HIF1a Modulates Cell Fate ReprogrammingThrough Early Glycolytic Shift and Upregulationof PDK1–3 and PKM2

ALESSANDRO PRIGIONE,a,b* NADINE ROHWER,c SHEILA HOFFMANN,b BARBARA MLODY,a

KATHARINA DREWS,a RAUL BUKOWIECKI,a,b KATHARINA BLUMLEIN,d ERICH E. WANKER,b

MARKUS RALSER,d THORSTEN CRAMER,c JAMES ADJAYEa,e

Key Words. Reprogramming • Induced pluripotent stem cells • Metabolism • Hypoxia-induciblefactor 1a • Pyruvate dehydrogenase kinase 1 • Pyruvate kinase isoform M2

ABSTRACT

Reprogramming somatic cells to a pluripotent state drastically reconfigures the cellular anabolicrequirements, thus potentially inducing cancer-like metabolic transformation. Accordingly, we andothers previously showed that somatic mitochondria and bioenergetics are extensively remodeledupon derivation of induced pluripotent stem cells (iPSCs), as the cells transit from oxidative to glyco-lytic metabolism. In the attempt to identify possible regulatory mechanisms underlying this meta-bolic restructuring, we investigated the contributing role of hypoxia-inducible factor one alpha(HIF1a), a master regulator of energy metabolism, in the induction and maintenance of pluripotency.We discovered that the ablation of HIF1a function in dermal fibroblasts dramatically hampers reprog-ramming efficiency, while small molecule-based activation of HIF1a significantly improves cell fateconversion. Transcriptional and bioenergetic analysis during reprogramming initiation indicated thatthe transduction of the four factors is sufficient to upregulate the HIF1a target pyruvate dehydrogen-ase kinase (PDK) one and set in motion the glycolytic shift. However, additional HIF1a activationappears critical in the early upregulation of other HIF1a-associated metabolic regulators, includingPDK3 and pyruvate kinase (PK) isoform M2 (PKM2), resulting in increased glycolysis and enhancedreprogramming. Accordingly, elevated levels of PDK1, PDK3, and PKM2 and reduced PK activity couldbe observed in iPSCs and human embryonic stem cells in the undifferentiated state. Overall, the find-ings suggest that the early induction of HIF1a targets may be instrumental in iPSC derivation via theactivation of a glycolytic program. These findings implicate the HIF1a pathway as an enabling regula-tor of cellular reprogramming. STEM CELLS 2014;32:364–376

INTRODUCTION

Oxygen concentration plays a critical role inmediating modifications of the cellular meta-bolic profile [1]. Under physiological normoxicconditions, human somatic cells are character-ized by active mitochondria and oxidativephosphorylation (OXPHOS)-based metabolism,while oxygen-deprived cells exhibit increasedconversion of glucose to lactate (the “Pasteureffect”). However, tumor cells fine-tune theircellular bioenergetics to respond to higher cel-lular demands and can shift to glycolysis-basedmetabolism even in the presence of high levelof oxygen, a phenomenon known as aerobicglycolysis or “Warburg effect” [2, 3]. This met-abolic shift seems apparently counter-intuitivegiven the low efficiency of glycolytic metabo-lism in terms of generation of ATP molecules.Nevertheless, several lines of evidence demon-

strate that this metabolic adaptation endowsproliferative cells with critical advantages.

First, anabolic pathways branching outfrom the glycolytic path supply the intermedi-ates for cell growth, including amino acids andlipid precursors [4, 5]. Thus, rapidly dividingcells, such as cancer cells, increase the fluxthrough glycolysis to satisfy their need of mac-romolecules, by enhancing glucose uptake andby slowing the entry of pyruvate into mito-chondria [6]. Second, the energy reconfigura-tion can provide protection against oxidativestress [5], by avoiding high levels of reactiveoxygen species (ROS), common by-products ofmitochondrial respiration, and by re-routingglycolytic intermediates into the pentose phos-phate pathway (PPP), which generates notonly essential nucleotide precursors but alsothe reducing factor NADPH, required for theactivity of antioxidant enzymes [7]. Recent

aMolecular Embryology andAgeing Group, Department ofVertebrate Genomics, MaxPlanck Institute for MolecularGenetics, Berlin, Germany;bDepartment ofNeuroproteomics, MaxDelbrueck Center forMolecular Medicine (MDC),Berlin, Germany;cMedizinische Klinik mitSchwerpunkt Hepatologieund Gastroenterologie,Campus Virchow-Klinikum,Charit�e-Universit€atsmedizinBerlin, Berlin, Germany;dDepartment of Biochemistryand Cambridge SystemsBiology Centre, University ofCambridge, Cambridge,United Kingdom; eInstitutefor Stem Cell Research andRegenerative Medicine,Medical Faculty, HeinrichHeine University, Duesseldorf,Germany

Correspondence: AlessandroPrigione, M.D., Ph.D., Robert-Roessle-Str. 10, D-13125 Berlin-Buch, Germany. Telephone:149-30-9406-2871; Fax:149-30-9406-3869; e-mail:alessandro.prigione@mdc-berlin.de.; or James Adjaye,Ph.D., Institute for Stem CellResearch and RegenerativeMedicine, Medical faculty, Hein-rich Heine University, Duessel-dorf, Germany. Telephone:149–211 81 08191; Fax:149–211 81 19147; e-mail:James.Adjaye@med.uni-duesseldorf.de

Received July 2, 2013; acceptedfor publication August 27, 2013;first published online in STEM

CELLS EXPRESS October 1, 2013.

VC AlphaMed Press1066-5099/2013/$30.00/0

http://dx.doi.org/10.1002/stem.1552

Stem Cells 2014;32:364–376 www.StemCells.com VC AlphaMed Press 2013

EMBRYONIC STEM CELLS/INDUCEDPLURIPOTENT STEM CELLS

data on pyruvate kinase (PK), which catalyzes the conversionof phosphoenolpyruvate into pyruvate in the last step of theglycolytic cascade, support the idea that the Warburg effectmay promote cellular redox homeostasis. Pyruvate kinase iso-form M2 (PKM2) is highly expressed in cancer cells [8, 9] andupon oxidation it looses activity, thereby reducing pyruvateformation and diverting the glycolytic flux into the PPP, whicheventually supports antioxidant activities [10–12].

A key mediator of the metabolic reconfiguration occurringunder low oxygen conditions is the transcription factorhypoxia-inducible factor one (HIF1) [13, 14]. HIF1 is a hetero-dimer consisting of a constitutively expressed HIF1 b subunitand an oxygen-regulated HIF1a, which is physiologicallydegraded under normoxic conditions by oxygen-dependentprolyl-hydroxylases (PHD1–3). When oxygen level decreases, orwhen PHD enzymes are pharmacologically inhibited, HIF1aprotein escapes degradation and translocates into the nucleus,where it initiates a gene expression program which leads to aswitch from OXPHOS to glycolysis. HIF1a target genes includeglucose transporters, to increase glucose uptake, and pyruvatedehydrogenase kinases (PDK1–3) [15–17], to shunt pyruvateaway from the mitochondria through the inhibition of pyru-vate dehydrogenase. In addition, HIF1a interacts with PKM2and promotes its gene transcription [18], further implying aninstructive role for HIF1a downstream signaling in theWarburg-like restructuring of glucose metabolism.

The derivation of induced pluripotent stem cells (iPSCs),allowing somatic cells to acquire embryonic stem cells (ESCs)-likefeatures [19], is also associated with a profound reconfigurationof anabolic demands. Indeed, iPSCs show high proliferation rateand distinct cell cycle features compared to parental somatic cells[20]. This suggests that a corresponding reprogramming of energymetabolism may also be in place. Accordingly, we previously dis-covered that somatic mitochondria and cellular bioenergetics areextensively remodeled upon cellular reprogramming as the cellsadopt a glycolytic metabolism [21]. Following these initial obser-vations, several groups further confirmed that a Warburg-likemetabolic reconfiguration takes place in both mouse and humaniPSCs [22–26]. Indeed, the process of “metabolic reprogramming”is now being recognized as an emerging important step of theinduction of pluripotency [27–30].

The metabolic switch of cell fate reprogramming does notappear to be due to dysfunctional mitochondria, which are infact capable of respiring and consume oxygen, as demon-strated by bioenergetic profiling studies [25, 26, 31]. However,like cancer cells, proliferating pluripotent stem cells (PSCs)may opt for glycolysis as they necessitate building biomassand at the same time maintaining redox homeostasis. Inaccordance, PSCs exhibit upregulation of genes involved inglucose uptake and the initial steps of glycolysis, increasedexpression of PDK1 [25, 31], suggesting the rerouting ofmetabolism outside of the mitochondria, and elevated levelsof glucose-6-phosphate [31], indicative of enhanced fluxthrough the pentose phosphate pathway. Moreover, ROS lev-els are also reduced in PSCs and so is the amount of oxidativedamage [21, 32]. Finally, exposure to hypoxic environmentfavorably supports self-renewal and pluripotency [33–35],enhances iPSC generation [36], and maintain hESCs in a moredevelopmentally immature state [37].

