Disruption of asxl1 results in myeloproliferative neoplasms in zebrafish

ASXL1 is one of three mammalian homologs of the Drosophila Asx gene, which is highly conserved across multiple species (Fisher et al., 2006). Its product is an epigenetic scaffolding protein that binds to chromatin and recruits polycomb repressive complex 2 (PRC2; consisting of EZH2, EED and SUZ12), which regulates the expression pattern of developmental genes in both hematopoietic and nonhematopoietic systems (Scheuermann et al., 2010). The PRC2 complex catalyzes locus-specific trimethylation of lysine 27 on histone H3 (H3K27me3), a hallmark repressive modification that recruits the PRC1 complex (Abdel-Wahab et al., 2012a), placing further repressive modifications on chromatin by catalyzing the monoubiquitination of lysine 119 on histone H2A (H2AK119). The transcriptional repression of polycomb group (PcG) target genes by PRC1 and PRC2 is important for the maintenance of lineage-specific gene expression programs (Müller and Verrijzer, 2009; Pietersen and van Lohuizen, 2008; Schuettengruber et al., 2007; Schwartz and Pirrotta, 2007). ASXL1 also associates with the deubiquitinating enzyme BRCA1-associated protein 1 (BAP1), which removes the ubiquitin moiety from H2AK119, thereby promoting the expression of key target genes (Abdel-Wahab et al., 2012a; Scheuermann et al., 2010).

ASXL1 is mutationally altered in several malignant myeloid diseases, including myeloproliferative neoplasms (MPNs; ∼10-15%), myelodysplastic syndromes (MDS; ∼15-25%), chronic myelomonocytic leukemia (CMML; ∼45%), and de novo (6.5%) or secondary (30%) cases of acute myeloid leukemia (AML) (Abdel-Wahab et al., 2011; Bejar et al., 2011; Boultwood et al., 2010a,b; Gelsi-Boyer et al., 2012, 2010; Inoue et al., 2013). These genetic alterations comprise either focal deletions, nonsense mutations or insertions/deletions (indels) that lead to frameshifts. Mutations in this gene are consistently associated with adverse outcomes, and thus serve as independent prognostic markers (Bejar et al., 2011; Gelsi-Boyer et al., 2012, 2010). In preclinical murine models, the combined loss of Asxl1 and Tet2 from specific hematopoietic cells resulted in an accelerated onset of MDS (Abdel-Wahab et al., 2013).

Recent reports have called attention to the fact that nonsense and frameshift mutations of ASXL1 in human myeloid malignancies often truncate the protein after 404 to 800 amino acids, retaining the ASX homology domain and truncating off the remainder of this 1541 amino acid protein, including the plant homeodomain (PHD) domain (Asada et al., 2018; Balasubramani et al., 2015; Hsu et al., 2017; Inoue et al., 2016; Kitamura, 2018; Metscher and Ahlberg, 1999; Nagase et al., 2018; Yang et al., 2018). A transgenic mouse model that expresses a truncated Asxl1 protein exhibited a gain-of-function alteration and induced myeloid malignancies (Yang et al., 2018). However, a knock-in mouse expressing a truncated ASXL1 mutant did not exhibit myeloid cell deficiency or malignancy (Hsu et al., 2017), raising the question of overexpression artifacts in the transgenic mouse model. Complicating the interpretation has been a problem in raising antibodies that recognize amino-terminal epitopes of ASXL1, so that western blotting has been very difficult and detailed biochemistry impossible, even in cell lines with classic exon 11 or 12 mutations. Heterozygous ASXL1 mutations that occur in earlier exons have been documented in primary chronic myelomonocytic leukemia samples (Abdel-Wahab et al., 2011), suggesting that this subset have true loss of one allele, and thus that haploinsufficiency also can contribute to the onset of hematologic malignancy. Thus, further work needs to be done to separate the roles of haploinsufficiency from possible dominant-negative or gain-of-function activities that may result from different classes of ASXL1 mutations.

