Oxygen Tension Regulates Human Mesenchymal Stem Cell Paracrine Functions

This study examined the shift of the human mesenchymal stem cell (hMSC) cytokine signature induced by oxygen tension. Conditioned media obtained from hMSCs cultured under near anoxia exhibited significantly enhanced chemotactic and proangiogenic properties and a significant decrease in the inflammatory mediator content. These results elucidate important aspects of using MSCs in regenerative medicine, contribute to improving the efficacy of such therapies, and highlight the interest in using conditioned media in therapeutic modalities.

Abstract

Mesenchymal stem cells (MSCs) have captured the attention and research endeavors of the scientific world because of their differentiation potential. However, there is accumulating evidence suggesting that the beneficial effects of MSCs are predominantly due to the multitude of bioactive mediators secreted by these cells. Because the paracrine potential of MSCs is closely related to their microenvironment, the present study investigated and characterized select aspects of the human MSC (hMSC) secretome and assessed its in vitro and in vivo bioactivity as a function of oxygen tension, specifically near anoxia (0.1% O₂) and hypoxia (5% O₂), conditions that reflect the environment to which MSCs are exposed during MSC-based therapies in vivo. In contrast to supernatant conditioned media (CM) obtained from hMSCs cultured at either 5% or 21% of O₂, CM from hMSCs cultured under near anoxia exhibited significantly (p < .05) enhanced chemotactic and proangiogenic properties and a significant (p < .05) decrease in the inflammatory mediator content. An analysis of the hMSC secretome revealed a specific profile under near anoxia: hMSCs increase their paracrine expression of the angiogenic mediators vascular endothelial growth factor (VEGF)-A, VEGF-C, interleukin-8, RANTES, and monocyte chemoattractant protein 1 but significantly decrease expression of several inflammatory/immunomodulatory mediators. These findings provide new evidence that elucidates aspects of great importance for the use of MSCs in regenerative medicine and could contribute to improving the efficacy of such therapies.

Significance

The present study investigated and characterized select aspects of the human mesenchymal stem cell (hMSC) secretome and assessed its in vitro and in vivo biological bioactivity as a function of oxygen tension, specifically near anoxia (0.1% O₂) and hypoxia (5% O₂), conditions that reflect the environment to which MSCs are exposed during MSC-based therapies in vivo. The present study provided the first evidence of a shift of the hMSC cytokine signature induced by oxygen tension, particularly near anoxia (0.1% O₂). Conditioned media obtained from hMSCs cultured under near anoxia exhibited significantly enhanced chemotactic and proangiogenic properties and a significant decrease in the inflammatory mediator content. These findings provide new evidence that elucidates aspects of great importance for the use of MSCs in regenerative medicine, could contribute to improving the efficacy of such therapies, and most importantly highlighted the interest in using conditioned media in therapeutic modalities.

Introduction

The results of studies with small and large animals have shown that mesenchymal stem cell therapies hold great promise for the treatment of human pathologies including cardiovascular [1], neurodegenerative [2], and osteoarticular diseases [3, 4]. These promising results, however, have not yet been translated into mainstream medical therapies in the clinical milieu. The underlying reasons for this situation are not fully known, but the rapid disappearance of mesenchymal stem cells (MSCs) after in vivo administration challenges the traditional view that MSCs play a direct role because of their capacity to differentiate into functional tissue cells when implanted in injured tissues (for review, see [5, 6]). In fact, the current paradigm is that MSCs support resident progenitor cells via indirect paracrine mechanisms including the release of multiple immunomodulatory, angiogenic, and chemotactic factors (for review, see [5–7]). This hypothesis is supported by recent preclinical studies that established improved tissue repair using MSC-conditioned medium (MSC-CM) without cell implantation in the context of cardiac [8, 9], skin [10], and bone repair [11]. Last but not least, the relative contributions of the two mechanisms have been estimated to be 20%–50% of the effects resulting from stem cell differentiation, whereas the actions of the MSC-CM account for 50%–80% of the effects in the context of cardiac cell therapy [12]. Collectively, these data indicate that the broad repertoire of secreted cytokines produced by human MSCs (referred to hereafter as the “hMSC secretome”) has considerable potential in regenerative medicine.

A rational design and implementation of the next generation of MSC-based therapeutics requires elucidation of microenvironmental cues (i.e., growth factor signaling, mechanical, biophysical, and biochemical signals) that affect the MSC paracrine signature and function. Such new insights into the regulation of cell phenotype are needed to harness the effect of the MSC secretome toward meaningful therapeutic outcomes.

Oxygen tension, an established key microenvironmental cue for hMSC functions, modulates their motility [13], maintains their undifferentiated state [14–16], and supports their ability to differentiate [14, 16–18] as well as to secrete angiogenic factors [19]. However, the degree and extent of hypoxia effects on the hMSC secretive functions, including balance between their angiogenic and immunomodulatory effects, has received little attention. This lack of data is surprising, considering that MSCs are a source of a broad array of cytokines whose secretion may (at least theoretically) be dramatically impacted by oxygen tension and result in secretomes with very different bioactivity.

For these reasons, the present study investigated and characterized select aspects of the hMSC secretome (with an extended assessment of the mediators released in their supernatant milieu using Luminex technology) and assessed its in vitro and in vivo bioactivity as a function of oxygen tension. Specifically, the study focused on the near-anoxia condition and assessed key functions of hMSCs, angiogenesis, and immunomodulation and thus provided new evidence regarding the role of oxygen tension in the modulation of the hMSC paracrine signature. Exposed to near-anoxia conditions, hMSCs increase their paracrine expression of the angiogenic mediators vascular endothelial growth factor (VEGF)-A, VEGF-C, interleukin-8 (IL-8), RANTES (regulated on activation, normal T-cell expressed and secreted), and monocyte chemoattractant protein 1 (MCP-1) but significantly decrease expression of several inflammatory/immunomodulatory mediators.

