Astrocytes strongly promote the formation and maturation of synapses by secreted proteins. To date, several astrocyte-secreted synaptogenic proteins controlling different stages of excitatory synapse development have been identified. However, the identities of astrocytic signals that induce inhibitory synapse formation remain elusive. Here, through a combination of in vitro and in vivo experiments, we identified Neurocan as an astrocyte-secreted inhibitory synaptogenic protein. Neurocan is a chondroitin sulfate proteoglycan that is best known as a protein localized to the perineuronal nets. However, Neurocan is cleaved into two after secretion from astrocytes. We found that the resulting N- and C- terminal fragments have distinct localizations in the extracellular matrix. While the N-terminal fragment remains associated with perineuronal nets, the Neurocan C-terminal fragment localizes to synapses and specifically controls cortical inhibitory synapse formation and function. Neurocan knockout mice lacking the whole protein or only its C-terminal synaptogenic region have reduced inhibitory synapse numbers and function. Through super-resolution microscopy and in vivo proximity labeling by secreted TurboID, we discovered that the synaptogenic domain of Neurocan localizes to somatostatin-positive inhibitory synapses and strongly regulates their formation. Together, our results unveil a mechanism through which astrocytes control circuit-specific inhibitory synapse development in the mammalian brain.

The formation of precise numbers of neuronal connections, known as synapses, is crucial for brain function. Therefore, synaptogenesis mechanisms have been one of the main focuses of cellular and molecular neuroscience. Immunohistochemistry is a common tool for labeling and visualization of synapses. Thus, quantifying the numbers of synapses from light microscopy images enables screening the impacts of experimental manipulations on synapse development. Despite its utility, this approach is paired with low throughput image analysis methods that are challenging to learn, and results are variable between experimenters. We developed a new open-source ImageJ-based software, SynBot, to address these technical bottlenecks by automating several stages of the analysis. SynBot incorporates the machine learning algorithm ilastik for accurate thresholding for synaptic puncta identification, and the code can easily be modified by users. The use of this software will allow for rapid and reproducible screening of synaptic phenotypes in healthy and diseased nervous systems.