5-Azacytidine Induction of Human γ-Globin Gene Expression: Experimental Evaluation of Current Models.

5-Azacytidine (5-Aza) was demonstrated to be a potent inducer of human fetal globin gene expression more than 20 years ago. More recently, 5-Aza-2-deoxycytidine has been shown to have similar properties. Since the 1980’s there have been two predominant hypotheses to explain the action of these agents. The first is based on the observation that these, and several other active inducing agents, are cytotoxic to differentiating erythroid cells and that drug treatment alters the kinetics of erythroid differentiation. This has been proposed to result in prolonged expression of the γ-globin genes which are normally expressed only early in differentiation. The second is based on the observation that both agents are DNA methyltransferase inhibitors and are presumed to cause demethylation of cellular DNA including the γ-globin gene promoters leading to activation of the genes. These two models lead to specific predictions that we have evaluated using an in vitro erythroid differentiation system. In this system, human adult CD34+ cells are cultured in SCF, Flt3 ligand and IL-3 for 7 days and then switched to Epo for 14 days. This results in an exponential expansion of erythroid cells. As has been described for normal human differentiation, these cells express small amounts of γ-globin mRNA early in differentiation followed by a much larger amount of β-globin mRNA. HPLC at the end of the culture period shows 99% HbA and 1% HbF. Treatment of cultures on a daily basis with 5-Aza starting on day 10 results in dose dependent increases in γ-globin mRNA, Gγ- and Aγ-chain production and HbF. The cytotoxicity model predicts that γ-globin expression will be prolonged to later in differentiation - and this is seen. However, a daily 5-Aza dose of 300 nM, which produces ~80% of the maximal response in γ-globin mRNA and HbF, has no effect on cell growth or differentiation kinetics. This argues against the toxicity model. We next examined the effect of 5-Aza on γ-globin promoter methylation using the bisulfite method. We studied CpGs at −344, −252, −162, −53, −50, +6, +19 and +50 relative to the start site. For untreated controls, all of the sites are nearly 100% methylated at day 1. By day 3, the upstream sites become ~50% methylated except the −53 CpG which was <20%. This pattern persisted at day 10. By day 14 the promoters had become largely remethylated. For cells treated with 5-Aza starting on day 10, there was no change in the levels of methylation seen on days 1,3 and 10, but at day 14 the low levels of upstream methylation persisted - just as γ-globin expression does. However, in both treated and untreated cells, down-stream CpG sites were highly methylated at all time points. This suggests that γ promoter demethylation may be due to a local and not a generalized effect of 5-Aza on cellular DNA methylation. We also made two unexpected observations. At a 300nM dose of 5-Aza, γ-globin mRNA is ~doubled while β-globin mRNA levels are ~halved - indicating that 5-Aza not only induces γ-globin expression also suppresses β-globin. Also despite only a doubling in γ-globin mRNA, there was an ~50-fold increase in HbF, from ~1% to more than 50%, while total per cell Hb levels were unchanged. Neither of these results are easily explained by current models of γ-globin gene induction. Our results raise the possibility that mechanisms beyond cytotoxicity and generalized DNA demethylation may be responsible for pharmacologic induction of γ-globin mRNA and HbF.