Hypoxia-induced reprogramming of glucose-dependent metabolic pathways maintains the stemness of human bone marrow-derived endothelial progenitor cells

The benefits of hypoxia for maintaining the stemness of cultured human bone marrow-derived endothelial progenitor cells (BM EPCs) have previously been demonstrated but the mechanisms responsible remain unclear. Growing evidences suggest that cellular metabolism plays an important role in regulating stem cell fate and self-renewal. Here we aimed to detect the changes of glucose metabolism and to explore its role on maintaining the stemness of BM EPCs under hypoxia. We identified the metabolic status of BM EPCs by using extracellular flux analysis, LC–MS/MS, and 13C tracing HPLC-QE-MS, and found that hypoxia induced glucose metabolic reprogramming, which manifested as increased glycolysis and pentose phosphate pathway (PPP), decreased tricarboxylic acid (TCA) and mitochondrial respiration. We further pharmacologically altered the metabolic status of cells by employing various of inhibitors of key enzymes of glycolysis, PPP, TCA cycle and mitochondria electron transport chain (ETC). We found that inhibiting glycolysis or PPP impaired cell proliferation either under normoxia or hypoxia. On the contrary, inhibiting pyruvate oxidation, TCA or ETC promoted cell proliferation under normoxia mimicking hypoxic conditions. Moreover, promoting pyruvate oxidation reverses the maintenance effect of hypoxia on cell stemness. Taken together, our data suggest that hypoxia induced glucose metabolic reprogramming maintains the stemness of BM EPCs, and artificial manipulation of cell metabolism can be an effective way for regulating the stemness of BM EPCs, thereby improving the efficiency of cell expansion in vitro.


Results
Cell stemness was maintained and metabolic reprogramming induced by hypoxia. The human BM EPCs were previously shown to express HLA-ABC, CD105, CD90, CD44, and CD29; moderately expressed HLA-DR, KDR, and CD54; and negatively express CD34, CD31, CD45, or CD144 3 . These cells were also capable of DiI-Ac-LDL uptake and UEA-1 binding, consistent with endothelial lineage cells. (Fig. 1A). Figure 1B shows the colony forming ability of human BM-derived EPCs under normoxia (~ 20% O 2 ) and hypoxia (1% O 2 ). Colony formation increased to 228 ± 41% of the normoxic rate when hypoxic conditions were employed (p < 0.05). Expression of the stemness markers, Nanog, Oct4, Klf4 and Sox2, was higher under hypoxia than under normoxia (Fig. 1C), indicating more favorable maintenance of stem-like characteristics. To evaluate the functional characteristics of BM EPCs, compared with normoxia, the abilities of in vitro angiogenesis were better maintained under hypoxia, including longer total segments length, longer total tube length, and more number of branches (both p < 0.05, Fig. 1D).

Metabolite levels.
Relative enrichment of specific metabolites was analyzed by Liquid Chromatography-Mass Spectrometry (LC-MS) to define metabolic profiles and a heatmap for intermediates of glucose metabolism is presented in Fig. 2A. Hypoxic conditions resulted in reduced D-glucose and increased glucose-6-P (all p < 0.05), indicating a shift to greater glucose consumption and enhanced glycolysis (Fig. 2B). Levels of Figure 1. Cell stemness and kinetic metabolic profiling assessments. (A) Typical BM EPCs under normoxia were characterized by double positive staining (merged in yellow) with both DiI-AcLDL (red) and FITC-UEA-I (green). Scale bar represents 100 μm. (B) Representative images of BM EPCs colonies. Scale bar represents 1 mm. The number of BM EPCs colonies under hypoxia was significantly higher than that under normoxia (n = 6). (C) qRT-PCR results showing higher expression of the stemness markers, Nanog, Oct4, Klf4 and Sox2 under hypoxia than under normoxia (n = 5). (D) Representative tube networks formed by BM EPCs under normoxia and hypoxia, respectively (n = 3). Images analyzed by the ImageJ plugin "Angiogenesis analyze. " There is an indication of master junctions (pink dots), master segments (yellow), meshes (light blue), branches (green), and isolated segments (blue). Longer total segments length, longer total tube length, and more number of branches were found under hypoxia compared with normoxia. Scale bar represents 200 μm. (E) Representative experiment showing extracellular acidification rate (ECAR) of BM EPCs cultured under normoxia or hypoxia. ECAR representing higher glycolytic rate and capacity in BM EPCs under hypoxia than under normoxia (n = 6). (F) Representative experiment showing oxygen consumption rate (OCR) of BM EPCs cultured under normoxia or hypoxia. OCR for basal and maximal respiration in BM EPCs under hypoxia was lower than under normoxia (n = 6). Data are presented as Mean ± SEM. *p < 0.05; **p < 0.01; ***p < 0.001 versus BM EPCs under normoxia. www.nature.com/scientificreports/  13 C glucose tracing. To ensure that the