Non-cell autonomous requirement for the bloodless gene in primitive hematopoiesis of zebrafish

Studies in vertebrates suggest that hematopoietic progenitors are derived from mesodermal tissue of the developing embryo, under the influence of genes such as Bmp4 and Mix1 (Davidson and Zon, 2000; Dzierzak and Medvinsky, 1995; Robertson et al., 1999; Zon, 1995). Subsequently, vertebrate hematopoiesis is characterized by successive waves of development, classified as primitive (embryonic) and definitive (fetal and adult) programs. Primitive and definitive phases of hematopoiesis are often distinguished on the basis of anatomic sites of development, time of initiation, cell types produced and cell morphology. Definitive hematopoiesis produces cells of erythroid, myeloid and lymphoid lineages, whereas primitive hematopoiesis is primarily erythroid with some macrophages produced as well.

Murine primitive hematopoiesis begins in the extra-embryonic mesoderm of the yolk sac around embryonic day 7.5 (E7.5) (Dzierzak and Medvinsky, 1995; Palis et al., 1999; Robb, 1997). In birds, primitive hematopoiesis initiates in yolk sac blood islands that arise in the posterior area opaca at the early somite stage (Dieterlen-Lievre, 1997; Lassila et al., 1982; Peault, 1996; Szenberg, 1977; Zagris, 1986). In amphibians such as Xenopus, blood is first formed in the intra-embryonic ventral mesoderm and migrates to form a ‘V-shaped’ hematopoietic blood island (Kelley et al., 1994; Rollins-Smith and Blair, 1990). In teleosts (bony fish) such as zebrafish, primitive hematopoiesis begins in the intra-embryonic mesodermal tissue called the intermediate cell mass (ICM), which is formed by medial migration of cells in the bilateral lateral plate mesoderm (Detrich et al., 1995; Thompson et al., 1998).

Several studies suggest that the first definitive hematopoietic cells arise from the yolk sac (Godin et al., 1995; Palis et al., 1999; Wong et al., 1986). Studies that correlate stem cell activity from in vitro clonal assays with expression of genes such as scl and gata1 demonstrate that primitive hematopoiesis takes place in the murine yolk sac, where definitive hematopoietic cells also originate (Palis et al., 1999). In mice, yolk sac cells from day 9 embryos can provide long-term multi-lineage reconstitution, which is capable of contributing to mature peripheral blood, thymus, spleen, and bone marrow lymphoid, myeloid and erythroid cell types (Yoder et al., 1997). Subsequently, the site of mouse definitive hematopoiesis moves to the aorta-gonad-mesonephros (AGM) region (Medvinsky and Dzierzak, 1996; Medvinsky et al., 1993; Muller et al., 1994). Multi-potential hematopoietic progenitors have also been detected in murine para-aortic splanchnopleura (Dieterlen-Lievre and Le Douarin, 1993; Godin et al., 1993).

Factors important for the development of both primitive and definitive hematopoiesis such as Scl, Lmo2, Gata1, Gata2 and Flk1 (Kdr) have been described (Pevny et al., 1995; Porcher et al., 1996; Robb et al., 1996; Shalaby et al., 1997; Shivdasani et al., 1995; Tsai et al., 1994; Warren et al., 1994). Genes required specifically for definitive but not primitive hematopoiesis, such as Myb, Kit, Slf (Pou3f4), Tel, Aml1 (Runx1), Cbfb and Epo have also been identified (Lin et al., 1996; Mucenski et al., 1991; Ogawa et al., 1993; Okada et al., 1998; Sasaki et al., 1996; Wang et al., 1997; Wang et al., 1996). However, a specific regulator of primitive hematopoiesis has not been reported. A primitive-specific hematopoietic defect in mammals would probably present with in utero lethality, and would not survive to demonstrate normal definitive hematopoiesis. By contrast, zebrafish is uniquely suited to uncover embryonic bloodless phenotypes, as severely anemic zebrafish embryos can be raised to adulthood in laboratory conditions (Brownlie et al., 1998; Liao et al., 2000a).

