Department of Medical Biochemistry and Molecular Biology, Biocenter Oulu and Oulu Center for Cell-Matrix Research, University of Oulu, Oulu, Finland
Correspondence: Peppi Koivunen, M.D., Ph.D., Department of Medical Biochemistry and Molecular Biology, Aapistie 7, P.O. Box 5000, University of Oulu, Oulu FIN-90014, Finland. Telephone: +358-8-5375822; Fax: +358-8-5375811; e-mail email@example.com
Author contributions: S.P.: conception and design, collection and/or assembly of data, data analysis and interpretation, manuscript writing, and final approval of manuscript; M.P.: conception and design, collection and/or assembly of data, data analysis and interpretation, and final approval of manuscript; P.L. and P.K.: conception and design, financial support, data analysis and interpretation, manuscript writing, final approval of manuscript, and contributed equally; S.L. and J.P.: provision of material or patients and final approval of manuscript; R.S.: collection and/or assembly of data and final approval of manuscript.
Human mesenchymal stem cells (hMSCs) are multipotent cells that have aroused great expectations in regenerative medicine. They are assumed to originate from hypoxic stem cell niches, especially in the bone marrow. This suggests that O2 is of importance in their regulation. In order to characterize regulation of the oxygen sensing pathway in these cells, we studied hMSCs isolated from three origins, adult and pediatric bone marrow and umbilical cord blood (UCB). Surprisingly, pediatric bone marrow and UCB MSCs showed normoxic stabilization of hypoxia-inducible factor-1α (HIF-1α) that is normally degraded completely by HIF prolyl 4-hydroxylases in the presence of oxygen. This was due to a high expression level of HIF-1α mRNA rather than inappropriate post-translational degradation of HIF-1α protein. HIF-1α mRNA was also induced in normoxic adult bone marrow MSCs, but 40% less than in the pediatric cells, and this was apparently not enough to stabilize the protein. The high normoxic HIF expression in all the hMSCs studied was accompanied by increased expression of a large number of glycolytic HIF target genes and increased glycolysis. Osteogenic differentiation of bone marrow-derived hMSCs reduced HIF-1α mRNA and protein expression and the expression of glycolytic mRNAs, resulting in decreased glycolysis and induction of oxidative metabolism. Induced mitochondrial biogenesis, changes in mitochondrial morphology and size indicative of increased oxidative phosphorylation, and induction of extracellular matrix synthesis were observed following osteogenic differentiation. Altogether, these data suggest that HIF-1α is a general regulator controlling the metabolic fate and multipotency of the hMSCs. Stem Cells2013;31:1902-1909
Human mesenchymal stem cells (hMSCs) are multipotent cells capable of osteogenic, adipogenic, and chondrogenic differentiation that have aroused great expectations in the field of regenerative cellular therapies [1-4]. They can be obtained from almost any tissue, but traditionally they are extracted from bone marrow and adipose tissue [1, 5]. hMSCs from adult sources have notable limitation in the clinical use, since they reach their division capacity relatively fast [6-8] and their in vivo survival is limited . The reason for their poor survival is not fully understood, although metabolic conditions have been shown to play a role in it . Umbilical cord blood-derived MSCs (UCBMSCs) have been suggested at least as a means of partly avoiding these limitations . Although all hMSCs possess the same cell surface antigen expression characteristics, cells derived from various tissues and individuals of different ages can show major differences in their morphology and functional properties .
The metabolic state has recently been shown to play a role in the maintenance of pluripotent cells in their undifferentiated state and in the differentiation process . Undifferentiated cells have been reported to rely for their energy production on aerobic glycolysis yielding lactate [14, 15], a pathway that has been documented as fuelling the anabolic and catabolic requirements for maintain stem cells in a pluripotent state. Mitochondrial biogenesis and maturation of the mitochondrial oxidative metabolism appear to match the energy demands of differentiation [14, 15]. It has also been suggested recently that glucose substitution may enhance hMSC survival after transplantation . All these data indicate that enhanced glycolytic metabolism could play a crucial role in the functional regulation of MSCs.
