Functional consequences of highly shared feedforward inhibition in the striatum

The striatum is crucial for motor control and reinforcement learning. The striatum has only 1% fast-spiking interneurons (FSI) which constitute feedforward inhibition (FFI). Therefore, striatal neurons receive a high degree of shared FFI, which makes the striatum a unique network in the brain. Here we ask, what role such high shared FFI may play in striatal function besides inducing synchrony. We found that with highly shared connectivity FSIs increase the trial-by-trial variability of striatal responses. Moreover, when the striatum is driven by correlated inputs, FSIs decorrelate the responses but when input is uncorrelated, FSIs correlate the responses. Thus, with their high connectivity to striatal neurons, FSIs create a correlation fixed point in the striatum. This effect disappears when we reduced the FSIs’ shared connectivity by increasing the number of FSIs in our model. This non-monotonic effect of FSIs in controlling striatal correlation underlies a temporal winner-take-all dynamics, especially in the early stages of learning e.g. in pattern separation. FSI connectivity is modulated by acetylcholine which can, in a context-dependent manner, dynamically control the effect of FSIs on trial-by-trial variability and pattern separation ability. Our results generalize to other small populations of neuron types with high connection probabilities.