A Hypoxia-Induced Positive Feedback Loop Promotes Hypoxia ...

MOLECULAR AND CELLULAR BIOLOGY, July 2011, p. 2696–2706 0270-7306/11/$12.00 doi:10.1128/MCB.01242-10 Copyright © 2011, American Society for Microbiology. All Rights Reserved.

Vol. 31, No. 13

A Hypoxia-Induced Positive Feedback Loop Promotes Hypoxia-Inducible Factor 1␣ Stability through miR-210 Suppression of Glycerol-3-Phosphate Dehydrogenase 1-Like䌤† Timothy J. Kelly,1,2,3 Amanda L. Souza,4 Clary B. Clish,4 and Pere Puigserver1,2* Department of Cancer Biology, Dana-Farber Cancer Institute,1 and Department of Cell Biology, Harvard Medical School,2 Boston, Massachusetts 02115; Department of Cell Biology, Johns Hopkins University School of Medicine, Baltimore, Maryland 212053; and Metabolite Profiling Initiative, Broad Institute of MIT and Harvard, 7 Cambridge Center, Cambridge, Massachusetts 02142 4 Received 26 October 2010/Returned for modification 2 December 2010/Accepted 26 April 2011

Oxygen-dependent regulation of the transcription factor HIF-1␣ relies on a family of prolyl hydroxylases (PHDs) that hydroxylate hypoxia-inducible factor 1␣ (HIF-1␣) protein at two prolines during normal oxygen conditions, resulting in degradation by the proteasome. During low-oxygen conditions, these prolines are no longer hydroxylated and HIF-1␣ degradation is blocked. Hypoxia-induced miRNA-210 (miR-210) is a direct transcriptional target of HIF-1␣, but its complete role and targets during hypoxia are not well understood. Here, we identify the enzyme glycerol-3-phosphate dehydrogenase 1-like (GPD1L) as a novel regulator of HIF-1␣ stability and a direct target of miR-210. Expression of miR-210 results in stabilization of HIF-1␣ due to decreased levels of GPD1L resulting in an increase in HIF-1␣ target genes. Altering GPD1L levels by overexpression or knockdown results in a decrease or increase in HIF-1␣ stability, respectively. GPD1Lmediated decreases in HIF-1␣ stability can be reversed by pharmacological inhibition of the proteasome or PHD activity. When rescued from degradation by proteasome inhibition, elevated amounts of GPD1L cause hyperhydroxylation of HIF-1␣, suggesting increases in PHD activity. Importantly, expression of GPD1L attenuates the hypoxic response, preventing complete HIF-1␣ induction. We propose a model in which hypoxia-induced miR-210 represses GPD1L, contributing to suppression of PHD activity, and increases of HIF-1␣ protein levels. decreased binding by VHL and subsequent protein accumulation (reviewed in references 28 and 46). Hypoxia-independent HIF-1␣ accumulation can be induced by inhibition of the PHDs through iron chelation or 2-oxoglutarate analogue supplementation (16, 26). Hypoxia-dependent HIF-1␣ translocation to the nucleus (29) results in binding to HIF-1␤ (ARNT) and hypoxic response elements (HREs) in promoter regions to induce transcription. HIF-1␣ targets a wide variety of genes, including genes involved in energy metabolism, angiogenesis, cell proliferation, and survival, among others (reviewed in reference 48). Over the past several years, microRNAs (miRNAs) have been shown to regulate many cellular processes and are predicted to target up to 30% of the human genes (32) through interaction within the 3⬘ untranslated region (UTR) of mRNAs. MicroRNAs have been shown to regulate protein levels by multiple mechanisms, including RNA degradation (5) and repression of protein translation (13, 30, 51). Recently, a number of microRNAs induced during hypoxia have been identified. One of these microRNAs, miR-210, is strongly induced by HIF-1␣ (31) and has pleiotropic effects. Expression of miR-210 in human umbilical vein endothelial cells (HUVEC) results in increased tubulogenesis and increased vascular endothelial growth factor (VEGF)-induced cell migration through the repression of the receptor tyrosine kinase ligand Ephrin-A3 (18). In stromal cells, miR-210 increased osteoblastic differentiation (37), and repression of MNT by miR-210 induces MYC activity (56). Interestingly, miR-210 overexpression modulates mitochondrial oxygen consumption through decreases in iron-sul-

Oxygen sensing in mammals is a tightly regulated process, and the response to low oxygen conditions is mediated primarily by gene expression changes induced by the hypoxia-inducible factor (HIF) family of transcription factors. Among this family, HIF-1␣ is the best characterized. Under normoxic conditions, HIF-1␣ protein has a very short half-life (less than 5 min under posthypoxic conditions in cell culture), and decreases in oxygen concentration cause its stability to increase almost immediately (24, 27). This regulation of protein stability is largely due to hydroxylation of two proline residues (prolines 402 and 564) by the prolyl hydroxylase (PHD) family of enzymes, which require oxygen, iron, and 2-oxogluarate for activity (reviewed in reference 21). In fact, previous data have shown that mutation of these prolines to alanine results in stabilized HIF-1␣ at normoxia (35). HIF-1␣ hydroxylation is modulated by the protein OS-9, which acts as a scaffold between HIF-1␣ and the PHDs (3). Prolyl hydroxylation of HIF-1␣ at normal oxygen tension results in subsequent binding by the von Hippel-Lindau (VHL) protein (36), which is part of an ubiquitin E3 ligase complex, resulting in HIF-1␣ polyubiquitination and degradation by the proteasome. When oxygen levels drop, HIF-1␣ is no longer hydroxylated, resulting in * Corresponding author. Mailing address: Dana-Farber Cancer Institute, 44 Binney Street, CLSB11144, Boston, MA 02115. Phone: (617) 582-7977. Fax: (617) 632-4770. E-mail: pere_puigserver@dfci .harvard.edu. † Supplemental material for this article may be found at http://mcb .asm.org/. 䌤 Published ahead of print on 9 May 2011. 2696

VOL. 31, 2011

POSITIVE FEEDBACK LOOP PROMOTES HIF-1␣ STABILITY

2697

FIG. 1. miR-210 overexpression induces HIF-1␣ protein stability and transcription. (A) 293A cells were transfected with either empty vector or FLAG–HIF-1␣, followed by transfection of the indicated amounts of negative control (NC) or miR-210 pre-miR approximately 24 h later. FLAG– HIF-1␣ was immunoprecipitated (IP) using FLAG antibody linked to agarose beads approximately 24 h after pre-miR transfection as described in Materials and Methods. ␣, anti. (B) HeLa cells were transiently transfected with either 5 nM negative control (NC) pre-miRs or miR-210 pre-miRs and placed under 0.2% or 1.0% oxygen for 8 h at 48 h posttransfection. Detection of protein levels was performed with Western blot analysis. (C) HeLa cells were transfected with hypoxic response element (HRE) containing firefly luciferase plasmid (10) along with control Renilla luciferase containing plasmid and 5 nM control or miR-210 pre-miRs and placed in hypoxia for 8 h. Values represent means ⫾ standard errors of the means (SEM) of results from two experiments, each in triplicate. Statistical significance was determined using one-way analysis of variance (ANOVA), followed by Bonferroni’s multiple comparison posttest. *, P ⬍ 0.05 for comparison to NC1, and #, P ⬍ 0.05 for comparison to NC2 at 8 h of hypoxia. (D) HeLa cells were transiently transfected with either 5 nM negative control (NC) pre-miRs or miR-210 pre-miRs and placed under 0.2% oxygen for 8 h at 48 h posttransfection. RNA was analyzed by RT-PCR as described in Materials and Methods. Values represent means ⫾ SEM of results from four experiments, each in triplicate. Statistical significance was determined using one-way ANOVA followed by Bonferroni’s multiple comparison posttest. *, P ⬍ 0.05, and ***, P ⬍ 0.001 for comparison to NC1, and ##, P ⬍ 0.01 for comparison to NC2 at 8 h of hypoxia.