The development of hematopoietic precursors is regulated by both extrinsic and intrinsic cues. Several studies describe the role of secreted growth factors such as bmp4 and stem cell factor (Steel) on the induction of blood and regulation of the hematopoietic stem cells, respectively (Mead and Zon, 1998; Whetton and Spooncer, 1998). Key intrinsic factors such as transcription factors Scl and Lmo2 are required for the formation of hematopoietic stem cells, and Gata1 is required for the differentiation of stem cells along the erythroid lineage (Orkin, 1996). In zebrafish, scl has been demonstrated to be sufficient for specifying hematopoietic progenitors (Gering et al., 1998). Overexpression of scl can rescue blood and endothelial cells in the cloche (clo) mutant, which specifically lacks those two cell types (Liao et al., 1998). With respect to hematopoiesis, the clo gene appears to act in a non-cell autonomous manner in the differentiation of embryonic blood cells, where reciprocal transplantation experiments show that wild-type donor cells were less likely to express gata1 in a mutant environment (Parker and Stainier, 1999). In addition, co-transplantation experiments show that clo is required cell autonomously in subsequent proliferation of embryonic blood cells, as wild-type donor cells always contribute a greater number of blood cells than mutant donor cells in the wild-type host.

We report our characterization of bloodless (bls), a dominant zebrafish mutation producing embryos that are severely anemic at the earliest time point that circulation can be detected. Analysis of the expression of early hematopoietic genes show that decreased number of primitive hematopoietic cells are formed from the lateral plate mesoderm. Those blood precursors that are formed fail to differentiate and undergo apoptosis. In addition to an absence of blood cells during embryogenesis, bls mutants also exhibit delayed initiation of lymphopoiesis. However, primitive macrophages develop normally in bls, illustrating distinct developmental regulation of erythroid and myeloid lineages during embryogenesis. Overexpression studies with bmp4, scl and gata1 suggest that bls may regulate primitive hematopoietic cell differentiation or survival, potentially by regulating the expression of scl. Cell transplant experiments between wild-type and bls mutant animals suggest that the bls gene is required in a non-cell autonomous manner for primitive hematopoiesis. Despite the lack of blood cells for the first 4 days of life, hematopoiesis recovers in bls mutants as the definitive blood program replaces the primitive wave. We present the first description of a primitive-specific mutant phenotype and show that the gene product is required in a non-cell autonomous manner for embryonic hematopoiesis, potentially regulating scl expression.

Zebrafish were raised and maintained as described (Westerfield, 1993), and staged as described (Kimmel et al., 1995). The spontaneous bloodless allele, bls^H75 was obtained from Carl Fulwiler and Walter Gilbert (Harvard University Biolabs), isolated in an insertional mutagenesis screen with a lacZ vector. The bls^H75 allele was outcrossed to the standard wild-type strain (AB), and homozygotes were inbred for five generations. Embryos used for in situ and transplantation experiments were collected from inbred bls^H75 homozygotes. The spontaneous cloche allele, clo^m39 (Stainier et al., 1995), was obtained from Mark Fishman (Cambridge, MA). Embryos used for in situ and micro-injection experiments were collected from pairwise matings between identified clo^m39 heterozygotes.

In situ hybridization and riboprobe synthesis were performed as described (Schulte-Merker et al., 1992), with modifications (Liao, 1998). Antisense riboprobes to cmyb, draculin, flk1, gata1, gata2, ikaros, ntl, rag1, scl, shh, spt and tbx6 have been described previously (Detrich et al., 1995; Griffin et al., 1998; Herbomel et al., 1999; Hug et al., 1997; Liao et al., 1998; Thompson et al., 1998; Willett et al., 1999). Digoxigenin or fluorescein-labeled riboprobes were detected with alkaline phosphatase conjugated anti-digoxigenin or anti-fluorescein antibody, respectively. Alkaline phosphatase substrates used to yield crimson, blue and purple are Vector Red, Vector Blue and BCIP/NBT, respectively (Vector Laboratories). To detect biotin-dextran labeled cells, the Vectastain peroxidase kit was used, where colors red and blue were developed using Vector NovaRed and VIP, respectively (Vector Laboratories). Whole-mount immunohistochemistry was performed essentially as described previously (Schulte-Merker et al., 1992) with anti-HCK-1 (1:100) followed by anti-rabbit HRP (1:300).

