ABSTRACT
The cardiovascular system develops early in embryogenesis from cells of mesodermal origin. To study the molecular and cellular processes underlying this transition, we have isolated mesodermal cells from murine embryos at E7.5 with characteristic properties of endothelial progenitors by using a combination of stromal cell layers and growth conditions. The isolated embryonic cells displayed unlimited stem-cell-like growth potential and a stable phenotype in culture. RNA analysis revealed that the embryonic cells express the endothelial-specific genes tie-2 and thrombomodulin (TM) as well as the early mesodermal marker fgf-3. The GSL I-B4 isolectin, a marker of early endothelial cells, specifically binds to the isolated cells. The in vitro differentiation with retinoic acid and cAMP led to a 5– to 10-fold induction of flk-1, von Willebrand Factor (vWF), TM, GATA-4 and GATA-6. Electron microscopy revealed that in vitro differentiation is associated with increased amounts of rER and Golgi, and a dramatic increase in secretory vesicles packed with vWF. When cultured in Matrigel, the embryonic cells assume the characteristic endothelial cobblestone morphology and form tubes. Injection into chicken embryos showed incorporation of the embryonic cells in the endocardium and the brain vasculature. The expression of TM, tie-2, GATA-4 and GATA-6 suggests that the isolated embryonic endothelial cell progenitors are derived from the proximal lateral mesoderm where the pre-endocardial tubes form. The properties of the endothelial cell progenitors described here provide a novel approach to analyze mediators, signaling pathways and transcriptional control in early vascular development.
INTRODUCTION
The vascular system arises early in embryogenesis to meet the nutritional needs of the developing organism (Risau, 1995; Wilting et al., 1995; Baldwin, 1996; Wilting and Christ, 1996). At E7.5 in the mouse, the extraembryonic mesodermal cells of the yolk sac aggregate into clusters that constitute the initial stage of blood island formation. Shortly thereafter, blood islands differentiate into an external layer of endothelial cells and an inner core of blood cells (Sabin, 1920). At about the same time, embryonic cells of the proximal lateral mesoderm assemble into pre-endocardial tubes that connect anteriorly to generate cardiac endocardium around the anterior intestinal portal (Sabin, 1917; DeRuiter et al., 1992). Posteriorly, angioblasts organize into the paired dorsal aortae that later assemble in the midline to form a single tube (Jolly, 1940). In parallel fashion, mesodermal cells in the allantois generate cords that give rise to the umbilical blood vessels. The allantois expands rapidly inside the exocoelomic cavity and eventually fuses with the chorionic extraembryonic mesoderm. This fusion triggers differentiation of the chorionic vasculature that indirectly connects with the maternal vascular system of the placenta. These early events are described by the term vasculogenesis, which signifies de novo formation of blood vessels from angiogenic precursor cells (Risau, 1991). Later in embryogenesis, the vascular tree grows by sprouting, cell division, migration and assembly of endothelial cells derived from pre-existing vessels through a process termed angiogenesis (Noden, 1989; Risau and Flamme, 1995).
Classical embryological investigations have sought to ascertain the relationship between the different sites of origin of embryonic blood vessels. According to His (1868), both intraembryonic and extraembryonic vessels are derived from the yolk sac vasculature, whereas Rabl (1889) suggested that all intraembryonic vessels sprout from the endocardium. Subsequently, Hahn (1909), Reagan (1915) and Sabin (1917) postulated that both somatic and splanchnic mesoderm possess angiogenic potential. More recently, transplantation experiments in chicken/quail chimeras (Le Douarin, 1969) and staining of embryos with endothelial-specific markers have established that most areas of early mesoderm contain endothelial cell precursors or angioblasts (Pardanaud et al., 1987, 1989; Coffin and Poole, 1988; Noden, 1988; Poole and Coffin, 1989; Coffin et al., 1991; Pardanaud and Dieterlen-Lièvre, 1993).
These morphological investigations have been complemented by recent molecular studies in which genes have been identified that are expressed in endothelial cell progenitors and embryonic endothelial cells. These genes encode the tyrosine kinase receptors flk-1, flt-1, flt-4, tie-1 and tie-2 (for a review and references on endothelial-specific tyrosine kinases, see Mustonen and Alitalo, 1995), the thrombin receptor thrombomodulin (TM; Weiler-Guettler et al., 1996), the transcription factors GATA-4 as well as GATA-6 (Kelley et al., 1993; Laverriere et al., 1994), and the adhesive protein von Willebrand Factor (vWF; Coffin et al., 1991). Flk-1 and flt-1 are receptors for vascular endothelial growth factor (VEGF; Breier et al., 1992; Terman et al., 1992; Millauer et al., 1993), whereas tie-2 is a receptor for angiopoietin-1 (Davis et al., 1996). The temporal and spatial patterns of expression of the above genes differ greatly during development. In the extraembryonic mesoderm, which gives rise to blood islands, flk-1 appears shortly after gastrulation followed sequentially by tie-2 and tie-1 over the next 24 hours (Dumont et al., 1995). While flk-1 is initially observed in extraembryonic mesoderm prior to its expression in embryonic mesoderm (Eichman et al., 1993; Yamaguchi et al., 1993; Flamme et al., 1995), TM is first detected intraembryonically in the proximal lateral mesoderm where pre-endocardial tubes form (Weiler-Guettler et al., 1996). From E8.5 on, TM is ubiquitously expressed in endothelial cells throughout extraembryonic and embryonic tissues. The transcription factors GATA-4 and GATA-6 are detected at about the same time and in the same location as TM, but are subsequently absent from other embryonic or extraembryonic blood vessels (Arceci et al., 1993; Kelley et al., 1993; Heikinheimo et al., 1994; Laverriere et al., 1994; Jiang and Evans, 1996; Morrisey et al., 1996). The expression of the vWF gene occurs in the same location as TM and tie-2 but is delayed by about 24 hours and is eventually noted in most extraembryonic and embryonic blood vessels (Dumont et al., 1992).
