Endothelial Differentiation and Vasculogenesis Induced by Three-Dimensional Adipose-Derived Stem Cells



Recently, an angiogenic therapy based on adipose-derived stem cells (ASCs) in an ischemic model has been reported. This study demonstrates the differentiation of human ASCs (hASCs) into endothelial cells clusters by culturing the cells in the form of three dimensional cell masses (3DCMs), which is based on the adherent activity of ASCs for a substrate. The 3DCM composed of hASCs induced hypoxic conditions and expressed angiogenic factors, such as vascular endothelial growth factor and interleukin-8. From immunochemical staining analysis, the 3DCMs of hASCs were CD31+, KDR+, and CD34+, whereas monolayer-cultured hASCs were negative for the these markers. To evaluate the ability of vasculature to form within 3DCMs, the 3DCMs were mixed in Matrigel/fibrin gel and injected into mice. Mature tubular microvessels perfused with blood were observed in the 3DCM/gel 20 days after injection, but not in the gel alone or hASC/gel mixture. Vasculature formed in the 3DCM/gel was recognized by antibodies against human α-smooth muscle actin, KDR, CD31, and CD34, but not by antibodies against murine antigens. These results suggest that the vasculatures originated from the embedded human cells. The 3DCMs of hASCs could function as a source of vascular cells for neovascularization, and could also be co-implanted with other cell types for regenerative medicine. Anat Rec, 2013. © 2012 Wiley Periodicals, Inc.

Endothelial cell (EC)-based repair has been applied to vascular tissue engineering and therapeutic angiogenesis, such as in the treatment of ischemic tissues. A decade ago, a report demonstrated that CD34+ hematopoietic progenitor cells purified from circulating adult blood could be differentiated into ex vivo endothelial lineage cells, which were termed endothelial progenitor cells (EPCs) (Asahara et al., 1997; Kalka et al., 2000; Kawamoto et al., 2001). More recently, it was reported that mesenchymal stem cells (MSCs) could be differentiated into endothelial lineage cells (Oswald et al., 2004; Silva et al., 2005). MSCs are found in the stromal-vascular fraction (SVF) of subcutaneous adipose tissue, as well as bone marrow (BM) and umbilical cord blood (Zuk et al., 2001; Safford et al., 2002). A population of adherent MSCs obtained from the SVF is termed adipose-derived stem cells (ASCs). There are numerous reports that human ASCs (hASCs) can both differentiate into endothelial lineage cells in vitro and have a proangiogenic action in ischemia models (Miranville et al., 2004; Planat-Benard et al., 2004; Ning et al., 2009). Postnatal EPCs and MSCs are found only at low levels in BM and circulation, therefore, limiting their therapeutic application in regenerative medicine, such as cell therapy and tissue engineering (Kim et al., 2011). However, hASCs are regarded as an attractive cell source for therapeutic angiogenesis because large numbers are easily obtained.

In animal models of hind-limb ischemia, administration of EPCs, BM mononuclear cells, or MSCs markedly improved blood flow and capillary density (Losordo and Dimmeler, 2004). The ability of implanted stem cells to be directly incorporated into newly formed vasculature has been a matter of debate, although cord blood-derived MSCs were found in reconstituted arterioles in a mouse model of leg ischemia (Kim et al., 2006). Transplanted ASCs or ASC-derived endothelial lineage cells resulted in host angiogenesis via the secretion of pro-angiogenic factors, and were partially involved in the neovascularization in mice hind-limb ischemia and in mice soft tissue injury model (Miranville et al., 2004; Planat-Benard et al., 2004; Cao et al., 2005; Altman et al., 2009; Flynn et al., 2009). However, it was not reported whether the cells themselves formed mature vascular networks in vivo.

