AML1-ETO reprograms hematopoietic cell fate by downregulating scl expression

Acute myelogenous leukemia (AML) is the most common form of leukemia. Each year, more than ten thousand new cases of AML are diagnosed in the United States and approximately a quarter of a million new cases are reported in the world (Stone et al., 2004). The (8;21)(q22;q22) chromosomal translocation can be found in 12-15% of AML patients. This translocation joins the AML1 (also known as CBFα2, RUNX1 and PEBPαB) gene and the ETO (also known as MTG8) gene(Downing, 1999; Peterson and Zhang, 2004). A majority of the patients carrying the AML1-ETO fusion gene have the M2 subtype of AML, according to the French-American-British classification,which is characterized by overproduction of granulocytic precursors. Granulocytes, like the cells of all blood lineages, are derived from multipotent hematopoietic stems cells (HSCs) through a series of cell fate decisions. It is thought that many leukemic oncogenes, including AML1-ETO, may contribute to the development of AML by affecting the specification and maturation of hematopoietic cells(Tenen, 2003).

AML1 by itself plays an important role in hematopoiesis. Normally, AML1 forms a complex with CBFβ, called the core-binding factor (CBF) complex. This complex binds the enhancer core motif and activates tissue-specific expression of a number of hematopoietic genes(Borregaard et al., 2001; Lutterbach and Hiebert, 2000). By contrast, ETO normally functions by recruiting the nuclear receptor co-repressor (N-CoR)/mSin3/histone deacetylase (HDAC) complex(Licht, 2001). The chromosomal translocation juxtaposes the region encoding the DNA-binding domain of the AML1 protein to the region encoding almost all of the ETO protein. Thus, the AML1-ETO fusion product is thought to antagonize the functions of AML1. In addition, previous studies indicate that this fusion protein has additional activities other than antagonizing AML1 function. For example, AML1-ETO knock-in mouse embryos contain abnormal hematopoietic progenitor cells that are not found in AML1-deficient mouse embryos(Okuda et al., 1998; Yergeau et al., 1997). Moreover, many of the AML1-ETO target genes that have been identified are not affected by AML1 (Shimada et al.,2000). At present, the full range of AML1-ETO target genes and their roles in AML pathogenesis remain poorly understood.

Many lines of evidence indicate that the (8;21) chromosomal translocation is likely to occur in a primitive hematopoietic progenitor cell capable of generating all hematopoietic lineages. For example, transcripts of the AML1-ETO fusion gene can be found in hematopoietic cells of non-myeloid lineages in some patients(Miyamoto et al., 2000). Moreover, it has been demonstrated that the cells capable of initiating leukemia are CD34⁺CD38^-, a characteristic of non-committed primitive hematopoietic progenitor cells(Bonnet and Dick, 1997). Thus,it is important to know how AML1-ETO affects the specification of various hematopoietic lineages from multipotent progenitor cells. Unfortunately,AML1-ETO knock-in mouse embryos die in early gestation, making it difficult to analyze the effect of AML1-ETO in these models(Okuda et al., 1998; Yergeau et al., 1997). Interestingly, it has been shown that AML1-ETO can promote myelopoiesis in adult mice (de Guzman et al.,2002; Fenske et al.,2004; Schwieger et al.,2002). However, these mouse models manifest phenotypes only after several months of latency. Thus, it is difficult to identify the immediate transcriptional and cytological changes that lead to the observed AML1-ETO effects.

During embryonic development of mammals, hematopoiesis starts and continues as two successive waves. The first wave, named primitive hematopoiesis, begins in the blood islands of the yolk sac at embryonic day 7 (E7.0) in the mouse. Although it is generally thought that only erythrocytes and macrophages are produced during primitive hematopoiesis in mammals, multilineage precursors are detected in the later, but still pre-circulation, yolk sac(Palis et al., 2001). The second wave of hematopoiesis, named definitive hematopoiesis, begins at E8.5 in the aorta-gonad-mesonephros (AGM) region. Definitive hematopoiesis produces HSCs that not only give rise to all blood lineages, but also possess self-renewal capabilities. Studies of zebrafish embryonic development have also identified two waves of hematopoiesis, a primitive wave beginning at 12 hours post-fertilization (hpf), and a definitive wave beginning at 24 hpf(Davidson and Zon, 2004). There is some evidence that the myeloerythoid progenitor cells (MPCs) arising from zebrafish primitive hematopoiesis are functionally equivalent to the common myeloid progenitors (CMPs) arising from mammalian definitive hematopoeisis, and many of the pathways governing hematopoietic cell fate decisions may be shared between these cells(Galloway et al., 2005; Rhodes et al., 2005). The MPCs of the primitive wave of hematopoiesis reside in two distinct embryonic locations that appear to influence their ultimate cell fates. MPCs of the rostral blood island (RBI) express the myeloid-specific transcription factor Pu.1 (Spi1 - Zebrafish Information Network) and produce cells of the myeloid lineage including macrophages and granulocytes, whereas MPCs of the intermediate cell mass (ICM) region express the erythroid-specific transcription factor Gata1 and produce erythrocytes(Davidson and Zon, 2004). It has been shown that abrogation of pu.1 expression in zebrafish embryos results in erythropoiesis in the RBI. Conversely, abrogation of gata1 expression results in myelopoiesis in the ICM region(Galloway et al., 2005; Rhodes et al., 2005). These results suggest that the primitive hematopoietic cells of the RBI and ICM are multipotent MPCs.

