Forward and Reverse Engineering of Cellular Decision-Making

Cells reside in highly dynamic environments to which they must adapt. Throughout its lifetime, an individual cell receives numerous chemical and mechanical signals, communicated through dense molecular networks, and eliciting a diverse array of responses. A large number of these signals necessitate discrete, all-or-none responses. For instance, a cell receiving proliferation signals must respond by committing to the cell-cycle and dividing, or by not initiating the process at all; that is, the cell must not adopt an intermediate route. Analogously, a stem-cell receiving signals for different lineages must commit exclusively to one of these lineages. How individual cells integrate multiple, possibly conflicting, noisy inputs, and make discrete decisions is poorly understood. Detailed insight into cellular decision-making can enable cell-based therapies, shed light on diseases arising out of dysregulation of control, and suggest practical design strategies for implementing this behavior in synthetic systems for research and industrial use. In this thesis, we have employed both mathematical modeling and experiments to further elucidate the mechanistic underpinnings of decision-making in cells. First, we describe a computational study that assesses the entire space of minimal networks to identify topologies that can not only make decisions but can do so robustly in the dynamic and noisy cellular environment. Second, via model-driven, quantitative experiments in a megakaryocyte erythroid progenitor line, we demonstrate that a simple network with mutual antagonism and autoregulation captures the dynamics of the master transcription factors at the level of individual cells. Expansion of this model to account for extrinsic cues reconciles the competing stochastic and instructive theories of hematopoietic lineage commitment, and implicates cytokine receptors in broader regulatory roles. Third, to assess the impact of specific genetic perturbations on the distribution of the population, and on commitment trajectories of individual cells, we implemented the core mutual antagonism and autoregulation topology synthetically in yeast cells. Our approach of using orthogonal variants of a single core protein represents a general, modular design strategy for building synthetic circuits, and model-driven experiments elucidate how gene dosage, repression strength, and promoter architecture can modulate decision-making behavior.