fur cluster stability and aconitase activity due to downregulation of the protein ISCU1/2 (11, 12, 19). Despite these studies defining the biological function of miR-210, identification of the key targets and its role during hypoxia is still unclear. Induction of glycolytic enzymes is an important part of the hypoxic response in order to counteract decreases in ATP production through oxidative metabolism. Many of the primary glycolytic enzymes are HIF-1␣ targets (reviewed in references 47 and 55). HIF-1␣ acts in a concerted manner to increase glycolytic generation of ATP by inducing transcription of glucose transporters (GLUT1 and GLUT3), lactate transporters (MCT-4), and glycolytic enzymes such as hexokinase (HK), aldolase (ALD), phosphoglycerate kinase (PGK), phosphofructokinase (PFK), and lactate dehydrogenase (LDH). Although there are significant data linking HIF-1␣ induction to principal glycolytic enzymes, it is unclear how other branches of glycolysis, such as the glycerol phosphate shuttle, are affected by hypoxia. The glycerol phosphate shuttle includes two enzymes, one cytosolic, and another mitochondrial, which together interconvert dihydroxyacetone phosphate (DHAP) and glycerol 3-phosphate (G3P), resulting in production of FADH2 in the mitochondria. The central enzyme involved in this shuttle, glycerol-3-phosphate dehydrogenase (GPD1), is cytosolic, and the crystal structure has been determined (40). Here, we have identified a new target of miR-210, glycerol-

3-phosphate dehydrogenase 1-like (GPD1L), which is directly regulated during hypoxia by a miR-210 site in the 3⬘ UTR. GPD1L, when partially purified by immunoprecipitation from cells, has glycerol-3-phosphate dehydrogenase enzymatic activity. Interestingly, wild-type (WT) GPD1L and catalytically inactive GPD1L mutants caused HIF-1␣ degradation and decreased HIF-1␣ transcriptional activity. Oppositely, knockdown of GPD1L by short hairpin RNA (shRNA) results in an increased accumulation of HIF-1␣ during hypoxia compared to the level for control shRNAs as well as increased HIF-1␣ transcriptional activity. HIF-1␣ degradation can be reversed by proteasome inhibition with MG132 or by prolyl hydroxylase inhibition with a 2-oxoglutarate analogue, dimethyloxalylglycine (DMOG). Inhibition of HIF-1␣ degradation by MG132 results in HIF-1␣ hyperhydroxylation when coexpressed with GPD1L. Furthermore, overexpression of miR-210 results in increased HIF-1␣ accumulation during hypoxia. Importantly, ectopic expression of GPD1L during hypoxia strongly attenuates induction of HIF-1␣ protein. We propose a model in which induction of miR-210 by HIF-1␣ results in decreased GPD1L protein, which in turn causes increased stability of HIF-1␣. We propose that the increased stability is due to a decrease in PHD activity, establishing a positive feedback loop which helps to increase HIF-1␣ levels during low-oxygen conditions.

2698

KELLY ET AL.

MOL. CELL. BIOL.

FIG. 2. GPD1L is downregulated by miR-210 through its 3⬘ UTR. (A) Diagram of the miR-210 binding site in the GPD1L 3⬘ UTR. In order to evaluate miR-210 binding, two 3⬘ UTR mutants were constructed. The “seed mutant” retained the 5⬘-most binding region of the miR-210 site; however, the seed region was mutated. The “full mutant” lost all apparent binding of miR-210. (B) Wild-type or “seed” or “full” mutant UTRs cloned 3⬘ of firefly luciferase were transfected along with control Renilla luciferase and increasing amounts of miR-210 pre-miR. Values represent means ⫾ SEM of results from two experiments, each in quadruplicate. Statistical significance was determined using one-way ANOVA with Dunnett’s multiple comparison posttest. ***, P ⬍ 0.001 for comparison to WT UTR “no-miR” control; #, P ⬍ 0.05, and ###, P ⬍ 0.001 for comparison to “seed mutant” UTR “no-miR” control; and , P ⬍ 0.01 for comparison to “full mutant” UTR “no-miR” control as determined by one-way ANOVA with Dunnett’s multiple comparison posttest.

MATERIALS AND METHODS Cell culture. Cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) (HyClone) with a high concentration of glucose, supplemented with 10% cosmic calf serum (HyClone) and penicillin-streptomycin solution (Mediatech). Cells were transfected with plasmids using PolyFect (Qiagen) overnight according to the manufacturer’s protocols. Pre-miRs (Ambion) were transfected using Lipofectamine 2000 (Invitrogen) according to the manufacturer’s protocol. Inhibitors. MG132 was obtained from Boston Biochemicals, and dimethyloxalylglycine (DMOG) was obtained from Frontier Scientific. Hypoxia. All hypoxia experiments were carried out for the indicated times under 1.0% or 0.2% oxygen using a Proox 110 gas regulator (BioSpherix). Plasmids and retrovirus/lentivirus. GPD1L and GPD1 were amplified by PCR from mouse cDNA and cloned into a pcDNA-FLAG, -hemagglutinin (HA), or -FLAG-HA (FLAGHA) backbone vector previously generated in the laboratory. Mutants of GPD1L were generated from this plasmid using turbo Cx Hotstart polymerase (Stratagene). The GPD1L 3⬘ UTR was PCR amplified from mouse gDNA so that the miR-210 binding site was located approximately in the middle of a 150-bp fragment. This was cloned 3⬘ of firefly luciferase that had previously been cloned into pcDNA 3. Mutant GPD1L UTRs were generated from this plasmid using Pfu turbo Cx Hotstart polymerase (Stratagene). The miR-210 overexpression vector was generated by amplifying a 400-bp fragment surrounding the miR-210 pre-miRNA from mouse gDNA (sequence available in the supplemental material). This fragment was cloned into the pSuper-retro-puro plasmid, which was cotransfected with pCL-Eco (38) into Phoenix cells to generate retrovirus. shRNA plasmids were generated using pLK0-puro (49), and lentivirus was produced by transfection into HEK 293T cells. shRNA sequence information is available in the supplemental material. Hypoxic response element (HRE) firefly luciferase was obtained as a gift from Lorenz Poellinger. Gene expression analysis. mRNA expression levels were analyzed by quantitative real-time PCR (qRT-PCR). Total RNA was prepared from HeLa cells by Trizol extraction (Invitrogen). cDNA was generated by a high-capacity cDNA reverse transcription kit (Applied Biosciences) and analyzed by qRT-PCR using Power Sybr green master mix (Applied Biosystems). All data were normalized to 36B4 levels. The oligonucleotide primers can be provided upon request. Cell-based luciferase reporter assays. Cells were transfected with plasmids encoding HRE-firefly luciferase (10) and cytomegalovirus (CMV)-Renilla using PolyFect. For luciferase assays containing microRNA overexpression, HRE and Renilla plasmids were transfected 24 h before pre-miR transfection. At 48 h after pre-miR transfection, cells were placed in 0.2% oxygen. For experiments involving overexpression of GPD1L, cells were transfected overnight and placed in 0.2% oxygen. For experiments involving shRNA-mediated knockdown, plasmids were transfected overnight and placed in 0.2% oxygen 72 h later. Luciferase activities were detected using the dual-luciferase reporter assay system from Promega and a FLUOstar Omega plate reader (BMG Labtech). All luciferase

