Stromal Cell-Derived Factor-1/CXCR4 Signaling Modifies the Capillary-Like Organization of Human Embryonic Stem Cell-Derived Endothelium In Vitro


  • Tong Chen,

    1. Center for Regenerative Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts, USA
    2. Department of Hematology, Huashan Hospital, Fudan University, Shanghai, China
    Search for more papers by this author
  • Hao Bai,

    1. Maine Medical Center Research Institute, Scarborough, Maine, USA
    Search for more papers by this author
  • Ying Shao,

    1. Center for Regenerative Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts, USA
    Search for more papers by this author
  • Melanie Arzigian,

    1. Maine Medical Center Research Institute, Scarborough, Maine, USA
    Search for more papers by this author
  • Viktor Janzen,

    1. Center for Regenerative Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts, USA
    Search for more papers by this author
  • Eyal Attar,

    1. Center for Regenerative Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts, USA
    Search for more papers by this author
  • Yi Xie,

    1. Department of Hematology, Huashan Hospital, Fudan University, Shanghai, China
    Search for more papers by this author
  • David T. Scadden,

    1. Center for Regenerative Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts, USA
    2. Harvard Stem Cell Institute, Harvard University, Cambridge, Massachusetts, USA
    Search for more papers by this author
  • Zack Z. Wang Ph.D.

    Corresponding author
    1. Center for Regenerative Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts, USA
    2. Maine Medical Center Research Institute, Scarborough, Maine, USA
    • Maine Medical Center Research Institute-Center for Regenerative Medicine & Technology, 81 Research Drive Scarborough, Maine 04074, USA. Telephone: 207-885-8313; Fax: 207-885-8179
    Search for more papers by this author


The molecular mechanisms that regulate human blood vessel formation during early development are largely unknown. Here we used human ESCs (hESCs) as an in vitro model to explore early human vasculogenesis. We demonstrated that stromal cell-derived factor-1 (SDF-1) and CXCR4 were expressed concurrently with hESC-derived embryonic endothelial differentiation. Human ESC-derived embryonic endothelial cells underwent dose-dependent chemotaxis to SDF-1, which enhanced vascular network formation in Matrigel. Blocking of CXCR4 signaling abolished capillary-like structures induced by SDF-1. Inhibition of the SDF-1/CXCR4 signaling pathway by AMD3100, a CXCR4 antagonist, disrupted the endothelial sprouting outgrowth from human embryoid bodies, suggesting that the SDF-1/CXCR4 axis plays a critical role in regulating initial vessel formation, and may function as a morphogen during human embryonic vascular development.


The study of early development in humans is hampered by the lack of experimental systems. However, the establishment of human ESCs (hESCs) provides a novel opportunity to study early human developmental events. Human ESCs are pluripotent cells capable of forming teratomas comprised of cells from all three germ layers in vivo [1]. They have the potential to differentiate in vitro into a variety of cell lineages, including cardiac tissue [2], neuronal tissue [3], β-islet pancreatic cells [4], hematopoietic cells [5], and endothelial cells [6, 7]. The hESC differentiation process follows a reproducible temporal pattern of development, which, to some extent, resembles early human embryogenesis [8].

During embryogenesis, primitive blood vessels are formed de novo by the aggregation of angioblasts, a process termed vasculogenesis [9]. In postnatal life, the development of new blood vessels is restricted to the female reproductive tract and to sites of wound healing, and occurs through angiogenesis, the growth of new vessels from pre-existing vessels [10]. Vasculogenesis begins in embryos at around day 18, during which fibroblast growth factors (FGFs), and other factors cause some cells in the mesoderm to undergo differentiation into endothelial progenitors, for example, the angioblasts that develop into embryonic endothelial cells [11, 12]. Many molecular mechanisms of embryonic endothelial cell migration and assembly of new vessels remain unknown. Recapitulating events in vessel formation offers further understanding of the molecular mechanisms of vascular development, including the roles of vascular endothelial growth factor (VEGF), FGF, angiopoietins and their respective receptors, as well as EphB tyrosine kinase receptors and ephrinB ligands [13, 14].

Cell migration is an essential element of embryogenesis because cell location in the embryo is a key determinant of cell fate. The process of cell localization and tissue patterning is largely orchestrated by extracellular gradients of morphogenetic proteins (morphogens), which are diffusible substances that alter tissue structures [15]. How embryonic endothelial cells migrate and fuse with other embryonic endothelial cells to form the initial vascular network remains to be determined. Among the chemokine receptors in the mouse embryo, the CXCR4 transcript is the most abundantly detected [16]. The ligand of CXCR4, stromal cell-derived factor-1 (SDF-1 or CXCL12), is a well documented and important participant in mammalian biology [17, [18], [19]–20]. In adults, SDF-1 is a highly effective lympho-monocyte and hematopoietic stem cell (HSC) chemoattractant, and it supports CD34+ progenitor proliferation and regulates HSC homing to the bone marrow [21, 22]. SDF-1/CXCR4 signaling plays an important role in the migration and branching of human umbilical vein endothelial cells (HUVECs) [23] and influences human adult endothelial cell chemotaxis, migration, and invasion [24]. Reducing endothelial expression of SDF-1 by inflammatory cytokines, such as tumor necrosis factor (TNF)-α and interferon (IFN)-γ, results in the disruption of endothelial-cell-branching morphogenesis in vitro and in vivo [25]. A critical role of SDF-1/CXCR4 in vascular network formation during early embryogenesis was indicated by the defective formation of large vessels supplying the gastrointestinal tract in mice lacking either CXCR4 or SDF-1. The mutant mice died in utero and were defective in the formation of vasculogenesis, hematopoiesis, and cardiogenesis [26, 27]. Whether SDF-1/CXCR4 signaling is necessary for vascular development in humans remains to be determined.

