Epigenetic Therapy Restores Normal Hematopoiesis in a Zebrafish Model

ORIGINAL ARTICLE Epigenetic therapy restores normal hematopoiesis in a zebraﬁsh model of NUP98–HOXA9-induced myeloid disease

AP Deveau1,2,15, AM Forrester1,2,15, AJ Coombs2,3, GS Wagner2,3, C Grabher4,5, IC Chute6, D Léger6, M Mingay7, G Alexe4,8, V Rajan1,2, R Liwski9,10, M Hirst7,11, K Steigmaier4,8, SM Lewis1,6,12,13, AT Look4 and JN Berman1,2,10,14

Acute myeloid leukemia (AML) occurs when multiple genetic aberrations alter white blood cell development, leading to hyperproliferation and arrest of cell differentiation. Pertinent animal models link in vitro studies with the use of new agents in clinical trials. We generated a transgenic zebrafish expressing human NUP98–HOXA9 (NHA9), a fusion oncogene found in high-risk AML. Embryos developed a preleukemic state with anemia and myeloid cell expansion, and adult fish developed a myeloproliferative neoplasm (MPN). We leveraged this model to show that NHA9 increases the number of hematopoietic stem cells, and that oncogenic function of NHA9 depends on downstream activation of meis1, the PTGS/COX pathway and genome hypermethylation through the DNA methyltransferase, dnmt1. We restored normal hematopoiesis in NHA9 embryos with knockdown of meis1 or dnmt1, as well as pharmacologic treatment with DNA (cytosine-5)-methyltransferase (DNMT) inhibitors or cyclo-oxygenase (COX) inhibitors. DNMT inhibitors reduced genome methylation to near normal levels. Strikingly, we discovered synergy when we combined sub-monotherapeutic doses of a histone deacetylase inhibitor plus either a DNMT inhibitor or COX inhibitor to block the effects of NHA9 on zebrafish blood development. Our work proposes novel drug targets in NHA9-induced myeloid disease, and suggests rational therapies by combining minimal doses of known bioactive compounds.

Leukemia (2015) 29, 2086–2097; doi:10.1038/leu.2015.126

INTRODUCTION in vivo studies to identify novel gene collaborators or to verify new Current treatment strategies for acute myeloid leukemia (AML) are drug discoveries. aggressive and use nonspecific cytotoxic drugs.1,2 These regimens In mouse models and human cell culture, overexpression of have led to modest improvements in survival over the last 10 either HOXA9 or NHA9 increases the number of hematopoietic stem cells (HSCs), suppresses myeloid differentiation and main- years, but are often associated with substantial morbidity and risk 10–13 of induction-related death.3,4 In the current genomic era, efforts tains survival of progenitor cells. NHA9 induces a preleukemic are underway to develop more rational therapeutic strategies by myeloproliferative neoplasm (MPN) in mouse models, in contrast to the myelodysplastic syndrome (MDS) seen in humans harboring targeting specific molecular abnormalities with small molecule NHA9.14 Co-overexpression of Hoxa9 and Meis1 in mice produces inhibitors improving efficacy and reducing toxicity. reduced latency over Hoxa9 alone (3 vs 8 months).15 Meis1 One particularly high-risk group of patients harbor the NUP98- overexpression also accelerates the progression to overt AML in HOXA9 (NHA9) fusion oncogene. NHA9 results from a t(7;11)(p15; NHA9-expressing mice, but maintains the preleukemic MPN p15) chromosomal translocation that fuses nucleoporin 98 kDa phase.16 Thus, HOXA9 and NHA9 appear to rely on multiple (NUP98) to homeobox A9 (HOXA9), an essential gene for 5 6 collaborating mutations to induce AML. However, other than hematopoiesis, leading to overexpression of the latter. Over MEIS1, few candidate gene or pathways have been identified. 80% of human AML cases show overexpression of HOXA9, and The zebrafish, Danio rerio, has been firmly established as a frequent upregulation of its cofactor, myeloid ecotropic integra- reliable in vivo tool for modeling human leukemia,17,18 with ease 7 tion site 1 (MEIS1). HOXA9 overexpression is the single most of genetic interrogation and inherent capacity for drug important predictor of treatment failure using traditional screening.19,20 We previously created a transgenic zebrafish chemotherapy.8 Moreover, the NHA9 oncogene confers inferior expressing human NHA9 downstream of a Cre/lox-inducible prognosis in de novo and treatment-related AML, and progression cassette under the 9.1-kb zebrafish spi1 myeloid promoter.21 to blast crisis in chronic myelogenous leukemia (CML).9 Targeted Nearly 25% of adult NHA9 fish develop MPN between 19 and therapies are needed, but there has been a relative paucity of 23 months of age. In embryos, NHA9 disrupts early blood