Materials and Methods

Cells and Cell Culture

hMSCs were isolated from bone marrow obtained as discarded tissue during routine bone surgery from five adult donors at the Lariboisiere Hospital Paris, France, according to the French bioethics laws These cells were isolated from each patient’s bone marrow using a procedure adapted from literature reports, characterized [20], pooled at an equal ratio at passage 1, and cultured in α minimum essential medium (αMEM; Dutscher, Brumath, France, http://www.dutscher.com) under standard cell culture conditions, that is, a humidified 37°C, 5% CO₂, 95% air environment. At 80%–85% confluence, the cells were trypsinized using trypsin-EDTA (Sigma) and passaged. Cell passages 4 and 5 were used for experiments. Endothelial progenitor cells (EPCs) were isolated from human umbilical cord blood, expanded, and characterized as previously described [21]. The study was approved by the local ethics committee of Hôpital des Instructions et des Armées de Begin (Saint-Mandé, France), and the protocol (number 201008043234797) conformed to the ethical guidelines of the Declaration of Helsinki. The endothelial cell phenotype was shown by double positivity for acetylated low density lipoprotein, labeled with 1,1′-dioctadecyl-3,3,3′,3′-tetramethyl-indocarbocyanine perchlorate (DiI-AcLDL) uptake and Bandeiraea simplicifolia agglutinin-1 (BS-1) lectin binding. All the following experiments were performed during the first 30 days of culture.

In Vitro Experiments

Hypoxic Conditions

Hypoxic conditions were achieved using a well-characterized, finely controlled proOx-C-chamber system (C-Chamber, C-374; Biospherix, Redfield, NY, http://www.biospherix.com). The oxygen concentration in this chamber was maintained at either 0.1%, 5%, or 21% with the residual gas mixture composed of 5% CO₂ and balance nitrogen for the duration of the experiments with cells.

For experiments under hypoxic conditions, the hMSCs were seeded at 12,000 cells per cm² in individual wells of 24-well cell-culture plasticware and were cultured in αMEM (PAN Biotech, Aidenbach, Germany, http://www.pan-biotech.com) containing 5 g/liter of glucose but no serum. In order to ensure constant oxygen levels, the hMSC cultures were maintained undisturbed and without supernatant medium change until the end of the experiments.

Determination of hMSC Viability and Apoptotic Rate

Evaluation of cell viability was performed at days 0, 1, 2, 3, 5, 7, 10, and 14 of culture unless otherwise stated. Each test was conducted in triplicate and repeated on three separate occasions. Cell viability was assessed using the Viacount assay according to the manufacturer’s instructions. Briefly, the Viacount assay is based on incorporation of fluorescent propidium iodide (PI) after loss of cell membrane integrity. The apoptotic rate of the hMSCs was assessed using the annexin V-PE/7 amino-actinomycin (7-AAD) assay, which is based on the fact that, in early apoptosis, annexin V binds to phosphatidylserine residues on the outer part of the cell membrane of the nonviable cells, but that 7-AAD is excluded by viable cells. The cell cycle phase of hMSCs was identified using the Guava cell cycle assay following the manufacturer’s instructions. Briefly, the cell cycle assay uses PI, a nuclear DNA stain, to identify the specific cell-cycle phase. Resting cells (at the G₀/G₁ phase) contain two copies of each chromosome. Cycling cells synthesize chromosomal DNA (S phase), which results in increased fluorescence intensity. When all chromosomal DNA is doubled (G₂/M phase), the cells fluoresce with twice the intensity of the initial population. Data for the determination of cell viability, apoptotic rate, and cell cycle were collected using a Attune cytometer (Life Technologies, Rockville, MD, http://www.lifetech.com) and were analyzed using Attune software.

Determination of hMSC Protein Content

The intracellular protein content was assessed in hMSCs cultured at either 0.1%, 5%, or 21% oxygen at different time points, specifically 3, 7, and 14 days of culture under the conditions of interest to the present study. Briefly, cell content was extracted using 100 μl of RIPA buffer per well of 24-well cell-culture plasticware, and the protein content was determined using the colorimetric DC Protein Assay (Bio-Rad, Hercules, CA, http://www.bio-rad.com). At the same time points, mediators released by the hMSCs in the respective supernatant media were quantified using the DC Protein Assay (Bio-Rad) and following the manufacturer’s instructions. These assessments were conducted in triplicate and repeated on three separate occasions.

Assessment of Released Bioactive Mediators

Release of bioactive mediators from hMSCs exposed to either 0.1%, 5%, or 21% oxygen for up to 14 days was assessed using Luminex technology (Millipore, Billerica, MA, http://www.millipore.com). The levels of 39 mediators, specifically, VEGF-A, VEGF-B, VEGF-C, VEGF-D, IL-1β, IL-1α, IP10, IL-2, IL-7, IL-6, IL-1Ra, IL-15, transforming growth factor α, basic fibroblast growth factor (FGF), hepatocyte growth factor (HGF), IL-4, IL-8, IL-10, IL-12p(70), IL-13, IL-17, angiopoietin 2, endothelin 1, BMP-9 (bone morphogenetic protein 9), leptin, granulocyte colony-stimulating factor, granulocyte-macrophage colony-stimulating factor, tumor necrosis factor α, epidermal growth factor, interferon γ, interferon α, RANTES, MCP-1, macrophage inflammatory proteins 1α and 1b, eotaxin, FGF-1, and FGF-2 were determined in the respective supernatant media at each time point of interest to the present study using a Milliplex MAP kit (Millipore), following the manufacturer’s instructions. These experiments were conducted in triplicate and repeated on three separate occasions.

