Selective early-acquired fear memories undergo temporary suppression during adolescence

Fear conditioning across adolescence.

To explore developmental distinctions in fear learning and memory, we used three separate cohorts of mice [postnatal day (P) 29, 39, and 49]. These ages approximate early, mid, and late adolescence, respectively (7, 22, 25). Ten-week-old adult mice (P70) were used as a basis for comparison with existing adult fear-conditioning literature. Mice of all ages were fear-conditioned with three tone-shock pairings in the conditioning context, context A (grid floor, peppermint odor). The age groups under study did not have differences in foot-shock pain sensitivity (22), showed similar baseline locomotor activity (Fig. S1A), and did not differ in freezing behavior during fear acquisition (Fig. S1B). We assessed hippocampal-dependent contextual fear by returning the mice to context A 24 h postconditioning and examining the levels of freezing behavior (as previously described in ref. 26) (Fig. 1A). P29 mice froze significantly less than older mice [P39, P49, and adult mice, ANOVA, F_{(3, 32)} = 15.073, P < 0.001] (Fig. 1B and Fig. S2A), indicating the absence of a contextual fear response. In an attempt to enhance contextual fear memory expression, we fear-conditioned a separate cohort of P29 mice using a longer preshock acclimation period of 5 min, as this leads to increased levels of contextual fear in adult mice (Fig. S2D) (20), and also conditioned a separate cohort of mice to the context only, without any tone cues, but the apparent absence of a contextual fear response persisted in both cases (Fig. S2 B and C). To assess amygdala-dependent cued fear, we presented the mice with three tones in a novel context, context B (green cylinder, lemon odor), 48 h postconditioning. In contrast to low levels of contextual fear, P29 mice exhibited high levels of freezing to tone cues (Fig. S1C). This finding is not surprising, as the somatosensory and auditory processing associated with amygdala-dependent cued conditioning lack the hippocampally mediated spatial complexity associated with contextual fear conditioning (16). The amygdala-dependent freezing observed in younger mice was not attributable to inherent differences in auditory stimuli sensitivity (Fig. S1B) and confirms that the physiological response systems required for freezing behavior are intact.

Spatial memory.

The lack of evidence for hippocampal-dependent contextual fear in early-adolescent (P29) mice could have been due to either hippocampal immaturity or immature connectivity in the medial prefrontal (mPFC)-amygdala-hippocampal fear circuit required for this task. To test these two possibilities, we used a hippocampal-dependent, spatial task optimized for the narrow developmental time period of interest and for which there was no underlying fear component: novel object placement (as previously described in refs. 27–29) (Fig. 1C). Although the novel object placement task assesses a short-term form of spatial memory, we were able to rely on it for a basic readout of hippocampal-dependent spatial learning. As P49 mice displayed hippocampal-dependent contextual fear indistinguishable from that of P70 mice, mice were tested at three time points (P29, P39, and P49) for preference of a novel object location. Acclimation trials confirmed that there were no differences in locomotor activity or exploratory behavior between age groups, and all age groups explored objects placed in novel locations significantly more than objects placed in familiar locations [P29: 55.65 ± 7.89 s novel to 22.90 ± 4.29 s familiar, trial 2 (T2) fraction of 62.40% ± 4.67; P39: 52.01 ± 5.02 s novel to 26.64 ± 3.99 s familiar, T2 fraction of 67.33% ± 3.86; P49: 51.38 ± 2.65 s novel to 22.92 ± 3.32 s familiar, T2 fraction of 71.22% ± 3.17; two-tailed t tests; P29: t₍₂₇₎ = 2.657, P < 0.05; P39: t₍₂₇₎ = 4.491, P < 0.001; P49: t₍₂₇₎ = 6.692, P < 0.001], indicating intact hippocampal-dependent spatial abilities. Thus, in contrast to the contextual fear test results, P29 mice do not differ from older mice in their ability to successfully perform a basic, hippocampal-dependent spatial task [ANOVA, F_{(1, 27)} = 1.1470, P = 0.248]. Lack of contextual fear, but spared performance on the novel object placement task, indicates a selective lack of expression for hippocampal-dependent emotional processes in early-adolescent mice and potential alterations in integration of the mPFC-amygdala-hippocampal fear circuitry.

Preadolescent fear conditioning.

Because previous groups have shown that aversive contextual learning is present in rats at ages younger than P29 (19, 20, 30), we wanted to further understand why contextual fear expression was absent or suppressed in the early-adolescent mice. To compare the observed contextual fear responses with previous studies, we tested separate cohorts of mice immediately before (P23–P27) and after (P31–P49) the transition into adolescence. Interestingly, mice fear-conditioned at early preadolescent ages displayed intact contextual fear expression that was indistinguishable from adult levels (Fig. 1D). As the mice transitioned into adolescence at P29, the expression of contextual fear was suppressed. Then, as the mice transitioned out of adolescence (P39), the expression of contextual fear reemerged (Fig. 1D). To assess whether a consolidated memory was suppressed during this transition into adolescence, separate cohorts of P27 mice were fear-conditioned and tested at P30 for contextual fear suppression. Whereas mice fear-conditioned at P27 and tested at P28 exhibit high levels of freezing behavior, mice conditioned at P27 and tested at P30 exhibit low levels of freezing behavior when tested for contextual fear (Fig. S3A). These results are consistent with the behavioral results observed in Fig. 1 and demonstrate that even previously consolidated fear memories are suppressed as the mice enter the adolescent transitional period (Fig. S3A).

