GluN2B and GluN2A-containing NMDAR are differentially involved in extinction memory destabilization and restabilization during reconsolidation

Extinction memory destabilized by recall is restabilized through mTOR-dependent reconsolidation in the hippocampus, but the upstream pathways controlling these processes remain unknown. Hippocampal NMDARs drive local protein synthesis via mTOR signaling and may control active memory maintenance. We found that in adult male Wistar rats, intra dorsal-CA1 administration of the non-subunit selective NMDAR antagonist AP5 or of the GluN2A subunit-containing NMDAR antagonist TCN201 after step down inhibitory avoidance (SDIA) extinction memory recall impaired extinction memory retention and caused SDIA memory recovery. On the contrary, pre-recall administration of AP5 or of the GluN2B subunit-containing NMDAR antagonist RO25-6981 had no effect on extinction memory recall or retention per se but hindered the recovery of the avoidance response induced by post-recall intra-CA1 infusion of the mTOR inhibitor rapamycin. Our results indicate that GluN2B-containing NMDARs are necessary for extinction memory destabilization whereas GluN2A-containing NMDARs are involved in its restabilization, and suggest that pharmacological modulation of the relative activation state of these receptor subtypes around the moment of extinction memory recall may regulate the dominance of extinction memory over the original memory trace.

Pre-recall intra-CA1 administration of the non-subunit specific NMDAR antagonist AP5 does not affect SDIA extinction memory expression or retention but hampers the amnesic effect of reconsolidation blockers. To evaluate the effect of pre-recall hippocampal NMDAR inhibition on SDIA extinction memory, we submitted SDIA-trained animals to the extinction protocol described above. Twenty-four (a) Animals were trained in SDIA (TR; 0.4 mA/2 s) and beginning 24 h later they were submitted to one daily extinction training trial for 5 consecutive days. Twenty-four hours after the last extinction training trial, SDIA extinction memory was reactivated (RA) and, 5 min thereafter, the animals received bilateral intra-dorsal CA1 infusions of vehicle (VEH; 1% DMSO in saline) or the NMDAR antagonist AP5 (5 µg/side). Retention was assessed 1 day and 7 days later (Test). (b) Animals were treated as in A except that they received intra-CA1 infusions of VEH or AP5 6 h after RA. (c) Animals were treated as in A, except that RA was omitted (No RA). (d) Animals were treated as in A, but a group of them received VEH or AP5 5 min after a pseudo-reactivation extinction session (pRA) carried out in an SDIA training box modified to be non-aversive (NA) for SDIAtrained animals (test latency in SDIA: Median = 162 s; IQR = 74-244.5 s; test latency in NA: Median: 10 s; IQR = 7.5-17.5 s; U = 3.00, p = 0.0003, SDIA vs NA in Mann-Whitney test). The non-aversive box was similar in dimensions to the SDIA-training apparatus but was painted gray and the elevated platform was made of transparent plexiglass instead of wood. Data are expressed as median ± IQR. (***) p < 0.001 versus VEH in Mann-Whitney test; n = 10-12 animals per group. www.nature.com/scientificreports/ hours after the last extinction training trial the animals received bilateral intra-dorsal CA1 infusions of VEH or AP5 and 20 min later were submitted to RA. We found that pre-RA AP5 did not affect SDIA extinction memory recall or retention ( Fig. 2a) but impeded the recovery of avoidance induced by post-RA intra-CA1 administration of rapamycin (RAP; 0.02 µg/side), an inhibitor of mammalian target of RAP (mTOR), a kinase that regulates synaptic protein synthesis through the phosphorylation of eukaryotic initiation factor 4E-binding protein 1 and p70 ribosomal S6 kinase 19 and is required for SDIA extinction memory reconsolidation in the hippocampus 4 ( Fig. 2b Hippocampal GluN2B-containing NMDARs mediate SDIA extinction memory destabilization during recall whereas GluN2A-containing NMDARs are necessary for SDIA extinction memory reconsolidation. GluN2A-containing and GluN2B-containing NMDAR regulate different cellular events in the amygdala during fear memory reconsolidation, controlling restabilization and destabilization phases, respectively 13,20 . Therefore, because our results with AP5 are consistent with a dual role of hippocampal NMDAR in extinction memory reconsolidation, we analyzed whether GluN2A-containing NMDAR and GluN2B-containing NMDAR also mediate SDIA extinction memory restabilization and destabilization differentially. To evaluate the involvement of these two NMDAR subtypes in extinction memory restabilization, we submitted SDIA-trained animals to the SDIA extinction protocol, as described above. One day after the last extinction training session, SDIA extinction memory was reactivated and 5 min later the animals received bilateral intradorsal CA1 infusions of VEH, the GluN2B-containing NMDAR antagonist RO25-6981 (RO; 2.5 µg/side), or the GluN2A-containing NMDAR antagonist TCN201 (TCN; 0.05 µg/side). As can be seen in Fig. 3, TCN, but not RO, impaired extinction memory retention and induced SDIA memory recovery 1 day and 7 days post-RA (Fig. 3a   RO did not affect extinction memory recall 20 min post-injection either but had no effect per se on retention and blocked the recovery of avoidance induced by post-RA RAP administration (Fig. 4b, 1 (Fig. 4c)

