Activation of hypoxia-inducible transcription factor depends primarily upon redox-sensitive stabilization of its alpha subunit.

Hypoxia-inducible factor 1 (HIF-1) is a heterodimeric transcription factor that is critical for hypoxic induction of a number of physiologically important genes. We present evidence that regulation of HIF-1 activity is primarily determined by the stability of the HIF-1α protein. Both HIF-1α and HIF-1β mRNAs were constitutively expressed in HeLa and Hep3B cells with no significant induction by hypoxia. However, the HIF-1α protein was barely detectable in normoxic cells, even when HIF-1α was overexpressed, but was highly induced in hypoxic cells, whereas HIF-1β protein levels remained constant, regardless of pO2. Hypoxia-induced HIF-1 binding as well as the HIF-1α protein were rapidly and drastically decreased in vivo following an abrupt increase to normal oxygen tension. Moreover, short pre-exposure of cells to hydrogen peroxide selectively prevented hypoxia-induced HIF-1 binding via blocking accumulation of HIF-1α protein, whereas treatment of hypoxic cell extracts with H2O2 had no effect on HIF-1 binding. These observations suggest that an intact redox-dependent signaling pathway is required for destablization of the HIF-1α protein. In hypoxic cell extracts, HIF-1 DNA binding was reversibly abolished by sulfhydryl oxidation. Furthermore, the addition of reduced thioredoxin to cell extracts enhanced HIF-1 DNA binding. Consistent with these results, overexpression of thioredoxin and Ref-1 significantly potentiated hypoxia-induced expression of a reporter construct containing the wild-type HIF-1 binding site. These experiments indicate that activation of HIF-1 involves redox-dependent stabilization of HIF-1α protein.

Hypoxia-inducible factor 1 (HIF-1) 1 was identified by Semenza, Wang, and co-workers (1-5) as a transcription factor in hypoxic cells that binds specifically to a 3Ј enhancer of the gene encoding erythropoietin and to promoters/enhancers in other genes important in adaptation to hypoxia, such as those encoding tyrosine hydroxylase (6), vascular endothelial growth factor (7), glycolytic enzymes (8 -10), and glucose transporters (11). HIF-1 activity can be induced by hypoxia in a wide variety of cells, as demonstrated by specific binding to oligonucleotides containing a HIF-1 response element (3) and by transactivation of reporter genes (12).
HIF-1 DNA binding, along with the expression of HIF-1responsive genes, can be induced by cobaltous ion and desferrioxamine with kinetics similar to that of hypoxia induction (3,13). HIF-1 activation, irrespective of stimulus, is blocked by pretreatment of cells with cycloheximide (2,3), an inhibitor of protein synthesis as well as by 2-aminopurine (13), a protein kinase inhibitor. Treatment of hypoxic cell extracts with alkaline phosphatase abolishes DNA binding (13). Thus phosphorylation may be required for binding. More recently, Wang et al. (5) reported that both HIF-1␣ mRNA and HIF-1␤ mRNA were barely detectable in normoxic cells but were highly induced following exposure to hypoxia, with rapid decay upon return to normoxia. Their results suggested that HIF-1 is regulated, at least in part, at the level of transcription and/or mRNA stability. However, these results are difficult to reconcile with the fact that both HIF-1␣ (15) and HIF-1␤ (ARNT) (14) genes were cloned independently from nonhypoxic cDNA libraries and that transcripts of both genes were detected at high levels. This concern prompted us to investigate the mechanism by which HIF-1 is regulated by first examining mRNA and protein expression of both ␣ and ␤ subunits. We found that levels of HIF-1␣ and HIF-1␤ mRNA were unaffected by changes in oxygen tension. In contrast, in vivo HIF-1 DNA binding activity induced by hypoxia closely paralleled the abundance of HIF-1␣ protein and that both HIF-1 activity and HIF-1␣ protein decreased markedly, not only by subsequent exposure of cells to oxygen, but also by pre-exposure of cells to hydrogen peroxide. The latter result led to our investigating the role of redox chemistry in HIF-1 activation.

