Characterization of zebrafish mutants with defects in embryonic hematopoiesis

Vertebrate hematopoietic cells are derived from self-renewing multipotential stem cells that are specified early during embryonic development (reviewed by Zon, 1995). Hematopoiesis involves the processes of cell proliferation, cell differentiation, commitment to particular cell lineages and protection from apoptosis. The lineages and stages of normal hematopoietic development in humans and mice have been well-characterized by in vivo transplantation and by in vitro culture assays (Moore and Metcalf, 1970). Multipotential stem cells differentiate into progenitors committed to erythroid, lymphoid or myeloid lineages. Stem cells and progenitors are rare components of bone marrow and have been defined by functional assays rather than morphology. In the case of the erythroid lineage, progenitors differentiate into precursors that progress from proerythroblast stages to erythroblasts and finally differentiate into mature erythrocytes. The morphology of blood cell precursors and mature cells have been extensively studied by light and electron microscopy in a wide range of vertebrates, including fish (Catton, 1951, Rowley et al., 1988, Glomski, et al., 1992). The identification of genes required for hematopoietic development is of considerable interest particularly given the clinical significance of factors that regulate blood cell production.

In mammals, hematopoiesis first occurs in ventral blood islands located on the yolk sac. This embryonic wave of differentiation is followed by a switch to fetal hematopoiesis in the liver and finally adult hematopoiesis in the bone marrow (Stamatoyannopoulos et al., 1987). Teleosts such as the zebrafish (Danio rerio) lack yolk sac blood islands and instead have a distinct site of embryonic erythropoiesis, called the intermedi-ate cell mass (Al-Adhami and Kunz, 1977; Colle-Vandevelde, 1963; Detrich et al., 1995). The intermediate cell mass is formed during early gastrulation from two bilateral stripes of paraxial mesoderm that converge medially in the tail and posterior trunk. By 20 hours of development, these stripes have fused into a single intraembryonic hematopoietic region located between the notochord and endoderm (Detrich et al., 1995). This region contains a population of presumptive stem cells that express the hematopoietic transcription factor GATA-2 and a more differentiated population of erythroid precursors that also express GATA-1. Embryonic/primitive erythroid progenitors develop to the early proerythroblast stage and are released into circulation. As development proceeds over the next 5 days, these cells undergo a nearly synchronous wave of differentiation to form embryonic erythrocytes (Al-Adhami and Kunz, 1977). During early larval development, zebrafish switch to adult/definitive hematopoiesis in the spleen and kidney rather than the bone marrow (Al-Adhami and Kunz, 1977).

Recently, several genes isolated by their biochemical or functional properties were found to be required for hematopoiesis by gene-targeting experiments in murine embryonic stem cell lines (Ness and Engel, 1994; Weiss and Orkin, 1995). Many of these genes appear to be required for fetal/adult hematopoiesis and not for the earlier embryonic wave of hematopoiesis. Mutations in the transcription factors myb (Mucenski et al., 1991), EKLF (Perkins et al., 1995), and Pu.1 (Scott et al., 1994), the receptor tyrosine kinase W (c-kit; Bernstein et al., 1990; Nocka et al., 1990b), and the growth factor Steel (c-kit ligand; Brannan et al., 1991; Huang et al., 1990; Nocka et al., 1990a) all cause defects in definitive hematopoiesis that takes place in the fetal liver and bone marrow, but primitive embryonic hematopoietic cells of the yolk sac are normal.

