The SCL gene encodes a basic helix–loop–helix (bHLH) transcription factor that is essential for the development of all haematopoietic lineages. SCL is also expressed in endothelial cells, but its function is not essential for specification of endothelial progenitors and the role of SCL in endothelial development is obscure. We isolated the zebrafish SCL homologue and show that it was co‐expressed in early mesoderm with markers of haematopoietic, endothelial and pronephric progenitors. Ectopic expression of SCL mRNA in zebrafish embryos resulted in overproduction of common haematopoietic and endothelial precursors, perturbation of vasculogenesis and concomitant loss of pronephric duct and somitic tissue. Notochord and neural tube formation were unaffected. These results provide the first evidence that SCL specifies formation of haemangioblasts, the proposed common precursor of blood and endothelial lineages. Our data also underline the striking similarities between the role of SCL in haematopoiesis/vasculogenesis and the function of other bHLH proteins in muscle and neural development.
During mammalian embryogenesis, haematopoiesis occurs sequentially at several different sites. Early data suggested that haematopoietic stem cells first arose in the yolk sac and subsequently colonized the fetal liver and bone marrow. However, the yolk sac progenitors appeared to have limited potential (Moore and Metcalf, 1970; Wong et al., 1986; Liu and Auerbach, 1991; Cumano et al., 1993; Huang and Auerbach, 1993). More recently, an intra‐embryonic source of multipotent haematopoietic stem cells has been identified in the posterior splanchnopleura (Godin et al., 1993, 1995; Cumano et al., 1996) and the region of the dorsal aorta, genital ridge and mesonephros (AGM) in mice (Medvinsky et al., 1993; Medvinsky and Dzierzak, 1996). The relationship between haematopoietic stem cells in the yolk sac and those arising within the embryo remains unclear, but it has been suggested that the intra‐embryonic haematopoietic stem cells colonize both yolk sac and fetal liver (Dzierzak and Medvinsky, 1995). Haematopoietic development is also highly conserved in lower vertebrates. Teleost fish produce differentiated cells analogous to most of the mature blood lineages found in mammals (Rowley et al., 1988). Embryonic haematopoiesis develops within the body of the embryo in many bony fish. Early haematopoietic stem cells first arise within the lateral mesoderm and subsequently migrate medially to form the intermediate cell mass (ICM), where they are closely associated with the dorsal aorta and axial vein (Al‐Adhami and Kunz, 1977; Detrich et al., 1995; Zon, 1995).
Several lines of evidence suggest a close relationship between the development of blood and endothelium in vertebrates. Both cell types emerge simultaneously during the formation of yolk sac blood islands (Sabin, 1920; Murray, 1932; Wagner, 1980), and early intra‐embryonic sites of haematopoiesis are closely associated with the dorsal aorta and other vessels (Godin et al., 1993, 1995; Medvinsky et al., 1993; Cumano et al., 1996; Medvinsky and Dzierzak, 1996; Tavian et al., 1996; Wood et al., 1997). Mice lacking Flk‐1, an early endothelial marker, fail to generate both vasculature and haematopoietic cells (Shalaby et al., 1995). Flk‐1 null embryonic stem (ES) cells also fail to contribute to either endothelium or haematopoiesis in chimaeric mice, showing that Flk‐1 is essential for the formation of both lineages (Shalaby et al., 1997). Consistent with this concept are observations demonstrating that VEGF, the ligand for Flk‐1, can stimulate formation of multipotent haematopoietic stem cells from ES cells (Kennedy et al., 1997). Furthermore, single Flk‐1‐positive (Flk‐1+) cells from avian embryos can develop into either haematopoietic or endothelial colonies (Eichmann et al., 1997). Moreover, in the zebrafish mutant cloche, the numbers of both endothelial and haematopoietic cells are severely reduced (Liao et al., 1997). These studies are all consistent with the notion of a common precursor for haematopoietic and endothelial cells, the haemangioblast.
The SCL gene (also known as TAL‐1) was first identified as a T‐cell oncogene (Begley et al., 1989a,b; Finger et al., 1989; Chen et al., 1990). It encodes a basic helix–loop–helix (bHLH) protein normally expressed in blood, endothelium and brain (Green et al., 1992; Hwang et al., 1993; Kallianpur et al., 1994; Drake et al., 1997). Antisense and overexpression studies have suggested that SCL modulates proliferation and self‐renewal of multipotent haematopoietic cells (Green et al., 1991) and also acts as a positive regulator of erythroid differentiation (Aplan et al., 1992). SCL null mice lacked yolk sac haematopoiesis (Robb et al., 1995; Shivdasani et al., 1995) but initially were reported to develop morphologically normal endothelium and defects of brain development were not seen (Shivdasani et al., 1995). SCL null ES cells also failed to contribute to any definitive haematopoietic lineage (Porcher et al., 1996; Robb et al., 1996). These data established an essential and cell‐autonomous role for SCL in the formation or subsequent behaviour of embryonic haematopoietic stem cells. However, the role of SCL in endothelial and neuronal development remained obscure.
In this study, we have isolated the zebrafish homologue of the SCL gene product and demonstrate that it is co‐expressed initially with markers for haematopoietic, endothelial and pronephric duct progenitors in the lateral mesoderm. Ectopic expression of SCL resulted in perturbation of vasculogenesis with excessive production of blood and endothelial precursors at the expense of somitic and pronephric duct cells. These data provide the first evidence that SCL acts to specify haemangioblast development from early mesoderm.