Tissue microenvironments consist of specific molecules, such as cytokines, cell–cell, and cell–extracellular matrix (ECM) contacts, and various physiological compositions, such as hypoxia. Cells undergo a variety of biological responses, such as proliferation, angiogenesis, and death when exposed to hypoxic conditions. Hypoxia stimulates the production of growth factors such as vascular endothelial growth factor (VEGF) that induce angiogenesis and EC survival (Tomanek and Schatteman, 2000; Calvani et al., 2006; Pilgaard et al., 2009). Stem cells have the potential to organize into vascular structures, and to secrete angiogenic factors in response to hypoxic environments (Potier et al., 2007). VEGF is a key regulator to induce physiological angiogenesis through VEGF receptors (Terman et al., 1992).

Three-dimensional spheroid cultures of tumor cells have been utilized to provide a hypoxic microenvironment for tumor angiogenesis studies. Solid tumors in the body induced hypoxia-dependent angiogenesis for tumor growth (Shweiki et al., 1995; Valcárcel et al., 2008). In this study, we sought to develop a three dimensional cell culture system to induce a hypoxic microenvironment for endothelial differentiation and in vivo vascularization, mimicking three-dimensional spheroid cultures of tumor cells. In our knowledge, this study is the first to demonstrate mature human microvessel formation from ASCs masses.


Isolation and Culture of Adipose-Derived Stem Cell

Human subcutaneous adipose tissue samples were obtained from the abdomen of seven different female donors at the ages between 35 and 54 under approval from the Catholic University of Korea Institutional Review Board. Tissue was processed to isolate hASCs as described previously (Freyberg et al., 2009; Park et al., 2009). hASCs obtained from different patients were expanded between passage 3 and 5 and randomly used in experiments.

Preparation of Cell Adhesion Substrates

Poly-[N-p-vinylbenzyl-O-α-D-glucopyranosyl-[1→4]-D-gluconamide] (PVMA) (50 μg/mL), which was synthesized as described previously (Kobayashi et al., 1994), was dissolved in ultrapure water, fibronectin (Sigma, St. Louis, MO) (20 μg/mL) was dissolved in phosphate buffered saline (PBS), and collagen Type I (Nitta Gelatin, Osaka, Japan) (20 μg/mL) was dissolved in acetic acid (pH 3) before coating the polystyrene (PS) plates. After 4 hr at 37°C, non-adsorbed polymer solutions were removed from the plates that were then washed with ultrapure water three times. The plates were treated for 1 hr at 37°C with 0.5% bovine serum albumin (BSA)/PBS to prevent non-specific cell adhesion, and then washed with ultrapure water three times.

Three Dimensional Cell Masses Formation Assay

For formation of a three dimensional cell masses (3DCM), hASCs were split and seeded on adhesion substrates in Dulbecco's modified Eagle's medium F-12 (DMEM F-12; Welgene, Daegu, Korea) with 10% fetal bovine serum (FBS; Welgene) at a density of 4 × 104 cells/cm2, and allowed to adhere at 37°C. After 3 days of culture, the 3DCM formation of adherent hASCs was observed using a phase contrast microscope (Nikon TE 2000-U, Tokyo, Japan). 3DCM sizes were measured by counting the area of individual cell clusters by image analysis. The diameters of 3DCMs were presented as median ± SD (N = 10 per group).

Reverse Transcriptase-Polymerase Chain Reaction

DNAse-treated total cellular RNA from cell cultures were prepared using TRIzol according to the manufacturer's instructions (Invitrogen life technologies, CA) and cDNA synthesis was performed in a mixture containing 5 μg RNA, oligo-(dt)12–18 primer, and reverse transcriptase (Promega, WI). The same reaction profile was used for all primers. The primer sequences are listed in Table 1. Polymerase chain reaction (PCR) was carried out for 30 cycles, each at 94°C for 45 sec, at the appropriate annealing temperature for 45 sec, and at 72°C for 1 min. The gel was visualized with ethidium bromide and photographed.