The presence and concentration of each of the aforementioned mediators were evaluated using the MasterPlex QT 1.0 system (MiraiBio, Alameda, CA, http://www.miraibio.com) and analyzed using Luminex-100 software version 1.7 (Luminex, Austin, TX, http://www.luminexcorp.com). A five-parameter regression formula was used to calculate concentrations of the respective mediators using standard curves for each mediator assessed [22]. The data were analyzed using either a five- or a four-parameter logistic or spline curve-fitting method as recommended by the manufacturer. The type of curve-fitting method for each mediator was chosen with respect to the lowest residual variance (<5%).

Assessment of the Angiogenic Effect of Supernatant CM From Cultured hMSCs

Formation of vascular-like structures were observed 20 hours after seeding 15 × 10³ EPCs per cm² on the top surface of Matrigel hydrogels (Matrigel Growth Factor Reduced, BD Falcon; BD Biosciences, San Diego, CA, http://www.bdbiosciences.com) and covered with 500 μl of supernatant CM. CM were obtained from hMSCs cultured at either 0.1%, 5%, or 21% oxygen for 1, 2, 3, 5, 7, 10, and 14 days. Serum-free αMEM (PAN Biotech) was used as a negative control, and EBM-2 medium (Lonza, Walkersville, MD, http://www.lonza.com) was used as a positive control. The vascular-like structures were then observed using light microscopy (magnification, ×10), photographed, and counted in three randomly chosen fields.

To assess VEGF involvement in the observed formation of vascular-like structures, VEGF-specific antibody (Life Technologies) was added (at the concentration of 1 μg/ml) to the supernatant CM collected from hMSCs cultured for 7 days under each condition of interest to the present study. The vascular-like formation tests were then repeated using these CM and following the procedures described in the previous paragraph. For these experiments, EBM-2 media (Lonza) and αMEM (PAN Biotech) media were used as the positive and negative controls, respectively. These experiments were conducted in triplicate and repeated on three separate occasions.

Chemotactic Cell Migration in Boyden Chambers

The chemoattractive potential of supernatant CM from hMSCs cultured under either 0.1%, 5%, or 21% oxygen levels after 0, 1, 2, 3, 5, 7, 10, and 14 consecutive days was determined using the Boyden chamber migration assay. Briefly, 600 μl of each respective CM were added in the bottom well, and 15 × 10³ EPCs were seeded on the top of the porous (8-μm pore diameter) membrane (previously coated with 0.5% gelatin) of the Transwell. After 6 hours, the EPCs that had migrated were fixed in situ using 11% glutaraldehyde, stained using hematoxylin eosine, and counted following visualization using light microscopy (magnification, ×10). In contrast, the cells that had not migrated from the original seeding location were removed by scrapping the top of the porous Transwell membrane.

To assess the involvement of IL-8, MCP-1, and RANTES in the chemotactic potential of the hMSC secretome, respective specific antibodies (Life Technologies) were separately added to the supernatant conditioned media at the concentration of 1 µg/ml each. Tests of hMSC migration were then conducted using Boyden Chambers and the procedure described earlier in this section.

For these experiments, EBM-2 medium (Lonza) was used as the positive control, and αMEM without serum (PAN Biotech) media was the negative control. These experiments were conducted in triplicate and repeated on three separate occasions.

In Vivo Experiments

Animals

The in vivo paracrine effect of hMSCs was assessed using a mouse (8-week-old male nude mice; Janvier, Saint Berthevin, France, http://www.janvier-labs.com) model. All procedures regarding housing, handling, surgery, and recovery of these animals after surgery were performed in compliance with the guidelines of the new European Directive 2010/63/EU on the protection of animals used for scientific purposes. The project was approved by the local ethics committee for animal experimentation. At the end of the experiments, the animals were sacrificed using overdoses of pentobarbital.

Chemotactic Effect of the Supernatant CM From Cultured hMSCs

Supernatant conditioned media from hMSCs cultured at either 0.1% or 21% oxygen for 7 days were lyophilized on collagen sponges that were implanted subcutaneously in mice for 7 and 14 days. αMEM (PAN Biotech; supplemented with 10% fetal calf serum [FCS]) and αMEM (PAN Biotech; without serum) that had been maintained in the absence of hMSCs but at either 0.1% or 21% oxygen for 7 days was also lyophilized on collagen sponges and implanted subcutaneously in mice for 7 and 14 days; the latter experiments served as positive and negative controls, respectively. Infiltration of collagen sponges by host cells, reflecting the chemotactic effect of the lyophilized bioactive content of the supernatant conditioned media was quantified by counting the cells present in each collagen sponge at the time of excision. At the prescribed times, the excised collagen sponges were fixed in 4% paraformaldehyde, dehydrated, and paraffin-embedded. Sections (each 5 μm thick) were stained using hematoxylin-eosin-safran (HES) and analyzed using light microscopy. The cells that had infiltrated into each collagen sponge were counted in three randomly chosen fields for each condition tested.

In vivo Angiogenic Effect of the Supernatant CM From Cultured hMSCs

To evaluate the angiogenic potential of the hMSC secretome, supernatant conditioned media from hMSCs cultured at either 0.1% or 21% oxygen for 7 days were lyophilized. EBM-2 medium (Lonza; containing 5% FCS) and αMEM (PAN Biotech; without serum) that had been maintained in the absence of hMSCs but at either 0.1% or 21% oxygen for 7 days were also lyophilized and used as the positive and negative controls, respectively. The lyophilized content of 1 ml of each supernatant CM tested in the present study was mixed with 500 μl of Matrigel (Matrigel growth factor reduced; BD Biosciences) and implanted subcutaneously in mice. After 14 days, the Matrigel plugs were excised, and their hemoglobin content was quantified using the Drabkin reagent (LaboModerne, Paris, France, http://www.labomoderne.com) following the manufacturer’s instructions.

