RUNX1/AML1/CBFA2 is essential for definitive hematopoiesis, and chromosomal translocations affecting RUNX1 are frequently involved in human leukemias. Consequently, the normal function of RUNX1 and its involvement in leukemogenesis remain subject to intensive research. To further elucidate the role of RUNX1 in hematopoiesis, we cloned the zebrafish ortholog (runx1) and analyzed its function using this model system. Zebrafish runx1 is expressed in hematopoietic and neuronal cells during early embryogenesis. runx1 expression in the lateral plate mesoderm co-localizes with the hematopoietic transcription factor scl, and expression of runx1 is markedly reduced in the zebrafish mutants spadetail and cloche. Transient expression of runx1 in cloche embryos resulted in partial rescue of the hematopoietic defect. Depletion of Runx1 with antisense morpholino oligonucleotides abrogated the development of both blood and vessels, as demonstrated by loss of circulation, incomplete development of vasculature and the accumulation of immature hematopoietic precursors. The block in definitive hematopoiesis is similar to that observed in Runx1 knockout mice, implying that zebrafish Runx1 has a function equivalent to that in mammals. Our data suggest that zebrafish Runx1 functions in both blood and vessel development at the hemangioblast level, and contributes to both primitive and definitive hematopoiesis. Depletion of Runx1 also caused aberrant axonogenesis and abnormal distribution of Rohon-Beard cells, providing the first functional evidence of a role for vertebrate Runx1 in neuropoiesis.
To provide a base for examining the role of Runx1 in leukemogenesis, we investigated the effects of transient expression of a human RUNX1-CBF2T1 transgene [product of the t(8;21) translocation in acute myeloid leukemia] in zebrafish embryos. Expression of RUNX1-CBF2T1 caused disruption of normal hematopoiesis, aberrant circulation, internal hemorrhages and cellular dysplasia. These defects reproduce those observed in Runx1-depleted zebrafish embryos and RUNX1-CBF2T1 knock-in mice. The phenotype obtained with transient expression of RUNX1-CBF2T1 validates the zebrafish as a model system to study t(8;21)-mediated leukemogenesis.
The mammalian hematopoietic system is established early during embryogenesis and is required for the continuous production of blood during fetal and adult life. During embryonic development, a transient wave of ‘primitive’ hematopoiesis that primarily gives rise to embryonic erythrocytes is followed by a second ‘definitive’ hematopoietic wave that establishes all blood lineages (reviewed by Orkin, 2000; Cumano and Godin, 2001). In mammals, hematopoietic stem cells (HSCs) are initially found in the extra-embryonic yolk sac blood islands and the intra-embryonic para-aortic splanchnopleura/aorta-gonad-mesonephros region (P-Sp/AGM). In successive waves of migration, HSCs colonize the fetal liver, spleen and bone marrow (reviewed by Dzierzak and Medvinsky, 1995; Robb, 1997). HSCs are mesodermal in origin, and this tissue also gives rise to angioblasts, the precursors of the vascular system. Although yet to be directly identified in vivo, increasing evidence points towards the existence of a mesodermally derived bipotential precursor termed the ‘hemangioblast’, from which both HSCs and angioblasts differentiate (Choi et al., 1998; Pardanaud and Dieterlen-Lievre, 1999). There is also evidence for the development of HSCs from ‘hemogenic endothelium’ in the ventral wall of the aorta (Jaffredo et al., 1998).
Recently, zebrafish (Danio rerio) genetics has contributed towards the understanding of early blood and vessel development (Amatruda and Zon, 1999; Paw and Zon, 2000). Hematopoietic programs are largely conserved between mammals and zebrafish (Zon, 1995), and the genetic amenity of the zebrafish has advanced its cause as a strong model system for developmental studies (Knapik, 2000). Several zebrafish mutants with defects in blood development exist. For example, the cloche mutant (Stainier et al., 1995) fails to develop mature blood or vessels, providing genetic evidence for the existence of a hemangioblast. Uncovering the mutation responsible for cloche is expected to provide insight into the molecular events that direct commitment of mesoderm towards blood and/or endothelial fates.
Studies in mammals have identified key transcriptional regulators that are involved in early commitment to a hematopoietic stem cell fate (Shivdasani and Orkin, 1996; Davidson and Zon, 2000) of which one example is RUNX1 (Downing, 1999; Tracey and Speck, 2000). RUNX1 is a member of the runt family of transcriptional regulators that are involved in many developmental processes, ranging from segmentation and sex determination in Drosophila to blood and bone development in mammals (reviewed by Westendorf and Hiebert, 1999; Canon and Banerjee, 2000). The RUNX1 gene (also known as AML1/CBFA2/PEBP2αB) was first isolated from the chromosome 21 breakpoint in t(8;21)(q22;q22) (Miyoshi et al., 1991). This translocation is found in approximately 40% of individuals with the M2 subtype of acute myeloid leukemia (AML) (Bitter et al., 1987) and results in the formation of a chimeric protein, now known as RUNX1-CBF2T1 (formerly AML1-ETO) (Erickson et al., 1992; Miyoshi et al., 1993). RUNX1 is involved in several other chromosomal translocations in acute leukemias, of which TEL-RUNX1 in t(12;21) and RUNX1-EVI1 in t(3;21) are the most common (reviewed by Lutterbach and Hiebert, 2000). In addition, point mutations in RUNX1 have been described in AML and myelodysplasia (Osato et al., 1999; Imai et al., 2000; Preudhomme et al., 2000), and haploinsufficiency at the RUNX1 locus causes familial thrombocytopenia with predisposition to AML (Song et al., 1999). Like other runt family members, RUNX1 contains a conserved runt domain (RD) that mediates binding to its target sequence, TGT/cGGT, and is also necessary for the physical interaction with the heterodimeric partner, CBFβ (Meyers et al., 1993; Ogawa et al., 1993b; Crute et al., 1996). Both of these functions are mediated by distinct, non-overlapping RD sites (Nagata and Werner, 2001). The structure of a RD-CBFβ complex bound to its cognate sequence has recently been solved, providing insight into the molecular basis of human disease involving RD mutations (Bravo et al., 2001).