Table 1. Sequence of primers used for PCR reaction
GeneSense primerAntisense primerPCR product (bp)

Human Angiogenic Protein Analysis

For analyzing the expression profiles of angiogenesis-related proteins, we used the Human Angiogenesis Array Kit (R&D Systems, Abingdon, UK). Cell samples (5 × 106 cells) were harvested and 150 μg of protein was mixed with 15 μL of biotinylated detection antibodies. After pre-treatment, the cocktail was incubated with the array overnight at 4°C on a rocking platform. Following a wash step to remove unbound material, streptavidin-horseradish and chemiluminescent detection reagents were added sequentially. The signals on membrane film were detected by scanning on an image reader LAS-3000 (Fujifilm, Tokyo, Japan). The positive signals seen on developed film can be identified by placing the transparency overlay on the array image and aligning it with the three pairs of positive control spots in the corners of each array.

Immunofluorescence Staining

Indirect immunofluorescence staining was performed using a standard procedure. In brief, tissues and 3DCMs cryosectioned at 4 μm were fixed with 4% paraformaldehyde, blocked with 5% BSA/PBS (1 hr, 24°C), washed twice with PBS, treated with 0.1% Triton X-100/PBS for 1 min, and washed extensively in PBS. The sections were stained with specific primary antibodies and fluorescent-conjugated secondary antibodies (Supporting Information Table 1) using M.O.M kit according to the manufacturer's instructions (Vector Laboratories, Burlingame, CA). To detect transplanted human cells, sections were immunofluorescently stained with anti-human nuclear antigen (HNA, Millipore, County Cork, Ireland). Nuclei were counterstained with 4,6-diamino-2-phenylindole dihydrochloride (DAPI; Vector Laboratories). Hypoxyprobe staining followed the manufacturer's protocol (Hypoxyprobe). Stained sections were viewed with a fluorescence microscope (DXM1200F, Nikon, Japan).

Confocal Microscopic Analysis

For whole mount staining, 3DCMs and implants were fixed for 24 hr in methanol containing 10% dimethyl sulfoxide (DMSO). Samples were double stained with anti-human CD31/α-smooth muscle actin (αSMA), anti-human hypoxia-inducible factor-1α (HIF-1α)/VEGF antibodies and analyzed by the EZ-C1 laser scanning confocal microscope (Nikon) and corresponding analyses software EZ-C1 Viewer (3D-software).

HIF-1 siRNA Transfection

Cells were grown to 75% confluence in each dish and were transfected for 24 hr with either a siRNAs specific for HIF-1α or a non-targeting siRNA as a negative control (Santa Cruz Biotechnology, Santa Cruz, CA) according to the manufacturer's instructions.

Western Blot Analysis

Cultured 3DCMs were solubilized in the RIPA buffer (Pierce, Rockford, IL). Protein was quantified using the bicinchoninic acid assay (BCA assay). Equal quantities of protein were separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) using 12%, 7.5% resolving gels. After transfer to nitrocellulose membranes (Millipore), the proteins were probed with antibodies against HIF-1α (Santa Cruz Biotechnology), VEGF (Abcam), and β-actin (Transduction Laboratories, Lexington, UK) following incubation with horseradish peroxidase conjugated antibody (1:1,000, Amersham Bioscience, Piscataway, NJ) for 1 hr at room temperature. The blots were developed with an enhanced chemiluminescence detection method (Amersham Bioscience).


Human VEGF production in the 3DCMs was assayed with a commercially available ELISA kit (R&D Systems, Minneapolis, MN) according to the manufacturer's protocols. Concentrations are expressed as the amount of growth factor per 104 cells at the time of harvest.

