DNA methylation and mechanism of action of 5-azacytidine

To the editor:

Mabaera et al¹  addressed the role of DNA hypomethylation in the induction of γ-globin expression by 5-azacytidine. On the basis of the failure of siRNA-mediated down-regulation of DNMT1 to induce expression of γ-globin in cultured human erythroid progenitors, the authors suggest “that the decrease in γ-globin promoter methylation seen with 5-azacytidine is not the primary cause of γ gene induction and fetal hemoglobin (HbF) production, but is a secondary effect related to gene activation by some other mechanism.” As an alternative, the authors propose changes in posttranscriptional RNA stability and translational efficiency possibly related to 5-azacytidine incorporation into RNA transcripts (5-azacytidine is a ribose). In proposing this mechanism, the authors fail to consider that decitabine (5-aza-2′-deoxycytidine), a deoxyribose analog of 5-azacytidine that is primarily incorporated into DNA,^2,–4  produces equivalent elevations of HbF in baboons and in patients with sickle cell disease at molar doses 10% to 20% that of 5-azacytidine^5,,,,,,–12  (corresponding to the approximate 10% of 5-azacytidine that is incorporated into DNA² ). Furthermore, the ribonucleotide reductase inhibitor hydroxyurea, which would be expected to block conversion of 5-azacytidine into the deoxyribose form and its incorporation into DNA but not into RNA, blocks the ability of 5-azacytidine to induce HbF in baboons.⁷  These results are most consistent with a requirement for 5-azacytidine incorporation into DNA, necessary for DNA methyltransferase inhibition, in its mechanism of action.

Consistent with a transcription activation mechanism of action, γ-globin promoter hypomethylation produced by decitabine treatment of baboons resulted in recruitment of RNA polymerase II and acetylation of histone H3.¹³  Although these results are from baboon erythroid cells, β-globin locus structure and developmental globin switching are similar in baboons and man and 5-azacytidine– and decitabine–related observations from this model have successfully translated into the clinic.^5,,,,,,–12  Furthermore, observations with primary erythroid cells isolated after in vivo decitabine treatment may be more representative of clinical effects than treatment of human cells in vitro, using high levels of fetal bovine serum and growth factors. In the baboon model, γ-globin promoter methylation during HbF expression was significantly lower than that observed by Mabaera et al.¹  We suggest that the siRNA and shRNA directed to DNMT1 may not have reduced γ-globin promoter DNA methylation to sufficient levels at an appropriate stage of erythroid differentiation to increase γ-globin expression. While the authors state that shRNA treatment reduced γ-globin methylation to levels observed in fetal liver erythroid cells (20%), this was only demonstrated on d17 of culture (Figure 6C¹ ), a time beyond peak globin gene transcription. At earlier times (d9, d13) γ-globin methylation was significantly higher (50%-60%). Other DNMTs may assist or act in a redundant manner to maintain γ-globin gene DNA methylation and knock-down of DNMT1 alone may be inadequate to mimic the effect of 5-azacytidine or decitabine.^14,15 

In conclusion, we suggest that the experimental evidence presented in the Mabaera paper is insufficient to support the conclusion that, with regards to the mechanism of induction of γ-globin expression by 5-azacytidine, decreased methylation of the γ-globin promoter is an unimportant secondary effect.