Involvement of Specific Mediators in the Chemotactic and Angiogenic Effects of CM

The involvement of VEGF, IL-8, MCP-1, and RANTES in the chemotactic and angiogenic effects of supernatant CM from cultured hMSCs was assessed using specific antibodies (Life Technologies) targeting VEGF, IL-8, MCP-1, and RANTES. All four antibodies (each at the concentration of 1 µg/m) were added simultaneously to the supernatant CM obtained from hMSCs exposed to either 0.1% or 21% oxygen for 7 days.

To assess the angiogenic effect, the lyophilized content from 1 ml of CM (for each condition of interest to the present study) was mixed with 500 μl of Matrigel and implanted subcutaneously in mice for 14 days. The lyophilized content of 1 ml of EBM-2 medium (Lonza; containing 5% FCS) and αMEM (PAN Biotech; without serum) that had been maintained in the absence of hMSCs but at either 0.1% or 21% oxygen for 7 days was also mixed with 500 μl of Matrigel and implanted subcutaneously in mice for 14 days; these experiments served as positive and negative controls, respectively. At the prescribed time, the Matrigel plugs were excised, and their hemoglobin content was quantified using the Drabkin reagent (LaboModerne) following the manufacturer’s instructions.

To assess the chemotactic effect, supernatant conditioned media (from hMSCs cultured for 7 days at either 0.1% or 21% oxygen) supplemented with antibodies (each at the concentration of 1 µg/ml each) to VEGF, IL-8, MCP-1, and RANTES were lyophilized on collagen sponges that were then implanted subcutaneously in mice for 7 and 14 days. At the prescribed times, the collagen sponges were excised, fixed in 4% paraformaldehyde, dehydrated, and paraffin-embedded. Sections (each 5 μm thick) were stained using HES and analyzed using light microscopy. Infiltration of the implanted collagen sponges by host cells, reflecting the chemotactic effect of the hMSC supernatant CM, was quantified using light microscopy (magnification, ×10) and by counting cells in three randomly chosen fields for each condition tested. Inflammatory response was assessed by quantifying reactive oxygen species (ROS) using the L012 chemiluminescence method. L012 is derived from luminol that has a high sensitivity for detecting superoxide radicals. Luminol does not show redox cycling itself.

Statistical Analyses

Numerical data were expressed as the means ± SD. The in vitro data were analyzed statistically using analysis of variance and the Bonferoni post-test. The in vivo data were analyzed using the Mann-Whitney tests and Prism software (GraphPad Software, Inc., San Diego, CA, http://www.graphpad.com). p values <.05 were considered significant.

Results

Low Impact of Oxygen Tension on the Amount of Proteins Secreted by hMSCs

When hMSCs were cultured in serum-free cell-culture medium at various oxygen levels (specifically, either 0.1%, 5%, or 21%) for up to 14 consecutive days, their survival rate (determined as the number of viable hMSCs) and morphology were similar under the 3 pO₂ conditions tested (supplemental online Fig. 1A, 1B). In addition, similar amounts of intracellular proteins were produced by hMSCs under the 3 oxygen levels tested during the 14 days of culture (supplemental online Fig. 1C). The total amount of proteins released by hMSCs was similar during the first 7 days of culture at all oxygen levels tested but was significantly (p < .05) decreased at 14 days when the hMSCs were cultured at 0.1% oxygen: specifically, this result was a 79% decrease in comparison with the proteins released by the hMSCs cultured at either 5% or 21% oxygen (supplemental online Fig. 1D). hMSCs cultured at either 5% or 21% oxygen for 14 days produced similar amounts of released protein (supplemental online Fig. 1C, 1D).

Near Anoxia Promotes the Chemotactic Potential of the hMSC Secretome

Because paracrine effects, particularly chemotactic and proangiogenic ones, are crucial aspects in the success of MSC-based regenerative medicine, the chemotactic potential of supernatant CM from hMSCs cultured at either 0.1%, 5%, or 21% oxygen for various periods of time was assessed in vitro using the Boyden chamber migration test (Fig. 1A). Supernatant CM from hMSCs cultured in near anoxic (0.1% O₂) conditions for up to 14 consecutive days promoted significantly (p < .05) increased cell migration (fold induction of 1.6, 2.4, and 2.25 with CM collected after 3, 7, and 14 days, respectively) compared with the respective supernatant CM from hMSCs cultured under either the normoxic (21% oxygen) or at 5% oxygen environments (Fig. 1A, 1B). Cell migrations induced by the contents of the supernatant CM from hMSCs cultured either under normoxia or under 5% oxygen were similar (Fig. 1B).

Figure 1.

Near anoxia promotes the chemotactic potential of the human mesenchymal stem cell (hMSC) secretome. (A, B): Representative light micrographs (A) and quantitation of endothelial progenitor cell (B) migration in Boyden chambers in the presence of supernatant CM from hMSCs cultured at either 21%, 5%, or 0.1% oxygen during a 14-day time period. Magnification, ×10, 0.4 inch = 50 μm (n = 9). ∗, p < .05. (C): Histological analysis of host mouse cells that had infiltrated collagen sponges containing lyophilized CM obtained from hMSCs cultured at either 21% (21% O₂ CM) or 0.1% (0.1% O₂ CM) oxygen for 7 days. Magnification, ×10. (D): Quantitation of cells that had infiltrated collagen sponges containing lyophilized CM (obtained from hMSCs cultured at either 21% (21% O₂ CM) or 0.1% (0.1% O₂ CM) oxygen for 7 days) at 7 and 14 days postimplantation in nude mice (n = 6). ∗, p < .05. Abbreviation: CM, conditioned media.

