Dynamics of dendritic spines in the mouse auditory cortex during memory formation and memory recall

Significance

Memories are believed to be stored as long-lasting structural changes in synapses. We tested the hypothesis that formation of a tone-shock association induces changes in the mouse auditory cortex by combining chronic in vivo imaging of dendritic spines with auditory fear conditioning. We find that memory formation is correlated with a transient increase in spine formation that leaves a long-lasting trace in the network. Memory recall does not lead to a recapitulation of structural remodeling as observed during initial memory formation. Our findings provide a potential synaptic mechanism underlying previously reported functional changes in auditory cortex following fear conditioning and challenge models in which memory traces are modified upon memory retrieval.

Abstract

Long-lasting changes in synaptic connections induced by relevant experiences are believed to represent the physical correlate of memories. Here, we combined chronic in vivo two-photon imaging of dendritic spines with auditory-cued classical conditioning to test if the formation of a fear memory is associated with structural changes of synapses in the mouse auditory cortex. We find that paired conditioning and unpaired conditioning induce a transient increase in spine formation or spine elimination, respectively. A fraction of spines formed during paired conditioning persists and leaves a long-lasting trace in the network. Memory recall triggered by the reexposure of mice to the sound cue did not lead to changes in spine dynamics. Our findings provide a synaptic mechanism for plasticity in sound responses of auditory cortex neurons induced by auditory-cued fear conditioning; they also show that retrieval of an auditory fear memory does not lead to a recapitulation of structural plasticity in the auditory cortex as observed during initial memory consolidation.

Mammalian brains are characterized by a tremendous level of plasticity. This plasticity is believed to underlie the ability to extract and store information about past experiences and is crucial for animals and humans to interact adaptively in a changing environment. Therefore, detection and localization of a physical representation of a memory has been an intriguing aspect for a long time (1). Plastic changes in synapses are believed to be substrates of memory (2). The development of imaging techniques that allow chronic monitoring of dendritic spines, the morphological correlates of excitatory synapses on pyramidal neurons, in the living animal has provided valuable insights in the dynamics of neuronal circuits (3–5). It has recently been shown that not only chronic perturbations of sensory inputs (6, 7), but also temporally restricted learning experiences, impact the turnover of synaptic structures in the motor cortex and frontal association cortex of the mouse (8–10) and the high vocal center in zebra finches (11).

Auditory-cued fear conditioning (ACFC) is an associative learning paradigm that has been widely used to analyze mechanisms of learning in the auditory modality (12). During a conditioning session, subjects quickly learn to associate a previously neutral sound cue [the conditional stimulus (CS)] with an aversive stimulus like a mild foot shock [unconditional stimulus (US)]. It is well established that memory traces after initial formation undergo several processes at different time scales that lead to their consolidation and render them to a stable state that is, e.g., resistant to trauma introduced by an electroconvulsive shock (13). Interestingly, similar molecular cascades are triggered not only during memory formation, but also when a memory trace is retrieved (14, 15). Furthermore, memory traces that were recently retrieved become sensitive again to manipulations like electroconvulsive shock (16), blockade of NMDA receptors (17), or blockade of protein synthesis (18). These similarities have been suggested to reflect remodeling of memory traces following recall (14, 19). However, data on the dynamics of synaptic structures during memory recall is lacking up to date.

A number of brain structures have been identified mediating the formation of a memory induced by ACFC (12). Whereas inputs via the auditory cortex (ACx) to the amygdala, an essential brain structure for this learning paradigm, appear to be always sufficient to support fear conditioning (20), their necessity can depend on the spectrotemporal properties of the auditory CS (21, 22). The ACx as the primary sensory cortical area for the auditory modality has been extensively analyzed in the past during classical conditioning to sound stimuli (23–25) or pairing of sounds with artificial stimulation of the cholinergic system, which can substitute for aspects of the US (26–28). These paradigms lead to changes in the receptive fields of ACx neurons that are specific to the conditioned sound. There is evidence based on local pharmacological or optogenetic manipulations that plasticity of the ACx itself is necessary for experience-induced alterations in sound responses and does not simply reflect plasticity elsewhere in the auditory pathway (29, 30). Indeed, there is evidence based on electrophysiological recordings that synaptic plasticity at intracortical synapses can be induced by pairing of a sound with stimulation of the cholinergic system in vivo (28). Structural plasticity following ACFC has been observed in the frontal association cortex (10). However, it remains elusive if plasticity in sound responses in the ACx induced by ACFC (23–25) also has a structural correlate at the synaptic level.