EXPERIMENTAL PROCEDURES
Human HeLa cells were grown in Dulbecco's modified Eagle's medium with 5% heat-inactivated fetal bovine serum, Hep3B cells in ␣-modified Eagle's medium with 10% heat-inactivated fetal bovine serum, and 293 cells in Dulbecco's modified Eagle's medium with 10% heat-inactivated fetal bovine serum. Cells were routinely cultured in 95% air and 5% CO 2 at 37°C and were made hypoxic by placing them in a gas-controlled chamber (Espec) maintained at 1% oxygen, 94% N 2 , and 5% CO 2 . In some experiments, cells were pretreated with 1 mM (or as indicated) hydrogen peroxide (Sigma) for 20 min before exposure to hypoxia. N-Ethylmaleimide, diamide, and dithiothreitol were purchased from Sigma, ␤-mercaptoethanol was from Bio-Rad, and thioredoxin was from Promega Corp.
Plasmid Constructions-pBluescriptSK/HIF-1a3.2T7 containing HIF-1␣ cDNA (kindly provided by G. L. Semenza) was modified by digestion with AvaII, filling in with Klenow fragment, and digestion again with XbaI. The blunt-ended AvaII-XbaI fragment was inserted into HincII and XbaI sites of pBluescriptII SK(ϩ) to create pBS-HIF-1␣. The resulting plasmid was further digested with HpaI and NsiI, bluntended as above, and religated to create pBS-HIF-1␣/HN with deletion of the 3Ј untranslated region. pBS-HIF-1␣ was also digested with SpeI and XbaI, blunt-ended as above, and religated to create pBS-HIF-1␣/ SX. Each plasmid contained appropriate stop codons for HIF-1␣. Eukaryotic expression vectors for HIF-1␣ were constructed by digestion of pBS-HIF-1␣ and pBS-HIF-1␣/HN with KpnI and XbaI, followed by insertion into KpnI and XbaI sites of pcDNA3 (Invitrogen). For HIF-1␣ riboprobe, a 604-base pair EcoRI-EcoRI fragment was internally deleted from pBS-HIF-1␣/SX, resulting in pT3-rHIF-1␣. The human erythropoietin 3Ј enhancer region (3453-3503) containing a functional HIF-1 binding site was polymerase chain reaction-amplified and cloned into KpnI and BglII sites of pGL3-promoter vector (Promega Corp.). The resultant plasmid, pEpoE-luc, carries a luciferase reporter gene driven by this 3Ј enhancer and SV40 promoter. To make pEpoEm1-luc, the HIF-1 site (5Ј-TACGTGCT-3Ј) was mutated to 5Ј-TAAAAGCT-3Ј using the Altered Sites II in Vitro Mutagenesis System (Promega Corp. Preparation of Whole-cell Extracts and Electrophoretic Mobility Shift Assay (EMSA)-Cells were placed on ice, rinsed with phosphate-buffered saline, and harvested. Whole-cell extracts were prepared essentially as described previously (17). Briefly, cell pellets were quickly frozen in liquid nitrogen for 5 min and then thawed on ice for 5-10 min with cell lysis buffer (17). Cells were lysed at 4°C by multiple (20 ϫ) passage through a 26-gauge needle, followed by centrifugation at 12,000 ϫ g for 15 min at 4°C. The supernatants were stored at Ϫ70°C. Protein concentrations were determined by using BCA protein assay reagent (Pierce). A typical DNA binding reaction was carried out by mixing 10 g of cell extracts with DNA binding buffer (17) in the presence of 150 ng of poly(dI⅐dC) and 1.75 pmol of M18 (mutant HIF-1 binding oligonucleotide) (2) for 5 min, followed by incubation for 10 min at room temperature with 17.5 fmol of [ 32 P]W18 (wild-type HIF-1 binding oligonucleotide) (2). The resulting complexes were resolved in 5% polyacrylamide gels in 0.5 ϫ Tris-borate EDTA buffer at 180 V for 100 min at cold room or room temperature. In final concentrations indicated in the text, N-ethylmaleimide, diamide, dithiothreitol, ␤-mercaptoethanol, or thioredoxin was incubated with cell extracts, respectively, or sequentially at room temperature for 10 min before addition to [ 32 P]W18 oligonucleotides. For DR-2, Oct-1, and HSF DNA binding assays, 1 g of poly(dI⅐dC) was used.
