Pharmacologically induced amnesia for learned fear is time and sleep dependent

The discovery in animal research that fear memories may change upon retrieval has sparked a wave of interest into whether this phenomenon of reconsolidation also occurs in humans. The critical conditions under which memory reconsolidation can be observed and targeted in humans, however, remain elusive. Here we report that blocking beta-adrenergic receptors in the brain, either before or after reactivation, effectively neutralizes the expression of fear memory. We show a specific time-window during which beta-adrenergic receptors are involved in the reconsolidation of fear memory. Finally, we observe intact fear memory expression 12 h after reactivation and amnesic drug intake when the retention test takes place during the same day as the intervention, but post-reactivation amnesia after a night of sleep (12 h or 24 h later). We conclude that memory reconsolidation is not simply time-dependent, but that sleep is a final and necessary link to fundamentally change the fear memory engram.

4) The authors do no report any exclusion of datasets in their analyses. Given that startle signals are not always perfect, I wonder if any participants were excluded or pre-selected?
5) The results reported for experiment 2 and 3 (employing two CSs paired with an US) are missing the comparison between the reactivated CS against the non-reactivated (e.g. CS1 vs CS2). Instead comparisons between the CS associated with the US and the control stimulus are employed. However, these comparisons do not reveal the effect of reactivation exerted on a CS that has been previously associated with the US. Hence, comparisons between these stimuli need to be included. 6) On page 10 line 201, the authors state that planed comparisons (CS1 vs CS3) revealed reduced memory retention in the 1hour pill group and intact differentiation in the 2hour pill group. Is the difference (or interaction) between groups significant, as well? Or is the interaction with the factor group only significant during extinction? The same is true for the planned comparisons (CS1 vs CS3) between the sleep and no sleep group on page 12 line 251. Here again, the crucial group comparison is not reported. 7) Propranolol is a classical example of a non-selective Antagonist at the beta Adrenoreceptor. However, Propranolol has antagonistic properties at 5-HT1 receptor (5-HT1B, 5-HT1A , 5-HT1c; Nishio H, Nagakura Y & Segawa T 1989; Arch Int Pharmacodyn 302:96-106), as well. Agren et al found an impact on 5HT on reconsolidation as well (Translational Psychiatry 2012). Hence, it might well be that 5HT contributes besides NA to reconsolidation in humans. In fact Nadolol has no affinity to 5HT receptors, which might have contributed to the current results.
Minor Comments: -The subjective evaluation of US intensities is around 3. Please provide the range of that scal.
-How many participants in each experiment were aware of the stimulus contingencies? -Please report the gender of the participants included in each sample and provide some evidence (e.g. testing in one of the experiments) that gender did not influence reconsolidation.
-Please provide information about the intake of food in each group, especially in the comparison between different time-points of drug administration. Food might have influenced the pharmacokinetic profiles.
-Please consider exchanging "biochemical signature" in the manuscript (e.g. page 8 line 173) with "pharmacokinetic profile/signature". While this might sound picky, it is just a meant to be a constructive help to use standardized terms.