In this study, we asked two major questions: Does ACFC in behaving mice induce structural plasticity in synaptic circuits of the ACx? To what extent do memory formation and memory recall share similarities at the level of synaptic structures? We addressed these questions by combining sound-cued fear conditioning and memory testing with chronic in vivo imaging of dendritic spines in the ACx.

Results

Auditory Cortex Lesions Impair Behavioral Expression of an Auditory Fear Memory.

As an indication for a role in memory storage, we first tested the necessity of the ACx for the expression of an auditory fear memory (21, 23, 29, 31, 32). In a previous study, we have identified a subset of specific, short, auditory broadband sound cues that required ACx function during the acquisition of the task when used as a CS (22). We chose one of those sounds as CS (complex 1; see Fig. S2A). Whereas in this previous study lesions of the ACx were performed before conditioning, we now asked if lesions of the ACx performed after conditioning would impact on memory recall and expression of the fear memory. First, two groups of mice were conditioned to a complex CS, and memory was tested on the next day (Fig. 1A; see Methods for details). We observed increased freezing during CS presentation in comparison with the silence that was similar in both groups, as expected [sham, silence: 4 ± 3%, lesion, silence: 6 ± 2%, Wilcoxon rank-sum test, P = 0.74, not significant (n.s.); sham, CS: 31 ± 5%; lesion, CS: 37 ± 7%, Wilcoxon rank-sum test, P = 0.63, n.s.]. On the next day, one group of mice underwent focal, bilateral lesions of the ACx (lesion: n = 8), whereas the other group underwent sham surgery (sham: n = 7). When testing both groups again, we observed a substantial reduction of freezing during the CS presentation in the lesioned group (sham, silence: 3 ± 1%, lesion, silence: 2 ± 1%, Wilcoxon rank-sum test, P = 0.75, n.s.; sham, CS: 26 ± 5%; lesion, CS: 12 ± 3%, Wilcoxon rank-sum test, P < 0.05). Finally, the mice were killed and the position and size of the lesion was verified (Fig. S1). On average, lesions covered ACx and extended partially to some of the surrounding areas.

Fig. 1.

ACx lesions impair expression of a fear memory. (A, Left) Mice were habituated, conditioned, and memory-tested on the following day. Subsequently, ACx was lesioned in one group (n = 8), and the other group (n = 7) underwent sham surgeries. (A, Right) Freezing levels were measured before and after surgery. Freezing during CS presentation was significantly reduced in the lesioned group compared with the sham group in the test following surgery, but not before or during silence. (B, Left) Mice were habituated, conditioned, and a week later underwent lesion (n = 8) or sham surgery (n = 9) and then memory-tested. Subsequently, the same mice were conditioned again to a 4-kHz pure tone, a CS that does not require ACx. (B, Right) Freezing level in memory test of lesioned group was significantly lower than sham controls but similar after reconditioning. Bars represent mean ± SEM; asterisks indicate significant differences between sham and lesioned mice.

