction error is too high, extinction may occur instead (Suzuki et al., 2004; Sevenster et al., 2014). Computational models suggest that, in models in which prediction error is low, memory updating/reconsolidation mechanisms are preferentially engaged, as the new experience is recognized as a new instance of the former one; however, as mismatch increases, the chances of it being attributed to a new latent cause (Gershman et al., 2017) or forming a new attractor in a neural network (Osan et al., 2011) increases.

This framework suggests that lowering the degree of mismatch between learning and reexposure might conceivably potentiate memory updating during extinction. This has been the rationale behind so-called retrieval-extinction procedures (Monfils et al., 2009; Kredlow et al., 2016) and has also been explored in short-term extinction protocols (Gershman et al., 2013). In this work, we propose a novel approach for long-term attenuation of traumatic memories that we term ‘deconditioning-update’. This strategy consists in weakening fear memories by updating the aversive information, substituting the original US by a mildly aversive stimulus during the plastic state induced by reactivation.

In order to approach our hypothesis, male Wistar rats were trained in auditory fear conditioning, where they received five conditioning trial tones (CS) that co-terminated with a 0.5-mA, 1-s footshock (US). On days 3 to 6 (reactivation), animals received 3 CSs during a 400-s daily session in a different context. In the no-footshock group, CSs were presented in the absence of shock, while in the deconditioning-update group, each tone co-terminated with a 0.1-mA footshock. A third group remained in the homecage (control group). On day 7, both groups were tested in the reactivation context with 3 CSs. On day 8, animals were tested in the training context for renewal, and, on day 28, they were retested for spontaneous recovery (Figure 1A).

Figure 1 with 4 supplements see all

Download asset Open asset

Figure 1—source data 1

Raw data of Figure 1.

https://cdn.elifesciences.org/articles/51207/elife-51207-fig1-data1-v2.xlsx

Download elife-51207-fig1-data1-v2.xlsx

Over the course of reactivation sessions, both groups presented a decrease in freezing, but this was more marked in the deconditioning-update group (Figure 1B), which presented a decrease in freezing of around 70% in comparison to the no-footshock group and 80% in comparison to the homecage control group in the test session (Figure 1C). Animals in the deconditioning-update group also had lower freezing responses in the renewal (Figure 1D) and spontaneous recovery (Figure 1E) tests, although it should be noted that memory recovery was not observed in the no-footshock group either, perhaps due to a ceiling effect caused by insufficient extinction (for complete statistics, see Supplementary file 1). Pre-CS freezing varied between 37% and 69% in both groups throughout the extinction sessions (Supplementary file 12), suggesting some degree of generalization between both contexts.

When performing the same experiment, but with each tone co-terminating with a 0.3-mA footshock in the reactivations, no fear reduction occurred (Figure 1—figure supplement 1); on the contrary, the footshock group presented higher levels of freezing than the no-footshock group throughout the reactivation sessions, as well as in the test. This result is in accordance with a recent study showing that fear memory may be strengthened by reactivation with a 0.3-mA footshock (Ferrara et al., 2019a). It also rules out that the suppression of freezing during reactivations by deconditioning-update may be due to inhibition of delay (i.e. animals learning to freeze only at the end of the CS with extended practice). When using a single reactivation session with 0.1mA, no difference was found in the test in comparison to the no-footshock group (Figure 1—figure supplement 2), suggesting that deconditioning-update requires several sessions to take place.

One could argue that the exposure to weak footshocks could simply lead to habituation and consequent devaluation of the US, without altering the conditioned association itself (Rescorla, 1973; Rescorla and Heth, 1975). In order to test this alternative interpretation, rats were conditioned as described above and the same 3 0.1-mA weak footshocks were given in another context without being paired with the tones (Figure 1—figure supplement 3A). In this case, no fear reduction was found in comparison to homecage controls. (Figure 1—figure supplement 3B). To further rule out the devaluation hypothesis, another set of animals was submitted to reinstatement after deconditioning, in order to test whether restoring the original footshock valence outside of the extinction context might lead to memory recovery. We found that the deconditioning group expressed a lower fear level when compared with the no-footshock group in the test even after reinstatement (Figure 1—figure supplement 3E), suggesting that the deconditioning-update effect is not due to US devaluation, but rather to updating of the CS-US association.

In an additional experiment, we investigated whether the stress induced by weak footshocks during reactivation could be enhancing extinction by itself. Animals underwent either the deconditioning-update protocol described in Figure 1A, or a similar protocol in which both tones and footshocks were applied during the reactivation sessions, but with no pairing between them. We found that this non-contingent tone-footshock protocol did not attenuate fear expression, leading to higher freezing than deconditioning-update during reactivations and in the test (Figure 1—figure supplement 4).