values are represented as firefly luciferase counts relative to Renilla luciferase counts. MicroRNA regulation of GPD1L 3ⴕ UTR. HEK 293A cells were transfected overnight with pre-miRs (Ambion), followed by transfection of UTR plasmids

FIG. 3. miR-210 regulates endogenous levels of GPD1L protein. (A) HeLa cells were transiently transfected with either control (NC) or miR-210 pre-miRs at the indicated concentrations, and proteins were detected by Western blotting 48 h later. (B) Stable C2C12 cells were generated using recombinant retrovirus with either empty or miR-210 vectors. Protein was detected using Western blotting after 15 h of hypoxia (H; 0.2% O2) treatment, and RNA was detected via Northern blotting as described in Materials and Methods. EtBr, ethidium bromide staining of total RNA; N, normoxia. (C) HEK 293A cells were transiently transfected with increasing amounts of plasmid encoding firefly luciferase (Luc) or firefly luciferase containing 8 miR-210 biding sites as a 3⬘ UTR (Luc-Sp8). The miR-210 sponge was generated using the miR-210 binding site cloned from the ISCU 3⬘ UTR. Protein was detected by Western blot analysis.

VOL. 31, 2011


2699

FIG. 4. GPD1L exhibits glycerol phosphate dehydrogenase activity. (A) HEK 293A cells were transiently transfected in duplicate with either empty, FLAGHA-GPD1L wild-type, G14A, or K206A vector and subjected to FLAG-mediated immunoprecipitation. Protein was eluted using 3⫻ FLAG peptide and assayed for glycerol phosphate dehydrogenase activity as described in Materials and Methods. (B) U2OS cells were transiently transfected in duplicate with either empty, FLAGHA-GPD1L wild-type, or K206A vector, and whole-cell lysate was assayed for glycerol phosphate dehydrogenase activity with or without the addition of dihydroxyacetone phosphate (DHAP) to the reaction. (C) HEK 293A cells were transiently transfected in triplicate with either empty, FLAGHA-GPD1L, or FLAGHA-GPD1 vector, and whole-cell lysate was assayed for glycerol phosphate dehydrogenase activity. Error bars represent standard deviations.

and CMV-Renilla luciferase the following day. The following day, cells were lysed and luciferase counts were determined using the dual-luciferase reporter assay system from Promega and FLUOstar Omega plate reader (BMG Labtech). All luciferase values are represented as firefly luciferase counts relative to Renilla luciferase counts. Immunoprecipitation. Cells were lysed using a buffer containing 0.1% Triton X-100, 20 mM Tris-Cl, pH 7.5, 125 mM NaCl, 0.5 mM EDTA, 1 mM dithiothreitol (DTT), 1⫻ complete protease inhibitor cocktail (Roche), and 5 ␮M MG132. Cells were vortexed 3 to 4 times while incubated on ice over a course of approximately 30 min. After centrifugation, anti-FLAG-linked agarose beads (Sigma) were added to clarified lysate and incubated at 4°C overnight with rotation. Antibodies. Anti-FLAG–horseradish peroxidase (HRP) and anti-HA–HRP antibodies were obtained from Sigma. Anti-HIF-1␣ antibody was obtained from BD Biosciences. Anti-hydroxy-HIF-1␣ (Pro564) antibody was obtained from Cell Signaling. Anti-PHD2 antibody was obtained from Novus Biologicals. Antitubulin antibody was obtained from Upstate, and secondary mouse and rabbit antibodies were obtained from Jackson Laboratories. Anti-GPD1L antibody was obtained as a gift from Barry London, Cardiovascular Institute, University of Pittsburgh Medical Center. Northern blotting. RNA was isolated using Trizol (Invitrogen) according to the manufacturer’s protocol. Northern blots were run according to the protocol previously described (41). Briefly, RNA was separated using the Bio-Rad Crite-

rion gel system with 15% polyacrylamide, 7 M urea, and 20 mM morpholinepropanesulfonic acid (MOPS)-NaOH, pH 7.0. After electrophoresis, gel was stained in distilled water (dH2O) to obtain an ethidium bromide (EtBr) picture. RNA from gel was transferred to Hybond-NX (Amersham) in dH2O using a semidry transfer (E&K Scientific). After transfer, the gel was restained with EtBr to ensure complete RNA transfer. Cross-linking was accomplished using a solution of 0.16 M 1-ethy-3-(3-dimethylaminopropyl) carbodiimide (Sigma) in 0.13 M 1-methylimidizole, pH 8.0 (MP Biomedicals), which was used to cover one piece of 3MM Whatman filter paper, upon which the membrane was placed, followed by incubation at 60°C for 2 h. After cross-linking, membranes were washed in dH2O. Blocking of the membrane was accomplished using ULTRAhyb-Oligo (Ambion) at 42°C for 30 min. Locked nucleic acid (LNA) miR-210 probe (Exiqon) was ␥-32P end labeled using T4 polynucleotide kinase (NEB) and purified using Illustra ProbeQuant G-50 micro columns (GE Healthcare). After overnight incubation with probe at 42°C, the membrane was washed twice in a solution consisting of 2⫻ SSC (1⫻ SSC is 0.15 M NaCl plus 0.015 M sodium citrate) and 0.5% SDS for 30 min. MicroRNA was visualized using HyBlot CL (Denville Scientific) film. MicroRNA 210 sponge. Firefly luciferase was cloned into pcDNA 3.1 followed by a region of the ISCU 3⬘ UTR of approximately 130 bp generated from PCR of mouse genomic DNA. This DNA fragment contains a miR-210 binding site approximately in the middle of this sequence. Primers were designed so as to

2700

MOL. CELL. BIOL.

KELLY ET AL.

