RESULTS

Absence of IRF-4 expression in leukemia cells is not due to promoter alterations
We have previously demonstrated a lack of IRF-4 expression in leukemia patients and specifically in CML T-cells (3). Here, we demonstrate the absence of IRF-4 expression in various hematopoietic cell lines, such as Jurkat, a T-cell leukemia, CML-T1, a bcr-abl-positive T-cell line, K-562, a bcr-abl-positve erythroleukemia, U-937, a monocytic leukemia, EM-2 and LAMA-84, bcr-abl-positve myeloid leukemia, but not in SD-1, a bcr-abl-positive acute lymphoblastic leukemia (pre B-ALL), RPMI-8226, a multiple myeloma and BV-173, a bcr-abl-positive B-cell line (Figures 1A and 5D). After sequencing of the IRF-4 promoter, it could be excluded that absence of IRF-4 expression in any of the above cell lines was due to genetic aberrations. However, 2 bp changes (nucleotide −1081, T→C and −1068, A→C) could be detected in both the IRF-4-positive BV-173 and the IRF-4-negative LAMA-84, EM-2 and K-562 (Figure 1B). At position −116 an A→C substitution was found in EM-2, K-562 and CML-T1, whereas Jurkat, BV-173 and SD-1 exhibited a mixed A/C sequence and U-937, LAMA-84 and RPMI-8226 no substitution at all (Figure 1B). Consequently, these alterations are unlikely to affect IRF-4 expression.

Increase of IRF-4 expression in hematopoietic cells after demethylating treatment
We next analyzed whether promoter methylation could be responsible for down-regulation of IRF-4 expression. A region including exon1 in the IRF-4 promoter exhibited a large number of CpG-rich sequences (Figure 3A). Several chemical substances such as 5-aza-2-deoxycytidine (AzadC) or 5-azacytidine (AzaC) inhibit de novo and maintenance methylation, and thus can be used to discern promoter methylation (32,33). We used AzadC to generate unmethylated DNA. A 72 h AzadC-treatment resulted in a concentration-dependent activation of IRF-4 transcription in Jurkat and CML-T1 T-cells as well as in U-937, K-562 and EM-2 cell lines (Figure 2A). IRF-4 transcription was induced in a time-dependent manner and was observed as early as 24 h after treatment with AzadC and increased over time until 72 h (Figure 2B). Time and strength of the appearance of IRF-4 transcripts varied among cell lines, i.e. CML-T1 responded strongest to AzadC-treatment (data not shown). In line with this, AzadC-treatment of CML-T1 and LAMA-84 cells also translated in an induction of IRF-4 protein expression (Figure 2C). Accordingly, treatment of the IRF-4-positive cell line BV-173, SD-1 and RPMI-8226 with AzadC had no effect on IRF-4 expression (Figure 2D). There was no difference in the effects of AzaC versus AzadC, as both increased the IRF-4 mRNA level in CML-T1 cells as well (data not shown). This implied that promoter methylation may control IRF-4 expression, but an alternative explanation may be activation of positive transcriptional regulators of IRF-4 by AzadC (or AzaC).

Methylation-sensitive enzymes do not cut specific sites in the IRF-4 promoter in hematopoietic cells
To further investigate promoter methylation as a regulatory mechanism of IRF-4 gene expression, restriction-PCR-assays were performed (20,24), where only methylated DNA would not be cut enabling subsequent PCR amplification and vice versa. Genomic DNA from leukemic cells Jurkat, CML-T1, U-937, K-562, EM-2 and BV-173 was digested with the methylation-sensitive enzymes HpaII, Bsh1236I and HaeII-isochizomer Bsp143II. EcoRI, which has no recognition site within the IRF-4 promoter, and the methylation-resistant enzyme MspI served as controls. Two separate amplification reactions were performed, generating two fragments, F1 and F2 (Figure 3A). After digestion with HpaII and Bsp143II a sufficient PCR amplification of F1 and F2 was detected in DNA from IRF-4-negative Jurkat, CML-T1, U-937, K-562 and EM-2 cells, suggesting a promoter methylation (and restriction protection) at the respective recognition sites (Figure 3B and C). Notably, in IRF-4-positive SD-1 cells digestion with the methylation-sensitive enzymes completely inhibited amplification of F1 and F2. In IRF-4-positive BV-173 cells a HpaII, but not a Bsh1236I digestion, significantly reduced the amplifiable DNA message of F2 (Figure 3C), whereas amplification of F1 was not affected (Figure 3B). This implied that IRF-4 transcription in SD-1 and BV-173 cells is associated with less promoter methylation (in BV-173 especially at HpaII sites) as compared with the tested IRF-4-negative cells.

