Acute myeloid leukemia (AML) is a clonal disorder evolving from myeloid progenitor cells on the background of a number of genetic changes including balanced translocations1,2 and alterations of tumor suppressor genes and proto-oncogenes.3Recently, transcriptional silencing of tumor suppressor genes by hypermethylation of promoter CpG islands has emerged as an event that may contribute to the leukemic phenotype. Among the genes found to be hypermethylated at significant rates in AML arep15INK4B,4-6 the estrogen receptor,7 and HIC1.8 By this letter we would like to draw attention to methodological features that may cause confusion in the interpretation of data on promoter methylation.
In a recent issue of Blood, Katzenellenbogen et al9reported that hypermethylation of the death-associated protein kinase (DAP-kinase) is a common alteration in B-cell malignancies. DAP-kinase is a 160-kd cytoskeleton-associated protein with serine/threonine kinase activity. It functions as a positive mediator of interferon-γ induced cell death10 and is furthermore involved in FAS and TNF-α induced apoptosis.11 Using a methylation-specific polymerase chain reaction (MSP) assay12Katzenellenbogen et al found that 100% (9 of 9) of samples from Burkitt's lymphoma and 84% (21 of 25) from other B-cell lymphoma patients were methylated in the DAP-kinase 5′ CpG island. In contrast, no methylation was found in T-cell lymphoma or T-cell acute lymphoblastic leukemia samples. In AML, 3.8% (1 of 26) of pediatric samples and 33% (2 of 6) of adult samples were found to contain methylated DAP-kinase alleles.
We recently developed a novel method, “bisulfite-DGGE,” that detects aberrant methylation and provides a detailed display of the composite pattern of clonotypic epigenotypes within a sample of DNA.6 This method combines bisulfite treatment of genomic DNA with nondiscriminatory PCR amplification of methylated and unmethylated sequences, followed by resolution of alleles with varying methylation density by denaturing gradient gel electrophoresis (DGGE). By using this method, we demonstrated a highly heterogeneous pattern of p15INK4Bmethylation in AML samples.6
Prompted by the above mentioned report,9 we analyzed leukemic blasts from AML patients and leukemic cell lines for methylation in the DAP-kinase CpG island using both MSP and bisulfite-DGGE. MSP was performed with primers identical to those described by Katzenellenbogen et al,9 while for bisulfite-DGGE, we designed a pair of primers to amplify a 103-bp region of the DAP-kinase CpG island, which contains 7 CpG sites and covers essentially the region examined by Katzenellenbogen et al.9
In concordance with the previous report,9 we found, by both MSP and bisulfite-DGGE analysis, that the Raji cell line is completely methylated in the DAP-kinase CpG island, and that the HL-60 cell line is partially methylated (Figure). MSP analysis of clinical AML samples revealed methylated alleles of the DAP-kinase in 42% (19 of 45) of de novo adult AML cases and in none of four de novo pediatric AML cases (Figure). Direct sequence analysis of the PCR products showed that all cytosines at CpG sites had remained as cytosine during bisulfite treatment and all cytosines at non-CpG sites were converted to thymine (data not shown), excluding the possibility that successful amplification could be due to unspecific primer annealing or incomplete bisulfite conversion. Bisulfite-DGGE was performed on the same 49 AML cases and Figure 1B shows bisulfite-DGGE experiments of the same samples as those analyzed by MSP in Figure 1A. For all AML samples, the vast majority of amplified DAP-kinase alleles migrated to a position in the gel corresponding to unmethylated DNA (Figure), and a band pattern consistent with some degree of DAP-kinase promoter methylation could be demonstrated in only 1 case of adult AML (data not shown). For comparison, the 10 AML samples represented in the Figure showed extensive methylation of the p15INK4B CpG island by bisulfite-DGGE analysis (Figure).