FIG. 5. Increased expression of GPD1L protein results in decreased levels of HIF-1␣ levels and transcription after incubation in hypoxia. (A) U2OS cells were transiently transfected with either empty or FLAGHA-GPD1L vectors and placed in hypoxia (0.2% O2) for the indicated times. Protein was detected using Western blot analysis. (B) U2OS cells were transiently transfected with HRE plasmid linked to firefly luciferase and empty or FLAGHA-GPD1L vectors. Cells were placed under hypoxia treatment (0.2% O2) for 8 h and luciferase activities determined. All wells were transfected with control Renilla luciferase. Values represent means ⫾ SEM of results from two experiments, each in triplicate. Statistical significance was determined using two-way Student’s t test. ***, P ⬍ 0.001 for comparison to the empty-vector control at 8 h of hypoxia.

allow repeated restriction digest and cloning of this fragment 8 times in succession. HEK 293A cells were transfected using CaCl2 and harvested 48 h later. Glycerol-3-phosphate dehydrogenase activity assays. Cells transfected with the indicated plasmids were lysed by rotation at 4°C for 30 min in a buffer composed of 0.25% Triton X-100, 50 mM Tris, pH 7.5, 100 mM NaCl, 0.5 mM EDTA, 1 mM DTT, and 1⫻ complete protease inhibitor cocktail (Roche). Depending on the experiment, either whole-cell lysates were used or FLAGHAGPD1L was immunoprecipitated using anti-FLAG-linked agarose beads (Sigma) and eluted using FLAG peptide (Sigma). The master mix was prepared so that the final concentration of the components after addition of the enzyme was 1 mM dihydroxyacetone phosphate (DHAP; provided as dilithium salt) (Sigma), 0.1 mM NADH (Sigma), and 50 mM Tris-Cl, pH 7.5. The dehydrogenase reaction was carried out at room temperature, where oxidation of NADH in the presence of DHAP results in the generation of NAD⫹ and glycerol-3phosphate. A master mix of reagents was added to a black 96-well plate, followed by addition of enzyme immediately before placement into a FLUOstar Omega plate reader (BMG Labtech). The oxidation state of NADH was monitored by fluorescence analysis (excitation at 355 nm and emission at 460 nm). Metabolite analysis. HEK 293A cells were transfected with either empty vector or vectors overexpressing HA-GPD1L WT, G14A, or K206A. Approximately 16 h later, metabolites were extracted using ice-cold 80% methanol, three times. Each extraction was pooled. Liquid chromatography-mass spectrometry (LC-MS) data were acquired using a model 4000 QTRAP triple quadrupole mass spectrometer (AB SCIEX, Foster City, CA) equipped with an HTS PAL autosampler (Leap Technologies, Carrboro, NC) and an Agilent 1200 series binary high-performance liquid chromatography (HPLC) pump (Santa Clara, CA). Central metabolites were separated using an Atlantis T3 column (3 by 100 mm; 3 ␮M particle; Waters, Milford, MA) that was eluted at a flow rate of 350 ␮l/min using 10 mM tributylamine-15 mM acetic acid (mobile phase A) and methanol (mobile phase B). The elution gradient program was as follows: 100% mobile phase A at initiation, 100% A at 2 min, 2% A at 22 min, and 2% A at 23.5 min. Multiple reaction monitoring (MRM) was used to acquire targeted MS data for specific metabolites in the negative ion mode. The electrospray ionization source voltage was ⫺4.5 kV, and the source temperature was 550°C. Declustering potentials and collision energies were optimized for each metabolite by infusion of reference standards prior to sample analyses. The scheduled MRM algorithm in the Analyst 1.5 software program (AB SCIEX; Foster City, CA) was used to automatically set dwell times for each transition. MultiQuant software (version 1.1; AB SCIEX; Foster City, CA) was used for automated peak integration, and metabolite peaks were manually reviewed for quality of integration and compared against a known standard to confirm identity.

RESULTS Overexpression of miR-210 results in increased HIF-1␣ accumulation. In order to investigate the regulatory role of miR210 during hypoxia, miR-210 was overexpressed and levels of

HIF-1␣ in both normoxia and hypoxia were assayed. Low-level overexpression of HIF-1␣ along with a transfection of a miR210 pre-miR under normoxic conditions resulted in increased HIF-1␣ protein compared to the level for control microRNA transfection (Fig. 1A). When miR-210 was transfected into HeLa cells, endogenous HIF-1␣ accumulation was significantly increased after incubation in both 0.2% and 1.0% oxygen for 8 h compared to the levels for two negative controls (Fig. 1B). Furthermore, ectopic expression of miR-210 caused increased endogenous HIF-1␣ transcriptional activity after exposure to low-oxygen conditions, as detected by cell-based luciferase reporter assays using a plasmid containing HIF-1␣ driven hypoxic response elements (HREs) (10) (Fig. 1C). This transcriptional activation was also detected when endogenous HIF-1␣ gene targets were measured after incubation under hypoxic conditions. Both VEGF and GLUT1 mRNAs showed significantly higher levels in cells overexpressing miR-210 after hypoxic treatment than either of two control miRNAs (Fig. 1D). These data suggest that an important function of miR-210 is to induce the protein levels of HIF-1␣ during hypoxia and that this function is maintained during normoxia. miR-210 directly regulates GPD1L through its 3ⴕ UTR, and this enzyme has glycerol-phosphate dehydrogenase activity. In order to identify which miR-210 target elicits the stabilizing effect on HIF-1␣, Targetscan, an in silico target prediction database, was utilized (32). One such target, glycerol-3-phosphate dehydrogenase 1-like (GPD1L), was selected for further study because of the established connection between glycolytic enzymes and hypoxia. To experimentally validate and confirm the Targetscan prediction of miR-210 targeting GPD1L, we assayed the expression of a luciferase reporter gene fused to the 3⬘ UTR of GPD1L when coexpressed with miR-210. Figure 2A shows the proposed microRNA binding site located within the 3⬘ UTR of the GPD1L gene as well as two mutant UTRs that were generated. The first mutant retained the 5⬘-most binding site of microRNA but lost the ability of miR-210 binding in the seed region (33) (bases 2 to 8; designated the “seed mutant”). The second mutant contained mutations in both the seed regions and the 5⬘-most binding region (“full mutant”). A

VOL. 31, 2011


2701

FIG. 6. Knockdown of GPD1L protein results in increased levels of HIF-1␣ and HIF-1␣-mediated transcription. (A) Stable HeLa cells were generated using lentivirus expressing one of three control shRNAs or one of five GPD1L-targeted shRNAs. Endogenous HIF-1␣ protein was detected by Western blot analysis in nuclear extract preparations under normoxic treatment or in whole-cell extract after 3 h of incubation in hypoxia (0.2% O2). (B) HeLa cells were transiently transfected with either one of three control shRNA plasmids or one of five GPD1L targeted shRNAs along with HRE containing firefly luciferase and control Renilla luciferase plasmids. At 72 h posttransfection, cells were placed in hypoxia (0.2% O2) for 8 h. Each point represents the mean of results from one experiment (completed in triplicate) for either one of three control shRNAs or one of five GPD1L-targeted shRNAs. ***, P ⬍ 0.001 for comparison to the sh-control at 8 h of hypoxia as determined by a two-sided stratified exact Wilcoxon test.

