Erythropoietin, a hypoxia-regulated factor, elicits a pro-angiogenic program in human mesenchymal stem cells

Experimental Hematology 35 (2007) 640–652

Erythropoietin, a hypoxia-regulated factor, elicits a pro-angiogenic program in human mesenchymal stem cells Kevin J. Zwezdaryka, Seth B. Coffelta, Yanira G. Figueroaa, Juliet Liua, Donald G. Phinneya,d, Heather L. LaMarcaa, Luisa Florezc, Cindy B. Morrisa, Gary W. Hoyleb, and Aline B. Scandurroa Departments of aMicrobiology & Immunology; bMedicine, and cPathology and dCenter for Gene Therapy, Tulane University Health Sciences Center, New Orleans, La., USA (Received 14 November 2006; revised 18 January 2007; accepted 19 January 2007)

Objective. The ability of erythropoietin (EPO) to elicit a pro-angiogenic effect on human mesenchymal stem cells (hMSC) was tested. hMSC are currently under study as therapeutic delivery agents that target tumor vessels. Hypoxia favors the differentiation of hMSC towards a pro-angiogenic program. However, the classical angiogenic factors, vascular endothelial growth factor and basic fibroblast growth factor, are not fully capable of restoring this effect. The hypoxia-regulated factor, EPO, induces angiogenesis in endothelial cells. Here, EPO’s pro-angiogenic effect on hMSC was analyzed. Methods. hMSC were tested for EPO receptor expression by western blot, immunofluorescence, and flow cytometry assays. Downstream receptor signaling components JAK and STAT were measured by standard assays. Pro-angiogenesis effects mediated by EPO treatment of hMSC were measured by proliferation, cytokine, or pro-angiogenesis factor secretion, metalloprotease activation, migration, invasion, wound healing, and tubule formation assays. Results. hMSC express the cognate EPO receptor and are capable of promoting angiogenesis following EPO treatment in all the angiogenesis assays tested. EPO-treated hMSC proliferate and secrete pro-angiogenesis factors more readily than untreated hMSC. EPO leads to increased hMSC chemotaxis, migration, and activation of matrix metalloprotease-2. This treatment causes greater recruitment of vessels as measured in an in vivo angiogenesis assay. Conclusion. EPO is capable of eliciting a pro-angiogenesis program in hMSC that instigates secretion of angiogenic factors and the subsequent recruitment of endothelium. This study defines a novel mechanism for tumor cell recruitment of blood vessels that is important to consider in the design of stem cell–based therapies. Ó 2007 International Society for Experimental Hematology. Published by Elsevier Inc.

The hypoxia-regulated factor erythropoietin (EPO) is a glycoprotein hormone that counteracts tissue hypoxia (lowoxygen) by stimulating the formation and differentiation of erythroid precursor cells in the bone marrow. EPO is synthesized in the fetal liver or adult kidney in response to various forms of diminished oxygen levels including anemia and hypoxia [1]. The brain, ovum, placenta, endometrium, and smooth muscles are known to secrete and/or respond to EPO [2–5]. Interestingly, this hypoxia-regulated factor appears to have various physiologic roles. The first, for which Offprint requests to: Aline B. Scandurro, Ph.D., Department of Microbiology and Immunology, Tulane University Health Sciences Center, 1430 Tulane Ave, SL38, New Orleans, LA 70112; E-mail: [email protected]

it was named, is characterized by the originally defined endocrine role on primitive hematopoietic cells. The second, more recently described, role for EPO is characterized by a paracrine and/or autocrine stress response [6,7]. Thus, exposure to hypoxia or neurotoxic glutamate in the brain releases erythropoietin from local glial cells that bind EPO receptors (EPOR) present on neurons initiating an antiapoptotic pro-survival response. While the former endocrine EPO role is mediated by high-affinity EPO receptors on immature hematopoietic cells, the latter paracrine/autocrine EPO effects are mediated by lower-affinity EPO receptors on the remaining organ sites [8]. Ribatti et al. and others described yet another role for erythropoietin as a pro-angiogenic factor comparable to

0301-472X/07 $–see front matter. Copyright Ó 2007 International Society for Experimental Hematology. Published by Elsevier Inc. doi: 10.1016/j.exphem.2007.01.044

K.J. Zwezdaryk et al./ Experimental Hematology 35 (2007) 640–652

the classical pro-angiogenesis factors like vascular endothelial growth factor (VEGF) and basic fibroblast growth factor (bFGF) [8–10]. Moreover, EPO or EPOR knockout mice studies underscore this role since homozygous knockout mice that perish on embryonic day 13.5 exhibit impaired circulation due to diminished vessel formation, especially around the heart. Further, close observation of early angiogenesis and vasculogenesis in murine EPO and EPOR knockout mice lead Muller-Ehmsen et al. to assign a distinct role for EPO in angiogenesis and dismiss a role for EPO in vasculogenesis [11]. Analogous genetic targeting experiments demonstrated that EPO and EPOR signaling promote angiogenesis and survival of several tumors of female reproductive organs [12–14]. The authors of this study concluded that deprivation of EPO signaling may represent an effective target for erythropoietin-producing malignant tumors [14]. Adult bone marrow–derived human mesenchymal stem cells (hMSC) are currently studied as chemotherapeutic delivery agents that specifically target neoplasms because hMSC are recruited by tumors and can contribute to their blood vessel development (angiogenesis) [15–19]. Angiogenesis, whether mediated by pro-angiogenesis factors like VEGF, bFGF, or EPO, is typically defined as the formation of new blood vessels from preexisting ones. This is a multi-step process divided into two phases. The first, ‘‘activation’’ phase, is characterized by endothelial cell– mediated degradation of the perivascular basement membrane, migration to extracellular space, and proliferation to enable tube formation. Differentiation of the endothelial cells, reestablishment of the basement membrane, and recruitment of supporting cells like smooth-muscle cells and pericytes characterize the second, ‘‘maturation’’ phase [20]. This paradigm of angiogenesis is now under scrutiny based on the finding that hMSC are locally recruited and contribute to de novo blood vessel formation [19]. In this regard, it was recently reported that hypoxia, or reduced oxygen tension, favors the differentiation of hMSC into cells capable of building new blood vessels [16]. However, the classical pro-angiogenesis factors VEGF and bFGF on their own did not fully restore this effect. In this report, the ability of erythropoietin to elicit a proangiogenic effect on primary human MSC cultures was tested as an unexplored yet alternate hypoxia-regulated pro-angiogenic factor to VEGF and bFGF. EPO was chosen because it is secreted from several hypoxic tumor microenvironments and is implicated in the autocrine growth of erythrocytic leukemia cells, hepatocellular carcinoma, and several female breast and reproductive organ carcinomas [14,21,22]. These cancers are candidates for genetically modified MSC-based therapies. Note that in this study, hypoxia exposure (2% O2) of the cells is done in parallel to or along with recombinant EPO treatment to compare results with those established in previous reports [16]. Here we report the first evidence that human adult bone marrow–

641

derived mesenchymal stem cells express erythropoietin receptors and are capable of promoting angiogenesis following erythropoietin treatment. The salient finding of the study is that hMSC can facilitate angiogenesis not necessarily by directly participating as the endothelial cells that line the new vessels but by secreting and recruiting the components for this process to occur following their stimulation. This finding may shed some light on previously unknown mechanisms exploited by cancer cells, providing a unique target for stem cell–based therapies.

