DNMT3A R882H Overexpression Acts in a Dominant Negative Manner to Cause DNA Hypomethylation and Increased Susceptibility to Hematopoietic Malignancies in Transgenic Mice

Somatic mutations in the DNA methyltransferase, DNMT3A, have been identified in >30% of de novo AML cases with a normal karyotype, and in >10% of patients with MDS and T-ALL. To understand whether mutations in DNMT3A alter hematopoietic development, we generated a transgenic mouse model capable of overexpressing either wild type human DNMT3A or the most common DNMT3A mutation found in AML cases (R882H, a hypomorphic variant that acts as a potent dominant negative inhibitor of WT DNMT3A, D. Germain et al., Cancer Cell 2014). Full-length human cDNAs encoding WT or R882H DNMT3A were cloned into a mammalian expression vector directly downstream from a tetracycline responsive element. This allows for the inducible expression of DNMT3A upon the expression of the rtTA coactivator, and the presence of Doxycycline (Dox). A single founder line for the WT DNMT3A allele, and two founder lines for the R882H DNMT3A allele, were established in the C57Bl6/J background. The WT DNMT3A transgene overexpressed 3.5x more human DNMT3A than endogenous murine DNMT3A in bone marrow cells; R882H DNMT3A transgenic line 1 expressed at a 4.5 fold excess, and R882H line 2 at a 16 fold excess.

To determine whether overexpression of the R882H allele was associated with focal DNA hypomethylation in the bone marrow cells of mice (similar to that observed in human AML samples), we used a novel CpG capture approach with bisulfite sequencing to assess 200,000 genomic regions containing ~3 million CpGs in the bone marrow cells of 3 WT C57Bl6/J mice, 3 Dnmt3a null mice, and healthy transgenic mice noted above that had been on Dox chow for either 6 months or 1 year (transgenic mice do not develop hematopoietic malignancies even after one year of transgene induction). We were able to assess 1.6 million CpGs with 10X or greater coverage in all 14 samples. The Dnmt3a null marrow samples contained 188,367 differentially methylated CpGs (average of >25% difference compared to WT bone marrow, q value=<0.01). Of these, 187,093 were hypomethylated (>99%); the hypomethylated CpGs were nearly identical in all three samples. Marrow cells from the two mice overexpressing the WT DNMT3A gene had only 338 differentially methylated CpGs compared to two matched rtTA control mice; of these, 337 were hypermethylated (>99%). For the two mice overexpressing the R882H allele in line 2 (16x overexpression), bone marrow cells had 2,356 differentially methylated CpGs, of which 2,316 were hypomethylated (98%). Of these CpGs, 1,745 (73%) overlapped with hypomethylated CpGs in the Dnmt3a null marrow samples, indicating that R882H overexpression causes hypomethylation in a subset of CpGs whose methylation in bone marrow cells is Dnmt3a dependent.

Because none of our mice developed hematologic malignancies even after one year, but had shown significant hypomethylation in the bone marrow, we hypothesized that cooperating mutations were necessary to produce malignancy. We transduced whole bone marrow cells from four transgenic mice: WT DNMT3A Tg x rtTA; R882H-1 Tg x rtTA; R882H-2 Tg x rtTA; and rtTA only (the same samples analyzed for methylation changes in the previous paragraph) with an MSCV-derived virus containing a human FLT3-ITD cDNA, and transplanted the transduced cells into 8-10 lethally irradiated recipients. Mice of all genotypes succumbed to myeloproliferative disease, T-cell lymphoma, T-lymphoma/ALL, or T-ALL. Overall median latencies were: rtTA=155 days, WT DNMT3A Tg x rtTA=164 days, R882H Tg-1 x rtTA=108.5 days, R882H-2 Tg x rtTA=135.5 days. The average latency for T cell malignancies demonstrated even greater differences among the four genotypes: rtTA n=4, 160.8 +/- 12.49 days (SEM), WT DNMT3A Tg x rtTA n=5, 167.3 +/- 4.854, R882H Tg-1 x rtTA n=3, 124.7 +/- 17.7, R882H-2 Tg x rtTA n=4 124.5 days +/- 22.14. T malignancies derived from R882H expressing cells were especially homogeneous compared to other groups; these tumors were CD4/CD8 double positive in all hematopoietic compartments. Despite the small sample size, these results demonstrate a trend towards a decreased latency for T malignancies in R882H expressing marrow cells, using a FLT3-ITD viral transduction model. We are confirming these data with additional mice. Taken together, our results demonstrate a clear focal hypomethyation phenotype in the bone marrow cells of DNMT3A R882H overexpressing mice, which may lead to increased susceptibility to neoplastic transformation.