Brain morphogenesis is an extremely complex process in which a large variety of different cellular processes take place. Although most neurons in the central nervous system are born during the embryonic and early postnatal periods, it is now well accepted that some regions of the brain keep producing new neurons throughout life. Understanding the molecular interactions that govern this process might provide key insights into future therapeutic treatments to a variety of pathological and neurodegenerative conditions. It is therefore of great interest to understand the processes that govern brain morphogenesis at the level of the cell or of the organ. Recent advances in light microscopy allowing neuroscientists to study the dynamics of living cells have changed our understanding of brain morphogenesis in profound ways. For instance, in cultured cells, studies into the dynamics of neurite outgrowth have given us insights into the differences between how axons and dendrites develop. Time-lapse images of the adult cortex of live animals have challenged previous models of brain plasticity, revealing that upon sensory experience, dendritic spines appear and disappear frequently, without observing many changes in dendritic and axonal processes, contrary to the previous hypothesis. However, even though it is now possible to efficiently image large volumes of dynamic processes in living cells and tissues, research is still limited by a lack of tools to analyze the massive datasets that are typically generated in such experiments. In most cases, biologically relevant features must be extracted by hand or through simple image processing techniques, limiting research to anecdotal rather than statistical evidence. In the worst case, important behaviors which cannot be discerned by human observers are overlooked. Thus, there is a great need for generalized computational methods to perform automatic, unbiased quantitative analyses of dynamic microscopic images. In this interdisciplinary proposal, two “wetlab” biology groups and two computational groups specializing in computer vision and machine learning will team up to develop such methods, as depicted in Fig. 1, which will provide a means to gain new insights into neurobiological problems that cannot be studied by conventional approaches.
Specifically, we propose to:
1.develop generalized computer vision methods to detect, segment, and track neurons and other types of cells in 2-D and 3-D time-lapse data sets, and extract meaningful morphodynamic signatures;
2.develop machine learning techniques to infer relationships between these morphodynamic signatures and various experimental factors regulating brain morphogenesis.
We will then apply these techniques to:
a.infer the network of signaling events regulate neurite outgrowth at a cellular level;
b.study a complex neuronal cell migration event in the olfactory bulb in a ex-vivo tissue.
a. While most of the structural and signaling components regulating neurite outgrowth are known, a holistic, systems biology view of how these components are wired is still lacking. Using a catalogue of 250 neurite-localized proteins as a template, we will systematically investigate the roles of these proteins using siRNA-mediated knockdown and time-lapse analysis in a cell culture assay. The developed computer vision algorithms will segment the neurite structures and extract morphodynamic signatures of the outgrowth process. Machine learning techniques will extract relevant information from these signatures and cluster this information from different gene perturbation experiments to infer functional interactions between different proteins. This will enable to these proteins are wired in signaling networks to regulate this complex cellular behavior. The predicted network will be validated using classical biochemical and cell biological techniques.
b. We will also study the migration of GFP-tagged newborn neurons in the adult olfactory bulb using 2-photon microscopy in their natural brain network using ex-vivo acute brain slices. This complex morphogenetic event involves two different cell migration events (tangential and radial) that respond to distinct extracellular cues and uses different cellular morphologies. Computer vision analysis will allow to segment and analyze multiple neurons that simultaneously migrate in a 3D tissue environment, to describe the morphological features associated with the two cell migration events, and how these are modified in response to pharmacological or genetic perturbations. In short, we expect that synergistic efforts between the computational and “wetlab” groups will produce powerful, efficient algorithms that will be available to the science community in Switzerland and abroad, and provide a better understanding of the processes that govern brain morphogenesis.