Cloning of the SCL cDNA from zebrafish
To clone the Danio rerio (zebrafish) SCL gene, an embryonic cDNA library was screened for cDNAs encoding SCL‐like bHLH transcription factors. The fragment used to screen the library had been amplified under low stringency conditions on genomic zebrafish DNA using two primers defined according to the 5′ and 3′ ends of the bHLH region of the SCL‐related Fugu rupripes gene SLP1 (Göttgens et al., 1998). Four different cDNAs were isolated and sequenced. All of them encoded bHLH proteins but only one exhibited extensive similarities to SCL proteins in other vertebrates. Its sequence of 324 amino acids shared 57–59% identical residues with the human, murine and chicken SCL proteins (compared with 36 and 40% identity with the closely related murine and human LYL‐1 proteins, respectively) (Figure 1a and b). Its bHLH domain (residues 185–244) was identical to that of the other vertebrate SCL proteins (Figure 1a). Outside the bHLH domain, significant conservation was also observed in both N‐ and C‐terminal regions, suggesting the existence of additional functionally important domains. Deletion of 53 amino acids including the conserved C‐terminal domain was associated with enhanced apoptosis of transformed human T cells (Leroy‐Viard et al., 1995). The N‐terminal region has been reported to contain a transactivation domain, the activity of which was modulated by phosphorylation of two serine residues (Figure 1a, asterisks; Cheng et al., 1993; Prasad and Brandt, 1997). Of these, Ser122 was conserved in zebrafish, whereas the second one, Ser172 (Prasad and Brandt, 1997), was replaced by an alanine residue.
Expression profile of SCL in early zebrafish development
Whole‐mount in situ hybridization analysis of zebrafish embryos was used to characterize SCL expression (Figure 2). SCL mRNA was first detected in 1–3 somite embryos (11 hpf) in two pairs of stripes (Figure 2a–d).
One pair of stripes was situated in the body, initially in the lateral margins of the mesoderm. These trunk stripes extended anteriorly and posteriorly (compare Figure 2a with e, c with g, and d with h). While the anterior third of these trunk stripes stayed lateral (Figure 2i–n, white arrowheads), the posterior two‐thirds moved towards the midline (compare Figure 2d, h and j). At this stage in development, cells of the lateral mesoderm are known to migrate to the midline to form the ICM just ventral to the notochord (Al‐Adhami and Kunz, 1977). We therefore believe that the moving SCL expression pattern reflects migration of cells rather than a wave of SCL gene expression traversing the mesoderm. The migrating stripes fused anteriorly and posteriorly at both ends at the 14 somite stage (15.5 hpf), resulting in a γ‐shaped pattern [shown in a 15 somite stage embryo (16.5 hpf) in Figure 2j]. The cells of the loop of this γ‐structure subsequently ‘zipped up’ in the craniocaudal direction to form a single stripe in the position of the ICM (Figure 2j, l and m, black arrowheads). This process was complete at the 20 somite stage (19 hpf) (data not shown). The ICM (between black arrowheads in Figure 2m) is known to give rise to haematopoietic cells which enter the circulation between 24 and 26 hpf (Zon, 1995). SCL expression was observed in circulating blood cells at 30 hpf (Figure 2o, arrowhead indicates SCL‐positive blood cells within the heart), but subsequently declined to undetectable levels by 50 hpf (Figure 2p). While these data suggest that SCL‐expressing cells within the ICM became blood cells, it was not clear from these initial results what the function of the SCL‐positive (SCL+) cells in the arms of the γ‐structure might be.
The second pair of stripes of SCL‐expressing cells was located in the head in the position of the head mesenchyme (Figure 2a, b, e and f). The most anterior of these cells ceased to express SCL between the 10 and the 15 somite stage (13.5 and 16.5 hpf) (compare Figure 2e and i), while the cells posterior to the eye anlagen continued to express SCL until ∼23 hpf (Figure 2n). At first, the nature of these cells was unclear. At 21 hpf, SCL was also noted transiently in the position of primary motoneurons of the spinal cord (data not shown) and at 50 hpf in the location of proliferative neuronal cells in the diencephalon, in the midbrain and in a few cells of the hindbrain (Figure 2p, arrowhead).
During mouse development, SCL has also been shown to be expressed in haematopoietic progenitors (yolk sac, AGM, fetal liver) as well as in the midbrain, hindbrain and the spinal cord (Green et al., 1992; Kallianpur et al., 1994; Silver and Palis, 1997). The pattern of SCL expression during development is therefore highly conserved from teleost fish to mammals.
SCL is expressed in multipotent progenitors in early lateral mesoderm
In zebrafish, the lateral mesoderm of the trunk and tail gives rise to at least three different tissues: the haematopoietic, endothelial and pronephric lineages. To investigate the relationship of the progenitors for these three lineages within the lateral mesoderm, the expression patterns of haematopoietic [GATA‐2 (Detrich et al., 1995; A.R.F.Rodaway, unpublished data)], endothelial [Flk‐1 (Fouquet et al., 1997; Liao et al., 1997; Sumoy et al., 1997)] and pronephric duct markers [Pax‐2 (Krauss et al., 1991)] were studied in pairwise comparisons with that of SCL. In double whole‐mount in situ hybridization experiments, embryos of different stages were first labelled for SCL‐expressing cells in red and then stained for GATA‐2, Flk‐1 or Pax‐2 expression in blue. Cells co‐expressing both SCL and the second marker therefore appeared purple. In addition, sequential staining of thin (10 μm) sections was performed to confirm co‐expression of genes in single cells. For this purpose, double in situ hybridizations were performed on embryos. The embryos were stained for SCL in red, sectioned, and the sections documented, before proceeding to stain the same section for expression of the second gene.
