A Novel iPSC Model Reveals a Role for RUNX1 in the Maintenance of AML Leukemia Stem Cells

Leukemia stem cells (LSCs) are believed to be a prominent source of relapse in acute myeloid leukemia (AML). Human AML LSCs are classically studied via xenotransplantation and defined as cells with (a) self-renewal ability, giving rise to leukemic engraftment that can be maintained over serial transplantation; and (b) ability to give rise to more differentiated progeny that are unable to engraft. LSCs are chemoresistant presumably because they have properties distinct from those of the bulk AML cells. A better understanding of the properties of LSCs can ultimately allow us to develop new therapies specifically targeting these cells to achieve lasting responses or even cures.

We recently reported that induced pluripotent stem cells (iPSCs) generated from AML patients (AML-iPSCs) exhibit leukemic features upon hematopoietic differentiation, including extensive proliferation, maintenance of hematopoietic stem cell (HSC) markers, and serial engraftment of a lethal leukemia in immunodeficient mice (Kotini et al. Cell Stem Cell 2017). More recently, we found that the hematopoietic stem/progenitor cells (HSPCs) derived from AML-iPSCs exhibit phenotypic and functional heterogeneity. We first observed the presence of two cell populations with distinct growth characteristics and morphology: a population exhibiting adherent growth (adherent, A) and a population growing in suspension (S). The A cell population contained cells with immature morphology and HSC immunophenotype (CD34⁺/CD38^-/CD90⁺/CD45RA^-/ CD49f⁺), while the S fraction contained more differentiated cells (CD34^-/low/CD38⁺/CD90^-/CD45RA⁺). Experiments utilizing serial replating, single-cell plating of GFP-labeled cells and mathematical modeling revealed a hierarchical organization, whereby the A cells continuously give rise to the S cell fraction. Time-lapse imaging showed that A cells divide both symmetrically and asymmetrically with the majority (~60%) of cell divisions giving rise to two A cells and less frequently generating either one A and one S (~15%) or two S cells (~15%). Serial transplantation experiments in NSG mice revealed that the engraftment potential was largely contained within the A cell fraction. Cell cycle analysis showed that adherent cells contained a lower S and higher G0/G1 phase fraction than suspension cells. Thus, these AML-iPSC-derived hematopoietic cells exhibit hallmarks of an LSC model, namely phenotypic and functional heterogeneity and hierarchical organization, with the A fraction containing LSCs that serially transplant leukemia and give rise to more differentiated cells (S fraction) without engraftment potential.

Transcriptome analyses showed that the A to S transition was predominantly mediated by gene upregulation with only 2 out of 418 differentially expressed genes being downregulated and revealed enrichment for HSC and LSC gene sets in the A fraction. Consistent with a general transcriptional upregulation characterizing the A to S transition, differentially accessible regions by ATAC-seq analysis were found predominantly in intragenic regions and enhancers in A cells, while they mostly localized in gene promoters in S cells. Time-course single cell-RNA-seq analyses and their integration with bulk RNA-seq and ATAC-seq data revealed cell clusters specific to the A cell fraction, which were enriched in HSC, LSC and hESC genes. These integrated analyses revealed candidate genes with a role in maintaining LSC properties. The transcription factor RUNX1 was particularly prominent, with RUNX1 motifs found highly accessible in A compared to S cells. Overexpression of RUNX1 in S cells conferred a more LSC-like immunophenotype. RUNX1 is a key transcriptional regulator of hematopoiesis. It has a well-characterized role during development of the hematopoietic system and is often mutated or translocated in AML. A role of RUNX1 in LSC maintenance was unanticipated in view of its known tumor-suppressor function. In further support for a role for RUNX1 in LSCs, we found RUNX1 expression to significantly correlate with survival in AML patient cohorts.

In summary, we developed a new model that enables us to prospectively isolate large numbers of genetically clonal human AML LSCs and perform genome-wide integrative molecular studies, with which we obtained new insights into the biology of AML LSCs.