A slow 5-HT1AR-mediated recurrent inhibitory network in raphe computes contextual value through synaptic facilitation

Abstract

AbstractSerotonin (5-HT) neurons in the dorsal raphe nucleus (DRN) receive a diverse constellation of long-range synaptic inputs, yet unifying principles of local circuitry and its dynamics are largely unknown - a crucial component of understanding how 5-HT output controls behavior. Here, we developed a formalism bridging optogenetic, electrophysiological, computational and behavioral strategies to reveal how the dynamics of local circuitry in DRN control the expression of reward associations. Using long-range input from lateral habenula (LHb) to interrogate functional DRN circuitry, we uncover 5-HT1A receptor-mediated local recurrent connections between 5-HT neurons, refuting classical theories of autoinhibition by 5-HT1A receptors. These inhibitory 5-HT connections were slow, stochastic, strongly facilitating, and gated spike output of 5-HT neurons. Targeted physiology and modeling approaches revealed that these functional connectivity features collectively support the emergence of a paradoxical excitation-driven inhibition in response to high frequency LHb activation, and of a winner-take-all computation over protracted timescales. In vivo, we found that optogenetic activation of LHb inputs to DRN transiently disrupted expression of a reward-conditioned response in an auditory conditioning task. In accordance with quantitative model predictions, this disruption occurred exclusively at the conjunction of high frequency LHb activation and high predicted reward value, and was not due to a modulation of the underlying reward association. Thus, we propose that recurrent dynamics in the DRN support a contextual value computation, where stable learned associations are integrated with dynamic environmental inputs to support sharp behavioral state transitions in changing environments.