At the six somite stage (12.5 hpf), SCL and the haematopoietic marker GATA‐2 were co‐expressed in the lateral mesoderm (cells between black arrowheads in Figure 3a). This was expected as both genes are known to be expressed in haematopoietic progenitors. However, SCL+ cells could also be stained for expression of the endothelial marker Flk‐1 (Figure 3d, cells between black arrowheads) and the pronephric duct marker Pax‐2 (Figure 3j). Sequential staining of embryos indicated that at the six somite stage, cells expressing SCL and Pax‐2 (compare Figure 3i and j, white arrowheads) were medial to cells expressing Pax‐2 only (Figure 3j, black arrowhead). This is seen more easily in 10 μm sections, which also confirm at the single cell level that SCL was expressed in a subpopulation of the Pax‐2‐positive (Pax‐2+) cells (compare Figure 3k and l). The co‐expression of SCL, GATA‐2, Flk‐1 and Pax‐2 within the same population of cells suggests that at this stage SCL+ cells have not yet committed themselves to the blood lineage and might still be multipotent.
At the 10 somite stage (13.5 hpf), SCL and Pax‐2 expression patterns had already begun to separate (data not shown) and were clearly distinct at the 15 somite stage (Figure 3m). By contrast, SCL, GATA‐2 and Flk‐1 were still co‐expressed in 10 somite stage embryos. Sequential staining of 10 μm sections for SCL and Flk‐1 expression confirmed co‐expression of both genes in single cells (compare Figure 3e and f). During the period of extension of the tail and convergence of the lateral mesoderm, SCL and GATA‐2 were still co‐expressed (Figure 3b and c) whereas distinct expression domains of SCL and Flk‐1 became apparent (Figure 3g). At the 18 somite stage (18 hpf), Flk‐1+ cells that no longer expressed SCL had reached the midline below the notochord where the dorsal aorta forms (the blue stripe indicated by the black arrow in Figure 3g). A stripe of cells expressing Flk‐1 and SCL remained in a more lateral position (purple cells indicated by the black arrowhead in Figure 3g). They were probably the progenitors of the axial vein that later ceased to express SCL (data not shown). Between the two stripes of Flk‐1‐expressing endothelial progenitors, a population of cells expressing SCL but not Flk‐1 was discernible on close inspection (the red stripe indicated by a white arrow in Figure 3g). These cells later came to lie in the centre of the ICM between the endothelial precursors, differentiated into red blood cells and entered circulation by 24–26 hpf.
These data suggest that the SCL+/Flk‐1+ cells seen in the 10 somite stage embryos are able to give rise to SCL+/Flk‐1− haematopoietic and SCL−/Flk‐1+ endothelial cells and, thus, represent common progenitors for both blood and endothelium, the proposed haemangioblast.
The double in situ hybridization experiments described above also shed light on the nature of the SCL+ cells in the head mesenchyme and in the arms of the γ‐structure. These cells were likely to be functionally different from those SCL+ cells that reached the ICM, in that previous studies have shown that, unlike in the ICM, there is no GATA‐1 expression and no blood formation in these more anterior areas (Detrich et al., 1995; Figure 5g). Our results showed that most of these more anterior cells also expressed GATA‐2 (white arrowheads in Figure 3a and b) and Flk‐1 (white arrowheads in Figure 3d, g and h) and therefore were likely to be endothelial progenitors involved in the formation of the head vessels and the vessels of the Ducti Cuvieri (Fouquet et al., 1997; Liao et al., 1997; Sumoy et al., 1997) that link the venous system to the heart. In the region of the Ducti Cuvieri, SCL+ cells (white arrowhead in Figure 3n) were adjacent to Pax‐2+ pronephric progenitors (black arrowhead in Figure 3n). This observation suggests that these SCL+ cells may participate in the vascularization of the pronephros or, alternatively, in the initiation of definitive haematopoiesis in the pronephros (Al‐Adhami and Kunz, 1977).
Overexpression of SCL in zebrafish embryos causes overproduction of blood
Microinjection of mRNA into Xenopus and, more recently, into zebrafish embryos has been used to produce ectopic expression of transcription factors that normally exhibit a restricted expression pattern (Hopwood and Gurdon, 1990; Hopwood et al., 1991; Ferreiro et al., 1994; Turner and Weintraub, 1994; Lee et al., 1995; Blader et al., 1997). To investigate the function of SCL in the development of early lateral mesoderm, SCL mRNA was injected into 2‐ or 4‐cell zebrafish embryos. Varying amounts of SCL mRNA were injected in pilot studies to determine the RNA concentration that produced a consistent phenotype with minimal non‐specific toxicity. By 30 hpf, the majority of embryos injected with 70 pg of SCL mRNA exhibited a dramatic abnormal phenotype (Table I). As shown in Figure 4, this was characterized by a marked excess of haemoglobinized blood cells in the ICM (Figure 4b, black arrowhead), a striking absence of blood circulation despite a beating heart (Figure 4b) and an abnormally curved axis. The presence of red blood cells and normal pigmentation (not visible in Figure 4 because embryos were treated with 1‐phenyl‐2‐thiourea to suppress pigmentation after 24 hpf) showed that the injected embryos were not retarded. The embryos were not ventralized since head and notochord structures which are the first to be affected in ventralized embryos (Neave et al., 1997) were morphologically normal. By 50 hpf, half of the affected embryos were able to establish restricted or normal blood circulation while the other half of the embryos never gained circulation, developed severe oedema and died after 6–7 days.
