Resources
Authors & Affiliations
Haleigh Mulholland,Matthias Kaschube,Gordon Smith
Abstract
Modular (columnar) activity is a fundamental mode of neural activity in the cortex of primates and carnivores. Work in ferret visual cortex has shown that already prior to visual experience, early spontaneous activity is modular, revealing large-scale correlated networks that are predictive of future functional networks representing stimulus orientation. However, the origin of modular activity in the early cortex is unclear at present. Computational models predict a striking ability of developing intracortical circuits to self-organize into modular, coordinated patterns of activity which reflect the spatial statistics of in vivo spontaneous activity. However, this prediction has not yet been tested in vivo during early development. Here, we combine widefield epifluorescence calcium imaging with excitatory optogenetics to simultaneously image and stimulate pyramidal neurons in layer 2/3 of developing ferret visual cortex (postnatal day 24-29, before eye opening). We found that optogenetic stimulation of the cortex with a large (~3 mm) spatially uniform stimulus led to the rapid emergence of non-uniform, modular neural activity, consistent with the predictions of the model. The proportion of strongly modular patterns increased with optogenetic light intensity, suggesting a network activity threshold to generate structured activity from this spatially uniform input. Repeated stimulations lead to a variety of spatial patterns, and opto-evoked activity resided in a moderately low-dimensional subspace that it appeared to share with ongoing spontaneous activity. Notably, these modular opto-evoked events resembled spontaneous activity in their spatial structure and revealed highly similar large-scale correlated networks. As predicted by the computational model, we found that the presence of modularity in opto-evoked events required activity propagating through local intracortical synapses to emerge, and that modularity persisted even in the absence of feedforward input from the LGN. Together, this provides strong evidence that modular patterned activity is an emergent property from intracortical interactions through a self-organizing network.