Thus far, five genes have been demonstrated to be essential for both primitive and definitive hematopoiesis in mice. Two members of the GATA family of transcription factors, GATA-1 and GATA-2 are required for the production of normal numbers of erythroid cells. GATA-1 deficient cells do not differentiate into embryonic erythroid cells in chimeric mice (Pevny et al., 1991). Progenitor studies of hematopoietic cells in GATA-1-deficient ES cell lines has demonstrated an arrest at the proerythroblast stage (Weiss et al., 1994). GATA-2 mutant mice (Tsai et al., 1994) have a decreased number of erythroid progenitors. The transcription factors SCL (Shivdasani et al., 1995) and TTG2 (Warren et al., 1994) are also required for embryonic hematopoiesis in mice. Mutants embryos with disruptions of these genes do not produce hematopoietic stem cells in the early embryo. Mice that lack the receptor tyrosine kinase FLK-1 have defects in both hematopoiesis and vasculogenesis (Shalaby et al., 1995). FLK-1 mutant mice fail to form yolk sac blood islands and have greatly decreased numbers of vascular endothelial cells. There are also a number of human hematopoietic disorders such as congenital dyserythropoietic anemia and Diamond Blackfan anemia, for which candidate genes have not been defined (Lux, 1995).

We are using the zebrafish as a model system to identify and characterize mutants of hematopoiesis. The optical clarity of the zebrafish embryo allows observation of the relative number and color of blood cells from the onset of circulation (Mullins et al., 1994). These observations serve as qualitative measurements of progenitor proliferation and erythroid differentiation. Blood cells obtained from zebrafish embryos display stagespecific cell morphological and histochemical characteristics that can be used to assess the developmental steps affected by the mutations.

Three zebrafish mutants have recently been investigated for their defects in hematopoietic development. Mutations in the gene cloche (clo^m39), affect early hematopoietic stem cell or progenitor populations. Homozygous clo^m39 mutant embryos do not express detectable levels of GATA-1 and GATA-2 in the intermediate cell mass (Stanier et al., 1995) and lack the endocardial lining of the heart. This represents the first identification of a gene that is essential for hematopoiesis in the zebrafish embryo. Other mutants such as spade tail (Ho and Kane, 1990) and sort-of-bloodless (W. Detrich and L. Zon, personal communication) express GATA-2 in early erythroid progenitors, but lack differentiated erythroid cells that express GATA-1. These mutants demonstrate that the zebrafish can be used to characterize the functions of genes that regulate early stem cell or progenitor differentiation.

In a large scale screen for mutations affecting development of the zebrafish (described by Haffter et al., 1996), we have isolated 33 mutants with defects in the formation or differentiation of hematopoietic cells. In this report we present a phenotypic analysis of these mutants. The zebrafish mutations presented here provide powerful tools for defining the molecular components involved in normal vertebrate hematopoietic development.

Zebrafish were bred and maintained as described by Mullins et al. (1994). The isolation of mutants and complementation analysis was performed as described by Haffter et al. (1996).

Whole embryo staining for heme expression was performed using odianisidine histochemistry according to previously described methods (Detrich et al., 1995; Iuchi and Yamamoto, 1983). After staining, embryos were postfixed in 4% paraformaldehyde for 1 hour.

Whole-mount in situ hybridization with digoxigenin-labeled RNA probes was performed as previously described (Schulte-Merker et al., 1992). Synthesis of GATA-1 and GATA-2 antisense RNA probes was performed as described by Detrich et al. (1995). Embryos were dehydrated in methanol and cleared in 2:1 benzyl benzoate: benzyl alcohol solution before mounting.

Photographs were taken on a Microphot-FXA photomicroscope or a SMZ-U dissecting microscope (Nikon) and processed using Photoshop 3.0 (Adobe).

Blood cells were collected by cardiac puncture on anesthetized embryos in a solution of 0.02% tricaine (Sigma), 1% bovine serum albumin in calcium- and magnesium-free phosphate-buffered saline, pH 7.4. Collected cells were concentrated by centrifugation at 700 rpm for 3 minutes onto slides using a Cytospin 3 (Shandon). Slides were then immediately fixed in methanol for 30 seconds and stained by the Wright-Giemsa method (Nelson and Morris, 1984).

Mutant embryos were examined under 330 to 380 nm ultra violet (UV) light using a Microphot photomicroscope (Nikon).

