Human Bone Marrow-Derived Mesenchymal Stem Cells Suppress Human Glioma Growth Through Inhibition of Angiogenesis§


  • Ivy A.W. Ho,

    1. Laboratory of Cancer Gene TherapyHumphrey Oei Institute of Cancer Research, National Cancer Center, Singapore
    Search for more papers by this author
  • Han C Toh,

    1. Laboratory of Cell Therapy and Cancer VaccineHumphrey Oei Institute of Cancer Research, National Cancer Center, Singapore
    Search for more papers by this author
  • Wai H. Ng,

    1. Department of Neurosurgery, National Neuroscience Institute, Singapore
    Search for more papers by this author
  • Yuan L. Teo,

    1. Laboratory of Cancer Gene TherapyHumphrey Oei Institute of Cancer Research, National Cancer Center, Singapore
    2. National Institute of Education, Singapore
    Search for more papers by this author
  • Chang M. Guo,

    1. Department of Orthopedic, Singapore General Hospital, Singapore
    Search for more papers by this author
  • Kam M. Hui,

    1. Bek Chai Heah Laboratory of Cancer Genomics, Humphrey Oei Institute of Cancer Research, National Cancer Centre, Singapore
    2. Cancer and Stem Cells Biology Program, Duke-NUS Graduate Medical School, Singapore
    Search for more papers by this author
  • Paula Y.P. Lam

    Corresponding author
    1. Laboratory of Cancer Gene TherapyHumphrey Oei Institute of Cancer Research, National Cancer Center, Singapore
    2. Department of Physiology, National University of Singapore, Singapore
    3. Cancer and Stem Cells Biology Program, Duke-NUS Graduate Medical School, Singapore
    • Laboratory of Cancer Gene Therapy, Division of Cellular and Molecular Research, Humphrey Oei Institute of Cancer Research, National Cancer Centre, 11, Hospital Drive, Singapore 169610, Singapore

    Search for more papers by this author
    • Telephone: 65-64368357; Fax: 65-62265694

  • Author contributions: I.A.W.H.: conception and design, data analysis and interpretation, and manuscript writing; H.C.T., C.M.G., and W.H.N: provision of study material; Y.L.T.: collection and assembly of data; K.M.H.: administrative support; P.Y.P.L.: conception and design, data analysis and interpretation, manuscript writing, financial support, and final approval of manuscript.

  • Disclosure of potential conflicts of interest is found at the end of this article.

  • §

    First published online in STEM CELLSEXPRESS October 3, 2012.


Tumor tropism of human bone marrow-derived mesenchymal stem cells (MSC) has been exploited for the delivery of therapeutic genes for anticancer therapy. However, the exact contribution of these cells in the tumor microenvironment remains unknown. In this study, we examined the biological effect of MSC on tumor cells. The results showed that MSC inhibited the growth of human glioma cell lines and patient-derived primary glioma cells in vitro. Coadministration of MSC and glioma cells resulted in significant reduction in tumor volume and vascular density, which was not observed when glioma was injected with immortalized normal human astrocytes. Using endothelial progenitor cells (EPC) from healthy donors and HUVEC endothelial cells, the extent of EPC recruitment and capacity to form endothelial tubes was significantly impaired in conditioned media derived from MSC/glioma coculture, suggesting that MSC suppressed tumor angiogenesis through the release of antiangiogenic factors. Further studies using antibody array showed reduced expression of platelet-derived growth factor (PDGF)-BB and interleukin (IL)-1β in MSC/glioma coculture when compared with controls. In MSC/glioma coculture, PDGF-BB mRNA and the corresponding proteins (soluble and membrane bound forms) as well as the receptors were found to be significantly downregulated when compared with that of glioma cocultured with normal human astrocytes or glioma monoculture. Furthermore, IL-1β, phosphorylated Akt, and cathepsin B proteins were also reduced in MSC/glioma. Taken together, these data indicated that the antitumor effect of MSC may be mediated through downregulation of PDGF/PDGFR axis, which is known to play a key role in glioma angiogenesis. STEM Cells2013;31:146–155


Human mesenchymal stem cells (MSC) are a population of nonhematopoeitic stem cells in the bone marrow microenvironment that have the ability to self-renew and differentiate into multiple lineages. MSC are increasingly being developed for potential clinical usage due to the ease of isolation from adult donors, thus obviating the ethical issues that plague embryonic stem cell research [1]. Because of its strong chemotactic response to injury, MSC have emerged as an attractive homing vehicle for the delivery of therapeutic genes to tumors and diseased tissues. MSC secrete a variety of cytokines that have both paracrine and autocrine functions in the tumor milieu. These include suppression of local immune response, inhibition of fibrosis and apoptosis, modulation of angiogenesis, and stimulation of mitosis and differentiation of cells [2]. MSC could also cause a direct effect through intercellular signaling via physical contacts with neighboring cells, which may consist of tumor cells, stromal fibroblast, infiltrating immune cells, blood, and lymphatic networks that ultimately determine the fate of the tumor growth kinetics.

Angiogenesis, the formation of new blood vessels from pre-existing vessels, is essential for normal development and maintenance of homeostasis. Tumor angiogenesis is also crucial for tumor growth; a tumor cannot grow beyond the size of 2–3 mm3 without blood supply [3]. Tumor angiogenesis could arise from the recruitment of existing endothelial cells (EC) or endothelial progenitor cells (EPC) in response to proangiogenic factors. In the latter example, vascular endothelial growth factor (VEGF) [4], fibroblast growth factor (FGF) [4], granulocyte macrophage-colony stimulating factor [5], platelet derived growth factor (PDGF) [6], hypoxia-induced macrophage migration inhibitory factor [7] have been shown to mobilize EPC. Alternatively, cytokine-stimulated MSC [8] and MSC that have undergone transdifferentiation into EC or smooth muscles cells [9–11] have been shown to participate in tumor angiogenesis and growth.