In Vivo Vascularization Assay

The animal studies were approved by the Korea Institute of Science and Technology Animal Use and Care Committee. A gel-based angiogenesis assay was slightly modified for an in vivo vascularization analysis of stem cells (Alajati et al., 2008; Traktuev et al., 2009). Briefly, monolayer-cultured hASCs were collected by trypsin/ethylenediaminetetraacetic acid (EDTA) treatment. 3DCMs were harvested by gentle suction without trypsin and/or EDTA treatment. hASCs (1 × 106 cells) and 3DCMs (12–13 masses) were centrifuged and mixed in 500 μL Matrigel (growth factor reduced; BD Biosciences, San Jose, CA) with 2 mg/mL of fibrinogen (Greencross, Youngin, Korea). Thrombin (0.4 U; Greencross) was added to the mixture and injected subcutaneously using a 22G needle on each side lateral to the abdominal midline region into 4-weeks-old BALB/c nude mice. An equivalent number of cells were injected in both conditions. Cell was quantified by using the BCA protein assay kit (Pierce, Rockford, IL). Mice were sacrificed 20 days after implantation and gel constructs were explanted for analyses.

Scanning Electron Microscope Analysis

The interior structure of 3DCMs and implants were observed with a scanning electron microscope (SEM) (Hitachi, Tokyo, Japan) operated at 15 kV. In brief, 3DCMs and implants fixed in 4% vol/vol formaldehyde for 24 hr were dehydrated using a graded ethanol series and dried. Samples were attached to SEM stubs using carbon tabs and the stub surfaces were then coated with gold using a sputter-coater (Eiko IB3, Tokyo, Japan).


Formation of Three-Dimensional hASC Masses and Hypoxic Induction

The formation of three-dimensional cell aggregation is affected by cell–matrix adhesion strength (Friedl and Brocker, 2000; Santini et al., 2000). We investigated the possibility of 3D cell aggregation of hASCs under various culture conditions using multiple cell adhesion substrates. When hASCs were cultured on fibronectin, Type I collagen, and tissue culture plate (TCP) substrates at a high cell-seeding density in the presence or absence of FBS, the cells were fully spread and extended 4 hr after seeding (data not shown). hASCs cultured on non-tissue culture plate (NTCP) in the presence of FBS formed a sheet 12 hr after seeding and were rolled to form a cellular aggregate 24 hr after seeding (Fig. 1A). The cellular aggregate was compacted and formed a floating 3DCM 3 days after seeding. The 3DCMs were formed with a mean size of 900 μm on the NTCP substrate (Fig. 1B). Although hASCs intermittently formed spheroids on BSA and PVMA substrates, which are anti-cell adhesive surfaces, in the presence of FBS (Fig. 1B) and on the NTCP substrate in the absence of FBS (data not shown), the sizes were below 200 μm. As seen by SEM analysis, the interior structure of 3DCMs was constructed with porous networks (Fig. 1C).

Figure 1.

Three-dimensional cell mass (3DCM) formation of hASCs and hypoxia induction. A: 3DCM formation process of hASCs cultured on NTCP. Scale bar = 500 μm. B: Diameter of cells clusters formed on various substrates. C: SEM analysis of 3DCM formed on NTCP with higher magnification image of the boxed region. D: Time dependent HIF-1α mRNA expression during 3DCM formation. E: Substrate dependent HIF-1α mRNA expression.

It was reported previously that hypoxia mediates the angiogenic switch in agglomerates of tumor cells larger than 200 μm in diameter (Kerbel and Folkman, 2002). To confirm that a hypoxic environment was developed in 3DCM cultures of hASCs, we detected the mRNA expression of HIF-1α, a key regulator for hypoxia-mediated angiogenesis. HIF-1α mRNA was expressed only in 3DCMs 3 days after culture (Fig. 1D), but not in the small spheroids formed on BSA or PVMA substrates or in monolayer-cultured hASCs on TCP (Fig. 1E). These results suggest that hypoxic environment was made in 3DCM composed of hASCs under the culture condition of normoxia.