The zebrafish may be a useful model for uncovering the in vivo effects of AML1-ETO expression. Zebrafish MPCs arise in predictable locations and differentiate in synchrony, making it possible to assess the effects of AML1-ETO on reprogramming cell fate decisions. Manipulation of gene expression and tracking of specific cell lineages are straightforward in zebrafish,making it possible to identify the direct effects of AML1-ETO expression in vivo. In addition, zebrafish embryos can be readily adapted for high-throughput chemical screens to identify compounds that modify the effects of AML1-ETO in vivo. Here, we report the generation of an inducible zebrafish AML1-ETO model. Previously, it had been shown that injection of the human AML1-ETO cDNA can cause hematopoietic defects in zebrafish embryos(Kalev-Zylinska et al., 2002). However, the effects of AML1-ETO expression on cell fate decisions had not been explored, and temporal control of AML1-ETO expression was not possible. Using our stable, inducible transgenic line, we demonstrate that AML1-ETO expression in zebrafish disrupts both primitive and definitive hematopoiesis. We show that AML1-ETO reprograms erythropoiesis to granulopoiesis, resulting in a robust phenotype that exhibits cytological and transcriptional characteristics similar to those seen in human AML. We also show that the cell fate changes caused by AML1-ETO expression are preceded by and dependent upon a rapid downregulation of the hematopoietic stem cell factor Scl (also known as Tal1), and that a small molecule histone deacetylase inhibitor is able to rescue many of the effects of AML1-ETO expression on hematopoiesis.

All zebrafish experiments were approved by the Massachusetts General Hospital Subcommittee on Research Animal Care. Zebrafish embryos were collected in Petri dishes and kept in a 23.0-28.5°C incubator until reaching the desired stages. The stages (hours post-fertilization) described in this report are based on the developmental stages of normal zebrafish embryos at 28.5°C (Kimmel et al.,1995).

To construct pHSP/AML1-ETO, we first amplified a 1.5-kb zebrafish hsp70 promoter fragment from pHSP70/4(Xiao et al., 2003) and cloned it into the HindIII and PstI sites of the pG1 vector. Subsequently, the GFP fragment in pG1 was removed and replaced with the XbaI fragment containing the human AML1-ETO gene from pCS2cmv-RUNX1-CBF2T1 (Kalev-Zylinska et al., 2002). The transgenic line is maintained in TL strain background. The zebrafish carrying the transgene were identified by fin clipping and genotyping, using PCR primers AML1-f(5′-GGAAGAGGGAAAAGCTTCAC-3′) and ETO-r(5′-GAGTAGTTGGGGGAGGTGG-3′).

Initially, zebrafish embryos were heat shocked in a 37-42°C incubator for 1 hour. The heat treatment was repeated three to four times every 12 hours from 4 hpf (Fig. 1, Fig. 2B,C, Figs 3, 4; Fig. S1 in the supplementary material). Later on, we found that incubation at 38°C for 1 hour once between 14 to 19.5 hpf is enough to reproducibly induce the AML1-ETO phenotypes, and, thus, this regimen was used in all subsequent experiments(Fig. 2A, Figs 5, 6, 7, 8 and 9; Fig. S2, Figs S3-S6, Table S1 in the supplementary material).

The morpholino antisense oligonucleotides hAML1-MO(5′-CTGGCATCTACGGGGATACGCATCA-3′) and pu.1-MO(5′-GATGATGTACCCCTCCATTCTGTAC-3′) were obtained from Gene-Tools,LLC. The Gata1-MO was purchased from Open Biosystems. For microinjection, 500μM hAML1-MO, 200 μM pu.1-MO or 250 μM Gata1-MO in 0.3×Danieau's buffer were prepared and injected as described(Nasevicius and Ekker, 2000). For scl mRNA injection, scl mRNA was transcribed from Danio rerio full-length IMAGE cDNA clone (catalog number MDR1734,Open Biosystems) using SP6 mMESSAGE mMACHINE (Ambion), subjected to a Poly (A)tailing reaction (Ambion), and was then injected at 100-200 ng/μl concentrations.

Fluorescence microangiography was performed as described(Weinstein et al., 1995).

