Connectomic analysis of astrocyte-synapse interactions in the cerebral cortex

Astrocytes, a main type of glia cells in the cortex, provide metabolic support to neurons, and their possible function as a synaptic partner has given rise to the notion of "tripartite" synapses, suggesting a contribution to neuronal computations. For astrocytes to serve such purposes, the interactions with synapses in neuronal circuits require a level of specificity beyond overall synaptic support. A systematic mapping of the astrocyte-connectome relationship would enable the testing of these hypotheses - such analysis is however still lacking, in particular for circuits in the cerebral cortex. Here, utilizing previously published connectomic data of more than 200,000 synapses, we systematically analyzed the spatial relation between astrocytes and synapses in mouse somatosensory cortex. We developed a quantitative assessment of astrocyte-synapse proximity, finding that only 22.7% of synapses are contacted by astrocytic processes for more than 50% of their synaptic circumference. This non-ubiquitous astrocytic attachment would render astrocyte-synapse specificity plausible. Astrocytic coverage depended strongly on synapse types, with thalamocortical shaft synapses being the most covered by astrocytic processes. We furthermore observed a strong dependence of astrocytic synaptic coverage on synapse size, which was exclusive for excitatory spine synapses. We then investigated the possible relation of astrocytic synaptic coverage to neuronal activity and synaptic plasticity, finding ultrastructural evidence for substantially reduced astrocytic support at synapses consistent with long-term depression, but not for astrocytic coverage dependence on baseline neuronal presynaptic activity. Together, our data demonstrate a high level of specificity of astrocyte-synapse interactions for particular synaptic types. They indicate the potential relevance of astrocytic coverage for synapse stability, in particular for large synapses, suggesting a contribution to long-term maintenance of learned synaptic states. These methods will allow a systematic testing of hypotheses about glial-neuronal interaction in various brain regions, disease models and species including human.