Production of Angiogenic Factors by hASCs in 3DCMs

A human angiogenesis protein array was performed to further examine the production of angiogenic factors in 3DCM 3 days after culture. As shown in Fig. 2, activin A (A-3), angiogenin (A-5), hepatocyte growth factor (HGF) (C-4), interleukin (IL)-1β (C-8), Leptin (C-11), MMP9 (D-3), and uPA (E-8) were upregulated (over twofold) in 3DCM as compared with monolayered hASCs cultured on TCP. In particular, VEGF (E-10) and IL-8 (C-9) proteins were over sixfold by 3DCMs (Fig. 2). IL-8 is an angiogenic cytokine expressed by early EPCs from BM. VEGF is a direct transcriptional target of HIF-1α (Carmeliet, 2003; Jiang et al., 2006). hASCs are differentiated into ECs by angiogenic factors such as VEGF and FGF (Cao et al., 2005). The extracellular matrix plays an important role in modulating vessel formation by changing its composition and structure to induce endothelial migration. MMP-9 and uPA are proteases that increase locally during ECM composition changes for vessel formation (John and Tuszynski, 2001; Mueller and Fusenig, 2004). These results suggest that hASCs produce various angiogenic factors in the 3DCM culture system.

Figure 2.

Expression of angiogenesis-related proteins by hASCs in 3DCMs. A: Angiogenesis-related proteins analysis of 3DCMs; an array template is shown in the lower panel, with dark gray indicating high expression. B: Quantification of angiogenesis-related proteins in 3DCMs. Abbreviations: EGF, epidermal growth factor; IL, interleukin; PDGF, platelet derived growth factor; TIMP, tissue inhibitor of matrix metalloproteinase; VEGF, vascular endothelial growth factor; GDNF, glial cell-derived neurotrophic factor; HGF, HGF; MMP, matrix metalloproteinases; uPA, urokinase plasminogen activator.

Correlation between HIF-1α and VEGF Protein Expression

To verify the protein expression of HIF-1α in 3DCMs 3 days after culture, immunofluorescence staining and western blot analysis for HIF-1α were performed. HIF-1α protein was expressed in 3DCMs, but not in monolayer cultured cells. In addition, HIF-1α protein was expressed strongly in the interior of 3DCMs, but not in the external surfaces around 3DCMs as shown in Fig. 3A, although the outer layer of 3DCMs was formed with highly dense DAPI-stained cells and VEGF-expressing cells. HIF-1α protein expression was supported by hypoxyprobe staining (Fig. 3B). The hypoxyprobe has been used to stain cells that are exposed to hypoxic conditions (Werno et al., 2010). This result indicates that oxygen diffusion was limited among these cells in the interior of 3DCMs. On the other hand, monolayer-cultured hASCs did not express HIF-1α and VEGF proteins (Fig. 3A). The transcriptional response of mammalian cells to hypoxia is largely mediated by HIF-1. HIF-1α was strongly upregulated under hypoxic conditions (Wang et al., 1995), its transcriptional activity is essential for the hypoxic induction of growth factor (VEGF, FGF2, and HGF) expression and the angiogenic signaling pathway (Pagh and Ratcliffe, 2003; Calvani et al., 2006).

Figure 3.

Correlation between HIF-1α and VEGF expression in 3DCMs. A: Confocal image of 3DCMs for HIF-1α and VEGF detection. Cryosectioned 3DCMs (NTCP) or hASCs monolayer-cultured on TCP were double stained with anti human HIF-1α and VEGF antibodies. Scale bar = 200 μm. B: Hypoxic cells in 3DCM were detected by immunofluorescence using hypoxyprobe (red). C: HIF-1α mRNA expression (72 hr) of 3DCMs transfected with siRNA-control or siRNA-HIF-1α. D: Western blot analysis of HIF-1α and VEGF expressed (72 hr) in 3DCMs transfected with siRNA-control or siRNA-HIF-1α. E: ELISA measurement of VEGF secretion in 3DCMs. Data, in triplicates, are corrected to 104 cells.

