RECURRENT NETWORKS ACROSS INTER- AND INTRA- REGIONS IN THE WHOLE BRAIN VERTEBRATE CONNECTOME

Increasing evidence highlights the importance of information processing with long time constants, or attractor states, which are thought to underlie cognitive functions, such as integration, working memory, decision-making, and prediction. In the vertebrate brain, recurrent networks are observed locally and globally by labeling subsets of neurons, genetically and chemically. However, characterizing synaptic-level recurrence across single neurons and populations remains challenging, due to technical limitations in achieving whole-brain, high-resolution reconstructions in the big brains.
Zebrafish larvae already lend themselves to neuroscience research, for their transparency, small size of the body and small number of neurons in the brain. In our dataset, Fish 2.0³, a single fish was exposed to a repertoire of visual stimuli while neural functional recording and muscle output recording were performed, followed by whole-brain reconstruction using electron microscopy. Extensive automated segmentation and manual proofreading enabled synapse-level annotation across the entire nervous system.
Using this dataset, we identified recurrent circuits at multiple scales, including both local intra-regional and long-range inter-regional loops. At the local level, we uncovered recurrent motifs within descending motor command circuits in the hindbrain, and show how these loops may contribute to the coordination of motor outputs using matched functional recordings. At the global level, we identified long-range recurrent connections between the preglomerular complex in the diencephalon and the pallium in the telencephalon, which might represent a teleost homolog of mammalian thalamocortical circuits. Our results provide new insights into how recurrent circuit architectures support the emergence of long time constants in sensorimotor integration and cognitive processing.