Amygdalar and Hippocampal Theta Rhythm Synchronization During Fear Memory Retrieval

See allHide authors and affiliations

Science  08 Aug 2003:
Vol. 301, Issue 5634, pp. 846-850
DOI: 10.1126/science.1085818


The amygdalohippocampal circuit plays a pivotal role in Pavlovian fear memory. We simultaneously recorded electrical activity in the lateral amygdala (LA) and the CA1 area of the hippocampus in freely behaving fear-conditioned mice. Patterns of activity were related to fear behavior evoked by conditioned and indifferent sensory stimuli and contexts. Rhythmically synchronized activity at theta frequencies increased between the LA and the CA1 after fear conditioning and became significant during confrontation with conditioned fear stimuli and expression of freezing behavior. Synchronization of theta activities in the amygdalohippocampal network represents a neuronal correlate of conditioned fear, apt to improve neuronal communication during memory retrieval.

Considerable progress has been made in our understanding of the synaptic circuits and plasticity that underlie emotional learning, specifically during Pavlovian fear conditioning, and the involvement of the amygdala therein (1, 2). Evidence suggests an alteration of neuronal responsiveness to fear-conditioned stimuli in the amygdala, sometimes paralleled by behavioral changes (3). During emotional arousal and various types of rhythmic activities during sleep, neurons in the amygdala produce theta activity (4, 5). Although these activities facilitate synaptic plasticity and memory in extended neuronal networks, their relevance for the expression of fear or fear memory remains unclear. Of the extensive afferent and efferent connections of the amygdala, interactions with the hippocampus are particularly important for memory formation (6, 7). Amygdala lesions attenuate hippocampal synaptic plasticity and block the memory-enhancing effects of direct hippocampal stimulation (8, 9). Further, behavioral stress as well as stimulation of the amygdala interferes with synaptic plasticity in the hippocampal formation (1012). This interaction appears to be bi-directional, given that tetanic stimulation of hippocampal efferent fibers can induce long-term potentiation in the LA (13). The present study was designed to characterize patterns of neural activity in amygdalohippocampal pathways related to the retrieval and expression of conditioned fear. The rationale was to use Pavlovian fear conditioning as a simple, well-established model of emotional learning (1) and to focus on interactions between the LA, which is the major input station of sensory signals to the amygdala, and the CA1 area of the hippocampus, with which the LA is mutually and prominently interconnected (14).

Mice were fear conditioned through the explicitly paired presentation of conditioned (CS+) and unconditioned (US) stimuli, and their responses during fear-memory retrieval were compared with those of control animals undergoing explicitly unpaired training (15). Conditioned freezing behavior was monitored in the retrieval session to assess fear memory and the emotional relevance of the CS+ and an indifferent control stimulus (CS). In addition, risk-assessment behavior (overt orienting and stretched attending) was examined as a control measure of species-specific defensive behavior with minimum locomotor activity. The results show a pronounced and selective fear response in conditioned mice, and a moderately aversive or ambiguous response of controls to the CS+. The behavioral response to the CS was not different between groups (fig. S1).

At the same time as the behavioral assessment, we determined electrophysiological activity by recording field potentials in both the LA and the CA1 of the dorsal hippocampus (15). In control animals, activity in the CA1 was distributed around the theta frequency, covering a relatively wide frequency range with only short periods of rhythmic patterns. No prominent pattern of activity was observed in the LA (Fig. 1, A to C), and cross-correlation analyses did not show any significant synchronization of activity between the two brain areas (Fig. 1, D and H). Activity in the CA1 or the LA did not differ during stimulus and prestimulus periods, nor could a difference be observed between CS+ and CS periods. In fear-conditioned animals, theta activity prevailed in the CA1 under all stimulus conditions (Fig. 2). Activity in the LA before and during CS presentation resembled the activity in control animals, in that there was no indication of a predominant pattern or frequency (Fig. 2, A to C). Upon presentation of the CS+, activity in the LA shifted into a highly rhythmic pattern centered at the theta frequency band (Fig. 2, F and G). Cross-correlation analyses revealed a progressive increase in synchronized activity at a frequency of 4 to 8 Hz during the CS+ (Fig. 2H and fig. S2). By averaging the cross-correlograms from four consecutive CS+ presentations and taking the second positive peak as a quantitative measure, a significant (P < 0.004, t = 3.531; Student's t test) increase in theta synchronization could be demonstrated in fear-conditioned animals (mean ± SEM: 0.125 ± 0.012; n = 8), as compared with control animals (0.048 ± 0.020; n = 6). A partial, insignificant increase was observed in fear-conditioned mice during the CS (0.085 ± 0.029 as compared with 0.044 ± 0.021 in controls) (16).