1Department of Microbiology & Immunology, Dalhousie University, Halifax, Nova Scotia, Canada; 2Department of Pediatrics, IWK Health Centre, Halifax, Nova Scotia, Canada; 3Department of Marine Biology, Dalhousie University, Halifax, Nova Scotia, Canada; 4Department of Pediatric Oncology, Dana-Farber Cancer Institute, Boston, MA, USA; 5Institute of Toxicology & Genetics, Karlsruhe Institute of Technology, Karlsruhe, Germany; 6Department of Biology, University of New Brunswick, Saint John, New Brunswick, Canada; 7Department of Microbiology & Immunology, Centre for High-Throughput Biology, University of British Columbia, Vancouver, British Columbia, Canada; 8Harvard Medical School, Boston, MA, USA; 9Department of Pathology, Queen Elizabeth II Health Science Centre, Halifax, Nova Scotia, Canada; 10Department of Pathology, Dalhousie University, Halifax, Nova Scotia, Canada; 11Canada's Michael Smith Genome Sciences Centre, BC Cancer Agency, Vancouver, British Columbia, Canada; 12Department of Chemistry & Biochemistry, Université de Moncton, Moncton, New Brunswick, Canada; 13Atlantic Cancer Research Institute, Moncton, New Brunswick, Canada and 14Department of Pediatrics, Dalhousie University, Halifax, Nova Scotia, Canada. Correspondence: Dr JN Berman, Berman Lab N103, Department of Pediatrics, Life Sciences Research Institute, Dalhousie University, 1348 Summer Street, Halifax, Nova Scotia B3H 4R2, Canada. E-mail: [email protected] 15These authors contributed equally to this work. Received 21 November 2014; revised 7 April 2015; accepted 22 April 2015; accepted article preview 28 May 2015; advance online publication, 3 July 2015 Epigenetic therapy blocks NHA9-mediated disease AP Deveau et al 2087 development with a gain in myeloid cells at the expense of Quantitative reverse-transcription PCR erythroid cells, with evidence of arrested myeloid differentiation. Embryos were heat-shocked at 24 h.p.f., grown to 28 h.p.f., then stabilized Here we show that NHA9 cooperates with zebrafish meis1 and with RNAlater (Ambion; Carlsbad, CA, USA). RNA was extracted using expands the number of early HSCs, which confirms the shared RNeasy Mini Kit, cDNA synthesized using QuantiTect Reverse Transcription features of our fish with previously studied mammalian models. Kit and quantitative reverse-transcription PCR (qRT-PCR) performed using We leveraged the blood developmental phenotypes in transgenic QuantiFast SYBR Green PCR Kit (QIAGEN; Valencia, CA, USA) with fi Stratagene Mx3000P QPCR thermocycler (Agilent; Santa Clara, CA, USA). zebra sh and found that NHA9 upregulates ptgs2 (prostaglandin Quantification used the 2 − ΔΔCt method (details in Supplementary synthase 2), which encodes a cyclo-oxygenase (COX) enzyme Methods). All gene expression studies included a minimum of three 22,23 isoform. Blocking this pathway with COX inhibitors restores independent experiments. normal hematopoiesis in our NHA9 embryos, which is in keeping with recent studies in AML1-ETO zebrafish20 and Hoxa9;Meis1 24 Fluorescence-activated cell sorting mice. Furthermore, using an unbiased microarray-based strat- Groups of 30–50 embryos were dissociated to a single cell suspension egy, we discovered that NHA9 activity depends on the over- (described by Covassin et al., 2009) and analyzed by FACSAriaIII (BD expression of DNA (cytosine-5)-methyltransferase 1 (dnmt1), which Biosciences; Mississauga, ON, Canada) at 4 °C. GFP+ stem cells were excited reveals a direct link between NHA9 and epigenetic regulation for using 488-nm emission and gated for the unique ‘GFP-low’ group. the first time. Finally, we demonstrate the use of synergetic drug Statistical analysis was done using the student’s t-test. combinations by combining submonotherapeutic doses of DNMT or COX inhibitors with histone deacetylase (HDAC) inhibitors to Drug treatments restore normal blood development in NHA9 embryos. Embryos were treated with either 0.3% dimethylsulfoxide (DMSO), or indicated doses of decitabine (DAC), zebularine (Zeb), valproic acid sodium salt (VPA), trichostatin A (TSA), indomethacin (Indo) or NS-398 (all MATERIALS AND METHODS compounds from Sigma-Aldrich; St Louis, MO, USA). Details of dose – fi optimization can be found in Supplementary Methods. Groups of 25 30 Zebra sh husbandry embryos were arrayed into six-well plates with 6 ml pure E3 medium. Zebrafish were maintained, bred and developmentally staged according to Compounds were added, plates were immediately heat-shocked, and then Westerfield.25 The Dalhousie University Animal Care Committee approved incubated at 28 °C until fixation. use of zebrafish (protocol no. 13–129). AB zebrafish were used as wild-type (WT) controls. Reporter lines Tg(cd41::eGFP) and Tg(runx1::eGFP) were gifts Microarray experiments from Drs Robert Handin (Brigham and Women’s Hospital, Boston, MA, USA) and Phil Crosier (University of Auckland, NZ), respectively. RNA was extracted using Trizol+RNeasy Mini Kit, then reverse transcribed and amplified using the Amino-Allyl Message Amp II kit (Life Technologies; Carlsbad, CA, USA). Purified aRNA samples were labeled with Alexa Fluor NUP98–HOXA9 activation by Cre recombinase and experimental dyes and hybridized at 65 °C overnight in a duplicate dye swap experiment time points to an Agilent 4 × 44 k zebrafish V3 slide (part# G2519F; Agilent) using an Generation of transgenic fish was previously described in Forrester et al.21 HS4800 Pro hybridization station (Tecan; Durham, NC, USA). Analysis Briefly, NHA9 expression is achieved by outcross of Tg(spi1::lox-eGFP-lox:: employed a GenePix 4200AL scanner and GenePix Pro 6.0, Acuity 4.0 NHA9) to Tg(heat-shock protein 70[hsp70]::Cre) and incubation of embryos software (Molecular Devices; Sunnyvale, CA, USA) Transcripts with expression changes ⩾ 2-fold were considered significant. Details of at 37 °C for 1 h (‘heat-shock’). Embryos were kept at 28 °C until fixation. hybridization and identifying mammalian homologs are contained in Unless otherwise stated in the text, staging and experimental endpoints Supplementary Methods. were as follows. To assess gene expression in posterior blood island (PBI), embryos were heat-shocked at 24 h post fertilization (h.p.f.) and fixed at 30 h.p.f. To assess gene expression in HSCs of dorsal aorta, embryos were Methylated DNA immunoprecipitation heat-shocked at 12 h.p.f. and fixed at 36 h.p.f. Genomic DNA isolated from 24 and 36 h.p.f. zebrafish embryo pools was sonicated to approximately 200–400 bp with a Covaris E-220 Sonicator (Covaris) and end-repaired using standard Illumina library preparation Morpholinos and mRNA constructs protocol. DNA was immunoprecipitated with anti-5-methylcytidine anti- Morpholino oligonucleotides (MOs) were purchased from Genetools body (1 mg/ml; Eurogentec) per microgram of DNA. Immunoabsorbed – (Philomath, OR, USA), and were previously published26 28 (Supple- DNA fragments were recovered by adding a rabbit anti-mouse IgG mentary Methods). All MOs targeted the ATG/5′UTR sequence. Primers secondary antibody (2.5 mg/ml, Jackson ImmunoResearch) and immuno- designed to clone human NHA9 (1.9-kb amplicon) into pCS2+ expression precipitation. To complete library preparation, two 10-cycle PCR reactions vector (Gateway sequence; forward: 5′-GGGGACAAGTTTGTACAAAAAAGCA using 16 μl of DNA were performed for each sample with paired-end GGCTTACCGAAGCGGCGGTCGGTG-3′; reverse: 5′-GGGGACCACTTTGTACAA Illumina PCR primers. A final size selection was performed to select all GAAAGCTGGGTATCACTCGTCTTTTGCTCG-3′). Plasmids were linearized with fragments o700 bp in size by electrophoresis in 8% polyacrylamide gel. NotI enzyme. mRNA was synthesized with mMESSAGE mMACHINE SP6 Kit, The quality of the final libraries was confirmed by Agilent DNA Bioanalyzer and cleaned using NucAway Spin Columns (Applied Biosystems/Ambion; analysis. The libraries were pooled and sequenced over two 2 × 75-bp runs Carlsbad, CA, USA). MOs and mRNA were diluted to working concentra- on the Illumina MiSeq (Illumina). Complete details of DNA preparation and bioinformatics are contained in Supplementary Methods. tions with sterilized milliQ H2O, mixed with 1% phenol red and injected into embryos at 1–4 cell stage. Associations of DNMT1 and PTGS2 gene expression with AML Whole-mount RNA in situ hybridization and scoring expression phenotypes in primary human tumor data In silico analysis was carried out on the Affymetrix U133 Plus2 gene Digoxigenin (DIG)-labeled RNA probes for lcp1, gata1, runx1, c-myb and 30 dnmt1 were synthesized from cDNA templates according to published expression data GSE144688 including 526 AML patient samples and RNAseq expression data from the The Cancer Genome Atlas (TCGA) LAML literature. Embryos were fixed in 4% paraformaldehyde overnight at 4 °C, data set,31 including 179 patient samples. Complete details of gene permeabilized with 10–50 μg/ml Proteinase K (Roche) for 10–45 min, rinsed associations are contained in Supplementary Methods. in 1X PBS-T and refixed in 4% paraformaldehyde for 20 min. Whole-mount RNA in situ hybridization (WISH) assays were adapted from standard protocol.29 Embryos were scored into ‘normal’, ‘high’ or ‘low’ expression Statistics categories to perform semiquantitative statistical analysis (details in Data for qRT-PCR reported as mean ± s.e.m. and statistics performed using Supplementary Methods). two-tailed student’s t-tests. χ2-goodness of fit was conducted for in situ

© 2015 Macmillan Publishers Limited Leukemia (2015) 2086 – 2097 Epigenetic therapy blocks NHA9-mediated disease AP Deveau et al 2088 gene comparisons. For each experiment the P-values are now included in embryos are a suitable surrogate for tNHA9 embryos when Supplementary Table S2. evaluating the hematopoietic phenotypes in our model.

RESULTS NHA9 expression results in an overproduction of HSCs – – NUP98–HOXA9 mRNA injection phenocopies the transgenic HSCs arise from the aorta gonad mesonephros region around model 30–36 h.p.f. and give rise to all hematopoietic lineages in the adult fi 26 We first wanted to determine if WT zebrafish embryos sh. Given that this population of blood cells is more stable than injected with NUP98–HOXA9 (NHA9) mRNA have a similar those early cells produced in the PBI, and have a greater potential fi phenotype to our transgenic NHA9 line to confirm findings to contribute to leukemia pathogenesis in the adult sh, we and facilitate studies of other cell lineages.21 We injected WT sought to evaluate changes in the HSC population in the context embryos with NHA9 mRNA (60 ng/μl concentration; referred to as of NHA9 expression. ‘rNHA9’) at the one-cell stage. At 28 h.p.f., rNHA9 embryos We performed a combined WISH assay for two zebrafish genes, 22 demonstrated an increase of the myeloid marker, lcp1 (30/62), runx1 and c-myb at 36 h.p.f. (described in North et al. ). This and decrease of the erythroid marker, gata1 (27/62) in the validated double-labeling strategy permits robust identification of posterior blood island (PBI) (Figures 1a–c), resembling the zebrafish HSCs in the dorsal aorta, so that changes to HSC phenotype that we previously showed in our transgenic embryos numbers can be better evaluated. Both tNHA9 (42/60) and rNHA9 (referred to as ‘tNHA9’).21 The smaller proportion of rNHA9 (31/60) embryos demonstrated increased expression of runx1/c- embryos with the hematopoietic phenotype compared with our myb in the region of the dorsal aorta where HSCs are located stable transgenic tNHA9 line was not unexpected. Thus, rNHA9 (Figures 1d and e). To quantify this increase, we crossed our NHA9