Postadolescent retrieval of earlier acquired fear memories.

In subsequent experiments, we targeted the potential physiological underpinnings of the suppressed contextual fear behavior at P29. Because preadolescent mice are capable of expressing a contextual fear response before P29, the lack of fear expression would likely not be explained by simple structural maturation in the fear-conditioning circuitry. Rather, a more complex series of events may be occurring during this sensitive developmental window. For successful expression of contextual fear, the conditioned stimulus (CS) and unconditioned stimulus (US) must first be successfully paired (acquisition). Subsequent to this pairing, the association must be consolidated such that it can be accessed later. Finally, when the CS is encountered again, the animal must be able to retrieve the associated memory in order for the appropriate response to be made (31). Because the lack of contextual fear expression at P29 could be due to selective failures in any of the aforementioned learning processes, we conditioned a separate cohort of P29 mice and tested them for contextual fear at both early (24 h) and late (between 3 and 14 d) time points, without the presence of reminder cues or additional US presentations (Fig. 2A). Because stable adult-like levels of contextual fear are not reached until between P39 and P49 (Fig. 1B), 14 d was chosen as the initial delay interval. Despite poor performance when tested 24 h postconditioning, mice that underwent fear conditioning at P29 demonstrated significantly higher levels of contextual fear expression 14 d postconditioning, as evidenced by increased freezing in context A [ANOVA, F_{(1, 15)} = 15.541, P = 0.001] (Fig. 2B). This increase in freezing response was also context-specific, as control and fear-conditioned mice had equivalently low levels of freezing when tested in context B 15 d postconditioning [ANOVA, F_{(1, 15)} = 0.545, P = 0.472] (Fig. S3B). To eliminate the possibility that the initial 24-h context test might be serving as a reminder and thus promoting an unmasking of the subsequent freezing after the 14-d delay, we fear-conditioned separate cohorts of P29 mice and tested them for contextual fear at various time points without administering the initial 24-h contextual fear test (Fig. S2 E and F). In addition, to confirm that this delayed expression of contextual fear memory occurs near the end of the adolescent period, separate cohorts of mice conditioned at P29 were tested throughout the 14-d interval (tested at P30–P43). Increased expression of contextual fear was not observed until 13 d postconditioning (Fig. S3C). This enhanced freezing response was specific to contextual fear conditioning, as control mice that had been exposed to the context and tone, but not the shock, did not show similarly elevated freezing levels when reexposed to context A 14 d later (Fig. 2B). Mice conditioned as adults froze significantly more than control mice at both test points, but did not show a similar augmentation of the fear response after the 14-d interval [ANOVA, 24 h: F_{(1, 18)} = 190.38, P < 0.001; 14 d: F_{(1, 18)} = 53.672, P < 0.001] (Fig. 2B). Adult fear did not generalize to the novel context after the delay interval [F_{(1, 18)} = 3.48, P = 0.065] (Fig. S3B), eliminating the possibility of a generalized, context-independent fear incubation artifact (32). Combined with the lack of contextual fear during early-adolescent time points, the delayed onset of contextual fear shows that the early-adolescent mice are indeed capable of contextual fear learning but that the fear undergoes a latent period in which the expression of the prototypical fear response associated with reexposure to the conditioning context is suppressed until early adulthood.

Basal amygdala electrophysiology.

