Learning from Cell-to-Cell Heterogeneity through Computer Vision: Fully Automated Image Analysis Identifies Relationships among Growth Cone Morphology, Rho GTPase Activation Motifs, and Neurite Elongation
Bagonis, Maria M.
MetadataShow full item record
CitationBagonis, Maria M. 2017. Learning from Cell-to-Cell Heterogeneity through Computer Vision: Fully Automated Image Analysis Identifies Relationships among Growth Cone Morphology, Rho GTPase Activation Motifs, and Neurite Elongation. Doctoral dissertation, Harvard University, Graduate School of Arts & Sciences.
AbstractThe growth cone, residing at the helm of elongating neurites, is characterized by the superposition of two geometrically distinct actin networks regulated by RhoGTPase signaling patterns. This complex, dynamic structure, plays an essential role in neuronal development, integrating a variety of extrinsic environmental cues, which serve to both guide the elongating neurite along its route, and ensure the appropriate temporal coordination of many simultaneous neurite elongation events. While much progress has been made in the identification of the molecular players important for neurite elongation and guidance, much less is known about how intra-cellular signaling events coordinate in space and time to control growth cone morphology and function. High resolution imaging of growth cones expressing fluorescent biosensors provide a means to monitor fundamental signaling nodes and directly measure cellular signaling relationships in real time. However, a dearth of automated tools for the extraction of biosensor measurements remains a primary bottleneck for application of this technology to higher throughput studies. Here we remove this bottleneck by developing Growth Cone Analyzer, a computer vision software tool that fully automates the extraction of fluorescent signal in relationship to biologically relevant morphological/dynamic landmarks, from growth cones exhibiting select behaviors. We showcase the tool by focusing first upon the morphological analysis of a structurally heterogeneous set of growth cones. Using these data, we demonstrate that morphodynamic features extracted using Growth Cone Analyzer can be employed to 1) identify latent correlations between growth cone architecture and neurite elongation rate, 2) distinguish subtleties in growth cone morphology/dynamics upon perturbation of seemingly redundant regulators of the RhoGTPase signaling network, and 3) detect the timing of significant transitions in growth cone morphology as a neurite progresses along its trajectory. We then illustrate how Growth Cone Analyzer’s utility may be extended beyond morphological studies to facilitate the extraction of high-fidelity spatial profiles of RhoGTPase activation patterns measured via FRET biosensors. Hence these tools provide a vital first step towards future studies aiming to quantify the RhoGTPase signaling patterns regulating neurite elongation and determine how specific RhoGTPase regulators may cooperate to maintain these motifs.
Citable link to this pagehttp://nrs.harvard.edu/urn-3:HUL.InstRepos:41140211
- FAS Theses and Dissertations