For embryos older than 26 hpf, hematopoietic cells were isolated as follows. Anesthetized embryos were transferred into phosphate buffered saline(PBS) containing 50 U/ml heparin, 1% bovine serum albumin and 0.006% tricaine. Tails were excised posterior to the yolk extension using a scalpel, and blood cells were extruded from the site of excision with the scalpel and collected using a micropipette.

For embryos younger than 26 hpf, hematopoietic cells were isolated as follows. Homozygous gata1-DsRed transgenic fish were crossed with wild-type or Tg(hsp:AML1-ETO) fish. Embryos were dechorionated and deyolked as described (Westerfield,2000). The embryos were then rinsed in calcium-free Ringer's solution (116 mM NaCl, 2.9 mM KCl and 5 mM HEPES, pH 7.2), and incubated in 0.05% Trypsin-EDTA solution (GIBCO) at 28.5°C for 30 minutes. Finally,dissociated cells were rinsed with PBS and filtered through 40 μm meshes to obtain a single cell suspension. The cell suspension was processed with a flow cytometer and fluorescent cells were collected at the Massachusetts General Hospital Flow Cytometry Core.

For cytological analyses, blood cells collected from the zebrafish embryos were transferred onto glass slides by cytospin and stained by Protocol®Wright-Giemsa stain (Fisher Diagnostics) following the manufacturer's instructions.

For hematopoietic cells collected from the flow cytometer, total RNA was isolated with Trizol (Invitrogen) from 5×10⁴ cells. The samples were then processed and hybridized to Zebrafish Oligo Microarrays(Agilent Technologies) at the Whitehead Institute Center for Microarray Technology. For blood cells extracted from 38- to 40-hour-old zebrafish embryos, total RNA was isolated with RNAqueous®-Micro (Ambion) and was used to synthesize fluorescent probes using the Agilent Low RNA Input Fluorescent Linear Amplification kit. The probes were hybridized to Zebrafish Oligo Microarrays (Agilent Technologies) following the manufacturer's processing protocol. After washing and drying steps, the microarrays were scanned in an Agilent DNA Microarray Scanner and the images were processed using Feature Extraction software.

Digoxigenin-labeled antisense riboprobes for gata1, mpo, l-plastin,scl, flk1, aml1, cmyb and c/ebpα were made according to previous publications (Bennett et al.,2001; Burns et al.,2005; Lyons et al.,2001; Thompson et al.,1998). Whole-mount in situ hybridization was performed as described (Ransom et al.,1996).

To create a zebrafish model of AML1-ETO expression, we generated an inducible transgenic zebrafish line Tg(hsp:AML1-ETO) in which the human AML1-ETO transgene is under the control of the zebrafish hsp70 promoter (Fig. 1A). To test the effect of AML1-ETO expression in zebrafish, we first crossed homozygous Tg(hsp:AML1-ETO) fish with wild-type fish and induced transgene expression at various times (see Materials and methods for heat-shock conditions). We then screened the embryos for any visible phenotypes at 24-40 hours post-fertilization (hpf). Wild-type embryos that have been subjected to the heat treatments do not show any abnormality and establish blood circulation(Fig. 1B,D). By contrast,Tg(hsp:AML1-ETO) embryos have no circulating blood cells even though their hearts are beating. Moreover, the majority of the blood cells in these embryos accumulate in a region that lies along the trunk ventral to the dorsal aorta but above the yolk extension, and in a region posterior to the yolk extension (Fig. 1C,E). The observed phenotype is consistent with the previously reported phenotype caused by injections of AML1-ETO DNA into zebrafish embryos(Kalev-Zylinska et al.,2002).

Using an antisense morpholino oligonucleotide against AML1-ETO, we have confirmed that this phenotype is dependent on the induced expression of AML1-ETO (see Fig. S1 in the supplementary material). In addition, we find that inducing AML1-ETO expression after 21 hpf greatly diminishes its effect (see Fig. S2 in the supplementary material).

In order to determine whether the loss-of-circulation and blood cell accumulation phenotypes are caused by cardiovascular defects in these embryos,we first analyzed the expression of the endothelial lineage marker flk1 at 21 and 36 hpf. We found that, at 21 hpf, flk1expression is slightly decreased in Tg(hsp:AML1-ETO) embryos(Fig. 2A). However, at 36 hpf,there are no significant differences between the expression levels and patterns of flk1 in wild-type embryos and those in Tg(hsp:AML1-ETO) embryos (Fig. 2A). We also crossed Tg(hsp:AML1-ETO) fish with an endothelial reporter line carrying the fli1-EGFP transgene(Lawson and Weinstein, 2002)and found that, in the double transgenic embryos, fli1-EGFPexpression is largely normal, except for a mildly reduced expression in the intersomitic vessels and an expansion in the ventral tail region(Fig. 2B). These results indicate that, if vascular development is affected in Tg(hsp:AML1-ETO) embryos, the defect is subtle. Finally, we employed fluorescent microangiography to test cardiovascular structure and function,and showed that fluorescein-coupled latex beads injected into the inflow tract of the atrium are able to perfuse the whole vascular system of Tg(hsp:AML1-ETO) embryos (Fig. 2C). These data indicate that Tg(hsp:AML1-ETO) embryos possess functional hearts, as well as lumenized and patterned circulatory systems. In contrast to our results, Kalev-Zylinska et al. have shown that injections of DNA containing the AML1-ETO cDNA elicit a vascular patterning defect (Kalev-Zylinska et al.,2002). The discrepancy between our results and those previously reported may be due to differences in the timing and duration of AML1-ETO expression in these two models.