Having said this, MSC are known for their unique biological property of expressing low levels of major histocompatibility complex antigens, which may contribute to their low immunogenicity and anti-inflammation effect when introduced exogenously [2]. Using chemically burned corneas in rat as a model, Oh et al. [12] showed that MSC mediated an anti-inflammatory and antiangiogenic effects on the injured corneal surface. Secchiero et al. [13] demonstrated that intraperitoneal injection of MSC improved the survival of lymphoma-bearing mice through induction of EC apoptosis. Murine MSC was observed to intercalate into the capillary networks of EC and caused capillary degeneration in a MSC/EC coculture system in vitro and abrogated melanoma growth in vivo [14]. More recently, the antitumor effect of cord blood-derived MSC was also demonstrated by Dasari et al. [15] in a glioma tumor model. Given the numerous conflicting reports on the ability of MSC to support tumor growth, it is important to further explore the role of MSC in the tumor microenvironment.

In this study, we demonstrated that MSC elicited an antitumor effect in patient-derived primary glioma cells as well as in human glioma xenograft model. The antitumor effect is mediated through the secretion of soluble factors that inhibit EPC recruitment and impaired tumor angiogenesis. Using antibody array, we have identified and subsequently confirmed that the antiangiogenic effect of MSC might be mediated through the downregulation of PDGF/PDGF Receptor (PDGFR) axis, which is known to play a key role in glioma angiogenesis


Cell Culture

Human MSC were isolated, cultured, and characterized as previously described [16]. The culturing of human glioma cells ΔGli36 was described previously [16]. Normal human astrocytes (NHA) was purchased from Lonza (Basel, Switzerland, and cultured in astrocyte basal medium (ABM) supplemented with recombinant human EGF, insulin, ascorbic acid, gentamycin sulfate, amphotericin, L-glutamine, and fetal bovine serum (FBS) as recommended by the supplier. Immortalized NHAs that overexpress E6, E7, and human telomerase reverse transcriptase (hTERT) were kindly provided by R.O. Pieper (University of California, San Francisco, CA). Human ECs, Human umbilical vein endothelial cells (HUVEC), was purchased from Lonza and cultured in endothelial growth medium supplemented with recombinant epidermal growth factor (EGF), hydrocortisone, VEGF, basic FGF, insulin growth factor (IGF), heparin, ascorbic acid, gentamycin sulfate, amphotericin-B, and 2% FBS.

Isolation of Primary Human Glioma Cells

This study has been approved by the SingHealth Centralized Institutional Review Board, Singapore. Primary human glioma cells were isolated, after informed consent, from the brain tumor tissues of patients undergoing brain tumor surgery at the National Neuroscience Institute, Singapore. The harvested tissue was separated into small pieces in the presence of complete medium (ABM) supplemented with 10% FBS, penicillin/streptomycin, normocin, and L-glucose (Cambrex Bio Science Walkersville, Inc., Walkersville, MD, The tissue suspensions were first passed through a 5 ml serological pipette, followed by a 1 ml pipette and finally a flame-polished pasteur pipette until no clumps were visible. Following trypsin digestion, the homogenate was filtered through a 70-μm cell strainer (BD Biosciences, San Jose, CA, and then subjected to centrifugation. The collected cells were cultured in complete ABM.

Isolation and Characterization of EPC

EPC was isolated according to Brunt et al. [17]. Briefly, peripheral blood mononuclear cells were isolated by centrifugation (400g for 20 minutes) in the presence of Histopaque-1077 (1.077 g/ml, Sigma, After centrifugation, the buffy coat was isolated, rinsed twice with phosphate buffered saline (PBS), and resuspended in endothelial growth medium (EGM-2MV) containing 5% serum, VEGF, FGF-2, EGF, IGF, hydrocortisone, and ascorbic acid. Cells were cultured at 37°C with 5% CO2 in a humidified incubator for 3 days. Nonadherent cells were removed after 3 days, and adherent cells were cultured for a further 10–20 days until colonies appeared.

EPC was characterized by the expression of CD133, CD34, Tie2, vascular endothelial (VE)-cadherin (CD144), VEGF receptor (VEGFR)-2 and for the uptake of acetylated-low density lipoprotein (Ac-LDL) [17, 18]. Briefly, DiI-labeled acetylated-LDL (DiI-Ac-LDL; 10 μg/ml; Invitrogen, Carlsbad, CA, was added to live cells and incubated at 37°C for 1 hour. Cells were rinsed twice with PBS to remove excess dye. The number of DiI-Ac-LDL positive cells was counted under a wide-field fluorescence microscope equipped with a CCD camera (Nikon Eclipse 90i; Nikon, Tokyo, Japan,

Flow Cytometry

For quantification of the glioma cells in the coculture experiments, carboxyfluorescein diacetate N-succinimidyl ester (CFDA-SE)-labeled-ΔGli36 glioma cells were coincubated with carboxymethyl-DiI-labeled MSC for 48 hours. Labeling of the cells was described previously [16]. Cells were detached with 0.25% trypsin/EDTA and washed twice with PBS. Cell pellet was then resuspended in 1 ml of PBS. Fluorescence-activated cell sorting analysis was performed using Cell Quest.

