Basic fibroblast growth factor positively regulates hematopoietic development

The establishment of blood islands in the extra-embryonic yolk sac marks the onset of hematopoiesis and vasculogenesis in the developing embryo. These blood islands develop from aggregates of mesodermal cells that colonize the presumptive yolk sac at approximately 7 days post-coitum (dpc). Despite great interest in understanding signals/factors governing the initiation of hematopoietic differentiation in the developing embryo, very little knowledge exists regarding mechanisms involved in hematopoietic specification. Most likely, factors involved in mesoderm induction/patterning events, such as bFGF, activin A, and bone morphogenetic proteins (BMPs), also regulate hematopoietic specification. However, methods previously available to address such an issue have been limited. For example, the use of mice lacking mesoderm-inducing factor genes to delineate the roles of specific function in hematopoietic commitment is complicated by early embryonic lethality prior to blood formation (Deng et al., 1994; Yamaguchi et al., 1994; Mishina et al., 1995; Winnier et al., 1995). Furthermore, tools to study hematopoietic commitment, i.e. the origin of the initial cell population committed to hematopoietic cell lineage, have also been limited.

In vitro differentiated ES cells (reviewed by Keller, 1995) are valuable for the study of early hematopoietic lineage differentiation, as demonstrated by the identification of the very primitive, hematopoietic committed cells which otherwise have been elusive in embryos (Choi et al., 1998). Blast-colony-forming cells (BL-CFCs), generated from in vitro differentiated progeny (embryoid bodies, EBs) of embryonic stem (ES) cells, represent the long sought common progenitor of hematopoietic and endothelial cells, the hemangioblast (Sabin, 1920; Murray, 1932; Wagner, 1980). They form blast colonies in response to vascular endothelial growth factor (VEGF) in methylcellulose cultures. Cells within the blast colonies, blast cells, express a set of genes common to both hematopoietic and endothelial lineages, including scl, CD34 and the VEGF receptor, Flk1. Most importantly, primitive, definitive hematopoietic, and endothelial cell progenitors are present within blast cells as revealed by replating in medium containing both hematopoietic and endothelial cell growth factors (Choi et al., 1998; Kennedy et al., 1997).

Several studies suggest that mesoderm-inducing factors are able to affect EB development, as well as hematopoietic differentiation. For example, the addition of bFGF or activin A to differentiating EBs enhances brachyury gene expression, a marker for mesodermal tissue (Yamada et al., 1994). This observation suggests that cells within EBs can respond to external signals, and therefore argues that one could utilize this system for examining factors involved in hematopoietic specification. Johansson and Wiles (1995) have demonstrated that BMP4 induced the expression of the embryonic globin gene, βH1, from day-5 EBs generated in serum free, chemically defined medium (CDM). Similarly, Huber et al. (1998) demonstrated that BMP4 in combination with activin A or bFGF induces the generation of erythroid cells in Xenopus animal cap cultures. These studies suggest that hematopoietic development could be governed by a combination of several, rather than individual, mesoderm inducing factors.

Since BL-CFCs represent the earliest committed hematopoietic precursors so far identified, they provide an ideal model for studying factors that regulate the onset of hematopoietic development. In this study, we investigated the effect of bFGF and Activin A on the generation of the BL-CFC and Flk1⁺ cells and analyzed fgfr1^−/− and scl^−/− ES cells to further characterize factors/signals regulating hemangioblast development. Our results suggest that bFGF or Activin A, singularly or in concert, positively regulate hematopoietic development and that Activin A-mediated BL-CFC generation requires bFGF-mediated signals. Our studies also suggest that the hemangioblast expresses both Flk1 and SCL.

ES cells were maintained on STO feeder cells in the presence of LIF. EBs were generated as described by Choi et al. (1998). Basic FGF or activin A was added at the initiation of EB differentiation. The resulting EBs (day-2.75) were analyzed for the blast colonies, as well as, FACS analysis for Flk1⁺ cells. Basic FGF was purchased from Upstate Biotechnology (Lake Placid, NY), and activin A was kindly provided by Dr A. Parlow at NIDDK’s National Hormone & Pituitary Program.