Materials and methods Cell culture Human MSC used in the study were obtained from our collaborators at the Tulane University Center for Gene Therapy, led by Darwin J. Prockop, M.D., Ph.D. Additionally, MSC were obtained from two other commercial suppliers, Cambrex BioScience (Walkersville, MD, USA) and Allcells (Emeryville, CA, USA), to ensure enough variability of the starting cell population and make certain that findings are universal and not unique to single donor pools derived from a unique source. These suppliers test the MSC for their homogeneity and provide test results for their differential potential to chondrogenic, osteogenic, and adipogenic lineages. Once obtained, expanded, and established in our lab, the MSC are also verified for several established MSC markers including but not limited to CD90, CD105, CD106, CD117, CD146, CD56, CD166, CD29, CD44, CD14, CD31, CD34, and HLA-DR, and by microscopic morphology. MSC cultivation medium consisted of minimum essential medium alpha with GlutaMax I (Invitrogen Corp., CA, USA) supplemented with 16.5 to 20% fetal bovine serum. For serum-free cultivation, growth medium without serum supplementation was used (Invitrogen Corp.). MSC of a passage number no greater than six was routinely used in all the experiments. Gelatin zymography Experimental cells were plated to 70% confluency in 24-well plates, serum-starved overnight, and further manipulated as described. Conditioned medium was then collected, protein concentrations were estimated with a Micro BCA Protein Assay Kit (Pierce, IL, USA), and approximately 50 mg of total protein were subjected to gelatin zymography. Precast gelatin or casein zymography gels and the reagents required for electrophoresis, renaturation, and development were obtained from commercial sources (Invitrogen Corp.). Conditioned medium samples were typically loaded without solute concentration and separated at 100 V prior to renaturation and development. Boyden chamber chemotaxis assay Recombinant human EPO, VEGF, and bFGF were the chemoattractants for migration (R&D Systems, Inc., Minneapolis, MN, USA) in a modified Boyden chamber chemotaxis assay. Growth factors were diluted in serum-free growth medium and added to the lower compartment of the Boyden chamber (Neuro Probe, Inc., MD, USA). Serum-free and 20% serum containing growth medium were added to lower compartments as negative and positive assay controls, respectively. Lower-chamber wells were overlaid with a porous

642

K.J. Zwezdaryk et al./ Experimental Hematology 35 (2007) 640–652

membrane filter (8 mm) that was precoated with 0.1 mg/mL gelatin (Biorad, CA, USA). The chamber was assembled and serum-starved experimental cells were added to the upper compartment of the apparatus at a density of 45  104 cells/well and incubated at 37 C overnight to allow for cell migration. Then the apparatus was disassembled, the cells on the upper side of the membrane were wiped off, and the migrated cells were visualized by staining (Diff-Quik, Dade Behring, Inc., DE, USA). Migration was quantified by counting the nuclei that passed through the filter. Stained nuclei from a minimum of 6 fields of view (200) for 3 replicates were counted and the data was expressed as the average number of migrated cells. Fluoroblok invasion assay Cellular invasion of experimental cells in response to factors described above was assessed using the quantitative Fluoroblok invasion assay (BD Discovery, MA, USA). Recombinant human EPO, VEGF, or bFGF was added to the bottom of a 24-well plate and serum-starved cells were seeded on top of Matrigel-coated (growth factor–reduced Matrigel, diluted 1:10 in serum-free medium, BD Biosciences, MA, USA) 8-mm Fluoroblok membrane inserts at a density of 2.5  105 cells per insert. Serum-free and 20% serum-containing growth medium were added to lower compartments as negative and positive assay controls, respectively. The inserts were placed into each well containing growth factors or controls and the plates were incubated at 37 C for 20 hours to allow for cell invasion. The invading cells were fluorescently labeled with calcein AM (Molecular Probes, OR, USA) for 1 hour at 37 C and relative fluorescence units were obtained with a fluorescence plate reader (FLUOstar optima, BMG Labtech, NC, USA) from a minimum of 3 replicates. Phase-contrast images of the invading cells were collected using a Nikon TE300 inverted epifluorescence microscope (DP Controller v1.2.1.108, Olympus Optical Company, TX, USA). Cell invasion was expressed from relative fluorescence units obtained for the Matrigel-coated filters over the control sample not containing Matrigel coat. Proliferation assays Experimental cells (approximately 7000 cells/well) were seeded onto 96-well plates in complete growth medium, allowed to adhere 3 to 4 hours at 37 C, then serum-starved overnight as standard. The medium was removed, the cells were washed twice with phosphate-buffered saline (PBS), and 8 replicates were treated as indicated. Cells were further incubated for 24, 48, 72, or 96 hours, prior to analyzing with CellTiter96 Assay proliferation kit (Promega, WI, USA). Cell proliferation within harvested plates was quantified as recommended in a plate reader (Bio-Tek Instruments, VT, USA). Wound-healing assay Cells were plated in 12-well plates to 70% confluence and serumstarved. The cells were washed two times with PBS, and a ‘‘wound’’ was made by scoring the confluent monolayer of MSC with a pipette tip across the plate well [23]. Scored wells were washed with PBS twice to remove cell debris prior to the addition of serum-free medium containing angiogenic factors to be tested. Samples containing an established angiogenic factor (VEGF) were included as positive controls and a sample with only serum-free medium and no angiogenic factor served as a negative control. Invasion or growth into the scored area was observed