Runx1 has been shown to play a critical role in blood development. Mice that lack Runx1 have no definitive hematopoiesis, display central nervous system (CNS) hemorrhages, and die in midgestation (Okuda et al., 1996; Wang et al., 1996a). P-Sp explant cultures from these mice exhibit defective vascular formation that can be rescued by HSCs (Takakura et al., 2000). Vascular growth factors were shown to augment Runx1 expression in an endothelial cell line, and expression of a CBFβ-MYH11 fusion inhibited the angiogenic activity of these cells (Namba et al., 2000). These observations suggest involvement of Runx1 in the process of angiogenesis.
Investigations directed at understanding how translocations involving RUNX1 contribute to the development of AML have shown that the RUNX1-CBF2T1 fusion protein is able to repress RUNX1 responsive genes (Okuda et al., 1998; Westendorf et al., 1998). CBF2T1/ETO normally interacts with the nuclear receptor co-repressor complex [N-CoR/mSin3/histone deacetylase (HDAC)], and thereby mediates transcriptional repression (Melnick et al., 2000; Hildebrand et al., 2001). Therefore, the RUNX1-CBF2T1 fusion protein could recruit HDAC to the promoters of genes that would normally be activated by RUNX1. Recently, it has been demonstrated that RUNX1-CBF2T1 influences the regulation of target genes with potential relevance in myeloid leukemogenesis. The fusion protein was shown to upregulate TIS11b, which induces myeloid cell proliferation when overexpressed (Shimada et al., 2000), and to downregulate the granulocytic differentiation factor C/EBPα (Pabst et al., 2001).
To investigate the role of RUNX1-CBF2T1 in leukemogenesis, RUNX1-CBF2T1 knock-in mice have been created (Yergeau et al., 1997; Okuda et al., 1998). These mice die in midgestation with a phenotype very similar to that seen in Runx1 and Cbfb null mice, providing further evidence that RUNX1-CBF2T1 abrogates the normal function of RUNX1. The embryonic lethality of RUNX1-CBF2T1 expression is problematic with respect to the development of a model system for the analysis of t(8;21) leukemogenesis. To overcome this obstacle, mice that conditionally express RUNX1-CBF2T1 under control of a tetracycline-responsive element (Rhoades et al., 2000), and mice in which transgene expression is regulated by the myeloid-specific human MRP8 promoter (Yuan et al., 2001) have been developed. Results show that RUNX1-CBF2T1 alone has a restricted capacity to transform cells and suggest that additional mutations are necessary to generate a leukemic phenotype. Inducible translocation of the Runx1 and Cbf2t1 genes in mice through Cre/loxP-mediated recombination has been described (Buchholz et al., 2000). Together, these mice provide valuable model systems for the analysis of t(8;21) leukemogenesis; however, it might be helpful to develop alternative animal models that are also highly amenable to genetic studies. The zebrafish offers genetic and developmental advantages that may facilitate analysis of the role of RUNX1 in normal hematopoiesis and in leukemia.
We consider the viability of the zebrafish as a model for the study of Runx1 function, and of RUNX1-CBF2T1-related disease. Initially, we cloned the zebrafish runx1 gene and examined its role in early embryonic development. Zebrafish runx1 shares a high degree of similarity with its mouse and human counterparts, and mapping data reflect conservation of synteny between the zebrafish and mammalian Runx1 genes. During zebrafish development, runx1 is expressed in hematopoietic and neuronal cells. The abrogation of Runx1 function using morpholino antisense oligonucleotides suggested a role for Runx1 in early hematopoiesis, vasculoangiogenesis and neuropoiesis. Our results confirm that zebrafish Runx1 functions in a similar manner to its mammalian orthologs. Furthermore, we found additional roles for zebrafish Runx1 in vasculogenesis and neuropoiesis. We then investigated the possibility that zebrafish could be used as a model for the study of RUNX1-CBF2T1-mediated leukemogenesis. Transient expression of a human RUNX1-CBF2T1 cDNA in zebrafish embryos caused a phenotype analogous to that observed in the RUNX1-CBF2T1 knock-in mice. Therefore, the zebrafish can serve as an alternative model system for examining additional genetic events that contribute to t(8;21)-mediated leukemia.
MATERIALS AND METHODS
Wild-type zebrafish (Danio rerio) were maintained, and embryos collected and staged as described (Kimmel et al., 1995; Westerfield, 1995). Embryos older than 24 hours post-fertilization (hpf) were usually incubated in 0.003% 1-phenyl-2-thiourea (PTU; Sigma) to inhibit pigmentation. cloche mutant embryos (clom39; obtained from L. Zon), were generated from pairwise crosses between identified heterozygous males and females.
Isolation of the zebrafish runx1 gene
Coding sequence of human AML1a/RUNX1a (X79549) encompassing the RD was amplified from cDNA derived from the Jurkat T-cell lymphoma line using 5′-CCGCTTCACGCCGCCTTCCACC-3′ (forward) and 5′-GGGCTGGGTGTGTGGGCTGAC-3′ (reverse) primers. This product was used as a probe to screen a 24 hour zebrafish embryo lambda cDNA library (Stratagene) at low stringency (2×SSC, 0.1% SDS at 55°C). One positive clone was isolated and the insert subcloned into pBluescriptII SK+ (Stratagene). Sequencing was done using an ABI 377XL sequencer.
Phylogenetic analysis was performed using the Phylogenetic Inference Package, PHYLIP 3.5 (Felsenstein, 1993). The sequences were aligned using the ClustalW method. Dendrograms were generated from amino acid sequences of the entire protein or RD and outgrouped to Drosophila runt. Tree reconstruction was done using the neighbor-joining algorithm and percent bootstrap values were derived from 1000 replications.