Reviewer #2 (Remarks to the Author):
In 3 independent experiments the authors show in human participants that the expression of conditioned fear memories (measured via startle response potentiation) can be abolished if propranolol (a β-adrenergic antagonist) is administered during a specific time window following memory reactivation. Moreover, the reduction of fear responses on the long-term was found to be dependent on intervening sleep. This is a nicely designed and wellcontrolled study that tackles a timely and important issue. The methods are sound and the manuscript is well-written. However, I have a few concerns that should be addressed.
1. My main concern is with regard to the authors' conclusion of a specific time window for β-AR activity of 2-3 hours post-reactivation (Fig. 3). In my understanding of the data, it cannot be excluded that β-AR activity during the first 1-2 hours plays a role for memory reconsolidation as well. What the data clearly show is that after 3 hours β-AR activity is (probably) no longer required. (Although even this conclusion is shaky because later intervals were not tested, e.g. it is possible that there is a second time window, e.g. around 6 hours post-reactivation, with a β-AR dependency.) But more importantly, based on Fig. 3 it cannot be excluded that propranolol acted on reconsolidation processes during the first 2 hours after reactivation. The drug administered 1 h before and right after reactivation could also have exerted an effect sooner than the 2-3 h time window. In order to show that in this phase β-AR activity is not required for reconsolidation, earlier drug administration would have to be tested, e.g. 2 vs. 5 hours before reactivation. This issue should be appropriately discussed and the conclusions toned down. 2. The observed sleep effect is particularly novel and interesting and I wonder about the possible underlying mechanisms of this effect. The authors discuss synaptic downscaling as one possibility, which seems to be reasonable. An alternative mechanism that came to my mind is the formation and preferential consolidation of a new memory trace. During the reactivation session, participants may acquire a new safety memory trace, i.e. learning that the CS1 is no longer paired with the US. Sleep may then foster the preferential consolidation of this new safety memory trace that interferes with the fear memory trace. 3. The sleep and no sleep groups of Exp. 3 differ in several regards. First, considering that propranolol was administered after reactivation, the active phase of the drug fell mainly in the sleep state for the sleep group and in the wake state for the no sleep group. Do the authors know of any evidence showing that propranolol may act differently during sleep and wakefulness? A number of hormones, neurotransmitters and plasticity-related processes differ between sleep and wakefulness and, thus, the effects of propranolol may also differ. Second, when did acquisition/extinction take place for the sleep and no sleep groups? If it occurred at the same time of day for both groups, this would mean that there were differences in the retention interval between acquisition and reactivation as well as between test and extinction for the two groups. Alternatively, if acquisition/extinction took place also in the evening and morning, respectively, can the authors exclude any potential circadian effects? 4. A few methodological details should be clarified: o In Exp. 1, in the instruction for participants it says that "an electric shock would follow one of the slides in most cases". Was the CS1 actually followed by an electric shock in all cases? Why were participants instructed otherwise? And why was the instruction different in Exp. 2, where it says "in all cases"? o Line 376: what does randomized within blocks (i.e., CS1 -CS2 -NA) mean? Were stimuli of each category always presented consecutively in one block, i.e. all CS1 one after the other etc.? o Line 385/396: were CSs presented without US here? o It is said that the drug was applied in a single blind fashion? What does that mean? Who was blind, the participants or the experimenter? o Missing startle responses (0.5%) were excluded. How did the authors define missing startle responses?
Minor comments: 1. Parts of the introduction, especially with regard to the details of the molecular pathways, do not seem to be essential for the rationale of the experiments. Perhaps these parts could be moved to the discussion section and the introduction could be shortened a bit. 2. It would be helpful to spell out β-AR in the title for non-experts. Also, the abbreviation PE in extended Figures 1, 2 and 3 should be defined. 3. Line 136: The abbreviation CS1-R is not defined here. Could it simply be termed CS1 here? 4. Line 190: after reactivation instead of retrieval? 5. Typos: line 305 "This in", line 322 "took play", line 364 "a ground references", line 709 "extinction trial group"