and documented by phase-contrast microscopy (Nikon TE300, Nikon, TX, USA). In vitro tubule formation assay Chamber slides were coated with undiluted Matrigel solution and allowed to solidify overnight in a humidified 37 C CO2 incubator. Primary human MSC cultures were plated in order to obtain at least 50,000 cells per test well and serum-starved overnight. Cells were fluorescently labeled (CellTracker, Invitrogen Corp.) prior to layering on top of Matrigel-coated slides. Serum-free medium containing angiogenic factors was added and slides were further incubated overnight. Tubule formation was monitored by phasecontrast and fluorescence microscopy. In vivo murine Matrigel plug angiogenesis assay Immunodeficient female BalbC nu/nu (nude) mice were obtained following protocol approval by the Institutional Animal Care and Use Committee of Tulane University and allowed 7 to 10 days to acclimate in sterile housing of vivarium before the start of the experiment. Five to 10 mice per treatment group were used to achieve statistical significance. Growth factor–reduced Matrigel was thawed on ice and 24-hour serum-starved MSC were grown to obtain at least 250,000 cells per test plug. Serum-starved cells were washed in PBS, trypsinized, neutralized in serum-free medium, and counted as typical. Washed cells were then fluorescently labeled with either or both DAPI and CellTracker. Cells and Matrigel were mixed and loaded in cold 1-mL syringes fitted with 25-g needles. Each mouse received bilateral injections of 500 mL growth factor and cell suspension at lower dorsal side just above hind legs (2–3 mm). Typically, one side was injected with the test sample and as assay control, the opposite side had Matrigel alone. Angiogenic lesions were allowed to form for 7 to 10 days postinoculation. At the end of the experiment, mice were sacrificed by CO2 inhalation. Plugs were harvested and quickly fixed in a freshly prepared cold 10% formalin solution. Subsequent to fixation, harvested plugs were processed for paraffin embedding or cryopreserved and sectioned as standard. A reference slide from each harvested plug was stained with hematoxylin-eosin (H&E). All antibodies used in this study were titered and analyzed for appropriate specificity in western blots prior to immunostaining of slides. New blood vessels were quantified by counting the number of vessels in five high-powered magnification fields and values were compared to those from a minimum of two ‘‘blind’’ readings from H&E-stained slides or immunostained slides as indicated. Western blot analysis Typically, 30 to 50 micrograms of total protein was resolved on 4 to 12% SDS-polyacrylamide gels, and transferred by blotting to nitrocellulose membranes (Amersham Biosciences, NJ, USA). Protein-free sites were blocked with 5% nonfat dried milk in PBS containing 0.1% Tween-20 (PBST) for 1 hour at room temperature and blots were incubated at 4 C overnight with the following antibodies: EPOR (#MAB307 R&D Systems, Inc., Minneapolis, MN, USA); JAK2 (#SC-294, Santa Cruz Biotechnology, CA, USA); phosphoJAK2(tyr1007/8) (#3771 Cell Signaling Technology, Beverly, MA, USA); STAT5 C-17 (#sc-835, Santa Cruz Biotechnology); phosphoSTAT5(tyr694) (#9351 Cell Signaling Technology); b-actin (#A2066 Sigma-Aldrich, St. Louis, MO, USA). The blots were then washed in PBST and incubated with

K.J. Zwezdaryk et al./ Experimental Hematology 35 (2007) 640–652

species-specific IgG conjugated to horseradish peroxidase (1:5000; Amersham Biosciences) for 1 hour at room temperature. Antigen-antibody complexes were visualized after exposure to x-ray film by enhanced chemiluminescence (Amersham Biosciences). Flow cytometry MSC were harvested following two PBS washes by scraping to avoid cell-surface marker losses due to conventional trypsin harvesting techniques. For each condition being tested 1  106 cells or a ratio of 0.5 ug antibody/1  106 cells were used. The Fc receptors/IgG binding sites of cells were blocked by incubation in 10% human serum or serum source of secondary antibody. The cells were resuspended in 500 uL of sheath fluid or PBS and analyzed with a BD-FACSCalibur flow cytometer (BD Biosciences). Primary antibodies used were: isotype-control FITC mouse IgG1K (BD #556649); isotype-control PE mouse IgG1K (BD #551436); isotype-control PE-Cy5 mouse IgG1K (BD #557224); CD105 (BD anti-human #555690); CD166 (BD PE-labeled antihuman #559263); CD90 (BD Cy-Chrome anti-human #555597); CD44 (BD FITC anti-human #555478); CD34 (BD PE anti-human #5507610; CD31 (BD PE anti-human #550761); CD106 (BD CyChrome anti-human #551148); and EPOR (R&D anti-hEPOR #MAB307). Immunohistochemical analysis Immunohistochemical staining was achieved with the horseradish peroxidase enhancement method of LSAB 2 kit (DakoCytomation Corp, CA, USA). Micrographs were taken on a Zeiss Axioplan 2 fluorescence microscope with Intelligent Imaging Innovations deconvolution hardware and software (SlideBook ver. 4). Statistical analysis Data are shown as average 6 standard error of the mean (SEM). Multiple-group comparison was performed by one-way analysis of variance (ANOVA) followed by the Bonferroni procedure for comparison of means. Comparison between any two groups was analyzed by the two-tailed Student’s t-test or two-way ANOVA (Prism4, GraphPad Software Inc., CA, USA). Values of p ! 0.05 were considered statistically significant.

Results Human mesenchymal stem cells express EPOR and the established downstream components of this receptor, JAK2 and STAT5 To establish a role for EPO in driving pro-angiogenic responses in primary human mesenchymal stem cell cultures, we first demonstrated the presence of the cognate EPO receptor on the surface of these cells. Specific antibodies to human EPO receptor were employed in flow cytometry, immunofluorescence, and western blot analysis for this purpose. Flow cytometry with antibodies specific for CD90, CD105, CD45, and CD34 served to characterize the experimental cell population consistent with established MSC markers (Fig. 1A). Our results indicate that primary cultures of hMSC do express EPOR (Fig. 1). Surprisingly, expression of EPOR was prevalent since typically greater

643

than 60% of the cells in hMSC preparations were positive for EPOR (n O 4). Note the relative uniform cytoplasmic staining for CD105 in MSC compared to only 3 of 5 cells positively stained MSC for cytoplasmic EPOR in the immunofluorescence micrographs (arrows, Fig. 1B). Not only was the EPOR present in these cells, but its established downstream signaling components, JAK2 and STAT5, were also present as determined by the use of specific antibodies in western blot analysis (Fig. 1C) [10]. Exposure to hypoxia (2% O2) modestly affected EPOR expression (1.5-fold to twofold) as established from both flow cytometry (data not shown) and western blot (Fig. 1C) analysis. Additionally, activation of JAK2 and STAT5 was affected by hypoxia given the consistently measured increase in phosphoJAK2 (2.1-fold) and phosphoSTAT5 (2.3-fold) after quantification by densitometric analyses. Measured b-actin levels served to standardize loading differences in these analyses.