Mapping was performed using SSCP on the MOP haploid mapping panel (Postlethwait et al., 1998). For display, the position of runx1 was intercalated into the HS panel (Woods et al., 2000). Comparative mapping with human was accomplished using the human chromosome 21 database at the Weizmann Institute (http://bioinformatics.weizmann.ac.il/chr21/). The mapping primers were: runx1, 5′-TCTATGCCTTTTCTATGGTTTCTTTTCTAA-3′ (forward) and 5′-GCTCGCCCGCTGATTGTG-3′ (reverse); app, 5′-ATGGAGCACCGTCACCCCTAACC-3′ (forward) and 5′-ACTTTGGCCATTGATTTGAACTGA-3′ (reverse); and gart, 5′-GAGCATTCCAGATCCAGACATTCC-3′ (forward) and 5′-CCTCTGAGAAGCTCCAGTTTTACTC-3′ (reverse).
Generation of expression constructs and microinjection of zebrafish embryos
The construct pCS2cmv-runx1 contained a 1442 bp EcoRI-DraI fragment of zebrafish runx1 cloned into the EcoRI and StuI sites of the expression vector pCS2+ (Rupp et al., 1994). A human RUNX1-CBF2T1 cDNA was kindly provided by Dr Scott W. Hiebert (Vanderbilt University, Nashville, TN). This was introduced into pCS2+ digested with XbaI, to generate pCS2cmv-RUNX1-CBF2T1. For expression in zebrafish embryos, approximately 100 pg of pCS2cmv-runx1 and 100-150 pg of pCS2cmv-RUNX1-CBF2T1, both linearized at a 3′ NotI site were injected per embryo (Westerfield, 1995). Microinjections were carried out using a Narishige micromanipulator (type GJ) and an MPPI-2 Milli-Pulse Pressure Injector (Applied Scientific Instrumentation) under a Leica MZ12 dissecting microscope.
Protein was isolated from pools of embryos injected with pCS2cmv-RUNX1-CBF2T1, after yolk removal as described (Westerfield, 1995). Samples (10 embryos equivalent per lane) were separated on 10% SDS-polyacrylamide gel and transferred to PVDF membrane. The Jurkat cell line that expresses RUNX1 (Takahashi et al., 1995), was used as a positive control. The blot was incubated with rabbit anti-human AML1/RD polyclonal antibody (1:40; Oncogene), and developed with goat anti-rabbit peroxidase conjugate (1:5000; Amersham) and the ECL detection system (Amersham).
Two morpholino antisense oligonucleotides (MO) targeting the runx1 transcript were obtained (Fig. 1A) (Gene-Tools, LLC): runx1-MO1, 5′-TGGCGTCCCAAAGAAAAACCATTT-3′; runx1-MO2: 5′-TTTGGTATGTTTTTGTCTCCGTGAG-3′. Sequence complementary to the predicted start codon is underlined. An antisense oligonucleotide with four base mismatches when compared with runx1-MO2 was used as a control: 5′-TTTGCTATGATTTTGACTCCCTGAG-3′ (mismatched bases are underlined). Solutions were prepared and injected as described (Nasevicius and Ekker, 2000).
Preparation of antisense RNA probes and runx1 mRNA
For runx1 in situ hybridization with zebrafish embryos, two probes were synthesized (Fig. 1A). Template 1 (1682 bp) was generated by PCR using 5′-GGTAAGCTTCGGGGAAGATGAGCGAGGGTTT-3′ (forward) and 5′-GGGGAATTCTGGGAGGAAACACTAGCTGTGC-3′ (reverse) primers. This template contained 1280 bp of coding sequence (nucleotides 77-1356 including the RD) and 403 bp of adjacent 3′-UTR. It excluded a further 3′-UTR with homology to a zebrafish mermaid repeat. Template 2 (981 bp; generated by ApaI digest of template 1 to exclude RD) contained nucleotides 779-1356 of the coding sequence and 403 bp of 3′-UTR. For synthesis of RNA probes, template 1 was linearized with HindIII and template 2 with KpnI, and both were transcribed with T7 polymerase using an RNA labeling kit (Boehringer Mannheim) according to the manufacturer’s instructions. Full-length antisense digoxigenin (DIG)- or fluorescein (FLU)- labeled riboprobes for zebrafish scl, flk-1 (kdr – Zebrafish Information Network) myb and βE3-globin (hbbe3 – Zebrafish Information Network) were synthesized using T7 polymerase from templates (provided by L. Zon) that had been linearized with SalI, SmaI, EcoRI and SmaI/KpnI respectively.
To generate synthetic runx1 mRNA, a pCS2cmv-runx1 construct was digested with NotI and capped mRNA was synthesized in vitro using the SP6 mMESSAGE mMACHINE Kit (Ambion). For rescue experiments, 10 pg of runx1 mRNA was injected, as described previously (Nasevicius and Ekker, 2000).
Whole-mount in situ hybridization and immunostaining
In situ hybridization using DIG- or FLU-labeled antisense riboprobes was performed as described (Broadbent and Read, 1999) with modifications. For in situ hybridization with runx1 probes, temperatures of 65-70°C were used for hybridization and wash steps (60-65°C for other probes). Washes were performed as follows: 2×SSCT/75%formamide, 2×SSCT/50%formamide, 2×SSCT/25%formamide, 2×SSCT (15 minutes), 0.2×SSCT (twice, 30 minutes), followed by three washes in PBST (5 minutes) and a rinse in MABT. Embryos were blocked in 2% blocking reagent (Boehringer Mannheim) for 3-4 hours. Hybridization was detected with anti-DIG or anti-FLU antibodies (Boehringer Mannheim) coupled to alkaline phosphatase (AP). Excess antibody was removed by eight washes with PBST (15 minutes). Bound antibody was visualized using the AP substrates BM Purple, Fast Red (Boehringer Mannheim) or NBT/BCIP (Promega). Double in situ hybridization was performed as described (Broadbent and Read, 1999).
After in situ hybridization, immunostaining of zebrafish embryos was performed as described (Macdonald, 1999). Mouse monoclonal anti-HNK-1/N-CAM (1:1000; Sigma) was used as a primary antibody. This was detected with a goat anti-mouse IgM peroxidase conjugate (1:300; Sigma).
Histology and cytology
Following runx1 in situ hybridization, embryos to be sectioned were re-fixed, mounted in agarose, dehydrated in ethanol and infiltrated with JB4 resin (Polysciences). Sections (5 μm) were cut using a RM2155 microtome and counterstained with nuclear fast red (Vector). Other embryos were fixed overnight, embedded in paraffin, sectioned at 3 μm on a Leica RM2135 microtome and stained with Hematoxylin and Eosin (Sigma). Cytological analysis of embryonic blood was performed essentially as described (Ransom et al., 1996). Cells were aspirated with a pulled glass capillary connected to a manual piston pump (CellTram Oil; Eppendorf).
