Understanding the regulation of vertebrate hematopoiesis and blood disorders: big lessons from a small fish

Abstract

Hematopoietic stem cells (HSCs) give rise to all differentiated blood cells. Understanding the mechanisms that regulate self-renewal and lineage specification of HSCs is key for developing treatments for many human diseases. Zebrafish have emerged as an excellent model for studying vertebrate hematopoiesis. This review will highlight the unique strengths of zebrafish and important findings that have emerged from studies of blood development and disorders using this system. We discuss recent advances in our understanding of hematopoiesis, including the origin of HSCs, molecular control of their development, and key signaling pathways involved in their regulation. We highlight significant findings from zebrafish models of blood disorders, and discuss their application for investigating stem cell dysfunction in disease and for developing new therapeutics.

Introduction

Over the last few decades, the zebrafish has emerged as a useful and versatile vertebrate model, bridging the gap between invertebrates and mammalian systems. The ability to study complex molecular and cellular events and importantly the onset and progression of pathological processes in a whole organism favors the use of zebrafish over in vitro cell culture settings and mammalian systems in many areas of research. This has been particularly valuable for large-scale and high-throughput mutagenesis screens and drug discovery approaches. Since the earliest forward genetic screens that identified multiple mutants with defects in hematopoietic development^1–3, there have been a growing number of publications describing the application of zebrafish for studying hematopoiesis. Despite 400 million years of divergence between zebrafish and humans, the development and function of the blood system is remarkably conserved, and the ability to modify the genome of zebrafish embryos with relative ease has allowed the identification and functional investigation of many important genes that are involved in hematopoietic development and disorders. The Danio rerio Sequencing Project has assisted in the identification of mutations causing disease and recent advances in genome editing technologies such as zinc finger nucleases (ZFNs), TALENs and the CRISPR/Cas9 system have allowed specific targeting of genes of interest to generate mutant lines and disease models. Hematological disorders such as myelodysplastic syndromes and acute leukemias can be modeled in zebrafish and have provided new insight into the molecular aspects of these diseases. Owing to the external fertilization and rapid development of small, optically transparent embryos, combined with the relatively straightforward engineering of transgenic reporter lines, zebrafish have enabled real-time in vivo visualization of blood cell emergence, migration, behavior and lineage analysis at the single cell level. This review will highlight some of the ways that the unique attributes of the zebrafish model have led to valuable discoveries in developmental hematopoiesis over the past three decades.

Molecular control of hematopoiesis in the zebrafish

The development of all fully differentiated hematopoietic cells involves a series of lineage decisions that are driven by the activity of multiple transcription factors. These function in a coordinated manner to control the activation and repression of distinct gene programs, which ultimately determine cell fate. The zebrafish is a particularly tractable system for studying the molecular mechanisms that regulate hematopoiesis and has enabled the discovery and investigation of many of the key genes involved in blood development. Vertebrate hematopoiesis proceeds in two waves: the first wave gives rise to a transient population of primitive blood cells that supports the embryo throughout early development, and the second wave produces definitive hematopoietic stem cells (HSCs)⁴. These unique cells are defined by their capacity to both self-renew and differentiate into all blood lineages. Hematopoietic development is controlled in a temporal, spatial, and molecular manner, and despite subtle differences in their origin, the genetic programs that regulate each distinct wave are highly conserved between zebrafish and mammals⁵.

Early specification of the zebrafish hematopoietic system

The entire hematopoietic system descends from distinct regions in the ventral and lateral mesoderm, which also give rise to the endothelial and cardiac lineages. The zebrafish draculin (drl) gene is the earliest marker of blood progenitor cells and can be detected as early as 9 hours post fertilization (hpf)⁶. Expression of the basic helix-loop-helix (bHLH) transcription factor scl, its LIM-domain containing binding partner lmo2, and the GATA binding factor gata2 can be observed in blood progenitors shortly after gastrulation⁷. These cells also express the vascular markers kdrl and fli1 and have become accepted to represent the zebrafish hemangioblast, as they have the potential to become specified into either hematopoietic or endothelial cells. The existence of the hemangioblast has been postulated for many years, but it still under debate as to whether one precursor gives rise to both blood and endothelial lineages, or if two precursors coexist. The concept was first introduced based on histological studies in chick embryos in the early 1920s, which described the presence of erythroblasts in close association with a network of endothelial cells, forming ‘blood islands’^8,9. In the 1990s, a clonal bipotent progenitor was identified in embryonic stem cell cultures, which was believed to be the in vitro equivalent of the hemangioblast¹⁰. In vivo studies demonstrating that single cells can give rise to both endothelial and hematopoietic cells have also supported the presence of the hemangioblast^11–14, however, some evidence indicates that the blood islands contain a more heterogeneous cell population in which progenitors cells are committed to either hematopoietic or vascular fate^15,16. In parallel, studies have described a linear pathway in which a subpopulation of hemangioblast cells give rise to blood via an intermediate cell type with hemogenic potential¹⁷. Cell fate mapping has confirmed the existence of individual bipotent cells in zebrafish; however, not all hematopoietic and endothelial progenitors appear to arise from these cells¹⁸. Further studies are required to establish the importance and relative contribution of bipotent hemangioblast cells to both hematopoiesis and vasculogenesis.