In the next series of experiments we asked if the ACx is essential only shortly after fear conditioning or if it is still required during extended periods following conditioning. The main difference with the previous experiment is that we performed sham (n = 9) and lesion (n = 8) surgeries a week after the conditioning session and omitted the memory test directly following conditioning to exclude potential effects induced by memory recall (Fig. 1B). In a memory test session, we again observed decreased levels of freezing in lesioned mice during the presentation of the CS (sham, silence: 9 ± 3%, lesion, silence: 3 ± 1%, Wilcoxon rank-sum test, P = 0.24, n.s.; sham, CS: 37 ± 5%; lesion, CS: 23 ± 5%, Wilcoxon rank-sum test, P < 0.05), indicating that ACx function is also required for expression of more remote auditory fear memories. It has been shown previously that conditioning to a pure tone CS is possible when the ACx is lesioned before conditioning (20, 22). We thus conditioned both groups of mice again to a 4-kHz pure tone. Freezing levels were similar in both groups, indicating that ACx lesions did not lead to a general impairment in freezing (sham, silence: 14 ± 3%, lesion, silence: 19 ± 5%, Wilcoxon rank-sum test, P = 0.41, n.s.; sham, CS: 41 ± 7%; lesion, CS: 50 ± 5%, Wilcoxon rank-sum test, P = 0.22, n.s.). In summary, both experiments demonstrate that lesions to the ACx lead to at least partial impairment in memory recall and expression.

Memory Recall Induces Expression of Immediate Early Genes in the Auditory Cortex.

Our lesion experiments demonstrate that the ACx is necessary for fear memory expression to specific sounds. Is the ACx storing at least part of the memory trace or is ACx necessary for successful memory readout from other brain regions? The temporary induction of immediate early genes (IEGs) for a few hours following learning experiences has been interpreted as a signature for the specific involvement of neurons during the formation of a memory trace (33, 34). We have previously demonstrated the induction of the IEGs c-fos and Arc/arg3.1 in the mouse ACx following fear conditioning to complex sounds (22). Interestingly, induction of IEGs has also been reported in specific brain structures following memory recall and was interpreted as a sign of reconsolidation processes occurring in those circuits storing the memory trace (35, 36).

To test the involvement of the ACx during memory recall, we analyzed the relative expression of the IEGs Arc and c-fos by quantitative PCR (qPCR) in ACx obtained from mice after memory retrieval (n = 10), exposure to the sound cue without prior conditioning (n = 8), and taken directly from the home cage (n = 7; Fig. 2A). We observed a significant increase in mRNA levels of Arc and c-fos in the group of mice that had undergone paired conditioning, indicating an induction of IEGs by memory retrieval triggered by memory recall (Arc: paired, 0.020 ± 0.004, context exposure, 0.014 ± 0.003, home cage, 0.009 ± 0.001; one-way ANOVA, P < 0.01; significant post hoc Wilcoxon rank-sum test: paired vs. home cage: P < 0.01; c-fos: paired, 0.010 ± 0.002, context exposure, 0.005 ± 0.001, home cage, 0.005 ± 0.001; one-way ANOVA, P < 0.01; significant post hoc Wilcoxon rank-sum tests: paired vs. home cage: P < 0.01, paired vs. context exposure: P < 0.01). To test the involvement of the ACx in retrieval of more remote memories, we performed a similar experiment with the memory test session scheduled a week after conditioning (Fig. 2B). We observed again a significant effect of behavioral treatment on mRNA levels of Arc and c-fos (Arc: paired, 0.022 ± 0.003, context exposure, 0.012 ± 0.002, home cage, 0.009 ± 0.001; one-way ANOVA, P < 0.001; significant post hoc Wilcoxon rank-sum tests: paired vs. home cage: P < 0.01, paired vs. context exposure: P < 0.001; c-fos: paired, 0.009 ± 0.001, context exposure, 0.007 ± 0.001, home cage, 0.005 ± 0.001; one-way ANOVA, P < 0.01; significant post hoc Wilcoxon rank-sum test: paired vs. home cage: P < 0.001). For both sets of experiments, presentation of the sound to naive mice caused an intermediate effect, with significant differences to memory retrieval in short- (c-fos) and long-time (Arc) paradigms (37). These findings, together with our previous observations (22), indicate a role of the ACx during acquisition as well as storage and retrieval of auditory fear memories as judged by the induction of IEGs.