Blast colonies, primitive erythroid, and myeloid colonies were generated as described previously (Keller et al., 1993; Choi et al., 1998). Briefly, blast colonies were generated from day-2.75-3 EBs in the presence of VEGF (5 ng/ml), KL (1% conditioned medium or 100 ng/ml purified) and D4T conditioned medium (CM) at 25%. Primitive erythroid progenitors from day-4 EBs were analyzed in the presence of plasma derived serum (PDS, Antech; 10%), ascorbic acid (12.5 ng/ml), L-glutamine (2 mM), transferrin (300 μg/ml; Boehringer Mannheim), protein free hybridoma media II (PFHMII, Gibco-BRL; 5%), and MTG (4.5×10⁻⁴ M) with erythropoietin (Amgen, 2 units/ml). EryP colonies were counted 4-6 days after replating. Other myeloid progenitors were analyzed by replating day-9 to -10 EBs in PDS (10%, Antech), ascorbic acid (12.5 ng/ml), L-glutamine (2 mM), transferrin (300 μg/ml; Boehringer Mannheim), protein free hybridoma medium II (PFHMII, Gibco-BRL; 5%), and MTG (4.5×10⁻⁴ M) containing Epo (2 units/ml), KL (1%), IL-3 (1%), IL-1 (10 ng/ml), IL-11 (25 ng/ml), GM-CSF (3 ng/ml), M-CSF (5 ng/ml), G-CSF (30 ng/ml), IL-6 (5 ng/ml), LIF (1 ng/ml), and VEGF (5 ng/ml). Hematopoietic colonies were counted 7-10 days after the replating. IL-1, IL-11, GM-CSF, M-CSF, IL-6, LIF and VEGF were purchased from R&D Systems. KL was obtained from medium conditioned by CHO cells transfected with a KL expression vector (kindly provided by Genetics Institute) or purchased from R&D Systems. Epo was purchased from Amgen (Thousand Oaks, CA). IL-3 was obtained from medium conditioned by X63 Ag8-653 myeloma cells transfected with a vector expressing IL-3 (Karasuyama and Melchers, 1988).

For Flk1 monoclonal antibody staining (Kabrun et al., 1997), EBs were treated with trypsin/EDTA for 3 minutes, cells were passed through a 20-gauge needle to produce a single cell suspension, counted, and centrifuged. Cells were resuspended to a concentration of 5×10⁶ cells/ml in 2.4G2 supernatant/2% HI rat serum to block antibody from binding to Fc gamma receptors II and III (CD16 and CD32, Unkeless, 1979). Cells were placed into each well of a V-shaped 96-well plate at 5×10⁵ cells/well followed by incubation on ice for 30 minutes. Subsequently, biotinylated anti-Flk1 antibody freshly diluted in wash buffer (20%FCS/PBS/0.1% sodium azide) was added and incubated on ice for 15 minutes. Cells were washed 3 times with 2% FCS/PBS/0.1% sodium azide. Streptavidin-phycoerythrin (secondary reagent) (Pharmingen) freshly diluted in wash buffer was added and incubated on ice in the dark for 10 minutes. Cells were washed for a total of 3 times, resuspended in wash buffer and transferred to 5 ml polypropylene tubes for analysis. Cells were analyzed on a FACSCaliber (Becton-Dickinson). FACS data were analyzed with CellQuest software (Becton-Dickinson). For Flk1⁺ cells sorting, cells were treated the same except that they were filtered through 40 μm nylon mesh prior to sorting. Flk1⁺ cells were sorted using FACSVantage (Becton-Dickinson).

β-galactosidase activity within Flk1⁺/^- EBs was measured as described by Schuh et al. (1999). Briefly, EBs were trypsinized to produce a single cell suspension. 1×10⁶ cells were resuspended in PBS, 10% FCS (Hyclone), 10 mM Hepes, pH 7.2 and 300 μM chloroquine and incubated at 37°C for 20 minutes. Pre-warmed 2 mM FDG (Molecular Probes) was added in equal volume to the cells and incubated for exactly 1 minute. The FDG was loaded into the cells (hypotonic state) and an isotonic shock was introduced with the addition of 1.8 ml of ice-cold staining medium (PBS, 10% Hyclone FBS, 300 μM chloroquine, 1 mM phenylethyl β-D-thiogalactopyranoside (PETG) and 1 μg/ml propidium iodide (PI)). Cells were maintained at 4°C in the dark until analysis. Endogenous β-galactosidase present in the lysosome was inhibited by pre-treatment with chloroquine. In addition, the competitive inhibitor, PETG, was added to stop the reaction by outcompeting FDG from being cleaved by β-galactosidase.

For endothelial cell marker staining, EBs were grown in methylcellulose or liquid culture, collected and treated with collagenase (0.25%/20%FCS/PBS) for 1 hour and gently disassociated to single cells by passage through a 20 gauge needle. After the Fc receptor blocking step (see above) cells were stained with the appropriate primary antibody. The following antibodies were used in these experiments: biotinylated anti-Flk1 (89B3A5), biotinylated anti-PECAM-1 (MEC 13.3; Pharmingen), anti-P-selectin (RB40.34; Pharmingen), anti-VE-cadherin (11D4.1; Pharmingen) and anti-CD34-FITC (RAM34; Pharmingen). The secondary reagent for the unconjugated antibodies was goat anti-rat IgG-FITC (Caltag Laboratories).

