Abstract
Abstract
Food reward valuation and decision making are key psychological processes regulating eating behaviour. Stress and negative affect can shape these processes, increasing vulnerability to overeating high-caloric, ultra-processed foods and thereby elevating obesity risk. Yet not all individuals under stress exhibit such behavioural shifts, highlighting the need to understand which modulating factors explain this inter-individual variability. Emerging evidence suggests that anxiety-/stress-induced changes in gut microbiota might alter reward sensitivity via the gut–brain axis. An integrative framework linking microbiota, neural, and affective processes is thus essential for explaining individual differences in emotional eating, yet such a framework is still lacking. To this end, supervised linked independent component analysis (Super-BigFLICA) was used to explore multivariate gut–brain latent components guided by affective and body composition indexes in healthy participants (N = 592) from the Healthy Brain Study. Next, general linear models were used to identify if any of the observed components were predictive of elevated unhealthy food valuation and choice in a food auction task. A multimodal component was predictive of both food valuation and choice. This component was characterized by co-occurring changes in several gut bacterial taxa at the genus level (e.g., UCG-010 and DTU089), specific connectivity between reward-related brain regions (e.g., nucleus accumbens, caudate), and was clearly linked to elevated anxiety sensitivity, life-event impact, and body mass index. These findings demonstrate the feasibility of supervised gut-brain integration for predicting behaviour and reveal a potential phenotype linking microbiota composition, brain function, and emotional state that should be further investigated in larger cohorts characterized by emotional eating and overweight.