LocoMat

Most animals are born with a minimal set of behaviors that is progressively enriched during development to accommodate new internal needs, physical and environmental constraints. The emergence of these new functional capacities reflects the concurrent maturation of the brain circuitry, as new neurons become functional and new connections are formed, but also results from changes in the synaptic strengths that modify the computational capacity of given networks. How innate developmental programs and on-going interactions with the environment concur to the self-organization of extended functional circuits remains an open question. In the present project, we will combine genetic developments, behavioral and functional monitoring, and data-driven graphical models to unveil fundamental mechanisms that orchestrate the maturation of brain-wide neural circuits leading to the progressive emergence of new computational capabilities and behaviors during early development. To address this question, we will leverage the assets of Danionella translucida (DT), a novel vertebrate model whose brain remains small and transparent up to the adult stage, and is thus amenable to large-scale monitoring of brain activity during a significant development period. The phylogenetic and anatomical proximity of DT with Zebrafish - a well-established model vertebrate - will facilitate the identification of functional neuronal circuits and allow for fruitful cross-species comparison. We will focus on the maturation of the neuronal circuits that organize spontaneous and light-driven exploration. We will first quantify how this behavior evolves from the larval to the juvenile stage using freely-swimming behavioral assays. We will characterize in particular the phenotypic transition, from continuous to bout-like navigation, that takes place during this developmental period. We will then use whole-brain calcium imaging and optogenetic activation to probe the collective dynamics of the neuronal circuits engaged in this process in partially tethered, fictively swimming animals across the same age range. We will use morphological registration between individuals of same age and across development to produce generic coarse-grained brain-scale dynamics, at successive stages of development. This unique dataset, that will include whole-brain neural activity, sensory and motor signals, will be used to train energy-based graphical models. Such models have generative powers: we will produce synthetic activity and fictive motor signals that will recapitulate in silico the brain-scale neural dynamics and associated sensory-motor computations. By training these models at various developmental stages, and comparing the inferred parameters, we aim to illuminate basic developmental rules that could account for the observed changes in behavior. The project will be carried out by a consortium of three teams, with highly complementary expertise. Filippo Del Bene is a biologist who has made key contributions to the understanding of molecular processes that control the development of neural circuits in the larval zebrafish brain. He has played a major role in establishing Danionella Translucida as an important emerging model system in neuroscience. Georges Debrégeas is an experimental physicist, who develops state-of-the art optical systems for whole-brain imaging in zebrafish larvae, high-throughput behavioral assays and who develops efficient data processing methods to analyze these high-dimensional datasets. Rémi Monasson is a theoretical physicist who applies statistical-physics and machine-learning based approaches to biological data analysis and modelling, particularly in neuroscience.