The direct comparison between the MSP and DGGE assays performed in this study prompts us to suggest that silencing of DAP-kinase by methylation may not be a biologically significant event in AML. This is evidenced by the fact that by MSP analysis, 42% of adult AML cases showed indication of DAP-kinase CpG island methylation, whereas bisulfite-DGGE of the same samples revealed that only an insignificant fraction of leukemic blasts, or possibly even a different cell type, contained methylated DAP-kinase alleles. These contrasting results may be reconciled by data in a recent study where we found that a small fraction of normal lymphocytes are methylated in the promoter region ofp15INK4B.6 The presence of rare nonmalignant methylated cells in clinical leukemia samples implies that highly sensitive methods for detection of methylation, such as the MSP, should be employed with caution and preferably in combination with other methods in order to obtain information on both the fraction and the extent of methylated alleles. In this setting, and when seen in conjunction with accepted clinical correlates, promoter methylation could, however, well turn out to be of significance in stratifying patients with hematologic diseases.
DAP-kinase methylation: methodology and biology
The above letter by Aggerholm et al examines the methylation of DAP-kinase in acute myelogenous leukemia (AML). As they have done for the tumor suppressor genep15/CDKN2B,1-1 they compare 2 methods of determining methylation patterns: Methylation specific PCR and DGGE, a method they have developed. The more sensitive method (MSP) yields a higher prevalence of methylation at this loci, and the authors rightly question the importance of this methylation. Why do these results differ?
In our original report,1-2 we focused primarily on acute lymphoblastic leukemia and lymphomas, as our initial studies suggested that methylation was particularly frequent in these malignancies. Furthermore, we observed that this methylation was limited to transformed cells, and was not observed in normal peripheral leukocytes or in EBV immortalized B lymphocytes,1-2 even with the sensitivity of MSP as carried out in our lab. This specificity for the transformed phenotype, the association of the lack of expression with inhibited γ-interferon induced programmed cell death, and the further specificity for B vs T-cell malignancies all suggested an important role for this gene in these malignancies.
We did examine other malignancies, including acute myelogenous leukemia, as studied by Aggerholm. In the 26 pediatric AML samples examined, only one exhibited DAP-kinase methylation, and though this leukemia was more correctly classified as a biphenotypic leukemia (B-cell markers were positive, including Leu12 and Leu14), we reported this as AML according to the original clinical diagnosis. The small number of adult AML samples studied—6—precluded us from making conclusions for this malignancy.
Aggerholm et al find methylation of DAP-kinase in 19 of 45 adult AML by MSP, but in only 1 of these is this confirmed by DGGE, suggesting the level of DAP-kinase methylation is minimal in the other samples. Differences in sensitivity between the techniques, including the potential for PCR bias in bisulfite modified DNA1-3 for techniques such as DGGE, could account for some of these differences. Unfortunately, the authors do not provide the details of the MSP analysis, including annealing temperatures, and most importantly, the number of cycles of PCR performed. In our study1-2 and other studies using MSP1-4,1-5 we have limited the cycle number to 35, while the authors in their study of p15 utilized 40 cycles.1-1 It is important to provide a comparison of the amplification of the methylated product with the unmethylated in MSP, so that a relative quantitation can be made. Higher cycle number or altered conditions could potentially lead to false priming or greatly increase the amplification of a minor population of methylated alleles. The authors address the possibility of false priming by sequencing the amplified methylated products in data not shown. While this should exclude the possibility of false priming, it is surprising that all CpG sites in DAP-kinase were methylated, given the authors',1-1our group's,1-6 and others'1-7,1-8 experience with the heterogeneity of methylation for p15/CDKN2B in primary leukemias and in other genes.1-8,1-9
The authors suggest that the methylation observed could be due to contamination of the sample with normal cells. However, for DAP-kinase, they did not examine normal cells. Our experience is that DAP-kinase is not methylated in normal cells.1-2,1-10 Rather, the transformed population could contain a small number of cells with DAP-kinase methylation that only become dominant within the population of cells if clonaly selected, as suggested by Peter Nowell 2 decades ago.1-11
Sensitive molecular methods, like MSP, must be used with caution. The authors may be correct in concluding that the methylation they observe in many of these adult AML samples may not be important for the transformation of these cells. One should determine the importance of methylation for any gene by examining the effect such methylation has on gene expression. Nonetheless, this result does not preclude the importance of inactivation of this gene in other malignancies.
We are grateful to Karin Brændstrup for expert technical assistance. This work was supported by a grant from The Danish Cancer Society.