The transcription factors Scl, Lmo2 and Gata2 play particularly important roles throughout hematopoietic development in context- and temporally-specific manners, and their functions are highly conserved⁵. As in mammalian systems^19–23, genetic depletion of either scl or lmo2 in zebrafish embryos results in a loss of both primitive and definitive hematopoietic cell lineages and defects in angiogenesis, indicating a requirement for these transcription factors in hemangioblast specification^7,24. Correspondingly, ectopic expression of scl leads to increased production of hematopoietic and endothelial precursors, which occurs at the expense of somitic and pronephric duct tissues^7,25,26. Interestingly, scl expression is reduced in the zebrafish cloche mutant²⁷, which is characterized by defects in blood and endothelial cell development¹². This mutant was identified in a large-scale mutagenesis screen but cloning the underlying gene was a significant challenge for researchers. Studies have indicated that the cloche mutation may act at the level of the hemangioblast, as the expression levels of many genes important for downstream hematopoietic and vascular development are affected in addition to scl, including gata1, kdrl, fli1, hhex and etv2^{12,24,28–31}. Very recently, after more than 20 years, the cloche gene has finally been revealed to encode the PAS-domain-containing bHLH transcription factor Npas4l, which acts upstream of Scl and Etv2³². The discovery of this master regulator of endothelial and hematopoietic fates will facilitate the identification of additional regulators important in early blood development and may be fundamental for future efforts to generate HSCs from pluripotent stem cells in vitro for regenerative medicine.

Mammalian studies have revealed a number of other genes that are required for hemangioblast specification including Flk1, Hhex and Etv2, deletion of which results in blood and endothelial abnormalities^13,33,34. In zebrafish, knockdown experiments suggest that etv2 plays an important role in vasculogenesis but is not required for hematopoiesis^31,35. When hhex is overexpressed in zebrafish embryos, there is an increase in blood and endothelial genes, but its knockdown does not affect vasculogenesis or hematopoiesis, indicating that it is not essential for formation of the hemangioblast³⁰. The zebrafish homolog of Flk1, kdrl, encodes the vascular endothelial growth factor (Vegf) receptor. Overexpression of vegf mRNA in zebrafish embryos results in the formation of ectopic vasculature and blood cells³⁶. However, kdrl knockdown leads to defects in vascular development without affecting hematopoiesis^37–39, suggesting that kdrl is not required at the level of the hemangioblast in zebrafish or that there may be a mechanism for compensation. Precisely how all of these genes interact to control hematopoiesis and vasculogenesis during early development remains unclear.

Primitive hematopoiesis

Primitive hematopoiesis occurs in the blood islands of the yolk sac in mammalian systems⁴⁰, and in the anterior and posterior lateral mesoderm in zebrafish⁴¹. In the anterior section, hemangioblast cells give rise to a primitive myeloid population that expresses pu.1, and these cells can be observed as they migrate across the yolk sac before maturing into either mpx-expressing neutrophils or l-plastin-expressing macrophages^42–44. This cell fate decision is determined downstream of pu.1 by the transcription factor Interferon regulatory factor-8 (Irf8), which promotes macrophage development⁴⁵. Other transcriptional regulators, including the Bmp receptor Alk8⁴⁶ and the co-repressors Ncor1 and Ncor2⁴⁷ are essential for pu.1 expression and primitive myelopoiesis.

At around 12 hpf, the posterior lateral mesoderm forms the intermediate cell mass (ICM), and primitive erythroid cells are specified⁴⁸. These cells express gata1 and begin circulating around 24 hpf. This transcription factor is the master regulator of erythropoiesis and is essential for the generation of erythroid progenitor cells during both primitive and definitive hematopoiesis. Expression of gata1 is controlled upstream by Homeobox transcription factors such as Meis1 and Pbx⁴⁹. When these factors are depleted, expression of gata1 and embryonic hemoglobin (hbae3) is down-regulated and the number of visible circulating blood cells is reduced⁴⁹. The zebrafish vlad tepes mutant, identified in a large-scale forward genetic screen³, carries a nonsense mutation in the gata1 gene, which results in translation of a truncated protein without DNA binding or transactivation activity⁵⁰. A hypomorphic gata1 mutation has also been described in zebrafish (gata1^T301K/T301K), which causes defects in primitive erythropoiesis whilst definitive hematopoiesis is unaffected⁵¹. By combining these two distinct gata1 mutants to generate an allelic series, it was possible to demonstrate that higher levels of Gata1 are required for erythropoiesis during primitive hematopoiesis compared to definitive blood development⁵¹.

Cell fate is not only determined by transcription factors, but also by the coordinated activity of co-regulators and chromatin factors. Researchers have recently begun using zebrafish to study the epigenetics of hematopoietic development. Our laboratory conducted a large-scale reverse genetic screen targeting over 400 chromatin factors in zebrafish, and found a number that regulated hematopoiesis, many of which were not previously implicated in blood development⁵². We identified smarca1, chrac1, actr2b and hdac9b as important regulators of primitive erythropoiesis, as knockdown of these chromatin factors reduced gata1 expression.

A transient population of bipotent erythromyeloid progenitors (EMPs) also emerges from the ICM around 24 hpf⁵³. These are the first definitive blood precursors, and their fate is controlled by cross-antagonism between lineage specific transcription factors: gata1 is required for erythroid differentiation whilst pu.1 is required for myeloid differentiation^54,55. Loss of function of either critical transcription factor in zebrafish embryos results in a fate change towards the alternative lineage, for example, knockdown of gata1 induces expansion of granulocytes at the expense of erythrocytes⁵⁵. Despite their differentiation potential, EMPs are distinct from definitive HSCs and are unaffected by a loss of Notch signaling⁵⁶, in contrast to HSCs that are absent in Notch mutants⁵⁷. In addition, EMPs can differentiate into definitive erythroid and myeloid colonies in vitro but do not possess lymphoid potential⁵³, even when cultured in media that supports lymphoid differentiation⁵⁸. They are also unable to generate lymphoid cells upon in vivo transplantation⁵³.

