Regulation of blood pressure by the central nucleus of the amygdala in dynamically changing appetitive and aversive classical conditioning tasks in rats

In an environment of dynamically changing conditions, humans and animals can determine whether a current situation is favorable to them and accordingly select actions. Autonomic cardiovascular tuning is as important as motor control for this function. However, neuronal mechanisms underlying the dynamic adjustments of autonomic cardiovascular responses remain unclear. In this study, we hypothesized that the amygdala plays a role in autonomic cardiovascular tuning in a dynamically changing situation. We recorded the blood pressures and heart rates of head-restrained rats during appetitive and aversive classical conditioning tasks. Rats learned varying associations between conditioned stimuli and unconditioned stimuli in appetitive, neutral, and aversive blocks. Blood pressure and heart rate in the appetitive block gradually increased in response to reward-predicting cue, preceded by a vigorously increased response to the actual reward. The predictive response was signicantly higher than the responses in the neutral and aversive condition blocks. Blood pressure and heart rate responses to the air puff-predicting cue in the aversive block were signicantly lower than that of the responses in the neutral block. The conditioned blood pressure response rapidly changed through condition switching. Furthermore, bilateral pharmacological inactivation of the central nucleus of the amygdala has signicantly decreased reward-predictive pressor responses in the latter phase, but not in the initial phase of block change. These results suggest that blood pressure is adaptively tuned by positive and negative conditioned stimuli and that the central nucleus of the amygdala likely assists in maintaining pressor response in dynamically changing situations.


Introduction
Predicting outcomes and preparing actions for future events based on current situation can often result in getting rewards ahead of competitors and ight from danger. Such prediction is critical for survival, especially in a dynamically changing environment. Surely motor control is essential for appropriate and rapid action, and autonomic cardiovascular tuning is likewise critical to supply energy to active skeletal muscles. However, the neuronal mechanisms underlying dynamic adjustments of autonomic cardiovascular responses are still unclear.
Previous studies show that the amygdala plays an essential role in emotional processing and defensive behavioral responses. The amygdala is classically believed to play a central role in a negative emotional aspect (Hilton and Zbrozyna 1963;Kapp et al. 1992;Pascoe and Kapp 1985); however, nuclei involvement in emotional, attentional, and learning processing from a positive aspect has also been reported in rodents (Calu et al. 2010;Everitt et al. 2003;Holland and Gallagher 1999;Kim et al. 2016;Roesch et al. 2012), macaques (Baxter and Murray 2002;Paton et al. 2006), and humans (Phelps and LeDoux 2005). Previous studies showed that the neuronal activities of the amygdala during appetitive and aversive conditioned stimuli are similar and correlated with autonomic arousal, such as changes in the blood pressure, suggesting an emotional arousal coding in the amygdala (Shabel and Janak 2009;Shabel et al. 2011). We previously showed that electrical and chemical stimulations of the central nucleus of the amygdala (CeA) in anesthetized rats induced bidirectional (facilitatory or inhibitory) cardiovascular responses in a region-speci c manner . In behavioral animals, we observed a sudden increase in the blood pressure immediately before exhaustion during high-intensity treadmill running with negative emotion and strong c-Fos expression in the amygdala (Kim et al. 2020).
Conditioned cardiovascular responses are classically recorded during anticipation of either appetitive or aversive outcomes (Cohen and Obrist 1975;Harris and Brady 1974). Because many of these studies have been conducted in free-moving animals, it is di cult to control animal behaviors such as approaching, freezing and eeing. Furthermore, when facing an environment where emotional context dynamically changes, it has not been demonstrated how autonomic cardiovascular response regulates and whether the amygdala is involved in the process.
Thus, we hypothesized in this study that the amygdala plays a role in autonomic cardiovascular tuning in accordance with the context of emotional arousal. In our experiments, we recorded blood pressure and heart rate from head-restrained rats to test the hypothesis during dynamically changing appetitive and aversive classical conditioning tasks and examined the effects of amygdala manipulation.