Sections through the severely affected 50 hpf embryos revealed high numbers of haemoglobinized cells in the ICM. Strikingly, the trunk vessels, dorsal aorta and axial vein, as well as one or both pronephric ducts, frequently were missing. In addition, a massive reduction of somitic tissue was observed, often more marked on one side of the embryo (Figure 4, compare c with d). By contrast, notochord and neural tube always appeared normal. Control injections were performed using similar amounts of mRNA for lacZ, green fluorescent protein (GFP), zebrafish E12 (Wülbeck et al., 1994) and a fusion protein consisting of the bHLH region of a neuronal bHLH protein fused to VP16. None of these produced a similar phenotype (data not shown). The specificity of the phenotype described here is also evidenced by the fact that it was not reported following injection of mRNA for multiple other bHLH proteins into frog or zebrafish embryos in other laboratories (Hopwood and Gurdon, 1990; Hopwood et al., 1991; Ferreiro et al., 1994; Turner and Weintraub, 1994; Lee et al., 1995; Blader et al., 1997).
The increased number of haemoglobinized cells in SCL‐injected embryos suggested an excess of blood in the region of the ICM that could reflect either excessive blood formation or abnormal distribution of blood cells. To distinguish between these possibilities, in situ hybridization on embryos prior to circulation was performed using markers for blood progenitors (SCL and GATA‐1). Endogenous SCL expression was assessed using a probe from a region of the 3′ untranslated region (UTR) absent from the injected SCL mRNA, and which therefore discriminates between endogenous and exogenous SCL mRNA. In SCL‐injected embryos, the number of cells expressing endogenous SCL in the lateral mesoderm was increased at the 10 somite stage (13.5 hpf) and was often asymmetrical (compare Figure 5a with d; Table I). The increased expression of endogenous SCL was also visible in whole mounts and sections just before the onset of circulation (compare Figure 5b with e, and c with f; Table I). SCL‐injected embryos also contained increased numbers of cells positive for GATA‐1 (compare Figure 5g with h; Table I). These observations therefore demonstrate that ectopic SCL expression resulted in overproduction of haematopoietic progenitors capable of differentiating into haemoglobinized cells.
SCL overexpression increased the number of haemangioblast‐like cells
The failure of circulation and lack of vessel formation in SCL‐injected embryos suggested that early endothelial development might be abnormal. To investigate this issue, SCL‐injected and control embryos were hybridized to probes for Flk‐1 (Fouquet et al., 1997; Liao et al., 1997; Sumoy et al., 1997) and Fli‐1 (L.Brown, T.F.Schilling, A.R.F.Rodaway, D.C.Hickleton, T.Jowett, P.W.Ingham, R.K.Patient and A.D.Scharrocks, in preparation), both early markers of endothelial development in zebrafish. In SCL‐injected embryos, there was a striking increase in the number of Fli‐1+ and Flk‐1+ cells (compare Figure 6a with b, and c with d: Table I) which were distributed aberrantly (compare Figure 6e with f and g).
The increased number of haematopoietic and endothelial progenitors in SCL‐injected 22 hpf embryos suggested that ectopic expression of SCL might be expanding a population of common precursors for both lineages. This would predict increased numbers of SCL+/Flk‐1+ cells in 10 somite stage embryos (13.5 hpf). To test this prediction, 10 μm sections of an SCL‐injected 10 somite stage embryo (shown in Figure 5d after staining of the whole mount for SCL) were stained sequentially for SCL and Flk‐1. These sections revealed a massive unilateral increase in the number of SCL‐expressing cells in the non‐axial mesoderm (Figure 6i). The vast majority of these SCL+ cells also expressed Flk‐1 (Figure 6j). This suggests that ectopic SCL expression increased the number of early SCL+/Flk‐1+ haemangioblast‐like progenitors.
To follow these cells later during development of the SCL‐injected embryo, we sequentially stained 22 hpf embryos for Flk‐1 (in red) and either SCL or GATA‐1 (in blue). Flk‐1 and SCL expression patterns were still overlapping at this stage (compare Figure 6k and l), but cells immediately above the yolk cell extension proved to express only SCL but not Flk‐1 (black arrowheads in Figure 6k and l). Cells in the same position were also found to express GATA‐1 (Figure 6n, black arrowhead), suggesting that these cells are the progenitors of the haemoglobinized cells found in injected embryos after 50 hpf (Figure 4b). The whole mounts indicated that Flk‐1 and GATA‐1 expression patterns were not overlapping (compare Figure 6m and n), and this was confirmed by sections through the trunk (Figure 6h).
Our data therefore suggest that the cells which initially co‐expressed Flk‐1 and SCL at the 10 somite stage developed into two domains in 22 hpf embryos. A ventral haematopoietic region (SCL+/GATA‐1+) contained cells which were capable of terminal differentiation as witnessed by excess production of haemoglobinized cells in SCL‐injected embryos. The more dorsal domain continued to express SCL and Flk‐1, suggesting the continued presence of haemangioblast‐like cells. The impairment of vessel formation in SCL‐injected embryos might reflect either a direct effect of SCL overexpression on the differentiation programme of individual endothelial precursors, or may be an indirect consequence of the mesodermal disorganization.