Next, we asked whether the deconditioning-update effect would also be efficient in reducing fear in female rats. As observed in males, females from the deconditioning-update group showed lower freezing levels throughout the reactivation sessions, as well as in the test session, when compared with the control group (Figure 1H). The same pattern was maintained in the renewal and spontaneous recovery tests (Figure 1I and J, respectively). As a further way to test for persistence of the original memory, we performed a retraining session in the original context with 3 0.5-mA CS-US pairings 24 hr after the spontaneous recovery test to assess savings. The deconditioning-update group had lower freezing compared with the other groups in a subsequent test session in the extinction context, suggesting that our protocol also lowers the rate of reacquisition of an aversive memory (Figure 1K).

Boundary conditions such as training intensity and memory age have been reported to prevent fear memories from being modified. Protocols with high training intensity make memory less prone to interference by pharmacological agents in the reconsolidation window (Frankland et al., 2006; Suzuki et al., 2004; Wang et al., 2009), while older memories are also less susceptible to modification than recent ones (Milekic and Alberini, 2002; Frankland et al., 2006; Suzuki et al., 2004; Bustos et al., 2009). Thus, our next experiments investigated whether deconditioning-update could attenuate fear expression in these cases. First, we trained the animals in the same protocol described above, but starting reactivations 40 days after conditioning (Figure 2A). Once more, the deconditioning-update group showed lower freezing expression throughout the reactivation sessions (Figure 2B). 24 hr after the last reactivation, both the footshock and no-footshock groups presented a comparable decrease in freezing behavior when compared to controls (Figure 2C). However, in the renewal and spontaneous recovery tests, fear expression reemerged in the no-footshock group, while the deconditioning-update group maintained its low freezing levels (Figure 2D and E, respectively; Supplementary file 2).

Figure 2 with 1 supplement see all

Download asset Open asset

Figure 2—source data 1

Raw data of Figure 2.

https://cdn.elifesciences.org/articles/51207/elife-51207-fig2-data1-v2.xlsx

Download elife-51207-fig2-data1-v2.xlsx

Next, we tested whether a stronger fear memory would constrain the deconditioning-update effect by training animals with 5 CS-US pairings using 1-mA shocks, while maintaining the rest of the protocol unchanged. In spite of the stronger shock intensity in the training session, the deconditioning-update group still presented reduced freezing levels in reactivation sessions when compared to the no-footshock group (Figure 2G). These lower freezing levels were maintained in the test, renewal and spontaneous recovery sessions, in which robust recovery was observed in no-footshock animals, but not in the deconditioning-update group (Figure 2H, I and J, respectively; Supplementary file 2). Similar results were observed in females in a slightly modified protocol with three reactivation sessions (Figure 2—figure supplement 1). These experiments suggest that the deconditioning-update induces a plastic state, allowing the fear memory trace to be altered even in boundary conditions that usually constrain memory updating (Pedraza et al., 2018).

In order to investigate whether the deconditioning-update approach is effective in attenuating other types of aversive memories, we trained animals in different fear-related tasks. First, we used a contextual fear conditioning paradigm, which is known to involve a set of brain regions that is partially distinct from auditory conditioning and includes the prefrontal cortex and hippocampus (Maren et al., 2013). Animals were placed in the conditioning chamber for 3 min before receiving two 0.5-mA, 2 s footshocks separated by a 30 s interval. On days 3 to 6, rats were reexposed to the same context for 4 min, with those in the deconditioning-update group receiving weak footshocks (0.1mA, 2 s) 3 min after being placed in the chamber (Figure 3A). The deconditioning-update group had lower fear expression during reactivations (Figure 3B) and maintained these lower freezing levels compared with the other groups in the test (Figure 3C). The same pattern was observed in the spontaneous recovery test, performed 20 days later, suggesting that the decrease in freezing caused by deconditioning-update is long-lasting (Figure 3D; Supplementary file 3).

Figure 3

Download asset Open asset

Figure 3—source data 1

Raw data of Figure 3.

https://cdn.elifesciences.org/articles/51207/elife-51207-fig3-data1-v2.xlsx

Download elife-51207-fig3-data1-v2.xlsx

Another set of animals underwent the step-through inhibitory avoidance task, in which training consists of applying 4 0.5-mA, 1-s footshocks with 5-s intervals when the animal enters the dark compartment of a shuttle box. During reactivations, animals were placed in the dark compartment for 30 s, with those in the deconditioning-update group receiving 2 0.1-mA shocks. In the test session, the animals were put in the light compartment, and the latency to enter the dark compartment was measured (Figure 3E). The deconditioning-update group showed a much shorter latency to reach the dark chamber (Figure 3F) and spent more time in the dark compartment over the 10-min session compared to the other groups (Figure 3G; Supplementary file 3), suggesting that memory was more robustly updated in this group during reactivation. Taken together, these results suggest that the deconditioning-update strategy is effective in weakening distinct types of fear-related memories.

