Generation of a Primitive Erythroid Cell Line and Promotion of Its Growth by Basic Fibroblast Growth Factor

HEMATOPOIESIS IS ESTABLISHED very early during mammalian embryogenesis. In mice, the first mature hematopoietic cells can be detected in extraembryonic blood islands of the yolk sac as early as day 7.5 of gestation.1,2 Mature blood cells produced at this stage are large, nucleated erythroids, known as primitive erythroid cells (EryP). These EryP cells produce embryonic forms of globins. Although EryP cells are the predominant mature lineage in the yolk sac, other hematopoietic progenitors can also be detected.1-7 As the hematopoietic site is shifted to the fetal liver and then to the bone marrow, newly produced erythroid cells become smaller and nonnucleated (definitive erythroid [EryD]). The shift from primitive to definitive erythropoiesis is accompanied by the switching of embryonic globin gene expression to adult forms.8 9 

Several studies suggest that regulation of primitive and definitive erythropoiesis is distinct. White spotting mutant (W) mice carry natural mutations in the receptor tyrosine kinase c-kit gene and are characterized by defects in melanocyte, germ cell, and hematopoietic development.10 The severity of the mutant phenotype coincides with the degree of impairment of Kit kinase activity. The most severe forms of mutations are W and W^19H, which result in a complete lack of Kit kinase activity. Mice homozygous for the W or W^19H mutations are not viable. They die either perinatally or in later stages of embryogenesis. However, yolk sac hematopoiesis in these mice appears to be normal, suggesting that Kit is not required for EryP development.1 Similarly, gene knock-out experiments demonstrate that the c-myb, erythropoietin (Epo), or erythropoietin receptor (EpoR) genes11-13 are essential for EryD development. While these studies provide some insights into EryD development, the mechanisms regulating primitive erythropoiesis are not as well established.

EryP and endothelial cells constituting blood islands of the yolk sac are the first mature progeny of mesoderm in developing mouse embryos. Fibroblast growth factors (FGFs) and factors that belong to transforming growth factor (TGF)-β group play a central role in mesoderm development in Xenopus and Drosophila.14-16 Mice carrying homozygous null mutations at the fgfr-1 locus die in utero due to the failure of embryonic cell proliferation and abnormal mesodermal patterning during gastrulation.17,18 Similarly, TGFβ-1^−/− mice die in utero due to defective extraembryonic hematopoiesis and endothelial differentiation.19 Even though these studies do not directly demonstrate the role of FGFs and TGF-β family growth factors in the establishment of the hematopoietic system, they may well be involved in the onset of hematopoietic development.

Here, we report the generation of a EryP cell line (EB-PE) using in vitro–differentiated progeny (embryoid bodies [EBs]) of embryonic stem (ES) cells20-22 and retroviral insertional mutation.23 24 Even though immortalized, they still require growth factors for their proliferation and differentiation. Among factors tested, basic (b) FGF and Epo show unique roles in EB-PE proliferation and differentiation depending on culture conditions. Furthermore, we demonstrate that fgfr-1 is the corresponding receptor for bFGF. Gene expression analysis indicates that primary EryP progenitors from EBs also express the fgfr-1 gene and that bFGF also stimulates these cells. These results demonstrate that bFGF is important in early hematopoietic lineage development, and further suggests for its possible role in the establishment of hematopoiesis.

CCE ES cells, obtained from Dr E. Robertson (Harvard University), were maintained on feeder cells, STO fibroblasts, in Dulbecco's modified Eagle's medium (DMEM) supplemented with fetal calf serum ([FCS] Gemini, Calabasas, CA; 15%), monothioglycerol ([MTG] Sigma, St Louis, MO; 1.5 × 10⁻⁴ mol/L), and leukemia inhibitory factor ([LIF] Genetics Institute, Boston, MA; 1.5% conditioned medium). ES cells were differentiated as described.20-22 Briefly, 2 days before initiating differentiation, ES cells were passaged in Iscove's modified Dulbecco's medium (IMDM) supplemented with FCS (15%), MTG (1.5 × 10⁻⁴ mol/L), and LIF (1.5% conditioned medium) without STO cells. For differentiation, cells were dissociated with trypsin, washed once, and plated in IMDM with FCS (15%; Hyclone, Logan, UT), l-glutamine (2 mmol/L), ascorbic acid (50 ng/mL; Sigma), and MTG (4.5 × 10⁻⁴mol/L). Cells were plated in a final volume of 10 mL at 5,000 cells/mL (day 2.5 to 3.5 EBs) or 3,000 to 4,000 cells/mL (day 3.5 to 5 EBs) in 100-mm bacterial-grade dishes. These cultures were maintained in a humidified chamber at 37°C containing 5% CO₂.