Acridine Orange staining of live embryos was performed as described, at five-somite, 15-somite, 23 hpf and 36 hpf (Seiler and Nicolson, 1999). Staining of hemoglobin by o-dianisidine was performed as previously described (Detrich et al., 1995). In brief, unfixed embryos were dechorionated and stained for 15 minutes in the dark, with a solution consisting of o-dianisidine (0.6 mg/ml), 0.01 M sodium acetate (pH 4.5), 0.65% hydrogen peroxide and 40% (vol/vol) ethanol. Embryos for histological sections were treated with acetone and embedded in Epon-Araldite (Polysciences) plastic resin, for histological sections.

Micro-injection of plasmid DNA was performed essentially as described (Westerfield, 1993), using a Nikon pico-injector and a Narishige micro-manipulator. The full-length bmp4 was a gift from Masataka Nikaido (Hokkaido University, Sapporo, Japan) and directionally cloned into EcoRI and XhoI of pCS2+ for overexpression. The expression construct for tolloid (pCS2:Ztld-3′MT) was a gift from Patrick Blader (IGBMC, Strasbourg, France), and gata1 (pCS2+) a gift from Sue Lyons (NIH, Bethesda, MD). The expression constructs for scl and GFP control has been described (Liao et al., 1998). Plasmid DNA expression constructs were purified (Qiagen), quantified spectrophometrically, and diluted to 100 ng/μl in sterile double distilled H₂O. Expression constructs of GFP control, bmp4, tolloid, scl and gata1 were micro-injected into wild-type, bls and clo embryos at the two- or four-cell stage. Approximately 50-80 pg of DNA was injected into the blastomeres of each embryo. Embryos were fixed at 23 hpf and analyzed by in situ hybridization using scl, gata1 or globin antisense riboprobes.

Cell transplant studies were carried out as described, with modifications herein specified (Westerfield, 1993). Donor cells were labeled with 1:1 mixture of 5% rhodamine and 5% biotin-dextran, resuspended in 0.2 M KCl. Micro-injection of embryos was performed from one- to four-cell stage, in 1× Danieus media. Donor and host embryos were dechorionated at the 16-cell stage with pronase treatment (4 mg/ml of pronase in 1× Danieus buffer), for exactly 2 minutes, followed by careful and thorough rinses in 1× Danieus buffer. Pronase dechorionation was performed in agarose (1.2% agarose in 1× Danieus) coated petri-dishes, as were all subsequent steps of embryo manipulation. Wild-type embryos used in transplant studies were derived from mating between wild-type AB fish. Mutant embryos used were derived from mating between bls homozygote adults, guaranteeing a clutch of embryos with mutant genotype. At the sphere stage, donor cells (25-40) were transplanted into the margin of sphere stage host animals. Donor and host animals were fixed at 23 hpf for in situ hybridization analysis and detection of biotin-dextran-labeled donor cells.

The bls^H75 allele was isolated as a spontaneous mutation. Although the original allele was discovered in a vector-based insertional mutagenesis screen, the mutation did not track with the lacZ vector (data not shown). The bls mutation is inherited in a dominant manner with incomplete penetrance, where the bloodless phenotype ranges from complete absence of circulating blood cells to severe anemia (10-20 cells). The bls phenotype can be detected morphologically with the onset of circulation at 26 hours post-fertilization (hpf), where the absence of blood cells can be illustrated by o-dianisidine staining (Fig. 1). However, as the mutant embryos continue to develop, they begin to accumulate blood cells after 5 dpf. The timing of the accumulation of red cells correlates with the onset of definitive hematopoiesis in zebrafish, which is believed to initiate around 4 dpf (Brownlie et al., 1998; Liao and Zon, 1999; Thompson et al., 1998). By 7.5 dpf, the embryos that were bloodless during embryogenesis become indistinguishable from wild-type (Fig. 1), and can be raised to adulthood. Peripheral blood and kidney of the adult bls mutants appear normal, and elaborates full range of blood lineages (data not shown). These studies demonstrate that the bls mutation leads to severe embryonic anemia but adult hematopoiesis appears unperturbed.