Transcriptional control of HSC specification

The first definitive, multi-lineage HSCs that can give rise to all differentiated blood cells are specified around 30 hpf from a specialized region in the ventral wall of the dorsal aorta, the hemogenic endothelium⁵⁹. This area is equivalent to the mammalian AGM. Live imaging of transgenic zebrafish lines has uniquely enabled the visualization of endothelial cells in the AGM as they acquire hematopoietic potential. These cells begin to express the key definitive hematopoietic transcription factor, runx1, and undergo budding from the vessel wall into the circulation, a process known as endothelial-to-hematopoietic transition (EHT)^60–62. Runx1 is one of the earliest known markers of HSCs in zebrafish^62,63. This factor is a critical regulator of EHT and is absolutely required for the emergence of definitive HSCs across species^64–68. In the absence of runx1, the process of EHT is aborted and HSCs do not survive in zebrafish embryos⁶⁰. The precise molecular dynamics by which runx1 controls EHT remain unknown.

A number of other transcription factors are fundamental for specification of hemogenic endothelium and the subsequent emergence of definitive HSCs (Figure 1). As described above, scl is required for definitive blood development^7,19–21,24. Interestingly, due to a whole genome duplication event, there are two distinct isoforms of scl in zebrafish and these are differentially required during definitive hematopoiesis⁶⁹. The sclβ isoform is expressed in the subset of endothelial cells in the AGM that gain hematopoietic potential and undergo EHT, whereas expression of sclα is not observed until later, in budding HSCs⁷⁰. It is thought that sclα is important for the maintenance of HSCs in the AGM, whilst scl is required for specification of hemogenic endothelium. Zebrafish studies have revealed an important role for adenosine signaling in the regulation of sclβ-positive endothelium⁷¹. Adenosine signals via the A2b receptor and a cAMP-PKA-dependent pathway to mediate production of the chemokine Cxcl8 and Sclβ expression in the hemogenic endothelium, which promotes HSC emergence. Scl is directly involved in EHT as part of a Hedgehog-Notch-Scl signaling axis, in which Hedgehog is required for hemogenic endothelial patterning⁷². Mutations in the Hedgehog pathway result in defects in definitive HSC production but do not affect primitive hematopoiesis⁷³. Notch has been shown to play an important role in specification of hemogenic endothelium by controlling expression of runx1. The zebrafish notch signaling mutant, mindbomb, is unable to specify HSCs but has normal primitive blood and EMP production^56,57. The mutant phenotype can be rescued by transient Notch activation, which induces runx1 expression⁵⁷. Expression of constitutively active Notch has been shown to increase HSC budding events⁷⁴.

Figure 1. Transcriptional control of hematopoietic stem cell development in zebrafish.

The zebrafish hemangioblast is derived from the lateral mesoderm, and further differentiates into endothelial cells, the hemogenic endothelium, or primitive hematopoietic cells. Hematopoietic stem cells are specified from hemogenic endothelium prior to budding. Key transcription factors that are important for each stage of development are indicated.

In addition to the Hedgehog pathway, Notch appears to be regulated via a number of distinct mechanisms that are dependent on non-autonomous signals from other cell types, all of which can impact the emergence of HSCs. Recent studies have highlighted an important role for the pro-inflammatory signaling molecule TNFα, which is released from primitive neutrophils and regulates HSC emergence via activation of Notch and NFκB⁷⁵. Secreted growth factors such as angiopoietin-like proteins mediate cleavage of the Notch receptor to induce its activation and transcription of downstream target genes⁷⁴. Notch is also regulated at the epigenetic level and it has been shown that the 5-methylcytosine dioxygenases Tet2 and Tet3, which actively demethylate DNA, are both necessary to activate Notch and promote the hemogenic potential of aortic endothelium⁷⁶. The effect of these enzymes is finely tuned to control Notch signaling specifically in the ventral wall of the dorsal aorta, highlighting their importance for the emergence of HSCs. Zebrafish studies have also revealed a requirement for non-canonical Wnt16 signaling, which indirectly regulates HSC emergence by inducing somite-derived expression of Notch ligands⁷⁷. Notch signaling is required to establish HSC fate much earlier in development than initially appreciated and prior to AGM formation, as precursors migrate over the ventral somite, and this is dependent on the junctional adhesion molecules Jam1a and Jam2a⁷⁸. The downstream mechanisms by which Notch induces runx1 expression in nascent HSCs are not well understood but requirements for the forkhead box transcription factor Foxc2⁷⁹, the proto-oncogene Myc⁷⁴ and the histone deacetylase Hdac1⁸⁰ have been demonstrated.

GATA2 is another key regulator of SCL expression and its essential role in both primitive and definitive hematopoiesis has been demonstrated using Gata2^−/− mice⁸¹. Zebrafish express two GATA2 orthologs, gata2a and gata2b, which differ in their function and expression pattern, suggesting subfunctionalization of the ancestral gene⁸². Zebrafish embryos with a gata2a mutation exhibit defects in endothelial integrity and circulation indicating a role in vascular development⁸³, whereas expression of gata2b is restricted to a subset of endothelial cells and is required for the generation of definitive HSCs⁸². In mice, Gata2 is known to directly regulate the hemogenic endothelial-specific Runx1 enhancer⁸⁴ and is essential for HSC survival⁸⁵. Recently, zebrafish gata2b has been identified as a specific marker of hemogenic endothelium and a key regulator of EHT upstream of runx1⁸². In gata2b deficient embryos, runx1 expression in the AGM is reduced, and this is reflected in mammalian systems, in which targeted deletion of either Gata2 or a critical upstream enhancer element, specifically in endothelial cells, results in a loss in the generation of functional HSCs^86,87. It is likely that the early somite-derived Notch signals induce gata2b expression to prime the HSC program in zebrafish progenitor cells⁸². Many of the factors that are essential for Notch-mediated hemogenic endothelium specification are not required for its role in arterial specification, and it is likely that the downstream effects of Notch activation are ligand dependent⁸⁸.

