Dopamine Dependent Modulation of Striatal Protein Kinase A Activity During Reinforcement Learning

Abstract

In the classical model of dopamine and basal ganglia dependent reinforcement learning, the patterns of midbrain dopamine neuron activities encode a reward prediction error, a signal that has been proven to be a useful teaching signal to update an animal’s understanding of its environment. These dopamine neurons, in turn, project to the input region of the basal ganglia, the striatum, and provides a teaching signal to striatal neurons in a form of dopamine. In particular, dopamine has been hypothesized to modulate striatal neurons’ protein kinase A (PKA) activity, which has been implicated in diverse synaptic plasticity mechanisms. However, this fundamental connection between dopamine and striatal PKA remains untested in behaving animals in the setting of reinforcement learning due to technical difficulties of monitoring intracellular pathway activities in-vivo. To directly investigate this connection, we developed a fluorescence lifetime photometry (FLiP) technique that allowed us to monitor dynamically changing PKA activity in genetically defined population of neurons in behaving mice. Combining FLiP with fiber photometry technique, we were able to monitor multiple steps of the dopamine signal transformation in the ventral tegmental area (VTA), a part of midbrain, and the nucleus accumbens (NAc), a ventral part of striatum: VTA dopamine neuron soma activity, dopamine neuron terminal activity in the NAc, release of dopamine in the NAc, and PKA activity in two distinct populations of striatal neurons in the NAc. With this approach, we found that the release of dopamine in the NAc is highly correlated with the soma and terminal activity of midbrain dopamine neuron, patterns of which are consistent with reward prediction error. Also, the modulation of dopamine level in NAc was correlated with and causally related to the asynchronous activation of PKA in type-1 (D1R) and type-2 DA receptor (D2R) expressing striatal neurons in the NAc in different stages of learning. In summary, we propose that FLiP can be a useful tool for measuring intracellular pathway activities in behaving animals and that the dopamine and basal ganglia dependent reinforcement learning model should incorporate the asynchronous activation of PKA in D1R and D2R expressing striatal neurons by the RPE encoding dopamine.