RNA isolation and RT-PCR were performed as described by Chomczynski and Sacchi (1987); Chelly et al. (1988). All RNA samples were treated with DNaseI (amplification grade from Gibco/BRL) before cDNA synthesis to eliminate any contaminating genomic DNA. Specific primers used are as follows (Keller et al., 1993; Johansson and Wiles, 1995).

HPRT, sense, 5′CACCTGGCACTCTCCACCTTC3′; antisense, 5′GCTGGTGAAAAGGACCTCT3′.

rex1, sense, 5′CGTGTAACATACACCATCCG3′, antisense, 5′GAAATCCTCTTCCAGAATGG3′.

fgf5, sense 5′GGCAGAAGTAGCGCGAGC3′; antisense 5′CCGTAAATTTGGCACTTG3′.

nodal, sense, 5′CCGTCCCCTCTGGCGTACATG3′; antisense 5′GACCTGAGAAGGAATGACGG3′.

GATA-4, sense 5′CTAAGCTGTCCCCACAAGGC3′; antisense 5′CAGAGCTCCACCTGGAAAGG3′.

Myf5, sense 5′GAAGGCTCCTGTATCCCCTCAC3′; antisense 5′GTTCTCCACCTGTTCCCTCAGC3′.

α-cardiac MHC, sense 5′CTCGCAGAACACCAGCC3′; antisense 5′GTAGGTGAGCTCCTTGAT3′.

β-H1, sense 5′AGTCCCCATGGAGTCAAAGA3′; antisense 5′CTCAAGGAGACCTTTGCTCA3′.

β-major, sense 5′CTGACAGATGCTCTCTTGGG3′; antisense 5′CACAAACCCCAGAAACAGACA3′.

T gene, sense, 5′CATGTACTCTTTCTTGCTGG3′; antisense, 5′GGTCTCGGGAAAGCAGTGGC3′.

BMP2, sense, 5′GAATCAGAACACAAGTCAGT; antisense, 5′GTTTGTGTTTGGCTTGACGC3′.

BMP4, sense, 5′TGTGAGGAGTTTCCATCACG; antisense, 5′CAGCGAAGGACTGCAGGGCT3′.

TGFβ1, sense, 5′CCAAAGACATCTCACACAGT3′; antisense, 5′TAACGCCAGGAATTGTTGCT3′.

VEGF, sense, 5′TCAGAGAGCAACATCACCAT3′; antisense, 5′ACCGCCTCGGCTTGTCACAT3′.

scl, sense, 5′ATTGCACACACGGGATTCTG3′; antisense, 5′GAATTCAGGGTCTTCCTTAG3′.

Flk1, sense, 5′CACCTGGCACTCTCCACCTTC3′; antisense, 5′GATTTCATCCCACTACCGAAAG3′.

KL, sense, 5′GACTGTGTGCTCTCTTCAAC3′; antisense, 5′CTTGCAAAACCTCCAGAGTC3′.

c-kit, sense, 5′TGTCTCTCCAGTTTCCCTGC3′; antisense, 5′GCAAGGACAAGGGAACCTCTG3′.

Single cell RT-PCR was performed following Brady et al. (1990). Briefly, single Flk1⁺ cells were picked into 4 μl of lysis/1st strand buffer by micromanipulation, reverse transcribed, poly(A) tailed using terminal transferase. Total cDNA was amplified using oligo(dT) as a primer (5′GTTAACTCGAGAATTCT₂₄3′). After one round of reamplification using 1 μl of the primary PCR products as template, PCR products were separated on an agarose gel, transferred to a nylon membrane, hybridized with ³²P randomly primed cDNA probe corresponding to the 3′ end of the L32, Flk1 or fgfr1 gene. After the hybridization, the blot was washed at high stringency and exposed to an X-ray film.

In an effort to determine if bFGF plays a role in hematopoietic development, we initially analyzed fgfr1^−/− ES cells in vitro for their potential to give rise to BL-CFCs and hematopoietic cells. fgfr1^−/− ES cells differentiated poorly in vitro (not shown), suggesting that the Fgfr1 mediated signal is important for EB cell development. Hematopoietic development was also defective in fgfr1^−/− EBs, such that the number of blast colonies, primitive erythroid and myeloid progenitors were all greatly reduced in fgfr1^−/− EBs compared to those in fgfr1^+/+ EBs (Fig. 1A-C). The number of secondary EBs, which arise from undifferentiated ES cells within EB cells, was greatly increased as shown in Fig. 1A’-C’. This suggested that the Fgfr1-mediated signal is necessary for ES cell differentiation and also for hematopoietic lineage development.