To address whether the deconditioning-update effect would also be observed within a single, long-lasting extinction session, we trained animals and exposed them 48 hr later to a session containing 12 CSs, with the deconditioning-update group receiving a 0.1-mA footshock at the end of each tone. Fear reduction was limited and largely similar across groups in the extinction session and in the subsequent tests (Figure 4—figure supplement 1), albeit with slightly lower renewal in the deconditioning-update group, suggesting that the pairing of the CS with a weak footshock is not as effective in accelerating single-session extinction. In order to further explore this possibility, we subjected animals to a 24-tone single-session extinction protocol (Figure 4A). This led to robust fear reduction both within the extinction session (Figure 4B) and in a subsequent test (Figure 4C) in the no-footshock group, while freezing remained largely unchanged in the footshock group. However, fear memory reemerged in the renewal and spontaneous recovery test among no-footshock animals (Figure 4D and E; Supplementary file 4), as typically occurs with extinction learning. These results show that the presence of a weak shock at the end of every CS not only does not enhance extinction occurring over a single behavioral session, but can actually impair it in the short-term.

Figure 4 with 2 supplements see all

Download asset Open asset

Figure 4—source data 1

Raw data of Figure 4.

https://cdn.elifesciences.org/articles/51207/elife-51207-fig4-data1-v2.xlsx

Download elife-51207-fig4-data1-v2.xlsx

Other studies have shown that activation of L-type voltage-gated Ca⁺⁺ channels (LVGCC) is necessary both for destabilizing a reactivated memory during reconsolidation (Suzuki et al., 2008; Lee and Flavell, 2014; Crestani et al., 2015; Haubrich et al., 2015) and for some forms of extinction (Cain et al., 2002). Thus, we used the LVGCC antagonist nimodipine as a pharmacological tool to investigate whether deconditioning-update involved memory destabilization mechanisms (Flavell et al., 2011; Sierra et al., 2013; Crestani et al., 2015; Haubrich et al., 2015). Animals were trained and divided into four groups: no-footshock + vehicle, no-footshock + nimodipine, deconditioning-update + vehicle and deconditioning-update + nimodipine, with nimodipine or vehicle administered systemically 30 min before each reactivation (Figure 4F). Nimodipine attenuated freezing decrease in both behavioral protocols, suggesting that the drug impaired both deconditioning-update and regular extinction (Figure 4G and H). In the renewal and spontaneous recovery sessions, high freezing levels reemerged both in the no-footshock vehicle group and in nimodipine-treated animals, while the deconditioning-update vehicle group maintained a lower freezing level (Figure 4I and J; Supplementary file 4). This suggests that deconditioning-update is mediated by memory destabilization processes requiring Ca⁺⁺ influx through LVGCCs.

In order to address whether the effect of nimodipine treatment before reactivation sessions might be explained by state-dependency of the extinction memory in the test session, we trained a new set of animals in the same condition, except that two tests were conducted on separate days in each animal, either in the presence or absence of nimodipine treatment 30 min before. We found that animals treated with nimodipine before the test kept expressing high freezing levels, suggesting that nimodipine injection before reactivations prevented memory destabilization instead of inducing state dependency (Figure 4—figure supplement 2).

One explanation for our findings is that pairing the CS with a weak footshock could lead to a smaller degree of prediction error during reexposure sessions, biasing them towards memory updating as opposed to new learning. Computational models using neural networks (Osan et al., 2011) or Bayesian approaches (Gershman et al., 2017) have explored how different degrees of mismatch between stored memories and new experiences can lead to these two outcomes, suggesting lower degrees of mismatch or prediction error could lead to greater destabilization of stored memories. With this in mind, we used an adaptation of one of these models (Osan et al., 2011) to explore if this framework could account for our main results – that is accelerated fear reduction over multiple sessions and lower memory recovery when mismatch is reduced during reexposure. This fully connected Hopfield-like network (Hopfield, 1984) of 100 neurons is capable of storing patterns using Hebbian learning rules and retrieving them according to the inputs presented, which in our simulations included neurons representing tone and context information, as well as shock/non-shock information (Figure 5A). Additionally, the network also updates synaptic weights according to mismatch between a cue input and the retrieved network pattern (Osan et al., 2011).

Figure 5 with 1 supplement see all

Download asset Open asset

As shown in Figure 5B, a lower degree of mismatch in the reexposure pattern, caused by weak activation of shock-related neurons (as would be expected from deconditioning-update), can indeed lead to a greater reduction in subsequent fear behavior than a ‘pure extinction’ pattern (sharing only the tone information with the original memory). This effect was sensitive to blockade of the mismatch-induced degradation term – the model equivalent of blocking memory destabilization mechanisms such as LVGCCs with a drug like nimodipine (Figure 5C). This occurs because the ‘minor footshock’ pattern leads to recovery of the original memory as the outcome of the initial sessions (Figure 5D), causing mismatch-induced updating of the weights supporting these memories and weakening of the mutual connections between shock and context neurons (Figure 5E). On the other hand, no-footshock extinction leads to immediate formation of a new attractor, leaving the connection weights related to the shock memory unaltered throughout the extinction process.