To further study the direct influence of hypoxia on VEGF release in 3DCMs, HIF-1α siRNA transfected hASCs were used to evaluate the formation of 3DCMs. HIF-1α mRNA was knocked down in hASCs transfected with HIF-1α siRNA (Fig. 3C). The effect of HIF-1α knockdown on the expression of VEGF was determined in hASCs cultured in 3DCMs. The expression of HIF-1α and VEGF proteins was monitored by Western blotting (Fig. 3D). Hypoxia strongly upregulated HIF-1α protein in 3DCMs; VEGF levels in 3DCM closely followed this pattern. Quantification of human VEGF proteins by using a commercially available ELISA kit confirmed that VEGF protein was expressed strongly in 3DCMs 3 days after seeding (Fig. 3E). Inhibition of the HIF-1α pathway by HIF-1α-specific siRNAs blocked the hypoxia-induced increase in VEGF expression. The results in Fig. 3 demonstrated that hypoxia-induced increases in VEGF synthesis were significantly blocked by HIF-1α siRNA. This suggests that HIF-1α contributes to the upregulation of VEGF in 3DCMs by hypoxia.

Endothelial Cell Differentiation of 3DCMs

Since various angiogenic factors were expressed in 3DCMs (Fig. 2), we evaluated 3DCM cells for endothelial gene expression. RT-PCR study verified their phenotype with the expression of known EC markers including CD31 (PECAM), CD34, CD144 (VE-cadherin), and E-selection (Fig. 4A). Undifferentiated cells did not express CD31, CD34, CD144, and E-selection. Parallel experiments with human umbilical vascular endothelial cells (HUVECs) served as a positive control.

Figure 4.

Endothelial phenotyping of 3DCMs. A: RT-PCR of endothelial markers; HUVECs were cultured in endothelial growth medium for being served as positive control. B: Phenotyping of expanded hASCs by immunofluorescence staining. hASCs (passage 3) were stained with CD29, CD90, and CD105 for mesenchymal stem cell identification and with KDR, CD31, CD34, and vWF for endothelial lineage cell identification. An anti Mouse IgG antibody was used as a negative control. Scale bar = 200 μm. C: Monitoring of cell surface markers by immunofluorescence staining. Cryosectioned 3DCMs were stained with anti-human CD29 (hCD29), CD34 (hCD34), CD31 (hCD29), and KDR (hKDR) antibodies. D: Confocal image analysis of 3DCM for hCD31 and hαSMA. Scale bars = 100 μm. Abbreviations: eNOS, endothelial nitroxide synthase; DAPI, 4′,6-diamidine-2′-phenylindole dihydrochloride; PECAM, platelet EC adhesion molecule.

The endothelial phenotype of 3DCM cells was also examined by immunofluorescence staining of a variety of EC surface markers. Previously, the surface markers of hASCs expanded in DMEM-F12/FBS were examined by immunofluorescence. The cells expressed CD29 (β1 integrin), CD90 (Thy-1), and CD105 (endoglin), MSC surface antigens, whereas they did not express CD34, CD31, KDR, and vWF (Fig. 4B). It is well known that undifferentiated hASCs are positive for CD29, CD90, and CD105 (Lee et al., 2004). This indicates that cells used in this study include a large population of hASCs without endothelial lineage cells contamination. 3DCMs were recognized by human CD29, CD34, CD31, and KDR (VEGF receptor) antibodies (Fig. 4C,D), whereas they were not recognized by human osteopontin (osteoblast marker), nestin, and MP (neuronal marker) antibodies (data not shown) and hASCs monolayer-cultured on TCP were not recognized by human CD34, CD31, KDR, osteopontin, and nestin antibodies (data not shown). The endothelial markers (CD34, CD31, and KDR) were expressed in 3DCMs 3 days after culture, but not in 3DCMs 1 day after culture (data not shown). These results are correlated to the time dependency of HIF-1α expression shown in Fig. 1C. In addition, 3DCMs were partially recognized by human αSMA, a cytoskeletal protein expressed in smooth muscle cells and pericytes, antibody (Fig. 4C), whereas hASCs monolayer-cultured on TCP were rarely recognized by the antibody. Confocal microscopy further revealed that the cell population with the CD31+ endothelial phenotype was widely distributed throughout 3DCMs. The cell population with the αSMA+ was distributed on the surface of 3DCMs (Fig. 4D). ASCs generally express some smooth muscle markers (Rodríguez et al., 2006). Also, a recent study demonstrated that SVF cells express both pericyte and mesenchymal markers (Stashower et al., 1999). Several studies suggest that vascular pericytes or solitary SMCs, which are progenitors of SMCs involved in blood vessel formation, have phenotypic and physiological characteristics similar to MSCs (Short et al., 2003). Therefore, 3DCMs may be composed of EC-lineage cells or SMC-lineage cells. In addition, confocal three-dimensional image assembly showed that 3DCMs had a porous network structure, correlating with the SEM images shown in Fig. 1C.