Animals
A total of 7 male Wistar-Kyoto rats (age: 7 weeks, weight: 235 ± 44 g at the time of rst surgery) were used in this study (Japan SLC, Inc., Japan). The animals were housed in a temperature-controlled room under a xed 12:12 h dark/light cycle (6:00-18:00/18:00-6:00). Animals were provided access to food and water ad libitum until full recovery from surgery. After the start of conditional training, water access was restricted to increase motivation for sucrose rewards. Animals permitted access to water ad libitum for one day a week (weekend) but could obtain water only in the behavioral task during the rest of the week. Body weights of rats were measured every day and given a few agar blocks (containing 15 mL water) in their home cages when they were less than 85% of their original weights. All experiments were approved by the Ethics Committee for Animal Experiments at Juntendo University and complied with the guidelines set by the Japan Physiological Society.

Surgery
We implanted radio transmitter to record blood pressure and then head plates to x animal's body during experiments. Recovery time after each surgery was over 1 week. During surgery, rectal temperature was monitored and maintained at 37°C using a heating pad (BWT-100, Bio Research Center, Japan). The level of anesthesia was checked by assessing limb withdrawal to noxious pinching. After surgery, antibiotics (benzylpenicillin, 1000U, i.m., Meiji Seika Pharma, Japan) and analgesics (meloxicam, 1mg/kg, s.c., Boehringer Ingelheim, Germany) were administered.
Implantation of a transmitter for telemetry A telemetric radio transmitter (HD-S10; Data Sciences International, USA) for chronic blood pressure recording from abdominal aorta was implanted, as described in previous studies (Waki et al. 2003;Yamanaka et al. 2018). Rats were anesthetized with pentobarbital sodium (50 mg/kg) by intraperitoneal (i.p.) administration and iso urane (2.0%-2.5% for maintenance) using an inhalation anesthesia apparatus (Univentor 400 anesthesia unit, Univentor, Sweden). After an abdominal midline incision was made in a rat in supine position, the intestines were moved aside to allow visualization, and the abdominal aorta was carefully isolated. After a temporal blockade using a sterilized string to prevent severe blood loss, the tip of the catheter of the transmitter was inserted into the abdominal aorta along a 21G needle guide. The transmitter catheter was then xed using a tissue adhesive (Vetbond, 3M, USA).
The transmitter was sutured to the ventral wall of the abdominal cavity.
The air puff was delivered through a stainless tube placed 8-10 cm from the rat's face. The interval of the time between CS offset to US onset (CS-US interval) was 15 s. The inter-trial interval was 60 ± 15 s. Each block consisted of 16 to 24 trials, and the order of trials (CS+ or CS− trial) was pseudo-randomly assigned. Animals could not predict the timing of block switching because none of the cues occurred with block changes. The order of blocks was xed as RW and AV were alternatively presented; NA was then deployed in between ( Fig. 1C; RW → NA → AV → NA → RW →…). RW or AV was randomly assigned to start the daily trials. As a learning process, rats were initially trained only in the RW block. The AV context was then added, with NA context training included last. Orofacial licking movements related to reward anticipation and consumption were monitored by measuring the strain on the reward spout tube ( Fig. 1A; KFG-2N-120-C1-23 ampli ed by DPM-911B, Kyowa, Japan).

Muscimol injection
Causality between conditioned blood pressure responses and activity of the CeA was assessed using pharmacological inactivation experiments. Four of the 7 animals were used in this experiment. Immediately before initiating the classical conditioning task of each experiment day, rats were microinjected with GABA A receptor agonist (muscimol, 80 pmol, 100 nL, M1523-10MG, Sigma-Aldrich, USA) into the bilateral CeA (1.8 mm caudal, 3.0 mm lateral from bregma, and 7.0 mm ventral from dura) using a glass micropipette (outside diameter of 20-30 μm; GC200F-10, Harvard Apparatus, USA). We injected saline (100 nL, Otsuka, Japan) to the control experiments for cardiovascular responses owing to volume effects of liquid injection. Micro-pipettes were then connected to a Hamilton microsyringe mounted on a syringe pump (LEGATO110, KD Scienti c, USA) to control the injection rate (500 nL/min).
