Neurometabolic signaling and control of policy complexity

Cognition and adaptive behavior emerge from neural information processing. This must operate within finite metabolic constraints, since neural information processing is metabolically expensive. While neural implementations of action selection and learning are well-studied, systems allocating the informational capacity required to encode complex behavioral policies remain unknown. We hypothesized that hypothalamic hypocretin/orexin neurons (HONs) are uniquely positioned to signal and control policy complexity, given that they are activated by metabolic depletion and influence decision-making systems. To explore this, we employed a set of cell/neurotransmitter-specific imaging and causal manipulations during a multi-armed bandit task where freely behaving mice learned probabilistic state-action-reward relationships (together [~]100,000 decisions from >100 mice). Miniscope recordings of HON activity revealed that pre-choice, but not post-choice activity correlates with decision policy, dissociating decisions from feedback. Furthermore, manipulating HON signals with optogenetics and pharmacology confirmed that they causally regulate the development of complex policies. Finally, neurotransmitter-specific sensors revealed that hypocretin/orexin receptors modulate decision policy-related dopamine and noradrenaline dynamics in the nucleus accumbens and medial prefrontal cortex. These findings identify HONs as subcortical regulators of policy complexity, encoding critical signals for decision-making adjustments. This opens a new window for the development of comprehensive mechanistic models of strategic learning which account for interplay between decision-making policies and metabolic/informational constraints, and may guide the development of potential treatments for disorders with policy deficits such as autism and schizophrenia.