Zebrafish erythropoiesis and the utility of fish as models of anemia

Erythrocytes contain oxygen-carrying hemoglobin to all body cells. Impairments in the generation of erythrocytes, a process known as erythropoiesis, or in hemoglobin synthesis alter cell function because of decreased oxygen supply and lead to anemic diseases. Thus, understanding how erythropoiesis is regulated during embryogenesis and adulthood is important to develop novel therapies for anemia. The zebrafish, Danio rerio, provides a powerful model for such study. Their small size and the ability to generate a large number of embryos enable large-scale analysis, and their transparency facilitates the visualization of erythroid cell migration. Importantly, the high conservation of hematopoietic genes among vertebrates and the ability to successfully transplant hematopoietic cells into fish have enabled the establishment of models of human anemic diseases in fish. In this review, we summarize the current progress in our understanding of erythropoiesis on the basis of zebrafish studies and highlight fish models of human anemias. These analyses could enable the discovery of novel drugs as future therapies.


Introduction
Red blood cells, or erythrocytes, carry hemoglobin to supply oxygen to all tissues and organs. Approximately 2 × 10 13 erythrocytes circulate throughout the whole body. In humans, more than 10 11 new erythrocytes are generated daily from bone marrow (BM) through a process known as erythropoiesis [1]. In the BM, the hierarchy of erythropoiesis is topped by hematopoietic stem cells (HSCs), which fi rst diff erentiate into common myeloid and common lymphoid progenitors. Common lymphoid and myeloid progenitors give rise to the adaptive and innate immune systems, respectively. Common myeloid progenitors diff erentiate into megakaryocyte/ erythroid progenitors and granulocyte/monocyte progeni tors. At the same time, common lymphoid progenitors diff erentiate into B lymphocytes, T lymphocytes, and natural killer cells. Megakaryocyte/erythroid progenitors later give rise to erythrocytes or thrombocytes (platelets), whereas granulocyte/monocyte progenitors give rise to granulocytes (neutrophils, eosinophils, and basophils), monocytes, and dendritic cells (Figure 1). Erythrocytes synthesize hemoglobin, which is composed of two βglobin subunits and two α-globin subunits that interact with an iron-containing heme moiety. Intrinsic transcription factors and extrinsic signaling molecules coordinately regulate erythroid diff erentiation and hemoglobin synthesis. Impaired erythrocyte production or hemoglobin synthesis results in anemia and decreases the oxygen supply throughout the body, a condition known as hypoxia. Much of what we know about human erythropoiesis and anemic diseases comes from studies using animal models such as Xenopus, zebrafi sh, chicks, and mice [2][3][4][5].
Zebrafi sh (Danio rerio) is a teleost freshwater fi sh widely distributed throughout tropical and subtropical areas of South Asia, including India, Nepal, Bangladesh, and Northern Burma [6]. Zebrafi sh are known worldwide as models for the study of development, cell biology, physiology, and genetics. In vivo analyses using zebrafi sh have some advantages over those using mice. First and foremost, zebrafi sh produce large numbers of small-sized embryos, permitting drug screen ing and functional analysis of specifi c genes on a large scale. Second, zebrafi sh have a short life span (42 to 66 months) [7] and develop rapidly, requiring 90 days to develop into adults [7], shortening periods required for experiments. Th ird, zebrafi sh embryos are transparent and develop outside a uterus, enabling researchers to view zebrafi sh development and genetically manipulate embryos under a microscope. Finally, many zebrafi sh gene functions are conserved in mice and humans, enabling researchers to trans late results obtained in zebrafi sh studies to mammalian Abstract Erythrocytes contain oxygen-carrying hemoglobin to all body cells. Impairments in the generation of erythrocytes, a process known as erythropoiesis, or in hemoglobin synthesis alter cell function because of decreased oxygen supply and lead to anemic diseases. Thus, understanding how erythropoiesis is regulated during embryogenesis and adulthood is important to develop novel therapies for anemia. The zebrafi sh, Danio rerio, provides a powerful model for such study. Their small size and the ability to generate a large number of embryos enable largescale analysis, and their transparency facilitates the visualization of erythroid cell migration. Importantly, the high conservation of hematopoietic genes among vertebrates and the ability to successfully transplant hematopoietic cells into fi sh have enabled the establishment of models of human anemic diseases in fi sh. In this review, we summarize the current progress in our understanding of erythropoiesis on the basis of zebrafi sh studies and highlight fi sh models of human anemias. These analyses could enable the discovery of novel drugs as future therapies.