metabolites listed above originated from free glucose (rather than from gluconeogenesis, for example), stable isotope labeling experiments using 13 C-labeled glucose were performed and isotopic enrichment quantified by High Pressure/Performance Liquid Chromatography-Q-Exactive-Mass Spectrometer (HPLC-QE-MS). No differences were found in 13 C-labeled glycolytic intermediates, such as m + 3 or m + 6 iso-topologues of glucose 6-P, phosphoenolpyruvate or pyruvate, between the two conditions ( Fig. 2E). However, there was a significant increase in 13 C-labeled intermediates of the PPP, such as m + 6, m + 5 or m + 7 iso-topologues of 6-phosphogluconate, ribose 5-P and sedoheptulose 7-P (all p < 0.05), in hypoxic cells (Fig. 2F). Moreover, decreased 13 C-labeling of TCA cycle intermediates, such as m + 2 iso-topologues of citrate, succinate and malate (all p < 0.05), was seen under hypoxia, suggesting a reduced contribution of glucose carbon to TCA cycle activity (Fig. 2G). Non-glucose-derived metabolites (m + 0 iso-topologues) of glycolysis, PPP and TCA cycle showed no significant differences between the two conditions. The above findings indicate greater flow of glucose carbon to the PPP and reduced flow to the TCA cycle under hypoxic conditions.

Effects of PPP inhibition on BM EPCs proliferation. PPP activity is vital for anabolic biosynthesis
and anti-oxidant defense 16 and depends on the diversion of glucose carbon away from the glycolytic pathway. Glucose-6-phosphate dehydrogenase (G6PD) is considered to be a key rate-controlling enzyme of the PPP and can be competitively inhibited by 6-AN ( Promotion of pyruvate oxidation reversed the effect of hypoxia in maintaining cell stemness. Pyruvate dehydrogenase (PDH) catalyzes the irreversible oxidative decarboxylation of the glycolytic product, pyruvate, to the TCA cycle substrate, acetyl coenzyme A and flux through PDH, is negatively regulated by pyruvate dehydrogenase kinase (PDK) enzymes which may be inhibited by the compound, AZD7545 ( Fig. 4A) 18,19 . Inhibition of PDK1, PDK2 and PDK3 by AZD7545, with the resulting stimulation of pyruvate oxidation 20,21 , suppressed cell proliferation (Fig. 4B), reduced colony formation (p < 0.01; Fig. 4C) and reduced expression of Nanog, Oct4, Klf4 and Sox2 (Fig. 4D) under hypoxic conditions. Glycolytic rate and capacity decreased (both p < 0.01, Fig. 4E) and basal and maximal respiration increased (both p < 0.05; Fig. 4F) under the same conditions. Use of AZD7545 under normoxic conditions produced the opposite results (Supplemental Fig. 1F-J). Overall, it appears that the promotion of pyruvate oxidation resulting from use of AZD7545 reversed the effects of hypoxia in maintaining stemness.
Inhibition of pyruvate oxidation promoted cell proliferation under normoxia. The glycolytic product, pyruvate, is transported into the mitochondrion by the mitochondrial pyruvate carrier (MPC) for further oxidation. UK5099 is a potent inhibitor of MPC (Fig. 4A) 22 . Under normoxic conditions, 20 uM UK5099 promoted cell proliferation (p < 0.01; Fig. 4G) and, although colony formation was unaffected (p > 0.05; Fig. 4H), Nanog, Oct4, Klf4 and Sox2, were all upregulated (all p < 0.05; Fig. 4I). Moreover, glycolytic rate (p < 0.05) and capacity (p < 0.001) were both higher in the presence of UK5099 (Fig. 4J) and basal (p < 0.05) and maximal respiration rates (p < 0.01) were both lower (Fig. 4K). By contrast, UK5099 had little effect on cell proliferation, glycolytic rate and capacity, basal or maximal respiration under hypoxic conditions (Supplemental Fig. 1A-E). In combination, the above data suggest that inhibition of pyruvate oxidation promotes glycolysis, reduces mitochondrial respiration and increases BM EPCs proliferation under normoxia. www.nature.com/scientificreports/ Succinyl phosphonate trisodium salt (SP) inhibit KGDHC isoforms from muscle, bacteria, brain and cultured human fibroblasts ( Fig. 5A) 25,26 . No chemical inhibitor of IDH3 was found. TCA cycle inhibition by PCA promoted cell proliferation under conditions of normoxia ( Fig. 5B), although colony formation was unaffected (p > 0.05, Fig. 5C). Expression of Nanog, Oct4, Klf4 and Sox2 was upregulated when PCA was present (all p < 0.05; Fig. 5D). Moreover, use of PCA increased glycolytic rate and capacity (both p < 0.05; Fig. 5E) and decreased basal (p < 0.05) and maximal respiration (p < 0.01; Fig. 5F) compared with controls. Very similar results were obtained with SP ( Fig. 5G-K). Thus, inhibition of the TCA cycle increases flux through glycolysis and proliferation under normoxic conditions. By contrast, use of PCA or SP had little effect on cell proliferation, glycolytic rate and capacity, basal or maximal respiration under hypoxic conditions (Supplemental Fig. 2). The implication from these data is that the TCA cycle may already be maximally suppressed under hypoxic conditions and further inhibition has little detectable effect.

