Angiogenesis in Differentiated Placental Multipotent Mesenchymal Stromal Cells Is Dependent on Integrin α5β1

Human placental multipotent mesenchymal stromal cells (hPMSCs) can be isolated from term placenta, but their angiogenic ability and the regulatory pathways involved are not known. hPMSCs were shown to express integrins αv, α4, α5, β1, β3, and β5 and could be induced to differentiate into cells expressing endothelial markers. Increases in cell surface integrins α5 and β1, but not α4, αvβ3, or αvβ5, accompanied endothelial differentiation. Vascular endothelial growth factor-A augmented the effect of fibronectin in enhancing adhesion and migration of differentiated hPMSC through integrin α5β1, but not αvβ3 or αvβ5. Formation of capillary-like structures in vitro from differentiated cells was inhibited by pre-treatment with function-blocking antibodies to integrins α5 and β1. When hPMSCs were seeded onto chick chorioallantoic membranes (CAM), human von Willebrand factor-positive cells were observed to engraft in the chick endothelium. CAMs transplanted with differentiated hPMSCs had a greater number of vessels containing human cells and more incorporated cells per vessel compared to CAMs transplanted with undifferentiated hPMSCs, and overall angiogenesis was enhanced more by the differentiated cells. Function-blocking antibodies to integrins α5 and β1 inhibited angiogenesis in the CAM assay. These results suggest that differentiated hPMSCs may contribute to blood vessel formation, and this activity depends on integrin α5β1.


Introduction
Multipotent mesenchymal stromal cells from different adult and fetal tissues have been shown to have the potential to differentiate into endothelial cells [1][2][3][4]. Human placental multipotent mesenchymal stromal cells (hPMSCs) express genes associated with ectoderm, endoderm and mesoderm, including hematopoietic/endothelial cell-related transcripts [5]. They can differentiate into osteogenic, adipogenic, chondrogenic, and neurogenic cell lineages [6][7][8], and, under the combined influence of growth factors and mechanical shear stress, have been reported to acquire aspects of the endothelial phenotype [9]. However, the angiogenic ability of these cells is not well characterized.
Angiogenesis is a complex process that involves extracellular matrix (ECM) remodeling, endothelial cell differentiation, migration and proliferation, and the functional maturation of new endothelial cell colonies into mature blood vessels [10]. There is evidence that fibronectin is a key ECM component at several stages, initially providing attachment sites for precursor cells [11,12], then promoting vascular endothelial growth factor (VEGF)-induced differentiation to endothelial cells [13]. Furthermore, fibronectin associated with VEGF-A enhances endothelial cell migration [14]. Targeted gene deletion studies have revealed that fibronectin functions in vascular stabilization and branching morphogenesis in the murine embryo [15][16][17][18], while a more restricted gene targeting approach that deletes alternatively spliced variants of fibronectin leads to defective placental angiogenesis [19]. Fibronectin is abundant in the mesenchymal compartment of human placenta where vasculogenesis and angiogenesis occur [20].
Integrin a 5 b 1 is a selective high affinity receptor for fibronectin, and a regulator of VEGF-A signaling [21]. Integrin a 5 b 1 is observed to have an essential role in pathological neovascularization in cornea [22] and is up-regulated in newly growing vessels in embryos and tumors [21,23,24]. Vascularization in the placenta is critical for normal delivery of nutrients to the fetus. Placental growth is most rapid in the first half of pregnancy, but development of the vascular tree continues to term [25,26]. Placental vascular defects, including reduced vessel density, are associated with fetal growth restriction [27,28]. VEGF-A is thought to play an important role in human placental vascularization, especially in the early stages [29].
Improved understanding of the cellular and molecular mechanisms of placental vasculogenesis and angiogenesis could potentially lead to treatments to achieve improved pregnancy outcome as well as the possibility of using placental progenitor cells in therapeutic applications. Thus, the aims of this study were to investigate if hPMSCs are capable of functional differentiation into endothelial cells, and to investigate the role in this process of integrin a 5 b 1 and its interaction with fibronectin in the presence of VEGF-A.

