Hypoxia-inducible Factor 1α (HIF-1α) Is a Non-heme Iron Protein

The hypoxia-inducible factor 1 complex (HIF-1) is involved in the transcriptional activation of several genes, like erythropoietin and vascular endothelial growth factor, that are responsive to the lack of oxygen. The HIF-1 complex is composed of two b-HLH proteins: HIF-1β, that is constitutively expressed, and HIF-1α, that is present only in hypoxic cells. The HIF-1α subunit is continuously synthesized and degraded by the ubiquitin-proteasome under oxic conditions. Hypoxia, transition metals, iron chelators, and several antioxidants stabilize the HIF-1α protein, allowing the formation of the transcriptionally active HIF-1 complex. The mechanisms of oxygen sensing and the pathways leading to HIF-1α stabilization are unclear. Because the involvement of a heme protein oxygen sensor has been postulated, we tested the heme sensor hypothesis by using a luciferase-expressing cell line (B-1), that is highly responsive to hypoxia. Exposure of B-1 cells to carbon monoxide and heme synthesis inhibitors failed to show any effect on the hypoxia responsiveness of these cells, suggesting that heme proteins are not involved in hypoxia sensing. Measurement of iron in recombinantly expressed HIF-1α protein revealed that this protein binds ironin vivo. Iron binding was localized to a 129-amino acid peptide between sequences 529 and 658 of the HIF-1α protein. Although the exact structure of the iron center has not been yet defined, a 2:1 metal/protein molar ratio suggests a di-iron center, probably similar to the one found in hemerythrin. This finding is compatible with a model where redox reaction may occur directly in the iron center of the HIF-1α subunit, affecting its survival inoxic conditions.

The adaptive responses to hypoxia features the activation of genes, like erythropoietin (Epo), 1 vascular endothelial growth factor, endothelin, glucose transporters, and glycolytic enzymes, that are specifically stimulated by the lack of oxygen (1,2). These genes share a common mechanism for oxygen sensing and transcriptional activation. The key step in gene activation by hypoxia is the formation of the HIF-1 (hypoxia-inducible factor-1) protein complex in the corresponding hypoxia-responsive enhancer sequences (HIF-1 sites) (3,4). The HIF-1 complex is a heterodimeric complex of two helix-loop-helix PAS proteins (5); HIF-1␤ or ARNT (aryl hydrocarbon nuclear receptor translocator), that is constitutively expressed, and HIF-1␣, that is rapidly degraded under normoxic conditions by the ubiquitin-proteasome system (6). Hypoxia induces the stabilization of the HIF-1␣ subunit, thus allowing the formation of the transcriptionally active complex. Stabilization of HIF-1␣ is also induced by transition metals such as cobalt, nickel, and manganese, by iron chelators and by antioxidants such as the thiol reducing agent N- (2-mercaptopropionyl)glycine and the oxygen radical scavenger 2-acetamidoacrylic acid (ADA-1) (6). The mechanisms by which cells sense oxygen tension and regulate the survival of the HIF-1␣ protein are currently unknown. Early work by Goldberg et al. (7) hypothesized the involvement of a heme protein as an oxygen sensor for hypoxic gene activation. Their model was based on the finding that carbon monoxide (CO), heme synthesis inhibitors, and iron chelators inhibited the response to hypoxia. They suggested that transition metals would substitute for iron in the heme porphyrin ring, and because they have low affinity for oxygen, they would lock the sensor in the deoxy conformation. To date, however, no such molecule has been identified in mammalian cells. Furthermore, desferrioxamine (iron chelator), rather than being an inhibitor, was later found to be a potent stimulator of HIF-1 activation. Similarly, the inhibitory effect of heme synthesis inhibitors and CO has not been found in all cells. Alternative models for oxygen sensing involving redox reactions have been proposed (8). These models are based on the observation that addition of H 2 O 2 blocks HIF-1 complex formation and from the recent finding that some antioxidants can stimulate HIF-1 in normoxic conditions (6,9,10). In this paper we report that CO and heme synthesis inhibitors have little effect on the hypoxia response on cells expressing a reporter gene under the control of the hypoxia-responsive enhancer. Furthermore, we report the finding that HIF-1␣ is itself a non-heme iron-binding protein and propose that oxygen sensing occurs by direct interaction of O 2 with this iron center.