Most of the genes discussed above have been inactivated by homologous recombination in mice. A common feature of these knockouts is embryonic lethality due to abnormal endothelial development. However, the corresponding phenotypes are distinctive. The flk-1 mutation affects both early endothelial development and hemopoiesis with endothelial cell precursors unable to undergo extraembryonic or intraembryonic vasculogenesis (Shalaby et al., 1995). On the contrary, tie-2 and angiopoietin-1 null embryos exhibit severe defects in angiogenesis that appear later in development (Dumont et al., 1994; Sato et al., 1995; Suri et al., 1996). Trabeculation in the heart is impaired and vasculature in brain as well as yolk sac remains homogeneous in size without undergoing reorganization to generate small and large vessels. The flt-1 knockout mice display defects in the assembly of endothelial cells into vascular channels (Fong et al., 1995), while tie-1 appears to be required for proper maintenance and integrity of the vascular system (Puri et al., 1995; Sato et al., 1995). In similar fashion, the VEGF knockout disrupts blood vessel formation (Carmeliet et al., 1996; Ferrara et al., 1996). Interestingly, mice heterozygous for VEGF exhibit aberrant vasculogenesis implying that ligand dose gradients are crucial for development of the vascular system. It is noteworthy that TM inactivation also leads to embryonic lethality at E8.5 suggesting that endothelial-specific functions such as thromboresistance may be critical for proper embryonic development (Healy et al., 1995).
The investigations described above have uncovered the existence of distinct networks of gene expression in intraembryonic and extraembryonic vasculogenesis. Although the ligands and receptors defined by these studies play a critical role in the developing vascular system, the precise intracellular events and altered cellular properties involved in endothelial cell differentiation remain obscure. We have established a potentially complementary approach for analyzing the molecular basis of vascular development that involves harvesting of endothelial cell progenitors which, through in vitro differentiation, can be used to define interactions that take place in vivo. In this communication, we describe the isolation and culture of early mesodermal embryonic cells that have properties of endothelial cell progenitors with unlimited, stem-cell-like growth. RNA and protein analyses demonstrate that the progenitor cells express a subset of mesodermal and early endothelial cell markers. In vitro differentiation leads to a dramatic induction of endothelial-specific genes and a morphological change to the mature endothelial cell phenotype. Transplantation experiments show that the isolated progenitor cells incorporate into the vasculature during embryonic development. These endothelial cell progenitors may allow investigation of mediators, signaling pathways and altered cellular properties that are important in early vascular development.
MATERIALS AND METHODS
Isolation of mouse embryonic endothelial progenitor cells
Mouse embryos were isolated at E7.5 to E7.8 of development. The midday of the vaginal plug was considered as day 0.5 post coitum. The egg cylinders with adjacent yolk sacs from 15–30 embryos were washed twice in phosphate-buffered saline (1× PBS; 100 mM NaCl, 4.5 mM KCl, 3 mM Na2HPO4, 3 mM KH2PO4) and incubated for 10–20 minutes with 0.05–0.1% trypsin/0.53 mM EDTA.4Na (Life Technologies). Dissociated cells were plated on γ-irradiated (3000 rads) mouse embryonic fibroblasts (MEFs) in ES culture medium (ESCM 20%) containing 20% heat-inactivated serum (55°C 30 minutes; Intergen), 0.1 mM β-mercaptoethanol, 1 mM MEM non-essential amino acids (Life Technologies), 100 u/ml penicillin and 100 μg/ml streptomycin (Life Technologies), 2 mM L-glutamine (Life Technologies) and 2 mM Hepes pH 7.5. Cells were grown in tissue culture incubators at 37°C and 5% CO2. Within 3-4 days, cell colonies of various morphologies emerged. When the above cells were similarly isolated from mouse embryos in which the TM gene was replaced by lacZ (β-galactosidase-producing gene; Weiler-Guettler et al., 1996), one type of colonies stained intensely blue with X-gal. These lacZ-positive colonies with round cell morphology were isolated using cloning rings and propagated separately on MEFs. After two passages on MEFs, the round cells could be maintained on 0.1% gelatin-coated plates in the absence of feeder layers. In all subsequent studies, endothelial progenitor cells were employed that had been extensively passaged on 0.1% gelatin-coated plates.
In vitro differentiation of embryonic endothelial cell progenitors
In vitro differentiation of embryonic endothelial cell progenitors was induced by addition of 1 μM all-trans retinoic acid (RA; Sigma) and 0.5 mM dibutyryl cyclic AMP (c-AMP; Sigma). RNA was isolated and antibody staining was performed following 4 days of treatment with RA and cAMP. For culture in Matrigel basement membrane matrix (Beckton Dickinson), endothelial progenitor cells cultured on gelatin-coated plates were resuspended in Matrigel, plated on 6-well plates and placed at 37°C until Matrigel polymerized. Subsequently, ESCM 20% was added and cultures observed for over 14 days.
RNA analysis
RNA was isolated using the RNA isolation kit from Stratagene. The following plasmids were linearized and used as templates to generate antisense RNA probes with Ambion’s MAXIscript kit; pTMPST674 (TM), pVEGF-1, pvWF, p421l1 (flk-1), pflt-1, pfgfa-a1, pfgfb-b1, pint-2c.28RI (fgf-3), pfgf5–5.3, pfgfk-k2, pkx10-32 (β-actin), pKFKF/PGK-Oct-3, pSK75 (brachyury), pAFPHB (α-fetoprotein), pPst232 (evx 1), pcM1x1.2Bam/BglIIA (Nkx-2.5), pGATA-1, pGATA-2, pGATA-4, pGATA-6 and p4A22.FR8 (cripto). RNase protections were conducted with Ambion’s RPA II kit following the manufacturer’s specifications except that overnight (O/N) hybridizations were carried out at 55–60°C. Products were resolved in 6% denaturing polyacrylamide gels. Gels were dried and autoradiographed.
Lectin and antibody binding
For the lectin staining, cells were grown in 8-chamber microscope slides. Chambers were washed twice with 1× PBS and fixed at 4°C in an 1:1 mixture of acetone:methanol for 5 minutes. Subsequently, cells were washed extensively with 1× PBS and processed for fluorescein-conjugated GSL I B4 binding (Vector Laboratories). Lectin concentration was 20 μg/ml in 10 mM Hepes pH 7.5, 150 mM NaCl. After binding for 1-2 hours at 24°C, cells were washed extensively and analyzed under a fluorescent microscope. For control, parallel cultures were stained in the presence of 0.5 M galactose, a specific inhibitor of GSL I B4 binding.