To better understand the hematopoietic defect observed in fgfr1^−/− EBs, wild-type, as well as, fgfr1^−/− EBs obtained from different time points were subjected to gene expression analysis using semi-quantitative RT-PCR. As shown in Fig. 2, rex1, expressed in the inner cell mass and undifferentiated ES cells, was down regulated as EBs developed in wild type cells. However, rex1 was still expressed in fgfr1^−/− EBs even at day 10, confirming the notion that fgfr1^−/− ES cells differentiated poorly in vitro and gave rise to secondary EBs when replated. The expression of fgf5, nodal and T gene in primitive ectoderm and the early primitive streak region and mesoderm, respectively, was still detectable in later fgfr1^−/− EBs but not in fgfr1^+/+ EBs. The expression of BMP2 and TGFβ1 was greatly reduced in fgfr1^−/− EBs compared to controls. However, BMP4 gene expression was not affected. This result suggested that the Fgfr1 signal is necessary for the BMP2 and TGFβ1 gene expression, but not for the expression of BMP4. The expression of VEGF and GATA4, in primitive endoderm, visceral endoderm and parietal endoderm (Miquerol et al., 1999) and visceral endoderm (Arceci et al., 1993), respectively, was down regulated in fgfr1^−/− EBs without kinetic change. The expression of the cardiac α-myosin heavy chain gene was down regulated in late fgfr1^−/− EBs (days 8 and 10), although the expression of myf5, the first myogenic marker, was up regulated in fgfr1^−/− EBs. The expression of genes including Flk1, scl, c-kit, kit ligand (KL), βH1, and β-major globin, was greatly reduced in fgfr1^−/− EBs compared to controls, confirming the finding that the hematopoietic differentiation is impaired in fgfr1^−/− EBs.

As the overall in vitro differentiation potential of fgfr1^−/− ES cells was impaired, it was difficult to evaluate whether bFGF plays a role, if any, in hematopoietic differentiation. Therefore, we assessed the effect of bFGF on hematopoietic development by adding it during EB differentiation and determining if it could affect BL-CFC development. As shown in Fig. 3A, EBs differentiated for 2.75 days in the presence of bFGF contained higher cell numbers compared to control EBs. This suggested that bFGF enhanced EB proliferation and/or differentiation and is consistent with the finding that Fgfr1 is important for embryonic cell proliferation (Deng et al., 1994; Yamaguchi et al., 1994) and EB differentiation. If hematopoietic progenitors expanded preferentially within EBs in response to bFGF, we expected that the BL-CFC frequency would increase as well. In contrast, if the effect of bFGF were to enhance EB cell proliferation non-specifically, one would predict that the BL-CFC frequency would remain the same. As shown in Fig. 3B and C, EBs grown for 2.75 days in the presence of bFGF contained a higher BL-CFC frequency compared to those from control EBs. The increase in blast colony number in the presence of bFGF was dose-dependent, such that the number of colonies increased with higher concentrations of bFGF used. The developmental potential of blast cells generated from bFGF-stimulated EBs remained the same, since a similar percentage of blast colonies gave rise to both hematopoietic and endothelial cells (30/48 from control and 34/48 blast colonies from EBs differentiated with bFGF). Taken together, it is likely that bFGF not only enhanced EB proliferation and/or differentiation, but also preferentially promoted the development of hematopoietic committed progenitors.

In Xenopus, it has been shown that basic FGF and activin A synergize in inducing mesoderm and that mesoderm induction by activin A requires FGF-mediated signals (Cornell and Kimelman 1994; LaBonne and Whitman 1994). To determine if bFGF and Activin A could also synergize in BL-CFC development, EBs were developed in the presence of bFGF, activin A, or bFGF and activin, replated, and the number of resulting blast colonies was compared. As shown in Fig. 4A,B, activin A also promoted generation of BL-CFC. The increase in BL-CFC frequency by activin A was more obvious at the lower concentrations of activin A used. Furthermore, the number of blast colonies from EBs obtained with activin A and bFGF together was much greater than that from EBs with activin or bFGF alone (Fig. 4C,D). Again, this increase was more apparent at lower concentrations of activin A. From these data, we concluded that bFGF or activin A singularly promoted BL-CFC development and that these two factors showed an additive or synergistic effect on BL-CFC generation.

