Response: 5-azacytidine induction of human fetal hemoglobin

We thank Dr Lavelle and his colleagues for their letter. Their comments are related to 2 major conclusions from our paper: that the induction of fetal globin gene expression by 5-azacytidine (5-Aza) is not the result of global DNA hypomethylation and that both transcriptional and posttranscriptional effects of 5-Aza contribute to increased fetal hemoglobin (HbF) levels.¹  While it may be a fine point, the first conclusion was not based on the inability of siRNA-mediated down regulation of DNMT1 to induce fetal globin gene expression, but on the inability of promoter and global DNA demethylation caused by DNMT1 siRNA knock-down to increase levels of fetal globin mRNA and HbF. This conclusion was also based on data showing that a greater decrease in γ-globin promoter and global methylation produced by stably expressed DNMT1 shRNA also failed to induce fetal gene expression and that 5-Aza (at doses that induced near-maximal fetal globin gene expression) caused a localized demethylation of the upstream promoter region but not of downstream or global CpGs. A similar result has been previously presented.²  Since the publication of our manuscript, Fathallah et al have reported that butyrate induction of HbF is also associated with γ-globin promoter hypomethylation.³  This finding is consistent with the idea that promoter demethylation can be a secondary effect and not the primary cause of γ-globin gene induction. While we agree that the maximal hypomethylating effect in the DNMT1 shRNA experiment occurred near the end of differentiation, both the siRNA and shRNA experiments produced hypomethylation equivalent to that seen with effective doses of 5-Aza earlier in differentiation. The fact that Lavelle and colleagues observed higher levels of promoter demethylation with decitabine in baboons is not inconsistent with our observations. Dose-response analyses in our system demonstrated that robust γ-globin gene induction occurred at concentrations of 5-Aza that are insufficient to cause downstream promoter or global DNA hypomethylation. Baboon erythropoietic cells may have been exposed to levels of decitabine that caused both γ-globin induction and higher levels of demethylation. Lavelle et al also take issue with our suggestion that 5-Aza incorporation into mRNA might explain the posttranscriptional effects we observed with the drug. This was only one of several possible alternative mechanisms that we discussed. Because we had not yet determined whether decitabine also exhibits posttranscriptional effects in our system, it would have been premature to rule out this hypothesis. Finally, we never stated or meant to imply that changes in γ-globin promoter methylation during the pharmacologic induction of fetal Hb are unimportant. Even if reduction of γ-globin promoter methylation is not the primary cause of fetal globin gene induction, it is still likely to be an important component of the induction process. 5-Aza and decitabine are potent, clinically active inducers of fetal hemoglobin. Determining the mechanisms of action of these compounds is likely to be an important step in the development of improved pharmacologic strategies for the β-hemoglobinopathies. Only further testing will determine whether our ideas concerning 5-Aza's mechanisms of action are correct. We hope that our report encourages other investigators to consider these issues.