Investigating the Control of Human Pluripotent Stem Cell Differentiation by Integrated BMP Signaling Dynamics

Abstract

Pluripotent stem cells generally differentiate to a heterogeneous mix of cell types in response to the uniform application of exogenous differentiation cues. Applications of stem cell technology to regenerative medicine, drug development, and disease modeling are impeded by uncontrolled cell fate heterogeneity and batch-to-batch variability with existing differentiation protocols. On the other hand, (organized) cell type heterogeneity is a desired feature of stem cell-based models of organ and embryo development. These systems are useful developmental models because they make use of mechanisms for self-organization used in development in vivo, notably regulation of paracrine signaling. However, we have an incomplete understanding of the cell signaling that takes place during differentiation in vivo and in vitro and how this signaling translates to cell fate. Recent work has shown that endogenous signaling patterns are not static; signaling varies both between cells and over time in a single cell. Cells may respond to different features of a dynamic signaling history such as the rate of signal change, raising the additional question of how the high dimensional space of possible signaling histories maps to cell fate. To understand the role of signaling in fate specification we must measure the history of signaling in individual cells and the fate of the same cells. To accomplish this, we developed a pipeline for time-lapse live-cell fluorescence microscopy followed by iterative immunofluorescence to follow signaling histories linked to cell fate in large numbers of human pluripotent stem cells (hPSCs). We apply this pipeline to understand how hPSCs interpret BMP signaling dynamics during differentiation in the absence of endogenous Wnt and Nodal signaling. We find that in disordered culture, BMP signaling in single cells is sigmoidal and varies primarily in the duration rather than amplitude of response. In this context the duration of signaling correlates more strongly with fate than the initial or final amplitude of signaling, but the time integral is more predictive of fate than any of these features. We find that time-integrated BMP signaling causally controls the decision to differentiate to amnion-like fate, and the signal level and duration control fate only by changing the signaling integral. This means that in this context, signal level and duration are interchangeable; a reduced signaling level can be compensated for with an increased duration and vice versa. In a stem cell model for patterning of the human embryo, we show that signaling histories predict the fate pattern and the integral model correctly predicts changes in cell fate domains when signaling is perturbed. We use bulk RNA sequencing and time series immunofluorescence to screen for genes that vary on the time scale of differentiation with a rate of change proportional to the BMP signaling level as potential BMP integrators. Among these genes, we identify the core pluripotency maintenance gene SOX2 as a strong candidate. We measure SOX2 dynamics at different BMP signaling levels in high temporal resolution with a hPSC line expressing a fluorescent fusion of GFP to SOX2 and develop a simple mathematical model explaining the regulation of SOX2 by BMP signaling and the control of amnion differentiation genes by BMP signaling and SOX2. Finally, we confirm that transient overexpression of SOX2 during BMP4-driven differentiation reduces amnion-like differentiation in agreement with our math model.