150-bp region of the 3⬘ UTR, containing either the wild-type or the mutant UTR, was cloned 3⬘ of firefly luciferase. As depicted in Fig. 2B, increasing miR-210 levels in HEK 293A cells transfected with plasmid containing the wild-type 3⬘ UTR resulted in significant repression of luciferase activity. Cells expressing the 3⬘ UTR with mutations in the seed region partially lost the ability of miR-210 to decrease luciferase activity. Those cells expressing a luciferase construct where miR-210 binding is completely lost responded only slightly to increasing miR210 levels. In each case, a plasmid expressing Renilla luciferase was cotransfected to act as a transfection control. Importantly, these results were further supported for HeLa cells transfected with pre-miRs in that miR-210-expressing cells exhibited decreased endogenous levels of GPD1L protein compared to cells expressing either one of two negative controls (Fig. 3A). Moreover, C2C12 myoblasts stably expressing miR-210 exhibited lower protein levels of GPD1L than control cells, which were further reduced following hypoxia treatment and concomitant miR-210 induction (Fig. 3B). Furthermore, cells transfected with increasing amounts of a miR-210 sponge vector (15) generated using a 3⬘ UTR fragment from the validated miR-210 target ISCU protein resulted in increases in endogenous GPD1L protein levels (Fig. 3C). Taken together, these data indicate that the predicted miR-210 binding site on

GPD1L 3⬘ UTR is a bona fide target and that miR-210 regulates GPD1L protein levels through direct interaction with the mRNA 3⬘ UTR as previously reported (18a). According to BLAST-directed pairwise alignment with cytosolic glycerol-3-phosphate dehydrogenase (GPD1), GPD1L is strikingly highly conserved at the protein level (71% of the amino acid sequence is identical). Therefore, we tested and confirmed that GPD1L contained glycerol-3-phosphate dehydrogenase enzyme activity (Fig. 4A). Furthermore, we generated GPD1L mutants that, based on the homology of this enzyme with GPD1, should no longer display glycerol-3-phosphate dehydrogenase activity. Mutation of lysine 206 to alanine (K206A) in the active site of GPD1L, previously suggested to be important in catalysis on the basis of the crystal structure of GPD1 (40), and glycine 14 to alanine, located in a conserved NAD⫹ binding domain (GXGXXG) (4), resulted in an inactive enzyme. Other glycine-to-alanine mutations within the NAD⫹ binding domain, involving glycine 12 or glycine 17, resulted in unstable GPD1L or did not change the GPD1L dehydrogenase activity, respectively (data not shown). Importantly, DHAP was absolutely required for oxidation of NADH in in vitro assays (Fig. 4B), indicating that this enzyme is a glycerol phosphate-dependent dehydrogenase. Compared to similar levels of GPD1, GPD1L exhibited significantly less

2702

KELLY ET AL.

MOL. CELL. BIOL.

FIG. 7. Increased expression of GPD1L results in decreased HIF-1␣ stability. (A) HEK 293A cells were transfected with either control or FLAG–HIF-1␣-expressing vectors along with either empty, wild-type, or K206A mutant HA-GPD1L vectors. FLAG–HIF-1␣ was immunoprecipitated using FLAG antibody-linked agarose at 48 h posttransfection, and protein was detected using Western blot analysis. (B) HEK 293A cells were transfected with empty vector or with FLAG–HIF-1␣ along with increasing amounts of HA-GPD1L. FLAG–HIF-1␣ was immunopurified using FLAG agarose at 48 h posttransfection, and protein was detected by Western blot analysis. NS, nonspecific band. (C) HEK 293A cells were transfected with increasing amounts of FLAG-GPD1L along with HA–HIF-2␣. HA–HIF-2␣ was immunoprecipitated using HA agarose at 48 h posttransfection and detected by Western blot analysis. (D) HEK 293A cells were transfected with either empty vector or HA-GPD1L wild-type, G14A, or K206A vector along with FLAG-p53. FLAG-p53 was immunoprecipitated using FLAG agarose at 48 h posttransfection and detected by Western blot analysis.

dehydrogenase activity than its counterpart (Fig. 4C). These data indicate that, while exhibiting considerably less activity than GPD1, GPD1L is in fact a glycerol-3-phosphate dehydrogenase enzyme. Altering GPD1L protein levels results in aberrant accumulation of endogenous HIF-1␣ and transcriptional activity. Next, to assess the functional role of miR-210-mediated decreases in GPD1L protein during hypoxia, cells were transiently transfected with GPD1L and placed under hypoxic conditions. Surprisingly, overexpression of GPD1L caused a noteworthy decrease in the accumulation of HIF-1␣ during hypoxia (Fig. 5A). Functionally, this resulted in a decrease in HIF-1␣ transcriptional activity, as assayed by cell-based luciferase reporter assays. Significant reductions in luciferase activity were observed at all time points of hypoxia treatment in the cells overexpressing GPD1L compared to the levels for those transfected with empty vector (Fig. 5B). In contrast, stable cell lines expressing one of five specific shRNAs directed toward GPD1L displayed increased amounts of HIF-1␣ both when detected under normoxic conditions in nuclear extracts and when detected after incubation in hypoxia, compared to cells expressing one of three control shRNAs (Fig. 6A). Similarly, cells expressing GPD1L shRNAs consistently had increased amounts of HIF-1␣-dependent luciferase activity in cell-based reporter assays after incubation in hypoxia (Fig. 6B). These data suggest that a putative role for GPD1L during hypoxia is as a regulator of HIF-1␣ protein levels and therefore transcriptional activity. GPD1L mediates decreases in HIF-1␣ protein levels through increased PHD-dependent HIF-1␣ proline hydroxylation. In an effort to determine the mechanism by which GPD1L regulates HIF-1␣ protein levels, we sought to deter-

mine whether GPD1L affected HIF-1␣ prolyl hydroxylation levels and subsequent protein degradation. We used a FLAGtagged HIF-1␣ (F–HIF-1␣) overexpression vector that recapitulated the effects of GPD1L overexpression on endogenous HIF-1␣. Briefly, F–HIF-1␣ was cotransfected in HEK 293A cells along with empty vector or with HA-GPD1L. Consistent with the effects on endogenous HIF-1␣, Fig. 7A and B show that HIF-1␣ protein levels were strongly decreased by both wild-type and mutant GPD1L and that these effects were amplified when GPD1L overexpression was increased. Interestingly, the catalytically inactive GPD1L K206A mutant was just as efficient or even more efficient at decreasing the protein amounts of HIF-1␣ (Fig. 7A and B). Overexpression of GPD1L also destabilized HIF-2␣ but did not result in increased degradation of p53, suggesting that GPD1L-mediated degradation of HIF-1␣ may be specific for certain proteins (Fig. 7C and D). These data indicate that GPD1L has a cellular function which is independent from its dehydrogenase activity. Taken together, these results indicate that the enzymatic dehydrogenase activity of GPD1L is not required to decrease HIF-1␣ protein levels. To gain insight into how GPD1L affects HIF-1␣ stability, we determined if proteasome and PHD activity were involved in this regulation. As expected, and consistent with previous results (26, 43), HIF-1␣ degradation can be reversed by treatment with the proteasome inhibitor MG132 or with the PHD inhibitor dimethyloxalylglycine (DMOG) (Fig. 8A and B). Interestingly, when HIF-1␣ degradation was blocked with MG132, the level of hydroxylation of proline 564 was consistently higher in samples expressing wildtype or mutant GPD1L (Fig. 8A to C), despite unchanged