As previously discussed, zebrafish studies have demonstrated the importance of epigenetic regulators in definitive HSC development. The chromatin factor screen conducted by our laboratory identified 29 factors that affected HSC number in the AGM, including the previously unknown factors brd8a, cbx6b, jmjd1c and nap1l4a⁵². The histone-modifying enzyme DNA methyltransferase 3bb.1 (Dnmt3bb.1) is expressed downstream of notch and runx1, and zebrafish mutants lacking functional Dnmt3bb.1 have low numbers of HSCs due to reduced methylation of the cmyb locus and reduced cmyb expression⁸⁹. A zebrafish forward genetic screen revealed an important role for another DNA methyltransferase, Dnmt1, which is essential for maintenance of HSCs via regulation of the downstream target gene cebpa⁹⁰.

After undergoing EHT, HSCs bud from the hemogenic endothelium into the sub-aortic space⁶⁰. The mechanisms that regulate budding are not well defined, although the Runx1 cofactor, Cbfβ, has been shown to play an important role in the release of HSCs⁹¹. Nascent HSCs then enter the circulation to begin their journey to their intermediate stem cell niche in the caudal hematopoietic tissue (CHT), which is equivalent to the mammalian fetal liver⁹². After a transient proliferation phase, definitive HSCs reenter the circulation and make their way to the final hematopoietic site, which is the bone marrow in mammals and the kidney marrow in zebrafish⁵⁹. From here, they produce a lifetime supply of mature blood cells.

The dynamics of HSC emergence and niche interactions

With its transparent ontogeny, the zebrafish provides a unique tool for studying the biology of HSCs in their native niche and over a defined period of time. Major advances in live cell imaging by time-lapse confocal microscopy have advanced the potential of this model system to explore the in vivo development of HSCs and their interaction with the environment at the cellular level.

In vivo visualization of HSC emergence

Live cell imaging has traced the fascinating journey of blood stem cells, from their birth in the hemogenic endothelium to their travel through the AGM to the CHT niche and beyond. This has been achieved using transgenic zebrafish lines expressing fluorescent proteins under hematopoietic stem and progenitor cell (HSPC)-specific promoters or enhancers such as cmyb, cd41 or runx1, combined with endothelial transgenics such as Tg(kdrl:GFP), in which vascular-specific GFP expression is controlled by the promoter of the kdrl gene from the angioblast stage^59–61,93. Two major works used these approaches to demonstrate the early stages of HSC origin. In one of these studies⁶⁰, the Tg(kdrl:GFP) line was used to trace the emergence of HSCs from aortic endothelium. By following GFP-positive endothelial cells from 18 hpf to around 100 hpf, it was observed that a particular endothelial cell undergoes a series of contractions and shape changes and eventually leaves the tightly woven endothelial field, whilst the rest of the surrounding cells maintain the integrity of the vessel wall. This phenomenon was termed the endothelial to hematopoietic transition, or EHT, as mentioned previously. The newly released cell, morphologically consistent with a typical hematopoietic progenitor, then travels in the sub-aortic space to the CHT, thymus, and kidney marrow. In the second study⁶¹, Tg(cmyb:eGFP;kdrl:mCherry) double-transgenic zebrafish were used to show that cmyb-positive HSPCs directly emerge from aortic endothelial cells. In their model, hemogenic endothelial cells begin to express the HSPC marker cmyb and then undergo a special movement of bending towards the sub-aortic space before eventually being released into the circulation, similar to the EHT described above⁶⁰. Once in the sub-aortic mesenchyme, HSPCs enter venous circulation between overlapping endothelial cells, in areas that appear to be specialized by weak cell junctions and are lined by stromal reticular cells⁵⁹. Nascent HSCs then make their way through the circulation to the CHT niche⁶¹.

The role of the HSC niche

A recent study demonstrated a critical role for primitive macrophages during HSC emergence⁹⁴. In Tg(cd41:eGFP;mpeg1:mCherry) zebrafish embryos, primitive macrophages were observed to accumulate in large numbers in the AGM. These cells were responsible for creating the structures between the extracellular matrix and the posterior cardinal vein allowing HSC entrance into venous circulation. Early depletion of macrophages from 25 hpf prevented stem cell entrance into the circulation and seeding of distant sites in the CHT, thymus, and kidney marrow, without disturbing EHT from the hemogenic endothelium. Importantly, the accumulation of primitive macrophages in the early HSC niche at the dorsal aorta/posterior cardinal vein is documented in human embryos, suggesting a conserved role in human hematopoietic development⁹⁴.

Our laboratory has studied the behavior and migration of blood stem cells through the transient CHT niche using runx1:GFP and runx1:mCherry transgenic lines. Using spinning disk confocal microscopy to capture high resolution imaging over 16 hours, runx1:GFP-positive HSPCs were visualized to exit the circulation and become rapidly surrounded by endothelial cells, which created a pocket for the cells in a process termed “endothelial cuddling”⁹⁵. In the endothelial pocket, HSPCs are situated proximal to cxcl12a-expressing stromal cells, which help to orient the cell during cellular divisions. Divisions appeared to be either symmetric, in which both daughter hematopoietic cells remained in the pocket and continued to interact with the stromal cells, or asymmetric, in which the stromal cell remained in contact with only one daughter cell, presumably the more stem-like cell, whilst the other daughter cell detached and entered circulation. The mechanisms by which stromal cells and other niche cells, such as macrophages, may affect the fate of HSCs is largely unknown and is under active investigation.