Development of zebrafi sh erythropoiesis
Like the generation of other blood cell types, zebrafi sh erythropoiesis takes place in the mesodermal germ layer and is classifi ed into two sequential waves: primitive and defi nitive. Th e primitive wave generates erythrocytes and macrophages during embryonic development, whereas the defi nitive wave produces defi nitive HSCs, which can diff erentiate into every blood cell type (namely, erythrocytes, granulocytes, lymphocytes, and platelets), and maintains homeostasis throughout the zebrafi sh lifetime ( Figure 2). To understand how erythropoiesis develops embryonically and is maintained in the adult, we also discuss the origin of erythrocytes from HSCs.

Primitive erythropoiesis
Th e fertilized egg, or zygote, divides and forms three germ layers: ectoderm, mesoderm, and endoderm; this process is known as gastrulation. Meso derm gives rise to muscle, notochord, hematopoietic cells, pronephros, and blood vessels. Mesoderm is divided into dorsal and ventral mesoderm. Dorsal meso derm develops into the notochord, whereas ventral mesoderm gives rise to hematopoietic cells, the pro nephros, and blood vessels. Primitive hematopoiesis intraembryonically starts in ventral mesoderm-derived tissue, known as the intermediate cell mass (ICM) (Figures 2 and 3). Th e ICM is located between the somites and yolk sac and consists of anterior and posterior ICM (Figure 3). At the two-somite stage, which is equivalent to 10 to 11 hours post fertilization (hpf ), genes encoding transcription factors required for hematopoietic cell specifi cation, such as T-cell acute lymphocytic leukemia 1 (tal1), GATAbinding protein 2a (gata2a), and LIM domain only 2 (lmo2), and vasculo genesis, such as ets variant gene 2 (etv2), are co-expressed in both the anterior and posterior ICM, implying the existence of a common ancestor of hematopoietic and endothelial cells, known as the hemangioblast [8][9][10] (Figure 4). Th ese genes are highly conserved among vertebrates [10][11][12][13]. Among proteins encoded by these genes, tal1, previously known as stem cell leukemia (scl), is a basic helix-loop-helix transcription factor required for both primitive and defi nitive hematopoiesis as well as endothelial cell diff erentiation, whereas gata2a is a zinc fi nger transcription factor functioning in proliferation and maintenance of hematopoietic progenitor cells (HPCs). Zebrafi sh lmo2 is an LIM domain transcription factor that interacts with tal1 and gata2a, forming a DNAbinding complex, which activates the transcription of both hematopoietic and endothelial genes [10]. Zebrafi sh etv2, previously known as ets1-related protein (etsrp), is an E-twenty six (ets) domain-containing factor that activates transcription of endothelial-specifi c genes essential for vasculogenesis [11]. Th e posterior ICM expresses GATA-binding protein 1a (gata1a), which encodes an erythroid-specifi c transcription factor, and spleen focus-forming virus (SFFV) proviral integration oncogene spi1 (spi1), which encodes a myeloid-specifi c transcription factor, whereas the anterior ICM expresses only spi1 [12,13]. Th ese studies suggest that the ICM has been committed to erythroid and myeloid lineage. Between 12 and 24 hpf, the anterior ICM develops into myeloid cells (macrophages), whereas the posterior ICM develops primarily into erythroid and some myeloid cells (Figure 3).
At 20 to 24 hpf, primitive erythroid cells (proerythroblasts and erythroblasts) expressing gata1a are present in the posterior ICM [14] (Figure 2). Similar to the mammalian yolk sac, primitive erythroid cells are surrounded by endothelial cells [15]. After the onset of blood circulation at 24 hpf, gata1a + primitive erythroid cells migrate throughout the embryo and diff erentiate into mature erythrocytes expressing aminolevulinate, delta-, synthetase 2 (alas2), which is an enzyme required for heme synthesis, and embryonic globin genes [16]. Unlike similar cells in humans and mice, zebrafi sh erythrocytes are nucleated and oval in shape ( Figure 4). Primitive erythropoiesis accounts for all circulating erythrocytes for the fi rst 4 days after fertilization [15].
In zebrafi sh, primitive erythropoiesis is regulated intrin sically by transcription factors expressed in erythroid cells and extrinsically by erythropoietin secreted from surrounding cells and tissues. Zebrafi sh gata1a, a zinc fi nger transcription factor, activates the expression of erythroid-specifi c genes functioning in hemoglobin synthesis, iron utilization, and cell membrane stabili zation. Gata1a also suppresses the expression of myeloidspecifi c genes [14]. As in the mouse, a nonsense mutation in the gene encoding zebrafi sh gata1a results in a lack of circulating erythrocytes at 26 hpf [17], suggesting a conservation of gata1a function among vertebrates. Unlike in the mouse, zebrafi sh primitive erythropoiesis depends on the extrinsic factor erythropoietin (epo). Epo and erythropoietin receptor (Epor) mRNA is detected in the ICM during 16 to 24 hpf. Knockdown of Epor impairs primitive erythropoiesis in fi sh [18].