Figure 1. NHA9–RNA injected and transgenic NHA9 embryos demonstrate a similar hematopoietic phenotype, including HSC gene overexpression. (a–c) Embryos were injected with NHA9 mRNA (rNHA9) at the 1–2-cell stage and assayed at 24 h.p.f. by WISH for myeloid and erythroid genes, lcp1 and gata1. Bar graph represents the percent of embryos with categorized expression levels. Boxed region contains the PBI. (d, e) WT, transgenic (tNHA9) and rNHA9 embryos were incubated at room temperature at 4 h.p.f. for 24 h then at 28 °C for 24 h to delay embryo development. tNHA9 embryos were heat shocked at 12 h.p.f. Embryos were assayed at 36 h.p.f. by WISH for combined HSC genes, runx1 and c-myb. Bar graph represents percent of embryos with normal or high HSC gene expression. Insets: magnified region of AGM. (f, g) Adult tNHA9 fish were crossed with a cd41::eGFP reporter line to create tNHA9; cd41::eGFP, whereas WT cd41::eGFP embryos were injected at the 1–2-cell stage with NHA9 mRNA. Embryos were dissociated at 48 h.p.f. and quantified using FACS. Images are composition of two brightfield images. Insets: magnified, fluorescent view of the CHT with GFP+ cells. Embryos displayed in side profile, anterior to the left, n values listed in bottom left corner. Experiments were conducted with 15–30 embryos per group × 3 trials. WISH, whole-mount in situ hybridization; PBI, posterior blood island; AGM, aorta–gonad–mesonephros; GFP, green fluorescent protein; CHT, caudal hematopoietic tissue. * denotes statistical significance (Po0.05).

Leukemia (2015) 2086 – 2097 © 2015 Macmillan Publishers Limited Epigenetic therapy blocks NHA9-mediated disease AP Deveau et al 2089 transgenic fish to each of the Tg(cd41::eGFP) and Tg(runx1::eGFP) normal gata1 expression (35/70 and 32/66, respectively; reporter lines, which label HSCs in the dorsal aorta.27,28 After Supplementary Figure S3A and C). However, only a mild reduction creating double homozygous Tg(NHA9;cd41::eGFP) and Tg(NHA9; in combined runx1/c-myb expression was demonstrated in tNHA9 runx1::eGFP) lines, we activated the NHA9 oncogene by out- embryos at 36 h.p.f. (Supplementary Figure S3B and D). Elevated crossing to the Tg(hsp70::Cre) line. This created the triple expression of zebrafish COX enzyme homologs and therapeutic heterozygous lines, Tg(NHA9;cd41::eGFP;hsp70::Cre or NHA9;runx1:: efficacy of small molecule COX inhibitors in our transgenic eGFP;hsp70::Cre). At 48 h.p.f., embryos expressing both tNHA9 and embryos suggests that the PTGS/COX pathway is activated and Tg(cd41::eGFP) presented a 3.4 ± 0.6-fold (Po0.05) increase in required by NHA9 for leukemic transformation. GFP+ cells compared to Tg(cd41::eGFP) controls (Figures 1f and g), which was replicated with injection of NHA9 mRNA (60 ng/μl) into Microarray identifies DNA methylation as a mechanism underlying + the Tg(cd41::eGFP) line (1.8 ± 1.2-fold increase GFP cells). Similarly, NHA9-induced myeloid disease embryos expressing both tNHA9 and Tg(runx1::eGFP) had a fi o + Given that our NHA9 zebra sh model shares gene/pathway 2.8 ± 0.6-fold (P 0.05) increase in GFP cells compared with Tg interactions with NHA9 and Hoxa9;Meis1 mammalian models, as (runx1::eGFP) controls (Supplementary Figure S1A and B). The well as other zebrafish models of AML, we initiated an unbiased increase in GFP-positive HSCs was demonstrated to 96 h.p.f. microarray to identify undiscovered downstream genes or path- (Supplementary Figure S1C). fi ways implicated in NHA9-induced myeloid disease. To optimize These data suggest that NHA9 impacts zebra sh HSC develop- the capture of genes responding directly to NHA9 expression, Cre- ment in addition to early myeloid cells in the PBI, implying that the activated tNHA9 embryos were compared with unactivated tNHA9 pathologic effects of NHA9 may continue to impact long-term embryos outcrossed to WT zebrafish. We found 189 genes blood cell populations, eventually giving rise to the MPN disease upregulated and 85 genes downregulated in embryos expressing fi 21 that we observe in adult NHA9 zebra sh. NHA9 (Figure 3a; Supplementary Table S1). One of the most upregulated genes (4 ± 2-fold) was dnmt1. Arrested differentiation 35 Endogenous cofactor, meis1, cooperates with NHA9 in zebrafish is a defining feature of AML. Interestingly, DNMT1 is part of the epigenetic machinery that represses genes necessary for myeloid myeloid disease 35 Next, we wanted to determine if NHA9 zebrafish shared cell differentiation. Upregulation of DNMT1 has been associated with loss-of-function C/EBPA and RUNX1 mutations in human features with previously studied mammalian systems, specifically, 35 that endogenous meis1 contributed to the hematopoietic AML, but it has not previously been linked to NHA9-induced 32 disease. Using qRT-PCR, we confirmed the upregulation of dnmt1 phenotype. qRT-PCR revealed a modest increase of meis1 (3 ± 1-fold, Po0.05; Figure 3b), and WISH performed at 18, 30 and expression at 28 h.p.f. in tNHA9 embryos compared with WT 48 h.p.f. found increased dnmt1 expression in tNHA9 embryos controls (1.5 ± 0.3-fold, P = 0.05; Supplementary Figure S1D). Using compared to WT controls (Supplementary Figure S4). the following embryo groups: uninjected WT; tNHA9 injected with In addition, we performed methylated DNA immunoprecipita- 1mM meis1 MO; rNHA9 alone; and rNHA9 coinjected with 0.5 mM tion (MeDIP) on tNHA9 embryos and found that dnmt1 over- meis1 MO, we found that loss of meis1 inhibited the myelopro- expression was accompanied by an increase in the extent of liferative phenotype (normal lcp1 expression) in both tNHA9 genomic DNA methylation (Figure 4c). These epigenetic modifica- (39/70 at 30 h.p.f.) and rNHA9 (32/46 at 24 h.p.f.) embryos (Figures tions were found in regions of gene promoters (5000 bp upstream, 2a, b and e) and normalized gata1 expression (Supplementary and 1000 bp downstream of the transcription start sites Figure S2). Similarly, loss of meis1 blocked HSC expansion in both (Figure 3c). More specifically, downregulated genes in tNHA9 tNHA9 and rNHA9 embryos at 36 h.p.f. (71/88 and 37/40, embryos, identified by microarray, demonstrate increased respectively, with normal-to-low runx1/c-myb expression; Figures 33 levels of gene promoter methylation (lGl vs tNHA9 P = 0.0003; 2c and f). Supplementary Figure S5A and B). Furthermore these downregulated genes also have higher CpG coverage around their The PTGS/COX pathway is also necessary for NHA9-induced promoters compared to the average of all the promoters disease (Supplementary Figure S5C and D). In contrast, genes over- fi Our NHA9 transgenic embryos share hematopoietic defects expressed in the microarray data have signi cantly lower CpG similar to the Tg(hsp70::AML1-ETO) zebrafish model.34 AML1-ETO coverage around their promoters compared to the total genome (total vs overexpressed P = 0.0005). Next we identified individual embryos show upregulation of the prostaglandin E2—prostaglan- 20 candidate genes that are downregulated in the microarray and din synthase/cyclo-oxygenase (PGE2—PTGS/COX) signaling axis. Small molecule COX inhibitors may be rational therapeutic options, determined that their promoter regions are hypermethylated in tNHA9 embryos compared to lGl controls (Supplementary Figure as they successfully block the hematopoietic defects in AML1-ETO fi zebrafish20 and also a mouse model of Hoxa9;Meis1-induced AML.24 S6). Lastly, overexpressed genes in tNHA9 embryos, identi ed by These studies prompted us toward a targeted candidate investiga- microarray, demonstrate elevated gene-body methylation com- tion of the PTGS/COX pathway in NHA9 zebrafish. pared with that of lGl controls (Supplementary Figure S5E and F). Using qRT-PCR to compare tNHA9 and WT controls, we observed that zebrafish ptgs1, the COX1 homolog, was expressed Inhibiting DNA binding and methyltransferase activity of dnmt1 near WT levels (1.15 ± 0.07-fold, P = 0.13; Supplementary Figure protein blocks myeloproliferation in tNHA9 embryos S1D). However, tNHA9 embryos showed a large, 64 ± 6-fold Given the association of DNMTs with stem cell development and upregulation (Po0.0001) of ptgs2a, a COX2 isoform and a myeloid leukemogenesis,36,37 we sought to determine the 1.98 ± 0.08-fold increase (Po0.0005) of ptgs2b. We then sought consequence of dnmt1 gene knockdown on hematopoiesis in to inhibit the PTGS/COX pathway through pharmacologic tNHA9 zebrafish. WT and tNHA9 embryos were injected with 20 means. Embryos were treated with 0.3% DMSO vehicle control, 0.75 mM dnmt1 MO and compared to uninjected controls. Knock- 10 μM indomethacin (Indo; a broad-spectrum COX inhibitor) or down of dnmt1 affected hematopoiesis in WT controls at 30 h.p.f. 10 μM NS-398 (a COX2 specific inhibitor). These doses did not with 28/60 embryos demonstrating low lcp1 expression and, 16/38 disrupt hematopoiesis in WT controls (Figures 2d and g and low gata1 expression. tNHA9+dnmt1 MO embryos demonstrated a Supplementary Figure S3). In tNHA9 embryos, Indo and NS-398 rescue of normal hematopoiesis with a relative reduction of lcp1 effectively blocked the lcp1 myeloproliferative phenotype at 30 h. expression (27/60) compared with uninjected tNHA9 controls p.f. (51/60 and 45/61, respectively; Figures 2d and g) and restored (41/60; Figures 4a and e). Similarly, tNHA9+dnmt1 MO embryos