To determine if activin A-mediated BL-CFC development also requires the Fgfr1 signal, fgfr1^+/+, fgfr1^+/−, and fgfr1^−/− ES cells were differentiated in vitro, in the presence of activin A, and the resulting blast colonies were examined. As shown in Fig. 4E, both fgfr1^+/+ and fgfr1^+/− EB cells gave rise to an increased number of blast colonies in the presence of activin A. However, fgfr1^−/− EBs developed in the presence of activin A still failed to generate any BL-CFCs, and only gave rise to secondary EBs when replated. This suggests that activin A cannot rescue the hematopoietic defect observed in fgfr1^−/− ES cells and that activin A-mediated BL-CFC generation requires the Fgfr1 signal.

We have previously demonstrated that vascular endothelial growth factor (VEGF) is critical for BL-CFCs to form blast colonies in methylcellulose cultures and that blast cells, progeny of BL-CFC, express the Flk1 gene (Choi et al., 1998; Kennedy et al., 1997). These observations would indicate that BL-CFCs express Flk1. To determine if this is the case, day-2.75 EB cells were stained with Flk1 monoclonal antibody (Kabrun et al., 1997) and Flk1⁺ cells were sorted by FACS. Sorted cell populations (Flk1⁺ vs. Flk1⁻) were replated in blast replating medium. As shown in Figs 5A and 7, Flk1⁺ cells were greatly enriched for BL-CFCs. These results suggest that BL-CFCs express Flk1 and therefore argue that Flk1 can be used as a marker for the hemangioblast.

To determine if Flk1-expressing cells also increased in response to bFGF, ES cells differentiated with or without bFGF were stained with monoclonal antibody against Flk1. As shown in Fig. 5B, the percentage of Flk1⁺ cells in EBs increased in the presence of bFGF. As with the gradual increase in blast colony number with increasing bFGF concentrations, the number of Flk1⁺ cells also increased in a bFGF dose-dependent manner (not shown). Together, our results suggested that bFGF not only increased the generation of BL-CFCs, but also Flk1⁺ cells.

To examine if we can detect the increase of Flk1⁺ cells at earlier time points than day 2.75, Flk1^+/− EB cells were differentiated with or without bFGF and subjected to staining for β-galactosidase. Flk1^+/− ES cells contain the lacZ gene under the transcriptional control of endogenous Flk1 regulatory elements, thus the β-galactosidase (β-gal, the lacZ gene product) activity acts as a histochemical marker of Flk1 expression (Shalaby et al., 1995). As shown in Table 1, the increase in Flk1⁺ cells by bFGF was observed as early as day 1.5 and then throughout EB development (up to day 3.5, the endpoint of the analysis). The increase in Flk1⁺ cells measured by β-galactosidase activity or by Flk1 monoclonal antibody staining was similar, as shown in Table 1.

To better elucidate the role of bFGF on hemangioblast development, we determined if Flk1⁺ cells expressed fgfr1. Thus, we flow sorted Flk1⁺ cells from day-2.75 EBs and picked individual Flk1⁺ cells into cell lysis buffer by micromanipulation. Single cell samples were then subjected to global amplification of mRNA transcripts, known as single-cell based polymerase chain reaction (PCR) analysis (Brady et al., 1990). The amplified PCR product from single Flk1⁺ cells was analyzed for the expression of Flk1 and fgfr1. As shown in Fig 6, cells sorted for the surface marker Flk1 all expressed the Flk1 gene. Furthermore, a majority of these cells (20/26) also expressed fgfr1, although the levels of fgfr1 expression were variable between samples.

The findings that the increase of Flk1⁺ cells can be detected as early as 1.5 days after the initiation of EB differentiation and that a majority of Flk1⁺ cells expressed fgfr1 suggested that bFGF conferred a proliferation signal on Flk1⁺ cells. To further confirm this notion, we next determined if the BL-CFC frequency remains the same among Flk1⁺ cells that were generated with or without bFGF. Thus, we flow sorted Flk1⁺ cells from day-2.75 EBs, differentiated with or without bFGF, and replated them for blast colonies. If BL-CFC differentiation among Flk1⁺ cells could be induced by bFGF, it was expected that the BL-CFC frequency would be higher from Flk1⁺ cells isolated from EBs differentiated with bFGF. However, if bFGF conferred overall proliferation of Flk1⁺ cells, the BL-CFC frequency would remain similar. As shown in Fig. 7A and B, the number of blast colonies developed from Flk1⁺ cells was similar whether the Flk1⁺ cells were derived with or without bFGF. Together, our data argue for the proliferative role of bFGF on Flk1⁺ cells.

