Cytogenetically normal acute myeloid leukemia (CN-AML) represents nearly 50% of human AML. Co-occurring mutations in the de novo DNA methyltransferase DNMT3A and the FMS related tyrosine kinase 3 (FLT3) are common in CN-AML and confer a poorer prognosis. We demonstrate that mice with Flt3-internal tandem duplication (Flt3ITD) and inducible deletion of Dnmt3a spontaneously develop a rapidly lethal, completely penetrant, and transplantable AML of normal karyotype. AML cells retain a single Dnmt3a floxed allele, revealing the oncogenic potential of Dnmt3a haploinsufficiency. FLT3ITD/DNMT3A-mutant primary human and murine AML exhibit a similar pattern of global DNA methylation associated with changes in the expression of nearby genes. In the murine model, rescuing Dnmt3a expression was accompanied by DNA remethylation and loss of clonogenic potential, suggesting that Dnmt3a-mutant oncogenic effects are reversible. Dissection of the cellular architecture of the AML model using single-cell assays, including single-cell RNA sequencing, identified clonogenic subpopulations that express genes sensitive to the methylation of nearby genomic loci and responsive to DNMT3A levels. Thus, Dnmt3a haploinsufficiency transforms Flt3ITD myeloproliferative disease by modulating methylation-sensitive gene expression within a clonogenic AML subpopulation.

Significance:DNMT3A haploinsufficiency results in reversible epigenetic alterations that transform FLT3ITD-mutant myeloproliferative neoplasm into AML. Cancer Discov; 6(5); 501–15. ©2016 AACR.

This article is highlighted in the In This Issue feature, p. 461

FLT3 is one of the most frequently mutated genes in cytogenetically normal acute myeloid leukemia (CN-AML), affecting 37% of cases (1–3). FLT3-internal tandem duplication (ITD) mutations are associated with a worse overall survival and constitute an independent prognostic factor for relapse and poor outcome in AML (4–7). Recently, recurrent mutations were found in the epigenetic regulators DNMT3A and TET2 (2, 3, 8–11). It is now recognized that DNA methyltransferase DNMT3A is one of the more commonly mutated genes in AML, affecting 23% of cases, and is associated with an unfavorable prognosis (3, 12–14). Intriguingly, 36% to 44% of patients with FLT3-mutant AML also carry DNMT3A mutations. Furthermore, the co-occurrence of DNMT3A and FLT3 mutants changes the risk classification to a poorer prognosis (3, 12, 13).

The mammalian DNA methyltransferases DNMT1, DNMT3A, and DNMT3B methylate the cytosine residue at 5 position (5mC) in 5′-C-phosphate-G-3′ (CpG) dinucleotides. Methylation of CpG sites at promoter-associated CpG islands is thought to silence expression of neighboring genes. In addition, methylation of CpG dinucleotides can also occur at the gene body or intergenic regions, where its role in gene regulation is less clear (15). Different molecular subtypes of CN-AML have highly divergent DNA methylation and transcription profiles (16, 17). Recurring missense mutations in DNMT3A identified in CN-AML act as loss of function or dominant negatives, significantly reducing the methyltransferase activity of DNMT3A by approximately 80% (11, 18, 19). Although the mechanisms of DNMT3A action in the pathogenesis of AML are still unclear, studies on stem cells demonstrate essential roles for all DNMT proteins. DNMT1 is required to maintain global DNA methylation through successive rounds of cell division, whereas the de novo methyltransferases DNMT3A and DNMT3B are required for mouse embryonic development (20). Serial transplantation of Dnmt3a knockout hematopoietic stem/progenitor cells (HSPC) results in impaired differentiation, enhanced proliferation, and differential methylation of distinct genetic loci, resulting in upregulation of stemness genes and proto-oncogenes and downregulation of differentiation genes (21). Despite this, Dnmt3a knockout HSPC-reconstituted mice do not develop AML.

FLT3ITD-mutant receptors undergo ligand-independent receptor dimerization, autophosphorylation, and constitutive activation (22–24). Mouse knock-in and human cell models of FLT3ITD demonstrate that FLT3-mutant receptors promote aberrant cell proliferation through activation of STAT5, AKT/ERK, and c-MYC pathways and impair differentiation by altering expression of myeloid transcription factors such as CEBPα (25–32). Despite the high frequency of FLT3 mutations in AML, murine models of Flt3ITD develop a mixture of myeloid and lymphoid neoplasms either as a single mutation or in combination with other oncogenic events, such as MLL partial tandem duplication (MllPTD; refs. 25, 27, 33–35). Importantly, Flt3ITD mice do not progress to leukemia; thus, other cooperating mutations must be required for development of AML.

To test whether Flt3 and Dnmt3a mutations cooperate to initiate leukemogenesis, we bred mice to combine Flt3ITD knock-in and homozygous Dnmt3a floxed alleles (Dnmt3afl/fl; Fig. 1A) with Mx1-Cre (MxCre). We found that Flt3+/ITD;Dnmt3afl/flMxCre mice (with one mutant Flt3 allele) had a long median survival of 227 days and exhibited splenomegaly (Supplementary Fig. S1A). Control Flt3+/ITD;Dnmt3afl/fl mice did not die in this time period. Conversely, doubling the number of Flt3ITD alleles significantly shortened survival, with a median of 43 days after birth (Fig. 1B, left and lower table). All Flt3ITD/ITD;Dnmt3afl/flMxCre mice exhibited splenomegaly (Fig. 1B, right) and were moribund by 60 days. Secondary and tertiary transplantation of Flt3ITD/ITD;Dnmt3afl/flMxCre splenocytes resulted in lethality, albeit with delayed latency (Supplementary Fig. S1B and data not shown). Because the disease occurred spontaneously (pIpC induction of MxCre expression was not required), we examined the Dnmt3afl/fl alleles (Fig. 1C). As expected, tail DNA from 2-week-old Flt3ITD/ITD;Dnmt3afl/flMxCre mice showed amplification of LoxP-flanked Dnmt3a exon 18 (Fig. 1C; “floxed” indicates no recombination; “deleted” indicates recombination). However, total or fractionated c-KIT+ (CD117) stem/progenitor enriched splenocytes from Flt3ITD/ITD;Dnmt3afl/flMxCre de novo or secondary transplant recipients demonstrated both floxed and deleted Dnmt3a alleles (Fig. 1C). In addition, we plated c-KIT+Flt3ITD/ITD;Dnmt3afl/flMxCre hematopoietic cells in methylcellulose, isolated DNA from individual colonies, and then analyzed floxed Dnmt3a alleles. Of 77 colonies analyzed, 28 lacked detectable floxed alleles (36%); thus, the majority of cells spontaneously deleted only one allele of Dnmt3a (Supplementary Fig. S1C). In agreement with this, the mRNA from c-KIT+ cells from Flt3ITD/ITD;Dnmt3afl/flMxCre mice showed sporadic low-level expression of Cre and approximately 50% reduction in Dnmt3a expression (Fig. 1D). These data suggest that sporadic MxCre expression leads to excision of floxed Dnmt3a alleles, and that Flt3ITD/ITD cells with a single Dnmt3a allele may have a selective advantage.

Figure 1.

A rapid, spontaneous, and fully penetrant model of cytogenetically normal AML utilizing Flt3ITD and Dnmt3a haploinsufficiency. A, schematic of mouse alleles with constitutive Flt3ITD knock-in (Flt3ITD) and Dnmt3a exon 18 flanked by LoxP sites (Dnmt3afl). B, Kaplan–Meier survival curves (left) and spleen (sp.) weights (right) of Flt3ITD/ITD;Dnmt3afl/fl mice without (blue squares) and with Mx1-Cre (MxCre) (red squares). The table beneath summarizes the number of animals, median survivals, and mean spleen weights of mice in B. Significant differences in survival were evaluated by log-rank (Mantel–Cox) test and by unpaired parametric t test. C, representative semiquantitative PCR detection of Dnmt3a LoxP sites from tail or c-KIT+ fractionated leukemic splenocyte genomic DNA from primary and secondary transplant recipient Flt3ITD/ITD;Dnmt3afl/flMxCre mice (“deleted” indicates deletion of a single floxed Dnmt3a exon 18; “floxed” indicates no recombination). Additional 5 independent Flt3ITD/ITD;Dnmt3afl/flMxCre showed similar results (see also Supplementary Fig. 1C). D, RT-qPCR for Dnmt3a (left) and Cre (right) expression in c-KIT+ leukemic splenocytes from individual Flt3ITD/ITD;Dnmt3afl/flMxCre and Flt3ITD/ITD;Dnmt3afl/fl mice (n = 5/group). Significance was determined by unpaired parametric t test. E, representative Wright Giemsa–stained bone marrow cytospins and hematoxylin and eosin (H&E)–stained bone marrow, spleen, and liver from WT, Flt3ITD/ITD;Dnmt3afl/fl, and Flt3ITD/ITD;Dnmt3afl/flMxCre mice. Top panels, cytospins: normal WT marrow cytospin shows a mixture of erythroid (e) and myeloid elements, including neutrophilic (N), intermediate (Int), and monocytic (Mo; lymphocytes present in other microscopic fields). Flt3ITD/ITD;Dnmt3afl/fl marrow is predominantly myelomonocytic, with increased monocytes, increased intermediate myeloid forms and few immature (Im) forms. Flt3ITD/ITD;Dnmt3afl/flMxCre marrow has numerous immature forms and intermediate myeloid forms. Second row panels: marrow H&E shows bone marrow histology consistent with the cellular constituents seen in cytospins; note numerous dark nuclei of erythroids in WT marrow, mixture of maturing myelomonocytic elements in Flt3ITD/ITD;Dnmt3afl/fl, and predominance of less mature forms (including immature forms/blasts) in Flt3ITD/ITD;Dnmt3afl/flMxCre. Third row panels, spleen H&E: In WT spleen, there is a clearly demarcated lymphoid area of white pulp (WP), and red pulp (RP) is predominantly erythroid. In Flt3ITD/ITD;Dnmt3afl/fl spleen, there is some alteration of lymphocyte morphology (more cytoplasm, including an increase in lymphocytes with a marginal zone morphology) in the white pulp, and the red pulp is expanded with mixed myeloid hyperplasia (erythroid, megakaryocytes, neutrophils, and monocytic cells). In Flt3ITD/ITD;Dnmt3afl/flMxCre spleen there is a loss of white pulp. The area indicated by white arrow is not lymphoid but rather is composed predominantly of myelomonocytic cells, including immature elements. The larger areas of red pulp (*) are expanded with mixed myeloid hyperplasia (erythroid, megakaryocytes, neutrophils, monocytic cells). F, average ± SEM number of colonies arising from c-KIT+ splenocytes from Flt3ITD/ITD;Dnmt3afl/fl and Flt3ITD/ITD;Dnmt3afl/flMxCre mice after 7 days in methylcellulose. Each bar represents individual mice (n = 2/group). G, spectral karyotyping analysis of a representative Flt3ITD/ITD;Dnmt3afl/flMxCre bone marrow illustrating the inverted image of the DAPI-stained metaphase cell (top left), the spectral image (bottom left), spectral karyotype (bottom right), and the karyotype from the classified image (top right). The karyotype was 40,XY. The table at bottom summarizes the karyotypes found in individual leukemic mice by spectral karyotyping (n = 5).

Figure 1.

A rapid, spontaneous, and fully penetrant model of cytogenetically normal AML utilizing Flt3ITD and Dnmt3a haploinsufficiency. A, schematic of mouse alleles with constitutive Flt3ITD knock-in (Flt3ITD) and Dnmt3a exon 18 flanked by LoxP sites (Dnmt3afl). B, Kaplan–Meier survival curves (left) and spleen (sp.) weights (right) of Flt3ITD/ITD;Dnmt3afl/fl mice without (blue squares) and with Mx1-Cre (MxCre) (red squares). The table beneath summarizes the number of animals, median survivals, and mean spleen weights of mice in B. Significant differences in survival were evaluated by log-rank (Mantel–Cox) test and by unpaired parametric t test. C, representative semiquantitative PCR detection of Dnmt3a LoxP sites from tail or c-KIT+ fractionated leukemic splenocyte genomic DNA from primary and secondary transplant recipient Flt3ITD/ITD;Dnmt3afl/flMxCre mice (“deleted” indicates deletion of a single floxed Dnmt3a exon 18; “floxed” indicates no recombination). Additional 5 independent Flt3ITD/ITD;Dnmt3afl/flMxCre showed similar results (see also Supplementary Fig. 1C). D, RT-qPCR for Dnmt3a (left) and Cre (right) expression in c-KIT+ leukemic splenocytes from individual Flt3ITD/ITD;Dnmt3afl/flMxCre and Flt3ITD/ITD;Dnmt3afl/fl mice (n = 5/group). Significance was determined by unpaired parametric t test. E, representative Wright Giemsa–stained bone marrow cytospins and hematoxylin and eosin (H&E)–stained bone marrow, spleen, and liver from WT, Flt3ITD/ITD;Dnmt3afl/fl, and Flt3ITD/ITD;Dnmt3afl/flMxCre mice. Top panels, cytospins: normal WT marrow cytospin shows a mixture of erythroid (e) and myeloid elements, including neutrophilic (N), intermediate (Int), and monocytic (Mo; lymphocytes present in other microscopic fields). Flt3ITD/ITD;Dnmt3afl/fl marrow is predominantly myelomonocytic, with increased monocytes, increased intermediate myeloid forms and few immature (Im) forms. Flt3ITD/ITD;Dnmt3afl/flMxCre marrow has numerous immature forms and intermediate myeloid forms. Second row panels: marrow H&E shows bone marrow histology consistent with the cellular constituents seen in cytospins; note numerous dark nuclei of erythroids in WT marrow, mixture of maturing myelomonocytic elements in Flt3ITD/ITD;Dnmt3afl/fl, and predominance of less mature forms (including immature forms/blasts) in Flt3ITD/ITD;Dnmt3afl/flMxCre. Third row panels, spleen H&E: In WT spleen, there is a clearly demarcated lymphoid area of white pulp (WP), and red pulp (RP) is predominantly erythroid. In Flt3ITD/ITD;Dnmt3afl/fl spleen, there is some alteration of lymphocyte morphology (more cytoplasm, including an increase in lymphocytes with a marginal zone morphology) in the white pulp, and the red pulp is expanded with mixed myeloid hyperplasia (erythroid, megakaryocytes, neutrophils, and monocytic cells). In Flt3ITD/ITD;Dnmt3afl/flMxCre spleen there is a loss of white pulp. The area indicated by white arrow is not lymphoid but rather is composed predominantly of myelomonocytic cells, including immature elements. The larger areas of red pulp (*) are expanded with mixed myeloid hyperplasia (erythroid, megakaryocytes, neutrophils, monocytic cells). F, average ± SEM number of colonies arising from c-KIT+ splenocytes from Flt3ITD/ITD;Dnmt3afl/fl and Flt3ITD/ITD;Dnmt3afl/flMxCre mice after 7 days in methylcellulose. Each bar represents individual mice (n = 2/group). G, spectral karyotyping analysis of a representative Flt3ITD/ITD;Dnmt3afl/flMxCre bone marrow illustrating the inverted image of the DAPI-stained metaphase cell (top left), the spectral image (bottom left), spectral karyotype (bottom right), and the karyotype from the classified image (top right). The karyotype was 40,XY. The table at bottom summarizes the karyotypes found in individual leukemic mice by spectral karyotyping (n = 5).

