Establishing high-content and machine learning-enhanced microscopy workflows to investigate spontaneous cell division defects in human Pluripotent Stem Cells and their potential role in tumourigenic potential.

Abstract

Human Pluripotent Stem Cells (hPSCs) are a breakthrough biomedical technology that provide a powerful tool for regenerative medicine. However, evidence of poorly understood tumourigenic potential in hPSCs restricts their therapeutic use. There is emerging evidence that karyotypic instability may contribute to tumourigenic potential in hPSCs. This thesis describes the development and use of a multi-day, multi-colour, timelapse high-content microscopy workflow in studying the nature and incidence of spontaneous chromosome segregation errors (CSEs) in dual-reporter hPSCs and their potential contribution to tumourigenic potential. Specifically, we visually screened 49,505 mitoses in hPSCs and a further 51,200 mitoses across varying early differentiating cells, from which we extracted qualitative and quantitative information relating to the causes, characteristics, and consequences of CSEs. To our knowledge, this constitutes the most comprehensive survey of spontaneous CSEs in normal human cells undertaken to date. I demonstrate that hPSCs consistently display lagging chromosomes, chromosomal bridges, and multipolar mitoses and that in the majority of cases these CSEs are non-lethal to cells which then produce apparently viable progeny with unperturbed cell cycle dynamics, indicating that the consequences of CSEs go undetected by cell cycle checkpoints and can propagate invisibly. I show that these CSEs consistently generate micronuclei, which may pose a chromothriptic tumourigenic risk, according to evidence in cancer cells. I found that CSEs persist and even amplify throughout the early differentiation of hPSCs. This finding indicates that a potential CSE-mediated window of tumourigenic potential extends beyond pluripotency, which has important consequences for the safe design of hPSC-derived tissues destined for therapeutic use. Finally, I describe the development and continued improvement of a deep-learning enhanced automated detection pipeline for the automated scoring of CSEs. This pipeline could be used, in future, to undertake large-scale screens to identify gene-disruptions or chemical interventions that could suppress tumourigenic potential in cell engineering applications.

Date of Award	3 Oct 2023
Original language	English
Awarding Institution	University of Bristol
Sponsors	Astra Zeneca Pharmaceuticals
Supervisor	Rafael E Carazo Salas (Supervisor)