In humans, de novo constitutive heterozygous nonsense mutations in ASXL1 have been identified in patients with Bohring-Opitz syndrome, a pediatric disease associated with developmental defects. In murine models, the hematopoietic phenotypes of animals with loss of Asxl1 are variable. Fisher et al. reported that haploinsufficiency of Asxl1 caused mildly perturbed myelopoiesis, while complete knockout induced severe hematopoietic defects and perinatal lethality but did not trigger the development of hematologic malignancies (Fisher et al., 2010). In another study, animals with hematopoietic-cell-specific homozygous loss of Asxl1 developed progressive myelodysplasia culminating in MDS, an outcome that was attributed to loss of PRC2-mediated trimethylation of H3K27 (Abdel-Wahab et al., 2012a, 2013). Moreover, Asxl1 loss cooperated with activated NRAS in this model to stimulate the rapid development of AML (Abdel-Wahab et al., 2012a). Finally, Wang et al. reported that surviving mice with ubiquitous loss of Asxl1 had features of MDS, while Asxl1 heterozygotes developed a milder form of the disease (Wang et al., 2014).

Here, we established zebrafish lines with germline inactivating mutations of asxl1 early in the coding sequence of the gene and analyzed the effect of asxl1 deficiency on hematopoietic cells in both embryonic and adult zebrafish. We show that asxl1 is important for the survival of newly formed hematopoietic stem and progenitor cells (HSPCs) as they migrate into the caudal hematopoietic tissue of the zebrafish embryo, and that homozygous loss of asxl1 induces mitochondrial apoptosis of HSPCs through the activation of bim and bid. In fish with heterozygous asxl1 inactivation, half of the 5-month-old adults exhibited MPN. Among fish that were compound heterozygous for asxl1 and tet2, 80% developed MPN, indicating that haploinsufficiency for tet2 potentiates the transforming activity of asxl1 during the pathogenesis of MPN.

The zebrafish genome contains a single zebrafish ortholog of human ASXL1. Alignment of the predicted zebrafish and human ASXL protein sequences showed a high level of conservation within the important functional domains (ASXN, ASXH, ASXM1, ASXM2 and PHD) that are crucial for the ability of the protein to promote PRC2-mediated repressive chromatin alterations through H3K27 trimethylation (Abdel-Wahab et al., 2012a; Fisher et al., 2006; Scheuermann et al., 2010) (Fig. S1). To generate loss-of-function alleles of the zebrafish asxl1 gene, we designed transcription activator-like effector nucleases (TALENs) to induce premature stop codons within the sequences encoding the catalytic ASXN domain (Fig. 1A). One-cell-stage embryos were microinjected with the TALEN mRNAs, grown to maturity and subsequently outcrossed. F1 progeny were then analyzed to identify fish harboring inherited indels that disrupted the asxl1 reading frame. We then generated two independent mutant alleles of asxl1 (asxl1Δ10 and asxl1Δ11) that possess 10- and 11-base-pair deletions, respectively, within exon 2 (Fig. 1B). Both TALEN-induced deletions led to premature stop codons (Fig. 1C) that disrupted the ASXN domain and abolished the ASXH, ASXM1, ASXM2 and PHD domains (Fig. 1D). Thus, the asxl1Δ10 and asxl1Δ11 lines are functionally null for asxl1. When all planned experiments were repeated with each mutant line, similar results were obtained. For simplicity, we present only data derived from the asxl1Δ10 line, hereafter referred to as asxl1^−/−.

To determine whether asxl1 is important for the growth and viability of developing zebrafish, we examined the morphology of clutches of fish derived from crosses of two heterozygous zebrafish. We found that asxl1^+/− fish develop normally and are indistinguishable from aslxl1^+/+ fish during development, although they died prematurely starting at about 18 weeks of age (Fig. 1F). By contrast, asxl1^−/− fish were viable but appeared abnormal by 7 days post-fertilization (dpf), when 100% of mutants (27 of 27) were shorter in length and had a reduced dorsal-ventral size overall. These homozygous mutant fish did not grow by 9 dpf (Fig. 1E), and the vast majority (25 out of 27) had died by 14 dpf (Fig. 1F). To explore the death mechanism of asxl1^−/− embryos, we studied the organ development at 6 dpf and 14 dpf by histopathologic analysis. At 6 dpf, the asxl1^−/− embryos exhibited normal morphology for the muscle and intestine as compared with asxl1^+/+ embryos (Fig. 2A,D and B,E). However, by 6 dpf the liver parenchyma appeared abnormal in asxl1^−/− embryos, with atypical hepatocytes containing poorly demarcated cells with vacuolated cytoplasm (Fig. 2C,F). By 14 dpf, the asxl1^−/− zebrafish embryos showed muscular atrophy with disorganization of muscle fibers and loss of differentiation as evidenced by round and not elongated nuclei compared with the normal striated muscle fibers and elongated nuclei in the muscle of asxl1^+/+ fish (Fig. 2G,J). The intestine was also abnormal in asxl1^−/− embryos, which showed intestinal architectural atypia with significant villus blunting and disorganized localization of cell nuclei, when compared to the normal cell nuclei and microvilli in the asxl1^+/+ intestinal epithelium (Fig. 2H,K). Additionally, the 14 dpf asxl1^−/− embryos examined demonstrated progressive architectural distortion of the liver parenchyma as evident by widespread atypical hepatocytes with cell crowding and poor cell border demarcation when compared to the asxl1^+/+ embryos (Fig. 2I,L).

