The zebrafish gene cloche acts upstream of a flk-1 homologue to regulate endothelial cell differentiation

Endothelial cells are the first cells to differentiate in the cardiovascular system; as such, they play an essential role in the formation and patterning of the vasculature in all vertebrates (Noden, 1991a; Poole and Coffin, 1991; Risau, 1991). The mesodermally derived endothelial cells and their progenitors, the angioblasts, are found in both extraembryonic and intraembryonic locations. On the yolk sac of most vertebrates, angioblasts in close association with hematopoietic progenitors form structures known as blood islands. Within the embryo itself, angioblasts give rise initially to the major blood vessels of the trunk, the dorsal aorta and axial vein, as well as to the endocardium of the heart. The in situ differentiation of angioblasts followed by their assembly into vascular channels, as observed in the trunk, is referred to as vasculogenesis.

The molecular mechanisms that regulate endothelial cell differentiation and vasculogenesis have remained largely unknown. The recent identification of two subfamilies of receptor tyrosine kinases whose expression is virtually restricted to endothelial cells and their progenitors, however, has provided an exciting entry point into these processes (reviewed by Mustonen and Alitalo, 1995). These receptor tyrosine kinases consist of the members of the vascular endothelial growth factor (VEGF) receptor family, namely Flk-1 and Flt-1, and the Tie (aka Tie-1) and Tek (aka Tie-2) orphan receptors. Mutational studies of Flk-1, Flt-1, Tie and Tek in the mouse indicate that these receptors play critical roles in vascular development (Dumont et al., 1994; Fong et al., 1995; Puri et al., 1995; Sato et al., 1995; Shalaby et al., 1995). Analyses of homozygous mutant mice indicate that Flk-1 is required for the formation of yolk-sac blood-islands and intraembryonic blood vessels as well as for hematopoiesis (Shalaby et al., 1995), whereas Flt-1 plays an essential role in regulating the assembly of vascular endothelium (Fong et al., 1995). Tek is also expressed in angioblasts and is required for vascular formation including angiogenesis (Dumont et al., 1994; Sato et al., 1995), a process defined as the sprouting of new blood vessels from pre-existing ones. Tie, which is expressed at later stages of endothelial cell differentiation, is required for the integrity and survival of endothelial cells (Puri et al., 1995; Sato et al., 1995). However, unlike Flk-1 or Flt-1, Tie does not seem to be essential for vasculogenesis.

We have recently initiated an effort to dissect the differentiation of the cardiovascular system using a genetic approach in the zebrafish (Stainier and Fishman, 1994), and have identified several mutations that appear to affect endothelial cell differentiation (Stainier et al., 1996). The cloche mutation is most notable in that it affects both the endothelial and hematopoietic lineages (Stainier et al., 1995). The endocardium is missing in cloche mutant embryos, but a small subset of endothelial cells, i.e. those lining the lower trunk and tail vessels, appear morphologically normal under the light microscope. In order to analyze further the endothelial defect in cloche, we have cloned the zebrafish homologues of flk-1 and tie and examined their expression in cloche mutants. We find that despite the similarity of the cloche and flk-1 mutant phenotypes, cloche and flk-1 are not linked and are thus different genes. Furthermore, we find that only cells in the lower trunk and tail regions of cloche mutant embryos express flk-1, but these cells do not go on to express tie. These data indicate that the cloche mutation deletes most endothelial cells and that the few endothelial cells that do differentiate to express flk-1 are blocked early in their differentiation pathway. Thus, cloche appears to affect endothelial cell differentiation by acting upstream of flk-1 either directly or indirectly.

Zebrafish were raised and handled as described by Westerfield (1993). Developmental time at 28.5°C was determined from the morphological features of the embryo (Kimmel et al., 1995). The original cloche allele, clo^m39 (Stainier et al., 1995), is a spontaneous allele identified in a semi-wild population from an Indonesian fish farm. This mutation is fully penetrant and exhibits complete and indistinguishable expressivity in a variety of genetic backgrounds tested. An ENU allele (clo^m378; Stainier et al., 1996), was used in this study (except for the linkage analysis where clo^m39 was used). Like clo^m39, clo^m378; is fully penetrant and exhibits complete and indistinguishable expressivity in a variety of backgrounds. In addition, clo^m39 and clo^m378; mutants are indistinguishable at both the cellular and molecular level further suggesting that these may represent null alleles.