To investigate if bFGF also plays a role in blast colony formation, we replated sorted Flk1⁺ cells with VEGF, bFGF, or VEGF and bFGF. As shown in Fig. 7C, the number of blast colonies developed was less when Flk1⁺ cells were replated with bFGF. The colony size was much smaller and blast cells did not look healthy in the presence of bFGF. Furthermore, bFGF did not show any additive or synergistic effect with VEGF as the number of blast colonies was similar whether the Flk1⁺ cells were replated in VEGF alone or VEGF and bFGF together. These results indicated that bFGF is important for BL-CFC expansion within EBs but is not critical for the subsequent formation of blast colonies from BL-CFCs in methylcellulose culture.

We have previously demonstrated that both Flk1 and scl are expressed in blast cells (Kennedy et al., 1997). In EBs, Flk1 gene expression can be detected as early as day 2 and scl at day 2.5 (Fig. 2). SCL is a transcription factor, which contains a basic-helix-loop-helix (bHLH) domain (Begley et al., 1989) and functional SCL has been shown to be required for all hematopoietic lineages to develop (Shivdasani et al., 1995; Robb et al., 1995; Porcher et al., 1996; Robb et al., 1996). To further define hemangioblast development, we determined if SCL function is required for BL-CFC/hemangioblast development. Thus, scl^−/− ES cells were analyzed for their potential to give rise to BL-CFCs and Flk1⁺ cells. While day-2.75 wild-type EBs gave rise to blast colonies when replated in response to VEGF, scl^−/− EBs did not generate any blast colonies. However, blast-like colonies, which are characterized by their tighter colony morphology compared to blast colonies, yet looser than EBs, and hence distinct from EBs and blast colonies, were readily detectable when scl^−/− EBs were replated (Fig. 8A-C). These blast-like colonies were also present in wild-type culture, although at a reduced frequency compared to the frequency of BL-CFCs (not shown). When the blast-like colonies were transferred to medium containing both hematopoietic and endothelial factors, wild-type blast-like colonies gave rise to both hematopoietic and endothelial cells, with a more vigorous growth in the endothelial component (not shown). However, scl^−/− blast-like colonies gave rise only to endothelial cells (47 out of 47 colonies analyzed). Since these blast-like colonies developed in cultures containing VEGF, it was expected that progenitors giving rise to blast-like colonies expressed Flk1 and thus Flk1-expressing cells still developed in scl^−/− EBs. Therefore, we analyzed scl^+/+ and scl^−/− EB cells for the presence of Flk1⁺ cells. As shown in Fig. 8D, scl^−/− EBs contained Flk1⁺ cells, confirming that Flk1⁺ cells develop in the absence of functional SCL. More importantly, when scl^−/− ES cells were differentiated in the presence of bFGF, the percentage of Flk1⁺ cells increased, as is also seen with wild-type ES cells (Fig. 8D).

As scl^−/− EBs contained a somewhat higher percentage of Flk1⁺ cells compared to the wild type EBs (Fig. 8D), we determined if the increase in Flk1⁺ cells within scl^−/− EBs reflected the increase in endothelial cell generation in the absence of SCL. Thus, scl^−/− EB cells were subjected to FACS analysis for endothelial markers. As shown in Fig. 9 and Table 2, cells expressing Flk1, PECAM-1, CD34, P-selectin, or VE-cadherin were present at higher levels within scl^−/− EBs compared to controls throughout EB differentiation (up to day 12). Together, our results suggest that blast-like colonies represent a transient Flk1⁺ population or pre-BL-CFCs, from mesoderm to BL-CFCs and that functional SCL is required for the formation of blast colonies from Flk1⁺ cells. Furthermore, our results suggest that Flk1⁺ cells readily give rise to endothelial cells in the absence of SCL.

We have previously demonstrated that BL-CFCs represent the long pursued common progenitor of hematopoietic and endothelial cells, the hemangioblast. As BL-CFCs form blast colonies in the presence of VEGF, a ligand for Flk1, we investigated if Flk1 is expressed on BL-CFCs. By sorting a cell population expressing the receptor tyrosine kinase Flk1 and replating for the generation of blast colonies, we have demonstrated that BL-CFCs indeed express Flk1. The observations that CD45⁺c-Kit⁺ hematopoietic cells develop from Flk1⁺VE-cadherin⁺CD45⁻ cells obtained from in vitro differentiated ES cells or yolk sac cells (Nishikawa et al., 1998) and that CD45⁺ hematopoietic cells develop from VEGF-R2⁺ (Flk1⁺) cells of the chick intra-aortic region (Jaffredo et al., 1998) are consistent with our notion that the hemangioblast express Flk1.