For detection of the von Willebrand Factor (vWF), an anti-human vWF rabbit polyclonal antibody (DAKO) that cross-reacts with murine antigen was used. Cells were washed twice with 1× PBS containing 1 mM MgCl2 and 0.1 mM CaCl2, fixed with 2% formaldehyde for 30 minutes, washed with 1× PBS and incubated with 1× PBS containing 5% fetal calf serum (FCS) for 30 minutes at 4°C to block non-specific binding. The cells were permeabilized with 0.2% saponin, incubated at 37°C with the anti-human vWF antibody (3 μg/ml) for 1 hour, washed several times with 1× PBS containing 1% albumin and then incubated with 1.5 μg/ml Cy3-conjugated donkey anti-rabbit IgG (Jackson Immunoresearch Laboratories Inc.) in 1× PBS containing 5% FCS for 1 hour. The samples were washed several times with 1× PBS containing 1% albumin, rinsed with distilled water and mounted in Mowiol (Calbiochem). The cells were viewed with a Zeiss Axioplan microscope or with a Biorad MRC 600 confocal microscope equipped with an argon krypton laser and photographed using Kodak film.
Electron microscopy
Endothelial progenitor cells were grown as described above for immunofluorescent microscopy. The cells were then rapidly rinsed with 1× phosphate buffer (PB), fixed with 2% glutaraldehyde in 0.1 M sodium cacodylate-HCl pH 7.2 for 60-90 minutes at 4°C, postfixed with 2% osmium tetraoxide for 60 minutes at 4°C followed by dehydration in graded alcohol up to 70% and en block staining with 70% alcoholic uranyl acetate for 30 minutes at 4°C. The samples were then processed for Epon embedding (Vasile et al., 1983). Thin Epon sections were poststained with uranyl acetate and Reynold’s lead citrate and viewed with a JEOL model 1200CX electron microscope operated at 80 kV.
For immunoelectron microscopy (Brown and Farquhar, 1989), cells were rapidly rinsed with 1× PB and fixed with 2% formaldehyde, 0.05 M lysine-HCl, 0.01 M sodium periodate in 0.75× PB. The cells were then incubated with 1× PB containing 5% normal goat serum for 30 minutes at 24°C to block non-specific binding and permeabilized with 0.2% saponin in the above solution. The cells were incubated with rabbit anti-human vWF antibody diluted to 3 μg/ml in PB containing 1% albumin and 0.02% saponin O/N at 4°C, washed in the same buffer and then incubated with 1.5 μg/ml goat anti-rabbit IgG conjugated to horse radish peroxidase at 24°C for 2 hours and finally washed in the same buffer. The cells were fixed with 3% glutaraldehyde, incubated with 0.015% diamino benzamidine-HCl and 0.01% H2O2 and processed for standard electron microscopy as described above.
ELISA assay
ELISA assays were carried out to measure vWF secretion using reagents from DAKO. Endothelial progenitor cells before and after stimulation with RA/cAMP for 4 days as described above were incubated O/N in a serum-free medium (ESCM) supplemented with 1× Nutridoma SP (Boehringer Mannheim). The following day, the supernatants were collected and centrifuged for 5 minutes at 1,100 revs/minute in a IEC model HN-SII table microfuge to remove floating cells. The supernatants were collected and processed for ELISA. The 96-well plates were coated with rabbit anti-human vWF antibody at 10 μg/ml for several days at 4°C in 10 mM PB pH 7.2 and 145 mM NaCl. The plates were washed extensively with 10 mM PB pH 7.2, 0.5 M NaCl and 0.1% Tween 20 (Buffer B). Supernatants diluted 2-5 fold with buffer B were placed in the wells and allowed to bind O/N at 4°C. The wells were then washed extensively with buffer B and the secondary peroxidase-conjugated rabbit anti-human vWF antibody was applied at 0.65 μg/ml in buffer B O/N at 4°C. Wells were extensively washed with buffer B and the chromogenic substrate (5 μl 30% H2O2 and 4 DAKOPATTS pellets in 12 ml 0.1 M citric acid-phosphate buffer pH 5.0) was applied for several hours at 24°C. When color was sufficiently intense, the reaction was stopped with 1.5× volume of 1 M sulfuric acid and optical density was measured in a DuPont microplate reader. For control, wells were incubated with either buffer B or ESCM that was incubated O/N at 37°C without cells and processed in parallel. Samples were prepared and measured in quadruplicates.
Xenograft of mouse cells in chicken embryos
Murine embryonic progenitor cells transfected with the PGKβ-geo construct were injected in chicken embryos at day 9 or 10 of development (β-geo is a fusion between lacZ (β-galactosidase gene) and neomycin resistance genes; Friedrich and Soriano, 1991). Murine embryonic fibroblasts, isolated from a ROSA gene-trap mouse line that ubiquitously expresses the lacZ gene, were used as a control. The two cell types displayed similar intensity of blue color when stained with X-gal. 150,000-350,000 cells in 3-7 μl were injected into an extraembryonic vein. The chicken embryos were isolated 1, 3 or 4 days later, dissected and frozen on dry ice. Subsequently, frozen tissue was mounted and cryosectioned. Sections were stored at −80°C until use. For X-gal staining, sections were fixed at 4°C in 1:1 acetone methanol for 5 minutes and washed twice for 10 minutes with 1× PBS. After fixation, sections were stained O/N at 37°C in a staining solution of 1× PBS containing 0.02% NP40, 0.01% SDS, 2 mM MgCl2, 5 mM potassium ferricyanide, 5 mM potassium ferrocyanide and 1 mg/ml X-gal (Boehringer Mannheim). After staining, sections were washed twice for 10 minutes with 1× PBS and photographed.