After completing the nal experiments, identi cation of the chemical inactivation site was con rmed by injections of uorescent microspheres (FluoSpheres, 100 nL, Thermo Fisher Scienti c, USA) at the same coordinates stereotaxically as muscimol injection.

Histology
Rats were deeply anesthetized with sodium pentobarbital and iso urane after completion of all experiments and intracardially perfused with saline followed by 4% paraformaldehyde (163-20145, FUJIFILM Wako Pure Chemical Corporation, Japan). The brains were then removed, post-xed for at least 48 h in 4% paraformaldehyde, and replaced with 30% sucrose. Brain tissue that settled out in the sucrose solution was sliced into 50-μm-thick serial sections on a freezing microtome (REM-710; Yamato Kohki Industrial, Japan). The sections were then mounted on slides and imaged using a uorescence microscope (EVOS FL Auto 2 imaging system; Thermo Fisher, USA) to map drug injection sites in the amygdala.

Data analysis and statistics
We recorded blood pressure and heart rate during the classical conditioning tasks (Fig. 2). These parameters were simultaneously recorded using the telemetry blood pressure recording system (PhysioTel, Data Sciences International, USA) with the PowerLab system (PowerLab/8s, ADInstruments, New Zealand). Mean blood pressure and heart rate were derived from pulsatile pressure signals using LabChart software (Version 8.0, AD Instruments). These data were subsequently analyzed in MATLAB (The MathWorks, USA). Arti cial drops or increases in blood pressure and heart rate signals were removed and treated as missing values in the dataset. We mainly focused on our analysis on blood pressure and heart rate from CS onset to US onset (CS-US interval), and changes in the mean blood pressure (ΔMean blood pressure; ΔMBP), whereas heart rate (ΔHeart rate; ΔHR) were calculated by subtracting mean values during baseline period 5-15 s before the CS onset. To examine the conditioning process in which animals learned the CS-US associations on each block, the magnitudes of cardiovascular responses during 6 s immediately before the US onset were calculated and compared using two-way analysis of variance (ANOVA) among the training sessions (day) and condition blocks (RW, NA, and AV) (Fig. 3). One-way ANOVA with Tukey-Kramer post-hoc test was used in comparing magnitudes of blood pressure and heart rate during 6 s immediately before the US onset among RW, NA, and AV blocks (Fig. 4). Considering that autonomic cardiovascular response is in uenced by movement as well as emotion, we attempted to distinguish whether the observed cardiovascular responses were caused by the orofacial movement with anticipatory licking behavior and/or emotion induced by outcome prediction. We divided the trials into two groups based on whether the amplitude of licking was lower or higher than the threshold level (3 s.d. from the baseline) during the CS-US interval. The licking amplitude was converted into absolute values, and then the change from the baseline value was calculated for each trial (ΔabsLC). Averaged blood pressure, heart rate, and licking movements were plotted and compared for low-and high-licking trials using one-way ANOVA with Tukey-Kramer post-hoc test (Fig. 5). Analysis of blood pressure and heart rate with switching of the context blocks was performed using ensemble responses to each CS during ve trials before and after block switch (Fig. 6). Finally, we quanti ed and compared CS-US intervals between muscimol-and saline-injected sessions to examine the effects of inactivation of the CeA. Average data of ΔMBP in response to CS+ for the effects of bilateral inactivation of the CeA before and after 4 trials (2 trials per bin) at condition block switching during the classical conditioning task were assessed. We then analyzed the data using two-way ANOVA and Mann-Whitney U-tests with trials from condition block change and drug (muscimol vs saline) factors (Fig. 7). Statistics analysis was conducted using MATLAB Statistics and Machine Learning Toolbox (The MathWorks). The criterion for statistical signi cance was p = 0.05.