Inhibition of the
Inhibition of the ETC. The complex III inhibitor, antimycin A (AA) was used to block the ETC 27  Adenosine triphosphate (ATP) production. Although a significant suppression of mitochondrial ATP production under-hypoxia (p < 0.001) was found, total cellular ATP production (p > 0.05) was unaffected, compared with normoxia (Supplemental Fig. 4A). Pharmacological manipulation of ATP level was then investigated, although inhibitors of glycolysis or the PPP could not be tested, since they had been shown to suppress BM EPCs growth. UK5099 reduced mitochondrial ATP production (p < 0.05) but did not affect total cellular ATP either under normoxia or hypoxia (p > 0.05, Supplemental Fig. 4B). AZD7545 reduced mitochondrial ATP under noxmoxia (p < 0.05) but did not affect cellular ATP either under normoxia or hypoxia (both p > 0.05; Supplemental  Fig. 4C). Similarly, the two TCA cycle inhibitors, PCA and SP, reduced mitochondrial ATP under normoxia (all p < 0.05) but did not affect cellular ATP level either under normoxia or hypoxia (all p > 0.05; Supplemental  Fig. 4D,E). Use of AA also reduced mitochondrial ATP level under both normoxic (p < 0.001) and hypoxic (p < 0.05) conditions but had no effect on total cellular ATP level under either condition (p > 0.05; Supplemental  Fig. 4F).

Discussion
The current study describes reprogramming of pathways related to glucose metabolism in human BM-derived EPCs under hypoxic conditions of 1% O 2 and demonstrates the relationship between metabolic reprogramming and stemness. Hypoxia was found to increase colony formation and expression of stemness markers compared with normoxia, findings consistent with previous studies 4 . Thus, hypoxic conditions are of benefit for maintenance of stemness in cultured BM EPCs. Experiments using the Seahorse XFe96 Flux Analyzer revealed enhanced glycolytic rate and capacity but reduced basal and maximal respiration under hypoxic conditions. Hypoxia may induce a metabolic switch away from mitochondrial respiration to glycolysis.
Metabolic profiles established by LC-MS were characterized by increased glycolytic and PPP intermediates (glucose-6-P and D-ribulose 5-P) and reduced TCA cycle intermediates (citrate, isocitrate, succinate, fumarate, malate and oxaloacetate) under hypoxic conditions. Increased flow of label from 13 C-glucose was found in PPP  Many rapidly-proliferating cell-types have been shown to undergo similar metabolic shifts. For example, stimulation of glycolysis via hypoxia or inhibition of mitochondrial respiration in pluripotent stem cells promotes stemness 28,29 while inhibition of glycolysis halts proliferation and precipitates cell death 30 . Hypoxia has also been shown to induce a metabolic switch between the PPP and glycolysis in glioma stem-like cells 31 . Glycolysis is far less efficient than mitochondrial respiration in terms of ATP-production, producing only 2ATP/glucose compared with 30-32ATP/glucose from full mitochondrial oxidation. Rapidly-proliferating cells are likely to have high requirements for ATP, as well as for carbon, nitrogen and hydrogen to support biosynthetic activity, which makes a shift towards the less efficient glycolysis pathway difficult to interpret 32 www.nature.com/scientificreports/ of the metabolic cost of synthesizing enzymes for fermentation versus aerobic respiration in E. coli found that energy expenditure in synthesizing proteins for mitochondrial respiration outweighed the energy gain from more efficient ATP-production. Thus, it may make more sense for fast-growing cells to utilize glycolysis for efficient cell growth and proliferation 34 . Inhibitors of glucose metabolism were used to assess the impact of the dependent pathways on cell stemness. Inhibition of the glycolytic enzymes, HK II, PFK and PKM, by BP, 3PO and PKM2-IN-1, respectively, reduced growth rates of BM EPCs under both hypoxia and normoxia. BP has previously been shown to inhibit ovarian cancer growth and to induce apoptosis of hepatocellular carcinoma cells and human breast cancer cells [35][36][37] . PKM2-IN-1 reduced glucose uptake and lactate production in esophageal and hepatocellular carcinoma cells 38,39  www.nature.com/scientificreports/ and 3PO inhibited growth of lung cancer, breast cancer and promyelocytic leukemia cells 14 . The above findings demonstrate the importance of glycolysis for proliferation of tumor cells.