In Vivo Tumor Model

To establish the subcutaneous tumor model for monitoring the effect of MSC on glioma growth, immunodeficient BALB/c-nu/nu mice (Animal Resource Center, Canningvale, Western Australia, were inoculated with equal ratio of ΔGli36 and carboxymethyl-DiI-labeled-MSC in the presence of Matrigel (BD Biosciences). Tumor volume was calculated using the formula: tumor volume (mm3) = 0.52 × (width [mm2] × (length [mm]) [19]. All animal experiments were performed according to the guidelines and protocols approved by the Institutional Animal Care and Use Committee at the Singapore General Hospital, Singapore.

Immunofluorescence and Immunohistochemistry Staining

Slides were visualized using wide-field microscopy using an upright microscope (Eclipse 90i; Nikon), and images were acquired on a CCD color digital camera (DS-U2; Nikon) using NIS Element AR v2.3 software (Nikon). To assess the cell localization with higher resolution, images were obtained with a confocal system (LSM 510 Meta; Carl Zeiss, Göttingen, Germany, Images were obtained using either a 40×/0.75 numerical aperture (N.A.) Plan-Neofluar or a 63×/1.25 N.A. Plan-Neofluar oil immersion objective (Carl Zeiss). The following antibodies were used for the various staining: CD31 (1 μg/ml; BD Biosciences), cathepsin B (5 μg/ml; R&D Systems, Minneapolis, MN,, single-stranded DNA (ssDNA; 1 μg/ml; clone F7-26, Millipore, Temecula, CA,, α-smooth muscle actin (SMA; 1:100 dilution; Sigma Aldrich, St. Louis, MO,, CD34 (1 μg/ml), CD133 (1:100 dilution), Tie2 (1 μg/ml), VE-cadherin (1 μg/ml) and VEGFR-2 (1 μg/ml) (Santa Cruz Biotechnology, Santa Cruz, CA,, and CD45 (1:100 dilution; BD Biosciences). Mouse IgG1 (Dako Denmark A/S, Glostrup, Denmark, and rabbit serum (Dako Denmark A/S) were used as isotype control.

Mice were sacrificed 21 days post-tumor implantation and immediately perfused through the heart with ice-cold saline. Tumors were harvested and kept in 4% paraformaldehyde (PFA) for 4 hours at 4°C, transferred to 30% sucrose in PBS overnight at 4°C, and embedded in Jung Tissue Freezing Medium (Leica Microsystems Nussloch, Heidelberg, Germany,, snap-frozen in isopentane immersed in liquid nitrogen, and stored at −80°C.

For immunofluorescence staining, frozen sections were fixed with ice-cold acetone or 4% PFA, blocked for 30 minutes with 10% goat serum, and incubated for 1 hour at room temperature with primary antibodies. To assess the pericyte coverage of microvessels, tumor sections were double-stained for α-SMA expression to detect pericytes and CD31 (1 μg/ml) to detect ECs. To determine the microvessels density (MVD), the number of CD31+ cells was counted. To express MVD counts microscope-independent, counts were transformed and expressed as the number of microvessels per mm2.

For immunohistochemistry staining, following incubation with primary antibodies, sections were stained with goat anti-mouse or anti-rabbit polymer, followed by detection using 3,3′ diaminobenzidine substrate solution (Dako Denmark A/S). Antibody detection in sections stained with mouse monoclonal antibodies was performed using the Zenon mouse IgG labeling kit (Invitrogen). Sections were then washed and coverslipped in mounting medium. Parallel samples were incubated with no primary antibody or an isotype as a specificity control.

Immunoblot Analysis

Tumor lysates (ΔGli36, ΔGli36/iNHA, and ΔGli36/MSC) (50 μg) were harvested for immunoblot analysis. The following antibodies were used: total and activated PDGFR-β (1:1,000 dilution; Cell Signaling Technology, Beverly, MA,, total and activated Akt (1:1,000 dilution; Cell Signaling Technology), β-tubulin (1:5,000 dilution; Sigma Aldrich), and interleukin (IL)-1R1 (1:100 dilution; Epitomics, Burlingame, CA, Chemiluminescence detection was performed using Western Lightning chemiluminescence kit (Perkin Elmer, Waltham, MA, The band intensity of specific proteins was quantified after normalization with the density of β-tubulin with MetaVue imaging software (version 6.1; Sunnyvale, CA).

Cytokine Array Analysis

RayBio Human Cytokine Array C2000 (RayBiotech, Inc., Norcross, GA, that contains antibodies targeting to 174 proteins were used to detect proteins that are differentially expressed in ΔGli36/MSC-CM in comparison to that of ΔGli36-CM and MSC-CM. One hundred microgram of conditioned media (CM) was used according to manufacturer's instructions. Signals were detected by chemiluminescence reaction and quantified using Quantity-One software (Bio-Rad Laboratories, Hercules, CA,

ECs Tube Formation Assay

Tube formation assay was performed according to manufacturer's instructions. Briefly, on the day of the assay, 10 μl of EC matrix (Millipore) was added to the ibidi μ-Slide angiogenesis dish (ibidi, Munich, Germany, and incubated for 1 hour at 37°C with 5% CO2. During the incubation time, HUVEC was harvested. The cells were counted and 50 μl of cells suspension containing 5,000 cells was added to the wells containing the EC matrix. The slides were then incubated at 37°C for 6 hours. The tubes formed are visualized and captured using the Zeiss 200M phase contrast microscope at ×50 original magnification (Carl Zeiss). Tube formation was quantified after 6 hours by measuring the total tube length using ImageJ software (National Institutes of Health, Bethesda, MD, For checking the effect of the CM on HUVEC tube formation abilities, cells were resuspended in the respective medium, namely, ΔGli36-CM, MSC-CM, NHA-CM, ΔGli36/NHA-CM, and ΔGli36/MSC-CM.