Reviewer 1:
Kindt and Soeter present an interesting set of experiments revealing three prerequisites for pharmacological interference with reconsolidation of fear-conditioned memory. Each experiment employs a systemic administration of a noradrenergic antagonist and employs a peripheral readout measurement (startle responses). The results delineate three central boundary conditions of reconsolidation processes in humans to the current literature. The authors are complimented on their elegant design, which was inspired by theories and rodent work on synaptic processes during memory consolidation. However, the authors need to be careful to avoid overinterpretation of the results (see comment 4) in terms of reverse inference (e.g. synaptic plasticity involves noradrenaline, but noradrenaline does not only work on synaptic plasticity). I am aware that systemic administration of drugs is "the best shot" to infer central processes in humans, hence the authors might reconsider framing the experiments. Another major concern is the small sample size, which can be easily addressed by providing a power calculation.
We agree with the reviewer that we should avoid reverse inference (i.e., inferring biological processes from behavioural data). We have now changed our text accordingly throughout the manuscript (see response to comment 4). Another major concern raised by this reviewer is the small sample size, which could indeed be easily addressed by a power calculation. Hence, following the suggestion of the reviewer, we have now added the power analysis in a footnote at the bottom of page 9. This yielded a sample size of 20 participants for a large effect size of f = 0.35 with an alpha level of 0.05 and a power level of 0.95. Given the large effect size, we believe that the current sample sizes are sufficient.
1 -In case of non significant findings, the authors provide no p-values, but F-values. In the cases of low F-values this strategy is understandable and even more informative (e.g. F<1). But for higher F-values (e.g. F(1,128)< 3.17 on page 7 or F<(1,9)=4.36 on page 10), it is hard to infer p-values (p=0.077 and p=0.065, respectively) from these F-values. Since p-values are still the standard in reporting statistical results, I would advocate for reporting all p-values below 0.1 in addition to the F-values. Actually, p-values below 0.1 in such small sample sizes are sometimes considered as "statistical trends".
We have now reported all p-values below 0.1 in addition to the F-values and commented on the statistical trends as well -see lines 138-139 and line 213-214.
2 -While the conclusion is parsimonious that sleep might be causal for the reconsolidation of the memory trace, the authors need to be more careful with this interpretation. In particular, it might well be that differences in endogenous regulation of neurotransmitters like cortisol, melatonine, etc through the circadian rhythm could have interfered with the reconsolidation. Hence, since sleep is one of a bunch of differences between groups, the authors should be careful with pinpointing a monocausal influence. Related to this, I did not understand the conclusion of this study, that "post retrieval amnesia was independent of time" (p.13 line 277).
We agree with the reviewer that we should be careful with our conclusions on the role of sleep and that post-reactivation amnesia is independent of time. We have now deleted this sentence from the manuscript and toned down the discussion (see lines 304-306 and 314-315). We have elaborated more extensively on the role of sleep in memory (re)consolidation in the general discussion (pages 16-17, lines 370-387, lines 396-400). Although the current study did not test for the effect of circadian rhythm, we refer to literature suggesting that it is not very likely that the relation of sleep and memory is explained by circadian rhythm (lines 312-315). Another example can be found on page 11, l.225. The description of the third experiment starts with : " To establish whether the beta-AR blocker propranolol acts specifically on the molecular mechanism mediating reconsolidation…". This has not been tested in experiment three.

-The authors
An example for a clear overstatement is found in the discussion (page 15, l.332): "Our findings suggest that that blocking beta-AR activity subsequent to memory destabilization prevents the synthesis of proteins". Protein synthesis has not been examined in this study. Rephrasing these examples and other parts in the manuscript would reframe the experiments into the right setting, i.e. systemic pharmacological challenge during a behavioural manipulation.
We have now changed the text on page 5-6, lines 100 to 104. Furthermore, we have replaced the word 'predictions' on line 76 in the introduction by 'hypotheses'. We also changed the text on page 11, lines 245 to 248.
Although we agree with the reviewer that we do not directly test whether propranolol affects the synaptic tag or protein synthesis, we actually formulated our hypotheses and design on basis of these processes. By simply ignoring these processes, the current designs and hypotheses would be hard to fathom. We agree though that we can only infer the critical conditions to target and observe memory reconsolidation from the underlying neurobiological processes. Therefore, we have now removed these inferences from the predictions and result descriptions in our manuscript.

-The authors do no report any exclusion of datasets in their analyses. Given that startle signals are not always perfect, I wonder if any participants were excluded or pre-selected?
Missing startle responses caused by recording artifacts or trials with excessive baseline activity (i.e., 0.5%) were excluded from the analyses. We explain this in lines 542-543.

-The results reported for experiment 2 and 3 (employing two CSs paired with an US) are
missing the comparison between the reactivated CS against the non-reactivated (e.g. CS1 vs CS2). Instead comparisons between the CS associated with the US and the control stimulus are employed. However, these comparisons do not reveal the effect of reactivation exerted on a CS that has been previously associated with the US. Hence, comparisons between these stimuli need to be included.
Following the suggestion of the reviewer, we have now included the comparisons between the reactivated CS1 and non-reactivated CS2 stimuli for both experiment_2 and experiment_3 -see pages 10-13 .