Defi nitive erythropoiesis
In mice, defi nitive HSCs are defi ned as having the ability to reconstitute all blood cell types in lethally irradiated adult mice. Defi nitive HSCs are fi rst detected in the aorta-gonad-mesonephros (AGM) region [19] and umbilical vessels [20] at mouse embryonic day 10 or 11. Defi nitive HSCs enter the blood circulation and colonize fetal liver, the fi rst site of defi nitive hematopoiesis, where they expand and diff erentiate into erythroid and myeloid cells.
In zebrafi sh, an AGM-like region exists along the trunk in the space between the dorsal aorta and the underlying axial vein [21][22][23][24] (Figure 3). Cell-tracking and time-lapse imaging analyses indicate that HSCs and HPCs originate directly from the ventral wall of dorsal aorta (VDA) at 26 hpf ( Figure 2) [24,25]. Th e morphology of endothelial cells lining the VDA changes from a long fl at shape to a round shape, and the cells egress into the subaortic space and enter the blood circulation via the axial vein ( Figure 3), a process called endothelial-to-hematopoietic transition [24]. As in mammals, runt-related trans cription factor 1 (runx1) is critical for the emergence of hematopoietic cells from endothelial cells [26]. At 24 hpf, Figure 2. Comparative timeline of embryonic zebrafi sh and mouse hematopoiesis. From 12 to 24 hours post fertilization (hpf ), primitive hematopoietic cells (proerythroblasts, erythroblasts, and macrophage precursors) appear in the intermediate cell mass, which is equivalent to the mouse yolk sac, from embryonic day (E) 7.5 to E10.0. After blood circulation starts at 24 hpf, primitive hematopoietic cells enter the circulation and mature. From 26 to 48 hpf, defi nitive hematopoietic precursors emerge from endothelial cells lining the ventral wall of dorsal aorta. The space between the dorsal aorta and axial vein is equivalent to the mouse aorta-gonad-mesonephros (AGM) region (E10.5 to E12.5). At 2 to 6 days post fertilization (dpf ), defi nitive hematopoietic precursors enter the circulation, colonize, and expand in the caudal hematopoietic tissue, which is equivalent to mouse fetal liver (E11.5 to E18.5). In fi sh, hematopoiesis then shifts to the kidney and thymus. In the mouse, hematopoiesis shifts from fetal liver to bone marrow at E18.0. At 3 dpf, lymphoblasts appear in the thymus and diff erentiate into T lymphocytes. In fi sh, all hematopoietic lineages appear in the kidney marrow at 4 dpf, whereas B lymphocytes appear at 19 dpf. Thus, the kidney marrow and thymus are major hematopoietic organs throughout the lifespan of adult zebrafi sh. Sites where erythropoiesis occurs are shown by red boxes. runx1-expressing HSCs and HPCs are observed in the VDA [27]. Later, at 26 hpf, dorsal aorta-derived runx1 + HSCs and HPCs start to express the transcription factor cmyb [28], which is required for HSC migration and diff er entiation but not for the endothelial-to-hematopoietic transition in zebrafi sh [29]. Later, cmyb + HSCs and HPCs express integrin alpha 2b (itga2b), also known as CD41 [30]. As in the mouse, CD41 is the earliest HSC and HPC surface marker seen in endothelial cells lining the dorsal aorta. Dorsal aorta-derived CD41 + HSCs and HPCs enter the blood circulation via the axial vein rather than the dorsal aorta and colonize caudal hematopoietic tissue by 48 hpf (Figure 3) [30]. Th is tissue, also known as the caudal vein plexus, is highly vascularized. From 48 hpf to 7 days post-fertilization (dpf ), cmyb + HSCs and HPCs expand and diff erentiate into erythrocytes, monocyte/macrophages, and thrombocytes in caudal hematopoietic tissue (Figures 2 and 3). Th erefore, caudal hemato poietic tissue exhibits properties similar to those of the mammalian fetal liver. Next, HSCs and HPCs migrate from the caudal hematopoietic tissue fi rst to the thymus and then to the pronephros [21][22][23] (Figure 2). CD41 + HSCs and HPCs fi rst appear in the thymus at 54 hpf (2.25 dpf ) [23], and by 4 dpf, cmyb + HSCs and HPCs appear in the pronephros (Figure 2), which later develops into the kidney and functions equivalently to mammalian BM [21,22]. Two routes of HSC migration from the AGM-like region to the pro nephros have been proposed: the fi rst to caudal hemato poietic tissue and pronephros via the circulation (Figure 3) and the second directly from the AGM-like region via the pronephric tubules [30]. At 4 dpf, only myeloerythroid lineages have progressively expanded in the kidney [15]. By 7 dpf, erythroblasts are found in the kidney [15], where they later become the major defi nitive hematopoietic organ of adult zebrafi sh.

