PROJECT SUMMARY/ABSTRACT Anorexia nervosa (AN) is an often-chronic eating disorder with the highest mortality rate of any mental illness, significant costs, and global disease burden. There is a critical need to identify brain-based factors that perpetuate AN symptoms and that may serve as mechanistic targets for existing and novel treatments. Most neurobiological studies in AN have focused on food-related behavior, and have specifically linked these symptoms to broad deficits in frontostriatal activation. However, biobehavioral research to date has failed to account for brain-based mechanisms that may maintain driven exercise, an alarming symptom experienced by a majority of adolescents with AN (59-80%). The goal of this K23 mentored patient-oriented research career development award is to better understand the neurocomputational underpinnings of reinforcement learning in adolescents with AN who engage in driven exercise (AN-DEx). Specifically, the proposed study leverages decision tasks to examine whether adolescents with AN-DEx demonstrate differences in reinforcement learning related to food or exercise reward stimuli. This study will compare task responses in 50 adolescents with AN-DEx, to those of 50 with AN, and 100 age-and activity-matched controls. As a secondary exploratory aim, this study will use functional magnetic resonance imaging (fMRI) to characterize neural activity substantiating task performance for a portion of each group (25 from each, 75 total). This study design will test the following hypotheses: Aim 1: H1a: Compared to controls, AN + AN-DEx will demonstrate deficits in model- based strategy (forward planning) in response to food and exercise stimuli. H1b: Compared to AN, AN-DEx will demonstrate deficits in model-based strategy in response to exercise stimuli; Aim 2: H2a: Compared to controls, AN + AN-DEx will demonstrate increased OFC - NAcc functional connectivity (frontal-limbic pathway, key brain regions implicated in inhibitory control). H2b: Compared to AN, AN-DEx will demonstrate increased OFC-NAcc functional connectivity in response to exercise stimuli. Data from this project will substantiate an explanatory model of DEx, pinpoint which components of reinforcement learning are altered in AN-DEx, and identify ways in which behavioral control-focused interventions may be most effective. This line of inquiry will ultimately inform targeted interventions that can more effectively interrupt DEx, and other compulsive AN symptoms. The current study will also serve as a vehicle for mentorship and training in concepts and skills that are critical to the candidate’s current project, and next steps. Specifically, the proposed training will allow the candidate to gain new knowledge in: (i) cognitive neuroscience and neural substrates specific to eating disorders, (ii) neurocomputational tasks and modeling, and (iii) preliminary training in fMRI. This project and fulfillment of the training goals will launch the candidate’s independent career in the translational neuroscience of AN and lay groundwork for future high-impact studies that combine sophisticated analytic approaches and neuroimaging to improve eating disorder treatment.

PROJECT NARRATIVE Anorexia nervosa (AN) is a severe and treatment-resistant psychiatric illness of unknown etiology, with pernicious and alarming symptoms such as driven exercise. This study is novel in its use of decision tasks and neuroimaging to index reward and reinforcement learning among adolescents with AN, specifically in an effort to elucidate biobehavioral factors that determine for whom and why driven exercise may arise and be maintained. A mechanistic understanding of the processes that contribute to heterogeneity in AN symptomology will assist in informing precision treatment development.