-On page 10 line 201, the authors state that planed comparisons (CS1 vs CS3) revealed reduced memory retention in the 1hour pill group and intact differentiation in the 2hour pill group. Is the difference (or interaction) between groups significant, as well? Or is the interaction with the factor group only significant during extinction? The same is true for the planned comparisons (CS1 vs CS3) between the sleep and no sleep group on page 12 line 251.
Here again, the crucial group comparison is not reported.  Mem. 17, 627-638). The study by Agren suggests that dopaminergic and serotonergic genes influence human fear memory reconsolidation, but these findings have neither been replicated, nor are they supported by animal research. More importantly, Agren et al did not manipulate memory reconsolidation by the administration of propranolol, but they tested the effect of the retrieval-extinction procedure, which is very different from pharmacologically inducing post-reactivation amnesia (see also Beckers & Kindt, 2017, Ann. Rev. Clin. Psychol. 13, 1).

The subjective evaluation of US intensities is around 3.
Please provide the range of that scale. US evaluation scores ranged from 0 to 5, where higher scores indicate more aversive, (see line 759).
2. How many participants in each experiment were aware of the stimulus contingencies? All participants were aware of the CS-US contingencies immediately after they underwent the fear conditioning procedure. We now clarify this in the legends of the Extended Data Figs. 1-3 on pages 34-36.
3. Please report the gender of the participants included in each sample and provide some evidence (e.g. testing in one of the experiments) that gender did not influence reconsolidation. We now report the gender of the participants included in each of the experiments (line 108, line 193 and line 256). However, the present sample sizes are too small to properly test for any gender effects. But note that in our previous studies we never detected differences between the gender groups. We have now clarified this in a footnote at the bottom on page 6.

Please provide information about the intake of food in each group, especially in the comparison between different time-points of drug administration. Food might have influenced the pharmacokinetic profiles.
After acquisition, participants were instructed to refrain from caffeine and alcohol during the 12 h and to avoid food and drinks other than water during the 2 h prior to memory reactivation. We now clarify this on page 19 and page 21. 5. Please consider exchanging "biochemical signature" in the manuscript (e.g. page 8 line 173) with "pharmacokinetic profile/signature". While this might sound picky, it is just a meant to be a constructive help to use standardized terms. We have now exchanged "biochemical signature" with "pharmacokinetic signature" throughout the manuscript (line 17, line 177, line 181, line 235, line 332, line 738).

Reviewer 2:
In 3 independent experiments the authors show in human participants that the expression of conditioned fear memories (measured via startle response potentiation) can be abolished if propranolol (a β-adrenergic antagonist) is administered during a specific time window following memory reactivation. Moreover, the reduction of fear responses on the long-term was found to be dependent on intervening sleep. This is a nicely designed and well-controlled study that tackles a timely and important issue. The methods are sound and the manuscript is well-written. However, I have a few concerns that should be addressed.
1 -My main concern is with regard to the authors' conclusion of a specific time window for β-AR activity of 2-3 hours post-reactivation (Fig. 3). In my understanding of the data, it cannot be excluded that β-AR activity during the first 1-2 hours plays a role for memory reconsolidation as well. What the data clearly show is that after 3 hours β-AR activity is (probably) no longer required. (Although even this conclusion is shaky because later intervals were not tested, e.g. it is possible that there is a second time window, e.g. around 6 hours post-reactivation, with a β-AR dependency.) But more importantly, based on Fig. 3 it cannot be excluded that propranolol acted on reconsolidation processes during the first 2 hours after reactivation. The drug administered 1 h before and right after reactivation could also have exerted an effect sooner than the 2-3 h time window. In order to show that in this phase β-AR activity is not required for reconsolidation, earlier drug administration would have to be tested, e.g. 2 vs. 5 hours before reactivation. This issue should be appropriately discussed and the conclusions toned down.
We inferred the specific time window of β-AR activity by combining the different timings of drug administration, the pharmacokinetics of propranolol and the presence/absence of changing the expression of fear memory. This has now been explained in more detail in the discussion on page 15, lines 331-349.