By measuring the frequency of the BL-CFC and Flk1⁺ cells as experimental readouts, we have assessed the capacity of various mesoderm-inducing factors to promote hematopoietic commitment. Our studies suggest that the generation of BL-CFCs and Flk1⁺ cells is regulated not by a single factor, but by several factors. Of these, bFGF-mediated signals are critical for hematopoietic development. The effect of Activin A was greater at lower concentrations. More importantly, bFGF together with activin A appear to have an additive or synergistic effect on BL-CFC generation and that activin A-mediated BL-CFC generation requires a Fgfr1 signal since the administration of activin A did not rescue the BL-CFC development in fgfr1^−/− EBs (Fig. 4E). These results are consistent with the findings in Xenopus that activin A and bFGF synergize in mesoderm induction and that activin A requires a bFGF-mediated signal (Cornell and Kimelman, 1994; LaBonne and Whitman, 1994). Our studies are also consistent with the recent study by Pardanaud and Dieterlen-Lievre (1999) who found that the treatment of somatopleural mesoderm, which otherwise contains angiopoietic potential only, with VEGF, bFGF or TGF-β1 could induce hemangiopoietic potential in this tissue in the quail/chick chimera model.

Previous studies have demonstrated that fgfr1^−/− embryos show growth arrest at an early stage of gestation and that fgfr1^−/− embryos die early in gestation and show abnormal mesodermal patterning. However, as fgfr1^−/− ES cells could differentiate into many different mesodermal cell types, including muscle cells, when teratomas were generated through subcutaneous injection into nude mice, mesodermal defects observed in fgfr1^−/− embryos are probably due to abnormal mesodermal patterning, rather than an intrinsic defect of mesoderm differentiation (Deng et al., 1994). Our in vitro analyses of fgfr1^−/− ES cells for their potential to differentiate and to give rise to hematopoietic progenitors are not only consistent with the in vivo findings but further reveal a role for bFGF on hemangioblast development. One, fgfr1^−/− ES cells differentiate poorly in vitro judged by the extended expression of the rex1 gene (Fig. 2) and the reduction in total EB cell numbers throughout EB differentiation. Two, the expression of the brachyury (T) gene, a marker for mesoderm, could be detected in fgfr1^−/− EBs, consistent with the in vivo findings (Deng et al., 1994; Yamaguchi et al., 1994). In wild-type EBs, T gene expression is temporally up-regulated in the early stage of EB differentiation and down-regulated in the later stage of EB differentiation (Fig. 2). As T gene expression was extended and still observed in day-10 fgfr1^−/− EBs, with rex1 being still expressed, unlike the control, it is possible that mesoderm differentiation is delayed in the absence of Fgfr1 or that cells that manage to differentiate to mesoderm are delayed in subsequent differentiation. Three, in vitro differentiated fgfr1^−/− EBs expressed, although at low levels, genes of both muscle and hematopoietic cells. Considering the fact that myf5 was up-regulated and that the expression of α-cardiac myosin heavy chain and other hematopoietic genes is greatly down-regulated, it is tempting to speculate that the major role of Fgfr1 signaling, at least in the in vitro differentiation system, is the proliferation of mesodermal progenitors once they are generated. Our data that a majority of Flk1⁺ cells expressed fgfr1 and expanded in response to bFGF, but that Flk1⁺ cells from day-2.75 EBs did not respond to bFGF in forming blast colonies further support this notion. Clearly VEGF is critical for blast colony formation from BL-CFCs (Choi et al., 1998; Schuh et al., 1999). However, since the addition of VEGF to EB differentiation does not increase the frequency of the BL-CFC or Flk1⁺ cells (not shown), we propose two distinct regulations on hemangioblast development, namely bFGF-mediated hemangioblast proliferation and VEGF-mediated hemangioblast migration and maturation to hematopoietic and endothelial cells (Shalaby et al., 1997; Choi et al., 1998; Schuh et al., 1999).