Murine ES cells have been used to study molecular mechanisms of early development in vitro, including hematopoietic, endothelial, muscle, and neuronal lineages [28]. Often however, rodent models provide inconsistent evidence for human models and diseases. In this study, we used human embryonic stem cells to examine the effect of SDF-1/CXCR4 on human embryonic vascular morphogenesis. By blocking SDF-1/CXCR4 cross-talking, we found that this signaling was required for vascular network formation during human embryonic vasculogenesis.

Materials and Methods

Maintenance and Differentiation of Human Embryonic Stem Cells

The hESC lines, H1 and H9, were maintained at an undifferentiated stage on irradiated low-passage mouse embryonic fibroblast (MEF) feeder layers on 0.1% gelatin-coated plates. The medium was changed daily, and it consisted of Dulbecco's modified Eagle's medium (DMEM)/F-12 (Invitrogen, Carlsbad, CA,, 20% knockout serum replacement (Invitrogen), 0.1 mM nonessential amino acids (Invitrogen), 2 mM l-glutamine (Mediatech, Inc, Herndon, VA,, 0.1 mM β-mercaptoethanol (Sigma, St. Louis,, and 4 ng/ml rhFGF-2 (FGF-2, R&D Systems Inc., Minneapolis, or PeproTech, Rocky Hill, NJ, The undifferentiated hESCs were treated by 1 mg/ml collagenase type IV (Invitrogen) in DMEM/F12 and scraped mechanically at the day of passage, as described previously [29, 30].

To induce hESC differentiation, undifferentiated hESCs were differentiated in hESC differentiation medium containing Iscove's modified Dulbecco's medium (IMDM) and 15% defined fetal bovine serum(FBS) (HyClone, Logan, UT,, 0.1 mM nonessential amino acids, 2 mM l-glutamine, 450 μM monothioglycerol (Sigma), 50 U/ml penicillin, and 50 μg/ml streptomycin, either in ultra-low attachment plates for the formation of suspended embryoid bodies (EBs) or directly on MEF feeders. For EB formation, hESCs after growing to ∼70%–80% confluence (at ∼day 7) were treated by 2 mg/ml dispase (Invitrogen) for 15 minutes at 37°C to loosen the colonies. The colonies were then scraped off, and transferred into ultra low-attachment plates (Corning Incorporated, Corning, NY, for EB formation [6, 31]. HUVECs (Clonetics, were grown on tissue culture plates in EGM-2 medium (Clonetics) containing 5% FBS, recombinant human (rh)VEGF, basic (b)FGF, R3-IGF-1, hydrocortisone, ascorbic acid, and heparin [32].

Magnetic CD34+ Cell Selection

Single cell suspensions from day 9 to 11 of differentiated hESCs were obtained by treatment with 2 mg/ml collagenase B at 37°C for 10–20 minutes, and the cells were passed through a 40-μm cell strainer (BD Falcon, San Diego, The CD34+ cells were positively selected using MACS immunomagnetic separation system (Miltenyi Biotec, Bergisch Gladbach, Germany, After incubation with FcR blocking reagent and Hapten-antibody, the cells were labeled with anti-Hapten MicroBeads for 15 minutes at 4°C and processed through LS+ and MS+ columns (Miltenyi Biotec). More than 90% of the recovered cells expressed CD34 (BD PharMingen, San Diego,, as determined by fluorescence-activated cell sorting (FACS). To generate hESC-derived endothelial cells, the isolated CD34+ cells from differentiated hESCs were grown on 0.1% gelatin-coated plates in differentiation medium containing 50 ng/ml rhVEGF and 5 ng/ml rhFGF-2 (R&D Systems) or in EGM-2 with 5% FBS. The medium was changed every 2–3 days.

Reverse Transcription-Polymerase Chain Reaction Analysis

At different time points, the total RNAs from undifferentiated hESCs and from differentiated EBs were extracted using TRIzol (Invitrogen). One μg of RNA was used for each reverse transcription (RT) reaction. To eliminate DNA contamination, the RNA samples were treated with DNase (Invitrogen) before RT reaction (SuperScript II RNase H-Reverse Transcriptase, Invitrogen). Oligonucleotide primers used are listed in the supplemental online Table.