Transwell EPC Recruitment Assay

For determining the recruitment potential of EPC, EPC (1 × 104) were cultured in EBM-2MV media with 0.5% FBS in a 24-well tissue culture insert with a 8-μm pore size membrane (BD Biosciences). CM from ΔGli36 human glioma cells (ΔGli36-CM), MSC (MSC-CM), or ΔGli36/MSC (ΔGli36/MSC-CM) coculture were added to the bottom well. After 6 hours, the filter membrane was fixed with 4% PFA, and cells were mounted in mounting medium containing propidium iodide (PI, 100 μg/ml; Sigma-Aldrich) containing RNase A. Migration of EPC was determined by counting the number of PI-stained nuclei on the underside of the membrane under ×200 magnification.

Quantitative Real Time Polymerase Chain Reaction and ELISA

Quantitative real time polymerase chain reaction (qPCR) was performed as described previously [20]. Primers used are listed in supporting information Table S1. Primers for the amplification of PDGF-BB and IL-1β were purchased from Qiagen (QuantiTech Primer Assay). The expression level of PDGF-BB, IGF-1, FGF-2, and IL-1β was quantified using QuantiTech SYBR Green PCR kit (Qiagen, Hilden, Germany, All qPCR reactions were performed in duplicate. Standard curves for each gene were generated independently. The relative copy number of each sample was calculated according to the corresponding standard curve using RotorGene software version 6.0. The relative expression levels were calculated by arbitrarily designating the lowest normalized value to 1. Quantification of PDGF-BB, IGF-1, FGF-2, and IL-1β protein expression was performed using Quantikine human ELISA kit (R&D Systems) according to manufacturer's suggestions.

Statistical Analysis

Statistical analysis was performed using Prism 3.0 (Graphpad Software Inc., San Diego, CA, One-way ANOVA followed by Bonferroni multiple comparisons test were used for comparing statistical significance for more than two groups. Two-way ANOVA was used for comparing statistical significance for more than two factors. p value <.05 was considered statistically significant.


MSC Induced Cell Death in Glioma Cells in an In Vitro and In Vivo System of Coculture

To evaluate the influence of MSC on the growth of human glioma cells, CFDA-SE-labeled MSC (green fluorescence) was cocultured with human glioma ΔGli36 cells prelabeled with carboxymethyl-DiI (red fluorescence) at equal ratio for 48 hours. Using flow cytometry analysis, we detected a 69.4% reduction in the number of glioma cells in ΔGli36/MSC coculture in comparison to that of ΔGli36/NHA coculture (ΔGli36/NHA; Fig. 1A). Similar findings were observed in primary glioma cells derived from human tumor biopsy materials (denoted as NNI32 henceforth). In comparison to NNI32/immortalized NHA (iNHA) coculture, the percentage of NNI32 was greatly reduced by 52.3% and 56.6% when the cells were cocultured with two independent MSC isolates, MSC-1 and MSC-8, respectively (Fig. 1B), suggesting that MSC exert an inhibitory effect on glioma cells.

Figure 1.

Coculture of MSC with glioma cell lines or patient-derived glioma cells induced tumor cell death in vitro. (A): Equal ratio of CFDA-SE-labeled-ΔGli36 glioma cells were cocultured with either carboxymethyl-DiI-labeled-MSC or carboxymethyl-DiI-labeled-NHA. Flow cytometry FL1-H and FL2-H analysis of the ΔGli36/MSC coculture demonstrating the separation of the two populations by color. Total number of CFDA-SE-labeled-ΔGli36 human glioma cells was counted using fluorescence-activated cell sorting analysis after 48 hours. (B): Equal ratio of patient-derived glioma cells NNI32 were cocultured with either carboxymethyl-DiI-labeled-MSC (MSC-1 and MSC-8) or carboxymethyl-DiI-labeled-iNHA . Images were captured from 10 random fields under ×100 original magnification using Nikon TE300 wide-field microscope equipped with a CCD camera after 48 hours. Total number of NNI32 glioma cells was counted using NIS Elements version 3.0. Data shown are averages of triplicates ± SEM. Abbreviations: CFDA-SE, carboxyfluorescein diacetate N-succinimidyl ester; iNHA, immortalized NHA; MSC, mesenchymal stem cell; NHA, normal human astrocytes. **, p < 0.01; ***, p < 0.0001.

