Abstract
Dysfunction in emotion regulation (ER) and autobiographical memory are components of major depressive disorder (MDD). However, little is known about how they mechanistically interact with mood disturbances in real time. Using machine learning-based neural signatures, we can quantify negative affect (NA), ER, and memory continuously to evaluate how these processes dynamically interact in MDD. Unmedicated individuals with MDD (N=45) and healthy volunteers (HV; N=38) completed a negative autobiographical memory functional magnetic resonance imaging task wherein they recalled, distanced from (an ER strategy), and immersed into memories. We used a negative affect signature (PINES) and an emotion regulation signature (ERS) to quantify moment-to-moment NA and ER. We then examined whether memory engagement, indexed by hippocampal activity, predicted subsequent change in PINES and ERS over time. During memory recall and immersion, greater hippocampal activity predicted increased PINES across groups. During distancing, greater hippocampal activity in HVs predicted increased ERS but not PINES. In MDD, greater hippocampal activity predicted increased PINES but not ERS. Findings suggest abnormalities in the real-time relationship between memory, NA, and ER in MDD. During distancing, as predicted, HVs showed an attenuation of the linkage between memory engagement and NA, and they had subsequent increases in ER following memory reactivation. In contrast, MDD was characterized by continued linkage between memory engagement and NA, without subsequent increases in ER. Deficits in engagement of ER and ineffective modulation of NA following negative memory recall may contribute to the mood disturbances in MDD and are potential targets for clinical intervention.