Revisiting the erythroid-myeloid lineage decision - a data-driven dynamical model analysis
It is widely conjectured that the mutually antagonizing pair of transcription factors GATA1 and PU.1, deter-mines the choice between the erythroid and myeloid lineages in hematopoiesis. In theoretical approaches, this appears natural with a bistable switch driving the decision. Recent extensive binding and gene expression experiments with some focus on the triad GATA1, GATA2 and PU.1 indicate that GATA2 may be more involved in this lineage decision than previously anticipated. Here, we analyze these experimental data by modeling regulatory sub-networks with deterministic rate equations. Using network dynamical parameters determined by the data, we deduce from increasing the self-interaction bindings in silico among the triad genes that GATA2 and PU.1 exhibit non-linear behavior with one unstable and one stable state. This is in contrast to GATA1, which shows smoother behavior. We extend the network to include the downstream regulators FOG1 and CEBPA, and extract the nature of the corresponding regulatory interactions, excitatory or suppressing, between this pair and the triad by fitting to experimental gene expression time series. Based on this extended network, we simulate and explore different knockout scenarios, providing insight into the role of these regulators in the process of lineage specification, as well as predictions for future experimental validation. We address the mechanism of GATA switching as a mechanism of lineage differentiation by investigating the dynamics of FOG1 regulation by GATA2 and GATA1. Overall, this analysis strongly suggests that within this network, GATA2 is the key driver of erythroid lineage specification through its repression of PU.1, whereas GATA1 appears to be more relevant for the downstream differentiation events.