An innovative approach to investigate HSPC and niche interaction is by parabiosis, in which two zebrafish embryos are fused during the blastula stage, such that they grow with shared circulation⁹⁶. Parabiotic zebrafish have allowed separation of cell-autonomous and non-cell-autonomous aspects of HSPC biology. By creating a genetic lesion in a presumed niche cell in just one of the embryos of a parabiotic pair, it is possible to observe how hematopoietic cells, which are circulating between the two fish, behave in wild-type or mutated environments. Using this technique, a recent study demonstrated that cadherin 5 (Cdh5) expression on endothelial cells was dispensable for HSPC development⁹⁷. The parabiosis was performed between cdh5-silenced embryos, which also carried the cd41:eGFP transgene, and transparent wild-type casper⁹⁸ embryos. In the parabiotic pair, cd41-positive HSPCs emerged from the cdh5-deficient aortic endothelium, migrated to the CHT of the wild-type casper fish, and were able to engraft and differentiate into circulating thrombocytes. These findings were validated in a mammalian hematopoietic system⁹⁷. As parabiotic fish share a common bloodstream, circulating soluble chemicals are also shared, making this a useful model to study the effect of overexpressed cytokines or other circulating signals on hematopoiesis.

Zebrafish as a model for blood disorders and therapeutics

Over the last few decades, zebrafish have risen to the forefront of small molecule drug discovery and human disease modeling. The zebrafish model has unique advantages for high-throughput drug screening, particularly over in vitro screening assays, as the use of a whole-organism system allows recognition of toxicity and embryonic lethality at an early stage of drug development. Furthermore, their small size enables the arraying of numerous larvae into multi-well plates, and their optical transparency facilitates phenotype-based screening approaches by allowing direct visualization of changes induced by compound treatment. This system provides the potential to identify previously undiscovered signaling pathways associated with human disease processes, which can be targeted for therapeutic benefit.

Zebrafish as a model for leukemia

Due to the high degree of similarity in hematopoietic development with mammalian systems, zebrafish have been widely used for studies of malignant and nonmalignant blood disorders in adult fish. With the powerful forward genetic and chemical screen technologies available in zebrafish, modeling blood disorders using this system has great potential for finding new modifier genes in disease pathophysiology and onset, and to identify new drugs for treatment.

T-cell acute lymphoblastic leukemia (T-ALL) and T-cell lymphoblastic lymphoma (T-LBL) were the first hematologic malignancies modeled in zebrafish, by overexpressing the mouse Myc gene fused to GFP, under the control of the rag2 promoter⁹⁹. These disease models recapitulated the human condition; with the added advantage of providing means of visualization and tracking of GFP-labeled leukemia cells. These cells could be isolated from primary leukemic zebrafish and transplanted into conditioned recipients, demonstrating full leukemia initiating potential of the cells. The T-LBL model was based on heat-inducible Myc expression, and over time progressed to T-ALL¹⁰⁰. This is a well-known transformation in humans, but the exact mechanism by which a localized lymphoma spreads to the blood in the form of leukemia was previously unknown. Mechanistic investigation using the T-LBL zebrafish model revealed that the cancerous cells are kept intact within the lymphoma by the apoptosis regulator Bcl2, the sphingosine 1-phosphate receptor 1 (S1pr1), and intracellular adhesion molecule 1 (Icam1), which together inhibit apoptosis and prevent the cell-cell interactions required for vascular invasion and spread. Upon Akt activation, however, this inhibition is overcome and the cells are able to seed the circulation, transforming into T-ALL.

Efforts to model myeloid leukemias have also shown promising results. Overexpression of AML1-ETO, a common oncogenic translocation found in human AML, had limited utility due to early lethality in fish¹⁰¹. However, a stable transgenic zebrafish line was generated that allowed expression of AML1-ETO in an inducible manner, which resulted in cytological and transcriptional similarity to human AML¹⁰¹. This model did not fully phenocopy the hematologic disease state in adult fish but provided useful insight into early hematopoietic development. Interestingly, the expression of the oncogenic protein during embryogenesis affected both primitive and definitive hematopoiesis, and promoted a disproportionate increase in granulopoiesis at the expense of erythropoiesis. The key to this regulation was the known HSC transcription factor scl which was down-regulated by AML1-ETO expression, leading to up-regulation of the myeloid factors pu.1 and mpx. The overexpression of scl could restore erythropoiesis in AML1-ETO expressing zebrafish. Furthermore, this zebrafish line was used to test the effect of chemical inhibition of AML1-ETO by a histone deacetylase inhibitor, trichostatin A (TSA), which prevented the down-regulation of scl and gata1, reversing the myeloid-biased hematopoietic program in early embryogenesis. Recruitment of HDACs by the ETO domain had been reported previously¹⁰², and this model presented the platform to test other known HDAC inhibitors in vivo. It could also be used for high-throughput chemical screening for novel compound identification with a simple in vivo readout. The AML1-ETO zebrafish model provided important mechanistic information about how this oncogenic protein may lead to cell fate changes, contributing to our understanding of the pathologic myeloid program switch in leukemia initiating cells.

