A multiscale model of striatum microcircuit dynamics

The striatum is the largest structure in the basal ganglia, and is known for its key role in functions such as learning and motor control. Studying these aspects requires investigating cellular/microcircuits mechanisms, in particular related to learning, and how these small-scale mechanisms affect large-scale behavior, and its interactions with other structures, such as the cerebral cortex. In this paper, we provide a multiscale approach to investigate these aspects. We first investigate striatum dynamics using spiking networks, and derive a mean-field model that captures these dynamics. We start with a brief introduction to the microcircuit of the striatum and we describe, step by step, the construction of a spiking network model, and its mean-field, for this area. The models include explicitly the different cell types and their intrinsic electrophysiological properties, and the synaptic receptors implicated in their recurrent interactions. Then we test the mean-field model by analyzing the response of the striatum network to the main brain rhythms observed experimentally, and compare this response to that predicted by the mean-field. We next study the effects of dopamine, a key neuromodulator in the basal ganglia, on striatal neurons. Integrating dopamine receptors in the spiking network model leads to emerging dynamics, which are also seen in the mean-field model. Finally, we introduce a basic implementation of reinforcement learning (one of the main known functions of the basal-ganglia) using the mean-field model of the striatum microcircuit. In conclusion, we provide a multiscale study of the striatum microcircuits and mean-field, that capture its response to periodic inputs, the effect of dopamine and can be used in reinforcement learning paradigms. Given that several mean-field models have been previously proposed for the cerebral cortex, the mean-field model presented here should be a key tool to investigate large-scale interactions between basal ganglia and cerebral cortex, for example in motor learning paradigms, and to integrate it in large scale and whole-brain simulations.