Flow Cytometry Analysis

The dissociated cells were washed and resuspended in phosphate-buffered saline (PBS) supplemented with 2% normal mouse serum to block nonspecific binding. For intracellular staining, the cells were permeabilized with a BD cytofix/Cytoperm solution for 20 minutes at 4°C. Direct staining of fluorochrome-conjugated anti-human monoclonal antibodies included: CD31-fluorescein isothiocyanate (FITC), CD34-allophycocyanin (APC), CD38-FITC, CD45-peridinin chlorophyll protein (PerCp), CXCR4 (CD184, Fusin)-phycoerythrin (PE) (all from BD PharMingen), Glycophorin A(CD235A, GlyA)-PE (Immunotech, Luminy, France,, VEGF receptor 2 KDR (Flk-1)-PE (R&D Systems), and AC133(CD133)-PE (Miltenyi Biotec). Unconjugated antibodies against VE-Cadherin (BD PharMingen), SDF-1 (R&D Systems) were stained by PE-labeled rat anti-mouse IgG1 (BD PharMingen) and PE-labeled donkey anti-goat IgG (Santa Cruz Biotechnology Inc., Santa Cruz, CA,, respectively. Isotype-matched controls (BD PharMingen) were used to determine the background staining. The samples were analyzed on a FACSCalibur (Becton, Dickinson and Company, Franklin Lakes, NJ, machine with CellQuest acquisition software. Data analysis was performed using CellQuest or FlowJo Software.

In Situ Immunofluorescence Staining

The undifferentiated hESC colonies were fixed in 4% paraformaldehyde/PBS for 15 minutes. Nonspecific binding was blocked with 4% normal goat serum for 30 minutes, following which the colonies were stained with antibodies to SSEA-1, SSEA-4, TRA-1-60, or TRA-1-81 (all from Chemicon International, Inc., Temecula, CA, and incubated with FITC-conjugated rat anti-mouse secondary antibodies (BD PharMingen) for 30 minutes.

For acetylated low-density lipoprotein (LDL) uptake, dil-acetylated LDL (Dil-AcLDL, Molecular Probes Inc., Eugene, OR, was diluted in differentiation medium to 10 μg/ml. hESC-derived endothelial cells were cultured in differentiation medium containing Dil-AcLDL for at least 4 hours. After washing with differentiation medium twice, the slides were fixed, mounted with mounting medium (Vector Laboratories, Burlingame, CA,, and covered with coverslips. Expression of von-Willebrand factor (vWF) in hESC-derived endothelial cells was detected by indirect staining with mouse anti-human vWF for 1 hour at room temperature, followed by PE-labeled rat anti-mouse IgG1 (all from BD PharMingen) for 30 minutes.

The sprouting EBs cultured on chamber slides were dehydrated and fixed in ethanol:methanol (1:3) for 20 minutes at 4°C. After incubation with 2% mouse serum for 20 minutes, the slides were incubated with anti-human CD31-PE or mouse IgG1-PE at 4°C overnight. The fluorescence images were obtained using either fluorescence microscopy (Nikon, Japan, or Zeiss confocal microscope system (

Three-Dimensional Sprouting EB Induction in Collagen Matrix

After 11 days of EB formation, the EBs were resuspended in 15% FCS, 100 ng/ml rhVEGF, and 100 ng/ml rhFGF-2 and mixed with an equal volume of 15% FCS and rat tail type-I collagen (BD Biosciences) in IMDM. One-hundred EBs in 1 ml of the mixture was plated in a two-well chamber slide (Nalgene Nunc, Rochester, NY,, maintained at 37°C without CO2 for 30 minutes to allow gelarization, and then incubated for 3–6 days at 37°C with humidity and 5% CO2 [33]. For the sprouting inhibition assay, 1 μg/ml AMD3100 (the antagonist of CXCR4, C28H54N8 · 8HCl · xH2O, Sigma), 15 μg/ml endostatin (Sigma) [34], or 1 μg/ml PF-4 peptide (provided by Dr. Zhongchao Han, Institute of Hematology, Chinese Academy of Medical Sciences and Peking Union of Medical College) was added to the EB suspension before being mixed with collagen medium.

Chemotaxis Assay

To investigate cell migration in response to SDF-1, a 96-well Boyden ChemoTx System (5 μm, Neuro Probe Inc., Gaithersburg, MD, was used in accordance with the manufacturer's instruction. Briefly, after growing on gelatin in the presence of VEGF (50 ng/ml) for 7–10 days, hESC-derived endothelial cells (∼3.2 × 103cells) in 25 μl of chemotaxis medium (IMDM with 0.5% BSA) were added to the upper chamber. Media (300 μl per well) with different working concentrations of SDF-1α were added to the lower chamber. After 5 hours of incubation at 37°C in 5% CO2, the cells were removed from the upper chamber, and the migrated cells present on the membrane were fixed and counted using microscopy.

Matrigel Tubular Formation

The assay was performed essentially as previously described [35, 36]. Twenty-four-well plates were coated with 200 μl per well Matrigel matrix (BD Biosciences) at room temperature for more than 30 minutes. Ten thousand to 50,000 of the endothelial cells in 200 μl of medium were replated on Matrigel-coated plates in differentiation medium at 37°C in 5% CO2. Ten ng/ml SDF-1α, 10 μg/ml neutralizing antibody against CXCR4 or isotype-matched control IgG2a was added to the culture medium. After 30 minutes of incubation, an additional 500 μl of medium with chemokine or antibody was gently added on top of the cells. The structures were photographed under phase-contrast microscopy (Nikon) after 16 hours of incubation.