Modeling pre-leukemic hematopoietic clonal disorders in zebrafish

Over the past decade, outstanding advances have been made in understanding the genetics of myeloid neoplasms, including disorders of benign clonal hematopoiesis¹⁰³, to pre-leukemic conditions, such as myelodysplastic syndromes (MDS) and myeloproliferative neoplasms (MPNs). The mutational landscape of known genes in myeloid neoplasms, about 100 by most reports¹⁰⁴, has been plotted in several large studies^{103,105–107}. Modeling these clonal disorders in zebrafish could provide an excellent platform for understanding the pathophysiology of leukemogenesis, from the early clonal expansion of abnormal stem cells to clinically apparent pre-leukemic states such as MDS.

Several zebrafish models of MDS and myeloproliferative disorders have been reported to date^108–111. An MDS model was generated using ZFN technology to establish a loss-of-function mutation in the tet2 gene¹⁰⁸. Homozygous mutants exhibited changes in their peripheral blood counts at approximately 11 months of age, which became more pronounced by 24 months, when dysplastic changes in kidney marrow cells, equivalent to mammalian bone marrow cells, were also observed. TET2 loss in human patients is found in clinically benign conditions associated with clonal hematopoietic proliferation, and requires the acquisition of additional mutations to progress into clinically apparent disease states like MDS or AML. Given the length of time needed for tet2-mutated zebrafish to show measurable changes in their hematopoietic system, it is possible that secondary mutations could have been acquired in the interim, similar to the pathophysiology of MDS in humans.

Myeloproliferative states with expansion of myeloid lineages are closest to the myeloproliferative neoplasms (MPNs) found in humans. An MPN-like model in zebrafish was identified as a result of a mutagenesis screen using ENU¹⁰⁹, which exhibited an increase in runx1-positive cells in early development and increased differentiation into myeloid lineages. The cbl gene, an ortholog of c-CBL implicated in human MPNs, was identified as the underlying mutation. The phenotype observed in the cbl-mutant zebrafish line was dependent on Flt3 signaling, validating similar pathophysiology to MPNs in patients. An increased myeloid phenotype was observed in another model created by mosaic overexpression of the human FLT3-ITD mutation, found in about one third of acute myeloid leukemias¹¹⁰. Injection of FLT3-ITD into the one-cell stage zebrafish embryos led to an expansion of primitive myeloid cells expressing pu.1, mpx and cebpa at 18 and 36 hpf. Sorted FLT3-ITD-positive cells appeared monocytic in morphology, recapitulating a similar phenotype observed in mice where overexpression of FLT3-ITD in hematopoietic cells results in myeloproliferative disorder¹¹². A third example of a myeloproliferative phenotype in zebrafish resulted from overexpression of the KRAS^G12D mutation, found in human MPNs and acute leukemias, using a heat shock inducible strategy¹¹¹. This was one of the earliest examples of the use of Cre/Lox technology in zebrafish. Kidney marrow cells from adult KRAS^G12D overexpressing zebrafish were collected and heat shocked ex vivo prior to transplantation into irradiated recipient fish. Two months post-transplant, 8 of 38 fish developed a myeloproliferative state, which resembled the induced phenotype observed in murine myeloproliferative disorder with KRAS^G12D expression^113,114.

Modeling non-malignant hematologic disorders in zebrafish

A large number of benign hematologic disorders affecting humans have been modeled in zebrafish. Some of these disorders result from impaired HSPC differentiation and can be treated using HSC transplantation. One such example is Diamond-Blackfan Anemia, a bone marrow failure syndrome caused by mutations in specific ribosomal protein genes. Patients with Diamond-Blackfan Anemia exhibit a predominantly erythroid progenitor defect, although multipotent progenitors are likely affected as well^115–117. Clinically, this manifests initially as a profound macrocytic anemia that may evolve over time to include thrombocytopenia and neutropenia. A number of zebrafish models of Diamond-Blackfan Anemia exist, including mutants and morphants for rps19 and rpl11^118–120, and these have been used to identify potential pharmacologic therapies for the disease^121–125. They have also been used to elucidate the mechanisms underlying Diamond-Blackfan Anemia, including up-regulation of p53 target genes^125–128. The connection between p53 and ribosomal protein mutations has also been observed in mammalian systems¹²⁹. A better understanding of the mechanisms leading to Diamond-Blackfan Anemia should permit the development of additional therapeutic options for patients with this disease.

It is worth noting that a significant number of human cases of bone marrow failure remain genetically uncharacterized, even though the clinical features are often consistent with well-recognized syndromes. For instance, it is not always possible to elucidate a ribosomal protein mutation in patients that are diagnosed clinically as having Diamond-Blackfan Anemia. The identification of zebrafish mutants with defects in HSPC specification, maintenance, or differentiation provides a list of candidate genes and pathways that may be implicated in as-yet genetically undefined human bone marrow failure syndromes^130–134.

Using zebrafish to assess HSC transplantation and engraftment

Hematopoietic stem cell transplantation in zebrafish was first reported in 2003, when it was shown that kidney marrow cells from a specific light-scatter flow cytometry gate, termed the lymphoid fraction, contained long-term reconstituting HSCs¹³⁵. Subsequent transplantation experiments using transgenic cd41:GFP donor fish revealed that at least some HSCs express cd41:GFP at low levels¹³⁶. More recently, transplantation experiments have demonstrated that a subset of runx1:mCherry and runx1:GFP cells isolated from transgenic kidney marrow have HSC activity⁹⁵. To date, monoclonal antibodies for identifying HSCs in zebrafish have not been developed despite significant efforts by multiple groups.

