A therapeutically crucial sub-set of leukemia cells are those that self-renew (leukemia stem-cells or LSC). An open question is whether LSC self-renewal is mediated by the same pathways or factors that self-renew normal hematopoietic stem-cells (HSC). A difference in self-renewal mechanisms could form the basis for LSC-specific therapy. Hematopoietic differentiation is dependent on lineage-determining factors such as Pu.1. Pu.1 can repress or activate genes, depending on its interaction partners. We show that Pu.1 mediated terminal differentiation involves sequential repression of genes associated with self-renewal (pro-SR) (HoxB4, Bmi-1, c-Kit) followed-by activation of genes associated with differentiation (pro-DIFF)(Mcsfr, Gmcsfr, F4/80). DNA methyl-transferase 1 (DNMT1) is an arbiter of transcription repression. DNMT1 depletion concurrent with Pu.1 introduction prevents the first step of pro-SR repression and maintains self-renewal. However, DNMT1 depletion 6 hours after Pu.1 introduction (at which point pro-SR repression by Pu.1 is complete), increases differentiation. Therefore, the phenotypic consequences of DNMT1 depletion critically depend on the differentiation chronology of the cell. We then examined the effects of a leukemia first-hit event, Runx1 disruption, on Pu.1 mediated differentiation. Runx1 knock-down allowed Pu.1 mediated pro-SR repression but prevented pro-DIFF up-regulation and terminal differentiation, resulting in persistent dysregulated proliferation (self-renewal) despite pro-SR repression. DNMT1 depletion in these abnormally self-renewing cells, with an epigenetic and transcript profile distinct from parental self-renewing cells, relieved pro-DIFF repression and restored the terminal differentiation response. Therefore, DNMT1 depletion had opposite effects in the cells containing the leukemia-associated abnormality versus parental cells. Suggesting that these findings were relevant to clinical disease, leukemic bone marrow (n=130), although consisting largely of myeloblasts, demonstrated a DNA methylome (1505 CpGs, Illumina) profile that resembled differentiated cells and not precursors, with aberrant methylation concentrated at precursor signature (pro-SR) promoters and hypomethylation at differentiation signature (pro-DIFF) promoters. To therapeutically exploit these observations, we demonstrate that the nucleoside analogue Decitabine, given at doses that deplete DNMT1 without causing DNA damage, and given frequently but intermittently to allow for cell-division, produces an ideal therapeutic profile with terminal differentiation of leukemia cells but increased self-renewal of human CD34+ HSC. This effect was seen in different models of human LSC (primary human CD34+ cells transduced with either the MLL-AF9 or RUNX1-ETO leukemia fusion genes) and in primary leukemia cell samples (n=15) containing a variety of chromosome abnormalities. This therapeutic approach increased survival in a murine xeno-graft model of aggressive human MLL-AF9 leukemia. In conclusion, leukemia and LSC self-renewal, in a number of examples covering a variety of transformation-initiating abnormalities, is programmatically distinct from HSC selfrenewal. Leukemia self-renewal is associated with an epigenetic and transcript profile that reflects part-way commitment into differentiation with aberrant gene repression preventing completion of the process. This difference could explain the opposite effects of DNMT1 depletion on HSC and leukemia self-renewal, and provides the scientific foundation for differentiation (frequent metronomic low-dose) instead of cytotoxic (infrequent high-dose) regimens of epigenetically active and clinically available agents such as Decitabine.

Disclosures: No relevant conflicts of interest to declare.

Author notes

Corresponding author