Figure 2. Knockdown of meis1 and the PTGS/COX pathway limit the effects of NHA9 on hematopoiesis. (a)tNHA9 embryos were injected with 1mM meis1 MO, heat shocked at 24 h.p.f. and assayed by WISH for lcp1 expression at 30 h.p.f. (b) WT embryos coinjected with NHA9 mRNA (rNHA9) and 0.5 mM meis1 MO at the 1–2-cell stage were assayed by WISH at 24 h.p.f. for lcp1 expression. (c) Both tNHA9 and rNHA9 embryos at 36 h.p.f. were assayed by WISH for the HSC genes, runx1 and c-myb (tNHA9 embryos were heat shocked at 12 h.p.f.). (d) Embryos were bathed in 0.3% DMSO, 10 μM Indo or 10 μM NS-398 starting at 24 h.p.f. and were heat shocked from 24–25 h.p.f. Bar graphs represent the percent of embryos with categorized expression for (e) lcp1 at 24 and 30 h.p.f. after meis1 MO injections, for (f) runx1/c-myb at 36 h.p.f. after meis1 MO injections and (g) lcp1 after PTGS/COX pathway drug treatments. Boxed region contains the PBI at 24 and 30 h.p.f. Insets: magniﬁed region of the AGM. Embryos displayed in side proﬁle, anterior to the left, n values listed in bottom left corner. Experiments were conducted with 15–30 embryos per group × 3 trials. WISH, whole-mount in situ hybridization; tNHA9, transgenic NUP98–HOXA9 embryos; MO, morpholino oligonucleotide; Indo, indomethacin; PBI, posterior blood island; AGM, aorto-gonad-mesonephros.

also showed rescue of near-normal gata1 levels (22/38) compared To further test whether zebrafish dnmt1 was directly involved in with uninjected tNHA9 controls (31/41; Supplementary Figure S7). NHA9 activity, we interrogated its cofactor, uhrf1, which encodes a Knockdown of dnmt1 had no effect on combined runx1/c-myb conserved DNA-binding protein. UHRF1 protein helps recruit and expression in WT controls at 36 h.p.f., but did restore WT levels of anchor DNMT1 protein to DNA for methylation of target runx1/c-myb in tNHA9 embryos (45/66; Figures 4b and f). These sequences.38 In human and mouse cell lines, as well as zebrafish findings suggest that blocking dnmt1 expression can block the embryos, UHRF1 knockdown leads to genome hypomethylation, effects of NHA9 on early myeloid and HSC development. which cannot be rescued by DNMT1 overexpression due to

Figure 3. Microarray analysis identifies overexpression of dnmt1. Global RNA expression was compared in tNHA9 embryos versus ‘lGl’ (NUP98- HOXA9 × WT) control embryos. Embryos were activated by heat shock at 24 h.p.f. and RNA was extracted at 28 h.p.f. (a) Microarray identified 85 downregulated genes (red) and 189 upregulated genes (yellow), including upregulation of dnmt1.(b) qRT-PCR confirmed a 3 ± 1-fold increase in dnmt1 expression in the tNHA9 embryos (black bars, n = 3) compared to WT controls (white bars, n = 5). Expression of indicated genes was normalized against the housekeeping gene, elongation factor 1a (ef1a). (c) MeDIP-sequencing and average DNA methylation signal is illustrated in promoter regions for all genes (defined as 5000 bp upstream and 1000 bp downstream from transcription start site) between lGl controls and tNHA9, calculated in 20 bp bins along the 6000 bp region (Unactivated tNHA9 (lGl)vstNHA9, P = 4.5200 × 10 − 108). tNHA9, transgenic NUP98-HOXA9 embryos; bp, base pairs; TSS, transcription start site. Student’s t-test results: *Po0.05; ***Po0.005.

38–41 impaired DNA binding. Therefore, we injected WT and tNHA9 36 h.p.f. (Figures 4d and f). Interestingly, following 5 h, 75 μM DAC embryos with 0.5 mM uhrf1 MO. Similar to knockdown of dnmt1, treatment reduced the MeDIP-seq signal around the transcription loss of uhrf1 modestly reduced both lcp1 (39/71) and gata1 (19/40) start sites in tNHA9 embryos to below WT levels (Figure 4g; expression at 30 h.p.f. in WT controls. Yet importantly, tNHA9 lGl vs tNHA9 DAC P = 9.1201 × 10 − 12). However, following +uhrf1 MO embryos showed a relative loss of lcp1 expression extended incubation over 12 h (overnight, labeled as o/n), gene (32/71) and a rescue of gata1 (21/40) compared with uninjected promoter methylation was restored to WT methylation levels tNHA9 controls (52/67 high lcp1, 47/60 low gata1; Supplementary (Supplementary Figure S9; lGl vs tNHA9 DAC o/n P = 0.6860). Both Figure S7). Thus, knockdown of either uhrf1 or dnmt1 in tNHA9 the DAC treatments (5 h and o/n) lead to promoter demethylation embryos produces a similar rescue of hematopoiesis, in keeping of downregulated genes identified in the microarray (Supple- with their interaction as DNA binding partners. mentary Figure S5A and B) and a decrease in gene body fi Next, we wanted to ensure that the activity of NHA9 speci cally methylation of overexpressed genes identified in the microarray fi required the enzymatic function of zebra sh dnmt1 protein. We (Supplementary Figure S5E and F). tested whether broad-spectrum small molecule DNMT inhibitors, fi 42 Altogether, these ndings suggest that both the binding of decitabine (DAC) and zebularine (Zeb), could inhibit NHA9 dnmt1 protein to DNA (via uhrf1 protein) and its catalytic function signaling and restore normal hematopoiesis. Following dose are required for NHA9 to induce its hematopoietic phenotypes. optimization (Supplementary Figure S8A and B), embryos were treated with 0.3% DMSO vehicle control, 75 μM DAC or 250 μM Zeb. These doses did not disrupt hematopoiesis in WT controls (Figures Combined epigenetic therapy demonstrates synergy in the rescue fi 4d and f and Supplementary Figure S8C–F). In tNHA9 embryos, of normal hematopoiesis in NHA9 zebra sh, as does HDAC plus DAC and Zeb effectively blocked the lcp1 myeloproliferative COX inhibition phenotype at 30 h.p.f. (39/60 and 32/60, respectively; Figures 4c Epigenetic regulation is mediated through chemical modifications and e) and rescued gata1 expression (41/60 and 32/60, to both DNA and its packaging proteins, which together constitute respectively; Supplementary Figure S8E and F). Furthermore, eukaryotic chromatin. In specific genomic contexts such a gene DAC restored WT levels of runx1/c-myb expression (81/103) at promoters, DNA methylation by DNMTs is an initial step that