For antibody staining, cryosections of chicken embryos injected with murine cells were removed from −80°C, air dried for 10-20 minutes and fixed at 4°C in 1:1 acetone methanol for 5 minutes. Slides were then washed twice for 10 minutes in 1× PBS and blocked with 1% BSA and 0.05% saponin at 24°C for 2-3 hours. Subsequently, primary antibodies were added after dilution in blocking buffer. The rabbit polyclonal anti-human vWF antibody from DAKO and a rat monoclonal anti-mouse TM antibody were used at 3 μg/ml and 30 ng/ml respectively. Staining was O/N at 4°C. The slides were then washed twice for 10 minutes in 1× PBS and secondary antibodies were applied in blocking buffer for 2-3 hours at 24°C. The secondary antibodies were fluorescein (DTAF)-conjugated donkey anti-rabbit IgG (H+L) at 3.75 μg/ml and Cy3-conjugated donkey anti-rat IgG (H+L) at 2.5 μg/ml (both from Jackson Immunoresearch Laboratories Inc.). The slides were washed twice for 10 minutes in 1× PBS, mounted with Vectashield (Vector Laboratories) and examined under either fluorescent or confocal microscopes.
RESULTS
Isolation of cells from early mouse embryos
We sought to isolate and grow endothelial cell progenitors from murine embryos at a stage when the first signs of vascular development are detected. Our approach was to identify cells from trypsin-dissociated E7.5–E7.8 embryos, confirm their identity with cell-type- and stage-specific genetic markers as well as characteristic morphological properties and then establish conditions for their maintenance in culture. Initial attempts to grow cells directly on plates coated with substrates such as laminin, gelatin and collagen failed. The addition to the growth media of various combinations of aFGF, bFGF, TGFβ, PDGF, LIF and VEGF also proved insufficient to support the proliferation of murine cells from early embryos. However, after incomplete trypsinization, cells were occasionally noted attached to the plate. This observation raised the possibility that a feeder or a stromal cell layer was required to provide an appropriate substratum for sustained cell proliferation. Indeed, in subsequent experiments, dissociated E7.5–E7.8 embryos were plated on γ-irradiated mouse embryonic fibroblasts and displayed a robust cell growth in standard tissue culture medium containing 20% serum.
The experimental scheme for this procedure is depicted in Fig. 1A employing murine embryos in which the lacZ gene had been targeted into the TM locus (Weiler-Guettler et al., 1996). At E7.5, X-gal staining demonstrates that expression is confined to the proximal lateral mesoderm where the first intraembryonic endothelial cell progenitors originate (insert photograph). TMlacZ embryos were isolated at E7.5–E7.8, and the entire decidua with trophoblast cells, ectoplacental cone, parietal endoderm and Reichert’s membrane was removed by microdissection. The remaining egg cylinders with the adjacent yolk sacs were trypsinized and plated on γ-irradiated mouse embryonic fibroblasts. Within 3 to 4 days, many morphologically distinct colonies were observed. After about a week, only round cells and fibroblastoid cells remained.
(A) Mouse embryos were isolated between E7.5 and E7.8. The photograph at the upper right corner shows an E7.5 TMlacZ embryo stained with X-gal. The decidua, trophectoderm, parietal endoderm and Reichert’s membrane, as well as the tissues above the indicated broken line including the ectoplacental cone and chorion, were removed. The remaining egg cylinder with the adjacent yolk sac were trypsinized and placed on γ-irradiated mouse embryonic fibroblasts as a feeder layer. Within a week, two types of colonies of round and fibroblastoid morphologies were observed. Colonies of distinct morphologic type were separated and grown individually. (B) Picture of a colony of round cell morphology isolated from E7.5–E7.8 TMlacZ murine embryos and stained with X-gal. (C) Endothelial cell progenitors isolated from E7.5– E7.8 wild-type mouse embryos after transfer to gelatin-coated plates.
(A) Mouse embryos were isolated between E7.5 and E7.8. The photograph at the upper right corner shows an E7.5 TMlacZ embryo stained with X-gal. The decidua, trophectoderm, parietal endoderm and Reichert’s membrane, as well as the tissues above the indicated broken line including the ectoplacental cone and chorion, were removed. The remaining egg cylinder with the adjacent yolk sac were trypsinized and placed on γ-irradiated mouse embryonic fibroblasts as a feeder layer. Within a week, two types of colonies of round and fibroblastoid morphologies were observed. Colonies of distinct morphologic type were separated and grown individually. (B) Picture of a colony of round cell morphology isolated from E7.5–E7.8 TMlacZ murine embryos and stained with X-gal. (C) Endothelial cell progenitors isolated from E7.5– E7.8 wild-type mouse embryos after transfer to gelatin-coated plates.
The round cell colonies derived from TMlacZ mice stained intensely blue after X-gal treatment, indicating origination from proximal lateral mesoderm (Fig. 1B). The round cell colonies were isolated using cloning rings and grown separately. Although the initial isolation required the use of feeder layers, once cultures were established, cell growth could be sustained on gelatin-coated plates. The morphology of the cloned round cells is shown in Fig. 1C.
Employing the isolation protocol described above, similar results have been obtained with other murine strains including C57BL, 129/J and BALBc. The round cells show unlimited growth potential, which is an unusual property of primary cultures. At the present time, the initially cloned round cells have been continuously propagated in culture for 3 years without diminished growth or phenotypic changes.
The isolated round cells express early endothelial cell markers
In earlier investigations of vascular development, the GSL I B4 isolectin has been demonstrated to bind to endothelial progenitors and mature endothelial cells between E7.5 and E9.0 (Goldstein and Hayes, 1978; Coffin et al., 1991). Therefore, we determined whether GSL I B4 binds to the isolated round cells. Fig. 2 documents a strong positive staining that is completely abolished by incubation with galactose, a specific inhibitor of GSL I B4 binding.
GSL I-B4 isolectin strongly binds to the isolated endothelial cell progenitors (upper panel). Binding is completely blocked by 0.5 M galactose, a specific inhibitor of GSL I-B4 binding (lower panel).
The above observations strengthen the notion that these cells might represent murine embryonic endothelial cell progenitors. To further examine this possibility, we investigated the gene expression profile of these cells by RNase protection assays with embryonic stem (ES) cells as a control. For these experiments, 25 genes were selected for study that were members of the following groups: (1) endothelial-specific genes expressed early in development, (2) genes expressed postgastrulation in lateral and primitive streak mesoderm, endoderm and ectoderm, and (3) genes expressed in cell lineages that develop in close association with endothelial cells, i.e., myocardium and blood.