To investigate the blood phenotype induced by AML1-ETO expression, we extracted blood cells at 40 hpf from wild-type and Tg(hsp:AML1-ETO) embryos that had been subjected to the same heat treatment. Using cytology, we determined that the accumulated blood cells in Tg(hsp:AML1-ETO) embryos are dramatically enriched for immature blast-like cells. As shown in Fig. 3, blood from both wild-type and transgenic fish contains a mixture of individual cells and clusters of cells, although cell clusters are more prevalent in samples from the transgenic fish than in samples from wild-type fish. The blood cells from wild-type fish are predominantly erythrocytes, with myeloid cells only occasionally observed(Fig. 3A,D,E). By contrast, the blood cells from the transgenic fish contain abundant blast-like cells, which are larger than the erythrocytes and have high nuclear to cytoplasmic ratios(Fig. 3B,C,F,G). Surprisingly,these blast-like cells remained in the immature state even 4 days after the heat treatment (see Fig. S3 in the supplementary material). Therefore, AML-ETO expression results in an accumulation of immature hematopoietic blast cells in zebrafish embryos.

To test whether transcriptional profiles in the zebrafish AML1-ETO model are similar to expression signatures of human AML, we performed microarray analysis using blood cells from wild-type and Tg(hsp:AML1-ETO)embryos isolated at 38-40 hpf. We found that the scl, gata1, cmyb,nfe2 and znfn1a1 genes are downregulated in Tg(hsp:AML1-ETO) zebrafish blood(Fig. 4). The scl,gata1 and cmyb genes are all expressed in primitive erythroid cells. The reduction of their expression is likely to reflect a reduction of erythropoiesis in Tg(hsp:AML1-ETO) zebrafish. Decreases in SCL expression have also been shown in leukemia samples harboring the t(8;21) translocation (Shimamoto et al.,1995). In addition, it has been shown that MYB expression is downregulated in human AML patients but upregulated in acute lymphoblastic leukemia (ALL) patients (Golub et al.,1999). The NFE2 and ZNFN1A1 (also called Ikaros) genes, both involved in hematopoietic differentiation, are downregulated by AML1-ETO in a human mononuclear cell line(Alcalay et al., 2003).

The lmo1, hoxa9, hoxa10, robo1 and caveolin 1 genes are upregulated in Tg(hsp:AML1-ETO) zebrafish blood(Fig. 4). lmo1upregulation by AML1-ETO has been shown in a cell culture study(Alcalay et al., 2003). In addition, it has been shown that both HOXA9 and HOXA10 genes are strongly upregulated in myeloid but not lymphoid leukemias(Golub et al., 1999; Lawrence et al., 1995). Moreover, the upregulation of ROBO1 and Caveolin 1 have been demonstrated as expression signatures of pediatric AML associated with AML1-ETO (Ross et al.,2004).

In all, these results show that in addition to the cytological similarities between our model and human AML, AML1-ETO exerts comparable transcriptional changes in key hematopoietic genes in zebrafish and in humans.

In zebrafish, definitive hematopoiesis begins between 24-48 hpf, as the hematopoietic stem cell markers aml1 and cmyb begin to be expressed in the ventral wall of the dorsal aorta. It has been shown that AML1-ETO disrupts definitive hematopoiesis in mice(Okuda et al., 1998; Yergeau et al., 1997). Thus,we examined whether AML1-ETO also affects definitive hematopoiesis in zebrafish. We incubated wild-type and Tg(hsp:AML1-ETO) embryos at 38°C for 1 hour at 18 hpf, and fixed the embryos at 33 hpf for aml1 and cmyb in situ hybridization. Although both aml1 and cmyb are expressed in wild-type embryos, they fail to be expressed in Tg(hsp:AML1-ETO) embryos(Fig. 5). These results indicate that AML1-ETO disrupts definitive hematopoiesis in zebrafish embryos.