Materials and Methods
Isolation and culture, of placenta-derived cells Clinically normal human placentas (37 to 40 weeks of gestation, n = 30) were obtained after cesarean section. Tissue was collected after written informed consent was obtained, and this study was approved by the Institutional Review Board of Mackay Memorial Hospital, Taipei. The animal studies were specifically approved by the ethics committee of Mackay Memorial Hospital for animal experimentation and were conducted following the institution's guidelines for animal husbandry.
Isolation of hPMSCs was performed as we described previously [30]. Briefly, about 100 g of tissue from central placental cotyledons was minced, trypsinized (0.05% trypsin-EDTA solution; Invitrogen), and treated with 10 U/ml DNase I (Sigma-Aldrich) in Dulbecco's modified Eagle's medium (DMEM; Gibco) at 37uC for 5 min several times, and finally filtered through a cell strainer (BD Biosciences). The supernatants were pooled and centrifuged, and the mononuclear cells in the supernatants were recovered by Percoll density gradient fractionation (1.073 g/ml, Sigma-Aldrich) [31]. The cells were resuspended and seeded in a flask, and were maintained in DMEM with 10% fetal bovine serum (FBS; Hyclone) at 37uC in a humidified atmosphere with 5% CO 2 . Approximately 2 to 3 weeks later, some colonies containing fibroblast-like cells were observed. These cells were trypsinized and replated for expansion after 70% confluence.

Flow cytometry
The cell phenotype of hPMSCs was characterized by a panel of PE-or FITC-conjugated antibodies purchased from Serotec, Calbiochem, Chemicon, R&D systems, GeneTex, or BD Biosciences using standard fluorescence-activated cell sorting analysis and CellQuest software. The cells were permeabilized using BD Cytofix/Cytoperm TM Fixation/Permeabilization Kit (BD) according to the manufacturer9s recommendation before flow cytometry assay for Oct-4 (Chemicon).
Alterations of integrin abundance on the cell surface of hPMSCs were evaluated by fluorescence intensity, an index of the surface concentration of integrin per cell. Mean fluorescence is the estimated x axis value at maximum peak height. This parameter is measured independently from the proportion of positive cells, estimated as the difference between the mean fluorescence (the surface concentration of integrin molecules) of positive cells and that of the control samples, regardless of the frequency of positive cells [32].

Differentiation induction
Induction of osteogenic or adipogenic differentiation.
Induction of endothelial cell differentiation. hPMSCs were seeded at a density of 1610 3 cells/cm 2 in petri dishes and cultured in endothelial cell growth medium 2 (EGM2; Promocell) supplemented with Supplement Mix (Promocell) which contains 1 mg/ml ascorbic acid, 10 ng/ml human recombinant basic fibroblast growth factor, 5 ng/ml human recombinant epidermal growth factor, 22.5 mg/ml heparin, 0.2 mg/ml hydrocortisone, 20 ng/ml long R3 insulin like growth factor-1, 0.62 ng/ml phenol red, 0.5 ng/ml human recombinant VEGF-A, and 2% FBS (Hyclone). Additionally, 50 ng/ml VEGF-A (Chemicon) was added to medium for induction of differentiation. The cultures were maintained for 14 to 21 days and the culture medium was replaced every three days [1,4].