hMSCs reside in hypoxic perivascular niches in the body, but little is known about the regulation of their oxygen sensing pathway at the molecular level . Hypoxia-inducible factor (HIF), the major regulator of hypoxia responses in all cells [18, 19], is a heterodimer consisting of an oxygen-regulated α-subunit (HIF-α) and a constitutively expressed β-subunit (HIF-β). Three distinct HIF-α subunits, HIF-1α, 2α, and 3α, exist in humans. The stability of the HIF-α subunits is regulated post-translationally by three HIF prolyl 4-hydroxylases, PHDs 1–3 (also known as HIF-P4H-1–3 or EglN2, 1, and 3, respectively). In the presence of oxygen, these PHDs hydroxylate-specific prolines in the HIF-α subunit, resulting in binding of the von Hippel Lindau protein, subsequent polyubiquitinylation, and proteosomal degradation, but under hypoxic conditions HIF-α escapes degradation and translocates to the nucleus, where it binds HIF-β and initiates the transcription of several hundred HIF target genes, such as erythropoietin, vascular endothelial growth factor, and a number of glycolytic genes such as glucose transporter 1 [18, 19]. Hypoxic conditions have also been reported to promote an undifferentiated state in hMSCs .
We focus here on analyzing at the molecular level the main components of the hypoxia response pathway in hMSCs extracted from UCB and the bone marrow of two differently aged groups of donors by comparison with a commercially available embryonic kidney cell line HEK293. We determined the expression of HIF-1α and PHDs in these cells at the mRNA and protein levels under normoxia and hypoxia. The expression of a set of glycolytic HIF target genes in the cells was determined under normoxia and hypoxia, along with glucose consumption and lactate production. The same parameters were also determined in the bone marrow-derived hMSCs after in vitro osteoinduction. Our data show that HIF-1α in hMSCs is stabilized in normoxia, and that it is this that drives aerobic glycolysis and the expression of glycolytic genes. Osteogenic differentiation reversed the normoxic HIF-1α stabilization and glycolytic metabolism and activated the mitochondria. We therefore argue that high normoxic HIF-1α plays a role in maintaining hMSCs in an undifferentiated state by supporting induced glycolysis.
Materials and Methods
Cell Culture and Osteoinduction
Bone marrow mesenchymal stem cells (BMMSCs) were harvested from patients operated on for osteoarthrosis (adult BMMSCs, aBMMSCs) and from pediatric patients operated for orthopaedic trauma or idiopathic scoliosis (paediatric BMMSC, pBMMSCs). All the samples were collected from unaffected bone sites. These BMMSCs samples were suspended in proliferation medium (supporting information Table S1) and seeded in a cell culture flask. Isolation of mononuclear cells from the idiopathic scoliosis patients was done by Ficoll-Hypaque (Amersham Biosciences, Piscataway, NJ, http://www.gelifesciences.com) gradient centrifugation before suspending the cells in proliferation medium. The primary cells were allowed to attach to the culture flask in the proliferation medium for two days at 37oC in 5% CO2 and 20% O2 before the medium was changed and non-attached cells were washed out.
The umbilical cord blood samples were collected at Helsinki University Central Hospital (HUCH), Department of Obstetrics and Gynaecology, and Helsinki Maternity Hospital. Mononuclear cells were isolated using Ficoll-Hypaque (Amersham Biosciences) gradient centrifugation. Tissue culture plates were coated with fibronectin (Sigma Aldrich, St Louis, MO, http://www.sigmaaldrich.com) and mononuclear cells were plated at a density of 1 x 106/cm2. The initial primary UCBMSC line was established under hypoxic conditions (5% CO2, 3% O2 at 37oC), and the cells were allowed to adhere overnight. Non-adherent cells were washed out when the medium was changed. Hence the UCBMSCs were cultured in the culture flasks without fibronectin under normoxic conditions (5% CO2 and 20% O2 at 37oC). The established primary lines were passaged when almost confluent and replated at 1000–3000 cells/cm2.