The bioactivity of the released chemotactic chemical compounds was also assessed in vivo. For this purpose, the contents of supernatant CM from hMSCs cultured under either 0.1% or 21% oxygen for 7 days were lyophilized on collagen sponges that were subcutaneously implanted in nude mice. Explants at 7 days postimplantation revealed a significantly (p < .05) larger number of cells that had infiltrated the collagen sponges that had been pretreated with the lyophilized content of supernatant CM obtained from hMSCs cultured at 0.1% oxygen for 7 days (Fig. 1C, 1D); in comparison, this effect was more pronounced (specifically, 29% higher) 14 days postimplantation (Fig. 1D). Histological and immunohistochemical analyses of explanted collagen sponges revealed that they were infiltrated mainly with fibroblastic cells. At time, neo-vessels with CD-31 positive cells and macrophages (ITGAM-positive cells; supplemental online Fig. 2) were observed. No lymphocytes, eosinophils, neutrophils, or basophils were detected using May-Grünwald Giemsa coloration. In contrast to what was observed with supernatant CM from hMSCs cultured under 1% oxygen, there was no cell infiltration of the collagen sponges pretreated with the content of supernatant CM from hMSCs cultured under 21% oxygen for 7 days (Fig. 1D).

Near Anoxia Promotes the Angiogenic Potential of the hMSC Secretome

The proangiogenic bioactivity of the hMSC secretome was assessed in vitro using the Matrigel test. CM obtained from hMSCs cultured under either near anoxia (0.1% oxygen), 5% oxygen, or normoxia (21% oxygen) for 3, 7, and 14 consecutive days were used to culture human EPCs. Formation of vascular-like structures was visualized using light microscopy. The supernatant CM produced when the hMSCs were exposed to near anoxic (0.1% oxygen) conditions resulted in significantly (p < .05) more vascular-like structures (Fig. 2A, 2B) than the CM obtained from hMSCs cultured at either 21% or 5% oxygen for 3, 7, and 14 consecutive days; specifically, these results were 2.4, 3.2, and 2.4 times higher at 3, 7, and 14 days, respectively. In contrast, the number of vascular-like structures induced by the supernatant CM from hMSCs cultured at either 21% or 5% oxygen for up to 14 days were similar (Fig. 2A, 2B).

Figure 2.

Near anoxia promotes the angiogenic potential of the hMSC secretome and decreases inflammatory mediator release. (A, B): Representative light micrographs illustrating the time course of formation (A) and quantitation of vascular-like structures formed by endothelial progenitor cells (B) on Matrigel matrices in the presence of supernatant CM obtained from human mesenchymal stem cells (hMSCs) cultured at either 21%, 5%, or 0.1% oxygen for up to 14 days. Magnification, ×10 (n = 9). ∗, p < .05. (C): Hemoglobin concentration contained in excised Matrigel plugs 14 days postimplantation in mice. These plugs were pretreated before implantation with the lyophilized content of supernatant CM obtained from hMSCs cultured at either 21% or 0.1% oxygen for 7 days (n = 6). ∗, p < .05. (D): Noninvasive, bioluminescence in vivo imaging of collagen sponges containing the lyophilized content of supernatant CM from hMSCs cultured either at 0.1% oxygen (on the left side of the nude mice) or 21% (on the right side of the nude mice) for 7 days following injection of L012 (a marker of inflammation) 2 days after subcutaneous implantation. (E): Quantification of the bioluminescent signal detected as described in (D). Abbreviations: CM, conditioned media; Max, maximum; Min, minimum.

The proangiogenic effects of supernatant CM were also assessed in vivo. To this aim, the lyophilized contents of supernatant CM from hMSCs cultured under either 0.1% or 21% oxygen for 7 days were added to Matrigel plugs (each 400 μl) that were subcutaneously implanted in nude mice. After 14 days, the vascular network in the excised Matrigel plugs containing lyophilized CM from hMSCs cultured under near anoxic (0.1% oxygen) conditions was higher than that observed in Matrigel plugs containing lyophilized CM from hMSCs cultured in normoxia (21% oxygen), as confirmed by the measurement of hemoglobin concentration in the respective Matrigel plugs (Fig. 2C). Specifically, the hemoglobin concentration was 5.6 times higher in Matrigel plugs containing the contents of supernatant CM from hMSCs cultured under near anoxia for 7 days than in Matrigel plugs containing the contents of supernatant CM from hMSCs cultured under normoxia for 7 days (Fig. 2C).

Near Anoxia Decreases Secretion of Inflammatory/Immunomodulatory Mediators by hMSCs

To assess the bioactivity of inflammatory/immunomodulatory mediators released by hMSCs in vivo, the contents of supernatant CM from hMSCs cultured under either 0.1% or 21% oxygen for 7 days were lyophilized on collagen sponges that were subcutaneously implanted in nude mice. The inflammatory response was followed noninvasively by bioluminescence using the L012 compound, a probe targeting ROS-mediated inflammation.

Two days postimplantation, the inflammatory response was significantly (p < .05) higher for the implanted collagen sponges containing the lyophilized content of supernatant CM from hMSCs cultured at 21% oxygen for 7 days in comparison with collagen sponges containing the lyophilized content of supernatant CM from hMSCs cultured at 0.1% oxygen for 7 days (Fig. 2D, 2E). This effect decreased but remained significant (p < .05), until 5 days postimplantation; no inflammatory response was detected thereafter (Fig. 2E).