We have identified 33 mutants with defects in the development of embryonic and larval hematopoietic cells. Complementation analysis placed 32 of these mutants into 17 complementation groups (Table 1). One mutant remains to be fully tested. Most of the 17 genes represented by the complementation groups were given names of wines, which reflects the classification of both by red and white colors. The exception is moonshine that was named for the rays of light reflected by the increased number of iridophores (Kelsh et al., 1996).

The blood displays a smaller number of morphological landmarks than most other developing tissues that are visible under the dissecting microscope. However, all of the circulating blood cells are visible during embryonic development. We have used the number of blood cells and the relative concentration of hemoglobin, as judged by the intensity of red cell color, as the primary criteria for the classification of blood mutants.

In addition, individual embryos were bled and peripheral blood cells were deposited on slides by centrifugation and stained by the Wright-Geimsa method (Nelson and Morris, 1984) to evaluate the morphology of circulating cells in greater detail. The classification of normal teleost blood cells into developmental stages based on morphology is well established (Catton, 1951, Rowley et al., 1988, Glomski et al, 1992). The morphology of erythroid cells in zebrafish has also been previously described and follows the common vertebrate pattern (Al-Adhami and Kunz, 1977). Proerythroblasts are round cells with large open nuclei and basophilic cytoplasm that gradually differentiate into mature erythrocytes which retain a compact nucleus and have cytoplasm filled with hemoglobin.

We also examined the mutants using o-dianisodine staining (Detrich et al., 1995; Iuchi and Yamamoto, 1983), which detects heme present in hemoglobin, and whole embryo in situ analysis to detect the expression of the hematopoietic transcription factors GATA-1 and GATA-2 (Detrich et al., 1995). Based on these analyses, we categorized the zebrafish mutants with hematopoietic defects into 4 phenotypic classes described below.

The 7 stronger alleles of moonshine (mon), including mon^tb222b and mon^tc246b, are the only mutations among the 33 that can be classified as bloodless at the onset of circulation. All of the mutant mon alleles are recessive, cause severe anemia and are lethal after 5 to 7 days except for mon^tc239b. Mutant mon^tc239b embryos have several dozen blood cells at the start of circulation but never produce more than approximately 100 cells.

These less severely affected larvae also form swim bladders but do not survive to adulthood. Mutants of the other seven alleles of mon have few if any cells in circulation initially but have 50 to 100 blood cells by 5 days of development. In addition to the hematopoietic defect, the mon mutants exhibit defects in fin morphology and in pigment cell proliferation and migration. The tail fins of mutants have jagged edges, visible in 2-day and older embryos but are not obviously necrotic. The mon mutants also have an abnormal pattern of iridophore proliferation and migration while melanophores are normal (described in Kelsh et al., 1996). The intensity of the blood phenotype correlates with the strength of the fin and the iridophore phenotypes. The weakest allele mon^tc239b produces the weakest blood, fin and iridophore phenotype.

We examined the expression patterns of the hematopoietic transcription factor GATA-1 in all of the 33 mutants and also tested the mon mutants for GATA-2 expression. The more severely affected mon mutants were the only ones with detectable differences in GATA-1 expression (Fig. 1A-D). Levels of GATA-2 expression in mon mutant embryos at the 18-somite stage are equivalent to those in wild type. 29% of the embryos from a mating between carriers of a strong allele (mon^tb222b) did not express detectable amounts of GATA-1 (Fig. 1B). 23% of the embryos derived from a mating between carriers of an intermediate allele (mon^tc246b) expressed reduced levels of GATA-1 (Fig. 1C), and all embryos from a cross between carriers of a weak allele (mon^tc239b) exhibited wild-type levels of GATA-1 (Fig. 1D). The expression of GATA-1 therefore correlates with the strength of the phenotype.

