Item

Abstract

In order to gain a mechanistic understanding of sensory signal flow in the rat vibrissal system, it is necessary to determine the activity of neurons in response to sensory inputs, and to identify the spatial distribution of synapses to and from these neurons. While the responses of neurons to whisker stimuli have been characterized extensively, a complete quantitative description of connectivity in the vibrissal system is still lacking. Therefore, we aim to determine all inputs and outputs to and from neurons in rat barrel cortex. Two widely used techniques for investigating neural connectivity are electron-microscopic (EM) reconstruction of brain circuits and tracing of bulk injections. However, so far these approaches have been limited to small circuits of hundreds of neurons (EM reconstruction) or only allow limited extraction of quantitative connectivity data (bulk injections). Here, we present a third, statistical approach for obtaining dense synaptic connectivity of neural networks from representative samples of 3D morphologies of single neurons labeled in vivo. To do so, we measure (i) the 3D geometry of rat barrel cortex, (ii) the 3D distribution of excitatory and inhibitory neuron somata, and (iii) the 3D morphology of dendrites and axons of 11 excitatory and 6 inhibitory neuron cell types. Combining these data in a common 3D reference frame allows estimating the 3D distribution of 5 billion boutons from 6,000 thalamocortical and 500,000 intracortical neurons in a volume of 6.5 cubic mm, at a resolution of 50 μm. Putative synaptic innervation at subcellular resolution is predicted from the overlap between reconstructed dendrites and axons. In the past, a variety of results obtained by statistical mapping deviated from direct connectivity measurements. Comparing the resultant cell type-specific innervation probabilities and cell-specific subcellular innervation patterns with previously reported in vivo paired-recordings and EM data, we demonstrate the validity and appropriate spatial resolution of this statistical approach. Finally, we show how this statistical approach can be used to generate a dense cell-to-cell connectome of barrel cortex. Analysis of the resultant connectivity reveals cell type- and location-specific connectivity patterns from the microscopic, subcellular level up to the macroscopic scale of the entire barrel cortex, which form the structural basis underlying sensory-evoked signal flow.