The PPP produces ribose-5-phosphate for DNA/RNA synthesis and NADPH for fatty acid synthesis and branches off from glycolysis during the preparatory phase. Shunting of glucose carbon through the PPP balances requirements for catabolic generation of ATP and reducing cofactors with production of biosynthetic substrates to meet anabolic needs 40,41 . Inhibition of the PPP enzyme, G6PD, by 6-AN, inhibited proliferation of glioblastoma cells and colony formation in embryonic stem cells 17,42 . The current study also found that 6-AN reduced EPC proliferation, demonstrating a critical role of the PPP in EPC growth.
Pyruvate is central to glucose metabolism, whether through lactate fermentation or the TCA cycle 43 , and also to synthesize lipids and amino acids. Use of the PDK inhibitor, AZD7545, to promote pyruvate oxidation www.nature.com/scientificreports/ has been shown to inhibit the growth of melanoma 44 and suppress proliferation of activated primary human CD4 + T cells and primary mouse T cells 45 . The current study found that AZD7545 decreased glycolytic rate and capacity, increased basal and maximal respiration and reduced EPC proliferation under hypoxic conditions. Thus, promotion of pyruvate oxidation appears to reverse the effect of hypoxia in promoting cell proliferation. By contrast, use of the MPC blocker, UK5099, has previously been shown to induce metabolic reprogramming and enhance stem-like properties in human prostate cancer cells 46 . Similarly, ovarian cancer cells switched to aerobic glycolysis with reduced ATP production and became more migratory and resistant to chemotherapy and radiotherapy in the presence of UK5099 47 . The current study found increased glycolytic rate and capacity, decreased basal and maximal respiration and enhanced BM EPCs proliferation under normoxic but not under hypoxic conditions when UK5099 was present.
The TCA cycle inhibitors, PCA and SP, increased glycolytic rate and capacity, decreased basal and maximal respiration and enhanced BM EPCs proliferation under normoxia. However, these inhibitors had little effect under hypoxia, a state in which the TCA cycle is suppressed. Similarly, the complex III inhibitor, AA, increased glycolysis, decreased maximal respiration, increased cell proliferation and stem cell marker mRNA levels under normoxia but not under hypoxia.
Metabolism has traditionally be regarded as fulfilling three roles: ATP-production for energy-requiring processes; production of precursors for anabolism and production of intermediates for enzyme-catalyzed reactions 48 . The first of these roles, ATP production, was found to be suppressed on a mitochondrial level but unaffected on a cellular level by hypoxic conditions. Similar results were found by use of inhibitors. Inhibition of pyruvate oxidation by UK5099 reduced mitochondrial, but not cellular, ATP under both normoxia and hypoxia. Inhibition of the TCA cycle by PCA or SP also reduced mitochondrial ATP under noxmoxic conditions but did not affect cellular ATP either under normoxia or hypoxia. The ETC inhibitor, AA, reduced mitochondrial ATP under both normoxia and hypoxia but had no effect on total ATP. The import of these data is that energy production does not have much significance in cell proliferation and production of biosynthetic and reaction intermediates may have greater responsibility for metabolic reprogramming.
Taken together, we have shown that hypoxic conditions of 1% O 2 maintain BM EPCs stemness and induce metabolic reprogramming, manifested as increased glycolysis and PPP and decreased TCA cycle and mitochondrial respiration. The current study has demonstrated that inhibition of glycolysis or the PPP or promotion of mitochondrial respiration resulted in a reversal of the effect of hypoxia in maintaining cell stemness. By contrast, inhibition of pyruvate oxidation, the TCA cycle or the ETC promoted BM EPCs proliferation under normoxic conditions mimicking hypoxic conditions. In conclusion, hypoxia induced glucose metabolic reprogramming, and such pattern of metabolic reprogramming maintains the stemness of BM EPCs, and artificial manipulation of cell metabolism can be an effective way for regulating the stemness of BM EPCs.