Studying HSC transplantation in the zebrafish system has been limited by several factors compared to its mammalian counterpart. Zebrafish are not isogenic and there is currently an incomplete knowledge of their relevant major histocompatibility complex (MHC) loci, which are more diverse than in humans^137–141. Lack of MHC matching has the potential to compromise the transplanted graft due to graft failure via rejection. To date, adequate results have been obtained without MHC matching, but for certain key studies the question lingers whether MHC matching would improve engraftment efficiency^95,136. It should be noted that the resources required to generate and maintain multiple transgenic donor strains in an MHC-matched background would be significant, not least because congenic strains in zebrafish suffer from a lack of fecundity^137,142,143. Strategies to increase engraftment besides MHC matching are being explored, including modulating the dose rate of irradiation¹⁴⁴ and subjecting recipients to cyclosporine A or other chemical treatments¹⁴⁵. There has also been interest in using genetically immunocompromised recipients rather than recipients subjected to gamma-irradiation, which may reduce the rate of graft rejection^146–148. The isolation of zebrafish HSCs by fluorescence-based cell sorting (FACS) also remains challenging due to the lack of characterized surface markers and their antibodies. Despite these limitations, clinically relevant advances have been made in the biology of HSPC engraftment using transgenic zebrafish with fluorescent protein expression under HSPC-specific promoters. HSC transplantation in zebrafish has been used to provide a functional definition of HSCs and to dissect the roles of various signaling pathways in vertebrate hematopoiesis^95,135,136. There has also been interest in using the zebrafish as a xenograft model of HSC transplantation^148,149.

Understanding the mechanisms that regulate HSC engraftment in vivo is fundamental for ensuring the success of bone marrow transplants in patients suffering from blood cancers, bone marrow failure and immunodeficiency syndromes. From a clinical standpoint, improving HSC engraftment after transplantation is critical for decreasing the risk of life-threatening infections and limiting transfusions of blood products for supportive care. General approaches to achieve this goal are to increase the number of stem cells transplanted, treat stem cells prior to transplantation to increase their engraftment capacity through improved homing or seeding of the marrow niche, or to treat the patient post-transplantation to create a more supportive environment for incoming stem cells. Increasing HSC number has been the focus of a number of preclinical studies that have since been translated into clinical trials for patients receiving umbilical cord blood stem cell transplants. A recent chemical screen in zebrafish directed towards identifying compounds that enhance HSC engraftment in transplantation assays was the first of its kind in transplantation biology¹⁵⁰. One interesting feature of the study was that engraftment was assessed using fluorescence intensity, which could be directly measured due to the use of a transparent casper zebrafish⁹⁸. This screen used whole kidney marrow cells isolated from transgenic donor fish that expressed either GFP or DsRed2 in a ubiquitous manner for competitive transplant experiments into casper. The GFP-positive cells were treated with a chemical library, and DsRed2-positive cells served as a control. The ratio of GFP/DsRed2 intensity was measured by direct imaging and FACS analysis of recipient kidney marrow. Two compounds showed an increase in the GFP/DsRed2 ratio: 11,12-epoxyeicosatrienoic acid (11,12-EET) and 14,15-EET, eicosanoids synthesized through the cytochrome P450 epoxygenase pathway. Neither of these compounds has previously been implicated in HSPC biology, although a P450 cytochrome enzyme, Cyp2j6, is known to be enriched at the mRNA level in murine long-term HSCs compared to multipotent progenitors¹⁵¹. The zebrafish study showed that treatment with 11,12-EET at the time of definitive hematopoietic initiation, between 24 and 36 hpf, increased runx1-positive HSPCs in the AGM, suggesting that it acts at the level of the hemogenic endothelium. Further molecular dissection of the mechanism revealed the involvement of multiple activator protein 1 (AP-1) family transcription factors such as fosl2, junb, and junbl, orthologs of human JUNB, which were up-regulated upon 11,12-EET treatment. In addition to activating runx1 expression and inducing HSC fate in the hemogenic endothelium, the AP-1 program has also been shown to be important in the cell-cell signaling required for cell migration, possibly explaining the positive effect of 11,12-EET on HSPC migration and homing¹⁵⁰. The study raised the possibility that 11,12-EET may be clinically relevant for improving engraftment rates in human HSC transplantation.

Zebrafish as a model for drug discovery

The zebrafish permits rapid screening of large numbers of pharmacologic compounds for their effect on a hematopoietic phenotype of interest¹⁵². This has resulted in the elucidation of pathways that impact HSC development and homeostasis and the identification of compounds that hold promise for the treatment of human disease. Screens have mainly been carried out using zebrafish embryos, but it is possible both to treat hematopoietic cells ex vivo prior to re-infusion and to perform screens in adults. Initial screens for compounds that alter hematopoiesis in zebrafish were conducted in embryos. One of the biggest success stories to date was the identification of prostaglandin E2 (PGE2) as a molecule that enhanced the formation of HSCs and resulted in faster zebrafish marrow recovery following irradiation injury¹⁵³. The initial screen was designed to identify compounds that affected the number of runx1/cmyb-positive HSPCs in zebrafish as measured by whole mount in situ hybridization. Approximately 2500 compounds from three compound libraries were screened, of which 5% were toxic to zebrafish embryos. The screen identified two compounds, linoleic acid and celecoxib, with opposite effects on HSPCs - the former increased and the latter decreased their number in the CHT. Both of these compounds acted on prostanoids and their effector, PGE2. PGE2 alone increased the number of runx1/cmyb HSPCs in embryos and enhances engraftment of HSCs in adult fish transplantations. These observations were validated in mouse transplantation assays, where exposure to PGE2 for only 2 hours enhanced the long-term reconstitution potential of murine HSPCs¹⁵⁴. A subsequent clinical trial in humans receiving umbilical cord blood transplants, which are complicated by relatively slow engraftment, has shown encouraging results. Accelerated neutrophil recovery, and early and sustained chimerism with long-term engraftment of the PGE2-treated cord blood unit were observed in 10 of 12 treated participants¹⁵⁵. Currently, studies are underway to fully understand the mechanism by which PGE2 regulates HSPCs. At least two of the four prostaglandin receptors, Ptger2 (EP2) and Ptger4 (EP4), are expressed on zebrafish HSCs, and knockdown of either leads to a decreased number of runx1/cmyb-positive cells in the AGM, an effect that is not rescued by PGE2 treatment¹⁵³. PGE2 has previously been shown to lead to activation of the GSK3/β-catenin pathway after EP2 or EP4 binding¹⁵⁶. PGE2 also interacts with Wnt and enhances its signaling by stabilizing β-catenin¹⁵⁷. It is possible that PGE2 may promote HSC development and an increase in HSC number via β-catenin-mediated expression of cycle proliferation and survival genes, whilst inhibiting cell death. Recently, cannabinoid receptor-2 was shown to regulate HSC development through PGE2 and P-selectin activity¹⁵⁸, but the precise mechanism by which PGE2 exerts its effects on HSCs has not yet been defined.