Fig. 2.

Retrieval of a memory induces IEG expression in ACx. (A, Left) Mice were habituated, and one group (n = 10) was conditioned and total RNA was extracted from ACx following a memory test. A second group (n = 8) was put in the conditioning context without sound or shock presentation. Subsequently, total RNA was extracted from ACx following a sound presentation. A third group of mice (n = 7) was kept in the home cage before total RNA extraction. (A, Right) Average Arc and c-fos mRNA expression levels in ACx measured by qPCR. Both Arc and c-fos mRNA levels were significantly increased in mice undergoing paired conditioning compared with home cage. (B, Left) Three groups of mice (n = 10, 10, and 7) underwent a similar paradigm as above; however, here the mice stayed in the home cage for a week before memory testing and RNA extraction. (B, Right) Average Arc and c-fos mRNA levels are again significantly increased in the paired conditioning group compared with the home cage. Bars represent mean ± SEM; asterisks indicate significant differences.

Intrinsic Imaging Allows Identification of GFP-Expressing Neurons in Cortex Areas Activated by the Conditional Stimulus.

Formation of long-term memories is thought to involve lasting structural changes in neuronal circuits (3–5, 8, 9). To directly assess the impact of ACFC on the structure of the circuits in the ACx, we combined behavioral training with in vivo two-photon imaging of dendritic spines through a chronically implanted glass window over the ACx. Imaging was performed in a transgenic mouse model that expresses GFP only in a small subset of neurons, primarily in layer 5 (Fig. 3A; Methods) (38). Because individual GFP-expressing neurons are stochastically distributed across the neocortex, it is important to identify those that are located in the ACx. Before spine imaging, we recorded the intrinsic signal from the cortex in response to a set of pure-tone stimuli to assess the tonotopic organization of the mouse ACx (Fig. 3B). Furthermore, we presented several broadband complex sound stimuli. Consistent with the tonotopic organization of the ACx, we observed that complex sounds elicited stronger responses and activated a larger area of the cortex (Fig. S2 A and B). Furthermore, a correlation analysis showed that the spatial structure of the intrinsic signals evoked by various complex sounds was more similar than that of different pure tones. Pure tone response patterns were more distinct and showed highest similarities to response patterns of pure tones of similar frequencies (Fig. S2C). Using the blood vessel pattern at the brain surface as reference, we were able to align functional maps obtained by intrinsic imaging and fluorescence images of the dendritic trees of individual GFP-expressing neurons (Fig. 3C). This approach allowed us to locate GFP-expressing neurons in functionally identified auditory cortical fields. GFP-expressing neurons were imaged from the area showing intrinsic signals evoked by sound complex 1 used as CS in conditioning experiments, which included primary ACx and additional auditory areas such as the anterior auditory field (39, 40).

Fig. 3.

Combining functional and structural imaging of the ACx. (A) Low-magnification image of the surface vessel pattern through a chronically implanted cranial window. (B, Upper) Intrinsic signals (average of 30 repetitions) in response to pure tone stimuli of increasing frequency shift systematically along cortical surface and reveal tonotopic organization of the ACx. (B, Lower) Complex stimuli, including sound used as CS in conditioning experiments (red label), activate ACx wider and more stereotypically than pure tones. (C) (i) Overlay of intrinsic signal elicited by a complex sound with vessel pattern; (ii) epifluorescence image of the apical tuft of a GFP-expressing neuron. Same vessels are visible as dark shadows. (iii) Maximum intensity projection of a two-photon image stack of the same dendritic tree. (iv) Two-photon images of the same dendritic section taken at different time points with examples of appearing (green arrows), disappearing (red arrows), and stable (blue arrows) spines.

Paired and Unpaired Conditioning Induce Specific Changes in the Turnover Rates of Dendritic Spines.

