Computational Modelling of Spinal Control of Larval Zebrafish Swimming Speeds

Abstract

The zebrafish (Danio rerio) is a commonly used animal model for biological study due to its available transgenic lines, fast life cycle, transparency, and relative simplicity. Zebrafish are also a great candidate to study locomotion, as most of the neuronal populations responsible for locomotion in zebrafish are evolutionarily conserved, including in mammals. From very early stages, embryonic and larval zebrafish display a wide range of swim-like behaviour following the neuronal development of their spinal cord. During the larval stages, they display a beat and glide swimming pattern consisting of frequent tail beats at various tail beat frequencies (TBFs) followed by a quiescent period. The literature strongly supports the idea that there are spinal cord speed microcircuits behind these different TBFs. However, the structure and mechanism of operation of these microcircuits are not well understood. Computational models are excellent tools for studying model animals as they enable tests which are not possible with the current experimental techniques. Previously, our lab has created computational models of various forms of swim-like behaviours of embryonic and larval zebrafish, displaying how different cellular populations interact to create the behavioural repository of the early stages of zebrafish development. This study aims to simulate the activation of speed microcircuits that can generate different speeds of the beat and glide swimming of larval stages. Based on the information presented in the literature, I hypothesize that there are speed microcircuits in the larval zebrafish spinal cord that are differentially recruited by the supraspinal regions and generate different tail beat frequencies. Different intrinsic properties, spatial distributions, projection lengths, patterns, and synaptic strengths are the building blocks of these microcircuits. There are currently close to twenty cell subgroups known to be differentially recruited at different swimming speeds. To be able to generate a model handling such complexity, a software tool, SiliFish, was implemented to create and test computational models of spinal control of swimming behaviour easily and quickly. Using SiliFish, a network of 440 cells and more than 8K synapses were created that can generate different ranges of tail beat frequencies, recruiting different cellular groups. The model is able to generate tail beats at different TBFs and replicates several features of spinal circuits observed experimentally, as reported in the literature.