We acknowledge some limitations to the current study. Firstly, we tested the effects of glucose metabolism on EPCs stemness by intervention at different stages of glucose metabolic pathways. However, this is just a preliminary blueprint study, the underlying molecular mechanisms of each specific enzyme involved in hypoxiainduced metabolic reprogramming and maintenance of BM EPCs stemness remain to be clarified. Secondly, www.nature.com/scientificreports/ we only detected glucose metabolism, and the other metabolism, such as amino acid metabolism and lipid metabolism, are worthy of future study on their relationship with EPCs stemness. Thirdly, in vivo experiments are required to clarify whether hypoxia or drug-induced enhancement of BM EPCs stemness has application to clinical transplantation therapy. Tube formation assay. The tube formation assay was carried out to estimate the angiogenic capacity of BM EPCs. Briefly, EPCs (3 × 10 4 cells) were seeded onto 96-well plates coated with growth factor-reduced matrigel (70 μl/well; BD Biocoat) and cultured in EBM-2 at 37 °C under hypoxia or normoxia. After incubation for 8 h, tubes in each well were visualized under an inverted microscope, photographed at 10 × magnification, and analyzed using the Angiogenesis Analyzer for ImageJ software. Extracellular flux analysis. Kinetic  with labeling medium containing 2 g/L [U6]-13 C-glucose (Sigma) followed by incubation for 24 h. 250 μL of water was added to 1 × 10 5 EPCs followed by vortex mixing for 30 s. Samples were pre-cooled in dry ice, freezethawed three times in liquid nitrogen, vortexed for 30 s and sonicated for 10 min in an ice-water bath. 750 μL of methanol was added to the sample solution with vortex mixing and vibration at 4 °C for 15 min, followed by incubation at − 40 °C for 1 h. Samples were centrifuged at 12,000 rpm (RCF = 13,800(× g), R = 8.6 cm) for 15 min at 4 °C. 800 μL of the clear supernatant was collected and dried in a vacuum concentrator. Residue was reconstituted with ultrapure water according to the cell count, vortexed and sonicated for 10 min in an ice-water bath before filtration through the centrifuge tube filter. Supernatant were transferred to inserts in injection vials for HPLC-QE-MS analysis.

Methods
HPLC separation was carried out using a Thermo Scientific Dionex ICS-6000 HPLC System (Thermo Scientific) equipped with Dionex IonPac AS11-HC (2 × 250 mm) and AG11-HC (2 mm × 50 mm) columns. Mobile phase A consisted of 100 mM NaOH in water and phase C of water. An additional pumping system was used to supply the solvent (2 mM acetic acid in methanol) and solvent mixed with eluent before loading of the ion source (flow rate of 0.15 mL/min). The column temperature was set at 30 °C, that of the auto-sampler at 4 °C and the injection volume was 5 μL. The QE mass spectrometer was used to acquire MS spectra in full MS mode with acquisition software (Xcalibur 4.0.27, Thermo) to continuously evaluate the full scan MS spectrum.
Proliferation assay. BM EPCs were seeded onto 96-well plates (4000 cells per well) and incubated under normoxia or hypoxia for 1 day in a final volume of 90 μL/well. 10 μL of inhibitor or vehicle (with or without DMSO) were added to each well and incubated under conditions of normoxia/hypoxia for 3 days. 10 μL Cell Counting Kit-8 (Dojindo, Japan) mixture was added and the plate incubated for 2 h before measurement of absorbance at 450 nm by microplate reader (Thermo FC). ATP measurement. Total ATP was measured using ATP Assay Kit (Beyotime) according to the manufacturer's instructions. Briefly, 5 × 10 5 BM EPCs were lysed with 200 μL lysis buffer and ultrasonicated after washing twice with PBS. The lysate was centrifuged at 12 000 × g for 15 min at 4 °C. 20 uL per well supernatant was added to a microplate together with 100 uL ATP detection solution. Chemiluminescence was detected by a luminescence reader (Synergy H1). Mitochondrial ATP was measured by using the XF Cell Mito Stress Test Kit (Seahorse Bioscience, North Billerica, MA) by subtracting the rate measurement before oligomycin addition from that after oligomycin. Statistical analysis. Data are presented as mean ± SEM and paired Student's t-test was used to compare parameter data, Wilcoxon test for nonparametric data and ANOVA for repeated measurement data. A value of p < 0.05 was considered significant.
Ethics approval and consent to participate. This study was approved by the Ethical Committee of Zhujiang Hospital, Southern Medical University (No. 2019-SJWK-007, 2019-3-21, Guangdong, China) and performed in accordance with the Declaration of Helsinki.