Near Anoxia Induces a Specific Secretome Profile

The secretome profile of hMSCs cultured under various oxygen tensions was determined by assessing the release of bioactive mediators known either for their pro- and anti-inflammatory, immunomodulative, chemotactic, or angiogenic properties, as well as of select growth factors. Release of only proangiogenic and chemotactic mediators was observed when the hMSCs were cultured under 0.1% oxygen; this result was observed as soon as 2 days and up to 14 days of culture (Fig. 3). Expression of the pro- and anti-inflammatory and immunomodulatory mediators, as well as of the growth factors by hMSCs exposed to near anoxia (0.1% oxygen), was similar to that observed in control medium (αMEM without serum) during the 14 days of the experiment and monitored in the present study. In contrast, a large number of mediators (specifically, IL-8, MCP-1, RANTES, IL-1b, IL-6, IL-1Ra, IL-15, VEGFA, basic FGF, and HGF) were significantly (p < .05 in comparison with results obtained in control medium) enhanced when these cells were maintained under either normoxic (21% oxygen) or at 5% oxygen conditions (Table 1; Figs. 3, 4).

Figure 3.

Near anoxia promotes the release of bioactive proangiogenic and chemotactic mediators from human mesenchymal stem cells (hMSCs). Shown are time courses of bioactive mediators released in the supernatants of hMSCs cultured at either 21%, 5%, or 0.1% oxygen for up to 14 consecutive days. (A–E): Released amounts of VEGF-A (A), VEGF-C (B), MCP-1 (C), IL-8 (D), and RANTES (E), respectively (n = 3). ∗, p < .05. Abbreviations: IL, interleukin; MCP-1, monocyte chemoattractant protein 1; VEGF, vascular endothelial growth factor.

Table 1.

Exposure of hMSCs to near anoxia promotes a specific secretome profile

Figure 4.

Near anoxia decreases the release of bioactive inflammatory mediators from human mesenchymal stem cells (hMSCs). (A–D): Time course of IL-1b (A), IL-6 (B), IL-15 (C), and IL-1Ra (D) released in the supernatants of hMSCs cultured at either 21%, 5%, or 0.1% oxygen for up to 14 consecutive days (n = 3). ∗, p < .05. Abbreviation: IL, interleukin.

Identification of the Released Bioactive Mediators

Analysis using Luminex technology provided evidence that the concentration of the proangiogenic growth factors VEGF-A and VEGF-C were significantly (p < .05) higher when the hMSCs were exposed to the 0.1% oxygen environment (Fig. 3A, 3B) as soon as 2 days and up to 14 days of culture. The amounts of these released mediators were significantly (p < .05) higher than those released from cells exposed to either normoxia (21%) or 5% oxygen (specifically, 8.8 ng/ml vs. 0.78 ng/ml and 0.75 ng/ml for VEGF-A and 78.2 pg/ml vs. 2.5 pg/ml and 1.2 pg/ml for VEGF-C, respectively).

Among the chemotactic mediators, MCP-1, IL-8, and RANTES were continuously released by the hMSCs cultured in the 0.1% oxygen environment over a period of 14 days (Fig. 3C–3E). The amount of these secreted mediators was significantly (p < .05) higher when the hMSCs were exposed to 0.1% oxygen compared with results obtained from hMSCs cultured under either normoxia (21%) or at 5% oxygen for the same time period (Fig. 3C–3E).

The immunomodulatory IL-1b, IL-6, IL-1Ra, IL-15, and basic FGF and HGF growth factors were significantly (p < .05) released only when hMSCs were exposed to either 5% or 21% oxygen for up to 14 days (Table 1; Fig. 4A–4D). The concentrations of the bioactive mediators released by the hMSCs exposed to either normoxia (21%) or 5% oxygen and monitored in the present study were similar (Table 1; Fig. 4A–4D).

Identification of the Mediators Responsible for the hMSC Secretome Functionality

The bioactivity of the mediators secreted by the hMSCs exposed to 0.1% oxygen was assessed by repeating the Matrigel tests and cell migration in Boyden chamber experiments (described in Materials and Methods), using appropriate antibodies to block the bioactivity of specific mediators. Blocking the VEGF activity in the supernatant CM from hMSCs cultured under 0.1% oxygen significantly (p < .05) decreased formation of vascular-like structures; in fact, there was a 50% inhibition of the observed vascular-like structure formation (Fig. 5A). In addition, blocking the VEGF activity in the supernatant CM from hMSCs cultured under 21% oxygen decreased vascular-like formation by 35% (Fig. 5A).

Figure 5.

Identification of the mediators responsible for the human mesenchymal stem cell (hMSC) secretome functionality. (A): In vitro quantitation of the formation of vascular-like structures formed by endothelial progenitor cells (EPCs) cultured on Matrigel plug in the presence of supernatant CM obtained from hMSCs cultured at either 21% or 0.1% oxygen with and without VEGF antibody (anti-VEGF) for 7 consecutive days. (B): Quantitation of EPC migration in Boyden chambers in the presence of supernatant CM obtained from hMSCs cultured at either 21% or 0.1% oxygen for 7 days and either supplemented or not with the respective anti-IL-8, anti-MCP-1, anti-RANTES, or the combination of these 3 antibodies (each at the concentration of 1 μg/ml). (C): Histological analysis of host cell colonization in collagen sponges containing lyophilized CM obtained from hMSCs cultured at either 21% or 0.1% oxygen for 7 days and either supplemented or not with anti-VEGF, anti-IL-8, anti-MCP-1, and anti-RANTES. (D): Hemoglobin concentration in Matrigel plugs containing the lyophilized content of 1 ml of supernatant CM obtained from hMSCs cultured at either 21% or 0.1% oxygen for 7 days and either supplemented or not with anti-VEGF, anti-IL-8, anti-MCP-1, and anti-RANTES 14 days postimplantation in nude mice (n = 6). ∗, p < .05. Abbreviations: 3AB, anti-IL-8 + anti-MCP-1 + and anti-RANTES; 4AB, anti-VEGF + anti-IL-8 + anti-MCP-1 + and anti-RANTES; AB, antibodies; CM, conditioned media; MCP, monocyte chemoattractant protein; VEGF, vascular endothelial growth factor.