Imaging and microangiography
Most images were captured on a MZ FLIII stereomicroscope using a Leica DC 200 camera and a PC Pentium III-600 equipped with Leica imaging software. For two-color runx1 in situ imaging and fluorescent photography, a ProgRes3008 digital camera (Zeiss Jenoptik) with associated software was used.
Microangiography of zebrafish embryos was performed as described (Weinstein et al., 1995) with modifications. Yellow-green fluoresceinated carboxylated latex beads (Molecular Probes) were used. After dilution, they were sonicated on ice using a Misonix sonicator for 25 minutes (five 5-minute cycles). Injection was performed as described elsewhere (http://mgchd1.nichd.nih.gov:8000/zfatlas/Intro%20Page/angiography.html) (Isogai et al., 2001). Image capture was performed with a Leica MZ FLIII stereomicroscope as described above, using a standard FITC filter set.
The cDNA sequence of the zebrafish runx1 gene reported in this manuscript has been deposited in GenBank under the accession number AF391125.
Isolation of zebrafish runx1
A human RUNX1a 620 bp cDNA probe encompassing the RD was used to screen a 24 hour old zebrafish cDNA library at low stringency, and a single 2.5 kb clone was isolated (Fig. 1A) (GenBank Accession Number, AF391125). Sequence analysis showed that this clone is homologous to human and mouse Runx1. The cDNA encodes an open reading frame of 451 amino acids that shows an overall 73% and 72% identity with the human and mouse Runx1 proteins, respectively. The highest degree of conservation is present within the RD, which shares 96% identity with both human and mouse proteins (Fig. 1B). All amino acid substitutions within the RD are conservative changes. At the DNA level, runx1 is highly similar to the human RUNX1 gene (71% overall identity and 84% within RD) and to the mouse Runx1 gene (70% overall and 82% within RD) in the coding region. Our zebrafish runx1 clone shares 99% sequence identity with a runxa clone isolated by Kataoka et al. (Kataoka et al., 2000). Five of the eight nucleotides that differ are located within the RD. Despite the nucleotide differences, both clones encode identical proteins. Phylogenetic analysis, based on amino acid sequences of either the entire protein or RD only, revealed that the closest relative of zebrafish runx1 is the Xenopus runx1 (Xaml), and that these two genes form separate outgroups from their mammalian Runx relatives (Fig. 1C). Mapping studies showed that runx1 is located on the MOP haploid mapping panel at a position equivalent to LG1_7.7 cM on the HS meiotic mapping panel (Woods et al., 2000) or LG1_3.5cM on the MGH mapping panel (Shimoda et al., 1999) (Fig. 1D). This region in the zebrafish genome exhibits conserved synteny with a region of human chromosome 21 that includes RUNX1 at 21q22.3 and murine Runx1 at 16_62.2 (Miyoshi et al., 1991; Bae et al., 1994). This conservation of synteny together with the phylogenetic analysis of the gene we have isolated indicates that it is very probably the zebrafish ortholog of RUNX1.
Zebrafish runx1 is expressed in hematopoietic and neuronal cells
We used whole-mount in situ hybridization analysis to characterize expression of the runx1 gene during zebrafish embryogenesis (Fig. 2). In all experiments, the two probes used yielded identical patterns of expression. Diffuse runx1 expression at 5 hpf suggests the presence of a maternal transcript (Fig. 2A). A specific pattern of zygotic runx1 expression can be detected from 12 hpf. Expression was first detected as bilateral stripes in the lateral plate mesoderm (LPM), and appeared to be most distinct at the posterior end of the embryo (Fig. 2B-D). At 18 hpf, runx1 expression was strongest at the end of the yolk extension where the bilateral stripes of runx1 expression appeared closer to the midline (Fig. 2E,F). From ∼18 hpf, cells of the LPM migrate to the midline to form the intermediate cell mass (ICM) (Al-Adhami and Kunz, 1977). Strong runx1 expression was observed in the developing ICM during this time (Fig. 2G). At 24 hpf, expression of runx1 in the posterior ICM weakened, while new patches of runx1 expression appeared in a more anterior position in the ventral wall of the dorsal aorta (Fig. 2H-J). A similar domain of Runx1 expression is found during mouse embryogenesis (Simeone et al., 1995). In addition, the temporal and spatial expression of runx1 is very similar to that of the hematopoietic transcription factor scl (Gering et al., 1998). Therefore, we investigated in more detail the relationship between scl and runx1 during the first 24 hours of development.
SCL is essential for the development of blood and vasculature (reviewed by Begley and Green, 1999), and zebrafish scl marks the earliest sites of commitment to these fates (Gering et al., 1998). Double in situ hybridization revealed co-expression of scl and runx1 within the LPM at 12 hpf (Fig. 3A). Dual expression of runx1 and scl was observed in individual cells (Fig. 3B,C). At 24 hpf, when runx1 expression weakens in the posterior ICM and appears in the ventral wall of the dorsal aorta, differences in the domains of expression of runx1 and scl within the posterior ICM were observed (Fig. 3D,E).