All of the mon mutants have 50 to 100 cells in circulation by 5 days of development in comparison to the 1000 to 3000 differentiated red blood cells found in wild-type larvae. These wild-type cells have a typical nucleated erythrocyte morphology in Wright-Giemsa stained preparations of isolated blood (Fig. 1E). The rare mutant cells such as those found in mon^tb222b mutants have a proerythroblast morphology with basophilic cytoplasm and non-condensed nuclei (Fig. 1F). The mon mutants also have no detectable expression of hemoglobin based on o-dianisidine staining. 2-day old wild-type embryos have robust circulation and express large amounts of hemoglobin (Fig. 1G) while mon^tb222b mutants have no detectable staining for hemoglobin (Fig. 1H). In summary, these data suggest that mutations in the gene moonshine affect the ability of hematopoietic stem cells to differentiate into erythroid progenitors that express GATA-1 and also affect a subset of pigmented neural crest cells.

Fourteen mutations cause severe anemia with initially normal levels of erythroid cells in circulation which then rapidly decrease in number. In these mutants, the level of hemoglobin expression decreases consistent with the peripheral blood cell number. Such a correlation is well documented in humans with iron deficiency anemia due to bleeding (Lux, 1995). The largest number of mutations are in this category, including chablis (cha, 2 alleles), frascati (frs, 4 alleles), merlot (mot, 2 alleles), retsina (ret, 2 alleles), thunderbird (tbr, 1 allele), riesling (ris, l allele), cabernet (cab, l allele) and grenache (gre, l allele). All of these mutants are difficult to distinguish from wild-type sibs during the first 2 days of development. They all have normally pigmented blood visible in living embryos and o-dianisidine stained whole-mounts indicating that their erythroid cells, although reduced in number, produce hemoglobin. All of these mutations are recessive and lethal in the first 1 to 2 weeks of larval life except a mutation in the gene thunderbird, tbr^ty118b, which is homozygous viable.

Mutations in the genes frs and ret cause severe anemia, visible after hatching, but display different types of morphological defects in the remaining erythroid cells (Fig. 2). Both groups of mutants have wild-type levels of GATA-2 and GATA-1 expression in 1-day old embryos (data not shown). The 4 mutations in frs vary with respect to the onset of the decrease in blood count. The phenotypes of the strongest mutations frs^tg221 and frs^tq223 are visible around hatching. The phenotypes of the slightly milder mutant alleles, frs^tg280a and frs^tm130d, are visible after 3 to 4 days of development. All of these mutations are lethal and result in few if any cells in circulation in 4-day old larvae. 2-day old wild-type embryos have cells in circulation with the color and morphology characteristic of erythroblasts with smaller more condensed nuclei than proerythroblasts (Fig. 2A). These cells also express large amounts of hemoglobin as measured by o-dianisidine staining (Fig. 2B). Blood collected from 2-day old frs^tq223 mutant embryos revealed cells at various stages of erythroblast maturation and none that have the morphology of differentiated erythrocytes. The cells are large with basophilic cytoplasm and many of the nuclei are abnormally fragmented into patches of chromatin (Fig. 2C). Whole-mount staining with o-dianisidine indicates that these cells, though reduced in number still express some hemoglobin (Fig. 2D). Rare frs mutant fish live 3 to 4 weeks and are severely anemic.

Two mutations in the gene retsina, ret^tr217 and ret^tr265 cause a similar decrease in blood count beginning after 3 to 4 days of development. These mutations are lethal during the first week of development. 3-day old wild-type larvae have blood cells with differentiated erythroblast morphology including condensed nuclei and less basophilic cytoplasm (Fig. 2E). These larvae have 1000 to 3000 thousand cells in circulation and stain strongly for hemoglobin (Fig. 2F). The remaining blood cells of ret^tr217 mutant larvae exhibit earlier proerythroblast morphology with large cells and large open nuclei (Fig. 2G). This suggests that they are blocked in erythroid cell differentiation. Whole-mount o-dianisidine staining indicates that the remaining blood cells of ret larvae express hemoglobin similarly to those of frs mutants. Based on the decreasing number of blood cells in this class of mutants and the levels of hemoglobin in the remaining circulating cells (Fig. 2D,H), these mutations may not directly affect the regulation of globin gene expression, but rather affect erythroid cell proliferation and differentiation.