Figure 4. dnmt1 has a critical role in NHA9-induced myeloid disease. Embryos were assayed by WISH for lcp1 expression at 30 h.p.f. (heat shock at 24 h.p.f.) and for the combined expression of runx1 and c-myb at 36 h.p.f. (heat-shock at 12 h.p.f.). (a, b) Embryos were injected at the 1–2-cell stage with dnmt1 MO and compared with uninjected embryos. (c, d) Embryos were bathed in either 0.3% DMSO or 75 μM DAC. Bar graphs represent the percent of embryos with categorized expression for (e) lcp1 at 30 h.p.f., and (f) runx1/c-myb at 36 h.p.f. (g) MeDIP sequencing and average DNA methylation signal is illustrated in promoter regions for all genes between lGl controls, tNHA9 and tNHA9-DAC treated calculated in 20 bp bins along the 6000 bp region (Unactivated tNHA9 (lGl)vstNHA9-DAC, P = 9.2100 × 10 − 12). Boxed region contains the PBI at 30 h.p.f. Insets: magniﬁed region of the AGM. Embryos displayed in side proﬁle, anterior to the left, n values listed in bottom left corner. Experiments were conducted with 15–30 embryos per group × 3 trials. WISH, whole-mount in situ hybridization; MO, morpholino oligonucelotide; DAC, decitabine; PBI, posterior blood island; AGM, aorta-gonad-mesonephros.

recruits other modifying enzymes, such as HDAC.43 Preclinical Figures 5a and g). However, both VPA and TSA only weakly in vitro and in vivo studies, as well as clinical data have rescued WT gata1 expression (12/51 and 18/60 low, respectively; demonstrated a therapeutic effect of HDAC inhibitors, such as Supplementary Figure S10 and S11). VPA also did not reduce VPA and TSA, in the treatment of myeloid diseases including runx1/c-myb expression in tNHA9 embryos compared with DMSO AML.44–46 Therefore, we sought to determine if we could block controls (41/68 vs 57/72 high expression, respectively; Figures 5b NHA9 by targeting HDACs as a component of the epigenetic and h). These ﬁndings suggest that HDAC inhibitors may not be as machinery that regulates gene transcription. Embryos were effective as DNMT inhibitors at blocking the full impact of NHA9 treated with 0.3% DMSO vehicle control, 250 μM VPA, or 600 nM activity, especially within the HSC compartment. TSA. VPA modestly reduced lcp1 expression in WT controls (15/60 Recent studies propose that combining HDAC and DNMT low). A relative reduction of lcp1 expression was seen in tNHA9 inhibitors may work better than either compound alone to restore +VPA embryos (29/60 normal) and tNHA9+TSA embryos (40/69 normal epigenetic regulation in patients with myeloid disease.47,48 normal) compared with tNHA9+DMSO controls (46/60 high; Thus, we sought to see if targeting both of these epigenetic

Figure 5. DAC and Indo function synergistically with VPA to restore hematopoiesis in NHA9 transgenic zebrafish. tNHA9 embryos were assayed by WISH for lcp1 expression at 30 h.p.f. (heat-shock at 24 h.p.f.) and for the combined expression of runx1 and c-myb at 36 h.p.f. (heat-shock at 12 h.p.f.) compared with WT embryos. (a, b) Embryos were bathed in either 0.3% DMSO or 250 μM VPA. (c, d) Embryos were bathed in combined treatment of 5 μM DAC and 25 μM VPA. (e, f) Embryos were bathed in combined treatment of 10 μM Indo and 75 μM VPA. Bar graphs represent the percent of embryos with categorized expression for (g) lcp1 at 30 h.p.f., and (h) runx1/c-myb at 36 h.p.f. (i) MeDIP sequencing and average DNA methylation signal is illustrated in promoter regions for all genes between lGl controls, tNHA9 and tNHA9-DAC+VPA treated calculated in 20 bp bins along the 6000 bp region (Unactivated tNHA9 (lGl)vstNHA9-DAC+VPA, P = 0.1130). (j) Human gene expression data sets of primary AML tumors determine human DNMT1 and PTGS2 overexpression correlates with high-risk features of leukemia. Boxed region contains the PBI at 30 h.p.f. Insets: magnified region of the AGM. Embryos displayed in side profile, anterior to the left, n values listed in bottom left corner. Experiments were conducted with 15–30 embryos per group × 3 trials. tNHA9 = transgenic NUP98–HOXA9 embryos; WISH, whole-mount in situ hybridization; PBI, posterior blood island; AGM, aorta-gonad-mesonephros. mechanisms simultaneously would yield a therapeutic response. embryos (Figure 5i). Moreover, promoter regions of the genes Indeed, when we combined as little as 10 μM DAC and 25 μM VPA down-regulated in the microarray were demethylated to lGl (13 and 10% of each monotherapy dose), we restored WT levels of control levels (Supplementary Figure S5A and B), whereas the lcp1 (39/60) at 30 h.p.f., and of runx1/c-myb (46/58) at 36 h.p.f. gene bodies of the genes overexpressed in the microarray also (Figures 5c, d, g and h), in tNHA9 embryos. A similar hematopoietic demonstrate control methylation levels (Supplementary Figure rescue was achieved with 25 μM DAC plus 125 nM TSA (10 and 21% S5E and F). Interestingly however, methylation levels were again of each monotherapy dose; 46/68 normal lcp1) (Supplementary higher following overnight incubation similar to DAC treatment Figure S10). alone overnight. (Supplementary Figure S11). Using MeDIP-seq, we found that a 5 h treatment of tNHA9 Given that there are potential risks to long-term therapy with embryos with DAC+VPA restored DNA methylation around gene HDAC inhibitors49 (discussed below), we examined whether a COX transcription start sites to those of DMSO-treated-lGl control inhibitor (2.5 μM Indo) combined with an HDAC inhibitor (75 μM

