Synaptic input architecture of visual cortical neurons revealed by large-scale synapse imaging without backpropagating action potentials

Abstract

How neurons integrate thousands of synaptic inputs to compute sharply tuned outputs is a critical question in sensory information processing. To answer this question, it is essential to record the location and activity of synaptic inputs in vivo. However, back-propagating action potential (BAP) calcium signals invade dendrites and spines, making accurate recording of spine responses difficult. In this study, we first developed a new method to record spine calcium responses without BAP signals. Using this method, we performed large-scale imaging of visually evoked spine activity from layer 2/3 pyramidal neurons and revealed three patterns of dendritic functional architectures of synaptic inputs: dendrites with clusters of spines of similar responses, dendrites with spines of diverse responses, and dendrites with spines where the majority of them show no visual response. Our model suggests that only a small fraction of spines on dendrites of clustered architectures are sufficient to generate sharply tuned output.