CD34⁻ AML models of differentiation arrest and genetics of samples. (A, left) Normal hemopoietic hierarchy. Initiating mutations in HSC or very early long-lived progenitors create preleukemic (Pre-L) stem cells with a clonal advantage. Model 1: mutations transform pre-LSCs into CD34⁺ progenitor-like LSCs that differentiate into CD34⁻ GM-pre–like LSCs, resulting in hierarchically arranged LSC populations. LSC populations then differentiate into CD34⁻ non-LSC blasts. Model 2: mutations transform pre-LSCs into CD34⁻ GM-pre–like LSCs that then differentiate into CD34⁻ non-LSC blasts. Here, CD34 would be aberrantly expressed on a small subset of LSCs. Model 3 combines models 1 and 2. Some clones acquire transforming mutations to create CD34⁺ progenitor-like LSCs that differentiate into CD34⁻ GM-pre–like LSCs; other clones acquire transforming mutations to create CD34⁻ precursor-like LSCs only. CMP, common myeloid progenitor; mat, mature granulocyte–monocytic effector cells; MEP, megakaryocyte-erythroid progenitor; MPP, MPP/short-term HSC; PL, preleukemic; pre, granulocyte–monocyte precursors. (B) Characteristics of 49 CD34⁻ AML samples: patient demographics, blast percentage, immunophenotype, karyotype, and mutational profile. (C) Karyotype and disease-associated nucleotide variants. The colored boxes denote either karyotype/risk stratification or mutation. No known disease-associated mutations were detected in CEBPA, CBL, ETV6, WT1, and Cohesin genes (see Table S1 B). (D, top) Frequency of the indicated mutations in our CD34⁻ AML (n = 49), our CD34⁺ AML (n = 84), and our TCGA unselected AML (n = 200) sample cohorts. (Bottom) Mutations within NPM1 mutant samples in all three cohorts. n, number of samples. Statistically significant comparisons assessed by χ² tests are highlighted and marked with asterisks: *, P < 0.05; **, P < 0.01. (E) Distribution of NPM1 wild-type and mutant AML samples across a continuum of CD34 expression (assessed by flow cytometry as a percentage of MNCs: CD34⁺ AML, n = 84; CD34⁻ AML, n = 49). The 2% threshold of CD34 expression used to select the CD34⁻ AML cohort in this study is marked. The difference in percentage of NPM1 mutant samples between the two cohorts was statistically significant (**, P < 0.01). PTEN, phosphatase and tensin homologue.