Isolation of RNA and Ribonuclease Protection Assay-Total RNA was isolated as described previously (18). The plasmid (pT3-rHIF-1␣) was linearized by SalI digestion and in vitro transcribed in the presence of [␣-32 P] GTP (DuPont NEN) with T3 polymerase (Boehringer Mannheim) to generate a 623-nucleotide riboprobe, which protects 370 and 253 nucleotides of HIF-1␣ transcripts. The Epo riboprobe was described previously (19). Ribonuclease protection assays were performed essentially as described previously (18) but modified as follows. Twenty g (for Epo) or 5 g (for HIF-1␣) of RNA were hybridized with the riboprobe overnight at 48°C and digested with 5 units of RNase One (Promega Corp.); the [ 32 P]RNA-RNA hybrids were precipitated with ethanol and resolved on a 6% polyacrylamide/8 M urea gel. In all of these analyses, riboprobe was present in substantial stoichiometric excess.
Western Blot-Western blotting was performed as described before (18). Briefly, samples (20 -30 g) of whole-cell extracts were resolved in a 12% SDS-polyacrylamide gel electrophoresis system and transblotted onto a piece of Immobilon-P transfer membrane (Millipore) using a Mini Trans-Blot Electrophoresis Transfer Cell (Bio-Rad). The membrane was probed with a mix of monoclonal anti-HIF-1␣ antibodies OZ12 and OZ15 (15) or 1:5000 dilution of polyclonal anti-ARNT antibody (20), followed by incubation, respectively, with anti-mouse or anti-rabbit (IgG) antibody conjugated with horseradish peroxidase. The antigenantibody complexes were detected by enhanced chemiluminescence (ECL, Amersham Corp.).
Transfection and Luciferase Assay-HeLa cells growing in log phase were pooled after trypsinization and transfected in suspension with the calcium phosphate method as described elsewhere (17). A typical transfection was performed by using 4 g of the luciferase reporter driven by human erythropoietin minimal enhancer and SV40 promoter (pEpoEluc) and 10 g of empty vector (pRc/CMV) or human thioredoxin gene driven by the cytomegalovirus enhancer and promoter (pCMV-ADF). The DNA calcium phosphate precipitates were incubated with suspended cells for 15-20 min and were resuspended with medium and split into two 60-mm dishes, one maintained at 21% O 2 and the other exposed to 1% O 2 . Cells were harvested 48 h after transfection. For hypoxia treatment, cells were incubated in the hypoxic chamber for 20 -24 h prior to harvest. Replicate transfection experiments were performed with two independent DNA preparations. Luciferase assays were performed as described previously (19) by using the Enhanced Luciferase assay kit (Analytical Luminescence Laboratory).

RESULTS
The Activation of HIF-1 Parallels the Abundance of ␣-Subunit Protein-When cell extracts prepared from normoxic and hypoxic cells were analyzed by electrophoretic mobility shift assays (EMSAs) using 32 P-labeled oligonucleotide containing the wild-type (W18) HIF-1 binding site (2), doublet bands (designated as HIF-1 binding) were detected in hypoxic extracts (Fig. 1, lane 2) consistent with previous reports (2,3). In agreement with the evidence that HIF-1 is an ␣␤ heterodimer, HIF-1 binding was supershifted with two different anti-HIF-1␣ monoclonal antibodies (lanes 5 and 6) and with a mixture of the two (lane 3), as well as with a polyclonal anti-HIF-1␤ antibody (lane 9).
To investigate the mechanism underlying regulation of HIF-1 activity, we first determined levels of HIF-1␣ and HIF-1␤ transcripts by ribonuclease protection analysis of HeLa cells maintained at 21% O 2 ( Fig. 2A, lane 1) or those incubated for 4 h under 1% O 2 (lane 2). Contrary to the initial report on the cloning of the HIF-1 subunits (5), we found that both HIF-1␣ ( Fig. 2A) and HIF-1␤ transcripts (data not shown) were readily detectable under normoxic conditions and were not significantly increased by hypoxia treatment. To exclude the possibility that longer hypoxic incubation might increase HIF-1␣ and HIF-1␤ mRNA levels, cells were incubated under 1% O 2 for 16 h. Again, no significant induction was observed. Moreover, RNA obtained from Hep3B cells gave the same results (data not shown).