Erythropoietin and hypoxia affect the proliferation and the secretion of angiogenesis factors by the MSC The direct effect of erythropoietin on hMSC growth was measured by CellTiter96 Aqueous One Assay proliferation assay kit (Promega). Human donor preparations of MSC were serum-starved overnight and treated with increasing amounts of recombinant human EPO added to serum-free growth medium; growth characteristics were measured following 0, 24, 48, 72, and 96 hours posttreatment. EPO treatment (0–100 U/mL) of MSC resulted in enhanced proliferation when compared to untreated controls (Fig. 2A) and was comparable to the growth effects measured for serum-treated MSC. Maximum effect was established by 24 to 48 hours posttreatment. Hypoxia exposure of primary MSC cultures did not appreciably affect the growth of these cells except when treatments were carried out with scarcely confluent plating (!25–50%). In this latter case, there was a significant increase in proliferation of hypoxia-treated MSC (O50%, data not shown). The conditioned medium from similarly treated hMSC cultures was tested for pro-angiogenic factor and matrix metalloproteinase (MMP) secretion. Primary hMSC cultures were treated with recombinant human EPO or exposed to hypoxia (2% O2) for 24 hours prior to harvesting the spent culture medium. Cytokine antibody array analysis revealed that the pro-angiogenic factor, VEGF, was consistently and significantly upregulated by EPO or hypoxia in agreement with published reports [10,16]. Additionally, hypoxia exposure also induced IL-6 and IL-8 secretion into the culture medium whereas EPO treatment did not. MSC incubation with EPO uniquely led also to upregulation of G-CSF, HGF, IL-1b, FGFb, leptin, and TIMP-2. Secretion of FGFa was notably repressed in EPO-treated MSC when compared to the untreated MSC. In hypoxiatreated MSC, TNF-a, FGFa,b, IL-12, IP-10, and TIMP-1

644

K.J. Zwezdaryk et al./ Experimental Hematology 35 (2007) 640–652

Figure 1. Human mesenchymal stem cells (MSC) express the erythropoietin receptor (EpoR). (A) Several human MSC donor pools were examined by flow cytometry for expression of known MSC cell-surface markers and the erythropoietin receptor, EPOR, all in red. Cells were harvested for flow cytometry and incubated with the indicated antibody and subsequent Alexa488-conjugated secondary antibody as detailed in Materials and methods. The grey filled-in line represents hMSC incubated with the corresponding isotype antibody control. Consistently these representative patterns were observed for each donor based on at least duplicate samples from several independently derived cell preparations (n O 6). (B) Human MSC cultured on chamber slides were examined by immunofluorescence (IF) for expression of CD105 or EPOR as indicated. Cells were incubated with the indicated primary (1 ) antibody and stained subsequently with Alexa488-conjugated secondary antibody as detailed in Materials and methods. Photomicrographs were taken with a Zeiss Axioplan 2 fluorescence microscope with Intelligent Imaging Innovations Deconvolution Software. Representative fields from recorded images (630) are shown. Similar observations were noted in three other chamber slide IF experiments (n O 2). Arrows point to positive EPOR staining. (C) Expression of EPOR, JAK2, and STAT5 in human mesenchymal stem cells (MSCs) exposed to hypoxia (2% O2) for 0 hours (N) or 24 hours (H). Cell lysates were analyzed by western blot (WB) and visualized by chemiluminescence. A representative autoradiogram is shown of results observed from triplicate assays from independently derived samples (n O 3). Actin immunodetection served as loading control. Diagram on the left is a linear representation of the signaling consequences of erythropoietin treatment.

K.J. Zwezdaryk et al./ Experimental Hematology 35 (2007) 640–652

secretion into the culture medium was consistently and significantly lower secondary to treatment (Fig. 2B). Zymograms revealed that pro-MMP2 and the activated MMP2 were present to a greater degree in EPO-treated and hypoxia-exposed MSC cultures (24 and 48 hours) when compared with results obtained from untreated condition medium (Fig. 2C). At 24 hours post EPO treatment there was a 3.2-fold increase in pro-MMP-2 secretion and 3.3-fold increase in active MMP-2 measured as compared to untreated controls. At 48 hours post EPO treatment there was an even greater increase, measured as 4.4-fold increase in pro-MMP-2 secretion and 4.7-fold increase in active MMP-2 secretion as compared to untreated controls. MSC cultures did not appear to secrete or activate the majority of other MMP since there was only minimal detectable activity in casein gels and only very slight activity observed for MMP9 following treatment as analyzed by zymography (data not shown). Erythropoietin treatment of MSC leads to in vitro capillary-like formation and improved ‘‘wound healing’’ capacity The pro-angiogenic program driven by EPO treatment of primary hMSC cultures was tested next by an in vitro Matrigel assay. This assay examines the potential for endothelial (or endothelial-like) cells to form capillary-like tubules when plated on a reconstituted basement membrane, Matrigel. Nonendothelial adherent cells generally are unable to attach to the Matrigel and do not form these tubules but remain compacted and quickly die following plating by anoikis. Human umbilical vein endothelial cells (HUVEC; shown) were plated as positive controls. To facilitate observation of the progress of the cells in the assay, cultured cells were transduced with an adenoviral vector encoding green fluorescence protein (GFP). Initial experiments revealed that primary hMSC cultures were equally as capable as endothelial cells to form tubules (Fig. 3A). Typically, the time required for hMSC cultures to aggregate into these tubules was longer than for HUVEC (6 hours vs 3 to 4 hours). The fluorescence images presented in Figure 3A illustrate our consistent observation of more complex tubules formed by MSC than those formed by equally treated HUVEC cultures. HUVEC cultures generally consisted of cells leading to single cell tubule projections (Fig. 3A). Hypoxia stimulated the rate of early tubule formation (0–6 hours), but after 6 hours, hypoxia-treated samples did not appreciably differ from those treated in room air (21% O2, data not shown). Tubule formation by HUVEC and MSC cells was most effective in serum-containing medium (20%), followed closely by recombinant EPO (40 U/mL), then recombinant VEGF (25 ng/mL). When given enough time for cell contact, tubes formed with MSC cultures grown in serum-free medium (Fig. 3B). The enhanced effect mediated by treatment of MSC with growth factors was evident