In situ hybridizations were performed using digoxigenin-labeled RNA probes essentially as described by Oxtoby and Jowett (1993). Clo mutant embryos are easily distinguished from wild-type by 24 hours postfertilization (hpf). To stain for flk-1 expression at earlier stages, pools of 20-30 embryos were examined, revealing a clear 3:1 segregation of the embryos based on their staining pattern. After 24 hpf, at least two wild-type embryos were always added to the sorted mutant embryos to serve as positive controls. Two different clones were used to probe for flk-1 expression: the original clone covers part of the tyrosine kinase domain and extends to the end of the 3′ untranslated region (Fig. 1); a second clone was obtained from Len Zon (Harvard Medical School) and covers most of the coding region (M. A. Thompson and L. I. Zon, personal communication; manuscript in preparation). Probes derived from both clones gave exactly the same staining pattern. Embryos used for histology (Fig. 5) were overstained using a different protocol (Westerfield, 1993; M. A. Thompson and L. I. Zon, personal communication; manuscript in preparation) which calls for a 55°C hybridization (instead of the more routinely used 70°C). Histology was carried out as described (Stainier et al., 1993) and sections counterstained with nuclear fast red.

Expression patterns were documented and photographed on 160 ASA Ektachrome Tungsten film after clearing the embryos in ben-zylbenzoate/benzyl alcohol (2:1). Images from photographic slides were scanned on a Polaroid Sprint Scan 35 Slide Scanner. Composite figures were assembled, and contrast enhanced when necessary, using Adobe Photoshop Software (Adobe Corporation).

A PvuII polymorphism in the flk-1 gene was identified between a clo^m39 fish and a polymorphic wild-type fish (+^HK/+ ^HK) on Southern blots using a 640 bp BstXI/XbaI fragment of the zebrafish flk-1 cDNA as a probe. These fish were bred, and individual F₁ fish (with the genotype clo^m39/+ ^HK) were identified as carrying the clo mutation and then mated to obtain mutant embryos. Four pools of 30 mutants each were then used to analyze the segregation pattern of the PvuII poly-morphism.

A partial flk-1 cDNA was isolated during a PCR-based search for receptor tyrosine kinases expressed during early development (B. H., Ellen Wilson, Virginia Walter and D. J. G., unpublished). A 210 bp PCR fragment with similarity to the flk-1 tyrosine kinase domain was used to screen a cDNA library prepared from 20-28 hpf embryos, and a 2.5 kb clone was isolated and sequenced (Fig. 1A). Sequence alignment of the protein encoded by this clone versus human and mouse Flk-1 and its close relative Flt-1 reveals several regions of high homology (Fig. 1B), closer to Flk-1 than to Flt-1 (e.g. the tyrosine kinase domain). However, this protein also contains a small carboxy-terminal domain (15 amino acids long) that is found in Flt-1 but not in Flk-1. Further molecular studies failed to reveal the existence of another closely related gene, other than flt-4 (M. A. Thompson and L. I. Zon, personal communication; manuscript in preparation). Thus, we presume that this gene represents an ancestral gene that gave rise to both flk-1 and flt-1; and rather than giving this gene a new name, we will refer to it as flk-1.

The flk-1 transcript was first detected by whole-mount in situ hybridization in embryos between the 5- and 7-somite stages; in the trunk region, two bilateral stripes of flk-1-positive cells appear in the lateral plate mesoderm flanking the embryonic axis (Fig. 2A,C). These two stripes of flk-1-expressing cells expand rostrally and caudally as the embryo develops. In the head region, flk-1-positive cells appear by the 7-somite stage and align themselves in bilateral stripes which will extend caudally and presumably lay out the foundation for the head vasculature (Fig. 2B). flk-1-expressing cells are also found in the dorsal part of the hindbrain region (Fig. 2A,B), although this particular expression quickly weakens (Fig. 2D).