Results
Blood pressure and heart rate responses during classical conditioning tasks We developed a classical conditioning task with switching among the blocks with different cue-outcome (CS-US) associations (Fig. 1). Head-restrained rats with implanted radio transmitters (n = 7) were trained to discriminate between two tone cues during appetitive reward (RW), neutral (NA), and aversive (AV) blocks. One tone (CS+, 10 kHz) cue was associated with reward delivery (reward CS+ predicts 5% sucrose, 0.08 mL, reward US+) in the RW block, no outcome in the NA block (neutral CS+), and punishment in the AV block (aversive CS+ predicts an air puff, 30-40 psi, 1 s, aversive US+). The other tone (CS−, 4 kHz) cue was associated with no outcome in all condition blocks (Fig. 1B, reward, neutral, and aversive CS−). We recorded blood pressure and heart rate as physiological parameters and licking movement as a behavioral parameter during the classical conditioning task.
For example, licking movement, blood pressure, and heart rate traces in the 2 trials in each RW and AV blocks on the rst day (day 1) and 12 days (day 12) of training are provided in Figure 2. In RW block on day 1 ( Fig. 2A), licking movement, blood pressure, and heart rate were phasically increased only immediately after reward delivery (US+) but not after CS+, CS−, and US−. On day 12 (Fig. 2B), the blood pressure and heart rate gradually increased after reward CS+, followed by a polyphasic pressor and vigorous tachycardiac response to the reward US+ but not reward CS− and US−. Conversely, in AV block on day 1 (Fig. 2C), blood pressure and heart rate increased immediately after aversive US+ (air puff). On day 12 (Fig. 2D), the blood pressure showed gradually depressive responses to aversive CS+ and polyphasic cardiovascular responses to aversive US+.
To examine how the predictive cardiovascular responses are shaped in each condition block in a dynamically changing classical conditioning task, learning curves of the mean blood pressure (ΔMBP) and heart rate (ΔHR) responses compared with the baseline activity during the CS-US interval were plotted from the start of learning in each condition (Fig. 3). A two-way ANOVA revealed that the mean ΔMBP value showed signi cant interaction of training session (1-12 days from the start of training) and . Therefore, we conducted subsequent analyses using data after su cient training (day 7 of learning AV block) when the CS responses were stable and distinguishable.
Although Until date, we have observed predictive cardiovascular responses in condition block-dependent manner, and it is possible that the activity re ects emotional responses induced by positive and negative outcome-associated cues. However, this activity may be caused by other possible events such as anticipatory licking movement. To eliminate this possibility, we divided trials into 2 groups in accordance with the amplitude of licking movement during the CS-US interval being lower or higher than the threshold value (3 s.d. from the baseline). Even in lower-licking trials ( Fig. 5A; totally 10,796 trials), ΔMBP to reward CS+ (+4.9 ± 0.2 mmHg) and aversive CS+ (−0.4 ± 0.1 mmHg) were signi cantly higher and lower than that of the neutral CS+ (+0.2 ± 0.1 mmHg), respectively (Fig. 5A, middle-left panel, F (2,4989) = 392.1, p < 0.001, one-way ANOVA with Tukey-Kramer post-hoc test). However, conditioned heart rate responses disappeared in lower licking trials ( Fig. 5A bottom panel, F (2,4617) = 2.8, p > 0.05). Conversely, ΔMBP (F (2,1053) = 53.4, p < 0.001) and ΔHR (F (2,972) = 9.4, p < 0.001) responses were increased not only in the RW block but also in the NA and AV blocks in parallel with licking amplitude as in the higher licking trials ( Fig. 5B; totally 1,505 trials), suggesting that the licking movement also strongly affected the cardiovascular responses. Therefore, at least for the blood pressure, it is possible that the conditiondependent characteristic responses are caused by emotional cue-outcome association, rather than by licking movement.