Numerous studies have reported that hypoxia induces the production of growth factors correlated with EC growth and function (Calvani et al., 2006; Potier et al., 2007; Pilgaard et al., 2009). In practice, hASCs can differentiate into functional ECs in vitro in the presence of angiogenic factors such as VEGF and FGF (Cao et al., 2005). Taken together, it is expected that the endothelial differentiation of hASCs in 3DCM might be upregulated by hypoxia-induced angiogenic factors. In 2D-cultures, cytokines secreted from hASCs are released and diluted in the excess volume of media, preventing the cells from responding to the released cytokines. On the other hand, in 3D-cultures, if cytokines were secreted from cells during or after cluster formation, the cytokines would be stored in the cluster and easily utilized by themselves or neighboring cells through the pore networks in an autocrine or paracrine manner.

In Vivo Human Vasculature Formation within 3DCMs in Mice

We carried out a gel-based in vivo assay to examine the vasculogenic activity of 3DCMs (Fig. 5). Figure 5A illustrates the appearance of gel mixtures dissected from nude mice. Dissected cell-free Matrigel/fibrin implants, which served as a negative control, were translucent. Monolayer cultured hASC- or 3DCM-embedded Matrigel/fibrin implants were opaque, indicating that cells or tissues might have grown in the gel. In particular, 3DCM-embedded implants were tinged with red, suggesting the formation of blood vessels in the gel. In SEM analysis, intact tubular structures were observed in the cross-sectioned 3DCM-embeded gel. We observed larger human vessels (40–50 μm in diameter) as well as on small capillaries (>10 μm in diameter) in Matrigel in vivo. Additionally, erythrocytes were identified in cross-sectioned tubular microvessels (Fig. 5B–E). These results indicate that the outgrowing human vasculature formed anastomoses with the mouse vasculature and perfused with mouse blood (as evidenced by the presence of erythrocytes).

Figure 5.

In vivo vascularization assay. A: PBS, hASCs (1 × 106 cells), and 3DCMs (12–13 masses) were mixed with Matrigel/fibrin (500 μL) and subcutaneously injected into the abdominal flanks. A small bulge is detectable in the skin. The injected gel mixtures were explanted from the transplanted region after 20 days. B–E: SEM images of blood vessels sectioned in the center of the grafted plug; longitudinal-sectioned microvessel (B) and cross-sectioned microvessel (C). The boxed region in (B) and the arrows in (C) indicate erythrocytes which are magnified in (D) and (E), respectively. Scale bar = 50 μm (B, C), 6 μm (D, E). F: Immunofluorescence staining of implants in mice; gel implants (control), hASCs-embedded gel implants (hASC), and 3DCMs-embedded gel implants were stained with anti hCD31, hCD34, hKDR, hαSMA, human nuclear antigen (HNA), and mouse CD31 (mCD31) antibodies. G: Three-dimensional structure of the humanized vascular networks by confocal microscopy. Scale bar = 100 μm.