Because of this, deconditioning-update was sensitive to blockade of mismatch-induced degradation, while no-footshock extinction – which depends basically on new learning – was not (Figure 5C). Blocking the Hebbian plasticity term, on the other hand, led to blockade of no-footshock extinction, as well as reconsolidation blockade in low-mismatch conditions (Figure 5—figure supplement 1). Interestingly, blocking Hebbian plasticity during deconditioning-update led to greater freezing decrease after the first reactivation session, suggesting that the procedure could be modulated by pharmacological agents such as protein synthesis inhibition; on the long run, however, the same manipulation decreased extinction due to interference with new learning.

A caveat here is that, unlike in the model, nimodipine did affect regular extinction in the experimental results. This suggests that the mechanistic distinction between deconditioning-update and classic extinction is not so clear-cut, and that these two processes might share common mechanisms, as has been proposed for reconsolidation and extinction (Almeida-Corrêa and Amaral, 2014). Capturing these subtleties, as well as other features of the data such as distinctions between single- and multiple-session extinction, likely requires model implementations that are more complex than this simple adaptation of the classic Hopfield formulation.

Taken as a whole, our findings demonstrate that presenting a tone followed by a weak footshock (deconditioning-update) was more effective in reducing fear memory than presenting a tone in the absence of shock (extinction training). Attenuation of fear responses following deconditioning-update was robust, long-lasting, and less sensitive to renewal and spontaneous recovery. Remarkably, this strategy was also effective in reducing fear within boundary conditions in which memories have been described to be less sensitive to modification (e.g., very strong training protocols and remote memories). The same effect was found in other types of fear-related memory (contextual fear conditioning and inhibitory avoidance tasks), and was sensitive to pharmacological blockade of LVGCCs. We suggest that the CS–US association is weakened during the deconditioning-update approach, leading to lower fear expression. Considering that (i) fear reduction is long-lasting, without spontaneous recovery, renewal and savings, (ii) weak-footshock pairings do not have the same effect in single-session extinction protocols or when stronger footshocks are applied, and (iii) this effect is prevented by LVGCC antagonism, we suggest that deconditioning-update is mediated by the memory destabilization effects commonly associated with reconsolidation.

These results are in line with predictions from a neural network model previously built to account for transitions between reconsolidation and extinction with increasing reexposure time in contextual fear conditioning (Osan et al., 2011). In this model, modification of existing connection weights is triggered by a protein synthesis-independent set of processes when there is mismatch between the memory pattern retrieved by the network (based on previously learned experiences) and the currently experienced sensory state, as when the prediction of a strong footshock is offset by a mild one. However, if mismatch is too extensive, as during extinction in the absence of a footshock, a new attractor is formed in the network at the first reexposure session already, preventing mismatch-induced weakening of the original memory.

Although the model is quite reductionist in its implementation, both due to the absence of realistic topology and to the non-spiking, continuous nature of neuronal activity, it is nevertheless able to account for transitions between simple retrieval, reconsolidation and extinction with increasing reexposure time (Suzuki et al., 2004) or number of non-reinforced CSs (Lee et al., 2006; Sevenster et al., 2014), as assessed by the effects of pharmacological agents such as protein synthesis inhibitors, NMDA antagonists and β-blockers. The model also helps to explain why, unlike pharmacological reconsolidation blockade, deconditioning-update takes several sessions to occur, as the impact of memory labilization mechanisms on a stored memory is much more pronounced when Hebbian mechanisms are blocked concomitantly (Osan et al., 2011).

Although we have not tested this explicitly, our results also appear compatible with the Bayesian inference framework proposed by Gershman et al. (2017), in which the probability that an experience is attributed to a new latent cause increases in proportion to the degree of prediction error generated by previous experience. This can explain why a lower degree of mismatch, such as that caused by reactivation followed by a weak footshock, can lead to a greater probability of memory updating and decrease the recovery of fear. If this is the case, the rationale for the greater effectiveness of deconditioning-update in reducing fear might be similar to that observed in so-called retrieval extinction paradigms (Monfils et al., 2009; Kredlow et al., 2016). In this case, the first retrieval trial, usually consisting of a single CS, is thought to induce retrieval-induced destabilization of the original memory, as the prediction error generated by this reexposure is not sufficient to form an extinction memory. Nevertheless, following this initial retrieval session with an extinction procedure leads to lower fear recovery than when extinction is performed without it, although this effect has been observed rather inconsistently across studies (Auber et al., 2013; Kredlow et al., 2016).