Fig. 3 shows representative data for a selected set of genes with RNase protection assays. The endothelial progenitors express high levels of endothelial cell markers such as tie-2 and TM as well as the early mesodermal marker fgf-3 (int-2; Wilkinson et al., 1988). These cells also exhibit small amounts of VEGF mRNA as well as trace quantities of vWF mRNA, but possess no detectable flk-1 mRNA or flt-1 mRNA. Finally, the endothelial cell progenitors show no expression of mesodermal markers associated with primitive streak and axial mesoderm such as brachyury (Wilkinson et al., 1990) and evx 1 (Bastian and Gruss, 1990), the endodermal marker α-fetoprotein (Tilghman, 1985), the ectodermal marker fgf-5 (Hébert et al., 1991), early myocardial markers such as Nkx-2.5 (Lints et al., 1993), MLC-2a (Kubalak et al., 1994) and cripto (Dono et al., 1993), early blood markers like GATA-1 (Pevny et al., 1991), GATA-2 (Tsai et al., 1994), embryonic or adult globins or other genes including fgf-1, fgf-2, kfgf (Hébert et al., 1990) and oct3/4 (Schöler et al., 1990).
RNase protection analysis of gene expression in isolated endothelial cell progenitors. The probe used in each experiment is indicated to the left. Asterisks mark protected fragments. The first two lanes of each panel represent incubation of the probe with yeast RNA in the absence (−) or presence (+) of RNases. ES, RNA from Embryonic Stem cells; MEEPs, RNA from mouse embryonic endothelial progenitors. In all cases, 10 μg of total RNA was incubated with about 50-100,000 cts/minute of 32P radioactively-labeled antisense RNA probes. β-actin was used as a control to monitor RNA levels in each lane.
RNase protection analysis of gene expression in isolated endothelial cell progenitors. The probe used in each experiment is indicated to the left. Asterisks mark protected fragments. The first two lanes of each panel represent incubation of the probe with yeast RNA in the absence (−) or presence (+) of RNases. ES, RNA from Embryonic Stem cells; MEEPs, RNA from mouse embryonic endothelial progenitors. In all cases, 10 μg of total RNA was incubated with about 50-100,000 cts/minute of 32P radioactively-labeled antisense RNA probes. β-actin was used as a control to monitor RNA levels in each lane.
In vitro differentiation of endothelial cell progenitors
Investigations mainly in Xenopus embryos have identified growth factors and other agents that regulate formation and differentiation of early mesoderm. These components include members of the FGF and TGFβ families, which induce ventralization, and retinoic acid (RA) and noggin, which induce dorsalization (Slack, 1994). Based upon these data, endothelial cell progenitors were exposed in culture to TGFβ, bFGF or RA together with cyclic AMP (RA/cAMP) to test whether these factors might influence growth and/or induce differentiation. These experiments demonstrated that TGFβ1 inhibits cell growth without affecting cell morphology or the pattern of gene expression, except for a moderate increase in VEGF mRNA (data not shown). These results also showed that bFGF had no obvious effect other than slightly enhancing the action of TGFβ1 (not shown). In contrast, exposure to RA/cAMP for 4 days led to dramatic morphological changes. The cells became flat, elongated and assumed an endothelial cell-like shape. RNA was isolated from RA/cAMP-treated cells as well as untreated cells, and RNase protection assays were carried out using the same battery of genes described above. In addition, two newly identified members of the GATA family, GATA-4 and GATA-6 were included in these studies because both genes are expressed early in pre-endocardial tubes.
We observed a dramatic elevation in the levels of vWF and TM mRNAs upon treatment with RA/cAMP (Fig. 4). Interestingly, the expression of flk-1 was activated whereas that of tie-2 was unaffected. The existing low levels of GATA-4 as well as GATA-6 were also greatly increased by exposure to RA/cAMP. Normalized for β-actin, the quantitative measurements demonstrate elevated message levels of about 5– fold for GATA-4, 8-fold for vWF, TM as well as GATA-6, and more than 10-fold for flk-1. The expression of GATA-4, GATA-6 and TM in uninduced endothelial cell progenitors suggest origination from the proximal lateral mesoderm where pre-endocardial tubes arise. The lack of expression of flk-1 in the uninduced cells implies that the isolated cells may be distinct from extraembryonic ventrolateral mesoderm, which gives rise to blood islands. The presence of fgf-3 mRNA substantiates the primitive nature of endothelial cell progenitors as compared to mature endothelial cells, which do not possess this marker. Table 1 contains a list of the genes analyzed with specific probes in the RNase protection assays.
A list of the genes analyzed by RNase protection assays in the undifferentiated and differentiated murine endothelial cell progenitors

RNase protection assays with RNA isolated from murine endothelial cell progenitors exposed to retinoic acid and cAMP. The probes used in each experiment are outlined at the top of each panel. UN, RNA from untreated endothelial cell progenitors; RC, RNA from RA and cAMP-treated cells. In all cases, 10 μg of total RNA was incubated with about 50-100,000 cts/minute of 32P radioactively-labeled antisense RNA probes. β-actin was used as a control to monitor RNA levels in each lane.
RNase protection assays with RNA isolated from murine endothelial cell progenitors exposed to retinoic acid and cAMP. The probes used in each experiment are outlined at the top of each panel. UN, RNA from untreated endothelial cell progenitors; RC, RNA from RA and cAMP-treated cells. In all cases, 10 μg of total RNA was incubated with about 50-100,000 cts/minute of 32P radioactively-labeled antisense RNA probes. β-actin was used as a control to monitor RNA levels in each lane.
Analysis of vWF gene induction
The induction of vWF gene expression constitutes an early molecular marker of endothelial cell development (Coffin et al., 1991). The biosynthesis, processing and secretion of this adhesive protein represents a characteristic property of mature endothelial cells (Jaffe et al., 1973; Wagner and Marder, 1984). To evaluate whether endothelial cell progenitors are able to carry out the full sequence of events, differentiated and undifferentiated cells were stained with a rabbit anti-human vWF antibody that recognizes the murine form of the adhesive protein. As shown in Fig. 5A, vWF protein accumulates in the cytoplasm of differentiated cells. There is strong perinuclear staining indicative of rough endoplasmic reticulum (rER) as well as the staining of large granules dispersed throughout the whole cytoplasm.