Here, we sought to investigate the mechanisms underlyingthe metabolic reprogramming occurring upon cell fate transi-

tion and specifically dissect the contribution of the HIF1apathway. We found that a small molecule mimicking HIF1aactivation enhances reprogramming, while the ablation ofHIF1a results in a dramatic loss of colony formation. By per-forming transcriptional and bioenergetic profiling during earlyreprogramming, we discovered that the transduction of thefour Yamanaka factors (4F: OCT4, SOX2, KLF4, and c-MYC) issufficient to upregulate PDK1 and thereby initiating a glyco-lytic shift. The exposure to hypoxia or to HIF1a activation fur-ther stimulates the expression PDK3 and PKM2, resulting inincreased early switch to glycolysis and more efficient iPSCgeneration. The additional upregulation of PDK3 and PKM2might be critical in enhancing reprogramming, since weobserved that their expression is elevated in undifferentiatedPSCs and is coupled to reduced PK activity, all traits associ-ated with a glycolytic state. Taken together, early induction ofHIF1a-associated glycolytic modulators may be instrumental inthe establishment of pluripotency and may possibly representan enabling regulatory step during cell fate conversion.

MATERIALS AND METHODS

Cell Lines and Culture Conditions

Neonatal foreskin fibroblasts (FFs) HFF1 and BJ were purchasedfrom ATCC (#SCRC-1041 and #CRL-2522, respectively), and der-mal fibroblasts (DFs) NFH2 were previously derived from an84-year-old woman [38]. All fibroblasts were cultured usingDulbecco’s modified Eagle’s medium (DMEM) supplementedwith 10% bovine serum, nonessential amino acids, L-glutamine,penicillin/streptomycin, and sodium pyruvate (all from Invitro-gen, Carlsbad, CA, http://www.invitrogen.com). All iPSC lineswere previously generated using the four Yamanaka factors (4F:OCT4, SOX2, KLF4, and c-MYC) retroviral cocktail: HFF1-derivediPSCs (lines iPS2 and iPS4) [21], BJ-derived iPSCs (lines iB4 andiB5) [31], NFH2-derived iPSCs (lines OiPS3, OiPS6, OiPS8, andOiPS16) [38]. Human embryonic stem cell (hESC) lines H1 andH9 (WiCell Research Institute, Madison, WI, http://www.wicell.org) and iPSCs were cultured in hESCs media containing KO-DMEM supplemented with 20% knockout serum replacement,nonessential amino acids, L-glutamine, penicillin/streptomycin,sodium pyruvate, 0.1 mM b-mercaptoethanol (all from Invitro-gen), and 8 ng/mL basic fibroblast growth factor (Peprotech,Rocky Hill, NJ, http://www.peprotech.com). PSCs were har-vested in feeder-free conditions using DMEM-F12 media sup-plemented with N2/B27. All cultures were normally kept in ahumidified atmosphere of 5% CO2 at 37�C under atmosphericoxygen condition (20%).

HIF1a Activation

To generate hypoxic conditions, the oxygen concentration wasset to 1% and the cells were maintained under hypoxia for 24hours. A small molecule activator of HIF1a was used, ethyl 3,4-dihydroxybenzoate (EDHB) (Sigma, #E24859), at a concentrationof 100 mM, as previously shown [39]. To test the effect of HIF1aactivation on the early reprogramming-initiating events, FFswere transduced twice with the four factor (4F) retroviral cock-tail, as previously described [40], alone or in combination with100 mM EDHB treatment. The cells were then harvested after24 hours from the first transduction (4F 24h and 4F EDHB 24h),after 48 hours from the first transduction (that means 24 hours

Prigione, Rohwer, Hoffman et al. 365

www.StemCells.com VC AlphaMed Press 2013

after the second transduction) (4F 48h and 4F EDHB 48h), andafter 72 hours from the first transduction (4F 72h and 4F EDHB72h). In addition, FFs were treated only with 100 mM EDHB for24 hours (EDHB 24h), 48 hours (EDHB 48h), and 72 hours(EDHB 72h).

HIF1a Knockdown

To stably knockdown HIF1a, BJ fibroblasts were transducedwith lentiviruses containing short hairpin RNA (shRNA)sequences against human HIF1a (BJ-HIF1-KD) and scrambledcontrol oligonucleotides (BJ-SCR-KD) (TIB MOLBIOL, Berlin,Germany, http://www.tib-molbiol.de) [41]. Oligonucleotideswere inserted into the lentiviral bicistronic vector pPR1, whichallows for coexpression of green fluorescent protein (GFP)[42]. Recombinant lentiviruses were produced in 293T cellsusing the calcium-phosphate method. Human BJ fibroblastsstably expressing shRNAs were generated by double transduc-tion with lentiviruses at a multiplicity of infection of 10 for 24hours. Transduction efficiency of target cells was determinedby flow cytometry analysis of GFP using a FACSCalibur (BectonDickinson, Heidelberg, Germany, http://www.bd.com).

Cellular Reprogramming

To test the consequences of HIF1a manipulation on the over-all efficiency of iPSC generation, BJ fibroblasts, SCR-KD BJfibroblasts, HIF1-KD BJ fibroblasts, and BJ fibroblasts treatedwith EDHB were reprogrammed to pluripotency using retrovi-ral vectors expressing the four Yamanaka factors, followingour previously published protocol [21]. To test the effect ofHIF1a knockdown on viral-free iPSC derivation, BJ fibroblasts,SCR-KD BJ fibroblasts, and HIF1-KD BJ fibroblasts were trans-fected using three nonintegrative episomal plasmids contain-ing a total of seven factors (4F plus NANOG, LIN28, and SVLT),as previously described [43]. Briefly, 8 3 105 fibroblasts (BJ,SCR-KD, and HIF1-KD) were nucleofected using the Amaxa CellLine Nucleofector Kit (Lonza, Basel, Switzerland, http://www.lonza.com). After nucleofection, fibroblasts were imme-diately mixed with 500 mL DMEM medium before seedinginto six-well plates. The following day, the cells were culturedusing hESCs medium supplemented with a small moleculecocktail composed of CHIR99021, A-83-01, PD0325901, andY-27632 (all from Stemgent, https://www.stemgent.com) [43].Four weeks after either retroviral or plasmid reprogramming,the cells were fixed and stained for NANOG expression usingthe ABC method (see below). The reprogramming efficiencywas defined as the number of NANOG positive colonies rela-tive the total starting number of fibroblasts.

Global Gene Expression Analysis

Biotin-labeled cRNA samples were produced as previouslydescribed [21] and hybridized onto Illumina human-8 Bead-Chips version 3 (Illumina, San Diego, CA). The following sam-ples were used: BJ, BJ-HIF1-KD, BJ-SCR-KD, 4F 24h, 4F 48h, 4F72h (hybridized in duplicate), EDHB 24h, EDHB 48h, EDHB72h, 4F EDHB 24h, 4F EDHB 48h, 4F EDHB 72h (hybridized insingle). In addition, previously generated array data wereincorporated in the analysis, including: amniotic fluid cells(AFCs), FFs (HFF1 and BJ), DFs (NFH2), hESC lines (H1 andH9), FFiPSC lines (iPS2, iPS4, iB4, and iB5), DFiPSC lines(OiPS3, OiPS6, OiPS8, and OiPS16) (all hybridized in duplicate),and AFiPSC (lines 4, 5, 6, 10, hybridized in single, and line 41,

hybridized in duplicate) [21, 31, 38, 44]. Microarray analysis,principal component analysis (PCA) plot, and the general heat-map were performed using the R/Bioconductor package.Genes were considered significantly expressed with detectionp values �.01. Differential expression analysis was performedusing the Illumina custom method, using differential p values�.01, fold change ratio >1.5. The heatmap for energy metab-olism was generated using Microarray Software Suite TM4(TMEV.bat) with an input list adapted from SA BiosciencesPCR arrays (Human Glucose Metabolism PCR Array, www.sabiosciences.com). Pathway analysis was determined by map-ping onto the kyoto encyclopedia of genes and genomes(KEGG) pathways using Database for Annotation, Visualizationand Integrated Discovery (DAVID) (http://david.abcc.ncifcrf.gov). Microarray results have been deposited in the geneexpression omnibus (GEO) database (accession numberGSE37709).

Quantitative Real-Time Polymerase Chain Reaction

Real-time polymerase chain reaction (PCR) was performed in384 or 96 Well Optical Reaction Plates (Applied Biosystems,Foster City, CA, http://www.appliedbiosystems.com) using SYBR-Green PCR Master Mix (Applied Biosystems). Reactions werecarried out on the ABI PRISM 7900HT Sequence Detection Sys-tem (Applied Biosystems). Duplicate or triplicate amplificationswere carried out for each target gene with at least three wellsserving as negative controls. Quantification was performedusing the comparative Ct method (ABI instruction manual), nor-malized over ACTB, and presented as a log2 values with respectto the biological controls. The list of all primers used in thisstudy is presented in Supporting Information Table 4.

Immunostaining and Western Blotting

Avidin-biotin complex (ABC) method was used to stainNANOG-positive hESC-like colonies, as described elsewhere[45]. Briefly, after primary incubation with NANOG antibody(1:100, #ab62734; Abcam, Cambridge, U.K., http://www.abcam.com), biotinylated universal secondary antibody wasapplied following the manufacturer’s instruction (ABC univer-sal kit #PK-6200; Vector Laboratories, Burlingame, CA, http://www.vectorlabs.com). The ABC reagent was then added tobind horseradish peroxidase (HRP) to the primary/secondarycomplex. HRP enzyme activity was then visualized using thechromogen substrate diaminobenzidine tetrachloride (Sigma #D5637). Senescence-associated b-galactosidase staining wasperformed according to the manufacturer’s protocol (Cell Sig-naling, Danvers, MA, www.cellsignal.com). Cells stained forNANOG or b-galactosidase were photographed using a digitalcamera (Canon).