MATERIALS AND METHODS
Cell Lines and Culture Conditions-Hep 3B and B-1 cells were grown in nucleoside-free ␣ minimal essential medium (Life Technologies, Inc.) supplemented with 10% heat-inactivated fetal calf serum (HyClone, Logan, VT), 2.5 mg/ml fungizone (Life Technologies, Inc.), 100 units/ml penicillin (Life Technologies, Inc.), 100 g/ml streptomycin (Life Technologies, Inc.). Cell lines were maintained in a well humidified incubator at 37°C in 5% CO 2 , 95% air. For hypoxic stimulation, the culture plates were incubated in a modular incubator chamber (Billups-Rothenburg, Del Mar, CA) and flushed with a gas mixture containing 1% O 2 , 5% CO 2 , and nitrogen balanced. For CO treatment, the cells were flushed with a gas mixture containing either 20% (normoxia) or 1% (hypoxia) O 2 , 6% CO, 5% CO 2 and balanced with N 2 . Cells were stim-* This work was supported in part by National Institutes of Health Grant DK-34642 and Juvenile Diabetes Foundation Grant 195009. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 (11). Luciferase Assay-Luciferase expression was assayed using a commercially available kit (Luciferase Assay System, Promega, Madison, WI). Briefly, cells were washed three times with cold phosphate-buffered saline prior to lysis with a 1 ϫ concentration of the supplied lysis buffer. Samples were collected, and 5-l aliquots were assayed using the luciferase assay reagent in a TD 20/20 luminometer (Promega). The results are expressed as relative light units per g of total protein. Protein concentrations were measured by the method of Bradford using a Bio-Rad kit (Bio-Rad, Hercules, CA), with bovine serum albumin (Sigma) as the standard.
Expression Vectors and Protein Production-All the HIF-1 expression plasmids were generated using the glutathione S-transferase fusion system (pGEX-4T-1 expression vector) from Amersham Pharmacia Biotech. The HIF-1␣ inserts were generated using polymerase chain reactions (PCR) and a human HIF-1␣ cDNA as template. The following PCR primers were synthesized at the Jefferson Nucleic Acid Facility: 529 GGATCCGAATTCAAGTTG, 589 GTCGACTCGAGTCATCAGCTTG-CGGA, 658 CTCGAGTCGACTTATGGTGATGATGT, and 826 GTCGACG-GATCCGTTAACTTGATC. PCR reactions were carried out in a Perkin-Elmer (DNA-thermal cycler 2.1) PCR machine for 30 cycles. The reaction products were digested with EcoRI and SalI and ligated, in frame, downstream of the 26-kDa GST protein cDNA, into EcoRI-SalIdigested pGEX-4T-1 vector. All the expression vectors were verified by DNA sequencing. The GST-B-filamin control plasmid was a gift of Dr. Toshiro Takafuta (Cardeza Foundation). Plasmids were utilized to transform Escherichia coli host strains (BL21) using standard procedures. Fusion proteins were generated by treating mid log phase cultures with 1 mM IPTG (isopropyl-␤-D-thiogalactopyranoside) to induce fusion protein expression. Cells were harvested by centrifugation and subsequently lysed by sonication. Solubilized extracts were applied to glutathione-Sepharose beads (Amersham Pharmacia Biotech) for 2 h, and the beads were subsequently washed four times with phosphatebuffered saline. Aliquots of beads were electrophoresed on SDS-polyacrylamide gels to verify protein production and recovery (Fig. 3). For studies using transition metals, either CoCl 2 , NiCl 2 , or MnCl 2 were added to the cultures at a final concentration of 500 M 1 h prior to IPTG induction.
