Metabolism and the mind: Investigating the link between glucose control and reinforcement learning in humans

Background: Signals from the body profoundly influence cognition. This process is known as interoception, and has been extensively studied in the cardiac, respiratory, and gastric domains; in contrast, metabolic influences remain poorly understood. Here, we focus on the link between glucose control and cognition, motivated by the observation that there is substantial, unexplained comorbidity between type-2 diabetes and depression. In rodents, insulin modulates dopamine signalling in the ventral striatum. We therefore hypothesised that, in humans, differences in glucose control would be associated with altered reward learning. Methods: To test this hypothesis, we recruited 48 participants from the general population, who each completed a glucose tolerance test, a monetary reward learning task known to relate to dopamine function, and mental health questionnaires. We fitted an established reinforcement learning model to the task data in order to obtain computational parameters characterising participants’ learning, and then examined the associations between these parameters and their glucose control. Results: We discovered that poorer glucose control was associated with greater reliance on recent rewards during learning which was, in turn, associated with higher depression symptoms. There was also more modest evidence for the association between glucose control and depression symptoms. Conclusions: Together, our results identify a specific neurocognitive process, reward learning, by which metabolic information may influence cognition, and which may explain the link between metabolic diseases like type-2 diabetes and depression.