In contrast, hypoxia had a marked effect at the level of HIF-1␣ protein expression. When HeLa cells were incubated for 4 h in 1% O 2 , the expected induction in HIF-1 DNA binding activity (Fig. 2B, lane 2) was accompanied by a striking increase in HIF-1␣ protein from barely detectable (Fig. 2C, top panel, lane 1) to abundant expression (lane 2), whereas HIF-1␤ protein was constitutively expressed at a high level with no significant change in abundance following exposure to hypoxia (lower panel). One mechanism for the accumulation of HIF-1␣ protein is that at low oxygen tension, the subunit is stablized but is rapidly degraded when oxygen is replete. To address this question, cells were incubated under 1% O 2 for 4 h, to produce maximal HIF-1 activation, and were then placed in a 21% O 2 chamber for 5, 10, 30, or 60 min. As shown in Fig. 2B, there was a precipitous drop in HIF-1 DNA binding after exposure to normoxia (lanes 2-6), as reported previously (13). The rapid decrease of HIF-1 DNA binding correlated closely with the rate of decay of HIF-1␣ abundance (T 1/2 , Ͻ5 min) as demonstrated by Western blot analyses of the same cell extract preparations, whereas HIF-1␤ abundance remained unchanged when the same blot was reprobed with anti-HIF-1␤ (Fig. 2C, lanes 2-6).
To confirm that the disappearance of HIF-1␣ was not due to lack of HIF-1␣ mRNA, total RNA was prepared from cells treated under the same conditions as above and analyzed by RNase protection. Once again, HIF-1␣ mRNA levels were unaffected ( Fig. 2A, lanes 2-6). Therefore, these results indicate that the HIF-1␣ protein is stablized by hypoxia and decays very rapidly upon reoxygenation. To investigate whether normoxia directly destabilizes HIF-1, hypoxic extracts were incubated at room temperature, 21% O 2 , for 1, 5, 10, 20, or 40 min and loaded onto an EMSA gel at the respective time points. Results in Fig. 2D showed that HIF-1 binding activity persisted throughout the course, suggesting that the instability of HIF-1␣ under normoxic conditions depends on cell integrity.
To test whether oxygen-dependent regulation of the level of HIF-1␣ protein is due to alterations in translation, HIF-1 was activated in HeLa cells by incubation for a total of 3 h in 1% O 2 , and at 120, 150, and 165 min cycloheximide (final concentration, 100 g/ml) was added anaerobically to separate plates in the chamber via a needle passed through a rubber stopper. No changes were noted at the 150-and 165-min time points, whereas at 120 min, this inhibitor of translation caused a modest reduction in the level of HIF-1␣ protein and in HIF-1 DNA binding (data not shown). These results indicate that in deoxygenated cells, HIF-1␣ protein cannot have a high turnover rate. Therefore, the rapid decay of HIF-1␣ protein seen with reoxygenation (Fig. 2C) cannot be due solely to suppres-sion of HIF-1␣ translation. Importantly, this decay was not affected by cycloheximide and, therefore, does not depend upon on-going protein synthesis.
The long 3Ј untranslated region of HIF-1␣ has been suggested to play a role in RNA stability (5). To confirm further that HIF-1␣ expression is controlled primarily at the protein level, full-length HIF-1␣ cDNA and HIF-1␣ cDNA lacking its 3Ј untranslated region were cloned into a eukaryotic expression vector downstream of the strong cytomegalovirus enhancer and promoter, designated pHIF-1␣ and pHIF-1␣/HN. To achieve high level expression of exogenous HIF-1␣, these constructs were transfected into 293 cells, which have a high degree of transfection efficiency, and analyzed by Western blot. As shown in Fig. 3, no significant amount of HIF-1␣ protein was detected under normoxic conditions (lanes 1, 3, and 5), whereas at low O 2 tension, significant increases in HIF-1␣ protein were detected in cells transfected with either pHIF-1␣ or pHIF-1␣ /HN (lanes 4 and 6). Accumulation of HIF-1␣ protein was not much affected by the 3Ј untranslated region in HIF-1␣ mRNA. Moreover, the expression of endogenous HIF-1␤ was unaffected by overexpression of HIF-1␣. These results show that oxygen decreases the levels of both overexpressed and endogenous HIF-1␣, further supporting the conclusion that regulation is independent of transcription and protein synthesis.