We analyzed dynamics of dendritic spines in three groups of transgenic mice with chronically implanted cranial windows over ACx (41) (Fig. 4A): following 3 d of habituation, two imaging sessions were scheduled on day 4. Three imaging sessions, again at a 2-h interval, were then performed on day 11. For each imaging session a mouse was anesthetized and image stacks of the same dendrites were acquired using two-photon microscopy. The three groups of mice differed only in the behavioral treatment during the first 2-h interval: The first group underwent paired conditioning and mice formed an association between CS and US. The second group underwent unpaired conditioning in which CS and US presentations were temporally separated such that the mice could not form an association. The third group of mice was transferred to the conditioning chamber without presentations of CS or US. All three groups of mice were transferred to their home cages during the first two imaging sessions on day 11; they were later transferred during the second and third imaging session to the memory test chamber, and the sound used for conditioning was presented. The sound presentation caused an increase in freezing levels specifically in the group of mice that previously had undergone paired conditioning (Fig. 4B; paired, n = 4: silence: 11 ± 4%, sound: 47 ± 14%, Wilcoxon rank-sum test, P < 0.05; unpaired, n = 5: silence: 14 ± 5%, sound: 19 ± 9%, Wilcoxon rank-sum test, P = 0.84, n.s.; context, n = 6: silence: 18 ± 6%, sound: 22 ± 9%, Wilcoxon rank-sum test, P = 1.0, n.s.). This behavior indicates two important points: first, the conditioning session a week before the memory test led to successful memory formation, and temporary anesthesia during imaging did not interfere with this process, because we observe similar freezing levels as before (Fig. 1). Second, the freezing indicated successful triggering of memory recall. In both remaining groups of mice, the sound presentation did not lead to increased freezing, as expected.

Fig. 4.

Conditioning correlates with transient changes in spine turnover. (A) Experimental design (see Results for details). (B) Freezing behavior measured in a test session between s4 and s5. Mice that underwent paired conditioning (n = 4) exhibit significant sound-induced freezing in contrast to unpaired (n = 5) and context exposure controls (n = 6). (C) Normalized spine count across sessions. Significant differences between groups were detected in imaging session 2. (D) Examples of spines formed (green arrows) and eliminated (red arrows) during 2-h intervals. (E) Analysis of the rate of spine formation in the three 2-h imaging intervals. The rate of spine formation is significantly increased between s1 and s2 in the paired conditioned group. (F) Analysis of the rate of spine elimination in the three 2-h intervals. The rate of spine elimination is significantly increased between s1 and s2 in the unpaired conditioned group. Bars represent mean ± SEM; asterisks indicate significant differences; number of analyzed dendrites is indicated above the bars.

As a first level of analysis we reconstructed dendrites using the best projection method (42), and appearing, disappearing, and persistent dendritic spines were identified (Fig. S3 A–C). Despite substantial turnover of spines, the total number of spines remained largely constant during the experiment in all three groups (Fig. 4C). However, a statistical analysis revealed small but significant differences in normalized spine numbers per dendrite in imaging session 2. The group of mice that underwent paired conditioning showed a gain in number of spines, whereas the unpaired group showed a loss of spines compared with the context exposure group [paired: session 2 (s2), 108.0 ± 2.6%, n = 18; s3, 100.1 ± 7.0%, n = 14; s4, 95.9 ± 7.5%, n = 14; s5, 96.4 ± 7.9%, n = 14; unpaired: s2, 95.1 ± 2.2%, n = 19; s3, 97.4 ± 4.9%, n = 12; s4, 95.4 ± 4.9%, n = 12; s5, 105.4 ± 5.7%, n = 7; context: s2, 100.8 ± 1.2%, n = 17; s3, 101.9 ± 3.1%, n = 15; s4, 99.5 ± 3.0%, n = 15; s5, 97.1 ± 3.5%, n = 13; two-way ANOVA significant interaction P < 0.05; significant post hoc Wilcoxon rank-sum tests: s2, paired vs. unpaired: P < 0.001; s2, context vs. unpaired: P < 0.01]. To analyze this effect in more detail, we calculated the rates of spine formation for the 2-h interval between imaging sessions 1 and 2. We observed that significantly more spines had been formed in the group that underwent paired fear conditioning compared with the two other groups (Fig. 4 D and E; s1→s2: paired: 16.7 ± 2.1%, n = 18; unpaired: 11.1 ± 1.5%, n = 19; context: 8.4 ± 1.3%, n = 17; one-way ANOVA, P < 0.05; significant post hoc Wilcoxon rank-sum tests: paired vs. context: P < 0.01, paired vs. unpaired: P < 0.05). When performing the same analysis for the 2-h intervals between session 3 and 4 and 4 and 5, similarly low spine formation rates were observed in all three groups of mice (s3→s4: paired: 7.2 ± 1.3%, n = 14; unpaired: 8.0 ± 1.4%, n = 12; context: 6.7 ± 1.1%; one-way ANOVA, P = 0.57, n.s.; s4→s5: paired: 7.1 ± 1.4%, n = 14; unpaired: 8.6 ± 1.8%, n = 7; context: 5.5 ± 1.1%; one-way ANOVA, P = 0.40, n.s.). This experimental design includes the group of conditioned mice in which a recall of the auditory fear memory was triggered, and also the context exposure group in which mice experience the novel sound cue for the first time.