Chemical screens in embryos can also incorporate transgenic technology. One example is the Tg(hsp70:AML1-ETO) zebrafish line that recapitulates some features of human acute myeloid leukemia (AML)¹⁰¹. This transgenic line has been used in a suppressor screen for compounds that inhibit the AML1-ETO-induced switch from erythropoiesis to granulopoiesis, which identified PGE2 and β-catenin pathways as important modulators of AML1-ETO-regulated hematopoietic differentiation¹⁵⁹. A role for these signaling pathways in mouse and human leukemia cells was subsequently confirmed¹⁶⁰. Pharmacologic screens in adult zebrafish are technically challenging, in part due to difficulty administering compounds. However, techniques to deliver compounds by serial gavage of adult zebrafish have recently been developed¹⁶¹. This should permit future screens that examine adult hematopoietic stem and progenitor cell function. Alternatively, hematopoietic cells marked by a fluorophore of choice can be harvested from zebrafish, treated ex vivo, and then re-infused into appropriately conditioned recipients. This strategy was recently employed in a zebrafish screen for compounds that improve HSC transplantation¹⁵⁰.

Conclusions and perspectives

Over the last few decades, the zebrafish has become a valuable tool for studying hematopoiesis, and many key discoveries in blood development and disorders that have been made using this system have been validated in mammalian models (Table 1). Zebrafish are particularly versatile for deciphering the molecular signaling pathways that are required for hematopoietic stem cell development. Subfunctionalization of zebrafish genes such as scl and gata2 makes it relatively straightforward to study gene function in the specific context of HSC specification without affecting vasculogenesis, which cannot be achieved in mammalian systems. The ease of high-throughput mutagenesis screening and advance of gene editing technologies has enabled the generation of many zebrafish mutants with hematopoietic defects, which have been fundamental for the identification and functional investigation of key transcription factors involved in HSC specification. These findings may inform future efforts to generate HSCs from human induced pluripotent stem cells in vitro. The unique advantages of the zebrafish also make it a particularly versatile model system for investigating the dynamics of HSC emergence and migration using real-time imaging approaches, and for unraveling the signaling pathways that regulate HSC engraftment. Zebrafish are now being widely used to model a range of malignant and non-malignant hematopoietic disorders and for investigating HSC transplantation. It is likely that these models will continue to provide valuable platforms for therapeutic drug discovery and for advancing our understanding of diseases affecting the hematopoietic system.

Table 1.

Discoveries in zebrafish steady state hematopoiesis and blood disorders that have been validated in mammalian model organisms or humans.

Zebrafish discoveries	Validation in mouse or human
Effect of adenosine signaling through the A2b receptor on HSC emergence from the AGM via a cAMP-PKA-dependent pathway and induction of sclβ expression in the hemogenic endothelium71.	Validated effect of adenosine in mouse ES cells in a hematopoietic differentiation system and explanted mouse AGM cells at E10.5 in colony forming assays. Human aortic endothelial cells used to confirm involvement of a cAMP-PKA pathway71.
HSC emergence from AGM using live imaging in transgenic zebrafish lines59–61.	HSC emergence from the aortic endothelium in mouse embryos68,84,93.
Role of primitive macrophages in the emergence of HSCs94.	Primitive CD68+ macrophages have been described in the aortic endothelium in human embryos94.
Cadherin 5 (Cdh5) in the endothelial niche is dispensable for HSC emergence and EHT97.	Conditional Cdh5 knockout mice have normal HSCs97.
Diamond-Blackfan Anemia like phenotype in zebrafish due to rsp29 mutation is rescued with p53 mutation, suggesting a critical role for p53 in ribosomal protein mutant phenotypes127,128.	Rps19Dsk3 allele mutant mice demonstrated decreased red cell count, macrocytosis, decreased reticulocyte count and increased apoptosis; the erythrocyte phenotype was rescued in the context of p53 mutation129.
Prostaglandin E2 (PGE2) modulates HSC homeostasis, enhancing formation and number of HSCs153.	PGE2 increases the potential of HSCs in colony forming assays and in vivo transplantation in mice153,154. PGE2 is the first chemical compound that emerged from zebrafish studies to be used in a human Phase I clinical trial. It showed enhanced and sustained long-term engraftment in umbilical cord transplantation155.
A suppressor screen in zebrafish identified the Cox-2 pathway to be involved in AML-ETO1 mediated hematopoietic dysregulation in zebrafish159.	Human myelogenous leukemia K562 cell line showed COX-2–dependent signaling upon AML-ETO1 transformation and effects on multipotent progenitor differentiation159.