Shortly after the 13-somite stage, the trunk flk-1 staining reaches the tailbud region. In the mid-trunk lateral plate mesoderm, the flk-1-positive cells have begun to converge towards the ventral midline (data not included). The convergence of these cells extends caudally as the tail structure develops and the cells eventually merge into a single stripe which terminates at the ventral region of the tail (Fig. 2E). The flk-1-positive cells in the ventral midline later form the dorsal aorta and axial vein (see below). In the anterior region of the trunk lateral plate mesoderm, staining continues to extend rostrally (Fig. 2E). Further rostrally, in the heart region, where the primitive myocardial tubes are being assembled, staining appears between the bilateral stripes of flk-1-expressing cells at the 15-somite stage (data not included). By the 18-somite stage, a small but dense group of flk-1-positive cells is found in the midline region (Fig. 2E, arrow). We presume that this group of cells gives rise to the endocardial cells (Stainier et al., 1993), while the transverse stripes of flk-1-positive cells on either side of the endocardial progenitors likely form the endothelium lining the first aortic arches, which connect the outflow tract of the heart to the arterial vasculature (Rieb, 1973).

At the 20-somite stage, the flk-1-positive cells from the anterior trunk region meet the flk-1-positive cells from the head (Fig. 2E). By the 24-somite stage, the dorsal aorta (DA) and axial vein (AV) are clearly distinct in the trunk region (Fig. 2F,G). Some flk-1-positive cells emerge between the somites, apparently sprouting from the dorsal aorta (Fig. 2F,G). These intersomitic vessels are being formed by angiogenesis. At 24 hpf, when the basic vascular system of the whole embryo has been laid out and circulation can be observed, flk-1 clearly labels all the endothelial cells lining the vasculature (Fig. 2H,I). On the basis of these staining patterns it appears that, at least until 36 hpf (data not included), all endothelial cells express flk-1.

Because of the similarity between Flk-1 deficient mice and zebrafish clo mutants, we wanted to analyze linkage between zebrafish flk-1 and clo. We initially identified a PvuII poly-morphism in the flk-1 gene between a clo^m39 fish and a poly-morphic wild-type fish (+ ^HK /+ ^HK) and followed the segregation of this polymorphism in a mapping cross. Fig. 3 shows that the 7.4 kb wild-type allele is present in two of the four batches of mutant embryos (30 embryos per batch), thus demonstrating that flk-1 and clo are not linked.

We used flk-1 as a molecular marker for endothelial cells and their progenitors to characterize further the clo mutant phenotype. Up to the 20-somite stage, the flk-1-positive cells found in the anterior head region and trunk lateral plate mesoderm of wild-type embryos appear to be missing in clo mutants (Fig. 4A). At the 20-somite stage, faint staining is observed in the tail region, near the caudal end of the yolk extension (Fig. 4B, large arrow). This staining becomes more distinct with time yet is always much weaker than that seen in wild-type siblings (Fig. 4C-G). In 30 hpf wild-type embryos, the endothelial cells of the entire vasculature, including the blood vessels in the head, dorsal aorta and axial vein in the trunk, clearly express flk-1 (Fig. 4D,E). In clo mutant embryos, however, flk-1 message is detected only in the lower trunk and tail regions (Fig. 4F,G). Examination of 36 hpf living mutant embryos with Nomarski optics reveals the presence of endothelial-like cells exclusively in this region (see Fig. 3B in Stainier et al., 1995). Thus, in clo mutant embryos, extensive aspects of flk-1 expression are missing; distinct levels of endothelial expression appear around the 20-somite stage only in cells of the lower trunk and tail regions. Faint expression is also seen in the dorsal part of the hindbrain (small arrows) and the tailbud (arrowheads) in both wild-type and mutant embryos (Fig. 4A-C). The exact nature of the flk-1-expressing cells in these regions is not clear, especially as flk-1 has recently been reported to be expressed outside the vasculature (e.g. in retinal progenitor cells; Yang and Cepko, personal communication).

Wild-type and clo mutant embryos were also processed for histology after overstaining for flk-1 expression. Examination of serial sections confirms that only cells in the lower trunk and tail regions of clo mutant embryos express flk-1 (Fig. 5). In the mid-trunk region for example, while vessels are clearly delineated in wild-type embryos (Fig. 5D), no flk-1 expression is seen in the mutants (Fig. 5G). In fact, somitic tissue appears to occupy the space where the dorsal aorta and axial vein usually form. Altogether, these data indicate that most endothelial cells are missing from clo mutant embryos.

