Male Sprague-Dawley rats (Orient Bio, Kyunggi-do, Republic of Korea), initially weighing 250–270 g, were used. All animals were individually housed in a climate-controlled vivarium with a reverse 12-h light/dark cycle (lights on at 9:00 PM). Experiments were conducted during the dark phase of the cycle and strictly followed the guidelines for the “Care and Use of Laboratory Rats” from Korea University, Seoul, Republic of Korea.

Rats were anesthetized with pentobarbital sodium (50 mg/kg, i.p.) and mounted on a stereotaxic frame. Two holes were drilled into the exposed cranium, and custom-made electrodes (stainless steel insect pins insulated with epoxy, except for 0.5–1 mm at the tip) were inserted through these holes to reach the target coordinates in the CeA. Bilateral lesions in the CeA were made by passing anodal currents, with coordinates and lesion parameters detailed in Supplementary Table 1 (Paxinos and Watson, 2006). Rats in the sham lesion group (SHAM) underwent the same surgical procedure except that the electrodes were inserted 1–2 mm dorsal to the target coordinates, and no current was passed. All rats were given a recovery period of 7–14 days after the surgery was completed and placed on a standard food deprivation schedule to maintain 80–90% of their normal body weights.

All rats underwent an approach-avoidance conflict task that had been previously developed (Kimm and Choi, 2018). The task involved a robot named Lobsterbot (Figure 1A) designed to mimic the prey-capturing motion of a predator, obstructing the rat’s access to a food pellet (1.5–2 g) by snapping its claws (Figure 1B).

The experimental protocol consisted of habituation and Lobsterbot sessions (Figure 1C). First, the rats were habituated to the experimental arena for 2–3 days without the Lobsterbot present. During this time, the rats were allowed to freely explore the arena for 5 min, and a single food pellet was available at the opposite end of the arena. The habituation stage for each rat ended when they consumed food pellets in 2 consecutive sessions.

After completing the habituation stage, the Lobsterbot sessions (5 min/day) began and lasted for 5 days. During these sessions, the Lobsterbot was positioned behind the food pellet and snapped its claws when it detected the rat’s presence. A breach in the infrared photobeam was considered an episode of encounter, and the Lobsterbot’s attack ended as soon as the rat escaped and the photobeam was no longer broken.

After the completion of the experiments, rats were overdosed with sodium pentobarbital and perfused transcardially with saline and 10% buffered formalin. The brains were stored in post-fix solution (30% sucrose in 10% buffered formalin) and sectioned on a microtome at 50 μm thickness. Sections were mounted on gelatin-coated slides, then stained with Cresyl violet and Prussian blue dyes. The brain lesion sites were histologically reconstructed for the analysis (Figures 1D, E).

The movements of the rats were tracked using ANY-maze video tracking software (Stoelting Co., USA). The software tracked the green dot, which represented the rat’s head, and the orange dot, which represented the center of the rat’s body. These movements were then used to calculate the angle and length of the stretched posture during the rat’s encounter with the Lobsterbot (as shown in Figures 2A, B). The frequency of encounters was monitored using the infrared photobeam sensors of the Lobsterbot (as shown in Figure 3A). Additionally, ANY-maze also monitored the position and movement of the rats to measure the time they spent hiding in the Gate zone (an area around the gate, as shown in Figures 3C, D).

The statistical analysis was conducted using IBM SPSS version 20.0. A two-way ANOVA with repeated measures was applied to examine behaviors (diagonal approach, stretched posture, encounter frequency, and duration in the Gate zone) for each session. Bonferroni’s post-hoc tests were conducted to compare both between-group and within-group differences in all the aforementioned behaviors. An independent t-test was utilized to evaluate the between-group difference in the mean percentage of diagonal approach and stretched posture across five sessions. Pearson’s correlation coefficient was employed to analyze the relationship between behavioral indices. Results are presented as the mean ± standard error of the mean (SEM).

SHAM (n = 11) and CeA-lesioned (n = 15) rats were tested in an approach-avoidance conflict task using the Lobsterbot. During the task, rats foraged for food pellets while a predator-like robot, Lobsterbot, obstructed access to the pellets by snapping its claws (Figure 1B and Supplementary Video 1). The rat’s head and body center were tracked to determine the approach angle and stretched posture (Figures 2A, B).

