Synaptic resolution whole-brain circuit mapping of molecularly defined cell types using a barcoded rabies virus 

Single-cell transcriptomics has revolutionized our understanding of neuronal diversity, revealing hundreds of molecularly defined neuronal cell types which are postulated to serve unique roles in sensory processing. While targeted studies of broad cell subclasses provide some evidence for cell type-specific local circuit motifs, a comprehensive map of circuit connectivity at the level of finely resolved cell types remains elusive, in part, due to a lack of high-throughput, cost-effective technologies for mapping connectivity among molecularly defined cell types. 

We and others have recently pioneered sequencing-based approaches to connectivity mapping that leverage high-complexity barcoded viral tools and single-cell sequencing to map connectivity at scale. Specifically, we developed a high-complexity barcoded rabies virus (RV) that can potentially be used to map the inputs to thousands of single neurons in parallel. Here, we address potential challenges that may arise when scaling barcoded RV to study brain-wide, densely labeled circuits (Aim 1), benchmark barcoded RV-inferred connectomes against other gold standard techniques (Aim 2), and develop new models using recurrent graph set transformers to reconstruct monosynaptic connectomes from barcoded viral datasets (Aim 3). By reducing the problem of synaptic connectivity into a problem of barcode sequencing, our approach has the potential to dramatically increase throughput, decrease costs and provide a direct link to the transcriptome of each mapped cell. 

As a proof of principle, we leverage barcoded RV to map the presynaptic inputs to thousands of single neurons in the mouse primary visual cortex (V1). The resulting cell type–resolved connectivity matrix confirms previously described motifs at the level of broad cell classes, and also reveals new circuit motifs at the level of finely resolved cell types. Using a combination of dissociated single-cell RNA-sequencing (scRNA-seq) and spatial transcriptomics, we identify both local and long-rangle inputs to mouse V1 neurons. Our approach provides a framework for scalable, sequencing-based connectomics in other brain regions and species.