Close modal

Examination of the blood, bone marrow, spleen, and liver of moribund Flt3ITD/ITD;Dnmt3afl/flMxCre mice demonstrated pathologic features of myeloid leukemia. The peripheral blood was notable for anemia and for leukocytosis of neutrophilic and monocytic cells, including immature forms (Supplementary Table S1 and data not shown). Bone marrow contained immature forms/blasts that were greater than 20% of nucleated cells along with numerous intermediate myeloid forms (i.e., non-blast myelomonocytic cells that have not differentiated sufficiently to be distinguished into monocytic or neutrophilic cells by morphology; ref. 36). Similar to Flt3ITD/ITD;Dnmt3afl/fl mice, there was a marked decrease in erythroid and lymphoid cells relative to wild-type (WT) marrow (Fig. 1E, top two rows). The massively enlarged spleens of the Flt3ITD/ITD;Dnmt3afl/flMxCre mice contained areas of predominantly immature cells (similar to those seen in the marrow), but most of the red pulp contained a mix of maturing neutrophilic, monocytic, erythroid, and megakaryocytic cells (Fig. 1E, third row). Concordant with the morphology, flow cytometric immunophenotyping of erythroid-depleted bone marrows and spleens of Flt3ITD/ITD;Dnmt3afl/flMxCre mice showed decreased lymphocytes (data not shown and Supplementary Fig. S1D) and increased monocytic cells (CD11b+, Gr1, B220) in comparison to WT mice, and increased immature cells [LINc-KIT+ (LK)] in comparison to Flt3ITD/ITD;Dnmt3afl/fl mice (Supplementary Fig. S1E–S1I; Supplementary Table S2). In contrast to the Flt3ITD/ITD;Dnmt3afl/fl mice, myelomonocytic cells aggressively invaded the perivascular and sinusoidal regions of the liver (Fig. 1E, fourth row). These pathologic findings are indicative of myelomonocytic leukemia. The presence of greater than 20% immature forms/blasts in the bone marrow of Flt3ITD/ITD;Dnmt3afl/flMxCre mice in combination with the rapid lethality observed in the primary animals warranted a diagnosis of acute leukemia (36). In agreement with previous reports, Flt3ITD/ITD;Dnmt3afl/fl control mice (Fig. 1E, second column) showed expansion of both mature and immature myeloid cells, consistent with myeloproliferative neoplasm (MPN; ref. 27). According to these diagnoses, we hereon refer to Flt3ITD/ITD;Dnmt3afl/fl mice as MPN and Flt3ITD/ITD;Dnmt3afl/flMxCre mice as AML.

To assess the ability of the leukemic cell populations to give rise to colony-forming units (CFU), we seeded purified c-KIT+ and c-KIT cells from moribund mice into methylcellulose. The c-KIT+ fraction from several different donor mice gave rise to significantly more colonies than the c-KIT fraction, compatible with c-KIT+ leukemic blasts (Supplementary Fig. S1J). Moreover, c-KIT+ splenocytes from AML mice had significantly increased colony-forming ability compared to c-KIT+ splenocytes from MPN mice (Fig. 1F). No differences in colony morphologies or numbers between c-KIT+ bone marrow and c-KIT+ splenocytes from AML mice were observed (data not shown). We concluded that loss of one Dnmt3a allele is sufficient to increase the clonogenic potential of Flt3ITD myeloid progenitors.

Because FLT3ITD and DNMT3A mutations are most common in patients with AML with normal karyotype, we performed spectral karyotyping analyses on bone marrow from five independent murine AMLs (Fig. 1G). Four of five AMLs exhibited normal karyotypes, and one had a small clone with loss of the Y chromosome (39,X,-Y[4]/40,XY[16]). Note that loss of the Y chromosome is observed in both malignant and nonmalignant states and is not considered to be a driver mutation. Like their human counterparts with similar mutations, the murine model tumors were cytogenetically normal AML.

Because the murine AML express less Dnmt3a mRNA, we next investigated whether our model recapitulated the epigenetic abnormalities seen in human DNMT3A-mutant AML, where activity of the enzyme is compromised due to the dominant-negative nature of the mutation (11, 18, 19). Using Enhanced Reduced Representation Bisulfite Sequencing (ERRBS; ref. 37), we derived the disease-specific DNA-methylation profiles in both human and murine AML with similar mutation profiles. First, we compared primary human AML carrying both FLT3ITD and DNMT3A mutations (FLT3ITD/DNMT3A-mutant) versus FLT3ITD AML lacking mutations in DNMT3A (FLT3ITD/DNMT3AWT). To ensure that the DNA methylation differences we observed could be explained only by the DNMT3A status, patients with mutations in other genes related to the DNA methylation pathway (TET2, IDH1, and IDH2) were excluded from both groups. Differentially methylated regions (DMR) with an FDR of <10% and a mean methylation difference of ≥25% were identified using a beta-binomial model. Patients with FLT3ITD/DNMT3A-mutant AML presented with a predominantly hypomethylated profile compared to FLT3ITD/DNMT3AWT cases (95.9% hypomethylated regions of 25,746 total DMR). These DMRs in FLT3ITD/DNMT3A-mutant AML were significantly enriched at CpG shores (22% of DMRs at CpG shores vs. 15% in background, P < 2.2 × 10−16) and enhancer regions (33% of DMRs at intragenic enhancers vs. 19% in background, P < 2.2 × 10−16 and 24% at intergenic enhancers versus 15% in background, P < 2.2 × 10−16; Fig. 2A–C; Supplementary Table S3). Next, we compared the murine AML versus MPN c-KIT+ splenocytes. Similar to our findings in human AML, Dnmt3a loss was associated with predominance of hypomethylation (80.7% hypomethylated regions of 503 total DMRs) with enrichment of DMRs at enhancer regions (9% of DMRs at intragenic enhancers vs. 3% background, P = 1.519 × 10−10 and 6% of DMRs at intergenic enhancers vs. 2% background, P = 6.005 × 10−7; Fig. 2D–F; Supplementary Table S3). Importantly, we found that 61.8% of the DMRs identified in the murine models (Flt3ITD/ITD;Dnmt3afl/flMxCre AML vs. Flt3ITD/ITD;Dnmt3afl/fl MPN) overlapped with the human FLT3ITD/DNMT3A-mutant signature.

Figure 2.

Genome-wide DNA hypomethylation associated with DNMT3A mutation in murine and human FLT3ITD AML. Human: AC, volcano plot (A) representation of mean methylation difference (x-axis) versus statistical significance (y-axis). Red dots designate significantly hypo- and hypermethylated regions in patients with FLT3ITD/DNMT3A-mutant AML compared to patients with FLT3ITD alone. B, heatmap representation of DMRs in FLT3ITD/DNMT3A-mutant AML compared to FLT3ITD/DNMT3AWT AML. Corresponding methylation levels are also shown in normal human bone marrow HSCs for normal baseline comparison. C, proportion of all measured CpGs by ERRBS (all tiles) and differentially methylated in FLT3ITD/DNMT3A-mutant versus FLT3ITD/DNMT3AWT AML (DMRs, hypermethylated, and hypomethylated) in genomic, CpG island (CpGi), and enhancer regions. Murine (n = 3 mice/group): DF, volcano plot (D) representation of mean methylation difference (x-axis) versus statistical significance (y-axis). Red dots designate significantly DMRs in Flt3ITD/ITD;Dnmt3afl/flMxCre AML compared to Flt3ITD/ITD;Dnmt3afl/fl MPN. E, heatmap representation of DMRs in Flt3ITD/ITD;Dnmt3afl/flMxCre AML compared to Flt3ITD/ITD;Dnmt3afl/fl MPN. Corresponding methylation levels are also shown in WT murine LINSCA-1+c-KIT+ (LSK) as a normal baseline comparison. HSPC-like genes share similar methylation patterns with WT LSK (indicated by gray sidebars), and AML-unique regions are differentially methylated from both MPN and WT LSK (indicated by gold sidebars). F, proportion of differentially methylated genomic, CpGi, and enhancer regions in all measured CpGs (all tiles) and DMRs (hypermethylated and hypomethylated) in Flt3ITD/ITD;Dnmt3afl/flMxCre AML compared to Flt3ITD/ITD;Dnmt3afl/fl MPN murine models. G, gene expression: hierarchical clustering of significantly differentially expressed genes in Flt3ITD/ITD;Dnmt3afl/flMxCre AML compared to Flt3ITD/ITD;Dnmt3afl/fl MPN. Gene expression is also shown for WT LSK, common myeloid progenitors (CMP), and granulocyte–monocyte progenitors (GMP) murine bone marrow cells as normal baseline comparison. “HSPC-like” genes are expressed in AML and LSK but not in CMP, GMP, or MPN. “AML-unique” genes are differentially expressed in AML compared to all other samples. H, GSEA: genes upregulated in murine AML versus MPN were ranked according to descending gene expression in human FLT3ITD/DNMT3A-mutant AML versus FLT3ITD/DNMT3AWT AML. I, dot plot representation of mean methylation difference for all differentially methylated genes in c-KIT+Flt3ITD/ITD;Dnmt3afl/flMxCre AML compared to Flt3ITD/ITD;Dnmt3afl/fl MPN (x-axis), versus their corresponding gene expression fold change (y-axis) (n = 3 mice/group). Genes whose gene expression changes overlapped with changes in methylation are highlighted in red. J, GSEA: genes corresponding to hypomethylated regions in murine Flt3ITD/ITD;Dnmt3afl/flMxCre AML (compared to MPN) were ranked according to descending gene expression in murine AML versus MPN. K, RT-qPCR validation of representative genes from AML-unique and HSPC-like groups in G that were hypomethylated in murine c-KIT+ splenocytes from Flt3ITD/ITD;Dnmt3afl/flMxCre AML (red boxes) versus Flt3ITD/ITD;Dnmt3afl/fl MPN (blue boxes) (n = 5 mice/group). Statistical significance was evaluated for each gene by unpaired parametric t test compared to Flt3ITD/ITD;Dnmt3afl/fl control. *, P < 0.05; ***, P < 0.001; ****, P < 0.0001.

Figure 2.