For Western blot analysis, nuclear protein extracts wereprepared as described before [46], then resolved byelectrophoresis on an 8% sodium dodecyl sulfate-polyacrylamidegel, and transferred to a nitrocellulose membrane (AmershamBiosciences, Freiburg, Germany). Blots were probed with antibod-ies against HIF1a (AB1536; R&D Systems, Minneapolis, MN,http://www.rndsystems.com), HIF2a (ab199; Abcam), and YY1 (sc-281; Santa Cruz Biotechnology, Santa Cruz, CA, http://www.scbt.com). Secondary antibodies were conjugated to HRP (Dianova,Hamburg, Germany) and peroxidase activity was visualized usingthe Western Lightning Chemiluminescence Reagent Plus (PerkinElmer Life Sciences, Boston, MA, http://www.perkinelmer.com).

366 HIF1a Associated Metabolic Switch in iPSCs

VC AlphaMed Press 2013 STEM CELLS

Absolute Quantification of PKM1 and PKM2

Mass spectrometry-based absolute quantification was used asdescribed previously [8] to quantify PKM1 and PKM2 isoforms,employing an AQUA method. Briefly, protein samples fromyeasts carrying p414TEF-PKM1 or p413TEF-PKM2, somaticcells, and pluripotent stem cells were separated on a 10%SDS-PAGE gels, and the mass region between 50 and 70 kDawas excised. The gel pieces were then subjected to an “in-gel”tryptic digest. For quantitation of the tryptic peptides of inter-est, their corresponding AQUA analog were spiked in the sam-ples after the tryptic digest. Analysis was performed on ananoLC (Eksigent, Ultra 2D) coupled online to a hybrid triplequadrupole/ion trap mass spectrometer (AB/SCIEX, QTRAP5500). The identity of the quantified peptides was confirmedby collecting of tandem mass spectrometry (MS/MS) spectraon the QTRAP instrument operating in iontrap mode. In orderto confirm specificity of the selected tryptic peptides, yeastsexpressing either only PKM1 or PKM2 were used. The quanti-tative values obtained for PKM1 and PKM2 were set in ratiowith PKM112, and all samples were corrected accordingly.

PK Activity

PK was assessed using the pyruvate kinase activity assay kit(MAK072; Sigma), according to the manufacturer’s instruction.Briefly, cells were rapidly homogenized and the rate of PKactivity was measured by assessing the fluorescence intensityevery 5 minutes until the value of the most active samplewas higher than the one of the highest standard. The resultswere then reported to the total number of cells calculatedaccording to the BCA protein assay kit (23225, Pierce, ThermoScientific, Rockford, IL, http://www.piercenet.com). Both PKactivity and protein measurements were obtained with aTecan reader (InfiniteM200, http://www.tecan.com).

Bioenergetic Profiling

Assessment of cellular energy metabolism was performedusing Seahorse XF24 extracellular flux analyzer (Seahorse Bio-science, www.seahorsebio.com), as previously described [31].The instrument allows the simultaneous quantification ofmitochondrial respiration (oxygen consumption rate, OCR)and glycolysis to lactic acid (extracellular acidification rate,ECAR). Four mitochondrial inhibitors (all from Sigma) wereused in succession. After three basal measurements, 1 mMoligomycin, a complex V blocker, was added to inhibitOXPHOS. After time point 6, the uncoupling agent carbonylcyanide 4-(trifluoromethoxy)phenylhydrazone (FCCP) wasinjected into the wells leading to the collapse of the mito-chondrial membrane potential and to the consumption ofoxygen in the absence of ATP production. The same FCCPconcentration (1 mM) was added again after time point 9 tomonitor the continuous mitochondrial uncoupling. Finally, 1mM rotenone (complex I blocker) and 1 mM antimycin A(complex III blocker) were simultaneously injected to com-pletely inhibit mitochondrial respiration, thus enabling thecalculations of both mitochondrial and non-mitochondrialrespiratory fractions. 40,000 fibroblasts were plated into eachwell of the XF-24 well plates approximately 18 hours beforethe analysis. Assays were initiated by removing the growthmedium and replacing it with unbuffered media, prepared aspreviously described [31].

Statistical Analysis

Data are expressed as mean and SEM, unless stated other-wise. Data were analyzed using GraphPad-Prism software(Prism 4.0, GraphPad Software, Inc.) and Windows XP Excel(Microsoft).

RESULTS

HIF1a Activation Stimulates Glycolysis and Enhancesthe Efficiency of Reprogramming

We previously discovered that the protein expression level ofPDK1 is elevated in human PSCs compared to somatic cells orto PSC-differentiated cells [31]. These findings have been inde-pendently demonstrated [25] and further supported by theobservation that small molecule-based PDK1 induction cansignificantly improve cellular reprogramming [22]. Since PDK1is a known downstream target of HIF1a [15, 16], we firstsought to determine the level of HIFa protein expression inundifferentiated PSCs. In agreement with previous dataobtained on hESCs [34], we verified that HIF1a and HIF2a arenot constitutively activated in human PSCs (Fig. 1A). The pat-tern of nuclear accumulation following hypoxic stimulationwas a bit different for HIF1a and HIF2a, as it was comparablein both fibroblasts and PSCs for HIF1a and instead only pres-ent in PSCs for HIF2a, suggesting that HIF2a could play amore important role in undifferentiated stem cells, as previ-ously described [47].

We then investigated whether HIF1a activation may influ-ence the induction of pluripotency. We observed that the dailyaddition of 100 mM of the PHD inhibitor Ethyl 3,4-dihydroxy-benzoate (EDHB)) to BJ foreskin fibroblasts could lead to signif-icant increase in the efficiency of reprogramming (from0.001% in BJ cells to 0.005% in BJ cells treated with EDHB).This was assayed based on the number of hESC-like coloniesexpressing the pluripotency-regulating protein NANOG 4 weeksafter retroviral transduction of the four Yamanaka factors (4F)(Fig. 1B, Student’s t test, 4F vs. 4F EDHB, p5 .0063). Thesefindings are in agreement with a previous report showing thata different PHD inhibitor, N-oxaloylglycine, and another HIF1ainducer, Quercetin, could lead to increased efficiency of cellu-lar reprogramming in human fibroblast cells [22]. Moreover,HIF1a overexpression has been found to improve the inductionof iPSC-like colonies in the A549 cancer cell line [48].

Since we previously demonstrated that reprogramming topluripotency is associated with a shift to glycolysis [21], wetested whether the HIF1a activator that facilitated reprogram-ming was capable of enhancing glycolysis. Indeed, short-termtreatment with EDHB was sufficient to decrease the rate ofcellular OXPHOS (Fig. 1C) and increase glycolytic metabolism(Fig. 1D). Several parameters related to mitochondrial respira-tion were strongly lowered by the treatment, including thebasal respiration, the ATP turnover, the maximal respirationrate, and the spare respiratory capacity (Supporting Informa-tion Fig. 1). Overall EDHB significantly reduced the ratiobetween OCR and ECAR (Fig. 1E, Student’s t test, EDHBtreated vs. untreated, p< .005), indicative of a switch to gly-colysis. Hence, mimicking HIF1a stimulation in somatic cellscan amplify the Warburg-like metabolic shift thereby improv-ing iPSC generation.

Prigione, Rohwer, Hoffman et al. 367

www.StemCells.com VC AlphaMed Press 2013

HIF1a Depletion Downregulates the Glycolytic Pathwayand Hampers iPSC Generation

In order to establish whether the activation of the HIF1apathway during reprogramming is critical for the generation

of iPSCs rather than simply supportive, we stably knocked-

down HIF1a in BJ fibroblasts (HIF1-KD) using a lentivirus-

based RNA interference approach. A scrambled knockdown

(SCR-KD) was included as control. Immunoblot analysis con-

firmed that HIF1-KD fibroblasts were incapable of accumulat-

ing HIF1a protein within the nucleus under hypoxic exposure

(Fig. 2A). Both HIF1-KD BJ and SCR-KD BJ exhibited normal

fibroblast-like growth features and did not show signs of early

senescence, as shown by b-galactosidase staining (Supporting

Information Fig. 2A).Strikingly, NANOG-positive colonies failed to appear in

HIF1-KD cells reprogrammed with the 4F retroviral transduc-tion (Fig. 2B), while SCR-KD cells retained the ability to giverise to NANOG-positive hESC-like colonies with an efficiencyapproximately similar to that of wild-type BJ fibroblasts(around 0.001%) (Fig. 2B). The addition of EDHB was not suffi-cient to overcome the reprogramming block in HIF1-KD cells(data not shown). In addition, we assessed the effect ofHIF1a depletion using an alternative nonintegrative episomalplasmid reprogramming approach [43]. Consistent with the

retroviral data, wild-type BJ and SCR-KD BJ fibroblasts trans-fected with the episomal plasmids could generate NANOG-positive colonies with a similar efficiency (around 0.0008%)after 4 weeks of reprogramming (Fig. 2B). However, HIF1-KDBJ fibroblasts failed to derive hESC-like colonies (Fig. 2B).These results imply that cells depleted of HIF1a may be inca-pable of achieving a pluripotent state regardless of thereprogramming method, pointing toward an instrumental roleof HIF1a in the induction of pluripotency.