Embryos of 9 mutants initially have normal numbers of blood cells which then decrease, accompanied by reduced levels of hemoglobin expression. These hypochromic blood phenotypes are caused by mutations in any of six genes including chardonnay (cdy 1 allele), chianti (cia, 1 allele), pinotage (pnt, 1 allele), sauternes (sau, 2 alleles), weißherbst (weh, 2 alleles), zinfandel (zin, 1 allele) and 1 unresolved mutation, te264. Three of these mutations, cdy^e216, cia^tu25f and pnt^tq209, are homozygous viable in the adult. The other mutations vary in viability as described below.

Embryos with either of two mutations in the gene weißherbst, weh^th238 and weh^tp85c are difficult to score before they are 2 days old. Around the hatching stage, however, these mutants clearly have hypochromic blood with nearly normal numbers of blood cells. Their blood cell count then decreases in number to approximately 50% of wild-type levels. Both of these mutations are recessive and lethal by 5 to 7 days of development. 2-day old wild-type embryos have circulating erythroblasts with pale blue cytoplasm and condensing nuclei (Fig. 3A). These normal cells express large amounts of hemoglobin as detected by o-dianisidine staining (Fig. 3B). In contrast the weh^tp85c mutant blood cells are not as differentiated and have the morphology of younger proerythroblasts (Fig. 3C). They are much larger than wild type and have large open nuclei typical of more immature erythroid cells. Whole mount o-dianisidine staining demonstrates that even though there are nearly wild-type numbers of cells in circulation the mutant cells express much less hemoglobin (Fig. 3D).

The single mutant allele of zinfandel (zin) also causes a hypochromic blood phenotype in the embryo. It is dominant and can be passaged through either male or female carriers. zin heterozygotes and homozygotes are indistinguishable in phenotype. 2-day old zin mutant embryos have distinctly pale colored blood. By 3 days of development, zin mutant carriers have hypochromic cells in circulation and slightly reduced blood counts. Blood cells isolated from 2-day old zin mutants are oval, larger, paler and have less condensed nuclei than normal 2-day old erythrocytes in Wright-Giemsa stained preparations (Fig. 3E). There is also an increased fraction of large cells with basophilic cytoplasm characteristic of earlier erythroblasts in the zin mutant embryos. The o-dianisidine staining indicates that the blood cells of zin mutant embryos do not express detectable levels of hemoglobin suggesting that they have a defect in globin production (Fig. 3F).

Two recessive mutations in the gene sauternes, sau^tb223 and sau^ty121, were identified that also cause hypochromic blood phenotypes. The sau^tb223 mutation is embryonic lethal while the sau^ty121 mutation is adult viable. Both of these mutations are visible when circulation begins. The blood is hypochromic and initially normal in amount. By 3 days of development, the blood count has decreased to 25 to 50% of wild-type levels. After 3 days of development wild-type embryos have mature erythrocytes in circulation (Fig. 3G) and stain strongly for hemoglobin (Fig. 3H). In contrast, blood cells from sau^ty121 mutant embryos have condensed nuclei like normal erythrocytes but are pale blue rather than red and there are numerous erythroblasts (Fig. 3I). O-dianisidine staining confirms that the sau^ty121 mutants (Fig. 3J) express less hemoglobin than wild-type embryos though relatively more than zin mutant embryos.

The mutations causing hypochromic blood may represent embryonic thalassemias – defects in one of the globin genes, producing a globin chain imbalance that destroys erythrocytes (Nagel, 1995). Alternatively, these may be mutations in genes that regulate hematopoietic cell differentiation.