645

after treatment for 3, 20, 48, and 96 hours. Similar growth factor–mediated effects were noted for HUVEC cultures in this assay (Fig. 3B). To ensure that this ability was universal to humanderived mesenchymal stem cells, several donors were equally tested in this Matrigel assay. MSC preparations from all the donors tested were capable of tubule formation (Fig. 3C); however, some quantitative and qualitative differences exist between each donor’s ability to form tubules. The ‘‘wound-healing’’ assay is another assay to analyze pro-angiogenic properties of MSC [23]. Primary hMSC cultures grown in serum-free medium to near confluency were treated with EPO (40 U/mL) in serum-free medium for indicated time prior to image capture microscopy. Following 5 hours of treatment, a change in alignment of the cells peripheral to the ‘‘wound’’ perpendicular to the angle of the cut was observed (Fig. 3D). After 20 hours, one of the donor MSC preparations completely sealed the originally scored area. MSC cultures from the other donor tested were slower to ‘‘heal’’ the wounded area. Erythropoietin induces chemotaxis and invasion of primary hMSC MSC migration was examined by the Boyden chamber chemotaxis assay. Recombinant human EPO, VEGF, or bFGF were loaded in the lower compartment and serum-starved primary hMSC cultures were added to the upper compartment (Fig. 4). After incubation, EPO (40 U/mL), bFGF (10 ng/mL), and VEGF (25 ng/mL) addition effectively induced cell migration in serum-starved MSC. At the concentrations employed for these factors it appeared that EPO was the best chemoattractant. By contrast, no MSC chemotaxis was measured in serum-free medium. As before, the best effect was measured when serum-containing growth medium, our positive control, was used as a chemoattractant. MSC invasion was measured by the Fluoroblok invasion assay. Serum-starved primary hMSC cultures were stimulated with EPO (40 U/mL) or VEGF (25 ng/mL). Cellular invasion was most efficient when 20% serum was used (bottom panels, 80% cell invasion) as stimulant for the serum-starved hMSC. In hMSC cultures incubated with EPO, a robust 73% cell invasion was consistently measured. Similar results were detected for VEGF samples (77% invasion) (Fig. 5, top panel). Erythropoietin treatment of mesenchymal stem cells leads to greater vessel density in an in vivo mouse Matrigel plug assay MSC pro-angiogenesis behavior was tested in an in vivo murine Matrigel plug assay. This assay tested the ability of endothelial (or endothelial-like) cells or growth factors like VEGF, bFGF, and EPO, trapped in the organoid-like matrix plug formed by the Matrigel substrate, to contribute to angiogenesis when inoculated into nude mice. Primary

646

K.J. Zwezdaryk et al./ Experimental Hematology 35 (2007) 640–652

Figure 2. Erythropoietin and hypoxia affect the proliferation and the secretion of angiogenesis factors by MSC. (A) MSC were treated with increasing amounts of erythropoietin (EPO, 0–100 U/mL) in serum-free medium or 20% serum-containing medium for 0, 24, 48, 72, and 96 hours. Cell proliferation from each sample was measured by the CellTiter96 Aqueous One Assay (Promega) as indicated in Materials and methods. Data shown are from 48-hour time point samples and are expressed relative to untreated controls. Cell proliferation assay was performed on at least three separate experiments with each sample repeated along 8 wells of the 96-well plate (n O 3). Error bars indicate SEM. (B) Angiogenesis factor antibody array analysis of the conditioned medium from treated hMSC as indicated in Materials and methods (Panomics Inc, CA, USA). Topmost panels are autoradiograms of arrayed blots. Representative blots from samples of 24-hour normoxia (21% O2), hypoxia (2% O2), and EPO (80 U/mL)-treated culture medium. The bottom panel is a map of the represented angiogenesis-related factors’ arrayed antibodies within the blots that significantly changed. Note there are two wells spotted for each arrayed antibody (duplicate spots for each factor). (C) The conditioned medium from treated hMSC was analyzed by zymography. Gelatin zymography analysis revealed MMP2 activation in the conditioned medium harvested from control (N), EPO-treated (40 U/mL) and hypoxia-exposed (2% O2) for 24 and 48 hours as indicated. Representative micrograph shown based on results observed from duplicate zymography analysis performed from three separate experiments (n O 3). Activated MMP2 form marked by asterisk in contrast to inactive pro-MMP2 band detected and marked by bar. The relative fold changes of each measured MMP form from the untreated ‘‘N’’ control sample are represented below the micrograph.

hMSC cultures were cultivated in serum-free medium overnight before harvesting. On the day of animal inoculation, approximately 250,000 serum-starved cells were labeled (CellTracker) and added to EPO-containing Matrigel prior to subcutaneous injection on one side of the mouse. The opposite side was injected with EPO-containing Matrigel without cells. There were also standard assay controls used in the experiments: Matrigel plugs containing no added pro-angiogenic factor or confirmed pro-angiogenic factors (VEGF and bFGF) injected subcutaneously into the animals with or without serum-starved hMSC prepara-

tions. Our initial pilot experiments revealed that hMSC were just as capable as endothelial cells to promote increased vessel density. Without exception, primary hMSC cultures from all tested donors consistently contributed to observed increased vessel density in this assay (Figs. 6 and 7). MSC-treated plugs consistently looked ‘‘bloodier’’ than untreated representative animals, whose plugs looked colorless. Additionally, hMSC-containing plugs were usually larger in size than those injected without added cells. This reflects more successful blood vessel recruitment by

K.J. Zwezdaryk et al./ Experimental Hematology 35 (2007) 640–652

647

Figure 3. EPO-treated MSC form capillary-like tubes in vitro and improve ‘‘wound healing.’’ (A) MSC and HUVEC were transfected with an adenoviral vector encoding for the green fluorescent protein (Ad-GFP, MOI 5 30) and serum-starved overnight, before being plated on reduced growth factor (RGF)Matrigel-coated chamber slides as indicated. Nuclei were visualized by staining with DAPI. Phase contrast (100 magnification, leftmost panel) and fluorescence (200 magnification, two right panels) photomicrographs were taken following 21 hours of incubation with a Nikon Eclipse TE300 inverted fluorescence microscope. Shown are representative images of tubule formation from greater than six independently executed experiments (n O 6). (B) MSC and HUVEC were serum-starved overnight before being plated on RGF-Matrigel-coated 24-well plates for in vitro Matrigel assay as before. MSC and HUVEC had accelerated capillary-like tubule formation in 10% serum-containing medium, followed by serum-free þ EPO (40 U/mL) or serum-free þ VEGF (25 mg/mL) medium. Serum-free culture conditions resulted in the slowest tubule formation. Phase-contrast micrographs were taken at 20 hours (100 magnification). Included are representative images of tubule formation from greater than three independently performed experiments (n O 3). (C) MSC obtained from the bone marrow of 4 human donors show capillary-like tubule formation on the in vitro Matrigel angiogenesis assay. MSC were serum-starved overnight before being plated on RGF-Matrigel-coated chamber slides. The cells were then treated with EPO 40 U/mL. Phase-contrast micrographs were taken after 21 to 23 hours as before. Depicted are representative images of tubule formation from greater than three independently derived experiments (n O 3) at a magnification of 100 or 200. (D) Confluent human MSC cultures from two donors in serum-free medium were wounded with a pipette tip and treated with EPO (40 U/mL) for the indicated times (0, 5 hours, and 20 hours). Black dashed lines mark initial wounded area. Yellow arrows point out cells growing towards indicated wounded site. Photomicrographs were taken by phase-contrast microscopy (40 magnification). Depicted are representative images from greater than three independently derived experiments (n O 3).

