The neural computation of affective internal states in the hypothalamus: A dynamical systems perspective

Internal affective states accompany evolutionarily ancient survival behaviors such as mating, aggression, and predator defense and may contribute to emotional feelings in humans. In this perspective, we introduce a dynamical system framework for thinking about such states. We synthesize evidence from recent studies suggesting that key state features, such as their intensity and duration, may be encoded by approximate line attractor manifolds in the hypothalamus. Evidence for these attractors arises from unsupervised data-driven dynamical system modeling of high-dimensional calcium imaging data from genetically identified cell populations in freely behaving mice. Dissection of the fit dynamical models and closed-loop modeling with experimental perturbations raise new questions regarding circuit- and cellular-level mechanisms of attractor implementation. These findings challenge prevailing views of hypothalamic behavioral control and afford a new avenue to study the emergence of slow state-encoding neural dynamics across scales, from single neurons to recurrent networks and neuromodulatory signaling.