Automated synapse-level reconstruction of neural circuits in the larval zebrafish brain

Abstract

Dense reconstruction of synaptic connectivity requires high-resolution electron microscopy images of entire brains and tools to efficiently trace neuronal wires across the volume. To generate such a resource, we sectioned and imaged a larval zebrafish brain by serial block-face electron microscopy at a voxel size of 14 × 14 × 25 nm3. We segmented the resulting dataset with the flood-filling network algorithm, automated the detection of chemical synapses and validated the results by comparisons to transmission electron microscopic images and light-microscopic reconstructions. Neurons and their connections are stored in the form of a queryable and expandable digital address book. We reconstructed a network of 208 neurons involved in visual motion processing, most of them located in the pretectum, which had been functionally characterized in the same specimen by two-photon calcium imaging. Moreover, we mapped all 407 presynaptic and postsynaptic partners of two superficial interneurons in the tectum. The resource developed here serves as a foundation for synaptic-resolution circuit analyses in the zebrafish nervous system.

A single-cell resolution gene expression atlas of the larval zebrafish brain

“…The mapzebrain atlas thus offers a platform for covisualization of gene expression and cellular architecture and a starting point for explorations of circuit function and structure. Future work will be directed at integrating brain-wide functional imaging data (e.g., from calcium or voltage recordings) and electron microscopic reconstructions ( 58 ) into the atlas.…”

Recurrent network interactions explain tectal response variability and experience-dependent behavior

“…Future extensions to the model, incorporating known heterogeneities in neurotransmission, morphology and intrinsic properties and with many more free parameters, should expand the accuracy and breadth of the model assuming such additions are adequately constrained by functional and/or anatomical data. In this regard, recent studies in zebrafish have characterised various aspects of molecular-genetic, morphological, and physiological diversity in OT ( Helmbrecht et al, 2018 ; Antinucci et al, 2019 ; Shainer et al, 2022 ) and high-quality ultrastructural data is now available, allowing definitive reconstruction of synaptic connectivity ( Svara et al, 2022 ). The model can also be extended to incorporate interconnections between OT and many other brain regions (e.g.…”

SyConn2: dense synaptic connectivity inference for volume electron microscopy

“…In a first step, a contact site instance segmentation was generated by iterating over the cell segmentation and storing adjacent supervoxel IDs. At every boundary voxel (6-connectivity) of the cell segmentation, a partner cell ID was identified by finding the majority ID within a window of [7,13] voxels (voxel size 10, 10, 25 nm). If a majority ID was found (background and the source boundary voxel ID were excluded), the contact site voxel was assigned a value that allowed the retrieval of the two partner cells (bit shift combination to uint64 in case of uint32 cell segmentation; tuple of uint64 in case of uint64 cell segmentation).…”

“…Despite these increases in acquisition speed and considerable advances in areas such as automated neuron reconstruction 4 , proofreading 5 , synapse and organelle detection 6,7 , cell type classification 8,9 and integrative processing in cloud environments 10,11 , a pipeline that creates an annotated connectome and can also be operated cost-efficiently on existing high-performance computing infrastructure is lacking. Here we introduce SyConn2, which requires existing dense neuron reconstructions and fundamentally upgrades our earlier software package 7 (see Supplementary Table 1 for a comparison) to allow neuroscientists to run queries against connectomes with millions of synapses 12,13 . To be able to handle the large amounts of data at reasonable cost, we focused on computationally efficient processing at every step, for example by operating on lightweight point cloud representations instead of dense data structures to analyze neuron morphology.…”