It seems plausible that the relative hypoplasia of the intestinal epithelium might account in part for the runted overall appearance of the juvenile asxl1^−/− fish, because the intestinal villus normally increases the mucosal surface 10-fold and the microvilli that make up the normal brush border increase the absorptive surface by 20-fold. Of course, we cannot establish the precise cause of the failure to thrive and premature death of the majority of asxl1^−/− embryos based on histology alone. However, we suspect that it can be reversible, because the small percentage of asxl1^−/− fish that survive are initially as small as the other asxl1^−/− fish, but, if they survive, they gradually attain normal size by 5 months of age and both males and females are fertile. Histopathological analysis of 17-month-old asxl1^−/− mutants revealed that the intestine and liver were normal (Fig. S2D-I), indicating that these fish had recovered from early developmental hypoplasia in these organs (Fig. 2). However, they did have decreased numbers of erythroid islands in the kidney marrow (Fig. S2A-C), indicating that the hematopoietic system was abnormal. Thus, asxl1 is almost always essential for development and viability, although rare individuals can recover normal size and survive well into adulthood.

We next asked whether asxl1 is required for normal embryonic and definitive hematopoiesis. In zebrafish with heterozygous or homozygous inactivation of asxl1, there was normal development of the embryonic erythroid, macrophage and granulocytic lineages (Fig. S3). The number of definitive HSPCs in asxl1 mutant embryos was determined by whole-mount in situ hybridization (WISH) with a cmyb riboprobe at both 36 hpf (Fig. 3A) and 3 dpf (Fig. 3B-E) on the progeny of a cross between heterozygous asxl1 mutant fish. At 36 hpf, the number of newly formed HSPCs budding from the ventral wall of the dorsal aorta appeared normal regardless of genotype (Fig. 3A); however, at 3 dpf there was an apparent decrease in the number of cells expressing cmyb in the caudal hematopoietic tissue (CHT) of the asxl1^−/− embryos (Fig. 3E), compared to both asxl1 wild-type and heterozygous embryos (Fig. 3C,D). To further confirm the cmyb WISH results shown in Fig. 3, we crossed asxl1^+/− fish with Tg(cmyb:EGFP), in which HSPCs could be visualized by EGFP. Incrosses of these fish showed a statistically significant (P=0.0002) reduction of EGFP+ cells at 3 dpf based on EGFP fluorescence (Fig. S4), similar to the results we observed by WISH at 3 dpf (Fig. 3). Because cmyb+ HSPCs were dramatically decreased in asxl1^−/− embryos between 36 hpf and 3 dpf (Fig. 3E), we postulated that the asxl1^−/− HSPCs may be undergoing apoptosis during this interval.

To address this hypothesis, we performed terminal transferase UTP nick end labeling (TUNEL) analysis coupled with immunohistochemistry (IHC) for cmyb/GFP in 48-hpf asxl1^+/+, asxl1^+/− or asxl1^−/− embryos generated in the background of the Tg(cmyb:GFP) reporter line (North et al., 2007). Compared with asxl1^+/+ and asxl1^+/− embryos, those with an asxl1^−/− genotype showed a significant increase in the number of cmyb/GFP+/TUNEL+ cells (Fig. 3F-I). These results indicate that, although asxl1 is not required for HSPC specification and budding from the ventral wall of the dorsal aorta at 36 hpf, it is important for the survival of newly formed HSPCs as they migrate into the CHT of the zebrafish embryo.

To determine whether the progeny of surviving asxl1^−/− adult fish had defects in hematopoiesis, we compared the progeny of incrosses of asxl1^+/− fish with those of incrosses of asxl1^−/− fish. Then, we performed WISH with a cmyb riboprobe at 3 dpf to assess the number of definitive HSPCs in embryos at that developmental stage. There was a similar decrease in the number of HSPCs expressing cmyb in the CHT of 3 dpf asxl1^−/− embryos, whether they arose from heterozygous (10 of 42; Fig. 4C) or homozygous (10 of 10; Fig. 4D) incrosses, as compared to wild-type (Fig. 4A) and heterozygous asxl1 (Fig. 4B) embryos.