Human embryonic kidney cells (HEK293) were obtained from the American Type Culture Collection and cultured at a density of 1 × 105 cells/cm2 under normoxic conditions (5% CO2 and 20% O2 at 37oC).
Osteoinduction was performed at 37oC in 5% CO2 and 95% air. The cells were plated at a density of 1500 cells/cm2 for a differentiation period of 14 or 18 days.
The medium was changed twice a week and the cells were cultured to 70–80% confluence. Three independent donors in each hMSC group were included in this study and all the cells were used before passage 6. The compositions of the proliferation and osteoinduction media are presented in supporting information Table S1.
Flow Cytometric Characterization of hMSCs
The minimal criterion panel of surface antigens proposed by the Mesenchymal and Tissue Stem Cell Committee of the International Society for Cellular Therapy for characterizing hMSCs was applied to the primary cell lines . The hMSCs were detached from the culture flasks and suspended in phosphate buffer saline – 0.5 % bovine serum albumin (PBS-0.5% BSA) (100,000 cell/mL), and the following antibodies were incubated for 20 min at room temperature: CD49e (phycoerythrin (PE), BD Biosciences, San Jose, CA, http://www.bdbiosciences.com), CD90 (FITC, Stem Cell Technologies, Grenoble, France, http://www.stemcell.com), CD73 (PE, BD Biosciences), Human Leukocyte Antigen -ABC (HLA-ABC) (Allophycocyanin (APC), BD Biosciences) and CD105 (FITC, Abcam). A negative panel of the following surface antigens was incubated simultaneously as a group in the same sample: HLA-DR (PE, BD Biosciences), CD34 (PE, BD Biosciences), CD45 (PE, BD Biosciences), CD14 (PE, BD Biosciences) and CD19 (PE, BD Biosciences). In addition, the following isotype controls were used: FITC Mouse IgG2a k (BD Biosciences) and PE Mouse IgG2a k (BD Biosciences). The cells were washed with PBS-0.5% BSA after incubation and analysed by FACSAria, equipped with laser emission at 488, 633 and 407 nm. The FITC, PE, and APC channels were used to detect the emission of conjugated surface antigens. The data were analysed using Cyflogic (CyFlo Ltd, Turku, Finland, http://www.cyflogic.com). Cell debris was gated out from all samples, the definition of a positive gate being based on isotype controls and unlabelled cells.
Hypoxic Exposure and Western Blot Analysis
Cells were exposed to hypoxia by culturing them in an InVivo2 400 hypoxia workstation (Ruskinn, Bridgend, UK, http://www.ruskinn.com) under 1% O2 balanced with N2. After 24 hours of exposure to hypoxia, the cell cultures were divided in half and one half was used for RNA isolation and the other for Western blot analysis. Cell extracts for the latter were obtained by lysing the cells in 150 mM NaCl, 0.1% SDS, 20 mM Tris-HCl, pH 6.8, and 0.25% Nonidet P-40. Qiagen TissueLyser LT (Qiagen, Hilden, Germany, http://www.qiagen.com) and 5 mm stainless steel beads (Qiagen) were used to homogenize the osteoinduced pBMMSCs. The supernatant protein concentrations were determined by the Bradford method (Bio-Rad protein assay, Bio-Rad, Hercules, CA, http://www.bio-rad.com), and the cell extracts were analysed by 10% SDS-PAGE, blocked with Tris-buffered saline with 5% non-fat dry milk and probed with the following primary antibodies: anti-HIF-1α (610959, BD Biosciences), anti-PHD1 (NB100-310, Novus Biologicals, Littleton, CO, http://www.novusbio.com), anti-PHD2 (NB100-2219, Novus Biologicals) and anti-α-tubulin (T6199, Sigma Aldrich). The samples were blotted onto Immobilon-P membranes (Millipore, Billerica, MA, http://www.millipore.com) following ECL™ immunodetection.