Interestingly, deconditioning-update only occurred when reactivation sessions were spaced across multiple days. When a massed extinction procedure with 24 non-reinforced CSs was performed within a single extinction session, on the other hand, a mild footshock at the end of the tones actually prevented within-session extinction, and increased freezing in a test performed on the following day. This seems to reinforce the notion that computations linking the degree of prediction error with memory destabilization occur only at the end of reexposure (Perez-Cuesta et al., 2007; Osan et al., 2011). It is also in line with the idea that within- and between-session extinction are distinct processes, with different dynamics and molecular requirements (Plendl and Wotjak, 2010; Almeida-Corrêa et al., 2015).

An important limitation of our computational framework, however, is that time is not explicitly modeled – thus, one cannot distinguish between massed and spaced extinction protocols in order to investigate the different results found in the two settings. An interesting challenge for future theoretical models, thus, would be to study whether and how within-session extinction relates mechanistically to mismatch-induced updating and between-session extinction. In this line, it is interesting to note that the model by Gershman et al. (2017) postulates that the probability that an experience is attributed to a new latent cause increases with time between initial learning and reactivation. This could plausibly account for why fear reduction with deconditioning-update is only observed in spaced sessions, in which new structural learning is more likely to take place in the absence of shocks (and thus prevent the original memory from updating in standard extinction conditions).

Nevertheless, in contrast to our work, Gershman et al. (2013) did find an effect of reducing mismatch at the start of a single 24-CS extinction session by pairing some of the initial tones with full-strength footshocks. That said, their effect was only observed in spontaneous recovery sessions and post-reinstatement sessions, and not within the extinction session itself. Moreover, the approach to induce lower degrees of mismatch in their experiment – gradually reducing the frequency of regular footshocks – was different from ours, in which this was achieved by providing a low-intensity shock at the end of every tone. Studying whether both approaches could be combined – by gradually decreasing footshock intensity, for example – could be an interesting topic for further investigation of the deconditioning-update effect. Another topic for future studies is whether mismatch between the training and reactivation context in auditory conditioning protocols also contributes to the deconditioning-update effect, as recently reported for reconsolidation (Ferrara et al., 2019b). In our study, animals showed relatively high pre-CS freezing levels, suggesting some degree of generalization between contexts, although the presence of renewal suggests that animals were also capable of differentiating them. It is thus an open question whether generalization might be important for deconditioning-update to occur.

Psychiatric disorders associated with pathological memories are prevalent, cause important social and economic burden, and approaches to translate basic knowledge of fear conditioning for potentiation of exposure therapy have met limited success so far in clinical settings (Monfils and Holmes, 2018). We believe that exploring the principles of deconditioning-update – that is, the notion that there is an ideal window of prediction error to potentiate reexposure effects – is a promising therapeutic avenue that could be explored in more depth in the setting of trauma-focused psychotherapy. The high efficacy, long-lasting effects, and drug-free nature of this approach make it particularly appealing for translation to human memory-related disorders, such as trauma, phobias and drug abuse.

A total of 323 male and female Wistar rats (2–3 months old, weighing between 300 and 400 g) from CREAL at the Federal University of Rio Grande do Sul (UFRGS) were used for the experiments. Only one animal was excluded (in the experiment in Figure 1A) due to health conditions. They were housed in plexiglass boxes, with four animals per cage, with block randomization using the cage as subgroup to ensure that each cage contained at least one animal per experimental group. Sample sizes ranged from 6 to 10 per group across experiments, yielding statistical power estimates between 69% and 89% to detect an absolute difference of 30% in freezing time (in line with the differences in memory recovery observed in Monfils et al. (2009)) with a standard deviation of 20% (our average value for test sessions in Figure 1) at α = 0.05 in a 2-tailed t test.

Animals were kept on 12/12 hr light/dark cycle under controlled temperature (21°C ± 2), with regular chow and water available ad libitum and humidity of approximately 65%. All experiments were performed during the light cycle. All procedures followed the Brazilian ethical guidelines for animal research set by the National Council for the Control of Experimental Animal Research (CONCEA). Methods and results are reported according to the revised ARRIVE guidelines (Percie du Sert et al., 2019).

All data generated or analysed during this study are included in the manuscript and supporting files.

1. 1. Almeida-Corrêa S
   2. Moulin TC
   3. Carneiro CF
   4. Gonçalves MM
   5. Junqueira LS
   6. Amaral OB

   (2015) Calcineurin inhibition blocks within-, but not between-session fear extinction in mice

   Learning & Memory 22:159–169.

   https://doi.org/10.1101/lm.037770.114

   • PubMed
   • Google Scholar
2. 1. Almeida-Corrêa S
   2. Amaral OB

   (2014) Memory labilization in reconsolidation and extinction--evidence for a common plasticity system?