The results showed that most rats in the SHAM group favored a diagonal approach angle (θ < 60°, 120° θ < 180°) over a straight approach angle (60° < θ < 120°) when approaching the Lobsterbot. The mean percentage of diagonal approaches was significantly affected by the lesion [Two-way repeated measures ANOVA; F_(1,23) = 8.336, p < 0.01; Figure 2C-left], but not by session [F_{(3.206,73.73)} = 1.671, p > 0.1] or lesion × session interaction [F_(4,92) = 0.6426, p > 0.6]. CeA-lesioned rats showed a significantly lower percentages of diagonal approaches compared to the SHAM rats in the third session (Bonferroni’s post hoc, p < 0.05; Figure 2C-left). On average, the CeA-lesioned rats displayed a significantly lower percentage of diagonal approaches over the course of 5 sessions compared to the SHAM rats [Independent t-test; t₍₂₄₎ = 2.656, p < 0.05; Figure 2C-right].

The stretched posture analysis revealed significant main effects of lesion [Two-way repeated measures ANOVA; F_(1,23) = 32.87, p < 0.0001; Figure 2D-left] and session [F_{(3.184,73.23)} = 5.282, p < 0.01], but not of lesion × session interaction [F_(4,92) = 0.6426, p > 0.6]. CeA-lesioned rats exhibited significantly shorter stretched postures compared to the SHAM rats across all sessions (Bonferroni’s post hoc, p-values < 0.05; Figure 2D-left) as well as in the 5-session average [Independent t-test; t₍₂₄₎ = 6.117, p < 0.05; Figure 2D-right].

To sum, both diagonal approach and stretched posture were reduced by CeA lesions, indicating that CeA is critical for the expression of CLBs.

In addition to the reduced CLBs, CeA lesions also altered the approach and avoidance. The frequency of encounters, used as an index of approach, was significantly increased by the lesion (Figures 3A, B). A Two-way repeated measures ANOVA revealed significant effects of lesion [F_(1,24) = 50.30, p < 0.0001], session [F_{(2.631,63.15)} = 30.11, p < 0.0001], and interaction [F_(4,96) = 18.73, p < 0.0001] (Figure 3B). Post hoc analysis further indicated that the frequency of encounters in the CeA-lesioned rats was significantly higher than in the SHAM-lesioned rats across all sessions (Bonferroni’s post hoc, p-values < 0.05; Figure 3B).

Avoidance was assessed by the time spent hiding in the Gate zone (a virtual 12 cm × 12 cm zone depicted in Figure 3C). The increased hiding time was observed when the rats avoided the Lobsterbot and preferred a location behind the walls in the Gate zone, as reported in a previous study (Kimm and Choi, 2018). This behavior is best characterized as ‘inhibitory avoidance’, which involves suppressing approach behaviors in the presence of the Lobsterbot. It represents a more passive reaction to threat, differing from active forms of avoidance or escape. CeA lesions significantly decreased the time spent hiding (Figures 3D, E). A Two-way repeated measures ANOVA found a significant effect of the lesion [F_(1,24) = 13.12, p < 0.0014] (Figure 3E). Post-hoc analysis also showed that the hiding time was lower in CeA-lesioned rats in sessions 1 and 5 (Bonferroni’s post hoc, p-values < 0.05; Figure 3E). Taken together, these results suggest that the CeA is critical for approach suppression and threat avoidance. Likewise, the pattern of time spent in the Nest area mirrored that of the Gate zone, even though the duration in the Nest area was approximately twice as long as in the Gate zone for both groups (Supplementary Figure 1).

Additionally, head-withdrawal, an active escape behavior from the Lobsterbot was measured (Supplementary Figure 2). The head-withdrawal latency measured at milliseconds scale was significantly slowed by the CeA lesion across all sessions, indicating CeA’s role in reflexive defensive behavior as well.

The significant impact of CeA lesions on CLBs and approach/avoidance behaviors raises the question of whether the decreased conflict behaviors are simply byproducts of reduced avoidance. To address this question, a correlational analysis was conducted among all behaviors (Figure 4). In the SHAM-lesioned rats, the diagonal approach showed a significant and positive correlation with the stretched posture (r = 0.27, p < 0.0001). These two conflict behaviors were not significantly correlated with any other behavioral indices of approach or avoidance, such as encounter frequency or hiding (p-values > 0.05). In the CeA-lesioned rats, the correlation between the two conflict behaviors remained robust (r = 0.58, p < 0.0001), without any significant correlation to other approach and avoidance behaviors (p-values > 0.05). These results suggest that the two conflict behaviors are not directly tied to approach suppression or avoidance behaviors and that CeA lesions have a direct impact on CLBs rather than an indirect influence through those behaviors. Interestingly, the correlation between the frequency of encounter (approach) and time spent hiding (avoidance), which was not significant in SHAM, became significantly negative in the CeA lesion (r = −0.54, p < 0.0001), suggesting that the inverse relationship between approach and avoidance becomes prominent when the CeA no longer dominantly regulates those behaviors.