Unlike a fully penetrant dominant mutation, mating of bls^H75 heterozygote to wild-type did not yield 50% phenotypic mutant progeny as would be expected. Instead, mating of bls^H75 heterozygote to wild type produced approximately 15% severely anemic embryos. Likewise, mating between two bls^H75 heterozygotes produced approximately 60% severely anemic embryos, whereas the expected progeny genotype would have predicted 75% mutants (50% bls^H75 heterozygotes and 25% bls^H75 homozygotes). However, as mutant bloodless embryos can be raised to adulthood, consecutive inbreeding of bls adults was carried out to obtain homozygous lines of bls^H75/bls^H75 mutants. After five generations of inbreeding, mating of bls^H75/bls^H75 with bls^H75/bls^H75 adults produced 100% of bloodless progeny. All of the studies herein reported were carried out using bloodless embryos derived from mating of inbred bls homozygotes.

At the five-somite stage, expression of scl in the lateral plate mesoderm marks the specification of primitive hematopoietic progenitors (Gering et al., 1998; Liao et al., 1998; Mead et al., 1998). By 23 hpf, scl is expressed in the ICM (region overlying the yolk tube and extending slightly caudally, forming a wedge) and an anatomically distinct tailbud derived population, referred to as posterior ICM (Detrich et al., 1995). Expression of scl was greatly reduced in the lateral plate mesoderm of bls mutants, suggesting that the bls gene product participates in the specification of hematopoietic progenitor cells (Fig. 2A). From 18-somite to 23 hpf, scl expression in the ICM of bls mutants decreased progressively, until only a few cells were detectable in the posterior ICM at 23 hpf (Fig. 2A, arrowhead with asterisk).

Expression of gata1 can be detected at the five-somite stage in wild-type embryos (Detrich et al., 1995). Similar to scl, gata1 expression at the eight-somite stage was greatly reduced in bls mutants, ranging from completely absent to a between five and ten detectable cells (Fig. 2A,B). At 23 hpf, wild-type gata1 expression is restricted to the anterior ICM and is not found in the posterior ICM (Detrich et al., 1995). In contrast to scl expression, gata1 transcripts were absent in bls mutants at 18-somites (Fig. 2B). Both scl and gata1 are absent in the anterior ICM at 23 hpf.

Transcripts of gata2 were detected in the wedge region of the anterior ICM and the posterior ICM at 23 hpf, albeit qualitatively reduced when compared with wild type (Fig. 2C, arrowhead and arrowhead with asterisk). With the reduced expression of gata2 in bls, it was difficult to determine by double RNA in situ whether the gata2-expressing cells co-expressed scl. As gata2 is expressed in both blood and endothelial progenitors, the reduced transcript level could be attributable to decreased number of hematopoietic but not endothelial cells. Additionally, transcripts of the zebrafish ikaros gene were detected in the hematopoietic progenitors of the ICM (Fig. 2C). Similar to observations of scl and gata1 expression, ikaros expression was absent in the ICM of bls mutants at 23 hpf.

As the ICM also includes cells that differentiate into the embryonic angioblasts, we examined flk1 expression in bls (Pardanaud et al., 1996). Expression of flk1 delineates the embryonic dorsal aorta, axial vein and inter-somitic vessels at 23 hpf (Fig. 2C). Expression of flk1 in the lateral plate mesoderm at eight somites was also unaffected (data not shown). This finding is corroborated by visual inspection of bls mutants, which showed morphologically intact vasculature.

The expression of scl but not gata1 at 18 somites suggests that the hematopoietic precursors that were specified early in embryogenesis of bls mutants failed to undergo normal differentiation and to maintain gata1 expression. It has been shown that hematopoietic precursors defective in gata1 expression undergo apoptosis (Weiss and Orkin, 1995). Consistent with this, staining of apoptotic cells with Acridine Orange demonstrated increased cell death in the ICM of bls mutants at 15 somites and 23 hpf (Fig. 3A). In addition to the ICM hematopoietic defects, bls embryos also exhibited marked apoptosis of cells lining the dorsal trunk and tailbud margins (Fig. 3A, arrowheads with asterisk).