the hMSC-containing plugs than those without hMSC (bar graph insert, Fig. 6). Representative micrographs of H&E-stained sections (Fig. 6) along with sections that were stained with PECAM (CD31) endothelial vessel marker antibody confirmed successful vessel recruitment into the plugs (Fig. 7). Quan-

tification of the vessel number recruited into each experimental plug was achieved by counting the vessels in at least five high-magnification fields from at least four independent experiments for representative samples (Fig. 6). For all experiments, plugs without cells resulted in the lowest vessel recruitment that is in agreement with published

648

K.J. Zwezdaryk et al./ Experimental Hematology 35 (2007) 640–652

Figure 4. EPO treatment of MSC leads to increased chemotaxis. Human MSC were serum-starved and subsequently added to a Boyden chamber that contained serum-free medium (top panel) or 20% serum growth medium (bottom panel) and the following growth factors added to the bottom well as indicated: 25 ng/mL of VEGF or 10 ng/mL of bFGF, or 40 U/mL EPO. The Boyden chamber apparatus was then incubated at 37 C to allow for cell migration. The cells that migrated through the porous membrane were stained and counted as outlined in Materials and methods. Included on bottom panel are representative images of migrated MSC from greater than three independently performed experiments done in triplicate (n O 3). Top panel bar graph represents quantification of the observed results. Error bars indicate SEM.

reports that addition of VEGF and bFGF to plugs leads to increased vascularization [10,24]. Hypoxia preexposure (20–24 hours) of hMSC resulted in improved recruitment of vessels into plugs (data not shown). Consistently, EPOcontaining plugs yielded enhanced vessel recruitment whether hMSC cells were present or not (Fig. 6). The obtained plug sections were incubated with a DAPI solution to stain nucleic acids (blue) and also indirectly stained with an antibody specific for PECAM (CD31) or EPOR followed by AlexaFluor 488–linked secondary antibody staining for fluorography as indicated (green). Interestingly, hMSC prelabeled with CellTracker (red) are found mostly surrounding plug vessels (Fig. 7, yellow arrows).

Discussion The exciting potential of stem cell–based therapies for the treatment and repair of diseased organs or tissues requires knowledge of how the stem cells, whether modified or not, will behave when reintroduced into the human host. In this study, the contribution of mesenchymal stem cells to facilitate angiogenesis was evaluated in the context of

hypoxia-regulated factors whose secretion is frequently enhanced in many tumor microenvironments. Our data demonstrate for the first time that hMSC have EPO receptors and that EPO can also be added to the list of hypoxiaregulated factors that drives a pro-angiogenesis program within the MSC. It was shown that EPO receptors are prevalent in MSC as documented by flow cytometry, immunofluorescence, and western blot analysis (Fig. 1). This observation may help explain the clinical finding that EPO treatment of patients results in increased mobilization and detection of MSC in peripheral blood [25–28]. It is plausible that improved recovery following ischemia and infarcts partially stem from EPO-activated MSC [29]. It is clear that EPOR on these cells serves a chemokine receptor–like function for the mobilization of the MSC from the bone marrow and tissue niches [11]. It may be that EPOR expression is vestigial to the role in early extra-embryonic angiogenesis for MSC progenitors [30]. The downstream signaling of the EPOR through JAK2 and STAT5 phosphorylation can explain the improved survival, proliferation, and mobilization of the treated hMSC since all of these functions have been assigned to these molecules [8–10].

K.J. Zwezdaryk et al./ Experimental Hematology 35 (2007) 640–652

649

Figure 5. EPO treatment of MSC leads to increased invasion. Human MSC were serum-starved and added to individual transwell inserts containing a Matrigel-coated FluoroBlok porous membrane (COAT) or, as a control, serum-starved MSC were also loaded on uncoated FluoroBlok porous membrane (NO COAT). The membranes were lowered into 24-well companion plates containing either serum-free medium (top panels), 20% serum growth medium (bottom panels), or 25 ng/mL of VEGF or 40 U/mL EPO in serum-free growth medium. Photomicrographs shown were taken following 16 hours of incubation with a Nikon Eclipse TE300 inverted fluorescence microscope. Included on the bottom panel are representative images of migrated MSC from greater than three independently performed experiments done in triplicate (n O 3). Top panel bar graph represents quantification of the observed results. Error bars indicate SEM.

EPOR engagement by erythropoietin led to increased proliferation in MSC by standard proliferation measurement assays. Similar effects have been reported for EPOtreated human vascular endothelial cells [10]. These latter cells, however, do not tolerate long-term hypoxia exposure so this parameter was not included in this report. Hypoxia only increased MSC proliferation when these cells were cultured at low confluency. This finding is in agreement with a recent report; however, a higher level of oxygen (8%) was employed when compared to the concentration used in this study (2%) [31]. It was consistently observed that low levels of erythropoietin (!10 U/mL) served a proliferative function whereas higher levels of EPO (O30 U/ mL) were more successful for differentiation assays like migration, invasion, and vessel formation in treated MSC. Similar occurrences have been reported for EPO-treated endothelial-like cells [8–10]. Our findings also agree with reports that both erythropoietin and hypoxia treatment of endothelial-like cells leads to increased pro-MMP2 activation and various pro-angiogenic factor secretion [10,16]. Notably, increased VEGF secretion and activation of MMP2 following erythropoietin treatment of MSC are usually observed in the early initiation phase of angiogenesis. The release of these angiogenic factors jus-

tifies the successful late-phase pro-angiogenesis recorded in the tubule and plug assays. Because this effect was consistently observed for all human donors tested, the results indicate that this is a common capability of human-derived MSC. The fact that these late-phase angiogenesis assays in our hands required greater incubation times might be explained by the notion that typical MSC preparations consist of a heterogeneous population of cells at various stages of differentiation. In this context reprogramming toward proangiogenesis pathways after EPO treatment could require longer incubation times. This is in contrast to the EPO treatment of more homogeneous, single-cell human endothelial cell cultures (e.g., HUVEC) that are already fully differentiated toward an angiogenesis program. This heterogeneity was consistently observed even in simple assays like the ‘‘wound-healing’’ assay presented, where one donor took longer than the other to close the wound (Fig. 3C). Perhaps the donor with slower wound closure time had a smaller proportion of pro-angiogenic MSC. Although longer incubation times were necessary for successful assay results, the consistent pro-angiogenesis effect of VEGF, bFGF, or EPO incubation within the MSC in all assays argues that this is a true MSC capability. EPO treatment of MSC in the collagen-based Boyden chamber