© 2015 Macmillan Publishers Limited Leukemia (2015) 2086 – 2097 Epigenetic therapy blocks NHA9-mediated disease AP Deveau et al 2094 VPA; 30 and 25% of each monotherapy dose, respectively) would can be replicated by NHA9 mRNA in WT embryos. The rescue of also block NHA9 activity. In tNHA9 embryos, this regimen restored normal hematopoiesis in NHA9 embryos following meis1 knock- wild-type lcp1 at 30 h.p.f. (37/60; Figures 5e and g) and also WT down seemed inconsistent with the modest increase of meis1 runx1/c-myb expression (34/68) at 36 h.p.f. (Figures 5f and h). expression by qRT-PCR. However, studies in NHA9-transformed human cell lines suggest that upregulation of MEIS1 may be a late- DNMT1 and PTGS2 cluster with high-risk features in human AML transformation event that coincides with increased cellular proliferation.50 Our previous21 and current work suggests that The power of the zebrafish preclinical model is its ability to rapidly fi altered differentiation is the predominant mechanism in early identify gene collaborators and drug-modi able phenotypes that fi fi blood cells from the PBI in tNHA9 zebra sh embryos, and that can be further investigated in high- delity mammalian models. hyperproliferation may not be a prominent feature until HSCs Thus, having identified that dnmt1 and ptgs/cox genes are fi develop, or possibly occur later in adulthood closer to the onset of important downstream of NHA9 in zebra sh, we interrogated MPN disease. Nevertheless, our findings suggest that meis1 human gene expression data sets of primary AML tumors to assess cooperates with NHA9 to dysregulate zebrafish hematopoiesis, whether the expression of human DNMT1 and PTGS2 correlates since the loss of meis1 can block the effects of NHA9. Taken with high-risk features of myeloid disease. The TCGA LAML fi 30 together, these data align our NHA9 zebra sh model to (RNASeq, 179 AML tumors) and Wouters et al. (Affymetrix, 526 mammalian models of NHA9- and HOXA9;MEIS1-induced myeloid AML tumors GSE14468) data sets both demonstrate that DNMT1 disease.12,51 and PTGS2 are indeed upregulated in human AML (Figure 5j). We subsequently embarked on two complementary approaches PTGS2 possesses a larger dynamic range of expression compared to discover contributing genes and pathways in NHA9 pathogen- to DNMT1, raising the possibility that a certain subpopulation of esis. First, a targeted candidate approach demonstrated upregula- fi AML tumors has high expression of PTGS2 as a uniquely de ning tion of a COX isoform gene, ptgs2a. The PTGS/COX pathway feature. We also found that AML tumors with high expression of appears to be required for the NHA9-induced phenotype, as either DNMT1 or PTGS2 share phenotypic and prognostic traits pharmacologic inhibition of COX enzymes abrogates the hema- with high-risk features of myeloid disease such as complex topoietic defects in tNHA9 embryos. The AML1-ETO zebrafish karyotype and TP53 mutations, but interestingly also with idt model also demonstrated this requirement for PTGS/COX (chromosome 16) anomalies commonly considered a prognos- signaling,20 which was subsequently linked to the activity of the tically lower risk group (Table 1). Wnt/β-catenin self-renewal pathway in blood cells.52 This associa- An important limitation of this in silico analysis is that these data tion between PTGS/COX and Wnt/β-catenin pathways was also sets do not annotate the NHA9 fusion oncogene, thus we could shown in a mouse model of Hoxa9;Meis1-induced AML, which not directly assess the correlation of DNMT1 or PTGS2 expression served as the basis for targeted therapy with COX inhibitors. The in tumors that harbor this mutation. However, our findings from Wnt/β-catenin pathway is active during early hematopoiesis to these data sets suggest that DNMT1 and PTGS2 may cluster in regulate HSC proliferation, self-renewal, and survival,23,50 and its high-risk myeloid disease and other subsets of AML, which reactivation occurs during AML pathogenesis to confer stem-like substantiates the findings from our genetic and pharmacological properties on differentiated cells.24 We have preliminary data to experiments performed in zebrafish. suggest that this pathway may be active in tNHA9 embryos, evidenced by a 1.7 ± 0.3-fold upregulation (Po0.005) of zebrafish cyclin D1 (ccnd1), a canonical target of Wnt/β-catenin signaling DISCUSSION (Supplementary Figure S1D). Our previous and current study demonstrates that NHA9 expres- Second, using an unbiased microarray approach, we identified sion in transgenic zebrafish alters the cell lineage assignment overexpression of a DNA methyltransferase gene, dnmt1, which during hematopoiesis. Namely, NHA9 disrupts the early myeloid– for the first time links NHA9 to an epigenetic regulator. In addition, erythroid balance, with an increase in myeloid precursors and a we determined that this NHA9-induced increase in dnmt1 concurrent decrease in erythroid cells.21 We further show that expression was associated with increased DNA methylation NHA9 increases the number of HSCs, endogenous meis1 is around transcription start sites. Furthermore we closely examined required to induce these defects, and that these phenotypes CpG coverage of up- and downregulated genes identified by

Table 1. Phenotypic associations for DNMT1 and PTGS2 overexpression in TCGA and Wouters et al. data sets based on two-tailed Fisher tests (cut-off P-value 0.05)

Category Phenotype/DNMT1 high TCGA Wouters et al.30

FAB — None None Classiﬁcation Complex Karyotype Positive correlation Positive correlation IDT16 Not signiﬁcant Positive correlation Risk High-risk Positive correlation N/A NPM1 WT Positive correlation Positive correlation TP53 mutation Positive correlation N/A Category Phenotype/PTGS2 high TCGA Wouters et al.30

FAB M4 Positive correlation Positive correlation Classiﬁcation Complex Karyotype Positive correlation Negative correlation IDT16 Positive correlation Positive correlation Risk High-risk Positive correlation — NPM1 WT Positive correlation — TP53 mutation Positive correlation NA Abbreviations: FAB, French-American-British; NA, not applicable; TCGA, The Cancer Genome Atlas; WT, wild type. Positive and negative associations are assigned for odds-ratios 41ando1, respectively.

Leukemia (2015) 2086 – 2097 © 2015 Macmillan Publishers Limited Epigenetic therapy blocks NHA9-mediated disease AP Deveau et al 2095 microarray and found less coverage in upregulated genes inhibitors exhibit synergy to block activity of the NHA9 fusion compared to the total genome or to downregulated genes. This oncogene, representing a potentially effective strategy for larger coverage may correspond to higher levels of methylation minimizing the risk of clonal resistance in high risk AML. around the promoter region of the down-regulated genes due to Human AML has a discouraging overall survival rate of less than the dmnt1 overexpression.53 Indeed, a series of candidate genes 60%, though discoveries of targeted therapies, such as imatinib down-regulated in the microarray demonstrated increased mesylate and all-trans retinoic acid, have great success rates methylation around their promoter regions. To our knowledge, against subtypes of myeloid disease with much lower toxicity to this is the first time gene array data and MeDIP have been used in the patient.61 However, targeted therapy for NHA9-induced AML parallel in zebrafish to identify potential genes that may benefit has remained elusive and overall success quite low. Targeted from targeted demethylating agents. Furthermore, we examined therapeutic strategies need to be directed towards actionable the level of methylation within the gene bodies of upregulated genetic aberrations. Our NHA9 zebrafish model is a novel tool for fi genes, identi ed by microarray. Recent work has revealed that the identification of collaborative genes and pathways, and thus increased levels of gene-body methylation can led to gene provides a strategic approach to discover targeted therapies. Here overexpression and may be targeted using demethylating agents fi 54 we have identi ed a number of genes and pathways that may such as DAC. We determined that gene bodies of the have a critical role in NHA9-induced disease, as well as possible upregulated genes were hypermethylated. Therefore, demethylat- modes of inhibition, providing compelling preclinical data to ing agents such as DAC may be working to reduce methylation pursue rational epigenetic-based interventions in AML patients levels in both the hypermethylated promoter regions of down- that harbor the NHA9 oncogene or that overexpress HOXA9. regulated genes and hypermethylated gene bodies in upregulated genes in our tNHA9 model. Ultimately, our studies provide a rationale to personalize We determined that the hematopoietic defects in NHA9 treatment that will provide better clinical outcomes with less zebrafish could be reversed by blocking dnmt1 protein function toxicity for this high-risk leukemia population. via direct knockdown and knockdown of its co-factor uhrf1.We could also block both the NHA9-induced hematopoietic and DNA CONFLICT OF INTEREST methylation defects by enzymatic inhibition using DAC or Zeb. fl These findings strongly suggest that the DNA methylation activity The authors declare no con ict of interest. of the dnmt1 enzyme helps drive NHA9-induced myeloid disease. Unfortunately, DNMT inhibition as a monotherapy presents ACKNOWLEDGEMENTS multiple issues. Clinically, DAC alone is used to treat MDS,55 but in We thank Angela Young, Jessica Hill and Emma Cummings for zebrafish care and overt AML, it may be more effective when combined with 56 maintenance; Jocelyn Jaques for administrative support. This work is supported by a cytotoxic chemotherapy. In addition, DAC poses carcinogenic Canadian Institutes of Health Research /Nova Scotia Health Research Foundation risks in a dose-dependent manner, as genome-wide DNA Regional Partnership Program Grant MED-Matching 2011–7509 (CIHR#.243778). APD is hypomethylation can lead to genomic instability, chromosome funded by The Cancer Research Training Program, supported by The Terry Fox Strategic 57 rearrangements and secondary tumors. In tNHA9 embryos, we Health Research Training Program in Cancer Research by the Beatrice Hunter Cancer observed toxicity at DAC doses greater than 75 μM, increased lcp1 Research Institute. AMF is funded by a Canadian Institutes of Health Research Banting and runx1/c-myb expression and decreased gata1 (data not and Best Graduate Student Award. The MeDIP work is supported by the MeDIP-seq shown). Following a 5 h treatment at 75 μM, we also observed a Program Project Grant funded by Terry Fox Foundation (TFF-122869) to MH. JNB is relative hypomethylation compared with lGl controls. These supported by a Cancer Care Nova Scotia Peggy Davison Clinician Scientist Award. consequences caution against the indiscriminate use of DNMT inhibitors and assert the value of incorporating lower doses of AUTHOR CONTRIBUTIONS these drugs as part of combination therapy. Clinical trials using DAC and VPA in combination have produced APD and AMF conceived and conducted experiments, analyzed the data and mixed results regarding response to treatment and toxicity.58,59 wrote the paper. AJC and GSW conducted experiments and analyzed the data. However, these studies applied these drug combinations to an CG generated the original NUP98-HOXA9 transgenic zebrafish line. ICC and DL unselected population of AML patients, which may dilute a performed microarray experiments and analyzed microarray data. MM targeted treatment response. This therapy may be more appro- performed MeDIP studies and analyzed methylation data. GA conducted priately delivered to a subgroup of patients with evidence of NHA9 human AML gene data set analysis. VR generated associated microarray figures. expression, a particularly high-risk group that responds quite RL performed cytospins and cell morphological analysis. MH oversaw MeDIP 6,51 poorly to current cytotoxic therapies. Our zebrafish data studies and analysis and edited the manuscript. SML oversaw microarray suggest that NHA9-induced myeloid disease is susceptible to studies and analysis and edited the manuscript. KS oversaw human AML gene combination therapy with DNMT and HDAC inhibitors, at data set analysis. ATL oversaw development of original NUP98–HOXA9- significantly lower doses than either agent as a monotherapy transgenic line and edited the manuscript. JNB conceived experiments, and more closely approximated WT methylation levels after 5 h oversaw all the zebrafish studies, wrote and edited the manuscript. exposure. Furthermore, our data suggest that this combination regimen is effective in the stem cell compartment.60 Finally, it may be advisable to include Wnt/β-catenin blockade REFERENCES into therapeutic strategies for NHA9-induced disease. HDAC 1 Burnett A, Wetzler M, Löwenberg B. Therapeutic advances in acute myeloid 49 inhibitors are known to activate this self-renewal pathway, and leukemia. J Clin Oncol 2011; 29: 487–494. there is a known association of this pathway with PTGS/COX 2 Redaelli A, Stephens JM, Brandt S, Botteman MF, Pashos CL. Short- and long-term signaling, which is targeted by NHA9. Our early data also suggest effects of acute myeloid leukemia on patient health-related quality of life. Cancer that a downstream target of Wnt/β-catenin may be upregulated in Treat Rev 2004; 30:103–117. NHA9 embryos. We hypothesize that leukemia cells harboring 3 Borthakur G, Estey EE. Therapy of acute myelogenous leukemia in adults. NHA9 might become resistant to epigenetic therapy through Wnt/ Nagarajan L. Acute Myelogenous Leukemia: Genetics, Biology and Therapy. Springer β New York: New York City, NY, USA, 2010 pp 257–270. -catenin signaling, eventually leading to disease relapse. Our 4 Gilliland DG, Tallman MS. Focus on acute leukemias. Cancer Cell 2002; 1:417–420. data show that the combination of HDAC plus COX inhibitors also 5 Argiropoulos B, Humphries RK. Hox genes in hematopoiesis and leukemogenesis. blocks the hematopoietic defects induced by NHA9. These striking Oncogene 2007; 26: 6766–6776. findings suggest that targeting both epigenetic and PTGS/COX 6 Borrow J, Shearman AA, Stanton VP, Becher R, Collins T, Williams AJ et al. The t pathways, using sub-monotherapeutic doses of small molecule (7;11)(p15;p15) translocation in acute myeloid leukmaemia fuses the genes for