EB cells were replated as described.20-22 Briefly, EBs were collected at different time points, dissociated with trypsin, and replated into methylcellulose cultures containing plasma-derived serum ([PDS] Antech, Tyler, TX; 10%), ascorbic acid (12.5 ng/mL), l-glutamine (2 mmol/L), transferrin (300 μg/mL; Boehringer, Indianapolis, IN), protein-free hybridoma media ([PFHM2] GIBCO-BRL, Gaithersburg, MD; 5%), and MTG (4.5 × 10⁻⁴ mol/L) with factors. bFGF was used at 30 pg to 1 ng/mL and Epo was used at 2 U/mL. One-milliliter quantities of methylcellulose cultures containing 3 × 10⁴cells were put in 30-mm bacterial-grade petridishes. EryP colonies were counted 4 to 6 days after replating.

EB-PE cells were maintained in IMDM containing 10% FCS with various factor combinations at the following concentrations: bFGF (10 ng/mL), c-kit ligand ([KL] 1% conditioned medium), interleukin (IL)-3 (1% conditioned medium), insulin-like growth-factor (IGF)-1 (10 ng/mL), IL-11 (10 ng/mL), IL-6 (10 ng/mL), and Epo (2 U/mL). For colony assay, EB-PE cells were washed, replated into methylcellulose cultures containing PDS (10%), ascorbic acid (12.5 ng/mL), l-glutamine (2 mmol/L), transferrin (300 μg/mL), PFHM2 (5%), and MTG (4.5 × 10⁻⁴ mol/L) with factors. The factors used were Kit ligand (10 ng/mL), IL-3 (10 ng/mL), granulocyte/macrophage colony-stimulating factor ([GM-CSF] 5 ng/mL), M-CSF (5 ng/mL), Epo (2 U/mL), vascular endothelial growth factor ([VEGF] 5 ng/mL), bFGF (30 pg to 30 ng/mL), IGF-1 (10 ng/mL), and activin A (0.1 to 10 ng/mL). Colonies were counted 4 to 6 days after replating.

Plasmid DNA provided by Dr A. Gudkov (University of Illinois at Chicago) was transfected into an ecotropic retroviral packaging cell line, BOSC 23,25 and the virus released was collected 24 hours after transfection and used for infection. Day 4 EBs were dissociated with trypsin and were infected with retroviruses in the presence of polybrene (5 μg/mL), VEGF (5 ng/mL), IGF-1 (10 ng/mL), and Epo (2 U/mL). After infection, cells were selected in methylcellulose culture containing VEGF (5 ng/mL), IGF-1 (10 ng/mL), Epo (2 U/mL), and G418 (500 μg/mL). The resulting colonies were picked and transferred to liquid culture after 7 to 10 days.