Endothelial Differentiation of Human Embryonic Stem Cells

To investigate the molecular mechanisms that regulate blood vessel formation and endothelial cell migration during early development in humans, we established a protocol to induce hESCs differentiation either to form sprouting EBs or to generate hESC-derived embryonic endothelial cells (Fig. 1A). Undifferentiated hES cell lines, H1 and H9, were maintained either on irradiated MEF feeder cells or on Matrigel-coated plates in the presence of MEF-conditioned medium [1, 37]. As expected, the undifferentiated hESCs expressed markers characteristic of hESCs, such as SSEA-4 (Fig. 1A, 1B), but not SSEA-1. Concurrent with the progression of hESC differentiation, the expression of SSEA-4 was reduced, whereas the expression of SSEA-1 was increased (Fig. 1B). Upon differentiation, hESCs showed a reduction in the transcription of the pluripotent gene, Oct-4 (Fig. 2B).

Figure Figure 1..

Differentiation of hESCs. (A): Schematic outline of the differentiation procedure. Undifferentiated hESCs were grown to 70%–80% confluence on mouse embryonic fibroblasts (MEFs) or on Matrigel in MEF-conditional medium, dated as day 0. The hEBs were induced in serum-containing medium. Day-11 hEBs were collected and replated in collagen matrix for sprouting outgrowth formation. CD34+ cells, which were CD45, were isolated by magnet beads from differentiated hESCs, and cultured in VEGF-containing medium to expand and induce endothelial maturation. (B): Undifferentiated hESCs and day-12 hEB cells from H1 were analyzed for SSEA-1 and SSEA-4 expression by flow cytometry. The histogram curves were overlaid on relative isotype-matched control antibody (tint). Data are representative of three independent experiments. Abbreviations: EB, embryoid body; FGF, fibroblast growth factor; hEB, human embryoid body; hESCs, human ESCs; VEGF, vascular endothelial growth factor.

Figure Figure 2..

KDR and CD34 expression in differentiated hESCs. Human ESCs (H1) were induced to be differentiated in 15% defined fetal bovine serum. Differentiated cells were harvested at various time points by collagenase B. The cells were analyzed by flow cytometry using antibodies: KDR-phycoerythrin (PE) (A) and CD34-allophycocyanin (APC) and CD38-fluorescein isothiocyanate (FITC) (C). (B): RNA samples from H1 were isolated at different time points and analyzed by reverse transcription-polymerase chain reaction. Primers used are shown in the supplemental Table. None of the samples showed genomic DNA amplification (data not shown). Amplification of the housekeeping gene GAPDH is indicated to demonstrate equivalent amount of RNA among samples. (D): CD34+ cells were isolated from day-10 hEB cells of H1 hESCs by passage through a MACS column twice and analyzed by flow cytometry using antibodies: CD34-APC, CD31-PE, and CD45-FITC. The line curve represented the background fluorescence assessed by staining with isotype-matched antibodies. Abbreviations: bp, base pair; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; MEF, mouse embryonic fibroblast.

Previous studies showed that differentiated EBs from hESCs contain embryonic endothelial cells which can be isolated based on either CD31 (PECAM1) [7], or CD45CD31+Flk-1+VE-cad+ 6. However, immunophenotypic markers for precursors to these endothelial cells have not been determined. To examine hESC-derived endothelial differentiation, we sought to isolate embryonic endothelial progenitor cells (EPCs). KDR (VEGF receptor two or Flk-1) and CD34 are potential EPC markers, based on their expression on circulating EPCs [38, 39]. Unlike murine ES cells, in which KDR is specifically expressed on hemangioblasts but not on undifferentiated ES cells [40, 41], KDR was highly expressed on undifferentiated hESCs. FACS analysis indicated that ∼15% of hESCs expressed KDR, and the expression was increased after 12 days of EB differentiation (Fig. 2A). The KDR expression pattern was confirmed by RT-PCR analysis (Fig. 2B) and was considered suboptimal for EPC isolation from hESCs. We decided to study CD34+ cells because they contain progenitors for both hematopoietic and endothelial cells in adult human peripheral blood, bone marrow, and cord blood [39, 42]. As shown in Figure 2C, CD34 was not expressed (or minimally expressed) on undifferentiated hESCs. However, upon hESC differentiation, CD34+ cell populations increased and peaked around day 12, whereas CD38+ cells remained low throughout the same time period. CD45+ cells were increased after 20 days of EB differentiation (supplemental Fig. 1).