Discussion
Our findings confirm that extinction learning does not erase SDIA memory but creates a new trace that competes with it for controlling behavior, and corroborate that SDIA extinction memory enters an instability phase when recalled and must be restabilized through reconsolidation to maintain the learned avoidance response inhibited. Importantly, our data demonstrate that SDIA extinction memory destabilization and restabilization necessitate hippocampal NMDARs activation and indicate that GluN2B-and GluN2A-containing NMDAR www.nature.com/scientificreports/ subtypes, respectively, are differentially involved in these processes. This assertion is based on experiments showing that post-recall infusion of AP5 and TCN, but not of RO, impaired extinction memory retention lastingly and induced the recovery of avoidance in a time-dependent manner whereas pre-recall administration of AP5 and RO, but not of TCN, rendered the reactivated extinction memory trace resistant to the amnesia caused by mTOR inhibition. Our conclusions are further supported by the fact that neither AP5 nor TCN affected extinction memory when its reactivation was omitted or when they were given after exploration of a non-aversive environment, as well as by results showing that pre-recall infusion of AP5, TCN and RO had no effect on SDIA memory expression. Moreover, our data agree with reports that pre-recall administration of non-subunit selective or GluN2B-selective NMDAR antagonists prevents contextual fear memory updating 10,21 and others showing that destabilized memories require GluN2A-containing NMDAR upregulation to regain stability and persist 13,22,23 . Milton and coworkers 13 attributed the differential involvement of GluN2B and GluN2A-containing NMDARs in the destabilization and restabilization of reconsolidating memories to the fact that GluN2B-NMDARs regulate the activation state of the protein degradation system that modulates memory lability 24 whereas GluN2A-NMDARs promote CREB phosphorylation 25 and long-term potentiation 26 , which are associated with memory re-encoding and maintenance [27][28][29] . Psychotherapeutic interventions based on memory extinction are initially effective in reducing the exacerbated recall of traumatic events that afflicts post-traumatic stress disorder (PTSD) patients. However, PTSD symptoms usually return spontaneously, or due to unexpected triggers, after the end of psychotherapy. Therefore, most research on the psychopharmacology of extinction has revolved around the search for tools to improve extinction learning, based on the premise that this could prevent fear recovery. Because it has been repeatedly shown that NMDAR blockade impairs extinction retention whereas NMDAR function enhancement facilitates fear extinction in experimental animals [30][31][32][33] , NMDARs have taken center stage in this enterprise. In this regard, it has been reported that the NMDAR partial agonist d-cycloserine (DCS) increases exposure therapy effects [34][35][36] , although the available data are inconclusive 37,38 and experimental animals as well as humans treated with DCS during extinction learning show substantial recovery of learned fear, suggesting that DCS may not prevent PTSD relapse [39][40][41][42] . These studies, together with several others, suggest that extinction can indeed be modulated by drugs, but not lastingly [43][44][45] . Our findings that SDIA memory is still able to assume control of behavior after www.nature.com/scientificreports/ undergoing an extinction procedure that generates an extinction memory resistant to spontaneous recovery, renewal and reinstatement but sensitive to recall-induced GluN2B-containing NMDAR-dependent destabilization indicate that it is this destabilization what enables the reappearance of avoidance, and lead us to propose that blockers of these receptors might be suitable tools to prevent PTSD relapse. It is crucial to emphasize here that memory destabilization can result not only from explicit recall, but also from exposure to subtle signals and cues unable to elicit discernible behavioral responses 46 . In addition, it is important to consider that because of local animal housing regulations and facility constraints our experiments were carried out exclusively on adult male rats although sex and age related differences in NMDAR endogenous modulation and subunit composition have been reported [47][48][49] , women are twice as likely to develop PTSD than men 50,51 , and PTSD management in children might require a differential therapeutic approach 52,53 . To do that, 2-4 h after the last behavioral test we infused 1 μl of 4% methylene-blue as described above; the extension of the dye 30 min thereafter was taken as an indicator of drug diffusion. Only data from animals with correct implants were analyzed (see Supplementary Fig. S1).

Methods
Step-down inhibitory avoidance (SDIA) training. The SDIA training box (50 × 25 × 25 cm) consisted of a Plexiglas chamber with a grid floor through which a scrambled electric shock could be delivered to the rat's feet and a wooden platform (5 × 8 × 25 cm) at the left end of the grid floor. For training (TR), each animal was placed on the platform and when it stepped down and placed its four paws on the grid, received a mild foot shock (0.4 mA/2 s). Immediately thereafter the rat was withdrawn from the training box.
SDIA memory extinction protocol. To extinguish SDIA memory, animals were submitted to five daily extinction sessions during which they were placed on the training box platform and allowed to freely explore the apparatus for 30 s after stepping down to the grid. To reactivate the SDIA extinction memory (RA), the animals were placed on the platform and after stepping down from it were immediately removed from the training box. Data analysis. Statistical analyses were performed using GraphPad Prism 8 software. Significance was set at p < 0.05. A 300 s ceiling was imposed on retention test session latency. Data were analyzed using two-tailed Mann-Whitney U test, Kruskal-Wallis test followed by Dunn's post hoc comparisons, or two-way RM ANOVA followed by Bonferroni's multiple-comparisons test.

Data availability
Data are available upon request by contacting the corresponding author.