All of the blood mutants were tested for photosensitivity and autofluorescence in a screen for porphyria syndromes. The offspring of identified heterozygotes were grown either in constant white light or in darkness and compared after 3 to 5 days of development. In a second experiment, all mutants were examined under UV light. This screen identified members of a class of mutants that have phenotypes analogous to the human congenital erythropoietic porphyrias. The blood of wild-type embryos is unaffected by UV light (Fig. 4A) whereas the red blood cells of two porphyria mutants, freixenet (frx^tu271; Fig. 4B) and yquem (yqe^tp61; Fig. 4C) are autofluorescent and bright orange to red.

frx mutant embryos that are exposed to light do not survive beyond 5 to 7 days of development. As little as 2 hours of exposure to ambient laboratory light is sufficient to eliminate all red blood cells from frx embryos that are older than 48 hours. When frx homozygous embryos are exposed to UV light the erythrocytes are orange-red in color (Fig. 4B). After 60 seconds of exposure, the blood cells of frx embryos stop circulating and the embryos die while wild-type siblings are unaffected. If frx larvae are exposed to white light they accumulate dark inclusions in their livers after 5 days of development caused by the accumulation of lysed red blood cells. Under UV illumination, the guts of 5-day old frx larvae are also fluorescent, presumably due to clearance of protoporphyrins from the blood stream.

Mutations in the gene yquem (yqe^tp61) also cause a porphyria syndrome. Under UV light their blood cells are autofluorescent like in frx mutants (Fig. 4C). When grown in constant light, yqe mutant embryos have no circulating cells while 1- and 2-day old frx mutant embryos have 300 to 500 blood cells. In addition, yqe mutant embryos show marked cardiac edema and autofluorescence in a screen for porphyria syndromes. The offspring of identified heterozygotes were grown either in constant white light or in darkness and compared after 3 to 5 days of development. In a second experiment, all mutants were examined under UV light. This screen identified members of a class of mutants that have phenotypes analogous to the human congenital erythropoietic porphyrias. The blood of wild-type embryos is unaffected by UV light (Fig. 4A) whereas the red blood cells of two porphyria mutants, freixenet (frx^tu271; Fig. 4B) and yquem (yqe^tp61; Fig. 4C) are autofluorescent and bright orange to red.

frx mutant embryos that are exposed to light do not survive beyond 5 to 7 days of development. As little as 2 hours of exposure to ambient laboratory light is sufficient to eliminate all red blood cells from frx embryos that are older than 48 hours. When frx homozygous embryos are exposed to UV light the erythrocytes are orange-red in color (Fig. 4B). After 60 seconds of exposure, the blood cells of frx embryos stop circulating and the embryos die while wild-type siblings are unaffected. If frx larvae are exposed to white light they accumulate dark inclusions in their livers after 5 days of development caused by the accumulation of lysed red blood cells. Under UV illumination, the guts of 5-day old frx larvae are also fluorescent, presumably due to clearance of protoporphyrins from the blood stream.

Mutations in the gene yquem (yqe^tp61) also cause a porphyria syndrome. Under UV light their blood cells are autofluorescent like in frx mutants (Fig. 4C). When grown in constant light, yqe mutant embryos have no circulating cells while 1- and 2-day old frx mutant embryos have 300 to 500 blood cells. In addition, yqe mutant embryos show marked cardiac edema and die after 3 to 5 days as a consequence of exposure to light. Both of these mutants display phenotypes that are reminiscent of the erythropoietic porphyria syndromes, which in humans result from defects in heme biosynthesis (Sassas and Kappas, 1995).

We are using the zebrafish as a model system to study hematopoietic cell development in a vertebrate embryo. We have used a mutational approach to identify genes that are indispensable and specifically required for hematopoietic development. The 33 mutants described here display a broad range of phenotypes during the first 5 days of development (Table 1). We have classified these mutations based on complementation analysis and specific phenotypic traits.

The hematopoietic program involves the induction of ventral mesoderm followed by events that regulate cell proliferation, cell differentiation and cell death. The mutations described here characterize several of these landmarks in the development of the embryonic hematopoietic system. The types of cells remaining in these anemic mutants suggest that all of these genes influence the differentiation of erythroblasts to normal erythrocytes. This does not exclude the possibility that these genes may primarily function at much earlier steps in differentiation. Studies with marker genes expressed at different developmental stages are needed to define the steps affected by these mutations.