Genome-wide DNA hypomethylation associated with DNMT3A mutation in murine and human FLT3ITD AML. Human: AC, volcano plot (A) representation of mean methylation difference (x-axis) versus statistical significance (y-axis). Red dots designate significantly hypo- and hypermethylated regions in patients with FLT3ITD/DNMT3A-mutant AML compared to patients with FLT3ITD alone. B, heatmap representation of DMRs in FLT3ITD/DNMT3A-mutant AML compared to FLT3ITD/DNMT3AWT AML. Corresponding methylation levels are also shown in normal human bone marrow HSCs for normal baseline comparison. C, proportion of all measured CpGs by ERRBS (all tiles) and differentially methylated in FLT3ITD/DNMT3A-mutant versus FLT3ITD/DNMT3AWT AML (DMRs, hypermethylated, and hypomethylated) in genomic, CpG island (CpGi), and enhancer regions. Murine (n = 3 mice/group): DF, volcano plot (D) representation of mean methylation difference (x-axis) versus statistical significance (y-axis). Red dots designate significantly DMRs in Flt3ITD/ITD;Dnmt3afl/flMxCre AML compared to Flt3ITD/ITD;Dnmt3afl/fl MPN. E, heatmap representation of DMRs in Flt3ITD/ITD;Dnmt3afl/flMxCre AML compared to Flt3ITD/ITD;Dnmt3afl/fl MPN. Corresponding methylation levels are also shown in WT murine LINSCA-1+c-KIT+ (LSK) as a normal baseline comparison. HSPC-like genes share similar methylation patterns with WT LSK (indicated by gray sidebars), and AML-unique regions are differentially methylated from both MPN and WT LSK (indicated by gold sidebars). F, proportion of differentially methylated genomic, CpGi, and enhancer regions in all measured CpGs (all tiles) and DMRs (hypermethylated and hypomethylated) in Flt3ITD/ITD;Dnmt3afl/flMxCre AML compared to Flt3ITD/ITD;Dnmt3afl/fl MPN murine models. G, gene expression: hierarchical clustering of significantly differentially expressed genes in Flt3ITD/ITD;Dnmt3afl/flMxCre AML compared to Flt3ITD/ITD;Dnmt3afl/fl MPN. Gene expression is also shown for WT LSK, common myeloid progenitors (CMP), and granulocyte–monocyte progenitors (GMP) murine bone marrow cells as normal baseline comparison. “HSPC-like” genes are expressed in AML and LSK but not in CMP, GMP, or MPN. “AML-unique” genes are differentially expressed in AML compared to all other samples. H, GSEA: genes upregulated in murine AML versus MPN were ranked according to descending gene expression in human FLT3ITD/DNMT3A-mutant AML versus FLT3ITD/DNMT3AWT AML. I, dot plot representation of mean methylation difference for all differentially methylated genes in c-KIT+Flt3ITD/ITD;Dnmt3afl/flMxCre AML compared to Flt3ITD/ITD;Dnmt3afl/fl MPN (x-axis), versus their corresponding gene expression fold change (y-axis) (n = 3 mice/group). Genes whose gene expression changes overlapped with changes in methylation are highlighted in red. J, GSEA: genes corresponding to hypomethylated regions in murine Flt3ITD/ITD;Dnmt3afl/flMxCre AML (compared to MPN) were ranked according to descending gene expression in murine AML versus MPN. K, RT-qPCR validation of representative genes from AML-unique and HSPC-like groups in G that were hypomethylated in murine c-KIT+ splenocytes from Flt3ITD/ITD;Dnmt3afl/flMxCre AML (red boxes) versus Flt3ITD/ITD;Dnmt3afl/fl MPN (blue boxes) (n = 5 mice/group). Statistical significance was evaluated for each gene by unpaired parametric t test compared to Flt3ITD/ITD;Dnmt3afl/fl control. *, P < 0.05; ***, P < 0.001; ****, P < 0.0001.

Close modal

We next determined the DNA methylation status of these DMRs in normal human and murine HSPC. We note that the methylation of a subset of DMR is similar in both DNMT3A-mutant leukemic and normal human hematopoietic stem cells (HSC) or murine AML and LINSCA-1+c-KIT+ (LSK) cells (Fig. 2B, 2E; “HSPC-like” DMRs; Supplementary Table S3), suggesting that Dnmt3a-mutant AML cells lack adequate DNMT3A levels to mediate normal differentiation-associated methylation of these genomic loci. In contrast, other DMRs were not found in normal progenitors and, instead, displayed methylation patterns unique to the DNMT3A-mutant human and mouse AMLs (Fig. 2B, 2E; “AML-unique” DMRs; Supplementary Table S3). These data suggested that although Dnmt3a loss resulted in disease-specific DMRs, it also allowed the persistence of HSPC-like methylation profiles in the leukemic blasts despite these cells expressing phenotypic differentiation markers that suggest maturity beyond HSPC.

Human AML is known to exhibit altered DNA methylation relative to normal HSPC (38, 39). As expected, we found that human FLT3ITD AML exhibits extensive global DNA methylation changes compared to normal human HSC, irrespective of DNMT3A status (Supplementary Fig. S2A and S2B). However, comparison of FLT3ITDDNMT3AWT versus FLT3ITDDNMT3A-mutant AML revealed a more restricted pattern of predominantly hypomethylated DNA associated with DNMT3A mutation (Supplementary Fig. S2C). Similarly, comparison of murine MPN or AML versus wild-type LSK marrow cells revealed dramatic differences in global DNA methylation patterns (Supplementary Fig. S2D–E); however, comparing murine MPN versus murine AML revealed a more subtle global consequence of Dnmt3a haploinsufficiency that predominantly resulted in DNA hypomethylation (Supplementary Fig. S2F).

To understand whether these changes in global DNA methylation correspond to changes in gene expression, we performed RNA sequencing on c-KIT+ splenocytes from murine AML and MPN (Fig. 2G; Supplementary Table S4). Similar to our findings in the methylome, the expression of some genes in murine AML blasts was similar to LSK marrow cells (“HSPC-like”; Fig. 2G; Supplementary Table S4), whereas a group of mostly upregulated genes were uniquely expressed in AML compared to LSK or MPN (“AML-unique”; Fig. 2G; Supplementary Table S4). In agreement with prior studies indicating that DNMT3A regulates HSPC function (21, 40), gene set enrichment analyses (GSEA) demonstrated that genes upregulated in murine AML are enriched for HSPC gene expression as well as genes downregulated during myeloid development (Supplementary Fig. S3A; Supplementary Table S4; refs. 41, 42). To evaluate the relevance of murine Flt3ITD/ITD;Dnmt3afl/flMxCre AML gene expression in human AML, we identified genes that are significantly upregulated in murine AML versus MPN, then curated a gene set of their human orthologs (Supplementary Table S4). When we compared gene expression data from human FLT3ITD AML with either wild-type or mutant DNMT3A, GSEA demonstrated that gene expression in DNMT3A-mutant AML is significantly enriched for this gene set (Fig. 2H). We then performed correlation (Fig. 2I) and GSEA analyses (Fig. 2J; Supplementary Table S4) between murine AML DMRs and gene expression and found that the majority of overexpressed genes (60%) overlap with sites of DNA hypomethylation events in the murine AML (Fig. 2I, upper left quadrant red dots). Finally, we validated candidate genes with Flt3ITD/ITD;Dnmt3afl/flMxCre AML-unique (Emilin2 and Pim1) and HSPC-like (Gata3) DMRs and associated gene upregulation (Fig. 2K). We concluded that deletion of a single Dnmt3a allele results in functional DNMT3A haploinsufficiency, and in the context of Flt3ITD/ITD recapitulates a pattern of methylation and gene expression seen in human AML with FLT3ITD and DNMT3A mutations.

GSEA also revealed that murine AML-upregulated genes are enriched for c-MYC–target gene sets (Supplementary Fig. S3B and Supplementary Table S4). Therefore, we genetically incorporated a knock-in allele that generates an eGFP–c-MYC fusion protein from the endogenous c-Myc locus (c-MyceGFP;Flt3ITD/ITD;Dnmt3afl/flMxCre AML vs. c-MyceGFP;Flt3ITD/ITD;Dnmt3afl/fl MPN; ref. 43). We found significantly increased levels of c-MYCeGFP expression in c-KIT+ splenocytes from AML versus MPN mice (Supplementary Fig. S4A); however, only a subpopulation of c-KIT+ splenocytes from either genotype expressed c-MYCeGFP. In contrast, when AML splenocytes were grown in cytokine-rich media in vitro, the majority expressed c-MYCeGFP, but treatment with an FLT3 inhibitor (Quizartinib/AC220) downregulated c-MYCeGFP (Supplementary Fig. S4B). Flow cytometric analysis of phospho-STAT5, a downstream effector of Flt3ITD (28), was uniformly activated in c-KIT+ AML splenocytes (Supplementary Fig. S4C). These data suggested that c-MYCeGFP is a potential biomarker of Flt3ITD signaling, but that there is significant heterogeneity in AML cellularity and/or integration of Flt3ITD and loss of DNMT3A signaling. To assess this possibility, we dissected the cellular architecture of the murine AML model using single-cell RNA sequencing (scRNA-seq) coupled to complementary single-cell assays. Using the Iterative Clustering and Guide-gene Selection (ICGS) algorithm in AltAnalyze (44), unsupervised analysis of 96 individual c-KIT+ AML splenocyte libraries (generated from two independent mice) identified 7 distinct cell populations. Based on ImmGen gene expression and ontology analysis in GO-Elite (45), the cell clusters were assigned identities similar to their normal counterparts [macrophage-dendritic precursors (MDP; ref. 46), neutrophil precursors, monocyte progenitors, and four distinct HSPC-like populations; Fig. 3A; Supplementary Table S5]. We noted that more mature cell types express c-Myc with low incidence and amplitude (e.g., MDP and neutrophil precursors), whereas the more immature populations show high c-Myc incidence and amplitude (Fig. 3B). However, the bioinformatics assignment of a mature cellular identity does not necessarily obviate the leukemogenic potential of an AML subpopulation. Therefore, we performed flow cytometric sorting and CFU assays to assess the clonogenic potential of ICGS-defined murine AML subpopulations. Using discovered candidate markers from the ICGS clusters (c-Kit, Cxcr4, c-Myc, and Il18r1;Fig. 3B), 10 AML subpopulations were sorted using combinations of these markers (Fig. 3C; representative flow plots). Although c-KIT+ AML cells contained the clonogenic fraction (Fig. 3D, compare 1 vs. 3), c-KIT+c-MYCeGFP cells were significantly enriched for clonogenicity (Fig. 3D, compare 2 and 3 vs. 4). Thus, although c-KIT+ splenocytes uniformly show activated STAT5 downstream of Flt3ITD signaling, the heterogeneity in integrating that signal to provide c-MYC protein accumulation appeared to be context dependent and most probably related to the progenitor versus precursor-like state of the Flt3ITD/ITD;Dnmt3afl/flMxCre AML cell.

Figure 3.

Single-cell RNA-seq to determine the cellular structure of the Flt3ITD/ITD;Dnmt3afl/flMxCre AML model. Ninety-six scRNA-seq libraries were constructed from the c-KIT+ leukemic splenocytes of 2 independent Flt3ITD/ITD;Dnmt3afl/flMxCre AML. A, ICGS HOPACH clustering of gene expression resolves AML subpopulations (clusters: Macrophage/Dendritic cell precursor, Neutrophil (Neu) precursor, HSPC-like-1, Monocyte progenitor (Mo prog), HSPC-like-2, HSPC-like-3, HSPC-like-4), based on gene ontology enrichment analysis (GO-Elite) and literature associations. Genes listed to the right represent the ICGS-selected guide genes (the most intracorrelated gene) within each of these clusters. B, column plots demonstrate amplitude [transcripts per million (TPM)] and incidence of selected gene expression in the same ICGS-derived cell ordering as in A. C, given the enrichment for c-Myc, Il18r1, and Cxcr4 expression in the HSPC-like subpopulations in B, c-MYCeGFP fusion–protein knock-in alleles were bred into the Flt3ITD/ITD;Dnmt3afl/flMxCre AML model and antibodies to the latter proteins were used to analyze c-KIT+Flt3ITD/ITD;Dnmt3afl/flMxCre AML splenocytes. Representative flow plots illustrating the gates for sorting AML subpopulations are shown. Each sorted AML subpopulation (numbered 1–10) was examined for clonogenicity in methylcellulose colony-forming assays. D, a representative experiment is shown depicting the average ± SEM number of colonies formed for each of the gated subpopulations in C after 7 days in methylcellulose. CFU assays were repeated three times with independent c-MyceGFP;Flt3ITD/ITD;Dnmt3afl/flMxCre AML, and produced similar results. Significant differences in colony numbers were evaluated by one-way ANOVA Tukey multiple comparisons test. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Figure 3.

Single-cell RNA-seq to determine the cellular structure of the Flt3ITD/ITD;Dnmt3afl/flMxCre AML model. Ninety-six scRNA-seq libraries were constructed from the c-KIT+ leukemic splenocytes of 2 independent Flt3ITD/ITD;Dnmt3afl/flMxCre AML. A, ICGS HOPACH clustering of gene expression resolves AML subpopulations (clusters: Macrophage/Dendritic cell precursor, Neutrophil (Neu) precursor, HSPC-like-1, Monocyte progenitor (Mo prog), HSPC-like-2, HSPC-like-3, HSPC-like-4), based on gene ontology enrichment analysis (GO-Elite) and literature associations. Genes listed to the right represent the ICGS-selected guide genes (the most intracorrelated gene) within each of these clusters. B, column plots demonstrate amplitude [transcripts per million (TPM)] and incidence of selected gene expression in the same ICGS-derived cell ordering as in A. C, given the enrichment for c-Myc, Il18r1, and Cxcr4 expression in the HSPC-like subpopulations in B, c-MYCeGFP fusion–protein knock-in alleles were bred into the Flt3ITD/ITD;Dnmt3afl/flMxCre AML model and antibodies to the latter proteins were used to analyze c-KIT+Flt3ITD/ITD;Dnmt3afl/flMxCre AML splenocytes. Representative flow plots illustrating the gates for sorting AML subpopulations are shown. Each sorted AML subpopulation (numbered 1–10) was examined for clonogenicity in methylcellulose colony-forming assays. D, a representative experiment is shown depicting the average ± SEM number of colonies formed for each of the gated subpopulations in C after 7 days in methylcellulose. CFU assays were repeated three times with independent c-MyceGFP;Flt3ITD/ITD;Dnmt3afl/flMxCre AML, and produced similar results. Significant differences in colony numbers were evaluated by one-way ANOVA Tukey multiple comparisons test. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Close modal

We found, in contrast to c-MYC, CXCR4+-expressing AML cells had reduced clonogenicity in all populations as compared to populations without surface CXCR4 expression (Fig. 3D, compare 3 vs. 6, 4 vs. 9, 5 vs. 7, and 8 vs. 10). Notably, although both c-Myc and Cxcr4 were expressed at high amplitude and incidence in several ICGS clusters, only c-Myc was expressed at high incidence in the HSPC-like-1 cluster (Fig. 3B). Searching for another HSPC-like-1–enriched marker with amenable reagents for flow cytometry, we focused upon interleukin 18 receptor 1 (Il18r1, IL18rα, or CD218a; Fig. 3B). Although IL18rα was expressed at very high levels in Flt3ITD/ITD;Dnmt3afl/flMxCre AML, a role for IL18rα in normal HSPC or AML has yet to be clearly defined (47). Similar to c-KIT+c-MYCeGFP cells, c-KIT+IL18rα+ cells were enriched for clonogenicity (Fig. 3D, population 3 vs. 5); however, AML cells that were triple-positive (c-KIT+c-MYCGFPIL18rα+CXCR4) were the most significantly enriched for leukemia clonogenicity (Fig. 3D, population 8). In summary, the Flt3ITD/ITD;Dnmt3afl/flMxCre AML model consisted of functionally heterogeneous c-KIT+ stem/progenitor/precursor populations with enriched clonogenicity corresponding to HSPC-like cluster gene expression.