Next, we sought to understand how AML1-ETO exerts the observed hematopoietic phenotype. Interestingly, the blood accumulation site in Tg(hsp:AML1-ETO) embryos corresponds to the posterior blood island,or intermediate cell mass (ICM) of zebrafish embryos. The ICM contains multipotent MPCs capable of producing cells of the erythroid and myeloid lineages. However, for reasons still to be identified, MPCs in the ICM homogeneously express Gata1, a transcription factor essential for erythropoiesis, and are developmentally programmed to adopt erythroid cell fates (Galloway et al., 2005; Rhodes et al., 2005). By in situ hybridization, we found that gata1 expression is completely abolished in the ICM region of Tg(hsp:AML1-ETO) embryos one hour after AML1-ETO induction (Fig. 6A). The reduction in gata1 expression suggests that AML1-ETO inhibits the normal erythropoietic process(Fig. 6B). In addition, three hours after we observed the downregulation of gata1, we detected increased pu.1 expression in the ICM region of Tg(hsp:AML1-ETO) embryos, as shown by an increase in the number and intensity of the fluorescent cells in AML1-ETO and zpu.1-EGFP double transgenic embryos(Hsu et al., 2004)(Fig. 6A). Pu.1 is a master regulator and marker of myeloid cells. Thus, these results suggest that AML1-ETO reprograms the cell fate decision of many of the multipotent hematopoietic progenitor cells, converting the erythroid cell fate to the myeloid cell fate (Fig. 6B).

We then sought to determine whether the pu.1⁺ myeloid cells arising in the ICM after AML1-ETO induction would go on to adopt the granulocytic and/or monocytic cell fates. After allowing embryogenesis to proceed to 33 hpf, we looked for the presence of granulocytes, which express the myeloid-specific peroxidase (Mpo), and monocytes, which express l-plastin. While both Mpo⁺ and l-plastin⁺ cells were observed in wild-type embryos, Tg(hsp:AML1-ETO) embryos exhibited a dramatic expansion in Mpo⁺ cells and a complete loss of l-plastin⁺ cells in the trunk region(Fig. 6A). We have confirmed that these Mpo⁺ cells are unlikely to be immature erythroid progenitors based on their lack of hemoglobin expression (see Fig. S4 in the supplementary material). Because there is no circulation in Tg(hsp:AML1-ETO) embryos, the abundant Mpo⁺ cells detected in the trunk region are likely to be derived from MPCs in the ICM of these embryos. In addition, the loss of the monocytic lineage appears specific to the ICM, because l-plastin⁺ cells are seen in the RBI, and some of them migrate into various tissues in Tg(hsp:AML1-ETO) embryos(Fig. 6A). These data suggest that AML1-ETO expression directs primitive hematopoietic progenitor cells to the granulocytic cell fate at the expense of erythropoietic and monocytic cell fates (Fig. 6B).

Moreover, it has been shown that the CCAAT/enhancer-binding protein α(C/ebpα) plays a crucial role in the maturation of granulocytes, and that human t(8;21) blasts express very low levels of CEBPA(Pabst et al., 2001a; Pabst et al., 2001b). We find that hematopoietic cells in the trunk region of Tg(hsp:AML1-ETO)embryos do not express C/ebpα (see Fig. S5 in the supplementary material). Thus, the loss of cebpa expression may cause a defect in the maturation of myeloblasts (Fig. 6B), resulting in the accumulation of hematopoietic blasts observed in Tg(hsp:AML1-ETO) embryos(Fig. 3).

It has been shown that, in zebrafish embryos, injections of gata1antisense morpholino oligonucleotides result in increased expression of pu.1 and myelopoiesis in the ICM(Galloway et al., 2005; Rhodes et al., 2005). These results suggest that pu.1 expression in the ICM is normally inhibited by Gata1. Subsequently, the hematopoietic cells in the ICM of gata1morpholino-injected embryos (morphants) express both the granulocytic cell marker mpo and the monocytic cell marker l-plastin(Galloway et al., 2005; Rhodes et al., 2005). We have shown that AML1-ETO also inhibits gata1 expression, leading to increased expression of pu.1 (Fig. 6A). However, in Tg(hsp:AML1-ETO) embryos, the accumulated hematopoietic cells in the ICM express only mpo and not l-plastin. Thus, it is tempting to hypothesize that AML1-ETO actively suppresses the specification of the monocytic cell fate in hematopoietic progenitor cells. To test this hypothesis, we injected gata1morpholinos into Tg(hsp:AML1-ETO) embryos and heat shocked half of the injected embryos only to induce AML1-ETO expression. We find that induced expression of AML1-ETO not only reduces l-plastinexpression in the gata1 morphants, but also enhances mpoexpression in these embryos (Fig. 7A). These effects can be seen in all of the embryos treated. In addition, it has been shown that injections of chordin antisense morpholino oligonucleotides result in an expanded pool of blood cells consisting of mostly erythrocytes and some monocytes in the ICM(Leung et al., 2005). We find that induced expression of AML1-ETO in the chordin morphants also results in downregulation of gata1 and l-plastin, but upregulation of mpo (see Fig. S6 in the supplementary material). These data strongly suggest that AML1-ETO suppresses the monocytic cell fate and promotes the granulocytic cell fate.