Given the dorsal position of the Acridine Orange stained cells, we attempted to determine whether they represented dorsal ganglia or Rohon Beard cells. Detection by in situ of neuronal derived dorsal ganglia and Rohon Beard cells with spt in situ and HNK-1 antibody, respectively, failed to uncover any abnormality in bls (Fig. 3B). Likewise, the expressions of spt and tbx6 in the tailbud were not perturbed by bls (Fig. 3B). Of interest, the paraxial mesoderm regulator spt is also expressed in the wedge region of the anterior ICM (Fig. 3B, arrowhead), whereas tbx6 is expressed along the anterior ICM like gata1 and ikaros. The functions of spt and tbx6 in hematopoiesis are unclear, although it has been suggested that the spt gene is required for hematopoiesis (Oates et al., 1999; Thompson et al., 1998). Additionally, to exclude the possibility of more global mesoderm abnormality in bls, we show that the expression of axial mesoderm markers such as no tail (ntl) and sonic hedgehog (shh) in the notochord and tailbud are undisrupted by bls (Fig. 3B). The tissue identity of the dorsal trunk and tailbud margin cells that undergo apoptosis remains uncertain, and the loss of these cells does not appear to affect normal development, as bls embryos develop into adult animals that appear grossly normal.

With the striking deficiency of primitive hematopoietic progenitors in bls, we next examined whether primitive myelopoiesis is also affected by the bls mutation. Myelopoiesis during zebrafish embryogenesis consists of expansion and differentiation of primitive macrophages (Herbomel et al., 1999). Unlike primitive erythroid cells that arise from the ICM, zebrafish primitive macrophages arise from a distinct rostral anlage that is derived from the anterior paraxial mesoderm (Herbomel et al., 1999; Bennet et al., 2001; Parichy et al., 2000). Primitive macrophages originating from the anterior paraxial mesoderm express genes such as pu.1 (spi1), cmyb and draculin (dra). In addition, cells in the ICM also express cmyb and dra but not pu.1 (Fig. 4). When the expression of these genes were examined in bls mutants, it was evident that the primitive macrophages were not affected by the mutation. Expression of pu.1, cmyb and dra were at wild-type levels in the primitive macrophages of bls mutants (Fig. 4).

Like gata1, the expression of cmyb was also absent in the ICM of bls mutants (Fig. 4B). By contrast, dra expression was reduced but not absent in the lateral plate mesoderm (Fig. 4C, asterisk) and persists in a few cells at 20 somites. The reduced expression of dra in bls is similar to that observed for scl. Although the function of dra in hematopoiesis remains to be determined, the similarity of reduced dra expression with that of scl suggests that dra may also be an early marker for hematopoiesis (Alan Davidson and L. Z., unpublished).

The thymic organs form by 65 hpf and are populated with rag1-expressing lymphocytes (Hansen and Zapata, 1998; Trede and Zon, 1998). Expression of rag1 is absent in bls embryos at 4.5 dpf (Fig. 5A). Thymic cytology revealed normal appearing thymic epithelium in bls and absence of lymphoblasts (Fig. 5B, arrowhead with asterisk). Interestingly, if mutant animals were raised to 7.5 dpf and then analyzed with rag1 in situ hybridization, lymphoid cells can be detected in the thymi (Fig. 5A, arrowhead with asterisk). Likewise, histology of thymi of wild-type and bls mutants appear similar at 7.5 dpf, populated with lymphocytes (Fig. 5B, arrowheads).

To place bls in the context of other known genes that participate in blood formation, we overexpressed bmp4, scl and gata1 into wild-type, bls and cloche (clo) zebrafish embryos. Injection of cmv-GFP control plasmid did not perturb normal development of the embryos, and did not lead to rescue of scl or gata1 expression in either bls or clo mutants (Fig. 6A). In the control injected bls mutants, scl expression is present in the posterior ICM at 23 hpf, as is seen in uninjected embyros. However, scl and gata1 are not detected in the anterior ICM of control injected embryos.