Southern, Northern, and reverse-transcription polymerase chain reaction (RT-PCR) analyses were performed as described.26-28 EryP colonies were obtained by replating day 4 EB cells into methylcellulose cultures with Epo. The resulting EryP colonies were pooled and RNA was extracted. Specific primers used for RT-PCR are as follows^20,29-32: β-actin: sense 5′ATGAAGATCCTGACCGAGCG3′, antisense, 5′TACTTGCGCTCAGGAGGAGC3′; β-H1: sense 5′AGTCCCCATGGAGTCAAAGA3′, antisense 5′CTCAAGGAGACCTTTGCTCA3′; SCL: sense 5′ATTGCACACACGGGATTCTG3′, antisense 5′CATACAGTACGACACTGACG3′; GATA-1: sense 5′ATGCCTGTAATCCCAGCACT3′, antisense 5′TCATGGTGGTAGCTGGTAGC3′; c-kit: sense 5′TGTCTCTCCAGTTTCCCTGC3′, antisense 5′TTCAGGGACTCATGGGCTCA3′;fgfr-1: sense 5′AGCCTGACCACCGAATTGGAG3′, antisense 5′CATCAACTCCACATTGCTGC3′; flk-1: sense 5′CACCTGGCACTCTCCAC3′, antisense 5′ATTTCATCCCACTACCG3′; flt-1: sense 5′CTCTGATGGTGATCGTGG3′, antisense 5′CATGCGTCTGGCCACTTG3′; and EpoR: sense 5′GGACACCTACTTGGTATTGG3′, antisense 5′GACGTTGTAGGCTGGAGTCC3′.

bFGF and Epo were purchased from Upstate Biotechnology (Lake Placid, NY) and Amgen (Thousand Oaks, CA), respectively. IL-11, IL-6, IGF-1, GM-CSF, M-CSF, and VEGF were purchased from R&D Systems. LIF was obtained from medium conditioned by CHO cells transfected with a LIF expression vector (kindly provided by Genetics Institute). KL was obtained either from medium conditioned by CHO cells transfected with a KL expression vectors (Genetics Institute) or from R&D Systems. IL-3 was obtained either from R&D Systems or from medium conditioned by X63 Ag8-653 myeloma cells transfected with a vector expressing IL-3.33 Recombinant human Activin A was kindly provided by National Hormone and Pituitary Program (NHPP), National Institute of Digestive & Kidney Diseases.

In an effort to understand how primitive hematopoietic development is regulated, we used retroviruses to mark individual EB cells to monitor their cell fate. Day 4 EBs were used since they contain a large number of hematopoietic progenitors.20,22 The retroviruses used in this study carry random cDNA fragments (200 to 400 bp), which serve as a unique tag and a neomycin-resistance gene for selection (Fig1A).³⁴ A majority of the cDNAs in these retroviruses were shown to be biologically inactive.34 To optimally target early hematopoietic progenitors, factors known to support immature progenitors such as VEGF, IGF-1, and Epo22 were included during retroviral infection and G418 selection. After G418 selection, several colonies developed in methylcellulose cultures, and when transferred to liquid culture media with the same growth factors, cells from one colony grew continuously. We did not obtain any continuously growing cells without retroviral infection. Southern blot analysis of the newly established line with multiple enzymes that cut only once within the retroviral genome indicated that a single retroviral genome was present within these cells (Fig 1B). To determine if the piece of cDNA present within the retroviral genome contributed to the immortalization, cDNA within the retroviral genome was amplified by PCR and sequenced. The cDNA sequence, which was oriented in an antisense configuration within the retroviral genome, showed 100% homology to mouse 18S rRNA when compared with those in the available data bases (Fig 1A). Since inhibition of 18S rRNA function should lead to the general inhibition of protein synthesis, it is highly unlikely that the rRNA gene fragment contributed to the immortalization phenotype. This result is consistent with the notion that a cellular gene near the retroviral integration has been activated and that this newly activated protein most likely contributed to the cell immortalization.

May-Grünwald/Giemsa staining of the immortalized cells demonstrated that they contained large nuclei with little cytoplasmic content (Fig 2A). As shown in Fig 2B, they expressed both ζ and βH1 globin genes (lanes 1 and 3). A low level of β-major globin gene expression (lane 2) was also observed. The observed cell morphology and gene-expression pattern suggested that they were EryP cells. To further investigate whether these immortalized cells were of erythroid origin, we tested several growth factors in semisolid media (methylcellulose cultures). As shown in Fig 2C, no colonies developed in the absence of growth factors, indicating that they were growth factor–dependent. Further, colonies developed only in cultures containing Epo, but not in IL-3, GM-CSF, or M-CSF. Two types of colonies were apparent: one that consisted of 20 to 40 cells and the other of 100 to 300 cells. Cells within both types of colonies showed strong hemoglobinization (Fig 2D). This observation was consistent with the idea that cells differentiated spontaneously in culture. Therefore, cells forming larger colonies would represent more immature progenitors. We have designated the immortalized cell line as EB-PE (embryoid body–derived primitive erythroid cells).