SCL directs non‐axial mesoderm into haematopoietic and endothelial lineages
Excessive production of haemangioblast‐like progenitors with concomitant loss of somitic and pronephric duct cells strongly suggested that ectopic SCL expression was altering the fate of non‐axial mesodermal cells. The abnormalities induced by injection of SCL mRNA were often more marked on one or the other side of SCL‐injected embryos, an observation likely to reflect asymmetrical distribution of the progeny of injected cells. Consistent with this explanation, co‐injection of lacZ and SCL mRNA demonstrated frequent unilateral β‐galactosidase activity in embryos with ipsilateral morphological abnormalities (Figure 7b). To investigate the timing of the events responsible for the observed abnormalities, embryos were hybridized with probes for somitic mesoderm (MyoD), paraxial mesoderm (Paraxis) or pronephric duct tissue (Pax‐2).
MyoD is an early marker for somitic mesoderm (Weinberg et al., 1996). In 15 somite stage embryos (16.5 hpf), MyoD expression was normal on the side of the embryo that displayed little β‐galactosidase activity (Figure 7b, white arrowhead), but was markedly reduced on the side that had received most of the injected mRNA (Figure 7b, black arrowhead; Table I). Moreover, Flk‐1 expression was up‐regulated in the opposite manner—increased expression was confined to the side of the embryo that exhibited β‐galactosidase activity [Figure 7c, purple stain, arrowheads indicate injected (black) and uninjected (white) sides]. These results suggest that SCL was altering the fate of cells normally destined to become somitic muscle.
In view of these results, we studied an even earlier marker, one for the paraxial mesoderm. Embryos were hybridized with a probe for Paraxis which is normally expressed both in early paraxial mesoderm and in newly formed somites (Burgess et al., 1995; S.Shanmugalingam and S.W.Wilson, in preparation). In six somite stage (12 hpf) SCL‐injected embryos, Paraxis expression in the paraxial mesoderm was markedly reduced, asymmetrically in the example shown, and somite boundaries were hardly visible (compare Figure 7d with e, arrowhead; Table I). These data suggest that ectopic SCL expression converted the fate of paraxial mesodermal cells at a very early stage.
Sections of severely affected SCL‐injected embryos showed that there was frequent loss of one or both pronephric ducts (Figure 4d). Pax‐2 was therefore used as an early marker for the pronephric lineage (Krauss et al., 1991). At the 10 somite stage (13.5 hpf), Pax‐2 expression appeared normal (Table I), suggesting that ectopic SCL expression was converting early mesoderm into blood and endothelial progenitors but not into pronephric precursors. However, Pax‐2 expression was clearly reduced in the region of one or both pronephric ducts by 22 hpf (26 somite stage) (compare Figure 7f and g arrowheads; Table I) and 30 hpf (Table I). These data may reflect diversion of putative multipotent progenitors within the lateral mesoderm away from the pronephric duct lineage, or alternatively may represent an effect of enforced SCL expression on subsequent pronephric duct differentiation.
These results are in accord with the concept that ectopic SCL expression converted non‐axial mesodermal cells into haematopoietic and endothelial lineages at the expense of somitic and pronephric duct fates. This suggests that the role of SCL in haematopoiesis and vasculogenesis is analogous to that of MyoD and NeuroD, which are able to convert fibroblasts into muscle cells in vitro (Davis et al., 1987) and non‐neural ectodermal cells into neuronal cells in Xenopus embryos (Lee et al., 1995), respectively. However, it was possible that instead of converting cell fate, SCL expression in non‐axial mesoderm resulted in apoptosis of somitic and pronephric duct progenitors followed by a secondary proliferation of normal haematopoietic and endothelial precursors. Two lines of evidence argue against this interpretation. First, the TUNEL assay was used to look for evidence of apoptosis in embryos at time points that preceded or coincided with the earliest molecular changes that we had observed. Only slightly increased apoptosis was observed in SCL‐injected embryos, but it was never particularly localized in the non‐axial mesoderm (Figure 7h–k). Secondly, we frequently observed aberrant expression of haematopoietic or endothelial markers in locations spatially separate from the main domain of haematopoietic/endothelial cells. This is difficult to reconcile with a model involving secondary proliferation of blood and endothelial progenitors, but is fully consistent with alterations in mesodermal cell fate. Interestingly, while in SCL‐injected embryos we consistently observed SCL+ (Figure 5f, arrowhead), Fli‐1+ (Figure 6b, arrowhead) or Flk‐1+ (Figure 6g, arrowhead) cells that were well separated from the main domain of SCL, Fli‐1 or Flk‐1 expression, we only rarely observed ectopic cells positive for the later haematopoietic marker GATA‐1 (Figure 5h), suggesting that environmental cues not present at the most ectopic locations are likely to be necessary for full haematopoietic differentiation. Taken together, these observations argue strongly for a dominant effect of SCL expression on mesodermal cell fate.
Previous studies have shown that SCL is a critical regulator of haematopoiesis, and is essential for the development of all haematopoietic lineages (Robb et al., 1995, 1996; Shivdasani et al., 1995; Porcher et al., 1996). However, these loss‐of‐function studies were unable to show whether SCL functioned to specify the formation of haematopoietic stem cells from mesoderm or whether SCL was required for the survival and subsequent differentiation of haematopoietic stem cells. The gain‐of‐function experiments reported here demonstrate for the first time that SCL can programme early mesodermal cells to form haematopoietic and endothelial progenitors at the expense of other non‐axial mesodermal cell fates (Figure 8).