Erythropoiesis in adult zebrafi sh
In mammals, adult erythropoiesis is maintained primarily in the BM. HSCs diff erentiate into erythroid progenitor cells and later erythroblasts. Subsequently, erythroblasts undergo terminal diff erentiation into mature erythrocytes. Mature erythrocytes are spherical and biconcave with a typical size of 7 to 8 μm. As in embryos, cellextrinsic cues, such as erythropoietin, and cell-intrinsic cues, such as erythroid-specifi c transcription factors Gata1 and Kruppel-like factor 1 (Klf1), coordinately regulate erythropoiesis in the BM [31].
Unlike in mammals, zebrafi sh erythropoiesis is maintained in the interstitium of the anterior and posterior kidney [2]. In adult kidney marrow, common myeloid progenitors diff erentiate into megakaryocyte/erythroid progenitors expressing the transcription factor gata1a, an ortholog of mouse Gata1. Th e megakaryocyte/erythroid progenitors diff erentiate into erythroblasts. Erythroblasts later diff erentiate into mature erythrocytes, which are elliptical and nucleated cells with a typical size of 7 × 10 μm [32].
Although kidney marrow erythrocytes can be fractionated from blood cell mixtures by fl ow cytometry based on forward and side scatter (which reveal cell size and granularity, respectively), their surface markers have not been fully identifi ed [33]. Although some potentially useful antibodies cross-react among species, it remains a challenge to analyze erythroid cells by using antibodybased techniques, owing to a shortage of reagents.
As in mammals, zebrafi sh erythrocytes contain hemoglobin. Human and mouse α-globin and β-globin genes are located on separate chromosomes and arranged in order of embryonic and adult expression. Th e change from embryonic to adult globin expression is known as globin switching [34]. By contrast, zebrafi sh α-globin and β-globin genes are located on the same chromomsome and found in embryonic and adult clusters separated by non-coding genomic DNA [35]. Th e embryonic cluster consists of hemoglobin alpha embryonic-1 (hbae1) and hemoglobin beta embryonic-1.1 (hbbe1.1), whereas the adult cluster contains hemoglobin alpha adult-1 (hbaa1) and beta adult-1 globin (ba1). Moreover, in the adult cluster, globin genes are oriented in a head-to-head pattern: 3'-5' in the case of α-globin and 5'-3' in the case of β-globin genes. Th us, they are transcribed in the opposite direction [36]. High-performance liquid chroma to graphy analysis shows that adult zebrafi sh erythrocytes in peripheral blood contain three major α-globin and two β-globin proteins [36]. As in mammals, globin switching also occurs during zebrafi sh development at a stage 10 dpf [36]. As in embryos, adult zebrafi sh erythropoiesis is regulated by extrinsic and intrinsic cues.