Given the association between DMR and gene expression (Fig. 2), and the considerable cellular heterogeneity in gene expression (Fig. 3), we next determined whether manipulating DNMT3A levels would reverse methylation, gene expression, and/or clonogenicity. To this end, we rescued Dnmt3a haploinsufficiency in murine AML by forced expression of DNMT3A (Fig. 4A). Compared with empty vector (EV) control–transduced AML, DNMT3A rescue did not alter differentiation or colony numbers in the initial plating (Fig. 4B, left bar graphs). Importantly, DNMT3A rescue ablated clonogenicity upon replating (Fig. 4B, right line graphs), and was associated with DNA hypermethylation (Fig. 4C; Supplementary Table S6). In particular, 74 of 260 of the DMRs in AML were reverted to methylation levels similar to MPN, and 40 new hypermethylated regions were identified (Fig. 4D). As in the de novo murine AML cells, there was a significant enrichment of DMRs overlapping with both intra- and intergenic enhancers (12% of DMRs at intragenic enhancers vs. 3% background, P < 2.2 × 10−16 and 8% of DMRs at intergenic enhancers versus 2% background, P < 2.2 × 10−16; Fig. 4E). In agreement with the association between HSPC-like gene expression and clonogenicity, the HSPC-like AML populations, prior to rescue, expressed genes proximal to remethylated regions at a significantly higher level compared to more mature precursors (Fig. 4F; Supplementary Fig. S4D, compare clusters c3-c7 versus c1 and c2). For example, we consistently observed DNA hypomethylation of the GATA3, CXXC5, and HDAC7 gene loci in both human and murine DNMT3A-mutant AML (Supplementary Table S3). In addition, GATA3 and CXXC5 were overexpressed in human FLT3ITD/DNMT3A-mutant AML (Supplementary Table S4 GSEA leading edge from Fig. 2H), whereas Gata3, Cxxc5, and Hdac7 were all significantly overexpressed in murine AML (Fig. 2I; Supplementary Table S4). Moreover, Gata3, Cxxc5, and Hdac7 were expressed almost exclusively in the HSPC-like murine AML subpopulations (Fig. 3B; Supplementary Fig. S4E) and were hypermethylated upon DNMT3A rescue (Fig. 4D). These data suggest that genes that respond in this manner may have central roles in CN-AML downstream of DNMT3A and FLT3. We conclude that DNA hypomethylation downstream of DNMT3A loss of function is reversible.

Figure 4.

Rescuing DNMT3A expression is associated with reversion of DNA hypomethylation and loss of clonogenicity. c-KIT+ AML splenocytes from two different donor Flt3ITD/ITD;Dnmt3afl/flMxCre AML mice were transduced with wild-type DNMT3A (DNMT3A) or EV control to rescue DNMT3A expression. A, average ± SEM of endogenous Dnmt3a and rescue DNMT3A expression in WT DNMT3A-transduced Flt3ITD/ITD;Dnmt3afl/flMxCre AML CFU relative to endogenous Dnmt3a in matched EV control (Sdha served as a control). Significance was evaluated by an unpaired parametric t test. B, average ± SEM relative proportions of granulocyte (CFU-G), monocyte (CFU-M), and mixed (CFU-GM) colonies formed after 7 days in methylcellulose (left stacked bar graphs) and average ± SEM number of colonies in two serial platings (right line graphs) with DNMT3A rescue or EV control Flt3ITD/ITD;Dnmt3afl/flMxCre AML cells. Significant reduction in CFU formed by DNMT3A rescued cells compared to EV control was determined by an unpaired parametric t test. Two independent mice are shown; averages are technical replicates per mouse. The experiment was repeated twice with similar results. C, volcano plot representation of mean methylation difference (x-axis) versus statistical significance (y-axis). Red dots designate significantly DMRs in DNMT3A-rescued Flt3ITD/ITD;Dnmt3afl/flMxCre AML versus EV control from B. D, heatmap representation of DMRs in DNMT3A-rescued versus EV control Flt3ITD/ITD;Dnmt3afl/flMxCre AML from B. Corresponding levels of methylation for those regions are also shown in the primary Flt3ITD/ITD;Dnmt3afl/fl MPN (1° MPN) and Flt3ITD/ITD;Dnmt3afl/fl AML (1° AML). E, proportion of differentially methylated genomic, CpGi, and enhancer regions in all measured CpGs (all tiles) and DMRs (hypermethylated and hypomethylated) in DNMT3A-rescued compared to EV control-transduced Flt3ITD/ITD;Dnmt3afl/flMxCre AML from B. F, violin plots represent the expression of genes associated with remethylated DMRs in D, within the ICGS clusters from Fig. 3A. G, average ± SEM number of colonies from c-KIT+ AML splenocytes from two different donor Flt3ITD/ITD;Dnmt3afl/flMxCre AML mice (two independent mice denoted by white or gray bars) transduced with Cxxc5, Emilin2, Il18r1 shRNA or nontargeting (NT) control. Significant reduction in CFU formed by Cxxc5 and Emilin2 knockdown AML cells compared to NT control was determined by an unpaired parametric t test (#, P = 0.07; *, P < 0.05). Two independent mice are shown; averages are technical replicates per mouse.

Figure 4.

Rescuing DNMT3A expression is associated with reversion of DNA hypomethylation and loss of clonogenicity. c-KIT+ AML splenocytes from two different donor Flt3ITD/ITD;Dnmt3afl/flMxCre AML mice were transduced with wild-type DNMT3A (DNMT3A) or EV control to rescue DNMT3A expression. A, average ± SEM of endogenous Dnmt3a and rescue DNMT3A expression in WT DNMT3A-transduced Flt3ITD/ITD;Dnmt3afl/flMxCre AML CFU relative to endogenous Dnmt3a in matched EV control (Sdha served as a control). Significance was evaluated by an unpaired parametric t test. B, average ± SEM relative proportions of granulocyte (CFU-G), monocyte (CFU-M), and mixed (CFU-GM) colonies formed after 7 days in methylcellulose (left stacked bar graphs) and average ± SEM number of colonies in two serial platings (right line graphs) with DNMT3A rescue or EV control Flt3ITD/ITD;Dnmt3afl/flMxCre AML cells. Significant reduction in CFU formed by DNMT3A rescued cells compared to EV control was determined by an unpaired parametric t test. Two independent mice are shown; averages are technical replicates per mouse. The experiment was repeated twice with similar results. C, volcano plot representation of mean methylation difference (x-axis) versus statistical significance (y-axis). Red dots designate significantly DMRs in DNMT3A-rescued Flt3ITD/ITD;Dnmt3afl/flMxCre AML versus EV control from B. D, heatmap representation of DMRs in DNMT3A-rescued versus EV control Flt3ITD/ITD;Dnmt3afl/flMxCre AML from B. Corresponding levels of methylation for those regions are also shown in the primary Flt3ITD/ITD;Dnmt3afl/fl MPN (1° MPN) and Flt3ITD/ITD;Dnmt3afl/fl AML (1° AML). E, proportion of differentially methylated genomic, CpGi, and enhancer regions in all measured CpGs (all tiles) and DMRs (hypermethylated and hypomethylated) in DNMT3A-rescued compared to EV control-transduced Flt3ITD/ITD;Dnmt3afl/flMxCre AML from B. F, violin plots represent the expression of genes associated with remethylated DMRs in D, within the ICGS clusters from Fig. 3A. G, average ± SEM number of colonies from c-KIT+ AML splenocytes from two different donor Flt3ITD/ITD;Dnmt3afl/flMxCre AML mice (two independent mice denoted by white or gray bars) transduced with Cxxc5, Emilin2, Il18r1 shRNA or nontargeting (NT) control. Significant reduction in CFU formed by Cxxc5 and Emilin2 knockdown AML cells compared to NT control was determined by an unpaired parametric t test (#, P = 0.07; *, P < 0.05). Two independent mice are shown; averages are technical replicates per mouse.

Close modal

Because DNMT3A rescue diminished the colony-forming ability of Flt3ITD/ITD;Dnmt3afl/flMxCre AML, we next sought to delineate genes downstream of DNMT3A loss of function that might be important for leukemic potential. To this end, we examined the clonogenicity of murine c-KIT+ AML cells upon shRNA-mediated knockdown of genes that are overexpressed and hypomethylated in murine AML, and/or DMR reverted upon DNMT3A rescue (e.g., Gata3, Emilin2, Hdac7, Cxxc5, Alpk3, and Ttyh2). We also examined knockdown of Il18r1 (Il18rα) on colony formation, because we found that this cell-surface receptor is expressed on HSPC-like populations with increased clonogenicity (Fig. 3D). Repression of Emilin2 and Cxxc5 by multiple independent hairpins caused significant reduction in colony formation, whereas Il18r1 knockdown had no effect (Fig. 4G; Supplementary Fig. S4F). Even though we did observe some reduction in colony formation with Gata3 and Alpk3 shRNA, these effects were less consistent across multiple hairpins in multiple mouse AMLs (data not shown); however, we cannot discount the possibility that these genes may also play a role in Flt3ITD/ITD;Dnmt3afl/flMxCre AML. In sum, these data suggest that IL18rα simply serves as a surface marker on AML cells with increased clonogenic capacity, whereas Cxxc5 and Emilin2, which are specifically expressed in HSPC-like AML populations, have an important functional role in self-renewal and proliferation in vitro. In conclusion, DNMT3A haploinsufficiency collaborates with Flt3ITD to form a rapidly lethal, fully penetrant AML. DNMT3A loss is associated with reversible changes in DNA methylation, leading to signaling programs distinct from Flt3ITD alone that promote leukemogenesis.

Flt3ITD mutation combined with other oncogenic lesions such as Tet2−/− (16), MllPTD (35), and Npmc+ (48) progresses to AML or mixed-lineage leukemias in mice. The uniquely rapid lethal nature of the Flt3ITD/ITD;Dnmt3afl/flMxCre AML model may be due to DNMT3A control of HSC self-renewal. Deletion of Dnmt3a (Dnmt3a−/−) specifically expands HSC and myeloproliferation in vivo, but does not produce acute leukemia (21, 49). The addition of KrasG12D mutation with Dnmt3a−/− in mice significantly accelerates MPN development compared to KrasG12D alone (50). Yet, only ∼30% of Dnmt3a−/−;KrasG12DMxCre mice progress to AML and require pIpC to induce complete Dnmt3a deletion. The DNMT3AR882H mutation found in human AML acts as a dominant negative that reduces, but does not completely block, the activity of DNMT3A (18). Moreover, the prognosis for human FLT3ITD AML is worsened with the addition of a DNMT3A mutation (3, 12, 13). In parallel to human FLT3ITD/DNMT3A-mutant CN-AML, Flt3ITD/ITD;Dnmt3afl/flMxCre mice spontaneously select for Dnmt3a haploinsufficiency—rather than a complete loss—that transforms Flt3ITD MPN into a rapidly lethal AML with 100% penetrance. Moreover, a high FLT3ITD allelic ratio predicts poor survival in human AML (51). Thus, the requirement for two Flt3ITD alleles to form AML (in cooperation with Dnmt3A loss) in this mouse model is consistent with human disease. The pattern of methylation was similar to human AML with FLT3ITD and mutations in DNMT3A. In addition, there was significant correlation between the expression of some genes with nearby changes in DNA methylation in Flt3ITD/ITD;Dnmt3afl/flMxCre AML compared to Flt3ITD/ITD;Dnmt3afl/fl MPN, indicating that these changes in DNA methylation do not reflect nonspecific methylation losses but rather play a functional role in the development of leukemia.

We used single-cell RNA-seq to dissect the cellular architecture of the leukemia.