To examine whether hESC-derived CD34+ cells contained endothelial progenitor cells, we isolated CD34+ cells from day-10 EBs using anti-CD34 antibodies and MACS magnetic-beads. The purity of the isolated CD34+ cells was greater than 90% after double selection (Fig. 2D). In day-10 EBs, ∼60%–85% of CD34+ cells also expressed CD31 but not CD45. The isolated CD34+ cells were further expanded with endothelial cell medium, EGM-2, or differentiation medium containing VEGF and FGF-2 for 7–10 days. The resulting cells morphologically resembled HUVECs, which were uniformly flat, adherent, and stellate in appearance (supplemental Fig. 2), and the majority of the cells coexpressed CD34 and CD31 (Fig. 3A). These cells also expressed other endothelial cell markers [39, 43], such as vWF (Fig. 3B), and they were capable of Dil-AcLDL uptake (Fig. 3C). These data suggest that CD34+ cells from hESCs contained embryonic endothelial progenitor cells, and they could mature into embryonic endothelial cells. When the CD34+ cells were cultured in the presence of hematopoietic growth factors, SCF (100 ng/ml), and Flt3 ligand (100 ng/ml), they differentiated into hematopoietic cells that expressed CD45 and formed CFCs on methylcellulose (supplemental Fig. 3).

Figure Figure 3..

Properties of hESC-derived endothelial cells. H1-derived CD34+ cells were cultured in the presence of vascular endothelial growth factor (VEGF) (50 ng/ml) for 7–10 days to generate embryonic endothelial cells. (A): The majority of embryonic endothelial cells expressed both CD31 and CD34. HUVECs were used as positive control. Isotype-matched antibodies were used in flow cytometry for background fluorescence. (B): Indirect staining with anti-vWF and phycoerythrin-conjugated secondary antibody was performed on embryonic endothelial cells and HUVECs, and the staining was overlaid onto phase images (right panel). Scale bar = 10 μm. Isotype-matched antibodies served as negative control in all immunofluorescence staining (data not shown). (C): The embryonic endothelial cells (top panels), HUVEC (middle panels), and cells from CD34 cells (low panels) cultured in the presence of VEGF (50 ng/ml) were stained with DAPI (blue), and Dil-acetylated-LDL (red). Scale bar = 100 μm. Data are representative from three independent experiments. Abbreviations: DAPI, 4′,6-diamidino-2-phenylindole; EC, endothelial cell; hESC, human ESC; HUVEC, human umbilical vein endothelial cell; LDL, low-density lipoprotein; vWF, von Willebrand factor.

Expression of CXCR4 on Embryonic Endothelial Cells

Mesoderm-derived angioblasts are thought to assemble into vasculature by migration and cohesion of embryonic endothelial cells [12]. In the mouse embryo, CXCR4 and SDF-1 expression correlates with areas of vascularization [16]. We observed that the CXCR4 transcript was expressed during hESC differentiation as assessed by RT-PCR (Fig. 4A). Furthermore, both CXCR4 and SDF-1 were highly expressed in embryonic endothelial cells (Fig. 4B). Human ESC-derived embryonic endothelial cells from both H1 and H9 cell lines expressed levels of cell surface CXCR4 and intracellular SDF-1, which was comparable to HUVECs. To investigate whether VEGF regulates the expression of CXCR4 and SDF-1, we cultured hESC-derived embryonic endothelial cells in the presence or absence of VEGF for 24 hours. Although the intracellular level of SDF-1 was not modulated by VEGF, the expression of CXCR4 on hESC-derived endothelial cells was enhanced by VEGF (Fig. 4C, red line). Taken together, these data suggest a possible role for SDF-1/CXCR4 in regulating human embryonic vascular development that may be modulated by VEGF.

Figure Figure 4..

CXCR4 and SDF-1 expression in embryonic endothelial cells. (A): Reverse transcription-polymerase chain reaction (RT-PCR) analysis of CXCR4 in hEBs. RNA samples from H1 hEBs were isolated at different time points and analyzed by RT-PCR. Primers used are shown in the supplemental Table. None of the samples showed genomic DNA amplification (data not shown). (B): Embryonic endothelial cells from H1 (H1-EC) and H9 (H9-EC) were stained with anti-CXCR4-phycoerythrin (PE) or anti-SDF-1 with PE-conjugated secondary antibodies for flow cytometric analysis. Data were gated to rule out background determined by isotype-matched antibodies (insets). HUVECs were used as a positive control. (C): The expression of CXCR4 (surface, intracellular) and SDF-1 (intracellular) in embryonic endothelial cells cultured in the presence (red) or absence (blue) of vascular endothelial growth factor (VEGF) for 24 hours, were analyzed by flow cytometry. The histogram curves were overlaid on respective isotype-matched antibody curves (tint). Data were representative of three independent experiments. Abbreviations: bp, base pair; EB, embryoid body; hESC, human ESC; HUVEC, human vascular endothelial cells; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; SDF, stromal cell-derived factor.