Quantitative Real Time RT-PCR (Q-PCR)
Total RNA was isolated and purified with an EZNA total RNA kit (OMEGA bio-tek, Norcross, GA, http://www.omegabiotek.com), and reverse transcription was performed with an iScript cDNA synthesis kit (Bio-Rad). Tata binding protein (TBP) or 18S ribosomal RNA (18S) was used to equalize the amounts of template in Q-PCR, which was performed with iTaq SYBR Green Supermix and ROX (Bio-Rad) in a Stratagene MX3005 thermocycler (Agilent Technologies, Santa Clara, CA, http://www.agilent.com). The sequences for the primers used in Q-PCR are listed in supporting information Table S2.
Glucose and Lactate Assays
Proliferation medium samples were collected on the day the cells were plated (day 0) and on days 3 and 4. In the osteogenic differentiation experiment the medium was collected on day 0 and day 18. Glucose and lactate in the proliferation medium were determined photometrically using a Glucose Hexokinase kit and a Lactate kit (DiaSys, Diagnostic Systems, Holzheim, Germany, http://www.diasys-diagnostics.com), employing the manufacturer's ‘sample start’ procedure for both assays. The Wallac 1420 VICTOR2 multilabel plate reader (PerkinElmer, Waltham, MA, http://www.perkinelmer.com) was used to measure the absorbance of the samples. Lactate production was compared with glucose consumption during proliferation and osteogenic differentiation.
Transmission Electron Microscopy (TEM)
Cells for TEM analysis of pBMMSC were harvested before and after the 14-day osteoinduction. The cells were fixed for 10 min in 1% glutaraldehyde and 4% formaldehyde mixture in 0.1 M phosphate buffer and then scraped off and pelleted, followed by a further fixation for 1 hour. The cell pellets were then immersed in 2% agarose in distilled water and postfixed in 1% osmium tetroxide, dehydrated in acetone and embedded in Epon LX 112 (Ladd Research Industries, Williston, VT, http://www.laddresearch.com). A Leica Ultracut UCT ultramicrotome was used to cut thin sections, which were then stained with uranyl acetate and lead citrate. A Philips CM100 transmission electron microscope (FEI Company, Hillsboro, OR, http://www.fei.com) with Morada CCD camera (Olympus Soft Imaging Solutions GMBH, Munster, Germany, http://www.olympus-sis.com) and Tecnai Spirit transmission electron microscope (FEI Company) with Quemesa CCD camera (Olympus, Center Valley, PA, http://www.olympus.com) were used to examine the samples, and image analysis was performed with the MCID Core 7.0 software (InterFocus Imaging Ltd, Linton, UK, http://www.mcid.co.uk).
Determination of pH of the culture media
The cells were suspended in the proliferation media and seeded at density of 1500 cells/cm2. The samples were collected on day 4. Gases were allowed to stabilize for 24 h before collecting the control samples from cell-free flasks. The pH of the proliferation media was measured immediately by Microprocessor pH-Meter CG 840 (SCHOTT Instruments, Mainz, Germany, http://www.schottinstruments.com) with Biotrode pH electrode (Hamilton Company, Reno, NV, http://www.hamiltoncompany.com).
The statistical analyses were performed and all the diagrams drawn using OriginPro 8.6 (OriginLab Corporation, Northampton, MA, http://www.originlab.com). All the hMSC data are presented as means ± standard deviation (SD) based on results from two distinct primary cell lines, unless otherwise indicated. Statistical significance was determined using the Two-Sample t-Test. Values of p < 0.05 were considered statistically significant.