The blood mutants can be classified as early or late disruptions of the normal hematopoietic program based on their expression of GATA-1 and GATA-2 (Detrich et al., 1995). Most mutants have wild-type levels of GATA-1 expression, suggesting that these mutants represent later defects in embryonic hematopoiesis. Of the mutations described here, only strong alleles of moonshine (mon) cause a lack of GATA-1 expression. Since mon embryos have wild-type levels of GATA-2 expression, but not GATA-1 expression, we conclude that this mutant has early hematopoietic cells that are blocked in differentiation. It is tempting to speculate, that mon may directly control the level of GATA-1 expression, although further studies would be required to confirm this hypothesis.

With regard to erythroid cell proliferation, there are a number of mutations including those in the genes chablis, frascati, grenache, merlot, pinotage, retsina, thunderbird, weißherbst and zinfandel that result in initially normal blood cell counts followed by severe anemia. This sudden drop in cell number indicates a destruction of the mutant blood cells. In mutants of the genes weißherbst and zinfandel, the number of circulating cells in 2-day old embryos is almost equivalent to wild type, but the level of globin expression is substantially decreased and the blood appears hypochromic. These mutations may be defects in globin genes or in globin gene regulation similar to the thalassemias found in humans (Nagel, 1995; Orkin, 1995b). Thalassemias are syndromes that result from abnormal globin synthesis. Unpaired or mutant globin chains can precipitate in erythrocytes and cause cell lysis (Nagel, 1995). Studies of peripheral blood cells of embryos mutant for weißherbst or frascati revealed cells that are blocked at the proerythroblast stage of erythroid differentiation. Regarding cell death, mutants such as frascati have blood cells with abnormal characteristics of the nucleus and vacuolated cytoplasm. It will be interesting to investigate the mechanism of cell death involved in the elimination of cells that fail to undergo the normal maturation process.

A number of mutants described in this report are severely anemic as embryos and larvae but develop into viable adults. These include mutants in chardonnay, chianti, pinotage, sauternes, thunderbird and zinfandel. This suggests that we may have identified genes that are specifically required for hematopoiesis in the embryo and not in the adult. Alternatively, these genes may be required for normal blood cell development in both embryos and adults. Further studies of adult fish that are homozygous for these mutations are underway to determine if they display defects in the adult hematopoietic system.

The hematopoietic mutants generated in this large scale screen may represent defects in novel genes or defects in homologous genes known to be required for normal hematopoiesis in mice or humans. Mutation in several known mouse genes cause phenotypes with similarities to the zebrafish mutants. Based on the synchronous wave of erythroid differentiation of circulating blood in the zebrafish, blocked erythroid maturation that is first visible after 2 days of development might be the phenotype of a GATA-1 defect (Pevny et al., 1991; Weiss et al., 1994). It is more difficult to predict a phenotype for zebrafish mutants deficient in GATA-2 expression. A defect in the GATA-2 gene, which in mice affects erythroid cell proliferation (Tsai et al., 1994), could cause relatively late phenotypes such as decreasing erythroid cell number in the zebrafish embryo. Alternatively, based on the expression in the putative hematopoietic stem cells, a mutation in GATA-2 could result in an earlier defect in the hematopoietic system of the zebrafish. Zebrafish mutants with defects visible on the first day or two of development could be analogous to the murine SCL (Shivdasani et al., 1995) and TTG2 mutants (Warren et al., 1994). Although the zebrafish mutants described here do not have vascular defects, it is possible that particular mutant alleles of FLK-1 could produce a hematopoietic phenotype.

The combination of hematopoietic and neural crest mutant phenotypes in mon mutants is reminiscent of the W and Steel mutations in mice, which are defects in c-kit and the c-kit ligand, respectively (Bernstein et al., 1990; Brannan et al., 1991; Huang et al., 1990; Nocka et al., 1990a,b). Mice with mutations in either of these two genes have dramatic reductions in the proliferation and migration of pigmented neural crest cells (melanocytes) as well as defects in the differentiation of definitive adult hematopoietic stem cells. However, the analogy between these genes and mon is not complete. mon mutants have no obvious defects in melanophore pigmentation (Fig. 1G) compared to wild type (Fig. 1H). Also, W and Steel mutants do not exhibit severe defects in embryonic erythropoiesis like those caused by mutations in mon.

