Fear-related disorders arise from inefficient fear extinction and have immeasurable social and economic costs. Here, we characterize mouse phenotypes that spontaneously show fear-independent behavioral traits predicting adaptive or maladaptive fear extinction. We find that, already before fear conditioning, specific morphological, electrophysiological, and transcriptomic patterns of cortical and amygdala pyramidal neurons predispose to fear-related disorders. Finally, by using an optogenetic approach, we show the possibility to rescue inefficient fear extinction by activating infralimbic pyramidal neurons and to impair fear extinction by activating prelimbic pyramidal neurons.