In addition to expression in hematopoietic tissues, we observed runx1 expression in neuronal cells within the head and the trunk (Figs 2, 3). Neuronal expression of runx1 first appeared at 14 hpf in individual cells along the anteroposterior axis that are located medial to the LPM (Fig. 2C,D). Immunolabeling with an antibody against HNK-1, a neuronal marker, indicated that these cells are Rohon-Beard neurons (sensory neurons related to the dorsal root ganglia) (Kruse et al., 1984; Artinger et al., 1999) (Fig. 3F,G). Expression of HNK-1 during early development in the zebrafish has been previously characterized (Metcalfe et al., 1990). The earliest neurons to express this tetrasaccharide are the primary sensory neurons mediating touch sensitivity, the Rohon-Beard neurons of the spinal cord and the trigeminal ganglion sensory neurons in the head. We found that only a subset of the Rohon-Beard cells display runx1 expression (Fig. 3F,G). runx1 transcripts were also detected in other neuronal tissues such as the olfactory placode (Fig. 2K,L). At 18 hpf, HNK-1 and runx1 mRNA co-localize in clusters of cells located against the anterior-medial surface of the otic vesicle (Fig. 3H,I). These cells may represent acoustico-vestibulo (VIII) cranial nerve ganglia.
runx1 acts downstream of spadetail and cloche
To determine whether runx1 is implicated in molecular pathways leading to the commitment of mesoderm to a blood or vascular fate, we analyzed runx1 expression in spadetail (sptb104) and cloche (clom39) zebrafish mutant embryos (Fig. 2O,P). The spt mutation abrogates the zebrafish tbx16 gene (Griffin et al., 1998), and results in defective differentiation of mesodermal cells such that they fail to converge during gastrulation (Ho and Kane, 1990). Consequently, spt embryos exhibit aberrant patterning of somites and the accumulation of mesodermal cells in the tail. Additional data shows that spt embryos fail to generate differentiated blood (Thompson et al., 1998). At 18 hpf, runx1 expression in the ICM region of spt embryos is maintained, although at significantly reduced levels (Fig. 2O). clo embryos are defective in the production of both blood and endothelial cells (Stainier et al., 1995), therefore the clo gene product is probably required for mesodermal commitment to both of these fates. In 24 hpf clo embryos, no runx1 expression was detected in the region that would normally form the dorsal aorta, and runx1 expression in the ICM region was markedly reduced compared with wild-type embryos (Fig. 2P). Furthermore, transient expression of runx1 partially rescued the clo hematopoietic defect (Fig. 4). 683 embryos obtained from clo heterozygote mating pairs were injected with 100 pg of pCS2cmv-runx1. These were scored for the mutant phenotype on the basis of the enlarged heart at 48 hpf. In 22% (36/163) of clo embryos, red blood cells were seen in the trunk (Fig. 4A). This was confirmed by a hbbe3 in situ (Fig. 4D). Overexpression of the same dose of pCS2cmv-runx1 in wild-type zebrafish embryos produced ectopic blood in up to 10% of injected embryos (Fig. 4C).
These findings are consistent with a role for runx1 downstream of spt and clo in pathways leading to the differentiation of blood and vasculature.
runx1 is required for the development of blood and vasculature and for neuropoiesis
Antisense, morpholino-modified oligonucleotides (morpholinos) have recently proved to be effective and specific translational inhibitors in zebrafish (Ekker, 2000; Nasevicius and Ekker, 2000). We used this targeted gene knock-down technique to generate runx1 loss-of-function embryos. Morpholinos were injected into the yolk of zebrafish embryos at one to eight cell stages. Two morpholino oligonucleotides targeting runx1 (Fig. 1A) were used to provide a control for specificity. Amounts of injected morpholino ranged from 0.5 to 16 ng. Injection of amounts higher than 4 ng were associated with nonspecific effects in addition to a specific phenotype. For runx1-MO1, nonspecific effects consisted of gastrulation defects that caused embryo death by 18 hpf. High levels of runx1-MO2 led to nonspecific necrosis within the CNS. To generate the data shown, we injected 2 ng of runx1-MO1. Eighty percent of embryos presented a specific phenotype, in comparison with 50% for the same dose of runx1-MO2. In both cases there were minimal nonspecific consequences, and the remaining embryos were normal. Injection of 2-6 ng of a runx1 control morpholino with four mismatched bases did not produce a phenotype. The effects of both active morpholinos (2 ng) were fully rescued with 10 pg runx1 mRNA, an amount that was insufficient to induce a phenotype on its own (data not shown).
We found that runx1-MO injected embryos (runx1-mo embryos) displayed dramatic defects in hematopoiesis, vasculogenesis and neuropoiesis in early embryonic development (Figs 5-8). Later embryos became severely edematous, and death occurred after 6-7 days. This edematous phenotype was similar to that seen in vegfa morphant embryos (Nasevicius et al., 2000).
Embryos injected with runx1-morpholino show disrupted vasculature and lack normal circulation
The most striking abnormality observed in the runx1-MO injected embryos was the lack of normal circulation at 48 hpf, with accumulation of red blood cells in the aorta and ventral tail (Fig. 5A-H). This phenotype occurred despite the presence of a beating heart. A small amount of blood was occasionally seen in the vitelline vessels, but was found to be stationary (data not shown). Consistent with the observed accumulation of blood cells in the dorsal aorta, microangiography performed at 52 hpf demonstrated a lack of circulation through the trunk and tail, with arrest of flow at the mid-trunk level (Fig. 5I,J).
Erythrocyte expression of hbbe3 was used to examine the functional integrity of the circulatory system of runx1-mo embryos (Fig. 6A,B). hbbe3 mRNA was limited to the posterior half of the injected embryos, with a marked reduction of blood cells seen in the vitelline and cranial vessels when compared with normal. The observed distribution of hbbe3 mRNA suggests that blood cells are unable to circulate and have become trapped in the dorsal aorta. To investigate the nature of the vascular defects observed in runx1-mo embryos, we used the expression of flk-1 (kdr – Zebrafish Information Network) (Yamaguchi et al., 1993; Liao et al., 1997) to mark developing vascular endothelial cells (Fig. 6E,F,I,J). In normal embryos, two stripes of flk-1-positive cells, corresponding to the developing aorta and axial vein were distinguishable in the trunk region. flk-1 expression was also present in the intersomitic vessels. By contrast, flk-1 expression in runx1-mo embryos was grossly perturbed and indicated missing segments of vasculature, abnormal axial vessels, formation of atypical or ectopic structures and deficient formation of intersomitic vessels. Furthermore, transverse sections through the trunk and tail revealed dilated and disorganized vascular channels in 48 hpf runx1-mo embryos (Fig. 6C,D,G,H,K,L). These findings imply that the lack of normal circulation can be at least partially attributed to defective development of the blood vessels in the trunk. Our results suggest that normal Runx1 function is essential for vasculoangiogenesis.