Specific CpG sites in the IRF-4 promoter are methylated in hematopoietic cells
In order to exactly map the methylation sites within the IRF-4 promoter, we treated DNA of Jurkat, CML-T1, U-937, K-562 and EM-2 cells as well as of SD-1, RPMI-8226 and BV-173 control cells with bisulfite, which chemically converts unmethylated cytosine to uracil, whereas it has no effect on methylated cytosine, i.e. in CpG (34). This technique is especially useful for detection of unknown methylation patterns. PCR amplification, cloning and sequencing of the bisulfite-treated DNA showed a specific methylation pattern of the analyzed 62 CpG sites in all cell lines (Figure 4 and Table 1). In general, the methylational status ranged from one cell line with a nearly non-methylated IRF-4 promoter (SD-1, IRF-4-positive) to a completely methylated IRF-4 promoter in CML-T1 (IRF-4-negative). Interestingly, the percentage of CpG methylation in the IRF-4 promoter from IRF-4-positive cells was very low (mean 24%) as compared with IRF-4-negative cells (mean 94%) (Figure 4A and Table 1). A 5′-region (R1) with 13 hypermethylated CpG sites (mean number of methylated clones 5.5 of 8 with 77% methylated CpGs) was found in most cells (except SD-1 and RPMI-8226) and a 3′-region (R3) of 6 hypomethylated CpG sites (mean number of methylated clones 1.7 of 8 with 33% methylated CpGs) was found in most cells (except CML-T1 and U-937) (Figure 4A and Table 1).
Intriguingly, a stretch of 13 CpG sites (#10–22; R2) was detected in between these regions, which were highly methylated in IRF-4-negative (mean number of methylated clones 7.1 of 8 with 89% methylated CpGs) but totally non-methylated in IRF-4-positive cells (Figure 4A and B). Furthermore, three CpG sites at the 5′ end (#54, 56, 58) and two CpG motifs at the 3′ end (#1, 2) showed this direct correlation between high methylation status and absence of IRF-4 expression. In addition, two CpG sites located in a NFκB (#48) and a SP1 element (#45) are less methylated in IRF-4-positive than in IRF-4-negative cells (mean number of methylated clones: 1/8 versus 8/8). These results indicate the involvement of CpG methylation in the regulation of IRF-4 expression in leukemic cells.

In vitro methylation of an IRF-4 promoter-reporter construct decreases its activity
To provide evidence for a direct effect of methylational status on IRF-4 promoter activity we performed reporter gene assays with IRF-4 promoter constructs before and after their in vitro methylation. A complete methylation of these constructs was checked via restriction assays with methylation-sensitive endonucleases (Figure 5A). Intriguingly, methylation of the IRF-4 promoter significantly decreased promoter activity in IRF-4-positive SD-1 cells by 85.0% (Figure 5B). The silencing effect of CpG methylation was not restricted to IRF-4-positive cells, since in vitro methylation led to a 92.9% abrogation of promoter activity in IRF-4-negative Jurkat cells (Figure 5C). In contrast, control methylation of a reporter construct with a different promoter (FasL) as well as an empty vector had no effect on the reporter activity (data not shown). These data proved a direct association between methylation and activity of the IRF-4 promoter.

mRNA expression of DNA methyltransferases and methyl-CpG-binding proteins may not be associated with IRF-4 promoter methylation
Since abundance of DNMT and MBP contribute to promoter regulation via methylation (25,26,28), we studied their mRNA expression to investigate a possible mechanism for the observed methylation differences in the IRF-4 promoter. To this end, we did not detect a significant difference in DNMT (DNMT1, DNMT3A and DNMT3B) or MBP (MBD1, MBD2, MBD4 and MeCP) mRNA expression between IRF-4-positive and -negative cells (Figure 5D). In fact, all analyzed cells had moderate to high mRNA levels of these tested DNMT/MBPs and differences in expression were not correlated with IRF-4 status. These results indicate a distinct cause of the methylation differences in IRF-4-positive and -negative cells rather than changes in the DNMT and MBP mRNA transcription.