Dynamic changes in cardiovascular responses triggered by condition block switching
We then assessed predictive cardiovascular responses during condition block switching using our dynamically changing appetitive and aversive classical conditioning task. The pseudo-color plots of averaged ΔMBP in response to CS+ and CS− in 5 trials before and after block switching (Fig. 6A, NA→RW→NA; Fig. 6B, NA→AV→NA) indicate that unanticipated association switching from neutral to reward condition blocks (NA→RW) produced ΔMBP to reward CS+ (−1.4 ± 0.3 mmHg, p = 0.42, Mann-Whitney U-test) similar to responses in the previous NA condition block (−1.1 ± 0.2 mmHg). However, once receiving reward US+, responses to reward CS+ were dramatically increased (4.4 ± 0.4 mmHg, p < 0.001, Mann-Whitney U-test). Pressor responses to reward CS+ were maintained until a few trials after switching back to the neutral block (RW→NA): 4.6 ± 0.3 mmHg in 5 trials before block switching, 3.9 ± 0.4 mmHg in the rst trial (p < 0.001, compared with ΔMBP to CS+ in NA block), 1.0 ± 0.4 mmHg in the second trial (p < 0.001), −0.2 ± 0.3 mmHg in the third trial (p = 0.005), and −0.8 ± 0.3 mmHg in the fourth trial (p = 0.52) after block switching. Immediately after the transition from neutral to aversive blocks (NA→AV), slight but signi cant depressor response to aversive CS+ (−0.7 ± 0.1 mmHg in 5 trials before block switching; −1.1 ± 0.3 mmHg in the rst trial, p = 0.06; −1.2 ± 0.2 mmHg in the second trial, p < 0.001) and prominent tri-phasic responses (increase-decrease-increase) to aversive US+ were observed (Fig. 6B). Cardiovascular responses to reward-and aversive-predictive cues rapidly changed, guided by reward and aversive expectations.
Bilateral inactivation of the CeA attenuated reward prediction-induced pressor response In this study, we observed that reward and aversion prediction induced higher and lower blood pressure responses compared with that in a neutral condition, respectively, in the classical conditioning task with dynamically changed cue-outcome associations. Previously, we reported that microstimulation of the CeA induced site-speci c, bidirectional (pressor and depressor) cardiovascular responses in anesthetized rats, suggesting involvement in autonomic tuning , through supporting the expression of defensive behaviors in emergency situations with emotional arousal. If the CeA plays a causal role in the regulation of blood pressure during emotional arousal, inactivation of CeA should abolish responses to emotional CSs.
We then examined the effects of cardiovascular responses to CSs to test this possibility via inactivation of bilateral CeA (1.8 mm caudal to bregma, 3.0 mm lateral to midline, and 7.0 mm ventral to dura). The GABA A receptor agonist muscimol (80 pmol, 100 nL) was microinjected while the rats performed the conditioning task. In total, we performed 23 injections (muscimol, 12 injections; saline, 11 injections) in 4 of the 7 rats. The injection site was con rmed using uorescent microspheres (FluoSpheres, 100 nL) based on the histology ( Fig. 7; right panel). All animals were injected with both muscimol and saline in different sessions. Inactivation of bilateral CeA using muscimol compared with saline injection caused signi cant decreases in ΔMBP before and after switching from the RW to NA blocks (Fig. 7, F (15,1287) = 2.64, p < 0.006, signi cant interaction, two-way ANOVA). Notably, the decrease in ΔMBP was observed during the second half of the RW block (p < 0.001) but not immediately after NA→RW switching (p > 0.05, Mann-Whitney U-test). CeA likely plays a more important role in maintaining rather than in acquiring emotional responses to salient events.