runx1-mo embryos accumulate immature hematopoietic progenitors
A characteristic feature observed in runx1-mo embryos was enlargement of the ICM region (Fig. 5A,B). This was first observed at 24 hpf and persisted until at least 48 hpf. We examined the developmental status of this cell population. The hematopoietic transcription factor scl is expressed in blood-generating regions and later in circulating cells during normal embryonic development, but declines markedly before 50 hpf (Gering et al., 1998). In runx1-mo embryos, in situ hybridization revealed that cells that accumulated in the ICM region from 24 hpf were scl positive and that this scl-positive cell population remained dominant at 48 hpf (Fig. 7A,B,I,J,M,N). Consistent with the vascular phenotype of runx1-MO injected embryos, there was no evidence of scl expression in circulation (Fig. 7E,F). Because scl expression marks early differentiation of hematopoietic cells, scl-positive cells that accumulate in the ventral tail are likely to represent immature hematopoietic progenitors. This is consistent with the observed immature morphology of these cells. Microscopic analysis revealed their blast-like features characterized by a high nuclear to cytoplasmic ratio, open chromatin and scanty basophilic cytoplasm (Fig. 7C,D). Immature erythroblasts were also seen (Fig. 7G,H). Together, these results support the idea that maturation of blood progenitors is delayed or arrested in runx1-mo embryos, resulting in decreased numbers of primitive erythrocytes.
runx1-mo embryos have a block in definitive hematopoiesis
The myb gene encodes a transcription factor that promotes the differentiation of definitive hematopoietic cell types (Mucenski et al., 1991). To investigate whether runx1 is upstream of myb in the regulatory cascade, we carried out in situ hybridization on runx1-mo embryos using a myb riboprobe (Fig. 7O,P). In normal embryos at 48 hpf, myb was expressed in clusters of cells in the ventral wall of the dorsal aorta, a region equivalent to the mammalian AGM that is likely to contain definitive hematopoietic stem cells (Thompson et al., 1998). In 48 hpf runx1-mo embryos, expression of myb in the dorsal aorta was markedly reduced by comparison. Consistent with the role of Runx1 in mammals, our results imply that runx1 function is required for definitive hematopoiesis in zebrafish.
runx1-mo embryos exhibit neurological abnormalities
In addition to the vascular and hematopoietic phenotypes described above, runx1-mo embryos also exhibit a number of neurological defects (Fig. 8).
Normally, otic vesicles of developing zebrafish embryos contain two otoliths. Strikingly, we observed otic vesicles in runx1-mo embryos that had irregular numbers of otoliths (commonly three) (Fig. 8A,B), indicating that the expression of runx1 adjacent to the forming otic vesicle is probably required for its development. Some embryos also exhibited head deformities such as enlargement of the ventricles in the brain, and, more rarely, head underdevelopment (Fig. 8A,B and Fig. 5D,E respectively). In addition, notochord abnormalities were observed, including missing cells (Fig. 8C,D), waviness, angulation and shortening of the anteroposterior body axis (data not shown).
To further investigate the role of runx1 in neuropoiesis, we determined the distribution of the neuronal marker HNK-1 in runx1-mo embryos. HNK-1 marks primary sensory neurons and the axons of these neurons become strongly positive for HNK-1 as they develop (Metcalfe et al., 1990). Immunostaining with anti-HNK-1 revealed abnormal neurogenesis in runx1-mo embryos. These embryos commonly displayed disrupted axonogenesis and abnormal localization of Rohon-Beard cells (Fig. 8E,F). These results provide the first functional evidence to our knowledge that Runx1 is necessary for development of the CNS.
Overexpression of human RUNX1-CBF2T1 cDNA in zebrafish embryos reproduces abnormalities seen in runx1-mo embryos and RUNX1-CBF2T1 transgenic mice
Studies in mice have shown that expression of a RUNX1-CBF2T1 fusion gene during embryogenesis causes embryonic lethality, probably by dominant interference with normal RUNX1 function (Yergeau et al., 1997; Okuda et al., 1998). To examine the validity of zebrafish as a model for leukemogenesis, we sought to determine whether the RUNX1-CBF2T1 fusion protein could produce similar effects in zebrafish embryos. To this end, a pCS2cmv-RUNX1-CBF2T1 construct was used to drive transient expression of human RUNX1-CBF2T1 during zebrafish embryogenesis (Figs 9-11). Embryos were injected with 100-150 pg of pCS2cmv-RUNX1-CBF2T1, and the presence of human RUNX1-CBF2T1 protein was confirmed by western analysis (Fig. 9A). Abnormalities were observed in 32-41% of injected embryos at 48 hpf and comprised defective development of blood and circulation and internal hemorrhaging (Fig. 9B). Depending on the amount of pCS2cmv-RUNX1-CBF2T1 injected, 13-41% of embryos lacked normal circulation and accumulated blood cells in the aorta and ventral tail (Fig. 9E,F). Some embryos also exhibited perturbed and/or reduced circulation (Fig. 9G,H). These defects were similar to those observed in runx1-mo embryos. Hemorrhages were found in the CNS and pericardium in a subset of embryos with established circulation (Fig. 9H-N). Within the CNS, both intracerebral and intraventricular areas of bleeding were observed. Extravasation of blood in the brain was confirmed by histological analysis (Fig. 10).
To investigate the molecular effects of RUNX1-CBF2T1 on hematopoiesis, in situ hybridization using selected hematopoietic and vascular markers were performed on pCS2cmv-RUNX1-CBF2T1-injected embryos with the phenotypes described above (Fig. 11A,B,E,F,I,J). Consistent with the diminished blood circulation, the expression of hbbe3 was significantly reduced in a subset of embryos (Fig. 11A,B). In embryos displaying CNS hemorrhages, ectopic hbbe3 expression was observed in the head (data not shown). The expression of flk-1 was perturbed in embryos that lacked circulation, indicating aberrant development of the trunk vasculature (Fig. 11E,F). A marked reduction in myb expression indicates defective definitive hematopoiesis (Fig. 11I,J).