In vitro angiogenesis
Induction of capillary tube formation was performed using an In Vitro Angiogenesis kit (Chemicon) as recommended by the manufacturer. VEGF-A-induced differentiated hPMSCs cells were preincubated (30 min at 37uC) in HBSS containing either 3 mg/ml non-specific mouse IgG (Dako) or function-blocking monoclonal antibody specific to integrin b 1 (HUTS-4; 1:1000; Chemicon), a 4 (P4G9; 1:1000; Chemicon), a 5 (SAM-1; 1:100; Chemicon), a n b 3 (LM609; 1:100; Chemicon), or a n b 5 (P1F6; 1:1000; Chemicon). After blocking with these antibodies, cells were resuspended to 1610 5 cells/ml in EGM2, after which 100 ml/well of cell suspension was added onto a gel of polymerized basement membrane-like material (ECMatrix TM ). The cells were incubated for 6 hours at 37uC for full development of capillary-like network structures. Tube formation was quantified by counting the number of polygonal tubes as well as cumulative tube length (long axis). In instances where several tube-like structures merged together or branched, the total length of the tubes was calculated as the sum of the length of the individual branches [39]. Results are represented as total tube length (mm) or number for six random photographic fields per experimental condition (magnification 506; Axiovert 200; Carl Zeiss MicroImaging). The experiment was independently repeated three times.

In vivo angiogenesis by chick chorioallantoic membrane assay
The chick chorioallantoic membrane (CAM) assay was modified from a prior report [40]. CAM was exposed by cutting a window (2 cm 2 ) on one side of 10-day-old specific pathogen free chicken eggs. A 3-mm-thick sterile straw disk, 8 mm in diameter, was placed on the CAM for 3-dimensional culture on an area with a minimum of small blood vessels. A total of 1610 5 of hPMSCs pretreated either with anti-integrin b 1 (HUTS-4; 1:1000; Chemicon), a 4 (P4G9; 1:1000; Chemicon), a 5 (SAM-1; 1:100; Chemicon) or non-specific mouse IgG (1:50; Dako) and resuspensed in 100 ml EGM2 were placed into the straw. The window in the shell was sealed with adhesive tape and the egg was incubated for 48 hours.
Representative CAMs from each treatment group were photographed under a dissecting microscope (106) and counted. The number of fine blood vessel branch points in the region of the sample was counted. As angiogenesis is characterized by the sprouting of new vessels in response to hPMSCs, counting blood vessel branching points is a useful quantitative means of obtaining an angiogenic index. At least 6 embryos were used per treatment group. Data were evaluated in terms of average number of blood vessel branching points per treatment group 6 SD. CAMs were further excised, cryopreserved, cut into 5-mm sections, and immunostained with vWF, CD31 and CD105 as described above. For cell tracking, hPMSCs with or without induction into endothelial differentiation by VEGF-A were labeled with green fluorescence dye CellTracker TM CMFDA (5 mM, Molecular Probes, Invitrogen) before implanting into CAM. The CAM was immunostained by primary antibody against human vWF (1:800, Sigma-Aldrich) after cell transplantation. Quantification of new vessel formation that contained hPMSC incorporation or the number of hPMSCs that incorporated into the endothelia of each vessel were conducted in 10 fields (4006) for each section from 3 randomly selected sections of each CAM tissue.