VOL. 31, 2011


2703

FIG. 8. Increased expression of GPD1L results in increased HIF-1␣ proline hydroxylation. (A) HEK 293A cells were cotransfected with FLAG–HIF-1␣ along with either empty vector or wild-type-HA-GPD1L overexpression vector. At 48 h posttransfection, cells were treated with 10 ␮M MG132 for 6 h to inhibit proteasome degradation. FLAG–HIF-1␣ was immunoprecipitated using FLAG antibody-linked agarose, and protein was detected by Western blot analysis. (B) HEK 293A cells were transfected with FLAG–HIF-1␣ along with either empty, wild-type, or mutant HA-GPD1L vectors. At 48 h posttransfection, cells were treated with 10 ␮M MG132 or 1 mM DMOG for 6 h to inhibit proteasome degradation or prolyl-hydroxylase activity, respectively. FLAG–HIF-1␣ was immunoprecipitated using FLAG antibody-linked agarose, and protein was detected by Western blot analysis. (C and D) HEK 293A cells were transfected with FLAG–HIF-1␣ along with vectors encoding either wild-type or mutant HA-GPD1L previously identified to be mutated in Bruggada and sudden infant death syndrome (C) or with vectors encoding GPD1L containing mutations in the conserved NAD⫹ binding domain (GXGXXG) or conserved catalytic lysine (D). At 48 h posttransfection, cells were treated with 10 ␮M MG132 for 6 h. FLAG–HIF-1␣ was immunoprecipitated using FLAG antibody-linked agarose, and protein was detected by Western blot analysis. (E) HEK 293A cells were transfected with FLAG–HIF-1␣ along with empty vector, HA-GPD1L, or HA-GPD1. At 48 h posttransfection, cells were treated with 10 ␮M MG132 or 1 mM DMOG for 6 h. FLAG–HIF-1␣ was immunoprecipitated using FLAG antibody-linked agarose, and protein was detected by Western blot analysis. (F) HEK 293A cells were transfected with FLAG–HIF-1␣ or FLAG–HIF-1␣ PP-A (Pro402 and Pro563 mutated to alanine) mutant along with empty vector or increasing amounts of wild-type HA-GPD1L. FLAG–HIF-1␣ or FLAG–HIF-1␣ PP-A was immunoprecipitated using FLAG antibody-linked agarose at 48 h posttransfection, and protein was detected by Western blot analysis.

levels of PHD2 protein (Fig. 7A), the primary prolyl hydroxylase of HIF-1␣ (2, 6). Previous studies have found endogenous genomic mutations in GPD1L that correlate with Brugada syndrome (34) and sudden infant death syndrome (53). Two of these mutations were shown to decrease the enzyme activity of GPD1L (52). In our cellular assay, these mutants are also able to decrease HIF-1␣ stability (Fig. 8D) and this destabilization was reversible upon treatment with MG132. GPD1 expression resulted in a similar destabilization of HIF-1␣ and a concomitant increase in proline 564 hydroxylation (Fig. 8E). To further confirm that the effects of GPD1L on HIF-1␣ levels were mediated through proline

hydroxylation, we transfected a mutant FLAG–HIF-1␣ vector which contains both prolines normally hydroxylated by PHD2 (prolines 402 and 564) mutated to alanine (PP-A mutant). When similar levels of wild-type and mutant HIF-1␣ were expressed, GPD1L had little effect on mutant HIF-1␣ stability, further suggesting that increased proline hydroxylation is a primary mechanism of HIF-1␣ degradation by GPD1L (Fig. 8F). Together, these results suggest that GPD1L affects HIF-1␣ protein stability by increasing the activity of the PHDs, independent of its dehydrogenase activity, resulting in targeting of HIF-1␣ to the proteasome for degradation.

2704

KELLY ET AL.

MOL. CELL. BIOL.

FIG. 9. Model of miR-210 action on HIF-1␣ protein levels. (A) During normal oxygen tension, HIF-1␣ is kept at low levels by proline hydroxylation and subsequent degradation by the proteasome. When HIF-1␣ protein levels are low, protein levels of GPD1L are high due to low levels of miR-210, a direct regulator of GPD1L through a 3⬘ UTR binding site within the mRNA. GPD1L acts to increase the activity of the PHDs, further ensuring HIF-1␣ proline hydroxylation. (B) As oxygen levels decrease, HIF-1␣ protein and transcriptional activity increase, triggering accumulation of miR-210. Increased miR-210 causes decreased GPD1L protein and a further inactivation of the PHDs, resulting in increased HIF-1␣ protein. This action results in a positive feedback regulatory loop where miR-210 acts to further induce and maintain HIF-1␣ protein levels.

DISCUSSION Understanding the full regulation of HIF-1␣ as it relates to cancer prognosis is vital. It has been shown that intertumoral hypoxia may lead to increased resistance to cancer treatment in solid tumors (8), and this may be partially linked to a multidrug resistance gene (MDR1) which is a gene target of HIF-1␣ (14). MDR1 encodes a protein which is a membrane transporter that has been shown to efficiently export a number of anticancer drugs and is associated with therapeutic resistance in some types of cancer (17). HIF-1␣ expression is increased in a number of human cancers (50, 57), correlates with poor prognosis in a number of different cancer types (reviewed in reference 45), and is likely a central factor contributing to cancer phenotypes (reviewed in reference 54). Similarly, increased miR210 expression is correlated with increased aggressiveness and metastatic capability (20) as well as poor survival rates and increased tumor size in breast cancer patients (9) and pancreatic cancer patients (23). Whereas hypoxic regulation of HIF-1␣ is thought to be primarily regulated by proline hydroxylation, it is becoming clear that there are a number of mechanisms for regulating HIF-1␣ at the protein level, some of which fall outside the canonical degradation pathway. The molecular chaperone HSP90 physically interacts with HIF-1␣ (22) and induces ubiquitination and degradation of HIF-1␣ by the proteasome, which is independent of oxygen concentration and VHL (25). Alternatively, it has recently been shown that when prolines 402 and 564, those residues responsible for regulating the stability of HIF-1␣ through hydroxylation, are mutated to alanine, VHL still controls the degradation of HIF-1␣ (1). This may suggest the presence of alternate binding sites for VHL apart from the

previously described hydroxylated proline residues. It seems clear that the full spectrum of HIF-1␣-regulating factors has yet to be determined. As outlined above, we have identified a new regulator of HIF-1␣, GPD1L, which is regulated by the HIF-1␣-inducible microRNA miR-210. Overexpression of miR-210 resulted in increased HIF-1␣ accumulation during hypoxia through decreased expression of GPD1L protein due to miR-210 targeting of the GPD1L mRNA 3⬘ UTR. Overexpression of GPD1L caused increased proline hydroxylation of HIF-1␣ at proline 564, which resulted in decreased HIF-1␣ protein stability. This degradation could be blocked by pharmacological inhibition of the proteasome as well as the PHDs. The data reported here support a model in which, under normal oxygen concentrations, when HIF-1␣ protein and transcriptional activity is suppressed, small amounts of miR-210 result in high GPD1L protein levels and high activity of the PHDs, further ensuring low levels of HIF-1␣ (Fig. 9A). During times of lowered oxygen tension or other induction of miR-210, decreased GPD1L protein potentiates inhibition of PHD activity, leading to increased HIF-1␣ stability and transcriptional activity (Fig. 9B). This model suggests that miR-210 may act to trigger a positive feedback loop where HIF-1␣ drives miR-210 expression, which further induces HIF-1␣ protein stability. In principle, these data correlate well with a recent report showing that miR-210 is induced in late-stage lung cancer, which results in increased HIF-1␣ protein due to decreased expression of subunit D of the succinate dehydrogenase complex (SDH), a miR-210 target (42). The model proposed in these studies is that decreases in SDHD results in increased succinate accumulation, which is a by-product and natural in-