Because in vitro coculture conditions are not representative of the tumor microenvironment, we performed in vivo coculture experiment to monitor the effect of MSC on the tumorigenicity of ΔGli36 glioma cells. For ease of monitoring changes in the tumor volume, both cell types were implanted in equal ratio subcutaneously. The tumor volumes and tumor weights in mice injected with equal ratio of MSC and ΔGli36 were significantly smaller than control animals injected with either ΔGli36 cells or ΔGli36/iNHA coculture (Fig. 2A, 2B, respectively). This phenomenon was consistently observed in two isolates of MSC, thus indicated that the antitumor effect was not pertaining to a particular MSC isolates. Two-way ANOVA analysis indicated that the differences among the six groups (ΔGli36, iNHA, MSC, ΔGli36/iNHA, ΔGli36/MSC-1, and ΔGli36/MSC-7) were statistically significant (p < .0001; Fig. 2A). By contrast, there was no statistically significant difference in the tumor volumes between ΔGli36/MSC-1 and ΔGli36/MSC-7 groups. It is important to note that none of the MSC isolates became tumorigenic; tumor nodule was not detectable in mice injected with the same MSC cell number in the presence of Matrigel up to 1 month. Gross anatomy of the tumors clearly showed that ΔGli36/MSC tumors were visibly less vascularized than those of the ΔGli36 control tumors (Fig. 2C). Hematoxylin and eosin (H&E) staining performed on the cocultured tumor sections revealed a peripheral rim of tumor cells cuffing a zone of necrosis (N) (Fig. 2C). Furthermore, analysis using antibodies against ssDNA indicated that the regions of the apoptotic cells in the cocultured tumor (Fig. 2E) coincided with the localization of the carboxymethyl-DiI-labeled-MSC (Fig. 2D, inset (I)), thus suggesting that the presence of MSC causes cell death in the glioma. On the contrary, apoptosis was not detected in ΔGli36 tumor and ΔGli36/iNHA tumor (Fig. 2E; supporting information Fig. S1). Taken together, the results demonstrated that specific interactions between MSC and ΔGli36 human glioma cells induced apoptosis of the glioma cells.

Figure 2.

Antitumor effect of MSC in glioma/MSC coculture in vivo. (A): ΔGli36 cells (5 × 105), MSC (5 × 105), iNHA (5 × 105), ΔGli36/iNHA (5 × 105:5 × 105), ΔGli36/MSC-1 (5 × 105:5 × 105), and ΔGli36/MSC-7 (5 × 105:5 × 105) were implanted into the flank of nude mice in the presence of Matrigel. Tumor volume was monitored at various time points. Data shown are averages of seven mice ± SEM. *, p < .05; **, p < .01 (B): Tumor weight comparison between ΔGli36 and ΔGli36/MSC-1 tumors. (C): Photomicrograph of ΔGli36 and ΔGli36/MSC tumors. H&E staining was performed on the tumor sections. Blue arrow indicated the presence of pockets of ΔGli36 tumor cells. Scale bar = 100 μm. (D): Photomicrograph of ΔGli36 tumor coinjected with carboxymethyl-DiI-labeled-MSC. Sections were counterstained with DAPI nuclear stain; red color indicates the presence of carboxymethyl-DiI-labeled-MSC. Photomicrograph showed composite image from three separate images (original magnification ×40, scale bar = 250 μm.). I, inset, showed the localization of carboxymethyl-DiI-labeled-MSC (original magnification ×200; scale bar = 100 μm). Images were captured using Nikon TE300 wide-field microscope equipped with a CCD camera. (E): Photomicrograph of tumor sections stained for the presence of apoptotic cells using anti-single-stranded DNA in ΔGli36, ΔGli36/MSC, and ΔGli36/iNHA tumor. Sections were counterstained with DAPI nuclear stain (blue); green color indicates anti-single-stranded-DNA-positive cells. Sections were shown at original magnification ×200. Scale bar = 50 μm. Abbreviations: H&E, hematoxylin and eosin; iNHA, immortalized NHA; MSC, mesenchymal stem cell; N, necrotic region; NHA, normal human astrocytes; T, tumor; DAPI, 4′,6-diamidino-2-phynylindole.

MSC Restrict Vascular Growth in Glioma Cells In Vivo and Reduced Angiogenesis In Vitro

Tumors obtained from coinjection of MSC and ΔGli36 human glioma cells were further evaluated for vessel morphology and microvessel density. The microvessel density as revealed by CD31 staining was significantly higher in ΔGli36 and ΔGli36/iNHA tumor when compared with that of ΔGli36/MSC (p < .01; Fig. 3A). To assess the functional status of the tumor neovasculature, tumor sections were simultaneously stained for CD31 expression and pericytes coverage (α-SMA). The results showed the presence of CD31+ cells in regional areas of the representative tumor sections of ΔGli36 and ΔGli36/MSC. Expression of α-SMA+ cells could only be found in ΔGli36/MSC but not in ΔGli36 (Fig. 3B), indicating the presence of a reactive microvasculature in the ΔGli36 tumors. In ΔGli36/MSC tumors, α-SMA+ pericytes were present within the tumor vasculature. These cells did not overlap with the staining against CD31 (Fig. 3B). As shown in Figure 3C, MSC (red cells) could be detected close to CD31+ regions, suggesting a possible role in angiogenesis.

Figure 3.

Reduction of angiogenesis in ΔGli36 cocultured with MSC. (A): Angiogenic vessels were identified by immunostaining using antibody against CD31 in ΔGli36, ΔGli36/MSC, and ΔGli36/iNHA tumor. Microvessel density of the tumor sections were determined by expressing the cumulative CD31+ vessels to the area of the tumor sections. Data shown are averages of five randomly selected fields ± SEM. *, p < .05; **, p < .01. Scale bar = 100 μm. (B): ΔGli36 and ΔGli36/MSC tumor sections were double immunostained with CD31 (green) and α-SMA (blue). White arrow indicates colocalization of CD31 and α-SMA staining. Scale bar = 20 μm. (C): Confocal images of ΔGli36/MSC tumor sections stained for the colocalization of CD31 (green), α-SMA (blue), and carboxymethyl-DiI-labeled MSC (red). Scale bar = 20 μm. Abbreviations: iNHA, immortalized NHA; NHA, normal human astrocytes; MSC, mesenchymal stem cells; SMA, α-smooth muscle actin.