Analysis of von Willebrand Factor induction. (A) Confocal microscopy of undifferentiated (UN) and RA/cAMP differentiated cells (RC) stained with a rabbit anti-human vWF polyclonal antibody. The secondary antibody is a donkey anti-rabbit IgG conjugated to Cy3. (B) Electron microscopy of undifferentiated (UN) and differentiated endothelial cell progenitors (RC). After differentiation, a dramatic increase of rER (arrow) and Golgi cisternae number in the cytoplasm indicates a transition to a secretory cell type (N, nucleus). Electron microscopy was performed following vWF antibody binding coupled to secondary anti-rabbit antibody-conjugated to horse radish peroxidase. The color reaction (shown as dark gray staining) reveals that vWF is found in large secretory vesicles (arrowheads). Bars are 1 μm. The insert shows an area of secretory vesicles containing vWF magnified 1.7× compared to the right panel. (C) The immunoassay of vWF in differentiated and undifferentiated endothelial cell progenitor tissue culture supernatants reveals increased secreted protein upon differentiation.
Analysis of von Willebrand Factor induction. (A) Confocal microscopy of undifferentiated (UN) and RA/cAMP differentiated cells (RC) stained with a rabbit anti-human vWF polyclonal antibody. The secondary antibody is a donkey anti-rabbit IgG conjugated to Cy3. (B) Electron microscopy of undifferentiated (UN) and differentiated endothelial cell progenitors (RC). After differentiation, a dramatic increase of rER (arrow) and Golgi cisternae number in the cytoplasm indicates a transition to a secretory cell type (N, nucleus). Electron microscopy was performed following vWF antibody binding coupled to secondary anti-rabbit antibody-conjugated to horse radish peroxidase. The color reaction (shown as dark gray staining) reveals that vWF is found in large secretory vesicles (arrowheads). Bars are 1 μm. The insert shows an area of secretory vesicles containing vWF magnified 1.7× compared to the right panel. (C) The immunoassay of vWF in differentiated and undifferentiated endothelial cell progenitor tissue culture supernatants reveals increased secreted protein upon differentiation.
The cytoplasmic localization of vWF protein was further defined using electron microscopy (EM). In the untreated cells, subcellular organelles associated with the secretory pathway, i.e. rER and the Golgi apparatus, are scarce suggesting that very little secretion takes place. Instead, a large number of free ribosomes are present (Fig. 5B, left panel). Following differentiation, a remarkable increase is observed in the number of both rER and Golgi cisternae documenting a shift to the secretory phenotype (Fig. 5B, right panel). A horse-radish-peroxidase-linked secondary antibody revealed the presence of vWF protein in large vesicles that appear ready for secretion. Although additional intracellular sites were stained, Weibel-Palade bodies were not conclusively identified (Wagner et al., 1982). The appearance of large secretory vesicles indicated that the differentiated cells secrete vWF protein. This surmise was verified by measuring the levels of vWF in the media of RA/cAMP-treated and untreated cells. As shown in Fig. 5C, a marked increase in the amount of secreted vWF protein was documented, substantiating the transition toward a functional endothelial cell phenotype. In vitro labeling with [35S]methionine and [35S]cysteine, followed by immunoprecipitation and SDS-PAGE demonstrated that vWF protein was correctly processed with removal of the prepeptides and propeptides (data not shown).
Isolated round cells assume endothelial morphology in Matrigel
The characteristics of endothelial cell progenitors were also assessed by culture in Matrigel, an extracellular matrix basement membrane material that can be used to promote differentiation of endothelial cells (Grant et al., 1991). As shown in Fig. 6, the cells gradually lose their round appearance and assume the characteristic cobblestone morphology of mature endothelial cells. Further incubation in Matrigel produced tube-like structures, a property of mature endothelial cells. These results provide additional support for the notion that the round cells represent endothelial cell progenitors.
The isolated endothelial cell progenitors were placed in Matrigel, a substratum that enhances differentiation of endothelial cells. (A) 2 days in culture, (B) 3 days in culture, (C) 5 days in culture and (D) 8 days in culture. The pictures show that the endothelial cell progenitors progressively lose their round shape, assume the characteristic cobblestone morphology of mature endothelial cells and form tube-like structures.
The isolated endothelial cell progenitors were placed in Matrigel, a substratum that enhances differentiation of endothelial cells. (A) 2 days in culture, (B) 3 days in culture, (C) 5 days in culture and (D) 8 days in culture. The pictures show that the endothelial cell progenitors progressively lose their round shape, assume the characteristic cobblestone morphology of mature endothelial cells and form tube-like structures.
The isolated endothelial cell progenitors incorporate into the embryonic vasculature
Transplantation experiments are difficult to perform in early mouse embryos because of the small size and surrounding thick decidual tissue. Chicken embryos are receptive to xenografts at early stages of development until a functional thymus has been formed and are easily accessible for transplantation studies. For these reasons, the developmental potential of the isolated cells was examined by injection into the extraembryonic veins of chicken embryos between day 9 and 10 of development. The endothelial cell progenitors employed for the injection experiments were previously transfected with a plasmid carrying the PGKβ−geo construct. After G418 selection, several neomycin-resistant colonies were isolated and RNA prepared. RNA analysis using most of the probes described above demonstrated that each colony exhibited the same expression profile as the wild-type isolated cells normalized to β-actin (data not shown). These observations showed that isolated endothelial progenitors represent a homogeneous population and provided an easily detected marker to identify areas engrafted with murine cells.
