NeuroImage (Feb 2020)
Clarifying the neural substrates of threat and safety reversal learning in humans
Abstract
Responding flexibly to sources of threat and safety is critical to the adaptive regulation of emotions, including fear. At a neural systems level, such flexibility is thought to rely on an extended neural circuitry involving the dorsal anterior cingulate cortex (dACC) and ventromedial prefrontal cortices (vmPFC), although precisely how this occurs remains unclear. Using a novel fear reversal task and functional magnetic resonance imaging (fMRI), we examined the neural correlates of threat and safety reversal learning and their associations with individual differences in anxious responding in a large sample of healthy adolescents and young adults. Overall, participants demonstrated successful threat and safety reversal learning, as indexed by subjective ratings. At a whole-brain level, threat reversal was associated with significant activation of the bilateral anterior insular cortex and dACC, in particular its rostral subregion. Conversely, safety reversal led to significant activation of the anterior vmPFC, together with posterior mid-line regions. Further analyses of regional responses suggested a more selective role for the rostral dACC in threat signal updating, as well as a direct association of its activity with participants’ change in subjective anxious arousal to the reversed threat. Taken together, our findings complement existing neurocircuitry models of human fear regulation, particularly regarding the importance of midline cortical regions, and provide further insights into their specific contribution to flexible threat-safety signal processing. In particular, our results suggest that rostral dACC function may be more centrally involved in regulating levels of anxious arousal when flexibility is required. They also raise important questions regarding the vmPFC’s role in safety learning, particularly involving its hypothesized subregional contributions to response inhibitory versus stimulus value processing functions.