VOL. 31, 2011


hibitor of PHD activity (44). We carried out metabolic profiling of cells overexpressing both wild-type and mutant GPD1L and found no difference in 2-oxoglutarate or succinate concentrations (see Fig. S1 in the supplemental material). Based on this analysis, it is unlikely that the increased hydroxylation of HIF-1␣ mediated by GPD1L is due to a change in either of these metabolites and is therefore distinct from destabilization mechanisms proposed in previous studies. We have also attempted to detect binding between GPD1L and PHD2 as well as between GPD1L and HIF-1␣; however, we have not been successful thus far. This may suggest that the mechanism of GPD1L regulation of HIF-1␣ stability may be divergent from the mechanism of increased hydroxylation due to expression of OS-9 (3). It remains unclear how changing levels of GPD1L alters HIF-1␣ hydroxylation, and this should be further studied. Interestingly, it has previously been reported that glycerol phosphate dehydrogenase activity has an inverse correlation with tumor growth rate (39), and ratios of LDH/GPD activity are increased in a number of tumors compared to the level for normal tissue (7). In this context, it is possible that inducing the expression of GPD1L, possibly through inhibition of miR-210 in tumors, may be advantageous during cancer treatments due to decreased HIF-1␣ protein and associated transcriptional targets. ACKNOWLEDGMENTS We thank John Dominy and Jorge Ruas for careful reviewing of the manuscript and Nika Danial and all members of the Puigserver laboratory for discussions relating to this project. A special thanks to Barry London for the GPD1L antibody. These studies were supported in part by an Ellison Medical Foundation New Scholar Award, American Diabetes Association, U.S. Department of Defense, and NIH grant R01 DK069966. We declare no conflict of interest. REFERENCES 1. Andre, H., and T. S. Pereira. 2008. Identification of an alternative mechanism of degradation of the hypoxia-inducible factor-1alpha. J. Biol. Chem. 283:29375–29384. 2. Appelhoff, R. J., et al. 2004. Differential function of the prolyl hydroxylases PHD1, PHD2, and PHD3 in the regulation of hypoxia-inducible factor. J. Biol. Chem. 279:38458–38465. 3. Baek, J. H., et al. 2005. OS-9 interacts with hypoxia-inducible factor 1alpha and prolyl hydroxylases to promote oxygen-dependent degradation of HIF1alpha. Mol. Cell 17:503–512. 4. Baker, P. J., K. L. Britton, D. W. Rice, A. Rob, and T. J. Stillman. 1992. Structural consequences of sequence patterns in the fingerprint region of the nucleotide binding fold. Implications for nucleotide specificity. J. Mol. Biol. 228:662–671. 5. Behm-Ansmant, I., et al. 2006. mRNA degradation by miRNAs and GW182 requires both CCR4:NOT deadenylase and DCP1:DCP2 decapping complexes. Genes Dev. 20:1885–1898. 6. Berra, E., et al. 2003. HIF prolyl-hydroxylase 2 is the key oxygen sensor setting low steady-state levels of HIF-1alpha in normoxia. EMBO J. 22:4082– 4090. 7. Boxer, G. E., and C. E. Shonk. 1960. Low levels of soluble DPN-linked alpha-glycerophosphate dehydrogenase in tumors. Cancer Res. 20:85–91. 8. Brown, J. M., and A. J. Giaccia. 1998. The unique physiology of solid tumors: opportunities (and problems) for cancer therapy. Cancer Res. 58:1408–1416. 9. Camps, C., et al. 2008. hsa-miR-210 Is induced by hypoxia and is an independent prognostic factor in breast cancer. Clin. Cancer Res. 14:1340–1348. 10. Carrero, P., et al. 2000. Redox-regulated recruitment of the transcriptional coactivators CREB-binding protein and SRC-1 to hypoxia-inducible factor 1alpha. Mol. Cell. Biol. 20:402–415. 11. Chan, S. Y., et al. 2009. MicroRNA-210 controls mitochondrial metabolism during hypoxia by repressing the iron-sulfur cluster assembly proteins ISCU1/2. Cell Metab. 10:273–284. 12. Chen, Z., Y. Li, H. Zhang, P. Huang, and R. Luthra. 2010. Hypoxia-regulated microRNA-210 modulates mitochondrial function and decreases ISCU and COX10 expression. Oncogene 29:4362–4368.