To further characterize EB-PE cells, the expression of several genes known to be expressed in erythroid cells,35 such as the transcription factors SCL, GATA-1, and a receptor tyrosine kinase, c-kit, was compared with that of primary EryP, obtained from EB cells, using semiquantitative RT-PCR. As shown in Fig3, βH1, SCL, and GATA-1 were expressed in both cell populations, although the levels of βH1 and SCL gene expression were much higher in primary EryP cells. c-kitexpression was barely detectable in EB-PE cells compared with that of EryP cells. Primary EryP cells obtained from day 4 EBs still expressed the fgfr-1 gene at low levels, even though they did not respond to bFGF at this stage (see below). Both EB-PE and primary EryP cells expressed flk-1 and flt-1, genes shown to be expressed in erythroid cells.36 These results indicated that EB-PE cells expressed genes characteristic of erythroid cells.

To determine how the EB-PE proliferation is regulated, we first analyzed their growth factor requirements. Factors known to act on early hematopoietic progenitors, such as activin A, bFGF, VEGF, IL-3, IL-6, IL-11, IGF-1, KL, and Epo were tested.22 37 The proliferation of EB-PE cells was determined either in liquid or methylcellulose cultures. As shown in Fig4A, EB-PE cells failed to grow and subsequently died in the absence of any growth factors. Activin A, bFGF, VEGF, IL-3, IL-6, IL-11, IGF-1, and KL all failed to support EB-PE cell proliferation when tested as the only exogenously provided growth factor. This suggested that Epo was absolutely required. Consistent with this notion, EB-PE cells proliferated in cultures containing Epo alone. Among factors tested in combination with Epo, only bFGF was able to support EB-PE proliferation significantly better than Epo alone (Fig 4A, data not shown). The response to bFGF was dose-dependent such that bFGF was growth stimulatory at levels as low as 30 pg/mL, even though cells proliferated better at higher bFGF concentrations (Fig 4B). From the growth curve, it appeared that EB-PE cells divided every 8 hours in the optimum growth-stimulatory conditions (bFGF at 10 ng/mL). The bFGF mitogenic effect on EB-PE was unique, such that cells failed to respond to other mitogenic signals, including Epo, when they were maintained in bFGF and Epo. Consistent with this, cells failed to form large colonies in methylcellulose cultures (Fig 4C). Large colonies that developed in the presence of bFGF and Epo (Fig 4C) were less hemoglobinized, suggesting that cells remained undifferentiated. To confirm if this was the case, cells grown in the presence or absence of bFGF were subjected to benzidine staining, an assay for hemoglobinization. As shown in Fig5, a higher portion of cells (∼60%) stained positive when cells were maintained in cultures in the absence of bFGF, while a smaller portion of cells (∼10% to 15%) were benzidine-positive in the presence of bFGF. When cells were transferred from bFGF/Epo to Epo alone, they started to differentiate such that approximately 40% of cells (after 24 hours) and approximately 80% of cells (after 48 hours) became benzidine-positive and subsequently died. These observations are consistent with the notion that while bFGF is a strong mitogen for EB-PE, Epo is required for EB-PE terminal differentiation.