We then sought to gain insights into the possible mecha-nisms responsible for the inability of HIF1-KD cells to undergoefficient reprogramming and performed global transcriptom-ics. Reassuringly, the results confirmed HIF1A as the mostdownregulated gene in HIF1-KD BJ compared to SCR-KD BJ(Supporting Information Table 1). Interestingly, pathway analy-sis revealed that among the most significantly downregulatedpathways in HIF1-KD compared to SCR-KD (fold change >1.5)there were the mammalian target of rapamycin (mTOR) sig-naling pathway, which is known to be associated with HIF1aand the regulation of energy metabolism [4], and the path-ways related to glycolysis and gluconeogenesis (Fig. 2C). Wethen assessed the bioenergetic profiling of fibroblasts andfound that the lack of HIF1a did not alter their basal glucosemetabolism, as indicated by the maintenance of OCR/ECARratio (Fig. 2D). Indeed, the rate of OXPHOS (Supporting

Figure 1. Mimicking HIF1a activation facilitates iPSC reprogramming. (A): HIF1a and HIF2a nuclear accumulation in FFs, hESCs, andFFiPSCs under normoxic conditions (N) and following 24 hours hypoxic incubation with 1% oxygen (H). The transcription factor YY1 wasused for normalization of nuclear extracts. (B): BJ fibroblasts were transduced with the four factor cocktail (4F) alone or in combinationwith daily treatment with 100 mM EDHB (4F EDHB). All cells were plated under reprogramming conditions, fixed 4 weeks later, andimmunostained against the pluripotency-associated protein NANOG, according to the Avidin-Biotin Complex (ABC) protocol. The experi-ments were repeated three times. Bar graphs represent the mean and SD of the average number of NANOG-positive hESC-like coloniesdetected. **, p5 .0063, two-tailed unpaired Student’s t test, 4F EDHB versus 4F. (C): OCR, indicative of OXPHOS activity, was assessedusing the Seahorse cellular flux analysis. Wild-type FFs (BJ cells) (black line) were compared to FFs treated with EDHB for 24 hours (yel-low line), 48 hours (orange line), and 72 hours (red line). (D): ECAR, indicative of glycolytic activity, was measured at the same time asOCR in the same samples. (E): OCR/ECAR ratio was calculated in order to generate a clear estimate of the overall metabolic state of thecells. ***, p< .005, two-tailed unpaired Student’s t test: EDHB 24h versus BJ (p5 .0043), EDHB 48h versus BJ (p5 .0025), and EDHB72h versus BJ (p5 .0032). Abbreviations: EDHB, ethyl 3,4-dihydroxybenzoate; ECAR, extracellular acidification rate; FFs, foreskin fibro-blasts; FFiPSCs, FF-derived iPSCs; H, hypoxic condition; hESCs, human embryonic stem cells; HIF1a, hypoxia inducible factor 1a; iPSC,induced pluripotent stem cells; N, normoxic condition; OCR, oxygen consumption rate.

368 HIF1a Associated Metabolic Switch in iPSCs

VC AlphaMed Press 2013 STEM CELLS

Information Fig. 2B) and glycolysis (Supporting Information Fig.2B) appeared similar in HIF1-KD fibroblasts, SCR-KD, and wild-type BJ fibroblasts. Overall, the data suggest that HIF1a abla-tion may not be sufficient to alter the basal metabolism offibroblasts but it might hinder reprogramming through thedownregulation of target genes that have to be activated inorder to enable the establishment of pluripotency. Since genesassociated with glycolysis and gluconeogenesis have been pre-viously found to be upregulated in undifferentiated PSCs com-pared to fibroblasts [25, 31], it may be conceivable that cellsthat are not capable of correctly activating a glycolytic pro-gram may also be refractory to efficient iPSC conversion.

Upregulation of HIF1a-Associated Metabolic RegulatorsDuring Initiation of Reprogramming

To follow-up the hypothesis that knockdown of HIF1a may alterthe transcriptional reconfiguration of energy metabolism andthat this may be crucial for reprogramming, we sought to focuson regulated gene expression occurring within the initiationphase of iPSC derivation. Indeed, we previously observed thatthe gene ontology biological process of “response to hypoxia”was significantly regulated during early reprogramming [40].We thus analyzed the transcriptome of FFs at 24 hours, 48hours, and 72 hours after (a) treatment with 100 mM EDHB(EDHB 24h, 28h, 72h), (b) transduction with the four reprog-ramming factors (4F 24h, 48h, 72h), and (c) both (4F EDHB24h, 48h, 72h). These data were compared to the transcrip-tome of fully reprogrammed iPSCs derived from FFs [21, 31],from adult DFs [38], and from AFCs [44], and of hESCs.

Principal component analysis revealed that at these earlytime points neither HIF1a manipulation nor 4F retroviraltransduction was sufficient to extensively alter the globaltranscriptional signature of somatic cells, which clusteredtogether and apart from all PSCs (Fig. 3A). The expressionlevel of the most significantly highly upregulated or downre-gulated genes (fold change >20) in PSCs compared to somaticcells remained unaffected in somatic fibroblasts exposed to 4Ftransduction and/or HIF1a activation (Fig. 3B). In particular,among these top-regulated genes, none appeared regulated inEDHB-treated fibroblasts (Supporting Information Table 2) andonly a few genes related to the viral-mediated introduction oftranscription factors (such as OCT4, the OCT4 pseudogene-1,and the cancer-associated H19) were upregulated in early 4F-transduced fibroblasts (Supporting Information Table 2).

We next focused on genes associated with HIF1a, energymetabolism, and mTOR signaling pathways (Supporting Infor-mation Table 3). The mRNA expression level of HIF1A andHIF2A genes was not altered upon early reprogramming orEDHB treatment (Supporting Information Table 3), as also con-firmed by quantitative real-time PCR (qPCR) analyses (Sup-porting Information Fig. 3A). Accordingly, no changes in HIF1Aor HIF2A gene expression could be observed in PSCs com-pared to somatic cells, both under normoxic and hypoxicexposure (Supporting Information Fig. 3B). This is in agree-ment with previous findings showing the lack of HIF1A tran-scriptional activation under hypoxia in hESCs [34].Interestingly, however, the expression of some of the genesassociated with energy metabolism underwent modifications

Figure 2. HIF1a knockdown inhibits reprogramming. (A): Immunoblot analysis confirmed that HIF1-KD BJ fibroblasts were incapable ofaccumulating HIF1a protein in the nucleus upon hypoxic stimulation while SCR-KD BJ fibroblasts retained this ability. (B): Wild-type BJfibroblasts, SCR-KD BJ fibroblasts, and HIF1-KD BJ fibroblasts were reprogrammed to pluripotency using either a classical retroviralapproach with the four Yamanaka factors (4F: OCT4, SOX2, KLF4, and c-MYC) or with episomal plasmids (expressing the four factors plusNANOG, LIN28, and SV40L). After 4 weeks of culturing under human embryonic stem cell (hESC) conditions, wild-type BJ and SCR-KD BJcells developed hESC-like colonies in a comparable fashion, as shown by the similar number of colonies that resulted positive for thepluripotency-associated marker NANOG (monitored with the Avidin-Biotin Complex method). However, NANOG-positive hESC-like colo-nies were not generated in HIF1-KD BJ cells, regardless of the reprogramming method used. (C): List of the most significantly downregu-lated pathways (fold change >1.5) in HIF1-KD BJ compared to SCR-KD BJ. (D): The ratio of OCR/ECAR, indicating the metabolic cellstate, was calculated in fibroblasts maintained under basal conditions using the seahorse bioanalyzer. Abbreviations: ECM, extracellularmatrix; ECAR, extracellular acidification rate; HIF1a, hypoxia inducible factor 1a; HIF1-KD, HIF-1a knockdown; H, hypoxic condition; N,normoxic condition; mTOR, mammalian target of rapamycin; OCR, oxygen consumption rate; SCR-KD, scrambled knockdown.

Prigione, Rohwer, Hoffman et al. 369

www.StemCells.com VC AlphaMed Press 2013

during early reprogramming (Supporting Information Table 3,Supporting Information Fig. 4). Among the metabolism-relatedgenes exhibiting early upregulation in 4F-transduced cells orEDHB-treated cells, we identified three HIF1a target factorsknown to regulate a reconfiguration of energy flux: PKM2,PDK1, and PDK3 (Supporting Information Fig. 4).

PKM2, PDK1, and PDK3 gene expression in fibroblastsshowed an increase following EDHB treatment, as confirmedby qPCR analysis (Fig. 4A). This is in agreement with previousdata linking HIF1a activation with PDK1–3 induction [15–17]and with PKM2 gene transcription [18]. Importantly, the soleintroduction of the four factors into fibroblasts resulted in theearly upregulation of PDK1 (Fig. 4A). On the other hand,PKM2 and PDK3 were not upregulated following 4F transduc-tion (Fig. 4A), indicating that the conventional Yamanaka pro-tocol may not be sufficient to induce the early expression ofthese two factors which instead require the additional activa-tion of the HIF1a pathway.

Finally, we confirmed that elevated expression of thesethree glycolytic regulators could occur in human fibroblasts fol-lowing both EDHB treatment (Fig. 4B) and hypoxic exposure(Fig. 4C). Importantly, PKM2, PDK1, and PDK3 were not upregu-lated in HIF1-KD fibroblasts treated with EDHB (Fig. 4B, one-way ANOVA, p< .005) or cultured under hypoxia (Fig. 4C, one-way ANOVA, p< .005). No alteration of the basal expression ofgenes associated with the pathways of HIF1a, energy metabo-lism, and mTOR could be detected in HIF1-KD compared toSCR-KD (with the exception of the downregulation of HIF1a)(Supporting Information Table 3). Hence, although fibroblastsbearing HIF1a knockdown showed a normal fibroblast-like basalmetabolism (Fig. 2D) and normal fibroblast-like metabolic-

related transcriptional signature (Supporting Information Table3), they appeared incapable of upregulating the expression ofkey glycolytic inducers (PDK1, PDK3, and PKM2). This implicatesthe Warburg-like transcriptional modulation of energy metabo-lism as a potential enabling step for initiating cellularreprogramming.