H 2 O 2 Blocks Accumulation of HIF-1␣, Resulting in Inactivation of HIF-1-The exogenous addition of hydrogen peroxide has been shown to inhibit hypoxic induction of erythropoietin protein (21). As indicated in Fig. 4, we show that this is associated with suppression of Epo mRNA levels. After treatment with increasing concentrations of H 2 O 2 for 15 min, Hep3B cells were exposed to 1% O 2 for 8 h prior to analysis of Epo mRNA by ribonuclease protection. Epo mRNA expression was slightly inhibited by 0.1 mM H 2 O 2 and was fully extinguished by 1 mM. This is a 3-5-fold higher concentration of peroxide than that generally required to elicit biological effects in cultured cells. In keeping with its suppression of the hypoxic induction of Epo mRNA, 1 mM H 2 O 2 inhibited HIF-1 DNA binding in extracts prepared from hypoxic Hep3B cells (data not shown). As shown in Fig. 5A, top panel, exposure of HeLa cells to 1 mM H 2 O 2 for 15 min, followed by incubation in 1% O 2 for 2 h, also resulted in nearly complete abolition of HIF-1 binding. The specificity of this inhibition was demonstrated by incubating the same cell extracts with 32 P-labeled oligonucleotides containing an Oct-1 binding site and the heat shock element, respectively. As shown in Fig. 5A, pretreatment with hydrogen peroxide had no effect on Oct-1 DNA binding (middle panel), but as expected (17), induced binding of heat shock transcription factor to the heat shock element (bottom panel).
To elucidate the mechanism by which pretreatment with hydrogen peroxide inhibits HIF-1 DNA binding, levels of both
HIF-1␣ and HIF-1␤ protein were quantified by Western blot analyses of these HeLa whole-cell extracts. The top panel of Fig. 5B shows that expression of HIF-1␣ protein was hypoxiainducible, consistent with the results in Fig. 2C, but this induction was greatly inhibited by hydrogen peroxide pretreatment. In contrast, the abundance of HIF-1␤ protein was not significantly affected (bottom panel). To further determine the effect of hydrogen peroxide on the expression of HIF-1␣, total RNA was isolated from cells treated under the same conditions as above and analyzed by ribonuclease protection. Despite the inhibition of HIF-1␣ protein accumulation, no significant changes in HIF-1␣ transcripts were observed (Fig. 5C).
To determine whether hydrogen peroxide has a direct effect on HIF-1 binding, hypoxic cell extracts were subjected to hydrogen peroxide in vitro prior to performing EMSAs. No loss of HIF-1 binding to DNA was observed at concentrations up to 1 mM (Fig. 5D), 2 indicating that hydrogen peroxide acts upstream of HIF-1 activation.
The inhibition of HIF-1␣ protein accumulation and HIF-1 DNA binding by hydrogen peroxide was fully reversible. After exposure of HeLa cells to H 2 O 2 for 20 min, the medium was removed, and the adherent cells were rinsed twice with normal (unmodified) medium prior to hypoxic incubation. As shown in Fig. 6A, removal of H 2 O 2 fully restored HIF-1 DNA binding (lane 4) with no significant effects on binding of nuclear receptors to the DR-2 site (Fig. 6B), a functionally important cisacting element located downstream of HIF-1 site in the Epo 3Ј enhancer (19). Moreover, removal of hydrogen peroxide also fully restored HIF-1␣ protein accumulation (Fig. 6C). Taken together, these results show that, just like normoxia, pretreatment with hydrogen peroxide inhibits hypoxia-induced activation of HIF-1 via blocking accumulation of HIF-1␣ protein, without affecting either HIF-1␣ mRNA expression or expression of HIF-1␤ protein.
As a test of whether hydrogen peroxide affects the translation process, in vitro translation of HIF-1␣ was performed in the presence and absence of hydrogen peroxide. Up to 10 mM H 2 O 2 had no effect on the formation of translation product. Although this cell-free system departs substantially from in vivo conditions, this result suggests that hydrogen peroxide does not impact directly on HIF-1␣ synthesis but instead targets upstream event(s) in the sensing/signaling pathway, thereby blocking accumulation of the HIF-1␣ protein.
Redox Regulation Appears to Be Necessary for HIF-1 DNA Binding-Although, as demonstrated above, the binding of HIF-1 to DNA is primarily controlled at the level of HIF-1␣ accumulation, there is little information on what posttranslational modifications are necessary for this activation. The inhibitory effects of H 2 O 2 on HIF-1 activity strongly suggest that redox chemistry is an important determinant of the stability of HIF-1␣. Therefore, we examined the effects of sulfhydryl reagents on in vitro HIF-1 activity. Whereas N-ethylmaleimide (NEM) irreversibly modifies thiol groups, diamide promotes reversible disulfide bond formation (22,23). In agreement with recently published results (24), when cell extracts were incubated with NEM or diamide prior to the addition to probe, HIF-1 DNA binding was significantly inhibited (Fig. 7, lanes 3  and 6) in a dose-dependent fashion (data not shown), suggest- ing that sulfhydryl alkylation or oxidation prevents DNA binding. Consistent with the chemical properties of these reagents, subsequent addition of dithiothreitol (DTT) fully reversed the inhibitory effect of diamide (lane 7) but not that of NEM (lane 4). Moreover, the monothiol reducing agent, ␤-mercaptoethanol, showed the same effect as DTT (lane 8). To further demonstrate that thiol groups were the target of these reagents, cell extracts were first treated with diamide followed by NEM to test whether the thiol groups were still accessible to NEM after oxidation. As shown in Fig. 7, DTT, in the presence of NEM, still reversed the inhibitory effect of diamide (lane 12), indicating that thiol groups involved in HIF-1 activity had been fully oxidized by diamide, thereby preventing alkylation by NEM.