Zebrafi sh erythropoietin
Erythropoietin (Epo) is a glycoprotein crucial for survival and proliferation of erythroid progenitor cells. In mammals, there is only one Epo that is primarily produced from kidney and BM [37]. Binding of EPO to its receptor activates Janus kinase/signal transducer and activator of trans crip tion 5 (JAK/STAT5) signaling pathway, which upregulates anti-apoptotic genes and promotes cell survival [31]. Mammalian EPO is not required for primitive erythro poiesis but is indispensable for defi nitive erythropoiesis [38,39].
Unlike the case in mice, erythropoiesis of both primitive and defi nitive zebrafi sh depends on erythropoietin signaling [18]. Unlike the case with the mammalian Epo gene, there are three splice variants of epo gene in fi sh: epo-L1, epo-L2, and epo-S [40]. epo-L1 and epo-L2 are expressed predominantly in the heart and liver, whereas epo-S is expressed in adult kidney marrow [40,41]. Th e C-terminal amino acid sequences of proteins encoded by these genes are identical, but the N-terminal signal peptides diff er. Misexpression studies in the monkey kidney fi broblast COS-1 cell line indicate that epo-L1 and epo-L2 are secreted but that epo-S is cytosolic [40].
In mammals, decreased blood oxygen because of anemia or hypoxia induces EPO production in the kidney and accelerates erythropoiesis. Similarly, anemia and hypoxia upregulate the expression of zebrafi sh epo mRNA in the heart [18]. Moreover, zebrafi sh erythro poietin signaling requires stat5.1 protein, an ortholog of human STAT5 [18]. Th ese observations demonstrate that epo/ epor function is highly conserved among vertebrates.

Zebrafi sh erythroid transcription factors
In the mouse, Gata1 regulates transcription of erythropoietic genes, including Klf1. Gata1 knockout mice die during gestation because of severe anemia [42]. Like Gata1, mouse Klf1 is essential for defi nitive erythropoiesis in fetal liver [43]. Klf1 reportedly regulates the expression of several erythroid-specifi c genes encoding (a) globin, (b) enzymes for heme biosynthesis, and (c) erythroid membrane and cytoskeletal proteins [44].
Little is known about the role of gata1a in adult erythropoiesis because of the lethal phenotype of null mutants at an early stage of development [17]. In addition, no functional ortholog of mouse Klf1 has been identifi ed in zebrafi sh. Although the zebrafi sh klfd gene is expressed in both primitive and defi nitive erythropoietic organs [45] and the amino acid sequence of zebrafi sh klfd is similar to that of mouse Klf1 [45], there is currently no direct evidence that klfd functions in defi nitive erythropoiesis. Zebrafi sh klf4 knockdown using antisense morpholino oligonucleotides downregulates the expres sion of embryonic β-globin and genes involved in heme bio synthesis, but no defect is seen in defi nitive erythro poiesis [46]. Th erefore, it is unlikely that klf4 is the ortholog of mouse Klf1. Th us, owing to the lack of a model, the roles of gata1a and klfd in adult erythropoiesis have not been clarifi ed. Th ese outcomes prompted us to establish transient and reversible downregulation of both transcription factors in adult zebrafi sh, as discussed in section 3.8.

Zebrafi sh as models of human erythropoiesis-related diseases
Zebrafi sh erythropoietic genes are functionally similar to those expressed in mice and humans and include genes encoding enzymes for heme biosynthesis, structural erythrocyte membrane proteins, epo/epo receptor, and globin [47]. Large-scale mutagenesis of zebrafi sh has enabled the identifi cation of genes regulating hematopoiesis/erythropoiesis [32,48]. Th e human homologs of some of these genes function in hematological diseases [3]. Zebrafi sh are advantageous for evaluating the function of genes underlying erythropoietic disease since fi sh embryos are resistant to severe anemic conditions because of passive diff usion of oxygen into the fi sh. Fish models of human anemias (Table 1) are described below. Many of the following mutants could be useful to test new drugs.