SDF-1α Induces Embryonic Endothelial Cell Chemotaxis and Tubular Formation

To further examine the functional relationship between the expression of SDF-1 and CXCR4 on embryonic endothelial cells and their migration towards SDF-1, chemotaxis assays were performed to SDF-1 gradients. Isolated CD34+ cells from hESCs were cultured in endothelial cell media for 7–10 days, and the resulting embryonic endothelial cells were harvested for chemotaxis assays in the Boyden ChemoTx System. As shown in Figure 5A, embryonic endothelial cells migrated to the lower chamber in response to SDF-1 in a dose-dependent fashion. AMD3100 is a selective CXCR4 antagonist that blocks the interaction of CXCR4 with SDF-1. It interferes with a number of pathological or physiological processes, which include blocking the homing of bone marrow-derived stem cells and preventing VEGF and FGF induced cell invasion [44]. The addition of AMD3100 at 10 μM concentration decreased endothelial cell migration; whereas the addition of anti-VEGFR antibodies (100 ng/ml) had no significant effects on hESC-endothelial cell migration (Fig. 5B). One of the characteristics of mature endothelial cells is their ability to rapidly form capillary-like networks in Matrigel [23, 45]. We investigated whether human embryonic endothelial cells were capable of similar processes and whether CXCR4 participated in the process. Embryonic endothelial cells were loaded onto Matrigel for 16 hours in the presence or absence of SDF-1α. The addition of SDF-1 enhanced connections between embryonic endothelial cells and their ability to assemble into a reticular network (cord formation) (Fig. 5C). In the presence of neutralizing CXCR4 antibodies or AMD3100, the cord formation of embryonic endothelial cells was significantly reduced and appeared in clumps of round cells or short cords attached to the Matrigel surface (Fig. 5C and supplemental Fig. 4). To determine whether blocking CXCR4 diminished vascular network formation in Matrigel was due to a decrease in embryonic endothelial cell proliferation, we assessed in vitro proliferation in the presence and absence of neutralizing CXCR4 antibodies. As shown in Figure 5D, the addition of anti-CXCR4 antibodies to cultures of hESC-derived endothelial cells for 48 hours did not alter cell numbers in the presence of VEGF. To test whether blockage of VEGF signaling would decrease SDF-1-induced migration of embryonic endothelial cells, we added anti-VEGFR antibodies to neutralize VEGF signaling in the chemotaxis assays and Matrigel assays. Our data showed that the addition of anti-VEGFR antibodies had no significant effects on hESC-endothelial cell migration and vascular cord formation, whereas the addition of AMD3100 decreased endothelial cell migration and cord formation significantly (Fig. 5B and supplemental Fig. 4). In addition, gene expression of VEGF and KDR remained unchanged in the presence or absence of SDF-1 for 48 hours (Fig. 5E). These data provide evidence for the critical role for SDF-1/CXCR4 in regulating embryonic endothelial cell movement and vessel formation.

Figure Figure 5..

Effect of modulating the SDF-1/CXCR4 axis in hESC-derived endothelial cell mobility. CD34+ cells isolated from day-10 hEBs (H1) were cultured in the presence of VEGF (50 ng/ml) for 7–10 days to generate hESC-derived embryonic endothelial cells. (A): Chemotaxis assay of embryonic endothelial cells towards SDF-1 by Boyden ChemoTx method. After a 5-hour chemotaxis towards SDF-1, migrated cells present on the membrane were quantified as described in the Materials and Methods. Data represent the average of three experiments. Error bars represent SD. (B): Chemotaxis assay of embryonic endothelial cells towards SDF-1 (100 ng/ml) with or without AMD3100 (10 μM) or anti-VEGFR (100 ng/ml). Data represent three experiments. (C): Embryonic ECs (10,000–20,000 cells per well) were plated on Matrigel in 24-well plates and incubated for 16 hours in serum-containing medium control (upper left), with SDF-1 (10 ng/ml) (upper right), with SDF-1 and neutralizing antibody to CXCR4 (10 μg/ml) (lower right), or with SDF-1 and isotype-matched control IgG2a (10 μg/ml) (lower left). Via phase-contrast microscopy, the images demonstrated cord formation after 16 hours of incubation. Scale bar = 100 μm. (D): Embryonic endothelial cells were cultured in the presence of VEGF (50 ng/ml), and with or without anti-CXCR4 antibodies (10 μg/ml) and control IgG2a for 48 hours. The cell numbers were counted after trypsin-treatment. Error bars represent standard deviation (n = 4). (E): Reverse transcription-polymerase chain reaction analysis. RNA samples were isolated from embryonic endothelial cells cultured in serum-free medium (hES-EC), with 100 ng/ml SDF-1 (hES-EC/SDF-1), 50 ng/ml VEGF (hES-EC/VEGF), or SDF-1 and VEGF (hES-EC/SDF-1/VEGF) for 48 hours. Undifferentiated hESCs (hESC), hEB differentiated for 10 days (hEB/10d), HUVEC and K562 cells were used as control. Abbreviations: hESC, human ESC; HUVEC, human vascular endothelial cells; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; SDF, stromal cell-derived factor; VE-cad, vascular endothelial cadherin; VEGF, vascular endothelial growth factor.