In zebrafish, scl is expressed in the lateral mesoderm, which gives rise to hematopoietic, endothelial, and pronephric lineages (Gering et al., 1998). Several studies might be interpreted to suggest that SCL/Tal-1 lies upstream of Flk1 in zebrafish. First, the zebrafish cloche (clo) mutation affects both hematopoietic and endothelial differentiation (Stainier et al., 1995; Liao et al., 1998). Furthermore, ectopically expressed SCL partially rescues hematopoietic and endothelial cell defects of clo mutants, as evidenced by the development of red cells, expression of hemoglobin, red cell markers (i.e. GATA-1) and endothelial markers (Flk1 and tie-2; Liao et al., 1998). These data suggest that SCL lies downstream of clo, yet upstream of Flk1 and GATA-1. However, many studies do not support the positioning of SCL/Tal-1 upstream of Flk1. First, we have demonstrated that expression of the Flk1 gene is detected prior to that of scl/tal-1 (day 2 versus day 2.5; Fig. 5). Similarly, while Flk1 gene expression in mouse embryos is seen in presumptive mesodermal yolk sac blood island progenitors as early as 7.0 dpc (Yamaguchi et al., 1993; Dumont et al., 1995), scl/tal-1 gene expression can be detected in embryonic and extra-embryonic mesoderm of 7.5 dpc embryos and in blood islands of the yolk sac of 8.5 dpc embryos (Kallianpur et al., 1994; Silver and Palis, 1997). Palis et al. (1999) also showed that Flk1 gene expression precedes that of scl in the developing embryo. Second, gene targeting experiments demonstrate that Flk1^−/− embryos die early in gestation (between 8.5-9.5) and the phenotype reflects the expression pattern of Flk1 such that Flk1^−/− embryos show defective blood islands, with both hematopoietic and endothelial cells being affected (Shalaby et al., 1995). However, scl^−/− embryos die between 8.5-10.5 days with the major defect being the absence of yolk sac hematopoiesis. In these mutant mice, endothelial cells are present (Robb et al., 1995; Shivdasani et al., 1995). Consistent with this, our studies demonstrate that in vitro differentiated scl^−/− EBs contained cells with endothelial cell markers. Since cells with endothelial cell markers are present at elevated levels in scl^−/− EBs, it is tempting to speculate that the fate of hemangioblast in the absence of functional SCL is skewed toward endothelial cell differentiation (Fig. 9 and Table 2). Third, we have demonstrated that Flk1⁺ cells develop in the absence of SCL/Tal1 and that scl^−/− EB cells respond appropriately to bFGF and give rise to a higher percentage of Flk1⁺ cells, compared to control EBs (Fig. 7D). Consistent with this, Flk1 expression was not impaired in scl^−/− EBs (Elefanty et al., 1997). Similarly, scl expression is unchanged in Flk1^−/− EBs compared to Flk1^+/+ EBs (unpublished data). Therefore, these studies suggest that the regulation of Flk1 and scl gene expression is independent, but developmentally coordinate.

The BL-CFC frequency is much lower (0.5-2%, Figure 3 and 4) compared to that of Flk1⁺ cells (15-25%, Fig. 5B and Table 1) in day-2.75 EBs. In addition, percentage change in Flk1⁺ (Fig. 5B and Table 1) cells in response to bFGF does not precisely correlate with the BL-CFC frequency (Figs 3, 4). Furthermore, only a fraction of Flk1⁺ cells give rise to blast colonies when replated (Figs 5A, 7). An explanation for these findings is that the Flk1⁺ cell population is heterogeneous and that only a fraction of Flk1⁺ cells represent true hemangioblasts, although we cannot rule out the possibility that the BL-CFC frequency is underestimated due to sub-optimal culture conditions. Based on the findings that blast colonies are absent, and yet blast-like colonies still develop from scl^−/− EBs, we postulate that Flk1⁺SCL⁺ cells represent hemangioblasts or a subset thereof (Fig. 10). According to this model, Flk1 single positive cells represent a transitional pre-BL-CFC cell population from mesoderm to BL-CFCs. Further identification of additional surface markers expressed on hemangioblasts, fractionation of Flk1⁺ cells based on these markers, and studies of their in vitro and in vivo developmental potential should be critical for characterizing Flk1⁺ cells and their contribution to hematopoietic and vascular establishment in the embryo.

We would like to thank A. Parlow at NIDDK’s National Hormone & Pituitary Program for Activin A, Hans Buhring at Eberhard-Karls University (Germany) for the Flk1 monoclonal antibody, and Andre Schuh for Flk1⁺/⁻ ES cells. We thank Kathleen Sheehan and Chris Nelson for help with purifying the Flk1 monoclonal antibody and Scott Robertson and Charles Wall for helpful discussion on single cell PCR. We thank Gordon Keller for helpful discussion and the L32 DNA probe used for single cell PCR analysis, and Mitchell Weiss for critically reading the manuscript. This work was partially supported by a grant from the National Institutes of Health, R29HL55337 (to K. C.).