Discussion
In this study, we demonstrated that reward-predictive and aversion-predictive cues induced pressor and depressor blood pressure responses, respectively. The characteristic autonomic responses were acquired by learning across sessions and were observed even with lower-licking trials. Switching condition block provokes rapid regulation of the blood pressure in accordance with its emotional valence. Furthermore, blockade of the activity of the CeA impaired pressor responses during switching from reward to neutral condition blocks. We demonstrate that cardiovascular responses are adaptively and rapidly tuned by positive and negative emotional stimuli, and CeA may contribute to the maintenance of adaptive regulations.
We con rmed the characteristic cardiovascular responses in our classical conditioning tasks. Blood pressure exhibited a phasic and gradual increase in response to reward-predictive cues and a phasic decrease to neutral-and aversive-predictive cues. The heart rate increased in response to cues under all conditions compared with the baseline values, but the level of tachycardia varied among the conditions: reward CS+ induced predominant tachycardia; however, aversive CS+ evoked a relatively small increase in heart rate compared with neutral CS+. These condition-dependent characteristic cardiovascular responses were progressively acquired through several sessions and consistently expressed after su cient learning. The blood pressure responses robustly discriminated conditions, but the increase in heart rate disappeared in lower-licking trials. Both the blood pressure and heart rate were increased in parallel with the amplitude of licking movement. These results suggest that both blood pressure and heart rate are affected by orofacial licking movement; however, at least the blood pressure may be regulated by additional factors, except orofacial movement per se, such as emotion. Previous studies have reported cardiovascular responses during classical conditioning task using free-moving and restrained animals in several species, such as pigeon (Cohen and Durkovic 1966), rabbit (Pascoe and Kapp 1985;Powell et al. 1997), dog (Anderson and Brady 1972;Dykman and Gantt 1956;Obrist and Webb 1967), rodent (Holdstock and Schwartzbaum 1965;Iwata and LeDoux 1988;Shabel and Janak 2009;Shabel et al. 2011), marmoset (Braesicke et al. 2005;Mikheenko et al. 2010), macaque (Randall et al. 1975), and humans (Hastings and Obrist 1967;Wexler and De Leon 1979). Appetitive conditioning using food or water as a reward US, induces pressor and tachycardia during the CS-US interval, whereas, in aversive conditioning, using electrical shock as punishment US often causes pressor or tachycardia, but some bradycardic responses are reported (Cohen and Obrist 1975;Harris and Brady 1974;Randall et al. 1975). In this study, small but signi cant lower blood pressure responses to air puff-predicting CS+ were also observed. Air puff is de nitely acting as the US based on the remarkable tri-phasic cardiovascular responses. Considering a pressor response temporally in the early stage of learning ( Fig. 3A; training session "Day 3"), one possible reason that air puff-predicting CS+ caused suppression but not facilitation of cardiovascular responses includes overtraining of the classical conditioning procedures, where the animals cannot avoid punishment (e.g., learned helplessness). The reasons for the lower response level of aversive CS+ compared with that of reward CS+ is that the air puff might be a mild stimulus compared with electrical shock and/or oor effect. In the present study, we used a neutral block in which CS+ was associated with no-outcome as a different measure from the reward block. As midbrain dopamine neurons encoding reward-prediction error signal have been reported to show a decrease in their activities in response to no-reward conditions (Matsumoto and Hikosaka 2009), no-outcome conditions may have functioned as an aversive rather than a neutral measure.
Interestingly, reward CS− induced relatively higher blood pressure and lower heart rate responses, although CS− was not associated with either reward or punishment in any condition blocks. Animals could get rewarded in our task after only reward CS+. The RW block with the highest value CS+ represented higher state value, whereas CS− in the block induced stronger negative emotion compared with other condition blocks. This possibility suggests that the value of an option (CS−) is assigned relatively, rather than absolutely, and is often in uenced by the value of the paired option (CS+) (Elliott et al. 2008).
