Mesoscale correlation structure with single cell resolution during visual coding
AbstractNeural circuitry represents sensory input with patterns of spiking activity. Across brain regions, initial representations are transformed to ultimately drive adaptive behavior. In mammalian neocortex, visual information is processed by primary visual cortex (V1) and multiple higher visual areas (HVAs). The interconnections of these brain regions, over which transformations can occur, span millimeters or more. Shared variability in spiking responses between neurons, called “noise correlations” (NCs), can be due to shared input and/or direct or indirect connectivity. Thus, NCs provide insight into the functional connectivity of neuronal circuits. In this study, we used subcellular resolution, mesoscale field-of-view two-photon calcium imaging to systematically characterize the NCs for pairs of layer 2/3 neurons across V1 and four HVAs (areas LM, LI, AL and PM) of mice. The average NCs for pairs of neurons within or across cortical areas were orders of magnitude larger than trial-shuffled control values. We characterized the modulation of NCs by neuron distance, tuning similarity, receptive field overlap, and stimulus type over millimeter scale distances in mouse visual cortex, within and across V1 and multiple HVAs. NCs were positively correlated with shared tuning and receptive field overlap, even across cortical areas and millimeter length scales. We compared the structure of these NCs to that of hypothetical networks to determine what network types can account for the results. We found that to reproduce the NC networks, neuron connectivity was regulated by both feature similarities and hub mechanism. Overall, these results revealed principles for the functional organization and correlation structure at the individual neuron level across multiple cortical areas, which can inform and constrain computational theories of cortical networks.