Amulti-tracklandscape of haematopoiesis informed by cellular barcoding and agent-based modelling

Abstract

In classic ball-and-stick models of haematopoiesis the implicit assumption is that all cells within each defined stem or progenitor cell population are equivalent in their fate. Instead, more recent models suggest a haematopoietic stem and progenitor cell (HSPC) continuum of lineage bias and commitment, which is largely inferred through snapshot analysis of single cell gene expression or clonal fate. However, the dynamic assessment of lineage commitment of specific HSPC populations and their clonal output over time in vivo is still lacking but is essential to fully inform accurate models of haematopoiesis. Here, using cellular barcoding we compare the single cell output of long-term haematopoietic stem cells (LT-HSCs), short-term HSCs (ST-HSCs), multipotent progenitors (LMPPs), common myeloid progenitors (CMPs), common lymphoid progenitors (CLPs), and macrophage/dendritic cell progenitors (MDPs). Each population was assessed for their output to multiple haematopoietic cell types spanning a subset of time points from 9 to 112 days of haematopoiesis after transplantation. These analyses revealed a wide range of clonal fate patterns that were inconsistent with their eponymous labels, i.e. stem and multipotent progenitors were rarely multi- or equipotent, and common progenitors were often highly restricted in their fate. To better describe how these clonal patterns integrate into a revised landscape, a novel agent-based mathematical modelling approach that explicitly accounts for haematopoiesis at a clonal level was developed to allow the simulation of growth, timing and branching of clonal trajectories that underlie the process. Rather than a continuum, the proposed model is suggestive of multiple tracks down which clonal trajectories progress, and where fate can branch to a track of lower potency at multiple points down the entire cascade of haematopoiesis.