Inhibition of SDF-1/CXCR4 Signaling Abolished Endothelial Sprouting

We and others previously used the sprouting EB model to investigate molecular mechanisms of vascular development [33, 34, 46, 47]. Generally, this model applies to murine ES cells, but we adapted it here to use as an in vitro model to study early vascular development in humans. Human ESC colonies were treated with dispase, and cultured in ultra low-attachment plates to form hESC EBs (hEBs). The hEBs began forming internal cystic structure around day 9 of cultivation (Fig. 6A). When the hEBs were cultured on a collagen I matrix for an additional 3 days, they formed sprouting outgrowths (Fig. 6B). The hEBs of day 11 or day 18 generated sprouting structures (supplemental Fig. 5A). To follow early developmental events, we used day-11 hEBs for our subsequent experiments. Sprouts from individual hEBs connected and formed vascular network-like structures that assimilated Dil-labeled Ac-LDL after longer culture in collagen matrix (supplemental Fig. 5B). The formation of sprouting structures was enhanced by the addition of the angiogenic growth factors, VEGF and FGF-2, whereas the addition of the angiogenic inhibitors, PF4 or endostatin, inhibited the formation of sprouting structures (Fig. 6B). These data suggest that the sprouting network-like structures were indeed endothelial in nature.

Figure Figure 6..

Endothelial sprouting outgrowths were disrupted by AMD3100. (A): Representative H1-EBs over 12 days in suspension aggregates. Internal cystic structures were formed at ∼9 days. Scale bar = 100 μm. (B): Representative sprouting hEBs. Day-11 H1-EBs were harvested and cultured for 3 days in collagen matrix in the absence of VEGF and FGF (FBS), in the presence of VEGF and FGF (VEGF/FGF), in the presence of VEGF, FGF, and endostatin (VEGF/FGF/endostatin), and in the presence of VEGF, FGF, and PF4 (VEGF/FGF/PF4). Scale bar = 100 μm. (C): Day-11 H1-EBs were harvested and cultured in collagen matrix in the presence of VEGF and FGF (control, upper panel), and with the addition of the antagonist of CXCR4, AMD3100 (AMD3100, bottom panel) for 3 days (left panel, Scale bar = 100 μm) and 6 days (right panel, Scale bar = 250 μm). Abbreviations: FBS, fetal bovine serum; FGF, fibroblast growth factor; VEGF, vascular endothelial growth factor.

To investigate the role of SDF-1/CXCR4 in embryonic endothelial outgrowth, we examined whether inhibition of SDF-1/CXCR4 signaling interrupts hEB sprouting. The addition of AMD3100 with the final concentration of 0.5 μg/ml to the sprouting hEB culture system decreased sprout density, reduced the percentage of sprouting EBs (supplemental Fig. 6), and impaired network formation (Fig. 6C).

Serial confocal photomicrographic sections of sprouting EBs after immunostaining with anti-CD31 revealed a three-dimensional organization of cells staining for CD31 (Fig. 7). The cylindrical shape and branching suggested that endothelial sprouts emerged from the initial primitive structure branched and generate an endothelial network. When the endothelial outgrowths were inhibited, nonsprouting EBs showed internal CD31+ cell clusters without branching vessel structures (Fig. 7).

Figure Figure 7..

Sprouting embryoid bodies (EBs) with channel-like vessel structures were formed in collagen sprouting assay. Channel-like structures were CD31+ by immunofluorescence staining. H1-EBs cultured in collagen matrix formed sprouting outgrowth. The human (h)EBs were dehydrated, fixed in a methanol-acetone solution, and stained with phycoerythrin-conjugated anti-CD31 as described in the Materials and Methods. Positive staining of CD31 confirmed that the sprouting structures are endothelial in origin (I). Serial images of endothelial sprouting EBs revealed spatial organization of CD31+ channel-like vessels (II–V). Nonsprouting EBs showed internal CD31+ cluster (VI). Arrows point to a vessel-like structure, which formed a lumen inside. Scale bar = 100 μm (I) or 25 μm (II–VI).


hESCs not only have enormous potential as a resource of therapeutic tissues but also provide a unique system for studying human embryonic development, including vascular development [6, 7]. A mammalian organized vascular system is established through three successive steps: (a) production of vascular endothelial cells, (b) formation of a primitive vascular plexus, and (c) remodeling to a higher order structure [48]. Evaluating the vascular development in humans is complicated, but it may be facilitated through the recapitulation of embryonic events by hESCs [1]. In this study, we used an embryonic stem cell model to demonstrate a role of the SDF-1/CXCR4 signaling in human embryonic endothelial cell migration, branching, and remodeling.

To investigate embryonic endothelial cell development, we developed a method to isolate EPCs and generate embryonic endothelial cells from hESCs. KDR, CD133, and CD34 are potential EPC markers based on their expression on human circulating EPCs [38, 39, 49]. We and others demonstrated that KDR and CD133 are expressed in undifferentiated hESCs [7, 30]. In this study, we chose CD34 as a marker of hESC-derived EPCs, because CD34 was not expressed (or only expressed at low levels) on undifferentiated hESCs. When CD34+ cells are isolated from day-10 hEBs and cultured in endothelial differentiation media (either in EGM-2 or IMDM differentiation medium with VEGF), they differentiate into embryonic endothelial cells able to form functional blood vessels in vivo (Z.Z. Wang, T. Chen, P. Au, et al., manuscript submitted). To further examine whether hESC-derived CD34+ cells are able to differentiate towards cardiomyocytes, we cultured CD34+ cells in endothelial growth medium for more than 3 weeks with weekly passage. No contracting cardiomyocytes were observed. However, the expression of endothelial markers, that is, CD31 and VE-Cad, decreased with prolonged culture, whereas the expression of the smooth muscle marker, α-SMA, increased (data not shown). It will be interesting to determine whether cardiomyocyte growth medium promotes cardiomyocytic differentiation from hESC-derived CD34+ cells.