When bmp4 was injected into wild-type embryos, ventral mesoderm derived ICM was expanded, expressing high levels of scl and gata1 (Fig. 6B). In the bls mutants, bmp4 ventralized animals lacked scl expression in the anterior ICM, although the posterior ICM expression of scl was moderately increased (Fig. 6B, arrowhead with asterisk). As the posterior ICM normally expresses scl in the bls mutant, only the presence of scl-positive cells in the anterior ICM is scored as hematopoietic rescue (Table 1). Additionally, only rare gata1 expressing cells were detected in the ICM of bmp4 ventralized tails of bls animals (Table 1). Furthermore, bmp4 failed to induce scl or gata1 expression in clo mutants (Fig. 6B). In contrast to bls, the ventralized posterior ICM of clo did not express scl. The same results were obtained when the endogenous level of bmp4 was increased with tolloid overexpression, which antagonizes chordin degradation of bmp4 (data not shown) (Blader et al., 1997).

When scl was expressed in wild-type animals, the ICM was slightly expanded (Fig. 6C) (Gering et al., 1998). In clo mutants, scl expression can lead to partial rescue of hematopoietic progenitors in the anterior ICM, suggesting that scl is sufficient to specify hematopoietic progenitors (Fig. 6C, arrowhead with asterisk) (Liao et al., 1998). When scl was overexpressed in bls mutants, partial rescue of gata1-expressing cells was detected in the anterior ICM (Fig. 6C, arrowheads). The gata1-expressing cells do differentiate to mature red cells as shown by globin staining (data not shown). This result is in contrast to bmp4 overexpression, which was unable to produce gata1-positive cells despite ventralization of the ICM. Although the percentage of animals exhibiting partial blood rescue with scl overexpression is similar between bls and clo, the number of cells rescued per embryo is significantly less in bls than in clo (Table 1). Furthermore, no rescue of blood cells was observed with overexpression of gata1 (Fig. 6D).

To determine whether bls acts in a cell autonomous or non-cell autonomous manner, reciprocal cell transplantation experiments were carried out between wild-type and bls mutant animals (Table 2). Analysis of gata1 expression in bls hosts carrying either bls or wild-type donor cells failed to detect any hematopoietic progenitors in the ICM (Fig. 7). By contrast, when bls mutant donor cells were transplanted into wild-type hosts, gata1-expressing hematopoietic progenitors derived from the mutant donor cells were found in the ICM (Fig. 7, white arrowheads). These studies demonstrate that bls acts in a non-cell autonomous manner, such that the mutant cells are competent to differentiate and express gata1 in a wild-type environment.

Our characterization of bls demonstrates that the mutation acts at an early developmental window of primitive hematopoiesis. Lacking a functional bls gene product, the number of hematopoietic cells specified from the lateral plate mesoderm is severely reduced, illustrated by the decreased number of scl-positive cells at eight somites. The primitive blood cells that are formed fail to differentiate, gata1 expression is not maintained, and the cells undergo apoptosis. Furthermore, although thymic organs form normally in bls, they are devoid of lymphocytes at 4 dpf. However, primitive macrophages develop normally in bls, as does the embryonic vasculature. Surprisingly, when the bloodless embryos are raised beyond 5 dpf, blood cells begin to accumulate and lymphocytes can be found in the thymi. The mutants can grow to adulthood, with no detectable abnormality in the definitive blood lineages. As such, bls is the earliest acting mutation specific to hematopoiesis that is reported in zebrafish. Other zebrafish mutations with early defects in hematopoiesis either affect other organ systems, such as vasculogenesis in clo, or have more general mesoderm defects, such as defect in paraxial mesoderm migration in spt (Stainier et al., 1995; Thompson et al., 1998). Additionally, bls is the only vertebrate mutation that exhibits severe anemia during embryogenesis but not adulthood.

Hematopoietic genes such as scl and gata1 are expressed in the zebrafish lateral plate mesoderm as early as five somites, evidence that hematopoietic precursors have been specified by that time point (Detrich et al., 1995; Gering et al., 1998; Liao et al., 1998; Thompson et al., 1998). The analysis of scl and gata1 expression in bls mutants show that the number of hematopoietic cells derived from the lateral plate mesoderm is greatly reduced. Meanwhile, expression of flk1 is unaffected by bls, and the vasculature is intact.