We found that HIF-1α mRNA was highly induced under normoxia in all the hMSCs studied, its expression being 30-fold higher in the UCBMSCs and pBMMSCs and 18-fold higher in the aBMMSCs than in the HEK cells. This resulted in normoxic stabilization of the HIF-1α protein in these primary cell lines, but not in the aBMMSCs. The HIF-1α stabilization was probably due to the increased transcription of HIF1A rather than deficient post-translation degradation of the protein by PHDs. The ratio of mRNAs of HIF-1α to those of PHD2, the main downregulator of HIF-α, was over 6.4 in all the hMSCs, for example, as compared to 0.49 in the HEK cells, but the PHD2 and PHD3 mRNAs were still hypoxia-inducible in the hMSCs, as they were in the HEK cells. It is likely that the PHDs in the aBMMSCs were able to degrade the induced HIF-1α, whereas the ever higher induction in the UCBMSCs and pBMMSCs, where the ratio of HIF-1α to PHD2 was 9.9 and 7.9, respectively, exceeded their degradation capacity. HIF is the master regulator of the hypoxia response, with the task of restoring the oxygen supply and adjusting its metabolism to the cell's needs under hypoxia. A delicate mechanism involving the HIF prolyl 4-hydroxylases, that is, PHDs, the activity of which requires molecular oxygen, has evolved to regulate HIF-α levels post-translationally, even though very little is known about the regulation of HIF-α at the transcriptional level. HIF1A transcription has been shown to be induced by NF-κB, IFNγ, BAF57, and PKM2 in macrophages and cancer cells [21-24], but the only report of HIF-1α mRNA induction in any multipotent cell line concerns hematopoietic stem cells (HSCs), in which it is driven by Meis1 [25, 26]. It was also noted recently in the context of therapeutic neovasculogenesis that mesenchymal stem/progenitor cells (MSPCs) display a low baseline signal for HIF-1α protein at ambient oxygen levels and stabilize HIF-1α at 5% O2, in contrast to endothelial colony-forming progenitor cells (ECFCs), and the HIF-1α in MSPCs is required for neovasculogenesis and the vitality of ECFCs . Our data thus suggest that hMSCs are also able to retain their hypoxic signaling outside the hypoxic stem cell niche environment through the high transcriptional activity of HIF1A.
Oxygen tension is indicated as one major determinant of the fate of hMSCs . MSCs originate from hypoxic niches, for example, bone marrow has approximately 5% O2 as compared to ≥10% O2 in arterial blood (Fig. 5) . At physiological oxygen tensions, the catalytic activity of the PHDs that downregulate HIF-α post-translationally is almost linearly dependent on the oxygen concentration, since their Km values for oxygen are 10% or higher [29, 30]. This means that the activity of PHDs in the hypoxic niches is low, leading to HIF stabilization. We made a surprising finding, however, in that hMSCs cultured under ambient conditions, that is, at 21% O2, have high HIF levels. This means that when hMSCs leave their hypoxic niche and face variable O2 tensions (≥16% O2 in the lung alveoli) they will still be able to maintain high HIF-α levels and glycolysis, thus most likely protecting their phenotype before homing on their final injury repair site (Fig. 5). At the final homing site, the role of oxygen appears to fulfill a dual role, for although hypoxic conditions have been reported in general to promote the undifferentiated state of hMSCs [17, 31], it is well established that hypoxia promotes the chondrogenic differentiation of MSCs and that this is at least partially regulated via HIF-1α , whereas it suppresses the adipogenic differentiation of BMMSCs. Also, hypoxia has been reported to both support and inhibit osteogenic differentiation [33, 34]. The current argument, however, is that bone healing requires oxygenation  and osteoclasts, bone-resorbing cells, are actually activated and osteoblasts inactivated under conditions of hypoxia [36, 37]. Our data support the latter picture, since osteogenic differentiation resulted in HIF-1α downregulation and reversal of the glycolytic metabolism toward oxidative phosphorylation (Fig. 5).
Our data show that the normoxic, stabilized HIF-1 in hMSCs is transcriptionally active and the expression of several glycolytic genes in normoxia is significantly higher in mesenchymal stem cells than in differentiated epithelial HEK cells. Interestingly, very small differences were seen between the hMSCs in the normoxic expression levels of the individual glycolytic HIF target genes, suggesting a common transcriptome for the hMSCs regardless of their tissue of origin. By contrast, osteogenic differentiation of bone marrow-derived hMSCs resulted in reversal of the induction of the glycolytic genes and of glycolysis, suggesting a linkage between the maintenance of multipotency and the glycolytic phenotype. There are also some reports by other groups suggesting that MSCs rely more on glycolysis than mitochondrial respiration [38, 39]. Our data coincide with these reports. Human pluripotent stem cells have been reported to maintain high glycolytic rates, for example, via high levels of hexokinase 2 and inactive pyruvate dehydrogenase , and in line with these data, we found a significant reduction in the expression of hexokinase 2 mRNA and also PDK1 mRNA after osteogenic differentiation, which indicates a metabolic shift toward oxidative phosphorylation and decreased glycolysis. HSCs are also reported to rely on glycolytic metabolism , under the control of Meis1, which upregulates both HIF1A and HIF2A (EPAS) via HIF-1α, whereas HIF-2α contributes to protection against reactive oxygen species and apoptosis .