To characterize further the flk-1-positive cells found in the lower trunk and tail regions of clo mutant embryos and to determine whether clo also plays a role in the differentiation of these cells, we examined the expression of tie (B. Bell, D. Y. R. Stainier and K. Peters, unpublished data). tie encodes an endothelial-specific receptor tyrosine kinase whose expression in mouse marks later stages of endothelial cell differentiation (Korhonen et al., 1994; Dumont et al., 1995). Similarly, in zebrafish, tie expression marks later stages of endothelial cell differentiation, first appearing around the 18-somite stage (data not included). In 30 hpf wild-type embryos, tie expression is detected in all endothelial cells, and similar to flk-1 expression, high levels of tie expression are found in the lower trunk and tail regions (Fig. 6A,B). In clo mutants, however, there are no tie-positive cells at this or later stages (Fig. 6A,C). Thus, the flk-1-positive cells found in the lower trunk and tail regions of clo mutant embryos do not go on to express tie-1, indicating a block early in their differentiation. On the basis of these data, we conclude that clo also functions in the endothelial cells of the lower trunk and tail regions.

We previously reported that the clo mutation affects blood cell differentiation (Stainier et al., 1995): The number of blood cells is greatly reduced in clo mutant embryos and, at early stages, the hematopoietic tissues show no expression of GATA-1 and GATA-2, two key hematopoietic transcription factor genes first expressed at the end of gastrulation. Because clo mutants display a delayed onset of flk-1 expression in a region where blood cell differentiation is known to occur (Detrich et al., 1995; Stainier et al., 1995), we wished to examine whether hematopoiesis was also delayed. In 30 hpf wild-type embryos, GATA-1 is expressed mainly in erythroid cells accumulated on the yolk sac, in the heart, and in a small group of cells in the lower trunk and tail regions just posterior to the caudal end of the yolk extension (Fig. 7A). (The location of this small group of GATA-1-positive cells within the stem cell population of the posterior intermediate cell mass (ICM) indicates that they are newly differentiated blood cells, not yet recruited into the circulation; Detrich et al., 1995; Stainier et al., 1995). In clo mutants, no GATA-1-positive cells are found on the yolk sac or in the heart; however, there are a few GATA-1-expressing cells in the ventral region of the tail (Fig. 7A-C). These GATA-1-positive cells are always within the region outlined by the flk-1-expressing cells (Fig. 7C,D).

Expression of flk-1 is, to date, the earliest marker for endothelial cells and their progenitors. In the mouse, flk-1 expression can be detected in presumptive mesodermal yolk-sac blood-island progenitors as early as 7.0 days postcoitum (i.e. early to mid-gastrulation) (Shalaby et al., 1995), and in avians, flk-1 (also called Quek1) is expressed in the mesoderm from the onset of gastrulation (Eichmann et al., 1993). In zebrafish, flk-1 expression is first detected during early somi-togenesis (a process which initiates much earlier in fish than in amniotes), at a time when differentiation markers for other car-diovascular cell types are also appearing, e.g. Nkx-2.5 in myocardial cells (Lee et al., 1996; Jon Alexander and D.Y.R.S., unpublished) and GATA-1 in blood cells (Detrich et al., 1995; Stainier et al., 1995). flk-1 expression identifies two physically separate populations of endothelial progenitor cells: one in the head and another in the trunk. In agreement with this pattern of flk-1 expression, lineage data indicate that endothelial prog-enitors are located throughout the marginal zone at the late blastula stage (Warga, 1996): trunk and tail endothelial progenitors are concentrated in the ventral half of the marginal zone whereas head endothelial progenitors span the entire marginal zone. Endocardial progenitors are themselves located in the ventral marginal zone (Lee et al., 1994; Warga, 1996).