Interestingly, when we examined pu.1 expression, we found that gata1 morphants express very high levels of pu.1, but induced expression of AML1-ETO reduces the extent of pu.1induction in gata1 morphants (Fig. 7A). Thus, pu.1 is kept at a moderate level of induction in the ICM of Tg(hsp:AML1-ETO) embryos compared with that of gata1 morphants. It has been shown that Pu.1 is expressed at a higher level in monocytic cells than its level in granulocytic cells(Dahl et al., 2003). Thus, a moderate level of pu.1 induction in Tg(hsp:AML1-ETO) embryos may not be able to support the monocytic cell specification in these embryos. To test whether this moderate level of pu.1 induction is required for the MPCs to adopt the granulocytic cell fate in Tg(hsp:AML1-ETO)embryos, we injected pu.1 morpholinos into Tg(hsp:AML1-ETO)embryos. We found that expression of the granulocytic lineage marker mpo in the ICM is reduced in pu.1 morphants, suggesting that pu.1 expression is required for granulocytic cell specification in Tg(hsp:AML1-ETO) embryos (Fig. 7B).

To identify the immediate downstream targets of AML1-ETO, we examined the transcriptional profiles of hematopoietic cells only two hours after the heat treatment. To isolate blood cells in the early embryos (∼22 hpf), we crossed wild-type and Tg(hsp:AML1-ETO) fish to the gata1-DsRed transgenic zebrafish line(Traver et al., 2003). The gata1-DsRed transgene is expressed transiently in all hematopoietic cells in the ICM before the heat treatment. Because of the stability of DsRed protein, the hematopoietic cells remain fluorescent after the heat treatment. We incubated embryos at 38°C for 1 hour at 19 hpf and isolated fluorescent cells using a flow cytometer. The experiments were done twice and the transcription profiles of these samples were analyzed and compared using DNA microarrays. We found that the most strongly downregulated gene in Tg(hsp:AML1-ETO) embryos is the hematopoietic stem cell gene scl, which is reduced 11-fold (see Table S1 in the supplementary material).

The scl gene encodes a basic helix-loop-helix (bHLH) transcription factor involved in the specification of both hematopoietic and endothelial cells (Gering et al., 1998; Liao et al., 1998). It has also been shown that Scl plays important roles in both primitive and definitive hematopoiesis in mice and in zebrafish(Begley and Green, 1999). The rapid and dramatic downregulation of scl observed in our microarray analyses prompted us to investigate whether it mediates some of the early effects of AML1-ETO expression. To confirm that scl is downstream of AML1-ETO, we heat-shocked wild-type and Tg(hsp:AML1-ETO) embryos at 18 hpf for 1 hour and harvested the embryos 1 hour after the heat treatment for scl in situ hybridization. As shown in Fig. 8A, scl expression is rapidly downregulated in Tg(hsp:AML1-ETO) embryos. Interestingly, this downregulation seems to be specific to hematopoietic cells in the ICM region, as sclexpression in the head (Fig. 8A, asterisk), which will contribute to both cranial vasculature and primitive macrophages, in the trunk above the yolk(Fig. 8A, arrowhead), which is likely to be the precursors of the endothelium of the ducts of Cuvier, and in the posterior tail region (Fig. 8A, arrow) is not affected(Zhang and Rodaway, 2007).

To investigate whether scl downregulation may mediate some of the early effects of AML1-ETO expression that we observed, we injected scl mRNA into Tg(hsp:AML1-ETO) embryos and induced AML1-ETO expression in these embryos. We found that injections of scl mRNA restores gata1 expression (in 47 out of 52 embryos injected; Fig. 8B). Next, we tested if scl expression could also rescue the effect of AML1-ETO on pu.1 and mpo expression. As shown in Fig. 8B,C, we found that injections of scl mRNA can abolish the upregulation of pu.1and mpo in the presence of AML1-ETO expression(Fig. 8B, 39 out of 46 embryos injected; Fig. 8C, 20 out of 56 embryos injected). These data indicate that AML1-ETO leads to the downregulation of gata1 and the upregulation of pu.1 and mpo through scl, and that scl is an important mediator of the effect of AML1-ETO on the specification of multipotent hematopoietic progenitors.