Although cancer stem cell markers are usually not conserved across different cancers (46), c-KIT expression is a known marker of leukemia stem cells (LSC) in other AML models. In agreement with this, our functional assays revealed tremendous enrichment for clonogenicity in the c-KIT+ population. However, the scRNA-seq analysis revealed significant heterogeneity within the c-KIT+ leukemic splenocytes, with many subpopulations expressing transcriptomes similar to myeloid precursors and possessing little clonogenicity. In contrast, expression of IL18rα (CD218a) significantly enriched clonogenic potential in our murine AML model. Manipulation of Il18r1 (IL18rα) by shRNA in Flt3ITD/ITD;Dnmt3afl/flMxCre AML cells had no significant effect on colony-forming ability. Therefore, IL18rα is a candidate surface marker of a rare LSC population, but is likely not driving the functional clonogenic output. In addition to IL18rα and c-KIT, the most highly clonogenic subpopulation also accumulated c-MYC protein and expressed significantly higher levels of HSPC-like genes, whose associated DMRs were reverted upon DNMT3A rescue. Therefore, this rare putative LSC population contained active oncogenic signaling downstream of mutant DNMT3A and FLT3ITD, demonstrating the power of single-cell analyses to resolve both biologic and molecular complexity.

The correlation between reverted DMR and significantly diminished clonogenicity in our murine AML suggests that the methylome changes downstream of mutant DNMT3A in human AML are probably disease-driving and reversible. Cxxc5 is an exemplary candidate because it conforms to many criteria of a functionally important putative DNMT3A target gene in AML; hypomethylated and overexpressed (HSPC-like) in murine and human AML, DMR reverted upon DNMT3A rescue, and knockdown of Cxxc5 (similar to DNMT3A rescue) significantly diminished AML colony-forming ability. The CXXC-type zinc-finger protein CXXC5 is a pleiotropic factor that can interact with Dishevelled to inhibit WNT/β-catenin signaling (52, 53), binds to CpG-rich DNA sequences to regulate gene expression (54), and is a biomarker and potential therapeutic target in leukemia (55–57). The extracellular matrix protein EMILIN2, also described as an inhibitor of WNT signaling (58), is hypomethylated and overexpressed in AML and is important for AML clonogenicity. Because both CXXC5 and EMILIN2 play a role in WNT signaling and c-MYC is a target of the WNT pathway, we monitored the level of c-MYCGFP and did not find that shRNA knockdown of either Cxxc5 or Emilin2 disrupted c-MYC accumulation (data not shown). Although this does not preclude a possible role for either CXXC5 or EMILIN2 in WNT signaling, more work needs to be done to fully understand the roles and therapeutic potential of CXXC5 and EMILIN2 in FLT3ITDDNMT3A-mutant AML.

Our data suggest that the functional implications of some cancer-specific epigenetic alternations (e.g., DNMT3A mutations) may extend beyond that of persisting clones from the premalignant state (59, 60) to a continuing role in AML disease pathogenesis and maintenance. We note recent progress in small-molecule inhibitors therapeutically targeting proteins that create and/or maintain epigenetic modifications in cancer. Given the acutely rapid and spontaneous development of AML in the Flt3ITD/ITD;Dnmt3afl/flMxCre mice, and the plasticity of DNA methylation within the resulting AML, we conclude that this new murine model is suitable for testing future therapeutic methods of manipulating DNMT3A, which may provide new opportunities for clinical intervention in CN-AML.

Animals

Mice were bred and housed by the Cincinnati Children's Hospital Medical Center (CCHMC) Veterinary Services, and mouse manipulations were performed according to the protocol reviewed and approved by the Children's Hospital Research Foundation Institutional Animal Care and Use Committee (Protocol Number IACUC 2013-0090). c-MyceGFP (43), Flt3ITD (27), Dnmt3aflox (61), and MxCre (62) mice were maintained on the C57BL/6-Ly5.2 (CD45.2) background. CD45.1+ C57BL/6-Ly5.1 (JAX and CCHMC Comprehensive Mouse Core) transplant recipient mice were between 1 and 3 months of age and given 700 Rads 5 hours prior to tail-vein injection of single-cell suspensions.

Bone Marrow and Leukemic Splenocyte Isolation

Moribund mice were euthanized with carbon dioxide and cervical dislocation. Cytospin buffer containing 10% FBS (HyClone), 2% BSA Fraction V (ThermoFisher Scientific), and Cell Dissociation Buffer (Gibco) in PBS was used for all cell isolations, washes, and cytospins under sterile conditions. Bone marrow cells were isolated from freshly sacrificed mice by crushing bones using a mortar and pestle. Splenocytes were isolated from freshly sacrificed mice and crushed between two glass slides. Red blood cells were lysed using Ammonium–Chloride–Potassium (ACK) buffer (Gibco), washed, and filtered to obtain single-cell suspensions for downstream applications.

Histology and Cytospins

Bone marrow, spleen, liver, and sternum were harvested from moribund Flt3ITD/ITD;Dnmt3afl/flMxCre and age-matched WT or Flt3ITD/ITD;Dnmt3afl/fl controls. Sternum was decalcified, then spleen, liver, and sternum were paraffin embedded, and sections were stained with hematoxylin and eosin (H&E; CCHMC pathology core) and cytospins were stained with Wright Giemsa using the Camco Stain Pak (Cambridge Diagnostic Products). Cytospins of bone marrow and spleen were performed both before and after red blood cell (RBC) lysis. Photographs were taken on a Nikon Eclipse 80i microscope with a Nikon Digital Sight camera using NIS-Elements F4.30 software at a resolution of 2,560 × 1,920. Using Adobe Photoshop CS2, auto contrast was applied to H&E images, unsharp mask was used to improve image clarity, images were resized and set at a resolution of 300 pixels/inch, and variations was used to adjust color balance and brightness. Scale bar represents 30 μm in each image. At least 3 to 5 mice per group were examined for histologic, morphologic, peripheral blood, and flow cytometric findings for acute myelomonocytic leukemia diagnosis.

Spectral Karyotyping Analyses

To characterize the cytogenetic pattern of AMLs from Flt3ITD/ITD;Dnmt3afl/flMxCre mice, G-banding and SKY analysis was performed using the ASI SkyPaint assay for mouse chromosomes as described previously (63) on fresh or viably frozen bone marrow from moribund mice with leukemia. Five independently derived AMLs (20 metaphases per case) from the Flt3ITD/ITD;Dnmt3afl/flMxCre model were analyzed. The analysis of 20 cells excludes an abnormal clone of ≥14% with 95% confidence.

Flow Cytometry and Cell Sorting

All flow cytometric staining, sorting, and analyses were performed by resuspending cells in 1× FACS buffer [1% FBS, 0.01% NaN3 in Dulbecco's Phosphate-Buffered Saline (DPBS)]. To isolate c-KIT+ leukemic stem/progenitor populations, freshly isolated single cell suspensions were incubated with CD117 MicroBeads and separated on an AutoMACS Pro separator (Miltenyi) according to the manufacturer's specifications.

To prepare for flow cytometric analysis of stem/progenitor populations, single-cell suspensions were first stained with a cocktail of biotin-conjugated antibodies against markers of lineage-positive cell types: anti-CD3 (clone 145-2C11, BioLegend), anti-CD4 (clone RM4-5, eBioscience), anti-CD8 (clone 53-6.7, Becton, Dickinson and Company), anti-CD11b (clone M1/70, Becton, Dickinson and Company), anti-CD19 (clone 6D5, BioLegend), anti-Gr1 (clone RB6-8C5, BioLegend), anti-Ter119 (clone Ter-119, BioLegend), and anti-CD45R (clone RA3-6B2, BioLegend). Following two washes in 1× FACS buffer, lineage-stained cells were stained with: streptavidin-conjugated APC–Cy7 (Becton, Dickinson and Company), PerCP–eFluor710-conjugated anti-CD16/32 (clone 93, eBioscience), APC-conjugated anti-CD117 (clone 2B8, Becton, Dickinson and Company), PE–Cy7-conjugated anti-Sca1 (clone D7, Becton, Dickinson and Company), and BV421-conjugated anti-CD34 (clone RAM34, Becton, Dickinson and Company). Alternatively, for the analysis of l-GMP populations, a modified lineage cocktail was used, which excluded anti-CD11b and anti-Gr1 biotin–conjugated antibodies. Furthermore, lineage-stained cells were incubated with the same antibodies as described above but with BV605-conjugated anti-CD11b (clone M1/70, Becton, Dickinson and Company) and PE–Cy5-conjugated anti-Gr1 (clone RB6-8C5, eBioscience) added and BV650-conjugated anti-CD117 (clone RAM34, Becton, Dickinson and Company) substituted for the APC conjugate. Antibodies were used at 1:100 dilutions, except for anti-CD34, which was used at 1:50 in a staining volume of 50 μL/5 × 106 cells. All results reported for LK, GMP, and LSK populations exclude CD11b and Gr1 along with other linage markers indicated above either as biotin or fluorochrome conjugates.

To prepare for flow cytometric analysis of mature populations, single-cell suspensions were incubated with FcγRII/III Block (clone 2.4G2, Becton, Dickinson and Company) and a combination of: Pacific Blue-conjugated anti-CD11b (cloneM1/70, BioLegend), PE–Cy7-conjugated anti-Gr1 (clone RB6-8C5, BioLegend), APC-conjugated anti-CD3 (clone 145-2C11, Becton, Dickinson and Company), or FITC-conjugated anti-CD3 (clone 145-2C11, Becton, Dickinson and Company), PE-conjugated anti-CD45R (clone RA3-6B2), or AlexaFluor 700–conjugated anti-CD45R (clone RA3-6B2, BioLegend), PerCP–Cy5.5-conjugated anti-CXCR4 (clone L276F12, BioLegend), PE-conjugated anti-IL18R (clone P3TUNYA, eBioscience), and APC-conjugated anti-CD117 (clone RAM34, Becton, Dickinson and Company), or BV650-conjugated anti-CD117 (clone RAM34, Becton, Dickinson and Company). Antibodies were used at 1:100 dilutions in a staining volume of 50 μL for up to 5 × 106 cells.

Cell sorting was performed on a MoFloXDP (Beckman Coulter) or BD FACSAria II with a 70-μm or 100-μm nozzle. Sorted populations of interest were collected in DPBS buffer containing 50% FBS (Hyclone). Flow cytometric analyses were performed on a FACS LSRII or LSR Fortessa (Becton, Dickinson and Company). Data were analyzed with FlowJo Software (TreeStar).

In preparation for flow cytometric analysis of phosphorylated (p) STAT5 quantitation, 1 million freshly isolated c-KIT+ splenocytes were added to each well of a 96-well round-bottom plate and centrifuged at 300 ×g at 4°C for 5 minutes. Cells were next fixed in 100 μL of 4% methanol-free formaldehyde (Pierce) for 20 minutes at room temperature and then centrifuged as before. Cells were washed in 200 μL 1× FACS buffer. Next, cells were permeabilized for 10 minutes at 4°C in 200 μL of 90% methanol (Fisher) and then washed twice. Cells were stained with PE-conjugated anti-pSTAT5Y694 antibody (clone 47/Stat5-pY694, Becton, Dickinson and Company) at a 1:100 dilution in 100 μL of 1× FACS buffer for 1 hour on ice. Cells were washed twice prior to flow analysis.

AC220 Treatment

Prior to in vitro treatment with AC220 (Ambit Biosciences), cryopreserved splenocytes were thawed and cultured overnight in StemSpan (Stem Cell Tech) supplemented with recombinant mouse IL3 (10 ng/mL), IL6 (20 ng/mL), and Stem Cell Factor (SCF; 25 ng/mL; Miltenyi). Cells were then replated at 1 million cells per mL with either 0.1% DMSO or 1 mmol/L AC220 and incubated for 5 hours at 37°C. Cells were then washed once in 1× FACS and stained with APC-conjugated anti-CD117 (clone RAM34, Becton, Dickinson and Company) at a 1:100 dilution in 100 μL of 1× FACS buffer for 1 hour on ice. Cells were washed twice prior to flow analysis.

RNA Isolation, cDNA Synthesis, and TaqMan Assay

RNA was extracted using TRIzol or TRIzol LS (Ambion) as previously described (64). Total RNA was converted into cDNA using the High Capacity cDNA Archive Kit (Life Technologies) for standard gene expression analyses. All gene expressions were quantified using TaqMan assays (Life Technologies) with the 2× Universal TaqMan Master Mix (Life Technologies) according to the manufacturer's instructions on the StepOne Plus Instrument (Applied Biosystems, Inc.) using equal amounts of starting cDNA. Standard inventoried gene expression TaqMan assays recommended by the manufacturer (Life Technologies) for best coverage were purchased when available; for human gene expression assays: DNMT3A (Hs01027166_m1); for mouse gene expression assays: Dnmt3a (Mm00432881_m1), Emilin2 (Mm00467425_m1), Pim1 (Mm00435712_m1), Gata3 (Mm00484683_m1), Cxxc5 (Mm00505000_m1), and Il18r1 (Mm00515178_m1). Gene expression data were analyzed using the comparative Ct method (2−ΔΔCt). Human SDHA (Hs00417200_m1) and mouse Sdha (Mm01352366_m1) served as loading controls.

