Probability-dependent scaling of fear responding, but not unconditioned stimulus inflation, requires the periaqueductal gray

Adaptive behavior requires animals to estimate threat, update these estimates following new aversive outcome experiences, and align fear responding to up-to-date estimates. The periaqueductal gray is traditionally viewed as a downstream relay that converts threat estimates into freezing. Here we tested whether the periaqueductal gray is necessary for probability-dependent scaling of fear responding, and for updating threat estimates through unconditioned stimulus inflation. We deleted neurons across the complete anterior-posterior span of the periaqueductal gray using a dual-viral approach, then quantified deletion extent and subregion specificity using machine-learning methods. In Experiment 1, 40 Long Evans rats (20 female) underwent fear discrimination in which a threat cue predicted foot shock with 30% or 10% probability, while a neutral cue never predicted foot shock. Ethograms revealed that control rats scaled a suite of reward, exploratory, and defensive behaviors to threat probability (threat > neutral, and 30% > 10%). Periaqueductal gray deletion abolished scaling (threat > neutral, but 30% ≈ 10%). In Experiment 2, 40 Long Evans rats (20 female) first received light and tone conditioning to a mild foot shock. During unconditioned stimulus inflation, half the rats received light-intense shock pairings, while the remaining rats received additional light-mild shock pairings. Periaqueductal gray deletion left inferred inflation intact (intense > mild; tone extinction responding), and enhanced direct inflation (intense > mild; light extinction responding). These results demonstrate that the periaqueductal gray is essential for scaling fear responding to threat probability, but dispensable for, and can even constrain, updating threat predictions through unconditioned stimulus inflation. Significance Statement This important work demonstrates an essential role for the periaqueductal gray (PAG), widely considered to simply organize freezing, in scaling behavior systems responding to threat probability. The evidence supporting these conclusions is compelling, including the use of innovative threat discrimination & unconditioned stimulus inflation procedures, machine-learning-based quantification of neuronal deletion, comprehensive ethograms, and bidirectional analyses linking PAG deletion to probability-insensitive fear responding. These results identify the PAG as a critical site where disruption of threat estimation may contribute to stress and anxiety disorders.