Once we identified bFGF to be a growth factor for EB-PE cells, we tested whether bFGF is also a growth factor for primary EryP cells. Our initial experiments to determine if bFGF has a mitogenic effect on EryP precursors were performed using day 4 EB cells, as they contain a large number of EryP progenitors. bFGF was used at 10 to 30 ng/mL for initial studies, since the highest growth-stimulatory effect on EB-PE cells was observed at these concentrations. When day 4 EB cells were tested in a methylcellulose assay, the number of EryP colonies was similar in cultures containing Epo or bFGF and Epo (Fig6, data not shown). Since it was possible that progenitors from later EBs contained more differentiated EryP progenitors that might not respond to bFGF growth stimulation, EB cells of an earlier stage of development were examined. However, the replating of early EBs with bFGF at 10 to 30 ng/mL resulted in increased secondary EB formation. Since bFGF was mitogenic for EB-PE cells, even at 30 pg/mL, and the number of secondary EBs was not increased from earlier EBs at these low bFGF concentrations (Fig 4B, data not shown), lower bFGF concentrations were tested to determine if bFGF stimulates primary EryP progenitor proliferation. One representative result obtained from replating EBs at different stages, with bFGF at various concentrations, on EryP progenitors is shown in Fig 6A and the relative increase of EryP numbers from three experiments is shown in Fig 6B. As shown, the number of EryP colonies was consistently higher in cultures containing bFGF at 30 to 100 pg/mL compared with cultures containing Epo only when early EB cells (up to day 3 to 3.5) were tested. The mitogenic effect of bFGF on EryP progenitors was heparin-dependent, since no enhancement in colony formation was observed in the absence of heparin (data not shown). The bFGF growth-stimulatory effect, at low concentrations (30 to 100 pg/mL), on EryP cells was less obvious when later EB (>day 3.5) cells were tested. Further, the effect of bFGF at higher concentrations (1 ng/mL) on EryP cells was not as consistent as that of lower concentrations (30 to 100 pg/mL), regardless of EB ages tested (Fig 6B, compare day 2.875 v 3). Taken together, it appeared that bFGF was growth stimulatory for early-stage EryP progenitors only at low concentrations and in the presence of heparin.

There are four different isoforms of FGF receptors currently known.38 To determine which of these is responsible for bFGF signaling, Northern blot analysis, as well as semiquantitative RT-PCR, was performed on RNA obtained from EB-PE cells grown in the presence or absence of bFGF. As shown in Fig7A and B, fgfr-1 was expressed at a low level when cells were grown in the absence of bFGF and was expressed at a much higher level when cells were grown in bFGF. This result indicated that fgfr-1 gene expression was induced by culturing cells with bFGF. flk-1 expression was also upregulated, while EpoR expression was downregulated in cells cultured with bFGF and Epo. The expression level of fgfr-2 was extremely low (can be detected only after a 5-day exposure) and the expression offgfr-3 and fgfr-4 was undetectable in EB-PE cells (data not shown). It is generally speculated that bFGF can upregulate its own expression in an autocrine manner in cells cultured with bFGF. However, bFGF gene expression in EB-PE cells grown in bFGF and Epo was not detectable, suggesting that bFGF worked as a paracrine growth factor for EB-PE cells (data not shown). Taken together, these data indicated that fgfr-1 was the corresponding receptor for bFGF and thatfgfr-1, flk-1, and EpoR gene expression could be regulated depending on culture conditions.

The study of EryP development has been limited by the transient nature of EryP cells and the lack of a representative cell line for ex vivo studies. We have generated an immortalized EryP cell line, EB-PE. The EB-PE cell line is important for the following reasons. First, these cells maintain EryP cell characteristics. They require growth factor(s) for survival and/or proliferation, yet can undergo differentiation. Therefore, these cells should be useful to investigate cellular and molecular events regulating EryP lineage development. Second, the generation of a EryP cell line from differentiated ES cells emphasizes the utility of the ES model system. In fact, EB-PE is, to our knowledge, the first EryP line to be generated. The manipulation of this in vitro ES system may enable us to access other primitive hematopoietic cells.

We demonstrated that Epo was absolutely required for EB-PE proliferation and differentiation and that bFGF is a mitogenic factor for EB-PE cells. Further, cells remained undifferentiated when bFGF was present in the culture medium. The response of EB-PE to a bFGF mitogenic signal correlated with the upregulation of fgfr-1 and the downregulation of EpoR gene expression. It is possible that the level of EpoR gene expression is directly related to the state of the cell, ie, proliferation versus differentiation. Further studies are required to resolve this issue.