To explore the potential physiological mechanism underlying the observed delay in contextual fear memory expression, we examined potential developmentally regulated changes in amygdala synaptic transmission during the periods in which we observed changes in behavioral performance. As the function of the hippocampus differs in recent versus remote memories (33) and because lesions to the basolateral amygdala complex have been shown to specifically impair long-term contextual fear memory (34), the amygdala was chosen as a stable indicator of long-term contextual fear-associated physiology. Based on previous characterization of amygdala subnuclei in fear learning (35–37) and intact cued fear observed in all age groups (Fig. S1C), we did not anticipate developmental differences in the lateral nucleus of the amygdala (LA). Given the specific lack of a contextual fear response at P29, followed by a recovery in contextual fear response after the 14-d delay interval, we hypothesized that there may be developmentally regulated functional differences in the basal nucleus (BA) of the amygdala, as this region has been strongly implicated in contextual fear (35–37). In vitro electrophysiological studies have demonstrated behaviorally induced modification of synaptic strength (38–40). Together with this finding and the known involvement of the BA in contextual fear conditioning, we tested whether contextual fear learning induces synaptic potentiation in the BA (36). We examined the field excitatory postsynaptic potential (fEPSP) slope recorded from the BA by stimulating the perirhinal cortex in brain slices (Fig. S4A) (41, 42). Consistent with intact fear memory in adult mice (Fig. S4D), we found an enhancement of fEPSP slope in the BA of fear-conditioned adult mice compared with the control mice, suggesting that conditioned fear led to long-term potentiation–like phenomena in BA synapses (Fig. 3A) [two-way ANOVA, F_{(1, 32)} = 17.45, P < 0.001]. This enhancement of BA fEPSP was also seen in P27 mice after fear conditioning (Fig. S4 B and C). Consistent with the observed suppression of contextual fear in P29 mice, fEPSP slopes from P29 fear-conditioned and control mice were not significantly different, suggesting the absence of synaptic potentiation in the BA of P29 animals after fear conditioning (Fig. 3B and Fig. S4E) [two-way ANOVA, P > 0.05]. This blunted response of the BA in P29 mice is in contrast to enhanced potentiation that is observed in the LA when P29 mice are tested for cued fear, confirming the specificity of the basal nucleus in the contextual fear memory (Fig. S5). Because the mice that underwent fear conditioning at P29 demonstrated significant improvements in contextual fear expression 14 d postconditioning (Fig. S4F), we hypothesized that the mice that underwent fear conditioning at P29 might exhibit synaptic potentiation in the BA 14 d postconditioning. Indeed, consistent with behavior, we observed a significantly higher fEPSP slope in the BA 14 d postconditioning in mice that underwent fear conditioning at P29, suggesting that improvements in contextual fear expression observed 14 d postconditioning involved a delayed synaptic potentiation in the BA (Fig. 3C) [two-way ANOVA, F_{(1, 28)} = 21.1, P < 0.001]. Therefore, the suppression of fear expression at P29 is associated with an absence of immediate synaptic potentiation in the basal amygdala which is restored after the 14-d delay interval.

Although early-adolescent mice show no evidence of learning and have an initially blunted BA synaptic response at the 24-h contextual fear test, intact fear expression and enhanced BA reactivity at later time points suggest that the contextual fear memory was acquired. Because the 14-d interval falls beyond the limited time window of protein synthesis–independent short-term memory (31), memory consolidation was also not likely disrupted. Furthermore, as the mice were returned to context A for a 24-h postconditioning test, reconsolidation does not appear to be disrupted either, leaving retrieval as a remaining memory process in question.

Retrieval-associated signaling in the hippocampus.

To examine the role of memory retrieval in the suppression of P29 contextual fear, we examined intracellular signaling within the hippocampus of P29 and adult mice, as contextual fear is mediated by both the amygdala and hippocampus. Unlike the previous experiments, which detailed electrophysiological recordings in the amygdala for a long-term memory of up to 2 wk, the hippocampus was chosen as the site to examine molecular signaling in the retrieval of 24-h contextual fear memory. Whereas it is well-established that the ERK/MAP kinase (ERK/MAPK) signaling cascade is involved in associative learning and the formation of contextual fear memories (43, 44), it has recently been shown that PI3K activity in the hippocampus is necessary and sufficient for the activation of p42 MAPK (ERK2) during contextual fear memory retrieval (45). To test the involvement of hippocampal PI3K and ERK/MAPK molecular cascades in contextual fear memory retrieval, and the potential lack thereof in early-adolescent mice, we fear-conditioned separate cohorts of adult and P29 mice and then divided them into two groups: a retrieval group that was tested for contextual fear 24 h postconditioning and a control group that was sacrificed 24 h postconditioning without experiencing a retrieval session. Fifteen minutes after contextual fear memory retrieval, mice in the experimental group were sacrificed for Western blot analysis. Consistent with previous results (45), phosphorylation levels of PI3K and MAPK downstream targets Akt and ERK2 were significantly increased in the hippocampi of adult mice after contextual fear memory retrieval [ANOVA, phosphorylated Akt (pAkt): F_{(1, 17)} = 8.993, P < 0.01; phosphorylated ERK2 (pERK2): F_{(1, 17)} = 10.595, P < 0.01] (Fig. 4 A and B). In contrast, pAkt and pERK2 levels remained unchanged in the hippocampi of P29 mice regardless of whether the mice underwent a retrieval session [ANOVA, pAkt: F_{(1, 12)} = 0.278, P = 0.607; pERK2: F_{(1, 12)} = 0.026, P = 0.876] (Fig. 4 A and B). The absence of a contextual fear response at the 24-h test combined with the blunted response of retrieval-associated signaling suggests that the suppression of contextual fear in early-adolescent mice is associated not only with a lack of BA activity but also with a lack of memory retrieval-associated signaling in the hippocampus.