2705

13. Chendrimada, T. P., et al. 2007. MicroRNA silencing through RISC recruitment of eIF6. Nature 447:823–828. 14. Comerford, K. M., et al. 2002. Hypoxia-inducible factor-1-dependent regulation of the multidrug resistance (MDR1) gene. Cancer Res. 62:3387–3394. 15. Ebert, M. S., J. R. Neilson, and P. A. Sharp. 2007. MicroRNA sponges: competitive inhibitors of small RNAs in mammalian cells. Nat. Methods 4:721–726. 16. Epstein, A. C., et al. 2001. C. elegans EGL-9 and mammalian homologs define a family of dioxygenases that regulate HIF by prolyl hydroxylation. Cell 107:43–54. 17. Fardel, O., V. Lecureur, and A. Guillouzo. 1996. The P-glycoprotein multidrug transporter. Gen. Pharmacol. 27:1283–1291. 18. Fasanaro, P., et al. 2008. MicroRNA-210 modulates endothelial cell response to hypoxia and inhibits the receptor tyrosine kinase ligand EphrinA3. J. Biol. Chem. 283:15878–15883. 18a.Fasanaro, P., et al. 2009. An integrated approach for experimental target identification of hypoxia-induced miR-210. J. Biol. Chem. 284:35134–35143. 19. Favaro, E., et al. 2010. MicroRNA-210 regulates mitochondrial free radical response to hypoxia and krebs cycle in cancer cells by targeting iron sulfur cluster protein ISCU. PLoS One 5:e10345. 20. Foekens, J. A., et al. 2008. Four miRNAs associated with aggressiveness of lymph node-negative, estrogen receptor-positive human breast cancer. Proc. Natl. Acad. Sci. U. S. A. 105:13021–13026. 21. Fong, G. H., and K. Takeda. 2008. Role and regulation of prolyl hydroxylase domain proteins. Cell Death Differ. 15:635–641. 22. Gradin, K., et al. 1996. Functional interference between hypoxia and dioxin signal transduction pathways: competition for recruitment of the Arnt transcription factor. Mol. Cell. Biol. 16:5221–5231. 23. Greither, T., et al. 2010. Elevated expression of microRNAs 155, 203, 210 and 222 in pancreatic tumors is associated with poorer survival. Int. J. Cancer 126:73–80. 24. Huang, L. E., Z. Arany, D. M. Livingston, and H. F. Bunn. 1996. Activation of hypoxia-inducible transcription factor depends primarily upon redox-sensitive stabilization of its alpha subunit. J. Biol. Chem. 271:32253–32259. 25. Isaacs, J. S., et al. 2002. Hsp90 regulates a von Hippel Lindau-independent hypoxia-inducible factor-1 alpha-degradative pathway. J. Biol. Chem. 277: 29936–29944. 26. Jaakkola, P., et al. 2001. Targeting of HIF-alpha to the von Hippel-Lindau ubiquitylation complex by O2-regulated prolyl hydroxylation. Science 292: 468–472. 27. Jewell, U. R., et al. 2001. Induction of HIF-1alpha in response to hypoxia is instantaneous. FASEB J. 15:1312–1314. 28. Kaelin, W. G., Jr., and P. J. Ratcliffe. 2008. Oxygen sensing by metazoans: the central role of the HIF hydroxylase pathway. Mol. Cell 30:393–402. 29. Kallio, P. J., et al. 1998. Signal transduction in hypoxic cells: inducible nuclear translocation and recruitment of the CBP/p300 coactivator by the hypoxia-inducible factor-1alpha. EMBO J. 17:6573–6586. 30. Kiriakidou, M., et al. 2007. An mRNA m7G cap binding-like motif within human Ago2 represses translation. Cell 129:1141–1151. 31. Kulshreshtha, R., et al. 2007. A microRNA signature of hypoxia. Mol. Cell. Biol. 27:1859–1867. 32. Lewis, B. P., C. B. Burge, and D. P. Bartel. 2005. Conserved seed pairing, often flanked by adenosines, indicates that thousands of human genes are microRNA targets. Cell 120:15–20. 33. Lewis, B. P., I. H. Shih, M. W. Jones-Rhoades, D. P. Bartel, and C. B. Burge. 2003. Prediction of mammalian microRNA targets. Cell 115:787–798. 34. London, B., et al. 2007. Mutation in glycerol-3-phosphate dehydrogenase 1 like gene (GPD1-L) decreases cardiac Na⫹ current and causes inherited arrhythmias. Circulation 116:2260–2268. 35. Masson, N., C. Willam, P. H. Maxwell, C. W. Pugh, and P. J. Ratcliffe. 2001. Independent function of two destruction domains in hypoxia-inducible factor-alpha chains activated by prolyl hydroxylation. EMBO J. 20:5197–5206. 36. Min, J. H., et al. 2002. Structure of an HIF-1alpha-pVHL complex: hydroxyproline recognition in signaling. Science 296:1886–1889. 37. Mizuno, Y., et al. 2009. miR-210 promotes osteoblastic differentiation through inhibition of AcvR1b. FEBS Lett. 583:2263–2268. 38. Naviaux, R. K., E. Costanzi, M. Haas, and I. M. Verma. 1996. The pCL vector system: rapid production of helper-free, high-titer, recombinant retroviruses. J. Virol. 70:5701–5705. 39. Ostro, M. J., and T. P. Fondy. 1977. Isolation and characterization of multiple molecular forms of cytosolic NAD-linked glycerol-3-phosphate dehydrogenase from normal and neoplastic rabbit tissues. J. Biol. Chem. 252: 5575–5583. 40. Ou, X., et al. 2006. Crystal structures of human glycerol 3-phosphate dehydrogenase 1 (GPD1). J. Mol. Biol. 357:858–869. 41. Pall, G. S., C. Codony-Servat, J. Byrne, L. Ritchie, and A. Hamilton. 2007. Carbodiimide-mediated cross-linking of RNA to nylon membranes improves the detection of siRNA, miRNA and piRNA by northern blot. Nucleic Acids Res. 35:e60. 42. Puissegur, M. P., et al. 2011. miR-210 is overexpressed in late stages of lung cancer and mediates mitochondrial alterations associated with modulation of HIF-1 activity. Cell Death Differ. 18:465–478.

2706

KELLY ET AL.

43. Salceda, S., and J. Caro. 1997. Hypoxia-inducible factor 1alpha (HIF1alpha) protein is rapidly degraded by the ubiquitin-proteasome system under normoxic conditions. Its stabilization by hypoxia depends on redoxinduced changes. J. Biol. Chem. 272:22642–22647. 44. Selak, M. A., et al. 2005. Succinate links TCA cycle dysfunction to oncogenesis by inhibiting HIF-alpha prolyl hydroxylase. Cancer Cell 7:77–85. 45. Semenza, G. L. 2010. Defining the role of hypoxia-inducible factor 1 in cancer biology and therapeutics. Oncogene 29:625–634. 46. Semenza, G. L. 2007. Hypoxia-inducible factor 1 (HIF-1) pathway. Sci. STKE 2007:cm8. 47. Semenza, G. L. 2009. Regulation of cancer cell metabolism by hypoxiainducible factor 1. Semin. Cancer Biol. 19:12–16. 48. Semenza, G. L. 2003. Targeting HIF-1 for cancer therapy. Nat. Rev. Cancer 3:721–732. 49. Stewart, S. A., et al. 2003. Lentivirus-delivered stable gene silencing by RNAi in primary cells. RNA 9:493–501. 50. Talks, K. L., et al. 2000. The expression and distribution of the hypoxiainducible factors HIF-1alpha and HIF-2alpha in normal human tissues, cancers, and tumor-associated macrophages. Am. J. Pathol. 157:411–421.

MOL. CELL. BIOL. 51. Thermann, R., and M. W. Hentze. 2007. Drosophila miR2 induces pseudopolysomes and inhibits translation initiation. Nature 447:875–878. 52. Valdivia, C. R., K. Ueda, M. J. Ackerman, and J. C. Makielski. 2009. GPD1L links redox state to cardiac excitability by PKC-dependent phosphorylation of the sodium channel SCN5A. Am. J. Physiol. Heart Circ. Physiol. 297: H1446–H1452. 53. Van Norstrand, D. W., et al. 2007. Molecular and functional characterization of novel glycerol-3-phosphate dehydrogenase 1 like gene (GPD1-L) mutations in sudden infant death syndrome. Circulation 116:2253–2259. 54. Vogelstein, B., and K. W. Kinzler. 2004. Cancer genes and the pathways they control. Nat. Med. 10:789–799. 55. Yeung, S. J., J. Pan, and M. H. Lee. 2008. Roles of p53, MYC and HIF-1 in regulating glycolysis—the seventh hallmark of cancer. Cell. Mol. Life Sci. 65:3981–3999. 56. Zhang, Z., et al. 2009. MicroRNA miR-210 modulates cellular response to hypoxia through the MYC antagonist MNT. Cell Cycle 8:2756–2768. 57. Zhong, H., et al. 1999. Overexpression of hypoxia-inducible factor 1alpha in common human cancers and their metastases. Cancer Res. 59:5830–5835.