After 3–4 days following injection, chicken embryos were isolated, frozen and cryosectioned. The sections were fixed and stained with X-gal to locate murine cells. The analysis showed that large numbers of endothelial cell progenitors had survived in the chicken embryos and that the majority of murine cells were present in the developing heart, brain and liver (not shown). The location and relationship of endothelial cell progenitors to the host chicken endothelium was analyzed using two endothelial-specific antibodies: (1) an anti-mouse TM rat monoclonal antibody that recognizes only mouse but not chicken antigen in conjunction with a secondary Cy3-conjugated donkey anti-rat antibody (red color), and (2) the previously mentioned rabbit polyclonal anti-human vWF antibody that recognizes both mouse and chicken antigens in association with a fluorescein-conjugated donkey anti-rabbit secondary antibody (green color). The cryosections were then examined for antibody staining by confocal microscopy as shown in Fig. 7. The endothelial cell progenitors are present within trabeculae in the ventricles of the developing heart, where murine cells became part of the host endocardium (Fig. 7A-C). Indeed, chicken myocardium is occasionally completely surrounded by murine cells. Many murine cells are also found around the developing retina, between the sensory and pigment layers, an area that harbors host endothelial cells (Fig. 7F). In certain cases, endothelial cell progenitors are integrated into the brain microvasculature (Fig. 7D,E). Murine cells in the liver did not appear to associate with endothelial cell sinusoids. No incorporation of progenitor cells was noted in the extraembryonic membranes or in large, well-established intraembryonic vessels. The experiment was repeated several times using different cell clones with identical results. As controls, similar numbers of murine embryonic fibroblasts expressing lacZ were injected at the same stage of development. Early time points (1 day) revealed the presence of fibroblasts throughout the embryo and the circulation whereas later points (4 days) showed no localization to specific organs and no incorporation into chick heart or brain. These results demonstrate that the isolated round cells exhibit the potential to form vasculature and support the conclusion that these cells represent an early progenitor cell.
Injection of murine endothelial cell progenitors into the extraembryonic veins of chicken embryos. The presence of murine cells in various organs was documented by confocal immunofluorescence microscopy on antibody stained cryosections. Chicken endothelial cells are marked with a primary rabbit polyclonal anti-vWF antibody in association with a secondary fluorescein-conjugated donkey anti-rabbit IgG (green color). Murine cells are stained with a rat anti-TM monoclonal antibody in conjunction with a secondary Cy3-conjugated donkey anti-rat antibody (red color). Computer-assisted superposition graphically reveals the incorporation of murine cells in the host circulatory system (yellow color). (A-C) Heart. The same section is stained with anti-vWF (A) and anti-TM (B) antibodies; computer-assisted superposition of A and B is shown in C (L, lumen; M, myocardium; bar is 40 μm). (D-F) Brain. In D and E, computer-assisted superposition of anti-vWF and anti-TM staining demonstrates incorporation of murine cells in the brain vasculature; bars are 20 μm? A large number of murine cells were present around the developing eye as shown in F; bar is 30 μm
Injection of murine endothelial cell progenitors into the extraembryonic veins of chicken embryos. The presence of murine cells in various organs was documented by confocal immunofluorescence microscopy on antibody stained cryosections. Chicken endothelial cells are marked with a primary rabbit polyclonal anti-vWF antibody in association with a secondary fluorescein-conjugated donkey anti-rabbit IgG (green color). Murine cells are stained with a rat anti-TM monoclonal antibody in conjunction with a secondary Cy3-conjugated donkey anti-rat antibody (red color). Computer-assisted superposition graphically reveals the incorporation of murine cells in the host circulatory system (yellow color). (A-C) Heart. The same section is stained with anti-vWF (A) and anti-TM (B) antibodies; computer-assisted superposition of A and B is shown in C (L, lumen; M, myocardium; bar is 40 μm). (D-F) Brain. In D and E, computer-assisted superposition of anti-vWF and anti-TM staining demonstrates incorporation of murine cells in the brain vasculature; bars are 20 μm? A large number of murine cells were present around the developing eye as shown in F; bar is 30 μm
DISCUSSION
During the past few years, considerable progress has been made in elucidating the molecular basis of vascular development. Specific tyrosine kinase receptors have been shown to be expressed in endothelial cell precursors. Inactivation of these genes and their corresponding ligands by homologous recombination has revealed the crucial importance of these signaling systems in the development, assembly and maintenance of the vascular system (Mustonen and Alitalo 1995). Moreover, components of the extracellular matrix and their cellular receptors such as fibronectin and integrins are also indispensable to the formation of blood vessels (George et al., 1993; Yang et al., 1993, 1995). However, it remains to be established if a single or multiple endothelial cell precursors exist for different vascular beds. It is also unclear whether a common set or multiple sets of molecular signals are required to induce differentiation of endothelial cell progenitors in the yolk sac, allantois and endocardial tubes, and if the various endothelial cell-specific signaling systems play the same role at each site. The nature of molecular interactions that establish the structural and functional diversity of endothelium as well as the linkage between endothelial cell development and hemopoiesis remain to be defined.
We have isolated and maintained in culture embryonic endothelial cell progenitors that arise on E7.5 to E7.8 prior to the formation of the cardiovascular system. The procedures developed to obtain cell populations with the characteristics of endothelial progenitors are simple as well as reproducible and generate homogenous cell populations with unlimited growth properties. The homogeneous nature of endothelial cell progenitors was established by isolation/expansion of single colonies obtained with cloning rings, and the analyses of gene expression in multiple independent clones following transfection with neomycin-resistant plasmids. The stem-cell-like unlimited growth of endothelial cell progenitors is shared by only one other primary culture system, the embryonic stem cells.
The isolated progenitor cells express high levels of early endothelial cell-specific genes, such as TM and tie-2, and stain with the GSL I B4 lectin, a marker for endothelial cells and their precursors (Coffin et al., 1991; Schnurch and Risau, 1993; Weiler-Guettler et al., 1996). These cell populations also express genes of early mesoderm such as fgf3, but not those specific for primitive streak mesoderm, notochord, endoderm, ectoderm, extraembryonic mesoderm or myocardium. The endothelial cell progenitors undergo in vitro differentiation with a dramatic induction of flk-1, vWF, GATA-4, GATA-6 and TM genes. Differentiation is accompanied by a transition to the mature endothelial cell phenotype with the correct processing and secretion of vWF. The endothelial cell progenitors when placed in Matrigel assume the characteristic cobblestone morphology of mature endothelial cells and assemble in tube-like structures. Finally, even after prolonged culture, the above cloned cell population retains the potential to contribute to the developing vascular system as shown by xenograft transplantation in chicken embryos.