Quantitation of Complexed Iron-Iron content was determined initially by the colorimetric method of Fish (12) with ferrous ethylenediammonium sulfate as the standard. About 500 g of protein was subjected to acid-permanganate (Sigma) treatment at 60°C for 2 h to facilitate iron release. Subsequently a solution containing Ferrozine (disodium 3-(2-pyridyl-5,6-bis(4-phenylsulfonate)-1,2,4-triazine, Sigma), neocuproine (2,9-dimethyl-1,10-bathophenanthroline, Sigma), ascorbic acid and buffered by ammonium acetate was added to detect the presence of iron by a magenta color formation. Iron was quantitated spectrophotometrically at 562 nm. The presence of complexed iron was further determined independently by atomic absorption spectrometry performed at Galbraith Laboratories Inc. (Knoxville, TN), using a graphite microfurnace assay (Perkin-Elmer model 4110 ZL AA spectrometer (13). Proteins were measured by the Bradford method (Bio-Rad kit) and independently by a micro-Kjeldahl method (14) at the Galbraith Laboratories Inc.

Effect of Heme Synthesis Inhibitors on Hypoxia Response-
The effect of the heme synthesis inhibitor 4,6-dioxoheptanoic acid (DHA) (15) on the hypoxia, cobalt, and desferrioxamine responses was studied in B-1 cells. These cells were derived from Hep 3B cells by stable transfection with a luciferase reporter under the control of a minimal Epo promoter and the hypoxia-responsive enhancer. As described previously (6), B-1 cells respond to hypoxia, cobalt chloride, and iron chelators, in a time-and concentration-dependent way, by increasing lucif-erase expression. B-1 cells were incubated with 2 mM AHA for various periods, from 6 to 48 h, and then exposed to hypoxia, cobalt chloride, or Dfx, for another 8 h. The results are expressed as the ratios between DHA-treated and untreated cells with or without stimulation. As shown in Fig. 1, although DHA was mildly inhibitory to control cells, it had no significant effect on the stimulation of luciferase expression by any of the three agents. That DHA was indeed inhibitory to heme synthesis was confirmed by measuring 59 Fe incorporation into heme, which showed 90 -95% inhibition.
Effect of CO on Luciferase Expression by B-1 Cells-The effect of CO was studied in B-1 cells exposed to normoxia, hypoxia, cobalt chloride, and Dfx in the presence of 6% CO. As shown in Fig. 2, although exposure to 6% CO was mildly inhibitory, this inhibition was the same for all experimental groups, suggesting a nonspecific effect. Similar results were observed when 10% CO was utilized (not shown).
HIF-1␣ as an Iron-binding Protein-The above results indicated that the oxygen sensor is unlikely a heme-containing protein. An alternative possibility to explain the stimulatory effect of iron chelators and antioxidants on HIF-1␣ stability would be the presence of a non-heme iron center in the HIF-1␣ protein itself. To evaluate the presence of iron in HIF-1␣ we expressed the protein as a GST fusion protein. HIF-1␣ cDNAs corresponding to amino acid sequences 529 -826, 529 -658, and 529 -589 were cloned, in frame, downstream of the 26-kDa GST domain of Schistosoma japonicum in the pGEX expression vector. The recombinant proteins were expressed in E. coli and purified by GST affinity chromatography using glutathione-Sepharose beads followed by glutathione elution. Iron was measured by utilizing the acid-permanganate Ferrozine method. Iron was released from the protein by treatment with an acid/permanganate mixture and the iron chelated by Ferrozine, which forms a water-soluble, highly stable, colored ferrous complex. This method provides reproducible sensitivity down to about 0.1-0.2 g of iron. For a hypothetical ironcontaining protein that binds 1 iron per M r 50,000, 100 g of protein would provide sufficient material for precise determinations. Interestingly, initial studies using proteins purified by glutathione elution failed to show the presence of iron in any of the peptide fragments. However, subsequent experiments utilizing uneluted proteins (still attached to the Sepharose beads) showed distinctively the presence of iron in fragments 529 -829 and 529 -658. Control glutathione-Sepharose beads, the 27-kDa GST-peptide, and a 99-kDa GST-B-filamin fusion protein,  Table I. The presence of iron in the HIF-1␣ fragments was confirmed by graphite furnace atomic absorption spectroscopy, as measured by an independent laboratory (Galbraith Laboratories Inc.). No heme was detected in the recombinant proteins (Drabkin's reagent, Fisher) Protein was measured by the Bradford method and by a micro-Kjeldahl nitrogen assay at Galbraith Laboratories Inc. The iron/protein molar ratio appears to be between 1:1 to 2:1. These estimates are based on six independent protein preparations.