Statistical analysis
The measurements of counting were conducted blind by two independent observers. The intra-class correlation coefficient for intra-and inter-rater reliability of cell counts .0.75 was considered good agreement. The data are described as means 6 SD. Differences were assessed using the independent-samples t test, paired-samples t test, or Mann Whitney U test when appropriate. A P value of less than 0.05 was considered significant. The statistical software used is SSPS version 12.0 (Chicago, IL,USA).
RT-PCR analysis of total RNA from cultured hPMSCs showed expression of Oct-4, Nanog and Sox-2, transcription factors associated with pluripotency [41][42][43] (Fig. 1B). hPMSC multipotency was demonstrated using standard osteocyte and adipocyte differentiation assays. In the former case hPMSCs formed nodules and stained with Alizarin Red S, indicating calcium salt crystallization and osteogenic differentiation (Fig. 1D). In the latter assay, hPMSCs developed Oil Red O-positive cytoplasmic lipid droplets indicating adipocyte differentiation (Fig. 1E). This was not seen in control cells (Fig. 1F).
Endothelial differentiation of hPMSCs was further revealed by mRNA analysis. Two different strains of hPMSCs were used for comparison before and after differentiation. Consistent with a previous report [5], the hPMSCs express various hematopoietic genes. Transcription from the CD34 and CD105 genes was variable between hPMSC isolates. However, after differentiation, the cells showed increased levels of mRNA encoding CD31, VEcadherin, VEGFR-1, VEGFR-2, and vWF (Fig. 2L).
Immunofluorescence staining demonstrated integrins a 4 , a 5 , b 1 , a v b 3 , and a v b 5 in undifferentiated hPMSCs. Integrin a 5 (Fig. 3E, 3F) and b 1 (Fig. 3A, 3B) were significantly increased after endothelial differentiation, but not integrin a 4 (Fig. 3C, 3D), a v b 3 (Fig. 3G, 3H), or a v b 5 (Fig. 3I, 3J). Cell surface levels of integrin a v , a 4 , a 5 , b 1 , b 3 , and b 5 subunits were assayed by flow cytometry before and after endothelial differentiation. Mean specific fluorescence (which corresponds to the increase in fluorescence intensity relative to second antibody alone) was higher for both integrins a 5 and b 1 in differentiated hPMSCs (Fig. 3K, 3L). This suggests a significant increase in cell surface integrin a 5 b 1 expression as the cells differentiated. In contrast, no increase was observed in expression of the avb3 or avb5 integrin, which have also been implicated in binding of fibronectin, vitronectin and other RGDcontaining ligands, and in angiogenesis [21,44,45].

promote VEGF-A-induced differentiated hPMSC adhesion and migration
Since integrin a 5 b 1 is a specific receptor for fibronectin, and angiogenesis often involves endothelial cell adhesion and migration within a fibronectin-rich ECM, we investigated the ability of differentiated cells to interact with fibronectin. As the cells also express integrins a v b 3 and a v b 5 , vitronectin was used as a control ligand. Differentiated hPMSCs adhered to fibronectin more efficiently than to vitronectin (P,0.001). VEGF-A significantly increased the adhesion of differentiated hPMSCs to fibronectin but not to vitronectin-or BSA-coated plates (Fig. 4A). Significant inhibition of differentiated hPMSC attachment to fibronectin-coated plates was observed in the presence of blocking antibodies to integrin a 5 and b 1 , whereas very limited inhibition effect was observed in the presence of antibody to a v b 3 and a v b 5. Attachment to vitronectin was low, and blocking antibodies had no significant effect (Fig. 4B). These experiments indicated that integrin a 5 b 1 on the surface of differentiated hPMSCs is important for mediating cell attachment to fibronectin, with a further contribution from a v b 3 .