Erythropoietic protoporphyria
Erythropoietic protoporphyria occurs worldwide, has a prevalence of 1:75,000 in Th e Netherlands [49], and is caused by ferrochelatase defi ciency. Ferrochelatase catalyzes the formation of heme by transferring iron to protoporphyrin, a heme intermediate. Mutations in the human ferrochelatase gene promote protoporphyrin accu mulation in the skin, erythrocytes, and liver, resulting in sensitivity to light exposed to the skin or even erythrocytes and skin burning and itching [49]. Among patients, 20% to 60% also exhibit anemia due to decreased heme synthesis and light-dependent erythrocyte lysis. Some patients (1% to 4%) show liver disease due to the accumulation of free protoporphyrin released from lysed erythrocytes [49]. Although the avoidance of sun exposure and treatment with light-protective substances such as β-carotene and melanin in skin can ameliorate symptoms, no curative treatment is yet available [49].
Th e zebrafi sh mutant dracula exhibits a point mutation in the ferrochelatase gene, creating an in-frame stop codon and expression of a dysfunctional enzyme. Dracula fi sh manifest autofl uorescent erythrocytes, light-dependent hemolysis, and liver malfunction, similar to conditions seen in humans [50]. Owing to the transparency of fi sh embryos and protoporphyrin autofl uorescence, protoporphyrin accumulation can be monitored microscopi cally in various organs of an intact fi sh, an analysis impossible in humans and mice, making dracula mutants a suitable model for human erythropoietic proto porphyria.

Hemolytic anemia
Abnormality of erythroid cell membrane leads to massive erythrocyte destruction in the spleen, a condition known as hemolytic anemia. Human hereditary elliptocytosis is characterized by elliptical erythrocytes, in which abnormal cell membranes lead to hemolytic anemia. Human hereditary elliptocytosis occurs worldwide but is prevalent in West Africa [51]. Often patients show no symptoms, and only 10% have mild to severe anemia [51].
In human hereditary elliptocytosis, many patients harbor point mutations in the gene encoding protein 4.1R, a major component of the erythrocyte cytoskeleton that maintains biconcave morphology. Th ese mutations promote decreased protein expression or impair protein interaction with other cyto skeletal proteins [52]. Owing to massive hemolysis, patients with hereditary elliptocytosis have complications such as cardiomegaly, splenomegaly, and gallstones. Only supportive treatments, such as folate therapy, blood trans fusion, splenectomy, and gallstone removal, are currently available [51,53].
Although HSC transplantation is one curative therapy, novel drugs are needed to antagonize hemolysis. Th e zebrafi sh merlot mutant exhibits severe hemolytic anemia due to mutation in the gene encoding 4.1R protein.
Unlike mammalian erythrocytes, wild-type mature zebrafi sh erythrocytes exhibit both spherical and elliptical morphologies. Merlot mutants show spiculated erythrocyte membranes, resulting in hemolytic anemia and conditions such as cardiomegaly and splenomegaly, pheno types similar to those seen in humans [54].

Congenital dyserythropoietic anemia type II
Human congenital dyserythropoietic anemia type II is an erythroid-specifi c abnormality in cell division, leading to multinuclear erythroblasts, erythroblast apoptosis (dyserythro poiesis), and anemia. Th e condition varies from mild to severe: approximately 15% of patients require blood transfusions during infancy and early childhood but not thereafter [55]. Splenomegaly occurs in 50% to 60% of patients, and gallstones are frequently observed. In the BM, 10% to 45% of erythroblasts are bi-and multinucleated [55]. Dyserythropoiesis is caused by a mutation in the anion exchanger protein band 3, which is present in the human erythrocyte membrane [55]. Th e zebrafi sh retsina mutant exhibits erythroid-specifi c defects in cell division because of mutation in the fi sh ortholog of the band 3 gene. Th ese defects resemble those of the human disease [56]. Severely aff ected anemic patients require blood transfusion and HSC transplantation [57].