-
The observed sleep effect is particularly novel and interesting and I wonder about the possible underlying mechanisms of this effect. The authors discuss synaptic downscaling as one possibility, which seems to be reasonable. An alternative mechanism that came to my mind is the formation and preferential consolidation of a new memory trace. During the reactivation session, participants may acquire a new safety memory trace, i.e. learning that the CS1 is no longer paired with the US. Sleep may then foster the preferential consolidation of this new safety memory trace that interferes with the fear memory trace.
It is highly unlikely that 1 unreinforced trial will trigger extinction learning thereby forming an inhibitory or safety memory. Actually the data in the non-effective conditions show that 1 unreinforced trial did not trigger extinction learning: The startle fear response from acquisition to test after one unreinforced trial did not decline at all (i.e., nadolol group in Fig. 1, pill_2h group in Figure 2, or the first retention test 12 h later without sleep in between in Fig. 4). In our previous work we have also demonstrated that a reinforced trial CS+ followed by propranolol neutralized the fear memory the following day, as long as the memory reactivation involves a prediction error (see Sevenster, Beckers & Kindt, 2013, Science, 339, 830-833). These observations can be considered as a falsification for the alternative hypothesis of inhibitory learning as opposed to weakening the excitatory fear memory.

-The sleep and no sleep groups of Exp. differ in several regards. First, considering that
propranolol was administered after reactivation, the active phase of the drug fell mainly in the sleep state for the sleep group and in the wake state for the no sleep group. Do the authors know of any evidence showing that propranolol may act differently during sleep and wakefulness? A number of hormones, neurotransmitters and plasticity-related processes differ between sleep and wakefulness and, thus, the effects of propranolol may also differ. Second, when did acquisition/extinction take place for the sleep and no sleep groups? If it occurred at the same time of day for both groups, this would mean that there were differences in the retention interval between acquisition and reactivation as well as between test and extinction for the two groups. Alternatively, if acquisition/extinction took place also in the evening and morning, respectively, can the authors exclude any potential circadian effects?
Although the sleep and non-sleep conditions differ indeed in several ways, there are strong arguments why these differences cannot explain the current observations. Even though the fear-reducing effects were initially not observed in the group who had the memory reactivation in the morning and a retention test 12 h later at the same day, this group also showed amnesia for learned fear when they were again tested 24 h later after a night of sleep. In view of the bioavailability of propranolol during the first hours of the night in the sleep-group, the drug was no longer available in the other no-sleep group, while they showed a similar fear reduction when tested 24 h later. This has now been extensively explained in the discussion section, see pages 16-17, lines 370-387, lines 396-400. Although the current study did not test for the effect of circadian rhythm, we refer to literature suggesting that it is not very likely that the relation of sleep and memory is explained by circadian rhythm (lines 312-315).

A few methodological details should be clarified:
o In Exp. 1, in the instruction for participants it says that "an electric shock would follow one of the slides in most cases". Was the CS1 actually followed by an electric shock in all cases? Why were participants instructed otherwise? And why was the instruction different in Exp. 2, where it says "in all cases"? In exp_1, the CS1 was in fact followed by the US on an 80% reinforcement scheme (i.e., non-asymptotic learning), which we have now clarified in lines 438-439. But in exp_2 and exp_3, the CS1 and CS2 were followed by the US in all cases (i.e., on a 100% reinforcement scheme or asymptotic learning).
o Line 376: what does randomized within blocks (i.e., CS1 -CS2 -NA) mean? Were stimuli of each category always presented consecutively in one block, i.e. all CS1 one after the other etc.? We now clarify that order of trial type was randomized within blocks of 3 trials (i.e., CS1, CS2, and NA) such that no more than two trials of the same type occurred in succession -see lines 441-443.
o Line 385/396: were CSs presented without US here? In lines 453-454 and lines 502-503 it is now clarified that during MR as well as during extinction learning the CSs were presented without the US.
o It is said that the drug was applied in a single blind fashion? What does that mean? Who was blind, the participants or the experimenter? We now clarify in lines 450-451 (double blind), lines 505-506 (single blind), lines 519-520 (single blind) that the participants were blind to medication assignment.
o Missing startle responses (0.5%) were excluded. How did the authors define missing startle responses? Missing startle responses were caused by recording artifacts, which is now clarified in lines 542-543.