Migration of differentiated hPMSC was enhanced by either VEGF-A (50 ng/ml) or fibronectin (50 mg/ml) alone, but the combination of VEGF-A and fibronectin produced an additive effect (Fig. 4C). In contrast, neither vitronectin alone nor the combination of VEGF-A and vitronectin promoted migration (Fig. 4C). In the presence of antibodies to integrin a 5 or b 1 , cell migration to VEGF-A and fibronectin together was suppressed, whereas antibodies to a 4 , a v b 3 , or a v b 5 had no effect (Fig. 4D). These results suggest strongly that enhanced adhesion and migration of differentiated hPMSC in the presence of VEGF-A and fibronectin are dependent on integrin a 5 b 1 .
The ability of differentiated hPMSCs to form capillary-like structures is mediated by integrin a 5 b 1 To study the molecular mechanisms underlying capillary-like morphogenesis in differentiated hPMSCs, either differentiated hPMSCs or human umbilical vein endothelial cells (HUVECs) were seeded onto a basement membrane-like gel. HUVEC (Fig. 5A) and undifferentiated hPMSCs (Fig. 5B) were used as positive and negative controls, respectively. The undifferentiated hPMSCs showed very few capillary-like structures after 6 hours, and most cells remained rounded (Fig. 5B). The differentiated hPMSCs typically showed cytoplasmic projections, spikes and extensions, and had elongated within 6 hours, with most cells becoming integrated into capillary-like structures (Fig. 5C). Formation of these structures was strongly inhibited in the presence of antbodies to integrin b 1 (Fig. 5D) or a 5 (Fig. 5F). Blocking antibodies to integrin a 4 , a v b 3 , or a v b 5 did not show inhibitory activity (Fig. 5E, 5G, 5H; quantified in Fig. 5M). Immunostaining confirmed that cells in the capillary-like structures express the specific endothelial markers including vWF (Fig. 5J), CD31 (Fig. 5K), and CD105 (Fig. 5L). Figure 5I was the cells stained by non-specific IgG as a control.  Integrin a 5 b 1 mediates the angiogenesis of differentiated hPMSCs in vivo Differentiated or undifferentiated hPMSCs were implanted onto CAMs of ten-day-old chick eggs, and two days later the CAM was imaged (Fig. 6A-G) and vessel branching points counted (Fig. 6H). In contrast to the short-term incubation in vitro, the transplanted undifferentiated hPMSCs augmented angiogenesis (Fig. 6B, 6H). However, there was a statistically significant increase of neovascularization in the CAM transplanted by differentiated hPMSCs compared to that of undifferentiated hPMSCs or control (Fig. 6A-C and 6H). The angiogenic activity of the differentiated hPMSCs was significantly reduced by preincubation with anti-integrin a 5 or b 1 prior to transfer to the CAM (Fig. 6D and 6F). Addition of antiintegrin a 4 or non-specific mouse IgG antibody had no significant effect on angiogenic activity (Fig. 6E and 6G). Immunostaining of the CAM further revealed cells within the neovessels were positive for human vWF, CD31 and CD105 (Fig. 6J-L). The CAM without hPMSC transplantation was vWF negative (Fig. 6I). These results demonstrated that the differentiated hPMSCs contributed to the neovascularization of CAM and the angiogenic activity is mediated through integrin a 5 and b 1 .