   Journal of Physiology-Paris 108:292–306.

   https://doi.org/10.1016/j.jphysparis.2014.08.006

   • PubMed
   • Google Scholar
3. 1. Archbold GE
   2. Bouton ME
   3. Nader K

   (2010) Evidence for the persistence of contextual fear memories following immediate extinction

   European Journal of Neuroscience 31:1303–1311.

   https://doi.org/10.1111/j.1460-9568.2010.07161.x

   • PubMed
   • Google Scholar
4. 1. Auber A
   2. Tedesco V
   3. Jones CE
   4. Monfils MH
   5. Chiamulera C

   (2013) Post-retrieval extinction as reconsolidation interference: methodological issues or boundary conditions?

   Psychopharmacology 226:631–647.

   https://doi.org/10.1007/s00213-013-3004-1

   • PubMed
   • Google Scholar
5. 1. Beckers T
   2. Kindt M

   (2017) Memory reconsolidation interference as an emerging treatment for emotional disorders: strengths, limitations, challenges, and opportunities

   Annual Review of Clinical Psychology 13:99–121.

   https://doi.org/10.1146/annurev-clinpsy-032816-045209

   • PubMed
   • Google Scholar
6. 1. Bouton ME

   (2002) Context, ambiguity, and unlearning: sources of relapse after behavioral extinction

   Biological Psychiatry 52:976–986.

   https://doi.org/10.1016/S0006-3223(02)01546-9

   • PubMed
   • Google Scholar
7. 1. Bouton ME
   2. Winterbauer NE
   3. Todd TP

   (2012) Relapse processes after the extinction of instrumental learning: renewal, resurgence, and reacquisition

   Behavioural Processes 90:130–141.

   https://doi.org/10.1016/j.beproc.2012.03.004

   • PubMed
   • Google Scholar
8. 1. Bustos SG
   2. Maldonado H
   3. Molina VA

   (2009) Disruptive effect of midazolam on fear memory reconsolidation: decisive influence of reactivation time span and memory age

   Neuropsychopharmacology 34:446–457.

   https://doi.org/10.1038/npp.2008.75

   • PubMed
   • Google Scholar
9. 1. Cain CK
   2. Blouin AM
   3. Barad M

   (2002) L-type voltage-gated calcium channels are required for extinction, but not for acquisition or expression, of conditional fear in mice

   The Journal of Neuroscience 22:9113–9121.

   https://doi.org/10.1523/JNEUROSCI.22-20-09113.2002

   • PubMed
   • Google Scholar
10. 1. Crestani AP
    2. Zacouteguy Boos F
    3. Haubrich J
    4. Ordoñez Sierra R
    5. Santana F
    6. Molina JM
    7. Cassini LF
    8. Alvares LO
    9. Quillfeldt JA

    (2015) Memory reconsolidation may be disrupted by a distractor stimulus presented during reactivation

    Scientific Reports 5:13633.

    https://doi.org/10.1038/srep13633

    • PubMed
    • Google Scholar
11. 1. De Oliveira Alvares L
    2. Crestani AP
    3. Cassini LF
    4. Haubrich J
    5. Santana F
    6. Quillfeldt JA

    (2013) Reactivation enables memory updating, precision-keeping and strengthening: exploring the possible biological roles of reconsolidation

    Neuroscience 244:42–48.

    https://doi.org/10.1016/j.neuroscience.2013.04.005

    • PubMed
    • Google Scholar
12. 1. Duvarci S
    2. Nader K

    (2004) Characterization of fear memory reconsolidation

    Journal of Neuroscience 24:9269–9275.

    https://doi.org/10.1523/JNEUROSCI.2971-04.2004

    • PubMed
    • Google Scholar
13. 1. Fernández RS
    2. Bavassi L
    3. Kaczer L
    4. Forcato C
    5. Pedreira ME

    (2016) Interference conditions of the reconsolidation process in humans: the role of Valence and different memory systems

    Frontiers in Human Neuroscience 10:641.

    https://doi.org/10.3389/fnhum.2016.00641

    • PubMed
    • Google Scholar
14. 1. Ferrara NC
    2. Jarome TJ
    3. Cullen PK
    4. Orsi SA
    5. Kwapis JL
    6. Trask S
    7. Pullins SE
    8. Helmstetter FJ

    (2019a) GluR2 endocytosis-dependent protein degradation in the amygdala mediates memory updating

    Scientific Reports 9:5180.

    https://doi.org/10.1038/s41598-019-41526-1

    • PubMed
    • Google Scholar
15. 1. Ferrara NC
    2. Trask S
    3. Pullins SE
    4. Helmstetter FJ

    (2019b) The dorsal Hippocampus mediates synaptic destabilization and memory lability in the amygdala in the absence of contextual novelty

    Neurobiology of Learning and Memory 166:107089.

    https://doi.org/10.1016/j.nlm.2019.107089

    • PubMed
    • Google Scholar
16. 1. Flavell CR
    2. Barber DJ
    3. Lee JL

    (2011) Behavioural memory reconsolidation of food and fear memories

    Nature Communications 2:504.