We also tested whether a complete loss of asxl1 leads to defects in hematopoiesis in adult zebrafish, as indicated by the reduction of erythroid islands in the kidney marrow, as shown for a representative asxl1^−/− fish at 17 months of age (Fig. S2A-C). Giemsa staining of the hematopoietic cells from kidney marrow touch preps of seven asxl1^−/− zebrafish sacrificed at 16- to 17-months of age showed a complete lack of myeloid maturation with increased numbers of myeloblasts (Fig. 4E, blue arrows) when compared to wild-type fish. In the absence of secondary causes, such as toxic exposure or drug administration, arrested myeloid maturation strongly suggests a myeloid neoplasm, and the presence of increased myeloblasts supports a diagnosis of a disease resembling AML. In peripheral blood smears from asxl1^−/− fish, only rare myeloblasts were present (Fig. 4E, blue arrows), which is not an unusual feature of AML. By close examination of blood smears from two asxl1^−/− fish, we identified circulating immature myeloid blast cells in the peripheral blood of the mutant zebrafish (3 of 250 blood cells in fish 1, and 4 of 250 blood cells in fish 2) that were not evident in wild-type fish (0 in 250 blood cells in fish 1, and 0 in 250 blood cells in fish 2; Fig. 4E, blue arrows). Thus, complete loss of asxl1 leads, after 16-17 months, to a disease resembling AML in zebrafish, likely after accumulation of additional mutations or epigenetic alterations.

To investigate the mechanism by which HSPCs undergo apoptosis in 48 hpf asxl1^−/− embryos, we performed quantitative real-time PCR (qPCR) on mRNA isolated from the CHT region of 48 hpf asxl1^+/+, asxl1^+/− and asxl1^−/− zebrafish embryos (Fig. 5A,B). We analyzed expression of p53 (tp53) as well as both pro-apoptotic (puma, bim, bid, bik and bax) and anti-apoptotic (bcl2, bcl-xL and mcl1a) members of the Bcl2 family. There was significant overexpression of bim and bid in asxl1^−/− compared to asxl1^+/+ and asxl1^+/− embryos (Fig. 5A; P<0.05) but no change in p53, puma, bik or bax expression. Additionally, we observed a significant decrease in bcl2 expression but no change in the levels of bcl-xL or mcl1a in asxl1^−/− embryos (Fig. 5B). These data suggest that changes in bim, bid and bcl2 mediate HSPC apoptosis in 48 hpf asxl1^−/− embryos, thereby leading to loss of cmyb+ cells in the CHT at 3 dpf.

To test whether bim is required for the loss of asxl1^−/− HSPCs, we used a bim-mutant zebrafish line (bcl2l11zdf19) that was generated by retroviral insertional mutagenesis (Golling et al., 2002). This line (hereafter referred to as the bim mutant line) harbors a retroviral insertion located within the coding sequence of bim exon 2 and lies upstream of the BH3 domain required to induce apoptosis, thereby creating a loss-of-function allele of bim. We crossed the bim mutant line into our asxl1^+/− zebrafish line and analyzed the progeny for cmyb expression by WISH at 3 dpf (Fig. 5C,D). Our data show that bim is required for the loss of cmyb+ HSPCs in the CHT of asxl1^−/− zebrafish at 3 dpf. We next asked whether bid is required for apoptosis in 48 hpf asxl1^−/− zebrafish HSPCs. Thus, after incrossing asxl1^+/− adult zebrafish and injecting the progeny with a bid-specific splice-blocking morpholino (MO) (Kratz et al., 2006; Pyati et al., 2011), we performed WISH for cmyb at 3 dpf, subsequently genotyped the embryos for the asxl1 mutation and quantified the results of the cmyb WISH (Fig. 5E,F). Remarkably, knockdown of bid also partially rescued the loss of cmyb+ HSPCs in the CHT of asxl1^−/− zebrafish at 3 dpf.