We show that SCL is expressed during zebrafish development in haematopoietic progenitors, as evidenced by the co‐expression of SCL and GATA‐2, and was also expressed in circulating blood cells. The haematopoietic expression of SCL is therefore conserved in mouse (Green et al., 1992; Kallianpur et al., 1994), Xenopus (Turpen et al., 1997; E.M.Read and R.K.Patient, unpublished data) and zebrafish. We also show co‐expression of SCL and the early endothelial marker Flk‐1 in the posterolateral mesoderm and the head mesenchyme in early zebrafish embryos. We provide evidence that in SCL‐injected embryos the population of SCL+/Flk‐1+ cells is massively increased in the posterior mesoderm. Studies in uninjected and SCL‐injected embryos suggest that these SCL+/Flk‐1+ cells are the progenitors for both blood and endothelium. In uninjected embryos, a separation of SCL and Flk‐1 expression domains coincides with segregation of cells into blood and endothelial precursors. This segregation occurs during their migration towards the midline where they form the ICM just below the notochord. Whereas SCL+ blood cells enter circulation, Flk‐1+ endothelial cells line the major vessels of the trunk. In SCL‐injected embryos, a separation of the apparently uniform population of early SCL+/Flk‐1+ cells into separate haematopoietic (SCL+/GATA‐1+/Flk‐1−) and endothelial/haemangioblast (SCL+/Flk‐1+) domains also supports the view that the early cells are bipotent. Thus, our results strongly suggest that SCL is capable of specifying haemangioblast formation, a concept which accords well with several lines of circumstantial evidence for the existence of haemangioblasts during development (Choi et al., 1998; Nishikawa et al., 1998 and references therein).
Our data also demonstrate striking similarities between the role of SCL in haematopoiesis and vasculogenesis, and the functions of other bHLH transcription factors such as MyoD and NeuroD in the development of muscle and neuronal tissues, respectively. Ectopic expression of these proteins was shown to cause a change of fate in certain susceptible tissues. While MyoD converts fibroblasts into myoblasts (Davis et al., 1987), NeuroD was reported to induce the differentiation of neurons from Xenopus non‐neural ectoderm (Lee et al., 1995). Ectopic SCL expression in zebrafish embryos resulted in expansion of haematopoietic and endothelial precursors at the expense of somitic and pronephric duct tissues. Since only a slight increase in apoptosis was observed and cell death was not particularly concentrated in the non‐axial posterior mesoderm, we believe that the reduction in somitic and pronephric duct tissues was not a consequence of cell death and replacement by proliferation but rather was due to the diversion of cells into the haematopoietic and endothelial lineages. This interpretation is supported by the expression of endogenous SCL, Flk‐1 and Fli‐1 in ectopic positions clearly separate from the main domains of haematopoietic and endothelial cells in SCL‐injected embryos.
Loss of SCL function in mice abolished the development of all haematopoietic lineages but did not prevent the formation of morphologically normal endothelial cells (Robb et al., 1995; Shivdasani et al., 1995). Defects in the formation of the nervous system, where SCL is expressed as well, were also not reported. These data suggested an essential role for SCL in haematopoietic stem cell formation or function, but could not ascribe a function to SCL in vasculogenesis or neurogenesis. The differences between the loss‐of‐function phenotype and the gain‐of‐function experiments reported here further emphasize the parallels between our data and the role of bHLH proteins in the control of cell fate in other tissues. In myogenesis and neurogenesis, it is clear that determination and differentiation are controlled by a cascade of bHLH proteins with overlapping expression patterns (Jan and Jan, 1993; Weintraub, 1993). Thus, ectopic expression of the myogenic bHLH proteins, MyoD, Myf5, MRF4 and myogenin, all induced muscle differentiation in transfected cells (Weintraub, 1993). MyoD and Myf‐5 individually were dispensible for skeletal muscle development (Braun et al., 1992; Rudnicki et al., 1992), but disruption of both genes resulted in the complete absence of skeletal muscle (Rudnicki et al., 1993). Similarly, ectopic expression of any of the four bHLH genes in the Drosophila achaete–scute complex produced an identical phenotype, increased numbers of microchaetae and macrochaetae (Rodriguez et al., 1990; Brand et al., 1993; Dominguez and Campuzano, 1993), whereas loss‐of‐function studies demonstrated different effects for each gene (Martinez and Modolell, 1991; Brand et al., 1993; Dominguez and Campuzano, 1993) In addition, NeuroD‐deficient mice developed an apparently normal nervous system (Naya et al., 1997), although NeuroD induced conversion of Xenopus non‐neural ectoderm into neuronal cells (Lee et al., 1995). These results are all consequences of the unique spatial and temporal patterns with which each bHLH gene is expressed during development. The loss‐of‐function phenotype associated with any particular bHLH protein is therefore thought to reflect the extent to which its expression pattern overlaps with functionally related bHLH proteins. In the case of SCL, functional redundancy with related bHLH proteins may explain the normal development of endothelial cells in SCL null mice.