Discussion
hPMSCs have the potential to differentiate into endothelial cells under appropriate conditions both in vitro and in vivo. The differentiated cells express a panel of endothelial cell makers including vWF, CD31, CD34, VE-cadherin and the endothelial cell receptors VEGFR-1 and VEGFR-2 in vitro and in vivo. They can form capillary tube-like structures in vitro and have a greater capacity to augment angiogenesis in the CAM assay than undifferentiated cells. The ability of differentiated hPMSCs to form new blood vessels involves integrin b 1 and a 5 , but not integrin a v , a 4 , b 3 , or b 5 .
Fibronectin and VEGF-A are important regulators of blood vessel growth [51,52]. Fibronectin is highly expressed within the hematopoietic microenvironment [53] and is involved in the adhesion of hematopoietic stem and progenitor cells [12,54,55]. Fibronectin acts as a ligand for integrins a 5 b 1 , a 4 b 1 and a v b 3 [21,44,[56][57][58]. However, hPMSC adhesion to, and migration on fibronectin are specifically dependent on subunits b 1 and a 5 , the expression of both subunits increases upon endothelial differentiation, and angiogenesis is inhibited by blocking either of these two subunits. Hence we suggest the heterodimer a 5 b 1 interacts with fibronectin in the pericellular matrix to mediate key steps in angiogenesis. VEGF-A and fibronectin together significantly promote the adhesion and migration of hPMSCs. This augmentation effect is specific to fibronectin and the a 5 b 1 integrin.
Alterations of integrin expression may contribute to angiogenesis. It has been found that VEGF-A increases the migration of human dermal microvascular endothelial cells through the upregulation of a v b 3 integrin expression [59]. Cells undergoing a TGF-b-induced angiogenic program up-regulate integrin a 5 [60]. A significant up-regulation of a 5 subunit expression in vascular cells participating in choroidal neovascularization of injured eye was observed [61]. Furthermore, during the early phase of vessel sprouting, when interacting with fibronectin-rich interstitial ECM, activated endothelial cells may utilize integrin a 5 b 1 or a v b 3 to mediate the angiogenic response, later switching to other integrin subunits once basement membrane ligands such as laminin have assembled around the new vessel [62][63][64][65]. These reports suggest that a 5 b 1 or a v b 3 may play a role in neovascularization and provide a target for therapeutic intervention [61,66]. We observed that VEGF-A increased the expression of integrin a 5 b 1 , but not a v b 3 or a v b 5 .
Targeted gene ablation reveals that successful vasculogenesis depends on integrin a 5 b 1 [18] and its ligand fibronectin [16,67], and is not strongly dependent on integrin a v b 3 [18]. Homozygotic integrin a 5 -deficient mouse embryos demonstrate vascularization disruption and die in utero with numerous morphological defects [15]. Fibronectin-deficient mice also develop defects in the yolk sac vasculature [16,67]. In integrin a 5 -null embryonic cells, development of a complex vasculature is hindered with reduced cell proliferation and increased apoptosis [17]. Integrin b 1 is required for the initiation of basement membrane formation. In integrin b 1null embryonic bodies, the complex vasculature formation is significantly delayed [68]. In contrast, vasculogenesis and angiogenesis in virtually all organs develop normally in a v -null embryos [69]. Mice lacking b 3 integrins or both b 3 and b 5 integrins not only enhance tumor growth, but have enhanced angiogenesis. Thus, neither b 3 nor b 5 integrins are essential for neovascularization [70]. Furthermore, both b 3 -null and b 5 -null mice are viable, with unaffected developmental angiogenesis and adult angiogenesis such as retinal neovascularization and wound healing [71,72]. These reports support our observations that angiogenesis by differentiated hPMSCs is independent of integrin a v , b 3 and b 5 .
In contrast to the short incubation period of in vitro angiogenesis assays, both undifferentiated and differentiated hPMSCs were observed to enhance angiogenesis in the CAM assay. hPMSCs with positive vWF staining were observed to engraft in the endothelium of CAM. Direct integration of hPMSCs into CAM vasculature may augment sprouting angiogenesis. Similar to the study of ischemic brain or heart, multipotent mesenchymal stromal cells widely incorporated into vasculature and a subset of them was capable of differentiating into endothelial cells [73,74]. However, hPMSCs after inducing endothelial cell differentiation have more cell numbers incorporated into CAM vessel and have a significantly greater angiogenic effect than that of undifferentiated hPMSC. Additionally, we cannot exclude other specific adhesion molecules or growth factors which may regulate distinct angiogenic responses. The paracrine factors expressed by hPMSCs may participate in enhancing angiogenesis other than the transdifferentiation into endothelial cells by hPMSCs [75]. Our results are the first to identify a role for integrins in the regulation of angiogenesis initiated by hPMSCs. hPMSCs are a useful model to study the role of VEGF-A in differentiation and maturation of endothelial cells, and the role of integrins during placental angiogenesis and vasculogenesis. Transplantation of hPMSCs may offer potential for treating ischemic diseases. Transplantation of hPMSCs to the ischemic limbs of SCID mice significantly improved blood vessel formation and blood flow in the affected limbs [76]. In addition these cells could be utilized in the engineering of complex tissues in whcih vascularization is an essential feature.

Author Contributions
Conceived and designed the experiments: CPC. Performed the experiments: MYL YHW CYC PCC. Analyzed the data: MYL JPH YYC JDA CPC. Contributed reagents/materials/analysis tools: JPH YYC CPC. Wrote the paper: MYL CPC.