Careful examination of flk-1 expression in embryos between the 14-to 18-somite stages reveals that the endocardial pro-genitor cells appear to originate from the anterior group of angioblasts (i.e. those that populate the head region). Indeed, several hours before the head and trunk populations of angioblasts join, flk-1 staining appears in the presumptive heart region between the bilateral stripes of flk-1-expressing cells. The subsequent aggregation of flk-1-positive cells in the midline region occurs at the precise location where we have previously observed and described the endocardial progenitor cells at later stages (Stainier et al., 1993). These data indicate that the endocardium originates from the cephalic paraxial and/or anterior lateral plate mesoderm, and undergoes minimal medial migration to reach its final destination. This interpretation is in agreement with previous studies in avian embryos which concluded that the endocardium arises from the head mesoderm (Noden, 1991b; Coffin and Poole, 1991). Thus, the endocardium does not appear to originate from the trunk lateral plate mesoderm as previously suggested (reviewed by Noden, 1991a). These data and the lineage data from Warga and Nüsslein-Volhard also indicate that, based on their migration pattern, there are at least two populations of cells in the ventral marginal zone of zebrafish blastulae. Myocardial and endocardial progenitors migrate dorso-anteriorly towards the head region during gastrulation whereas blood as well as trunk and tail endothelial progenitors take a more posterior route. Deter-mining the exact migration pathways of these cells and proving unequivocally that endocardial cells indeed come from the head region will require careful lineage and cell tracking analyses.

Finally, these data also shed some light on the postulated existence of the hemangioblast, defined as the common meso-dermal precursor for both endothelial and hematopoietic cells (Sabin, 1920). In zebrafish blastulae, blood progenitors lie in the ventral marginal zone together with trunk and tail endothelial progenitors while head endothelial progenitors are found throughout the marginal zone. Thus, while trunk and tail endothelial cells may share part of their lineage with blood cells, it is clear that some head endothelium is already distinct from the blood lineage at the blastula stage. These data indicate that some, but not all, endothelial cells may originate from hemangioblasts and that there may in fact be two distinct endothelial lineages, only one of them related to hematopoiesis. Recent transplantation experiments in chick also indicate the presence of two distinct endothelial lineages, only one of them related to hematopoiesis (Pardanaud et al., 1996).

The clo mutation affects the differentiation of both the endothelial and blood cell lineages. Similarly, Flk-1 deficient mice lack most endothelial and blood cells (Shalaby et al., 1995). We therefore analyzed linkage between zebrafish flk-1 and clo and found no linkage between these two genes, indicating that clo does not encode zebrafish Flk-1.

As assessed by flk-1 expression, the clo mutation appears to block the generation of most endothelial cells. In clo mutants, flk-1 expression is not detected in regions of the early embryo where endothelial cells and their progenitors normally reside. Around the 20-somite stage, a few cells in the tail region are clearly-expressing flk-1, although at a much reduced level as compared to wild-type (Fig. 4B). This staining becomes more pronounced by 24 hpf (Fig. 4C) and at 30 hpf, the extent of flk-1 staining in mutant embryos (Fig. 4F,G) can be fully appreciated as compared to wild-type (Fig. 4D,E). flk-1-positive cells are only present in the lower trunk and tail regions and this staining is weaker than in wild-type, apparently because there are fewer flk-1-expressing cells in mutant embryos (Fig. 5E,H). Examination of serial sections of flk-1-stained embryos indicates that all endothelial cells in both wild-types and mutants express flk-1. Thus, most endothelial cells appear to be missing in clo mutant embryos confirming the original Nomarski optics observations in live embryos (Stainier et al., 1995).

In the lower trunk and tail regions of clo mutant embryos, a few cells express flk-1 (Fig. 3D,E). These cells do not go on to express tie (Fig. 5A,C) indicating that they are blocked at an early stage of their differentiation. Thus, clo appears to affect the differentiation of all endothelial cells, although the (presumed) absence of its function has different effects in the head and trunk versus the tail endothelial progenitor populations. The molecular basis for this spatial distinction is not clear at this point. It is possible, for example, that clo acts in conjunction with one or more genes to regulate directly or indirectly the expression of flk-1, and that in the absence of clo function these other genes are capable of inducing limited flk-1 expression in only the lower trunk and tail regions of the embryo. It is also interesting to note that in clo mutant embryos endothelial cells differentiate mostly in the tail region, and that tail formation has been hypothesized to involve mechanisms distinct from primary body formation (which gives rise to the head and trunk regions) (Holmdahl, 1925; Griffith et al., 1992).