Elevated Expression of the HIF1a-Related GlycolyticRegulators PKM2, PDK1, and PDK3 in PSCs

We next investigated the association of the three HIF1a-related metabolic regulators with pluripotency. The expressionlevel of PDK1, PDK3, and PKM2 was measured in FFiPSCs,DFiPSCs and hESCs, under both normoxia and 24 hours 1%hypoxia, and compared it to that in FFs (for FFiPSCs), DFs (forDFiPSCs), and all somatic fibroblasts (for hESCs) grown undernormoxic conditions. In agreement with previous proteinexpression data [25, 31], the transcriptional level of PDK1appeared upregulated in all PSCs (Fig. 5A). However, PDK1induction was statistically significant in iPSCs only underhypoxic conditions (Fig. 5A, Student’s t test, hypoxic iPSCs vs.normoxic fibroblasts, p< .05), while both normoxic and hypoxichESCs exhibited significant PDK1 upregulation (Fig. 5A, Stu-dent’s t test, normoxic/hypoxic hESCs vs. normoxic fibroblastsp< .005). PDK3 was significantly upregulated in both iPSCs(Fig. 5B, Student’s t test, normoxic/hypoxic iPSCs vs. normoxicfibroblasts p< .05) and hESCs (Fig. 5B, Student’s t test, nor-moxic/hypoxic hESCs vs. normoxic fibroblasts p< .005) undernormoxia and hypoxia, although it was more elevated in iPSCsunder hypoxic growth (Fig. 5B, Student’s t test, hypoxic iPSCsvs. normoxic fibroblasts p< .005). Hence, the exposure tohypoxia led to increased upregulation of PDK1 and PDK3 in all

Figure 3. Transcriptional modulation during reprogramming initiation. (A): Principal component analysis (PCA) showing the clustering ofthe transcriptomes of the following somatic cells: FFs, DFs, AFCs, FFs knocked-down for HIF1a (HIF-KD) and knocked-down for ascrambled transcript (SCR-KD), FFs transduced with the four factors only for 24, 48, and 72 hours (4F 24h, 4F 48h, 4F 72h) or in combi-nation with daily 100 mM EDHB (4F EDHB 24h, 4F EDHB 48h, 4F EDHB 72h), and FFs only treated with EDHB (EDHB 24h, EDHB 48h,EDHB 72h), and the following pluripotent stem cells: hESCs, FFiPSCs, DFiPSCs, and AFiPSCs. (B): Heatmap depicting the genes mosthighly downregulated and upregulated (fold change >20) in somatic-derived and embryonic-derived pluripotent stem cells compared towild-type untreated somatic cells. Different iPSCs were compared to their respective somatic cells, while hESCs were compared to theaverage of all wild-type somatic cells. The samples include wild-type untreated somatic cells (gray bar), FFs harvested 24 hours, 48hours, and 72 hours after 4F transduction (green bar), FFs treated with EDHB for the same time points (yellow bar), FFs exposed toboth 4F and EDHB treatment (purple bar), iPSCs (black bar), and hESCs (red bar). Values indicate row-normalized log2 average expres-sion values; downregulated genes are indicated in green, upregulated genes in red. Abbreviations: AFCs, amniotic fluid cells; AFiPSCs,AFC-derived iPSCs; DFs, dermal fibroblasts; DFiPSCs, DF-derived iPSCs; EDHB, ethyl 3,4-dihydroxybenzoate; FFs, foreskin fibroblasts;FFiPSCs, FF-derived fibroblasts; hESCs, human embryonic stem cells; HIF1a, hypoxia inducible factor 1a; iPSCs, induced pluripotent stemcells; PC, principal component; SCR-KD, scrambled knockdown.

370 HIF1a Associated Metabolic Switch in iPSCs

VC AlphaMed Press 2013 STEM CELLS

iPSCs but not in hESCs, which showed elevated expression of thetwo genes already under normoxic growth (Fig. 5A, 5B). This dif-ferential response to hypoxic stimuli between somatic-derivedand embryonic-derived PSCs may be explained by recent evi-dence demonstrating that they may not be equivalent in termsof their pluripotency [49–51]. Alternatively, it may be simply dueto the known heterogeneity displayed by different PSC lines [52].In any case, further studies are warranted to dissect this differen-tial response to hypoxic stimuli between iPSCs and hESCs.

In order to dissect the role of PKM2 in PSCs, weemployed mass spectrometry-based targeted proteomic analy-sis. This approach allows the quantification of the absoluteamount of the two alternatively spliced isoforms encoded bythe PKM2 gene, PKM1 and PKM2, and it was previously car-ried out in healthy and tumor-associated tissues [8]. Remark-ably, in comparison to fibroblasts, both iPSCs and hESCsexhibited reduced PKM1 (Fig. 5C, Student’s t test, PSCs vs. FFsp< .05) coupled with significantly elevated amounts of PKM2(Fig. 5D, Student’s t test, PSCs vs. FFs p< .005). Overall, thePKM2/PKM1 ratio appeared significantly increased in PSCsunder both normoxic and hypoxic conditions (Fig. 5E, Stu-dent’s t test, normoxic/hypoxic PSCs vs. normoxic/hypoxic FFsp< .01). The PKM2 isoform is known to display reduced enzy-matic activity [12]. Thus, we next measured the PK functional-ity in fibroblasts and PSCs and identified a diminished PKactivity in all PSCs (Fig. 5F, Student’s t test, PSCs vs. wild-typeBJ, p< .005). This implies that the elevated expression ofPKM2 in PSCs may be functionally relevant and could contrib-ute to the maintenance of the glycolytic phenotype of PSCs.

Taken together, undifferentiated iPSCs and hESCs expresshigh levels of three HIF1a downstream targets, whose functionmediates a Warburg-like effect by increasing the energy flux inthe upstream glycolytic branches and PPP, eventually leading tobiomass stimulation and redox maintenance [6, 7]. Furthermore,since reprogramming efficiency is increased upon EDHB treat-ment (Fig. 1B) or hypoxic stimulation [36], it is tempting tospeculate that the elevated expression of PKM2 and PDK3detected in fibroblasts exposed to EDHB or hypoxia (Fig. 4B, 4C)may possibly play a role in improving the conversion to iPSCsby additional metabolic modulation toward a glycolytic state.

Metabolic Remodeling During Initiation ofReprogramming

Finally, we asked whether the transcriptional activation of meta-bolic regulators during reprogramming initiation was sufficientto induce a metabolic reconfiguration. Bioenergetic profiling offibroblasts following 4F transduction indicated that mitochon-drial respiration increased over the first 3 days of reprogram-ming (Fig. 6A). However, the elevation of glycolysis appearedmore pronounced (Fig. 6B). The additional treatment with EDHBresulted in enhanced glycolytic conversion, with drastic OCRreduction (Fig. 6C) and higher ECAR values (Fig. 6D). Overall, 4Ftransduction led to progressive decrease in the OCR/ECAR ratio,indicative of initial conversion to glycolytic metabolism (Fig. 6E,Student’s t test, 4F-transduced BJ vs. wild-type BJ, p< .05). Inaccordance, the amount of extracellular lactate in 4F-transducedfibroblasts showed a similar gradual increase during the firstdays of reprogramming (Fig. 6F, Student’s t test, 4F-transduced

Figure 4. Upregulation of HIF1a-related metabolic regulators during early reprogramming. (A): Quantitative real-time polymerase chainreaction analysis of HIF1a targets known to play a role in regulating glycolytic metabolism (PKM2, PDK1, and PDK3) during the first 3days of reprogramming initiation. Relative mRNA level to ACTB is presented in comparison to wild-type untreated FFs. Green line: FFstransduced with the four factors; yellow line: FFs treated with EDHB; purple line: FFs transduced with the 4F and treated with EDHB atthe same time. (B): Expression of PKM2, PDK1, and PDK3 after 24 hours of EDHB treatment in wild-type FFs (BJ and HFF1), SCR-KDfibroblasts, and HIF1-KD fibroblasts. ***, p< .005, one-way ANOVA single factor: PKM2 (p5 5E2 07), PDK1 (p5 9E2 06), and PDK3(p5 1 E2 07). (C): Transcriptional level of the three glycolytic regulators after 24 hours of exposure to 1% hypoxia. ***, p< .005, one-way ANOVA single factor: PKM2 (p5 4E2 05), PDK1 (p5 9E2 08), and PDK3 (p5 1E2 07). Abbreviations: EDHB, ethyl 3,4-dihydroxy-benzoate; FFs, foreskin fibroblasts; HIF1a, hypoxia inducible growth factor; HIF1-KD, HIF-1a knockdown; PKM2, pyruvate kinase isoformM2; PDK1, pyruvate dehydrogenase kinase 1; PDK3, pyruvate dehydrogenase kinase 3; SCR-KD, scrambled knockdown.

