Digital Restriction Enzyme Analysis of Methylation (DREAM) by Next Generation Sequencing Uncovers Epigenetic Disturbances In Acute Myelogenous Leukemia.

Abstract 3635

DNA methylation is a key epigenetic mark affecting the configuration of chromatin and the potential for gene expression. Disorganization of DNA methylation contributes to the development of leukemia. There is a need for high resolution, quantitative and cost effective methods to investigate changes of methylome in leukemia. To achieve this goal, we have recently developed a digital restriction enzyme analysis of methylation (DREAM) for quantitative mapping of DNA methylation at approximately 50,000 CpG sites across the whole genome (Jelinek et al., ASH 2009, abstract 567). The method is based on creating distinct DNA signatures at unmethylated or methylated CCCGGG sites by sequential restriction digests of genomic DNA with the SmaI and XmaI endonucleases and on resolving these signatures by massively parallel sequencing. Using the DREAM method, we have analyzed DNA methylation in bone marrow cells from 2 patients with AML, 3 samples of white blood cells from healthy adults and 2 myeloid leukemia cell lines (K562 and HEL). The first patient (Pt#1) was a 72 year-old male with AML transformation of the myelodysplastic syndrome (MDS). He had 32% blasts in the bone marrow and a complex karyotype. He had received lenalidomide treatment only. The second AML patient (Pt#2) was a 28 year-old male suffering from a relapse of an AML FAB M1. The bone marrow showed 87% of blasts and a complex karyotype. The patient was heavily pretreated with daunorubicin, ara-C, etoposide, 6-thioguanine, dexamethasone and l-asparaginase. Neither of the patients received demethylating drugs. Using typically 2 sequencing lanes per sample and paired-end reads of 36 bases on the Illumina Gene Analyzer II platform, we acquired 20–38 (median 33) million sequence tags per sample; of these, 7–17 (median 12) million were mapped to SmaI/XmaI sites unique in the human genome. With a threshold of minimum 20-fold coverage, we obtained quantitative information on the DNA methylation level of 39,603-53,312 (median 44,490) CpG sites associated with 8,939-10,735 (median 9,517) genes. In general, methylation was largely absent within CpG islands (CGI). The CpG sites most protected from methylation were in CGI and within 1 kb from gene transcription start sites (TSS). These regions were represented by 13,474 CpG sites. Focusing our analysis on these CpG sites, methylation >10% was detected only in 268 sites in normal controls (1.9%). The numbers of sites with methylation >10% were significantly higher (P<.0001, chi-square test) in both AML patients: 397 sites in Pt#1 (2.9%) and 2,143 sites in Pt#2 (15.6%), respectively. Leukemia cell lines mirrored the pattern of CGI hypermethylation seen in primary AML cells. Methylation >10% in CGI within 1 kb from TSS was observed at 2,331 sites (17.0%) in K562 and at 2,484 sites (18.1%) in HEL. Differential hypermethylation in AML patients affected 906 genes, including multiple genes previously shown to be methylated in cancer, such as CDKN1B, FOXO3, GATA2, GATA4, GDNF, HOXA9, IGFBP3, SALL1 and WT1. Methylated genes were significantly enriched in canonical pathways affecting embryonic stem cell signaling, Wnt-beta-catenin signaling and pluripotency suggesting an important role in AML stem cells. In contrast to CGI, it is known that CpG sites outside of CpG islands (NCGI) are generally fully methylated in normal cells. We analyzed 11,220 NCGI sites that were >1 kb from gene TSS. Methylation >90% was observed at 5,217 (46%) sites in normal controls, in 5,380 sites (48%) in Pt#1, while only in 1,873 sites (17%) in Pt#2 (P<.0001). Leukemia cell lines also showed this NCGI hypomethylation with only 1,422 (13%) fully methylated sites in K562 and 4,200 sites (37%) in HEL. Thus, significant degrees of hypomethylation in NCGI were observed in Pt#2, and in K562 and HEL cell lines, but not in Pt#1. In conclusion, high resolution quantitative mapping of DNA methylation changes in leukemia is feasible using the DREAM method. Relatively small alterations in DNA methylation observed in the MDS/AML Pt#1 contrasted with extensive hyper and hypomethylation found in Pt#2 with relapsed AML M1. Our results illustrate the complexity and diverse extent of DNA methylation changes in leukemia.