Quantitative Methods for 3D Gastruloids Image Analysis

Abstract

The study of early developmental stages in humans was limited by technical difficulties and ethical concerns. As a consequence, the scientific community does not have an experimental system able to generate enough reliable data about human development. For this reason, new models need to be developed. Recently, multi lineage organoids were generated in vitro to model post-implantation events (~2 weeks post-fertilization). These 3D aggregates derived from human embryonic stem cells are called gastruloids because they mimic events of the early stages of gastrulation. Although the visualization of such samples produced informative results, the qualitative assessment of events remains a limitation intrinsic to the method. In this work, we describe a workflow useful to extract meaningful quantitative information from Z-stacks of human gastruloids. Performing nuclear segmentation with Cellpose, we identified 144,592 nuclei expressing FOXA2 and SOX17, two transcription factors involved in early development. Processing iteratively single optical sections of 23 gastruloids, we were able to measure markers intensity over time. Furthermore, combining tools in Fiji and custom functions in R, we recorded nuclear shape changes and cellular alignment events. The pipeline developed allows quantification of features which would have been only described by looking at the images. This study serves as a collection of open access tools for researchers to perform accurate image analysis from 3D digital images.