To determine the effect of RUNX1-CBF2T1 on hematopoiesis at a cellular level, we examined the cytology of blood cells collected from 48 hpf embryos injected with pCS2cmv-RUNX1-CBF2T1 (Fig. 11C,D,G,H,K,L). In these embryos, immature blood cell precursors were seen to accumulate in the ventral tail region. The immature cells had a blast morphology similar to that described for runx1-mo embryos (Fig. 11C). In addition, some early cells showed atypical features of nuclear cleavage and perinuclear cytoplasmic clearing, more characteristic of atypical myeloid than erythroid precursors (Fig. 11D). In the injected embryos, we also observed circulating erythroid cells with dysplastic features which included nuclear bridges, doughnut-shaped and irregular nuclei, binucleation and delayed maturation (Fig. 11G,H,K,L). Our results demonstrate that the expression of RUNX1-CBF2T1 in developing zebrafish embryos leads to a phenotype that shares similarities with the runx1-mo embryos, and also aligns closely with the data obtained in RUNX1-CBF2T1 knock-in mice (Yergeau et al., 1997; Okuda et al., 1998).
We have isolated a zebrafish ortholog of the runx1 gene. In this study, we describe the expression of zebrafish runx1 during early embryogenesis, and show that it is involved in hematopoiesis, vasculogenesis and neuropoiesis. Furthermore, we provide evidence that the zebrafish may be used as an alternative model for studies of t(8;21)-mediated leukemogenesis.
The zebrafish runx1 gene we isolated shares 99% nucleotide sequence identity with a previously described gene, runxa (Kataoka et al., 2000). Although Kataoka et al. obtained a similar pattern of neuronal expression, they did not observe runx1 expression in hematopoietic tissues. The probe used in their studies encompassed 483 bp of sequence 3′ of the RD and contained 248 bp that overlapped with our shorter probe. Within this region of overlap, one nucleotide differed (H. Kataoka, personal communication). When we reproduced their experimental conditions, the results were unchanged from those shown in Fig. 2. The reasons for the different hybridization patterns are unclear. However, it is possible that their probe detects an alternatively spliced exon of runx1 that is only found in neuronal but not in hematopoietic tissues, and that this probe is too short to detect hematopoietic runx1 expression. Intron/exon boundaries are present in this region in the human RUNX1 gene (Levanon et al., 2001).
The hematopoietic transcription factor SCL has previously been shown to be required for the correct differentiation of blood and vascular tissue in the zebrafish (Gering et al., 1998; Liao et al., 2000). Recent evidence suggests that SCL is likely to be directly involved in hemangioblast specification in mammalian cells (Robertson et al., 2000). Therefore, the co-expression of SCL and runx1 at 12 hpf observed in our studies suggests a role also for Runx1 at the hemangioblast level of stem cell differentiation. In addition, we found that runx1 expression was lost in the zebrafish mutants spadetail and cloche, and the forced expression of runx1 in clo resulted in partial rescue of hematopoiesis. These findings indicate that runx1 acts downstream of spadetail and cloche genes and support an involvement of runx1 in the development of blood and vasculature. This is strengthened by the morpholino data, showing that the depletion of Runx1 leads to aberrant vasculoangiogenesis as well as the accumulation of immature hematopoietic progenitors. Together, these results support the existence of a hemangioblast and argue that runx1 is involved in its development in the zebrafish.
Previously, Runx1 function was shown to be required for definitive, but not primitive hematopoiesis in mice (Okuda et al., 1996; Wang et al., 1996a). It is therefore surprising that zebrafish runx1 expression was detected so early in mesodermal cells, as they give rise to the primitive blood. Nevertheless, consistent with its presence in differentiating mesoderm, analysis of Runx1-depleted embryos supports a role for Runx1 in zebrafish primitive hematopoiesis. An expanded population of blast-like, scl-positive cells accumulated in the ventral tail of runx1-mo embryos. These cells failed to differentiate and maintained scl expression for an extended period of time (Fig. 7). Because runx1-mo embryos also displayed aberrant vasculature, it is likely that the accumulation of progenitors resulted from an inability of these cells to circulate and/or differentiate, rather than from increased proliferation. To support this, an increase in the scl-positive population was not observed in runx1-mo embryos at 14 hpf, prior to vascularization (M. K., unpublished). Therefore, we conclude that there is a delay or arrest in the maturation of primitive erythrocytes in Runx1-depleted embryos.
There is evidence that Runx1 may be involved in primitive hematopoiesis in other vertebrates. The Xaml is expressed in sites of primitive hematopoiesis and functional studies indicated its requirement for the formation of primitive blood (Tracey et al., 1998). The involvement of Runx1 in zebrafish and Xenopus primitive hematopoiesis may therefore represent an ancestral condition from which the mammalian state was derived. In the mouse, all newly emerging primitive erythrocytes transiently express Runx1 until around 10.5 dpc (North et al., 1999), but it does not appear to be required for their development. However, circumstantial evidence hints that Runx proteins may be involved in primitive hematopoiesis in mammals. CBFβ is absolutely required for Runx1 function (Wang et al., 1996b) and can heterodimerize with all three Runx proteins (Ogawa et al., 1993a; Wang et al., 1993). A dominant negative form of CBFβ resulting from the CBFβ-MYH11 fusion impaired primitive hematopoiesis in vivo (Castilla et al., 1996), implying involvement of a CBFβ-Runx complex in primitive hematopoiesis. It has been proposed that there is functional redundancy amongst runt family members, which would reconcile the absence of a defect in primitive hematopoiesis in Runx1-null mice with Runx1 expression in primitive lineages (Tracey et al., 1998).
Consistent with its role in mammals, Runx1 is required for definitive hematopoiesis in zebrafish. Zebrafish runx1 is expressed in the ventral wall of the dorsal aorta, a site associated with the development of early definitive progenitors in mice (Dzierzak and Medvinsky, 1995; North et al., 1999; Mukouyama et al., 2000). Furthermore, Runx1-depleted embryos showed a marked reduction in the expression of myb (Fig. 7).
