Neither DNA Hypomethylation or Changes in the Kinetics of Erythroid Differentiation Account for 5-Azacytidine’s Ability To Induce Human Fetal Hemoglobin.

5-Azacytidine (5-Aza) is among the most potent inducers of fetal hemoglobin (HbF) in patients with β-thalassemia and sickle cell disease. Two models have been proposed to explain this activity. The first model is based on the drug’s ability to inhibit DNA methyltransferase enzymes causing global DNA hypomethylation, including the promoters of the fetal globin genes, resulting in their expression during adult erythropoiesis. The second model is based on the drug’s other well-know property - cytotoxicity. In this model, a cohort of differentiating erythroblasts either dies or undergoes growth arrest during drug exposure. The marrow then responds with a rapid proliferation of erythroblasts which express high levels of fetal globin mRNA and HbF. To determine which model best explains HbF induction by 5-Aza, we used an in vitro differentiation system in which CD34+ cells from normal donors are cultured with SCF, IL-3 and Flt-3 ligand for 7 days and then switched to Epo for an additional 14 days. This results in >90% erythroid cells, a 2,000-fold expansion in cell number and Hb expression similar to that seen in normal red cells (∼99% HbA, ∼1% HbF). To model common 5-Aza dosing schedules, we treated cultures daily with doses ranging from 0 to 1000 nM. Near-maximal induction of γ-globin mRNA (∼2-fold increase in γ/γ+β) and HbF (from 1% to 40%) occurred at the 300 nM dose level. While this dose was associated with hypomethylation of the 7 CpGs between −256 and +50 bp of the γ promoter (59% +/− 14% methylated CpGs vs. 92% +/− 8.1% in untreated controls, p < 0.001), no changes were found in cell expansion rate, differentiation kinetics (as judged by glycophorin A expression) or cell cycle distribution. These results argue against the cytotoxicity model. We also observed that global DNA methylation was not significantly changed by the 300 nM dose of 5-Aza. Because 5-Aza is incorporated into both RNA and DNA, it affects many cellular processes beyond DNA methylation. To determine if DNA hypomethylation is sufficient for γ-globin mRNA and HbF induction, we treated differentiating cells with DNMT1 siRNA. This resulted in transiently decreased DNMT1 mRNA and protein and lowered global DNA methylation. γ-globin promoter methylation decreased to levels equivalent to those seen with 300 nM 5-Aza (61% +/− 7% of CpGs methylated vs. 88% +/− 8% in controls, p < 0.001) but did not induce γ-globin mRNA or HbF. To further evaluate this unexpected finding we used a lentiviral vector expressing DNMT1 shRNA. This suppressed DNMT1 mRNA and protein throughout differentiation and decreased total γ-globin promoter methylation from 98% +/− 4% in control cells, to 20% +/− 11% (p<0.001) in treated cells. This level of γ promoter methylation is similar to that seen in fetal erythroid cells (13% +/− 7%). Global CpG methylation was also decreased vs. controls (30% +/− 10% vs. 83% +/− 11%, p<0.001). Again, despite significant γ promoter and global hypomethylation, no induction of γ-globin mRNA or HbF was observed. These results suggest that neither model explains 5-Aza’s ability to induce fetal Hb and that alternative mechanisms await discovery. If verified, these findings have important implications for the future development of clinically useful HbF inducing agents.