We also analyzed the rate of spine elimination in the three groups of mice for the 2-h intervals (Fig. 4 D and F). We found that rates of spine elimination were generally in balance with the observed spine formation rates. However, we observed a specific, transient increase in the rate of spine elimination in the group of mice that underwent unpaired conditioning in imaging session 2 (paired: 10.8 ± 1.3%, n = 18; unpaired: 15.7 ± 1.9%, n = 19; context: 7.8 ± 1.1%, n = 17; one-way ANOVA, P < 0.01; significant post hoc Wilcoxon rank-sum test: unpaired vs. context: P < 0.001). Mice do not form a specific association between the CS and US during unpaired conditioning. However, it has been shown that foot shocks have an impact on the ACx even in the absence of a sound (22, 29), and the increased rate of spine elimination may be related to more general effects on dendritic morphology associated with stress or initial inhibitory conditioning (43, 44). For the other 2-h intervals we observed generally low spine elimination rates, again including the group of mice that underwent a memory retrieval test and the group of mice exposed to a novel sound cue (s3→s4: paired: 11.1 ± 1.8%, n = 14; unpaired: 9.9 ± 1.5%, n = 12; context: 8.8 ± 1.2%, n = 15; one-way ANOVA, P = 0.59, n.s.; s4→s5: paired: 7.2 ± 0.8%, n = 14; unpaired: 6.6 ± 1.2%, n = 7; context: 7.2 ± 1.1%, n = 13; one-way ANOVA, P = 0.81, n.s.). The analysis of spine formation and elimination rates were performed on a dendrite-based level, because single dendritic branches are believed to act as units of dendritic computation (45). However, similar results were obtained when pooling all spines per condition and time point (Fig. S3 D–J). When analyzing the survival rates of spines that already existed at the beginning of the experiment, we observed in all groups a similar level of survival rates at the short timescale (2 h). For a longer interval (1 wk), a slightly decreased survival probability for the unpaired group was observed (Fig. S4). Together, our analysis of turnover rates indicates that paired and unpaired conditioning, in contrast to memory recall, have specific effects on the synaptic network in the ACx.

Memory Formation Leaves a Long-Lasting Trace in the Network.