Prigione, Rohwer, Hoffman et al. 371

www.StemCells.com VC AlphaMed Press 2013

BJ vs. wild-type BJ, p< .05). Nonetheless, this metabolic remod-eling does not appear to be completed at these early stages, asshown by the much higher lactate secretion occurring in fullyreprogrammed iPSCs and hESCs (Fig. 6F, Student’s t test, PSCsvs. BJ, p< .005). The additional exposure to a drug mimickingHIF1a activation may thus improve the efficiency of iPSC gener-ation through the early enhancement of glycolytic activation, asshown by a dramatic OCR/ECAR decrease (Fig. 6E, Student’s ttest, 4F-EDHB treated BJ vs. BJ, p< .005).

On the basis of data presented here, we suggest that theintroduction of the four Yamanaka factors in human fibro-blasts may be sufficient to upregulate the HIF1a target PDK1,which can in turn initiate a glycolytic program. On the otherhand, the inclusion of hypoxia or EDHB treatment may stimu-

late the expression of additional HIF1a-related metabolic regu-lators PDK3 and PKM2, enhancing the glycolytic shift at thevery early stages of reprogramming and eventually leading toimproved iPSC derivation. Finally, reprogramming may beinhibited in cells that are incapable of activating HIF1a andupregulating these three metabolic regulators, further under-lying the importance of HIF1a-associated metabolic restructur-ing in the induction of pluripotency in somatic cells (Fig. 7).

DISCUSSION

We and others previously demonstrated that cell fate reprog-ramming is associated with a transition from respiratory toglycolytic metabolism [21–26]. Tumor cells are believed to

Figure 5. PKM2, PDK1, and PDK3 are highly expressed in pluripotent stem cells (PSCs). (A): Relative PDK1 mRNA expression in: FFiPSCs(iPS2, iPS4, iB4, and iB5) under normoxia and hypoxia in relation to normoxic FFs (HFF1 and BJ); DFiPSCs (OiPS6, OiPS8, and OiPS16)under normoxia and hypoxia in relation to normoxic DFs (NFH2); and hESCs (H1 and H9) under normoxia and hypoxia in relation to allfibroblasts (HFF1, BJ, and NFH2) grown under normoxic conditions. **, p5 .0098, two-tailed unpaired Student’s t test, hypoxic FFiPSCsversus normoxic FFs. *, p5 .043, two-tailed unpaired Student’s t test, hypoxic DFiPSCs versus normoxic DFs. ***, p< .005, two-tailedunpaired Student’s t test: normoxic hESCs versus all normoxic fibroblasts (p5 .0005), hypoxic hESCs versus all normoxic fibroblasts(p5 .0011). (B): Relative expression of PDK3 in PSCs under normoxia and hypoxia compared to normoxic somatic fibroblasts. *,p5 .023, two-tailed unpaired Student’s t test, normoxic FFiPSCs versus normoxic FFs. **, p5 .0061, two-tailed unpaired Student’s t test,normoxic DFiPSCs versus normoxic DFs. ***, p< .005, two-tailed unpaired Student’s t test: normoxic hESCs versus all normoxic fibro-blasts (p5 .0007), hypoxic FFiPSCs versus normoxic FFs (p5 .0009), hypoxic DFiPSCs versus normoxic DFs (p5 .0021), hypoxic hESCs ver-sus all normoxic fibroblasts (p5 .0002) (C): Absolute protein quantification of PKM1 in FFs, FFiPSCs, and hESCs. *, p< .05, two-tailedunpaired Student’s t test, FFiPSCs versus FFs and hESCs versus FFs. (D): Absolute protein quantification of PKM2 in FFs (HFF1 and BJ),FFiPSCs (iPS2, iPS4, iB4, and iB5), and hESCs (H1 and H9). ***, p< .005, two-tailed unpaired Student’s t test, FFiPSCs versus FFs andhESCs versus FFs. (E): PKM2/PKM1 ratio in FFs, FFiPSCs, and hESCs grown under normoxic and hypoxic conditions. **p< .01, two-tailedunpaired Student’s t test, normoxic hESCs versus normoxic FFs. ***, p< .005, two-tailed unpaired Student’s t test, normoxic FFiPSCs ver-sus normoxic FFs, hypoxic FFiPSCs versus hypoxic FFs, and hypoxic hESCs versus hypoxic FFs. (F): Rate of PK activity, normalized overthe total protein amount, was measured in wild-type FFs (BJ and HFF1), SCR-KD, and HIF1-KD fibroblasts, FFiPSCs (iB4, and iB5), andhESCs (H1 and H9). ***, p< .005, two-tailed unpaired Student’s t test, iB4 versus BJ, iB5 versus BJ, H1 versus BJ, and H9 versus BJ.Abbreviations: DFs, dermal fibroblasts; DFiPSCs, DF-derived iPSCs; FFs, foreskin fibroblasts; FFiPSCs, FF-derived iPSCs; hESCs, humanembryonic stem cells; HIF-1a, hypoxia inducible factor 1a; HIF1-KD, HIF-1a knockdown; PDK, pyruvate dehydrogenase kinase; PKM2,pyruvate kinase isoform M2; SCR-KD, scrambled knockdown.

372 HIF1a Associated Metabolic Switch in iPSCs

VC AlphaMed Press 2013 STEM CELLS

undergo similar metabolic transformation events in order tosustain the cost of anabolic reactions caused by their highproliferative rate and resistance to stress signals [4]. Thus,metabolic adaptation appears as a crucial mechanism for pro-liferating cells, which require to build biomass and the sametime preserve the redox balance [6]. Here, we sought touncover possible mechanistic pathways underlying the meta-bolic reprogramming of iPSCs and demonstrated that HIF1a, amaster regulator of glucose metabolism [13, 14], plays a criti-cal role during the induction of pluripotency by modulatingthe early establishment of a glycolytic program.

Hypoxia and HIF1a have been previously implicated in themaintenance of pluripotency. Indeed, hypoxic conditions havebeen found to lead to improved stemness and reprogramming[33, 36]. HIF1a activation could induce hESC-like signature incancer cell lines [48] and drive mouse ESCs to acquire glycolyticfeatures upon transition toward an epiblast stem cell-like state[53]. Moreover, stabilization of hypoxia and HIF1a inhibit themetabolic shift to OXPHOS that is required for efficient differen-tiation of human mesenchymal stem cells [54]. Hypoxic exposure

may also increase the protein expression uncoupling protein two(UCP2) [55], a regulator of energy metabolism in undifferenti-ated PSCs [26]. Finally, hypoxia might lead to OCT4 reactivationto promote the dedifferentiation of PSC-derived progenies [56].Our results support all these findings and demonstrate that cellsincapable of activating a HIF1a response might also be refractoryto reprogramming, thus underlying the critical importance ofthis pathway for the induction and maintenance of pluripotency.

Our data suggest that a glycolytic shift may be instrumentalfor reprogramming as it may be set into motion during thereprogramming initiation stage through the early upregulationof the HIF1a target PDK1. Indeed, PSCs exhibit elevated PDK1protein expression [25, 31]. Moreover, small molecule-basedactivation of PDK1 can improve iPSC derivation [22], while PDK1inhibitors lead to reduced hESC-like colony formation [23].Hence, PDK1 may possibly represent an early marker of reprog-ramming involved in the Warburg-like metabolic restructuringassociated with the conversion to pluripotency.

The findings also provide additional support to previousdata showing that during reprogramming, glycolysis-associated

Figure 6. Metabolic shift during early reprogramming. (A): OCR profile of wild-type FFs (BJ) and FFs after 24 hours, 48 hours, and 72hours from 4F transduction. (B): ECAR profile of FFs at the basal level and after 4F introduction. (C): OCR profile in BJ fibroblasts treatedwith the HIF1a mimicker EDHB in addition to the 4F transduction. (D): ECAR profile in FFs and 4F-EDHB FFs. (E): OCR/ECAR ratio gradu-ally decreases during reprogramming initiation while it is drastically reduced upon the additional introduction of EDHB to the reprog-ramming cocktail. *, p5 .0370, two-tailed unpaired Student’s t test, 4F 48h versus BJ. **, p5 .0062, two-tailed unpaired Student’s ttest, 4F 72h versus BJ. ***, p< .005, two-tailed unpaired Student’s t test: 4F EDHB 24h versus BJ (p5 .0045), 4F EDHB 48h versus BJ(p5 .0001), and EDHB 72h versus BJ (p5 .0002). (F): Production of extracellular lactate in FFs (BJ, SCR-KD, HIF1-KD), FFs transducedwith the 4F, hESCs (H1 and H9), and BJ-FFiPSCs (iB4 and iB5). The values are reported to the amount of lactate generated in controlwild-type BJ fibroblasts. *, p< .05, two, two-tailed unpaired Student’s t test: 4F 24h versus BJ (p5 .035), 4F 48h versus BJ (p5 .021).**, p< .01, two-tailed unpaired Student’s t test, 4F 72h versus BJ (p5 .008). ***, p< .005, two-tailed unpaired Student’s t test: hESCsversus BJ (p5 .0035), BJ FFiPSCs versus BJ (p5 .0041). Abbreviations: ECAR, extracellular acidification rate; EDBH, ethyl 3,4-dihydroxy-benzoate; FFiPSCs, foreskin fibroblast-derived induced pluripotent stem cells; hESCs, human embryonic stem cells; HIF-1a, hypoxia induc-ible factor 1a; HIF1-KD, HIF-1a knockdown; OCR, oxygen consumption rate; SCR-KD, scrambled knockdown.