In Flk-1 deficient mice as in clo mutants, tie is not expressed, indicating the absence of more mature endothelial cells (Shalaby et al., 1995). The absence of tie expression is not an indirect consequence of the lack of circulation seen in these mutants; indeed, we find normal tie expression in silent heart mutants (a mutation originally identified in Eugene, Oregon by Walker and Kimmel, which blocks the initiation of the heart beat and causes a lack of circulation). Furthermore, in Flk-1 deficient mice, flk-1 expression is activated appropriately in the developing mesoderm, in contrast to the lack of early flk-1 expression observed in clo mutants. Finally, the necrosis present in Flk-1 deficient mice prevents analysis of older mutant embryos, but it is possible that in these mutants, as in clo homozygotes, a small subset of vessels would form in the tail region at late stages.

In clo mutants, early flk-1 expression is mainly absent. One model is that clo plays a role within the endothelial lineage in the transduction of the signal that induces flk-1 expression, or in the differentiation of the cells that normally go on to express flk-1. Several additional lines of evidence support such a model. First, clo acts cell-autonomously within the endothelial lineage (Stainier et al., 1995). Second, using GATA-1 and -2 as probes, a molecular defect can be detected in clo mutant embryos by the end of gastrulation (i.e. several hours before flk-1 expression is first detected), arguing that the clo gene product first functions before flk-1 is normally expressed. We therefore propose that clo acts upstream of flk-1, directly or indirectly, to regulate endothelial cell differentiation.

Analysis of clo mutant embryos at early stages revealed no GATA-1 or -2 expression, indicating that clo affects hematopoietic differentiation (Stainier et al., 1995). Because clo also affects endothelial cell differentiation, we presented several models to explain this phenotype. In one of these models, we hypothesized that endothelial cells or their pro-genitors induce the differentiation and/or maturation of the physically adjacent hematopoietic progenitors. The observation of flk-1-positive cells in 24 hpf clo mutant embryos led us to examine GATA-1 expression at these late stages. In agreement with an induction model, we find some GATA-1-expressing cells within the region lined by the flk-1-positive endothelial cells. Additional experiments, however, must be performed to test the validity of this model. For example, cell transplantation experiments to analyze the cell-autonomy of the blood defect in clo mutants would be useful, but the very low frequency with which one can transplant cells that give rise to either trunk endothelial cells or blood cells (due to their close lineal relationship) makes this approach impractical.

The restricted differentiation of blood in the ventroposterior-most region of clo mutant embryos also argues for a ventral source of blood inducing signal(s) as has been previously proposed in Xenopus (Kelley et al., 1994; Walmsley et al., 1994). In this model, blood cell differentiation would occur in clo mutants only in the region (or derivatives of the region) exposed to the highest concentration of inducing signal(s).

We must also consider the possibility that our clo mutations are weak alleles, and that residual clo function leads to limited endothelial and blood cell differentiation and/or maturation. Arguing against this possibility is the fact that the two existing clo alleles are phenotypically indistinguishable at the cellular and molecular levels, even when outcrossed into widely poly-morphic strains, suggesting that they represent loss-of-function mutations.

In summary, we have shown that clo is an early regulator of endothelial cell differentiation and that it is likely to function upstream of flk-1, the earliest endothelial marker isolated to date. We have also collected additional data consistent with a model in which endothelial cells play a role in blood cell differentiation and/or maturation. The isolation and characterization of the clo gene as well as the analysis of its expression pattern will shed further light on the mechanisms regulating the differentiation of the endothelial and hematopoietic lineages.

We thank Marina Gerson and Virginia Walter for excellent technical help, Jon Alexander, Leon Parker, Debbie Yelon and other members of the lab for comments on the manuscript, Rachel Warga and Len Zon and his group for sharing unpublished data and discussions. This work was supported in part by the NIHLB Program of Excellence in Molecular Biology (W. L.), a Postdoctoral Fellowship from NSERC Canada (B. W. B.), a James S. McDonnell Foundation Scholarship and NIH HL55265 (K. P.), the Primary Children’s Medical Center Foundation (Utah) and NSF (D. J. G.), a University of California CRCC grant, the UCSF Simon Fund, the March of Dimes, the David and Lucile Packard Foundation, and NIH HL54737 (D. Y. R. S.).