Fig. 2: Statistical inference of HSC differentiation pathways.

a Model scheme, with subdivision of HSCs and progenitors according to Sca-1 expression and, for HSCs, CD201 expression. Straight arrows denote cell differentiation (vertical), or change in Sca-1 expression (horizontal), or both (diagonal); curved arrows denote cell proliferation. All possible differentiation are superimposed. b Model selection by bias-corrected Akaike information criterion (AICc) against experimental data for HSC fate mapping, mitotic history, and relative cell numbers. Models are classified (and color-coded accordingly) by two features: parallel differentiation pathways and irreversible loss of Sca-1. The best-fitting model c has ∆AICc = 0. None of the remaining 143 models is within ∆AICc < 2. Further models with acceptable evidence (∆AICc < 10) are all variants of the best-fitting model (see Supplementary Fig. 5b). c Best model selected by experimental data. Arrow width signifies values of the corresponding differentiation or proliferation rates, while node size quantifies cell number (relative to lin⁻ cells) in each compartment at 100 days of chase; see Supplementary Table 1. Fits of the best model for RFP labeling frequency propagation (d), compartment size (e), and H2B-GFP dilution (f) (experimental data: mean ± SEM; mice in d and e are the same as in Fig. 1c–e; for f, n = 3, 4, 4, 2, 2, 4, 4 mice/consecutive timepoint (0−21 weeks of chase) were analyzed; model fit: straight line, best fit; gray-shaded areas, 95% confidence bounds). g Inferred lineage pathways to CMPs and their relative contributions to generation of CMPs (showing the probability that a CMP was generated by each pathway; error bars, 95% prediction bands). h As predicted by the model, transplanted Sca-1^hi HSCs, but not Sca-1^lo HSCs, yielded lymphoid progeny. 100 Sca-1^hi or Sca-1^lo donor HSCs were competitively transplanted and contribution to peripheral blood neutrophils, B and T cells was determined (data re-analyzed from Morcos et al.²⁵). Lineage bias of donor-derived leukocytes was calculated (Neutrophil chimerism/[B + T cell chimerism]/2; n = 4 recipient mice for Sca-1^hi HSCs and n = 3 for Sca-1^lo HSCs in primary and secondary transplantation; mean and SEM are shown). i Ratio of cell fluxes between Sca-1^lo HPC-1 → CMP and Sca-1^hi HPC-1 → CLP for young and old mice (error bars, 95% prediction bands).