Prigione, Rohwer, Hoffman et al. 373

www.StemCells.com VC AlphaMed Press 2013

genes may be upregulated prior to genes involved in self-renewal and pluripotency [23]. Furthermore, transcriptionalchanges in the processes related to cellular metabolism havebeen detected during the first initial wave of reprogramming,which has been suggested to be under the control of KLF4 andc-MYC [57]. It is thus tempting to speculate that KLF4 and c-MYC may be driving the transcriptional and bioenergetic modu-lation observed during the initiation of cellular reprogramming.Indeed, the oncogene c-MYC can co-operate with HIF1a toinduce a transcriptional program leading to stimulated glycolyticactivity [58, 59], although a glycolytic shift has been observedeven in iPSCs derived in the absence of c-MYC [23]. KLF4 maybe capable of activating glycolytic metabolism in cancer cells[60], and KLF5, which can substitute KLF4 in the reprogrammingcocktail [61], can regulate energy metabolism and upregulateUCP2 [62], a protein recently linked to the glycolytic state ofPSCs [26]. Nonetheless, it may as well be that the key stemnessfactor OCT4 could regulate downstream targets implicated inOXPHOS and glycolysis [28, 63]. In fact, OCT4 expression inter-mingles with HIF signaling pathways [47] and may positivelyinteract with PKM2 [64]. Further studies are warranted toclearly dissect the specific roles of the Yamanaka factors in theremodeling of energy metabolism of reprogrammed cells.

Finally, since reconfiguration of glucose metabolism couldhave significant advantages in preventing redox imbalance [5,65], it is conceivable that iPSC generation may require such met-abolic reprogramming in order to safeguard the genome integ-rity. Nuclear and mitochondrial genetic defects have beenreported in human iPSCs [31, 66, 67]. Perhaps, the suppressionof metabolic resetting may function as a reprogramming road-

block by inducing an uncontrolled rise of genomic damage whichmay be incompatible with continuous cell growth. In accordance,pro-oxidant reactions can promote PSC differentiation [68].

It is important to note that our analyses were performedon whole cultures and may thus not necessarily mirror the sit-uation occurring in the small percentage of cells actuallyundergoing reprogramming. Reassuringly, a recent work usingsingle-cell analysis showed that reprogramming is initiated inthe majority of virally transduced fibroblasts, although onlycompleted in a smaller cellular subpopulation [69]. Since wemainly focused on early reprogramming initiation, it is thenpossible that our data may reflect real reprogramming-relatedcellular conditions. Nevertheless, given the known heteroge-neity of single PSC lines [49, 52], it would be of interest torepeat our relatively small-scale investigation (which included2 hESC lines and 13 human iPSC lines derived from 3 differentcell sources, i.e., FFs, DFs, and AFCs) using larger datasets ofseveral human and murine PSCs to clearly demonstrate therole of HIF1a and glycolytic regulation in cell fate conversion.

CONCLUSION

Overall, our results indicate that HIF1a-mediated reconfigura-tion of glucose metabolism may represent an early enablingstep of cellular reprogramming, a barrier that has to be over-come in order to make somatic cells capable of sustaining theirnewly acquired proliferative and biosynthetic needs. We antici-pate that the study of metabolism in stem cells may unveil crit-ical mechanisms governing the induction of pluripotency,eventually elucidating the pathways responsible for allowingthis remarkable example of cellular plasticity.

ACKNOWLEDGMENTS

The authors would like to thank Beata Lukaszewska for help inthe PKM quantification and Claudia Vogelgesang and AydahSabah at the microarray facility. We acknowledge support fromthe Max Planck Society. A.P. acknowledges support from the FritzThyssen Foundation (grant AZ. 10.11.2.160). M.R. is a WellcomeTrust Research Career Development and Wellcome-Beit fellow.J.A. acknowledges support from the German Federal Ministry ofEducation and Research (BMBF) grants (01GN1005, 01GN0807),and 0315717A, which is a partner of the ERASysBio1 initiativesupported under the EU ERA-NET Plus scheme in FP7.

AUTHOR CONTRIBUTIONS

A.P.: conception and design, collection and assembly of data,data analysis and interpretation, financial support, and manu-script writing; N.R. and K.D.: collection and assembly of data;S.H., B.M., R.B., and K.B.: collection of data; E.E.W.: financialsupport; M.R. and T.C.: data analysis and interpretation andfinancial support; J.A.: conception and interpretation, financialsupport, and editing and final approval of the manuscript.

DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST

The authors indicate no potential conflicts of interest.

Figure 7. HIF1a-associated metabolic reconfiguration duringreprogramming initiation. Cartoon depicting the potential mecha-nisms through which the introduction of the 4F or the modula-tion of HIF1a pathway might regulate glycolytic metabolismduring the early stages of somatic cell reprogramming. The intro-duction of the 4F in somatic fibroblasts upregulates the HIF1a tar-get PDK1, which re-routes the energy flux outside themitochondria, thereby enhancing the glycolytic metabolism. HIF1aactivation upregulates additional HIF1a targets PKM2 and PDK3,further increasing the glycolytic shift and eventually resulting inimproved conversion to pluripotency. Abbreviations: EDHB, ethyl3,4-dihydroxybenzoate; HIF1a, hypoxia inducible factor 1a;OXPHOS, oxidative phosphorylation; PDK, pyruvate dehydrogenasekinase; PDH, pyruvate dehydrogenase; PEP, phosphoenolpyruvate;PKM2, pyruvate kinase M2.

374 HIF1a Associated Metabolic Switch in iPSCs

VC AlphaMed Press 2013 STEM CELLS

REFERENCES1 Semenza GL. Life with oxygen. Science2007;318:62–64.2 Warburg O, On the origin of cancer

cells. Science 1956;123:309–314.3 Koppenol WH, Bounds PL, Dang, CV.Otto Warburg’s contributions to current con-cepts of cancer metabolism. Nat Rev Cancer2011;11:325–337.4 Levine AJ, Puzio-Kuter AM. The control

of the metabolic switch in cancers by onco-genes and tumor suppressor genes. Science2010;330:1340–1344.5 Cairns RA, Harris IS, Mak TW. Regulation

of cancer cell metabolism. Nat Rev Cancer2011;11:85–95.6 Vander Heiden MG, Cantley LC,Thompson CB. Understanding the Warburgeffect: The metabolic requirements of cellproliferation. Science 2009;324:1029–1033.7 Ralser M, Wamelink MM, Kowald A

et al. Dynamic rerouting of the carbohydrateflux is key to counteracting oxidative stress. JBiol 2007;6:10.8 Bluemlein K, Gruning NM, FeichtingerRG et al. No evidence for a shift in pyruvatekinase PKM1 to PKM2 expression duringtumorigenesis. Oncotarget 2011;2:393–400.9 Christofk HR, Vander Heiden MG, HarrisMH et al. The M2 splice isoform of pyruvatekinase is important for cancer metabolismand tumour growth. Nature 2008;452:230–233.10 Anastasiou D, Poulogiannis G, Asara JMet al. Inhibition of pyruvate kinase M2 byreactive oxygen species contributes to cellu-lar antioxidant responses. Science 2011;334:1278–1283.11 Gruning NM, Rinnerthaler M, BluemleinK et al. Pyruvate kinase triggers a metabolicfeedback loop that controls redox metabo-lism in respiring cells. Cell Metab 2011;14:415–427.12 Chaneton B, Gottlieb E. Rocking cellmetabolism: Revised functions of the key gly-colytic regulator PKM2 in cancer. Trends Bio-chem Sci 2012;37:309–316.13 Semenza GL. Hypoxia-inducible factorsin physiology and medicine. Cell 2012;148:399–408.14 Rohwer N, Zasada C, Kempa S et al. Thegrowing complexity of HIF-1alpha’s role intumorigenesis: DNA repair and beyond.Oncogene 2013;32:3569–3576.15 Kim JW, Tchernyshyov I, Semenza GLet al. HIF-1-mediated expression of pyruvatedehydrogenase kinase: A metabolic switchrequired for cellular adaptation to hypoxia.Cell Metab 2006;3:177–185.16 Papandreou I, Cairns RA, Fontana Let al. HIF-1 mediates adaptation to hypoxiaby actively downregulating mitochondrialoxygen consumption. Cell Metab 2006;3:187–197.17 Lu CW, Lin SC, Chen KF et al. Inductionof pyruvate dehydrogenase kinase-3 byhypoxia-inducible factor-1 promotes meta-bolic switch and drug resistance. J Biol Chem2008;283:28106–28114.18 Luo W, Hu H, Chang R et al. Pyruvatekinase M2 is a PHD3-stimulated coactivatorfor hypoxia-inducible factor 1. Cell 2011;145:732–744.