Other murine mutations that predominantly affect definitive and not embryonic hematopoiesis are less likely to have counterparts among the zebrafish mutants described here. These included mutations in the transcription factors myb (Mucenski et al., 1991), EKLF (Perkins et al., 1995; Huang et al., 1990; Nocka et al., 1990a), and Pu.1 (Scott et al., 1994).

These zebrafish mutants have defects that also resemble diseases of the hematopoietic system in human congenital disorders. As described above, some of the hypochromic mutants may represent embryonic thalassemias (Higgs et al., 1990). We have isolated and characterized the embryonic globin genes from zebrafish and we are testing polymorphisms for linkage to these mutants (A. Brownlie and L. Zon, unpublished data). Other hypochromic blood mutants may be the counterpart to heme biosynthetic or iron utilization disorders, such as the congenital sideroblastic anemias or iron deficiency (Lux, 1995). For instance frx and yqe mutant embryos have porphyria phenotypes with intense fluorescence in the red blood cells and severe hemolysis upon exposure to light. These mutations may therefore be the equivalent of a congenital erythropoietic protoporphyria or erythropoietic porphyria, although formal biochemical studies of the enzymes and porphyrin bi-products would be required to secure such a diagnosis (Sassa and Kappas, 1995). Some zebrafish mutants had abnormally large erythroid cells reminiscent of the dyserythropoietic or megaloblastic erythroid cells characteristic of such human disorders as congenital dyserythropoietic anemia, Diamond Blackfan anemia, and B12 or folate deficiency (Lux, 1995). It should however be noted that only fetal and adult hematopoiesis have been extensively characterized in human hematopoietic diseases. The zebrafish mutants described here may represent defects of embryonic erythropoiesis, and it is possible that embryonic signaling cascades and enzymes (and not the adult homologs) are affected in these mutants. Nevertheless, the genes identified by the mutations described here are likely to play the same or similar roles in the hematopoietic system of other vertebrates.

We have placed 32 of the 33 mutations affecting the hematopoietic system into complementation groups representing 17 genes. The allelism of te264 remains unresolved to date.

Thus the allele frequency in this group of mutants is relatively low. The large scale mutational screen was not specifically designed to identify mutations affecting the hematopoietic system. Instead, a large variety of mutations affecting different tissues and organs was isolated (Haffter et al., 1996). The mutations described here either affect the number of circulating cells and/or the intensity of the red color of the blood cells. Mutations affecting myeloid or lymphoid lineages specifically or mutations affecting the red blood cell lineage without affecting the cell number or the intensity of the red color were not identified in the screen. This would suggest that additional mutants may be found by performing a screen aimed at systematically identifying the genes involved in the different aspects of zebrafish hematopoiesis.

In this report, we have shown that the zebrafish offers a powerful new method to screen for mutations in genes required for the differentiation of erythroid cells. It will be interesting to determine if any of these mutations function in a cell autonomous manner or whether factors secreted by other tissues are involved in erythroid cell differentiation, given the fact that terminal differentiation of erythroblasts takes place in circulation. Further studies are also necessary to determine what myeloid or lymphoid lineages may be affected by these mutations. We have begun to search for zebrafish molecular markers to identify the various lymphoid and myeloid cell types in order to further characterize these mutants. Characterization of the genes affected by the mutations described in this report will further our understanding of the molecular mechanisms underlying hematopoiesis in the vertebrate embryo.

We thank Leonard Zon for his generous support of D. G. R. and A. B., for critical review of this manuscript and for many helpful discussions. We also thank Lauren Barone and Stephen Pratt for technical assistance. D. G. R. is a postdoctoral fellow of the Howard Hughes Medical Institute.