Bulk RNA-Sequencing Analyses

c-KIT+ splenocytes were harvested from three independent Flt3ITD/ITD;Dnmt3afl/flMxCre or age-matched Flt3ITD/ITD;Dnmt3afl/fl mice. RBCs were lysed with ACK, and 5 × 106 leukemia cells were lysed with TRIzol (Invitrogen). RNA quality and quantity were measured by Agilant Bioanalyzer at CCHMC Microarray Core. Then, samples with an RNA Integrity Number of >8.0 were processed by poly-A selection, stranded True-seq protocol, and RNA library sequencing using paired-end 75-bp reads at a depth of 40 million paired-reads per sample (Illumina HiSeq 2500) by the CCHMC Sequencing Core. Sorted LSK, CMP, and GMP bulk RNA-sequenced samples, obtained using the same library preparation and sequencing protocol, were obtained from the gene expression omnibus dataset GSE70235. The fastq files were aligned to mouse genome Mm10 using BowTie2 and TopHat 2 (65, 66). Gene-level Reads Per Kilobase of transcript per Million mapped reads (RPKM) were obtained using AltAnalyze 2.0.9 (44) to identify differentially expressed genes between Flt3ITD/ITD;Dnmt3afl/flMxCre AML and Flt3ITD/ITD;Dnmt3afl/fl MPN using default parameters, including a moderated t test with Benjamini–Hochberg multiple testing correction, raw P value cutoff of 0.05, RPKM > 1 and a differential expression cutoff of 2-fold. WT LSK samples were also displayed as normal baseline comparators. Hierarchical Ordered Partitioning And Collapsing Hybrid (HOPACH) clustering in AltAnalyze was used to identify clusters of gene expression in AML, MPN, and WT LSK.

Single-Cell RNA-Sequencing Analyses

Single CD117+ (c-KIT+) cells were prepared using the C1 Single-Cell Auto Prep System (Fluidigm), according to the manufacturer's instructions. In short, sorted cells were counted and resuspended at a concentration of 333,333 cells per 1 mL DPBS, and then loaded onto a primed C1 Single-Cell Auto Prep Integrated Fluidic Chip for mRNA-seq (5–10 μm). After the fluidic step, cell separation was visually scored to identify wells containing single live cells. In two independent rounds, 52 and 56 cells were captured. Cells were lysed on the chip, and reverse transcription was performed using the Clontech SMARTer Kit with the mRNA Seq: RT + Amp (1771×) according to the manual. After the reverse transcriptase step, cDNAs were transferred to a 96-well plate and diluted with 6 μL of C1 DNA Dilution Reagent. cDNAs were quantified using the Quant-iT PicoGreen dsDNA Assay Kit (Life Technologies) and Agilent High Sensitivity DNA Kit (Agilent Technologies). Libraries for the first 48 visually validated cells from each of the two independent captures were prepared using the Nextera XT DNA Library Preparation Kit (Illumina Inc.). In each single-cell library preparation, a total of 125 pg cDNA was tagmented at 55°C for 20 minutes. Next, libraries were pooled and purified on the AMPure bead-based magnetic separation kit prior to a final quality control check using both the Qubit dsDNA HS Assay Kit (Life Technologies) and Agilent High Sensitivity DNA Kit. The majority of cDNA fragments were confirmed to be between 200 and 602 base pairs (bp), qualifying the libraries for sequencing. All 96 single-cell RNA-seq libraries were sequenced on one HiSeq 2500 gel. Gene expression was quantified as transcripts per million (TPM) using RNA-Seq by Expectation-Maximization (RSEM) alignment and quantification (67). The predominant gene expression signatures were obtained from the ICGS module of AltAnalyze 2.0.9 (http://altanalyze.org/ICGS.html). For ICGS analysis, genes with a minimum TPM > 1 and fold >4 in at least 3 cells were retained for analysis, with a minimum Pearson correlation cutoff of 0.4 and conservative exclusion of cell-cycle effects selected as the parameters. Associated cell population predictions were obtained the GO-Elite (45) module in AltAnalyze and verified based on prior literature. All associated expression plots display gene expression values as the median subtracted or unsubtracted log2 TPM values. The genes associated with DNMT3A rescued AML reverted DMRs are displayed in Violin plots comparing the expression of these genes in the original single-cell RNA-seq AML dataset across all clusters using the vioplot library in R.

Human AML Specimens

Genomic DNA was isolated from bone marrow mononuclear cells from patients with AML either seen at Erasmus Medical Center, Rotterdam, or enrolled in the E1900 clinical trial from the Eastern Cooperative Oncology Group. Informed consent was obtained prior to sample banking. Institutional Review Board (IRB) approval for this study was obtained at the University of Michigan and Weill Cornell Medical Center. This study was performed in accordance with the Declaration of Helsinki. All specimens analyzed age-matched consisted of >90% blasts. A total of 32 patients were included (Supplementary Table S7). Sixteen patients were FLT3ITD/DNMT3AWT and 13 patients were FLT3ITD/DNMT3A-mutant. All patients were wild-type for TET2, IDH1, and IDH2. Patients with FLT3-TKD were also excluded. Bone marrow from normal donors was obtained after IRB approval from the Johns Hopkins Sidney Kimmel Cancer Center Specimen Accessioning Core. Mononuclear cells were isolated by density centrifugation using Ficoll–Paque Plus (GE Healthcare) and then labeled for CD34-positive selection using the CD34 MicroBead Kit (Miltenyi Biotec) and frozen viable in FBS + 10% DMSO. Upon thawing, CD34+ cells were labeled using the Aldefluor Kit (Stem Cell Technologies), CD34-BV421 or CD34-PE-Cy7, CD45-PE, CD38-APC (BD Biosciences), and 7-Aminoactinomycin D (7-AAD; Life Technologies). HSCs were identified as ALDHhiCD34+CD38CD45+ and sorted on a Beckman Coulter MoFlo in the Cell Sorting Core Facility at the Johns Hopkins Bloomberg School of Public Health. Gene expression data for these patients with AML were previously published (ref. 68; GEO accession number: GSE6891).

DNA Isolation and DNMT3A Genotyping

DNA was isolated from c-KIT+ splenocytes (ERRBS) or single-cell suspensions from individual colonies grown for 7 days in methylcellulose (genotyping) using the Qiagen Puregene Kit according to the manufacturer's instructions. For small cell numbers, glycogen was added to help precipitate DNA. ERRBS data for WT murine LSK cells were previously published (ref. 69; GEO accession number: GSE57114). For genotyping, the DNMT3Afl/fl-F 5′-TGT GGC ATC TCA GGG TGA T-3′, DNMT3Afl/fl-R 5′-CTC AGG CCC TCT AGG CAA G-3′, and DNMT3AKO-R 5′-GAA GCA GCA GTC CCA GGT AG-3′ were used.

DNA Methylation by ERRBS

ERRBS libraries were generated from 25 ng of genomic DNA as previously described (37) for the human specimens, and sequenced on a HiSeq-2000 sequencer. Murine libraries were prepared in the same way but using the indexed adaptors and sequenced with two libraries per lane on a HiSeq-2500 sequencer (70). Reads were aligned against a bisulfite-converted version of the hg19 or mm9 builds of the genome using Bowtie followed by methylation percentage calling using Bismark (71). Conversion rates were >99.8 for all samples. Only genomic regions with coverage between 10× and 500× were included for downstream data analysis.

DNA Methylation Data Analysis

ERRBS data analysis was performed using R statistical software version 3.0.3 and the MethylSig 0.1.3 package (72) for differential methylation calling using the beta binomial method. The MethylKit 0.9.2.5 package (73) was used for downstream gene-, CpG-, and enhancer region annotation analysis and chromosome ideogram representations. DMRs were identified by first summarizing the methylation status of genomic regions into 25-bp tiles and then identifying regions with absolute methylation difference ≥25% and false discovery rate (FDR) <10%. DMRs were annotated to GENCODE genes using a custom script with the following criteria: (i) DMRs overlapping with a gene were annotated to that gene, (ii) intergenic DMRs were annotated to all neighboring genes within a 50-kb window, and (iii) if no gene was detected within a 50-kb window, then the DMR was annotated to the nearest transcription start site (TSS; ref. 74). Statistical significance for comparisons between all measured CpGs (all tiles) and DMRs at genomic, CpG, and enhancer regions was calculated using upper and lower tail binomial tests, and percentage of total DMR and P values are reported.

Virus Production and Transduction

For virus production, Lenti-X 293T cells (Clonetech) were seeded at 4 × 106 cells/10-cm dish in standard growth medium DMEM (ThermoFisher Scientific) with 10% heat-inactivated FBS (Atlanta Biologics), penicillin–streptomycin (ThermoFisher Scientific), and l-Glutamine (ThermoFisher Scientific). The following day, cells were replenished with fresh Opti-MEM GlutaMAX media (ThermoFisher Scientific) supplemented with 10% heat-inactivated FBS (Atlanta Biologics) at least 2 hours prior to transfection. MSCV-DNMT3A-GFP and EV MSCV-GFP retroviral plasmids were a kind gift from Dr. Grant Challen (Washington University, St. Louis, MO). For retroviruses, 10 μg of DNMT3A and retroviral vectors were cotransfected with 10 μg retroviral packaging and envelope plasmid pCL-Eco (75) into Lenti-X 293T cells using Transit-LT1 (Mirus) according to the manufacturer's instructions. Virus-containing supernatant was collected 48 and 72 hours after transfection, and either used immediately or frozen at −80°C in aliquots. For lentiviruses, gene-targeting shRNA or nontargeting control lentiviral vectors were cotransfected with lentiviral packaging plasmid Δ8.9 (76) and envelope plasmid VsVg (77) using TransIT-LT1 (Mirus Bio) into Lenti-X 293T cells (Clonetech). Virus-containing supernatant was collected 48 and 72 hours after transfection, and either used immediately or frozen at −80°C in aliquots. Two independent shRNAs against Cxxc5 (TRCN0000143369, TRCN0000219528), Emilin2 (TRCN0000257587, TRCN0000246910), and Il18r1 (TRCN0000068103, TRCN0000068105) were purchased from the CCHMC Lenti shRNA viral Core or Sigma.

Transduction of c-KIT+ leukemic splenocytes with retroviruses was performed using the spinfection method. Tissue culture plates were precoated with 5 μg/cm2 recombinant Retronectin (Takara) overnight at 4°C, then blocked with 2% BSA (ThermoFisher Scientific) in PBS (ThermoFisher Scientific) for 30 minutes at room temperature. Unconcentrated fresh or freeze/thawed retroviral supernatant was spun onto the retronectin at 1,000 g for 2 hours at 4°C. Supernatant was removed, and a second spin was performed with additional retrovirus supernatant. Viral supernatant was removed, and 1 × 106 c-KIT+ AML cells in 0.5 mL standard growth medium were spun onto the retrovirus/retronectin-coated wells at 1,000 g for 10 minutes at room temperature. Plates were then incubated overnight at 37°C/5% CO2. The spinfection protocol was repeated again the next day. Forty-eight hours after transduction, cells were sorted for GFP+ expression, and 2,000 cells were plated in methylcellulose M3434 (supplemented) in triplicate. After 7 days, colonies were enumerated, single-cell suspensions were made, and 2,000 cells were serially replated in triplicate. Extra cells not replated were mixed with cell lysis solution (Qiagen; for DNA isolation) or lysed in TRIzol (for RNA isolation).

For lentiviral transduction, c-KIT+ leukemic splenocytes freshly harvested from moribund AML mice were cultured overnight in standard growth medium RPMI-1640 (Thermo Fisher Scientific) containing 20% FBS (HyClone) and supplemented with mouse recombinant SCF (60 ng/mL), IL6 (20 ng/mL), and IL3 (20 ng/mL; Miltenyi Biotech). c-KIT+ leukemia cells (1 × 106) in 400 μL standard growth medium containing FBS were mixed with 100 μL lentiviral supernatant. Cytokines (60 ng/mL SCF, 20 ng/mL IL6 and IL3) and polybrene (10 μg/mL, Sigma) were added to the final volume. Cells were incubated with virus for 6 hours at 37°C with 5% CO2, and then virus supernatant was replaced with standard growth medium containing FBS and cytokines. Cells were incubated for 48 hours, then for an additional 48 hours with 1.5 μg/mL puromycin in fresh growth medium. Puromycin-selected cells (4,000) were plated in triplicate in methylcellulose M3434 (Stem Cell Technologies) for colony-forming assay. Colonies were enumerated after 7 days. Extra cells (not plated for colony assays) were lysed in TRIzol (for RNA isolation).

GSEA Analyses

Log2 RNA-seq expression derived from c-KIT+ leukemic splenocytes from three independent Flt3ITD/ITD;Dnmt3afl/flMxCre AML mice and three Flt3ITD/ITD;Dnmt3afl/fl MPN mice was imported into GSEA v.2.2.0 (41, 42). A fold-change ranked gene list was created comparing the AML with MPN using the Diff_of_Classes ranking metric, and genes were sorted by descending order real mode. GSEA analyses were run using 1,000 gene_set permutations of gene sets in the website gene matrix database with the weighted enrichment statistic. Results are reported for the most highly positively correlated and negatively correlated hematopoietic-related gene sets with leading-edge analyses.

Genes upregulated in Flt3ITD/ITD;Dnmt3afl/flMxCre AML versus murine MPN were tested for enrichment in human FLT3ITD/DNMT3A-mutant AML in GSEA as described above using a preranked fold-change gene expression list in descending order comparing FLT3ITD/DNMT3A-mutant AML versus FLT3ITD/DNMT3AWT AML. Genes upregulated in Flt3ITD/ITD;Dnmt3afl/flMxCre AML, both legacy and unique groups, were used as the testing gene set. A gene set of nearest gene neighbors to hypomethylated regions in Flt3ITD/ITD;Dnmt3afl/flMxCre AML versus MPN was tested for enrichment against a fold-change ranked gene expression list from RNA-seq in Flt3ITD/ITD;Dnmt3afl/flMxCre AML versus Flt3ITD/ITD;Dnmt3afl/fl MPN as described above.