Two recent lines of evidence support the novel concept that SCL plays an important role in endothelial development. Visvader et al. (1998) have partially rescued the haematopoietic phenotype of SCL null mice and thereby revealed a requirement for SCL in angiogenic remodelling of the yolk sac capillary network. In zebrafish, Liao et al. (1998) have shown that ectopic expression of SCL partially rescued the blood and vascular defects of cloche mutant embryos. Our results are consistent with both of these reports but, in addition, demonstrate that SCL specifies a common progenitor for both blood and endothelium at the expense of other mesodermal fates.
Our data also raise the issue of the relationship between haematopoietic/endothelial and pronephric precursors. In six somite stage (12 hpf) zebrafish embryos, SCL, GATA‐2, Flk‐1 and Pax‐2 expression domains overlapped in the posterior lateral mesoderm, consistent with the existence of a population of cells capable of giving rise to blood, endothelium and pronephric duct cells. Ectopic SCL expression did not alter the number of Pax‐2‐expressing cells at the 10 somite stage (13.5 hpf), suggesting that SCL did not change the number of putative multipotent progenitors in early lateral mesoderm. However, the subsequent loss of Pax‐2 expression and the abnormalities of pronephric duct development suggest that SCL was diverting the fate of the putative multipotent cells from pronephric duct to blood and endothelial fates or, alternatively, that enforced SCL expression perturbed the differentiation of pronephric duct cells.
The pattern of SCL expression was also remarkably conserved in the nervous system. Our data demonstrate SCL expression in the diencephalon, midbrain and hindbrain, together with transient expression in primary motoneurons of the spinal cord in zebrafish embryos, an expression pattern very similar to that described in mice (Green et al., 1992; Kallianpur et al., 1994; A.M.Sinclair, B.Göttgens, M.L.Stanley, S.Bahn, M.Gering, L.Pardanaud, M.Klaine, A.Bench, S.Aparicio, M.‐J.Sanchez, J.L. Fordham and A.R.Green, in preparation) This remarkable degree of conservation suggests an important role for SCL in the development of the nervous system. The studies reported here have focused on the consequences of ectopic SCL expression for the posterolateral mesoderm development. Future experiments with various specific neuronal markers may reveal aspects of SCL function in the developing brain.
Materials and methods
cDNA cloning and plasmid constructs
To screen zebrafish cDNA libraries for SCL‐related bHLH genes, we used a 160 bp DNA fragment amplified under low stringency conditions (annealing temperature of 45°C for 1 min) on zebrafish genomic DNA between primers A (5′ GTCTTCACCAACAGCCG 3′) and B (5′ CCAGGAAGTTGATGTACTT 3′) defined according to the 5′ and 3′ ends of the bHLH region of the SCL‐related bHLH gene SLP‐1 (Göttgens et al., 1998), respectively. First, 1.2×106 recombinant plaques of a somite stage zebrafish cDNA library (9–16 hpf) (David Grunwald) prepared in the vector λ ZAP II (Stratagene, La Jolla, CA) were screened at moderate stringency (hybridization at 65°C, washes at 65°C in 2× SSC and 0.1% SDS). From a tertiary screen, Bluescript phagemids were rescued from 13 positive λ ZAPII clones and their inserts partially sequenced. The original 13 clones represented three different cDNA clones and one genomic DNA clone. Only one of the cDNAs (plasmid pZE6 containing a 1.9 kb insert), found three times, encoded a protein very similar to SCL in other vertebrates as judged by a BLAST search (Altschul et al., 1990) of GenBank and EMBL databases. However, it turned out that the cDNA was not full‐length. Therefore, a second gastrula and somite stage cDNA library (Qi‐ling Xu) was screened under the same conditions described previously with a 400 bp SmaI–KpnI subfragment of pZE6 encoding the C‐terminus of the SCL protein behind the bHLH region. Two identical cDNA clones were obtained with 2527 bp inserts encoding a full‐length SCL reading frame (in plasmid pZN2). The DNA sequence is available from DDBJ/EMBL/GenBank under accession number DR_SCL AF038873.
DNA sequences were processed with the help of the Wisconsin Sequence Analysis Package of GCG. The BLAST program was applied to compare partial cDNA sequences with GenBank (Altschul et al., 1990) and EMBL databases. Multiple sequence alignments were done with the Clustal_X (Thompson et al., 1997) and Boxshade 3.21 programs. Radial phylogenetic trees generated with the PHYLIB program (version 3.572) (Felsenstein, 1989) were based on ClustalX alignments. Photographs of embryos and sections were scanned and processed using the Adobe Photoshop and Freehand programs.
Whole‐mount in situ hybridizations
Breeding zebrafish were maintained and embryos were raised and staged according to Westerfield (1993). Whole‐mount in situ hybridizations were carried out as previously described (Jowett and Yan, 1996). All RNA probes used were labelled with digoxigenin (DIG) except the Flk‐1, GATA‐2 and Pax‐2 probes which were used in double in situ hybridizations and were labelled with fluorescein. Detection of the antibody–alkaline phosphatase conjugate was done using BM‐Purple, Fast Red or 5‐bromo‐4‐chloro‐3‐indolylphosphate (BCIP, blue) (Boehringer Mannheim, Germany). After in situ hybridization, embryos were re‐fixed in 4% paraformaldehyde, transferred into 80% glycerol and photographed. Embryos that were to be sectioned were rehydrated and either embedded in 1.5% agarose/5% sucrose for cryostat sectioning or transferred into ethanol and embedded in JB4 methacrylate (Agar Scientific, UK) for microtome (Leica Jung RM2055) sectioning. Some of the sections were stained with toluidine blue. For sequential staining of thin sections, embryos were co‐hybridized with DIG and fluorescein‐labelled probes, stained for one marker in red, embedded in agarose and sectioned. Sections were documented and stained for the expression of the second gene in purple using BCIP/NBT (4‐nitro blue tetrazolium chloride).