To further investigate whether MSC may be involved in EPC recruitment during angiogenesis, we isolated EPC from peripheral blood obtained from healthy donor using density gradient centrifugation. EPC expressed high level of CD133, which is indicative of early population of EPC [17]. In addition, the isolated cells were immune-positive for CD34, Tie2, VE-cadherin (CD144), VEGFR2 (CD309) but expressed very low level of CD45 (Fig. 4A; supporting information Fig. S2). Moreover, cellular uptake of DiI-Ac-LDL, which is indicative of functional EPC, was also observed in these cells and was absent in MSC (Fig. 4B). Collectively, these markers confirmed the endothelial lineage of the isolated cells [17, 18]. As shown in Figure 4C, the recruitment of EPC was severely impaired by ΔGli36/MSC-CM in comparison to ΔGli36-CM. Similar findings were obtained with another isolate of MSC (data not shown), indicating that the suppressive effect of MSC on EPC recruitment is not specific to one MSC isolate. Interestingly, the extent of EPC recruitment in MSC-CM was similar to those in ΔGli36/MSC-CM, thus suggesting that the observed reduced tumor vasculature in the presence of MSC is likely due to an inhibition on the recruitment of EPC and its associated angiogenic inducing factors.

Figure 4.

Conditioned medium from ΔGli36/MSC coculture prevented recruitment of EPC and endothelial tube formation. (A): EPCs were immunostained with antibodies against CD133, CD34, VE-cadherin, VEGFR-2, Tie2, and CD45. Mouse IgG1 and rabbit serum were used as negative control for the first five antibodies and CD45, respectively. Scale bar = 100 μm. (B): EPC and MSC were assessed for their ability to uptake DiI-Ac-LDL (red fluorescence). Cells were counterstained with DAPI (blue fluorescence). Representative images were presented. Scale bar = 50 μm. (C): EPC recruitment assay was performed using ΔGli36-CM, MSC-CM, and ΔGli36/MSC-CM. The number of recruited cells to the bottom chamber were counterstained with propidium iodide and counted. Data shown are averages of triplicates ± SEM. *, p < .05. (D): Angiogenesis assay was performed to determine whether ΔGli36-CM, MSC-CM, NHA-CM, ΔGli36/NHA-CM, or ΔGli36/MSC-CM could induce tube formation of HUVEC. 50 ng/ml of recombinant human VEGF was used as positive control. Tubes formed were visualized and captured using the Zeiss 200M wide-field microscope at ×50 original magnification. Scale bar = 200 μm. (E): Tube formation index was scored using ImageJ software (National Institutes of Health, Bethesda, MD). Data were obtained from four replicates of two independent experiments and presented as ratio of coculture to monoculture. Abbreviations: CM, conditioned medium; EPC, endothelial progenitor cells; LDL, low density lipoprotein; NHA, normal human astrocytes; MSC, mesenchymal stem cell; VE, vascular endothelial; VEGFR, vascular endothelial growth factor receptor; DAPI, 4′,6-diamidino-2-phenylindole; HUVEC, Human umbilical vein endothelial cells.

Next, we investigated whether the antitumor effect was mediated by antiangiogenic factors in ΔGli36/MSC coculture using in vitro angiogenesis assay. Our results showed that incubation of HUVECs in ΔGli36-CM, MSC-CM, NHA-CM, and ΔGli36/NHA-CM stimulated tube formation (Fig. 4D), which is comparable with the known angiogenic molecule, recombinant human VEGF (data not shown). As shown in Figure 4E, increased tube formation index was observed in ΔGli36/NHA coculture when compared with that of NHA monoculture (ratio of ΔGli36/NHA vs. NHA is 1.5), whereas tube formation index is similar for ΔGli36/NHA and ΔGli36 (ratio of ΔGli36/NHA vs. ΔGli36 is 1.0). However, the presence of MSC in the ΔGli36 coculture inhibited tube formation by 50% when compared with either ΔGli36 or MSC monoculture (ratio of ΔGli36/MSC vs. ΔGli36 or MSC is 0.5), thus indicating that MSC impaired tumor angiogenesis through the release of antiangiogenic factors.

MSC Downregulate the Expression of Angiogenic Molecules in Glioma Cells

To determine which factors are expressed by ΔGli36/MSC coculture, a proteome array was performed using CM harvested from ΔGli36, MSC, and ΔGli36/MSC. Cytokines that are differentially expressed in ΔGli36/MSC coculture were shown in Table 1. These candidates included activated leukocyte cell adhesion molecule (also known as CD166 antigen), PDGF-BB, FGF-2, macrophage inflammatory protein-1-α, Eotaxin-3 (also known as CCL26), regulated on activation normal T-cell expressed and secreted (also known as CCL5), tissue inhibitors of metalloproteinases-2, interleukin-1β (IL-1β), and IGF-1 (Table 1). Because angiogenic index was reduced in the in vivo coculture, we narrowed down the candidate proteins to those that were downregulated in angiogenesis. The mRNA expression of these genes was further validated by qPCR in the coculture cells. Significant downregulation of PDGF-BB and IL-1β mRNA expression was observed in ΔGli36/MSC coculture in comparison to that of ΔGli36/NHA and ΔGli36 (Fig. 5A). In line with the qPCR results, lower level of PDGF-BB proteins were also detected in ΔGli36/MSC when compared with ΔGli36/NHA and ΔGli36, thus confirmed our antibody array results (Fig. 5B). However, IGF-1, IL-1β, and FGF-2 protein expression failed to show similar trend as observed in qPCR. IGF-1 and FGF-2 protein expressions were lower in ΔGli36/MSC when compared with the controls, while the expression of IL-1β protein was similar for the three culture conditions (Fig. 5B).