In Vitro Colony-Forming Assays

Equal numbers of cells were mixed with methylcellulose M3434 complete with cytokines SCF, IL6, IL3, and Erythropoietin (Stem Cell Technologies) in triplicate according to the manufacturer's instructions. c-KIT+ or c-KIT cells (2,000) where indicated were seeded per plate. On day 7, immortalized/transformed colonies >500 cells were enumerated. Colonies were dissociated with PBS, single-cell suspensions were counted, and equal numbers of cells were either replated or used in other indicated downstream applications. Each group was performed in triplicate and results were repeated in independent experiments at least twice.

Statistical Analyses

The absolute cell numbers of mouse normal, MPN, and AML total bone marrow and spleen counts were converted to natural log prior to statistical evaluation to normalize the variance between groups. Significant differences between normal, MPN, and AML groups for each population were examined by two-way ANOVA with Tukey multiple comparisons correction. The average cell counts are shown for at least 3 mice per group.

The percentage of total bone marrow and spleen LK, GMP, and LSK and spleen CD11b+Gr1+CD3, CD11b+Gr1CD3, and CD11b+Gr1B220 populations are shown for at least 3 mice per group. Significant differences among normal, MPN, and AML groups for each population were examined by two-way ANOVA with Tukey multiple comparisons correction. Descriptions of statistical analyses used can also be found in the figure legends.

Accession Numbers

RNA sequencing (GEO accession GSE77849) and ERRBS data (GEO accession GSE77026) can be accessed in the Gene Expression Omnibus.

A. Melnick reports receiving commercial research grants from Janssen, Roche, and Eli Lilly and is a consultant/advisory board member for Boehringer, Epizyme, and Roche. No potential conflicts of interest were disclosed by the other authors.

Conception and design: E. Paietta, M.E. Figueroa, H.L. Grimes

Development of methodology: N. Salomonis, H.L. Grimes

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.E. Meyer, D.E. Muench, K. Masuda, E. Orr, L. Suarez, S.D. Gore, R. Delwel, E. Paietta, M.S. Tallman, H. Fernandez, A. Melnick, M.M. Le Beau

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.E. Meyer, T. Qin, D.E. Muench, K. Masuda, M. Venkatasubramanian, R. Delwel, M.S. Tallman, A. Melnick, S. Kogan, N. Salomonis, M.E. Figueroa, H.L. Grimes

Writing, review, and/or revision of the manuscript: S.E. Meyer, T. Qin, D.E. Muench, L. Suarez, S.D. Gore, E. Paietta, H. Fernandez, A. Melnick, M.M. Le Beau, S. Kogan, M.E. Figueroa, H.L. Grimes

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Delwel, H.L. Grimes

Study supervision: H. Fernandez, M.E. Figueroa, H.L. Grimes

The authors acknowledge the Cornell University Epigenetics Core, the University of Michigan Epigenomics Core facility, and the Cincinnati Children's Hospital Pathology, Flow Cytometry, Gene Expression, and Comprehensive Mouse Cores. They also thank Dr. Peter Valk (Erasmus University Medical Center, the Netherlands) for human AML samples and gene expression data. c-MyceGFP knock-in mice were a kind gift from Barry Sleckman (Washington University, St. Louis, MO). Dnmt3afl/fl mice were a kind gift from David Hildeman (CCHMC). DNMT3A retroviral constructs were a kind gift from Grant Challen (Washington University, St. Louis, MO).

This research was supported by a Ladies Auxiliary to the Veterans of Foreign Wars (LAVFW) Postdoctoral Fellowship (to S.E. Meyer); CancerFree KIDS (Loveland, Ohio); University of Cincinnati training grant NIEHS T32ES007250 (to S.E. Meyer); CCTST Just-In-Time Core Grant Program (University of Cincinnati and CCHMC); a Leukemia and Lymphoma Society Translational Research Program Award, NHLBI Progenitor Cell Biology Consortium U01HL099997 (to N. Salomonis); and NIH R01CA159845, CA196658, and R21CA186945 (to H.L. Grimes).

