Specific Patterns of DNA Remethylation in the Bone Marrow Cells of Dnmt3a Deficient Mice after Induced Expression of Wild Type Human DNMT3A

We previously identified recurrent mutations in the de novo DNA methyltransferase DNMT3A in patients with acute myeloid leukemia (AML).  The most common DNMT3A mutation in AML (R882H) creates a dominant negative protein that reduces DNA methylation activity by ~80% in AML cells, and causes canonical patterns of DNA hypomethylation in the AML genome (Russler-Germain et al, Cancer Cell 2014).  Approaches to restore DNMT3A activity in these AML genomes may be therapeutically relevant, but only if remethylation can return these genomes to their native methylation state.  To begin to address whether DNA methylation can be restored in hematopoietic cells that are deficient for Dnmt3a, we performed an Òadd-backÓ experiment using a transgenic mouse model system.

First, to characterize the genome-wide effect of Dnmt3a loss on DNA methylation in hematopoietic cells, we carried out whole-genome bisulfite sequencing (WGBS) on the total bone marrow cells of wild type (WT) mice, vs. marrow derived from littermates homozygous for a germline Dnmt3a null mutation (Dnmt3a^-/-, Okano et al Cell, 1999), which have overtly normal hematopoiesis. Total bone marrow cells from Dnmt3a^-/- mice have a canonical pattern of DNA hypomethylation at specific CpG residues and regions in the genome; many of the CpGs are virtually unmethylated in specific regions, suggesting that the normal Dnmt3a-dependent methylation ÒmarkÓ was added in stem/progenitor cells, and then maintained in all lineages.

To define the timing and specificity of remethylation in Dnmt3a-/- mice, we crossed heterozygotes from this line with transgenic mice containing a tetracycline-inducible human wild type DNMT3A cDNA (DNMT3A Tg mice), and a second transgenic mouse containing the rtTA coactivator, expressed from the Rosa26 locus. When DNMT3A Tg+, rtTA+ mice are fed Doxycycline (Dox) chow for one week, WT human DNMT3A is expressed in the marrow at a level ~4 times higher that of endogenous murine Dnmt3a. Since Dnmt3a^-/- mice die of severe runting at about three weeks of age, we harvested the marrow cells from Dnmt3a^-/-, DNMT3A Tg+, rtTA+ mice at 2 weeks of age, and transplanted them into lethally irradiated C57Bl/6 recipients. The marrow was allowed to engraft for four weeks. Half of the mice were then given Dox chow, and half were untreated. Whole bone marrow was isolated from pairs of mice (+ vs. - Dox), DNA was purified, and then subjected to whole genome bisulfite sequencing (WGBS).

WGBS produced methylation data on >93% of the CpGs in the mouse reference sequence with a median coverage of 10-12x per sample. Differential methylation analysis was performed on 2 kb tiled windows across the whole genome, revealing 108, 797 differentially methylated regions (DMRs) that were hypomethylated (Table 1). Dnmt3a^-/-, DNMT3A Tg+, rtTA+ bone marrow from mice without Dox (i.e. no DNMT3A was expressed) demonstrated no evidence for remethylation at any time after transplant.  However, if mice were treated with Dox for only 2 weeks, 59% of these DMRs were remethylated, increasing to 70% at 4 weeks, and 83% by week 9.  Data from weeks 12 and 24 are pending.

Patterns of remethylation for a subset of 560 differentially methylated CpGs (DMCpGs) are shown in Figure 1. These CpGs all had methylation values of  >= 90% in WT mouse bone marrow cells, and <=10% in Dnmt3a^-/- derived marrow. The methylation patterns for these CpGs is shown as a function of time, after Òadd-backÓ of WT DNMT3A (induced by Dox chow, upper panel). Individual CpGs are tracked by color-coded lines, which show the % methylation at each time point (red = > 66% methylated at a time point, yellow = 33%-66%, and green = < 33%).  Most CpGs are remethylated after only 2 weeks of DNMT3A expression, but some are delayed, and become remethylated later (yellow at week 2, changing to red at 4 or 9 weeks).  A small subset of these CpGs remethylate very slowly, if at all (green). Without the induction of DNMT3A (-Dox, lower panel), little or no remethylation is detected.  Inspection of specific regions of the genome showed that regions that are normally unmethylated in the bone marrow cells of WT mice are rarely methylated by adding back human DNMT3A in this system (data not shown), showing that remethylation is highly specific. Although the mechanisms involved in specifying remethylation patterns are not yet clear, these data may have important implications for therapeutically restoring DNMT3A activity in AML patients with DNMT3A mutations that reduce its activity.