650

K.J. Zwezdaryk et al./ Experimental Hematology 35 (2007) 640–652

Figure 6. EPO treatment of MSC leads to increased vessel density in an in vivo angiogenesis plug assay. Human MSC (MSC 0.25  106 cells/plug) in EPO (40 U/mL or 80 U/mL) containing serum-free media were mixed with 400 mL of RGF-Matrigel on ice and then injected subcutaneously into the lower back (right side) of BALB/c nu/nu athymic mice. Mice were sacrificed 7 days later and hardened Matrigel plugs formed at the site of injection (depicted by bar graph inset wet mount picture EPO alone or EPO þ MSC right side) were surgically removed and examined histologically after hematoxylin-eosin staining (bottom panel). Shown in lower panels are representative images of H&E-stained sections demonstrating vessel formation ability from greater than six independently executed experiments (n O 6). Panels included: No cells, harvested Matrigel plugs containing no added cells or added pro-angiogenic factor; No cells þ EPO, Matrigel plug containing no MSC, only pro-angiogenic EPO, 40 U/mL; No cells þ VEGF, Matrigel plug containing no MSC, only proangiogenic VEGF, 25 ng/mL; MSC alone, Matrigel plug containing only MSC, no pro-angiogenic factors; MSC þ EPO, Matrigel plug containing MSC and pro-angiogenic EPO, 80 U/mL; MSC þ VEGF, Matrigel plug containing MSC and pro-angiogenic VEGF, 25 ng/mL. Top panel bar graph represents quantification of the observed results. Error bars indicate SEM. Both EPO 40 U/mL and 80 U/mL treatment of MSC containing plugs resulted in statistically significant (p ! 0.05) increased blood vessel formation when compared to serum-free control.

assay was more effective than the other pro-angiogenesis factors, VEGF or bFGF (Fig. 4); however, this observation was not found in the Matrigel-based Fluoroblok invasion assay where equal effects were noted for all three factors (Fig. 5). EPO diffusion differences among the matrices used in these assays may explain these results. In other reports, these three factors have not been similarly used side by side in chemotaxis and invasion assays. EPO treatment of MSC led to pro-angiogenesis comparable to the effects recorded for VEGF- and bFGF-derived plugs in the in vivo plug assay. This observation can similarly be explained by the EPO diffusion differences in the Matrigelbased plugs. In this assay, it might be predicted that if collagen-based plugs had been employed, enhanced vessel recruitment would be noted for EPO-treated plugs when compared to VEGF- and bFGF-treated plugs. Regardless of the amount of pro-angiogenesis factor tested, EPOcontaining plugs effectively recruited blood vessels as recorded in immunofluorescence assays (Fig. 7).

The most interesting observation made was that although EPO-, VEGF-, and bFGF-treated MSC within the plugs effectively enhanced recruitment of vessels, seldom were those vessels composed of the MSC (Fig. 7). There is only one vessel cell stained (as the yellow arrows indicate) that is itself a member of the recruited vessel. Instead the originally stained MSC are mostly found surrounding the vessels (Fig. 7), suggesting that MSC serve a supportive role in secretion of trophic endothelial growth factors. This observation would dispute the implied role for these cells in tumor angiogenesis previously reported [19]. However, this notion is consistent with the classical hematopoietic stem cell supportive role assigned to MSC. Although the results presented here and elsewhere concerning the invasive and migratory ability of MSC argues for an endothelial-like behavior, it may be that this parallel endothelial-like behavior is a prerequisite for endothelial migration and invasion for new blood vessel formation. MSC-mediated growth factor secretion, migration, and invasion may lead the way for the later-arriving endothelial

K.J. Zwezdaryk et al./ Experimental Hematology 35 (2007) 640–652

651

Figure 7. Immunofluorescence assay of sections obtained from the in vivo angiogenesis plug assay. Fluorescence immunohistochemistry was done on corresponding sections with primary antibodies to the endothelial cell marker PECAM (CD31, top panels) and erythropoietin receptor (EPOR, middle panels) both labeled GREEN after Alexa488-linked secondary antibody incubation. Nuclei were stained BLUE with DAPI and prior to incorporation into the Matrigel plugs, MSC were stained RED with CellTracker (Molecular Probes). Yellow arrows mark some of the RED CellTracker-labeled MSC found peripheral to the vessels. Note that double labeling of CellTracker RED and AlexaFluor 488 that appears yellow is also evident for both CD31 and EPOR-stained MSC. As a control, an isotype primary antibody was used for incubations shown in the bottom panel sections. Photomicrographs were taken with a Zeiss Axioplan 2 fluorescence microscope with Intelligent Imaging Innovations Deconvolution Software. Representative fields from 400-magnified images are included. Similar observations were noted in IF experiments from greater than six independently performed in vivo plug assays (n O 6).

cell components of the vessels. Experiments to support this notion are currently under way. The experiments presented here conclusively demonstrate the ability of erythropoietin to elicit a pro-angiogenesis program on primary human MSC cultures. As such, erythropoietin becomes yet another hypoxia-regulated factor that drives a pro-angiogenesis program within the MSC that then promotes local angiogenesis.

Acknowledgments This work was supported by the following grants: NIH AI056229, NIH 1P20RR20152-01, and a research grant award from Cancer Association of Greater New Orleans (CAGNO). The authors would like to thank the laboratory of Donald Phinney for their generous assistance with initially acquiring and cultivating the human mesenchymal stem cells. The laboratories of Cindy B. Morris and Deborah E. Sullivan were also invaluable resources in the optimization of the angiogenic assays. We would also like to thank Joanna DeSalvo for her assistance in the preparation of

figures and Elizabeth Norton and Dr. Ruth Waterman for insightful discussions of our work.