Hereditary spherocytosis
Human hereditary spherocytosis is a hemolytic anemia common in Caucasians and has a prevalence of 1:2,000 to 1:5,000 [53]. Hereditary spherocytosis is characterized by abnormal erythrocyte morphology. Normally, the shape of human erythrocytes is a biconcave disk. In hereditary spherocytosis, the erythrocytes exhibit a spherical shape, leading to their massive destruction in the spleen. Muta tion in the gene encoding the cytoskeletal protein spectrin has been identifi ed as a cause of human here di tary spherocytosis. Erythroid spectrin stabilizes mem brane bilayers [58]. Anemia ranging from mild (blood trans fusionindependent) to severe (blood transfusion-depen dent) is the main clinical feature of this condition. Patients also exhibit hyperbilirubinemia, causing jaun dice, and splenomegaly. Th e zebrafi sh riesling carries a mutant β-spectrin gene and exhibits anemia due to erythrocyte hemolysis, similar to con ditions seen in humans [59]. Zebrafi sh β-spectrin shares 62.3% identity with the human ortholog. In addition to exhibiting abnormal cell morphology-induced hemolysis, zebrafi sh riesling erythrocytes undergo apoptosis, which had not been observed in human hereditary sphero cytosis. Th us, analysis of the zebrafi sh riesling mutant has revealed a novel mechanism of erythrocyte hemolysis [59].

Congenital sideroblastic anemia
Human congenital sideroblastic anemia is characterized by iron deposition in mitochondria of erythroblasts in the BM and is caused by a mutation in the gene encoding δaminolevulinate synthase, or ALAS2, which catalyzes the fi rst step of heme biosynthesis. A lack of heme promotes increases in free iron levels and subsequent iron deposition in erythroblast mitochondria, causing insuffi cient production of mature erythrocytes [60]. Generally, patients have symptoms of anemia, such as skin paleness, fatigue, dizziness, and enlargement of the spleen and liver. In addition to a decrease of mature erythrocytes, hypochromic microcytic erythrocytes are observed in the patients. Heme reportedly promotes gene expression of βglobin through binding Batch1, a transcriptional repressor of β-globin gene [61,62]. Th ere fore, decreased intracellular heme because of mutated ALAS2 may contribute to hypochromic microcytic anemia. Mutation in the zebrafi sh gene sauternes, which encodes alas2 protein, results in a condition similar to hypochromic microcytic anemia in humans [63]. Sauternes mutant embryos show delayed erythrocyte maturation and decreased β-globin expression. Th ese mutants represent the fi rst animal model to allow the investigation of mechanisms underlying hemeinduced globin synthesis.

Hypochromic anemia
Hypochromic anemia is a general term for anemia in which erythrocytes look paler and smaller than they do normally. In humans, hypochromic anemia results from reduction in either globin synthesis (as occurs in thalassemia) or iron absorption (as occurs in iron defi ciency) or from vitamin B 6 defi ciency. In mammals and fi sh, diff erent globin subtypes are expressed in embryonic and adult stages. In adult mammals, embryonic globin synthesis is suppressed whereas adult globin synthesis is activated [34]. Th alassemic patients show reduced adult globin synthesis and re-activate fetal globin expression, which can ameliorate anemia severity. Th e zebrafi sh zinfandel mutant exhibits hypochromic microcytic anemia due to defective embryonic globin production. Th is condition is rescued in adult fi sh once adult globin is produced [35].

Type IV hemochromatosis (iron overload)
Type IV hemochromatosis, or iron overload, is characterized by increased intestinal iron absorption and progressive iron deposition in various tissues, resulting in hepatic cirrhosis, arthritis, cardiomyopathy, diabetes, hypopituitarism, and/or hyperpigmentation [64]. Its primary cause is mutation in genes encoding membrane proteins functioning in iron transportation, such as hemochroma tosis protein, the transferrin receptor, and hemojuvelin. Analysis of zebrafi sh weissherbst mutants revealed mutations in a novel iron transporter, named ferroportin 1 [65]-mutations later identifi ed in patients with type IV hemochromatosis [66,67].