© 2015 Macmillan Publishers Limited Leukemia (2015) 2086 – 2097 Epigenetic therapy blocks NHA9-mediated disease AP Deveau et al 2096 nucleoporin NUP98 and class 1 homeoprotein HOXA9. Nat Genet 1996; 12: 31 Network TCGAR. Genomic and Epigenomic Landscapes of Adult De Novo Acute 159–167. Myeloid Leukemia The Cancer Genome Atlas Research Network. N Engl J Med 7 Lawrence HJ, Rozenfeld S, Cruz C, Matsukuma K, Kwong A, Kömüves L et al. 2013; 368:2059–2074. Frequent co-expression of the HOXA9 and MEIS1 homeobox genes in human 32 Thorsteinsdottir U, Kroon E, Jerome L, Blasi F, Sauvageau G. Defining Roles for myeloid leukemias. Leukemia 1999; 13:1993–1999. HOX and MEIS1 Genes in Induction of Acute Myeloid Leukemia. Mol Cell Biol 2001; 8 Golub TR, Slonim DK, Tamayo P, Huard C, Gaasenbeek M, Mesirov JP. Molecular 21:224–234. classification of cancer: class discovery and class prediction by gene expression 33 Cvejic A, Serbanovic-Canic J, Stemple DL, Ouwehand WH. The role of meis1 in monitoring. Science 1999; 286:531–537. primitive and definitive hematopoiesis during zebrafish development. Haemato- 9 Giles FJ, Keating A, Goldstone AH, Avivi I, Willman CL, Kantarjian HM. Acute logica 2011; 96:190–198. myeloid leukemia. Hematology Am Soc Hematol Educ Program 2002: 73–110. 34 Yeh J-RJ, Munson KM, Chao YL, Peterson QP, Macrae CA, Peterson RT. AML1-ETO 10 Calvo KR, Knoepfler PS, Sykes DB, Pasillas MP, Kamps MP. Meis1a suppresses reprograms hematopoietic cell fate by downregulating scl expression. Develop- differentiation by G-CSF and promotes proliferation by SCF: potential mechan- ment 2008; 135: 401–410. isms of cooperativity with Hoxa9 in myeloid leukemia. Proc Natl Acad Sci USA 35 Saunthararajah Y, Triozzi P, Rini B, Singh A, Radivoyevitch T, Sekeres M et al. P53- 2001; 98: 13120–13125. independent, normal stem cell sparing epigenetic-differentation therapy for 11 Faber J, Krivtsov AV, Stubbs MC, Wright R, Davis TN, van den Heuvel-Eibrink M myeloid and other malignancies. Semin Oncol 2012; 39:97–108. et al. HOXA9 is required for survival in human MLL-rearranged acute leukemias. 36 Thol F, Damm F, Lüdeking A, Winschel C, Wagner K, Morgan M et al. Incidence Blood 2009; 113: 2375–2385. and prognostic influence of DNMT3A mutations in acute myeloid leukemia. J Clin 12 Ohno Y, Yasunaga S, Janmohamed S, Ohtsubo M, Saeki K, Kurogi T et al. Hoxa9 Oncol 2011; 29: 2889–2896. transduction induces hematopoietic stem and progenitor cell activity through 37 Trowbridge JJ, Sinha AU, Zhu N, Li M, Armstrong SA, Orkin SH. Haploinsufficiency direct down-regulation of geminin protein. PLoS One 2013; 8: e53161. of Dnmt1 impairs leukemia stem cell function through derepression of bivalent 13 Calvo KR, Sykes DB, Pasillas MP, Kamps MP. Nup98-HoxA9 immortalizes myeloid chromatin domains. Genes Dev 2012; 26: 344–349. progenitors, enforces expression of Hoxa9, Hoxa7 and Meis1, and alters cytokine- 38 Tittle RK, Sze R, Ng A, Nuckels RJ, Swartz ME, Anderson RM et al. Uhrf1 and Dnmt1 specific responses in a manner similar to that induced by retroviral co-expression are required for development and maintenance of the zebrafish lens. Dev Biol of Hoxa9 and Meis1. Oncogene 2002; 21: 4247–4256. 2011; 350:50–63. 14 Hatano Y, Miura I, Nakamura T, Yamazaki Y, Takahashi N, Miura AB. Molecular 39 Chu J, Loughlin Ea, Gaur Na, SenBanerjee S, Jacob V, Monson C et al. UHRF1 heterogeneity of the NUP98/HOXA9 fusion transcript in myelodysplastic syn- phosphorylation by cyclin A2/cyclin-dependent kinase 2 is required for zebrafish dromes associated with t(7;11)(p15;p15). Br J Haematol 1999; 107:600–604. embryogenesis. Mol Biol Cell 2011; 23:59–70. 15 Kroon E, Krosl J, Thorsteinsdottir U, Baban S, Buchberg AM, Sauvageau G. Hoxa9 40 Arita K, Ariyoshi M, Tochio H, Nakamura Y, Shirakawa M. Recognition of hemi- transforms primary bone marrow cells through specific collaboration with Meis1a methylated DNA by the SRA protein UHRF1 by a base-flipping mechanism.pdf. but not Pbx1b. EMBO J 1998; 17: 3714–3725. Nature 2008; 455:818–821. 16 Kroon E, Thorsteinsdottir U, Mayotte N, Nakamura T, Sauvageau G. NUP98-HOXA9 41 Bostick M, Kim JK, Estève PO, Clark A, Pradhan S, Jacobsen SE. UHRF1 plays a role expression in hemopoietic stem cells induces chronic and acute myeloid leuke- in maintaining DNA methylation in mammalian cells. Science 2007; 317: mias in mice. EMBO J 2001; 20: 350–361. 1760–1764. 17 Langenau DM, Traver D, Ferrando AA, Kutok JL, Aster JC, Kanki JP et al. 42 Ceccaldi A, Rajavelu A, Champion C, Rampon C, Jurkowska R, Jankevicius G et al. Myc-induced T cell leukemia in transgenic zebrafish. Science 2003; 299:887–890. C5-DNA Methyltransferase Inhibitors: From Screening to Effects on Zebrafish 18 Sabaawy HE, Azuma M, Embree LJ, Tsai H-J, Starost MF, Hickstein DD. TEL-AML1 Embryo Development. ChemBioChem 2011; 12: 1337–1345. transgenic zebrafish model of precursor B cell acute lymphoblastic leukemia. Proc 43 Guidotti A, Dong E, Kundakovic M, Satta R, Grayson DR, Costa E. Characterization Natl Acad Sci USA 2006; 103: 15166–15171. of the action of antipsychotic subtypes on valproate-induced chromatin remo- 19 Streisinger G, Walker C, Dower N, Knauber D, Singer F. Production of clones deling. Trends Pharmacol Sci 2009; 30:55–60. of homozygous diploid zebra fish (Brachydanio rerio). Nature 1981; 291: 44 Bellos F, Mahlknecht U. Valproic acid and all-trans retinoic acid: meta-analysis of a 293–296. palliative treatment regimen in AML and MDS patients. Onkologie 2008; 31: 20 Yeh J-RJ, Munson KM, Elagib KE, Goldfarb AN, Sweetser DA, Peterson RT. 629–633. Discovering chemical modifiers of oncogene-regulated hematopoietic differ- 45 Göttlicher M, Minucci S, Zhu P, Krämer OH, Schimpf A, Giavara S et al. Valproic entiation. Nat Chem Biol 2009; 5: 236–243. acid defines a novel class of HDAC inhibitors inducing differentiation of 21 Forrester AM, Grabher C, McBride ER, Boyd ER, Vigerstad MH, Edgar A et al. transformed cells. EMBO J 2001; 20: 6969–6978. NUP98-HOXA9-transgenic zebrafish develop a myeloproliferative neoplasm and 46 Marks P, Rifkind Ra, Richon VM, Breslow R, Miller T, Kelly WK. Histone deacetylases provide new insight into mechanisms of myeloid leukaemogenesis. Br J Haematol and cancer: causes and therapies. Nat Rev Cancer 2001; 1:194–202. 2011; 155: 167–181. 47 Fuks F, Burgers WA, Brehm A, Hughes-Davies L, Kouzarides T. DNA methyl- 22 North TE, Goessling W, Walkley CR, Lengerke C, Kopani KR, Lord AM et al. transferase Dnmt1 associates with histone deacetylase activity. Nat Genet 2000; 24: Prostaglandin E2 regulates vertebrate haematopoietic stem cell homeostasis. 88–91. Nature 2007; 447: 1007–1011. 48 Savickiene J, Treigyte G, Borutinskaite V-V, Navakauskiene R. Antileukemic activity 23 Goessling W, Allen RS, Guan X, Jin P, Uchida N, Dovey M et al. Prostaglandin E2 of combined epigenetic agents, DNMT inhibitors zebularine and RG108 with enhances human cord blood stem cell xenotransplants and shows long-term HDAC inhibitors, against promyelocytic leukemia HL-60 cells. Cell Mol Biol Lett safety in preclinical nonhuman primate transplant models. Cell Stem Cell 2011; 8: 2012; 17: 501–525. 445–458. 49 Wiltse J. Mode of Action: Inhibition of Histone Deacetylase, Altering WNT- 24 Wang Y, Krivtsov AV, Sinha AU, North TE, Goessling W, Feng Z et al. The Wnt/beta- Dependent Gene Expression, and Regulation of Beta-Catenin—Developmental catenin pathway is required for the development of leukemia stem cells in AML. Effects of Valproic Acid. Crit Rev Toxicol 2005; 35:727–738. Science 2010; 327: 1650–1653. 50 Lord AM, North TE, Zon LI. Prostaglandin E2: Making more of your bone marrow. 25 Westerfield M. The zebrafish book. A Guide For The Laboratory Use of Zebrafish Cell Cycle 2007; 6: 3054–3057. (Danio rerio). 4th edn University of Oregon Press: Eugene, Oregon, 2000. 51 Iwasaki M, Kuwata T, Yamazaki Y, Jenkins Na, Copeland NG, Osato M et al. 26 Kissa K, Herbomel P. Blood stem cells emerge from aortic endothelium by a novel Identification of cooperative genes for NUP98-HOXA9 in myeloid leukemogenesis type of cell transition. Nature 2010; 464:112–115. using a mouse model. Blood 2005; 105:784–793. 27 Lin H-F, Traver D, Zhu H, Dooley K, Paw BH, Zon LI et al. Analysis of thrombocyte 52 Goessling W, North TE, Loewer S, Lord AM, Lee S, Stoick-Cooper CL. Genetic development in CD41-GFP transgenic zebrafish. Blood 2005; 106: 3803–3810. interaction of PGE2 and Wnt signaling regulates developmental specification of 28 Lam EYN, Chau JYM, Kalev-Zylinska ML, Fountaine TM, Mead RS, Hall CJ et al. stem cells and regeneration. Cell 2009; 136: 1136–1147. Zebrafish runx1 promoter-EGFP transgenics mark discrete sites of definitive blood 53 Arand J, Spieler D, Karius T, Branco MR, Meilinger D, Meissner A et al. In vivo progenitors. Blood 2009; 113: 1241–1249. Control of CpG and Non-CpG DNA Methylation by DNA Methyltransferases. PLoS 29 Bennett CM, Kanki JP, Rhodes J, Liu TX, Paw BH, Kieran MW et al. Myelopoiesis in Genet 2012; 8: e1002750. the zebrafish, Danio rerio. Blood 2001; 98:643–651. 54 Yang X, Han H, De Carvalho DD, Lay FD, Jones PA, Liang G. Gene body methy- 30 Wouters BJ, Löwenberg B, Erpelinck-Verschueren CaJ, Van Putten WLJ, Valk PJM, lation can alter gene expression and is a therapeutic target in cancer. Cancer Cell Delwel R. Double CEBPA mutations, but not single CEBPA mutations, define a 2014; 26: 577–590. subgroup of acute myeloid leukemia with a distinctive gene expression profile 55 Kantarjian H, Issa J-PJ, Rosenfeld CS, Bennett JM, Albitar M, DiPersio J et al. that is uniquely associated with a favorable outcome. Blood 2009; 113: Decitabine improves patient outcomes in myelodysplastic syndromes: results of a 3088–3091. phase III randomized study. Cancer 2006; 106:1794–1803.