Given that the isolated progenitor cells possess the X-gal reaction product when derived from TMlacZ mice and express GATA-4 as well as GATA-6, it appears likely that this cell population originates from the proximal lateral mesoderm. In previous studies of embryonic development, two patterns of vasculogenesis have been described (Noden, 1989). Intraembryonic round cells have been observed that assemble locally to form endothelium in the midline of the embryo (Dumont et al., 1992). A mesenchymal fibroblast-like cell from the splanchnopleure or ventral mesoderm has been identified that migrates widely and forms vascular cords (Jolly, 1940). We presume that endothelial cell progenitors isolated in this study represent the first type, i.e., cells that assemble to form pre-endocardial tubes and eventually differentiate into endocardium, dorsal aorta and other portions of the central vascular system. In the light of the above discussion, the extensive incorporation of endothelial cell progenitors in the trabeculae of the developing chicken heart may reflect an affinity for myocardium and/or the presence of the tie-2 ligand, angiopoietin-1, in this tissue. It is also interesting to note that accumulation of injected cells around the developing eye might be explained by the recent observation that angiopoietin-1 is also highly expressed in neighboring epithelium (Davis et al., 1996).
A second endothelial progenitor cell type might then be responsible for generating blood islands. Consistent with this notion, we have isolated from E7.5 to E7.8 embryos a fibroblastoid cell population that responds to bFGF and expresses endothelial markers including flk-1 and flt-1, as well as scavenger receptor that leads to acetylated Low Density Lipoprotein uptake (ac-LDL; unpublished data). This embryonic cell population resembles early endothelial precursors isolated from quail blastodiscs with regard to their morphology, ac-LDL uptake and dependence on bFGF for growth (Flamme and Risau, 1992). It is noteworthy that transplantation experiments in chicken/quail chimeras support the existence of two distinct populations of endothelial cell progenitors; one that contributes to formation of the dorsal aorta and dorsal vessels and the other that contributes to ventral vessels and also gives rise to hematopoietic precursors (Pardanaud et al., 1996).
The absence of flk-1 expression in endothelial cell progenitors with induction of the gene upon differentiation in vitro is somewhat surprising. It is possible that flk-1 expression in endothelial cell progenitors is maintained in vivo by an exogenous factor and thus lost upon isolation. Alternatively, it is conceivable that, unlike blood islands, the expression of flk-1 in endocardial tubes occurs later than other endothelial cell-specific genes. From published reports and our own whole-mount in situ data with flk-1 probes, this later possibility appears reasonable because widespread expression of this receptor gene in the intraembryonic mesoderm obscures its early endocardial tube expression. Indeed, the staining of parallel sections of E7.8 embryos with antibodies against TM and flk-1 reveal that TM expression precedes flk-1 expression in endocardial tubes (data not shown).
While this manuscript was in preparation, the isolation of putative human endothelial cell progenitors was reported from adult human blood (Asahara et al., 1997). In contrast to these findings, the endothelial cell progenitors described in the present study could only be isolated from E7.5 to E7.8 embryos. Employing identical conditions of isolation, we have failed to obtain the same cells from older embryos or adult tissues. At present, the relationship between the two progenitor cell types is unclear. Differences exist in morphology and patterns of gene expression. The isolated adult endothelial cell progenitors are spindle shaped, express both flk-1 and tie-2, have a finite life span in vitro and appear to differentiate to cells positive for ac-LDL uptake. Thus, these progenitors resemble the cells dissociated from quail blastodiscs (Flamme and Risau, 1992). The adult progenitor cells isolated from blood integrate into sites of active angiogenesis after venous injection. In similar fashion, it has been previously shown that human umbilical vein endothelial cells can also incorporate around angiogenic sites after injection (Ojeifo et al., 1995). It will be of great interest to compare the behavior of the embryonic endothelial cell progenitors in response to angiogenic signals in adult pathological situations (Folkman, 1995).
The endothelial cell progenitors isolated in this investigation should provide a new tool for elucidating regulatory events in vasculogenesis. The ease of isolation, unlimited growth without detectable phenotypic changes, feasibility of genetic manipulation and in vivo developmental potential of progenitors constitute major advantages of this approach. We are currently searching for signals, besides RA/cAMP, that induce genes like flk-1 and vWF and hope to determine whether such factors have comparable functions in vivo. In this regard, signals of this type are known to originate from the anterior intestinal portal endoderm that induce both endocardial- and myocardial-specific genes in isolated mesodermal cells (Sugi and Markwald 1996). The endothelial cell precursors can be employed in differential screens to identify new genes that are induced or suppressed after differentiation in vitro and similar changes can then be sought in vivo. It should also be possible to isolate progenitors from mice with knockouts of endothelial cell-specific receptors which cause later embryonic lethality; the behavior of these genetically altered cells can be then assessed by a combination of cell culture and transplantation experiments. The above studies should help to define the regulation of vasculogenesis and shed light on angiogenic mechanisms activated in adult pathological situations affecting the vascular system, e.g., tumor-induced angiogenesis, diabetic retinopathy, psoriasis, myocardial infarction and endothelial cell growth during tissue remodeling.
ACKNOWLEDGEMENTS
We are greatly indebted to many of our colleagues for sending us cDNA probes and antibodies for analysis: Dr S. Kennel for the TM antibody; Dr B. G. Hermann for the brachyury cDNA, Dr D. Dumont for flk-1 and tie-2, Drs G. Breier and W. Risau for flt-1 and VEGF, Drs G. Martin and U. Deutch for fgf1, fgf2, fgf3, kfgf and fgf5, Dr S. H. Orkin for GATA-1 and GATA-2, Dr M. S. Parmacek for GATA-4 and GATA-6, Drs P. Gruss and H. Bastian for evx 1, Dr P. Soriano for β-geo, Drs K. Chien and S. Kubalak for MLC-2a, Dr R. P. Harvey for Nkx-2.5, Dr M. Persico for cripto, Dr H. Schöler for oct3/4 and Dr Tilghman for α-fetoprotein. We would like to thank Drs D. Fischman and T. Mikawa for helpful comments and suggestions, Dr H. Rayburn for advice on animal husbandry, Dr Cesario Bianchi for his help and advice on the lectin and antibody stainings, Dr S. Chatterji and D. Smith for their help in using the confocal microscope, and J. Jackson and Mrs C. Battlefield for preparing chicken embryos and cells for microinjection. We would like to thank Drs M. Houssain, W. Aird, M. Krieger and E. W. Knapik for insightful comments on the manuscript. This work was supported by HL 41434.