As embryogenesis progresses in bls, the number of scl- and gata1-expressing cells decreases. By 23 hpf, no scl or gata1 transcripts can be detected in the ICM and only a handful of cells expressing scl can be found in the posterior ICM. Concurrent with the decreasing number of scl- and gata1-expressing cells, apoptosis is noted in the ICM. Furthermore, the persistence of gata2 and scl and the absence of gata1 and ikaros in the anterior ICM of the mutant animal suggests that these cells are able to express early hematopoietic genes, but fail to differentiate and express genes associated with progressive hematopoietic lineage differentiation. These observations suggest that the hematopoietic progenitors that do form in bls fail to undergo normal differentiation and undergo apoptosis (Weiss and Orkin, 1995). Of note, scl, gata1 and flk1 transcripts are absent in the lateral plate mesoderm of clo mutants (Liao et al., 1997, Liao et al., 1998). Unlike clo, some scl-positive hematopoietic progenitors are specified in the lateral plate mesoderm. Therefore, bls appears to regulate, but is not absolutely required, for the specification of scl-expressing blood progenitors. Subsequently, bls is involved in maintaining scl and gata1 expression during differentiation of the embryonic blood cells.

Analysis of macrophage markers pu.1, cmyb and draculin demonstrate that primitive macrophages are produced at wild-type levels in the bilateral rostral paraxial mesoderm. This suggests that distinct hematopoietic programs regulate the development of primitive macrophages from the anterior paraxial mesoderm, and the elaboration of primitive hematopoietic progenitors from the ICM. These different regulatory signals are anatomically segregated, where the ICM may receive signals from the trunk paraxial mesoderm, whereas the rostral macrophage anlage is influenced by anterior cues.

The delay in lymphopoiesis in bls suggests that either the gene is required in the development of lymphoblasts, or that the ICM hematopoietic progenitors contribute to the initiation of lymphopoiesis. Moreover, studies in mice underscore the symbiotic dependence between thymic epithelium and lymphoblasts in their mutual development (Manley, 2000; Nehls et al., 1996; Ritter and Boyd, 1993). It is not clear whether this interdependence is also present in the development of lymphoblasts and thymic epithelium in zebrafish. Further work in lineage tracing and electron-microscopic analysis of the thymus in bls mutants will better characterize the lymphoid defect.

Unlike gata1 and ikaros, the expression of scl and gata2 is not restricted to the anterior ICM and is also found in the posterior ICM at 23 hpf. Fate-mapping experiments have shown that cells of the posterior ICM arise as a result of complex migration from the extending tailbud (Kanki and Ho, 1997). Among vertebrates, scl and gata2 are expressed in both hematopoietic and endothelial cells, whereas gata1 and ikaros are expressed in only hematopoietic cells (Georgopoulos et al., 1992; Orkin, 1995). Additionally, the mutant clo lacks scl expression in both the anterior and posterior ICM, and fails to produce hematopoietic or endothelial cells. Last, endothelial markers such as hhex, flk1 and fli1 are highly expressed in the posterior ICM (Liao et al., 1997; Liao et al., 2000b; Thompson et al., 1998). Taken together, these observations suggest that the posterior ICM represents endothelial tissue and not hematopoietic progenitors. Therefore, expression of scl in the posterior ICM of bls is consistent with its phenotype of compromised hematopoiesis but intact vasculogenesis.

Cell transplantation studies showed that bls donor cells were able to contribute to gata1-expressing cells in the ICM of wild-type hosts. Conversely, bls hosts did not support the differentiation of wild-type donor cells to gata1-expressing cells. Given the non-cell autonomous action of the bls gene, one might speculate that the gene product could be a secreted factor or a cell surface receptor required for proper development of primitive hematopoietic cells. The dominant but incompletely penetrant aspect of the bls mutation supports a secreted factor hypothesis, where the proposed cytokine may be limiting during embryogenesis, manifesting in a haplo-insufficient phenotype.

