Abstract B2-50: Stochastic population genetics modeling of the evolution of the myelodysplastic syndrome

Abstract

Abstract We are developing new models to understand the evolution of Myelodysplastic Syndromes (MDS), a group of pre-leukemic blood diseases, which are increasing in incidence as the population is aging and which remain difficult to treat. In the early stage of MDS, exaggerated apoptosis of myeloid blood cells occurs, but as the disease evolves, survival of undifferentiated cells dominates. As a model we study progression to MDS/AML from severe congenital neutropenia (SCN), an inherited bone marrow failure with known germ-line mutations, e.g. ELANE which encodes the neutrophil elastase. Administration of granulocyte colony-stimulating factor (GCSF) improves the neutropenia but is associated with transformation to MDS and acute myeloid leukemia (AML), defined by greater than 20% blasts in the marrow. Of note, ~ 70% of SCN patients who progress to MDS acquire a truncating D715 mutation in the GCSF Receptor. We previously developed a model of the SCN -> MDS -> AML sequence [1], which however did not include the MDS stage. We hypothesize that the propagation of the D715 mutant in MDS secondary to SCN follows a newly developed Moran model with directional selection and co-localization. Selection is caused by alterations in the STAT/SOCS feedback loop with the truncated GCSF receptor, leading to changes in proliferation and differentiation [2]. Usually, mathematical models assume that bone marrow is a well-mixed environment and the normal and malignant precursors freely interact. However, these subpopulations occupy distinct niches [3] and their competition is limited by physical separation, which is a factor in our variant of the Moran model. We estimate the selective advantage of mutant cells based on competitive repopulation assay and flow cytometry cell-cycle distribution and find the time at onset of MDS consistent with the published median of 13 years [1]. We built an ODE model of the STAT/SOCS circuit, and investigated its sensitivity to reduced binding of JAK/STAT complexes to the D715 mutant receptor, even that the circuit exhibits extremely strong stabilizing properties. A link to cell cycle characteristics via distributions of cell-cycle-phase dependent activated STATs suggests connection with differentiation and proliferation. Further, to investigate a fuller model in which a number of driver mutations may occur at a different order, we developed a new version of the infinite-allele branching process of Griffiths and Pakes [4, 5]. Mathematical modeling of SCN->MDS -> AML will lead to a greater understanding of homeostatic and aberrant myelopoiesis as well as designing better predictive biomarkers for malignant clonal evolution.