To determine whether the decrease in bcl2 expression could contribute to the apoptosis observed in 48 hpf asxl1^−/− HSPCs, we incrossed asxl1^+/− adult zebrafish and injected the progeny with mRNA encoding bcl2 or a control gene. We again performed WISH for cmyb at 3 dpf, genotyped the embryos and quantified the results (Fig. 5G,H). We found that overexpression of bcl2 rescued most of the apoptotic asxl1^−/− HSPCs. To show these findings by confocal fluorescence microscopy, we repeated the rescue assay in asxl1^−/−, asxl1^+/− and asxl1^+/+ embryos bred into the Tg(cmyb:EGFP) background, in which HSPCs could be visualized by expression of EGFP. Statistically significant partial rescue of EGFP+ HSPCs was observed at 3 dpf in asxl1^−/− embryos injected with each of the bim MO and bid MO, and also with bcl2 mRNA (Fig. S5). This experiment indicated that the apoptosis that occurred shortly after budding of asxl1^−/− HSPCs from the hemogenic endothelium was due to the combined action of the BH3-only proteins Bim and Bid, and that it could be blocked by overexpression of bcl2. Thus, death of asxl1^−/− HSPCs occurred through programmed cell death mediated by the intrinsic mitochondrial apoptotic pathway.

We recently described a zebrafish model of MDS in which both tet2^−/− and tet2^+/− fish develop trilineage dysplasia at 11 months of age and full-blown MDS with anemia by 24 months (Gjini et al., 2015). In human myeloid malignancies, inactivating mutations of ASXL1 and TET2 often occur together in the same patient's malignant cells (Abdel-Wahab et al., 2012b; Patnaik et al., 2016), suggesting that loss of these two tumor suppressors may act synergistically in myeloid transformation. To pursue this hypothesis, we intercrossed the tet2^+/− and asxl1^+/− lines and then inbred the compound heterozygous progeny to generate each of the genotypes that was relevant to our experiment (asxl1^+/+tet2^+/+, asxl1^+/−tet2^+/+, asxl1^+/−tet2^+/− and asxl1^+/−tet2^−/−). These progenies were identified through genotyping at 2 months of age and were closely monitored for survival. At 5 months of age, May–Grünwald–Giemsa (MGG) staining of kidney marrow and peripheral blood smears representing 9-11 individual fish per genotype (Fig. 6A-H) revealed an increase in the number of myelomonocytes (Fig. 6A-D, light-blue arrows) in the kidney marrow of a subset of animals with heterozygosity for asxl1 [asxl1^+/−tet2^+/+ (5 of 11); asxl1^+/−tet2^+/− (8 of 11) and asxl1^+/−tet2^−/− (3 of 10)], which is diagnostic of MPN. Moreover, 2 of 10 asxl1^+/−tet2^−/− fish had high numbers of immature myeloid cells lacking differentiation, which is indicative of AML (Fig. 6D, black arrows). Analysis of peripheral blood smears showed an aberrant increase in circulating immature red blood cells (Fig. 6E-H, orange arrows) in a subset of animals heterozygous for asxl1 [asxl1^+/−tet2^+/+ (5 of 11), asxl1^+/−tet2^+/− (8 of 11) and asxl1^+/−tet2^−/− (3 of 10)]. These results demonstrate the rapid onset of MPN in fish with heterozygous loss of asxl1 and tet2, with AML apparently being restricted to asxl1^+/− fish with loss of both tet2 alleles.

To assess the consequences of asxl1 heterozygosity on hematopoiesis more rigorously, we quantified 4 major categories of blood cell populations (erythrocytes, lymphocytes, myelomonocytes and progenitor cells) with respect to their relative numbers in the kidney marrow and their absolute numbers per liter of peripheral blood. Forward- and side-scatter flow cytometry was performed individually for wild type or asxl1 heterozygous mutants, alone or in combination with hetero- or homozygosity for tet2 (i.e. 6 genotypes, 9-11 fish per group). Analysis of the relative numbers of each cell type in the kidney marrow (Fig. 6I-L and Fig. S6) revealed a significant decrease in the erythroid cell fraction in asxl1^+/−tet2^+/+ and asxl1^+/−tet2^+/− fish, but not in asxl1^+/−tet2^−/− fish, compared to wild type (Fig. 6I). The lymphocyte fraction, by contrast, did not differ appreciably, consistent with findings in Fig. 6A-D. Myelomonocytes increased significantly in asxl1^+/−tet2^+/+, asxl1^+/−tet2^+/− and asxl1^+/−tet2^−/− fish compared to wild type (Fig. 6K). Finally, there was a significant decrease in the progenitor cell fraction of asxl1^+/−tet2^+/− fish but not in those with other genotypes, compared to wild type (Fig. 6L). Analysis of absolute cell counts in peripheral blood for each blood cell type (Fig. 6M-P) supported the results based on cell fraction, with contribution of homozygous loss of tet2 becoming more evident. Thus, regardless of tet2 status, heterozygosity for asxl1 consistently increased myelomonocytes in the kidney marrow and lowered absolute numbers of circulating erythrocytes, both characteristic features of human MPN. Notably, however, homozygous loss of tet2 leads to AML in 20% of asxl1 heterozygotes, suggesting a synergistic interaction between these two genes when they are both mutated in the same myeloid progenitors.