When IL-8, MCP-1, and RANTES were selectively blocked (Fig. 5B) using specific antibodies, inhibition of EPC migration in the Boyden chamber was significantly (p < .05) inhibited: the respective inhibition was 65%, 47%, and 50% for these 3 chemical compounds, respectively. Blocking all 3 of these compounds simultaneously induced a 67% inhibition of cell migration (Fig. 5B). Inhibition of cell migration by blocking separately each of these 3 mediators in the supernatant CM from hMSCs exposed to normoxia (21% O₂) for 7 days was also significant (p < .05), specifically, inhibition of 55%, 16%, and 25% for IL-8, MCP-1, and RANTES, respectively (Fig. 5B).

The effect of blocking the bioactivity of VEGF, IL-8, MCP-1, and RANTES in in vivo tests was also assessed. In these cases, Matrigel plugs and collagen sponges pretreated with the lyophilized content from the supernatant CM of hMSCs cultured either under near anoxia (0.1% O₂) or under normoxia (21% O₂) for 7 days were used. Select bioactive mediators were blocked in some experiments using specific antibodies before lyophilization. Matrigel plug vascularization and collagen sponge infiltration by host cells were determined 14 days postimplantation in nude mice. Analysis of collagen sponge infiltration revealed that specific inhibition of VEGF, IL-8, MCP-1, and RANTES significantly (p < .05) inhibited host cell infiltration (Fig. 5C); this inhibition was 72% and 47% in collagen sponges pretreated with the lyophilized content from the supernatant CM of hMSCs cultured either under near anoxia (0.1% O₂) or under normoxia (21% O₂) for 7 days, respectively. Moreover, the hemoglobin concentration in Matrigel plugs containing lyophilized contents from the supernatant CM of hMSCs cultured either under near anoxia (0.1% O₂) or under normoxia (21% O₂), when VEGF, IL-8, MCP-1, and RANTES bioactivity were selectively blocked, was significantly (p < .05) decreased by 80% and 40%, respectively (Fig. 5D).

Discussion

Considering the increasing attention paid to bioactive mediators pertinent to new tissue formation that are secreted by hMSCs and the variability of oxygen tensions in body tissues and organs, knowledge of the paracrine properties of stem cells under various oxygen levels is essential for planning appropriate clinical strategies in regenerative-medicine and tissue-engineering applications. The major gap in current scientific knowledge in this field is that previous studies on the hMSC secretome addressed the effects of specific oxygen tensions on a small number of growth factors; consequently, the role of oxygen tension on the extensive repertoire of bioactive mediators that affect various functions of hMSCs is, at best, incomplete.

The present study provides the first evidence of a shift of the hMSCs cytokine signature induced by oxygen tension, particularly near anoxia (0.1% O₂). Indeed, when exposed to 0.1% oxygen, chemotactic and angiogenic mediators are increasingly produced (p < .05) by hMSCs, whereas inflammatory/immunomodulatory cytokine secretion is significantly (p < .05) decreased (Fig. 6A). It should be noted that only angiogenic and chemotactic mediators were released from hMSCs under 0.1% O₂, whereas immunomodulative mediators and growth factors were significantly (p < .05) expressed at 5% and 21% oxygen (Table 1). This selectivity in mediators produced when the hMSCs are cultured in a near anoxic environment challenges some of the functions currently attributed to hMSCs [23, 24], specifically, the immunomodulative properties of mediators expressed under normoxic (21% O₂) conditions [25, 26].

Figure 6.

Comparison and classification of the bioactive mediators released in the supernatant by human mesenchymal stem cells (hMSCs) as a function of oxygen tension. (A): Concentrations of the bioactive mediators expressed by hMSCs cultured under the various oxygen tension conditions (either 0.1%, 5%, or 21%) tested after 14 days of culture (n = 6). (B): Relative distribution of the various bioactive mediators tested as described for (A) but grouped according to their known bioactivity. Abbreviations: FGFbasic, basic fibroblast growth factor; IL, interleukin; MCP, monocyte chemoattractant protein; VEGF, vascular endothelial growth factor.

MCP-1, IL-8, VEGF, and RANTES Are Crucial Paracrine Mediators Released Under Near Anoxia

Among the mediators assessed in the present study, MCP-1, IL-8, VEGF-A, VEGF-C and RANTES proved to be the crucial regulators of the hMSCs paracrine bioactivity potential in near anoxic conditions. Stem cells produce a broad spectrum of cytokines, chemokines, growth factors, antioxidants, microRNA, ECM molecules, and chaperone proteins [27, 28]. In the present study, which focused on select cytokines, the bioactivity of IL-8, MCP-1, RANTES, and VEGF in vitro and in vivo was confirmed by using specific, pertinent antibodies to block these four mediators (Fig. 5); the result was a significant (p < .05) subsequent inhibition of chemotaxis and angiogenesis both in vitro and in vivo. This is the first time such results are reported for the effect of the near anoxia milieu on hMSC function pertinent to wound healing and new tissue formation. VEGF, IL-8, and MCP-1, which are known for their chemotactic and angiogenic effect on endothelial cells [10, 26, 29, 30], are released in significant amounts by hMSCs under normoxic (21% O₂) conditions (Fig. 6A). The evidence that these mediators are released in increased (10-fold) concentrations (Fig. 6A, 6B) when the hMSCs are exposed to near anoxia and that these concentrations represent more than 99% of the total released (Fig. 6B) bioactive agents monitored in the present study emphasizes the key role of VEGF, IL-8, MCP-1, and RANTES in regenerative-medicine and tissue-engineered applications.