The recruitment of histone deacetylase (HDAC) by the ETO domain of AML1-ETO is believed to play an important role in AML1-ETO-mediated pathogenesis(Wang et al., 1999). To test whether the observed AML1-ETO effects can be suppressed pharmacologically, we added TSA, an HDAC inhibitor, to the embryo media two hours before the heat treatment. We found that 0.5 μM TSA blocks AML1-ETO-mediated downregulation of both scl and gata1, as shown by in situ hybridization(Fig. 9A,B). In addition,adding 0.5 μM TSA to embryos before (data not shown) or after 1-hour heat treatment also reverses the ability of AML1-ETO to induce the accumulation of Mpo⁺ cells (Fig. 9C). These results indicate that the zebrafish AML1-ETO phenotype may be dependent on HDAC activity. Moreover, the zebrafish AML1-ETO phenotype can be reversed by treatment with small molecules and may enable facile identification by high-throughput screens of novel compounds that suppress the effect of AML1-ETO in vivo.

We report here the generation of a highly tractable in vivo model of AML1-ETO expression. We show that induced expression of AML1-ETO in zebrafish embryos results in rapid manifestation of a robust phenotype that exhibits cytological and transcriptional hallmarks of human AML, suggesting that AML1-ETO signaling pathways are likely to be conserved between human and zebrafish. Most importantly, using the zebrafish model of AML1-ETO expression enabled us to track the molecular changes that take place well before morphological phenotypes can be detected,and to determine the roles of candidate AML1-ETO target genes. We demonstrate that AML1-ETO regulates scl and several lineage-specific transcription factors, reprogramming hematopoietic cell fate in vivo.

Induced expression of AML1-ETO in zebrafish results in an accumulation of non-circulating hematopoietic cells. Our results suggest that the loss-of-circulation phenotype is likely to be due to intrinsic defects in hematopoietic cells rather than to a general disruption of cardiovascular function. By histological analyses, we know that at least some hematopoietic cells are able to leave their niche and enter the vasculature (data not shown). In addition, we occasionally observe some circulating cells bypassing a pool of non-circulating blood cells. Interestingly, granulocytic sarcomas(chloromas or myeloblastomas) are a common clinical manifestation of t(8;21)AML (Schwyzer et al., 1998; Tallman et al., 1993). Granulocytic sarcoma is a solid tumor composed of myeloblasts. In this situation, the myeloblasts are very adhesive and have a high tendency to form aggregates. Whether the blood cells in Tg(hsp:AML1-ETO) embryos are more adhesive than those in wild-type embryos is not clear at present.

We show that induced expression of AML1-ETO results in gata1 downregulation and pu.1 upregulation in multipotent hematopoietic progenitor cells, suppressing erythropoiesis and promoting myelopoiesis. These results corroborate the finding that AML1-ETO inhibits erythropoiesis of purified human hematopoietic progenitor cells(Choi et al., 2006), and the finding that it suppresses erythropoiesis and stimulates granulopoiesis in mice (Schwieger et al., 2002). Our results indicate that the level of pu.1 expression is determined by the ability of AML1-ETO to regulate both gata1 and pu.1. AML1-ETO downregulates gata1. As a result, pu.1 is induced. Conversely, AML1-ETO may directly suppress pu.1 expression. It has been shown that AML1-ETO can bind to Pu.1 and inhibit its function(Vangala et al., 2003). Such interaction may pose an inhibitory effect on the autoregulation of the pu.1 gene (Chen et al.,1995). The ability of AML1-ETO to fine tune the level of pu.1 induction is likely to be very important for its leukemogenic effect. Supporting this idea, it has been shown that the hypomorphic allele,but not the null allele, of the Pu.1 gene in mice leads to AML-like phenotypes, suggesting the importance of Pu.1 gene dose in leukemogenesis (Rosenbauer et al.,2005).

Although we demonstrate that AML1-ETO causes cell fate changes, converting erythropoiesis to myelopoiesis, there are clearly still some erythrocytes in Tg(hsp:AML1-ETO) embryos (Fig. 3). We have observed that both induced expression of AML1-ETO and the loss of gata1 expression last only for a few hours after the heat shock (data not shown). Thus, some hematopoietic cells may eventually commit to the erythroid cell fate as a result of the restoration of gata1 expression.

During primitive hematopoiesis, zebrafish mpo expression initiates between 18 and 20 hpf, first in the ICM and then in the RBI. At 24 hpf,myeloblasts can be identified morphologically only in the RBI and not in the ICM, although granulocytes are reliably found in the circulation by 48 hpf(Bennett et al., 2001; Lieschke et al., 2001). The source of these circulating granulocytes is not totally clear at this moment. For example, it is not known whether the Mpo⁺ cells observed in the ICM at 20 hpf can become mature in situ. In our model, induced expression of AML1-ETO causes an enrichment of Mpo⁺ cells and morphologically immature hematopoietic blasts in the ICM. The accumulation of hematopoietic blasts is only partially reversed as AML1-ETO expression ceases (see Fig. S3 in the supplementary material). What causes the long-lasting effect of transient induction of AML1-ETO on the blockade of maturation of these cells is not clear. It could be that these cells cannot migrate to the environment that supports their maturation, that the environmental signals that induce the maturation of these cells no longer exist, or that a factor that is required for maturation, such as C/ebpα, can no longer be expressed after AML1-ETO expression diminishes.