1.
Dohner
K
,
Dohner
H
. 
Molecular characterization of acute myeloid leukemia
.
Haematologica
2008
;
93
:
976
82
.
2.
CGARN CGARN
. 
Genomic and epigenomic landscapes of adult de novo acute myeloid leukemia
.
N Engl J Med
2013
;
368
:
2059
74
.
3.
Patel
JP
,
Gonen
M
,
Figueroa
ME
,
Fernandez
H
,
Sun
Z
,
Racevskis
J
, et al
Prognostic relevance of integrated genetic profiling in acute myeloid leukemia
.
N Engl J Med
2012
;
366
:
1079
89
.
4.
Kiyoi
H
,
Naoe
T
,
Nakano
Y
,
Yokota
S
,
Minami
S
,
Miyawaki
S
, et al
Prognostic implication of FLT3 and N-RAS gene mutations in acute myeloid leukemia
.
Blood
1999
;
93
:
3074
80
.
5.
Schnittger
S
,
Schoch
C
,
Dugas
M
,
Kern
W
,
Staib
P
,
Wuchter
C
, et al
Analysis of FLT3 length mutations in 1003 patients with acute myeloid leukemia: correlation to cytogenetics, FAB subtype, and prognosis in the AMLCG study and usefulness as a marker for the detection of minimal residual disease
.
Blood
2002
;
100
:
59
66
.
6.
Thiede
C
,
Steudel
C
,
Mohr
B
,
Schaich
M
,
Schakel
U
,
Platzbecker
U
, et al
Analysis of FLT3-activating mutations in 979 patients with acute myelogenous leukemia: association with FAB subtypes and identification of subgroups with poor prognosis
.
Blood
2002
;
99
:
4326
35
.
7.
Kottaridis
PD
,
Gale
RE
,
Frew
ME
,
Harrison
G
,
Langabeer
SE
,
Belton
AA
, et al
The presence of a FLT3 internal tandem duplication in patients with acute myeloid leukemia (AML) adds important prognostic information to cytogenetic risk group and response to the first cycle of chemotherapy: analysis of 854 patients from the United Kingdom Medical Research Council AML 10 and 12 trials
.
Blood
2001
;
98
:
1752
9
.
8.
Tefferi
A
,
Lim
KH
,
Levine
R
. 
Mutation in TET2 in myeloid cancers
.
N Engl J Med
2009
;
361
:
1117
;
author reply -8
.
9.
Delhommeau
F
,
Dupont
S
,
Della Valle
V
,
James
C
,
Trannoy
S
,
Masse
A
, et al
Mutation in TET2 in myeloid cancers
.
N Engl J Med
2009
;
360
:
2289
301
.
10.
Ley
TJ
,
Ding
L
,
Walter
MJ
,
McLellan
MD
,
Lamprecht
T
,
Larson
DE
, et al
DNMT3A mutations in acute myeloid leukemia
.
N Engl J Med
2010
;
363
:
2424
33
.
11.
Yan
XJ
,
Xu
J
,
Gu
ZH
,
Pan
CM
,
Lu
G
,
Shen
Y
, et al
Exome sequencing identifies somatic mutations of DNA methyltransferase gene DNMT3A in acute monocytic leukemia
.
Nat Genet
2011
;
43
:
309
15
.
12.
Marcucci
G
,
Metzeler
KH
,
Schwind
S
,
Becker
H
,
Maharry
K
,
Mrozek
K
, et al
Age-related prognostic impact of different types of DNMT3A mutations in adults with primary cytogenetically normal acute myeloid leukemia
.
J Clin Oncol
2012
;
30
:
742
50
.
13.
Ribeiro
AF
,
Pratcorona
M
,
Erpelinck-Verschueren
C
,
Rockova
V
,
Sanders
M
,
Abbas
S
, et al
Mutant DNMT3A: a marker of poor prognosis in acute myeloid leukemia
.
Blood
2012
;
119
:
5824
31
.
14.
Im
AP
,
Sehgal
AR
,
Carroll
MP
,
Smith
BD
,
Tefferi
A
,
Johnson
DE
, et al
DNMT3A and IDH mutations in acute myeloid leukemia and other myeloid malignancies: associations with prognosis and potential treatment strategies
.
Leukemia
2014
;
28
:
1774
83
.
15.
van Vlodrop
IJ
,
Niessen
HE
,
Derks
S
,
Baldewijns
MM
,
van Criekinge
W
,
Herman
JG
, et al
Analysis of promoter CpG island hypermethylation in cancer: location, location, location!
Clin Cancer Res
2011
;
17
:
4225
31
.
16.
Shih
AH
,
Abdel-Wahab
O
,
Patel
JP
,
Levine
RL
. 
The role of mutations in epigenetic regulators in myeloid malignancies
.
Nat Rev Cancer
2012
;
12
:
599
612
.
17.
Schoofs
T
,
Berdel
WE
,
Muller-Tidow
C
. 
Origins of aberrant DNA methylation in acute myeloid leukemia
.
Leukemia
2014
;
28
:
1
14
.
18.
Russler-Germain
DA
,
Spencer
DH
,
Young
MA
,
Lamprecht
TL
,
Miller
CA
,
Fulton
R
, et al
The R882H DNMT3A mutation associated with AML dominantly inhibits wild-type DNMT3A by blocking its ability to form active tetramers
.
Cancer Cell
2014
;
25
:
442
54
.
19.
Holz-Schietinger
C
,
Matje
DM
,
Reich
NO
. 
Mutations in DNA methyltransferase (DNMT3A) observed in acute myeloid leukemia patients disrupt processive methylation
.
J Biol Chem
2012
;
287
:
30941
51
.
20.
Okano
M
,
Bell
DW
,
Haber
DA
,
Li
E
. 
DNA methyltransferases Dnmt3a and Dnmt3b are essential for de novo methylation and mammalian development
.
Cell
1999
;
99
:
247
57
.
21.
Challen
GA
,
Sun
D
,
Jeong
M
,
Luo
M
,
Jelinek
J
,
Berg
JS
, et al
Dnmt3a is essential for hematopoietic stem cell differentiation
.
Nat Genet
2012
;
44
:
23
31
.
22.
Nakao
M
,
Yokota
S
,
Iwai
T
,
Kaneko
H
,
Horiike
S
,
Kashima
K
, et al
Internal tandem duplication of the flt3 gene found in acute myeloid leukemia
.
Leukemia
1996
;
10
:
1911
8
.
23.
Kiyoi
H
,
Towatari
M
,
Yokota
S
,
Hamaguchi
M
,
Ohno
R
,
Saito
H
, et al
Internal tandem duplication of the FLT3 gene is a novel modality of elongation mutation which causes constitutive activation of the product
.
Leukemia
1998
;
12
:
1333
7
.
24.
Kiyoi
H
,
Ohno
R
,
Ueda
R
,
Saito
H
,
Naoe
T
. 
Mechanism of constitutive activation of FLT3 with internal tandem duplication in the juxtamembrane domain
.
Oncogene
2002
;
21
:
2555
63
.
25.
Mead
AJ
,
Kharazi
S
,
Atkinson
D
,
Macaulay
I
,
Pecquet
C
,
Loughran
S
, et al
FLT3-ITDs instruct a myeloid differentiation and transformation bias in lymphomyeloid multipotent progenitors
.
Cell Rep
2013
;
3
:
1766
76
.
26.
Sexauer
A
,
Perl
A
,
Yang
X
,
Borowitz
M
,
Gocke
C
,
Rajkhowa
T
, et al
Terminal myeloid differentiation in vivo is induced by FLT3 inhibition in FLT3/ITD AML
.
Blood
2012
;
120
:
4205
14
.
27.
Lee
BH
,
Tothova
Z
,
Levine
RL
,
Anderson
K
,
Buza-Vidas
N
,
Cullen
DE
, et al
FLT3 mutations confer enhanced proliferation and survival properties to multipotent progenitors in a murine model of chronic myelomonocytic leukemia
.
Cancer Cell
2007
;
12
:
367
80
.
28.
Hayakawa
F
,
Towatari
M
,
Kiyoi
H
,
Tanimoto
M
,
Kitamura
T
,
Saito
H
, et al
Tandem-duplicated Flt3 constitutively activates STAT5 and MAP kinase and introduces autonomous cell growth in IL-3-dependent cell lines
.
Oncogene
2000
;
19
:
624
31
.
29.
Mizuki
M
,
Fenski
R
,
Halfter
H
,
Matsumura
I
,
Schmidt
R
,
Muller
C
, et al
Flt3 mutations from patients with acute myeloid leukemia induce transformation of 32D cells mediated by the Ras and STAT5 pathways
.
Blood
2000
;
96
:
3907
14
.
30.
Spiekermann
K
,
Bagrintseva
K
,
Schwab
R
,
Schmieja
K
,
Hiddemann
W
. 
Overexpression and constitutive activation of FLT3 induces STAT5 activation in primary acute myeloid leukemia blast cells
.
Clin Cancer Res
2003
;
9
:
2140
50
.
31.
Fiskus
W
,
Sharma
S
,
Qi
J
,
Shah
B
,
Devaraj
SG
,
Leveque
C
, et al
BET protein antagonist JQ1 is synergistically lethal with FLT3 tyrosine kinase inhibitor (TKI) and overcomes resistance to FLT3-TKI in AML cells expressing FLT-ITD
.
Mol Cancer Ther
2014
;
13
:
2315
27
.
32.
Tickenbrock
L
,
Schwable
J
,
Wiedehage
M
,
Steffen
B
,
Sargin
B
,
Choudhary
C
, et al
Flt3 tandem duplication mutations cooperate with Wnt signaling in leukemic signal transduction
.
Blood
2005
;
105
:
3699
706
.
33.
Kelly
LM
,
Liu
Q
,
Kutok
JL
,
Williams
IR
,
Boulton
CL
,
Gilliland
DG
. 
FLT3 internal tandem duplication mutations associated with human acute myeloid leukemias induce myeloproliferative disease in a murine bone marrow transplant model
.
Blood
2002
;
99
:
310
8
.
34.
Lee
BH
,
Williams
IR
,
Anastasiadou
E
,
Boulton
CL
,
Joseph
SW
,
Amaral
SM
, et al
FLT3 internal tandem duplication mutations induce myeloproliferative or lymphoid disease in a transgenic mouse model
.
Oncogene
2005
;
24
:
7882
92
.
35.
Zorko
NA
,
Bernot
KM
,
Whitman
SP
,
Siebenaler
RF
,
Ahmed
EH
,
Marcucci
GG
, et al
Mll partial tandem duplication and Flt3 internal tandem duplication in a double knock-in mouse recapitulates features of counterpart human acute myeloid leukemias
.
Blood
2012
;
120
:
1130
6
.
36.
Kogan
SC
,
Ward
JM
,
Anver
MR
,
Berman
JJ
,
Brayton
C
,
Cardiff
RD
, et al
Bethesda proposals for classification of nonlymphoid hematopoietic neoplasms in mice
.
Blood
2002
;
100
:
238
45
.
37.
Akalin
A
,
Garrett-Bakelman
FE
,
Kormaksson
M
,
Busuttil
J
,
Zhang
L
,
Khrebtukova
I
, et al
Base-pair resolution DNA methylation sequencing reveals profoundly divergent epigenetic landscapes in acute myeloid leukemia
.
PLoS Genet
2012
;
8
:
e1002781
.
38.
Figueroa
ME
,
Abdel-Wahab
O
,
Lu
C
,
Ward
PS
,
Patel
J
,
Shih
A
, et al
Leukemic IDH1 and IDH2 mutations result in a hypermethylation phenotype, disrupt TET2 function, and impair hematopoietic differentiation
.
Cancer Cell
2010
;
18
:
553
67
.
39.
Figueroa
ME
,
Lugthart
S
,
Li
Y
,
Erpelinck-Verschueren
C
,
Deng
X
,
Christos
PJ
, et al
DNA methylation signatures identify biologically distinct subtypes in acute myeloid leukemia
.
Cancer Cell
2010
;
17
:
13
27
.
40.
Challen
GA
,
Sun
D
,
Mayle
A
,
Jeong
M
,
Luo
M
,
Rodriguez
B
, et al
Dnmt3a and dnmt3b have overlapping and distinct functions in hematopoietic stem cells
.
Cell stem cell
2014
;
15
:
350
64
.
41.
Mootha
VK
,
Lindgren
CM
,
Eriksson
KF
,
Subramanian
A
,
Sihag
S
,
Lehar
J
, et al
PGC-1alpha-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes
.
Nat Genet
2003
;
34
:
267
73
.
42.
Subramanian
A
,
Tamayo
P
,
Mootha
VK
,
Mukherjee
S
,
Ebert
BL
,
Gillette
MA
, et al
Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles
.
Proc Natl Acad Sci U S A
2005
;
102
:
15545
50
.
43.
CY
Huang
,
AL
Bredemeyer
,
LM
Walker
,
CH
Bassing
,
BP
Sleckman
. 
Dynamic regulation of c-Myc proto-oncogene expression during lymphocyte development revealed by a GFP-c-Myc knock-in mouse
.
Eur J Immunol
2008
;
38
:
342
9
.
44.
Emig
D
,
Salomonis
N
,
Baumbach
J
,
Lengauer
T
,
Conklin
BR
,
Albrecht
M
. 
AltAnalyze and DomainGraph: analyzing and visualizing exon expression data
.
Nucleic Acids Res
2010
;
38
:
W755
62
.
45.
Zambon
AC
,
Gaj
S
,
Ho
I
,
Hanspers
K
,
Vranizan
K
,
Evelo
CT
, et al
GO-Elite: a flexible solution for pathway and ontology over-representation
.
Bioinformatics
2012
;
28
:
2209
10
.
46.
Visvader
JE
,
Lindeman
GJ
. 
Cancer stem cells: current status and evolving complexities
.
Cell Stem Cell
2012
;
10
:
717
28
.
47.
Bourdeau
A
,
Trop
S
,
Doody
KM
,
Dumont
DJ
,
Tremblay
ML
. 
Inhibition of T cell protein tyrosine phosphatase enhances interleukin-18-dependent hematopoietic stem cell expansion
.
Stem Cells
2013
;
31
:
293
304
.
48.
Rau
R
,
Magoon
D
,
Greenblatt
S
,
Li
L
,
Annesley
C
,
Duffield
AS
, et al
NPMc+ cooperates with Flt3/ITD mutations to cause acute leukemia recapitulating human disease
.
Exp Hematol
2014
;
42
:
101
13
e5
.
49.
Guryanova
OA
,
Lieu
YK
,
Garrett-Bakelman
FE
,
Spitzer
B
,
Glass
JL
,
Shank
K
, et al
Dnmt3a regulates myeloproliferation and liver-specific expansion of hematopoietic stem and progenitor cells
.
Leukemia
2015
.
50.
Chang
YI
,
You
X
,
Kong
G
,
Ranheim
EA
,
Wang
J
,
Du
J
, et al
Loss of Dnmt3a and endogenous Kras(G12D/+) cooperate to regulate hematopoietic stem and progenitor cell functions in leukemogenesis
.
Leukemia
2015
;
29
:
1847
56
.
51.
Schlenk
RF
,
Kayser
S
,
Bullinger
L
,
Kobbe
G
,
Casper
J
,
Ringhoffer
M
, et al
Differential impact of allelic ratio and insertion site in FLT3–ITD-positive AML with respect to allogeneic transplantation
.
Blood
2014
;
124
:
3441
9
.
52.
Kim
MS
,
Yoon
SK
,
Bollig
F
,
Kitagaki
J
,
Hur
W
,
Whye
NJ
, et al
A novel Wilms tumor 1 (WT1) target gene negatively regulates the WNT signaling pathway
.
J Biol Chem
2010
;
285
:
14585
93
.
53.
Andersson
T
,
Sodersten
E
,
Duckworth
JK
,
Cascante
A
,
Fritz
N
,
Sacchetti
P
, et al
CXXC5 is a novel BMP4-regulated modulator of Wnt signaling in neural stem cells
.
J Biol Chem
2009
;
284
:
3672
81
.
54.
Liu
N
,
Wang
M
,
Deng
W
,
Schmidt
CS
,
Qin
W
,
Leonhardt
H
, et al
Intrinsic and extrinsic connections of Tet3 dioxygenase with CXXC zinc finger modules
.
PLoS One
2013
;
8
:
e62755
.
55.
Treppendahl
MB
,
Mollgard
L
,
Hellstrom-Lindberg
E
,
Cloos
P
,
Gronbaek
K
. 
Downregulation but lack of promoter hypermethylation or somatic mutations of the potential tumor suppressor CXXC5 in MDS and AML with deletion 5q
.
Eur J Haematol
2013
;
90
:
259
60
.
56.
Kuhnl
A
,
Valk
PJ
,
Sanders
MA
,
Ivey
A
,
Hills
RK
,
Mills
KI
, et al
Downregulation of the Wnt inhibitor CXXC5 predicts a better prognosis in acute myeloid leukemia
.
Blood
2015
;
125
:
2985
94
.
57.
Astori
A
,
Fredly
H
,
Aloysius
TA
,
Bullinger
L
,
Mansat-De Mas
V
,
de la Grange
P
, et al
CXXC5 (retinoid-inducible nuclear factor, RINF) is a potential therapeutic target in high-risk human acute myeloid leukemia
.
Oncotarget
2013
;
4
:
1438
48
.
58.
Marastoni
S
,
Andreuzzi
E
,
Paulitti
A
,
Colladel
R
,
Pellicani
R
,
Todaro
F
, et al
EMILIN2 down-modulates the Wnt signalling pathway and suppresses breast cancer cell growth and migration
.
J Pathol
2014
;
232
:
391
404
.
59.
Genovese
G
,
Kahler
AK
,
Handsaker
RE
,
Lindberg
J
,
Rose
SA
,
Bakhoum
SF
, et al
Clonal hematopoiesis and blood-cancer risk inferred from blood DNA sequence
.
N Engl J Med
2014
;
371
:
2477
87
.
60.
Xie
M
,
Lu
C
,
Wang
J
,
McLellan
MD
,
Johnson
KJ
,
Wendl
MC
, et al
Age-related mutations associated with clonal hematopoietic expansion and malignancies
.
Nat Med
2014
;
20
:
1472
8
.
61.
Kaneda
M
,
Okano
M
,
Hata
K
,
Sado
T
,
Tsujimoto
N
,
Li
E
, et al
Essential role for de novo DNA methyltransferase Dnmt3a in paternal and maternal imprinting
.
Nature
2004
;
429
:
900
3
.
62.
Kuhn
R
,
Schwenk
F
,
Aguet
M
,
Rajewsky
K
. 
Inducible gene targeting in mice
.
Science
1995
;
269
:
1427
9
.
63.
Le Beau
MM
,
Bitts
S
,
Davis
EM
,
Kogan
SC
. 
Recurring chromosomal abnormalities in leukemia in PML-RARA transgenic mice parallel human acute promyelocytic leukemia
.
Blood
2002
;
99
:
2985
91
.
64.
Meyer
SE
,
Zinser
GM
,
Stuart
WD
,
Pathrose
P
,
Waltz
SE
. 
The Ron receptor tyrosine kinase negatively regulates mammary gland branching morphogenesis
.
Develop Biol
2009
;
333
:
173
85
.
65.
Kim
D
,
Pertea
G
,
Trapnell
C
,
Pimentel
H
,
Kelley
R
,
Salzberg
SL
. 
TopHat2: accurate alignment of transcriptomes in the presence of insertions, deletions and gene fusions
.
Genome Biol
2013
;
14
:
R36
.
66.
Trapnell
C
,
Pachter
L
,
Salzberg
SL
. 
TopHat: discovering splice junctions with RNA-Seq
.
Bioinformatics
2009
;
25
:
1105
11
.
67.
Li
B
,
Dewey
CN
. 
RSEM: accurate transcript quantification from RNA-Seq data with or without a reference genome
.
BMC Bioinformatics
2011
;
12
:
323
.
68.
Verhaak
RG
,
Wouters
BJ
,
Erpelinck
CA
,
Abbas
S
,
Beverloo
HB
,
Lugthart
S
, et al
Prediction of molecular subtypes in acute myeloid leukemia based on gene expression profiling
.
Haematologica
2009
;
94
:
131
4
.
69.
Shih
AH
,
Jiang
Y
,
Meydan
C
,
Shank
K
,
Pandey
S
,
Barreyro
L
, et al
Mutational cooperativity linked to combinatorial epigenetic gain of function in acute myeloid leukemia
.
Cancer Cell
2015
;
27
:
502
15
.
70.
Garrett-Bakelman
FE
,
Sheridan
CK
,
Kacmarczyk
TJ
,
Ishii
J
,
Betel
D
,
Alonso
A
, et al
Enhanced reduced representation bisulfite sequencing for assessment of DNA methylation at base pair resolution
.
J Vis Exp
2015
:
e52246
.
71.
Krueger
F
,
Andrews
SR
. 
Bismark: a flexible aligner and methylation caller for Bisulfite-Seq applications
.
Bioinformatics
2011
;
27
:
1571
2
.
72.
Park
Y
,
Figueroa
ME
,
Rozek
LS
,
Sartor
MA
. 
MethylSig: a whole genome DNA methylation analysis pipeline
.
Bioinformatics
2014
;
30
:
2414
22
.
73.
Akalin
A
,
Kormaksson
M
,
Li
S
,
Garrett-Bakelman
FE
,
Figueroa
ME
,
Melnick
A
, et al
methylKit: a comprehensive R package for the analysis of genome-wide DNA methylation profiles
.
Genome Biol
2012
;
13
:
R87
.
74.
Meldi
K
,
Qin
T
,
Buchi
F
,
Droin
N
,
Sotzen
J
,
Micol
JB
, et al
Specific molecular signatures predict decitabine response in chronic myelomonocytic leukemia
.
J Clin Invest
2015
;
125
:
1857
72
.
75.
Naviaux
RK
,
Costanzi
E
,
Haas
M
,
Verma
IM
. 
The pCL vector system: rapid production of helper-free, high-titer, recombinant retroviruses
.
J Virol
1996
;
70
:
5701
5
76.
Naldini
L
,
Blomer
U
,
Gage
FH
,
Trono
D
,
Verma
IM
. 
Efficient transfer, integration, and sustained long-term expression of the transgene in adult rat brains injected with a lentiviral vector
.
Proc Natl Acad Sci U S A
1996
;
93
:
11382
8
77.
Burns
JC
,
Friedmann
T
,
Driever
W
,
Burrascano
M
,
Yee
JK
. 
Vesicular stomatitis virus G glycoprotein pseudotyped retroviral vectors: concentration to very high titer and efficient gene transfer into mammalian and nonmammalian cells
.
Proc Natl Acad Sci U S A
1993
;
90
:
8033
7
.

Supplementary data