Since oxidation or alkylation of thiols prevented HIF-1 DNA binding and sulfhydryl reduction reversed the inhibitory effect of oxidation, we investigated whether sulfhydryl reductants could stimulate HIF-1 DNA binding. Contrary to the conclusions of others (24), we found that when oxidized thioredoxin purified from E. coli was added in the presence of DTT, a clear stimulation in HIF-1 DNA binding was reproducibly detected in hypoxic extracts (Fig. 8A, lane 4). Moreover, in the presence of DTT, thioredoxin not only reversed the inhibitory effect of diamide but also enhanced HIF-1 DNA binding over that in untreated extracts (compare lanes 5 with 7 and 8). In contrast, oxidized thioredoxin had no effect in the absence of DTT (data not shown).
To assay the functional effect of redox potential as a modulator of HIF-1 activity in intact cells, we transfected HeLa cells with a vector expressing human thioredoxin (pCMV-ADF) along with a luciferase reporter (pEpoE-luc) containing the human erythropoietin HIF-1 binding site. It has been shown that exogenous expression of thioredoxin stimulates AP-1 activity but inhibits nuclear factor-B (25,26). As shown in Fig.  8B, overexpression of thioredoxin markedly potentiated the hypoxic induction of luciferase gene expression. Moreover, this effect correlated with the dose of vector (data not shown). Ref-1, a nuclear protein possessing both redox and apurinic endonuclease DNA repair activities (27,28), has also been shown to facilitate AP-1 binding activity (29). When the Ref-1 expression vector (pCMV-APE) was cotransfected with the reporter, like thioredoxin, it further enhanced the hypoxic induction of the reporter (Fig. 8B). Moreover, when the HIF-1 binding site was mutated, no potentiation was observed with either thioredoxin or Ref-1. Thus, the effect of these proteins was HIF-1-dependent. Because of high redox buffering capacity of the cells, it is likely that both thioredoxin and Ref-1 were in the reduced form.

DISCUSSION
Diverse mechanisms have evolved to activate transcription factors in response to specific environmental signals. There is growing evidence that hypoxia regulates transcription of a broad repertoire of biologically important genes through a common mechanism of O 2 sensing, signal transduction, and transactivation (30). The oxygen sensor is likely to be a heme protein (31,32), perhaps a cytochrome b-like flavo-heme NAD(P)H oxidase that signals by altering intracellular peroxide and activated oxygen compounds (21,(33)(34)(35). Hypoxia results in the activation of heterodimeric HIF-1, enabling it to bind to cis elements of responsive genes. HIF-1␣ interacts specifically with p300/CBP, a transcriptional co-activator, thereby up-regulating gene expression (15).
Our studies address the mechanism responsible for HIF-1 activation by hypoxia. Others (5) have reported up-regulation of both HIF-1␣ and HIF-1␤ mRNA following exposure of Hep3B cells to hypoxia, cobalt, and desferrioxamine and rapid decay of both transcripts following transfer of cells from a hypoxic to a normoxic environment. However, the Northern blots in this report showed striking variability in both ␣ and ␤ signals in unstimulated cells as well as unexpected discontinuities in time course experiments. The conclusion that HIF-1␣ and HIF-1␤ mRNA levels are tightly regulated by cellular O 2 tension (5) is not borne out by our analyses by ribonuclease protection, showing no significant induction of either HIF-1␣ or HIF-1␤ mRNA with hypoxia and no significant decay with reoxygenation.
