(200a) Cell Cycle Transition in Embryonic Stem Cells during Self-Renewal and Differentiation

Embryonic stem cells are known to have a unique cell cycle, with a shortened G1 phase resulting in shorter doubling time. This G1 phase lengthens with differentiation, leading to an overall longer doubling time. A long-standing question in stem cell differentiation is the coupling of the cell cycle with differentiation, which still remains unanswered. Extensive experimental studies have offered interesting insights, in particular the establishment of the FUCCI reporter in human embryonic stem cells (hESCs). However, mechanistic information for this complex, non-linear interaction would be difficult to obtain with a purely experimental approach. Coupling experiments with mathematical modeling can aid in obtaining a more comprehensive understanding of the system. Our objective is to represent these unique dynamics of cell cycle transition in hESC using an integrated experimental and computational approach.

An important component in the analysis of this system is the heterogeneity in cell population and cell cycle state of each individual cell. Hence, we start off with a stochastic cellular automaton model to describe the behavior of the cells at the population level along with its associated heterogeneity. In this model, we track individual cells and their progress through the cell cycle, with the state of the cell being defined by the cell cycle phase which the cell is in (G1, S, or G2/M). The residence times for phases are described by probability distributions and their moments (e.g. normal distribution described by its mean and standard deviation). Individual cells are tracked in time, thereby simulating the dynamics of the cell population. Information from these dynamics can be obtained by comparing simulations to experimentally synchronized cells, achieved by arresting the cellular population, either hESC or hESC-derived pancreatic progenitor cells, in the G2/M phase with nocodazole.

Our population model, when applied to a self-renewing population of human pluripotent stem cells, is able to accurately predict phase resident time distributions from experimentally synchronized hESC. Interestingly, only one combination of specific distributions and their moments, representing phase residence times, were able to recapitulate the experimental data. These results were validated by a completely separate analysis, showing the accuracy of the modeling approach. In the next step this analysis was extended to explain the desynchronization dynamics of pancreatic progenitor phenotype. Fourier analysis predicts multiple subpopulations contributing to cell cycle desynchronization, and it was necessary to break the dynamics into these two subpopulations to capture the experimental behavior. The predicted phase distributions show that, in this more mature pancreatic progenitor phenotype, the G1 phases become both longer and more variable (as compared to undifferentiated cells).

To mechanistically explain this lengthening of G1 phase, we developed an ODE model of G1-S transition. We developed a skeletal model including key proteins reported to drastically change during differentiation: cyclin D, cyclin E, and p21 (CIP). Incorporation of these equation-based models in the automaton model results in a hybrid framework, where the G1 transition is governed purely by equations, while the S and G2/M phases are driven by the aforementioned probability distributions. The hybrid model is able to capture the overall trends of cell cycle transition and change in G1 population percentage during induced differentiation, in addition to the dependency on the differentiation molecule concentration. Furthermore, pertinent features of this transition, including dependency of lengthening on time and cell cycle position, probability of lengthening and maximum G1 residence time, were predicted.

Together, the developed single cell and stochastic population models can describe cell cycle behavior both at various stages of maturation as well as during differentiation. The population model and synchronization experiments are able to extract single cell information from simple population analyses, which can be mechanistically explained by the G1-S ODE model. By integrating the single cell model into the individual automatons of the population model, a more comprehensive understanding of the ESC cycle transition during differentiation is obtained.