Deciphering cell fate: a mathematical model of life and death decisions in erythroid progenitors

May 2024, Model of the Month by Dr Lorenz Adlung, Marcel Schilling’s Lab, DKFZ, Germany

Introduction

Every second you are reading this (actually every second of your life), two to three million new red blood cells are being produced in your body (1). The key players in the production of red blood cells are called erythroid progenitors. They operate under a tightly regulated network of signalling pathways in which the JAK2/STAT5 axis plays a central role. Our understanding of erythroid progenitors has been primarily at the population level, but this perspective fails to capture the intricate cell-to-cell variability that is fundamental to individual cell fate decisions.

The variability in erythroid fate decisions is a critical factor influencing the efficacy of therapeutic interventions in diseases such as polycythemia vera and other myeloproliferative neoplasms. Current models, while valuable, are limited in their ability to analyse the nuances of individual cellular responses within a heterogeneous population. This challenge is compounded by the dynamic and stochastic nature of cellular signalling, where even genetically identical cells can exhibit markedly different behaviours under similar conditions.

Our research aimed to bridge this gap by developing a comprehensive mathematical model of the JAK2/STAT5 signalling pathway involved in cell survival. By exploring the variability in signalling, we hoped to gain new insights into how individual erythroid progenitor cells make the critical choice between life (proliferation/differentiation) and death (apoptosis).

Model

Complementing our previous publication detailing the nuanced regulation of cell proliferation by the AKT and ERK pathways (2), this model extends our understanding of the regulatory mechanisms involved in cell fate decisions. Our model combines coupled ordinary differential equations (ODEs) and nonlinear mixed effects modelling of the JAK2/STAT5 pathway (Figure 1).

By integrating different types of biological data, from targeted proteomics to flow cytometry measurements, the model provides complex insights into the biological system at the population level and with respect to cell-to-cell variability (3). We believe that our approach is applicable beyond the JAK2/STAT5 pathway and erythroid cells.

Results

With our calibrated model, we were able to fit the dynamics of STAT5 activation (Figure 2). We then predicted a critical threshold for activated STAT5, identifying a specific number of molecules in the nucleus of an (erythroid progenitor) cell that is required to initiate an anti-apoptotic gene expression programme. This threshold, estimated to be between 24 and 118 molecules, is critical in determining whether a cell will survive or undergo apoptosis.

We have validated our model predictions using a range of experimental techniques, from Western blotting, which helps to determine protein levels, to confocal fluorescence microscopy, which provides insights into the spatial distribution of signalling molecules within the cytoplasm or nucleus of individual cells.

In addition, our model allows us to predict how external stimuli would affect the signalling dynamics within erythroid progenitor cells. This aspect of the model opens up the possibility of exploring how external factors, such as drug treatments or changes in the cellular microenvironment, could influence cell fate decisions, providing a valuable tool for therapeutic intervention in erythropoiesis.

Discussion

Our approach focuses on the protein level. By assessing cell-to-cell variability using flow cytometry, we have gained valuable insights into the dynamic range of protein abundance and its impact on cell fate decisions in erythroid progenitors. This approach allows us to capture the real-time dynamics of protein translocation and signalling pathways, providing a more accurate representation of cellular processes as they occur in living cells.

While our current approach provides robust insights at the protein level, complementing it with mRNA data could provide a more comprehensive view of the cellular machinery. Longitudinal single-cell mRNA sequencing would allow us to track changes in mRNA expression over time, adding a temporal dimension to our understanding of cell fate decisions. We have recently explored such an approach in the context of inflammatory bowel disease (4).

In summary, our approach combines protein-level analysis with advanced computational modelling. Looking ahead, the integration of mRNA data through cutting-edge technologies such as single-cell mRNA sequencing and spatial transcriptomics promises to further revolutionise our understanding of cellular dynamics and disease mechanisms to finally answer critical questions about life and death at the molecular level of individual cells.

References

1. Blood cell turnover in human weighing 70kg - Human Homo sapiens - BNID 106129. https://bionumbers.hms.harvard.edu/bionumber.aspx?id=106129&ver=4&trm=turnover+red+blood+cells+human&org=.

2. L. Adlung, S. Kar, M.-C. Wagner, B. She, S. Chakraborty, J. Bao, S. Lattermann, M. Boerries, H. Busch, P. Wuchter, A. D. Ho, J. Timmer, M. Schilling, T. Höfer, U. Klingmüller, Protein abundance of AKT and ERK pathway components governs cell type-specific regulation of proliferation. Mol Syst Biol 13, 904 (2017).

3. L. Adlung, P. Stapor, C. Tönsing, L. Schmiester, L. E. Schwarzmüller, L. Postawa, D. Wang, J. Timmer, U. Klingmüller, J. Hasenauer, M. Schilling, Cell-to-cell variability in JAK2/STAT5 pathway components and cytoplasmic volumes defines survival threshold in erythroid progenitor cells. Cell Rep 36 (2021).

4. C. Kilian, H. Ulrich, V. Zouboulis, P. Sprezyna, J. Schreiber, T. Landsberger, M. Büttner, M. Biton, E. J. Villablanca, S. Huber, L. Adlung, Longitudinal single-cell data informs deterministic modelling of inflammatory bowel disease. bioRxiv [Preprint] (2023). https://doi.org/10.1101/2023.10.27.561846.

Figure Legends

Figure 1: Model scheme of the ODE model of the JAK2/STAT5 pathway in erythroid progenitor cells.

Figure 2: CoPaSi simulation with 20 U/ml Epo.