hMSCs and cancer cells share certain features: both have a high proliferation demand and downregulation of differentiation, and we have now identified another common denominator, induced glycolysis. Cancer cells have been reported to rely on the Warburg effect for their energy production, with induced aerobic glycolysis and repressed mitochondrial oxidative metabolism . It has been suggested that this enables them to convert glucose efficiently into biomass rather than ATP, the former supporting their rapid proliferation. Induced glycolysis may also offer hMSCs a proliferation advantage. Of the cells studied here UCBMSCs had the highest proliferation rate followed by pBMMSCs and aBMMSCs (Fig. 1B). Interestingly, the lack of normoxic HIF-1α stabilization in bone marrow-derived hMSCs from adult donors was also accompanied here by a notably slower proliferation capacity in these cells. No data on age-related differences in multipotency were obtained for the cells studied here.
In this work, we have identified novel markers for hMSCs: namely high normoxic HIF, induced glycolysis, and suppressed oxidative phosphorylation. The therapeutic use of hMSCs in regenerative medicine has been restricted to date by their limited proliferation and vitality in vitro. We suggest that high endogenous HIF-1α and induced glycolysis are novel markers of the multipotency and vitality of hMSCs, and that these noted endogenous properties may support their proliferation, so that culture conditions of low oxygen tension and high glucose concentration, as suggested recently by Deschepper et al., may be of benefit for their in vitro multiplication . The upstream driver of HIF1A induction remains to be identified, as does the time span of the induction: is the high normoxic HIF an infinite property of hMSCs or a restricted one? HIF can be stabilized in normoxia by PHD inhibitors, which are currently undergoing clinical trials for the treatment of anemia . It remains to be determined whether the treatment of differentiated cells with these inhibitors could actually reverse their metabolism toward glycolysis, and whether this could influence their cell fate, promoting multipotency. Another possibility might be to improve the quality of stem cells to be used in regenerative medicine by treating them with PHD inhibitors to promote their multipotency. It has actually been shown that dimethyl oxalylglycine, a pan 2-oxoglutarate-dependent enzyme inhibitor, attenuates apoptosis and cell death, stabilizes HIF-1α, and induces GLUT-1 protein levels in murine MSCs .
Despite their shared surface antigens and multipotency, the hMSCs are considered a somewhat heterogenous group. Our data show that high normoxic expression levels of HIF-1α, induced glycolysis, and suppressed oxidative metabolism are novel characteristics of hMSCs that may be used as additional markers for their characterization. We also show that high HIF-1α is associated with hMSCs being in an undifferentiated state, suggesting that agents that can increase HIF levels might be beneficial in retaining the multipotency of hMSCs, especially in therapeutic settings. Interestingly, induction of HIF-1α was observed in hMSCs at the transcriptional level even though its upstream regulator remains to be identified.
We thank Tanja Aatsinki, Henna Ek, and Minna Savilampi for their expert technical support, the Finnish Red Cross Blood Service for the UCBMSCs, and Biocenter Oulu for the use of its EM core facility. This work was funded by Academy of Finland Grants 120156, 140765, and 218129 (P.K.), the S. Jusélius Foundation (P.K.), The Finnish Cultural Foundation (P.K.), The Finnish Cancer Organizations (P.K.), and Tekes—the Finnish Funding Agency for Technology and Innovation (P.L.).
Disclosure of Potential Conflicts of Interest
The authors indicate no potential conflicts of interest.