The ASXL1 gene is often somatically mutated in patients with hematopoietic malignancies such as MDS, MPN and AML, generally through acquired heterozygous nonsense or frame-shift mutations (Abdel-Wahab et al., 2011; Bejar et al., 2011; Boultwood et al., 2010a,b; Gelsi-Boyer et al., 2012, 2010; Inoue et al., 2013). Here, we report a zebrafish model with germline asxl1 inactivation and the unique property that half of the fish with heterozygous mutations develop MPNs by 5 months of age, characterized by increased numbers of myelomonocytes in the kidney marrow and peripheral blood together with anemia. In this context, our preclinical model shows the considerable consequences in vivo of haploinsufficiency for a true null allele of asxl1 in terms of myeloid malignancy, because the mutations we have made truncate the protein after amino acids 39 or 62, which is upstream of or internal to each of the conserved domains of ASXL1 (see Fig. 1D). Based on recent publications suggesting that an amino-terminal portion of ASXL1 with dominant-negative or neomorphic activity could be expressed in some patients with hematologic malignancy (Asada et al., 2018; Balasubramani et al., 2015; Hsu et al., 2017; Inoue et al., 2016; Kitamura, 2018; Nagase et al., 2018; Yang et al., 2018), it will also be important in the future to produce zebrafish models that introduce asxl1 frame-shift mutations and premature termination in exon 11 or early in exon 12 to assess additional phenotypes that may accompany a truncated protein arising from this gene.

Abdel-Wahab and colleagues (2013) used an elegant conditional Asxl1 knockout approach inserting two loxP sites that flanked exons 5-10 to generate their murine model. Hematopoietic-cell-specific homozygous deletion of Asxl1 resulted in MDS characterized by progressive, multilineage cytopenias and dysplasia accompanied by increased numbers of less differentiated hematopoietic stem and progenitor cells. The MDS phenotype in Asxl1 haploinsufficient mice at 6-12 months (Abdel-Wahab et al., 2013) was different from the very penetrant and early MPN phenotype (5 months) that we observed in the asxl1 mutant zebrafish model. Thus, subtle differences exist in the myeloid phenotypes and possibly the affected downstream pathways disrupted by asxl1 haploinsufficiency in these two vertebrate models of myeloid malignancy. The disease was transplantable, with a shorter latency for myeloid malignancies with Asxl1 null hematopoietic cells versus those with haploinsufficiency. ASXL1 loss in this model leads to a global reduction of PRC2-mediated trimethylation of H3K27 and thus causes dysregulated expression of key regulators of hematopoiesis, including HOXA9 (Abdel-Wahab et al., 2013). In results similar to those of Abdel-Wahab et al. in mice (Abdel-Wahab et al., 2013), we show that tet2 and asxl1 cooperate in the pathogenesis of myeloid malignancies in zebrafish, although the phenotypes in zebrafish that we report are focused on MPN progressing to AML, rather than progressive MDS as observed in the murine model.

Wang and colleagues (2014) developed a constitutive Asxl1 murine knockout line by disrupting the translation start site of Asxl1. A small fraction of the homozygous knockout mice survived for a maximum of 18-42 days, and some of these developed MDS with anemia, thrombocytopenia and neutropenia. However, heterozygous Asxl1^+/− animals in this model developed only minimal abnormalities in myeloid cell morphology without significant alterations in blood counts, similar to the constitutive Asxl1 knockout mice reported by Fisher et al. (2010) and Wang et al. (2014). Thus, haploinsufficiency for ASXL1, the main acquired genetic abnormality in human myeloid malignancies, does not by itself seem to reproducibly induce hematopoietic malignancies in mice.

