Specific behaviors during auditory fear conditioning and postsynaptic expression of AMPA receptors in the basolateral amygdala predict interindividual differences in fear generalization in male rats

Abstract

Auditory fear conditioning in rats is a widely used method to study learning, memory, and emotional responding. Despite procedural standardizations and optimizations, there is substantial interindividual variability in fear expression during test, notably in terms of fear expressed toward the testing context alone. To better understand which factors could explain this variation between subjects, we here explored whether behavior during training and expression of AMPA receptors (AMPARs) after long-term memory formation in the amygdala could predict freezing during test. We studied outbred male rats and found strong variation in fear generalization to a different context. Hierarchical clustering of these data identified two distinct groups of subjects that independently correlated with a specific pattern of behaviors expressed during initial training (i.e., rearing and freezing). The extent of fear generalization correlated positively with postsynaptic expression of GluA1-containing AMPA receptors in the basolateral nucleus of the amygdala. Our data thus identify candidate behavioral and molecular predictors of fear generalization that may inform our understanding of some anxiety-related disorders, such as posttraumatic stress disorder (PTSD), that are characterized by overgeneralized fear.

Figure 1.

Wide range of fear responses to the testing context following fear conditioning. (A) Schematic representation of the experiment. Animals were habituated to context A for 3 d (5 min/day), received two tone–shock pairings in context B, and were tested by presenting the tone without shock in context A 24 h later. (B) Average individual freezing before (pre-CS) and during (CS) tone presentation during the testing session. Wilcoxon matched-pairs signed rank test: pre-CS versus CS, W = 262. (***) P < 0.001. (C) During the 2 min preceding presentation of the tone (pre-CS phase) in context A, there was an overall unexpected widening of fear responses to the context. Horizontal dashed lines indicate cutoff at 10%. (D) Freezing (in percentage) to context B before the first tone–shock pairing (Pre phase) at the training session was minimal. Same cutoff as in C is indicated.

Figure 2.

Behavioral profiles during fear memory conditioning. (A) We split the auditory fear conditioning session into five phases (“Pre”: before conditioning, “TS1”; first tone–shock presentation, “ITI”: intertrial interval, “TS2”: second tone–shock presentation, and “Post”: after conditioning) with different durations (in seconds) and measured for each phase both absolute bout frequency and time spent freezing and rearing, as well as time spent grooming. (B–H) Behavior profiles included time spent grooming (B), freezing (C), and rearing (D); absolute bout frequency of freezing (E) and rearing (F); and average bout duration of freezing (G) and rearing (H). Freezing, rearing, and grooming were recalculated as phase percentage for phase comparison (gray traces in B–D). For the latter we ran a Friedman's test with Dunn–Bonferroni multiple comparisons test as post-hoc analysis. (*) P < 0.05, (**) P < 0.01, (***) P < 0.001. All data points represent mean ± 1 SEM.

Figure 3.

Training parameters poorly predict pre-CS freezing during test. Only four out of 120 Pearson correlations were statistically significant. Training ITI freezing (A), TS2 freezing (B), and Post rearing (D) average bout duration (in seconds per bout), as well as Pre rearing duration (in seconds) (C) linearly correlated with testing absolute bout frequency of freezing parameter during pre-CS phase. Squared Pearson coefficient (r²) and P-value indicate that only between 17% and 23% of the pre-CS freezing bouts during testing can be explained by any one of these four training parameters.

Figure 4.

Hierarchical cluster analysis produces two major clusters of animals. We analyzed dependent variables freezing, rearing, and grooming duration (in seconds) for all training phases (Pre, TS1, ITI, TS2, and Post) and freezing duration for both testing phases (pre-CS and CS) for each rat. Here, we show a scree plot (A) taken from the agglomeration schedule (Supplemental Table S5) and a clustering dendrogram (B) indicating at least two meaningful clusters. (C) Freezing and rearing during the training ITI are the best predictors for pre-CS freezing during the test and thus are displayed labeled by cluster as a 3D scatter plot with cluster centroid spikes. (D) Providing more detail, these data also are presented as a 2D matrix plot.

Figure 5.

High freezing and low rearing after first shock presentation predict fear generalization during test. Displayed are behavioral profiles of animals in clusters 1 and 2 for time (in seconds) spent freezing (A), rearing (B), and grooming (E), as well as average bout duration (in seconds per bout) of freezing (C) and absolute bout frequency of rearing (D) during training. We also show freezing duration (F), absolute bout frequency (G), and average bout duration (H) during test. We performed two-sample hypothesis testing using Student's t-test on parameters with normal distribution and Mann–Whitney U-test on nonnormal data (see Supplemental Table S15). Data points represent mean ± 1 SEM. Student's t-test: (**) P < 0.01, (***) P < 0.001; Mann–Whitney U-test: (#) P < 0.05, (###) P < 0.001.

Figure 6.

Amygdala punches for BLA tissue collection. In order to analyze the expression of AMPARs at amygdala synapses, we collected amygdala tissue from selected animals with a 1-mm hollow needle directed at the basolateral amygdala nuclei (BLA). Here we show a representative microscopy image of a rat brain slice after punching, illustrating that most of the tissue was collected from the BLA and lateral amygdala (LA) while sparing other nuclei, including the central (CeA) and the basomedial (BMA) nuclei.

Figure 7.

Synaptic GluA1 expression at the BLA correlates with fear generalization during testing. Differential centrifugation of BLA tissue taken from selected animals displaying varying degrees of testing pre-CS freezing yielded synaptic and extrasynaptic fractions containing PSD proteins. Quantification of these fractions via optical densitometry revealed dissimilar expression of synaptic (sGluA1 and sGluA2) (A,C) and extrasynaptic (eGluA1 and eGluA2) (B,D) AMPARs, with GluA1 showing the highest degree of linear correlation with pre-CS freezing duration (in seconds) during test. Squared Pearson coefficient (r²) and respective P-values are indicated as scatter plot insets. (E) Blots of all proteins analyzed on each fraction are shown, with pre-CS freezing levels displayed as phase percentage above the blots.