Effect of Transition Metals and Iron Chelators on the Iron Content of HIF-1␣ Fragments-
To evaluate the effect of transition metals and iron chelators on the iron content of the peptide fragments, E. coli cells expressing the 529 -658 fragment were incubated with Dfx (100 M) and the transition metals cobalt, manganese, nickel (500 M) 30 min before induction with IPTG. The recombinant peptides were purified by glutathione-Sepharose and the iron content analyzed by atomic absorption spectroscopy. As shown in Fig. 3, Dfx and manganese did not significantly affect protein expression. However, the iron content in those samples was significantly reduced, as shown in Table I. Furthermore, there was an increase manganese content in the manganese-treated samples, suggesting a replacement of iron by manganese. No effect on iron content was observed in the samples treated with cobalt or nickel.
Removal of Iron by Chelating Agents in Vitro-To determine the oxidation state of the iron in HIF-1␣ and the ability of chelators to interact with it, we incubated the recombinant 529 -568 fragment with the Fe 2ϩ chelator 1,10-bathophenanthroline or the Fe 3ϩ chelators desferrioxamine and tiron. Both Fe 3ϩ chelators completely removed the iron from the peptide, whereas the Fe 2ϩ chelator had little effect. The finding that chelators remove iron from the proteins explains the absence of detectable iron in the glutathione-treated peptides, because glutathione is a potent iron chelator (29). DISCUSSION The hypoxic activation of several genes is mediated by the hypoxia-responsive enhancer, first described in the 3Ј end of the Epo gene (16 -18). A hypoxia-inducible protein complex, termed HIF-1, binds to the enhancer and stimulates transcription. Of the two proteins that form the complex, HIF-1␤ (ARNT) is constitutively expressed, whereas HIF-1␣ is only present in hypoxic cells (5). Recent evidence has shown that HIF-1␣ protein is continually synthesized, but rapidly de-graded in normoxia. Normoxic degradation of HIF-1␣ appears to be mediated by the ubiquitin-proteasome system by still yet undetermined signals (6). Similarly to hypoxia, HIF-1␣ degradation is inhibited by iron chelators, transition metals such as cobalt, nickel, and manganese and several antioxidants. Based on the findings that Epo gene expression by hypoxia could be inhibited by CO and heme synthesis inhibitors, Goldberg et al. (7) postulated the presence of a rapidly turning over heme protein as an oxygen sensor. In their model, transition metals would be incorporated into the heme molecule, and because they bind oxygen with low affinity, they would lock the sensor in the deoxy conformation. So far, however, there has been no identification of such sensor, and furthermore, there are several experimental observations that cannot be accounted by the heme sensor model (reviewed in Ref. 19). Our failure to show any effect of heme synthesis inhibitors or CO in the response to hypoxia, using the sensitive and specific response of B-1 cells, is very suggestive that heme proteins are not involved in oxygen sensing. These results are in line with similar findings reported by Eckardt et al. (20) in freshly isolated hepatocytes and by a recent report by Graven et al. (21) on the lack of effect of CO and AHA in the response to hypoxia by pulmonary endothelial cells.
