Optogenetic perturbations to unveil brain-scale functional organization in zebrafish larva

In spite of their apparent stochasticity, neuronal circuits display a highly structured spontaneous activity even in the absence of sensory stimulation. Large-scale recordings have shown that neuronal network dynamics is characterized by the coordinated activity of so-called cell assemblies, which can be identified using dimensionality reduction techniques. This complex activity is well modelled by attractor states associated to specific behavioral patterns. Understanding the computational origin of these states and the mechanisms underlying spontaneous transitions between them is a longstanding major issue in systems neuroscience. As part of my M2 internship project, I developed an experimental approach in order to study evoked neuronal activity in the vertebrate brain. I used a recently developed platform in my laboratory enabling both local perturbation of neuronal activity with one-photon optogenetics and whole-brain imaging with light-sheet microscopy. These methods allowed me to induce brain-scale perturbations in the zebrafish larva brain, a widely used vertebrate model organism in neuroscience, and concurrently record the activity of almost all the 100,000 neurons constituting its brain. Perturbing brain activity is of major importance for testing causal interactions in neuronal networks, understand the contribution of small neuronal populations and establish a functional connectivity between different regions. I developped a stimulation protocol testing the effect of different stimulation intensities and durations over the whole brain activity. Brain-wide responses were specific of the stimulation region. The computation of correlation between the activities of different regions enabled me to establish a brain-scale functional connectivity map. Besides allowing an unprecedented exploration of the zebrafish brain dynamics at all spatial scales, these results will allow to directly connect neurophysiological circuit properties with a recently probabilistic graphical model developed in collaboration with my laboratory, called compositional Restricted Boltzmann Machines, aiming at capturing neuronal activity statistics in a physiologically interpretable way. This in silico model will allow to predict perturbations of neuronal activity at whole brain scale and will sharply contrast to currently used blackbox-like machine learning techniques.