Development of bioimage analysis tools for the study of neurogenesis and neurodegeneration

Resumen

The bioimaging field has quickly evolved during the last years with the implementation of hardware and software for greater automation at the different stages that compose a microscopy-based experiment. The topmost representatives of these forefront methodologies are those combining high-content imaging with automated bioimage analysis protocols, although there is a whole spectrum of intermediate levels of throughput up to the manual analysis that is still practiced in many fields. Therefore, a big effort is still to be made in order to spread the use of cutting-edge imaging technologies throughout the life science fields relying on image-based experiments. Our aim has been to generate bioimage analysis workflows, in some cases including the development and optimization of the experimental and imaging setups, for the study of brain-specific processes in the field of regeneration. On one hand, we investigated the possibilities to adopt microscopy-based screenings combined with automated bioimage analysis for the in vitro study of the subependymal neural stem cells (NSCs). On the other hand, we looked at the possibility to semi-automate the transmission electron microscopy (TEM) analysis of in vivo experiments for the assessment of remyelination. We have shown that automated bioimage analysis protocols developed for the study of cultured NSCs improve the throughput and the reproducibility of the in vitro assays of different kinds and analytical complexity. Specifically, we have deployed high-content imaging and bioimage analysis protocols for i) static, cell-to-cell adhesion-based assays, ii) cell-based proliferation and apoptosis assays and iii) screening-like pseudo-clonal neurosphere formation assays. Emphasis has been made in the establishment of well-conducted pre-processing steps, which have involved the adoption of different illumination correction strategies or the search for more optimised protocols to generate rules for the detection of fields-of-view affected by blur. We also have shown that the combination of supervised machine learning with semiautomated segmentation of fibre cross-sections on TEM images contributes to enhance the analysis throughput of experiments aimed to quantify classic metrics just as the g-ratio, whereas it also empowers the inclusion of novel metrics by including inner-tongue-based features to assess new parameters.