Circuit-based framework for fine spatial scale clustering of orientation tuning in mouse V1

Recent population imaging studies in mouse primary visual cortex (V1) have revisited and questioned the traditional view of 'salt-and-pepper' organization of orientation tuning preference. A controversy has emerged on whether the organization is truly spatially disorganized, or if there is long-range (~100 μm) or short range (~10 μm) spatial clustering of orientation tuning. We imaged and analyzed the response profile (to orientated gratings) of in vivo populations of layer 2/3 (L2/3) and layer 4 (L4) neurons with a neuropil-free nuclear-targeted GCaMP6s indicator. These datasets provide an unprecedented picture of spatial responses, without the contamination of neuropil responses that can unduly bias correlated responses. A significant similarity of orientation tuning properties constrained over a fine spatial scale (around 5-20 μm) was revealed in both L2/3 and L4 -- we term this structure a 'micro-clustered' organization. To explore the novel elements of intracortical connectivity in mouse V1 that underlie the emergence and propagation of micro-clusters in L2/3, we considered an excitatory-inhibitory balanced network with spatially dependent connectivity. To cover the fine spatial scale correlations measured in vivo, in addition to the baseline recurrent synaptic wiring over hundreds of microns in mouse V1, we hypothesize a second, narrow scale wiring comparable to the size of the micro-clusters, so that the network model could quantitatively match the micro-clustered organization from L2/3 recordings. To validate this hypothesis, we leveraged techniques of rabies tracing and single-cell resolution holographic optogenetics to map the intracortical synaptic wiring of mouse V1 L2/3 pyramidal cells. In agreement with our model predictions, large neuron-to-neuron influences were restricted to cells that were within 20 μm of each other. Together, our work takes an important step in building a circuit-based theory of visual processing in mouse V1 over spatial scales that are often ignored, yet contain powerful synaptic interactions.