Homozygous loss of asxl1 in our zebrafish model caused profound hypoplasia of the gastrointestinal system, with underdeveloped epithelium, resulting in very small larvae and near-complete lethality by 14 dpf. Interestingly, about 8% of the asxl1^−/− zebrafish larvae in our study survived, regained normal size and became fertile by 3 months of age. Thus, in the zebrafish, the total absence of asxl1 appears to delay or prevent the emergence of pathways important for gastrointestinal development but, with time, these pathways are restored in a small fraction of the fish, enabling their unimpeded progression to normal adulthood. The hematopoietic phenotype of adult asxl1^−/− fish was profound in that all of the surviving animals that we examined had kidney marrow morphology suggestive of AML by 17 months of age.

We did not observe craniofacial abnormalities or anophthalmia, both of which have been reported in the murine models with Asxl1 deletion. For example, homozygous inactivation of Asxl1 by Abdel-Wahab and colleagues (2013) in nearly all tissues with EIIa-cre caused embryonic lethality and craniofacial abnormalities, including microophthalima/anophthalmia. Similarly, Wang and colleagues (2014) reported that homozygous constitutive inactivation of Asxl1 in mice led to dwarfism and anophthalmia. Fisher et al. (2010) also developed a constitutive loss-of-function murine model, with insertion of a pgk promoter-driven neomycin-resistance cassette into exon 5. In retrospect, this alteration appears to be hypomorphic, as homozygous insertion produced only modest homeotic transformation of the axial skeleton and runting.

We observed that asxl1^−/− fish have defects in the earliest phases of definitive hematopoiesis, in that homozygous loss of asxl1 led to reduced numbers of HSPCs by 72 hpf. We also noted that specification of definitive HSCs from the ventral wall of the dorsal aorta occurs normally in asxl1^−/− fish, whereas the newly formed HSCs rapidly succumb to apoptosis, such that only 40% of normal numbers remain by 72 hpf. Importantly, the increased apoptosis that we observed is mediated through the mitochondrial pathway, as it can be rescued by Bcl2. We further show that the death of migratory HSPCs is induced by upregulation of the BH3-only pro-death proteins Bim and Bid, as knockout of either of these genes could partially rescue the affected HSPCs in asxl1^−/− animals.

Because of the optical clarity of zebrafish embryos and juvenile fish, the asxl1 mutant zebrafish lines reported here provide definite imaging advantages for assessing the pathways involved in the initiation and maintenance of MPNs arising from this deletion. These same advantages permit the analysis of chemical libraries of US Food and Drug Administration (FDA)-approved drugs for their ability to inhibit zebrafish HSPCs in order to discover drugs that are synthetic lethal with asxl1 loss for the suppression of HSPC growth and survival. A very successful screen of HSPCs in zebrafish embryos by North and coworkers (2007) identified PGEII as a promoter and indomethacin as a suppressor of the growth of HSPCs in normal embryos. This same approach could be applied to identify drugs that specifically suppress HSPCs with haploinsufficiency for ASXL1. Such drugs might prove active in human MPN cases with ASXL1-inactivating mutations. Moreover, clonal hematopoiesis, arising in 10% of individuals over 65 years of age, has been linked to clonal expansion of HSPCs with mutations of several genes, including the epigenetic modifiers DNMT3A, TET2 and ASXL1 (Genovese et al., 2014; Jaiswal et al., 2014). Importantly, this age-related defect predisposes to the development of hematologic malignancy (Corces-Zimmerman et al., 2014; Hirsch et al., 2016; Shlush et al., 2014) and more recently was related to a doubling of the risk of coronary artery disease (Jaiswal et al., 2017). Currently, there is a lack of safe and efficient drugs for inhibiting mutant but not wild-type HSPCs. The zebrafish model described here could be exploited to identify small-molecule drugs already in human use that could be repurposed for the suppression of ASXL1 mutant HSPCs and their progeny.

Wild-type stocks of AB fish, and transgenic and mutant lines were maintained according to a previously reported protocol (Bolli et al., 2010). Animal handling was approved by the Dana-Farber Institutional Animal Care and Use Committee. The Tg(cmyb-EGFP) line is described elsewhere (North et al., 2009), as are the details of a developmental staging system (Metscher and Ahlberg, 1999). For all experiments, zebrafish embryos were cultured in ‘egg water’ consisting of 0.03% sea salt and 0.002% Methylene Blue as a fungicide. To inhibit pigment formation and facilitate in situ hybridization, we incubated embryos with 0.0045% 1-phenyl-2-thiourea (Sigma).