Surprisingly, basic FGF, which is known to act synergistically with VEGF to promote angiogenesis [31], was not detected in the present study despite the angiogenic effects observed in functional tests in vitro and in vivo. Nevertheless, basic FGF takes part in a late step of angiogenesis, and the present study suggests that this mediator does not play a pivotal role in the angiogenic effects observed in our study.

Near Anoxia Induces Decreased Release of Inflammatory Mediators by hMSCs

The present study also provided evidence for the significant (p < .05) decrease of inflammatory/immunomodulatory mediators secreted by hMSCs in near anoxic conditions. The immunomodulatory functions of hMSCs have great potential for therapeutic applications [24], because of their effect on the immune system in the context of inflammatory-mediated disorders [32, 33]. In some cases, however, the reported data that MSCs enhance immune cell survival and function conflict other published reports that MSCs inhibit inflammation and encourage wound healing [34, 35]. The present study provides new information that elucidates aspects of this controversy, specifically that oxygen tension is an important regulator of hMSC paracrine functions. In fact, oxygen tension differently modulated the inflammatory and angiogenic effects of hMSCs, specifically decreased release of the inflammatory/immunomodulatory mediators by hMSCs occurred only under near anoxia (0.1% O₂). Consequently, an important outcome of the present study concerns the definition of “hypoxia.” Literature reports used the range between 1% and 5% oxygen as the definition of an “hypoxic environment” [13, 36]; in the present study, hMSCs cultured under either 5% or 21% oxygen contained similar amounts of intracellular proteins (supplemental online Fig. 1) and released similar concentrations of bioactive mediators in the supernatant media (Figs. 3, 4), which exhibited comparable chemotactic and angiogenic potentials (Figs. 1, 2).

Perspectives for the Use of CM in Regenerative Medicine

The present study is the first to assess the impact of a range of oxygen tensions on a large panel of bioactive mediators released by hMSCs that have crucial implications in tissue regeneration and, thus, for tissue engineering applications. The challenge is now to adapt the insights gained from these in vitro cell culture and in vivo animal studies to the field of regenerative medicine.

A perspective opened by the results of the present study concerns the use, and efficacy, of CM in therapeutic modalities. The results provide encouraging insight into how to improve in vitro and particularly in vivo (Figs. 1B–1D, 2C) angiogenesis and cell chemotaxis by using CM obtained by exposing hMSCs to near anoxic conditions. However, with respect to potential future applications, use of supernatant CM from hMSCs cultured under near anoxia for tissue repair applications necessitates further investigations in clinically pertinent animal models. Nevertheless, recent studies support this approach: in fact, Ando et al. [11] reported that CM from MSCs cultured for 72 hours accelerated osteogenesis in a murine model, and Chen et al. [37] and Jun et al. [38] showed that CM from hMSCs kept under 1% oxygen for 2 days enhanced skin-wound healing in mice. The shift in the released-cytokine signature as a function of oxygen tension observed in the present study could help endeavors to optimize CM preparation for several potential therapeutic applications. In addition, the present study provided evidence that the angiogenic and chemotactic potentials of CM from hMSCs were a function of time. Indeed, the angiogenic and chemotactic potential of this CM reached a maximum after 7 days and remained at this maximal level for up to 14 days of hMSC culture (Figs. 1B, 2B).

Another aspect evidenced by the present study concerns the use of MSCs in engineered tissue constructs. The assessed in vitro model of near anoxia reflects conditions encountered by hMSCs implanted in the core of large-volume tissue constructs, such as the ones used in the treatment of clinically relevant bone defects, and contributes valuable information regarding the expression and paracrine bioactivity of the specific secretome profile of hMSCs in an oxygen environment simulating that of the in vivo situation [39–41].

It has been claimed that during the first postimplantation days of tissue-engineered constructs, and until these implants and the surrounding newly formed tissues become vascularized, the most crucial aspect for MSC survival and function in this microenvironment is oxygen tension [42]; moreover, massive death of cells was reported under these conditions [43]. In this respect, the results of the present study could be used in planning and establishing novel and appropriate strategies to overcome the negative impact of low oxygen content in the outcome of wound healing and new tissue formation postimplantation of, for example, tissue-engineered constructs. However, extrapolations of the findings of the present study to in vivo events that involve hMSC exposure not only to near anoxia but also to other crucial bioactive signals (including inflammatory cytokines) undoubtedly require further investigation.

Conclusion

The present study demonstrated for the first time that the hMSC secretome profile under near anoxia (0.1% O₂) conditions specifically affects select chemotactic and angiogenic mediators, such as VEGF-A, VEGF-C, IL-8, MCP-1, and RANTES, which are of crucial importance in tissue regeneration and tissue engineering applications. Moreover, hMSCs exposed to the near-anoxia milieu tested do not release inflammatory cytokines. These findings emphasize the importance of oxygen tension, which must be taken into account when evaluating the hMSC paracrine potential and the use of conditioned media in therapeutic modalities.

Author Contributions

J.P.: conception and design, collection and assembly of data, data analysis and interpretation, manuscript writing; M.D: conception and design, data analysis and interpretation; A.M.: provision of study material, data analysis and interpretation; D.L.-A.: data analysis and interpretation; C.B.-V.: provision of study material; H.P.: conception and design, data analysis and interpretation, financial support, manuscript writing, final approval of manuscript.

Disclosure of Potential Conflicts of Interest

C.B.-V. is a compensated former employee of the CNRS, National Center of Scientific Research. H.P. has compensated research funding from Biobank. The other authors indicated no potential conflicts of interest.