Sites of runx1 expression coincide with the expression of zebrafish cbfb (Blake et al., 2000). Like runx1, zebrafish cbfb expression was also detected in the LPM and ICM regions. Interestingly, the neural expression of cbfb is more extensive than runx1, implying that it may interact with other members of the runt family. The neural expression of runx1 was very similar to that previously described for runxa (Kataoka et al., 2000). We observed expression of runx1 in a subset of Rohon-Beard cells, olfactory placode and the presumptive cranial nerve VIII ganglia. The RUNX1 homolog lozenge is required for olfactory sense organ and eye development in Drosophila (Flores et al., 1998; Gupta et al., 1998), and also for hematopoiesis (Lebestky et al., 2000). In addition, both mouse Runx1 and Xaml are expressed in neural structures, including neural crest and Rohon-Beard cells, respectively (Simeone et al., 1995; Tracey et al., 1998). However, a neuronal function for Runx1 had not previously been demonstrated in a vertebrate. Strikingly, Runx1-depleted zebrafish embryos showed gross abnormalities in neuropoiesis as revealed by a disorganized distribution of Rohon-Beard cells and perturbed axonogenesis (Fig. 8). In addition, Runx1 was found to be required for specification of the number of otoliths (two) in the developing otic vesicle. Embryos compromised for Runx1 function were found to have supernumerary otoliths. This is the first functional evidence that runx genes contribute to neuronal and sensory organ development in vertebrates. The CNS hemorrhages that occurred in Runx1 null mice primarily affected cranial nerves (especially VII-VIII complex) and dorsal root ganglia. Moreover, necrosis was observed in neural crest and endothelial cells (Wang et al., 1996a). As Rohon-Beard cells are developmentally related to neural crest cells (Artinger et al., 1999), the CNS hemorrhages may reflect a role for Runx1 in neurogenesis.
A zebrafish model for RUNX1-CBF2T1-mediated leukemogenesis
The RUNX1-CBF2T1 fusion gene product that results from t(8;21) in humans has been studied in mice as a model for t(8;21)-derived leukemia (Yergeau et al., 1997; Okuda et al., 1998; Rhoades et al., 2000; Yuan et al., 2001). Because the zebrafish is highly amenable to genetic and developmental studies (Amatruda and Zon, 1999), we sought to create the basis for a RUNX1-CBF2T1 leukemogenic model using zebrafish. As a starting point, we overexpressed a human RUNX1-CBF2T1 cDNA construct in zebrafish embryos.
This transient overexpression caused disorganized, reduced or absent circulation, along with internal hemorrhages (CNS and pericardial). Abnormal vascular development and defective hematopoiesis contributed to the observed defects, as revealed by aberrant flk-1 and reduced myb expression, and the presence of dysplastic and immature blood cells. Although misexpression of a heterologous fusion protein might induce artefacts that affect the observed phenotype, these abnormalities were also observed in the zebrafish runx1-mo embryos, consistent with the idea that the RUNX1-CBF2T1 fusion protein dominantly inhibits endogenous runx1 function. The phenotype generated by RUNX1-CBF2T1 in zebrafish varied with the amount of the construct injected, consistent with the observation that RUNX1 functions in a dose-dependent manner (Wang et al., 1996a; Song et al., 1999; Cai et al., 2000). Significantly, this phenotype aligns with the changes observed in RUNX1-CBF2T1 knock-in mice (Yergeau et al., 1997; Okuda et al., 1998). Furthermore, fetal liver cells from RUNX1-CBF2T1 knock-in mice form dysplastic hematopoietic progenitors with a high self-renewal capacity in vitro (Okuda et al., 1998). Consequently, it was suggested that RUNX1-CBF2T1 mediates inappropriate cell proliferation in addition to repression of normal Runx1 function. The RUNX1-CBF2T1 protein was also shown to block myeloid differentiation in vitro (Rhoades et al., 2000). Consistent with these data, we observed accumulation of immature hematopoietic precursors in the ventral tail ICM and dysplastic erythroid cells.
Because the expression of RUNX1-CBF2T1 in zebrafish generates a phenotype that mimics the equivalent situation in mice, we conclude that the zebrafish presents a valid model system for the evaluation of t(8;21)-mediated leukemogenesis. There are precedents for the use of zebrafish as a model for human hematopoietic disease with the alas-e mutation in sauternes, urod in yquem, ferrochelatase in dracula and sptb in riesling offering models for congenital sideroblastic anemia, hepatoerythropoietic porphyria, erythropoietic protoporphyria and hereditary spherocytosis, respectively (Brownlie et al., 1998; Wang et al., 1998; Childs et al., 2000; Liao et al., 2000). We believe that the zebrafish offers a promising, genetically amenable alternative to the mouse for examining the role of t(8;21) in leukemogenesis. The phenotype generated by expressing human RUNX1-CBF2T1 cDNA in zebrafish embryos represents the first demonstration of a human oncogenic protein modeled in zebrafish. To further explore this exciting possibility, we are currently generating a zebrafish transgenic line with inducible RUNX1-CBF2T1 expression. We intend to use this line in a genetic screen for molecules that modulate the effects of RUNX1-CBF2T1.
We thank Anne Bardsley for help in screening the zebrafish cDNA library, Ross Bland for advice on the in situ work, Chris Hall for facilitating microangiography and Latifa Khan for technical assistance. This study would not have been possible without the excellent fish facility run by Dr Peter Cattin, who also helped with embryo sectioning. We appreciate the help of Natalie Potter and Isobel Early with histology, Michelle Petrasich with cytology and Jeff Greenwood with Western blotting. Selected digital imaging was carried out at The University of Auckland Biomedical Imaging Research Unit with the help of Hilary Holloway. We are grateful to Dr Len Zon for providing fixed cloche and spadetail embryos, clom39 line and plasmids for the scl, flk-1, myb and hbbe3 in situ probes. Our sincere thanks go to Dr Scott Hiebert for providing the human RUNX1-CBF2T1 cDNA, Dr Hiroko Kataoka for helpful discussion and to Dr Gordon Keller for comments on the manuscript. M. K. holds a Ruth Spencer Medical Research Fellowship from the Auckland Medical Research Foundation and the New Zealand Guardian Trust. This work was also supported by the Health Research Council of New Zealand, the Leukaemia and Blood Foundation of New Zealand and grant RR10715 from NIH to J. H. P.