Other anemias
As noted, most zebrafi sh mutant lines with defects in erythropoiesis were generated by mutagenesis by using ethylnitrosourea [32,48]. Th ese pioneering studies provided useful animal models of anemia and identifi ed numerous genes underlying human anemias. None theless, such screens take time and are expensive, prompting us to establish a novel, simpler anemic model.
It is known that temperature regulates hematopoiesis [68]. High temperature increases the number of hematopoietic cells in the BM of rats and the nine-banded armadillo (Dasypus novemcinctus), whereas low temperature has opposite eff ects [68]. However, in nature, zebrafi sh survive in a wide range of temperatures, from 6°C in winter to 38°C in summer [6], making them useful to investigate the eff ect of temperature on hematopoiesis. Previously, we established a novel reversible anemic model by keeping zebrafi sh at 17°C (cold zebrafi sh) [41]. By comparison with fi sh kept at a higher temperature (26.5°C), 'cold' zebrafi sh appear paler starting at the fi rst week of cold exposure (Figure 5a, right panel). Th e kidney Similarly aged zebrafi sh kept in 17°C water for 7 months exhibit abnormal RTs (right). Erythrocyte clusters (arrow) were observed only in the kidney marrow of cold zebrafi sh. (c) Expression of erythropoiesis-related genes in the kidney marrow of normal (26.5°C, white bars) and cold-exposed (17°C, gray bars) zebrafi sh. Gene expression was quantifi ed by real-time polymerase chain reaction during the second week of cold exposure. Expression levels of genes encoding transcription factors important for erythropoiesis (gata1a and klfd), the adult hemoglobin gene α-globin (hbaa1), and β-globin (ba1) were downregulated, whereas the epo and epor expression levels were slightly decreased. Low expression of hbaa1, ba1, klfd, and gata1a indicates anemic status of cold zebrafi sh. These frames are modifi ed from our previous report [41]. ba1, globin ba1; epo, erythropoietin; epor, erythropoietin receptor; gata1a, GATA-binding protein 1a; hbaa1, hemoglobin alpha adult-1; klfd, Kruppel-like factor d. These fi gures are modifi ed from [41].
marrow of cold zebrafi sh shows abnormally structured renal tubules and erythrocyte clusters (Figure 5b, right panel). We examined the expression of genes essential for HSC maintenance, erythropoiesis, and myelopoiesis by real-time polymerase chain reaction (PCR). Interestingly, only HSC-regulated genes (runx1, cmyb, gata2a, and tal1) (data not shown) and erythropoietic genes (gata1a, klfd, epo, epor, hbaa1, and ba1) (Figure 5c) were downregulated the fi rst week of exposure. Th ese genes were expressed at generally lower levels by the second week, suggestive of anemia. However, expression of colonystimulating factor 1a, which is required for macrophage diff erentiation, and of colony-stimulating factor 3 (granulocyte), which is required for granulocyte diff erentiation, was unchanged [41]. Th ese results suggest that cold exposure specifi cally suppresses erythropoiesis. Moreover, cold-induced anemia was reversed when fi sh were returned to 26.5°C conditions, implying that phenotypes are directly caused by temperature [41].
To further identify novel genes regulating erythropoiesis mediated by cold exposure, we assessed global changes in gene expression in the kidney marrow by DNA microarray (unpublished data). Although decreased temperature suppressed enzymatic activity and resulted in a global decrease in gene expression, we observed both increased and decreased gene expression in the kidney marrow of cold zebrafi sh, as confi rmed by real-time PCR. We are now undertaking loss-of-function analysis of candidate factors by using antisense morpholino oligonucleotides and small interference RNAs in fi sh, and we are conducting gain-of-function analysis in mouse and human cell lines.

Conclusions
Defects in erythropoiesis result in various anemic diseases. To gain an understanding of these diseases, it is necessary to determine how erythropoiesis is regulated in normal conditions as well as to establish in vivo models. We propose that, in addition to several anemic mutants, our cold zebrafi sh model is a useful tool to explore novel genes functioning in erythropoiesis. Given the advantages of zebrafi sh models, it is feasible to reach these objectives. Knowledge of the etiology and molecular mechanisms underlying these conditions will lead to the development of novel therapies.