Ventral pallidum GABA and glutamate neurons drive approach and avoidance through distinct modulation of VTA cell types

Abstract

The ventral pallidum (VP) contains GABA and glutamate neurons projecting to ventral tegmental area (VTA) whose stimulation drives approach and avoidance, respectively. Yet little is known about the mechanisms by which VP cell types shape VTA activity and drive behavior. Here, we found that both VP GABA and glutamate neurons were activated during approach to reward or by delivery of an aversive stimulus. Stimulation of VP GABA neurons inhibited VTA GABA, but activated dopamine and glutamate neurons. Remarkably, stimulation-evoked activation was behavior-contingent such that VTA recruitment was inhibited when evoked by the subject’s own action. Conversely, VP glutamate neurons activated VTA GABA, as well as dopamine and glutamate neurons, despite driving aversion. However, VP glutamate neurons evoked dopamine in aversion-associated ventromedial nucleus accumbens (NAc), but reduced dopamine release in reward-associated dorsomedial NAc. These findings show how heterogeneous VP projections to VTA can be engaged to shape approach and avoidance behaviors.

Ventral tegmental area interneurons revisited: GABA and glutamate projection neurons make local synapses

“…It is now understood that DA signals can induce or correlate with distinct behavioral responses depending on their projection targets ( Azcorra et al, 2023 ; Badrinarayan et al, 2012 ; de Jong et al, 2019 ; Faget et al, 2024 ). This is true also for VTA GABA and glutamate neurons.…”

“…Local intra-VTA GABA modulation of VTA output, particularly DA output, may underlie key aspects of behavioral reinforcement. For example, inhibitory inputs to VTA from lateral hypothalamus, bed nucleus of stria terminalis, or VP can drive positive reinforcement and approach behaviors through inhibition of VTA GABA neurons and disinhibition of VTA DA neurons ( Faget et al, 2024 ; Nieh et al, 2015 ; Soden et al, 2020 ; Soden et al, 2023 ). VTA GABA circuits also appear to be critical for the generation of DA reward prediction error signals ( Eshel et al, 2015 ; Keiflin & Janak, 2015 ).…”