We were able to identify several early downstream targets of AML1-ETO by isolating hematopoietic cells only two hours after the induction of AML1-ETO expression. We show that induced expression of AML1-ETO rapidly downregulates scl expression. Even though decreased scl expression has been shown in t(8;21) leukemia samples, a direct link between AML1-ETO and scl expression had not been established previously (Shimamoto et al.,1995). We have not been able to identify potential Aml1 binding sites upstream of the coding region of scl, indicating that AML1-ETO may inhibit scl expression through binding with other factors. Interestingly, the downregulation of scl is seen only in hematopoietic progenitor cells and not in cells that will give rise to the endothelial lineage in Tg(hsp:AML1-ETO) embryos. It has been shown that the heat shock promoter used to drive the human AML1-ETO gene results in ubiquitous expression upon induction(Xiao et al., 2003). Thus,these results suggest that other hematopoietic-specific cofactors may be required for AML1-ETO function.

Overexpression of scl can block the ability of AML1-ETO to reprogram hematopoietic cell fate, as shown by the reversal of gata1,pu.1 and mpo dysregulation in Tg(hsp:AML1-ETO) embryos. However, Scl-mediated rescue of the early effects of AML1-ETO on hematopoiesis does not completely rescue the circulation defect in Tg(hsp:AML1-ETO)embryos. This may be because of additional defects caused by ubiquitous scl overexpression, or by non-Scl-mediated effects of AML1-ETO at later stages of development. Nevertheless, these results indicate that scl is an important effector that mediates the earliest observable effects of AML1-ETO in hematopoietic progenitor cells.

Induced expression of AML1-ETO in zebrafish embryos before 21 hpf disrupts definitive hematopoiesis but also enables us to study the in vivo effects of AML1-ETO in MPCs generated during primitive hematopoiesis. The transcriptional profile of Tg(hsp:AML1-ETO) embryo blood at 40 hpf presents expressional signatures of human AML, suggesting that the regulatory mechanisms of cell specification and maturation in these cells resemble those in progenitors of adult human blood. However, MPCs in zebrafish embryos may not have the self-renewal capability that is also essential for leukemogenesis. Thus, the effects of AML1-ETO on the self-renewal capability of hematopoietic progenitor cells may not be accessible using this model. Additionally, in humans and mice, AML1-ETO is not sufficient to induce leukemia in the absence of secondary mutations, so the blast cells observed in AML1-ETO-expressing fish are unlikely to possess full leukemogenic potential. In future studies, it will be interesting to combine AML1-ETO expression with known collaborating mutations, and to characterize the effects of AML1-ETO expression in hematopoietic stem cells by inducing AML1-ETO expression at later stages of development, or in adults.

Understanding the molecular mechanisms by which AML1-ETO exerts its influence on hematopoietic progenitor cells may help us develop targeted therapeutics. For example, compounds that block SCL or GATA1downregulation in the presence of AML1-ETO expression may prove useful in treating AML associated with t(8;21) translocation. It has been shown that compounds that enhance Cebpa (C/ebpα) transcriptional activity induce differentiation of AML cell lines(Jiang et al., 2005). Moreover, high-throughput screening is feasible in the zebrafish(Yeh and Crews, 2003; Zon and Peterson, 2005). We have shown that some of the effects of AML1-ETO can be suppressed by a histone deacetylase inhibitor Trichostatin A. Therefore, in addition to the fundamental insights into the mechanism of AML1-ETO function already provided by this model, it may ultimately provide a unique opportunity to conduct whole-organism chemical suppressor screens to identify compounds that can reverse AML1-ETO function in vivo.

We thank Drs K. E. Crosier and J. Y. Kuwada for providing AML1-ETOand zebrafish hsp70 plasmids; and Drs A. T. Look and L. I. Zon for providing Tg(zpu.1:EGFP) and Tg(gata1:DsRed) fish. We also thank Elma Feric for technical assistance; and Drs K. R. Chien, I. A. Drummond, D. A. Haber, D. A. Sweetser and Developmental Biology Laboratory members for helpful suggestions on this work and on the preparation of this manuscript. We are grateful to Brian Patterson of Agilent Technologies for assistance with the microarray experiments. The authors received financial support from the Ned Sahin Research Fund for Restoring Developmental Plasticity, the Novartis Institutes for BioMedical Research, and the National Cancer Institute. J.-R.J.Y. was supported by a Tosteson Fellowship from the Massachusetts Biomedical Research Corporation.