Molecular interactions of oxygen are essentially of two kinds; reversible liganding, as in hemoglobin, and redox-based, were oxygen acts as an electron acceptor. Oxygen-sensing mechanisms implying redox reactions have been already proposed (8). Treatment of cells with H 2 O 2 will inhibit the activation of HIF-1 and the hypoxic expression of Epo and vascular endothelial growth factor genes (9). Conversely, antioxidants such as the thiol donor N-(2-mercaptopropionyl)glycine or oxygen radical scavengers like ADA-1 or mannitol would induce HIF-1␣ and gene expression in normoxic cells (6). Acker

Hypoxia-inducible Protein-1␣ 18021
survival is unclear. Although hydrogen peroxide can react with Fe 2ϩ , by way of the Fenton reaction generating hydroxyl radicals (OH ⅐ ), these radicals are so reactive that it is difficult to conceive their role as specific signal transducers. Furthermore, studies using iodonium compounds, potent inhibitors of NADPH oxidases (22), have failed to show a stimulatory effect on HIF-1␣ expression (23). 2 The finding that HIF-1␣ is an iron-containing protein provides an alternative mechanism for oxygen sensing; localized Fenton reactions could occur in the core of the protein itself, leading to oxidative in situ modification of critical amino acid residues or changing the conformation of the protein, thus targeting it for proteasomal degradation. In this model, iron chelators would act by either decreasing the availability of a labile iron pool or by directly removing iron from HIF-1␣. Transition metals may compete for iron for the metal binding site in the protein, as was shown for the case of manganese in our bacterial studies, and antioxidants may inhibit the localized redox reactions.
Although the exact sequences involved in the normoxic degradation of HIF-1␣ have not been completely defined, early work by Jiang et al. (24) indicated that the C-terminal end of HIF-1␣ contained the putative degradation domain. Further work by Pugh et al. (25) provided evidence suggesting that fragment amino acids 530 -634 and within that fragment, amino acids 549 -582, were involved in the oxygen-regulated degradation of the protein. Our finding of an iron binding site in fragment 529 -658 provides a plausible mechanism for this oxygen-regulated degradation. As mentioned, iron can react with H 2 O 2 to generate OH radicals. These highly reactive molecules can react with certain amino acids in their vicinity to produce oxidized adducts, which could then become the target for ubiquitination. A rather similar situation has been found in the case of IRP-2 (iron regulatory protein-2), where iron binding generates carbonyls and induces the proteasomal degradation of the protein (26). The sources of H 2 O 2 remain unclear. Because mitochondria, one of the main sources of H 2 O 2 in cells, appear not to be involved in oxygen sensing, one possible source is cytoplasmic oxidase. Indeed, Acker and colleagues (10) have postulated the involvement of a NADPH-linked oxidase, similar to the one present in neutrophils, as a possible oxygen sensor. However, inhibitors of NADPH-dependent oxidases do not activate HIF-1, suggesting that they are not the sources of peroxides. Alternatively, H 2 O 2 could be produced in situ by autoxidation of Fe 2ϩ in the presence of oxygen. This phenomenon has been described in the case of ferritin and more recently by Biaglow and Kachur (27) in the reaction of molecular oxygen with polyphosphate complexes of ferrous ions. The structure and ligands of the iron pocket in HIF-1␣ are, to this time, undetermined. Our initial data suggest 1-2 mol of iron/ mol of protein. The iron center seems very labile, because iron could be easily removed by chelators, both in vitro and in vivo. Furthermore, treatment with manganese markedly reduced iron incorporation into the protein. Because there are no cysteine residues in the 529 -658 fragment, an iron-sulfur cluster is unlikely. A possible structure may be a di-iron center, similar to the ones found in the oxygen carrying protein hemerythrin and in the enzymes diribonucleotide reductase and methane mono-oxygenase (28). These proteins have the common property of binding oxygen, by way of oxidizing the Fe 2ϩ to Fe 3ϩ . Preliminary studies using electron paramagnetic resonance spectroscopy (EPR) analysis at low temperature (10 K), conducted in Dr. P. L. Dutton's laboratory (University of Pennsylvania), showed no paramagnetic signal. However, this is not unusual in di-iron centers, which usually have both irons antiparamagnetically coupled (28). In summary, we provide evidence that suggests that oxygen sensing is mediated by an iron binding site(s) in the HIF-1␣ protein. The interaction of the iron center, either with H 2 O 2 or directly with oxygen, may provide a signal for the ubiquitin-proteasomal degradation of the protein, thus controlling the transcriptional activation of hypoxia-responsive genes.