How long-lasting is the effect of a transient modulation of turnover rates on the network structure? We quantified the fraction of spines in the population at imaging session 3 that has been newly formed a week earlier between imaging sessions 1 and 2 (Fig. 5A). This fraction was significantly higher in the group of mice that experienced paired conditioning compared with the two other groups, indicating a long-lasting effect of memory formation on the ACx (paired: 10.0 ± 2.1%, n = 14; unpaired: 4.5 ± 1.3%, n = 12; context: 2.4 ± 0.6%, n = 15; one-way ANOVA, P < 0.01; significant post hoc Wilcoxon rank-sum test: paired vs. context, P < 0.01). Two factors can contribute to this effect: A transiently increased rate of formation, as observed (Fig. 4 D and E). In addition, the survival probability of the spines formed during learning could be particularly high, as previously shown in other brain regions (8, 9). In our dataset, spines formed during paired conditioning showed highest survival probabilities, but this trend was not significant (Fig. 5B; paired: 47.3 ± 8.2%, n = 11; unpaired: 35.0 ± 9.6%, n = 11; context: 34.4 ± 8.5%, n = 14; one-way ANOVA, P = 0.28, n.s.), which indicates that both the larger number of newly formed spines and a higher survival rate likely contribute to an increased fraction of spines in the network that were formed during learning.

Fig. 5.

Spines formed during conditioning remain stable in the population and are not affected by memory retrieval. (A) Spines formed between s1 and s2 represent a significantly larger fraction of all spines present at s3 in the group of mice that underwent paired conditioning compared with the other groups. (B) Analysis of the probability of spines formed between s1 and s2 to persist for a week until s3. Probability that these spines will persist during s3 and s4 (home cage) (C) and during s4 and s5 (sound presentation) (D). Survival rates are high in all three groups in both time points, including the group of mice that memory retrieval was triggered. Bars represent mean ± SEM; number of analyzed dendrites is indicated above the bars.

Memory Recall Does Not Affect Survival of Spines Formed During Conditioning.

At the behavioral level it was shown that reconsolidation processes are selective for those memories that have been recently retrieved (46). In our previous analysis of turnover rates, we were not able to detect effects triggered by memory recall at the level of synapses. However, we considered the whole population of dendritic spines present at the time of the memory recall, and this may dilute effects specific to a smaller subpopulation of spines. To address the question of whether memory recall selectively affects spines that have been formed during memory formation and that persisted for a week, we quantified the survival rates of this specific subpopulation of spines under basal conditions and memory recall. We found in all three groups of mice that the survival probability of spines formed during imaging sessions 1 and 2 was very high during imaging sessions 3 and 4 when mice were just brought back to their home cage (Fig. 5C; paired: 89.9 ± 4.9%, n = 11; unpaired: 100 ± 0%, n = 8; context: 86.7 ± 10.7%, n = 10; one-way ANOVA, P = 0.20, n.s.). Interestingly, the survival probability of these spines remained at a similarly high level also between imaging sessions 4 and 5 in all three groups, and did not reveal any effects of memory recall or exposure to a novel sound (Fig. 5D; paired: 92.4 ± 4.2%, n = 11; unpaired: 100 ± 0%, n = 7; context: 87.5 ± 13.7%, n = 8; one-way ANOVA, P = 0.35, n.s.). These results do not support models that propose structural remodeling of memory traces during memory recall, at least at the level of layer 5 pyramidal neurons in the ACx (19).

Discussion

In this study we investigated the impact of memory formation and memory recall on the structure of neural circuits in the mouse ACx in the context of ACFC. We made two main observations, summarized in Fig. S5. First, paired as well as unpaired conditioning leads to distinct, transient changes in the turnover rates of spines, which leave a long-lasting trace in the network. An increase of spine formation was observed after paired conditioning, and an increase of spine elimination after unpaired conditioning. Second, lesion experiments and induction of IEGs demonstrate a role of the ACx for the recall of the fear memory a week after conditioning. However, at the level of the structure of synapses, memory recall did not trigger processes similar to those during initial memory formation.