Diaminofluorene staining and TUNEL assay
The staining of haemoglobinized cells with diaminofluorene was done as described by Weinstein et al. (1996). TUNEL assays on one and 10 somite stage, as well as 24 and 30 hpf zebrafish embryos, were performed. The assay was based on the original method (Gavrieli et al., 1992) that was adapted for zebrafish embryos by Juliet Williams using the ApopTag Terminal Transferase Kit (Oncor, USA).
Preparation of mRNA for injection and antisense RNA probes
To generate mRNA for injection, the entire SCL reading frame was first cloned into the vector pβUT‐MT2. The vector pβUT‐MT2 is a derivative of plasmid pβUT2 which consists of the 5′ and 3′ UTRs of Xenopus β‐globin (from pSP64T; Krieg and Melton, 1984) ligated into pBluescript (Stratgene, USA) with a polylinker replacing the BglII site of pSP64T. pβUT2‐MT was derived from pβUT2 by replacing its 3′ UTR with the 3′ UTR of pSP64T‐MEnT (a gift from Kathy Weston; Badiani et al., 1994) which also encodes a Myc epitope tag. To clone the SCL gene in‐frame with the Myc tag reading frame of the vector, we performed a PCR between the primers C (5′ CTCCTCTAGAGCCACCATGGAAAAACTGAAATCCGAGCAATT 3′) and D (5′ CTCCTCGAGACCGCTGGGCATTTCCGTCCAGCGG 3′) using Vent DNA polymerase (New England Biolabs, USA). The PCR fragment was gel purified, cut with XbaI and XhoI that were introduced into the sequence through the primers, and ligated into the XbaI‐ and XhoI‐digested plasmid. The DNA sequence of the insert was confirmed. The plasmid was linearized with EcoRI and the gene was transcribed in vitro with the help of the T3 mMessage mMachine Kit (Ambion, USA) yielding capped RNA for injection. The linearized plasmids pCSGFP2 [a derivative of pbGFP/RN3P (Zernicka‐Goetz et al., 1996) generated by J.Haseloff and J.Pines] and pSP64Tβ (a gift from D.Wilkinson) were transcribed with T7 and SP6 (Ambion, USA) to yield synthetic GFP and nuclear lacZ RNA for injection, repectively.
To prepare SCL antisense RNA probes, plasmid pZE62 was used. This plasmid was derived from pZE6 by restriction digestion with EcoRI and religation to yield pZE61 that was then cut with EcoRV and XhoI, made blunt‐ended with T4 DNA polymerase, and religated. The new plasmid named pZE62 thus only carried the EcoRI–EcoRV fragment 3′ of the SCL reading frame, allowing us to pick up the expression of the endogenous SCL gene only.
To make DIG‐ and fluorescein‐labelled antisense RNA probes, plasmids containing fragments of the SCL (pZE62), GATA‐1 (Detrich et al., 1995), Flk‐1 (Fouquet et al., 1997), Fli‐1 (L.Brown, T.F.Schilling, A.R.F.Rodaway, D.C. Hickleton, T.Jowett, P.W.Ingham, R.K.Patient and A.D.Scharrocks, in preparation), Paraxis (cloned by S.Shanmugalingam, gift from S.W.Wilson), Pax2 (Krauss et al., 1991) and MyoD (Weinberg et al., 1996) cDNAs were linearized, and run‐off transcripts were generated using either T3 or T7 RNA polymerase (Promega, USA) and digoxigenin‐11‐UTP or fluorescein‐12‐UTP (Boehringer Mannheim, Germany).
Two‐ or 4‐cell stage zebrafish embryos were injected with 200 pl of capped SCL RNA in the range of 180–500 ng/μl. mRNAs encoding GFP and nuclear lacZ were co‐injected to monitor injection efficiences by fluorescence in blue light (480 nm) using an FITC filter set or staining embryos with X‐Gal (Griffin et al., 1995), respectively. Injection of SCL mRNA at a concentration of 350 ng/μl was found to be optimal, giving a high number of embryos with the described phenotype and only ∼10% clearly abnormal or dying embryos that were not taken for further studies.
The authors thank Pamela K.Stockham and James G.R.Gilbert for assistance with the preparation of the manuscript and the figures, respectively. We would like to thank Anthony J.Bench and Juliet Williams for help with the screening of the zebrafish cDNA libraries and the TUNEL assays, respectively. We are grateful to Stephen W.Wilson and Nigel Holder, and the members of their laboratories as well as Marianne Bienz for helpful discussions and comments. We thank the BBSRC for providing the fish facility. We appreciate the help of Leonard I.Zon, Andrew D.Sharrocks, Stephen W.Wilson, Jose A.Campos‐Ortega and Mark C.Fishman who sent us the GATA‐1, Fli‐1, Paraxis, E12 and Flk‐1 expression plasmids, respectively. Work in the authors' laboratories was supported by the Wellcome Trust (A.R.G. and A.R.F.R.) and the Leukemia Research Fund (A.R.G.). M.G. had a fellowship from the Deutsche Forschungsgemeinschaft.
- Copyright © 1998 European Molecular Biology Organization