The pair of TALENs recognizing exon 2 of the zebrafish asxl1 gene was designed with ZiFiT Targeter software (http://zifit.partners.org/ZiFiT/), and the TAL effector repeats were constructed by the ‘unit assembly’ method described previously (Sander et al., 2011). TALEN mRNA was synthesized by in vitro transcription using the SP6 mMESSAGE mMACHINE Kit (Ambion). Approximately 50 pg of mRNAs encoding each of the two TALENs were injected into 1-cell-stage zebrafish embryos. The genomic DNA was isolated from a mixture of 3 embryos at 48 hpf, and the fragments containing the TALEN target site were amplified by PCR and sequenced with specific primers to examine the efficiency of the TALENs.

P53 morpholino, bid morpholino and control morpholinos were purchased from Gene Tools LLC (all the sequences are listed in Table S1). Capped mRNAs (bcl2-mCherry) were transcribed from linearized PCS2+ plasmids (mMessage Machine; Ambion), purified, and diluted to 100 ng/ml for injection at the 1-cell stage of development.

Digoxigenin-labeled RNA probes were transcribed using linearized constructs with T3 or T7 polymerase (Ambion). Embryos at the desired time points were fixed overnight in 4% paraformaldehyde (PFA) at 4°C. Whole-mount in situ hybridization (WISH) was performed as described (North et al., 2009).

Adult fish were anesthetized with 0.02% tricaine before kidney removal. The kidney was dissected and placed in ice-cold 0.9×PBS containing 5% fetal calf serum (FCS). Single-cell suspensions were generated by aspiration, followed by mild teasing of the kidney on a 40-μm nylon mesh filter with a pipette tip. Blood cell populations were analyzed on a BD FACS Aria with high forward scatter/side scatter (FSC/SSC), as previously described (Traver et al., 2003). Data analyses were performed with FlowJo software (TreeStar, Ashland, OR, USA). Statistical analysis was done with Prism software with unpaired Student's t-tests to determine the P-value for each genotype group compared with controls.

Adult fish were anesthetized with 0.02% tricaine, and 1 µl peripheral blood was obtained by cardiac puncture diluted in 499 µl of 0.9×PBS containing 5% FCS. Ten microliters of heparin sodium (1000 units/ml) was added to prevent blood coagulation, and 20 µl of Flow-Check fluorospheres (10⁶ fluorospheres/ml; Beckman Coulter, Inc.) was added to this solution. The number of cells corresponding to 1000 counted fluorospheres was obtained with use of the BDFACSAria with high FSC/SSC, as previously described (Gjini et al., 2015). Data analyses were performed with FlowJo software (TreeStar).

RNA from the clipped caudal hematopoietic tissue region from asxl1^+/+, asxl1^+/− and asxl1^−/− embryos was extracted with TRIZOL reagent (Invitrogen). cDNA was synthesized with Superscript III (Invitrogen), while quantitative real-time PCR (qPCR) was carried out on a ViiA 7 real-time PCR system using SYBR Green. The fold change was calculated by the ΔΔCT method normalized to β-actin (see Table S1 for the primer sequences).

The kidney was dissected and smeared on glass slides, which were then fixed and stained with MGG stain (Sigma-Aldrich) according to the manufacturer's instructions and visualized with an Olympus BX51 microscope (Olympus).

TUNEL was performed with the In-Situ Cell Death Detection kit (POD: Roche) according to the manufacturer's recommendations. Phosphorylated histone H3 labeling of fixed embryos was performed with the rabbit anti-phosphohistone H3 antibody (Santa Cruz) at 4°C overnight and visualized with Alexa-Fluor-488 goat anti-rabbit secondary antibody (Invitrogen).

Images of zebrafish immunofluorescence staining or live transgenic embryos were taken with a Leica SP5X scanning confocal microscope. The embryos were mounted in 1% low-melt agarose and the confocal images were captured with 20× objective. Fluorescence-positive cells were counted in each individual slice and sum numbers were analyzed with the GraphPad Prism 7 software using the two-tailed Student's t-test. The optical slice thickness is 3 µm.

For embryo stages and adult stages, asxl1 mutant fish were individually genotyped using one forward (5′-GGTGAATGTCTTTGCCGTTC-3′) and one reverse (5′-GAGAGTGAAGCATGGTGACAAG-3′) primer. The zebrafish bim mutant line (insertion mutation) was crossed with the asxl1 mutant line in the rescue experiment. Genotyping primers are listed in Table S1.

We thank Cicely Jette and John Gilbert for editorial assistance and critical comments.