19 Takahashi K, Yamanaka, S. Induction ofpluripotent stem cells from mouse embryonicand adult fibroblast cultures by defined fac-tors. Cell 2006;126:663–676.20 Ruiz S, Panopoulos AD, Herrerias A et al.A high proliferation rate is required for cellreprogramming and maintenance of humanembryonic stem cell identity. Curr Biol 2011;21:45–52.21 Prigione A, Fauler B, Lurz R et al. Thesenescence-related mitochondrial/oxidativestress pathway is repressed in humaninduced pluripotent stem cells. Stem Cells2010;28:721–33.22 Zhu S, Li W, Zhou H et al. Reprogram-ming of human primary somatic cells byOCT4 and chemical compounds. Cell StemCell 2010;7:651–655.23 Folmes CD, Nelson TJ, Martinez-Fernandez A et al. Somatic oxidative bioener-getics transitions into pluripotency-dependent glycolysis to facilitate nuclearreprogramming. Cell Metab 2011;14:264–271.24 Panopoulos AD, Yanes O, Ruiz S et al.The metabolome of induced pluripotentstem cells reveals metabolic changes occur-ring in somatic cell reprogramming. Cell Res2011;22:168–177.25 Varum S, Rodrigues AS, Moura MB et al.Energy metabolism in human pluripotentstem cells and their differentiated counter-parts. PLoS One 2011; 6:e20914.26 Zhang J, Khvorostov I, Hong JS et al.UCP2 regulates energy metabolism and dif-ferentiation potential of human pluripotentstem cells. EMBO J 2011;30:4860–4873.27 Zhang J, Nuebel E, Daley GQ et al. Met-abolic regulation in pluripotent stem cellsduring reprogramming and self-renewal. CellStem Cell 2012;11:589–595.28 Folmes CD, Dzeja PP, Nelson TJ et al.Metabolic plasticity in stem cell homeostasisand differentiation. Cell Stem Cell 2012;11:596–606.29 Xu X, Duan S, Yi F et al. Mitochondrialregulation in pluripotent stem cells. CellMetab 2013;18:325–332.30 Bukowiecki R, Adjaye J, Prigione A.Mitochondrial function in pluripotent stemcells and cellular reprogramming. Gerontol-ogy 2013 (in press).31 Prigione A, Lichtner B, Kuhl H et al.Human induced pluripotent stem cells harborhomoplasmic and heteroplasmic mitochon-drial DNA mutations while maintaininghuman embryonic stem cell-like metabolicreprogramming. Stem Cells 2011;29:1338–1348.32 Armstrong L, Tilgner K, Saretzki G et al.Human induced pluripotent stem cell linesshow stress defense mechanisms and mito-chondrial regulation similar to those ofhuman embryonic stem cells. Stem Cells2010;28:661–673.33 Ezashi T, Das P, Roberts RM et al. LowO2 tensions and the prevention of differen-tiation of hES cells. Proc Natl Acad Sci USA2005;102:4783–4788.34 Forristal CE, Wright KL, Hanley NA et al.Hypoxia inducible factors regulate pluripo-tency and proliferation in human embryonicstem cells cultured at reduced oxygen ten-sions. Reproduction 2010;139:85–97.

35 Mohyeldin A, Garzon-Muvdi T,Quinones-Hinojosa A. Oxygen in stem cellbiology: A critical component of the stemcell niche. Cell Stem Cell 2010;7:150–161.36 Yoshida Y, Takahashi K, Okita K et al.Hypoxia enhances the generation of inducedpluripotent stem cells. Cell Stem Cell 2009;5:237–41.37 Lengner CJ, Gimelbrant AA, Erwin JAet al. Derivation of pre-X inactivation humanembryonic stem cells under physiological oxy-gen concentrations. Cell 2010;141:872–883.38 Prigione A, Hossini AM, Lichtner B et al.Mitochondrial-associated cell death mecha-nisms are reset to an embryonic-like state inaged donor-derived iPS cells harboring chro-mosomal aberrations. PLoS One 2011;6:e27352.39 Warnecke C, Griethe W, Weidemann Aet al. Activation of the hypoxia-induciblefactor-pathway and stimulation of angiogene-sis by application of prolyl hydroxylase inhibi-tors. FASEB J 2003;17:1186–1188.40 Mah N, Wang Y, Liao MC et al. Molecu-lar insights into reprogramming-initiationevents mediated by the OSKM gene regula-tory network. PLoS One 2011;6:e24351.41 Sowter HM, Raval RR, Moore JW et al.Predominant role of hypoxia-inducible tran-scription factor (Hif)21alpha versus Hif-2alpha in regulation of the transcriptionalresponse to hypoxia. Cancer Res 2003;63:6130–6134.42 Leurs C, Jansen M, Pollok KE et al. Com-parison of three retroviral vector systems fortransduction of nonobese diabetic/severecombined immunodeficiency mice repopulat-ing human CD341 cord blood cells. HumGene Ther 2003;14:509–519.43 Yu J, Chau KF, Vodyanik MA et al. Effi-cient feeder-free episomal reprogrammingwith small molecules. PLoS One 2011;6:e17557.44 Wolfrum K, Wang Y, Prigione A et al.The large principle of cellular reprogram-ming: Lost, acquired and retained geneexpression in foreskin and amniotic fluid-derived human iPS cells. PLoS One 2010;5:e13703.45 Kawamura T, Suzuki J, Wang YV et al.Linking the p53 tumour suppressor pathwayto somatic cell reprogramming. Nature 2009;460:1140–1144.46 Rohwer N, Welzel M, Daskalow K et al.Hypoxia-inducible factor 1alpha mediatesanoikis resistance via suppression of alpha5integrin. Cancer Res 2008;68:10113–10120.47 Covello KL, Kehler J, Yu H et al. HIF-2alpha regulates Oct-4: Effects of hypoxia onstem cell function, embryonic development,and tumor growth. Genes Dev 2006;20: 557–570.48 Mathieu J, Zhang Z, Zhou W et al. HIFinduces human embryonic stem cell markersin cancer cells. Cancer Res 2011;71:4640–4652.49 Yamanaka S. Induced pluripotent stemcells: Past, present, and future. Cell Stem Cell2012;10:678–684.50 Ohi Y, Qin H, Hong C et al. IncompleteDNA methylation underlies a transcriptionalmemory of somatic cells in human iPS cells.Nat Cell Biol 2011;13:541–549.

Prigione, Rohwer, Hoffman et al. 375

www.StemCells.com VC AlphaMed Press 2013

51 Kim K, Zhao R, Doi A et al. Donor celltype can influence the epigenome and differ-entiation potential of human induced pluri-potent stem cells. Nat Biotechnol 2011;29:1117–1119.52 Cahan P, Daley, GQ. Origins and implica-tions of pluripotent stem cell variability andheterogeneity. Nat Rev Mol Cell Biol 2013;14:3573–68.53 Zhou W, Choi M, Margineantu D et al.HIF1alpha induced switch from bivalent toexclusively glycolytic metabolism during ESC-to-EpiSC/hESC transition. EMBO J 2012;31:2103–2116.54 Hsu SH, Chen CT, Wei YH. Inhibitoryeffects of hypoxia on metabolic switch andosteogenic differentiation of human mesen-chymal stem cells. Stem Cells 2013 Jun 4.doi: 10.1002/stem.1441. [Epub ahead ofprint].55 Deng S, Yang Y, Han Y et al. UCP2 inhib-its ROS-mediated apoptosis in A549 underhypoxic conditions. PLoS One 2012;7:e30714.56 Mathieu J, Zhang Z, Nelson A et al.Hypoxia induces re-entry of committed cellsinto pluripotency. Stem Cells 2013 [Epubahead of print].

57 Polo JM, Anderssen E, Walsh RM et al.A molecular roadmap of reprogrammingsomatic cells into iPS cells. Cell 2012;151:1617–1632.

58 Dang CV, Le A, Gao, P. MYC-inducedcancer cell energy metabolism and therapeu-tic opportunities. Clin Cancer Res 2009;15:6479–6483.

59 Shim H, Dolde C, Lewis BC et al. c-Myctransactivation of LDH-A: Implications fortumor metabolism and growth. Proc NatlAcad Sci USA 1997;94:6658–6563.

60 Moon JS, Kim HE, Koh E et al. Kruppel-like factor 4 (KLF4) activates the transcriptionof the gene for the platelet isoform of phos-phofructokinase (PFKP) in breast cancer. JBiol Chem 2011;286:23808–23816.

61 Nakagawa M, Koyanagi M, Tanabe Ket al. Generation of induced pluripotent stemcells without Myc from mouse and humanfibroblasts. Nat Biotechnol 2008;26:101–106.

62 Oishi Y, Manabe I, Tobe K et al. SUMOy-lation of Kruppel-like transcription factor 5acts as a molecular switch in transcriptionalprograms of lipid metabolism involving PPAR-delta. Nat Med 2008;14:656–666.

63 Chen X, Xu H, Yuan P et al. Integrationof external signaling pathways with the coretranscriptional network in embryonic stemcells. Cell 2008;133:1106–1117.64 Lee J, Kim HK, Han YM et al. Pyruvatekinase isozyme type M2 (PKM2) interactsand cooperates with Oct-4 in regulating tran-scription. Int J Biochem Cell Biol 2008;40:1043–1054.65 Gruning NM, Ralser M. Cancer: Sacrificefor survival. Nature 2011;480:190–191.66 Gore A, Li Z, Fung HL et al. Somatic cod-ing mutations in human induced pluripotentstem cells. Nature 2011;471:63–67.67 Hussein SM, Batada NN, Vuoristo Set al. Copy number variation and selectionduring reprogramming to pluripotency.Nature 2011;471: 58–62.68 Yanes O, Clark J, Wong DM et al. Meta-bolic oxidation regulates embryonic stem celldifferentiation. Nat Chem Biol 2010;6:411–417.69 Tanabe K, Nakamura M, Narita M et al.Maturation, not initiation, is the major road-block during reprogramming toward pluripo-tency from human fibroblasts. Proc NatlAcad Sci USA 2013;110:12172–12179.

See www.StemCells.com for supporting information available online.

376 HIF1a Associated Metabolic Switch in iPSCs

VC AlphaMed Press 2013 STEM CELLS