Study of the regulatory programs controlling pluripotency and differentiation in Embryonic Stem Cells

Pluripotent stem cells (PSCs) are defined by two core properties: pluripotency, the ability to differentiate into all cell types, and self-renewal, the capacity to proliferate indefinitely in vitro while maintaining an undifferentiated state. These features enable their use in disease modeling, drug testing, and the production of healthy cells for transplantation. This dissertation addresses how PSCs transition to a differentiated state, focusing on two open questions: (i) how gene regulatory network (GRN) perturbations reshape differentiation trajectories, and (ii) how exit from pluripotency can be characterized by examining the role of population-level, intrinsic cellular, and genealogical properties. To address the first question, we developed IGNITE (Inference of Gene Networks using Inverse kinetic Theory and Experiments), an unsupervised machine learning framework that infers directed, weighted, and signed GRNs directly from unperturbed single-cell RNA sequencing (scRNA-seq) data. It is based on the inverse problem for a kinetic Ising model, a statistical physics framework previously applied to biological network studies. By reconstructing effective networks, IGNITE reproduces wild-type expression patterns and generates perturbations, predicting how single or multiple gene knockouts reshape cellular phenotypes. Unlike previous approaches, it does not require perturbation data or prior knowledge for training. We benchmarked IGNITE using two distinct PSC systems: mouse PSCs transitioning from the naïve to formative state, and human PSCs differentiating toward definitive endoderm. This allowed us to test the method across different species, developmental trajectories, and single-cell platforms (10X vs. Fluidigm C1), evaluating the method’s performance and generalizability. IGNITE aligned with independent experimental findings, and outperformed current gold-standard methods in predictive accuracy. To address the second question, we analyzed long-term live-cell imaging of mouse Rex1-GFPd2 embryonic stem cells (ESCs) undergoing differentiation, to investigate lineage relationships and features of exit from naïve pluripotency. Because exit occurred asynchronously at the population level, we tested whether this heterogeneity could reflect stable inherited cellular properties. Lineage analysis revealed transient genealogical correlations in exit timing, which dissipated across generations and were absent among unrelated cells. To characterize the exit process, we defined exit operationally as a threshold-crossing event in total Rex1 fluorescence. According to this criterion, exit typically followed a variable lag period, occurred independently of cell-cycle phase or colony size, was not predetermined by founder-cell fluorescence, and proceeded symmetrically between sisters. Together, these two approaches provide complementary insights into exit from naïve pluripotency across molecular and cellular scales, offering a framework to link regulatory networks with single-cell and lineage behaviors in stem cell fate decisions.