Figure 5.

Crosstalk between MSC and ΔGli36 human glioma cells downregulate the expression of angiogenic molecules. (A): QPCR was performed on total RNA isolated from ΔGli36, ΔGli36/MSC, and ΔGli36/NHA coculture using primers against PDGF-BB, IGF-1, IL-1β, and FGF-2. (B): ELISA was performed on CM harvested from ΔGli36, ΔGli36/MSC, and ΔGli36/NHA coculture. For both qPCR and ELISA, the results shown are normalized to ΔGli36. Data shown are averages of triplicate ± SEM. *, p < .05. Abbreviations: CM, conditioned medium; FGF-2, fibroblast growth factor-2; IL-1β, interleukin-1β; IGF, insulin growth factor; NHA, normal human astrocytes; PDGF-BB, platelet-derived growth factor.

Table 1. Cytokines that are differentially expressed in ΔGli36/mesenchymal stem cells coculture as determined by antibody array analysis
inline image

Impaired IL-1β and PDGF Signaling Cascades in Glioma Cocultured with MSC

IL-1β, a proinflammatory cytokine, binds to its cell-surface receptor and initiates signaling cascade that involves the activation of nuclear factor kappa B, which in turn activates the cysteine protease, cathepsin B. Cathepsin B has been shown to be overexpressed in malignant gliomas, which are characterized by aberrant neovascularization [21]. Thus, we first examined whether MSC may downregulate angiogenesis through IL-1β signaling. IL-1β protein expression was 40% and 80% lower in ΔGli36/MSC tumor in comparison to that of ΔGli36/iNHA and ΔGli36, respectively (Fig. 6A). However, using a representative of two animals per group, we detected significantly lower level of IL1-R1 in one of the ΔGli36/MSC sample (Fig. 6A). Immunohistochemistry staining confirmed that cathepsin B expression was lost in the ΔGli36/MSC tumor, whereas abundant cathepsin B expression was detected in ΔGli36 and ΔGli36/iNHA tumor alone (Fig. 6B), indicating that MSC exert an inhibitory effect on tumor vasculature, in part, through the downregulation of IL-1β signaling cascade.

Figure 6.

Glioma/MSC coculture resulted in downregulation of PDGF and IL-1β signaling. (A): IL-1β and IL1-R1 expression were determined using ELISA and Western blot, respectively, on tumor lysate harvested from ΔGli36, ΔGli36/MSC, and ΔGli36/iNHA tumors. Amount of protein loaded was normalized against β-tubulin. (B): Photomicrograph of tumor sections stained for the presence of cathepsin B in ΔGli36, ΔGli36/MSC, and ΔGli36/iNHA tumor. Sections were counterstained with hematoxylin and shown at original magnification ×200. Scale bar = 50 μm. (C): PDGF-BB expression was determined using ELISA on ΔGli36, ΔGli36/MSC, and ΔGli36/iNHA tumor lysates. Data shown are averages of triplicate ± SEM. (D): Immunoblot was performed on tumor lysates harvested from ΔGli36, ΔGli36/MSC, and ΔGli36/iNHA tumors using antibodies against Akt, phospho-Akt, PDGFR-β, and phospho-PDGFR-β. Amount of protein loaded was normalized against β-tubulin. *, p < .05. Immunoblots were performed using tumor lysates harvested from two representative animals per group. Abbreviations: IL-1β, interleukin-1β; iNHA, immortalized NHA; MSC, mesenchymal stem cells; PDGF-BB, platelet-derived growth factor.

In a recent report, the downregulation of cathepsin B was shown to decrease the expression of PDGFR-β expression and correspondingly increase the apoptotic index of glioma cells [22]. Since PDGF-BB is also known to play an important role in glioma angiogenesis [23], we further examined whether the reduction in microvessel density may be modulated through PDGF-BB/PDGFR-β signaling. As shown in Figure 6C, PDGF-BB protein expression was significantly reduced in the ΔGli36/MSC tumor section when compared with that of ΔGli36 and ΔGli36/iNHA tumors (Fig. 6C). Immunoblot analysis further showed that the levels of activated PDGFR-β were correspondingly reduced in ΔGli36/MSC tumor in comparison to control tumors (using a representative of two animals per group as shown in Fig. 6D). This reduced PDGFR-β activation correlated with downregulation of activated Akt (Fig. 6D) in ΔGli36/MSC tumors, suggesting the involvement of PDGF/PDGFR signaling in MSC-induced tumor suppression.


In the field of cancer therapy, the most attractive feature of MSC is its inherent ability to track tumor cells and its immunosuppressive nature, suggesting that MSC modified with the appropriate therapeutic gene of interest could home to microscopic tumors and deliver long-term transgene expression prior to eradication by the immune system. However, the multifaceted role of MSC in the tumor microenvironment is still not fully understood. Results of this study suggested that the inhibitory effect of MSC on glioma tumor angiogenesis might be mediated through paracrine pathways that resulted in impaired EPC recruitment and downregulation of proangiogenic factors such as PDGF-BB, IGF-1, FGF-2, and IL-1β.

