Using multivariate live-cell imaging to build a cue-signal-response model of endometrial stromal cell decidualization

Introduction: The human endometrium undergoes continuous remodeling in response to fluctuating levels of ovarian hormones estradiol and progesterone. These cyclic changes prepare the endometrium for embryo implantation, but in its absence resolve with menstruation. Among the various cellular processes that occur throughout the menstrual cycle, decidualization of endometrial stromal cells (ESCs) is paramount for normal endometrial and reproductive physiology. Decidualization occurs over several days and involves the transformation of ESCs from a mesenchymal phenotype to a more epithelial-like secretory cell that regulates epithelial and immune functions. However, compromised decidualization, sometimes referred to as progesterone resistance, is a hallmark of different gynecological disorders such as endometriosis. While decidualization is triggered by rising levels of circulating progesterone and intracellular spikes of protein kinase A (PKA) activity, this differentiation process involves multiple other signaling pathways and autocrine cascades that dictate the success of the decidualization program. Yet, to date, most of these pathways have been studied in isolation and with low temporal resolution, which poses a significant barrier to understanding the complex multipathway regulation that underlies decidualization.

Materials and

Methods: We propose using a data-driven systems biology approach that leverages multivariate measurements and a diverse set of experimental conditions to bridge this gap. Specifically, by measuring signaling dynamics obtained across several experimental perturbations, we will build mathematical models that link signaling events to cellular responses (i.e. decidualization). The large number of combinatorial cues (e.g. hormones, cytokines) that can modulate decidualization necessitates high-throughput experimental approaches to acquire sufficient data for training a model. Therefore, we have built a panel of genetically engineered reporter cells to monitor the activity of several critical signaling pathways and cell phenotypes that enabled us to build a cue-signal-response model of decidualization based on functional regression analysis. We then performed feature selection to determine a set of key time points and signaling features to be measured in primary cells using antibody-based assays that will allow us to apply our reporter-based model to these patient-derived ESCs.

Results, Conclusions, and Discussions: Our computational model captures the quantitative relation between the dynamic and coordinated role of multiple signaling pathways and the success of the decidualization program in engineered cell lines. Moreover, our transfer learning approach based on key features of the computational model allows us to make quantitative predictions about primary ESCs to evaluate potential therapeutic targets to restore progesterone responsiveness. We anticipate that our approach will be useful to advance our understanding of decidualization as a complex process. Furthermore, we envision that applying our model to patient-derived cells will be a useful diagnostic tool to determine potential interventions to remediate impaired decidualization in patients.

Acknowledgements (Optional):