In the hASC-embedded gel implants, there were few cells that were stained by human CD31 and αSMA antibodies, although many DAPI-stained cells were identified (Fig. 5F). In the 3DCM-embedded gel implants, many cells reacted with human α-SMA, CD31, CD34, KDR, and vWF antibodies (Fig. 5F). Additionally, the positive cells appeared to form vascular networks. To identify mouse-originated vasculatures, various regions of dissected gels were double-stained with antibodies against mouse or human CD31. There were no positive signals for the anti-mouse CD31 antibody. This indicates that host-derived cells were not incorporated into the vascular network structure formed in the gel. However, the vasculature was stained with HNA. Besides, the human antibodies used in this experiment did not cross-react with the vessels of nude mice (data not shown). To further confirm the structure of the vascular networks stained with antibodies against human CD31 and α-SMA, confocal microscopy was used to obtain 3D images at depths up to 60 μm from the middle of specimen. Confocal 3D image assembly revealed a dense network of human vessels of various calibers fused with each other (Fig. 5G). These results indicate that the vascular network structure in the 3DCM-embeded gel originated from human cells.

To our knowledge, this is the first report describing the mature human microvessel formation of expanded hASCs. EC-lineage stem cell-derived EPCs and stem cells were previously examined for therapeutic angiogenesis for the treatment of cardiac infarction and ischemia. Mostly single cells were injected into the target site. The injected cells expressed angiogenic factors and induced host angiogenesis through a paracrine effect or were partially incorporated during angiogenesis. In particular, in cardiac infarction treatment, injected cells affected the early state; however, they did not induce angiogenesis or vasculogenesis. Most cells die within the hypoxic and low-mass transfer microenvironment of cardiac infarctions (Qutub and Popel, 2008). In our studies, 3DCMs directly formed vasculature perfused by host blood within the gel where host cells were not present. Interestingly, 3DCMs formed microvessels in vivo independently of the host cell. How the cells formed vasculature and fused to host vessels is the focus of future studies. However, injection of 3DCMs shows advantages for vascularization within poor physiological microenvironments, such as cardiac infarction. It remains unclear whether the endothelial differentiation of hASCs was due to the formation of 3DCMs or to the hypoxic environment. Nevertheless, we emphasize the significance of the 3DCM culture for endothelial differentiation. The transplantation of 3D cell aggregates provided better results than that of single cells in in vivo experiments.

It has been reported that HUVECs can be implanted as spheroids into a matrix to give rise to a complex three-dimensional network of human neovessels in mice (Alajati et al., 2008). Their angiogenesis prevention and intervention experiments validated the assay for antiangiogenic drug screening purposes and demonstrated its usefulness in relating experimental findings to individual steps of the angiogenic cascade. We believe that 3DCM allows for the study of human vascularization processes in a murine and may lead to drug discovery related to angiogenesis. During the past decade, numerous studies have provided preclinical data on the safety and efficacy of hASCs, supporting the use of these cells in future clinical applications (Zuk et al., 2001; Gimble and Guilak, 2003). Various clinical trials have shown the regenerative capability of hASCs in subspecialties of medical fields such as plastic surgery, cardiac surgery, and orthopedic surgery (Lindroos et al., 2011). Several experimental strategies for improving the engraftment of stem cells in ischemic tissue have been developed, including transplantation in combination with growth factor delivery, genetic modification of stem cells, and cell spheroid culture. There are numerous reports that spheroids of stem cells have improved therapeutic efficacy due to enhanced cell survival and retention in several different forms of ischemia (Valcárcel M et al., 2008; Bhang et al., 2011). We suggest that 3DCM is a promising cell source for therapeutic angiogenesis, although the safety of 3DCM should thoroughly be examined prior to extensive use in clinical applications.


We hypothesized that 3D stem cell cultures might lead to angiogenic differentiation of hASCs in vitro due to the hypoxic microenvironment. Therefore, we developed a substrate-dependent 3D hASC culture method to examine hypoxia-dependent angiogenic factor expression and endothelial differentiation of cells. hASCs can be safely harvested in large quantities from fat tissue without ethical drawbacks. Therefore, this culture system provides the possibility of using 3DCM of hASCs as a source of vascular cells for vascular tissue engineering, as well as therapeutic neovascularization.