In this study, we have shown that aberrant DNA methylation of multiple genes characterizes advanced stages of CML and the disease when resistant to imatinib. Our interpretation is that the disease progression is associated with epigenetic changes including aberrant DNA methylation. We speculate that epigenetically mediated silencing of genes involved in drug transport may also affect the responsiveness of leukemic cells to imatinib. Given that CML starts as a genetically homogeneous disease, one can hypothesize that disease progression and clinical heterogeneity in CML are related to epigenetic factors including DNA hypermethylation. Our data extend previous observations on DNA methylation in CML and identify multiple new biomarkers in the disease. Of the genes specifically studied here, some may in fact contribute to the pathophysiology of disease progression. For instance, PDLIM4
, also known as RIL, is a LIM domain protein that has tumor-suppressor and pro-apoptotic properties. We have previously described a significant correlation between methylation and silencing of this gene 
was found hypermethylated and silenced in prostate cancer. Restoration of its expression suppressed tumor growth in xenografts 
. A recent report suggests that PDLIM4 is important for inactivation of Src and that epigenetic silencing of PDLIM4
may contribute to aberrant activation of Src in cancer 
. Hypermethylation of PDLIM4
in our sets of CML patients had a negative prognostic impact independent of the response to imatinib. We suggest that CML patients with epigenetically silenced PDLIM4
may particularly benefit from treatment with Src/Abl inhibitors.
, or transcription factor AP-2 epsilon, is a gene that we have found to have a potential tumor-suppressor function (and correlation between methylation and expression) 
, or H-cadherin, is a cell adhesion molecule with tumor-suppressor properties and an established correlation between methylation and expression 
. Methylation of CDKN2B
or the p15
tumor suppressor gene is frequently reported in myeloid malignancies 
. However, CDKN2B
was methylated only in 11% of the CML patients in our study, and its methylation did not show an association with shortened survival.
Aberrrant methylation of other genes likely represents passenger epigenetic defects that reflect the pressures to increase promoter DNA methylation in neoplastic cells. NPM2
, or nucleophosmin 2, is involved in forming nucleolus-like bodies in oocytes 
. It is aberrantly methylated in patients with acute myeloid leukemia 
and melanoma 
. Progesterone receptor methylation has been reported in leukemia 
and solid cancers 
. The role of DPYS
, or dihydropyrimidinase, in the hematopoietic system is currently unknown, and it may simply be a marker of methylation defects in CML as is the case in prostate and breast cancer 
Methylation of the OSCP1
gene (formerly known as C1orf102
) and its strong association with resistance to imatinib is intriguing. Aberrant methylation of OSCP1
has been shown in nasopharyngeal carcinoma 
and acute myeloid leukemia 
. This gene codes for an organic solute carrier protein with broad substrate specificity 
. The gene product may be involved in the transport of imatinib to target cells and its silencing may thus contribute to imatinib resistance.
We found strong concordant methylation for several of the genes tested, which cannot be explained simply by phase-specific methylation as concordant methylation also occurred within each phase. The genes involved do not share features, such as structure, chromosomal location, or function, and so this concordant methylation was likely caused by patient-specific pressures to increase DNA methylation, a phenomenon akin to the CpG island methylator phenotype described in colon cancer 
. The causes of this phenotype remain unknown, and whether the same factors that lead to it in solid tumors are involved in leukemia pathogenesis or progression remains to be determined.
Complex changes of DNA methylation in cancer can be summarized as focal hypermethylation of promoter CpG islands and global hypomethylation elsewhere, including repetitive elements 
. We have previously shown that methylation of the LINE-1 repetitive element in cancer and leukemia was highly variable 
. In this paper, we found that LINE-1 was methylated above the normal range in 55% of CML patients and only 3% of patients showed LINE-1 hypomethylation. This is in contrast with reports of LINE-1 hypomethylation and transcriptional activation in CML and progressive hypomethylation in the advanced phase of the disease 
Further analysis of the data on higher DNA methylation of OSCP1
and other genes in imatinib-resistant patients is warranted. It is possible that gene silencing provides an alternative to BCR-ABL
mutations in conferring imatinib resistance. Indeed, only about 50% of imatinib resistance can be conclusively traced to acquired mutations 
. It will therefore be worthwhile to study mutations and methylation simultaneously and determine whether there is an inverse correlation between the two events, and whether it is relevant to resistance to other tyrosine kinase inhibitors. It is also interesting to consider the fact that DNA methylation can be partially reversed by treatment with decitabine or azacitidine. Decitabine has demonstrated single-agent activity in CML 
, and a combination of decitabine and imatinib has shown a promising response rate in AP and BP 
. Given that many patients with blastic-phase CML continue to die of their disease, such drug combinations may be relevant even after therapy with second-generation tyrosine kinase inhibitors.