DISCUSSION
Many genetic lesions are known to influence gene expression of tumor suppressor genes. Whereas mutations and deletions or insertions have permanent effects, reversible mechanisms are gene methylation, or expression and activation of transcription factors, respectively. We studied a putative cause for absent IRF-4 expression in leukemia cells and first focused on genetic aberrations of the promoter. We observed no genetic alterations in the IRF-4 promoter, which can account for the lack of IRF-4 expression: The detected base pair changes at position -1081 (T-->C substitution), at position -1068 (A-->C substitution) and at position -116 (A-->C substitution) are unlikely responsible for absent IRF-4-expression since the first two mutations were found both in IRF-4-positive and -negative cells whereas the latter change was not detected consistently in all IRF-4-negative or -positive cells and may thus be a polymorphism. All three substitutions did not change any known putative transcription factor binding sites (30,31) and also do not affect any restriction sites or primer binding sites of the used assays. However, permanent genetic variations in the IRF-4 coding sequence, such as deletions or mutations resulting in stop-codons have not been excluded by sequence analysis. Since IRF-4 expression in cell lines and CML can be induced by demethylation and successful IFN-alpha therapy (3), respectively, the existence of such genetic aberrations seems unlikely. 
We then investigated whether the previously described down-regulation of IRF-4 expression in human myeloid leukemias was due to a differential hypermethylation of the promoter, since the presented re-expression due to AzadC-treatment might also be a result of activation of positive transcriptional regulators of IRF-4. Methylation of CpG sites is a common mechanism of silencing genes in leukemia and has also been shown for another IRF, IRF-7 (35) and for PU.1 (36), an interacting partner of IRF-4. To elucidate the relevance of this mechanism for the regulation of IRF-4 expression, various leukemic cells were treated with demethylating agents and promoters were sequenced after bisulfite treatment. We found that IRF-4 expression could indeed be connected to the methylation status of distinct CpG motifs in the IRF-4 promoter. In Figure 4A, those CpG sites are shown (bottom line), whose hypermethylation may account for the absence of IRF-4 expression in the respective cells. One of them (#54) is adjacent to an identified regulatory element (NFkappaB-site), indicating a possible involvement of this site. At two further CpG sites (#48, 45) the methylation status in IRF-4-positive was lower than that of IRF-4-negative cells. These CpG sites are located in an NFkappaB and an SP1 element (31) and thus may also play a role in regulation of IRF-4 expression. It has been shown that NFkappaB elements play an important role in IRF-4 induction as IRF-4 expression depends on binding of the transactivator c-Rel to these elements in the IRF-4 promoter (31,37). Furthermore, methylation of the central CpG in the NFkappaB element inhibits binding of the NFkappaB protein complexes (38), promoting the significance of the observed methylation differences in IRF-4-positive and -negative cells. 
Via in vitro methylation and reporter gene assays we could clearly appoint the silencing of the IRF-4 promoter to a methylation effect, which may thus be the mechanism of IRF-4 deregulation in vivo. One possible cause for the aberrant methylation in tumorigenesis is an increased level of DNMTs during the pathogenetic process. In colon, lung and hematologic malignancies, overexpression of DNMT1, a maintenance DNMT, has been detected (39-41). Furthermore, it has been shown that CML cells in the acute phase exhibited elevated levels of the three known DNMTs, while CML cells in chronic phase expressed normal levels of DNMTs if compared with normal bone marrow cells (25). Interestingly, a positive correlation between DNMT1 expression levels and hypermethylation of p15INK4b has been detected in AML (25). In this work, we did not detect significant mRNA expression differences of selected DNMT or MBP, making it an unlikely cause for the observed methylation and thus IRF-4 expression differences in leukemia cells. 
The finding that IRF-4 expression is silenced by promoter hypermethylation might represent a mechanism that accounts for the previously observed loss of IRF-4 expression in CML. Indeed, several clinical trials with leukemia patients and patients with myelodysplastic syndromes demonstrated the potential clinical benefit of a treatment with demethylating agents (42-45). 
The expression of another IRF, IFN consensus sequence binding protein (ICSBP/IRF-8), is impaired in myeloid leukemias especially CML (27,46,47). But in contrast to IRF-4, the loss of this IRF could not be reverted in ICSBP-negative cell lines (EM-2, CML-T1, K-562 and LAMA-84) by treatment with AzadC (Figure 6) and AzadC has no effect on ICSBP levels in ICSBP-positive U-937 cells (Figure 6). These data suggest a distinct regulatory mechanism for these two IRFs. 
IRF-4, similar to many other classical tumor suppressor genes p15INK4b, p16INK4a or p53, may thus be a subject of alterations in the promoter methylation status leading to expression changes, which might contribute to the initiation and/or progression of cancer. Still, the obvious functional diversity of IRF-4 remains remarkable and cannot be fully explained by the IRF-4 promoter methylation status. For example, IRF-4 is primarily known for its oncogenic features. In multiple myeloma (MM) a translocation on chromosome 14q was reported to lead to a fusion gene of immunoglobulin heavy-chain (IgH) and IRF-4 resulting in a subsequent overexpression of IRF-4 (48,49). In addition, abundant IRF-4 expression was found to be a marker for various subsets of lymphomas, such as diffuse large B-cell lymphomas, primary effusion lymphoma, and marginal zone lymphoma, and adult T-cell leukemia (11,31,50-52). This draws a more complex picture of the role of IRF-4. Down-regulation of IRF-4 may promote leukemogenesis in myeloid cell context (3), which was recently confirmed in IRF-4-/- ICSBP-/- double knock-out mice (53), while IRF-4 up-regulation may induce a growth advantage in lymphomas or MM (48). 
Taken together, our data suggest that IRF-4 promoter methylation regulates IRF-4 expression, and that aberrant expression of IRF-4 in certain types of leukemia may be a consequence of IRF-4 promoter hypermethylation. 