Several other studies have demonstrated that bFGF is a hematopoietic growth factor.39-43 bFGF has been shown to directly stimulate myeloid progenitors when either low-density adherence-depleted bone marrow mononuclear cells or CD34⁺cell population were tested in a methylcellulose colony assay.40 The direct mitogenic effect has also been observed on the growth of megakaryocyte progenitors.41,42 Our finding that bFGF is a potent, novel growth factor for EryP cells suggests that bFGF may have a broader role in hematopoiesis. The fact that EryP are the first mature blood cells to develop further argues that bFGF may even have a role in the establishment of hematopoiesis. Contrary to our observation that bFGF is a mitogen for EryP progenitors, studies by Beradi et al,40 using bone marrow–derived cells, did not show any growth-stimulatory effect of bFGF on erythroid cells. Given that bone marrow cells contain only EryD progenitors, it is difficult to compare our results with theirs. However, it is possible that bFGF confers a mitogenic effect only on early-stage EryP progenitors. In fact, the ability to respond to a bFGF mitogenic effect could reflect the different regulation mechanism(s) involved in primitive versus definitive erythropoiesis.

The bFGF mitogenic effect on primary EryP progenitors appears to be dependent on concentration and differentiation stage of the progenitor, as the bFGF mitogenic effect was observed only when early EBs were replated (<day 3 to 3.5) at lower bFGF concentrations (30 to 100 pg/mL). Further, heparin was required for the bFGF mitogenic effect on primary EryP cells, while the response of EB-PE cells to bFGF was independent of heparin. Since heparin, upon binding to bFGF, augments the fgfr-1 activation, it is possible that cells expressing high levels of fgfr-1 may not require heparin. It is not clear why bFGF at higher concentrations failed to confer a mitogenic signal for early EryP progenitors, especially since the bFGF mitogenic effect on EB-PE cells was dose-dependent (Fig 4B). Factors including activin A and TGF-β1 exhibit their mitogenic effects in a concentration-dependent manner.44 45 Therefore, it is possible that bFGF also exerts its mitogenic effect in a similar manner. Further studies are required to resolve this issue.

Gene expression analysis indicated that βH1, SCL, and c-kitexpression is much higher in primary EryP cells. The difference in βH1 and c-kit gene expression levels in these two cell populations could reflect immortalized versus primary cells. Alternatively, this gene expression difference could be due to culture conditions. While EB-PE cells were maintained in FCS, primary EryP cells were generated with PDS. Both fgfr-1 and flk-1gene expression were induced when EB-PE cells were grown in bFGF. Thefgfr-1 gene induction clearly conferred a further mitogenic stimulus on EB-PE cells. However, the significance of the flk-1induction by bFGF in EB-PE cells is less clear, since VEGF did not confer any growth stimulus on these cells. The similar observation thatflk-1 gene expression is induced by bFGF was also made by Flamme et al.46 In this study, flk-1 expression was induced within 24 hours when bFGF was added to the in vitro quail blastodiscs culture system, which gives rise to blood islands and vascular structure formation. Further studies are required to determine whether VEGF/Flk-1/Flt-1 interaction is necessary for primitive erythropoiesis.

The finding that bFGF is a growth factor for EB-PE and thatfgfr-1 expression is induced when cells were grown in bFGF and Epo is intriguing, since bFGF has been shown to be a growth factor for a number of different leukemic cells.47,48 Receptors for various isoforms of FGFs have also been shown to be expressed in many different leukemic cells.47 48 Based on our observation, it will be interesting to see whether bFGF/Fgfr-1 interaction plays a role in erythroleukemic cell proliferation. Our data indicate that a single retroviral genome is present in EB-PE cells. Since the retroviral genome does not carry an oncogene, it is speculated that the EB-PE immortalization was a result of the deregulation of the expression of a cellular gene. We are currently characterizing the insertion site to investigate the nature of the interrupted cellular sequence.

We are grateful to G. Keller, G. Longmore, and D.M. Pardoll for helpful discussions and advice throughout this work. We thank P. Faloon, A. Gudkov, D. Kalvakolanu, A. Kazarov, and M. Shin for critically reading the manuscript. We thank A. Kazarov for help with figure preparation. We also thank D.H. Fremont for encouragement throughout this work.