Fig. 1.

Neural activity in the CA 1 and the LA of a control animal during the presentation of CS [(A) to (D)] and CS+ [(E) to (H)]. (A and E) Original traces of field-potential recordings in the CA 1 (upper traces) and the LA (bottom traces) before and during CS or CS+ presentation (bars above the traces). (B and F) Color-coded power spectra of the traces in (A) and (E) demonstrate CA 1 theta activity in a frequency band of 4to 12 Hz during the entire stimulus (white bar) and prestimulus phase. LA activity lacks such a prominent pattern [the time scales in (B) and (F) differ from those in (A) and (E)]. Behavior (r, risk assessment; x, exploration) is indicated near the bottom of the diagrams. (C and G) Autocorrelation analyses indicate only short periods of rhythmic activities in the CA 1 and the lack of rhythms in the LA. The correlation coefficient is indicated in the upper right corner. (D and H) Cross-correlation analyses of activities in the LA and the CA 1 during stimulus presentation reveal a low level of synchronization. Four successive 3-s intervals are shown [as indicated by numbers 1 to 4 in (B) and (F)], starting 1 s before presentation of the stimuli.

Fig. 2.

Neural activity in the CA 1 and LA of a fear-conditioned animal during presentation of CS [(A) to (D)] and CS+ [(E) to (H)]. (A and E) Original traces of field-potential recordings in the CA 1 (upper traces) and the LA (bottom traces) before and during CS or CS+ presentation (bars above the traces). (B and F) Color-coded power spectra of the traces in (A) and (E) [which have time scales that differ from those in (B) and (F)]. Similar to control animals, broad-range theta activity is seen in the CA 1 but not in the LA during the CS. However, CS+ presentation is associated with highly rhythmic theta activity at around 5 Hz in both brain areas and the expression of freezing behavior (f) (s, stereotypic behavior; x, exploration; r, risk assessment). (C and G) Autocorrelation analyses indicate short epochs of rhythmic activities in the CA 1 alone during exploratory behavior and an increased rhythmic activity in the LA during freezing. (D and H) Cross-correlation analyses reveal a progressive increase of correlated theta activity in the two brain areas during presentation of the CS+ but not the CS. Four successive 3-s intervals are shown [(as indicated by numbers 1 to 4 in (B) and (F)], starting 1 s before presentation of the stimuli.

Atropine-sensitive type 2 theta activity (4 to 8 Hz) has been shown to occur in the hippocampal formation during periods of immobility, whereas atropine-resistant type 1 theta activity (8 to 14 Hz) is observed during exploration (17, 18). Type 2 theta can be elicited by strongly arousing stimuli, such as confrontation with predators or noxious stimuli (19). We therefore determined electrophysiological activity during periods of defined defensive behavior, such as freezing and risk assessment (15). Indeed, freezing of fear-conditioned mice during the CS+ was associated with a significantly stronger theta synchronization (0.112 ± 0.019 at 5.0 ± 1.5 Hz, P < 0.05; n = 8) than the risk-assessment response of controls (0.050 ± 0.030 at 4.4 ± 0.6 Hz; n = 6) (Fig. 3A). An association of theta synchronization and freezing behavior was also evident in three control animals, which displayed freezing intervals of sufficient length for a cross-correlation analysis (0.141 ± 0.034). Both training groups showed low levels of synchronized theta activity when displaying risk-assessment behavior during the CS (0.068 ± 0.016 in fear-conditioned animals; 0.055 ± 0.009 in controls) and strong hippocampal theta activity without synchronization to the LA during exploration (Fig. 3B).

Fig. 3.

(A) Averaged cross-correlograms of CA1/LA activity in the population of fear-conditioned animals (n = 8; a and c) and control animals (n = 6; b and d), during presentation of the first CS (a and b) and the first CS+ (c and d). Synchronization increases as a result of fear conditioning and becomes significant during the presentation of the CS+. The increase is particularly evident when comparing the predominant behavioral responses: freezing in fear-conditioned animals (e; n = 5) and risk assessment in control animals (f; n = 5). Asterisks indicate significant differences in the amplitude of the second peak (representing theta activity at about 5 Hz) between the fear-conditioned and control group. (B) Power spectra (a) and cross-correlogram (b) of extracellular field recordings in the CA 1 and the LA during exploratory behavior. Profound theta activity is apparent in the CA 1 but not the LA, and there is a lack of theta synchrony between the two brain areas.

