The Amygdala and Aberrant Fear
Material below summarizes the article Disrupted Prediction Error Links Excessive Amygdala Activation to Excessive Fear, published on January 13, 2016, in JNeurosci and authored by Auntora Sengupta, Bryony Winters, Elena E. Bagley, and Gavan P. McNally.
The amygdala, a group of nuclei in the medial temporal lobe, is a key brain structure for fear and anxiety. Several neuropsychiatric disorders are characterized by heightened activation of the amygdala. fMRI studies in clinically anxious people consistently show increased amygdala activity compared to controls during anxiety symptom provocation or presentations of negative emotional stimuli.
This heightened amygdala activation has been hypothesized to underlie the core psychopathologies in anxiety: increased emotional reactivity; over generalization of fear so that safe people, places, or things are perceived as dangerous; and deficits in fear inhibition so that fear lingers on well after danger has passed.
Despite the robustness of these neuroimaging findings, the mechanisms linking heightened amygdala activation to heightened emotional learning remain elusive.
Attempts have been made to mimic heightened activity in the amygdala in the laboratory via complex environmental events, such as acute or chronic stress, that have strong face validity for the aetiology of human clinical anxiety. However, these events are imprecise and poorly defined in terms of their impact at cellular and circuit levels on the amygdala.
Further complicating the matter, different amygdala nuclei serve complementary but different roles in fear, and each nucleus is composed of a variety of classes of neurons. These different classes of neurons have distinct roles in regulating excitability and plasticity of amygdala principal neurons, the key amygdala neuron class for learning.
We addressed these issues using an approach developed by Bryan Roth and colleagues. We expressed a receptor, formally called the hM3Dq DREADD, not normally present in mammals within principal neurons of the basolateral amygdala. This allowed us to selectively activate these neurons by injecting a selective agonist, CNO, that does not bind to any endogenous receptors. The advantages of this approach were that we could control the activity of amygdala cell types directly and that the method of exciting these cells exploited well-defined cell signaling pathways.
We first verified that application of CNO increased the excitability of basolateral amygdala principal neurons in vitro and in vivo. We then determined the effect of this heightened activation in vivo on fear learning and responding.
The obvious hypothesis was that excitation of basolateral amygdala principal neurons would augment fear learning. It did not. Regardless of whether we established weak or strong fear memories, chemogenetic excitation of basolateral amygdala principal neurons had no effect on fear learning or responding.
Although our experiments were in rats, even sufferers of clinical anxiety do not always show heightened fear learning and memory in simple conditioning preparations. This lack of effect is puzzling. The amygdala is the hub for fear learning and fear memory formation, and principal cells within the basolateral amygdala are the key cellular locus.
Why does heightened activity of these cells not yield stronger or more persistent fear memories, particularly when this phenomenon is readily confirmed in our experiments at the cellular level and is an accepted signature of pathological fear and anxiety?
The answer rests in a more sophisticated understanding of learning. Fear learning proceeds as a function of prediction error, the discrepancy between actual outcomes and expected outcomes. When an unexpected aversive event occurs, prediction error is positive and fear is learned. When an expected event is omitted, prediction error is negative and fear is lost. When actual outcomes and expectancies are matched, there is no prediction error and no further learning.
In short, prediction error drives increases in fear during unexpected danger and reduces fear when danger has passed. Our next step was to determine the impact of heightened amygdala activation on aversive prediction errors during fear learning.
To isolate the action of positive prediction error, we utilized a blocking protocol. Rats were trained to fear a visual stimulus (CSA) via pairings with shock in phase two. In phase one, rats were presented with an audiovisual compound stimulus (CSAB) paired with a shock. Rats were then tested for their fear to the auditory stimulus (CSB). In this classic design, rats do not learn to fear the auditory stimulus because the visual stimulus CSA already predicts the shock. In other words, the CSA blocks learning to CSB.
The hypothesis was that excitation of basolateral amygdala principal neurons would attenuate blocking by reinstating positive prediction error in phase two when it would otherwise be absent.
This was the case. Rats with heightened activity in the basolateral amygdala showed attenuated blocking compared to control rats. We were able to show that heightened amygdala activation is able to restore positive prediction error to produce aberrant fear learning.
Does a similar mechanism prevent the inhibition of fear when expectations of danger are excessive relative to the danger posed? To answer this question, we turned to a behavioral paradigm that isolates the action of negative prediction error on fear learning. We used an overexpectation paradigm: Rats were trained to fear CSA and CSB separately via pairings with shock. These stimuli were then compounded (CSAB) and paired with a shock. Rats were then tested for their fear to CSA and CSB. In this experiment, rats typically demonstrate a loss of fear to CSA and CSB, presumably driven by a negative prediction error that results from an overexpectation of danger when these stimuli are compounded in phase two.
The hypothesis was that excitation of basolateral amygdala principal neurons would attenuate overexpectation by counteracting the negative prediction that drives inhibitory learning to both stimuli in phase two.
This was the case. Rats with heightened activity in the basolateral amygdala showed attenuated overexpectation relative to control rats.
We think our manuscript makes two contributions. First, we were able to isolate a specific learning mechanism linking heightened amygdala activity to heightened fear: disrupted prediction error signals. The disruption of these error signals, which normally constrain and regulate association formation, leads to inappropriate fear associations. Indeed, animals feared stimuli that were not valid predictors of danger and were not able to reduce fear when it was excessive relative to the danger posed.
Second, we were able to identify this mechanism by adopting a more nuanced approach to assessing fear learning. Most studies of fear learning in animals focus on the simple acquisition and extinction of contextual or auditory fear conditioning. This focus is understandable. These behaviors are robust and easily studied in any laboratory. The difficulty is that learning is complex. The simple acquisition and extinction of fear are not always able to isolate the specific learning processes underpinning this complexity which, as our work shows, may go awry in human clinical anxiety.
Disrupted Prediction Error Links Excessive Amygdala Activation to Excessive Fear. Auntora Sengupta, Bryony Winters, Elena E. Bagley, Gavan P. McNally. The Journal of Neuroscience Jan 2016, 36(2): 385-395; DOI: 10.1523/JNEUROSCI.3670-15.2016