Cellular decision making: from phage lambda to stem cells

Cellular decision making is the process by which cells choose among functionally-distinct cell states. Heritable cell states are typically maintained and stabilized by the activity of specific genes, but cells can also be induced to switch to alternative states given the appropriate stimulus. Underlying decision making processes that result in different cell states are temporally regulated gene expression cascades. The decision-making process for switching between cell states can be biased by environmental factors, or can be driven solely by biochemical noise due to the stochastic nature of the cell. The inherent stochastic nature of biochemical reactions in the cell has been highlighted by recent quantitative single-cell measurements. When examined at the single-cell level, the decision-making process often appears noisy, where individual cells choose different cell states even when subject to identical conditions. The mixed outcomes of these decisions have been used to demonstrate that molecular noise can dominate whole-cell processes. However, there may also exist previously unaccounted-for cell parameters that affect the decision making process, making the decision appear more random than it really is. Additionally, the maintenance of a heritable cell state is also subject to the stochastic nature of gene expression. For example, the gene expression programs associated with stabilized cell states often contain a self-regulating protein. However, characterization of the effect to which fluctuations in gene expression of a fate-determining protein modulate the stability of the cell state has not been accomplished. Questions about the effect of the stochastic nature of gene expression on decision making include: In the face of gene expression stochasticity, can decision-making processes appear more precise when the proper variables are taken into account? Does the level of gene expression noise dictate the stability of a gene expression state? To answer questions such as these, we investigate two systems that exhibit cellular decision making and cell-state maintenance, bacteriophage lambda and mouse embryonic stem cells. Bacteriophage lambda (phage lambda) is a bacterial virus that, upon infection of its host bacterium, Escherichia coli, decides between two alternative pathways: The phage can replicate and kill the host cell, or it can integrate into the host chromosome and passively replicate as part of the host. This integrated phage can spontaneously switch to replicate and kill the host either by random chance or induction by specific stimuli. We investigated this decision-making process of phage lambda via microscopy, at single-cell and single-phage resolution. We observed that the decision-making process is first made at the level of individual phages, and then integrated into a whole-cell decision. Additionally, we investigated the stability of the integrated phage in the replicating host. With single-molecule resolution measurements of gene activity and the measurements of cell-state switching rates, we were able to determine the relationship between stochastic gene activity and cell-state stability. In order to extend these techniques to a higher system, we chose to study mouse embryonic stem cells, which are often used because they closely resemble human biology. Embryonic stem cells are extracted from the developing embryo and can be maintained in vitro indefinitely while still remaining pluripotent. Pluripotency is the ability to assume any cell state in the adult body and is the hallmark of embryonic stem cells. The molecular mechanisms for the stability of pluripotency have been narrowed to three fate-determining proteins. Two of these proteins are thought to be tightly regulated, Oct4 and Sox2, while the third, Nanog, exhibits large variability among the population. Additionally, the level of Nanog has been correlated with the stability of pluripotency. The reasons for the variability in Nanog level are not known, but stochastic gene expression has been hypothesized as a possible source. We measured the gene activity of Oct4 and Nanog and found that while Nanog did exhibit a higher degree of heterogeneity at the mRNA level, both genes exhibited intermittent transcription activity. Additionally, when we used phenomenological models to extract the kinetics of transcription, we found that the cause of Nanog’s higher heterogeneity was due to a slower rate of transcriptional activation. Our experiments demonstrate that high-resolution measurements paired with modeling of stochastic processes is a powerful approach for studying cellular decision making. The techniques developed here allow for better resolution of the precision of cellular decision making by accounting for sources of measurement noise. Our techniques also give us the ability to connect the stochastic events of gene expression to the whole-cell phenotypes of cell-state stability