Observation of structural changes in neuronal circuits during learning experiences has provided in recent years an important link between synaptic plasticity and the formation of memories. Chronic imaging has revealed a transient increase in the rate of formation of spines in the mouse motor cortex during motor skill learning (8, 9) and during song learning in the zebra finch (11). Our observation of structural remodeling of spines in the ACx induced by a behavioral learning paradigm provides direct evidence for synaptic plasticity as a cellular mechanism that could underlie alterations in sound responses following fear conditioning (23). The observed transient increase in spine formation in the ACx during paired conditioning resembles previous reports on the impact of motor skill learning on spines in the motor cortex. Interestingly, the frontal associative cortex showed increased spine elimination during initial classical conditioning to sound cues but an increase in spine formation during extinction training (10), paralleling earlier studies in which the specific role of higher-order cortical areas has been demonstrated in extinction (47). This observation indicates that the observed effect on synaptic structure during ACFC appears to be specific to the cortical area. Our findings are consistent with a distributed network involving multiple brain areas for memory formation during auditory fear conditioning (12, 23, 24, 48).

Our experimental paradigm allowed us to correlate spine dynamics with the formation of a memory, but also to correlate it with memory recall triggered by a memory retention test. Our findings that structural synaptic plasticity as observed during paired conditioning does not recapitulate upon memory retrieval is consistent with a scenario in which memory retrieval does not lead to a modification of the memory trace per se (49). Alternatively, our observations could imply that reconsolidation processes do not primarily affect synaptic structures but are predominant at the molecular level (50). Furthermore, if the retrieval of previously formed memory triggers structural remodeling of synapses, this process could be restricted to different brain structures, such as the amygdala (18).

What could be the functional relevance of the induction of IEGs in the ACx associated with memory recall? Induction of IEGs has been shown previously to be correlated with synaptic plasticity, memory formation, and memory retrieval; however, the role of individual effector genes is only fragmentarily understood. It is conceivable that these genes are contributing to (i) the active modification of synapses, (ii) the stabilization of recent modifications, and (iii) homeostatic processes not directly associated with the expression of synaptic plasticity (51). Our observations are consistent with a model in which the induction of IEGs during memory recall reflect primarily processes that do not lead to structural modification of synapses but to further stabilization of a memory trace, perhaps similar to the psychophysical effect that repeated presentations lead to the formation of stronger long-term memories (52).

The direct observation of long-lasting structural changes of synapses in the ACx during fear conditioning represents an important entry point toward the analysis how memory traces are embedded in neuronal circuits in the context of a fast and well-controlled behavioral learning paradigm. It will be interesting to investigate in the future how these specific changes alter function of the cortical circuit and lead to auditory representations that are linked to behavioral experiences.

Methods

Animals.

Ten- to 14-wk-old male (Charles River) mice were used for IEG expression and lesion experiments. Three- to 6-mo-old transgenic mice of line GFP-M (Tg(Thy1-EGFP)MJrs/J) (38) were used for spine imaging experiments. All experiments were performed in accordance with the Austrian laboratory animal law guidelines for animal research and approved by the Viennese Magistratsabteilung 58 (approval no. M58/004995/2010/6).

ACFC.

We applied a slightly modified behavioral protocol used previously (22). Briefly, animals were habituated to handling and the experimental environment for at least 3 d and subsequently underwent auditory paired or unpaired conditioning. Finally, memory was tested upon exposure to conditioned sound. Freezing level was scored automatically and used as an estimate of the memory output. See SI Methods for details of behavioral experiments and their combination with qPCR and lesion experiments.

In Vivo Two-Photon Imaging of Dendritic Spines.

We applied modified protocol used previously (41) for in vivo two-photon imaging and image analysis. GFP-labeled dendrites of anesthetized mice were imaged using 950 nm with a pulsed Ti:Sapphire laser (Chameleon Ultra; Coherent) and a two-photon microscope (Ultima IV; Prairie Technologies) equipped with a 20× objective (XLUMPlan Fl, N.A. = 0.95; Olympus). Image stacks (512 × 512 pixels, pixel size ∼0.1 × 0.1 µm) were acquired every 0.5 µm along the z axis. Best projections of dendrites (41, 42) were reconstructed (Photoshop) from single frames, allowing detection and indexing of the same dendritic spines over time. See SI Methods for intrinsic imaging and quantitative image analysis.