    https://doi.org/10.1038/ncomms1515

    • PubMed
    • Google Scholar
17. 1. Frankland PW
    2. Ding HK
    3. Takahashi E
    4. Suzuki A
    5. Kida S
    6. Silva AJ

    (2006) Stability of recent and remote contextual fear memory

    Learning & Memory 13:451–457.

    https://doi.org/10.1101/lm.183406

    • Google Scholar
18. 1. Gershman SJ
    2. Jones CE
    3. Norman KA
    4. Monfils MH
    5. Niv Y

    (2013) Gradual extinction prevents the return of fear: implications for the discovery of state

    Frontiers in Behavioral Neuroscience 7:164.

    https://doi.org/10.3389/fnbeh.2013.00164

    • PubMed
    • Google Scholar
19. 1. Gershman SJ
    2. Monfils MH
    3. Norman KA
    4. Niv Y

    (2017) The computational nature of memory modification

    eLife 6:e23763.

    https://doi.org/10.7554/eLife.23763

    • PubMed
    • Google Scholar
20. 1. Haubrich J
    2. Crestani AP
    3. Cassini LF
    4. Santana F
    5. Sierra RO
    6. Alvares LO
    7. Quillfeldt JA

    (2015) Reconsolidation allows fear memory to be updated to a less aversive level through the incorporation of appetitive information

    Neuropsychopharmacology 40:315–326.

    https://doi.org/10.1038/npp.2014.174

    • PubMed
    • Google Scholar
21. 1. Hopfield JJ

    (1984) Neurons with graded response have collective computational properties like those of two-state neurons

    PNAS 81:3088–3092.

    https://doi.org/10.1073/pnas.81.10.3088

    • PubMed
    • Google Scholar
22. 1. Kredlow MA
    2. Unger LD
    3. Otto MW

    (2016) Harnessing reconsolidation to weaken fear and appetitive memories: a meta-analysis of post-retrieval extinction effects

    Psychological Bulletin 142:314–336.

    https://doi.org/10.1037/bul0000034

    • PubMed
    • Google Scholar
23. 1. Lee JL
    2. Milton AL
    3. Everitt BJ

    (2006) Reconsolidation and extinction of conditioned fear: inhibition and potentiation

    The Journal of Neuroscience 26:10051–10056.

    https://doi.org/10.1523/JNEUROSCI.2466-06.2006

    • PubMed
    • Google Scholar
24. 1. Lee JL

    (2009) Reconsolidation: maintaining memory relevance

    Trends in Neurosciences 32:413–420.

    https://doi.org/10.1016/j.tins.2009.05.002

    • PubMed
    • Google Scholar
25. 1. Lee JL
    2. Flavell CR

    (2014) Inhibition and enhancement of contextual fear memory destabilization

    Frontiers in Behavioral Neuroscience 8:144.

    https://doi.org/10.3389/fnbeh.2014.00144

    • PubMed
    • Google Scholar
26. 1. Maren S
    2. Phan KL
    3. Liberzon I

    (2013) The contextual brain: implications for fear conditioning, extinction and psychopathology

    Nature Reviews Neuroscience 14:417–428.

    https://doi.org/10.1038/nrn3492

    • PubMed
    • Google Scholar
27. 1. Milekic MH
    2. Alberini CM

    (2002) Temporally graded requirement for protein synthesis following memory reactivation

    Neuron 36:521–525.

    https://doi.org/10.1016/S0896-6273(02)00976-5

    • PubMed
    • Google Scholar
28. 1. Monfils MH
    2. Cowansage KK
    3. Klann E
    4. LeDoux JE

    (2009) Extinction-reconsolidation boundaries: key to persistent attenuation of fear memories

    Science 324:951–955.

    https://doi.org/10.1126/science.1167975

    • PubMed
    • Google Scholar
29. 1. Monfils MH
    2. Holmes EA

    (2018) Memory boundaries: opening a window inspired by reconsolidation to treat anxiety, trauma-related, and addiction disorders

    The Lancet Psychiatry 5:1032–1042.

    https://doi.org/10.1016/S2215-0366(18)30270-0

    • PubMed
    • Google Scholar
30. 1. Nader K
    2. Schafe GE
    3. Le Doux JE

    (2000) Fear memories require protein synthesis in the amygdala for reconsolidation after retrieval

    Nature 406:722–726.

    https://doi.org/10.1038/35021052

    • PubMed
    • Google Scholar
31. 1. Nader K
    2. Einarsson EO

    (2010) Memory reconsolidation: an update

    Annals of the New York Academy of Sciences 1191:27–41.

    https://doi.org/10.1111/j.1749-6632.2010.05443.x

    • PubMed
    • Google Scholar
32. 1. Osan R
    2. Tort AB
    3. Amaral OB

    (2011) A mismatch-based model for memory reconsolidation and extinction in attractor networks

    PLOS ONE 6:e23113.

