Introduction: Neural mechanisms underlying behavior therapy in obesity are poorly understood, particularly in patients with comorbid depression. Putative targets include neural circuits for conscious and nonconscious emotion regulation. Hypothesis: Changes in a priori neural circuits for emotion regulation from baseline (0) to 2 mo (mediators, M) mediate changes in BMI, physical activity, and eating behaviors from 0 to 6 or 12 mo (outcomes, Y). Methods: In a trial of integrated behavior therapy, adults with obesity and depression were randomized to intervention (n=59) or usual care (n=49). The yearlong intervention included 6 in-person problem solving therapy (PST) sessions for depression by 2 mo, 3 additional PST sessions for depression and 11 home videos on lifestyle changes for weight loss by 6 mo, and 6 maintenance calls by 12 mo. Participants completed validated conscious and nonconscious facial emotion viewing tasks in fMRI at 0 and 2 mo (time point corresponding to initial PST for depression in the intervention) and weight and height (baseline only) measurements, 7-day physical activity recalls, and 24-hour dietary recalls at 0, 6, and 12 mo. Analyses tested mediation as showing both a significant treatment effect (X) of the intervention vs. usual care on the mediator (X->M, path a ) and a significant correlation of the mediator with the outcome either as a main effect or an interaction effect by treatment group (M->Y, path b ). Results: Compared with usual care, the intervention led to significantly reduced nonconscious reactivity to threat-related facial emotions (fear, anger) at 2 mo for the left (-0.89, 95% CI -1.64 to -0.14; P =.02) and right amygdala (-0.84, 95% CI -1.54 to -0.13; P =.02) ( path a ). These reductions correlated significantly with increased leisure-time physical activity (MET minutes/wk) at 12 mo: main effect -686.0 (95% CI -1127.9 to -244.1; P =.003) and interaction effect 664.7 (95% CI 73.1 to 1256.4; P =.03) for the left amygdala and main effect -692.9 (95% CI -1365.3 to -20.6; P= .04) for the right amygdala. Similarly, the reductions correlated significantly with increased energy expenditure (kcal/kg/d) at 12 mo: main effect -1.72 (95% CI -2.49 to -0.95; P <.001) and interaction effect 1.89 (95% CI 0.84 to 2.95; P <.001) for the left amygdala; and main effect -1.82 (95% CI -3.08 to -0.56; P =.006) and interaction effect 1.94 (95% CI 0.50 to 3.38; P =.01) for the right amygdala ( path b ). Nonconscious reactivity to threat in bilateral amygdala did not correlate significantly with changes in BMI or eating behaviors. Conclusions: Reduced nonconscious reactivity to threat in bilateral amygdala in response to initial PST for depression significantly mediated improvements in leisure-time physical activity and energy expenditure. Results advance the understanding of neural mechanisms underlying behavior therapy in obesity and inform intervention optimization to enhance efficacy.