The roles of the amygdala in classical conditioning have already been examined in previous studies, particularly through fear conditioning in free-moving animals. The CeA projects into various regions such as the lateral hypothalamus, the periaqueductal grey and brainstem to regulate autonomic sympathetic and parasympathetic activities and endocrine responses to promote defensive behaviors (such as attack, escape, and freeze) (Davis 1992). Lesions in the bilateral amygdala attenuated conditioned acceleration of heart rate in both appetitive (Braesicke et al. 2005) and aversive (Iwata et al. 1986;LeDoux et al. 1990;Mikheenko et al. 2010) conditioning using free-moving animals. These ndings suggest that the amygdala may be involved in the autonomic arousal in emotional processing. Several past studies have examined either appetitive or aversive condition blocks; however, little is known concerning autonomic tuning in different conditions that changes dynamically between appetitive and aversive. Our results are also consistent with previous studies in which the amygdala (CeA) plays an important role in accelerating cardiovascular tuning. Notably, the attenuation of pressor responses by amygdala inactivation did not appear immediately after switching to the appetitive (RW) condition block but appeared when switching from the appetitive to neutral (RW→NA) block, suggesting a more crucial role in the maintenance, than acquisition, of emotional autonomic tuning.
Lack of signi cant changes in the cardiovascular responses to aversive-predicting cue by amygdala inactivation is a limitation of the present study. Conditioned responses to fear-predictive stimuli are taskdependent (Harris and Brady 1974), and decelerated responses observed in our study were not su ciently strong. In addition, based on previous electrophysiological studies, several amygdala neurons encode emotional arousal that respond to both appetitive-and aversive-predicting CS (Shabel and Janak 2009;Shabel et al. 2011). As results in the present study, we reported that CeA microstimulation induces sitespeci c bidirectional (pressor and depressor) cardiovascular responses . Blockade of the amygdala may diminish autonomic responses if accelerated or more decelerated responses are evoked by aversive CS.
Our ndings demonstrated cardiovascular tuning and the involvement of the CeA in dynamically changing the CS-US associations. The characteristic blood pressure responses rapidly adapt in a few trials immediately after condition block switching. Such rapid switching is reminiscent of the activity of neurons of the amygdala (Belova et al. 2007), striatum (Lauwereyns et al. 2002), and midbrain dopamine neurons (Roesch et al. 2007). Blockade of dopamine D1 receptor in the striatum impaired contextdependent behavioral responses (Nakamura and Hikosaka 2006;Ueda et al. 2017). Arti cial suppression of the midbrain dopaminergic neurons during fear-to-safety context switching induced a delay in the extinction of fear conditioning behavior (Luo et al. 2018). In addition, the activity of dopamine neurons has also been reported to contribute to the prediction of reward as well as aversion stimuli (Matsumoto and Hikosaka 2009;Moriya et al. 2018). Anatomically, the amygdala also receives dopaminergic inputs (Garris and Wightman 1995). There is a direct projection from the amygdala (basolateral region) to the striatum, whereas projection from the striatum to the amygdala is believed to be polysynaptic. Functionally, activities of the striatal neurons increased in response to reward-predicting cue, but their activities were attenuated by the blockade of the amygdala (Ambroggi et al. 2008). Based on these evidences, condition-dependent cardiovascular responses observed in the present study may re ect the differences in the outcome expectation with the functional interactions between the amygdala and striatum via dopaminergic inputs (Averbeck and Costa 2017). On the other hand, the neuronal circuits of aversion-predictive depressive responses are still unclear. Based on behavioral and autonomic aspects and functional connections to the amygdala, possible candidates are the midbrain periaqueductal gray which is involved in negative emotional processing and freezing behavior induced by fearful stimuli (Amorapanth et al. 1999;Buhle et al. 2013), and the claustrum which shows depressive responses to arti cial stimulation and functionally connected with the amygdala (Kim et al. 2020). Thus, future studies should further examine the functions of neuronal circuits that may support and maintain the emotional expression suitable for a dynamically changing environment.