References 1. Bunn HF, Gu J, Huang LE, Park JW, Zhu H. Erythropoietin: a model system for studying oxygen-dependent gene regulation. J Exp Biol. 1998;201:1197–1201. 2. Chikuma M, Masuda S, Kobayashi T, Nagao M, Sasaki R. Tissuespecific regulation of erythropoietin production in the murine kidney, brain, and uterus. [In Process Citation]. Am J Physiol Endocrinol Metab. 2000;279:E1242–E1248. 3. Masuda S, Kobayashi T, Chikuma M, Nagao M, Sasaki R. The oviduct produces erythropoietin in an estrogen- and oxygen-dependent manner. Am J Physiol Endocrinol Metab. 2000;278:E1038–E1044. 4. Ogilvie M, Yu X, Nicolas-Metral V, et al. Erythropoietin stimulates proliferation and interferes with differentiation of myoblasts. [In Process Citation]. J Biol Chem. 2000;275:39754–39761. 5. Sasaki R, Masuda S, Nagao M. Pleiotropic functions and tissue-specific expression of erythropoietin. News Physiol Sci. 2001;16:110–113.

652

K.J. Zwezdaryk et al./ Experimental Hematology 35 (2007) 640–652

6. Sakanaka M, Wen TC, Matsuda S, et al. In vivo evidence that erythropoietin protects neurons from ischemic damage. Proc Natl Acad Sci U S A. 1998;95:4635–4640. 7. Junk AK, Mammis A, Savitz SI, et al. Erythropoietin administration protects retinal neurons from acute ischemia-reperfusion injury. Proc Natl Acad Sci U S A. 2002;99:10659–10664. 8. Anagnostou A, Lee ES, Kessimian N, Levinson R, Steiner M. Erythropoietin has a mitogenic and positive chemotactic effect on endothelial cells. Proc Natl Acad Sci U S A. 1990;87:5978–5982. 9. Anagnostou A, Liu Z, Steiner M, et al. Erythropoietin receptor mRNA expression in human endothelial cells. Proc Natl Acad Sci U S A. 1994;91:3974–3980. 10. Ribatti D, Presta M, Vacca A, et al. Human erythropoietin induces a pro-angiogenic phenotype in cultured endothelial cells and stimulates neovascularization in vivo. Blood. 1999;93:2627–2636. 11. Muller-Ehmsen J, Schmidt A, Krausgrill B, Schwinger RH, Bloch W. Role of erythropoietin for angiogenesis and vasculogenesis: from embryonic development through adulthood. Am J Physiol Heart Circ Physiol. 2006;290:H331–H340. 12. Wu H, Lee SH, Gao J, Liu X, Iruela-Arispe ML. Inactivation of erythropoietin leads to defects in cardiac morphogenesis. Development. 1999;126:3597–3605. 13. Yu X, Shacka JJ, Eells JB, et al. Erythropoietin receptor signalling is required for normal brain development. Development. 2002;129:505– 516. 14. Acs G, Acs P, Beckwith SM, et al. Erythropoietin and erythropoietin receptor expression in human cancer. Cancer Res. 2001;61:3561– 3565. 15. Aicher A, Brenner W, Zuhayra M, et al. Assessment of the tissue distribution of transplanted human endothelial progenitor cells by radioactive labeling. Circulation. 2003;107:2134–2139. 16. Annabi B, Lee YT, Turcotte S, et al. Hypoxia promotes murine bone marrow–derived stromal cell migration and tube formation. Stem Cells. 2003;21:337–347. 17. Liu W, Reinmuth N, Stoeltzing O, et al. Antiangiogenic therapy targeting factors that enhance endothelial cell survival. Semin Oncol. 2002;29:96–103. 18. Moore MA. Putting the neo into neoangiogenesis. J Clin Invest. 2002; 109:313–315.

19. Rafii S, Lyden D, Benezra R, Hattori K, Heissig B. Vascular and haematopoietic stem cells: novel targets for anti-angiogenesis therapy? Nat Rev Cancer. 2002;2:826–835. 20. Lamouille S, Mallet C, Feige JJ, Bailly S. Activin receptor–like kinase 1 is implicated in the maturation phase of angiogenesis. Blood. 2002; 100:4495–4501. 21. Yasuda Y, Fujita Y, Masuda S, et al. Erythropoietin is involved in growth and angiogenesis in malignant tumours of female reproductive organs. Carcinogenesis. 2002;23:1797–1805. 22. Yasuda Y, Okano M, Nagao M, Masuda S, Fujita Y, Sasaki R. Erythropoietin and erythropoietin-receptor producing cells demonstrated by in situ hybridization in mouse visceral yolk sacs. Anat Sci Int. 2002; 77:58–63. 23. Lee H, Goetzl EJ, An S. Lysophosphatidic acid and sphingosine 1– phosphate stimulate endothelial cell wound healing. Am J Physiol Cell Physiol. 2000;278:C612–C629. 24. Al-Khaldi A, Eliopoulos N, Martineau D, Lejeune L, Lachapelle K, Galipeau J. Postnatal bone marrow stromal cells elicit a potent VEGFdependent neoangiogenic response in vivo. Gene Ther. 2003;10:621–629. 25. Bahlmann FH, De Groot K, Spandau JM, et al. Erythropoietin regulates endothelial progenitor cells. Blood. 2004;103:921–926. 26. George J, Goldstein E, Abashidze A, et al. Erythropoietin promotes endothelial progenitor cell proliferative and adhesive properties in a PI 3-kinase-dependent manner. Cardiovasc Res. 2005;68:299–306. 27. George J, Patal S, Wexler D, et al. Circulating erythropoietin levels and prognosis in patients with congestive heart failure: comparison with neurohormonal and inflammatory markers. Arch Intern Med. 2005;165:1304–1309. 28. Heeschen C, Aicher A, Lehmann R, et al. Erythropoietin is a potent physiologic stimulus for endothelial progenitor cell mobilization. Blood. 2003;102:1340–1346. 29. Tsai PT, Ohab JJ, Kertesz N, et al. A critical role of erythropoietin receptor in neurogenesis and post-stroke recovery. J Neurosci. 2006; 26:1269–1274. 30. Kertesz N, Wu J, Chen TH, Sucov HM, Wu H. The role of erythropoietin in regulating angiogenesis. Dev Biol. 2004;276:101–110. 31. Ren H, Cao Y, Zhao Q, et al. Proliferation and differentiation of bone marrow stromal cells under hypoxic conditions. Biochem Biophys Res Commun. 2006;347:12–21.