In spite of virtually constant levels of HIF-1␣ mRNA, we do observe marked changes in the abundance of HIF-1␣ protein that correlates closely with HIF-1 DNA binding activity. Both parameters decay rapidly when hypoxic cells are exposed to normoxia. Moreover, the pretreatment of cells with hydrogen  (5), who also noted a rapid decay of HIF-1␣ protein with reoxygenation but reported a slight decrease in HIF-1␤ protein with redistribution from the nucleus to the cytoplasm. Regulation of HIF-1␣ protein levels could occur via alteration in either the rate of translation or in the stability of the protein product. The fact that pretreatment with cyclohexamide blocks induction of HIF-1 DNA binding (2,13) is consistent with the former mechanism. However, we found that the addition of cyclohexamide after maximal hypoxic induction had no effect on the levels of HIF-1␣ protein or HIF-1 DNA binding. Moreover, overexpression of HIF-1␣ (Fig. 3) failed to induce any detectable HIF-1␣ protein in oxygenated cells. Finally, we have been unable to show any effect of peroxide on the rate of synthesis of the HIF-1␣ protein in a cell-free transcription-translation system. These three independent results argue strongly against a significant role of alterations in translation and favor the second mechanism: posttranslational alterations in stability. HIF-1␣ may be intrinsically unstable, owing to a pair of PEST-rich sequences (36) in the C-terminal portion of the protein (5). In this respect, it is of interest to note relevant differences between HIF-1␣ and the AH receptor, which also heterodimerizes with HIF-1␤. The AH protein, when unliganded, has a half-life of 8 h in hepatoma cells (37). In contrast to HIF-1␣, the highly stable AH receptor lacks canonical PEST sequences.
Regardless of the contribution of PEST sequences to the catabolism of HIF-1␣ protein, an issue of particular biological interest is the mechanism responsible for the modification of its stability by oxygen. Our experiments with hydrogen peroxide may provide insights into the signaling pathway. The pretreatment of cells with H 2 O 2 markedly inhibited HIF-1 DNA binding and accumulation of HIF-1␣ protein (Fig. 5, A and B). In contrast, even higher concentrations of H 2 O 2 had no effect on the ability of HIF-1 to bind to DNA (Fig. 5D). In like manner, hydrogen peroxide can activate NF-B when added to intact cells but has no effect on cell-free extracts (38). In both cases, the action of H 2 O 2 is indirect, presumably via upstream signaling.
Although hydrogen peroxide has pleotropic effects on cell metabolism and function, impacting on various signaling pathways (39 -41), including triggering of protein phosphorylation (42), its mechanism of action and that of reactive oxygen intermediates generated from peroxide depend on redox chemistry. In further support of the importance of the intracellular redox environment on the activation of HIF-1, we have shown that intracellular expression of either thioredoxin or a nuclear reducing protein Ref-1 up-regulates HIF-1-dependent hypoxic induction of a reporter gene (Fig. 8B). However, these experiments alter the redox status of the cell in an artificial manner. Physiologic signaling in a cell through a redox mechanism is likely to be channeled within the confines of specific subcellular compartments. Moreover, experimental manuevers that affect upstream signaling in a redox-dependent manner may have no relevance to direct modification of HIF-1 by sulfhydryl reagents. Again, nuclear factor-B may be an apt analogy; peroxide enhances its activation via upstream signaling steps involving release from its inhibitor (25,26,38,43), whereas direct modification of nuclear factor-B by sulfhydryl reagents inhibits DNA binding (44). Our experiments suggest the need for detailed structure-function studies of HIF-1␣ to ascertain whether, like Fos/Jun (45), USF (46), Rel/B (47), and bovine papilloma type 1 E2 protein (48), reduction of critical cysteine sulfhydryls plays an important functional role.
Taken together, as depicted in Fig. 9, our results indicate that the pathway leading from the sensing of hypoxia to the activation of HIF-1 is critically dependent on the relative abundance of its ␣ subunit. The striking difference in the oxygendependent expression of HIF-1␣ and HIF-1␤ at the protein level is consonant with the apparent specificity of HIF-1␣, the only known function of which is in hypoxia-specific regulation, whereas HIF-1␤ can heterodimerize with the aryl hydrocarbon receptor and perhaps with other transcription factors. The effect of oxygen in acutely lowering levels of HIF-1␣ is mimicked by hydrogen peroxide, a finding which is consistent with a large body of experimental work implicating reactive oxygen intermediates in the signaling process (30). We present both in vitro and in vivo experiments that indicate that redox chemistry contributes importantly to HIF-1 activation.