VEGF and its cellular receptors, VEGFR1 (Flt-1) and VEGFR2 (KDR or Flk-1), were implicated in the formation of the embryonic vasculature by their colocalized expression during embryogenesis and by impaired vessel formation in Flk-1- and Flt-1-deficient embryos [50, [51], [52]–53]. During early development, FGF and VEGF secreted by the endoderm may support the differentiation of mesodermal cells to angioblasts [11]. In vitro, VEGF induces not only endothelial cell proliferation but also promotes endothelial cells to form tubule-like structures [11]. However, the molecular mechanism of VEGF regulating migratory and sprouting activity of angioblasts is largely unknown. In this study, we showed that VEGF enhanced the formation of tubular structures in a human ES cell model. Recent studies demonstrated that SDF-1/CXCR4 signaling plays a critical role in endothelial cell migration for vascular branching morphogenesis and angiogenesis [23, 54, 55]. Here, we demonstrated that SDF-1 and CXCR4 were expressed in embryonic endothelial cells and that VEGF induced CXCR4 expression. Thus, our data provide direct evidence for signaling and functional cooperation between CXCR4 and VEGF in the formation of branching tubes from embryonic endothelial cells and the migratory events that go with it. Hematopoietic and endothelial cells are derived from a common precursor, the hemangioblast, in yolk sac blood islands [56, [57]–58]. Although the properties of the hemangioblast need to be identified, recent studies demonstrated that KDR (Flk-1) provides an excellent marker for hemangioblasts during early mouse development because Flk-1+ populations give rise to hematopoietic and endothelial cells in mouse ES cells [41, 59]. However, the expression of KDR on undifferentiated hESCs raises the possibility that VEGF and its receptors play a broader role than just hemangioblast and endothelial development.

Vascularization is a complex process including the growth and the migration of angioblasts, followed by their organization into vascular tubes by a poorly understood morphogeneic process [34]. It is not clear how embryonic endothelial cells migrate and fuse with other embryonic endothelial cells to form the initial vascular network. Our in vitro morphogenesis assays, in which hESC-derived embryonic endothelial cells assembled into capillary-like structures in Matrigel, revealed that SDF-1 enhanced extracellular matrix-dependent endothelial cell branching. Inhibition of SDF-1/CXCR4 signaling interrupted hESC-derived endothelial cell migrate as well as SDF-1-induced formation of tubular structures. Salcedo et al. [54] reported that subcutaneous serial injections of SDF-1 into mouse skin induce formation of local small blood vessels and that SDF-1 treatment enhances VEGF release from HUVECs in vitro. We examined whether SDF-1 regulated VEGF and KDR expression in embryonic endothelial cells. We found that the transcripts of VEGF and KDR remained unchanged in the presence of SDF-1 for 48 hours, and blockage of CXCR4 signaling did not affect the proliferation of hESC-endothelial cells. The long-term effect of SDF-1/CXCR4 on the proliferation and survival of embryonic endothelial cells is unknown. Whether the endogenous expression of SDF-1 in hESC-endothelial cells is sufficient for recruitment of other endothelial cells to form vessels is unclear. In the sprouting EB model, the endothelial outgrowth and networks between hEBs were CXCR4-dependent. However, the addition of SDF-1 did not enhance network formation between hEBs (data not shown), suggesting that hEBs express sufficient endogenous SDF-1 to attract endothelial outgrowths from neighboring EBs. Several signaling pathways are involved in vasculogenesis and angiogenesis, including VEGF/Flk-1, Angiopointin/Tie-2, ephrins B2/EphB4, and Notch pathway [60]. Among these, EphB4 receptor tyrosine kinase and its ephrinB2 ligand regulate endothelial cell mobility by cell-cell contact [61]. In our previous study of mouse ES cells, we demonstrated that EphB4 regulates vasculogenesis and hematopoiesis at the level of an extremely primitive, common precursor population [33]. It will be interesting to examine whether other signaling pathways, such as EphB4/ephrinB2, interact with SDF-1 to promote vascular network formation in hESC model.

In summary, our data indicated a morphogenic role for SDF-1/CXCR4 in human embryonic vasculogenesis. The potential for a translational application of SDF-1/CXCR4 signaling for human neovascularization in vivo remains to be confirmed.


The authors indicate no potential conflicts of interest.


T.C. and H.B. contributed equally to this work. We thank for Dr. Zhongchao Han (Institute of Hematology, Chinese Academy of Medical Sciences and Peking Union of Medical College) for the generous gift of PF4-peptide. This study was partially supported by NIH Grants K01DK064696 and P20RR018789 (to Z.W.).