Distinct Processing of Aversive Experience in Amygdala Subregions

Abstract

Background: The amygdala is an anatomically complex medial temporal brain structure whose subregions are considered to serve distinct functions. However, their precise role in mediating human aversive experience remains ill understood.

Methods: We used functional magnetic resonance imaging in 39 healthy volunteers with varying levels of trait anxiety to assess distinct contributions of the basolateral amygdala (BLA) and centromedial amygdala to anticipation and experience of aversive events. Additionally, we examined the relationship between any identified functional subspecialization and measures of subjective reported aversion and trait anxiety.

Results: Our results show that the centromedial amygdala is responsive to aversive outcomes but insensitive to predictive aversive cues. In contrast, the BLA encodes an aversive prediction error that quantifies whether cues and outcomes are worse than expected. A neural representation within the BLA for distinct threat levels was mirrored in self-reported subjective anxiety across individuals. Furthermore, high trait-anxious individuals were characterized by indiscriminately heightened BLA activity in response to aversive cues, regardless of actual threat level.

Conclusions: Our results demonstrate that amygdala subregions are distinctly engaged in processing of aversive experience, with elevated and undifferentiated BLA responses to threat emerging as a potential neurobiological mediator of vulnerability to anxiety disorders.

Keywords: Anxiety; Basolateral amygdala; Centromedial amygdala; Emotional processing; Threat; fMRI.

Figure 1

Experimental task. At cue presentation, one of two insects (mosquito or bug) appearing next to a hand indexed an objective probability, learned before the scanning session, of an upcoming electrical shock. One insect indicated a high probability and one insect indicated a low probability of receiving a shock. At outcome, an appearance of a red dot superimposed on the hand indicated receipt of concurrent shock. By contrast, a red dot next to the hand indicated no shock. Following a jittered fixation, subjects were asked to report how anxious they remembered feeling during cue presentations, i.e., while the insect had been present. After another jittered fixation, one of the two insects appeared again to indicate the beginning of the next trial.

Figure 2

Amygdala responses to different levels of threat at cue presentation. (A) Greater basolateral amygdala (BLA) activity at time of cue presentation was associated with enhanced objective threat levels, i.e., high vs. low probability of upcoming shock. (B) Trend-level interaction between subregion and expectation at cue, with significant threat modulation (high vs. low probability of upcoming shock) in BLA and no effect in centromedial amygdala (CMA). Mean β values for bilateral BLA and CMA masks. *p < .05, (*)p = .074. Error bars indicate SEM. Neural results are presented as SPM activation maps overlaid on a default structural brain in MRIcron . a.u., arbitrary units; n.s., not significant.

Figure 3

Dissociation between basolateral amygdala (BLA) and centromedial amygdala (CMA) in response to aversive events. (A) Shock vs. no shock outcomes are associated with increased outcome-related activity in the CMA. (B) Low- vs. high-expectation cues are associated with increased activity in the BLA at the time of outcomes, contrasting with expectation-related modulation at the time of cue presentation. (C) Response to all 4 outcome types in CMA. Activity in CMA represents aversive events, which are not modulated by expectations. Mean β values for bilateral CMA mask. High probability and low probability indicate high and low probability of shock. (D) Response to all 4 outcome types in BLA. Activity in BLA represents an aversive prediction error that depends on both aversive events and expectations about those events. Stronger activation for less predicted (low probability of shock) compared with highly predicted (high probability of shock) aversive events. Stronger attenuation of responses for less predicted (high probability of shock) compared with highly predicted (low probability of shock) omission of aversive events. Mean β values for bilateral BLA mask. High probability and low probability indicate high and low probability of shock. (E) Significant interaction between subregion and outcome as indicated by greater shock responses in CMA than BLA. Significant interaction between subregion and expectation as indicated by significant effect of expectation in BLA and no effect in CMA. Mean β values for bilateral BLA and CMA masks. ***p < .001, **p < .01, *p < .05. Error bars indicate SEM. a.u., arbitrary units, n.s., not significant.

Figure 4

Amygdala activity and subjective aversive experience. (A) Subjective reports of remembered anticipatory anxiety at cue presentation. Anxiety ratings both were influenced by objective threat level in the cue period, i.e., greater for high vs. low expectation of upcoming shock, and were biased by experienced outcomes, i.e., greater for shock vs. no shock trials. Error bars indicate SEM. (B) A greater neural difference between cue-elicited basolateral amygdala (BLA) responses (high vs. low probability of upcoming shock) was linked to a greater dissociation between threat levels in anxiety ratings (high vs. low probability of shock). *p < .05. (C) Positive correlation between trial-by-trial variability in centromedial amygdala (CMA) activity at time of reporting on a visual analog scale and retrospective reports of subjective anxiety at cue presentation. a.u., arbitrary units.

Figure 5

Threat signals in basolateral amygdala (BLA) and trait anxiety. (A) A greater overall cue-related BLA response (high and low probability of upcoming shock) was associated with greater trait anxiety. (B) A greater neural difference between cue-related BLA responses (high vs. low probability of upcoming shock) was associated with lower trait anxiety. *p < .05. a.u., arbitrary units.