The clo mutant has combined defect in hematopoietic and endothelial progenitors, and the clo gene is thought to function at the level of the hemangioblast, a proposed transient bi-potential cell that gives rise to endothelial and hematopoietic lineages (Liao et al., 1997; Stainier et al., 1995). Overexpression of scl and hhex can lead to partial rescue of hematopoietic progenitors in clo mutants (Liao et al., 1998; Liao et al., 2000b). Interestingly, the clo gene appears to act non-cell autonomously early in the differentiation of embryonic blood cells, before the expression of gata1 (Parker and Stainier, 1999). Therefore, the relationships between scl, hhex, clo and bls in early hematopoiesis deserve further investigation.

Overexpression studies in Xenopus have shown that bmp4 expands ventral mesoderm and induces the expression of scl (Mead et al., 1998). Overexpression of tolloid leads to ectopic maintenance of endogenous bmp4, and leads to ventralized phenotype with expanded gata1-expressing cells (Blader et al., 1997). Although the number of cells expressing scl is increased in the posterior ICM of bmp4 ventralized tails of bls mutants, no scl-expressing cells were observed in the anterior ICM. Only rare gata1-expressing cells were seen in the anterior ICM of bmp4-injected embryos (1.8% of injected embryos, Table 1). Ventralization of the ICM by bmp4 in clo also did not rescue scl- or gata1-expressing cells. Failure of bmp4 to rescue hematopoietic cells in the anterior ICM suggests that the bls and clo gene products are required for hematopoiesis at a stage downstream of ventral mesoderm induction.

By contrast, gata1-expressing cells can be detected in the anterior ICM of bls and clo mutants that have been injected with scl. Either the gata1-positive cells are rescued hematopoietic progenitors or they represent blood cells specified from the lateral plate mesoderm. The percentage of scl injected embryos with gata1-positive blood cells is comparable between bls and clo mutants (48% and 41%, respectively) (Table 1). However, there are much fewer gata1-expressing cells in the anterior ICM of scl-injected bls animals than similarly injected clo mutants (Fig. 6C). This observation implies that bls may play additional roles in maintaining scl expression during embryogenesis, where the absence of the bls gene product attenuates the partial hematopoietic rescue when compared with that observed in clo. This view is consistent with expression analysis of scl and gata1, which show that bls mutants are able to produce some scl-positive blood progenitors in the lateral plate mesoderm, but such cells fail to express gata1 and undergo apoptosis. Furthermore, overexpression of gata1 is not sufficient for rescue of primitive in blood in neither bls nor clo. This suggests that other downstream target genes of bls and scl are necessary for primitive hematopoiesis, and that gata1 represents one such target gene. Taken together with the absence of scl induction by bmp4 in the bls mutant, we propose that bls acts downstream of bmp4 in initiating or maintaining expression of scl.

The unique role of bls as a non-cell autonomous regulator of primitive hematopoiesis exemplifies the utility of zebrafish as a genetic model system for the study of blood development. Possible function of bls as a stromal signal in activating and maintaining scl expression underscores the importance of extrinsic cues during early hematopoietic differentiation. Moreover, the requirement for bls during primitive hematopoiesis but not definitive blood development may provide insight into regulatory cues that differ between embryonic development and adult homeostasis.

Our sincere thanks go to Alan Davidson, Barry Paw and David Trevor for review of this manuscript. Our gratitude goes to Walt Saganic and Neal White for excellent fish care and husbandry. We thank Jarema Malicki and Kristin Artinger for advice on cell transplantation experiments, Chirs Amemiya for the ikaros cDNA, Masataka Nikaido for the bmp4 cDNA, Sue Lyons for the gata1 expression construct, and Patrick Blader for the tolloid expression construct. E. C. L. has received support from the NIH Medical Scientist Training Program, American Cancer Society Stone Fellowship, and Yamaguchi Award from Children’s Hospital Division of Hematology/Oncology. E. C. L. is a Pre-doctoral Fellow and L. I. Z. an Investigator of the Howard Hughes Medical Institute. This work is also supported by grants from the NIH. This work was funded by the Howard Hughes Medical Institute, NIH grant P50 DK49216.