MSC has been shown to have the capacity to transdifferentiate into EC-like [10] and pericytes-like cells [24] to stimulate angiogenesis and promote tumor growth [25]. Direct coculture of adipose-derived MSC (AMSC) with prostate cancer was shown to enhance tumor growth through increase in tumor vascularity mediated by FGF-2 [26] and differentiation of AMSC into endothelial-like cells [27]. By contrast, Otsu et al. [14] showed that coculture containing 1:1 ratio of EC to MSC resulted in EC apoptosis and degeneration of EC capillaries due to the generation of reactive oxygen species. In our system, neither differentiation of MSC into EC (Fig. 3) nor pericytes-like cells (Fig. 3) was observed in glioma/MSC cocultured tumors. Rather, our results showed decreased microvessel density in vivo that correlated with reduced tube formation index and recruitment of EPC in vitro, suggestive of an antiangiogenic response elicited by the MSC. The presence of the α-SMA+ cells, which is indicative of pericytes, and corresponding reduction of CD31+ microvessel density in glioma/MSC when compared to glioma alone suggested that the presence of MSC in the EC environment reduces vascular permeability, a finding that is consistent with that of Pati et al. [28].

Pericytes are known to contribute to the integrity of mature blood vessels. Absence of pericytes results in a more reactive vessel that is irregular, leaky and supports metastasis [29, 30]. It is interesting to note that although α-SMA+ cells are present in glioma/MSC coculture, the level of PDGF-BB, which has been shown to recruit pericytes to glioma [23], is downregulated. PDGF-BB is expressed in the endothelial tip cells or cell that leads the new sprout at the forefront during angiogenesis, which stimulates both recruiting pericytes with PDGFR-β expressed on their surface and the subsequent creation of the wall of the newly formed capillary [31, 32]. However, PDGF-BB is not the only factor that mediates pericyte recruitment because mice that are deficient in either PDGF or PDGFR-β have been shown to have normal pericyte coverage in the perisinusoidal capillaries [33], suggesting the involvement of other factors, including secreted protein acidic and rich in cysteine [34] and endothelial nitric oxide [35]. On the other hand, the reduced PDGF expression in the presence of MSC may be a consequence of the downregulation of angiogenesis pathways. Gliomas are highly invasive and malignant tumors and displayed aberrant PDGF signaling [36]. PDGFs bind and signal through the receptor tyrosine kinases, PDGFR-α and −β. Binding of PDGFs to its receptors trigger receptor dimerization that leads to receptor autophosphorylation and subsequent signal propagation. Inhibition of PDGF signaling with the receptor tyrosine kinase inhibitor SU6668 [37] or SU5416 [38] has been shown to inhibit tumor growth through vessel regression with or without pericyte involvement.

Inflammation and angiogenesis are pivotal processes in tumor progression. Inflammatory molecules such as IL-1β are frequently upregulated in cancer cells and macrophages found in the tumor microenvironment. In the coculture of ΔGli36/MSC, the IL-1β mRNA levels were significantly reduced when compared with that of ΔGli36 or ΔGli36/NHA (Fig. 5A). Although we observed a slight decrease in IL-1β proteins in the CM derived from ΔGli36/MSC when compared with the controls, the difference was not significant possibly due to the relatively low levels of proteins (8.74–9.93 pg/ml). However, the lower IL-1β proteins in ΔGli36/MSC versus ΔGli36 or ΔGli36/NHA were statistically significant in vivo (Fig. 6A). IL-1β has been shown to interact with PDGF-BB to induce sustained phosphorylation of PDGFR-β and its association with IL1-R1 [39], which may produce synergistic effects in downstream signaling. Moreover, IL-1β signaling regulates cathepsin B expression via the NF-κB pathway. Cathepsin B has been shown to promote remodeling of the ECM to permit neovascularization [40, 41]. Because cathepsin B cleaves and processes IGF-1 following receptor internalization [42], a process that is required for cell signaling via the Shc/MAP kinase pathway, the marked inhibition of cathepsin B expression by IL-1β signaling suggested that MSC in the tumor milieu prevented ECM remodeling that results in increase invasion and metastasis. Additionally, downregulation of cathepsin B, either alone or with urokinase-type plasminogen activator receptor silencing, decrease the expression of PDGFR-β, increase the apoptotic index of glioma cells [22], and downregulate integrin expression [43]. Furthermore, disruption of the tube-like structure of EC in vitro and reduced microvasculature in vivo was also observed in cathepsin B-silenced SNB19 glioma cells [44], thus supports our results that reduced cathepsin B expression retarded ECM-dependent tumor expansion and associated microvascular growth.

In summary, our results showed that MSC could prevent EPC recruitment to tumor vasculature, and at the same time, releases soluble factors that inhibit ECs tube formation and reduced microvessel density. The antitumor effect might be mediated, in part, through paracrine pathways that resulted in impaired EPC recruitment and downregulation of proangiogenic factors such as PDGF-BB, IGF-1, FGF-2, and IL-1β. Although there are many factors that could influence the fate of the tumor growth kinetic, it is important to continue in-depth investigations toward the understanding of MSC biology as they could serve as an invaluable tool in the detection or possibly, treatment of cancer when modified with appropriate therapeutic gene.


Our results have clearly demonstrated that MSC exhibited an antiglioma effect. Analysis indicated that PDGF signaling, which is known to play a key role in glioma angiogenesis, is one of the pathways that is likely involved in the observed reduced glioma angiogenesis.


This research is supported by grants from the SingHealth Foundation and National Medical Research Council, Singapore.


The authors indicate no potential conflicts of interest.