Together, our observations indicate that elicitation of conditioned freezing behavior is associated with type 2 theta activity and theta synchronization in amygdalohippocampal pathways (20). Theta synchronization may thus be functionally related to the retrieval and/or expression of conditioned fear.

Both the hippocampus and the amygdala are known to participate in the formation of fear memories (1, 6), and their relative contribution appears to depend on both the conditioning paradigm and training intensity used (21). Conditional freezing involves both the LA and the dorsal hippocampus, and although the latter may be more concerned with configural processing and contextual tasks, a lesion of the area still reduces cued conditioned freezing performance (22). We thus extended our analysis to amygdalohippocampal synchronization during contextually induced freezing behavior (Fig. 4). Again, we observed significant freezing behavior (47.0 ± 7.3% of recording time; P < 0.0001, t = 14.24, Student's t test, n = 5; compared to the neutral context) that was associated with pronounced theta activity in the LA. As during cued retrieval, these theta oscillations in the LA were significantly synchronized with hippocampal rhythms (0.207 ± 0.051 at 4.3 ± 0.8 Hz; P < 0.05, t = 2.627, Student's t test, n = 5). No comparable rhythmicity and synchronicity were observed during risk assessment in the shock context or the neutral context. Although it has been suggested that foreground context (fig. S3) and background context conditioning are differently dependent on the dorsal hippocampus (23), the two conditioning methods were equally efficient in evoking synchronized rhythmic activity. Hence, communication along amygdalohippocampal pathways may be involved in the development and expression of fear-related emotions under different training conditions. Indeed, the temporally structured relay of signals between the amygdala and hippocampus during theta synchronization may allow a parallel processing of unitary and configural stimulus information related to cued and contextual fear memories.

Fig. 4.

Neural activity in the CA 1 and the LA after background context conditioning. (A) Original traces of field-potential recordings in the CA 1 (upper trace) and the LA (bottom trace). (B) Color-coded power spectrum of the traces shown in (A) [(A) and (B) have different time scales]. Theta activity at 7 to 10 Hz prevails in the CA 1 during risk-assessment behavior (r, as indicated near the bottom of the diagrams), and theta activity at 4 to 5 Hz appears in the LA and the CA 1 during freezing (f). (C) Autocorrelation analysis reveals epochs of rhythmic activities during both risk-assessment and freezing behavior in the CA 1 and a high level of rhythmic activity in the LA during freezing. Numbers in the upper right corner indicate correlation coefficients. (D) Conditioned mice displayed significant freezing (Freez) and risk-assessment (RA) behavior during reexposure to the training context (mean + SEM) but no freezing behavior in the neutral context. Data of foreground (n = 2; fig. S3) and background contextual conditioning (n = 3) were similar and were therefore pooled. Lower conditional freezing compared with cued retrieval sessions is probably due to contextual preexposure in all groups. Asterisk indicates significant differences between contexts. (E to G) Averaged cross-correlograms of electrical neural activity in the CA 1 and the LA (pooled from foreground and background conditioning experiments). Asterisk in (G) indicates significant difference in the amplitude of the second peak (representing theta activity at about 4Hz) between freezing periods in the shock context and periods of risk-assessment behavior in either the neutral or the shock context.

The LA/CA1 network system seems to be well suited to rhythmically oscillate at theta frequencies: The basolateral amygdaloid complex receives synaptic inputs from the hippocampus (24), where theta waves have been observed (17, 18), and from the anterior thalamic nuclei (25), which could transfer hippocampal theta rhythms to the amygdala. The intrinsic oscillatory properties of LA projection neurons (26, 27), in turn, may provide adequate recurring time windows for the facilitated integration of synaptic inputs at theta frequencies (supporting online material text). Consistent with this notion is the observation that cellular theta activities in the perirhinal cortex and amygdala can be phase locked to entorhinal theta waves (28, 29), and thus most likely also to hippocampal theta waves. Given evidence that theta waves or thetafrequency stimulation facilitates synaptic plasticity, such as long-term potentiation (and depotentiation) in the hippocampus (30) or long-term depression in the LA (31), the increase in coherent theta activities in amygdalohippocampal circuits may represent an increase in neuronal communication apt to promote or stabilize synaptic plasticity in these areas in relation to the retention of fear memory.

Supporting Online Material

Materials and Methods

Figs. S1 to S3


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article