    https://doi.org/10.1371/journal.pone.0023113

    • PubMed
    • Google Scholar
33. 1. Pedraza LK
    2. Sierra RO
    3. Lotz FN
    4. de Oliveira Alvares L

    (2018) Periodical reactivation under the effect of caffeine attenuates fear memory expression in rats

    Scientific Reports 8:7260.

    https://doi.org/10.1038/s41598-018-25648-6

    • PubMed
    • Google Scholar
34. Preprint

    1. Percie du Sert N
    2. Hurst V
    3. Ahluwalia A

    (2019) The ARRIVE guidelines 2019: updated guidelines for reporting animal research

    bioRxiv.

    https://doi.org/10.1101/703181

    • Google Scholar
35. 1. Perez-Cuesta LM
    2. Hepp Y
    3. Pedreira ME
    4. Maldonado H

    (2007) Memory is not extinguished along with CS presentation but within a few seconds after CS-offset

    Learning & Memory 14:101–108.

    https://doi.org/10.1101/lm.413507

    • Google Scholar
36. 1. Plendl W
    2. Wotjak CT

    (2010) Dissociation of within- and between-session extinction of conditioned fear

    Journal of Neuroscience 30:4990–4998.

    https://doi.org/10.1523/JNEUROSCI.6038-09.2010

    • PubMed
    • Google Scholar
37. 1. Rescorla RA

    (1973) Effect of US habituation following conditioning

    Journal of Comparative and Physiological Psychology 82:137–143.

    https://doi.org/10.1037/h0033815

    • PubMed
    • Google Scholar
38. 1. Rescorla RA
    2. Heth CD

    (1975) Reinstatement of fear to an extinguished conditioned stimulus

    Journal of Experimental Psychology: Animal Behavior Processes 1:88–96.

    https://doi.org/10.1037/0097-7403.1.1.88

    • Google Scholar
39. 1. Sevenster D
    2. Beckers T
    3. Kindt M

    (2014) Prediction error demarcates the transition from retrieval, to reconsolidation, to new learning

    Learning & Memory 21:580–584.

    https://doi.org/10.1101/lm.035493.114

    • PubMed
    • Google Scholar
40. 1. Sierra RO
    2. Cassini LF
    3. Santana F
    4. Crestani AP
    5. Duran JM
    6. Haubrich J
    7. de Oliveira Alvares L
    8. Quillfeldt JA

    (2013) Reconsolidation may incorporate state-dependency into previously consolidated memories

    Learning & Memory 20:379–387.

    https://doi.org/10.1101/lm.030023.112

    • PubMed
    • Google Scholar
41. 1. Suzuki A
    2. Josselyn SA
    3. Frankland PW
    4. Masushige S
    5. Silva AJ
    6. Kida S

    (2004) Memory reconsolidation and extinction have distinct temporal and biochemical signatures

    Journal of Neuroscience 24:4787–4795.

    https://doi.org/10.1523/JNEUROSCI.5491-03.2004

    • PubMed
    • Google Scholar
42. 1. Suzuki A
    2. Mukawa T
    3. Tsukagoshi A
    4. Frankland PW
    5. Kida S

    (2008) Activation of LVGCCs and CB1 receptors required for destabilization of reactivated contextual fear memories

    Learning & Memory 15:426–433.

    https://doi.org/10.1101/lm.888808

    • PubMed
    • Google Scholar
43. 1. Wang SH
    2. de Oliveira Alvares L
    3. Nader K

    (2009) Cellular and systems mechanisms of memory strength as a constraint on auditory fear reconsolidation

    Nature Neuroscience 12:905–912.

    https://doi.org/10.1038/nn.2350

    • PubMed
    • Google Scholar

Author details

1. Bruno Popik

   Neurobiology of Memory Lab, Biophysics Department, Biosciences Institute, Federal University of Rio Grande do Sul, Porto Alegre, Brazil

   Contribution

   Methodology, Project administration

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-0385-2749

2. Felippe Espinelli Amorim

   Institute of Medical Biochemistry Leopoldo de Meis, Federal University of Rio de Janeiro, Rio de Janeiro, Brazil

   Contribution

   Investigation, Methodology

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-1922-8211

3. Olavo B Amaral

   Institute of Medical Biochemistry Leopoldo de Meis, Federal University of Rio de Janeiro, Rio de Janeiro, Brazil

   Contribution

   Formal analysis, Supervision, Writing - original draft

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-4299-8978

4. Lucas De Oliveira Alvares

   Neurobiology of Memory Lab, Biophysics Department, Biosciences Institute, Federal University of Rio Grande do Sul, Porto Alegre, Brazil

   Contribution

   Conceptualization, Supervision, Funding acquisition, Writing - original draft, Writing - review and editing

   For correspondence

   lucas.alvares@ufrgs.br

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-0472-903X

• 2,994

  views

• 291

  downloads

• 29

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.