Interacting evolutionary pressures drive mutation dynamics and health outcomes in aging blood
A small population of self-renewing, hematopoietic stem cells continuously reconstitutes our immune system. As we age, these cells, or their pluripotent descendants, accumulate somatic mutations; some of these mutations provide selection advantages and increase in frequency in the peripheral blood cell population. This process of positive selection, deemed age-related clonal hematopoiesis (ARCH), is associated with increased risk for cardiac disease and blood malignancies, like acute myeloid leukemia (AML). However, it remains unclear why some people with ARCH do not progress to AML, even when their blood cells harbor well-known AML driver mutations. Here, we examine whether negative selection can play a role in determining AML progression by modelling the complex interplay of positive and negative selective processes. Using a novel approach combining deep learning and population genetic models, we detect pervasive negative selection in targeted sequence data from the blood of 92 pre-AML individuals and 385 healthy controls. We find that the relative proportion of passenger to driver mutations is critical in determining if the selective advantage conferred to a cell by a known driver mutation is able to overwhelm negative selection acting on passenger mutations and allow clones harbouring disease-predisposing mutations to rise to dominance. We find that a subset of non-driver genes is enriched for mildly damaging mutations in healthy individuals fitting purifying models of evolution suggesting that mutations in these genes might confer a protective role against disease-predisposing clonal expansions. Through exploring non drivercentric models of evolution, we show how different classes of evolution act to shape hematopoietic dynamics and subsequent health outcome which may better inform disease prediction and unveil novel therapeutic targets. We anticipate that our results and modelling techniques can be broadly applied to identify both driver mutations and those mildly damaging passenger mutations, as well as help understand the early evolution of cancer in other cells and tissues.