Leukemia (2015) 2086 – 2097 © 2015 Macmillan Publishers Limited Epigenetic therapy blocks NHA9-mediated disease AP Deveau et al 2097 56 Saba HI. Decitabine in the treatment of myelodysplastic syndromes. Ther Clin Risk 59 Detich N, Bovenzi V, Szyf M. Valproate induces replication-independent active Manag 2007; 3:807–817. DNA demethylation. J Biol Chem 2003; 278: 27586–27592. 57 Maslov a Y, Lee M, Gundry M, Gravina S, Strogonova N, Tazearslan C et al. 5-aza-2’- 60 Misaghian N, Ligresti G, Steelman LS, Bertrand FE, Bäsecke J, Libra M et al. Tar- deoxycytidine-induced genome rearrangements are mediated by DNMT1. geting the leukemic stem cell: the Holy Grail of leukemia therapy. Leukemia 2009; Oncogene 2012; 31: 5172–5179. 23:25–42. 58 Blum W, Klisovic RB, Hackanson B, Liu Z, Liu S, Devine H et al. Phase I study of 61 Tallman MS, Andersen JW, Schiffer CA, Appelbaum FR, Feusner JH, Ogeden A et al. decitabine alone or in combination with valproic acid in acute myeloid leukemia. All-Trans-Retinoic Acid in Acute Promyelocytic Leukemia. N Engl J Med 1997; 337: J Clin Oncol 2007; 25: 3884–3891. 1021–1028.

Supplementary Information accompanies this paper on the Leukemia website (http://www.nature.com/leu)